On Thu, Mar 3, 2016 at 12:37 AM, Ronald S. Bultje <rsbul...@gmail.com> wrote: > Hi, > > On Wed, Mar 2, 2016 at 2:32 PM, foo86 <fooba...@gmail.com> wrote: > >> Limit the maximum length of unary part of Rice code by the number of >> available bits instead of using an arbitrary constant that happens to be >> just large enough to work. >> >> This effectively limits amount of data that can be overread per segment >> by maximum length of binary code per sample multiplied by maximum >> segment size. >> >> Increase size of padding area after the end of input buffer according to >> this limit and add some extra overread checks to make reading past end >> of buffer safe. >> --- >> libavcodec/dca_xll.c | 25 +++++++++++++++---------- >> libavcodec/dcadec.h | 2 +- >> 2 files changed, 16 insertions(+), 11 deletions(-) > > > Completely random comment, but isn't it easier to just enable the safe > bitstream reader? >
This is actually enabled by default, so it probably should never have overread anywhere, but if the code can be made a bit safer in itself it probably does no harm either way. - Hendrik _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
