[cssom] Resolved value for width et al should be computed value for display:contents

[ghost]

The CSSOM spec currently says this should be the case for display:none. Should probably be the same for display:contents.

[zcorpan]

Yeah. That also seems to match what Gecko does.
http://software.hixie.ch/utilities/js/live-dom-viewer/saved/4470

[MatsPalmgren]

FYI, the relevant Gecko code is:
https://hg.mozilla.org/mozilla-central/annotate/8a494adbc5cc/layout/style/nsComputedDOMStyle.cpp#l4863

So, yes, when the element doesn't have a box, or is a non-replaced or SVG inline,
then 'calcWidth' is false and we use the else-branch on line 4887, which uses
the computed values. But note that min-/max-* are used as well, so it seems to be
a little more than just "use the computed 'width' value".

A few examples:
data:text/html,<div style="display:none; min-width:10px">
I get a resolved 'width' of 1293px (the content-width of the implicit )
data:text/html,<div style="display:none; max-width:10px">
results in width = 'auto'
data:text/html,<div style="display:none; width:20px; max-width:10px">
width = 10px
data:text/html,<div style="display:none; min-width:20px; max-width:10px;">
width = 'auto'
data:text/html,<div style="display:none; width:0; min-width:20px; max-width:10px;">
width = 20px

What do other UAs do in these cases?

[MatsPalmgren]

Correction,
data:text/html,<div style="display:none; min-width:10px">
is also width = 'auto'

So, it appears that a specified width:auto always resolves to 'auto' (when there is no box),
otherwise min/max is also applied. Here's the code that does this:
https://hg.mozilla.org/mozilla-central/annotate/8a494adbc5cc/layout/style/nsComputedDOMStyle.cpp#l5341

[upsuper]

@MatsPalmgren note that, we are changing our behavior of min-width for computed style, because per spec, it should be computed value, not used value. And some other engines also return computed value for {min,max}-{width,height} conforming the spec. See bug 1292447.

[zcorpan]

Fixed the display:contents case in c65b958. Please open a new issue if there's something that needs to be fixed in the spec for min/max-width/height.

[zcorpan]

Reopening per #540 (comment)