Add content::SurfaceCapturer

media/base/video_frame.cc

Index: media/base/video_frame.cc
diff --git a/media/base/video_frame.cc b/media/base/video_frame.cc
index 08e7e1ad7a236f3ab29d5e7ee6fc9da2098fb630..2bf9b8ffdf8a81393400065c6a5875c8ab9986a9 100644
--- a/media/base/video_frame.cc
+++ b/media/base/video_frame.cc
@@ -124,9 +124,13 @@ scoped_refptr<VideoFrame> VideoFrame::WrapExternalSharedMemory(
     const gfx::Rect& visible_rect,
     const gfx::Size& natural_size,
     uint8* data,
+    size_t data_size,
     base::SharedMemoryHandle handle,
     base::TimeDelta timestamp,
     const base::Closure& no_longer_needed_cb) {
+  if (data_size < AllocationSize(format, coded_size))
+    return NULL;
+
   switch (format) {
     case I420: {
       scoped_refptr<VideoFrame> frame(new VideoFrame(
@@ -243,6 +247,30 @@ size_t VideoFrame::NumPlanes(Format format) {
   return 0;
 }
 
+// static
+size_t VideoFrame::AllocationSize(Format format, const gfx::Size& coded_size) {

[ncarter (slow), 2013/08/15 20:46:59]

The behavior of this function does not match the padding used by AllocYUV() and
AllocRGB() in this class, and I worry that this is misleading.

It seems that what is really implemented here is the padding convention that is
assumed by the video capture code. VideoFrame::WrapExternalSharedMemory makes
the same assumption for YV12 data. How can we move to a more consistent
behavior?

In general, I think the right approach is for whoever allocates the memory to
pick strides/padding, and for this decision to be explicitly communicated to all
subsequent users. 

[sheu, 2013/08/15 21:30:48]

The problem is that I'm doing a lot of "if format then packed size = " kind of
thing all over for the capture pipeline, and I'd like to centralize it in one
place.  I could rename this "PackedAllocationSize" or something if it makes
things clearer.

This is essentially a helper for whoever that is which picks the
strides/padding.  When the VideoFrame is actually constructed with
WrapExternalSharedMemory, then width/stride are exposed to subsequent user as
visible_rect/coded_size, respectively.

+  switch (format) {
+    case VideoFrame::RGB32:
+      return coded_size.GetArea() * 4;
+    case VideoFrame::YV12:
+    case VideoFrame::I420:
+      return coded_size.GetArea() * 3 / 2;
+    case VideoFrame::YV16:
+      return coded_size.GetArea() * 2;
+    case VideoFrame::YV12A:
+      return coded_size.GetArea() * 5 / 4;
+    case VideoFrame::INVALID:
+    case VideoFrame::EMPTY:
+    case VideoFrame::NATIVE_TEXTURE:
+#if defined(GOOGLE_TV)
+    case VideoFrame::HOLE:
+#endif
+      break;
+  }
+  NOTREACHED() << "Unsupported video frame format: " << format;
+  return 0;
+}
+
 static inline size_t RoundUp(size_t value, size_t alignment) {
   // Check that |alignment| is a power of 2.
   DCHECK((alignment + (alignment - 1)) == (alignment | (alignment - 1)));