13:03:06 <andreasn> #startmeeting
13:03:07 <zodbot> Meeting started Mon May 25 13:03:06 2015 UTC.  The chair is andreasn. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:03:07 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
13:03:14 <andreasn> .hello andreasn
13:03:15 <zodbot> andreasn: andreasn 'Andreas Nilsson' <anilsson@redhat.com>
13:03:23 <stefw> .hello stefw
13:03:24 <zodbot> stefw: stefw 'Stef Walter' <stefw@redhat.com>
13:04:27 <mvollmer> .hello mvo
13:04:28 <zodbot> mvollmer: mvo 'Marius Vollmer' <marius.vollmer@gmail.com>
13:04:51 <andreasn> anyone else?
13:05:23 <andreasn> #topic Agenda
13:05:24 <stefw> holiday in US and germany
13:05:27 <stefw> er
13:05:32 <andreasn> today?
13:05:36 <stefw> yup
13:05:38 <andreasn> oooh
13:05:49 <stefw> * Docker exec and kubernetes containers view
13:05:51 <andreasn> that explains why I didn't get hold of sue :)
13:05:55 <andreasn> * branding
13:06:31 <stefw> * frame message protocol
13:06:32 <mvollmer> * Fedora rawhide
13:07:08 <stefw> * use of less
13:08:02 <andreasn> sounds good
13:08:02 <stefw> * outstanding pull requests
13:08:51 <andreasn> ok, lets start
13:08:53 <andreasn> #topic Docker exec and kubernetes containers view
13:09:12 <stefw> for the kubernetes containters view, i've been working on being able to show the console *and* spawn a shell inside a container
13:09:30 <stefw> to do that
13:09:40 <stefw> we need to have docker exec support
13:09:45 <stefw> which is a way of starting a process inside a running container
13:09:53 <stefw> this means using a later version of the docker API at least for those calls
13:09:58 <stefw> are we okay with that?
13:10:14 <mvollmer> yes
13:10:29 <stefw> example here: https://rawgit.com/cockpit-project/cockpit-design/master/kubernetes/pods.html
13:10:31 <mvollmer> we should move to a newer version in general, I'd say.
13:10:47 <stefw> i think 1.15 is the earliest version with docker exec
13:10:58 <stefw> but i can't find out what actual docker version that equates to
13:11:52 <stefw> in addition is this spawn functionality something we should retrofit into our docker specific containers view?
13:12:20 <mvollmer> hmm
13:12:43 <mvollmer> it would be natural to have individual terminals
13:13:11 <mvollmer> instead of a "execute this string" dialog.
13:13:43 <stefw> yes, i've hard coded this to spawn /bin/sh
13:13:44 <mvollmer> could we do "docker exec /bin/bash" and have a shell session in a temrinal?
13:13:44 <stefw> well
13:13:46 <stefw> /bin/sh -i
13:13:47 <stefw> yup
13:13:51 <stefw> see the link i posted above
13:13:57 <stefw> whin you click on _Shell_
13:14:04 <stefw> it basically does
13:14:10 <mvollmer> ahh, that's what it is.
13:14:13 <stefw> 'docker exec -ti containerid /bin/sh -i'
13:14:20 <stefw> but via the REST API
13:14:26 <mvollmer> yep
13:14:54 <mvollmer> yes, I think that fits into the 'normal' docker UI very well
13:15:06 <stefw> yeah, need to figure out with andreasn where the action butonts fit
13:15:08 <stefw> if we have terminal tabs
13:15:15 <stefw> but shouldn't be insurmountable
13:15:31 <stefw> not *real* terminal tabs
13:15:34 <mvollmer> might be a better experience than attach/log
13:15:44 <stefw> attach/log is really useful
13:15:47 <stefw> to see what's happening
13:15:50 <mvollmer> but needs a /bin/sh in the container
13:15:54 <stefw> and the exec shell is useful to make something happen
13:15:56 <mvollmer> which most should have, no?
13:16:05 <stefw> yeah, it's pretty much required by unix software
13:16:14 <stefw> otoh, the various commands like /bin/ps
13:16:16 <stefw> are often not present
13:16:18 <stefw> but such is life
13:16:21 <stefw> only so much we can do
13:17:23 <andreasn> I imagine for the docker container view it could only be a switch between console and shell, since we have more space for the status than in the pod list
13:18:16 <stefw> yeah could be
13:18:25 <andreasn> but lets see
13:19:23 <andreasn> next topic?
13:20:06 <andreasn> #topic branding
13:20:48 <andreasn> so we did a bunch of changes based on feedback from fedora that we can't ship the branding, but that needs to be shipped in another package (fedora-logos)
13:21:33 <stefw> the mechanism we currently use to display *any* logo is a dangling symlink
13:21:43 <andreasn> so yeah
13:21:45 <stefw> so if the file happens to be in the right place in /usr/share/pixmaps ... it gets used
13:22:15 <stefw> in my opinion this mechanism shouldn't be used for more than the logo
13:22:28 <stefw> and not to build complex branding
13:22:42 <andreasn> I think longer term we need to make sure the right elements are shipped with the fedora-logos package, and then make sure that that is shipped with all RHELs, CentOS etc
13:23:08 <stefw> fedora-logos will be shipped by all distros?
13:23:20 <andreasn> only the ones that are based on fedora :/
13:23:24 <stefw> that seems like a challenging task
13:23:52 <stefw> not sure how that works
13:23:58 <stefw> i think due to trademarks only fedora ships fedora-logos
13:24:06 <stefw> not that i care
13:24:49 <stefw> anyway, when the right logo is not present for the upper right corner, we substitute in the cockpit logo
13:24:57 <andreasn> right
13:24:57 <stefw> that's how the current implementation works
13:26:01 <stefw> anyway, i think the current implementation is good enough for now right?
13:26:19 <stefw> i don't think we can mark it stable yet
13:26:37 <stefw> and therefore we can't get too comilicated with depending on external images/logos, in anything but a very rudimentary way
13:26:40 <andreasn> possibly, not sure if it's using the right RHEL logo, but we can take that elsewhere
13:27:02 <stefw> yeah
13:27:31 <andreasn> good enough for now. Thanks for fixing that up
13:27:33 <stefw> the rhel logos package contains files called 'fedora-xxxx'
13:27:41 <stefw> and those contain red hat logos
13:28:01 <andreasn> same for centos?
13:28:03 <stefw> no idea
13:28:11 <andreasn> I'll check
13:28:12 <stefw> i posted information about the new branding rework on cockpit-devel
13:28:21 <andreasn> I have about a dozen OSes in VMs now
13:28:43 <andreasn> (actually 5)
13:29:03 <andreasn> ok, lets move on
13:29:17 <andreasn> #topic frame message protocol
13:29:55 <stefw> the current way we send messages to parent windows does't work for embedding well
13:30:06 <stefw> in particular it doesn't work when the parent window is not from the same domain
13:30:25 <stefw> if we want to support that case, then we have to change how the component detects the parent frame
13:30:32 <stefw> perhaps by using window.name or something like that
13:30:43 <stefw> instead of inspecting properties on window.parent.xxxx
13:31:10 <stefw> if we do change this, it needs to be done in a backwards compatible way, since our backwards compatibility (to 0.48) depends on this)
13:31:33 <stefw> that's one reason i'm again waiting on https://github.com/cockpit-project/cockpit/pull/1978
13:32:00 <stefw> that's it on that topic
13:32:42 <andreasn> all right
13:32:47 <andreasn> #topic Fedora Rawhide
13:33:18 <mvollmer> yep
13:33:39 <mvollmer> I am working on the CI machinery again a bit
13:33:50 <mvollmer> based on the script pointed out by jan last time
13:34:13 <mvollmer> the idea is to create the test machine imahes from scratch, without the magic base tarball
13:34:26 <mvollmer> that works ok, expect with hubbot
13:34:37 <stefw> there's always something :S
13:34:39 <mvollmer> because virt-install seems to need a pty
13:34:55 <mvollmer> i think we can give it one with "expect"
13:35:26 <mvollmer> anyway, when that works, we can use it to make rawhide images
13:35:33 <stefw> ssh -t
13:35:35 <mvollmer> if not, we make them old-style
13:35:53 <mvollmer> hmm, worth a try
13:35:56 <stefw> 'ssh -t' can be used to create a pty
13:36:02 <stefw> even if you have to do silly tihngs like
13:36:06 <mvollmer> yeah, I read the bug comments. :)
13:36:24 <mvollmer> somebody said ssh itself needs a pty for that, though.
13:36:29 <stefw> oh right
13:36:37 <stefw> is this just about sudo?
13:36:45 <mvollmer> no
13:36:47 <stefw> ok
13:36:52 <mvollmer> it's weird
13:37:05 <mvollmer> virt-install runs "virsh console"
13:37:12 <mvollmer> but cirsh console needs a tty
13:37:16 <mvollmer> *virsh
13:37:43 <mvollmer> alternative is to just wait until the domain has shut down itself, but then we lose all output.
13:38:04 <stefw> interesting
13:38:08 <stefw> thanks for cleaning this stuff up
13:38:12 <stefw> the pull requests so far look good
13:38:37 <mvollmer> if only they would work :-)
13:39:01 <mvollmer> i have done some rawhide testing already, and there are many failures, but with a pattern.
13:39:17 <mvollmer> i haven't looked seriously yet, though.
13:40:21 <stefw> to be honest rawhide is a real heap of compost
13:40:38 <stefw> it's going to be hard to have stable CI against it
13:41:00 <mvollmer> i guess
13:41:02 <stefw> but worth seeing what it's like, i guess
13:41:06 <mvollmer> eraly warning system
13:41:09 <stefw> true
13:41:26 <mvollmer> we shouldn't let hubbot make it all red because of rawhide
13:41:35 <stefw> right
13:41:39 <mvollmer> next topic?
13:41:43 <andreasn> yup
13:41:46 <stefw> likely a report posted?
13:41:51 <stefw> about failures on rawhide eventually?
13:41:58 <stefw> like the ones we get from Coverity?
13:42:23 <mvollmer> yes
13:42:31 * stefw nods to next topic
13:42:47 <andreasn> #topic Use of less
13:43:30 <stefw> i'm wondering if we can start using less to build patternfly with our font paths and see how that goes
13:43:44 <stefw> beyond that i notice that our CSS is very repetitive with colors and such
13:43:56 <stefw> and i wonder if a CSS compiler like less would help us
13:44:16 <stefw> the big issue is that we like to have a development model where people can contribute without a build step
13:44:19 <stefw> at least that was the idea
13:44:21 <andreasn> also adds a bit of building complexity. Does it bring in any dependencies?
13:44:27 <stefw> well it depends on nodejs
13:44:33 <stefw> which we use during development
13:44:44 <stefw> the big issue is that if we make it part of our normal 'make' workflow
13:44:46 <andreasn> I always compile after any css change anyway
13:44:52 <andreasn> because that'
13:44:54 <stefw> then we wouldn't be able to run cockpit right out of our cockpit repository
13:44:58 <stefw> but would always have to do an install
13:44:58 <andreasn> s how I have things set up
13:45:07 <stefw> and that's a non-starter as far as i'm concerned
13:45:33 <stefw> if we had a separate 'update-css' target that changed files in the repository
13:45:39 <stefw> and then you committed the less and css files together?
13:45:44 <stefw> anyway, just ideas to think about
13:45:59 <stefw> more concretely i'd like to start using less to bringi n patternfly and bootstrap in the 'make update-lib' target
13:46:04 <stefw> rather than the silly sed script we have now
13:46:10 <stefw> that edits the files and replaces font paths
13:46:19 <andreasn> right
13:46:51 <mvollmer> we used to use sass
13:47:06 <mvollmer> i think that went away when we switched to patternfly
13:47:28 <mvollmer> http://sass-lang.com/
13:47:39 <stefw> right
13:47:39 <mvollmer> or did we use less?  can't remember tbh
13:47:44 <stefw> but patternfly has less in its build system
13:47:50 <mvollmer> yes
13:47:50 <stefw> and its bower component, ready to use
13:48:50 * stefw gives the mic back to andreasn
13:49:05 <andreasn> #topic outstanding pull requests
13:49:17 <andreasn> I need to review this one https://github.com/cockpit-project/cockpit/pull/2303
13:49:48 <andreasn> but https://github.com/cockpit-project/cockpit/pull/2075 for example I have no idea what to do about
13:49:51 <stefw> we should try to help fridex's pull requests move along
13:50:18 <andreasn> or https://github.com/cockpit-project/cockpit/pull/2062
13:50:23 <stefw> does #2075 that need a Feature page?
13:50:36 <stefw> again, does it need a feature page?
13:50:51 <stefw> worth asking fridex to put them together?
13:51:01 <stefw> to understand what is really trying to accomplish here?
13:51:12 <andreasn> yes, I would need some rationale on why to add them
13:51:20 <andreasn> I can ask him for that
13:51:32 <stefw> i feel ilke passward expiry should be managed centrally
13:51:34 <stefw> in IPA
13:51:37 <stefw> and not on individual servers
13:51:41 <stefw> as a best-practice use case
13:51:44 <andreasn> right
13:51:46 <stefw> obviously you *can* do it individually
13:51:53 <stefw> but we don't need to have everything that's possible to do in cockpit
13:51:59 <stefw> but #2062 doesn't seem so bad
13:51:59 <andreasn> exactly
13:52:10 <stefw> oh i mean #2075 doesn't seem to bad
13:52:13 <stefw> the advanced options
13:52:33 <stefw> still needsdesign though
13:52:43 <andreasn> yes
13:52:57 <stefw> what about: https://github.com/cockpit-project/cockpit/pull/1965
13:53:01 <stefw> i've done some work on blankslate
13:53:04 <stefw> that i'll post
13:53:19 <stefw> this is blankslate: https://github.com/cockpit-project/jquery-amend/pulls
13:53:24 <stefw> and i showed it in a demo recently
13:53:38 <stefw> where you replace content with a centered icon and text
13:53:43 <andreasn> that makes sense
13:53:45 <stefw> is that a standard patternfly/bootstrap pattern?
13:53:50 <stefw> i thought i saw it somewhere
13:53:53 <stefw> but then i lost it
13:54:09 <stefw> so i just used https://github.com/cockpit-project/jquery-amend/pulls as inspiration
13:54:42 <andreasn> yeah, I kind of recall something like that too
13:54:53 <stefw> anyway, if you see it i'd be interested
13:54:56 <stefw> if not, it's not hard to get working
13:55:37 <andreasn> I probably have it around somewhere
13:55:48 <andreasn> so that's those
13:56:04 <andreasn> what else is in that list?
13:56:08 <stefw> mvollmer, are you replacing https://github.com/cockpit-project/cockpit/pull/2231 ?
13:56:26 <stefw> andreasn, https://github.com/cockpit-project/cockpit/pull/2303
13:56:27 <mvollmer> i am not sure...
13:56:47 <mvollmer> stefw, I have to check it more carefully
13:56:50 <andreasn> stefw: yes, that one I'm revieweing today
13:56:51 <stefw> ok, no worries, just poking at the pull requests to see if they're alive
13:56:53 <stefw> like jelly fish on the beach
13:57:00 <andreasn> I am going to review today...
13:57:28 <stefw> i'm going to close this for now: https://github.com/cockpit-project/cockpit/pull/1917
13:58:44 <github> [cockpit] stefwalter closed pull request #1917: WIP: Simpler reauthorize (master...simpler-reauthorize) http://git.io/hUIZ
13:59:19 * stefw is done poking
13:59:35 <andreasn> I'm going to take another shot at https://github.com/cockpit-project/cockpit/pull/2290
13:59:41 <andreasn> it just fell apart last week
14:00:15 <andreasn> and I guess fridex will update https://github.com/cockpit-project/cockpit/pull/2307
14:00:30 <stefw> well it's very confusing
14:00:36 <stefw> we should probably summarize what still needs to be done
14:01:53 <andreasn> what it should be fixing is the behavior that occurs when you click the box, then click outside it
14:02:29 <andreasn> because it freaks out because it's suddently empty and gives an error
14:02:47 <andreasn> but what the patch ended up fixing was to just reset for everything
14:02:55 <andreasn> instead of ever giving errors
14:03:14 <stefw> right
14:03:17 <stefw> i guess explain that
14:03:20 <stefw> and it should help
14:03:31 <andreasn> yeah, I'll try to be more clear
14:04:19 <stefw> alright, all done?
14:04:26 <andreasn> anything else before we hang up the hat for the day?
14:04:29 <andreasn> I'm good
14:04:34 <stefw> oh one more about the branding
14:04:52 <stefw> two things.
14:05:02 <stefw> once you log in, there's usually a lot of padding on the left of the OS text, in the navbar
14:05:04 <stefw> it seems out of alignment
14:05:15 <stefw> and secondly, sometimes when the OS text is not available, the top level of the navbar just collapses
14:05:24 <stefw> and the user name overlaps the second layer of navbar
14:05:56 <stefw> hmmm, interesting the OS name is aligned with the server icon
14:05:57 <andreasn> oh
14:05:59 <stefw> and not with its button
14:06:05 <andreasn> I think I know some css for that
14:06:09 <stefw> so depending on whether you are focused on a server, or focused on the dashboard
14:06:14 <stefw> it'll look out of alignment either way
14:06:18 <stefw> or maybe not?
14:06:46 <andreasn> I'll try and fix it and make a pull request
14:06:51 <stefw> cool
14:07:36 <andreasn> #endmeeting