Adds SVGImageElement as a CanvasImageSource

third_party/WebKit/Source/core/svg/SVGImageElement.cpp

 /*
  * Copyright (C) 2004, 2005, 2006, 2008 Nikolas Zimmermann <zimmermann@kde.org>
  * Copyright (C) 2004, 2005, 2006, 2007, 2008, 2009 Rob Buis <buis@kde.org>
  * Copyright (C) 2006 Alexander Kellett <lypanov@kde.org>
  *
  * This library is free software; you can redistribute it and/or
  * modify it under the terms of the GNU Library General Public
  * License as published by the Free Software Foundation; either
  * version 2 of the License, or (at your option) any later version.
  *
@@ skipping 177 matching lines @@
     // A previous loader update may have failed to actually fetch the image if
     // the document was inactive. In that case, force a re-update (but don't
     // clear previous errors).
     if (!imageLoader().image())
       imageLoader().updateFromElement();
   }
 
   return InsertionDone;
 }
 
+FloatSize SVGImageElement::sourceDefaultSize() {
+  return FloatSize(0, 0);

[fs, 2017/03/07 22:01:25]

To be somewhat equivalent with the HTMLImageElement version, I think this would
be something like:

if (layoutObject())
  return toLayoutSVGImage(layoutObject())->objectBoundingBox().size();
SVGLengthContext lengthContext(this);
return FloatSize(m_width->currentValue()->value(lengthContext),
                 m_height->currentValue()->value(lengthContext));

I wonder if these width/heights shouldn't be the <canvas> width/height - reading
step 3 of https://html.spec.whatwg.org/multipage/scripting.html#drawing-images

[fserb, 2017/03/08 18:31:59]

Thanks for the code snippet, I'll implement it like this.

Regarding the spec thing, those functions are messed up
(sourceWidth/sourceHeight). They are not called by the CanvasImageSource code
path at all (they are used for BitmapImageSource). On the canvas side, we always
get called through getSourceImageForCanvas() which receives a
"defaultObjectSize" (which is the canvas size) -- I've added tests to make sure
that step 3 works (it's the 6th id on the test, iirc). Since those are not used
for the actual canvas rendering, I interpreted them as "the inherit dimension of
the image" and therefore I think your code makes sense (but I'll go over it
again for Bitmap source, maybe even remove it completely). 

Does this make sense to you?

[fs, 2017/03/08 18:57:16]

Ok (ImageBitmapSource though, right?). Codesearch seems to only show them being
called in modules/shapedetection though. Did it miss something?
If they are indeed only used by the shapedetection code, it makes perfect sense
to revisit later. So, if you want to go with the above, that would be fine with
me. I suspect that these are buggy/broken though, and should probably be
replaced/removed.

[fserb, 2017/03/21 19:15:22]

yep. Only for shapedetection, (which I assumed is only used for
ImageBitmapSource, no?)
Anyway, moot point. I've used your snippet and will look into those again for
ImageBitmapSoure.

[fs, 2017/03/21 21:05:41]

Acknowledged.

+}
+
 const AtomicString SVGImageElement::imageSourceURL() const {
   return AtomicString(hrefString());
 }
 
 void SVGImageElement::didMoveToNewDocument(Document& oldDocument) {
   imageLoader().elementDidMoveToNewDocument();
   SVGGraphicsElement::didMoveToNewDocument(oldDocument);
 }
 
 }  // namespace blink