[Webkit-unassigned] [Bug 126432] REGRESSION (r161195): Acid2 regression tests frequently fail

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=126432





--- Comment #14 from Alexey Proskuryakov <ap at webkit.org>  2014-01-07 10:40:16 PST ---
It is slightly easier to reproduce with WebKit2, but WebKit1 is also affected. I believe you now saw that on flakiness dashboard (but there are also failures on http versions of the tests, so it's more than "once").

On a Mac Pro, I could reproduce the WebKit1 failure very reliably by running this command in two terminal windows at once to create high CPU load. Likely easier on a notebook:

run-webkit-tests fast/css/acid2-pixel.html --repeat-each 10000 -f

So yes, there are many tasks in WebKit development that are much harder than reproducing a failure on an isolated test, no need to be snarky about it.

> > - change the delay from 0 to some positive value.
> 
> Which delay?

Your patch made object load failing asynchronous, i.e. use a zero delay timer. You could change the timer to a non-zero one to simulate stress conditions. This is a fairly common technique.

> > 1. In this case, there are 4 flaky tests, each with a much higher rate of failure.
> 
> The bug mentions two copies of acid2. What are the other two?

Those are http versions of the same tests.

You could see that by looking at <http://trac.webkit.org/r161371> that's mentioned in comment 8. Another way to find put what's going on is to check the flakiness dashboard for "acid2": <http://webkit-test-results.appspot.com/dashboards/flakiness_dashboard.html#tests=acid2&revision=153686>.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.