Rasterizing an in-document SVG to Canvas

In order to answer this question properly I thought that I would:

  1. Convert an in-document SVG file to a data URL
  2. Load it into an <img>
  3. Draw that <img> to a <canvas>
  4. Convert that <canvas> to a PNG data URL
  5. Load that data URL into an image.

I have an example of this attempt here:
http://phrogz.net/SVG/svg_to_png.xhtml

In Firefox I get (NS_ERROR_NOT_AVAILABLE) [nsIDOMCanvasRenderingContext2D.drawImage] when attempting to draw the image to the canvas in step 3.

Why do I get this error in Firefox, or how do I work around it?

In Chrome I get a SECURITY_ERR exception when I try to call toDataURL() in step 4.

Why would I get this error in Chrome, or how do I work around it?

The WhatWG specs state that the origin for an image "generated from a data: URL found in another Document or in a script" should be the same as that document. All data in this test is from the same document.


Solution 1:

In Firefox, SVG images taint the canvas but we're working on removing that restriction in https://bugzilla.mozilla.org/show_bug.cgi?id=672013 when that bug lands what you're trying to do should be possible. This restriction has been removed in FFv12.

All browsers implement similar restrictions and all are working on removing them; generally by making what you can do in SVG images more restricted. For instance we don't want to get back to the bad old days of being able to work out what sites you've visited by creating an SVG image with links in it and then reading off the colour of the links using canvas.

In addition at the moment firefox requires that the svg element have width and height attributes in order to render to canvas. This restriction may be removed in future.

Edit by Phrogz: as noted in my comment above, I received independent confirmation that the security issue in Chrome is the same issue: Chrome (currently) always taints a canvas when an SVG document is drawn to it.