Bug 80348 – [chromium] Remove texture readback from WebViewImpl::paint

• Component: New Bugs
• Status: RESOLVED WONTFIX

[Alexandre Elias]

[chromium] Remove texture readback from WebViewImpl::paint

[Alexandre Elias]

Created attachment 130248 [details]
Patch

[WebKit Review Bot]

Please wait for approval from [e-mail], [e-mail] or [e-mail] before submitting because this patch contains changes to the Chromium platform API.

[James Robinson]

This will produce odd-looking thumbnails for pages like poster circle. Can we do that?

[Alexandre Elias]

I understand the thumbnailer will be moving to capturing the browser-side texture/bitmap on most platforms anyway, addressing that issue in the long run.  I guess we could hold off on submitting this until that transition is complete everywhere, although my inclination would be to go ahead and delete it now.

The other uses are corner cases like Chrome for Android's link disambiguation popup feature, and it's probably not worth maintaining this codepath for a corner-case-on-a-corner-case.

[W. James MacLean]

(In reply to comment #4)
> I understand the thumbnailer will be moving to capturing the browser-side texture/bitmap on most platforms anyway, addressing that issue in the long run.  I guess we could hold off on submitting this until that transition is complete everywhere, although my inclination would be to go ahead and delete it now.
> 
> The other uses are corner cases like Chrome for Android's link disambiguation popup feature, and it's probably not worth maintaining this codepath for a corner-case-on-a-corner-case.

What's the rush to delete it?

When the browser-texture patch is possible then it seems much simpler, but until then I think we need to provide some reasonable looking form of thumbnail.

[Alexandre Elias]

I trust the software path to give reasonable thumbnails more often than the hardware readback one in any case.  In Chrome for Android, we ran into many bugs where we'd get an entirely black thumbnail returned.  They went away when we switched to browser-side thumbnailing.  I'd wager the other platforms have similar bugs, but they're just hidden because compositing is usually not enabled.

[WebKit Review Bot]

Comment on attachment 130248 [details]
Patch

Attachment 130248 [details] did not pass chromium-ews (chromium-xvfb):
Output: http://queues.webkit.org/results/11837103

New failing tests:
compositing/geometry/fixed-in-composited.html
compositing/geometry/vertical-scroll-composited.html
compositing/geometry/layer-due-to-layer-children.html
compositing/direct-image-compositing.html
compositing/geometry/fixed-position-transform-composited-page-scale-down.html
compositing/geometry/layer-due-to-layer-children-deep.html
compositing/geometry/fixed-position-transform-composited-page-scale.html
compositing/geometry/fixed-position.html
animations/opacity-transform-animation.html
accessibility/aria-disabled.html
compositing/backface-visibility.html
compositing/geometry/fixed-position-iframe-composited-page-scale.html
compositing/geometry/tall-page-composited.html
compositing/culling/filter-occlusion-alpha-large.html
animations/missing-values-last-keyframe.html
compositing/flat-with-transformed-child.html
compositing/geometry/horizontal-scroll-composited.html
compositing/images/direct-image-background-color.html
compositing/geometry/ancestor-overflow-change.html
compositing/scrollbar-painting.html
animations/missing-values-first-keyframe.html
compositing/backface-visibility-hierarchical-transform.html
compositing/geometry/clipping-foreground.html
compositing/culling/filter-occlusion-alpha.html
compositing/geometry/fixed-position-iframe-composited-page-scale-down.html
compositing/masks/direct-image-mask.html
compositing/culling/filter-occlusion-blur-large.html
compositing/culling/filter-occlusion-blur.html
compositing/color-matching/image-color-matching.html
compositing/video-page-visibility.html

[WebKit Review Bot]

Created attachment 130259 [details]
Archive of layout-test-results from ec2-cr-linux-02

The attached test failures were seen while running run-webkit-tests on the chromium-ews.
Bot: ec2-cr-linux-02  Port: <class 'webkitpy.common.config.ports.ChromiumXVFBPort'>  Platform: Linux-2.6.35-28-virtual-x86_64-with-Ubuntu-10.10-maverick

[James Robinson]

Oh, and DumpRenderTree uses this path for all of the compositing layout tests.  A bunch of browser_tests do the same thing.  I forgot about those :)

I would really like to get rid of this codepath ASAPly, it's a pain in the butt to maintain.

[James Robinson]

Comment on attachment 130248 [details]
Patch

I don't think we can do this yet.

[Stephen Chenney]

https://code.google.com/p/chromium/issues/detail?id=231293