Video Capture: extract storage info from pixel format in VideoCaptureFormat.

media/base/video_capture_types.h

Index: media/base/video_capture_types.h
diff --git a/media/base/video_capture_types.h b/media/base/video_capture_types.h
index 8483a2dd42f5cb03015dffda309f08efdae20727..34845290f96124d890720abf0d3495bbc5934462 100644
--- a/media/base/video_capture_types.h
+++ b/media/base/video_capture_types.h
@@ -29,12 +29,19 @@ enum VideoPixelFormat {
   PIXEL_FORMAT_RGB32,
   PIXEL_FORMAT_ARGB,
   PIXEL_FORMAT_MJPEG,
-  PIXEL_FORMAT_TEXTURE,  // Capture format as a GL texture.
-  PIXEL_FORMAT_GPUMEMORYBUFFER,
   PIXEL_FORMAT_UNKNOWN,  // Color format not set.
   PIXEL_FORMAT_MAX,
 };
 
+// Storage type for the pixels. In principle, all combinations of Storage and
+// Format are possible, though some are very typical, such as texture + ARGB,

[hubbe, 2015/06/18 19:23:11]

I don't think this is clear enough / struct.
What does texture + MJPEG mean? Is it allowed? Is the format ignored for
textures?

[mcasas, 2015/06/19 02:58:36]

Same as with VideoFrame, the specification of 
storage and pixel format are independent, so all
combinations are meaningful although perhaps some
are just not implemented. A texture MJPEG means
exactly what is says: a formatless buffer allocated
in the GPU (==texture) with data representing a
video frame in MJPEG pixel format.
No it isn't. But the only one used so far is ARGB.

[hubbe, 2015/06/23 23:35:40]

A texture is not a formatless buffer.
OpenGL doesn't give you direct access to the bytes of a texture, so it's not
possible to store arbitrary data in it.

+// and others are only available if the platform allows it e.g. GpuMemoryBuffer.
+enum VideoPixelStorage {

[hubbe, 2015/06/18 19:23:11]

Shouldn't this really be media::VideoFrame::StorageType?

[mcasas, 2015/06/19 02:58:36]

Again is only partially coincidental with
media::VideoFrame::StorageType, so although
it would be good, this is more reduced.

+  PIXEL_STORAGE_CPU,
+  PIXEL_STORAGE_TEXTURE,
+  PIXEL_STORAGE_GPUMEMORYBUFFER,
+};
+
 // Policies for capture devices that have source content that varies in size.
 // It is up to the implementation how the captured content will be transformed
 // (e.g., scaling and/or letterboxing) in order to produce video frames that
@@ -70,9 +77,15 @@ struct MEDIA_EXPORT VideoCaptureFormat {
   VideoCaptureFormat(const gfx::Size& frame_size,
                      float frame_rate,
                      VideoPixelFormat pixel_format);
+  VideoCaptureFormat(const gfx::Size& frame_size,
+                     float frame_rate,
+                     VideoPixelFormat pixel_format,
+                     VideoPixelStorage pixel_storage);
 
+  // TODO(mcasas): Remove this method http://crbug.com/501134.
   std::string ToString() const;
   static std::string PixelFormatToString(VideoPixelFormat format);
+  static std::string PixelStorageToString(VideoPixelStorage storage);
 
   // Returns the required buffer size to hold an image of a given
   // VideoCaptureFormat with no padding and tightly packed.
@@ -91,6 +104,7 @@ struct MEDIA_EXPORT VideoCaptureFormat {
   gfx::Size frame_size;
   float frame_rate;
   VideoPixelFormat pixel_format;
+  VideoPixelStorage pixel_storage;
 };
 
 typedef std::vector<VideoCaptureFormat> VideoCaptureFormats;