dbu wrote:

The problem is indeed in function formatHeadings in Parser.php
The following line always unindents from the current depth, altough we might did
not reach that depth.

if( $toclevel<$wgMaxTocLevel ) {

$toc .= $sk->tocUnindent( $prevtoclevel - $toclevel );

}

I fixed it in my code using
$toc .= $sk->tocUnindent( min($prevtoclevel, $wgMaxTocLevel-1) - $toclevel );

From the svn repository, i think the bug is still present in current versions.
Can anybody confirm or deny this?

Fixed in r21725 along with a couple related bugs.

r21725 fixed the issue with $wgMaxTocLevel = 3 but it caused the same
bug with the default $wgMaxTocLevel = 999 (see bug #9760).

I reverted a small part of r21725 to solve bug #9760, thus reopening this one.

I meant bug #9764 instead of bug #9760.

Added a parser test in r21833 :

TOC with wgMaxTocLevel=3 (bug 6204)

Fixed in r21837.

We unindented line without checking if we did an indent before, that
was not always the case cause $wgMaxTocLevel actually prevent the indenting
of heading above its level.

brandono wrote:

This is not fixed yet (as of 1.11). This is the corner case.

for example when wgMaxTocLevel = 3

Section 1

Section 1.1

Section 1.1.1==== <<-- this is hidden

Section 1.1.1.1==== <<-- this is hidden

Section 2== <<-- this only gets UnIndented ONCE... not twice.

The problem is $prevtoclevel is not kept track of properly. The fix is to track it whenever we indent/unindent. You cannot (should not) track it with toclevel changes because those can fluctuate outside of the $wgMaxTocLevel area.

around line 3561, remove $prevtoclevel set

if( $toclevel ) {

$prevlevel = $level;
//$prevtoclevel = $toclevel;  //<<---commented out

}

around line 3605, track $prevtoclevel.

if( $toclevel<$wgMaxTocLevel ) {

if($prevtoclevel < $wgMaxTocLevel) {
    # Unindent only if the previous toc level was shown :p
    $toc .= $sk->tocUnindent( $prevtoclevel - $toclevel );
    $prevtoclevel = $toclevel; //<<--added.
} else {
    $toc .= $sk->tocLineEnd();
}

}

This fixed it for us by tracking $prevtoclevel only with actual indentation/unindentations. Might not be the right fix for everybody though. Thoughts?

Corrected example text (was missing a =):

Section 1

Section 1.1

Section 1.1.1

Section 1.1.1.1

Section 2

Should add some parser test cases for this?

conrad.irwin wrote:

Fixed (finally) in r72336.

brandono wrote:

I've long since moved on from this project but it's nice to see you guys still care :P.