Rebased on latest blender2.8.

Change the value of CMP_NODE_DISTANCE_TRANSFORM from 323 to 324. 323 was used for CMP_NODE_CRYPTOMATTE.

Rebased on the latest master.

@Robert Guetzkow (rjg)
The major part of the "source code" is under GPL2+, but Cycles is Apache2.0 that's incompatible with GPL2.
Consequently, the Blender "binary" built with Cycles becomes GPL3+ automatically.
COPYING file is not included in the binary distribution.

Rebased on latest blender2.8.

readfile.c: Correct initialization of wmWindow

Rebase on latest master, also delete trailing whitespace

@Sergey Sharybin (sergey) @Julian Eisel (Severin) @Bastien Montagne (mont29)
Hey reviewers, please don't leave it untended for a long time.

Rebased on latest master.

Rebased on the latest master.

@Sergey Sharybin (sergey) Sergey, can you check on this again? You are now the main reviewer.

@Michael P. (forest-house) You forgot to set parameters of the Dilate/Erode Node. Setting the Distance -1 removes the artifact.

Add brief comments, no functional change.
Also rebase on latest master.

• Forgot to update extern/smaa_areatex/README.blender

Also update the test build for Windows 64bit.

• Improve render quality
• Fix several render issues
• Remove unimportant options such as Predicated thresholding
• Remove "Blending Weights" output that is useful only for debugging
• Rename "Depth" to "Value"
• Modify UI texts
• Move smaa_areatex.cpp to extern/smaa_areatex/
• Optimization
• Style cleanup
• Update copyright year

Style and comment fixes, remove own utility functions

Here is a test build for Windows 64bit.

Add new API to set modal, handle line height, update comments

Update after personal communication with @Campbell Barton (campbellbarton).

@Julian Eisel (Severin) Can you make sure if this patch doesn't change behavior of the Win32 IME support? (I already confirmed that on my Win10 PC, though.)

Generalize to one API, don't force redraws, deactivate XIM when it's apparently unneeded

• Update for latest master, dropping SConscript

@Campbell Barton (campbellbarton)
Er, perhaps you misunderstand the behavior of the over-the-spot style input:

• Fix segfault that happens with no available input style, correct active region check

Thanks for your review.

Ray mirror, AO, and ray/buffer shadows have the same problem.

BTW, I noticed the only shadow lamp seems to be not working correctly in Blender Render.
If "Only Shadow" is enabled, disabling "Specular" option takes no effect.

Thanks for the problem report.

Closed by commit rB8008d9bdfd57.

Indeed I didn't consider BGE when I added support for X input method.

Er, maybe @Brecht Van Lommel (brecht) misunderstands the current behavior of the transparent shadow... The shadow color always becomes black (not the "combined" color!) because the shader output node doesn't set shr->diff. (I've already noticed this strange behavior several months ago and plan to fix it.)

This is caused by "if(shi->nodes==0 &&" in rayshade.c L.544, which has been added in order to prevent infinite recursion.
Removing it resolves the issue though I'm not sure if it's still needed.

Committed rB6ec2d72eca61. Thanks!

The behavior change is intentional. I'd like to fix the existing behavior that's confusing and a bit buggy. Currently, if active layer is disabled, the rendered viewport shows all layers for actlay=0, and shows primary layer (or isn't updated) for actlay>0.

Ah, got it. Thanks for the investigation.

Thanks for the fix, the crash doesn't occur anymore.

That's curious, I confirmed this on both Ubuntu and MS-Windows...

Probably this already has been fixed by updating the OSL library package in Mar 26. Now it uses the LLVM static library to avoid symbol conflicts with the graphics driver.

I believe the issue has been resolved by latest update of OSL package, which now uses LLVM as static library.

Committed d81f492d08e6 to bring back the previous behavior that rendered viewport always uses render visibility, because 162d6c73e3d0 caused some problems. For example, mesh deform cage is drawn as a solid/textured mesh (not a wireframe mesh) and its unnecessary surfaces and shadows mess up the preview.

Thanks, it works fine now. :)

The problem still isn't resolved, deleting/adding render layer doesn't update viewport...
The following functions should need DAG_id_tag_update():

I made a patch D346. So far it works fine for me though it's only workaround.

Um, the crash seems to be fixed but another problem happens.

Thanks, but rB277126881d69 fixed only the case 1.
The crash (case 2) still happens, unfortunately.

@Dalai Felinto (dfelinto) Ah you are right, the cases 1 and 2 are similar but different bugs...

Ah OK, the remaining dead-lock is quite rare, so we can archive this.

Any update on this? Many users want the problem to be resolved.

Tested with FFmpeg 2.0.2 and 2.1.3. Totem, MPlayer, and VLC could correctly play movie files that the Vorbis encoder was used.

My PPA build provides a debug symbol package "blender-dbg". The symbols are contained in /usr/lib/debug/usr/lib/blender/blender. nm command should be able to collect symbols from that file.

Any update on this?

On Ubuntu 13.04, the Gallium graphics driver uses LLVM 3.3, while Blender and OSL in my PPA use LLVM 3.2 shared library. I guess the different versions of LLVM shared libraries cause the problem.

Oh, that's good news. Thanks for the report. :)

Step to reproduce:

Thanks, I'm now testing Py_BEGIN_ALLOW_THREADS + Py_END_ALLOW_THREADS, but there is another case not related to the Python global interpreter lock and the deadlock still happens. BLI_ticket_mutex_lock() in sub thread also causes the deadlock.

Fixed in git 69cbf3d.

I confirmed GLSL view is not essential. Switching the viewport from "Solid" to "Rendered" also can cause the crash.