1844627 - (webgpu-v1-windows-blocklist-intermittent) `WebGPU is disabled by blocklist` in Windows CI

Status: RESOLVED FIXED

(Core :: Graphics: WebGPU, defect, P1)

Description

[Erich Gubler [:ErichDonGubler]]

There appears to be an odd variance in Windows CI environments where WebGPU does not run, and the diagnostic emitted blames our blocklist:

EXCEPTION: WebGPU device failed to initialize with NotSupportedError "WebGPU is disabled by blocklist."; not retrying

When this error is encountered, all WebGPU CTS tests in the job fail. EDIT: Actually, it appears that test runs can recover from this. 🤔⁉️

Comment 3

[Intermittent Failures Robot]

2 failures in 3520 pushes (0.001 failures/push) were associated with this bug in the last 7 days.

Repository breakdown:

• autoland: 1
• mozilla-central: 1

Platform and build breakdown:

• windows11-32-2009-shippable-qr: 2
  • opt: 2

For more details, see:
https://treeherder.mozilla.org/intermittent-failures/bugdetails?bug=1844627&startday=2023-07-17&endday=2023-07-23&tree=all

Comment 5

[BugBot [:suhaib / :marco/ :calixte]]

The severity field for this bug is set to S3. However, the following bug duplicate has higher severity:

• Bug 1851402: S2

:jimb, could you consider increasing the severity of this bug to S2?

For more information, please visit BugBot documentation.

Comment 6

[Jim Blandy :jimb]

Raising the priority on this to P1: "CTS impact widespread enough that it obscures other issues", based on Matrix discussion with EDG.

Leaving the severity as S3, per WebGPU project policy.

Comment 8

[Erich Gubler [:ErichDonGubler]]

Attachment: WIP: Bug 1844627 - test(webgpu): set `dom.webgpu.enabled` to `true` via `mozharness`, instead of WPT r=#webgpu-reviewers!

Comment 9

[Erich Gubler [:ErichDonGubler]]

Not sure that I have a strong lead on resolving this yet, but I did want to note an interesting observation: It appears we may legitimately have different graphics adapters. In most cases (where this issue isn't happening), a log message is occasionally printed by glean with the following text:

[WARN  glean_core::error_recording] gfx.adapter.primary.driver_files: Value length 811 exceeds maximum of 100

However, in tests that do run into this issue, there is no such thing, implying that the set of driver files is small enough to fit within the limit.

Comment 10

[Jim Blandy :jimb]

One way to debug this would be to start with that error message:

EXCEPTION: WebGPU device failed to initialize with NotSupportedError "WebGPU is disabled by blocklist."; not retrying

and then work backwards to find the sufficient conditions for getting that message, and add logging at each point so we can see in detail how we're arriving at that message. For example, it sure would be nice to know which blocklist entry we're hitting.

This would entail repeated submissions to the try server, so it would need to be a background project.

Comment 11

[Intermittent Failures Robot]

2 failures in 3995 pushes (0.001 failures/push) were associated with this bug in the last 7 days.

Repository breakdown:

• autoland: 2

Platform and build breakdown:

• windows11-64-2009-qr: 2
  • debug: 2

For more details, see:
https://treeherder.mozilla.org/intermittent-failures/bugdetails?bug=1844627&startday=2024-05-20&endday=2024-05-26&tree=all

Comment 12

[Erich Gubler [:ErichDonGubler]]

A patch stack now exists against dependency bug 1899536. Once it's landed, this should be resolved. 🙌🏻

Comment 13

[Intermittent Failures Robot]

1 failures in 4202 pushes (0.0 failures/push) were associated with this bug in the last 7 days.

Repository breakdown:

• try: 1

Platform and build breakdown:

• windows11-64-2009-qr: 1
  • opt: 1

For more details, see:
https://treeherder.mozilla.org/intermittent-failures/bugdetails?bug=1844627&startday=2024-05-27&endday=2024-06-02&tree=all