Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83; site wgivax.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!bellcore!decvax!mcnc!unccvax!wgivax!mo
From: mo@wgivax.UUCP (Michael O'Shea)
Newsgroups: net.unix-wizards
Subject: Re: flock(2) on 4.2bsd
Message-ID: <156@wgivax.UUCP>
Date: Thu, 27-Mar-86 09:16:39 EST
Article-I.D.: wgivax.156
Posted: Thu Mar 27 09:16:39 1986
Date-Received: Mon, 31-Mar-86 23:58:47 EST
References: <2721@sdcrdcf.UUCP>
Lines: 64

>I am using the LOCK_EX (exclusive lock) operation on a file.
>The manual states that if a lock already exists, the call should
>block until lock is available.  But in testing, 
>if the lock exits, flock returns with an error, EWOULDBLOCK.
>
>It behaves as if I specified (LOCK_EX|LOCK_NB) for the operation,
>the LOCK_NB being no blocking, but I have NOT!
>
>Am I doing something wrong?

You don't give an extract from the code which is doing this, but
assuming you are doing it as the manual states, it should work fine.

I use the following type of construct in several pieces of code on
a Vax 11/780 running vanilla 4.2, and have never had any problems
with them acting as expected (program waits for 5 minutes to get
lock; if it can't get lock, prints appropriate message and exits):


#include <setjmp.h>
#include <signal.h>

jmp_buf lockjump;

xyz()
{
	int  alarmtrap();
	FILE *fopen();

	/* open file */
	if((file = fopen(NAME,"a+")) == NULL)
	{
		perror(NAME);
		exit(-1);
	}

	/* wait to get lock for 5 minutes, then exit */

	signal(SIGALRM, alarmtrap);
	alarm(300);

	if(setjmp(lockjump) == 0)
	{
		if(flock(file->_file,LOCK_EX))
		{
			fprintf(stderr,"ERROR LOCKING %s\n",NAME);
			exit(-1);
		}
		alarm(0);
	}
	else
	{
		fprintf(stderr,"\n\n\n\n");
		fprintf(stderr,"Please try at a later time, ");
		fprintf(stderr,"someone else is hogging program right now!\n");
		fprintf(stderr,"\n\n\n\n");
		exit(1);
	}
}

alarmtrap() {
	alarm(0);
	longjmp(lockjump,1);
}