REGRESSION (r167845): ASSERT(!m_renderView.needsLayout()) in svg/custom/bug79798.html

• Component: Tools / Tests
• Status: RESOLVED FIXED

[Tim Horton]

The following layout test is failing on all Mac WebKit2 bots:

svg/custom/bug79798.html

Probable cause:

http://trac.webkit.org/changeset/167845 causes us to finish layout with dirty layout, somehow. It seems to involve doing a subtree layout of a subframe (the <object>?) but having dirtied layout on the main frame (during deletion-within-selection), or some such (haven't debugged extensively).

No idea why it would only happen in WebKit2.

[Tim Horton]

I'm tempted to skip the test instead of rolling out because the user facing

[Tim Horton]

(In reply to comment #1)
> I'm tempted to skip the test instead of rolling out because the user facing
impact of the original bug is so much worse, but we've done so much work to flush out all of the ASSERT(!needsLayout) issues that it would be sad to reintroduce one.

[Tim Horton]

http://trac.webkit.org/changeset/167898 skips it for WebKit2 (still no idea why it is specific to those bots, but at least this means the test is still covered somewhere)

[Tim Horton]

<rdar://problem/16738249>

[Daniel Bates]

I just hit this assertion failure in RenderLayerCompositor::requiresCompositingForScrollableFrame() using a debug build of WebKit r216439 on <https://www.icloud.com> by performing the following:

1. Visit <https://www.icloud.com>.
2. Type 'a' in the password field.
3. Press the delete key on the keyboard.

Then the WebProcess crashes because of the assertion failure.

[Daniel Bates]

Created attachment 309454 [details]
Stacktrace

Stacktrace after following the reproduction steps in comment #5.

[Simon Fraser (smfr)]

I think Andy was going to fix this (or did recently).

[Andy Estes]

Created attachment 309641 [details]
Patch

[Alexey Proskuryakov]

Comment on attachment 309641 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=309641&action=review

> LayoutTests/platform/wk2/TestExpectations:-196
> -webkit.org/b/132297 svg/custom/bug79798.html [ Skip ]

There is one more expectation for this:

LayoutTests/platform/ios-wk2/TestExpectations:272:webkit.org/b/132297 [ Debug ] svg/custom/bug79798.html [ Skip ]
LayoutTests/platform/wk2/TestExpectations:196:webkit.org/b/132297 svg/custom/bug79798.html [ Skip ]

[Andy Estes]

Created attachment 309652 [details]
Patch

[Andy Estes]

(In reply to Alexey Proskuryakov from comment #9)
> Comment on attachment 309641 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=309641&action=review
> 
> > LayoutTests/platform/wk2/TestExpectations:-196
> > -webkit.org/b/132297 svg/custom/bug79798.html [ Skip ]
> 
> There is one more expectation for this:
> 
> LayoutTests/platform/ios-wk2/TestExpectations:272:webkit.org/b/132297 [
> Debug ] svg/custom/bug79798.html [ Skip ]

If iOS was hitting this assertion, then this patch probably won't fix it, since iOS sets ScrollableInnerFrameTrigger.

Let's see what EWS says.

[WebKit Commit Bot]

Comment on attachment 309652 [details]
Patch

Clearing flags on attachment: 309652

Committed r216643: <http://trac.webkit.org/changeset/216643>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.