Re: Slight simplification to README-release

[Reuben Thomas]

On 6 October 2011 21:04, Bruno Haible <address@hidden> wrote:
> Reuben Thomas wrote:
>> the expectation is that you already have all tests
>> passing before you start the release procedure.
>
> Exactly. It makes no sense to work on the packaging details of
> code that doesn't even pass its self-tests. So, if the developer
> is not yet in the phase where releasing is appropriate, he will
> waste time by doing "make distcheck": because "make distcheck"
> will take some time running "make dist", "configure", and "make",
> before it comes to the "make check" that will fail.

This is README-release! A developer who is not yet ready to release
should not be following the procedure of README-release.

I'm trying to compress the currently long and tortuous README-release,
which I never manage to get through without messing up, into a short
and simple procedure that can eventually [English, not French sense!]
be automated. It looks as though this should be possible: a series of
checks are made, followed by the release tagging and tarball making.
It's then left to the developer to upload and announce it (and there's
help with that too). It seems reasonable to me to assume that the
developer following this procedure considers their project to be in a
fit state to release, so while we should try to make the procedure
fail as early as possible if it's not going to work, and hence leave
the longest-running tasks until last, that should not come at the
expense of simplicity.

-- 
http://rrt.sc3d.org