Bug 97673 – MediaStream API: Allow RTCPeerConnection to pass down its Document reference to the embedder

• Status: RESOLVED
• Resolution: INVALID
• Component: WebCore Misc.

[Tommy Widenflycht]

This patch tries to accomplish the same thing as my recent UserMediaRequest patch (97095), namely to pass down a Document reference to the embedder. 
However since this code lives under WebCore/platform there isn't a clear way to do so. Creating a frame client and sending it down that way has been suggested but that would require even more cludges. In that light something like this might be a compromise?

[Tommy Widenflycht]

Created attachment 165786 [details]
Patch

[Tommy Widenflycht]

Note, this patch is for discussion only at this stage.

[WebKit Review Bot]

Please wait for approval from [e-mail], [e-mail], [e-mail], [e-mail] or [e-mail] before submitting, as this patch contains changes to the Chromium public API. See also https://trac.webkit.org/wiki/ChromiumWebKitAPI.

[Adam Barth]

Comment on attachment 165786 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=165786&action=review

> Source/Platform/chromium/public/WebRTCPeerConnectionHandler.h:35
> +class WebDocument;

You can't reference WebDocument in Source/Platform.  Document is a DOM-level concept, not a Platform-level concept.

> Source/WebCore/platform/mediastream/RTCPeerConnectionHandler.cpp:50
> -    virtual bool initialize(PassRefPtr<RTCConfiguration>, PassRefPtr<MediaConstraints>) OVERRIDE;
> +    virtual bool initialize(Document*, PassRefPtr<RTCConfiguration>, PassRefPtr<MediaConstraints>) OVERRIDE;

This is also a layering violation.  Code in WebCore/platform shouldn't know anything about Document.

> Source/WebKit/chromium/src/RTCPeerConnectionHandlerChromium.cpp:249
> +namespace WebCore {
> +
> +PassOwnPtr<RTCPeerConnectionHandler> RTCPeerConnectionHandler::create(RTCPeerConnectionHandlerClient* client)
> +{
> +    return adoptPtr(new WebKit::RTCPeerConnectionHandlerChromium(client));
> +}
> +
> +} // namespace WebCore

This is wrong.  We shouldn't have code in Source/WebKit/chromium that's in the WebCore namespace.  I know other features do this, but we're trying to unwind all these layering violations so we can split WebCore and WebKit into separate DLLs.

[Adam Barth]

Maybe we should talk in a more interactive medium to work out this issue?