[DISCUSS] Removing Jython

[Stephen Mallette <sp...@gmail.com>]

If you build the gremlin-python module you will get an error related to
pypi. Robert Dale was good enough to provide me the link that help firm up
in my mind something I'd already run into on the last release with pypi:

https://mail.python.org/pipermail/distutils-sig/2017-October/031712.html

This breaks our build as Jython - perhaps more specifically the Jython
build plugin - is not updated to support this change. Given that there's
been no activity on that plugin in the last 12 months, I'm not sure we can
expect to see a fix quickly. While I looked into what might be necessary to
offer a patch, I wondered why we need the added build of Jython at all. It
was necessary when we first got started with gremlin-python because we
didn't have a good native build, but now we have a really nice one and
actually test both Python 2 and 3. And now with the work I've been doing to
build the GLV test suite we can get full round trip testing of Gremlin in
native python which has already proven to be more effective than the Jython
stuff as I've uncovered python bugs that the Jython tests didn't.   If we
dropped the jython build we could save some build processing time too.

So, to summarize, I'm wondering if there is any reason anyone can think of
to keep the Jython build around given the good testing we have with the
native python system. I don't want to rush this discussion, but given that
we can't currently build our software, this is a fairly critical situation.
I will immediately "ignore" the Jython build sequence and CTR that in so
that we can at least build without error. If there are no objections in the
next 72 hours I'll assume lazy consensus and do a full removal. Then when
the GLV test suite is ready we can all feel comfortable again.

Re: [DISCUSS] Removing Jython

[Stephen Mallette <sp...@gmail.com>]

Sorry - one more post, I hit send too soon. I also realized that the next
version of Jython (the dependency) must have a bug or incompatibility of
some sort. When I tried to go to that newer version one of our tests
started failing (works fine on the current version). Anyway, not sure what
kind of trouble that will be in the future, but just thought I'd mention it
here.

On Fri, Oct 27, 2017 at 2:39 PM, Stephen Mallette <sp...@gmail.com>
wrote:

> Just a quick update, gremlin-python is now building on tp32 and master
> without the jython enabled build. Note that the intent here is NOT to
> remove the jython dependency, but to remove the jython aided build of the
> python code. Doing that had the nice effect of speeding up our build a bit
> - we gained back about 1.5 minutes i think.
>
> On Fri, Oct 27, 2017 at 12:45 PM, Stephen Mallette <sp...@gmail.com>
> wrote:
>
>> If you build the gremlin-python module you will get an error related to
>> pypi. Robert Dale was good enough to provide me the link that help firm up
>> in my mind something I'd already run into on the last release with pypi:
>>
>> https://mail.python.org/pipermail/distutils-sig/2017-October/031712.html
>>
>> This breaks our build as Jython - perhaps more specifically the Jython
>> build plugin - is not updated to support this change. Given that there's
>> been no activity on that plugin in the last 12 months, I'm not sure we can
>> expect to see a fix quickly. While I looked into what might be necessary to
>> offer a patch, I wondered why we need the added build of Jython at all. It
>> was necessary when we first got started with gremlin-python because we
>> didn't have a good native build, but now we have a really nice one and
>> actually test both Python 2 and 3. And now with the work I've been doing to
>> build the GLV test suite we can get full round trip testing of Gremlin in
>> native python which has already proven to be more effective than the Jython
>> stuff as I've uncovered python bugs that the Jython tests didn't.   If we
>> dropped the jython build we could save some build processing time too.
>>
>> So, to summarize, I'm wondering if there is any reason anyone can think
>> of to keep the Jython build around given the good testing we have with the
>> native python system. I don't want to rush this discussion, but given that
>> we can't currently build our software, this is a fairly critical situation.
>> I will immediately "ignore" the Jython build sequence and CTR that in so
>> that we can at least build without error. If there are no objections in the
>> next 72 hours I'll assume lazy consensus and do a full removal. Then when
>> the GLV test suite is ready we can all feel comfortable again.
>>
>
>

Re: [DISCUSS] Removing Jython

[Stephen Mallette <sp...@gmail.com>]

Just a quick update, gremlin-python is now building on tp32 and master
without the jython enabled build. Note that the intent here is NOT to
remove the jython dependency, but to remove the jython aided build of the
python code. Doing that had the nice effect of speeding up our build a bit
- we gained back about 1.5 minutes i think.

On Fri, Oct 27, 2017 at 12:45 PM, Stephen Mallette <sp...@gmail.com>
wrote:

> If you build the gremlin-python module you will get an error related to
> pypi. Robert Dale was good enough to provide me the link that help firm up
> in my mind something I'd already run into on the last release with pypi:
>
> https://mail.python.org/pipermail/distutils-sig/2017-October/031712.html
>
> This breaks our build as Jython - perhaps more specifically the Jython
> build plugin - is not updated to support this change. Given that there's
> been no activity on that plugin in the last 12 months, I'm not sure we can
> expect to see a fix quickly. While I looked into what might be necessary to
> offer a patch, I wondered why we need the added build of Jython at all. It
> was necessary when we first got started with gremlin-python because we
> didn't have a good native build, but now we have a really nice one and
> actually test both Python 2 and 3. And now with the work I've been doing to
> build the GLV test suite we can get full round trip testing of Gremlin in
> native python which has already proven to be more effective than the Jython
> stuff as I've uncovered python bugs that the Jython tests didn't.   If we
> dropped the jython build we could save some build processing time too.
>
> So, to summarize, I'm wondering if there is any reason anyone can think of
> to keep the Jython build around given the good testing we have with the
> native python system. I don't want to rush this discussion, but given that
> we can't currently build our software, this is a fairly critical situation.
> I will immediately "ignore" the Jython build sequence and CTR that in so
> that we can at least build without error. If there are no objections in the
> next 72 hours I'll assume lazy consensus and do a full removal. Then when
> the GLV test suite is ready we can all feel comfortable again.
>