https://treeherder.mozilla.org/#/jobs?repo=pine&filter-tier=1&filter-tier=2&filter-tier=3&selectedJob=26776 and the reftest analyzer: http://hg.mozilla.org/mozilla-central/raw-file/tip/layout/tools/reftest/reftest-analyzer.xhtml#logurl=https://queue.taskcluster.net/v1/task/Mqff13WTQpikI61u7IRPag/runs/0/artifacts/public/logs/live_backing.log&only_show_unexpected=1 https://treeherder.mozilla.org/#/jobs?repo=pine&filter-tier=1&filter-tier=2&filter-tier=3&selectedJob=26791 and the reftest analyzer: http://hg.mozilla.org/mozilla-central/raw-file/tip/layout/tools/reftest/reftest-analyzer.xhtml#logurl=https://queue.taskcluster.net/v1/task/GaBEY-wfSAOwOcf7qgQCzg/runs/0/artifacts/public/logs/live_backing.log&only_show_unexpected=1 At a first glance, it looks like this only involves the "display: contents" instances that also make use of "blue" color ?!

William, do you have any idea?

The reftest e10s s2 and s3 errors on opt and debug are the same as those.

ok, it seems to relates to shadow dom + display: contents more than the color. Reverting the only pine-specific patch we have regarding ShadowRoot does not help.

William, I am not able to see any difference between central and pine touching either the related prefs, or ShadowDOM or CSS :/ Can you help?

Yeah, it's really strange that these tests are failing if the Shadow DOM platform code hasn't been changed. I'll look into it.

It looks like the failures is caused by this patch: http://hg.mozilla.org/projects/pine/rev/e417ce4b94e4 That patch traverses across the shadow root when getting the parent node that is added to the undisplayed map. The problem is that we don't do the same when we do a lookup in some other code so we end up using the wrong key in the map lookup. This patch should fix the test failures you are seeing.

That's surprising, this is a patch I tested reverting, and I was still having the issue :/

Ok, I got lurred by antialiasing :/

(In reply to Alexandre LISSY :gerard-majax from comment #9) > https://treeherder.mozilla.org/#/jobs?repo=try&revision=3ffdc596fbc4 Looks like I pushed the wrong branch :/ https://treeherder.mozilla.org/#/jobs?repo=try&revision=4bb395d797db

Thanks William, all reftests are green!

Comment on attachment 8757117 [details] [diff] [review] Use GetParentCrossingShadowRoot() instead of GetParent() in GetStyleContextInMap when looking up undisplayed nodes You should probably also add the MOZ_ASSERT(parent || !aContent->GetParent(), "no non-elements"); that Kyle's patch has. And this should land at the same time as bug 1261351. r=dbaron

Pushed by wchen@mozilla.com: https://hg.mozilla.org/integration/mozilla-inbound/rev/daecf88f0427 Use GetParentCrossingShadowRoot() instead of GetParent() in GetStyleContextInMap when looking up undisplayed nodes. r=dbaron