Closed Bug 1878389 Opened 2 years ago Closed 2 years ago

Intermittent layout/reftests/pagination/table-nested-1308876-1.xhtml == layout/reftests/pagination/table-nested-1308876-1-ref.html | single tracking bug

Categories

(Core :: Layout, defect)

defect

Tracking

()

RESOLVED DUPLICATE of bug 1878425

People

(Reporter: imoraru, Unassigned)

Details

(Keywords: intermittent-failure, intermittent-testcase)

Failure line does not appear in Treeherder.

''' INFO '''

Images are identical.

''' INFO '''

The log shows

INFO | REFTEST fuzzy test (23, 1) <= (0, 0) <= (23, 1)
[task 2024-02-02T18:50:15.614Z] 18:50:15     INFO - REFTEST TEST-UNEXPECTED-PASS

The test has a strictly-must-be-fuzzy macOS annotation, but it seems the fuzzy-failure didn't happen in this case.

Failure line does not appear in Treeherder.

That's because this log has an extremely large number of reftest fuzzy-failures. This particular failure just happens to be the first one reported in the log.

Here's the reftest-analyzer view:
http://hg.mozilla.org/mozilla-central/raw-file/tip/layout/tools/reftest/reftest-analyzer.xhtml#logurl=https://firefoxci.taskcluster-artifacts.net/X9u7fC_5QfWNC5KvDSd51g/0/public/logs/live_backing.log

There are other test failures in there that look more interesting/real than this spurious pass. In particular, there are some SVG-as-image test failures where the testcase's image is rendered at a different size than the reference case, e.g.:
layout/reftests/svg/as-image/context-stroke-08.html and img-foreignObject-iframe-1a.html and layout/reftests/svg/as-image/img-foreignObject-embed-1.html

and there are a lot of test failures in there where the reference case looks like it's shifted 1px (or a fraction of a pixel) with respect to the testcase, particularly in layout/reftests/pixel-rounding/.

I wonder if we're somehow being rendered at a higher-than-expected display resolution, which is responsible both for the unexpected-pass in comment 0 - comment 1 and also responsible for these unexpected failures from slight shifts and size-mismatches?

Depends on: 1878425

I filed bug 1878425 as an umbrella bug for this issue (since there's at least one other bug/failure that looks to be the same issue).

I guess we might as well dupe these to the umbrella bug, as emilio did in bug 1878707.

Status: NEW → RESOLVED
Closed: 2 years ago
No longer depends on: 1878425
Duplicate of bug: 1878425
Resolution: --- → DUPLICATE
You need to log in before you can comment on or make changes to this bug.