Tighten tests on iinfo/finfo

[crusaderky]

• test iinfo/finfo vs array input
• test output object

NOTE

This makes array-api-compat explode. Needs prerequisite PR data-apis/array-api-compat#294
CI will fail here until the next array-api-strict release, when it will pick up data-apis/array-api-strict#135

array_api_strict git tip
================================================================
(all green)

array_api_strict 2.3.1 (CI below)
================================================================
- array inputs fail (https://github.com/data-apis/array-api-strict/pull/143)
- dtype is a numpy dtype (https://github.com/data-apis/array-api-strict/pull/135)

array_api_compat.numpy (array-api-compat git tip, numpy 2.2)
================================================================
FAILED array_api_tests/test_data_type_functions.py::test_finfo[float32-dtype] - AssertionError: type(out.eps)=<class 'numpy.float32'>, but should be float [finfo(float32)]
FAILED array_api_tests/test_data_type_functions.py::test_finfo[complex64-dtype] - AssertionError: type(out.eps)=<class 'numpy.float32'>, but should be float [finfo(complex64)]
FAILED array_api_tests/test_data_type_functions.py::test_finfo[*-array] - ValueError: data type <class 'numpy.object_'> not inexact
FAILED array_api_tests/test_data_type_functions.py::test_iinfo[*-array] - ValueError: Invalid integer data type 'O'.

array_api_compat.dask.array (array-api-compat git tip)
================================================================
FAILED array_api_tests/test_data_type_functions.py::test_finfo[float32-*] - AssertionError: type(out.eps)=<class 'numpy.float32'>, but should be float [finfo(float32)]
FAILED array_api_tests/test_data_type_functions.py::test_finfo[complex64-*] - AssertionError: type(out.eps)=<class 'numpy.float32'>, but should be float [finfo(complex64)]

array_api_compat.cupy (array-api-compat git tip)
================================================================
FAILED array_api_tests/test_data_type_functions.py::test_finfo[float32-*] - AssertionError: type(out.eps)=<class 'numpy.float32'>, but should be float [finfo(float32)]
FAILED array_api_tests/test_data_type_functions.py::test_finfo[complex64-*] - AssertionError: type(out.eps)=<class 'numpy.float32'>, but should be float [finfo(complex64)]

array_api_compat.torch (array-api-compat git tip)
================================================================
FAILED array_api_tests/test_data_type_functions.py::test_finfo[dtype0-dtype] - AssertionError: assert 'float32' == torch.float32
FAILED array_api_tests/test_data_type_functions.py::test_finfo[dtype1-dtype] - AssertionError: assert 'float64' == torch.float64
FAILED array_api_tests/test_data_type_functions.py::test_finfo[dtype2-dtype] - AssertionError: assert 'float32' == torch.float32
FAILED array_api_tests/test_data_type_functions.py::test_finfo[dtype3-dtype] - AssertionError: assert 'float64' == torch.float64
FAILED array_api_tests/test_data_type_functions.py::test_iinfo[dtype0-dtype] - AssertionError: assert 'int8' == torch.int8
FAILED array_api_tests/test_data_type_functions.py::test_iinfo[dtype1-dtype] - AssertionError: assert 'int16' == torch.int16
FAILED array_api_tests/test_data_type_functions.py::test_iinfo[dtype2-dtype] - AssertionError: assert 'int32' == torch.int32
FAILED array_api_tests/test_data_type_functions.py::test_iinfo[dtype3-dtype] - AssertionError: assert 'int64' == torch.int64
FAILED array_api_tests/test_data_type_functions.py::test_iinfo[dtype4-dtype] - AssertionError: assert 'uint8' == torch.uint8
FAILED array_api_tests/test_data_type_functions.py::test_iinfo[dtype5-dtype] - AssertionError: assert 'uint16' == torch.uint16
FAILED array_api_tests/test_data_type_functions.py::test_iinfo[dtype6-dtype] - AssertionError: assert 'uint32' == torch.uint32
FAILED array_api_tests/test_data_type_functions.py::test_iinfo[dtype7-dtype] - AssertionError: assert 'uint64' == torch.uint64
FAILED array_api_tests/test_data_type_functions.py::test_iinfo[*-array] - TypeError: iinfo(): argument 'type' (position 1) must be torch.dtype, not Tensor

[crusaderky]

Suggested merge order:

1. merge array-api-compat PR
2. wait for next array-api-strict release
3. merge array-api-tests PR

[ev-br]

Not sure we want an out-of-band array-api-strict release just for this. How about adding the failing tests to the skip list for the time being.

[crusaderky]

Done. pre-commit error is unrelated.

[ev-br]

Thanks Guido, merged now.

I'll take a look at the lintrunner error. These things are meant to be helpful, no?