13:06:30 <mvollmer> #startmeeting weekly meeting
13:06:30 <zodbot> Meeting started Mon Mar 27 13:06:30 2017 UTC.  The chair is mvollmer. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:06:30 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
13:06:30 <zodbot> The meeting name has been set to 'weekly_meeting'
13:06:40 <mvollmer> .hello mvo
13:06:41 <zodbot> mvollmer: mvo 'Marius Vollmer' <marius.vollmer@gmail.com>
13:06:42 <dperpeet> .hello dperpeet
13:06:44 <zodbot> dperpeet: dperpeet 'None' <dperpeet@redhat.com>
13:06:47 <pitti> .hello martinpitt
13:06:50 <zodbot> pitti: martinpitt 'Martin Pitt' <martin@piware.de>
13:08:19 <mvollmer> #topic Agenda
13:10:01 <pitti> sounds like a very fast meeting today :)
13:10:02 <mvollmer> :-)
13:10:35 <pitti> actually
13:10:41 <pitti> * testing on different architectures
13:10:59 <mvollmer> okay
13:11:06 <mvollmer> #topic  testing on different architectures
13:11:28 <pitti> as part of https://trello.com/c/y9wrQjg1/375-epic-unit-and-integration-tests-on-a-non-x86-architecture I am currently investigating testing cockpit on a big-endian architecture
13:11:56 <dperpeet> pitti, when I last looked at this, emulation speed was an issue
13:11:58 <mvollmer> which one?
13:11:59 <pitti> the most well known/common ones are s390x, ppc64 (NOT -le), and the old powerpc (by and large  dead now)
13:12:18 <pitti> I'm making progress with cross-building and running unit tests in qemu
13:12:23 <dperpeet> ppc64 was the most promising one last I checked
13:12:25 <pitti> but this will be too slow for the integration tests
13:12:41 <dperpeet> I think it would be fine to run daily
13:12:44 <pitti> well, AFAIK with Power8 the -le variant is much more common these days
13:13:00 <pitti> but anyway, my question was which of those you think we should aim for
13:13:09 <pitti> considering that at some point we should get some real iron for that
13:13:17 <pitti> i. e. is it easier to get ppc64 or s390x?
13:13:24 <mvollmer> maybe we can find some real hardware, from red hat
13:13:57 <dperpeet> we can see if either is available
13:13:57 <pitti> a VM instance would suffice, of course
13:13:59 <mvollmer> if we have real hw, do we still need to emulate the unit tests?
13:14:17 <pitti> no, we wouldn't need to any more
13:14:25 <mvollmer> pitti, did you talk to jscotka?
13:14:30 <pitti> this is to get some coverage until we get hardware, and also to possibly find/fix some initial issues
13:14:46 <pitti> mvollmer: that would be my next question -- who to talk to; so jscotka?
13:14:51 <mvollmer> pitti, then I would say we first sort out the hw, and don't 'waste' time on cross compiling and emulation
13:15:21 <mvollmer> pitti, yeah, jscotka probably knows most about how to get hw
13:15:22 <jscotka> mvollmer, pitti what is question?
13:15:25 <mvollmer> or dperpeet
13:15:47 <pitti> jscotka: if we can get a ppc64 machine/VM or s390x LPAR or VM for  cockpit testing
13:16:12 <mvollmer> pitti, are you thinking permanently?
13:16:20 <pitti> yes, otherwise that doesn't make sense
13:16:29 <dperpeet> let's discuss that outside this meeting
13:16:32 <mvollmer> yeah, hmm.
13:16:38 <dperpeet> I think the question for here is: do we want to have that permanently?
13:16:39 <pitti> ok
13:16:47 <pitti> well, I won't work on it as an one-shot project
13:16:55 <pitti> the point is to ensure we don't regress on big-endian
13:17:05 <pitti> if we don't care about these architectures, let's not waste time on those at all
13:17:43 <pitti> (I understood the card as "yes, we do want it")
13:18:09 <mvollmer> and run it on all pull requests?
13:18:16 <mvollmer> that would be awesome
13:18:24 <pitti> that's a capacity issue of course
13:18:38 <pitti> ideally yes, practically we might need to start more slowly
13:19:34 <dperpeet> ok, let's try to get hardware
13:19:36 <pitti> so I'll read up on https://fedoraproject.org/wiki/QA/Beaker
13:19:42 <dperpeet> not sure if we can test every pr though
13:20:39 <pitti> e. g. https://github.com/cockpit-project/cockpit/issues/6041 is an instance of what goes wrong on BE
13:21:26 <pitti> ok, I think I'm done, thanks for the input!
13:21:26 <mvollmer> uhh
13:22:35 <mvollmer> (The "uhh" was about the swapped IP address issue)
13:23:23 <pitti> mvollmer: heh, yes
13:24:09 <mvollmer> alright
13:24:23 <mvollmer> arm is little endian as well, right?
13:24:28 <pitti> network vs. host order, byte arrays vs ints, there's plenty of places for swappiness
13:24:30 <pitti> right
13:24:40 * mvollmer digs out his old amiga
13:25:11 <pitti> oh, the m68k was BE? I don't remember
13:25:22 <mvollmer> pitti, we spend some thought on getting the IP swapping right, but yeah, code is either tested or broken.
13:25:37 <mvollmer> pitti, yep.
13:25:41 <pitti> mips is, but I doubt we can cobble together enough WRT54 routers for this :)
13:25:57 <mvollmer> next topic?
13:26:28 <mvollmer> well...
13:26:34 <mvollmer> #topic Open floor
13:30:37 <mvollmer> okay, looks like we are done.
13:30:40 <mvollmer> Thanks everyone!
13:30:44 <mvollmer> #endmeeting