+++ This bug was initially created as a clone of Bug #1370430 +++ I'm filing this as a placeholder bug for the next webrender update. I may be running a cron script [1] that does try pushes with webrender update attempts, so that we can track build/test breakages introduced by webrender on a rolling basis. This bug will hold the try push links as well as dependencies filed for those breakages, so that we have a better idea going into the update of what needs fixing. I might abort the cron job because once things get too far out of sync it's hard to fully automate fixing all the breakages. When we are ready to actually land the update, we can rename this bug and use it for the update, and then file a new bug for the next "future update". [1] https://github.com/staktrace/moz-scripts/blob/master/try-latest-webrender.sh

WR @ 6752684fcc7402b0a5480e0b9f73152b2f9ed1e5 https://treeherder.mozilla.org/#/jobs?repo=try&revision=4c9c5828545ee09756afd2fc760c8ae548736808 https://treeherder.mozilla.org/#/jobs?repo=try&revision=7079d8e926f29e22397c9d6a067ee5c7533cc9bf Linux is green, windows still pending.

WR @ 0b0ee6357d2a096370b8c1b66994c9b2a5edff67 https://treeherder.mozilla.org/#/jobs?repo=try&revision=5ecac6c053f14fc9c770a9059195dea0dd3b7a28 https://treeherder.mozilla.org/#/jobs?repo=try&revision=f2ff0ce03a507f6401da96a4c1027b46f8f305b4 Linux is showing a lot of crashes. Regression range: * 0b0ee63 Auto merge of #1362 - glennw:gpu-cache-resource-rects-2, r=kvark * 0f3df68 Remove resolve pass, port resource rects to GPU cache. * cc7a041 Add a reftest for border images. Regression is from servo/webrender#1362, and the crash stack seems to be related.

WR @ 0b0ee6357d2a096370b8c1b66994c9b2a5edff67 https://treeherder.mozilla.org/#/jobs?repo=try&revision=6b774b9d89edb905c5101b6535cee3ee4e9fa67b https://treeherder.mozilla.org/#/jobs?repo=try&revision=75aec87fa737ac9a0e5c2aa3668c309eb6c29c63 Same as before

servo/webrender#1383 is supposed to fix the crashes from comment 2. Also of note is that bholley cherry-picked servo/webrender#1382 into m-c (well it's on autoland right now) in order to avoid build bustage from two incompatible rayon versions coexisting in m-c. Both of these PRs are included in this next try push: WR @ 568c727b7375961d08f0d979bd5e78b88ecde72f https://treeherder.mozilla.org/#/jobs?repo=try&revision=2c0b747baaabf0224d7a08384be678b448620b73 https://treeherder.mozilla.org/#/jobs?repo=try&revision=da52e565573419071862efd351be9b2c98f4fb37

That one had bustage from the push_built_display_list API change. Now with bustage fix: https://treeherder.mozilla.org/#/jobs?repo=try&revision=83e0a2fd450d56fe5b47955378d4bb9ba6cae76e https://treeherder.mozilla.org/#/jobs?repo=try&revision=a43878c3f0ba350071c522a96643559c0e733cdb Green again. I'll want to land this relatively soon so that we don't have an extended period of time where we have a cherry-picked webrender commit sitting in m-c.

One more try push to verify the rebase onto latest master: https://treeherder.mozilla.org/#/jobs?repo=try&revision=861d73aa1fbe2b0be958b2d16cebb88954aec2e7

The try push in comment 8 is showing two reftests failing. Known good: m-c cset 035c25bef7b5 + WR cset 568c727b7375961d08f0d979bd5e78b88ecde72f [1] Failing: m-c cset 75be6742abb9 + WR cset 568c727b7375961d08f0d979bd5e78b88ecde72f [2] Also from current m-c tip: Known good: m-c cset 75be6742abb9 + WR cset 6752684fcc7402b0a5480e0b9f73152b2f9ed1e5 [3] So the m-c pushlog is [4] and the WR range is [5]. The most suspicious thing in the m-c pushlog is my own patches from bug 1372603. I'll see if I can verify that. [1] https://treeherder.mozilla.org/#/jobs?repo=try&revision=83e0a2fd450d56fe5b47955378d4bb9ba6cae76e [2] https://treeherder.mozilla.org/#/jobs?repo=try&revision=861d73aa1fbe2b0be958b2d16cebb88954aec2e7 [3] https://treeherder.mozilla.org/#/jobs?repo=mozilla-central&revision=75be6742abb94d79f3bcb731ab7fafa3d42ac4da&filter-searchStr=linux64-qr&group_state=expanded [4] https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=035c25bef7b5&tochange=75be6742abb9 [5] * 568c727 Auto merge of #1383 - glennw:fix-blob-assert, r=kvark |\ | * 62a3c6b Fix a bug with UV rects for blob images in the GPU cache. * | 7f7cd45 Auto merge of #1380 - mrobinson:push-nested-display-list, r=glennw |\ \ | * | 4eea6ae Make push_prebuilt_display_list a memcpy | |/ * | 181759d Auto merge of #1382 - bholley:bump_rayon_0_8, r=glennw |\ \ | |/ |/| | * e392895 Bump rayon to 0.8. |/ * 0b0ee63 Auto merge of #1362 - glennw:gpu-cache-resource-rects-2, r=kvark * 0f3df68 Remove resolve pass, port resource rects to GPU cache. * cc7a041 Add a reftest for border images.

Try pushes just before [1] and just after [2] my patches. [1] https://treeherder.mozilla.org/#/jobs?repo=try&revision=dc48853649a206480b0f4576ccb007e1ad48acc2 [2] https://treeherder.mozilla.org/#/jobs?repo=try&revision=506eb8b6108a2a0f366add2f35c9cd733da859df

Comment on attachment 8878193 [details] Bug 1372608 - Update call to push_built_display_list due to API change in WR cset 4eea6ae. https://reviewboard.mozilla.org/r/149558/#review154360

That indicates that it was in fact my patches that caused the reftest failures on the m-c side. Not sure which PR on the webrender side is at fault yet. Bisections: https://treeherder.mozilla.org/#/jobs?repo=try&revision=59a44e6c0bdcc47bd0c24bd886272e0242bf6428 https://treeherder.mozilla.org/#/jobs?repo=try&revision=5c2fba195c5cb89a873732b205bced1b464b7bf3

So that confirms the reftest failure is a combination of bug 1372603 and WR cset 4eea6ae (the push-nested-display-list change).

I investigated this further. It looks like there's a scenario where if the nested display list is wrapped by a stacking context with a transform, but without a clip in between the stacking context and the nested display list, then the stacking context's transform gets ignored. This can be fixed either by restoring the clip I took out of WebRenderDisplayItemLayer (in [1]), or by applying the one-line change at [2]. Seeing as [2] is the more correct/desirable fix, and is already in a PR that will land in upstream WR soon, I think I'll just wait for that to get merged and then pick up a WR that includes that change. [1] https://hg.mozilla.org/integration/autoland/rev/7ece58ca5ca3 [2] https://github.com/servo/webrender/pull/1386/files#diff-054157e919dfed0477381da46476b556L638

Comment on attachment 8878192 [details] Bug 1372608 - Update webrender to cset 568c727b7375961d08f0d979bd5e78b88ecde72f. Dropping this review flag for now.

b0d2024b5743cce762e2392ab8e42a27423114cd is the commit that includes servo/webrender#1386 so we should update to that. I'm working on fixing the build up.

Forward-duping to bug 1373381 (the next WR update bug) since there's a bunch of work on that bug that's also relevant to the update that will land.