Meeting:Packaging IRC log 20060907

From FedoraProject

Jump to: navigation, search
[10:57:07]  Join scop has joined this channel (n=scop@cs181056070.pp.htv.fi).
[10:58:35]  <tibbs> So we have 1/2 scop and 0 spot. Probably no racor either.
[10:59:02]  <rdieter> ok, let's party hard then. (:
[10:59:28]  <rdieter> (or run around in circles, not sure which way will accomplish more)
[10:59:52]  <scop> yep, I need to run @1630 UTC, and will be only half present also until that
[10:59:57]  <tibbs> I wonder if we shouldn't consider what we'll do if Ralf quits the project as he keeps threatening to do.
[11:00:36]  <rdieter> like when he doesn't show up for meetings? (:
[11:00:55]  <tibbs> He has told us that the time doesn't work for him.
[11:02:07]  <f13> I'm here.
[11:02:17]  <rdieter> maybe we can talk about new meeting time(s) first then?
[11:02:34]  <tibbs> It would be nice to have Toshio here.
[11:02:54]  <tibbs> As regards to the times, I believe that everyone has acked the new time except for Ralf.
[11:03:04]  <tibbs> Sorry, it's not really "the new time".
[11:04:27]  <thimm> Jose hasn't ack'd, too
[11:04:32]  <rdieter> I remember seeing a reply from Ralf, saying the new time isn't any better than the old (or possibly worse)
[11:04:40]  <thimm> I'll be sporadically available today
[11:05:38]  <scop> I'll mail Jose and ask what's up
[11:05:42]  <tibbs> I have not seen Jose participate in any meetings or discussion yet. If this is because of meeting times then it would be really good if he could at least weigh in on a time that works for him.
[11:07:47]  <tibbs> We have to start getting sufficient people here or there's really no point in having meetings.
[11:07:48]  <rdieter> I guess spot (or us, someone) should ping Jose again, and for lack of response, a replacement be found.
[11:07:56]  <rdieter> tibbs++
[11:08:54]  * scop already promised to ping Jose a few lines up
[11:09:44]  <rdieter> FYI, Ralf's reply (1): https://www.redhat.com/archives/fedora-packaging/2006-September/msg00018.html
[11:09:49]  <scop> so, anything we want to actually discuss today?
[11:10:10]  <tibbs> scop, can you still not edit ReviewGuidelines?
[11:10:15]  <f13> does somebody want to formalize some of the .pc discissions that have been going on?
[11:10:38]  <scop> tibbs, I can, haven't just found round tuits for it
[11:10:55]  <rdieter> f13: didn't the (mostly) consensus end up being simply sticking with the current guideline(s)?
[11:11:22]  <tibbs> scop: I'll update the schedule to remove the note that you can't update them.
[11:11:28]  <scop> ok
[11:11:46]  <f13> rdieter: except we still get one file subpackages for no good reason
[11:11:55]  <rdieter> Or is there interest to remove the MUST: .pc files be in -devel pkg and change it to say:
[11:12:27]  <f13> gtk-sharp2-gapi-devel for one file.
[11:12:31]  <rdieter> (something ilke) MUST: .pc files must be packaged with the other -devel-type stuff (which *usually* is a -devel subpkg).
[11:12:58]  <tibbs> I disagree with "no good reason".
[11:13:18]  <f13> rdieter: also, there should be some text about .pc files USUALLY being used to determine how to compile a package, but with mono, its also somewhat how to RUN a package.
[11:13:33]  <tibbs> It seems that there's a pretty good reason for keeping -devel stuff, especially given what automatic dependency generation can do.
[11:13:59]  <tibbs> We don't want to say it's OK now and then change back if automatic dependency generation from .pc files gets turned on.
[11:14:04]  <rdieter> agreed, .pc files can also convey runtime info too (despite jeremy's assertion that this is crack-talk)
[11:15:07]  Join abadger1999 has joined this channel (n=abadger1@108.164-78-65.ftth.swbr.surewest.net).
[11:15:30]  <rdieter> It's starting to look like we can't have a simple cut-n-dried rule (as-it-is now) anymore...
[11:16:01]  <f13> indeed.
[11:16:11]  <tibbs> That is indeed the case. Perhaps we need to think about blanket allowing reasonable exceptions instead of completely formalizing everything.
[11:16:30]  <rdieter> tibbs++
[11:16:35]  <f13> tibbs: if the .pc is already _in_ a development type package, there isn't really a good reason to further split it into -devel
[11:17:01]  <tibbs> Compilers already violate all of the rules about what should be in -devel.
[11:17:13]  <rdieter> perhaps even, go one step further? Until we come up with something better, drop the rule: "MUST: .pc files be in -devel"
[11:17:28]  <tibbs> So we've been making a reasonable exception there for ages.
[11:17:46]  <tibbs> rdieter: At that point we should start dropping all kinds of rules.
[11:17:53]  <rdieter> Either drop, or change MUST -> SHOULD?
[11:18:06]  <tibbs> It's obviously good policy that .pc files need to be in -devel packages.
[11:18:28]  <rdieter> *most* .pc files need to be in -devel pkgs, yes (atm anyway).
[11:18:34]  <tibbs> The problem is that there will always be necessary and reasonable exceptions to pretty much every rule.
[11:19:05]  <rdieter> ok, how do we communicate that to packagers/reviewers?
[11:19:12]  <tibbs> If you think you can list out the necessary and reasonable exceptions for the .pc rule, then go ahead.
[11:19:18]  <f13> perhaps the mono packaging guide should list reasonable exception to the .pc rule.
[11:19:31]  <rdieter> f13++ (that's a start anyway)
[11:19:38]  <f13> to prevent folks like noid from mass filing bugs for improper packaging.
[11:20:16]  <abadger1999> f13: Or have a place to point him when he does....
[11:20:25]  <tibbs> The problem with some reviewers and some package submitters is that one or the other loses sight of the fact that a review is a discussion.
[11:20:52]  <abadger1999> (Sorry I'm late. Anyone have logs they can pastebin quickly?)
[11:22:13]  <tibbs> abadger1999: http://www.math.uh.edu/~tibbs/irc-log.txt
[11:22:26]  <abadger1999> tibbs: Thanks
[11:22:42]  <rdieter> OK, does anyone disagree that the mono guidelines should mention this "reasonable exception"?
[11:23:09]  <rdieter> (or should we just vote?)
[11:23:36]  <f13> +
[11:23:37]  <f13> 1
[11:24:03]  <f13> but perhaps we should write up the proposed reasonable exceptions first? (:
[11:24:28]  <rdieter> f13: darn, I was trying to give you a blank check...
[11:24:48]  <tibbs> Who knows enough about this to propose something?
[11:24:48]  <abadger1999> f13: Yes. I think change is in order but we should know what the change is.
[11:25:14]  <tibbs> runtime stuff in .pc seems to violate decency, but then this is mono.
[11:25:34]  <f13> heh
[11:26:31]  <abadger1999> Did everyone see alexl's reply to my email this morning?
[11:27:02]  <abadger1999> It has a pretty formal rule but I'm not sure we're discussing it or not.
[11:27:21]  <f13> yes, he still had issues with the creation of something like gtk-sharp2-gapi-devel
[11:29:28]  <scop> duty calls, I need to go now, sorry
[11:29:49]  Part scop has left this channel ("Leaving").
[11:29:51]  <tibbs> scpo shorts us one minute.
[11:29:59]  <rdieter> (just re-read alex's reply), yep, he's got it, right on.
[11:30:06]  <abadger1999> I have no problem with something allowing gtk-sharp2-gapi as it appears to be runtime.
[11:30:27]  <rdieter> proposal: .pc files must be in a -devel pkg, unless the pkg itself is a devel tool not installed in a user runtime.
[11:30:58]  <rdieter> ... If the pkg has a -devel subpkg for other reasons, the place for the .pc file is up to the package owner (depending on the usecase for the .pc file)
[11:31:37]  <tibbs> I can buy that.
[11:32:47]  <tibbs> There are other rules that are bent in the same way. For example, gcc has all kinds of devel stuff but no -devel package (as it would be pointless).
[11:34:41]  Join abadger1991 has joined this channel (n=abadger1@108.164-78-65.ftth.swbr.surewest.net).
[11:35:02]  <tibbs> How many abadgers are there?
[11:36:25]  <rdieter> We could expand on this, and define when/if -devel subpkgs are appropriate.
[11:36:27]  <abadger1991> It's actually a date, so you're now speaking to my younger self.
[11:37:15]  <tibbs> rdieter: The problem is that at some point we get away from nice guidelies and get into something hilarious like the tax code.
[11:37:41]  <tibbs> We will never be able to define in advance all of the necessary exceptions.
[11:37:47]  <rdieter> fair enough, so are we happy with alex's proposal?
[11:39:32]  <rdieter> I'd say a slight ammendment: ... If the pkg has a -devel subpkg for other reasons, the place for the .pc file depends on the usecase for the .pc file
[11:40:25]  <rdieter> (not that it matters, I 'don't think we have enough folks to pass anything official anyway)
[11:40:49]  <rdieter> I'll put that on our ToDo, maybe we can vote/pass next week.
[11:41:35]  <abadger1991> I'm reasonably happy. But let's extend it a bit... If gcc were to grow a pkgconfig file, would we be happy with gcc dragging in glibc-devel?
[11:41:58]  <rdieter> (It doesn't already?)
[11:42:05]  <abadger1991> (Not sure.)
[11:42:20]  <abadger1991> It does.
[11:42:25]  <rdieter> So what exactly did you have in mind then?
[11:42:40]  <abadger1991> So based on current Core thinking, we'd be happy with it.
[11:42:45]  <abadger1991> :-)
[11:44:05]  <rdieter> afk (for a few...)
[11:45:56]  Join abadger2000 has joined this channel (n=badger@108.164-78-65.ftth.swbr.surewest.net).
[11:47:45]  Quit abadger1999 has left this server (Read error: 110 (Connection timed out)).
[11:47:52]  Join abadger1999 has joined this channel (n=badger@108.164-78-65.ftth.swbr.surewest.net).
[11:50:32]  <rdieter> anything else to discuss? (I'm sure we must be forgetting something...)
[11:51:47]  <rdieter> fyi, just added "update .pc file guideline" to GuidelinesToDo
[11:52:11]  <tibbs> We still need to revisit the cross-compiler issue. I've given up on having all of the interested parties here at once.
[11:52:38]  Quit abadger1991 has left this server (Read error: 60 (Operation timed out)).
[11:53:25]  <abadger1999> I'm +1 for tibbs+Ralf naming :-)
[11:53:59]  <tibbs> Ralf has objected to "cross-", but I can't fathom why it would be a problem for anyone.
[11:54:12]  <abadger1999> cross-[Whatever it was Ralf said] 
[11:54:25]  <tibbs> He said he wants things to sort together, but then doesn't want all of the cross-compilers to sort together.
[11:54:28]  <rdieter> Ralf just didn't like the forced/artificial cross- prefix.
[11:54:43]  <tibbs> I'll write up a proposal for full discussion.
[11:54:52]  <rdieter> Personally, I'm ok with either.
[11:55:46]  <rdieter> any jpackage movement?
[11:56:01]  <rdieter> (or are we done with that?)
[11:56:17]  Quit abadger2000 has left this server (Read error: 60 (Operation timed out)).
[11:57:27]  <rdieter> Oh, I remember now (as we run out of time), the mailing lists had more discussion on "files generated by scriptlets".
[11:58:53]  <tibbs> I think it's rare enough that it may be reasonable to deal with on a case-by-case basis.
[11:59:11]  <rdieter> ok, before I go... anybody grabbing the irc log? (If not, I can do it)
[11:59:12]  <tibbs> Do we have an example of a package that does this but does not create certs?
[11:59:37]  <abadger1999> rdieter: Please do.
[12:00:10]  <abadger1999> I usually do but I've been disconnected more often than not today.