Error converting masked time variable to python datetime

[emilycollier]

Linux
Python 3.7.7
cftime 1.1.2
netCDF4 1.5.3
Data to reproduce the error: https://www.ncdc.noaa.gov/ibtracs/index.php?name=ib-v4-access > netCDF > IBTrACS.SI.v04r00.nc

I have a masked time variable ("times"; fill_value=-9999000.0) sliced from a 2D time array ("times_tc") in a netCDF4 file that I can't convert to a python datetime object using:
datetimes = netCDF4.num2date(times, times_tc.units, times_tc.calendar,
only_use_cftime_datetimes=False, only_use_python_datetimes=True)
The error message I receive is "OverflowError in python datetime, probably because year < datetime.MINYEAR."

It works if I pass non-masked values:
datetimes = nc.num2date(times[0:np.argmax(times.mask)-1], times_tc.units, times_tc.calendar,
only_use_cftime_datetimes=False, only_use_python_datetimes=True)

I am new to python, but looking at lines 346 to 350 at https://github.com/Unidata/cftime/blob/master/cftime/_cftime.pyx, it seems the problem could be that passed times are converted to a numpy array before checking for a mask, which removes the mask.

[jswhit]

Looks like it works if only_use_cftime_datetimes=True,only_use_python_datetimes=False so it must be a flaw in the logic to treat masked values when creating python datetimes. I'll have a look.

[jswhit]

Fix in PR #177

[emilycollier]

Great, thanks!