[chromium] Enable fixed position compositing on Android DRT
Created attachment 170419 [details] Patch
Having some trouble seeing the effects of this locally, so let's run it through the bots. I did check with debug logs that the appropriate flags are getting set.
Comment on attachment 170419 [details] Patch Attachment 170419 [details] did not pass chromium-ews (chromium-xvfb): Output: http://queues.webkit.org/results/14563296 New failing tests: compositing/geometry/fixed-position-transform-composited-page-scale-down.html
This is likely a patch for jamesr to review.
Looks like we don't run Android layout tests on the bots yet. Running locally, I see 3 types of failures: - Red/blue color channels are swapped. - Repaint tests don't work with the compositor. - Some compositor tiles have have different low-order bits, most likely because of inaccuracies in linear interpolation. I think this is expected because GL doesn't guarantee bit-exactness. I'll try to fix the red/blue issue first.
Created attachment 170611 [details] Patch Fix red/blue color channel issue. I also checked the tile texel filtering problem on three different devices and got three different results, so I think we'll have to live with it. Same goes for repaint tests, because all of them need to be rewritten to use the new mechanism from bug 97801.
Comment on attachment 170611 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=170611&action=review Cool, thanks Sami! Do we know how many tests are failing after the color adjustments? > Tools/DumpRenderTree/chromium/TestShell.cpp:685 > + // in software rendering mode but BGRA if accelerated compositing is active. Isn't Alexandre working on software rendering (albeit through the compositor)?
Comment on attachment 170611 [details] Patch Attachment 170611 [details] did not pass chromium-ews (chromium-xvfb): Output: http://queues.webkit.org/results/14566587 New failing tests: compositing/geometry/fixed-position-transform-composited-page-scale-down.html
(In reply to comment #7) > (From update of attachment 170611 [details]) > View in context: https://bugs.webkit.org/attachment.cgi?id=170611&action=review > > Cool, thanks Sami! Do we know how many tests are failing after the color adjustments? I'm still running the tests locally, but from the failures so far I'd estimate about 7100 :) > > Tools/DumpRenderTree/chromium/TestShell.cpp:685 > > + // in software rendering mode but BGRA if accelerated compositing is active. > > Isn't Alexandre working on software rendering (albeit through the compositor)? Yes he is, but at least initially that's not going to be the mode used by Chrome on Android. And as you said, it's still going through the compositor, so we'd need some new mechanism to detect that mode. Thinking about this a little more, the root issue here is that the compositor readback gives back pixels with the wrong component order. Let me take another pass to figure out why that is happening. Also, I'll see what's up with that failing test on cr-linux.
Comment on attachment 170611 [details] Patch Attachment 170611 [details] did not pass chromium-ews (chromium-xvfb): Output: http://queues.webkit.org/results/14544604 New failing tests: compositing/geometry/fixed-position-transform-composited-page-scale-down.html
Created attachment 170653 [details] Patch Fixing R/B ordering issue separately in bug 98647. compositing/geometry/fixed-position-transform-composited-page-scale-down.html fails with and without this patch.
Comment on attachment 170653 [details] Patch Attachment 170653 [details] did not pass chromium-ews (chromium-xvfb): Output: http://queues.webkit.org/results/14566692 New failing tests: compositing/geometry/fixed-position-transform-composited-page-scale-down.html
Anyone care to review?
Comment on attachment 170653 [details] Patch Sure
Comment on attachment 170653 [details] Patch The failing test should be unrelated, the CQ should verify that. Setting CQ+ on request.
Comment on attachment 170653 [details] Patch Rejecting attachment 170653 [details] from commit-queue. New failing tests: compositing/geometry/fixed-position-transform-composited-page-scale-down.html Full output: http://queues.webkit.org/results/14623401
Comment on attachment 170653 [details] Patch Trying again.
Comment on attachment 170653 [details] Patch Clearing flags on attachment: 170653 Committed r133607: <http://trac.webkit.org/changeset/133607>
All reviewed patches have been landed. Closing bug.