We'll need to move crashreporter.ini into toolkit/locale and whatever other build-fu is necessary to make the crashreporter client fully localizable.

We should switch this from an ini file to a properties file, since we're just parsing it for key=value pairs anyway. That should make l10n easier.

Yeah, what Ted said, as per discussion on IRC.

The parser at http://mxr.mozilla.org/mozilla/source/toolkit/airbag/client/crashreporter.cpp#130 doesn't do any kind of format checking. It won't eat everything that we consider to be a .properties file, though, so in order to not stumble over that, we should fix the parser to be a little more generous. Sample implementations are at http://lxr.mozilla.org/mozilla/source/xpcom/ds/nsPersistentProperties.cpp#146 and I'm using regexps at http://lxr.mozilla.org/mozilla/source/testing/tests/l10n/lib/Mozilla/Parser.py#95, too. Basically, the difference is leading whitespace is ignored, the ':' can be a separator char, and whitespace around the separator char is ignored. Trailing spaces and tabs, too. The worst part is, we're still unescaping unicode escapes in .properties, which for whatever reason some tools out there might still serialize. If you don't want to support all of that, we could just stick with the ini file, or preprocess a .properties file into an ini-ish file, but that would likely require moving some of the python test code out of testing into build. Thinking about it some more, at least translate toolkit seems to do ini import and export, we should probably just use the ini file with its constrained grammar. MozillaTranslator has issues with integrating it, but offers a plain text translation of it, which should be good enough.

Some more info for my sanity: http://developer.mozilla.org/en/docs/Localization_notes bug 382128 may or may not be a blocker here, it's something we should definitely consider. We'll need a note that the two replacements here cannot be reordered: http://mxr.mozilla.org/mozilla/source/toolkit/airbag/client/crashreporter.ini#6

Update to the links, http://mxr.mozilla.org/mozilla/source/toolkit/crashreporter/client/crashreporter.cpp#130 http://bonsai.mozilla.org/cvsblame.cgi?file=/mozilla/toolkit/crashreporter/client/crashreporter.ini&rev=1.7&mark=6 I'm not sure if bug 382128 should block, but maybe it doesn't matter. I'll comment there. Ted, as M8 is out, care to retriage this one? Thanks

Ok, this patch moves crashreporter.ini to toolkit/locales/crashreporter and does a little bit of makefile stuff to make it work. It also adds l10n notes to the crashreporter.ini file.

Can we have the crashreporter.ini strings in a .properties format? Just like the NSIS installer stuff?

You can see Pike and I discussed this, he seemed to think it would be ok. We'd have to do some work to make the crashreporter app parse real properties files as per comment 4.

Comment on attachment 283391 [details] [diff] [review] move crashreporter.ini to toolkit/locales/crashreporter + build-fu to make it work + localization notes r=me. We have .ini files in l10n already, so I'm OK with adding another one, and parsing .properties is just a lot more work than .ini. Nice localization notes, thanks.

Comment on attachment 283391 [details] [diff] [review] move crashreporter.ini to toolkit/locales/crashreporter + build-fu to make it work + localization notes Pretty trivial patch, but necessary for localization of the crash reporter client. Also note the blocking1.9+.

Checked in since we're back in orange and it's a blocker.