Four tests still "failing" but inconsequential

[mpkeightley]

Bug raised in Fingerprint fail on generating TMD #522 appears to be fixed, though flagging that four tests still "fail" on Windows, but all appear to be floating point differences between Windows and Mac rather than code bugs. Pointing this out because other users on Windows might run into this and think it is a true failure.

FAILED tests/test_weights.py::test_weights - ValueError: WEIGHT FINGERPRINT MISMATCH USING 2022 DATA:
p50: actual=398.079155, expected=398.097

FAILED tests/test_tax_expenditures.py::test_tax_exp_diffs - ValueError:
paytax,act,exp,atol,rtol= 1431.7 1429.5 0.1 1e-05

FAILED tests/test_imputed_variables.py::test_obbba_deduction_tax_benefits - ValueError:
(small numerical differences across OTM, TIP, ALI, ALL deductions)

FAILED tests/test_tmd_file_fingerprint.py::test_tmd_file_fingerprint - Failed: 34 statistic mismatch(es)
(fingerprint generated under taxcalc 6.5.2, currently installed 6.6.2):

Mac: total=189765003.27, mean=880.604580
Windows: total=189765156.11, mean=880.605289

These remaining differences appear to be platform-related rather than bugs. Happy to provide any additional diagnostic information if useful.

[martinholmer]

@mpkeightley, Thank you for posting documentation issue #527 about your experience generating TMD data on Windows.

As you know, the TMD code generates the same ("no errors") test results on both Linux and Mac computers. The differences on Windows that you report here have lead us to add a recommendation in PR #526 for Windows users to work with the TMD code using Microsoft's Windows Subsystem for Linux running Ubuntu Linux. Doing this is free and, according to @donboyd5, fairly easy. See his account of his own experiences with WSL in PR #526.