Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 13 Feb 2006 00:23:10 -0700 (MST)
From:      "M. Warner Losh" <imp@bsdimp.com>
To:        ertr1013@student.uu.se
Cc:        des@des.no, stefanf@freebsd.org, arch@freebsd.org
Subject:   Re: [releng_6 tinderbox] failure on sparc64/sparc64
Message-ID:  <20060213.002310.125802352.imp@bsdimp.com>
In-Reply-To: <20060205220211.GA5151@falcon.midgard.homeip.net>
References:  <20060205084813.GN21806@wombat.fafoe.narf.at> <867j89n71d.fsf@xps.des.no> <20060205220211.GA5151@falcon.midgard.homeip.net>

next in thread | previous in thread | raw e-mail | index | archive | help
In message: <20060205220211.GA5151@falcon.midgard.homeip.net>
            Erik Trulsson <ertr1013@student.uu.se> writes:
: On Sun, Feb 05, 2006 at 04:45:34PM +0100, Dag-Erling Sm=F8rgrav wrote=
:
: > Stefan Farfeleder <stefanf@FreeBSD.org> writes:
: > > On Sat, Feb 04, 2006 at 03:58:56PM -0700, M. Warner Losh wrote:
: > > > This driver wants to access these structures as arrays of uint3=
2_t.
: > > > It used to cast directly, but that isn't allowed.  So, I've pas=
sed
: > > > the cast through a (void *).  Is that allowed?  Eg:
: > > > =

: > > > 	struct foo foo;
: > > > 	((uint32_t *)(void *)&foo)[3] =3D 12;
: > > > =

: > > > is that OK?
: > > I'm afraid that only silences the warning without solving the und=
erlying
: > > problem.  I don't think there's a Standard conforming way to trea=
t a
: > > struct foo as an uint32_t array.
: > =

: > A union should do the trick.
: =

: No, it will not.  If you have a struct foo and try to access it as an=
 array
: of int, the program will have unspecified (and maybe even undefined)
: behaviour.  It does not matter if you do it with a union or by castin=
g
: pointers.
: =

: In general, if you have an object of type X, then the only ways it ca=
n be
: accessed is either as an object of type X, or as an array of [unsigne=
d]
: char.  =


So the proper fix for the above code is:

	struct foo foo;
	uint32_t value[sizeof(foo) / sizeof(uint32_t)];

	memcpy(value, &foo);
	// write out value one 32-bit word at a time

Is that right?  Or at least 'proper' here means defined.

Warner
	=


	=




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20060213.002310.125802352.imp>