config.guess and config.sub are generated by autogen.sh and the automake tools, so no changes to CVS are required. I downloaded the latest versions.
In the CVS repo, make sure configure.ac has the version set to 0.17.4,
and also add an entry into ChangeLog saying "released version 0.17.4"
and commit these changes.
Hopefully at this point, the CVS repo has everything it needs, and won't
need to change again before the release. That being the case, I would
set a v0-17-4 tag in the repo. That way, years from know we'll know
exactly which code was used for the 0.17.4 release.
Next, I'd delete my tn5250 sandbox and check out a fresh one from CVS
using the v0-17-4 tag -- this guarantees that you're using the code you
think you are :)
I'll do all these steps after we have verified that the code is what we actually want to release. In light of that, I've created a potential release tarball by fixing the slang Makefile.am and running 'make dist'. Could you and anyone else please check that this behaves the way we expect? Download this test tarball at:
If there are any problems with the build, I'd like to fix them now before making any change to the CVS tags.
autogen.sh, configure, make, make dist and make distcheck to verify that
it's all healthy. (Though, I just tried this, and it's not healthy.
SLang support appears to still be broken from when you reorganized the
I made a small commit to CVS to fix the SLang issue as far as make dist and distcheck go, though trying to build with slang support active still doesn't work. I don't know if it is because of the reorganization or if the newer SLang libraries I have on my machine don't work with the rather old slang code we have.
to summarize, this is what I think the build process (for a release)
1. Get the latest config.sub & config.guess
2. Verify the versions & add a "version XXX" released message to ChangeLog
3. Commit Updates from Step 2.
4. Set the v0-17-4 tag.
5. Update the configure.ac file with 0.17.5 for the version so it's
ready for future development, and commit that.
6. Check out a fresh CVS from tag v0-17-4
7. cd tn5250 && ./autogen.sh && ./configure
8. make dist
9. make distcheck
10. Send the output from make dist (tn5250-0.17.4.tar.gz) to me
11. I'll verify that the release works on FreeBSD (I'm assuming you
verified it on Linux)
12. I'll build Windows binaries.
13. I'll upload your tarball, plus the Windows binaries to Sourceforge.
If you can verify that my test tarball behaves the way it should then I agree that these are the right steps to take.
I don't know how to create/test RPMs and no longer have easy access to a
Linux box, so I can't help with that. If nobody volunteers, I agree
that we shouldn't worry about RPMs. If someone does volunteer, I'll
send that person the tarball after verifying that it works on FreeBSD,
and I'll wait for the RPMs before uploading to SF.
Considering the amount of traffic we get on this list, I think we need to set a deadline for which we will wait for replies. For example, if no one responds by the end of the week then we just assume no one is going to.
Don't worry about FreeBSD packages, we don't normally keep those on the
TN5250 site, rather FreeBSD takes care of them.
Ok, I won't worry then.
if you want to understand why that is, there are many good books on
the design of operating systems. please pass them along to redmond
when you're done reading them :)
- Paul Davis on ardour-dev