ReleaseEngineering/Meetings/2007-aug-20

From FedoraProject

Jump to: navigation, search

Contents

Fedora Release Engineering Meeting :: Monday 2007-08-20

Mass Rebuild

  • f13
  • we need to rebuild for the build-id stuff, ppc32 binaries for not needing execmem, and that's about the extent of the "technical" reasons.
  • we could also pick up license clarifications to spec files at that time, but it's not technical necessary
  • concerns about PPC blocking test2 release
  • see IRC log for more details

Decision: NONE. f13 will send mail to list with additional information.

String Freeze Policy Draft

Decision:

  • The Fedora Localization Team, potentially together with the release team, will evaluate the request and provide feedback. If the string change is rejected, then you'll have to wait for the next development cycle to begin to get the change into Fedora packages. Disputes over rejected changes can be escalated to FESCo.
  • Adjust policy as needed after seeing it in action for F8.
  • Glezos to put the written policy at a permanent location on the wiki and write an announcement for fedora-devel-announce

Open Thoughts & Discussion

  • Consequences of slipping F8 schedule more than 1 week
  • PPC as primary architecture
  • see IRC log

Decision: NONE

IRC Transcript

Idea.png f13 changed the topic of #fedora-meeting to: Fedora Release Engineering Meeting - Roll Call<a href="#t13:01" class="time">13:01</a>
* poelcat here<a href="#t13:03" class="time">13:03</a>
f13jeremy: spot: jwb: rdieter: wwoods: warren: poelcat: notting: ping<a href="#t13:04" class="time">13:04</a>
rdieterhere<a href="#t13:04" class="time">13:04</a>
jwbhere (sorta)<a href="#t13:04" class="time">13:04</a>
* jeremy is here<a href="#t13:04" class="time">13:04</a>
jeremyf13: notting's on vacation still<a href="#t13:04" class="time">13:04</a>
warrensorry, here<a href="#t13:04" class="time">13:04</a>
f13hrm, I thought Bill was back from vaca today.  I see emails from him..  oh well<a href="#t13:04" class="time">13:04</a>
warrenWFH = work from hotel<a href="#t13:05" class="time">13:05</a>
f13ok, lets get started<a href="#t13:06" class="time">13:06</a>
Idea.png f13 changed the topic of #fedora-meeting to: Fedora Release Engineering Meeting - Mass Rebuild<a href="#t13:06" class="time">13:06</a>
f13So we've got a rebuild coming up.  I'd like to have it done after Test2 and before the freeze for Test3<a href="#t13:07" class="time">13:07</a>
f13we need to rebuild for the build-id stuff, ppc32 binaries for not needing execmem, and that's about the extent of the "technical" reasons.<a href="#t13:07" class="time">13:07</a>
poelcatwhy wouldn't we want to have it done before Test2 to get more testing time?<a href="#t13:08" class="time">13:08</a>
jeremyf13: I really think we need it before test2<a href="#t13:08" class="time">13:08</a>
f13We could also pick up license clarifications to spec files at that time, but it's not technical necessary<a href="#t13:08" class="time">13:08</a>
warrenRebuild absolutely everything, or only stuff not built since X date?<a href="#t13:08" class="time">13:08</a>
jeremyf13: otherwise, test2 is DOA for ppc<a href="#t13:08" class="time">13:08</a>
f13jeremy: test1 was DOA too, how many reports did we get?<a href="#t13:08" class="time">13:08</a>
f13poelcat: I wanted to give the build-id stuff some more time to settle so that we're building for that correctly instead of finding a problem half way througha nd restarting<a href="#t13:08" class="time">13:08</a>
jeremyf13: if ppc is an arch that we're shipping, then we need to bloody ship it.  not hide behind excuses like that<a href="#t13:08" class="time">13:08</a>
f13warren: personally I think just the stuff that needs it.  Let mdomsch's rebuild results point out other packages that need to rebuild on a case by case basis.<a href="#t13:09" class="time">13:09</a>
warrensounds good<a href="#t13:10" class="time">13:10</a>
f13so to do it before test2, would mean to coordinate and do it in the next... week?<a href="#t13:10" class="time">13:10</a>
jwbare we going to have maintainers do it, or rel-eng?<a href="#t13:10" class="time">13:10</a>
f13I'd like maintainers to do it, because I'm lazy (:<a href="#t13:11" class="time">13:11</a>
f13but if we want to get all binary packages built in a week, I don't think we can rely upon that<a href="#t13:11" class="time">13:11</a>
warrenIf we tell people to begin NOW, we can spread out the load.<a href="#t13:11" class="time">13:11</a>
mdomschbuild-id hit somewhere in the middle of my last rebuild run - need I redo it all from scratch?<a href="#t13:11" class="time">13:11</a>
f13mdomsch: that would be helpful, or at least redo anything that was done before build-id stuff landed<a href="#t13:11" class="time">13:11</a>
mdomschok<a href="#t13:12" class="time">13:12</a>
f13lets talk abit about how to find out what packages to build.<a href="#t13:12" class="time">13:12</a>
jeremyf13: my big thing is that it is absolutely insane to say that we're feature frozen _before_ a rebuild.  and also to be feature frozen with a completely busted arch.<a href="#t13:12" class="time">13:12</a>
f13it's my understanding that only binary packages, which produce debuginfo, need the build-id stuff?<a href="#t13:12" class="time">13:12</a>
f13jeremy: these arguments might have been helpful last week when we talked and you seemed in agreement that we needed to wait.  We could have been working on this all week instead of fiddling around with pungi :(<a href="#t13:13" class="time">13:13</a>
jeremyf13: I didn't realize how busted ppc was until the end of last week when I was doing the livecd stuff.  I thought it was sporadic apps with problems, not the whole world is complete shite.<a href="#t13:14" class="time">13:14</a>
f13AFAIK it's just selinux.  I thought I made that clear.<a href="#t13:15" class="time">13:15</a>
f13sorry if I didn't.<a href="#t13:15" class="time">13:15</a>
jeremyf13: selinux is part of what makes fedora fedora...  telling people to go off and do "selinux=0" is the wrong answer<a href="#t13:16" class="time">13:16</a>
f13so is there anything wrong with my assumption of what needs to be rebuilt for build-id stuff?<a href="#t13:16" class="time">13:16</a>
jeremyf13: afaik, that's correct<a href="#t13:16" class="time">13:16</a>
warrenCould we tell maintainers, if your package was built after X date, please rebuild it now?<a href="#t13:17" class="time">13:17</a>
warrenAnd we figure out at the end of this week which need rebuilding by rel-eng.<a href="#t13:17" class="time">13:17</a>
warrenThat way we spread out the rebuilding load<a href="#t13:17" class="time">13:17</a>
warrenerr.... before X date<a href="#t13:17" class="time">13:17</a>
f13jeremy: yes, I understand that, but it /is/ a workaround.  It's not like the platform is completely unusable.  But this isn't helpful, lets stop arguing over our 20 users.<a href="#t13:18" class="time">13:18</a>
* warren wonders why we kept it as a first tier arch.<a href="#t13:18" class="time">13:18</a>
jeremyf13: if it's twenty users and we don't care, then let's stop pretending and wasting our time on it.  THAT is the crux of my argument<a href="#t13:18" class="time">13:18</a>
f13warren: I'd like to spend today figuring out what needs to be built.  From the ppc side (a window of builds that need to be corrected) and from the build-id side (a list of binary packages that haven't been built after X)<a href="#t13:18" class="time">13:18</a>
jwbhey<a href="#t13:18" class="time">13:18</a>
f13jeremy: why yell at me about it?  I'll happily dro pthe fucking arch yesterday.<a href="#t13:18" class="time">13:18</a>
jeremyeither it's an arch in which case, selinux working damned well better be part of what has to be the case for shipping a test release.  or, it's *NOT* in which case why do we keep pretending it is?<a href="#t13:19" class="time">13:19</a>
EvilBobwarren: I have wondered that for quite some time<a href="#t13:19" class="time">13:19</a>
jwbwait<a href="#t13:19" class="time">13:19</a>
jwbi have something to say about ppc<a href="#t13:19" class="time">13:19</a>
jwbi agree with jeremy<a href="#t13:19" class="time">13:19</a>
jeremyf13: I know.  and the decision from the board and fesco is that it's still a primary arch.  so *WE* have to follow that, we can't just go off and pretend it's not except when it suits us<a href="#t13:19" class="time">13:19</a>
f13I agree, I just have a hard time getting motivated to spend a ton of effort that doesn't really matter much.<a href="#t13:20" class="time">13:20</a>
jeremyf13: it's not up to us to make the rules, though<a href="#t13:20" class="time">13:20</a>
EvilBobdoesn't matter much to most users<a href="#t13:20" class="time">13:20</a>
jwbthat's not quite what i was agreeing to jeremy<a href="#t13:20" class="time">13:20</a>
f13but that's neither here nor there.  I'm willing to do the work, so lets move on.<a href="#t13:20" class="time">13:20</a>
jwbbut yes, you're correct<a href="#t13:20" class="time">13:20</a>
* wwoods will be back shortly<a href="#t13:21" class="time">13:21</a>
f13so if I can get a list of what needs to be built later today, how do we want to handle this?  DO we ask maintainers to build them, give them until a date, and if it hasn't been completed by $date, we step in and do it ourselves?<a href="#t13:22" class="time">13:22</a>
f13do we just do it ourselves in the interest of having it done in a week?<a href="#t13:22" class="time">13:22</a>
* nirik wonders if the window of broken ppc builds is known? so could just rebuild those before test2? <a href="#t13:22" class="time">13:22</a>
jeremyf13: to some extent, I think it depends on how big the list is<a href="#t13:23" class="time">13:23</a>
jwbare jakub and roland ready for the rebuild?<a href="#t13:23" class="time">13:23</a>
warrenf13, how long will it take to generate a list?<a href="#t13:23" class="time">13:23</a>
f13you know, I'm not even actually sure we've fixed the ppc issue in the buildsystem yet.  I never got conformation from the bug about this.<a href="#t13:23" class="time">13:23</a>
f13warren: a good chunk of the afternoon probably<a href="#t13:23" class="time">13:23</a>
jeremyf13: that would be important too :)<a href="#t13:23" class="time">13:23</a>
f13warren: to do it in a way that can be repeated for continually updated lists of needs doing.<a href="#t13:23" class="time">13:23</a>
* f13 does some digging<a href="#t13:24" class="time">13:24</a>
warrenf13, do we know the "built after" date for both issues?<a href="#t13:32" class="time">13:32</a>
f13warren: not specifically, still digging.<a href="#t13:35" class="time">13:35</a>
jeremyso, what else do we have<a href="#t13:38" class="time">13:38</a>
f13yea, lets move on while I try to gather that info, we'll follow up on list when I have numbers.<a href="#t13:38" class="time">13:38</a>
Idea.png f13 changed the topic of #fedora-meeting to: Fedora Release Engineering Meeting - String Freeze Policy Draft<a href="#t13:39" class="time">13:39</a>
f13<a href="http://fedoraproject.org/wiki/ReleaseEngineering/StringFreezePolicy/Draft">http://fedoraproject.org/wiki/ReleaseEngineering/StringFreezePolicy/Draft</a>  glezos would like us to read/ratify this and add it to our list of Policies.<a href="#t13:39" class="time">13:39</a>
* f13 spots another typo and fixes.<a href="#t13:41" class="time">13:41</a>
warren    *<a href="#t13:42" class="time">13:42</a>
warrenThe Fedora Localization Team, potentially together with the release team, will evaluate the request and provide feedback. If the string change is rejected, then you'll have to wait for the next development cycle to begin to get the change into Fedora packages. Disputes over rejected changes can be escalated to FESCo.<a href="#t13:42" class="time">13:42</a>
jeremyseems fine<a href="#t13:42" class="time">13:42</a>
jeremya slight optimization might be trying to get someone from the l10n team on the rel-eng team and then we could make it a little easier for people<a href="#t13:42" class="time">13:42</a>
warrenI know that glezos is very active, but if approval is possibly blocking a key important fix that could block release, shouldn't there be someone a little more guaranteed to be accountable there?<a href="#t13:43" class="time">13:43</a>
warren(i.e. a l10n employee)<a href="#t13:44" class="time">13:44</a>
f13are there any l10n employees on the FEdora Local Team?<a href="#t13:44" class="time">13:44</a>
f13I would hope so<a href="#t13:45" class="time">13:45</a>
warrenglezos, are any of the RH l10n employees in the fedora l10 team?<a href="#t13:46" class="time">13:46</a>
glezosSorry I'm late for the meeting. Hope I wasn't needed for the String freeze policy<a href="#t13:47" class="time">13:47</a>
glezoswarren, yeah, a bunch of them AFAIK<a href="#t13:47" class="time">13:47</a>
f13glezos: we're discussing that right now<a href="#t13:47" class="time">13:47</a>
f13given that, I'm +1 for adopting this policy<a href="#t13:47" class="time">13:47</a>
glezosI did send a link to the policy a week ago, no complaints raised on -trans-list.<a href="#t13:48" class="time">13:48</a>
* runa_b is part of the RH l10n team<a href="#t13:48" class="time">13:48</a>
jeremyanyone against?  if not, let's keep this moving.<a href="#t13:49" class="time">13:49</a>
warren+1<a href="#t13:49" class="time">13:49</a>
glezosf13, my only concern about the policy is who decides about breaking<a href="#t13:49" class="time">13:49</a>
warrenDoes it require one member of rel-eng and one member of l10n?<a href="#t13:50" class="time">13:50</a>
glezosthe FLP doesn't have a FLoSCo or something<a href="#t13:50" class="time">13:50</a>
f13warren: I'm happy with just two l10ns<a href="#t13:50" class="time">13:50</a>
warrenAnd how quickly do the package maintainers pick up l10n changes?<a href="#t13:51" class="time">13:51</a>
f13I don't necessarily want two rel-eng's and nobody from l10n stepping up to take it.<a href="#t13:51" class="time">13:51</a>
f13but I'm happy with this as a first try at written string freeze policy and adjusting to suit as needed after we see it in action<a href="#t13:51" class="time">13:51</a>
warrenlet;s do it<a href="#t13:51" class="time">13:51</a>
Idea.png f13 changed the topic of #fedora-meeting to: Fedora Release Engineering Meeting - F8 Test2 Freeze<a href="#t13:51" class="time">13:51</a>
* glezos is happy about 2 l10ns as well -- don't want to put more stuff on releng than necessary<a href="#t13:51" class="time">13:51</a>
glezosthanks guys.<a href="#t13:52" class="time">13:52</a>
f13Freeze is Tuesday, a week and a day away<a href="#t13:52" class="time">13:52</a>
f13we need to start jumping on the broken deps and broken upgrade paths.  BOFH mode engaged.<a href="#t13:52" class="time">13:52</a>
f13also I need to start doing full media spins to make sure all that code is still working,a nd since new code is landing there it's going to be touch and go<a href="#t13:53" class="time">13:53</a>
warrenglezos, put the written policy at a permanent location on the wiki and write an announcement for fedora-devel-announce<a href="#t13:53" class="time">13:53</a>
f13I'd like to get a LiveCD and a full iso set up on the torrents right around freeze time too<a href="#t13:53" class="time">13:53</a>
f13warren: I'll move it to where it needs to go.<a href="#t13:53" class="time">13:53</a>
glezoswarren, ACK<a href="#t13:53" class="time">13:53</a>
f13(by renaming the page)<a href="#t13:53" class="time">13:53</a>
warrenk<a href="#t13:53" class="time">13:53</a>
poelcatf13: re: freeze... I just sent mail to fedora-devel about <a href="http://fedoraproject.org/wiki/Releases/8/FeatureList">http://fedoraproject.org/wiki/Releases/8/FeatureList</a><a href="#t13:54" class="time">13:54</a>
f13poelcat: noticed that, thanks<a href="#t13:54" class="time">13:54</a>
* poelcat realizes we are mostly a time based release, but wonders if any of those could change freeze date<a href="#t13:55" class="time">13:55</a>
rdietermaybe a tiny bit, if lives of any cute kittens are at stake. :)<a href="#t13:55" class="time">13:55</a>
f13mostly it would be for misbehaviing features that are already "in"<a href="#t13:56" class="time">13:56</a>
f13like buildID maybe<a href="#t13:56" class="time">13:56</a>
f13I don't know if we'd wait for IceTea<a href="#t13:56" class="time">13:56</a>
f13but that's not really a discussion for this meeting<a href="#t13:56" class="time">13:56</a>
wwoodsIf we push more than a week or two, we start hitting vacations and outages<a href="#t13:56" class="time">13:56</a>
f13any other thoughts on the freeze?<a href="#t13:56" class="time">13:56</a>
f13glezos: <a href="http://fedoraproject.org/wiki/ReleaseEngineering/StringFreezePolicy">http://fedoraproject.org/wiki/ReleaseEngineering/StringFreezePolicy</a> is your new page home<a href="#t13:57" class="time">13:57</a>
glezosf13, just noticed, thanks. I'll write a quick announce<a href="#t13:57" class="time">13:57</a>
f13Ok, that's all I have for this week, any open discussion items?<a href="#t13:58" class="time">13:58</a>
Idea.png f13 changed the topic of #fedora-meeting to: Fedora Release Engineering - Open<a href="#t13:58" class="time">13:58</a>
warrenDid you find out "built after" dates?<a href="#t13:58" class="time">13:58</a>
f13warren: no, I'm still looking, can't find the goddamn binutils bug<a href="#t13:59" class="time">13:59</a>
warrenKnowing the built after dates would be useful to get some maintainers started on rebuilding.<a href="#t13:59" class="time">13:59</a>
wwoods"*if* F8 slips more than 1 week, then we'll  have to delay release until the week of December 10th<a href="#t13:59" class="time">13:59</a>
warrenthey don't need our list of pakcages to rebuild<a href="#t13:59" class="time">13:59</a>
jeremywwoods: maybe, maybe not<a href="#t14:00" class="time">14:00</a>
wwoodsSo yeah - if we slip for more than 1 week, we end up slipping for 6 weeks. Or, at least, there'll be some wrasslin' to get it done.<a href="#t14:00" class="time">14:00</a>
jeremywwoods: like I said, maybe, maybe not.  I have a discussion pending on it<a href="#t14:00" class="time">14:00</a>
warrenf13, given the schedule pressure, we need to decide upon a rebuild plan now?<a href="#t14:00" class="time">14:00</a>
EvilBobI know this is not the right place to talk about it but what exactly does fedora benefit from having PPC as a primary arch?<a href="#t14:00" class="time">14:00</a>
f13warren: please give me some time to find the package list?   Thanks.<a href="#t14:01" class="time">14:01</a>
skvidalEvilBob: it doesn't<a href="#t14:01" class="time">14:01</a>
warrenf13, something like 1) Ask maintainers to rebuild packages if they haven't been built since X date.  2) Figure out a way to generate a needs-rebuild list quickly and figure out what's left 3) Poke people 4) Do it ourselves beginning Thursday?<a href="#t14:01" class="time">14:01</a>
wwoodsjeremy: right, that's good. I'm just trying to make the point that, if we have to decide between deferring features for F9 or slipping the release, we should strongly consider the former<a href="#t14:01" class="time">14:01</a>
warrenf13, we don't need the list is what I'm saying, the date is more important to get people started while we figure it out.<a href="#t14:01" class="time">14:01</a>
f13warren: can't do anything until we have a list, can't get the list until we have dates, can't get dates until I find the information about the binutils thing.  So please, let me get this done.<a href="#t14:02" class="time">14:02</a>
EvilBobskvidal: so is it fair to say that having it as a primary arch creates more work for many to benefit a niche group of users?<a href="#t14:03" class="time">14:03</a>
warrenI mostly agree, but I don't think we really need a list to get people started.<a href="#t14:03" class="time">14:03</a>
warrenI'm done.<a href="#t14:03" class="time">14:03</a>
f13warren: fine, we don't need a list, we need the dates to build before/after.  DOn't ahve that yet either, precurser to finding the list.<a href="#t14:03" class="time">14:03</a>
nirikf13: is it: <a href="https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=184446">https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=184446</a> ?<a href="#t14:04" class="time">14:04</a>
warrenf13, ok.<a href="#t14:04" class="time">14:04</a>
f13EvilBob: some will say that we benifit by building on nonx86, a big endian platform that does catch some bugs.<a href="#t14:04" class="time">14:04</a>
warrenf13, do we agree on Thursday or Friday as the time to begin building things ourselves?<a href="#t14:04" class="time">14:04</a>
f13or catches some bugs quicker than if we didn't build there.<a href="#t14:04" class="time">14:04</a>
f13(I don't buy this argument myself)<a href="#t14:04" class="time">14:04</a>
f13nirik: that's from '06<a href="#t14:04" class="time">14:04</a>
f13nirik: I'm starting to suspect that no bug was ever filed on this.<a href="#t14:04" class="time">14:04</a>
f13(hurray for ppc users!)<a href="#t14:05" class="time">14:05</a>
nirikyeah, so it is. :( Sorry.<a href="#t14:05" class="time">14:05</a>
jwbf13, rawhide ppc32 users are rare<a href="#t14:05" class="time">14:05</a>
jwbeven among ppc users<a href="#t14:05" class="time">14:05</a>
* nirik is one, but thought it was already filed. ;( <a href="#t14:06" class="time">14:06</a>
f13jwb: it's not just ppc32, most of ppc64 is made up of 32bit packages<a href="#t14:06" class="time">14:06</a>
wwoodsyeah, as we've discussed before, pretty much every ppc32 rawhide user is in #fedora-devel<a href="#t14:06" class="time">14:06</a>
f13jwb: it's any 32bit binary<a href="#t14:06" class="time">14:06</a>
jwbsorry, s/ppc32/ppc<a href="#t14:06" class="time">14:06</a>
Idea.png f13 changed the topic of #fedora-meeting to: Channel is used by various Fedora groups and committees for their regular meetings | Note that meetings often get logged | For questions about using Fedora please ask in #fedora | See <a href="http://fedoraproject.org/wiki/Communicate/FedoraMeetingChannel">http://fedoraproject.org/wiki/Communicate/FedoraMeetingChannel</a> for meeting schedule<a href="#t14:06" class="time">14:06</a>

Generated by irclog2html.py 2.3 by Marius Gedminas - find it at mg.pov.lt!