Binding media stream audio track to speech recognition [renderer]

content/common/speech_recognition_messages.h

 // Copyright (c) 2012 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 // Multiply-included message file, hence no include guard.
 
 #include <string>
 
+#include "base/memory/shared_memory.h"
+#include "base/native_sync_socket.h"
+#include "base/process/process_handle.h"
 #include "content/public/common/speech_recognition_error.h"
 #include "content/public/common/speech_recognition_grammar.h"
 #include "content/public/common/speech_recognition_result.h"
 #include "ipc/ipc_message_macros.h"
 #include "ipc/ipc_param_traits.h"
+#include "media/audio/audio_parameters.h"
 #include "ui/gfx/rect.h"
 
+#if defined(OS_POSIX)
+#include "base/file_descriptor_posix.h"
+#endif
+
 #define IPC_MESSAGE_START SpeechRecognitionMsgStart
 
 IPC_ENUM_TRAITS_MAX_VALUE(content::SpeechAudioErrorDetails,
                           content::SPEECH_AUDIO_ERROR_DETAILS_LAST)
 IPC_ENUM_TRAITS_MAX_VALUE(content::SpeechRecognitionErrorCode,
                           content::SPEECH_RECOGNITION_ERROR_LAST)
 
 IPC_STRUCT_TRAITS_BEGIN(content::SpeechRecognitionError)
   IPC_STRUCT_TRAITS_MEMBER(code)
   IPC_STRUCT_TRAITS_MEMBER(details)
@@ skipping 29 matching lines @@
   // Speech grammars to use.
   IPC_STRUCT_MEMBER(content::SpeechRecognitionGrammarArray, grammars)
   // URL of the page (or iframe if applicable).
   IPC_STRUCT_MEMBER(std::string, origin_url)
   // Maximum number of hypotheses allowed for each results.
   IPC_STRUCT_MEMBER(uint32, max_hypotheses)
   // Whether the user requested continuous recognition or not.
   IPC_STRUCT_MEMBER(bool, continuous)
   // Whether the user requested interim results or not.
   IPC_STRUCT_MEMBER(bool, interim_results)
+  // Wheter user has set an audio track as input
+  IPC_STRUCT_MEMBER(bool, using_audio_track)
+  // Process handle for the renderer
+  IPC_STRUCT_MEMBER(base::ProcessHandle, peer_process_handle)
 IPC_STRUCT_END()
 
 
 // Requests the speech recognition service to start speech recognition.
 IPC_MESSAGE_CONTROL1(SpeechRecognitionHostMsg_StartRequest,
                      SpeechRecognitionHostMsg_StartRequest_Params)
 
 // Requests the speech recognition service to abort speech recognition on
 // behalf of the given |render_view_id| and |request_id|. If there are no
 // sessions associated with the |request_id| in the render view, this call
@@ skipping 32 matching lines @@
 
 IPC_MESSAGE_ROUTED1(SpeechRecognitionMsg_AudioStarted, int /* request_id */)
 
 IPC_MESSAGE_ROUTED1(SpeechRecognitionMsg_SoundStarted, int /* request_id */)
 
 IPC_MESSAGE_ROUTED1(SpeechRecognitionMsg_SoundEnded, int /* request_id */)
 
 IPC_MESSAGE_ROUTED1(SpeechRecognitionMsg_AudioEnded, int /* request_id */)
 
 IPC_MESSAGE_ROUTED1(SpeechRecognitionMsg_Ended, int /* request_id */)
+
+IPC_MESSAGE_ROUTED5(SpeechRecognitionMsg_AudioTrackReady,

[tommi (sloooow) - chröme, 2014/08/29 11:25:30]

is this approach the same as what we do elsewhere where we use SyncSocket? (I
ask since NativeSyncSocket is a new class but we've been using SyncSocket for a
while)

[no longer working on chromium, 2014/08/29 12:23:06]

I have the same concern here, Tommi, FYI, our existing code is using #if to
handle the case:
#if defined(OS_WIN)
IPC_MESSAGE_CONTROL4(AudioMsg_NotifyStreamCreated,
                     int /* stream id */,
                     base::SharedMemoryHandle /* handle */,
                     base::SyncSocket::Handle /* socket handle */,
                     uint32 /* length */)
#else
IPC_MESSAGE_CONTROL4(AudioMsg_NotifyStreamCreated,
                    int /* stream id */,
                    base::SharedMemoryHandle /* handle */,
                    base::FileDescriptor /* socket handle                       
                
                    uint32 /* length */)
#endif

Though, I personally don't like having a NativeSyncSocket in base.

[burnik, 2014/08/29 13:26:16]

I think it would be ok if we had |base::SyncSocket::SharedDescriptor|. It would
make more sense and indicate that this descriptor is to be used by the peer
process.
On 2014/08/29 11:25:30, tommi wrote:

+                    int /* request_id */,
+                    media::AudioParameters /* params */,
+                    base::SharedMemoryHandle /* memory */,
+                    base::NativeSyncSocket::Descriptor /* socket */,
+                    int /* length */)