When you invert the coordinate system using PDFGraphics2D.scale(1,-1), to make it compatible with AWT's coordinate system, the inversion is not obeyed by the painter routine that processes unknown paints, see example below. That is, the raster is the same as if you didn't define the transform. See attached example, as well as current and expected output. I used gradient paint to force rasterization, so you must disable my recent patch (bug #39033) that prevents rasterization, to see the error.
Created attachment 18051 [details] test example
Created attachment 18052 [details] expected output
Created attachment 18053 [details] current incorrect output (with patch #39033 disabled)
resetting P2 open bugs to P3 pending further review
this is fixed and can be closed.
(In reply to comment #5) > this is fixed and can be closed. luis, do you have any information about which revision fixed it? or is it that you are unable to reproduce it now?
the thread of comments above gives the answer. the patch for bug 39033, which has been applied, also fixes this bug. I also ran the test case and the output is the expected.
(In reply to comment #7) > the thread of comments above gives the answer. the patch for bug 39033, which > has been applied, also fixes this bug. I also ran the test case and the output > is the expected. not really, but i tracked down the answer via bug 39033; see http://svn.apache.org/viewcvs?rev=392917&view=rev for fix