Bugzilla

Appreciate the insights - thanks!

(In reply to Glenn Watson [:gw] from comment #9)
> If it renders correctly in sw-wr but is off by 1/255 in hw-wr> [...] 

It does, yeah.

> [...] it's unlikely to be a simple fix.

OK, good to know.

> I think we're best to mark these as fuzzy, for now, it's unlikely I'll have time to look in to this in depth this half

Cool.  Yeah, this is visually imperceptible, so it's hard to make a case for making this a priority to fix, other than from a "that seems really-unexpected, computationally" sort of perspective.

I'm happy to mark these as fuzzy for the time being, though (when time allows).  A `maxDifference=1` mismatch along the edge of a rotated shape seems pretty trivially fine from a fuzzy-annotation perspective, and unrelated to what the test is trying to test for (which is does the transform work *at all*; and it does work.)

Appreciate the insights - thanks!

(In reply to Glenn Watson [:gw] from comment #9)
> If it renders correctly in sw-wr but is off by 1/255 in hw-wr> [...] 

It does / it is, yeah.

> [...] it's unlikely to be a simple fix.

OK, good to know.

> I think we're best to mark these as fuzzy, for now, it's unlikely I'll have time to look in to this in depth this half

Cool.  Yeah, this is visually imperceptible, so it's hard to make a case for making this a priority to fix, other than from a "that seems really-unexpected, computationally" sort of perspective.

I'm happy to mark these as fuzzy for the time being, though (when time allows).  A `maxDifference=1` mismatch along the edge of a rotated shape seems pretty trivially fine from a fuzzy-annotation perspective, and unrelated to what the test is trying to test for (which is does the transform work *at all*; and it does work.)