fast/css/sticky/sticky-left.html Failure on Big Sur wk2 ARM64. First appearance 6/3/21 r278415. History: https://results.webkit.org/?suite=layout-tests&test=fast%2Fcss%2Fsticky%2Fsticky-left.html Unable to access diff due to 404 error.
<rdar://problem/79885861>
Reproduced test results on build r278417. run-webkit-tests --root 278417 fast/css/sticky/sticky-left.html --iterations 500 --exit-after-n-failures 2
13:44:16.831 21360 worker/4 Test marked as failed, ImageDiff crashed because of no data while reading stdout for the server process. 13:44:16.831 21360 worker/4 fast/css/sticky/sticky-left.html : ImageDiff produced stderr output: Error: test and reference images have different sizes. Test image is 800x700, reference image is 800x600 ImageDiff crashed 13:44:16.831 21360 worker/4 Test marked as failed, ImageDiff crashed because of a broken pipe when writing to stdin of the server process. 13:44:16.831 21360 worker/4 ref test mismatch did not produce an image diff. 13:44:16.832 21360 [10756/59269] fast/css/sticky/sticky-left.html failed unexpectedly (reference mismatch) 13:44:16.832 21360 worker/4 fast/css/sticky/sticky-left.html failed: 13:44:16.832 21360 worker/4 reference mismatch
Bisected test failure to range of 227412 (results not reproduced) to 227417 (reproduced results).
(In reply to Eric Hutchison from comment #4) > Bisected test failure to range of 227412 (results not reproduced) to 227417 > (reproduced results). So I revisited the bisection with Eric. It seems that the initial reproduction case was actually reproducing a crash in the test. We have not been able to reproduce this crashing image diff issue that the bot is seeing. But we found that the crash is only happening on the release builds and started on r278417. a debug build of r278417 does not seem to crash. I am uncertain how a version bump in r278417 could have anything to do with this, but our archived builds seem to point to this revision as starting some kind of problem.
Created attachment 432530 [details] Patch
Comment on attachment 432530 [details] Patch Clearing flags on attachment: 432530 Committed r279386 (239251@main): <https://commits.webkit.org/239251@main>
All reviewed patches have been landed. Closing bug.
reopen bug
Probably has something to do with this test, which runs just before sticky-left.html LayoutTests/fast/css/sticky/sticky-left-percentage.html: window.resizeTo(800, 700)
Or perhaps even more relevant is sticky-left-percentage-expected.html, which would be the one loaded right before. Note that it performs window.resizeTo(800, 700) from an inline script, before onload. I wonder if this confuses IPC.
Pull request: https://github.com/WebKit/WebKit/pull/316
Adding fast/css/sticky/sticky-left.html is a flaky image failure on [ Monterey Release wk2 arm64 ] as well. Description: fast/css/sticky/sticky-left.html The first failure that I see on the dashboard was 3/18/2022 at r291490 , but this is not relevant to causing this issue. This test seems to have been flaky since the start on Monterey. History: https://results.webkit.org/?suite=layout-tests&test=fast%2Fcss%2Fsticky%2Fsticky-left.html&platform=mac&version_name=Monterey&architecture=arm64&flavor=wk2&style=release&limit=50000 Image Diff: https://build.webkit.org/results/Apple-Monterey-Release-AppleSilicon-WK2-Tests/r249750@main%20(2764)/fast/css/sticky/sticky-left-diffs.html I have marked this test as flaky image failure while this issue is investigated.
Committed r293029 (249768@main): <https://commits.webkit.org/249768@main> Reviewed commits have been landed. Closing PR #316 and removing active labels.
Reopening since the landed change was just updating test expectations.