On 14.12.2016 02:46, Ronald S. Bultje wrote: > Not wanting to discourage you, but I wonder if there's really a point to > this...?
These patches are prerequisites for enforcing the validity of codec parameters [1]. > I don't see how the user experience changes. Users won't see ffmpeg claiming nonsense bit rates like -1184293205235990 kb/s anymore. > This isn't specifically intended at this patch, but rather at the sort of > rabbit hole this change might lead to, I have a pretty good map of this rabbit hole (i.e. lots of samples triggering UBSan errors) and one day I might try to dig it up, but for now I'm limiting myself to the codec parameters. > which would cause the code to be uber-full of such checks, none of which > really have any significance. But maybe others disagree... Not relying on undefined behavior is a significant improvement. And doing these checks consequently where the values are set makes it possible for other code to rely on their validity without further checks. Best regards, Andreas 1: https://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20161026/c0151e90/attachment.patch _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel