Bug 222343 – [GPU Process] Some DisplayList items may not be replayed back before calling getImageData()

• Component: Canvas
• Status: RESOLVED FIXED

[Said Abou-Hallawa]

Ensuring the DisplayList items are replayed to the ImageBuffer before calling getImageData() will fix these tests:

fast/canvas/canvas-arc-connecting-line.html [ Failure ]
fast/canvas/canvas-arc-zero-lineto.html [ Failure ]
fast/canvas/canvas-bezier-same-endpoint.html [ Failure Pass ]
fast/canvas/canvas-blend-image.html [ Failure Pass ]
fast/canvas/canvas-blend-solid.html [ Failure Pass ]
imported/w3c/web-platform-tests/html/canvas/element/imagebitmap/createImageBitmap-origin.sub.html [ Failure ]

[Said Abou-Hallawa]

Created attachment 421383 [details]
Patch

[Said Abou-Hallawa]

Created attachment 421710 [details]
Patch

[Tim Horton]

Comment on attachment 421710 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=421710&action=review

> Source/WebKit/ChangeLog:10
> +        Make sure all the pending DisplayList items are pushed to the GPUP and
> +        submitted to RemoteImageBuffer before calling the asynchronous method
> +        getImageData().

Is there an assertion we could add somewhere that would have alerted us to this?

[Said Abou-Hallawa]

Comment on attachment 421710 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=421710&action=review

>> Source/WebKit/ChangeLog:10
>> +        getImageData().
> 
> Is there an assertion we could add somewhere that would have alerted us to this?

This is going to be costly because we have to know the state of RemoteImageBuffer playing back in WebP. This will require sending a confirmation message from GPUP to WebP for every playback.

[Tim Horton]

(In reply to Said Abou-Hallawa from comment #4)
> Comment on attachment 421710 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=421710&action=review
> 
> >> Source/WebKit/ChangeLog:10
> >> +        getImageData().
> > 
> > Is there an assertion we could add somewhere that would have alerted us to this?
> 
> This is going to be costly because we have to know the state of
> RemoteImageBuffer playing back in WebP. This will require sending a
> confirmation message from GPUP to WebP for every playback.

That's probably beyond the scope of a reasonable assertion, then.

[EWS]

Committed r273672: <https://commits.webkit.org/r273672>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 421710 [details].

[Radar WebKit Bug Importer]

<rdar://problem/74882498>