2016-09-26 8:42 GMT+02:00 Jan Ekstrom <jee...@gmail.com>: > On Sep 26, 2016 09:29, "Carl Eugen Hoyos" <ceffm...@gmail.com> wrote: >> >> 2016-09-26 1:52 GMT+02:00 Jan Ekström <jee...@gmail.com>: >> >> If this fixes anything (I am not sure I understand: Are you changing a >> type which introduces a possible undefined behaviour, so you add an >> additional check at the same time?) it should be part of above new >> patch imo (or an independent patch). > > The specification notes that width and height are uint32 in MXF. Thus the > old type of "int" was incorrect. The width and height of codecpar are "int" > so by fixing the type it can overflow (you'd get negative values), and thus > have to be capped. > > The old code would have had thus an overflow happen naturally in the > case of a large enough unsigned integer as the value would have been > read into an int.
I thought no overflow happens in that case but it does not really matter: FFmpeg ignores even 16bit width / height (for msb set). Carl Eugen _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
