Added video capture capabilities retrieval and caching to VideoCaptureManager

content/browser/renderer_host/media/video_capture_manager.cc

Index: content/browser/renderer_host/media/video_capture_manager.cc
diff --git a/content/browser/renderer_host/media/video_capture_manager.cc b/content/browser/renderer_host/media/video_capture_manager.cc
index feb556df2267d56df7cdcf365e6aca857265ce77..56f5230c5e75ed748d53ef2694dc3f00e188aaa8 100644
--- a/content/browser/renderer_host/media/video_capture_manager.cc
+++ b/content/browser/renderer_host/media/video_capture_manager.cc
@@ -150,6 +150,12 @@ void VideoCaptureManager::DoStartDeviceOnDeviceThread(
         video_capture_device.reset(use_fake_device_ ?
             media::FakeVideoCaptureDevice::Create(*found) :
             media::VideoCaptureDevice::Create(*found));
+
+        // Filter capture capabilities: remove all except the current one.
+        media::VideoCaptureCapabilities& formats =

[perkj_chrome, 2013/10/23 13:19:45]

What happen after the camera has been stopped? Then you can restart the camera
with all capabilities.

+            video_capture_capabilities_[entry->id];
+        formats.clear();
+        formats.push_back(capture_params);
       }
       break;
     }
@@ -252,6 +258,19 @@ void VideoCaptureManager::StopCaptureForClient(
   DestroyDeviceEntryIfNoClients(entry);
 }
 
+const media::VideoCaptureCapabilities*
+VideoCaptureManager::EnumerateDeviceCapabilities(
+    const StreamDeviceInfo& device_info) {
+  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));

[perkj_chrome, 2013/10/23 13:19:45]

What is the plan when it comes to multiple opening of a camera - maybe from
multiple processes?

Shouln't the |device_info| that actually started the device also continue to get
the full capability list?

[perkj_chrome, 2013/10/23 13:19:45]

So this is done on the IO thread but you modify |video_capture_capabilities_| on
the device thread, thats a no no.

On what thread will the request for the cabilities come from? I suspect the IO
thread. Can you do all processing of |video_capture_capabilities_| on the IO
thread? 

+
+  // Find the device we are looking for and return its associated capabilities.
+  std::map<std::string, media::VideoCaptureCapabilities>::iterator it =
+      video_capture_capabilities_.find(device_info.device.id);
+  if ( it != video_capture_capabilities_.end())
+    return &(it->second);
+  return NULL;
+}
+
 void VideoCaptureManager::DoStopDeviceOnDeviceThread(DeviceEntry* entry) {
   SCOPED_UMA_HISTOGRAM_TIMER("Media.VideoCaptureManager.StopDeviceTime");
   DCHECK(IsOnDeviceThread());
@@ -321,8 +340,21 @@ VideoCaptureManager::GetAvailableDevicesOnDeviceThread(
       // enumerate the devices again.
       if (!use_fake_device_) {
         media::VideoCaptureDevice::GetDeviceNames(&result);
+        // Add to the cache the devices' supported capture formats.
+        media::VideoCaptureDevice::Names::iterator name_it;

[perkj_chrome, 2013/10/23 13:19:45]

See how GetAvailableDevicesOnDeviceThread is used for an idea of how to get the
capabilities to the IO thread. 

I think you will need a separate function for retrieving the capabilities back
to the IO thread. 

+        for (name_it = result.begin(); name_it != result.end(); ++name_it) {
+          media::VideoCaptureDevice::GetDeviceSupportedFormats(
+              *name_it, &video_capture_capabilities_[name_it->id()] );
+        }
       } else {
         media::FakeVideoCaptureDevice::GetDeviceNames(&result);
+        // Add to the cache the devices' supported capture formats.
+        media::VideoCaptureDevice::Names::iterator name_it;
+        media::VideoCaptureCapabilities formats;
+        for (name_it = result.begin(); name_it != result.end(); ++name_it) {
+          media::FakeVideoCaptureDevice::GetDeviceSupportedFormats(
+              *name_it, &video_capture_capabilities_[name_it->id()] );
+        }
       }
 
       // TODO(nick): The correctness of device start depends on this cache being