=== mgagne is now known as Guest51737 [15:24] smoser, I'm curious how soon you think 0.7.9 might drop...I see that there are a ton of changes to the systemd units since 0.7.8, and thinking about carrying patches vs. just waiting a bit. === another_larsks is now known as larsks === Guest51737 is now known as mgagne [18:42] harlowja, systemd dependencies question: if cloud-init-local.service is Before=NetworkManager, shouldn't it also be Before=networking.service? [18:43] And similarly, if cloud-init.service is After=networking.service, shouldn't it also be After=NetworkManager.service? [18:43] This may just be me being unfamiliar with the naming of ubuntu/debian systemd units... [18:54] nrezinorn ^ [18:55] larsks there was dragons there that nrezinorn hopefully remembers [18:56] harlowja, because I need to set deps correctly on the redhat legacy "network.service" as well as NetworkManager, and want to be consistent... [18:56] draggggggons [18:56] lol [19:00] * harlowja poked nrezinorn internally, he should be able to answer [19:01] all i remember is that on the newer cloud-init the systemd files were wrong for RHEL 7 [19:02] doyou have the link to the bug or patch? This was a while ago, i’ll have tp page this stuff back into my brain [19:02] because its called network.service (rhel) vs networking.service (ubuntu) [19:03] thats all i remember, and i think i just made a specific file for rhel builds, but i dont think any of that is merged back upstream yet [19:03] i think taht was the only difference, too [19:04] once i fixed the rpm build for network.service things just worked - for my super limited testing that i did lol - other things have taken my time away [19:05] larsks u trying to fix it for good i hope :-P [19:05] never again! [19:05] lol [19:09] nrezinorn, where are your fixes? [19:10] harlowja, I don't know. Because rhel and ubuntu have different service names, we can't really have a single unit file. [19:10] I think unit files should be part of the packages, not part of the upstream code, but smoser and I disagree on that point. [19:10] larsks right, we may need 2 [19:10] larsks prob disagree with me also :) [19:11] i like it when code-bases can build themselves into packages [19:11] if we have 2 unit files, doesn't seem to bad [19:11] harlowja, I think it's a terrible idea because it requires the upstream code maintainers to also be expert package maintainers for multiple distributions. [19:11] Or they end up being experts in one distribution, and everyone else has to carry a list of patches a mile long. [19:11] sounds like package creation is to much a PITA then [19:11] and if it requires a expert then its not done right :-P [19:12] packaging software shouldn't require experts imho ;) [19:12] Which is why it should just be left to the package maintainers. [19:12] (not in the 21st century) [19:12] Why make upstream developers worry about those things? [19:12] Let the upstream concentrate on quality code, rather than making sure they have the names of depdendencies correct. [19:12] i get that POV [19:13] it still feels like a messed up situation [19:14] I dunno. I mean, ubuntu/debian/redhat/suse/fedora/alpine/etc...everyone has there own packaging system. I'm not sure it's possible to get everyone to settle on a single set of package metadata. [19:14] ...until systemd introduces a packaging format, I guess. [19:14] sounds like we need a package trump for president [19:14] lol [19:15] make packages great again! [19:15] So here's another idea...we remove *all* distribution specific dependencies out of the unit files, and those go into separate systemd drop-in files. This lets the distro-specific stuff be managed without needing to duplicate the entire unit file(s). [19:16] k [19:17] nrezinorn, you mentioned that you were working on the rpm packages...what did you end up with for service dependencies? Are those rpms somewhere I can grab them and look? [19:29] larsks: i last poked things 2 months ago- and i dont have anything to share . let me see if i can drop a file somewhere you can grab...you can poke it yourself (if i remember where i did all this stuff) === mpontillo_ is now known as mpontillo [19:35] i cant find where i put the stuff i was doing lol, it may have been on a server that got wiped...so thats fun [19:40] nrezinorn, thanks for looking! [19:41] i really only added the rhel files for systemd and editing some other thing, prob the makefile macros to feed in the rhel systemd files [19:41] there was also other things not working right lol [19:46] i looked under a few rocks, sorry i couldnt find it. [19:56] harlowja, debian packaging question: what part of the packages/debian/* files causes the systemd unit files from the systemd/ directory to get copied into the right place? [19:56] eck, ask smoser not me :-P [19:57] Darn. smoser seems to be afk today. Or at least af irc. [19:57] ill find it for you [19:58] i don't do that debian stuff, ha [19:59] https://git.launchpad.net/cloud-init/tree/packages/bddeb?id=166df605dc9864eb163007300db7a611feb309d6 something in that lol [19:59] jsut trace it all the way down to where its copying the systemd/ folder i guess :) [20:00] ah i see [20:00] so the debian packager just takes ther tarball of the source [20:00] so "it's just there" [20:04] nrezinorn, ugh, right, we've abused setup.py as a general installation tool. Thanks. I keep forgetting that. [20:05] yeah im not a fan of that stuff lol [20:06] @harlowja: around? [20:08] Ugh, so why does the spec file copy things in manually? HEAD ASPLODE. [20:48] rpm spec? [20:48] it copies things becuase it has to be copied into the RPM BUILDROOT, thats how rpms work