It works on Windows XP (mostly)

[unpingco]

I had to make the changes to the completion function to the "old" way that is documented in your code instead of the version that Min recommends:

starting from line 400:

 m = get_child_msg(msg_id)
# get rid of unicode (sporadic issue, haven't tracked down when it happens)
matches = [str(u) for u in m['content']['matches']]
# mirk sez do: completion_str = '[' + ', '.join(msg['content']['matches'] ) + ']'
# because str() won't work for non-ascii characters
#matches = m['content']['matches']
#end = len(base)
#completions = [m[end:]+findstart+base for m in matches]
matches.insert(0,base) # the "no completion" version
#echo(str(matches))
completions = matches

otherwise, it would crash when I would try to do completions and mysteriously would recognize the Ctrl+u part of the completion sequence as the letter 'u' instead of as Ctrl+u. This obviously caused a number of downstream errors that I couldn't isolate.

Great work!

[ivanov]

Thanks for the bug report, but the str version will still fail if there are unicode filenames among the completions. Please try the alternative version from the commit which will close this bug in a few minutes, and reopen the bug if it does not fix the issue for you.

[unpingco]

It works!