Re: [css3-flexbox] transitions w/ min-width:auto not working in gecko

On 02/14/2013 03:08 PM, Tab Atkins Jr. wrote:
> That's a spec violation. :/  http://dev.w3.org/csswg/css3-flexbox/#min-size-auto

FWIW, dbaron actually started a thread in May about that chunk of spec
text being potentially-problematic.[1]  He and bz[2] suggested that the
special behavior applies to the used value, not the computed value.

You had one post in that thread[3] responding to dbaron, but I'm not
clear on whether you were agreeing with them or not.

[1] http://lists.w3.org/Archives/Public/www-style/2012May/0655.html
[2] http://lists.w3.org/Archives/Public/www-style/2012May/0663.html
[3] http://lists.w3.org/Archives/Public/www-style/2012May/0659.html

> Also, I don't understand what you mean.  Why do the flex containers
> have to care?  It's the children that have the value, and which are
> computing to min-content.

The flex containers perform the flex layout algorithm, which takes the
children's min-width values as inputs. That doesn't really answer your
question, though -- here's a more complete response:

[warning: gecko implementation details below]

I think you're imagine that a UA determines the computed value of
"min-width:auto" on a node by inspecting its parent's computed style,
and resolving "auto" to either "min-content" or "0" depending on the
parent's "display" and "flex-direction" values.

That makes sense, I think, but in practice, that can't (currently) work
in Gecko. When we're computing the value of a given property on a given
node, we only allow it to inspect a subset of the parent's computed
style (specifically, the properties that we store in the same "style
struct" -- basically, a group of closely-related properties).  This
allows inheritance to work (because we can see the value of our property
on the parent) without the child needing access to *all* of the parent's
computed style.

As it happens, min-width is stored in a different style-struct than
'display' and 'flex-direction', so it can't inspect those properties on
the parent.  So, we can't actually resolve "min-width:auto" during
style-computation -- so we leave it at "auto" and convert it
as-necessary when it's used.

Anyway, this will all be moot if min-width:auto is removed, so it may
not be worth worrying too much about.

Received on Thursday, 14 February 2013 23:25:20 UTC