[Webkit-unassigned] [Bug 189063] REGRESSION(r235408): GTK bots exiting early

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=189063

--- Comment #11 from Claudio Saavedra <csaavedra at igalia.com> ---
I debugged this a bit today, have some findings that might shed some light.

- This only happens if you run a series of tests sharing the DumpRendererTree. If you use --run-singly, the tests pass. So it probably has to be with the remainder state of DumpRendererTree.

- I noticed that there's a 20_s in the AsyncTask that is added for the world leaks in the offending commit. I'm not familiar with this but I'm guessing that this is the timeout for the task. The timeout for the tests is 5 s, AFAIR, so it's probable that the world leaks task is timing out. I increased the tests timeout when running them locally to 20s, and then the tests pass. Similarly, if I change the 20_s to 5_s, the tests also pass.

- But what I don't understand is why this task is being called at all? Isn't it so that --world-leaks needs to be called explicitly for the check to run? What I see is that this is done unconditionally in willDestroyWebView(). Is this intended?

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.webkit.org/pipermail/webkit-unassigned/attachments/20180904/7649ade2/attachment.html>