54261 - HR's can become infinite, and are the wrong size always

Status: VERIFIED DUPLICATE of bug 53568

(Core :: Layout, defect, P3)

Description

[Randell Jesup [:jesup] (needinfo me)]

HR's with size < 3 (the default) become weird.  HR's with size 1 (or width 1)
can cause some serious problems, especially if they're NOSHADE.  Since <hr
size=1 noshade> is a typical "give me a one-pixel line" trick, this is bad. 
I've seen this lots of places.  Amazon used to use it (and may still).

The problem is caused because of how HR interacts with -moz-bg-inset (see
html.css).  The inset takes away pixels from the base size of the HR, and can
take more than the HR has to give.  <hr size=2 noshade> is 30 tall - but the
borders are each 15, leaving mComputedHeight as 0.  For <hr size=1 noshade>,
mComputedHeight becomes -15, and the FillRect goes to infinity (and lots of
things get screwed up, including other elements not appearing at all).

-moz-bg-inset (and outset I believe) should add the margins to the mRect of the
element, so the content size remains the same.

I'm uploading a simple example.

Comment 1

[Randell Jesup [:jesup] (needinfo me)]

Attachment: testcase

Comment 2

[Randell Jesup [:jesup] (needinfo me)]

Viewing the differences between this testcase and ns4.7 or ieeee makes the
problem VERY evident.

I suspect this is a very easy fix.  It might make sense to double-check how
regular borders are working (make sure they aren't stealing from the content
area) as well.

Updated subject to be more precise

BTW, I note that HR's w/ noshade are rounded-ends in mozilla (if tall enough),
and square in ns4.7.  Is this "correct"?

Comment 3

[Randell Jesup [:jesup] (needinfo me)]

This seems to be a recent regression.  CNN is showing this as well.

Comment 4

[tor]

*** This bug has been marked as a duplicate of 53568 ***

Comment 5

[Scott Brodmerkle]

Verified dupe.