AttributeError: module 'distutils' has no attribute 'version' : with setuptools 59.6.0

[keskival]

🐛 Describe the bug

# python3 -m pip install --upgrade setuptools torch tensorboard`
# python3
Python 3.8.10 (default, Sep 28 2021, 16:10:42) 
[GCC 9.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> from torch.utils.tensorboard import SummaryWriter
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/local/lib/python3.8/dist-packages/torch/utils/tensorboard/__init__.py", line 4, in <module>
    LooseVersion = distutils.version.LooseVersion
AttributeError: module 'distutils' has no attribute 'version'

Versions

Collecting environment information...
PyTorch version: 1.10.0+cu102
Is debug build: False
CUDA used to build PyTorch: 10.2
ROCM used to build PyTorch: N/A

OS: Ubuntu 20.04.3 LTS (x86_64)
GCC version: (Ubuntu 9.3.0-17ubuntu1~20.04) 9.3.0
Clang version: Could not collect
CMake version: Could not collect
Libc version: glibc-2.31

Python version: 3.8.10 (default, Sep 28 2021, 16:10:42)  [GCC 9.3.0] (64-bit runtime)
Python platform: Linux-5.4.0-88-generic-x86_64-with-glibc2.29
Is CUDA available: False
CUDA runtime version: No CUDA
GPU models and configuration: No CUDA
Nvidia driver version: No CUDA
cuDNN version: No CUDA
HIP runtime version: N/A
MIOpen runtime version: N/A

Versions of relevant libraries:
[pip3] numpy==1.21.4
[pip3] torch==1.10.0
[conda] Could not collect

cc @seemethere @malfet @pytorch/pytorch-dev-infra

[keskival]

As per PEP-632 migration advice, distutils.version should migrate to using packaging.version instead:
https://www.python.org/dev/peps/pep-0632/#migration-advice

[keskival]

distutils.version is also still used in several other places in testing tools and build utilities:

# find . -name "*.py" | xargs grep distutils\.version
./test/run_test.py:        if distutils.version.LooseVersion(torch.version.cuda) >= "11.5":
./test/test_sparse.py:from distutils.version import LooseVersion
./test/test_spectral_ops.py:LooseVersion = distutils.version.LooseVersion
./torch/testing/_internal/common_methods_invocations.py:                                    active_if=TEST_SCIPY and distutils.version.LooseVersion(scipy.__version__) < "1.4.0"),
./torch/testing/_internal/common_methods_invocations.py:                                    active_if=TEST_SCIPY and distutils.version.LooseVersion(scipy.__version__) < "1.4.0"),
./torch/testing/_internal/common_methods_invocations.py:                                    active_if=TEST_SCIPY and distutils.version.LooseVersion(scipy.__version__) < "1.4.0"),
./torch/testing/_internal/common_cuda.py:CUDA11OrLater = torch.version.cuda and distutils.version.LooseVersion(torch.version.cuda) >= "11.0"
./tools/setup_helpers/cmake.py:        if cmake3 is not None and CMake._get_version(cmake3) >= distutils.version.LooseVersion("3.10.0"):
./tools/setup_helpers/cmake.py:        elif cmake is not None and CMake._get_version(cmake) >= distutils.version.LooseVersion("3.10.0"):
./tools/setup_helpers/cmake.py:                return distutils.version.LooseVersion(line.strip().split(' ')[2])

These are outside the scope of the pull request, which only fixes the immediate problem of inability to import TensorBoard.

[keskival]

Apparently there's an alternative pull request for this same bug here: #69823

[malfet]

#69904 should fix the immediate fallback of setuptools update.
As I've already mentioned in #69823 it would be good to have a more detailed comparison of packaging.Version vs distuitls.version.LooseVersion (as former more closely resembles distutils.version.StrictVersion, isn't it?)

Also, version is not removed from setuptool, it simply no-longer accidentally re-imported. Proper way to import it always been from distutils.version import LooseVersion, not sure what started from setuptools import distutils; distuils.version.LooseVersion anti-pattern