Make Python 3.10 kwarg peephole less restrictive - #8044

I did a quick check. It does resolve the failure in numba.tests.test_ndarray_subclasses.TestNdarraySubclasses

smoketest BFID: numba_smoketest_cpu_yaml_83

smoketest BFID: numba_smoketest_cpu_yaml_83

This passed on multiple build jobs with Python 3.10, and all other build jobs not impacted by some other unrelated issue.

Just checked into the few failing Python 3.10 jobs, there's still a couple that are reporting the issue in #8043. Need to take a closer look as to what has happened.

smoketest BFID: numba_smoketest_cpu_yaml_83

This passed on multiple build jobs with Python 3.10, and all other build jobs not impacted by some other unrelated issue.

Just checked into the few failing Python 3.10 jobs, there's still a couple that are reporting the issue in #8043. Need to take a closer look as to what has happened.

@stuartarchibald If you post the list of tests I'm happy to take a look.

@stuartarchibald If you post the list of tests I'm happy to take a look.

Thanks @njriasan it's these two tests, same failures as in #8043:

• numba.tests.test_ndarray_subclasses.TestNdarraySubclasses.test_my_array_return
• numba.tests.test_ndarray_subclasses.TestNdarraySubclasses.test_my_array_allocator_override

What I'm not sure about is if this is caused by a genuine issue with this patch or there's something unusual going on in the farm builds.

Set up was Python 3.10, np 1.21, linux 64 bit system.

Thanks @stuartarchibald I couldn't reproduce locally on my M1, so I'm install docker to test with a linux 64 bit system. Will update after I've tested that.

@stuartarchibald I reran the whole test file manually using a ubuntu x86 docker image. I could not reproduce any failures. I'm inclined to believe this may be a build farm issue.

@stuartarchibald, I think we might be mixing up the result of two builds. I'm going to combine the two PRs (#8044 and #8046) to tests.

@stuartarchibald, I think we might be mixing up the result of two builds. I'm going to combine the two PRs (#8044 and #8046) to tests.

Thanks @sklam

@stuartarchibald I reran the whole test file manually using a ubuntu x86 docker image. I could not reproduce any failures. I'm inclined to believe this may be a build farm issue.

@njriasan Thanks for checking.

tested as part of #8053 and its all green