Once this patch lands, all content drawn by WebRender is drawn into
a picture cache surface.

This will incur some extra GPU memory overhead since there are extra
GPU texture buffers. Much of this can be reduced by adding a couple
of simple optimizations in future to detect tiles that are solid
colors only.

With this change, we'll now be able to provide exact dirty rects for
the entire screen without any hacks, and start the work to draw into
OS compositor surfaces directly.

Backed out changeset 4f126cfd8012 (bug 1584439) for causing wrench bustage

failure: https://treeherder.mozilla.org/#/jobs?repo=autoland&selectedJob=269194066&resultStatus=testfailed%2Cbusted%2Cexception%2Cusercancel%2Crunnable&revision=4f126cfd8012ff538ea8e414bfc3cc0c4e852e11

backout: https://hg.mozilla.org/integration/autoland/rev/af4f4ed1a70ce59288c24131244f6a420fb86b99

I've updated the fuzziness annotations for those tests on pixel2, thanks!

Backed out changeset 3eba60b905c0 (bug 1584439) for causing wrench bustages CLOSED TREE

Backout revision https://hg.mozilla.org/integration/autoland/rev/d958fe844b16815800d7c5a97b05f1849a42f7e8

Failed push https://treeherder.mozilla.org/#/jobs?repo=autoland&selectedJob=269198042&searchStr=wrench&revision=3eba60b905c06bdb6b39367ac941326a8499c5ab

Failuire logs https://treeherder.mozilla.org/logviewer.html#?job_id=269198042&repo=autoland

Glen can you please take a look?

Hmm, there doesn't seem to be anything obviously wrong in the log files - I wonder if it's crashing but not including any crash information in the log output. Can you see anything in the logs that looks like an application error or crash?

Otherwise, I'll try to add some extra logging information and debug on try and/or a local android build tomorrow.

It does look like it stops running before all the reftests complete, since the output finishes earlier than I'd expect. But then it doesn't seem to include any crash details, and just seems to continue on with the script.

I ran this locally on a pixel2 device.

I found that there was one more test that was failing due to fuzziness. The output is being truncated on the try server for some reason, which is why I couldn't see a failure.

I've updated the fuzziness, and confirmed that they all pass on my pixel2 device. I've kicked off another try run [1] to be safe, then this should hopefully be ready to land again.

[1] https://treeherder.mozilla.org/#/jobs?repo=try&revision=fbe2ecef8e9d808d585e06c35e5370d15e52a7da

https://hg.mozilla.org/mozilla-central/rev/9b42bdc4889f

Seems to have caused bug 1585898.