test_c_parser_only on linux py_3.8_32 failing on MacPython.pandas-wheels

[simonjayhawkins]

https://dev.azure.com/pandas-dev/pandas-wheels/_build/results?buildId=42365&view=logs&j=bb1c2637-64c6-57bd-9ea6-93823b2df951&t=9c62e7ff-391e-517d-ede1-af63bc6a25da

=========================== short test summary info ============================
FAILED ../../venv/lib/python3.8/site-packages/pandas/tests/io/parser/test_c_parser_only.py::test_float_precision_options[c_high]
FAILED ../../venv/lib/python3.8/site-packages/pandas/tests/io/parser/test_c_parser_only.py::test_float_precision_options[c_low]
= 2 failed, 66511 passed, 7439 skipped, 1144 xfailed, 6 xpassed, 15 warnings in 434.87s (0:07:14) =

[jreback]

these are still failing?

[simonjayhawkins]

yes.

on latest nightly build (https://dev.azure.com/pandas-dev/pandas-wheels/_build/results?buildId=48368&view=logs&j=79f3a53a-4a6d-509c-98b5-ff6e9111f67c) failed on windows py_3.8_32 and windows py_3.9_32

some wheels are failing to build. so there maybe other envs this fails on.

I am no longer able to select the build history from the breadcrumb bar. permission error. need to log in to Azure DevOps. am using sign-in with GitHub. are you seeing this too?

[jreback]

these look like we are mishandling 32 bit pointers

these are failing nightly but not on commit pushes?

[simonjayhawkins]

these are not failing on commit pushes.

IIUC on commit pushes we use a docker container for the 32bit environment where on MacPython the environment is selected using PYTHON_ARCH environment variable.

I think MacPython may need some updates on versions, maybe NP_BUILD_DEP and CYTHON_BUILD_DEP. have opened MacPython/pandas-wheels#111 to troubleshoot the failures. this failure may or may not be related.

[simonjayhawkins]

forgot to mention. 1.1.x wheel builds are OK. so this is probably not an environment issue. MacPython/pandas-wheels#109

[simonjayhawkins]

forgot to mention. 1.1.x wheel builds are OK. so this is probably not an environment issue. MacPython/pandas-wheels#109

forget that.

test_float_precision_options was added in #36228 which is not in 1.1.x