18:00:46 #startmeeting Copr 18:00:47 Meeting started Wed Apr 7 18:00:46 2010 UTC. The chair is skvidal. Information about MeetBot at http://wiki.debian.org/MeetBot. 18:00:48 Useful Commands: #action #agreed #halp #info #idea #link #topic. 18:01:02 #chair abadger1999 18:01:03 Current chairs: abadger1999 skvidal 18:02:42 Alrighty, spot, adamw, mizmo, any other interested parties, we're starting 18:02:52 okie doke 18:03:36 I guess I'll start? 18:03:42 Sounds good. 18:03:45 more work on sync to repo: http://skvidal.fedorapeople.org/misc/sync_to_repo.py 18:04:07 this is mostly to help folks setting up a repo and wanting to add pkgs to it but not change anything else 18:04:15 (this is more or less the chromium problem) 18:04:27 since i'm doing other stuff, i'll just throw this in - still didn't hear from jono, though i pinged him. at this point probably worth others trying their ubuntu contacts. 18:04:29 I sent email to all the folks with repodata on fedorapeople.org 18:04:36 Will also help the people uploading packages for reviews 18:04:39 and asked a few questions 18:04:51 specifically these questions 18:04:53 1. What do you keep in your repo? 18:04:53 2. If the contents of the repo are not in fedora, why not? 18:04:53 3. How did you build the pkgs in the repo? 18:04:53 4. Do you update the pkgs in the repo? If so, how often and what process do you use to update them? 18:04:53 5. If we were working on the tools to maintain your own repos like this - what things would you suggest to 18:04:54 adamw: Thanks 18:04:55 include in those tools - or in short - what are the unfun things about maintaining a repo now? 18:05:06 the email was sent to 65people 18:05:12 and I got back about 23 responses 18:05:19 which is not bad, considering 18:05:28 the responses varied somewhat widely 18:05:35 but the themes I noticed were 18:05:56 1. most folks are doing scratch builds with modified srpms 18:06:31 2. the folks doing stuff with different build deps are using mock or smock.pl to build their pkgs and then createrepoing and pushing them up 18:07:03 3. a few cases wanted to be able to work more easily with more than one person to maintain a repo 18:07:21 4. so it means having a shared signing key between them and a shared space to upload files to 18:07:37 AND a way to upload to that space w/o needing to keep their ssh private key semi-exposed 18:07:58 one person wanted launchpad.... which was... interesting 18:08:32 so it seemed to me 18:08:39 we can help multiple users a bit at a time 18:08:55 the long term goals of coprs help all of these users (well, not the launchpad person but...) 18:09:18 in the short term - making it easier to build pkgs with arbitrary builddeps from arbitrary repos helps users 18:09:40 and making it easier to add/remove a few pkgs from a repo w/o having to have a complete copy of the repo locally 18:09:58 now there are two ways to deal with that problem. 18:10:22 one thing I've been working on is the sync_to_repo - thing I linked above -it's let me play with a few ideas which is fine and it works - which is fine 18:10:47 but the other way is to just install the (albeit somewhat old) ver of createrepo on fedorapeople.org and if the machine runs out of ram during a run then 'oh well' 18:11:22 as far as making builds easier from arbitrary repos 18:11:48 and easy-ish win is seeing if there is a nice way to integrate what smock.pl is doing into mock itself so you don't have to have 2 languages and 2 codebases to do it 18:12:34 I think that's most of what I have 18:12:48 18:12:49 if anyone wants to see the responses I got back from the email I have them all saved 18:14:13 * skvidal waits through the echoing applause..... 18:14:17 Okay. So this week, I started writing some code. Decided that basing things around func would get us to a proof of concept quickest. 18:15:36 Worked through some of my misconceptions of how func works with skvidal and I have func setup in a way that will work for us. 18:15:48 yay 18:15:54 Certmaster runs on one box, in Fedora Infra, that's probably puppet01. 18:16:05 The task scheduler will run on a different box. 18:16:29 That way the task scheduler will only have access to run func commands that we explicitly grant via acls. 18:16:40 (the certmaster box can run any commands) 18:16:43 abadger1999: you don't have to use the same certmaster, of course 18:16:58 you can have an entirely different set of ports, etc, if you want to 18:17:23 that's true... we just need to kep the ca cert away from the func commandline client. 18:17:46 abadger1999: ah - new func takes func -c config 18:17:54 I think for infra, keeping puppet01 just a box where we do configuration is probably best. 18:18:00 and in that config file you specify the ssl certs you want to use for auth 18:18:05 abadger1999: agreed 18:18:07 skvidal: Ah, that will be nice for what we're doing. 18:18:57 So that piece of work done, I then started coding a func module to list the commands that we're allowed to run. 18:20:30 I've gone down a few false paths but I think I've figured out how to do it -- the method will need to take a hostname. The module will look up the client cert for that hostname and then pass it into the existing acl checking function. 18:21:20 Once that's done, I'll start writing a module to invoke a mock build on a remote host. 18:22:12 ok 18:22:29 With those two, we can move on to someone working on the task scheduler, someone working on the buildsystem server, and someone continuing to add func modules for creating repos, etc. 18:23:19 Which brings me to the Fedora Summer Coding effort. 18:23:22 #link https://fedoraproject.org/wiki/Summer_Coding_2010 18:23:49 should we be counting on that happening at this point? 18:24:04 There's only a small amount of funding for summer coding projects this summer, but there is some. 18:24:40 So I've started working on organizing the docs we have; getting rid of kopers thoughts that we've discarded. 18:25:02 I'll be putting copr on the list of projects that can be worked on. 18:25:31 I'm hoping that we can attract one of the people that's worked on fedora infra projects before. 18:26:34 As I could assign them to work on one of the big chunks of functionality and get a result that we'll like. 18:27:39 If we get less experienced people interested, it'll be harder to know what we can do with them but we'll see what we get. 18:29:19 I'm thinking of putting the design and implementation of the packager facing code as the main project... should be a reasonably straightforward TG2 app + command line client using the json interface to get us off the ground. 18:29:58 My deadline for getting copr submitted as a project is next wed. 18:30:06 .... 18:30:10 Alright, that's all I've got. 18:30:51 okay where to from here? 18:32:12 Getting enough code to do something useful. 18:32:30 indeed 18:32:35 Getting docs updated so people know what we are and aren't doing 18:32:52 bah - docs 18:32:55 :-) 18:33:29 okay 18:33:29 so 18:33:44 since it seems like just you and me talking to each other... 18:33:55 Oh -- what's the test hardware setup? 18:34:04 cnode5 is all we have 18:34:19 and it's just for dinking with, really 18:34:41 do we have something that needs to be tested there? 18:35:09 Not yet. But All my work is currently on my local network. 18:35:20 I've documented what to do to replicate it. 18:35:45 But it's always nice to see it start working on a second system. 18:36:49 spot: ping - if we think we might need some hw 18:37:13 spot: can you help? 18:39:03 I'll take the echoing silence as a "spot is in a meeting try back later" 18:39:38 abadger1999: anything else? 18:39:58 Nope. seems like we've updated each other now :-) 18:40:50 okie doke 18:40:55 #endmeeting