On Wed, 18 Apr 2018 23:01:00 +0200 Nicolas George <[email protected]> wrote:
> wm4 (2018-04-18): > > There was a simple patch that would have solved that issue locally in > > the UDP code. > > This is not true. The patch would have fixed the author's use case but > would have broken all applications that rely on empty UDP packets. Since > it is about public API, we cannot assume anything about what > applications do. A solution is only acceptable if it works for all cases > and do not break existing applications. > Which applications (in context of FFmpeg) need to detect empty UDP packets? Which applications would be broken by a trivial patch in the UDP code? From what I remember, the problem was that empty packets were recognized as EOF, and this issue was fixed with a trivial patch (which was rejected). On the other hand, the current code still gives you no way to detect empty UDP packets, because the retry_wrapper thing in the avio generic code basically drops them. A solution that might have worked better (if empty packets really need to be detected) would have been returning a special error code, that applications could choose to interpret. _______________________________________________ ffmpeg-devel mailing list [email protected] http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
