In gecko, the attached test case will read "東 京" in kanji with ruby on top. The "東" and "京" will be at the 1/3 and 2/3 mark of the page. In stylo, the "東" and "京" will be at opposite ends of the page, but the ruby will be in the same spot. I thought this might be due to us not handling mTextAlignLastUnsafe, but that's not getting set on this page. This causes ./layout/reftests/writing-mode/1135361-ruby-justify-1.html

I also verified that the getComputedStyle() trees are identical (which only points more strongly to something weird going on with the disabled unsafe/true value support), but that's not the case either.

So this is because the SHOULD_SUPPRESS_LINEBREAK flag is not correctly set to the text frame in this case, so it breaks the shortcut check at https://searchfox.org/mozilla-central/rev/e5b13e6224dbe3182050cf442608c4cb6a8c5c55/layout/generic/nsLineLayout.cpp#3133,3142 So in this case, inheriting flags doesn't work, because the text frame inherits from <ruby>, which doesn't have the flag we want. We need to do an additional check in StyleAdjuster::adjust_for_text() to apply this flag when its parent has any ruby display type.

Servo PR: https://github.com/servo/servo/pull/18075

Pushed by xquan@mozilla.com: https://hg.mozilla.org/integration/autoland/rev/aeef23d80cf2 Update reftest expectation.