[Numpy-discussion] Scheduling the 1.7.1 and 1.8 releases

[Nathaniel Smith]

On Thu, Apr 11, 2013 at 11:32 AM, Sebastian Berg
<sebastian at sipsolutions.net> wrote:
> On Wed, 2013-03-06 at 11:43 -0700, Charles R Harris wrote:
>> Hi All,
>>
> <snip>
>
>> The development branch has been accumulating stuff since last summer,
>> I suggest we look to get it out in May, branching at the end of this
>> month.
>
> Hey,
>
> maybe it is a bit early, but I was wondering. What are the things that
> should be done before branching? Most of the blockers should be
> resolved? I think the largest of that are maybe the deprecations (but
> since 1.7.0 is not too long ago many of them maybe don't need touching).

The "known blockers" list is here:
  https://github.com/numpy/numpy/issues?milestone=1&page=1&state=open

#3194: this has a PR and will be closed shortly.

#3008: this is fundamentally trivial, but annoying and confusing. I
*think* we understand generally what we need to do here, just need
re-arrange some header files so it works:
  https://github.com/numpy/numpy/issues/3008#issuecomment-14316385

#2905: trivial, needs doing

#2830: all this actually needs is an INSTALL.txt update.

#2801: need to rip NPY_CHAR out of master. trivial, needs doing.

#596: needs bumping to 1.9, and a check whether the deprecation
warning message explicitly says "1.8". if so then it should be made
more vague.

#456: same as #596

#378: need to merge #452.

#294: same as #596

> Aside from that, I think these are lurking/in-progress:
>
> Fix the piled up mapping.c things:
>     * This means resolving https://github.com/numpy/numpy/pull/436 .
>       It will be even worse now, but maybe at least most of it can
>       be put in without too much work?
>     * After that, non-integer deprecations should be done fully before
>       1.8. I think, but these need to change mapping.c too (the
>       most work is probably the tests...):
>         - https://github.com/numpy/numpy/pull/2891
>         - https://github.com/numpy/numpy/pull/2825
>     * (https://github.com/numpy/numpy/pull/2701)
>
> Are the 2to3 fixers supposed to be finished for 1.8? If yes there is
> probably no point in thinking about branching, etc. since back porting
> them all is just pain.

My feeling is that as a matter of policy, we should branch as soon as
the blockers are fixed, and anything that isn't ready can go into 1.9.
If you want a change into 1.8, you should hurry instead of making 1.8
(and everything else in it) wait ;-).

-n