Bug 171832 – [iOS Simulator] LayoutTest/fast/multicol/newmulticol/multicol-inside-multicol.html is flaky timeout

• Status: NEW
• Component: New Bugs

[Matt Lewis]

The test fast/multicol/newmulticol/multicol-inside-multicol.html is a flaky Timeout

The test also seems to be a flaky image failure.

History:
https://webkit-test-results.webkit.org/dashboards/flakiness_dashboard.html#showAllRuns=true&tests=fast%2Fmulticol%2Fnewmulticol%2Fmulticol-inside-multicol.html

Timeout Build:
https://build.webkit.org/builders/Apple%20iOS%2010%20Simulator%20Debug%20WK2%20(Tests)/builds/1175

Image failure Build:
https://build.webkit.org/builders/Apple%20iOS%2010%20Simulator%20Debug%20WK2%20(Tests)/builds/1046

[Alexey Proskuryakov]

Looks like when this test times out, it does that super quickly (in a couple seconds), which should be impossible. This looks like some terrible infrastructure bug.


13:38:58.288 57493 worker/6 fast/multicol/newmulticol/multicol-clip-rounded-corners.html failed:
13:38:58.289 57493 worker/6  reference mismatch
13:38:59.335 57493 Wait on notifyDone timed out, process com.apple.WebKit.WebContent.Development pid = 57533
13:38:59.536 57493 worker/6 fast/multicol/newmulticol/multicol-inside-multicol.html output stderr lines:
13:38:59.537 57493   Wait on notifyDone timed out, process com.apple.WebKit.WebContent.Development pid = 57533
13:38:59.539 56025 [15908/41695] fast/multicol/newmulticol/multicol-inside-multicol.html failed unexpectedly (test timed out)

[Alexey Proskuryakov]

Ryan, Matt, is this reproducible by any chance? Running tests in the same order as in a failing bot run would probably be the way to go.

[Jonathan Bedard]

(In reply to Alexey Proskuryakov from comment #1)
> Looks like when this test times out, it does that super quickly (in a couple
> seconds), which should be impossible. This looks like some terrible
> infrastructure bug.
> 
> 
> 13:38:58.288 57493 worker/6
> fast/multicol/newmulticol/multicol-clip-rounded-corners.html failed:
> 13:38:58.289 57493 worker/6  reference mismatch
> 13:38:59.335 57493 Wait on notifyDone timed out, process
> com.apple.WebKit.WebContent.Development pid = 57533
> 13:38:59.536 57493 worker/6
> fast/multicol/newmulticol/multicol-inside-multicol.html output stderr lines:
> 13:38:59.537 57493   Wait on notifyDone timed out, process
> com.apple.WebKit.WebContent.Development pid = 57533
> 13:38:59.539 56025 [15908/41695]
> fast/multicol/newmulticol/multicol-inside-multicol.html failed unexpectedly
> (test timed out)

Without investigating, I suspect that this test is deliberately changing it's timeout.  That's almost exactly a second between the reference mis-match line and the timeout.  That doesn't seem unintended.

[Alexey Proskuryakov]

This test is synchronous, and doesn't use timeouts.

[Radar WebKit Bug Importer]

<rdar://problem/32173015>

[Alexey Proskuryakov]

I couldn't reproduce after some trying (running just multicol tests, running tests in the same order as a failing run, enabling GuardMalloc).

We may be able to add assertions to catch this, e.g. confirming that notifyDone timer isn't running when a new test begins.

[Alexey Proskuryakov]

<https://webkit-test-results.webkit.org/dashboards/flakiness_dashboard.html#showAllRuns=true&tests=fast%2Fmulticol%2Fnewmulticol%2Fmulticol-inside-multicol.html> suggests that this hasn't been happening recently.