EPEL/Meetings/20070318

From FedoraProject

Jump to: navigation, search

Contents

Meeting 20070318

Attending

Summary

The current proposal works with just two repos and suggested to build packages for devel/testing (against the latest updates that might become the next quarterly update) and then build the pacakge anew for stable if everything looks fine

Please discuss on the list!

Note: there are some takes on the schedule at http://fedoraproject.org/wiki/EPEL/Schedule that could need some help.

Full log

00:00            --- | thl has changed the topic to: EPEL SIG meeting
00:00 <         thl> | ping EPEL SIG members
00:00 <         thl> | who's around?
00:00              * | quaid is here
00:00              * | thimm is
00:00 <     entr0py> | bjohnson here
00:00              * | nirik is here.
00:00            --- | thl has changed the topic to: EPEL SIG meeting -- http://fedoraproject.org/wiki/EPEL/Schedule
00:01 <         thl> | hi everyone
00:01              * | kanarip is here
00:01 <         thl> | dgilmore mentioned it would be likely that he could not join us today
00:01 <       thimm> | :/
00:01 <       thimm> | Isn't he the only one that insists on Sunday meetings?
00:02 <       nirik> | yeah, sunday or off business hours during the week...
00:02 <         thl> | nirik, yes, something like that was what dgilmore wants
00:03 <         thl> | and even if someone asks for this is doesn't mean that he has free time each week ;-)
00:03 <       nirik> | true...
00:03 <       thimm> | So, how about we move the meeting time into the work week?
00:03 <         thl> | I pinged mmcgrath in #fedora-devel
00:03            --- | thl has changed the topic to: EPEL SIG meeting -- http://fedoraproject.org/wiki/EPEL/Schedule -- meeting time
00:04 <         thl> | thimm, well, "off business hours" in the US means probably middle of the night for us europeans
00:04 <       nirik> | thimm: where was the link to that wiki timeschedule page you made? I was meaning to look at it and add, perhaps we could see what time works best based on that?
00:04 <       thimm> | Isn't he down-under?
00:04 <       thimm> | http://fedoraproject.org/wiki/EPEL/NewMeetingTime
00:04 <         thl> | thimm, dgilmore? no
00:04 <       thimm> | But noone entered anything in there
00:04 <         thl> | he comes from down under iirc, but is in the US for some years now
00:05 <       nirik> | do we have any west coast people at all? perhaps we could use that time as well?
00:05 <       nirik> | (for now)
00:05 <       nirik> | yeah, I think he's in MN.us.
00:05 <         thl> | quaid, are you on the west coast?
00:05 <       quaid> | I am
00:06 <       nirik> | ah, sorry, thought there weren't any west coast folks. :) My mistake
00:07 <       quaid> | we used this to figure out likely slots for our far-flung FDSCo members:
00:07 <       quaid> | http://fedoraproject.org/wiki/BobJensen/MeetingTimes
00:07 <         thl> | well, I stated my preferred times on the list in https://www.redhat.com/archives/epel-devel-list/2007-March/msg00203.html
00:07 <       quaid> | each person put in their unavailability by their initial, and we looked for the ligthest coverage
00:07 <       thimm> | similar to what we did with FPC
00:07 <       quaid> | thl: right, we tried that for a while, but it failed to make it obvious when the best times were
00:08 <       quaid> | so, if passing ideas back and forth doesn't help, me may need a matrix like this
00:08 <       thimm> | See http://fedoraproject.org/wiki/Packaging/NewMeetingTime
00:08 <       thimm> | This is what I used as a template for http://fedoraproject.org/wiki/EPEL/NewMeetingTime
00:08 <       nirik> | I would prefer to have dgilmore around, since he is dealing with so much of the infrastructure, but most times are ok with me otherwise.
00:09 <       nirik> | thimm: sundays are usually bad for you?
00:09 <         thl> | nirik, +1, we need to make sure the certain key driers so far can join the meeting normally
00:09 <       thimm> | Yes
00:09 <       thimm> | Sundays seems to be bad for many people including @RH.com
00:09 <       nirik> | yeah, true...
00:10 <       nirik> | from dgilmore's email to the list: " It would need to be between 23:00-3:00 UTC  or 11:30-12:30 UTC or on the weekends."
00:10 <       thimm> | We'd like more participation from the enterprise world and people with jobs like to work on Mo-Fr
00:10            --> | che (Rudolf Kastl)  has joined #fedora-meeting
00:10 <         thl> | thimm, we fist need to lift of...
00:10 <       nirik> | thimm: agreed.
00:11 <       thimm> | thl: If we close doors, how will we lift off?
00:11 <         thl> | anyway, can somebody fill out http://fedoraproject.org/wiki/EPEL/NewMeetingTime with what we know so far
00:11 <         thl> | then we can ask the others to put up their time
00:11 <         thl> | and then we can look further
00:11 <         thl> | thimm, it's not that bad currently
00:12 <         thl> | and those that showed most interested seems to agree on sunday
00:12 <         thl> | for now
00:12              * | thimm looks around
00:12 <       nirik> | I think we are left with sunday or not having dgilmore available at the meetings.
00:12 <         thl> | nirik, +1
00:12 <       thimm> | nirik: Or both like today ...
00:13 <       thimm> | I won't be able to make it on other Sundays
00:13 <       thimm> | This is my first Sunday I could make it anyway
00:13 <       nirik> | what about saturday?
00:13 <       thimm> | Well, weekends in general are bad, but Saturday early (in UTC speech) are the lesser evil
00:14 <       nirik> | perhaps we can propose that to dgilmore. Or just pick a weekday meeting time and move on.
00:15            --> | engwnbie (Leo Canale)  has joined #fedora-meeting
00:15              * | thl still votes that somebody that pre-fills http://fedoraproject.org/wiki/EPEL/NewMeetingTime with what is publically known; then we ask others to add their info, and then we'll look at it next week
00:15 <       thimm> | It would make sense for everyone interested to fill out his own slots
00:15 <       thimm> | That's how it worked at the FPC
00:16 <       thimm> | Cross out what you can't attend and ++ the lsots you prefer
00:16 <         thl> | thimm, go ahead and to it
00:16 <       thimm> | s/lsots/slots/
00:16 <         thl> | thimm, somebody needs to start
00:16 <         thl> | so others know how it shall look like
00:16 <         thl> | add mine and dglimores times, too
00:16 <         thl> | and others will join
00:16 <       nirik> | in my case as long as it's not fesco and at times I am awake it's ok with me. ;)
00:17 <         thl> | let's fill out the form until next week and then look at it?
00:17 <         thl> | and move on now?
00:18 <       nirik> | ok
00:18 <         thl> | somebody just needs to start afaics
00:18              * | thl will move on
00:18            --> | glezos (Dimitris Glezos)  has joined #fedora-meeting
00:18            --- | thl has changed the topic to: EPEL SIG meeting -- http://fedoraproject.org/wiki/EPEL/Schedule -- RHEL5 on the builders
00:19 <         thl> | we have some issues getting RHEL5 in place
00:19 <         thl> | but they hopefully sort out soon
00:19 <       nirik> | I think dgilmore was working on that. ;)
00:19 <         thl> | dgilmore, yes, he and mmcgrath are working on it
00:19 <       nirik> | FYI, centos has a 4.92 beta out to test with...
00:19 <         thl> | the big problem is:
00:19 <         thl> | do we need one merged tree
00:20 <         thl> | or different ones for Client and server
00:20 <         thl> | I'd say:
00:20 <       thimm> | thl: OK, my part is done
00:20 <       nirik> | I think we agreed on one merged tree in the past. I think Client and Server would make things too complicated.
00:20 <         thl> | nirik, +1
00:20 <         thl> | nirik, neertheless we need to split stuff somehow
00:20 <         thl> | otherwise Client users will run into missing dep issues
00:20 <       nirik> | split how?
00:21 <         thl> | if a pacakge requires something from the server
00:21 <         thl> | split the repo into Client (Desktop, VT, Workstation, VT+Workstation) and Server
00:21 <         thl> | or use a yum plugin
00:22 <         thl> | that prevents that people run into missing dep issues
00:22 <       thimm> | yum plugin sounds nicer
00:22 <         thl> | thimm, +1
00:22 <         thl> | that's my plan, too
00:22 <       nirik> | yuck. Yeah, yum plugin would be better...
00:22 <         thl> | as we would hae just one repo
00:22 <       thimm> | It will deal with any furture layering of RHEL products, too
00:22 <         thl> | and centos users simply could use all
00:22 <       nirik> | but who is going to write it? :)
00:22 <         thl> | nirik, that's the problem...
00:22 <       thimm> | Well, is it really a plugin?
00:23 <         che> | thl, well but the plugin cant just ignore missing deps
00:23 <       thimm> | Maybe it is yum core functionality
00:23 <         thl> | che, we'd need to ship list of packages
00:23 <       thimm> | Like a switch to deal with missing/broekn deps
00:23 <         thl> | where we know that deps will be missing
00:23 <         thl> | well, or something like that...
00:23 <       thimm> | This switch would be usefull for rawhide as well
00:23 <         thl> | I'll try to look into this during the week
00:23 <         thl> | and post to the list
00:24 <         thl> | but well, if somebody else is interested in solving this I won't mind :)
00:24 <         thl> | is anyone?
00:24 <       nirik> | didn't someone post a list of packages that were just in server/client? I can't find the list now...
00:24 <       thimm> | I did on epel and rhelv5 lists
00:24 <         thl> | nirik, is more complicated then just server vs client
00:25 <       thimm> | Just grep for evolution in the body
00:25 <         thl> | nirik, there are different client versions
00:25 <       nirik> | thl: in 4, but not 5, right?
00:25 <       thimm> | thl: sure?
00:25 <       nirik> | 5 just has a since client/workstation I thought.
00:25 <       thimm> | thl: difference in premium etc is support, not content
00:25 <         thl> | only RHEL5-Client-Workstation for example ships PHP and eclipse
00:26 <         thl> | thimm, not 100% sure, but quite sure, yes
00:26 <         thl> | you have a isntall-number
00:26 <       nirik> | ah, found the list... it's pretty small...
00:26 <         thl> | you have to fill in during install
00:26 <         thl> | and that decides about the package set you can install
00:26 <       thimm> | I see
00:27 <       quaid> | are the "not available" packages not oon RHN for those install types?
00:27 <       thimm> | so one media, several outputs
00:27 <         thl> | well, as I said, I'll try to look at this closer during the week
00:27 <         thl> | thimm, yes
00:27 <         thl> | quaid, are they?
00:27 <       quaid> | I don't know
00:27 <       thimm> | Makes organisation quite messy
00:27              * | thl has no RHEL5 at hand ATM
00:27 <       quaid> | this is a bit unexpected thinking to me ...
00:27 <       thimm> | quaid: It wouldn't make much sense, or not?
00:27 <       quaid> | well
00:28 <       quaid> | from a support perspective, sure, if you have Server and call about Eclipse on it ...
00:28 <       quaid> | but does that mean there is no way to get the package?
00:28 <       quaid> | I'm thinking about  dep resoling in particular
00:28 <         thl> | let's stop here
00:28 <         thl> | and further investigate during the week
00:28 <       quaid> | and not being to replace packages i nthe distro, etc.
00:28 <       quaid> | ok
00:28 <       thimm> | If the install number prohibit installtion of eclipse from the media they should also do so from the channels
00:28 <         thl> | does that sound sane?
00:28 <       nirik> | yeah, I think further investigation would be good.
00:29 <       nirik> | I would really prefer to just have one repo per release of epel... otherwise it's getting very complex for maintainers.
00:29 <         thl> | nirik, +1
00:29 <         thl> | and for us, too
00:29              * | thl will move on soon if nobody yells
00:29            --- | thl has changed the topic to: EPEL SIG meeting -- http://fedoraproject.org/wiki/EPEL/Schedule -- http://fedoraproject.org/wiki/EPEL/GuidelinesAndPolicies/PackageMaintenanceAndUpdates
00:30 <         thl> | do we all agree on that this is the way forward?
00:30              * | nirik re-reads it again quickly
00:30 <         thl> | not much changed in the past week
00:31 <         thl> | actually, the changes since the initial posting were quite small in general
00:31 <       thimm> | I disagree on kernel module policy
00:32 <         thl> | thimm, suggestions please :)
00:32 <         thl> | this was discussed on the list, wasn#t it?
00:32 <         thl> | did you speak up there? (/me wonders if he missed that)
00:32 <       thimm> | I did
00:32 <       thimm> | And you replied
00:32 <       thimm> | But I still disagree :)
00:32 <       nirik> | kernel modules are a nice can of worms. I think we want to at least avoid them until everything else is up...
00:32 <         thl> | nirik, +100 :)
00:33 <       thimm> | Also: why rebuild packages?
00:33 <         che> | from a usability point of view i like the dkms approach
00:33 <         thl> | thimm, ?
00:33 <       thimm> | From testing to stable transition
00:33 <       thimm> | Doesn't that defeat any testing/QA done?
00:33 <         thl> | thimm, that's not written there afaics
00:34 <         thl> | the testing branch simply becomes the stable branch when the quarterly update happens
00:34 <       thimm> | "go to a testing branch for a short time period and then build a second time for the stable branch"
00:34 <       nirik> | yeah, shouldn't need to do that... since ABI/API should keep stable in minor releases of RHEL, right?
00:34 <         thl> | thimm, that's something different
00:34 <         thl> | that's more a: "updates-testing" meachnism
00:34 <         thl> | and only for those rare cases where updates to stable are needed
00:34 <       thimm> | Yes, and this should be rebuilt as well
00:35 <       thimm> | Otherwise the "testing" part gets reset
00:35 <       nirik> | isn't that more "there were problems in testing, so a new release was needed to fix them before pushing to stable" ?
00:35 <       thimm> | That's different
00:35 <       thimm> | If the package has a bug it needs to get fixed in testng
00:35 <       nirik> | agreed.
00:35 <         thl> | thimm, can you please outline your thoughts on the list
00:35 <       thimm> | And when the bugs are off move the final testing package to stable
00:35 <         thl> | I think that would be the better place
00:36 <       thimm> | Well, these are no "thoughts", just general pratice
00:36 <         thl> | thimm, don#t forget that packages in testing might get build agianst something else then the ones in stable
00:36 <       nirik> | thimm: I agree, perhaps we need to clarify the document in that?
00:36 <         thl> | stable=rh-latest
00:36 <       thimm> | thl: that should not happen
00:37 <         thl> | testing=rh-what-becomes-next-quarterly-update
00:37 <         thl> | thimm, let's move this to the list
00:37 <         thl> | it becomes to complicated here on IRC in my opinion
00:37 <       thimm> | thl: There are two concepts mixes up then
00:37 <       nirik> | The only time I would think that would happen is if there were several packages in epel that depended on each other and needed to be pushed at once.
00:37 <       thimm> | You want "rawhide" and "updates-testing" in one repo
00:37 <       thimm> | That won't wokr
00:38 <       thimm> | For development we need to separate them, just like FC does
00:38 <         thl> | development is fedora
00:38 <       thimm> | security and major bugfxies go to the "updates-testing"
00:38 <         thl> | security and major bugfxies go to the "updates-testing" +1
00:38 <       thimm> | Any the prepeation for quarterly go to epel's "rawhide"
00:39 <         thl> | no
00:39              * | thl stopps here, this doens't lead to anything
00:39 <         thl> | lets move it to the list
00:39 <       thimm> | Why?
00:39 <       nirik> | security and major bugfixes go to "updates-testing" for a short time only and then get pushed to fix the security/major bug, right?
00:39 <       thimm> | Yes
00:39 <       thimm> | W/o rebuilding
00:39 <       thimm> | Otherwise the QA gets lost
00:39 <       nirik> | the minor updates/versions just keep sitting in updates-testing until the next minor
00:40 <         thl> | I'd say with rebuilding
00:40 <       thimm> | Yes, but that needs to be a separate repo
00:40 <       nirik> | thimm: +1 to no rebuild.
00:40 <         thl> | at leat if we build against rhel-what-becomes-update soon
00:40 <       thimm> | E.g. one for quarterly long-term updates and one for near-term updates to the current minor
00:40 <       nirik> | updates-testing and updates-nextrel ?
00:41            <-- | engwnbie  has left #fedora-meeting ( "I'm Done")
00:41 <         thl> | thimm, then we'd need to have 3 repos
00:41              * | thl thinks that's to much
00:41 <       thimm> | Whch three?
00:41 <         thl> | that will fail just as updates-testing fail
00:41              * | thl repeats: let's discuss it on the list
00:41 <         thl> | we don#t understnad each other here
00:41 <       thimm> | Let's talk in FC naming:
00:41              * | thl votes for moving on
00:42 <         thl> | and reisiting this next week and on the list
00:42 <         thl> | revisit
00:42 <         thl> | stupid keyboard
00:42 <       nirik> | so for each release, say RHEL5... there is a "5" repo with main packages, stable, pushed out. A 'updates-testing' thats usually empty, but sometimes has major security/bugfix packages, and a 'updates-nextrel' that has the packge updates for 5.1 waiting in it.
00:42 <       nirik> | thl: ok.
00:43 <       thimm> | nirik: something like that, yes
00:43 <       thimm> | Where the two latter repo are consdiered our development repos
00:43 <       thimm> | Not development as in == rawhide, but development as in where we do QA and shape packages for the next quarterly
00:43 <       thimm> | Anyway, let's move on
00:44            --- | thl has changed the topic to: EPEL SIG meeting -- http://fedoraproject.org/wiki/EPEL/Schedule -- what else?
00:44 <       nirik> | right. I think thats all doable, and mostly autopmagically so maintainers don't need to mess with it too much.
00:44 <         thl> | anything else for today?
00:44 <         thl> | repotag?
00:44 <       thimm> | Voting body & fesco mandate?
00:44 <         thl> | thimm, was discussed last week
00:44            --> | ChitleshGoorah (Chitlesh GOORAH)  has joined #fedora-meeting
00:44 <         thl> | but we can bring it up again
00:44 <       thimm> | I'm sure it wasn't voted on ;)
00:45 <       thimm> | repotag, fedr-ausrmgmt etc need a final voting somewhere
00:45 <         thl> | sure, because we'd need to ask FESCo first how to set up a voting body
00:45 <       nirik> | thimm: so what are you looking for? voting on tech issues where there is no consensus?
00:45 <       thimm> | It's either the sig or fesco
00:45 <         thl> | and the fest members around just said "try to continue without a oting body"
00:45            --> | lancelan (Lance Davis)  has joined #fedora-meeting
00:45 <       thimm> | thl: we need to suggest something and get iot ratified
00:45 <         thl> | thimm, seems a lot of people disagreed last week
00:46 <       thimm> | Last week?
00:46 <         thl> | last meeting
00:46 <       thimm> | On this irc?
00:46 <       nirik> | I think repotag should fall under packaging...
00:46 <         thl> | nirik, +1
00:46 <       thimm> | Where less than half of the sig can participate?
00:46 <         thl> | the summaries get send to the list
00:46 <         thl> | so people can look at it
00:46 <         thl> | and yell if they don#t like something
00:46 <         thl> | nobody yelled
00:47 <       thimm> | That's not to be deducted into everybody agrees
00:47 <       quaid> | fwiw, we aren't near ready to be a formal project
00:47 <         thl> | thimm, see https://www.redhat.com/archives/epel-devel-list/2007-March/msg00210.html
00:47 <       quaid> | so it is SIG or nothing , afaict
00:47 <       nirik> | I would much rather get to some tech consensus than brute force voting on issues where everyone gets unhappy in the end.
00:47 <       thimm> | Well, every sig does have to make decisions, right?
00:47 <         thl> | thimm, "some discussion about steering issues that got discussed on the list"
00:48 <         thl> | nirik, +1
00:48 <       quaid> | right, a SIG can vote, can't it?  I mean, anyone can hjold a vote, unless they are somewhere they will get shot for doing it :/
00:49 <         thl> | quaid, well, yes, but who is allowed to vote
00:49 <       thimm> | So the N people mentioned on the SIg list already form a voting body, is that what you're saying?
00:49 <         thl> | that can quickly become confusing if people just join the sig just to be able to vote
00:49 <         thl> | and how many votes does something need to get accepted
00:49 <       thimm> | 50%
00:49 <       thimm> | >50%
00:50 <         thl> | to much
00:50 <         thl> | that will deadlock soon
00:50 <       thimm> | That's how voting works
00:50 <       nirik> | so this is all for fedora-usermgmt banning? or ?
00:50 <       thimm> | No, its for everything
00:50 <         thl> | I'd say a group that decides about how epel moves forward should be 5 or 7 people
00:50 <         thl> | not more
00:50 <       thimm> | Then make a suggestion and have fesco ratify the list
00:50 <       nirik> | well, what I mean is what "everything" do we need voting for now? aside from usermgmt?
00:51 <       thimm> | repotag?
00:51 <       thimm> | repo layout?
00:51 <       thimm> | guidelines?
00:51 <       thimm> | policies?
00:51 <         thl> | thimm, we just do it the "if nobody yells it's okay" way currently
00:51 <         thl> | and that was what FESCo members in the last meeting prefered
00:51 <       thimm> | Which means 100% consensus or people are not paing attention
00:51 <         thl> | and non-FESCo members, too afaics
00:51 <         thl> | thimm, yes
00:51 <       thimm> | That's not a healthy way of a project
00:52 <         thl> | it worked so far
00:52 <         thl> | sure, if we grow
00:52 <       thimm> | It depends on the view I guess
00:52 <         thl> | then we need a committee
00:52 <       thimm> | so how do you deal with disagreement?
00:52 <         thl> | we try to avoid it for now
00:53 <       thimm> | Discuss until RHEL6?
00:53 <         thl> | or move it to FESCo
00:53 <       thimm> | Burry it under the carpet?
00:53 <       thimm> | Oh fesco will be grateful for esalating any decision to them
00:53 <       thimm> | For example repotag
00:53 <       thimm> | I would agree that pushing it to FPc is the right thing to do
00:54 <       nirik> | well, I would prefer to avoid the overhead, but if you want voting on issues, how about: anyone with a EPEL package can vote... voting to take place on a wiki page somewhere on each issue?
00:54 <       thimm> | At least not w/o a preference form EPEL
00:54 <         thl> | thimm, just out of interest, what's your opinion on repotag?
00:54 <         thl> | nirik, that has other disadantages
00:54 <         thl> | okay, I'll write something up for FESCo
00:54 <       thimm> | I don't really care, I would like to see one, but I wopuld rather see it decided upon either way before it gets an infininte  thread
00:54 <         thl> | I'll of course send it to the list beforehand
00:55 <       thimm> | thl: thanks! :)
00:55 <         thl> | anything else?
00:55 <       thimm> | Off the meeting: How do I import all my packages to EPEL?
00:55              * | thl will be afk soon anyway
00:56 <       thimm> | (All but a few non-EPELizable)
00:56 <       thimm> | Can someone here do a mass branching for me?
00:56 <         thl> | thimm, I asked last week for people to write a script to simplyfy mass-branching in cvs
00:56 <       thimm> | Who's doing the branches for EPEL, dgilmore?
00:56 <         thl> | thimm, I'll do this script properly myself later as no one showed interest
00:57 <       nirik> | might be best to generate a list and mail cvsadmins@fedoraproject.org?
00:57 <         thl> | thimm, cvsadmins, that includes dgilmore
00:57 <       thimm> | OK, I will try cvsadmins, thanks
00:57 <         thl> | anything else?
00:57              * | thl will close the meeting in 30
00:58 <       thimm> | bye all!
00:58              * | thl will close the meeting in 10
00:58 <         thl> | -- MARK -- Meeting end