Re: Am I using PRIxPTR wrong? Musl-libc complains, glibc doesn't

[Rich Felker <dalias@aerifal.cx>]

On Tue, Jan 14, 2014 at 08:36:36PM +0000, David Wuertele wrote:
> /* test.c - demo difference between glibc and musl-libc wrt PRIxPTR
> **
> ** Both native (x86_64 glibc) and target (arm musl-libc) define
> ** PRIxPTR as "lx", but uintptr_t as unsigned int:
> **
> **   $ gcc -E test.c | grep uintptr_t
> **   typedef unsigned long int uintptr_t;
> **     printf ("main is at 0x%""l" "x""\n", (uintptr_t) main);
> **
> **   $ arm-linux-musleabishf-gcc -E test.c | grep uintptr_t
> **   typedef unsigned int uintptr_t;
> **     printf ("main is at 0x%""lx""\n", (uintptr_t) main);
> **
> ** While native gcc doesn't complain, target gcc does:
> ** 
> **   $ gcc -c test.c -Wformat
> **
> **   $ arm-linux-musleabishf-gcc -c test.c -Wformat
> **   test.c: In function ‘main’:
> **   test.c:6:3: warning: format ‘%lx’ expects argument of type
> ** ‘long unsigned int’, but argument 2 has type ‘unsigned int’ [-Wformat]
> */
> 
> #include <stdio.h>
> #include <inttypes.h>
> int
> main (int ac, char *av[])
> {
>   printf ("main is at 0x%"PRIxPTR"\n", (uintptr_t) main);
>   return 0;
> }
> 

I suspect your compiler was miscompiled for a non-EABI ARM
configuration; perhaps putting "shf" on the end of the maching tuple
string messed it up. I can't find the exact logic for UINTPTR_TYPE,
but my copy of gcc-4.7.3/gcc/config/arm/arm.h contains:

#define SIZE_TYPE (TARGET_AAPCS_BASED ? "unsigned int" : "long unsigned int")

suggesting that for AAPCS-based (EABI) targets, the correct type for
pointer-sized integer types is int, whereas on legacy targets, a long
type was used.

Note that if your compiler was miscompiled for non-EABI, many other
things will be wrong, like alignment/padding of 64-bit arguments, etc.

Rich