open_dataset crash with long filenames

[sbiner]

Code Sample

import xarray as xr
import shutil
import numpy as np

# create netcdf file
data = np.random.rand(4, 3)
foo = xr.DataArray(data)
foo.to_netcdf('test.nc')

f_nc = 'a.nc'
shutil.copy('test.nc', f_nc)
while 1:
    print '{:05n} characteres'.format(len(f_nc))
    ds1 = xr.open_dataset(f_nc)
    ds1.close()
    nf_nc = 'a' + f_nc
    shutil.move(f_nc, nf_nc)
    f_nc = nf_nc

    if len(f_nc) == 100:
        break

Problem description

On my linux machine (CentOS) this code crashes (memory corrruption) when the filename length hits 32 characters.

On my OSX machine it is fine until 255 character and stops with an IOError

Output of xr.show_versions()

# Paste the output here xr.show_versions() here

INSTALLED VERSIONS

commit: None
python: 2.7.11.final.0
python-bits: 64
OS: Linux
OS-release: 3.10.0-514.2.2.el7.x86_64
machine: x86_64
processor: x86_64
byteorder: little
LC_ALL: None
LANG: en_CA.UTF-8
LOCALE: None.None

xarray: 0.10.4
pandas: 0.22.0
numpy: 1.14.2
scipy: 0.16.1
netCDF4: 1.2.2
h5netcdf: None
h5py: 2.5.0
Nio: None
zarr: None
bottleneck: None
cyordereddict: None
dask: 0.17.2
distributed: None
matplotlib: 1.5.0
cartopy: 0.13.1
seaborn: 0.8.1
setuptools: 19.2
pip: 10.0.1
conda: None
pytest: None
IPython: 4.0.1
sphinx: 1.7.2

[shoyer]

Try updating netCDF4-python to a newer version if possible.

This looks similar to a previously reported issue when using old versions of netCDF4-Python with 88 character filenames in #1745

[sbiner]

Hi @shoyer

updating netCDF4 to 1.4.0 works. I now obtain a proper IOError message. Help appreciated. Thanks.

I intent to close the issue, hope it's proper "etiquette".