Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!swrinde!elroy.jpl.nasa.gov!usc!orion.oac.uci.edu!cedman
From: cedman@golem.ps.uci.edu (Carl Edman)
Newsgroups: comp.sys.hp
Subject: Re: The "getdtablesize" command
Message-ID: <CEDMAN.90Dec19093804@lynx.ps.uci.edu>
Date: 19 Dec 90 17:36:59 GMT
References: <28687@mimsy.umd.edu> <4310152@hpindda.cup.hp.com>
Organization: University of California, Irvine, USA.
Lines: 39
Nntp-Posting-Host: lynx.ps.uci.edu
In-reply-to: ash@hpindda.cup.hp.com's message of 19 Dec 90 01:45:04 GMT

In article <4310152@hpindda.cup.hp.com> ash@hpindda.cup.hp.com (Art Harkin) writes:
   Try:

   #include <sys/types.h>

   #ifdef hpux
       t = FD_SETSIZE;
   #else /* hpux */
       t = getdtablesize();
   #endif /* hpux */


   Art Harkin
   Hewlett Packard

How come that _no_one_ in the world, not even HP, uses its own system
call to do this ? (sysconf, a solution with which I posted a few days ago).

Also this solution is, if I understand correctly, incorrect or at
least wastefull. While FD_SETSIZE is guaranteed to be greater than
FILENO, thus it probably wouldn't cause any harm for the most common
applications, it is at least wasteful.

Also imagine, e.g., an application which has a variable number of
sockets open. This application has usually many low-priority
connections, which can be closed at a moments notice, without any harm
but must be able to accept some uncommon high-priority messages
quickly. So the application closes one low-priority connection when it
notices that its number of open connections approaches FILENO. Using
FD_SETSIZE instead of this breaks this application.

        Carl Edman



Theorectical Physicist,N.:A physicist whose  | Send mail
existence is postulated, to make the numbers |  to
balance but who is never actually observed   | cedman@golem.ps.uci.edu
in the laboratory.                           | edmanc@uciph0.ps.uci.edu