On Tue, 20 Nov 2007 17:24:42 -0500, Brion Vibber
wrote:
Steve Sanbeg wrote:
On Tue, 20 Nov 2007 21:26:48 +0000, brion wrote:
> Revision: 27696
> Author: brion
> Date: 2007-11-20 21:26:48 +0000 (Tue, 20 Nov 2007)
>
> Log Message:
> -----------
> Revert r27694 -- if you're seeing this problem, it's probably because
> you're pulling things out of $wgParser->mFunctionHooks and copying it
> to another parser without knowing the internal format has changed.
> That should be fixed whereever that's being done (as it was fixed on
> the parser tests.)
For labeled section transclusion, this was fixed by having the
extension call a recursive parse, instead of passing back the wiki
text to let the parser do its thing, which seems to be pretty broken
now; so the old way at least gives partial functionality.
So we probably won't be able to sort out any of the regressions until
we can figure out where that's being done.
The behavior of the LST code before
being changed was wrong (incorrect
parsing), but didn't trigger a fatal error message due to an invalid
callback, which is what r27694 was trying to protect against (by taking
an invalid callback entry and trying to extract a correct callback from
it, but without the correct flags being in the state array).
Can you clarify what triggered r27694 exactly?
-- brion vibber (brion @
wikimedia.org)
My bad, I think I had a bad update.
It looks like the significant regressions are mostly because
$headingOffset was removed from the parser. So if we can reimplement
that, things should be working again.
LST can be reimplemented by using the DOM representation of the
transcluded template to scan for <section> tags. I'm in the process of
doing this.
-- Tim Starling