Revisiting the mass rebuild plans for EPEL5

Thu Apr 12 18:26:25 UTC 2007

Hi,

FESCo today nacked the rebuild plan from voting "1" in
https://www.redhat.com/archives/epel-devel-list/2007-April/msg00055.html
See the end of this mail for details. In short: FESCo afaics didn't like
to leave release unchanged.

So I'd like to propose we switch back to the alternate plan (should
still be in the wiki at
http://fedoraproject.org/wiki/EPEL/SteeringCommittee/Voting
but the wiki is down currently aaics :-(  ) that I proposed in the vote.

That would mean: We announce when RHEL5 is on the builders to the public
and give people round about 72 (?) hours to rebuild their EPEL5 packages
manually on their own (in case they want to do it themselves).
Everything that didn't get rebuild gets a ".1" added to release in cvs;
that change gets committed, tagged and build -- we afaik have a script
that should be able to do that and I can take care of running it if
that's okay for everyone.

Does that sounds sane to everyone?

CU
thl

P.S.:here is the relevant part from the FESCo meeting:

19:42            --- | bpepple has changed the topic to: FESCO-Meeting
-- EPEL
19:42 <     bpepple> | anything in regard to EPEL need to be discussed?
19:42 <         jwb> | yes
19:42              * | thl send some notes from this week to the list
19:43 <         thl> | one hour or so ago
19:43 <       tibbs> | notting: If you do have any ideas about that,
please let us know.
19:43 <         jwb> | there was the buildroot issue Axel wanted acked
by FESCo
19:43 <     notting> | tibbs: get out baseball bats and beat upstream? :)
19:43              * | bpepple hasn't had a chance to read his e-mail
today. :(
19:44 <         thl> | bpepple, it's about the "mass-rebuild of EPEL5
now that we soon have RHEL5 final on the builders"
19:44 <         thl> | bpepple, it was voted to delete everything and
just rebuild
19:44 <     bpepple> | thl: ok.
19:44 <         jwb> | -1
19:44 <         thl> | everything, without chaning ENVR
19:44 <         f13> | er..
19:44 <         jwb> | yeah, -1
19:44 <         f13> | that may not bode well for clients whom already
ahve stuff installed
19:45 <       nirik> | (note: only EPEL-5... not 4)
19:45 <         thl> | f13, tell those that voted like that
19:45 <         f13> | as noted many times before, packages changing
checksums and such get messy
19:45 <         jwb> | f13, it was noted in their discussion.
apparently it didn't seem that big of a deal
19:45              * | thl disliked that plan, too
19:45 <         jwb> | are we voting on this yet?
19:45 <         f13> | *shrug*  I don't run rhel5 so I won't get
effected by it.
19:45 <         jwb> | f13, more than rhel5
19:45 <     bpepple> | jwb: Yeah, we should do a quick vote.
19:46 <       tibbs> | I'm still not understanding why you wouldn't want
to bump, and I read the IRC logs.
19:46 <         jwb> | tibbs, me either
19:46 <    dgilmore> | tibbs: becaue people did not want to fork the spec
19:46 <         jwb> | that is just lazy
19:46 <         thl> | jwb, +1
19:46 <         jwb> | you're pissing on your users because you don't
want to make a 2 character change
19:46 <    dgilmore> | i wanted to add a .1 and rebuild
19:46 <       tibbs> | Ah, that is a point, but I don't think it's a
terribly good point.
19:46 <         jwb> | dgilmore, that would be very acceptable
19:46 <     c4chris> | yea, .1 and rebuild
19:46 <     rdieter> | dgilmore: +1
19:46 <       tibbs> | The spec will diverge pretty much immediately anyway.
19:47 <         jwb> | right
19:47 <         thl> | dgilmore, why did you vote for deleting the
packges then?
19:47              * | thl is confused
19:47 <     notting> | ? you don't need to fork the spec. just b/c the
release changes, doesn't mean you have to build and push for older releases
19:47 <         jwb> | notting, fork it vs. the fedora spec
19:47            <-- | sankarshan has quit (Connection timed out)
19:47 <         f13> | notting: er, they have to bump the spec there,
but nowhere else, so now the specs are diverged
19:48 <     notting> | *horrors*
19:48 <       tibbs> | As I understand things, EPEL has no reason to
attempt to keep any kind of release ordering with Fedora.
19:48 <    dgilmore> | thl: i was confused by then.
19:48 <         f13> | not that I find anything _wrong_ with that.
19:48 <       tibbs> | So it's not even appending ".1"; just bump the
release.
19:48 <         f13> | nod
19:48 <         thl> | dgilmore, np, I was just confused now
19:48 <     notting> | thl: well, two issues. i'd be all for 'rebuild
and delete all old packages', but with a release bump
19:48 <         thl> | tibbs, some people prefer to appending ".1" ovefr
bumpin the release
19:48 <         f13> | is there a call for fesco vote?
19:48 <         thl> | I think they have a point
19:49 <         jwb> | f13, axel requested one
19:49 <         f13> | or a point 1
19:49 <         f13> | (:
19:49 <         thl> | notting, sounds fine for me
19:49 <     c4chris> | :-)
19:49 <       tibbs> | OTOH, not rebuilding at all seems to be working
for Fedora at this point.  What's the reason they absolutely must be
rebuilt?
19:49 < abadger1999> | tibbs: If they want to use the vanilla spec
later, using .1 lets them come back on the next Fedora Release rather
than the next upstream bump
19:49 <       nirik> | tibbs: they were build against beta1
19:49 <       tibbs> | abadger1999: Extremely good point.
19:49 <         f13> | abadger1999: but that actually overwrites history
19:50 <         f13> | unless they merge that .1 somewhere into the
history of hte FEdora spec
19:50 <       nirik> | abadger1999: yeah, changelog is lost then if you
merge
19:50 <       tibbs> | nirik: And we have .fc6 packages in F7; surely F7
diverges from FC6 more than rhel5b1 diverges from rhel5release.
19:50 <         thl> | f13, is that really a big problem if it was just
a "rebuild" in the chanelog?
19:50 <     notting> | dgilmore, this is only rebuilding things actually
built for EPEL, not everything in EPEL cvs, right?
19:50 <         thl> | notting, yes, only what has been build up to now
19:51 <         f13> | thl: it's not a really big problem, but I
generally don't like to see history get stomped
19:51 <       nirik> | tibbs: yeah, you would think so... dunno for sure.
19:51 <         thl> | f13, agreed; I think in this case it's still not
nice, but acceptable
19:51 <         f13> | and who k nows what happens with the rebuild,
something may end up needing changed to build again against RHEL5 GA
19:51 <         f13> | tibbs: you'd be surprised what all changed from
B1 to GOLD
19:52              * | rdieter thinks we're not here to (re)make epel's
decision for them (or not?), just ack or nack it.
19:52 <         f13> | -1
19:52 <         f13> | (for their current plan)
19:52 <         jwb> | -1
19:52 <     c4chris> | (plan == rebuild and no bump, right)
19:52 <         jwb> | rdieter, but we can nack with a suggested improvement
19:52 <         thl> | c4chris, yes
19:52 <         f13> | c4chris: yep
19:53 <     bpepple> | c4chris: correct.
19:53 <     c4chris> | k, -1 then
19:53 <     notting> | -1
19:53 <     bpepple> | -1 here also.
19:53 <       tibbs> | Yeah, I hate to be an obstruction, but -1 to
rebuilding with no bump.
19:53 < abadger1999> | -1
19:53 <         thl> | jwb, I can take care of that if you want; i was
against this in any case ;-)
19:53 <     bpepple> | so it looks like we against EPEL suggested plan.
19:53 <         jwb> | thl, great
19:54            <-- | gregdek has quit (Read error: 104 (Connection
reset by peer))
19:54 <         thl> | bpepple, I'll get that out to epel and will take
care of it
19:54 <    dgilmore> | notting: yeah just whats built
19:54 <     bpepple> | thl: great, thanks.