16:05:50 <abadger1999> #startmeeting fpc
16:05:50 <zodbot> Meeting started Thu Jun 26 16:05:50 2014 UTC.  The chair is abadger1999. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:05:50 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
16:05:55 <abadger1999> #meetingname fpc
16:05:55 <zodbot> The meeting name has been set to 'fpc'
16:05:59 <abadger1999> #topic Roll Call
16:06:05 * RemiFedora here
16:06:08 <abadger1999> #chair geppetto RemiFedora
16:06:08 <zodbot> Current chairs: RemiFedora abadger1999 geppetto
16:06:46 <abadger1999> spot, SmootherFrOgZ, tibbs: FPC ping
16:07:05 * SmootherFrOgZ is  here
16:07:28 <abadger1999> #chair SmootherFrOgZ
16:07:28 <zodbot> Current chairs: RemiFedora SmootherFrOgZ abadger1999 geppetto
16:07:45 <SmootherFrOgZ> one more to quorum
16:09:12 <abadger1999> #topic #339     software collections in Fedora
16:09:17 <abadger1999> https://fedorahosted.org/fpc/ticket/339
16:09:30 <abadger1999> I've started work on adding the scldevel section.
16:10:13 <abadger1999> https://fedoraproject.org/wiki/User:Toshio/SCL_Guidelines_(draft)#SCL_scldevel_subpackages
16:11:09 <abadger1999> Needs some work as the language about why scldevel should be used is unclear.
16:11:23 <geppetto> I'm not sure that can be fixed :)
16:12:26 <abadger1999> And after jotting it all down I remembered that we had come to the conclusion that we were going to mandate that no General SCL Package could require packages from another SCL.  Instead they needed to be placed in their own SCL.
16:12:53 <abadger1999> So I need to figure out how to integrate that (probably needs to have another new section that the scldevel section links to).
16:13:29 <abadger1999> Rathann: Welcome!
16:13:34 <abadger1999> #chair Rathann
16:13:34 <zodbot> Current chairs: Rathann RemiFedora SmootherFrOgZ abadger1999 geppetto
16:13:36 <Rathann> sorry guys, got tied up doing some shopping
16:14:28 <abadger1999> I'm asking mattdm about whether the intrascl naming thing needs to be addressed before the rails scl can be approved or not.
16:14:45 <abadger1999> He's looking at it and thinking about it.
16:15:41 <abadger1999> RemiFedora: I got your email about further scldevel thoughts but didn't quite understand it.  Want to explain that to me on IRC after the meeting or tomorrow?
16:15:47 <abadger1999> (Or open floor if we have time)
16:16:01 <RemiFedora> yes, later (and I'd like to get slavek feedback)
16:16:06 <abadger1999> k
16:16:28 <abadger1999> RemiFedora: you can also read what I've written in the scldevel section and let me know if I'm off on the wrong track ;-)
16:17:30 <abadger1999> #topic #382     Go Packaging Guidelines Draft
16:17:35 <abadger1999> https://fedorahosted.org/fpc/ticket/382
16:17:39 <abadger1999> Still no reply from vbatts.
16:18:41 <abadger1999> I'm tempted to draft up something that would require rebuilds whenever the Go compiler is updated since I tend to agree with geppetto's rationale as to why that's preferred.
16:19:00 <abadger1999> But -ENOTIME at least as long as I'm working over the SCL stuff.
16:19:23 <abadger1999> If anyone else would like to do that or figure out a better Go solution...?
16:19:47 <abadger1999> #topic #414     Please consider requiring AppData for all desktop
16:19:51 <abadger1999> https://fedorahosted.org/fpc/ticket/414
16:21:21 <abadger1999> So.... It seems like we had agreement in the ticket.
16:21:24 <abadger1999> Where does that leave us?
16:24:03 <Rathann> https://fedoraproject.org/wiki/User:Rhughes/DraftAppDataGuidelines is a bit inconsistent
16:24:11 <geppetto> I think https://fedoraproject.org/wiki/User:Rhughes/DraftAppDataGuidelines has all the changes we asked for, right?
16:24:25 <Rathann> at the beginning, it says: ... must correctly validate using appdata-validate.
16:24:32 <Rathann> but then at the end:
16:24:43 <Rathann> you SHOULD run appdata-validate (in %check or %install)
16:24:54 <Rathann> hm... on second thought, it's not contradictory
16:25:40 * abadger1999 would be okay with changing the second SHOULD to a MUST
16:26:07 <abadger1999> tibbs|w: Greetings, you here for the meeting?
16:26:09 <geppetto> abadger1999: didn't we ask him to change that the other way though?
16:26:12 <RemiFedora> I'd really like to see an example "with" translation
16:26:19 <abadger1999> Hmm..
16:26:21 <tibbs|w> Sorry, folks; Thursday has been bad for me lately.
16:26:23 * abadger1999 checks ticket comments
16:26:29 <geppetto> abadger1999: I thought a concern we had was if changes to the validation stopped rebuilds.
16:26:43 <RemiFedora> appdata without translation doesn't make any sense imho
16:26:45 <Rathann> any idea about the timeline for this? "SPDX is planning on adding IDs for all Fedora Free licenses, so this should not be an issue. "
16:26:47 <abadger1999> geppetto: I think that was in relation to license files.
16:26:50 <abadger1999> err
16:26:55 <Rathann> https://fedorahosted.org/fpc/ticket/414#comment:18
16:26:56 <abadger1999> license tags of the metadata
16:27:04 * geppetto nods
16:27:07 <Rathann> spot: ^
16:27:35 <geppetto> I think I'm +1 on it as-is.
16:28:03 <abadger1999> geppetto: We didn't want a license tag to stop rebuilds as we don't have permission to change the license, only upstream does.
16:28:11 * geppetto nods … yeh
16:28:53 <geppetto> If you really want to change the second should to a must, I'll go along with it … but I'm on sick leave today, so you might be bullying me into making bad choices ;p
16:29:01 <abadger1999> heh.
16:29:54 <Rathann> also, s/desktop-file-install usage/appdata-validate usage/
16:29:58 <abadger1999> I think we probably need to know the status of SPDX adding fedora licenses anyway... otherwise we could have to omit AppData files that are okay by Fedora but not in the SPDX yet.
16:31:13 <Rathann> with that, I'm +1, assuming the SPDX inclusion of Fedora licences is resolved
16:32:45 <abadger1999> Okay, desktop-files-validate is a MUST so I'll change the app-data-calidate to a MUST in the draft
16:33:44 <abadger1999> I'll update the ticket to ask about SPDX timeline and explain our rationale on the timeline.
16:34:01 <abadger1999> hopefully we can vote on this next week
16:34:06 <abadger1999> #chair tibbs|w
16:34:06 <zodbot> Current chairs: Rathann RemiFedora SmootherFrOgZ abadger1999 geppetto tibbs|w
16:34:26 <RemiFedora> abadger1999, about the translation of the summary/description ?
16:36:17 <abadger1999> RemiFedora: ah... Right.  What's the translation issue?
16:36:32 <RemiFedora> <RemiFedora> I'd really like to see an example "with" translation
16:36:46 <abadger1999> RemiFedora: Just want the example appdata file to contain a sample translation?
16:37:06 <RemiFedora> <RemiFedora> appdata without translation doesn't make any sense imho
16:37:27 <RemiFedora> abadger1999, yes, because it just a nightmare to find "how" to add translation
16:37:31 <abadger1999> <nod>
16:37:40 <RemiFedora> (old discussion with richard on devel @)
16:40:10 <abadger1999> Okay, ticket updated.
16:40:16 <RemiFedora> here is one with translation http://svnweb.tuxfamily.org/filedetails.php?repname=qet%2Fqet&path=%2Ftrunk%2Fmisc%2Fqelectrotech.appdata.xml
16:40:19 <abadger1999> #topic #419     ruby193 in SCL
16:40:24 <abadger1999> https://fedorahosted.org/fpc/ticket/419
16:40:33 <abadger1999> langdon: Anything from you on naming?
16:40:53 * abadger1999 notes that the question I asked mattdm for input on also bears on this... in particular the rails case.
16:44:04 * abadger1999 adds question about whether ruby1.9.3 and v8 scls need to be added to the rails "packages in platform"
16:44:59 <abadger1999> Okay... if no langdon I guess we'll move on for now.
16:45:12 <abadger1999> #topic #435     %py3dir not removed by rpmbuild --clean
16:45:15 <abadger1999> https://fedorahosted.org/fpc/ticket/435
16:45:25 * abadger1999 looks if there's an update in the bugzilla
16:46:25 <abadger1999> No update in bugzilla.
16:46:33 <abadger1999> I'll ping it.
16:46:41 <abadger1999> #topic #436     Bundled code advise for shiboken
16:46:42 <langdon> im here... sorry .. was afk for a minutr
16:46:46 <abadger1999> https://fedorahosted.org/fpc/ticket/436
16:46:49 <abadger1999> Ah..
16:46:55 <abadger1999> #topic #419     ruby193 in SCL
16:47:19 <abadger1999> langdon: Okay, any updates to the naming question? (Or other scl-related topics)
16:48:04 <langdon> well.. after all that, i think no.. the discussion drags on
16:48:10 <abadger1999> Okay.
16:48:27 <abadger1999> anything else or should I move on?
16:48:49 <langdon> i think you should move on, Ill try to come to resolution by next meeting
16:48:56 <abadger1999> Okee dokee
16:49:06 <abadger1999> #topic #436     Bundled code advise for shiboken
16:49:11 <abadger1999> https://fedorahosted.org/fpc/ticket/436
16:49:55 <abadger1999> I see that limb voted in ticket so this has passed.
16:50:08 * abadger1999 will close it out after the meeting
16:50:16 <abadger1999> #topic #437     add to packaging:guidelines section 'AppStream files'
16:50:22 <abadger1999> https://fedorahosted.org/fpc/ticket/437
16:51:36 <abadger1999> I'm.... not sure I see the point of this guideline
16:52:15 <abadger1999> If we require maintainers to run appstream-util  validate on the file but also don't error out if the file doesn't validate.... it seems like an overall null effect.
16:53:20 <Rathann> also, it looks like a superset of the appdata guideline
16:54:31 <abadger1999> yeah... maybe placing it in a subsection of the appdata guidelines?
16:54:35 <geppetto> yeh, I think it was like a patch to the other guildines
16:54:40 <abadger1999> The ticket doesn't make it clear what the relationship is.
16:54:54 <abadger1999> (although the email message is somewhat more informative)
16:55:04 <geppetto> I assume the person didn't know the other guidelines were still going on, or something.
16:55:29 <abadger1999> Or that they existed at all..... he refers to the desktop-utils guideline but not the appdata guideline
16:57:14 <abadger1999> Okay -- so I'll ask for a draft that's a patch to or subsection of the appdata proposal.
16:57:49 <abadger1999> And to either choose to make running validdate a SHOULD or make it cause failure to the build (and note that existing validate's go the failure route)
16:58:09 <abadger1999> #topic #438     Please recommend usage of PrivateDevices= and  PrivateNetwork= for long-running system services
16:58:17 <abadger1999> https://fedorahosted.org/fpc/ticket/438
16:58:53 * zbyszek is here
17:00:52 <tibbs|w> Where would this go?  It's kind of long for a guideline.
17:01:19 * Rathann guesses: in systemd guidelines
17:01:23 <abadger1999> Yeah
17:01:39 <Rathann> seems well written to me
17:02:06 <Rathann> but a bit too verbose
17:02:24 <Rathann> maybe cut down on the details and provide direct links to documentation?
17:03:30 <geppetto> I'm concerned about the performance inpact of privatenetwork
17:03:52 <Rathann> geppetto: oh? please elaborate
17:04:11 <Rathann> why would there be one?
17:04:37 <geppetto> Rathann: AIUI it's using more than just name mapping … you actaully bouce all your data through a local interface?
17:05:25 <geppetto> It also seems much more likely to be enabled and mostly work, but break something (Eg. the NSS/LDAP stuff).
17:06:23 <geppetto> In contrast the privatedevices thing seems like it'd be almost free, and very unlikely to break anything.
17:06:26 <Rathann> geppetto: the guideline explicitly says that if you service requires non-files naming services then don't enable it
17:06:57 <Rathann> "Services which might need to resolve non-system user and group names. Since on some setups resolving non-system users might require network access to an LDAP or NIS server, enabling this option on might break resolving of these user names. "
17:07:07 <abadger1999> Hmm... then perhaps we want this to be a Fedora 21/22 Change and we add the Guidelines to do it as part of that Change?
17:07:25 <Rathann> abadger1999: that's mentioned at the top and at the bottom
17:07:27 <geppetto> Rathann: yeh, but if we are recommending it people are going to enable it without doing all the testing needed (not many home users need/use LDAP or NSS, for example).
17:07:50 <Rathann> "This is in-line with the accepted F21 feature posted here (which includes a rationale why): "
17:08:00 <Rathann> "(Oh, and please only recommend this for Fedora 21 and newer, before that the two options didn't work the way described above, or didn't exist.) "
17:08:10 <geppetto> So saying that … I'm very happy tot +1 recommending privatedevices … and leave the privatenetwork stuff to sysadmins, or packagers who really know they want it by default.
17:08:52 <abadger1999> k... If it's a Fedora Change... then I suppose I'm +1 to both changes.
17:08:55 <Rathann> I'm +1 to both parts
17:09:04 <RemiFedora> also +1
17:09:45 <abadger1999> fesco should probably weigh on the Change in if there's a performance vs security question.
17:10:54 <tibbs|w> I'm certainly +1 to telling packagers they should consider turning these on.
17:11:10 <SmootherFrOgZ> +1 here
17:12:04 <abadger1999> #info Adding privatedevice recommendation to systemd guidelines PASSED: (+1: 6, 0:0, -1:0)
17:12:15 <abadger1999> #info Adding privatenetwork recommendation to systemd guidelines PASSED: (+1: 5, 0:0, -1:1)
17:13:06 <abadger1999> If someone wants to work on making it shorter/more concise, I think they could work on that as an update to the guideline?
17:13:20 <abadger1999> (at antoher meeting)
17:14:58 <abadger1999> geppetto: Note -- there's an FPC "bylaw" that if you're concerned abouta guideline, it can be brought to fesco to discuss and approve... want me to bring the privatenetwork thing to them under that?  (Or do you want to file a fesco ticket about the privatedevice change on your own)?
17:15:13 <abadger1999> (where your == any FPC member)
17:15:27 <geppetto> abadger1999: you can do it, if you don't mind.
17:15:49 <abadger1999> will do.  I may not be able to aruge it effectiveely in the meeting but I'll file the ticket and get it started.
17:15:52 <abadger1999> #chair limburgher
17:15:52 <zodbot> Current chairs: Rathann RemiFedora SmootherFrOgZ abadger1999 geppetto limburgher tibbs|w
17:15:57 <abadger1999> Hey limburgher :-)
17:16:07 <abadger1999> #topic #439     update for Packaging:Tmpfiles.d
17:16:11 <abadger1999> https://fedorahosted.org/fpc/ticket/439
17:16:24 <tibbs|w> Sorry, folks, my burglar alarm at home was set off so I have to leave.
17:16:29 <tibbs|w> Crappy day so far.
17:17:35 <abadger1999> :-(
17:17:40 <abadger1999> tibbs: Good luck
17:17:52 * limburgher waves
17:18:03 <abadger1999> Here's the current tmpfiles.d guideline: https://fedoraproject.org/wiki/Packaging:Tmpfiles.d
17:21:23 <zbyszek> I forgot to mention it in the ticket, but currently 78 packages seem to use /etc/tmpfiles.d/ instead of /usr/lib/tmpfiles.d.
17:21:25 * abadger1999 wonders if this is the place where we had a whole huge discussion about the effects of %ghost and %config here.
17:24:25 <geppetto> not sure
17:25:22 <abadger1999> ah... and this is a directory, not a file.
17:25:30 <abadger1999> that vaguely jogs my memory too.
17:25:47 <abadger1999> Maybe it's that files are fine to %ghost but directories aren't.
17:25:58 <geppetto> But, yeh, surely the package should ghost the /etc override?
17:26:11 <Rathann> wait, what's a directory?
17:26:12 <zbyszek> geppetto: no, of course not
17:26:17 <geppetto> zbyszek: why?
17:26:37 <zbyszek> Because its an admin generated file, manually crafted.
17:26:46 <geppetto> zbyszek: if the sysadmin puts a file there, you want rpm -qf to show which file it belongs to.
17:27:02 <geppetto> zbyszek: yeh, it still belongs to that file though … it's just generated by the admin.
17:28:06 <zbyszek> But following this rule, there should also be a %ghost for /run/
17:28:53 <Rathann> geppetto: /etc/resolv.conf is unowned, for example
17:29:07 <zbyszek> I always thought that %ghost is only for (semi-)automatically generated files... that it helps answer the question "where did this file come from?"
17:29:15 <abadger1999> Hmmmmmmmmm
17:29:37 <abadger1999> geppetto: I think that's getting into a different idea.... which I have to admit I've been uncomfortable about a lot recently.
17:32:15 <abadger1999> geppetto: Files under /etc/ are config files.  To some people that's a categorization "config files" and to others it's only a definition of OS requirements that hte file needs "unsharable, static"
17:32:48 <geppetto> Rathann: That's always annoyed me
17:33:11 <geppetto> Rathann: And I assumed at least some of the history there was Ulrich
17:33:50 * geppetto shrugs … if nobody else thinks we should ghost I'll let it go. Change seems fine to me, otherwise.
17:34:20 <abadger1999> geppetto: so now that several big, important daemons consider their defaults to not be "config", there's no files in /etc/ for sysadmins to base off of.  they have to know to look in /usr/share, copy files to some other directory in /etc/ and modify those there.
17:34:40 * geppetto nods
17:34:58 <zbyszek> abadger1999: it's easy enough if they all follow the same pattern
17:34:59 <abadger1999> I'm not sure I like that trend... but it seems to be separate from %ghost of /var/run/files
17:34:59 <Rathann> if you want to %ghost it, then it should also be marked as %config
17:35:06 <Rathann> not sure if that combination works
17:35:25 <geppetto> It's crappy … but what else can you do if you want a solution but want to ignore the packaging system as much as possible.
17:36:23 <geppetto> I'm just looking at it from the point of view of someone looking in /etc and seeing a random file and thinking "wtf does this belong to" and running rpm -qf … and getting no usable output.
17:36:47 <abadger1999> zbyszek: So... looking at the change... I think we could instead say "Files (but '''not''' directories) that the program places directly into /run or into its subdirectories may be listed in the %files section as %ghost but you may also omit them entirely as the files will be cleaned up on every reboot. "
17:37:48 <abadger1999> geppetto: Yeah.... question -- do a lot of these files have numbers in their copied name to do ordering?
17:37:55 <zbyszek> Would that mean that they would not be created immediately when the package is installed?
17:38:07 <abadger1999> geppetto: if so, those probably don't work with the %ghost strategy
17:38:37 <geppetto> abadger1999: I assume they have to be named identically to whatever the name is in /usr/lib/tmpfies.d
17:38:58 <abadger1999> zbyszek: yes they should be created when the program runs (argubly a bug if the program can't create a local state *file*)
17:39:10 <zbyszek> abadger1999: so, yeah, that causes problems.
17:39:39 <abadger1999> zbyszek: Explain?
17:39:47 <zbyszek> abadger1999: Basically, the choice is either %post %tmpfiles_create, or no %ghost, and everything, even in /run, installed in package
17:40:13 <abadger1999> zbyszek: Point to a bug report that backs that up please, so I can look at it.
17:40:18 <zbyszek> moment
17:40:26 <geppetto> rdieter: what is your comment #2 referring to … that is an alternative to the tmpfile_create post?
17:40:49 <geppetto> abadger1999: we might be talking about the same thing.
17:40:57 <zbyszek> yeah, the screen bug is good example https://bugzilla.redhat.com/show_bug.cgi?id=1112016
17:40:57 <abadger1999> yeah.
17:41:08 <abadger1999> zbyszek: that one is a directory.
17:41:35 <abadger1999> zbyszek: So that should be fixed by clarifying the guidelines to make clear that %ghost cannot be used with directories.
17:42:22 <zbyszek> I see your point.
17:43:10 <zbyszek> But OTOH, I don't see why %ghost files if they will be created by tmpfiles later on.
17:43:10 <abadger1999> Cool.
17:44:17 <zbyszek> Just seems simpler and less error prone to use the same 'formula' for everything.
17:45:58 <abadger1999> So... for a file, we need the file to be created when the package is installed.  But the file's contents will change as the program runs so we do not want rpm to track the contents.
17:46:21 <abadger1999> So %ghost + a %post script to create the file seems correct.
17:46:27 <abadger1999> However, all it needs is
17:46:30 <abadger1999> %post
17:46:38 <abadger1999> touch %{_localstatedir}/file
17:46:58 <abadger1999> No need for a tmpfilesd provided macro
17:47:52 <zbyszek> It depends how many such files you have. If one, sure, if more, than it becomes annoying to synchronize the scriptlets, especially if you have some permissions attached to those files.
17:49:11 <abadger1999> Hmm... I'm not sure if we can omit those either.
17:49:18 <abadger1999> files in a subdirectory, we can omit
17:49:37 <abadger1999> but files directly in /var/run probably need to be pre-created.
17:50:30 <abadger1999> (The difference is that ownership of the directory being written to can be controlled in the subdir case but not in the direct-to-/run case)
17:55:36 <limburgher> Sorry, folks, something came up and I have to go.  Ping me directly if you need my vote in a ticket. . .
17:55:54 <abadger1999> tangent -- tmpfilesdir is available on rhel7 so changes related to that should be okay.
17:57:28 * Rathann notes we're almost at two hour mark
17:57:43 <abadger1999> So.... the draft portion about %ghosting and files is wrong.
17:58:52 <abadger1999> I think that we assumed a bit too much about what people already knew (or could find elsewhere in the guidelines) about using %ghost.
17:59:25 <abadger1999> To someone not familiar with %ghost, they'll likely do the wrong thing with the current guideline in that section aa well.
17:59:47 <abadger1999> Let me work on clarifying that section and we can vote next week.
17:59:57 <abadger1999> #topic Open Floor
18:00:03 <abadger1999> Anyone have anything to bring up?
18:00:15 * SmootherFrOgZ has to leave. will be back in about 30min.
18:00:36 <SmootherFrOgZ> Oh open floor, that should be quick :)
18:00:39 <Rathann> I seem to have stumbled across a missing guideline for pkgconfig files in packages using MPI
18:00:49 <Rathann> but this is last minute, so no draft yet
18:01:14 * RemiFedora won't assist to next week meeting, Holidays \o/ !
18:01:15 <Rathann> also, I hope to coax smani into writing it ;)
18:02:09 <Rathann> ok, I need to leave now
18:02:39 <Rathann> thanks guys
18:03:07 <abadger1999> Alright.
18:03:15 <abadger1999> So long Everyone!
18:03:21 <abadger1999> #endmeeting