Re: [htdig3-dev] Alarm clock


Subject: Re: [htdig3-dev] Alarm clock
From: Geoff Hutchison (ghutchis@wso.williams.edu)
Date: Sat Feb 26 2000 - 17:04:13 PST


At 7:11 PM +0200 2/23/00, Valdas Andrulis wrote:
>~/htdig-3.2.0/bin$ ./htdig -v -i
>
>ht://dig Start Time: Wed Feb 23 19:05:08 2000
>New server: humoro.centras.lt, 80
>Alarm clock
>
>~/htdig-3.2.0/bin$
>
>What does it mean? Shoudn't program just retry or forget this server and
>continue with others? I think this alarm is in htnet code.

Hmm. The alarm isn't supposed to kill it! The alarm comes from here:

        for (;;)
          {
            // Set an alarm to make sure the connect() call times out
appropriately
            // This ensures the call won't hang on a dead server or
bad DNS call
            alarm(timeout_value);
            status = ::connect(sock, (struct sockaddr *)&server,
sizeof(server));
            alarm(0);
            if (status < 0 && errno == EINTR && !allow_EINTR)
              {
                ::close(sock);
                open();
                continue;

Previously, people reported that the ::connect would never time out,
or at least until the OS decided to kill off that socket (often quite
long). So on the advice of a few people, I put in the alarm. Now I
haven't used alarms too often--am I missing something?

-Geoff

------------------------------------
To unsubscribe from the htdig3-dev mailing list, send a message to
htdig3-dev-unsubscribe@htdig.org
You will receive a message to confirm this.



This archive was generated by hypermail 2b28 : Sat Feb 26 2000 - 17:18:23 PST