16:11:36 #startmeeting EPEL-8 Plan 16:11:36 Meeting started Fri Mar 1 16:11:36 2019 UTC. The chair is smooge. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:11:36 Useful Commands: #action #agreed #help #info #idea #link #topic. 16:11:36 Meeting started Fri Mar 1 16:11:36 2019 UTC. 16:11:36 This meeting is logged and archived in a public location. 16:11:36 The chair is smooge. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:11:36 Useful Commands: #action #agreed #halp #info #idea #link #topic. 16:11:37 The meeting name has been set to 'epel-8_plan' 16:12:00 #chair tdawson kanarip 16:12:00 Current chairs: kanarip smooge tdawson 16:12:00 Current chairs: kanarip smooge tdawson 16:12:12 hahahahah double the meetings in one time 16:12:45 #topic Review of packages 16:12:49 Whooo hoo :) 16:13:16 so currently I have come up with the following info 16:13:45 #info RHEL-7.6 has 2728 src RPMs in base 16:13:45 #info RHEL-8beta has 2679 src RPMs in base 16:13:45 #info There are 1939 src RPMs which overlap 16:14:35 #info EPEL-6 has 5930 src RPMs currently (many removed from peak) 16:14:35 #info EPEL-7 has 6891 src RPMs currently (some removed from peak) 16:14:35 #info RHEL-8beta has 419 EPEL-6 RPM names 16:14:35 #info RHEL-8beta has 399 EPEL-7 RPM names 16:15:12 #info 6492 EPEL-7 Packages are not in RHEL-8-beta 16:15:12 #info 5511 EPEL-6 Packages are not in RHEL-8-beta 16:15:12 #info 8149 EPEL-6/7 Packages are not in RHEL-8-beta 16:15:34 #info 7945 of that 8149 are still in Fedora 30 16:15:59 there are also some packages which are in RHEL-6 and RHEL-7 which aren't in RHEL-8. Those usually get requested to be put in EPEL 16:16:20 #info 1017 RHEL-6/7 packages not in RHEL-8-beta but in Fedora 30 16:16:34 So, I had similar numbers, so I agree with yours. But if a package was in EPEL6, but not EPEL7 ... do we really want to consider it for EPEL8? 16:16:46 a lot of people leap frog 16:17:08 we have things in EL7 which were EL5 16:17:09 True 16:17:16 so I am going to assume that 16:17:18 huh ... ok 16:17:45 When I did my EPEL8 build tests, I only used the 6492 EPEL7 packages not in RHEL8. 16:18:02 #topic Missing Build Requirements 16:18:29 tdawson, a bunch of the 1017 items have been ones which are needed to build those 6492 packages 16:18:54 That's understandable. 16:19:45 #info big culprits python2-sphinx (and deps), groff-perl, qt4-devel, maven tools, gcc-objc, gcc-gnat, compat-openssl10-devel, /usr/bin/python 16:20:21 Ugg ... sphinx :( 16:20:33 sphinx is horrible to rebuild 16:20:36 I tried 16:20:52 and will need help because it has a weird set of dependencies 16:20:57 I believe it is easiest to just change it to python3-sphinx 16:21:19 so I am currently recompiling everything from F30 packages 16:21:26 All it's doing it documentation ... you can use the pyhon3-sphinx in RHEL8 16:21:39 smooge: meaning you are getting your source rpm's from F30? 16:21:42 so I was hoping that these would be patched already but most aren't 16:21:46 yeah. 16:22:04 Ya ... it's because Fedora hasn't really pushed to get python2 out 16:22:10 I started with f29 and then pulled the equivs out of f30 because the python2 to python3 should have been flushed 16:22:18 I'm worried that when python2 goese away, it's going to be a Y2K scenario 16:22:39 goese -> goes 16:22:48 however it looks like a lot of stuff was correctly rebuilt because the rawhide buildroot had enough python2 still that it worked 16:23:01 it is definitely a Y2k 16:23:25 #topic Tdawson's rebuild 16:23:36 so I figured you could go over what you did and found 16:24:28 Well, there have been two major rebuilds that I did. The first sounds similar to what you are doign right now, except I wasn't trying to fix anything, just find numbers and packages. 16:24:38 same here 16:24:51 I was asked to see if F30 was better 16:24:51 I didn't know we were going to be talking about this, so I don't have my numbers in front of me. 16:25:15 no problem.. I was looking for more of a 'oh I did this and it is at this website and my god the patches' 16:25:24 and sorry for not telling you that before i did this 16:25:38 Well, the big takeaway from my rebuild, is what has gone into CRB ... those are/were the packages that were in RHEL8 that we needed in order to build as many of the EPEL8 packages as we could. 16:26:46 I *think* (and I'll double check in a few minutes) that I got 4000 packages to build, just with RHEL8, CRB, and as many F29 packages (F29 was rawhide at the time) 16:27:39 I did some blanket changes ... like sphinx, I believe I changed them all to python3-sphinx, and that solved 90% of the problems 16:27:41 thanks I really should have sent my agenda earlier and asked if you could give some info 16:28:18 And I believe there were one or two major packages that I had to tweek to get to build. 16:28:33 yeah I was hoping to make patching as small as possible since a lot of these packages are done by people who say 'I branched it and tried to build it.. if it works cool otherwise i will retire' 16:28:55 but it is clear we will need to a bunch 16:29:04 smooge: Yep, I bet that's going to be the common theme ... unless their work requires it. 16:29:16 on the other hand, the pythong fixes can be fed into Fedora for the f31 rebuild 16:29:25 even when their work requires it 16:29:32 The second major rebuild I did was KDE. And that, the first time, was much more invasive. 16:29:44 yeah I was looking at that 16:29:50 There are several packages that require QT4 16:30:00 i have yet to get that to compile 16:30:01 I spent alot of work pulling QT4 out of RHEL8. 16:30:40 Personally, I think it's time for it to retire. And I noticed, that in Rawhide, there are starting to be %if qt4 stuff going on. 16:30:42 and there are a lot of packages in epel which use it which I never thought of as kde 16:31:55 for my initial KDE rebuild, I did 3 things. 1) bootstrap .. there are alot of circular dependencies 2) remove any qt4, and drop packages that absolutely have to have it, and 3) whever possible, I took out python2 references. 16:32:06 when I looked for missing packages in the failed EPEL builds.. the qt4-devel was number 3 next to maven 16:32:12 Oh, and /usr/bin/python references, if possible 16:32:26 yeah a ton of those 16:33:19 I was looking at a couple failures and such which had them deep inside some script. Again lots of patches 16:33:32 I change /usr/bin/python to /usr/bin/python3 whenever possible, and most of them ran just fine. 16:34:44 smooge: Most of those scripts have just "python" ... which is fixable with that little script I put in the spec file. 16:35:49 smooge: But honestly, I'm surprised at how many of those there are ... I thought Fedora had done a pretty good sweep of pulling out /usr/bin/python from everything. 16:36:00 yeah. I expect one of the first requests will be to ship something like the python{2,3}-unversioned-command which can be installed to do the wrong thing 16:36:17 tdawson, I think a lot of it is that the upstreams haven't 16:36:48 and since many packagers just take what upstream does to not have a lot of extra patches.. it shows up 16:36:50 smooge: Oh, someting I recently hit, is that with F30, it now checks to see if there are any /usr/bin/python or /usr/bin/env python 16:37:07 is that mock or rpmbuild? 16:37:41 smooge: I had a package that was happily building on F29, went to build it on F30 (koji) and there were several scripts that had /usr/bin/env in them ... 16:37:46 smooge: koji 16:38:09 smooge: Hopefully that went in before the mass rebuild 16:39:24 ah ok 16:39:46 thanks again for the info 16:39:48 Here is an example of it failling my package, it's "mangling shebang"'s - https://kojipkgs.fedoraproject.org//work/tasks/9726/33109726/build.log 16:40:14 smooge: So hopefully, at least by F31 timeframe, we've got all those /usr/bin/python 's cleaned up. 16:40:38 that is a whole lot of mangling 16:40:44 #topic Where do we go next? 16:41:04 That's a good question 16:41:56 You/I/Others could just take all the packages that we know build, and rebuild them on EPEL8 ... but that defeats the "community" part of EPEL 16:42:15 And leaves You/I/Others with the burdon of supporting all those packages. 16:42:15 OK so I was going to look at what would be required to set up a /pub/alt/epel/ and the rules for koji to look at rhel-8 and then build against it 16:42:37 so for the RHEL-7 beta that is what we ended up doing 16:42:43 and I think the 6 beta also 16:42:44 Oh, have it as an 'alt' ... ok 16:43:04 dgilmore did all the builds for that target to show what was available and what wasn't 16:44:17 then when RHEL-7 started we put in official branches in CVS? (were we on git yet?) and people could branch their packages if they wanted to have a RHEL-8 one 16:44:39 That sounds like a good plan. 16:44:50 this led to a lot of missing packages but it also had us not have the 'I HAVE NO PLANS TO SUPPORT RHEL-X' like we did with EPEL-6 16:45:03 when we did it without asking 16:45:58 I was thinking that if we have a /pub/alt bare bones that people can fall back on for a while that will let them get working for things and allow packagers time since most of them aren't going to do anything until CentOS-8 is out 16:46:05 Ya ... I don't think doing the real builds without asking is a good idea ... definatly ask, or be prepared to support it. 16:46:42 smooge: Something like mock, fedpkg ... etc ... and their dependencies? 16:46:45 and maybe put together an epel-packagers sig which can support more packages like the java/python/etc 16:47:19 smooge: Ohh ... that's a very good idea 16:47:54 sorry I didn't mean bare-bones. I meant the beta stuff 16:48:12 smooge: Although ... maybe it should be specific ... cuz I'm not very good with those java packages ... though maybe that would be the whole idea. 16:48:28 specific as in epel-java-packagers 16:49:06 well I think it would need to be less specific.. we are probably going to only have 1 person who does java 16:49:23 but we can hum a tune until they get back from walk-about 16:49:24 smooge: Ah, OK. You give a try at those 8000 package, and KDE, for the /alt/ repo 16:49:55 smooge: :) OK ... ya. And sometime the java problems are within my skill range ... sometimes not. 16:50:07 did we call 8beta 7.99995 or anything like that 16:50:30 so we have most of the packages we can build in /pub/alt/epel/8.beta/Stuff (or 7.99995 ) 16:50:44 s/have/would have/ 16:51:00 smooge: ick :) ... I prefer 8.beta 16:51:16 okie dokie 8.beta 16:51:57 i don't have much else to say at the moment. 16:52:02 smooge: The debate on doing 7.999 for 8 alpha/beta/rc came up ... it got shot down really quick. 16:52:38 smooge: So ... what about putting up some epel pages with helpful hints on common problems that the packagers might hit 16:52:49 7.98.1 for alpha1 7.98.2 for alpha2.. that is old school rhl versioning 16:53:38 smooge: If we get some type of wiki page up, I could add my findings and solutions, and you could add yours as well. 16:53:49 yeah. I think we need to write that out first as I am trying to figure out how we are going to fix them. The .local/python patch you used won't fly with our rules 16:54:27 tdawson, I have put that down as a deliverable by Monday 16:54:57 smooge: Well, I had to write a 'find | sed" and "grep | sed" section for the package I showed earlier ... so there are different ways to fix the problem, some better than others. 16:55:51 smooge: Sounds like a good deliverable. 16:56:26 smooge: Are we still somewhat on track for the end of March that infrastructure will be ready to start these /alt/ rebuilds? 16:56:39 I expect we will need to have some sort of epel-bootstrap-8.beta rpm which we put in the compose root which has various gotchas in it 16:57:09 tdawson, I don't know. I have had some priorities shifted this week 16:57:26 I am hoping so.. but I won't know until the 15th 16:57:37 I will say we are Yellow/Orange 16:57:51 Well, better than red :) 16:58:07 this meeting and the plans above moved us from Orange-Red 16:58:28 :( 16:59:05 so we are improving 16:59:16 now I just need some more blueshifting 16:59:48 Make everything a nice fuzzy green :) 16:59:55 ok that is all I have. did you have anything you wanted from this meeting 17:00:39 nope, I think that's all I have. If you could get a wiki page up (or tell me where a good place to put one is) ... I can start shifting my notes to that. 17:02:14 yeah I am trying to think where is a good place. 17:03:04 I don't know if EPEL8_beta_known_problems_and_solutions is the right thing or /epel/EPEL8 or ... 17:03:18 which is why nothing ever gets documented 17:03:30 that other woodshed looks so much nicer than mine 17:03:43 *laughs* 17:03:44 ok I will come up with something and drop some starting data in it 17:03:55 will email you by end of business today 17:04:04 I think I can end this meeting 17:04:05 smooge: OK, thank you 17:04:08 #endmeeting