IRC logs for #baserock for Thursday, 2015-05-07

*** zoli__ has joined #baserock05:04
*** zoli__ has quit IRC05:47
*** zoli__ has joined #baserock06:06
* straycat meows06:58
*** paulw has quit IRC07:07
*** paulw has joined #baserock07:07
*** sambishop has joined #baserock07:13
*** Albert_ has joined #baserock07:20
*** a1exhughe5 has joined #baserock07:25
*** Albert_ has quit IRC07:31
*** Albert_ has joined #baserock07:33
*** bashrc has joined #baserock07:54
*** mariaderidder has joined #baserock07:59
pedroalvarezOk, looks like the problem with gcc5.1 is syslinux08:12
*** rdale has joined #baserock08:14
pedroalvarezmy experiment:08:14
pedroalvarez1) using a 15.10 devel system, upgrade morph, override syslinux with the artifacts of syslinux built with gcc5.1, deploy a base system08:14
pedroalvarez2) base sytem doesn't boot08:14
juergbipedroalvarez: in case this helps, syslinux 6.02 works for me when built with gcc 5 but syslinux 6.03 doesn't08:17
persiaCool.  Can we use an EFI bootloader?08:17
pedroalvarezjuergbi: that might be relevant :)08:20
KinnisonOnly a minority of our systems can use EFI08:20
juergbipedroalvarez: i'm not sure. i wanted to do a quick bisect but had inconsistent results so i gave up for now08:21
persiaReally?  I thought it worked for A9+, POWER7+, and i686+08:21
* persia can't figure out when MIPS started supporting EFI, but at least some years back in shipping hardware08:22
Kinnisonpersia: It depends on the bootloaders08:22
Kinnisonpersia: e.g. nominally the AARCH64 moonshot should support EFI, but because HP build their u-boot without the requisite support, you can't run the EFI stubs08:23
Kinnisonpersia: as for x86/x86_64 - again it depends on the systems08:23
Kinnisonfor virtualised systems under KVM it depends on OVMF which has licence issues08:24
* persia murmurs about flashing the desired version into the preboot environment flash, and givs up08:24
KinnisonNow if only all of our target systems let us affect the preboot environment08:24
* Kinnison would love it if we could move from extlinux to grub208:25
*** jonathanmaw has joined #baserock08:25
KinnisonBut again, some of our targets lack even the hooks necessary for grub2 to hook in08:25
Kinnisonagain e.g. moonshot which doesn't have the u-boot API built in (or at least didn't, I've not checked the most recent firmware drop)08:25
* pedroalvarez wonders when syslinux is being used in the deployer system08:29
Kinnisonto install the extlinux bootloader08:29
Kinnisononto the disk image08:29
pedroalvarezKinnison: only extlinux.conf, right? or something else?08:30
pedroalvarez(also menu.c32, but that is not relevant)08:30
persiaKinnison: Hrm?  What system doesn't allow one to write to the preboot environment flash?08:30
Kinnisonpedroalvarez: No, the actual boot sector IIRC08:31
Kinnisonpersia: IIRC, the HP moonshot doesn't let you write the u-boot environment08:31
pedroalvarezah, ok, as I expected it couldn't be that easy, thanks08:32
Kinnisonpersia: hence we had to jump through interesting hoops08:32
persiaKinnison: It has no means to flash updated firmware?08:32
Kinnisonpersia: Sure it does, but they're signed lumps applied by the HP iLO08:32
persiaAh, so only HP can produce firmware contents that can be flashed?08:32
Kinnisonit's possible that might not be the case, but I've not investigated08:33
KinnisonCertainly only HP-produced firmware will be in the least bit supported08:33
Kinnisoni.e. they likely will claim user-produced firmware voids hardware warranty08:33
persiaIt's worth asking.  I know lots of other HP products allow you to flash arbitrary firmware, although support is only available with HP-provided firmware.08:35
radiofreeThere were a couple of commands left over in the default environment that looked like you could replace the version of uboot08:36
radiofreeIf you were brave enough to test them08:36
persiaThat seems a sensible model.  Not being brave is well understood :)08:37
persiaIn other hardware (I've not used a Moonshot that way), there are often multiple versions of firmware that can be stored (or one loaded at initialisation and wiped on finalisation), making it much safer to use features like that than one might imagine.08:38
*** ssam2 has joined #baserock08:42
*** ChanServ sets mode: +v ssam208:42
* pedroalvarez sees a 'sleep(2)' call after installing extlinux, as a hack to make the install finish08:47
*** CTtpollard has joined #baserock08:47
Kinnisonx86 bootloaders are sadmaking08:48
pedroalvarezhehe, sleep(10) didn't do anything :)08:51
*** tiagogomes_ has joined #baserock08:56
tiagogomes_it would be nice to be able to `morph build system1 system2 …`08:58
* Kinnison remains desireous of morph build foocluster.morph [optional system list]09:00
persiaKinnison: What do you have against x86 bootloaders?09:00
persiaAlso, +1 to `morph build foocluster.morphP09:01
* persia thinks rEFIt and Clover are both interesting projects for x8609:04
* richard_maw wants to sort out the build graphing code to support graphing more than one system at once09:19
*** gary_perkins has joined #baserock09:27
straycatmaybe we should fix the existing graphing code before thinking about additional functionality?09:27
persiaI generally try to understand the long-term goals before attempting to make the short-term fix, unless there is urgent pain, in which case the activities can be parallelised.09:29
*** ssam2 has quit IRC09:29
richard_mawstraycat: it needs to be fixed to allow graphing multiple systems09:29
*** lachlanmackenzie has joined #baserock09:30
*** ssam2 has joined #baserock09:30
*** ChanServ sets mode: +v ssam209:30
straycatrichard_maw, what about rationalising the source pool creation and artifact resolving stages09:31
straycatthis was the basis for blocking the merge of the fix i made a month ago to detect cycles in the build graph?09:31
richard_mawah yes, the inability of it handling loops09:32
straycatto be fair, i've failed to improve the change in the past month for lack of time09:33
persiaI wonder if it might be worth having a monthly development meeting, where people talk about their priorities for the next month, so that we can have a better idea when various bits will land.09:34
persiaOr if people don7t have time to work on things, this can be reported, and if someone else wants the feature, they can pick it up without as high a burden as poking the last submitter.09:35
pedroalvarezI like that idea09:35
* SotK does too09:36
persiaI've seen it work well in other environments, but it depends on two things.09:37
*** rdale has quit IRC09:37
persia1) someone volunteering to mangage the task list (usually on the wiki) and run the meetings [this can be a rotating role]09:37
*** rdale has joined #baserock09:38
persia2) Everyone having some degree of confidence in what they will be doing over the next month09:38
persiaNote that I don't believe we can cover *all* the activities, because there are always those quick patches one does on a sunday afternoon because one doesn't want to take a fourth bath in the day, but we can probably have some idea what larger things are on the horizon.09:39
radiofreepedroalvarez: did you ever submit upstream?09:39
pedroalvarezradiofree: yes09:40
radiofreepedroalvarez: it didn't seem to land09:41
pedroalvarezradiofree: nope, let me find the bug link09:41
pedroalvarezthey actually closed the bug, but I re-opened it09:44
radiofreeand then they ignored you :(09:45
straycatmore likely they just forgot?09:45
pedroalvarezwell, they answered me 22 days after I reported the bug09:45
straycat / were preoccupied with other things09:46
*** Krin has quit IRC09:47
richard_mawpedroalvarez: so… they brushed you off on a policy reason, rather than fixing the bugs that cause it to fail with -Werror set?09:47
pedroalvareznot sure about how bad sounds "brush you off", but yes09:47
straycatpersia, seems like a good idea but in practice i'm not sure we have that level of contribution reliability09:48
pedroalvarezI guess I should have sent a patch to fix the warnings, instead of adding --disable-werror support09:48
straycatand then when the compiler gets updated it will just break with some new warning?09:49
persiastraycat: You mean that you don't think most of the more active developers know what they are working on that far in advance?09:49
persiaWe could do more often, but that means more time from everyone, which means either finding times that everyone is off work, or getting permission from employers, etc.09:50
pedroalvarezpersia: note that I really like the idea09:50
pedroalvarez1 month sounds ok to me09:50
pedroalvarezradiofree: any suggestions about what can I do with that bug report?09:52
persiapedroalvarez: Are you up for volunteering to run it?09:53
persiaGiven your contributions to the ops team, I'd think someone else might have more time, but there's the "if you want something done, ask a busy person" rule :)09:53
* pedroalvarez is participating in around 5 projects everyday09:55
pedroalvarezbut I can allocate some time for this09:56
straycatI didn't know about until April 5th when i spotted the bug09:56
pedroalvarezpersia: I'll send an email to baserock-dev to discuss the format, content and process of these meetings, is that ok?09:57
persiaSounds reasonable to me.09:57
* straycat sighs and goes back to whatever he was doing09:58
persiaI've gotten involved in very complex processes around such things, but I'd strongly suggest starting with something like a list of tasks, loose assignment (by volunteering), and quick status by task in the meeting.  I'd further suggest having the tasklist on the wiki as a set of links into storyboard.09:59
straycatyou're not listening are you09:59
persiaMost importantly, be sure that the set of things is not intended to be all-inclusive.  There's always going to be lots of stuff that can't make the list, for various reasons.10:00
persiastraycat: Right.  I think there's *loads* of stuff that we'll end up doing that won't make a meeting.10:01
persiaBut if there are things people want to do longer-term, I think it makes sense to share those at a time when we're all paying attention.10:01
persiaThat said, if folk think it's better to do that by sending RFCs to the list, we can do it that way10:01
* persia has a tendency to do too much on IRC10:02
straycatA meeting would help as a forum to bring things up maybe, but I don't see what the difference would be between me raising something in a meeting and raising something on gerrit? Persistence perhaps. For example we know we need to tidy up some code, the only reason it's not been done is because no one's had the time for it, really, not because there's an absence of planning10:13
Kinnisonstraycat: submitting things on gerrit quickly gets them lost in a crowd I think10:14
Kinnisonstraycat: similarly to patches sent to a busy ML10:14
* straycat nods10:16
KinnisonA meeting, if well attended by people who attribute it with some level of authority and thus vest a level of dedication to it, might be better10:17
jjardonpersia: you can build with support for bios and/or efi when you build a (recent) syslinux10:21
straycatKinnison, if they have the time to dedicate then sure10:27
Kinnisonstraycat: indeed10:28
Kinnisonstraycat: But those without the time are unlikely to attend such a meeting, surely?10:28
*** Krin has joined #baserock10:28
persiaI've also seen folk whose time is dedicated have time for occasional meetings, which can be useful, so that the agenda of the folk behind the scenes (who rarely emerge in open source communities) can be glimpsed, at least in part.10:29
straycatKinnison, i guess :)10:32
jjardonI also agree the meeting is a good idea. I've created a skeleton page here to add topics the people are interested on: Maybe is a good idea to create a doodle poll to set the date?10:36
Kinnisonjjardon: I think that's rather jumping the gun10:41
Kinnisonjjardon: pedroalvarez was going to send a mail to the ML to discuss things10:41
straycatperhaps we should have a meeting to organise our meeting10:41
* pedroalvarez will write the email as soon as I finish something else10:42
jjardonKinnison: jumping the gun?10:46
persiaTo date we've taken decisions on the mailing list, and more recently in gerrit.  Taking a decision in IRC to take more decisions in IRC without consulting the traditional forums is likely to leave many existing members of our community feeling disenfranchised.10:49
jjardonpersia: dont think anyone is out: the meeting / doodle was meant to be announce in the mailing list10:52
persiaI'm just a bit more conservative than that: I'd like to see loose consensus that there *should* be an IRC meeting before announcing a doodle.10:52
persiaBut I don't assert that my conservatism should inhibit others.10:53
mwilliams_ctAdd a "do not want" option to the doodle?10:55
*** zoli__ has quit IRC11:22
*** zoli__ has joined #baserock11:23
rdaleis an infinite loop in a known bug11:36
ssam2yes, it happens if there is a loop in the build graph11:40
ssam2 would fix it I think11:40
rdaleall i've done is to reference the openssh chunk from a stratum, as i don't want to pull all of foundation in, just for one chunk11:41
pedroalvarezrdale: not possible to do that atm11:46
rdalei've renamed as 'openssh-wrt' and that's fixed it11:47
pedroalvarezoh, so you created a stratum for openssh?11:48
pedroalvarezor you added openssh into that stratum?11:48
rdalei added openssh to an existing cpe-wrt stratum11:49
ZaraSotK: thanks for the fix. I've +1'd.11:51
pedroalvarezah, then if that gives you loop error might be because openssh is already a dependency?11:51
rdaleah yes, i've found the problem - cpe-wrt depends on connectivity which depends on foundation11:52
*** rdale has quit IRC12:08
*** rdale has joined #baserock12:10
radiofreefound the cause of the vt switching issue...12:26
radiofreeusing vga=...12:26
*** gary_perkins has quit IRC12:29
*** gary_perkins has joined #baserock12:31
*** rdale has quit IRC12:33
*** rdale has joined #baserock12:38
ssam2radiofree: so remove that and it works?13:02
SotKZara: thanks13:04
radiofreessam2: no framebuffer device then13:04
radiofreeusing uvesafb instead fixed nothing13:05
ssam2oh, so you've found the cause but not the solution?13:07
*** pedroalvarez has quit IRC13:10
*** pedroalvarez has joined #baserock13:11
*** ChanServ sets mode: +v pedroalvarez13:11
*** pedroalvarez has quit IRC13:11
radiofreessam2: yeah13:12
*** pedroalvarez has joined #baserock13:12
*** ChanServ sets mode: +v pedroalvarez13:12
radiofreethough that only works with uvesafb not vesafb13:13
radiofreei think.. maybe...13:13
radiofreedidn't work this time13:14
radiofreeright so if you switch to the vt before you initilise the uvesafb driver (and i assume vesa here as well) it will work once you load the driver13:16
Kinnisonjjardon: do you care about ? It hasn't been touched in 5 or so weeks13:27
pedroalvarezlooks like tlsa updated that patch after jjardon13:31
tlsai doubt it; that was for manifest generation, whihc is now done differently13:32
pedroalvarezthen, can we abandon it?13:33
tlsaI suspect so, unless anyone wants more info in the metadata13:33
*** sambishop has quit IRC13:37
*** paulw has quit IRC13:40
*** paulw has joined #baserock13:40
*** sambishop has joined #baserock13:52
radiofreeand now it's working completly fine....13:55
pedroalvarezwell, this is baserock, redeploy and it should be broken again13:56
persiaWeren't there some patches recently that let one test reproducibility to avoid that?14:08
* radiofree has no idea wtf he did to get this working14:13
radiofreei didn't edit any config files.. build new software...14:13
rjekDid you sacrifice a lower mammal?14:19
rjekAny pentagrams?14:19
rjekwin 2214:19
* radiofree has successfully broken it again by redeploying14:32
rjekNow you get to say "bah"14:32
radiofreenow this is working14:35
radiofreeand now it is not14:37
pedroalvarez- Son, take a look at the 'turn signals' and tell me if they work14:40
pedroalvarez- they work!14:40
pedroalvarez- now they don't14:40
pedroalvarez-they work!14:40
pedroalvarez.. etc14:40
pedroalvarez(spanish joke)14:40
* richard_maw prefers the Father Ted version with the Fairy lights on the Christmas tree14:41
*** sambishop has quit IRC14:45
*** sambishop has joined #baserock15:14
*** zoli___ has joined #baserock15:17
*** zoli__ has quit IRC15:17
ssam2Patch to update Morph in definitions for the 15.19 release:
*** a1exhughe5 has quit IRC16:02
Zaraif I've made commits before adding gerrit as a remote and getting the commit-msg hook, is there any way to push to gerrit? or will I need to go back and make the commits again?16:02
rdalehas anyone time to review this patch before the release and corresponding 588 patch, as minimal systems won't work currently without it16:03
persiaI'd recommend doing a rebase, which will recreate all the commits.16:03
*** jonathanmaw has quit IRC16:09
jjardonKinnison: Iv'e abandoned it know16:11
ssam2Zara: git-review fixes it for you automatically16:11
ssam2rdale: thanks, i'll make sure it gets in16:12
rdalethanks a lot16:13
rdaleis there an easy way to amend a commit message for a patch on gerrit?16:20
rdaledo i amend it locally and push again?16:20
ssam2you can edit it in the web UI16:20
ssam2'Edit Message' button just above the commit message and to the right16:20
rdaleoh yes - that's neat!16:21
ssam2on the subject of tagging Morph whenever it understands a new definitions version: I'll create tags named baserock-definitions-1, baserock-definitions-2 and baserock-definitions-3  unless anyone has a strong opinion on how else it should be done16:30
ssam2I'm not going to tag versions where support was dropped at this point because I don't have a clear idea how to do it yet, others are welcome to do it16:31
jjardonssam2: +116:32
persiaDo we have any versions where support was dropped yet?16:40
ssam2actually no16:40
*** tiagogomes_ has quit IRC16:41
persiaThen I don't mind that nothing is tagged that way :)16:41
persia+1 on the tag changes16:41
*** zoli___ has quit IRC16:45
*** mariaderidder has quit IRC16:49
ssam2tags complete! I'll also try to add a bit to the release process about how to do this in future16:55
*** flatmush has quit IRC16:59
ssam2argh! Seems Gerrit has resumed its game of losing merged commits17:00
ssam2 is marked as merged but is not in 'master' of morph.git on or git.baserock.org17:00
*** flatmush has joined #baserock17:00
SotKis there any chance could be merged in time for the release?17:01
*** bashrc has quit IRC17:01
ssam2ha, good idea17:02
*** Krin has quit IRC17:03
SotKssam2: thanks :)17:03
ssam2oh, part of my confusion is that 588 is for definitions.git, not morph.git17:08
ssam2oh! it's for fhs-dirs.git17:10
ssam2so actually nothing is wrong, but we need to update the ref for fhs-dirs in definitions.git17:11
ssam2i'll do that, update morph.git, set builds running, then go home!17:11
paulsherwoodsounds like a plan :)17:12
* SotK sends and heads off too17:15
*** ssam2 has quit IRC17:30
*** zoli__ has joined #baserock17:33
*** gary_perkins has quit IRC17:40
*** edcragg has quit IRC18:17
*** lachlanmackenzie has quit IRC18:20
*** zoli__ has quit IRC18:49
*** zoli__ has joined #baserock19:14
*** zoli__ has quit IRC19:23
*** zoli__ has joined #baserock19:29
*** zoli__ has quit IRC21:04
*** Albert_ has quit IRC22:23

Generated by 2.15.3 by Marius Gedminas - find it at!