Based on a piece of layout/reftests/bugs/364066-1.html (bug 364066).

The image changes width fine. What doesn't change is the table cell. The problem is that in this case a height change in fact leads to an intrinsic width change. This looks like a regression from bug 502288.

Except of course we do clear ancestor intrinsics in this case...

Er, no. We in fact do not. Doing that should fix this bug; testing.

Er, and I take that back once again. For height we have: return NS_CombineHint(hint, NS_SubtractHint(nsChangeHint_ReflowFrame, nsChangeHint_ClearDescendantIntrinsics));

Aha! The key is that the <div> that is a _child_ of the <div> we're changing the height on has an intrinsic width that depends on the height of the ancestor. If we want to have that to be the case (and currently it is because GetPercentHeight in nsLayoutUtils calls itself on the containing block), then changes to height do in fact need to clear _descendant_ intrinsics in general. David, does that sound ok?

Yes, that sounds fine.

Is this broken on 1.9.2 as well? It seems like something we should get fixed there.

> Is this broken on 1.9.2 as well? Yes. I'll try to get together a patch for this today; then we can land it on m-c and work on getting on branch. If you think this should block, please nominate? I'm being a little torn about that at this late date...

At this point, I'd be ok with fixing it in a .1, although I don't seem to be able to nominate it for that.

Comment on attachment 419343 [details] [diff] [review] m-c fix r=dbaron

Comment on attachment 419344 [details] [diff] [review] 1.9.2 merge r=dbaron

Pushed http://hg.mozilla.org/mozilla-central/rev/89d33cf3490c

This is a layout regression caused by slight over-optimization. The fix is very low risk, and I think we should really get it in before shipping a major update to 3.6.

Comment on attachment 419344 [details] [diff] [review] 1.9.2 merge a=beltzner for 1.9.2.2, in a fantastic deal with dbaron by which way everyone wins!

Pushed http://hg.mozilla.org/releases/mozilla-1.9.2/rev/4a8ddcbeef48

verified FIXED on build using the testcase provided: Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.5; en-US; rv:1.9.2.2) Gecko/20100316 Firefox/3.6.2