Fixes failure reported by t/config/socketoptions.t testsuite

[Sashan]

there are both (SO_KEEPALIVE and TCP_KEEPIDLE) TCP socket options defined on Solaris. In order to make test suite happy I've decided to let TCP_KEEPIDLE precedence over SO_KEEPALIVE when building proftpd on OS which supports both those options.

this pull request just highlights the issue. perhaps there is better way to fix it.

[Castaglia]

Does Solaris have both TCP_KEEPALIVE and SO_KEEPALIVE defined? Are they defined as the exact same thing for some reason?

[Sashan]

Does Solaris have both TCP_KEEPALIVE and SO_KEEPALIVE defined? Are they defined as the exact same thing for some reason?

Sorry there was typo in my earlier command. The typo really confuses things. The issue is not about SO_KEEPALIVE vs TCP_KEEPALIVE it is TCP_KEEPIDLE vs. TCP_KEEPALIVE. Both those TCP_KEEP* options are handled here in src/inet.c

 978   /* We only try to set the TCP keepalive specifics if SO_KEEPALIVE was
 979    * enabled successfully.
 980    */
 981   pr_trace_msg(trace_channel, 15, "enabled SO_KEEPALIVE on socket fd %d",
 982     c->listen_fd);
 983 
 984   /* On Mac OS, the socket option is TCP_KEEPALIVE rather than
 985    * TCP_KEEPIDLE.
 986    */
 987 #if defined(TCP_KEEPIDLE) || defined(TCP_KEEPALIVE)
 988   val = tcp_keepalive->keepalive_idle;
 989   if (val != -1) {
 990     int option_name;
 991 
 992 # if defined(TCP_KEEPALIVE)
 993     option_name = TCP_KEEPALIVE;
 994 # else
 995     option_name = TCP_KEEPIDLE;
 996 # endif /* TCP_KEEPALIVE or TCP_KEEPIDLE */
 997 
 998 # ifdef __DragonFly__
 999     /* DragonFly BSD uses millsecs as the KEEPIDLE unit. */
1000     val *= 1000;
1001 # endif /* DragonFly BSD */
1002     if (setsockopt(c->listen_fd, IPPROTO_TCP, option_name, (void *) &val,
1003         sizeof(int)) < 0) {
1004       pr_log_pri(PR_LOG_NOTICE,
1005         "error setting TCP_KEEPIDLE %d on fd %d: %s", val, c->listen_fd,
1006        strerror(errno));
1007 
1008     } else {
1009       pr_trace_msg(trace_channel, 15,
1010         "enabled TCP_KEEPIDLE %d on socket fd %d", val, c->listen_fd);
1011     }
1012   }
1013 #endif /* TCP_KEEPIDLE */

The thing is the Solaris uses both options TCP_KEEPALIVE and TCP_KEEPIDLE The TCP_KEEPALIVE is used on Solaris to change keep alive TCP timer. While TCP_KEEPIDLE option tunes keep-alive timer per-socket (https://docs.oracle.com/cd/E88353_01/html/E37851/tcp-4p.html)