Requirements gathering for new package source control

[Kevin Kofler]

Christoph Höger <choeger <at> cs.tu-berlin.de> writes:
> It would be a great thing, if I could control all that from my own
> versioned repository copy with some 'make <USUAL_STEP>' commands and
> finally merge my changes back especially when testing builds for
> multiple releases.

Why don't you just commit the changes even if you don't know yet whether they 
build? Especially for a small package where you're the sole maintainer, it 
won't really matter if CVS HEAD isn't always buildable.

For better or for worse, you'll also notice that many maintainers with many 
packages won't even bother building locally first, we usually just commit, tag 
and build, and after we get the failure mail from Koji, fix, commit, force-tag 
and resubmit, and repeat that until the build succeeds. Decentralized commits 
don't help with that at all (they only add an extra "push" step after the 
commit, which is just one more chance to screw up).

        Kevin Kofler