Bug 104502 – [chromium] webkit-unit-tests tests are failing on win

• Status: RESOLVED
• Resolution: FIXED
• Component: Tools / Tests

[noel gordon]

[chromium] webkit_unit_tests tests are failing on win

http://trac.webkit.org/changeset/137092 disabled updateLayerPreserves3DWithAnimations.

[noel gordon]

http://build.webkit.org/builders/Chromium%20Win%20Release%20%28Tests%29/builds/31871/steps/webkit-unit-tests/logs/stdio

applyScrollToScrollableArea is failing on the win bots.

[----------] 1 test from GraphicsLayerChromiumTest
[ RUN      ] GraphicsLayerChromiumTest.applyScrollToScrollableArea

GMOCK WARNING:
Uninteresting mock function call - returning directly.
    Function call: scheduleComposite()
Stack trace:

GMOCK WARNING:
Uninteresting mock function call - returning directly.
    Function call: scheduleComposite()
Stack trace:

GMOCK WARNING:
Uninteresting mock function call - returning directly.
    Function call: scheduleComposite()
Stack trace:
Backtrace:
	(No symbol) [0x021C1D9C]
	(No symbol) [0x021C2A7B]
	(No symbol) [0x021AB477]
	(No symbol) [0x021AB68B]
	(No symbol) [0x0219E053]
	(No symbol) [0x013304EB]
	(No symbol) [0x014FDB89]
	(No symbol) [0x014FDD0B]
	(No symbol) [0x014FE041]
	(No symbol) [0x014FE1DA]
	(No symbol) [0x01418B81]
	(No symbol) [0x0227B58A]
	BaseThreadInitThunk [0x76B2339A+18]
	RtlInitializeExceptionChain [0x77579EF2+99]
	RtlInitializeExceptionChain [0x77579EC5+54]
Backtrace:
	RtlInitializeExceptionChain [0x77579EC5+54]
Backtrace:
	RtlInitializeExceptionChain [0x77579EC5+54]
Backtrace:
	RtlInitializeExceptionChain [0x77579EC5+54]
Backtrace:
	RtlInitializeExceptionChain [0x77579EC5+54]
Backtrace:
	RtlInitializeExceptionChain [0x77579EC5+54]
Backtrace:
	RtlInitializeExceptionChain [0x77579EC5+54]
Backtrace:
	RtlInitializeExceptionChain [0x77579EC5+54]
Backtrace:
	RtlInitializeExceptionChain [0x77579EC5+54]

[noel gordon]

http://trac.webkit.org/changeset/137103 to disable the applyScrollToScrollableArea test.

[noel gordon]

Result of http://trac.webkit.org/changeset/137103 ??  The poblem appeared elsewhere.

http://build.webkit.org/builders/Chromium%20Win%20Release%20%28Tests%29/builds/31872/steps/webkit-unit-tests/logs/stdio

[----------] 5 tests from ScrollingCoordinatorChromiumTest
[ RUN      ] ScrollingCoordinatorChromiumTest.fastScrollingByDefault
Backtrace:
	(No symbol) [0x010B1D9C]
	(No symbol) [0x010B2A7B]
	(No symbol) [0x0109B477]
	(No symbol) [0x0109B68B]
	(No symbol) [0x0108E053]
	(No symbol) [0x0039A7AE]
	(No symbol) [0x00395F3E]
	(No symbol) [0x003EDB89]
	(No symbol) [0x003EDD0B]
	(No symbol) [0x003EE041]
	(No symbol) [0x003EE1DA]
	(No symbol) [0x00308B81]
	(No symbol) [0x0116B58A]
	BaseThreadInitThunk [0x76B2339A+18]
	RtlInitializeExceptionChain [0x77579EF2+99]
	RtlInitializeExceptionChain [0x77579EC5+54]
Backtrace:
	RtlInitializeExceptionChain [0x77579EC5+54]
Backtrace:
	RtlInitializeExceptionChain [0x77579EC5+54]
Backtrace:
	RtlInitializeExceptionChain [0x77579EC5+54]
Backtrace:
	RtlInitializeExceptionChain [0x77579EC5+54]
Backtrace:
	RtlInitializeExceptionChain [0x77579EC5+54]
Backtrace:
	RtlInitializeExceptionChain [0x77579EC5+54]
Backtrace:
	RtlInitializeExceptionChain [0x77579EC5+54]
Backtrace:
	RtlInitializeExceptionChain [0x77579EC5+54]

[noel gordon]

updateLayerPreserves3DWithAnimations  started failing in this range:

http://trac.webkit.org/log/?action=stop_on_copy&mode=follow_copy&rev=137083&stop_rev=137080&limit=100&verbose=on

[noel gordon]

(In reply to comment #3)
> Result of http://trac.webkit.org/changeset/137103 ??  The poblem appeared elsewhere.

I rolled out r137103 on bug 104503, fyi.

[Stephen White]

(In reply to comment #5)
> (In reply to comment #3)
> > Result of http://trac.webkit.org/changeset/137103 ??  The poblem appeared elsewhere.
> 
> I rolled out r137103 on bug 104503, fyi.

I tried to fix this by disabling applyScrollToScrollableArea in http://trac.webkit.org/changeset/137092, but I don't think it worked -- either I got the syntax wrong, or I got the wrong test.

At any rate, my guess is that this started with the Chromium change http://src.chromium.org/viewvc/chrome?view=rev&revision=171850.  You can see it on the downstream WebKit bots, and it doesn't show up on the build.webkit.org bots until Chromium DEPS rolled into WebKit at http://trac.webkit.org/changeset/137083.

[James Robinson]

They started failing on the build.webkit.org bot here:
http://build.webkit.org/builders/Chromium%20Win%20Release%20%28Tests%29/builds/31863/

which proves that it's a chromium side patch.  The tightest chromium range I can find is http://build.chromium.org/f/chromium/perf/dashboard/ui/changelog.html?url=/trunk/src&mode=html&range=171857:171840

[Stephen White]

Chrome r171850 was reverted, but that didn't seem to fix the problem.  So my next guess is Chrome r171847.

[Stephen White]

I've reverted my disable of updateLayerPreserves3DWithAnimations in WebKit r137242, since that just play whack-a-mole with the tests anyway.

[Dana Jansens]

Better backtrace

Backtrace:
	cc::LayerImpl::verticalScrollbarLayer [0x619F47FB+610521]
	cc::LayerImpl::verticalScrollbarLayer [0x619F0399+593015]
	cc::LayerImpl::verticalScrollbarLayer [0x619EFE56+591668]
	cc::LayerImpl::verticalScrollbarLayer [0x619F45DF+609981]
	cc::LayerImpl::verticalScrollbarLayer [0x61A1E230+781070]
	cc::LayerImpl::verticalScrollbarLayer [0x61A1C1D7+772789]
	webkit_glue::CppBoundClass::GetProperty [0x62E78260+2266620]
	webkit_glue::CppBoundClass::GetProperty [0x62E76732+2259662]
	webkit_glue::CppBoundClass::GetProperty [0x62E76676+2259474]
	WebKit::WebPerformance::domainLookupEnd [0x665AF297+1364499]
	WebKit::WebPerformance::domainLookupEnd [0x66796D31+3361965]
	WebKit::WebPerformance::domainLookupEnd [0x667956C7+3356227]
	WebKit::WebPerformance::domainLookupEnd [0x667971E6+3363170]
	WebKit::WebPerformance::domainLookupEnd [0x66797196+3363090]
	WebKit::WebFilterOperation::WebFilterOperation [0x66DE6D7D+5249597]
	WebKit::WebFilterOperation::WebFilterOperation [0x66DFC31F+5337055]
	WebKit::WebFilterOperation::WebFilterOperation [0x66DE6CF1+5249457]
	WebKit::WebFilterOperation::WebFilterOperation [0x66DE744F+5251343]
	WebKit::WebFilterOperation::WebFilterOperation [0x66DEDDBD+5278333]
	WebKit::WebFilterOperation::WebFilterOperation [0x66DFCFB7+5340279]
	WebKit::WebFilterOperation::WebFilterOperation [0x66DEC620+5272288]
	WebKit::WebFilterOperation::WebFilterOperation [0x66DDDED3+5213075]
	WebKit::WebPerformance::domainLookupEnd [0x668B0E99+4517397]
	(No symbol) [0x00D810B3]
	(No symbol) [0x00E45ADF]
	(No symbol) [0x00E4590F]
	BaseThreadInitThunk [0x75D9339A+18]
	RtlInitializeExceptionChain [0x77579EF2+99]
	RtlInitializeExceptionChain [0x77579EC5+54]
Backtrace:
	RtlInitializeExceptionChain [0x77579EC5+54]
Backtrace:
	RtlInitializeExceptionChain [0x77579EC5+54]
Backtrace:
	RtlInitializeExceptionChain [0x77579EC5+54]

[James Robinson]

Pretty sure that stack is also incorrectly symbolized.

[Dana Jansens]

Darn, linux+valgrind is happy with these tests.

[Dana Jansens]

I wonder if landing https://bugs.webkit.org/show_bug.cgi?id=104046 will fix it? Maybe it's a case of different perspectives of the old classes?

[noel gordon]

See http://build.webkit.org/builders/Chromium%20Win%20Release%20%28Tests%29/builds/31898 seems bug 104046 fixed it.