Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(224)

Unified Diff: content/renderer/media/audio_device.h

Issue 9104043: Monitor the IO message loop in the AudioDevice classes. (Closed) Base URL: svn://svn.chromium.org/chrome/trunk/src
Patch Set: Created 8 years, 11 months ago
Use n/p to move between diff chunks; N/P to move between comments. Draft comments are only viewable by you.
Jump to:
View side-by-side diff with in-line comments
Download patch
« no previous file with comments | « no previous file | content/renderer/media/audio_device.cc » ('j') | no next file with comments »
Expand Comments ('e') | Collapse Comments ('c') | Show Comments Hide Comments ('s')
Index: content/renderer/media/audio_device.h
diff --git a/content/renderer/media/audio_device.h b/content/renderer/media/audio_device.h
index de6ba3b0e2df9da6bd6a11b4fe5c44c5bee67b56..0d48fc84e3ad4dd0ebc51af8c43f4293bc31a314 100644
--- a/content/renderer/media/audio_device.h
+++ b/content/renderer/media/audio_device.h
@@ -66,7 +66,9 @@
#include <vector>
#include "base/basictypes.h"
+#include "base/bind.h"
#include "base/memory/scoped_ptr.h"
+#include "base/message_loop.h"
#include "base/shared_memory.h"
#include "base/synchronization/lock.h"
#include "base/threading/simple_thread.h"
@@ -75,14 +77,77 @@
#include "media/audio/audio_parameters.h"
#include "media/base/audio_renderer_sink.h"
-namespace base {
-class WaitableEvent;
-}
+// A common base class for AudioDevice and AudioInputDevice that manages
+// the IO message loop pointer and monitors it for destruction.
+// When the loop goes away, we automatically stop any running streams.
+template <class T>
+class AudioDeviceIOLoopObserver
+ : public MessageLoop::DestructionObserver {
+ public:
+ AudioDeviceIOLoopObserver()
+ : io_loop_(
+ ChildProcess::current()->io_message_loop()->message_loop_proxy()) {
+ // Watch the IO loop and make sure that we stop the audio thread before we
+ // lose the IO thread.
+ ObserveIOLoopDestruction(true, NULL);
+ }
+
+ ~AudioDeviceIOLoopObserver() {
+ ObserveIOLoopDestruction(false, NULL);
+ }
+
+ protected:
+ // Accessor to the IO loop that's used by the derived class.
+ base::MessageLoopProxy* loop() { return io_loop_; }
+
+ private:
+ // MessageLoop::DestructionObserver implementation.
+ // Called on the IO thread when the thread is going away.
+ virtual void WillDestroyCurrentMessageLoop() OVERRIDE {
+ DCHECK(io_loop_->BelongsToCurrentThread());
+ // The IO thread is terminating. We must stop the audio thread before the
+ // IO thread goes away.
+ static_cast<T*>(this)->ShutDownOnIOThread(NULL);
+ }
+
+ // Call from the IO thread to add or remove ourselves from the list of
+ // destruction observers for the IO message loop.
+ void ObserveIOLoopDestruction(bool enable, base::WaitableEvent* done) {
+ if (io_loop_->BelongsToCurrentThread()) {
+ MessageLoop* loop = MessageLoop::current();
+ enable ? loop->AddDestructionObserver(this) :
enal1 2012/01/31 17:23:54 Why not regular if statement?
tommi (sloooow) - chröme 2012/02/01 14:50:49 Done.
+ loop->RemoveDestructionObserver(this);
+ } else {
+ base::WaitableEvent event(false, false);
+ if (io_loop_->PostTask(FROM_HERE,
enal1 2012/01/31 17:23:54 Are you sure we can post anything on IO loop if lo
tommi (sloooow) - chröme 2012/02/01 14:50:49 It's OK if the loop is being deleted and PostTask
+ base::Bind(
+ &AudioDeviceIOLoopObserver<T>::ObserveIOLoopDestruction,
+ base::Unretained(this), enable, &event))) {
+ event.Wait();
+ } else {
+ // The IO thread has already terminated, so no need to wait.
+ }
+ }
+
+ if (done)
+ done->Signal();
+ }
+
+ // A pointer to the IO thread's message loop. That's here we do control the
+ // stream from. It is possible during shutdown that the IO thread has gone
+ // away before the user of the AudioDevice instance has called Stop().
+ // So, we need to be prepared to stop and take down the audio thread if
+ // the IO thread ever goes away while we're still playing.
+ // So, when this pointer is NULL, the IO thread has gone away and all methods
+ // that post to it, must check for this first.
enal1 2012/01/31 17:23:54 I don't see how this pointer can become NULL. It i
tommi (sloooow) - chröme 2012/02/01 14:50:49 It cannot become NULL - did you see any checks for
+ scoped_refptr<base::MessageLoopProxy> io_loop_;
+};
class CONTENT_EXPORT AudioDevice
: NON_EXPORTED_BASE(public media::AudioRendererSink),
public AudioMessageFilter::Delegate,
- public base::DelegateSimpleThread::Delegate {
+ public base::DelegateSimpleThread::Delegate,
+ public AudioDeviceIOLoopObserver<AudioDevice> {
public:
// Methods called on main render thread -------------------------------------
@@ -154,6 +219,8 @@ class CONTENT_EXPORT AudioDevice
void InitializeOnIOThread(const AudioParameters& params);
void PlayOnIOThread();
void PauseOnIOThread(bool flush);
+ // Allow AudioDeviceIOLoopObserver to call ShutDownOnIOThread.
+ friend class AudioDeviceIOLoopObserver<AudioDevice>;
void ShutDownOnIOThread(base::WaitableEvent* signal);
void SetVolumeOnIOThread(double volume);
« no previous file with comments | « no previous file | content/renderer/media/audio_device.cc » ('j') | no next file with comments »

Powered by Google App Engine
This is Rietveld 408576698