[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Strictness

From: Ralf Wildenhues
Subject: Re: Strictness
Date: Tue, 14 Aug 2007 00:54:37 +0200
User-agent: Mutt/1.5.13 (2006-08-11)

* Carl Fürstenberg wrote on Mon, Aug 13, 2007 at 12:53:39AM CEST:
> On 8/13/07, Robert Collins <address@hidden> wrote:
> > On Sat, 2007-08-11 at 22:06 +0200, Carl Fürstenberg wrote:
> > > On 8/11/07, Noah Slater <address@hidden> wrote:
> > > > > I think you misunderstanding me, it's the generation if the changelog
> > > > > that will take too long time.
> > > > How long does it take to generate?
> > >
> > > it all depends on subjects connection and computer power, but for a
> > > normal person, perhaps 10 min.

Wow.  That's a long time.

Nevertheless, if you release your software as a distribution tarball,
then that tarball will not contain version control meta data.  In that
case, it should contain change descriptions in some other form, the
ChangeLog file(s) being a suitable (and the preferred GNU) form.

However, the way things are now, the check for ChangeLog is done at
automake run time, not at 'make dist' time.  So, if you don't ever do
source code releases, or if you do, but you only want to invest those
10 min when releasing/release testing, we could think about relaxing
the test for a ChangeLog file until 'make dist' time, or omit the test
if the Automake option 'no-dist' has been used.

None of these changes seem particularly convincing, as they seem
somewhat ad-hoc; the next package has different requirements again.

> NEWS: this is one of the most empty files. As source code often are
> part of a bigger "project", usually it's more logical to have news
> binded to the project instead of the tar ball.

Why not
  echo "Please see the ../../NEWS file for changes of this package." > NEWS

> AUTHORS: This might from a legal standpoint be needed, but I would
> like sometimes to merge it with COPYING to a COPYRIGHT file.

See above.  It's humans that parse these files, so giving a pointer is a
helpful thing.

> README: A really important file, I would say, in --gnu, it should
> check that the file is not empty.

Then people who override the policy will add a dot, or other nonsensical
stuff.  I don't think these checks can ever work semantically, all they
can help with is unintentional omission.

> INSTALL: Here is the problem file with --foreign. with gnu, and
> --add-missing, it will install a generic file, but for --foreign, the
> upstream must craft an own, thum many chooses to add empty "gnu
> required" files, just to get INSTALL to be generated.
> As it is now, either you are a hardcore gnu, or a total outsider.

Hmm, maybe automake should have a way to install some of its files only.
Or maybe just 'automake --print-lib-dir' would be sufficient, so you can
copy the file yourself.  Unsure.


reply via email to

[Prev in Thread] Current Thread [Next in Thread]