Pass camera facing to WebKit

content/common/media/media_devices.h

Index: content/common/media/media_devices.h
diff --git a/content/common/media/media_devices.h b/content/common/media/media_devices.h
index 6ac1e2c5d7e5de67e65b00002ac59b7cb7a59ed9..c5b563db495fc5372226431a2b892d34edea1d61 100644
--- a/content/common/media/media_devices.h
+++ b/content/common/media/media_devices.h
@@ -12,6 +12,7 @@
 
 namespace media {
 struct AudioDeviceDescription;
+struct VideoCaptureDeviceDescriptor;
 }
 
 namespace content {
@@ -31,9 +32,13 @@ struct CONTENT_EXPORT MediaDeviceInfo {
   explicit MediaDeviceInfo(
       const media::AudioDeviceDescription& device_description);
 
+  explicit MediaDeviceInfo(
+      const media::VideoCaptureDeviceDescriptor& descriptor);
   std::string device_id;
   std::string label;
   std::string group_id;
+  // Format is "vid:pid". Ex. "E2A1:34D0".
+  std::string model_id;

[jochen (gone - plz use gerrit), 2017/01/05 12:17:59]

why not a pair of ints then?

[shenghao, 2017/01/06 09:43:01]

We use the formatted "vid:pid" string as identifier in camera config files. When
we get model_id from VideoCaptureDeviceDescriptor, it's already in this format,
so converting it to 2 ints and back seems redundant. Plus that separated vid or
pid are not useful anywhere in the system.  

[wuchengli, 2017/01/09 08:28:00]

Another reason is we'll have non-USB cameras soon. Those cameras do not have vid
and pid.

 };
 
 using MediaDeviceInfoArray = std::vector<MediaDeviceInfo>;