Per the spec on layout containment [1] all overflow is supposed to be treated as ink (visual) overflow. This implementation works as of bug 1476495, but fails when a layout-contained element has scrollable overflow produced by floated children. See attached test/reference case.

[1] https://drafts.csswg.org/css-contain/#containment-layout " 3. If the contents of the element overflow the element, they must be treated as ink overflow. "

This seems like the sort of thing that's relatively straightforward to debug with rr, by starting from the bad overflow area and working backwards to how it ended up that way.

Actually, I think the behavior here is correct. I'm assuming that what you thought was the bug was that the second box has a scrollbar. However, the child of that box, <div class="inner-sm contain border"> has contain:layout, and *it* has a large floating child. Because of: > 1. The element establishes an independent formatting context. the size of the div with class inner-sm will expand to contain its floating descendants. Then *it* (not its contents) contributes to the scrollable overflow of its parent. In order to make the test correct, I think there is a need to add something like: .inner-sm { height: 50px; width: 50px; } (which was, maybe, from the name and from the reference, intended?). I still don't know what the first box in the reference was intended to match, though.

Yeah, it looks like we match Chrome on the attached testcase, and it's not obvious to me that the float is directly creating any scrollable overflow here (beyond just propping up the height of its contain:layout parent, which *is* allowed to create scrollable overflow). So maybe this is working as-intended. > I still don't know what the first box in the reference was intended to match, though. I think that was left in by mistake - I think the testcase/reference here are shorter versions of other similar testcases, and Morgan just left in a bit in the reference case that she'd removed in the testcase.

[Setting ni=me to land a modified version of the attached patch as a reftest. Once we've got a test for layout-containment on the parent of a float, for thoroughness/coverage, then I think we can close this out, probably as INVALID.]

This test is basically a copy of its -001 variant, with some "float:left" sprinkled around on contained descendants. Before this patch, this test had an additional arbitrary sizing difference as compared to the -001 version -- there's one element that arbitrarily has class "outer" in the -002 test whereas it has class "inner-lg" in the -001 version. These classes have different sizing characteristics, which makes a difference to whether scrollbars show up, because this element is not contained (though it is a layout container itself). This patch undoes this arbitrary difference and also adds a "float" class to make it easier to see which elements we've sprinkled float styling onto. Depends on D3825

Comment on attachment 9002576 [details] Bug 1481951 part 1: Adjust contain-layout-overflow-* reftests to remove unused rule for nonexistent class "inner-md". r=dbaron David Baron :dbaron: 🏴󠁵󠁳󠁣󠁡󠁿 ⌚UTC-7 has approved the revision.

Comment on attachment 9002577 [details] Bug 1481951 part 2: Adjust reftest contain-layout-overflow-002.html to be clearer & have accurate expectations. r=dbaron David Baron :dbaron: 🏴󠁵󠁳󠁣󠁡󠁿 ⌚UTC-7 has approved the revision.

Try run: https://treeherder.mozilla.org/#/jobs?repo=try&revision=046950ff0ee41b8e73450f6e257cd9c1ace54311

Pushed by dholbert@mozilla.com: https://hg.mozilla.org/integration/mozilla-inbound/rev/e47a225b907e part 1: Adjust contain-layout-overflow-* reftests to remove unused rule for nonexistent class "inner-md". r=dbaron https://hg.mozilla.org/integration/mozilla-inbound/rev/6c3db80981da part 2: Adjust reftest contain-layout-overflow-002.html to be clearer & have accurate expectations. r=dbaron

https://hg.mozilla.org/mozilla-central/rev/e47a225b907e https://hg.mozilla.org/mozilla-central/rev/6c3db80981da

web-platform-tests import at https://github.com/web-platform-tests/wpt/pull/12596

Marking this as qe-verify -, per comment 16.

Pushed by dholbert@mozilla.com: https://hg.mozilla.org/integration/autoland/rev/8f159b8c631d late-breaking followup: Remove mistaken comment with a reference to this (closed) bug. r=TYLin