299315 - WARNING: Positioned frame that does not handle positioned kids

Status: RESOLVED WORKSFORME

(Core :: Layout, defect)

Description

[Mats Palmgren (inactive)]

Spawned from bug 299188 comment 14.

With the attached testcase we hit this warning in
nsCSSFrameConstructor::GetAbsoluteContainingBlock:

"WARNING: Positioned frame that does not handle positioned kids; looking further
up the parent chain, file nsCSSFrameConstructor.cpp, line 7841"

and the wrong frame is returned from GetAbsoluteContainingBlock() AFAICT.
We return the DIV when we should have returned the child area frame of the
FIELDSET, but because it's NULL at this point we continue traversing up...

A couple of alternatives to fix this:
1. in GetAbsoluteContainingBlock, store the result from GetFieldSetAreaFrame
   in a temp and don't asssign it to 'containingBlock' if it's NULL
   ('containingBlock' is correct at this point.)
http://bonsai.mozilla.org/cvsblame.cgi?file=mozilla/layout/base/nsCSSFrameConstructor.cpp&rev=1.1108&root=/cvsroot&mark=7794,7817-7821#7792

2. Somehow make the area frame the first child of the fieldset *before*
   processing children in nsCSSFrameConstructor::ConstructFieldSetFrame:
http://bonsai.mozilla.org/cvsblame.cgi?file=mozilla/layout/base/nsCSSFrameConstructor.cpp&rev=1.1108&root=/cvsroot&mark=5062,5073,5088,5125,5132-5151,5158#5060

Comment 1

[Mats Palmgren (inactive)]

Attachment: Testcase 1

Comment 2

[Mats Palmgren (inactive)]

Attachment: Trace for Testcase 1

Comment 3

[Boris Zbarsky [:bzbarsky]]

> 2. Somehow make the area frame the first child of the fieldset *before*

This should be the right solution, I think.  The problem is doing it without
triggering SetInitialChildList asserts and without triggering unnecessary
reflows...  The latter might be doable better once reflow refactoring lands.

Comment 4

[Mats Palmgren (inactive)]

Cannot reproduce with a recent m-c debug build.