From: Michel L. <wa...@zo...> - 2001-08-01 20:34:25
|
On Wed, Aug 01, 2001 at 08:26:18PM +0100, James Courtier-Dutton wrote: > I attach the output I get from gdb. > stream not demultiplexed ? > stream not demultiplexed ? > stream not demultiplexed ? This tells me that the stream was not correctly demuxed, and the startcode search in libmpeg2 found some startcodes that should never be present in a demuxed stream. > Program received signal SIGSEGV, Segmentation fault. > [Switching to Thread 2051 (LWP 10084)] > MC_avg_y8_mmxext (dest=0x417c9ce0 "", > ref=0x418fed14 <Address 0x418fed14 out of bounds>, stride=720, height=4) > at motion_comp_mmx.c:625 > 625 movq_m2r (*ref, mm0); One of the common things that happen right now if libmpeg2 sees a corrupted stream, is that it will try to apply motion compensation with a corrupted motion vector, and copy pixels from somewhere outside the reference frame, which can cause segfaults if that memory is not mapped in. This is what happened here. But the bug is still somewhere in the demuxer (or it could be a corrupted transport stream too) -- Michel "Walken" LESPINASSE Of course I think I'm right. If I thought I was wrong, I'd change my mind. |