drop jython support in 3.0.0?

Question

drop jython support in 3.0.0?

cjw296 opened this issue 5 years ago · comments

@rbtcollins / @mfoord / @voidspace - thoughts?

We have no CI for it, so it's likely to have bit-rotted...

Michael Foord · Answer 1 · Tue Apr 30 2019 18:28:44 GMT+0800 (China Standard Time)

Without CI we can't guarantee it. As far as I recall it was no extra work, but unless someone will test/maintain it I'm happy to drop official support. (No idea how I ended up with two GitHub users.)

Chris Withers · Answer 2 · Tue Apr 30 2019 18:36:57 GMT+0800 (China Standard Time)

Yeah, I can't find any official or support docker images for Jython, so...

Berker Peksag · Answer 3 · Tue Apr 30 2019 19:37:40 GMT+0800 (China Standard Time)

+1 for dropping it. I've never run mock's test suite on Jython before, so I guess it may have already been broken.

Karthikeyan Singaravelan · Answer 4 · Tue Apr 30 2019 20:49:14 GMT+0800 (China Standard Time)

There is no straightforward to run tests in Jython from Travis like pypy as far as I have searched around. The only example was astroid that downloaded and built it themselves. They also removed it from their CI later. Jython also doesn't support Python 3 yet so I am not sure if it's worth the effort given Python 2 EoL.

pylint-dev/astroid@3591182

Chris Withers · Answer 5 · Tue Apr 30 2019 20:50:59 GMT+0800 (China Standard Time)

Well, I'm planning to move the backport to CircleCI for various reasons, so if we had a stable, publicly maintainer docker image, that would do it.

But, as you say, if they have no Py3, then I think we're safe to drop it.

Chris Withers · Answer 6 · Wed May 01 2019 15:54:01 GMT+0800 (China Standard Time)

Okay, 'tis done, ed3bc8b.

Robert Collins · Answer 7 · Wed May 01 2019 18:27:18 GMT+0800 (China Standard Time)

I don't think we should drop the Jython glue; it is platform checked to not affect other interpreters and has approx 0 maintenance overhead for us. The whole point of the backport is to make current mock features more widely available.

Chris Withers · Answer 8 · Wed May 01 2019 18:34:44 GMT+0800 (China Standard Time)

Anyone who needs this can use 2.0.0 though, right?

I guess my point in saying that is that we're offering a backport that is maintained; I have no idea whether the current master of the backport will blow up when used with Jython or not, and it makes me uncomfortable keeping untested code around that's an indication it might be an expectation.

Robert Collins · Answer 9 · Wed May 01 2019 18:56:35 GMT+0800 (China Standard Time)

Kindof defeats the purpose of doing the backport is all. We can equally say that folk can just use the mock in their Python interpreter's stdlib. We do the backport to allow projects that use mock to use newer features without having to mandate a very latest release of CPython; to a large degree we should be driven by our users, and in this case its worth remembering that we're a second order dependency: a project with developers or deployers using Jython may depend on mock working there, without it enabling us to CI it. We'll know its broken if they report bugs, but we can't tell the difference between 'not used' or 'working fine' with the feedback systems we have in place today.

…

-Rob

Chris Withers · Answer 10 · Wed May 01 2019 19:00:22 GMT+0800 (China Standard Time)

Totally agree, and if we see evidence of Jython users or Jython users complaining that things no longer work, we can very easily revert the one commit above and bring it back :-)

Robert Collins · Answer 11 · Wed May 01 2019 19:01:34 GMT+0800 (China Standard Time)

Sure, I'll leave it with you; my bias is to not remove it in the first place, because then there's no interrupt to run around and un-break, but if you're up for that, all good.

Karthikeyan Singaravelan · Answer 12 · Wed May 01 2019 19:20:14 GMT+0800 (China Standard Time)

Not sure if this is the issue to discuss this but is it agreed upon to remove it in upstream unittest.mock too? I feel upstream CPython has a larger userbase. Though there are no signs of Python 3 support for Jython or CI in CPython for Jython I think removal could at the least have a NEWS entry/whatsnew/porting section entry.

Karthikeyan Singaravelan · Answer 13 · Wed May 01 2019 19:21:07 GMT+0800 (China Standard Time)

Sorry forgot to paste the merged PR link : python/cpython#13033

Robert Collins · Answer 14 · Wed May 01 2019 19:23:05 GMT+0800 (China Standard Time)

The upstream one gets copied into Jython when they update their stdlib, so if its still needed at all, they'll still need it there. I see value in keeping the delta low, so yeah, removing it upstream makes sense to me too.

…

-Rob

On Wed, 1 May 2019 at 23:20, Xtreak ***@***.***> wrote: Not sure if this is the issue to discuss this but is it agreed upon to remove it in upstream unittest.mock too? I feel upstream CPython has a larger userbase. Though there are no signs of Python 3 support for Jython or CI in CPython for Jython I think removal could at the least have a NEWS entry/whatsnew/porting section entry. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#453 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AADZ7XRNQTNNTFQIEEJP7KDPTF4G5ANCNFSM4HJJ4BAA> .