Re: Gcc 3.0, printf() and exim-3.30

[Nix <nix at esperi dot demon dot co dot uk>]

On Wed, 20 Jun 2001, Zack Weinberg spake:
> On Wed, Jun 20, 2001 at 11:45:18PM +0100, Nix wrote:
>> Nope; printf is not a macro, and GCC does not define printf :)
> 
> But GNU libc does, when used with GCC in optimizing mode.  The C
> standard permits the library to define any standard function as a
> macro, provided that there is a real function as well.

I checked for this case, and couldn't find any sign of a printf() macro.

(I plead hay-fever induced semi-blindness and simple stupidity, m'lud.)

> The C standard also says
> 
> 	If there are sequences of preprocessing tokens within the list
> 	of arguments that would otherwise act as preprocessing
> 	directives, the behavior is undefined.
> 
> (C99 6.10.3 paragraph 11).  This is just the case you have tripped
> over.  The error message means exactly what it says - you cannot use
> any directive in text which is read as the arguments of a macro.

I noticed that, but I couldn't convince GCC to give me a printf() macro,
and I missed the definition in bits/stdio.h somehow.

Of course the failure to replicate is because my try-by-default
optimization flag is -Os :(

> You can #undef printf immediately after #include <stdio.h>, or write
> 
> 	(printf)(" -d ..."
> 		 ...
> 		);
> 
> which will suppress macro expansion for that use only.

(Random aside: This is less portable than #undef, of course, but since
the macro is only defined if __GNUC__, this is not a problem.)

>                                                          There doesn't
> seem to be a way to prevent glibc from defining all its stupid macros.

We need a feature macro _LEAVE_OPTIMIZATION_TO_THE_COMPILER...

>> I think your #if itself is triggering a bug in GCC; a workaround may be
>> to say
>> 
>> #if !defined(NOPOSIX)
> 
> I don't think that has anything to do with it.

Neither do I, now.

>> #if #else
>> 
>> which is obvious nonsense; but it doesn't seem to me that this should be
>> triggered here.
> 
> #if #else is actually acceptable to GCC, it counts as a use of the
> deprecated "assertions" extension.

Whenever I manage to erase assertions from my mind, they come back and
bite me. I tried to forget about digraphs recently too, only to find
them attacking in turn...

-- 
`This will immediately become a flamewar.' --- Mark Mitchell