18:01:10 #startmeeting EPEL (2019-05-15) 18:01:10 Meeting started Wed May 15 18:01:10 2019 UTC. 18:01:10 This meeting is logged and archived in a public location. 18:01:10 The chair is smooge. Information about MeetBot at http://wiki.debian.org/MeetBot. 18:01:10 Useful Commands: #action #agreed #halp #info #idea #link #topic. 18:01:10 The meeting name has been set to 'epel_(2019-05-15)' 18:01:10 #meetingname epel 18:01:10 The meeting name has been set to 'epel' 18:01:10 #topic Chair and Introductions 18:01:10 #chair bstinson Evolution nirik smooge pgreco tdawson 18:01:10 Current chairs: Evolution bstinson nirik pgreco smooge tdawson 18:01:25 morning 18:01:27 hello 18:01:32 Hello 18:01:37 * jsmith lurks 18:01:39 sorry 2 phone calls and trying to get something opened all at :58 to the hour 18:01:40 hello 18:02:21 OK so I have a google doc I was trying to make into a etherpad doc or something for people to read 18:02:22 * bstinson waves 18:02:58 #topic Agenda 18:02:58 #info https://etherpad.gnome.org/p/EPEL-8# 18:02:58 #info CentOS-8 18:03:01 \o 18:03:22 #topic EPEL-8 or why Smooge can't read 18:03:33 #info https://etherpad.gnome.org/p/EPEL-8 18:03:57 this is temp doc until I can get wiki or something not to completely explode 18:04:38 * nirik can try and read and feedback... but it's a bit long to just read right now 18:05:36 nirik, this is the key for me "Currently EPEL-7 has 6915 src.rpms of which ~4000 require libraries that are inside of modules and can not be built without them" 18:06:18 a slightly longer view is https://smooge.fedorapeople.org/EPEL-8/ 18:06:52 sure, but there's of course no assurance that people want all those src.rpms build for epel8... but it's an interesting metric in any case. 18:06:58 nirik, you have write/feedback rights on the original google doc 18:07:17 I'd rather us not *own* all the epel packages in 8 as well. 18:07:49 yes, -1000000 to us mass branching and maintaing things. This is a community effort. ;) 18:07:59 * nirik afk for a min, fedex at the door. 18:09:05 the main issue from my tests is that there is various stuff which was in RHEL-6/7 which is not in EL-8 but is needed for many leaf packages to build 18:09:33 do we have any statistic of packagers that use the same spec for all branches? 18:09:42 not really 18:10:08 because, if we could somehow "strongly suggest" that.... 18:10:14 I expect it is a 2 peak average. There is a group who do it and there is a group who never do it 18:10:19 So, if we don't do *something*, EPEL starts with 0 packages, and nobody knows which ones will build. If we do something like do a mass branch to rhel8-stage, or something temporary, we could see which ones build, which ones don't, and have a staging repo. 18:10:24 and they each hate each other 18:11:14 tdawson: well, it's gonna start with a few packages... epel-release, etc 18:11:41 smooge, my main example is rust 18:11:45 .whoowns rust 18:11:50 pgreco: owner: jistone; admin: ignatenkobrain - Fedora: @rust-sig; Fedora EPEL: @rust-sig 18:11:52 I have been in the 'use one spec for all branches' camp, but I think I am moving toward not doing that anymore. 18:11:53 * orionp uses common specs unless impossible, FWIW 18:12:28 rust is pretty special... even in fedora land. 18:12:31 the admin who maintains rust has said they have 0 interest in doing anything with EPEL in the past. It would require some other group to have it in EPEL 18:12:56 but it is available there, and he en accepted my patch for epel/armhfp 18:13:22 the base rust package or ? 18:13:59 yes, the base package, https://src.fedoraproject.org/rpms/rust/ 18:14:15 the base rust package is in RHEL-8 as a module 18:14:37 yes, I meant it as an example of a package that uses the same spec for all branches 18:14:55 well I think they are moving to it being only a module 18:15:01 .hello2 18:15:02 ignatenkobrain: ignatenkobrain 'Igor Gnatenko' 18:15:05 which, in my opinion would be the way to make epel easier 18:15:42 do you need any answers about rust? 18:15:57 well, it means spec files are more complex all around and full of conditionals... 18:16:11 ignatenkobrain, mainly I 'spoke for you' when you said you had 0 interest in rust being in EPEL in the past 18:16:12 ignatenkobrain, just using its spec as an example 18:17:12 which I took from various comments in the past where you don't have time/interest in supporting it in EPEL 18:17:29 smooge: yeah, we can't build any crates in EPEL7, but it might be not bad idea to be able to use modularity to deliver rust-based apps in EPEL8 18:17:45 I don't understand where this falls into the discussion of whether we do some type of staging repo for EPEL8. of the ~7000 packages about ~2000 (possibly 3000) don't need any change from the F30 spec file. 18:17:52 yes, I don't have any interest in supporting epel, but if it will be just yet another platform in modulemd, I will not mind 18:18:39 tdawson: people will then use those things. They will try and report bugs and ask for new updates. Are we doing that? 18:19:37 For EPEL8 staging ... no ... someone shouldn't be able to file a bug on EPEL8 staging packages. 18:19:41 smooge: but actually jistone is primary person to support compiler so he does that for all fedora/epel. I'm pretty sure he will be happy to change conditionals as needed 18:20:15 tdawson: do you want to take epel7 packages and rebuild them or fedora ones ? 18:20:22 ignatenkobrain, RHEL-8 comes with a rust so this would be additional modules which add things which relied on that main module 18:20:29 sure, but they will against other versions... 18:20:43 ignatenkobrain: I would take fedora 30 packages, the epel7 ones are too old. 18:21:10 as a means to show people what builds I suppose it's useful, but as always maintainers should make sure the things they need are there and maintained before they branch anyhow... 18:21:28 tdawson: then just do `rpmspec --undefine fedora --define 'rhel 8'` for all specs and compare with fedora results 18:21:30 But, in the end, if we do a staging, and we dont' support the ones in staging, it will be up to the developer who takes the packages to determine if they want an older epel7 package or a newer fedora30 package. 18:21:39 in case of discrepancy, check more carefully 18:22:09 ie, as a Xfce co-maintainer, I will likely want to build the Xfce stack... but even if I see them in a staging build I still need to make sure all the things I depend on are also branched and maintained... so I will be building it myself to figure that out and the staging thing isn't very useful to me. Perhaps I am missing something tho 18:22:59 nirik: You have a point there. 18:23:05 the maintainer may want all kinds of differnet things, we have no way of knowing. 18:23:16 the issue is whether you want to know if you have a chance of building it or not 18:24:02 personally, thats not too valuable to me, as I could just build myself or in copr or whatever... but I suppose it might be to some folks 18:24:28 but I don't think it deserves a ton of our attention, we should just get things to where people can branch and build. 18:25:02 ok in that case we can remove a bunch of stuff 18:25:14 I think one other thing it might do, if we do a EPEL8-staging, is help us test everything out, our RHEL8 repo's and default modules, that sort of thing. 18:25:54 ignatenkobrain++ thank you for answering questions in our meeting 18:25:55 smooge: Karma for ignatenkobrain changed to 2 (for the current release cycle): https://badges.fedoraproject.org/tags/cookie/any 18:25:55 But would we have to rebuilt everything to find that out *shrugs* maybe not. 18:26:34 the main reason for doing anything in staging is to make sure we can build anything before we put it in production and find we can't build anything 18:27:37 So, what if we take a big handfull of packages and work through those. Maybe something that each of us know we will put in EPEL8. 18:28:10 the reason I want to test branching is because we will need to do branches and have it go over to the CentOS side with repoSpanner. 18:29:09 sure. +1 to testing in stg 18:29:19 I just don't think we should do 7000 builds... 18:29:40 My list of KDE packages is about 350 packages ... and covers a wide variety of languages. Would that be a good sample size. 18:29:51 yeah, that might be good. 18:29:56 ok I am trying to find where I said we were doing 7000 builds. 18:29:58 Xfce is a lot smaller... 18:30:13 smooge: Right under "Open Questions" 18:30:28 but I think if we can build a set of things we can then quickly move to prod... 18:31:12 OK there is some confusion then. I don't mean we will do that. I am saying that out of our current set of packages, we know only N will build 18:31:24 Oh, ok. 18:31:27 but I am clearly communicating something else 18:33:07 smooge: Ohh ... I think we didn't move down to EPEL-8 Proof of Concept 18:33:42 There you say "Is not a complete rebuild of EPEL-7 but a subset of 1200 packages known to mostly compile from previous builds" 18:33:52 I'm just saying we shouldn't do a ton of builds in stg or publish them, we should only do enough so we are reasonably sure prod will work and then move to prod to actually let people start doing builds. 18:34:13 And that this is just for tests, that the branches will not show up anywhere. 18:34:29 so let us go with a list of packages we want to test: 18:35:31 tdawson: they would be in our stg src.stg.fedoraproject.org... but otherwise yeah 18:35:33 I just thought of something, that my KDE list will not be a good list. Except for a handful, they all have to be build serially, in chain-build fashion ... so I withdraw that we shoudl use them. 18:35:43 XFCE and KDE are both unlikely to build with RHEL-8.0 due to some -devel packages not available 18:36:16 they would require stuff from CentOS buildroot 18:36:26 I can probibly dig up a number of one-off type packages that I maintain that would be easy to build... apg, etc 18:37:49 or once we have a very small base we could ask interested folks to try builds? 18:38:07 apg will compile. ansible will not 18:38:23 most of the items rsc supports will compile 18:38:47 huh, what does ansible need? thats another good one to test... 18:39:03 but I can work thru that myself... don't need meeting time on it 18:39:14 I believe ansible will be part of RHEL8 ... as for what it needs, I'd need to dig it up, but it definatly needs things not in RHEL8 18:39:30 https://paste.fedoraproject.org/paste/LeTna-6iz9aRjsDG6A2LEQ 18:39:41 actually it will be an addon channel 18:40:10 Ya ... not regular RHEL8 18:40:33 nirik, and those end up needing more packages.. and a couple of them loop. 18:40:43 but can be fixed with F28 18:40:47 anyway 18:41:18 many of those are needed for tests only... but yeah. 18:42:37 please read and send comments on the document... the https://smooge.fedorapeople.org/EPEL-8/ will get updates until I figure out how to make that work in wiki 18:42:55 the etherpad will go away probably when my browser reboots 18:43:57 my main goal this week is to work on grobisplitter to make it work with the published trees vs dvd. 18:44:58 anything else on the doc? [Beyond make it clear we are only going to test a dozen or so packages and then move to production] 18:45:28 smooge: currently, are we on Proof of Concept, or EPEL-8.0 Beta ? 18:45:29 once you have grobisplitter going I can add the koji stuff in stg 18:46:03 I'm looking at the document not on etherpad - https://smooge.fedorapeople.org/EPEL-8/ 18:46:11 We are Proof of Concept still 18:46:39 OK 18:47:08 though I think POC and Beta need a rewrite from this meeting 18:47:24 So "a subset of 1200 packages" could mean as small as a dozen packages ... ok. 18:48:33 I will reword that "Of the 1200 known compilable packages, we will focus on a smaller list to be published at <>" 18:48:47 *nods* 18:49:39 that is all I have.. anything else? 18:49:58 not from me, not very helpful today :( 18:50:35 pgreco, I will send you a copy for you to tear apart at your leisure 18:50:49 I think we covered everything I had questions about. 18:50:57 #topic Free Food 18:51:14 OK if there isn't anything else on this I open to the Free Lunch segemnt 18:54:10 and if there isn't anything else I am going to close this as I need coffee 18:54:12 #endmeeting