net-im/telegram-desktop: add patch to fix crash on voice recording

- Don't crash on voice recording error.
- Fix voice recoding with FFmpeg 4.4.

See also: https://github.com/telegramdesktop/tdesktop/issues/16217

Package-Manager: Portage-3.0.19, Repoman-3.0.3
Signed-off-by: Stefan Strogin <steils@gentoo.org>

2 files changed, 287 insertions, 0 deletions

diff --git a/net-im/telegram-desktop/files/tdesktop-2.7.4-voice-crash.patch b/net-im/telegram-desktop/files/tdesktop-2.7.4-voice-crash.patch
new file mode 100644
index 000000000000..41e447989cb2
--- /dev/null
+++ b/net-im/telegram-desktop/files/tdesktop-2.7.4-voice-crash.patch
@@ -0,0 +1,262 @@
+From 71deaa48afab2bcf9df67b9b347b1f44aad3a9ce Mon Sep 17 00:00:00 2001
+From: John Preston <johnprestonmail@gmail.com>
+Date: Thu, 13 May 2021 15:17:54 +0400
+Subject: [PATCH] Don't crash on voice recording error.
+
+Fixes #16217.
+---
+ .../media/audio/media_audio_capture.cpp       | 89 +++++++++++--------
+ 1 file changed, 54 insertions(+), 35 deletions(-)
+
+diff --git a/Telegram/SourceFiles/media/audio/media_audio_capture.cpp b/Telegram/SourceFiles/media/audio/media_audio_capture.cpp
+index a5965e0d1988..deb2474d7891 100644
+--- a/Telegram/SourceFiles/media/audio/media_audio_capture.cpp
++++ b/Telegram/SourceFiles/media/audio/media_audio_capture.cpp
+@@ -47,17 +47,17 @@ class Instance::Inner final : public QObject {
+ 	void start(Fn<void(Update)> updated, Fn<void()> error);
+ 	void stop(Fn<void(Result&&)> callback = nullptr);
+ 
+-	void timeout();
+-
+ private:
+-	void processFrame(int32 offset, int32 framesize);
++	void process();
++
++	[[nodiscard]] bool processFrame(int32 offset, int32 framesize);
+ 	void fail();
+ 
+-	void writeFrame(AVFrame *frame);
++	[[nodiscard]] bool writeFrame(AVFrame *frame);
+ 
+ 	// Writes the packets till EAGAIN is got from av_receive_packet()
+ 	// Returns number of packets written or -1 on error
+-	int writePackets();
++	[[nodiscard]] int writePackets();
+ 
+ 	Fn<void(Update)> _updated;
+ 	Fn<void()> _error;
+@@ -150,6 +150,7 @@ struct Instance::Inner::Private {
+ 	AVCodec *codec = nullptr;
+ 	AVCodecContext *codecContext = nullptr;
+ 	bool opened = false;
++	bool processing = false;
+ 
+ 	int srcSamples = 0;
+ 	int dstSamples = 0;
+@@ -217,7 +218,7 @@ struct Instance::Inner::Private {
+ 
+ Instance::Inner::Inner(QThread *thread)
+ : d(std::make_unique<Private>())
+-, _timer(thread, [=] { timeout(); }) {
++, _timer(thread, [=] { process(); }) {
+ 	moveToThread(thread);
+ }
+ 
+@@ -226,10 +227,10 @@ Instance::Inner::~Inner() {
+ }
+ 
+ void Instance::Inner::fail() {
+-	Expects(_error != nullptr);
+-
+ 	stop();
+-	_error();
++	if (const auto error = base::take(_error)) {
++		InvokeQueued(this, error);
++	}
+ }
+ 
+ void Instance::Inner::start(Fn<void(Update)> updated, Fn<void()> error) {
+@@ -384,13 +385,21 @@ void Instance::Inner::stop(Fn<void(Result&&)> callback) {
+ 	}
+ 	_timer.cancel();
+ 
+-	if (d->device) {
++	const auto needResult = (callback != nullptr);
++	const auto hadDevice = (d->device != nullptr);
++	if (hadDevice) {
+ 		alcCaptureStop(d->device);
+-		timeout(); // get last data
++		if (d->processing) {
++			Assert(!needResult); // stop in the middle of processing - error.
++		} else {
++			process(); // get last data
++		}
++		alcCaptureCloseDevice(d->device);
++		d->device = nullptr;
+ 	}
+ 
+ 	// Write what is left
+-	if (!_captured.isEmpty()) {
++	if (needResult && !_captured.isEmpty()) {
+ 		auto fadeSamples = kCaptureFadeInDuration * kCaptureFrequency / 1000;
+ 		auto capturedSamples = static_cast<int>(_captured.size() / sizeof(short));
+ 		if ((_captured.size() % sizeof(short)) || (d->fullSamples + capturedSamples < kCaptureFrequency) || (capturedSamples < fadeSamples)) {
+@@ -414,11 +423,13 @@ void Instance::Inner::stop(Fn<void(Result&&)> callback) {
+ 
+ 			int32 framesize = d->srcSamples * d->codecContext->channels * sizeof(short), encoded = 0;
+ 			while (_captured.size() >= encoded + framesize) {
+-				processFrame(encoded, framesize);
++				if (!processFrame(encoded, framesize)) {
++					break;
++				}
+ 				encoded += framesize;
+ 			}
+-			writeFrame(nullptr); // drain the codec
+-			if (encoded != _captured.size()) {
++			// Drain the codec.
++			if (!writeFrame(nullptr) || encoded != _captured.size()) {
+ 				d->fullSamples = 0;
+ 				d->dataPos = 0;
+ 				d->data.clear();
+@@ -436,14 +447,14 @@ void Instance::Inner::stop(Fn<void(Result&&)> callback) {
+ 	_captured = QByteArray();
+ 
+ 	// Finish stream
+-	if (d->device) {
++	if (needResult && hadDevice) {
+ 		av_write_trailer(d->fmtContext);
+ 	}
+ 
+ 	QByteArray result = d->fullSamples ? d->data : QByteArray();
+ 	VoiceWaveform waveform;
+ 	qint32 samples = d->fullSamples;
+-	if (samples && !d->waveform.isEmpty()) {
++	if (needResult && samples && !d->waveform.isEmpty()) {
+ 		int64 count = d->waveform.size(), sum = 0;
+ 		if (count >= Player::kWaveformSamplesCount) {
+ 			QVector<uint16> peaks;
+@@ -472,11 +483,7 @@ void Instance::Inner::stop(Fn<void(Result&&)> callback) {
+ 			}
+ 		}
+ 	}
+-	if (d->device) {
+-		alcCaptureStop(d->device);
+-		alcCaptureCloseDevice(d->device);
+-		d->device = nullptr;
+-
++	if (hadDevice) {
+ 		if (d->codecContext) {
+ 			avcodec_free_context(&d->codecContext);
+ 			d->codecContext = nullptr;
+@@ -528,12 +535,17 @@ void Instance::Inner::stop(Fn<void(Result&&)> callback) {
+ 		d->waveform.clear();
+ 	}
+ 
+-	if (callback) {
++	if (needResult) {
+ 		callback({ result, waveform, samples });
+ 	}
+ }
+ 
+-void Instance::Inner::timeout() {
++void Instance::Inner::process() {
++	Expects(!d->processing);
++
++	d->processing = true;
++	const auto guard = gsl::finally([&] { d->processing = false; });
++
+ 	if (!d->device) {
+ 		_timer.cancel();
+ 		return;
+@@ -582,7 +594,9 @@ void Instance::Inner::timeout() {
+ 		// Write frames
+ 		int32 framesize = d->srcSamples * d->codecContext->channels * sizeof(short), encoded = 0;
+ 		while (uint32(_captured.size()) >= encoded + framesize + fadeSamples * sizeof(short)) {
+-			processFrame(encoded, framesize);
++			if (!processFrame(encoded, framesize)) {
++				return;
++			}
+ 			encoded += framesize;
+ 		}
+ 
+@@ -597,13 +611,13 @@ void Instance::Inner::timeout() {
+ 	}
+ }
+ 
+-void Instance::Inner::processFrame(int32 offset, int32 framesize) {
++bool Instance::Inner::processFrame(int32 offset, int32 framesize) {
+ 	// Prepare audio frame
+ 
+ 	if (framesize % sizeof(short)) { // in the middle of a sample
+ 		LOG(("Audio Error: Bad framesize in writeFrame() for capture, framesize %1, %2").arg(framesize));
+ 		fail();
+-		return;
++		return false;
+ 	}
+ 	auto samplesCnt = static_cast<int>(framesize / sizeof(short));
+ 
+@@ -650,7 +664,7 @@ void Instance::Inner::processFrame(int32 offset, int32 framesize) {
+ 		if ((res = av_samples_alloc(d->dstSamplesData, 0, d->codecContext->channels, d->dstSamples, d->codecContext->sample_fmt, 1)) < 0) {
+ 			LOG(("Audio Error: Unable to av_samples_alloc for capture, error %1, %2").arg(res).arg(av_make_error_string(err, sizeof(err), res)));
+ 			fail();
+-			return;
++			return false;
+ 		}
+ 		d->dstSamplesSize = av_samples_get_buffer_size(0, d->codecContext->channels, d->maxDstSamples, d->codecContext->sample_fmt, 0);
+ 	}
+@@ -658,7 +672,7 @@ void Instance::Inner::processFrame(int32 offset, int32 framesize) {
+ 	if ((res = swr_convert(d->swrContext, d->dstSamplesData, d->dstSamples, (const uint8_t **)srcSamplesData, d->srcSamples)) < 0) {
+ 		LOG(("Audio Error: Unable to swr_convert for capture, error %1, %2").arg(res).arg(av_make_error_string(err, sizeof(err), res)));
+ 		fail();
+-		return;
++		return false;
+ 	}
+ 
+ 	// Write audio frame
+@@ -670,45 +684,50 @@ void Instance::Inner::processFrame(int32 offset, int32 framesize) {
+ 
+ 	avcodec_fill_audio_frame(frame, d->codecContext->channels, d->codecContext->sample_fmt, d->dstSamplesData[0], d->dstSamplesSize, 0);
+ 
+-	writeFrame(frame);
++	if (!writeFrame(frame)) {
++		return false;
++	}
+ 
+ 	d->fullSamples += samplesCnt;
+ 
+ 	av_frame_free(&frame);
++	return true;
+ }
+ 
+-void Instance::Inner::writeFrame(AVFrame *frame) {
++bool Instance::Inner::writeFrame(AVFrame *frame) {
+ 	int res = 0;
+ 	char err[AV_ERROR_MAX_STRING_SIZE] = { 0 };
+ 
+ 	res = avcodec_send_frame(d->codecContext, frame);
+ 	if (res == AVERROR(EAGAIN)) {
+-		int packetsWritten = writePackets();
++		const auto packetsWritten = writePackets();
+ 		if (packetsWritten < 0) {
+ 			if (frame && packetsWritten == AVERROR_EOF) {
+ 				LOG(("Audio Error: EOF in packets received when EAGAIN was got in avcodec_send_frame()"));
+ 				fail();
+ 			}
+-			return;
++			return false;
+ 		} else if (!packetsWritten) {
+ 			LOG(("Audio Error: No packets received when EAGAIN was got in avcodec_send_frame()"));
+ 			fail();
+-			return;
++			return false;
+ 		}
+ 		res = avcodec_send_frame(d->codecContext, frame);
+ 	}
+ 	if (res < 0) {
+ 		LOG(("Audio Error: Unable to avcodec_send_frame for capture, error %1, %2").arg(res).arg(av_make_error_string(err, sizeof(err), res)));
+ 		fail();
+-		return;
++		return false;
+ 	}
+ 
+ 	if (!frame) { // drain
+ 		if ((res = writePackets()) != AVERROR_EOF) {
+ 			LOG(("Audio Error: not EOF in packets received when draining the codec, result %1").arg(res));
+ 			fail();
++			return false;
+ 		}
+ 	}
++	return true;
+ }
+ 
+ int Instance::Inner::writePackets() {
diff --git a/net-im/telegram-desktop/files/tdesktop-2.7.4-voice-ffmpeg44.patch b/net-im/telegram-desktop/files/tdesktop-2.7.4-voice-ffmpeg44.patch
new file mode 100644
index 000000000000..4156956032d4
--- /dev/null
+++ b/net-im/telegram-desktop/files/tdesktop-2.7.4-voice-ffmpeg44.patch
@@ -0,0 +1,25 @@
+From a8807bc915f2439acc7c84f06d931d96d6ca602a Mon Sep 17 00:00:00 2001
+From: John Preston <johnprestonmail@gmail.com>
+Date: Thu, 13 May 2021 15:33:42 +0400
+Subject: [PATCH] Fix voice recoding with FFmpeg 4.4.
+
+Fixes #16217.
+---
+ Telegram/SourceFiles/media/audio/media_audio_capture.cpp | 4 ++++
+ 1 file changed, 4 insertions(+)
+
+diff --git a/Telegram/SourceFiles/media/audio/media_audio_capture.cpp b/Telegram/SourceFiles/media/audio/media_audio_capture.cpp
+index deb2474d789..d129168783d 100644
+--- a/Telegram/SourceFiles/media/audio/media_audio_capture.cpp
++++ b/Telegram/SourceFiles/media/audio/media_audio_capture.cpp
+@@ -679,6 +679,10 @@ bool Instance::Inner::processFrame(int32 offset, int32 framesize) {
+ 
+ 	AVFrame *frame = av_frame_alloc();
+ 
++	frame->format = d->codecContext->sample_fmt;
++	frame->channels = d->codecContext->channels;
++	frame->channel_layout = d->codecContext->channel_layout;
++	frame->sample_rate = d->codecContext->sample_rate;
+ 	frame->nb_samples = d->dstSamples;
+ 	frame->pts = av_rescale_q(d->fullSamples, AVRational { 1, d->codecContext->sample_rate }, d->codecContext->time_base);
+