Updated Regression Output

[Germain Garand]

Le Mardi 05 Octobre 2004 12:26, Stephan Kulow a écrit :
> Hi!
>
> I was quite some time absent/busy, so I took today to
> update the current regression state and found 312
> failed test cases. 

Hi Stephan!

> Most of them are either shrinked
> layers

yes... that's a required change. If you remember: they were expanded some time 
ago to provide a correct scrolling size in overflow: scroll situations.

But it turned out we can't do that systematically, because of a IE 
compatibility hack we have on table cells (we make those grow to encompass 
their own overflow, which is quite stupid, but IE does that). This broke 
nvidia.com for instance.

So I made the layers encompass children only if they hidesOverflow().

> or additional paintBackground flags.

This one is interesting. It showed up when Leo changed table's default 
border-width, but I did some investigations this afternoon and it turns out 
the bug is not there.

The problem is that the hasBorder() test is broken... it is currently 
returning true even when no border has been set at all.
 (see attached patch, that I'd like to commit tomorrow - it also fixes the 
operator== of inherited flags, that was broken by the recent enlargement to 
64 bits)

as RenderObject::setStyle says:

    setShouldPaintBackgroundOrBorder( ... ||  m_style->hasBorder() || ... );

we are almost always setting the paintBackground flag, and thus trying to 
paint some kind of inexisting background at every repaint.

Of course, this will change a lot of [R] output... ;(

Greetings,
Germain
-------------- next part --------------
A non-text attachment was scrubbed...
Name: hasBorder.diff
Type: text/x-diff
Size: 1328 bytes
Desc: not available
URL: <https://mail.kde.org/mailman/private/kfm-devel/attachments/20041007/c1316ccf/attachment.diff>