[cairo] Use 8-bit colour values internally

[Bill Spitzak]

Jason Dorje Short wrote:

>> -    color->alpha_short = color->alpha * CAIRO_COLOR_ONE_MINUS_EPSILON;
>> +    color->alpha_char = floor (color->alpha * 255.0 + 0.5);
> 
> Looks like the uniform-distribution hack here is trying to do the same 
> thing you're doing with rounding (otherwise 0.{9} would never give 
> COLOR_MAX, always COLOR_MAX-1).  Clearly rounding is more accurate 
> though (if I understand rightly what this code does).

Actually the uniform-distribution hack produces more desirable results 
in that the range 0.0-1.0 is divided into equal-sized bins that land in 
each resulting pixel value. The rounding version makes the bin for the 
largest and smallest pixel values only half-size.

PS: having made this mistake personally many times, I should know. Use 
the multiply and floor, not rounding.