Rawhide stability

[Nicolas Mailhot]

Le jeudi 22 février 2007 à 10:37 -0500, Andrew Overholt a écrit :
> Thanks for your thoughts, everyone.  I'm glad rawhide stability isn't an
> issue.  I'm also in agreement that we shouldn't impose limits or
> anything which will slow down progress bleeding-edge-ness.

Actually, even though rawhide stability is not a huge issue (speaking as
someone who's been running rawhide systems continuously since ~ RHL
6-7), there are lots of room for improvement. Most of rawhide pains are
not difficult-to-diagnose stuff that requires days of analysis and would
slow progress if we tried to avoid it, but trivial mistakes, incomplete
rebuilds, bad interdependency tracking, etc that could be avoided with
better tooling/processes. Taking care of those would make rawhide hugely
more attractive for testers, and leave the difficult/subtle problems for
test releases. Focus is good.

Stuff that would make testers life easier:
- better rebuild staging (do not expose half-rebuilt repos, wait till
they're complete)
- trivial stuff hot-fixes (allow maintainers to push trivial fixes
before the nightly push when the fix is known and every tester complains
about the same problem)
- rollback of problem packages (put back the previous package when a
problem has been identified with the new one and we know it will take
days to fix)

I hate to write this, but even though rawhide should be allowed to eat
babies occasionally, this argument is often used to justify slacking on
small fixes. Letting small problems accumulate only means there is
little time left for difficult stuff during test phases.

-- 
Nicolas Mailhot
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 197 bytes
Desc: Ceci est une partie de message num?riquement sign?e
URL: <http://listman.redhat.com/archives/fedora-maintainers/attachments/20070224/bbbe74e0/attachment.sig>