string-related chromium browser_tests failing around webkit rev 126164

As of http://build.chromium.org/p/chromium.webkit/builders/Mac10.6%20Tests/builds/12967 , the following browser_tests are consistently failing in http://build.chromium.org/p/chromium.webkit/waterfall?show=Mac10.6%20Tests :

FillProfileCrazyCharacters
PrefsStringSavedAsIs
WhitespacesAndSeparatorCharsStrippedForValidCCNums

blamelist:
webkit 126160-126164 inclusive
chrome 152566-152579 inclusive

My guess is that it is one of the following CLs:

http://trac.webkit.org/changeset/126164 ('TextIterator takes O(n2) to iterate over n empty blocks') : [e-mail]
http://trac.webkit.org/changeset/126162 ('Web Inspector: break on exceptions decoration was lost, restoring.') : [e-mail]
http://src.chromium.org/viewvc/chrome?view=rev&revision=152569 ('Extract PrefServiceBase into chrome/browser/api.  Use in api and autofill.'): [e-mail]

I don't see the failure in https://chromium-build.appspot.com/p/chromium/console , so it's probably not a chromium change.

Maybe it's a webkit change, or a problem with this particular builder...

Without knowing what those tests do and how they’re failing, I am not sure what to do.

(In reply to comment #2)
> Without knowing what those tests do and how they’re failing, I am not sure what to do.

Welcome to the club. :-)

My hope was that one of you who committed those CLs would have a "lightbulb moment".  But maybe that won't happen.

But this is going to block the webkit_revision roll in Chromium if not resolved, so hopefully we can figure something out.

FWIW, the strings “FillProfileCrazyCharacters”, “PrefsStringSavedAsIs”, “WhitespacesAndSeparatorCharsStrippedForValidCCNums” do sound related to the concepts mentioned in <http://src.chromium.org/viewvc/chrome?view=rev&revision=152569>: “PrefServiceBase”, “autofill”.

(In reply to comment #4)
> FWIW, the strings “FillProfileCrazyCharacters”, “PrefsStringSavedAsIs”, “WhitespacesAndSeparatorCharsStrippedForValidCCNums” do sound related to the concepts mentioned in <http://src.chromium.org/viewvc/chrome?view=rev&revision=152569>: “PrefServiceBase”, “autofill”.

Agreed!  In fact, these are all autofill tests.  From the stdio:

Failing tests:
AutofillTest.FillProfileCrazyCharacters
AutofillTest.PrefsStringSavedAsIs
AutofillTest.WhitespacesAndSeparatorCharsStrippedForValidCCNums

joi???

[Kenneth Russell]

I tried reverting mitz's change http://trac.webkit.org/changeset/126164 locally but it didn't help.

Filed http://code.google.com/p/chromium/issues/detail?id=143992 regarding these failures.

I have no idea why they aren't failing on the TOT Chromium waterfall, nor why they started failing on this bot.

I haven't built the tests Release mode yet, only Debug, so only have seen the assertion failure in the above Chromium bug report. I'm almost done with a Debug build at WebKit r120619 to see the behavior at that revision.

(In reply to comment #6)
> I tried reverting mitz's change http://trac.webkit.org/changeset/126164 locally but it didn't help.
> 
> Filed http://code.google.com/p/chromium/issues/detail?id=143992 regarding these failures.
> 
> I have no idea why they aren't failing on the TOT Chromium waterfall, nor why they started failing on this bot.

Certainly http://src.chromium.org/viewvc/chrome?view=rev&revision=152569 ('Extract PrefServiceBase into chrome/browser/api.  Use in api and autofill.') seems spot-on.  So it's worth reverting that locally to see if it helps...

But if so, I don't understand why this didn't show up in the Chromium waterfall.

[Ilya Sherman]

At a glance, this is likely caused by [ https://chromiumcodereview.appspot.com/10855253/ ], which added these tests.  See also [ http://code.google.com/p/chromium/issues/detail?id=143992 ].

[Kenneth Russell]

It looks like the timeouts have been fixed in http://code.google.com/p/chromium/issues/detail?id=143992 , and that it was a Chromium-side change affecting just these bots.