[css-multicol] definition of `column-span` should say what happens without an ancestor multicol

[dbaron]

The definition of column-span: all should say what happens if there is no "nearest multicol ancestor in the same block formatting context". This is relatively straightforward; the column-span property doesn't do anything in that case. However, the spec should say so.

/cc @neerjapancholi

[mstensho]

Also probably worth noting that invalid spanners don't force creation of new block formatting contexts and things like that, so that e.g. the following passes:

<!-- Not inside a multicol container. -->
<p>No red should be seen.</p>
<div style="float:left; width:20px; height:20px; background:green;"></div>
<div style="column-span:all; width:20px; height:20px; background:red;"></div>

[tantek]

Agenda+ing just to make sure the WG is ok with this proposal. Figuring we're going to talk a lot of multicol this week.

[astearns]

Removing agenda+ until #1072 is resolved

[dbaron]

This isn't particularly related to #1072, and I think it's trivial enough that it can be resolved at editor discretion if the existing implementations do the obvious thing.

[neerjapancholi]

Tested the case suggested by @mstensho on different browsers and posting results with screenshots:

Test case:
https://jsfiddle.net/Ld86cuwk/

Chrome 56.0.2924.87 (64-bit)	Safari 10.0.3 (11602.4.8.0.1)	Edge 38.14393.0.0
Does not create BFC	Creates BFC	Creates BFC

Screenshot of Chrome:

Screenshot of Safari (same for Edge):

[frivoal]

Firefox agrees with Chrome.
Ignore me, that's just because FF does not implement column-span yet.

[frivoal]

it's trivial [...] if the existing implementations do the obvious thing

Unfortunately, not all do. So we do need to talk…

[mstensho]

So it looks like Chrome is the odd one here, right? In a way I find it strange that a property meant to be used inside multicol containers only has any effect on elements not inside such a container, but the spec does say that column-span applies to "in-flow block-level elements". The spec needs to become clearer. I have no strong opinions when it comes to the exact resolution, other than the fact that it should either be:

A: Chrome has a bug - fix the spec to say that column-span:all outside of multicol still establishes a new block formatting context (on that note: issue #1071 points out that "block formatting context" isn't good wording)

B: Beef up the spec and have the other browsers align with Chrome's behavior

[fantasai]

The case to think about here is, what if the designer had a two-column layout and then switched to a one-column layout. What's least likely to cause breakage? BFCs have to main behaviors that are relevant here: containing floats, and collapsing margins with children. Floats inside column-spanners might be a bit rare, but margins are common. It would imho make sense to preserve the BFC-ness to ensure the children of the spanner don't start collapsing with their parent.