21:00:42 #startmeeting 21:00:42 Meeting started Wed Nov 28 21:00:42 2012 UTC. The chair is pwhalen. Information about MeetBot at http://wiki.debian.org/MeetBot. 21:00:42 Useful Commands: #action #agreed #halp #info #idea #link #topic. 21:00:42 #chair pwhalen jonmasters bconoboy ctyler pbrobinson dgilmore 21:00:42 Current chairs: bconoboy ctyler dgilmore jonmasters pbrobinson pwhalen 21:00:51 good day all 21:00:55 howdy 21:00:58 Howdy paul 21:01:03 .fas pwhalen 21:01:03 pwhalen: pwhalen 'Paul Whalen' 21:01:26 .fas Frojoe 21:01:26 Frojoe: jacwang 'Jordan Cwang' - burningfool 'Jordan Cwang' 21:01:42 .fas agreene 21:01:42 agreene: agreene 'Andrew Greene' - tag4fedora 'Tim Greene' 21:01:53 .fas ahs3 21:01:54 ahs3: ahs3 'Al Stone' 21:02:27 .fas darthjava 21:02:28 DarthJava: darthjava 'Dmitry Kozunov' 21:02:42 .fas jcapik 21:02:42 jcapik: jcapik 'Jaromír Cápík' 21:03:05 #topic 1) Current Problem packages 21:03:17 Peter sent an email to the list: texlive and ruby for both rawhide and F-18 are the current blockers. 21:03:18 The former is critical for rawhide and becoming more so for F-18 21:03:52 I removed the hsv builders from the largebuild channel, so texlive can build locally 21:04:45 #link ruby failure - http://arm.koji.fedoraproject.org/koji/taskinfo?taskID=1269138 21:06:04 is anyone currently looking at ruby? 21:06:16 or any volunteers? 21:07:05 .fas jonmasters 21:07:05 jonmasters: jcm 'Jon Masters' 21:07:13 .fas dmarlin 21:07:13 dmarlin: dmarlin 'David A. Marlin' 21:07:26 I could take a look 21:07:48 hey pwhalen going back one...on texlive, is the plan now to build it locally and import, or fix a builder to handle the 1.5GB srpm? 21:07:58 #action jcapik to look at ruby failure 21:08:03 tnx jcapik! 21:08:03 it looks like its failing during the tests 21:08:14 10283 tests, 1911515 assertions, 1 failures, 0 errors, 49 skips 21:08:22 jonmasters, I removed the hsv builders, its currently building on a local trimslice 21:08:43 pwhalen: then you'll manually import? I know pbrobinson was ok with that given it's likely to bump again 21:08:43 that build log is huge :D 21:08:43 however, armv5 its waiting 21:08:57 pwhalen: so it still needs an ARMv5 solution 21:09:00 its building in koji right now 21:09:13 there's a v5 builder for it too. perhaps it's just busy at the moment. 21:09:19 #link texlive build - http://arm.koji.fedoraproject.org/koji/taskinfo?taskID=1269858 21:09:23 pwhalen: I wonder if a solution is to ask dmarlin to setup an HSV highbank system with an sfp image to be a v5 builder 21:09:45 no buildroots on the v5 builder, perhaps a space issue? 21:09:47 jonmasters: no, squid seems to be part of the problem, so must be local 21:09:48 appears 13-1 is down 21:09:52 can somebody at seneca look at that? 21:10:03 enabled and checking it, not ready 21:10:08 13-1 will not be qavailable for another day 21:10:11 checking in 21:10:20 pwhalen: I can't ping it from hongkong 21:10:25 dmarlin: Peter and I discussed squid. I was going to ask Brendan (but haven't gotten around to it - thanks for the reminder) to check our squid is configured with a 2GB file limit 21:10:28 darthjava: why? 21:10:36 dmarlin: by default squid doesn't cache files that large 21:10:41 jonmasters: it's not a squid filesize limit. 21:10:47 odd, its still checking in 21:10:57 bconoboy: thoughts on what it is? 21:10:59 We have students working on it and currently in the process of moving them to free panda 21:11:15 after that it needs to be wiped and brought to spec 21:11:27 ya we've been strapped for hw as Chris's class has been using our trimslices 21:11:39 we're trying to get them back in order tho 21:12:13 jonmasters: no. dgilmore was going to investigate. 21:12:14 *have been 21:12:14 fossjon: how many trimslices are out of action? 21:12:18 bconoboy: ok 21:12:28 three are out at the moment 21:12:37 only one is in the farm 21:12:42 we're trying to get at least 1 back in working order jonmasters 21:12:48 which is 14-3 21:12:57 we need 2 in working order. 21:12:58 hopefully DarthJava can do that soon 21:13:11 fossjon, pwhalen: let's back up a topic then and ask what the current build system status is right now? 21:14:03 #topic 2) Koji status update 21:14:26 ok 21:14:30 bconoboy I'll get you another one around noon on friday. 21:14:40 DarthJava postgres update ^^^ 21:14:50 oh my... 21:15:16 Database performance issues are partially solved 21:15:42 Vacuum procedure stardted working, but autovacuum still not functional 21:16:02 i set up cron job to do manual vacuum meanwhile 21:16:22 what build hardware is out of action? 21:16:27 is it just trimslices? 21:16:33 hosts page has been great, thanks DarthJava 21:16:50 +1 21:16:59 trimslices and panda 5-2 21:16:59 yea, loads quickly now 21:18:04 hmmm...there does seem to be a good ratio of v5/v7 builders currently 21:18:05 jonmasters: seneca have re-purposed most of the trimslices for student stuff which is why we're one short. we'll have a second trimslice back on friday. 21:18:26 bconoboy: indeed, I actually didn't know that. I'll make a note to check the hosts page more regularly now it's loading quickly 21:18:30 if texlive builds on armv7hl successfully we'll know it's a squid issue. 21:18:35 ok 21:18:47 jonmasters: I didn't know that until just now either. 21:18:55 any other koji status updates? 21:19:05 looking at the other issues Peter mentions: 21:19:11 repo issues (the generally perl based build failures due to repo 21:19:11 issues). I reported I thought I had found the offending host but the 21:19:11 issue appears to have come back. Was the host re-enabled, what testing 21:19:12 has Seneca done? 21:19:23 * jonmasters doesn't want to sound irritated, because it's understandable that Seneca use their hardware as they like, but could we get a head's up next time, ctyler ? 21:19:33 i sent an email to the mailing list regarding peters problems because there were no details or examples in it to look at 21:19:52 #info Database performance issue partially solved by dumping/reloading db 21:20:07 #info autovacuum not yet running, but a cron job is doing regular vacuums now 21:21:43 Thanks fossjon read the email 21:22:21 I looked into the repo issue myself after the meeting last week and couldnt see that it was attributed to a specific host 21:22:42 pwhalen: Let's take the rest of that to email since the thread is already active 21:23:00 #action Followup on other koji issues will take place on arm-list where the thread has already begun 21:23:21 #topic 3) F18 ARM Beta - target release date, vfad planning 21:23:36 I'd like to suggest we shoot for F18 Beta in the next 7 days. 21:23:46 +1 21:24:09 The F18 primary beta went out yesterday so we should pull together what we have, resolve any pressing issues, and get it out. 21:24:09 when is a good time to vfad? 21:24:28 we then should look at doing the vfad this week, or early next depending on image creation 21:24:45 Once images are available- Suggest a tentative Friday. 21:25:17 bconoboy: we have issues of unsigned packages hitting the repos (mirrors) 21:25:33 packages need to be signed before images can be created 21:25:54 #action Image creation is stuck pending all packages on the mirrors being signed. 21:25:57 #undo 21:25:57 Removing item from minutes: 21:25:59 friday is not a good choice for people from Europe :] 21:26:01 #info Image creation is stuck pending all packages on the mirrors being signed. 21:26:07 Monday? 21:26:25 Monday sounds better :] 21:26:41 Friday is great, Monday is good. 21:26:58 I've got 2 mondays are good. Any objections? 21:27:07 * jonmasters notes that Dennis is going on vacation, so we should try to have beta images by Monday *latest* 21:27:08 so the package signing, will it be solvaed in a day, or a weekend? 21:27:27 (Other than iobjecting to monday on general princicple) 21:27:38 right, scheduling the vfad before we have images is tenative at best. 21:27:54 what's left for images, other than sfp? 21:27:54 Okay, let's shoot for Monday. That should be plenty of time to get images generated. 21:28:11 Right- so, assuming images are made, do we believe all beta blockers are resolved? 21:28:25 bconoboy: where's the list 21:28:26 pwhalen: did you get ahold of that Kirkwood cable you needed? 21:28:32 (of beta blockers) 21:28:35 we need some testing on the Kirkwood image 21:28:37 pwhalen: so you can test, right? 21:28:37 if the images ar emade before monday, I'd be happy to test one or two before monday. 21:28:39 jonmasters, not yet 21:28:57 ok, so who has a kirkwood they can test? 21:29:01 jonmasters i wasnt able to give pwhalen the cable he needed :/ 21:29:03 #action Tentative date of Monday, for F18 Beta Test Image VFAD 21:29:03 Seneca has one, the other is out with a student 21:29:05 you can blame me for that 21:29:13 specifically a guru plug 21:29:20 fossjon, np , perhaps someone there could test? 21:29:26 not all kirkwood share the same image 21:29:48 pwhalen: link to beta release criteria? 21:29:55 pwhalen is DarthJava able to test it? 21:30:10 fossjon, sure! 21:30:10 pwhalen what is needed to be done?? 21:30:19 I'll send an image 21:30:22 done then 21:30:33 I have a personal Dreamplug 21:30:46 jonmasters: IIUC that will need a different image 21:30:50 if nobody else can do it in time, I will test the Dreamplug over the weekend 21:30:52 #link Beta release criteria - https://fedoraproject.org/wiki/Architectures/ARM/Fedora_18_Beta_Release_Criteria 21:30:56 * dmarlin has no plug experience 21:31:34 pwhalen: do we have a list of specific things from alpha that needed to be fixed for beta? 21:31:55 we have a guy in the channel that used dmarlin's image creation wiki to build himself a plug image... I'm sure he's be happy to test... cannot think of hsi name right now 21:32:04 can we just agree on what else needs doing beyond kirkwood? 21:32:13 AFAIK we're good beyond that, right? 21:32:33 #link Alpha VFAD results - https://fedoraproject.org/wiki/Architectures/ARM/Quality_Assurance/2012-10-26-VFAD-Fedora_18_Test_Day 21:32:51 there were small issues with xfce, some of which appeared on PA 21:33:09 Only outstanding issue I"m aware of is the selinux relabel thing. 21:33:31 which you think is %post fixable 21:33:34 bconoboy: as soon as we can make images (signed packages) I'll test your suggestion 21:33:35 y 21:33:41 dmarlin: tnx 21:33:42 so if that's tested in an image, then we're good, right? 21:33:49 no blockers that I am aware of 21:34:01 jonmasters: is the prelink issue something we need fixed before beta? 21:34:04 I'm just of the opinion that in theory we ought to be able to get this done right now 21:34:20 if not sooner? ;-) 21:34:22 does dgilmore do the signing? 21:34:32 dmarlin, pwhalen: can you point me to the kirkwood images you made for testing? I can try on a dreamplug this evening, might be better than waiting 21:34:32 can anybody else do the signing? 21:34:57 dmarlin, pwhalen: I do have a SheevaPlug if necessary, though it's in use 21:35:01 jonmasters: once again, it is for a guruplug... dreamplug needs a different image 21:35:28 dmarlin: ah, yea, thanks for pointing that out 21:35:33 :) 21:36:00 masta: good question 21:36:04 ok, well never mind. Did someone commit to testing that on a guru? Was it masta? 21:36:24 Darthjava will test the image on the Guruplug 21:36:27 dmarlin: for prelink, it probably needs fixing. I'll make that priority number one, then the rc7 kernel issue 21:36:30 DarthJava can test an image for you 21:36:48 DarthJava: thanks 21:36:48 seems prelink is failing noisily rather than corrupting stuff, which is at least a benefit over the last time 21:36:53 :) 21:37:03 jonmasters: not me, but I was talking to a fellow in our IRC who was testing our images with his guru 21:37:52 masta: the image I created is based on Quentin Armitage's work on plug images 21:38:09 masta: IIRC he was testing on dreamplug 21:38:21 * masta stands corrected 21:38:29 masta: the images are slightly different 21:39:56 anything else for the Beta? I think we all agreed on Monday for the vfad provided images are ready? 21:40:08 so the plan is to get packages signed, prelink fixed, and make images, correct? 21:40:27 (in that order) 21:40:41 +1 21:41:16 is prelink a blocker for the beta? 21:41:19 jonmasters: is working prelink. who has ensuring packages are signed? 21:41:43 pwhalen: I'm not sure if it's a blocker... maybe nth 21:42:35 does anyone think it should be a blocker? 21:42:56 if it breaks things 21:43:16 afaik it doesn't actualyl have any impact... ti spews out stuff durring a kernel install... 21:43:23 it doesn't seem to and was noticed during the alpha 21:43:56 perhaps earlier even.. 21:43:58 * jonmasters has ordered a GuruPlug, with a JTAG adapter, on two day delivery 21:44:08 the fact that it was identified in alpha is what made me think it shuold be fixed in beta 21:44:16 jonmasters: why? 21:44:19 need certainty on having an option to test that before next week 21:44:37 jonmasters: I like how you operate, no BS... just be done 21:45:17 I will undertake, if prelink is the same issue again, to have that fixed this evening 21:45:25 then the rc7 kernel 21:45:36 I can get you a working prelink by the morning Eastern 21:46:06 so the prelink thing, does anybody think it causes any harm, besides wasting time redoing it's prelink thing... or to some extent even existing in the first palce? :-p 21:46:07 okay, then with signed packages we should be covered 21:46:32 masta: it can cause harm if prelink writes to a binary with the wrong linking data 21:46:46 masta: e.g. we had that before and on SPARC they had a really good one a few years ago 21:46:59 masta: so I should triage this at least now. I will come back if it's harmless noise 21:47:00 Monday 11am Eastern for the VFAD work for everyone? 21:47:13 if it's harmless but ok for beta I'll tell you within a couple hours 21:47:34 jonmasters: sounds great 21:47:43 #action jonmasters to fix prelink 21:47:44 thank you jonmasters 21:47:49 pwhalen: monday 11 EST, yes... looks good 21:47:51 #action jonmasters to fix 3.7 kernel 21:47:59 side note, my Google Chromebook arrived today. It's the carrot. If I get all these things done, I get to play with an A15 laptop ;) 21:48:25 #info Fedora ARM VFAD tentatively scheduled for Monday Dec 3rd 11am EST in #fedora-arm (dependent on image creation) 21:48:34 * masta has his on the way 21:48:48 they even sent me one with a SIM slot - that was super extra nice 21:49:10 #action pwhalen to email the list confirming VFAD 21:49:23 (anyway talk about that on #fedora-arm later) 21:49:28 ok, so we need somebody with package signing powers. 21:49:35 #topic 4) Raspberry Pi F18 Remix 21:49:49 does anybody want to say the best way to locate the packages to short-list for signing our images? 21:49:52 Is Seneca planning on doing an F18 remix? 21:49:52 well we know Dennis can do it and he's back Stateside in a day or so 21:50:11 Dennis is somewhere over the midwest even now. 21:50:14 * jonmasters is eager to test. My model B Pi arrived this week 21:50:20 I guess it woudl be scriptable to run through our repo and look for unhappy signing, right? 21:50:24 ...in a cute pink case 21:50:28 fossjon/darthjava: f18? pi? whatcha think? 21:51:15 * jcapik is still waiting for his model B .... 21:51:22 bconoboy, im working on the alpha image now 21:51:54 I should have an image by tomorrow 21:52:22 agreene, awesome.. 21:53:00 im still working on v6hl 21:53:10 ctyler and i are looking at the remaining problem packages 21:53:12 1500 left 21:53:15 xulrunner 21:53:21 thats about it for me 21:53:37 #info agreene composing F18 Alpha remix for the Raspberry Pi 21:53:43 agreene: armv5tel f18 rpi image? 21:54:06 bconoboy, yeah that is correct 21:54:12 cool 21:54:21 fossjon: you're doing f17 or f18? 21:54:53 um f17-v6hl as a base but we're trying to get a potential f18-v6hl release one day 21:56:17 has anybody measured the speed difference between v6hl and v5tel ? 21:56:19 #info fossjon working on f17-v6hl, about 1500 packages to go 21:56:47 agreene, please send an email to the list if the image is good for testing 21:57:21 pwhalen, thats on my todo list once its completed 21:57:31 we havent sped test anything yet 21:57:34 will it be for both models? 21:58:03 fossjon: are you able to compose an image yet? 21:58:26 fossjon: Would be good to know if anything is faster at all... bzip2, that sort of thing 21:59:00 that is true, i havent had a chance yet to see if we have a basic pkg set all completed 21:59:05 pwhalen, it will be for both models 21:59:11 i need to get a list of pkgs that we install on the pi and see if i have them built 21:59:15 awesome, thanks 21:59:16 the tests can be done with different distro .... using some benchmarks 21:59:18 im pretty sure i have most of the f17 ones ready 21:59:38 chris wanted me to do a sample f17-v6hl compose soon 21:59:42 fossjon: It'd be good to do some testing to find out if it really is worth the trouble 21:59:57 #action agreene to email the list once F18 Alpha Remix ready for the Raspberry Pi 22:00:14 we're at the hour, moving on 22:00:18 #topic 5) 3.7 Kernel and Device tree - support plan 22:00:45 I remember my benchmarks of i386 and i586 ..... surprisingly i386 builds were faster in most cases 22:00:57 on 3.7 22:01:02 and smaller :] 22:01:02 so I will fix the rc7 problem 22:01:07 final 3.7 is coming this weekend 22:01:21 the variant kernels will still exist so we should expect those as the upgrade path in F18 22:01:40 there will also be a unified kernel package that folks can test, but I expect per-variant upgrades for most users still in F18, with a unified kernel in F19 22:01:51 unless Peter has re-thought that, that was the understanding we agreed recently 22:02:14 that sounds good to me. keep it consistent until f19 22:02:15 however, we should test whether a unified kernel upgrade works, once we have a 3.7 kernel 22:02:18 so the tentative plan for F18 with 3.7 is to ignore device tree? 22:02:21 dmarlin: that's the plan 22:02:37 bconoboy: well no, just not to require folks to move to a unified kernel 22:03:08 bconoboy: we may need per-variant device trees, and we'll need to look at the impact of that once we have kernels 22:03:25 bconoboy: so there is potential for some impact. Good news is we should have a 3.7 kernel for testing well before F18 final 22:03:35 The 3.7 kernel should boot a armada 370 SoC, right? 22:03:38 so we can know if we need to e.g. ship some device tree thing in the final images 22:03:49 * masta is thinking to get a miraboard to test the unified kernel 22:04:32 aka with DT different than highbank 22:04:53 masta: I don't know how well the 370 is supported with drivers, but the core code is there 22:05:22 bconoboy: so in summary, we keep the variants, but we might have devicetree stuff that's needed to be in the final F18 image in preparation for future upgrade to a 3.7 kernel 22:05:43 bconoboy: and determining that should be the priority post-beta as I mentioned to you privately yesterday 22:06:09 #info F18 3.7 plan: keep the variants, but we might have devicetree stuff that's needed to be in the final F18 image in preparation for future upgrade to a 3.7 kernel 22:06:31 ok so that means we need a device-tree-blobs package made 22:06:51 masta, I think dgilmore is working on that one 22:07:28 yes 22:07:28 #topic 6) Mirror syncing - missing packages on the mirrors 22:07:44 we 22:07:54 we're in OT so speeding it up a little 22:08:56 so the secondary arch are providing rsync access to the down stream mirrors, right? 22:09:19 maybe some of them hang durring their rsync 22:09:40 do all the mirrors have missing package, or just some of them? 22:10:47 my understanding was there were packages on the master mirror that have yet to be sync'd, an example being the latest Kernel 22:11:20 well as long as that does not impact image creation 22:11:26 ... 22:11:45 pwhalen: I had seen that, but not sure if it is still an issue 22:11:53 dmarlin: have you seen this on the mirror you use for images? 22:12:02 dmarlin, ok 22:12:23 #topic 7) Your topic here 22:12:31 masta: I saw it on several mirrors at one point, but it may have been addressed 22:12:54 masta: it may have been due to some outage 22:14:06 anything else for today? 22:14:14 pwhalen: please, action me to do few v6hl-v5tel speed tests 22:15:01 jcapik, there has yet to be an image for v6hl... 22:15:23 pwhalen: I can do that on different distro 22:15:49 the v6hl koji repo is publically accessible :) 22:15:49 pwhalen: AFAIK ppisar has a running v6hl gentoo 22:15:56 via japan.proximity.on.ca/koji 22:16:23 pwhalen: I'm curious if it is worthy 22:17:25 #action jcapik to do initial testing on v6hl vs v5tel 22:17:34 pwhalen: thx 22:17:39 np :) 22:17:50 any other action items requests ? 22:17:58 pwhalen: thanks, good meeting 22:18:07 #endmeeting