Hello, I've noticed a somewhat subtle problem with mplayer's direct rendering on my osx ppc machine with corevideo output (using latest svn, although the problem has been around for quite a while). I'm guessing it's platform-specific (or big-endian-specific) because otherwise it's noticeable enough that it probably would have been caught and fixed long ago. :-) The problem presents itself as horizontal stripes of color at the right edge of the image. They are 16 pixels wide and occur on every other line of chroma (two pixels tall). Two cropped examples: http://img44.imageshack.us/img44/4952/mplayerdr1bug1.png http://img24.imageshack.us/img24/1373/mplayerdr1bug2.png Notes: (1) Clearly the chroma from the next-to-last column of 16 pixels is affecting the last column of 16. (2) The second example image shows that the luma is also affected somehow, because the background is black. It seems like there's some kind of compositing effect. (3) It only occurs when the movie width is congruent to 16 modulo 32 (I created test movies with ffmpeg to check this). (4) It's not codec-specific (I tested mpeg-2 and mpeg-4). (5) As I've already implied, the problem disappears if the direct rendering flag "do_dr1" is not set (in libmpcodecs/vd_ffmpeg.c). Questions: (1) Does anyone recognize these symptoms? (2) Any suggestions for further tests? (3) Where do I start if I want to look at buffer contents and compare the behavior of do_dr1=0 and do_dr1=1? best, Ivan
