Message ID | 20220505230458.1024-1-gsun@roblox.com |
---|---|
State | New |
Headers | show |
Series | [FFmpeg-devel] avformat/oggparsevorbis: Fix oggvorbis duration parsing for small files | expand |
Context | Check | Description |
---|---|---|
yinshiyou/make_loongarch64 | success | Make finished |
yinshiyou/make_fate_loongarch64 | success | Make fate finished |
andriy/make_x86 | success | Make finished |
andriy/make_fate_x86 | success | Make fate finished |
On Thu, May 5, 2022 at 4:05 PM Guangyu Sun <sunguangyucn@gmail.com> wrote: > The decoded Data is not returned from the first frame; it must be used > to 'prime' the decode engine. The duration of the first packet should > be ignored also by the container, to make sure the total duration is > trimmed correctly. > > If the file has multiple pages for the data packets. The duration of the > last packet can be correctly adjusted without being affected by the first > packet. But if the first packet and the last packet are located in the > same page, the duration of the last packet will have an unwanted offset. > > This commit fixes https://trac.ffmpeg.org/ticket/6367. > > dd if=/dev/zero of=./silence.raw count=1 bs=500 > oggenc --raw silence.raw --output=silence.ogg > oggdec --raw --output silence.oggdec.raw silence.ogg > ffmpeg -codec:a libvorbis -i silence.ogg -f s16le -codec:a pcm_s16le > silence.libvorbis.ffmpeg.raw > ffmpeg -i silence.ogg -f s16le -codec:a pcm_s16le silence.native.ffmpeg.raw > ls -l *.raw > > Signed-off-by: Guangyu Sun <gsun@roblox.com> > --- > libavformat/oggparsevorbis.c | 11 ++++++++++- > 1 file changed, 10 insertions(+), 1 deletion(-) > > diff --git a/libavformat/oggparsevorbis.c b/libavformat/oggparsevorbis.c > index 289900f7cf..9a486c9fe8 100644 > --- a/libavformat/oggparsevorbis.c > +++ b/libavformat/oggparsevorbis.c > @@ -412,6 +412,7 @@ static int vorbis_packet(AVFormatContext *s, int idx) > struct ogg_stream *os = ogg->streams + idx; > struct oggvorbis_private *priv = os->private; > int duration, flags = 0; > + int first_page; > > if (!priv->vp) > return AVERROR_INVALIDDATA; > @@ -420,7 +421,8 @@ static int vorbis_packet(AVFormatContext *s, int idx) > * here we parse the duration of each packet in the first page and > compare > * the total duration to the page granule to find the encoder delay > and > * set the first timestamp */ > - if ((!os->lastpts || os->lastpts == AV_NOPTS_VALUE) && !(os->flags & > OGG_FLAG_EOS) && (int64_t)os->granule>=0) { > + first_page = !os->lastpts || os->lastpts == AV_NOPTS_VALUE; > + if (first_page && !(os->flags & OGG_FLAG_EOS) && > (int64_t)os->granule>=0) { > int seg, d; > uint8_t *last_pkt = os->buf + os->pstart; > uint8_t *next_pkt = last_pkt; > @@ -479,6 +481,13 @@ static int vorbis_packet(AVFormatContext *s, int idx) > flags = 0; > } > os->pduration = duration; > + /* Data is not returned from the first frame; it must be used to > 'prime' > + * the decode engine. The duration of the first packet should be > ignored > + * also by the container, to make sure the total duration is > trimmed > + * correctly. > + */ > + if (first_page && os->segp == 1) > + os->pduration = 0; > } > > /* final packet handling > -- > 2.30.1 > > Ping.
On Mon, May 9, 2022 at 11:19 AM Guangyu Sun <sunguangyucn@gmail.com> wrote: > On Thu, May 5, 2022 at 4:05 PM Guangyu Sun <sunguangyucn@gmail.com> wrote: > >> The decoded Data is not returned from the first frame; it must be used >> to 'prime' the decode engine. The duration of the first packet should >> be ignored also by the container, to make sure the total duration is >> trimmed correctly. >> >> If the file has multiple pages for the data packets. The duration of the >> last packet can be correctly adjusted without being affected by the first >> packet. But if the first packet and the last packet are located in the >> same page, the duration of the last packet will have an unwanted offset. >> >> This commit fixes https://trac.ffmpeg.org/ticket/6367. >> >> dd if=/dev/zero of=./silence.raw count=1 bs=500 >> oggenc --raw silence.raw --output=silence.ogg >> oggdec --raw --output silence.oggdec.raw silence.ogg >> ffmpeg -codec:a libvorbis -i silence.ogg -f s16le -codec:a pcm_s16le >> silence.libvorbis.ffmpeg.raw >> ffmpeg -i silence.ogg -f s16le -codec:a pcm_s16le >> silence.native.ffmpeg.raw >> ls -l *.raw >> >> Signed-off-by: Guangyu Sun <gsun@roblox.com> >> --- >> libavformat/oggparsevorbis.c | 11 ++++++++++- >> 1 file changed, 10 insertions(+), 1 deletion(-) >> >> diff --git a/libavformat/oggparsevorbis.c b/libavformat/oggparsevorbis.c >> index 289900f7cf..9a486c9fe8 100644 >> --- a/libavformat/oggparsevorbis.c >> +++ b/libavformat/oggparsevorbis.c >> @@ -412,6 +412,7 @@ static int vorbis_packet(AVFormatContext *s, int idx) >> struct ogg_stream *os = ogg->streams + idx; >> struct oggvorbis_private *priv = os->private; >> int duration, flags = 0; >> + int first_page; >> >> if (!priv->vp) >> return AVERROR_INVALIDDATA; >> @@ -420,7 +421,8 @@ static int vorbis_packet(AVFormatContext *s, int idx) >> * here we parse the duration of each packet in the first page and >> compare >> * the total duration to the page granule to find the encoder delay >> and >> * set the first timestamp */ >> - if ((!os->lastpts || os->lastpts == AV_NOPTS_VALUE) && !(os->flags & >> OGG_FLAG_EOS) && (int64_t)os->granule>=0) { >> + first_page = !os->lastpts || os->lastpts == AV_NOPTS_VALUE; >> + if (first_page && !(os->flags & OGG_FLAG_EOS) && >> (int64_t)os->granule>=0) { >> int seg, d; >> uint8_t *last_pkt = os->buf + os->pstart; >> uint8_t *next_pkt = last_pkt; >> @@ -479,6 +481,13 @@ static int vorbis_packet(AVFormatContext *s, int idx) >> flags = 0; >> } >> os->pduration = duration; >> + /* Data is not returned from the first frame; it must be used to >> 'prime' >> + * the decode engine. The duration of the first packet should be >> ignored >> + * also by the container, to make sure the total duration is >> trimmed >> + * correctly. >> + */ >> + if (first_page && os->segp == 1) >> + os->pduration = 0; >> } >> >> /* final packet handling >> -- >> 2.30.1 >> >> > Ping. > Ping again. This patch fixes an oggvorbis transcode issue. https://devforum.roblox.com/t/recent-changes-to-uploaded-audio-compressionprocessing-causing-timing-drift/1735440/10
diff --git a/libavformat/oggparsevorbis.c b/libavformat/oggparsevorbis.c index 289900f7cf..9a486c9fe8 100644 --- a/libavformat/oggparsevorbis.c +++ b/libavformat/oggparsevorbis.c @@ -412,6 +412,7 @@ static int vorbis_packet(AVFormatContext *s, int idx) struct ogg_stream *os = ogg->streams + idx; struct oggvorbis_private *priv = os->private; int duration, flags = 0; + int first_page; if (!priv->vp) return AVERROR_INVALIDDATA; @@ -420,7 +421,8 @@ static int vorbis_packet(AVFormatContext *s, int idx) * here we parse the duration of each packet in the first page and compare * the total duration to the page granule to find the encoder delay and * set the first timestamp */ - if ((!os->lastpts || os->lastpts == AV_NOPTS_VALUE) && !(os->flags & OGG_FLAG_EOS) && (int64_t)os->granule>=0) { + first_page = !os->lastpts || os->lastpts == AV_NOPTS_VALUE; + if (first_page && !(os->flags & OGG_FLAG_EOS) && (int64_t)os->granule>=0) { int seg, d; uint8_t *last_pkt = os->buf + os->pstart; uint8_t *next_pkt = last_pkt; @@ -479,6 +481,13 @@ static int vorbis_packet(AVFormatContext *s, int idx) flags = 0; } os->pduration = duration; + /* Data is not returned from the first frame; it must be used to 'prime' + * the decode engine. The duration of the first packet should be ignored + * also by the container, to make sure the total duration is trimmed + * correctly. + */ + if (first_page && os->segp == 1) + os->pduration = 0; } /* final packet handling
The decoded Data is not returned from the first frame; it must be used to 'prime' the decode engine. The duration of the first packet should be ignored also by the container, to make sure the total duration is trimmed correctly. If the file has multiple pages for the data packets. The duration of the last packet can be correctly adjusted without being affected by the first packet. But if the first packet and the last packet are located in the same page, the duration of the last packet will have an unwanted offset. This commit fixes https://trac.ffmpeg.org/ticket/6367. dd if=/dev/zero of=./silence.raw count=1 bs=500 oggenc --raw silence.raw --output=silence.ogg oggdec --raw --output silence.oggdec.raw silence.ogg ffmpeg -codec:a libvorbis -i silence.ogg -f s16le -codec:a pcm_s16le silence.libvorbis.ffmpeg.raw ffmpeg -i silence.ogg -f s16le -codec:a pcm_s16le silence.native.ffmpeg.raw ls -l *.raw Signed-off-by: Guangyu Sun <gsun@roblox.com> --- libavformat/oggparsevorbis.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-)