khtml patch review

[Germain Garand]

Le Lundi 29 Octobre 2007 01:44, Charles Samuels a écrit :
> Hi Germain,
>
> Germain Garand wrote, on Sunday 2007 October 28:
> > no this patch is wrong I'm afraid.
> > Only positioned objects are allowed to have z-order.
>
> Maks pointed out to me that requiresLayer() should return true if the
> object is positioned. 

> Why is it that the iframe, being positioned in the 
> css, still has its requiresLayer() return false? (If you don't know, I'll
> investigate that myself)

it is? ah right, I thought only the container was positioned in your testcase 
but I see the class applies to the iframe itself too.

you are right then, the bug might be simple.
I don't know what could cause that. It might be related to how positioned 
inline objects have their display forcibly mutated to block (so here the 
iframe is certainly display block) just an idea.

>
> > You can't fix this bug in branch as it is non-trivial - due to the way we
> > implemented z-order on iframes by way of masks, which are a bit expensive
> > to calculate and thus need to skip some cases. Actually, it would be
> > possible to improve the mask calculating algorithm, but I'm not likely to
> > do that before some time. Please feel free if you feel like diving into
> > trouble waters :)
>
> What kind of regressions could be caused by my patch, out of curiosity?
> Especially in 3.5.

it will make some kitten cry? seriously I have no idea, but I'd not willfully 
break the CSS specification and hope nothing will happen.

>
> > As for trunk, my recent painting patch should have taken care of it?
>
> I'm afraid not

now I'm seriously puzzled as to why it works in my tree... it shouldn't make 
any difference in that respect.. oh well.

Greetings,
Germain