Date: Mon, 2 Jul 2001 04:41:47 +1000 (EST) From: Bruce Evans <bde@zeta.org.au> To: Mike Barcroft <mike@q9media.com> Cc: audit@FreeBSD.org Subject: Re: src/bin/ed patch Message-ID: <Pine.BSF.4.21.0107020417590.33742-100000@besplex.bde.org> In-Reply-To: <200107011506.f61F6P904227@coffee.q9media.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, 1 Jul 2001, Mike Barcroft wrote:
> Bruce Evans <bde@zeta.org.au> writes:
> > On Fri, 29 Jun 2001, Mike Barcroft wrote:
> > > o Since the rcsids are generating warnings, switch to __RCSID().
> >
> > Please don't. obrien and I are trying to decide whether ids (initially
> > in libc) should be changed to use __RCSID() when they are cleaned up.
> > I think we are unlikely to use it. One problem is that it is also used
> > in other BSDs, so it can't be used to hide messy ifdefs to control the
> > visibility of the id string in a system-dependent way.
>
> Why not just add a switch in /etc/{defaults/,}make.conf to allow one
> to control the visibility of RCSIDs? What concern is it of ours if the
> other BSDs embed RCSIDs unconditionally?
It would be silly to have a switch to fully control FreeBSD rcsids if
OtherBSD rcsids get embedded (in object files) unconditionally.
> In any event, __RCSID() doesn't appear ready for general use. An
> updated version of my patch is available at the end of this message
> and also at: http://testbed.q9media.net/freebsd/ed.20010701.patch
Thanks. I like most of this version.
> Index: ed/main.c
> ===================================================================
> RCS file: /home/ncvs/src/bin/ed/main.c,v
> retrieving revision 1.20
> diff -u -r1.20 main.c
> --- ed/main.c 2001/06/28 22:06:27 1.20
> +++ ed/main.c 2001/07/01 14:29:06
> @@ -27,18 +27,10 @@
> */
>
> #ifndef lint
> -static char * const copyright =
> -"@(#) Copyright (c) 1993 Andrew Moore, Talke Studio. \n\
> - All rights reserved.\n";
> -#endif /* not lint */
> -
> -#ifndef lint
> -#if 0
> -static char * const rcsid = "@(#)main.c,v 1.1 1994/02/01 00:34:42 alm Exp";
> -#else
> -static char * const rcsid =
> +static const char copyright[] =
> + "@(#) Copyright (c) 1993 Andrew Moore, Talke Studio. All rights reserved.";
> +static const char rcsid[] =
> "$FreeBSD: src/bin/ed/main.c,v 1.20 2001/06/28 22:06:27 dd Exp $";
> -#endif
> #endif /* not lint */
>
> /*
Why change the original copyright string (source and oupput format)?
The copyright normally goes in a separate "#ifdef lint".
> ...
> @@ -1385,12 +1377,13 @@
> {
> char *hup = NULL; /* hup filename */
> char *s;
> + char ed_hup[] = "ed.hup";
> int n;
>
> if (!sigactive)
> quit(1);
> sigflags &= ~(1 << (signo - 1));
> - if (addr_last && write_file("ed.hup", "w", 1, addr_last) < 0 &&
> + if (addr_last && write_file(ed_hup, "w", 1, addr_last) < 0 &&
> (s = getenv("HOME")) != NULL &&
> (n = strlen(s)) + 8 <= PATH_MAX && /* "ed.hup" + '/' */
> (hup = (char *) malloc(n + 10)) != NULL) {
The correct fix is to declare write_file()'s first arg as const.
BTW, I think we're only seeing the tip of the iceberg for constant
poisoning of function args. -Wwrite-strings triggers a few warnings
for "char *" args, but there are zillions of interfaces that take
pointers to data that is not changed by the interface yet not declared
const. Change a top-level interface to declare things as const and
the poison will spread to lower levels.
Bruce
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-audit" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.4.21.0107020417590.33742-100000>