4 years agoMove allocation of imbuf from array to allocimbuf.
Antony Riakiotakis [Mon, 23 Feb 2015 16:19:06 +0000 (17:19 +0100)]
Move allocation of imbuf from array to allocimbuf.
Skip allocation of temporary imbuf

4 years agoCleanup: Simplify brick texture code a bit.
Thomas Dinges [Mon, 23 Feb 2015 15:49:17 +0000 (16:49 +0100)]
Cleanup: Simplify brick texture code a bit.

4 years agoRNA: use 1/0 for move layers operator
Dalai Felinto [Mon, 23 Feb 2015 15:17:28 +0000 (12:17 -0300)]
RNA: use 1/0 for move layers operator

internal api Consistency.

For the records, I suspect there are still a few cases of this. I found
this by chance (moving an object to a different layer), and I ran into
the previous one (fixed by Campbell) as well (by adding a cube).

Anyways, since it only happens when building with crash on asserts is
not a big issue. But by the time we change RNA to use bool instead of
int it would be nice to do a call for test to prevent those breaks.

4 years agoGPU module: get rid of gluscaleimage - game engine also uses that but
Antony Riakiotakis [Mon, 23 Feb 2015 15:33:18 +0000 (16:33 +0100)]
GPU module: get rid of gluscaleimage - game engine also uses that but
commit to remove from there will be done separately due to issues with

4 years agoAfter discussion on irc change new function to only allocate from a
Antony Riakiotakis [Mon, 23 Feb 2015 15:30:37 +0000 (16:30 +0100)]
After discussion on irc change new function to only allocate from a
buffer only. scaling can be done separately

4 years agoIMB library: Add function that scales an array of byte or float pixels.
Antony Riakiotakis [Mon, 23 Feb 2015 14:51:01 +0000 (15:51 +0100)]
IMB library: Add function that scales an array of byte or float pixels.

Function just wraps the array in an imbuf and does regular imbuf

4 years agoFix compilation error after recent GPU debug changes
Sergey Sharybin [Mon, 23 Feb 2015 13:09:28 +0000 (18:09 +0500)]
Fix compilation error after recent GPU debug changes

4 years agoFix mismatch (missing 'const' to mactch funcs declarations).
Bastien Montagne [Mon, 23 Feb 2015 12:55:11 +0000 (13:55 +0100)]
Fix mismatch (missing 'const' to mactch funcs declarations).

Was breaking windows compile, reported by bdancer over IRC, thanks.

Also, quite some annoying 'unused vars' warnings (debug-only vars).

4 years agoDebug GPU functionality from soc-viewport_fx by Jason Wilkins
Antony Riakiotakis [Mon, 16 Feb 2015 20:19:12 +0000 (21:19 +0100)]
Debug GPU functionality from soc-viewport_fx by Jason Wilkins

patch number D706 with changes:

- WITH_GPU_DEBUG just creates a debug context (and enables the debug messaging
system functions) but leaves the checks we had intact. Old patch
added the debug functionality only if we had the flag on to save some

Rationale here is that we might not want to recompile blender just to get
the extra information, and having users start blender with a -d flag to
get the extra information is also useful for bug reports. Those checks already
existed and most expensive ones are hidden behind a debug mode check
so performance should not be that bad.

- Did some cleanup of existing functionality:
When things go wrong blender side, just print the error,
don't check for GL errors first.

- Did not port changes needed for GLES to regular glew.h

- Got rid of duplicate or very similar new functionality.

Generally, code is more moving things around/cleanup and should work exactly
as before apart from the debug context, so it's safe to add even now.

It also provides a nice substitute function for glu error descriptions

4 years agoSnap to Active: missing armature/pose/mball/curve
Campbell Barton [Mon, 23 Feb 2015 12:05:54 +0000 (23:05 +1100)]
Snap to Active: missing armature/pose/mball/curve

de-duplicate active center calc between transform & snap-to-selected.

4 years agoFix T43774: Snap to Cursor ignores active pivot
Campbell Barton [Mon, 23 Feb 2015 11:24:48 +0000 (22:24 +1100)]
Fix T43774: Snap to Cursor ignores active pivot

4 years agoFix T43776: Rigging : Clear User transform is not working when mesh is in weight...
Bastien Montagne [Mon, 23 Feb 2015 11:23:13 +0000 (12:23 +0100)]
Fix T43776: Rigging : Clear User transform is not working when mesh is in weight paint mode.

Use same trick as for regular 'clear transform' operators to get a valid armature
in weight paint mode (where active object is weight-painted one).

4 years agoFix T43740: 'Back to Previous' failed on reload
Campbell Barton [Mon, 23 Feb 2015 10:33:14 +0000 (21:33 +1100)]
Fix T43740: 'Back to Previous' failed on reload

Would restore with the file-selector where the 'info' header should be.

4 years agoScreen: remove redundant NULL check in ED_area_prevspace
Campbell Barton [Mon, 23 Feb 2015 10:06:55 +0000 (21:06 +1100)]
Screen: remove redundant NULL check in ED_area_prevspace

4 years agoPartial fix for T43740
Campbell Barton [Mon, 23 Feb 2015 09:02:54 +0000 (20:02 +1100)]
Partial fix for T43740

ED_area_data_swap would put the screen in an invalid state and crash.
(SpaceLink.spacetype didn't match ScrArea.type)

However behavior is still odd in the instance of the report.

4 years agoFix T43769: Envmap Texture copy looses image
Campbell Barton [Mon, 23 Feb 2015 07:22:09 +0000 (18:22 +1100)]
Fix T43769: Envmap Texture copy looses image

4 years agoCustomData: const correctness
Campbell Barton [Mon, 23 Feb 2015 04:55:48 +0000 (15:55 +1100)]
CustomData: const correctness

4 years agoBLI_utildefines: correct comment
Campbell Barton [Mon, 23 Feb 2015 05:16:23 +0000 (16:16 +1100)]
BLI_utildefines: correct comment

4 years agoEdge/Vert Slide: avoid redundant loop angle calls
Campbell Barton [Mon, 23 Feb 2015 04:44:28 +0000 (15:44 +1100)]
Edge/Vert Slide: avoid redundant loop angle calls

No need to calculate loop angle for each layer

4 years agoError in last commit
Campbell Barton [Mon, 23 Feb 2015 04:40:43 +0000 (15:40 +1100)]
Error in last commit

Accidentally left in malloc/free

4 years agoFix error in recent vert/edge-slide commits
Campbell Barton [Mon, 23 Feb 2015 02:43:09 +0000 (13:43 +1100)]
Fix error in recent vert/edge-slide commits

`CustomData_bmesh_interp_n` was expecting the 'dest' arg not to have its offset applied.

This was a bit confusing since the source args have it applied,
and in some cases we only have the destination with the offset.

4 years agoCustomData: replace calloc -> malloc
Campbell Barton [Mon, 23 Feb 2015 02:55:33 +0000 (13:55 +1100)]
CustomData: replace calloc -> malloc

stack array is uninitialized, relying on calloc here would just hide bugs.

4 years agoCustomData: const correctness for interp()
Campbell Barton [Mon, 23 Feb 2015 02:51:55 +0000 (13:51 +1100)]
CustomData: const correctness for interp()

4 years agoBLI_utildefines: add pointer offset macro
Campbell Barton [Mon, 23 Feb 2015 02:50:12 +0000 (13:50 +1100)]
BLI_utildefines: add pointer offset macro

Handy since it keeps the type of the original.

4 years agoSubsurf: missing NULL checks on free
Campbell Barton [Sun, 22 Feb 2015 21:33:32 +0000 (08:33 +1100)]
Subsurf: missing NULL checks on free

4 years agoFix T43768: Pointcache end frame ignores preview range
Julian Eisel [Sun, 22 Feb 2015 16:14:19 +0000 (17:14 +0100)]
Fix T43768: Pointcache end frame ignores preview range

Fix for T43768

This way it works consistent with cloth, softbodys, etc.

Reviewers: lukastoenne

Differential Revision:

4 years agoCycles: Make it survive if preview_cycles.blend uses persistent data
Sergey Sharybin [Sat, 21 Feb 2015 14:17:09 +0000 (19:17 +0500)]
Cycles: Make it survive if preview_cycles.blend uses persistent data

Currently should be no functional changes, but nice thing to have fixed
for some further experiments.

4 years agoFix T43748: MouseLook actuator raises an Assert error
Sybren A. Stüvel [Sat, 21 Feb 2015 13:39:04 +0000 (14:39 +0100)]
Fix T43748: MouseLook actuator raises an Assert error

An assertion seems to be inverted by mistake, as MT_assert(!m_mouse); is followed
by using m_mouse->XXXX.

4 years agoGPencil: Another fix for the "big square dots" problem
Joshua Leung [Sat, 21 Feb 2015 13:04:20 +0000 (02:04 +1300)]
GPencil: Another fix for the "big square dots" problem

Strokes containing only a single point are now drawn so that they respect the
pressure values too. Previously, they were getting drawn at a fixed size based
on the thickness of strokes. Thanks to @kekeljevic for posting a screenshot which
made the problem here easy to identify!

(On a side note: Perhaps it's a combination of the drivers I'm using, or the
multisampling goodness at work, but after applying this change, the dots are
now appearing as round circles as originally intended, and can also be scaled
up quite a bit too. And that's for 3D points which couldn't be affected like
this earlier! I'm not sure what's going on here, but whatever it is, I like it ;)

4 years agoCode cleanup: Use bools instead of int
Sergey Sharybin [Sat, 21 Feb 2015 12:50:05 +0000 (17:50 +0500)]
Code cleanup: Use bools instead of int

4 years agoCycles: Add note about autodiff in OSL wireframe shader
Sergey Sharybin [Sat, 21 Feb 2015 12:31:41 +0000 (17:31 +0500)]
Cycles: Add note about autodiff in OSL wireframe shader

4 years agoCycles: Use lower progressive update timeout for preview rendering
Sergey Sharybin [Sat, 21 Feb 2015 12:29:27 +0000 (17:29 +0500)]
Cycles: Use lower progressive update timeout for preview rendering

This ways previews are refreshing with the same ratio as job was expecting
this to happen, giving more instant feedback on the changes.

4 years agoFix T43755: Wireframe attribute doesn't work with displace
Sergey Sharybin [Sat, 21 Feb 2015 09:49:55 +0000 (14:49 +0500)]
Fix T43755: Wireframe attribute doesn't work with displace

This attribute missed derivatives calculation.

Not totally sure what's the proper approach for algebraic derivative
calculation, so calculating them by definition. This isn't fastest
way to do it in this case and could be replaced with some smarter magic
in the wireframe calculation loop.

At least currently implemented approach is better than nothing.

4 years agoscene conversion: style cleanup
Ines Almeida [Sun, 3 Aug 2014 17:13:06 +0000 (19:13 +0200)]
scene conversion: style cleanup

4 years agogamengine: removing useless comment
Ines Almeida [Wed, 6 Aug 2014 08:40:09 +0000 (10:40 +0200)]
gamengine: removing useless comment

4 years agogameengine: removing addInitFromFrame hack
Ines Almeida [Wed, 23 Jul 2014 12:07:01 +0000 (13:07 +0100)]
gameengine: removing addInitFromFrame hack

4 years agogameengine: dataconversion import and whitespace cleanup
Ines Almeida [Wed, 23 Jul 2014 10:22:33 +0000 (11:22 +0100)]
gameengine: dataconversion import and whitespace cleanup

4 years agodoxygen: updating blender version and link to tracker
Ines Almeida [Wed, 23 Jul 2014 10:12:09 +0000 (11:12 +0100)]
doxygen: updating blender version and link to tracker

4 years agopython bge.types module - Updating documentation to close T40778
Ines Almeida [Fri, 4 Jul 2014 19:41:32 +0000 (20:41 +0100)]
python bge.types module - Updating documentation to close T40778

4 years agoFix T33762 - texture fonts were not using the material color (gameengine)
HG1 [Wed, 18 Jun 2014 15:55:11 +0000 (16:55 +0100)]
Fix T33762 - texture fonts were not using the material color (gameengine)

4 years agomakeScreenshot - adding frame number to the filename
HG1 [Sat, 14 Jun 2014 07:30:09 +0000 (08:30 +0100)]
makeScreenshot - adding frame number to the filename

4 years agocommiting patch to fix T30173
Benoit Bolsee [Mon, 26 May 2014 09:38:12 +0000 (10:38 +0100)]
commiting patch to fix T30173

4 years agogameengine: KX_Scene comment cleanup
Ines Almeida [Tue, 1 Jul 2014 09:00:20 +0000 (10:00 +0100)]
gameengine: KX_Scene comment cleanup

4 years agoBGE cleanup: removing unused async logic bricks
Ines Almeida [Thu, 12 Jun 2014 10:54:23 +0000 (11:54 +0100)]
BGE cleanup: removing unused async logic bricks

This code was never run and is an outdated copy from the code above.
If someday this is needed again, it is better to start clean.

4 years agoDocumentation update: references to OpenGL tutorials in the python bgl module
Ines Almeida [Wed, 4 Jun 2014 19:37:01 +0000 (20:37 +0100)]
Documentation update: references to OpenGL tutorials in the python bgl module

4 years agoCleanup: Put all Bump dx/dy code in the beginning here, same as with other nodes.
Thomas Dinges [Sat, 21 Feb 2015 11:55:19 +0000 (12:55 +0100)]
Cleanup: Put all Bump dx/dy code in the beginning here, same as with other nodes.

4 years agoCorrection for RNA getter of use_save_buffers
Sergey Sharybin [Sat, 21 Feb 2015 07:47:29 +0000 (12:47 +0500)]
Correction for RNA getter of use_save_buffers

4 years agoCycles: Workaround for NVidia OpenCL compilation
Sergey Sharybin [Sat, 21 Feb 2015 07:33:21 +0000 (12:33 +0500)]
Cycles: Workaround for NVidia OpenCL compilation

It was complaining about explicit __constant to __private memory conversion,
which is now worked around using implicit conversion.

It's not a real fix i'm afraid and i'm still failing to build OpenCL kernel
with latest Linux drivers, but maybe it'll let someone else to investigate
what causes compiler to run out of memory?

4 years agoBGE: Fix T43215: Rigidbody constraints deletion.
Thomas Szepe [Fri, 20 Feb 2015 22:37:32 +0000 (23:37 +0100)]
BGE: Fix T43215: Rigidbody constraints deletion.

Deleting an object was deleting all rigidbody constraints in the scene.
Bug was introduced with D701.

Reviewers: agoose77, sergof, moguri, lordloki, sybren

Reviewed By: lordloki, sybren

Subscribers: sybren, hbar, blueprintrandom, sergof, agoose77

Differential Revision:

4 years agoFix T43550: Bug in Cycles Bake Texture
Sergey Sharybin [Fri, 20 Feb 2015 20:56:48 +0000 (01:56 +0500)]
Fix T43550: Bug in Cycles Bake Texture

Issue was caused by the changes in 7b16fda which changed the initial
state for rng. This commit makes it so the same initial hash is used
(which solves the regression without distorting AA-looking image.

It also makes it so OpenCL compiler is happy about this code (before
this change it'll complain about trying to cast private variable to
global one).

4 years agoRNA: use 1/0 for object layers
Campbell Barton [Fri, 20 Feb 2015 15:08:42 +0000 (02:08 +1100)]
RNA: use 1/0 for object layers

internal api Consistency.

4 years agoYet another commit to get rid of missing faces in fill brushes.
Antony Riakiotakis [Fri, 20 Feb 2015 14:11:12 +0000 (15:11 +0100)]
Yet another commit to get rid of missing faces in fill brushes.

We are now guarding against some divisions by small values.
There are still issues here but they are not on boundary faces
anymore so they must be related to some other issue such as
the triangle intersection test.

4 years agoSequencer: allow scene volume above 1
Campbell Barton [Fri, 20 Feb 2015 13:08:27 +0000 (00:08 +1100)]
Sequencer: allow scene volume above 1

4 years agoFix T43651: New pointiness attribute doesn't work with displacement
Sergey Sharybin [Fri, 20 Feb 2015 12:20:24 +0000 (17:20 +0500)]
Fix T43651: New pointiness attribute doesn't work with displacement

Simple fix: just make pointiness aware of bump offset.

4 years agoFix T43739: bad documentation for bmesh py `intersect_face_point()`
Bastien Montagne [Fri, 20 Feb 2015 09:54:26 +0000 (10:54 +0100)]
Fix T43739: bad documentation for bmesh py `intersect_face_point()`

Previous text could make think this func was checking whether a point was
inside and on the face, while it actually checks whether the projection
of that point onto the face is inside its boundary...

4 years agoRNA: correct bool callbacks
Campbell Barton [Fri, 20 Feb 2015 08:54:28 +0000 (19:54 +1100)]
RNA: correct bool callbacks

4 years agoRevert "Cast result of rna_RenderSettings_use_game_engine_get to {0, 1}"
Sybren A. Stüvel [Fri, 20 Feb 2015 07:41:21 +0000 (08:41 +0100)]
Revert "Cast result of rna_RenderSettings_use_game_engine_get to {0, 1}"

There are more flags that trigger this assertion, so I want to discuss
an appropriate way to fix it before changing them all.

This reverts commit b1191e5caf1a888bfad2185ed41db2e91fbf46e2.

4 years agoCast result of rna_RenderSettings_use_game_engine_get to {0, 1}
Sybren A. Stüvel [Fri, 20 Feb 2015 07:36:33 +0000 (08:36 +0100)]
Cast result of rna_RenderSettings_use_game_engine_get to {0, 1}

rna_RenderSettings_use_game_engine_get() in rna_scene.c would return 0 or 2,
whereas the assertion in rna_access.c:1843 indicates that boolean values
should only be 0 or 1.

4 years agoRNA: calloc -> malloc when fixbuf is uninitialized
Campbell Barton [Fri, 20 Feb 2015 06:16:58 +0000 (17:16 +1100)]
RNA: calloc -> malloc when fixbuf is uninitialized

4 years agoFix for unlikely NULL pointer dereference
Campbell Barton [Fri, 20 Feb 2015 06:17:36 +0000 (17:17 +1100)]
Fix for unlikely NULL pointer dereference

Potential crash reading freestyle modifiers from future blend-files

4 years agoFix possible (unlikely) memory leak
Campbell Barton [Fri, 20 Feb 2015 06:09:40 +0000 (17:09 +1100)]
Fix possible (unlikely) memory leak

4 years agoCycles: Use size_t for aligned allocator
Sergey Sharybin [Thu, 19 Feb 2015 17:17:42 +0000 (22:17 +0500)]
Cycles: Use size_t for aligned allocator

Seems it got changes by accident to int during guarded allocation work,
causing bad memory allocations.

4 years agoSkip flip check in cases where it's not needed.
Antony Riakiotakis [Thu, 19 Feb 2015 15:56:05 +0000 (16:56 +0100)]
Skip flip check in cases where it's not needed.

4 years agoRemove flipping for colinear case/ In such cases it is doubtful whether
Antony Riakiotakis [Thu, 19 Feb 2015 15:51:51 +0000 (16:51 +0100)]
Remove flipping for colinear case/ In such cases it is doubtful whether
face orientation is really meaningful

4 years agoProjective painting: Only dither if factor is more than zero
Antony Riakiotakis [Thu, 19 Feb 2015 15:46:05 +0000 (16:46 +0100)]
Projective painting: Only dither if factor is more than zero

4 years agoFix T43735 painting on perpendicular to the screen polygons is
Antony Riakiotakis [Thu, 19 Feb 2015 15:38:56 +0000 (16:38 +0100)]
Fix T43735 painting on perpendicular to the screen polygons is
problematic (still!)

It seems due to floating point precision issues we have to account for
even more cases.

Added colinear check for one more triangle edge-vertex and assumed that
if no collision test is found then triangle is completely inside box.

Such cases are mostly from float precision not catching triangle just
outside bucket. Now tests are much better for simple tests (paint/flood
fill cube from ortho view, paint sphere), however there are still some
rare cases of stray pixels.

4 years agoCycles: Cleanup for indentation in device_cpu.cpp
Sergey Sharybin [Thu, 19 Feb 2015 14:05:04 +0000 (19:05 +0500)]
Cycles: Cleanup for indentation in device_cpu.cpp

Perhaps became broken after rather recent change about which entry point
to kernel to use.

4 years agoCycles: Free memory used by intermediate BVH vectors earlier
Sergey Sharybin [Thu, 19 Feb 2015 13:58:21 +0000 (18:58 +0500)]
Cycles: Free memory used by intermediate BVH vectors earlier

Ideally we should get rid of those temporary vectors anyway, but
it's not so trivial because of the alignment. For untl then we'll
just have a bit worse solution. This part of code is not the root
of the issue of memory spikes for now anyway.

But since we're getting rid of temporary memory earlier actual spike
is a bit smaller as now. For example in franck_sheep file it's now
5489.69MB vs. previously 5599.90MB.

4 years agoCycles: Reduce memory used by background light update
Sergey Sharybin [Thu, 19 Feb 2015 13:15:07 +0000 (18:15 +0500)]
Cycles: Reduce memory used by background light update

Simple change: just get rid of intermediate data a bit earlier, before
final pixels array is being allocated. This gives around 30% of memory
save during light update (this is about 60meg in the frank sheep file
i'm using here).

This isn't really visible by artists a lot, because actual spike happens
on BVH construction. But it doesn't mean we shouldn't be accurate with
memory usage in other areas.

4 years agoCleanup: Restructure curve settings code a bit.
Thomas Dinges [Thu, 19 Feb 2015 11:36:13 +0000 (12:36 +0100)]
Cleanup: Restructure curve settings code a bit.

This is a bit more logical, as it better represents the 3 primitive types we have.

4 years agoFix 2d paint issue with textured brushes: Multiply brush color with
Antony Riakiotakis [Thu, 19 Feb 2015 11:37:35 +0000 (12:37 +0100)]
Fix 2d paint issue with textured brushes: Multiply brush color with
texture in the right space

4 years agobge physics: disallowing compound collisions for soft bodies
Ines Almeida [Tue, 24 Jun 2014 14:44:34 +0000 (15:44 +0100)]
bge physics: disallowing compound collisions for soft bodies

Closes D610, T40427, T30630

Cherrypicked from local squash

4 years agoBGE physics: adding links to bullet documentation
Ines Almeida [Tue, 13 Jan 2015 18:28:36 +0000 (19:28 +0100)]
BGE physics: adding links to bullet documentation

4 years agogameengine physics: removing triangle mesh collision option for character controllers
Ines Almeida [Sun, 17 Aug 2014 17:55:57 +0000 (19:55 +0200)]
gameengine physics: removing triangle mesh collision option for character controllers

4 years agogameengine physics: removing compound option for character controllers
Ines Almeida [Sun, 17 Aug 2014 17:40:33 +0000 (19:40 +0200)]
gameengine physics: removing compound option for character controllers

Cherrypicking c31d921 onto 117edbb

4 years agoAnimation ghosting: when tweaking start-end values don't clamp user
Antony Riakiotakis [Thu, 19 Feb 2015 11:10:44 +0000 (12:10 +0100)]
Animation ghosting: when tweaking start-end values don't clamp user
input, instead make sure the end-start values are sane instead

4 years agoGuardedalloc: Reset peak memory should set peak to currently allocated memory
Sergey Sharybin [Thu, 19 Feb 2015 08:14:06 +0000 (13:14 +0500)]
Guardedalloc: Reset peak memory should set peak to currently allocated memory

Otherwise statistics could be really funny looking.

4 years agoCycles: Make sphere and tube image mapping friendly with OpenCL
Sergey Sharybin [Thu, 19 Feb 2015 07:52:48 +0000 (12:52 +0500)]
Cycles: Make sphere and tube image mapping friendly with OpenCL

OpenCL doesn't let you to get address of vector components, which
is kinda annoying. On the other hand, maybe now compiler will have
more chances to optimize something out.

4 years agoCycles: Make util_math_fast.h compatible with OpenCL
Sergey Sharybin [Thu, 19 Feb 2015 07:29:06 +0000 (12:29 +0500)]
Cycles: Make util_math_fast.h compatible with OpenCL

4 years agoRNA: pass only 0/1 to RNA_property_boolean_set
Campbell Barton [Thu, 19 Feb 2015 01:33:14 +0000 (12:33 +1100)]
RNA: pass only 0/1 to RNA_property_boolean_set

4 years agoRNA: poll functions restricting object types
Campbell Barton [Wed, 18 Feb 2015 23:02:51 +0000 (10:02 +1100)]
RNA: poll functions restricting object types

D1100,D1105,D1106 by @lichtwerk

4 years agoGenerate correct Cycles node setup for quick smoke "fire" preset.
Thomas Dinges [Wed, 18 Feb 2015 23:17:21 +0000 (00:17 +0100)]
Generate correct Cycles node setup for quick smoke "fire" preset.

Patch by Gottfried Hofmann.
Differential Revision:

4 years agoFix T43652: X11 "_NET_WM_PID" not set by Ghost
Campbell Barton [Wed, 18 Feb 2015 22:19:12 +0000 (09:19 +1100)]
Fix T43652:  X11 "_NET_WM_PID" not set by Ghost

D1107 by @kevindietrich

4 years agoGhost/X11 cleanup (reduce variable scope)
Campbell Barton [Wed, 18 Feb 2015 22:02:36 +0000 (09:02 +1100)]
Ghost/X11 cleanup (reduce variable scope)

4 years agoBGE: Fix T41502 Path following jumping
Jorge Bernal [Wed, 18 Feb 2015 22:24:02 +0000 (23:24 +0100)]
BGE: Fix T41502 Path following jumping

New Lock Z velocity parameter was added. This parameter avoid the micro-jumping.
By default it is actived except when you load an old file that it is deactived to keep former behaviour.

Additionally it was solved another issue related with the acceleration: That is the acceleration value was not taked into account and we had always the maximum linear velocity from the beginning of movement. Now the acceleration is taken into account until we reach the maximum velocity.
When you load an old file, the acceleration value is set to the maximum range (1000.f). This way we simulate a maximum velocity constant from the beginning of movement (former behaviour).


Reviewers: moguri, dfelinto, campbellbarton

Reviewed By: campbellbarton

Subscribers: sergey

Differential Revision:

4 years agoFix Cycles OpenCL error with scons, util_math_fast.h was missing.ø
Thomas Dinges [Wed, 18 Feb 2015 21:43:24 +0000 (22:43 +0100)]
Fix Cycles OpenCL error with scons, util_math_fast.h was missing.ø

4 years agoFix minor mistake - only increase size of compositing when stencil -is-
Antony Riakiotakis [Wed, 18 Feb 2015 21:16:04 +0000 (22:16 +0100)]
Fix minor mistake - only increase size of compositing when stencil -is-

4 years agoReport total unfreed memory size on exit
Sergey Sharybin [Wed, 18 Feb 2015 21:08:09 +0000 (02:08 +0500)]
Report total unfreed memory size on exit

Previously only number of unfreed blocks would have been printed,
which might not be totally enough during investigation process.

4 years agoGuardedalloc: Add extra logging and checks in MEM_freeN()
Sergey Sharybin [Wed, 18 Feb 2015 20:58:49 +0000 (01:58 +0500)]
Guardedalloc: Add extra logging and checks in MEM_freeN()

We don't like when NULL is send to MEM_freeN(), but there was some
differences between lockfree and guarded allocators:

- Lockfree would have silently crash, in both release and debug modes
- Guarded allocator would have printed error message, abort in debug
  but keep working in release build.

This commit makes lockfree allocator behavior to match guarded one.

4 years agoCycles: Make aligned allocation to respect WITH_BLENDER_GUARDEDALLOC
Sergey Sharybin [Wed, 18 Feb 2015 20:48:59 +0000 (01:48 +0500)]
Cycles: Make aligned allocation to respect WITH_BLENDER_GUARDEDALLOC

Title pretty says it all actually. Can only briefly mention that we're
indeed entering that state when after applying some WIP patches having
much fuller statistics about memory usage would help giving exact memory

4 years agoCMake: remove expression in endif(...)
Campbell Barton [Wed, 18 Feb 2015 20:44:40 +0000 (07:44 +1100)]
CMake: remove expression in endif(...)

4 years agoCMake: indent & missing header
Campbell Barton [Wed, 18 Feb 2015 20:15:55 +0000 (07:15 +1100)]
CMake: indent & missing header

4 years agoCMake: picky style edit
Campbell Barton [Wed, 18 Feb 2015 20:15:00 +0000 (07:15 +1100)]
CMake: picky style edit

'' relies on this formattng.

4 years agoRNA: assert when non 0/1 values used as bool
Campbell Barton [Wed, 18 Feb 2015 20:08:10 +0000 (07:08 +1100)]
RNA: assert when non 0/1 values used as bool

Prepare for using 'bool' type.

4 years agoRNA: assert on over-sized array index access
Campbell Barton [Wed, 18 Feb 2015 20:01:49 +0000 (07:01 +1100)]
RNA: assert on over-sized array index access

4 years agoRNA: use __func__ for alloc-id's
Campbell Barton [Wed, 18 Feb 2015 19:57:45 +0000 (06:57 +1100)]
RNA: use __func__ for alloc-id's

4 years agoBGE: Fix T42437: Physics/Compound bug
Jorge Bernal [Wed, 18 Feb 2015 18:59:21 +0000 (19:59 +0100)]
BGE: Fix T42437: Physics/Compound bug

Null check to verify that parent has a character controller. Otherwise (i.e empty) it will crash.

Reviewers: moguri, campbellbarton

Reviewed By: campbellbarton

Differential Revision:

4 years agoBGE: Fix T41570: Blender crash when physics createConstraint
Jorge Bernal [Wed, 18 Feb 2015 18:52:54 +0000 (19:52 +0100)]
BGE: Fix T41570: Blender crash when physics createConstraint

Move physicsid type to unsigned long long to avoid crashes on Windows 8.1 64bits.
Other systems also modified to put them inline with this solution.

Reviewers: dfelinto, brita_, moguri, juicyfruit, campbellbarton

Reviewed By: juicyfruit, campbellbarton

Subscribers: juicyfruit

Differential Revision:

4 years agoFix T43716 regression in mask modifier - edges could access out of
Antony Riakiotakis [Wed, 18 Feb 2015 16:58:57 +0000 (17:58 +0100)]
Fix T43716 regression in mask modifier - edges could access out of
bounds indices of vertices.

Issue here is that if we copy edge customdata the MEdge data that were
previously written get overwritten, solution is to write indices and
flags last (probably writing flags is superfluous here).

4 years agoCycles: Initialize "headless" flags on engine initialization
Sergey Sharybin [Wed, 18 Feb 2015 16:16:52 +0000 (21:16 +0500)]
Cycles: Initialize "headless" flags on engine initialization

This flag is global for all the sessions and never changes. so it doesn't
really make sense to pass it around to all sessions and synchronization

Switched to a static member of BlenderSession now, but it's probably more
logical to introduce some sort of BlenderGlobals. Doesn't currently worth
a hassle for a single boolean flag tho.