NetNews Usenet Archive 1993 #3/NN_1993_3.iso/spool/comp/os/linux/25750

Text File | 1993-01-28 | 1.5 KB | 44 lines

Newsgroups: comp.os.linux
Path: sparky!uunet!charon.amdahl.com!pacbell.com!decwrl!concert!gatech!swrinde!zaphod.mps.ohio-state.edu!howland.reston.ans.net!usc!news.service.uci.edu!ucivax!ucla-cs!twinsun!eggert
From: [e-mail] (Paul Eggert)
Subject: Linux sigismember bug
Message-ID: <bkOFi=[e-mail]>
Sender: [e-mail]
Nntp-Posting-Host: farside
Organization: Twin Sun Inc, El Segundo, CA, USA
X-Newsreader: NN version 6.4.19
Date: Wed, 27 Jan 1993 03:56:28 GMT
Lines: 31

I'm not a Linux user, just an RCS maintainer.  But [e-mail]
(Ian Jackson) tells me that Linux has a bug that causes CVS's
configuration package to go into a loop.  Here's an example of the
problem.  Linux has only 32 signals, numbered 1 through 32.  The
program enclosed below invokes sigismember(&s,33).  Since 33 is not a
valid signal, Posix 1003.1-1990 sections 3.3.3.3 and 3.3.3.4 say that
sigismember(&s,33) must yield either -1 (setting errno to EINVAL), or
0.  Unfortunately, under Linux, sigismember(&s,33) yields 1, causing
the program to print `33 is a defined signal'.  That's incorrect, since
33 is not a member of any signal set.


#include <signal.h>
#include <stdio.h>

int
main()
{
    sigset_t s;
    if (sigfillset(&s) != 0) {
        perror("sigfillset");
        return 1;
    }
    switch (sigismember(&s, 33)) {
        case -1: perror("sigismember"); break;
        case 0: printf("33 is not a defined signal\n"); break;
        case 1: printf("33 is a defined signal\n"); break;
        default: printf("sigismember yields bad value\n"); return 1;
    }
    return 0;
}