Bug 220699 – [DumpRenderTree] Some js tests generating a large page are randomly timing out on AppleWin EWS

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

[Fujii Hironori]

[Win][DumpRenderTree] Some js tests are randomly timing out

AppleWin EWS layout tests are observing random timeout for some js tests.
For exameple,

attachment 417783 [details]
https://ews-build.s3-us-west-2.amazonaws.com/Windows-EWS/r417783-72338/results.html
js/dfg-float64array.html

attachment 417792 [details]
https://ews-build.s3-us-west-2.amazonaws.com/Windows-EWS/r417792-72345/results.html
js/dfg-inline-arguments-out-of-bounds.html

These js tests generates very long page.
It takes long time to resetWebViewToConsistentStateBeforeTesting after testing.
See also Bug 220145.

[Fujii Hironori]

It's easy to reproduce this issue by invoking DumpRenderTree.exe directly.

.\WebKitBuild\Debug\bin64\DumpRenderTree.exe file:///C:/webkit/LayoutTests/js/dfg-float64array.html

It takes some time before DumpRenderTree.exe quits after reporitng #EOF.

[Fujii Hironori]

Created attachment 417800 [details]
Patch

[Alexey Proskuryakov]

Comment on attachment 417800 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=417800&action=review

> Tools/ChangeLog:9
> +        (runTest): Reset preferences after loading the empty page because

Is there any history of this line is here? I remember that we adjusted some of these before for other reasons, maybe this very one, but am not in front of a large enough screen to check now.

[Fujii Hironori]

r177542 and r177613 changed Windows DRT to align with Mac DRT.

r42056 changed Mac DRT.

> * DumpRenderTree/mac/DumpRenderTree.mm:
> (runTest): Added an additional call to resetWebViewToConsistentStateBeforeTesting just
> before loading an empty page. This prevents extra policy delegate calls from being logged.

[Fujii Hironori]

Created attachment 417859 [details]
Patch

[Alexey Proskuryakov]

Making Windows DRT more like Mac one is reasonable. I've been thinking of recent changes like r267860, r256232, r244668.

Looks like the latest patch breaks a couple tests on Mac.

[Fujii Hironori]

(In reply to Alexey Proskuryakov from comment #6)
> Making Windows DRT more like Mac one is reasonable. I've been thinking of
> recent changes like r267860, r256232, r244668.

IIUC those changes, my patch is irrelevant to them.

> Looks like the latest patch breaks a couple tests on Mac.

Oh no. Here is the callstack of the assertion failure of
accessibility/add-children-pseudo-element.html and accessibility/svg-shape-labelled.html.

ASSERTION FAILED: !createIfNecessary || rootSVGObject
./accessibility/AccessibilityRenderObject.cpp(3255) : WebCore::AccessibilitySVGRoot *WebCore::AccessibilityRenderObject::remoteSVGRootElement(WebCore::AccessibilityRenderObject::CreationChoice) const

1   0x10f9baa59 WTFCrash
2   0x12b679f0b WTFCrashWithInfo(int, char const*, char const*, int)
3   0x12de67d4f WebCore::AccessibilityRenderObject::remoteSVGRootElement(WebCore::AccessibilityRenderObject::CreationChoice) const
4   0x12de5b96a WebCore::AccessibilityRenderObject::detachRemoteSVGRoot()
5   0x12de5b8ee WebCore::AccessibilityRenderObject::detachRemoteParts(WebCore::AccessibilityDetachmentType)
6   0x12dddc51f WebCore::AXCoreObject::detach(WebCore::AccessibilityDetachmentType)
7   0x12dddc0be WebCore::AXObjectCache::~AXObjectCache()
8   0x12dddc7c5 WebCore::AXObjectCache::~AXObjectCache()
9   0x12e61a39b std::__1::default_delete<WebCore::AXObjectCache>::operator()(WebCore::AXObjectCache*) const
10  0x12e61a31f std::__1::unique_ptr<WebCore::AXObjectCache, std::__1::default_delete<WebCore::AXObjectCache> >::reset(WebCore::AXObjectCache*)
11  0x12e5b8897 std::__1::unique_ptr<WebCore::AXObjectCache, std::__1::default_delete<WebCore::AXObjectCache> >::operator=(std::nullptr_t)
12  0x12e5a953c WebCore::Document::clearAXObjectCache()
13  0x12e5b6b2c WebCore::Document::destroyRenderTree()
14  0x12e5b7137 WebCore::Document::willBeRemovedFromFrame()
15  0x12f4deef0 WebCore::Frame::setView(WTF::RefPtr<WebCore::FrameView, WTF::RawPtrTraits<WebCore::FrameView>, WTF::DefaultRefDerefTraits<WebCore::FrameView> >&&)
16  0x10b24f7bd WebFrameLoaderClient::transitionToCommittedForNewPage()
17  0x12f2bb2a8 WebCore::FrameLoader::transitionToCommitted(WebCore::CachedPage*)
18  0x12f2b9f00 WebCore::FrameLoader::commitProvisionalLoad()
19  0x12f22e12c WebCore::DocumentLoader::commitIfReady()
20  0x12f22e8c2 WebCore::DocumentLoader::finishedLoading()
21  0x12f23a47d WebCore::DocumentLoader::maybeLoadEmpty()
22  0x12f23a603 WebCore::DocumentLoader::startLoadingMainResource()
23  0x12f2e7535 WebCore::FrameLoader::continueLoadAfterNavigationPolicy(WebCore::ResourceRequest const&, WebCore::FormState*, WebCore::NavigationPolicyDecision, WebCore::AllowNavigationToInvalidURL)::$_11::operator()()
24  0x12f2e6e59 WTF::Detail::CallableWrapper<WebCore::FrameLoader::continueLoadAfterNavigationPolicy(WebCore::ResourceRequest const&, WebCore::FormState*, WebCore::NavigationPolicyDecision, WebCore::AllowNavigationToInvalidURL)::$_11, void>::call()
25  0x12b688c1a WTF::Function<void ()>::operator()() const
26  0x12b6f2eaa WTF::CompletionHandler<void ()>::operator()()
27  0x12f2b70a8 WebCore::FrameLoader::continueLoadAfterNavigationPolicy(WebCore::ResourceRequest const&, WebCore::FormState*, WebCore::NavigationPolicyDecision, WebCore::AllowNavigationToInvalidURL)
28  0x12f2e4344 WebCore::FrameLoader::loadWithDocumentLoader(WebCore::DocumentLoader*, WebCore::FrameLoadType, WTF::RefPtr<WebCore::FormState, WTF::RawPtrTraits<WebCore::FormState>, WTF::DefaultRefDerefTraits<WebCore::FormState> >&&, WebCore::AllowNavigationToInvalidURL, WTF::CompletionHandler<void ()>&&)::$_8::operator()(WebCore::ResourceRequest const&, WTF::WeakPtr<WebCore::FormState, WTF::EmptyCounter>&&, WebCore::NavigationPolicyDecision)
29  0x12f2e420c WTF::Detail::CallableWrapper<WebCore::FrameLoader::loadWithDocumentLoader(WebCore::DocumentLoader*, WebCore::FrameLoadType, WTF::RefPtr<WebCore::FormState, WTF::RawPtrTraits<WebCore::FormState>, WTF::DefaultRefDerefTraits<WebCore::FormState> >&&, WebCore::AllowNavigationToInvalidURL, WTF::CompletionHandler<void ()>&&)::$_8, void, WebCore::ResourceRequest&&, WTF::WeakPtr<WebCore::FormState, WTF::EmptyCounter>&&, WebCore::NavigationPolicyDecision>::call(WebCore::ResourceRequest&&, WTF::WeakPtr<WebCore::FormState, WTF::EmptyCounter>&&, WebCore::NavigationPolicyDecision)
30  0x12f3196e0 WTF::Function<void (WebCore::ResourceRequest&&, WTF::WeakPtr<WebCore::FormState, WTF::EmptyCounter>&&, WebCore::NavigationPolicyDecision)>::operator()(WebCore::ResourceRequest&&, WTF::WeakPtr<WebCore::FormState, WTF::EmptyCounter>&&, WebCore::NavigationPolicyDecision) const
31  0x12f30e1e5 WTF::CompletionHandler<void (WebCore::ResourceRequest&&, WTF::WeakPtr<WebCore::FormState, WTF::EmptyCounter>&&, WebCore::NavigationPolicyDecision)>::operator()(WebCore::ResourceRequest&&, WTF::WeakPtr<WebCore::FormState, WTF::EmptyCounter>&&, WebCore::NavigationPolicyDecision)

I just reordered loading an empty page and resetting prefs.
It's weird. Anyway, I should look into it.

[Darin Adler]

Comment on attachment 417859 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=417859&action=review

> Tools/DumpRenderTree/mac/DumpRenderTree.mm:2042
> +        // Reset preferences after loading an empty page.

This comment needs to say *why* we do it after loading an empty page. Just like the one on Windows does.

> Tools/DumpRenderTree/win/DumpRenderTree.cpp:1343
> +    // Reset preferences after loading an empty page because it takes a long time for a large page.

I think the use of "it" here makes this hard to understand. I would say "because changing some settings is very slow if there is a large page already loaded."

[Radar WebKit Bug Importer]

<rdar://problem/73558724>

[Fujii Hironori]

r274407 (Bug 223109) fixed this bug.

*** This bug has been marked as a duplicate of bug 223109 ***