Date: Mon, 16 Dec 1996 21:25:18 -0700 (MST) From: Marc Slemko <marcs@znep.com> To: Jason Downs <downsj@teeny.org> Cc: freebsd-chat@freebsd.org Subject: snprintf vs. strncpy (was: Re: crontab security hole) Message-ID: <Pine.BSF.3.95.961216210309.10949D-100000@alive.ampr.ab.ca> In-Reply-To: <199612170316.TAA22717@threadway.teeny.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, 16 Dec 1996, Jason Downs wrote:
> It's bad programming practice to blindly use snprintf() without considering
> your options. (Just as it's bad to use strncpy() without terminating the
> string yourself, or call strncat() with the wrong length.)
I think this discussion is far too sane. Let's see who can make up the
most meaningless numbers.
marcs@alive:/tmp/str$ ls
snprintf strncpy strncpy-2.c
snprintf.c strncpy-2 strncpy.c
marcs@alive:/tmp/str$ cat snprintf.c
#include <stdio.h>
int main () {
char s[] = "jjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjj";
char t[200];
int c;
for (c = 0; c < 500000; c++) {
snprintf(t, sizeof(t), "%s", s);
}
}
marcs@alive:/tmp/str$ cat strncpy.c
#include <string.h>
int main () {
char s[] = "jjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjj";
char t[200];
int c;
for (c = 0; c < 500000; c++) {
strncpy(t, s, sizeof(t));
t[sizeof(t)-1] = '\0';
}
}
marcs@alive:/tmp/str$ cat strncpy-2.c
#include <string.h>
int main () {
char s[] = "jjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjj";
char t[200];
int c;
for (c = 0; c < 500000; c++) {
strncpy(t, s, sizeof(t));
}
}
marcs@alive:/tmp/str$ time ./snprintf ; time ./strncpy ; time ./strncpy-2
6.86 real 5.65 user 0.02 sys
11.50 real 9.85 user 0.00 sys
9.54 real 9.15 user 0.01 sys
marcs@alive:/tmp/str$
Hmm. Seems like snprintf is nearly twice as fast in this case.
What was that about considering your options? Considering that we
already have sprintf in Vixie's code...
(No, I don't think snprintf is better than strncpy all the time, or
perhaps even most of the time but I don't see any reason to complain about
using snprintf instead of strncpy. It is bad programming practice to do
_ANYTHING_ without considering your options. What I do think is that we
need a freebsd-silly-numbers mailing list... I'm sure someone can reverse
these numbers for me.)
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.3.95.961216210309.10949D-100000>