311557 - [FIX]No reflow after setting -moz-box-align property (in onload handler)

Status: RESOLVED FIXED

(Core :: XUL, defect, P3)

Description

[Jens Bannmann]

When a new style rule including the -moz-box-align property is applied to a box
via an onload handler, its children do not reflow. They do reflow when the
align= attribute is added to and removed from the box, though.

Comment 1

[Jens Bannmann]

Attachment: test case

Test case with two boxes, setting a class name on the left box in an onload
handler. The rule gets applied (background color is lightgray), but the box
alignment doesn't have any effect. The right box has the class name from the
start, and displays correctly.

Comment 2

[Jens Bannmann]

Note: I stumbled upon this while creating the patch for bug 133527. As the CSS
rule didn't work, I additionally set the align=center attribute in the JS. Once
this bug gets fixed, the workaround (and the comment referencing this bug) can
be removed from alert.js and the CSS files.

Comment 3

[Boris Zbarsky [:bzbarsky]]

Attachment: Fix

Note that all this stuff can die once we do XUL attribute mapping, I think...

Comment 4

[Boris Zbarsky [:bzbarsky]]

Fixed.

Comment 5

[John Mellor (Jomel)]

Setting -moz-box-orient at runtime used to have no effect either, but now works in current trunk builds, probably thanks to this patch.

Comment 6

[Jens Bannmann]

Comment on attachment 218601 [details] [diff] [review]
Fix

Can we land this on the 1.8.1 branch? It would be great if branch and trunk code would be synced here, because otherwise, my patch for bug 133527 would have to include a workaround that is no longer needed on trunk.

Comment 7

[Brian Ryner (not reading)]

Comment on attachment 218601 [details] [diff] [review]
Fix

Wait, don't you need to return something from this??

Comment 8

[Boris Zbarsky [:bzbarsky]]

I'm not completely happy with putting this on branch; I don't have a good feel for either regression risk or interaction with other parts of branch code here.  Also, I have no time to either sort the above out or to maintain this on branch.  If someone wants to put this on branch, they'll have to do it without my participation, basically.

And yes, a "return NS_OK" is needed there; see rev 1.302 of nsBoxFrame.cpp.

Comment 9

[Mike Schroepfer]

Comment on attachment 218601 [details] [diff] [review]
Fix

Given comment 8 minusing for the 1.8 branch