[Webkit-unassigned] [Bug 98504] [CSS Shaders] Implement overlay, color-dodge, color-burn, hard-light, soft-light blend modes.

bugzilla-daemon at webkit.org bugzilla-daemon at webkit.org
Fri Oct 12 23:23:04 PDT 2012


https://bugs.webkit.org/show_bug.cgi?id=98504





--- Comment #17 from Huang Dongsung <luxtella at company100.net>  2012-10-12 23:23:48 PST ---
(In reply to comment #14)
> (In reply to comment #13)
> > (In reply to comment #12)
> > > (From update of attachment 168290 [details] [details] [details])
> > > Hmm... I see that Max did in fact make this suggestion "For the simpler blend modes that don't need to blend by component, we will overwrite this default value."

Done. I did misunderstand what Max instructed.

> > If Huang makes the previous change, then it should be something like:
> > // Cs: is the source color in css_BlendColor() and the source color component in css_BlendComponent()
> > ...

Done.

> > I think the % maps nicely to the hand calculations above, but sometimes the % causes color rounding problems and we have to use the 8-bit number variants. (I wish we didn't have to! It would be nice if we could specify a color tolerance per test.)> 
> A color tolerance per test is holy grail. We really need it to reduce creating tests and maintenance cost for each platform.

Currently, we can set the tolerance of image differences per execution of new-run-webkit-tests. I wish we could specify a color tolerance per test as Max said.

We need the way to send meta data per test to new-run-webkit-tests. I think there are two solutions.
1. Specify the tolerance in a file name. For example, custom-filter-blend-modes-TOLERANCE=5.html
2. Create a meta file. If a test does not have a meta file, apply the default setting. For example, custom-filter-blend-modes-meta.txt including TOLERANCE=5.

IMHO the pixel test is the only way to check the regression of CSS Shaders. All platforms output slightly different pixels. We need a solution soon.

> > > Also, this shows that we may need a more complex test for burn, since a completely black result doesn't really exercise much.
> > 
> > Good point- perhaps we could use different input colors whose result isn't black.

Done.

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



More information about the webkit-unassigned mailing list