206704 - ASSERTION: if you have an mtable in a table

Status: RESOLVED FIXED

(Core :: MathML, defect)

Description

[steve.swanson]

For each cell in an mtable in a table, I'm getting the following ASSERT

###!!! ASSERTION: not implemented: '!IsBorderCollapse(aParentFrame)', file 
c:/Mozilla/mozilla/layout/html/style/src/nsCSSFrameConstructor.cpp, line 1181

I don't understand what the ASSERTION is guarding against, but it seems a bit 
eager.  Or some style on the table is leaking into the mtable.

Comment 1

[steve.swanson]

Attachment: testcase

A very simple mtable in a single cell table.

Comment 2

[Boris Zbarsky [:bzbarsky]]

Yep, the <mtable> is ending up with a computed border-collapse of "collapse"....
that seems wrong, since I don't see any style rules explicitly telling it to
inherit the border-collapse value.

Comment 3

[rbs]

The collapse is being inherited from the fact that there is <table rules="all">. 

It is defined in the spec as:
http://www.w3.org/TR/1998/REC-CSS2-19980512/tables.html#borders
'border-collapse'
    Value:   collapse | separate | inherit
    Initial:   collapse
    Applies to:   'table' and 'inline-table' elements
    Inherited:   yes
    Percentages:   N/A
    Media:   visual

Aside from this bug, there seems to be a discrepancy with the code, though.
Why isn't |NS_STYLE_BORDER_COLLAPSE| the default in nsStyleStruct?

// -----------------------
// nsStyleTableBorder

nsStyleTableBorder::nsStyleTableBorder(nsIPresContext* aPresContext) 
{ 
  mBorderCollapse = NS_STYLE_BORDER_SEPARATE;

  nsCompatibility compatMode = eCompatibility_FullStandards;
  if (aPresContext)
    aPresContext->GetCompatibilityMode(&compatMode);
  mEmptyCells = (compatMode == eCompatibility_NavQuirks)
                  ? NS_STYLE_TABLE_EMPTY_CELLS_SHOW_BACKGROUND     
                  : NS_STYLE_TABLE_EMPTY_CELLS_SHOW;
  mCaptionSide = NS_SIDE_TOP;
  mBorderSpacingX.Reset();
  mBorderSpacingY.Reset();
}

Comment 4

[Boris Zbarsky [:bzbarsky]]

Oh, right. I forgot that border-collapse inherits by default....

As for the initial value, see
http://www.w3.org/Style/css2-updates/REC-CSS2-19980512-errata#s-17-6 (the
"section 17.6 borders" item).  And in general, it's just easier to look at the
CSS2.1 spec, which incorporates the CSS2 errata....

Comment 5

[steve.swanson]

Sorry, I'm confused.  Is the firing of the ASSERTION a bug, or not?  If it's 
not a bug, why does mtable have problems with border-collapse?

Comment 6

[rbs]

Attachment: patch - reset the two sole border properties that are inherited

it is weird (and counter-intuitive) that |border-collapse| and |border-spacing|
set on a table get inherited down to other deeper tables. All the other border
related properties are not inherited, which is more in line with the intuition.


Oh well. Anyway, here is patch to reset what is irrelevant to the <mtable>
case.

Comment 7

[rbs]

Comment on attachment 123979 [details] [diff] [review]
patch - reset the two sole border properties that are inherited

asking a= for some simple CSS rules to reset certain properties not relevant to
MathML <mtable>s.

Comment 8

[Asa Dotzler [:asa]]

Comment on attachment 123979 [details] [diff] [review]
patch - reset the two sole border properties that are inherited

a=asa (on behalf of drivers) for checkin to 1.4

Comment 9

[rbs]

Checked in.