This is the mail archive of the
mailing list for the Cygwin project.
Re: Pending packages status
Igor Pechtchanski wrote:
Worse, my pretest versions of libtool are based on *different* CVS
snapshots. So they differ not only in REL, but also in VER, from the
packages on the cygwin mirrors.
Umm, Chuck, the above suggestion was intended only for different
pre-releases of the package with the *same* VER number. If you have
different VER numbers, you already have a way of distinguishing various
pre-releases, and no need to do anything extra to that end.
True -- but I brought it up simply to emphasize that the "problem space"
is larger than I sensed was being considered. I wanted to head off the
argument that pre-releases (or pretests) following a given version be
numbered using tags to the preceeding release.
e.g. foo-1.3.2-3 is official
Somebody was SURE to suggest that pretests for -4 be named
"foo-1.3.2-3a" -3b, -3c, etc.
so, I was trying to point out the problem THAT naming scheme runs into
when I'm working on pre-release versions of foo-1.3.3. Should they
follow the -1.3.2-3X rule (since they 'succeed' 1.3.2-3), or the -0.x
rule (since they will precede the first official release of 1.3.3)?
BLECH. *Yet another amendment to the rule*. All because it's too
difficult to write rules that cover every conceivable case. Sometimes,
it's okay to just say "Rules? We don' need no steenking rules!" and
just tell folks to use their brains. As cgf says, maintainers are
supposed be a cut above, and should be able to handle these issues in
the way most appropriate for THEIR package. [Yes, I also recognize the
political implications of this statement; you *don't* want to get me
started on that. I might start quoting the Federalist papers and George
Washington's farewell address -- and nobody wants that; incl. me. I've
got real work to do...]
Heck, they might use one method for their foo package, and a different
method for their baz package. Fine by me -- as long as it makes sense.
I'm sorry I didn't fully explain what I was trying to get across with
that libool-DATE paragraph; I had hoped the problems would be clear by
implication, so I wasn't explicit. I should have been.
FWIW, I think the practice of naming the initial releases -1 is related to
the absense of release notes for packages in setup. If there were a way
to access the release notes (or the announcement, which should amount to
the same thing), it wouldn't matter what the release number is. This is
more than just a "so patch setup" issue, since there is no connection
currently between upset/setup and the announcements.
Well, that's a whole 'nother issue. I have some ideas, but they involve
upset changes AND setup changes AND automated monitoring of
cygwin-announce. Icky stuff which I have no intention of coding, so
I'll just shut up now.