13568 - {css2} not correctly solving for 'margin-top' when container has height:auto

Status: VERIFIED FIXED

(Core :: Layout, defect, P3)

Description

[Hixie (not reading bugmail)]

When the containing block has height:auto, we are not correctly solving the
equality given in CSS2 section 10.6.4 for margin-top.

In this example, the relevant properties have been set like this:

   top:                 0;
   margin-top:          auto;
   border-width-top:    0;
   padding-top:         0;
   height:              1em;
   padding-bottom:      0;
   border-width-bottom: 0;
   margin-bottom:       0;
   bottom:              0;

The equality says that all of those should add up to the height of the
containing block (distance from top padding edge to bottom padding edge of
nearest non position:static ancestor, in this case the parent).

Somehow, we do not perform this calculation correctly if the height in question
is 'auto'. (The current result is that the block being positioned goes off to
Mars somewhere and we don't render it).

If the height is explicitly given, or if we are solving for margin-bottom or
just bottom instead of margin-top, then it all works.

This can also be seen on this test case:
http://www.bath.ac.uk/~py8ieh/internet/projects/mozilla/absolute-margin-top.html

Comment 1

[Hixie (not reading bugmail)]

See bug 12945, which is also related to problems in section 10.6.4.

Comment 2

[troy]

It wasn't just that the containing block has height:auto, it's also that it is
relatively positioned (not absolutely positioned). That means we can't compute
the containing block height in a top-down fashion and we need to use the shrink
wrap height based on the content.

The spec is kind of vague on whether these things are calculated top-down (as
usual for in flow elements, e.g. percentage heights become 'auto' if the
parent's height is specified as 'auto') or bottom-up, but your example seems
reasonable and should work so I changed the code to use the shrink wrap height
in that case

Comment 3

[Hixie (not reading bugmail)]

Works fine with Viewer on Win32 1999-10-04. Marking verified.

Comment 4

[Hixie (not reading bugmail)]

Migrating from {css2} to css2 keyword. The {css1}, {css2}, {css3} and {css-moz}
radars should now be considered deprecated in favour of keywords.
I am *really* sorry about the spam...