there is a very small chance that a stream which is damaged && unseekable && very small packet && damage to the flags && unlucky buffer state could have a valid packet after a damaged packet lost, but i think this is so unlikely that the simplification outweights it
mans if you disagree with any commit to mpeg.c just say so and ill revert it

Originally committed as revision 7562 to svn://svn.ffmpeg.org/ffmpeg/trunk