Xref: utzoo alt.sys.sun:4084 comp.unix.wizards:25832
Path: utzoo!utgpu!news-server.csri.toronto.edu!bonnie.concordia.ca!thunder.mcrcim.mcgill.edu!mouse
From: mouse@thunder.mcrcim.mcgill.edu (der Mouse)
Newsgroups: alt.sys.sun,comp.unix.wizards
Subject: Re: Maybe I am missing something....but why doesn't this work?
Keywords: TCP SOCKETS
Message-ID: <1991Jun1.193159.24881@thunder.mcrcim.mcgill.edu>
Date: 1 Jun 91 19:31:59 GMT
References: <1991May29.145535.23801@bradley.bradley.edu>
Organization: McGill Research Centre for Intelligent Machines
Lines: 33

In article <1991May29.145535.23801@bradley.bradley.edu>, brad@bradley.bradley.edu (Bradley E. Smith) writes:

> I am not sure why but the following program has problems.....

[program deleted - problem with bind()]

> HARDWARE USED:

> SUN 470 SunOS 4.1.1		- NO
> SUN SLC SunOS 4.1.1		- NO
> IBM RT BSD OS			- YES
> AT&T 3B2 WIN TCP 3.2		- YES
> AT&T 386 WIN TCP 3.0		- YES

What you need to do is zero the supposedly-unused area of the struct
sockaddr_in before calling bind().  As far as I know SunOS is the only
system that imposes this requirement, but the range of systems I have
actually checked this on is fairly small - 4.3 BSD is the only one that
I'm sure doesn't require it.

Try doing a bzero((char *)&backupserver,sizeof(backupserver)) before
storing anything into it.  That generally does it for me.

The reason calling x() changes things is that backupserver (which is an
auto variable) is allocated at a slightly different place on the stack,
so the unused portion contains different stack trash.  The working way
places it so the unused portion contains zeros.

					der Mouse

			old: mcgill-vision!mouse
			new: mouse@larry.mcrcim.mcgill.edu