abs() macro in Arduino.h inverts signedness of (positive) 0.0F

[dok-net]

With regard to arduino/ArduinoCore-avr#406,
I found out that the C macro defined in our Arduino.h, flips the signedness of 0.0F to -0.0F. I think it would be more natural, if one has to make a choice, to let the error (if it is considered as such) be such that abs() returns (negative) -0.0F for the input (negative) -0.0F but not flips the signedness of the positive 0.0F.

MCVE, abstest is an identical copy of the abs() macro in Arduino.h:

float one = 1.0F;
float minusone = -1.0F;
float zero = 0.0F;
float minuszero = -0.0F;

#define abstest(x) ({ __typeof__(x) _x = (x); _x > 0 ? _x : -_x; })

void setup()
{
	Serial.begin(74880);
	delay(500);
	Serial.println("abstest");
	Serial.print("one = "); Serial.println(*reinterpret_cast<uint32_t*>(&one));
	Serial.print("minusone = "); Serial.println(*reinterpret_cast<uint32_t*>(&minusone));
	Serial.print("zero = "); Serial.println(*reinterpret_cast<uint32_t*>(&zero));
	Serial.print("minuszero = "); Serial.println(*reinterpret_cast<uint32_t*>(&minuszero));

	float res;
	res = abstest(one);
	Serial.print("abs(one) = "); Serial.println(*reinterpret_cast<uint32_t*>(&res));
	res = abstest(minusone);
	Serial.print("abs(minusone) = "); Serial.println(*reinterpret_cast<uint32_t*>(&res));
	res = abstest(zero);
	Serial.print("abs(zero) = "); Serial.println(*reinterpret_cast<uint32_t*>(&res));
	res = abstest(minuszero);
	Serial.print("abs(minuszero) = "); Serial.println(*reinterpret_cast<uint32_t*>(&res));
}

void loop()
{
}

Output:

abstest
one = 1065353216
minusone = 3212836864
zero = 0
minuszero = 2147483648
abs(one) = 1065353216
abs(minusone) = 1065353216
abs(zero) = 2147483648
abs(minuszero) = 0