422089 - -moz-columns inside <table> make content disappear and don't change height when downsizing window

Status: RESOLVED FIXED

(Core :: Layout, defect)

Description

[Radek 'sysKin' Czyz]

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9b5pre) Gecko/2008031004 Firefox/3.0b5pre
Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9b5pre) Gecko/2008031004 Minefield/3.0b5pre

If number of -moz-columns shrinks because of window resizing, the following content might disappear

Reproducible: Always

Steps to Reproduce:
1. Open webpage from URL or testcase attached below
2. Make sure browser is wide enough to see columns.
3. Resize browser to make it smaller and reduce number of columns

Actual Results:  
Content under columns no longer displays and takes no height.

Expected Results:  
No magic :)

Not 100% reproduceable, and I can sometimes make the content come back. Lots of magic going on.

Comment 1

[Radek 'sysKin' Czyz]

Attachment: testcase

resize browser window to see magic

Comment 2

[John P Baker]

See also Bug 417487 - this may be a better component though.

Comment 3

[j.j.]

Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9.1a2pre) Gecko/2008080403 Minefield/3.1a2pre

Confirming this. Some notes:

1. After downsizing the window, the height of the div don't change, so also
   some of div's content is lost.
2. Page reload clears all problems.
3. The bug appears even without the <table> container, but less 	
   reproducible.

Comment 4

[Hans Schmucker]

I can confirm this behavior also in an independent testcase, so it's really quite widespread. Apparently, the columns really "lose" data during resizes. I've got a seperate DIV that appears right next to the columned text (behind it in the tree). It's absolutely positioned to be at the top right of the document. Resizing the page (really fast, for some reason that bug won't trigger in my case when resizing slowly), causes it to disappear.

Comment 5

[Daniel.S]

On horizontal resize, the h1 element vanishes.

This is a regression from Firefox 2.

Comment 6

[Radek 'sysKin' Czyz]

Regression window:
20061207 works
20061208 fails, and additionally freezes on the testcase.

-> reflow branch landing :(

Comment 7

[j.j.]

Attachment: testcase 2

bottom-border of table diappears when bug occurs.

Comment 9

[Daniel.S]

Attachment: References list in Wikipedia

This may not happen rarely, Wikipedia is suffering from this problem. Look what it does with the reference list of http://en.wikipedia.org/wiki/Mozilla_Firefox

Comment 10

[Ryo Onodera]

Attachment: another testcase

This attached testcase might be related to this bug.
Load it and change the height of window, then reload it. An rendering difference appears, but shouldn't. Apparently, columns aren't taking account for height changes while reflowing because of resizing. Some assertions are alarming.

I'll investigate this issue lightly...

Comment 11

[Ryo Onodera]

Attachment: half baked patch.

I tracked down what part of code causing this bug. I show the troubling part with this attached patch. Apparently, flags for optimization are too loose. Because of naively evaluating these flags to false with this patch, Firefox get a bit sluggish from my sense. 

Also, with this patch applied, columns are correctly reflowed only when I resize the window by grabbing the corner. If I change the height of the window, sometimes it works, sometimes it doesn't work.

If no one enlighten me, I'll deeply investigate the real problem behind this bug in some days or weeks...

Comment 12

[John P Baker]

I am having trouble duplicating behaviour of the testcase on 3.5, but I can in 3.0.11, so someone else may need to check.

I was rechecking them as bug 417487 has been fixed by the checkin for bug 389462.

Comment 13

[philippe (part-time)]

I cam easily reproduce the issue in testcase 1 and 2, along with the Wikipedia page with a recent Shiretoko build (Gecko 1.9.1)
The latest Minefield build works correctly - no vanishing content etc. I guess this can be marked fixed by bug 389462.

Comment 14

[j.j.]

Confirming comment 13
I can reproduce the bug with Firefox 3.5 and testcase2.

No bug in 1.9.2 trunk
Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9.2a1pre) Gecko/20090707 Minefield/3.6a1pre

Comment 15

[philippe (part-time)]

Ok, marking as FIXED by bug 389462, per comment 13 and 14.