IRC logs for #baserock for Tuesday, 2016-09-13

*** rdale has quit IRC01:46
*** gtristan has joined #baserock02:10
*** rdale has joined #baserock03:19
*** rdale has quit IRC04:23
*** gtristan has quit IRC05:03
*** gtristan has joined #baserock05:27
*** gtristan has quit IRC05:34
*** gtristan has joined #baserock06:32
*** rdale has joined #baserock06:57
*** CTtpollard has joined #baserock07:23
paulsherwoodradiofree: the kbas is working for uploads/downloads... there's something up with the status page though07:37
*** toscalix has joined #baserock07:57
*** locallycompact has joined #baserock08:06
*** CTtpollard has quit IRC08:25
*** CTtpollard has joined #baserock08:27
*** locallycompact has quit IRC08:43
*** tiagogomes has joined #baserock09:04
*** CTtpollard has quit IRC09:04
*** CTtpollard has joined #baserock09:04
*** gtristan has quit IRC09:33
*** locallycompact has joined #baserock09:37
*** gtristan has joined #baserock09:53
*** jonathanmaw has joined #baserock10:05
jonathanmawWhen building armv7 from scratch, I get the error "No rule to make target `mpn/add_n.lo', needed by `libgmp.la'" when building stage1-gcc.10:10
jonathanmawI'm building from a debian system, so it's possible I'm missing dependencies in the build environment, but I'm not sure what.10:11
richard_mawnever seen that before10:11
*** rdale has quit IRC10:14
pedroalvarezjonathanmaw: silly question: current definitions or are you upgrading something (gcc?)10:23
jonathanmawpedroalvarez: I oughtn't to have changed gcc10:25
jonathanmawthough the ref is different to the one in definitions master10:26
jonathanmawstage1-gcc using the baserock/build-essential version10:27
jonathanmawI think it's just old, though.10:27
pedroalvarezyes, just old10:30
pedroalvarezi thought that the dependencies were all installed using the install_dependencies.sh script10:31
pedroalvarezbut I guess something is missing10:31
*** jonathanmaw has quit IRC10:34
*** jonathanmaw has joined #baserock10:34
*** toscalix has quit IRC10:38
*** toscalix has joined #baserock10:45
*** edcragg has quit IRC11:07
*** edcragg has joined #baserock11:09
*** rdale has joined #baserock11:35
radiofreewhy don't changes to system-integration commands in a chunk result in a rebuilt of the chunk?11:58
pedroalvarezbecause these commands are executed when assembling the System artifact, and nothing has to be done when building the chunk (ybd)12:04
pedroalvarezMorph's implementation of system-integration commands will trigger chunk rebuilds12:04
* richard_maw can't remember if he fixed that or just thought of fixing it12:05
pedroalvarezrichard_maw: hm.. fixing morph's implementation?12:07
pedroalvarezI don't know what else can be fixed :/12:07
radiofreewell i have a chunk here where i have to add something like "echo hello" to the build-commands to ever run the change the system-integration12:12
radiofreeso it's a ybd issue?12:12
radiofreesystem-integration commands seem to be run when assembling the system artifact12:13
radiofreenot at deploy (thought i can try that if you'd like)12:13
radiofrees/thought/though12:13
persiaWhen assembling the system artifact is the right time to run the system-integration commands, so changes to those should cause reassembly, but not rebuild.12:14
pedroalvarezradiofree: I would call it feature the fact that ybd doesn't have to rebuild the chunk, just the system.12:14
radiofreepersia: they appear to cause neither12:14
pedroalvarezif ybd is not rebuilding the system, then yes, it's a bug12:14
persiaradiofree: Then that's a bug :)12:14
radiofreethe system artifact cache key is made up from all the chunks/strata in a system?12:15
persiaMy memory is that morph has a different bug, in that it rebuilds both, but I may be mistaken, or this may be an artifact of an old version of morph that encouraged putting the system integration commands in the chunk git repo, so the tree hash of that would change.12:15
radiofreeif the chunk cache key doesn't change when you change a system-integration command there, how would ybd know to re-run the command?12:16
persiaThe system artifact cache key should consist of all the components of the system + all the code that would be executed when assembling the system.12:16
* radiofree doesn't really like system-integration commands in chunks12:16
persiaWhere would you put them?12:17
radiofreethey should be more like "essential steps" (e.g gtk3) not something i might not want to do (e.g enable opensshd in the sshd chunk)12:17
persiaAnd, if "system", how would you reuse things that are chunk-related.12:17
radiofreewell if on my system i want to enable sshd by default i'd do that in something in the system morph12:17
persiaWhy would you want to ship a disabled sshd?12:18
* persia would expect it to be more secure to just not ship it12:18
radiofreebecause that's what the client wants12:18
radiofreeif they need ssh the connect via a serial console and enable it12:18
persiaOK.  Why would the client want to ship a disabled sshd?12:18
radiofreebut, ok, poor example12:19
radiofreeactually system-integration commands are used rather sparingly in definitions12:20
persiaYep.  I believe they are only used when there is no other choice.12:20
radiofreeand they seem to be only for things that are essential for the chunk to work12:21
persiaOne of the interesting things about system-integration commands is that the results are not available in the build environments, so they end up being used much less than maintainer scripts in package-based systems.12:21
radiofreeso shouldn't that be part of the chunks cache key?12:21
persiaWell, technically, you don't need to rebuild the chunk if you change them, only the system.12:22
radiofreedoes morph still work?12:22
* radiofree might try the test with morph12:22
persiaI expect someone optimised them out of the cache key to reduce cache misses, but may not have remembered to add the aggregation of all the system-integration commands for all the components in the system cache key.12:22
persiaI believe morph considers system-integration commands to be part of the cache key (by considering the entire chunk morphology to be part of the cache key), so it wouldn't expose this, but would rebuild things that didn't necessarily need it.12:23
radiofreewhat's the cache server for morph, artifacts.baserock.org?12:23
SotKit was cache.baserock.org once upon a time12:24
radiofreeERROR: Definitions format version 6 is not supported12:31
radiofreei guess i need to git clone morph?12:31
SotKprobably12:33
SotKwhat does `morph --version` give you?12:33
radiofreeERROR: Definitions format version 6 is not supported after cloning from gbo12:33
radiofreea3d3e81914b4bcb157ce862e15196294ac943b5212:33
locallycompactversion 6 is pretty old no?12:34
* radiofree pulls definitions12:34
locallycompactI thought submodules was version 812:34
locallycompactthat was the last12:34
radiofreeyes, november 23 201512:34
radiofreenot sure why i have such an old definitions12:34
radiofreeah12:34
radiofreechecked out old tag12:34
* radiofree likes the progress indicator for downloading cache artifacts12:35
*** gtristan has quit IRC13:00
pedroalvarezmorph does still work13:03
*** edcragg has quit IRC13:33
*** edcragg has joined #baserock13:35
*** gtristan has joined #baserock13:38
* radiofree comes back to the terminal with morph running13:47
radiofreeit froze on stage2-gcc-misc[##################            ] 303.8/499.0 MB13:47
radiofreethough it seems to resume quite nicely13:47
paulsherwoodradiofree: what's precisely broken in ybd here?13:48
* paulsherwood has missed most of this, as a result of being dragged into non-software things13:49
paulsherwoodsystem-integration-commands are part of ybd cache-key afaik13:49
paulsherwood(in each chunk, hence in systems that contain chunks that have s-i commands)13:50
persiapaulsherwood: radiofree was reporting a lack of rebuild on changing system-integration commands, which pedroalvarez and I thought was a feature (excepting the need for system-cache-key integration).14:00
pedroalvarezbut radiofree reported that the system wasn't being rebuild either, which if that is the case, I thought that was a bug14:01
* paulsherwood confirms that system-integration-commands have somehow slipped through :/14:02
paulsherwoodthis week is really starting to annoy me :/14:02
pedroalvarez:S14:04
* paulsherwood wonders what that emoticon means14:06
radiofreepaulsherwood: i haven't managed to test either yet14:07
radiofreethe ybd cache server didn't work for me14:07
persiapaulsherwood: In my opinion, system-integration-commands belong in the cache key for the system, rather than for the chunk, if that can be done sensibly.14:08
*** edcragg has quit IRC14:32
*** brlogger has joined #baserock15:11
*** faybrocklebank has quit IRC15:48
radiofreewhat does OSError: [Errno 18] Invalid cross-device link mean?16:13
radiofreedoes morph use sandbox lib?16:14
richard_mawradiofree: it means something's using os.link() across two different filesystems. The only place I can think of that being used directly is staging area creation, which needs to have the staging areas on the same filesystem as the hardlink cache or it defeats the purpose of having the cache16:18
radiofreehmm16:19
radiofreei'm doing this in a chroot, so i guess they are on the same filesystem?16:19
richard_mawit depends on how your chroot is constructed, but being chrooted doesn't inherently mean you're on a different filesystem16:20
richard_mawradiofree: what does running `findmnt` from outside the chroot say?16:24
locallycompactxpost from #rust on mozilla: <dtolnay> locallycompact: also I really like your use case so I filed https://github.com/serde-rs/serde-rs.github.io/issues/22 to add it to the website16:25
radiofreerichard_maw: ooh that's a lovely command16:25
radiofreehttps://paste.codethink.co.uk/?82116:25
richard_mawradiofree: sounds like it's picking a staging area directory that's not in your artifacts mount then16:31
radiofreehmm16:31
richard_mawthat looks like artifacts, gits and your root are on the same deviceā€¦ are you using btrfs subvolumes for br64 and brsys?16:32
radiofreei'm chrooting into an extracted baserock chroot image16:33
radiofreethe 32-bit chroot seems to work16:33
radiofreehold on i'll scratch it and try again16:33
* radiofree assumes he did something wrong16:34
richard_mawwhat *are* brsys and *br64*?16:34
radiofreebrsys is a 32-bit baserock chroot16:35
radiofreeoh, is it because i'm mount binding sys, proc/ in both?16:35
richard_mawsys dev and proc should not matter16:35
radiofreebr64 is a 64-bit baserock chroot16:35
richard_mawand you've bind-mounted the 32-bit chroot's gits and artifacts into the 64-bit one's16:36
radiofreei don't think so16:36
radiofreeyes16:36
radiofreeyes i have16:36
richard_mawthat's what your findmnt output is telling me whether you think it or not16:36
paulsherwoodradiofree: can you be more precise about 15:07 < radiofree> the ybd cache server didn't work for me16:36
radiofreei think i did that to share gits between the two chroots without having to copy16:36
radiofreei clearly don't need the artifacts from a 32-bit chroot....16:37
* richard_maw shrugs16:37
paulsherwoodif you were trying to build 32-bit artifacts, there are none cached16:37
radiofreethat fixed it, thanks richard_maw16:37
radiofreepaulsherwood: it didn't appear to work with 64-bit either16:38
radiofreegive me a second and i'll try it again16:38
richard_mawthere are cases when you'd want them  (e.g. deploying a different architecture or on systems where you are capable of building multiple target architectures from the same host architecture)16:38
* paulsherwood has fixed the status error, and confirmed that it appears to work16:38
radiofreepaulsherwood: is that with master of definitions?16:39
paulsherwoodyup16:39
radiofreepaulsherwood: looks like it's working, thanks16:47
paulsherwoodgreat!16:48
paulsherwoodnow i'll look at the s-i commands16:49
*** jonathanmaw has quit IRC16:51
*** toscalix has quit IRC17:15
*** tiagogomes has quit IRC17:42
*** locallycompact has quit IRC18:19
*** paulwaters_ has joined #baserock18:29
*** gtristan has quit IRC18:59
*** rdale has quit IRC19:16
*** CTtpollard has quit IRC19:16
*** persia has quit IRC19:16
*** inara` has quit IRC19:16
*** mwilliams_ct has quit IRC19:16
*** trn has quit IRC19:16
*** dabukalam_ has quit IRC19:16
*** leeming has quit IRC19:16
*** cyndis has quit IRC19:16
*** perryl has quit IRC19:16
*** SotK has quit IRC19:16
*** pedroalvarez has quit IRC19:16
*** ChanServ has quit IRC19:16
*** radiofree has quit IRC19:16
*** anahuelamo_ has quit IRC19:17
*** rjek has quit IRC19:17
*** benbrown_ has quit IRC19:17
*** bfletcher has quit IRC19:17
*** jjardon_matrix has quit IRC19:17
*** persia_ has quit IRC19:17
*** lachlanmackenzie has quit IRC19:17
*** gary_perkins has quit IRC19:17
*** cosm has quit IRC19:17
*** _longines has quit IRC19:17
*** zoli_ has quit IRC19:17
*** bjdooks has quit IRC19:17
*** paulsherwood has quit IRC19:17
*** vgrade has quit IRC19:17
*** juergbi has quit IRC19:17
*** jjardon has quit IRC19:17
*** jmacs has quit IRC19:17
*** JPohlmann has quit IRC19:17
*** gary_perkins has joined #baserock19:23
*** vgrade has joined #baserock19:23
*** paulsherwood has joined #baserock19:23
*** inara` has joined #baserock19:23
*** persia has joined #baserock19:23
*** CTtpollard has joined #baserock19:23
*** rdale has joined #baserock19:23
*** mwilliams_ct has joined #baserock19:23
*** bfletcher has joined #baserock19:23
*** benbrown_ has joined #baserock19:23
*** rjek has joined #baserock19:23
*** anahuelamo_ has joined #baserock19:23
*** radiofree has joined #baserock19:23
*** lachlanmackenzie has joined #baserock19:23
*** persia_ has joined #baserock19:23
*** jjardon_matrix has joined #baserock19:23
*** ChanServ has joined #baserock19:23
*** leeming has joined #baserock19:23
*** cyndis has joined #baserock19:23
*** perryl has joined #baserock19:23
*** bjdooks_ has joined #baserock19:23
*** cosm has joined #baserock19:23
*** _longines has joined #baserock19:23
*** dabukalam_ has joined #baserock19:23
*** hitchcock.freenode.net sets mode: +o ChanServ19:23
*** juergbi has joined #baserock19:23
*** jmacs has joined #baserock19:23
*** JPohlmann has joined #baserock19:23
*** SotK has joined #baserock19:23
*** pedroalvarez has joined #baserock19:23
*** hitchcock.freenode.net sets mode: +v pedroalvarez19:23
*** rdale has quit IRC19:28
*** jjardon has joined #baserock19:32
*** zoli_ has joined #baserock19:34
*** trn has joined #baserock19:36
*** gtristan has joined #baserock21:01

Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!