| Summary: | REGRESSION: [ EWS iOS wk2 ] css2.1/20110323/replaced-intrinsic-ratio-001.htm is a flakey text failure | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Product: | WebKit | Reporter: | Robert Jenner <jenner> | ||||||
| Component: | Layout and Rendering | Assignee: | Said Abou-Hallawa <sabouhallawa> | ||||||
| Status: | NEW --- | ||||||||
| Severity: | Normal | CC: | ap, bfulgham, sabouhallawa, simon.fraser, webkit-bot-watchers-bugzilla, webkit-bug-importer, zalan | ||||||
| Priority: | P2 | Keywords: | InRadar | ||||||
| Version: | WebKit Nightly Build | ||||||||
| Hardware: | Unspecified | ||||||||
| OS: | Unspecified | ||||||||
| See Also: |
https://bugs.webkit.org/show_bug.cgi?id=208577 https://bugs.webkit.org/show_bug.cgi?id=206254 |
||||||||
| Attachments: |
|
||||||||
|
Description
Robert Jenner
2021-04-02 11:59:15 PDT
Was able to reproduce the failure at iOS 14 Simulator Release ToT using the following test: run-webkit-tests --ios-simulator --iterations 1000 -f css2.1/20110323/replaced-intrinsic-ratio-001.htm The test is flakey, and I can't bisect to a specific regression point because there are a lot of revisions that did not get built that I would need in order narrow down where it started. The first failure occurred at r275303. The closest available build to that was r275285, and when I tested on that build, I could not reproduce the failure.. I suspect a change was introduced somewhere between r275303 and r275285 that caused the test to start flakey failing, but I'm not sure which one. Updated test expectations here to Pass Failure: https://trac.webkit.org/changeset/275435/webkit This test doesn't appear to have any potentially flaky content like timers, so this has to be a layout stability regression. This bug is similar to bug 208577 and bug 206254 where some elements in the page reference an external SVG image. In some case, the layout and the rendering of this SVG image is racy such that it gets laid out and rendered using the inline replaced box default size which is 300x150. Created attachment 429212 [details]
Patch
Created attachment 429214 [details]
Patch
Comment on attachment 429214 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=429214&action=review > LayoutTests/css2.1/20110323/replaced-intrinsic-ratio-001.htm:49 > + if (window.testRunner) > + testRunner.waitUntilDone(); > + window.addEventListener('load', () => { > + document.body.offsetHeight; > + if (window.testRunner) > + testRunner.notifyDone(); > + }); I would expect this to be a no-op, but perhaps there's a reason it isn't. WTR waits for the load event before it finishes the test, and it forces a layout before dumping the results. So what does this trigger that doesn't already happen? Comment on attachment 429214 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=429214&action=review > LayoutTests/ChangeLog:9 > + Sometimes there is a race between the layout and the rendering of the SVG Is this race correct behavior, or a WebKit bug? As Simon said, it is not super clear why the fix changes the behavior. |