Today I upgraded the hhvm variant of Quibble docker image (used by MW Jenkins jobs) from Debian Jessie to Debian Stretch (dbf2347bbc049b1, d48921fb8f30e).
The job for MW plain and individual extensions was passing still, but the job for shared-gate extensions (which includes Navigation Timing, and also runs on MW core commits) started failing, like so:
✔ Load topics
ext.flow.dm mw.flow.dm.Board
✔ Create board
Thanks thank
✔ thanked cookie
✔ gets user gender
ext.navigationTiming
✔ Basic
✔ First view
✔ Repeat view
✔ Reloaded view
✔ Without Navigation Timing API
✔ Oversample config and activation
✔ emitOversampleNavigationTiming tests
✔ onMwLoadEnd - plain
✔ onMwLoadEnd - controlled
✔ Oversample Geo integration tests
✔ Optional APIs
✔ makeResourceTimingEvent
✔ emitTopImageResourceTiming
✔ emitCentralNoticeTiming
10 02 2019 04:05:24.902:WARN [HeadlessChrome 0.0.0 (Linux 0.0.0)]: Disconnected (1 times), because no message in 60000 ms.
HeadlessChrome 0.0.0 (Linux 0.0.0) ERROR
Disconnected, because no message in 60000 ms.
Finished in 1 min 0.319 secs / 0.292 secs @ 04:05:24 GMT 0000 (UTC)
10 02 2019 04:05:24.905:DEBUG [karma]: Run complete, exiting.
10 02 2019 04:05:24.906:DEBUG [launcher]: Disconnecting all browsers
10 02 2019 04:05:24.926:DEBUG [launcher]: Process Chrome exited with code 0
10 02 2019 04:05:24.927:DEBUG [temp-dir]: Cleaning temp dir /tmp/karma-74277250
10 02 2019 04:05:24.933:DEBUG [launcher]: Finished all browsers
Warning: Task "karma:main" failed. Use --force to continue.
Aborted due to warnings.FAILED TESTS:
ext.navigationTiming
✖ emitCpuBenchmark
HeadlessChrome 0.0.0 (Linux 0.0.0)
Test took longer than 60000ms; test timed out.It basically just times out mid-way the "ext.navigationTiming" test suite, at the some point consistently so. By looking at the console output of an older build of that same Jenkins job I found that the test that normally runs after this one is emitCpuBenchmark.
Disabling that made the build pass again.