re: iphlpapi: Don't rely on the HAVE_XXX macros having a numeric values.

[Dan Kegel]

-#elif defined(HAVE_SYS_SYSCTL_H) && defined(IPCTL_STATS) &&
(HAVE_STRUCT_IPSTAT_IPS_TOTAL || HAVE_STRUCT_IP_STATS_IPS_TOTAL)
+#elif defined(HAVE_SYS_SYSCTL_H) && defined(IPCTL_STATS) &&
(defined(HAVE_STRUCT_IPSTAT_IPS_TOTAL) ||
defined(HAVE_STRUCT_IP_STATS_IPS_TOTAL))

Isn't this a no-op?  I thought that an undefined symbol was
treated as false by the preprocessor.

When I compile

#if defined(HAVE_SYS_SYSCTL_H) && defined(IPCTL_STATS) &&
(HAVE_STRUCT_IPSTAT_IPS_TOTAL || HAVE_STRUCT_IP_STATS_IPS_TOTAL)
defined
#else
not defined
#endif

with gcc -E or cl /E, I get

not defined

iphlpapi: Don't rely on the HAVE_XXX macros having a numeric values.

[Francois Gouget]

On Fri, 8 Jun 2012, Dan Kegel wrote:
> -#elif defined(HAVE_SYS_SYSCTL_H) && defined(IPCTL_STATS) &&
> (HAVE_STRUCT_IPSTAT_IPS_TOTAL || HAVE_STRUCT_IP_STATS_IPS_TOTAL)
> +#elif defined(HAVE_SYS_SYSCTL_H) && defined(IPCTL_STATS) &&
> (defined(HAVE_STRUCT_IPSTAT_IPS_TOTAL) ||
> defined(HAVE_STRUCT_IP_STATS_IPS_TOTAL))
>
> Isn't this a no-op?  I thought that an undefined symbol was
> treated as false by the preprocessor.

Actually you may be right. But then this is the only place in Wine where 
we rely on this and I see not reason to make this assumption. Even more 
so since we don't for the other macros in the same test which makes this 
locally inconsistent.

iphlpapi: Don't rely on the HAVE_XXX macros having a numeric values.

[Dan Kegel]

On Fri, Jun 8, 2012 at 11:51 AM, Francois Gouget  wrote:
> On Fri, 8 Jun 2012, Dan Kegel wrote:
> > Isn't this a no-op?  I thought that an undefined symbol was
> > treated as false by the preprocessor.
>
> Actually you may be right. But then this is the only place in Wine where
> we rely on this and I see not reason to make this assumption. Even more
> so since we don't for the other macros in the same test which makes this
> locally inconsistent.

It feels funny to avoid using a basic feature of C...
maybe we should remove the extra defined() calls.

iphlpapi: Don't rely on the HAVE_XXX macros having a numeric values.

[Alexandre Julliard]

Dan Kegel  writes:
> On Fri, Jun 8, 2012 at 11:51 AM, Francois Gouget  wrote:
> > On Fri, 8 Jun 2012, Dan Kegel wrote:
> > > Isn't this a no-op?  I thought that an undefined symbol was
> > > treated as false by the preprocessor.
> >
> > Actually you may be right. But then this is the only place in Wine where
> > we rely on this and I see not reason to make this assumption. Even more
> > so since we don't for the other macros in the same test which makes this
> > locally inconsistent.
>
> It feels funny to avoid using a basic feature of C...
> maybe we should remove the extra defined() calls.

It's common practice to avoid it, to make things like -Wundef more useful.

iphlpapi: Don't rely on the HAVE_XXX macros having a numeric values.

[Dan Kegel]

Alexandre wrote:
> It's common practice to avoid it, to make things like -Wundef more useful.

Good point.

Reading http://ozlabs.org/~rusty/index.cgi/tech/2008-01-04.html makes
me sad.  -Wundef would be a lot more useful if autoconf defined symbols
to zero instead of undefining them; then we could get rid of all the defined()
calls, and the compiler could tell us when we'd misspelled a HAVE_XXX.
 Oh, well.
- Dan