40283 - Table miscalculation with cell width = 100% , NoWrap, and <input>

Status: VERIFIED FIXED

(Core :: Layout: Tables, defect, P1)

Description

[Juan Pablo Alcaraz]

I can not to check/uncheck a checkbox form. Its name is 'Mostrar mensajes
ampliados'.

Steps to Reproduce:

1) Go to http://www.patagon.com.ar/forlf.asp?fId=2&pi=1&ci=20&i=2&f=18&o=
fechaHora&s=desc&a=false&rnd=0.6461603

2) Try to click in 'Mostrar mensajes ampliados' in the upper right corner.


Actual Results:
I can not click in it.

Expected Results:
I need to click it.

Build Date & Platform:
2000052208 Windows NT Version

Additional Information:
I try it on Windows NT 4 Englsi Workstation + SP 5

Comment 1

[R.K.Aa.]

Tested on Linux M16 2000-052208:
There's something odd there: Only parts of the checkbox is visible.
Clicking IN the visible checkbox area has no effect.
However: Clicking on the edge/corner of it does toggle it checked/unchecked.

Comment 2

[WD]

Making simplified test case.   Looks like a layout problem, actually.

Comment 3

[WD]

Attachment: Simplified Testcase

Comment 4

[WD]

The test case is a table with 2 cells.

---------------------------------------
| width=100% | nowrap "text" <input> |
|            |                        |
---------------------------------------

It appears that in such a case, the <input> is not considered in the calculation
for the cell width.    In my example, the checbox is not clickable because it
falls on (under?) the border of the table.   Other types of <input> give the
same problem.

Comment 5

[rods (gone)]

reassigning

Comment 6

[karnaze (gone)]

rod, can you do a little research and if it is not a <input> problem then 
reassign to buster. The problem is that block 011F2204 is returning a max 
element width of 1005 (#1 below) but then when it is reflowed with that as an 
avail width it is returning a desired width of 1305. The block or its contents 
appear to be not right the first time.

  TO::Rfl en 011F1F50 rea=0 av=(8940,UC) comp=(0,0) count=0
     T::Rfl en 011F1FA4 rea=0 av=(8940,UC) comp=(UC,UC) count=1
       TRG::Rfl 011F2010 rea=0 av=(UC,UC) comp=(UC,UC) count=2
         TR::Rfl en 011F2054 rea=0 av=(UC,UC) comp=(UC,UC) count=3
           TC::Rfl 011F209C rea=0 av=(UC,UC) comp=(UC,UC) count=4
             Area::Rfl en 011F20F8 rea=0 av=(UC,UC) comp=(UC,UC) count=5
             Area::Rfl ex 011F20F8 des=(0,0) maxElem=(0,0)
          TC::Rfl ex 011F209C des=(165,180) maxElem=(165,180)
          TC::Rfl 011F21A8 rea=0 av=(UC,UC) comp=(UC,UC) count=6
             Area::Rfl en 011F2204 rea=0 av=(UC,UC) comp=(UC,UC) count=7
             Area::Rfl ex 011F2204 des=(1305,285) maxElem=(1005,285) // #1
          TC::Rfl ex 011F21A8 des=(1485,465) maxElem=(1185,465)
        TR::Rfl ex 011F2054 des=(1770,0) maxElem=(1350,465)
      TRG::Rfl ex 011F2010 des=(UC,30) maxElem=(1350,465)
      TRG::Rfl 011F2010 rea=2 av=(8910,UC) comp=(8910,UC) count=8
         TR::Rfl en 011F2054 rea=2 av=(8910,UC) comp=(8910,UC) count=9
           TC::Rfl 011F209C rea=2 av=(7635,UC) comp=(7455,UC) count=10
             Area::Rfl en 011F20F8 rea=2 av=(7485,UC) comp=(7485,UC) count=11
             Area::Rfl ex 011F20F8 des=(7485,0)
          TC::Rfl ex 011F209C des=(7635,180)
          TC::Rfl 011F21A8 rea=2 av=(1185,UC) comp=(1005,UC) count=12
             Area::Rfl en 011F2204 rea=2 av=(1005,UC) comp=(1005,UC) count=13
             Area::Rfl ex 011F2204 des=(1305,285) // #2
          TC::Rfl ex 011F21A8 des=(1185,465)
        TR::Rfl ex 011F2054 des=(8880,465)
      TRG::Rfl ex 011F2010 des=(8910,465)
    T::Rfl ex 011F1FA4 des=(8940,555)
  TO::Rfl ex 011F1F50 des=(8940,555)

Comment 7

[rods (gone)]

The problem is the line is not taking into account that it is in "nowrap" when 
it returns its max-element-size. It is incorrectly returning the MES of the 
largest frame it holds (which is fine for wrap mode), in this case the text 
"Check this:" which is 1005 twips wide (the input is 195 twips wide).

BRS: setting compute-MES to 0
BRS: setting compute-MES to 0
has multiple monitor apis is 0
BRS: setting compute-MES to 0
BRS: setting compute-MES to 0
BRS: setting compute-MES to 0
BRS: setting compute-MES to 1
       PASS1 Block(td)(0)@012057F4: line.floaters= band.floaterCount=0
nsBlockFrame::PostPlaceLine: 012057F4 setting line 0120587C MES 0
nsBlockFrame::PostPlaceLine block 012057F4 line 0120587C caching max width 0
nsBlockFrame::ReflowLine block 012057F4 line 0120587C setting 
aLine.mMaxElementWidth to 0
nsBlockFrame::CFS: 012057F4 returning MES 0
       PASS1 Block(td)(0)@012057F4: max-element-size:0,0 desired:0,0 
maxSize:1073741824,1073741824
block 012057F4 returning with maxElementSize=0,0
BRS: setting compute-MES to 1
       PASS1 Block(td)(1)@01205900: line.floaters= band.floaterCount=0
       PASS1 Block(td)(1)@01205900: old max-element-size=0,0 new=1005,285 (set 
wrong here)
nsBlockFrame::PostPlaceLine: 01205900 setting line 01205A3C MES 1005
nsBlockFrame::PostPlaceLine block 01205900 line 01205A3C caching max width 1305
nsBlockFrame::ReflowLine block 01205900 line 01205A3C setting 
aLine.mMaxElementWidth to 1005
nsBlockFrame::CFS: 01205900 returning MES 1005
       PASS1 Block(td)(1)@01205900: max-element-size:1005,285 desired:1305,285 
maxSize:1073741824,10737
41824
block 01205900 returning with maxElementSize=1005,285 
BRS: setting compute-MES to 0
BRS: setting compute-MES to 0

Comment 8

[buster]

Rod and Chris did great research: this should be easy to fix now.

Comment 9

[buster]

Attachment: better test case derived from the first. legal html, fewer elements

Comment 10

[buster]

I've got the fix for this.  Rod and Chris made it easy for me.
Recommend nsbeta2+ for these reasons:
1) trivial fix, no risk
2) potentially effects many pages that use tables.  it's a fix to the underlying 
block max-element-size measuring code that was just plain wrong in the case of 
nowrap being applied (via HTML attribute or CSS.)  max-element-size is used by 
tables to determine the width of columns.  if it's wrong, elements end up in the 
wrong place and/or columns end up with the wrong size.

Comment 11

[buster]

adding chris and rod to the cc list.  can one of you please review the patch?

Comment 12

[buster]

Attachment: patch. simply accounts for nowrap in computing maxElementWidth

Comment 13

[Peter Trudelle]

nsbeta2+ until 6/15. 

Comment 14

[buster]

fix checked in

Comment 15

[WD]

Verified