A multi-column element should establish a new block formatting context. At the moment, it doesn't and may cause issues like bug 616722. There's a test over at IE Test Center, see http://samples.msdn.microsoft.com/ietestcenter/css3/css_harness.htm?type=multicolumn&url=column-block-formatting-context-001 that is passed by IE10, but not by WebKit.

(In reply to comment #0) > ... but not by WebKit. Fwiw, that testcase displays correctly in Chrome / Linux - 12.0.742.112 and Webkit nightlies ( r91108), Chrome dev channel on OS X 10.6.8

(In reply to comment #1) > Fwiw, that testcase displays correctly in Chrome / Linux - 12.0.742.112 and > Webkit nightlies ( r91108), Chrome dev channel on OS X 10.6.8 Oh, I didn't notice. However, if you look at the second testcase of bug 616722, you'll see that Webkit (Chrome 12 tested) doesn't establish a BFC if column-count is only 1, while IE does (the spec says a multi-column container is established if column-count is not auto). I can see the advantage of WebKit's behaviour, but it's not per spec.

(In reply to comment #2) The latest WebKit nightly (r91611) now behaves per spec and inhibits margin-collapse through (mostly, column-count:0 seems to inhibit margin-collapse through, which is wrong I think) webkit bug: https://bugs.webkit.org/show_bug.cgi?id=64975

(In reply to comment #3) > The latest WebKit nightly (r91611) now behaves per spec and inhibits > margin-collapse through (mostly, column-count:0 seems to inhibit > margin-collapse through, which is wrong I think) Yes, because 0 is not a valid column-count value. IE, Opera and Fx don't accept 0 (see their respective dev tools). Though, we should not discuss other browsers issues in this bug; I actually reported this issue to have a report to refer to in bug 477462.

URL field set to http://samples.msdn.microsoft.com/ietestcenter/css3/multicolumn/column-block-formatting-context-001.htm

Chrome 28.0.1500.63 and Opera 12.15 pass the test. Adding [parity-Chrome] [parity-Opera] to Whiteboard

Daniel, I get expected results when trying the URL test with Firefox 43 buildID=20151210085006 and Firefox 46.0a1 buildID=20151220080149 . Resolving as WORKSFORME