When running native android fennec with the patch on bug 607417, and with MOZ_FIXED_POSITION_LAYERS defined, fixed position layers draw areas of the tile that have not been rendered into (and should not be visible).
This can be seen on http://bradfrostweb.com/demo/fixed/index.html
Turning off tiled layer drawing fixes the problem. Turning off retained tiles makes no difference to the behaviour.
Most likely because we're not respecting the region, should be an easy fix. I'll look at this soon.
Taking the bug. This should be at least a release blocker.
Created attachment 619153 [details] [diff] [review]
Created attachment 619162 [details] [diff] [review]
Comment on attachment 619162 [details] [diff] [review]
Review of attachment 619162 [details] [diff] [review]:
This looks correct to me.
Created attachment 619196 [details] [diff] [review]
I r+'d the original patch, but it doesn't apply to current central - here it is, rebased (non-trivial, so this should be reviewed if it's used). Works, but I've only done extremely limited testing.
Rebase looks good to me. Landing.
Comment on attachment 619196 [details] [diff] [review]
[Approval Request Comment]
Regression caused by (bug #): Bug 739679 - Add a Shadowable TiledThebesLayer implementation
User impact if declined: Visual glitches for complex layers (non rectangular).
Testing completed (on m-c, etc.): Tested locally to fix a few test cases. I'm looking to allow a few days of bake time on m-c.
Risk to taking this patch (and alternatives if risky): Low, changed the draw region to reflect accurately the input value.
String changes made by this patch: None
Recommending a few days of bake time on m-c to be certain this fixes the problem to make sure we only need to uplift on patch for this problem.
*** Bug 749793 has been marked as a duplicate of this bug. ***
Is there a link to the inbound changeset?
Verified fixed on brad's aurora try build.