17:00:05 #startmeeting FESCO (2011-08-29) 17:00:06 Meeting started Mon Aug 29 17:00:05 2011 UTC. The chair is sgallagh. Information about MeetBot at http://wiki.debian.org/MeetBot. 17:00:06 Useful Commands: #action #agreed #halp #info #idea #link #topic. 17:00:10 #meetingname fesco 17:00:10 The meeting name has been set to 'fesco' 17:00:26 #chair notting nirik ajax cwickert mjg59 mmaslano t8m pjones sgallagh#topic init process 17:00:26 Current chairs: ajax cwickert init mjg59 mmaslano nirik notting pjones process sgallagh sgallagh#topic t8m 17:00:30 #topic init process 17:00:37 * nirik waves. 17:00:51 * notting is here 17:00:51 Afternoon 17:01:28 * jsmith lurks 17:01:52 * abadger1999 sits in the sheap seats 17:02:03 * sgallagh hands abadger1999 some shears 17:02:09 *grin* 17:03:05 I don't think we have quorum, if I'm counting correctly. 17:03:24 wonder if some folks are unburrying from the weather. ;( 17:04:05 We have a number of people who voted on the Java topic in the ticket, but I'm not sure that's sufficient (since it needs plenty of discussion) 17:04:38 well, if they are not around we could use their votes... 17:04:44 and the rest of us discuss. 17:05:14 ok, so: 17:05:18 #topic #663 Late F16 Feature Java7 17:05:20 .fesco 663 17:05:21 sgallagh: #663 (Late F16 Feature Java7) - FESCo - Trac - https://fedorahosted.org/fesco/ticket/663 17:05:22 * dbhole is still in the process of going through the 400+ Java packages to sort out which ones will not work with Java 7 as the default 17:05:48 my personal vote is: ship 7, but leave things built against 6. 17:06:00 and mass rebuild/fix for 7 in f17. 17:06:06 nirik: +1 to that, the way things look right now 17:06:12 nirik: I agree with that plan, but we need to know how feasible that is 17:06:24 Since some things have been built since J7 entered the buildroot 17:06:27 this would mean that we would have to rebuild some things for f16 that were built against 7. 17:06:32 yeah. 17:06:53 sgallagh: There haven't been that may afaik 17:07:03 one of those (icedtea-web) I altready rebuilt for 1.6 17:07:12 s/altready/already 17:07:26 and I commented on a bug for xalan-j2 to do the same 17:07:27 Has J7 been removed from the buildroot? 17:07:51 I expired the buildroot override immediately after last weeks thread.. is there anything else that needs to be done? 17:08:07 * nirik looks 17:08:31 dbhole: well, eventually the packge in stable will be 'available' for the buildroot 17:08:34 dbhole: what do most java packages require/ 17:08:37 java-devel ? 17:08:52 nirik: Yes, java for Requires and java-devel for BR 17:08:57 we need to make sure 1.7.0 does not provide whatever packages BuildRequire. 17:09:02 nirik: yes, the package is currently in stable. 17:09:10 so, its in the buildroot. 17:09:12 yep. 17:09:19 nirik: That would require a significant amount of rebuilds 17:09:30 nirik: Isn't it possible to blacklist it somehow? 17:09:31 Perhaps the approach should be to force Java7 to Provides: java7-devel 17:09:51 So that apps that are not compliant with J7 can 'Conflicts: java7-devel' 17:09:52 it would? 17:10:25 nirik: Ah sorry, you mean modify the java 7 package.. sorry nevermind my comment 17:11:11 dbhole: Right, so that in F16 it doesn't 'Provides: java-devel' 17:11:18 Okay, that is doable, having java-1.7.0-openjdk provide java7, java7-devel, etc... but if some package does BR it, it will still get into the root 17:11:21 So instead packages will pull in 1.6 17:11:25 a) change java-1.7.0 to not provide java-devel or anything that packages buildrequire for java. b) Push that and get it added to the buildroot asap c) identify those packages that were built with 1.7.0 and rebuild them with 1.6.0 17:11:27 Yes 17:11:47 if a package specifically wants java7, I would think thats ok. 17:11:56 ie, doesn't work with 6 for some reason. 17:11:59 sgallagh: Once it is in the root, it will do the building 17:12:04 nirik: c) As long as they were not INTENTIONALLY built against 1.7, in which case they need to update their [Build]Requires: 17:12:12 sgallagh: yeah 17:12:23 17:12:26 e.g. foo requires bar, and bar requires java7.. building foo will pull in bar which will pull in 1.7, causing foo to get built with 1.7 17:12:58 which means author of foo now needs to worry about it 17:13:00 dbhole: Hmm, good point. 17:13:23 yeah, so perhaps we should just drop the provides for now, and in f17/rawhide add provides java7-devel/java7 ? 17:13:26 dbhole: Can we modify the java 1.7 package to NOT ship /usr/bin/javac (or rather, only ship the versioned variant)? 17:13:30 (they could of course still require the package) 17:14:01 sgallagh: We can, but that would make it harder for anyone to use it 17:14:22 sgallagh, nirik: If it could be blacklisted from the buildroot, it can ensure that no package can ever pull it in for build 17:14:33 there's no way to do that I know of. 17:14:37 I am not sure how doable that is though 17:14:38 ah :( 17:15:03 nirik: Well, if we forcibly remove the 'Provides: java-devel', at least it won't happen by accident 17:15:30 * sgallagh wishes java separated the compiler from the runtime in subpackages 17:15:41 sgallagh, nirik: Yes, for now I think that may be the best way then.. removing all java and java-devel requirements altogether 17:15:51 Because then pulling in the runtime for a dependency wouldn't mean that the dependent package would build against 7 17:16:14 dbhole: yeah, remove provides in f16... but have them in f17... and you can work on mass rebuild there (in a side tag? ) 17:16:18 sgallagh: It does.. java-1.7.0-openjdk is the runtime, the -devel is the compiler 17:16:32 sgallagh: But there are packages that need -devel at runtime too :/ 17:16:53 Isn't that a violation of the package guidelines? 17:16:55 * sgallagh goes to look 17:17:13 sgallagh: some need it.. e.g. eclipse which is an IDE.. it is unusable without the compiler 17:17:27 * sgallagh sighs 17:17:34 nirik: I tried to rebuild all java packages on Friday.. 130 of 448 failed.. very few are due to java 7 though.. most are just broken :/ 17:17:35 That makes sense, of course 17:17:46 nonetheless, it is a pain to sort and I am sitll going through it 17:18:11 right, which is too much pain for f16. ;) 17:18:48 * dbhole agrees 17:19:07 I'm really wondering whether we should just give Java 7 the boot at this time. 17:19:29 sgallagh: well, it might be nice to ship it for people to test with/port code/try out? 17:19:31 nirik, sgallagh: Okay, so I will go ahead and do a rebuild to make java-1.7 no longer provides java or java-devel 17:19:41 That seems to be the safest 17:19:55 nirik: I understand that, but the chaining issue concerns me 17:19:56 sgallagh: I think F-17 is too far to not have java 1.7. till then.. people have been asking for it since pre-release days :/ 17:20:32 dbhole: It's too late to say this now, but we probably should have started a Python 3.0-ish behavior of parallel builds back in pre-release days. 17:20:32 proposed: remove provides from 1.7.0, add buildroot override, identify packages built with it and rebuild them for f16, work on java-1.7.0 mass rebuild in f17. 17:20:48 sgallagh: There were too many changes happening to the spec 17:20:59 sgallagh: The spec didn't finalize till mid-late july 17:21:05 they were tinkering with it till the end 17:21:22 nirik: sure, +1 17:21:23 nirik: +1 from me for that 17:21:39 does that match up with any votes from the ticket? 17:21:40 nirik: I'd like to amend that to say that the policy for F16 needs to be updated to disallow building packages for bodhi update against Java 7 17:21:52 So we don't have things chain-building into 7 accidentally 17:22:49 As an aside, we've been discussing this for fifteen minutes. Is it safe to assume we want to continue for another 15? 17:22:56 well, we could do that.. but might be hard to notice. 17:23:28 nirik, sgallagh: Can't buildsys be modified to conflict java-1.7.0-openjdk? 17:23:31 nirik: I'm not expecting people to necessarily follow it. I want that in place so that if (when) things break, we can point to it and say "This is why we told you not to do that" 17:24:05 That would prevent someone from accidentally pulling it in 17:24:13 It also gives us precedent in advance to block an update if it breaks things 17:24:25 dbhole: It also would prevent people from intentionally pulling it in 17:24:42 sgallagh: Isn't that what we want? prevent anything from building against 1.7 for F-16? 17:24:54 even intentionally 17:25:22 I don't know of any way to do that... aside from java-1.7.0 just not providing the devel bits... 17:25:26 dbhole: If we're preventing it intentionally, then why aren't we blocking Java 7 entirely 17:25:33 but then it's less usefull for people to test with. 17:27:01 I'm ok with announcing I guess that folks shouldn't build against java-1.7.0 in f16 for offical updates. 17:27:12 I'm fine with nirik's proposal, provided that we strongly advise against building packages against Java 7 if they have a runtime Requires: of java7-devel 17:27:21 nirik: Ah sorry.,. nevermind. Just went through the logs. In the old system, a buildsys rpm was installed and it pulled in the core root packages. I was refering to making that conflict with java 1.7. However it seems yum groups are used now 17:28:07 yes 17:28:33 nirik: I'm amending my previous statement of 'never build against J7' to "Don't build against J7 if it has Requires: java7-devel at runtime" 17:28:37 Then we should be okay, I think 17:30:04 I don't think we want anything to build against it in f16. 17:30:16 ok 17:30:17 because if foo requires bar and bar rebuilds against it, foo is messed up. 17:30:18 Okay, so then the resolution is to change all provides by java 1.7 to provide java7*, and have a Fedora policy against building for 1.7? 17:30:31 er, F-16 policy 17:30:41 dbhole: For official updates 17:30:58 sgallagh: Okay, that sounds good to me. 17:31:26 do we have enough folks to actually agree on something here? or add proposal to the ticket and collect votes? 17:32:21 #proposal Remove Provides: java[*] from Java 7, require that all F16 official updates are built against Java 1.6 17:32:58 nirik: Looks like this would match a +1 from mmaslano as described in the ticket 17:33:22 * nirik is +1 to that. 17:33:40 Yeah, I think I'm +1 17:33:41 The other comments appear not to directly apply 17:33:50 I'm +1 as well 17:34:25 notting: You out there? 17:34:34 seems ok 17:34:36 +1 17:35:06 ok, thats 5. 17:35:13 I count 4 +1 in here, with what I'm assuming to be another +1 in the ticket 17:35:21 * nirik nods 17:35:47 #agreed Remove Provides: java[*] from Java 7, require that all F16 official updates are built against Java 1.6 17:36:23 Anything else to discuss before we move one? 17:36:25 *on 17:36:28 I take it that the above applies to leaf packages as well? 17:36:41 e.g. icedtea-web -- nothing is going to rely on it 17:36:49 it provides a java plug-in and webstart 17:37:00 I am fine with going with 1.7 or 1.6.. just want to be clear :) 17:37:26 dbhole: Yes, the official updates must all be built on 1.6 and be capable of running on 1.6 17:37:35 sgallagh: Got it.. thanks 17:37:47 Anything else? 17:37:50 and do submit a feature for f17. ;) 17:38:09 dbhole: Do you want to take that as an action item? 17:38:15 Probably best to get the ball rolling now 17:38:27 sgallagh: Yes, I will build it before the end of the day 17:38:33 and submit the update 17:39:03 #action dbhole to update java-openjdk-1.7.0 to remove the Provides: 17:39:17 dbhole: I was also speaking of starting the F17 feature page 17:40:18 sgallagh: Sure. I will reuse the one we had for 16 17:40:39 #action dbhole to file Fedora 17 Feature Page for Java 7 17:40:49 sgallagh: Is there any other way to popularize it though, that we now ship java 1.7? it'd be nice to get the word out 17:40:56 for 16 I mean 17:41:12 Can it still go in the release notes? 17:41:16 I suppose it could be a prevew type thing in f16? 17:41:22 +1 17:41:22 tech-preview 17:41:31 yeah, that sounds good to me 17:42:17 Is there a process for it (tech-reviews) that I can read up on? or would that be just another feature page? 17:42:50 we have done it before in the past... 17:42:56 * nirik tries to recall details. 17:44:14 I guess add it as a release note and/or talking point? 17:44:17 I'm going to suggest we take this part outside the meeting 17:44:20 ie, talk to docs and marketing 17:44:58 sgallagh, nirik: Sure.. i will open a separate docs ticket for it then 17:45:17 thanks everyone! 17:45:22 Ok, shall we move on? There are no Eng Services tickets today, so I'll suggest we move to Open Floor. 17:45:33 Oh sorry, next week's chair 17:45:37 #topic Next week's chair 17:45:59 It should be noted that next Monday is a US Federal Holiday. 17:46:46 oh yeah, perhaps we should cancel? 17:47:27 How many of our members are US-based? 17:47:45 Probably enough to break quorum. 17:48:09 * nirik is, but will likely be on-line anyhow. ;) 17:48:40 * sgallagh wonders if anyone else is still here 17:48:52 i'll likely be out on mon 17:48:58 As will I 17:49:39 how about we ping all fesco folks in a ticket or list post... ask who wants to chair and if we will have quorum next week or should cancel. 17:50:49 +1 17:51:02 wfm 17:51:43 Ok 17:51:48 #topic Open Floor 17:52:03 Does anyone have anything for Open Floor? 17:52:24 * nirik does not 17:52:30 Nor I 17:53:16 not i 17:53:34 Ok, I'll end the meeting in one minute if there is nothing else. 17:54:43 #endmeeting