Merge pull request #10004 from nextcloud/fix-trying-to-use-an-audio-s…

…ink-as-source

Fix trying to use an audio sink as source

recording/src/nextcloud/talk/recording/Benchmark.py

@@ -145,8 +145,9 @@ def run(self, args):
             self._display.start()
 
             # Start new audio sink for the audio output of the player.
-            self._audioModuleIndex, audioSinkIndex = newAudioSink("nextcloud-talk-recording-benchmark")
+            self._audioModuleIndex, audioSinkIndex, audioSourceIndex = newAudioSink("nextcloud-talk-recording-benchmark")
             audioSinkIndex = str(audioSinkIndex)
+            audioSourceIndex = str(audioSourceIndex)
 
             env = self._display.env()
             env['PULSE_SINK'] = audioSinkIndex
@@ -166,7 +167,7 @@ def run(self, args):
             recorderArgumentsBuilder.setFfmpegOutputAudio(args.audio_args.split())
             recorderArgumentsBuilder.setFfmpegOutputVideo(args.video_args.split())
             recorderArgumentsBuilder.setExtension(f".{extension}")
-            self._recorderArguments = recorderArgumentsBuilder.getRecorderArguments(status, self._display.new_display_var, audioSinkIndex, args.width, args.height, extensionlessFileName)
+            self._recorderArguments = recorderArgumentsBuilder.getRecorderArguments(status, self._display.new_display_var, audioSourceIndex, args.width, args.height, extensionlessFileName)
 
             self._fileName = self._recorderArguments[-1]
 

recording/src/nextcloud/talk/recording/RecorderArgumentsBuilder.py

@@ -36,13 +36,14 @@ def __init__(self):
         self._ffmpegOutputVideo = None
         self._extension = None
 
-    def getRecorderArguments(self, status, displayId, audioSinkIndex, width, height, extensionlessOutputFileName):
+    def getRecorderArguments(self, status, displayId, audioSourceIndex, width, height, extensionlessOutputFileName):
         """
         Returns the list of arguments to start the recorder process.
 
         :param status: whether to record audio and video or only audio.
         :param displayId: the ID of the display that the browser is running in.
-        :param audioSinkIndex: the index of the sink for the browser audio output.
+        :param audioSourceIndex: the index of the source for the browser audio
+               output.
         :param width: the width of the display and the recording.
         :param height: the height of the display and the recording.
         :param extensionlessOutputFileName: the file name for the recording, without
@@ -51,7 +52,7 @@ def getRecorderArguments(self, status, displayId, audioSinkIndex, width, height,
         """
 
         ffmpegCommon = ['ffmpeg', '-loglevel', 'level+warning', '-n']
-        ffmpegInputAudio = ['-f', 'pulse', '-i', audioSinkIndex]
+        ffmpegInputAudio = ['-f', 'pulse', '-i', audioSourceIndex]
         ffmpegInputVideo = ['-f', 'x11grab', '-draw_mouse', '0', '-video_size', f'{width}x{height}', '-i', displayId]
         ffmpegOutputAudio = self.getFfmpegOutputAudio()
         ffmpegOutputVideo = self.getFfmpegOutputVideo()

recording/src/nextcloud/talk/recording/Service.py

@@ -42,15 +42,18 @@ def newAudioSink(baseSinkName):
     Start new audio sink for the audio output of the browser.
 
     Each browser instance uses its own sink that will then be captured by the
-    recorder. Otherwise several browsers would use the same default sink, and
-    their audio output would be mixed.
+    recorder (using the monitor of that sink as the source). Otherwise several
+    browsers would use the same default sink, and their audio output would be
+    mixed.
 
     The sink is created by loading a null sink module. This module needs to be
-    unloaded once the sink is no longer needed to remove it.
+    unloaded once the sink is no longer needed to remove it (which also removes
+    the monitor of that sink).
 
     :param baseSinkName: the base name for the sink; it is expected to have been
            sanitized and to contain only alpha-numeric characters.
-    :return: a tuple with the module index and the sink index, both as ints.
+    :return: a tuple with the module index, the sink index and the source index
+             (the monitor of the sink), all as ints.
     """
 
     # A random value is appended to the base sink name to "ensure" that there
@@ -83,7 +86,16 @@ def newAudioSink(baseSinkName):
         if not sinkIndex:
             raise Exception(f"New audio sink {sinkName} not found ({sinkList})")
 
-    return moduleIndex, sinkIndex
+        sourceIndex = None
+        sourceList = pacmd.source_list()
+        for source in sourceList:
+            if source.monitor_of_sink == sinkIndex:
+                sourceIndex = source.index
+
+        if not sourceIndex:
+            raise Exception(f"Audio source (monitor) of sink {sinkName} not found ({sourceList})")
+
+    return moduleIndex, sinkIndex, sourceIndex
 
 def processLog(loggerName, pipe, level = logging.INFO):
     """
@@ -180,8 +192,9 @@ def start(self, actorType, actorId):
                 raise Exception("Display started after recording was stopped")
 
             # Start new audio sink for the audio output of the browser.
-            self._audioModuleIndex, audioSinkIndex = newAudioSink(f"{sanitizedBackend}-{self.token}")
+            self._audioModuleIndex, audioSinkIndex, audioSourceIndex = newAudioSink(f"{sanitizedBackend}-{self.token}")
             audioSinkIndex = str(audioSinkIndex)
+            audioSourceIndex = str(audioSourceIndex)
 
             if self._stopped.is_set():
                 raise Exception("Audio sink started after recording was stopped")
@@ -210,7 +223,7 @@ def start(self, actorType, actorId):
             extensionlessFileName = f'{fullDirectory}/Recording {datetime.now().strftime("%Y-%m-%d %H-%M-%S")}'
 
             recorderArgumentsBuilder = RecorderArgumentsBuilder()
-            recorderArguments = recorderArgumentsBuilder.getRecorderArguments(self.status, self._display.new_display_var, audioSinkIndex, width, height, extensionlessFileName)
+            recorderArguments = recorderArgumentsBuilder.getRecorderArguments(self.status, self._display.new_display_var, audioSourceIndex, width, height, extensionlessFileName)
 
             self._fileName = recorderArguments[-1]