[Qemu-devel] error compiling hw/sh7750.c

[Qemu-devel] error compiling hw/sh7750.c

[Ben Taylor]

I'm getting this error compiling hw/sh7750.c

I believe the problem comes from the include in hw/sh_intc.h
with the macro

#define _INTC_ARRAY(a) a, sizeof(a)/sizeof(*a)

in line 587 of hw/7750.c, I see

        sh_intc_register_sources(&s->intc,
                                 _INTC_ARRAY(vectors_irlm),
                                 _INTC_ARRAY(NULL));

So the macro turns the last _INTC_ARRAY(NULL) into

"NULL, sizeof(NULL)/sizeof(*NULL)

Since sizeof(*NULL) doesn't make any sense, the compiler barfs
on Solaris 10/x86_64.

gcc -Wall -O2 -g -fno-strict-aliasing -m64 -I. -I.. -I/export-1/bent/src/qemu/111107/qemu/target-sh4 -I/export-1/bent/src/qemu/111107/qemu -MMD -MP -DNEED_CPU_H -D_GNU_SOURCE -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE -I/export-1/bent/src/qemu/111107/qemu/fpu -I/usr/gnu/include   -I/export-1/bent/src/qemu/111107/qemu/slirp  -m64  -c -o sh7750.o /export-1/bent/src/qemu/111107/qemu/hw/sh7750.c
/export-1/bent/src/qemu/111107/qemu/hw/sh7750.c: In function `error_access':
/export-1/bent/src/qemu/111107/qemu/hw/sh7750.c:184: warning: unsigned int format, different type arg (arg 5)
/export-1/bent/src/qemu/111107/qemu/hw/sh7750.c: In function `ignore_access':
/export-1/bent/src/qemu/111107/qemu/hw/sh7750.c:190: warning: unsigned int format, different type arg (arg 5)
/export-1/bent/src/qemu/111107/qemu/hw/sh7750.c: In function `sh7750_init':
/export-1/bent/src/qemu/111107/qemu/hw/sh7750.c:576: error: invalid type argument of `unary *'
/export-1/bent/src/qemu/111107/qemu/hw/sh7750.c:589: error: invalid type argument of `unary *'
gmake[1]: *** [sh7750.o] Error 1
gmake[1]: Leaving directory `/export-1/bent/src/qemu/111107/qemu/sh4-softmmu'
gmake: *** [subdir-sh4-softmmu] Error 2

In this case, I've replaced _INTC_ARRAY(NULL) with "NULL, NULL" and it
succesfully compiles, but I don't have any sh emulation to do.

Ben

Re: [Qemu-devel] error compiling hw/sh7750.c

[Carlo Marcelo Arenas Belon]

On Sun, Nov 11, 2007 at 09:30:26AM -0500, Ben Taylor wrote:
> So the macro turns the last _INTC_ARRAY(NULL) into
> 
> "NULL, sizeof(NULL)/sizeof(*NULL)

in my 64bit linux using gcc-4.1.2 it becomes instead :

  ((void *)0), sizeof(((void *)0))/sizeof(*((void *)0))

what version of gcc (gcc -v) are you using in that Solaris 10 box? (HINT: not
all available versions of gcc compile qemu correctly) :

  http://www.opensolaris.org/os/project/qemu/host/gcc-failures/

Carlo

Re: [Qemu-devel] error compiling hw/sh7750.c

[Carlo Marcelo Arenas Belon]

On Sun, Nov 11, 2007 at 10:06:34AM -0600, Carlo Marcelo Arenas Belon wrote:
> On Sun, Nov 11, 2007 at 09:30:26AM -0500, Ben Taylor wrote:
> > So the macro turns the last _INTC_ARRAY(NULL) into
> > 
> > "NULL, sizeof(NULL)/sizeof(*NULL)
> 
> in my 64bit linux using gcc-4.1.2 it becomes instead :
> 
>   ((void *)0), sizeof(((void *)0))/sizeof(*((void *)0))
> 
> what version of gcc (gcc -v) are you using in that Solaris 10 box? (HINT: not
> all available versions of gcc compile qemu correctly) :
> 
>   http://www.opensolaris.org/os/project/qemu/host/gcc-failures/

That didn't sound as clear as it should, so I booted an OpenSolaris Nevada 70b
using the sun provided gcc :

$ uname -a
SunOS dell 5.11 snv_70b i86pc i386 i86pc
$ gcc -v
Reading specs from /usr/sfw/lib/gcc/i386-pc-solaris2.11/3.4.3/specs
Configured with: /builds2/sfwnv-gate/usr/src/cmd/gcc/gcc-3.4.3/configure
--prefix=/usr/sfw --with-as=/usr/sfw/bin/gas --with-gnu-as
--with-ld=/usr/ccs/bin/ld --without-gnu-ld --enable-languages=c,c++,f77,objc
--enable-shared
Thread model: posix
gcc version 3.4.3 (csl-sol210-3_4-20050802)

and run a test, the problem is in /usr/include/iso/stdio_iso.h that defined
NULL as an int and not a pointer as shown by :

fndef NULL
#if defined(_LP64)
#define NULL    0L
#else
#define NULL    0
#endif
#endif

contradicting (at least in spirit) the C99 (ISO/IEC 9899:TCS) standard that
says :

  "A pointer to void shall have the same representation and alignment 
   requirements as a pointer to a character type"

and therefore makes sizeof(*NULL) == 1 solving the problem to compile as it is
done if applying the following patch :

Carlo

---
--- qemu/hw/sh7750.c	Sun Oct  7 09:31:11 2007
+++ qemu/hw/sh7750.c	Sun Nov 11 16:13:13 2007
@@ -29,6 +29,9 @@
 #include "sh7750_regnames.h"
 #include "sh_intc.h"
 
+#undef NULL
+#define NULL ((void *)0)
+
 #define NB_DEVICES 4
 
 typedef struct SH7750State {

Re: [Qemu-devel] error compiling hw/sh7750.c

[Paul Brook]

> +#undef NULL
> +#define NULL ((void *)0)

Absolutely not.

Paul

Re: [Qemu-devel] error compiling hw/sh7750.c

[Carlo Marcelo Arenas Belon]

On Sun, Nov 11, 2007 at 05:24:06PM +0000, Paul Brook wrote:
> > +#undef NULL
> > +#define NULL ((void *)0)
> 
> Absolutely not.

this was not meant to be a final solution, or even be committed in qemu, just
a stop gap solution answering Ben so that he can get their build to complete 
and have a somehow equivalent definition of what the code is doing in linux
(where I presume was tested) until it can be fixed in a portable way.

Ben's solution was to replace _INTC_ARRAY(NULL) with "NULL, NULL" but it
should had been instead "NULL, 4" or "NULL, 8" depending on what ISA he is
compiling it for, doing the "#undef/#define" leads to that and, yes I agree
it is awful.

Carlo

Re: [Qemu-devel] error compiling hw/sh7750.c

[M. Warner Losh]

In message: <20071111172122.GA9387@tapir>
            Carlo Marcelo Arenas Belon <carenas@sajinet.com.pe> writes:
: On Sun, Nov 11, 2007 at 10:06:34AM -0600, Carlo Marcelo Arenas Belon wrote:
: > On Sun, Nov 11, 2007 at 09:30:26AM -0500, Ben Taylor wrote:
: > > So the macro turns the last _INTC_ARRAY(NULL) into
: > > 
: > > "NULL, sizeof(NULL)/sizeof(*NULL)
: > 
: > in my 64bit linux using gcc-4.1.2 it becomes instead :
: > 
: >   ((void *)0), sizeof(((void *)0))/sizeof(*((void *)0))
: > 
: > what version of gcc (gcc -v) are you using in that Solaris 10 box? (HINT: not
: > all available versions of gcc compile qemu correctly) :
: > 
: >   http://www.opensolaris.org/os/project/qemu/host/gcc-failures/
: 
: That didn't sound as clear as it should, so I booted an OpenSolaris Nevada 70b
: using the sun provided gcc :
: 
: $ uname -a
: SunOS dell 5.11 snv_70b i86pc i386 i86pc
: $ gcc -v
: Reading specs from /usr/sfw/lib/gcc/i386-pc-solaris2.11/3.4.3/specs
: Configured with: /builds2/sfwnv-gate/usr/src/cmd/gcc/gcc-3.4.3/configure
: --prefix=/usr/sfw --with-as=/usr/sfw/bin/gas --with-gnu-as
: --with-ld=/usr/ccs/bin/ld --without-gnu-ld --enable-languages=c,c++,f77,objc
: --enable-shared
: Thread model: posix
: gcc version 3.4.3 (csl-sol210-3_4-20050802)
: 
: and run a test, the problem is in /usr/include/iso/stdio_iso.h that defined
: NULL as an int and not a pointer as shown by :
: 
: fndef NULL
: #if defined(_LP64)
: #define NULL    0L
: #else
: #define NULL    0
: #endif
: #endif
: 
: contradicting (at least in spirit) the C99 (ISO/IEC 9899:TCS) standard that
: says :
: 
:   "A pointer to void shall have the same representation and alignment 
:    requirements as a pointer to a character type"

NULL isn't a pointer.  It is a integral constant, per section
6.3.2.3.3 of C99.  So *NULL is undefined.  If you want a constant
thats an actual pointer, you need to define one special.

       6.3.2.3  Pointers
...
       [#3] An integer constant expression with  the  value  0,  or
       such  an  expression  cast  to type void *, is called a null
       pointer constant.  If a null pointer constant is assigned
       to or compared for equality to a pointer,  the  constant  is
       converted to a pointer of that type.  Such a pointer, called
       a null pointer,  is  guaranteed  to  compare  unequal  to  a
       pointer to any object or function.

       [#4]  Conversion  of  a null pointer to another pointer type
       yields a null pointer of that type.  Any two  null  pointers
       shall compare equal.

       7.17  Common definitions <stddef.h>
...
       [#3] The macros are
               NULL
       which  expands  to  an  implementation-defined  null pointer
       constant; and
..

The root cause is of this whole problem is *NULL, an undefined,
unportable construct.  That's what should be fixed, not the definition
of NULL if you want this code to be portable.

Warner

Re: [Qemu-devel] error compiling hw/sh7750.c

[Andreas Schwab]

Carlo Marcelo Arenas Belon <carenas@sajinet.com.pe> writes:

> and run a test, the problem is in /usr/include/iso/stdio_iso.h that defined
> NULL as an int and not a pointer as shown by :

NULL is not a pointer, it is a null pointer constant, and 0 is a valid
null pointer constant.

> contradicting (at least in spirit) the C99 (ISO/IEC 9899:TCS) standard that
> says :
>
>   "A pointer to void shall have the same representation and alignment 
>    requirements as a pointer to a character type"

A null pointer constant is not a pointer.  It can be convert to a
pointer, but that happens only in the right context (through explicit or
implicit conversions).

> and therefore makes sizeof(*NULL) == 1 solving the problem to compile as it is
> done if applying the following patch :
>
> Carlo
>
> ---
> --- qemu/hw/sh7750.c	Sun Oct  7 09:31:11 2007
> +++ qemu/hw/sh7750.c	Sun Nov 11 16:13:13 2007
> @@ -29,6 +29,9 @@
>  #include "sh7750_regnames.h"
>  #include "sh_intc.h"
>  
> +#undef NULL
> +#define NULL ((void *)0)

This does not make it better, since sizeof(void) is still not allowed in
C.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux Products GmbH, Maxfeldstraße 5, 90409 Nürnberg, Germany
PGP key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."