[webkit-reviews] review denied: [Bug 44539] API changes for Video Frame sharing between WebKit and Chromium : [Attachment 65439] Patch

[bugzilla-daemon at webkit.org]

Darin Fisher (:fishd, Google) <fishd at chromium.org> has denied Victoria
<vrk at google.com>'s request for review:
Bug 44539: API changes for Video Frame sharing between WebKit and Chromium
https://bugs.webkit.org/show_bug.cgi?id=44539

Attachment 65439: Patch
https://bugs.webkit.org/attachment.cgi?id=65439&action=review

------- Additional Comments from Darin Fisher (:fishd, Google)
<fishd at chromium.org>
WebCore/platform/graphics/chromium/VideoFrameChromium.h:38
 +  
nit: extra new line here

WebCore/platform/graphics/chromium/VideoFrameChromium.h:71
 +  public:
nit: no need for the redundant 'public:' section

WebCore/platform/graphics/chromium/VideoFrameProvider.h:39
 +  
nit: extra new line here

WebCore/platform/graphics/chromium/VideoFrameProvider.h:41
 +	virtual VideoFrameChromium* getCurrentFrame() = 0;
please document the ownership model.  who owns the VideoFrameChromium?
how do I free it?  it looks like i have to call putCurrentFrame to
free it, but is that the only way?  it might help if these methods
were more fully documented.

WebKit/chromium/public/WebMediaPlayer.h:133
 +	// FIXME: Will be made pure virtual when API changes are fully landed.
WebKit API interfaces implemented by the embedder are intended to have
non-pure virtual methods.  so this FIXME is backwards.	the other methods
should have empty implementations :)

WebKit/chromium/public/WebVideoFrame.h:40
 +  
nit: kill the extra new line here

WebKit/chromium/public/WebVideoFrame.h:43
 +	enum Format {
nit: our convention for enums at the WebKit API level is to use the following
format:

  enum Foo {
      FooBar,
      FooBaz,
      ...
  };

So, I'd expect to see:

  enum Format {
      FormatInvalid,
      FormatRGB555,
      ...
  };

  enum SurfaceType {
      SurfaceTypeSystemMemory,
      ...
  };

WebKit/chromium/public/WebVideoFrame.h:66
 +	virtual SurfaceType type() const = 0;
nit: type should be surfaceType to match the type name SurfaceType

WebKit/chromium/public/WebVideoFrame.h:71
 +	virtual int stride(unsigned plane) const = 0;
why does stride return signed instead of unsigned?

WebKit/chromium/public/WebVideoFrame.h:72
 +	virtual void* data(unsigned plane) const = 0;
can this pointer be a |const void*| instead?  it should be const
if you only expect the caller to get read access to the buffer.
otherwise, non-const implies mutability.  i think you only want
to provide read access, right?

WebKit/chromium/src/VideoFrameChromiumImpl.cpp:55
 +  VideoFrameChromiumImpl::VideoFrameChromiumImpl(WebVideoFrame*
webVideoFrame) :
nit: the colon should be on the next line

WebKit/chromium/src/VideoFrameChromiumImpl.h:42
 +  
nit: no new line here

WebKit/chromium/src/VideoFrameChromiumImpl.cpp:47
 +  WebVideoFrame*
VideoFrameChromiumImpl::toWebVideoFrame(WebCore::VideoFrameChromium*
videoFrame)
nit: i recommend adding a 'using namespace WebCore' at the top of this file,
and then remove the WebCore:: prefixes in the body of this file.

WebKit/chromium/src/VideoFrameChromiumImpl.cpp:56
 +	m_webVideoFrame(webVideoFrame)
what is the ownership model for m_webVideoFrame?  it looks like
VideoFrameChromiumImpl
is not responsible for deleting m_webVideoFrame.  should it?

WebKit/chromium/src/WebMediaPlayerClientImpl.h:119
 +	// This function returns a pointer to a VideoFrameChromium, which is
this looks like documentation that should be in VideoFrameProvider.h instead of
here.