[FFmpeg-devel] [patch] mjpeg restart marker parsing

[Tom Harper]

re:so why do you check against next_restart_marker at all?

If you don't have the next valid restart marker then it is probable that 
the bitstream is corrupted-   I would imagine they invented this scheme 
so that you could detect if there was data loss- otherwise why would 
there be more than one restart marker type?  To be flexible this patch 
allows the decoding to proceed with the incorrect next marker- which it 
should do- but it probably means that there will be artifacts in the 
decoded image.  So in this case I would want ffmpeg to make its best 
effort but also to tell me that something bad happened, such that I 
won't blame the decoder for something the encoder did wrong, or that 
happened in transport.

Tom