15:23:18 #startmeeting EPEL-python36-flag-day 15:23:18 Meeting started Thu Mar 7 15:23:18 2019 UTC. 15:23:18 This meeting is logged and archived in a public location. 15:23:18 The chair is smooge. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:23:18 Useful Commands: #action #agreed #halp #info #idea #link #topic. 15:23:18 The meeting name has been set to 'epel-python36-flag-day' 15:23:19 Meeting started Thu Mar 7 15:23:18 2019 UTC. The chair is smooge. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:23:19 Useful Commands: #action #agreed #help #info #idea #link #topic. 15:23:31 since we have a lot of stuff being said here which needs to be captured 15:23:43 #topic General Meeting On What Happens 15:24:08 Replay of previous conversation 15:24:11 [2019-03-07-10:08] I just wanted to get step 1 (merge the pull request to python-rpm-macros, and built it) started, so the next two steps can start. Because the start of this is going to be slow until we get to the big rebuild section. 15:24:11 [2019-03-07-10:17] mhroncok: For the macros update, you've tested this pull request in copr, correct? And it's the right thing? https://src.fedoraproject.org/rpms/python-rpm-macros/pull-request/16 15:24:11 [2019-03-07-10:19] tdawson: yes and yes 15:24:11 [2019-03-07-10:19] tdawson: you build this 15:24:13 [2019-03-07-10:20] tdawson: then you rebuild python36 and python34 packages 15:24:15 [2019-03-07-10:20] mhroncok: OK, merging pull request now 15:24:17 [2019-03-07-10:20] tdawson: and then you rebuild the rest 15:24:19 [2019-03-07-10:20] tdawson: I can amend the commit message 15:24:21 [2019-03-07-10:20] mhroncok: Can I rebuild python36 and 34 at the same time? 15:24:23 [2019-03-07-10:20] mhroncok: If you dont' mind, that would be great 15:24:27 [2019-03-07-10:21] mhroncok: So we dont' have something merged with a big old "DO NOT MERGE" sighn around it's neck. 15:24:30 [2019-03-07-10:22] tdawson: amended. and yes, you can build python34 and 36 at the same time 15:24:32 [2019-03-07-10:22] just make sure this landed in the buildroot 15:24:34 [2019-03-07-10:22] disclamer: if only python36 is built but python34 is not, they will conflict 15:24:36 [2019-03-07-10:23] but once both are built with this change, it is going to be ok (until some magical bug appears) 15:25:15 mhroncok: Yep, I'll be overriding it, and then waiting for the repo to regen before I start the python34 and python36 builds. 15:25:32 tdawson: is this going to be in a side tag? 15:25:35 why do those need rebuilt? oh... to get the new macros...duh 15:26:13 mhroncok: From what I know, no. I'll just be setting the karma on the builds really high. 15:26:30 mhroncok: So they'll be in epel-testing for as long as we need. 15:26:40 I get it 15:27:05 people who build stuff might get a surprise, there has not been any announcement other than the calendar invite 15:27:54 when we submit them to epel-testing we should group them together 15:28:10 mhroncok, what kind of announcement should I do? 15:29:11 Good point ... I need to at least announce that I'm starting the macro rebuild, and the list of packages I intend on changing/rebuilding I guess. 15:29:48 well, we have discussed on epel-devel... 15:30:43 Here's a question ... how long should I put the overrides in? 15:31:07 Should I do it just until I'm done with the builds today? Or should I put it in as long as I do the karma? 15:33:22 For now, I'm just doing one day, which should give me enough time to do the builds. I can always extend it, but I don't see anything easy for shortening an override. 15:38:00 tdawson: python3Y builds for a while and there might be test failures to deal with 15:38:41 I guess you want the override to last as long as possible so you don't get new builds with old macros in before we manage to rebuild everything 15:39:39 yeah, I'd say set them really long... 15:39:53 rhel-7 ends in 2024 15:40:40 something like that, yes :D 15:49:14 mhroncok: Another question ... do I have to wait until the python34 and python36 builds are done before rebuilding the other python packages? 15:49:43 Part of me says "of course", but another part of me thinks I just have to have the macros package done and in override. 15:57:48 I'm around to help if there's anything anyone needs me to do. just ping me. ;) 16:01:22 nirik: Will do. 16:01:42 Right now the macros are done and in buildroot. python34 and python36 are rebuilding. 16:02:30 I'm not positive if I can/should start the other python builds, or wait for python34 and 36. 16:03:09 I'm going AFK for a little bit 16:08:08 I guess wait to be safe 16:13:58 tdawson: you do need to wait 16:15:14 tdawson: otherwise you'll get mixed content (macros meaning vs. actual content of stuff like /usr/bin/python3) 16:24:57 #chair tdawson nirik 16:24:57 Current chairs: nirik smooge tdawson 16:24:57 Current chairs: nirik smooge tdawson 16:25:28 thanks for advice mhroncok 16:37:30 mhroncok: As smooge said, thank you for letting us know. 17:09:38 tdawson, how does it look? 17:11:39 smooge: Both python34 and python36 are built on everything but aarch64 17:18:23 oh yeah we don't do s390 for epel 17:21:17 good afternoon 17:21:22 a meeting is going on? 17:21:50 kanarip: yes but we are cheering form aarch64 builds of python 3.X 17:22:20 that's the flag activity day i presume, that's fine, i'll shut up 17:35:15 going offline, we'll cehck again later 17:35:20 *will 17:44:29 for those wanting to follow along on the python34 and python36 builds, here are the task links - https://koji.fedoraproject.org/koji/taskinfo?taskID=33270347 https://koji.fedoraproject.org/koji/taskinfo?taskID=33270367 17:44:50 Still waiting for aarch64 on both 17:48:32 that doesn't give me much hope for the armhfp rebuild.....:( 17:49:03 on the other hand... the strategy worked and x86_64 and others are OK? 18:05:26 tdawson: I see the python3X builds are finished 18:11:16 Tagging them into override now 18:12:39 Interesting ... seems you can only tag into override for a max of 31 day ... at least from the command line 18:13:20 same for bodhi web 18:13:28 at least the calendar doesn't allow more 18:13:46 Well, that should get the packages built ... we'll see about extending it later 18:16:46 well, once they are in stable they are automatically in... 18:19:19 I'm going to start with the packages that have pull requests, then I guess I'll go through everything else alphabetically. 18:21:41 tdawson: get setuptools six and pytest early 18:22:03 mhroncok: python3-pytest? 18:22:34 yes 18:22:57 OK, will do. And I suppose those need to be tagged into override 18:23:12 mhroncok: Do you think I should tag everything I build into override? 18:25:03 tdawson: either everything or explicitly just the stuff that's going to be missed by others 18:25:30 at least setuptools six and pytest will be definitively needed 18:25:38 mhroncok: OK. I'll do setuptools, six and pytest ... and then whatever else is needed 18:26:08 tdawson: the copr builds section might give you a hint for order 18:26:16 https://copr.fedorainfracloud.org/coprs/g/python/epel-python3/builds/ 18:26:18 can I help with any of this? 18:27:04 mhroncok: OK, cool, thanks. 18:27:27 smooge: I'd like to say yes, but I don't think you have the permissions. 18:27:42 ok 18:27:58 I will do the frantic manager thing then 18:28:11 .fasinfo smooge 18:28:12 mhroncok: User: smooge, Name: Stephen J Smoogen, email: smooge@gmail.com, Creation: 2005-05-03, IRC Nick: smooge, Timezone: US/Eastern, Locale: en, GPG key ID: D027E793, Status: active 18:28:15 mhroncok: Approved Groups: +sysadmin-calendar @sysadmin-teleirc sysadmin-test @sysadmin-backup wikiadmin @sysadmin-veteran @vendor-support fedorabugs @sysadmin-logs cla_done cla_fedora packager +sysadmin-noc sysadmin-tools sysadmin-cloud sysadmin-securit +gitfedora-zikula @sysadmin @sysadmin-web gitsecstate sysadmin-devel @sysadmin-qa sysadmin-dba @sysadmin-troubleshoot @sysadmin-hosted @sysadmin-main accounts (1 more message) 18:29:23 hihi @ sysadmin-calendar 18:35:18 https://bodhi.fedoraproject.org/overrides/?user=tdawson 18:35:26 https://koji.fedoraproject.org/koji/tasks?owner=tdawson&state=all 18:36:10 OK, I'm going offline again. will reconnect on a train in ~1 hour 18:37:13 good luck 20:04:11 I'm back 20:05:29 hi mhroncok 20:06:52 smooge: hi. anything problematic so far? 20:07:26 tdawson: ^ 20:08:08 mhroncok: nothing major ... there are a few that are failling for things other than dependencies, but they aren't critical packages. 20:09:06 mhroncok: Just so you know, if there are any BuildRequires or Requires python34-*, I'm fixing those to be python%{python3_pkgversion}-* 20:09:34 I figure if there is ever a time to do it, now is the time. 20:09:59 tdawson, are you fixing those in the epel-7 branch? 20:10:05 pgreco: Yes 20:10:31 actually, does the epel-7 branch differ much from f30/master? 20:10:41 good 20:10:57 pgreco: it totally depends on the packager, and with how many I have to fix, I'm not checking master 20:11:32 tdawson, if you want, send me the list of "modified" and I can take a look at how much those differ from master 20:11:43 to see if it is worth unifying 20:11:55 specially thinking about epel-8 20:13:26 tdawson: generally yes, but there might be cases when they BR python34 and use python3.4 (which is OK). I don't think there are such cases, but when you change the package, maybe grep the specfile for 4 20:13:48 pgreco: That shouldn't be too hard to check, just get the list of packages, checkout master (I'm checked out epel7) and do a grep for python34 ... right now even this conversation is slowing me down. 20:14:15 mhroncok: What do you mean grep it for 4? incase there is a 3.4? 20:14:41 tdawson, ack, go on ;) 20:17:28 tdawson: yes 20:17:53 tdawson: or just don't and we can fix it if the build fails 20:18:17 mhroncok: Ok ... cuz ... well, grepping for 4 will have alot of positives 20:20:35 Ie seen crazy stuff 20:20:41 like %{__python3}.4 20:25:02 mhroncok: Hmm ... Just ran into one that had python34 listed specifically, and it failed, becuase it was looking for /usr/bin/python3.4: in the build ... looks like make wants it ... maybe when I get cmake done I'll put it in overrides 20:28:09 that's why %{__python3} should be "%{_bindir}/python3 " ;-) 20:31:36 tdawson: which one? 20:32:03 mhroncok: lhapdf 20:32:50 tdawson: https://src.fedoraproject.org/rpms/lhapdf/c/2f68c283e7326b63c59e0bd8b66fb35117ca7380?branch=epel7 20:32:55 tdawson: it required both 20:33:50 mhroncok: Ha ... I just got in the groove and didn't notice that. 20:33:52 mhroncok: Wow, that argues for defining %__python3 to end in a space. 20:34:18 tibbs: anyway, %__python3 in EPEL is actually /usr/bin/python3.6 20:34:28 Ah, right, it changed. 20:34:39 so this was maybe somewhere else, maybe it was %{__python}2 or something like that 20:34:43 \o> \o> \o> 20:36:23 mhroncok: Do you know offhand if we have a global variable for 34 ... or should I just put it back to the way it was? 20:37:11 tdawson: python%{python3_other_pkgversion}-devel 20:37:21 mhroncok: cool ... thanks 20:38:14 tdawson: and since they are using the macros everywhere, I'd use them in BRs as well for consistency and also if we ever go with other being 3.8 (probably not I guess) 20:38:47 mhroncok: Yep ... save us some pain in the future. 20:50:50 pgreco: nordugrid-arc was the last one the had python34 as a build requires in it ... I'm a little curious if it's the same in master, but I've already kicked off my build so my console is in watching mode and I can't look. 20:54:49 20:54:50 BuildRequires: python2-devel 20:54:50 %if %{?fedora}%{!?fedora:0} >= 13 || %{?rhel}%{!?rhel:0} >= 8 20:54:50 BuildRequires: python3-devel 20:54:50 %endif 20:54:50 %if %{?rhel}%{!?rhel:0} == 7 20:54:50 BuildRequires: python34-devel 20:54:51 %endif 20:54:53 from master 20:55:23 So ... yep, looks like we'll need to go through the master stuff for rhel8 20:56:04 I'll see if I can at least flag which packages I fix from here on out. 20:58:43 stuff from root logs that will need override: 20:59:51 Cython, dateutil, PyYAML, gobject, cached_property, productmd, docopt, texttable 21:00:13 websocket-client, docker, dockerpty, jsonschema 21:00:38 docker-compose has: No matching package to install: 'python36-backports-ssl_match_hostname >= 3.5' -- that is not needed with 3.6 21:00:54 setuptools_scm, llfuse, msgpack 21:01:12 dbus, 21:01:41 mhroncok: Yep, got them all on my list 21:02:08 mhroncok: The problem I have with doing this alphabetically, is that "python" comes after alot of things. :) 21:02:13 tdawson: oh, ok, sorry for the noise 21:02:34 mhroncok: Not a problem, it's good for others to be checking. 21:02:47 tdawson: pushing everyhting at once in the first batch is also an option 21:02:59 https://github.com/hroncok/mini-mass-rebuild 21:04:51 mhroncok: Huh ... though the biggest reason I'm doing it semi-automated is so I can look at the spec files if there is a python34 in them. I've got a few manual checks so I feel better. 21:04:55 i can fix docker-compose, that's mine anyways 21:05:06 carlwgeorge: good! 21:05:20 tdawson, checking 21:05:21 carlwgeorge: yet there are still dependencies missing for that one 21:05:48 i'll at least get the spec file squared away so it will be ready to build once the deps are there 21:06:20 at one point i was working on logic to handle the need for ssl_match_hostname, but it was just too messy 21:06:41 mhroncok: python36-Cython is probrubly the most requested 36 package ... and I started it quite a while ago, but it's been sitting on the aarch64 builder for a long time. 21:10:08 tdawson: sorry for not recommending that one earlier 21:10:22 I didn't know it was so heavily used 21:11:13 docker-compose should work now once the deps are present https://src.fedoraproject.org/rpms/docker-compose/c/517fd02f8e2deae8c8c1b5730794daa4d8f059db 21:12:28 mhroncok: 3 or 4 thus far ... not *heavy* use, but so far, the most requested 21:14:13 tdawson, wrt nordugrid-arc, it should be really easy to sync master and epel7 21:14:19 carlwgeorge, tdawson: also python3-backports-ssl_match_hostname -> I guess not worth rebuilding at all. either retire it or switch it to py3_other 21:14:23 just 2 patches added and your change 21:16:03 tdawson: I'm trying to get python-distlib imported to epel7 and built fresh. once I push it, should you trigger the build, so it is not forgotten in the update? 21:16:30 (new 3.6 only epel7 package) 21:19:17 * nirik thanks everyone here working away at this. ;) 21:20:01 tdawson, applying this against master should work for epel https://paste.fedoraproject.org/paste/W6wXilaVkHkj11mUD11LcA/ 21:22:50 mhroncok: python3-cryptography pull request ... merge it? - https://src.fedoraproject.org/rpms/python3-cryptography/pull-request/3 21:22:50 tdawson: anyway the PR is here https://src.fedoraproject.org/rpms/python-distlib/pull-request/4 21:23:28 tdawson: first this one https://src.fedoraproject.org/rpms/python3-cryptography-vectors/pull-request/2 21:24:05 tdawson: once that one is merged and build, you should merge and build the one you've linked 21:24:39 tdawson: it was not merged yet, we waited for maintainer's feedback, this is potentially a breaking change (yet this entire thing is breaking, so why not do it now) 21:25:20 I'd say go ahead and if they get mad, I'll take the flak. ;) 21:25:20 mhroncok: OK ... I didn't see that one and just got done building -vectors ... but ... hopefully the -vectors one will rebase cleanly. 21:25:30 tdawson: you need to rebase that PR 21:25:34 .whoowns python3-cryptography 21:25:35 nirik: owner: abompard 21:25:46 tdawson: yes, only changelog conflict 21:25:48 ah, I don't think he will mind. 21:27:03 he was mostly silent in Fedora recently :( 21:28:11 he's been busy on other stuff, but I know he wants to move to python36 for mailman3 21:29:10 good 21:32:22 i'll switch python3-backports-ssl_match_hostname to the other macros 21:34:45 carlwgeorge: https://src.fedoraproject.org/rpms/python3-backports-ssl_match_hostname/c/83297c2f11530b4deb82b1707a735f98d2b3d1fe?branch=epel7 https://koji.fedoraproject.org/koji/taskinfo?taskID=33279961 21:39:23 yeah python36-backports-ssl_match_hostname-3.5.0.1 doesn't make any sense 21:46:44 python3-backports-ssl_match_hostname is ready to rebuild again to keep it on py34 https://src.fedoraproject.org/rpms/python3-backports-ssl_match_hostname/c/fae4ee81a5b9cea65445a54611d7d3cec11bb177?branch=epel7 21:48:05 tdawson: ^ 21:48:21 tdawson: I see you've created a buildroot override for it 21:49:10 carlwgeorge: Wait ... why do we want to keep it at py34? This rebuild it to transition to 36, and there are several packages (two thus far) that need the 36 version. 21:49:21 tdawson: they don't 21:49:27 this is in 3.6 standard library 21:49:43 tdawson: one of them is docker compose, carlwgeorge removed the dependency 21:49:47 what is the other? 21:50:14 python3-docker 21:50:16 :) 21:50:20 makes sense 21:50:22 (22:14:19) mhroncok: carlwgeorge, tdawson: also python3-backports-ssl_match_hostname -> I guess not worth rebuilding at all. either retire it or switch it to py3_other 21:50:30 you must have missed this one 21:50:42 mhroncok: Ya ... I've been pretty heads down 21:50:53 yeah, backporting a 3.5 change to 3.6... not so much 21:51:37 So ... I take it I should cancel the override :) 21:52:00 tdawson: yes please 21:52:13 tdawson: also, I suspect thet nose will be used 21:52:23 so you may create one for python3-nose-1.3.7-4.el7 21:52:37 OK, override canceled for python3-backports-ssl_match_hostname-3.5.0.1-2.el7 21:53:03 tdawson: if we have stuff to build, should we build it or throw it at you? 21:53:29 mhroncok: Is it on your original list? 21:53:39 no 21:53:51 mhroncok: Then feel free to build it. 21:53:54 but it will need to be on the update list 21:54:10 carlwgeorge: you can build the 34 backport ssl match hostname 21:55:32 for the record https://koji.fedoraproject.org/koji/taskinfo?taskID=33280590 21:55:49 that is distlib a new epel7 python36 only package 21:58:46 tdawson: python3-mock is also a good candidate for override 21:59:26 cool, i'll fix python3-docker too 21:59:39 I'm getting frustrated with python36-Cython ... it's been 3 hours in teh build, and just barely failed on aarch64 :( 21:59:47 tdawson, wrt nordugrid-arc, if you apply my patch, then git checkout epel7;git revert e9fb5e9ca93db493769918fb49724a8f0d93f8d1;git merge master 21:59:49 There are at least 6 packages that want it 22:03:44 tdawson: our copr didn't have aarch 22:03:48 tdawson: let me check the failure 22:04:42 that's ppc64 failure 22:05:10 It's the 3 hours of waiting that makes me sad ... well ... the failure too. 22:06:02 tdawson: https://github.com/cython/cython/issues/1982 22:06:19 tdawson: it's a test failure only, on big endian, fix available upstream 22:07:19 python3-docker is ready to be rebuilt again when the deps are available 22:07:40 tdawson: workaround (if patching is too hard) is viisble in https://src.fedoraproject.org/rpms/Cython/c/8fbaae6d8482329a132e94734c13726a070cd87d?branch=master line 105 22:07:55 tdawson: cython in Fedora builds for 4-6 hours on arm 22:08:37 what are we doing with builds that are done? one big update or sidetag stuff? 22:08:54 python3-backports-ssl_match_hostname rebuilt to stay on py34 -> https://koji.fedoraproject.org/koji/buildinfo?buildID=1226550 22:09:08 * carlwgeorge doesn't know much about the sidetag approach 22:11:44 carlwgeorge: one big bodhi update 22:12:18 how are we keeping track of all the builds that are going to be included? 22:12:36 carlwgeorge: no idea 22:12:42 😀 22:12:49 carlwgeorge: but python3-backports-ssl_match_hostname doesnẗ need to be included :D 22:13:22 mhroncok: python3-pytz ... is this pull request ready for merging? I'm not really sure why it's waiting for 36 - https://src.fedoraproject.org/rpms/python3-pytz/pull-request/1 22:13:23 should i submit it separately then? or keep it together so it's released at the same time? 22:14:09 carlwgeorge: together 22:14:20 carlwgeorge: otherwise your package could end up being there sooner 22:14:52 tdawson: it was waiting for the maintainer to respond 22:15:02 tdawson: but I think it's good to be merged 22:15:27 mhroncok: It's been a month, with no response ... I'm ok pulling it in. 22:15:40 sure thing 22:17:01 this was probably broken on 3.4 as well for some time now. no idea why nobody realized 22:21:01 tdawson: python3-Cython-0.28.4-2.el7 aarch64 built for 7 hours. I suggest you skip the tests for now to get it over with and reenable them immediatelly after 22:24:24 tdawson, I am ok with that in case you need a second opinion 22:24:58 *nods* If nothing depended on it, I'd do something else, but right now, there are quite a few packages that need it, some that have other dependencies. 22:26:50 * mhroncok is approaching Prague, will go offline again in couple minutes 22:31:15 bye 22:32:06 tdawson, when do you want to call it a day today? 22:33:14 smooge: well, probrubly another hour and a half. 22:33:44 ok no problem 23:20:49 I am going to end this meeting for today. will start it up tomorrow with a recap and what people can do 23:20:53 #endmeeting