14:15:03 <haraldh> #startmeeting Fedora Base Design Working Group (2015-07-27)
14:15:03 <zodbot> Meeting started Mon Jul 27 14:15:03 2015 UTC.  The chair is haraldh. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:15:03 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
14:15:07 <haraldh> #meetingname  Fedora Base Design Working Group
14:15:07 <zodbot> The meeting name has been set to 'fedora_base_design_working_group'
14:15:12 <haraldh> #chair haraldh msekleta jreznik dgilmore vpavlin masta dwalsh lnykryn
14:15:12 <zodbot> Current chairs: dgilmore dwalsh haraldh jreznik lnykryn masta msekleta vpavlin
14:15:17 * masta is here
14:15:22 <haraldh> ping msekleta jreznik dgilmore vpavlin masta dwalsh lnykryn
14:15:29 <haraldh> hi masta
14:15:37 <masta> hey haraldh =)
14:18:25 * jreznik is here, looking outside - that's nice heavy rain :)
14:18:35 <haraldh> hi jreznik
14:18:40 <jreznik> hi haraldh
14:22:09 <haraldh> #topic define the minimal install
14:22:45 <haraldh> IIRC, dgilmore wanted to start a thread on fedora-devel about defining the core comps group
14:22:59 <haraldh> haven't seen one, did you?
14:24:31 <masta> not yet
14:24:53 <haraldh> jreznik, do you have something to add to the discussion of last week to this topic?
14:25:19 <jreznik> nope
14:25:23 <masta> I got curious about what the core group really entails, so I wrote a script to show the full graph... but it's a bash script. http://fpaste.org/248562/80070501/
14:26:33 <masta> I noticed DNF has a plugin to output DOT files for graphiz, but so far I've not noticed that in the fedora repos.
14:27:59 <haraldh> nice shell script.. let me try it out
14:28:16 <masta> anyways, seems like being conservative with minimal install might require we look at he full dep graph of packages down stream of the top level comps group....
14:28:27 <masta> hope that makes sense
14:30:31 <masta> haraldh: it's slow, bash script... might take 3 minutes or so
14:37:21 * jreznik is looking
14:40:50 <masta> I think the new feature of weak dependencies might help a bit here.
14:41:34 <masta> certainly the core and base comps groups are something we should review for weak deps, but I think for now the compose tools will not honor those things.
14:42:13 <haraldh> here is a faster shell script :)
14:42:16 <haraldh> https://paste.fedoraproject.org/248577/80081171/
14:42:35 <haraldh> repoquery has a "--recursive" option :)
14:43:04 <masta> oh cool, haraldh!
14:44:10 <jreznik> there was discussion about weak deps and how it should be honored in compose
14:44:15 <jreznik> on devel list
14:45:57 <dgilmore> hi all here now
14:46:03 <masta> jreznik: yeah, I think (more like hope) for now we would choose to not honor weak deps in the compose, BUT we would begin to curate the package set of the most minimal comps groups to accommodate them for future.
14:46:17 <jreznik> hi dgilmore
14:46:23 <masta> hey dgilmore
14:46:44 <haraldh> updated script with sort -u https://paste.fedoraproject.org/248577/80081171/
14:47:08 <haraldh> 230 packages
14:47:22 <masta> haraldh++
14:47:42 <haraldh> package list:  http://paste.fedoraproject.org/248580/80084481
14:48:12 <haraldh> bind99-license-0:9.9.7-5.fc23.noarch wow
14:48:24 <haraldh> a package for a license file
14:48:43 <masta> haraldh: huh... my original script might have been in error; It had over 300 findings.
14:48:43 <dgilmore> haraldh: the bind packaging is odd
14:48:58 * jreznik has to swap to phone, will be right back
14:49:08 <haraldh> masta, we need to double check, which script is wrong
14:49:24 <haraldh> masta, a dnf install into a chroot would give valid numbers
14:49:41 <haraldh> dnf groupinstall
14:51:11 <dgilmore> dnf --installroot /tmp/test install @core
14:51:12 <dgilmore> error: cannot open Packages database in /tmp/test/var/lib/rpm
14:51:22 <masta> haraldh: indeed, that is what I was trying to avoid, but it's the ultimate test.
14:51:32 <haraldh> dgilmore, well
14:51:35 <dgilmore> how fun
14:51:39 <haraldh> dgilmore, you have to mkdir manually
14:51:48 <haraldh> BUT
14:51:55 <dgilmore> haraldh: that is really dumb
14:51:58 <haraldh> yes
14:51:58 <dgilmore> but okay
14:52:51 <masta> meh... I think dnf is improving, but for now has much to improve ;-)
14:52:55 <dgilmore> dnf --installroot /tmp/test groupinstall core --disablerepo * --enablerepo fedora
14:52:58 <dgilmore> Error: Unknown repo: 'anaconda-ks.cfg'
14:53:03 <dgilmore> I really dislike dnf most days
14:54:59 <haraldh> # dnf groupinstall --releasever=23  --disablerepo=* --enablerepo=fedora   --installroot /var/tmp/test core
14:55:04 <haraldh> Install  249 Packages
14:55:19 <haraldh> Is this ok [y/N]:
14:55:39 <haraldh> so, neither 230 package nor 300 packages..
14:55:40 <haraldh> hmm
14:56:02 <dgilmore> that is always fun
14:56:21 <dgilmore> I get 249 also for that
14:56:50 <masta> haraldh: interesting... this is the kind of thing I can relentlessly obsess upon... unfortunately   =(
14:57:13 <masta> dgilmore: well I guess 249 is *the* correct number
14:57:42 <dgilmore> I wonder if yum is the same
14:57:43 <masta> I wonder if there is treatment for default items, or arch=
14:57:56 <masta> dgilmore: oh that is a great idea for comparison.
14:58:38 <dgilmore> Install  43 Packages (+201 Dependent packages)
14:58:45 <dgilmore> 244
14:58:57 <dgilmore> dnf picks 5 extra packages
14:59:51 <masta> that is approximately the number of default packages. We should try to diff the two lists.
15:00:19 <masta> but it could be dnf, doign what hawkey does
15:00:39 <masta> err... libsolve, or whatever it's based on
15:02:02 <haraldh> ah, oh.. called repoquery without --releasever=
15:02:21 <haraldh> fixed http://paste.fedoraproject.org/248590/43800933
15:03:15 <haraldh> anyway, here is the diff
15:03:28 <haraldh> https://paste.fedoraproject.org/248591/80093681/
15:04:07 <haraldh> hmm, seems like reqoquery does not honor --releasever=
15:04:53 <haraldh> so, let me recheck with rawhide
15:09:18 <dgilmore> http://paste.fedoraproject.org/248597/14380097/
15:09:38 <dgilmore> that is the difference between yum and dnf
15:09:57 <haraldh> repoquery vs dnf for rawhide: https://paste.fedoraproject.org/248600/14380097/
15:10:36 <haraldh> not correct
15:11:39 <dgilmore> haraldh: I got my diff by doing the installroot, chrooting in and doing rpm -qa |sort -u >file and diffing the files
15:12:02 <masta> slightly demoralizing, the lack of deterministic results. =(
15:12:17 <masta> dgilmore: I think that is a pretty good approach
15:12:33 * dgilmore has another meeting in 18 minutes
15:12:42 * masta too
15:13:26 <haraldh> correct diff https://paste.fedoraproject.org/248604/09993143/
15:13:32 <haraldh> repoquery vs dnf
15:13:47 <haraldh> this is strange
15:14:05 <masta> haraldh: looks like we have homework for the week ahead of next meeting.
15:14:17 <haraldh> maybe "Wants:" and "Recommends:"
15:15:42 <masta> haraldh: I think defaults too
15:16:01 <haraldh> right
15:16:23 <masta> firewalld, uboot, etc...
15:16:53 <masta> those get magically pulled in anyways, depending on the kickstart
15:17:03 <dgilmore> masta: not really
15:17:20 <dgilmore> well grubby on arm requires uboot-tools
15:17:39 <masta> for example anaconda will install authconfig, if the kickstart has auth lines. same for firewalld
15:18:09 <dgilmore> yes
15:18:21 <dgilmore> but not all installs are done by anaconda
15:18:25 <dgilmore> at least today
15:18:49 <masta> dgilmore: hrm... yeah your're right, but not sure I noticed a line of deps of grubby--> uboot. I'll take another look to make sure
15:22:52 <haraldh> ok, then let's close this meeting and as a homework look at the package list more closely.
15:23:16 <haraldh> #endmeeting