Duplicate virtual server "xxx" always hits on HUP

[Alan DeKok]

Bjørn Mork wrote:
> Which implies that commit 5a710e98 is completely bogus.

  Well, no.  It's still a good idea.  The implementation is just too
specific.

>  We cannot check
> for existing servers that way.  It would be possible to eliminate those
> with matching file names and line numbers, but that would still fail for
> no good reason if you modified the virtual server configuration and
> HUPed the server.

  The solution is to use cf_top_section().  If the two virtual servers
share the same top-level CONF_SECTION, then they're duplicates.
Otherwise, they're from different HUPs.

  Alan DeKok.