1626188 - load_shader.h is not generated deterministically

Status: RESOLVED FIXED

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

Description

[Mike Hommey [:glandium]]

This should have shown up whenever changes to the build script generated these changes, but didn't because of bug 1626163. It only recently surfaced as a generated code difference because the variance in load_shader.h now induces some difference in the order some functions appear in the binary.
See https://firefoxci.taskcluster-artifacts.net/ObjR4GipRpCsvtRm_VED2w/0/public/diff.html

Comment 1

[Mike Hommey [:glandium]]

(incidentally, it also means possible cache misses when building with ccache or sccache)

Comment 2

[Intermittent Failures Robot]

15 failures in 5015 pushes (0.003 failures/push) were associated with this bug in the last 7 days.

Repository breakdown:

• mozilla-central: 15

Platform breakdown:

• diff: 15

For more details, see:
https://treeherder.mozilla.org/intermittent-failures.html#/bugdetails?bug=1626188&startday=2020-03-30&endday=2020-04-05&tree=all

Comment 3

[Andreea Pavel [:apavel]]

Jessie can you assign someone to take a look at this? There are15 failures in the last 7 days, all on diff opt.

Recent failure log: https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=295739518&repo=mozilla-central&lineNumber=6812

[task 2020-04-01T12:00:54.469Z] |########################## | 97% xpcom/ds/nsGkAtomConsts.h ETA: 0:00:00
[task 2020-04-01T12:00:54.472Z] |########################## | 97% xpcom/ds/nsGkAtomList.h ETA: 0:00:00
[task 2020-04-01T12:00:54.473Z] |########################## | 98% …parser/xpidl/xpidllex.py ETA: 0:00:00
[task 2020-04-01T12:00:54.474Z] |########################## | 98% …arser/xpidl/xpidlyacc.py ETA: 0:00:00
[task 2020-04-01T12:00:54.479Z] |########################## | 98% …lect/xptinfo/xptdata.cpp ETA: 0:00:00
[task 2020-04-01T12:00:54.480Z] |########################## | 99% …s/gtest/dafsa_test_1.inc ETA: 0:00:00
[task 2020-04-01T12:00:54.481Z] |########################## | 99% xpcom/xpcom-config.h ETA: 0:00:00
[task 2020-04-01T12:00:54.482Z] |########################## | 99% xpcom/xpcom-private.h ETA: 0:00:00
[task 2020-04-01T12:00:54.484Z] |###########################| 100% ETA: 0:00:00
[task 2020-04-01T12:00:54.484Z] |###########################| 100% ETA: 0:00:00
[task 2020-04-01T12:00:54.484Z] |###########################| 100% ETA: 0:00:00
[task 2020-04-01T12:00:54.484Z] |###########################| 100% ETA: 0:00:00
[task 2020-04-01T12:00:54.825Z] |###########################| 100% Time: 0:00:09
[task 2020-04-01T12:00:54.825Z] + report_error diff
[task 2020-04-01T12:00:54.825Z] + python3
[task 2020-04-01T12:00:54.835Z] TEST-UNEXPECTED-FAIL | target.generated-files.tar.gz/target.generated-files.tar/i686-unknown-linux-gnu/release/build/swgl-77a71763a275eb7b/out/load_shader.h differs. See the diff.html or diff.txt artifact
[task 2020-04-01T12:00:54.836Z] + fail
[task 2020-04-01T12:00:54.836Z] + exit 1
[taskcluster 2020-04-01 12:00:55.456Z] === Task Finished ===
[taskcluster 2020-04-01 12:00:56.147Z] Unsuccessful task run with exit code: 1 completed in 33.47 seconds

Please note there is a conversation on this here also: https://bugzilla.mozilla.org/show_bug.cgi?id=1596350

Comment 4

[Lee Salzman [:lsalzman]]

Attachment: Bug 1626188 - ensure shader keys are in sorted order. r?jimb

Comment 5

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/faa791f721b8
ensure shader keys are in sorted order. r=jimb

Comment 6

[Cristian Brindusan [:cbrindusan]]

https://hg.mozilla.org/mozilla-central/rev/faa791f721b8

Comment 7

[Intermittent Failures Robot]

14 failures in 4823 pushes (0.003 failures/push) were associated with this bug in the last 7 days.

Repository breakdown:

• mozilla-central: 14

Platform breakdown:

• diff: 14

For more details, see:
https://treeherder.mozilla.org/intermittent-failures.html#/bugdetails?bug=1626188&startday=2020-04-06&endday=2020-04-12&tree=all

Comment 8

[Alexandru Michis [:malexandru]]

Mike, there are recent diffoscope failures on Linux32g, looking similar to the ones this bug was tracking:
https://treeherder.mozilla.org/#/jobs?repo=mozilla-central&searchStr=diff&fromchange=2fd61eb5c69ce9ac806048a35c7a7a88bf4b9652&tochange=2203d818a3b488bb86ec28ef264d63fae053d3ea&selectedJob=298730900

Failure log: https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=298730900&repo=mozilla-central&lineNumber=6835

Should we reopen this bug or file a new one?

Comment 9

[Mike Hommey [:glandium]]

Please file a new one. Same component (webrender). Please Cc me.

Comment 10

[Mike Hommey [:glandium]]

And please mark that new bug as regressed by bug 1604615, which is the likely cause.