_readdir_pyx extension cannot be compiled with MSVC

[Talden]

>> John, you just released 1.7.1. If this was fixed in trunk, why it's not
>> in the release?
>> It's annoying. Sorry, but it's really annoying.

> Because we didn't know the problem existed. I had 1.7.1rc1 up for a
> week, which is the perfect time to test this sort of thing and make sure
> it compiles everywhere with all compilers.
>
> It isn't really possible for one person to do this, which is why we have
> release candidates. To give other people time to test what will become a
> final release.

Can we not we a more formal signed/unsigned release process.

We take a candidate release through a waiting period during which we
collect signatures detailing success in a defined set of combinatorial
testing.

Assuming we reach the end of the RC period without needing another RC -

   If we reach the minimal combinatorial set we make a signed release.

   If we don't reach the minimal combinatorial set we make an unsigned release.

In both cases we list the combinatorial testing that was reported in
release notes.

We don't delay release. Though if we repeatedly fail to meet our
testing targets we either revisit them or spend a cycle getting that
done.

Some people may decide only to use signed releases in production.

Signatures could initially be accepted from any of the full
committers.  We should get potential signers to detail the
combinations they feel likely to be able to test since, as not every
tester is going to be available to test every release, gaps and
limited redundancy in release testing coverage is made visible.

People interested in helping the project but unable provide coding
effort might feel inclined to help fill in testing gaps for the
project.

--
Talden