Created attachment 118568 [details] test output On all tested platforms, the following failures have been bisected to a single commit: piglit.spec.!opengl 2_1.pbo piglit.spec.!opengl 1_0.gl-1.0-beginend-coverage Additionally, the same commit regressed the following test ONLY on 32 bit platforms: piglit.spec.!opengl 1_0.gl-1.0-no-op-paths The problematic commit: Author: Brian Paul <brianp@vmware.com> Date: Wed Sep 30 11:29:13 2015 -0600 mesa: consolidate texture binding code Before, we were doing the actual _mesa_reference_texobj() call and ctx->Driver.BindTexture() and misc housekeeping in three different places. This consolidates the common code in a new bind_texture() function. Reviewed-by: Tapani Pälli <tapani.palli@intel.com> Output for each test saved in attachments.
Created attachment 118569 [details] test output
Created attachment 118570 [details] test output
It is _mesa_BindTexture that is broken, I can't quite see why though. Will try to debug.
Created attachment 118583 [details] [review] hack that fixes issue The bug always occurs whenever default texture object get used, it's textureIndex is always 0 so textureIndex used in new bind_texture function is incorrect. Attached patch fixes the issue, maybe real fix would be to add index as one of the parameters, what do you think Brian?
*** Bug 92231 has been marked as a duplicate of this bug. ***
The hack does not fix the issue on g33, fwiw
*** Bug 92263 has been marked as a duplicate of this bug. ***
(In reply to Tapani Pälli from comment #4) > Created attachment 118583 [details] [review] [review] > hack that fixes issue > > The bug always occurs whenever default texture object get used, it's > textureIndex is always 0 so textureIndex used in new bind_texture function > is incorrect. Attached patch fixes the issue, maybe real fix would be to add > index as one of the parameters, what do you think Brian? I can't reproduce the regressions on a 64-bit nor 32-bit host with llvmpipe. I did full piglit runs before/after my patch series and there were no regressions (modulo the BindTextureUnit expected error code). Has anyone else been able to repro?? In any case, we should initialize the default textures' TargetIndex fields but I think a different patch/approach would be better. I'll post a patch for that in a bit.
I just posted a patch series to mesa-dev. The 3rd patch changes the TextureTarget initialization code and covers the case which the "hack that fixes the issue" did. I have no idea, though, if this will help with the regression that Mark hit.
I tested this patch and found: regressions: spec.arb_shader_storage_buffer_object.layout-std140-write-shader (BDW only) expected[1] = 1.00. Read value: 0.00 spec.arb_copy_image.arb_copy_image-srgb-copy (assertion) arb_copy_image-srgb-copy: src/mesa/main/texobj.c:1739: _mesa_BindTexture: Assertion `newTexObj->TargetIndex == targetIndex' failed. fixes: spec.!opengl 1_0.gl-1.0-beginend-coverage spec.!opengl 1_0.gl-1.0-no-op-paths spec.arb_copy_image.arb_copy_image-formats
*** Bug 92325 has been marked as a duplicate of this bug. ***
fixed by mesa 7d7dd18
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.