CI: npdev SystemError

[jbrockmendel]

cc @seberg was there a change in the ufunc code in the last day or so that might lead to a TypeError getting turned into a SystemError?

[seberg]

Hmm, yeah, it could be numpy/numpy#18450 which is backported to the 1.20.x branch, so we may want to take care here.

I would expected the opposite, SystemError getting turned into TypeError (i.e. bad/weird error being made OK), if the opposite happened here, we should look into it maybe. Do you have a reproducer?!

[jbrockmendel]

Do you have a reproducer?!

https://dev.azure.com/pandas-dev/pandas/_build/results?buildId=55279&view=logs&j=eab14f69-13b6-5db7-daeb-7b778629410b&t=ce687173-08c6-5301-838d-71b2dda24510&l=43

I haven't tried it locally, but this is a copy-pasteable version of whats failing on the CI.

import pandas as pd
import pandas._testing as tm

data = pd.array([True, True, pd.NA, pd.NA, False, False, True])
df = pd.DataFrame({"A": [1, 1, 2, 2, 3, 3, 1], "B": data})
result = df.groupby("A").B.apply(lambda x: x.array)
expected = pd.Series(
            [
                df.B.iloc[[0, 1, 6]].array,
                df.B.iloc[[2, 3]].array,
                df.B.iloc[[4, 5]].array,
            ],
            index=pd.Index([1, 2, 3], name="A"),
            name="B",
        )
tm.assert_series_equal(result, expected)

[seberg]

Thanks, that was perfect, already found the issue. (If you run it in a python debug version, by default this already causes an assert, and that means you immediately get a gdb stacktrace, which trivially finds the issue – it was a simple oversight in that PR I linked.)