Thanks to both Ole and Ronny for enlightening me. :)

-- 
Joshua Butikofer
Cluster Resources, Inc.

[EMAIL PROTECTED]
Voice: (801) 717-3707
Fax:   (801) 717-3738
--------------------------


Ole Holm Nielsen wrote:
> Ronny T. Lampert wrote:
>>> I'm not sure I understand what you mean by "daemon should duplicate
>>> all standard descriptors to
>>> /dev/null." Perhaps you can help me in my ignorance.
>>
>> Ah, yes. Usually, fully background daemons close 0, 1 and 2 (stdin,
>> out and err) and dup() them to /dev/null, and also do a setsid().
>>
>> This prevents accidentially opened ttys (and sessions) to be held open
>> or msgs (e.g. perror()) going somewhere where they possibly shouldn't.
>> One also could dup() them to a file, so all output goes there in case
>> some library functions really wants to use stdout or -err.
>>
>> I think closing stdin is quite crucial to correct operation, else the
>> daemon could accidentially get blocked.
> 
> I've also noticed that if I restart the maui daemon, my interactive
> SSH-login shell to the central server refuses to terminate.  I have
> to kill the shell from adifferent window.  Fortunately, the maui
> daemon has always survived this harsh treatment of its open file
> descriptors.
> 
> I have a (possibly naive) snippet of code from another type of
> daemon process where the files are properly closed:
> 
>         /* Become a daemon */
>         pid = fork();
>         if (pid > 0) {          /* Parent */
>                 /* Close parent's streams */
>                fclose (stdin);
>                fclose (stdout);
>                fclose (stderr);
>                 return(0);
>         } else if (pid < 0) {   /* Error */
>                 printf("%s: Could not fork child process\n", argv[0]);
>                 return(-1);
>         }
> 
> The Maui code in src/server/OServer.c doesn't close file descriptors,
> which IMHO it ought to do.  Ronny has even more detailed suggestions
> above which I don't claim to understand fully.
> 
> Best regards,
> Ole Holm Nielsen
> Technical University of Denmark
_______________________________________________
mauiusers mailing list
mauiusers@supercluster.org
http://www.supercluster.org/mailman/listinfo/mauiusers

Reply via email to