On Tue, Feb 6, 2018 at 5:40 PM, Josh Allmann <joshua.allm...@gmail.com> wrote: > If the encoder has its avctx fields set by > avcodec_parameters_to_context, then the extradata is deep-copied.
But it shouldn't have. Thats not a common pattern. What source codec parameters would you use for an encoder? What if an encoder doesn't actually generate any extradata (in audio, very few actually do), you could end up writing entirely unrelated extradata (say aac extradata with an ac3 stream), which results in all sorts of weird files. This could be repeated for all sorts of odd fields which are not necessarily used by every encoder. For encoding, you should allocate a context, and set the relevant properties manually (for audio, mostly sample rate and channel count). Everything else should remain empty. Its not intended for another context to be copied into an encoder context. Every field is documented who gets to write to it, and for extradata for encoding its the encoder, not the user, so if you write into it for an encoding context, you are using the API wrong. - Hendrik _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
