X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=ffmpeg_capture.cpp;h=1ffa9d5418dc8d16cf215fb84892346c89084cb8;hb=9e47a2f661b9d292598ef0277e507458e3dad62f;hp=5e6104f82c628be57cf543ee06599ac639ff3f6a;hpb=d205e9c826b4a4e1290cc4160067aad818e70081;p=nageru

diff --git a/ffmpeg_capture.cpp b/ffmpeg_capture.cpp
index 5e6104f..1ffa9d5 100644
--- a/ffmpeg_capture.cpp
+++ b/ffmpeg_capture.cpp
@@ -32,6 +32,7 @@ extern "C" {
 #include "ffmpeg_util.h"
 #include "flags.h"
 #include "image_input.h"
+#include "ref_counted_frame.h"
 #include "timebase.h"
 
 #define FRAME_SIZE (8 << 20)  // 8 MB.
@@ -228,7 +229,8 @@ void FFmpegCapture::configure_card()
 		set_video_frame_allocator(owned_video_frame_allocator.get());
 	}
 	if (audio_frame_allocator == nullptr) {
-		owned_audio_frame_allocator.reset(new MallocFrameAllocator(65536, NUM_QUEUED_AUDIO_FRAMES));
+		// Audio can come out in pretty large chunks, so increase from the default 1 MB.
+		owned_audio_frame_allocator.reset(new MallocFrameAllocator(1 << 20, NUM_QUEUED_AUDIO_FRAMES));
 		set_audio_frame_allocator(owned_audio_frame_allocator.get());
 	}
 }
@@ -319,7 +321,7 @@ void FFmpegCapture::send_disconnected_frame()
 		video_frame.len = width * height * 4;
 		memset(video_frame.data, 0, video_frame.len);
 
-		frame_callback(-1, AVRational{1, TIMEBASE}, timecode++,
+		frame_callback(-1, AVRational{1, TIMEBASE}, -1, AVRational{1, TIMEBASE}, timecode++,
 			video_frame, /*video_offset=*/0, video_format,
 			FrameAllocator::Frame(), /*audio_offset=*/0, AudioFormat());
 	}
@@ -407,12 +409,13 @@ bool FFmpegCapture::play_video(const string &pathname)
 		if (process_queued_commands(format_ctx.get(), pathname, last_modified, /*rewound=*/nullptr)) {
 			return true;
 		}
-		FrameAllocator::Frame audio_frame = audio_frame_allocator->alloc_frame();
+		UniqueFrame audio_frame = audio_frame_allocator->alloc_frame();
 		AudioFormat audio_format;
 
+		int64_t audio_pts;
 		bool error;
 		AVFrameWithDeleter frame = decode_frame(format_ctx.get(), video_codec_ctx.get(), audio_codec_ctx.get(),
-			pathname, video_stream_index, audio_stream_index, &audio_frame, &audio_format, &error);
+			pathname, video_stream_index, audio_stream_index, audio_frame.get(), &audio_format, &audio_pts, &error);
 		if (error) {
 			return false;
 		}
@@ -422,6 +425,12 @@ bool FFmpegCapture::play_video(const string &pathname)
 				fprintf(stderr, "%s: Rewind failed, not looping.\n", pathname.c_str());
 				return true;
 			}
+			if (video_codec_ctx != nullptr) {
+				avcodec_flush_buffers(video_codec_ctx.get());
+			}
+			if (audio_codec_ctx != nullptr) {
+				avcodec_flush_buffers(audio_codec_ctx.get());
+			}
 			// If the file has changed since last time, return to get it reloaded.
 			// Note that depending on how you move the file into place, you might
 			// end up corrupting the one you're already playing, so this path
@@ -434,7 +443,7 @@ bool FFmpegCapture::play_video(const string &pathname)
 		}
 
 		VideoFormat video_format = construct_video_format(frame.get(), video_timebase);
-		FrameAllocator::Frame video_frame = make_video_frame(frame.get(), pathname, &error);
+		UniqueFrame video_frame = make_video_frame(frame.get(), pathname, &error);
 		if (error) {
 			return false;
 		}
@@ -444,12 +453,15 @@ bool FFmpegCapture::play_video(const string &pathname)
 				pts_origin = frame->pts;	
 			}
 			next_frame_start = compute_frame_start(frame->pts, pts_origin, video_timebase, start, rate);
-			video_frame.received_timestamp = next_frame_start;
+			video_frame->received_timestamp = next_frame_start;
 			bool finished_wakeup = producer_thread_should_quit.sleep_until(next_frame_start);
 			if (finished_wakeup) {
-				frame_callback(frame->pts, video_timebase, timecode++,
-					video_frame, 0, video_format,
-					audio_frame, 0, audio_format);
+				if (audio_frame->len > 0) {
+					assert(audio_pts != -1);
+				}
+				frame_callback(frame->pts, video_timebase, audio_pts, audio_timebase, timecode++,
+					video_frame.get_and_release(), 0, video_format,
+					audio_frame.get_and_release(), 0, audio_format);
 				break;
 			} else {
 				if (producer_thread_should_quit.should_quit()) break;
@@ -460,7 +472,6 @@ bool FFmpegCapture::play_video(const string &pathname)
 				}
 				// If we just rewound, drop this frame on the floor and be done.
 				if (rewound) {
-					video_frame_allocator->release_frame(video_frame);
 					break;
 				}
 				// OK, we didn't, so probably a rate change. Recalculate next_frame_start,
@@ -527,7 +538,9 @@ namespace {
 
 }  // namespace
 
-AVFrameWithDeleter FFmpegCapture::decode_frame(AVFormatContext *format_ctx, AVCodecContext *video_codec_ctx, AVCodecContext *audio_codec_ctx, const std::string &pathname, int video_stream_index, int audio_stream_index, FrameAllocator::Frame *audio_frame, AudioFormat *audio_format, bool *error)
+AVFrameWithDeleter FFmpegCapture::decode_frame(AVFormatContext *format_ctx, AVCodecContext *video_codec_ctx, AVCodecContext *audio_codec_ctx,
+	const std::string &pathname, int video_stream_index, int audio_stream_index,
+	FrameAllocator::Frame *audio_frame, AudioFormat *audio_format, int64_t *audio_pts, bool *error)
 {
 	*error = false;
 
@@ -536,6 +549,7 @@ AVFrameWithDeleter FFmpegCapture::decode_frame(AVFormatContext *format_ctx, AVCo
 	AVFrameWithDeleter audio_avframe = av_frame_alloc_unique();
 	AVFrameWithDeleter video_avframe = av_frame_alloc_unique();
 	bool eof = false;
+	*audio_pts = -1;
 	do {
 		AVPacket pkt;
 		unique_ptr<AVPacket, decltype(av_packet_unref)*> pkt_cleanup(
@@ -554,6 +568,9 @@ AVFrameWithDeleter FFmpegCapture::decode_frame(AVFormatContext *format_ctx, AVCo
 					return AVFrameWithDeleter(nullptr);
 				}
 			} else if (pkt.stream_index == audio_stream_index) {
+				if (*audio_pts == -1) {
+					*audio_pts = pkt.pts;
+				}
 				if (avcodec_send_packet(audio_codec_ctx, &pkt) < 0) {
 					fprintf(stderr, "%s: Cannot send packet to audio codec.\n", pathname.c_str());
 					*error = true;
@@ -565,17 +582,23 @@ AVFrameWithDeleter FFmpegCapture::decode_frame(AVFormatContext *format_ctx, AVCo
 		}
 
 		// Decode audio, if any.
-		int err = avcodec_receive_frame(audio_codec_ctx, audio_avframe.get());
-		if (err == 0) {
-			convert_audio(audio_avframe.get(), audio_frame, audio_format);
-		} else if (err != AVERROR(EAGAIN)) {
-			fprintf(stderr, "%s: Cannot receive frame from audio codec.\n", pathname.c_str());
-			*error = true;
-			return AVFrameWithDeleter(nullptr);
+		if (*audio_pts != -1) {
+			for ( ;; ) {
+				int err = avcodec_receive_frame(audio_codec_ctx, audio_avframe.get());
+				if (err == 0) {
+					convert_audio(audio_avframe.get(), audio_frame, audio_format);
+				} else if (err == AVERROR(EAGAIN)) {
+					break;
+				} else {
+					fprintf(stderr, "%s: Cannot receive frame from audio codec.\n", pathname.c_str());
+					*error = true;
+					return AVFrameWithDeleter(nullptr);
+				}
+			}
 		}
 
 		// Decode video, if we have a frame.
-		err = avcodec_receive_frame(video_codec_ctx, video_avframe.get());
+		int err = avcodec_receive_frame(video_codec_ctx, video_avframe.get());
 		if (err == 0) {
 			frame_finished = true;
 			break;
@@ -622,10 +645,15 @@ void FFmpegCapture::convert_audio(const AVFrame *audio_avframe, FrameAllocator::
 	}
 	audio_format->num_channels = 2;
 
+	int64_t channel_layout = audio_avframe->channel_layout;
+	if (channel_layout == 0) {
+		channel_layout = av_get_default_channel_layout(audio_avframe->channels);
+	}
+
 	if (resampler == nullptr ||
 	    audio_avframe->format != last_src_format ||
 	    dst_format != last_dst_format ||
-	    av_frame_get_channel_layout(audio_avframe) != last_channel_layout ||
+	    channel_layout != last_channel_layout ||
 	    av_frame_get_sample_rate(audio_avframe) != last_sample_rate) {
 		avresample_free(&resampler);
 		resampler = avresample_alloc_context();
@@ -634,7 +662,7 @@ void FFmpegCapture::convert_audio(const AVFrame *audio_avframe, FrameAllocator::
 			exit(1);
 		}
 
-		av_opt_set_int(resampler, "in_channel_layout",  av_frame_get_channel_layout(audio_avframe), 0);
+		av_opt_set_int(resampler, "in_channel_layout",  channel_layout,                             0);
 		av_opt_set_int(resampler, "out_channel_layout", AV_CH_LAYOUT_STEREO,                        0);
 		av_opt_set_int(resampler, "in_sample_rate",     av_frame_get_sample_rate(audio_avframe),    0);
 		av_opt_set_int(resampler, "out_sample_rate",    OUTPUT_FREQUENCY,                           0);
@@ -648,7 +676,7 @@ void FFmpegCapture::convert_audio(const AVFrame *audio_avframe, FrameAllocator::
 
 		last_src_format = AVSampleFormat(audio_avframe->format);
 		last_dst_format = dst_format;
-		last_channel_layout = av_frame_get_channel_layout(audio_avframe);
+		last_channel_layout = channel_layout;
 		last_sample_rate = av_frame_get_sample_rate(audio_avframe);
 	}
 
@@ -657,7 +685,7 @@ void FFmpegCapture::convert_audio(const AVFrame *audio_avframe, FrameAllocator::
 
 	uint8_t *data = audio_frame->data + audio_frame->len;
 	int out_samples = avresample_convert(resampler, &data, 0, num_samples_room,
-		audio_avframe->data, audio_avframe->linesize[0], audio_avframe->nb_samples);
+		const_cast<uint8_t **>(audio_avframe->data), audio_avframe->linesize[0], audio_avframe->nb_samples);
 	if (out_samples < 0) {
                 fprintf(stderr, "Audio conversion failed.\n");
                 exit(1);
@@ -691,12 +719,12 @@ VideoFormat FFmpegCapture::construct_video_format(const AVFrame *frame, AVRation
 	return video_format;
 }
 
-FrameAllocator::Frame FFmpegCapture::make_video_frame(const AVFrame *frame, const string &pathname, bool *error)
+UniqueFrame FFmpegCapture::make_video_frame(const AVFrame *frame, const string &pathname, bool *error)
 {
 	*error = false;
 
-	FrameAllocator::Frame video_frame = video_frame_allocator->alloc_frame();
-	if (video_frame.data == nullptr) {
+	UniqueFrame video_frame(video_frame_allocator->alloc_frame());
+	if (video_frame->data == nullptr) {
 		return video_frame;
 	}
 
@@ -722,17 +750,17 @@ FrameAllocator::Frame FFmpegCapture::make_video_frame(const AVFrame *frame, cons
 	uint8_t *pic_data[4] = { nullptr, nullptr, nullptr, nullptr };
 	int linesizes[4] = { 0, 0, 0, 0 };
 	if (pixel_format == bmusb::PixelFormat_8BitBGRA) {
-		pic_data[0] = video_frame.data;
+		pic_data[0] = video_frame->data;
 		linesizes[0] = width * 4;
-		video_frame.len = (width * 4) * height;
+		video_frame->len = (width * 4) * height;
 	} else if (pixel_format == PixelFormat_NV12) {
-		pic_data[0] = video_frame.data;
+		pic_data[0] = video_frame->data;
 		linesizes[0] = width;
 
 		pic_data[1] = pic_data[0] + width * height;
 		linesizes[1] = width;
 
-		video_frame.len = (width * 2) * height;
+		video_frame->len = (width * 2) * height;
 
 		const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(sws_dst_format);
 		current_frame_ycbcr_format = decode_ycbcr_format(desc, frame);
@@ -743,7 +771,7 @@ FrameAllocator::Frame FFmpegCapture::make_video_frame(const AVFrame *frame, cons
 		int chroma_width = AV_CEIL_RSHIFT(int(width), desc->log2_chroma_w);
 		int chroma_height = AV_CEIL_RSHIFT(int(height), desc->log2_chroma_h);
 
-		pic_data[0] = video_frame.data;
+		pic_data[0] = video_frame->data;
 		linesizes[0] = width;
 
 		pic_data[1] = pic_data[0] + width * height;
@@ -752,7 +780,7 @@ FrameAllocator::Frame FFmpegCapture::make_video_frame(const AVFrame *frame, cons
 		pic_data[2] = pic_data[1] + chroma_width * chroma_height;
 		linesizes[2] = chroma_width;
 
-		video_frame.len = width * height + 2 * chroma_width * chroma_height;
+		video_frame->len = width * height + 2 * chroma_width * chroma_height;
 
 		current_frame_ycbcr_format = decode_ycbcr_format(desc, frame);
 	}