[cairo] [Patch] Some work on cairo-xcb

[Andrea Canciani]

On Fri, Dec 17, 2010 at 10:05 PM, Uli Schlachter <psychon at znc.in> wrote:
> Am 17.12.2010 21:43, Uli Schlachter wrote:
>> Am 17.12.2010 21:02, Uli Schlachter wrote:
>>> Chris Wilson said he didn't like the memory allocation that is added here, but I
>>> didn't succeed in avoid it (although I haven't searched too hard). Feel free to
>>> tell me how this could be done.
>>
>> A big thanks goes to Andrea Canciani for telling me how and criticizing some
>> more stuff. Attached is v2 of the first patch.
>
> Sorry for the spam, but here is v3 of the patch. The last version compared the
> wrong pointers to check if it has to do any cleanup
> "if (clip == &extents_region)" instead of "if (clip_region == &extents_region)".
>
> Hopefully I didn't break anything this time.

I tested your patches and I can confirm that 0001-...-v3 fixes
clip-{fill-eo,fill-nz,stroke}-unbounded tests and 0002 fixes clear-source.

Sorry for complaining about your patches again, but I think that the commit
messages could be improved.

For the 0001-...-v3, the first line should state what is the purpose of the
commit (i.e. correcting the extents computation) instead of the effect
(fixing the unbounded tests).

For the 0002, it would be nice to explain why the optimization is incorrect
NB: the problem is not actually just with source-bounded operations, the
code is incorrect because it might be replacing a none-extended black
surface with a solid color, which is not valid no matter what the operator
is (actually it would be ok for DEST and CLEAR, but you should never
get here if the op is one of them).

If you want to, you can add Reviewed-by: Andrea Canciani <ranma42 at gmail.com>
to the commit messages, but that's only valid for the code, the commit messages
can still be improved ;)

Andrea