1702463 - glxtest should try choosing both single and double buffered visuals

Status: RESOLVED FIXED

(Core :: Graphics, defect)

Description

[Simo Melenius]

User Agent: Mozilla/5.0 (X11; Ubuntu; Linux x86_64; rv:87.0) Gecko/20100101 Firefox/87.0

Steps to reproduce:

Running Firefox under PRIME render offload on a system with no EGL available/configured might fail the GL detection in toolkit/xre/glxtest.cpp and leave WebGL unavailable for the end-user.

Actual results:

When running Firefox as a render offload client the glxtest.cpp can fail to detect GL properly because single-buffered visuals might not be available (render offload visuals can differ from the larger set of visuals of the primary desktop). The attributes passed in for glXChooseVisual() in get_glx_status() will only match single-buffered visuals.

Expected results:

At the glxtest stage it shouldn't matter which buffering the visual comes with as its purpose is to find a valid GL implementation for WebGL. If it also checked ChooseVisual() with GLX_DOUBLEBUFFER it would find a GLX implementation behind render offload, too. The actual WebGL initialization works correctly regardless of the pre-test in glxtest.cpp.

Comment 1

[Robert Mader [:rmader]]

I'm confused - according to the docs (1) the default value for GLX_DOUBLEBUFFER is GLX_DONT_CARE - and we don't specify GLX_DOUBLEBUFFER:false anywhere. Do I miss something? Is this simply a driver bug maybe?

1: https://www.khronos.org/registry/OpenGL-Refpages/gl2.1/xhtml/glXChooseFBConfig.xml

Comment 2

[Simo Melenius]

(In reply to Robert Mader [:rmader] from comment #1)

I'm confused - according to the docs (1) the default value for GLX_DOUBLEBUFFER is GLX_DONT_CARE - and we don't specify GLX_DOUBLEBUFFER:false anywhere. Do I miss something? Is this simply a driver bug maybe?

1: https://www.khronos.org/registry/OpenGL-Refpages/gl2.1/xhtml/glXChooseFBConfig.xml

It's the glXChooseVisual() that fails on https://searchfox.org/mozilla-central/source/toolkit/xre/glxtest.cpp#718. From https://www.khronos.org/registry/OpenGL-Refpages/gl2.1/xhtml/glXChooseVisual.xml:

GLX_DOUBLEBUFFER

If present, only double-buffered visuals are considered. Otherwise, only single-buffered visuals are considered.

Comment 3

[Robert Mader [:rmader]]

Attachment: Bug 1702463 - Also check for double buffered visual in glxtest, r=aosmond

glXChooseVisual requires us to choose whether we want single or
double buffered visuals. We don't care here, so check for both
cases.

Comment 4

[Robert Mader [:rmader]]

Simo: the patch above should fix the issue - can you try the following build once it's ready?

https://treeherder.mozilla.org/jobs?repo=try&revision=b016371b31ae0cc659dda45b22e66d9fc81d1e9c

If you don't know how to run download it, I can post a direct link once its build.

Edit: https://firefox-ci-tc.services.mozilla.com/api/queue/v1/task/Ma9jnKOJTuyN2FUcaAhNEg/runs/0/artifacts/public/build/target.tar.bz2

Comment 5

[Simo Melenius]

Thanks for the direct download link. I set up PRIME render offload and tested that a vanilla version of Firefox fails to enable webgl. Then I tested your build and it does correctly identify my NVIDIA GL that I've configured to be the render offload source. Both browsers correctly enable webgl with MESA drivers when I run them without PRIME render offload. So, this fixes it. Thanks!

The patch looks good, too.

Comment 6

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/c1b8b60b2e52
Also check for double buffered visual in glxtest, r=aosmond

Comment 7

[Sandor Molnar[:smolnar]]

https://hg.mozilla.org/mozilla-central/rev/c1b8b60b2e52

Comment 8

[Robert Mader [:rmader]]

Comment on attachment 9213076 [details]
Bug 1702463 - Also check for double buffered visual in glxtest, r=aosmond

Beta/Release Uplift Approval Request

• User impact if declined: A small set of dri prime nvidia users won't have WebGL support. They don't get WR, so normal browsing is not affected.
• Is this code covered by automated tests?: Yes
• Has the fix been verified in Nightly?: Yes
• Needs manual test from QE?: No
• If yes, steps to reproduce:
• List of other uplifts needed: None
• Risk to taking this patch: Low
• Why is the change risky/not risky? (and alternatives if risky): This patch introduces a fallback path. All users who currently have working WebGL/WR do not hit it, so only quite broken setups are affected.
• String changes made/needed:

Comment 9

[Ryan VanderMeulen [:RyanVM]]

Comment on attachment 9213076 [details]
Bug 1702463 - Also check for double buffered visual in glxtest, r=aosmond

Approved for 88.0b7.

Comment 10

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/releases/mozilla-beta/rev/9c12707a59ff