1887909 - Update WGPU to upstream (week of 2024-03-26)

Status: RESOLVED FIXED

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

Comment 1

[Erich Gubler [:ErichDonGubler] (he/him)]

I'll take on this one, since :teoxoy is out.

Comment 2

[Erich Gubler [:ErichDonGubler] (he/him)]

Attachment: Bug 1887909 - update WGPU to 0c5bebca514eb06d9387f87666c1c658f3f673b4 (`trunk` as of 2024-04-02) r=#webgpu-reviewers!

Comment 3

[Erich Gubler [:ErichDonGubler] (he/him)]

Attachment: Bug 1887909 - test(webgpu): update lone subtest `TIMEOUT` or `NOTRUN` `expectations` to be `[TIMEOUT, NOTRUN]` r=#webgpu-reviewers!

Comment 4

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/5bdc4db1e756
Update WGPU to b34219ca214caea2af86dbb64a2a2b613ea5198c (`trunk` as of 2024-03-26) r=webgpu-reviewers,supply-chain-reviewers,nical
https://hg.mozilla.org/integration/autoland/rev/18a3f41af0ed
test(webgpu): update lone subtest `TIMEOUT` or `NOTRUN` `expectations` to be `[TIMEOUT, NOTRUN]` r=webgpu-reviewers,nical,teoxoy

Comment 5

[Noemi Erli[:noemi_erli]]

Backed out 2 changesets (Bug 1887909) for causing failures in test_command_buffer_creation.html CLOSED TREE

Push with failures: https://treeherder.mozilla.org/jobs?repo=autoland&group_state=expanded&selectedTaskRun=Ku4HFrrESLm3FWSV2ciBpg.0&resultStatus=testfailed%2Cbusted%2Cexception%2Cretry%2Cusercancel&revision=18a3f41af0eda3c87aa96bd2f542583fd9148228

Log: https://treeherder.mozilla.org/logviewer?job_id=452864938&repo=autoland&lineNumber=1896

Backout: https://hg.mozilla.org/integration/autoland/rev/d87381ca434918941fe3670ad182962d1933afb0

Comment 6

[Erich Gubler [:ErichDonGubler] (he/him)]

Investigating posthaste!

Comment 7

[Jim Blandy :jimb]

In the logs, dom/webgpu/mochitest/test_command_buffer_creation.html times out. This is because the mochitest that ran immediately before it, dom/webgpu/mochitest/test_buffer_mapping_invalid_device.html, has hung the GPU process trying to clean up device resources.

Specifically, wgpu_core::device::Device::maintain takes a &SnatchGuard argument, meaning that its caller must have a read guard on the device's snatch lock. However, mantain calls Device::release_gpu_resources, which calls Buffer::destroy, which tries to acquire a write guard on the snatch lock. Probably a similar problem would occur with Texture::destroy.

Comment 8

[Jim Blandy :jimb]

Filed upstream: https://github.com/gfx-rs/wgpu/issues/5471

Comment 9

[Erich Gubler [:ErichDonGubler] (he/him)]

Attachment: Bug 1887909 - test(webgpu): update current expected outcomes for CTS tests r=#webgpu-reviewers!

Comment 10

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/3ae3d83aae40
test(webgpu): update lone subtest `TIMEOUT` or `NOTRUN` `expectations` to be `[TIMEOUT, NOTRUN]` r=webgpu-reviewers,nical,teoxoy
https://hg.mozilla.org/integration/autoland/rev/f6e62f23a6dc
test(webgpu): update current expected outcomes for CTS tests r=webgpu-reviewers,nical
https://hg.mozilla.org/integration/autoland/rev/4fbd7bf93c96
update WGPU to 0c5bebca514eb06d9387f87666c1c658f3f673b4 (`trunk` as of 2024-04-02) r=webgpu-reviewers,supply-chain-reviewers,nical

Comment 11

[pstanciu]

https://hg.mozilla.org/mozilla-central/rev/3ae3d83aae40
https://hg.mozilla.org/mozilla-central/rev/f6e62f23a6dc
https://hg.mozilla.org/mozilla-central/rev/4fbd7bf93c96