DataFrame.prod after transpose on Windows results in incorrect result

[devin-petersohn]

System information

• OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Windows only
• Modin version (modin.__version__): 0.7.1+4.ga39d66a
• Python version: All
• Code we can use to reproduce:

Describe the problem

Recently with #1081, a new behavior has emerged caused by updating pandas. For DataFrame.prod in Windows, pytest will generate incorrect numbers for df.T.prod. This can be reproduced by running the tests after removing the @pytest.markskipif command on these lines:

modin/modin/pandas/test/test_dataframe.py

Lines 3214 to 3217 in a39d66a

	@pytest.mark.skipif(
	os.name == "nt",
	reason="Windows has a memory issue for large numbers on this test",
	)

modin/modin/pandas/test/test_dataframe.py

Lines 3269 to 3272 in a39d66a

	@pytest.mark.skipif(
	os.name == "nt",
	reason="Windows has a memory issue for large numbers on this test",
	)

Note that this is only happening on prod and on Windows.

[vnlitvinov]

@devin-petersohn are you sure this is fixed? I see the same issue hitting CI, for example here: https://github.com/modin-project/modin/pull/1801/checks?check_run_id=908354983

[devin-petersohn]

I am not sure. There were some CI issues yesterday so #1801 may have been merged by mistake.

[devin-petersohn]

Reverted commit.

[mvashishtha]

Looks like we're no longer skipping the prod test even in the transposed cases:

modin/modin/pandas/test/dataframe/test_reduce.py

Line 294 in 3578288

def test_prod(

So I'll close this issue.