IRC logs for #baserock for Monday, 2014-08-25

*** franred [] has joined #baserock07:30
*** franred [] has quit [Quit: Leaving]07:48
*** franred [] has joined #baserock07:59
paulsherwoodrjek: morning :)09:26
persiapaulsherwood: Looking at your jetson upgrade: what precisely does "no_console_suspect=1" mean?  Without background, that looks like it suspends the machine until/unless it has console, but I may be misreading it.09:36
paulsherwoodpersia: as i've just commented on the ml, i just took this directly from the wiki instructions09:50
paulsherwoodi have no idea what's actually going on :)09:50
paulsherwoodnote this is my weak, belated attempt to save future jetsons from the lobotomy i gave to pris :)09:51
paulsherwoodby having a jetson-upgrade explicitly in definitions, folks may be less tempted to run morph --update with the default devel-update, as i did.09:53
persiaHeh.  Oh well.  If you do it without HOSTNAME, remember to specify on the command line (as with upgrade-devel).09:53
persiaIndeed.  I think the goal is good: I just worry as to whether that particular argument means it doesn't work without serial console and/or attached HDMI+usb.09:53
* persia will have to read docs09:53
paulsherwoodi'll try. i'm steadfastly refusing to attach a serial console until i've killed another jetson09:54
persiaGood choice.09:55
persiaTurns out that no_console_suspend does almost the opposite of what I would expect from the name: it *enables* continuous console output during the suspend process, rather than suspending it for delivery at resume (the default).09:56
paulsherwoodmy flashing process is depressingly slow, still, though. my usb passthrough speed varies between bad and appalling. looks like it may take well over an hour at this rate09:57
persiaYou either need to install another operating system on your laptop or morph should support it (unfortunately, it doesn't have the syscalls upon which morph depends).09:58
paulsherwoodand give up my beloved macos? surely you are joking :)09:59
persiaAgain, "or"10:00
persiaHmm.  Seems jail(8) aren't available in OS X, and chroot(8) isn't strong enough, although I do wonder why morph uses linux-user-chroot and still requires root (as the point of linux-user-chroot was to allow for non-root).10:05
persiaErr, rather, without the substitution10:07
* persia decides to postpone understanding man sections yet again10:07
persiasandbox(7) looks interesting, but I can't find much documentation about it (although I did find some complains about lack of documentation :) )10:21
persiaIt's the protection used to separate iOS apps, so at least some believe it to be safe.10:22
persiaAlso probably requires using ZFS, as HFS+ can't support building some systems, and btrfs has no support.10:24
paulsherwoodnever mind. i've rebooted10:24
persiaWithout completing the flash?10:24
persiaYou might be able to get away with a live environment: if you have an OS that can run over EFI from USB, just hold down the option key on boot.10:25
paulsherwoodcan baserock do that?10:25
persiaI don't know: I haven't been following the boot-from-USB work.10:27
* paulsherwood remembers richard_maw had something running a while ago10:27
persiaBut Apple EFI has some hacks to more easily boot legacy USB sticks, which may mean that it works better on Macs than other places.10:27
paulsherwoodafter reboot, my usb is running at least 10 times faster than before10:27
persiaSo 6 minutes then?10:28
paulsherwoodno, i reckon it'll be 30 mins now. it was on for taking most of today, before10:28
persiaOh my.10:28
* paulsherwood 'loves' software10:28
* paulsherwood wonders why franred has responded to richard_maw's patches the workflow tweaks thread now10:45
paulsherwoodthe more time goes by, the more i think persia may be right to insist on killing morph edit entirely10:46
franredpaulsherwood, they were on review and they fix some behavior that we want to fix. Remove morph edit would be a future task10:49
paulsherwoodfranred: have you tried the change? i did, at the time, but then i got distracted10:53
paulsherwoodthe actual change, not the yarns10:53
franredpaulsherwood, not yet, I've just read the thread and reviewed the code which looks sensible10:55
paulsherwoodi wonder if it still works. i'll try it :)10:55
paulsherwoodall tests pass, so that's a start :)11:08
paulsherwoodfranred: do you have supercowpowers to revert a commit in morph?11:23
franredpaulsherwood, I think so11:23
paulsherwoodi can't see any record of a review of 23e86c643a82731aba6561321b8bb5b168e80598 on the list. and afaict it breaks morph edit for me11:25
paulsherwoodi'm not suggesting we revert unilaterally, but if others can confirm it causes breakage, i suggest it gets taken out11:26
paulsherwoodso i tried the same sequence three times. first time with the always-use stuff. broke. then with head. still broken. then with 35f3a8779a316ed6a44ce5bb62bea9e2d7c67a16. works11:29
persiapaulsherwood: From your log, I don't see how always-use-current differens from not-always-use-current11:34
paulsherwoodi could have been clearer with my naming. the only change in each case is i used a different version of morph.11:36
paulsherwoodHEAD is broken HEAD~1 is not, for me11:36
persiaI believe you: I just don't see anything in your paste indicating switching beween versions of morph11:37
paulsherwoodi did it in a different window11:38
persiaMaybe next time, useful to run `morph --version` midstream :)11:38
paulsherwoodyes indeed11:39
persiaSo, anyway, rather than fixing this, I think we should just remove morph edit: this would mean you couldn't replicate the bug, and everyone would be happy, right?11:39
paulsherwoodi hope you are being ironic?11:40
persiaHard to say.  I've yet to find anyone willing to tell me good things about morph edit, but I've also failed to find anyone who is willing to remove it, so whether I'm being ironic or hopeful is a difficult question.11:43
paulsherwoodi'm ok with the general idea of removing morph edit, but i don't think this bug should be the trigger11:43
paulsherwoodremoving morph edit means fixing quite a lot of documentation. fixing this bug involves reverting one commit11:44
persiaI can accept that.11:45
persiaOn the other hand, the implications of reverting are worrying: without this patch, we can't expect the content of a ref to be consistent, which rather breaks the definitions model.11:46
petefoth /me volunteers for 'fixing quite a lot of documentation' when required11:46
persiaThis is not a documentation issue, sadly.  It's a repeatability issue.11:46
persiaSince it has become possible to change the content available from a ref, without this class of hackery, a ref is insufficient to be able to have confidence in stable source.11:47
paulsherwoodi think it was always possible? is this a recent thing?11:51
persiaRelatively, yes.11:52
persiaErr, Hmm..  Rather, it has become more prominent lately: it seems that `git replace` has been in the codebase for quite some time.11:55
paulsherwoodin our codebase? or git's?11:55
persiaSo in fact, every version of morph prior to HEAD was unable to reproduce builds if the user happened to have run git replace locally.11:56
paulsherwoodiiuc from richard_maw's explanation there would be evidence of replace in any actual repo affected by this?11:57
persiaYes.  We should be able to detect that this was happening, and so detect if it was happening *differently* than it was happening before.12:00
persiaWhat I'm less sure about is if we can cause a git-replaced environment to be git-replaced in the manner it was the last time we did something, rather than in the manner it is replaced currently.12:00
franredpaulsherwood, I've reproduced the error, I will send the error log to the mailing list but I wouldn't like to revert the patch if it can wait until tomorrow12:02
paulsherwoodpersia: no, what i meant was, for an affected repo, wouldn't there be evidence in refs/replace/ ?12:04
paulsherwoodif so we can visit all repos and see if there are any? :)12:04
persiaYes, but I'm not sure that if we find one, we can know at what point in time any given entry was added.12:05
persiaWhich makes me agree with richard_maw's initial decision to just make it not work until more thought is applied.12:05
paulsherwoodpersia: remember that algorithm script i mentioned last week? this would be a good usecase for that - check for replace on all upstreams12:06
persiaBut I'm not sure how knowing that helps anything.12:06
paulsherwoodi'm ok with richard_maw's initial decision. i'm not ok with him breaking the default current workflow12:06
paulsherwoodwell, if there are none, it's a moot point :)12:07
persiaNo, because lorry could add one in an hour or two.12:07
paulsherwoodat least for gbo repos to-date.12:07
paulsherwoodfair enough - so has lorry been fixed to spot this?12:08
persiaHow do you mean "fixed"?12:08
* persia wouldn't expect any lorry behaviour changes around this issue12:08
paulsherwoodok, i may be off track here. if upstream does git replace, how do we ensure repeatability?12:10
persiaThat is the problem that remains to be solved.12:10
persiaPerhaps there is a means to track when a given replacement entry was added.12:10
persiaPerhaps we need to cache the state of refs/replace at the time of build.12:11
persiaPerhaps there's another solution that doesn't come to mind immediately for me, but would be obvious to someone with a deeper understanding of git internals.12:11
paulsherwoodso, what's the benefit of the patch, if the above remains to be solved?12:11
persiaNote that this may also affect definitions notation, because one may want to build the same SHA with different replace contents, and may need a way to indicate that.12:12
paulsherwoodi'm pretty sure normal users of morph would not be doing the fancy git trickery we're afraid of12:12
persiaThis patch ensures that no replace is ever followed: morph just ignores it entirely.12:13
persiaUsers need not: if someone pushes a replace upstream, then everyone gets it, even if they don't understand.12:13
paulsherwoodok. and now, suddenly, morph starts ignoring any existing replaces? amazing.12:14
radiofreepersia: it will work without serial console and/or attached HDMI+usb12:14
radiofreehowever there's no default password, so you won't be able to login12:14
radiofreeso you'll need.... a serial console and/or attached HDMI+usb keyboard to set the password12:14
paulsherwoodaha, understood, thanks radiofree12:15
persiaradiofree: Can't I just inject an ssh key?12:15
radiofreesure, but you'd have to do that at the deployment stage right?12:15
persiaAh, good point.12:15
* persia should probably just try this, rather than speculating12:16
paulsherwoodback to the replace discussion - if there are any official upstream replaces, in what morph has been building, it will now build something different?12:17
persiapaulsherwood: Hrm.  Interesting point: we go from having no idea whether we can reproduce, to being able to reproduce from now, but no idea whether we can reproduce the current cache objects (as they may have used replace).12:17
paulsherwoodthis was not thought through, and not discussed. it may be a real problem, but this patch is not an actual solution12:18
persiadefinitions is still at 35f3a8779a316ed6a44ce5bb62bea9e2d7c67a16, so only a few folk have ever used this new broken morph.12:19
* paulsherwood would prefer if a real git wizard could comment, since our interpretation may be wrong12:19
paulsherwoodtrue, persia.12:19
persiaHaving gone through it, despite my initial discomfort, I think I agree that unless discovery was performed to ensure there were no replace users for at least the refs in current definitions (but ideally every repo referenced by every trove in use), I agree that perhaps this is dangerous (even aside from the minor issue of it breaking a tool for which I don't see the point).12:21
persiaIf such discovery *was* performed, then I'd rather not revert the patch, as it ensures reproducibility for at least g.b.o history.12:22
persia(although it ought get adjusted to not break things)12:22
* paulsherwood wonders if franred fancies helping him create a discovery script12:23
paulsherwoodin other news, the actual branch i was testing does not seem to behave as i expected it would. i'm thinking it would probably be better to spend time agreeing what our desired workflow should be, and get to that asap12:23
* paulsherwood wonders what to do, to ensure the replae discussion doesn't get lost12:34
*** genii [~quassel@ubuntu/member/genii] has joined #baserock14:22
*** richard_1aw [] has joined #baserock14:30
*** richard_maw [] has quit [Ping timeout: 260 seconds]14:32
* paulsherwood fiddles with putting a trove into a public OpenStack, while building a new jetson image on one jetson and braining another....15:40
persiaSo, how well does morph cope with doing multiple things at once?15:41
paulsherwoodi wouldn't try that. the flashing (non-morph) and openstack (morph) stuff on one br machine - the jetson build is directly on the jetson15:42
paulsherwoods/wouldn't try/haven't tried/15:42
paulsherwoodi notice that recent changes to morph (needing to specify relative paths to morph files) have invalidated some of the wiki documentation15:45
persiaOutdated, rather than invalidated, I'd say.15:46
persiapetefoth: Did you mention earlier you wanted to fix docs?  There's an opportunity there.15:46
paulsherwoodi've fixed the ones i've hit15:48
* franred is still trying to fix error in gerrit, nss is not installed in the system and to compile, it we needs libc6-dev...16:19
franreds/error in gerrit/the java configuration in gerrit to have access/16:20
persiaI'd expect NSS to build out of morphologies: do you have a log?16:27
franred - this is not an error but when I try to add an user I've gotten - I've try to solve as [1] (no luck) [2] which tells me that so I've tried to compile nss from debian packages16:44
franred getting
franredand looking for gnu/stubs-32.h on google, I found that I need libc6-dev16:45
juergbifranred: in case that's not clear, libc6-dev are just the dev files of (e)glibc16:57
juergbifranred: if you're building nss on x86-64, you need16:58
juergbiexport USE_64=116:58
juergbiotherwise it tries to build a 32-bit version which won't work with a pure 64-bit glibc16:58
juergbi(yes, the nss build system is not particularly smart)16:59
franredjuergbi, cheers :) - I got nss compiled (yep, it is not very smart)17:12
*** franred [] has quit [Quit: Leaving]17:52
*** franred [] has joined #baserock18:10
*** franred_ [] has joined #baserock18:10
*** franred_ [] has quit [Client Quit]18:10
*** flatmush [] has quit [Ping timeout: 240 seconds]18:43
*** franred [] has quit [Ping timeout: 260 seconds]18:44
*** flatmush [] has joined #baserock18:45
*** franred [] has joined #baserock18:46
*** franred [] has quit [Ping timeout: 246 seconds]18:50
*** franred [] has joined #baserock18:50
*** flatmush [] has quit [Ping timeout: 250 seconds]18:51
*** franred [] has quit [Ping timeout: 245 seconds]18:56
*** franred [] has joined #baserock18:59
*** flatmush [] has joined #baserock19:00
*** franred [] has quit [Ping timeout: 240 seconds]19:04
*** flatmush [] has quit [Ping timeout: 264 seconds]19:06
*** flatmush [] has joined #baserock19:12
*** franred [] has joined #baserock19:16
*** flatmush [] has quit [Ping timeout: 260 seconds]19:19
*** franred [] has quit [Ping timeout: 264 seconds]19:33
*** flatmush [] has joined #baserock19:44
*** flatmush [] has quit [Ping timeout: 260 seconds]19:56
*** flatmush [] has joined #baserock20:12
*** flatmush [] has quit [Ping timeout: 240 seconds]20:16
*** flatmush [] has joined #baserock20:17
*** flatmush [] has quit [Ping timeout: 240 seconds]20:22
*** flatmush [] has joined #baserock20:23
*** flatmush [] has quit [Ping timeout: 245 seconds]20:27
*** flatmush [] has joined #baserock20:28
*** flatmush [] has quit [Ping timeout: 245 seconds]20:32
*** flatmush [] has joined #baserock20:37
*** flatmush [] has quit [Ping timeout: 250 seconds]20:42
*** flatmush [] has joined #baserock20:51
*** flatmush [] has quit [Ping timeout: 255 seconds]20:56
*** flatmush [] has joined #baserock21:07
*** flatmush [] has quit [Ping timeout: 240 seconds]21:11
*** flatmush [] has joined #baserock21:16
*** flatmush [] has quit [Ping timeout: 246 seconds]21:32
*** flatmush [] has joined #baserock21:34
*** flatmush [] has quit [Ping timeout: 240 seconds]21:54
*** flatmush [] has joined #baserock22:00
*** flatmush [] has quit [Ping timeout: 255 seconds]22:09
*** flatmush [] has joined #baserock22:11
*** flatmush [] has quit [Ping timeout: 255 seconds]22:17
*** flatmush [] has joined #baserock22:23
*** flatmush [] has quit [Ping timeout: 260 seconds]22:30
*** flatmush [] has joined #baserock22:33
*** flatmush [] has quit [Ping timeout: 240 seconds]22:51
*** flatmush [] has joined #baserock22:57
*** genii [~quassel@ubuntu/member/genii] has quit [Remote host closed the connection]23:04
*** flatmush [] has quit [Ping timeout: 245 seconds]23:12
*** flatmush [] has joined #baserock23:25
*** flatmush [] has quit [Ping timeout: 250 seconds]23:37
*** flatmush [] has joined #baserock23:38
*** flatmush [] has quit [Ping timeout: 240 seconds]23:53
*** flatmush [] has joined #baserock23:54

Generated by 2.15.3 by Marius Gedminas - find it at!