[css-tables] Specify the containing block of a table cell

[mwenge]

There is disagreement among browsers, and no specification in CSS2.1, concerning the containing block of a table cell.

Blink consider the row the containing block. I believe Firefox and Edge consider the table the containing block.

Can the editor clarify for me and maybe add something to the spec?

Others have encountered the same question.

[FremyCompany]

How do you test that exactly? Percentages resolution in tables is special and does not obey the general rules which reference containing blocks, but there might be another side effect I didn't think of that helped you determine that Blink uses the row and Edge uses the table?

[Nadya678]

Greater problem is table-cell with position:relative and inside block absolute-positioned. The IE badly interpret this.

Next problem is no colspan and row-span equivalent.

[mwenge]

I had it the wrong way round: Blink considers the table the cell's containing block. I don't have a good test that displays the difference in behaviour between browsers because I think we've all hacked our way around to the right answer where differences existed. I think the spec would benefit from defining the cell's containing block because, for example, it's not obvious that the right answer is either the row, section, or table.

[FremyCompany]

I think the spec would benefit from defining the cell's containing block because, for example, it's not obvious that the right answer is either the row, section, or table.

I'm not confident going in any direction here because I don't understand what impact this would have on layout.

That being said, it does not seem crazy to me to say the containing block of a table-cell is the table-cell itself, but I'm afraid it might depend on which layout pass we are talking about. In the first layout passes, the table-cell doesn't have any size yet and I guess if there are layout steps that need the containing block for fit-content sizing purposes, and maybe they would want to use the table one for these purposes, though maybe they don't care at all.

It's difficult to make a normative statement on a claim I cannot back up with a test.