Cannot use non-ascii chars in Milestone group name

[trac-ja@…]

If trac.ini includes with non-ascii(i.e. Japanese) characters in [milestone-groups] section, Trac dumps stacktrace as UnicodeDecodeError.

[milestone-groups]
NonAsciiClosed = closed
NonAsciiClosed.order = 0
NonAsciiClosed.css_class = closed
NonAsciiClosed.query_args = group=resolution
NonAsciiClosed.overall_completion = true

NonAsciiActive = *
NonAsciiActive.order = 1
NonAsciiActive.css_class = open

NonAsciiNew = new,reopened
NonAsciiNew.order = 2
NonAsciiNew.css_class = new

I think, it should be able to append label to milestone group name, and I wrote a patch (source:trunk@6429​).

When apply this patch, we can write above case to:

[milestone-groups]
closed = closed
closed.order = 0
closed.query_args = group=resolution
closed.overall_completion = true
closed.label = NonAsciiClosed

active = *
active.order = 1
active.css_class = open
active.label = NonAsciiActive

new = new,reopened
new.order = 2
new.label = NonAsciiNew

# s/NonAscii[A-Z][a-z]+/Japanese or non-ascii chars/

[trac-ja@…]

Patch against source:trunk@6429​

[Christian Boos]

Patch applied in [6571], together with some docs (the TracIni#milestone-groups-section was updated as well). By the way, I'd really like to know the real name behind trac-ja, as I think you deserve your entry in our THANKS file!

Note that the problem with the UnicodeDecodeError will surface again, I'm sure. We would have to make some quite extensive changes in trac/config.py in order to cope with unicode keys. Ideally, that's what we should do, but I'm not really sure if it would not be quicker to completely drop the remaining bits of ConfigParser and do everything ourselves. Large parts are already duplicated / overriden anyway to take care of the [inherit] file logic. The other main advantages of that would be to preserve the ordering of entries, drop the special behavior of % in the values, and last but not least, don't throw away the comments when rewriting the .ini file.