16:01:45 #startmeeting fpc 16:01:45 Meeting started Thu Apr 16 16:01:45 2015 UTC. The chair is geppetto. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:01:45 Useful Commands: #action #agreed #halp #info #idea #link #topic. 16:01:45 #meetingname fpc 16:01:45 #topic Roll Call 16:01:45 The meeting name has been set to 'fpc' 16:02:32 geppetto limburgher mbooth orionp racor Rathann SmootherFr0gZ spot tibbs|w tomspur: FPC ping 16:02:38 Howdy. 16:02:49 #chair tibbs 16:02:49 Current chairs: geppetto tibbs 16:02:56 hey 16:03:05 Hi 16:03:05 did you have a meeting last week? 16:03:09 #chair tomspur 16:03:09 Current chairs: geppetto tibbs tomspur 16:03:09 Yes. 16:03:25 But I suck at the whole meeting thing and failed to send an agenda. 16:03:30 ahh 16:03:34 We did get some stuff done. 16:03:39 didn't see any minutes either … so wasn't sure :) 16:03:48 The minutes are auto-sent now. 16:03:53 Oh 16:03:56 Just not to that mailing list. 16:04:00 ahh 16:04:05 Haven't figured out what we're supposed to do there. 16:04:33 blame nirik ?? ;) 16:04:41 Works for me. 16:05:10 I accept your blame and raise you guilt! 16:05:21 hi 16:05:38 #chair Rathann 16:05:38 Current chairs: Rathann geppetto tibbs tomspur 16:05:44 nirik: ha :) 16:05:59 orionp and mbooth said they couldn't make it. 16:06:18 hmmm 16:06:51 limburgher racor Rathann SmootherFr0gZ spot: FPC ping 16:07:23 Bah sorry for all the pings Rathann 16:08:19 :) no problem 16:10:37 doesn't look like we'll get 5 16:11:15 Schedule was: https://lists.fedoraproject.org/pipermail/packaging/2015-April/010558.html 16:11:21 So you get off easy. 16:11:25 Is there anything we should talk about anyway? 16:11:31 That UID thing. 16:11:37 for ceph? 16:11:39 Yeah. 16:12:07 I'm thinking that may be one of the few valid reasons, except that I don't understand why it would actually help their use case. 16:12:20 But I don't know squat about ceph. 16:12:49 #topic #524 static UID for ceph 16:12:55 https://fedorahosted.org/fpc/ticket/524 16:13:04 BTW, .fpc works now 16:13:07 .fpc 524 16:13:09 tibbs|w: #524 (static UID for ceph) – fpc - https://fedorahosted.org/fpc/ticket/524 16:13:23 nirik set that up for us last week. 16:13:30 oh, cool 16:13:53 So I just do #topic + .fpc now? 16:14:06 I think so; saves you from pasting the URL. 16:14:08 So, yeh, the ceph thing looks like a pretty normal shared storage thing 16:14:44 So what would having a static UID actually give them? 16:15:13 If they're a filesystem, they could just specify the UID at mount time like FAT does. 16:15:26 I assume the ceph stuff needs write access to the files 16:15:28 Or is it some kind of layered thing where there's really ext4 underneath. 16:15:51 But if they're a filesystem, they can give write access to whatever they want. 16:15:57 That's what I'm not understanding. 16:16:34 I mean, if I mount fat or tmpfs, I can pass uid= whatever I want. 16:16:52 AIUI the back store for the ceph filesystem is giant files stored on ext4 16:17:16 So you plug the drive in with ext4 files on it, and then ceph needs to be able to rw those files 16:17:21 OK, that makes more sense. I figured it had to be something like that. 16:17:42 And according to the ticket they're trying to get away from using root. 16:17:52 yeh 16:18:10 Even a shim to do a chown would work. 16:18:14 if it's ext4 then it's easy to mount it temporarily somewhere and fix the uid before mounting it in the final place 16:18:19 But... 16:18:45 IMHO they should start a discussion on the cross-distro corrdination mailing list first 16:19:18 Actually suse and debian just copied what ktdreyer put in the FPC ticket. 16:19:32 Thing is, they're going to have to do the chown thing anyway, to support existing filesystems. 16:20:23 ext4 doesn't seem to support forcing the UID on mount, which is unfortunate as it would obviate the need for this. 16:21:14 So, I guess I'm not particularly opposed to this, but it seems like the least creative way to handle this. 16:21:44 And if distros don't agree on the actual UID chosen then I'm not entirely sure what the point is. 16:22:48 it's still useful within a distro. family though 16:22:57 But I guess the actual choice of the UID would be up to the setup maintainer. 16:24:00 If I can ever get the time to work on this idea I have about fixing UIDs before package installation, the "within the distro family" thing wouldn't be much of a concern for Fedora. 16:24:28 orion showed ma a glorious hack that means I don't have to add a new kickstart section to anaconda. 16:25:41 that would be cool 16:25:58 Yeah. I just wonder if anyone would accept the hack. 16:26:03 although that means a custom kickstart for every machine you want to share ceph files with, right? 16:26:21 Not really, no. 16:26:42 Well, for machines that are already installed, you can just change the UID it uses. 16:27:19 Fresh kickstarts are always the issue, so allowing UID fixing there would solve the majority of the issues. 16:27:25 We have five now, don't we? 16:27:44 hi, sorry for being late. 16:28:15 racor: We were just discussing the static UID for ceph. 16:28:18 .fpc 524 16:28:21 tibbs|w: #524 (static UID for ceph) – fpc - https://fedorahosted.org/fpc/ticket/524 16:28:22 ok 16:28:48 We could vote now, but I'm not sure there's consensus. 16:29:06 #chair racor 16:29:06 Current chairs: Rathann geppetto racor tibbs tomspur 16:29:17 I'm +1 on letting them have it 16:29:29 I as well. 16:29:30 +1 16:31:22 have one for "all Fedora/RHEL systems" or wait for a generall distribution outcome? 16:31:25 0 ... I am lacking sufficient background to be able to have an opion 16:31:40 tomspur: Really the choice of UID isn't up to us. 16:31:55 If we allow it, I'm sure they'll do their best to find one everyone can agree on. 16:32:48 yeh 16:32:53 that's what I'm assuming anyway 16:33:29 I wonder how they achieve "exchanging drives" without being root to set uids 16:33:47 RIght now they are root. 16:34:09 In the future they won't be, so they want to pick a UID so that they can exchange drives that way. 16:34:20 the request is about doing the same "unprivileged" 16:34:27 there are a bunch of situations where you might want to let someone be able to insert/remove drives from a hot swap bay but not have root on the drive 16:34:34 on the host* 16:37:18 racor: Yes, if they can depend on the UID in use, they don't need to be privileged. 16:37:32 Supposedly. I mean, something actually has to mount the filesystem. 16:37:43 I guess they use udisks or whatever we're using now to handle that. 16:38:16 yeh, I assumed auto mount 16:39:06 It's all still kind of confusing to me. I'm assuming they've thought about this quite a bit, but nobody communicated that to us. 16:39:21 At least a link to a list discussion or docs or something would have been useful. 16:39:32 Anyway, no way we'll get to +5 today. 16:40:17 yeh 16:40:57 #action A link/summary of the upstream discussion of how this will be used, might be nice. Currently split on how this is useful. 16:41:24 #topic #522 Should -static packages require -devel 16:41:27 We are at +2, though, and could collect votes in the ticket. 16:41:36 .fpc 522 16:41:37 geppetto: #522 (Should -static packages require -devel) – fpc - https://fedorahosted.org/fpc/ticket/522 16:41:54 Hopefully we'll have more people next week and can just look at it again, then 16:42:15 To me this 522 is just one of those things I don't care about. 16:42:44 I still think my assumption there is not that "wrong"? 16:43:11 This would make it more complicated to actually link to the static library, but maybe that's also intended :) 16:43:17 Yeh, I don't see why we want to add must here 16:43:39 I don't see why we want to spend time on this at all for what would be, what, ten packages? 16:44:02 +1 Actually, I wonder people came up with this. It used to be "common sense" and is inevitable in practice (Header must be in *devel, so no *static without devel) 16:44:21 tomspur: The only reason I can think of is that relying on that then breaks the ~0 people who still use rpmbuild on a host that has the -static and -devel installed 16:44:53 although > 0 for epel, I'd guess 16:45:08 I was willing to +1 it last week because of not caring enough either way. Why not save a line in a couple of specs if it makes sense? 16:46:05 so what is the proposed rewording? 16:46:22 I thought it was obvious that static must require -devel if headers are there 16:46:29 There wasn't one, making it more fun. 16:46:40 meh 16:46:44 Rathann: the ticket seems to be more like "static always requires devel" 16:46:49 geppetto: In fedora, the implicit default is to link dynamical. static linkage only takes place if no shared libs are available or if static linkage has explicitly been specified. 16:46:53 Wait.... 16:47:03 " If the *-static subpackage requires headers or other files from *-devel in order to be useful it must require the *-devel subpackage. " 16:47:30 orionp wrote that in there last week. 16:47:43 Did I forget to update the ticket after a vote? Probably. 16:47:43 hmmm 16:48:06 #action was voted on last week, and policy changed. 16:48:15 Just so happened that the past three weeks have been incredibly busy for me. 16:48:21 tibbs|w: +1 to that wording FWIW 16:48:32 Rathann: That's the current wording, actually. 16:49:00 Let me dig the vote out of the logs. I swear I pasted it in there already but might not have hit submit. 16:49:02 tibbs|w: not on the wiki yet 16:49:12 I just pasted it from the wiki.... 16:49:16 #topic #515 Bundling determination and exception request 16:49:18 I still think it silly to have to explicitly say in guidelines: "if pkg A needs pkg B to be functional, then pkg A MUST Requires: B" , in order to get maintainer to add the dep :( 16:49:21 .fpc 515 16:49:22 geppetto: #515 (Bundling determination and exception request) – fpc - https://fedorahosted.org/fpc/ticket/515 16:49:32 rdieter: Don't disagree, but meh. 16:49:34 rdieter: yeh, but it's there now 16:49:49 for -devel case 16:50:02 Sometimes people want to play lawyer. "But the guidelines don't say that I have to, so pbftttt." 16:50:05 (it should be common sense) 16:50:31 indeed, that's why this happened, which is just dumb, sorry for the side-track 16:51:32 We didn't talk about 515 in the last two weeks. 16:53:12 Seems like his questions are valid, but I don't know the answers. 16:53:40 I don't see anything in the last comment to change my mind … static lib. that *&%$ 16:54:00 for his three questions: 16:54:06 Probably just a case of someone needing to help him do it. 16:55:15 1) Don't care … pick one. 2) Not really, it's still static. 3) the size is based on the bundling code (not the clients) … which is pretty big from what I can see. 16:55:46 Anyone else have a different opinion? 16:56:18 Not me. 16:56:38 I've never done that static library packaging thing, though. How difficult is it? 16:56:58 +1 16:57:00 Just like a normal library, but you don't have to care about ABIs as much :) 16:58:58 #action Create a static. lib., decide among yourself who will own it. Size exception is based on the bundling size (and xmlrpcpp is big). 16:59:10 #topic #520 [Guidelines Draft] Per-Product Configuration Defaults v2 16:59:16 .fpc 520 16:59:17 geppetto: #520 ([Guidelines Draft] Per-Product Configuration Defaults v2) – fpc - https://fedorahosted.org/fpc/ticket/520 16:59:37 well, why not create a shared library while he's unbundling it anyway? 16:59:39 This was hung up because we asked for something to be macroized and the redhat-rpm-config maintainers haven't responded to the ticket. 16:59:53 Rathann: that's much harder if upstream don't care about ABI 17:00:04 ok 17:00:23 fair enough 17:00:37 tibbs: do we leave it in needinfo then, or keep it in meeting and check the BZ every week? 17:00:51 move it to needinfo* 17:00:55 I suggest to sgallagh that he just go ahead and add the damn macro. 17:01:01 :) 17:01:16 If the macro is good, it doesn't matter what package has it as long as the guidelines match. 17:01:29 where would he add it to? 17:01:34 And if the redhat-rpm-config maintainers balk, he can move it. 17:01:45 tibbs|w: OK, I'll just push it to redhat-rpm-config later today 17:02:10 Unless there's some other package which all of these packages will need. And I don't think there is. 17:02:26 For language stuff you'd stick it in the appropriate -devel or compiler package or whatever. 17:03:03 We can drop macros into /usr/lib/rpm/macros.d trivially now, unlike in the dark old days. 17:03:15 * geppetto nods 17:03:43 tibb|w: ACK 17:04:28 this would also make packages which are using these macros identiable at the rpm level. 17:05:40 So do we need to do anything, or are we waiting on the package changes that sgallagh will do? 17:05:51 sgallagh: So, honestly, if you don't want to mess with redhat-rpm-config, just put it into an appropriate package and tweak the guidelines to say that packages which will do this kind of thing need to BR: that package. 17:05:53 * nirik wonders if we may want some more active redhat-rpm-config maintainers 17:06:09 nirik: I kind of assume it's community-maintained at this point. 17:06:22 I added some macro a while back and nobody seemed to care at all. 17:06:26 nirik: Aren't they the rpm maintainers? 17:06:49 It's ffesti, I think 17:06:53 * geppetto nods 17:07:02 so, yeh 17:07:03 geppetto: yeah, but I think ffesti has wanted as little to do with it as possible 17:07:05 It feels to me like that package should be group-maintained by FPC 17:07:05 Haven't seen anything at all from him in a long time, though. 17:07:28 in fact I recall in a bug him saying for me to make changes he didn't have time/desire. 17:07:33 ffesti, jcm, and Panu. 17:07:36 Yeh, I wouldn't be shocked if it was way down his TODO list … and he's busy since Panu left 17:07:45 ajax wanted in but hasn't been given access. 17:08:09 weird 17:08:21 He asked for access a long time ago. 17:08:40 Before the pkgdb timeline, at least. 17:08:53 Anyway, I say just do it and lets go forward. 17:09:02 sgallagh: Is use of that macro documented in the current draft? 17:09:05 Hmmm … that's a long time ago :) 17:09:32 Doesn't look like it. 17:09:47 tibbs|w: No, sorry. 17:09:51 I'll do that after Go/No-Go 17:10:05 And wasn't there a question recently about some use of %ghost? 17:10:11 that kind of fell off my radar with beta release shuffle 17:10:27 16:21:26 how about %ghost'ing the symlinks? 17:10:31 tibbs|w: Right, I remember 17:10:33 16:21:49 racor: Did I forget to write that? 17:10:35 I will add that too 17:10:35 16:21:55 Yeah, that's necessary. Oops 17:10:38 Cool. 17:11:04 amusingly ajax was the last person to build redhat-rpm-config 17:12:25 So I guess we can revisit this next week when the macro should be in and the draft fixed up. 17:13:16 I'm working on it right now in parallel. If there are other agenda items, I may be able to finish it before you're done 17:13:34 Ok 17:13:59 #action sgallagh just do the macro change, and update the draft. We'll vote on it next week. 17:14:06 #topic #513 Use python -Es in shbang 17:14:14 .fpc 513 17:14:16 geppetto: #513 (Use python -Es in shbang) – fpc - https://fedorahosted.org/fpc/ticket/513 17:14:40 I think we are waiting on tibbs to do macros here? 17:14:47 Well, there's recent discussion. 17:14:55 And there are macros in 281. 17:15:00 .fpc 281 17:15:01 tibbs|w: #281 (New Python Macros for Easier Packaging) – fpc - https://fedorahosted.org/fpc/ticket/281 17:15:15 * geppetto nods … but they needed tweaking IIRC 17:15:52 I'm happy to delay it another week or so 17:16:20 esp. as the meat of the ticket has been voted on already 17:17:10 Yeah, this is no big deal. Plus there's so much confusion about the right thing to do with the unversioned macros. 17:17:18 * geppetto nods 17:17:19 #topic Open Floor 17:17:50 I did a quick look at 508, but I'd rather give tibbs a few more weeks to do his thing 17:18:16 would be cool if he/we got an instant gratification for the work. 17:18:53 Anything else before lunch? 17:20:28 geppetto: Page updated now 17:20:34 https://fedoraproject.org/wiki/User:Sgallagh/Per-Product_Configuration_Packaging_Draft 17:21:03 I haven't built the updated redhat-rpm-config yet, but that should be the presumed final form of the draft 17:21:07 Sorry, someone at the door. 17:21:35 If I can finish this current project (implementing a security camera system, including writing the software) then I'll have some actual free time. 17:22:04 Sorry, also someone at the door. Need to go... 17:22:42 sgallagh: ok, cool 17:23:38 I'm out. Will try to do a pass over the existing tickets and see what I forgot to record and close up over the past two weeks. 17:25:48 Ok, I guess we're done 17:25:49 #endmeeting