Date: Sun, 19 Sep 2004 15:10:22 GMT From: Ruslan Ermilov <ru@FreeBSD.org> To: freebsd-bugs@FreeBSD.org Subject: Re: bin/71855: making kdump WARNS=6 clean Message-ID: <200409191510.i8JFAM28032870@freefall.freebsd.org>
next in thread | raw e-mail | index | archive | help
The following reply was made to PR bin/71855; it has been noted by GNATS.
From: Ruslan Ermilov <ru@FreeBSD.org>
To: Divacky Roman <xdivac02@stud.fit.vutbr.cz>
Cc: bug-followup@FreeBSD.org
Subject: Re: bin/71855: making kdump WARNS=6 clean
Date: Sun, 19 Sep 2004 18:01:29 +0300
On Sat, Sep 18, 2004 at 11:03:57AM +0200, Divacky Roman wrote:
>
> >Description:
> This makes kdump partially WARNS=6 clean... its also necessary to resolve
> conflict between pccard/cardinfo.h and sys/memrange.h MDF_ACTIVE define...
>
> >How-To-Repeat:
> apply this patch and do "something" with that MDF_ACTIVE define
> >Fix:
> diff -urN /usr/src/usr.bin/kdump/Makefile kdump/Makefile
> --- /usr/src/usr.bin/kdump/Makefile Tue Jul 23 11:15:45 2002
> +++ /usr/src/usr.bin/kdump/Makefile Sat Sep 18 10:52:49 2004
> @@ -6,6 +6,7 @@
> PROG= kdump
> SRCS= kdump.c ioctl.c subr.c
> CFLAGS+= -I${.CURDIR}/../ktrace -I${.CURDIR}/../..
> +WARNS?= 6
>
> CLEANFILES= ioctl.c
>
This is unsafe, please see below.
> diff -urN /usr/src/usr.bin/kdump/kdump.c kdump/kdump.c
> --- /usr/src/usr.bin/kdump/kdump.c Thu Mar 25 13:33:55 2004
> +++ /usr/src/usr.bin/kdump/kdump.c Sat Sep 18 10:52:49 2004
> @@ -79,6 +79,7 @@
> void ktrcsw(struct ktr_csw *);
> void ktruser(int, unsigned char *);
> void usage(void);
> +extern const char *ioctlname(register_t);
>
> int timestamp, decimal, fancy = 1, tail, maxdata;
> const char *tracefile = DEF_TRACEFILE;
> @@ -285,7 +286,6 @@
> {
> int narg = ktr->ktr_narg;
> register_t *ip;
> - const char *ioctlname(u_long);
>
> if (ktr->ktr_code >= nsyscalls || ktr->ktr_code < 0)
> (void)printf("[%d]", ktr->ktr_code);
>
I've committed a variation of this, thanks!
> --- /usr/src/sys/sys/diskpc98.h Wed Apr 7 06:19:49 2004
> +++ /usr/src/sys/sys/diskpc98.h Sat Sep 18 10:57:10 2004
> @@ -36,9 +36,11 @@
> #include <sys/ioccom.h>
>
> #define DOSBBSECTOR 0 /* DOS boot block relative sector number */
> +#ifdef PC98
> #define DOSPARTOFF 0
> #define NDOSPART 16
> #define DOSPTYP_386BSD 0x94 /* 386BSD partition type */
> +#endif
>
> struct pc98_partition {
> unsigned char dp_mid;
>
This won't be committed, as it's bogus. It's bogus because these
defines are correct, and the problem you see is elsewhere, not in
this header.
A kdump(1) is special in that it tries to show as much ioctl names
as possible. For this, it dynamically builds a list of all ioctls
by scanning the system headers and looking for "_IO" defines. As
you never know if there will be a conflict in names or not (the
exact list of headers used is unknown and unpredictable), bumping
a WARNS level to 6 will likely cause breakages in the future.
While "it would be nice" to have kdump(1) compile without warnings,
guaranteeing this in a long term is troublesome.
Cheers,
--
Ruslan Ermilov
ru@FreeBSD.org
FreeBSD committer
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200409191510.i8JFAM28032870>