Closed Bug 1429995 Opened 6 years ago Closed 6 years ago

[Ubuntu 16.04] TEST-UNEXPECTED-PASS | dom/canvas/test/webgl-mochitest/ensure-exts/test_WEBGL_compressed_texture_s3tc.html | fail-if condition in manifest - We expected at least one failure

Categories

(Core :: Graphics: CanvasWebGL, defect, P5)

defect

Tracking

()

RESOLVED FIXED
mozilla59
Tracking Status
firefox59 --- fixed

People

(Reporter: glandium, Assigned: glandium)

References

Details

(Whiteboard: [gfx-noted][stockwell fixed:other])

Attachments

(1 file)

+++ This bug was initially created as a clone of Bug #1429934 +++

Filed by: shindli [at] mozilla.com

https://treeherder.mozilla.org/logviewer.html#?job_id=155760208&repo=autoland

https://queue.taskcluster.net/v1/task/fA6cGoPnT32bDG3OWSRncw/runs/0/artifacts/public/logs/live_backing.log

---- Additional information below:

This is permafail when running on up-to-date Ubuntu 16.04. CI is currently running on an image built 13 days ago, something changed since then, and updating the image triggers the error. IOW, this is preventing any change happening on those docker images.

This can be reproduced on automation by doing the following:

Run `mach try fuzzy --full`


Take a try build that runs gl3 (take a green one), such as:
https://treeherder.mozilla.org/#/jobs?repo=try&revision=fc6c8fd82a6951671e243128c6a4ae486b646926&selectedJob=155794739

Select the gl3 job. In the bottom-left section, click on the … and select One-click loaner.

Be sure to be logged-in, and click "Create task". Wait for the task to start and click on the "shell" button when it comes up.

In the wizard that shows up in the shell session, select "4" (Exit)

When you get a shell prompt, run `apt-get update` then `apt-get -y upgrade`

Then `run-wizard`, and on its prompt, choose "1" (Resume task).
Blocks: 1429996
The unexpected pass seems suspicious, because we do indeed fail the extension checks as expected? Maybe I'm missing something to the test.

> [task 2018-01-11T22:56:53.816Z] 22:56:53     INFO - Buffered messages logged at 22:56:53
> [task 2018-01-11T22:56:53.819Z] 22:56:53     INFO - TEST-FAIL | dom/canvas/test/webgl-mochitest/ensure-exts/test_WEBGL_compressed_texture_s3tc.html | Should have webgl extension WEBGL_compressed_texture_s3tc. 
> [task 2018-01-11T22:56:53.820Z] 22:56:53     INFO - TEST-FAIL | dom/canvas/test/webgl-mochitest/ensure-exts/test_WEBGL_compressed_texture_s3tc.html | Should have webgl2 extension WEBGL_compressed_texture_s3tc. 
> [task 2018-01-11T22:56:53.822Z] 22:56:53     INFO - Buffered messages finished
> [task 2018-01-11T22:56:53.823Z] 22:56:53     INFO - TEST-UNEXPECTED-PASS | dom/canvas/test/webgl-mochitest/ensure-exts/test_WEBGL_compressed_texture_s3tc.html | fail-if condition in manifest - We expected at least one failure
> [task 2018-01-11T22:56:53.824Z] 22:56:53     INFO - TEST-INFO | expected FAIL
Flags: needinfo?(jgilbert)
Whiteboard: [gfx-noted]
Is this about a new failure here after updating the os/drivers? If so, this is somewhat expected. We should expect to need to regenerate our expected failure list after any os/driver changes.

I can help with this, if that's what's going on here.
Flags: needinfo?(jgilbert) → needinfo?(mh+mozilla)
You tell me. Comment 1 shows one thing that is not making sense in the log.
Flags: needinfo?(mh+mozilla) → needinfo?(jgilbert)
I believe bug 1337674 regressed behavior for fail-if unexpected passes, which is now causing 'PASS' in fail-if=<true> test runs to claim any ok(true, ...) assertion is TEST-FAIL.

I suspect that s3tc is supported on post-update linux machines, and so if we are updating all linux machines, we should remove linux from the fail-if list here, and that doing so will resolve this perma-failure.
Flags: needinfo?(jgilbert)
I confirmed WEBGL_compressed_texture_s3tc is supported in the new images, and not supported in the old images.
See Also: → 1430946
GL_EXT_compressed_texture_s3tc was seen supported by the driver in both new and old images, but we only expose the webgl extension in the new image.

We block s3tc on llvmpipe and some mac drivers:
https://dxr.mozilla.org/mozilla-central/rev/e4107773cffb1baefd5446666fce22c4d6eb0517/gfx/gl/GLContext.cpp#1749

:glandium saw both drivers as llvmpipe, so we're probably not detecting llvmpipe properly anymore.
That said, newer llvmpipe probably fixes most of the issues we were having previously, so no emergency there.
llvmpipe 4.0 vs. 5.0, FYI.
Assignee: nobody → mh+mozilla
Comment on attachment 8943094 [details]
Bug 1429995 - Mark test_WEBGL_compressed_texture_s3tc.html as not failing on Linux anymore.

https://reviewboard.mozilla.org/r/213388/#review219110

::: taskcluster/docker/desktop1604-test/Dockerfile:4
(Diff revision 1)
>  FROM          ubuntu:16.04
>  MAINTAINER    Joel Maher <joel.maher@gmail.com>
>  
> +

Spurious added newline?
Attachment #8943094 - Flags: review?(jgilbert) → review+
Comment on attachment 8943094 [details]
Bug 1429995 - Mark test_WEBGL_compressed_texture_s3tc.html as not failing on Linux anymore.

https://reviewboard.mozilla.org/r/213388/#review219114
Pushed by mh@glandium.org:
https://hg.mozilla.org/integration/autoland/rev/05028d9a5e62
Mark test_WEBGL_compressed_texture_s3tc.html as not failing on Linux anymore. r=jgilbert
https://hg.mozilla.org/mozilla-central/rev/05028d9a5e62
Status: NEW → RESOLVED
Closed: 6 years ago
Resolution: --- → FIXED
Target Milestone: --- → mozilla59
Whiteboard: [gfx-noted] → [gfx-noted][stockwell fixed:other]
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: