Re: [Qemu-devel] [PATCH 03/11] configure: Fix compiler warning in config.log (integer from pointer)

From: Eric Blake <eblake@redhat.com>
To: Peter Maydell <peter.maydell@linaro.org>
Cc: qemu-trivial@nongnu.org, Stefan Weil <sw@weilnetz.de>,
	qemu-devel@nongnu.org
Subject: Re: [Qemu-devel] [PATCH 03/11] configure: Fix compiler warning in config.log (integer from pointer)
Date: Sat, 17 Dec 2011 07:18:55 -0700	[thread overview]
Message-ID: <4EECA4CF.5030407@redhat.com> (raw)
In-Reply-To: <CAFEAcA-Dpy1a-hsN6oxPn6M=3LTwtfpR2c8J2Gz_a7-TfYQv-A@mail.gmail.com>

[-- Attachment #1: Type: text/plain, Size: 1086 bytes --]

On 12/17/2011 07:09 AM, Peter Maydell wrote:
>> Typically NULL is defined to be ((void *)0 for C
>> (that's the reason why I used a type cast, too). Only for
>> C++ it is defined without a type cast.
> 
> The C standard permits plain "0" as a definition of NULL.

POSIX concurs with the C standard that a plain "0" can be used for
representing a null pointer, but requires that the macro NULL be defined
as ((void*)0) as a tighter requirement than C.

http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_chap03.html#tag_03_244
http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/stddef.h.html#tag_13_47

> We use plain 0 for a null pointer constant in various
> existing tests in configure; this is simpler and avoids
> dragging in stddef.h. I would prefer that.

I agree that configure tests should be as small as possible while still
being correct, and use of a 0 constant without a cast, rather than NULL,
is appropriate in that context.

-- 
Eric Blake   eblake@redhat.com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 620 bytes --]