295315 - table-cell height calculation uses border-box sizing; should use content-box

Status: RESOLVED DUPLICATE of bug 248239

(Core :: Layout: Tables, defect)

Description

[Jon Levell]

At the URL given I would expect the red table to be as tall as it is wide (And
the offsetHeight reported in the alert to be strictly greater than than the
style.height reported).

The width and height of the cell are bigger than its contents, setting the width
to 50 pixels means the overall width of the cell (including a cellpadding of
5px) is 60px, which is correct (as I understand the box model).

Setting the height of the cell to 50, causes the overall height of the cell
(including the cellpadding of 5px) to be (much to my surprise) 50 pixels.

It doesn't matter if I set the height via <td height="" or <td style="height:.."

I don't understand why the calculation of the width and height of the cell are
different.

This is a such a simple bug, that if it really was a bug, it would have been
report ad-infinitum but I don't know what I'm misunderstanding. I don't think
this is a dup of bug 22274 as giving the img a display style of block or
changing it to a few characters gives the same results - help?

(Tested with Mozilla Suite and Firefox 0.9.x upto latest 1.1 nightlies on Win
and Lin).

Comment 1

[Jon Levell]

I should have mentioned that IE6 treats the Height equivalently to the width (so
the red box is square) as I (mistakenly?) expect.  (Konq. 3.3.1 has a more
oblong box than Gecko and seems to ignore the height attribute entirely so that
the offsetHeight is only 30px).

Comment 2

[Bernd]

yeah it seems that the height is applied including the padding

Comment 3

[Bernd]

Attachment: strict testcase

Comment 4

[Bernd]

Attachment: quirks mode testcase

IE does differentiate between quirks and standards mode. So we need to follow
we should fix the behaviour for standards mode rendering and leave it as it is
for quriks mode.

Comment 5

[Bernd]

Attachment: patch

Comment 6

[Jon Levell]

This seems to work just fine to me - assuming people with the relevant authority
r & sr the patch should it be checked in pre or post Firefox 1.1?

Comment 7

[Bernd]

the patch needs to be rtest`ed

Comment 8

[Bernd]

Comment on attachment 185354 [details] [diff] [review]
patch

I rtested the patch, the only testcases that changed the behaviour, have been
the cases where in standards mode the height was defined, but that was
expected.

Comment 9

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 185354 [details] [diff] [review]
patch

>Index: nsTableRowFrame.cpp
>+    borderPadding = nsTableFrame::GetBorderPadding(nsSize(0, 0),
>+      GetPresContext()->PixelsToTwips(), (nsTableCellFrame*) aCellFrame);

Why 0x0? Won't this mess up percentage paddings?

>     case eStyleUnit_Percent: {
>+        SetPctHeight(position->mHeight.GetPercentValue() + 
borderPadding.top + borderPadding.bottom);

No, this is wrong -- it's mixing percent and twips.

Comment 10

[Bernd]

Boris if I go for fixed height only patch similar to the one that you correctly r-ed, do I have chance to get this reviewed while leaving the pct case unfixed?

Comment 11

[Boris Zbarsky [:bzbarsky]]

To be honest, I think David would be better-qualified to review this; he's been doing a lot more table layout stuff recently.

Comment 13

[gadjo]

isn't it a dupe of bug 248239?

Comment 15

[Bernd]

Attachment: strict pct testcase

Comment 16

[Bernd]

Attachment: comparative rendering

Comment 17

[Gérard Talbot]

(In reply to Bernd from comment #15)
> Created attachment 556233 [details]
> strict pct testcase

td.target {height:50%;}

"
CSS 2.1 does not define how the height of table cells and table rows is calculated when their height is specified using percentage values. 
"
17.5.3 Table height algorithms
http://www.w3.org/TR/CSS21/tables.html#height-layout

------------

There is a spec clarification bug on this btw:

Does auto-height table row height depend on padding+border of cells or just height?
https://www.w3.org/Bugs/Public/show_bug.cgi?id=15462

and their bug _15461_ is also somewhat related.

Such bug _15462_ is more a confirmation that what we are all thinking is correct than a real spec bug since there is no interoperability right now. 

------------

(In reply to gadjo from comment #13)
> isn't it a dupe of bug 248239?

Yes it is. From examining 
https://bug248239.bugzilla.mozilla.org/attachment.cgi?id=152910
bug 248239 is a DUPLICATE of this bug.

Gérard

Comment 18

[Tal Aloni]

(In reply to Gérard Talbot from comment #17)
> bug 248239 is a DUPLICATE of this bug.
Correct, and now that 248239 is resolved, this could be marked resolved as well.

p.s.
"strict pct testcase" is not a bug, the percentage height overrides the pixel height, and we render it just like Chrome / Safari / Opera.