https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71611
--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Certainly not here.
int
f1 ()
{
enum I { c = -__LONG_MAX__ - 1 } x = c;
x = -x;
return x;
}
int
f2 ()
{
enum J { c = -__INT_MAX__ - 1 } x = c;
x = -x;
return x;
}
int
f3 ()
{
int x = -__INT_MAX__ - 1;
x = -x;
return x;
}
int
f4 ()
{
enum { c = -__LONG_MAX__ - 1 } x = c;
x = -x;
return x;
}
int
f5 ()
{
enum { c = -__INT_MAX__ - 1 } x = c;
x = -x;
return x;
}
int
main ()
{
f1 ();
f2 ();
f3 ();
f4 ();
f5 ();
return 0;
}
reports:
pr71611.c:5:5: runtime error: negation of -9223372036854775808 cannot be
represented in type 'I'; cast to an unsigned type to negate this value to
itself
pr71611.c:13:5: runtime error: negation of -2147483648 cannot be represented in
type 'J'; cast to an unsigned type to negate this value to itself
pr71611.c:21:5: runtime error: negation of -2147483648 cannot be represented in
type 'int'; cast to an unsigned type to negate this value to itself
pr71611.c:29:5: runtime error: negation of -9223372036854775808 cannot be
represented in type '<unknown>'; cast to an unsigned type to negate this value
to itself
pr71611.c:37:5: runtime error: negation of -2147483648 cannot be represented in
type '<unknown>'; cast to an unsigned type to negate this value to itself
GCC in the middle-end considers enums and integer types with the same sign and
precision as compatible types, so there is no guarantee the types the sanitizer
sees are the original ones, and changing that (making for -fsanitize=undefined
conversion between different types not useless) would be way too costly for the
generated code quality.
