On Fri, Apr 14, 2006 at 10:33:20PM -0500, Jerry Amundson wrote:
> Is this going to be the direction for the lwp, rvm, and coda modules too?
> My reason for asking is I've been working on spec file changes for all
> four, not realizing until today that CVS commits have been done. I'm adding
> fairly involved changes for requirements, and feel that I'll be trying to
> hit a moving target by continuing. (I'll subscribe to "changelog" to
> prevent this in the future...)
Possibly, but I shouldn't have much trouble merging, I would probably
simply replace my version of the spec files.
> Regarding independence from configure, I think this should be taken a step
> further - have bootstrap.sh edit the spec file in the top of the source
> tree (or move it there), and not under pkgs/tools. This is where "rpmbuild
> -ta" expects it to "rpm directly from tarball", and has the benefit of
> removing the need of the src.rpm.
A released tarball is actually not a simple CVS checked out copy. I use
some scripts, which actually do a clean build first, and only construct
the tarball if that works.
i.e. something like the following,
(actually a bit more involved since I do this initial build in a
chrooted tree to make sure the build is as clean as possible).
#!/bin/sh -e
cvs checkout $comp
cd $comp
./bootstrap.sh
./configure --prefix=/usr
make
make distclean
make dist
This results in a tarball that is copied over to a couple of machines
where I try to build the result on different distributions/OS's.
So as long as make distclean leaves the rpm.spec file around, and make
dist includes the spec file in the tarball we shouldn't have a problem
for people who try to build RPMs from the released source tarballs.
This is what that CVS commit tries to achieve.
> Let me know so I can submit my changes, or hold off and work with what I
> have.
I'd say, keep on trucking.
Jan
