Tests failing on recent master

[Jean-Marc Lasgouttes]

Le 18/05/2024 à 17:45, Jürgen Spitzmüller a écrit :
> Am Samstag, dem 18.05.2024 um 16:28 +0200 schrieb Kornel Benko:
>> How should I know? I am reacting to the output of
>> valgrind only.
> 
> I'd like to understand why the patch is needed. And why only for this
> boolean. Otherwise it seems just a shot in the dark.

You are right. What happens is that, since ad6aa8ba4b9, updateIndexes() 
relies on is_long_tabular, but this boolean is initialized *after* 
calling updateIndexes() in Tabular::init().

I could just put the initialization on line before but I prefer to let 
you have a look.

JMarc