Created attachment 8712 [details]
Output of the PDF backend.

Created attachment 8713 [details]
Output of the PS backend.

Forgot to mention how I render and how I create the pattern:

pattern creation:

            pixbuf_swap_red_and_blue(image->pixbuf);

            image->surface = cairo_image_surface_create_for_data(
                gdk_pixbuf_get_pixels(image->pixbuf), CAIRO_FORMAT_ARGB32,
                gdk_pixbuf_get_width(image->pixbuf), gdk_pixbuf_get_height(image->pixbuf),
                gdk_pixbuf_get_rowstride(image->pixbuf));

            image->pattern = cairo_pattern_create_for_surface(image->surface);
            cairo_pattern_set_extend(image->pattern, CAIRO_EXTEND_NONE);
            cairo_pattern_set_filter(image->pattern, CAIRO_FILTER_BEST);


rendering sequence for the image:

        cairo_save(cr);

        cairo_translate(cr, width/2, height/2);
        cairo_rotate(cr, M_PI * image->orientation/180);

        cairo_scale(cr, sx, sy);
        cairo_translate(cr, ix, iy);
        cairo_set_source(cr, image->pattern);

        if (image->opacity < 1.0) {
            cairo_paint_with_alpha(cr, image->opacity);
        } else {
            cairo_paint(cr);
        }

        cairo_restore(cr);

Created attachment 8721 [details]
Primitive test image

Created attachment 8723 [details]
Minimal test case

Created attachment 8724 [details]
Correct output of the image backend for the testcase.

Created attachment 8725 [details]
Correct output of the Postscript backend for the testcase.

Created attachment 8726 [details]
Incorrect output of the PDF backend for the testcase.

Carl, isn't this supposed to be fixed?

(In reply to comment #9)
> Carl, isn't this supposed to be fixed?

Yes, it was supposed to be---that's why I was surpised to find this bug here, (and very surprised because the bug report hadn't gotten mailed out to the cairo-bugs list when the bug was originally submitted).

I haven't looked closely at this yet to see what we missed.

-Carl

Reading in the release notes that this issue shall be fixed reminded me to check if I can deavivate abuse of the image backend in my code.

Created attachment 8947 [details]
Correct output showing things are working in 1.3.15

I just verified that yes, indeed, this bug is already fixed in the 1.3.15 series, (see newly attached output.pdf).

Thanks for the report,

-Carl

This is fixed in 1.3.15 already and the fix will of course be in 1.4.0.

-Carl