15:02:13 <pknirsch> #startmeeting Fedora Base Design Working Group (2014-10-24)
15:02:13 <zodbot> Meeting started Fri Oct 24 15:02:13 2014 UTC.  The chair is pknirsch. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:02:13 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
15:02:22 <pknirsch> #meetingname  Fedora Base Design Working Group
15:02:22 <zodbot> The meeting name has been set to 'fedora_base_design_working_group'
15:02:28 <pknirsch> Alright, lets chair everyone :)
15:02:35 <vpavlin> hi
15:02:58 <pknirsch> #chair haraldh vpavlin dgilmore masta msekleta jreznik
15:02:58 <zodbot> Current chairs: dgilmore haraldh jreznik masta msekleta pknirsch vpavlin
15:03:06 <jreznik> hi!
15:03:18 <msekleta> hi!
15:03:25 <pknirsch> hey everyone!
15:03:28 <ffesti> Moin!
15:03:40 <pknirsch> so lets quickly start with the topics for today
15:03:45 <pknirsch> #topic Status buildrequires cleanup work (davids & nils!)
15:03:57 <pknirsch> Any updates from nphilipp or dazo_afk
15:03:58 <pknirsch> ?
15:04:24 <pknirsch> (and lets make it a bit shorter than last week as we have probably a bigger discussion with the etc stuff ahead of us ;))
15:05:08 <nphilipp> We've both worked on an initial shot at the results database, dazo_afk seems to prefer raw SQL and PL/PSQL stored procedures, I'm more the "high-level" ORM kinda guy.
15:05:32 <nphilipp> But we've hashed out a schema that should enable us to get going.
15:06:21 <nphilipp> I've pinged Karsten about the beefy build machine, but haven't followed up. The PSU of our results machine kinda died and we swapped in the one from it's sibling.
15:06:42 <nphilipp> Short enough?
15:06:44 <nphilipp> :D
15:07:02 <haraldh> <-
15:07:24 <nphilipp> haraldh: ?!
15:07:40 <haraldh> sorry... disconnect somehow
15:08:39 <haraldh> "<-" was still in the out buffer of xchat it seems
15:09:55 <pknirsch> sounds good
15:10:06 <pknirsch> so lets move on to topic 2! :)
15:10:29 <pknirsch> #topic Revisit/update factory-reset/config subpackage
15:10:46 <pknirsch> First up, thanks ffesti for being able to make it to our meeting today
15:11:02 <pknirsch> with that i'll hand it over to haraldh to start the discussion :)
15:11:25 <haraldh> so, yeah, basically we want two features
15:11:33 <haraldh> more or less independent of each other
15:11:47 <haraldh> 1. factory reset and diff
15:11:51 <haraldh> 2. config subpackages
15:11:52 <ffesti> .oO(only two months til Xmas)
15:12:00 <haraldh> ffesti, and then
15:13:02 <haraldh> http://paste.fedoraproject.org/144892/14163549/
15:13:18 <haraldh> for everybody to read as a starting point.
15:13:30 <haraldh> will post some of it on the fedora devel list soon
15:13:44 <ffesti> Wrt feature 1: I am still not sure if packaging the files is the right solution
15:13:49 * jreznik is reading
15:13:58 <ffesti> or if doing some runtime solution is more practical
15:14:14 <ffesti> especially in the light of 3rd party packages
15:14:40 <haraldh> I am fine with either thing
15:14:59 <ffesti> wrt feature 2: I am not sure if creating this sub packages automatically is a good idea
15:15:12 <ffesti> but I agree that we need an easer way of creating them
15:15:13 <haraldh> although inspecting the package contents is different from what is on disk... but that is the case for scripts also
15:15:33 <haraldh> ffesti, yeah, a conversion script could be handy
15:15:37 <ffesti> easier
15:16:16 <haraldh> combined with a rpmlint, which denies mixed non-config/config packages
15:16:24 <ffesti> What I have in mind is rather having some kind of template that can be added to the package with a one-liner
15:16:52 <ffesti> I really do not think that such a crow bar approach will work well
15:17:31 <haraldh> vpavlin, would that work for you?
15:17:43 <haraldh> sgallagh, had an additional point
15:17:59 <ffesti> My guess is that the split up is probably best done package by package
15:18:00 <haraldh> how to build different config sub variants in one build?
15:18:09 <ffesti> but it is not my distribution
15:18:21 <haraldh> not?
15:18:24 <sgallagh> haraldh: That was in fact the original question :)
15:18:36 <sgallagh> Right now, we're sort of emulating it with symlinks, but that's ugly.
15:18:44 <ffesti> sgallagh, I already have a patch solving that
15:19:01 <sgallagh> Right, I remember that. It was just not covered in this particular exchange, so I was unsure
15:19:10 <vpavlin> haraldh: Yes, that sounds acceptable
15:19:15 <ffesti> actually I have the patch for quite a while but to solve other use cases, too, it got a bit more complicated
15:19:48 <ffesti> (use case being rebuild parts of the package with different params)
15:21:19 <ffesti> sgallagh, I think we can make this short: Your use case being considered and should be solved with the next rpm release
15:21:28 <sgallagh> Thanks!
15:21:33 * sgallagh goes back to Beta validation
15:22:22 <haraldh> so, that should solve Feature 2 completely.... I think
15:22:34 <ffesti> haraldh, can you elaborate what of the mentioned systemd functions already do exist and what is their status
15:22:53 <ffesti> haraldh, not really
15:23:36 <ffesti> it will only allow conflicting files between subpackages
15:23:56 <haraldh> systemd-template has all the bits in it ... rawhide at least
15:24:18 <ffesti> if you have enough man power to do all the sub packages by hand you are right and feature 2 is solved
15:24:43 <haraldh> we don't have to convert every package immediately
15:26:10 <ffesti> so you already have a systemd-tmpfiles and systemd-delta command implemented?
15:26:16 <haraldh> yep
15:26:30 <haraldh> delta not yet with the factory, but that is easy
15:26:42 <ffesti> ok, is there a command to actually store files?
15:26:54 <haraldh> "store" ?
15:27:23 <ffesti> like make a copy from a file in /etc/ and store in under factory/
15:27:49 <ffesti> or how are files supposed to get there right now?
15:28:06 <haraldh> in /usr/share/factory?
15:28:14 <ffesti> yup
15:28:17 <msekleta> ffesti, I think that should be other way around
15:28:21 <haraldh> no rpm is packaged that way
15:28:33 <haraldh> IIRC
15:29:09 <ffesti> so there have never been files in /usr/share/factory?
15:29:11 <haraldh> in the ideal case, an rpm does not ship anything in etc
15:29:20 <haraldh> ffesti, nope
15:29:45 <haraldh> no rpm ships anything in  /usr/share/factory
15:30:03 <ffesti> and you do not have any other tools that does anything with it
15:30:07 <haraldh> /usr/share/factory/etc/nsswitch.conf
15:30:08 <haraldh> /usr/share/factory/etc/pam.d/other
15:30:08 <haraldh> /usr/share/factory/etc/pam.d/system-auth
15:30:10 <haraldh> oopsie
15:30:13 <haraldh> systemd already does :)
15:30:29 <haraldh> tstststs
15:31:42 <haraldh> systemd-tmpfiles is the only tool right now
15:31:57 <haraldh> http://www.freedesktop.org/software/systemd/man/tmpfiles.d.html
15:32:03 <haraldh> C
15:32:03 <haraldh> Recursively copy a file or directory, if the destination files or directories do not exist yet. Note that this command will not descend into subdirectories if the destination directory already exists. Instead, the entire copy operation is skipped. If the argument is omitted, files from the source directory /usr/share/factory/ with the same name are copied.
15:33:32 <haraldh> ffesti, why do you ask?
15:33:58 <ffesti> I wonder how much the whole concept is tested
15:34:10 <haraldh> http://particles.surfsite.org/
15:34:23 <haraldh> Here is my small test /usr
15:34:29 <haraldh> you can install and boot it
15:34:39 <haraldh> with the mechanisms mentioned
15:34:51 <haraldh> worksforme
15:35:14 <haraldh> systemd-tmpfiles is used
15:35:28 <haraldh> and sysusers
15:35:36 <haraldh> http://0pointer.net/blog/projects/stateless.html
15:36:10 <haraldh> + the talk Lennart gave on LinuxCon
15:36:57 <ffesti> ok, as we have a set of patches that implement file triggers we'll just uses this feature as one of the test cases
15:37:18 <ffesti> that should backup the files without the need of altering all the packages
15:37:33 <haraldh> Slides: http://0pointer.de/public/linuxconeurope2014.pdf
15:37:42 <ffesti> and is compatible to moving the files in the packages by hand
15:37:56 * jreznik has to leave now to move office -> home before no-go meeting
15:38:06 <pknirsch> cya jreznik !
15:38:11 <haraldh> cya
15:38:29 <haraldh> I have to leave also right now.. but I see things progress, which makes me a happy man :)
15:39:09 <haraldh> thanks ffesti
15:39:26 <pknirsch> cool, sounds great, thanks haraldh and ffesti !
15:39:35 <haraldh> pknirsch, I am sorry, have to leave also
15:39:41 <pknirsch> np :)
15:39:45 <haraldh> bye
15:39:50 <pknirsch> cya haraldh
15:40:50 <pknirsch> so i think that looks like we have a plan to move forward. And you can obviously always hijack haraldh ffesti when you have questions :)
15:41:41 <dgilmore> hey guys i just got to fudcon
15:41:53 <vpavlin> Hey dgilmore
15:42:22 <pknirsch> hey dgilmore, glad you had a safe flight. enjoy your time at fudcon! Where is it this time?
15:43:22 <dgilmore> pknirsch: Nicaragua
15:43:27 <pknirsch> dgilmore: cool :)
15:43:39 <dgilmore> pknirsch: i got here two days ago, just got to the event for the day
15:44:20 <pknirsch> nice :)
15:44:33 <vpavlin> Nicaragua, cool..
15:44:40 <pknirsch> alright, i think we covered the etc stuff, so lets quickly open the floor for other topics
15:44:43 <pknirsch> #topic Open Floor
15:44:47 <vpavlin> dgilmore: any news about base images from RCM side?
15:49:35 <msekleta> can I do quick update on situation around systemd-in-docker in the meantime?
15:49:46 <pknirsch> sure, go ahead msekleta !
15:50:34 <msekleta> so we discussed this topic last week in Dusseldorf quite a bit
15:50:49 <dgilmore> vpavlin: they failed ofr build for Beta RC1
15:50:53 <dgilmore> I need to fix it
15:52:13 <msekleta> I have all patches from systemd ready, actually already pushed the most important one last week, and got ACK from lennart for other two patches
15:52:59 <pknirsch> nice!
15:53:01 <msekleta> so on systemd we are in good shape, but PR for docker (/run on tmpfs) got nowhere so far
15:53:09 <pknirsch> :(
15:53:25 <msekleta> and that PR is pretty much showstopper
15:54:11 <pknirsch> did you discuss that with Lennart and the rest of the folks as well?
15:54:43 <msekleta> yes, we discussed all this on hackfest
15:55:26 <msekleta> on top of /run on tmpfs patch we need there is probably one more thing docker should do and doesn't right know and that is /dev/mqueue
15:55:58 <msekleta> but compared to /run issue it is less serious one
15:56:09 <pknirsch> is anyone working on the tmpfs patch yet?
15:56:21 <msekleta> patch is done
15:57:37 <msekleta> code is ready to merge, but from what I saw on github it seems that there is a lack of interest from docker people
15:58:25 <msekleta> vpavlin, do you think that Alex or Dan could help and maybe ping some folks to draw more attention to this PR?
15:58:30 <pknirsch> have you or anyone else had discussions with them about i?
15:59:21 <vpavlin> msekleta: Sure, I'll ask Dan to do that
15:59:37 <msekleta> pknirsch, Alex Larsson and Dan Walsh
15:59:47 <dgilmore> msekleta: we are way to late for new features in f21
16:00:00 <dgilmore> so I hope you're targeting f22
16:00:16 <vpavlin> Well, we've already talked about it and he continuosly nudges Docker people about his PRs...but....
16:00:47 <msekleta> dgilmore, no F21
16:02:15 <msekleta> dgilmore, and this is not a feature per say, one docker patch and couple of systemd patches, no big deal imho, shouldn't put any burden on QA or RCM
16:02:57 <msekleta> I'd call it enablement
16:04:37 <msekleta> pknirsch, I don't have anything else to add
16:04:47 <pknirsch> okidokie
16:05:28 <dgilmore> msekleta: its too late for f21
16:05:40 <dgilmore> msekleta: seriously right now is bug fix only
16:06:11 <dgilmore> msekleta: small little changes people think are no big deal end up causing lots of work for others
16:06:31 <dgilmore> look at what happened with changing fedora-release-standard to fedora-release-nonproduct
16:06:55 <dgilmore> msekleta: seriously I will block it getting into F21
16:07:09 <dgilmore> i will do everything I can so it only gets in rawhide at this point
16:08:12 <msekleta> dgilmore, systemd patches are very minimal, all reviewed by lennart and others on systemd-devel, you'll not even notice they are there
16:08:49 <msekleta> as for docker patch it is not even in upstream and who knows when it will get there
16:10:09 <pknirsch> yea, lets get the upstream work done first. until thats done there's no point in assuming where the changes will land. but i agree with dgilmore, it's really late for F21 for any non-bugfix changes for core components like systemd
16:10:34 <pknirsch> there's always the possibility to provide testing versions in a COPRS repo though msekleta
16:10:44 <pknirsch> so people can experiment with it if they like
16:11:24 <msekleta> pknirsch, sure
16:11:49 <dgilmore> msekleta: getting it in rawhide and testing it there is perfect
16:11:51 <pknirsch> and right now would be the best time to get things into rawhide for testing
16:11:53 <pknirsch> aye
16:11:58 <pknirsch> exactly
16:12:09 <dgilmore> im not at all saying don't do it
16:12:15 <dgilmore> just don't do it in f21
16:12:18 <pknirsch> :)
16:12:24 <dgilmore> get it in rawhide
16:13:20 <msekleta> dgilmore, right know we can't do it anywhere due to missing functionality in docker
16:14:20 <msekleta> and even once we have the patch in upstream docker, it may never hit F21 because maybe Lokesh will not backport it for one reason or other
16:14:41 <pknirsch> jup, so lets get the work done first, get it into rawhide, work out the any issues that might come up and then decide what to do
16:15:14 * msekleta nods
16:15:35 <pknirsch> alright, sounds we got a plan then as well for the docker/systemd work.
16:17:38 <pknirsch> oki, then lets wrap it up for today. Thanks everyone and have a great weekend!
16:20:08 <pknirsch> #endmeeting