BUG: segfault from json.dumps in Python 3.10

[fangchenli]

The segfault is from pandas/tests/io/json/test_pandas::TestPandasContainer::test_tz_is_utc and pandas/tests/io/json/test_pandas::TestPandasContainer::test_tz_range_is_utc

[fangchenli]

#42518 The segfault seems disappeared somehow.

[kleschenko]

@fangchenli It seems like this (or related) issue is still happening in the final Python 3.10 builds. Here is a reproducible scenario using the docker image:

$ docker run -it --rm python:3.10.1-slim-buster bash
Unable to find image 'python:3.10.1-slim-buster' locally
...
root@4064d1f7d5c9:/# pip install pandas ipython
...
root@4064d1f7d5c9:/# ipython
Python 3.10.1 (main, Dec 21 2021, 09:50:13) [GCC 8.3.0]
Type 'copyright', 'credits' or 'license' for more information
IPython 8.0.0 -- An enhanced Interactive Python. Type '?' for help.

In [1]: import pandas as pd

In [2]: df = pd.read_json('{"columns":["timestamp","forecast_distance","forecast_point","prediction"],"index":[0],"data":[["200
   ...: 8-07-31T21:04:19.800000Z",100,"2008-07-31T21:04:19.700Z",9.30289263e+123]]}', orient='split', convert_dates=False, conv
   ...: ert_axes=False)

In [3]: df.to_json(orient="split", date_format="iso", index=False)
Out[3]: '{"columns":["timestamp","forecast_distance","forecast_point","prediction"],"data":[["2008-07-31T21:04:19.800000Z",100,"2008-07-31T21:04:19.700Z",9.30289263e+123]]}'

In [4]: df["forecast_point"] = pd.to_datetime(df["forecast_point"])

In [5]: df.to_json(orient="split", date_format="iso", index=False)
Segmentation fault (core dumped)
root@4064d1f7d5c9:/# 

[lithomas1]

Hi @kleschenko, I think my PR fixes this.
This is what I get on my PR. Is this correct?

>>> import pandas as pd
>>> df = pd.read_json('{"columns":["timestamp","forecast_distance","forecast_point","prediction"],"index":[0],"data":[["2008-07-31T21:04:19.800000Z",100,"2008-07-31T21:04:19.700Z",9.30289263e+123]]}', orient='split', convert_dates=False, convert_axes=False)
>>> df.to_json(orient="split", date_format="iso", index=False)
'{"columns":["timestamp","forecast_distance","forecast_point","prediction"],"data":[["2008-07-31T21:04:19.800000Z",100,"2008-07-31T21:04:19.700Z",9.30289263e+123]]}'
>>> df["forecast_point"] = pd.to_datetime(df["forecast_point"])
>>> df.to_json(orient="split", date_format="iso", index=False)
'{"columns":["timestamp","forecast_distance","forecast_point","prediction"],"data":[["2008-07-31T21:04:19.800000Z",100,"2008-07-31T21:04:19.700Z",9.30289263e+123]]}'

[kleschenko]

@lithomas1 yes, it is correct. Thanks for looking into this!