Do not call SendMessageToNativeLog on Android in AudioInputSyncWriter

content/browser/renderer_host/media/audio_input_sync_writer.cc

 // 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.
 
 #include "content/browser/renderer_host/media/audio_input_sync_writer.h"
 
 #include <algorithm>
 
 #include "base/memory/shared_memory.h"
 #include "content/browser/renderer_host/media/media_stream_manager.h"
+#include "content/public/browser/browser_thread.h"
 
 static const uint32 kLogDelayThreadholdMs = 500;
 
 namespace content {
 
 AudioInputSyncWriter::AudioInputSyncWriter(
     base::SharedMemory* shared_memory,
     int shared_memory_segment_count)
     : shared_memory_(shared_memory),
       shared_memory_segment_count_(shared_memory_segment_count),
@@ skipping 22 matching lines @@
     base::TimeDelta interval = base::Time::Now() - creation_time_;
     oss << "Audio input data received for the first time: delay = "
         << interval.InMilliseconds() << "ms.";
   } else {
     base::TimeDelta interval = base::Time::Now() - last_write_time_;
     if (interval.InMilliseconds() > kLogDelayThreadholdMs) {
       oss << "Audio input data delay unexpectedly long: delay = "
          << interval.InMilliseconds() << "ms.";
     }
   }
-  if (!oss.str().empty())
+  if (!oss.str().empty()) {
+#if defined(OS_ANDROID)

[Ami GONE FROM CHROMIUM, 2014/05/21 18:03:38]

vrk@ points out that logging is not available on android anyway b/c of lack of
extension API, so how about you just conditionalize l.65 on !defined(OS_ANDROID)
and call it a day?

+    // MediaStreamManager::SendMessageToNativeLog posts a task to the UI thread,
+    // which will attach the audio thread to the Android java VM. Unlike chrome
+    // created threads, the audio thread is owned by the OS and does not detach
+    // itself from the VM on exit, causing a crash (crbug/365915). So we post to
+    // the IO thread first to avoid attaching the audio thread to VM.

[Ami GONE FROM CHROMIUM, 2014/05/21 17:51:42]

I don't get it. 
How is it that the PostTask that MediaStreamManager::SendMessageToNativeLog does
(and which is _all_ MSM::SMTNL() does) triggers an attachment of the calling
thread to the JVM, while the PostTask here does not?

[jiayl, 2014/05/21 17:59:07]

See https://code.google.com/p/chromium/issues/detail?id=365915#c45.
It seems only posting to the UI thread will cause VM attached.

[Feng Qian, 2014/05/21 18:08:33]

jiayl, can you build and test the new patch locally? It should give a definitive
answer to Ami's question by testing the patch using instructions I sent you.

On 2014/05/21 17:59:07, jiayl wrote:

[jiayl, 2014/05/21 18:13:58]

Code removed for Android now.

+    BrowserThread::PostTask(
+        BrowserThread::IO, FROM_HERE,
+        base::Bind(&MediaStreamManager::SendMessageToNativeLog, oss.str()));
+#else
     MediaStreamManager::SendMessageToNativeLog(oss.str());
+#endif
+  }
 
   last_write_time_ = base::Time::Now();
 
   uint8* ptr = static_cast<uint8*>(shared_memory_->memory());
   ptr += current_segment_id_ * shared_memory_segment_size_;
   media::AudioInputBuffer* buffer =
       reinterpret_cast<media::AudioInputBuffer*>(ptr);
   buffer->params.volume = volume;
   buffer->params.size = size;
   buffer->params.key_pressed = key_pressed;
@@ skipping 33 matching lines @@
     base::ProcessHandle process_handle,
     base::FileDescriptor* foreign_handle) {
   foreign_handle->fd = foreign_socket_->handle();
   foreign_handle->auto_close = false;
   return (foreign_handle->fd != -1);
 }
 
 #endif
 
 }  // namespace content