Undefined behavior about integer overflow

IntegerOverflowUndefinedBehavior.md

Overflowing an integer is undefined behavior. Then the compiler is allowed to assume you are not going to overflow integers and optimize accordingly.

I was doing some delicate bit-fiddling code and got a surprise, that if you put in a file this

#include <stdint.h>

void consume(bool);

extern long maxInt;

bool noUndefinedBehavior() {
    maxInt = static_cast<unsigned long>(long{-1}) >> 1;
    consume(INT64_MAX == maxInt);
    long minInt = static_cast<unsigned long>(maxInt) + 1;
    return minInt < 0;
}

bool undefinedBehavior() {
    maxInt = static_cast<unsigned long>(long{-1}) >> 1;
    consume(INT64_MAX == maxInt);
    auto minInt = maxInt + 1;
    return minInt < 0;
}

then define in another .cpp file some consume function and long maxInt; then build with -O2, and you'll see undefinedBehavior returns false!