Issue is that the real default context is NULL, however, in python and RNA, this value can't be used easily. So we use a specific string instead ("*"), defined as BLF_I18NCONTEXT_DEFAULT_BPYRNA.
From now on, all bpy/rna code should only use the BLF_I18NCONTEXT_DEFAULT_BPYRNA value, while all "usual" C code should use the BLF_I18NCONTEXT_DEFAULT value (BLF_pgettext is still able to "understand" both, anyway).
Also added BLF_is_default_context helper func, so that we can keep that check in a single place!
Finally, we should no need anymore to understand the void string "" as default context too - two values for a same thing are more than enough!
This also revealed another bug, as you could not explicitely set default context to text_ctxt UI func parameter (None is not accpeted by RNA string props), so I had to change default context from py POV to "*" instead of None.
Anyway, that physics UI translation remains weak, as the trick used here (helper func) prevents message extractor script to directly find them. Currently it works because specified labels are also defined elsewhere, but it would be nice to have some kind of "translation markers" in py code too (similar to our N_/CTX_N_ C macros, unfortunately python does not have preprocessing ;) )...
(Did not add those when I created that module, because I did not thought we would actually need them in usual UI code, but turned out I was wrong).
Also made some optimizations in those py gettext funcs, when i18n is disabled at build time, no need to do pyobject -> cstring -> pyobject conversions!.
Also now display nice "area" units (maybe using "length" units was a perf matter, but anyway, you can't have more than a few tens of values displayed at a time, after that they become unreadable). Easy to undo anyway if we really want to keep ugly "10m" as area display!
This should not be needed, I really see no reason for this linking error, but I'd rather use this hack than wasting more time over this issue. When will we get rid of this "two build systems to maintain" time-wasting situation?
This bug did not appear earlier because the "" default context was actually never used, always NULL context was passed instead. But bpy.app.translations uses "" as default context in its keys (simplifies the hash/comp functions of internal py messages cache)...
So now, Blender prefers NULL (None in python) as default context value, but understands also "" as such.
This commit adds:
* A new bpy.app.translations module giving some info about locales/translation stuff (current active locale, all locales currently known by blender, all translation contexts currently defined, etc.).
* The ability for addons to feature translations, using the (un)register functions of above module.
* Also cleans up "translate py string when storing into RNA prop" by removing "PROP_TRANSLATE" string's subtype, and adding a PROP_STRING_PY_TRANSLATE flag instead (this way it is no more exposed to python...).
Addon translations work with py dictionaries: each addon features a dict {lang: {(context, message): translation, ...}, ...}, which is registered when the addon is enabled (and unregistered when disabled).
Then, when a key (context, message) is not found in regular mo catalog, a cache dict for current locale is built from all registered addon translations, and key is searched in it.
Note: currently addons writers have to do all the work by hand, will add something (probably extend "edit translation" addon) to automate messages extraction from addons soon(ish)! To get a look to expected behavior from addons, have a look at render_copy_settings/__init__.py and render_copy_settings/translations.py (rather stupid example currently, but...). Once we have a complete process, I'll also update relevant wiki pages.
