gradint-build/server/gradint.cgi → server/gradint.cgi

 #!/usr/bin/env python
 # -*- coding: utf-8 -*-
+#  (either Python 2 or Python 3)
 
-program_name = "gradint.cgi v1.06 (c) 2011 Silas S. Brown.  GPL v3+"
+program_name = "gradint.cgi v1.38 (c) 2011,2015,2017-25 Silas S. Brown.  GPL v3+"
 
 #    This program is free software; you can redistribute it and/or modify
 #    it under the terms of the GNU General Public License as published by
@@ -13,15 +14,40 @@ program_name = "gradint.cgi v1.06 (c) 2011 Silas S. Brown.  GPL v3+"
 #    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 #    GNU General Public License for more details.
 
-gradint_dir = "/home/ssb22/gradint" # include samples/prompts
-path_add = "/home/ssb22/gradint/bin" # include sox, lame, espeak, maybe oggenc
-lib_path_add = "/home/ssb22/gradint/lib"
-espeak_data_path = "/home/ssb22/gradint"
+# If your Python is 3.13 or above (expected in Ubuntu 2.26 LTS)
+# then you will need: pip install legacy-cgi
 
-import os, os.path, sys, commands, cgi, cgitb, urllib ; cgitb.enable()
-import Cookie, random
+gradint_dir = "$HOME/gradint" # include samples/prompts
+path_add = "$HOME/gradint/bin" # include sox, lame, espeak, maybe oggenc
+lib_path_add = "$HOME/gradint/lib"
+espeak_data_path = "$HOME/gradint"
+
+import os, os.path, sys, cgi, urllib, time, re # if this fails, see note above
+import tempfile, getpass
+myTmp = tempfile.gettempdir()+os.sep+getpass.getuser()+"-gradint-cgi"
+try: from commands import getoutput # Python 2
+except: from subprocess import getoutput # Python 3
+try: from urllib import quote,quote_plus,unquote # Python 2
+except: from urllib.parse import quote,quote_plus,unquote # Python 3
+try: from importlib import reload # Python 3
+except: pass
+home = os.environ.get("HOME","")
+if not home:
+  try:
+    import pwd
+    home = os.path.expanduser("~{0}".format(pwd.getpwuid(os.getuid())[0]))
+  except: home=0
+  if not home: home = ".." # assume we're in public_html
+gradint_dir = gradint_dir.replace("$HOME",home)
+path_add = path_add.replace("$HOME",home)
+lib_path_add = lib_path_add.replace("$HOME",home)
+espeak_data_path = espeak_data_path.replace("$HOME",home)
+try: import Cookie # Python 2
+except: import http.cookies as Cookie # Python 3
+import random
 if "QUERY_STRING" in os.environ and "&" in os.environ["QUERY_STRING"] and ";" in os.environ["QUERY_STRING"]: os.environ["QUERY_STRING"]=os.environ["QUERY_STRING"].replace(";","%3B") # for dictionary sites to add words that contain semicolon
-query = cgi.parse()
+try: query = cgi.FieldStorage(encoding="utf-8") # Python 3
+except: query = cgi.FieldStorage() # Python 2
 os.chdir(gradint_dir) ; sys.path.insert(0,os.getcwd())
 os.environ["PATH"] = path_add+":"+os.environ["PATH"]
 if "LD_LIBRARY_PATH" in os.environ: os.environ["LD_LIBRARY_PATH"] = lib_path_add+":"+os.environ["LD_LIBRARY_PATH"]
@@ -29,60 +55,76 @@ else: os.environ["LD_LIBRARY_PATH"] = lib_path_add
 os.environ["ESPEAK_DATA_PATH"] = espeak_data_path
 
 cginame = os.sep+sys.argv[0] ; cginame=cginame[cginame.rindex(os.sep)+1:]
-sys.stderr=open("/dev/null","w") ; sys.argv = [] ; import gradint
-
-lDic = {}
-for l in gradint.ESpeakSynth().describe_supported_languages().split():
-    abbr,name = l.split("=")
-    lDic[abbr]=name
+sys.stderr=open("/dev/null","w") ; sys.argv = []
+gradint = None
+def reinit_gradint(): # if calling again, also redo setup_userID after
+    global gradint,langFullName
+    if gradint:
+      if sys.version_info[0]>2: gradint.map,gradint.filter,gradint.chr=gradint._map,gradint._filter,gradint.unichr # undo Python 3 workaround in preparation for it to be done again, because reload doesn't do this (at least not on all Python versions)
+      gradint = reload(gradint)
+    else: import gradint
+    gradint.waitOnMessage = lambda *args:False
+    langFullName = {}
+    for l in gradint.ESpeakSynth().describe_supported_languages().split():
+        abbr,name = gradint.S(l).split("=")
+        langFullName[abbr]=name.replace("_","-")
+    # Try to work out probable default language:
+    lang = os.environ.get("HTTP_ACCEPT_LANGUAGE","")
+    if lang:
+        for c in [',',';','-']:
+            if c in lang: lang=lang[:lang.index(c)]
+        if not lang in langFullName: lang=""
+    global noGTranslate
+    if lang:
+        gradint.firstLanguage = lang
+        if lang=="en": noGTranslate = True
+        else:
+          gradint.secondLanguage="en" # (most probable default)
+          noGTranslate = lang in gradint.GUI_translations # (unless perhaps any are incomplete)
+    elif " zh-" in os.environ.get("HTTP_USER_AGENT",""): # Chinese iPhone w/out Accept-Language
+      gradint.firstLanguage,gradint.secondLanguage = "zh","en"
+      noGTranslate = True # (don't know if it even pops up on that browser, but anyway)
 
-# Try to work out probable default language:
-lang = os.environ.get("HTTP_ACCEPT_LANGUAGE","")
-if lang:
- for c in [',',';','-']:
-   if c in lang: lang=lang[:lang.index(c)]
- if not lang in lDic: lang=""
-if lang:
- gradint.firstLanguage = lang
- if not lang=="en": gradint.secondLanguage="en"
-elif " zh-" in os.environ.get("HTTP_USER_AGENT",""): gradint.firstLanguage,gradint.secondLanguage = "zh","en" # Chinese iPhone
+reinit_gradint()
 
 def main():
+  if "id" in query: # e.g. from redirectHomeKeepCookie
+    queryID = query.getfirst("id")
+    if not re.match("[A-Za-z0-9_.-]",queryID): return htmlOut("Bad query.  Bad, bad query.") # to avoid cluttering the disk if we're being given random queries by an attacker.  IDs we generate are numeric only, but allow alphanumeric in case server admin wants to generate them.  Don't allow =, parens, etc (likely random SQL query)
+    os.environ["HTTP_COOKIE"]="id="+queryID
+    print ('Set-Cookie: id=' + queryID+'; expires=Wed, 1 Dec 2036 23:59:59 GMT') # TODO: S2G
   if has_userID(): setup_userID() # always, even for justSynth, as it may include a voice selection (TODO consequently being called twice in many circumstances, could make this more efficient)
-  if "id" in query:
-    os.environ["HTTP_COOKIE"]="id="+query["id"][0]
-    print 'Set-Cookie: id=' + query["id"][0]+'; expires=Wed, 1 Dec 2036 23:59:59 GMT'
   filetype=""
-  if "filetype" in query: filetype=query["filetype"][0]
+  if "filetype" in query: filetype=query.getfirst("filetype")
   if not filetype in ["mp3","ogg","wav"]: filetype="mp3"
   for k in query.keys():
     if k.startswith("del-"):
-     k=urllib.unquote(urllib.unquote(k)) # might be needed
+     k=unquote(unquote(k)) # might be needed
      if '=' in k:
        l2,l1 = k[4:].split('=')
        setup_userID()
        gradint.delOrReplace(gradint.ensure_unicode(l2),gradint.ensure_unicode(l1),"","","delete")
        return listVocab(True)
   if "js" in query: # just synthesize (js=text jsl=language)
-    if "jsl" in query: justSynth(query["js"][0], query["jsl"][0],filetype=filetype)
-    else: justSynth(query["js"][0],filetype=filetype)
+    if "jsl" in query: justSynth(query.getfirst("js"), query.getfirst("jsl"),filetype=filetype)
+    else: justSynth(query.getfirst("js"),filetype=filetype)
   elif "spk" in query: # speak (l1,l2 the langs, l1w,l2w the words)
     gradint.justSynthesize="0"
-    if "l2w" in query and query["l2w"][0]:
+    if "l2w" in query and query.getfirst("l2w"):
       gradint.startBrowser=lambda *args:0
-      if query["l2"][0]=="zh" and gradint.sanityCheck(query["l2w"][0],"zh"): gradint.justSynthesize += "#en Pinyin needs tones.  Please go back and add tone numbers." # speaking it because alert box might not work and we might be being called from HTML5 Audio stuff (TODO maybe duplicate sanityCheck in js, if so don't call HTML5 audio, then we can have an on-screen message here)
-      else: gradint.justSynthesize += "#"+query["l2"][0].replace("#","").replace('"','')+" "+query["l2w"][0].replace("#","").replace('"','')
-    if "l1w" in query and query["l1w"][0]: gradint.justSynthesize += "#"+query["l1"][0].replace("#","").replace('"','')+" "+query["l1w"][0].replace("#","").replace('"','')
-    if gradint.justSynthesize=="0": return htmlOut('You must type a word in the box before pressing the Speak button.'+backLink) # TODO maybe add a Javascript test to the form also, IF can figure out if window.alert works
+      if query.getfirst("l2")=="zh" and gradint.generalCheck(query.getfirst("l2w"),"zh"): gradint.justSynthesize += "#en Pinyin needs tones.  Please go back and add tone numbers." # speaking it because alert box might not work and we might be being called from HTML5 Audio stuff (TODO maybe duplicate generalCheck in js, if so don't call HTML5 audio, then we can have an on-screen message here)
+      else: gradint.justSynthesize += "#"+query.getfirst("l2").replace("#","").replace('"','')+" "+query.getfirst("l2w").replace("#","").replace('"','')
+    if "l1w" in query and query.getfirst("l1w"): gradint.justSynthesize += "#"+query.getfirst("l1").replace("#","").replace('"','')+" "+query.getfirst("l1w").replace("#","").replace('"','')
+    if gradint.justSynthesize=="0": return htmlOut(withLocalise('You must type a word in the box before pressing the Speak button.')+backLink) # TODO maybe add a Javascript test to the form also, IF can figure out if window.alert works
     serveAudio(stream = len(gradint.justSynthesize)>100, filetype=filetype)
   elif "add" in query: # add to vocab (l1,l2 the langs, l1w,l2w the words)
-    if "l2w" in query and query["l2w"][0] and "l1w" in query and query["l1w"][0]:
+    if "l2w" in query and query.getfirst("l2w") and "l1w" in query and query.getfirst("l1w"):
       gradint.startBrowser=lambda *args:0
-      if query["l2"][0]=="zh": scmsg=gradint.sanityCheck(query["l2w"][0],"zh")
-      else: scmsg=None
-      if scmsg: htmlOut(scmsg+''+backLink)
-      else: addWord(query["l1w"][0],query["l2w"][0],query["l1"][0],query["l2"][0])
-    else: htmlOut('You must type words in both boxes before pressing the Add button.'+backLink) # TODO maybe add a Javascript test to the form also, IF can figure out a way to tell whether window.alert() works or not
+      if query.getfirst("l2")=="zh": gcmsg=gradint.generalCheck(query.getfirst("l2w"),"zh")
+      else: gcmsg=None
+      if gcmsg: htmlOut(gradint.B(gcmsg)+gradint.B(backLink))
+      else: addWord(query.getfirst("l1w"),query.getfirst("l2w"),query.getfirst("l1"),query.getfirst("l2"))
+    else: htmlOut(withLocalise('You must type words in both boxes before pressing the Add button.')+backLink) # TODO maybe add a Javascript test to the form also, IF can figure out a way to tell whether window.alert() works or not
   elif "bulkadd" in query: # bulk adding, from authoring options
     dirID = setup_userID()
     def isOK(x):
@@ -94,13 +136,13 @@ def main():
     def mycmp(x,y): return cmp(int(x[1:]),int(y[1:]))
     keyList = sorted(filter(lambda x:isOK(x),query.keys()),mycmp)
     for k in keyList:
-      l2w,l1w = query[k][0].split('=',1)
-      addWord(l1w,l2w,query["l1"][0],query["l2"][0],False)
+      l2w,l1w = query.getfirst(k).split('=',1)
+      addWord(l1w,l2w,query.getfirst("l1"),query.getfirst("l2"),False)
     redirectHomeKeepCookie(dirID,"&dictionary=1") # '1' is special value for JS-only back link; don't try to link to referer as it might be a generated page
   elif "clang" in query: # change languages (l1,l2)
     dirID = setup_userID()
-    if (gradint.firstLanguage,gradint.secondLanguage) == (query["l1"][0],query["l2"][0]) and not query["clang"][0]=="ignore-unchanged": return htmlOut('You must change the settings before pressing the Change Languages button.'+backLink) # (external scripts can set clang=ignore-unchanged)
-    gradint.updateSettingsFile(gradint.settingsFile,{"firstLanguage": query["l1"][0],"secondLanguage":query["l2"][0]})
+    if (gradint.firstLanguage,gradint.secondLanguage) == (query.getfirst("l1"),query.getfirst("l2")) and not query.getfirst("clang")=="ignore-unchanged": return htmlOut(withLocalise('You must change the settings before pressing the Change Languages button.')+backLink) # (external scripts can set clang=ignore-unchanged)
+    gradint.updateSettingsFile(gradint.settingsFile,{"firstLanguage": query.getfirst("l1"),"secondLanguage":query.getfirst("l2")})
     redirectHomeKeepCookie(dirID)
   elif "swaplang" in query: # swap languages
     dirID = setup_userID()
@@ -108,7 +150,7 @@ def main():
     redirectHomeKeepCookie(dirID)
   elif "editsave" in query: # save 'vocab'
     dirID = setup_userID()
-    if "vocab" in query: vocab=query["vocab"][0]
+    if "vocab" in query: vocab=query.getfirst("vocab")
     else: vocab="" # user blanked it
     open(gradint.vocabFile,"w").write(vocab)
     redirectHomeKeepCookie(dirID)
@@ -116,13 +158,25 @@ def main():
     dirID = setup_userID()
     try: v=open(gradint.vocabFile).read()
     except: v="" # (shouldn't get here unless they hack URLs)
-    htmlOut('<form action="'+cginame+'" method="post"><textarea name="vocab" style="width:100%;height:80%" rows="15" cols="50">'+v+'</textarea><br><input type=submit name=editsave value="Save changes"> | <input type=submit name=dummy value="Cancel"></form>',"Text edit your vocab list")
-  elif "lesson" in query: # make lesson
+    htmlOut('<form action="'+cginame+'" method="post"><textarea name="vocab" style="width:100%;height:80%" rows="15" cols="50">'+v+'</textarea><br><input type=submit name=editsave value="Save changes"> | <input type=submit name=placeholder value="Cancel"></form>',"Text edit your vocab list")
+  elif "lesson" in query: # make lesson ("Start lesson" button)
     setup_userID()
-    gradint.maxNewWords = int(query["new"][0]) # (shouldn't need sensible-range check here if got a dropdown; if they really want to hack the URL then ok...)
-    gradint.maxLenOfLesson = int(float(query["mins"][0])*60)
+    gradint.maxNewWords = int(query.getfirst("new")) # (shouldn't need sensible-range check here if got a dropdown; if they really want to hack the URL then ok...)
+    gradint.maxLenOfLesson = int(float(query.getfirst("mins"))*60)
     # TODO save those settings for next time also?
     serveAudio(stream = True, inURL = False, filetype=filetype)
+  elif "bigger" in query or "smaller" in query:
+    u = setup_userID() ; global zoom
+    if "bigger" in query: zoom = int(zoom*1.1)
+    else: zoom = int(zoom/1.1 + 0.5)
+    open(u+"-zoom.txt","w").write("%d\n" % zoom)
+    listVocab(True)
+  elif any("variant"+str(c) in query for c in range(max(len(gradint.GUI_translations[v]) for v in gradint.GUI_translations.keys() if v.startswith("@variants-")))):
+    for c in range(max(len(gradint.GUI_translations[v]) for v in gradint.GUI_translations.keys() if v.startswith("@variants-"))): #TODO duplicate code
+      if "variant"+str(c) in query: break
+    u = setup_userID()
+    gradint.updateSettingsFile(u+"-settings.txt",{"scriptVariants":{gradint.GUI_languages.get(gradint.firstLanguage,gradint.firstLanguage):c}})
+    setup_userID() ; listVocab(True)
   elif "voNormal" in query: # voice option = normal
     setup_userID()
     gradint.voiceOption=""
@@ -131,21 +185,32 @@ def main():
   elif "vopt" in query: # set voice option
     setup_userID()
     for v in gradint.guiVoiceOptions:
-      if v.lower()=="-"+query["vopt"][0].lower():
+      if v.lower()=="-"+query.getfirst("vopt").lower():
         gradint.voiceOption = v
         gradint.updateSettingsFile(gradint.settingsFile,{"voiceOption":v})
         break
     listVocab(True)
+  elif "lFinish" in query:
+    dirID = setup_userID()
+    try: os.rename(gradint.progressFile+'-new',gradint.progressFile)
+    except: pass # probably a duplicate GET
+    try: os.remove(gradint.progressFile+'-ts') # the timestamp file
+    except: pass
+    redirectHomeKeepCookie(dirID)
   elif not isAuthoringOption(query): listVocab(has_userID()) # default screen
 
+def U(x):
+  try: return x.decode('utf-8')
+  except: return x
+
 def isAuthoringOption(query):
   # TODO document the ?author=1 option
   if "author" in query:
     htmlOut('<form action="'+cginame+'" method="post"><h2>Gradint word list authoring mode</h2>This can help you put word lists on your website. The words will be linked to this Gradint server so your visitors can choose which ones to hear and/or add to their personal lists.<p>Type any text in the box below; use blank lines to separate paragraphs. To embed a word list in your text, type:<br><em>phrase 1</em>=<em>meaning 1</em><br><em>phrase 2</em>=<em>meaning 2</em><br><em>phrase 3</em>=<em>meaning 3</em><br>etc, and <b>make sure there is a blank line before and after the list</b>. Then press <input type=submit name="generate" value="Generate HTML">.<p>Language for phrases: '+langSelect('l2',gradint.secondLanguage)+' and for meanings: '+langSelect('l1',gradint.firstLanguage)+'<p><textarea name="text" style="width:100%;height:80%" rows="15" cols="50"></textarea><br><input type=submit name="generate" value="Generate HTML"></form>',"Word list authoring",links=0)
     # TODO maybe langSelect for mand+cant together ? (but many wordlists wld be topolect-specific)
   elif "generate" in query:
-    l1,l2,txt = query["l1"][0],query["l2"][0],query["text"][0]
-    paras = "\n".join([l.strip() for l in txt.replace("\r\n","\n").replace("\r","\n").decode('utf-8').split("\n")]).split("\n\n")
+    l1,l2,txt = query.getfirst("l1"),query.getfirst("l2"),query.getfirst("text")
+    paras = "\n".join([l.strip() for l in U(txt).replace("\r\n","\n").replace("\r","\n").split("\n")]).split("\n\n")
     need_h5a = False
     for i in xrange(len(paras)):
         lines = filter(lambda x:x,paras[i].split("\n")) # filter needed for trailing newline on document
@@ -163,14 +228,16 @@ def allLinesHaveEquals(lines):
     for l in lines:
         if not '=' in l: return False
     return True
+gradintUrl = os.environ.get("SCRIPT_URI","") # will be http:// or https:// as appropriate
+if not gradintUrl and all(x in os.environ for x in ["REQUEST_SCHEME","SERVER_NAME","SCRIPT_NAME"]): gradintUrl = os.environ["REQUEST_SCHEME"]+"://"+os.environ["SERVER_NAME"]+os.environ["SCRIPT_NAME"]
+if not gradintUrl: gradintUrl = "gradint.cgi" # guessing
 def authorWordList(lines,l1,l2):
-    gradintUrl = os.environ["SCRIPT_URI"]
     r=[] ; count = 0
     # could have target="gradint" in the following, but it may be in a background tab (target="_blank" not recommended as could accumulate many)
     r.append('<form action="%s" method="post" accept-charset="utf-8"><table style="margin-left:auto;margin-right:auto;border:thin solid blue"><tr><td colspan=3 style="text-align:center"><em>Click on each word for audio</em></td></tr>' % gradintUrl)
     for l in lines:
         l2w,l1w = l.split('=',1)
-        r.append('<tr><td><input type="checkbox" name="W%d" value="%s=%s" checked></td><td>%s</td><td>%s</td></tr>' % (count,l2w,l1w,justsynthLink(l2w.encode('utf-8'),l2).replace('HREF="'+cginame+'?','HREF="'+gradintUrl+'?').decode('utf-8'),justsynthLink(l1w.encode('utf-8'),l1).replace('HREF="'+cginame+'?','HREF="'+gradintUrl+'?').decode('utf-8')))
+        r.append('<tr class="notranslate"><td><input type="checkbox" name="W%d" value="%s=%s" checked></td><td>%s</td><td>%s</td></tr>' % (count,l2w,l1w,U(justsynthLink(l2w.encode('utf-8'),l2)).replace('HREF="'+cginame+'?','HREF="'+gradintUrl+'?'),U(justsynthLink(l1w.encode('utf-8'),l1)).replace('HREF="'+cginame+'?','HREF="'+gradintUrl+'?')))
         count += 1
     # could have target="gradint" in the following href, but see comment above
     r.append('<tr><td colspan=3><input type="submit" name="bulkadd" value="Add selected words"> to your <a href="%s">personal list</a></td></tr></table><input type="hidden" name="l1" value="%s"><input type="hidden" name="l2" value="%s"></form>' % (gradintUrl,l1,l2))
@@ -193,52 +260,107 @@ def justSynth(text,lang="",filetype=""):
 def justsynthLink(text,lang=""): # assumes written function h5a
   if lang in gradint.synth_partials_voices and gradint.guiVoiceOptions: cacheInfo="&curVopt="+gradint.voiceOption
   else: cacheInfo=""
-  return '<A HREF="'+cginame+'?js='+urllib.quote_plus(text)+'&jsl='+urllib.quote_plus(lang)+cacheInfo+'" onClick="javascript:return h5a(this);">'+text+'</A>'
+  return '<A HREF="'+cginame+'?js='+gradint.S(quote_plus(text))+'&jsl='+quote_plus(lang)+cacheInfo+'" onClick="return h5a(this);">'+gradint.S(text)+'</A>'
 # TODO if h5a's canPlayType etc works, cld o/p a lesson as a JS web page that does its own 'take out of event stream' and 'progress write-back'.  wld need to code that HERE by inspecting the finished Lesson object, don't call play().
 
+zoom = 100 # in case browser device lacks a zoom UI, we'll provide one
+noGTranslate = False
 def htmlOut(body_u8,title_extra="",links=1):
-    print "Content-type: text/html; charset=utf-8" ; print
+    if noGTranslate: print ("Google: notranslate")
+    print ("Content-type: text/html; charset=utf-8\n")
     if title_extra: title_extra=": "+title_extra
-    print '<html><head><title>Gradint Web edition'+title_extra+'</title>'
-    print '<meta name="viewport" content="width=device-width">'
-    print '</head><body>'+body_u8
-    print '<HR>'
+    print ('<html lang="en"><head><title>Gradint Web edition'+title_extra+'</title>')
+    print ('<meta name="mobileoptimized" content="0"><meta name="viewport" content="width=device-width">')
+    print ('<script>if(window.matchMedia && window.matchMedia("(prefers-color-scheme: dark)").matches)document.write("<style>body,input,textarea { background-color: black; color: #c0c000; } select,input[type=submit],input[type=button] { background-color: #300020; color: #c0c000; } input[type=text] { border: grey groove; } select[disabled],input[disabled] { background-color: #101010; color: #b0b000; } a:link { color: #00b000; } a:visited { color: #00c0c0; } a:hover { color: red; }</style>");</script>')
+    if not zoom==100: print('<style>body{font-size:%d%%;}body *{font-size:100%%;}</style>' % zoom)
+    print ('</head><body>')
+    if type(body_u8)==type(u""): body_u8=body_u8.encode('utf-8')
+    if hasattr(sys.stdout,'buffer'): # Python 3
+      sys.stdout.flush()
+      sys.stdout.buffer.write(body_u8)
+      sys.stdout.flush()
+    else: print(body_u8)
+    print ('<HR>')
     if links:
-        print 'This is Gradint Web edition.  If you need recorded words or additional functions, please <A HREF="http://people.pwf.cam.ac.uk/ssb22/gradint/">download the full version of Gradint</A>.'
+        print ('This is Gradint Web edition.  If you need recorded words or additional functions, please <A HREF="//ssb22.user.srcf.net/gradint/">download the full version of Gradint</A>.')
         # TODO @ low-priority: Android 3 <input type="file" accept="audio/*;capture=microphone"></input>
-        if "iPhone" in os.environ.get("HTTP_USER_AGENT","") and gradint.secondLanguage=="zh": print '<p>You can also try the Open University <A HREF="http://itunes.apple.com/gb/app/chinese-characters-first-steps/id441549197?mt=8#">Chinese Characters First Steps</A> iPhone application.'
-    print '<p>'+program_name[:program_name.index("(")]+"using "+gradint.program_name[:gradint.program_name.index("(")]
-    print "</body></html>"
-backLink = ' <A HREF="'+cginame+'" onClick="javascript:history.go(-1);return false">Back</A>' # TODO may want to add a random= to the non-js HREF
+    print ('<p>'+program_name[:program_name.index("(")]+"using "+gradint.program_name[:gradint.program_name.index("(")])
+    print ("</body></html>")
+backLink = ' <A HREF="'+cginame+'" onClick="history.go(-1);return false">Back</A>' # TODO may want to add a random= to the non-js HREF
 
 def serveAudio(stream=0, filetype="mp3", inURL=1):
   # caller imports gradint (and sets justSynthesize or whatever) first
-  if filetype=="mp3": print "Content-type: audio/mpeg"
-  else: print "Content-type: audio/"+filetype # ok for ogg, wav?
+  if os.environ.get("HTTP_IF_MODIFIED_SINCE",""):
+    print ("Status: 304 Not Modified\n\n") ; return
+  httpRange = re.match("bytes=([0-9]*)-([0-9]*)$",os.environ.get('HTTP_RANGE','')) # we MUST support Range: for some iOS players (Apple did not follow the HTTP standard of having a sensible fallback if servers respond with 200, and Apache will not do Range for us if we're CGI).  Single Range should be sufficient.
+  if httpRange: httpRange = httpRange.groups()
+  if httpRange==('',''): httpRange = None # must spec one
+  if httpRange:
+    if not httpRange[0]: httpRange=[-int(httpRange[1]),None]
+    elif not httpRange[1]: httpRange=[int(httpRange[0]),None]
+    else: httpRange=[int(httpRange[0]),int(httpRange[1])+1]
+    print ("Status: 206 Partial Content")
+    stream = 0
+  if filetype=="mp3": print ("Content-type: audio/mpeg")
+  else: print ("Content-type: audio/"+filetype) # ok for ogg, wav?
   if inURL:
-    print "Last-Modified: Sun, 06 Jul 2008 13:20:05 GMT"
-    print "Expires: Wed, 1 Dec 2036 23:59:59 GMT"
+    print ("Last-Modified: Sun, 06 Jul 2008 13:20:05 GMT")
+    print ("Expires: Wed, 1 Dec 2036 23:59:59 GMT") # TODO: S2G
+  print ("Content-disposition: attachment; filename=gradint."+filetype) # helps with some browsers that can't really do streaming
   gradint.out_type = filetype
+  gradint.waitBeforeStart = 0
   def mainOrSynth():
+    oldProgress = None ; rollback = False
+    if not gradint.justSynthesize and 'h5a' in query:
+      try: oldProgress = open(gradint.progressFile,'rb').read()
+      except: pass
+      rollback = True
+      if "lesson" in query: random.seed(query.getfirst("lesson")) # so clients that re-GET same lesson from partway through can work
     try: gradint.main()
     except SystemExit:
       if not gradint.justSynthesize:
+        o1,o2 = gradint.write_to_stdout,gradint.outputFile
+        reinit_gradint() ; setup_userID()
+        gradint.write_to_stdout,gradint.outputFile = o1,o2
+        gradint.setSoundCollector(gradint.SoundCollector())
         gradint.justSynthesize = "en Problem generating the lesson. Check we have prompts for those languages." ; gradint.main()
+        if oldProgress: open(gradint.progressFile,'wb').write(oldProgress)
+        rollback = oldProgress = None
+    if rollback: # roll back pending lFinish
+      os.rename(gradint.progressFile,gradint.progressFile+'-new')
+      if oldProgress: open(gradint.progressFile,'wb').write(oldProgress)
+    # end of def mainOrSynth
   if stream:
-    print "Content-disposition: attachment; filename=gradint.mp3" # helps with some browsers that can't really do streaming
-    print ; sys.stdout.flush()
+    print ("")
+    sys.stdout.flush()
     gradint.write_to_stdout = 1
     gradint.outputFile="-."+filetype ; gradint.setSoundCollector(gradint.SoundCollector())
     mainOrSynth()
   else:
-    tempdir = commands.getoutput("mktemp -d")
     gradint.write_to_stdout = 0
-    gradint.outputFile=tempdir+"/serveThis."+filetype ; gradint.setSoundCollector(gradint.SoundCollector())
-    gradint.waitBeforeStart = 0
-    mainOrSynth()
-    print "Content-Length:",os.stat(tempdir+"/serveThis."+filetype).st_size
-    print ; sys.stdout.flush()
-    os.system("cat "+tempdir+"/serveThis."+filetype)
+    tempdir = tempfile.mkdtemp()
+    fn,fn2 = tempdir+"/I."+filetype, tempdir+"/O."+filetype
+    if httpRange and "lesson" in query: # try to cache it
+      try: os.mkdir(myTmp)
+      except: pass # exist ok
+      for f in os.listdir(myTmp):
+        if os.stat(myTmp+os.sep+f).st_mtime < time.time()-4000:
+          os.remove(myTmp+os.sep+f)
+      fn = gradint.outputPrefix+str(int(query.getfirst("lesson")))+"."+filetype # (don't be tricked into clobbering paths with non-int lesson IDs)
+    if not os.path.exists(fn):
+      gradint.outputFile=fn
+      gradint.setSoundCollector(gradint.SoundCollector())
+      mainOrSynth()
+    if httpRange:
+      total = os.stat(fn).st_size
+      open(fn2,"wb").write(open(fn,"rb").read()[httpRange[0]:httpRange[1]])
+      if httpRange[0]<0: httpRange[0] += total
+      if not httpRange[1]: httpRange[1] = total
+      print("Content-Range: bytes %d-%d/%d" % (httpRange[0],httpRange[1]-1,total))
+    else: fn2 = fn
+    print ("Content-Length: "+repr(os.stat(fn2).st_size)+"\n")
+    sys.stdout.flush()
+    os.system("cat "+fn2) # components already validated so no quoting required
     os.system("rm -r "+tempdir)
 
 def addWord(l1w,l2w,l1,l2,out=True):
@@ -247,56 +369,88 @@ def addWord(l1w,l2w,l1,l2,out=True):
       if not ((gradint.firstLanguage,gradint.secondLanguage) == (l2,l1) and "HTTP_REFERER" in os.environ and not cginame in os.environ["HTTP_REFERER"]): gradint.updateSettingsFile(gradint.settingsFile,{"firstLanguage": l1,"secondLanguage":l2})
       gradint.firstLanguage,gradint.secondLanguage = l1,l2
     if (l1w+"_"+l1,l2w+"_"+l2) in map(lambda x:x[1:],gradint.parseSynthVocab(gradint.vocabFile,forGUI=1)):
-      if out: htmlOut('This word is already in your list.'+backLink)
+      if out: htmlOut(withLocalise('This word is already in your list.')+backLink)
       return
-    gradint.appendVocabFileInRightLanguages().write(l2w+"="+l1w+"\n")
+    gradint.appendVocabFileInRightLanguages().write(gradint.B(l2w)+gradint.B("=")+gradint.B(l1w)+gradint.B("\n"))
     if not out: return
-    if "HTTP_REFERER" in os.environ and not cginame in os.environ["HTTP_REFERER"]: extra="&dictionary="+urllib.quote(os.environ["HTTP_REFERER"])
+    if "HTTP_REFERER" in os.environ and not cginame in os.environ["HTTP_REFERER"]: extra="&dictionary="+quote(os.environ["HTTP_REFERER"])
     else: extra=""
     redirectHomeKeepCookie(dirID,extra)
 
 def redirectHomeKeepCookie(dirID,extra=""):
-    print "Location: "+cginame+"?random="+str(random.random())+"&id="+dirID[dirID.rindex("/")+1:]+extra ; print
+    dirID = gradint.S(dirID) # just in case
+    print ("Location: "+cginame+"?random="+str(random.random())[2:]+"&id="+dirID[dirID.rindex("/")+1:]+extra+"\n")
 
 def langSelect(name,curLang):
     curLang = gradint.espeak_language_aliases.get(curLang,curLang)
-    return '<select name="'+name+'">'+''.join(['<option value="'+abbr+'"'+gradint.cond(abbr==curLang," selected","")+'>'+localise(abbr)+' ('+abbr+')'+'</option>' for abbr in sorted(lDic.keys())])+'</select>'
+    return '<select name="'+name+'">'+''.join(['<option value="'+abbr+'"'+gradint.cond(abbr==curLang," selected","")+'>'+localise(abbr)+' ('+abbr+')'+'</option>' for abbr in sorted(langFullName.keys())])+'</select>'
 
 def numSelect(name,nums,curNum): return '<select name="'+name+'">'+''.join(['<option value="'+str(num)+'"'+gradint.cond(num==curNum," selected","")+'>'+str(num)+'</option>' for num in nums])+'</select>'
 
-def localise(x):
+def localise(x,span=0):
     r=gradint.localise(x)
-    if r==x: return lDic.get(gradint.espeak_language_aliases.get(x,x),x)
-    else: return r.encode('utf-8')
+    if r==x: return langFullName.get(gradint.espeak_language_aliases.get(x,x),x)
+    if span==1: r="<span lang=\""+gradint.firstLanguage+"\">"+r+"</span>"
+    elif span==2: r+='" lang="'+gradint.firstLanguage
+    if type(r)==type("")==type(u""): return r # Python 3
+    else: return r.encode('utf-8') # Python 2
 for k,v in {"Swap":{"zh":u"交换","zh2":u"交換"},
             "Text edit":{"zh":u"文本编辑"},
             "Delete":{"zh":u"删除","zh2":u"刪除"},
+            "Really delete this word?":{"zh":u"真的删除这个词?","zh2":u"真的刪除這個詞?"},
+            "Your word list":{"zh":u"你的词汇表","zh2":u"你的詞彙表"},
+            "click for audio":{"zh":u"击某词就听声音","zh2":u"擊某詞就聽聲音"},
+            "Repeats":{"zh":u"重复计数","zh2":u"重複計數"},
+            "To edit this list on another computer, type":{"zh":u"要是想在其他的电脑或手机编辑这个词汇表，请在别的设备打","zh2":u"要是想在其他的電腦或手機編輯這個詞彙表，請在別的設備打"},
+            "Please wait while the lesson starts to play":{"zh":u"稍等本课正开始播放","zh2":u"稍等本課正開始播放"},
+            "Bigger":{"zh":u"大"},"Smaller":{"zh":u"小"},
+            'You must type a word in the box before pressing the Speak button.':{"zh":u"按‘发音’前，应该框里打字。","zh2":u"按‘發音’前，應該框裡打字。"},
+            'You must type words in both boxes before pressing the Add button.':{"zh":u"按‘添加’前，应该在两框里打字。","zh2":u"按‘添加’前，應該在兩框裡打字。"},
+            'You must change the settings before pressing the Change Languages button.':{"zh":u"按‘选择其他语言’前，应该转换语言设定。","zh2":u"按‘選擇其他語言’前，應該轉換語言設定。"},
+            'This word is already in your list.':{"zh":u"本词已经在您的词汇表。","zh2":u"本詞已經在您的詞彙表。"},
+            "Your word list is empty.":{"zh":u"词汇表没有词汇，加一些吧","zh2":u"詞彙表沒有詞彙，加一些吧"}
             }.items():
   if not k in gradint.GUI_translations: gradint.GUI_translations[k]=v
 
+def withLocalise(x): return x+"  "+localise(x,1)
+
 def h5a():
-    body = """<script language="Javascript"><!--
-function h5a(link) { if (document.createElement) {
+    body = """<script><!--
+function h5a(link,endFunc) { if (document.createElement) {
    var ae = document.createElement('audio');
-   if (ae.canPlayType && function(s){return s!="" && s!="no"}(ae.canPlayType('audio/mpeg'))) {
-     if (link.href) ae.setAttribute('src', link.href);
-     else ae.setAttribute('src', link);
-     ae.play();
-     return false; }"""
-    if gradint.got_program("oggenc"): body += """else if (ae.canPlayType && function(s){return s!="" && s!="no"}(ae.canPlayType('audio/ogg'))) {
-     if (link.href) ae.setAttribute('src', link.href+"&filetype=ogg");
-     else ae.setAttribute('src', link+"&filetype=ogg");
-     ae.play();
-     return false; }"""
+   function cp(t,lAdd) { if(ae.canPlayType && function(s){return s!="" && s!="no"}(ae.canPlayType(t))) {
+       if (link.href) ae.setAttribute('src', link.href+lAdd);
+       else ae.setAttribute('src', link+lAdd);
+       if (typeof endFunc !== 'undefined') { ae.addEventListener("ended", endFunc, false); ae.addEventListener("timeupdate",function(e){t=ae.currentTime;m=Math.floor(t/60);t=Math.floor(t%60);document.forms[0].lesson.value=m+(t<10?":0":":")+t},false) }
+       ae.play(); return true; // Safari can say "Unhandled Promise Rejection: AbortError: The operation was aborted." in console log, but plays anyway when loaded
+    } return false; }
+   if (cp('audio/mpeg','')) return false;"""
+    if gradint.got_program("oggenc"): body += """else if (cp('audio/ogg',"&filetype=ogg")) return false;"""
     body += """} return true; }
 //--></script>"""
     return body
+def hasVoiceOptions(l):
+    if not l in gradint.synth_partials_voices: return False
+    if not gradint.guiVoiceOptions: return False
+    try: voices = os.listdir(gradint.partialsDirectory+os.sep+l)
+    except: voices = []
+    for v in voices:
+        if "-" in v and v[:v.index("-")] in voices: return True
 def listVocab(hasList): # main screen
     firstLanguage,secondLanguage = gradint.firstLanguage, gradint.secondLanguage
     # TODO button onClick: careful of zh w/out tones, wld need to JS this
     body = h5a() + '<center><form action="'+cginame+'">'
-    gotVoiceOptions = (gradint.secondLanguage in gradint.synth_partials_voices or gradint.firstLanguage in gradint.synth_partials_voices) and gradint.guiVoiceOptions
-    # TODO what if it's in synth_partials_voices but NOT the one that has guiVoiceOptions ? (e.g. Cantonese when both Mandarin voices are installed) (currently displaying 'non-functional' voice option buttons when that happens)
+    body += '<input type=submit name=bigger value="%s"> | <input type=submit name=smaller value="%s">' % (localise("Bigger"),localise("Smaller"))
+    GUIlang = gradint.GUI_languages.get(firstLanguage,firstLanguage)
+    if "@variants-"+GUIlang in gradint.GUI_translations:
+      body += " -- " # separating from big/small
+      count = 0
+      for variant in gradint.GUI_translations["@variants-"+GUIlang]:
+        if count: body += " | "
+        body += '<input type=submit name="variant'+str(count)+'" value="'+gradint.cond(type("")==type(u""),variant,variant.encode('utf-8'))+'"'+gradint.cond(gradint.scriptVariants.get(GUIlang,0)==count,' disabled="disabled"',"")+'>'
+        count += 1
+    body += "<br>"
+    gotVoiceOptions = (hasVoiceOptions(gradint.secondLanguage) or hasVoiceOptions(gradint.firstLanguage))
     if gotVoiceOptions:
       body += 'Voice option: <input type=submit name=voNormal value="Normal"'+gradint.cond(gradint.voiceOption=="",' disabled="disabled"',"")+'>'
       for v in gradint.guiVoiceOptions: body += ' | <input type=submit name=vopt value="'+v[1].upper()+v[2:]+'"'+gradint.cond(gradint.voiceOption==v,' disabled="disabled"',"")+'>'
@@ -305,53 +459,59 @@ def listVocab(hasList): # main screen
     # must have autocomplete=off if capturing keycode 13
     if gotVoiceOptions: cacheInfo="&curVopt="+gradint.voiceOption
     else: cacheInfo=""
-    body += (localise("Word in %s") % localise(secondLanguage))+': <input type=text name=l2w autocomplete=off onkeydown="if(event.keyCode==13) {document.forms[0].spk.click();return false} else return true"> <input type=submit name=spk value="'+localise("Speak")+'" onClick="javascript: if (!document.forms[0].l1w.value && !document.forms[0].l2w.value) return true; else return h5a(\''+cginame+'?spk=1&l1w=\'+document.forms[0].l1w.value+\'&l2w=\'+document.forms[0].l2w.value+\'&l1=\'+document.forms[0].l1.value+\'&l2=\'+document.forms[0].l2.value+\''+cacheInfo+'\');"><br>'+(localise("Meaning in %s") % localise(firstLanguage))+': <input type=text name=l1w autocomplete=off onkeydown="if(event.keyCode==13) {document.forms[0].add.click();return false} else return true"> <input type=submit name=add value="'+(localise("Add to %s") % localise("vocab.txt").replace(".txt",""))+'"><script language="Javascript"><!--\nvar emptyString="";document.write(\' <input type=submit name=dummy value="'+localise("Clear input boxes")+'" onClick="javascript:document.forms[0].l1w.value=document.forms[0].l2w.value=emptyString;document.forms[0].l2w.focus();return false">\')\n//--></script><p>'+localise("Your first language")+': '+langSelect('l1',firstLanguage)+' '+localise("second")+': '+langSelect('l2',secondLanguage)+' <nobr><input type=submit name=clang value="'+localise("Change languages")+'"><input type=submit name=swaplang value="'+localise("Swap")+'"></nobr>'
+    body += (localise("Word in %s",1) % localise(secondLanguage))+': <input type=text name=l2w autocomplete=off onkeydown="if(event.keyCode==13) {document.forms[0].spk.click();return false} else return true" onfocus="document.forms[0].onsubmit=\'document.forms[0].onsubmit=&quot;return true&quot;;document.forms[0].spk.click();return false\'" onblur="document.forms[0].onsubmit=\'return true\'"> <input type=submit name=spk value="'+localise("Speak",2)+'" onClick="if (!document.forms[0].l1w.value && !document.forms[0].l2w.value) return true; else return h5a(\''+cginame+'?spk=1&l1w=\'+document.forms[0].l1w.value+\'&l2w=\'+document.forms[0].l2w.value+\'&l1=\'+document.forms[0].l1.value+\'&l2=\'+document.forms[0].l2.value+\''+cacheInfo+'\');"><br>'+(localise("Meaning in %s",1) % localise(firstLanguage))+': <input type=text name=l1w autocomplete=off onkeydown="if(event.keyCode==13) {document.forms[0].add.click();return false} else return true" onfocus="document.forms[0].onsubmit=\'document.forms[0].onsubmit=&quot;return true&quot;;document.forms[0].add.click();return false\'" onblur="document.forms[0].onsubmit=\'return true\'"> <input type=submit name=add value="'+(localise("Add to %s",2) % localise("vocab.txt").replace(".txt",""))+'"><script><!--\nvar emptyString="";document.write(\' <input type=submit name=placeholder value="'+localise("Clear input boxes",2)+'" onClick="document.forms[0].l1w.value=document.forms[0].l2w.value=emptyString;document.forms[0].l2w.focus();return false">\')\n//--></script><p><nobr>'+localise("Your first language",1)+': '+langSelect('l1',firstLanguage)+'</nobr> <nobr>'+localise("second",1)+': '+langSelect('l2',secondLanguage)+'</nobr> <nobr><input type=submit name=clang value="'+localise("Change languages",2)+'"><input type=submit name=swaplang value="'+localise("Swap",2)+'"></nobr>' # onfocus..onblur updating onsubmit is needed for iOS "Go" button
     def htmlize(l,lang):
        if type(l)==type([]) or type(l)==type(()): return htmlize(l[-1],lang)
-       if "!synth:" in l: return htmlize(l[l.index("!synth:")+7:l.rfind("_")],lang)
+       l = gradint.B(l)
+       if gradint.B("!synth:") in l: return htmlize(l[l.index(gradint.B("!synth:"))+7:l.rfind(gradint.B("_"))],lang)
        return justsynthLink(l,lang)
     def deleteLink(l1,l2):
        r = []
        for l in [l2,l1]:
-         if type(l)==type([]) or type(l)==type(()) or not "!synth:" in l: return "" # Web-GUI delete in poetry etc not yet supported
-         r.append(urllib.quote(l[l.index("!synth:")+7:l.rfind("_")]))
-       r.append(localise("Delete"))
-       return '<TD><input type=submit name="del-%s%%3d%s" value="%s" onClick="javascript: return confirm(\'Really delete this word?\');"></TD>' % tuple(r)
+         if type(l)==type([]) or type(l)==type(()) or not gradint.B("!synth:") in gradint.B(l): return "" # Web-GUI delete in poetry etc not yet supported
+         l = gradint.B(l)
+         r.append(gradint.S(quote(l[l.index(gradint.B("!synth:"))+7:l.rfind(gradint.B("_"))])))
+       r.append(localise("Delete",2))
+       return ('<td><input type=submit name="del-%s%%3d%s" value="%s" onClick="return confirm(\''+localise("Really delete this word?")+'\');"></td>') % tuple(r)
     if hasList:
        gradint.availablePrompts = gradint.AvailablePrompts() # needed before ProgressDatabase()
        # gradint.cache_maintenance_mode=1 # don't transliterate on scan -> NO, including this scans promptsDirectory!
        gradint.ESpeakSynth.update_translit_cache=lambda *args:0 # do it this way instead
        data = gradint.ProgressDatabase().data ; data.reverse()
-       if data: hasList = "<p><TABLE style=\"border: thin solid green\"><caption><nobr>Your word list</NOBR> <NOBR>(click for audio)</NOBR> <input type=submit name=edit value=\""+localise("Text edit")+"\"></caption><TR><TH>Repeats</TH><TH>"+localise(gradint.secondLanguage)+"</TH><TH>"+localise(gradint.firstLanguage)+"</TH></TR>"+"".join(["<TR><TD>%d</TD><TD>%s</TD><TD>%s</TD>%s" % (num,htmlize(dest,gradint.secondLanguage),htmlize(src,gradint.firstLanguage),deleteLink(src,dest)) for num,src,dest in data])+"</TABLE>"
+       if data: hasList = "<p><table style=\"border: thin solid green\"><caption><nobr>"+localise("Your word list",1)+"</nobr> <nobr>("+localise("click for audio",1)+")</nobr> <input type=submit name=edit value=\""+localise("Text edit",2)+"\"></caption><tr><th>"+localise("Repeats",1)+"</th><th>"+localise(gradint.secondLanguage,1)+"</th><th>"+localise(gradint.firstLanguage,1)+"</th></tr>"+"".join(["<tr class=\"notranslate\"><td>%d</td><td lang=\"%s\">%s</td><td lang=\"%s\">%s</td>%s" % (num,gradint.secondLanguage,htmlize(dest,gradint.secondLanguage),gradint.firstLanguage,htmlize(src,gradint.firstLanguage),deleteLink(src,dest)) for num,src,dest in data])+"</table>"
        else: hasList=""
     else: hasList=""
-    if hasList: body += '<P><table style="border:thin solid blue"><tr><td>'+numSelect('new',range(2,10),gradint.maxNewWords)+' '+localise("new words in")+' '+numSelect('mins',[15,20,25,30],int(gradint.maxLenOfLesson/60))+' '+localise('mins')+' <input type=submit name=lesson value="'+localise("Start lesson")+'"></td></tr></table>'
+    if hasList: body += '<p><table style="border:thin solid blue"><tr><td>'+numSelect('new',range(2,10),gradint.maxNewWords)+' '+localise("new words in")+' '+numSelect('mins',[15,20,25,30],int(gradint.maxLenOfLesson/60))+' '+localise('mins')+""" <input type=submit name=lesson value="""+'"'+localise("Start lesson",2)+"""" onClick="document.forms[0].lesson.disabled=1; document.forms[0].lesson.value=&quot;"""+localise("Please wait while the lesson starts to play")+"""&quot;;document.d0=new Date();return h5a('"""+cginame+'?lesson='+str(random.random())[2:]+"""&h5a=1&new='+document.forms[0].new.value+'&mins='+document.forms[0].mins.value,function(){if(new Date()-document.d0>60000)location.href='"""+cginame+'?lFinish='+str(random.random())[2:]+"""';else{document.forms[0].lesson.value='PLAY ERROR'}})"></td></tr></table>""" # when lesson ended, refresh with lFinish which saves progress (interrupts before then cancel it), but don't save progress if somehow got the ended event in 1st minute as that could be a browser issue
     if "dictionary" in query:
-        if query["dictionary"][0]=="1": body += '<script language="Javascript"><!--\ndocument.write(\'<p><a href="javascript:history.go(-1)">Back to referring site</a>\')\n//--></script>' # apparently it is -1, not -2; the redirect doesn't count as one (TODO are there any JS browsers that do count it as 2?)
-        else: body += '<p><a href="'+query["dictionary"][0]+'">Back to dictionary</a>' # TODO check for cross-site scripting
-    if not hasList: hasList="<P>Your word list is empty."
+        if query.getfirst("dictionary")=="1": body += '<script><!--\ndocument.write(\'<p><a href="javascript:history.go(-1)">'+localise("Back to referring site",1)+'</a>\')\n//--></script>' # apparently it is -1, not -2; the redirect doesn't count as one (TODO are there any JS browsers that do count it as 2?)
+        else: body += '<p><a href="'+query.getfirst("dictionary")+'">'+localise("Back to dictionary",1)+'</a>' # TODO check for cross-site scripting
+    if hasList:
+      if "://" in gradintUrl: hasList += "<p>"+localise("To edit this list on another computer, type",1)+" <kbd>"+gradintUrl.replace(".","<wbr>.").replace("/","<wbr>/")+"?id="+re.sub("([0-9]{4})(?!$)",r"\1<wbr><span><!-- (this is not a phone number) --></span>",getCookieId())+"</kbd>" # span needed for iOS at least
+    else: hasList="<p>"+localise("Your word list is empty.",1)
     body += hasList
-    htmlOut(body+'</form></center><script language="Javascript"><!--\ndocument.forms[0].l2w.focus()\n//--></script>')
+    htmlOut(body+'</form></center><script><!--\ndocument.forms[0].l2w.focus()\n//--></script>')
 
-def has_userID():
+def has_userID(): # TODO: can just call getCookieId with not too much extra overhead
     cookie_string = os.environ.get('HTTP_COOKIE',"")
     if cookie_string:
         cookie = Cookie.SimpleCookie()
         cookie.load(cookie_string)
         return 'id' in cookie
 
+def getCookieId():
+    cookie_string = os.environ.get('HTTP_COOKIE',"")
+    if not cookie_string: return
+    cookie = Cookie.SimpleCookie()
+    cookie.load(cookie_string)
+    if 'id' in cookie: return cookie['id'].value.replace('"','').replace("'","").replace("\\","")
+
 def setup_userID():
     # MUST call before outputting headers (may set cookie)
     # Use the return value of this with -settings.txt, -vocab.txt etc
     if cginame=="gradint.cgi": dirName = "cgi-gradint-users" # as previous versions
     else: dirName = cginame+"-users" # TODO document this feature (you can symlink something-else.cgi to gradint.cgi and it will have a separate user directory) (however it still reports gradint.cgi on the footer)
     if not os.path.exists(dirName): os.system("mkdir "+dirName)
-    userID = None
-    cookie_string = os.environ.get('HTTP_COOKIE',"")
-    if cookie_string:
-        cookie = Cookie.SimpleCookie()
-        cookie.load(cookie_string)
-        if 'id' in cookie: userID = cookie['id'].value.replace('"','').replace("'","").replace("\\","")
+    userID = getCookieId()
     need_write = (userID and not os.path.exists(dirName+'/'+userID+'-settings.txt')) # maybe it got cleaned up
     if not userID:
         while True:
@@ -359,14 +519,28 @@ def setup_userID():
             if not os.path.exists(dirName+'/'+userID+'-settings.txt'): break
         open(dirName+'/'+userID+'-settings.txt','w') # TODO this could still be a race condition (but should be OK under normal circumstances)
         need_write = 1
-        print 'Set-Cookie: id=' + userID+'; expires=Wed, 1 Dec 2036 23:59:59 GMT'
-    userID = dirName+'/'+userID
+        print ('Set-Cookie: id=' + userID+'; expires=Wed, 1 Dec 2036 23:59:59 GMT') # TODO: S2G
+    userID0, userID = userID, dirName+os.sep+userID # already validated
     gradint.progressFileBackup=gradint.pickledProgressFile=None
     gradint.vocabFile = userID+"-vocab.txt"
     gradint.progressFile = userID+"-progress.txt"
     gradint.settingsFile = userID+"-settings.txt"
+    gradint.outputPrefix = myTmp+os.sep+userID0+"-"
     if need_write: gradint.updateSettingsFile(gradint.settingsFile,{'firstLanguage':gradint.firstLanguage,'secondLanguage':gradint.secondLanguage})
     else: gradint.readSettings(gradint.settingsFile)
+    gradint.auto_advancedPrompt=1 # prompt in L2 if we don't have L1 prompts on the server, what else can we do...
+    if os.path.exists(userID+"-zoom.txt"):
+      global zoom ; zoom = int(open(userID+"-zoom.txt").read().strip())
     return userID
 
-main()
+try: main()
+except Exception as e:
+  print ("Content-type: text/plain; charset=utf-8\n")
+  sys.stdout.flush()
+  import traceback
+  try: traceback.print_exc(file=sys.stdout)
+  except: pass
+  sys.stdout.flush()
+  if hasattr(sys.stdout,"buffer"): buf = sys.stdout.buffer
+  else: buf = sys.stdout
+  buf.write(repr(e).encode("utf-8"))

gradint-build/server/lesson-table.py → server/lesson-table.py

 #!/usr/bin/env python
+# (compatible with both Python 2 and Python 3)
 
 # Script to generate an HTML table of the contents of a lesson
 # for summarizing it to a teacher or native speaker.
 # Reads from progressFile and progressFileBackup.
 
-# Version 1.03 (c) 2011 Silas S. Brown.  License: GPL
+# Version 1.07 (c) 2011, 2020-21, 2025 Silas S. Brown.  License: GPL
 
 # Example use:
-# python lesson-table.py | ssh some-server 'mutt -e "set record = \"\";" -e "set charset=\"utf-8\"; set send_charset=\"utf-8\"; set content_type=\"text/html\";" to-address -s "Gradint report"' || echo Send failed
-
-samples_url = None # or "http://example.org/path/to/samples/"
+# export samples_url=http://example.org/path/to/samples/ # or omit
+# python lesson-table.py [gradint-params] | ssh some-server 'mutt -e "set record = \"\";" -e "set charset=\"utf-8\"; set send_charset=\"utf-8\"; set content_type=\"text/html\";" to-address -s "Gradint report"' || echo Send failed
 
 import gradint, os
+samples_url = os.getenv("samples_url","")
+
+from gradint import B,S
 newpf = gradint.progressFile
 gradint.progressFile = gradint.progressFileBackup
 gradint.pickledProgressFile=None
@@ -31,35 +34,68 @@ newProg = gradint.ProgressDatabase(alsoScan=0)
 gradint.mergeProgress(newProg.data,mergeIn)
 del mergeIn
 changes = [] ; count=0
-newProg.data.sort()
+gradint.sort(newProg.data,gradint.cmpfunc)
 for tries,l1,l2 in newProg.data:
   if not tries: continue
   key = gradint.norm_filelist(l1,l2)
   oldTries = opd.get(key,0)
-  if not oldTries==tries: changes.append((oldTries,count,tries-oldTries,l1,l2))
+  if not oldTries==tries: changes.append((oldTries,count,tries-oldTries,S(l1),S(l2)))
   count += 1
 del newProg,opd
 changes.sort()
-print '<html><head><meta http_equiv="Content-type" value="text/html; charset=utf-8"><title>Gradint lesson report</title></head><body><h2>Gradint lesson report</h2><table border><tr><th>Repeats before</th><th>Repeats today</th><th>Question</th><th>Answer</th></tr>' # (have Question/Answer order rather than Word/Meaning, because if it's L2-only poetry then the question is the previous line, which is not exactly "meaning")
+print ('<html><head><meta http-equiv="Content-Type" content="text/html; charset=utf-8"><title>Gradint lesson report</title><meta name="mobileoptimized" content="0"><meta name="viewport" content="width=device-width"><script>if(window.matchMedia && window.matchMedia("(prefers-color-scheme: dark)").matches)document.write("<style>body { background-color: black; color: #c0c000; } a:link { color: #00b000; } a:visited { color: #00c0c0; } a:hover { color: red; }</style>");if(navigator.languages && navigator.languages.indexOf("en")>=0) document.write(\'<meta name="google" content="notranslate">\')</script></head><body><h2>Gradint lesson report</h2>')
+if gradint.unix and gradint.got_program("zgrep"):
+  print (os.popen("zgrep '^# collection=' \"%s\"" % gradint.progressFile).read()[2:].rstrip())
+print ('<table border><tr><th>Repeats before</th><th>Repeats today</th><th>Question</th><th>Answer</th></tr>') # (have Question/Answer order rather than Word/Meaning, because if it's L2-only poetry then the question is the previous line, which is not exactly "meaning")
+  
+had_h5a = False
+def h5aCode(filename):
+  r = real_h5aCode(filename)
+  if r:
+    global had_h5a
+    if not had_h5a:
+      had_h5a = True
+      print ("""<script language="Javascript"><!--
+function h5a(link,type) { if (document.createElement) {
+   var ae = document.createElement('audio');
+   if (ae.canPlayType && function(s){return s!="" && s!="no"}(ae.canPlayType(type))) {
+     if (link.href) ae.setAttribute('src', link.href);
+     else ae.setAttribute('src', link);
+     ae.play();
+     return false; } } return true; }
+//--></script>""")
+  return r
+def real_h5aCode(filename):
+  if filename.endswith(gradint.dotmp3): return ' onClick="javascript:return h5a(this,\'audio/mpeg\')"'
+  elif filename.endswith(gradint.dotwav): return ' onClick="javascript:return h5a(this,\'audio/wav\')"'
+  else: return ""
+
 def wrappable(f):
-  z = u'\u200b'.encode('utf-8') # zero-width space
+  z = u'\u200b' # zero-width space
+  if not type(u"")==type(""): z=z.encode('utf-8') # Py2
   return f.replace(os.sep,os.sep+z).replace('_',z+'_')
-def checkVariant(l):
+def checkVariant(l,ensureTxt=0):
+  l=S(l)
   if os.sep in l: fname=l[l.rindex(os.sep)+1:]
   else: fname=l
-  variants = gradint.variantFiles.get(gradint.samplesDirectory+os.sep+l,[fname])
+  variants = map(S,gradint.variantFiles.get(B(gradint.samplesDirectory+os.sep+l),[fname]))
   if fname in variants: return l # ok
   # else no default variant, need to pick one for the link
-  if not os.sep in l: return variants[0]
-  return l[:l.rindex(os.sep)+1]+variants[0]
+  for v in variants:
+    if ensureTxt:
+      if not v.endswith(gradint.dottxt): continue
+    elif v.endswith(gradint.dottxt): continue
+    if not os.sep in l: return v
+    return l[:l.rindex(os.sep)+1]+v
 def link(l):
   if type(l)==type([]): return link(l[-1])
-  if l.lower().endswith(gradint.dottxt): l="!synth:"+gradint.u8strip(gradint.read(gradint.samplesDirectory+os.sep+l)).strip(gradint.wsp)+"_"+gradint.languageof(l)
+  l = S(l)
+  if l.lower().endswith(gradint.dottxt): l="!synth:"+S(gradint.u8strip(gradint.read(gradint.samplesDirectory+os.sep+checkVariant(l,1)))).strip(gradint.wsp)+"_"+gradint.languageof(l)
   if "!synth:" in l:
-    if gradint.languageof(l) not in [gradint.firstLanguage,gradint.secondLanguage]: l=gradint.textof(l)+" ("+gradint.languageof(l)+")"
-    else: l=gradint.textof(l)
+    if gradint.languageof(l) not in [gradint.firstLanguage,gradint.secondLanguage]: l=S(gradint.textof(l))+" ("+gradint.languageof(l)+")"
+    else: l=S(gradint.textof(l))
     return l.replace('&','&amp;').replace('<','&lt;')
-  if samples_url: return '<A HREF="'+samples_url+checkVariant(l)+'">'+wrappable(l)+'</A>'
+  if samples_url: return '<A HREF="'+samples_url+checkVariant(l)+'"'+h5aCode(checkVariant(l))+'>'+wrappable(l)+'</A>'
   return wrappable(l).replace('&','&amp;').replace('<','&lt;')
-for b4,pos,today,l1,l2 in changes: print '<tr><td>%d</td><td>%d</td><td>%s</td><td>%s</td></tr>' % (b4,today,link(l1),link(l2))
-print '</table></body></html>'
+for b4,pos,today,l1,l2 in changes: print ('<tr><td>%d</td><td>%d</td><td class="notranslate">%s</td><td class="notranslate">%s</td></tr>' % (b4,today,link(l1),link(l2)))
+print ('</table></body></html>')

gradint-build/server/safety-check-progressfile.py → server/safety-check-progressfile.py

+#!/usr/bin/env python
+# (either Python 2 or Python 3)
+
 # safety-check-progressfile.py:
 # The purpose of this script is to check
 # progress.txt for safety.  Because it's

gradint-build/server/samples.cgi → server/samples.cgi

 #!/bin/bash
 
-# Gradint online samples browser v1.0 (c) 2011 Silas S. Brown.  License: GPL
+# Gradint online samples browser v1.1 (c) 2011,2013 Silas S. Brown.  License: GPL
 
 # Works as an "indexing" CGI.
 # To set up in Apache, make .htaccess with:
@@ -28,7 +28,11 @@ if ! test "a$REQUEST_URI" == a; then
   cd "/home/ssb22/public_html/$(echo "$REQUEST_URI"|sed -e 's/?.*//')"
 fi # else assume the server put us in the right directory, like mathopd does
 
-export Filename="$(pwd|sed -e 's,.*/,,').zip"
+if echo "$SERVER_SOFTWARE"|grep Apache >/dev/null; then
+  echo "Status: 200 OK" # overriding the 403
+fi # (mathopd doesn't need this, and not tested with all mathopd versions)
+
+Filename="$(pwd|sed -e 's,.*/,,').zip"
 
 if test "$QUERY_STRING" == zip || test "a$(echo "$REQUEST_URI"|sed -e 's/.*?//')" == azip; then
   echo Content-type: application/zip
@@ -66,8 +70,8 @@ function h5a(link) {
 //--></script>
 EOF
   for N in *; do
-    export Size=$(du -h --apparent-size -s "$N"|cut -f1)
-    if echo "$N"|grep '\.txt$'>/dev/null && echo $Size|grep '^[0-9]*$' >/dev/null;then export Size="$(cat "$N")";else export Size="($Size)"; fi
+    Size=$(du -h --apparent-size -s "$N"|cut -f1)
+    if echo "$N"|grep '\.txt$'>/dev/null && echo $Size|grep '^[0-9]*$' >/dev/null;then Size="$(cat "$N")";else Size="($Size)"; fi
     echo "<LI><A HREF=\"$N\" onClick=\"javascript:return h5a(this)\">$N</A> $Size</LI>"
   done
   echo "</UL></BODY></HTML>"

gradint-build/server/vocab2html.py → server/vocab2html.py

@@ -13,34 +13,35 @@
 # be used.  E.g.: export ESPEAK_CGI_URL="/~userID/espeak.cgi"
 # (TODO: this script ignores the possibility of synthesizing phrases from partials)
 
-# Version 1.1, (c) Silas S. Brown, License: GPL
+# Version 1.2, (c) Silas S. Brown, License: GPL
 
 from gradint import *
 if not synthCache: synthCache_contents = []
 langs=[secondLanguage,firstLanguage]
 o=open(vocabFile,"rU")
 justHadP=1
-print '<html><HEAD><META HTTP-EQUIV=Content-type CONTENT="text/html; charset=utf-8"><meta name="viewport" content="width=device-width"></HEAD><body>' # (assume utf8 in case there's any hanzi, but TODO what if using another charset for another language?)
-for l in o.readlines():
+print ('<html><HEAD><META HTTP-EQUIV=Content-type CONTENT="text/html; charset=utf-8"><meta name="viewport" content="width=device-width"></HEAD><body>') # (assume utf8 in case there's any hanzi, but TODO what if using another charset for another language?)
+for l in o:
   l2=l.lower()
   if l2.startswith("set language ") or l2.startswith("set languages "): langs=l.split()[2:]
   if not l.strip():
     # blank line
-    if not justHadP: print "<P>"
+    if not justHadP: print ("<P>")
     justHadP=1 ; continue
-  if not justHadP: print "<BR>"
+  if not justHadP: print ("<BR>")
   if l2.startswith("set language ") or l2.startswith("set languages ") or l2.startswith("limit on") or l2.startswith("limit off") or l2.startswith("begin poetry") or l2.startswith("end poetry"):
-    print "<EM>%s</EM>" % (l,)
+    print ("<EM>%s</EM>" % (l,))
   elif l2.startswith("#"):
     # comment (and may be part of multi-line comment)
-    if not l[1:].strip().startswith("<!--"): print "<small>#</small> "
-    print l[1:]
+    if not l[1:].strip().startswith("<!--"): print ("<small>#</small> ")
+    print (l[1:])
   else:
     # vocab line
     langsAndWords=zip(langs,map(lambda x:x.strip(),l.split("=")))
     out = []
     for lang,word in langsAndWords:
-      fname=synthCache_transtbl.get(word.lower()+"_"+lang+dotwav,word.lower()+"_"+lang+dotwav)
+      lang,word = S(lang),S(word)
+      fname=S(synthCache_transtbl.get(word.lower()+"_"+lang+dotwav,word.lower()+"_"+lang+dotwav))
       found = 0
       for fn2 in [fname,fname.replace(dotwav,dotmp3)]:
           if fn2 in synthCache_contents:
@@ -48,9 +49,10 @@ for l in o.readlines():
               found = 1 ; break
       if not found:
           if os.getenv("ESPEAK_CGI_URL"):
-              import urllib
-              out.append("<A HREF=\""+os.getenv("ESPEAK_CGI_URL")+"?"+urllib.urlencode({"t":word,"l":lang})+"\">"+word+"</A>")
+              try: from urllib import urlencode # Python 2
+              except: from urllib.parse import urlencode # Python 3
+              out.append("<A HREF=\""+os.getenv("ESPEAK_CGI_URL")+"?"+urlencode({"t":word,"l":lang})+"\">"+word+"</A>")
           else: out.append(word)
-    print " = ".join(out)
+    print (" = ".join(out))
   justHadP=0
-print "</body></html>"
+print ("</body></html>")

src/Makefile

+../gradint.py:
+	make -C .. gradint.py

gradint-build/src/booktime.py → src/booktime.py

-# This file is part of the source code of
-# gradint v0.9979 (c) 2002-2011 Silas S. Brown. GPL v3+.
+# This file is part of the source code of Gradint
+# (c) Silas S. Brown.
 #    This program is free software; you can redistribute it and/or modify
 #    it under the terms of the GNU General Public License as published by
 #    the Free Software Foundation; either version 3 of the License, or
@@ -23,17 +23,14 @@
 
 def initialGlue(): return Glue(0,maxLenOfLesson)
 
-try: import bisect
+try: from bisect import insort
 except:
-    class bisect: pass
-    bisect=bisect()
     def insort(l,item):
         l.append(item) ; l.sort()
-    bisect.insort = insort
 class Schedule(object):
     # A sorted list of (start,finish) times that are booked
     def __init__(self): self.bookedList = []
-    def book(self,start,finish): bisect.insort(self.bookedList,(start,finish))
+    def book(self,start,finish): insort(self.bookedList,(start,finish))
 
 earliestAllowedEvent = 0 # for "don't start before" hacks, so can keep all initial glue starting at 0
 
@@ -123,10 +120,7 @@ class Glue (GlueOrEvent):
     def __init__(self,length,plusMinus):
         GlueOrEvent.__init__(self,length,plusMinus,1)
 
-def sgn(a):
-    # Not all versions of Python have this built-in
-    if a: return a/abs(a)
-    else: return 1
+def sgn(a): return [1,-1][a<0]
 
 class StretchedTooFar(Exception): pass
 class GluedEvent(object):
@@ -244,9 +238,9 @@ class Lesson(object):
         else:
             # less confusing message for a beginner
             return l("Today we will learn %d words\nThis will require %d %s %d %s\nFollow the spoken instructions carefully") % (self.newWords,finish/60,singular(finish/60,"minutes"),finish%60,singular(finish%60,"seconds"))
-    def addSequence(self,gluedEventList):
+    def addSequence(self,gluedEventList,canTrack=True):
         bookIn(gluedEventList,self.schedule)
-        if not gluedListTracker==None: gluedListTracker.append(gluedEventList)
+        if not gluedListTracker==None and canTrack: gluedListTracker.append(gluedEventList)
         glueStart = 0 ; lastI = None
         for i in gluedEventList:
             i.event.setOnLeaves("sequenceID",self.eventListCounter) # for max_lateness stuff
@@ -268,7 +262,7 @@ class Lesson(object):
         latenessCap = {} ; nextStart = 0
         for t,event in self.events:
             if nextStart:
-                for k in latenessCap.keys(): latenessCap[k] += (nextStart-(t+event.length)) # the gap
+                for k in list(latenessCap.keys()): latenessCap[k] += (nextStart-(t+event.length)) # the gap
             nextStart = t
             if not hasattr(event,"importance"): continue # (wasn't added via addSequence, probably not a normal lesson)
             event.max_lateness=min(event.max_lateness,latenessCap.get(event.importance,maxLenOfLesson))
@@ -294,13 +288,16 @@ class Lesson(object):
         global sequenceIDs_to_cancel ; sequenceIDs_to_cancel = {}
         global copy_of_runner_events ; copy_of_runner_events = []
         global lessonStartTime ; lessonStartTime = 0 # will be set to time.time() on 1st event
-        # make the runner as late as possible
-        if soundCollector: runner = sched.scheduler(collector_time,collector_sleep)
-        else: runner = sched.scheduler(time.time,mysleep)
-        for (t,event) in self.events: copy_of_runner_events.append((event,runner.enter(t,1,play,(event,)),t))
-        # TODO what if Brief Interrupt appears during that events loop and someone presses it (will act as a Cancel and go back to main)
-        try: runner.run()
-        except KeyboardInterrupt: handleInterrupt()
+        disable_lid(0)
+        try:
+          # make the runner as late as possible
+          if soundCollector: runner = sched.scheduler(collector_time,collector_sleep)
+          else: runner = sched.scheduler(time.time,mysleep)
+          for (t,event) in self.events: copy_of_runner_events.append((event,runner.enter(t,1,play,(event,)),t))
+          # TODO what if Brief Interrupt appears during that events loop and someone presses it (will act as a Cancel and go back to main)
+          try: runner.run()
+          except KeyboardInterrupt: handleInterrupt()
+        finally: disable_lid(1)
         runner = None
         if soundCollector: soundCollector.finished()
         if logFileHandle: logFileHandle.close()
@@ -312,17 +309,17 @@ def decide_subst_synth(cache_fname):
 def subst_some_synth_for_synthcache(events):
     # turn SOME synthcache events back into synth events (for testing new synths etc)
     reverse_transTbl = {}
-    for k,v in synthCache_transtbl.items(): reverse_transTbl[v]=k
+    for k,v in list(synthCache_transtbl.items()): reverse_transTbl[v]=k
     for i in range(len(events)):
         if hasattr(events[i][1],"file") and events[i][1].file.startswith(synthCache+os.sep):
-            cache_fname = events[i][1].file[len(synthCache+os.sep):]
+            cache_fname = B(events[i][1].file[len(synthCache+os.sep):])
             cache_fname = reverse_transTbl.get(cache_fname,cache_fname)
-            if cache_fname[0]=="_": continue # a sporadically-used synthCache entry anyway
+            if cache_fname[:1]==B("_"): continue # a sporadically-used synthCache entry anyway
             if type(synthCache_test_mode)==type([]):
                 found=0
                 for str in synthCache_test_mode:
-                    if (re and re.search(str,cache_fname)) or cache_fname.find(str)>-1:
+                    if (re and re.search(str,cache_fname)) or cache_fname.find(str)>=0:
                         found=1 ; break
                 if found: continue
             lang = languageof(cache_fname)
-            if get_synth_if_possible(lang) and decide_subst_synth(cache_fname): events[i] = (events[i][0],synth_event(lang,cache_fname[:cache_fname.rindex("_")]))
+            if get_synth_if_possible(lang) and decide_subst_synth(cache_fname): events[i] = (events[i][0],synth_event(lang,cache_fname[:cache_fname.rindex(B("_"))]))

gradint-build/src/filescan.py → src/filescan.py

-# This file is part of the source code of
-# gradint v0.9979 (c) 2002-2011 Silas S. Brown. GPL v3+.
+# This file is part of the source code of Gradint
+# (c) Silas S. Brown.
 #    This program is free software; you can redistribute it and/or modify
 #    it under the terms of the GNU General Public License as published by
 #    the Free Software Foundation; either version 3 of the License, or
@@ -17,7 +17,7 @@ def init_scanSamples():
   dirsWithIntros = []
   filesWithExplanations = {}
   singleLinePoems = {} # keys are any poem files which are single line only, so as to avoid saying 'beginning' in prompts
-  variantFiles = {} # careful with clearing this if prompts is using it also (hence called only below and in loop.py before prompt scan)
+  variantFiles = {} # maps dir+fname to (no dir+) fname list, main use is in fileToEvent.  Careful with clearing this if prompts is using it also (hence init_scanSamples is called only below and in loop.py before prompt scan)
 init_scanSamples() ; emptyCheck_hack = 0
 def scanSamples(directory=None):
     if not directory: directory=samplesDirectory
@@ -59,7 +59,7 @@ def import_recordings(destDir=None):
                 if checkFirst:
                   for lang in [firstLanguage,secondLanguage]:
                    for ext in [dotwav,dotmp3]:
-                    if f[:f.rfind(extsep)]+"_"+lang+ext in curFiles: raise CannotOverwriteExisting()
+                    if checkIn(f[:f.rfind(extsep)]+"_"+lang+ext,curFiles): raise CannotOverwriteExisting()
                   continue
                 if not destDir:
                     if not getYN("Import the recordings that are in "+importDir+"?"): break
@@ -74,7 +74,7 @@ def import_recordings(destDir=None):
                     try:
                         import shutil
                         shutil.copy2(importDir+os.sep+f,destDir+os.sep+f)
-                    except: open(destDir+os.sep+f,"wb").write(read(importDir+os.sep+f))
+                    except: write(destDir+os.sep+f,read(importDir+os.sep+f))
                     os.remove(importDir+os.sep+f)
                 numFound += 1
     if numFound: open(destDir+os.sep+"settings"+dottxt,"w").write("firstLanguage=\""+firstLanguage+"\"\nsecondLanguage=\""+secondLanguage+"\"\n")
@@ -83,7 +83,7 @@ def import_recordings(destDir=None):
 def exec_in_a_func(x): # helper function for below (can't be nested in python 2.3)
    # Also be careful of http://bugs.python.org/issue4315 (shadowing globals in an exec) - better do this in a dictionary
    d={"firstLanguage":firstLanguage,"secondLanguage":secondLanguage}
-   exec x in d
+   exec (x,d)
    return d["secondLanguage"],d["firstLanguage"]
 def check_has_variants(directory,ls):
     if directory==promptsDirectory: return True
@@ -101,9 +101,9 @@ def getLsDic(directory):
     if not (directory.find(exclude_from_scan)==-1): return {}
     try: ls = os.listdir(directory)
     except: return {} # (can run without a 'samples' directory at all if just doing synth)
-    if "settings"+dottxt in ls:
+    if checkIn("settings"+dottxt,ls):
         # Sort out the o/p from import_recordings (and legacy record-with-HDogg.bat if anyone's still using that)
-        oddLanguage,evenLanguage = exec_in_a_func(u8strip(read(directory+os.sep+"settings"+dottxt).replace("\r\n","\n")).strip(wsp))
+        oddLanguage,evenLanguage = exec_in_a_func(wspstrip(u8strip(read(directory+os.sep+"settings"+dottxt).replace("\r\n","\n"))))
         if oddLanguage==evenLanguage: oddLanguage,evenLanguage="_"+oddLanguage,"-meaning_"+evenLanguage # if user sets languages the same, assume they want -meaning prompts
         else: oddLanguage,evenLanguage="_"+oddLanguage,"_"+evenLanguage
         for f in ls:
@@ -115,15 +115,15 @@ def getLsDic(directory):
             os.rename(directory+os.sep+f,directory+os.sep+f[:i]+(("%0"+str(len(str(len(ls))))+"d") % (int((int(num)-1)/2)*2+1))+cond(int(num)%2,oddLanguage,evenLanguage)+f[f.rfind(extsep):])
         os.remove(directory+os.sep+"settings"+dottxt)
         ls = os.listdir(directory)
-    ls.sort() ; lsDic = {}
+    ls.sort()
+    lsDic = {} # key is file w/out extension but INCLUDING any variant number.  Value is full filename if it's an extension we know about, "" if it's a file we can't process, or None if it's a directory (in which case key includes any 'extension' if the directory has one)
     has_variants = check_has_variants(directory,ls)
     for file in ls:
         filelower = file.lower()
-        # in lsDic if it's in the list (any extension); =filename if it's an extension we know about; =None if it's a directory (in which case the key is the full filename), ottherwise =""
-        if has_variants and file.find("_",file.find("_")+1)>-1: languageOverride=file[file.find("_")+1:file.find("_",file.find("_")+1)]
+        if filelower.endswith(dottxt) and checkIn((file+extsep)[:file.rfind(extsep)],lsDic): continue # don't let a .txt override a recording if both exist with same variant number
+        if has_variants and file.find("_",file.find("_")+1)>=0: languageOverride=file[file.find("_")+1:file.find("_",file.find("_")+1)] # for can_be_synthesized below
         else: languageOverride=None
-        if filelower.endswith(dottxt) and (file+extsep)[:file.rfind(extsep)] in lsDic: continue # don't let a .txt override a recording if both exist
-        if (filelower.endswith(dottxt) and file.find("_")>-1 and can_be_synthesized(file,directory,languageOverride)) or filelower.endswith(dotwav) or filelower.endswith(dotmp3): val = file
+        if (filelower.endswith(dottxt) and file.find("_")>=0 and can_be_synthesized(file,directory,languageOverride)) or filelower.endswith(dotwav) or filelower.endswith(dotmp3): val = file
         else:
             val = ""
             if filelower.endswith(extsep+"zip"): show_warning("Warning: Ignoring "+file+" (please unpack it first)") # so you can send someone a zip file for their recorded words folder and they'll know what's up if they don't unpack it
@@ -131,37 +131,49 @@ def getLsDic(directory):
                 lsDic[file]=None # a directory: store full name even if it has extsep in it.  Note however that we don't check isDirectory() if it's .wav etc as that would take too long.  (however some dirnames can contain dots)
                 # (+ NB need to store the directories specifically due to cases like course/ and course.pdf which may otherwise result in 2 traversals of "course" if we check isDirectory on 'extension is either none or unknown')
                 continue
-            elif (file+extsep)[:file.rfind(extsep)] in lsDic: continue # don't let a .txt~ or other unknown extension override a .txt
+            elif checkIn((file+extsep)[:file.rfind(extsep)],lsDic): continue # don't let a .txt~ or other unknown extension override a .txt
         lsDic[(file+extsep)[:file.rfind(extsep)]] = val # (this means if there's both mp3 and wav, wav will overwrite as comes later)
     if has_variants:
-        ls=list2set(ls) ; newVs = []
-        for k,v in lsDic.items():
+        ls=list2set(ls)
+        newVs = {} # variantFiles keys we added or changed
+        for k,v in list(lsDic.items()):
             # check for _lang_variant.ext and take out the _variant,
             # but keep them in variantFiles dict for fileToEvent to put back
-            if not v or (not directory==promptsDirectory and v.find("_explain_")>-1): continue # don't get confused by that
+            if not v or (not directory==promptsDirectory and v.find("_explain_")>=0): continue # skip directories, and don't get confused by explanation files
             last_ = v.rfind("_")
             if last_==-1: continue
             penult_ = v.rfind("_",0,last_)
             if penult_==-1: continue
-            del lsDic[k]
-            newK,newV = k[:k.rfind("_")], v[:v.rfind("_")]+v[v.rfind(extsep):]
-            if not newK in lsDic: lsDic[newK] = newV
-            else: # variants of different file types? better store them all under one (fileToEvent will sort out).  (Testing if the txt can be synth'd has already been done above)
-                if v.endswith(dottxt) and not lsDic[newK].endswith(dottxt): # if any variants are .txt then we'd better ensure the key is, so transliterate etc finds it. So move the key over to the .txt one.
-                    old_dirV = directory+os.sep+lsDic[newK]
+            # Now k = filename without extension but including a variant number, and v = full filename
+            del lsDic[k] # we don't want variant numbers in lsDic, we want them in variantFiles instead
+            newK,newV = k[:k.rfind("_")], v[:v.rfind("_")]+v[v.rfind(extsep):] # = k and v without the variant number (we'll add the real v to variantFiles[dir+newV] below, so it will be findable without variant number)
+            new_dirV = B(directory)+B(os.sep)+B(newV)
+            if not checkIn(newK,lsDic): # filename without variant number doesn't exist (for any extension)
+                lsDic[newK] = newV # so start it
+                assert not checkIn(new_dirV,variantFiles)
+                variantFiles[new_dirV] = [v]
+            elif v.endswith(dottxt) and not lsDic[newK].endswith(dottxt): # filename without variant number DOES exist (or we made the key when we saw a previous variant), and this new variant is .txt but the one without variant number is not.  If any variants are .txt then we'd better ensure the key maps to a .txt file (so transliterate etc finds it) and recordings are counted as variants of this .txt file, rather than .txt as variants of recordings.
+                old_dirV = B(directory+os.sep+lsDic[newK]) # the variantFiles key for the recording(s) we've already put in lsDic (but it'll be in variantFiles only if it HAD a variant number when we saw it, which won't be the case if the first variant had no number)
+                if checkIn(old_dirV,variantFiles):
                     d = variantFiles[old_dirV]
                     del variantFiles[old_dirV]
-                    lsDic[newK] = newV
-                    variantFiles[directory+os.sep+newV] = d
-                    lsDic[newK] = newV # just add to the previous key
-                else: newV = lsDic[newK]
-            dir_newV = directory+os.sep+newV
-            if not dir_newV in variantFiles:
-                variantFiles[dir_newV] = []
-                if newV in ls: variantFiles[dir_newV].append(newV) # the no-variants name is also a valid option
-            variantFiles[dir_newV].append(v)
-            newVs.append(dir_newV)
-        for v in newVs: random.shuffle(variantFiles[v])
+                    variantFiles[new_dirV] = d
+                else: variantFiles[new_dirV] = [B(lsDic[newK])] # the recording had no variant number, but now we know it does have variants, so put in the recording as first variant of the .txt key
+                variantFiles[new_dirV].append(v)
+                if checkIn(old_dirV,newVs):
+                    del newVs[old_dirV]
+                newVs[new_dirV] = 1
+                lsDic[newK] = newV
+            else: # filename without variant number does exist (or we made the key), and we need to add new variant
+                newV = lsDic[newK]
+                new_dirV = B(directory)+B(os.sep)+B(newV)
+                if not checkIn(new_dirV,variantFiles): # without variant number exists but isn't in variantFiles, so we need to add it as a variant before we add this new variant.  We know the key from lsDic.
+                    variantFiles[new_dirV] = [B(newV)]
+                variantFiles[new_dirV].append(v)
+            newVs[new_dirV]=1
+        for v in list(newVs.keys()):
+            assert checkIn(v,variantFiles), repr(sorted(list(variantFiles.keys())))+' '+repr(v)
+            random.shuffle(variantFiles[v])
     return lsDic
 
 def scanSamples_inner(directory,retVal,doLimit):
@@ -169,14 +181,14 @@ def scanSamples_inner(directory,retVal,doLimit):
     secLangSuffix = "_"+secondLanguage
     lsDic = getLsDic(directory)
     intro = intro_filename+"_"+firstLanguage
-    if intro in lsDic: dirsWithIntros.append((directory[len(samplesDirectory)+len(os.sep):],lsDic[intro]))
-    if not doLimit: doLimit = limit_filename in lsDic
-    doPoetry = poetry_filename in lsDic
+    if checkIn(intro,lsDic): dirsWithIntros.append((directory[len(samplesDirectory)+len(os.sep):],lsDic[intro]))
+    if not doLimit: doLimit = checkIn(limit_filename,lsDic)
+    doPoetry = checkIn(poetry_filename,lsDic)
     if doPoetry:
         # check which language the poetry is to be in (could be L1-to-L2, L2-to-L3, L2-only, or L3-only)
-        def poetry_language():
+        def poetry_language(firstLangSuffix,secLangSuffix,lsDic):
          ret = ""
-         for file,withExt in lsDic.items():
+         for file,withExt in list(lsDic.items()):
           if withExt:
             if file.endswith(secLangSuffix): ret=secLangSuffix # but stay in the loop
             elif (not file.endswith(firstLangSuffix)):
@@ -184,11 +196,11 @@ def scanSamples_inner(directory,retVal,doLimit):
                 for l in otherLanguages:
                     if not l in llist and file.endswith("_"+l): return "_"+l
          return ret
-        doPoetry = poetry_language()
+        doPoetry = poetry_language(firstLangSuffix,secLangSuffix,lsDic)
     prefix = directory[len(samplesDirectory)+cond(samplesDirectory,len(os.sep),0):] # the directory relative to samplesDirectory
     if prefix: prefix += os.sep
     lastFile = None # for doPoetry
-    items = lsDic.items() ; items.sort()
+    items = list(lsDic.items()) ; items.sort()
     for file,withExt in items:
         swapWithPrompt = 0
         if not withExt:
@@ -204,7 +216,7 @@ def scanSamples_inner(directory,retVal,doLimit):
                 wordSuffix=None
                 for l in otherLanguages:
                     if not l in [firstLanguage,secondLanguage] and file.endswith("_"+l):
-                        if l in otherFirstLanguages: swapWithPrompt=1
+                        if checkIn(l,otherFirstLanguages): swapWithPrompt=1
                         wordSuffix="_"+l ; break
                 if not wordSuffix: continue # can't do anything with this file
             if swapWithPrompt or firstLanguage==secondLanguage: promptFile=None
@@ -227,7 +239,7 @@ def scanSamples_inner(directory,retVal,doLimit):
                 # poetry without first-language prompts
                 if lastFile:
                     promptToAdd = prefix+lastFile[-1]
-                    if promptToAdd in singleLinePoems: del singleLinePoems[promptToAdd]
+                    if checkIn(promptToAdd,singleLinePoems): del singleLinePoems[promptToAdd]
                 else:
                     promptToAdd = prefix+withExt # 1st line is its own prompt
                     singleLinePoems[promptToAdd]=1
@@ -236,7 +248,7 @@ def scanSamples_inner(directory,retVal,doLimit):
             retVal.append((0,promptToAdd,prefix+withExt))
             if emptyCheck_hack: return
             if explanationFile: filesWithExplanations[prefix+withExt]=explanationFile
-            if doLimit: limitedFiles[prefix+withExt]=prefix
+            if doLimit: limitedFiles[B(prefix+withExt)]=prefix
             lastFile = [promptFile,withExt]
 
 cache_maintenance_mode=0 # hack so cache-synth.py etc can cache promptless words for use in justSynthesize, and words in prompts themselves
@@ -249,62 +261,62 @@ def parseSynthVocab(fname,forGUI=0):
     if not fileExists(fname): return []
     if not emptyCheck_hack: doLabel("Reading "+fname)
     allLangs = list2set([firstLanguage,secondLanguage]+otherLanguages)
-    for l in u8strip(read(fname)).replace("\r","\n").split("\n"):
+    for l in u8strip(read(fname)).replace(B("\r"),B("\n")).split(B("\n")):
         # TODO can we make this any faster on WinCE with large vocab lists? (tried SOME optimising already)
-        if not "=" in l: # might be a special instruction
+        if not B("=") in l: # might be a special instruction
             if not l: continue
-            canProcess = 0 ; l2=l.strip(wsp)
-            if not l2 or l2[0]=='#': continue
+            canProcess = 0 ; l2=bwspstrip(l)
+            if not l2 or l2[0:1]==B('#'): continue
             l2=l2.lower()
-            if l2.startswith("set language ") or l2.startswith("set languages "):
-                langs=l.split()[2:] ; someLangsUnknown = 0
+            if l2.startswith(B("set language ")) or l2.startswith(B("set languages ")):
+                langs=map(S,l.split()[2:]) ; someLangsUnknown = 0
                 maxsplit = len(langs)-1
                 for l in langs:
-                    if not l in allLangs: someLangsUnknown = 1
-            elif l2.startswith("limit on"):
+                    if not checkIn(l,allLangs): someLangsUnknown = 1
+            elif l2.startswith(B("limit on")):
                 doLimit = 1 ; limitNo += 1
-            elif l2.startswith("limit off"): doLimit = 0
-            elif l2.startswith("begin poetry"): doPoetry,lastPromptAndWord,disablePoem = True,None,False
-            elif l2.startswith("end poetry"): doPoetry = lastPromptAndWord = None
-            elif l2.startswith("poetry vocab line:"): doPoetry,lastPromptAndWord = 0,cond(lastPromptAndWord,lastPromptAndWord,0) # not None, in case we're at the very start of a poem (see "just processed"... at end)
+            elif l2.startswith(B("limit off")): doLimit = 0
+            elif l2.startswith(B("begin poetry")): doPoetry,lastPromptAndWord,disablePoem = True,None,False
+            elif l2.startswith(B("end poetry")): doPoetry = lastPromptAndWord = None
+            elif l2.startswith(B("poetry vocab line")): doPoetry,lastPromptAndWord = 0,cond(lastPromptAndWord,lastPromptAndWord,0) # not None, in case we're at the very start of a poem (see "just processed"... at end)
             else: canProcess=1
             if not canProcess: continue
-        elif "#" in l and l.strip(wsp)[0]=='#': continue # guard condition '"#" in l' improves speed
+        elif B('#') in l and bwspstrip(l)[0:1]==B('#'): continue # guard condition "'#' in l" improves speed
         if forGUI: strCount=""
         else:
             strCount = "%05d!synth:" % (count,)
             count += 1
-        langsAndWords = zip(langs,l.split("=",maxsplit)) # don't try strip on a map() - it's faster to do it as-needed below
+        langsAndWords = list(zip(langs,l.split(B("="),maxsplit))) # don't try strip on a map() - it's faster to do it as-needed below
         # (maxsplit means you can use '=' signs in the last language, e.g. if using SSML with eSpeak)
-        if someLangsUnknown: langsAndWords = filter(lambda x:x[0] in allLangs, langsAndWords)
+        if someLangsUnknown: langsAndWords = filter(lambda x,a=allLangs:checkIn(x[0],a), langsAndWords)
         # Work out what we'll use for the prompt.  It could be firstLanguage, or it could be one of the other languages if we see it twice (e.g. if 2nd language is listed twice then the second one will be the prompt for 2nd-language-to-2nd-language learning), or it could be the only language if we're simply listing words for cache maintenance
         if firstLanguage==secondLanguage: langsAlreadySeen = {}
         else: langsAlreadySeen = {firstLanguage:True}
-        def findPrompt():
+        def findPrompt(langsAndWords,langsAlreadySeen,doPoetry,strCount):
             i=0
             while i<len(langsAndWords):
                 lang,word = langsAndWords[i] ; i += 1
                 isReminder = cache_maintenance_mode and len(langsAndWords)==1 and not doPoetry
-                if (lang in langsAlreadySeen or isReminder) and (lang in getsynth_cache or can_be_synthesized("!synth:"+word+"_"+lang)): # (check cache because most of the time it'll be there and we don't need to go through all the text processing in can_be_synthesized)
+                if (lang in langsAlreadySeen or isReminder) and (lang in getsynth_cache or can_be_synthesized(B("!synth:")+B(word)+B("_")+B(lang))): # (check cache because most of the time it'll be there and we don't need to go through all the text processing in can_be_synthesized)
                     if not word: continue
-                    elif word[0] in wsp or word[-1] in wsp: word=word.strip(wsp) # avoid call if unnecessary
-                    return strCount+word+"_"+lang, cond(isReminder,0,i)
+                    elif word[0:1] in bwsp or word[-1:] in bwsp: word=bwspstrip(word) # avoid call if unnecessary
+                    return B(strCount)+word+B("_"+lang), cond(isReminder,0,i)
                 langsAlreadySeen[lang]=True
             return None,0
-        prompt,onePastPromptIndex = findPrompt()
+        prompt,onePastPromptIndex = findPrompt(langsAndWords,langsAlreadySeen,doPoetry,strCount)
         if not prompt and len(langsAndWords)>1: # 1st language prompt not found; try 2nd language to 3rd language etc
-            langsAlreadySeen = list2dict(otherFirstLanguages) ; prompt,onePastPromptIndex = findPrompt()
+            langsAlreadySeen = list2dict(otherFirstLanguages) ; prompt,onePastPromptIndex = findPrompt(langsAndWords,langsAlreadySeen,doPoetry,strCount)
             if not prompt:
-                langsAlreadySeen = {secondLanguage:True} ; prompt,onePastPromptIndex = findPrompt()
+                langsAlreadySeen = {secondLanguage:True} ; prompt,onePastPromptIndex = findPrompt(langsAndWords,langsAlreadySeen,doPoetry,strCount)
         prompt_L1only = prompt # before we possibly change it into a list etc.  (Actually not necessarily L1 see above, but usually is)
         if doPoetry:
             if prompt and lastPromptAndWord:
-                if lastPromptAndWord[0]: prompt=[lastPromptAndWord[0],prompt,lastPromptAndWord[1]] # L1 for line 1, L1 for line2, L2 for line 1
-                else: prompt=[lastPromptAndWord[1],prompt] # line 1 doesn't have L1 but line 2 does, so have L2 for line 1 + L1 for line 2
+                if lastPromptAndWord[0]: prompt=[S(lastPromptAndWord[0]),S(prompt),S(lastPromptAndWord[1])] # L1 for line 1, L1 for line2, L2 for line 1
+                else: prompt=[S(lastPromptAndWord[1]),S(prompt)] # line 1 doesn't have L1 but line 2 does, so have L2 for line 1 + L1 for line 2
             elif not prompt:
                 if lastPromptAndWord:
                     prompt=lastPromptAndWord[-1]
-                    if lastPromptAndWord[-1] in singleLinePoems: del singleLinePoems[lastPromptAndWord[-1]]
+                    if checkIn(lastPromptAndWord[-1],singleLinePoems): del singleLinePoems[lastPromptAndWord[-1]]
                 else:
                     prompt = 1 # file itself (see below)
         if prompt:
@@ -312,53 +324,55 @@ def parseSynthVocab(fname,forGUI=0):
             while i<len(langsAndWords):
                 lang,word = langsAndWords[i] ; i+=1
                 if i==onePastPromptIndex or (lang==firstLanguage and not firstLanguage==secondLanguage) or not word: continue # if 1st language occurs more than once (target as well as prompt) then don't get confused - this vocab file is probably being used with reverse settings
-                elif word[0] in wsp or word[-1] in wsp: word=word.strip(wsp) # avoid call if unnecessary
-                if lang in getsynth_cache or can_be_synthesized("!synth:"+word+"_"+lang):
+                elif word[0:1] in bwsp or word[-1:] in bwsp: word=bwspstrip(word) # avoid call if unnecessary
+                if checkIn(lang,getsynth_cache) or can_be_synthesized(B("!synth:")+word+B("_"+lang)):
                   if not (doPoetry and disablePoem):
-                    f=strCount+word+"_"+lang
+                    f=B(strCount)+word+B("_"+lang)
                     if prompt==1 or prompt==f: # a file with itself as the prompt (either explicitly or by omitting any other prompt)
                         prompt=f
                         singleLinePoems[f]=1
-                    ret.append((0,prompt,f))
+                    ret.append((0,S(prompt),S(f)))
                     if emptyCheck_hack: return ret
-                    if doLimit: limitedFiles[f]="synth:"+str(limitNo)
+                    if doLimit: limitedFiles[f]=B("synth:"+str(limitNo))
                     if doPoetry: lastPromptAndWord = [prompt_L1only,f]
                 elif doPoetry: disablePoem=1 # if one of the lines can't be synth'd, disable the rest of the poem (otherwise get wrongly connected lines, disconnected lines, or re-introduction of isolated lines that were previously part of a poem but can't be synth'd on this platform)
-        if not lastPromptAndWord==None: doPoetry = 1 # just processed a "poetry vocab line:" (lastPromptAndWord is either the real last prompt and word, or 0 if we were at the start)
+        if not lastPromptAndWord==None: doPoetry = 1 # just processed a "poetry vocab line" (lastPromptAndWord is either the real last prompt and word, or 0 if we were at the start)
     return ret
 
 def sanitise_otherLanguages():
     for l in otherFirstLanguages:
-        if not l in otherLanguages: otherLanguages.append(l)
+        if not checkIn(l,otherLanguages): otherLanguages.append(l)
     for l in otherLanguages:
-        if not l in possible_otherLanguages: possible_otherLanguages.append(l)
+        if not checkIn(l,possible_otherLanguages): possible_otherLanguages.append(l)
 sanitise_otherLanguages()
 
 # Prompt file syntax: word_language.wav
 # or: word_language_2.wav .. (alternatives chosen at random)
 # ('word' can also be a language name)
-class PromptException(Exception):
+class MessageException(Exception):
     def __init__(self,message): self.message = message
     def __repr__(self): return self.message
+class PromptException(MessageException): pass
+auto_advancedPrompt=0 # used by gradint.cgi
 class AvailablePrompts(object):
     reservedPrefixes = list2set(map(lambda x:x.lower(),["whatmean","meaningis","repeatAfterMe","sayAgain","longPause","begin","end",firstLanguage,secondLanguage] + possible_otherLanguages))
     def __init__(self):
         self.lsDic = getLsDic(promptsDirectory)
         self.prefixes = {}
-        for k,v in self.lsDic.items():
+        for k,v in list(self.lsDic.items()):
             if v: self.prefixes[k[:k.rfind("_")]]=1 # delete language
             else: del self.lsDic[k] # !poetry etc doesn't make sense in prompts
-        self.prefixes = self.prefixes.keys()
+        self.prefixes = list(self.prefixes.keys())
         self.user_is_advanced = None
     def getRandomPromptList(self,promptsData,language):
         random.shuffle(self.prefixes)
         for p in self.prefixes:
-            if p.lower() in self.reservedPrefixes: continue
+            if checkIn(p.lower(),self.reservedPrefixes): continue
             try:
                 theList = self.getPromptList(p,promptsData,language)
                 return theList
             except PromptException: pass
-        raise PromptException("Can't find a non-reserved prompt suitable for language '%s'" % (language))
+        raise PromptException("Can't find a non-reserved prompt suitable for language '%s'. Try creating tryToSay_%s%s etc in %s" % (language,language,dotwav,promptsDirectory))
     def getPromptList(self,prefix,promptsData,language):
         # used for introducing foreign-language prompts to
         # beginners.  language is the suffix of the language we're *learning*.
@@ -368,11 +382,11 @@ class AvailablePrompts(object):
                 if p > advancedPromptThreshold2:
                     self.user_is_advanced = 1 ; break # got a reasonably advanced user
         beginnerPrompt = prefix+"_"+firstLanguage
-        if not beginnerPrompt in self.lsDic:
+        if not checkIn(beginnerPrompt,self.lsDic):
             if self.user_is_advanced and not language==secondLanguage and prefix+"_"+secondLanguage in self.lsDic: beginnerPrompt=prefix+"_"+secondLanguage # No first language prompt, but in advanced mode may be able to find a second-language prompt for a 3rd language
             else: beginnerPrompt = None
         advancedPrompt = prefix+"_"+language
-        if not advancedPrompt in self.lsDic:
+        if not checkIn(advancedPrompt,self.lsDic):
             # Must use beginnerPrompt
             if beginnerPrompt: r=[self.lsDic[beginnerPrompt]]
             else:
@@ -380,7 +394,7 @@ class AvailablePrompts(object):
                 else: raise PromptException("Can't find "+prefix+"_"+language+", "+prefix+"_"+firstLanguage+" or "+prefix+"_"+secondLanguage)
         elif not beginnerPrompt:
             # Must use advancedPrompt
-            if (not self.user_is_advanced) and cond(language==secondLanguage,advancedPromptThreshold,advancedPromptThreshold2): raise PromptException("Prompt '%s' is too advanced; need '%s_%s' (unless you set %s=0 in advanced%stxt)" % (advancedPrompt,prefix,firstLanguage,cond(language==secondLanguage,"advancedPromptThreshold","advancedPromptThreshold2"),extsep))
+            if (not self.user_is_advanced) and not auto_advancedPrompt and cond(language==secondLanguage,advancedPromptThreshold,advancedPromptThreshold2): raise PromptException("Prompt '%s' is too advanced; need '%s_%s' (unless you set %s=0 in advanced%stxt)" % (advancedPrompt,prefix,firstLanguage,cond(language==secondLanguage,"advancedPromptThreshold","advancedPromptThreshold2"),extsep))
             r=[self.lsDic[advancedPrompt]]
         elif promptsData.get(advancedPrompt,0) >= cond(language==secondLanguage,advancedPromptThreshold,advancedPromptThreshold2): r=[self.lsDic[advancedPrompt]]
         elif promptsData.get(advancedPrompt,0) >= cond(language==secondLanguage,transitionPromptThreshold,transitionPromptThreshold2): r=[self.lsDic[advancedPrompt], self.lsDic[beginnerPrompt]]
@@ -393,7 +407,7 @@ class AvailablePrompts(object):
         # Increment advancedPrompt, taking care not to go
         # past the threshold if it's not available yet
         adv = promptsData.get(advancedPrompt,0)
-        if advancedPrompt in self.lsDic or adv <= cond(language==secondLanguage,transitionPromptThreshold,transitionPromptThreshold2):
+        if checkIn(advancedPrompt,self.lsDic) or adv <= cond(language==secondLanguage,transitionPromptThreshold,transitionPromptThreshold2):
             adv += 1
         promptsData[advancedPrompt] = adv
         # and finally,
@@ -406,14 +420,14 @@ def introductions(zhFile,progressData):
     for d,fname in dirsWithIntros[:]:
         found = 0
         for p in progressData:
-            if p[-1].startswith(d) and p[0]:
+            if B(p[-1]).startswith(B(d)) and p[0]:
                 # this dir has already been introduced
                 found=1 ; dirsWithIntros.remove((d,fname)) ; break
         if found: continue
-        if zhFile.startswith(d): toIntroduce.append((d,fname))
+        if B(zhFile).startswith(B(d)): toIntroduce.append((d,fname))
     toIntroduce.sort() # should put shorter ones 1st
-    return map(lambda (x,fname): fileToEvent(cond(x,x+os.sep,"")+fname), toIntroduce)
+    return map(lambda x: fileToEvent(cond(x[0],x[0]+os.sep,"")+x[1]), toIntroduce)
 
 def explanations(zhFile):
-    if zhFile in filesWithExplanations: return fileToEvent(zhFile.replace(dotmp3,dotwav).replace(dottxt,dotwav).replace(dotwav,"_explain_"+firstLanguage+filesWithExplanations[zhFile][-len(dotwav):]))
+    if checkIn(zhFile,filesWithExplanations): return fileToEvent(zhFile.replace(dotmp3,dotwav).replace(dottxt,dotwav).replace(dotwav,"_explain_"+firstLanguage+filesWithExplanations[zhFile][-len(dotwav):]))
 

gradint-build/src/frontend.py → src/frontend.py

-# This file is part of the source code of
-# gradint v0.9979 (c) 2002-2011 Silas S. Brown. GPL v3+.
+# This file is part of the source code of Gradint
+# (c) Silas S. Brown.
 #    This program is free software; you can redistribute it and/or modify
 #    it under the terms of the GNU General Public License as published by
 #    the Free Software Foundation; either version 3 of the License, or
@@ -23,17 +23,19 @@ appTitle += time.strftime(" %A") # in case leave 2+ instances on the desktop
 def waitOnMessage(msg):
     global warnings_printed
     if type(msg)==type(u""): msg2=msg.encode("utf-8")
-    else: msg2=msg
+    else:
+        try: msg2,msg=msg,msg.decode("utf-8")
+        except AttributeError: msg2=msg # Python 2.1 has no .decode
     if appuifw:
         t=appuifw.Text() ; t.add(u"".join(warnings_printed)+msg) ; appuifw.app.body = t # in case won't fit in the query()  (and don't use note() because it doesn't wait)
-        appuifw.query(u""+msg,'query')
+        appuifw.query(msg,'query')
     elif android:
         # android.notify("Gradint","".join(warnings_printed)+msg) # doesn't work?
         android.dialogCreateAlert("Gradint","".join(warnings_printed)+msg)
         android.dialogSetPositiveButtonText("OK")
         android.dialogShow() ; android.dialogGetResponse()
     elif app:
-        if not (winsound or winCEsound or mingw32 or cygwin): show_info(msg2+"\n\nWaiting for you to press OK on the message box... ",True) # in case terminal is in front
+        if not (winsound or winCEsound or mingw32 or cygwin): show_info(msg2+B("\n\nWaiting for you to press OK on the message box... "),True) # in case terminal is in front
         app.todo.alert = "".join(warnings_printed)+msg
         while True:
             try:
@@ -42,8 +44,8 @@ def waitOnMessage(msg):
             time.sleep(0.5)
         if not (winsound or winCEsound or mingw32 or cygwin): show_info("OK\n",True)
     else:
-        if clearScreen(): msg2 = "This is "+program_name.replace("(c)","\n(c)")+"\n\n"+msg2 # clear screen is less confusing for beginners, but NB it may not happen if warnings etc
-        show_info(msg2+"\n\n"+cond(winCEsound,"Press OK to continue\n","Press Enter to continue\n"))
+        if clearScreen(): msg2 = B("This is "+program_name.replace("(c)","\n(c)")+"\n\n")+msg2 # clear screen is less confusing for beginners, but NB it may not happen if warnings etc
+        show_info(msg2+B("\n\n"+cond(winCEsound,"Press OK to continue\n","Press Enter to continue\n")))
         sys.stderr.flush() # hack because some systems don't do it (e.g. some mingw32 builds), and we don't want the user to fail to see why the program is waiting (especially when there's an error)
         try:
             raw_input(cond(winCEsound,"See message under this window.","")) # (WinCE uses boxes for raw_input so may need to repeat the message - but can't because the prompt is size-limited, so need to say look under the window)
@@ -54,13 +56,14 @@ def waitOnMessage(msg):
 def getYN(msg,defaultIfEof="n"):
     if appuifw:
         appuifw.app.body = None
-        return appuifw.query(u""+msg,'query')
+        return appuifw.query(ensure_unicode(msg),'query')
     elif android:
         android.dialogCreateAlert("Gradint",msg)
         android.dialogSetPositiveButtonText("Yes") # TODO do we have to localise this ourselves or can we have a platform default?
         android.dialogSetNegativeButtonText("No")
         android.dialogShow()
-        return android.dialogGetResponse().result['which'] == 'positive'
+        try: return android.dialogGetResponse().result['which'] == 'positive'
+        except KeyError: return 0 # or raise SystemExit, no 'which'
     elif app:
         app.todo.question = localise(msg)
         while app and not hasattr(app,"answer_given"): time.sleep(0.5)
@@ -74,7 +77,7 @@ def getYN(msg,defaultIfEof="n"):
         while not ans=='y' and not ans=='n':
             try: ans = raw_input("%s\nPress y for yes, or n for no.  Then press Enter.  --> " % (msg,))
             except EOFError:
-                ans=defaultIfEof ; print ans
+                ans=defaultIfEof ; print (ans)
         clearScreen() # less confusing for beginners
         if ans=='y': return 1
         return 0
@@ -89,7 +92,7 @@ def primitive_synthloop():
         old_js = justSynthesize
         if appuifw:
             if not justSynthesize: justSynthesize=""
-            justSynthesize=appuifw.query(u"Say:","text",u""+justSynthesize)
+            justSynthesize=appuifw.query(u"Say:","text",ensure_unicode(justSynthesize))
             if justSynthesize: justSynthesize=justSynthesize.encode("utf-8")
             else: break
         else:
@@ -102,18 +105,18 @@ def primitive_synthloop():
             if (winCEsound or riscos_sound or android) and not justSynthesize: break # because no way to send EOF (and we won't be taking i/p from a file)
             if interactive and not readline:
               interactive="('a' for again) Say: "
-              if justSynthesize=="a": justSynthesize=old_js
+              if B(justSynthesize)==B("a"): justSynthesize=old_js
         oldLang = lang
-        if justSynthesize: lang = just_synthesize(interactive,lang)
+        if justSynthesize: lang = S(just_synthesize(interactive,lang))
         # and see if it transliterates:
-        if justSynthesize and lang and not "#" in justSynthesize:
-            if justSynthesize.startswith(lang+" "):
+        if justSynthesize and lang and not B('#') in B(justSynthesize):
+            if B(justSynthesize).startswith(B(lang)+B(" ")):
                 t = transliterates_differently(justSynthesize[len(lang+" "):],lang)
                 if t: t=lang+" "+t
             else: t = transliterates_differently(justSynthesize,lang)
             if t:
                 if appuifw: justSynthesize = t
-                else: show_info("Spoken as "+t+"\n")
+                else: show_info(B("Spoken as ")+t+B("\n"))
         if warnings_printed: # at end not beginning, because don't want to overwrite the info message if appuifw
             if appuifw:
                 t=appuifw.Text()
@@ -124,6 +127,9 @@ def primitive_synthloop():
             warnings_printed = []
         if not lang: lang=oldLang
 
+if android:
+  if not isDirectory("/mnt/sdcard/svox") and not isDirectory("/system/tts/lang_pico"): waitOnMessage("English voice might not be installed. Check under Home > Menu > Settings > Voice output > text to speech > Pico > English")
+
 def startBrowser(url): # true if success
   if winCEsound: return None # user might be paying per byte! + difficult to switch back if no Alt-Tab program
   try:
@@ -131,7 +137,7 @@ def startBrowser(url): # true if success
       g=webbrowser.get()
   except: g=0
   if g and (winCEsound or macsound or (hasattr(g,"background") and g.background) or (hasattr(webbrowser,"BackgroundBrowser") and g.__class__==webbrowser.BackgroundBrowser) or (hasattr(webbrowser,"Konqueror") and g.__class__==webbrowser.Konqueror)):
-      return g.open_new(url)
+      return g.open_new(S(url))
   # else don't risk it - it might be text-mode and unsuitable for multitask-with-gradint
   if winsound: return not os.system('start "%ProgramFiles%\\Internet Explorer\\iexplore.exe" '+url) # use os.system not system here (don't know why but system() doesn't always work for IE)
   # (NB DON'T replace % with %%, it doesn't work. just hope nobody set an environment variable to any hex code we're using in mp3web)
@@ -144,7 +150,7 @@ def clearScreen():
         warnings_printed = []
         return
     if winsound or mingw32: os.system("cls")
-    else: os.system("clear 1>&2") # (1>&2 in case using stdout for something else)
+    else: os.system("clear >&2") # (>&2 in case using stdout for something else)
     return True
 
 cancelledFiles = []
@@ -164,17 +170,17 @@ def handleInterrupt(): # called only if there was an interrupt while the runner
         # cancelledEvent = runner.queue[0][-1][0] worked in python 2.3, but sched implementation seems to have changed in python 2.5 so we're using copy_of_runner_events instead
         if hasattr(cancelledEvent,"wordToCancel") and cancelledEvent.wordToCancel: cancelledFiles.append(cancelledEvent.wordToCancel)
     if not app and needCountItems and cancelledFiles: show_info("(%d cancelled items)...\n" % len(cancelledFiles))
-    global repeatMode ; repeatMode = 0 # so Ctrl-C on justSynth-with-R works
+    global repeatMode ; repeatMode = "interrupted"
 
 tkNumWordsToShow = 10 # the default number of list-box items
 
 def addStatus(widget,status,mouseOnly=0):
     # Be VERY CAREFUL with status line changes.  Don't do it on things that are focused by default (except with mouseOnly=1).  Don't do it when the default status line might be the widest thing (i.e. when list box is not displayed) or window size could jump about too much.  And in any event don't use lines longer than about 53 characters (the approx default width of the listbox when using monospace fonts).
     # (NB addStatus now takes effect only when the list box is displayed anyway, so OK for buttons that might also be displayed without it)
-    widget.bind('<Enter>',lambda *args:app.set_statusline(status))
+    widget.bind('<Enter>',lambda e=None,status=status:app.set_statusline(status))
     widget.bind('<Leave>',app.restore_statusline)
     if not mouseOnly:
-        widget.bind('<FocusIn>',lambda *args:app.set_statusline(status))
+        widget.bind('<FocusIn>',lambda e=None,app=app,status=status:app.set_statusline(status))
         widget.bind('<FocusOut>',app.restore_statusline)
 def makeButton(parent,text,command):
     button = Tkinter.Button(parent)
@@ -199,12 +205,17 @@ def addLabel(row,label):
 def CXVMenu(e): # callback for right-click
     e.widget.focus()
     m=Tkinter.Menu(None, tearoff=0, takefocus=0)
-    ctrl=cond(macsound,"<Command-","<Control-")
-    m.add_command(label="Cut",command=(lambda e=e: e.widget.event_generate(ctrl+'x>')))
-    m.add_command(label="Copy",command=(lambda e=e: e.widget.event_generate(ctrl+'-c>')))
-    m.add_command(label="Paste",command=(lambda e=e: e.widget.event_generate(ctrl+'-v>')))
-    m.add_command(label="Delete",command=(lambda e=e: e.widget.event_generate('<Delete>')))
-    m.add_command(label="Select All",command=(lambda e=e: selectAll(e)))
+    if macsound:
+        cut,copy,paste = "<<Cut>>","<<Copy>>","<<Paste>>"
+    else:
+        ctrl="<Control-"
+        cut,copy,paste = ctrl+'x>',ctrl+'c>',ctrl+'v>'
+    def evgen(e,cmd): e.widget.event_generate(cmd)
+    funclist = [("Paste",paste),("Delete",'<Delete>')]
+    if not macsound:
+        funclist = [("Cut",cut),("Copy",copy)]+funclist # doesn't work reliably on Mac Tk
+    for l,cmd in funclist: m.add_command(label=l,command=(lambda e=e,c=cmd,evgen=evgen: e.widget.after(10,evgen,e,c)))
+    m.add_command(label="Select All",command=(lambda e=e: e.widget.after(10,selectAll,e)))
     m.tk_popup(e.x_root-3, e.y_root+3,entry="0")
 def selectAll(e):
     e.widget.event_generate('<Home>')
@@ -218,13 +229,15 @@ def addTextBox(row,wide=0):
     text = Tkinter.StringVar(row)
     entry = Tkinter.Entry(row,textvariable=text)
     entry.bind('<ButtonRelease-3>',CXVMenu)
-    if macsound: entry.bind('<Control-ButtonRelease-1>',CXVMenu)
+    if macsound:
+        entry.bind('<Control-ButtonRelease-1>',CXVMenu)
+        entry.bind('<ButtonRelease-2>',CXVMenu)
     if winCEsound:
       if WMstandard: # non-numeric inputs no good on WMstandard Tkinter
         def doRawInput(text,entry):
             app.input_to_set = text
             app.menu_response="input"
-        entry.bind('<Return>',lambda e:doRawInput(text,entry))
+        entry.bind('<Return>',lambda e,doRawInput=doRawInput,text=text,entry=entry:doRawInput(text,entry))
         if wide: # put help in 1st wide textbox
           global had_doRawInput
           try: had_doRawInput
@@ -233,10 +246,10 @@ def addTextBox(row,wide=0):
             text.set("(Push OK to type A-Z)") # (if changing this message, change it below too)
             class E: pass
             e=E() ; e.widget = entry
-            entry.after(10,lambda *args:selectAll(e))
+            entry.after(10,lambda _=None,e=e:selectAll(e))
       else: # PocketPC: try to detect long clicks. This is awkward. time.time is probably 1sec resolution so will get false +ves if go by that only.
         def timeStamp(entry): entry.buttonPressTime=time.time()
-        entry.bind('<ButtonPress-1>',lambda e:timeStamp(entry))
+        entry.bind('<ButtonPress-1>',lambda e,timeStamp=timeStamp,entry=entry:timeStamp(entry))
         global lastDblclkAdvisory,lastDblclk
         lastDblclkAdvisory=lastDblclk=0
         def pasteInstructions(t):
@@ -248,8 +261,8 @@ def addTextBox(row,wide=0):
         def doPaste(text,entry):
             text.set(entry.selection_get(selection="CLIPBOARD"))
             global lastDblclk ; lastDblclk=time.time()
-        entry.bind('<ButtonRelease-1>',lambda e:pasteInstructions(time.time()-getattr(entry,"buttonPressTime",time.time())))
-        entry.bind('<Double-Button-1>',lambda e:doPaste(text,entry))
+        entry.bind('<ButtonRelease-1>',lambda e,entry=entry,pasteInstructions=pasteInstructions:pasteInstructions(time.time()-getattr(entry,"buttonPressTime",time.time())))
+        entry.bind('<Double-Button-1>',lambda e,doPaste=doPaste,text=text,entry=entry:doPaste(text,entry))
     # Tkinter bug workaround (some versions): event_generate from within a key event handler can be unreliable, so the Ctrl-A handler delays selectAll by 10ms:
     entry.bind(cond(macsound,'<Command-a>','<Control-a>'),(lambda e:e.widget.after(10,lambda e=e:selectAll(e))))
     bindUpDown(entry,False)
@@ -286,14 +299,14 @@ def make_output_row(parent):
     # if there aren't any options then return None
     # we also put script-variant selection here, if any
     row = None
-    def getRow(row):
+    def getRow(row,parent):
       if not row:
         row = Tkinter.Frame(parent)
         row.pack(fill=Tkinter.X,expand=1)
       return row
     GUIlang = GUI_languages.get(firstLanguage,firstLanguage)
-    if "@variants-"+GUIlang in GUI_translations: # the firstLanguage has script variants
-        row=getRow(row)
+    if checkIn("@variants-"+GUIlang,GUI_translations): # the firstLanguage has script variants
+        row=getRow(row,parent)
         if not hasattr(app,"scriptVariant"): app.scriptVariant = Tkinter.StringVar(app)
         count = 0
         for variant in GUI_translations["@variants-"+GUIlang]:
@@ -301,7 +314,7 @@ def make_output_row(parent):
             count += 1
         app.scriptVariant.set(str(scriptVariants.get(GUIlang,0)))
     if synth_partials_voices and guiVoiceOptions:
-        row=getRow(row)
+        row=getRow(row,parent)
         if not hasattr(app,"voiceOption"): app.voiceOption = Tkinter.StringVar(app)
         Tkinter.Radiobutton(row, text=u" Normal ", variable=app.voiceOption, value="", indicatoron=forceRadio).pack({"side":"left"})
         for o in guiVoiceOptions: Tkinter.Radiobutton(row, text=u" "+o[1].upper()+o[2:]+u" ", variable=app.voiceOption, value=o, indicatoron=forceRadio).pack({"side":"left"})
@@ -310,9 +323,9 @@ def make_output_row(parent):
     if not hasattr(app,"outputTo"):
         app.outputTo = Tkinter.StringVar(app) # NB app not parent (as parent is no longer app)
         app.outputTo.set("0") # not "" or get tri-state boxes on OS X 10.6
-    row=getRow(row)
+    row=getRow(row,parent)
     rightrow = addRightRow(row) # to show beginners this row probably isn't the most important thing despite being in a convenient place, we'll right-align
-    def addFiletypeButton(fileType):
+    def addFiletypeButton(fileType,rightrow):
         ftu = fileType.upper()
         t = Tkinter.Radiobutton(rightrow, text=cond(forceRadio,""," ")+ftu+" ", variable=app.outputTo, value=fileType, indicatoron=forceRadio)
         bindUpDown(t,True)
@@ -326,13 +339,13 @@ def make_output_row(parent):
     addStatus(t,"Select this to send all sounds to\nthe speaker, not to files on disk")
     bindUpDown(t,True)
     t.pack({"side":"left"})
-    if got_program("lame"): addFiletypeButton("mp3")
-    if got_windows_encoder: addFiletypeButton("wma")
-    if got_program("neroAacEnc") or got_program("faac") or got_program("afconvert"): addFiletypeButton("aac")
-    if got_program("oggenc"): addFiletypeButton("ogg")
-    if got_program("toolame"): addFiletypeButton("mp2")
-    if got_program("speexenc"): addFiletypeButton("spx")
-    addFiletypeButton("wav")
+    if got_program("lame"): addFiletypeButton("mp3",rightrow)
+    if got_windows_encoder: addFiletypeButton("wma",rightrow)
+    if got_program("neroAacEnc") or got_program("faac") or got_program("afconvert"): addFiletypeButton("aac",rightrow)
+    if got_program("oggenc") or got_program("oggenc2"): addFiletypeButton("ogg",rightrow)
+    if got_program("toolame"): addFiletypeButton("mp2",rightrow)
+    if got_program("speexenc"): addFiletypeButton("spx",rightrow)
+    addFiletypeButton("wav",rightrow)
     # "Get MP3 encoder" and "Get WMA encoder" changed to "MP3..." and "WMA..." to save width (+ no localisation necessary)
     if unix and not got_program("lame") and got_program("make") and got_program("gcc") and (got_program("curl") or got_program("wget")): addButton(rightrow,"MP3...",app.getEncoder,status="Press this to compile an MP3 encoder\nso Gradint can output to MP3 files") # (checking gcc as well as make because some distros strangely have make but no compiler; TODO what if has a non-gcc compiler)
     # (no longer available) elif (winsound or mingw32) and not got_windows_encoder and not got_program("lame"): addButton(rightrow,"WMA...",app.getEncoder,status="Press this to download a WMA encoder\nso Gradint can output to WMA files")
@@ -341,19 +354,19 @@ def make_output_row(parent):
 def updateSettingsFile(fname,newVals):
     # leaves comments etc intact, but TODO does not cope with changing variables that have been split over multiple lines
     replacement_lines = []
-    try: oldLines=u8strip(read(fname)).replace("\r\n","\n").split("\n")
+    try: oldLines=u8strip(read(fname)).replace(B("\r\n"),B("\n")).split(B("\n"))
     except IOError: oldLines=[]
     for l in oldLines:
         found=0
-        for k in newVals.keys():
-            if l.startswith(k):
-                replacement_lines.append(k+"="+repr(newVals[k]))
+        for k in list(newVals.keys()):
+            if l.startswith(B(k)):
+                replacement_lines.append(B(k+"="+repr(newVals[k])))
                 del newVals[k]
                 found=1
         if not found: replacement_lines.append(l)
-    for k,v in newVals.items(): replacement_lines.append(k+"="+repr(v))
-    if replacement_lines and replacement_lines[-1]: replacement_lines.append("") # ensure blank line at end so there's a \n but we don't add 1 more with each save
-    open(fname,"w").write("\n".join(replacement_lines))
+    for k,v in list(newVals.items()): replacement_lines.append(B(k+"="+repr(v)))
+    if replacement_lines and replacement_lines[-1]: replacement_lines.append(B("")) # ensure blank line at end so there's a \n but we don't add 1 more with each save
+    writeB(open(fname,"w"),B("\n").join(replacement_lines))
 
 def asUnicode(x): # for handling the return value of Tkinter entry.get()
     try: return u""+x # original behaviour
@@ -362,16 +375,27 @@ def asUnicode(x): # for handling the return value of Tkinter entry.get()
         except: return x.decode("iso-8859-1") # TODO can we get what it actually IS? (on German WinXP, sys.getdefaultencoding==ascii and locale==C but Tkinter still returns Latin1)
 
 def setupScrollbar(parent,rowNo):
+    onLeft = winCEsound or olpc
     s = Tkinter.Scrollbar(parent,takefocus=0)
-    s.grid(row=rowNo,column=cond(winCEsound or olpc,0,1),sticky="ns"+cond(winCEsound or olpc,"w","e"))
+    s.grid(row=rowNo,column=cond(onLeft,0,1),sticky="ns"+cond(onLeft,"w","e"))
+    try: parent.rowconfigure(rowNo,weight=1)
+    except: pass
     c=Tkinter.Canvas(parent,bd=0,width=200,height=100,yscrollcommand=s.set)
-    c.grid(row=rowNo,column=cond(winCEsound or olpc,1,0),sticky="nsw")
+    c.grid(row=rowNo,column=cond(onLeft,1,0),sticky="nsw")
     s.config(command=c.yview)
     scrolledFrame=Tkinter.Frame(c) ; c.create_window(0,0,window=scrolledFrame,anchor="nw")
     # Mousewheel binding.  TODO the following bind_all assumes only one scrolledFrame on screen at once (redirect all mousewheel events to the frame; necessary as otherwise they'll go to buttons etc)
-    scrolledFrame.bind_all('<Button-4>',lambda *args:c.yview("scroll","-1","units"))
-    scrolledFrame.bind_all('<Button-5>',lambda *args:c.yview("scroll","1","units"))
-    # DON'T bind <MouseWheel> on Windows - our version of Tk will segfault when it occurs. See http://mail.python.org/pipermail/python-bugs-list/2005-May/028768.html but we can't patch our library.zip's Tkinter anymore (TODO can we use newer Tk DLLs and ensure setup.bat updates them?)
+    app.ScrollUpHandler = lambda e=None,c=c:c.yview("scroll","-1","units")
+    app.ScrollDownHandler = lambda e=None,c=c:c.yview("scroll","1","units")
+    if macsound:
+        def ScrollHandler(event):
+            if event.delta>0: app.ScrollUpHandler()
+            else: app.ScrollDownHandler()
+        scrolledFrame.bind_all('<MouseWheel>',ScrollHandler)
+        # DON'T bind <MouseWheel> on Windows - our version of Tk will segfault when it occurs. See http://mail.python.org/pipermail/python-bugs-list/2005-May/028768.html but we can't patch our library.zip's Tkinter anymore (TODO can we use newer Tk DLLs and ensure setup.bat updates them?)
+    else: # for X11:
+        scrolledFrame.bind_all('<Button-4>',app.ScrollUpHandler)
+        scrolledFrame.bind_all('<Button-5>',app.ScrollDownHandler)
     return scrolledFrame, c
 
 # GUI presets buttons:
@@ -379,8 +403,8 @@ shortDescriptionName = "short-description"+dottxt
 longDescriptionName = "long-description"+dottxt
 class ExtraButton(object):
     def __init__(self,directory):
-        self.shortDescription = u8strip(read(directory+os.sep+shortDescriptionName)).strip(wsp)
-        if fileExists(directory+os.sep+longDescriptionName): self.longDescription = u8strip(read(directory+os.sep+longDescriptionName)).strip(wsp)
+        self.shortDescription = wspstrip(u8strip(read(directory+os.sep+shortDescriptionName)))
+        if fileExists(directory+os.sep+longDescriptionName): self.longDescription = wspstrip(u8strip(read(directory+os.sep+longDescriptionName)))
         else: self.longDescription = self.shortDescription
         self.directory = directory
     def add(self):
@@ -397,7 +421,7 @@ class ExtraButton(object):
         try: ls = os.listdir(samplesDirectory)
         except: os.mkdir(samplesDirectory)
         name1=newName
-        while newName in ls: newName+="1"
+        while checkIn(newName,ls): newName+="1"
         name2=newName
         newName = samplesDirectory+os.sep+newName
         os.rename(self.directory,newName)
@@ -406,13 +430,13 @@ class ExtraButton(object):
             which_collection = localise(" has been added to your collection.")
             o=open(vocabFile,"a")
             o.write("# --- BEGIN "+self.shortDescription+" ---\n")
-            o.write(u8strip(read(newName+os.sep+"add-to-vocab"+dottxt)).strip(wsp)+"\n")
+            o.write(wspstrip(u8strip(read(newName+os.sep+"add-to-vocab"+dottxt)))+"\n")
             o.write("# ----- END "+self.shortDescription+" ---\n")
             if hasattr(app,"vocabList"): del app.vocabList # so re-reads
             os.remove(newName+os.sep+"add-to-vocab"+dottxt)
         if fileExists(newName+os.sep+"add-to-languages"+dottxt):
             changed = 0
-            for lang in u8strip(read(newName+os.sep+"add-to-languages"+dottxt)).strip(wsp).split():
+            for lang in wspstrip(u8strip(read(newName+os.sep+"add-to-languages"+dottxt))).split():
                 if not lang in [firstLanguage,secondLanguage]+otherLanguages:
                     otherLanguages.append(lang) ; changed = 1
             if changed: sanitise_otherLanguages(), updateSettingsFile("advanced"+dottxt,{"otherLanguages":otherLanguages,"possible_otherLanguages":possible_otherLanguages})
@@ -450,8 +474,8 @@ def focusButton(button):
             try: button.config(state=state)
             except: pass # maybe not a button
         for t in range(250,1000,250): # (NB avoid epilepsy's 5-30Hz!)
-          app.after(t,lambda *args:flashButton(button,"active"))
-          app.after(t+150,lambda *args:flashButton(button,"normal"))
+          app.after(t,lambda e=None,flashButton=flashButton,button=button:flashButton(button,"active"))
+          app.after(t+150,lambda e=None,flashButton=flashButton,button=button:flashButton(button,"normal"))
         # (Don't like flashing, but can't make it permanently active as it won't change when the focus does)
 
 if WMstandard: GUI_omit_statusline = 1 # unlikely to be room (and can disrupt nav)
@@ -462,16 +486,24 @@ def startTk():
             Tkinter.Frame.__init__(self, master)
             class EmptyClass: pass
             self.todo = EmptyClass() ; self.toRestore = []
+            self.ScrollUpHandler = self.ScrollDownHandler = lambda e=None:True
             global app ; app = self
             make_extra_buttons_waiting_list()
             if olpc: self.master.option_add('*font',cond(extra_buttons_waiting_list,'Helvetica 9','Helvetica 14'))
-            elif macsound and Tkinter.TkVersion>=8.6: self.master.option_add('*font','System 13') # ok with magnification.  Note >13 causes square buttons.  (Including this line causes "Big print" to work)
+            elif macsound:
+                if Tkinter.TkVersion>=8.6: self.master.option_add('*font','System 13') # ok with magnification.  Note >13 causes square buttons.  (Including this line causes "Big print" to work)
+                if "AppTranslocation" in os.getcwd(): self.todo.alert="Your Mac is using \"app translocation\" to stop Gradint from writing to its folder. This will cause many problems. Quit Gradint, drag it to a different folder and run it again."
             elif WMstandard: self.master.option_add('*font','Helvetica 7') # TODO on ALL WMstandard devices?
             if winsound or cygwin or macsound: self.master.resizable(1,0) # resizable in X direction but not Y (latter doesn't make sense, see below).  (Don't do this on X11 because on some distros it results in loss of automatic expansion as we pack more widgets.)
+            elif unix:
+                if getoutput("xlsatoms|grep COMPIZ_WINDOW").find("COMPIZ")>=0: # (not _COMPIZ_WM_WINDOW_BLUR, that's sometimes present outside Compiz)
+                  # Compiz sometimes has trouble auto-resizing our window (e.g. on Ubuntu 11.10)
+                  self.master.geometry("%dx%d" % (self.winfo_screenwidth(),self.winfo_screenheight()))
+                  if not GUI_always_big_print: self.todo.alert = "Gradint had to maximize itself because your window manager is Compiz which sometimes has trouble handling Tkinter window sizes"
             self.extra_button_callables = []
             self.pack(fill=Tkinter.BOTH,expand=1)
             self.leftPanel = Tkinter.Frame(self)
-            self.leftPanel.pack(side="left",fill=Tkinter.X,expand=1) # "fill" needed so listbox can fill later
+            self.leftPanel.pack(side="left",fill=Tkinter.BOTH,expand=1)
             self.rightPanel = None # for now
             self.cancelling = 0 # guard against multiple presses of Cancel
             self.Label = Tkinter.Label(self.leftPanel,text="Please wait a moment")
@@ -483,33 +515,37 @@ def startTk():
                 nominalSize = intor0(f[-1])
                 if nominalSize: f=" ".join(f[:-1])+" %d"
                 else: # Tk 8.5+ ?
-                    f=str(self.tk.eval('set font [font actual '+' '.join(f)+']')).split()
-                    upNext = 0
-                    for i in range(len(f)):
+                    for f2 in ['set font [font actual '+' '.join(f)+']', # Tk 8.5
+                               'set font [font actual default]']: # Tk 8.6
+                      f=str(self.tk.eval(f2)).split()
+                      upNext = 0
+                      for i in range(len(f)):
                         if f[i]=="-size": upNext=1
                         elif upNext:
                             nominalSize=intor0(f[i])
                             if nominalSize<0: nominalSize,f[i] = -nominalSize,"-%d"
                             else: f[i]="%d"
                             break
+                      if nominalSize==long(32768)*long(65536): nominalSize = 0 # e.g. Tk 8.6 on Ubuntu 16.04 when using the first eval string above
+                      elif f2=='set font [font actual default]': nominalSize *= 0.77 # kludge for Tk 8.6 on Ubuntu 16.04 to make large-print calculation below work
+                      if nominalSize: break
                     f=" ".join(f)
-                    if not "%d" in f: raise Exception("wrong format") # caught below
+                    if (not checkIn("%d",f)) or not nominalSize: raise Exception("wrong format") # caught below
                 pixelSize = self.Label.winfo_reqheight()-2*int(str(self.Label["borderwidth"]))-2*int(str(self.Label["pady"]))
                 # NB DO NOT try to tell Tk a desired pixel size - you may get a *larger* pixel size.  Need to work out the desired nominal size.
                 approx_lines_per_screen_when_large = 25 # TODO really? (24 at 800x600 192dpi 15in but misses the status line, but OK for advanced users.  setting 25 gives nominal 7 which is rather smaller.)
                 largeNominalSize = int(nominalSize*self.Label.winfo_screenheight()/approx_lines_per_screen_when_large/pixelSize)
                 if largeNominalSize >= nominalSize+3:
                     self.bigPrintFont = f % largeNominalSize
+                    self.bigPrintMult = largeNominalSize*1.0/nominalSize
                     if GUI_always_big_print:
-                        self.master.option_add('*font',self.bigPrintFont)
-                        self.Label["font"]=self.bigPrintFont
-                        del self.bigPrintFont ; self.isBigPrint=1
+                        self.bigPrint0()
                 else: self.after(100,self.check_window_position) # (needs to happen when window is already drawn if you want it to preserve the X co-ordinate)
             except: pass # wrong font format or something - can't do it
             if winCEsound and ask_teacherMode: self.Label["font"]="Helvetica 16" # might make it slightly easier
             self.remake_cancel_button(localise("Cancel lesson"))
             self.Cancel.focus() # (default focus if we don't add anything else, e.g. reader)
-            self.copyright_string = u"This is "+(u""+program_name).replace("(c)",u"\n\u00a9").replace("-",u"\u2013")
+            self.copyright_string = u"This is "+ensure_unicode(program_name).replace("(c)",u"\n\u00a9").replace("-",u"\u2013")
             self.Version = Tkinter.Label(self.leftPanel,text=self.copyright_string)
             addStatus(self.Version,self.copyright_string)
             if olpc: self.Version["font"]='Helvetica 9'
@@ -625,7 +661,9 @@ def startTk():
                     updateUserRow(1)
                 if hasattr(self,"bigPrintFont"):
                     self.BigPrintButton = addButton(self.leftPanel,localise("Big print"),self.bigPrint)
-                    self.BigPrintButton["font"]=self.bigPrintFont
+                    try: self.BigPrintButton["font"]=self.bigPrintFont
+                    except:
+                        self.BigPrintButton.pack_forget() ; del self.BigPrintButton, self.bigPrintFont
                 self.remake_cancel_button(localise("Quit"))
                 if not GUI_omit_statusline: self.Version.pack(fill=Tkinter.X,expand=1)
                 if olpc or self.todo.set_main_menu=="test" or GUI_for_editing_only: self.showtest() # olpc: otherwise will just get a couple of options at the top and a lot of blank space (no way to centre it)
@@ -634,7 +672,7 @@ def startTk():
                 self.restore_copyright()
             if hasattr(self.todo,"alert"):
                 # we have to do it on THIS thread (especially on Windows / Cygwin; Mac OS and Linux might get away with doing it from another thread)
-                tkMessageBox.showinfo(self.master.title(),self.todo.alert)
+                tkMessageBox.showinfo(self.master.title(),S(self.todo.alert))
                 del self.todo.alert
             if hasattr(self.todo,"question"):
                 self.answer_given = tkMessageBox.askyesno(self.master.title(),self.todo.question)
@@ -699,7 +737,7 @@ def startTk():
             addLabel(self.LessonRow,localise("mins"))
             self.MakeLessonButton=addButton(self.LessonRow,localise("Start lesson"),self.makelesson,{"side":"left"},status="Press to create customized lessons\nusing the words in your collection")
             self.lastOutTo=-1 # so it updates the Start Lesson button if needed
-            self.MakeLessonButton.bind('<FocusIn>',(lambda *args:app.after(10,lambda *args:app.MinsEntry.selection_clear())))
+            self.MakeLessonButton.bind('<FocusIn>',(lambda e=None,app=app:app.after(10,lambda e=None,app=app:app.MinsEntry.selection_clear())))
         def sync_listbox_etc(self):
             if not hasattr(self,"vocabList"):
                 if hasattr(self,"needVocablist"): return # already waiting for main thread to make one
@@ -730,18 +768,18 @@ def startTk():
             if synthCache:
                 cacheManagementOptions = [] # (text, oldKey, newKey, oldFile, newFile)
                 for t,l in [(text1.encode('utf-8'),secondLanguage),(text2.encode('utf-8'),firstLanguage)]:
-                    k,f = synthcache_lookup("!synth:"+t+"_"+l,justQueryCache=1)
+                    k,f = synthcache_lookup(B("!synth:")+t+B("_")+B(l),justQueryCache=1)
                     if f:
-                      if (partials_langname(l) in synth_partials_voices or get_synth_if_possible(l,0)): # (no point having these buttons if there's no chance we can synth it by any method OTHER than the cache)
-                        if k in synthCache_transtbl and k[0]=="_": cacheManagementOptions.append(("Keep in "+l+" cache",k,k[1:],0,0))
-                        elif k[0]=="_": cacheManagementOptions.append(("Keep in "+l+" cache",0,0,f,f[1:]))
-                        if k in synthCache_transtbl: cacheManagementOptions.append(("Reject from "+l+" cache",k,"__rejected_"+k,0,0))
+                      if (checkIn(partials_langname(l),synth_partials_voices) or get_synth_if_possible(l,0)): # (no point having these buttons if there's no chance we can synth it by any method OTHER than the cache)
+                        if checkIn(k,synthCache_transtbl) and B(k[:1])==B("_"): cacheManagementOptions.append(("Keep in "+l+" cache",k,k[1:],0,0))
+                        elif B(k[:1])==B("_"): cacheManagementOptions.append(("Keep in "+l+" cache",0,0,f,f[1:]))
+                        if checkIn(k,synthCache_transtbl): cacheManagementOptions.append(("Reject from "+l+" cache",k,"__rejected_"+k,0,0))
                         else: cacheManagementOptions.append(("Reject from "+l+" cache",0,0,f,"__rejected_"+f))
                     else:
-                      k,f = synthcache_lookup("!synth:__rejected_"+t+"_"+l,justQueryCache=1)
-                      if not f: k,f = synthcache_lookup("!synth:__rejected__"+t+"_"+l,justQueryCache=1)
+                      k,f = synthcache_lookup(B("!synth:__rejected_")+t+B("_"+l),justQueryCache=1)
+                      if not f: k,f = synthcache_lookup(B("!synth:__rejected__")+t+B("_"+l),justQueryCache=1)
                       if f:
-                        if k in synthCache_transtbl: cacheManagementOptions.append(("Undo "+l+" cache reject",k,k[11:],0,0))
+                        if checkIn(k,synthCache_transtbl): cacheManagementOptions.append(("Undo "+l+" cache reject",k,k[11:],0,0))
                         else: cacheManagementOptions.append(("Undo "+l+" cache reject",0,0,f,f[11:]))
                       elif l==secondLanguage and mp3web and not ';' in t: cacheManagementOptions.append(("Get from "+mp3webName,0,0,0,0))
                 if not hasattr(self,"cacheManagementOptions"):
@@ -756,16 +794,16 @@ def startTk():
                 if not (text1 or text2): self.ListBox.selection_clear(0,'end') # probably just added a new word while another was selected (added a variation) - clear selection to reduce confusion
                 else: return # don't try to be clever with searches when editing an existing item (the re-ordering can be confusing)
             text1,text2 = text1.lower().replace(" ",""),text2.lower().replace(" ","") # ignore case and whitespace when searching
-            l=map(lambda (x,y):x+"="+y, filter(lambda (x,y):text1 in x.lower().replace(" ","") and text2 in y.lower().replace(" ",""),self.vocabList)[-tkNumWordsToShow:])
+            l=map(lambda x:x[0]+"="+x[1], filter(lambda x,text1=text1,text2=text2:x[0].lower().replace(" ","").find(text1)>-1 and x[1].lower().replace(" ","").find(text2)>-1,self.vocabList)[-tkNumWordsToShow:])
             l.reverse() ; synchronizeListbox(self.ListBox,l) # show in reverse order, in case the bottom of the list box is off-screen
         def doSynthcacheManagement(self,oldKey,newKey,oldFname,newFname):
             # should be a quick operation - might as well do it in the GUI thread
             if (oldKey,oldFname) == (0,0): # special for mp3web
                 self.menu_response="mp3web" ; return
-            if oldKey in synthCache_transtbl:
+            if checkIn(oldKey,synthCache_transtbl):
                 if newKey: synthCache_transtbl[newKey]=synthCache_transtbl[oldKey]
                 else: del synthCache_transtbl[oldKey]
-                open(synthCache+os.sep+transTbl,'w').write("".join([v+" "+k+"\n" for k,v in synthCache_transtbl.items()]))
+                open(synthCache+os.sep+transTbl,'wb').write(B("").join([v+B(" ")+k+B("\n") for k,v in list(synthCache_transtbl.items())]))
             if oldFname:
                 del synthCache_contents[oldFname]
                 if newFname:
@@ -777,7 +815,12 @@ def startTk():
             for control,current,restoreTo in self.toRestore:
                 if asUnicode(control.get())==current: control.set(restoreTo)
             self.toRestore = []
-        def stripText(self,*args): self.Text1.set(fix_commas(hanzi_and_punc(asUnicode(self.Text1.get()))))
+        def stripText(self,*args):
+            t = self.Text1.get()
+            u = asUnicode(t)
+            v = fix_commas(hanzi_and_punc(u))
+            if t==u: v=asUnicode(v)
+            self.Text1.set(v)
         def thin_down_for_lesson(self):
             if hasattr(self,"OutputRow"): self.OutputRow.pack_forget()
             if hasattr(self,"CopyFromButton"):
@@ -804,12 +847,17 @@ def startTk():
             self.Label.pack() ; self.CancelRow.pack()
             self.Label["text"] = "Working..." # (to be replaced by time indication on real-time, not on output-to-file)
             self.Cancel["text"] = localise("Quit")
-        def bigPrint(self,*args):
-            self.thin_down_for_lesson()
+        def bigPrint0(self):
             self.master.option_add('*font',self.bigPrintFont)
-            self.Version["font"]=self.Label["font"]=self.bigPrintFont
+            self.sbarWidth = int(16*self.bigPrintMult)
+            self.master.option_add('*Scrollbar*width',self.sbarWidth) # (works on some systems; usually ineffective on Mac)
+            self.Label["font"]=self.bigPrintFont
             del self.bigPrintFont # (TODO do we want an option to undo it?  or would that take too much of the big print real-estate.)
             self.isBigPrint=1
+        def bigPrint(self,*args):
+            self.thin_down_for_lesson()
+            self.Version["font"]=self.bigPrintFont
+            self.bigPrint0()
             if self.rightPanel: # oops, need to re-construct it
                 global extra_buttons_waiting_list
                 extra_buttons_waiting_list = []
@@ -885,8 +933,8 @@ def startTk():
             self.L2Entry.bind('<Return>',self.changeLanguages)
             for e in [self.L1Entry,self.L2Entry]: e.bind('<Button-1>',(lambda e:e.widget.after(10,lambda e=e:selectAll(e))))
             self.ChangeLanguageButton = addButton(self.row3,"",self.changeLanguages,status="Use this button to set your\nfirst and second languages") # will set text in updateLanguageLabels
-            self.ChangeLanguageButton.bind('<FocusIn>',(lambda *args:app.after(10,lambda *args:app.L2Entry.selection_clear())))
-            self.AddButton.bind('<FocusIn>',(lambda *args:app.after(10,lambda *args:app.L1Entry.selection_clear()))) # for backwards tabbing
+            self.ChangeLanguageButton.bind('<FocusIn>',(lambda e=None,app=app:app.after(10,lambda e=None,app=app:app.L2Entry.selection_clear())))
+            self.AddButton.bind('<FocusIn>',(lambda e=None,app=app:app.after(10,lambda e=None,app=app:app.L1Entry.selection_clear()))) # for backwards tabbing
             if GUI_omit_settings and (vocabFile==user0[1] or fileExists(vocabFile)): self.row3.pack_forget()
             if textEditorCommand:
                 self.RecordedWordsButton = addButton(self.row4,"",self.showRecordedWords,{"side":"left"},status="This button lets you manage recorded\n(as opposed to computer-voiced) words")
@@ -933,8 +981,8 @@ def startTk():
             m=Tkinter.Menu(None, tearoff=0, takefocus=0)
             for i in range(len(lastUserNames)):
                 if lastUserNames[i] and not i==intor0(self.userNo.get()):
-                    if fileExists(addUserToFname(user0[1],i)): m.add_command(label=u"Copy vocab list from "+lastUserNames[i],command=(lambda e=None,i=i:self.copyVocabFrom(i)))
-                    m.add_command(label=u"Copy recordings to/from "+lastUserNames[i],command=(lambda e=None,i=i:self.setToOpen((addUserToFname(user0[0],i),addUserToFname(user0[0],intor0(self.userNo.get()))))))
+                    if fileExists(addUserToFname(user0[1],i)): m.add_command(label=u"Copy vocab list from "+lastUserNames[i],command=(lambda e=None,i=i,self=self:self.copyVocabFrom(i)))
+                    m.add_command(label=u"Copy recordings to/from "+lastUserNames[i],command=(lambda e=None,i=i,self=self:self.setToOpen((addUserToFname(user0[0],i),addUserToFname(user0[0],intor0(self.userNo.get()))))))
             m.tk_popup(self.CopyFromButton.winfo_rootx(),self.CopyFromButton.winfo_rooty(),entry="0")
         def setToOpen(self,toOpen): self.menu_response,self.toOpen = "samplesCopy",toOpen
         def copyVocabFrom(self,userNo):
@@ -944,11 +992,12 @@ def startTk():
             select_userNumber(intor0(self.userNo.get()),updateGUI=0)
             vCurrent = list2set(vocabLinesWithLangs())
             o=appendVocabFileInRightLanguages()
+            if not o: return # IOError
             langs = (secondLanguage,firstLanguage)
             for newLangs,line in vCopyFrom:
-                if (newLangs,line) in vCurrent: continue # already got it
-                if not newLangs==langs: o.write("SET LANGUAGES "+" ".join(list(newLangs))+"\n")
-                o.write(line+"\n")
+                if checkIn((newLangs,line),vCurrent): continue # already got it
+                if not newLangs==langs: o.write(B("SET LANGUAGES ")+B(" ").join(list(newLangs))+B("\n"))
+                o.write(B(line)+B("\n"))
                 langs = newLangs
             o.close()
             if hasattr(self,"vocabList"): del self.vocabList # re-read
@@ -963,15 +1012,15 @@ def startTk():
             global firstLanguage,secondLanguage
             firstLanguage1=asUnicode(self.L1Text.get()).encode('utf-8')
             secondLanguage1=asUnicode(self.L2Text.get()).encode('utf-8')
-            if (firstLanguage,secondLanguage) == (firstLanguage1,secondLanguage1): # they didn't change anything
+            if (B(firstLanguage),B(secondLanguage)) == (firstLanguage1,secondLanguage1): # they didn't change anything
                 langs = ESpeakSynth().describe_supported_languages()
                 msg = (localise("To change languages, edit the boxes that say '%s' and '%s', then press the '%s' button.") % (firstLanguage,secondLanguage,localise("Change languages")))+"\n\n"+localise("Recorded words may be in ANY languages, and you may choose your own abbreviations for them.  However if you want to use the computer voice for anything then please use standard abbreviations.")
                 if langs:
-                    if tkMessageBox.askyesno(self.master.title(),msg+"  "+localise("Would you like to see a list of the standard abbreviations for languages that can be computer voiced?")): self.todo.alert = localise("Languages that can be computer voiced:")+"\n"+langs
+                    if tkMessageBox.askyesno(self.master.title(),msg+"  "+localise("Would you like to see a list of the standard abbreviations for languages that can be computer voiced?")): self.todo.alert = localise("Languages with computer voices (some better than others):")+"\n"+langs
                 else: self.todo.alert = msg+"  "+localise("(Sorry, a list of these is not available on this system - check eSpeak installation.)")
                 return
-            need_redisplay = "@variants-"+GUI_languages.get(firstLanguage,firstLanguage) in GUI_translations or "@variants-"+GUI_languages.get(firstLanguage1,firstLanguage1) in GUI_translations # if EITHER old or new lang has variants, MUST reconstruct that row.  (TODO also do it anyway to get the "Speaker" etc updated?  but may cause unnecessary flicker if that's no big problem)
-            firstLanguage,secondLanguage = firstLanguage1,secondLanguage1
+            need_redisplay = checkIn("@variants-"+GUI_languages.get(firstLanguage,firstLanguage),GUI_translations) or checkIn("@variants-"+GUI_languages.get(S(firstLanguage1),S(firstLanguage1)),GUI_translations) # if EITHER old or new lang has variants, MUST reconstruct that row.  (TODO also do it anyway to get the "Speaker" etc updated?  but may cause unnecessary flicker if that's no big problem)
+            firstLanguage,secondLanguage = S(firstLanguage1),S(secondLanguage1)
             updateSettingsFile(settingsFile,{"firstLanguage":firstLanguage,"secondLanguage":secondLanguage})
             if need_redisplay:
                 self.thin_down_for_lesson()
@@ -1031,7 +1080,10 @@ def startTk():
             self.menu_response="add"
         def zap_newlines(self): # in case someone pastes in text that contains newlines, better not keep them when adding to vocab
             text1,text2 = asUnicode(self.Text1.get()),asUnicode(self.Text2.get())
-            t1,t2 = text1.replace("\n"," ").replace("\r","").strip(wsp), text2.replace("\n"," ").replace("\r","").strip(wsp)
+            # (also remove the simple visual markup that Wenlin sometimes adds)
+            t1,t2=text1,text2
+            for zap in ["\n","\r","<b>","</b>","<i>","</i>","<u>","</u>"]: t1,t2=t1.replace(zap,""),t2.replace(zap,"")
+            t1,t2 = wspstrip(t1),wspstrip(t2)
             if not t1==text1: self.Text1.set(t1)
             if not t2==text2: self.Text2.set(t2)
         def getEncoder(self,*args):
@@ -1083,7 +1135,7 @@ def startTk():
         appThread(Application)
 
 def hanzi_only(unitext): return u"".join(filter(lambda x:0x3000<ord(x)<0xa700 or ord(x)>=0x10000, list(unitext)))
-def hanzi_and_punc(unitext): return u"".join(filter(lambda x:0x3000<ord(x)<0xa700 or ord(x)>=0x10000 or x in '.,?;":\'()[]!0123456789-', list(remove_tone_numbers(fix_compatibility(unitext)))))
+def hanzi_and_punc(unitext): return u"".join(filter(lambda x:0x3000<ord(x)<0xa700 or ord(x)>=0x10000 or x in '.,?;:\'()[]!0123456789-', list(remove_tone_numbers(fix_compatibility(unitext))))) # no " as it could be from SGML markup
 # (exclusion of 3000 in above is deliberate, otherwise get problems with hanzi spaces being taken out by fix-compat+strip hence a non-functional 'delete non-hanzi' button appears)
 def guiVocabList(parsedVocab):
     # This needs to be fast.  Have tried writing interatively rather than filter and map, and assume stuff is NOT already unicode (so just decode rather than call ensure_unicode) + now assuming no !synth: (but can still run with .txt etc)
@@ -1100,24 +1152,26 @@ def guiVocabList(parsedVocab):
         if b.endswith(fl2): b=b[:fl2Len]
         elif b.endswith(fl3): b=readText(b)
         else: continue
-        ret.append((unicode(c,"utf-8"),unicode(b,"utf-8")))
+        ret.append((ensure_unicode(c),ensure_unicode(b)))
     return ret
 def readText(l): # see utils/transliterate.py (running guiVocabList on txt files from scanSamples)
-    l = samplesDirectory+os.sep+l
-    if l in variantFiles: # oops. just read the 1st .txt variant
-        if os.sep in l: lp=(l+os.sep)[:l.rfind(os.sep)]+os.sep
-        else: lp = ""
-        varList = filter(lambda x:x.endswith(dottxt),variantFiles[l])
+    l = B(samplesDirectory)+B(os.sep)+B(l)
+    if checkIn(l,variantFiles): # oops. just read the 1st .txt variant
+        if B(os.sep) in l: lp=(l+B(os.sep))[:l.rfind(B(os.sep))]+B(os.sep)
+        else: lp = B("")
+        varList = filter(lambda x:x.endswith(B(dottxt)),variantFiles[l])
         varList.sort() # so at least it consistently returns the same one.  TODO utils/ cache-synth.py list-synth.py synth-batchconvert-helper.py all use readText() now, can we get them to cache the other variants too?
         l = lp + varList[0]
-    return u8strip(read(l)).strip(wsp)
+    return bwspstrip(u8strip(read(l)))
 
 def singular(number,s):
   s=localise(s)
   if firstLanguage=="en" and number==1 and s[-1]=="s": return s[:-1]
   return s
 def localise(s):
-  d = GUI_translations.get(s,{}) ; s2 = 0
+  if s=="zh-yue" or s=="zhy": k="cant"
+  else: k=s
+  d = GUI_translations.get(k,{}) ; s2 = 0
   GUIlang = GUI_languages.get(firstLanguage,firstLanguage)
   if scriptVariants.get(GUIlang,0): s2 = d.get(GUIlang+str(scriptVariants[GUIlang]+1),0)
   if not s2: s2 = d.get(GUIlang,s)
@@ -1166,10 +1220,7 @@ if useTK:
     textEditorCommand=explorerCommand=None
     if winsound or mingw32 or cygwin:
         textEditorName="Notepad" ; textEditorWaits=1
-        # Try Notepad++ first, otherwise plain notepad
-        textEditorCommand = programFiles+os.sep+"Notepad++"+os.sep+"notepad++.exe"
-        if fileExists(textEditorCommand): textEditorCommand='"'+textEditorCommand+'" -multiInst -notabbar -nosession'
-        else: textEditorCommand="notepad"
+        textEditorCommand="notepad"
         explorerCommand="explorer"
     elif macsound:
         textEditorName="TextEdit"
@@ -1177,21 +1228,27 @@ if useTK:
         if got_program("bbedit"):
             textEditorName="bbedit"
             textEditorCommand="bbedit -w" ; textEditorWaits=1
+        elif got_program("edit"): # TextWrangler
+            textEditorName="edit"
+            textEditorCommand="edit -w" ; textEditorWaits=1
         if sys.version.startswith("2.3.5") and "DISPLAY" in os.environ: explorerCommand = None # 'open' doesn't seem to work when running from within Python in X11 on 10.4
         else: explorerCommand="open"
     elif unix:
-        if "KDE_FULL_SESSION" is os.environ and got_program("kfmclient"):
+        if "KDE_FULL_SESSION" in os.environ and got_program("kfmclient"):
             # looks like we're in a KDE session and can use the kfmclient command
             textEditorCommand=explorerCommand="kfmclient exec"
         elif not olpc and got_program("gnome-open"):
             textEditorCommand=explorerCommand="gnome-open"
+        elif got_program("nautilus"): explorerCommand="nautilus"
+        elif got_program("pcmanfm"): explorerCommand="pcmanfm" # LXDE, LXQt
+        elif got_program("pcmanfm-qt"): explorerCommand="pcmanfm-qt" # might not work as well as pcmanfm on 24.04
         elif got_program("rox"):
             # rox is available - try using that to open directories
             # (better not use it for editor as it might not be configured)
-            # (TODO if both rox and gnome-open are available, can we tell which one the user prefers?  currently using gnome-open)
+            # (TODO if both rox and gnome are available, can we tell which one the user prefers?)
             explorerCommand="rox"
         # anyway, see if we can find a nice editor
-        for editor in ["leafpad","gedit","nedit","kedit","xedit"]:
+        for editor in ["leafpad","featherpad","gedit","nedit","kedit","xedit"]:
             if got_program(editor):
                 textEditorName=textEditorCommand=editor
                 textEditorWaits = 1
@@ -1201,7 +1258,12 @@ if useTK:
                 break
     # End of finding editor - now start GUI
     try:
-        import thread,Tkinter,tkMessageBox
+        try: import thread
+        except ImportError: import _thread as thread
+        try: import Tkinter,tkMessageBox
+        except:
+            import tkinter as Tkinter
+            from tkinter import messagebox as tkMessageBox
         forceRadio=(macsound and 8.49<Tkinter.TkVersion<8.59) # indicatoron doesn't do very well in OS X 10.6 (Tk 8.5) unless we patched it
         if olpc:
             def interrupt_main(): os.kill(os.getpid(),2) # sigint
@@ -1221,31 +1283,35 @@ if useTK:
 def openDirectory(dir,inGuiThread=0):
     if winCEsound:
         if not dir[0]=="\\": dir=os.getcwd()+cwd_addSep+dir # must be absolute
-        ctypes.cdll.coredll.ShellExecuteEx(ctypes.byref(ShellExecuteInfo(60,File=u"\\Windows\\fexplore",Parameters=u""+dir)))
+        ctypes.cdll.coredll.ShellExecuteEx(ctypes.byref(ShellExecuteInfo(60,File=u"\\Windows\\fexplore",Parameters=ensure_unicode(dir))))
     elif explorerCommand:
+        if ' ' in dir: dir='"'+dir+'"'
         cmd = explorerCommand+" "+dir
         if winsound or mingw32: cmd="start "+cmd # (not needed on XP but is on Vista)
         elif unix: cmd += "&"
         os.system(cmd)
     else:
-        msg = "Don't know how to start the file explorer.  Please open the %s directory (in %s)" % (dir,os.getcwd())
+        msg = ""
+        if not dir.startswith(os.sep): msg=" (in %s)" % os.getcwd()
+        msg = "Don't know how to start the file explorer.  Please open the %s directory%s" % (dir,msg)
         if inGuiThread: tkMessageBox.showinfo(app.master.title(),msg)
         else: waitOnMessage(msg)
 
-def sanityCheck(text,language,pauseOnError=0): # text is utf-8; returns error message if any
+def generalCheck(text,language,pauseOnError=0): # text is utf-8; returns error message if any
     if not text: return # always OK empty strings
     if pauseOnError:
-        ret = sanityCheck(text,language)
+        ret = generalCheck(text,language)
         if ret: waitOnMessage(ret)
         return ret
     if language=="zh":
-        allDigits = True
-        for t in text:
+        allDigits = True ; text=B(text)
+        for i in xrange(len(text)):
+            t = text[i:i+1]
             if ord(t)>127: return # got hanzi or tone marks
-            if t in "12345": return # got tone numbers
-            if t not in "0123456789. ": allDigits = False
+            if t in B("12345"): return # got tone numbers
+            if t not in B("0123456789. "): allDigits = False
         if allDigits: return
-        return "Pinyin needs tones.  Please go back and add tone numbers to "+text+"."+cond(startBrowser("http://www.pristine.com.tw/lexicon.php?query="+fix_pinyin(text,[]).replace("1","1 ").replace("2","2 ").replace("3","3 ").replace("4","4 ").replace("5"," ").replace("  "," ").strip(wsp).replace(" ","+"))," Gradint has pointed your web browser at an online dictionary that might help.","")
+        return B("Pinyin needs tones.  Please go back and add tone numbers to ")+text+B(".")+cond(startBrowser(B("http://www.mdbg.net/chinese/dictionary?wdqb=")+bwspstrip(fix_pinyin(text,[])).replace(B("5"),B("")).replace(B(" "),B("+"))),B(" Gradint has pointed your web browser at an online dictionary that might help."),B(""))
 
 def check_for_slacking():
     if fileExists(progressFile): checkAge(progressFile,localise("It has been %d days since your last Gradint lesson.  Please try to have one every day."))
@@ -1260,12 +1326,12 @@ def checkAge(fname,message):
     if days>=5 and (days%5)==0: waitOnMessage(message % days)
 
 def s60_addVocab():
-  label1,label2 = u""+localise("Word in %s") % localise(secondLanguage),u""+localise("Meaning in %s") % localise(firstLanguage)
+  label1,label2 = ensure_unicode(localise("Word in %s") % localise(secondLanguage)),ensure_unicode(localise("Meaning in %s") % localise(firstLanguage))
   while True:
     result = appuifw.multi_query(label1,label2) # unfortunately multi_query can't take default items (and sometimes no T9!), but Form is too awkward (can't see T9 mode + requires 2-button save via Options) and non-multi query would be even more modal
     if not result: return # cancelled
     l2,l1 = result # guaranteed to both be populated
-    while sanityCheck(l2.encode('utf-8'),secondLanguage,1):
+    while generalCheck(l2.encode('utf-8'),secondLanguage,1):
         l2=appuifw.query(label1,"text",u"")
         if not l2: return # cancelled
     # TODO detect duplicates like Tk GUI does?
@@ -1273,7 +1339,7 @@ def s60_addVocab():
     appendVocabFileInRightLanguages().write((l2+"="+l1+"\n").encode("utf-8"))
 def s60_changeLang():
     global firstLanguage,secondLanguage
-    result = appuifw.multi_query(u""+localise("Your first language")+" (e.g. "+firstLanguage+")",u""+localise("second")+" (e.g. "+secondLanguage+")")
+    result = appuifw.multi_query(ensure_unicode(localise("Your first language")+" (e.g. "+firstLanguage+")"),ensure_unicode(localise("second")+" (e.g. "+secondLanguage+")"))
     if not result: return # cancelled
     l1,l2 = result
     firstLanguage,secondLanguage = l1.encode('utf-8').lower(),l2.encode('utf-8').lower()
@@ -1287,7 +1353,7 @@ def s60_runLesson():
 def s60_viewVocab():
     global justSynthesize
     doLabel("Reading your vocab list, please wait...")
-    vList = map(lambda (l2,l1):l2+u"="+l1, guiVocabList(parseSynthVocab(vocabFile,1)))
+    vList = map(lambda x:x[0]+u"="+x[1], guiVocabList(parseSynthVocab(vocabFile,1)))
     if not vList: return waitOnMessage("Your computer-voiced vocab list is empty.")
     while True:
       appuifw.app.body = None
@@ -1297,8 +1363,8 @@ def s60_viewVocab():
       action = appuifw.popup_menu([u"Speak (just "+secondLanguage+")",u"Speak ("+secondLanguage+" and "+firstLanguage+")",u"Change "+secondLanguage,u"Change "+firstLanguage,u"Delete item",u"Cancel"], vList[sel])
       if action==0 or action==1:
         doLabel("Speaking...")
-        justSynthesize = secondLanguage+" "+l2.encode('utf-8')
-        if action==1: justSynthesize += ("#"+firstLanguage+" "+l1.encode('utf-8'))
+        justSynthesize = B(secondLanguage)+B(" ")+l2.encode('utf-8')
+        if action==1: justSynthesize += (B('#')+B(firstLanguage)+B(" ")+l1.encode('utf-8'))
         just_synthesize()
         justSynthesize = ""
       elif action==5: pass
@@ -1307,11 +1373,11 @@ def s60_viewVocab():
           oldL1,oldL2 = l1,l2
           if action==2:
               first=1
-              while first or (l2 and sanityCheck(l2.encode('utf-8'),secondLanguage,1)):
-                  first=0 ; l2=appuifw.query(u""+secondLanguage,"text",l2)
+              while first or (l2 and generalCheck(l2.encode('utf-8'),secondLanguage,1)):
+                  first=0 ; l2=appuifw.query(ensure_unicode(secondLanguage),"text",l2)
               if not l2: continue
           elif action==3:
-              l1 = appuifw.query(u""+firstLanguage,"text",l1)
+              l1 = appuifw.query(ensure_unicode(firstLanguage),"text",l1)
               if not l1: continue
           doLabel("Processing")
           delOrReplace(oldL2,oldL1,l2,l1,cond(action==4,"delete","replace"))
@@ -1319,36 +1385,81 @@ def s60_viewVocab():
               del vList[sel]
               if not vList: return # empty
           else: vList[sel] = l2+"="+l1
+def android_addVocab():
+  while True:
+    l2 = None
+    while not l2 or generalCheck(l2.encode('utf-8'),secondLanguage,1):
+      l2 = android.dialogGetInput("Add word","Word in %s" % localise(secondLanguage)).result
+      if not l2: return # cancelled
+    l1 = android.dialogGetInput("Add word","Meaning in %s" % localise(firstLanguage)).result
+    if not l1: return # cancelled
+    # TODO detect duplicates like Tk GUI does?
+    android.makeToast(u"Added "+l2+"="+l1)
+    appendVocabFileInRightLanguages().write((l2+"="+l1+"\n").encode("utf-8"))
+def android_changeLang():
+    global firstLanguage,secondLanguage
+    l1 = android.dialogGetInput("Gradint","Enter your first language",firstLanguage).result
+    if not l1: return # cancelled
+    l2 = android.dialogGetInput("Gradint","Enter your second language",secondLanguage).result
+    if not l2: return # cancelled
+    firstLanguage,secondLanguage = l1.encode('utf-8').lower(),l2.encode('utf-8').lower()
+    updateSettingsFile(settingsFile,{"firstLanguage":firstLanguage,"secondLanguage":secondLanguage})
 
 def delOrReplace(L2toDel,L1toDel,newL2,newL1,action="delete"):
     langs = [secondLanguage,firstLanguage]
-    v=u8strip(read(vocabFile)).replace("\r\n","\n").replace("\r","\n")
-    o=open(vocabFile,"w") ; found = 0
-    if last_u8strip_found_BOM: o.write('\xef\xbb\xbf') # re-write it
-    v=v.split("\n")
+    v=u8strip(read(vocabFile)).replace(B("\r\n"),B("\n")).replace(B("\r"),B("\n"))
+    if paranoid_file_management:
+        fname = os.tempnam()
+        o = open(fname,"w")
+    else: o=open(vocabFile,"w")
+    found = 0
+    if last_u8strip_found_BOM: writeB(o,LB('\xef\xbb\xbf')) # re-write it
+    v=v.split(B("\n"))
     if v and not v[-1]: v=v[:-1] # don't add an extra blank line at end
     for l in v:
         l2=l.lower()
-        if l2.startswith("set language ") or l2.startswith("set languages "):
-            langs=l.split()[2:] ; o.write(l+"\n") ; continue
-        thisLine=map(lambda x:x.strip(wsp),l.split("=",len(langs)-1))
+        if l2.startswith(B("set language ")) or l2.startswith(B("set languages ")):
+            langs=map(S,l.split()[2:]) ; writeB(o,l+B("\n")) ; continue
+        thisLine=map(bwspstrip,l.split(B("="),len(langs)-1))
         if (langs==[secondLanguage,firstLanguage] and thisLine==[L2toDel.encode('utf-8'),L1toDel.encode('utf-8')]) or (langs==[firstLanguage,secondLanguage] and thisLine==[L1toDel.encode('utf-8'),L2toDel.encode('utf-8')]):
             # delete this line.  and maybe replace it
             found = 1
             if action=="replace":
-                if langs==[secondLanguage,firstLanguage]: o.write(newL2.encode("utf-8")+"="+newL1.encode("utf-8")+"\n")
-                else: o.write(newL1.encode("utf-8")+"="+newL2.encode("utf-8")+"\n")
-        else: o.write(l+"\n")
+                if langs==[secondLanguage,firstLanguage]: writeB(o,newL2.encode("utf-8")+B("=")+newL1.encode("utf-8")+B("\n"))
+                else: writeB(o,newL1.encode("utf-8")+B("=")+newL2.encode("utf-8")+B("\n"))
+        else: writeB(o,l+B("\n"))
     o.close()
+    if paranoid_file_management:
+        write(vocabFile,read(fname))
+        os.remove(fname)
     return found
 
-def maybeCanSynth(lang): return lang in synth_partials_voices or get_synth_if_possible(lang,0) or synthCache
+def maybeCanSynth(lang): return checkIn(lang,synth_partials_voices) or get_synth_if_possible(lang,0) or synthCache
+def android_main_menu():
+  while True:
+    menu=[]
+    if maybeCanSynth(secondLanguage):
+        menu.append((unicode(localise("Just speak a word")),primitive_synthloop))
+        doVocab = maybeCanSynth(firstLanguage)
+        if doVocab: menu.append((unicode(localise("Add word to my vocab")),android_addVocab))
+        menu.append((unicode(localise("Make lesson from vocab")),lesson_loop))
+        # if doVocab: menu.append((u"View/change vocab",android_viewVocab)) # (TODO but lower priority because SL4A has an editor)
+    else: menu.append((unicode(localise("Make lesson")),lesson_loop))
+    menu += [(unicode(localise("Record word(s) with mic")),android_recordWord),(unicode(localise("Change languages")),android_changeLang)]
+    menu.append((unicode(localise("Quit")),None))
+    android.dialogCreateAlert("Gradint","Choose an action")
+    android.dialogSetItems(map (lambda x:x[0], menu))
+    android.dialogShow()
+    try: function = menu[android.dialogGetResponse().result['item']][1]
+    except KeyError: break # probably an error condition: don't try to redisplay, just quit
+    if function: function() # and redisplay after
+    else: break # quit
 def s60_main_menu():
   while True:
     appuifw.app.body = None # NOT text saying version no etc - has distracting blinking cursor
     menu=[]
     if maybeCanSynth(secondLanguage):
-        menu.append((u"Just speak a word",primitive_synthloop))
+        menu.append((u"Just speak a word",primitive_synthloop)) # no localise() as S60 is not guaranteed to be able to display the characters
         doVocab = maybeCanSynth(firstLanguage)
         if doVocab: menu.append((u"Add word to my vocab",s60_addVocab))
         menu.append((u"Make lesson from vocab",s60_runLesson))
@@ -1364,20 +1475,18 @@ def s60_main_menu():
 
 def downloadLAME():
     # Sourceforge keep making this harder!
-    return not system("""if which curl >/dev/null 2>/dev/null; then export Curl="curl -L"; else export Curl="wget -O -"; fi
-if ! test -e lame*.tar.gz; then
-  export Link="$($Curl "http://sourceforge.net/project/showfiles.php?group_id=290&package_id=309"|grep tar.gz|head -1)"
-  echo "Got HTML: $Link" 1>&2
-  export Link="$(echo "$Link"|sed -e 's,href="/,href="http://sourceforge.net/,' -e 's/.*http:/http:/' -e 's/.tar.gz.*/.tar.gz/')"
-  echo "Following link to $Link" 1>&2
-  if ! $Curl "$Link" > lame.tar.gz; then
+    # Removed code to check for latest version, as we
+    # can't use v3.100 due to Lame bug 488.
+    return not system("""if which curl >/dev/null 2>/dev/null; then Curl="curl -L"; else Curl="wget -O -"; fi
+if ! [ -e lame*.tar.gz ]; then
+  if ! $Curl "https://sourceforge.net/projects/lame/files/lame/3.99/lame-3.99.5.tar.gz/download" > lame.tar.gz; then
     rm -f lame.tar.gz; exit 1
   fi
   if grep downloads.sourceforge lame.tar.gz 2>/dev/null; then
-    export Link="$(cat lame.tar.gz|grep downloads.sourceforge|head -1)"
-    echo "Got HTML 2: $Link" 1>&2
-    export Link="$(echo "$Link"|sed -e 's/.*http/http/' -e 's,.*/projects,http://sourceforge.net/projects,' -e 's/".*//')"
-    echo "Following link 2 to $Link" 1>&2
+    Link="$(cat lame.tar.gz|grep downloads.sourceforge|head -1)"
+    echo "Got HTML: $Link" >&2
+    Link="$(echo "$Link"|sed -e 's/.*http/http/' -e 's,.*/projects,http://sourceforge.net/projects,' -e 's/".*//')"
+    echo "Following link to $Link" >&2
     if ! $Curl "$Link" > lame.tar.gz; then
       rm -f lame.tar.gz; exit 1
     fi
@@ -1391,9 +1500,8 @@ def gui_event_loop():
       disable_once_per_day = cond(getYN(localise("Do you want Gradint to start by itself and remind you to practise?")),0,1)
       updateSettingsFile("advanced"+dottxt,{"disable_once_per_day":disable_once_per_day})
       if disable_once_per_day: # signal the background process to stop next time
-        for f in ["background1"+dottxt,"background2"+dottxt]:
-          try: os.remove(f)
-          except: pass
+        try: os.remove("background"+dottxt)
+        except: pass
     if orig_onceperday&2: check_for_slacking()
     while app:
         while not hasattr(app,"menu_response"):
@@ -1406,13 +1514,16 @@ def gui_event_loop():
             if emulated_interruptMain: check_for_interrupts()
             time.sleep(0.3)
         menu_response = app.menu_response
+        del app.menu_response
         if menu_response=="input": # WMstandard
             app.todo.input_response=raw_input()
         elif menu_response=="go":
             gui_outputTo_start()
             if not soundCollector: app.todo.add_briefinterrupt_button = 1
             try: lesson_loop()
-            except PromptException,prEx: waitOnMessage("Problem finding prompts:\n"+prEx.message) # and don't quit, user may be able to fix
+            except PromptException:
+                prEx = sys.exc_info()[1]
+                waitOnMessage("Problem finding prompts:\n"+prEx.message) # and don't quit, user may be able to fix
             except KeyboardInterrupt: pass # probably pressed Cancel Lesson while it was still being made (i.e. before handleInterrupt)
             if app and not soundCollector: app.todo.remove_briefinterrupt_button = 1 # (not app if it's closed by the close box)
             gui_outputTo_end()
@@ -1427,7 +1538,7 @@ def gui_event_loop():
                     waitOnMessage("WARNING: Word may not save non-Western characters properly.  Try an editor like MADE instead (need to set its font).") # TODO Flinkware MADE version 2.0.0 has been known to insert spurious carriage returns at occasional points in large text files
                     if not app.fileToEdit[0]=="\\": app.fileToEdit=os.getcwd()+cwd_addSep+app.fileToEdit # must be absolute
                     if not fileExists(app.fileToEdit): open(app.fileToEdit,"w") # at least make sure it exists
-                    ctypes.cdll.coredll.ShellExecuteEx(ctypes.byref(ShellExecuteInfo(60,File=u""+app.fileToEdit)))
+                    ctypes.cdll.coredll.ShellExecuteEx(ctypes.byref(ShellExecuteInfo(60,File=ensure_unicode(app.fileToEdit))))
                     waitOnMessage("When you've finished editing "+app.fileToEdit+", close it and start gradint again.")
                     return
             elif textEditorCommand:
@@ -1448,7 +1559,7 @@ def gui_event_loop():
                         if time.time() < t+3: waitOnMessage(textEditorName+" returned control to Gradint in less than 3 seconds.  Perhaps you already had an instance running and it loaded the file remotely.  Press OK when you have finished editing the file.")
                         newContents = read(fileToEdit)
                         if not newContents==oldContents:
-                            if paranoid_file_management: open(app.fileToEdit,"w").write(newContents)
+                            if paranoid_file_management: write(app.fileToEdit,newContents)
                             if app.fileToEdit==vocabFile:
                                 app.wordsExist=1 ; del app.vocabList # re-read
                             else: waitOnMessage("The changes you made to "+app.fileToEdit+" will take effect when you quit Gradint and start it again.")
@@ -1475,18 +1586,21 @@ def gui_event_loop():
             text1 = asUnicode(app.Text1.get()).encode('utf-8') ; text2 = asUnicode(app.Text2.get()).encode('utf-8')
             if not text1 and not text2: app.todo.alert=u"Before pressing the "+localise("Speak")+u" button, you need to type the text you want to hear into the box."
             else:
-              msg=sanityCheck(text1,secondLanguage)
-              if msg: app.todo.alert=u""+msg
+              if text1.startswith(B('#')): msg="" # see below
+              else: msg=generalCheck(text1,secondLanguage)
+              if msg: app.todo.alert=ensure_unicode(msg)
               else:
                 app.set_watch_cursor = 1 ; app.toRestore = []
                 global justSynthesize ; justSynthesize = ""
                 def doControl(text,lang,control):
-                    global justSynthesize
+                    global justSynthesize ; text=B(text)
                     restoreTo = asUnicode(control.get())
-                    if text:
-                        if can_be_synthesized("!synth:"+text+"_"+lang): justSynthesize += ("#"+lang+" "+text)
-                        else: app.todo.alert="Cannot find a synthesizer that can say '"+text+"' in language '"+lang+"' on this system"
-                        t=transliterates_differently(text,lang)
+                    if text.startswith(B('#')): justSynthesize = B(justSynthesize)+text # hack for direct control of just_synthesize from the GUI (TODO document it in advanced.txt? NB we also bypass the GUI transliteration in the block below)
+                    elif text:
+                        if can_be_synthesized(B("!synth:")+text+B("_")+B(lang)):
+                            justSynthesize=B(justSynthesize)+(B("#")+B(lang)+B(" ")+B(text))
+                        else: app.todo.alert=B("Cannot find a synthesizer that can say '")+text+B("' in language '")+B(lang)+B("' on this system")
+                        t=S(transliterates_differently(text,lang))
                         if t: # (don't go straight into len() stuff, it could be None)
                           if unix and len(t)>300 and hasattr(app,"isBigPrint"): app.todo.alert="Transliteration suppressed to work around Ubuntu bug 731424" # https://bugs.launchpad.net/ubuntu/+bug/731424
                           else:
@@ -1502,7 +1616,6 @@ def gui_event_loop():
                   elif ask_teacherMode: # Do the L2, then ask if actually WANT the L1 as well (might be useful on WinCE etc, search-and-demonstrate-L2)
                     doSynth()
                     if app and not getYN("Also speak the %s?" % firstLanguage):
-                      if app: del app.menu_response
                       continue
                 doControl(text2,firstLanguage,app.Text2)
                 doSynth()
@@ -1527,7 +1640,7 @@ def gui_event_loop():
             if not app: break
             found=0
             for f in scanDirs()[0].keys():
-              if not f in oldLs and (f.lower().endswith(dotmp3) or f.lower().endswith(dotwav)) and getYN("Use "+f[f.rfind(os.sep)+1:]+"?"): # TODO don't ask this question too many times if there are many and they're all 'no'
+              if not checkIn(f,oldLs) and (f.lower().endswith(dotmp3) or f.lower().endswith(dotwav)) and getYN("Use "+f[f.rfind(os.sep)+1:]+"?"): # TODO don't ask this question too many times if there are many and they're all 'no'
                 system("mp3gain -r -s r -k -d 10 \""+f+"\"") # (if mp3gain command is available; ignore errors if not (TODO document in advanced.txt)) (note: doing here not after the move, in case synthCache is over ftpfs mount or something)
                 uf=scFile=text1.encode("utf-8")+"_"+secondLanguage+f[-4:].lower()
                 try:
@@ -1560,30 +1673,29 @@ def gui_event_loop():
                 if not getYN("Download failed.  Try again?"): break
               if worked:
                 app.setLabel("Compiling...")
-                if system("""tar -zxvf lame*.tar.gz && cd lame-* && if ./configure && make; then ln -s $(pwd)/frontend/lame ../lame || true; else cd .. ; rm -rf lame*; exit 1; fi"""): app.todo.alert = "Compile failed"
+                if system("""tar -zxvf lame*.tar.gz && cd lame-* && if ./configure && make; then ln -s $(pwd)/frontend/lame ../lame || true; else cd .. ; rm -rf lame*; exit 1; fi"""):
+                    app.todo.alert = "Compile failed"
+                    if macsound:
+                        app.todo.alert += ". Check the system has Xcode with command-line license accepted (try running gcc from the Terminal)"
+                        # might be asked to run: sudo xcodebuild -license
           app.todo.set_main_menu = 1
         elif (menu_response=="add" or menu_response=="replace") and not (app.Text1.get() and app.Text2.get()): app.todo.alert="You need to type text in both boxes before adding the word/meaning pair to "+vocabFile
-        elif menu_response=="add" and hasattr(app,"vocabList") and (asUnicode(app.Text1.get()),asUnicode(app.Text2.get())) in app.vocabList:
+        elif menu_response=="add" and hasattr(app,"vocabList") and checkIn((asUnicode(app.Text1.get()),asUnicode(app.Text2.get())),app.vocabList):
             # Trying to add a word that's already there - do we interpret this as a progress adjustment?
             app.set_watch_cursor = 1
             t1,t2 = asUnicode(app.Text1.get()),asUnicode(app.Text2.get())
             lang2,lang1=t1.lower(),t2.lower() # because it's .lower()'d in progress.txt
             d = ProgressDatabase(0)
-            l1find = "!synth:"+lang1.encode('utf-8')+"_"+firstLanguage
+            l1find = S(B("!synth:")+lang1.encode('utf-8')+B("_"+firstLanguage))
             found = 0
-            msg=(u""+localise("%s=%s is already in %s.")) % (t1,t2,vocabFile)
+            msg=(ensure_unicode(localise("%s=%s is already in %s.")) % (t1,t2,vocabFile))
             for listToCheck in [d.data,d.unavail]:
               if found: break
               for item in listToCheck:
-                if (item[1]==l1find or (type(item[1])==type([]) and l1find in item[1])) and item[2]=="!synth:"+lang2.encode('utf-8')+"_"+secondLanguage:
+                if (item[1]==l1find or (type(item[1])==type([]) and checkIn(l1find,item[1]))) and item[2]==S(B("!synth:")+lang2.encode('utf-8')+B("_"+secondLanguage)):
                     if not item[0]: break # not done yet - as not-found
+                    newItem0 = reviseCount(item[0])
                     app.unset_watch_cursor = 1
-                    # suggested reduction:
-                    thresholds=[1,2,knownThreshold,reallyKnownThreshold,meaningTestThreshold,randomDropThreshold,randomDropThreshold2] ; thresholds.sort() ; thresholds.reverse()
-                    newItem0 = 0
-                    for i in range(len(thresholds)-1):
-                        if item[0]>thresholds[i]:
-                            newItem0=thresholds[i+1] ; break
                     if getYN(msg+" "+localise("Repeat count is %d. Reduce this to %d for extra revision?" % (item[0],newItem0))):
                         app.set_watch_cursor = 1
                         listToCheck.remove(item)
@@ -1596,12 +1708,15 @@ def gui_event_loop():
                 app.todo.alert=msg+" "+localise("Repeat count is 0, so we cannot reduce it for extra revision.")
         elif menu_response=="add":
             text1 = asUnicode(app.Text1.get()).encode('utf-8') ; text2 = asUnicode(app.Text2.get()).encode('utf-8')
-            msg=sanityCheck(text1,secondLanguage)
-            if msg: app.todo.alert=u""+msg
+            msg=generalCheck(text1,secondLanguage)
+            if msg: app.todo.alert=ensure_unicode(msg)
             else:
                 o=appendVocabFileInRightLanguages()
-                o.write(text1+"="+text2+"\n") # was " = " but it slows down parseSynthVocab
+                if not o: continue # IOError
+                writeB(o,text1+B("=")+text2+B("\n")) # was " = " but it slows down parseSynthVocab
                 o.close()
+                if paranoid_file_management:
+                    if filelen(vocabFile)<filelen(vocabFile+"~") or chr(0) in readB(open(vocabFile,"rb"),1024): app.todo.alert="Vocab file corruption! You'd better restore the ~ backup."
                 if hasattr(app,"vocabList"): app.vocabList.append((ensure_unicode(text1),ensure_unicode(text2)))
                 app.todo.clear_text_boxes=app.wordsExist=1
         elif menu_response=="delete" or menu_response=="replace":
@@ -1614,19 +1729,19 @@ def gui_event_loop():
             if found and menu_response=="replace": # maybe hack progress.txt as well (taken out of the above loop for better failsafe)
                 d = ProgressDatabase(0)
                 lang2,lang1=lang2.lower(),lang1.lower() # because it's .lower()'d in progress.txt
-                l1find = "!synth:"+lang1.encode('utf-8')+"_"+firstLanguage
+                l1find = S(B("!synth:")+lang1.encode('utf-8')+B("_"+firstLanguage))
                 for item in d.data:
-                    if (item[1]==l1find or (type(item[1])==type([]) and l1find in item[1])) and item[2]=="!synth:"+lang2.encode('utf-8')+"_"+secondLanguage and item[0]:
+                    if (item[1]==l1find or (type(item[1])==type([]) and checkIn(l1find,item[1]))) and item[2]==S(B("!synth:")+lang2.encode('utf-8')+B("_"+secondLanguage)) and item[0]:
                         app.unset_watch_cursor = 1
-                        if not getYN(localise("You have repeated %s=%s %d times.  Do you want to pretend you already repeated %s=%s %d times?") % (lang2,lang1,item[0],t2,t1,item[0])):
+                        if not getYN(localise("You have repeated %s=%s %d times.  Do you want to pretend you already repeated %s=%s %d times?") % (S(lang2),S(lang1),item[0],S(t2),S(t1),item[0])):
                             app.set_watch_cursor = 1 ; break
                         d.data.remove(item)
-                        l1replace = "!synth:"+t2.encode('utf-8')+"_"+firstLanguage
+                        l1replace = S(B("!synth:")+t2.encode('utf-8')+B("_"+firstLanguage))
                         if type(item[1])==type([]):
                             l = item[1]
                             l[l.index(l1find)] = l1replace
                         else: l=l1replace
-                        item = (item[0],l,"!synth:"+t1.encode('utf-8')+"_"+secondLanguage)
+                        item = (item[0],l,S(B("!synth:")+t1.encode('utf-8')+B("_"+secondLanguage)))
                         d.data.append(item)
                         app.set_watch_cursor = 1
                         for i2 in d.unavail:
@@ -1639,35 +1754,39 @@ def gui_event_loop():
             app.todo.clear_text_boxes=1
             app.unset_watch_cursor = 1
             if not found: app.todo.alert = "OOPS: Item to delete/replace was not found in "+vocabFile
-        if app: del app.menu_response
 
 def vocabLinesWithLangs(): # used for merging different users' vocab files
     langs = [secondLanguage,firstLanguage] ; ret = []
-    try: v=u8strip(read(vocabFile)).replace("\r","\n")
-    except IOError: v=""
-    for l in v.split("\n"):
+    try: v=u8strip(read(vocabFile)).replace(B("\r"),B("\n"))
+    except IOError: v=B("")
+    for l in v.split(B("\n")):
         l2=l.lower()
-        if l2.startswith("set language ") or l2.startswith("set languages "): langs=l.split()[2:]
+        if l2.startswith(B("set language ")) or l2.startswith(B("set languages ")): langs=map(S,l.split()[2:])
         elif l: ret.append((tuple(langs),l)) # TODO what about blank lines? (currently they'd be considered duplicates)
     return ret
 
 def appendVocabFileInRightLanguages():
     # check if we need a SET LANGUAGE
     langs = [secondLanguage,firstLanguage]
-    try: v=u8strip(read(vocabFile)).replace("\r","\n")
-    except IOError: v=""
-    for l in v.split("\n"):
+    try: v=u8strip(read(vocabFile)).replace(B("\r"),B("\n"))
+    except IOError: v=B("")
+    for l in v.split(B("\n")):
         l2=l.lower()
-        if l2.startswith("set language ") or l2.startswith("set languages "): langs=l.split()[2:]
-    o=open(vocabFile,"a")
-    if not v.endswith("\n"): o.write("\n")
-    if not langs==[secondLanguage,firstLanguage]: o.write("SET LANGUAGES "+secondLanguage+" "+firstLanguage+"\n")
+        if l2.startswith(B("set language ")) or l2.startswith(B("set languages ")):
+            langs=l.split()[2:]
+            for i in range(len(langs)): langs[i]=S(langs[i])
+    try: o=open(vocabFile,"ab") # (ensure binary on Python 3)
+    except IOError:
+        show_warning("Cannot write to "+vocabFile+" (current directory is "+os.getcwd()+")")
+        return
+    if not v.endswith(B("\n")): o.write(B("\n"))
+    if not langs==[secondLanguage,firstLanguage]: o.write(B("SET LANGUAGES "+secondLanguage+" "+firstLanguage+"\n"))
     return o
 
 def transliterates_differently(text,lang):
     global last_partials_transliteration ; last_partials_transliteration=None
     global partials_are_sporadic ; o=partials_are_sporadic ; partials_are_sporadic = None # don't want to touch the counters here
-    if synthcache_lookup("!synth:"+text+"_"+lang):
+    if synthcache_lookup(B("!synth:")+B(text)+B("_")+B(lang)):
         partials_are_sporadic = o
         if last_partials_transliteration and not last_partials_transliteration==text: return last_partials_transliteration
         else: return # (don't try to translit. if was in synth cache - will have no idea which synth did it)
@@ -1691,13 +1810,13 @@ def gui_outputTo_start():
         except: pass
         gui_output_counter = 1 # now local because we also got prefix
         if justSynthesize:
-            if '#' in justSynthesize[1:]: prefix="" # multiple languages
+            if B('#') in B(justSynthesize)[1:]: prefix=B("") # multiple languages
             else: # prefix the language that's being synth'd
-                prefix=justSynthesize.split()[0]
-                if prefix.startswith('#'): prefix=prefix[1:]
-        else: prefix = "lesson"
+                prefix=B(justSynthesize).split()[0]
+                if prefix.startswith(B('#')): prefix=prefix[1:]
+        else: prefix = B("lesson")
         while not outputFile or fileExists(outputFile):
-            outputFile=gui_output_directory+os.sep+prefix+str(gui_output_counter)+extsep+app.outputTo.get()
+            outputFile=gui_output_directory+os.sep+S(prefix)+str(gui_output_counter)+extsep+app.outputTo.get()
             gui_output_counter += 1
         global write_to_stdout ; write_to_stdout = 0
         global out_type ; out_type = app.outputTo.get()
@@ -1730,9 +1849,10 @@ def gui_outputTo_end(openDir=True):
                 # NB we're passing this to cmd, NOT bash:
                 cmd = "cscript \""+pFiles+"\\Windows Media Components\\Encoder\\WMCmd.vbs\" -input \""+o+"\" -output \""+f+"\" -profile a20_1 -a_content 1"
             elif t=="aac": cmd="afconvert \""+o+"\" -d aac \""+f+"\"" # could also use "afconvert file.wav -d samr file.amr", but amr is bigger than aac and not as good; don't know if anyone has a device that plays amr but not aac.
+            # afconvert default is 64kbit AAC. if want 96+ for music, use -b 96000 after the -d aac (and if want iTunes to be able to accept it, specify extension mp4 instead of aac to afconvert; do not rename aac to mp4, but tell afconvert it's mp4)
             else: assert 0
             if cygwin:
-                assert not "'" in cmd, "apostrophees in pathnames could cause trouble on cygwin"
+                assert not "'" in cmd, "apostrophes in pathnames could cause trouble on cygwin"
                 cmd="echo '"+cmd+" && exit' | cmd" # seems the only way to get it to work on cygwin
             system(cmd)
             os.remove(outputFile)
@@ -1748,11 +1868,15 @@ def gui_outputTo_end(openDir=True):
         except: pass
 
 def main():
-    global useTK,justSynthesize,waitBeforeStart,traceback,appTitle
+    global useTK,justSynthesize,waitBeforeStart,traceback,appTitle,app,warnings_toprint
     if useTK:
-        if justSynthesize and not justSynthesize[-1]=='*': appTitle=cond('#' in justSynthesize,"Gradint","Reader") # not "language lesson"
+        if justSynthesize and not B(justSynthesize)[-1:]==B('*'): appTitle=cond(B('#') in B(justSynthesize),"Gradint","Reader") # not "language lesson"
         startTk()
-    else: rest_of_main()
+    else:
+        app = None # not False anymore
+        if not appuifw and not android: # REALLY output them to stderr
+            for w in warnings_toprint: show_warning(w)
+        warnings_toprint = [] ; rest_of_main()
 def rest_of_main():
     global useTK,justSynthesize,waitBeforeStart,traceback,appTitle,saveProgress,RM_running
     exitStatus = 0 ; RM_running = 1
@@ -1762,20 +1886,27 @@ def rest_of_main():
         except NameError: ceLowMemory=0
         if ceLowMemory and getYN("Low memory! Python may crash. Turn off progress saving for safety?"): saveProgress=0
         
-        if justSynthesize=="-": primitive_synthloop()
-        elif justSynthesize and justSynthesize[-1]=='*':
+        if B(justSynthesize)==B("-"): primitive_synthloop()
+        elif justSynthesize and B(justSynthesize)[-1:]==B('*'):
             justSynthesize=justSynthesize[:-1]
             waitBeforeStart = 0
             just_synthesize() ; lesson_loop()
         elif justSynthesize: just_synthesize()
         elif app and waitBeforeStart: gui_event_loop()
         elif appuifw: s60_main_menu()
+        elif android: android_main_menu()
         else: lesson_loop()
-    except SystemExit: pass
+    except SystemExit:
+        e = sys.exc_info()[1]
+        exitStatus = e.code
     except KeyboardInterrupt: pass
-    except PromptException,prEx:
+    except PromptException:
+        prEx = sys.exc_info()[1]
         waitOnMessage("\nProblem finding prompts:\n"+prEx.message+"\n")
         exitStatus = 1
+    except MessageException:
+        mEx = sys.exc_info()[1]
+        waitOnMessage(mEx.message+"\n") ; exitStatus = 1
     except:
         w="\nSomething has gone wrong with my program.\nThis is not your fault.\nPlease let me know what it says.\nThanks.  Silas\n"+exc_info()
         try: import traceback
@@ -1799,7 +1930,7 @@ def rest_of_main():
         except: pass
         waitOnMessage(w.strip())
         if not useTK:
-            if tracebackFile: sys.stderr.write(read("last-gradint-error"+extsep+"txt"))
+            if tracebackFile: writeB(sys.stderr,read("last-gradint-error"+extsep+"txt"))
             elif traceback: traceback.print_exc() # will be wrong if there was an error in speaking
         exitStatus = 1
         if appuifw: raw_input() # so traceback stays visible
@@ -1812,7 +1943,10 @@ def rest_of_main():
     elif not app==None: pass # (gets here if WAS 'app' but was closed - DON'T output anything to stderr in this case)
     elif appuifw: appuifw.app.set_exit()
     elif riscos_sound: show_info("You may now close this Task Window.\n")
-    elif not android: show_info("\n") # in case got any \r'd string there - don't want to confuse the next prompt
+    elif not android:
+        try:
+            doLabelLastLen ; show_info("\n") # if got any \r'd string there - don't want to confuse the next prompt
+        except NameError: pass # no doLabelLastLen - no \r
     RM_running = 0
     if exitStatus: sys.exit(exitStatus)
 

gradint-build/src/lessonplan.py → src/lessonplan.py

-# This file is part of the source code of
-# gradint v0.9979 (c) 2002-2011 Silas S. Brown. GPL v3+.
+# This file is part of the source code of Gradint
+# (c) Silas S. Brown.
 #    This program is free software; you can redistribute it and/or modify
 #    it under the terms of the GNU General Public License as published by
 #    the Free Software Foundation; either version 3 of the License, or
@@ -25,29 +25,34 @@ class ProgressDatabase(object):
           self.unavail = mergeProgress(self.data,scanSamples()+parseSynthVocab(vocabFile))
           if not cache_maintenance_mode:
             doLabel("Checking transliterations")
+            global tList # for Python 2.1
             tList = {}
-            def addVs(ff,dirBase):
-                if dirBase: dirBase += os.sep
-                if dirBase+ff in variantFiles:
-                   if os.sep in ff: ffpath=ff[:ff.rfind(os.sep)+1]
-                   else: ffpath=""
-                   variantList=map(lambda x:ffpath+x,variantFiles[dirBase+ff])
+            def addVs(ff,dirBase): # add variants of ff to tList which we might need to transliterate
+                dirBase,ff = B(dirBase),B(ff)
+                if dirBase: dirBase += B(os.sep)
+                if checkIn(dirBase+ff,variantFiles):
+                   if B(os.sep) in ff: ffpath=ff[:ff.rfind(B(os.sep))+1]
+                   else: ffpath=B("")
+                   variantList=map(lambda x,f=ffpath:f+B(x),variantFiles[dirBase+ff])
                 else: variantList = [ff]
                 l=languageof(ff)
                 for f in variantList:
-                  if f.lower().endswith(dottxt): text=u8strip(read(dirBase+f)).strip(wsp)
-                  elif f.find("!synth")==-1: continue # don't need to translit. filenames of wav's etc
+                  f = B(f)
+                  if f.lower().endswith(B(dottxt)):
+                      text=bwspstrip(u8strip(read(dirBase+f)))
+                  elif f.find(B("!synth"))==-1: continue # don't need to translit. filenames of wav's etc
                   else: text = textof(f)
-                  if not l in tList: tList[l]={}
+                  if not checkIn(l,tList): tList[l]={}
                   tList[l][text]=1
             for ff in availablePrompts.lsDic.values(): addVs(ff,promptsDirectory)
             for _,l1,l2 in self.data:
                 if not type(l1)==type([]): l1=[l1]
                 for ff in l1+[l2]: addVs(ff,samplesDirectory)
             doLabel("Transliterating")
-            for lang,dic in tList.items():
+            for lang,dic in list(tList.items()):
                 s = get_synth_if_possible(lang,0)
-                if s and hasattr(s,"update_translit_cache"): s.update_translit_cache(lang,dic.keys())
+                if s and hasattr(s,"update_translit_cache"): s.update_translit_cache(lang,list(dic.keys()))
+            del tList
         self.didScan = alsoScan
     def _load_from_binary(self):
         if pickledProgressFile and fileExists(pickledProgressFile):
@@ -55,11 +60,12 @@ class ProgressDatabase(object):
                 global firstLanguage, secondLanguage, otherLanguages
                 if compress_progress_file or (unix and got_program("gzip")):
                     if paranoid_file_management: open(pickledProgressFile) # ensure ready
-                    f = os.popen('gzip -fdc "'+pickledProgressFile+'"',"rb")
+                    f = os.popen('gzip -fdc "'+pickledProgressFile+'"',popenRB)
                 else: f=open(pickledProgressFile,"rb")
                 try: thingsToSet, tup = pickle.Unpickler(f).load()
                 except: return False # probably moved to a different Python version or something
                 exec(thingsToSet)
+                self._py3_fix()
                 return True
             # otherwise drop out and return None
     def _load_from_text(self,fromString=0):
@@ -67,11 +73,11 @@ class ProgressDatabase(object):
         elif fileExists(progressFile):
             if compress_progress_file or (unix and got_program("gzip")):
                 if paranoid_file_management: open(progressFile) # ensure ready
-                expr = os.popen('gzip -fdc "'+progressFile+'"',"rb").read()
+                expr = readB(os.popen('gzip -fdc "'+progressFile+'"',popenRB))
             else: expr = read(progressFile)
         else: expr = None
         if expr:
-            expr = u8strip(expr) # just in case progress.txt has been edited in Notepad
+            expr = u8strip(expr).replace(B("\r\n"),B("\n")) # just in case progress.txt has been edited in Notepad
             # First, try evaluating it as self.data (legacy progress.txt from older versions).  If that doesn't work, execute it (newer versions).
             global firstLanguage, secondLanguage, otherLanguages
             try: self.data = eval(expr)
@@ -81,49 +87,80 @@ class ProgressDatabase(object):
                 except: codeop = 0
                 if codeop: # try a lower-memory version (in case text file has been edited by hand and we're on NSLU2 or something) - don't compile all of it at once
                     lineCache = []
-                    for l in expr.replace("\r\n","\n").split("\n"):
+                    for l in expr.replace(B("\r\n"),B("\n")).split(B("\n")):
                         lineCache.append(l)
-                        if lineCache[-1].endswith(","): continue # no point trying to compile if it's obviously incomplete
-                        code = codeop.compile_command("\n".join(lineCache))
+                        if lineCache[-1].endswith(B(",")): continue # no point trying to compile if it's obviously incomplete
+                        code = codeop.compile_command("# coding=utf-8\n"+S(B("\n").join(lineCache)))
                         if code:
                             lineCache = []
-                            exec code
-                else: exec(expr)
+                            exec(code)
+                else: exec(B("# coding=utf-8\n")+expr)
             del expr
         # Remove legacy extentions in promptsData (needed only when loading from text, as this was before pickledProgressFile was added)
-        for k in self.promptsData.keys():
+        for k in list(self.promptsData.keys()):
             if k.endswith(dotwav) or k.endswith(dotmp3):
                 self.promptsData[k[:-len(dotwav)]]=self.promptsData[k]
                 del self.promptsData[k]
+        self._py3_fix()
+    def _saved_by_py3(self):
+        # NB the Windows version of Gradint is still Python 2.3 so generator expressions (new in 2.4) would be a syntax error even though this code is never reached in that version, so:
+        for l in [self.data,self.unavail]:
+            for i in l:
+                for j in i[1:]:
+                    if type(j)==str: j=[j]
+                    for k in j:
+                        for c in k:
+                            if ord(c) > 255: return True # must have been written by the Python 3 version
+    def _py3_fix(self):
+        if not type("")==type(u""): return
+        # we're Python 3, and we might have just loaded data from Python 2.  Might have to encode as Latin-1 then decode as UTF-8.  But don't do this if file was in fact saved by Python 3.
+        if self._saved_by_py3(): return
+        for l in [self.data,self.unavail]:
+            for i in range(len(l)):
+                for j in [1,2]:
+                    if type(l[i][j])==str: l[i]=l[i][:j]+(S2(LB(l[i][j])),)+l[i][j+1:]
+                    elif type(l[i][j])==list: l[i]=l[i][:j]+(map(lambda x:S2(LB(x)),l[i][j]),)+l[i][j+1:]
+    def _py3_fix_on_save(self):
+        if type("")==type(u"") and not self._saved_by_py3(): self.unavail.append((1,u"\u2014","[Py3]")) # ensure there's at least one, to prevent a py3_fix redo
     def save(self,partial=0):
         if need_say_where_put_progress: show_info("Saving "+cond(partial,"partial ","")+"progress to "+progressFile+"... ")
         else: show_info("Saving "+cond(partial,"partial ","")+"progress... ")
+        self._py3_fix_on_save()
         global progressFileBackup
         # Remove 0-repeated items (helps editing by hand)
         data = [] # don't use self.data - may want to make another lesson after saving
         for a,b,c in self.data:
             if a: data.append(denumber_filelists(a,b,c))
-        data.sort(cmpfunc) # to normalise when using diff etc
+        sort(data,cmpfunc) # to normalise when using diff etc
         if progressFileBackup:
             try:
                 import shutil
                 shutil.copy2(progressFile,progressFileBackup) # preserve timestamp etc if shutil is available
             except:
-                try:
-                    open(progressFileBackup,'wb').write(read(progressFile))
+                try: write(progressFileBackup,read(progressFile))
                 except IOError: pass # maybe progressFile not made yet
             progressFileBackup = None
         while True:
           try:
-            if compress_progress_file: f=os.popen('gzip -9 > "'+progressFile+'"','w')
+            if compress_progress_file:
+              if paranoid_file_management: fn=os.tempnam() # on some ftpfs setups gzip can fail causing silent corruption
+              else: fn=progressFile
+              f=os.popen('gzip -9 > "'+fn+'"','w')
             else: f = open(progressFile,'w')
+            global progressFileHeader
+            if type(u"")==type(""): # Python 3: ensure UTF-8
+                import codecs
+                realF,f = f,codecs.getwriter("utf-8")(f.buffer)
+                progressFileHeader=progressFileHeader.replace("mode: python ","mode: python; coding: utf-8")
+            else: realF = f
             f.write(progressFileHeader)
             f.write("firstLanguage=\"%s\"\nsecondLanguage=\"%s\"\n# otherLanguages=%s\n" % (firstLanguage,secondLanguage,otherLanguages)) # Note: they're declared "global" above (and otherLanguages commented out here for now, since may add to it in advanced.txt) (Note also save_binary below.)
-            if self.didScan: f.write("# collection=%d done=%d left=%d lessonsLeft=%d\n" % (len(self.data),len(data),len(self.data)-len(data),(len(self.data)-len(data)+maxNewWords-1)/maxNewWords))
+            if self.didScan and maxNewWords: f.write("# collection=%d done=%d left=%d lessonsLeft=%d\n" % (len(self.data),len(data),len(self.data)-len(data),(len(self.data)-len(data)+maxNewWords-1)/maxNewWords))
             prettyPrintLongList(f,"self.data",data)
             f.write("self.promptsData=") ; pprint.PrettyPrinter(indent=2,width=60,stream=f).pprint(self.promptsData)
             prettyPrintLongList(f,"self.unavail",self.unavail)
-            f.close()
+            realF.close()
+            if compress_progress_file and paranoid_file_management: write(progressFile,read(fn)),os.remove(fn)
             self.save_binary(data)
           except IOError: # This can happen for example on some PocketPC devices if you reconnect the power during progress save (which is likely if you return the device to the charger when lesson finished)
             if app or appuifw or android:
@@ -135,10 +172,17 @@ class ProgressDatabase(object):
         if not app and not appuifw and not android: show_info("done\n")
     def save_binary(self,data): # save a pickled version if possible (no error if not)
         if not (pickledProgressFile and pickle): return
+        self._py3_fix_on_save()
         try:
-            if compress_progress_file: f=os.popen('gzip -9 > "'+pickledProgressFile+'"','wb')
+            if compress_progress_file:
+              if paranoid_file_management: fn=os.tempnam()
+              else: fn=pickledProgressFile # TODO near-duplicate code with above
+              f=os.popen('gzip -9 > "'+fn+'"',popenWB)
+              if hasattr(f,'buffer'): _,f = f,f.buffer
             else: f = open(pickledProgressFile,'wb')
             pickle.Pickler(f,-1).dump(("self.data,self.promptsData,self.unavail,firstLanguage,secondLanguage = tup", (data,self.promptsData,self.unavail,firstLanguage,secondLanguage)))
+            f.close()
+            if compress_progress_file and paranoid_file_management: write(pickledProgressFile,read(fn)),os.remove(fn)
         except IOError: pass # OK if not got permissions to do it (NB need to catch the write as well because popen won't throw, and don't have to worry about a corrupted partial binary because loader would ignore it)
     def savePartial(self,filesNotPlayed):
         curPD,curDat = self.promptsData, self.data[:] # in case want to save a more complete one later
@@ -146,7 +190,7 @@ class ProgressDatabase(object):
         if hasattr(self,"previous_filesNotPlayed"):
             i=0
             while i<len(filesNotPlayed):
-                if filesNotPlayed[i] in self.previous_filesNotPlayed: i+=1
+                if checkIn(filesNotPlayed[i],self.previous_filesNotPlayed): i+=1
                 else: del filesNotPlayed[i] # cumulative effects if managed to play it last time but not this time (and both lessons incomplete)
         self.previous_filesNotPlayed = filesNotPlayed = list2set(filesNotPlayed)
         if not filesNotPlayed:
@@ -160,7 +204,7 @@ class ProgressDatabase(object):
             l.append(self.data[i][2])
             found=0
             for ii in l:
-              if ii in filesNotPlayed:
+              if checkIn(ii,filesNotPlayed):
                   self.data[i] = self.oldData[i]
                   found=1 ; break
             if not found and not self.data[i] == self.oldData[i]: changed = 1
@@ -170,7 +214,7 @@ class ProgressDatabase(object):
     def makeLesson(self):
         global maxLenOfLesson
         self.l = Lesson()
-        self.data.sort(cmpfunc) ; jitter(self.data)
+        sort(self.data,cmpfunc) ; jitter(self.data)
         self.oldData = self.data[:] # for handling interrupts & partial progress saves
         self.exclude = {} ; self.do_as_poem = {}
         # First priority: Recently-learned old words
@@ -192,28 +236,34 @@ class ProgressDatabase(object):
         for p in poems:
             for l in p: self.do_as_poem[self.responseIndex[l]] = p
         self.addToLesson(reallyKnownThreshold,-1,1,1,-1)
+        if not self.l.events:
+            # desperado if someone really wants extra revision of few words
+            global randomDropLevel, randomDropLevel2
+            rdl,rdl2,randomDropLevel,randomDropLevel2 = randomDropLevel,randomDropLevel2,0,0
+            self.addToLesson(reallyKnownThreshold,-1,1,1,-1)
+            randomDropLevel, randomDropLevel2 = rdl,rdl2
         l = self.l ; del self.l, self.responseIndex, self.do_as_poem
         if not l.events: raise Exception("Didn't manage to put anything in the lesson")
-        if commentsToAdd: l.addSequence(commentSequence())
+        if commentsToAdd: l.addSequence(commentSequence(),False)
         if orderlessCommentsToAdd:
             for c in orderlessCommentsToAdd:
                 try:
-                    l.addSequence([GluedEvent(Glue(1,maxLenOfLesson),fileToEvent(c,""))])
+                    l.addSequence([GluedEvent(Glue(1,maxLenOfLesson),fileToEvent(c,""))],False)
                 except StretchedTooFar:
-                    sys.stderr.write("Was trying to add %s\n" % (c,))
+                    show_info(("Was trying to add %s\n" % (c,)),True)
                     raise
         # Add note on "long pause", for beginners
         longpause = "longpause_"+firstLanguage
-        if not advancedPromptThreshold and not longpause in availablePrompts.lsDic: longpause = "longpause_"+secondLanguage
+        if not advancedPromptThreshold and not checkIn(longpause,availablePrompts.lsDic): longpause = "longpause_"+secondLanguage
         o=maxLenOfLesson ; maxLenOfLesson = max(l.events)[0]
-        if longpause in availablePrompts.lsDic and self.promptsData.get(longpause,0)==0:
+        if checkIn(longpause,availablePrompts.lsDic) and self.promptsData.get(longpause,0)==0:
             try:
-                def PauseEvent(): return fileToEvent(availablePrompts.lsDic[longpause],promptsDirectory)
-                firstPauseMsg = PauseEvent()
+                def PauseEvent(longpause): return fileToEvent(availablePrompts.lsDic[longpause],promptsDirectory)
+                firstPauseMsg = PauseEvent(longpause)
                 # the 1st potentially-awkward pause is likely to be a beepThreshold-length one
-                l.addSequence([GluedEvent(Glue(1,maxLenOfLesson),CompositeEvent([firstPauseMsg,Event(max(5,beepThreshold-firstPauseMsg.length))]))])
+                l.addSequence([GluedEvent(Glue(1,maxLenOfLesson),CompositeEvent([firstPauseMsg,Event(max(5,beepThreshold-firstPauseMsg.length))]))],False)
                 while True:
-                    l.addSequence([GluedEvent(Glue(1,maxLenOfLesson),CompositeEvent([PauseEvent(),Event(50)]))])
+                    l.addSequence([GluedEvent(Glue(1,maxLenOfLesson),CompositeEvent([PauseEvent(longpause),Event(50)]))],False)
                     self.promptsData[longpause] = 1
             except StretchedTooFar: pass
         maxLenOfLesson = o
@@ -246,7 +296,7 @@ class ProgressDatabase(object):
             numFailures = 0 ; startTime = time.time() # for not taking too long
             for i in xrange(len(self.data)):
                 if maxNumToAdd>-1 and numberAdded >= maxNumToAdd: break # too many
-                if i in self.exclude: continue # already had it
+                if checkIn(i,self.exclude): continue # already had it
                 (timesDone,promptFile,zhFile)=self.data[i]
                 if timesDone < minTimesDone or (maxTimesDone>=0 and timesDone > maxTimesDone): continue # out of range this time
                 if timesDone >= knownThreshold: thisNumToTry = min(random.choice([2,3,4]),numToTry)
@@ -255,7 +305,7 @@ class ProgressDatabase(object):
                     # dropping it at random
                     self.exclude[i] = 1 # pretend we've done it
                     continue
-                if i in self.do_as_poem:
+                if checkIn(i,self.do_as_poem):
                     # this is part of a "known poem" and let's try to do it in sequence
                     self.try_add_poem(self.do_as_poem[i]) ; continue
                 oldPromptsData = self.promptsData.copy()
@@ -264,8 +314,8 @@ class ProgressDatabase(object):
                 global earliestAllowedEvent ; earliestAllowedEvent = 0
                 if not timesDone and type(promptFile)==type([]):
                     # for poems: if any previously-added new word makes part of the prompt, try to ensure this one is introduced AFTER that one
-                    for f,t in newWordTimes.items():
-                        if f in promptFile: earliestAllowedEvent = max(earliestAllowedEvent,t)
+                    for f,t in list(newWordTimes.items()):
+                        if checkIn(f,promptFile): earliestAllowedEvent = max(earliestAllowedEvent,t)
                 if not timesDone: newWordTimes[zhFile] = maxLenOfLesson # by default (couldn't fit it in).  (add even if not type(promptFile)==type([]), because it might be a first line)
                 try: self.l.addSequence(seq)
                 except StretchedTooFar: # If this happens, couldn't fit the word in anywhere.  If this is "filling in gaps" then it's likely that we won't be able to fit in any more words this lesson, so stop trying.
@@ -314,7 +364,7 @@ class ProgressDatabase(object):
             covered = 0
             for timesDone,promptFile,zhFile in self.data:
                 if timesDone: covered += 1
-            x = (covered > 1000) # arbitrary
+            x = (covered > veryExperiencedThreshold)
             self.cached_very_experienced = x
         return x
     def message(self):
@@ -323,8 +373,8 @@ class ProgressDatabase(object):
         for timesDone,promptFile,zhFile in self.data:
             if timesDone:
                 covered += 1
-                if zhFile.find(exclude_from_coverage)==-1: actualCovered += 1
-            if zhFile.find(exclude_from_coverage)==-1: actualTotal += 1
+                if B(zhFile).find(B(exclude_from_coverage))==-1: actualCovered += 1
+            if B(zhFile).find(B(exclude_from_coverage))==-1: actualTotal += 1
         l=cond(app,localise,lambda x:x)
         toRet = (l("You have %d words in your collection") % total)
         if not total==actualTotal: toRet += (" (actually %d)" % actualTotal)
@@ -339,15 +389,23 @@ def prettyPrintLongList(f,thing,data):
     if winCEsound: p=0 # don't use WinCE's PrettyPrinter here - it inconsistently escapes utf8 sequences (result can't reliably be edited in MADE etc)
     else: p=pprint.PrettyPrinter(indent=2,width=60,stream=f)
     for start in range(0,len(data),step):
+        dat = data[start:start+step]
+        if type("")==type(u""): # Python 3: probably best to output strings rather than bytes
+            for i in range(len(dat)):
+                for j in [1,2]:
+                    if type(dat[i][j])==bytes:
+                        dat[i]=dat[i][:j]+(S2(dat[i][j]),)+dat[i][j+1:]
+                    elif type(dat[i][j])==list:
+                        dat[i]=dat[i][:j]+(map(S2,dat[i][j]),)+dat[i][j+1:]
         if start: f.write(thing+"+=")
         else: f.write(thing+"=")
         if p:
             t = time.time()
-            p.pprint(data[start:start+step])
+            p.pprint(dat)
             if not start and (time.time()-t)*(len(data)/step) > 5: p=0 # machine is too slow - use faster version on next iteration
         else: # faster version - not quite as pretty
             f.write("[")
-            for d in data[start:start+step]: f.write("  "+repr(d)+",\n")
+            for d in dat: f.write("  "+repr(d)+",\n")
             f.write("]\n")
 
 def calcDropLevel(timesDone):
@@ -372,15 +430,23 @@ def cmpfunc(x,y):
     if r: return r # skipping the rest if x[0]!=y[0]
     if x[0]: return cmpfunc_test(x,y) # our special order is needed only for new words (to ensure correct order of introduction)
     def my_toString(x):
-        if type(x)==type([]): return "".join(x)
-        else: return x
-    x2 = (my_toString(x[1]).replace(os.sep,chr(0)), my_toString(x[2]).replace(os.sep,chr(0)))
-    y2 = (my_toString(y[1]).replace(os.sep,chr(0)), my_toString(y[2]).replace(os.sep,chr(0)))
+        if type(x)==type([]): return B("").join(map(B,x))
+        else: return B(x)
+    x2 = (my_toString(x[1]).replace(B(os.sep),chr(0)), my_toString(x[2]).replace(B(os.sep),chr(0)))
+    y2 = (my_toString(y[1]).replace(B(os.sep),chr(0)), my_toString(y[2]).replace(B(os.sep),chr(0)))
     return cmpfunc_test(x2,y2)
 def cmpfunc_test(x,y):
-    if x < y: return -1
-    elif x > y: return 1
-    else: return 0
+    try:
+        if x < y: return -1
+        elif x > y: return 1
+        else: return 0
+    except: # probably Python 3 can't compare list vs string
+        if x[0] < y[0]: return -1
+        elif x[0] > y[0]: return 1
+        x,y = repr(x),repr(y)
+        if x < y: return -1
+        elif x > y: return 1
+        else: return 0
 
 def denumber_filelists(r,x,y):
     if type(x)==type([]): x=map(lambda z:denumber_synth(z),x)
@@ -389,18 +455,18 @@ def denumber_filelists(r,x,y):
     else: y=denumber_synth(y)
     return (r,x,y)
 def denumber_synth(z,also_norm_extsep=0):
-    zf = z.find("!synth:")
-    if zf>-1:
+    z=B(z) ; zf = z.find(B("!synth:"))
+    if zf>=0:
         z=lower(z[zf:]) # so ignores the priority-number it had (because the vocab.txt file might have been re-organised hence changing all the numbers).  Also a .lower() so case changes don't change progress.  (Old versions of gradint said .lower() when parsing vocab.txt, but this can cause problems with things like Mc[A-Z].. in English espeak)
-        if z.endswith(dotwav) or z.endswith(dotmp3): return z[:z.rindex(extsep)] # remove legacy extensions from synth vocab
-    elif also_norm_extsep: return z.replace("\\","/").replace(".","/") # so compares equally across platforms with os.sep and extsep differences
+        if z.endswith(B(dotwav)) or z.endswith(B(dotmp3)): return z[:z.rindex(B(extsep))] # remove legacy extensions from synth vocab
+    elif also_norm_extsep: return z.replace(B("\\"),B("/")).replace(B("."),B("/")) # so compares equally across platforms with os.sep and extsep differences
     return z
 
 def norm_filelist(x,y):
-    def noext(x): return (x+extsep)[:x.rfind(extsep)] # so user can change e.g. wav to mp3 without disrupting progress.txt
-    if type(x)==type([]): x=tuple(map(lambda z:denumber_synth(noext(z),1),x))
+    def noext(x): return (B(x)+B(' '))[:B(x).rfind(B(extsep))] # so user can change e.g. wav to mp3 without disrupting progress.txt (the ' ' is simply removed if rfind returns -1)
+    if type(x)==type([]): x=tuple(map(lambda z,noext=noext:denumber_synth(noext(z),1),x))
     else: x=denumber_synth(noext(x),1)
-    if type(y)==type([]): y=tuple(map(lambda z:denumber_synth(noext(z),1),y))
+    if type(y)==type([]): y=tuple(map(lambda z,noext=noext:denumber_synth(noext(z),1),y))
     else: y=denumber_synth(noext(y),1)
     return (x,y)
 def mergeProgress(progList,scan):
@@ -426,7 +492,7 @@ def mergeProgress(progList,scan):
     renames = {}
     for (_,j,k) in scan:
         key = norm_filelist(j,k)
-        if key in proglistDict:
+        if checkIn(key,proglistDict):
             # an existing item - but in the case of synth'd vocab, we need to take the capitals/lower-case status from the scan rather than from the progress file (see comment above in denumber_synth) so:
             progList[proglistDict[key]]=(progList[proglistDict[key]][0],j,k)
         elif type(key[0])==type("") and (key[0]+key[1]).find("!synth")==-1 and ("_" in key[0] and "_" in key[1]):
@@ -439,16 +505,16 @@ def mergeProgress(progList,scan):
                 while ki>lastDirsep and not "0"<=normK[ki]<="9": ki -= 1
                 if ki<=lastDirsep: break
                 key2 = (key[0][:ki+1]+key[0][key[0].rindex("_"):],key[1][:ki+1]+key[1][key[1].rindex("_"):])
-                if key2 in proglistDict:
-                    if not key2 in renames: renames[key2] = []
+                if checkIn(key2,proglistDict):
+                    if not checkIn(key2,renames): renames[key2] = []
                     renames[key2].append((j,k))
                     found=1 ; break
                 while ki>lastDirsep and "0"<=normK[ki]<="9": ki -= 1
             if not found: progList.append((0,j,k)) # new item
         else: progList.append((0,j,k)) # ditto
         scanlistDict[key]=1
-    for k,v in renames.items():
-        if k in scanlistDict or len(v)>1: # can't make sense of this one - just add the new stuff
+    for k,v in list(renames.items()):
+        if checkIn(k,scanlistDict) or len(v)>1: # can't make sense of this one - just add the new stuff
             for jj,kk in v: progList.append((0,jj,kk))
         else: progList[proglistDict[k]]=(progList[proglistDict[k]][0],v[0][0],v[0][1])
     # finally, separate off any with non-0 progress that are
@@ -458,7 +524,7 @@ def mergeProgress(progList,scan):
     n = 0 ; unavailList = []
     while n<len(progList):
         i,j,k = progList[n]
-        if not norm_filelist(j,k) in scanlistDict:
+        if not checkIn(norm_filelist(j,k), scanlistDict):
             unavailList.append((i,j,k))
             del progList[n]
         else: n += 1
@@ -514,11 +580,11 @@ def jitter(list):
     # need to work on them more quickly, and can limit manually
     limitCounts = {} ; i = 0 ; imax = len(list)
     while i < imax:
-        if list[i][0]==0 and (list[i][-1] in limitedFiles): # or not languageof(list[i][2])==secondLanguage):
+        if list[i][0]==0 and checkIn(list[i][-1],limitedFiles): # or not languageof(list[i][2])==secondLanguage):
             # if not languageof(list[i][2])==secondLanguage: countNo="other-langs"
             # else:
             countNo = limitedFiles[list[i][-1]]
-            if not countNo in limitCounts: limitCounts [countNo] = 0
+            if not checkIn(countNo,limitCounts): limitCounts [countNo] = 0
             limitCounts [countNo] += 1
             # (below is a hack: if already moved something, set limit_words to 1.  May want to do it better than that e.g. go back and ensure the first thing only left 1 as well, or share out limit_words among any concurrently available new items that are just about to be introduced)
             if limitCounts [countNo] > cond(imax==len(list),limit_words,1) or (countNo=="other-langs" and limitCounts [countNo] > 1):
@@ -543,14 +609,14 @@ def find_known_poems(progressData):
             nextLineDic[line]=response # TODO check what would happen if 2 different poems in vocab.txt share an identical line (or if responseIndex is ambiguous in any way)
             hasPreviousLine[response]=True
     poems = []
-    for poemFirstLine in filter(lambda x:not x in hasPreviousLine,nextLineDic.keys()):
+    for poemFirstLine in filter(lambda x,hasPreviousLine=hasPreviousLine:not x in hasPreviousLine,nextLineDic.keys()):
         poemLines = [] ; line = poemFirstLine
         poem_is_viable = True
         while True:
             poemLines.append(line)
-            if not line in responseIndex or progressData[responseIndex[line]][0] < reallyKnownThreshold:
+            if not checkIn(line,responseIndex) or progressData[responseIndex[line]][0] < reallyKnownThreshold:
                 poem_is_viable = False ; break # whole poem not in database yet, or is but not well-rehearsed
-            if not line in nextLineDic: break
+            if not checkIn(line,nextLineDic): break
             line = nextLineDic[line]
         if poem_is_viable: poems.append(poemLines)
     return poems, responseIndex

gradint-build/src/loop.py → src/loop.py

-# This file is part of the source code of
-# gradint v0.9979 (c) 2002-2011 Silas S. Brown. GPL v3+.
+# This file is part of the source code of Gradint
+# (c) Silas S. Brown.
 #    This program is free software; you can redistribute it and/or modify
 #    it under the terms of the GNU General Public License as published by
 #    the Free Software Foundation; either version 3 of the License, or
@@ -18,7 +18,12 @@ def doOneLesson(dbase):
         lesson = dbase.makeLesson()
     else:
         soFar = "Re-loading saved lesson, so not scanning collection."
-        if compress_progress_file: lesson=pickle.Unpickler(os.popen('gzip -fdc "'+saveLesson+'"','rb')).load()
+        if compress_progress_file:
+            pp = os.popen('gzip -fdc "'+saveLesson+'"',popenRB)
+            if hasattr(pp,'buffer'): ppb = pp.buffer
+            else: ppb = pp
+            lesson=pickle.Unpickler(ppb).load()
+            del ppb,pp
         else: lesson=pickle.Unpickler(open(saveLesson,'rb')).load()
     if app and not dbase: app.setNotFirstTime()
     while 1:
@@ -37,18 +42,38 @@ def doOneLesson(dbase):
             app.setLabel("Starting lesson")
             app.cancelling = 0
         lesson.play()
+      if not gluedListTracker==None:
+          global lastLessonMade ; lastLessonMade = lesson # used by utils/trace.py
       if dbase and saveProgress and not dbase.saved_completely: # justSaveLesson is a no-op if not first time through lesson (because scripts that use it probably mean "save if not already save"; certainly don't mean "play if is saved")
           if cancelledFiles: dbase.savePartial(cancelledFiles)
           else: dbase.save()
           if dbase.saved_completely and app: app.setNotFirstTime() # dbase.saved_completely could have been done by EITHER of the above (e.g. overlapping partial saves)
           if saveLesson:
-              if compress_progress_file: pickle.Pickler(os.popen('gzip -9 > "'+saveLesson+'"','wb'),-1).dump(lesson)
+              if compress_progress_file:
+                  pp = os.popen('gzip -9 > "'+saveLesson+'"',popenWB)
+                  if hasattr(pp,'buffer'): ppb=pp.buffer
+                  else: ppb = pp
+                  pickle.Pickler(ppb,-1).dump(lesson) # TODO: paranoid_file_management ? (c.f. saveProgress)
+                  del ppb,pp
               else: pickle.Pickler(open(saveLesson,"wb"),-1).dump(lesson)
               saveLesson = None # so saves only the first when doing multiple lessons
               if justSaveLesson: break
       if not app and not app==None: break # close box pressed
       if not waitBeforeStart or not getYN(cond(not askAgain_explain and (not dbase or not saveProgress or dbase.saved_completely),"Hear this lesson again?",askAgain_explain+"Start this lesson again?")): break
 
+def disable_lid(restore): # for portable netbooks (like eee), can close lid & keep listening
+  if unix:
+   if app and not outputFile:
+    global oldLid,warnedAC
+    try: warnedAC
+    except: warnedAC=0
+    if (not restore) and getoutput("cat /proc/acpi/ac_adapter/AC*/state 2>/dev/null").find("off-line")>=0 and not warnedAC:
+      waitOnMessage("Some quirky Linux battery managers turn speakers off mid-lesson, so AC power is recommended.") ; warnedAC=1 # (TODO what if pull out AC during the lesson without looking at the screen?  Spoken message??)
+    ls = "et org.gnome.settings-daemon.plugins.power lid-close-" ; src=["ac","battery"]
+    if restore and oldLid[0]: return [getoutput("gsettings s"+ls+p+"-action "+q+" 2>/dev/null") for p,q in zip(src,oldLid)]
+    oldLid = [getoutput("gsettings g"+ls+p+"-action 2>/dev/null").replace("'","") for p in src]
+    if oldLid[0]: [getoutput("gsettings s"+ls+p+"-action blank 2>/dev/null") for p in src]
+
 if loadLesson==-1: loadLesson=(fileExists(saveLesson) and time.localtime(os.stat(saveLesson).st_mtime)[:3]==time.localtime()[:3])
 
 def lesson_loop():
@@ -65,7 +90,7 @@ def lesson_loop():
         dbase = ProgressDatabase()
         if not dbase.data:
             msg = "There are no words to put in the lesson."
-            if app or appuifw:
+            if app or appuifw or android:
                 drop_to_synthloop = False
                 msg = localise(msg)+"\n"+localise("Please add some words first.")
             else:

gradint-build/src/makeevent.py → src/makeevent.py

-# This file is part of the source code of
-# gradint v0.9979 (c) 2002-2011 Silas S. Brown. GPL v3+.
+# This file is part of the source code of Gradint
+# (c) Silas S. Brown.
 #    This program is free software; you can redistribute it and/or modify
 #    it under the terms of the GNU General Public License as published by
 #    the Free Software Foundation; either version 3 of the License, or
@@ -17,7 +17,7 @@ def filesToEvents(files,dirBase=None):
     # (e.g. when learning poetry, may have a composite
     # prompt)
     if not type(files)==type([]): files = [files]
-    return CompositeEvent(map(lambda x:fileToEvent(x,dirBase),files))
+    return CompositeEvent(map(lambda x,d=dirBase:fileToEvent(x,d),files))
 
 class Partials_Synth(Synth):
     # text is really a list of lists of filenames
@@ -40,11 +40,12 @@ class Partials_Synth(Synth):
         # the first syllable in 1st list can optionally be the header file to use
         fname = os.tempnam()+dotwav
         o=open(fname,"wb")
-        if not (text and text[0] and text[0][0].endswith(dotwav)): o.write(read(partialsDirectory+os.sep+"header"+dotwav))
+        if not (text and text[0] and B(text[0][0]).endswith(B(dotwav))): o.write(read(partialsDirectory+os.sep+"header"+dotwav))
         for phrase in text:
             datFileInUse = 0 ; assert type(phrase)==type([])
             for f in phrase:
-                if f in audioDataPartials:
+                f = S(f)
+                if checkIn(f,audioDataPartials):
                     datFile,offset,size = audioDataPartials[f]
                     if not datFileInUse: datFileInUse = open(partialsDirectory+os.sep+datFile,"rb")
                     datFileInUse.seek(offset) ; o.write(datFileInUse.read(size))
@@ -58,16 +59,20 @@ class Partials_Synth(Synth):
 
 def fileToEvent(fname,dirBase=None):
     if dirBase==None: dirBase=samplesDirectory
-    if dirBase: dirBase += os.sep
-    orig_fname = fname
-    if os.sep in fname and fname.find("!synth:")==-1: dirBase,fname = dirBase+fname[:fname.rindex(os.sep)+1], fname[fname.rindex(os.sep)+1:]
-    if "_" in fname: lang = languageof(fname)
+    dirBase,fname = B(dirBase),B(fname)
+    if dirBase: dirBase += B(os.sep)
+    orig_DB,orig_fname = dirBase,fname
+    if B(os.sep) in fname and fname.find(B("!synth:"))==-1: dirBase,fname = dirBase+fname[:fname.rindex(B(os.sep))+1], fname[fname.rindex(B(os.sep))+1:]
+    if B("_") in fname: lang = languageof(fname)
     else: lang="-unknown-" # so can take a simple wav file, e.g. for endAnnouncement
-    if dirBase+fname in variantFiles:
+    if checkIn(dirBase+fname,variantFiles):
         variantFiles[dirBase+fname]=variantFiles[dirBase+fname][1:]+[variantFiles[dirBase+fname][0]] # cycle through the random order of variants
-        fname=variantFiles[dirBase+fname][0]
-    if fname.lower().endswith(dottxt) and "_" in fname: fname = "!synth:"+u8strip(read(dirBase+fname)).strip(wsp)+'_'+lang
-    if fname.find("!synth:")>-1:
+        fname=B(variantFiles[dirBase+fname][0])
+    if fname.lower().endswith(B(dottxt)) and B("_") in fname:
+        ftxt = bwspstrip(u8strip(read(dirBase+fname)))
+        if not ftxt: raise MessageException(B("File ")+fname+B(" in ")+dirBase+B(" has no text in it; please fix this")) # nicer message than catching it at a lower level
+        fname = B("!synth:")+B(ftxt)+B('_')+B(lang)
+    if fname.find(B("!synth:"))>=0:
         s = synthcache_lookup(fname)
         if type(s)==type([]): # trying to synth from partials
             if filter(lambda x:not type(x)==type([]), s): # but not completely (switching between partials and synth in a long text), this is more tricky:
@@ -88,9 +93,11 @@ def fileToEvent(fname,dirBase=None):
                     e.append(Event(betweenPhrasePause))
                 e=CompositeEvent(e[:-1]) # omit trailing pause
             if not lessonIsTight(): e.length=math.ceil(e.length) # (TODO slight duplication of logic from SampleEvent c'tor)
-        elif s: e=SampleEvent(synthCache+os.sep+s) # single file in synth cache
-        else: e=synth_event(languageof(fname),textof(fname))
-        e.is_prompt=(dirBase==promptsDirectory+os.sep)
+        elif s: e=SampleEvent(synthCache+os.sep+S(s)) # single file in synth cache
+        else:
+            e=synth_event(languageof(fname),textof(fname))
+            e.file = orig_DB+orig_fname # for trace.py check_for_pictures
+        e.is_prompt=(dirBase==B(promptsDirectory+os.sep))
     else: e=SampleEvent(dirBase+fname)
     e.setOnLeaves('wordToCancel',orig_fname)
     return e
@@ -98,27 +105,30 @@ def fileToEvent(fname,dirBase=None):
 transTbl = "TRANS"+extsep+"TBL"
 if mp3web: # synth-cache must exist
     if not synthCache: synthCache = "synth-cache"
-    if not isDirectory(synthCache): os.mkdir(synthCache)
+    try: os.mkdir(synthCache)
+    except: pass # already exists, temporarily-dangling symlink, etc
 
 if synthCache:
     # this listdir() call can take ages on rpcemu if it's large
     if riscos_sound: show_info("Reading synthCache... ")
-    try: synthCache_contents = os.listdir(synthCache)
+    try: synthCache_contents = map(B,os.listdir(synthCache))
     except: synthCache_contents = synthCache = []
     for i in synthCache_contents:
-        if i.upper()==transTbl: # in case it's a different case
-            transTbl=i ; break
+        if i.upper()==B(transTbl): # in case it's a different case
+            transTbl=S(i) ; break
     synthCache_contents = list2dict(synthCache_contents) # NOT 2set, as the GUI can delete things from it
     if riscos_sound: show_info("done\n")
 synthCache_transtbl = {}
-if synthCache and transTbl in synthCache_contents:
+if synthCache and checkIn(B(transTbl),synthCache_contents):
     ensure_nodups = {} # careful of duplicate filenames being appended to trans.tbl, make sure they override previous entries
-    for l in open(synthCache+os.sep+transTbl).readlines():
-        v,k = l.strip(wsp).split(None,1)
-        if v in ensure_nodups: del synthCache_transtbl[ensure_nodups[v]]
+    for l in open(synthCache+os.sep+transTbl,'rb').readlines():
+        v,k = bwspstrip(l).split(None,1)
+        if checkIn(v,ensure_nodups): del synthCache_transtbl[ensure_nodups[v]]
         ensure_nodups[v]=k ; synthCache_transtbl[k]=v
     del ensure_nodups
-def textof(fname): return fname[fname.find('!synth:')+7:fname.rfind('_')]
+def textof(fname):
+    fname = B(fname)
+    return fname[fname.find(B('!synth:'))+7:fname.rfind(B('_'))]
 last_partials_transliteration = None
 synth_partials_cache = {} ; scl_disable_recursion = 0
 def synthcache_lookup(fname,dirBase=None,printErrors=0,justQueryCache=0,lang=None):
@@ -126,8 +136,9 @@ def synthcache_lookup(fname,dirBase=None,printErrors=0,justQueryCache=0,lang=Non
     if dirBase==None: dirBase=samplesDirectory
     if dirBase: dirBase += os.sep
     if not lang: lang = languageof(fname)
-    if fname.lower().endswith(dottxt):
-        try: fname = fname[:fname.rfind("_")]+"!synth:"+u8strip(read(dirBase+fname)).strip(wsp)+"_"+lang
+    fname = B(fname)
+    if fname.lower().endswith(B(dottxt)):
+        try: fname = fname[:fname.rfind(B("_"))]+B("!synth:")+bwspstrip(u8strip(read(S(B(dirBase)+B(fname)))))+B("_")+B(lang)
         except IOError: return 0,0 # probably trying to synthcache_lookup a file with variants without first choosing a variant (e.g. in anticipation() to check for sporadic cache entries in old words) - just ignore this
     text = textof(fname)
     useSporadic = -1 # undecided (no point accumulating counters for potentially-unbounded input)
@@ -136,10 +147,10 @@ def synthcache_lookup(fname,dirBase=None,printErrors=0,justQueryCache=0,lang=Non
     if synthCache:
       for init in "_","":
         for ext in "wav","mp3":
-            k=init+text.lower()+"_"+lang+extsep+ext
-            s=synthCache_transtbl.get(k,k)
-            if s in synthCache_contents: ret=s
-            elif s.lower().endswith(dotwav) and s[:-len(dotwav)]+dotmp3 in synthCache_contents: ret=s[:-len(dotwav)]+dotmp3
+            k=B(init)+text.lower()+B("_"+lang+extsep+ext)
+            s=B(synthCache_transtbl.get(k,k))
+            if checkIn(s,synthCache_contents): ret=s
+            elif s.lower().endswith(B(dotwav)) and checkIn(s[:-len(dotwav)]+B(dotmp3),synthCache_contents): ret=s[:-len(dotwav)]+B(dotmp3)
             else: ret=0
             if ret:
                 if justQueryCache==1: ret=(k,ret)
@@ -149,8 +160,8 @@ def synthcache_lookup(fname,dirBase=None,printErrors=0,justQueryCache=0,lang=Non
                     elif tryHarder: tryHarder=ret
                 else: return ret
     if justQueryCache==1: return 0,0
-    if lang not in synth_partials_voices: l,translit=None,None # don't bother trying to transliterate here if there aren't even any partials for that language
-    elif (lang,text) not in synth_partials_cache:
+    if not checkIn(lang,synth_partials_voices): l,translit=None,None # don't bother trying to transliterate here if there aren't even any partials for that language
+    elif not checkIn((lang,text),synth_partials_cache):
         # See if we can transliterate the text first.
         synth,translit = get_synth_if_possible(lang,0,to_transliterate=True),None
         if espeak_language_aliases.get(lang,lang) in ["zhy","zh-yue"]:
@@ -163,17 +174,17 @@ def synthcache_lookup(fname,dirBase=None,printErrors=0,justQueryCache=0,lang=Non
         if lang=="zh": t2=sort_out_pinyin_3rd_tones(pinyin_uColon_to_V(t2)) # need to do this BEFORE stripPuncEtc, for correct sandhi blocking
         phraseList = stripPuncEtc(t2.lower())
         l = [synth_from_partials(phrase,lang) for phrase in phraseList] # TODO do we really want to be able to pick new voices at every phrase?  if not, would have to pass the pause points into synth_from_partials itself
-        if None in l: # at least one of the partials-phrases failed
+        if checkIn(None,l): # at least one of the partials-phrases failed
           global scl_disable_recursion
           if len(t2)<100 or not filter(lambda x:x,l) or scl_disable_recursion: l=None # don't mix partials and synth for different parts of a short phrase, it's too confusing (TODO make the 100 configurable?)
           elif type(get_synth_if_possible(lang,0))==EkhoSynth: l=None # some faulty versions of Ekho are more likely to segfault if called on fragments (e.g. if the fragment ends with some English), so don't do this with Ekho (unless can confirm it's at least ekho_4.5-2ubuntu10.04 .. not all versions of ekho can report their version no.)
           else: # longer text and SOME can be synth'd from partials: go through it more carefully
             t2=fix_compatibility(ensure_unicode(text2.replace(chr(0),"")).replace(u"\u3002",".").replace(u"\u3001",",")).encode('utf-8')
-            for t in ".!?:;,": t2=t2.replace(t,t+chr(0))
+            for t in ".!?:;,": t2=t2.replace(B(t),B(t)+chr(0))
             l=[]
             scl_disable_recursion = 1
             for phrase in filter(lambda x:x,t2.split(chr(0))):
-              ll=synthcache_lookup("!synth:"+phrase+"_"+lang,dirBase,0,0,lang)
+              ll=synthcache_lookup(B("!synth:")+phrase+B("_"+lang),dirBase,0,0,lang)
               if type(ll)==type([]): l += ll
               else: l.append(synth_event(lang,phrase,0))
             scl_disable_recursion = 0
@@ -188,7 +199,7 @@ def synthcache_lookup(fname,dirBase=None,printErrors=0,justQueryCache=0,lang=Non
     if l: return l
     if tryHarder and not tryHarder==True: return tryHarder
     if printErrors and synthCache and not (app and winsound):
-        r = repr(text.lower()+"_"+lang)
+        r = repr(text.lower()+B("_"+lang))
         if len(r)>100: r=r[:100]+"..."
         global NICcount
         try: NICcount += 1
@@ -205,9 +216,10 @@ def can_be_synthesized(fname,dirBase=None,lang=None):
     else: return get_synth_if_possible(lang) # and this time print the warning
 def stripPuncEtc(text):
     # For sending text to synth_from_partials.  Removes spaces and punctuation from text, and returns a list of the text split into phrases.
-    for t in " -_'\"()[]": text=text.replace(t,"")
-    for t in ".!?:;": text=text.replace(t,",")
-    return filter(lambda x:x,text.split(","))
+    text = B(text)
+    for t in " -_'\"()[]": text=text.replace(B(t),B(""))
+    for t in ".!?:;": text=text.replace(B(t),B(","))
+    return filter(lambda x:x,text.split(B(",")))
 
 for zipToCheck in ["yali-voice","yali-lower","cameron-voice"]:
     if riscos_sound:
@@ -215,7 +227,10 @@ for zipToCheck in ["yali-voice","yali-lower","cameron-voice"]:
     elif not winsound: # ok if mingw32, appuifw etc (unzip_and_delete will warn)
         for d in [os.getcwd()+cwd_addSep,".."+os.sep,samplesDirectory+os.sep]:
             f=d+zipToCheck+".exe"
-            if fileExists(f): unzip_and_delete(f,ignore_fail=1) # ignore the error exit status from unzip, which will be because of extra bytes at the beginning
+            if fileExists(f):
+                unzip_and_delete(f,ignore_fail=1) # ignore the error exit status from unzip, which will be because of extra bytes at the beginning
+                try: os.unlink("setup.bat")
+                except: pass
 
 # Filename / Unicode translation - need some safety across filesystems.  synthCache(+utils) could be done this way also rather than having TRANS.TBL (however I'm not sure it would save that much code)
 non_normal_filenames = {} ; using_unicode_filenames=0
@@ -227,7 +242,7 @@ def filename2unicode(f):
     def u8_or_raw(s):
         try: return unicode(s,"utf-8")
         except UnicodeDecodeError: return unicode(s,"latin1") # (actually should try the local codepage on Windows for correct display, but at least this stops a crash)
-    if f.find("_u")>-1 or f.find("_U")>-1:
+    if f.find("_u")>=0 or f.find("_U")>=0:
         try: return unicode(f.replace("_u","\\u").replace("_U","\\u"),"unicode_escape")
         except UnicodeDecodeError: # oops, need to be more careful
             ret = []
@@ -248,7 +263,7 @@ def filename2unicode(f):
     return u
 def unicode2filename(u):
     if using_unicode_filenames: return u
-    if u in non_normal_filenames: return non_normal_filenames[u]
+    if checkIn(u,non_normal_filenames): return non_normal_filenames[u]
     f=u.encode("unicode_escape").replace("\\u","_u")
     for unsafe_char in "?+*<=": f=f.replace(unsafe_char,"_u%04x" % ord(unsafe_char))
     return f
@@ -263,7 +278,7 @@ if partialsDirectory and isDirectory(partialsDirectory):
     try:
         ela = espeak_language_aliases
         format,values = pickle.Unpickler(open(partials_cache_file,"rb")).load()
-        if format==partialsCacheFormat: exec format+"=values"
+        if format==partialsCacheFormat: exec (format+"=values")
         if not (ela==espeak_language_aliases and dirsToStat[0][0]==partialsDirectory): espeak_language_aliases,dirsToStat=ela,[]
         del ela,format,values
     except MemoryError: raise # has been known on winCEsound when we're a library module (so previous memory check didn't happen)
@@ -277,7 +292,7 @@ if partialsDirectory and isDirectory(partialsDirectory):
     langs = os.listdir(partialsDirectory)
     dirsToStat.append((partialsDirectory,os.stat(partialsDirectory)))
     audioDataPartials = {} ; synth_partials_voices = {}
-    partials_raw_mode = "header"+dotwav in langs
+    partials_raw_mode = checkIn("header"+dotwav,langs)
     for l in langs:
         try: voices = os.listdir(partialsDirectory+os.sep+l)
         except: voices = []
@@ -286,7 +301,7 @@ if partialsDirectory and isDirectory(partialsDirectory):
         for v in voices:
             if "-" in v and v[:v.index("-")] in voices:
               suffix=v[v.index("-"):]
-              if not suffix in guiVoiceOptions: guiVoiceOptions.append(suffix)
+              if not checkIn(suffix,guiVoiceOptions): guiVoiceOptions.append(suffix)
             start,mid,end = [],[],[] ; flags=0
             try: files = os.listdir(partialsDirectory+os.sep+l+os.sep+v)
             except: files = []
@@ -306,10 +321,11 @@ if partialsDirectory and isDirectory(partialsDirectory):
                     while True:
                         fftell = ff.tell()
                         char = ff.read(1)
-                        if not "0"<=char<="9": break
-                        size,fname = (char+ff.readline(256)).strip(wsp).split(None,1)
+                        if not B("0")<=char<=B("9"): break
+                        size,fname = bwspstrip(char+ff.readline(256)).split(None,1)
                         try: size=int(size)
                         except: break # binary just happened to start with "0"-"9"
+                        fname = S(fname)
                         addFile(fname)
                         amend.append(l+os.sep+v+os.sep+fname)
                         audioDataPartials[l+os.sep+v+os.sep+fname] = (f,offset,size)
@@ -319,41 +335,41 @@ if partialsDirectory and isDirectory(partialsDirectory):
                 if partials_raw_mode:
                     if not f.endswith(extsep+"raw"): return
                 elif not f.endswith(dotwav) or f.endswith(dotmp3): return
-                if f.find("-s")>-1 or f.find("-i")>-1: start.append(f) # 'start' or 'initial'
-                elif not "-" in f or f.find('-m')>-1: mid.append(f)
-                elif f.find('-e')>-1 or f.find('-f')>-1: end.append(f) # 'end' or 'finish'
+                if f.find("-s")>=0 or f.find("-i")>=0: start.append(f) # 'start' or 'initial'
+                elif not "-" in f or f.find('-m')>=0: mid.append(f)
+                elif f.find('-e')>=0 or f.find('-f')>=0: end.append(f) # 'end' or 'finish'
             for f in files: addFile(f)
             def byReverseLength(a,b): return len(b)-len(a)
-            start.sort(byReverseLength) ; mid.sort(byReverseLength) ; end.sort(byReverseLength) # important if there are some files covering multiple syllables (and do it to start,mid,end not to files initially, so as to catch files within audiodata.dat also)
+            sort(start,byReverseLength) ; sort(mid,byReverseLength) ; sort(end,byReverseLength) # important if there are some files covering multiple syllables (and do it to start,mid,end not to files initially, so as to catch files within audiodata.dat also)
             def toDict(l): # make the list of filenames into a dict of short-key -> [(long-key, filename) list].  short-key is the shortest possible key.
                 if not l: return {}
                 l2 = [] ; kLen = len(l[0])
                 for i in l:
                     if "-" in i: key=i[:i.index("-")]
                     else: key=i[:i.rindex(extsep)]
-                    if key.find("_u")>-1 or key.find("_U")>-1: # a unicode partial with a portable filename?
+                    if key.find("_u")>=0 or key.find("_U")>=0: # a unicode partial with a portable filename?
                         key = filename2unicode(key).encode('utf-8')
                     l2.append((key,i))
                     kLen=min(kLen,len(key))
                 l = {}
                 for k,i in l2:
-                    if not k[:kLen] in l: l[k[:kLen]]=[]
+                    if not checkIn(k[:kLen],l): l[k[:kLen]]=[]
                     l[k[:kLen]].append((k,i))
                 return l
             thisLangVoices.append((v,toDict(start),toDict(mid),toDict(end),flags))
         synth_partials_voices[l] = thisLangVoices
-        if l in espeak_language_aliases: partials_language_aliases[espeak_language_aliases[l]]=l
+        if checkIn(l,espeak_language_aliases): partials_language_aliases[espeak_language_aliases[l]]=l
     if riscos_sound or winCEsound: show_info("done\n")
     if pickle:
       try: pickle.Pickler(open(partials_cache_file,"wb"),-1).dump((partialsCacheFormat,eval(partialsCacheFormat)))
       except IOError: pass # ignore write errors as it's only a cache
       except OSError: pass
   if partials_raw_mode:
-    (wtype,wrate,wchannels,wframes,wbits) = sndhdr.what(partialsDirectory+os.sep+"header"+dotwav)
-    partials_raw_0bytes = int(betweenPhrasePause*wrate)*wchannels*(wbits/8)
+    (wtype,wrate,wchannels,wframes,wbits) = swhat(partialsDirectory+os.sep+"header"+dotwav)
+    partials_raw_0bytes = int(betweenPhrasePause*wrate)*wchannels*int(wbits/8)
 else: synth_partials_voices,partials_raw_mode = {},None
 
-if "cant" in synth_partials_voices: synth_partials_voices["zhy"]=synth_partials_voices["zh-yue"]=synth_partials_voices["cant"]
+if checkIn("cant",synth_partials_voices): synth_partials_voices["zhy"]=synth_partials_voices["zh-yue"]=synth_partials_voices["cant"]
 
 def partials_langname(lang):
     lang = espeak_language_aliases.get(lang,lang)
@@ -362,29 +378,30 @@ def partials_langname(lang):
 
 def synth_from_partials(text,lang,voice=None,isStart=1):
     lang = partials_langname(lang)
-    text=text.strip(wsp) # so whitespace between words is ignored on the recursive call
+    text=bwspstrip(B(text)) # so whitespace between words is ignored on the recursive call
     if lang=="zh": # hack for Mandarin - higher tone 5 after a tone 3 (and ma5 after 4 or 5 also)
         lastNum = None
         for i in range(len(text)):
-            if text[i] in "123456":
-                if text[i]=="5" and (lastNum=="3" or (lastNum>"3" and i>2 and text[i-2:i+1]=="ma5")): # (TODO ne5 also? but only if followed by some form of question mark, and that might have been dropped)
+            if text[i:i+1] in B("123456"):
+                if text[i:i+1]==B("5") and (lastNum==B("3") or (lastNum and lastNum>B("3") and i>2 and text[i-2:i+1]==B("ma5"))): # (TODO ne5 also? but only if followed by some form of question mark, and that might have been dropped)
                     # see if we got a "tone 6" (higher tone 5)
                     # don't worry too much if we haven't
-                    r=synth_from_partials(text[:i]+"6"+text[i+1:],lang,voice,isStart)
+                    r=synth_from_partials(text[:i]+B("6")+text[i+1:],lang,voice,isStart)
                     if r: return r
                     else: break
                 elif lastNum: break # don't look beyond 1st 2
-                lastNum = text[i]
+                lastNum = text[i:i+1]
     if not voice: # try all voices for the language, see if we can find one that can say all the necessary parts
-        if not lang in synth_partials_voices: return None
+        if not checkIn(lang,synth_partials_voices): return None
         needCalibrated=False
         if lang=="zh": # hack for Mandarin - avoid consecutive 1st tones on non-calibrated voices
             # (DON'T do 3rd tone sandhi here - that's the caller's responsibility and we don't want it done twice now there's sandhi-blocking rules)
             lastNum=None
-            for c in text:
-                if c=="1" and lastNum=="1":
+            for i in xrange(len(text)):
+                c = text[i:i+1]
+                if c==B("1") and lastNum==B("1"):
                     needCalibrated=True ; break # TODO: unless this syllable is exactly the same as the last syllable (a repeated syllable is always ok to use even if uncalibrated)
-                if c in "123456": lastNum=c
+                if c in B("123456"): lastNum=c
             # end of hack for Mandarin
         vTry = synth_partials_voices[lang]
         if voiceOption:
@@ -398,12 +415,13 @@ def synth_from_partials(text,lang,voice=None,isStart=1):
             r = synth_from_partials(text,lang,v)
             if r:
                 if partials_raw_mode and v[-1]&2: r.insert(0,"header"+dotwav)
-                return map(lambda x:lang+os.sep+v[0]+os.sep+x,r)
+                return map(lambda x,v=v,lang=lang:lang+os.sep+v[0]+os.sep+x,r)
         return None
     dir, start, mid, end, flags = voice
     def lookup_dic(text,dic):
+        text = S(text)
         if dic:
-            for k,v in dic.get(text[:len(dic.keys()[0])],[]):
+            for k,v in dic.get(text[:len(list(dic.keys())[0])],[]):
                 if text.startswith(k): return k,v
         return None,None
     if not text: return [] # (shouldn't happen)
@@ -434,12 +452,12 @@ def optimise_partial_playing(ce):
     for e in ce.eventList[1:]:
         if not soundFileType(e.file)==fileType: return ce # must be all the same type for this optimisation
     s = None
-    if fileType=="mp3" and madplay_program and not macsound and not cygwin: # (don't do this on cygwin because cygwin will require changeToDirOf and that could get awkward)
+    if fileType=="mp3" and madplay_path and mp3Player==madplay_path and not macsound and not cygwin: # (don't do this on cygwin because cygwin will require changeToDirOf and that could get awkward)
         # mp3 probably has encoding gaps etc, but we can try our best
-        if playProgram=="aplay": s=ShellEvent(madplay_program+' -q -A $Vol$'+''.join(map(lambda x:' "'+x.file+'"', ce.eventList))+' -o wav:-|aplay -q',True) # (set retryOnFail=True)
-        else: s=ShellEvent(madplay_program+' -q -A $Vol$'+''.join(map(lambda x:' "'+x.file+'"', ce.eventList)),True)
+        if wavPlayer=="aplay": s=ShellEvent(mp3Player+' -q -A $Vol$'+''.join(map(lambda x:' "'+x.file+'"', ce.eventList))+' -o wav:-|aplay -q',True) # (set retryOnFail=True)
+        else: s=ShellEvent(mp3Player+' -q -A $Vol$'+''.join(map(lambda x:' "'+x.file+'"', ce.eventList)),True)
         s.VolReplace="soundVolume_dB"
-    elif (not fileType=="mp3") and playProgram in ["aplay","sox"]:
+    elif (not fileType=="mp3") and (wavPlayer in ["aplay","sox"] or wavPlayer.strip().endswith("<")):
         # if they're all the same format, we can use sox concatenation (raw, with an unspecified-length wav header at start)
         # (don't try to do that if different formats - the low-end hardware may not take the rate conversion)
         ok=gotSox
@@ -449,21 +467,25 @@ def optimise_partial_playing(ce):
                 if not simplified_header(e.file)==format:
                     ok=False ; break
         if ok:
-            s=ShellEvent('set -o pipefail;('+'&&'.join(['cat "%s" | sox -t %s - -t wav - $Vol$ 2>/dev/null' % (ce.eventList[0].file,fileType)]+['cat "%s" | sox -t %s - -t raw - $Vol$'%(e.file,fileType) for e in ce.eventList[1:]])+')'+sox_ignoreLen+'|'+cond(playProgram=="aplay",'aplay -q','sox -t wav - '+sox_type+' '+oss_sound_device),True)
+            if wavPlayer=="aplay": wpMod="aplay -q"
+            elif wavPlayer.strip().endswith("<"): wpMod=wavPlayer.strip()[:-1] # nc etc
+            else: wpMod='sox -t wav - '+sox_type+' '+oss_sound_device
+            s=ShellEvent('set -o pipefail;('+'&&'.join(['cat "%s" | sox -t %s - -t wav - $Vol$ 2>/dev/null' % (S(ce.eventList[0].file),fileType)]+['cat "%s" | sox -t %s - -t raw - $Vol$'%(S(e.file),fileType) for e in ce.eventList[1:]])+')'+sox_ignoreLen+'|'+wpMod,True)
             s.VolReplace="sox_effect"
-        elif playProgram=="aplay" and not sox_effect: s=ShellEvent('aplay -q '+''.join(map(lambda x:' "'+x.file+'"', ce.eventList)),True) # (which is not quite as good but is the next best thing) (and hope they don't then try to re-play a saved lesson with a volume adjustment)
+        elif wavPlayer=="aplay" and not sox_effect: s=ShellEvent('aplay -q '+''.join(map(lambda x:' "'+x.file+'"', ce.eventList)),True) # (which is not quite as good but is the next best thing) (and hope they don't then try to re-play a saved lesson with a volume adjustment)
     if s:
         s.length = ce.length
         s.equivalent_event_list = ce.eventList
         return s
     else: return ce # can't figure out an optimisation in these circumstances
 def simplified_header(fname):
-    h=sndhdr.what(fname)
+    # called by optimise_partial_playing(_list)
+    h=swhat(fname)
     # ignore num frames i.e. h[3], just compare formats
     if h: return h[:3]+h[4:]
 def optimise_partial_playing_list(ceList):
     # similar to above, but returns a ShellEvent for a list of ce's that are to be separated by short pauses, or None if can't do this optimisation.  This is because sox on NSLU2's etc has too much latency for the short pauses.
-    if (soundCollector and not saveLesson) or not playProgram=="aplay" or not gotSox: return
+    if (soundCollector and not saveLesson) or not wavPlayer=="aplay" or not gotSox: return
     format = None ; l = [] ; theLen = 0
     for ce in ceList:
         for e in ce.eventList:

gradint-build/src/play.py → src/play.py

-# This file is part of the source code of
-# gradint v0.9979 (c) 2002-2011 Silas S. Brown. GPL v3+.
+# This file is part of the source code of Gradint
+# (c) Silas S. Brown.
 #    This program is free software; you can redistribute it and/or modify
 #    it under the terms of the GNU General Public License as published by
 #    the Free Software Foundation; either version 3 of the License, or
@@ -27,7 +27,7 @@ def play(event):
     timeout_time = time.time() + max(10,event.length/3) # don't loop *forever* if unable to start playing (especially if we're being used in a reminder system etc, it may be best to exit eventually)
     if lessonStartTime and not soundCollector:
         if hasattr(event,"max_lateness"): timeout_time = min(timeout_time, lessonStartTime + (copy_of_runner_events[0][2]+event.max_lateness))
-        if hasattr(event,"sequenceID") and event.sequenceID in sequenceIDs_to_cancel: timeout_time = 0
+        if hasattr(event,"sequenceID") and checkIn(event.sequenceID,sequenceIDs_to_cancel): timeout_time = 0
     play_error = "firstTime"
     while play_error and time.time()<=timeout_time: # use <= rather than < in case we have only 1sec precision
         if not play_error=="firstTime":
@@ -41,18 +41,19 @@ def play(event):
         if hasattr(event,"sequenceID"): sequenceIDs_to_cancel[event.sequenceID]=True # TODO what if its last event has "endseq" attribute, do we want to decrement wordsLeft early?
     if hasattr(event,"endseq"): wordsLeft[event.endseq] -= 1
     del copy_of_runner_events[0]
-    if soundCollector: doLabel("%d%% completed" % (soundCollector.tell()*100/lessonLen))
-    else:
-        line2 = "" # report what you'd lose if you cancel now (in case you're deciding whether to answer the phone etc)
-        new,old=wordsLeft[True],wordsLeft[False]
-        if new: line2="%d new " % new
-        if old:
-          if line2: line2 += ("+ %d old " % old)
-          else: line2="%d old words " % old
-        elif new: line2 += "words "
-        if line2: line2=cond(app or appuifw or android,"\n",", ")+line2+"remain"
-        if not lessonStartTime: lessonStartTime = time.time() # the actual time of the FIRST event (don't set it before as there may be delays).  (we're setting this at the END of the 1st event - the extra margin should be ok, and can help with start-of-lesson problems with slow disks.)
-        if finishTime and time.time() >= emergency_lessonHold_to: doLabel("%s (finish %s)%s" % (time.strftime("%H:%M",time.localtime(time.time())),time.strftime("%H:%M",time.localtime(finishTime)),line2)) # was %I:%M but don't like leading '0' in PM times.  2nd condition added because might press 'brief interrupt' while playing.
+    if soundCollector: return doLabel("%d%% completed" % (soundCollector.tell()*100/lessonLen))
+    line2 = "" # report what you'd lose if you cancel now (in case you're deciding whether to answer the phone etc), + say how many already cancelled (for diagnosing results of interruptions caused by phone events etc on those platforms)
+    new,old=wordsLeft[True],wordsLeft[False]
+    if new: line2="%d new " % new
+    if old:
+      if line2: line2 += ("+ %d old " % old)
+      else: line2="%d old words " % old
+    elif new: line2 += "words "
+    if line2:
+      line2=cond(app or appuifw or android,"\n",", ")+line2+"remain"
+      if cancelledFiles: line2 += "\n("+str(len(cancelledFiles))+" cancelled)"
+    if not lessonStartTime: lessonStartTime = time.time() # the actual time of the FIRST event (don't set it before as there may be delays).  (we're setting this at the END of the 1st event - the extra margin should be ok, and can help with start-of-lesson problems with slow disks.)
+    if finishTime and time.time() >= emergency_lessonHold_to: doLabel("%s (finish %s)%s" % (time.strftime("%H:%M",time.localtime(time.time())),time.strftime("%H:%M",time.localtime(finishTime)),line2)) # was %I:%M but don't like leading '0' in PM times.  2nd condition added because might press 'brief interrupt' while playing.
 def doLabel(labelText):
     labelText = ensure_unicode(labelText)
     if app: app.setLabel(labelText)
@@ -96,77 +97,105 @@ if ask_teacherMode:
 def maybe_unicode(label):
     if app or appuifw or android:
         try: return unicode(label,'utf-8')
-        except: return label # ??
+        except: return label # e.g. Python 3
     else: return repr(label)
 
-madplay_program = 0
-if (winsound or mingw32) and fileExists("madplay.exe"): madplay_program = "madplay.exe"
+madplay_path = None
+if (winsound or mingw32) and fileExists("madplay.exe"): madplay_path = "madplay.exe"
+elif riscos_sound:
+  amplay = "$.!Boot.Resources.!System.350.Modules.Audio.MP3.AMPlayer"
+  if fileExists(amplay): os.system(amplay) # seems it doesn't always load at boot; no-op if run again
+  else: amplay = 0
 elif unix and hasattr(os,"popen"):
-    madplay_program = os.popen("PATH=$PATH:. which madplay 2>/dev/null").read().strip(wsp)
-    if not fileExists(cond(cygwin,madplay_program+".exe",madplay_program)): madplay_program=0 # in case of a Unix 'which' returning error on stdout
-if madplay_program and not winsound and not mingw32: madplay_program='"'+madplay_program+'"' # in case there's spaces etc in the path
+  madplay_path = os.popen("PATH=$PATH:. which madplay 2>/dev/null").read()
+  try: madplay_path = wspstrip(madplay_path)
+  except: madplay_path = madplay_path.strip()
+  if not fileExists(cond(cygwin,madplay_path+".exe",madplay_path)): madplay_path="" # in case of a Unix 'which' returning error on stdout
+  if madplay_path and not winsound and not mingw32: madplay_path='"'+madplay_path+'"' # in case there's spaces etc in the path
+if madplay_path and not mp3Player: mp3Player=madplay_path
 
 def intor0(v):
     try: return int(v)
     except ValueError: return 0
+def digitPrefix(v): # in case we don't have regexp library
+    l = []
+    for d in list(v):
+        if '0' <= d <= '9': l.append(d)
+        else: break
+    return intor0(''.join(l))
 
-playProgram = mpg123 = "" ; sox_effect=""
-sox_8bit, sox_16bit, sox_ignoreLen = "-b", "-w", ""
+sox_effect=""
+sox_8bit, sox_16bit, sox_ignoreLen, sox_signed = "-b", "-w", "", "-s"
 # Older sox versions (e.g. the one bundled with Windows Gradint) recognise -b and -w only; sox v14+ recognises both that and -1/-2; newer versions recognise only -1/-2.  We check for newer versions if unix.  (TODO riscos / other?)
 soundVolume_dB = math.log(soundVolume)*(-6/math.log(0.5))
 if unix:
   if macsound: got_afplay = got_program("afplay") # 10.5+, use in preference to the bundled qtplay which requires PowerPC or Rosetta
-  sox_formats=os.popen("sox --help 2>&1").read() # NOT .lower() yet
-  if sox_formats.lower().startswith("sox: sox v"):
-    soxMaj = intor0(sox_formats[10:sox_formats.index('.')])
-    if soxMaj>=14:
-      sox_8bit, sox_16bit = "-1", "-2" # see comment above
-      if soxMaj==14 and sox_formats[13]<'3': pass
-      else: sox_ignoreLen = "|sox --ignore-length -t wav - -t wav - 2>/dev/null"
-  if sox_formats.lower().find("wav")>-1: gotSox=1
-  else:
-    gotSox=0
-    if got_program("sox"): show_warning("SOX found, but it can't handle WAV files. Ubuntu users please install libsox-fmt-all.")
+  def sox_check():
+   global sox_8bit, sox_16bit, sox_ignoreLen, sox_signed, sox_formats
+   sox_formats=os.popen("sox --help 2>&1").read() # NOT .lower() yet
+   sf2 = ' '.join(sox_formats.lower().split())
+   if sf2.startswith("sox: sox v"):
+    if sf2[10]==' ': soxMaj=15 # guess (broken HomeBrew install)
+    else: soxMaj = intor0(sf2[10:sf2.index('.')])
+   else: soxMaj=0
+   if soxMaj>=14:
+    if soxMaj==14 and sf2[13]<'3': pass
+    else: sox_ignoreLen = "|sox --ignore-length -t wav - -t wav - 2>/dev/null"
+    if soxMaj==14 and sf2[13]<'4': sox_8bit, sox_16bit = "-1", "-2" # see comment above
+    else: sox_8bit, sox_16bit, sox_signed = "-b 8", "-b 16", "-e signed-integer" # TODO: check if 14.3 accepts these also (at least 14.4 complains -2 etc is deprecated)
+   return sf2.find("wav")>=0, sf2.find("mp3")>=0
+  gotSox,soxMp3 = sox_check()
+  if macsound:
+      if not gotSox and not os.system("mv sox-14.4.2 sox && rm sox.README"): gotSox,soxMp3 = sox_check() # see if that one works instead (NB must use os.system here: our system() has not yet been defined)
+  if not gotSox and got_program("sox"):
+      if macsound or cygwin: xtra=""
+      else: xtra=". Ubuntu users please install libsox-fmt-all."
+      show_warning("SoX found but can't handle WAV, so you won't be able to write lessons to files for later"+xtra)
 else: gotSox = got_program("sox")
+wavPlayer_override = not (not wavPlayer)
 if winsound or mingw32:
-    # in winsound can use PlaySound() but better not use it for LONGER sounds - find a playProgram anyway for those (see self.length condition in play() method below)
+    # in winsound can use PlaySound() but better not use it for LONGER sounds - find a wavPlayer anyway for those (see self.length condition in play() method below)
     # (TODO sndrec32.exe loads the whole of the file into memory before playing.  but mplayer/mplay32 sometimes halts on a yes/no dialogue about settings, and Media Player can't take files on command line so needs correct file association and executable permissions.  And many of the freeware command-line players have the same limitations as winsound.)
     # TODO now that we (usually) have tkSnack bundled with the Windows version, can we try that also (with file=) before sndrec32?
-    if fileExists(os.environ.get("windir","C:\\Windows")+"\\system32\\sndrec32.exe"): playProgram = "start /min sndrec32 /play /close" # TODO could also use ShellExecute or some other utility to make it completely hidden
+    if not wavPlayer and fileExists(os.environ.get("windir","C:\\Windows")+"\\system32\\sndrec32.exe"): wavPlayer = "start /min sndrec32 /play /close" # TODO could also use ShellExecute or some other utility to make it completely hidden
 elif unix and not macsound:
-    sox_type = "-t ossdsp -s "+sox_16bit # (we will check that sox can do ossdsp below) (always specify 16-bit because if we're adjusting the volume of 8-bit wav's then we could lose too many bits in the adjustment unless we first convert to 16-bit)
+    sox_type = "-t ossdsp "+sox_signed+" "+sox_16bit # (we will check that sox can do ossdsp below) (always specify 16-bit because if we're adjusting the volume of 8-bit wav's then we could lose too many bits in the adjustment unless we first convert to 16-bit)
     if not soundVolume==1: sox_effect=" vol "+str(soundVolume)
     if sox_effect and not gotSox:
         show_warning("Warning: trying to adjust soundVolume when 'sox' is not on the system might not work")
         # (need a warning here, because if using 'aplay' then sox o/p is 2>/dev/null (see below) so a missing sox won't be obvious)
+    if sox_formats.find("alsa")>=0 and isDirectory("/dev/snd"):
+        sox_type=sox_type.replace("ossdsp","alsa")
+        oss_sound_device = " " # older versions could take "hw:0,0" but just leave at -t alsa now?
     if not oss_sound_device:
         dsps_to_check = []
-        if sox_formats.find("ossdsp")>-1: dsps_to_check += ["/dev/sound/dsp","/dev/dsp"]
-        if sox_formats.find("sunau")>-1: dsps_to_check += ["/dev/audio"]
+        if sox_formats.find("ossdsp")>=0:
+            dsps_to_check += ["/dev/sound/dsp","/dev/dsp"]
+        if sox_formats.find("sunau")>=0: dsps_to_check += ["/dev/audio"]
         for dsp in dsps_to_check:
             if fileExists_stat(dsp):
                 oss_sound_device = dsp
-                if dsp=="/dev/audio": sox_type="-t sunau -s "+sox_16bit
+                if dsp=="/dev/audio": sox_type="-t sunau "+sox_signed+" "+sox_16bit
                 break
-    if sox_formats.find("-q")>-1: sox_type="-q "+sox_type
-    # Try to find playProgram (and maybe mpg123, for use if no madplay or mp3-playing playProgram)
-    if oss_sound_device and not cygwin and gotSox: playProgram = "sox"
-    elif cygwin and got_program("sndrec32"): # XP's Sound Recorder (vista's is called soundreorder.exe but won't do this) (+ don't have to worry about the >2G memory bug as not applicable to playing)
-        playProgram = "sndrec32 /play /close" # prefer this to esdplay due to cygwin esdplay delaying every other call and being asynchronous
-        if got_program("cmd"): playProgram = "cmd /c start /min "+playProgram # TODO could also use ShellExecute or some other utility to make it completely hidden
-    elif cygwin and oss_sound_device and got_program("play"): playProgram = "play" # this is part of sox, but it'll be the sox installed in cygwin rather than any sox.exe in gradint directory from Windows version
-    else:
+    if sox_formats.find("-q")>=0: sox_type="-q "+sox_type
+    if not wavPlayer:
+      if oss_sound_device and not cygwin and gotSox: wavPlayer = "sox"
+      elif cygwin and got_program("sndrec32"): # XP's Sound Recorder (vista's is called soundreorder.exe but won't do this) (+ don't have to worry about the >2G memory bug as not applicable to playing)
+        wavPlayer = "sndrec32 /play /close" # prefer this to esdplay due to cygwin esdplay delaying every other call and being asynchronous
+        if got_program("cmd"): wavPlayer = "cmd /c start /min "+wavPlayer # TODO could also use ShellExecute or some other utility to make it completely hidden
+      elif cygwin and oss_sound_device and got_program("play"): wavPlayer = "play" # this is part of sox, but it'll be the sox installed in cygwin rather than any sox.exe in gradint directory from Windows version
+      else:
         otherPrograms = ["aplay","esdplay","auplay","wavp","playmus","mplayer","playwave","alsaplayer"] # alsaplayer is a last resort because the text-mode version may or may not be installed; hopefully they'll have alsa-utils installed which includes 'aplay'. (playwave has been known to clip some files)
         for otherProgram in otherPrograms:
             if got_program(otherProgram):
-                playProgram = otherProgram
+                wavPlayer = otherProgram
                 break
-    if not cygwin and not madplay_program:
+    if not cygwin and not mp3Player:
         for mpg in ["mpg123","mpg321","mad123","mplayer"]:
             if got_program(mpg):
-                mpg123 = mpg ; break
-    if not playProgram and not outputFile: show_warning("Warning: no known "+cond(mpg123,"non-MP3 ","")+"sound-playing command found on this system\n  (checked for sox with /dev/dsp etc, also checked for play "+" ".join(otherPrograms)+")\n - expect problems with realtime lessons"+cond(mpg123," unless everything is MP3",""))
-may_need_mp3_warning = ((playProgram or winsound or riscos_sound or mingw32) and not (mpg123 or gotSox or madplay_program))
+                mp3Player = mpg ; break
+    if not wavPlayer and not outputFile: show_warning("Warning: no known "+cond(mp3Player,"non-MP3 ","")+"sound-playing command found on this system\n  (checked for sox with /dev/dsp etc, also checked for play "+" ".join(otherPrograms)+")\n - expect problems with realtime lessons"+cond(mp3Player," unless everything is MP3",""))
+may_need_mp3_warning = ((wavPlayer or winsound or riscos_sound or mingw32) and not (mp3Player or gotSox))
 def maybe_warn_mp3():
     global may_need_mp3_warning
     if may_need_mp3_warning:
@@ -179,7 +208,7 @@ sox_same_endian = sox_little_endian = ""
 if gotSox and unix:
     # should only have to run this test if macsound (don't bother on NSLU2's etc):
     # (wav is little-endian, so if it doesn't pass the string through then it interpreted the i/p as big-endian)
-    if macsound and os.popen('echo "This is a test" | sox -t raw -r 8000 '+sox_16bit+' -s -c 1 - -t wav - 2>/dev/null').read().find("This is a test")==-1:
+    if macsound and readB(os.popen('echo "This is a test" | sox -t raw -r 8000 '+sox_16bit+' '+sox_signed+' -c 1 - -t wav - 2>/dev/null')).find(B("This is a test"))==-1:
         sox_little_endian = " -x"
         if not big_endian: sox_same_endian = " -x"
     elif big_endian: sox_little_endian = " -x"
@@ -188,35 +217,47 @@ def changeToDirOf(file,winsound_also=0):
     # used before running a non-cygwin program in the cygwin environment (due to directory differences etc)
     # and (with winsound_also) before running a program on Windows without needing to quote the filename (e.g. because some versions of eSpeak won't write to a quoted wav file when called from popen).  Note windows os.chdir DOES change the drive also.  Use this only if filename will not contain special characters (e.g. should be able to use it for temp files).
     # NB if winsound_also is set, will return file "quoted" on other systems (so can set winsound_also and not worry about whether or not it should be quoted)
+    file = S(file)
     if winCEsound and not ' ' in file: return file # don't need to quote
     elif winsound_also and not (winsound or mingw32 or cygwin): return '"'+file+'"'
     elif (cygwin or ((winsound or mingw32) and winsound_also)) and os.sep in file:
         os.chdir(file[:file.rfind(os.sep)])
         return file[file.rfind(os.sep)+1:]
-    else: return file
+    elif riscos_sound:
+        if file.find("..")==-1:
+            c = os.getcwd()
+            return c[c.index('$'):]+os.sep+file
+    return file
 
 def system(cmd):
     # Don't call os.system for commands like sound playing, because if you do then any Control-C interrupt will go to that rather than to gradint as we want, and it will pop up a large blank console window in Windows GUI-only version
-    if riscos_sound or not hasattr(os,"popen"): return os.system(cmd) # no popen
-    if unix and ';' in cmd: cmd='/bin/bash -c "'+cmd.replace('\\','\\\\').replace('"','\\"').replace('$','\\$')+'"' # not /bin/sh if it's complex
+    if riscos_sound and type("")==type(u""): # work around memory issues with os.system() in RISC OS Python 3.8 (not needed if the command is a module rather than a program)
+        import subprocess
+        return subprocess.Popen(S(cmd).replace('"','').split()).wait() # must be S() not B() here (TODO: what if space in a filename?  TODO: catch swi.error and say please install TaskRunner module?)
+    if not hasattr(os,"popen"): return os.system(cmd)
+    if unix and (';' in cmd or '<' in cmd): cmd='/bin/bash -c "'+cmd.replace('\\','\\\\').replace('"','\\"').replace('$','\\$')+'"' # not /bin/sh if it's complex
     try: r=os.popen(cmd)
     except: return os.system(cmd) # too many file descriptors open or something
     r.read() ; return r.close()
+signal=0
 if unix:
-    # Unix: make sure "kill" on gradint's pid includes the players:
+  # Unix: make sure "kill" on gradint's pid includes the players:
+  try:
     os.setpgrp()
     import signal
     def siggrp(sigNo,*args):
-        os.killpg(os.getpgrp(),sigNo)
-        os.abort() # in case still here
+        signal.signal(sigNo,signal.SIG_IGN)
+        os.killpg(os.getpgrp(),sigNo) # players etc
+        raise KeyboardInterrupt # clean up, rm tempfiles etc
     signal.signal(signal.SIGTERM,siggrp)
-else: signal=0
+  except: pass
 
 # Event(len) gives a pause of that length
 # SampleEvent extends this to actually play something:
 
 def soundFileType(file):
-    if extsep in file: return file[file.rindex(extsep)+1:].lower()
+    file,sep = B(file),B(extsep)
+    if sep in file: return S(file[file.rindex(sep)+1:].lower())
     else: return "wav"
 
 def lessonIsTight(): return maxLenOfLesson <= 10*60 * min(1.8,max(1,maxNewWords/5.0)) # ?
@@ -231,8 +272,8 @@ class SampleEvent(Event):
         if not lessonIsTight() and not useExactLen: approxLen = math.ceil(self.exactLen) # (if <=10min in lesson, don't round up to next second because we want a tighter fit)
         Event.__init__(self,approxLen)
     def __repr__(self):
-        if use_unicode_filenames: return self.file.encode('utf-8')
-        else: return self.file
+        if use_unicode_filenames: return self.file.encode('utf-8') # winCEsound, will be Python 2
+        else: return S(self.file)
     def __del__(self):
       if hasattr(self,"isTemp"):
         import time,os # in case gc'd
@@ -240,14 +281,17 @@ class SampleEvent(Event):
           try: return os.unlink(self.file)
           except: time.sleep(0.2) # may have taken extra time for the player to load
           if not fileExists_stat(self.file): break # unlink suceeded and still threw exception ??
-    def makesSenseToLog(self): return not self.file.startswith(promptsDirectory) # (NB "not prompts" doesn't necessarily mean it'll be a sample - may be a customised additional comment)
+    def makesSenseToLog(self):
+        if hasattr(self,"is_prompt"): return not self.is_prompt # e.g. prompt from synth-cache
+        return not B(self.file).startswith(B(promptsDirectory)) # (NB "not prompts" doesn't necessarily mean it'll be a sample - may be a customised additional comment)
     def play(self): # returns a non-{False,0,None} value on error
-        if paranoid_file_management and not hasattr(self,"isTemp"): open(self.file) # ensure ready for reading
+        if paranoid_file_management:
+            if not hasattr(self,"isTemp"): open(self.file) # ensure ready for reading
         fileType=soundFileType(self.file)
         if soundCollector: soundCollector.addFile(self.file,self.exactLen)
         elif appuifw:
             fname = self.file
-            if not fname[1]==":": fname=os.getcwd()+cwd_addSep+fname # must be full drive:\path
+            if not B(fname[1:2])==B(":"): fname=B(os.getcwd()+cwd_addSep)+B(fname) # must be full drive:\path
             sound = audio.Sound.open(ensure_unicode(fname))
             sound.play()
             try: time.sleep(self.length) # TODO or exactLen?
@@ -256,56 +300,61 @@ class SampleEvent(Event):
             return
         elif android:
             fname = self.file
-            if not fname[0]=='/': fname=os.getcwd()+'/'+fname
-            android.mediaPlay("file://"+fname)
+            if not B(fname[0])==B('/'): fname=B(os.getcwd()+'/')+fname
+            android.mediaPlay("file://"+S(fname))
             return
-        elif fileType=="mp3" and madplay_program and not macsound and not playProgram=="aplay":
+        elif fileType=="mp3" and madplay_path and mp3Player==madplay_path and not macsound and not wavPlayer=="aplay":
             oldcwd = os.getcwd()
-            play_error = system(madplay_program+' -q -A '+str(soundVolume_dB)+' "'+changeToDirOf(self.file)+'"') # using changeToDirOf because on Cygwin it might be a non-cygwin madplay.exe that someone's put in the PATH.  And keeping the full path to madplay.exe because the PATH may contain relative directories.
+            play_error = system(mp3Player+' -q -A '+str(soundVolume_dB)+' "'+changeToDirOf(self.file)+'"') # using changeToDirOf because on Cygwin it might be a non-cygwin madplay.exe that someone's put in the PATH.  And keeping the full path to madplay.exe because the PATH may contain relative directories.
             os.chdir(oldcwd)
             return play_error
         elif winCEsound and fileType=="mp3":
             # we can handle MP3 on WinCE by opening in Media Player.  Too bad it ignores requests to run minimized.
             fname = self.file
-            if not fname[0]=="\\": fname=os.getcwd()+cwd_addSep+fname # must be full path
-            r=not ctypes.cdll.coredll.ShellExecuteEx(ctypes.byref(ShellExecuteInfo(60,File=u""+fname)))
+            if not B(fname[0])==B("\\"): fname=os.getcwd()+cwd_addSep+fname # must be full path
+            r=not ctypes.cdll.coredll.ShellExecuteEx(ctypes.byref(ShellExecuteInfo(60,File=ensure_unicode(fname))))
             time.sleep(self.length) # exactLen may not be enough
-        elif (winsound and not (self.length>10 and playProgram)) or winCEsound: # (don't use winsound for long files if another player is available - it has been known to stop prematurely)
+        elif (winsound and not (self.length>10 and wavPlayer)) or winCEsound: # (don't use winsound for long files if another player is available - it has been known to stop prematurely)
             if fileType=="mp3": file=theMp3FileCache.decode_mp3_to_tmpfile(self.file)
             else: file=self.file
             try:
                 if winsound: winsound.PlaySound(file,winsound.SND_FILENAME)
                 else: # winCEsound
                     fname = self.file
-                    if not fname[0]=="\\": fname=os.getcwd()+cwd_addSep+fname # must be full path
-                    ctypes.cdll.coredll.sndPlaySoundW(u""+fname,1) # 0=sync 1=async
+                    if not B(fname[0])==B("\\"): fname=os.getcwd()+cwd_addSep+fname # must be full path
+                    ctypes.cdll.coredll.sndPlaySoundW(ensure_unicode(fname),1) # 0=sync 1=async
                     time.sleep(self.exactLen) # if async.  Async seems to be better at avoiding crashes on some handhelds.
             except RuntimeError: return 1
         elif macsound:
-          if got_afplay: return system("afplay \"%s\"" % (self.file,))
-          else:
-            try: unicode(self.file,"ascii")
-            except UnicodeDecodeError: # qtplay can't always handle non-ASCII
-              t=os.tempnam()+self.file[self.file.rindex(extsep):]
-              open(t,"w").write(open(self.file).read())
-              ret=system("qtplay \"%s\"" % (t,))
-              os.remove(t)
-              return ret
-            return system("qtplay \"%s\"" % (self.file,))
+          if got_afplay: player="afplay"
+          else: player="qtplay"
+          try: unicode(self.file,"ascii")
+          except UnicodeDecodeError: # Mac command line can't always handle non-ASCII
+            t=os.tempnam()+self.file[self.file.rindex(extsep):]
+            write(t,open(self.file).read())
+            ret=system(player+" \"%s\"" % (t,))
+            os.remove(t)
+            return ret
+          return system(player+" \"%s\"" % (S(self.file),))
         elif riscos_sound:
-            if fileType=="mp3": file=theMp3FileCache.decode_mp3_to_tmpfile(self.file) # (TODO find a RISC OS program that can play the MP3s directly?)
+            if fileType=="mp3":
+                if amplay: return os.system("AMPlay \"%s\"" % (S(self.file),)) # module call, so no need for subprocess
+                file=theMp3FileCache.decode_mp3_to_tmpfile(self.file)
             else: file=self.file
-            system("PlayIt_Play \"%s\"" % (file,))
-        elif playProgram.find('sndrec32')>-1:
+            global playit_loaded
+            try: playit_loaded
+            except: playit_loaded=not os.system("Run PlayIt:Ensure -e 1.68") # relies on PlayIt$Path having been set by !PlayIt.!Boot when seen by filer
+            os.system("PlayIt_Play \"%s\"" % (S(file),)) # module call, so no need for subprocess; TODO: not ARMv7 compatible apparently (crashes on Pi400, sox does also, AMPlay can't play wav), saying "use mp3" in index.html for now
+        elif wavPlayer.find('sndrec32')>=0:
             if fileType=="mp3": file=theMp3FileCache.decode_mp3_to_tmpfile(self.file)
             else: file=self.file
             oldDir = os.getcwd()
             t=time.time()
-            os.system(playProgram+' "'+changeToDirOf(file)+'"') # don't need to call our version of system() here
-            if playProgram.find("start")>-1: time.sleep(max(0,self.length-(time.time()-t))) # better do this - don't want events overtaking each other if there are delays.  exactLen not always enough.  (but do subtract the time already taken, in case command extensions have been disabled and "start" is synchronous.)
+            os.system(wavPlayer+' "'+changeToDirOf(file)+'"') # don't need to call our version of system() here
+            if wavPlayer.find("start")>=0: time.sleep(max(0,self.length-(time.time()-t))) # better do this - don't want events overtaking each other if there are delays.  exactLen not always enough.  (but do subtract the time already taken, in case command extensions have been disabled and "start" is synchronous.)
             os.chdir(oldDir)
-        elif fileType=="mp3" and mpg123 and not sox_effect and not (playProgram=="aplay" and madplay_program): return system(mpg123+' "'+self.file+'"')
-        elif playProgram=="sox":
+        elif fileType=="mp3" and mp3Player and not sox_effect and not (wavPlayer=="aplay" and mp3Player==madplay_path): return system(mp3Player+' "'+S(self.file)+'"')
+        elif wavPlayer=="sox" and (soxMp3 or not fileType=="mp3"):
             # To make it more difficult:
             # sox v12.x (c. 2001) - bug when filenames contain 2 spaces together, and needs input from re-direction in this case
             # sox 14.0 on Cygwin - bug when input is from redirection, unless using cat | ..
@@ -314,7 +363,7 @@ class SampleEvent(Event):
             # riscos can't do re-direction (so hope not using a buggy sox) (but again don't have to worry about this if playing because will use PlayIt)
             # + on some setups (e.g. Linux 2.6 ALSA with OSS emulation), it can fail without returning an error code if the DSP is busy, which it might be if (for example) the previous event is done by festival and is taking slightly longer than estimated
             t = time.time()
-            play_error = system('cat "%s" | sox -t %s - %s %s%s >/dev/null' % (self.file,fileType,sox_type,oss_sound_device,sox_effect))
+            play_error = system('cat "%s" | sox -t %s - %s %s%s >/dev/null' % (S(self.file),fileType,sox_type,oss_sound_device,sox_effect))
             if play_error: return play_error
             else:
                 # no error, but did it take long enough?
@@ -323,18 +372,20 @@ class SampleEvent(Event):
                 if timeDiff==0 and self.exactLen < 1.5: return 0 # (we'll let that one off for systems that have limited clock precision)
                 if not app: show_info("play didn't take long enough - maybe ") # .. problem playing sound
                 return 1
-        elif playProgram=="aplay" and ((not fileType=="mp3") or madplay_program or gotSox):
-            if madplay_program and fileType=="mp3": return system(madplay_program+' -q -A '+str(soundVolume_dB)+' "'+self.file+'" -o wav:-|aplay -q') # changeToDirOf() not needed because this won't be cygwin (hopefully)
-            elif gotSox and (sox_effect or fileType=="mp3"): return system('cat "'+self.file+'" | sox -t '+fileType+' - -t wav '+sox_16bit+' - '+sox_effect+' 2>/dev/null|aplay -q') # (make sure o/p is 16-bit even if i/p is 8-bit, because if sox_effect says "vol 0.1" or something then applying that to 8-bit would lose too many bits)
+        elif wavPlayer=="aplay" and ((not fileType=="mp3") or madplay_path or gotSox):
+            if madplay_path and fileType=="mp3": return system(madplay_path+' -q -A '+str(soundVolume_dB)+' "'+S(self.file)+'" -o wav:-|aplay -q') # changeToDirOf() not needed because this won't be cygwin (hopefully)
+            elif gotSox and (sox_effect or fileType=="mp3"): return system('cat "'+S(self.file)+'" | sox -t '+fileType+' - -t wav '+sox_16bit+' - '+sox_effect+' 2>/dev/null|aplay -q') # (make sure o/p is 16-bit even if i/p is 8-bit, because if sox_effect says "vol 0.1" or something then applying that to 8-bit would lose too many bits)
             # (2>/dev/null to suppress sox "can't seek to fix wav header" problems, but don't pick 'au' as the type because sox wav->au conversion can take too long on NSLU2 (probably involves rate conversion))
-            else: return system('aplay -q "'+self.file+'"')
+            else: return system('aplay -q "'+S(self.file)+'"')
         # May also be able to support alsa directly with sox (aplay not needed), if " alsa" is in sox -h's output and there is /dev/snd/pcmCxDxp (e.g. /dev/snd/pcmC0D0p), but sometimes it doesn't work, so best stick with aplay
         # TODO: auplay can take -volume (int 0-100) and stdin; check esdplay capabilities also
-        elif fileType=="mp3" and mpg123: return system(mpg123+' "'+self.file+'"')
-        elif playProgram:
-            if fileType=="mp3" and not playProgram=="mplayer": file=theMp3FileCache.decode_mp3_to_tmpfile(self.file)
-            else: file=self.file
-            return system(playProgram+' "'+file+'"')
+        elif fileType=="mp3" and mp3Player and not sox_effect: return system(mp3Player+' "'+S(self.file)+'"')
+        elif wavPlayer:
+            if fileType=="mp3" and not wavPlayer=="mplayer": file=theMp3FileCache.decode_mp3_to_tmpfile(self.file)
+            else: file=S(self.file)
+            if sox_effect and wavPlayer.strip().endswith("<"): return system('sox "%s" -t wav - %s | %s' % (file,sox_effect,wavPlayer.strip()[:-1]))
+            return system(wavPlayer+' "'+file+'"')
+        elif fileType=="mp3" and mp3Player: return system(mp3Player+' "'+S(self.file)+'"') # ignore sox_effect
         else: show_warning("Don't know how to play \""+self.file+'" on this system')
 
 br_tab=[(0 , 0 , 0 , 0 , 0),
@@ -357,19 +408,25 @@ def rough_guess_mp3_length(fname):
   try:
     maybe_warn_mp3() # in case there's no mp3 player
     # (NB this is only a rough guess because it doesn't support VBR
-    # and doesn't even check all sync bits or scan beyond 128 bytes.
-    # It should be fairly quick though.)
-    head=open(fname).read(128)
-    i=head.find('\xFF')
-    b=ord(head[i+1])
+    # and doesn't even check all sync bits.  It should be fairly quick though.)
+    o = open(fname,"rb") ; i = -1
+    while True:
+      head=o.read(512)
+      if len(head)==0: raise IndexError # read the whole file and not found a \xFF byte??
+      i=head.find(LB('\xFF'))
+      if i==-1: continue
+      if i+2 < len(head): head += o.read(3)
+      o.seek(o.tell()-len(head)+i+2) ; b=ord(head[i+1:i+2])
+      if b >= 0xE0: break # valid frame header starts w. 11 1-bits (not just 8: some files with embedded images could throw that off)
+    s = o.tell() ; o.close()
     layer = 4-((b&6)>>1)
     if b&24 == 24: # bits are 11 - MPEG version is 1
       column = layer-1 # MPEG 1 layer 1, 2 or 3
     elif layer==1: column = 3 # MPEG 2+ layer 1
     else: column = 4 # MPEG 2+ layer 2+
-    bitrate = br_tab[ord(head[i+2])>>4][column]
+    bitrate = br_tab[ord(head[i+2:i+3])>>4][column]
     if bitrate==0: bitrate=48 # reasonable guess for speech
-    return filelen(fname)*8.0/(bitrate*1000)
+    return (filelen(fname)-s)*8.0/(bitrate*1000)
   except IndexError: raise Exception("Invalid MP3 header in file "+repr(fname))
 
 def filelen(fname):
@@ -378,15 +435,30 @@ def filelen(fname):
     return fileLen
 
 def lengthOfSound(file):
-    if file.lower().endswith(dotmp3): return rough_guess_mp3_length(file)
+    if B(file).lower().endswith(B(dotmp3)): return rough_guess_mp3_length(file)
     else: return pcmlen(file)
 
+if type("")==type(u""): # Python 3
+    import wave
+    def swhat(file):
+        if file.lower().endswith(os.extsep+"wav"):
+            o = wave.open(file,'rb')
+            return "wav",o.getframerate(),o.getnchannels(),o.getnframes(),8*o.getsampwidth()
+        else: # fallback non-WAV
+            import sndhdr # before Python 3.13
+            return sndhdr.what(file)
+else: # Python 2
+    import sndhdr
+    swhat = sndhdr.what
 def pcmlen(file):
-    header = sndhdr.what(file)
-    if not header: raise IOError("Problem opening file '%s'" % (file,))
+    header = swhat(file)
     (wtype,wrate,wchannels,wframes,wbits) = header
-    divisor = wrate*wchannels*wbits/8 # do NOT optimise with (wbits>>3), because wbits could be 4
-    if not divisor: raise IOError("Cannot parse sample format of '%s'" % (file,))
+    if android:
+        if wrate==6144: # might be a .3gp from android_recordFile
+            d = open(file).read()
+            if 'mdat' in d: return (len(d)-d.index('mdat'))/1500.0 # this assumes the bitrate is roughly the same as in my tests, TODO figure it out properly
+    divisor = wrate*wchannels*int(wbits/8) # do NOT optimise with (wbits>>3), because wbits could be 4
+    if not divisor: raise IOError("Cannot parse sample format of '%s': %s" % (file,repr(header)))
     return (filelen(file) - 44.0) / divisor # 44 is a typical header length, and .0 to convert to floating-point
 
 ##########################################################
@@ -395,30 +467,32 @@ class SoundCollector(object):
     def __init__(self):
         self.rate = 44100 # so ok for oggenc etc
         if out_type=="raw" and write_to_stdout: self.o=sys.stdout
-        elif out_type=="ogg": self.o=os.popen("oggenc -o \"%s\" -r -C 1 -q 0 -" % (cond(write_to_stdout,"-",outputFile),),"wb") # oggenc assumes little-endian, which is what we're going to give it
+        elif out_type=="ogg": self.o=os.popen(oggenc()+" -o \"%s\" -r -C 1 -q 0 -" % (cond(write_to_stdout,"-",outputFile),),popenWB) # oggenc assumes little-endian, which is what we're going to give it
         elif out_type=="aac":
-            if got_program("neroAacEnc"): self.o=os.popen("sox %s - -t wav - | neroAacEnc -br 32000 -if - -of \"%s\"" % (self.soxParams(),cond(write_to_stdout,"-",outputFile)),"wb") # (TODO optionally use -2pass, on a physical input file like the afconvert code)
-            else: self.o=os.popen("faac -b 32 -P%s -C 1 -o \"%s\" -" % (cond(big_endian,""," -X"),cond(write_to_stdout,"-",outputFile)),"wb") # (TODO check that faac on big-endian needs the -X removed when we're giving it little-endian.  It SHOULD if the compile is endian-dependent.)
-        elif out_type=="mp3": self.o=os.popen("lame -r%s%s -m m --vbr-new -V 9 - \"%s\"" % (lame_endian_parameters(),lame_quiet(),cond(write_to_stdout,"-",outputFile)),"wb") # (TODO check that old versions of lame won't complain about the --vbr-new switch.  And some very old hardware players may insist on MPEG-1 rather than MPEG-2, which would need different parameters)
+            if got_program("neroAacEnc"): self.o=os.popen("sox %s - -t wav - | neroAacEnc -br 32000 -if - -of \"%s\"" % (self.soxParams(),cond(write_to_stdout,"-",outputFile)),popenWB) # (TODO optionally use -2pass, on a physical input file like the afconvert code)
+            else: self.o=os.popen("faac -b 32 -P%s -C 1 -o \"%s\" -" % (cond(big_endian,""," -X"),cond(write_to_stdout,"-",outputFile)),popenWB) # (TODO check that faac on big-endian needs the -X removed when we're giving it little-endian.  It SHOULD if the compile is endian-dependent.)
+        elif out_type=="mp3": self.o=os.popen("lame -r%s%s -m m --vbr-new -V 9 - \"%s\"" % (lame_endian_parameters(),lame_quiet(),cond(write_to_stdout,"-",outputFile)),popenWB) # (TODO check that old versions of lame won't complain about the --vbr-new switch.  And some very old hardware players may insist on MPEG-1 rather than MPEG-2, which would need different parameters)
         # Older versions of gradint used BladeEnc, with these settings: "BladeEnc -br 48 -mono -rawmono STDIN \"%s\"", but lame gives much smaller files (e.g. 3.1M instead of 11M) - it handles the silences more efficiently for a start).
+        # Size regression would probably also happen if we used shineenc or comprec: fixed-point, so likely faster than floating-point lame on low-end ARM etc, but won't reduce bitrate for silences.
+        # Anyway, on a Raspberry 1 B+, oggenc runs ~5x faster than lame (probably due to the type of floating-point in use), and speexenc 10-15x faster than lame, and both produce smaller files.  So if MP3-compatibility is not required (and you're not concerned about further quality loss of any samples already stored in mp3 format) then ogg or spx would be a better choice.
         # Typical file sizes for a 30-minute lesson: OGG 2.7M, neroAacEnc 3.0M at 32000 (you might be able to put up with 1.8M at 18000 or 2.2M at 24000), MP3 3.1M, MP2 3.4M, faac 3.7M, WAV 152M
         # TODO try AAC+?  aacplusenc wavfile(or -) aacfile kbits, 10,12,14,18,20,24,32,40 (or 48 for stereo), but will need a player to test it
         # (mp2 could possibly be made a bit smaller by decreasing the -5, but don't make it as low as -10)
         elif out_type=="spx":
             self.rate = 32000 # could also use 16000 and -w, or even 8000, but those are not so good for language learning
-            self.o=os.popen("speexenc -u --vbr --dtx - "+cond(write_to_stdout,"-",outputFile),"wb") # and write 16-bit little-endian mono
+            self.o=os.popen("speexenc -u --vbr --dtx - "+cond(write_to_stdout,"-",outputFile),popenWB) # and write 16-bit little-endian mono
         elif out_type=="mp2":
             self.rate = 22050
-            self.o=os.popen("toolame %s -s %f -v -5 -p 4 -m m - \"%s\"" % (cond(big_endian,"-x",""),self.rate/1000.0,cond(write_to_stdout,"-",outputFile)),"wb") # TODO check that toolame compiled on big-endian architectures really needs -x to accept little-endian input
+            self.o=os.popen("toolame %s -s %f -v -5 -p 4 -m m - \"%s\"" % (cond(big_endian,"-x",""),self.rate/1000.0,cond(write_to_stdout,"-",outputFile)),popenWB) # TODO check that toolame compiled on big-endian architectures really needs -x to accept little-endian input
         elif not out_type=="raw":
             if out_type=="wav": self.rate=22050 # try not to take TOO much disk space
-            self.o=os.popen("sox %s - -t %s \"%s\"" % (self.soxParams(),out_type,cond(write_to_stdout,"-",outputFile)),"wb")
+            self.o=os.popen("sox %s - -t %s \"%s\"" % (self.soxParams(),out_type,cond(write_to_stdout,"-",outputFile)),popenWB)
         else: self.o = open(outputFile,"wb")
         self.theLen = 0
         self.silences = []
     def soxParams(self):
         # Have 16-bit mono, signed, little-endian
-        return ("-t raw "+sox_16bit+" -s -r %d -c 1" % (self.rate,))+sox_little_endian
+        return ("-t raw "+sox_16bit+" "+sox_signed+" -r %d -c 1" % (self.rate,))+sox_little_endian
     def tell(self):
         # How many seconds have we had?  (2 because 16-bit)
         return 1.0*self.theLen/self.rate/2
@@ -429,16 +503,16 @@ class SoundCollector(object):
         sampleNo = int(0.5+seconds*self.rate)
         if not sampleNo: sampleNo=1 # so don't lock on rounding errors
         byteNo = sampleNo*2 # since 16-bit
-        outfile_writeBytes(self.o,"\0"*byteNo)
+        outfile_writeBytes(self.o,chr(0)*byteNo)
         self.theLen += byteNo
-    def addFile(self,file,length):
+    def addFile(self,file,length): # length ignored in this version
         fileType=soundFileType(file)
-        if fileType=="mp3": file,fileType = theMp3FileCache.decode_mp3_to_tmpfile(file),"wav" # in case the system needs madplay rather than sox
+        if fileType=="mp3": file,fileType = theMp3FileCache.decode_mp3_to_tmpfile(file),"wav" # in case the system needs madplay etc rather than sox
         if riscos_sound:
-            os.system("sox -t %s \"%s\" %s tmp0" % (fileType,file,self.soxParams()))
+            system("sox -t %s \"%s\" %s tmp0" % (fileType,file,self.soxParams()))
             handle=open("tmp0","rb")
-        elif winsound or mingw32: handle = os.popen(("sox -t %s - %s - < \"%s\"" % (fileType,self.soxParams(),file)),"rb")
-        else: handle = os.popen(("cat \"%s\" | sox -t %s - %s -" % (file,fileType,self.soxParams())),"rb")
+        elif winsound or mingw32: handle = os.popen(("sox -t %s - %s - < \"%s\"" % (fileType,self.soxParams(),file)),popenRB)
+        else: handle = os.popen(("cat \"%s\" | sox -t %s - %s -" % (S(file),fileType,self.soxParams())),popenRB)
         self.theLen += outfile_writeFile(self.o,handle,file)
         if riscos_sound:
             handle.close() ; os.unlink("tmp0")
@@ -448,9 +522,9 @@ class SoundCollector(object):
             t1 = self.tell()
             self.addSilence(betweenBeeps/2.0)
             if riscos_sound:
-                os.system(beepCmd() % (self.soxParams(),"tmp0"))
+                system(beepCmd(self.soxParams(),"tmp0"))
                 data=read("tmp0") ; os.unlink("tmp0")
-            else: data=os.popen((beepCmd() % (self.soxParams(),"-")),"rb").read()
+            else: data=readB(os.popen(beepCmd(self.soxParams(),"-"),popenRB))
             outfile_writeBytes(self.o,data)
             self.theLen += len(data)
             self.addSilence(betweenBeeps/2.0)
@@ -469,7 +543,7 @@ class SoundCollector(object):
         if not app: show_info("Lengths of silences: %s (total %s)\n" % (self.silences,ttl))
         if not outputFile=="-": outfile_close(self.o)
 def outfile_writeBytes(o,bytes):
-    try: o.write(bytes)
+    try: writeB(o,bytes)
     except IOError: outfile_write_error()
 def outfile_close(o):
     try: o.close()
@@ -477,24 +551,30 @@ def outfile_close(o):
 def outfile_writeFile(o,handle,filename):
     data,theLen = 1,0
     while data:
-        data = handle.read(102400)
+        data = readB(handle,102400)
         outfile_writeBytes(o,data)
         theLen += len(data)
-    if not filename.startswith(partialsDirectory+os.sep): assert theLen, "No data when reading "+filename+": check for sox crash" # (but allow empty partials e.g. r5.  TODO if it's from EkhoSynth it could be a buggy version of Ekho)
+    if not B(filename).startswith(B(partialsDirectory+os.sep)): assert theLen, "No data when reading "+S(filename)+": check for sox crash" # (but allow empty partials e.g. r5.  TODO if it's from EkhoSynth it could be a buggy version of Ekho)
     return theLen
 def outfile_write_error(): raise IOError("Error writing to outputFile: either you are missing an encoder for "+out_type+", or the disk is full or something.")
-
+def oggenc(): # 2016: some Windows builds are now called oggenc2
+    global cached_oggenc
+    try: return cached_oggenc
+    except: pass
+    if got_program("oggenc"): cached_oggenc = "oggenc"
+    else: cached_oggenc = "oggenc2"
+    return cached_oggenc
 def lame_endian_parameters():
   # The input to lame will always be little-endian regardless of which architecture we're on and what kind of sox build we're using.
   # lame 3.97 has -x (swap endian) parameter, needed with little-endian i/p on little-endian architecture
   # lame 3.98+ has changed the default of -x and introduced explicit --big-endian and --little-endian.
   # (Note: None of this would be needed if we give lame a WAV input, as email-lesson.sh does.  But lame 3.97 on Windows faults on wav inputs.)
   lameVer = os.popen("lame --version").read()
-  if lameVer.find("version ")>-1:
+  if lameVer.find("version ")>=0:
     lameVer = lameVer[lameVer.index("version "):].split()[1]
-    if lameVer and '.' in lameVer and (lameVer[0]>'3' or intor0(lameVer[2:4])>97):
+    if lameVer and '.' in lameVer and (intor0(lameVer[:lameVer.index('.')])>3 or digitPrefix(lameVer[lameVer.index('.')+1:])>97):
       # Got 3.98+ - explicitly tell it the endianness (but check for alpha releases first - some of them don't deal with either this or the 3.97 behaviour very well)
-      if lameVer.find("alpha")>-1 and lameVer[0]=="3" and intor0(lameVer[2:4])==98: show_warning("Warning: You have a 3.98 alpha release of LAME.\nIf the MP3 file is white noise, try a different LAME version.")
+      if lameVer.find("alpha")>=0 and lameVer[0]=="3" and intor0(lameVer[2:4])==98: show_warning("Warning: You have a 3.98 alpha release of LAME.\nIf the MP3 file is white noise, try a different LAME version.")
       return " --little-endian"
   # otherwise fall-through to older lame behaviour:
   if big_endian: return "" # TODO are we sure we don't need -x on lame 3.97 PPC as well?
@@ -509,11 +589,15 @@ beepType = 0
 beepCmds = ["sox -t nul - %s %s synth trapetz 880 trim 0 0:0.05",
 "sox -t nul - %s %s synth sine 440 trim 0 0:0.05"]*3+["sox -t nul - %s %s synth trapetz 440 trim 0 0:0.05",
 "sox -t nul - %s %s synth sine 440 trim 0 0:0.05"]*2+["sox -t nul - %s %s synth 220 trim 0 0:0.05"]
-def beepCmd():
+def beepCmd(soxParams,fname):
   global beepType
   r = beepCmds[beepType]
   beepType += 1
   if beepType==len(beepCmds): beepType=0
+  if unix:
+      # not all versions of sox support -t nul; /dev/zero is safer on Unix
+      r=r.replace("-t nul -","%s /dev/zero" % (soxParams,))
+  r = r % (soxParams,fname)
   return r
 
 # -----------------------------------------------------
@@ -522,14 +606,14 @@ def beepCmd():
 class ShSoundCollector(object):
     def __init__(self):
         self.file2command = {}
-        self.commands = ["C() { echo -n $1% completed $'\r' 1>&2;}"]
+        self.commands = ["C() { echo -n $1% completed $'\r' >&2;}"]
         self.seconds = self.lastProgress = 0
         if write_to_stdout: self.o=sys.stdout
         else: self.o = open(outputFile,"wb")
         start = """#!/bin/bash
-if echo "$0"|grep / >/dev/null; then export S="$0"; else export S=$(which "$0"); fi
-export P="-t raw %s -s -r 44100 -c 1"
-tail -1 "$S" | bash\nexit\n""" % (sox_16bit,) # S=script P=params for sox (ignore endian issues because the wav header it generates below will specify the same as its natural endian-ness)
+if echo "$0"|grep / >/dev/null; then S="$0"; else S=$(which "$0"); fi
+P="-t raw %s %s -r 44100 -c 1"
+tail -1 "$S" | bash\nexit\n""" % (sox_16bit,sox_signed) # S=script P=params for sox (ignore endian issues because the wav header it generates below will specify the same as its natural endian-ness)
         outfile_writeBytes(self.o,start)
         self.bytesWritten = len(start) # need to keep a count because it might be stdout
         self.commands.append("sox $P - -t wav - </dev/null 2>/dev/null") # get the wav header with unspecified length
@@ -547,7 +631,7 @@ tail -1 "$S" | bash\nexit\n""" % (sox_16bit,) # S=script P=params for sox (ignor
         while gap > betweenBeeps+0.05:
             t1 = self.tell()
             self.addSilence(betweenBeeps/2.0)
-            self.commands.append(beepCmd() % ("$P","-"))
+            self.commands.append(beepCmd("$P","-"))
             self.seconds += 0.05
             self.addSilence(betweenBeeps/2.0)
             gap -= (self.tell()-t1)
@@ -555,9 +639,9 @@ tail -1 "$S" | bash\nexit\n""" % (sox_16bit,) # S=script P=params for sox (ignor
     def addFile(self,file,length):
         fileType=soundFileType(file)
         self.seconds += length
-        if not file in self.file2command:
+        if not checkIn(file,self.file2command):
             if fileType=="mp3": fileData,fileType = decode_mp3(file),"wav" # because remote sox may not be able to do it
-            elif compress_SH and unix: handle=os.popen("cat \""+file+"\" | sox -t "+fileType+" - -t "+fileType+" "+sox_8bit+" - 2>/dev/null","rb") # 8-bit if possible (but don't change sample rate, as we might not have floating point)
+            elif compress_SH and unix: handle=os.popen("cat \""+S(file)+"\" | sox -t "+fileType+" - -t "+fileType+" "+sox_8bit+" - 2>/dev/null",popenRB) # 8-bit if possible (but don't change sample rate, as we might not have floating point)
             else: handle = open(file,"rb")
             offset, length = self.bytesWritten, outfile_writeFile(self.o,handle,file)
             self.bytesWritten += length
@@ -579,7 +663,7 @@ tail -1 "$S" | bash\nexit\n""" % (sox_16bit,) # S=script P=params for sox (ignor
     def finished(self):
         if outputFile_appendSilence: self.addSilence(outputFile_appendSilence,False)
         outfile_writeBytes(self.o,"\n") # so "tail" has a start of a line
-        self.commands.append("C 100;echo 1>&2;exit")
+        self.commands.append("C 100;echo >&2;exit")
         for c in self.commands: outfile_writeBytes(self.o,c+"\n")
         outfile_writeBytes(self.o,"tail -%d \"$S\" | bash\n" % (len(self.commands)+1))
         if not write_to_stdout:
@@ -594,30 +678,44 @@ def dd_command(offset,length):
 warned_about_sox_decode = 0
 def warn_sox_decode():
     global warned_about_sox_decode
-    if not warned_about_sox_decode and not sox_ignoreLen:
-        show_warning("Had to use sox to decode MP3 (as no madplay etc); some versions of sox truncate the end of MP3s") # but 14.3+ (sox_ignoreLen set) should be OK
+    if not warned_about_sox_decode:
+        r = []
+        if macsound: r.append("the sox bundled with Mac Gradint was not compiled with MP3 support (please install madplay or a better sox)") # (or upgrade to a version of Mac OS that has afconvert)
+        if not sox_ignoreLen: r.append("some versions of sox truncate the end of MP3s (please upgrade sox or install madplay/mpg123)") # sox 14.3+ (sox_ignoreLen set) should be OK
+        if r: r.insert(0,"Had to use sox to decode MP3")
+        if r: show_warning('; '.join(r))
         warned_about_sox_decode = 1
-def decode_mp3(file):
+def decode_mp3(file): # Returns WAV data including header.  TODO: this assumes it's always small enough to read the whole thing into RAM (should be true if it's 1 word though, and decode_mp3 isn't usually used unless we're making a lesson file rather than running something in justSynthesize)
+    file = S(file)
     if riscos_sound:
-        warn_sox_decode()
-        os.system("sox -t mp3 \""+file+"\" -t wav"+cond(compress_SH," "+sox_8bit,"")+" tmp0")
+        warn_sox_decode() # TODO: can use madplay or AMPlay to decode if correctly installed
+        system("sox -t mp3 \""+file+"\" -t wav"+cond(compress_SH," "+sox_8bit,"")+" tmp0")
         data=read("tmp0") ; os.unlink("tmp0")
         return data
-    elif madplay_program or got_program("mpg123"):
+    elif madplay_path:
+        oldDir = os.getcwd()
+        d=readB(os.popen(madplay_path+cond(compress_SH," -R 16000 -b 8","")+" -q \""+changeToDirOf(file)+"\" -o wav:-",popenRB))
+        if winsound: d=d.replace(B("data\xFF"),B("data\x7F"),1) # sox bug workaround
+        os.chdir(oldDir) ; return d
+    elif got_program("mpg123"): # do NOT try to read its stdout (not only does it write 0 length, which we can fix, but some versions can also write wrong bitrate, which is harder for us to fix)
         oldDir = os.getcwd()
-        if madplay_program: d=os.popen(madplay_program+cond(compress_SH," -R 16000 -b 8","")+" -q \""+changeToDirOf(file)+"\" -o wav:-","rb").read()
-        else: d=os.popen("mpg123 -q -w - \""+changeToDirOf(file)+"\"","rb").read()
-        os.chdir(oldDir)
-        # fix length (especially if it's mpg123)
-        wavLen = len(d)-8 ; datLen = wavLen-36 # assumes no other chunks
-        if datLen<0: raise IOError("decode_mp3 got bad wav") # better than ValueError for the chr() in the following line
-        return d[:4] + chr(wavLen&0xFF)+chr((wavLen>>8)&0xFF)+chr((wavLen>>16)&0xFF)+chr(wavLen>>24) + d[8:40] + chr(datLen&0xFF)+chr((datLen>>8)&0xFF)+chr((datLen>>16)&0xFF)+chr(datLen>>24) + d[44:]
+        tfil = os.tempnam()+dotwav
+        system("mpg123 -q -w \""+tfil+"\" \""+changeToDirOf(file)+"\"")
+        if compress_SH and gotSox: dat = readB(os.popen("sox \""+tfil+"\" -t wav "+sox_8bit+" - ",popenRB))
+        else: dat = open(tfil,"rb").read()
+        os.unlink(tfil) ; os.chdir(oldDir) ; return dat
+    elif macsound and got_program("afconvert"):
+        tfil = os.tempnam()+dotwav
+        system("afconvert -f WAVE -d I16@44100 \""+file+"\" \""+tfil+"\"")
+        if compress_SH and gotSox: dat = readB(os.popen("sox \""+tfil+"\" -t wav "+sox_8bit+" - ",popenRB))
+        else: dat = open(tfil,"rb").read()
+        os.unlink(tfil) ; return dat
     elif unix:
-        if gotSox:
+        if soxMp3:
             warn_sox_decode()
-            return os.popen("cat \""+file+"\" | sox -t mp3 - -t wav"+cond(compress_SH," "+sox_8bit,"")+" - ","rb").read()
+            return readB(os.popen("cat \""+S(file)+"\" | sox -t mp3 - -t wav"+cond(compress_SH," "+sox_8bit,"")+" - ",popenRB))
         else:
-            show_warning("Don't know how to decode "+file+" on this system")
+            show_warning("Don't know how to decode "+file+" on this system.  Try installing madplay or mpg123.")
             return ""
     else: raise Exception("decode_mp3 called on a setup that's not Unix and doesn't have MADplay.  Need to implement non-cat sox redirect.")
 
@@ -625,14 +723,17 @@ def decode_mp3(file):
 class Mp3FileCache(object):
     def __init__(self): self.fileCache = {}
     def __del__(self):
-        import os # as it might already have been gc'd
-        for v in self.fileCache.values():
+        try: import os # as it might already have been gc'd
+        except: pass
+        try:
+          for v in self.fileCache.values():
             try: os.remove(v)
             except: pass # somebody may have removed it already
+        except: pass
     def decode_mp3_to_tmpfile(self,file):
-        if not file in self.fileCache:
+        if not checkIn(file,self.fileCache):
             self.fileCache[file] = os.tempnam()+dotwav
-            open(self.fileCache[file],"wb").write(decode_mp3(file))
+            write(self.fileCache[file],decode_mp3(file))
         return self.fileCache[file]
 theMp3FileCache = Mp3FileCache()
 
@@ -659,7 +760,7 @@ if outputFile:
     if out_type=="sh": soundCollector,sample_table_hack = ShSoundCollector(), 1
     else: soundCollector = SoundCollector()
     waitBeforeStart = 0
-    if unix and out_type in ["ogg","mp3"] and os.uname()[4].startswith("arm"): show_warning("Note: On armel, compile lame or oggenc with -fno-finite-math-only, or use lame -b 64 (or higher).  See http://martinwguy.co.uk/martin/debian/no-finite-math-only")
+    if unix and out_type in ["ogg","mp3"] and os.uname()[4].startswith("arm"): show_warning("Note: On armel, compile lame or oggenc with -fno-finite-math-only, or use lame -b 64 (or higher)")
 if not (soundCollector and out_type=="sh"): compress_SH = False # regardless of its initial setting (because it's used outside ShSoundCollector)
 def collector_time(): return soundCollector.tell()
 def collector_sleep(s): soundCollector.addSilence(s)

gradint-build/src/recording.py → src/recording.py

-# This file is part of the source code of
-# gradint v0.9979 (c) 2002-2011 Silas S. Brown. GPL v3+.
+# This file is part of the source code of Gradint
+# (c) Silas S. Brown.
 #    This program is free software; you can redistribute it and/or modify
 #    it under the terms of the GNU General Public License as published by
 #    the Free Software Foundation; either version 3 of the License, or
@@ -31,25 +31,33 @@ class MicInput(InputSource):
           self.rate = max(rates)
         else: self.rate = None
     def startRec(self,outFile,lastStopRecVal=None):
-        if not self.rate:
-            app.todo.alert="Cannot record on this system (try aoss?)"
-            return
-        self.sound = tkSnack.Sound(file=outFile, rate=self.rate, channels=1, encoding="Lin16")
-        self.sound.record()
-    def stopRec(self): self.sound.stop()
+        if not self.rate: return self.err("Cannot record on this system (try aoss?)")
+        try: self.sound = tkSnack.Sound(file=outFile, rate=self.rate, channels=1, encoding="Lin16")
+        except: return self.err("Cannot write to sound file '"+outFile+"' with tkSnack")
+        try: self.sound.record()
+        except: # e.g. waveInOpen failed on Windows 7 (driver problems?)
+            self.err("sound.record() failed")
+            try: self.sound.stop()
+            except: pass
+            try: os.remove(outFile)
+            except: pass
+            del self.sound
+    def err(self,msg): app.todo.alert=msg
+    def stopRec(self):
+        if hasattr(self,"sound"): self.sound.stop()
 
 class PlayerInput(InputSource): # play to speakers while recording to various destinations
     def __init__(self,fileToPlay,startNow=True,startTime=0): # (if startNow=False, starts when you start recording)
         global paranoid_file_management
         if use_unicode_filenames: fileToPlay=ensure_unicode(fileToPlay)
         else:
-            assert not type(fileToPlay)==type(u"")
+            assert type(fileToPlay)==type("")
             if not paranoid_file_management and filter(lambda x:ord(x)>=128,list(fileToPlay)): paranoid_file_management = True # hack to try to work around a Tkinter fault on Linux with utf-8 filenames
         if paranoid_file_management: # try to ensure it's ready for reading
             if filelen(fileToPlay)<1048576:
                 # only small - copy to temp 1st
                 self.fileToDel = os.tempnam()+fileToPlay[fileToPlay.rfind(extsep):]
-                open(self.fileToDel,"wb").write(read(fileToPlay))
+                write(self.fileToDel,read(fileToPlay))
                 fileToPlay=self.fileToDel
             else: open(fileToPlay)
         if fileToPlay.lower().endswith(dotwav) and filelen(fileToPlay)<1048576: self.sound=tkSnack.Sound(load=fileToPlay) # in-memory if <1M (saves problems with Windows keeping them open even after object deleted), TODO is this still needed now that .destroy() is called properly?  (but might be a good idea to keep it in anyway)
@@ -78,11 +86,12 @@ class PlayerInput(InputSource): # play to speakers while recording to various de
         elapsedTime = self.elapsedTime()
         if elapsedTime>=self.length-self.startSample*1.0/self.sampleRate: self.close()
         else:
-            import thread
-            def stopMe(self,thread_id):
+            try: import thread
+            except ImportError: import _thread as thread
+            def stopMe(self,thread_id,elapsedTime):
                 time.sleep(max(0.5,self.length-self.startSample*1.0/self.sampleRate-elapsedTime))
                 self.autostop(thread_id)
-            thread.start_new_thread(stopMe,(self,thread_id))
+            thread.start_new_thread(stopMe,(self,thread_id,elapsedTime))
     def elapsedTime(self):
         try: t=tkSnack.audio.elapsedTime()
         except: t=0.0
@@ -124,35 +133,16 @@ class PlayerInput(InputSource): # play to speakers while recording to various de
                 theRecorderControls.current_recordFrom_button = theRecorderControls.old_recordFrom_button
             app.todo.undoRecordFrom=True # we might not be the GUI thread
 
-if not tkSnack:
-  if macsound: # might still be able to use Audio Recorder
-    if fileExists("AudioRecorder.zip"): unzip_and_delete("AudioRecorder.zip")
-    if fileExists("Audio Recorder.app/plist"): # Audio Recorder with our special preferences list
-        runAudioRecorderYet = 0
-        def MacStartRecording():
-            global runAudioRecorderYet
-            if not runAudioRecorderYet: os.system("mv ~/Library/Preferences/com.benshan.AudioRecorder31.plist ~/Library/Preferences/com.benshan.AudioRecorder31.plist-OLD 2>/dev/null ; cp Audio\\ Recorder.app/plist ~/Library/Preferences/com.benshan.AudioRecorder31.plist; open Audio\\ Recorder.app")
-            os.system("osascript -e 'Tell application \"Audio Recorder\" to Record'")
-            runAudioRecorderYet = 1
-        def MacStopRecording(): os.system("osascript -e 'Tell application \"Audio Recorder\" to Stop'")
-        MacRecordingFile = "/tmp/audiorec-output-for-gradint.wav" # specified in the plist
-        def quitAudioRecorder():
-            if runAudioRecorderYet: os.system("osascript -e 'Tell application \"Audio Recorder\" to quit' ; rm ~/Library/Preferences/com.benshan.AudioRecorder31.plist ; mv ~/Library/Preferences/com.benshan.AudioRecorder31.plist-OLD ~/Library/Preferences/com.benshan.AudioRecorder31.plist 2>/dev/null")
-        import atexit ; atexit.register(quitAudioRecorder)
-        del MicInput
-        class MicInput(InputSource): # Mac Audio Recorder version
-            def startRec(self,outFile,lastStopRecVal=None):
-                self.fileToWrite = outFile
-                MacStartRecording()
-            def stopRec(self):
-                MacStopRecording()
-                os.rename(MacRecordingFile,self.fileToWrite)
-        tkSnack = "MicOnly"
-  elif unix and useTK and isDirectory("/dev/snd") and got_program("arecord"): # no tkSnack, but can record via ALSA (but no point doing the tests if not useTK)
+if useTK and not tkSnack:
+  if unix and isDirectory("/dev/snd") and got_program("arecord"): # can record via ALSA
+      recParams = (os.P_NOWAIT,got_program("arecord"),"arecord","-f","S16_LE","-r","22050")
+  elif gotSox: recParams = (os.P_NOWAIT,got_program("sox"),"sox","-d","-r","22050","-c","1")+tuple(sox_16bit.split()+sox_signed.split())
+  else: recParams = None
+  if recParams:
     del MicInput
     class MicInput(InputSource):
         def startRec(self,outFile,lastStopRecVal=0.5):
-            self.pid = os.spawnl(os.P_NOWAIT,"/bin/bash","/bin/bash","-c","arecord -f S16_LE -r 22050 "+shell_escape(outFile))
+            self.pid = os.spawnl(*(recParams+(outFile,)))
             time.sleep(lastStopRecVal) # allow process to start
         def stopRec(self):
             os.kill(self.pid,2) # INT
@@ -200,7 +190,7 @@ def wavToMp3(directory):
 
 def makeMp3Zips(baseDir,outDir,zipNo=0,direc=None):
     zipSplitThreshold = 5*1048576 # to be safe (as will split when it goes OVER that)
-    if baseDir==outDir: return zipNo # omit
+    if baseDir==outDir or baseDir.endswith(extsep+"zip"): return zipNo # omit the output dir, plus any existing ZIP files
     elif not direc:
         for f in os.listdir(baseDir): zipNo = makeMp3Zips(baseDir,outDir,zipNo,f)
     elif isDirectory(baseDir+os.sep+direc): zipNo = makeMp3Zips(baseDir+os.sep+direc,outDir,zipNo)
@@ -222,7 +212,7 @@ def getAmplify(directory):
     for f in os.listdir(directory):
         factor = None
         if f.endswith(dotwav) and not system("sox \""+directory+os.sep+f+"\" "+out2nul+" stat 2> \""+statfile+"\""):
-            for l in read(statfile).replace("\r","\n").split("\n"):
+            for l in S(read(statfile)).replace("\r","\n").split("\n"):
                 if l.startswith("Volume adjustment:"): factor=l.split()[2]
         if not factor: continue
         tmplist.append([float(factor),f,factor])
@@ -252,7 +242,7 @@ def doAmplify(directory,fileList,factor):
 class ButtonScrollingMixin(object):
     # expects self.ourCanvas
     def bindFocusIn(self,b):
-        b.bind('<FocusIn>',lambda *args:self.scrollIntoView(b))
+        b.bind('<FocusIn>',lambda e=None,self=self,b=b:self.scrollIntoView(b))
         if not hasattr(app,"gotFocusInHandler"):
             # (see scrollIntoView method's use of justGotFocusIn)
             app.gotFocusInHandler=1
@@ -277,7 +267,7 @@ class ButtonScrollingMixin(object):
             if by+bh<=cy+ch: return # make this the last one - don't loop consuming CPU on bottom of list
         elif by < cy: self.ourCanvas.yview("scroll","-1","units")
         else: return # done
-        app.after(10,lambda *args:self.continueScrollIntoView(button))
+        app.after(10,lambda e=None,self=self,button=button:self.continueScrollIntoView(button))
 
 class RecorderControls(ButtonScrollingMixin):
     def __init__(self):
@@ -315,7 +305,7 @@ class RecorderControls(ButtonScrollingMixin):
         self.frame.pack_forget()
         theISM.setInputSource(None)
     def addButton(self,row,col,text,command,colspan=None):
-        if (row,col) in self.coords2buttons: self.coords2buttons[(row,col)].grid_forget()
+        if checkIn((row,col),self.coords2buttons): self.coords2buttons[(row,col)].grid_forget()
         b = makeButton(self.grid,text=text,command=command)
         self.bindFocusIn(b)
         self.coords2buttons[(row,col)] = b
@@ -328,17 +318,56 @@ class RecorderControls(ButtonScrollingMixin):
         if col: self.coords2buttons[(row,col)].grid(row=row,column=col,columnspan=colspan)
         else: self.coords2buttons[(row,col)].grid(row=row,column=0,columnspan=colspan,sticky="w")
     def addLabel(self,row,col,utext):
-        if (row,col) in self.coords2buttons: self.coords2buttons[(row,col)].grid_forget()
-        self.coords2buttons[(row,col)] = self.makeLabel_lenLimit(utext)
-        self.coords2buttons[(row,col)].grid(row=row,column=col,sticky="w")
-        if col==0: self.coords2buttons[(row,col)].bind('<Button-1>',lambda *args:self.startRename(row,col,utext))
+        if checkIn((row,col),self.coords2buttons): self.coords2buttons[(row,col)].grid_forget()
+        rc = self.coords2buttons[(row,col)] = self.makeLabel_lenLimit(utext)
+        rc.grid(row=row,column=col,sticky="w")
+        if col==0:
+          rc.bind('<Button-1>',lambda e=None,self=self,row=row,col=col,utext=utext:self.startRename(row,col,utext))
+          if not winCEsound:
+            def contextMenu(e,row=row,col=col,self=self,utext=utext):
+                # TODO: document this context menu?
+                m=Tkinter.Menu(None, tearoff=0, takefocus=0)
+                m.add_command(label="Rename",command=lambda e=None,self=self,row=row,col=col,utext=utext:self.startRename(row,col,utext))
+                if self.currentDir.startswith(samplesDirectory): m.add_command(label="Add extra revision",command=lambda e=None,self=self,utext=utext:self.addRevision(utext))
+                m.add_command(label="Delete",command=lambda e=None,self=self,utext=utext:self.delete(utext))
+                m.tk_popup(e.x_root-3, e.y_root+3,entry="0")
+            rc.bind('<ButtonRelease-3>',contextMenu)
+            if macsound:
+                rc.bind('<Control-ButtonRelease-1>',contextMenu)
+                rc.bind('<ButtonRelease-2>',contextMenu)
+    if not winCEsound:
+      def delete(self,filename):
+        toDel = [] ; fs=filename.encode('utf-8')
+        for f in os.listdir(self.currentDir):
+            if f.startswith(S(fs)): toDel.append(f)
+        if not toDel: return tkMessageBox.showinfo(filename,"No files found") # shouldn't happen
+        if tkMessageBox.askyesno(filename,"Really delete "+" ".join(toDel)+"?"):
+            for d in toDel: os.remove(self.currentDir+os.sep+d)
+            self.undraw() ; self.draw() # TODO incremental update? (need to check really does affect just that row; careful with variants, synth, etc)
+      def addRevision(self,filename):
+        # c.f. gui_event_loop menu_response=="add" when already in vocabList
+        app.set_watch_cursor = 1
+        d = ProgressDatabase(0)
+        found = 0
+        curDir = self.currentDir[len(samplesDirectory)+len(os.sep):]
+        if curDir: curDir += os.sep
+        for item in d.data: # TODO: don't run this loop in the GUI thread!
+            if not item[2].startswith(curDir+filename.encode('utf-8')+"_"): continue
+            if not item[0]: break # not done yet
+            newItem0 = reviseCount(item[0])
+            if tkMessageBox.askyesno(filename,localise("Repeat count is %d. Reduce this to %d for extra revision?" % (item[0],newItem0))):
+                d.data.remove(item)
+                d.data.append((newItem0,item[1],item[2]))
+                d.save()
+            found = 1 ; break
+        if not found: tkMessageBox.showinfo(filename,localise("Repeat count is 0, so we cannot reduce it for extra revision."))
     def makeLabel_lenLimit(self,utext): return Tkinter.Label(self.grid,text=utext,wraplength=int(self.ourCanvas.winfo_screenwidth()/(1+len(self.languagesToDraw))))
     def addSynthLabel(self,filename,row,col):
-        try: ftext = ensure_unicode(u8strip(read(filename).strip(wsp)))
+        try: ftext = ensure_unicode(u8strip(bwspstrip(read(filename))))
         except IOError: return False
         l = self.makeLabel_lenLimit(ftext)
         l.grid(row=row,column=col,columnspan=2,sticky="w")
-        l.bind('<Button-1>',lambda *args:self.startSynthEdit(l,row,col,filename))
+        l.bind('<Button-1>',lambda e=None,self=self,l=l,row=row,col=col,filename=filename:self.startSynthEdit(l,row,col,filename))
         return True # do NOT put it in self.coords2buttons (not to do with space bar stuff etc)
     def startSynthEdit(self,l,row,col,filename):
         if hasattr(self,"renameToCancel"):
@@ -346,29 +375,29 @@ class RecorderControls(ButtonScrollingMixin):
           self.cancelRename(rr,cc)
         if l: l.grid_forget()
         editText,editEntry = addTextBox(self.grid,"nopack")
-        try: editText.set(ensure_unicode(u8strip(read(filename).strip(wsp))))
+        try: editText.set(ensure_unicode(u8strip(bwspstrip(read(filename)))))
         except IOError: pass
         editEntry.grid(row=row,column=col,sticky='we',columnspan=2)
-        editEntry.bind('<Return>',lambda *args:self.doEdit(editText,editEntry,row,col,filename))
-        editEntry.bind('<Escape>',lambda *args:self.cancelEdit(editEntry,row,col,filename))
+        editEntry.bind('<Return>',lambda e=None,self=self,editText=editText,editEntry=editEntry,row=row,col=col,filename=filename:self.doEdit(editText,editEntry,row,col,filename))
+        editEntry.bind('<Escape>',lambda e=None,self=self,editEntry=editEntry,row=row,col=col,filename=filename:self.cancelEdit(editEntry,row,col,filename))
         focusButton(editEntry)
         if hasattr(self.coords2buttons.get((row-1,col+1),""),"is_synth_label"):
             self.addLabel(row-1,col+1,localise("(synth'd)"))
             self.coords2buttons[(row-1,col+1)].is_synth_label = True
     def doEdit(self,editText,editEntry,row,col,filename):
-        text = asUnicode(editText.get()).encode("utf-8").strip(wsp)
-        if text: open(filename,"w").write(text+"\n")
+        text = bwspstrip(asUnicode(editText.get()).encode("utf-8"))
+        if text: writeB(open(filename,"w"),text+B("\n"))
         else:
             try: os.remove(filename)
             except: pass
         self.cancelEdit(editEntry,row,col,filename)
-        if row+1 < self.addMoreRow and (row+1,col+1) in self.coords2buttons: focusButton(self.coords2buttons[(row+1,col+1)]) # focus the next "synth" button if it exists (don't press it as well like file renaming because it might be a variant etc, TODO can we skip variants?)
+        if row+1 < self.addMoreRow and checkIn((row+1,col+1),self.coords2buttons): focusButton(self.coords2buttons[(row+1,col+1)]) # focus the next "synth" button if it exists (don't press it as well like file renaming because it might be a variant etc, TODO can we skip variants?)
     def cancelEdit(self,editEntry,row,col,filename):
         editEntry.grid_forget()
         labelAdded = self.addSynthLabel(filename,row,col)
         if hasattr(self.coords2buttons.get((row-1,col+1),""),"is_synth_label"):
             if labelAdded: self.addLabel(row-1,col+1,localise("(synth'd)"))
-            else: self.addButton(row-1,col+1,text=localise("Synthesize"),command=(lambda *args:self.startSynthEdit(None,row,col,filename)))
+            else: self.addButton(row-1,col+1,text=localise("Synthesize"),command=(lambda e=None,self=self,row=row,col=col,filename=filename:self.startSynthEdit(None,row,col,filename)))
             self.coords2buttons[(row-1,col+1)].is_synth_label = True
     def amplify(self,*args):
         self.AmplifyButton["text"] = localise("Please wait") # TODO not in the GUI thread !! (but lock our other buttons while it's doing it)
@@ -390,7 +419,7 @@ class RecorderControls(ButtonScrollingMixin):
             numZips = makeMp3Zips(self.currentDir,self.currentDir+os.sep+"zips")
             if numZips:
                 openDirectory(self.currentDir+os.sep+"zips",1)
-                if numZips>1: app.todo.alert=localise("Please send the %d zip files as %d separate messages, in case one very large message doesn't get through.") % (zipNo,zipNo)
+                if numZips>1: app.todo.alert=localise("Please send the %d zip files as %d separate messages, in case one very large message doesn't get through.") % (numZips,numZips)
                 else: app.todo.alert=localise("You may now send the zip file by email.")
             else: app.todo.alert=localise("No recordings found")
         self.undraw() ; self.draw()
@@ -398,11 +427,11 @@ class RecorderControls(ButtonScrollingMixin):
         if hasattr(self,"renameToCancel"):
           rr,cc = self.renameToCancel
           self.cancelRename(rr,cc)
-        if self.has_variants and filename.find(" (")>-1:
+        if self.has_variants and filename.find(" (")>=0:
             app.todo.alert=self.renamevar_msg
             return
         self.renameToCancel = (row,col)
-        if (row,col) in self.coords2buttons: self.coords2buttons[(row,col)].grid_forget()
+        if checkIn((row,col),self.coords2buttons): self.coords2buttons[(row,col)].grid_forget()
         renameText,renameEntry = addTextBox(self.grid,"nopack")
         renameEntry['width']=min(8,len(filename)+2)
         renameEntry.theText = renameText
@@ -419,20 +448,20 @@ class RecorderControls(ButtonScrollingMixin):
             selectAllFunc = selectAll
         class E: pass
         e=E() ; e.widget = renameEntry
-        self.ourCanvas.after(50,lambda *args:(e.widget.focus(),self.scrollIntoView(e.widget),selectAllFunc(e)))
-        renameEntry.bind('<Return>',lambda *args:self.doRename(row,col))
-        renameEntry.bind('<Escape>',lambda *args:self.cancelRename(row,col))
+        self.ourCanvas.after(50,lambda _=None,e=e,self=self,selectAllFunc=selectAllFunc:(e.widget.focus(),self.scrollIntoView(e.widget),selectAllFunc(e)))
+        renameEntry.bind('<Return>',lambda e=None,self=self,row=row,col=col:self.doRename(row,col))
+        renameEntry.bind('<Escape>',lambda e=None,self=self,row=row,col=col:self.cancelRename(row,col))
     def doRename(self,row,col):
         if hasattr(self,"renameToCancel"): del self.renameToCancel
         try: origName = self.coords2buttons[(row,col)].origName
         except AttributeError: return # event must have fired twice for some reason?
-        newNames = filter(lambda x:x,asUnicode(self.coords2buttons[(row,col)].theText.get()).split("\n")) # multiline paste, ignore blank lines
+        newNames = filter(lambda x,self=self,row=row,col=col:x,asUnicode(self.coords2buttons[(row,col)].theText.get()).split("\n")) # multiline paste, ignore blank lines
         for newName in newNames:
             if not origName: # extra lines - need to get their origNames
                 if row==self.addMoreRow: self.addMore()
-                elif not (row,col) in self.coords2buttons: row += 1 # skip extra row if there are notes
+                elif not checkIn((row,col),self.coords2buttons): row += 1 # skip extra row if there are notes
                 origName=self.coords2buttons[(row,col)]["text"]
-            if self.has_variants and origName.find(" (")>-1:
+            if self.has_variants and origName.find(" (")>=0:
                 app.todo.alert=self.renamevar_msg
                 break
             if len(newNames)>1 and not '0'<=newName[0]<='9': # multiline paste and not numbered - we'd better keep the original number
@@ -444,7 +473,7 @@ class RecorderControls(ButtonScrollingMixin):
                 except:
                     tkMessageBox.showinfo(app.master.title(),localise("Could not rename %s to %s") % (origName,newName))
                     return
-                self.addButton(row,col,text=newName,command=(lambda e=None,f=self.currentDir+os.sep+newName:self.changeDir(f)))
+                self.addButton(row,col,text=newName,command=(lambda e=None,self=self,f=self.currentDir+os.sep+newName:self.changeDir(f)))
             else: # not a directory - rename individual files
                 self.doStop() # just in case
                 for lang in list2set([firstLanguage,secondLanguage]+otherLanguages+self.languagesToDraw): # not just self.languagesToDraw, as a student of more languages than these might not have them all showing and still expect renames to work
@@ -455,53 +484,53 @@ class RecorderControls(ButtonScrollingMixin):
                         except:
                             tkMessageBox.showinfo(app.master.title(),localise("Could not rename %s to %s") % (origName+"_"+lang+ext,newName+"_"+lang+ext)) # TODO undo any that did succeed first!  + check for destination-already-exists (OS may not catch it)
                             return
-                        if not lang in self.languagesToDraw: continue
+                        if not checkIn(lang,self.languagesToDraw): continue
                         self.updateFile(unicode2filename(newName+"_"+lang+ext),row,self.languagesToDraw.index(lang),cond(ext==dottxt,0,2)) # TODO the 2 should be 1 if and only if we didn't just record it
                         updated=True
-                    if not updated and lang in self.languagesToDraw: self.updateFile(unicode2filename(newName+"_"+lang+dotwav),row,self.languagesToDraw.index(lang),0)
+                    if not updated and checkIn(lang,self.languagesToDraw): self.updateFile(unicode2filename(newName+"_"+lang+dotwav),row,self.languagesToDraw.index(lang),0)
                 self.addLabel(row,col,newName)
             # TODO what about updating progress.txt with wildcard changes (cld be going too far - we have the move script in utilities)
             origName = None # get any others from the form
             row += 1
         if len(newNames)==1 and row<self.addMoreRow: # put cursor on the next one
-            if not (row,col) in self.coords2buttons: row += 1 # skip extra row if there are notes
+            if not checkIn((row,col),self.coords2buttons): row += 1 # skip extra row if there are notes
             if row<self.addMoreRow:
               origName=self.coords2buttons[(row,col)]["text"]
               if not isDirectory(unicode2filename(self.currentDir+os.sep+origName)): self.startRename(row,0,origName)
     def cancelRename(self,row,col):
         if hasattr(self,"renameToCancel"): del self.renameToCancel
         origName = self.coords2buttons[(row,col)].origName
-        if isDirectory(unicode2filename(self.currentDir+os.sep+origName)): self.addButton(row,col,text=origName,command=(lambda e=None,f=ensure_unicode(self.currentDir+os.sep+origName).encode('utf-8'):self.changeDir(f)))
+        if isDirectory(unicode2filename(self.currentDir+os.sep+origName)): self.addButton(row,col,text=origName,command=(lambda e=None,self=self,f=ensure_unicode(self.currentDir+os.sep+origName).encode('utf-8'):self.changeDir(f)))
         else: self.addLabel(row,col,origName)
     def updateFile(self,filename,row,languageNo,state,txtExists="unknown"): # state: 0 not exist, 1 already existed, 2 we just created it
         if not os.sep in filename: filename = self.currentDir+os.sep+filename
         recFilename = filename
         if recFilename.lower().endswith(dotmp3): recFilename=recFilename[:-len(dotmp3)]+dotwav # always record in WAV; can compress to MP3 after
-        if state: # exists
-            if not tkSnack or tkSnack=="MicOnly": self.addButton(row,2+3*languageNo,text=localise("Play"),command=(lambda e=None,f=filename:(self.doStop(),SampleEvent(f).play())))  # but if got full tkSnack, might as well use setInputSource instead to be consistent with the non-_ version:
-            else: self.addButton(row,2+3*languageNo,text=localise("Play"),command=(lambda e=None,f=filename:(self.doStop(),theISM.setInputSource(PlayerInput(f,not self.syncFlag)),self.setSync(False))))
+        if state: # sound file exists
+            if not tkSnack or tkSnack=="MicOnly" or wavPlayer_override: self.addButton(row,2+3*languageNo,text=localise("Play"),command=(lambda e=None,f=filename,self=self:(self.doStop(),SampleEvent(f).play())))  # but if got full tkSnack, might as well use setInputSource instead to be consistent with the non-_ version:
+            else: self.addButton(row,2+3*languageNo,text=localise("Play"),command=(lambda e=None,f=filename,self=self:(self.doStop(),theISM.setInputSource(PlayerInput(f,not self.syncFlag)),self.setSync(False))))
             if tkSnack and (state==2 or self.always_enable_rerecord):
-                self.addButton(row,3+3*languageNo,text=localise("Re-record"),command=(lambda e=None,f=recFilename,r=row,l=languageNo:self.doRecord(f,r,l,needToUpdatePlayButton=(not filename==recFilename))))
+                self.addButton(row,3+3*languageNo,text=localise("Re-record"),command=(lambda e=None,f0=filename,f=recFilename,r=row,l=languageNo,self=self:self.doRecord(f,r,l,needToUpdatePlayButton=(not f0==f))))
             else:
                 self.addLabel(row,3+3*languageNo,"")
                 self.need_reRecord_enabler = not (not tkSnack)
-        else: # does not exist
+        else: # sound file does not exist
             synthFilename = filename[:filename.rfind(extsep)]+dottxt
             if txtExists=="unknown": txtExists=fileExists(synthFilename)
             if txtExists: self.addLabel(row,2+3*languageNo,localise("(synth'd)"))
-            elif self.always_enable_synth and get_synth_if_possible(self.languagesToDraw[languageNo],0): self.addButton(row,2+3*languageNo,text=localise("Synthesize"),command=(lambda *args:self.startSynthEdit(None,row+1,1+3*languageNo,synthFilename)))
+            elif self.always_enable_synth and get_synth_if_possible(self.languagesToDraw[languageNo],0): self.addButton(row,2+3*languageNo,text=localise("Synthesize"),command=(lambda e=None,self=self,row=row,languageNo=languageNo,synthFilename=synthFilename:self.startSynthEdit(None,row+1,1+3*languageNo,synthFilename)))
             else: self.addLabel(row,2+3*languageNo,localise("(empty)"))
             self.coords2buttons[(row,2+3*languageNo)].is_synth_label = True
             if winCEsound and not tkSnack: self.addLabel(row,3+3*languageNo,"")
-            else: self.addButton(row,3+3*languageNo,text=localise("Record"),command=(lambda e=None,f=recFilename,r=row,l=languageNo:self.doRecord(f,r,l)))
+            else: self.addButton(row,3+3*languageNo,text=localise("Record"),command=(lambda e=None,f=recFilename,r=row,l=languageNo,self=self:self.doRecord(f,r,l)))
     def add_addMore_button(self):
         if winCEsound and not tkSnack: pass # no 'add more words' button on WinCE; use PocketPC record button instead
-        else: self.addButton(self.addMoreRow,0,text=localise("Add more words"),command=(lambda *args:self.addMore()),colspan=cond(self.need_reRecord_enabler,2,4))
-        if self.need_reRecord_enabler: self.addButton(self.addMoreRow,2,text=localise("Re-record"),command=(lambda *args:self.global_rerecord()),colspan=2)
-        self.addButton(self.addMoreRow,4,text=localise("New folder"),command=(lambda *args:self.newFolder()),colspan=3)
+        else: self.addButton(self.addMoreRow,0,text=localise("Add more words"),command=(lambda e=None,self=self:self.addMore()),colspan=cond(self.need_reRecord_enabler,2,4))
+        if self.need_reRecord_enabler: self.addButton(self.addMoreRow,2,text=localise("Re-record"),command=(lambda e=None,self=self:self.global_rerecord()),colspan=2)
+        self.addButton(self.addMoreRow,4,text=localise("New folder"),command=(lambda e=None,self=self:self.newFolder()),colspan=3)
     def del_addMore_button(self):
-        if (self.addMoreRow,0) in self.coords2buttons: self.coords2buttons[(self.addMoreRow,0)].grid_forget() # old 'add more' button
-        if (self.addMoreRow,2) in self.coords2buttons: self.coords2buttons[(self.addMoreRow,2)].grid_forget() # old 're-record' button
+        if checkIn((self.addMoreRow,0),self.coords2buttons): self.coords2buttons[(self.addMoreRow,0)].grid_forget() # old 'add more' button
+        if checkIn((self.addMoreRow,2),self.coords2buttons): self.coords2buttons[(self.addMoreRow,2)].grid_forget() # old 're-record' button
         self.coords2buttons[(self.addMoreRow,4)].grid_forget() # old 'new folder' button
     def addMore(self,*args):
         self.del_addMore_button()
@@ -514,7 +543,7 @@ class RecorderControls(ButtonScrollingMixin):
                 self.gridLabel(lang,self.addMoreRow)
             self.addMoreRow += 2 ; self.maxPrefix += 1
         self.add_addMore_button()
-    def gridLabel(self,lang,row): Tkinter.Label(self.grid,text=" "+localise(cond(lang.find("-meaning_")>-1,"meaning",lang))+": ").grid(row=row,column=1+3*self.languagesToDraw.index(lang))
+    def gridLabel(self,lang,row): Tkinter.Label(self.grid,text=" "+localise(cond(lang.find("-meaning_")>=0,"meaning",lang))+": ").grid(row=row,column=1+3*self.languagesToDraw.index(lang))
     def doRecord(self,filename,row,languageNo,needToUpdatePlayButton=False):
         if not tkSnack: return tkMessageBox.showinfo(app.master.title(),localise("Sorry, cannot record on this computer because the tkSnack library (python-tksnack) is not installed."))
         theISM.startRecording(filename)
@@ -522,7 +551,7 @@ class RecorderControls(ButtonScrollingMixin):
         self.coords2buttons[(row,3+3*languageNo)]["text"]=localise("Stop")
         self.updateForStopOrChange()
         self.currentRecording = (filename,row,languageNo)
-        self.coords2buttons[(row,3+3*languageNo)]["command"]=(lambda *args:self.doStop())
+        self.coords2buttons[(row,3+3*languageNo)]["command"]=(lambda e=None,self=self:self.doStop())
         if app.scanrow.get()=="2": # "stop"
           focusButton(self.coords2buttons[(row,3+3*languageNo)])
         else:
@@ -530,14 +559,14 @@ class RecorderControls(ButtonScrollingMixin):
           if app.scanrow.get()=="1": # move along 1st
             while languageNo+1<len(self.languagesToDraw):
               languageNo += 1
-              if (row,3+3*languageNo) in self.coords2buttons:
+              if checkIn((row,3+3*languageNo),self.coords2buttons):
                   focusButton(self.coords2buttons[(row,3+3*languageNo)])
                   return
             languageNo = 0 # start of the row
           # fall-through - vertical movement
           for r in [row+1,row+2]:
             if r==self.addMoreRow: self.addMore()
-            if (r,3+3*languageNo) in self.coords2buttons:
+            if checkIn((r,3+3*languageNo),self.coords2buttons):
                 return focusButton(self.coords2buttons[(r,3+3*languageNo)])
     def doStop(self,*args):
         theISM.stopRecording()
@@ -555,9 +584,9 @@ class RecorderControls(ButtonScrollingMixin):
           if hasattr(self,"oldCanvasBbox") and bbox==self.oldCanvasBbox: pass
           else:
               self.oldCanvasBbox = bbox
-              c.config(scrollregion=bbox,width=bbox[2],height=min(c["height"],c.winfo_screenheight()/2,bbox[3]))
+              c.config(scrollregion=bbox,width=bbox[2],height=min(int(c["height"]),int(c.winfo_screenheight()/2),int(bbox[3])))
         if hasattr(self,"currentRecording") and not theISM.currentOutfile: self.doStop() # ensure GUI updates the recording button after player auto-stop (for want of a better place to put it)
-        app.after(cond(winCEsound,3000,600),lambda *args:self.reconfigure_scrollbar())
+        app.after(cond(winCEsound,3000,600),lambda e=None,self=self:self.reconfigure_scrollbar())
     def setSync(self,syncFlag): self.syncFlag = syncFlag
     def newFolder(self,*args):
         count=0
@@ -577,12 +606,12 @@ class RecorderControls(ButtonScrollingMixin):
         self.doStop()
         theISM.setInputSource(PlayerInput(filename,not self.syncFlag))
         self.current_recordFrom_button = (row, self.coords2buttons[(row,0)])
-        self.addButton(row,0,text=localise("Stop"),command=(lambda *args:(self.doStop(),theISM.setInputSource(MicInput()))),colspan=1)
+        self.addButton(row,0,text=localise("Stop"),command=(lambda e=None,self=self,theISM=theISM:(self.doStop(),theISM.setInputSource(MicInput()))),colspan=1)
         col = 1
         for inc in [-30, -5, 5, 30]:
             if inc<0: text="<"+str(-inc)
             else: text=str(inc)+">"
-            self.addButton(row,col,text=text,command=(lambda e=None,i=inc:self.handleSkip(filename,i)))
+            self.addButton(row,col,text=text,command=(lambda e=None,i=inc,self=self,filename=filename:self.handleSkip(filename,i)))
             col += 1
     def handleSkip(self,filename,i):
         self.protect_currentRecordFrom()
@@ -596,7 +625,7 @@ class RecorderControls(ButtonScrollingMixin):
         if hasattr(self,"current_recordFrom_button") and self.current_recordFrom_button:
             row, button = self.current_recordFrom_button
             for col in range(1+3*len(self.languagesToDraw)):
-                if (row,col) in self.coords2buttons:
+                if checkIn((row,col),self.coords2buttons):
                     self.coords2buttons[(row,col)].grid_forget()
                     del self.coords2buttons[(row,col)]
             button.grid(row=row,column=0,columnspan=1+3*len(self.languagesToDraw),sticky="w")
@@ -635,17 +664,23 @@ class RecorderControls(ButtonScrollingMixin):
         else: self.languagesToDraw = [secondLanguage,firstLanguage] # each lang cn take 3 columns, starting at column 1 (DO need to regenerate this every draw - languages may have changed!)
         if self.currentDir==samplesDirectory: app.master.title(localise("Recordings manager"))
         else: app.master.title(localise("Recordings manager: ")+filename2unicode((os.sep+self.currentDir)[(os.sep+self.currentDir).rindex(os.sep)+1:]))
+        if hasattr(app,"isBigPrint") and winsound:
+            # Vista sometimes has window placement problems here
+            try: app.master.geometry("+0+0")
+            except: pass
         if not self.snack_initialized:
             if tkSnack and not tkSnack=="MicOnly":
                 tkSnack.initializeSnack(app)
-                if paranoid_file_management and tkSnack.audio.playLatency()<500: tkSnack.audio.playLatency(500) # at least 500ms latency if we're paranoid_file_management, just in case (since tkSnack.audio.elapsedTime does not account for hold-ups)
+                if paranoid_file_management:
+                    if tkSnack.audio.playLatency()<500: tkSnack.audio.playLatency(500) # at least 500ms latency if we're paranoid_file_management, just in case (since tkSnack.audio.elapsedTime does not account for hold-ups)
             self.snack_initialized = True
         if not hasattr(app,"scanrow"):
             app.scanrow = Tkinter.StringVar(app) # (keep StringVars in app not self to avoid d'tor confusion when close box pressed)
             app.scanrow.set("0")
             self.reconfigure_scrollbar()
         if tkSnack: theISM.setInputSource(MicInput())
-        self.frame=Tkinter.Frame(app.leftPanel) ; self.frame.pack()
+        self.frame=Tkinter.Frame(app.leftPanel)
+        self.frame.pack(fill=Tkinter.Y,expand=1)
 
         self.need_reRecord_enabler = 0 # no previously-existing words yet (when we get existing words we 'lock' them and have to unlock by pressing a global 'rerecord' button 1st, just in case)
 
@@ -653,6 +688,15 @@ class RecorderControls(ButtonScrollingMixin):
         else:
           r = Tkinter.Frame(self.frame)
           r.grid(row=1,sticky="e",columnspan=2)
+          if hasattr(app,"isBigPrint") and macsound:
+              # Try to make up for the fact that we can't always increase the width of the scrollbar (and the keyboard often loses focus).  Add extra up/down buttons. (TODO: does any other platform need this?)
+              r2 = Tkinter.Frame(r)
+              r2.pack({"side":"right"})
+              addButton(r2,unichr(8593),lambda e=None,app=app:app.ScrollUpHandler(),"left")
+              addButton(r2,unichr(8595),lambda e=None,app=app:app.ScrollDownHandler(),"left")
+              Tkinter.Label(r,text="    ").pack({"side":"right"}) # TODO: more flexible spacer
+              r = Tkinter.Frame(r)
+              r.pack({"side":"right"})
           Tkinter.Label(r,text=localise("Action of spacebar during recording")).pack()
           r=Tkinter.Frame(r) ; r.pack()
           for button in [
@@ -669,7 +713,7 @@ class RecorderControls(ButtonScrollingMixin):
         maxPrefix = 0 ; self.has_recordFrom_buttons = False
 
         if not self.currentDir==samplesDirectory and os.sep in self.currentDir:
-            self.addButton(curRow,0,text=localise("(Up)"),command=(lambda e=None,f=self.currentDir[:self.currentDir.rindex(os.sep)]:self.changeDir(f)))
+            self.addButton(curRow,0,text=localise("(Up)"),command=(lambda e=None,f=self.currentDir[:self.currentDir.rindex(os.sep)],self=self:self.changeDir(f)))
             curRow += 1
         l = os.listdir(self.currentDir)
         def cmpfunc(a,b): # sort alphabetically but ensure L2 comes before L1 for tab order
@@ -681,19 +725,19 @@ class RecorderControls(ButtonScrollingMixin):
             if a>b: return 1
             elif b>a: return -1
             else: return 0
-        l.sort(cmpfunc)
+        sort(l,cmpfunc)
         self.has_variants = check_has_variants(self.currentDir,l)
         allLangs = list2set([firstLanguage,secondLanguage]+possible_otherLanguages)
         hadDirectories = False
         for fname in l:
             flwr = fname.lower() ; isMeaning=0
             if firstLanguage==secondLanguage and firstLanguage+"-meaning_"+secondLanguage in fname: isMeaning,languageOverride = True, firstLanguage+"-meaning_"+secondLanguage # hack for re-loading a dir of word+meaning in same language.  TODO hope not combining -meaning_ with variants
-            elif self.has_variants and fname.find("_",fname.find("_")+1)>-1 and not fname.find("_explain_")>-1: languageOverride=fname[fname.find("_")+1:fname.find("_",fname.find("_")+1)]
+            elif self.has_variants and fname.find("_",fname.find("_")+1)>=0 and not fname.find("_explain_")>=0: languageOverride=fname[fname.find("_")+1:fname.find("_",fname.find("_")+1)]
             else: languageOverride=None
             if isDirectory(self.currentDir+os.sep+fname):
                  if not flwr in ["zips","utils","advanced utilities"]: # NOT "prompts", that can be browsed
                     newDir = self.currentDir+os.sep+fname
-                    self.addButton(curRow,0,text=filename2unicode(fname),command=(lambda e=None,f=newDir:self.changeDir(f)))
+                    self.addButton(curRow,0,text=filename2unicode(fname),command=(lambda e=None,f=newDir,self=self:self.changeDir(f)))
                     # TODO if _disabled have an Enable button ?
                     # if not have a Disable ??
                     # (NB though the above button will have a column span)
@@ -701,12 +745,14 @@ class RecorderControls(ButtonScrollingMixin):
                         focusButton(self.coords2buttons[(curRow,0)])
                         dirToHighlight = None # done
                     curRow += 1
-                    if fileExists(self.currentDir+os.sep+fname+os.sep+longDescriptionName): description=u8strip(read(self.currentDir+os.sep+fname+os.sep+longDescriptionName)).strip(wsp)
-                    elif fileExists(self.currentDir+os.sep+fname+os.sep+shortDescriptionName): description=u8strip(read(self.currentDir+os.sep+fname+os.sep+shortDescriptionName)).strip(wsp)
+                    if fileExists(self.currentDir+os.sep+fname+os.sep+longDescriptionName): description=wspstrip(u8strip(read(self.currentDir+os.sep+fname+os.sep+longDescriptionName)))
+                    elif fileExists(self.currentDir+os.sep+fname+os.sep+shortDescriptionName): description=wspstrip(u8strip(read(self.currentDir+os.sep+fname+os.sep+shortDescriptionName)))
                     else: description=None
                     if description:
-                        l = Tkinter.Label(self.grid,text="     "+description,wraplength=self.ourCanvas.winfo_screenwidth())
-                        l.grid(row=curRow,column=0,columnspan=1+3*len(self.languagesToDraw),sticky="w")
+                        try: sbarWidth = app.sbarWidth
+                        except: sbarWidth = 16 # default
+                        ll = Tkinter.Label(self.grid,text="     "+description,wraplength=self.ourCanvas.winfo_screenwidth()-sbarWidth-50) # allow for borders on Windows (TODO: is 50px always right?)
+                        ll.grid(row=curRow,column=0,columnspan=1+3*len(self.languagesToDraw),sticky="w")
                         curRow += 1
                     if not flwr=="prompts": hadDirectories = True
             elif "_" in fname and (languageOverride in allLangs or languageof(fname) in allLangs): # something_lang where lang is a recognised language (don't just take "any _" because some podcasts etc will have _ in them)
@@ -731,24 +777,24 @@ class RecorderControls(ButtonScrollingMixin):
               if iprefix>maxPrefix: maxPrefix=iprefix # max existing numerical prefix
               
               if (flwr.endswith(dotwav) or flwr.endswith(dotmp3) or flwr.endswith(dottxt)): # even if not languageOverride in self.languagesToDraw e.g. for prompts - helps setting up gradint in a language it doesn't have prompts for (creates blank rows for the prefixes that other languages use). TODO do we want to add 'and languageOverride in self.languagesToDraw' if NOT in prompts?
-                if not prefix in prefix2row:
+                if not checkIn(prefix,prefix2row):
                     self.addLabel(curRow,0,utext=filename2unicode(prefix))
                     foundTxt = {}
                     for lang in self.languagesToDraw:
-                        if realPrefix+"_"+lang+afterLang+dottxt in l: foundTxt[lang]=(self.currentDir+os.sep+realPrefix+"_"+lang+afterLang+dottxt,2+3*self.languagesToDraw.index(lang))
+                        if checkIn(realPrefix+"_"+lang+afterLang+dottxt,l): foundTxt[lang]=(self.currentDir+os.sep+realPrefix+"_"+lang+afterLang+dottxt,2+3*self.languagesToDraw.index(lang))
                     prefix2row[prefix] = curRow
                     for lang in self.languagesToDraw: # preserve tab order
                         if lang==languageOverride and not flwr.endswith(dottxt):
                             self.updateFile(fname,curRow,self.languagesToDraw.index(lang),state=1)
                             languageOverride=None # so not done again
-                        else: self.updateFile(prefix+"_"+lang+dotwav,curRow,self.languagesToDraw.index(lang),state=0,txtExists=(lang in foundTxt))
+                        else: self.updateFile(prefix+"_"+lang+dotwav,curRow,self.languagesToDraw.index(lang),state=0,txtExists=checkIn(lang,foundTxt))
                         self.gridLabel(lang,curRow)
                     for filename,col in foundTxt.values(): self.addSynthLabel(filename,curRow+1,col)
                     curRow += 2
-                if languageOverride in self.languagesToDraw and not flwr.endswith(dottxt):
+                if checkIn(languageOverride,self.languagesToDraw) and not flwr.endswith(dottxt):
                     self.updateFile(fname,prefix2row[prefix],self.languagesToDraw.index(languageOverride),state=1)
             elif (flwr.endswith(dotwav) or flwr.endswith(dotmp3)) and tkSnack and not tkSnack=="MicOnly": # no _ in it but we can still play it for splitting
-                self.addButton(curRow,0,text=(localise("Record from %s") % (filename2unicode(fname),)),command=(lambda e=None,r=curRow,f=self.currentDir+os.sep+fname:self.doRecordFrom(f,r)))
+                self.addButton(curRow,0,text=(localise("Record from %s") % (filename2unicode(fname),)),command=(lambda e=None,r=curRow,f=self.currentDir+os.sep+fname,self=self:self.doRecordFrom(f,r)))
                 self.has_recordFrom_buttons = True
                 curRow += 1
         self.addMoreRow = curRow ; self.maxPrefix = maxPrefix+1
@@ -767,14 +813,24 @@ class RecorderControls(ButtonScrollingMixin):
         addButton(r2,localise("Record from file"),self.do_recordFromFile,"left")
         if got_program("lame"): self.CompressButton = addButton(r2,localise("Compress all"),self.all2mp3_or_zip,"left") # was "Compress all recordings" but it takes too much width
         # TODO else can we see if it's possible to get the encoder on the fly, like in the main screen? (would need some restructuring)
-        elif got_program("zip") and (explorerCommand or winCEsound): self.CompressButton = addButton(r2,localise("Zip for email"),lambda *args:self.all2mp3_or_zip,"left")
+        elif got_program("zip") and (explorerCommand or winCEsound): self.CompressButton = addButton(r2,localise("Zip for email"),lambda e=None,self=self:self.all2mp3_or_zip(),"left")
         addButton(r2,localise(cond(recorderMode,"Quit","Back to main menu")),self.finished,"left")
         
         if winCEsound and not tkSnack: msg="Click on filenames at left to rename; click synthesized text to edit it"
         else: msg="Choose a word and start recording. Then press space to advance (see control at top). You can also browse and manage previous recordings. Click on filenames at left to rename (multi-line pastes are allowed); click synthesized text to edit it."
-        Tkinter.Label(self.frame,text=msg,wraplength=cond(hasattr(app,"isBigPrint") or olpc or winCEsound,self.ourCanvas.winfo_screenwidth(),min(int(self.ourCanvas.winfo_screenwidth()*.7),512))).grid(columnspan=2) # (512-pixel max. so the column isn't too wide to read on wide screens, TODO increase if the font is large)
+        if olpc or winCEsound: labelwidth = self.ourCanvas.winfo_screenwidth()
+        elif hasattr(app,"isBigPrint"): labelwidth = self.ourCanvas.winfo_screenwidth()-50 # allow for borders on Windows (TODO: is 50px always right?)
+        else: labelwidth=min(int(self.ourCanvas.winfo_screenwidth()*.7),512) # (512-pixel max. so the column isn't too wide to read on wide screens, TODO increase if the font is large)
+        Tkinter.Label(self.frame,text=msg,wraplength=labelwidth).grid(columnspan=2)
         # (Don't worry about making the text files editable - editable filenames should be enough + easier to browse the result outside Gradint; can include both languages in the filename if you like - hope the users figure this out as we don't want to make the instructions too complex)
 
+def reviseCount(num):
+    # suggested reduction for revision
+    thresholds=[1,2,knownThreshold,reallyKnownThreshold,meaningTestThreshold,randomDropThreshold,randomDropThreshold2] ; thresholds.sort() ; thresholds.reverse()
+    for i in range(len(thresholds)-1):
+        if num>thresholds[i]: return thresholds[i+1]
+    return 0
+
 def doRecWords(): # called from GUI thread
     if hasattr(app,"LessonRow"): app.thin_down_for_lesson() # else recorderMode
     app.Label.pack_forget() ; app.CancelRow.pack_forget()
@@ -784,18 +840,39 @@ def doRecWords(): # called from GUI thread
     theRecorderControls.draw()
     app.wordsExist = 1 # well not necessarily, but see comments re "Create word list"
 
-# Functions for recording on S60 phones:
+# Functions for recording on Android and S60 phones:
+
+def android_recordFile(language):
+ fname = os.getcwd()+os.sep+"newfile_"+language+dotwav
+ while True:
+  android.recorderStartMicrophone(fname) # TODO: python-for-android's MediaRecorderFacade.java startAudioRecording uses default output format and encoder, which likely means that so-called .wav file is really a .3gp file.  Have worked around in pcmlen for now, but don't know if the assumptions made there are universal, plus we don't want to name these files .wav if they're not really .wav
+  android.dialogCreateAlert("Recording",language)
+  android.dialogSetPositiveButtonText("Stop")
+  android.dialogShow() ; android.dialogGetResponse()
+  android.recorderStop()
+  android.mediaPlay("file://"+fname)
+  if not getYN("Are you happy with this?"):
+    os.remove(fname) ; continue
+  return fname
+
+def android_recordWord():
+    if not getYN("Ready to record "+secondLanguage+" word?"): return
+    def ipFunc(prompt,value=u""): return android.dialogGetInput("Gradint",prompt,value).result
+    droidOrS60RecWord(android_recordFile,ipFunc)
 def s60_recordWord():
+    def ipFunc(prompt,value=u""): return appuifw.query(prompt,"text",value)
+    droidOrS60RecWord(s60_recordFile,ipFunc)
+def droidOrS60RecWord(recFunc,inputFunc):
  if secondLanguage==firstLanguage: l1Suffix, l1Display = firstLanguage+"-meaning_"+firstLanguage, "meaning"
  else: l1Suffix, l1Display = firstLanguage, firstLanguage
  while True:
-  l2 = s60_recordFile(secondLanguage)
+  l2 = recFunc(secondLanguage)
   if not l2: return
   l1 = None
   while not l1:
-    if (not maybeCanSynth(firstLanguage)) or getYN("Record "+l1Display+" too? (else computer voice)"): l1 = s60_recordFile(l1Suffix) # (TODO what if maybeCanSynth(secondLanguage) but not first, and we want to combine 2nd-lang synth with 1st-lang recorded? low priority as if recording will prob want to rec L2)
+    if (not maybeCanSynth(firstLanguage)) or getYN("Record "+l1Display+" too? (else computer voice)"): l1 = recFunc(l1Suffix) # (TODO what if maybeCanSynth(secondLanguage) but not first, and we want to combine 2nd-lang synth with 1st-lang recorded? low priority as if recording will prob want to rec L2)
     else:
-       l1txt = appuifw.query(u""+firstLanguage+" text:","text")
+       l1txt = inputFunc(u""+firstLanguage+" text:")
        if l1txt:
           l1 = "newfile_"+firstLanguage+dottxt
           open(l1,"w").write(l1txt.encode("utf-8"))
@@ -803,21 +880,22 @@ def s60_recordWord():
        os.remove(l2) ; break
   if not l1: continue
   ls = list2set(os.listdir(samplesDirectory))
-  def inLs(prefix):
-    for ext in [dotwav,dotmp3,dottxt]:
-      for l in [firstLanguage,secondLanguage]:
-        if prefix+"_"+l+ext in ls: return 1
-  c = 1
-  while inLs("%02d" % c): c += 1
-  origPrefix = prefix = u""+("%02d" % c)
+  def inLs(prefix,ls):
+    for l in ls:
+        if l.startswith(prefix) and len(l) > len(prefix) and l[len(prefix)] not in "0123456789": return True
+  global recCount
+  try: recCount += 1
+  except: recCount = 1
+  while inLs("%02d" % recCount,ls): recCount += 1
+  origPrefix = prefix = ensure_unicode("%02d" % recCount)
   while True:
-    prefix = appuifw.query(u"Filename:","text",prefix)
+    prefix = inputFunc(u"Filename:",prefix)
     if not prefix: # pressed cancel ??
       if getYN("Discard this recording?"):
-        os.remove(l1) ; os.remove(l2) ; return
+        recCount-=1;os.remove(l1);os.remove(l2);return
       else:
         prefix = origPrefix ; continue
-    if not inLs(prefix) or getYN("File exists.  overwrite?"): break
+    if not inLs(prefix,ls) or getYN("File exists.  overwrite?"): break
   if samplesDirectory: prefix=samplesDirectory+os.sep+prefix
   os.rename(l1,prefix+l1[l1.index("_"):])
   os.rename(l2,prefix+l2[l2.index("_"):])
@@ -826,15 +904,15 @@ def s60_recordFile(language):
  fname = "newfile_"+language+dotwav
  while True:
   S=audio.Sound.open(os.getcwd()+os.sep+fname)
-  def forgetS():
+  def forgetS(fname,S):
     S.close()
     try: os.remove(fname)
     except: pass
-  if not getYN("Press OK to record "+language+" word"): return forgetS()
+  if not getYN("Press OK to record "+language+" word"): return forgetS(fname,S)
   S.record()
   ret = getYN("Press OK to stop") ; S.stop()
   if not ret:
-    forgetS() ; continue
+    forgetS(fname,S) ; continue
   S.play()
   ret = getYN("Are you happy with this?")
   S.stop() ; S.close()

gradint-build/src/sequence.py → src/sequence.py

-# This file is part of the source code of
-# gradint v0.9979 (c) 2002-2011 Silas S. Brown. GPL v3+.
+# This file is part of the source code of Gradint
+# (c) Silas S. Brown.
 #    This program is free software; you can redistribute it and/or modify
 #    it under the terms of the GNU General Public License as published by
 #    the Free Software Foundation; either version 3 of the License, or
@@ -42,7 +42,7 @@ def anticipation(promptFile,zhFile,numTimesBefore,promptsData):
     # work out number of repetitions needed.  not sure if this should be configurable somewhere.
     first_repeat_is_unessential = 0
     if not numTimesBefore: # New word.  If there are L2 variants, introduce them all if possible.
-        numVariants = min(3,len(variantFiles.get(samplesDirectory+os.sep+zhFile,[0]))) # TODO really max to 3? or 4? or .. ?
+        numVariants = min(3,len(variantFiles.get(B(samplesDirectory)+B(os.sep)+B(zhFile),[0]))) # TODO really max to 3? or 4? or .. ?
         if numVariants>1 and lessonIsTight(): numVariants = 1 # hack
         numRepeats = numVariants + cond(numVariants>=cond(availablePrompts.user_is_advanced,2,3),0,1)
     elif numTimesBefore == 1: numRepeats = 3
@@ -53,7 +53,7 @@ def anticipation(promptFile,zhFile,numTimesBefore,promptsData):
     else: numRepeats = 1
     if numRepeats==1:
       k,f = synthcache_lookup(zhFile,justQueryCache=1)
-      if f and k[0]=="_" and not textof(zhFile) in subst_synth_counters:
+      if f and B(k[:1])==B("_") and not checkIn(textof(zhFile),subst_synth_counters):
         # Hack: an experimental cache entry but only 1 repetition - what do we do?
         c=random.choice([1,2,3])
         if c==1: pass # do nothing
@@ -65,7 +65,7 @@ def anticipation(promptFile,zhFile,numTimesBefore,promptsData):
     theList = []
     if instrIsPrefix: theList = instructions
     theList.append(promptEvent)
-    if promptFile==zhFile and not promptFile in singleLinePoems:
+    if promptFile==zhFile and not checkIn(promptFile,singleLinePoems):
         # A multi-line poem with no first-language prompts, so we're using each fragment as a prompt for the next, but the first fragment is its own prompt, which means that 'prompt' is ambiguous.  Say "beginning" to disambiguate it.
         theList = theList + map(lambda x:fileToEvent(x,promptsDirectory), availablePrompts.getPromptList("begin",promptsData,languageof(zhFile)))
     if not instrIsPrefix: theList += instructions
@@ -111,10 +111,11 @@ def reverseAnticipation(promptFile,zhFile,promptsData):
     return CompositeEvent(theList)
 
 def languageof(file):
-    assert "_" in file, "no _ in %s" % (file,)
-    s=file[file.rindex("_")+1:]
-    if extsep in s: return s[:s.rindex(extsep)]
-    else: return s
+    file = B(file)
+    assert B("_") in file, "no _ in %s" % (repr(file),)
+    s=file[file.rindex(B("_"))+1:]
+    if B(extsep) in s: return S(s[:s.rindex(B(extsep))])
+    else: return S(s)
 
 def commentSequence():
     sequence = []
@@ -128,7 +129,7 @@ def anticipationSequence(promptFile,zhFile,start,to,promptsData,introList):
     # (try number from 'start' to 'to', EXCLUDING 'to')
     sequence = []
     # First one has initialGlue() whatever the value of 'start' is
-    if meaningTestThreshold and to==start+1 and start>meaningTestThreshold and random.choice([1,2])==1 and not type(promptFile)==type([]) and promptFile.find("_"+firstLanguage+extsep)>=0:
+    if meaningTestThreshold and to==start+1 and start>meaningTestThreshold and random.choice([1,2])==1 and not type(promptFile)==type([]) and B(promptFile).find(B("_"+firstLanguage+extsep))>=0:
         # *** not sure about that condition - should the random be more biased?
         # (the type() and following condition is a hack that ensures this is not used for poetry etc where there are composite prompts or the prompt is the previous line.  TODO would be better to keep track of which samples are poetic, because the above breaks down on the first line of a poem that has a translation into the first language because that looks like a normal prompt/response - but ok for now)
         firstItem = reverseAnticipation(promptFile,zhFile,promptsData)

gradint-build/src/synth.py → src/synth.py

-# This file is part of the source code of
-# gradint v0.9979 (c) 2002-2011 Silas S. Brown. GPL v3+.
+# This file is part of the source code of Gradint
+# (c) Silas S. Brown.
 #    This program is free software; you can redistribute it and/or modify
 #    it under the terms of the GNU General Public License as published by
 #    the Free Software Foundation; either version 3 of the License, or
@@ -26,15 +26,18 @@ class Synth(object):
     ################## don't have to re-implement below
     def __init__(self): self.fileCache = {}
     def __del__(self):
-        import os # as it might already have been gc'd
-        for v in self.fileCache.values():
+        try: import os # as it might already have been gc'd
+        except: pass # might still be able to use it
+        try:
+          for v in self.fileCache.values():
             try: os.remove(v)
             except: pass # someone may have removed it already, e.g. cache-synth.py's renaming
+        except: pass
         self.fileCache = {} # essential for the globalEspeakSynth hack (or get crash when try to make multiple lessons to file)
     def makefile_cached(self,lang,text):
         if type(text)==type([]): textKey=repr(text)
         else: textKey=text
-        if (lang,textKey) in self.fileCache: return self.fileCache[(lang,textKey)]
+        if checkIn((lang,textKey),self.fileCache): return self.fileCache[(lang,textKey)]
         t = self.makefile(lang,text)
         self.fileCache[(lang,textKey)] = t
         return t
@@ -67,45 +70,85 @@ class OSXSynth_Say(Synth):
     def __init__(self): Synth.__init__(self)
     def works_on_this_platform(self):
         if not (macsound and fileExists("/usr/bin/say")): return False
-        self.voices = self.scanVoices() ; return True
-    def supports_language(self,lang): return lang in self.voices
+        global osxSayVoicesScan
+        try: osxSayVoicesScan # singleton
+        except: osxSayVoicesScan = self.scanVoices()
+        self.voices = osxSayVoicesScan ; return True
+    def supports_language(self,lang): return checkIn(lang,self.voices)
     def guess_length(self,lang,text): return quickGuess(len(text),12) # TODO need a better estimate
-    def play(self,lang,text): return system("say %s\"%s\"" % (self.voices[lang],text.replace('"','')))
+    def play(self,lang,text): return system("say %s\"%s\"" % (S(self.voices[lang]),S(self.transliterate(lang,text)).replace('"','')))
     # TODO 10.7+ may also support -r rate (WPM), make that configurable in advanced.txt ?
     def makefile(self,lang,text):
         fname = os.tempnam()+extsep+"aiff"
-        system("say -o %s %s\"%s\"" % (fname,self.voices[lang],text.replace('"','')))
+        system("say %s-o %s \"%s\"" % (S(self.voices[lang]),fname,S(self.transliterate(lang,text)).replace('"','')))
         return aiff2wav(fname)
+    def transliterate(self,lang,text,forPartials=0):
+        if not self.voices[lang] in ['-v "Ting-Ting" ','-v "Tingting" ']: return text
+        # The hanzi-to-pinyin conversion in the Ting-Ting voice is not always as good as eSpeak's, but it can be controlled with pinyin.
+        ut = ensure_unicode(text)
+        if u"\u513f" in ut or u"\u5152" in ut: return text # might be erhua - better pass to the synth as-is
+        es = ESpeakSynth()
+        if not es.works_on_this_platform() or not es.supports_language('zh'): return text
+        return es.transliterate('zh',text,forPartials)
+    def can_transliterate(self,lang):
+        if not self.voices.get(lang,0) in ['-v "Ting-Ting" ','-v "Tingting" ']: return 0
+        es = ESpeakSynth()
+        return es.works_on_this_platform() and es.supports_language('zh')
     def scanVoices(self):
         d = {}
-        try: from AppKit import NSSpeechSynthesizer
-        except: return {"en":""} # no -v parameter at all
-        for vocId in NSSpeechSynthesizer.availableVoices():
-            vocAttrib = NSSpeechSynthesizer.attributesForVoice_(vocId)
-            lang = vocAttrib['VoiceLanguage']
-            if lang.startswith("en-"): lang="en" # TODO do any others need hyphen dropping?  careful / check 10.7's Cantonese etc
-            if not lang in d: d[lang]=[]
-            d[lang].append(vocAttrib['VoiceName'].encode('utf-8'))
-        found=0
-        for k,v in d.items()[:]:
+        try:
+            from AppKit import NSSpeechSynthesizer
+            voiceAttrs=[NSSpeechSynthesizer.attributesForVoice_(vocId) for vocId in NSSpeechSynthesizer.availableVoices()]
+        except: # maybe we're running under Homebrew Python instead of /usr/bin/python; in at least some recent OS X versions we should be able to get a voice list with 'say -v ?' instead (I'm not sure how far back that goes, so leaving in the above NSSpeechSynthesizer method as well)
+            voiceAttrs = []
+            for l in os.popen('say -v "?" </dev/null 2>/dev/null').readlines():
+                if not '#' in l: continue
+                name,lang=l[:l.index('#')].rsplit(None,1)
+                voiceAttrs.append({'VoiceName':name,'VoiceLanguage':lang.replace('_','-')})
+            if not voiceAttrs: return {"en":""} # maybe we're on ancient OS X: don't use a -v parameter at all
+        for vocAttrib in voiceAttrs:
+            if not checkIn('VoiceName',vocAttrib): continue
+            if not checkIn('VoiceLanguage',vocAttrib):
+                lang={"Damayanti":"id","Maged":"ar","Stine":"nb"}.get(vocAttrib['VoiceName'],None) # TODO: can sometimes use VoiceLocaleIdentifier instead, dropping the _ part (but can't even do that with Damayanti on 10.7)
+                if not lang: continue # TODO: output VoiceName in a warning?
+            else: lang = vocAttrib['VoiceLanguage']
+            if '-' in lang: lang=lang[:lang.index("-")]
+            d.setdefault(lang,[]).append(vocAttrib['VoiceName'].encode('utf-8'))
+        found=0 ; d2=d.copy()
+        class BreakOut(Exception): pass
+        # First, check for voice matches in same language beginning
+        for k,v in list(d.items()):
             if k in macVoices:
+              try:
                 for m in macVoices[k].split():
-                    if m in v:
-                        d[k] = [m] ; found=1 ; break
-            if len(d[k])>1: d[k]=[d[k][0]]
-        if d.keys()==['en'] and not found: return {"en":""} # just use the default
-        for k,v in d.items()[:]: d[k]='-v "'+v[0]+'" '
-        return d
+                  for vv in v:
+                    if B(m.lower()) == B(vv.lower()):
+                        d2[k] = [S(vv)] ; found=1 ; del macVoices[k] ; raise BreakOut()
+              except BreakOut: pass
+            if len(d2[k])>1: d2[k]=[d2[k][0]]
+        # Then check across languages (e.g. cant -> zh-...)
+        for k,v in list(macVoices.items()):
+         try:
+          for kk,vv in list(d.items()):
+            for m in v.split():
+              for vvv in vv:
+                if B(m.lower()) == B(vvv.lower()):
+                  d2[k] = [S(vvv)] ; found=1 ; raise BreakOut()
+         except BreakOut: pass
+        if list(d.keys())==['en'] and not found: return {"en":""} # just use the default
+        for k,v in list(d2.items()): d2[k]='-v "'+S(v[0])+'" '
+        return d2
 
 def aiff2wav(fname):
     if not system("sox \"%s\" \"%s\"" % (fname,fname[:-4]+"wav")):
         # good, we converted it to wav
         os.remove(fname)
         fname=fname[:-4]+"wav"
-    # else just return aiff and hope for the best (TODO won't work with cache-synth)
+    # else just return aiff and hope for the best (TODO won't work with cache-synth; TODO can get here when 'say' gave empty output, e.g. just a dot, and the returned aiff might raise IOError when constructing a SampleEvent)
     return fname
 
 class OSXSynth_OSAScript(Synth):
+    # for old Macs that don't have a "say" command
     def __init__(self): Synth.__init__(self)
     def supports_language(self,lang): return lang=="en"
     def works_on_this_platform(self): return macsound and fileExists("/usr/bin/osascript")
@@ -115,16 +158,15 @@ class OSXSynth_OSAScript(Synth):
         fname = os.tempnam()+extsep+"aiff"
         os.popen("osascript","w").write('say "%s" saving to "%s"\n' % (text,fname))
         return aiff2wav(fname)
-# TODO: if the user has installed an OS X voice that supports another language, can use say -v voicename  ( or 'using \"voicename\"' for the osascript version )  (but I have no access to a suitably-configured Mac for testing this)
 
 class OldRiscosSynth(Synth):
     def __init__(self): Synth.__init__(self)
     def supports_language(self,lang): return lang=="en"
-    def works_on_this_platform(self): return riscos_sound and not os.system("sayw .")
+    def works_on_this_platform(self): return riscos_sound and got_program("*sayw")
     def guess_length(self,lang,text): return quickGuess(len(text),12) # TODO need a better estimate
     def play(self,lang,text): return system("sayw %s" % (text,))
 
-class S60Synth(Synth):
+class S60Synth(Synth): # TODO: figure out if S60 Python can call external programs; might be able to use eSpeak http://pvagner.webranet.sk/espeak/espeak.sisx
     def __init__(self): Synth.__init__(self)
     def supports_language(self,lang): return lang=="en" # (audio.say always uses English even when other languages are installed on the device)
     def works_on_this_platform(self): return appuifw and hasattr(audio,"say")
@@ -137,7 +179,7 @@ class S60Synth(Synth):
 
 class AndroidSynth(Synth):
     def __init__(self): Synth.__init__(self)
-    def supports_language(self,lang): return lang=="en" # TODO others?
+    def supports_language(self,lang): return lang==systemVoice # TODO others? (but ttsSpeak can't set language)
     def works_on_this_platform(self): return android
     def guess_length(self,lang,text): return quickGuess(len(text),12) # TODO need a better estimate
     def play(self,lang,text): android.ttsSpeak(text)
@@ -165,7 +207,7 @@ class PttsSynth(Synth):
                 # must keep the full path even on non-cygwin because we're adding ,1 to changeToDirOf (hope we don't hit a Windows version that doesn't like this).  But we can keep relative paths if tempdir_is_curdir. (TODO if this breaks when not tempdir_is_curdir, could try copying ptts.exe to temp, but would need to delete it afterwards)
                 if cygwin or not tempdir_is_curdir: self.program='"'+os.getcwd()+cwd_addSep+i+'"'
                 else: self.program = i
-                self.offlineOnly = 'offline' in i
+                self.offlineOnly = checkIn('offline',i)
                 break
         if not self.program:
             # (in case someone's running on Windows from source)
@@ -174,15 +216,15 @@ class PttsSynth(Synth):
         else: self.lily_file = lily_file
         if fileExists(self.lily_file):
             self.old_lily_data=read(self.lily_file)
-            if "zh" in sapiVoices and sapiVoices["zh"][0].lower()=="vw lily": del sapiVoices["zh"] # because we don't want to bypass our own interface to lily if a user set that without realising it's not needed
+            if checkIn("zh",sapiVoices) and sapiVoices["zh"][0].lower()=="vw lily": del sapiVoices["zh"] # because we don't want to bypass our own interface to lily if a user set that without realising it's not needed
         else: self.lily_file = None
-    def supports_language(self,lang): return lang in sapiVoices or lang=="en" or (self.lily_file and lang=="zh")
+    def supports_language(self,lang): return checkIn(lang,sapiVoices) or lang=="en" or (self.lily_file and lang=="zh")
     # Voice list: os.popen("echo | "+self.program+" -vl").read().split("\n").  If any .lower() contain "espeak-" then after the "-" is an espeak language code see ESpeakSynth (it may not want unicode).  Other voices may also have support for specific languages - may sometimes be able to use <lang langid="locale-hex-code"/> (e.g. 809 UK, 804 Chinese (PRC) 404 Taiwan, 411 Japan) but no way for gradint to tell if successful
     def works_on_this_platform(self): return self.program
     def guess_length(self,lang,text): return quickGuess(len(text),12) # TODO need a better estimate, especially if they're going to set the speed in the control panel!
     def play(self,lang,text):
         if self.offlineOnly: return SampleEvent(self.makefile_cached(lang,text)).play()
-        if lang in sapiVoices:
+        if checkIn(lang,sapiVoices):
             d=os.getcwd()
             ret=self.sapi_unicode(sapiVoices[lang][0],ensure_unicode(text),speed=sapiSpeeds.get(lang,None))
             os.chdir(d) ; return ret
@@ -198,7 +240,7 @@ class PttsSynth(Synth):
     def sapi_unicode(self,voice,unicode_string,toFile=None,sampleRate=None,speed=None):
         # Speaks unicode_string in 'voice'.  toFile (if present) must be something that was returned by tempnam.  May change the current directory.
         if voice=="Ekho Cantonese": unicode_string = preprocess_chinese_numbers(fix_compatibility(unicode_string),isCant=2) # hack to duplicate the functionality of EkhoSynth
-        unifile=os.tempnam() ; open(unifile,"wb").write(codecs.utf_16_encode(unicode_string)[0])
+        unifile=os.tempnam() ; write(unifile,codecs.utf_16_encode(unicode_string)[0])
         if not toFile: extra=""
         else:
             extra=' -w '+changeToDirOf(toFile,1)+' -c 1'
@@ -211,7 +253,7 @@ class PttsSynth(Synth):
     def makefile(self,lang,text):
         fname = os.tempnam()+dotwav
         oldcwd=os.getcwd()
-        if lang in sapiVoices: r=self.sapi_unicode(sapiVoices[lang][0],ensure_unicode(text),fname,sapiVoices[lang][1],speed=sapiSpeeds.get(lang,None))
+        if checkIn(lang,sapiVoices): r=self.sapi_unicode(sapiVoices[lang][0],ensure_unicode(text),fname,sapiVoices[lang][1],speed=sapiSpeeds.get(lang,None))
         elif lang=="en":
             p=os.popen(self.program+' -c 1 -w '+changeToDirOf(fname,1)+self.speedParam(sapiSpeeds.get(lang,None))+toNull,"w") # (can specify mono but can't specify sample rate if it wasn't mentioned in sapiVoices - might make en synth-cache bigger than necessary but otherwise no great problem)
             p.write(text+"\n")
@@ -222,7 +264,7 @@ class PttsSynth(Synth):
         else: r=0 # shouldn't get here
         os.chdir(oldcwd)
         assert not r,"ptts.exe failed"
-        d = sapi_sox_bug_workaround(read(fname)); open(fname,"wb").write(d)
+        d = sapi_sox_bug_workaround(read(fname)); write(fname,d)
         if cygwin: os.system("chmod -x '"+fname+"'")
         return fname
     def preparePinyinPhrase(self,pinyin):
@@ -256,9 +298,9 @@ class PttsSynth(Synth):
             kVal="_%d" % count ; count += 1 # better not make kVal too long otherwise the voice can insert awkward pauses
             dicWrite.append('"%s","%s","p"\r\n' % (kVal,p2))
             rVal.append(p.replace(p2,kVal)) # (leave in full stops etc; assumes p2 is a substring of p, which is why hyphens are taken out before stripPunc)
-        open(self.lily_file,"wb").write(''.join(dicWrite))
+        write(self.lily_file,''.join(dicWrite))
         return ''.join(rVal).replace('@','') # (WITHOUT spaces, otherwise pauses far too much)
-    def restore_lily_dict(self): open(self.lily_file,"wb").write(self.old_lily_data) # done ASAP rather than on finalise, because need to make sure it happens (don't leave the system in an inconsistent state for long)
+    def restore_lily_dict(self): write(self.lily_file,self.old_lily_data) # done ASAP rather than on finalise, because need to make sure it happens (don't leave the system in an inconsistent state for long)
 def sapi_sox_bug_workaround(wavdata):
     # SAPI writes an 'EVNT' block after the sound data, and some versions of sox don't recognise this.  NB this hack is not very portable (relies on SAPI5 guaranteeing to write exactly one EVNT chunk and the bytes 'EVNT' never occur inside it, otherwise would need proper parsing)
     f=wavdata.rfind("EVNT")
@@ -268,13 +310,14 @@ py_final_letters="aeginouvrAEGINOUVR:" # (don't just pick up on tone numbers, bu
 def sort_out_pinyin_3rd_tones(pinyin):
     # Tone sandhi blocking rules: Need to stop 3rd-tones sortout at end of any 2-syllable word + "gei3 ni3" + "wo3 xiang3".
     # Also need to stop at phrase breaks and any English word (or hanzi, although may get awkward cases with 3rd-tone hanzi mixed with pinyin, but that's no big worry as lily isn't too reliable anyway and with partials it'll be transliterated)
-    segments = [] ; thisSeg = "" ; syls = 0
-    def endsWithSpecialWordpair(segLower): return segLower.endswith("gei3 ni3") or segLower.endswith("gei3 wo3") or segLower.endswith("ni3 xiang3") or segLower.endswith("wo3 xiang3")
-    for c in pinyin:
-        if ord(c)>128 or c in ".,?;" or (c==" " and syls==2) or endsWithSpecialWordpair(thisSeg.lower()):
-            segments.append(thisSeg) ; thisSeg="" ; syls = 0
-        elif c==" ": syls = 0
-        elif c in "12345": syls += 1
+    segments = [] ; thisSeg = B("") ; syls = 0 ; pinyin=B(pinyin)
+    def endsWithSpecialWordpair(segLower): return segLower.endswith(B("gei3 ni3")) or segLower.endswith(B("gei3 wo3")) or segLower.endswith(B("ni3 xiang3")) or segLower.endswith(B("wo3 xiang3"))
+    for i in xrange(len(pinyin)):
+        c = pinyin[i:i+1]
+        if ord(c)>128 or c in B(".,?;") or (c==B(" ") and syls==2) or endsWithSpecialWordpair(thisSeg.lower()):
+            segments.append(thisSeg) ; thisSeg=B("") ; syls = 0
+        elif c==B(" "): syls = 0
+        elif c in B("12345"): syls += 1
         thisSeg += c
     segments.append(thisSeg)
     # Now go for each segment
@@ -282,19 +325,19 @@ def sort_out_pinyin_3rd_tones(pinyin):
     for seg in segments:
       i=0
       while i<len(seg):
-        while i<len(seg) and seg[i] not in '12345': i+=1
-        if i<len(seg) and seg[i]=='3' and i and seg[i-1] in py_final_letters:
+        while i<len(seg) and seg[i:i+1] not in B('12345'): i+=1
+        if i<len(seg) and seg[i:i+1]==B('3') and i and seg[i-1:i] in B(py_final_letters):
             toneToChange = i ; numThirdsAfter = 0
             j = i
             while True:
                 j += 1
-                while j<len(seg) and seg[j] not in '12345': j+=1
-                if j<len(seg) and seg[j]=='3' and seg[j-1] in py_final_letters: numThirdsAfter+=1
+                while j<len(seg) and seg[j:j+1] not in B('12345'): j+=1
+                if j<len(seg) and seg[j:j+1]==B('3') and seg[j-1:j] in B(py_final_letters): numThirdsAfter+=1
                 else: break
-            if numThirdsAfter % 2: seg=seg[:toneToChange]+'2'+seg[toneToChange+1:]
+            if numThirdsAfter % 2: seg=seg[:toneToChange]+B('2')+seg[toneToChange+1:]
         i += 1
       ret.append(seg)
-    return "".join(ret)
+    return B("").join(ret)
 
 class FliteSynth(Synth):
     def __init__(self): Synth.__init__(self)
@@ -310,10 +353,10 @@ if winsound or mingw32 or cygwin: del FliteSynth.play # because on some (even hi
 
 if macsound:
     # See if we need to unpack eSpeak and/or set the variables
-    import commands
-    f=commands.getoutput("echo espeak*-OSX.zip")
+    f=getoutput("echo espeak*-OSX.zip")
     if fileExists(f): unzip_and_delete(f)
-    f=commands.getoutput("echo espeak*/speak")
+    if not os.system("espeak*/speak64 -h >/dev/null 2>/dev/null"): os.system("mv espeak*/speak64 espeak*/speak") # because some newer macOS releases want 64bit-only apps, and we won't need the dual PPC/i386 one on those
+    f=getoutput("echo espeak*/speak")
     if fileExists(f) and fileExists(f[:-5]+"espeak-data/phontab"):
       os.environ["ESPEAK_DATA_PATH"]=os.getcwd()+cwd_addSep+f[:-6]
       os.environ["PATH"]=os.getcwd()+cwd_addSep+f[:-6]+":"+os.environ["PATH"]
@@ -338,22 +381,26 @@ espeak_language_aliases = { "cant":"zhy" }
 
 class SimpleZhTransliterator(object): # if not got eSpeak on system
     def can_transliterate(self,lang): return lang=="zh"
-    def transliterate(self,lang,text,forPartials=1,from_espeak=0):
-        if lang=="zh" and not text.find("</")>-1: # (not </ - don't do this if got SSML)
-            text = preprocess_chinese_numbers(fix_compatibility(ensure_unicode(text))).encode("utf-8")
-            found=0
-            for t in text:
-                if ord(t)>=128:
-                    found=1 ; break
-            if not found and text.lower()==fix_pinyin(text,[]): return text # don't need espeak
-            elif from_espeak: return [text] # This transliterate() and ESpeakSynth's transliterate() work together - don't call espeak if there aren't any special characters (this saves launching a lot of espeak processes unnecessarily when synthing from partials), but DO proceed if fix_pinyin changes something, as in this case we need to check for embedded en words so fix_pinyin doesn't add spurious 5's, + embedded letters etc.
-            elif not found: return fix_pinyin(text,[]) # No ESpeak on system and fix_pinyin needed to do something - best we can do is hope there aren't any embedded English words (because if there are, they'll have spurious 5's added)
+    def transliterate(self,lang,text,forPartials=1,for_espeak=0):
+        text = B(text)
+        if not lang=="zh": return text
+        if text.find(B("</")) > -1: return text # (SSML)
+        text = preprocess_chinese_numbers(fix_compatibility(ensure_unicode(text))).replace(u'\u0144g','ng2').replace(u'\u0148g','ng3').replace(u'\u01f9g','ng4').encode("utf-8") # (ng2/3/4 substitution here because not all versions of eSpeak can do it)
+        found=0
+        for i in xrange(len(text)):
+            if ord(text[i:i+1])>=128:
+                found=1 ; break
+        if not found and text.lower()==fix_pinyin(text,[]): return text # don't need espeak if no non-ASCII (but DO need espeak if fix_pinyin changes something, as in this case we need to check for embedded en words so fix_pinyin doesn't add spurious 5's, + embedded letters etc)
+        elif for_espeak:
+            for s,r in [('\xc4\x80', '\xc4\x81'), ('\xc3\x81', '\xc3\xa1'), ('\xc7\x8d', '\xc7\x8e'), ('\xc3\x80', '\xc3\xa0'), ('\xc4\x92', '\xc4\x93'), ('\xc3\x89', '\xc3\xa9'), ('\xc4\x9a', '\xc4\x9b'), ('\xc3\x88', '\xc3\xa8'), ('\xc5\x8c', '\xc5\x8d'), ('\xc3\x93', '\xc3\xb3'), ('\xc7\x91', '\xc7\x92'), ('\xc3\x92', '\xc3\xb2')]: text = text.replace(LB(s),LB(r)) # work around espeak bug where capital pinyin letters with tone marks can result in bad transiterations
+            return [text] # as list so ESpeakSynth's transliterate_multiple will further process it
+        elif not found: return fix_pinyin(text,[]) # No ESpeak on system and fix_pinyin needed to do something - best we can do is hope there aren't any embedded English words (because if there are, they'll have spurious 5's added)
 simpleZhTransliterator = SimpleZhTransliterator()
 
 def shell_escape(text):
-    text = text.replace('\\','\\\\').replace('"','\\"')
-    if unix: text=text.replace("$","\\$").replace("`","\\`").replace("!","! ")
-    return '"'+text+'"'
+    text = B(text).replace(B('\\'),B('\\\\')).replace(B('"'),B('\\"'))
+    if unix: text=text.replace(B("$"),B("\\$")).replace(B("`"),B("\\`")).replace(B("!"),B("! "))
+    return B('"')+text+B('"')
 
 espeakTranslitCacheFile = "espeak-translit-cache"+extsep+"bin" # TODO to advanced.txt?
 class ESpeakSynth(Synth):
@@ -361,11 +408,14 @@ class ESpeakSynth(Synth):
         Synth.__init__(self)
         self.languages = {} ; self.program=""
         tryList = []
-        if riscos_sound and 'eSpeak$dir' in os.environ: tryList=[os.environ['eSpeak$dir']+'.espeak-dat',os.environ['eSpeak$dir']+'.espeak-data']
+        if riscos_sound:
+            if checkIn('eSpeak$Dir',os.environ):
+                os.system("RMEnsure speak 3.49  Run <eSpeak$Dir>.speakmod")
+                tryList=[os.environ['eSpeak$Dir']+'.espeak-dat',os.environ['eSpeak$Dir']+'.espeak-data']
         elif winsound or mingw32: tryList=[programFiles+"\\eSpeak\\espeak-data"]
         elif winCEsound: tryList=["\\espeak-data"] # Can't try \\Storage Card because our eSpeak compile can't cope with spaces (and quoting it does not work)
         else:
-            tryList=[os.environ.get("HOME","")+"espeak-data","/usr/share/espeak-data","/usr/local/share/espeak-data"]
+            tryList=[os.environ.get("HOME","")+"espeak-data","/usr/share/espeak-data","/usr/local/share/espeak-data","/usr/lib/x86_64-linux-gnu/espeak-data"]
             if cygwin: tryList.append(programFiles+"/eSpeak/espeak-data")
         if os.environ.get("ESPEAK_DATA_PATH",""): tryList.insert(0,os.environ["ESPEAK_DATA_PATH"]+os.sep+"espeak-data")
         langList = []
@@ -375,6 +425,13 @@ class ESpeakSynth(Synth):
                 langList = os.listdir(self.place+os.sep+"voices")
             except: self.place = None
             if langList: break
+        if unix: # espeak might know where its data is
+          if not self.place:
+            versionLine = (filter(lambda x:x.strip(),os.popen("(speak --help||espeak --help) 2>/dev/null").read().split("\n"))+[""])[0]
+            if versionLine.find("Data at:")>=0:
+              self.place = versionLine[versionLine.index("Data at:")+8:].strip()
+              try: langList = os.listdir(self.place+os.sep+"voices")
+              except: self.place = None
         for l in langList[:]:
             if l in ["default","!v","mb"]: langList.remove(l)
             elif isDirectory(self.place+os.sep+"voices"+os.sep+l):
@@ -383,40 +440,42 @@ class ESpeakSynth(Synth):
             else: self._add_lang(l,l)
         self.theProcess = None
         self.translitCache = {}
-        if pickle and fileExists(espeakTranslitCacheFile):
+        if self.place:
+          if pickle and fileExists(espeakTranslitCacheFile):
             try: placeStat,tc = pickle.Unpickler(open(espeakTranslitCacheFile,"rb")).load()
             except: placeStat,tc = (),{}
             if placeStat==tuple(os.stat(self.place)): self.translitCache = tc # otherwise regenerate it because eSpeak installation has changed (TODO if you overwrite an existing _dict file in-place, it might not update the stat() of espeak-data and the cache might not be re-generated when it should; espeak's --compile seems ok though)
-        if self.place: self.place=self.place[:self.place.rindex(os.sep)] # drop the \espeak-data, so can be used in --path=
+          self.place=self.place[:self.place.rindex(os.sep)] # drop the \espeak-data, so can be used in --path=
     def _add_lang(self,lang,fname):
+        lang,fname = S(lang),S(fname)
         if "~" in lang: return # emacs backup files
         self.languages[lang]=fname
-        for l in open(self.place+os.sep+"voices"+os.sep+fname).read(256).replace("\r","\n").split("\n"):
-            if l.startswith("language "):
-                l=l[9:].strip(wsp)
-                if not l==lang:
-                    if l in espeak_language_aliases.values(): # aliasing to an alias - update it
-                        for k,v in espeak_language_aliases.items():
-                            if v==l: espeak_language_aliases[k]=lang
-                    espeak_language_aliases[l] = lang
+        for l in open(self.place+os.sep+"voices"+os.sep+fname,"rb").read(256).replace(B("\r"),B("\n")).split(B("\n")):
+            if l.startswith(B("language ")):
+                l=bwspstrip(l[9:])
+                if not l==B(lang):
+                    Sl = S(l)
+                    if Sl in espeak_language_aliases.values(): # aliasing to an alias - update it
+                        for k,v in list(espeak_language_aliases.items()):
+                            if v==Sl: espeak_language_aliases[k]=lang
+                    espeak_language_aliases[Sl] = lang
     def describe_supported_languages(self):
         ret=[]
-        items=self.languages.items() ; items.sort()
+        items=list(self.languages.items()) ; items.sort()
         for k,v in items:
             if "-" in k and not k=="zh-yue": continue # skip variants in the report (but do recognise them)
-            o=open(self.place+os.sep+"espeak-data"+os.sep+"voices"+os.sep+v)
-            line=""
+            o=open(self.place+os.sep+"espeak-data"+os.sep+"voices"+os.sep+v,'rb')
             for t in range(10):
                 line=o.readline()
-                if line.find("name")>-1:
-                    lname = line.split()[1].replace("_test","").replace("-test","").replace("-experimental","").replace("-expertimental","") # (delete the -test etc for more screen real-estate, as this is used only for explaining what the language abbreviations mean)
+                if line.find(B("name"))>=0:
+                    lname = S(line.split()[1]).replace("_test","").replace("-test","").replace("-experimental","").replace("-expertimental","") # (delete the -test etc for more screen real-estate, as this is used only for explaining what the language abbreviations mean)
                     if not lname: continue
                     lname=lname[0].upper()+lname[1:]
                     ret.append(k+"="+lname)
                     break
         return " ".join(ret)
-    def supports_language(self,lang): return espeak_language_aliases.get(lang,lang) in self.languages
-    def not_so_good_at(self,lang): return lang not in prefer_espeak
+    def supports_language(self,lang): return checkIn(espeak_language_aliases.get(lang,lang),self.languages)
+    def not_so_good_at(self,lang): return not checkIn(lang,prefer_espeak)
     def works_on_this_platform(self):
         if len(self.languages.items())==0: return 0
         if winCEsound:
@@ -428,7 +487,7 @@ class ESpeakSynth(Synth):
         elif cygwin: toTry=[programFiles+"/eSpeak/command_line/espeak.exe"]
         else: toTry = []
         if toTry: # windows or cygwin
-            if "ESPEAK_DATA_PATH" in os.environ:
+            if checkIn("ESPEAK_DATA_PATH",os.environ):
                 toTry.insert(0,os.environ["ESPEAK_DATA_PATH"]+os.sep+"espeak.exe")
                 toTry.insert(0,os.environ["ESPEAK_DATA_PATH"]+os.sep+"command_line"+os.sep+"espeak.exe")
             for t in toTry:
@@ -439,9 +498,8 @@ class ESpeakSynth(Synth):
             return False
         else: # not windows or cygwin
             self.program="speak"
-            if riscos_sound: return True # we've already confirmed <eSpeak$dir> works in the constructor
-            import commands
-            loc=commands.getoutput("locale -a|grep -i 'utf-*8$'|head -1").strip(wsp)
+            if riscos_sound: return True # we've already confirmed <eSpeak$Dir> works in the constructor (if it doesn't, we'll just have an empty language list)
+            loc=wspstrip(getoutput("locale -a 2>/dev/null|grep -i 'utf-*8$'|head -1"))
             if loc: loc="LC_CTYPE="+loc+" " # in case espeak can't find a utf-8 locale by itself
             self.program=loc+"speak"
             if got_program("speak"): return True
@@ -449,26 +507,38 @@ class ESpeakSynth(Synth):
             self.program = loc+"espeak"
             return got_program("espeak")
     def guess_length(self,lang,text):
-        if text.find("</")>-1: # might be SSML - don't count inside <...>
+        text=B(text)
+        if text.find(B("</"))>=0: # might be SSML - don't count inside <...>
             l=inSsml=0
-            for c in text:
-                if c=="<": inSsml=1
-                elif c==">": inSsml=0
+            for i in xrange(len(text)):
+                c = text[i:i+1]
+                if c==B("<"): inSsml=1
+                elif c==B(">"): inSsml=0
                 elif not inSsml: l += 1
         else: l=len(text)
-        return quickGuess(l,12)+cond(winCEsound,1.3,0) # TODO need a better estimate.  Overhead on 195MHz Vario (baseline?) >1sec (1.3 seems just about ok)
+        latency = 0
+        if winCEsound: latency = 1.3 # TODO need a better estimate.  Overhead on 195MHz Vario (baseline?) >1sec (1.3 seems just about ok)
+        elif unix:
+          if espeak_pipe_through and not outputFile:
+            if not hasattr(self,"latency"):
+              t = time.time()
+              self.play("en","")
+              self.latency = time.time() - t # 2secs on eeePC Ubuntu 11.10, mostly AFTER the utterance
+              if self.latency > 0.5: show_info("espeak_pipe_through latency is "+str(int(self.latency*10)/10.0)+"\n",True)
+            latency = self.latency
+        return quickGuess(l,12)+latency
     def can_transliterate(self,lang): return espeak_language_aliases.get(lang,lang) in ["zh","zhy","zh-yue"] and not riscos_sound # TODO it's OK on RISC OS if the eSpeak version is recent enough to do --phonout=filename; TODO aliases for zhy (but not usually a problem as can_transliterate is called only for preference)
     def winCE_run(self,parameters,expectedOutputFile,infileToDel=None):
         self.winCE_start(parameters)
         time.sleep(0.3) # 0.2 not always long enough for transliterations (get empty output file if try to read too soon, then loop waiting for it to have contents)
         return self.winCE_wait(expectedOutputFile,infileToDel)
     def winCE_start(self,parameters):
-        s = ShellExecuteInfo(60,File=u""+self.program,Parameters=u"--path="+self.place+" "+parameters,fMask=0x40)
+        s = ShellExecuteInfo(60,File=ensure_unicode(self.program),Parameters=u"--path="+self.place+" "+parameters,fMask=0x40)
         ctypes.cdll.coredll.ShellExecuteEx(ctypes.byref(s))
         self.hProcess = s.hProcess # TODO check it's not NULL (failed to run)
     def winCE_wait(self,expectedOutputFile,infileToDel=None,needDat=1):
         # won't always work: if app and not app.Label["text"].strip(): app.setLabel("Waiting for eSpeak") # in case it doesn't produce output
-        ctypes.cdll.coredll.WaitForSingleObject(self.hProcess,4294967295) # i.e. 0xFFFFFFFF but that throws up a warning on Python 2.3
+        ctypes.cdll.coredll.WaitForSingleObject(self.hProcess,long(65535)*long(65537)) # i.e. 0xFFFFFFFF but that throws up a warning on Python 2.3; Python 2.1 won't accept 4294967295 without L but Python 3 says syntax error if L, so need to multiply
         ctypes.cdll.coredll.CloseHandle(self.hProcess)
         # In some rare circumstances, that command won't wait (e.g. process hadn't started despite the fact we delayed), so check the output files also.
         # (Leave WaitForSingleObject in as it could save some CPU cycles / potential OS crashes on some WinCE versions)
@@ -477,11 +547,11 @@ class ESpeakSynth(Synth):
             if firstIter: firstIter -= 1
             else: time.sleep(0.2),check_for_interrupts() # (latter needed in case it gets stuck)
             try:
-              if needDat: dat=read(u""+expectedOutputFile)
-              else: dat=open(u""+expectedOutputFile).read(8)
+              if needDat: dat=read(ensure_unicode(expectedOutputFile))
+              else: dat=open(ensure_unicode(expectedOutputFile)).read(8)
             except: continue # error on trying to read output
             if not dat: continue # output read as empty
-            if expectedOutputFile.endswith(dotwav) and (len(dat)<8 or dat[6:8]=="\xff\x7f"): continue # length field not yet written
+            if expectedOutputFile.endswith(dotwav) and (len(dat)<8 or dat[6:8]==LB("\xff\x7f")): continue # length field not yet written
             # (TODO how could we make sure a long transliteration has finished, if the OS lets us open the file before done and if WaitForSingleObject doesn't work?)
             if not firstIter: time.sleep(0.2) # just in case
             if infileToDel:
@@ -490,7 +560,7 @@ class ESpeakSynth(Synth):
             return dat
     def update_translit_cache(self,lang,textList): # forPartials=1 assumed
         if not lang=="zh": return # TODO if expand 'transliterate' to do other languages, make sure to update this also, and the cache format
-        if self.translitCache: textList=filter(lambda x:x not in self.translitCache, textList)
+        if self.translitCache: textList=filter(lambda x,self=self:not checkIn(x,self.translitCache), textList)
         step = 1000 # should be about right?
         for i in range(0,len(textList),step):
             tl = textList[i:i+step]
@@ -502,56 +572,86 @@ class ESpeakSynth(Synth):
             try: pickle.Pickler(open(espeakTranslitCacheFile,"wb"),-1).dump((tuple(os.stat(self.place+os.sep+"espeak-data")),self.translitCache))
             except IOError: pass # 'permission denied' is ok
     def transliterate(self,lang,text,forPartials=1):
-        if lang=="zh" and text in self.translitCache: return self.translitCache[text] # (TODO add "and forPartials"? but don't need to bother with this extra condition on slow systems)
+        if lang=="zh" and checkIn(text,self.translitCache): return self.translitCache[text] # (TODO add "and forPartials"? but don't need to bother with this extra condition on slow systems)
         return self.transliterate_multiple(lang,[text],forPartials)[0] # and don't cache it - could be experimental, and we don't want cache to grow indefinitely
+    if unix:
+        def check_dicts(self,lang,txt):
+            if not hasattr(self,"dictsChecked"): self.dictsChecked = {}
+            if checkIn(lang,self.dictsChecked) or not lang in ["zh","zhy","ru"]: return
+            if type(txt)==list: txt=B("").join(txt)
+            if re.match(B("^[ -~]*$"),txt): return # don't need to warn about dictionaries if we were given all-ASCII input (TODO: and tone marks?)
+            if filelen(self.place+os.sep+"espeak-data"+os.sep+lang+"_dict")<100000: show_warning("Warning: the eSpeak on this system has only a short dictionary for language '"+lang+"' - please install the Additional Data at espeak.sourceforge.net/data")
+            # TODO: Ubuntu 18.04's espeak-data package does NOT include the additional data; IS included by espeak-ng-data; on unix prefer the espeak-ng command and ask it for its own data path?
+            self.dictsChecked[lang]=1
+    else:
+        def check_dicts(self,lang,txt): pass
     def transliterate_multiple(self,lang,textList,forPartials=1,keepIndexList=0):
       # Call eSpeak once for multiple transliterations, for greater efficiency (especially on systems where launching a process is expensive e.g. WinCE).
       # Note: Don't make textList TOO long, because the resulting data must fit on the (RAM)disk and in memory.
       retList = [] ; write_to_espeak = [] ; indexList = []
       split_token = "^^^" # must be something not defined in the _rules files
+      textList = map(lambda x:B(x), textList)
+      self.check_dicts(lang,textList)
       for text in textList: # DON'T escape_jyutping (treat as en words)
         if lang=="zh":
          if keepIndexList: # making the cache - can we go a bit faster?
            try: t = unicode(text,"ascii") # if no utf, know is OK (but ONLY if keepIndexList, as the result is imprecise)
-           except UnicodeDecodeError: t = simpleZhTransliterator.transliterate(lang,text,from_espeak=1)
-         else: t = simpleZhTransliterator.transliterate(lang,text,from_espeak=1)
+           except UnicodeDecodeError: t = simpleZhTransliterator.transliterate(lang,text,for_espeak=1)
+         else: t = simpleZhTransliterator.transliterate(lang,text,for_espeak=1)
         else: t=[fix_compatibility(ensure_unicode(text)).encode("utf-8")]
         if t and not riscos_sound: # same TODO as above re RISC OS
             if type(t)==type([]):
                 indexList.append(len(retList))
                 retList.append(None) # result not filled in yet
-                if lang=="zh": tt=pinyin_uColon_to_V(t[0].replace("-","/")) # NB fix_compatibility has already been done (as has preprocess_chinese_numbers), by simpleZhTransliterator above
+                if lang=="zh": tt=pinyin_uColon_to_V(t[0].replace(B("-"),B("/"))) # NB fix_compatibility has already been done (as has preprocess_chinese_numbers), by simpleZhTransliterator above
                 else: tt=t[0]
-                write_to_espeak.append(fix_commas(tt).replace(split_token," "))
+                write_to_espeak.append(fix_commas(tt).replace(B(split_token),B(" ")))
                 # (replacing - with / because espeak zh voice treats / as a silent word separator but - is ignored; - is used as a word separator in MeiLing etc.  so if you want to write the hanzi for wei2ren2 but you want it to be wei4+ren2, you can hack in this way.  TODO document?)
             else: retList.append(t)
         else: retList.append(None)
       if keepIndexList: self.lastIndexList = indexList
       if not indexList: return retList
+      overruns = [] # elements that need to be merged with their following elements (duplicates allowed because indices change after each merge), used when we're transliterating very long texts (not usually as part of a lesson) because some versions of espeak truncate very long lines
+      i = 0
+      while i < len(write_to_espeak):
+          if len(write_to_espeak[i]) > 500:
+              x = write_to_espeak[i].decode('utf-8')
+              write_to_espeak[i] = x[:150].encode('utf-8')
+              write_to_espeak.insert(i+1,x[150:].encode('utf-8'))
+              overruns.append(i-len(overruns))
+          i += 1
       fname = os.tempnam()
-      open(fname,"w").write((".\n"+split_token+" ").join(write_to_espeak))
+      writeB(open(fname,"w"),(B(".\n")+B(split_token)+B(" ")).join(write_to_espeak))
       oldcwd=os.getcwd()
       if winCEsound:
           translit_out = os.tempnam()
           data=self.winCE_run(' -v%s -q -X -f %s --phonout=%s' % (espeak_language_aliases.get(lang,lang),fname,translit_out),translit_out)
           os.remove(translit_out)
-      else: data=os.popen(self.program+' -v%s -q -X -f %s%s' % (espeak_language_aliases.get(lang,lang),changeToDirOf(fname,1),cond(unix," 2>&1","")),"rb").read() # popen2 might not work, so had better do it this way:
+      elif checkIn(lang,espeak_preprocessors): data=readB(os.popen('%s < %s | %s -v%s -q -X %s' % (espeak_preprocessors[lang],changeToDirOf(fname,1),self.program,espeak_language_aliases.get(lang,lang),cond(unix," 2>&1","")),popenRB))
+      else: data=readB(os.popen('%s -v%s -q -X -f %s%s' % (self.program,espeak_language_aliases.get(lang,lang),changeToDirOf(fname,1),cond(unix," 2>&1","")),popenRB)) # popen2 might not work, so had better do it this way
       os.chdir(oldcwd) ; os.remove(fname)
-      data = data.replace("\r\n","\n").split("\nTranslate '"+split_token+"'\n")
-      if len(data)==2*len(indexList)-1:
+      data = data.replace(B("\r\n"),B("\n")).split(B("\nTranslate '")+B(split_token)+B("'\n"))
+      if len(data)==2*(len(indexList)+len(overruns))-1:
         # split points are doubled - better take every ODD item.  (NB the text in between is NOT necessarily blank - espeak can flush its sentence cache there)
         d2 = []
         for i in xrange(0,len(data),2): d2.append(data[i])
         data = d2
+      for o in overruns:
+          data[o] += data[o+1]
+          del data[o+1]
       if not len(data)==len(indexList):
           if not (winsound or macsound): show_warning("Warning: eSpeak's transliterate returned wrong number of items (%d instead of %d).  Falling back to separate runs for each item (slower)." % (len(data),len(indexList)))
           return None
       for index,dat in zip(indexList,data):
           en_words={} # any en words that espeak found embedded in the text
           r=[] ; lastWasBlank=False
-          delete_last_r_if_blank = 0
+          delete_last_r_if_blank = 0 ; appendNext = 0
           thisgroup_max_priority,thisgroup_enWord_priority = 0.5,0
-          for l in dat.strip(wsp).split("\n"):
+          for l in bwspstrip(dat).split(B("\n")):
+              if appendNext: # (see below)
+                  r.append(l[l.index(B("["))+1:l.index(B("]"))])
+                  appendNext = 0 ; continue
+              # print "Debugger:",l.strip()
               # get en_words for fix_pinyin (and for making sure we embed them in cant)
               lWords = l.split()
               if lWords: int0 = intor0(lWords[0])
@@ -559,40 +659,46 @@ class ESpeakSynth(Synth):
               if int0:
                   if int0 > thisgroup_max_priority:
                       thisgroup_max_priority = int0
-                      if lWords[-1]=="[_^_]": thisgroup_enWord_priority = int0 # so far it looks like this is going to be an English word
+                      if lWords[-1]==B("[_^_]"): thisgroup_enWord_priority = int0 # so far it looks like this is going to be an English word
               else: # a split between the groups
                   if thisgroup_enWord_priority == thisgroup_max_priority: # the choice with the highest priority was the one containing the [_^_] to put the word into English
                       en_words[r[-1]]=1
                   thisgroup_max_priority,thisgroup_enWord_priority = 0.5,0
               # end of getting en_words
-              if lang=="zh" and not lastWasBlank and r and (l.startswith("Replace") or l.startswith("Translate") or l.startswith("Found")): r[-1]+="," # (because not-blank is probably the line of phonemes)
-              elif not lang=="zh" and l.startswith("_|") and r: r[-1] += "," # works for zh-yue
+              if lang=="zh" and r and ((not lastWasBlank and (l.startswith(B("Replace")) or l.startswith(B("Translate")) or l.startswith(B("Found")))) or l.find(B("';'"))>1 or l.find(B("','"))>1): r[-1]+=B(",") # (because not-blank is probably the line of phonemes)
+              elif not lang=="zh" and l.startswith(B("_|")) and r: r[-1] += B(",") # works for zh-yue
               if delete_last_r_if_blank and not l: r=r[:-1] # "Translate" followed by blank line is probably corner-brackets or something; don't want that confusing the transliteration (especially if it's for partials)
               delete_last_r_if_blank = 0
               foundLetter=0
-              if l.startswith("Translate "):
-                  toAppend=l[l.index("'")+1:-1].replace("\xc3\xbc","v")
-                  if not (toAppend in en_words and r and toAppend==r[-1]):
+              if l.startswith(B("Translate ")):
+                  toAppend=l[l.index(B("'"))+1:-1].replace(LB("\xc3\xbc"),B("v"))
+                  if toAppend==LB("\xc2\xa0"): continue # stray no-break space (don't let this interfere with being able to do partials)
+                  if not (checkIn(toAppend,en_words) and r and toAppend==r[-1]):
                     # TODO what about partial English words? e.g. try "kao3 testing" - translate 'testing' results in a translate of 'test' also (which assumes it's already in en mode), resulting in a spurious word "test" added to the text box; not sure how to pick this up without parsing the original text and comparing with the Replace rules that occurred
                     r.append(toAppend)
-                    delete_last_r_if_blank = 1
+                    delete_last_r_if_blank = 1 # this depends on original Jonathan Duddington eSpeak and is not compatible with eSpeak NG which does not always give us blank lines.  TODO: if want eSpeak NG compatibility (which might be a good idea as mainstream GNU/Linux distributions now ship eSpeak NG and just call it "espeak"), we might want to pick up on THIS l ending with B("[_^_]") and next Translate being an identical syllable.  That might not be the only problem.
                   else: en_words[toAppend]=1
               else: # not Translate
-                  if lang=="zh" and l.startswith("Found: ") and ((l[7]==l[9]=="'" and "a"<=l[8]<="z") or (l[8]==" " and "a"<=l[7]<="z")): # an alphabetical letter - we can say this as a Chinese letter and it should be compatible with more partials-based synths.  But DON'T do this if going to give it to a unit-selection synth - 'me1' and 'ne1' don't have hanzi and some synths will have difficulty saying them.
-                      if l[8]==' ': letter=l[7]
-                      else: letter=l[8]
-                      if forPartials: r.append("a1 bo1 ci1 de1 e1 fou1 ge1 he1 yi1 ji1 ke1 le1 me1 ne1 wo1 po1 qi1 ri4 si1 te4 yu1 wei4 wu1 xi1 ye1 zi1".split()[ord(letter)-ord('a')])
-                      else: r.append(letter)
+                  if lang=="zh" and l.startswith(B("Found: ")) and ((l[7:8]==l[9:10]==B("'") and B("a")<=l[8:9]<=B("z")) or (l[8:9]==B(" ") and B("a")<=l[7:8]<=B("z"))): # an alphabetical letter - we can say this as a Chinese letter and it should be compatible with more partials-based synths.  But DON'T do this if going to give it to a unit-selection synth - 'me1' and 'ne1' don't have hanzi and some synths will have difficulty saying them.
+                      if l[8:9]==B(' '): letter=l[7:8]
+                      else: letter=l[8:9]
+                      if forPartials: r.append(B("a1 bo1 ci1 de1 e1 fou1 ge1 he1 yi1 ji1 ke1 le1 me1 ne1 wo1 po1 qi1 ri4 si1 te4 yu1 wei4 wu1 xi1 ye1 zi1").split()[ord(letter)-ord('a')])
+                      else: # a letter in something we're transliterating for a pinyin-driven unit-selection synth
+                          r.append(letter)
+                          en_words[r[-1]]=1
                       foundLetter = 1
-                  elif not lang=="zh" and l.startswith("Found: ") and (ord(l[7])>127 or (l[7]=="'" and ord(l[8])>127)): # (espeak 1.40 puts in l[7], 1.44 surrounds in quotes)
-                      r.append(l[l.index("[")+1:l.index("]")])
-              lastWasBlank=(l.startswith("Replace") or not l or foundLetter) # (take 'Replace' lines as blank, so 'Translate' doesn't add a second comma.  ditto letters thing.)
-          while r and r[-1] and r[-1][-1]==',': r[-1]=r[-1][:-1] # strip any trailing commas
-          if lang=="zh": retList[index]=fix_pinyin(" ".join(r),en_words)
-          else: retList[index]=" ".join(r)
+                  elif not lang=="zh" and l.startswith(B("Found: ")) and (ord(l[7:8])>127 or (l[7:8]==B("'") and ord(l[8:9])>127)): # (espeak 1.40 puts in l[7], 1.44 surrounds in quotes)
+                      if not B("[") in l: appendNext=1 # probably a spurious newline in the Found quote (espeak 1.48.03)
+                      else: r.append(l[l.index(B("["))+1:l.index(B("]"))])
+              lastWasBlank=(l.startswith(B("Replace")) or not l or foundLetter) # (take 'Replace' lines as blank, so 'Translate' doesn't add a second comma.  ditto letters thing.)
+          while r and r[-1] and r[-1][-1:]==B(','): r[-1]=r[-1][:-1] # strip any trailing commas
+          if lang=="zh": retList[index]=fix_pinyin(B(" ").join(r),en_words)
+          else: retList[index]=B(" ").join(r)
       return retList
-    def escape_jyutping(self,text): return re.sub(r"([abcdefghjklmnopstuwz][a-z]*[1-7])",r"[[\1]]",text) # TODO what if user already escaped it?
+    def escape_jyutping(self,text): return re.sub(B(r"([abcdefghjklmnopstuwz][a-z]*[1-7])"),B(r"[[\1]]"),B(text)) # TODO what if user already escaped it?
     def play(self,lang,text):
+        lang = S(lang)
+        self.check_dicts(lang,text)
         if espeak_language_aliases.get(lang,lang) in ["zhy","zh-yue"]: text=self.escape_jyutping(preprocess_chinese_numbers(fix_compatibility(ensure_unicode(text)),isCant=1).encode("utf-8"))
         elif lang=="zh": text=fix_commas(preprocess_chinese_numbers(fix_compatibility(ensure_unicode(text))).encode('utf-8'))
         if winCEsound: # need to play via makefile, and careful not to leave too many tempfiles or take too long
@@ -617,12 +723,13 @@ class ESpeakSynth(Synth):
             finally:
               if hasattr(self,"winCEhint"): os.unlink(self.makefile(lang,"")) # make sure to clean up on interrupt
             return ret
-        elif winsound or mingw32 or cygwin:
-            # Windows command line is not always 100% UTF-8 safe, so we'd better use a pipe.  (Command line ok on other platforms, and must do it on riscos - no pipes.)  (espeak_pipe_through only needs supporting on non-Windows - it's for aplay etc)
-            p=os.popen(self.program+cond(text.find("</")>-1," -m","")+' -v%s -a%d' % (espeak_language_aliases.get(lang,lang),100*soundVolume),"wb")
-            p.write(text+"\n") ; return p.close()
-        else: return system(self.program+cond(text.find("</")>-1," -m","")+' -v%s -a%d %s %s' % (espeak_language_aliases.get(lang,lang),100*soundVolume,shell_escape(text),espeak_pipe_through)) # (-m so accepts SSML tags)
+        elif unix or winsound or mingw32 or cygwin:
+            # Windows command line is not always 100% UTF-8 safe, so we'd better use a pipe.  Unix command line OK but some espeak versions have a length limit.  (No pipes on riscos.)
+            p=os.popen(self.program+cond(B(text).find(B("</"))>=0," -m","")+' -v%s -a%d %s' % (espeak_language_aliases.get(lang,lang),100*soundVolume,espeak_pipe_through),"w")
+            writeB(p,B(text).replace(B(". "),B(".\n"))+B("\n")) ; return p.close() # (see comment below re adding newlines)
+        else: return system(B(self.program+cond(B(text).find(B("</"))>=0," -m","")+' -v%s -a%d ' % (espeak_language_aliases.get(lang,lang),100*soundVolume))+shell_escape(text)+B(' '+espeak_pipe_through)) # (-m so accepts SSML tags)
     def makefile(self,lang,text,is_winCEhint=0):
+        self.check_dicts(lang,text)
         if espeak_language_aliases.get(lang,lang) in ["zhy","zh-yue"]: text=self.escape_jyutping(preprocess_chinese_numbers(fix_compatibility(ensure_unicode(text)),isCant=1).encode("utf-8"))
         elif lang=="zh": text=fix_commas(preprocess_chinese_numbers(fix_compatibility(ensure_unicode(text))).encode('utf-8'))
         if hasattr(self,"winCEhint"): # waiting for a previous async one that was started with is_winCEhint=1
@@ -632,7 +739,8 @@ class ESpeakSynth(Synth):
             return fname
         fname = os.tempnam()+dotwav
         oldcwd=os.getcwd()
-        sysCommand = cond(winCEsound,"",self.program)+cond(text.find("</")>-1," -m","")+' -w %s -v%s' % (changeToDirOf(fname,1),espeak_language_aliases.get(lang,lang))
+        sysCommand = cond(winCEsound,"",self.program)+cond(B(text).find(B("</"))>=0," -m","")+' -v%s -w %s%s' % (espeak_language_aliases.get(lang,lang),cond(unix,"/dev/stdout|cat>",""),changeToDirOf(fname,1))
+        # (Unix use stdout and cat because some espeak versions truncate the output file mid-discourse)
         # (eSpeak wavs are 22.05k 16-bit mono; not much point down-sampling to 16k to save 30% storage at expense of CPU)
         if winsound or mingw32: os.popen(sysCommand,"w").write(text+"\n") # must pipe the text in
         elif riscos_sound: os.system(sysCommand+' '+shell_escape(text))
@@ -643,11 +751,11 @@ class ESpeakSynth(Synth):
                 self.winCE_start(sysCommand+' -f '+fnameIn)
             else: self.winCE_run(sysCommand+' -f '+fnameIn,fname,fnameIn)
         else:
-            # we can make it asynchronously
-            if cygwin: sysCommand='echo "'+text.replace('"','\\"')+'"|'+sysCommand # (still need to pipe)
-            else: sysCommand += (' "'+text.replace('"','\\"')+'"')
+            # we can make it asynchronously (still need to pipe)
+            # (add end-of-sentence newlines due to short line buffer in some versions of espeak)
+            sysCommand=B('echo ')+shell_escape(B(text).replace(B(". "),B(".\n")))+B('|')+B(sysCommand)
             if not self.theProcess: self.theProcess = os.popen("/bin/bash","w")
-            self.theProcess.write('cd "'+os.getcwd()+'"\n'+sysCommand+"\n")
+            writeB(self.theProcess,B('cd "'+os.getcwd()+'"\n')+sysCommand+B("\n"))
             self.theProcess.flush()
         os.chdir(oldcwd)
         return fname
@@ -658,12 +766,12 @@ class ESpeakSynth(Synth):
 def fix_commas(text):
   # some versions of espeak won't honour ordinary commas in among Chinese text if the ordinary commas don't have spaces after
   # also put 2 spaces after full stop, and make sure capitalised
-  i=0
+  i=0 ; text=B(text)
   while i<len(text)-1:
-    if text[i] in '.,?;!':
-      tRest = text[i+1:].strip(wsp)
-      if tRest and (ord(tRest[0])>=128 or 'a'<=tRest[0].lower()<='z'):
-        text=text[:i+1]+cond(text[i] in ".?!","  "+tRest[0].upper()," "+tRest[0])+tRest[1:]
+    if text[i:i+1] in B('.,?;!'):
+      tRest = bwspstrip(text[i+1:])
+      if tRest and (ord(tRest[:1])>=128 or B('a')<=tRest[:1].lower()<=B('z')):
+        text=text[:i+1]+cond(text[i:i+1] in B(".?!"),B("  ")+tRest[:1].upper(),B(" ")+tRest[:1])+tRest[1:]
     i+=1
   return text
 
@@ -672,33 +780,33 @@ def fix_pinyin(pinyin,en_words):
   if en_words:
     ret=[]
     def stripPunc(w):
-      i=0 ; j=len(w) ; w=w.lower()
-      while i<len(w) and not 'a'<=w[i]<='z': i+=1
-      while j>1 and not ('a'<=w[j-1]<='z' or '1'<w[j-1]<='5'): j-=1
+      w=B(w) ; i=0 ; j=len(w) ; w=w.lower()
+      while i<len(w) and not B('a')<=w[i:i+1]<=B('z'): i+=1
+      while j>1 and not (B('a')<=w[j-1:j]<=B('z') or B('1')<w[j-1:j]<=B('5')): j-=1
       return w[i:j]
     for w in pinyin.split():
-      if stripPunc(w) in en_words: ret.append(w)
+      if checkIn(stripPunc(w),en_words): ret.append(w)
       else: ret.append(fix_pinyin(w,[]))
-    return ' '.join(ret)
+    return B(' ').join(ret)
   i=0
-  pinyin=pinyin_uColon_to_V(pinyin)+"@@@" # (includes .lower; @@@ for termination)
+  pinyin=pinyin_uColon_to_V(pinyin)+B("@@@") # (includes .lower; @@@ for termination)
   while i<len(pinyin):
-    if pinyin[i] in "12345":
+    if pinyin[i:i+1] in B("12345"):
       moveBy=0
-      if pinyin[i+1] in "iuv": moveBy=1 # these are never initial letters
-      elif pinyin[i+1]=="o" and not pinyin[i+2] in "u12345": moveBy=1 # "o" and "ou" are valid syllables, but a number before "o" is likely to be premature especially if the "o" is not itself followed by a number (or "u")
-      elif pinyin[i+1:i+3]=="ng" and not pinyin[i+3] in "aeiouv": moveBy=2 # before an -ng, but NOT before a -n g-(vowel)
-      elif pinyin[i+1] in "nr" and not pinyin[i+2] in "aeiouv" and not (pinyin[i+1]=="r" and i and not pinyin[i-1]=="e") and not pinyin[i+1:i+3]=="r5": moveBy=1 # before -n or -r (as final not as initial) (but NB -r is only on -er, otherwise it's an r5.  and if it already says r5, leave it.)
-      if moveBy: pinyin=pinyin[:i]+pinyin[i+1:i+moveBy+1]+pinyin[i]+pinyin[i+moveBy+1:]
+      if pinyin[i+1:i+2] in B("iuv"): moveBy=1 # these are never initial letters
+      elif pinyin[i+1:i+2]==B("o") and not pinyin[i+2:i+3] in B("u12345"): moveBy=1 # "o" and "ou" are valid syllables, but a number before "o" is likely to be premature especially if the "o" is not itself followed by a number (or "u")
+      elif pinyin[i+1:i+3]==B("ng") and not pinyin[i+3:i+4] in B("aeiouv"): moveBy=2 # before an -ng, but NOT before a -n g-(vowel)
+      elif pinyin[i+1:i+2] in B("nr") and not pinyin[i+2:i+3] in B("aeiouv") and not (pinyin[i+1:i+2]==B("r") and i and not pinyin[i-1:i]==B("e")) and not pinyin[i+1:i+3]==B("r5"): moveBy=1 # before -n or -r (as final not as initial) (but NB -r is only on -er, otherwise it's an r5.  and if it already says r5, leave it.)
+      if moveBy: pinyin=pinyin[:i]+pinyin[i+1:i+moveBy+1]+pinyin[i:i+1]+pinyin[i+moveBy+1:]
     i+=1
   i=0
   while i<len(pinyin): # check for missing 5's
-    if (pinyin[i] in "aeiouvr" and pinyin[i+1] not in "aeiouv12345") or (ord('a')<=ord(pinyin[i])<=ord('z') and not (ord("a")<=ord(pinyin[i+1])<=ord("z") or pinyin[i+1] in "12345")): # ("alnum and next is not alnum" is not strictly necessary, but we do need to add 5's after en-like words due to 'fix_pinyin(t)==t' being used as a do-we-need-proper-translit. condition in SimpleZhTransliterator, otherwise get problems with things like "c diao4" going to eSpeak when it could go to partials-with-letter-substitutions)
-      if pinyin[i+1:i+3]=="ng" and not pinyin[i+3] in "aeiouv":
-        if pinyin[i+3] not in "12345": pinyin=pinyin[:i+3]+"5"+pinyin[i+3:]
-      elif (pinyin[i+1]=="n" or pinyin[i:i+2]=="er") and not pinyin[i+2] in "aeiouv" and not pinyin[i]=="r":
-        if pinyin[i+2] not in "12345": pinyin=pinyin[:i+2]+"5"+pinyin[i+2:]
-      else: pinyin=pinyin[:i+1]+"5"+pinyin[i+1:]
+    if (pinyin[i:i+1] in B("aeiouvr") and pinyin[i+1:i+2] not in B("aeiouv12345")) or (ord('a')<=ord(pinyin[i:i+1])<=ord('z') and not (ord("a")<=ord(pinyin[i+1:i+2])<=ord("z") or pinyin[i+1:i+2] in B("12345"))): # ("alnum and next is not alnum" is not strictly necessary, but we do need to add 5's after en-like words due to 'fix_pinyin(t)==t' being used as a do-we-need-proper-translit. condition in SimpleZhTransliterator, otherwise get problems with things like "c diao4" going to eSpeak when it could go to partials-with-letter-substitutions)
+      if pinyin[i+1:i+3]==B("ng") and not pinyin[i+3:i+4] in B("aeiouv"):
+        if pinyin[i+3:i+4] not in B("12345"): pinyin=pinyin[:i+3]+B("5")+pinyin[i+3:]
+      elif (pinyin[i+1:i+2]==B("n") or pinyin[i:i+2]==B("er")) and not pinyin[i+2:i+3] in B("aeiouv") and not pinyin[i:i+1]==B("r"):
+        if pinyin[i+2:i+3] not in B("12345"): pinyin=pinyin[:i+2]+B("5")+pinyin[i+2:]
+      else: pinyin=pinyin[:i+1]+B("5")+pinyin[i+1:]
     i+=1
   return pinyin[:-3] # remove the @@'s
 
@@ -711,8 +819,8 @@ def remove_tone_numbers(utext): # for hanzi_and_punc to take out numbers that ca
 def preprocess_chinese_numbers(utext,isCant=0): # isCant=1 for Cantonese, 2 for hanzi (and if 1 or 2, also assumes input may be jyutping not just pinyin)
     # Hack for reading years digit by digit:
     for year in ["nian2",u"\u5e74"]: # TODO also " nian2" to catch that? what of multiple spaces?
-        while utext.find(year)>=4 and 1200 < intor0(utext[utext.find(year)-4:utext.find(year)]) < 2300: # TODO is that range right?
-            yrStart = utext.find(year)-4
+        while utext.find(year)>=4 and 1200 < intor0(utext[utext.index(year)-4:utext.index(year)]) < 2300: # TODO is that range right?
+            yrStart = utext.index(year)-4
             utext = utext[:yrStart] + " ".join(list(utext[yrStart:yrStart+4]))+" "+utext[yrStart+4:]
     # End of hack for reading years
     i=0
@@ -776,9 +884,9 @@ def espeak_stdout_works():
     except ValueError: return False
 def espeak_volume_ok():
     # if has "zh", should be recent enough
-    return "zh" in ESpeakSynth().languages
-if unix and not macsound and not (oss_sound_device=="/dev/sound/dsp" or oss_sound_device=="/dev/dsp"):
-    if playProgram=="aplay" and espeak_stdout_works(): espeak_pipe_through="--stdout|aplay -q" # e.g. NSLU2
+    return checkIn("zh",ESpeakSynth().languages)
+if wavPlayer_override or (unix and not macsound and not (oss_sound_device=="/dev/sound/dsp" or oss_sound_device=="/dev/dsp")):
+    if wavPlayer=="aplay" and espeak_stdout_works(): espeak_pipe_through="--stdout|aplay -q" # e.g. NSLU2
     else: del ESpeakSynth.play # because we have no way of sending it to the alternative device, so do it via a file
     if hasattr(FliteSynth,"play"): del FliteSynth.play
 if hasattr(ESpeakSynth,"play") and (soundVolume<0.04 or (soundVolume<0.1 and not espeak_volume_ok()) or soundVolume>2): del ESpeakSynth.play # old versions of espeak are not very good at less than 10% volume, so generate offline and use sox
@@ -824,19 +932,78 @@ class FestivalSynth(Synth):
     if oss_sound_device:
       def play(self,lang,text):
         if not self.theProcess: self.startProcess()
-        self.theProcess.write("(Parameter.set 'Audio_Command \"play --device=%s \$FILE vol %.1f\")\n(tts_text \"%s\" nil)\n" % (oss_sound_device,5*soundVolume,text)) # (tts_text text nil) can be better than (SayText text) because it splits into multiple utterances if necessary
+        self.theProcess.write("(Parameter.set 'Audio_Command \"play --device=%s \\$FILE vol %.1f\")\n(tts_text \"%s\" nil)\n" % (oss_sound_device,5*soundVolume,text)) # (tts_text text nil) can be better than (SayText text) because it splits into multiple utterances if necessary
         self.theProcess.flush()
     # else send it via a file, because we haven't got code to give it to play to the other devices directly
     def makefile(self,lang,text):
         if not self.theProcess: self.startProcess()
         fname = os.tempnam()+dotwav
-        self.theProcess.write("(Parameter.set 'Audio_Command \"sox \$FILE %s vol 5\")\n(SayText \"%s\")\n" % (fname,text))
+        self.theProcess.write("(Parameter.set 'Audio_Command \"sox \\$FILE %s vol 5\")\n(SayText \"%s\")\n" % (fname,text))
         self.theProcess.flush()
         return fname
     def finish_makefile(self):
         if self.theProcess: self.theProcess.close()
         self.theProcess = None
 
+class CoquiSynth(Synth):
+    def __init__(self):
+        Synth.__init__(self)
+        self.synths = {}
+    def works_on_this_platform(self):
+        if not unix: return 0 # I'm unable to test elsewhere
+        self.base = os.environ.get("HOME","")+"/.local/share/tts"
+        return isDirectory(self.base) # Voices require large downloads the first time they are used, so we'll use only already-downloaded voices
+    def supports_language(self,lang):
+        for a in os.listdir(self.base): # don't use any() with a generator func because we need to be Python 2.3 compatible
+            if a.startswith("tts_models--"+lang+"-"): return True # TODO: might not want to use all downloaded models, or might not want to use for all input types (e.g. zh does not support pinyin)
+    def guess_length(self,lang,text): return quickGuess(len(text),cond(lang in ["zh"],6,12)) # need better estimate
+    def makefile(self,lang,text):
+        text = ensure_unicode(text)
+        if lang=="zh": text += u"\u3002" # otherwise that model can glitch and repeat the last word of the phrase
+        if not lang in self.synths:
+            import torch;from TTS.api import TTS # shouldn't fault if models are downloaded to ~/.local/share/tts (unless uninstalled and not cleaned up...)
+            # We can assume Python 3 by this point, but must still use syntax compatible with Python 2
+            for a in sorted(os.listdir(self.base)):
+                if a.startswith("tts_models--"+lang+"-"):
+                    self.synths[lang]=TTS(a.replace("--","/")).to(cond(torch.cuda.is_available(),"cuda","cpu"))
+                    break
+        fname = os.tempnam()+dotwav
+        self.synths[lang].tts_to_file(text,file_path=fname)
+        return fname
+
+class PiperSynth(Synth):
+    def __init__(self):
+        Synth.__init__(self)
+        self.lCache = {}
+    def works_on_this_platform(self):
+        if not unix: return 0 # I can't test on other platforms
+        for self.program in ["piper/piper","./piper"]:
+            if fileExists(self.program): return True
+    def supports_language(self,lang):
+        if lang in self.lCache: return self.lCache[lang]
+        for d in [".","piper"]:
+            foundSubdir = False
+            for f in os.listdir(d):
+                if f=="piper": foundSubdir=True
+                if (f.startswith(lang+"_") or f.startswith(lang+"-")) and f.endswith('.onnx'):
+                    self.lCache[lang] = d+"/"+f
+                    return self.lCache[lang]
+            if not foundSubdir: break
+    def guess_length(self,lang,text): return quickGuess(len(text),cond(lang in ["zh"],6,12)) # need better estimate
+    def transliterate(self,lang,text,forPartials=0):
+        # Piper TTS models are controlled by eSpeak phonemes, so we should be able to get eSpeak to do this
+        es = ESpeakSynth()
+        if not es.works_on_this_platform() or not es.supports_language(lang): return text
+        return es.transliterate(lang,text,forPartials)
+    def can_transliterate(self,lang):
+        es = ESpeakSynth()
+        return es.works_on_this_platform() and es.supports_language(lang)
+    def makefile(self,lang,text):
+        fname = os.tempnam()+dotwav
+        f=os.popen(self.program+' --model "'+self.supports_language(lang)+'" --output_file "'+fname+'"',popenWB)
+        f.write(text+"\n") ; f.close()
+        return fname
+
 class GeneralSynth(Synth):
     def __init__(self): Synth.__init__(self)
     def supports_language(self,lang):
@@ -859,7 +1026,7 @@ class GeneralFileSynth(Synth):
         return 0
     def works_on_this_platform(self): return extra_speech_tofile
     def guess_length(self,lang,text):
-        if not lang in self.letters: self.letters[lang]=self.duration[lang]=0
+        if not checkIn(lang,self.letters): self.letters[lang]=self.duration[lang]=0
         if self.letters[lang]<25:
             self.letters[lang] += len(text)
             self.duration[lang] += SampleEvent(self.makefile_cached(lang,text)).exactLen
@@ -876,6 +1043,7 @@ class GeneralFileSynth(Synth):
                 return fname
 
 all_synth_classes = [GeneralSynth,GeneralFileSynth] # at the beginning so user can override
+all_synth_classes += [CoquiSynth,PiperSynth] # override espeak if present (especially PiperSynth)
 for s in synth_priorities.split(): # synth_priorities no longer in advanced.txt (see system.py above) but we can still support it
     if s.lower()=="ekho": all_synth_classes.append(EkhoSynth)
     elif s.lower()=="espeak": all_synth_classes.append(ESpeakSynth)
@@ -883,7 +1051,7 @@ for s in synth_priorities.split(): # synth_priorities no longer in advanced.txt
        all_synth_classes.append(OSXSynth_Say)
        all_synth_classes.append(OSXSynth_OSAScript) # (prefer _Say if >=10.3 because it's faster)
     elif s.lower()=="sapi": all_synth_classes.append(PttsSynth)
-all_synth_classes = all_synth_classes + [FestivalSynth,FliteSynth,OldRiscosSynth,S60Synth,AndroidSynth]
+all_synth_classes += [FestivalSynth,FliteSynth,OldRiscosSynth,S60Synth,AndroidSynth]
 prefer_espeak = prefer_espeak.split()
 
 viable_synths = []
@@ -894,9 +1062,10 @@ def setSoundCollector(sc):
     global soundCollector, viable_synths, getsynth_cache
     soundCollector,viable_synths,getsynth_cache = sc,[],{}
 def get_synth_if_possible(language,warn=1,to_transliterate=False):
-    if language in getsynth_cache and not to_transliterate: return getsynth_cache[language] # most common case (vocab.txt parse)
+    language = S(language)
+    if checkIn(language,getsynth_cache) and not to_transliterate: return getsynth_cache[language] # most common case (vocab.txt parse)
     if language==None:
-        if not None in getsynth_cache: getsynth_cache[None]=Partials_Synth()
+        if not checkIn(None,getsynth_cache): getsynth_cache[None]=Partials_Synth()
         return getsynth_cache[None]
     global viable_synths, warned_about_nosynth
     if not viable_synths:
@@ -906,18 +1075,24 @@ def get_synth_if_possible(language,warn=1,to_transliterate=False):
     if to_transliterate: # for partials: return a synth that can transliterate the language, if possible
         for synth in viable_synths:
             if synth.supports_language(language) and synth.can_transliterate(language): return synth
-        if language=="zh": return simpleZhTransliterator # in case haven't got eSpeak
-    for synth in viable_synths:
+        if language=="zh": return simpleZhTransliterator # fallback if no eSpeak
+    for synth in viable_synths: # find a good one ?
         if synth.supports_language(language) and not synth.not_so_good_at(language):
             getsynth_cache[language]=synth ; return synth
-    for synth in viable_synths:
+    for synth in viable_synths: # OK, not-so-good one ?
         if synth.supports_language(language):
-            getsynth_cache[language]=synth ; return synth
+            if warn and language not in synth_partials_voices and not language==firstLanguage and language in ["zh","cant","zhy","zh-yue"] and not language in warned_about_nosynth:
+                # Some people fail to install Yali etc because web advertising taught them to ignore sidebars :-(
+                warned_about_nosynth[language] = 1
+                show_warning("You don't have "+cond(language=="zh","Yali Cheng's Mandarin","Cameron Wong's Cantonese")+" voice installed, only a more basic robot voice. Please see the sidebar on the Gradint website for downloads.")
+                if app: waitOnMessage("In the meantime, you'll have to bear with this....")
+                getsynth_cache[language]=synth # only if warn (otherwise wait until we're called again, then warn)
+            return synth
     if (not warn) or language not in [firstLanguage,secondLanguage]+possible_otherLanguages: return None # without printing a warning
-    if not language in warned_about_nosynth:
+    if not checkIn(language,warned_about_nosynth):
         warned_about_nosynth[language] = 1
         canSay = []
-        if language in synth_partials_voices: canSay.append("recorded syllables (partials)")
+        if checkIn(language,synth_partials_voices): canSay.append("recorded syllables (partials)")
         if synthCache: canSay.append("recorded phrases (synthCache)")
         if canSay: canSay="\n  - can use only "+" and ".join(canSay)
         else: canSay="\n  (did you read ALL the comments in vocab.txt?)"
@@ -932,28 +1107,28 @@ def pinyin_uColon_to_V(pinyin):
     pinyin = pinyin.lower()
 
     pristineU = unichr(0xfc).encode('utf-8')
-    pinyin = pinyin.replace("j"+pristineU,"ju").replace("q"+pristineU,"qu").replace("x"+pristineU,"xu").replace(pristineU,"v").replace(unichr(0xea).encode('utf-8'),"e") # for pristine's pinyin
+    pinyin = B(pinyin).replace(B("j")+pristineU,B("ju")).replace(B("q")+pristineU,B("qu")).replace(B("x")+pristineU,B("xu")).replace(pristineU,B("v")).replace(unichr(0xea).encode('utf-8'),B("e")) # for pristine's pinyin
     
-    return pinyin.replace("u:","v").replace("leu","lv").replace("neu","nv")
+    return pinyin.replace(B("u:"),B("v")).replace(B("leu"),B("lv")).replace(B("neu"),B("nv"))
 
 class SynthEvent(Event):
     def __init__(self,text,synthesizer,language,is_prompt=0):
         assert text,"Trying to speak zero-length text"
-        self.text = text ; self.synthesizer = synthesizer
+        self.text = B(text) ; self.synthesizer = synthesizer
         self.modifiedText = self.text
         if language=="en":
-            self.modifiedText = self.modifiedText.replace("\xE2\x80\xA7","").replace("\xE2\x80\xB2","") # remove syllable boundaries and primes (usually just confuse speech synths)
-            if not self.text[-1] in ";.!?-" and not (';' in self.text and ';' in self.text[self.text.index(';')+1:]): self.modifiedText += ';' # prosody hack (some synths sound a bit too much like 'disjointed strict commands' without this)
+            self.modifiedText = self.modifiedText.replace(LB("\xE2\x80\xA7"),B("")).replace(LB("\xE2\x80\xB2"),B("")) # remove syllable boundaries and primes (usually just confuse speech synths)
+            if not self.text[-1:] in B(";.!?-") and not (B(';') in self.text and B(';') in self.text[self.text.index(B(';'))+1:]): self.modifiedText += B(';') # prosody hack (some synths sound a bit too much like 'disjointed strict commands' without this)
         elif language=="zh":
             # normalise pinyin
             # (note - this code is NOT used for partials synth, only for passing to espeak etc.  see elsewhere for partials synth)
             self.modifiedText = pinyin_uColon_to_V(self.modifiedText) # includes .lower()
             # and put space between every syllable of w, if it's one word only (the Lily voice seems to stand a better chance of getting it right that way, and occasionally other voices do too, e.g. "chang2yuan3" in at least some versions of eSpeak, not to mention Loquendo Lisheng
-            self.modifiedText = self.modifiedText.replace("-"," ") # for Lily, Lisheng etc.  NB replace hyphen with space not with "", otherwise can get problems with phrases like "wang4en1-fu4yi4"
-            if (not " " in self.modifiedText) and ("1" in self.modifiedText or "2" in self.modifiedText or "3" in self.modifiedText or "4" in self.modifiedText or "5" in self.modifiedText):
+            for t in ["1","2","3","4","5"]: self.modifiedText = self.modifiedText.replace(B(t+"-"),B(t+" ")) # for Lily, Lisheng etc.  NB replace hyphen with space not with "", otherwise can get problems with phrases like "wang4en1-fu4yi4".  DON'T do it except after tone marks, because for hanzi we might want to use hyphens for word-boundary disambiguation.
+            if (not B(" ") in self.modifiedText) and (B("1") in self.modifiedText or B("2") in self.modifiedText or B("3") in self.modifiedText or B("4") in self.modifiedText or B("5") in self.modifiedText):
                 self.modifiedText=fix_pinyin(self.modifiedText,[]) # better call that before doing the following (in case any digits in the wrong place)
                 for f in py_final_letters:
-                    for t in "12345": self.modifiedText=self.modifiedText.replace(f+t,f+t+" ")
+                    for t in "12345": self.modifiedText=self.modifiedText.replace(B(f+t),B(f+t+" "))
             if synthesizer.__class__ in [GeneralSynth, GeneralFileSynth]:
                 # some e.g. eSpeak use capitals to start a new sentence, so need to undo some of the .lower() that pinyin_uColon_to_V did.
                 # (ESpeakSynth already calls fix_commas in play() and makefile() so don't need to do it here.)
@@ -988,7 +1163,7 @@ class SynthEvent(Event):
             self.sound = self.synthesizer.makefile_cached(self.language,self.modifiedText)
             self.synthesizer.finish_makefile()
         if sample_table_hack:
-            if not self.sound in sample_table_hack_lengthDic: sample_table_hack_lengthDic[self.sound]=SampleEvent(self.sound).exactLen
+            if not checkIn(self.sound,sample_table_hack_lengthDic): sample_table_hack_lengthDic[self.sound]=SampleEvent(self.sound).exactLen
             soundCollector.addFile(self.sound,sample_table_hack_lengthDic[self.sound])
             open(self.sound,"wb") # i.e. truncate at 0 bytes to save space (but keep around so no name clashes)
         elif self.sound:
@@ -1048,7 +1223,7 @@ def abspath_from_start(p): # for just_synthesize to check for paths relative to
     os.chdir(d)
     return r
 
-def just_synthesize(callSanityCheck=0,lastLang_override=None):
+def just_synthesize(callGeneralCheck=0,lastLang_override=None):
     # Handle the justSynthesize setting (see advanced.txt)
     global startAnnouncement,endAnnouncement,logFile,synth_partials_cache
     synth_partials_cache = {} # to stop 'memory leak' when running from the GUI
@@ -1058,7 +1233,7 @@ def just_synthesize(callSanityCheck=0,lastLang_override=None):
     called_synth = 0
     # we re-generate the lesson on each repeat, so sporadic-synthcache stuff works
     global repeatMode ; repeatMode = 1
-    while repeatMode:
+    while repeatMode and not repeatMode=="interrupted":
       repeatMode = 0
       less = Lesson()
       lastStartTime = lastEndTime = lastWasDelay = 0
@@ -1068,20 +1243,23 @@ def just_synthesize(callSanityCheck=0,lastLang_override=None):
           ret=can_be_synthesized(fname)
           if ret: return fileToEvent(fname)
           else: show_warning("Can't say "+repr(fname)) # previous warnings should have said why (e.g. partials-only language)
-      for line in justSynthesize.split("#"):
-        line = line.strip(wsp) ; l = line.split(None,1)
-        if extsep in line and fileExists(line): event = fileToEvent(line,"")
-        elif extsep in line and fileExists(abspath_from_start(line)): event = fileToEvent(abspath_from_start(line),"")
-        elif line=='R':
+      for line in B(justSynthesize).split(B('#')):
+        line = bwspstrip(line)
+        l = line.split(None,1)
+        if B(extsep) in line and fileExists(line): event = fileToEvent(line,"")
+        elif B(extsep) in line and fileExists(abspath_from_start(line)): event = fileToEvent(abspath_from_start(line),"")
+        elif line==B('R'):
             repeatMode=1 ; continue
         elif len(l)==1:
             try: delayVal = float(l[0])
             except ValueError: delayVal = None
             if delayVal==None:
                 # no float value; assume it's a single word to synth in secondLanguage or whatever was the last language used
-                show_warning("Assuming that %s is a word to synthesize in language '%s'" % (repr(l[0]),lastLanguage))
-                if callSanityCheck and sanityCheck(l[0],lastLanguage,1): return
-                event = checkCanSynth("!synth:"+l[0]+"_"+lastLanguage)
+                r = repr(l[0])
+                if r[:1]=="b": r=r[1:]
+                show_warning("Assuming that %s is a word to synthesize in language '%s'" % (r,lastLanguage))
+                if callGeneralCheck and generalCheck(l[0],lastLanguage,1): return
+                event = checkCanSynth("!synth:"+S(l[0])+"_"+S(lastLanguage))
                 if not event: continue # couldn't synth
                 called_synth = 1
             else:
@@ -1093,17 +1271,17 @@ def just_synthesize(callSanityCheck=0,lastLang_override=None):
             lang, text = l
             if lang=="sh:": event = ShellEvent(text)
             else:
-                fname = "!synth:"+text+"_"+lang
+                fname = B("!synth:")+B(text)+B("_")+B(lang)
                 if not can_be_synthesized(fname):
                     if lang in [firstLanguage,secondLanguage]+otherLanguages:
                         show_warning("Can't say %s in %s" % (repr(text),repr(lang)))
                         lastLanguage=lang ; continue
                     # otherwise, user might have omitted lang by mistake
-                    show_warning("Assuming %s was meant to be synthesized in language '%s'" % (cond("#" in justSynthesize or len(repr(line))<10,"that '"+repr(line)+"'","this line"),lastLanguage))
-                    if callSanityCheck and sanityCheck(line,lastLanguage,1): return
-                    event = checkCanSynth("!synth:"+line+"_"+lastLanguage)
+                    show_warning("Assuming %s was meant to be synthesized in language '%s'" % (cond(B('#') in B(justSynthesize) or len(repr(line))<10,"that '"+repr(line)+"'","this line"),lastLanguage))
+                    if callGeneralCheck and generalCheck(line,lastLanguage,1): return
+                    event = checkCanSynth("!synth:"+S(line)+"_"+S(lastLanguage))
                 else:
-                    if callSanityCheck and sanityCheck(text,lang,1): return
+                    if callGeneralCheck and generalCheck(text,lang,1): return
                     event = checkCanSynth(fname)
                     lastLanguage = lang
                 if not event: continue
@@ -1117,5 +1295,6 @@ def just_synthesize(callSanityCheck=0,lastLang_override=None):
       global dbase ; dbase = None # for handleInterrupt
       less.play()
     startAnnouncement,endAnnouncement,logFile = oldStart,oldEnd,oldLogfile
+    if repeatMode=="interrupted": sys.exit(1) # better tell the calling script
     if not called_synth: return None
     return lastLanguage

gradint-build/src/system.py → src/system.py

-# This file is part of the source code of
-# gradint v0.9979 (c) 2002-2011 Silas S. Brown. GPL v3+.
+# This file is part of the source code of Gradint
+# (c) Silas S. Brown.
 #    This program is free software; you can redistribute it and/or modify
 #    it under the terms of the GNU General Public License as published by
 #    the Free Software Foundation; either version 3 of the License, or
@@ -63,24 +63,31 @@ if winsound:
         try: ctypes.cdll.commdlg
         except: WMstandard = True
 
-if macsound and __name__=="__main__": os.system("clear 1>&2") # so warnings etc start with a clear terminal (1>&2 just in case using stdout for something else)
+if macsound and __name__=="__main__": os.system("clear >&2") # so warnings etc start with a clear terminal (>&2 just in case using stdout for something else)
 if riscos_sound: sys.stderr.write("Loading Gradint...\n") # in case it takes a while
 
-try: import android
-except: android = 0
+try: import androidhelper as android
+except:
+  try: import android
+  except: android = 0
 if android:
-    android = android.Android()
-    android.makeToast("Loading Gradint")
+    try: android = android.Android()
+    except:
+        print ("\n"*50+" *** Your SL4A server has crashed ***\n  Please restart SL4A\n  (or restart your phone)\n  and try running Gradint again.\n\n\n")
+        raise SystemExit
 
-wsp = '\t\n\x0b\x0c\r ' # whitespace characters - ALWAYS use .strip(wsp) not .strip(), because someone added \xa0 (iso8859-1 no-break space) to string.whitespace on WinCE Python, and that can break processing of un-decoded UTF8 strings, e.g. a Chinese phrase ending "\xe5\x86\xa0"!  (and assign to string.whitespace does not work around this.)
+wsp = '\t\n\x0b\x0c\r ' ; bwsp=B(wsp) # whitespace characters - ALWAYS use .strip(wsp) not .strip(), because someone added \xa0 (iso8859-1 no-break space) to string.whitespace on WinCE Python, and that can break processing of un-decoded UTF8 strings, e.g. a Chinese phrase ending "\xe5\x86\xa0"!  (and assign to string.whitespace does not work around this.)
 # As .split() can't take alternative characters (and re-writing in Python is probably slow), just be careful with using it on un-decoded utf-8 stuff.  (split(None,1) is ok if 1st word won't end in an affected character)
 
-warnings_printed = [] ; app = None
+warnings_printed = [] ; app = False # False is a hack for "maybe later"
+warnings_toprint = []
 def show_warning(w):
-    if not app and not appuifw and not android:
+    if w+"\n" in warnings_printed: return
+    if not app and not app==False and not appuifw and not android:
         if winCEsound and len(w)>100: w=w[:100]+"..." # otherwise can hang winCEsound's console (e.g. a long "assuming that" message from justSynthesize)
         sys.stderr.write(w+"\n")
     warnings_printed.append(w+"\n")
+    if app==False: warnings_toprint.append(w) # may need to output them if app/appuifw/android turns out not to be created
 
 def show_info(i,always_stderr=False):
     # == sys.stderr.write(i) with no \n and no error if closed (+ redirect to app or appuifw if exists)
@@ -88,21 +95,16 @@ def show_info(i,always_stderr=False):
     if not riscos_sound and not always_stderr and hasattr(sys.stderr,"isatty") and not sys.stderr.isatty(): return # be quiet if o/p is being captured by cron etc (but isatty() might always return false on RISC OS
     if winCEsound and len(i)>101: i=i[:100]+"..."+i[-1] # otherwise can hang winCEsound's console
     if type(i)==type(u""): i=i.encode('utf-8')
-    try: sys.stderr.write(i)
+    try: writeB(sys.stderr,i)
     except IOError: pass
 
-# For pre-2.3 versions of Python (e.g. 2.2 on Symbian S60 and Mac OS 10.3):
-try: True
-except: exec("True = 1 ; False = 0")
-# TODO make sure to avoid writing "string1 in string2" without thinking - if string1 is multiple characters it won't work on pre-2.3
-# TODO check all lambda functions for Python2.2 compatibility
 # (TODO: GUI_translations, if not set in advanced.txt, won't work properly on pre-2.3 - it'll take them as Latin-1)
 # (TODO: and if it *IS* set in advanced.txt, will 2.2's exec() correctly exec a unicode string?)
 
 # Check if we're on big-endian architecture (relevant to sox etc)
 try: import struct
 except: struct=0
-if struct and struct.pack("h",1)[0]=='\x00': big_endian = 1
+if struct and B(struct.pack("h",1)[0])==B('\x00'): big_endian = 1
 else: big_endian = 0
 
 # RISC OS has a different extension separator because "." is used as a directory separator (from the original 1982 BBC Micro DFS with 1-character directories)
@@ -120,6 +122,14 @@ def list2dict(l):
   return d
 try: list2set = set
 except NameError: list2set = list2dict
+def checkIn(k,obj):
+    try: return k in obj # dict or set
+    except:
+        try: return obj.has_key(k) # Python 2.1 (may raise TypeError, AttributeError etc if try to use the "in" operator as above, but has_key rm'd from Python3)
+        except: return obj.find(k) > -1 # Python 2.1 strings
+try: object
+except:
+    class object: pass # Python 2.1
 
 # settings.txt and advanced.txt
 # (done here before the variables start to be used in
@@ -168,16 +178,23 @@ if use_unicode_filenames:
 
 def u8strip(d):
     global last_u8strip_found_BOM ; last_u8strip_found_BOM = 0
-    if d.startswith('\xef\xbb\xbf'):
+    d = B(d)
+    if d.startswith(LB('\xef\xbb\xbf')):
         last_u8strip_found_BOM = 1
         return d[3:] # ignore Notepad's UTF-8 BOM's
     else: return d
+def bwspstrip(s):
+    try: return s.strip(bwsp)
+    except: return s.strip() # Python 2.1
+def wspstrip(s):
+    try: return s.strip(wsp)
+    except: return s.strip() # Python 2.1
 GUI_translations_old = GUI_translations
 configFiles = map(lambda x:x+dottxt,["advanced","settings"]) # MUST have settings last so can have per-user override of scriptVariants
 if not hasattr(sys,"argv"): sys.argv=" " # some Symbian versions
 starting_directory = os.getcwd()
 if not fileExists(configFiles[0]):
-  if macsound and "_" in os.environ:
+  if macsound and checkIn("_",os.environ):
     s=os.environ["_"] ; s=s[:s.rfind(os.sep)]
     os.chdir(s)
     if not fileExists(configFiles[0]):
@@ -186,9 +203,14 @@ if not fileExists(configFiles[0]):
         os.chdir(s)
   if not fileExists(configFiles[0]) and sys.argv and (os.sep in sys.argv[0] or (os.sep=='\\' and '/' in sys.argv[0])):
     # try the sys.argv[0] directory, in case THAT works
-    if os.sep=="\\" and '/' in sys.argv[0] and fileExists(sys.argv[0].replace('/','\\')): sys.argv[0]=sys.argv[0].replace('/','\\') # hack for some Windows Python builds accepting / in command line but reporting os.sep as \
+    if os.sep=="\\" and '/' in sys.argv[0] and fileExists(sys.argv[0].replace('/','\\')): sys.argv[0]=sys.argv[0].replace('/','\\') # hack for some Windows Python builds accepting slash in command line but reporting os.sep as backslash
     os.chdir(starting_directory)
     os.chdir(sys.argv[0][:sys.argv[0].rfind(os.sep)])
+  if not fileExists(configFiles[0]): # argv[0] might be a symlink
+    os.chdir(starting_directory)
+    try: rp = os.path.realpath(sys.argv[0])
+    except: rp = 0 # e.g. no os.path, or no os.path.realpath
+    if rp: os.chdir(rp[:rp.rfind(os.sep)])
   if not fileExists(configFiles[0]):
     # Finally, try the module pathname, in case some other Python program has imported us without changing directory.  Apparently we need to get this from an exception.
     try: raise 0
@@ -202,7 +224,7 @@ if not fileExists(configFiles[0]):
 
 # directory should be OK by now
 if sys.platform.find("ymbian")>-1: sys.path.insert(0,os.getcwd()+os.sep+"lib")
-import time,sched,sndhdr,random,math,pprint,codecs
+import time,sched,random,math,pprint,codecs
 
 def exc_info(inGradint=True):
     import sys # in case it's been gc'd
@@ -214,29 +236,32 @@ def exc_info(inGradint=True):
     while tbObj and hasattr(tbObj,"tb_next") and tbObj.tb_next: tbObj=tbObj.tb_next
     if tbObj and hasattr(tbObj,"tb_lineno"): w += (" at line "+str(tbObj.tb_lineno))
     if inGradint:
-        if tbObj and hasattr(tbObj,"tb_frame") and hasattr(tbObj.tb_frame,"f_code") and hasattr(tbObj.tb_frame.f_code,"co_filename") and not tbObj.tb_frame.f_code.co_filename.find("gradint"+extsep+"py")>-1: w += (" in "+tbObj.tb_frame.f_code.co_filename+"\n")
-        else: w += (" in "+program_name[:program_name.index("(c)")]+"\n")
+        if tbObj and hasattr(tbObj,"tb_frame") and hasattr(tbObj.tb_frame,"f_code") and hasattr(tbObj.tb_frame.f_code,"co_filename") and not tbObj.tb_frame.f_code.co_filename.find("gradint"+extsep+"py")>=0: w += (" in "+tbObj.tb_frame.f_code.co_filename)
+        else: w += (" in "+program_name[:program_name.index("(c)")])
+        w += " on Python "+sys.version.split()[0]+"\n"
     del tbObj
     return w
 
 def read(fname): return open(fname,"rb").read()
+def write(fname,data): open(fname,"wb").write(data)
 def readSettings(f):
-   try: fdat = u8strip(read(f)).replace("\r","\n")
+   try: fdat = u8strip(read(f)).replace(B("\r"),B("\n"))
    except: return show_warning("Warning: Could not load "+f)
    try: fdat = unicode(fdat,"utf-8")
    except: return show_warning("Problem decoding utf-8 in "+f)
-   try: exec(fdat) in globals()
+   try: exec(fdat,globals())
    except: show_warning("Error in "+f+" ("+exc_info(False)+")")
 synth_priorities = "eSpeak MacOS SAPI Ekho" # old advanced.txt had this instead of prefer_espeak; we can still support it
-dir1 = list2set(dir()+["dir1","f","last_u8strip_found_BOM"])
+dir1 = list2set(dir()+["dir1","f","last_u8strip_found_BOM","__warningregistry__"])
 for f in configFiles: readSettings(f)
 for d in dir():
-  if not d in dir1 and eval(d) and not type(eval(d))==type(lambda *args:0): # (ignore unrecognised options that evaluate false - these might be an OLD unused option with a newer gradint rather than vice versa; also ignore functions as these could be used in command-line parameters)
+  if not checkIn(d,dir1) and eval(d) and not type(eval(d))==type(lambda *args:0): # (ignore unrecognised options that evaluate false - these might be an OLD unused option with a newer gradint rather than vice versa; also ignore functions as these could be used in command-line parameters)
     show_warning("Warning: Unrecognised option in config files: "+d)
 del dir1
 GUI_translations_old.update(GUI_translations) ; GUI_translations = GUI_translations_old # in case more have been added since advanced.txt last update
 
 def cond(a,b,c):
+    # Python 2.4 can inline "b if a else c" but Python 2.3 can't
     if a: return b
     else: return c
 
@@ -259,24 +284,44 @@ if paranoid_file_management:
   # For ftpfs etc.  Retry on errno 13 (permission denied), and turn append into a copy.  Otherwise occasionally get vocab.txt truncated.
   _old_open = open
   def tryIO(func):
-    for tries in range(10)+["last"]:
+    for tries in list(range(10))+["last"]:
         try: return func()
-        except IOError,err:
-            if tries=="last" or not err.errno in [5,13]: raise
+        except IOError:
+            err = sys.exc_info()[1]
+            if tries=="last" or not err.errno in [5,13,None]: raise
             time.sleep(0.5)
   def read(file): return tryIO(lambda x=file:_old_open(x,"rb").read())
-  def open(file,mode="r"):
+  def _write(fn,data):
+    tryIO(lambda x=fn,y=data:_old_open(x,"wb").write(y))
+    time.sleep(0.5)
+    if not filelen(fn)==len(data):
+      # might be a version of curlftpfs that can't shorten files - try delete and restart (although this can erase permissions info)
+      os.remove(fn)
+      tryIO(lambda x=fn,y=data:_old_open(x,"wb").write(y))
+      if not filelen(fn)==len(data): raise IOError("wrong length")
+    if not read(fn)==data: raise IOError("verification failure on "+repr(fn))
+  def write(fn,data): return tryIO(lambda x=fn,y=data:_write(x,y))
+  def open(file,mode="r",forAppend=0):
     if "a" in mode:
-        try: dat = open(file,mode.replace("a","r")).read()
-        except IOError,err:
+        try: dat = open(file,"rb").read()
+        except IOError:
+            err = sys.exc_info()[1]
             if err.errno==2: dat = "" # no such file or directory
             else: raise
+        if len(dat) < filelen(file): raise IOError("short read")
         try: os.rename(file,file+"~") # just in case!
         except: pass
-        o=open(file,mode.replace("a","w"))
+        o=open(file,"wb",1)
         o.write(dat)
         return o
-    return tryIO(lambda x=file,m=mode:_old_open(x,m))
+    r=tryIO(lambda x=file,m=mode:_old_open(x,m))
+    if "w" in mode and not forAppend and filelen(file): # it's not truncating (see _write above)
+        r.close()
+        os.unlink(file)
+        r=tryIO(lambda x=file,m=mode:_old_open(x,m))
+    return r
+
+if seedless: random.seed(0)
 
 # Different extension separators again
 if not extsep==".":
@@ -298,11 +343,11 @@ for p in [progressFile,progressFileBackup,pickledProgressFile]:
     if extsep in p[1]: p=(p[0],p[1][:p[1].rfind(extsep)]) # here rather than earlier to cover cases where extsep is in a directory name but not in the filename
     if oldDir==None: oldDir=p
     elif not oldDir==p:
-        sys.stderr.write("ERROR: progressFile, progressFileBackup and pickledProgressFile, if not None, must have same directory and major part of filename.  Gradint will not run otherwise.  This sanity-check was added in case some script sets progressFile to something special but forgets to set the others.\n")
+        sys.stderr.write("ERROR: progressFile, progressFileBackup and pickledProgressFile, if not None, must have same directory and major part of filename.  Gradint will not run otherwise.  This coherence check was added in case some script sets progressFile to something special but forgets to set the others.\n")
         sys.exit(1)
 
 # Check for RISC OS pre-1970 clock problem (actually quite likely if testing on the rpcemu emulator without setting the clock)
-if riscos_sound and hex(int(time.time())).find("0xFFFFFFFF")>-1 and not outputFile:
+if riscos_sound and hex(int(time.time())).find("0xFFFFFFFF")>=0 and not outputFile:
     sys.stderr.write("ERROR: time.time() is not usable - gradint cannot run interactively.\n")
     sys.stderr.write("This error can be caused by the RISC OS clock being at 1900 (the Unix time functions start at 1970).\nClose this task window, set the clock and try again.\n")
     sys.exit()
@@ -325,7 +370,7 @@ Tk_might_display_wrong_hanzi = wrong_hanzi_message = "" ; forceRadio=0
 if macsound:
   try: os.remove("_tkinter.so") # it might be an old patched version for the wrong OS version
   except: pass
-  def tkpatch(): # patch Mac OS Tk to the included v8.6 (as v8.4 on OS10.5 has hanzi problem and v8.5 on 10.6 has fontsize problems etc)
+  def tkpatch(): # (called only on specific older versions of Mac OS X) patch Mac OS Tk to the included v8.6 (as v8.4 on OS10.5 has hanzi problem and v8.5 on 10.6 has fontsize problems etc)
     f="/System/Library/Frameworks/Python.framework/Versions/"+sys.version[:3]+"/lib/python"+sys.version[:3]+"/lib-dynload/_tkinter.so"
     if fileExists(f): # we might be able to patch this one up
      if not isDirectory("Frameworks") and fileExists("Frameworks.tbz"): os.system("tar -jxvf Frameworks.tbz && rm Frameworks.tbz && chmod -R +w Frameworks")
@@ -340,6 +385,7 @@ if macsound:
     elif sys.version[:5] == "2.5.1": # 10.5
       if not tkpatch(): Tk_might_display_wrong_hanzi="10.5"
     elif sys.version[:5] == "2.6.1": tkpatch() # 10.6 (still has Tk8.5, hanzi ok but other problems)
+    elif sys.version[:5] == "2.7.5": tkpatch() # 10.9 (problems with "big print" button if don't do this)
   if Tk_might_display_wrong_hanzi: wrong_hanzi_message = "NB: In Mac OS "+Tk_might_display_wrong_hanzi+", Chinese\ncan display wrongly here." # so they don't panic when it does
 
 # Handle keeping progress file and temp directories etc if we're running from a live CD
@@ -355,8 +401,8 @@ def progressFileOK():
         except: return 0
 if winsound:  # will try these dirs in reverse order:
     tryList = ["C:\\TEMP\\gradint-progress.txt", "C:\\gradint-progress.txt", "C:gradint-progress.txt"]
-    if "HOMEDRIVE" in os.environ and "HOMEPATH" in os.environ: tryList.append(os.environ["HOMEDRIVE"]+os.environ["HOMEPATH"]+os.sep+"gradint-progress.txt")
-elif "HOME" in os.environ: tryList=[os.environ["HOME"]+os.sep+"gradint-progress.txt"]
+    if checkIn("HOMEDRIVE",os.environ) and checkIn("HOMEPATH",os.environ): tryList.append(os.environ["HOMEDRIVE"]+os.environ["HOMEPATH"]+os.sep+"gradint-progress.txt")
+elif checkIn("HOME",os.environ): tryList=[os.environ["HOME"]+os.sep+"gradint-progress.txt"]
 elif riscos_sound: tryList=["$.gradint-progress/txt"]
 else: tryList = []
 foundPF = okPF = 0 ; defaultProgFile = progressFile
@@ -400,29 +446,33 @@ if winsound or winCEsound or mingw32 or riscos_sound or not hasattr(os,"tempnam"
         global tempnam_no ; tempnam_no += 1
         return tmpPrefix+str(tempnam_no)
     os.tempnam = os.tmpnam = tempnam
+elif (macsound or sys.platform.lower().find("bsd")>0) and os.environ.get("TMPDIR",""): # BSD tempnam uses P_tmpdir instead, override
+    tempnam0 = os.tempnam
+    os.tempnam=lambda *args:tempnam0(os.environ["TMPDIR"])
 
 if disable_once_per_day==1:
   if once_per_day==3: sys.exit()
   else: once_per_day=0
 if once_per_day&2 and not hasattr(sys,"_gradint_innerImport"): # run every day
     currentDay = None
-    # markerFile logic to avoid 2 background copies etc (can't rely on taskkill beyond WinXP)
-    # (however this doesn't protect against uninstall + immediate reinstall)
-    markerFile,toDel="background1"+dottxt,"background2"+dottxt
-    if fileExists(markerFile): markerFile,toDel=toDel,markerFile
-    try: os.remove(toDel)
-    except OSError: pass
-    open(markerFile,"w").write("(delete this file to make the background process quit on next check)\n")
-    while fileExists(markerFile):
+    # markerFile logic to avoid 2+ background copies (can't rely on taskkill beyond WinXP)
+    myID = str(time.time())
+    try: myID += str(os.getpid())
+    except: pass
+    markerFile="background"+dottxt
+    open(markerFile,"w").write(myID)
+    def reador0(f):
+        try: return read(f)
+        except: return 0
+    while reador0(markerFile)==myID:
      if not currentDay == time.localtime()[:3]: # first run of day
       currentDay = time.localtime()[:3]
       if __name__=="__main__": # can do it by importing gradint
         sys._gradint_innerImport = 1
-        try:
-            try: reload(gradint)
-            except NameError: import gradint
-            gradint.orig_onceperday = once_per_day
-            gradint.main()
+        try: reload(gradint)
+        except NameError: import gradint
+        gradint.orig_onceperday = once_per_day
+        try: gradint.main()
         except SystemExit: pass
       elif winsound and fileExists("gradint-wrapper.exe"): # in this setup we can do it by recursively calling gradint-wrapper.exe
         s=" ".join(sys.argv[1:])
@@ -435,6 +485,8 @@ if once_per_day&2 and not hasattr(sys,"_gradint_innerImport"): # run every day
         show_warning("Not doing once_per_day&2 logic because not running as main program")
         # (DO need to be able to re-init the module - they might change advanced.txt etc)
         break
+      if len(sys.argv)>1: sys.argv.append(";")
+      sys.argv.append("disable_once_per_day=0") # don't let a disable_once_per_day=2 in argv result in repeated questioning
      time.sleep(3600) # delay 1 hour at a time (in case hibernated)
 if once_per_day&1 and fileExists(progressFile) and time.localtime(os.stat(progressFile).st_mtime)[:3]==time.localtime()[:3]: sys.exit() # already run today
 try: orig_onceperday
@@ -443,9 +495,9 @@ except: orig_onceperday=0
 if winsound:
     # check for users putting support files/folders in the desktop shortcuts folder and thinking it's the gradint folder
     # We can't do much about detecting users on non-English Windows who have heeded the warning about moving the "Desktop" folder to the real desktop but then mistook this for the gradint folder when adding flite (but hopefully they'll be using ptts/espeak anyway, and yali has an installer)
-    if "HOMEDRIVE" in os.environ and "HOMEPATH" in os.environ: dr=os.environ["HOMEDRIVE"]+os.environ["HOMEPATH"]
+    if checkIn("HOMEDRIVE",os.environ) and checkIn("HOMEPATH",os.environ): dr=os.environ["HOMEDRIVE"]+os.environ["HOMEPATH"]
     else: dr="C:\\Program Files" # as setup.bat (location for gradint on Win95 etc)
-    if "USERPROFILE" in os.environ: dr=os.environ["USERPROFILE"]
+    if checkIn("USERPROFILE",os.environ): dr=os.environ["USERPROFILE"]
     if not dr[-1]=="\\": dr += "\\"
     try: dirList = os.listdir(dr+"Desktop\\gradint\\") # trailing \ important, otherwise it can include gradint.zip etc on Desktop
     except: dirList = []
@@ -461,11 +513,42 @@ elif macsound:
        os.system('open ../Gradint.app')
        sys.exit(0)
     elif fileExists_stat("../Gradint 2.app/deleteme"):
-       import thread ; thread.start_new_thread(lambda *x:(time.sleep(2),os.system('rm -rf "../Gradint 2.app"')),())
+       try: import thread
+       except ImportError: import _thread as thread
+       thread.start_new_thread(lambda *x:(time.sleep(2),os.system('rm -rf "../Gradint 2.app"')),())
 
 def got_program(prog):
-    # Test to see if the program 'prog' is on the system, as portable as possible.  NB some Unix 'which' output an error to stdout instead of stderr, so check the result exists.
-    return (winsound and fileExists(prog+".exe")) or (unix and fileExists_stat(os.popen("which "+prog+" 2>/dev/null").read().strip(wsp)))
+    if winsound:
+        if fileExists(prog+".exe"): return prog+".exe"
+    elif riscos_sound:
+        if prog[:1]=="*": # module
+            os.system("help "+prog[1:]+" { > _tstCmd_ }")
+            got = open("_tstCmd_").read().find(prog[1:].upper())>-1
+            os.unlink("_tstCmd_") ; return got
+        return checkIn("Alias$"+prog,os.environ) # works in Python 3.8 but not 2.7 (Alias$ vars hidden)
+    elif unix:
+        try:
+            try: from shutil import which as find_executable # PEP 632
+            except: from distutils.spawn import find_executable
+            if (":"+os.environ.get("PATH","")).find(":.")>-1:
+                prog = find_executable(prog)
+            else: # at least some distutils assume that "." is in the PATH even when it isn't, so do it ourselves without checking "."
+                oldCwd = os.getcwd()
+                pList = os.environ.get("PATH","").split(':')
+                if pList:
+                  done=0
+                  for p in pList:
+                    try: os.chdir(p)
+                    except: continue
+                    done=1 ; break
+                  if done:
+                    prog = find_executable(prog)
+                    os.chdir(oldCwd)
+        except ImportError:
+            # fall back to running 'which' in a shell (probably slower if got_program is called repeatedly)
+            prog = wspstrip(os.popen("which "+prog+" 2>/dev/null").read())
+            if not fileExists_stat(prog): prog=None # some Unix 'which' output an error to stdout instead of stderr, so check the result exists
+        return prog
 
 def win2cygwin(path): # convert Windows path to Cygwin path
     if path[1]==":": return "/cygdrive/"+path[0].lower()+path[2:].replace("\\","/")
@@ -497,17 +580,11 @@ def check_for_interrupts(): # used on platforms where thread.interrupt_main won'
         raise KeyboardInterrupt
 
 # If forking, need to do so BEFORE importing any Tk module (we can't even verify Tk exists 1st)
-if outputFile or justSynthesize or appuifw or not (winsound or winCEsound or mingw32 or macsound or riscos_sound or cygwin or "DISPLAY" in os.environ): useTK = 0
-if useTK and runInBackground and not (winsound or mingw32) and hasattr(os,"fork") and not "gradint_no_fork" in os.environ:
-    import fcntl, termios
+if outputFile or justSynthesize or appuifw or not (winsound or winCEsound or mingw32 or macsound or riscos_sound or cygwin or checkIn("DISPLAY",os.environ)): useTK = 0
+if useTK and runInBackground and not (winsound or mingw32) and hasattr(os,"fork") and not checkIn("gradint_no_fork",os.environ):
     if os.fork(): sys.exit()
-    os.setpgid(0,0)
+    os.setsid()
     if os.fork(): sys.exit()
-    try: tty = os.open("/dev/tty", os.O_RDWR)
-    except: tty = None
-    if not tty==None:
-        fcntl.ioctl(tty, termios.TIOCNOTTY, 0)
-        os.close(tty)
     devnull = os.open("/dev/null", os.O_RDWR)
     for fd in range(3): os.dup2(devnull,fd)
 else: runInBackground = 0
@@ -531,8 +608,8 @@ try:
   import locale
   locale.setlocale(locale.LC_ALL, 'C')
 except: pass
-if not '\xc4'.lower()=='\xc4': # buggy setlocale (e.g. S60) can create portability issues with progress files
-  lTrans="".join([chr(c) for c in range(ord('A'))]+[chr(c) for c in range(ord('a'),ord('z')+1)]+[chr(c) for c in range(ord('Z')+1,256)])
+if not LB('\xc4').lower()==LB('\xc4'): # buggy setlocale (e.g. S60) can create portability issues with progress files
+  lTrans=B("").join([chr(c) for c in range(ord('A'))]+[chr(c) for c in range(ord('a'),ord('z')+1)]+[chr(c) for c in range(ord('Z')+1,256)])
   def lower(s): return s.translate(lTrans) # (may crash if Unicode)
 else:
   def lower(s): return s.lower()

gradint-build/src/top.py → src/top.py

 #!/usr/bin/env python
 # -*- coding: utf-8 -*-
+#   (Python 2 or Python 3, but more fully tested on 2)
 
-program_name = "gradint v0.9979 (c) 2002-2011 Silas S. Brown. GPL v3+."
+program_name = "gradint v3.095 (c) 2002-25 Silas S. Brown. GPL v3+."
 
 #    This program is free software; you can redistribute it and/or modify
 #    it under the terms of the GNU General Public License as published by
@@ -21,8 +22,48 @@ appTitle = "Language lesson"
 
 import sys,os
 
-if not sys.version_info[0]==2: # oh panic, someone's probably trying to run us on Py3k
-    sys.stderr.write("Sorry, Gradint cannot run on Python "+repr(sys.version_info[0])+"\nPlease install a 2.x version of Python (must be 2.2+).\n")
-    sys.exit(1)
+if sys.version_info[0]>2:
+    _map,_filter = map,filter
+    def map(*args): return list(_map(*args))
+    def filter(*args): return list(_filter(*args))
+    from functools import cmp_to_key
+    def sort(l,c): l.sort(key=cmp_to_key(c))
+    raw_input,unichr,xrange,long = input,chr,range,int
+    def chr(x): return unichr(x).encode('latin1')
+    from subprocess import getoutput
+    popenRB,popenWB = "r","w"
+    def unicode(b,enc):
+        if type(b)==str: return b
+        return b.decode(enc)
+else: # Python 2
+    def sort(l,c): l.sort(c)
+    popenRB,popenWB = "rb","wb"
+    bytes = str
+    try: from commands import getoutput
+    except ImportError: pass
+    # For pre-2.3 versions of Python (e.g. 2.2 on Symbian S60 and Mac OS 10.3):
+    try: True
+    except: exec("True = 1 ; False = 0")
+def readB(f,m=None):
+    if hasattr(f,"buffer"): f0,f=f,f.buffer # Python 3 non-"b" file
+    if m: return f.read(m)
+    else: return f.read() # no "None" in Python 2
+def writeB(f,b):
+    if hasattr(f,"buffer"): f0,f=f,f.buffer # Python 3 non-"b" file
+    f.write(b)
+def B(x):
+    if type(x)==bytes: return x
+    try: return x.encode('utf-8')
+    except: return x # maybe not a string
+def LB(x):
+    if type(x)==bytes: return x
+    try: return x.encode('latin1')
+    except: return x
+def S(x):
+    if type(x)==bytes and not bytes==str: return x.decode('utf-8')
+    return x
+def S2(s):
+    try: return S(s)
+    except: return s # coding errors OK in unavail, leave as byte-string
 
 # --------------------------------------------------------

gradint-build/src/users.py → src/users.py

-# This file is part of the source code of
-# gradint v0.9979 (c) 2002-2011 Silas S. Brown. GPL v3+.
+# This file is part of the source code of Gradint
+# (c) Silas S. Brown.
 #    This program is free software; you can redistribute it and/or modify
 #    it under the terms of the GNU General Public License as published by
 #    the Free Software Foundation; either version 3 of the License, or
@@ -43,21 +43,21 @@ def setup_samplesDir_ifNec(d=0): # if the user doesn't have a samples directory,
   if not d: d=samplesDirectory
   if not isDirectory(d):
     os.mkdir(d)
-    if fileExists(user0[0]+os.sep+"README"+dottxt): open(d+os.sep+"README"+dottxt,'wb').write(read(user0[0]+os.sep+"README"+dottxt))
+    if fileExists(user0[0]+os.sep+"README"+dottxt): write(d+os.sep+"README"+dottxt,read(user0[0]+os.sep+"README"+dottxt))
 
 def get_userNames(): # list of unicode user names or []
   ret=[]
   u=userNameFile ; c=0
   while fileExists(u):
-    ret.append(unicode(u8strip(read(u)).strip(wsp),'utf-8'))
+    ret.append(unicode(bwspstrip(u8strip(read(u))),'utf-8'))
     c += 1 ; u=addUserToFname(userNameFile,c)
   global lastUserNames ; lastUserNames = ret
   return ret
 
-def set_userName(N,unicodeName): open(addUserToFname(userNameFile,N),"w").write(unicodeName.encode("utf-8")+"\n") # implicitly adds if N=num+1
+def set_userName(N,unicodeName): writeB(open(addUserToFname(userNameFile,N),"w"),unicodeName.encode("utf-8")+B("\n")) # implicitly adds if N=num+1
 
 def wrapped_set_userName(N,unicodeName):
-  if unicodeName.strip(wsp): set_userName(N,unicodeName)
+  if wspstrip(unicodeName): set_userName(N,unicodeName)
   else: app.todo.alert="You need to type the person's name in the box before you press "+localise("Add new name") # don't waitOnMessage because we're in the GUI thread
 
 GUI_usersRow = lastUserNames = None
@@ -91,7 +91,7 @@ def updateUserRow(fromMainMenu=0):
         userBSM = ButtonScrollingMixin() ; userBSM.ourCanvas = c
     else: userBSM = None
     for i in range(len(names)):
-      if names[i].strip(wsp):
+      if wspstrip(names[i]):
         r=Tkinter.Radiobutton(row, text=names[i], variable=app.userNo, value=str(i), takefocus=0)
         r.grid(row=i+1,column=0,sticky="w")
         r["command"]=cmd=lambda e=None,i=i: select_userNumber(i)
@@ -107,41 +107,42 @@ def updateUserRow(fromMainMenu=0):
         r=Tkinter.Frame(row) ; r.grid(row=i+1,column=0,columnspan=4)
         text,entry = addTextBox(r)
         if not fromMainMenu: entry.focus() # because user has just pressed the "add other students" button, or has just added a name and may want to add another
-        l=lambda *args:(wrapped_set_userName(i,asUnicode(text.get())),updateUserRow())
+        l=lambda e=None,wrapped_set_userName=wrapped_set_userName,i=i,text=text:(wrapped_set_userName(i,asUnicode(text.get())),updateUserRow())
         addButton(r,localise("Add new name"),l)
         entry.bind('<Return>',l)
         if not i: Tkinter.Label(row,text="The first name should be that of the\nEXISTING user (i.e. YOUR name).").grid(row=i+2,column=0,columnspan=4)
       if userBSM: userBSM.bindFocusIn(r) # for shift-tab from the bottom
       if hasattr(row,"widgetsToDel"): row.widgetsToDel.append(r)
       if not names[i]: break
-    if userBSM: c.after(cond(winCEsound,1500,300),lambda *args:c.config(scrollregion=c.bbox(Tkinter.ALL),width=c.bbox(Tkinter.ALL)[2],height=min(c["height"],c.winfo_screenheight()/2,c.bbox(Tkinter.ALL)[3]))) # hacky (would be better if it could auto shrink on resize)
-  else: row.widgetsToDel.append(addButton(row,localise("Family mode (multiple user)"),lambda *args:(set_userName(0,""),updateUserRow())))
+    if userBSM: c.after(cond(winCEsound,1500,300),lambda e=None,c=c:c.config(scrollregion=c.bbox(Tkinter.ALL),width=c.bbox(Tkinter.ALL)[2],height=min(c["height"],c.winfo_screenheight()/2,c.bbox(Tkinter.ALL)[3]))) # hacky (would be better if it could auto shrink on resize)
+  else: row.widgetsToDel.append(addButton(row,localise("Family mode (multiple user)"),lambda e=None:(set_userName(0,""),updateUserRow())))
 
 def renameUser(i,radioButton,parent,cancel=0):
-    if hasattr(radioButton,"in_renaming"):
+    if hasattr(radioButton,"in_renaming"): # finish the rename
         del radioButton.in_renaming
         n=asUnicode(radioButton.renameText.get())
         if cancel: pass
-        elif not n.strip(wsp) and len(lastUserNames)>1: tkMessageBox.showinfo(app.master.title(),"You can't have blank user names unless there is only one user.  Keeping the original name instead.")
+        elif not wspstrip(n) and (len(lastUserNames)>1 and not (len(lastUserNames)==2 and not lastUserNames[-1])): tkMessageBox.showinfo(app.master.title(),"You can't have blank user names unless there is only one user.  Keeping the original name instead.")
         else:
             set_userName(i,n)
+            lastUserNames[i] = n
             radioButton["text"]=n
         radioButton.renameEntry.grid_forget()
         radioButton.grid(row=i+1,column=0,sticky="w")
-    else:
+    else: # start the rename
         radioButton.in_renaming = 1
         radioButton.grid_forget()
         radioButton.renameText,radioButton.renameEntry = addTextBox(parent,"nopack")
         radioButton.renameEntry.grid(row=i+1,column=0)
         radioButton.renameText.set(lastUserNames[i])
         radioButton.renameEntry.focus()
-        radioButton.after(10,lambda *args:radioButton.renameEntry.event_generate('<End>'))
-        radioButton.renameEntry.bind('<Return>',lambda *args:renameUser(i,radioButton,parent))
-        radioButton.renameEntry.bind('<Escape>',lambda *args:renameUser(i,radioButton,parent,cancel=1))
+        radioButton.after(10,lambda e=None,radioButton=radioButton:radioButton.renameEntry.event_generate('<End>'))
+        radioButton.renameEntry.bind('<Return>',lambda e=None,radioButton=radioButton,i=i,parent=parent:renameUser(i,radioButton,parent))
+        radioButton.renameEntry.bind('<Escape>',lambda e=None,i=i,radioButton=radioButton,parent=parent:renameUser(i,radioButton,parent,cancel=1))
 
 def deleteUser(i):
     for n in ["Are you sure","Are you REALLY sure","This is your last chance: Are you REALLY SURE"]:
-        if not tkMessageBox.askyesno(app.master.title(),u""+n+" you want to delete "+lastUserNames[i]+" permanently, including any vocabulary list and recordings?"): return
+        if not tkMessageBox.askyesno(app.master.title(),ensure_unicode(n)+" you want to delete "+lastUserNames[i]+" permanently, including any vocabulary list and recordings?"): return
     numUsers=len(lastUserNames)
     for fileOrDir in user0+(userNameFile,):
         d=addUserToFname(fileOrDir,i)