387202 - [FIX]Canvas shouldn't use direct same-origin check

Status: RESOLVED FIXED

(Core :: Graphics: Canvas2D, defect, P2)

Description

[Boris Zbarsky [:bzbarsky]]

Attachment: Proposed patch

Instead, this should be an Equals() or Subsumes() check.  I think Subsumes() is right, so that a chrome canvas won't become readonly when something is painted into it.

Comment 1

[Boris Zbarsky [:bzbarsky]]

On the other hand, this will fail for a signed cert script which is grabbing an image from the same jar.  Do we care?  Unfortunately, we can't tell for the cert without doing the image load...

Perhaps images should just expose the channel principal on the image?

Comment 2

[Boris Zbarsky [:bzbarsky]]

To be exact, we could expose the principal of the channel that loaded the image on imgIRequest and possible on nsIImageLoadingContent as a shortcut.  That would be a much better setup than using the image URI, in my opinion.

Comment 3

[Boris Zbarsky [:bzbarsky]]

Then again, that raises the question of how document.domain should be handled.  Right now, it's handled badly.  Perhaps we'd want the sort of principal version of CheckConnect I mention in bug 387216 comment 1 here?

Comment 4

[Boris Zbarsky [:bzbarsky]]

And nsCanvasRenderingContextGLPrivate really has the same issues...

Comment 5

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

Comment on attachment 271312 [details] [diff] [review]
Proposed patch

Ignore GLPrivate, it needs some more work :)

Comment 6

[Boris Zbarsky [:bzbarsky]]

Checked in.  Filed bug 389048 on imagelib exposing a better setup here.