107185 - Relative (1*) (cell)sizes is broken (between 0.9.5 and nightly 1027)

Status: RESOLVED INVALID

(Core :: Layout: Tables, defect)

Description

[Carl Lunde]

The screenshots says it all,
a table with two cells, one 1* and one 100px wide will be shown more like 100px
and 1*..

Comment 1

[Boris Zbarsky [:bzbarsky]]

OK.  This works in the 2001-10-21-05 nightly.  It does not in the 2001-10-22-10
nightly.

Over to HTML tables.  This is likely fallout from the fix for bug 52531 or bug
96343.

Comment 2

[Carl Lunde]

Seems like the code was removed
http://bugzilla.mozilla.org/show_bug.cgi?id=97027 here, but I can't see how this
was "Misuse of ParseValueOrPercentOrProportional"

Here's the CVS diff:
http://bonsai.mozilla.org/cvsview2.cgi?diff_mode=context&whitespace_mode=show&file=nsHTMLTableCellElement.cpp&root=/cvsroot&subdir=mozilla/content/html/content/src&command=DIFF_FRAMESET&rev1=1.51&rev2=1.52

W3C:
http://www.w3.org/TR/html4/types.html#type-multi-length
I can't see anything about this being invalid for cells? I'd rather think the
opposite

Comment 3

[Boris Zbarsky [:bzbarsky]]

Carl, the relevant part of the spec is here:

http://www.w3.org/TR/html4/struct/tables.html#h-11.2.6

The definition of the "width" attribute is:

  width = length [CN]
    Deprecated. This attribute supplies user agents with a recommended cell width.

Note that it's "length", not "multilength". So "1*" is indeed invalid for cells.

What confuses me is why this changed our rendering....  ccing bernd.

Comment 4

[Bernd]

taking

Comment 5

[Bernd]

Attachment: testcase

Comment 6

[Bernd]

as demonstrated by the testcases we ignored before my checkin width="1*", now we
render it as width="1" or in other words ignore the *. With my checkin our
rendering is now identical to IE and NN. Hey I didn't know that this even fixed
a bug. We didn't honour the prop width as visible by my testcase, the first two
columns did not proportionaly distribute the space. The html is invalid marking so.