- User Since
- Apr 27 2010, 10:07 PM (330 w, 4 d)
Actually the line drawing is not wireframe, it's for the selection outline. My mistake!
This test build changes version to 2.78. New prefs, new startup file.
Ok, that tells us the very first OpenGL function call is failing. Trying to run code at address 0, which might mean it's being called without an active GL context.
Try running this one from Terminal.app. It should tell us which line of code leads to the crash, or if the problem is somewhere else.
This commit also adds trailing whitespace—the invisible disease—to blank lines.
rB9d3813e602b8 Added new functions to Gawain.
Thanks for this! Yes it's useful toward T49043
No I haven't seen the problem on my end.
Fri, Aug 26
This is derived from work I did spring 2015 in GPU_data_request branch.
Thu, Aug 25
Tue, Aug 23
Mon, Aug 22
Sun, Aug 21
Fri, Aug 19
Thu, Aug 18
Also tested working on Windows & Linux.
Wed, Aug 17
Tue, Aug 16
Reported bug to Intel graphics folks.
Crash is inside glLinkProgram, called from GPU_shader_create_ex. The vertex & fragment shaders compiled successfully with nothing in their info logs. Intel driver bug??
On Intel HD 4600 I'm seeing the same bad behavior @Aaron Carlisle (Blendify) described in IRC. Crash in driver, no GL error reported.
driver stuff ... 20: GPU_shader_create_ex - 0xDBC721C0 19: GPU_generate_pass - 0xDBC77EE0 18: GPU_material_construct_end - 0xDBC5FF50 17: GPU_material_from_blender - 0xDBC5B360 16: GPU_begin_object_materials - 0xDBC43A70 15: draw_mesh_object - 0xDAFF82A0 14: draw_object - 0xDAFE5880 13: view3d_draw_objects - 0xDAFCB890 ...
Still need to fix the separate error I see on AMD with the same file. But the Intel one first!
Mon, Aug 15
Here's my system-info.txt GLSL section
nVidia has a good FAQ about these limits. Reading carefully now...
Actual problem is with glActiveTexture(GL_TEXTURE0 + number), not glEnable.
Using your file I get an error inside GPU_shader_uniform_texture, might originate further up.
Did something similar in master to aid debugging in 2.78 rB2c03d0186a01
Caught up with the half float vs uint16 debate in IRC just after you left.
Sun, Aug 14
Fri, Aug 12
Good point @Brecht Van Lommel (brecht). I'll try factory settings for both versions and see if results still differ.
Thu, Aug 11
The crash report doesn't offer many details. Only that some command leads GPUSupport (an Apple library called internally by the drivers) to peace out. Problem could be in the Intel client driver or Blender's misuse of OpenGL. We have not seen this problem on other drivers, so that driver is suspect.
Wed, Aug 10
My bad... was using GitHub client for Windows, which has a shiny "Sync" button instead of pull --rebase and push. I'll be more careful in the future.
Tue, Aug 9
Mon, Aug 8
Sun, Aug 7
Added APPLE_flush_buffer_range to Mac requirements. Useful when updating part of a buffer. This extension has been around a long time and all GPUs support it.
Fri, Aug 5
According to Apple's OpenGL matrix the following helpful extensions are available on all GPUs that are capable of running Blender 2.8:
Thu, Aug 4
Wed, Aug 3
This implements the Windows part of T49012
Tue, Aug 2
Thanks for the info!