[PATCH] support for redirections in testregression

[Germain Garand]

Le Jeudi 29 Juillet 2004 16:20, Tobias Anton a écrit :
> > What matters right now is if you have
> > http://ktown.kde.org/~coolo/khtml_reverts.diff applied
>
> Huh?
>
> Should I or shouldn't I have this patch applied to generate a baseline for
> a regression test?
> What does this patch do at all?

As Stephan said, it's a revert. When the parser finishes its work, the canvas 
element is closed, which means the document should be layouted for good.

 With this revert applied, the document is immediately relayouted upon 
closing, without any scheduling. Apparently, this is dangerous and can 
trigger some reentrancy or assert failures... you'd have to ask Dirk for 
details.
Without it, the document is just marked as dirty, and will be definetly 
layouted when the next layout timer elapses.

Now the problem: we have been plagued for a long time by cyclic recalculations 
of vertical scrollbars. A number of hacks in RenderCanvas::layout have been 
made as a turnaround, but the root of the problem is still unsolved, and 
difficult to solve.

By just marking the document as dirty, we return earlier to the event loop and 
allow scrollbars to change their state when the document root isn't 
layouted(), thus resurecting some scrollbar issues.

So here we are. The patch should be avoided, but we need to get rid of the 
anti-flashing scrollbars hacks AND still solve the scrollbar flickering 
problem.
I've already given several tries to that, with little success up to now, so 
feel free to help if you have ideas :-)

Germain
----
Back from the beautiful landscapes of Bretagne. Head still in the clouds.