IRC logs for #baserock for Tuesday, 2015-10-27

*** gtristan has quit IRC02:24
*** gtristan has joined #baserock02:45
*** richard_maw has quit IRC02:47
gtristanpaulsher1ood, we can close #137 if you like... obviously it's not what I thought it was05:53
gtristanI'm sure I saw it block (dont have the log anymore), but seems to be unrelated to what I thought it was05:54
gtristanok branch up:
* gtristan goes for lunch06:13
*** gtristan has quit IRC06:13
*** gtristan has joined #baserock07:04
*** paulwaters_ has joined #baserock07:23
KinnisonHmm, richard_maw has vanished07:28
petefothwas there a puff of magical smoke?07:43
KinnisonNo idea, I have joins and parts hidden07:43
* petefoth spots the smoke at 2:47am: caused by Mr Ping TImeout07:45
KinnisonSadly his absence, and the lack of coffee (wdutch has put a pot on, but it's brewing yet) means that I've just gone and added hide=1 support to master of gitano07:45
* Kinnison growls at users and their being nice07:45
petefothLAck oif coffee can do terrrible things :)07:45
KinnisonWhen Richard turns up, can someone tell him (a) I did the hide=1 thing and (b) he owes me a yarn for it :-)07:55
* Kinnison was a naughty moo and pushed an untested (well, un-validatable) change to master07:55
* petefoth will keep an eye out for maw[0]07:56
paulsher1oodgtristan: ack08:10
* gtristan gets nostalgic08:20
* gtristan runs cvs08:20
jjardongtristan: rebase your branch against master and submit again to avoid conflicts (webkitgtk is already in GNOME stratum)08:44
* gtristan did rebase :-/08:45
gtristandid it upload that patch again ?08:45
* gtristan double checks his tree08:45
gtristanjjardon, I pushed that branch off of a branch freshly rebased off of master08:50
gtristanjust pulled master and rebased said branch again, both say I'm still "up to date"08:51
* gtristan can git review again, just for the shits and giggles :)08:51
gtristanjjardon, the push fails with a "(no new changes)" message08:53
gtristan ! [remote rejected] HEAD -> refs/publish/master/gnome-initial-setup (no new changes)08:53
gtristanjjardon, gerrit is just on crack again I'll bet... maybe it doesnt understand that the preceding 'glib-networking' patch must absolutely be applied *before* the gnome-initial-setup patch; as it is ordered in the branch08:57
pedroalvarezI see glib-networking as the first one to be merged09:02
pedroalvarezMorning guys!09:03
pedroalvarez(and girls)09:03
pedroalvarez(and cats)09:03
*** bashrc has joined #baserock09:04
gtristanmorning pedro09:05
gtristanI see it that way too, well, "the gerrit UI shows it on top", if that is any indication at all at what it's trying to do09:06
gtristananyway, I expect it to conflict if one would attempt to apply them in the wrong order; I had to reorder webkitgtk so that it comes before online accounts - so there is churn there09:09
*** gary_perkins has quit IRC09:13
pedroalvarezYes, you can expect that09:14
pedroalvarezI normally use gertty if I want to see a more sane UI09:15
*** nowster has joined #baserock09:15
*** jonathanmaw has joined #baserock09:15
gtristanok so... I'm drawing a blank here: Am I expected to do something more about this ?09:16
gtristanAre we blocking on some mysterious thing ?09:16
gtristanWant me to upload clean 'git format-patch' output somewhere ?09:16
*** toscalix has joined #baserock09:17
jjardonI think the one that it's conflicting is the one where you added webkitgtk, as webkitgtk is already in master; I will take a closer look later09:18
*** gary_perkins has joined #baserock09:18
gtristanjjardon, which patch adds webkit ?09:18
gtristanjjardon, eek, that looks like a problem09:19
gtristanactually; I have no idea *how* it manages to even build here if webkit is listed twice09:19
* gtristan pulls hair impatiently and tries to figure this out09:21
gtristanjjardon, that patch on gerrit is clearly not the same patch created by a regular git format-patch09:24
gtristanhowever; pushing the same patch through the gerrit bottleneck; rejects it as there are "no changes" to it09:24
gtristansigh, looks like gerrit has exhausted the lifetime of this checkout09:25
* gtristan clones a separate definitions and git am's the patches over09:25
* tiagogomes__ wonders if there is a newer version of Gerrit that we should update to09:26
pedroalvarezthere are newer versions09:27
pedroalvarezand we should update yes09:27
pedroalvarezupdating might not be easy though, but we should try09:28
gtristanwow, even with a fresh checkout09:28
gtristanok, /me discards patches in the UI and tries again09:29
* gtristan removes the magic numbers and rebase/rewords for good measure09:30
gtristanlets not gerrit think they are the *same* ones09:30
gtristan*lets not let...09:30
tiagogomes__mm, that's not the way to do it, we lose history09:31
*** dabukalam has quit IRC09:32
gtristanhah !09:33
gtristantake that gerrit !09:33
* gtristan smacks gerrit09:33
* gtristan smacks gerrit harder09:34
gtristanstory is, I upload the patch set, and gerrit mangles it, and *doesnt get the same patch*09:34
pedroalvarezit's removing enchant and it shouldn't?09:36
gtristan... is clearly not the same patch as:
* gtristan lost the 'xclip' line from bash history... had to go googling for how xclip works again09:39
gtristanno its fine09:39
gtristanpatch is totally fine09:39
pedroalvarezit is fine yes09:40
* gtristan wastes time09:40
* gtristan receives pedropalm in face09:40
gtristanthank you09:40
*** dabukalam has joined #baserock09:41
gtristanjjardon, gerrit is just dumb, the patch applies fine, if the glib-networking patch is properly applied first09:41
rjekThat sounds faintly smutty09:41
gtristanwhich gerrit just doesnt realize09:41
gtristantime to check if that gnome-terminal build works09:41
*** richard_maw has joined #baserock09:49
*** ChanServ sets mode: +v richard_maw09:49
pedroalvarez<Kinnison> When Richard turns up, can someone tell him (a) I did the hide=1 thing and (b) he owes me a yarn for it :-)09:53
*** ssam2 has joined #baserock09:54
*** ChanServ sets mode: +v ssam209:54
richard_mawpedroalvarez: you're not the only person whose IRC nick starts "^pe" who has told me that09:55
* richard_maw doesn't know why his IRC died09:55
pedroalvarezwell, my english name is "peter"09:55
rjekpedroalvarez: What is richard_maw's spanish name?09:56
* rjek always maps "Enrico" to "Eric", but it's actually rather perplexingly "Henry".09:56
petefothrjek is now known as roberto09:57
richard_mawrjek: drop the hard H on henry and it's 'enry, which can easily become enri09:57
* rjek nods09:58
*** edcragg has joined #baserock10:05
*** franred has joined #baserock10:07
gtristanCan I get a vote on this lorry: ?10:14
pedroalvarezmerged 8 minutes ago?10:14
gtristanoh thats possible :)10:14
* gtristan prepares second branch now10:15
gtristanhmmm, ok still has problems with locale, but after fixing that terminal should run10:48
jjardongtristan: I will +1 you branch so you can keep going, but I will prepare another branch to move gnutls, nettle and p11-kit to network-security. Is that ok by you?10:48
gtristanI'm pushing the branch anyway as I'm sure the fix for terminal lies at another level of the stack10:49
gtristanjjardon, you need to bring nettle along10:49
gtristanif that's what you want10:49
gtristanjjardon, still, I rather feel it's a time sink for you10:50
gtristanjjardon, with the present stratum model, I would avoid doing extra work unless those packages are particularly required by another "system" maintainer10:51
jjardonmmm, maybe you are rigth10:52
jjardonmore now that we have to rebuild webkitgtk :)10:52
radiofreehow long does webkitgtk take to build?10:53
gtristanwith the current model we basically have *some* common core system components, where organization of chunks is questionable (does every system need every chunk in 'core' and 'foundation' ?)... and then we have huge splits and possibly acceptable duplication in the higherlevel stratum10:53
gtristanjjardon, one might be inclined to just duplicate further down the stack, making it much more clear what goes into a "gnome" system vs another; also it encapsulates the build so that nothing in a separate system ever effects another; they just become entirely separate trees10:54
gtristanradiofree, about an hour give or take10:55
pedroalvarezoh, webkitgtk in gnome and gnome in the CI?10:55
*** paulwaters_ has quit IRC10:55
pedroalvarezwell, mason is very broken anyway :)10:56
gtristannautilus file browser now works in gnome, gnome-terminal is there; but it doesnt start because 'locale -a' gives garbage10:56
gtristanlooks like something has to be fixed in the glibc build10:57
* gtristan just needs to fix that bug, and ensure that gnome-initial-setup creates a sudoer at least as the primary account10:57
pedroalvarezgtristan: to have locales?10:57
gtristanthen I can default enable gdm and lock it down :)10:57
gtristanpedroalvarez, I haz no locales :)10:58
pedroalvarezI used to know what was needed to have locales working10:58
gtristanpedroalvarez, good thing you mentioned it in the WONTFIX bugzilla report so I can find it ! ;-P10:59
* gtristan shuts up10:59
pedroalvarez`localedef -v -c -i en_GB -f UTF-8 en_GB.UTF-8`10:59
pedroalvarezthat installs en_GB iirc11:00
pedroalvarezbut it will fail becasue it's missing a folder11:00
*** paulwaters_ has joined #baserock11:00
pedroalvarezwhich is "/usr/lib/locale/"11:01
pedroalvarezcreate that folder, and then run the localedef command with the locales you want11:01
gtristanmaybe that's why yeah11:01
gtristanpedroalvarez, that does it indeed11:02
gtristanpretty sure it's a hack and we're missing stuff, but it does it11:02
pedroalvareznobody else cared to create some discussion about how to include this in baserock11:02
pedroalvarezhahah, i thought it wasn't a hack at all :?11:03
gtristanwell, does it satisfy everything ?11:03
pedroalvarezhm.. what's everything11:04
gtristanpedroalvarez, I guess maybe it works for en_GB and en_US ? and maybe fr_FR ?11:04
gtristanits a good question, what is everything11:05
gtristanfirst there is the locale definitions, which I think you create that way11:05
pedroalvarezgtristan: I think it will work for anything in /usr/share/locale/11:05
gtristanI'm pretty sure the glyphs dont magically come along with that, for greek or such11:05
pedroalvarezthere are a lot of locales11:05
gtristanalthough the utf8 in the .mo files will still indicate the correct glyph codes11:06
* gtristan wonders if the terminal as is can read a .po file in greek and show something other than garbage11:06
gtristanthat tends to work in, say POSIX locale11:07
*** paulwaters_ has quit IRC11:07
gtristansigh, ok tomorrow stuff11:08
*** paulwaters_ has joined #baserock11:08
* gtristan is upset that he is not a sudoer11:08
rjekThis incident will be reported, gtristan.11:08
* gtristan knows it will11:08
* gtristan shuts down the VM before he gets noticed11:09
gtristanwell actually, with a little locale fix, I can at least enable gdm automatically11:10
gtristanand after creating the first user, one can do "su -" in the terminal to become root11:10
gtristanshould have the regular no password11:10
gtristanit's a start11:10
gtristanpedroalvarez, I'll take a closer look at that localedef line, perhaps I'll add it in system integration of something for now11:12
gtristanand see how things boot tomorrow11:12
gtristanthanks for the merge jjardon !11:13
jjardongtristan: no, thanks to you :) there are some errors though, fixing them atm11:14
gtristanjjardon, if you want to try it out; create /usr/lib/locale and run localedef as pedroalvarez indicated to create a few locales11:15
gtristanthen run systemctl start gdm and enjoy :)11:15
pedroalvarezwoho! :)11:15
gtristanoh there are still errors in the journal too :)11:15
jjardongtristan: I tihnk its because you use ybd and is less strict than morph, so no really your fault11:15
gtristanjjardon, I've been trying to order them by dependency as morph likes, but I may have got some wrong11:16
gtristandoing it by eye11:16
gtristanjjardon, oh one more thing !11:16
gtristanjjardon, there is still a problem, I think you still didnt do gnutls gitmodules like you said you would ;-)11:17
jjardongtristan: true, I will do it asap11:18
gtristangnutls had an old baserock branch from way way back when... but needs a branch off of gnutls_3_4_6 with the gitmodules11:18
ssam2i hadn't realised that YBD allows definitions to be out-of-order. We should really make a decision one way or the other and document it in
paulsher1oodssam2: why shouldn't it?11:19
ssam2i've no opinion one way or the other11:19
ssam2if we choose the 'strict' option, maybe we should have a tool that does the sorting, though11:20
ssam2i like things in strata to be sorted in a logical order. Having the tools break if things aren't sorted probably isn't the way to achieve that, though ..11:20
gtristanssam2, I think really; ultimately the chunks each go in a separate morph file and specify their direct dependencies, in which case that's a non-issue11:20
ssam2so i guess permissive is better11:20
ssam2gtristan: I think we can still have some kind of grouping type11:21
ssam2gtristan: but yeah, that would make it much less of an issue!11:21
gtristansure, but I think those can overlap too11:21
gtristanone thing that's been in the back of my mind...11:22
gtristanis it may be an interesting side project for someone with time to kill.... to generate lfs-style HTML documents based on a cluster or system target11:22 --generate-build-story systems/gnome-system-x86_64.morph --output /path/to/html11:23
paulsher1oodgtristan: i think ssam2 already did something like that using owl11:23
gtristancool :)11:23
tlsaor try scraping lfs to definitions11:23
gtristantlsa, I seriously doubt there is any plausible way to consume system integration work automatically11:24
gtristantlsa, the product really is the integration work that we do, and it varies with each test and package upgrade11:26
tlsaI suppose it depends on whether all baserock users will ever want to do is base their work off our reference definitions, or whether it would be good to have alternative options11:28
gtristanI dont see myself even trying baserock myself if I dont have a stable baseline provided by upstream baserock to expand on11:29
ssam2tlsa: it would be plausible if anyone else published integration instructions as Baserock definitions11:29
ssam2which hopefully they will some day11:29
gtristantlsa, in other words, of course every $baserock_using_product will differ slightly from the stable branch, but nobody needs to put in one years work of system integration into something from scratch if they can just; say; downstream debian instead11:30
*** gtristan has quit IRC11:42
persia I think we need a sensible derivation model before it makes sense to try to define the meaning of "stable baseline".11:46
persiaI also think we are approaching that.11:47
paulsher1oodwhat is a 'derivation model'?11:47
persiaSome sensible way to derive from definitions and consume updates to definitions I'm that derivation.11:50
* persia annoyingly has an autocorrecting keyboard11:51
*** mariaderidder has joined #baserock11:55
ssam2the thing that the migrations/ folder and the `morph diff` command have been working towards11:57
jjardonmmm, Do we have any idea what that "UnicodeEncodeError" is about?11:59
ssam2oh, it's back12:03
ssam2basically, no idea12:04
ssam2from the debugging stuff I added, it seems like the string is actually valid Unicode, but Python is determined to represent it as ASCII12:04
ssam2maybe removing the .decode() call now that the log file is opened with encoding='utf-8' will help12:05
ssam2i'll try that12:05
pedroalvarezssam2: I left some comments on that patch12:05
pedroalvarezwhich is basically what you said12:06
jjardonseems its always with cogl, if that helps, but I can not reproduce here12:06
pedroalvarezstrata/gnome.morph|gnome-devel, chunk gnome-shell references its dependency network-manager-applet before it is defined12:07
ssam2wouldn't be hard to just remove that check from Morph, if people agree12:08
ssam2wouldn't even need to be a new version of definitions12:08
pedroalvarezoh, that error has been already fixed12:08
*** gtristan has joined #baserock12:15
ssam2in morph or definitions ?12:22
pedroalvarezin definitions12:22
ssam2ah. seems it would save gtristan time to do remove the check in Morph. i might have time to do a patch12:23
paulsher1oodgtristan is using ybd, afaik12:26
ssam2but our current CI system uses Morph12:30
ssam2s/CI/automated build/12:30
*** paulwaters_ has quit IRC12:41
tiagogomes__Hi, I want to resurrect . The thing I only tested clang on x86_{64,32}. Is it ok to add it to only these archs, or should I take the risk and add it to the other archs as well. I can also just abandon the patch12:59
ssam2since nothing will be using it yet, i see no harm in adding it to all architectures13:02
* gtristan gets back from fetching a pizza13:02
gtristanit would save *us* time in general, in post-review fixes of my work, yes :)13:02
tiagogomes__ssam2 it could break the build in the other archs13:06
gtristanIs it possible to add 'system-integration' hooks to a stratum ?13:06
* gtristan notes that system-integration is not defined here:
gtristanperhaps it's not a good idea anyway, even if it 'works'13:07
ssam2tiagogomes__: oh, true13:08
ssam2 is probably more up to date13:08
ssam2didn't realise that info was in 2 places13:08
ssam2gtristan: it's not possible to do that currently. Might make sense in theory13:09
gtristanssam2, I would avoid it as it might tangle things up further in the case of a refactor13:09
gtristanssam2, right now chunks could easily be split out and leave nothing else but 'grouping' to be desired from the stratum13:10
gtristanAlso, I blame evolution for breaking the ML threads13:10
gtristan <--- that was not a reply to a message in the thread13:10
gtristanbut click on link in evo, anyway, sorry for that :)13:10
lostducktiagogomes__, Given the comments on that change would it be better to put clang in its own stratum I wonder?13:14
tiagogomes__lostduck that is the idea, but that stratum needs to be included somewhere13:33
lostduckokay i'm a bit confused, 916 shows 3 new chunks being added to devtools?13:35
tiagogomes__lostduck I haven't submitted an updated patch yet13:36
lostduckahh okay13:37
pedroalvarezheheh looks like Mason can't cope with webkit14:45
pedroalvarezc++: internal compiler error: Killed (program cc1plus)14:45
pedroalvarezotoh, it doesn't have the Unicode error anymore14:46
ssam2actually, morph's autodetected value for max-jobs is suspect, isn't it?14:47
SotKI thought we set max-jobs: 1 to stop that kind of madness with webkit?14:47
ssam2that was qtwebkit14:47
ssam2it would be nice to avoid that because it makes the build way slower14:47
pedroalvarezI removed that max-jobs=1 yesterday14:47
ssam2maybe if we fixed Morph's default max-jobs to the value Paul suggested (I forget what it was, but it was lower than the current one) it would help14:48
pedroalvarezbut this is another webkit14:48
pedroalvarezssam2: it was the lowest of "number of cores" or 914:48
pedroalvarezI hope i can english14:48
richard_mawssam2: the gist of it is that 1½ × CPUs isn't such a useful metric as it once was since HDD speeds have improved and since we're constructing the staging area just before the build, much is already cached in RAM14:49
pedroalvarezwe are doing now "cores" * 1.514:49
jmacs;/win 3114:49
paulsher1oodwebkit builds break with cores > 1 unless the machine has plenty of memory (>2GB i think)14:49
pedroalvarezit only has 2GB14:50
pedroalvarezand 2 cores14:50
ssam2it definitely wouldn't work with -j 2 or -j 3 or -j 4?14:50
ssam2I guess at the moment it's -j 314:50
ssam2we really need a way to override this sort of thing in the build machine rather than in the definitions...14:50
paulsher1oodj > cores is a waste of time. it will slow it down14:50
richard_mawssam2: morph had --max-jobs=, but that would change it for every build14:51
ssam2sounds like it makes sense to set max-jobs=2 in /etc/morph.conf on that mason, at least14:51
ssam2I will do and see if it fixes the webkit build14:51
radiofreeYou could add some swap14:52
ssam2oh, swap would make sense14:53
ssam2if -j 2 doesn't fix it I'll try doing that14:53
ssam2you even suggested this before and we did it for a customer and it fixed things14:53
* lostduck is a little surprised mason only has 2GB14:55
jjardoncan we add more RAM to mason?14:58
ssam2resizing existing OpenStack instances is tricky14:59
ssam2it's been super broken in the past14:59
* pedroalvarez can redeploy if we agree14:59
ssam2ah, cool (i don't have time)14:59
jjardonplease! we all benefit from the cache generated by mason15:00
jjardoncan you add some more cores as well? :)15:02
ssam2if we add more cores then webkit will break again :-)15:04
pedroalvarezlet's see how this build goes first15:30
pedroalvarez12% built :)15:30
*** paulwaters_ has joined #baserock16:00
pedroalvarezjonathanmaw: Hi, I'm currently trying to figure out why we are patching here:
pedroalvarezand also wondering if we could submit it upstream16:26
jonathanmawpedroalvarez: that's because ilm_client and ilm_control have a mutual dependency16:26
jonathanmawand that change is how you make it link against one while testing the other16:27
jonathanmawi.e. because wayland-ivi-extension's code at this specific point in time is a bit of a mess, and version hell prevents us using anything better16:29
jonathanmawupstreaming is unlikely, because wayland-ivi-extension's upstream is already way ahead of this version, but the GDP is stuck with is because of version hell (specifically, there was an interface change without backwards compatibility, which breaks the GDP HMI)16:31
pedroalvarezthanks! I think I'll include that patch in the branch we are using, and remove this from the  chunk16:43
* pedroalvarez is going to research about gocd this week16:44
richard_mawwhat's gocd?16:45
richard_mawor haven't you got that far yet ;¬)16:45
rjekCD, in Go.16:45
richard_mawrjek: ptraces the parent process does it?16:46
* richard_maw has occasionally needed a chdir program analogous to chroot, which chdirs then exec's the rest of the argv16:47
paulsher1oodrichard_maw: http://www.go.cd16:48
pedroalvarezI don't know much about it yet, but looks interesting16:49
richard_mawso… It's a Java/JRuby on Rails project, which shares its name with a programming language also used for web stuff16:50
pedroalvarezheh, that's the first thing you get when looking at it, yes16:53
pedroalvarezbut it might be interesting and it may make things easier16:53
pedroalvarezHm, it doesn't have support for elastic agents yet, but they want to include that on the next release16:56
paulsher1oodoh? i may have been misinformed, then16:58
pedroalvarezgithub issues may be wrong though, I'll ask them16:59
* richard_maw wonders how pipelines are configured16:59
pedroalvarezthey also want to add suport for "Pipeline configuration from source control" on the next release17:01
richard_mawpedroalvarez: that's nice, but what matters is that there's an API for dynamically configuring them17:02
richard_mawhm, Java-based agent required on clients17:03
*** paulwaters_ has quit IRC17:04
*** paulwaters_ has joined #baserock17:04
pedroalvarezI expected that one17:05
richard_mawthat was one of the things that caused us problems when we tried to do things with Jenkins17:06
pedroalvarezregarding API for reconfiguring pipelines I don't think there is17:07
pedroalvarezI need to research about the possibilities of this but bearing in mind our needs for ciat17:07
richard_mawthere's the web UI17:07
pedroalvarezgood catch17:10
richard_mawnot sure whether the API documentation is unclear, or whether they don't actually use it, since it doesn't seem to fit their data model that you have to provide a version control as the input, since it talks about being able to trigger builds from other pipelines or artifacts17:12
pedroalvarezwell, mason failed to build webkit:
pedroalvarezI will put more ram on it then17:23
richard_mawhm, that pipeline creation API is deprecated, but it was historically used internally17:25
*** franred has quit IRC17:28
pedroalvarezI wonder where did you see that17:28
richard_mawhold on, I'll get you a url17:28
pedroalvarezI also wonder why we need to dynamically configuring them17:28
pedroalvarezI wokred on CIAT, but didn't get some concepts and needs17:29
*** mariaderidder has quit IRC17:29
richard_mawpedroalvarez: the config for which candidate branches are created and what feeds into them is defined in a git repository17:31
*** nowster has quit IRC17:31
richard_mawso potentially one pipeline gocd would run would be to trigger on changes to that, which reconfigured gocd'd build and test pipelines17:32
richard_mawthough ideally I'd like it if whatever we use could be configured to run individual builds, so we get a pipeline step for each build step17:33
richard_mawwhich if it supports elastic pipelines gives a free distbuild17:34
richard_mawwell, not free, you've got to split out individual builds from the build tool17:34
* richard_maw has been looking at what the components required for a Baserock-style are, and implementing the individual bit bottom-up17:38
richard_mawso far I have which was based on an article I wrote about locking, and, which I need to tidy up further17:38
*** paulwaters_ has quit IRC17:40
ssam2for sandboxing, feel free to do stuff to sandboxlib17:40
richard_mawyep, I was intending to17:40
richard_mawnot there yet though17:41
ssam2then we just need a library for parsing definitions!17:41
richard_mawhah, there's still a ways to go before that17:41
richard_mawI'd like to make cloning the git repository as fast as possible, which means using --shared, which makes it have a .git/objects/info/alternates file saying where the object store actually is17:42
richard_mawwhich of course isn't inside the chroot17:42
richard_mawso there'll also be something to hard-link or tell you how to bind-mount the object stores in to make a shared repository chroot safe17:43
* richard_maw is hoping to provide useful individual projects by working from the bottom up17:44
richard_mawthough since I've been working on it in my own time, python-gitcache is rather other-thought and over-engineered17:45
richard_mawanyway, this relates to gocd, since I intend that a single-build would be its own component making use of the git cache, artifact cache, sandboxlib and chroot-safe-git-repo, as its own command-line application17:46
ssam2sounds like a good plan17:48
* richard_maw thinks `python -m gitcache clone-temporary $REPO_URL -x ${ARGV[@]}` is fun, since it runs a command in a temporary clone of the git repository and holds locks to prevent anyone removing the backing store from the cache while the repo is in use17:54
richard_mawpart of tidying that up will involve documenting whil this is necessary17:55
*** jonathanmaw has quit IRC18:00
ssam2smarter caching will be really good18:00
* richard_maw wants to do away with the theoretical race where at parse time a ref is resolved in the repository, and another concurrent build starts which also fetches, including a delete and gc of the object you used, so later at build time this object you needed no longer exists18:02
pedroalvarezerr.. is this because.. python3 or something like that?18:03
*** bashrc has quit IRC18:03
richard_mawcurrently python-gitcache does that by taking a shared lock on the repository and running with autogc turned off when it doesn't have an exclusive lock, but I'd like to have it create a ref namespace for that to have finer-grained ref locking18:04
richard_mawpedroalvarez: aye, looks like it18:04
* richard_maw packs up to go to a comedy night18:04
* SotK assumes the backscroll means we are throwing away the current CIAT implementation?18:12
pedroalvareznot just yet18:18
toscalix is not running?18:29
toscalixI cannot connect18:29
*** ssam2 has quit IRC18:33
pedroalvareztoscalix: it is not running, and I don't know who turned it off18:38
toscalixpedroalvarez: I will find out18:39
pedroalvarezI sleep better since it's off tbh18:39
*** edcragg has quit IRC18:56
*** locallycompact has joined #baserock18:59
*** gtristan has quit IRC19:11
*** toscalix has quit IRC19:26
lostduckpedroalvarez, lol19:50
*** edcragg has joined #baserock20:12
*** edcragg has quit IRC20:55
*** cosm has joined #baserock21:26
persiaOn system testing when a client is required on the target to run the tests: create a test worker that has the client and uses Ansible to coordinate the testing of the target system.21:42
persiaThe test worker needs the client to report to the overall test dispatch system, but the target system requirements are controlled entirely by the test itself.21:42
*** edcragg has joined #baserock23:23
*** edcragg has quit IRC23:56

Generated by 2.14.0 by Marius Gedminas - find it at!