Packaging:Minutes/20090512

From FedoraProject

Jump to: navigation, search

Contents

Fedora Packaging Committee Meeting 2009-05-12

Present

  • Denis Leroy (delero)
  • Dominik Mierzejewski (Rathann)
  • Jason Tibbitts (tibbs)
  • Rex Dieter (rdieter)
  • Tom Callaway (spot)
  • Toshio Kuratomi (abadger1999)
  • Xavier Lamien (SmootherFrOgZ)

Regrets

  • Hans de Goede (hansg)
  • Ralf Corsepius (racor)

Votes

The following proposals were considered; please note that racor's votes were on-list and available at https://www.redhat.com/archives/fedora-packaging/2009-May/msg00056.html

IRC Logs

* spot gets ready for another exciting FPC meeting 12:04
tibbs Indeed. 12:05
spot abadger1999, SmootherFrOgZ, tibbs, rdieter: ping 12:05
rdieter here 12:05
abadger1999 spot: here 12:05
* SmootherFrOgZ here 12:06
--> delero has joined this channel (n=delero@nat/sun/x-cf0eacf74c9de57d). 12:07
* delero is here 12:07
spot i pinged Rathann too, but i think he's idle 12:07
spot we're not expecting hans, racor 12:07
spot so, i think everyone that's coming is here 12:08
spot First item: https://fedoraproject.org/wiki/GConf_Scriptlets_%28draft%29 12:08
spot This is abadger's refined draft of handling the GConf scriptlets 12:08
spot abadger1999: have you tested the macros? 12:09
abadger1999 spot: I tested them last night. they seemd to work. 12:09
spot okay. 12:09
abadger1999 spot: I may have left out some corner cases though so anyone else that wants to give them a try is welcome to. 12:10
SmootherFrOgZ np 12:10
abadger1999 Copy the macros into a file in /etc/rpm/ 12:10
* spot knows < 0 about GConf 12:10
delero negative knowledge 12:10
tibbs Me too, but it sure looks nicer. 12:10
--> Rathann has joined this channel (n=rathann@fedora/rathann). 12:10
spot that said 12:11
Rathann sorry for being late 12:11
spot the macros look nice, the logic is sound 12:11
spot and the GNOME folks seem to like it 12:11
abadger1999 Rathann: We're discussing this: https://fedoraproject.org/wiki/GConf_Scriptlets_%28draft%29 12:11
Rathann yup, I gathered as much from what spot said just now 12:11
spot I'm inclined to +1 on this one. 12:11
Rathann +1 from me as well 12:12
rdieter +1 12:12
SmootherFrOgZ +1 from me 12:12
delero +1 12:12
abadger1999 +1 12:12
tibbs Did the desktop team get a chance to chime in? 12:13
abadger1999 Ah one thing -- anyone want to weigh in on what package to add this to? 12:13
abadger1999 mclasen commented on the early drafts. He liked it. 12:13
spot abadger1999: perhaps redhat-rpm-config ? 12:13
tibbs I recall that they liked the idea of changing things. Just wanted to make sure that we didn't go off in the wrong direction for them. 12:14
spot that would get it in the buildroot by default 12:14
abadger1999 He's working with upstream to get schemas to move to %{_datadir} and wanted to have that change encoded in macros. 12:14
abadger1999 <nod> 12:14
abadger1999 redhat-rpm-config will work. Do we want the macro and the directory to be owned by the same package? 12:15
tibbs Is that always present at runtime? 12:15
tibbs It is not. 12:15
abadger1999 If so, GConf using packages need to Require(pre): Require(post): redhat-rpm-config 12:15
spot hrm. 12:16
abadger1999 correct, it's not installed by default at runtime. 12:16
spot it seems like GConf2 should own /var/lib/rpm-state/gconf 12:16
abadger1999 yeah. 12:16
tibbs Yes, if possible, gconf should own this. 12:16
spot but i don't want redhat-rpm-config to dep on GConf2 12:16
rdieter spot: +2 12:16
abadger1999 The macros shouldn't need the directory at build time. 12:17
abadger1999 Just when the scriptlets run. 12:17
tibbs Do things fail gracefully (or at least understandably) if that gets fouled up? 12:17
abadger1999 tibbs: Not gracefully. But understandably for someone who knows what's in the macros. 12:18
spot Well, since anything that tries to use these macros is going to need Requires(pre): GConf2 12:18
tibbs Someone somewhere will miss a dependency eventually. 12:18
abadger1999 A bit obtuse for someone who doesn't know what's in the macros (that's the problem anytime you hide the intricate details of something) 12:18
spot i think it is safe for GConf2 to own the dir and have redhat-rpm-config own the macros 12:19
spot without an explicit dep between the two 12:19
--> mattiasellert has joined this channel (n=ellert@c213-100-175-41.swipnet.se). 12:19
abadger1999 Do we want to break up /var/lib/rpm-state and /var/lib/rpm-state/gconf (or use a different /var/lib/$DIRECTORY) 12:19
* spot doesn't care about that really 12:20
* abadger1999 doesn't care either 12:20
rdieter macros could be in GConf2 as well, since gconf-using apps will have that BR'd anyway. 12:20
tibbs Anyway, +1 to this; as long as the last couple of details get worked out sanely I don't see any problem. 12:20
spot rdieter: thats a good point 12:20
spot so, I see +7 12:21
spot it passes 12:21
spot Next item: https://fedoraproject.org/wiki/PackagingDrafts/Globus 12:21
spot the old macro cruft is gone from the examples (and presumably, the tool as well) 12:22
tibbs I thought that was going to the list after things were changed but I don't recall seeing it. 12:22
mattiasellert yes 12:22
* spot tries to remember what the other concerns were 12:22
tibbs Anyone else getting no response from fedoraproject.org all of a sudden? 12:22
spot yep. 12:22
rdieter tibbs: here too 12:22
SmootherFrOgZ same 12:23
spot i have cached copies of the pending drafts... 12:23
spot if push comes to shove. :) 12:23
delero ye 12:25
Rathann looks like it's back 12:25
tibbs Back up for me, at least. 12:25
spot Okay, so the three items from last meeting were... 12:25
spot the spec tool (and templates) needs to limit disttag macro use to rhel and fedora >= 9 12:25
spot (done) 12:25
spot the spec tool should use pushd/popd to simplify the %install section 12:25
spot and documenting the versions of doxygen that need to be cleaned up after 12:26
rdieter dumb/late question perhaps, but any chance to ask globus upstream to distribute stuff a little more sanely? 12:26
tibbs  %install is simplified with a macro and looks much better. 12:26
f13 hrm, package committee meeting? 12:27
spot tibbs: well, its simplified with a shell variable, which is a little incongruous 12:27
SmootherFrOgZ f13: yep 12:27
spot but its not wrong 12:27
f13 can somebody ping me when the java stuff comes up? 12:27
spot f13: yeah 12:27
tibbs We can't complain as long as we allow $RPM_BUILD_ROOT, of course. 12:27
Rathann this flavor macro should be moved into rpm macros 12:28
spot tibbs: yeah, i know. 12:28
tibbs But I'd personally use something much shorter. Still, it looks better than it did. 12:28
Rathann it's on top of every specfile in the examples 12:28
spot mattiasellert: did you figure out which versions of doxygen required the cleanups? 12:29
mattiasellert Yes. 12:29
mattiasellert It is needed for RHEL4 12:29
spot okay. 12:30
mattiasellert I added a comment about that in the examples. 12:30
spot I see that now 12:30
Rathann GLOBUSPACKAGEDIR could be an rpm macro too 12:31
spot well, i'm okay with it with the changes. it looks like a nasty beast to have to tame (globus), but this seems mostly sane. 12:31
spot Rathann: thats what I was talking about, but i won't lose sleep over it not being one 12:31
tibbs Yes, this seems to be making the best of a rather bad situation. 12:31
Rathann those fedora/rhel conditionals in BRs 12:31
Rathann could be simplified by requiring the necessary binaries directly 12:32
Rathann i.e. BR: %{_bindir}/latex 12:32
tibbs True. 12:32
spot  %if %{?rhel}%{!?rhel:0} >= 6 ... predictive? :) 12:32
Rathann since %{_bindir} and %{_sbindir} are in base metadata, this doesn't mean downloading filelists to resolve them 12:33
tibbs That's not terribly important for build dependencies anyway. 12:33
Rathann yup 12:33
* Rathann doesn't like distro-specific conditionals 12:34
spot Either with that change or without it, my vote here is +1. 12:34
tibbs I don't like them because they make the spec look like line noise. 12:34
delero don't like them either, but I wouldn't vote this down for it 12:34
Rathann yeah 12:35
SmootherFrOgZ tibbs: <nod> 12:35
Rathann +1 from me 12:35
abadger1999 +1 12:35
rdieter +1 12:35
delero +1 12:35
spot thats +5. 12:36
tibbs I'm torn 12:36
tibbs I support this, but I think it could be simpler. 12:36
tibbs So I guess +1. 12:36
spot mattiasellert: if you decide to move to using the file dependencies rather than the conditionalized ones, please let us know so we can update the templates in the guidelines 12:37
mattiasellert OK 12:37
spot Next item: https://fedoraproject.org/wiki/Pre-review_Guidelines_%28draft%29 12:37
spot Its worth noting that this is something we've been asked to do by FESCo 12:38
Rathann frankly I don't like the whole idea of prereviews at all 12:38
spot There is apparently a big pile of java packages that need to come in at once 12:38
dbhole f13: This is us 12:38
Rathann why can't they go through regular reviews? 12:38
dbhole overholt: ^ 12:38
tibbs I dislike this idea as well. 12:38
tibbs However, that doesn't come from my committee position. 12:38
Rathann the draft doesn't answer the question why we need them at all 12:39
f13 so FESCo already approved them coming in 12:39
f13 FESCo just asked the PC to come up with the minim required things to check off 12:39
f13 such as build from source and proper licenses 12:39
spot f13: is this only permitted for a specific set of java packages 12:39
tibbs If we're going to do this (and admit that we don't care about package quality at that stage) then we just need to verify legal status and that the source is unadulterated. 12:39
spot or is this for anything that wants to? 12:39
f13 spot: no, this is a one time deal for a specific set of java packages for this time critical maven update 12:40
tibbs I don't even agree with "build from source". 12:40
spot tibbs: you don't? 12:40
abadger1999 tibbs: I do since it's toolchain related. 12:40
f13 the people who have the time/capability to work on the maven update have a small window of opportunity to work on it, and they are different folks than the ones that will work on the fine tuning of the packaging layout 12:40
tibbs If we cared about that we wouldn't be doing this. 12:40
abadger1999 If it wasn't toolchain, then I'd let that be caught at the full review stage. 12:41
spot Honestly, i'm not sure that this is in the FPC domain. 12:41
tibbs FESCo seems to have made it FPC domain. 12:41
f13 spot: abadger1999 thought it was and asked it to be 12:41
abadger1999 Well, I asked that the things to review for come here. 12:41
tibbs So what can we agree on? 12:41
spot These aren't really guidelines. They're asking us to tell them how to do "prereviews" before permitting cvs commit. 12:41
tibbs "check the license", obviously. 12:42
tibbs "check that the upstream source is unadulterated"? 12:42
spot given that these apply only to Java, i think the key points on the draft are especially valid 12:42
abadger1999 If we don't think it is FPC purview, I can just toss that page back out -- it satisfies notting, nirik, and the other FESCo member's criteria. 12:42
f13 I as a reviewer need to know the minimum things to check before allowing the import and build on the side of these packages. 12:42
tibbs I don't know how we can verify the other stuff when we can't even build the package. 12:42
abadger1999 Err... toss the page back to FESCo/java people to start using. 12:42
spot if it prevents java from dragging along prebuilt crufty jar crap, and we can catch it at this "prereview" stage, i think its a good thing 12:43
spot binaries of something solely for bootstrapping purposes, okay, fine. I understand that. 12:43
dbhole spot: Certain jars will still be pulled in at this preview stage, but only for bootstrap 12:44
tibbs You can't properly check that without building the package, which you can't at this stage. 12:44
dbhole spot: nvm, you already know that then :) 12:44
spot tibbs: well, you can check that the jars in the package match the package 12:44
tibbs Sure, you can look to see that it deletes some jars but that doesn't really tell you with any certainty. 12:44
spot e.g. no "eclipse-1.2.3.jar" in "jboss-super-awesome-snake" 12:44
* overholt calls dibs on "super-awesome-snake" project name 12:45
spot unless its documented clearly as bootstrapping (and jboss-super-awesome-snake has a proper dep on eclipse) 12:45
spot this falls under the "no system libraries" clause 12:45
spot which is why i think its important to be covered in the prereview 12:46
spot the builds from source is necessary to help explain the bootstrapping 12:46
* Rathann is looking at the relevant FESCO meeting log 12:46
spot and that the bootstrapping is TEMPORARY. :) 12:46
Rathann http://bpepple.fedorapeople.org/fesco/FESCo-2009-05-08.html 12:46
spot so, even with the shortcomings, this draft seems like a good enough plan for "prereviews" 12:48
spot +1 from me 12:48
abadger1999 +1 from me naturally 12:49
spot (even if i still don't think this is the FPC domain, i'd be happy to just pass these recommendations back to FESCo for implementation) 12:49
tibbs I'll go along with it, but I maintain that it's not possible to properly follow these guidelines if you can't build the package. 12:49
abadger1999 k. Build from source is too hard to verify? 12:49
spot looks like it is attempting to build from source? ;) 12:49
tibbs Indeed. 12:50
tibbs But +1 anyway. 12:50
rdieter +1 too 12:50
tibbs My primary complaint is that this screws with (and dumps on) the review process even more. 12:50
SmootherFrOgZ +1 12:50
spot tibbs: its a one-off, thankfully. 12:50
tibbs I mean, it's dysfunctional enough as it is. 12:50
Rathann 0 from me, I'm against such exceptions, even one-offs 12:50
spot if f13 had said it was a new practice, i'd have been much more concerned. 12:51
tibbs It should have been a zero-off, because there's a better way to do this that was ignored. 12:51
Rathann spot: this will get used as an argument of "we did this before" later, you'll see 12:51
tibbs The earlier representation was that this wasn't a one-off. 12:51
f13 representation by whom? 12:51
spot Rathann: well, its not going in the guidelines, we're just handing this back to FESCo. 12:51
spot with +5, it passes. f13, here you go. ;) 12:52
f13 thanks. 12:52
spot Next item: https://fedoraproject.org/wiki/MAN_pages_which_exists_in_other_places%28draft%29 12:52
tibbs -1. 12:52
<-- overholt has left this channel. 12:52
spot the English is pretty rough in this one 12:52
tibbs I simply disagree with the proposal in its entirety. 12:52
spot there are a lot of SHOULDs in there 12:53
spot but no MUST 12:53
tibbs We can document somewhere outside the guidelines that maintainers might want to look for debian-specific manpages. 12:53
tibbs But this directly contradicts our whole "do the work upstream" ethic. 12:53
* Rathann agrees with tibbs 12:53
spot yeah, i'm inclined to agree with this. 12:54
Rathann -1 12:54
spot it sure would be nice if debian could figure out how to submit things upstream. 12:54
spot -1 12:54
tibbs That doesn't stop our packagers from taking them and submitting them upsteam, of course. 12:54
abadger1999 0 12:54
tibbs But that doesn't need to be in the guidelines. 12:55
spot tibbs: i was about to make the same point 12:55
spot tibbs: or even carrying them while waiting for upstream to add them 12:55
tibbs It would be nice in general to have some documentation about how to find other distro patches and fixes. 12:55
spot (xpdf is carrying about 15 debian patches that upstream knows about) 12:55
tibbs I have no idea how to even fathom debian's packaging format. 12:56
spot tibbs: i could probably document that 12:56
* spot adds it to his todo list 12:56
tibbs I tried, but the patch of patches was just bizarre. 12:56
Rathann f13: who is on the "Java Maven team"? 12:57
spot theres a lot of crackrock there. 12:57
tibbs I have a performance review at 1PM, BTW, so I need to leave very soon. 12:57
rdieter tibbs: nod, makes me laugh whenever I hear folks tout .deb's being so much better than rpm 12:57
--> delero1 has joined this channel (n=delero@nat/sun/x-13744750215f7990). 12:57
spot okay, the only other item on the agenda is trying to find a better meeting time/day 12:57
spot https://fedoraproject.org/wiki/TomCallaway/FPC_Availability 12:57
delero1 just had a nice Xorg hang, with no ctrl-alt-del to escape it... 12:57
spot that has everyone's data except SmootherFrOgZ... *cough* 12:58
f13 Rathann: I don't have a full name list. dbhole might be able to come up with names. 12:58
spot delero1: then turn it on. silly. ;) 12:58
delero1 indeed 12:58
tibbs I hate to mention it, but didn't hansg join a committee when he explicitly could not make the meetings? 12:58
spot tibbs: he's made some of the meetings. 12:58
SmootherFrOgZ spot: i will 12:58
tibbs But the meeting time lies outside of the only times when he's available. 12:58
tibbs I mean, it's possible that his availability changed, sure. 12:59
tibbs Anyway, I'm logging but I need to go. 12:59
spot okay, we can talk about this item on the list 12:59
spot any other items for today's meeting? 12:59
Rathann my availability times are going to change next month, BTW 13:00
Rathann no idea to what yet 13:00
spot i think if we can find a time where 8 of 9 can regularly attend it will have to be good enough 13:01
Rathann f13: ok, it's fine if someone knows who can be held responsible for the mess ;) 13:01
spot we'll invite the other individual to vote via email in advance or as a tie breaker later 13:01
f13 Rathann: if all else fails, I will be responsible for it. I brought the request to FESCo in the first place. 13:02
Rathann I'll keep that in mind 13:02
* abadger1999 trusts f13 to clean up and to yell at relevant people if it is a mess ;-) 13:03
Rathann spot: sounds reasonable, I think that's the best we can do in the circumstances 13:03
spot okay, i think we're done for today, thanks everyone. 13:04
Rathann thanks spot 13:04