From: Gyan Doshi <ffmpeg@gyani.pro> To: ffmpeg-devel@ffmpeg.org Subject: Re: [FFmpeg-devel] [PATCH] ffmpeg: add per-stream input option drop_changed Date: Sun, 16 Mar 2025 12:52:11 +0530 Message-ID: <c1ac6511-2c79-4413-ac2c-56bd5d78e75b@gyani.pro> (raw) In-Reply-To: <b6528a6f-fd16-5fed-cc07-2343627e1948@passwd.hu> On 2025-03-16 01:21 am, Marton Balint wrote: > > > On Sat, 15 Mar 2025, Gyan Doshi wrote: > >> This is a replacement in ffmpeg for the deprecated avcodec flag >> AV_CODEC_FLAG_DROPCHANGED. >> >> This option is meant to be used when the filtergraph should not be >> reinited upon input parameter changes as that leads to loss of state >> in the filtergraph potentially leading to broken or aborted output, >> e.g. inserting of silence with first_pts specified in aresample. >> >> Generally useful to avoid corrupted yet decodable packets in live >> streaming inputs. >> >> This option when enabled takes precedence over reinit_filters >> --- >> doc/ffmpeg.texi | 6 ++++++ >> fftools/ffmpeg.h | 2 ++ >> fftools/ffmpeg_demux.c | 13 ++++++++++++- >> fftools/ffmpeg_filter.c | 10 ++++++++++ >> fftools/ffmpeg_opt.c | 3 +++ >> 5 files changed, 33 insertions(+), 1 deletion(-) >> >> diff --git a/doc/ffmpeg.texi b/doc/ffmpeg.texi >> index fca220a334..a73db79f94 100644 >> --- a/doc/ffmpeg.texi >> +++ b/doc/ffmpeg.texi >> @@ -1373,6 +1373,12 @@ The properties where a change triggers >> reinitialization are, >> for video, frame resolution or pixel format; >> for audio, sample format, sample rate, channel count or channel layout. >> >> +@item -drop_changed[:@var{stream_specifier}] @var{integer} >> (@emph{input,per-stream}) >> +This boolean option determines whether a frame with differing frame >> parameters mid-stream >> +gets dropped instead of leading to filtergraph reinitialization, as >> that would lead to loss >> +of filter state. Generally useful to avoid corrupted yet decodable >> packets in live streaming >> +inputs. Default is false. >> + >> @item -filter_threads @var{nb_threads} (@emph{global}) >> Defines how many threads are used to process a filter pipeline. Each >> pipeline >> will produce a thread pool with this many threads available for >> parallel processing. >> diff --git a/fftools/ffmpeg.h b/fftools/ffmpeg.h >> index 86a3e10c6b..5869979214 100644 >> --- a/fftools/ffmpeg.h >> +++ b/fftools/ffmpeg.h >> @@ -232,6 +232,7 @@ typedef struct OptionsContext { >> SpecifierOptList filter_scripts; >> #endif >> SpecifierOptList reinit_filters; >> + SpecifierOptList drop_changed; >> SpecifierOptList fix_sub_duration; >> SpecifierOptList fix_sub_duration_heartbeat; >> SpecifierOptList canvas_sizes; >> @@ -262,6 +263,7 @@ enum IFilterFlags { >> IFILTER_FLAG_REINIT = (1 << 1), >> IFILTER_FLAG_CFR = (1 << 2), >> IFILTER_FLAG_CROP = (1 << 3), >> + IFILTER_FLAG_DROPCHANGED = (1 << 4), >> }; >> >> typedef struct InputFilterOptions { >> diff --git a/fftools/ffmpeg_demux.c b/fftools/ffmpeg_demux.c >> index b8171e336e..292b718f14 100644 >> --- a/fftools/ffmpeg_demux.c >> +++ b/fftools/ffmpeg_demux.c >> @@ -67,6 +67,7 @@ typedef struct DemuxStream { >> int reinit_filters; >> int autorotate; >> int apply_cropping; >> + int drop_changed; >> >> >> int wrap_correction_done; >> @@ -1099,7 +1100,8 @@ int ist_filter_add(InputStream *ist, >> InputFilter *ifilter, int is_simple, >> return AVERROR(ENOMEM); >> >> opts->flags |= IFILTER_FLAG_AUTOROTATE * !!(ds->autorotate) | >> - IFILTER_FLAG_REINIT * !!(ds->reinit_filters); >> + IFILTER_FLAG_REINIT * !!(ds->reinit_filters) | >> + IFILTER_FLAG_DROPCHANGED* !!(ds->drop_changed); >> >> return 0; >> } >> @@ -1410,6 +1412,15 @@ static int ist_add(const OptionsContext *o, >> Demuxer *d, AVStream *st, AVDictiona >> ds->reinit_filters = -1; >> opt_match_per_stream_int(ist, &o->reinit_filters, ic, st, >> &ds->reinit_filters); >> >> + ds->drop_changed = 0; >> + opt_match_per_stream_int(ist, &o->drop_changed, ic, st, >> &ds->drop_changed); >> + >> + if (ds->drop_changed && ds->reinit_filters) { >> + if (ds->reinit_filters > 0) >> + av_log(ist, AV_LOG_ERROR, "drop_changed and >> reinit_filters both set, which is contradictory; ignoring >> reinit_filters.\n"); > > I'd rather simply return an error, instead of continuing, unless there > is a very good reason to allow conflicting options and fall back to > one or the other. Done > >> + ds->reinit_filters = 0; >> + } >> + >> ist->user_set_discard = AVDISCARD_NONE; >> >> if ((o->video_disable && ist->st->codecpar->codec_type == >> AVMEDIA_TYPE_VIDEO) || >> diff --git a/fftools/ffmpeg_filter.c b/fftools/ffmpeg_filter.c >> index ef35c3988f..fddfbceec1 100644 >> --- a/fftools/ffmpeg_filter.c >> +++ b/fftools/ffmpeg_filter.c >> @@ -126,6 +126,7 @@ typedef struct InputFilterPriv { >> >> int eof; >> int bound; >> + uint64_t nb_dropped; >> >> // parameters configured for this input >> int format; >> @@ -2898,6 +2899,13 @@ static int send_frame(FilterGraph *fg, >> FilterGraphThread *fgt, >> } else if (ifp->downmixinfo_present) >> need_reinit |= DOWNMIX_CHANGED; >> >> + if (need_reinit && (ifp->opts.flags & IFILTER_FLAG_DROPCHANGED)) { >> + ifp->nb_dropped++; >> + av_log(fg, AV_LOG_DEBUG, "Avoiding reinit; dropping >> frame pts: %s bound for %s\n", av_ts2str(frame->pts), ifilter->name); > > Maybe it would make sense to use AV_LOG_WARNING for the first dropped > frame. Done. > >> + av_frame_unref(frame); >> + return 0; >> + } >> + >> if (!(ifp->opts.flags & IFILTER_FLAG_REINIT) && fgt->graph) >> need_reinit = 0; >> >> @@ -3140,6 +3148,8 @@ read_frames: >> ret = read_frames(fg, &fgt, fgt.frame); >> if (ret == AVERROR_EOF) { >> av_log(fg, AV_LOG_VERBOSE, "All consumers returned EOF\n"); >> + if (ifp->opts.flags & IFILTER_FLAG_DROPCHANGED) >> + av_log(fg, AV_LOG_VERBOSE, "Total changed input >> frames dropped : %"PRId64"\n", ifp->nb_dropped); > > I think you should make this AV_LOG_INFO and it is enough to print if > the number of dropped frames is non-zero. Level promoted. I've kept the drop count as it is a crude indicator of 'signal quality'. Regards, Gyan > >> break; >> } else if (ret < 0) { >> av_log(fg, AV_LOG_ERROR, "Error sending frames to >> consumers: %s\n", >> diff --git a/fftools/ffmpeg_opt.c b/fftools/ffmpeg_opt.c >> index 1f5e6050c8..6ec325f51e 100644 >> --- a/fftools/ffmpeg_opt.c >> +++ b/fftools/ffmpeg_opt.c >> @@ -1718,6 +1718,9 @@ const OptionDef options[] = { >> { "reinit_filter", OPT_TYPE_INT, OPT_PERSTREAM | >> OPT_INPUT | OPT_EXPERT, >> { .off = OFFSET(reinit_filters) }, >> "reinit filtergraph on input parameter changes", "" }, >> + { "drop_changed", OPT_TYPE_INT, OPT_PERSTREAM | >> OPT_INPUT | OPT_EXPERT, >> + { .off = OFFSET(drop_changed) }, >> + "drop frame instead of reiniting filtergraph on input >> parameter changes", "" }, >> { "filter_complex", OPT_TYPE_FUNC, OPT_FUNC_ARG | >> OPT_EXPERT, >> { .func_arg = opt_filter_complex }, >> "create a complex filtergraph", "graph_description" }, >> -- >> 2.46.1 > > Thanks, > Marton > _______________________________________________ > ffmpeg-devel mailing list > ffmpeg-devel@ffmpeg.org > https://ffmpeg.org/mailman/listinfo/ffmpeg-devel > > To unsubscribe, visit link above, or email > ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe". _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-devel To unsubscribe, visit link above, or email ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
prev parent reply other threads:[~2025-03-16 7:22 UTC|newest] Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top 2025-03-15 12:20 Gyan Doshi 2025-03-15 19:51 ` Marton Balint 2025-03-16 7:22 ` Gyan Doshi [this message]
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=c1ac6511-2c79-4413-ac2c-56bd5d78e75b@gyani.pro \ --to=ffmpeg@gyani.pro \ --cc=ffmpeg-devel@ffmpeg.org \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: link
Git Inbox Mirror of the ffmpeg-devel mailing list - see https://ffmpeg.org/mailman/listinfo/ffmpeg-devel This inbox may be cloned and mirrored by anyone: git clone --mirror https://master.gitmailbox.com/ffmpegdev/0 ffmpegdev/git/0.git # If you have public-inbox 1.1+ installed, you may # initialize and index your mirror using the following commands: public-inbox-init -V2 ffmpegdev ffmpegdev/ https://master.gitmailbox.com/ffmpegdev \ ffmpegdev@gitmailbox.com public-inbox-index ffmpegdev Example config snippet for mirrors. AGPL code for this site: git clone https://public-inbox.org/public-inbox.git