14:33:23 <dgilmore> #startmeeting RELENG (2014-11-03)
14:33:23 <zodbot> Meeting started Mon Nov  3 14:33:23 2014 UTC.  The chair is dgilmore. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:33:23 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
14:33:38 <dgilmore> #meetingname releng
14:33:38 <zodbot> The meeting name has been set to 'releng'
14:33:38 <dgilmore> #chair dgilmore nirik tyll sharkcz bochecha masta pbrobinson
14:33:38 <zodbot> Current chairs: bochecha dgilmore masta nirik pbrobinson sharkcz tyll
14:33:39 <dgilmore> #topic init process
14:33:47 <bochecha> hi
14:33:50 <nirik> morning
14:33:52 * masta is here
14:35:00 * pbrobinson is here
14:35:14 <dgilmore> #topic Secondary Architectures updates
14:35:15 <dgilmore> #topic Secondary Architectures update - ppc
14:36:14 * danofsatx is observing
14:37:31 <pbrobinson> S
14:38:07 <pbrobinson> So we're looking reasonably good, was hoping to get a beta RC out late last week but there was a few issues, hope to have it out today
14:38:18 <dgilmore> cool
14:38:19 <pbrobinson> that's about it really
14:38:31 <dgilmore> Alpha was decent?
14:38:56 <masta> gettext got fixed I believe, right?
14:38:56 <pbrobinson> yes, the major issue appears to be multipathing issues
14:39:13 <pbrobinson> but I believe it's be the same on mainline if it was being tested
14:39:32 <dgilmore> lovely
14:39:45 <dgilmore> #topic Secondary Architectures update - s390
14:39:47 <pbrobinson> encouraging IBM to get involved with the Server product WG to define and deal with that moving forward, likely too late for F-21
14:39:49 <dgilmore> sharkcz:
14:39:58 <dgilmore> you're up
14:41:02 <sharkcz> yep, s390 looks good, I made a new compose last in Wed, but want (and need) to do a newer one to include the latest anaconda roughly the same as beta in primary, but all looks good
14:41:35 <dgilmore> okay
14:41:43 <sharkcz> seems we can install in kvm on s390x
14:41:57 <dgilmore> that has to be nice :)
14:42:03 <sharkcz> yes :-)
14:42:07 <pbrobinson> sharkcz: we'll have a f21-beta tag with most of the core packages to compose against if you want to use that to make it easier to sync to primary
14:43:22 <sharkcz> pbrobinson: ah, thanks for info, the daily branched mash from Saturday should do it too I think
14:43:45 <nirik> I did a stable push friday before the tag was cloned, so it's a bit of a mess. ;( Sorry for that...
14:44:16 <pbrobinson> sharkcz: I plan to compose aarch64/ppc against that tag once we have it so to minimise the moving target
14:44:42 <sharkcz> pbrobinson: good idea
14:46:56 <dgilmore> okay
14:47:19 <dgilmore> #topic Secondary Architectures update - arm
14:47:39 <pbrobinson> so we're looking pretty good there as well
14:47:51 <dgilmore> :) great
14:47:54 <pbrobinson> got to the bottom of the gettext build failure
14:48:21 <pbrobinson> which ended up being a kernel futex bug affecting a number of arches. Props to jcm for finding that one
14:48:48 <pbrobinson> similar to ppc I plan on Beta RC1 at the same time
14:49:22 <pbrobinson> and alpha looked pretty good
14:49:22 <dgilmore> awesome
14:49:43 <pbrobinson> so with luck RC1 might end up as beta in both ppc/arm cases
14:50:44 <dgilmore> :) hope so
14:51:17 <dgilmore> #topic #5931 [Proposal] Move new branch and unretire requests to pkgdb2
14:51:30 <dgilmore> https://fedorahosted.org/rel-eng/ticket/5931
14:51:38 <dgilmore> pingou: around?
14:51:44 <pingou> I am
14:52:35 <pingou> so there are two things left: a) do we have other checks than the ones listed? and b) fixing the new branch script
14:53:00 <pingou> I looked at b) but found out it's a shell script and I got scarred of breaking it more than anything else
14:53:27 <pingou> as for a) dgilmore and I still need to seat at one point and see what we can do :)
14:53:59 <dgilmore> pingou: later in the week we can sit down
14:54:15 <pingou> dgilmore: cool, just let me know when it's good for you
14:54:29 <pingou> just, I'll be offline next week, all but Monday
14:54:50 <pingou> for b), I think tyll_ offered to help
14:55:26 <dgilmore> pingou: ping me wednesday when you start your day
14:55:39 <pingou> dgilmore: will do :)
14:56:28 <dgilmore> #topic #6016 Use fedpkg-minimal in Fedora buildroots
14:56:37 <dgilmore> https://fedorahosted.org/rel-eng/ticket/6016
14:56:45 <dgilmore> so the package will need review
14:56:53 <dgilmore> and we will have to ship it
14:57:31 * nirik is fine with the concept, just needs doing
14:57:34 <pbabinca> There won't be any problem with neither of those, will it?
14:58:00 <dgilmore> pbabinca: just that the package has to be reviewed
14:59:29 <masta> it works in epel7, so there is that. Has it changed any for Fedora use?
14:59:39 <pbabinca> If nobody is against that I can create new package and ask for a review.
14:59:53 <dgilmore> masta: it was used because we had to to bootstrap epel
15:00:04 <dgilmore> it needs to be submitted for review
15:01:27 <dgilmore> #action pbabinca to submit fedpkg-minimal for review
15:01:36 <dgilmore> #topic #5329 RFE: Mention in branched report when we are in a freeze
15:01:48 <dgilmore> https://fedorahosted.org/rel-eng/ticket/5329
15:02:21 <nirik> for this (and a number of other things) we would need some 'official' place to show we are in freeze... if we had a composedb it could be there, but failing that not sure.
15:02:33 <dgilmore> anyone want to take on writing a patch?
15:02:56 <dgilmore> nirik: that would make it easier
15:03:11 <dgilmore> nirik: bochecha is working on composedb
15:03:46 <nirik> because otherwise how do we know? I guess it could be a setting in the script... and just manually changed.
15:03:48 <masta> nirik: yeah I was just wondering how the script would know about the freeze
15:04:38 <dgilmore> nirik:we we could say if there is no change it could be due to freeze
15:04:56 <dgilmore> but to be more specific we need an api
15:06:40 <masta> let's stall on that ticket while we think about the API in #fedora-releng
15:06:51 <dgilmore> huh
15:07:16 <masta> I mean, we just have to decide how it knows about the freeze.
15:08:21 <masta> I'd do something hackish like touch a file and have the script check if that exists, say /etc/freeze or whatever... but an API in a database would be great/better.
15:08:27 <dgilmore> we do not have to
15:08:47 <dgilmore> we could just always add a message
15:09:12 <dgilmore> anyway i removed the meeting keyword and added needs-patch
15:09:58 <dgilmore> #topic #6024 Clear old ACLs when unretiring a package
15:10:09 <dgilmore> https://fedorahosted.org/rel-eng/ticket/6024
15:10:25 <pbrobinson> Th2.K44ks
15:10:32 <pingou> so atm, pkgdb keeps all the ACLs when a package is orphaned
15:10:48 <dgilmore> I am really not sure we should change much here
15:10:49 <pingou> recently we changed it so that the person orphaning the package loose commit and watchcommits ACL
15:11:23 <pingou> I guess the question is: should we drop the ACLs for *all* when a package is orphaned?
15:11:39 <pingou> and tbh, I don't really have an opinion on this
15:11:41 <dgilmore> orphaned no
15:11:46 <dgilmore> retired probably
15:12:06 <dgilmore> i do not have a stong opinion here
15:12:39 * nirik neither.
15:12:48 <pingou> well, retired: the page keeps the info, but it's blocked for its integration with gitolite
15:13:02 <nirik> I guess we could drop all, and if they needed to adjust the dead.package file or something, they can take it, change it, and orphan it again?
15:13:13 <pingou> example: https://admin.fedoraproject.org/pkgdb/package/aeolus-audrey-agent/ but only rel-eng will be able to push to the repo
15:13:43 <pingou> nirik: and that would be only for non-provepackager even :)
15:13:57 <nirik> yeah
15:14:25 <nirik> I guess IMHO, it would make most sense to leave all acls when orphaning, then clear them all at retire?
15:14:36 <pingou> sounds good to me
15:17:26 <dgilmore> nirik: i think so yeah
15:18:32 <dgilmore> anything else here>
15:18:35 <dgilmore> ?
15:19:13 <dgilmore> #action pingou to remove acls on retire
15:19:29 * pingou on it
15:19:32 <dgilmore> #topic #5963 Orphaned vulnerable packages in EPEL
15:19:44 <dgilmore> https://fedorahosted.org/rel-eng/ticket/5963
15:21:15 <dgilmore> tyll_: is going though all of these
15:22:00 <nirik> epel folks were looking at doing orphan->retired in a while.... tyll_ has been sending out lists of orphaned packages.
15:22:06 <nirik> these should get caught in that too.
15:23:09 <pingou> btw, python-gitpython is in this list, which means: fedpkg, pkgdb-cli won't work anymore on el5
15:23:11 <dgilmore> yep
15:23:41 <dgilmore> fedpkg on epel5 is ancient and likely doesnt work
15:23:50 <pingou> ok
15:23:55 <pingou> then I'll drop pkgdb-cli on el5 as well
15:24:12 <dgilmore> current fedpkg neecds a newer python
15:25:01 <dgilmore> #topic #6027 secondary arch old mash trees cleanup
15:25:12 <dgilmore> https://fedorahosted.org/rel-eng/ticket/6027
15:25:23 <dgilmore> pbrobinson: i see you took this
15:26:03 <dgilmore> what we really need is something running as part of bildbranched and buildrawhide that cleans up old mashes automatically
15:26:12 <dgilmore> ideally we keep 2 weeks worth
15:27:18 <pbrobinson> dgilmore: OK, do we do something similar in mainline? I was going to ask what our expected retention was
15:27:26 <dgilmore> pbrobinson: we do not
15:27:29 <dgilmore> we should
15:27:32 * nirik nods.
15:27:43 <pbrobinson> dgilmore: OK, I can look at that
15:27:47 <dgilmore> pbrobinson: right now we go clean up when we remeber
15:28:10 <pbrobinson> dgilmore: any exceptions I should take into account?
15:28:25 <pbrobinson> ie during freeze periods or similar?
15:28:26 <dgilmore> i dont think so
15:28:34 <pbrobinson> or just keep 2 weeks rolling?
15:28:45 <dgilmore> we do not ever fail a compose
15:29:08 <dgilmore> but maybe making sure we do not breal the rawhide/branched symlink
15:29:28 <dgilmore> i think 2 weeks rolling is fine
15:29:43 <dgilmore> we need bodhi to clean up old updates mashes alo
15:29:45 <dgilmore> also
15:29:51 <nirik> yep. 2 weeks +1
15:30:06 <pbrobinson> OK, will initially do a manual clean up on secondaries to recover space and then investigate adding it to the scripts as a rolling removal
15:30:17 <dgilmore> sounds good
15:30:38 <dgilmore> #action pbrobinson  will initially do a manual clean up on secondaries to recover space and then investigate adding it to the scripts as a rolling removal
15:30:55 <nirik> I can clean primary once we are out of freeze.
15:30:56 <masta> cool
15:31:15 <dgilmore> okay, we are at an hour going to do a quick open floor after this, im falling asleep here
15:31:57 <nirik> I had one item for open floor.
15:32:03 <dgilmore> #topic Open Floor
15:32:09 <dgilmore> nirik: :)
15:32:17 <nirik> https://fedorahosted.org/fedora-infrastructure/ticket/4575
15:32:27 <nirik> (was misfiled, should have been in releng trac)
15:32:45 <nirik> but basically is there any reason we still need url and description in scm requests?
15:32:49 <nirik> pingou: ^
15:33:02 <dgilmore> should be a releng ticket i think
15:33:17 <pingou> nirik: not really
15:33:19 <dgilmore> nirik: to initially populate
15:33:26 <pingou> nirik: it basically means new packages won't have the info
15:33:35 <pingou> for max a week
15:33:45 <dgilmore> which is not okay
15:33:46 <nirik> ah true...
15:33:51 <pingou> and anyway iirc, the data is optional already now
15:34:00 <pingou> (as in the script won't break if the data isn't there)
15:34:01 <nirik> unless we could trigger it to run on create?
15:34:20 <nirik> well, url is... I think it errors on no description
15:34:33 <pingou> nirik: it relies on yum's metadata
15:34:42 <dgilmore> i think nirik is right there
15:34:55 <pingou> nirik: summary or description ?
15:35:09 <nirik> short description...
15:35:17 <pingou> so summary :)
15:35:42 <nirik> anyhow, should we just say "needed for initial import" now? or ?
15:35:52 <dgilmore> nirik: i thin so yeah
15:35:56 <pingou> +1
15:36:14 <dgilmore> i guess we could look at extracting it automatically down the road
15:36:21 <nirik> this might change when we move new stuff into pkgdb no?
15:36:31 <dgilmore> yep
15:36:50 <pingou> nirik: yes it will
15:37:00 <pingou> well, people will still have to fill it manually actually
15:37:11 <nirik> pingou: in their request?
15:37:15 <pingou> yes
15:37:17 <nirik> ok
15:37:23 <pingou> nirik: it will change when we move the review off bugzilla
15:37:43 <dgilmore> so going forward there will be change
15:37:53 <dgilmore> but for now please provide the data
15:38:20 <pingou> +1
15:38:50 <nirik> yep. sounds good.
15:38:51 <dgilmore> anything else?
15:38:52 <nirik> thats all I had
15:39:06 <pingou> there is https://fedorahosted.org/rel-eng/ticket/6024
15:39:13 <pingou> that we should look at, at one point
15:39:46 <dgilmore> pingou: we went over that earlier
15:40:08 <pingou> ah ok, sorry
15:40:21 <dgilmore> i think so
15:40:27 <dgilmore> im really really tired
15:40:42 <pingou> dgilmore: just a side note: nirik and I sponsored bochecha into sysadmin so that he can directly help with porting dist-git in ansible :)
15:40:57 <dgilmore> pingou: okay
15:41:00 <dgilmore> :)
15:41:11 <pingou> dgilmore: in other words, he is paying for the drinks :)
15:41:28 <dgilmore> nirik: we need to get pbrobinson access to the builders also
15:41:30 <bochecha> pingou, yeah, I'm paying tonight, too bad you left last week ;)
15:41:37 <pingou> bochecha: damn!
15:42:10 <dgilmore> okay of nothing else
15:42:20 <dgilmore> #endmeeting