- From: Tab Atkins Jr. <jackalmage@gmail.com>
- Date: Mon, 23 Apr 2012 17:51:04 -0700
- To: Daniel Holbert <dholbert@mozilla.com>
- Cc: www-style list <www-style@w3.org>, Alex Mogilevsky <alexmog@microsoft.com>
On Mon, Apr 23, 2012 at 5:04 PM, Daniel Holbert <dholbert@mozilla.com> wrote: > Hi www-style, > > I just noticed one circular dependency in the spec that could stand some > clarification. > > So -- the definition for 'flex-align: stretch' [1] says that it makes > flexbox items' cross sizes resolve to their line's cross size. > > Meanwhile, in the Flexbox Layout Alg section about computing a line's > cross size, we use its items' cross sizes as inputs. [2] > > So: when we're computing a line's cross size, what should we be using as > the cross sizes for its auto-sized items with flex-align:stretch? I see > two possibilities: > (a) 0 (since it's not resolved yet) > (b) the "normal" auto size (disregarding the effects of "stretch") > > I'd advocate (b), because otherwise, flexboxes whose children are all > auto-sized and have flex-align:stretch (the defaults) would have a cross > size of 0 & not render anything, which is pretty bad default behavior. > > Could we clarify this in the spec? I think the chunk that most needs to > address this is [2]. (where it talks about using the items' cross sizes > as inputs) Yup, it should be (b). fantasai and I are right now rewriting the layout algorithm to ground it in extreme specificity, and we're finding that all the layout algorithms are giant piles of hands, blindly waving. ~TJ
Received on Tuesday, 24 April 2012 00:51:53 UTC