414190 - getBoundingClientRect is broken for tables

Status: RESOLVED FIXED

(Core :: DOM: CSS Object Model, defect, P1)

Description

[Robert O'Callahan (:roc) (email my personal email if necessary)]

getBoundingClientRect currently uses the table outer frame to get geometry, which is wrong because it includes margins.

IE basically seems to return the union of the caption and inner table border boxes, so for compatibility we should do that too. For getClientRects() we could return one rect that's the union of those two border boxes, for compatibility with IE, or for more precise geometry we could return two rectangles, one for the main table and one for the caption if any. I'm not so sure about that but I think I'll go for returning two rects with the table rect as the first one ... cases where the getClientRects incompatibility matters would be rather obscure.

Comment 2

[Anne (:annevk)]

Would this be HTMLTableElement specific or apply to all elements with display:table (and maybe inline-table?)?

Comment 3

[Robert O'Callahan (:roc) (email my personal email if necessary)]

All elements with display:table I guess.

The underlying problem is that the drawn border of a display:table element doesn't enclose its display:table-caption child (if any). So we have similar issues as in bug 363183 (getComputedStyle for tables), except they're easier to solve here because we're dealing with border-box metrics instead of content-box metrics, so it's not absurd to add the height of the inner table to the height of the caption.

Comment 4

[Robert O'Callahan (:roc) (email my personal email if necessary)]

BTW our offset* property implementation ignores the caption and returns values only for the inner table. I haven't tested what IE does. But we've always done that so I don't plan to change it now. Right now I'm more concerned about getting the shiny new API right.

Comment 5

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Having getClientRects return separate rects for the inner table and caption could actually be important ... otherwise there would be no easy way to compute the rect for the inner table.

In pagination situations my current implementation will return all the rectangles for the table pieces followed by all the rectangles for the caption pieces.

Comment 6

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Attachment: fix

This does what I suggested.

Comment 7

[Mike Schroepfer]

^

Comment 8

[Mats Palmgren (inactive)]

Comment on attachment 299582 [details] [diff] [review]
fix

In AddRectanglesForFrames():
>+    if (!TryGetSVGBoundingRect(aFrame, &r)) {

s/aFrame/f/

>+      r = nsRect(GetOffsetFromInitialContainingBlock(f), f->GetSize());

r.SetRect(GetOffsetFromInitialContainingBlock(f), f->GetSize()) is better?

r+sr=mats with the first nit fixed.

Comment 9

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Those rect operations should compile to the same code. I slightly prefer the first one.

Checked in.