Mans Rullgard <[EMAIL PROTECTED]> added the comment: Michael Niedermayer <[EMAIL PROTECTED]> writes:
>> Michael, in r7562 you made some changes to the bounds checking in the >> header parsing. Care to comment? > > r7562 was just intended as simplification it wasnt intended to change > actual behavior. and i think it also does not change actual behavior > for valid streams (and for invalid streams behavor change requires a large > number of things to be true ...) > the idea behind r7562 was that checking after every read byte complicates > the code alot and simply checking at the end would achive the same, that > is (in the current code at least) seek back to after the last start code > and search for the next start code > > is there a problem with r7562 ? No, there isn't. I'm asking since the bad files could be played if care was taken to only read the header_len bytes from the header, regardless of what the flags say. > as far is i understand this bug is caused by the r7853 code causing a > correct detection of an invalidly set flag and then trying to resync from > the last startcode, while e7852 did ignore the invalid flag Yes, that's what's happening. I'm happy to leave it this way, and fail to play broken files. However, you are usually of the opinion that everything that can be played at all should be supported, even if the files are invalid. It's your call. ______________________________________________________ FFmpeg issue tracker <[EMAIL PROTECTED]> <https://roundup.mplayerhq.hu/roundup/ffmpeg/issue198> ______________________________________________________
