This patch set is a little more involved than the last. The basic idea
is to reduce the usage of frame_size, which has loosely-defined usage
and is set in various places. It's practical use is primarily for
duration guessing in libavformat. For demuxing, this duration guess is
only needed to fill-in missing timestamps from demuxers which do not set
them.
The first part of the patch set consolidates reliable calculation of
packet duration based on codec_id and basic AVCodecContext params into a
single function. While this is not ideal in terms of maintenance when we
add new codecs, it avoids such hacks as forcing
avformat_find_stream_info() to wait for the decoder to possibly set
frame_size for select codecs in order to fill-in missing timestamps and
not break stream copy and/or seeking.
The second goal is related to muxing. In the muxer we want to avoid
using frame_size for muxer time_base and/or header fields indicating
packet duration because, when encoding, frame_size does not always
directly correspond to the output packet duration. Rather, it is used to
request the number of samples to be sent to the encoder (which may
decide to delay, split, or combine input frames). So instead, it makes
more sense to only use frame_size as a backup value when there is no
known constant frame size for the codec_id being muxed.
Questions and suggestions are welcome!
Thanks,
Justin
Justin Ruggles (22):
avcodec: add av_get_exact_bits_per_sample() function
avcodec: add av_get_audio_frame_duration() function.
riffenc: use av_get_audio_frame_duration()
rtpenc: use av_get_audio_frame_duration() for max_frames_per_packet
swfenc: use av_get_audio_frame_duration() instead of
AVCodecContext.frame_size
avformat: do not require frame_size in avformat_find_stream_info()
for AAC
avformat: do not require frame_size in avformat_find_stream_info()
for MP1/2/3
avformat: do not require frame_size in avformat_find_stream_info()
for CELT
rdt: remove workaround for infinite loop with aac
ape: do not set AVCodecContext.frame_size.
mov: do not set AVCodecContext.frame_size
aiffdec: do not set AVCodecContext.frame_size
amr demuxer: do not set AVCodecContext.frame_size.
siff: do not set AVCodecContext.frame_size
alsdec: do not set AVCodecContext.frame_size
libopencore-amr: do not set AVCodecContext.frame_size
libspeexdec: do not set AVCodecContext.frame_size
dca_parser: parse the sample rate and frame durations
lavf: Use av_get_audio_frame_duration() in get_audio_frame_size()
avformat: do not fill-in audio packet duration in
compute_pkt_fields()
avformat: do not require a pixel/sample format if there is no decoder
cosmetics: reindent
doc/APIchanges | 6 ++
libavcodec/Makefile | 3 +-
libavcodec/alsdec.c | 1 -
libavcodec/avcodec.h | 20 +++++
libavcodec/dca.c | 47 +----------
libavcodec/dca_parser.c | 80 ++++++++++++++++++-
libavcodec/libopencore-amr.c | 1 -
libavcodec/libspeexdec.c | 3 -
libavcodec/utils.c | 176 ++++++++++++++++++++++++++++++++++++++---
libavformat/aiffdec.c | 30 ++++---
libavformat/amr.c | 2 -
libavformat/ape.c | 1 -
libavformat/avformat.h | 1 +
libavformat/mov.c | 13 ---
libavformat/rdt.c | 3 -
libavformat/riff.c | 28 ++++++-
libavformat/rtpenc.c | 11 ++-
libavformat/siff.c | 8 +-
libavformat/swfenc.c | 6 +-
libavformat/utils.c | 135 ++++++++++++++++++-------------
tests/ref/fate/dxa-feeble | 1 +
tests/ref/fate/wmv8-drm-nodec | 40 +++++-----
tests/ref/seek/adpcm_qt_aiff | 36 ++++----
23 files changed, 450 insertions(+), 202 deletions(-)
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel
