[23:53:52]  <halfline> f13: i'm sure we've had this conversation before,
   but i don't remember. why does rawhide point at the frozen repo instead of
   latest?
   [23:55:22]  <airlied> is there a static repo that works at the moment?
   [23:55:28]  <halfline> yea
   [23:56:13]  <halfline> and i would tell you if my ssd wasn't stalling my
   system
   [23:56:42]  <halfline>
   http://koji.fedoraproject.org/static-repos/dist-f11-build-current/x86_64/
   [23:57:07]  <airlied> ah dist-f11 I think I had dist-rawhide
   [23:57:41]  <halfline> airlied: yea i think that's basically my question.
   "Why is dist-rawhide not pointed at dist-f11-build-current"
   ...
   [00:03:03]  <f13> halfline: to make it easy for everybody to test the
   frozen bits
   [00:03:33]  <f13> and to keep the unwanted changes out of the way
   [00:03:51]  <halfline> f13: you mean so the beta gets some testing before
   it's released to the masses?
   [00:03:58]  <f13> right
   [00:04:08]  <f13> particularly the installation aspect
   [00:04:42]  <halfline> ahh
   [00:04:58]  <halfline> it's to make it easy to test things that are broken
   and are blocking the beta from going out
   [00:05:19]  <f13> and to verify that the "fixes" actually fix things
   [00:05:30]  <f13> without having other unwanted regressions or changes get
   in the way
   [00:06:48]  <halfline> so it's sort of a shame, because the number of
   people testing specific fixes is much smaller than the number of people
   that want to test rawhide
   [00:07:04]  <halfline> but i guess it beats having two separate trees
   nightly...
   [00:07:21]  <f13> before making two trees was nearly impossible due to the
   time it took
   [00:07:26]  <f13> these days that's less of a case
   [00:07:40]  <halfline> might be worth revisiting then
   [00:07:51]  <f13> I've been infrequently thinking about how best to
   provide two trees
   [00:08:03]  <f13> how to mirror it and such
   [00:08:31]  <halfline> clearly we need rawhide and rarehide
   [00:08:59]  <halfline> and we should rename rhel to welldonehide
   [00:09:24]  * halfline tries to cvs annotate that Requires(pre) and fails
   [00:10:15]  <f13> and name centos Naugahyde ?
   [00:10:59]  <mclasen> having some buffer zone between what gets pushed
   into koji and what lands in rawhide would be nice
   [00:11:02]  <halfline> f13: haha
   [00:11:30]  <f13> mclasen: I have a good plan for that actually
   [00:11:34]  <halfline> mclasen: hmm, that's sort of the other side of the
   coin
   [00:11:38]  <f13> mclasen: you can do koji builds with --skip-tag
   [00:11:50]  <f13> the build gets done, imported and whatever, it just
   doesn't get the tag applied
   [00:12:06]  <f13> you could download it, fiddle with it, do whatever, and
   when happy, apply the tag and it'll show up wherever
   [00:12:57]  <halfline> could we apply multiple tags to it?
   [00:13:03]  <f13>
   https://fedoraproject.org/wiki/Automated_QA_Testing_Project#Build_acceptance_testing
   [00:13:05]  <halfline> desktop-staging
   [00:13:14]  <f13> halfline: sure, any given build can have multiple tags
   [00:13:32]  <f13> like many of the builds right now have "dist-f11" tags
   as well as "f11-beta" tags
   [00:13:42]  <halfline> so we could have our own "get the latest greatest
   desktop stuff at your own risk" tag
   [00:13:48]  <halfline> and compose
   [00:13:59]  <halfline> and then when it's in good shape
   [00:14:06]  <halfline> tag it for the general rawhide masses
   [00:14:13]  <f13> yep
   [00:14:17]  <halfline> and likewise kernel could do the same thing
   [00:14:30]  <f13> there are logistics issues such as where to compose to,
   and what to do with the output
   [00:14:33]  <halfline> and then we wouldn't end up with unbootable rawhide
   days
   [00:14:41]  <f13> we'll still end up with those
   [00:14:43]  <halfline> or days where gdm crashes at startup
   [00:14:46]  <f13> due to merges
   [00:15:05]  <f13> or due to the "masses" not using the rawerhide stuff and
   missing the brokenness until it hits "mainline"
   [00:15:28]  <whot> mjg59`: thx
   [00:15:31]  <halfline> i guess it would just be a filter for the obvious
   broken for everyone stuff
   [00:15:33]  <f13> I think we'd avoid a lot of those scenarios if we get
   automated testing up and running
   [00:15:44]  <halfline> right
   [00:16:00]  <f13> which we can do as builds happen in theory, regardless
   of what tag they get
   [00:16:06]  <halfline> so maybe stuff should get tagged as desktop-staging
   or kernel-staging until it passes the unit test suite
   [00:16:10]  <halfline> and then maybe autotagged
   [00:16:17]  <f13> or just not tagged until it passes the tests
   [00:16:21]  <f13> (as the wiki page outlines)
   [00:16:32]  <halfline> oh i should probably read the link
   [00:17:38]  <halfline> yea page makes sense
   [00:17:53]  <airlied> how difficult would it be to automate an installer
   compose
   [00:17:58]  <airlied> and bitch if it fails
   [00:18:32]  <f13> not very difficult, but often the problem isn't the
   compose itself
   [00:18:36]  <halfline> installer is one of the few things that already has
   a framework for automated operation
   [00:18:52]  <f13> the problem is installing the results of the compose
   [00:18:54]  <halfline> (kickstart)
   [00:18:57]  <f13> right
   [00:19:09]  <airlied^ well boot it inside a kvm somewhere as well
   [00:19:11]  <f13> RHT has a ton of software built up to run such tests,
   but it's not free
   [00:19:25]  <halfline> lame
   [00:19:25]  <f13> so Fedora has to write their own
   [00:19:44]  <f13> or piggy back on RHT rewrites of the software and force
   it to go OSS
   [00:20:10]  <f13> james laska, will woods and I are working on the autoqa
   project, and will have more to show for it once beta goes out
   [00:20:34]  <f13> We've already got some simple code to notice when new
   repos show up to run repo sanity tests, such as dep closure and what not
   [00:20:49]  <f13> we're hoping to have simple kvm install testing hooked
   up quite soon as well
   [00:21:01]  <f13> cobbler/koan is a big help here
   [00:22:21]  <f13> I'm going to go kill some brain cells by playing video
   games.  g'night all!
   [00:22:22]  <halfline> does fedora have the hardware it needs to do the
   testing?
   [00:22:26]  <halfline> g'night
   [00:23:04]  <f13> it has some now, and budget for more later
   [00:24:00]  <mclasen^ are such staging tags free-for-all ? could I e.g.
   use a tag to collect all builds of an unstable gnome push to test them
   together before moving them to rawhide en masse ?
   [00:24:55]  <halfline> mclasen: yup, that's what we were talking about
   earlier
   [00:25:28]  <halfline> i guess we would have to cooridinate with him
   though
   [00:25:48]  <halfline> i don't think every step is self-service
   [00:25:56]  <mclasen> right
   [00:25:58]  <halfline> he'd have to fire off the compose
   [00:26:02]  <mclasen> I think that was my question