Incorrect handling of overflow

[Ghabry]

Test case:

@> Control Variables: V[0001] = 9999999
@> Control Variables: V[0001] *= 300

This overflow a signed 32bit integer.

RPG_RT: 9999999
Player: -9999999

[Ghabry]

For MSVC... well.........

---

For gcc and clang this is quite trivial, something like this (uses compiler intrinsics to create optimal code like checking the overflow flag of the CPU):

constexpr Var_t VarAdd(Var_t l, Var_t r) {
	Var_t res = 0;
	if (__builtin_add_overflow(l, r, &res)) {
		// FIXME: Determine if under- or overflow
		// Only possible for + + or - -: ++ overflow else underflow
		return std::numeric_limits<Var_t>::max();
	}
	return res;
}

constexpr Var_t VarSub(Var_t l, Var_t r) {
	Var_t res = 0;
	if (__builtin_sub_overflow(l, r, &res)) {
		// FIXME: Determine if under- or overflow
		// Only possible for + - or - +: + - overflow, else underflow
		return std::numeric_limits<Var_t>::max();
	}
	return res;
}

constexpr Var_t VarMult(Var_t l, Var_t r) {
	Var_t res = 0;
	if (__builtin_mul_overflow(l, r, &res)) {
		// FIXME: Determine if under- or overflow
		// All cases possible here:
		// When + + or - -: Overflow, else underflow
		return std::numeric_limits<Var_t>::max();
	}
	return res;
}

[Ghabry]

MSVC could use "safeint.h", is part of the Windows SDK.

If you wonder why I want built-ins: SPEEEEEEEEEEEED. ;)

Better wait Post-0.7 for this because this needs not too old compilers.

[carstene1ns]

Since it is clamped later anyway, return std::numeric_limits<Var_t>::max(); could be return _max;

Player/src/game_variables.cpp

Lines 90 to 91 in 93d1f32

	value = op(v, value);
	v = Utils::Clamp(value, _min, _max);

[Ghabry]

Since it is clamped later anyway, return std::numeric_limits<Var_t>::max(); could be return _max;

Player/src/game_variables.cpp

Lines 90 to 91 in 93d1f32

	value = op(v, value);
	v = Utils::Clamp(value, _min, _max);

Couldn't use this btw because the Op is a static function and has no access to the configured limits