Blender Git Loki

Moving back sr locales to sr_RS (else, with recent changes (42830), loading sr was no more working). My mistake, using only sr wasn't useful here (unlike spanish, we only have one version of serbian, in cyrilic and latin forms...)!

Fix #29652: operator tab can not be maximised

Clamp region's size on expand to max allowed size, so it wouldn't
be automatically hidden just after expand.

svn merge ^/trunk/blender -r42778:42839

give errors when using unsupported configurations with mingw`

Cleaning up the GPU_extensions_init/exit() code a bit to keep the Blenderplayer from crashing on exit and restart.

fix for own mistake [#29031] Importing BGL Can Crash the BGE and Blender

was adding the Buffer type to the module without increffing

add DNA support for int64_t. nothing uses this currently.

recent matrix row/col swap broke matrix assignment in the BGE, fix provided by Andrew Hale

DNA genfile - de-duplicate some checks and replace magic type numbers with an enum

Another Fix after Cucumber Merge:
* Cycles did not displayed the Scene panels anymore, due to addition of COMPAT_ENGINE to them.

update the api changelog 2.60 --> 2.61

Code cleanup: fix some clang static checker warnings.

Fix #29494: Problem loading translations at Blender's startup

Problem was caused by trying to set locale to short named "es" locale
which failed. It's not really obvious which full locale name should
be used here (there are plenty of dialects), so rather than keeping
locale stuff on state when both of locale and utf-8 locale names
failed, restore default settings (restore environment variables and
use default locale for gettext).

This will resolve cases when spanish language is native on the system,
but it will fail in cases when somebody will want to have spanish
interface in non-spanish system. This might be worked around by setting
LANG and LANGUAGE environment variables to es_<dialect>.UTF-8 and it
should work fine,

Translation context for RNA properties

This commit implements a way to define context of property which is used by
localization stuff and which is needed to resolve translation context when
some word wit the same english spelling is used in different meanings
(like Manual in meaning of tutorial, and Manual in meaning of something is
setting up by hand).

To define property's context there's a function RNA_def_property_translation_context.
If property doesn't have context, regular BLF_gettext function is used to get
translation of property name, otherwise BLF_pgettext is used for this.

Hence, for correct translation, messages in .po files should be marked
by "msgctxt" context, otherwise property with context declared wouldn't
be translated at all. Toolchain scripts from bf-translation project
would be updated soon.

If context for some values of enumerator property, property itself should
be moved to other context and all items from this enum would be moved to
this context automatically (it's impossible to move one few items to
another context).

P.S. Think context like "BRUSH" or "MODIFIER" are preferable than "NOUN" and "VERB"
because in some cases the same english noun used in different areas better be
translated differently to make translation more native.

Fix/workaround for russian ui crash on exit

Crash was caused by data segment corruption near "string" in view3d_modeselect_pup().
Enlarged size of this static buffer, so it's now enough to store translated modeselect string.

It's not actually fixes because in some other language this modeline might be much longer
and it'll lead to corruptions again.

Some small perf tweaks to weightvg proximity: do not use indices in case all vertices are affected...

update axis conversion for recent matrix changes

fix for matrix assignment with recent changes which broke for eg:
ob.matrix_world = matrix

remove commented index drawing code (now its a debug option), and remove code in editmesh stat drawing which was left over from when text drawing was immediate rather then cached.

patch [#29673] Visualize Indices (developer aid)
by Howard Trickey (howardt)