This is the mail archive of the
java@gcc.gnu.org
mailing list for the Java project.
re: long longlong in libffi (powerpc)
- From: Andrew Haley <aph at redhat dot com>
- To: Andreas Tobler <toa at pop dot agri dot ch>
- Cc: green at redhat dot com, java at gcc dot gnu dot org
- Date: Tue, 21 Jan 2003 11:19:04 +0000 (GMT)
- Subject: re: long longlong in libffi (powerpc)
- References: <3DEE86F600004959@mail02.agrinet.ch>
=?iso-8859-1?Q?Andreas=5FTobler?= writes:
> To follow up. (Sorry can't quote, webmail)
>
> The problem I have is also visible on powerpc-linux.
>
> Given this testcase:
>
> static void closure_test_fn4(ffi_cif* cif,void* resp,void** args,
> void* userdata)
> {
> *(ffi_arg*)resp =
> (int)*(long *)args[0] +(*(int *)args[1]) +
> /* (int)(*(float *)args[2]) + *(int *)args[3]*/ + (int)(long)userdata;
>
> printf("%d %d %d %d: %d\n",
> (int)*(long *)args[0], (*(int *)args[1]),
> /* (int)(*(float *)args[2]), *(int *)args[3],*/ (int)(long)userdata,
> *(int*)resp);
> }
>
> typedef int (*closure_test_type4)(long, int);
>
> I run it with this definition:
>
> cl_arg_types[0] = &ffi_type_ulong;
> cl_arg_types[1] = &ffi_type_uint;
> cl_arg_types[2] = NULL;
>
> Here I get a failure since ulong is 'mapped' onto longlong (8bytes)
>
> setting the arg[0] type to &ffi_type_uint32 makes the testcase pass.
>
> So I'm a bit confused. My long is 4 byte, my longlong 8. And the ABI from
> sysv & darwin handles the long and the longlong different. (long !=longlong)
>
>
> Do I follow a phantom or my one explain to me what I forget?
SIZEOF_LONG is determined by an autoconf test. Here it is:
--------------------------------------------------------------------------------------
echo $ac_n "checking size of long""... $ac_c" 1>&6
echo "configure:3135: checking size of long" >&5
if eval "test \"`echo '$''{'ac_cv_sizeof_long'+set}'`\" = set"; then
echo $ac_n "(cached) $ac_c" 1>&6
else
for ac_size in 4 8 1 2 16 12 ; do # List sizes in rough order of prevalence.
cat > conftest.$ac_ext <<EOF
#line 3141 "configure"
#include "confdefs.h"
#include "confdefs.h"
#include <sys/types.h>
int main() {
switch (0) case 0: case (sizeof (long) == $ac_size):;
; return 0; }
EOF
if { (eval echo configure:3151: \"$ac_compile\") 1>&5; (eval $ac_compile) 2>&5; }; then
rm -rf conftest*
ac_cv_sizeof_long=$ac_size
else
echo "configure: failed program was:" >&5
cat conftest.$ac_ext >&5
fi
rm -f conftest*
if test x$ac_cv_sizeof_long != x ; then break; fi
done
fi
--------------------------------------------------------------------------------------
Please try to figure out if this is returning the wrong answer, and if
so why. It could be s shell bug.
Andrew.