225126 2021-04-27 16:11:37 -0700 [Stress Tests EWS] WPT tests failing in `run-layout-tests-in-guard-malloc-stress-mode` 2021-04-28 17:05:01 -0700 1 1 1 Unclassified WebKit Tools / Tests WebKit Nightly Build Unspecified Unspecified NEW https://bugs.webkit.org/show_bug.cgi?id=224358 InRadar P2 Normal --- 1 ryanhaddad webkit-unassigned aakash_jain ap gsnedders jbedard rego rniwa webkit-bot-watchers-bugzilla webkit-bug-importer oldest_to_newest 1754827 0 ryanhaddad 2021-04-27 16:11:37 -0700 It appears that whenever stress tests EWS runs the `run-layout-tests-in-guard-malloc-stress-mode ` step for tests in the imported/w3c/web-platform-tests directory, the wptwk server isn't started, which causes the tests to fail Example build showing a WPT test passing in the `run-layout-tests-in-stress-mode` step, but failing in in the `run-layout-tests-in-guard-malloc-stress-mode` step: https://ews-build.webkit.org/#/builders/62/builds/1853 Test result showing the test failing with a render tree dump for about:blank: https://ews-build.s3-us-west-2.amazonaws.com/macOS-Release-WK2-Stress-Tests-EWS/r427118-1853-guard-malloc/results.html 1754828 1 webkit-bug-importer 2021-04-27 16:12:24 -0700 <rdar://problem/77236114> 1754848 2 ryanhaddad 2021-04-27 16:50:47 -0700 The `run-layout-tests-in-stress-mode` test run is indeed starting wptwk, as seen in the logs (and the fact that the test passes in that mode) https://ews-build.webkit.org/#/builders/62/builds/1853/steps/14/logs/stdio The only difference for the `run-layout-tests-in-guard-malloc-stress-mode ` step is the addition of `--guard-malloc`, so maybe something is broken in `run-webkit-tests` itself? 1754858 3 jbedard 2021-04-27 17:04:46 -0700 (In reply to Ryan Haddad from comment #2) > The `run-layout-tests-in-stress-mode` test run is indeed starting wptwk, as > seen in the logs (and the fact that the test passes in that mode) > https://ews-build.webkit.org/#/builders/62/builds/1853/steps/14/logs/stdio > > The only difference for the `run-layout-tests-in-guard-malloc-stress-mode ` > step is the addition of `--guard-malloc`, so maybe something is broken in > `run-webkit-tests` itself? Very strongly suspect that is the case. It's also probably a very easy fix. I'll take a look tomorrow. 1755106 4 rniwa 2021-04-28 11:54:37 -0700 I've seen regular layout test step reproducing the same issue too. 1755229 5 jbedard 2021-04-28 16:40:55 -0700 This seems to be about guard-malloc specifically, not a general problem with all invocations of run-webkit-tests. 1755238 6 jbedard 2021-04-28 17:05:01 -0700 We are starting the WPT server in guard-malloc mode, seems like the web-platforms tests may be genuinely timing out.