NumPy and cuNumeric behave differently in 5 cases in API concatenate

[robinwnv]

NumPy and cuNumeric behave differently for the following cases.
Some of the cases need to be fixed while the others are expected divergences.

Case-1

arrays = (1,)
num.concatenate(arrays, axis=0)
arrays = (1, 2)
num.concatenate(arrays, axis=0)

In Numpy, in both cases, it raises ValueError
In cuNumeric, in both cases, it raises IndexError

Case-2

a = [[1, 2], [3, 4]]
b = [[5, 6]]
axis = 0
out_np = np.zeros((3, 2))
out_num = num.array(out_np)
dtype = np.float32
num.concatenate((num.array(a), num.array(b)), axis=axis, out=out_num, dtype=dtype)

In Numpy, it raises TypeError as dtype cannot be provided together with out.
In cuNumeric, it pass.

Case-3

a = [[1, 2], [3, 4]]
b = [[5, 6]]
axis = 1
num.concatenate((num.array(a), num.array(b)), axis=axis)

In Numpy, it raises ValueError as along axis=0, the size of a and b are not equal.
In cuNumeric, it pass.

Case-4

a = [1, 2]
b = [5, 6]
num.concatenate((num.array(a), num.array(b)), axis=axis)

For axis = -2 or 1,
In Numpy, it raises ValueError
In cuNumeric, it raises IndexError

Case-5

a = [[1, 2], [3, 4]]
b = [[5, 6]]
num.concatenate(num.array(a), num.array(b)), axis=0, casting="unknown")

For invalid casting,
In Numpy, raise ValueError
In cuNumeric, pass

[robinwnv]

PR-844 and PR-866 fixed this issue. It's verified all tests pass. So close this issue.