Blender Git Loki

Fix #33040: baking selected to active could miss at pixels at the edge of faces
when there was a tiny mismatch between low and high poly models, maybe because
of float precision when editing the mesh. Added a small epsilon now to avoid this.

RNA C++: fixes for lookup_int/lookup_string which were not working in some cases,
and don't use ReportList for function arguments.

UV editor: change unwrap warning about object non-unit size to only warning about
non-uniform scale, otherwise it gives warnings for cases when it's not needed.

fix for own recent commit checking bounds off-by-one.

fix for compositor regression where blur note offset the image one pixel to the top right.

correct operator preset tooltip and dont save their settings for re-use

own recent commit to add pointer passing between py-rna and our C code, broke function calls for C++ rna api.

reported as [#33048] Cycles crashes when rendering

OSX: suppress meaningless /net and /home in fileselector also on 10.5+

all remove functions now invalidate the RNA objects passed, to help script authors to avoid bugs with accessing removed data.

Bugfix #24030

Greasepencil mode 'hold d' was also inserting drivers (hotkey D) on mouse over.
The modal operator was default passing on all events, I made it swallowing it.
Doesn't seem to be affecting use at all.

all library data now gets the PointerRNA's invalidated on removal.

Fix for Hue/Sat/Val compositor node. This was using the wrong input socket for the color constant (Fac instead Image).

fix for release building

Bugfix [#33036] Invalid memory read in wm_handlers_do (valgrind warning)

fix for long standing problem with blender 2.5x py api.
Removing data then accessing would allow invalid memory access and often crash.

Example:
import bpy
image = bpy.data.images.new(name="a", width=5, height=5)
bpy.data.images.remove(image)
print(image.name)

Now access to the removed data raises an error:
ReferenceError: StructRNA of type Image has been removed

This is the same level of error checking that was done in blender 2.4x but was made difficult by RNA functions not having access to the PyObject's.

Bugfix #33038

TIMER events could get keymodifier set - in this case the user assigned
spacebar modifier for setting views (running smoothview timer)

Also: cleaned op old hacks from event checking code.
The rule should be:

1) generate event properly, frozen state
2) pass on to handlers without exceptions or changing internal state

The only exception currently is for the "CLICK" (map key-release to unhandled
key-press).

Also: made --debug-event print OK. Slight cleanup in eventprinting in general.
It was putting the wmEvent state print in wrong place, doing it 4 or 8 times.

remove BLI_utildefines from BKE_DerivedMesh.h header

Bugreport - Christian Krupa in irc:

Curves behaved totally bad suddenly. Caused by Campbell code cleanup, replacing
this:

/* this is for float inaccuracy */
if (t < knots[0])
t = knots[0];
else if (t > knots[opp2])
t = knots[opp2];


with:

t = (t < knots[0]) ? knots[0] : knots[opp2];

Tss!

fix for possible buffer overflow in gpu_nodes_get_vertex_attributes() and hair_velocity_smoothing()
and a unlikely NULL pointer dereference in unlink_material_cb().

style cleanup