290129 - Messed up table rendering with dynamic css/js

Status: RESOLVED WORKSFORME

(Core :: Layout: Tables, defect)

Description

[Andreas Lange]

User-Agent:       Mozilla/5.0 (X11; U; SunOS sun4u; en-US; rv:1.8b2) Gecko/20050412
Build Identifier: Mozilla/5.0 (X11; U; SunOS sun4u; en-US; rv:1.8b2) Gecko/20050412

By using JS to toggle display:none for classes used on table rows/cols it is
possible to end up with a table rendering full with holes.

Reproducible: Always

Steps to Reproduce:
1. Load testcase
2. Click Col A, Col B, Col C
3. Click Col A again

Actual Results:  
One cell in the Col A header disappears.


This testcase is a minimized version of a rather large dhtml application. For
various reasons, we used an abstracted way to toggle visible parts of tables by
inserting/removing display:none css rules for classes used in the tables. Not
browser friendly, but convinient. 

We have had problems with messed up tables with this method from Mozilla1.7 (and
the FF branch). Making a minimized testcase proved difficult though.

Comment 1

[Andreas Lange]

Attachment: Medium size testcase

This testcase is reproducable on current trunk builds, tested with own Solaris
build plus a windows nightly to rule out build issues.

It is not reproducable on Moz1.7 or FF1 even though we have hade the same
issues on those products with the application. I guess there is some kind of
testcase size and randomness issue involved here.

Use the steps above: uncheck the Col checkboxes LTR, then check the first
again. Random clicking might provoke even further errors.

I suspect that the style rule "border-collapse: collapse" on the table probably
is significant, removing that breaks the testcase.

Comment 2

[Boris Zbarsky [:bzbarsky]]

Bernd, any idea what's going on here?  When I hide col C or reshow col A I get:

###!!! ASSERTION: invalid BC damage area: 'PR_FALSE', file
../../../mozilla/layout/tables/nsTableFrame.cpp, line 4563

Comment 3

[Bernd]

BC blues I would guess

Comment 4

[Andreas Lange]

Tested with 3.0pre9 2007-10-13 and 2007-10-22 (thought bug 244135 checked in during that timespan might have fixed it, but it seems to have been fixed previously).

Comment 5

[Boris Zbarsky [:bzbarsky]]

Can we get a reftest in for this?

Comment 6

[Andreas Lange]

Attachment: Reftest-1

I don't have an environment right now where I can make a patch, but here are reftest files anyway. Hope someone can check them in.

I was able to minimize the test and automate it with the body.offsetWidth trick. Fails consistently in Fx2.

Comment 7

[Andreas Lange]

Attachment: Reftest-1-ref (js removed)

Comment 8

[Boris Zbarsky [:bzbarsky]]

Andreas, thanks!  Test checked in.