[poppler] [PATCH] Improve linear gradients in cairo backend

[Albert Astals Cid]

A Dilluns, 27 de juliol de 2009, Carlos Garcia Campos va escriure:
> El dom, 26-07-2009 a las 20:17 +0200, Albert Astals Cid escribió:
> > A Dissabte, 25 de juliol de 2009, Carlos Garcia Campos va escriure:
> > > Hi all,
> > >
> > > when drawing gradients with cairo, we don't need to fill a path for
> > > every region, we can just fill the bounding box with the gradient
> > > pattern, but this only works when shading->getExtend0() ==
> > > shading->getExtend1(), since we can use CAIRO_EXTEND_NONE when they are
> > > FALSE and CAIRO_EXTEND_PAD when they are TRUE. When they are different,
> > > we could just use the current approach, adding a path for every region.
> > > This is exactly what the attached patch does.
> >
> > shading->getExtend0() == shading->getExtend1() in Gfx.cc code seems like
> > exposing a Cairo bug/feature to the rest of the world and doesn't seem a
> > good idea imho.
>
> yes, indeed, the whole useFillSolorStop() thing is a little bit tricky
> to avoid duplicate code.
>
> > What about a making useFillColorStop have shading as input parameter and
> > needFill as output parameter?
>
> or it might be an output parameter of axialShadedFill() since it doesn't
> depend on every useFillColorStop() call, and shading is already a
> parameter of axialShadedFill(). It still looks weird to me, though.

Another variant, switch axialShadedFill from returning GBool to return

enum axialShadedFillResult
{
    AxialShadingDone,
    AxialShadingNotDoneNeedsFill,
    AxialShadingNotDoneDoesNotNeedFill
}

What do you think?

Albert

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