Bug 35453 – Need to move items that CanvasRenderingContext2D depends on into CanvasSurface.

• Status: RESOLVED
• Resolution: FIXED
• Component: DOM

[David Levin]

Created attachment 49638 [details]
Here's the changes inplace in HTMLCanvasElement (to make the diff clearer).

This is step 2a from https://bugs.webkit.org/show_bug.cgi?id=34909#c2.

[David Levin]

Created attachment 49646 [details]
Proposed fix.

See the previous patch for the same changes but in an easier to diff format.

[Oliver Hunt]

Comment on attachment 49646 [details]
Proposed fix.

> +    // FIXME: Make isSupportedImageMIMETypeForEncoding threadsafe (to allow this method to be used on a worker thread).
> +    if (mimeType.isNull() || !MIMETypeRegistry::isSupportedImageMIMETypeForEncoding(mimeType))

It may be worth (for now) adding an isOnMainThread() assertion to this code path so that we can't accidentally forget to fix this.

> +        return buffer()->toDataURL("image/png");
> +
> +    return buffer()->toDataURL(mimeType);

I wonder how thread safe the image _encoders_ are.


i wonder what the DPI of an offscreen canvas should be...

> +IntRect CanvasSurface::convertLogicalToDevice(const FloatRect& logicalRect) const
> +{


I'm okay r+ing this, but i thought i'd see if you have any comments on these before i do.

[David Levin]

Comment on attachment 49646 [details]
Proposed fix.

I'll do something for Oliver's comments and upload a new patch.

[David Levin]

Created attachment 49859 [details]
Proposed fix.

The only difference between this patch and the last one is adding a few isMainThread() asserts (and corresponding changelog entries).

[David Levin]

(In reply to comment #2)
> (From update of attachment 49646 [details])
> > +    // FIXME: Make isSupportedImageMIMETypeForEncoding threadsafe (to allow this method to be used on a worker thread).
> > +    if (mimeType.isNull() || !MIMETypeRegistry::isSupportedImageMIMETypeForEncoding(mimeType))
> 
> It may be worth (for now) adding an isOnMainThread() assertion to this code
> path so that we can't accidentally forget to fix this.

done. I also added bug https://bugs.webkit.org/show_bug.cgi?id=35616 for this.

> I wonder how thread safe the image _encoders_ are.

I looked at the code (for a while and for several different platforms) and it appeared pretty safe.

I added a few asserts in some places that may need fixes. I also added bug https://bugs.webkit.org/show_bug.cgi?id=35615 and made it block the "master" bug.


> i wonder what the DPI of an offscreen canvas should be...

Me too :)

Something I'll have to figure out as I get there but my instinct is to make it match the document that the worker came from.

[David Levin]

Created attachment 49864 [details]
Proposed fix (added the modified changelog).

[Oliver Hunt]

Comment on attachment 49864 [details]
Proposed fix (added the modified changelog).

r=me

[David Levin]

Committed as http://trac.webkit.org/changeset/56447