[poppler] Radial shading in poppler

[Thomas Freitag]

Am 23.01.2011 19:09, schrieb Albert Astals Cid:
> A Diumenge, 23 de gener de 2011, Andrea Canciani va escriure:
>> On Wed, Jan 19, 2011 at 8:22 PM, Albert Astals Cid<aacid at kde.org>  wrote:
>>> A Dimecres, 19 de gener de 2011, Thomas Freitag va escriure:
>>>> I don't know, if we need to wait until Albert commit my last
>>>> patch, at least we need it, when I have to call a different method than
>>>> getColor.
>>> FWIW I don't think i'll be able to finish the regtesting of your patch
>>> before the weekend.
>> I have a working implementation of the simple approach suggested by
>> Thomas, I'm now working on improving it as I suggested.
>>
>> I had a look at the two radial gradient implementations and I would
>> advise against having both of them in the code with one #ifdef'd out.
>> If you think that the bresenham implementation should stay in the
>> code, it would be better to at least always compile it, so that it will
>> not bitrot too much. I think that anyway independent implementations
>> (Adobe Reader, Ghostscript) are to be preferred for regtesting,
>> because they are not affected by bugs shared in the two poppler
>> implementations (example: the bug in the range computation in
>> GfxState.cc)
After all that discussion and keeping in mind You're working on an 
improvement, I've no problem when we remove the #ifdef'd code. You can 
do that with your improvement patch, or Albert can do it before 
committing my latest patch. I agree that we probably don't need it anymore.
>> The newer implementation is more complicated than it is actually
>> needed (in particular in the 5 nested if's part). If you are willing to
>> use this implementation, I can clean it up a little.
I've no problem if You simplify my code. Please do that with Your 
improvement patch.
> I'm willing to use wathever you guys give me that is better than we have and
> produces no regressions ;-)
Sorry once again for the inconvenience  I caused the last weeks, Albert. 
It wasn't such easy to implement it, but You should at least agree, that 
also the old Gfx implementation has a lot of regressions. But please 
test my "newest" patch, hopefully this runs without any problems through 
Your regression test. I'm quite sure, that it will, or at least we only 
will have little problems with the quite complicated if's part, but 
after all the weeks with a lot of work I first want to see success. We 
can improve it (and I think we should do that to have always the best 
speed for this quality, but this is not caused by the if's part) later.

Thomas
> Albert
>
>> Andrea
>> _______________________________________________
>> poppler mailing list
>> poppler at lists.freedesktop.org
>> http://lists.freedesktop.org/mailman/listinfo/poppler
> _______________________________________________
> poppler mailing list
> poppler at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/poppler
>
> .
>