2012/2/9 Anton Khirnov <[email protected]>: > I'm basically ok with allowing the caller to override the duration, but > we should make the API semantics a bit clearer first. If the user set > the duration when encoding, do we want to > a) assume he knows what he's doing and write this duration always > OR > b) only use this duration if the file is not seekable. > > If we pick a), then your patch also needs to test for file seekability, > since input might be shorter than the supplied -t value, so we end up > writing wrong duration even for seekable files.
In case of flvenc, on write_trailer it seeks back and rewrites duration if it can. > (though we might want to change the semantics of -t to duplicate frames > /pad with silence to get requested duration). I wonder what the duration difference we might need to fix, as long as -t option serves its initial purpose of setting target file duration. -- Andrey Utkin _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
