We don't need to wait for the pref to be flipped by default; now that we're building with vertical mode enabled in the code, we can just set the pref from within the reftest manifest, and begin running these tests regularly.

Try run with this patch: https://treeherder.mozilla.org/#/jobs?repo=try&revision=a51d32b80d2b. I'm expecting 1131013-vertical-bidi.html to fail here, as it needs the followup patch (bug 1131013 comment 7) to work across all platforms.

Backed out in https://hg.mozilla.org/integration/mozilla-inbound/rev/338cb106bfc4 - you'll be wanting to spend rather a lot of time on try with debug builds, since at the very least ua-style-sheet-input-number-1.html has between zero and sixteen "Shouldn't be incomplete if availableBSize is UNCONSTRAINED" assertion failures.

Oh, drat. That's thanks to bug 1123299, which is also why that test was already marked as known-failing. I've disabled the offending part of the test for the time being (as we've also done in some other comparable cases), and will add a comment in bug 1123299 to remind us to restore it once it's implemented. Re-landed without that problem, I trust: https://hg.mozilla.org/integration/mozilla-inbound/rev/e37d57033b13

sorry had to back this out for breaking reftests like https://treeherder.mozilla.org/logviewer.html#?job_id=6874976&repo=mozilla-inbound

Ugh - sorry about that. :( Filed bug 1135622 for the unexpected failure, and restored the failure annotation that we had in the original landing (the one that was backed out for assertions). Tryserver says we should be ok now: https://treeherder.mozilla.org/#/jobs?repo=try&revision=fa27c631c2e7 (the failures there are a new test that's not actually present in the tree, or in this landing). So here we go again: https://hg.mozilla.org/integration/mozilla-inbound/rev/45d4cc092de1

Backed out for WinXP reftest leaks. https://hg.mozilla.org/integration/mozilla-inbound/rev/b8f973e91242 https://treeherder.mozilla.org/logviewer.html#?job_id=6885298&repo=mozilla-inbound