Error number (errno) of OSError is None

I am wondering if it’s expected that error.errno in raised exception is None when a hdf5 file is opened in mode w- but already exist?

My scenario looks like this:

import h5py;

try:
    with h5py.File("file-that-already-exist.hdf5", "w-") as f:
        # Do something with the file here
except OSError as error:
    print(error.errno)
    # --> None
    print(str(error))
    # --> OSError: Unable to create file (unable to open file: name = 'file-that-already-exist.hdf5', errno = 17, error message = 'File exists', flags = 15, o_flags = a02)

Ideally I would like to be able to do:

expecting_error_no_17 = maybe_yes_maybe_no(my_magic_conditions)
try:
    with h5py.File("file-that-already-exist.hdf5", "w-") as f:
        # Do something with the file here
except OSError as error:
    if error.errno == 17 and expecting_error_no_17:
        pass
    else:
        raise

The above code doesn’t work. I can achieve it by parsing the error message for errno but I am wondering why error.errno is None instead of the error number reported in the error message? Is it expected that error.errno is None or is this a bug? If it’s expected is there a preferred way of getting the error number or is that up to the user?

h5py.version.info:

h5py    2.8.0
HDF5    1.10.3
Python  3.6.6 | packaged by conda-forge | (default, Jul 26 2018, 09:55:02)
[GCC 4.2.1 Compatible Apple LLVM 6.1.0 (clang-602.0.53)]
sys.platform    darwin
sys.maxsize     9223372036854775807
numpy   1.15.3

I am running macOS 10.12.6.

[tacaswell]

I do not think we currently get enough information out of the hdf5 layer to provide this (see https://forum.hdfgroup.org/t/get-posix-errno-from-hdf5-errors/5003 ).

[miccoli]

Opened issue #1226 : I understand that changing established behaviour is tricky, but this is clearly an issue that should be reported.