286143 - Content gets doubled when hovering over tbody which has style tbody:hover{display:table-caption;}

Status: VERIFIED FIXED

(Core :: Layout: Tables, defect)

Description

[Martijn Wargers (dead)]

The testcase comes from bug 281527.

The content gets doubled when hovering over the text in the testcase. This
should not happen.

Comment 1

[Bernd]

Attachment: patch

the patch does not fix the crash in the dependent bug 286137 but testcases 2
and 3 in that bug

Comment 2

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 177636 [details] [diff] [review]
patch

So... it seems this code is just pretty confused, all told.

Notice the call to GetCaptionAdjustedParent in the "aParent is not a table"
branch of this if.  That seems to be pretty bogus, no?

Not only that, but how are we ending up in AppendFrames with a caption?  What
This function appends an entire list, of frames, no?  Is it only called with
lists of length 1?  What happens when a caption is not the first element of the
list?  Or when a caption is first but a non-caption follows?

All this apart from AppendFrames clearly saying it deals with _principal_ child
lists, so we shouldn't be entering it with captions, table columns, etc!  If we
ever end up supporting :after on tables (which we need to do for CSS2.1
compliance), the very first block of code in this method will screw up the
append for those frame types.

If only lists of length one are passed in and AppendFrames is expected to
handle non-principal child lists for some reason, I think some comments
(asserts?) to that effect are in order, and I think the right logic here would
be:

nsIFrame* outerTableFrame;
if (nsLayoutAtoms::tableFrame == aParentFrame->GetType() &&
    nsLayoutAtoms::tableColFrame == aFrameList->GetType()) {
  // Do whatever with columns
}
else if (GetCaptionAdjustedParent(aParentFrame, aFrameList, &outerTableFrame))
{
  // append to outerTableFrame
}
else {
  // append to aParentFrame
}

Also, as long as we're here, GetCaptionAdjustedParent should probably call
AdjustCaptionParentFrame or something...  or we should only have one function
that does this.

Comment 3

[Bernd]

fixed by the checkin for bug 286137

Comment 4

[Ben Fowler]

(In reply to comment #0)
> The testcase comes from bug 281527.

What I see is an oscillating effect in which the line of text is formatted
as a thin column, and then immediately re-formatted as a single line.

I can certainly see how this happens, but is it the expected behaviour?

Comment 5

[Bernd]

hmmm expected, the caption tries to minimize it size as the table that it should
cover has 0 width (is nonexistent).That might be another bug but this one is fixed.