Actually, the problem is with the CEXPR_EH macro; it always returns 1.
If I replace the definition with '#define CEXPR_EH
__builtin_constant_p', all of your tests pass.
I guess the problem is that the behaviour of this clause (from 6.5.15p6)
is not correctly implemented:
if one operand is a null pointer constant, the result has the type of
the other operand
-E
On Thu, 20 May 2021, Petr Skocik wrote:
Hi.
Not sure if this is news but tinycc treats certain values which are
definitely not integer constant expressions as if they were.
Examples:
#define CEXPR_EH(X) _Generic((1? (void *) ((!!(X))*0ll) : (int *) 0),
int*: 1, void*: 0)
//ok
void should_be_ok1(void) { enum{x=0}; _Static_assert(CEXPR_EH(x),""); }
//each of these fails on tinycc
void should_be_ok2(void) { extern int x;
_Static_assert(!CEXPR_EH(x),""); }
void should_be_ok3(void) { int x=0; _Static_assert(!CEXPR_EH(x),""); }
void should_be_ok4(int x) { _Static_assert(!CEXPR_EH(x),""); }
void should_be_ok5(void) { double x=0;
_Static_assert(!CEXPR_EH(x),""); }
void should_be_ok6(void) { extern double y;
_Static_assert(!CEXPR_EH(y),""); }
void should_be_ok7(void) { char *x=0 ;
_Static_assert(!CEXPR_EH(x),""); }
void should_be_ok8(char *x) { _Static_assert(!CEXPR_EH(x),""); }
Regards,
Petr S.
_______________________________________________
Tinycc-devel mailing list
Tinycc-devel@nongnu.org
https://lists.nongnu.org/mailman/listinfo/tinycc-devel
------------------------------------------------------------------------
_______________________________________________
Tinycc-devel mailing list
Tinycc-devel@nongnu.org
https://lists.nongnu.org/mailman/listinfo/tinycc-devel