scoder

Given that Py3.4 is now out of maintenance, and that the test suite takes excessively long to run in Py3.4, I strongly favour removing also the support for Py3.4 in...

I don't think it's something for right now. There will be more changes along the way for Cython 3.0, and Py2 support will only be removed afterwards. Better keep the...

Update: we missed 2019, and will hardly make 2020 for the Cython 3.0/1 releases. Since Python 3.5 is also officially out of maintenance now, I think it's best to remove...

Whatever the Python devs say about their support intentions, here are the PyPI download starts, averaged over the last month: ``` 2.6 : 12.8 / day ( 0.0%) 2.7 :...

Looking back through the ticket communication, I noticed that we basically already [decided to remove Py3.5 support](https://github.com/cython/cython/issues/2800#issuecomment-703072864), also for Cython 3.0, and to make 2.7/3.6 the minimum versions. That was...

> I strongly believe that _intersection_ of users on very old Python versions like 2.7 expecting and using new cython features is very small. I rather think that allowing users...

It's ok to reject `cpdef int func() -> int: ...` with a parser error. It's simply confusing to write something like this. `cpdef func() -> int: ...` is ok, though,...

> ```cython > IF INT_MIN != -INT_MAX: > ``` This is decidedly _NOT_ a use case for conditional compilation. It should be handled at C compile time.

To clarify, what takes up code space is the separate calculation of constant values in the parser (i.e. `DEF`). The conditional statements (' IF' etc.) are trivial compared to that....

I added a deprecation warning for both `DEF` and `IF`, as well as some explanations and examples in the docs. Please keep commenting in this ticket if you have a...