NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/lang/c-part2/14088

Internet Message Format | 1996-08-05 | 2.7 KB

Path: anvil.ugrad.cs.ubc.ca!not-for-mail
From: [e-mail] (Kazimir Kylheku)
Newsgroups: comp.lang.c
Subject: Re: ANSI C and POSIX (was Re: C/C++ knocks the crap out of Ada)
Date: 11 Apr 1996 11:27:04 -0700
Organization: Computer Science, University of B.C., Vancouver, B.C., Canada
Message-ID: <[e-mail]>
References: <[e-mail]> <[e-mail]> <dewar.829096975@schonberg> <[e-mail]>
NNTP-Posting-Host: anvil.ugrad.cs.ubc.ca

In article <[e-mail]>,
Gareth Rees <[e-mail]> wrote:
 >Robert Dewar <[e-mail]> wrote:
 >> Highly imprecise thinking here I fear. There is no practical way for
 >> any implementation to do the check you mention (this = value in 3rd
 >> argument that is greater than the available buffer size).
 >
 >But the *programmer* can do the check while writing the program!  The
 >programmer must have declared or allocated a buffer of some size.
 >Consider: if I wrote code like
 >
 >    #include<stdio.h>
 >
 >    int main () {
 >        char buf[80];
 >        /* Input lines are always 78 characters or less, so the limit of
 >           100 can never cause a buffer overflow */
 >        while (fgets(buf,100,stdin))
 >            puts(buf);
 >        return 0;
 >    }
 >
 >you would quite rightly point out that this is gratuitously broken and
 >that it would cost me nothing (and gain a lot) to change `100' to `80'.
 >It's the same with the buffer size in `read'.  Why make the byte count
 >larger than the buffer size when you don't have to?

Because only careful rationality can convince you not to it, not the vague
POSIX specification for the read function.

What about the select() call? Would you assume that the timeval structure is
modified by select() if every system you ever used select() with left the
structure alone?  This is not as clear-cut as the read() case.

On some systems, the parameter is even declared as ``const struct timeval
*timeout'', which is a promise that it won't me modified. Yet under Linux it
_is_ modified.

It's safer to write code that assumes the structure will be modified, but I
would not have even thought of it until coming to Linux.

To this date, I have personally ``fixed'' two or three programs because of this
difference, the most recent of which was the old BSD game ``hunt''.

It comes with Slackware, compiled for AF_UNIX sockets---unlike me who was
desperate for a good match with some buddies, they apparently just gave up
getting it work properly with AF_INET. Due to the select() difference, the
client would not wait enough for a server reply over a slow network, since the
game would use the same struct timeval each time select() was called without
re-initializing its contents.
--