13:02:14 <mvollmer> #startmeeting weekly meeting
13:02:15 <zodbot> Meeting started Mon Apr 11 13:02:14 2016 UTC.  The chair is mvollmer. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:02:15 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
13:02:15 <zodbot> The meeting name has been set to 'weekly_meeting'
13:02:23 <larsu> andreasn: yes (can be a few other things as well, though, like "window pane")
13:02:51 <mvollmer> .hello mvo
13:02:52 <zodbot> mvollmer: mvo 'Marius Vollmer' <marius.vollmer@gmail.com>
13:03:22 <mvollmer> many people will have to leave in half an hour, I guess
13:03:27 <andreasn> .hello andreasn
13:03:28 <zodbot> andreasn: andreasn 'Andreas Nilsson' <anilsson@redhat.com>
13:03:32 <stefw> .hello stefw
13:03:34 <zodbot> stefw: stefw 'Stef Walter' <stefw@redhat.com>
13:03:38 <mulhern> .hello mulhern
13:03:39 <zodbot> mulhern: mulhern 'None' <amulhern@redhat.com>
13:03:49 <andreasn> slightly late, had to get coffee in the mug with the green clothed dude from Moomi
13:03:54 <andreasn> Snusmumriken?
13:04:08 <mvollmer> hmm
13:04:17 <mvollmer> nurskamuikuinen
13:04:26 <mvollmer> have to look up the spelling
13:04:27 <andreasn> Snufkin
13:04:38 <andreasn> in English
13:04:43 <andreasn> anyway, sorry to divert :)
13:05:06 <mvollmer> :-)
13:05:21 <mvollmer> is anyone around for more than 30 minutes today?
13:05:43 <larsu> yeah. What's in 30 minutes?
13:05:54 <mvollmer> som secret meeting
13:06:32 <mvollmer> okay, let's collect agenda
13:06:36 <mvollmer> #topic Agenda
13:07:02 <mvollmer> * storage tests on debian, udev confusion
13:07:11 <mvollmer> * docker storage setup
13:07:37 <larsu> * test- vs release-tarballs and build system
13:09:23 <mvollmer> okay!
13:09:33 <mvollmer> #topic  storage tests on debian, udev confusion
13:09:47 <mvollmer> the PR just got merged that enables storage tests on Debian
13:09:54 <mvollmer> with storaged from Experimental
13:10:03 <mvollmer> iscsi isn't supported yet
13:10:18 <mvollmer> multipath works differently enough to need some extra effort from us
13:10:34 <mvollmer> and then Debian exposes a race condition between udev and storaged
13:10:53 <mvollmer> that has always been there, I think, and recently it got worked around in udisks2 storaged.
13:11:05 <mvollmer> i think this is all aweful
13:11:39 <mvollmer> udevd issues a re-read-partitions ioctl, and the kernel sends remove/add events for all partitions, even if nothing has changed
13:12:06 <larsu> storaged should be able to handle that, no?
13:12:27 <mvollmer> yes, it should, but it didn't until recently.
13:12:30 <mvollmer> same for udisks2
13:13:23 <mvollmer> details are here: https://github.com/storaged-project/storaged/commit/eb4b994c8a12c77744431f1a8459126889c647f9
13:14:35 <mvollmer> i think there were a lot of errors during formatting, even earlier, but it so happens that we don't show them in the UI.
13:14:45 <mvollmer> since formatting happens in the background as a job
13:14:56 <mvollmer> and we don't have error reporting for failed jobs
13:15:21 <mvollmer> mostly the error probably was "failed to synch after mkfs"
13:15:33 <mvollmer> which doesn't really matter since the main job is done
13:15:43 <mvollmer> okay, too many details. :)
13:16:46 <mvollmer> summary: Debian tests are running, still some work to do.
13:16:55 <mvollmer> next?
13:16:55 <larsu> great :)
13:17:22 <mvollmer> #topic docker storage setup
13:17:36 <mvollmer> so, progress is being made :-)
13:18:04 <mvollmer> hard to say exactly where we are, but no major blockers are in sight
13:18:15 <andreasn> nice
13:18:28 <mvollmer> right now I am thinking we need some support from docker-storage-setup
13:18:39 <mvollmer> maybe a "--check" mode, or a "--status" mode
13:18:52 <mvollmer> so that we can put reliable information into the UI.
13:19:05 <mvollmer> and not just the config file, which can be anything
13:19:24 <mvollmer> we might also need help to identify the root device
13:19:37 <mvollmer> i have to think about that
13:20:41 <mvollmer> i think i can make a little screencast in the next few days
13:22:10 <mvollmer> next?
13:22:26 <andreasn> yup
13:22:27 <mvollmer> #topic test- vs release-tarballs and build system
13:22:31 <cockpitbot> 2 tests failed - http://fedorapeople.org/groups/cockpit/logs/master-27781a8b-verify-debian-unstable/log.html
13:22:57 <larsu> yeah. I changed the debian package in tools/debian to be a non-native package (as per debian policy)
13:23:07 <larsu> which of course broke testing
13:23:30 <larsu> looking into that, I noticed that we're testing a weird tarball (made by tools/make-source) instead of what we would release
13:24:00 <mvollmer> yes
13:24:10 <larsu> and I really think we should change that. We should test exactly what would be released if I were to tag the commit immeditaly after getting it in
13:24:36 <mvollmer> so, make dist?
13:25:00 <larsu> stefw commented that running configure and make dist might be too slow, but I think it should be negligible compared to running the tests themselves
13:25:30 <larsu> mvollmer: right. Ideally, the exact same scripts we use for releasing (minus publishing and tagging)
13:25:35 <larsu> well, and adding the testing bits
13:25:37 <mvollmer> we might need to run "make all dist"
13:25:47 <larsu> what's the difference there?
13:25:59 <mvollmer> first, it builds the whole thing
13:25:59 <larsu> doesn't dist depend on all?
13:26:03 <mvollmer> maybe
13:26:16 <larsu> automake docs say it should
13:26:25 <mvollmer> my point was that we do need to build, probably, which is the slow part
13:26:37 <larsu> to make dist?
13:26:38 <mvollmer> but yeah
13:26:41 <mvollmer> yes
13:26:46 <larsu> ah, I see what you mean now
13:26:54 <mvollmer> anyway, if you are aware of that, then all is good.
13:27:06 <larsu> mvollmer: do you think we should build inside the vm?
13:27:14 <larsu> make-source is run outside right now
13:27:18 <larsu> and the tarball is copied
13:27:22 <mvollmer> i think we can do that in the container
13:27:26 <larsu> ok
13:27:38 <mvollmer> like the release script
13:27:41 <mvollmer> right?
13:27:54 <mvollmer> maybe we can have a single container image for both?
13:27:58 <larsu> yes, the release script does it in a container
13:28:04 <mvollmer> would that make things simpler, or more complex?
13:28:24 <larsu> that would be great - but quite the project considering the release scripts are in a separate repository
13:28:47 <larsu> I have the feeling it would make things simpler overall
13:29:00 <larsu> and we would test closer to what we release, hopefully
13:29:00 <mvollmer> yeah
13:29:21 <mvollmer> we could also run "make distcheck"
13:29:29 <mvollmer> travis runs that now, I think.
13:29:52 <larsu> can we use the tarball it produces?
13:30:04 <mvollmer> I don't think so.
13:30:05 <larsu> how does travis fit into all of this?
13:30:17 <larsu> hm, maybe we should have a separate meeting about htis
13:30:25 <mvollmer> it runs the unit tests, with and without valgrind
13:30:38 <larsu> ah ok
13:30:39 <mvollmer> we might be able to let the container that job
13:31:00 <mvollmer> but that's another topic
13:31:23 <larsu> want to talk about this in more detail after the meeting? (maybe in a call even?)
13:31:47 <mvollmer> yes, we need stefw also
13:31:51 <mvollmer> and not today
13:31:52 <larsu> indeed
13:31:54 <larsu> ok
13:32:22 <mvollmer> when I did the first Debian bits, my initial idea was to not do any packaging, actually
13:32:42 <mvollmer> and just do "configure; make; make install" right in the test vm.
13:32:55 <mvollmer> but then we deciced to use packages to get closer to the real thing.
13:33:00 <mvollmer> so let's go all the way
13:33:02 <mvollmer> IMO
13:33:14 <larsu> right, I saw the remove-with-autoconf sed :)
13:33:21 <larsu> but now they've diverged enough
13:34:29 <mvollmer> it will be good to really sort out what goes into a release and what doesn't
13:35:02 <mvollmer> okay, I think I have to leave...
13:35:10 <mvollmer> should we close the meeting?
13:35:22 <larsu> yes
13:35:32 <mvollmer> yep, thanks!
13:35:36 <mvollmer> #endmeeting