Re: [Qemu-devel] [PATCH 2/2] target-ppc: Cast ssize_t to size_t before printing with %zx

[Eric Blake <eblake@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 2835 bytes --]

On 12/23/2014 03:22 PM, Peter Maydell wrote:
> The mingw32 compiler complains about trying to print variables of type
> ssize_t with the %z format string specifier. Since we're printing it
> as unsigned hex anyway, cast to size_t to silence the warning.

Hmm.  I wonder if mingw headers are mixing 'long' and 'unsigned int' (or
maybe 'unsigned long' and 'int') for the underlying definitions of
'ssize_t' vs. 'size_t'; both are 32-bits on the platform, but a
difference in underlying type would definitely explain why you get a
compiler warning if you use %zd with size_t or %zu with ssize_t.  On the
other hand, the warning I've most often seen is that native windows
completely lacks %z support, so gcc is smart enough to warn that ANY use
of %z is not going to work, for a function marked
__attribute__((__printf__)) on mingw.

> 
> Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
> ---
> I suspect that this is a compiler bug, but this is the only instance
> in the codebase so I'm prepared to work around it to get to a point
> where we can turn on warnings-as-errors for w32, because some of the
> w32-specific errors really are flagging up issues we need to fix.

With new enough mingw headers, the alternative solution is to guarantee
that #__USE_MINGW_ANSI_STDIO is defined before including standard
headers, then %z, %j, %lld, and other required constructs magically work
(because you are no longer using the native printf, but the fixed mingw
shim).  Once you do that, you want to use
__attribute__((__gnu_printf__)) instead of __attribute__((__printf__))
in order to teach gcc that yes, your use of %z is really going to work.

For some more background reading:
http://sourceforge.net/p/mingw-w64/wiki2/gnu%20printf/
https://lists.gnu.org/archive/html/bug-gnulib/2014-12/msg00029.html
> ---
>  hw/ppc/spapr.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
> index 08401e0..9aaa800 100644
> --- a/hw/ppc/spapr.c
> +++ b/hw/ppc/spapr.c
> @@ -1438,7 +1438,7 @@ static void ppc_spapr_init(MachineState *machine)
>      }
>      if (spapr->rtas_size > RTAS_MAX_SIZE) {
>          hw_error("RTAS too big ! 0x%zx bytes (max is 0x%x)\n",
> -                 spapr->rtas_size, RTAS_MAX_SIZE);
> +                 (size_t)spapr->rtas_size, RTAS_MAX_SIZE);

What's the actual compiler warning you got? This feels like it won't
work if the real problem is the compiler telling you that %z is unknown;
conversely, if it is a mismatch between 'size_t' vs. 'ssize_t', we
probably ought to report it to the mingw folks to fix their environment
to use the same underlying type for signed vs. unsigned sizes.

-- 
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: 604 bytes --]