Switch WASAPI playout to timer-driven mode with AUTOCONVERTPCM

  Use WAVEFORMATEXTENSIBLE and AUDCLNT_STREAMFLAGS_AUTOCONVERTPCM to
  support devices whose native format differs from 48kHz stereo PCM,
  such as Bluetooth HFP endpoints (16kHz mono) and surround sound
  devices (5.1/7.1).

  Replace the event-driven render loop (AUDCLNT_STREAMFLAGS_EVENTCALLBACK)
  with a timer-driven polling loop to avoid a known issue where
  AUTOCONVERTPCM combined with EVENTCALLBACK causes the audio engine to
  stop signaling render events, resulting in premature thread termination.

  Also adds 192kHz to the supported sample rate list, uses the first
  closest-match format from IsFormatSupported as a fallback when no
  exact match is found, and applies the same changes to the recording
  path.

Author: roxanneskelly

build/patches/windows_add_192k.patch

@@ -1,5 +1,5 @@
 diff --git a/modules/audio_device/win/audio_device_core_win.cc b/modules/audio_device/win/audio_device_core_win.cc
-index f1cc0474fc..f8e4f25bb9 100644
+index f1cc0474fc..2aacd1d309 100644
 --- a/modules/audio_device/win/audio_device_core_win.cc
 +++ b/modules/audio_device/win/audio_device_core_win.cc
 @@ -1845,8 +1845,8 @@ int32_t AudioDeviceWindowsCore::InitPlayout() {
@@ -142,14 +142,13 @@ index f1cc0474fc..f8e4f25bb9 100644
      RTC_LOG(LS_VERBOSE) << "Additional settings:";
      RTC_LOG(LS_VERBOSE) << "_playAudioFrameSize: " << _playAudioFrameSize;
      RTC_LOG(LS_VERBOSE) << "_playBlockSize     : " << _playBlockSize;
-@@ -1980,12 +2007,14 @@ int32_t AudioDeviceWindowsCore::InitPlayout() {
+@@ -1980,12 +2007,12 @@ int32_t AudioDeviceWindowsCore::InitPlayout() {
    }
    hr = _ptrClientOut->Initialize(
        AUDCLNT_SHAREMODE_SHARED,  // share Audio Engine with other applications
 -      AUDCLNT_STREAMFLAGS_EVENTCALLBACK,  // processing of the audio buffer by
-+      AUDCLNT_STREAMFLAGS_EVENTCALLBACK |  // processing of the audio buffer by
-                                           // the client will be event driven
-+          AUDCLNT_STREAMFLAGS_AUTOCONVERTPCM |
+-                                          // the client will be event driven
++      AUDCLNT_STREAMFLAGS_AUTOCONVERTPCM |  // let WASAPI handle format conversion
 +          AUDCLNT_STREAMFLAGS_SRC_DEFAULT_QUALITY,
        hnsBufferDuration,  // requested buffer capacity as a time value (in
                            // 100-nanosecond units)
@@ -159,7 +158,19 @@ index f1cc0474fc..f8e4f25bb9 100644
        NULL);              // session GUID
 
    if (FAILED(hr)) {
-@@ -2031,7 +2060,7 @@ int32_t AudioDeviceWindowsCore::InitPlayout() {
+@@ -2016,11 +2043,6 @@ int32_t AudioDeviceWindowsCore::InitPlayout() {
+                         << bufferFrameCount * _playAudioFrameSize << " bytes)";
+   }
+ 
+-  // Set the event handle that the system signals when an audio buffer is ready
+-  // to be processed by the client.
+-  hr = _ptrClientOut->SetEventHandle(_hRenderSamplesReadyEvent);
+-  EXIT_ON_ERROR(hr);
+-
+   // Get an IAudioRenderClient interface.
+   SAFE_RELEASE(_ptrRenderClient);
+   hr = _ptrClientOut->GetService(__uuidof(IAudioRenderClient),
+@@ -2031,7 +2053,7 @@ int32_t AudioDeviceWindowsCore::InitPlayout() {
    _playIsInitialized = true;
 
    CoTaskMemFree(pWfxOut);
@@ -168,7 +179,7 @@ index f1cc0474fc..f8e4f25bb9 100644
 
    RTC_LOG(LS_VERBOSE) << "render side is now initialized";
    return 0;
-@@ -2039,7 +2068,7 @@ int32_t AudioDeviceWindowsCore::InitPlayout() {
+@@ -2039,7 +2061,7 @@ int32_t AudioDeviceWindowsCore::InitPlayout() {
  Exit:
    _TraceCOMError(hr);
    CoTaskMemFree(pWfxOut);
@@ -177,7 +188,7 @@ index f1cc0474fc..f8e4f25bb9 100644
    SAFE_RELEASE(_ptrClientOut);
    SAFE_RELEASE(_ptrRenderClient);
    return -1;
-@@ -2163,7 +2192,7 @@ int32_t AudioDeviceWindowsCore::InitRecording() {
+@@ -2163,7 +2185,7 @@ int32_t AudioDeviceWindowsCore::InitRecording() {
    HRESULT hr = S_OK;
    WAVEFORMATEX* pWfxIn = NULL;
    WAVEFORMATEXTENSIBLE Wfx = WAVEFORMATEXTENSIBLE();
@@ -186,7 +197,7 @@ index f1cc0474fc..f8e4f25bb9 100644
 
    // Create COM object with IAudioClient interface.
    SAFE_RELEASE(_ptrClientIn);
-@@ -2201,7 +2230,7 @@ int32_t AudioDeviceWindowsCore::InitRecording() {
+@@ -2201,7 +2223,7 @@ int32_t AudioDeviceWindowsCore::InitRecording() {
    Wfx.Samples.wValidBitsPerSample = Wfx.Format.wBitsPerSample;
    Wfx.SubFormat = KSDATAFORMAT_SUBTYPE_PCM;
 
@@ -195,7 +206,7 @@ index f1cc0474fc..f8e4f25bb9 100644
    hr = S_FALSE;
 
    // Iterate over frequencies and channels, in order of priority
-@@ -2218,8 +2247,10 @@ int32_t AudioDeviceWindowsCore::InitRecording() {
+@@ -2218,8 +2240,10 @@ int32_t AudioDeviceWindowsCore::InitRecording() {
        // If the method succeeds and the audio endpoint device supports the
        // specified stream format, it returns S_OK. If the method succeeds and
        // provides a closest match to the specified format, it returns S_FALSE.
@@ -208,7 +219,7 @@ index f1cc0474fc..f8e4f25bb9 100644
        if (hr == S_OK) {
          break;
        } else {
-@@ -2228,10 +2259,14 @@ int32_t AudioDeviceWindowsCore::InitRecording() {
+@@ -2228,10 +2252,14 @@ int32_t AudioDeviceWindowsCore::InitRecording() {
                             << ", nSamplesPerSec=" << Wfx.Format.nSamplesPerSec
                             << " is not supported. Closest match: "
                                "nChannels="
@@ -227,7 +238,7 @@ index f1cc0474fc..f8e4f25bb9 100644
          } else {
            RTC_LOG(LS_INFO) << "nChannels=" << Wfx.Format.nChannels
                             << ", nSamplesPerSec=" << Wfx.Format.nSamplesPerSec
-@@ -2242,7 +2277,20 @@ int32_t AudioDeviceWindowsCore::InitRecording() {
+@@ -2242,7 +2270,20 @@ int32_t AudioDeviceWindowsCore::InitRecording() {
      if (hr == S_OK)
        break;
    }
@@ -249,7 +260,7 @@ index f1cc0474fc..f8e4f25bb9 100644
    if (hr == S_OK) {
      _recAudioFrameSize = Wfx.Format.nBlockAlign;
      _recSampleRate = Wfx.Format.nSamplesPerSec;
-@@ -2270,6 +2318,8 @@ int32_t AudioDeviceWindowsCore::InitRecording() {
+@@ -2270,6 +2311,8 @@ int32_t AudioDeviceWindowsCore::InitRecording() {
        AUDCLNT_SHAREMODE_SHARED,  // share Audio Engine with other applications
        AUDCLNT_STREAMFLAGS_EVENTCALLBACK |  // processing of the audio buffer by
                                             // the client will be event driven
@@ -258,7 +269,7 @@ index f1cc0474fc..f8e4f25bb9 100644
            AUDCLNT_STREAMFLAGS_NOPERSIST,   // volume and mute settings for an
                                             // audio session will not persist
                                             // across system restarts
-@@ -2321,7 +2371,7 @@ int32_t AudioDeviceWindowsCore::InitRecording() {
+@@ -2321,7 +2364,7 @@ int32_t AudioDeviceWindowsCore::InitRecording() {
    _recIsInitialized = true;
 
    CoTaskMemFree(pWfxIn);
@@ -267,7 +278,7 @@ index f1cc0474fc..f8e4f25bb9 100644
 
    RTC_LOG(LS_VERBOSE) << "capture side is now initialized";
    return 0;
-@@ -2329,7 +2379,7 @@ int32_t AudioDeviceWindowsCore::InitRecording() {
+@@ -2329,7 +2372,7 @@ int32_t AudioDeviceWindowsCore::InitRecording() {
  Exit:
    _TraceCOMError(hr);
    CoTaskMemFree(pWfxIn);
@@ -276,3 +287,42 @@ index f1cc0474fc..f8e4f25bb9 100644
    SAFE_RELEASE(_ptrClientIn);
    SAFE_RELEASE(_ptrCaptureClient);
    return -1;
+@@ -2661,7 +2704,6 @@ DWORD WINAPI AudioDeviceWindowsCore::WSAPICaptureThreadPollDMO(LPVOID context) {
+ 
+ DWORD AudioDeviceWindowsCore::DoRenderThread() {
+   bool keepPlaying = true;
+-  HANDLE waitArray[2] = {_hShutdownRenderEvent, _hRenderSamplesReadyEvent};
+   HRESULT hr = S_OK;
+   HANDLE hMmTask = NULL;
+ 
+@@ -2777,20 +2819,16 @@ DWORD AudioDeviceWindowsCore::DoRenderThread() {
+   // >> ------------------ THREAD LOOP ------------------
+ 
+   while (keepPlaying) {
+-    // Wait for a render notification event or a shutdown event
+-    DWORD waitResult = WaitForMultipleObjects(2, waitArray, FALSE, 500);
+-    switch (waitResult) {
+-      case WAIT_OBJECT_0 + 0:  // _hShutdownRenderEvent
+-        keepPlaying = false;
+-        break;
+-      case WAIT_OBJECT_0 + 1:  // _hRenderSamplesReadyEvent
+-        break;
+-      case WAIT_TIMEOUT:  // timeout notification
+-        RTC_LOG(LS_WARNING) << "render event timed out after 0.5 seconds";
+-        goto Exit;
+-      default:  // unexpected error
+-        RTC_LOG(LS_WARNING) << "unknown wait termination on render side";
+-        goto Exit;
++    // Sleep for half the device period, then poll for available buffer space.
++    // We use the shutdown event wait as the sleep mechanism so we can still
++    // wake immediately on shutdown.
++    DWORD sleepMs = static_cast<DWORD>(devPeriod / 10000 / 2);
++    if (sleepMs < 1)
++      sleepMs = 1;
++    DWORD waitResult = WaitForSingleObject(_hShutdownRenderEvent, sleepMs);
++    if (waitResult == WAIT_OBJECT_0) {
++      keepPlaying = false;
++      break;
+     }
+ 
+     while (keepPlaying) {