glm::rotation() produce non-identity quaternions for equal vectors

[WinterDog]

For equal or almost equal vectors (when cosTheta is ~= 1.0f) glm::rotation() can produce slightly incorrect quaternions. You can easily test it with this code:

glm::vec3 testVec = glm::normalize(glm::vec3(5.0f, 5.0f, 5.0f));
glm::quat rotQuat = glm::rotation(testVec, testVec); // Result will be (0.0f, 0.0f, 0.0f, 0.99999994f)

It is OK in most cases, but there are specific exceptions. I think it would be better to add check to glm::rotation() for such cases, something like that:

if (cosTheta >= static_cast<T>(1) - epsilon<T>())
    return quat();

This would also prevent unnecessary extra calculations like glm::cross() (which returns zero vector for that case).

[Groovounet]

It looks like I missed that bug... sorry about that. It should be fixed in master branch for GLM 0.9.7.1 release.

Thanks,
Christophe

[OlivierSohn]

Hello, I think the fix was not correct, as it produced not an identity quaternion but a zero quaternion. #691 should fix it I think.