glm::equal(val1, val2, int ULPs) gives the wrong answer

[petersteneteg]

This test

glm/glm/ext/scalar_relational.inl

Lines 27 to 31 in ea678fa

	if(a.negative() != b.negative())
	{
	// Check for equality to make sure +0==-0
	return a.mantissa() == b.mantissa() && a.exponent() == b.exponent();
	}

will return true for any two number of different sign but with equal magnitude.

the test on line 30 should be return x==y;

Example
https://godbolt.org/z/iR8unh

[Groovounet]

This issue should be fixed in master branch.

Thanks for contributing,
Christophe

[petersteneteg]

Should not -0.0f be equal to 0.0f?

[petersteneteg]

Btw this code is also in vector_relational.inl

[petersteneteg]

@Groovounet see above

[Groovounet]

Do you have an idea how to implement this correctly?

[petersteneteg]

This is definitely not in my expertise, but looking at the article referenced from
https://github.com/g-truc/glm/blob/master/glm/detail/type_float.hpp

https://randomascii.wordpress.com/2012/02/25/comparing-floating-point-numbers-2012-edition/

I would guess something similar to this...

    // Different signs means they do not match.
    if (uA.Negative() != uB.Negative())
    {
        // Check for equality to make sure +0==-0
        if (A == B)
            return true;
        return false;
    }