490182 - [FIX]Removing width attribute of <col> doesn't affect layout in this case

Status: RESOLVED FIXED

(Core :: Layout: Tables, defect)

Description

[Jesse Ruderman]

Attachment: testcase (dynamic)

Comment 1

[Jesse Ruderman]

Attachment: reference (static)

Comment 2

[Boris Zbarsky [:bzbarsky]]

Attachment: Proposed fix

This is great fun.  Basically, when we insert the <col> into the table we put the colframe in the colgroup and then call FrameNeedsReflow() on the _table_.  We don't set any dirty bits on the colgroup.  The table's reflow ends up not reflowing the colgroup, so in particular we never reflow the <col> so we don't clear the NS_FRAME_IS_DIRTY bit or NS_FRAME_FIRST_REFLOW bit on it.

Then we change the "width" attribute, which computes a reflow change hint for the col, which calls nsCSSFrameConstructor::StyleChangeReflow, which does nothing because NS_FRAME_FIRST_REFLOW is set.

The bounce() call is not involved here; the bug is still there even if that's taken out.

The proposed patch just makes us mark the right dirty bits on the colgroup.  The removal of the FrameNeedsReflow call in nsTableColGroupFrame::RemoveFrame is ok because the RemoveChild call we make immediately above that handles calling it.

Comment 3

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

Comment on attachment 374864 [details] [diff] [review]
Proposed fix

Maybe leave a comment in RemoveFrame pointing out that the call in question handles the necessary marking dirty?

sr=dbaron

Comment 4

[Boris Zbarsky [:bzbarsky]]

Good idea.  Done.

Comment 5

[Boris Zbarsky [:bzbarsky]]

Pushed http://hg.mozilla.org/mozilla-central/rev/c532f17489f7