17:01:36 <geppetto> #startmeeting fpc
17:01:36 <zodbot> Meeting started Thu Jan 27 17:01:36 2022 UTC.
17:01:36 <zodbot> This meeting is logged and archived in a public location.
17:01:36 <zodbot> The chair is geppetto. Information about MeetBot at https://fedoraproject.org/wiki/Zodbot#Meeting_Functions.
17:01:36 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
17:01:36 <zodbot> The meeting name has been set to 'fpc'
17:01:37 <geppetto> #meetingname fpc
17:01:37 <geppetto> #topic Roll Call
17:01:37 <zodbot> The meeting name has been set to 'fpc'
17:01:50 * GwynCieslasheher here
17:02:06 <carlwgeorge> .hi
17:02:07 <zodbot> carlwgeorge: carlwgeorge 'Carl George' <carl@redhat.com>
17:02:10 <geppetto> #chair GwynCieslasheher
17:02:10 <zodbot> Current chairs: GwynCieslasheher geppetto
17:02:15 <geppetto> #chair carlwgeorge
17:02:15 <zodbot> Current chairs: GwynCieslasheher carlwgeorge geppetto
17:03:15 <tibbs> Hey.
17:04:05 <geppetto> #chair tibbs
17:04:05 <zodbot> Current chairs: GwynCieslasheher carlwgeorge geppetto tibbs
17:04:23 <decathorpe> .hello2
17:04:24 <zodbot> decathorpe: decathorpe 'Fabio Valentini' <decathorpe@gmail.com>
17:06:19 <geppetto> #chair decathorpe
17:06:19 <zodbot> Current chairs: GwynCieslasheher carlwgeorge decathorpe geppetto tibbs
17:11:27 <geppetto> #topic Schedule
17:11:30 <geppetto> #link https://lists.fedoraproject.org/archives/list/packaging@lists.fedoraproject.org/message/QFDIKXHIYBBA5UI7JT5SPVZILD2CXMRA/
17:12:00 <geppetto> #topic #pr-1152 Add BLAS/LAPACK guidelines
17:12:00 <geppetto> https://pagure.io/packaging-committee/pull-request/1152
17:12:26 <geppetto> There are a lot of +1's here already
17:12:47 <GwynCieslasheher> I missed t, I'll +1 there
17:13:07 <geppetto> I think we can just merge it now, unless anyone has any objections or requested changes?
17:13:13 <GwynCieslasheher> Ship it
17:13:54 <geppetto> Ok, I pressed the button
17:14:03 <decathorpe> yeah I don't think I plussed it yet, but it makes sense to have guidelines for this
17:14:39 <GwynCieslasheher> Agreed, it' sone of those things that seem to come up every few years and it's important to do it right and consistently across the board as much as possible.
17:14:58 <geppetto> #topic #1150 request for clarification wrt. base version / compat package naming
17:14:59 <geppetto> .fpc 1150
17:14:59 <geppetto> https://pagure.io/packaging-committee/issue/1150
17:15:00 <zodbot> geppetto: Issue #1150: request for clarification wrt. base version / compat package naming - packaging-committee - Pagure.io - https://pagure.io/packaging-committee/issue/1150
17:15:19 <GwynCieslasheher> Linear Algebra: It's Everybody's Problem.
17:16:05 <geppetto> We talked about this last week, but had a couple of comments
17:16:26 <geppetto> I think the golang problem could be fixed by just changing the macros
17:16:41 <geppetto> Although migration will be a problem, I guess
17:17:02 <decathorpe> geppetto: not really, as upstreams are identified by those paths anyway
17:17:50 <decathorpe> i.e. something like "use github.com/foo/bar/v2" or something like that ... I don't think it would make sense to break expectations by doing this differently on the RPM side for no good reason
17:18:22 <carlwgeorge> meant to reply to this one, but my response the to "but what about libsoup3 on a stable branch" is that should be discouraged by way of it being annoying/difficult (but not impossible)
17:18:30 <geppetto> eh, I'm not sure we need to keep the devel github urls in sync. with rpm package names
17:19:13 <decathorpe> but the github URLs are how the project is identified
17:19:20 <decathorpe> there's literally no other way to do that
17:20:52 <geppetto> decathorpe: Yeh, I'm just not sure that needs to leak into Fedora package names
17:21:33 <geppetto> decathorpe: I'd assume most golang devs. could work out what syncthing1 and syncthing2 mean.
17:22:28 <geppetto> But then the opinion from last week was that it would be a difficult project to unify this across the distro. … so ¯\_(ツ)_/¯
17:22:45 <decathorpe> yeah, I don't think there's a one-size-fits-all solution here
17:23:21 <GwynCieslasheher> Ultimately different upstreams do different things, so all we can really attempt is to acheive as much consistency as possible while not concentrating the pain all in one spot.
17:23:35 <GwynCieslasheher> s/acheive/achieve/
17:24:55 <geppetto> i before e, except when not ;)
17:24:57 <decathorpe> I think something like this would make sense: "In general, the package without the version suffix SHOULD be the one with the latest version. If there are more specific naming guidelines for your language or ecosystem, apply those instead."
17:25:32 <GwynCieslasheher> Fabio Valentini That could work.
17:25:52 <GwynCieslasheher> geppetto English is the best language, exect for all the others.
17:26:05 <GwynCieslasheher> s/exect/except/
17:26:10 <tibbs> In the end it has to be up to the packager, but it doesn't hurt to state that at least the common case is unversioned = "latest".
17:26:22 * geppetto nods
17:26:31 <tibbs> Of course, "latest" doesn't usually mean "git head".
17:26:35 <GwynCieslasheher> yep
17:26:38 <geppetto> decathorpe: You want to propose an MR with that somewhere?
17:26:52 <geppetto> latest released?
17:27:04 <decathorpe> latest packaged?
17:27:43 <decathorpe> I can try to add something like that to the guidelines .. though I'm not sure where would be the best place for it.
17:27:44 <geppetto> packaged seems like a confusing thing to say in this context
17:28:02 <tibbs> If I want to package the latest stable release and a development build, I'd make the unersioned one the stable version.
17:28:29 <carlwgeorge> latest upstream version/snapshot
17:28:44 <GwynCieslasheher> I was thnking so. Maybe released, it's not like we're going to have libsoup3 and libsoup both following git HEADs....I hope,....
17:28:47 <decathorpe> tibbs: See, this is where things are getting complicated again :)
17:29:16 <tibbs> That's precisely why this is left out of the current guidelines.
17:29:32 <tibbs> It's ultimately up to the packager.
17:29:48 <GwynCieslasheher> Maybe. We won't hit perfect but we can have least worst.
17:30:33 <geppetto> Yeh, I think we are aiming for "If you aren't sure what to do, try this." and less "This is what you should do."
17:30:58 <geppetto> So something will hopefully be better than nothing
17:31:05 <carlwgeorge> instead of just have several apparently equal SHOULDs, can we phrase it to prioritize them?
17:31:23 <GwynCieslasheher> "this way will hurt least"
17:36:03 <decathorpe> I can make a flowchart :)
17:37:24 <geppetto> decathorpe++
17:37:24 <zodbot> geppetto: Karma for decathorpe changed to 4 (for the current release cycle):  https://badges.fedoraproject.org/tags/cookie/any
17:37:41 <geppetto> Maybe a couple of examples will be better than a paragraph of text?
17:37:54 <geppetto> But I'm +1 on a flowchart
17:38:22 <GwynCieslasheher> Me too
17:38:41 <decathorpe> (flowchart was supposed to be a joke, but if you want one ...)
17:39:13 <geppetto> decathorpe: I know :)
17:39:23 <carlwgeorge[m]> seems there is still some lag on the matrix/irc bridge
17:39:24 <carlwgeorge[m]> instead of just have several apparently equal SHOULDs, can we phrase it to prioritize them?
17:39:25 <decathorpe> I was more thinking along the lines of "here's a list of things to consider, and if any of those match, you know what to do"
17:40:06 <geppetto> decathorpe: yeh, seems good
17:40:20 <decathorpe> coincidentally I think the way we name compat packages is broken, but I don't want to open that particular can of worms yet ...
17:40:31 <decathorpe> I'll try to work out something.
17:40:34 <geppetto> carlwgeorge[m]: I saw that … I'm less sure how to prioritize
17:40:56 <geppetto> #action decathorpe will solve everything.
17:41:10 <geppetto> #topic Open Floor
17:41:49 <geppetto> Nothing else, tagged as meeting, is new or has changed since last week
17:41:50 <decathorpe> geppetto: That's a tall order, but I'll manage.
17:42:03 <geppetto> So … anything anyone wants to talk about?
17:42:35 <decathorpe> Not me.
17:42:47 <tibbs> How about just saying we leave it to the packager's best judgement but if the best option isn't clear, then do X.
17:43:01 <geppetto> decathorpe: I've got my official "solved" stamp ready, whatever you do will get stamped ;)
17:43:04 <carlwgeorge> i had one
17:43:14 <geppetto> carlwgeorge: Go for it
17:43:15 <tibbs> Sorry I'm slow.
17:43:27 <carlwgeorge> i thought this was already covered in guidelines, but i can only find it for libraries https://docs.fedoraproject.org/en-US/packaging-guidelines/Python/#_explicit_lists
17:43:50 <carlwgeorge> "don't over-glob", e.g. `%{_bindir}/*`
17:44:11 <carlwgeorge> anyone have thoughts on that before i submit a pr to make that generic, vs just for libs?
17:44:58 <tibbs> I would generally be against doing anything other than suggesting that packagers be careful with globbing.
17:44:59 <GwynCieslasheher> No.
17:44:59 <carlwgeorge> ah, wrong link https://docs.fedoraproject.org/en-US/packaging-guidelines/#_listing_shared_library_files
17:45:22 <tibbs> The library thing is due to that being a constant source of breakage.
17:45:26 <carlwgeorge> the python link is the example i want to follow
17:45:28 <decathorpe> I think that's a good rule of thumb. Don't use globs for anything where it can cause problems if packages add or remove stuff.
17:45:37 <carlwgeorge> > Packagers SHOULD NOT simply glob everything under a shared directory.
17:45:41 <tibbs> Is there a constant source of breakage due to new binaries appearing?
17:46:10 <carlwgeorge> my point is this is already in the python guidelines, but i would like to make it generic for all packages
17:46:30 <tibbs> This makes packaging more complex in some cases so it would need to have an actually perceptible gain instead of just making people list more things in %files.
17:47:00 <geppetto> carlwgeorge: How do you want to make it more generic?
17:47:12 <decathorpe> tibbs: I know literally zero packages which use globs like this correctly.
17:47:26 <carlwgeorge> copy the python guideline on this topic and put it in the main guidelines, removing the python specific examples
17:47:27 <geppetto> Like the shared lib. thing is w/e … but it's common to do the same looking thing for man pages.
17:47:31 <decathorpe> so I'm all for saying "don't do that"
17:47:57 <carlwgeorge> tibbs: it prevents new upstream commands that conflict with other packages being shipped unknowingly
17:48:02 <tibbs> I guess it depends on how you define "correctly".
17:48:43 <tibbs> And how often does that happen in practice, as opposed to the library thing which caused unknown ABI bumps all the time?
17:49:16 <decathorpe> would you consider using "*" as the only thing in %files a good thing? because that's the next step down the slippery slope of "how often can this cause new problems" :)
17:49:39 <decathorpe> (I've seen packagers do that, btw.)
17:49:48 <tibbs> That generally runs afoul of rules on directory ownership.
17:50:12 <geppetto> decathorpe: To be fair I sympathize, and kind of wish we could fix it so that would work.
17:50:18 <decathorpe> well who's checking that outside of fedora-review? nobody ...
17:50:38 <geppetto> Eg. have rpm know that random packages shouldn't own /usr/bin or whatever
17:51:36 <geppetto> I have no problem having carlwgeorge make more generic "don't use globs in these cases, because it cases problems" … but less so if it's just more pain for possible problems.
17:52:53 <carlwgeorge> and like the python one, it'll be a SHOULD NOT, so packages can say "i know better in this case" as needed
17:53:05 <geppetto> Personally I much prefer globs over having %if blocks in the %files list.
17:53:16 * geppetto nods
17:53:42 <geppetto> I'm probably fine with what you want to propose, I'm just old grumpy and like globs ;)
17:53:57 <carlwgeorge> i'll work up a pr and wait for comments
17:54:02 * geppetto nods
17:55:13 <geppetto> Anything else anyone wants to talk about?
17:56:01 <GwynCieslasheher> Not I said the duck.
17:57:11 <geppetto> Ok, see you next week.
17:57:15 <geppetto> #endmeeting