I found this while looking into web platform tests that stylo passes but gecko fails in bug 1389187. This causes `/css/CSS2/ui/outline-width-096.xht` failure on non-stylo build. Currently gecko does a fixup to computed value of outline-width[1] and stores that value inside `mActualOutlineWidth` field instead of `mOutlineWidth` in `nsStyleOutline`. But inside nsRuleNode.cpp, we are still getting mOutlineWidth of parent[2] and that's causing to get the wrong inherited value. [1] http://searchfox.org/mozilla-central/rev/e5b13e6224dbe3182050cf442608c4cb6a8c5c55/layout/style/nsStyleStruct.cpp#569-578 [2] http://searchfox.org/mozilla-central/rev/6482c8a5fa5c7446e82ef187d1a1faff49e3379e/layout/style/nsRuleNode.cpp#7943

Comment on attachment 8897320 [details] Bug 1390432 - Get the right parent outline-width value during computation Patch looks trivial, but I don't know the cascade logic in the old style system, so deferring to dbaron.

Comment on attachment 8897320 [details] Bug 1390432 - Get the right parent outline-width value during computation https://reviewboard.mozilla.org/r/168624/#review174124 So there are two differences between mOutlineWidth and mActualOutlineWidth (see nsStyleOutline::RecalcData): 1. the fixup for outline-style:none 2. the rounding to device pixels (1) should be part of the computed value, but (2) should not. bug 598080 already covers this problem. See my review comments there, e.g., bug 598080 comment 14.