[Webkit-unassigned] [Bug 42273] Convolution computation causes bad alpha channel values

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=42273


Adam Langley <agl at chromium.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |agl at chromium.org




--- Comment #2 from Adam Langley <agl at chromium.org>  2010-07-14 13:29:53 PST ---
There's a comment in the layout test HTML saying that this test simply shouldn't crash. However, I think that comment should exist in the output, otherwise the expected values are pretty mysterious.

Secondly, if it still crashes as a text-only layout test, that would be a plus. (I could well believe that it doesn't, but it would be nice to check.)

The fix for the crash looks good.

I'm afraid that I don't know enough to really comment on the change to the convolution. I don't think values should be un-premultiplied before convolving (which I don't think this patch does): It seems right that adding $x times an almost transparent pixel doesn't do a lot.

So this code looks sane, but I could imagine other sane solutions.

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