From Fedora Project Wiki

Fedora Packaging Committee Meeting of {2008-03-11}


  • DominikMierzejewski (Rathann|work)
  • HansdeGoede (hansg)
  • JasonTibbitts (tibbs)
  • RalfCorsepius (racor)
  • RexDieter (rdieter)
  • TomCallaway (spot)
  • ToshioKuratomi (abadger1999)
  • VilleSkyttä (scop)


The following proposals were considered:

  • Ban unicode in package names (no draft submitted)
  • Not accepted
  • Voting for: racor rdieter
  • Voting against: abadger1999
  • Abstaining: tibbs

Next Meeting

Unless the time is changed to adjust for DST or the schedules of the new committee members, the next meeting will be March 25 at 17:00 UTC in

  1. fedora-meeting.

IRC Logs

[12:05]  <hansg> Hi all, I don't know how involved in any discussions I'll be as I'm home alone with my 2 (awake) daughters of 10 months resp. 4 years old.
[12:06]  <scop> unfortunately I'll have to leave in about 10 minutes today
[12:06]  * rwmjones just wants to discuss any issues with
[12:07]  <hansg> rwmjones, I've taken a good look at it this week, and it looks fine to me
[12:07]  <-- svahl has left this channel.
[12:07]  <hansg> But don't  shouldn't we have something like a schedule and someone leading the meeting (making sure we sortof stick to the schedule)
[12:08]  <rdieter> spot: ping, around to lead FPC meeting?
[12:09]  <hansg> Does anyone have the right to set the topic here?
[12:09]  <tibbs> Yes.
[12:09]  <tibbs> The running agenda is
[12:10]  <spot> sorry, lunch ran late
[12:10]  * spot is here now
[12:12]  <spot> who else is still around? abadger1999, racor, hansg, rdieter, Rathann|work, scop?
[12:12]  * abadger1999 half here.  Also busy with FAS2 migration.
[12:12]  <rdieter> here
[12:12]  * hansg is here (sortof also babysitting at the same time :)
[12:12]  <racor> here
[12:12]  <scop> still around, but will need to leave in just a couple of minutes, so count me out
[12:13]  <tibbs> Can we get opinions on the ocaml changes really quick?
[12:13]  <rwmjones> any questions, just ask me
[12:13]  * spot is still reading the ocaml draft
[12:13]  <hansg> ocaml gets +! from me
[12:13]  <tibbs>
[12:13]  <hansg> make that +1
[12:14]  <tibbs> I wish the draft indicated where the changes are.
[12:14]  <spot> you're disabling debuginfo
[12:14]  <spot> is that really what we want to do?
[12:14]  <tibbs> You have to disable debuginfo for ocaml.
[12:14]  <tibbs> Otherwise it just comes out empty.
[12:15]  <rwmjones> debuginfo doesn't really contain anything useful for ocaml programs
[12:15]  <rwmjones> you _can_ debug them under gdb
[12:15]  <rwmjones> but gdb doesn't know the language so you end up having to debug assembler
[12:15]  <rwmjones> (if I've understood the purpose of debuginfo, that is)
[12:15]  <scop> "Binaries should be stripped, as per ordinary Fedora packaging guidelines." is misleading
[12:15]  <tibbs> Not to mention that the compiler doesn't put in any symbols that you could strip out.
[12:15]  * Rathann|work is here
[12:15]  <abadger1999>
[12:16]  <rwmjones> scop I didn't understand that -- I just meant that ordinary Fedora policy about stripping binaries should apply, except in that one special circumstance (which in practice affects precisely two binaries at the moment)
[12:17]  <scop> ok
[12:17]  <spot> +1 from me
[12:17]  <Rathann|work> rwmjones: as far as I know this stripping bug is being addressed upstream
[12:17]  <rwmjones> Rathann|work, afaiaa upstream rejected it?
[12:18]  <rwmjones> if you follow that debian bug report you'll see some messages from upstream on the subject
[12:18]  <Rathann|work> yes
[12:18]  <tibbs> If in the future is addressed upstream then we can update the guidelines to indicate releases where it is no longer necessary to avoid stripping.
[12:18]  <scop> sorry, I need to go now, and am not familiar enough with the draft to vote
[12:18]  <Rathann|work> > Rather than muck with ELF, a simpler solution would be to embed the
[12:18]  <Rathann|work> > bytecode executable as initialized C arrays in the executable
[12:18]  <Rathann|work> > generated by ocamlc -custom.  That's what ocamlc -output-obj does, and
[12:18]  <Rathann|work> > I believe it shouldn't be too hard to adapt the existing -output-obj
[12:18]  <Rathann|work> > code to the -custom case.
[12:18]  <Rathann|work> Mmm. Ok, it will be ocaml 3.09 stuff though.
[12:18]  <rwmjones> ah ok, upstream did a partial fix in ocaml 3.09
[12:18]  <rwmjones> but they didn't fix the ocamlc -custom case
[12:18]  <rwmjones> it used to be that you couldn't strip any bytecode binaries at all, but that's not true since 3.09
[12:19]  <abadger1999> Do we need to/want to discuss new meeting times?  (Since we have new members)?
[12:19]  <-- scop has left this channel ("Leaving").
[12:19]  <spot> abadger1999: lets do that on the mailing list
[12:19]  * spot tries to keep us focused on the ocaml topic
[12:19]  <spot> if you haven't voted, please do so. issue needs +5 to pass.
[12:19]  <Rathann|work> +1 from me
[12:19]  <rdieter> +1
[12:20]  <rwmjones> that's +4 so far by my count ...
[12:20]  <tibbs> +1 all seems reasonable.
[12:20]  <hansg> still +1 from me
[12:20]  <racor> 0
[12:20]  <Rathann|work> I'd like the find-provides/requires to hack to find its way into our rpm-build package
[12:20]  <tibbs> Plus I've reviewed a bunch of ocaml stuff and this will help to clarify.
[12:20]  <abadger1999> rwmjones: What is the META file you reference?
[12:20]  <spot> abadger1999: want to vote?
[12:20]  <rwmjones> Rathann|work, there's still a bug that I'm working with upstream about
[12:20]  <Rathann|work> cool
[12:20]  <tibbs> I agree with Rathann|work that it should get upstream.
[12:21]  <abadger1999> I want to +1 as son as I know what META is.
[12:21]  <rwmjones> abadger1999, META is a file which describes how libraries are linked together (kind of link pkg-config)
[12:21]  <tibbs> But we can't really wait.
[12:21]  <rwmjones> abadger1999, debian agreed to standardize and have _all_ their libraries require a valid META file
[12:21]  <tibbs> abadger1999: There's a file named META that contains some package description.
[12:21]  <rwmjones> so I copied their policy for us
[12:21]  <abadger1999> rwmjones: What should the packager/reviewer "check" in it?
[12:21]  <rwmjones> not much, basically just the 'requires = "..."' lists all required libraries
[12:22]  <rwmjones> I should reference some upstream documentation about META if I can find any ...
[12:22]  <rwmjones> hmm
[12:22]  * rwmjones checks
[12:22]  <tibbs> Hmm, yeah, might be nice to say what should be there.
[12:22]  <tibbs> I've only been checking that it exists.
[12:22]  <abadger1999> Ocaml +1
[12:22]  <spot> ok, it passes.
[12:22]  <rwmjones> do I need to take it to fesco now?
[12:23]  <spot> rwmjones: we'll take it
[12:23]  <tibbs> It will be in the summary and fesco will discuss it on Thursday.
[12:23]  <rwmjones> ok thanks
[12:23]  <spot> Next item:
[12:23]  <spot> diff since we last discussed it:
[12:25]  <spot> the main concern from before was that we weren't sure how unpacking the file in advance saved disk space
[12:25]  <spot> caolan updated the draft with explanatio
[12:25]  <spot> ...n. :)
[12:25]  <spot> it also includes an example, so I'm +1 on this.
[12:26]  <tibbs> I don't remember whether we discussed it last time, but is there nothing compiled in these extensions?
[12:26]  <spot> tibbs: i'm pretty sure the answer is no
[12:27]  <Rathann|work> what language are they written in?
[12:27]  <rwmjones> sorry to interrupt out of turn, I've just made this edit which adds a link describing META files:
[12:27]  <tibbs> Most of these are in Java, I think.
[12:27]  <Rathann|work> I mean, they're not binaries, are they?
[12:28]  <abadger1999> is the new example correct?  It looks like the unpackaed extension is going in a directory with a .zip extension.
[12:28]  <hansg> Hmm, IMHO the openoffice-foo naming should be made mandatory
[12:28]  <spot> i think they're java goop
[12:28]  <Rathann|work> hm
[12:28]  <tibbs> Do we still want to see Java guidelines first?
[12:28]  <Rathann|work> shouldn't they be compiled using our javac then?
[12:29]  <tibbs> It's possible that compiling them would have no use.
[12:29]  <tibbs> But honestly someone who knows needs to be around to answer these questions.
[12:29]  <abadger1999> hansg: +1
[12:30]  <rdieter> no precompiled binaries is pretty much already covered by existing guidelines.
[12:30]  <spot> ok, i'll pose these items to caolan and we'll revisit it in two weeks
[12:30]  <tibbs> Also, there's probably no point in having Orion in the template changelog.
[12:30]  <Rathann|work> rwmjones: you should add a footnote link, not just tell people to look at the bottom
[12:30]  <hansg> Indeed I see no compiling / build step in the current .spec example. Since its mandatory that everything gets compiled from source, this is not acceptable
[12:30]  <abadger1999> rdieter: True, but the examples should show that then.
[12:30]  <spot> next item:
[12:31]  <spot> since last discussed:
[12:31]  <tibbs> hansg: Many specs have no %build and that's not problematic.
[12:31]  <spot> the concern was around the ambivalent naming case
[12:31]  <spot> he has since resolved it
[12:31]  <tibbs> If there's nothing to compile, the rule about compiling from source doesn't really apply.
[12:31]  <Rathann|work> tibbs: yes, but they contain scripts, not binaries
[12:31]  <Rathann|work> or firmware
[12:31]  <hansg> tibbs I know, but these plugins are code afaik, not data files / nor in an intepreted language
[12:31]  <racor> where in $PATH is unopkg? fc8 has /usr/lib64/, so scriptlets won't work
[12:32]  <spot> [spot@localhost logjam-4.5.3] $ rpm -qf /usr/bin/unopkg
[12:32]  <spot>
[12:32]  <hansg> I was just about to complement spot that he is keeping the tempo up, but it seems we're still at ooo extensions?
[12:32]  <tibbs> I guess doing to tcl was optimistic.
[12:33]  * spot waves his hands about
[12:33]  <racor> spot: => R: is not sufficient
[12:33]  <tibbs> racor: The scriptlets do work.
[12:33]  <tibbs> I can't say just how.
[12:33]  <hansg> racor, you;re probably at F-8, and this is F-9 and up only
[12:33]  <tibbs> See the existing writer2latex package, btw.
[12:33]  <rdieter> in particular, ooo >= 2.4
[12:33]  <racor> tibbs: fc8 doesn't have /usr/bin/unopkg => this proposal is incompatible to fc8
[12:34]  <tibbs> racor: Who said it was supposed to be compatible with f8?
[12:34]  <spot> i will have caolan clarify how to do this for F-8 and older.
[12:34]  <Rathann|work> -1 from me on openoffice extensions for packaging java binaries without compiling them
[12:34]  <tibbs> caolan indicated that he may push 2.4.0 to the release branches which would bring in a proper unopkg.
[12:34]  <hansg> Rathann|work,  no voting needed I think, this cleary needs more work
[12:34]  <spot> Rathann|work: we're not voting on this draft, it still has obvious issues
[12:35]  <racor> hansg: Yes, FC8. This proposal apparently is only applicable for fc >9.
[12:35]  <spot> we are, however, now looking at the tcl draft.
[12:35]  <spot>
[12:35]  <spot> diff since last discussed:
[12:35]  <spot> from last time, the concern was around the ambivalent naming case, which the author has resolved.
[12:36]  <spot> I'm +1 on this.
[12:37]  * spot assumes everyone is reading (or i've fallen off the internet)
[12:37]  * Rathann|work is reading
[12:37]  <rdieter> "This rule applies even for Tcl/Tk packages that are already prefixed with tcl in the name (see examples below)", really?
[12:37]  <abadger1999> all the things I remember being issues have been cleaned up.
[12:37]  <tibbs> I guess EPEL people might appreciate some comments on some of the version-specific issues in those releases.
[12:37]  <tibbs> But we can add that later.
[12:38]  <rdieter> I'm only curious, I'm +1 either way.
[12:39]  <abadger1999> +1
[12:39]  <hansg> +1
[12:39]  <spot> thats +4. +5 needed. :)
[12:39]  <tibbs> FYI, yum list tcl\*|grep -v tcl- gives 21 packages.
[12:40]  <tibbs> So we'd have tcl-tclxml according to these new naming guidelines.
[12:40]  <tibbs> And I picked the example that's already in there.  Duh.
[12:41]  <Rathann|work> looks sensible, +1
[12:41]  <tibbs> Anyway, +1.
[12:41]  <tibbs> But we need to think about whether we want existing packages to be renamed.
[12:41]  <spot> tibbs: we've never mandated that in the past.
[12:41]  <hansg> tibbs: renaming -1
[12:41]  <rdieter> I was wonderying why tcl- was mandatory.  That's contrary to our guidelines wrt python, for example.
[12:41]  <spot> our guidelines don't forbid it.
[12:42]  <tibbs> I personally dislike the python guidelines because of that.
[12:42]  <spot> anyways, it passes
[12:42]  <spot> PackagingDrafts/Lisp still isn't ready, i need to spend some time working on that one with the author
[12:42]  <rdieter> tibbs: I'd prefer consistency, one way or the other.
[12:43]  <spot> thats all the items that I have on the schedule
[12:43]  <Rathann|work> rdieter: I've been told to rename a subpackage from foo-python to python-foo a couple of times
[12:43]  <spot> the floor is open to any other issues
[12:43]  <tibbs> I only bring up the renaming because I don't think we've made guideline changes that would cause so much renaming before.
[12:43]  <hansg> I would like to start working on the java guidelines
[12:43]  <tibbs> Yes.
[12:43]  <spot> hansg: great!
[12:43]  <rdieter> Rathann|work: been there, done that. :)
[12:43]  <spot> you don't need our permission to do that. ;)
[12:43]  <tibbs> hansg: Did you get the message about the conference call?
[12:44]  <hansg> So any people who are willing to assist (as in review what ever ugly thing I come up with) ?
[12:44]  <tibbs> I think the idea of a conference call is kind of pointless but I'm going to try to be up early enough to attend anyway.
[12:44]  <tibbs> hansg: Did you see the in-progress draft?
[12:44]  <hansg> Or shall I just post a message to the packaging-list when I've got a new draft ready?
[12:44]  <tibbs> Folks have been working on it.
[12:44]  <spot> hansg: the packaging-list might not be a bad idea, a wider audience
[12:45]  <abadger1999> We have unicode naming from last wekk.
[12:45]  <hansg> tibbs: yes and I'm in contact with Lubu erm whats his name?
[12:45]  <abadger1999> I'm still -1 to banning unicode in package names.
[12:45]  <hansg> Conference call?
[12:45]  <Rathann|work> abadger1999: well, I was half-joking in my posts on that topic ;)
[12:45]  <abadger1999> But there might be enough people to passit.
[12:45]  <Rathann|work> -1 from me too
[12:45]  <tibbs> hansg: Yes, Andrew Overholt mailed a bunch of people about having a conference call on Friday.
[12:46]  <hansg> I'm -1 to unicode in packagenames , package name == filename, and our mirrors aren't ready, I'm not even sure we are ready.
[12:46]  <tibbs> I'm happy to simply not vote until we know the infrastructure is ready.
[12:47]  <abadger1999> hansg: The proposal was from racor:  "Ban unicode in package names"
[12:47]  <hansg> tibbs, a conference call about the java guidelines?
[12:47]  <tibbs> Although I think we should have one inconsequential unicode package name just to test our infrastructure.
[12:47]  <hansg> abadger1999,  I know
[12:47]  <tibbs> hansg: Yes.  Andrew Overholt == big Java guy in Red Hat.
[12:47]  <abadger1999> Okay.  But you'd be +1 to the proposal then.
[12:47]  <Rathann|work> écolier-fonts?
[12:47]  <tibbs> Lubomir isn't really the person to be talking to as far as I know.
[12:48]  <hansg> tibbs, which sounds like a good idea until createrepo explodes on updates, or half of the mirrors fail to sync
[12:48]  <hansg> abadger1999, yes
[12:48]  <tibbs> If we keep our heads in the sand and never test it, we'll never test it.  At some point we have to try sometthing.
[12:48]  <tibbs> But that's more in the court of the infrastructure folks.
[12:50]  <hansg> tibbs, about the java things, Lubomir  was asking people to take a look at the guidelines on fedora-devel list, but I'll get in contact with Andrew too.
[12:50]  <tibbs> hansg: We're seeing the typical tichotomy between people who work better in person or via voice and the people who want to work by email, wiki and IRC.
[12:51]  <tibbs> Personally I don't understand why we need a conference call at all, but I didn't call for it, so....
[12:51]  <abadger1999> Maybe we need to create a little repo on our infrastructure that gets mirrored but isn't the main repo.  Then we can start trying things and see what breaks.
[12:52]  <abadger1999> drop écolier-fonts in there and see what breaks.
[12:52]  <hansg> abadger1999, sounds like a plan
[12:52]  <tibbs> A pain for someone besides us, though.
[12:52]  * abadger1999 opens a ticket
[12:53]  <tibbs> I say we can ask infrastructure to tell us when they think things are ready.  Until then we obviously can't have utf8 package names for fear of breaking things.
[12:53]  <racor> sorry, i got distracted
[12:54]  <racor> abadger1999: -1
[12:54]  <tibbs> racor: -1 to what?
[12:55]  <racor> to abadger1999's drop écolier-fonts in there and see what breaks
[12:55]  <tibbs> That's an infrastructure team matter.
[12:55]  <racor> tibbs: my issue is not the infrastructure, my point is usability
[12:55]  <abadger1999> racor: Because of usability?
[12:56]  <racor> abadger1999: let me reiterate: ban non acsii, because many users will not be able to type such package names
[12:56]  <hansg> racor I think you misunderstand / misread, abadger proposed a seperate unicode-test repo, and then request some mirrors to carry this (seperately) and then we can get some idea if utf-8 names are technicall feasible
[12:57]  <tibbs> You know, I think we all understand the point that you're against this.  Simply throwing out -1 whenever someone wants to investigate part of the issue really isn't productive.
[12:57]  <racor> hansg: this would be OK
[12:57]  <spot> ok, so the proposal is to ban unicode (non-ascii) in package names
[12:57]  <hansg> racor: system -> preferences -> hardware -> keyboard
[12:57]  <spot> can i see votes on that specific proposal?
[12:58]  <racor> tibbs:  drop écolier-fonts in there and see what breaks to me means : release this package under this name and see who complains.
[12:58]  <hansg> layoutoptions-> composekey -> choose one
[12:58]  <hansg> compose " e ->ë
[12:58]  <tibbs> I wish I had a compose key.
[12:58]  <racor> tibbs: gotcha
[12:58]  <abadger1999> racor: Sorry.  My thought was split over two lines.
[12:58]  * spot whistles to himself
[12:58]  <hansg> tibbs, you can choose one, I use left alt, or one of those never used windows keys
[12:59]  <abadger1999> spot: -1
[12:59]  <hansg> make that right alt
[12:59]  <racor> hansg: I give you 5 secs to answer: type a cyrillic "k" (small k)
[12:59]  <hansg> anyways the point is latin1 chars aren't that hard, I agree chinese chars is a different story
[13:00]  <tibbs> spot: 0
[13:00]  <hansg> racor that isn't laten1 but latin2
[13:00]  <tibbs> I don't think we can really consider the matter as long as a vote allowing utf8 package names is pointless because infrastructure blocks it.
[13:00]  * spot calls for votes again...
[13:00]  <racor> hansg: to you! to aunt tillie "é" or "ß" is a problem
[13:00]  <hansg> aunt tillie will use a gui -> no problem
[13:01]  <tibbs> The aunt tillie argument really gets old.
[13:01]  <tibbs> She'll just click.
[13:01]  <racor> sorry, but this is going to be ridiculous
[13:01]  <-- racor has left this server ("Leaving").
[13:02]  <tibbs> "takes his ball...."
[13:02]  <rdieter> spot: +1 (at least until infrastructure signs off)
[13:02]  <hansg> spot votes for what, I say lets vote to end this meeting :)
[13:02]  <spot> since we're down to 6 members, and two have voted either against or to abstain...
[13:02]  <spot> the measure fails
[13:03]  <spot> are there any other items for today?
[13:03]  <tibbs> Did we make any progress on the remaining vacancy?
[13:03]  <hansg> am I one of those 2? and what measure I'm confused now
[13:03]  <spot> <spot> ok, so the proposal is to ban unicode (non-ascii) in package names
[13:03]  <spot> <abadger1999> spot: -1
[13:03]  <spot> <tibbs> spot: 0
[13:03]  <hansg> ok
[13:04]  <spot> if you don't have anything additional for today, please indicate that no, you do not. :)
[13:04]  <spot> my psychic powers only work on thursday.
[13:04]  <hansg> I don't
[13:04]  <tibbs> Nothing besides the question I just posed.
[13:04]  <rdieter> no
[13:04]  <spot> tibbs: we have some volunteers, i'm going to send an email out for people to vote on them.
[13:04]  <tibbs> Cool, thanks.
[13:05]  <spot> ok guys, we're done for today. thanks for your patience. we'll meet again in two weeks.
[13:05]  <abadger1999> no