https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121209
--- Comment #9 from Halalaluyafail3 <luigighiron at gmail dot com> --- (In reply to uecker from comment #8) > No, it has to make a decision and document it. For GCC this is here: > > https://gcc.gnu.org/onlinedocs/gcc/Types-implementation.html C23 defines implementation-defined behavior as "unspecified behavior where each implementation documents how the choice is made", so it just has to document how the choice is made not which choice is made. Furthermore, implementation-defined value is defined as "unspecified value where each implementation documents how the choice is made". So I don't see how "If one type is an enumerated type and the other is an integer type other than an enumerated type, it is implementation-defined whether or not the composite type is an enumerated type." can be interpreted to require that it has to be chosen consistently for all enumerated types. An example of when GCC doesn't choose the same result consistently is when right shifting a negative integer: "The result of E1 >> E2 is E1 right-shifted E2 bit positions. If E1 has an unsigned type or if E1 has a signed type and a nonnegative value, the value of the result is the integral part of the quotient of E1/2^E2. If E1 has a signed type and a negative value, the resulting value is implementation-defined." Here GCC doesn't choose the same value consistently so expressions such as -4>>1 and -2>>1 result in different values.
