[cairo] [PATCH] CAIRO_FILTER_BEST uses NEAREST filter when possible

[Bill Spitzak]

On 08/01/2014 10:52 PM, Chris Wilson wrote:
> On Fri, Aug 01, 2014 at 07:25:22PM -0700, Bill Spitzak wrote:
>> This avoids slow filtering for identity transforms when BEST is used.
>> Pixman is already handling the GOOD case.
>
> See _cairo_pattern_analyze_filter
> -Chris
>
Yes that looks like it should be doing this, and explains why the 
identity transform never hit my code. I think the problem is the same 
one I encountered, which is that sin(M_PI_2) does not produce exactly 
one. Rounding to the nearest fixed value fixes this.