759144 - [BC]Dynamic changes to cell border widths can't seem to shrink the used border width

Status: NEW

(Core :: Layout: Tables, defect)

Description

[Andy Earnshaw]

User Agent: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/535.19 (KHTML, like Gecko) Ubuntu/12.04 Chromium/18.0.1025.151 Chrome/18.0.1025.151 Safari/535.19

Steps to reproduce:

I have a table, of which several rows are hidden by default for the purposes of animating when the table takes focus. These rows are hidden by setting the font-size and border-width of their child TD elements to 0 in order to work around the limitation of not being able to animate the CSS display property.




Actual results:

Once a cycle of showing and hiding the elements again has completed, one or more of the TD elements now has a computed height and border-width of 1px. This occurs regardless of animation or choice of pseudo-class used. 


Expected results:

The computed style values should return to their original state once the pseudo-class no longer applies.

An example is available at http://jsfiddle.net/NprYn/. Click the table once to expand several hidden rows. Click anywhere else to hide the rows again and look at the top of the table.

Comment 1

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

Attachment: reporter's testcase (unsimplified)

Comment 2

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

Attachment: reporter's testcase with jsfiddle boilerplate removed

Comment 3

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

Attachment: reporter's testcase with jsfiddle boilerplate removed

td { padding: 0px } was essential to the testcase but only included because it was part of the JSFiddle normalize.css

(Please don't use JSFiddle for testcases for bugs.)

Comment 4

[Andy Earnshaw]

Sorry, I'm very lazy and it won't happen again :-)

Comment 5

[Boris Zbarsky [:bzbarsky]]

The basic issue is that BCMapCellInfo::SetBottomBorderWidths and BCMapCellInfo::SetTopBorderWidths can never shrink the border widths of the cell, as far as I can see:

5293        mCell->SetBorderWidth(NS_SIDE_BOTTOM, NS_MAX(aWidth,
5294                              mCell->GetBorderWidth(NS_SIDE_BOTTOM)));

and

5306         mCell->SetBorderWidth(NS_SIDE_TOP, NS_MAX(aWidth,
5307                               mCell->GetBorderWidth(NS_SIDE_TOP)));

The code has been this way (modulo some renaming) since this checkin:

  3.471 <[e-mail]> 2002-02-19 07:48
  bug 41262 - new table border collapsing code. sr=attinasi, r=alexsavulov.

We really need to take an axe to the [BC] stuff.  :(

Comment 6

[Gérard Talbot]

Test
----

http://www.gtalbot.org/BugzillaSection/Bug759144-Dynamic-changes-border-width.html

The border-top of the topmost cell is 0px. But since the table's border-top is 50px, then the resulting border-top for that cell is 50px: the wider border of conflicting cells wins. This happens in both tables. 

Now, after the document has finished loading, the border-top of the topmost cell of the first table, of the dynamic table is dynamically set from 0px to 50px. Since the table's border-top is already also 50px, then the collapsed border of the cell should still remain 50px too. But this execution in Firefox 47 and in Firefox 50.0a1 (buildID=20160701121753) unexpectedly changes the rendering (rightmost part) of the blue border-top of the 3rd cell.

There are other tests possible to do on dynamic changes to cell border widths but this one is probably the simpliest I could do.