For posterity, I *think* the bug comes from this block of code, http://mxr.mozilla.org/mozilla-central/source/layout/style/nsCSSDataBlock.cpp#151 Although Gijs has pointed out that this would mean that the background-color would take the high-contrast color but the text color wouldn't, so I'm not so sure.

No colors ever work in HCM. We just end up picking nice ones for foreground/background, but we seem to pick the background color for outline, instead of the foreground. I don't know why this is.

Comment on attachment 8598769 [details] [diff] [review] Fix the initial value of outline-color to actually be correct I wonder how we missed this for so long. (It seems to predate the rule tree landing.)

Backed out in https://hg.mozilla.org/integration/mozilla-inbound/rev/b899a9295076 for mochitest-5 orange: https://treeherder.mozilla.org/logviewer.html#?job_id=9382841&repo=mozilla-inbound

Yeah, looks like the big switch statement in: $("one").addEventListener("transitionend", function(event) { switch (event.propertyName) { needs a case for outline-color just like the ones for everything other than border-right-color.

> Yeah, looks like the big switch statement in: That, plus adding it to the block with explicit styles passes try.

One thing I noticed while auditing -- the patch in bug 1141607 should have added a started_test() call -- as, probably, should some other changes to the test in the past. I think we're two short for the listener on #one, and also two short for the listener on the root. But we're two over for #seven since that listener doesn't call finished_test (and also doesn't check for duplicate events on ::after). It might be worth seeing if adding the necessary calls makes failures that are currently intermittent more reliable.

Er, I meant that for bug 1156451, although it's not entirely inappropriate here.