[cairo] alpha blending

[Bill Spitzak]

I'm pretty certain the above proposal is for modifying the color values, 
not the alpha.

What I was specifically saying is that treatment of the alpha does not 
change, it should already be considered linear.

Premultiplied is a problem as that is a combination of both a gamma 
color and a linear alpha, and it is difficult to say what space the 
multiplication was done in.

On 03/25/2015 12:33 PM, James Cloos wrote:
>>>>>> "BS" == Bill Spitzak <spitzak at gmail.com> writes:
>
> BS> There are a couple problems:
>
> BS> First there is "coverage" values which are calculated by the filling
> BS> algorithm. These in fact are not gamma corrected, .5 should mean
> BS> .5. All masks and alpha channels are coverage values
>
> In case I wasn't clear, the bug I referenced (and therefor my query
> here) were only about changing the way A is combined, and leaving the
> R, G & B as they are.
>
> Would that still lead to the issues you noted were the existing OPs
> changed?
>
> That being asked, the bug's proposal, then, is little different than
> converting everything to linear on input and back to gamma values on
> output, yes?
>
> Except that cairo explicitly does not treat alpha as gamma-corrected
> because unlike PDF it uses the one value for a combination of coverage
> and transparancy?
>
> Therefore gamma-correcting alpha is contrary to cairo's design?
>
> -JimC
>