I hit bloom filter subtraction assertions on a clean stylo build with the commits from [1] applied, browsing trivial testcases. Presumably this is because we don't fix up the bloom filter in Gecko. Do we need a bottom-up traversal now? It's kind of unfortunate to add that, but we can certainly do it if necessary. We'll need to implement store_children_to_process and did_process_child, probably by just moving the atomic into ElementData. This will add another blocker to hoisting the computed style into the frame and dropping the ElementData, but I'm ok with that for now. This will presumably block any incubator updates, so is pretty high-priority. Emilio, do you have time to look at this? [1] https://github.com/servo/servo/pull/14353

I guess I can take a look. Without having dug into it, I bet this is because in Gecko `children` doesn't imply `depth += 1`, which is kind of sad. Post-traversal shouldn't be relevant. Leaving ni? open.

This is a one-line fix in https://github.com/servo/servo/pull/14439.

Emilio fixed this in https://github.com/servo/servo/pull/14439 Thanks Emilio!

Pushed by bobowencode@gmail.com: https://hg.mozilla.org/integration/mozilla-inbound/rev/a608c5cc4ff8 Change USER_NON_ADMIN access token level from whitelist to blacklist containing Admin SIDs. r=jimm

Backout by cbook@mozilla.com: https://hg.mozilla.org/integration/mozilla-inbound/rev/c7e9ab08a73a Backed out changeset a608c5cc4ff8 for landing with wrong bug number