configure: Separate subsystem for Immersive Audio Model

This change allows users to build libavfomat without support
for Immersive Audio Model by specifying --disable-iamf.
It helps to save on binary size in cases where it's important.

Co-authored-by: James Almer <jamrial@gmail.com>
Signed-off-by: Eugene Zemtsov <eugene@chromium.org>
Signed-off-by: James Almer <jamrial@gmail.com>

configure

@@ -140,6 +140,7 @@ Component options:
   --disable-error-resilience disable error resilience code
   --disable-lsp            disable LSP code
   --disable-faan           disable floating point AAN (I)DCT code
+  --disable-iamf           disable support for Immersive Audio Model
   --disable-pixelutils     disable pixel utils in libavutil
 
 Individual component options:
@@ -2070,6 +2071,7 @@ SUBSYSTEM_LIST="
     error_resilience
     faan
     fast_unaligned
+    iamf
     lsp
     pixelutils
     network
@@ -2851,6 +2853,8 @@ h264_sei_select="atsc_a53 golomb"
 hevcparse_select="golomb"
 hevc_sei_select="atsc_a53 golomb"
 frame_thread_encoder_deps="encoders threads"
+iamfdec_deps="iamf"
+iamfenc_deps="iamf"
 inflate_wrapper_deps="zlib"
 intrax8_select="blockdsp wmv2dsp"
 iso_media_select="mpeg4audio"
@@ -3590,8 +3594,8 @@ gxf_muxer_select="pcm_rechunk_bsf"
 hds_muxer_select="flv_muxer"
 hls_demuxer_select="aac_demuxer ac3_demuxer adts_header ac3_parser eac3_demuxer mov_demuxer mpegts_demuxer"
 hls_muxer_select="mov_muxer mpegts_muxer"
-iamf_demuxer_select="iamfdec"
+iamf_demuxer_deps="iamfdec"
-iamf_muxer_select="iamfenc"
+iamf_muxer_deps="iamfenc"
 image2_alias_pix_demuxer_select="image2_demuxer"
 image2_brender_pix_demuxer_select="image2_demuxer"
 imf_demuxer_deps="libxml2"
@@ -3606,9 +3610,9 @@ matroska_demuxer_suggest="bzlib zlib"
 matroska_muxer_select="mpeg4audio riffenc aac_adtstoasc_bsf pgs_frame_merge_bsf vp9_superframe_bsf"
 mlp_demuxer_select="mlp_parser"
 mmf_muxer_select="riffenc"
-mov_demuxer_select="iso_media riffdec iamfdec"
+mov_demuxer_select="iso_media riffdec"
 mov_demuxer_suggest="zlib"
-mov_muxer_select="iso_media riffenc rtpenc_chain vp9_superframe_bsf aac_adtstoasc_bsf ac3_parser iamfenc"
+mov_muxer_select="iso_media riffenc rtpenc_chain vp9_superframe_bsf aac_adtstoasc_bsf ac3_parser"
 mp3_demuxer_select="mpegaudio_parser"
 mp3_muxer_select="mpegaudioheader"
 mp4_muxer_select="mov_muxer"
@@ -4079,6 +4083,7 @@ enable asm
 enable debug
 enable doc
 enable faan faandct faanidct
+enable iamf iamfdec iamfenc
 enable large_tests
 enable optimizations
 enable ptx_compression

libavformat/mov.c

@@ -863,6 +863,7 @@ static int mov_read_dac3(MOVContext *c, AVIOContext *pb, MOVAtom atom)
     return 0;
 }
 
+#if CONFIG_IAMFDEC
 static int mov_read_iacb(MOVContext *c, AVIOContext *pb, MOVAtom atom)
 {
     AVStream *st;
@@ -1042,6 +1043,7 @@ fail:
 
     return ret;
 }
+#endif
 
 static int mov_read_dec3(MOVContext *c, AVIOContext *pb, MOVAtom atom)
 {
@@ -4810,6 +4812,7 @@ static void fix_timescale(MOVContext *c, MOVStreamContext *sc)
     }
 }
 
+#if CONFIG_IAMFDEC
 static int mov_update_iamf_streams(MOVContext *c, const AVStream *st)
 {
     const MOVStreamContext *sc = st->priv_data;
@@ -4853,6 +4856,7 @@ static int mov_update_iamf_streams(MOVContext *c, const AVStream *st)
 
     return 0;
 }
+#endif
 
 static int mov_read_trak(MOVContext *c, AVIOContext *pb, MOVAtom atom)
 {
@@ -4917,11 +4921,13 @@ static int mov_read_trak(MOVContext *c, AVIOContext *pb, MOVAtom atom)
 
     mov_build_index(c, st);
 
+#if CONFIG_IAMFDEC
     if (sc->iamf) {
         ret = mov_update_iamf_streams(c, st);
         if (ret < 0)
             return ret;
     }
+#endif
 
     if (sc->dref_id-1 < sc->drefs_count && sc->drefs[sc->dref_id-1].path) {
         MOVDref *dref = &sc->drefs[sc->dref_id - 1];
@@ -8573,7 +8579,9 @@ static const MOVParseTableEntry mov_default_parse_table[] = {
 { MKTAG('i','p','r','p'), mov_read_iprp },
 { MKTAG('i','i','n','f'), mov_read_iinf },
 { MKTAG('a','m','v','e'), mov_read_amve }, /* ambient viewing environment box */
+#if CONFIG_IAMFDEC
 { MKTAG('i','a','c','b'), mov_read_iacb },
+#endif
 { 0, NULL }
 };
 
@@ -9045,8 +9053,10 @@ static void mov_free_stream_context(AVFormatContext *s, AVStream *st)
     av_freep(&sc->coll);
     av_freep(&sc->ambient);
 
+#if CONFIG_IAMFDEC
     if (sc->iamf)
         ff_iamf_read_deinit(sc->iamf);
+#endif
     av_freep(&sc->iamf);
 }
 
@@ -9951,6 +9961,7 @@ static int mov_read_packet(AVFormatContext *s, AVPacket *pkt)
 
         if (st->codecpar->codec_id == AV_CODEC_ID_EIA_608 && sample->size > 8)
             ret = get_eia608_packet(sc->pb, pkt, sample->size);
+#if CONFIG_IAMFDEC
         else if (sc->iamf) {
             int64_t pts, dts, pos, duration;
             int flags, size = sample->size;
@@ -9973,7 +9984,9 @@ static int mov_read_packet(AVFormatContext *s, AVPacket *pkt)
             }
             if (!ret)
                 return FFERROR_REDO;
-        } else
+        }
+#endif
+        else
             ret = av_get_packet(sc->pb, pkt, sample->size);
         if (ret < 0) {
             if (should_retry(sc->pb, ret)) {

libavformat/movenc.c

@@ -318,6 +318,7 @@ static int mov_write_sdtp_tag(AVIOContext *pb, MOVTrack *track)
     return update_size(pb, pos);
 }
 
+#if CONFIG_IAMFENC
 static int mov_write_iacb_tag(AVFormatContext *s, AVIOContext *pb, MOVTrack *track)
 {
     AVIOContext *dyn_bc;
@@ -343,6 +344,7 @@ static int mov_write_iacb_tag(AVFormatContext *s, AVIOContext *pb, MOVTrack *tra
 
     return update_size(pb, pos);
 }
+#endif
 
 static int mov_write_amr_tag(AVIOContext *pb, MOVTrack *track)
 {
@@ -1386,8 +1388,10 @@ static int mov_write_audio_tag(AVFormatContext *s, AVIOContext *pb, MOVMuxContex
         ret = mov_write_wave_tag(s, pb, track);
     else if (track->tag == MKTAG('m','p','4','a'))
         ret = mov_write_esds_tag(pb, track);
+#if CONFIG_IAMFENC
     else if (track->tag == MKTAG('i','a','m','f'))
         ret = mov_write_iacb_tag(mov->fc, pb, track);
+#endif
     else if (track->par->codec_id == AV_CODEC_ID_AMR_NB)
         ret = mov_write_amr_tag(pb, track);
     else if (track->par->codec_id == AV_CODEC_ID_AC3)
@@ -5520,6 +5524,7 @@ static int mov_write_ftyp_tag(AVIOContext *pb, AVFormatContext *s)
     int has_h264 = 0, has_av1 = 0, has_video = 0, has_dolby = 0;
     int has_iamf = 0;
 
+#if CONFIG_IAMFENC
     for (int i = 0; i < s->nb_stream_groups; i++) {
         const AVStreamGroup *stg = s->stream_groups[i];
 
@@ -5529,6 +5534,7 @@ static int mov_write_ftyp_tag(AVIOContext *pb, AVFormatContext *s)
             break;
         }
     }
+#endif
     for (int i = 0; i < mov->nb_streams; i++) {
         AVStream *st = mov->tracks[i].st;
         if (is_cover_image(st))
@@ -6655,6 +6661,7 @@ static int mov_write_subtitle_end_packet(AVFormatContext *s,
     return ret;
 }
 
+#if CONFIG_IAMFENC
 static int mov_build_iamf_packet(AVFormatContext *s, MOVTrack *trk, AVPacket *pkt)
 {
     int ret;
@@ -6705,6 +6712,7 @@ static int mov_build_iamf_packet(AVFormatContext *s, MOVTrack *trk, AVPacket *pk
 
     return ret;
 }
+#endif
 
 static int mov_write_packet(AVFormatContext *s, AVPacket *pkt)
 {
@@ -6718,6 +6726,7 @@ static int mov_write_packet(AVFormatContext *s, AVPacket *pkt)
 
     trk = s->streams[pkt->stream_index]->priv_data;
 
+#if CONFIG_IAMFENC
     if (trk->iamf) {
         int ret = mov_build_iamf_packet(s, trk, pkt);
         if (ret < 0) {
@@ -6728,6 +6737,7 @@ static int mov_write_packet(AVFormatContext *s, AVPacket *pkt)
             return ret;
         }
     }
+#endif
 
     if (is_cover_image(trk->st)) {
         int ret;
@@ -7067,10 +7077,12 @@ static void mov_free(AVFormatContext *s)
         ff_mov_cenc_free(&track->cenc);
         ffio_free_dyn_buf(&track->mdat_buf);
 
+#if CONFIG_IAMFENC
         ffio_free_dyn_buf(&track->iamf_buf);
         if (track->iamf)
             ff_iamf_uninit_context(track->iamf);
         av_freep(&track->iamf);
+#endif
 
         avpriv_packet_list_free(&track->squashed_packet_queue);
     }
@@ -7145,6 +7157,7 @@ static int mov_create_dvd_sub_decoder_specific_info(MOVTrack *track,
     return 0;
 }
 
+#if CONFIG_IAMFENC
 static int mov_init_iamf_track(AVFormatContext *s)
 {
     MOVMuxContext *mov = s->priv_data;
@@ -7204,6 +7217,7 @@ static int mov_init_iamf_track(AVFormatContext *s)
 
     return 0;
 }
+#endif
 
 static int mov_init(AVFormatContext *s)
 {
@@ -7342,6 +7356,7 @@ static int mov_init(AVFormatContext *s)
         s->streams[0]->disposition |= AV_DISPOSITION_DEFAULT;
     }
 
+#if CONFIG_IAMFENC
     for (i = 0; i < s->nb_stream_groups; i++) {
         AVStreamGroup *stg = s->stream_groups[i];
 
@@ -7362,6 +7377,7 @@ static int mov_init(AVFormatContext *s)
         if (!mov->nb_tracks) // We support one track for the entire IAMF structure
             mov->nb_tracks++;
     }
+#endif
 
     for (i = 0; i < s->nb_streams; i++) {
         AVStream *st = s->streams[i];
@@ -7447,9 +7463,11 @@ static int mov_init(AVFormatContext *s)
         }
     }
 
+#if CONFIG_IAMFENC
     ret = mov_init_iamf_track(s);
     if (ret < 0)
         return ret;
+#endif
 
     for (int j = 0, i = 0; j < s->nb_streams; j++) {
         AVStream *st = s->streams[j];