257955 - A horizontal bar is surely displayed by MozillaZine Japan.

Status: RESOLVED FIXED

(Core :: Layout: Form Controls, defect)

Description

[Hiro]

The trunk version of 2004-09-03 In nightly build, since a horizontal bar is
always displayed, access of a page becomes difficult.

The trunk version of 2004-08-31 There is no problem in nightly build.

Reproducible: Always
Steps to Reproduce:
1. Open http://ryuzi.dyndns.org/mozillazine/html/modules/news/

Actual Results:
display the horizontal bar

Expected Results:  
don't display the horizontal bar

Mac OS X 10.3.5

Comment 1

[Hiro]

Attachment: Screen shot of Mozilla nightly build 20040903

Comment 2

[Hiro]

Attachment: Screen shot of Mozilla nightly build 20040831

There is no problem in this build.

Comment 3

[Boris Zbarsky [:bzbarsky]]

What about the 09-01 and 09-02 nightlies?  Having a one-day regression range
would help a lot.

This is very unlikely to be a parser bug.  Over to layout, but we need a minimal
testcase to figure out what's up.

Comment 4

[Hiro]

The Mac version of 09-01 There was no nightly build.
And tt reproduced also by 09-02.

Comment 5

[Boris Zbarsky [:bzbarsky]]

One more question... When are those nightly builds made?  That is, what's the
hour in the timestamp?

Comment 6

[Hiro]

The time stump of mozilla is as follows.
It is displaying by TZ of Japan.

-rwxr-xr-x    1 sek  unknown  115708  3  9 01:59 mozilla-bin

Comment 7

[Hiro]

oops...
comment#6 is 09-02 nightly build.

Comment 8

[Boris Zbarsky [:bzbarsky]]

I meant the timestamp in the title bar of the browser..

Comment 9

[Hiro]

It is displayed on the title bar as follows.
   Build ID:2004090208

And this was also reproduced, when it checked now, since there was Firefox of
2004-09-01-08-trunk .

Comment 10

[Bernd]

I see it with 2004090105 winxp.
A reduced testcase would be nice.

Comment 11

[Bernd]

Attachment: testcase

Comment 12

[Bernd]

Attachment: patch

Comment 13

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

One question: should nsHTMLReflowMetrics::SetMEWToActualWidth include
borderPadding for the percentage-width case?

Shouldn't nsFieldSetFrame::Reflow initialize aDesiredSize.mMaxElementWidth to
the borderpadding instead of zero, in particular for the case where there's no kid?

What happens when the legend is percentage width?

Comment 14

[Bernd]

>Should nsHTMLReflowMetrics::SetMEWToActualWidth include borderPadding for the
percentage-width case?

That might be a good idea, if the content width is zero  the box has anyway the
dimensions of borderpadding.left + borderpadding.right.

Comment 15

[Bernd]

Attachment: minimal fieldset

there is no empty fieldset

Comment 16

[Bernd]

line 024E3350: count=1 state=block,clean,prevmarginclean,not impacted,not
wrapped,nobr[0x804] {24,0,7104,259} <
                FieldSet(fieldset)(1)@024E3128 {24,0,7104,259} [state=00800004]
[content=024EDBB0] [sc=024E307C]<
                  Area(fieldset)(1)@024E317C {144,91,0,0} [state=00900004]
sc=024E31FC(i=1,b=0)<
                    line 024E32F0: count=1
state=inline,clean,prevmarginclean,not impacted,not wrapped,nobr[0x800] {0,0,0,0} <
                      Text(0)@024E32B0[0,2,T]  {0,0,0,0} [state=51100024]
sc=024E3284<
                        "\n "
                      >
                    >
                  >
                >
              >

Comment 17

[Bernd]

Attachment: minimal legend

Comment 18

[Bernd]

content model for the legend case:
  body@024E8D18 refcount=3<
    Text@024BD540 refcount=2<\n  >
    fieldset@024BC6C0 refcount=3<
      legend@024BC7B8 refcount=2<
        Text@024BCAE8 refcount=2<\n >
      >
    >
  >

Comment 19

[Bernd]

legend frames dont honour any width spec. They are currently pure content driven.
See the regression testcases that I checked in.

http://lxr.mozilla.org/seamonkey/source/layout/html/tests/formctls/base/legend_strict.html

Comment 20

[Mats Palmgren (inactive)]

*** Bug 258945 has been marked as a duplicate of this bug. ***

Comment 21

[Bernd]

Attachment: revised patch

this also defaults the mew to borderpadding. As the previous testcases show
there is always a child to a fieldset frame. This might change.

Comment 22

[Bernd]

Attachment: the real patch

Comment 23

[Bernd]

fix checked in