While the folder icon was originally planned when in append/link mode, it's easier to distinguish with a blender icon, so the folder icon is now replaced.
Also fixed issue introduced in rev. 27491 where filter settings were incorrectly set when moving out of .blend file again.
Added automatic generation of lookup_int callbacks for collections, for
quicker lookup by index instead of looping over the whole thing. Import
is still quite slow, though now it only takes a few seconds.
The next bottleneck seems to be running update (depsgraph, notifiers, ..)
on setting every property. I fixed part of that by avoiding a notifier
to be added each time, now it checks for duplicates.
After watching 2.5 from a distance,
i did review the soft body module
to match in 2.5 every thing can be animated rule.
Until now i did not realize, that, by default, every property
is 'fcurve'-able unless told to be not.
I really like it that way.
However SB code did assume some things not to be
changing after birth of the SB object.
After spending some hours with softbody.c
/* as may be read in its history */
I think most of the SB properties are ready to go.
For those that do not, some of them never will,
i did reset the flag in the RNA definitions.
There is one not completely resolved:
bending stiffness
which will work if the initial value was non zero,
because only in this case the secondary set of springs
needed is built at all.
Duh, and there a zillions of cases to test ..
please do so.
because the mipmap was not being refreshed. Also this will be problematic
to support when I add tile/mipmap cache, so would not rather not try to.
Can be added back afterwards if someone wants to make it work.
#21506 'sort by' while in append/link file browser causes files to disappear
- issue was that the objects were still filtered
- temporary fix until refactoring of the append/link integration
This fixes [#21087] Opacity of 0 turns off effect rather than affecting transparency
and makes the whole early_out-business in strip stack a lot more readable.
The actual fix is just using the composited result in layer fall through
case (se1->ibuf_comp instead of se1->ibuf).
Mesh's boundbox should be re-calculated after curve->mesh conversion.
To avoid troubles with displaying texture space i've used
tex_space_mesh() for this.
encoding format
Bypassed existing hardcoded ffmpeg presets that executed when changing
format, replaced with bpy presets.
Leaving old code there for now, haven't got python/rna access to the ffmpeg
id properties.. Anyone know how to do this?
Code snippets here: http://www.pasteall.org/11657/c
- Replaced hardcoded nurbcol array with theme colors
- Send notification in duplicate curve operator (this operator could
reset/change active nurb)
- Edge seam color added to the user preferences dialog
This fixes a bug where the texture coordinates were wrong, and also
makes filtering now work even at the borders of the tiles into which
the image is split. Also see bug #20933.
