[Pixman] Radial gradients (again)

[Andrea Canciani]

On Mon, Oct 18, 2010 at 9:32 PM, Soeren Sandmann <sandmann at daimi.au.dk> wrote:
> Andrea Canciani <ranma42 at gmail.com> writes:
>
>> I don't remember the timeline for pixman 0.20,
>
> Basically, I wanted to get a new release out tomorrow and call it a
> release candidate, but it looks like that's not going to happen.
>
> I think pixman 0.20 by now is feature frozen, and the next development
> release will become 0.20 unless major bugs turn up.
>
> The one thing that could still make it is the linear gradient fix. As
> I recall, there was some issues about the classify method. If someone
> wants to take a look at that patch:
>
> http://cgit.freedesktop.org/~ranma42/pixman/commit/?h=wip/linear-master&id=7a9433f3942cf4f0c929fb3907d89af87a2d477c
>
> it would be very helpful.

I just pushed another branch to cleanup a little that commit
(whitespace and line length fixes) and to do the same change
in the classification function:

http://cgit.freedesktop.org/~ranma42/pixman/log/?h=wip/linear-master2

This branch passes pixman and cairo tests (well, it does not cause
regressions and fixes linear-gradient-large). The classification
is based on the same code as the rasterization (thus I'm quite
confident it should be correct) but might be suboptimal for
some projective matrices (if a matrix is projective, it will never
try to return CLASS_HORIZONTAL).

I didn't remove the VERTICAL flag from pixman-private.h even
if it is not used anymore, but that change should be trivial (if
desired).

Andrea