"incompatible implicit declaration" warnings with GCC 4.2.1

"incompatible implicit declaration" warnings with GCC 4.2.1

[Juanma Barranquero]

  C:\emacs> gcc --version
  gcc (GCC) 4.2.1-sjlj (mingw32-2)
  Copyright (C) 2007 Free Software Foundation, Inc.

Compilation produces these warnings. They don't happen with gcc 3.4.5 (MinGW):

  gcc -I. -DWIN32_LEAN_AND_MEAN -D_WIN32_WINNT=0x0400 -D_X86_=1 -c
-gstabs+ -g3 -mno-cygwin -mtune=pentium4 -O2  -Di386 -D_CRTAPI1=_cdecl
-Demacs=1 -DWINDOWSNT -DDOS_NT -DHAVE_CONFIG_H -I../nt/inc -D_UCHAR_T
-DHAVE_NTGUI=1 -DUSE_CRT_DLL=1 -DPURESIZE=5000000 -o
oo-spd/i386/w32fns.o w32fns.c
  w32fns.c: In function 'gamma_correct':
  w32fns.c:1241: warning: incompatible implicit declaration of
built-in function 'pow'
  w32fns.c: In function 'w32_abort':
  w32fns.c:9127: warning: incompatible implicit declaration of
built-in function 'abort'

             Juanma

Re: "incompatible implicit declaration" warnings with GCC 4.2.1

[Eli Zaretskii]

> Date: Mon, 22 Oct 2007 17:24:49 +0200
> From: "Juanma Barranquero" <lekktu@gmail.com>
> 
>   gcc -I. -DWIN32_LEAN_AND_MEAN -D_WIN32_WINNT=0x0400 -D_X86_=1 -c
> -gstabs+ -g3 -mno-cygwin -mtune=pentium4 -O2  -Di386 -D_CRTAPI1=_cdecl
> -Demacs=1 -DWINDOWSNT -DDOS_NT -DHAVE_CONFIG_H -I../nt/inc -D_UCHAR_T
> -DHAVE_NTGUI=1 -DUSE_CRT_DLL=1 -DPURESIZE=5000000 -o
> oo-spd/i386/w32fns.o w32fns.c
>   w32fns.c: In function 'gamma_correct':
>   w32fns.c:1241: warning: incompatible implicit declaration of
> built-in function 'pow'
>   w32fns.c: In function 'w32_abort':
>   w32fns.c:9127: warning: incompatible implicit declaration of
> built-in function 'abort'

Does it help to include <math.h> and <stdlib.h> in w32fns.c?

Re: "incompatible implicit declaration" warnings with GCC 4.2.1

[Juanma Barranquero]

On 10/22/07, Eli Zaretskii <eliz@gnu.org> wrote:

> Does it help to include <math.h> and <stdlib.h> in w32fns.c?

<math.h> helps for the `abs' warning.

<stdlib.h> does not help for the `abort' warning. w32fns.c has a dozen
abort() calls, and then

 #undef abort

followed by w32_abort, which has the `abort' call that causes the
warning. Why so?

             Juanma

Re: "incompatible implicit declaration" warnings with GCC 4.2.1

[Jason Rumney]

Juanma Barranquero wrote:
> <stdlib.h> does not help for the `abort' warning. w32fns.c has a dozen
> abort() calls, and then
>
>  #undef abort
>
> followed by w32_abort, which has the `abort' call that causes the
> warning. Why so?
>   

stdlib.h needs to be included before whichever header it is that contains

#define abort w32_abort

Re: "incompatible implicit declaration" warnings with GCC 4.2.1

[Juanma Barranquero]

On 10/22/07, Jason Rumney <jasonr@gnu.org> wrote:

> stdlib.h needs to be included before whichever header it is that contains
>
> #define abort w32_abort

That's src/s/ms-w32.h, which is included from config.h. Including
<stdlib.h> before config.h produces this new error:

  w32fns.c:9105: warning: conflicting types for 'w32_abort'
  w32fns.c:496: warning: previous implicit declaration of 'w32_abort' was here

I'd really like to know why this include shuffling is suddenly needed
with GCC 4.2.1...

             Juanma

Re: "incompatible implicit declaration" warnings with GCC 4.2.1

[Jason Rumney]

Juanma Barranquero wrote:
> That's src/s/ms-w32.h, which is included from config.h. Including
> <stdlib.h> before config.h produces this new error:
>
>   w32fns.c:9105: warning: conflicting types for 'w32_abort'
>   w32fns.c:496: warning: previous implicit declaration of 'w32_abort' was here
>
> I'd really like to know why this include shuffling is suddenly needed
> with GCC 4.2.1...
>   

I think the original warning you reported is a new warning.

Try the latest trunk (without any manual changes to include files).

Re: "incompatible implicit declaration" warnings with GCC 4.2.1

[Juanma Barranquero]

On 10/23/07, Jason Rumney <jasonr@gnu.org> wrote:

> I think the original warning you reported is a new warning.

Aha.

> Try the latest trunk (without any manual changes to include files).

The `abort' warning has been fixed. The `pow' one is still there (you
haven't included <math.h> in w32fns.c).

             Juanma

Re: "incompatible implicit declaration" warnings with GCC 4.2.1

[Juanma Barranquero]

On 10/23/07, Juanma Barranquero <lekktu@gmail.com> wrote:

> The `abort' warning has been fixed. The `pow' one is still there (you
> haven't included <math.h> in w32fns.c).

Now it is fixed. Thanks.

             Juanma

Re: "incompatible implicit declaration" warnings with GCC 4.2.1

[Eli Zaretskii]

> Date: Mon, 22 Oct 2007 23:55:29 +0200
> From: "Juanma Barranquero" <lekktu@gmail.com>
> Cc: "Eli Zaretskii" <eliz@gnu.org>, emacs-devel@gnu.org
> 
> I'd really like to know why this include shuffling is suddenly needed
> with GCC 4.2.1...

Evidently, GCC 4.2.1 now has more functions implemented as built-ins,
among them `pow' and `abort'.  This means that GCC knows their
prototypes even without seeing the header file in which they are
supposed to be declared.  Thus, it can now issue warnings where it
previously couldn't.

Re: "incompatible implicit declaration" warnings with GCC 4.2.1

[Juanma Barranquero]

On 10/23/07, Eli Zaretskii <eliz@gnu.org> wrote:

> Evidently, GCC 4.2.1 now has more functions implemented as built-ins,
> among them `pow' and `abort'.

OK, I understand. Thanks.

             Juanma

Re: "incompatible implicit declaration" warnings with GCC 4.2.1

[Juanma Barranquero]

On 10/23/07, Jason Rumney <jasonr@gnu.org> wrote:

> I think the original warning you reported is a new warning.
>
> Try the latest trunk (without any manual changes to include files).

BTW, the fixes should be backported to EMACS_22_BASE.

             Juanma

Re: "incompatible implicit declaration" warnings with GCC 4.2.1

[Jason Rumney]

Juanma Barranquero wrote:
> BTW, the fixes should be backported to EMACS_22_BASE.
>   

They are not fixing a bug, only getting rid of compiler warnings, and
the changes were substantial enough to carry a risk of breaking
something, so I'd rather not.

Re: "incompatible implicit declaration" warnings with GCC 4.2.1

[Juanma Barranquero]

On 10/24/07, Jason Rumney <jasonr@gnu.org> wrote:

> They are not fixing a bug, only getting rid of compiler warnings, and
> the changes were substantial enough to carry a risk of breaking
> something, so I'd rather not.

That's true for the `abort' warning. The one for `pow' can be cured
just by including <math.h>, I think.

             Juanma