Bug 155092 – js/arraybuffer-wrappers.html is a flaky time out

• Status: NEW
• Component: JavaScriptCore

[Alexey Proskuryakov]

js/arraybuffer-wrappers.html is a new test that times out very frequently (and when it doesn't, it runs fast).

https://webkit-test-results.webkit.org/dashboards/flakiness_dashboard.html#showAllRuns=true&tests=js%2Farraybuffer-wrappers.html

[Keith Miller]

Is the test still flakey? The link appears to be dead and the test doesn't show up on https://webkit-test-results.webkit.org/dashboards/flakiness_dashboard.html

[Keith Miller]

Wait, nevermind. My internet was acting up and the page was not loading properly.

[Ryan Haddad]

Marked test as flaky in <http://trac.webkit.org/projects/webkit/changeset/197696>

[Alexey Proskuryakov]

Keith, does this look more like a test bug, or a JSC bug?

[Keith Miller]

(In reply to comment #4)
> Keith, does this look more like a test bug, or a JSC bug?

This looks like a JSC bug. I would expect this test to finish in < 20ms and it definitely should not take more than 1s.

[Radar WebKit Bug Importer]

<rdar://problem/25130361>

[Alexey Proskuryakov]

*** Bug 155043 has been marked as a duplicate of this bug. ***

[Alexey Proskuryakov]

Still happening very frequently (also on iOS and Gtk, where we don't have the expectation yet).

Keith, would you mind taking a look?

[Matt Lewis]

Marked this test as flaky on iOS for now:
https://trac.webkit.org/changeset/221687/webkit

[Keith Miller]

(In reply to Alexey Proskuryakov from comment #8)
> Still happening very frequently (also on iOS and Gtk, where we don't have
> the expectation yet).
> 
> Keith, would you mind taking a look?

Oh, I looked into this a bit. It seems like it's some kind of GC deadlock based on the spindumps on the bots. I'm not sure what causes it or how to reproduce it. I tried running the test (and the 10 tests leading up to it) on all the cores on my machine repeatedly but the issue never reproduced.

[Alexey Proskuryakov]

Is there any instrumentation that can be added to get more information when this happens on the bots?

[Keith Miller]

(In reply to Alexey Proskuryakov from comment #11)
> Is there any instrumentation that can be added to get more information when
> this happens on the bots?

I can't think of anything off the top of my head but maybe @pizlo has some ideas. It's also possible that this isn't a GC deadlock but that's just my current hypothesis.