Re: Patch for sawfish.spec.in

> I may not have made it clear - the Fedora spec in the tarball (loaded 
> from the .src.rpm) was clearly NOT derived from the sawfish.spec.in
> in the tarball.

Actually it's vice versa - I didn't copy those ChangeLog-changes etc. to
sawfish repo while updating Fedora's spec. Those are not that
important and I'm lazy... 

> >> +/bin/sh autogen.sh --nocfg
> >
> > As far as I remember, autogen was only needed for snapshots.
> 
> True - there appears to be some tweaking of the snapshot on its way
> to the release, generating configure and a Makefile, and NEWS below.

Yes, and that is the problem here. Spec file should not re-generate
files unless really needed. What I meant in my previous message was
that Fedora's spec is for release-tarballs only and therefore it
doesn't run autogen to re-generate configure etc. Regenerating
stuff in configure / Makefile is normal.

> > 1) There should be some kind of test is autogen is needed or not.
> 
> Sounds good, but how?   I'm a mere dabbler in rpmbuild specs.

Normal shell commands are OK, so something like:

[ ! -f configure ] || ./autogen.sh --nocfg

> > 2) Don't update NEWS.
> 
> NEWS already exists as the .texi, why shouldn't it pass to the
> install? The release tarball includes it, and it in fact includes
> instructions on how to generate it from the .texi :-)

Don't re-generate stuff in spec... Again something:

[ ! -f NEWS ] || /usr/bin/makeinfo --no-validate --no-headers
man/news.texi > NEWS

> >> +* Tue Apr 22 2014 Allan Duncan
> > This line is missing version number.
> 
> I thought about that and decided that it is up to the maintainer to
> bump the version if it is
> warranted, so the version number is unchanged from below.  Should I
> have included a repeat?

Yes, you must repeat the current version.

-- 
Sawfish ML