[chromium] Tiled compositor crash after recovering from lost GPU process

[Adrienne Walker]

From http://crbug.com/68151:
1. Go to http://webkit.org/blog/386/3d-transforms/ with the compositor enabled and scroll down a bit.
2. From the task manager kill the GPU process.
3. Move the mouse cursor in the window to trigger redraw.

This appears to be due to the fact that the layer is not being resized properly after the GPU process recovers and there's an assertion on an out-of-range indexing operation.

[Adrienne Walker]

Created attachment 77560 [details]
Patch

[Kenneth Russell]

Comment on attachment 77560 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=77560&action=review

> WebCore/platform/graphics/chromium/LayerTilerChromium.cpp:402
> +    IntSize newSize(std::max(m_layerSize.width(), layerRect.right()),
> +                    std::max(m_layerSize.height(), layerRect.bottom()));

Could you use a preexisting method on IntSize to compute this? For example IntSize::expandedTo()?

[Adrienne Walker]

Created attachment 77568 [details]
Patch

[Adrienne Walker]

(In reply to comment #2)
> (From update of attachment 77560 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=77560&action=review
> 
> > WebCore/platform/graphics/chromium/LayerTilerChromium.cpp:402
> > +    IntSize newSize(std::max(m_layerSize.width(), layerRect.right()),
> > +                    std::max(m_layerSize.height(), layerRect.bottom()));
> 
> Could you use a preexisting method on IntSize to compute this? For example IntSize::expandedTo()?

Good point.

[Kenneth Russell]

Comment on attachment 77568 [details]
Patch

Looks good.

[WebKit Commit Bot]

Comment on attachment 77568 [details]
Patch

Clearing flags on attachment: 77568

Committed r74722: <http://trac.webkit.org/changeset/74722>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Abhishek Arya]

Created attachment 103405 [details]
Ignore, wrong upload

[Abhishek Arya]

sorry, wrong upload, please ignore.