From Fedora Project Wiki
Fedora Release Engineering Meeting :: Monday 2009-02-23
General Summary
- the only difference between last week's discussion and this weeks is the introduction of 'dist-f11-rebuild' tag, and tagging every build into there before moving the whole kit over to dist-f11
- every box in the buildsystem is running rpm-4.6.0
- http://koji.fedoraproject.org/koji/buildinfo?buildID=83912 is the first build to use the new redhat-rpm-config
- EL-5 scratch builds are now possible
- we will need to work out how to do EPEL releng in bodhi
- post mass rebuild
- f13 following up on Orphan process
- Need a new meeting minutes taker starting at the next meeting (2009-03-03)
IRC Transcript
| f13 | ping: jwb rdieter lmacken spot wwoods notting poelcat warren | 10:29 |
|---|---|---|
| f13 | warren: do you still want to be pinged for these meetings? | 10:29 |
| jwb | here | 10:29 |
| * notting is vaguely here | 10:29 | |
| warren | f13: for the moment yes. | 10:29 |
| rdieter | yo | 10:30 |
| -!- f13 changed the topic of #fedora-meeting to: Fedora Release Engineering - Mass Rebuild | 10:31 | |
| f13 | short topic list today, only one. The rebuild. | 10:31 |
| f13 | koji outage is over, and redhat-rpm-config to turn on sha256 just got build | 10:31 |
| f13 | builds done after Mon, 23 Feb 2009 18:31:07 UTC will have sha256 | 10:31 |
| f13 | I'm going to let a build or two finish, and then let the script rip | 10:32 |
| f13 | any thoughts/comments on the rebuild sprocess? | 10:33 |
| f13 | the only difference between last week's discussion and this weeks is the introduction of 'dist-f11-rebuild' tag, and tagging every build into there before moving the whole kit over to dist-f11 | 10:33 |
| f13 | this will help save build loop issues with automated provides/requires on pkgconfig(foo) stuff | 10:34 |
| f13 | and add some logic to what gets tagged over to lower the liklyhood of an automated rebuild overriding a manual rebuild | 10:34 |
| warren | f13: what if they have the same N-V-R between the two? | 10:38 |
| f13 | pardon? | 10:38 |
| f13 | that's not possible. | 10:38 |
| warren | f13: what does "automated rebuild overriding a manual rebuild" mean? | 10:38 |
| f13 | the CVS system won't allow the same N-V-R to be used twice | 10:39 |
| f13 | warren: here is the scenario. | 10:39 |
| f13 | the automated rebuilds get put in a background build, which means they have the lowest prioirty | 10:39 |
| f13 | priority that is. | 10:39 |
| f13 | if at some point, after an automated build was submitted, the maintainer goes and does a further build of a pcakge for some other reason, the maintiainer issued build may get built first, at a higher priority | 10:39 |
| f13 | then the automated build will get done, and if it were tagged normally, it would appear as the "latest" build of that package. | 10:40 |
| f13 | hiding the maintainer done build. | 10:40 |
| warren | oh. because latest-pkg really is latest, not "newest" | 10:40 |
| f13 | so instead, we are tagging every automated build into dist-f11-rebuild and will be using a script I wrote to verify that the automated build was indeed the last build completed before tagging it over. | 10:41 |
| f13 | warren: that's correct. Latest == 'last tagged' | 10:41 |
| f13 | which is even more specific than the time the build was done | 10:41 |
| f13 | the only thing that matters to koji is when the tag action occurred. | 10:41 |
| warren | f13: the buildroot used to build dist-f11-rebuild will be dist-f11 without partially gcc-4.4 built packages? | 10:44 |
| f13 | there is no "buildroot" for dist-f11-rebuild | 10:44 |
| f13 | dist-f11-rebuild is just a holding tag. The builds themselves use the dist-f11-build buildroot, same as builds from dist-f11 | 10:44 |
| warren | but do dist-f11-rebuild tags go into dist-f11-build? | 10:45 |
| warren | A is built. B is built against A. A isn't the gcc-4.4 version. | 10:45 |
| f13 | no, they don't. | 10:46 |
| f13 | the goal isn't necessarily to have everything built against something that was built with gcc-4.4, the goal is just to have everything built with gcc-4.4, and with bigger rpm checksums, and with the new build flags. | 10:47 |
| f13 | in order to build everything against something that was built with gcc-4.4, we'd have to do multiple passes. | 10:47 |
| warren | I know. | 10:47 |
| warren | I'm just wondering if there are differences or problems if things are not built against gcc-4.4 built deps. Probably not. | 10:47 |
| f13 | matt's FTBFS will pick up some of that | 10:48 |
| warren | No real complaints here, was just wondering. | 10:49 |
| f13 | nod | 10:49 |
| dgilmore | please report odd build failures | 10:49 |
| dgilmore | there were alot of changes | 10:50 |
| * wwoods lurks | 10:50 | |
| f13 | dgilmore: we will | 10:50 |
| dgilmore | every box in the buildsystem is running rpm-4.6.0 | 10:50 |
| f13 | http://koji.fedoraproject.org/koji/buildinfo?buildID=83912 is the first build to use the new redhat-rpm-config | 10:50 |
| dgilmore | :) | 10:50 |
| dgilmore | after resolving some issues with F-9 and srpm chroots. all releases seem to be building fine | 10:51 |
| f13 | cool | 10:51 |
| dgilmore | not that it matters here EL-5 scratch builds are now possible | 10:52 |
| f13 | nice | 10:53 |
| dgilmore | we will need to work out how to do EPEL releng in bodhi | 10:53 |
| dgilmore | but thats post mass rebuild | 10:53 |
| -!- f13 changed the topic of #fedora-meeting to: Fedora Release Engineering - Open Floor | 10:54 | |
| f13 | anything else anybody woudl like to discuss? | 10:54 |
| dgilmore | i changed the task capacity of the ppc builders | 10:54 |
| mbonnet | dgilmore: do we want to rebuild something that gets pulled into buildroots (bash, for example) with larger hashes, to ensure that's working before the mass rebuild starts? | 10:54 |
| dgilmore | so please keep an eye on them also | 10:54 |
| f13 | I know that poelcat is no longer going to manage the meeting minutes, after beta, so we'll need to find a volunteer for that. | 10:54 |
| f13 | mbonnet: I can do that | 10:54 |
| dgilmore | mbonnet: bash was my local test case | 10:54 |
| mbonnet | f13: ok, seems like a good idea to make sure that's working | 10:54 |
| f13 | yep. | 10:55 |
| f13 | I'll do bash, and once there is a newrepo with it, I'll do pungi. | 10:55 |
| dgilmore | :) | 10:55 |
| f13 | if those work, I'll let the script loose on the package set. | 10:55 |
| mbonnet | f13: sounds good | 10:55 |
| poelcat | f13: orphan process? | 10:57 |
| * poelcat can't remember where we left off w/ it, but remembers seeing an email | 10:57 | |
| f13 | oh yeah, I had a couple people comment on fedora-devel-list saying that they were going ot do this | 10:58 |
| f13 | however I haven't seen any work done. I'll try to follow up. | 10:58 |
| mitr | (xmoto looks fine) | 10:59 |
| f13 | woo! | 10:59 |
| f13 | mitr: well it hasn't been imported yet | 10:59 |
| f13 | that's the real test | 10:59 |
| warren | f13: question | 11:00 |
| f13 | warren: shoot | 11:01 |
| warren | f13: how long will we wait between start of mass rebuilds and moving -rebuild to - | 11:01 |
| warren | f13: maintainers might be confused if they see commits and build success mail but a few days later the package isn't in rawhide | 11:01 |
| f13 | warren: basically I'm going to wait the builds from the first round to finish | 11:02 |
| f13 | and I don't know how long that will take. We have more builders than ever before, but we also have dynamic build weighing which we didn't have before, and we have more packages, more packagers, not sure what other load will be like | 11:02 |
| f13 | so it'll be a bit, but I'll be likely making some reports of the progress each day to fedora-devel or fedora-devel-announce. | 11:02 |
| warren | cool | 11:03 |
| f13 | We're at meeting end time, I'll happily discuss further in #fedora-devel | 11:03 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!