Make kaeru insert the aac_adtstoasc filter when needed.

[nageru] / nageru / kaeru.cpp

diff --git a/nageru/kaeru.cpp b/nageru/kaeru.cpp
index 7a058e47e5e0d4c1733953acf077d9d37c485186..3d2db1c3ba810e3ffab95ef4e785b06c53dc4bbc 100644 (file)
--- a/nageru/kaeru.cpp
+++ b/nageru/kaeru.cpp
@@ -47,11 +47,12 @@ int write_packet(void *opaque, uint8_t *buf, int buf_size, AVIODataMarkerType ty
                type = AVIO_DATA_MARKER_SYNC_POINT;
        }
 
+       HTTPD::StreamID stream_id{ HTTPD::MAIN_STREAM, 0 };
        if (type == AVIO_DATA_MARKER_HEADER) {
                stream_mux_header.append((char *)buf, buf_size);
-               httpd->set_header(HTTPD::MAIN_STREAM, stream_mux_header);
+               httpd->set_header(stream_id, stream_mux_header);
        } else {
-               httpd->add_data(HTTPD::MAIN_STREAM, (char *)buf, buf_size, type == AVIO_DATA_MARKER_SYNC_POINT, time, AVRational{ AV_TIME_BASE, 1 });
+               httpd->add_data(stream_id, (char *)buf, buf_size, type == AVIO_DATA_MARKER_SYNC_POINT, time, AVRational{ AV_TIME_BASE, 1 });
        }
        return buf_size;
 }
@@ -71,8 +72,13 @@ unique_ptr<Mux> create_mux(HTTPD *httpd, AVOutputFormat *oformat, X264Encoder *x
 
        string video_extradata = x264_encoder->get_global_headers();
 
+       // If audio is disabled (ie., we won't ever see any audio packets),
+       // set nullptr here to also not include the stream in the mux.
+       AVCodecParameters *audio_codecpar =
+               global_flags.enable_audio ? audio_encoder->get_codec_parameters().release() : nullptr;
+
        unique_ptr<Mux> mux;
-       mux.reset(new Mux(avctx, global_flags.width, global_flags.height, Mux::CODEC_H264, video_extradata, audio_encoder->get_codec_parameters().get(),
+       mux.reset(new Mux(avctx, global_flags.width, global_flags.height, Mux::CODEC_H264, video_extradata, audio_codecpar,
                get_color_space(global_flags.ycbcr_rec709_coefficients), COARSE_TIMEBASE,
                /*write_callback=*/nullptr, Mux::WRITE_FOREGROUND, { &stream_mux_metrics }));
        stream_mux_metrics.init({{ "destination", "http" }});
@@ -132,9 +138,33 @@ void video_frame_callback(FFmpegCapture *video, X264Encoder *x264_encoder, Audio
        }
 }
 
-void audio_frame_callback(Mux *mux, const AVPacket *pkt, AVRational timebase)
+void raw_packet_callback(Mux *mux, int stream_index, const AVPacket *pkt, AVRational timebase)
+{
+       mux->add_packet(*pkt, pkt->pts, pkt->dts == AV_NOPTS_VALUE ? pkt->pts : pkt->dts, timebase, stream_index);
+}
+
+void filter_packet_callback(Mux *mux, int stream_index, AVBSFContext *bsfctx, const AVPacket *pkt, AVRational timebase)
 {
-       mux->add_packet(*pkt, pkt->pts, pkt->dts == AV_NOPTS_VALUE ? pkt->pts : pkt->dts, timebase, /*stream_index=*/1);
+       AVPacket *in_pkt = av_packet_clone(pkt);
+       unique_ptr<AVPacket, decltype(av_packet_unref) *> in_pkt_cleanup(in_pkt, av_packet_unref);
+       int err = av_bsf_send_packet(bsfctx, in_pkt);
+       if (err < 0) {
+               fprintf(stderr, "av_bsf_send_packet() failed with %d, ignoring\n", err);
+       }
+       for ( ;; ) {
+               AVPacket out_pkt;
+               unique_ptr<AVPacket, decltype(av_packet_unref) *> pkt_cleanup(&out_pkt, av_packet_unref);
+               av_init_packet(&out_pkt);
+               err = av_bsf_receive_packet(bsfctx, &out_pkt);
+               if (err == AVERROR(EAGAIN)) {
+                       break;
+               }
+               if (err < 0) {
+                       fprintf(stderr, "av_bsf_receive_packet() failed with %d, ignoring\n", err);
+                       return;
+               }
+               mux->add_packet(out_pkt, out_pkt.pts, out_pkt.dts == AV_NOPTS_VALUE ? out_pkt.pts : out_pkt.dts, timebase, stream_index);
+       }
 }
 
 void adjust_bitrate(int signal)
@@ -173,9 +203,9 @@ int main(int argc, char *argv[])
        parse_flags(PROGRAM_KAERU, argc, argv);
        if (optind + 1 != argc) {
                usage(PROGRAM_KAERU);
-               exit(1);
+               abort();
        }
-       global_flags.num_cards = 1;  // For latency metrics.
+       global_flags.max_num_cards = 1;  // For latency metrics.
 
 #if LIBAVFORMAT_VERSION_INT < AV_VERSION_INT(58, 9, 100)
        av_register_all();
@@ -199,14 +229,34 @@ int main(int argc, char *argv[])
        if (global_flags.transcode_audio) {
                audio_encoder->add_mux(http_mux.get());
        }
-       x264_encoder->add_mux(http_mux.get());
+       if (global_flags.transcode_video) {
+               x264_encoder->add_mux(http_mux.get());
+       }
        global_x264_encoder = x264_encoder.get();
 
        FFmpegCapture video(argv[optind], global_flags.width, global_flags.height);
        video.set_pixel_format(FFmpegCapture::PixelFormat_NV12);
-       video.set_frame_callback(bind(video_frame_callback, &video, x264_encoder.get(), audio_encoder.get(), _1, _2, _3, _4, _5, _6, _7, _8, _9, _10, _11));
-       if (!global_flags.transcode_audio) {
-               video.set_audio_callback(bind(audio_frame_callback, http_mux.get(), _1, _2));
+       if (global_flags.transcode_video) {
+               video.set_frame_callback(bind(video_frame_callback, &video, x264_encoder.get(), audio_encoder.get(), _1, _2, _3, _4, _5, _6, _7, _8, _9, _10, _11));
+       } else {
+               video.set_video_callback(bind(raw_packet_callback, http_mux.get(), /*stream_index=*/0, _1, _2));
+       }
+       if (!global_flags.transcode_audio && global_flags.enable_audio) {
+               AVBSFContext *bsfctx = nullptr;
+               if (strcmp(oformat->name, "mp4") == 0 && strcmp(audio_encoder->get_codec()->name, "aac") == 0) {
+                       // We need to insert the aac_adtstoasc filter, seemingly (or we will get warnings to do so).
+                       const AVBitStreamFilter *filter = av_bsf_get_by_name("aac_adtstoasc");
+                       int err = av_bsf_alloc(filter, &bsfctx);
+                       if (err < 0) {
+                               fprintf(stderr, "av_bsf_alloc() failed with %d\n", err);
+                               exit(1);
+                       }
+               }
+               if (bsfctx == nullptr) {
+                       video.set_audio_callback(bind(raw_packet_callback, http_mux.get(), /*stream_index=*/1, _1, _2));
+               } else {
+                       video.set_audio_callback(bind(filter_packet_callback, http_mux.get(), /*stream_index=*/1, bsfctx, _1, _2));
+               }
        }
        video.configure_card();
        video.start_bm_capture();