The poppler cairo backend produces only blank pages (i.e. evince is unusable) on powerpc if used with the image backend, it works fine with the xlib backend. The first thing poppler does for each page is to clip to the paper area. Removing that gives a non blank page. I'll attach the (image backend) output of the clip-nesting, clip-operator and clip-twice tests.
Created attachment 2931 [details] clip-nesting-image-out.png
Created attachment 2932 [details] clip-operator-image-out.png
Created attachment 2933 [details] clip-twice-image-out.png
If you aren't testing with Cairo CVS, can you test with Cairo CVS?
This was with a fresh CVS checkout.
Correction: the test files were generated with a fresh checkout. I haven't tested poppler with this version of cairo.
I get a bunch of failures on my Mac, although I don't think I've seen these blank-image failures.
clip-nesting is not completely blank, some parts are white, some are transparent. I didn't produce these on a Mac, but a G4 Pegasos II (shouldn't matter, I guess). I'm using Ubuntu Breezy. I looked at this problem after being notified of this: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=315133 Does this pdf http://bugs.debian.org/cgi-bin/bugreport.cgi/evince-test.pdf?bug=315133;msg=15;att=2 work on your Mac with evince and poppler using the cairo backend? Can evince/poppler-cairo display pdfs at all on your Mac (since the debian bug has been filed, a bug where poppler dropped all winding-rule clip operations has been fixed, resulting in blank pages for every pdf file)?
OK, we're seeing the same failures then, I think.
Fixed in HEAD. The problem was in pixman_fill_rect_8bpp. It was converting the value of a 32 bit pixel to 8 bits like this: *(char*)pixel When it really meant to do: (char) *pixel
Move bugs against "cvs" version to "0.9.3" so we can remove the "cvs" version.
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.