Re: [Tinycc-devel] Many new errors on RPi because of ssize_t incompatible redefinition

[Christian JULLIEN]

Here is the results of quick study for ptrdiff_t

VC++:

// Definitions of common types
#ifdef _WIN64
  typedef unsigned __int64 size_t;
  typedef __int64 ptrdiff_t;
  typedef __int64 intptr_t;
#else
  typedef unsigned int size_t;
  typedef int ptrdiff_t;
  typedef int intptr_t;
#endif

gcc 5.4.0 64bits on Windows:

ptrdiff_t: long int

size_t: long unsigned int

gcc 5.4.0 32bits on Windows:

ptrdiff_t: int

size_t: unsigned int

gcc 4.4.7 64bits on Linux x64:

ptrdiff_t: long int

size_t: long unsigned int

gcc 4.4.7 64bits on Linux x86:

ptrdiff_t: int

size_t: unsigned int

gcc 4.9.2 32bits on RPi

ptrdiff_t: int

size_t: unsigned int

Please note that VC++ does not define __PTRDIFF_TYPE__ __SIZE_TYPE__ which are gcc specific.

AT LEAST FOR 32bits, it looks safe to replace long -> int,. Fot PE it depends if we want to be VC++ compatible or gcc compatible. For other 64bits system I've no clear opinion atm.

Looks safe:

#else
  /* Other 32bit systems. */
  tcc_define_symbol(s, "__SIZE_TYPE__", "unsigned int");
  tcc_define_symbol(s, "__PTRDIFF_TYPE__", "int");
  tcc_define_symbol(s, "__ILP32__", NULL);
#endif

>  Le : 25 juillet 2017 à 23:01 (GMT +02:00)
> De : "Michael Matz" <address@hidden>
> À : "address@hidden" <address@hidden>
> Objet : Re: [Tinycc-devel] Many new errors on RPi because of ssize_t
>  incompatible redefinition
>
>
> Hi,
>
> On Tue, 25 Jul 2017, grischka wrote:
>
> > uso ewin wrote:
> >> I guess ptrdiff_t should indeed be a long int, but ssize_t should be an 
> >> int,
> >> same for intptr_t and uintptr_t should be an unsigned int.
> >
> > gcc 6.3.0 on win32 (32bits):
> >
> > test.c:9:14: error: '_Generic' selector of type 'int' is not compatible with 
> > any association
> >     _Generic((ptrdiff_t)0, void*:0);
> >              ^
> >
> > So at least this gcc thinks that ptrdiff_t is of type 'int'.
>
> You can definitely find out with:
>
> % echo __PTRDIFF_TYPE__ | gcc -E
>
> That is the value that tcc must use as well (and then stddef.h is supposed 
> to simply use that macro, but only to define ptrdiff_t, not any related 
> types, which shouldn't come from stddef.h at all actually).  And yes, 
> GCCs use 'int' for -m32 normally and 'long int' for -m64.
>
> So the state with commit 28b7c9 was actually correct and was unfixed with 
> the later 52622c.
>
>
> Ciao,
> Michael.
>
> _______________________________________________
> Tinycc-devel mailing list
> address@hidden
> https://lists.nongnu.org/mailman/listinfo/tinycc-devel

‍