Drop Python 3.5

[jrbourbeau]

This PR drops support for Python 3.5 and is a companion PR to dask/dask#5528 in which we dropped Python 3.5 in dask/dask

[TomAugspurger]

cc @mrocklin. We assume you're OK with dropping for 3.5 from distributed as well as dask, but just wanted to confirm.

[mrocklin]

I'm happy to defer to the group opinion here. I suspect that you all have thought things through well.

…

On Tue, Oct 29, 2019 at 1:19 PM Tom Augspurger ***@***.***> wrote: ***@***.**** approved this pull request. cc @mrocklin <https://github.com/mrocklin>. We assume you're OK with dropping for 3.5 from distributed as well as dask, but just wanted to confirm. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#3179?email_source=notifications&email_token=AACKZTEWBR6HRYFVPXK2AWDQRCLFXA5CNFSM4JGMRTUKYY3PNVWWK3TUL52HS4DFWFIHK3DMKJSXC5LFON2FEZLWNFSXPKTDN5WW2ZLOORPWSZGOCJUAGCY#pullrequestreview-308806411>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AACKZTC5PVYDZGYAEF5C7NDQRCLFXANCNFSM4JGMRTUA> .

[jrbourbeau]

We're getting a couple of test failures on Appveyor I haven't seen before:

================================== FAILURES ===================================
____________________________ test_connection_args _____________________________
Traceback (most recent call last):
  File "C:\projects\distributed\distributed\tests\test_security.py", line 200, in test_connection_args
    assert len(tls_12_ciphers) == 1
AssertionError: assert 0 == 1
  -0
  +1
______________________________ test_listen_args _______________________________
Traceback (most recent call last):
  File "C:\projects\distributed\distributed\tests\test_security.py", line 253, in test_listen_args
    assert len(tls_12_ciphers) == 1
AssertionError: assert 0 == 1
  -0
  +1

These checks were previous being skipped on Windows which was only run with Python 3.5

distributed/distributed/tests/test_security.py

Line 197 in 9429ffe

if sys.version_info >= (3, 6):

@jcrist if you have a moment would you be able to take a look at this?

[mrocklin]

Given that @jcrist is out, I wonder if @mariusvniekerk can take a look at the TLS failures in Appveyor here:

================================== FAILURES ===================================
____________________________ test_connection_args _____________________________
Traceback (most recent call last):
  File "C:\projects\distributed\distributed\tests\test_security.py", line 203, in test_connection_args
    assert len(tls_12_ciphers) == 1
AssertionError: assert 0 == 1
  -0
  +1
---------------------------- Captured stderr call -----------------------------
[{'alg_bits': 128,
  'description': 'ECDHE-RSA-AES128-GCM-SHA256 TLSv1.2 Kx=ECDH     Au=RSA  '
                 'Enc=AESGCM(128) Mac=AEAD',
  'id': 50380847,
  'name': 'ECDHE-RSA-AES128-GCM-SHA256',
  'protocol': 'TLSv1/SSLv3',
  'strength_bits': 128}]
______________________________ test_listen_args _______________________________
Traceback (most recent call last):
  File "C:\projects\distributed\distributed\tests\test_security.py", line 259, in test_listen_args
    assert len(tls_12_ciphers) == 1
AssertionError: assert 0 == 1
  -0
  +1
---------------------------- Captured stderr call -----------------------------
[{'alg_bits': 128,
  'description': 'ECDHE-RSA-AES128-GCM-SHA256 TLSv1.2 Kx=ECDH     Au=RSA  '
                 'Enc=AESGCM(128) Mac=AEAD',
  'id': 50380847,
  'name': 'ECDHE-RSA-AES128-GCM-SHA256',
  'protocol': 'TLSv1/SSLv3',
  'strength_bits': 128}]
============================== warnings summary ===============================

[mrocklin]

OK, so what should we do here? It seems like things are in an unpleasant state given that dask/dask has dropped 3.5, and so now our tests fail hard in some way. We don't seem to be able to find someone to debug familiar enough to debug TLS issues. Some options:

1. We can revert the 3.5 drop in dask/dask
2. We can live with the TLS failure, merge this, and hope that it's not significant
3. We can reach out to others that might know more about TLS and can help? (@TomAugspurger , is there anyone in Anaconda who can help here?)

[mrocklin]

@jacobtomlinson can you look at the windows CI failures here?

[jrbourbeau]

Yeah, I apologize, the current unfortunate state of things is my fault. I should have had PRs dropping 3.5 in both dask/dask and here ready to go concurrently instead of doing them one at a time.

My preference is to wait until the maintainer call tomorrow (maybe someone with TLS experience who has been pinged can investigate between now and then) and discuss this on the call. If nobody's able to debug, we mark the failing TLS tests with xfail on Windows, merge this, and open an issue.

[mrocklin]

OK, that sounds like a good plan to me. Thanks James

…

On Mon, Nov 4, 2019 at 2:57 PM James Bourbeau ***@***.***> wrote: Yeah, I apologize, the current unfortunate state of things is my fault. I should have had PRs dropping 3.5 in both dask/dask and here ready to go concurrently instead of doing them one at a time. My preference is to wait until the maintainer call tomorrow (maybe someone with TLS experience who has been pinged can investigate between now and then) and discuss this on the call. If nobody's able to debug, we mark the failing TLS tests with xfail on Windows, merge this, and open an issue. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#3179?email_source=notifications&email_token=AACKZTEF3RTK57T3RTIAPR3QSCSGTA5CNFSM4JGMRTUKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEDBAGQI#issuecomment-549585729>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AACKZTELXYERVZ4RLCFB6A3QSCSGTANCNFSM4JGMRTUA> .