Jesse Ruderman Reporter
	Description • 13 years ago

###!!! ASSERTION: continuations should have the same style context: 'oldContext == nextIBSiblingContext || oldContext->GetPseudo() != nextIBSiblingContext->GetPseudo() || oldContext->GetParent() != nextIBSiblingContext->GetParent()', file layout/base/nsFrameManager.cpp, line 1198

This assertion was added in bug 670467. It is similar to the assertion in bug 615944.

	Boris Zbarsky [:bzbarsky]
	Comment 2 • 13 years ago

So this is an interesting case.  What happens here is that nextIBSiblingContext and oldContext are in fact identical (same parent, same rulenode, same bits, same pseudo tag, same if-visited style).  But they are not the same object, because the parent has 13 kids so we ran off the sharing threshold in nsStyleContext::FindChildWithRules.

I would think that a similar assert can be triggered for the existing in-flow code without any ib issues in sight, and in fact bug 615944 confirms that.

David, we need to either relax the assertions here or make the sharing more exact....

	Robert O'Callahan (:roc) (email my personal email if necessary)
	Comment 3 • 13 years ago

It would probably make sense to use an equality check that's not pointer equality.

	Boris Zbarsky [:bzbarsky]
	Comment 4 • 13 years ago

That would be the relaxing the assertion I referred to, yes.

	Jesse Ruderman Reporter
	Comment 5 • 11 years ago

The testcase in this bug no longer triggers the assertion, but other (even uglier) testcases do.  Is the explanation in comment 2 still relevant?

	David Baron :dbaron:
	Comment 6 • 11 years ago

Not sure.

After recent changes in layout/style/RestyleManager.cpp the pointers should be kept equal as a result of being copied over every time we reresolve style, so the sharing optimization being inexact should no longer matter.