On 1/5/2024 1:42 PM, Anton Khirnov wrote:
This avoids the requirement to always have a decoder context. --- fftools/ffmpeg_demux.c | 43 +++++++++++++++++++++--------------------- 1 file changed, 21 insertions(+), 22 deletions(-)diff --git a/fftools/ffmpeg_demux.c b/fftools/ffmpeg_demux.c index cacdc76a71..892094c512 100644 --- a/fftools/ffmpeg_demux.c +++ b/fftools/ffmpeg_demux.c @@ -899,19 +899,18 @@ static int choose_decoder(const OptionsContext *o, AVFormatContext *s, AVStream } }-static int guess_input_channel_layout(InputStream *ist, int guess_layout_max)+static int guess_input_channel_layout(InputStream *ist, AVCodecParameters *par, + int guess_layout_max) { - AVCodecContext *dec = ist->dec_ctx; - - if (dec->ch_layout.order == AV_CHANNEL_ORDER_UNSPEC) { + if (par->ch_layout.order == AV_CHANNEL_ORDER_UNSPEC) { char layout_name[256];- if (dec->ch_layout.nb_channels > guess_layout_max)+ if (par->ch_layout.nb_channels > guess_layout_max) return 0; - av_channel_layout_default(&dec->ch_layout, dec->ch_layout.nb_channels); - if (dec->ch_layout.order == AV_CHANNEL_ORDER_UNSPEC) + av_channel_layout_default(&par->ch_layout, par->ch_layout.nb_channels); + if (par->ch_layout.order == AV_CHANNEL_ORDER_UNSPEC) return 0; - av_channel_layout_describe(&dec->ch_layout, layout_name, sizeof(layout_name)); + av_channel_layout_describe(&par->ch_layout, layout_name, sizeof(layout_name)); av_log(ist, AV_LOG_WARNING, "Guessed Channel Layout: %s\n", layout_name); } return 1; @@ -1145,16 +1144,6 @@ static int ist_add(const OptionsContext *o, Demuxer *d, AVStream *st) ist->user_set_discard = ist->st->discard; }- ist->dec_ctx = avcodec_alloc_context3(ist->dec);- if (!ist->dec_ctx) - return AVERROR(ENOMEM); - - ret = avcodec_parameters_to_context(ist->dec_ctx, par); - if (ret < 0) { - av_log(ist, AV_LOG_ERROR, "Error initializing the decoder context.\n"); - return ret; - } - if (o->bitexact) av_dict_set(&ist->decoder_opts, "flags", "+bitexact", AV_DICT_MULTIKEY);@@ -1181,7 +1170,7 @@ static int ist_add(const OptionsContext *o, Demuxer *d, AVStream *st)case AVMEDIA_TYPE_AUDIO: { int guess_layout_max = INT_MAX; MATCH_PER_STREAM_OPT(guess_layout_max, i, guess_layout_max, ic, st); - guess_input_channel_layout(ist, guess_layout_max); + guess_input_channel_layout(ist, par, guess_layout_max); break; } case AVMEDIA_TYPE_DATA: @@ -1190,7 +1179,7 @@ static int ist_add(const OptionsContext *o, Demuxer *d, AVStream *st) MATCH_PER_STREAM_OPT(fix_sub_duration, i, ist->fix_sub_duration, ic, st); MATCH_PER_STREAM_OPT(canvas_sizes, str, canvas_size, ic, st); if (canvas_size) { - ret = av_parse_video_size(&ist->dec_ctx->width, &ist->dec_ctx->height, + ret = av_parse_video_size(&par->width, &par->height, canvas_size); if (ret < 0) { av_log(ist, AV_LOG_FATAL, "Invalid canvas size: %s.\n", canvas_size); @@ -1201,8 +1190,8 @@ static int ist_add(const OptionsContext *o, Demuxer *d, AVStream *st) /* Compute the size of the canvas for the subtitles stream. If the subtitles codecpar has set a size, use it. Otherwise use the maximum dimensions of the video streams in the same file. */ - ist->sub2video.w = ist->dec_ctx->width; - ist->sub2video.h = ist->dec_ctx->height; + ist->sub2video.w = par->width; + ist->sub2video.h = par->height; if (!(ist->sub2video.w && ist->sub2video.h)) { for (int j = 0; j < ic->nb_streams; j++) { AVCodecParameters *par1 = ic->streams[j]->codecpar; @@ -1226,6 +1215,16 @@ static int ist_add(const OptionsContext *o, Demuxer *d, AVStream *st) default: av_assert0(0); }+ ist->dec_ctx = avcodec_alloc_context3(ist->dec);+ if (!ist->dec_ctx) + return AVERROR(ENOMEM); + + ret = avcodec_parameters_to_context(ist->dec_ctx, par); + if (ret < 0) { + av_log(ist, AV_LOG_ERROR, "Error initializing the decoder context.\n"); + return ret; + } + ist->par = avcodec_parameters_alloc(); if (!ist->par) return AVERROR(ENOMEM);
This does not fix the issue of the guessed channel layout not being present on the output stream, but might be a change in the right direction.
before:
$ ./ffmpeg -i ~/samples/wav/200828-005.wav out.wav [aist#0:0/pcm_s16le @ 00000252b1d2d420] Guessed Channel Layout: stereo Input #0, wav, from '../samples/wav/200828-005.wav': Duration: 00:00:12.80, bitrate: 1536 kb/s Stream #0:0: Audio: pcm_s16le ([1][0][0][0] / 0x0001), 48000 Hz, 2 channels, s16, 1536 kb/s Stream mapping: Stream #0:0 -> #0:0 (pcm_s16le (native) -> pcm_s16le (native)) Press [q] to stop, [?] for help Output #0, wav, to 'out.wav': Stream #0:0: Audio: pcm_s16le ([1][0][0][0] / 0x0001), 48000 Hz, stereo, s16, 1536 kb/s $ ./ffmpeg -i out.wav [aist#0:0/pcm_s16le @ 000001d616b1d3e0] Guessed Channel Layout: stereo Duration: 00:00:12.80, bitrate: 1536 kb/s Stream #0:0: Audio: pcm_s16le ([1][0][0][0] / 0x0001), 48000 Hz, 2 channels, s16, 1536 kb/s
After:
$ ./ffmpeg -i ../samples/wav/200828-005.wav out.wav [aist#0:0/pcm_s16le @ 0000024603c9d420] Guessed Channel Layout: stereo Input #0, wav, from '../samples/wav/200828-005.wav': Duration: 00:00:12.80, bitrate: 1536 kb/ Stream #0:0: Audio: pcm_s16le ([1][0][0][0] / 0x0001), 48000 Hz, stereo, s16, 1536 kb/s Stream mapping: Stream #0:0 -> #0:0 (pcm_s16le (native) -> pcm_s16le (native)) Press [q] to stop, [?] for help Output #0, wav, to 'out.wav': Stream #0:0: Audio: pcm_s16le ([1][0][0][0] / 0x0001), 48000 Hz, stereo, s16, 1536 kb/s $ ./ffmpeg -i out.wav [aist#0:0/pcm_s16le @ 000001a1467ed3e0] Guessed Channel Layout: stereo Duration: 00:00:12.80, bitrate: 1536 kb/s Stream #0:0: Audio: pcm_s16le ([1][0][0][0] / 0x0001), 48000 Hz, stereo, s16, 1536 kb/s
So the printed output shows the guessed layout for the input stream, but the written output stream is still unspec.
_______________________________________________ ffmpeg-devel mailing list [email protected] https://ffmpeg.org/mailman/listinfo/ffmpeg-devel To unsubscribe, visit link above, or email [email protected] with subject "unsubscribe".
