Bug 5518

Created attachment 4248 [details]
testcase

Created attachment 4249 [details]
rendered with the png backend

As you can see the test is rendered correctly with the png backend.

Created attachment 4250 [details]
result using xlib

Created attachment 4251 [details]
also using xlib

Sometimes I see this as a result of the testcase.

Just tested without XRENDER and it is looking correctly then.

Anyone looking into this?

Ubuntu seems afected:
https://launchpad.net/distros/ubuntu/+source/xserver-xorg-driver-ati/+bug/34435
https://launchpad.net/distros/ubuntu/+source/xserver-xorg-driver-ati/+bug/38198

Created attachment 8144 [details]
the gradients are not rendered in Gnome

OK, Benjamin, I'm not yet too familiar with this.
Do you think this is the bug I'm also experiencing with my icons?
For everybody, who's not one of my brothers, I before attached an SVG (at least
I think I did) that is rendered badly in Gnome and what I blame Cairo for.
If this isn't the bug, is it Cairo at all?

As for my _part_ of the bug, I figured out, that it is transformation matrices
of gradients that cause it.
to give you an example:

<linearGradient
       inkscape:collect="always"
       xlink:href="#linearGradient3813"
       id="linearGradient4074"
       gradientUnits="userSpaceOnUse"
gradientTransform="matrix(3.786847e-6,1.045166e-6,-1.045166e-6,3.786847e-6,-53.0238,-21.8569)"
       x1="37179696"
       y1="16832558"
       x2="19937692"
       y2="5225754" />

without the gradientTransform the display is just like in inkscape, but
obviously it's not the correct dislay.

This is almost certainly a server bug, not a Cairo bug (per se). So 
if any reporters can still reproduce this,the relevant information
is:

 - What version of the X server
 - What driver
 - What driver version
 - Any driver options? (For example, is EXA enabled instead of
   XAA)

As a data point, the test case works fine for me with git Xorg as
of 2008-04-01, with the patches from bug 15333 and bug 15345 applied
on an ATI rv350 using EXA.

(From the age of the bug I suspect that those EXA-specific patches
aren't relevant to the original bug report, though they do cause misrendering of 
cairo gradients.)

Old bug in the ATI driver.