Created attachment 272844 [details] reproducible example C23 §7.12.1 p5: > A floating result overflows if a finite result value with ordinary accuracy would have magnitude > (absolute value) too large for the representation with full precision in the specified type. A result > that is exactly an infinity does not overflow. If a floating result overflows and default rounding > is in effect, then the function returns the value of the macro HUGE_VAL, HUGE_VALF, or HUGE_VALL > according to the return type, **with the same sign as the correct value of the function.** However, when an overflow or an underflow occurs in powl, it returns a wrong sign in some patterns. Wrong patterns: powl(-3.0L, odd) = inf should be -inf powl(-1.0L / 3.0L, odd) = 0 should be -0 powl(-3.0L, -odd) = 0 should be -0 powl(-1.0L / 3.0L, -odd) = inf should be -inf Correct patterns: powl(3.0L, even) = inf should be inf powl(INFINITY, even) = inf should be inf powl(3.0L, odd) = inf should be inf powl(INFINITY, odd) = inf should be inf powl(1.0L / 3.0L, even) = 0 should be 0 powl(1.0L / INFINITY, even) = 0 should be 0 powl(1.0L / 3.0L, odd) = 0 should be 0 powl(1.0L / INFINITY, odd) = 0 should be 0 powl(-3.0L, even) = inf should be inf powl(-INFINITY, even) = inf should be inf powl(-INFINITY, odd) = -inf should be -inf powl(-1.0L / 3.0L, even) = 0 should be 0 powl(-1.0L / INFINITY, even) = 0 should be 0 powl(-1.0L / INFINITY, odd) = -0 should be -0 powl(3.0L, -even) = 0 should be 0 powl(INFINITY, -even) = 0 should be 0 powl(3.0L, -odd) = 0 should be 0 powl(INFINITY, -odd) = 0 should be 0 powl(1.0L / 3.0L, -even) = inf should be inf powl(1.0L / INFINITY, -even) = inf should be inf powl(1.0L / 3.0L, -odd) = inf should be inf powl(1.0L / INFINITY, -odd) = inf should be inf powl(-3.0L, -even) = 0 should be 0 powl(-INFINITY, -even) = 0 should be 0 powl(-INFINITY, -odd) = -0 should be -0 powl(-1.0L / 3.0L, -even) = inf should be inf powl(-1.0L / INFINITY, -even) = inf should be inf powl(-1.0L / INFINITY, -odd) = -inf should be -inf You can easily reproduce the issue with the attached file. > clang -lm test.c > ./a.out
Someone grabbed powl() from the CEPHES math library. I suspect limited testing was performed.