http/tests/misc/acid3.html sometimes fails on Windows due to a subtest taking too long

[Adam Roben (:aroben)]

http/tests/misc/acid3.html sometimes fails on Windows. Here's an example:

http://build.webkit.org/builders/Windows%207%20Release%20%28Tests%29/builds/9903
http://build.webkit.org/builders/Windows%207%20Release%20%28Tests%29/builds/9904
http://build.webkit.org/builders/Windows%207%20Release%20%28Tests%29/builds/9905

The test only failed in build 9904. The test output said:

ALERT: Test 69 failed: timeout -- could be a networking issue

I don't think this is a new issue, but only recently have the bots been green enough to notice it.

[Adam Roben (:aroben)]

Here's test 69: http://trac.webkit.org/browser/trunk/LayoutTests/http/tests/misc/acid3.html?rev=60999#L2437

[Adam Roben (:aroben)]

Apparently for a test to time out, it must return "retry" 500 times in a row. Test 69 returns "retry" in this case:

if (kungFuDeathGrip.title.length < 7)
    return "retry";

Looks like kungFuDeathGrip.title gets set up in test 65 by loading a bunch of iframes: http://trac.webkit.org/browser/trunk/LayoutTests/http/tests/misc/acid3.html?rev=60999#L2306

So I guess we're stalling while waiting for those iframes to load?

[Adam Roben (:aroben)]

<rdar://problem/9075086>

[Adam Roben (:aroben)]

This just happened again:

http://build.webkit.org/builders/Windows%207%20Release%20%28Tests%29/builds/10417

[Adam Roben (:aroben)]

And again:

http://build.webkit.org/results/Windows%207%20Release%20(Tests)/r81355%20(10512)/results.html

[Adam Roben (:aroben)]

And again: http://build.webkit.org/results/Windows%207%20Release%20(Tests)/r81363%20(10517)/http/tests/misc/acid3-pretty-diff.html

[Adam Roben (:aroben)]

This is one of the most-frequent causes of flakiness on the Windows 7 bots.