ReleaseEngineering/Meetings/2007-aug-13

From FedoraProject

Jump to: navigation, search

Contents

Fedora Release Engineering Meeting :: Monday 2007-08-13

Release Naming

  • see IRC log for specific discussion
  • Decision: Release Engineering is going to get an early start on F8 naming

1. query maintainers-list for suggestions 1. filter them into a list for legal 1. take results from legal shove it into a voting application

Feature Freeze and Package QA

  • see IRC log for discussion
  • Decisions: NONE

IRC Transcript

Idea.png f13 changed the topic of #fedora-meeting to: Fedora Release Engineering<a href="#t13:02" class="time">13:02</a>
* warren here<a href="#t13:02" class="time">13:02</a>
* poelcat here<a href="#t13:03" class="time">13:03</a>
* bpepple|lt lurking about.<a href="#t13:04" class="time">13:04</a>
* wwoods here... ish<a href="#t13:04" class="time">13:04</a>
f13well, we don't have anything on the agenda for today<a href="#t13:04" class="time">13:04</a>
jwbadjourn<a href="#t13:04" class="time">13:04</a>
warrenso done yet?<a href="#t13:04" class="time">13:04</a>
jwb')<a href="#t13:04" class="time">13:04</a>
f13but it was requested that we start on the naming process now<a href="#t13:04" class="time">13:04</a>
jwber, ;)<a href="#t13:04" class="time">13:04</a>
spoti like "Aurora"<a href="#t13:04" class="time">13:04</a>
f13get it knocked out and finished so it can't catch us later.<a href="#t13:04" class="time">13:04</a>
jwbspot, connection with moonshine?<a href="#t13:04" class="time">13:04</a>
f13I drank a lot of moonshine to produce Aurora?<a href="#t13:05" class="time">13:05</a>
jwbheh, just thought of that<a href="#t13:05" class="time">13:05</a>
jwbi like it<a href="#t13:05" class="time">13:05</a>
wwoodscelestial light?<a href="#t13:05" class="time">13:05</a>
warrenf13, can a name be two words?<a href="#t13:05" class="time">13:05</a>
jwbspot, would be even better if Sparc was there for f-8.  how close?<a href="#t13:05" class="time">13:05</a>
spotjwb: close.<a href="#t13:05" class="time">13:05</a>
spotaurora and moonshine are both waterfalls<a href="#t13:06" class="time">13:06</a>
f13warren: I don't see an immediate reason why not, but has there been prior art there?<a href="#t13:06" class="time">13:06</a>
jwbspot, any idea on the buildsys side of things?<a href="#t13:06" class="time">13:06</a>
jeremywarren: spaces could break things<a href="#t13:06" class="time">13:06</a>
spotalmost ready. literally, waiting on me to test.<a href="#t13:06" class="time">13:06</a>
jeremybut let's just get names starting to be submitted so we can run a pile past legal soon<a href="#t13:06" class="time">13:06</a>
jeremy(brb)<a href="#t13:06" class="time">13:06</a>
wwoodswhere should names be submitted?<a href="#t13:06" class="time">13:06</a>
spoti need to get python & db4 up to rawhide levels to test the builder<a href="#t13:06" class="time">13:06</a>
f13previous releases we just pinged maintainers- for name suggestions with link explenation<a href="#t13:07" class="time">13:07</a>
f13and I went through them after a week or so and gathered up the sane ones that had proper links and submitted them as a batch to Legal to pre-approve.<a href="#t13:07" class="time">13:07</a>
warrenNames are no longer a surprise<a href="#t13:07" class="time">13:07</a>
jwbnaming is boring<a href="#t13:07" class="time">13:07</a>
* warren doesn't care about names anymore.<a href="#t13:07" class="time">13:07</a>
warrenIt became boring after we finally got to use Zod.<a href="#t13:08" class="time">13:08</a>
f13We can still keep the result of the voting a secret until late.<a href="#t13:08" class="time">13:08</a>
warrenNo more reason for living.<a href="#t13:08" class="time">13:08</a>
* spot shoots warren<a href="#t13:08" class="time">13:08</a>
f13all we need to do is get a name legally approved<a href="#t13:08" class="time">13:08</a>
f13we can have abadger1999 do up a voting app for us again<a href="#t13:09" class="time">13:09</a>
warrenf13, I guess do that on the list again?<a href="#t13:09" class="time">13:09</a>
* warren wants "Charon" as the next name.<a href="#t13:09" class="time">13:09</a>
jwbshizzle<a href="#t13:09" class="time">13:09</a>
* warren doesn't really care though.<a href="#t13:09" class="time">13:09</a>
f13warren: query the list for suggestions, filter them into a list for legal, take results from legal shove it into a voting app<a href="#t13:09" class="time">13:09</a>
f13(spiderpig shall win though.  I decree it)<a href="#t13:09" class="time">13:09</a>
jwbshizzlepig<a href="#t13:10" class="time">13:10</a>
wwoodsbetcha Rupert Murdoch owns that trademark<a href="#t13:10" class="time">13:10</a>
f13I'll get started on that this week.<a href="#t13:10" class="time">13:10</a>
jwbf13, i can send the email and filter if you want<a href="#t13:11" class="time">13:11</a>
f13poelcat: Decision: Release Engineering is goign to get an early start on F8 naming:  query maintainers-list for suggestions, filter them into a list for legal, take results from legal shove it into a voting app.<a href="#t13:11" class="time">13:11</a>
f13jwb: that would rock!<a href="#t13:11" class="time">13:11</a>
jwbf13, it's one of the few things someone not @RH can do, so no problem<a href="#t13:11" class="time">13:11</a>
jeremyf13: have we made any progress on getting a machine in the colo that we can do the actual composes on?<a href="#t13:11" class="time">13:11</a>
f13jeremy: that may be stalled by getting ram upgrades in the colo<a href="#t13:12" class="time">13:12</a>
f13jeremy: which a budget for that is being worked out by max and co today IIRC<a href="#t13:12" class="time">13:12</a>
jeremyf13: okay, so tied up with that<a href="#t13:12" class="time">13:12</a>
wwoodswe've got.. two weeks 'til feature freeze at this point?<a href="#t13:13" class="time">13:13</a>
f13yea, not a long time<a href="#t13:13" class="time">13:13</a>
jeremysounds right<a href="#t13:13" class="time">13:13</a>
wwoodsdo we need to sound some alarms about that?<a href="#t13:13" class="time">13:13</a>
warrenyes<a href="#t13:14" class="time">13:14</a>
warrenfedora-devel-announce and internally<a href="#t13:14" class="time">13:14</a>
jwbfedora-devel and fedora-devel-announce<a href="#t13:14" class="time">13:14</a>
f13jwb: fedora-devel gets all fedora-devel-announce mails<a href="#t13:14" class="time">13:14</a>
jwboh good<a href="#t13:15" class="time">13:15</a>
wwoods"remember kids: no need to force it. F9 feature freeze is only 6 months away!"<a href="#t13:15" class="time">13:15</a>
jeremywwoods: that's an important one to emphasize<a href="#t13:15" class="time">13:15</a>
wwoodsThe F9 one? Because that's what RHEL6 will be based from?<a href="#t13:16" class="time">13:16</a>
warrenAny volunteers to do the mail?  Otherwise I'll do it.<a href="#t13:16" class="time">13:16</a>
warrenwwoods, possibly, uncertain<a href="#t13:16" class="time">13:16</a>
jeremywwoods: no, the fact that it's 6 months out<a href="#t13:17" class="time">13:17</a>
f13warren: go for it!<a href="#t13:17" class="time">13:17</a>
wwoodsjeremy: ah, yes<a href="#t13:17" class="time">13:17</a>
f13oh a different subject, do we want to put up Fedora spin isos when we put up rawhide live images (on the torrent)?<a href="#t13:18" class="time">13:18</a>
wwoodsrawhide installer isos?<a href="#t13:19" class="time">13:19</a>
f13warren: yeah.<a href="#t13:20" class="time">13:20</a>
warrenclarification...<a href="#t13:20" class="time">13:20</a>
warrenfeature freeze is for the stuff that makes it onto spins?<a href="#t13:20" class="time">13:20</a>
f13more than jsut boot.iso, but a self contained iso that would test media installation targets<a href="#t13:20" class="time">13:20</a>
warrenWe do continue to add more stuff to the devel collection until some point later<a href="#t13:20" class="time">13:20</a>
f13we typically only find media install problems at the test release points, and maybe we could find issues earlier if there were more people consuming media install images.<a href="#t13:20" class="time">13:20</a>
f13warren: feature freeze is across the board, as "stuff in the spins" is really really vauge when you can add the everything repo at install time.<a href="#t13:21" class="time">13:21</a>
wwoodsif we've got disk space and bandwidth I'm all for it<a href="#t13:21" class="time">13:21</a>
f13s/vague/ambiguous/<a href="#t13:21" class="time">13:21</a>
warrenf13, so we stop additions of new packages to devel in 2 weeks as well?<a href="#t13:21" class="time">13:21</a>
warrenWe have to more clearly define this.<a href="#t13:21" class="time">13:21</a>
f13wwoods: diskspace is probably ok on the torrent box, and bandwidth is torrent.<a href="#t13:21" class="time">13:21</a>
warrenmore than 2 months of no additions to devel is WAY too long.<a href="#t13:21" class="time">13:21</a>
wwoodsI guess this means we also have two weeks left to define the Default Spin<a href="#t13:22" class="time">13:22</a>
f13warren: we've rarely stopped adding new packages, except for near final deep freeze.  We just don't want feature changes in the packages we already have<a href="#t13:22" class="time">13:22</a>
wwoodswe've never closed devel for new packages<a href="#t13:22" class="time">13:22</a>
wwoodsexcept.. right<a href="#t13:22" class="time">13:22</a>
warrenWe have to explicitly define that the feature freeze is for stuff on the standard spins, and go into detail of what those spins are.<a href="#t13:22" class="time">13:22</a>
f13and adding new packages which would cause the need to rebuild a bunch of other stuff is somewhat frowned upon.<a href="#t13:22" class="time">13:22</a>
f13warren: uh no.<a href="#t13:22" class="time">13:22</a>
warrenIf we don't, this is too vague.<a href="#t13:22" class="time">13:22</a>
f13warren: again, 'on the spin' is meaningless.<a href="#t13:22" class="time">13:22</a>
wwoodsso here's a question. what defines the Standard Spin? comps + some package list in pungi?<a href="#t13:22" class="time">13:22</a>
f13wwoods: + whatever repo you add at install time, and we're talking about making it one-click access to the Everything tree<a href="#t13:23" class="time">13:23</a>
f13so really, Everything.<a href="#t13:23" class="time">13:23</a>
wwoodsright, but there is a de facto core vs. non-core split<a href="#t13:23" class="time">13:23</a>
warrenOK, so rather, feature freeze means = No major changes to things already in devel, but you can add new things as long as it doesn't require major changes to existing packages.<a href="#t13:23" class="time">13:23</a>
warren?<a href="#t13:23" class="time">13:23</a>
f13warren: that sounds better<a href="#t13:23" class="time">13:23</a>
warrenthis definition of feature freeze is a bit unintuitive.<a href="#t13:23" class="time">13:23</a>
wwoodse.g. the stuff that's in a default install is better tested and needs to be more stable than the whole package universe<a href="#t13:23" class="time">13:23</a>
f13warren: *shrug* if you want clarity we can just block all new packages period.<a href="#t13:24" class="time">13:24</a>
warrenThat isn't an option.<a href="#t13:24" class="time">13:24</a>
f13warren: it is if we change how we branch<a href="#t13:24" class="time">13:24</a>
warrenPerhaps I should define the goal of the feature freeze<a href="#t13:24" class="time">13:24</a>
f13if we branch F-8 at feature freeze time and allow for rawhide to continue on...<a href="#t13:24" class="time">13:24</a>
warrenGoals...<a href="#t13:24" class="time">13:24</a>
f13but then we really shoot f-8 in the foot<a href="#t13:24" class="time">13:24</a>
warren1) Stuff in default install especially needs to be better tested.<a href="#t13:25" class="time">13:25</a>
f13the problem is that it seems many many contributors just don't care about having releases, they just want a constantly moving package repo.<a href="#t13:25" class="time">13:25</a>
warren2) The collection must be in a known and tested state.  Big changes after the feature freeze are discouraged.<a href="#t13:25" class="time">13:25</a>
f13wwoods: what's on the "Fedora" spin (which is the only choose your own adventure spin we have right now) is a pungi manifest that lists some comps groups and some extra packages.<a href="#t13:25" class="time">13:25</a>
jwbf13, poelcat: FYI naming email sent<a href="#t13:25" class="time">13:25</a>
warren3) New packages are allowed as long as they don't require major changes to existing packages.<a href="#t13:25" class="time">13:25</a>
wwoodsWe have a bunch of package subsets: @core, @base, default install, default spin contents, Everything<a href="#t13:26" class="time">13:26</a>
wwoodswhere, IIRC, each set is a subset of the subsequent one<a href="#t13:26" class="time">13:26</a>
warren4) Contact FESCo if you need clarification or permission for an exception to the Feature Freeze.<a href="#t13:26" class="time">13:26</a>
wwoodsso. that list is ordered by size/priority. and we need a clear cutoff point for "no significant updates to packages inside this set"<a href="#t13:26" class="time">13:26</a>
f13wwoods: and unfortunately those sets are somewhat fluid as depsolving is involved<a href="#t13:27" class="time">13:27</a>
f13wwoods: and what else you select at installtime as effect upon /how/ those deps are resolved out.<a href="#t13:28" class="time">13:28</a>
jwbwarren, that all looks fine<a href="#t13:28" class="time">13:28</a>
f13(who know, doing anything less than 'everything in the collection' for install spins introduces a lot of complexity...)<a href="#t13:28" class="time">13:28</a>
f13s/know/knew/<a href="#t13:29" class="time">13:29</a>
wwoodswell yeah, but at any given moment you should be able to calculate the precise package list for any of those sets<a href="#t13:29" class="time">13:29</a>
wwoodsobviously they change over time, but they're deterministic<a href="#t13:29" class="time">13:29</a>
f13wwoods: "precise" ?<a href="#t13:29" class="time">13:29</a>
warrenFAQ: When will we stop adding new packages to F8?<a href="#t13:30" class="time">13:30</a>
warrenDeep freeze, which is typically ________<a href="#t13:30" class="time">13:30</a>
warrenOctober 23rd?<a href="#t13:30" class="time">13:30</a>
f13wwoods: you can get close, but again since we have various sets of packages which all provide the same thing, you can wind up with different stuff for your selection of one of these groups<a href="#t13:30" class="time">13:30</a>
jwbwarren, don't we already have wiki pages explaining all this stuff?<a href="#t13:31" class="time">13:31</a>
jwbi'm pretty sure we do<a href="#t13:31" class="time">13:31</a>
warrenI don't know<a href="#t13:31" class="time">13:31</a>
f13we have the Overview page<a href="#t13:31" class="time">13:31</a>
f13but I don't think it gets this much into detail.<a href="#t13:31" class="time">13:31</a>
f13but that's where you'd change it.<a href="#t13:31" class="time">13:31</a>
wwoodshow can set of packages that are in @base change if you're adding other groups?<a href="#t13:31" class="time">13:31</a>
f13<a href="http://fedoraproject.org/wiki/ReleaseEngineering/Overview">http://fedoraproject.org/wiki/ReleaseEngineering/Overview</a><a href="#t13:32" class="time">13:32</a>
f13wwoods: it's not the packages in @base, but what the packages in @base will pull in for their Requires.<a href="#t13:32" class="time">13:32</a>
f13wwoods: and the Requires of /those/ Requires<a href="#t13:32" class="time">13:32</a>
f13/usr/bin/mail is a good one to think about<a href="#t13:32" class="time">13:32</a>
wwoodsah. could be sendmail, could be postfix, &c<a href="#t13:33" class="time">13:33</a>
f13somehow in your install set, picking postfix instead of sendmail would change what you get a bit.<a href="#t13:33" class="time">13:33</a>
wwoodsright, but since sendmail is the default I think we'd define the @base group (for purposes of releng/testing) to include sendmail instead of postfix<a href="#t13:33" class="time">13:33</a>
f13there are a few other cases, but just something to keep in mind.<a href="#t13:34" class="time">13:34</a>
wwoodswe need deterministic package lists for a few tiers of packages to set things like testing priority / freeze strictness<a href="#t13:34" class="time">13:34</a>
wwoodsthey don't need to be static lists of package names<a href="#t13:35" class="time">13:35</a>
f13wwoods: sounds like a job for the QA team!  (:<a href="#t13:35" class="time">13:35</a>
wwoodsQA and releng<a href="#t13:35" class="time">13:35</a>
wwoodsand so here we are<a href="#t13:35" class="time">13:35</a>
f13so how can we help you?<a href="#t13:36" class="time">13:36</a>
wwoodsI guess I need help coming up with good definitions (i.e. something with a deterministic algorithm) for package tiers<a href="#t13:36" class="time">13:36</a>
wwoodsthe first thing I should do is write up a rough draft of the tiers<a href="#t13:37" class="time">13:37</a>
jwbtier 1) kernel<a href="#t13:37" class="time">13:37</a>
wwoodsand we'll bounce the algorithms back and forth until we have something that takes, basically, repodata/comps/other assorted datafiles as input<a href="#t13:37" class="time">13:37</a>
jwbtier 2) glibc<a href="#t13:37" class="time">13:37</a>
jwbtier 3) everything else<a href="#t13:37" class="time">13:37</a>
wwoodsand emits package lists<a href="#t13:37" class="time">13:37</a>
wwoodsso we can definitely say "don't touch package X at all because it is hell of deep frozen.. you can touch packages outside this set"<a href="#t13:38" class="time">13:38</a>
wwoodsetc<a href="#t13:38" class="time">13:38</a>
wwoodsso, okay, what are the "assorted datafiles"? pungi config files, comps, ..?<a href="#t13:38" class="time">13:38</a>
f13pungi manifest, comps, livecd kickstart file<a href="#t13:39" class="time">13:39</a>
f13and yum repodata<a href="#t13:39" class="time">13:39</a>
f13'yum deplist' is your friend<a href="#t13:39" class="time">13:39</a>
wwoodsokay, those are the inputs<a href="#t13:39" class="time">13:39</a>
wwoodsI'll hack up a draft or something<a href="#t13:39" class="time">13:39</a>
f13but unfortunately I don't think it takes group names.  Might be easy enough to make that happen<a href="#t13:40" class="time">13:40</a>
wwoodsyeah, I'll probably need to make some custom tool, but this is something that needs to be able to talk to pkgdb etc so we can say "hello $maintainer, your package $p has been moved to tier $n ..."<a href="#t13:42" class="time">13:42</a>
f13urg.<a href="#t13:42" class="time">13:42</a>
wwoodsgroundwork for QA/Releng stuff<a href="#t13:42" class="time">13:42</a>
* f13 smells the regression from here<a href="#t13:42" class="time">13:42</a>
wwoodswell. that's probably not going to happen anytime soon, if ever<a href="#t13:42" class="time">13:42</a>
wwoodsso pretend I never said that because I don't want to hear the pissing and moaning<a href="#t13:42" class="time">13:42</a>
jwbyou will hear it anyway<a href="#t13:43" class="time">13:43</a>
f13heh<a href="#t13:43" class="time">13:43</a>
wwoodsbut we need to be able to define small subsets of the package universe to target testing on<a href="#t13:43" class="time">13:43</a>
f13having these lists and data is great to let the maintainer make their own decision about it<a href="#t13:43" class="time">13:43</a>
wwoodssaying "TEST EVERYTHING WOOO" doesn't work so good. it's mostly supposed to be informative.<a href="#t13:43" class="time">13:43</a>
wwoodsin the future people might want to make policy based on this information.<a href="#t13:43" class="time">13:43</a>
jwbwwoods, i don't want to be a prick, but if you do that it'll be the opposite from what you've said in the past<a href="#t13:43" class="time">13:43</a>
wwoodsoh?<a href="#t13:44" class="time">13:44</a>
jwbyes.  during the bodhi "behavior" discussions<a href="#t13:44" class="time">13:44</a>
jwbabout auto-promotion, etc<a href="#t13:44" class="time">13:44</a>
jwbanyway, immaterial to the matter at hand<a href="#t13:45" class="time">13:45</a>
wwoodssure, we can talk about that elsewhere<a href="#t13:45" class="time">13:45</a>
wwoodsbut I'm not sure how this is contradicting stuff I've said in the past<a href="#t13:45" class="time">13:45</a>
f13wwoods: so let me know what you might want from me to thelp<a href="#t13:46" class="time">13:46</a>
wwoodsf13: I'll let you know when I have a draft, or if I've got gaps in info/input data<a href="#t13:47" class="time">13:47</a>
f13ok.<a href="#t13:47" class="time">13:47</a>
f13Is there anything else anybody would like to discuss for this week?<a href="#t13:47" class="time">13:47</a>
jwbf13, how long do you want me to collect names for?<a href="#t13:48" class="time">13:48</a>
f13jwb: a week?  2?  Your choice.<a href="#t13:48" class="time">13:48</a>
jwbf13, give me a due date<a href="#t13:48" class="time">13:48</a>
jwbi can adjust from there depending on new submissions<a href="#t13:48" class="time">13:48</a>
jwbbut a week or two works i guess<a href="#t13:49" class="time">13:49</a>
f13In two weeks we should have a prepped list for Legal<a href="#t13:49" class="time">13:49</a>
jwbok<a href="#t13:49" class="time">13:49</a>
f13adjust from there for how much time you'll need to filter requests and such<a href="#t13:49" class="time">13:49</a>
jwbyep<a href="#t13:49" class="time">13:49</a>
warrensent announce<a href="#t13:53" class="time">13:53</a>
f13approved<a href="#t13:54" class="time">13:54</a>
* ricky giggles at <a href="http://en.wikipedia.org/wiki/Monstrous_moonshine">http://en.wikipedia.org/wiki/Monstrous_moonshine</a><a href="#t13:55" class="time">13:55</a>
warrenI had a proxy error when I tried to approve it<a href="#t13:55" class="time">13:55</a>
f13warren: I get that from time to time.<a href="#t13:58" class="time">13:58</a>
f13Ok, I think meeting is over.<a href="#t13:59" class="time">13:59</a>

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