[css-tables] height on a table-row-group

[gregwhitworth]

Compat Issues: Yes

Description:
When a height is specified on a table-row-group:

• Chrome ignores the height.
• Firefox respects the height (as if a cell spanned the rows with that height).
• Edge applies the height to each cell (definitely buggy).

Proposed Option:
Firefox behavior looks good.

CSS 2.1 Option:
Height distribution is undefined

Other options:
Blink/Webkit currently ignore the height on a table-row-group we could go that route as well.

Testcases:
http://codepen.io/FremyCompany/pen/MKVPGR?editors=1100

[FremyCompany]

The CSSWG resvoled this issue during TPAC2016:

User agents should work like Gecko, and distribute the height among the rows spanned by the tbody/thead/tfoot.

[gregwhitworth]

Glancing through the specification there isn't an obvious place to drop this in, looking through we'll need to add a section similar to "Computing Column Measures" but for "Computing row measures" at which point we'll place this in there. This is not an easy thing to just swap, we'll want to verify them all with tests.

[gregwhitworth]

So @FremyCompany and myself starting testing the "Computing Column Measures" to see if we could replace the directionality terms within the section and make this generic for both columns and rows. At this point, we've proved that "intermediate min-content width for span 1" works the same for "intermediate min-content height for span 1".

[atotic]

So @FremyCompany and myself starting testing the "Computing Column Measures" to see if we could replace the directionality terms within the section and make this generic for both columns and rows. At this point, we've proved that "intermediate min-content width for span 1" works the same for "intermediate min-content height for span 1".

There are significant differences in how percentage size is resolved. For columns, %ge cells grow to fill percentage, and make table wider. For rows, %ge cell do not grow initially, but do grow to fill their size if excess height is getting distributed.

[atotic]

Chrome's TablesNG implementation will be very close to FF. The tests I've used will land shortly in wpt as
css/css-tables/tentative/tbody-height-redistribution.html
The main differences with FF is treatment of empty rows.