unnecessary 'int' in s_conf.c

[Lubomir Radev]

A smart compiler (which gcc supposedly is) would optimize this code in such
way that ret will never be allocated and therefore not referenced - even
with the default -O2 only


> -----Original Message-----
> From: hybrid-bounces at lists.ircd-hybrid.org [mailto:hybrid-
> bounces at lists.ircd-hybrid.org] On Behalf Of Rachel Llorenna
> Sent: Friday, January 07, 2005 7:02 PM
> To: Mailing List: ircd-hybrid
> Subject: unnecessary 'int' in s_conf.c
> 
> >From ircd-hybrid-current, serial number 20041220_0. The 'int ret' is
> unnecessary in s_conf.c:lookup_confhost (lines 2161-2205),
> 
> static void
> lookup_confhost(struct ConfItem *conf)
> {
>   struct AccessItem *aconf;
>   struct addrinfo hints, *res;
>   int ret;
> 
>   aconf = map_to_conf(conf);
> 
>   if (EmptyString(aconf->host) ||
>       EmptyString(aconf->user))
>   {
>     ilog(L_ERROR, "Host/server name error: (%s) (%s)",
>          aconf->host, conf->name);
>     return;
>   }
> 
>   if (strchr(aconf->host, '*') ||
>       strchr(aconf->host, '?'))
>     return;
> 
>   /* Do name lookup now on hostnames given and store the
>    * ip numbers in conf structure.
>    */
>   memset(&hints, 0, sizeof(hints));
> 
>   hints.ai_family   = AF_UNSPEC;
>   hints.ai_socktype = SOCK_STREAM;
> 
>   /* Get us ready for a bind() and don't bother doing dns lookup */
>   hints.ai_flags = AI_PASSIVE | AI_NUMERICHOST;
> 
>   if ((ret = irc_getaddrinfo(aconf->host, NULL, &hints, &res)))
>   {
>     conf_dns_lookup(aconf);
>     return;
>   }
> 
>   assert(res != NULL);
> 
>   memcpy(&aconf->ipnum, res->ai_addr, res->ai_addrlen);
>   aconf->ipnum.ss_len = res->ai_addrlen;
>   aconf->ipnum.ss.ss_family = res->ai_family;
>   irc_freeaddrinfo(res);
> }
> 
> Note that 'ret' is not referenced beyond that single if statement.
> Since the return value of irc_getaddrinfo is only used by that one
> 'if' statement, why not collapse that line to:
>   if (irc_getaddrinfo(aconf->host, NULL, &hints, &res)) ...
> 
> And remove the declaration of ret? This would only save a small and
> insignificant amount of memory, but at least it doesn't come at the
> cost of CPU cycles or other limited resources.
> --
> Regards,
> 
> Rachel Llorenna (frequency)