1296345 - WebGL2 failure on a new driver, with "Error during native OpenGL init" on Windows

Status: RESOLVED INCOMPLETE

(Core :: Graphics: CanvasWebGL, defect, P3)

Description

[Milan Sreckovic [:milan] (needinfo for best results)]

Windows 10, Intel HD Graphics 530, driver 20.19.15.4474 from June 2016, and we  get the "WebGL2 Renderer: WebGL creation failed: * Error during native OpenGL init".

Given our whitelisting strategy for GL/Windows/WebGL2, we should understand why.

See bug 1295215 comment 5 for more details.

Comment 1

[Benjamin Smedberg]

Attachment: bsmedberg's DxDiag.txt

From bug 1294529, I see this on Windows 10 with a Surface Book and I'm attaching an edited summary of information from DxDiag.

If it matters, I'm using an external 4k monitor attached via displayport, and I have the laptop itself closed so I'm not using that screen currently.

Comment 2

[Stebs]

Attachment: about:support.txt from affected new profile

Not really sure if I experience really the same bug, but it could be related and I found no better bug.

On my linux nightly 52.0a1 (2016-10-09) (64-bit), WebGL 2 AND 1 are no longer working on any page.

webglreport.com says: This browser supports WebGL 1, but it is disabled or unavailable. (same with WebGl2)

Browser console: Error: WebGL: Failed to create WebGL context: WebGL creation failed: 
* Error during native OpenGL init.
...
* Exhausted GL driver caps.
* Exhausted GL driver options.

Yet, about:support looks normal:
WebGL Renderer	NVIDIA Corporation -- GeForce GTX 650 Ti BOOST/PCIe/SSE2
WebGL2 Renderer	NVIDIA Corporation -- GeForce GTX 650 Ti BOOST/PCIe/SSE2
Driver Version	4.5.0 NVIDIA 370.28

glxinfo looks normal, 3d works, FF49: WebGL 1 works, opengl-compositing seems to also work?, also tested with a new profile.

Imho this regression is fairly recent, going to pinpoint it with mozregression...

Comment 3

[Stebs]

Last good revision: 3e9a2031152fa07b088d0cb5e168eb53a2c882c0
First bad revision: c838d2546cadd65bf8d5579db20a268c8b6e4b87
Pushlog:
https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=3e9a2031152fa07b088d0cb5e168eb53a2c882c0&tochange=c838d2546cadd65bf8d5579db20a268c8b6e4b87

INFO: Looks like the following bug has the changes which introduced the regression:
https://bugzilla.mozilla.org/show_bug.cgi?id=1289718

Comment 4

[Stebs]

So seccomp-bpf (linux) related and not with this bug, going to post it there...

Comment 6

[Gian-Carlo Pascutto [:gcp]]

Milan, this bug was originally reported against Windows. The bug you duped is Linux only.

Comment 7

[Milan Sreckovic [:milan] (needinfo for best results)]

The original patch broke multiple platforms (see comment 3), but the fix in bug 1308851 was for Linux only.  Got it.

Comment 8

[Gian-Carlo Pascutto [:gcp]]

Well, you created this bug 2016-08-10 for Windows. My Linux breakage landed 2016-10-06. Consequently, the original problem on Windows must've been something else.

Comment 9

[Milan Sreckovic [:milan] (needinfo for best results)]

Clarifying this bug is Windows only.  We can ignore comment 3 in that context.
Chris, does your about:support still show WebGL failure?

Comment 10

[Chris Peterson [:cpeterson]]

(In reply to Milan Sreckovic [:milan] from comment #9)
> Clarifying this bug is Windows only.  We can ignore comment 3 in that
> context.
> Chris, does your about:support still show WebGL failure?

Did you intend to needinfo bsmedberg? He saw this error in comment 1 whereas I have not seen this error.

That said, I have had WebGL2 issues and my about:support shows bug 1242659:

> failures: CP+[GFX1]: Potential driver version mismatch ignored due to missing DLLs 0.0.0.0 and 0.0.0.0

Comment 11

[Milan Sreckovic [:milan] (needinfo for best results)]

Right, I know about the mismatched driver.  But, yes, I did mean you (Chris), given this in your about:support mentioned in bug 1295215 comment 5:

WebGL2 Renderer: WebGL creation failed: * Error during native OpenGL init. * Error during native OpenGL init. * Error during native OpenGL init. * Error during native OpenGL init. * Error during native OpenGL init. * Exhausted GL driver options.

Sounds like that one is gone for you, let's see if Benjamin still sees it.

Comment 12

[Chris Peterson [:cpeterson]]

(In reply to Milan Sreckovic [:milan] from comment #11)
> Right, I know about the mismatched driver.  But, yes, I did mean you
> (Chris), given this in your about:support mentioned in bug 1295215 comment 5:
> 
> WebGL2 Renderer: WebGL creation failed: * Error during native OpenGL init. *
> Error during native OpenGL init. * Error during native OpenGL init. * Error
> during native OpenGL init. * Error during native OpenGL init. * Exhausted GL
> driver options.
> 
> Sounds like that one is gone for you, let's see if Benjamin still sees it.

I forgot about that. You are correct: I no longer see those "Error during native OpenGL init" errors or glitches.

Comment 13

[Inigo]

I am having this error in Firefox 51, Windows 10, GeForce 980 GTX, latest drivers. Any help is welcome

Comment 14

[Milan Sreckovic [:milan] (needinfo for best results)]

What does the graphics section of about:support contain?

Comment 15

[Rob Thijssen [:grenade (EET/UTC+0300)]]

This exception also occurs on TaskCluster Windows 10 tests causing failures in reftest-no-accel, chunk 4. These tests have not previously run regularly on Windows 10 in CI/automation.

test runs:
https://treeherder.mozilla.org/#/jobs?repo=try&revision=009a98aea54aa6a9ec573b0bb791396a9a926ac0&group_state=expanded&exclusion_profile=false&filter-searchStr=Ru4

fail log:
https://treeherder.mozilla.org/logviewer.html#?job_id=96194472&repo=try&lineNumber=10670

Comment 16

[Joel Maher ( :jmaher ) (UTC -8)]

are we running on the GPU instances?  win10 is on vm and reftest requires the GPU instances.

Comment 17

[Rob Thijssen [:grenade (EET/UTC+0300)]]

i spoke too soon. Ru4 failed on Win 10 while the test was running on an ec2 instance without hardware acceleration (c4.2xlarge). When this was corrected to use an instance with GPU (g2.2xlarge), the test succeeded, the error was not thrown.

See https://treeherder.mozilla.org/#/jobs?repo=try&revision=009a98aea54aa6a9ec573b0bb791396a9a926ac0&group_state=expanded&exclusion_profile=false&filter-searchStr=Ru4 for the runs against both instance types.

Comment 18

[Milan Sreckovic [:milan] (needinfo for best results)]

Peter, can somebody take a look?

Comment 19

[Kelsey Gilbert [:jgilbert]]

(In reply to Rob Thijssen (:grenade - EEST) from comment #17)
> i spoke too soon. Ru4 failed on Win 10 while the test was running on an ec2
> instance without hardware acceleration (c4.2xlarge). When this was corrected
> to use an instance with GPU (g2.2xlarge), the test succeeded, the error was
> not thrown.
> 
> See
> https://treeherder.mozilla.org/#/
> jobs?repo=try&revision=009a98aea54aa6a9ec573b0bb791396a9a926ac0&group_state=e
> xpanded&exclusion_profile=false&filter-searchStr=Ru4 for the runs against
> both instance types.

In that case, please file a different bug for this.

Comment 20

[Rob Thijssen [:grenade (EET/UTC+0300)]]

apologies, i wasn't clear in my correction. there is no issue. the test passes correctly on gpu instances.

Comment 21

[Joel Maher ( :jmaher ) (UTC -8)]

I vote for closing this as wontfix