> Gagh -%N0 wasn't in that one. Here's 3.45-1 again,
> with -%N0. Same behavior, -%N0 seems to have no
> effect whatsoever

Humm, I'm seeing it now. This is not related to %N0 however, the backing
should also work despite delayed type checks.

The bug might be located in back_fill() (htscore.c), the following test looks
buggy to me:

      if (ok && liens[p]->sav != NULL && liens[p]->sav[0] != '\0' 
        && hash_read(opt->hash,liens[p]->sav,"",0,opt->urlhack) >= 0)     
lookup in liens_record

I'm now trying to figure out why in hell I did put this check. I wish I had
the idea to put the sources on svn at this time, damn it.