sincosf returns (0, 1) on PI instead of (0, -1)

[CryZe]

I'm currently porting euclid to no_std and I noticed that the test for 180 degree rotations seems to fail. The problem does indeed seem to be due to the sincosf implementation in libm which returns (0, 1) on sincosf(PI) instead of (0, -1). I didn't test much further yet. sincos seems unaffected. cosf itself also returns the proper -1.

Update: It calls k_cosf(PI - S2PIO2) where S2PIO2 = PI, therefore it simplifies to k_cosf(0.0), which returns 1, which is then directly returned without modification, which seems wrong.