Incorrect double.pow for large exponents

[ghost]

LibTest/core/Math/pow/Math/pow/A11/t01
LibTest/core/Math/pow/Math/pow/A13/t01

[ghost]

And for NaN:
LibTest/core/Math/pow/Math/pow/A18/t01

[iposva-google]

Added Accepted label.

[ghost]

I think that several tests are flawed. Evaluating if co19 issues should be filed.

LibTest/core/Math/pow/Math/pow/A11/t01

---

This test is wrong since it exceeds double precision:

Expect.equals(double.NEGATIVE_INFINITY, Math.pow(-.0, -9223372036854775809.0))

Javascript disagrees with this expected results, should we also?

Expect.equals(double.NEGATIVE_INFINITY, Math.pow(-.0, -1.001));

LibTest/core/Math/pow/Math/pow/A13/t01

---

All tests expect NaN instead of 1 in cases like Math.pow(1, double.INFINITY).

LibTest/core/Math/pow/Math/pow/A18/t01 is fixed.

---

Remaining issue is that there is not Bigint Math.pow version to handle:

Expect.equals(double.NEGATIVE_INFINITY, Math.pow(-.0, -9223372036854775809));

---

cc @floitschG.

[floitschG]

Definitely agree with A11/01 and -9223372036854775809.0. Exceeds double-precision and the closest number is actually even.

I would be ok with saying that Math.pow transforms both arguments to double if one of them is a double.
That still requires a bigint version for Math.pow on integers, but in your example I would simply return Math.pow(-.0, -9223372036854775809.0).

[ghost]

The math libraries have been moved and pow has been implemented even for bigints (tested).

---

Added Fixed label.