[cairo] [PATCH] Support image filtering in xlib and xcb backends

Bill Spitzak spitzak at gmail.com
Fri Aug 15 15:33:42 PDT 2014

On 08/15/2014 12:12 AM, Chris Wilson wrote:
> On Thu, Aug 14, 2014 at 04:30:01PM -0700, Bryce Harrington wrote:
>> On Tue, Aug 12, 2014 at 08:06:35PM -0700, Bill Spitzak wrote:
>>> _cairo_pattern_analyze_filter() will change GOOD to BILINEAR if
>>> it thinks bilinear is acceptable. Removed the equivalent test
>>> from _pixman_image_set_properties.
>>> The xlib and xcb backends will use the image fallback for GOOD
>>> and BEST so that the convolution can be done.
>>> Notes:
>>> - I did not test the xcb changes.
>>> - I am assuming other backends treat bilinear and good identically,
>>> at least for scales larger than 1/1.35.
>>> - A CAIRO_RENDER option maybe should be added in case pixman ever
>>> supports GOOD/BEST.
>> Should the above notes be taken care of before this is merged?
>> Functionally this looks good, and I gather this is outcome from recent
>> discussions with Chris and others.  I'd appreciate a third set of eyes
>> on this just for sanity checking, but otherwise:
>> Reviewed-by: Bryce Harrington <b.harrington at samsung.com>
> No. We have to make the fallback happen inside the drivers otherwise it
> will never be fixed.
> -Chris

I'm not sure I understand what you are asking for.

Do you mean that this should not be committed because that will 
discourage XRender from ever doing GOOD/BEST?

I agree it would be nicer if pixman (and thus XRender) was doing this 
but there seems to be some resistance to my patch to pixman by Søren.

Also my previous accepted patch for cairo is pretty much a work-around 
for pixman not doing the filtering. If pixman was fixed such that 
XRender works, this could be removed as well. So I am confused as to why 
that was accepted when this is not.

More information about the cairo mailing list