[Webkit-unassigned] [Bug 110357] ClippPath artifacts due to mask image clamping

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=110357





--- Comment #4 from Florin Malita <fmalita at chromium.org>  2013-02-23 10:39:26 PST ---
(In reply to comment #3)
> It shows the same clipping artifact but the untransformed clippath is small enough to fit in a 4096x4096 buffer.  It is applied to shapes that are themselves inside the same transform group.  It would seem you could apply the clippath pre-transform and then transform the result without any of these issues.  It would solve my particular issue though not the more general problem that has been discussed here.  Is this reasonable?

Unfortunately, due to the current mask-based clipping approach, I don't think that would help: for results to be accurate, the mask bitmap needs to match the device resolution (hence it needs to be sized in absolute coordinates). Otherwise whatever intermediate bitmap we get by applying it in user coordinates would still have to be scaled when drawn, with the same undesirable results.

Basically the root cause is the mask resolution not matching the device resolution; whether this is because we clamp its size to 4096^2 or because we compute it in a different coordinate space doesn't really make a difference.


> Note that Safari 5.1.7 doesn't have a problem with either case.  That was surprising to some but it is verifiable and it would be nice to understand why it works there.

Yes, agreed. I can't easily check the WK version used in 5.1.7, but I presume it's < 96155 (where the 4096 clamping was introduced - http://trac.webkit.org/changeset/96155). So I'm guessing it works because in 5.1.7 we allocate an appropriately sized mask.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.