IRC logs for #baserock for Friday, 2015-04-24

*** zoli__ has joined #baserock00:17
*** zoli__ has quit IRC00:22
*** zoli__ has joined #baserock02:06
*** zoli__ has quit IRC02:10
*** zoli__ has joined #baserock04:00
*** zoli__ has quit IRC04:32
*** zoli__ has joined #baserock04:51
*** edcragg has joined #baserock06:41
*** Albert_ has joined #baserock07:18
*** mariaderidder has joined #baserock07:42
*** a1exhughe5 has joined #baserock07:44
*** edcragg has quit IRC08:06
*** mauricemoss_ has joined #baserock08:12
*** ssam2 has joined #baserock08:16
*** ChanServ sets mode: +v ssam208:16
*** jonathanmaw has joined #baserock08:25
*** tiagogomes_ has joined #baserock08:41
*** edcragg has joined #baserock08:42
jonathanmawhrm, I have systemd user sessions, but shutdown can take up to 90 seconds because it shuts things down in the wrong order. fedora has a workaround
jonathanmawdo we want changes in the repo, or in the chunk morphology?08:44
jonathanmawfor that matter, what's the preferred way to add config files, like systemd units and pam config? I've been using heredocs in install-commands, but that's starting to look very messy.08:46
*** gary_perkins has joined #baserock08:46
franred_jonathanmaw, IMO adding the custom systemd unit to the chunk would be enough - adding a comment for why we overwrite it/them.09:01
jonathanmawfranred_: to the chunk morphology, or in the source code?09:02
franred_chunk morphology09:02
*** franred_ is now known as franred09:03
pedroalvarezcould we persuade systemd to add this workaround by default?09:07
*** rdale has quit IRC09:08
*** rdale has joined #baserock09:09
*** lachlanmackenzie has joined #baserock09:14
KinnisonI imagine they'd rather "fix it properly"09:33
* richard_maw has discovered that bits of nfs-utils were broken by our gcc update, and is preparing a patch09:49
richard_mawalso, for the ostree patches we need to go through all the write extensions and have them deliberately break hardlinks09:54
richard_mawI currently only know that rsync isn't doing this09:55
*** franred has quit IRC09:57
pedroalvarezhrm.. that bug may be the cause of my failed attempts of using pxeboot.write10:10
richard_mawpedroalvarez: that may be the case if your devel system is sufficiently new10:13
tlsafreetype's auto hinting can be improved when it uses harfbuzz, but harfbuzz deps on freetype.  So to get nice hinting, you have to build freetype, then build harfbuzz, then rebuild freetype with FT_CONFIG_OPTION_USE_HARFBUZZ10:18
rjektlsa: Eugh10:19
KinnisonYay intrinsic build-dep loops10:19
Kinnisonthey make me so happy10:19
rjekThese are tears or pure joy10:19
richard_mawthis makes me want multi-source chunks10:19
KinnisonTo produce a freetype+harfbuzz chunk which deals with that internally?10:20
richard_mawKinnison: yeah10:20
richard_mawif the two projects are that incestuously linked then I'm not sure it's sane to build them separately10:21
richard_mawthough wasn't harfbuzz split out of freetype anyway?10:21
tlsafreetype is C10:21
tlsaharfbuzz is C++10:21
rjekWhat Harfbuzz does is outside of Freetype's scope.10:21
rjekFreetype is just a renderer and font metric library, basically10:21
tlsaharfbuzz is layout10:22
rjekHarfbuzz is about typesetting and layout10:22
tlsaand also glyph sustitution10:22
jjardonIf someone have some time:
ssam2also, patch to update Morph to latest:
ssam2after that is merged, I'm going to tag definitions with 'baserock-15.17-rc' as a kind of release candidate10:54
ssam2I need to deploy something stable next week, so need a tag, but I don't think it's worth (or practical) doing a full release today10:55
ssam2jjardon: that patch looks good, thanks!10:55
*** franred has joined #baserock10:56
radiofreessam2: regarding the rc, can you patch the jetson clusters to use BOOT_DEVICE and ROOT_DEVICE11:04
radiofreeso release and jetson-upgrade11:04
ssam2yes, ok11:04
ssam2do you have a patch?11:04
radiofreehold on11:04
ssam2radiofree: thanks, pushed to master11:11
pedroalvarezoh, thanks for that patch11:22
jjardonmwilliams_ct: Zara when you have some time, it would be great if you can test some of the armv5 patches: and
jjardon(follow the instructions from
*** tiagogomes_ has quit IRC12:57
*** tiagogomes_ has joined #baserock12:57
pedroalvarezanyone with a 4.0 kernel baserock system (x86_64) can run this? ` zcat /proc/config.gz | grep ROOT_NFS`13:26
* pedroalvarez will see the default config13:27
franredpedroalvarez, same here13:29
pedroalvarezpaulsherwood, franred  t hanks13:29
* pedroalvarez discards a theory13:29
* straycat ♥ partial builds13:36
paulsherwoodstraycat: is that in master now?13:40
paulsherwoodcool ;)13:42
* richard_maw just hopes it won't be misused13:44
KinnisonOf course it will13:45
Kinnisonwe're humans aren't we?13:45
paulsherwoodis there documentation, or a video tutorial maybe?13:45
Kinnisonpaulsherwood: documentation is always wrong, just read the code13:45
* Kinnison turns down the sass13:45
straycatmight be useful to have something on the wiki somewhere13:46
radiofreeif you document it then it absolutely will be misused13:46
richard_mawhas anyone else seen doffm's 'google compute engine' support patches?13:46
paulsherwoodKinnison: i love it when you hurl my words back at me. note however that i do make videos, which hopefully help to convey the approach, even if the details change13:46
richard_maw and
Kinnisonpaulsherwood: :)13:47
* persia worries that "partial deploys" will break things, and hopes that the existence of this feature doesn't dissuade people from writing `morph hack`13:53
straycati thought partial builds was morph hack?13:54
franredrichard_maw, I did, I was waiting for him to answer your suggestions14:00
jjardonoh, gcc 5.1 has been released!14:04
radiofreetime to upgrade :D14:06
tiagogomes_or no14:08
tiagogomes_upgrading gcc will normally break things, and sometimes you only realize it weeks after the upgrade14:09
radiofreei was, of course, being entirely serious14:09
pedroalvarezHehe, maybe with it we can remove gmp workarounds in gcc chunks14:10
* jjardon files
jonathanmawsomething weird is going on with systemd and pam. I tried to improve on the hacky way I did it yesterday (have login start, as systemd attempts to do it 'properly' by having user@.service use the systemd-user pam config14:11
jonathanmawas far as I can tell, this doesn't work because user@.service won't get started unless is called by something else14:12
franredjjardon, that was easy... the one which follows it will be trickier14:13
Kinnisonjonathanmaw: what are you trying to achieve?14:15
jonathanmawKinnison: yesterday I had in login's pam config. I was seeing if it was possible to get systemd's pam config to hold that14:16
jonathanmawuser@.service has PAMName=systemd-user, which is where ought to be put14:16
Kinnisoneach pam-aware thing runs its own root "thing"14:16
Kinnisonso login always runs the 'login' pam.d entry14:16
jonathanmawah, I see how yocto did it14:18
jonathanmawtheir system-session always does pam_systemd.so14:18
jonathanmawhence login will have inherited it, etc. etc.14:18
*** zoli__ has quit IRC14:28
doffmrichard_maw: Basically agree with all comments on google-compute-engine. How do I go about addressing them an uploading a new patch?14:47
doffmI actually tried setting the mtu via a network file. Didn't work, which is why i have the wierd systemd files.14:48
doffmI'll try again with your suggestion.14:48
persiastraycat: `morph hack` was supposed to parse definitions, and run the build commands for a chunk against the local working tree (or just bare directory, to support things like build-from-tarball).  I thought partial builds just built chunks for part of a system, rather than an entire system, but from the git sources as usual.14:52
persiaThat said, I may have misunderstood robtaylor's description of `morph hack`, so I may be entierly wrong.14:52
KinnisonThat's not what I understood morph hack to be14:54
KinnisonBut then there's been many attempts at describing 'hack'14:54
persiaWhat was your understanding of `morph hack`, and to whom do you attribute it?14:54
Kinnisonmy understanding was that it would elevate a chunk into a hack stratum at the top of the system (farthest from the leaves) and apply edits there, rather than replacing the unedited chunk deeper in the stack.  Anything "hacked" would build with hacked variants of their build-deps (in case you're doing a multichunk hack) and during system artifact construction the hacked chunks would replace their unhacke14:56
Kinnisond variants during unpack/link14:56
KinnisonI attribute it to myself and robtaylor discussing it, and further ponderings between myself and richard_maw14:57
jjardonwow, the kernel used in openbmc is so old that the Makefile fails to build with a modern make; more than 6 years old!14:58
rjekjjardon: Erk!14:59
KinnisonErm, urgh15:02
ZaraHi, I'm trying to pull some changes from gerrit into my checkout. I ran git-review -d <change Id> but got an error: . What's going wrong?15:11
jjardon"2015-04-24 15:10:36 system minimal-system-armv5l-wedge is cached at /src/cache/artifacts/e3a3f3bed1804b3e1359b0c217c636f3078e6adde676aebd3ae95da9faad3acd.system.minimal-system-armv5l-aspeed-rootfs" :) now lets investigate how to deploy this15:12
persiaKinnison: Hrm.  That interpretation also sounds very useful.  I suppose both need better names :)15:16
Kinnisonjjardon: congrats15:16
pedroalvarezjjardon: minimal? meh! put openstack on armv5l!15:17
persiaIt's been done before.  Should be relatively easy.15:18
persiaErr, nevermind.  It's been done on armv6l: my mistake.15:18
jmacsDo we have baserock on armv6l now?15:19
persiaI seem to recall someone talking about putting it on a PI, which would imply the capability exists, but I don't think the relevant work was shared with the rest of us, so for most interpretations of "we", I think the answer is "no".15:20
persiaThe OpenStack on PI stuff I saw was non-Baserock.15:20
jmacsI know pedroalvarez was working on it15:20
jjardonjmacs: not yet15:21
pedroalvarezopenstack on PI?15:21
persiaThis was a while ago, when it seemed like a better idea than it does now.15:21
jmacsTune in next week for OpenStack on a Commodore 64.15:22
persia(where "while" is measured in "years")15:22
rjekAnd the week after that an Acorn System 1.15:22
persiajmacs: Is there a working 8-bit python interpreter?15:22
pedroalvarezI was joking about openstack on armv5, I assume that jjardon's arv5 board is tiny15:22
rjekPython's core is ANSI C, but I don't think it's as portable as Lua's ANSI C core (which runs quite happily on 6502)15:23
jmacspersia: I doubt it, but there are some odd people out there.15:23
rjekI've certainly seen Python running on small 16-bit systems15:23
KinnisonYou don't need one15:24
KinnisonJust run a 32bit CPU emulation on the 8bit system15:24
rjekThat's how somebody booted Linux on an Arduino.15:25
jmacsWith cassette-backed swap.15:25
rjekARM emulator + bit-banged SIMM15:25
rjekAVR there15:26
flatmushrjek: that's disgusting, hasn't that guy heard of FPGAs?15:39
pedroalvarezfranred: so, ceilometer has a dedicated service for the storage node?15:46
franredpedroalvarez, no, depends on if this exists or not ceilometer will run in the storage node or in the compute node (if this is the node where cinder runs)15:47
pedroalvarezfranred: I will rephrase it a bit15:48
*** CTtpollard has quit IRC15:48
pedroalvarezfranred: if the openstack setup has a compute node to run nova-compute, and a storage node to run cinder, do you need to run anything from ceilometer on the storage node?15:48
franredin that case ceilometer-compute should run in the storage node and not in nova-compute15:49
franredpedroalvarez, ^^15:50
pedroalvarezfranred: hm.. then sounds like they named it wrongly15:51
franredpedroalvarez, ... Im lying to you...15:52
franredand to everybody, I will fix the patches and resend them15:53
jonathanmawaha! slow shutdown is because systemd-exit.service failed, because it was looking for /usr/bin/kill, when we have /bin/kill15:53
Albert_Having trouble with Genivi AudioManager. I'm building from source. The one currently running is v6.0.1. The repo only has 6.0, or 6.1, etc. No 6.0.1. I try to build 6.0 and it fails to buildv(v6.2 builds OK). Is there a known issue?15:53
*** a1exhughe5 has quit IRC15:54
pedroalvarezfranred: then, do we need ceilometer services on the storage node?15:55
franredpedroalvarez, no, we don't15:55
franredpedroalvarez, patches re-sent16:00
jonathanmawso, the reason why systemd gets the wrong place for kill, is that the base-system-x86_64-generic does not include strata/coreutils-common, which installs /usr/bin/kill16:05
jonathanmawbut coreutils-common is in the staging area16:05
*** Albert_ has quit IRC16:09
richard_mawI'm bloody sick of having split /usr16:30
persiaThen remove it.16:44
nowsterpersia: Lennart, is that you?16:47
*** mariaderidder has quit IRC16:49
ssam2Release candidate baserock-15.17-rc is released!16:50
ssam2it's up to you what you want 'release candidate' to mean :) there's a tag, and x86 builds.16:50
ssam2'ostree' the tool is in this release, but Morph does not have the code to use OSTree, and it turns out quite a bit more work is needed on that16:53
*** jonathanmaw has quit IRC16:53
ssam2current version of the code is:, and it largely works, but there are some big isues. Many write extensions can't handle deploying from a hardlink checkout without creating a broken system. It also breaks cross-bootstrap16:54
paulsherwoodouch :)17:00
*** gary_perkins has quit IRC17:06
ssam2time to go ride bikes!17:06
*** ssam2 has quit IRC17:07
*** tiagogomes_ has quit IRC17:13
* pedroalvarez is adding support to pxeboot write to choose the target interface device to boot from17:13
pedroalvarezI found that one machine was loading the kernel from one interface, and then the rootfs from the other interface just because dhcp was faster on the other one17:14
*** mauricemoss_ has quit IRC17:26
*** edcragg has quit IRC17:32
*** zoli__ has joined #baserock17:39
*** rdale has quit IRC17:43
*** straycat has quit IRC17:54
*** zoli__ has quit IRC17:57
*** lachlanmackenzie has quit IRC18:19
*** zoli__ has joined #baserock18:24
*** zoli__ has quit IRC23:04

Generated by 2.14.0 by Marius Gedminas - find it at!