Re: [PATCH 0/2] Update pixman reference images

On Fri, Oct 10, 2014 at 08:03:28AM +0100, Chris Wilson wrote:
> On Thu, Oct 09, 2014 at 06:38:10PM -0700, Bryce Harrington wrote:
> > The downscaling tests check all the different filter levels, to three
> > different scales.  Two of the scales, 96:96 and 96:24, are easy scaling
> > settings and thus should produce essentially identical results.  The
> > third scaling, 96:95, requires blurring the lines bordering solid
> > colors, and we generally are fairly permissive as to what the renderer
> > decides to do.  Unfortunately the various backends seem to vary quite a
> > bit in what they choose to do.
> > 
> > Because of this, I'm tempted to just drop the 96:95 scale test since it
> > will always be finicky.  However, I do think it's important that we test
> > "off" scales like this.
> 
> Don't drop it. The image/pixman backend serves as a reference. If the
> other backends have equivalent quality but slightly different output,
> they define their own reference.

Alright, then I'll go ahead and commit these patches.

> If they have unacceptable differences,
> mark them as xfail so that we can detect any changes.

Will do.  All of these looked fine.  The 96:24 BEST cases looked
legitimately broken but I think that needs separate investigation.

Thanks,
Bryce
-- 
cairo mailing list
[email protected]
http://lists.cairographics.org/mailman/listinfo/cairo