IRC logs for #baserock for Friday, 2015-09-25

*** zoli___ has quit IRC01:27
*** zoli__ has joined #baserock01:29
*** zoli__ has joined #baserock01:38
*** paulw has joined #baserock06:12
*** paulw has left #baserock06:12
*** Walkerdine has quit IRC07:00
*** Walkerdine has joined #baserock07:12
*** Walkerdine has quit IRC07:15
*** zoli__ has quit IRC07:16
*** bashrc_ has joined #baserock08:01
pedroalvarezhey tlsa, your tester scripts need cliapp. Anything else they need?08:12
tlsawdutch: python-novaclient08:13
tlsapedroalvarez: ^^08:13
wdutchsorry?08:13
*** mariaderidder has joined #baserock08:13
pedroalvarezand glanceclient I'd guess08:13
tlsapedroalvarez: yeah, although it just needs the cli tool for that08:13
pedroalvarezcool08:14
wdutchciat-orchestration depends on virtualenv, sandboxlib and python-request08:14
wdutchybd doesn't yet run in a virtualenv which is something I'd like to fix08:15
tlsahttp://ciat.baserock.org:8010/waterfall  <-- looks like it ran out of disc space08:15
* wdutch nods08:15
tlsahttp://ciat.baserock.org:8010/json/help  <--  looks like we can get various bits of info out of it from there08:17
tlsaany ideas on how to visualise it, or what we want to visulise08:18
tlsa?08:18
wdutchwe're going to make our own webview?08:20
tlsathat's the plan08:21
pedroalvarezMy idea: One box per column, green if the latest build was successful, red if not, and yellow but with some "movement" if it's building08:21
tlsapedroalvarez: that's exactly what the top row of the waterfall shows08:22
wdutchby box you men slave?08:22
tlsathe "last build" row shows that08:23
tlsaexcept "active" is yellow rather than moving08:23
pedroalvarezwdutch: no, a square08:23
pedroalvarezhehe08:23
pedroalvareztlsa: also, some lines between boxes representing the triggers08:23
pedroalvarez(that's just my idea)08:24
wdutchyes seeing the flow of the triggers would be useful08:24
pedroalvarezI don't know if you can get that information from the api08:25
*** toscalix__ has joined #baserock08:28
*** rdale has joined #baserock08:33
Kinnisonwdutch: Hey, the build on ciat.baserock.org seems broken with ENOSPC again, and afaict it's *still* running in /src rather than /moresrc08:34
Kinnisonwdutch: is there a reason it hasn't been moved to the disc 10 times larger than the one it's on?08:34
* wdutch patched it to tell ybd to use /moresrc08:35
* wdutch will have a look at what's going on08:35
* Kinnison wonders why it's ENOSPC'ing then08:36
*** jonathanmaw has joined #baserock08:39
wdutchhuh I did a commit to switch back to /src but didn't leave a reason in the commit message08:40
wdutchshould I revert it?08:41
tlsai think it continued to use /src, after pdar added code to delete unpacked system artifacts once they were copied to somewhere else (/archive?) for testing08:41
tlsabut I could be wrong08:41
wdutchthis was last friday, it hasn't been a problem in the meantime08:42
KinnisonIt ENOSPC'd earlier in the week too08:43
pedroalvarezI think we went back to /src becasue it was faster? or we thought it was?08:45
KinnisonErm /moresrc is ssd, /archive is spinning rust08:45
KinnisonThough this is all slightly moot in that I could just detach and remove /src and put /moresrc over the top08:46
Kinnisonif there's nothing precious in /src08:46
pedroalvareznothing that cannot be re-created08:46
Kinnisonwdutch: would you prefer if I dump /src ?08:47
wdutchI have no preference08:47
KinnisonChoose now08:47
wdutchI don't really know what you mean by 'dump'08:49
Kinnisondelete08:50
Kinnisonchuck08:50
Kinnisonthrow away08:51
Kinnisonremove08:51
Kinnisonerase08:51
Kinnisonblast from the face of the earth08:51
Kinnison(terminate the EBS volume)08:51
wdutchand the alternative that I have to choose between is not doing that?08:52
wdutchI don't understand why there must be a preference and why it must be mine08:52
wdutchthere's nothing on /src I care about, I have no idea if other people care about it08:53
paulsherwoodi want to push a branch to g.b.o. definitions - is that still possible since gerrit?08:54
* Kinnison chooses to remove /src08:54
Kinnisonplease hold08:54
tiagogomes_paulsherwood I do it all the time, before realizing that I pushed to the wrong git server08:55
paulsherwoodtiagogomes_: origin git://git.baserock.org/baserock/baserock/definitions.git (push) ?08:55
Kinnisonwdutch: Okay, the disc behind /src is gone and the disc behind /moresrc is now mounted on /src08:55
wdutchk08:56
Kinnisonwdutch: It seems to have some cruft in which might need cleaning up, but I assume (I hope) that things will cope08:56
* paulsherwood is getting could not read from remote repo08:56
tiagogomes_paulsherwood it needs to be ssh:// to push I believe08:56
Kinnisonpaulsherwood: You can't push to git://08:56
Kinnisonpaulsherwood: (yet)08:56
paulsherwoodaha!08:56
* Kinnison will be adding support for signed pushes to gitano08:56
Kinnisonand at that point, supporting push to git:// will be possible08:56
Kinnison\o/08:56
pedroalvarez\o/08:56
pedroalvarezso, tlsa, did you ever had problems using glanceclient against DC?08:57
paulsherwoodtiagogomes_: pls could you provide the full url after the ssh:// ?08:58
tiagogomes_paulsherwood, ssh://git@git.baserock.org/baserock/baserock/definitions08:59
paulsherwoodtiagogomes_: tvm08:59
paulsherwoodhmm... maybe i've been revoked for bad behaviour :)08:59
tlsaI need to imporve my understanding of the firehose driven end of the ciat pipeline09:00
tiagogomes_paulsherwood what is the error?09:00
tlsadoes firehose poll the trove for changes?09:00
paulsherwoodtiagogomes_: denied09:00
paulsherwoodtlsa: i need to improve it too. i can't believe it has to be as complicated as what richard_maw has described09:01
tiagogomes_paulsherwood I assume that you pushing from a machine containing your ssh key, or you log on the the machine using -A09:01
paulsherwoodyup09:01
paulsherwoodah, but there's an error 'ssh: Could not resolve hostname add: nodename nor servname provided, or not known'09:02
* paulsherwood thinks this used to work09:02
*** ssam2 has joined #baserock09:02
*** ChanServ sets mode: +v ssam209:02
paulsherwoodmust be my mac09:02
tlsawhat's the recurring scriptbot thing on the left column of http://ciat.baserock.org:8010/waterfall09:02
tlsawdutch: ^^09:03
wdutchtlsa: I don't really know, it seems to be an arbitrary name for the sendchange, it came from the example code and I just left it in, it's given by the -W flag which is not documented in the manpage09:04
paulsherwoodwhere is the code for this stuff?09:05
paulsherwood(ciat, i mean)09:05
tlsawdutch: OK, so does firehose poll the trove for changes?09:05
wdutchpaulsherwood: http://cu010-trove.codethink.com/cgi-bin/cgit.cgi/cu010-trove/br6/orchestration.git/09:05
wdutchtlsa: trove has a hook which sends json to orchestration (bottlerock) which triggers firehose09:06
wdutchpaulsherwood: also http://cu010-trove.codethink.com/cgi-bin/cgit.cgi/cu010-trove/br6/buildslave-scripts.git/ for code executed on the slaves09:07
tlsawdutch: so why are there so many firehose events that don't start a build?  Changes in upstream branches that our definitions don't track?09:07
wdutchtlsa: not every change in upstream is relevent09:07
wdutchnot every change in upstream will make firehose change definitions and not every change in definitions is a change we care about, I don't know how changes in definitions are evaluated these days, I think all my logic for that has been reverted09:09
paulsherwoodwdutch: thanks!09:10
Kinnisontlsa: it's the "blame" indicator09:23
Kinnisontlsa: meant to be so you can see why an operation was triggered IIRC09:23
tlsaok09:24
* wdutch sees no documentation for that but maybe we could make use of it09:24
KinnisonThe JSON calls it 'blame'09:26
Kinnison(I spent some time yesterday reading the JSON)09:26
wdutchcool09:27
*** lachlan75 has joined #baserock09:48
ssam2initial GNOME system sent for review!09:52
ssam2https://gerrit.baserock.org/#/c/1134/09:52
ssam2it builds, that's about it. But I think we should merge it now and put it in the CI09:53
ssam2otherwise it will just bitrot again09:53
ssam2it may actually work under X in fact, it's only been tested with Wayland (which doesn't work for some reason)09:54
pedroalvareztlsa: so, how did you solve this problem? http://paste.baserock.org/levisufono09:54
* pedroalvarez hopes he did09:54
pedroalvarezI don't get why it would work with one cloud and not with another09:55
pedroalvarezssam2: "put it in the CI" :)09:55
pedroalvarezI need to automate the cleanups in mason :)09:55
paulsherwoodmason will die soon, surely?10:00
pedroalvarezI hope so10:03
pedroalvarezIt has been working hard though10:03
ssam2if it works, I see little reason to kill it, until we have a production quality replacement10:04
edcraggjjardon: why the change to linux-stable for the recent 4.2 update?10:04
pedroalvarezI haven't heard  people complaining about long waits for builds for a long time :)10:05
pedroalvarezso, mason has been a success :)10:05
edcraggunless i've missed something, i don't think we want to base our kernels off linux stable10:06
pedroalvarezedcragg: why?10:06
edcraggit's a stable tree, meant for people who for some reason *need* to stick to a particular kernel version for a long time10:07
jjardonssam2: \o/10:07
jjardonshould I or you vote +1? :)10:07
*** toscalix__ has quit IRC10:07
edcraggpedroalvarez: and why not just stick with mainline?10:08
jjardonedcragg: linux-stable has all the "stable" releases; for example 4.2.110:08
pedroalvarezedcragg: I was just asking why, I don't have opinions yet10:08
pedroalvarezwe have moved to linux-stable some times10:08
edcraggfair enough10:09
edcraggit does mean that technically it will receive updates it wouldn't from mainline all the while 4.2 is still being used10:09
edcraggwhich isn't a bad thing10:10
*** toscalix__ has joined #baserock10:10
edcraggalthough that wouldn't change since baserock is using a SHA as a ref10:12
edcraggso i guess it makes no difference10:12
edcragg(changes to stable branches are hand-picked by a maintainer for each version stable branch, and retroactively applied to the stable branches - for stability and other critical fixes)10:15
jjardonyeah, at some point I would like to track linux-4.2.y instead a fixed sha, but yeah now doesnt make difference10:16
edcraggyeah, makes sense (especially with integration). i still don't see that there's any particular advantage in basing it off a stable branch, particularly, though10:17
*** lachlan75 has quit IRC10:18
pedroalvarezjjardon: I hope that CIAT will do that for us (track a branch, change the sha1, test the change, and send the patch for review)10:20
pedroalvarezsame for gnome ;)10:20
jjardonI hope it can handle it :) this is a first iteration of the GNOME system, and still some bug components are missing (like webkitgtk)10:22
*** Lachlan1975 has joined #baserock10:24
ssam2*big10:24
ssam2I guess we can both +1 but neither of us should +210:24
ssam2maybe that is fair10:24
Kinnisonwdutch: orchestration on ciat seems to believe its address is 0.0.0.0 -- can you teach it to believe that it is ciat.baserock.org instead?10:24
Kinnisonwdutch: otherwise the URLs in the JSON output are broken10:24
wdutchokay10:24
pedroalvarezand port 80 :)10:24
pedroalvarezor not10:24
pedroalvareznevermind10:24
pedroalvarezwe better put something nicer in port 8010:25
wdutchmay I take it offline briefly?10:25
pedroalvarezwe should assume that it is going to be an intermitent service10:26
KinnisonNo, I'm going to put other stuff on port 8010:27
ssam2the announcement email Daniel didn't actually *say* that it was a prototype and shouldn't be relied upon for important work yet, but I kind of assumed that it still is10:28
pdaris there a convention of what ports to put things on?10:29
richard_mawhttp is 80, 80XY is usually used for testing web services10:30
*** zoli__ has joined #baserock10:30
richard_mawsee /etc/services10:30
wdutchpdar: if you're interested in general:http://overapi.com/static/cs/common_ports.pdf10:30
pdarand what happened to moresrc?10:30
pdarthanks richard_maw wdutch10:31
pdarwhoops, it seems I put the image server test thing on the internet radio port10:32
pedroalvarezare you listening your favourite images?10:33
*** gary_perkins has quit IRC10:34
*** toscalix_ has joined #baserock10:34
pedroalvareztlsa: yay, the problem with deploying to DC is because an eventlet bug.10:35
pedroalvarezI should upgrade the eventlet version of baserock to 0.17.410:35
* pedroalvarez makes a note10:35
pdarhas src breen replaced with moresrc on the ciat?10:37
*** toscalix__ has quit IRC10:38
*** gary_perkins has joined #baserock10:39
wdutchpdar: I believe that Kinnison did that, yes10:39
pdarwdutch: okay, ta i shall put the cache archive thing elsewhere then10:40
pdarare we alowed to play with the master buildslave scripts again?10:42
Kinnisonpdar: Put it in /src, but ideally in a subdir somewhere10:42
Kinnison:-)10:42
pdarKinnison: okeydokey10:43
* wdutch has problems replacing 0.0.0.0 with ciat.baserock.org, ciat may be down a bit longer10:44
Kinnisonwdutch: if it wants it to be an A record not a CNAME then use ciat-aws-0.baserock.org10:48
wdutchlooks to be working http://ciat.baserock.org:8010/waterfall10:52
wdutchbottle didn't like it though10:52
Kinnisonbottle is after a bind address so 0.0.0.0 is fine for that10:52
Kinnisonthe buildbot needs a proper address to tell others10:52
wdutchokeedoke10:52
*** toscalix_ has quit IRC10:52
*** toscalix_ has joined #baserock10:52
Kinnisonthough the json still says 0.0.0.010:52
Kinnisongah10:52
Kinnisonconsider http://ciat.baserock.org:8010/json/builders/1.%20Integration/builds/_all?as_text=110:53
Kinnisonsee the log URLs near the top?10:53
*** CTtpollard has joined #baserock10:53
CTtpollardWhat is the genivi system baserock is building with the new CIAT?10:54
pedroalvarezCTtpollard: GDP10:55
pedroalvarezfrom an old branch that jonathanmaw had around I believe10:56
KinnisonCTtpollard: genivi-demo-platform-x86_64-generic10:56
CTtpollardvery nice10:56
pedroalvarezi'm currently thinking that maybe the network-id to use in the ciat testing belongs more to the o hosts configuration than to the systems configuration10:59
pedroalvarezbut the flavours...  meh11:00
Kinnisonmmm flavours11:00
paulsherwoodi'm about to forward kinnison's email to a lot of folks... can i assume  the ciat url will remain reasonably solid?11:01
*** toscalix_ has quit IRC11:02
KinnisonYes, if you want11:04
paulsherwood:-)11:04
KinnisonI arranged the DNS so that when we migrate the work to a different system, we can flip the CNAME trivially11:05
paulsherwoodKinnison: i assume you don't object to the forwarding? i think my cover email will be less entertaining, but more disruptive11:05
Kinnisonpaulsherwood: my only concern is that it points at the known-pants UI rather than the nicer one we hope to create in the coming week if we can find time11:05
paulsherwoodsure11:06
Kinnisonpaulsherwood: If you're okay with the pig going out sans-lippy then forward it11:06
paulsherwoodi'm ok with it11:06
* Kinnison wonders who paulsherwood intends to disrupt :-)11:07
paulsherwood:-)11:07
pedroalvarezoh! finally spotted the problem with `nova boot`11:17
pedroalvarez-                '--nic', 'net-id', self.net_id,11:18
pedroalvarez+                '--nic', 'net-id=' + self.net_id,11:18
Kinnisonpedroalvarez: yay11:19
paulsherwoodwhat about using meeor for the ciat ui?11:19
paulsherwoodthat + colour + rectangles with rounded corners == shiny.11:20
wdutchpaulsherwood: the advantage would be the reactivity11:20
paulsherwoodwdutch: plus it's fast to hack in, iirc?11:20
wdutchthe main disadvantage of meteor would be the database11:20
wdutchI din't know if they support sqlite11:20
Kinnisonmeteor seems to be entirely pointless11:20
Kinnisonfor what we need11:20
Kinnisonwhich is consuming some semi-static data and then a JSON API11:21
Kinnisonto render data11:21
paulsherwoodwe just need shiny :)11:21
wdutchit's fast to hack in if you are comfortable with javascript, which a lot of people here are not :P11:21
paulsherwoodKinnison: and in other conversations you've admitted you are the wrong person to talk about graphical ui :-)11:21
wdutchmaybe CTtpollard has an opinion, he knows meteor O:)11:21
paulsherwoodCTtpollard: cmon, say yes!11:22
Kinnisonpaulsherwood: I'm not the wrong person for frameworks though11:22
Kinnisonpaulsherwood: I just dislike the shiny11:22
pedroalvarezhe can say yes if he is going to do it :)11:22
CTtpollard.....11:22
paulsherwoodmaybe i should do it myself :-)11:22
pedroalvarez+111:22
pedroalvarez:{P11:22
paulsherwoodi'm pretty busy next week, though11:23
ssam2is the JSON API going to be exposed as well?11:23
KinnisonIt already is11:23
ssam2so there could theoretically be multiple UIs ?11:23
paulsherwoodyup11:23
Kinnisonssam2: indeed11:23
ssam2(not that it's a good idea to try to do 2 for next week, just curious)11:23
* wdutch is reminded of http://www.commitstrip.com/en/2015/09/16/how-to-choose-the-right-javascript-framework/11:23
Kinnisonssam2: http://ciat.baserock.org:8010/json/help11:23
ssam2ah, cool11:24
Kinnisonssam2: The lipstick for the pig would be something consuming a description of the pipeline (which we could generate as we generate the buildbot and firehose configs) and consuming that JSON API in some fashion11:24
ssam2so it'd kind of visualise the pipeline, and show status?11:25
Kinnisonssam2: at least initially yes11:25
ssam2this sort of thing is tricky, I've not actually seen a UI that I like for a CI tool, ever11:25
Kinnisonssam2: eventually it could also consume descriptions of the elastic resources etc11:25
Kinnisonssam2: and show things like the build/test farm size and direction of growth/shrinkage11:25
Kinnisonssam2: etc.11:25
ssam2actually http://build.gnome.org/ is quite nice, but probably because it's so simple. CIAT couldn't really be so simple11:25
paulsherwoodyup. i was hoping we could use concourse, but from my initial explore there it's way more code/complexity than i expected11:26
Kinnisonssam2: a white page with a list of builds down the left11:26
Kinnisonpaulsherwood: concourse is an entire CI system afaict11:26
paulsherwoodyup11:27
ssam2Kinnison: it used to show the test results in a more fancy way :-)11:27
paulsherwoodand it's dockertastic11:27
rjekIs there a guide to how to interpret the waterfall page?  I find it quite confusing11:27
Kinnisonpaulsherwood: and there's a strong reason not to use it11:27
paulsherwoodyup11:27
Kinnisonrjek: that the waterfall page is confusing is exactly why we're trying to re-skin things11:27
rjekAh11:27
* rjek will be quiet, then11:27
ssam2but actually, once it goes beyond a list of builds, it becomes really confusing to me, always. Maybe there's a clever way to make more complex systems easily comprehensible, but I've not seen one yet11:27
rjekEnjoy it while it lasts11:27
Kinnisonrjek: in brief, the columns are the pipeline stages, (kinds of work to do) the vertical axis is time, and there's no visualised linkage for triggers11:28
ssam2but since there is an API exposed, we enable research in this area :-)11:28
Kinnisonssam2: :-)11:29
SotKIs there a common opinion (however vague) on what the shiny UI should have in it?11:30
rjekMirrors.11:31
KinnisonSotK: Currently the person with the strongest vision on what shiny we need is paulsherwood11:31
KinnisonSotK: I am the person with the weakest vision wrt. the UI and possibly the strongest opinions wrt. the backend complexity11:31
wdutchdoes ybd record which artefacts it produced when it builds a system?11:35
KinnisonI think pdar was working on something to do that11:35
wdutchhaving CIAT keep track of which systems are built in order to trigger deploys of clusters depending on multiple systems will depend on the artefacts not being removed, is this okay for now or too hacky?11:36
wdutchthen again a deploy will just build them again ...11:37
wdutchor try to11:37
KinnisonI'm fine if for now we assume things just built won't be thrown away11:39
Kinnisonlater we can consider ways to manage locking to ensure things11:39
pedroalvareztlsa: can you explain to me the TODO of the line 236?11:42
pdarwdutch: regards 12:35. It doesnt, I have a branch that can do that if you need it11:46
wdutchpdar: awesome, I'll spin up a test server and we'll give it a go, I need some lunch first though, I'll ping you in a bit :)11:47
*** CTtpollard has quit IRC11:56
*** tiagogomes_ has quit IRC11:59
*** tiagogomes_ has joined #baserock11:59
pedroalvarezI just had a weird backtrace when pushing to the trove that ciat is using12:05
pedroalvarezhttp://paste.baserock.org/manoxayana12:05
paulsherwoodi've pushed back on pdar's change so far.. https://github.com/devcurmudgeon/ybd/pull/12412:15
paulsherwoodam i missing something? wouldn't this be best in the meta file?12:16
paulsherwoodwdutch: ^^12:16
KinnisonThe big issue is that we need to know, for a given run of ybd, the exact cache key of the system built12:17
straycatfor ui, maybe worth looking at  http://hydra.nixos.org/project/nixos#tabs-project for inspiration? pretty similar system?12:18
*** straycat is now known as sleepycat12:21
KinnisonSo I like that, but I fear paulsherwood will think it doesn't match his vision12:22
wdutchunless I'm missing something it's just lists which doen't do any of the work of helping you visualise it12:27
tlsapedroalvarez: I don't have a todo on that line12:31
tlsahave you pulled the last version I pushed?12:31
tlsapedroalvarez: or just paste the full comment here?12:31
wdutchpdar: paulsherwood: The metafile where where I looked when wondering if it kept track of the system. It would make sense to me12:31
*** sleepycat is now known as tiredcat12:43
paulsherwoodKinnison: that's in the log from ybd?12:46
Kinnisonpaulsherwood: I'd far prefer a proper interface rather than loggrepping12:47
paulsherwoodKinnison: interface how? where?12:48
* Kinnison wants a tool which answers the question "Given these definitions, what is the cachekey for this thing"12:48
KinnisonBut failing that, having ybd be able to be explicitly told "Write out the cache keys of what you build to here" would do12:48
paulsherwoodok i understand12:48
KinnisonMy eventual goal is a suite of smaller tools (smaller even than ybd) which cover the use-cases12:49
paulsherwoodi think the nixos approach is simple and sensible. i have a short term interest to see something shiny since we're off to ELCE with it, that's all12:49
paulsherwoodKinnison: ybd is already useable as a library12:50
paulsherwoodso tool could parse definitions and ask cache_key(target)12:50
Kinnisonpaulsherwood: if you can give pdar some hints as to how he might achieve that, that'd be great12:51
Kinnisonpaulsherwood: or even knock up a tiny tool for it yourself if you think it's trivial12:51
* Kinnison doesn't have the brainspace to try and grok ybd's "library"12:51
paulsherwoodssam2: could you explain the approach?12:52
ssam2this is as far as I got with using YBD as a library: https://github.com/CodethinkLabs/research-git-active-days/blob/master/measure.py12:58
ssam2you can get a ybd.Definitions() instance, which lets you do a couple of things, including get the results of YBD's cache key calculation I think12:58
ssam2hmm that example just uses the ybd.repos module, not the cache keys12:59
paulsherwoodi guess it's noisy (outputs all the cache-keys while parsing defintions)12:59
KinnisonI don't care about noise12:59
Kinnisonlogging can be shuffled elsewhere12:59
paulsherwoodybd.cache.cache_key(system) would be the magic call after following same setup as ssam2's measure.py12:59
ssam2yeah, that should work13:00
ssam2where you can find system with Definitions.get('systems/foo.morph') I think13:00
*** bashrc_ is now known as bashrc13:03
bashrcdoes baserock have a definition for lttng ?13:04
KinnisonI don't believe we currently include that anywhere, no13:05
bashrcok, was just wondering13:06
bashrchaving a good tracing system might be useful for embedded13:06
richard_mawis lttng one of those that require non-mainline kernel patches?13:10
richard_maw'cause there's already upstreamed tracing tools AIUI13:10
bashrcwhich are those?13:15
* richard_maw can't remember the names13:15
richard_mawbut there's talk of traces being {e,}BPF filters13:16
Kinnisonthere's things like ftrace13:16
bashrcyes, ftrace is built in13:16
bashrcbut afaik is kernel specific, not a general tracing system13:17
* richard_maw vaguely remembers talk of one of the kernel tracing things being extended to userland13:18
*** petefoth has quit IRC13:20
pedroalvareztlsa: I was refering to this chunk of code: http://paste.baserock.org/ruwesoxidu13:31
Kinnisonrichard_maw: I think it's perf13:31
tlsapedroalvarez: looks redundant13:34
tlsa(the TODO)13:34
pedroalvarezok, it's working anyway :)13:35
tlsareally? the whole thing works?13:35
* tlsa is amazed13:35
pedroalvarezwell, it fetched the images, created them, and then created the instances and waited for ssh13:36
pedroalvarezgiven that they were 0byte images, nothing else happened :P13:36
radiofreewhy does gerrit need flash13:39
radiofreenever mind, "copy to the clipboard"13:40
paulsherwoodthinking about this need for cache_key(system).... is the deploy step being given a completely new checkout, or using the same one as the build?13:48
pedroalvarezpaulsherwood: I would assume a new checkout. It may happen in a different machine13:49
paulsherwoodok13:49
*** jonathanmaw has quit IRC13:49
pedroalvarezI may be wrong, though13:49
paulsherwoodyou're probably right13:50
tlsapedroalvarez: still, it's got further than I could have hoped, given it was hastily thrown together and untested when I handed it over :)13:51
pedroalvarezwell, I found a couple of tiny bugs :)13:52
pedroalvareztesting now with a real image13:53
tlsapedroalvarez: ah, ok :)13:53
tlsathe image you deploy needs to have an authorized_keys with the public key from the system/user running the test runner13:54
paulsherwoodpdar: i've merged your cache-log chang13:54
pedroalvareztlsa: oops, good to know13:54
* pedroalvarez ^Cs13:54
*** jonathanmaw has joined #baserock13:56
pdarpaulsherwood: cool, thanks paul!13:56
paulsherwoodnp... it's a workable solution - if i think if something better we can improve later14:03
paulsherwoodwhich was the systemdd patch that required a fix in definitions? can someone provide links please?14:14
pedroalvarezyup, one sec14:15
pedroalvarezpaulsherwood: http://git.baserock.org/cgi-bin/cgit.cgi/delta/systemd.git/commit/?id=d379d44255469f03994832ab5821bf1b9034f4dc14:16
*** zoli__ has quit IRC14:22
paulsherwoodpedroalvarez: tvm14:23
paulsherwoodand the problem was what precisely, in one line?14:23
pedroalvarezsystemd requred a new version of util-linux14:24
pedroalvareznewer version of util-linux triggered a problem in our integration of readline14:26
pedroalvarezso there were 2 changes in definitions: fix readline, and upgrade util-linux14:26
paulsherwoodtvm14:26
paulsherwoodnow... it just occurs to me ... is that new version or readline GPLv3?14:27
paulsherwoods/or/of/14:27
pedroalvarezpaulsherwood: no no, I didn't upgrade readline, I fixed the integration14:28
paulsherwoodaha :)14:28
paulsherwoodtvm14:28
pedroalvarezI'm aware of the readline and GPLv3 problem :)14:28
pedroalvarezanytime :)14:28
*** franred has joined #baserock14:31
pedroalvareztlsa: testing images work :) I've added a warning when the tester tries to test in systems that weren't defined14:49
pedroalvareztlsa: I'll try to make it show the output of the tests14:50
tlsacool15:03
pedroalvarez:)15:03
tlsastdout=self.output15:04
tlsaadd that to the runcmd call?15:04
pedroalvarezyup, I was going to try that15:04
pedroalvarezI wonder why we are not using cliapp.ssh_runcmd (or however it's called)15:06
richard_mawpedroalvarez: insufficient support for temporary auth options15:06
pedroalvarezaha! makes sense15:07
*** zoli__ has joined #baserock15:16
pedroalvarezyay! output!15:26
tlsa\o/15:27
pedroalvarezthings left: run commands as a given user (should be easy)15:27
pedroalvarezand proper clean ups15:27
tlsadid you need to change much to get it going?15:28
pedroalvareznope15:29
tlsacool15:29
*** fay_ has quit IRC15:38
pedroalvareznot sure if  I should redirect stderr too. I'll leave it as it is for now15:43
*** rdale has quit IRC15:52
*** fay_ has joined #baserock16:24
*** mariaderidder has quit IRC16:31
tlsapedro: The only way to run it without the output order messing up when the output is captured as a log is to run it with `python -u tester`16:36
tlsacos stdout gets buffered and stderr doesn't unless you force it to run unbuffered like that16:37
*** fay_ has quit IRC16:39
*** ssam2 has quit IRC16:45
pedroalvareztlsa: aha, thanks for the info!16:51
* jjardon goes to celebrate that the first chunk of the GNOME system has been integrated in definitions :)16:53
pedroalvarezright, if redirect the output to a file then the output is messed up :)16:54
* pedroalvarez tries `python -u`16:54
pedroalvarezhm.. that didn't solve the problem16:56
pedroalvarezI'll forget about this for now16:56
*** bashrc has quit IRC16:57
*** jonathanmaw has quit IRC17:00
*** petefoth has joined #baserock17:00
*** fay_ has joined #baserock17:16
*** fay_ has quit IRC17:18
*** franred has quit IRC17:26
*** tiagogomes_ has quit IRC17:59
*** Lachlan1975 has quit IRC18:28
*** zoli__ has quit IRC20:26
*** zoli__ has joined #baserock20:35
*** zoli__ has quit IRC21:05
*** zoli__ has joined #baserock21:17
*** zoli__ has quit IRC22:35

Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!