open_dataset segfaults when opening hdf5 with array of strings

What happened?

An attempt to open non netCDF4 hdf5 file with xarray.open_dataset results in a segmentation fault.

What did you expect to happen?

Code execute and not die a horrible death.

Minimal Complete Verifiable Example

import h5py
import numpy as np
from xarray import open_dataset

filename = "foobar.h5"

with h5py.File(filename, "w") as fp:
    fp["test"] = np.array(["foo", "bar", "baz"], dtype="|S3")

with open_dataset(filename, engine="netcdf4") as ds:
    print("yay")

MVCE confirmation

• Minimal example — the example is as focused as reasonably possible to demonstrate the underlying issue in xarray.
• Complete example — the example is self-contained, including all data and the text of any traceback.
• Verifiable example — the example copy & pastes into an IPython prompt or Binder notebook, returning the result.
• New issue — a search of GitHub Issues suggests this is not a duplicate.

Relevant log output

$ python3 -Wd -X faulthandler bug.py 
Fatal Python error: Segmentation fault

Current thread 0x00007f2283947740 (most recent call first):
  File "/home/xarth/codes/xarthisius/xarray/xarray/backends/netCDF4_.py", line 106 in _getitem
  File "/home/xarth/codes/xarthisius/xarray/xarray/core/indexing.py", line 816 in explicit_indexing_adapter
  File "/home/xarth/codes/xarthisius/xarray/xarray/backends/netCDF4_.py", line 93 in __getitem__
  File "/home/xarth/codes/xarthisius/xarray/xarray/core/indexing.py", line 527 in __array__
  File "/home/xarth/codes/xarthisius/xarray/xarray/core/variable.py", line 250 in _as_array_or_item
  File "/home/xarth/codes/xarthisius/xarray/xarray/core/variable.py", line 510 in values
  File "/home/xarth/codes/xarthisius/xarray/xarray/core/variable.py", line 337 in data
  File "/home/xarth/codes/xarthisius/xarray/xarray/core/common.py", line 1866 in contains_cftime_datetimes
  File "/home/xarth/codes/xarthisius/xarray/xarray/core/common.py", line 1875 in _contains_datetime_like_objects
  File "/home/xarth/codes/xarthisius/xarray/xarray/conventions.py", line 345 in decode_cf_variable
  File "/home/xarth/codes/xarthisius/xarray/xarray/conventions.py", line 521 in decode_cf_variables
  File "/home/xarth/codes/xarthisius/xarray/xarray/backends/store.py", line 27 in open_dataset
  File "/home/xarth/codes/xarthisius/xarray/xarray/backends/netCDF4_.py", line 567 in open_dataset
  File "/home/xarth/codes/xarthisius/xarray/xarray/backends/api.py", line 495 in open_dataset
  File "/home/xarth/codes/xarthisius/yt/bug.py", line 10 in <module>
Segmentation fault (core dumped)

Anything else we need to know?

This regression was introduced in #6489

Environment

INSTALLED VERSIONS

commit: None python: 3.9.5 (default, Nov 23 2021, 15:27:38) [GCC 9.3.0] python-bits: 64 OS: Linux OS-release: 5.4.0-121-generic machine: x86_64 processor: x86_64 byteorder: little LC_ALL: en_US.utf8 LANG: en_US.utf8 LOCALE: (‘en_US’, ‘UTF-8’) libhdf5: 1.12.2 libnetcdf: 4.9.0

xarray: 2022.3.1.dev60+g5f01c115 pandas: 1.4.3 numpy: 1.23.0 scipy: 1.9.0rc1 netCDF4: 1.6.0 pydap: None h5netcdf: None h5py: 3.7.0 Nio: None zarr: None cftime: 1.6.0 nc_time_axis: None PseudoNetCDF: None rasterio: None cfgrib: None iris: None bottleneck: None dask: 2022.6.1 distributed: None matplotlib: 3.5.2 cartopy: None seaborn: None numbagg: None fsspec: 2022.5.0 cupy: None pint: None sparse: None setuptools: 44.0.0 pip: 20.0.2 conda: None pytest: 7.1.2 IPython: 8.4.0 sphinx: None

[Xarthisius]

Can you reproduce this without xarray? It should probably be reported upstream? (In the netCDF4 repo?)

Yeah, it can be triggered with:

from netCDF4 import Dataset
import h5py
import numpy as np
import operator

filename = "foobar.h5"
with h5py.File(filename, "w") as fp:
    fp["test"] = np.array(["foo", "bar", "baz"], dtype="|S3")

rootgrp = Dataset("foobar.h5", "r", format="NETCDF4")
# This works
print(rootgrp["/test"][slice(None)])

# This segfaults
operator.getitem(rootgrp["/test"], slice(None))

[dcherian]

Thanks @kmuehlbauer