sdl video push conversion stream and toxav video sink

src/CMakeLists.txt

@@ -115,6 +115,8 @@ target_sources(tomato PUBLIC
 	./frame_streams/sdl/sdl_audio2_frame_stream2.hpp
 	./frame_streams/sdl/sdl_audio2_frame_stream2.cpp
 	./frame_streams/sdl/video.hpp
+	./frame_streams/sdl/video_push_converter.hpp
+	./frame_streams/sdl/video_push_converter.cpp
 
 	./stream_manager_ui.hpp
 	./stream_manager_ui.cpp

src/frame_streams/sdl/video_push_converter.cpp

@@ -0,0 +1,58 @@
+#include "./video_push_converter.hpp"
+
+SDL_Surface* convertYUY2_IYUV(SDL_Surface* surf) {
+	if (surf->format != SDL_PIXELFORMAT_YUY2) {
+		return nullptr;
+	}
+	if ((surf->w % 2) != 0) {
+		SDL_SetError("YUY2->IYUV does not support odd widths");
+		// hmmm, we dont handle odd widths
+		return nullptr;
+	}
+
+	SDL_LockSurface(surf);
+
+	SDL_Surface* conv_surf = SDL_CreateSurface(surf->w, surf->h, SDL_PIXELFORMAT_IYUV);
+	SDL_LockSurface(conv_surf);
+
+	// YUY2 is 4:2:2 packed
+	// Y is simple, we just copy it over
+	// U V are double the resolution (vertically), so we avg both
+	// Packed mode: Y0+U0+Y1+V0 (1 plane)
+
+	uint8_t* y_plane = static_cast<uint8_t*>(conv_surf->pixels);
+	uint8_t* u_plane = static_cast<uint8_t*>(conv_surf->pixels) + conv_surf->w*conv_surf->h;
+	uint8_t* v_plane = static_cast<uint8_t*>(conv_surf->pixels) + conv_surf->w*conv_surf->h + (conv_surf->w/2)*(conv_surf->h/2);
+
+	const uint8_t* yuy2_data = static_cast<const uint8_t*>(surf->pixels);
+
+	for (int y = 0; y < surf->h; y++) {
+		for (int x = 0; x < surf->w; x += 2) {
+			// every pixel uses 2 bytes
+			const uint8_t* yuy2_curser = yuy2_data + y*surf->w*2 + x*2;
+			uint8_t src_y0 = yuy2_curser[0];
+			uint8_t src_u = yuy2_curser[1];
+			uint8_t src_y1 = yuy2_curser[2];
+			uint8_t src_v = yuy2_curser[3];
+
+			y_plane[y*conv_surf->w + x] = src_y0;
+			y_plane[y*conv_surf->w + x+1] = src_y1;
+
+			size_t uv_index = (y/2) * (conv_surf->w/2) + x/2;
+			if (y % 2 == 0) {
+				// first write
+				u_plane[uv_index] = src_u;
+				v_plane[uv_index] = src_v;
+			} else {
+				// second write, mix with existing value
+				u_plane[uv_index] = (int(u_plane[uv_index]) + int(src_u)) / 2;
+				v_plane[uv_index] = (int(v_plane[uv_index]) + int(src_v)) / 2;
+			}
+		}
+	}
+
+	SDL_UnlockSurface(conv_surf);
+	SDL_UnlockSurface(surf);
+	return conv_surf;
+}
+

src/frame_streams/sdl/video_push_converter.hpp

@@ -0,0 +1,89 @@
+#pragma once
+
+#include "./video.hpp"
+#include "../frame_stream2.hpp"
+
+#include <cassert>
+
+#include <iostream> // meh
+
+static bool isFormatYUV(SDL_PixelFormat f) {
+	return
+		f == SDL_PIXELFORMAT_YV12 ||
+		f == SDL_PIXELFORMAT_IYUV ||
+		f == SDL_PIXELFORMAT_YUY2 ||
+		f == SDL_PIXELFORMAT_UYVY ||
+		f == SDL_PIXELFORMAT_YVYU ||
+		f == SDL_PIXELFORMAT_NV12 ||
+		f == SDL_PIXELFORMAT_NV21 ||
+		f == SDL_PIXELFORMAT_P010
+	;
+}
+
+SDL_Surface* convertYUY2_IYUV(SDL_Surface* surf);
+
+template<typename RealStream>
+struct PushConversionVideoStream : public RealStream {
+	SDL_PixelFormat _forced_format {SDL_PIXELFORMAT_IYUV};
+
+	template<typename... Args>
+	PushConversionVideoStream(SDL_PixelFormat forced_format, Args&&... args) : RealStream(std::forward<Args>(args)...), _forced_format(forced_format) {}
+	~PushConversionVideoStream(void) {}
+
+	bool push(const SDLVideoFrame& value) override {
+		SDL_Surface* surf = value.surface.get();
+		if (surf->format != _forced_format) {
+			//std::cerr << "DTC: need to convert from " << SDL_GetPixelFormatName(converted_surf->format) << " to SDL_PIXELFORMAT_IYUV\n";
+			if (surf->format == SDL_PIXELFORMAT_YUY2 && _forced_format == SDL_PIXELFORMAT_IYUV) {
+				// optimized custom impl
+
+				//auto start = Message::getTimeMS();
+				surf = convertYUY2_IYUV(surf);
+				//auto end = Message::getTimeMS();
+				// 3ms
+				//std::cerr << "DTC: timing " << SDL_GetPixelFormatName(converted_surf->format) << "->SDL_PIXELFORMAT_IYUV: " << end-start << "ms\n";
+			} else if (isFormatYUV(surf->format)) {
+				// TODO: fix sdl rgb->yuv conversion resulting in too dark (colorspace) issues
+				// https://github.com/libsdl-org/SDL/issues/10877
+
+				// meh, need to convert to rgb as a stopgap
+
+				//auto start = Message::getTimeMS();
+				SDL_Surface* tmp_conv_surf = SDL_ConvertSurfaceAndColorspace(surf, SDL_PIXELFORMAT_RGB24, nullptr, SDL_COLORSPACE_RGB_DEFAULT, 0);
+				//auto end = Message::getTimeMS();
+				// 1ms
+				//std::cerr << "DTC: timing " << SDL_GetPixelFormatName(converted_surf->format) << "->SDL_PIXELFORMAT_RGB24: " << end-start << "ms\n";
+
+				//start = Message::getTimeMS();
+				surf = SDL_ConvertSurfaceAndColorspace(tmp_conv_surf, _forced_format, nullptr, SDL_COLORSPACE_YUV_DEFAULT, 0);
+				//end = Message::getTimeMS();
+				// 60ms
+				//std::cerr << "DTC: timing SDL_PIXELFORMAT_RGB24->" << SDL_GetPixelFormatName(_forced_format) << ": " << end-start << "ms\n";
+
+				SDL_DestroySurface(tmp_conv_surf);
+			} else {
+				surf = SDL_ConvertSurface(surf, _forced_format);
+			}
+
+			if (surf == nullptr) {
+				// oh god
+				std::cerr << "DTC error: failed to convert surface to IYUV: " << SDL_GetError() << "\n";
+				return false;
+			}
+		}
+		assert(surf != nullptr);
+		if (surf != value.surface.get()) {
+			// TODO: add ctr with uptr
+			SDLVideoFrame new_value{value.timestampUS, nullptr};
+			new_value.surface = {
+				surf,
+				&SDL_DestroySurface
+			};
+
+			return RealStream::push(std::move(new_value));
+		} else {
+			return RealStream::push(value);
+		}
+	}
+};
+

src/tox_av_voip_model.cpp

@@ -9,8 +9,18 @@
 #include "./frame_streams/multi_source.hpp"
 #include "./frame_streams/audio_stream_pop_reframer.hpp"
 
+#include "./frame_streams/sdl/video.hpp"
+#include "./frame_streams/sdl/video_push_converter.hpp"
+
+#include <cstring>
+
 #include <iostream>
 
+// fwd
+namespace Message {
+uint64_t getTimeMS(void);
+} // Message
+
 namespace Components {
 	struct ToxAVIncomingAV {
 		bool incoming_audio {false};
@@ -21,12 +31,15 @@ namespace Components {
 		ObjectHandle o;
 		// ptr?
 	};
-	// vid
+	struct ToxAVVideoSink {
+		ObjectHandle o;
+	};
 	struct ToxAVAudioSource {
 		ObjectHandle o;
-		// ptr?
 	};
-	// vid
+	struct ToxAVVideoSource {
+		ObjectHandle o;
+	};
 } // Components
 
 struct ToxAVCallAudioSink : public FrameStream2SinkI<AudioFrame2> {
@@ -84,6 +97,63 @@ struct ToxAVCallAudioSink : public FrameStream2SinkI<AudioFrame2> {
 	}
 };
 
+// exlusive
+struct ToxAVCallVideoSink : public FrameStream2SinkI<SDLVideoFrame> {
+	using stream_type = PushConversionVideoStream<LockedFrameStream2<SDLVideoFrame>>;
+	ToxAVI& _toxav;
+
+	// bitrate for enabled state
+	uint32_t _video_bitrate {2}; // HACK: hardcode to 2mbits (toxap wrongly multiplies internally by 1000)
+
+	uint32_t _fid;
+	std::shared_ptr<stream_type> _writer;
+
+	ToxAVCallVideoSink(ToxAVI& toxav, uint32_t fid) : _toxav(toxav), _fid(fid) {}
+	~ToxAVCallVideoSink(void) {
+		if (_writer) {
+			_writer = nullptr;
+			_toxav.toxavVideoSetBitRate(_fid, 0);
+		}
+	}
+
+	// sink
+	std::shared_ptr<FrameStream2I<SDLVideoFrame>> subscribe(void) override {
+		if (_writer) {
+			// max 1 (exclusive, composite video somewhere else)
+			return nullptr;
+		}
+
+		auto err = _toxav.toxavVideoSetBitRate(_fid, _video_bitrate);
+		if (err != TOXAV_ERR_BIT_RATE_SET_OK) {
+			return nullptr;
+		}
+
+		// toxav needs I420
+		_writer = std::make_shared<stream_type>(SDL_PIXELFORMAT_IYUV);
+
+		return _writer;
+	}
+
+	bool unsubscribe(const std::shared_ptr<FrameStream2I<SDLVideoFrame>>& sub) override {
+		if (!sub || !_writer) {
+			// nah
+			return false;
+		}
+
+		if (sub == _writer) {
+			_writer = nullptr;
+
+			/*auto err = */_toxav.toxavVideoSetBitRate(_fid, 0);
+			// print warning? on error?
+
+			return true;
+		}
+
+		// what
+		return false;
+	}
+};
+
 void ToxAVVoIPModel::addAudioSource(ObjectHandle session, uint32_t friend_number) {
 	auto& stream_source = session.get_or_emplace<Components::VoIP::StreamSources>().streams;
 
@@ -94,13 +164,11 @@ void ToxAVVoIPModel::addAudioSource(ObjectHandle session, uint32_t friend_number
 	incoming_audio.emplace<Components::FrameStream2Source<AudioFrame2>>(std::move(new_asrc));
 	incoming_audio.emplace<Components::StreamSource>(Components::StreamSource::create<AudioFrame2>("ToxAV Friend Call Incoming Audio"));
 
-	std::cout << "new incoming audio\n";
 	if (
 		const auto* defaults = session.try_get<Components::VoIP::DefaultConfig>();
 		defaults != nullptr && defaults->incoming_audio
 	) {
 		incoming_audio.emplace<Components::TagConnectToDefault>(); // depends on what was specified in enter()
-		std::cout << "with default\n";
 	}
 
 	stream_source.push_back(incoming_audio);
@@ -135,6 +203,55 @@ void ToxAVVoIPModel::addAudioSink(ObjectHandle session, uint32_t friend_number)
 	_os.throwEventConstruct(outgoing_audio);
 }
 
+void ToxAVVoIPModel::addVideoSource(ObjectHandle session, uint32_t friend_number) {
+	auto& stream_source = session.get_or_emplace<Components::VoIP::StreamSources>().streams;
+
+	ObjectHandle incoming_video {_os.registry(), _os.registry().create()};
+
+	auto new_vsrc = std::make_unique<FrameStream2MultiSource<SDLVideoFrame>>();
+	incoming_video.emplace<FrameStream2MultiSource<SDLVideoFrame>*>(new_vsrc.get());
+	incoming_video.emplace<Components::FrameStream2Source<SDLVideoFrame>>(std::move(new_vsrc));
+	incoming_video.emplace<Components::StreamSource>(Components::StreamSource::create<SDLVideoFrame>("ToxAV Friend Call Incoming Video"));
+
+	if (
+		const auto* defaults = session.try_get<Components::VoIP::DefaultConfig>();
+		defaults != nullptr && defaults->incoming_video
+	) {
+		incoming_video.emplace<Components::TagConnectToDefault>(); // depends on what was specified in enter()
+	}
+
+	stream_source.push_back(incoming_video);
+	session.emplace<Components::ToxAVVideoSource>(incoming_video);
+	// TODO: tie session to stream
+
+	_video_sources[friend_number] = incoming_video;
+
+	_os.throwEventConstruct(incoming_video);
+}
+
+void ToxAVVoIPModel::addVideoSink(ObjectHandle session, uint32_t friend_number) {
+	auto& stream_sinks = session.get_or_emplace<Components::VoIP::StreamSinks>().streams;
+	ObjectHandle outgoing_video {_os.registry(), _os.registry().create()};
+
+	auto new_vsink = std::make_unique<ToxAVCallVideoSink>(_av, friend_number);
+	outgoing_video.emplace<ToxAVCallVideoSink*>(new_vsink.get());
+	outgoing_video.emplace<Components::FrameStream2Sink<SDLVideoFrame>>(std::move(new_vsink));
+	outgoing_video.emplace<Components::StreamSink>(Components::StreamSink::create<SDLVideoFrame>("ToxAV Friend Call Outgoing Video"));
+
+	if (
+		const auto* defaults = session.try_get<Components::VoIP::DefaultConfig>();
+		defaults != nullptr && defaults->outgoing_video
+	) {
+		outgoing_video.emplace<Components::TagConnectToDefault>(); // depends on what was specified in enter()
+	}
+
+	stream_sinks.push_back(outgoing_video);
+	session.emplace<Components::ToxAVVideoSink>(outgoing_video);
+	// TODO: tie session to stream
+
+	_os.throwEventConstruct(outgoing_video);
+}
+
 void ToxAVVoIPModel::destroySession(ObjectHandle session) {
 	if (!static_cast<bool>(session)) {
 		return;
@@ -153,6 +270,18 @@ void ToxAVVoIPModel::destroySession(ObjectHandle session) {
 			_audio_sources.erase(it_asrc);
 		}
 	}
+	if (session.all_of<Components::ToxAVVideoSource>()) {
+		auto it_vsrc = std::find_if(
+			_video_sources.cbegin(), _video_sources.cend(),
+			[o = session.get<Components::ToxAVVideoSource>().o](const auto& it) {
+				return it.second == o;
+			}
+		);
+
+		if (it_vsrc != _video_sources.cend()) {
+			_video_sources.erase(it_vsrc);
+		}
+	}
 
 	// destory sources
 	if (auto* ss = session.try_get<Components::VoIP::StreamSources>(); ss != nullptr) {
@@ -237,6 +366,39 @@ void ToxAVVoIPModel::tick(void) {
 			}
 		}
 	}
+
+	for (const auto& [oc, vsink] : _os.registry().view<ToxAVCallVideoSink*>().each()) {
+		if (!vsink->_writer) {
+			continue;
+		}
+
+		for (size_t i = 0; i < 10; i++) {
+			auto new_frame_opt = vsink->_writer->pop();
+			if (!new_frame_opt.has_value()) {
+				break;
+			}
+			const auto& new_frame = new_frame_opt.value();
+
+			if (!new_frame.surface) {
+				// wtf?
+				continue;
+			}
+
+			// conversion is done in the sink's stream
+			SDL_Surface* surf = new_frame.surface.get();
+			assert(surf != nullptr);
+
+			SDL_LockSurface(surf);
+			_av.toxavVideoSendFrame(
+				vsink->_fid,
+				surf->w, surf->h,
+				static_cast<const uint8_t*>(surf->pixels),
+				static_cast<const uint8_t*>(surf->pixels) + surf->w * surf->h,
+				static_cast<const uint8_t*>(surf->pixels) + surf->w * surf->h + (surf->w/2) * (surf->h/2)
+			);
+			SDL_UnlockSurface(surf);
+		}
+	}
 }
 
 ObjectHandle ToxAVVoIPModel::enter(const Contact3 c, const Components::VoIP::DefaultConfig& defaults) {
@@ -311,6 +473,8 @@ bool ToxAVVoIPModel::accept(ObjectHandle session, const Components::VoIP::Defaul
 	// bitrate cb or what?
 	assert(!session.all_of<Components::ToxAVAudioSink>());
 	addAudioSink(session, friend_number);
+	assert(!session.all_of<Components::ToxAVVideoSink>());
+	addVideoSink(session, friend_number);
 
 	if (const auto* i_av = session.try_get<Components::ToxAVIncomingAV>(); i_av != nullptr) {
 		// create audio src
@@ -321,6 +485,8 @@ bool ToxAVVoIPModel::accept(ObjectHandle session, const Components::VoIP::Defaul
 
 		// create video src
 		if (i_av->incoming_video) {
+			assert(!session.all_of<Components::ToxAVVideoSource>());
+			addVideoSource(session, friend_number);
 		}
 	}
 
@@ -419,6 +585,11 @@ bool ToxAVVoIPModel::onEvent(const Events::FriendCallState& e) {
 				}
 
 				// video
+				if (s.is_accepting_v() && !o.all_of<Components::ToxAVVideoSink>()) {
+					addVideoSink(o, e.friend_number);
+				} else if (!s.is_accepting_v() && o.all_of<Components::ToxAVVideoSink>()) {
+					// remove vsink?
+				}
 
 				// add/update sources
 				// audio
@@ -429,6 +600,11 @@ bool ToxAVVoIPModel::onEvent(const Events::FriendCallState& e) {
 				}
 
 				// video
+				if (s.is_sending_v() && !o.all_of<Components::ToxAVVideoSource>()) {
+					addVideoSource(o, e.friend_number);
+				} else if (!s.is_sending_v() && o.all_of<Components::ToxAVVideoSource>()) {
+					// remove vsrc?
+				}
 			}
 		}
 	}
@@ -470,7 +646,66 @@ bool ToxAVVoIPModel::onEvent(const Events::FriendAudioFrame& e) {
 	return true;
 }
 
-bool ToxAVVoIPModel::onEvent(const Events::FriendVideoFrame&) {
+bool ToxAVVoIPModel::onEvent(const Events::FriendVideoFrame& e) {
+	auto vsrc_it = _video_sources.find(e.friend_number);
+	if (vsrc_it == _video_sources.cend()) {
+		// missing src from lookup table
+		return false;
+	}
+
+	auto vsrc = vsrc_it->second;
+
+	if (!static_cast<bool>(vsrc)) {
+		// missing src to put frame into ??
+		return false;
+	}
+
+	assert(vsrc.all_of<FrameStream2MultiSource<SDLVideoFrame>*>());
+	assert(vsrc.all_of<Components::FrameStream2Source<SDLVideoFrame>>());
+
+	auto* new_surf = SDL_CreateSurface(e.width, e.height, SDL_PIXELFORMAT_IYUV);
+	assert(new_surf);
+	if (SDL_LockSurface(new_surf)) {
+		// copy the data
+		// we know how the implementation works, its y u v consecutivlely
+		// y
+		for (size_t y = 0; y < e.height; y++) {
+			std::memcpy(
+				//static_cast<uint8_t*>(new_surf->pixels) + new_surf->pitch*y,
+				static_cast<uint8_t*>(new_surf->pixels) + e.width*y,
+				e.y.ptr + e.ystride*y,
+				e.width
+			);
+		}
+
+		// u
+		for (size_t y = 0; y < e.height/2; y++) {
+			std::memcpy(
+				static_cast<uint8_t*>(new_surf->pixels) + (e.width*e.height) + (e.width/2)*y,
+				e.u.ptr + e.ustride*y,
+				e.width/2
+			);
+		}
+
+		// v
+		for (size_t y = 0; y < e.height/2; y++) {
+			std::memcpy(
+				static_cast<uint8_t*>(new_surf->pixels) + (e.width*e.height) + ((e.width/2)*(e.height/2)) + (e.width/2)*y,
+				e.v.ptr + e.vstride*y,
+				e.width/2
+			);
+		}
+
+		SDL_UnlockSurface(new_surf);
+	}
+
+	vsrc.get<FrameStream2MultiSource<SDLVideoFrame>*>()->push({
+		// ms -> us
+		Message::getTimeMS() * 1000, // TODO: make more precise
+		new_surf
+	});
+
+	SDL_DestroySurface(new_surf);
 	return false;
 }
 

src/tox_av_voip_model.hpp

@@ -16,11 +16,13 @@ class ToxAVVoIPModel : protected ToxAVEventI, public VoIPModelI {
 
 	// for faster lookup
 	std::unordered_map<uint32_t, ObjectHandle> _audio_sources;
+	std::unordered_map<uint32_t, ObjectHandle> _video_sources;
 
 	// TODO: virtual? strategy? protected?
 	virtual void addAudioSource(ObjectHandle session, uint32_t friend_number);
 	virtual void addAudioSink(ObjectHandle session, uint32_t friend_number);
-	// TODO: video
+	virtual void addVideoSource(ObjectHandle session, uint32_t friend_number);
+	virtual void addVideoSink(ObjectHandle session, uint32_t friend_number);
 
 	void destroySession(ObjectHandle session);