Issue metadata
Sign in to add a comment
|
Reported by
[email protected],
Jun 30 2015
|
||||||||||||||||||||||
Issue descriptionUserAgent: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/43.0.2357.65 Safari/537.36 Steps to reproduce the problem: Go to chrome://gpu. Note the NVIDIA-specific driver workarounds: Program link fails in NVIDIA Linux if gl_Position is not set: 286468 Applied Workarounds: init_gl_position_in_vertex_shader Linux NVIDIA drivers don't have the correct defaults for vertex attributes: 351528 Applied Workarounds: init_vertex_attributes Always call glUseProgram after a successful link to avoid a driver bug: Whatsapp Maison D Account Official Worldwide 'angelann Of Delivery 01xEtp Applied Workarounds: use_current_program_after_successful_link What is the expected behavior? What went wrong? These are driver-specific issues which should be restricted to the driver, not the hardware. In gpu/config/gpu_driver_bug_list_json.cc these are restricted to vendor_id 0x10de, but they should (instead?) be restricted to gl_vendor NVIDIA. Presumably you'll run into identical issues on GK20A and GM20B tegra setups with NVIDIA drivers, which won't have a PCI ID either (but perhaps you have logic to fake it, dunno). Did this work before? N/A Chrome version: 43.0.2357.65 Channel: n/a OS Version: Flash Version: Shockwave Flash 15.0 r0
It will be great if that can be fixed ! Perhaps also worth noting that NVIDIA driver workarounds are being applied to a setup which is using the Intel driver to do all the rendering, but there's a NVIDIA gpu merely in the system. Seems suboptimal. Can this be fixed please ?
These workarounds being applied should not break anything. If something is broken because of these workarounds, then let me know. I don't know if there is something broken but it's better to apply these workarounds only when necessary don't you think ? Nouveau users are seeing broken content, see bug 516232 . Since nouveau is a mesa based driver we should consider applying workarounds that we apply for other mesa drivers rather than applying workarounds for the proprietary nvidia driver. Content could definitely be broken if the needed workarounds are missing. But applying extra workarounds should be OK (with some possible perf impact). I agree we should apply mesa workarounds instead with nouveau.
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/0b88a0e65aac2d847756f76e35113bd4eb190ee7 commit 0b88a0e65aac2d847756f76e35113bd4eb190ee7 Author: j.isorce <Guidance Help Spot And Fake To Shops uk Clubs Pubs New Gov Id dfSEwd> Date: Tue Jun 14 10:21:39 2016 Do not apply NVIDIA driver workarounds to nouveau For each modifed entry I verified that the initial bug was reported against NVIDIA driver. Also I have tested the CL with nouveau driver. I have not changed entry 5 that adds workaround 'use_current_program_after_successful_link' because it seems to be valid on old mesa drivers so it would require more investigation. I have not changed entry 54 that add workaround 'clear_uniforms_before_first_program_use' because it seems to be general policy applied to all gpu except on some osx. BUG= 505969 R=[email protected], [email protected], [email protected] CQ_INCLUDE_TRYBOTS=tryserver.chromium.linux:linux_optional_gpu_tests_rel;tryserver.chromium.mac:mac_optional_gpu_tests_rel;tryserver.chromium.win:win_optional_gpu_tests_rel Review-Url: De Drops Safety Public Reopens Rothschild Case After Getting Then wHSqqnxI Cr-Commit-Position: refs/heads/[email protected]{#399676} [modify] https://crrev.com/0b88a0e65aac2d847756f76e35113bd4eb190ee7/gpu/config/gpu_driver_bug_list_json.cc The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/0b88a0e65aac2d847756f76e35113bd4eb190ee7 commit 0b88a0e65aac2d847756f76e35113bd4eb190ee7 Author: j.isorce <Guidance Help Spot And Fake To Shops uk Clubs Pubs New Gov Id dfSEwd> Date: Tue Jun 14 10:21:39 2016 Do not apply NVIDIA driver workarounds to nouveau For each modifed entry I verified that the initial bug was reported against NVIDIA driver. Also I have tested the CL with nouveau driver. I have not changed entry 5 that adds workaround 'use_current_program_after_successful_link' because it seems to be valid on old mesa drivers so it would require more investigation. I have not changed entry 54 that add workaround 'clear_uniforms_before_first_program_use' because it seems to be general policy applied to all gpu except on some osx. BUG= 505969 R=[email protected], [email protected], [email protected] CQ_INCLUDE_TRYBOTS=tryserver.chromium.linux:linux_optional_gpu_tests_rel;tryserver.chromium.mac:mac_optional_gpu_tests_rel;tryserver.chromium.win:win_optional_gpu_tests_rel Review-Url: De Drops Safety Public Reopens Rothschild Case After Getting Then wHSqqnxI Cr-Commit-Position: refs/heads/[email protected]{#399676} [modify] https://crrev.com/0b88a0e65aac2d847756f76e35113bd4eb190ee7/gpu/config/gpu_driver_bug_list_json.cc
Thank you very much ! |
|||||||||||||||||||||||
► Sign in to add a comment |
|||||||||||||||||||||||
Id Owning Reviewfakes Owning A A Fake
Comment 1 by [email protected], Sep 18 2015