fedora-meeting-1
LOGS
20:00:31 <pwhalen> #startmeeting
20:00:31 <zodbot> Meeting started Wed Jul 25 20:00:31 2012 UTC.  The chair is pwhalen. Information about MeetBot at http://wiki.debian.org/MeetBot.
20:00:31 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
20:00:39 <bconoboy> .fas blc@
20:00:39 <zodbot> bconoboy: blc '' <blc@redhat.com>
20:00:41 <pwhalen> .fas pwhalen
20:00:42 <zodbot> pwhalen: pwhalen 'Paul Whalen' <pwhalen@redhat.com>
20:00:43 <dmarlin> .fas dmarlin
20:00:48 <zodbot> dmarlin: dmarlin 'David A. Marlin' <dmarlin@redhat.com>
20:00:58 <djdelorie> .fas djdelorie
20:00:59 <zodbot> djdelorie: djdelorie 'DJ Delorie' <dj@redhat.com>
20:01:21 <pwhalen> #chair pwhalen jonmasters bconoboy
20:01:21 <zodbot> Current chairs: bconoboy jonmasters pwhalen
20:01:50 <pwhalen> #topic 1) F18 Mass rebuild status update
20:02:22 <fossjon> https://twitter.com/FedoraARMStats/status/228215255627616256
20:02:28 <fossjon> 76.3% complete!
20:02:32 <pwhalen> pbrobinson was unable to join us today, but shared the rebuild is coming along as expected. Not sure if anyone else has something they would like to discuss
20:02:45 <jcapik> .fas jcapik
20:02:47 <zodbot> jcapik: jcapik 'Jaromír Cápík' <jcapik@redhat.com>
20:02:48 <bconoboy> #link https://twitter.com/FedoraARMStats/status/228215255627616256
20:02:56 <djdelorie> when do we expect to "catch up" again?
20:02:56 <bconoboy> #info according to the link, we're 76.3% complete
20:03:01 <pwhalen> that is encouraging
20:03:14 <fossjon> assuming my calculations are correct, i just did a list-tagged ratio
20:03:21 <fossjon> for f18 no inheritence
20:03:24 <fossjon> :)
20:04:13 <bconoboy> we're doing 500 or so builds a night
20:04:27 <bconoboy> so if all goes well we'll be done in about 6 or 7 more nights
20:04:51 <bconoboy> #info builders are doing 500-600 builds per night
20:04:57 <djdelorie> estimated ratio of "PA build time" to "ARM build time" then?  Overall, I mean?
20:04:58 <bconoboy> we could use more builders
20:05:07 <dmarlin> bconoboy: do we have all available builders working on this?
20:05:14 <pwhalen> so lets all just keeping on eye on koji and hosts
20:05:44 <fossjon> well according to a list-tagged, there's 11781 pkgs in f18
20:05:45 <bconoboy> dmarlin: We have one more panda board we could add, but that wouldn't really make a meaningful difference
20:05:51 <fossjon> source rpms i believe
20:06:26 <dmarlin> bconoboy: ok, running near 100%.  good.
20:06:31 <jonmasters> .fas jcm
20:06:32 <zodbot> jonmasters: jcmoore 'Curt Moore' <jcmoore@nuvio.com> - jcmartin 'James C Martin' <jcmarti4@gmail.com> - jcmoralesc 'Juan Morales' <jcmoralesc@ice.co.cr> - nagarajcm 'Nagaraj' <nagarajcm@gmail.com> - jcmontero 'Juan Carlos Montero' <jcmontero@uach.mx> - jcmannam 'MANNAM JAYACHAND' <jcmannam@gmail.com> - jcmex '' <dedocmc@hotmail.com> - jcmcderacg 'James McDermott' <jmcdermott@acg-hq.com> - jmasters 'Jon Masters' (1 more message)
20:06:32 <bconoboy> if seneca have any extra builders they can bring online that'd help
20:06:33 * jonmasters is in
20:06:40 <bconoboy> ctyler: anything pending?
20:07:10 <pwhalen> ctyler wasnt sure if he could join us today either.. fossjon, Frojoe, amaxm ^
20:07:32 <fossjon> our team is supposed to respond on behalf of chris
20:07:42 <fossjon> im trying to ask them what builds arent in the farm
20:07:49 <fossjon> builders*
20:07:50 <maxam> bconoboy: we have some very weak efikas that are going down consistantly
20:07:55 <jonmasters> fossjon: hey, did you setup those FedoraARMStats just since I suggested it yesterday? That was fast!
20:08:16 <maxam> bconoboy: other than those we are at full capacity
20:08:21 <bconoboy> maxam: okay, definitely not those
20:08:22 <fossjon> jonmasters heh ya, i was bored last night
20:08:27 <fossjon> found a twitter based py script
20:08:32 <jonmasters> fossjon: kudos!
20:08:33 <fossjon> library*
20:08:36 <bconoboy> pwhalen: sounds like that's it
20:08:47 <pwhalen> #topic 2) Raspberry Pi remix status
20:09:09 <pwhalen> fossjon, agreene - how goes it?
20:09:24 <Frojoe> Agreene has made some fixes to a bunch of packages in our override repo, and we've just composed a new image that fossjon is testing now
20:09:49 <Frojoe> We've had some blockage on our test armv6 build for the pi, so we're holding of on that for the meanwhile
20:09:52 <pwhalen> is that image ready for those here with Pi's to play with?
20:09:56 <djdelorie> can/should any of the remix work be rolled into our nightlies?
20:10:13 <bconoboy> frojoe: pointer to the most recent test image?
20:10:21 <Frojoe> Can host it shortly
20:10:38 <bconoboy> frojoe: ctyler published one yesterday, would be good to add to the minutes
20:10:47 <Frojoe> Righto
20:10:48 <fossjon> we'll scp it to scotland, we're testing the latest image now
20:10:59 <bconoboy> URL?
20:11:03 <fossjon> i fixed up our firstboot problems and hopefully the other scripts work
20:11:20 <fossjon> bconoboy just one min :)
20:11:27 <bconoboy> #info f17 nightly images are not being updated, it is better to use the seneca test images
20:11:28 <fossjon> we're choosing a dir on scotland
20:12:01 <bconoboy> #link ftp://ftp.ges.redhat.com/raspberrypi/
20:12:17 <bconoboy> #info this is a mirror of the scotland test images seneca is producing
20:12:27 <fossjon> thanks bconoboy :)
20:12:48 <bconoboy> fossjon: I'm not sure what the equivalent url is for scotland though
20:13:10 <bconoboy> oh, here we are
20:13:12 <bconoboy> #link http://scotland.proximity.on.ca/raspberrypi/
20:13:23 <bconoboy> #info This is the primary point where new images are being released
20:13:48 <bconoboy> fossjon/frojoe: What is the plan for these images? IE, when will there be a final? or a beta?
20:13:52 <jonmasters> bconoboy: thanks for the mirror setup
20:14:41 <fossjon> well we're testing the latest image now, im lagging behind on this splash screen as it seems to be taking up a lot of resources on boot and slowing things down
20:14:41 <bconoboy> the ges mirror is updated nightly so it won't always be as current, but it's stupid fast.
20:14:56 <fossjon> for the most part it seems fairly stable in text mode at least and probably gui mode
20:15:25 <fossjon> I think if this image works smoothly (verified by chris) that we'll either ship it or make one more final one just to be sure
20:15:25 <bconoboy> fossjon: do you want feedback on user experiences of these test images?
20:15:42 <Frojoe> Some user feedback would be nice
20:15:55 <bconoboy> #info 1-2 more test images likely to come out before ready for general consumption
20:15:55 <maxam> http://scotland.proximity.on.ca/raspberrypi/test-releases/raspberrypi-fedora-remix-17/latest/
20:15:56 <fossjon> we'll have the link in one min so that everyone can test it
20:15:57 <Frojoe> We're kind of limited in our time to test this, alongside doing other things
20:16:01 <bconoboy> #info pelease test and provide feedback
20:16:03 * jonmasters will test one before the weekend on my Pi
20:16:05 <fossjon> ^^^ there it is :)
20:16:09 <maxam> #link http://scotland.proximity.on.ca/raspberrypi/test-releases/raspberrypi-fedora-remix-17/latest/
20:16:18 <fossjon> on another note, I may need someones help with armv6hl
20:16:27 <fossjon> thats the last task im currently failing at
20:16:33 <jonmasters> fossjon: it's the "test-releases" I want, right?
20:16:37 <bconoboy> pwhalen: next topic? (perhaps fossjon's)
20:16:39 <pwhalen> #action - those with Raspberry Pi's please download an test the latest image. Provide feedback in #fedora-arm
20:16:46 <Frojoe> The link will be the correct one shortly
20:16:52 <Frojoe> The image is still copying over
20:17:02 <pwhalen> #topic 3) Your topic here/open floor
20:17:13 <jwb> i have a quick one
20:17:24 <bconoboy> fossjon: what's this about armv6hl?
20:17:44 <fossjon> sorry i didnt mean to jump topics
20:17:51 <fossjon> it was related to my status on the rasp pi
20:17:59 <jonmasters> bconoboy: they're considering a VFPv2 ABI based rPi rebuild for v6
20:18:05 <fossjon> umm i cant seem to get the rcs/macros set right for armv6hl on a armv7hl builder
20:18:21 <fossjon> even tho i copied a lot of them from armv7hl to try and duplicate it
20:18:31 <jonmasters> I guess pretty much everything v6 has a VFPv2 so that can make sense
20:18:46 <fossjon> mock/rpmbuild arent pulling in the set variables needed during build for example %initrddir
20:18:59 <bconoboy> fossjon: so this is a problem with the rpm macros?
20:19:24 <fossjon> im not sure where this is failing, but it *feels* like mock or rpmbuild isn't seeing my armv6hl macros/rc files
20:19:36 <fossjon> because a bunch of the build env variables are reporting empty or non-existentt
20:19:50 <fossjon> i dont have that much knowledge on what all the little bits are needed for mock/rpmb to work
20:20:00 <bconoboy> dgilmore did the changes for armv7hl originally, you should ask him for help
20:20:04 <jonmasters> fossjon: I thought Peter was looking at that for you?
20:20:19 <bconoboy> I think peter just provided the right flags for gcc
20:20:19 <fossjon> i asked pbr for gcc flags but not really about mock
20:20:29 <fossjon> i can ask dgilmore when he gets back
20:20:33 <fossjon> :)
20:20:38 <jonmasters> ok, yea, email Dennis. He's in Australia for a while, so email him
20:20:39 <fossjon> chris also said he could help me possibly
20:20:52 <bconoboy> anybody else want to volunteer to help fossjon?
20:21:06 * jonmasters would love to but I don't have time this week
20:21:38 <bconoboy> #action fossjon to ask dgilmore for help on armv6hl rpm macros
20:21:43 <Frojoe> The link has been updated
20:21:52 <Frojoe> The new rpfr compose is up
20:21:57 <fossjon> I think we got a link for everyone :)
20:22:16 <djdelorie> fossjon: are you folks willing to use distcc/cross to speed up the builds?  Or are you sticking with "honest" builds?
20:22:59 <fossjon> djdelorie to be honest im not really sure, chris just assigned me the task and i assumed (maybe wrongly) that i could use one of our "bigger" armv7hl machines to compile armv6hl with mock
20:23:18 <fossjon> but im open to any kind of ideas / thoughts that could be automated
20:23:34 * jonmasters was discussing optimizing mock with pbrobinson last night, so that it could be taught to do cross-building for some architectures during bootstrap, etc.
20:23:35 <fossjon> you guys know more about this stuff than i do :)
20:23:37 <djdelorie> the latest bootstrap scripts know how to build and set up a distccd on the fast host, but I haven't pushed that into an RPM for mock to use
20:23:43 <bconoboy> djdelorie: they're trying to build armv6hl using an armv7hl environment rather than boot strap from the beginning
20:23:57 <djdelorie> right, but they still need to rebuild everything with an armv6 compiler
20:24:01 <bconoboy> #info (Seneca is working on an experiemtnal armv6hl build for the Pi)
20:24:12 <djdelorie> the bootstrap scripts make the *right* compiler, and run it with the native distcc without hacks
20:24:37 <fossjon> i dont have that much knowledge of gcc but i assumed that an armv7hl builder could compile code for a vfpv2 based machine
20:24:41 <jonmasters> I think for now, fossjon, I would advocate a simple non-distcc approach because it's more straightforward. But if you're after fun research projects, djdelorie's idea is a fun one
20:24:41 <djdelorie> so "all you need" is to add an rpm to mock that sets up $PATH
20:25:20 <djdelorie> I suspect using arm armv7hf gcc might "accidentally" build you an armv7hf binary once in a while.  Not all builds are sane.
20:25:28 <fossjon> ok, i can look into other options if im still failing at it :)
20:25:57 <jonmasters> :)
20:26:08 <bconoboy> we set with this topic for now?
20:26:10 <bconoboy> jwb?
20:26:22 <jwb> one sec
20:26:37 <bconoboy> I have one if more than a sec is needed
20:26:39 <jwb> sorry.  quick statement/question
20:26:58 <jwb> 1) i'm removing arm support from the kernel in f16 unless someone comes up with a really good reason why i shouldn't
20:27:19 <jwb> 2) are you guys actively building and distributing f17 kernel updates for arm?
20:27:29 <djdelorie> how about "why waste your time doing it?" ;-)
20:27:38 <bconoboy> jwb: We are actively building fedora 17 update kernels, yes.
20:27:45 <jonmasters> but not f16
20:27:48 <jwb> djdelorie, because arm wastes my time by being present in f16 when we rebase.  a lot of time.
20:27:56 <bconoboy> jwb: I don't think we care about any kernel earlier than 3.4.
20:28:18 <jonmasters> jwb: we don't care about f16 kernels so feel free to not care
20:28:40 <jwb> jonmasters, my form of not caring is removing it because otherwise it eats time for the kernel team.
20:28:47 <jonmasters> does anyone object?
20:28:52 <bconoboy> no, do it
20:29:13 <jonmasters> #agreed f16 kernel support for arm to be removed (not that we ever supported f16 anyway)
20:29:18 <jwb> ok, thank you.  f17 is getting rebased to 3.5 this week.  i'm sure the configs will need looking at.
20:29:19 <pwhalen> #agreed - remove arm from the kernel in F16. Its not being used.
20:29:43 <jonmasters> jwb: yea, there's some possible fallout from that switch to 3.5 in terms of device tree. We'll let you know.
20:29:43 <bconoboy> okay, next topic: linker path in fedora 17
20:29:59 * jonmasters suggests we drop it
20:30:01 <bconoboy> It's been 4 weeks since we moved to the new linker path in f18
20:30:05 <jonmasters> no need to change the path in F17
20:30:18 <pwhalen> #topic - linker path in f17
20:30:19 <bconoboy> Evidently upgrades from the f17 glibc to the f18 glibc do not work
20:30:31 <bconoboy> #info upgrades from the f17 glibc to the f18 glibc do not work
20:30:44 <jonmasters> we'll blow something up. We'll have all the best intentions, but something will go wrong and we'll break f17 systems for little gain
20:31:04 <bconoboy> shouldn't we support updating from f17 to f18?
20:31:08 <jonmasters> we have an f18 mass rebuild with it all nicely switched over ready for f18
20:31:29 <bconoboy> because we can't do that right now.
20:31:33 <jonmasters> bconoboy: sure, that needs fixing, and I mailed Jeff about it, but I dont' think we need to switch the linker path in F17 absent a compelling justification
20:32:10 <bconoboy> jonmasters: I think we all agreed 4 weeks ago that we would make the switch if we could.  Are we changing our minds now?
20:32:29 <jonmasters> it's one thing to make upgrades to F18 work (they should, and they don't) but it's another to change the linker path for F17 when we're on track to have an F18 release with it switched
20:32:52 <bconoboy> What has changed between now and 4 weeks ago?
20:32:54 <jonmasters> I'll let others speak, but my suggestion is if it aint broke...
20:33:13 <djdelorie> fyi, the F17 bootstrap scripts have to put in the symlink because there's a disconnect between what's built and what's expected.
20:33:28 <djdelorie> This doesn't affect the RPMs we use, but it points out that it's not "all one way by default" in F17
20:33:30 <bconoboy> but it is broke. we can't run binaries in f17 from other diistros.
20:34:34 <jonmasters> so, I've nothing against making sure a symlink is in place on F17 from the new ld-linux-armhf.so.3 to the legacy ld-linux.so.3 but I object to switching over
20:34:39 <jwb> jonmasters, confused.  3.5 has been sitting in rawhide for weeks.  why wouldn't it already be solved there?
20:34:56 <jwb> jonmasters, sorry, backtracking on your device tree comment)
20:35:19 <jonmasters> by having a symlink in the reverse direction we can run binaries from other distros in F17, we just can't necessarily compile something that will run on e.g. Ubuntu
20:35:37 <jonmasters> (unless the Ubuntu user installs one of their compatibility packages AIUI)
20:35:53 <bconoboy> jonmasters: no, you need a glibc patch to run at all, even with a reverse symlink.
20:36:07 <jonmasters> oh for the dynamic linker, true, yes
20:36:14 <bconoboy> so it *is* broken
20:36:21 <jonmasters> I agree, that is broken
20:36:53 <bconoboy> So I think *after* the symlink problem is fixed, we should update f17.
20:36:55 <jonmasters> I just disagree on the need to switch the primary linker location. I think I favor a symlink to the new ld-linux-armhf.so.3 from the legacy one on F17 and then a fix to the linker to make that work
20:37:24 <bconoboy> That's fine, we still need a patched glibc
20:37:38 <jonmasters> but if you're totally convinced we need to switch over, we could. I think it'll break something we haven't anticipated, so I prefer the lower risk option
20:37:46 <jonmasters> bconoboy: ok, so we agree on that plan?
20:38:19 <bconoboy> Well, if the plan is:
20:38:33 <bconoboy> 1. Get the symlink problem fixed in f18 so f17->f18 upgrades work
20:38:39 <jonmasters> jwb: more testing has been done on F17 than on rawhide for various boards. It ought to be ok, I'm just saying I know a whole boatload of devicetree stuff landed in 3.5 so a head's up that we might have some cleanup
20:38:50 <bconoboy> 2. Update f17 glibc to accept either linker path
20:39:11 <bconoboy> 3. Update f17 glibc to provide a symlink to the legacy lnker path for f17 only
20:39:22 <bconoboy> If those 3 things are true, I agree :-)
20:39:37 <jwb> jonmasters, and i'm saying that cleanup should have already been done in rawhide.  unless you're essentially ignoring rawhide kernels and using f17 as basically the tip of the spear
20:39:46 <jonmasters> where 3 means /lib/ld-linux-armhf.so.3 -> /lib/ld-linux.so.3 then I violently agree with you :)
20:39:53 <bconoboy> yes, exactly
20:40:09 <jonmasters> I just lost all power, brb
20:40:14 <jonmasters> (UPS units kicked in)
20:40:16 <bconoboy> Okay, let's get this in the minutes...
20:40:43 <bconoboy> #action Update f18 glibc to allow f17->f18 glibc rpm upgrades
20:41:10 <bconoboy> #action Update f17 glibc to accept either ld-linux-armhf.so.3 or ld-linux.so.3 linker names to allow cross-distro compatibility
20:41:32 <bconoboy> #action Update f17 glibc to provide a symlink from lib/ld-linux-armhf.so.3 -> /lib/ld-linux.so.3
20:41:56 <bconoboy> any objections?
20:41:58 <pwhalen> #topic 3) Your topic here/open floor
20:42:05 <pwhalen> bconoboy, sorry
20:42:05 <fossjon> just to quickly clarify from before since we were fumbling, the link for the new image we're testing should be ready -> http://scotland.proximity.on.ca/raspberrypi/test-releases/raspberrypi-fedora-remix-17/latest/rpfr-727.img
20:42:17 <fossjon> just wanted to say that again in case anyone was confused
20:42:40 <jonmasters> ok, breakers reset
20:42:47 <pwhalen> #link Current Raspberry Pi test image -  http://scotland.proximity.on.ca/raspberrypi/test-releases/raspberrypi-fedora-remix-17/latest/rpfr-727.img
20:42:55 <fossjon> thanks pwhalen
20:43:21 <jonmasters> jwb: we're not ignoring rawhide kernels, but we are still working on a nightly testing process there
20:43:42 <jonmasters> jwb: personally, I'm making sure my boards are running F18, and I know others are too...it's not perfect, it's getting better
20:43:56 <jwb> are those boards running rawhide kernels too?
20:44:04 <jwb> because if they are, they should already be on 3.5
20:44:06 <bconoboy> #agreed f17 linker path partial update plan is a-go
20:44:26 <jonmasters> jwb: yes, I have 3.5 kernels running, I'm just saying there might be some boards that haven't had as much coverage
20:44:35 <jonmasters> jwb: and that's improviing
20:45:19 <pwhalen> any other items for today meeting? the floor is open
20:45:42 <jonmasters> pwhalen is doing a good job so far on QE. Hey, Paul, can you ping jwb after the meeting to get input on things we can improve for testing rawhide kernels?
20:45:52 <pwhalen> jonmasters, will do
20:46:02 <bconoboy> I think we're all set
20:46:09 * jonmasters discussed this with rmk too when I saw him (the need to get good ongoing rawhide coverage)
20:46:38 <jonmasters> jwb: please work with pwhalen to ensure your concerns are at least well known though I think we get it already
20:47:04 <jwb> if pwhalen isn't responsible for the kernel configs, that might not be the most productive
20:47:13 <jwb> but i'd be happy to discuss it anyway and with whomever
20:47:44 <bconoboy> pbrobinson and dgilmore are primarily responsible for the kernel configs
20:48:56 <jwb> pwhalen i'll bug you about different but related things in the arm channel
20:49:15 <pwhalen> jwb, sounds good, and I appreciate any input you may share
20:49:23 <pwhalen> thanks all!
20:49:27 <pwhalen> #endmeeting