From FedoraProject

Jump to: navigation, search


Fedora Release Engineering Meeting :: Monday 2008-05-05

F9 Release Candidate

  • A few things holding up the release candidate
  • texlive
  • X crashing on evdev removal during installer (kvms)
  • packagekit that just got built.
  • emacs has broken alternatives and is still on the blocker list
  • conflict between rsyslog and syslog-ng

Posting F9 Release Candiate

  • Once we have RC spun and are testing it we should pre-process the pending F9 updates and get them pushed into directories
  • We should also have tickets filed for releases tasks that we need to make sure happen like export approval filing, etc.

Rawhide Switch to F10

  • BugZappers want to cooridnate with bugzilla versioning and rawhide rebase
  • Most likely rawhide composes starting on May 8th or 9th will become F10
  • f13 to keep poelcat updated

IRC Transcript

-!- f13 changed the topic of #fedora-meeting to: Fedora Release Engineering - Roll Call01:00
f13ping: notting jeremy jwb spot lmacken rdieter wwoods warren01:00
jwbsort of here01:00
* notting is here01:00
* poelcat here01:02
* lmacken here01:02
f13ok, lets get started.01:03
-!- f13 changed the topic of #fedora-meeting to: Fedora Releng - Release Candidate F901:03
f13We know of a few things holding up the release candidate.  texlive, X crashing on evdev removal during installer (kvms), and a packagekit that just got built.01:04
nottingpk is tagged01:04
nottingf13: emacs having broken alternatives is still on the blocker list01:04
f13I thought that got fixed.  Guess not01:04
warrenis the PK build related to the GPG problems?01:04
nottingit's a traceback in the yum backend01:05
spoti can't reproduce that emacs failure01:05
spotat all01:05
spotjrb says he can, but he's not sure how he got his system into that state01:05
* wwoods testing stuff01:05
spotsomething about starting at F8 and updating01:05
wwoodsX-crashing-on-evdev-removal is not all machines01:05
wwoodsI can't produce that bug01:06
f13wwoods: you're likely not getting detected as evdev01:06
nottingf13: have you successfully cornered ajax?01:06
f13notting: was waiting for him to finish lunch01:06
f13and now I need to go find where he actually sits.01:06
wwoodsit should only happen on machines with devices that are keyboard *and* mouse combos, *and* you have to insert/remove a keyboard/mouse01:06
nottingwwoods: hm, would a ps/2 converter count?01:07
f13notting: that may pop as evdev01:07
warrenI have a two PS/2 to USB converter at home01:07
f13wwoods: this isn't a combo, it's a real mouse, usb, plugged into the keyboard, also usb01:07
nottingand is this anaconda only or post-install too?01:07
f13notting: I've only see it during anaconda01:08
f13but I don't spend much time out of anaconda in this setup01:08
f13and from earlier fun with evdev, we know htat the combo detection is a little bad, there are lots of systems that get picked up as combos when they aren't01:08
f13this can also lead to bad keyboard shortcuts post-install if the detection is wrong01:09
f13ajax had said something about neutering evdev to not claim combo devices01:09
nottingyes... but... not a blocker. x crashing may be a blocker01:09
* spot will try to reproduce that emacs blocker again today01:09
f13other than those, I think we're in OK shape.  We're working our way through the testing matrix01:11
f13wwoods: would you consider the syslog-ng and rsyslog conflict a blocker?01:11
warrenf13: is the fix simple?01:11
f13I don't know if wwoods filed a bug on it.01:12
f13warren: don't know.01:12
* notting says 'probably not, as there are certainly multilib conflicts somewhere we haven't noticed'01:12
f13so ideally we'll get the fixes in today for the things that are blockers, and spin the real RC from tomorrow's rawhide.01:14
wwoodsI'm doing a small-e "everything" install from the x86_64 DVD right now01:14
f13also I can start pre-syncing content to PHX since everything is signed and there are going to be minimal changes01:14
wwoodsit's essentially01:14
buggbotBug 441664: low, low, ---, Peter Vrabec, NEW , rsyslog should explicitly conflict with syslog-ng01:14
wwoodsthey both ship /etc/logrotate.d/syslog01:14
wwoodswhich, wtf01:14
f13explicitly conflict is verboten01:14
f13maybe they need to be taught about alternatives01:14
wwoodswhy doesn't rsyslog just ship /etc/logrotate.d/rsyslog01:15
f13wwoods: or that01:15
nottingwwoods: i'm inclined to leave it, it's not a regression, right?01:16
wwoodsno, don't think it's a regression01:16
wwoodsstupid, yes. probably a simple fix, yes. but not a huge problem either01:16
wwoodssolution: don't install both packages01:17
wwoodsI guess we could also fix it by dropping syslog-ng from the media01:17
nottingit's on the media????01:17
warrenthat doesn't belong there01:17
nottingit shouldn't be01:17
wwoodsyeah, that's the whole problem01:17
nottingf13: can you fix *that*?01:17
wwoodsinstalling everything in comps on the media (i.e. checking every "install all optional packages" box for everything in the installer)01:17
wwoodsgives you "file conflict" crashy badness01:18
nottingwwoods: from reading it, the f8 packages ATM have an explicit Conflicts:. so i'd rather not rush a special fix for f901:18
wwoodssolution: uncheck syslog-ng01:18
wwoodsbetter solution: don't put it on the stupid media01:18
wwoodsshipping a package known to conflict with something in @base seems obviously bad01:19
metherwhy did we put it on the media before?01:19
f13need to find out why it's on the media01:19
nottingf13: this is presumably b/c of pungi pulling all providers of 'syslog'?01:19
f13mether: we didn't explicitly01:19
f13notting: oh yeah, I bet that's it.01:19
f13we can just drop a -rsyslog in the manifest01:19
nottingf13: otherway01:19
f13er yea01:19
notting+1 to -syslog-ng in the manifest01:20
warrentotally +101:20
f13I'll toss that into pungi today and ask for a tag request.01:20
f13incidentally that'll require a bumped fedora-release too01:20
f13(since we ship the config there as well)01:20
warrennot a big loss01:21
nottingf13: tangent: did the ppc compose box survive the move?01:21
f13notting: hrm, I haven't seen clumens in yet01:21
f13I could hook it up in my cube for now.01:21
wwoodsthought he was on a plane?01:21
f13long plane ride01:22
wwoods      <packagereq type="optional">syslog-ng</packagereq>01:22
wwoodsyeah it's in comps?01:22
f13wwoods: optional though01:22
f13wwoods: we don't pull optional packages into the media by default, only specific ones/groups01:22
nottingf13: aside from that, are there any other infrastructural issues preventing a rc compose?01:22
wwoodsis -syslog-ng sufficient?01:22
f13this is just a case of "give me everything that provides 'syslog'"01:22
f13wwoods: should be.01:22
f13notting: none that I know of.01:23
f13jwb: ping; have you done any ppc composes lately or installs from rawhide?01:23
warrenf13: is a newer tree coming on reducto soon?01:23
jwbf13, i was supposed to have a machine to do so this week.  no idea if that will actually happen01:23
jwb(so no)01:23
jwbalso, i've been out quite a bit over the last 2 weeks.  family issues01:23
f13warren: there has been only 1 package change since the last reducto tree.01:24
f13jwb: ok.01:25
warrenah, ok01:25
f13we'll fly a bit blind with ppc then.01:25
f13I no longer have machines to test with01:25
jwbf13, understandable.  we fixed the dvd size issue, yes?01:25
f13I do believe so01:25
jwbk.  we should pester dwmw2_gone01:25
f13those excludes are in the pungi package now.  For reals.01:25
jwbfo shizzle01:25
f13alright, anything else regarding the release candidate?01:26
warrenwwoods:  how did people on the outside get the RC in order to test this?01:26
nottingf13: one? there's PK and allegro01:26
notting(at least)01:26
f13warren: they used rawhide, which is acceptable.01:26
f13notting: those didn't make today's rawhide01:27
nottingf13: sorry, confusing last tree/rawhide/post-rawhide-today01:27
f13notting: I meant changes in rawhide.  Pulling packages down and making a local repo is possible, manually figuring out what should be multilib.01:27
wwoodswarren: presumably they used rawhide, or built their own images01:27
nottingf13: ok, have box with ps/2 keyboard, usb keyboard, usb mouse, and two ps/2->usb converters attached. :)01:28
warrenInstallSourceHardDrive being broken is a little sad01:28
f13warren: some people do use pungi locally to create media too, which is also acceptable/encouraged.01:28
f13notting: scary01:28
f13warren: yeah, but without a clumens, I'm not sure we'll see a fix for that soon01:28
-!- f13 changed the topic of #fedora-meeting to: Fedora releng - post-release candidate01:29
warrenwhere did clumens go?01:29
f13once we have RC spun and are testing it, we should pre-process the pending F9 updates and get them pushed into directories to make sure that's all hot to trot01:29
f13warren: I don't know01:29
* warren tries it01:30
f13also, we should have a ticket somewhere for release tasks, or af ew tickets.  I probably need to look into those and make some things happen, like export approval filing and such01:31
nottingf13: hey, i broke it!01:31
f13notting: crash?01:31
nottingf13: yeah01:31
nottingnow, how to debug???01:31
f13nod.  I'll hunt ajax down after the meeting01:31
f13notting: good luck with that.01:31
f13it's evdev, I seriously think the answer here is stop evdev from claiming the device in this situation01:32
nottingf13: evdev didn't claim anything according to the log01:32
f13notting: really?!01:32
nottinglemme try again, but it didn't appear to01:32
f13notting: the last thing I see in the log I attached was evdev removal.01:32
f13er, last thing before the backtrace01:32
warrenhm, do we have any means to make a USB installer now?01:33
warren(not live)01:34
wwoodsthere's a script.. somewhere01:34
f13warren: nothing official.01:35
warrenwhere's the unofficial?01:35
f13the script I posted to anaconda-devel list many moons ago01:35
-!- f13 changed the topic of #fedora-meeting to: Fedora releng - Open Floor01:36
warrenf13: if we find a fix for the install from hard drive in the next day is it too late?01:37
f13warren: probably not.01:38
warrenwwoods: is a bug filed for that?01:38
wwoods was the NFS iso one01:39
buggbotBug 443291: medium, low, ---, Anaconda Maintenance Team, CLOSED RAWHIDE, ISO image via NFS, got "Missing ISO 9660 image" and failed to install01:39
f13nfsiso is confirmed fixed01:39
buggbotBug 443580: low, low, ---, Chris Lumens, CLOSED DUPLICATE, Hard disk installation fails01:39
wwoodsis the hdiso variant01:39
wwoodswhich was closed as a dupe of the former01:39
f13hdiso is now failing for some other reason01:39
warrenf13: could you throw that into source control anywhere?01:39
wwoodsguess I'll reopen it?01:39
poelcatf13: did you see my mail asking when rawhide will start composing from devel branch again?01:39
f13wwoods: probably best, and throw in the latest logs which should show something else.01:39
warrenwwoods: do we have a full anaconda log of the failure?01:39
warrenwwoods: the new failure01:39
f13poelcat: I did not, that'll likely be once I start syncing the GOLD bits to the mirrors01:40
poelcatf13: want to coordinate it w/ bugzilla versioning01:40
* f13 thinks.01:40
f13We kind of have to wait until F9 goes public due to mirror manager redirection01:40
f13or else people will just get yum failures for a few days01:41
wwoodswarren: they're not interesting01:41
wwoodsbut yeah, I can get some01:41
poelcatf13: so that would mean, not until 2008-05-13 ?01:42
f13poelcat: actually, what we might be able to do, is create empty updates repos and get those synced out, and point all f9 requests to the udpates directory01:43
f13poelcat: that way we can let rawhide move on earlier than that, this week, and not break people who are on F9.01:44
f13poelcat: I'll start a conversation with mdomsch and co about this01:44
f13poelcat: I'd say either the 8th or the 9th rawhide will become F1001:44
poelcatf13: okay, that is helpful... can you keep me in the loop?01:44
poelcati'll re-fwd that mail01:45
nottingf13: crashing in core server code01:46
f13Ok, if there is nothing else, I'm going to go hunt down ajax.01:48
f13Ok, rock and roll.01:49

Generated by 2.3 by Marius Gedminas - find it at!