18:01:13 <smooge> #startmeeting EPEL (2019-01-09)
18:01:13 <zodbot> Meeting started Wed Jan  9 18:01:13 2019 UTC.
18:01:13 <zodbot> This meeting is logged and archived in a public location.
18:01:13 <zodbot> The chair is smooge. Information about MeetBot at http://wiki.debian.org/MeetBot.
18:01:13 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
18:01:13 <zodbot> The meeting name has been set to 'epel_(2019-01-09)'
18:01:13 <smooge> #meetingname epel
18:01:13 <smooge> #topic Chair and Introductions
18:01:13 <zodbot> The meeting name has been set to 'epel'
18:01:13 <smooge> #chair avij bstinson Evolution nirik smooge pgreco
18:01:13 <zodbot> Current chairs: Evolution avij bstinson nirik pgreco smooge
18:01:19 <nirik> morning
18:01:24 <Evolution> I made it for once.
18:01:50 <tdawson> Hi
18:01:57 <smooge> ha
18:02:01 <smooge> missed that
18:02:03 <pgreco> hello
18:02:05 <smooge> sorry for the other ping
18:02:36 <smooge> now the scary part.. Evolution is here and bstinson is not. No one has seen them in the same place for a long time
18:03:15 * Evolution slides the mask out of sight
18:03:47 <smooge> #topic Agenda
18:03:47 <smooge> #info Alternative Architectures
18:03:47 <smooge> #info Project Documentation
18:03:47 <smooge> #info EPEL-8 planning
18:06:38 <smooge> #topic Alternative Architectures
18:08:29 <smooge> pgreco, I am going to roll the arm/i386 under your bailiwick .
18:08:45 <smooge> I figure not much has happened due to well.. no one being around
18:08:53 <smooge> and no path forward
18:08:54 <pgreco> well, not much from me here
18:09:02 <pgreco> everything is up to date in armhfp
18:09:03 <bstinson> hey, i'm around
18:09:12 <bstinson> sorry, i wasn't paying attention to IRC
18:09:17 <avij> hi
18:09:26 <smooge> hello avij bstinson
18:09:42 <pgreco> and I'm planning on restarting the i386 version once everything comes back to normal
18:09:54 <pgreco> after new year I mean
18:11:26 <smooge> ok cool and understood
18:11:43 <smooge> #topic Documentation
18:11:43 <smooge> #info smooge is back from PTO and has round-tuits
18:11:43 <smooge> #info https://en.wiktionary.org/wiki/round_tuit
18:12:20 <smooge> So I will be prioritizing and getting a list of docs to rewrite from wiki to pagure/docs in the next week. Will have the list for people to look at next week.
18:12:53 <smooge> #topic EPEL-8 planning
18:12:53 <smooge> #info smooge is back from PTO and has round-tuits
18:12:53 <nirik> we may want to just merge into the normal docs at this point... not use our own pagure one.
18:13:40 <smooge> that is true. I was doing it there because we didn't have docs at that point and we were too 'different' from what they were doing
18:14:06 <nirik> probibly under here: https://docs.fedoraproject.org/en-US/engineering/
18:14:50 <nirik> anyhow, that doesn't matter, we still need content no matter what
18:15:16 <smooge> yep. I was going to get the content done and then see where they wanted it
18:15:36 <smooge> the pagure would be short erm
18:16:06 <smooge> on the other item.. I am going to start a document for the EPEL-8 planning and have people start dropping what needs to be done.
18:16:16 <smooge> I can do so in google docs or gobby or ...
18:17:21 <nirik> whatever works... or wiki page.
18:18:14 <smooge> I was wanting something people could collaborate on.. not for where ideas go to die :)
18:18:28 <nirik> I am exploring if we can enable things for modules somewhat easy... but I need to find some mbs folks to tell me whats involved there.
18:18:28 <smooge> i'll be here all week folks
18:19:00 <tdawson> :)
18:19:01 <Evolution> lets not do wiki, lets use the new tooling/markdown please.
18:20:34 <Evolution> speaking of epel8 and modularity
18:20:36 <Evolution> given the topic.
18:20:40 <Evolution> https://pagure.io/fesco/issue/2003
18:20:48 <Evolution> we need to lean on this a little.
18:21:03 <nirik> yeah.
18:21:11 <Evolution> we need this, or otherwise similar functionality
18:21:32 <nirik> I am hoping we can at least get modules working first, but we do need that for non modular rpms
18:22:01 <smooge> is that all non-modular rpms or just  non-modular rpms which require modules
18:22:06 <mizdebsk> i don't think that ursa-major is necessary for epel8 any more than it is for fedora
18:23:02 <smooge> mizdebsk, so the mbs/ursa people say it is, you say it isnt
18:23:04 <nirik> mizdebsk: well, contyk and others I have talked to disagree...
18:23:36 <nirik> if we just enable the appstream repo it's like enabling all modules at once... and they do not promise parallel installability
18:24:13 <nirik> smooge: well, without modules there's no php or python or any of the other stacks... so rpms we could build would be pretty limitied
18:24:38 <mizdebsk> nirik, ursa-major doesn't solve this problem; for it to work for this you would need to import rhel 8 module into fedora koji, not use external repos
18:25:15 <nirik> well, it solves the problem of which modules are availablle in the buildroot... but yeah, thats another problem we would have to import them somehow
18:25:16 <Evolution> mizdebsk: can you elaborate on that a bit please? This is something I'm going to have to tell management
18:25:37 <smooge> ugh.. that was always a no-no-never-allowed
18:26:02 * contyk looks
18:26:16 <contyk> yes, we would need RHEL modules in our MBS
18:26:21 <mizdebsk> i have a feeling that there is some widespread misunderstanding about what ursa-major really is
18:26:35 <mizdebsk> ursa-major is a very simple tool that just manages koji tag inheritance
18:26:49 <mizdebsk> it can enable individual modules, but then these modules need to have their own tags in fedora koji
18:27:35 <contyk> I agree UM seems to cause a lot of confusion
18:27:36 <mizdebsk> which means either importing rhel 8 builds into fedora koji, or splitting rhel 8 repos into multiple repos (one per module) and adding these repos to different tags as external repos
18:27:44 <Evolution> okay, in the vein of "explain it like I'm 5", what's needed for EPEL to build modules and packages against rhel8 modules/appstream ?
18:27:45 <contyk> I suspect all of us will be at devconf?
18:27:54 * mizdebsk will be there
18:27:58 <Evolution> yeah.
18:28:09 <nirik> yep
18:28:36 <contyk> to build stuff properly on top of RHEL 8, you RHEL modules represented as tags and recorded in MBS (could be very simple records)
18:28:48 <contyk> and the actual tag representing non-modular RHEL shouldn't include any modular packages
18:29:01 <smooge> so it sounds like RHEL-8 needs to be imported into koji versus an external repo which means those packages are now downloadable to anyone
18:29:11 <contyk> then you use UM to add some modules to that tag's inheritance
18:29:51 <mizdebsk> smooge, not necessarily downloadable; they could be blocked like cisco coded
18:30:09 <nirik> they could likely be blocked the same way rhel6/7 are now.
18:30:52 <smooge> nirik, but I didn't think rhel-6/7 were imported . At least when I said that once dgilmore said I was wrong (or using terms wrongly)
18:30:53 <mizdebsk> contyk, but you don't need even UM for this; assuming rhel doesn't change very often, releng could simply adjust tag inheritance manually when a new rhel release comes out
18:31:23 <nirik> smooge: true, they are not.
18:31:31 <contyk> mizdebsk: sort of true
18:31:34 <mizdebsk> doing this may require less work than maintaining UM and its config
18:31:43 <contyk> mizdebsk: MBS knows how to work with UM to deal with conflicting streams
18:31:59 <contyk> mizdebsk: without UM we would need to invent another mechanism or have some fake UM configs around
18:32:32 <contyk> mizdebsk: quick example -- platform includes foo:a, I'm building bar:1 that BRs foo:b
18:32:49 <contyk> mizdebsk: MBS maska foo:a's artifacts so that it can pull foo:b in and let you build against it
18:33:03 <contyk> mizdebsk: it knows that foo:a is in platform because it reads UM's configs
18:33:25 <contyk> and it masks the artifacts via conflicts in module-build-macros
18:33:51 <Evolution> contyk: that brings up a question of how epel would get platform updates in UM's config if a package/module gets added/removed
18:34:13 <contyk> Evolution: yes, it's a good question -- even how do you get the initial configuration?
18:34:42 * nirik ponders the 'split appstream into different external repos' idea... I guess that might be tricky, but could let us keep using external repos
18:34:44 <mizdebsk> it depends whether epel8 wants to build against default module streams or arbitrary module streams
18:35:15 <mizdebsk> if just default (as proposed in fesco ticket) then rhel 8 module defaults can be translated into ursa-major config
18:35:22 <nirik> and it's hard to answer that....
18:35:33 <mizdebsk> if arbitrary then epsco decides which?
18:35:48 <smooge> the proposal is that EPEL doesn't just build against default but against all arbitrary and possibly ship competing module streams
18:36:21 <smooge> s/is/was/
18:36:41 <Evolution> yes, one of the requests was that EPEL carry "supplemental" modules that compete/conflict with base/appstream
18:36:49 <nirik> well, for modular content it just tells what it builds against
18:36:53 <Evolution> to enable features that may be disabled in rhel.
18:37:02 <nirik> it's the non modular rpms that are the tricky part
18:37:12 * contyk nods
18:37:27 <contyk> so as mizdebsk says, you could just enable all defaults
18:37:43 <contyk> at minimum you need to enable modules that provide runtime deps for the non-modular ones
18:37:47 <nirik> if we can make modular things as easy as a package to create/maintain, we could just say 'all epel8 content are modules'
18:37:57 <contyk> because yes, non-modular packages can depend on packages provided by default streams
18:38:11 <nirik> right
18:38:54 <contyk> "all epel8 content are modules", heh
18:39:07 <contyk> that's probably not the way to go
18:39:20 <nirik> well, it would solve the non modular package issues. ;) but yes, it's a pretty heavy hammer.
18:39:54 <smooge> I think that if a package requires a non default module.. it needs to be a module itself
18:40:01 <contyk> yes
18:40:21 <contyk> if it requires something provided by a non-default stream, that is
18:40:29 <tdawson> All epel8 is in modules, would solve the long term support issues.
18:40:31 <contyk> as packages don't depend on modules
18:40:45 <nirik> tdawson: right, you could have EOL's for each module
18:41:33 <Evolution> that is largely in-line with what mattdm requested
18:41:45 <pgreco> "You don't build packages, you build modules", something that got stuck in my head after watching a video about modules
18:42:03 <pgreco> also, I think it is the only thing I actually understood
18:42:14 <contyk> :)
18:42:26 <avij> pgreco: you know this stuff much better than I do
18:42:30 <smooge> contyk, sorry the terms for streams, modules, etc have gotten very muddled in my head.
18:42:31 <contyk> what I would like to see if module stream expansion across Fedora and EPEL
18:42:40 <contyk> if the two are vastly different, no one will do that
18:43:08 <mizdebsk> contyk, i'd like to see that too
18:43:12 <nirik> yep.
18:43:20 <mizdebsk> (a single module build that works on fedora and epel8)
18:43:28 <contyk> so for instance I have this very important module, dwm; it depends on "any platform", which means that it would be produced for EPEL automatically
18:43:43 <nirik> ideally we could get mbs working for epel8 and then all the existing fedora modules (that wanted to) could just add epel8 to their targets and bam
18:43:58 <contyk> it doesn't have to be a single build, could be different binaries, but from the maintainer's perspective it could seem as a "single" build
18:44:30 <contyk> you don't need that
18:44:46 <contyk> if you use the same MBS we already have, you just define platform:el8 and point it to the right base tag
18:44:50 <contyk> and that's it
18:45:36 <nirik> sure, but we have to set that up...
18:45:59 <smooge> and we need to know how to do that without 'breaking' koji or leaking rhel
18:46:42 <Evolution> we can (in theory) bring in someone from the modularity team and/or koji to help with this
18:47:03 * contyk points at tdawson
18:47:28 <Evolution> tdawson: thank you for volunteering
18:47:31 * tdawson waves
18:48:09 <nirik> being volenteered. :)
18:48:14 <Evolution> nirik | smooge can you work with tdawson on getting what we can set up on our side?
18:48:29 <Evolution> I know it won't be complete, but as far as we can move, we should.
18:48:55 <Evolution> tdawson: use *all* your fingers when you wave, please.
18:48:57 <Evolution> :-P
18:49:05 <nirik> sure, happy to...
18:49:05 <tdawson> *laughs*
18:49:16 <nirik> and we can try and move ursa-major forward in fesco.
18:49:31 <avij> it's lovely that we have volunteers such as tdawson who are willing to help us with this
18:49:35 <smooge> as an idea.. I would like to have any epel-8beta show up in /pub/alt versus /pub/epel
18:49:56 <contyk> I'll try to meet with lsedlar tomorrow and ask him for ideas to make the buildroot compose cheap
18:50:03 <contyk> cheap to make, cheap to mirror
18:50:05 <smooge> people tend to think whatever is in /pub/epel is STONE SET.. and we are probably going to break down and cry a couple of times
18:50:30 <nirik> smooge: +1
18:50:49 <smooge> It would allow us to experiment with say even making all of EPEL-7 modules :P
18:51:13 <contyk> you'd need updated dnf in epel7
18:51:22 <contyk> but it's possible
18:51:57 <contyk> I think there was an SCL with updated yum
18:51:59 <smooge> I am expecting that will be a Register article
18:52:17 <smooge> EPEL moving to ALL MODULES, USERS HAVE TO LUMP IT
18:52:29 <smooge> ok anyway.. it is 13:52.
18:52:36 <avij> contyk: that SCL yum4 thing got nuked, it's no longer maintained
18:52:39 <smooge> anything else on this
18:53:12 <contyk> :/
18:53:20 <Evolution> contyk: more updated than the dnf that dropped in 7.6?
18:53:36 <avij> contyk: but it's available elsewhere, CentOS Extras has it (no idea which channel on RHEL)
18:53:38 <Evolution> that claims modular support, but it *may* just be able to read the data without erroring
18:53:41 <Evolution> not install.
18:54:14 <contyk> Evolution: I don't actually know what's included in 7.6
18:54:33 <contyk> but things are still changing slightly, so maybe 7.7 :)
18:54:50 <avij> "dnf update" does work on my test CentOS 7.6 box. dunno about modules.
18:55:56 <contyk> add the rawhide repo and try "dnf module list"
18:56:06 <tdawson> On real RHEL 7.6 it's in rhel-7-server-extras-rpms
18:56:29 <tdawson> dnf-2.7.5-19.el7_6
18:57:04 <smooge> #topic Open Floor
18:57:16 <smooge> ok I am going to close this out and send people home in 2 minutes
18:57:32 <avij> "dnf module list" says "Nothing to show" at the moment. no rawhide repo in use, unsure if that'd be wise to mix in.
18:57:40 <smooge> thank you for coming.. and I hope you had a good week. For those who will be in Brno at the end of the month we will show
18:57:41 <avij> good plan
18:58:11 <smooge> s/show/see each other soon/
18:58:25 <smooge> #endmeeting