IRC logs for #baserock for Thursday, 2015-01-08

*** zoli_ [~zoli_@linaro/zoli] has quit [Remote host closed the connection]01:00
*** zoli_ [] has joined #baserock01:00
*** zoli_ [] has quit [Changing host]01:00
*** zoli_ [~zoli_@linaro/zoli] has joined #baserock01:00
*** zoli__ [] has joined #baserock08:00
*** zoli_ [~zoli_@linaro/zoli] has quit [Read error: Connection reset by peer]08:00
*** wdutch [] has joined #baserock08:14
*** zoli__ [] has quit [Remote host closed the connection]08:22
*** zoli_ [] has joined #baserock08:42
*** zoli_ [] has quit [Changing host]08:42
*** zoli_ [~zoli_@linaro/zoli] has joined #baserock08:42
*** zoli_ [~zoli_@linaro/zoli] has quit [Remote host closed the connection]08:43
*** zoli_ [] has joined #baserock08:43
*** zoli_ [] has quit [Changing host]08:43
*** zoli_ [~zoli_@linaro/zoli] has joined #baserock08:43
*** mariaderidder [] has joined #baserock08:43
*** bashrc [] has joined #baserock09:01
perrylpedroalvarez: i'm having trouble connecting to my firehose machine on DC, is there an issue with VMs or has it lost its floating IP?09:04
persiaI believe floating IPs changed, but I may be misinterpreting.  Did it have a logical name?09:05
perrylIIRC, it was just firehose09:07
perrylssam2 was the one who set it up09:08
* perryl isn't sure09:11
perrylit may be worth waiting for ssam2 to arrive and double checking with him09:11
persiaHeh, possibly.  Unless you want to spend time querying DNS to see if you can figure it out first (dig can be fun).09:12
pedroalvarezperryl: sorry about that, I'll give you now the new IP09:16
*** tiagogomes [] has joined #baserock09:16
*** franred [] has joined #baserock09:18
perrylsorted, thanks pedroalvarez!09:18
* pedroalvarez checks other urls09:19
*** jonathanmaw [] has joined #baserock09:25
*** ssam2 [] has joined #baserock09:35
Mode #baserock +v ssam2 by ChanServ09:35
pedroalvarezjust discovered that lorry controller is failing to lorry because it cannot resolve (e.g.) github.com09:45
pedroalvarezI see that the nameserver specified in its resolv.conf differs from other machines pluged in the same network09:46
pedroalvarezI don't know why this is happening09:51
persiaI'd suggest fixing for now, and then troubleshooting in a test environment.09:51
pedroalvarezbut, won't the resolv.conf be regenerated?09:52
persiaGiven the recent migration, I presume there to be data available to allow easy generation of the test environment, although I may be mistaken.09:52
persiaDepends on the system configuration.  What environment?  How is resolv.conf being populated now?09:52
pedroalvarezpersia: you are right, easy to prepare a test environment09:52
persiaAh, good.  So we can have our cake and eat it too :)09:53
* persia hugs virtualisation09:53
pedroalvarezsystemd-resolved is what creates resolv.conf09:55
persiaAnd how is that configured?  How does this configuration differ from instances that don't have the issue?09:56
tiagogomeshey ssam2, do you remember vaguely when you disabled the C++ stuff for binutils?09:58
pedroalvarezok, differences between a sane system and In the sane one /etc/resolv.conf is a symlinkt to /etc/run/systemd/resolve/resolv.conf. And in g.b.o this is a static file.09:58
ssam2when? will have been about 2 years ago09:58
ssam2if I didn't leave a comment in the morphology or the commit message, shoot me09:59
ssam2looking at the morphology, I don't see any C++ stuff being disabled, what do you mean?10:00
pedroalvarezpersia: looks like if you modify /etc/resolv.conf with a static file, systemd-resolved won't try to replace it with the symlink to the autogenerated resolv.conf10:00
persiapedroalvarez: Aha, in that case, make it a symlink, and it ought be good.10:00
pedroalvarezyes. I've also double checked that the file to be symlinked is sane10:01
persiaThat's best behaviour for systemd-resolved: it lets people with specialised needs statically configure the file in the traditional way.10:01
ssam2i'm wondering what the best version of Morph would be to build baserock-14.4610:02
persiaHow about the version used in 14.46?10:02
ssam2would make sense, but it's not tagged10:02
* persia queries10:02
ssam2I can dig and find the sha1, but the fact that it's not tagged suggests that it's not so simple10:03
tiagogomesssam2, you disabled it in and configure e72775a. I was just wondering if you hacked configure manually or auto-generated it from configure.in10:03
tiagogomesssam2, the reason that I am asking is that I tried to regenerate it but I got: error: Please use exactly Autoconf 2.64 instead of 2.69. (no retro-compatibility ¬ ¬)10:04
ssam2tiagogomes: I will have generated it using autoconf10:04
ssam2we have updated autoconf since then, probably10:04
ssam2persia: definitions is tagged, so I can use the version of Morph listed in the baserock-14.46 tag of definitions10:05
tiagogomesso or either I downgrade autoconf or I try to manually edit configure10:05
ssam2just seems strange not to be tagging releases of Morph10:05
ssam2tiagogomes: I'd advise against trying to edit configure10:05
ssam2at least if you generated it with autoconf you can blame problems that you have on autoconf, not yourself10:05
Kinnisonssam2: we haven't tagged morph for some time (at least not as baserock releases)10:06
persiassam2: As I've said before, I think morph ought have a release model, even though I think definitions should not.  Getting there is confusing.10:06
persiaAnyway, 384dc8d2eb9222746085f0274c625eb163eda88a should work10:06
franredssam2, I though we aim to use latest version of morph always... but this means that we need backward compatibility, though10:07
ssam2tiagogomes: I see what you mean now10:07
ssam2tiagogomes: does latest binutils still use autoconf 2.64? that's pretty unfortunate if so10:08
tiagogomesssam2, it seems easier to change configure, than downgrade autoconf. This will would be temporary because when we have a bootstrap c++ compiler, I won't need to disable the c++ stuff in bintutils10:09
ssam2your funeral :)10:09
persiaCan the autoconf stuff be upgraded to 2.69?10:10
persiaOr does upstream have another target version they might like?10:10
persiaDo any of the distros have patches that build with alternate autoconf?10:11
*** locallycompact [] has joined #baserock10:12
ssam2tiagogomes: remember you don't have to downgrade autoconf in Baserock10:13
ssam2just install autoconf 2.64 from a release tarball, run it, and commit the result10:13
ssam2the whole point of using a tarball is that you don't need autoconf (or any other autotool) at build time, because 'configure' exists and is committed to git10:14
tiagogomeson AC_PREREQ(2.64)10:14
tiagogomesMaybe I could try to change AC_PREREQ(2.69), but that could break thinks in other ways10:14
ssam2I believe you can parellel-install different versions of autoconf on a machine, by setting a configure option when you configure autoconf10:15
tiagogomesthat's a good idea ssam210:15
KinnisonI think that'd imply updating devel systems to have both first10:15
Kinnisonsince early stages are built using the enclosing devel system10:16
ssam2Kinnison: not at all, because everything is built from tarballs at that point10:16
ratmice__haven't found the patch (e72775a) but, i believe you should be able to just --disable-dirname for any directory that uses c++ in the binutils root?10:16
persiassam2: Yes, but unless you actively dirty the cache, you can't cause to do that sort of hack.10:16
Kinnisonssam2: Oh cool10:17
franredtiagogomes, why don't try to update the autoconf prereq to 2.69 and fix errors and submit the change to upstream? (talking always if you have time)10:17
ssam2persia: i don't understand how has anything to do with this10:17
ssam2ratmice__: that's interesting!10:17
persiassam2: Ah, right, you're just changing source so that you don't need the tools to be present.  I was thinking about the case where the tools were being used as part of a live build.10:18
tiagogomesratmice__, e72775acb5326e948d2542622afb9bed4fd67c15 in git://
ratmice__in general gcc/binutils use a specific version of autoconf because if everyone used whatever version of autoconf there would be endless churn/large diffs whenever someone changes it, and it keeps everyone running the same version, just be glad it uses a release version :D at one time it did not :)10:19
ssam2ratmice__: is --disable-dirname a binutils ./configure flag? I don't see it (but I have an old clone of binutils)10:19
tiagogomesI was wondering the same, I couldn't find it either10:20
tiagogomesNot even in the newest repo10:21
ratmice__s/dirname/gdb/ or s/dirname/ld/, (or in this case likely --disable-gold) i forget where it is, I didn't know about it either until they switched to git10:23
* ratmice__ hopes he is remembering correctly :)10:23
ssam2oh, so e.g. if only 'gold' uses C++ we can just --disable-gold ?10:23
ssam2that's worth investigating10:23
ssam2as long as we don't end up with no linker or something :)10:24
*** locallycompact [] has quit [Ping timeout: 240 seconds]10:24
tiagogomesssam2 that wouldn't be good10:24
tiagogomesah, they use cc for c++ files. I was looking for cpp, c++, hpp10:25
ratmice__ahh, so i see this is actually in ld/configure disabling the c++ test cases, so it may not in fact help at all10:28
* Kinnison says rude words about nested autoconf10:29
* richard_maw dislikes it because the result is usually that the nested ./configure scripts are run when you use `make`, rather than when you run the top-level ./configure10:30
* ratmice__ dislikes it because --help=recursive outputs a bunch of junk over and over10:34
Kinnisonin general, 'recursive $BUILD_FOO' is a bad plan IMO10:36
persiaMost of the reasons people do it is because they don't understand the loop instructions in their build tool.10:37
persiaThen again, most of the reasons there are so many build tools is because people don't understand their build tool.10:37
paulsherwoodthat certainly explains why i started doing that foyr ybd :)10:37
ratmice__I'll admit i like the concept behind binutils build tree specifically, where you have a top-level directory it just builds everything underneath it, so you can add and remove modular components10:37
ratmice__but ideally the top-level directory should be stupid of the components underneath it, and etc, etc...10:38
persiaFor make, I'd prefer to have a text file named "components" containing a list of the components I wanted to enable.10:38
persiaI'm not as familiar with other build tools, but I presume they can support a similar construct.10:38
*** locallycompact [] has joined #baserock10:39
franredI need a repo to compile another repo, what is the neatest way to add it without having to copy the repo into the system and them copy it to the build directory for the second repository to compile this latest one?10:41
franreds/to compile/for compiling/10:42
ssam2not sure there is a neat way ... how big is it?10:50
franredssam2, I need the apache portable runtime to build apache httpd10:51
franrednot sure how big it is :S10:51
persiaIs there a set of artifacts from the apr that can be considered a "chunk" to be a build-dependency of apache?10:52
persiaThat's the most semantically compatible way to do it, but it may be messier than that.10:52
franredpersia, Im afraid apache-httpd needs apr source code10:55
persiaAll of it?10:56
persiaYou could make a chunk that contains a snapshot of a repo content, but I'd be surprised if that was the best way.10:56
persiaHow do the distros do it?10:57
persiaAt least Debian seems to build apr separately, and then build apache against it10:59
franredpersia, thanks, I will give a go :)11:00
*** jonathanmaw [] has quit [Remote host closed the connection]11:00
*** jonathanmaw [] has joined #baserock11:01
persiaIf anyone knows how to get similar information for how Fedora builds things, it would be useful to look at both techniques (I've forgotten most of the right Fedora URLs)11:01
persiaBut I can't find apache2 there: it may just be me.11:13
ssam2I think it's called 'httpd' in Fedora11:13
persiaAlso not found in that cgit search interface11:14
ssam2ah, not sure then11:14
radiofree ?11:15
persiaradiofree: Thanks!11:16
persiafranred: ^^11:16
jjardonHi, current NetworkManager requires readline6 but baserock only have readline5 for (I think) license reasons. What would be the best way to proceed here?11:16
persiaDo we have any information about which systems have license restrictions?11:17
jjardon(readline is in core because is a dependency of cpython)11:17
radiofreegenivi system, no gpl311:17
persiaI suspect the best way is to fork the strata, so some systems get one, and others get the other.11:17
persiaThen we split core into core-gpl2 and core-gp3?11:17
radiofreethere's no network manager in genivi systems right?11:18
jjardonpersia: that would mean fork everything on top of core ...11:18
radiofreecouldn't you just put a newever version of readline in the network manager strata?11:18
jjardonradiofree: no, they use conman11:18
radiofreecan you --disable-cli or something?11:18
radiofree(in NM)11:18
persiajjardon: How much commonality is there between the gpl2-only and gpl3-ok systems?  I suspect that after a short time, we don't need to fork anymore.11:18
jjardonradiofree: no, seems its a hard dependency. 11:21
pedroalvarezradiofree's suggestion of adding readline in network manager strata sounds good to me11:22
jjardonpersia: all the current systems in definitions master are gpl2-only, I think11:22
persiaI don't like the idea of having two different readlines on network-manager enabled systems.11:22
persiaI'd rather see python done with the same readline.11:22
persiajjardon: But are they all required to be so?11:22
jjardonyeah, me neither, but if its the only solution Id use it11:22
radiofreeit's better than doubling the number of strata11:22
jjardonpersia: no, only genivi AFAIK11:23
persiaI think that only the GENIVI systems are required to be gpl2-only, and suspect that we only have to fork a few strata to achieve that.11:23
pedroalvarezpersia: I also liked your suggestion of a core-gpl3, but then all the artifacts above core will differ11:23
persiapedroalvarez: Yes, but how many artifacts do we need in both GENIVI and non-GENIVI systems?11:23
jjardonthats why I have to put coreutils is in a differect stratum, dor example11:23
radiofreemultimedia, wayland, graphics, input, mesa, weston (audio? connectivity?)11:24
pedroalvarezshould the genivi definitions be a fork of definitions? 11:25
persiaThat might be easier, and helps ensure the tooling can usefully support multiple definitions repos.11:26
persiaSince we anticipate every user having a separate definitions repo, this is a key feature.11:26
radiofreepedroalvarez: that sounds like a good idea11:26
ssam2pedroalvarez: that's a really good idea !11:26
persiaThe only detriment is that if work is done that can benefit both definitions, it may need to be duplicated.11:27
ssam2merging improvements from baserock/definitions.git into downstream forks of definitions.git is a problem we need to solve11:28
jjardonId like that, because I personally do not care about genivi systems, but that would mean a lot of duplicated strata everywhere11:28
persiaThe other direction as well, to ease sharing updates upstream.11:28
ssam2I'm on a project currently where one of the things we need to do is solve the former problem (at least, make things easier). The latter would be nice too.11:29
ssam2currently `git merge baserock/master` presents a comically unsolvable merge conflict :)11:29
ssam2(assuming `baserock` is a remote that points to git://
*** jonathanmaw [] has quit [Remote host closed the connection]11:30
paulsherwoodssam2: if i may ask, how many commits are in the fork branch?11:31
*** jonathanmaw [] has joined #baserock11:32
ssam2paulsherwood: I've no idea, but 9+ months of work11:32
ssam2so probably lots11:32
paulsherwoodok, that's enough of an answer :)11:32
jjardonok, I will go for the "readline6 in NetworkManager stratum" Thanks for the feedback!11:34
persiaradiofree: pedroalvarez: Do you feel comfortable forking defintions from before that patch lands to maintain GENIVI?11:37
pedroalvarezwell, maintaining a fork, will be a bit more effort, but I think is something that we need to face11:39
persiaOne of the areas of complication is CI: how do we want to handle CI for this?11:41
persiaI know of another project that is considering using Baserock for musl-based systems that wants to be upstream: do we think they should have yet another fork, or does that present different problems to definitions?11:41
persia(the same applies for folk who might want to use ulibc, but I don't have any special knowledge there)11:42
ssam2persia: different question, I think (although it might have the same answer, for now)11:43
pedroalvarezI'm now thinking that if the genivi systems were in a fork, then I wouldn't be able to use the current infra to build them11:43
ssam2pedroalvarez: why?11:44
ssam2morph checkout baserock:baserock/genivi master11:44
pedroalvarezssam2: yeah, but my current infra has distbuild networks11:44
pedroalvarezso I have to reconfigure them to build from genivi, and then switch them back to continue doing CI?11:45
persiassam2: Yes, changing readline vs. the C library is different, but both are very deep changes, so have the same class of issues.11:45
persiapedroalvarez: Why?11:45
ssam2pedroalvarez: baserock:baserock/definitions isn't hardcoded anywhere :)11:45
pedroalvarezyeah, that's true and good news to me11:46
*** sambishop [] has quit [Ping timeout: 264 seconds]11:51
tiagogomesI am having build failures when compiling binutils 2.25. A comment found in the web says that binutils needs flex to compile. But flex is not on build essential. Linux From Scratch seems to not need flex to compile binutils. Anyone has any insight about this?11:54
persiaHow does LFS compile binutils?11:55
persiaThe current LFS book says to use binutils 2.17, and that "versions greater than 2.24 are not recommended"), so it may not be useful to rely on LFS as a source for binutils 2.25 integration.11:58
tiagogomespersia the online documentation has intructios for 2.2511:59
tiagogomesbut flex on LFS is built after the final build of binutils12:00
tiagogomesin Baserock happens the same12:00
persiaYou're better at finding LFS docs than I :)12:00
persiaMaybe there is an alternate target of some sort that doesn't need flex?12:01
tiagogomesI am getting this error:
persiaMaybe it is special configure arguments, or similar?12:01
persiaHow are sysinfo.c and syslex_wrap.c being generated?12:02
tiagogomesNeither baserock or LFS seems to use a configure option to disable flex requirement12:03
bwhI don't see how you could avoid a dependency on flex12:03
bwhbinutils does kind of need to parse stuff12:03
ssam2tiagogomes: probably the generated flex files are present in the tarball12:04
persiaThat seems the easiest bootstrap workaround.12:04
ssam2but the makefile will try to rebuild them if the timestamps of the generated files are older than the timestamps of the flex input files12:04
ssam2I've no idea why you get "multiple definition of `main'" though!12:05
persiaAnd git doesn't preserve timestamps, so lorried tarballs are annoying.12:05
ssam2persia: morph resets all timestamps when checking out a source tree, to work around that problem12:05
*** sambishop [] has joined #baserock12:06
persiassam2: How does morph resetting everything work around git resetting everything?12:06
ssam2it does so after running `git checkout`12:06
ssam2I assume the code was added because git resetting everything wasn't working right. I've never checked this assumption.12:06
persiaYes, but 1) how is that useful, and 2) why does this work around the issue?12:07
persiaUnless morph has instructions to only update timestamps for generated files, it causes all the same problems.12:07
ssam2no, if the timestamps are all the same then nothing is newer than anything else12:07
ssam2it's always worked fine for me, so I've never dug into it to work out why it works12:07
persiaBut git sets all the timestamps to be the same, doesn't it?12:07
persiaI think it's a no-op that coincidentally worked for some sources because of file ordering.12:08
ssam2that wouldn't surprise me. but I'm not going to start fixing something that isn't broken until I see it break :)12:08
persiaIf not, it's deep magic, and morph probably shouldn't do it (e.g. setting timestamps of all lex files, then all C files, etc.)12:09
persiatiagogomes: Is your failure in a morph-controlled biuld?12:09
tiagogomespersia correct12:09
ssam2but seems totally unrelated to timestamps12:10
ssam2or are we talking about different failures ?12:10
persiaI agree, which is why I asked how they were generated.12:11
persiaIn the event that this is broken because morph is being silly, we should fix that.  If not, we can let morph be silly longer.12:11
tiagogomessyslex.c and sysinfo.c are present in the tarball, so flex and bison shouldn't be needed12:11
persiaDo they both have main()?12:13
tiagogomescontext: this happens building the final version of binutils, (but still using and older binutils on stage1 and stage2)12:13
tiagogomesbut stage1-bintutils built fine12:14
persiaDo we have an earlier-stage flex at this point?12:14
pdarHiya, Im having trouble starting a baserock vm in virtualbox after following the instruction here:
pdarI get the following error:
tiagogomesmaybe building binutils 2.25 with 2.25 will fix the problem, but I don't know how12:16
pedroalvarezpdar: have you tried `modprobe vboxnetflt`?12:16
pdarI did, and the suggestions here:
pdarand it is still kaput :(12:17
franredpdar, have a look at your ipaddr or ifconfig and check that your ethernet name is eth0 or another12:18
tiagogomespdar or be lazy and change your network connection to bridge or NAT12:20
jmacsWe can't, host-only networking has to work for this project12:20
pdarfranred: it is called em112:21
persiaNot en1?12:21
pdarpersia: yep em112:22
franredpdar, replace eth0 by em112:22
tiagogomespdar, have you tried file -> preferences -> network -> and then add a host only network12:22
persiaInteresting.  I've not seen that before.12:22
franredpdar, in the previous workaround12:22
pdarfranred: I have tried that too and nada12:23
franredummm odd12:23
kejiahuHi, any suggestion on why 'Failed to start Remount Root and Kernel File Systems.' happens during boot?12:29
Zara_Is there a simpler gem we could use as the basis for the rubygems import tool tutorial? I think it'd help to have a quickstart tutorial- so people can get to grips with the tool itself, and then a different, more detailed tutorial for rails (or another difficult gem). Or a 'common errors' section just for the importer!12:30
Zara_I feel like we'll be adding things to this tutorial indefinitely, otherwise, because many gems need slightly different approaches.12:33
Zara_It might even be worth it to have a table of gems and their respective approaches, though that's probably overkill, and could get out of date fast. :P12:33
persiaCan the approaches be generalised to help make it easier to use?12:34
persiaI also seem to remember hearing some complaints about the number of strata produced: are there ways we can change things to be easier to use that also simplify the documentation?12:35
rdale_i have got quite a lot of gems built in order to get one of codethink's internal apps working - we could have a look at exactly what build commands i've needed to use and see if it could be automated more12:35
persiaThat sounds like a good strategy12:35
Zara_persia: some can, but I think most things that can be completely generalisable can be automated, which then means you don't need the tutorial for the manual steps.12:35
SotKI want to send some patches, but don't have access to the machine which has my ssh key on it. Would people prefer I just don't push the branches until tomorrow, or for me to put them up on github?12:35
persiaAlthough I'm a fan of documentation, I think I prefer when the documentation is short.12:36
persiaSotK: There's been a few patchsets sent without branch references, and others sent with github references, so you have precedent for both.12:36
pedroalvarezjjardon: OOI, where you looking at the upgrade of systemd to 218? IIRC it wasn't a trivial upgrade?12:37
Zara_persia: I like a lot of documentation, but not too much on any single page. :) I think that when it gets too long, it's a sign that either things are too complicated, or large sections are irrelevant to any single user.12:38
persiaOr that the author has become distracted by writing the authoritative book, rather than providing tutorial and reference information for the code.12:38
jjardonpedroalvarez: it failed to compile with a weird compilation error12:39
Zara_heh, I think the tutorial in this case is straightforward and well-written, though I came across some horrors on my Philosophy degree...12:40
petefothor that the author is unclear about *why* they are writing what they are writing and *who* it is intended for12:40
persiapetefoth: Better point.  Sometimes it is correct to write the authoritative book.12:41
petefothpersia: indeed. But probably not today :)12:41
pedroalvarezjjardon: thanks for the info.12:41
* paulsherwood finds 'no bootable medium found' on latest build-system image from d.b.o?12:52
paulsherwoodpdar: is this similar to your problem?12:52
ssam2kejiahu: does the journal have any useful info? do you have a weird /etc/fstab? do you have weird kernel commandline arguments that might effect the root disk ?12:52
ssam2re. the import tool tutorial, it's certainly not ideal12:53
pedroalvarezpaulsherwood: :(12:53
ssam2it tries to be both an introduction plus an in-depth 'all you need to know for rubygems' guide12:53
paulsherwoodoh, actually. it's probably me being an idiot12:53
ssam2if we had time to rewrite it into two separate documents, I'd be happy12:53
paulsherwoodpedroalvarez: pleae hold, i think it's me12:54
jjardonmmm, my system is trying to build curl, is not working?12:56
Zara_ssam2: Yeah, I'd prefer two separate documents; if someone can give me an example gem that can be used for an simple guide then I could make (or at least start!) a quickstart guide today.12:56
Zara_otherwise I can look around for an example gem, but I suspect that'd take longer. 12:57
jjardon(I didnt change anything in core or lower strata)12:57
persiaWhat cache key did your curl attempt generate?12:58
persiaComparing that to the recent value from a mason log might give some guidance.12:58
jjardonpersia: how I check that?12:59
persiaYours should have appeared in your morph log just before the build content (in a failure to fetch message), I think.13:00
persiaI thought would work, but it doesn't.13:01
persiaMaybe someone else knows the right hostname (as I'd like to know it as well)13:01
kejiahu+ssam2, seems caused by /etc/fstab, will check further, thank you.13:03
radiofreekejiahu: oooh13:03
radiofreeif you're using the jetson image13:04
radiofreei think you'll need to delete /etc/fstab13:04
radiofreeit's btrfs specific13:04
kejiahuradiofree, yes, that's what I found follow sam's suggestion..13:05
pedroalvarezjjardon: I was debugging it before reading your comment about cache.b.o13:12
pedroalvarezjjardon: it looks ok now13:13
jjardonpedroalvarez: yep, morph stop building and fetching from the cache automatically :), thanks!13:32
*** simonh_ [] has joined #baserock13:40
*** mauricemoss_ [] has quit [Read error: Connection reset by peer]13:40
simonh_ is now known as mauricemoss_13:55
straycatssam2, I think it may be worth making the comparison carried out in lorryset more flexible, one other reason I've had problems adjusting lorries is because the importer is comparing the strings using ==, rather than a case-insensitive comparison that also treats - and _ as equivalents13:59
straycatyeah let's do this14:07
ssam2Zara: not sure about what Gem would be best, I picked Rails because it had a suitable number of complications for an in-depth tutorial :) (and because it's a well-known thing)14:51
ssam2straycat: i'm not sure what you mean14:52
ssam2which strings?14:52
straycatssam2, I think the importer should optionally provide a comparison function for comparing package names14:52
*** franred [] has quit [Ping timeout: 255 seconds]14:53
straycatwhen calling enable_importer14:53
ssam2yes, makes sense (after in-person discussion)14:56
tiagogomesI am increasingly confused with my binutils problem. The Internet says in more than one place that the fix is to install flex; but both baserock and LFS don't install flex as part of the toolchain14:56
ssam2is the problem that the generated flex files in the tarball are out of date?14:56
ssam2the .c files generated by flex, rather14:57
radiofreedoes the latest release of baserock have the systemd upgrade?14:58
ssam2no, but there's a release being done tomorrow that will have it14:59
ssam2and I believe the release is ready already, if you need it for something ..14:59
tiagogomesssam2, looking at the modification times, it seems so. I'll try to see how binutils generates them15:00
radiofreeok, thanks ssam215:01
ssam2tiagogomes: it's really strange that building from the binutils release tarball doesn't work15:01
ssam2it suggests either we're doing something wrong, or the binutils release tarball is broken15:01
ssam2you can make your own tarball with `make dist` and try that, perhaps15:01
Zara_ssam2: I think we should definitely keep rails as an in-depth tutorial, but possibly use a gem with fewer complications as an intro to the rubygems tool. Otherwise, someone completely new to the tool and new to baserock won't know if they've made a mistake, or if the instructions they're following aren't sufficient. That leads to a lot of repeating steps, starting from scratch, etc.15:03
tiagogomesssam2 I will try that, it seems that the release is broken, because stage1-binutils and stage2-binutils compile fine; as they use flex from the host, they don't use the generated C files15:03
*** zoli_ [~zoli_@linaro/zoli] has quit [Remote host closed the connection]15:04
ssam2zara: ok, but remember that each tutorial adds a cost to changing the import tool15:05
ssam2there are a fair few changes that would improve the tool, and if each change requires updating 3 tutorials then it makes it a bit of a pain...15:07
Zara_ssam2: yeah, I think whatever we do there's going to be more maintainence. :( I'd really like it if we could get a guinea pig and see if it's just me being dense, tbh.15:07
ssam2I'm not trying to stop you from doing stuff, I just don't want to have out of date tutorials :)15:07
*** franred [] has joined #baserock15:08
petefothThe commandment whxcih says 'Code LESS: every line creates a work-chain. Code is a liability, not an asset' applies equalyy to to documentation. As do all the other commandments sice documentation is a form of software. We should only produce the dpocumentation that the spec says is required.15:10
petefothOh wait - there isn't a spec...15:10
petefothSee commandment #1715:10
KinnisonYou may need to link that for anyone here who isn't a Codethinker15:11
petefothThat public at
tiagogomesssam2, it sucks a but having to use generated C files. Do you think it would be hard to bring flex and bison to build essential15:13
ssam2I don't think it sucks that much, it's what every other distro does15:13
Kinnisontiagogomes: I think that'd imply bringing autoconf, automake etc down into b-e if it's not there already15:13
ssam2we're the weird ones for building stuff from git instead of tarballs15:13
Kinnisontiagogomes: which is a big can'o'worms thanks to m4 and perl15:14
ssam2well, flex and bison can be built from tarballs too15:14
ssam2but I think build-essential should be as small as possible15:14
ssam2makes the bootstrap faster and simpler15:14
ssam2and reduces the number of things we have to build from tarballs15:14
ssam2(or rebuild from git afterwards)15:14
* Kinnison concurs15:15
*** zoli_ [] has joined #baserock15:24
*** zoli_ [] has quit [Changing host]15:24
*** zoli_ [~zoli_@linaro/zoli] has joined #baserock15:24
bwhI did an upgrade yesterday, rebooted and journalctl doesn't show anything new since then15:27
bwhAny ideas?15:28
paulsherwoodof a baserock system?15:31
paulsherwood(sorry, just to be sure)15:31
paulsherwoodmaybe ssam2 can help?15:32
bwhI started with the x86_64 image and upgraded using baserock/baserock/definitions commit 2918868b8dbf15:32
paulsherwoodbut it boots ok, other things work?15:33
bwhIt has networking and sshd15:34
pedroalvarezbwh: I've seen that, but I don't know why this happens15:34
*** franred [] has quit [Ping timeout: 245 seconds]15:34
pedroalvarezbwh: I know how to solve it though: `systemctl restart systemd-journald`15:35
pedroalvarez"solve it"15:35
pedroalvarezlooks like the systemd-journald unit starts writing in /var before the system mounts the /var btrfs subvolume15:36
* paulsherwood finds his own vm has the same issue15:37
KinnisonDoesn't the systemd-journald unit have a guard in it for that kind of problem?15:37
tiagogomescan I tell morph to keep the staging areas15:37
paulsherwoodtiagogomes: you mean to diagnose a build? or something else?15:37
bwhYeah, if I move the /var mount out of the way then journalctl shows the missing log entries15:38
pedroalvarezKinnison: no, and talked to some #systemd people about adding RequiresMountsFor=/var/log/journal15:38
pedroalvarezbut they didn't agree15:38
tiagogomespaulsherwood something else, I know that the staging area is kept when building a chunk fails. But I am interested in all stating areas15:39
Kinnisonpedroalvarez: :-(15:39
Kinnisonpedroalvarez: did they suggest an alternative?15:39
ratmice__tiagogomes: so ld/ has a relevent comment right before overriding am__skiplex & friends, tried commenting those out?15:39
pedroalvarezKinnison: I think they suggested an upgrade to 21815:42
Kinnisonpedroalvarez: aah15:42
paulsherwoodtiagogomes: i agree it would be better if it did, but i don't think it's currently possible. you could hack morph to do so easily...15:42
paulsherwoodcomment out 'self.remove_staging_area(staging_area)' in buildcommand.py15:42
*** franred [] has joined #baserock15:43
paulsherwoods/if it did/if it could/15:43
robtaylo1 is now known as robtaylor15:43
tiagogomesthanks paulsherwood15:44
paulsherwoodwin 1515:45
tiagogomesratmice__, I didn't, but I think the problem that I am facing is that the generated C files on the tarball are older than the .y and .l, which could be the root cause why binutils fails to compile15:45
ratmice__tiagogomes: yeah, the am__skiplex/yacc macros disable regeneration of the lex/yacc files unless maintainer mode is enabled, and for some unexplained reason ld is avoiding that15:47
tiagogomesratmice__, the C files will not be generated anyway, because there isn't an available flex/bison when the final version of binutils is built15:48
ratmice__ahh, I thought that they existed but the timestamp was wrong15:50
tiagogomesmmm, empty commits seems to not cause a rebuild15:58
persiaNope.  Morph looks at the tree sha, not the commit sha.16:00
tiagogomesratmice__, they exist because they come from the tarball, and the timestamp is wrong, but I cannot regenerate them as flex/bison are not available16:01
*** cosm [] has joined #baserock16:01
ssam2tiagogomes: you can regenerate them on a system with flex/bison available and then commit them to Git16:02
ssam2that's the point of having all the files from the tarball commited to Git16:02
tiagogomesssam2 yap, but when I tried to that it failed with some link error :'( . That's why I asked whether it was possible to keep the staging areas, so I can import those files from the staging are to the git repo16:04
ratmice__tiagogomes: okay, so if you get rid of the am__skiplex hack in the, and regenerate makefiles and such, then hopefully it shouldn't care about the timestamp unless maintainer mode is enabled (which i'd have to look and see if that gets automagically enabled when building from a git repo or not)16:04
ratmice__but you could probably --disable-maintainer-mode regardless of that16:05
ssam2tiagogomes: if you want a staging area to be kept around after a build, set the last line of the 'install-commands' to `false`16:06
ssam2it's a bit of a trick, but it means the build will fail so the staging area gets kept around16:06
ratmice__ahh, yeah libfl probably?16:06
tiagogomesssam2, ah true, it is better than hacking on morph to not bin them16:07
rdale_is it allowed for one chunk in a stratum to overwrite an executable created by an earlier chunk in the stratum?16:16
paulsherwoodit works, if that's what you mean16:19
paulsherwoodshouldn't be done, though, i think16:19
rdale_ok thanks16:19
rdale_yes, i'm trying to avoid it16:19
paulsherwoodwhy would you need that?16:19
rdale_i need gem version 1.8.25, but ruby 1.9 builds gem version and i can't seem to get the option to stop ruby 1.9 from building the gem executable16:20
ssam2we use that feature in build-essential, and it's horrid in theory, but I can't think of a better approach that's simple, and it does work16:21
*** sambishop [] has quit [Ping timeout: 264 seconds]16:23
locallycompactHow do I force all my lorries to rerun?16:32
pedroalvarezUsing the lorry-controller API ( you can do the following:16:34
pedroalvarezcurl -X POST -d 'path=delta/libflangrt' http://localhost:12765/1.0/move-to-top 16:35
pedroalvarezto promote delta/libflangrt in this case16:35
persiaDoesn't that just promote the lorry, rather than force a re-run?16:36
pedroalvarezhm.. then I don't know what locallycompact wants to do16:37
locallycompactI changed the urls in several hundred lorries from http to https due to url mistake and the interval is set to one week, I want them all to rerun now.16:38
straycatlocallycompact, you can certainly stop new lorries from being added to the run queue, I can't remember if you can clear the run queue so that you can then restart it16:38
straycatI guess you could take a look at the api and see if there's anything of use16:38
straycatlast time I tried doing stuff with lorry controller I found it could with some methods to provide more fine grained control16:39
straycatnoteably, please stop everything now, clear the run queue, and lorry foo, bar and baz NOW16:39
straycat*could do with16:40
*** sambishop [] has joined #baserock16:45
locallycompactThat POST -d path=delta/blah just tells me the path doesn't exist16:48
Zara_email re: rubygems tutorial sent. still hoping I'm being overly pessimistic and we don't actually need to change anything anywhere.16:49
pedroalvarezlocallycompact: People may not recommend this, but in the past, in some trove tests I've removed /home/lorry/webapp.db to clear the queue16:50
locallycompactpedroalvarez: wait, I got it16:51
locallycompactI was using the name of the lorry file without .lorry rather than the name in the json16:51
locallycompactit does make it run now16:51
locallycompactnow to do it for six million files16:51
*** jonathanmaw [] has quit [Quit: Leaving]17:03
franredummm, looks like that apr-util and apache needs --with-apr (where that flag means that we need to specify the source directory :S)17:15
KinnisonEven if apr is installed on the system?17:15
franredKinnison, I think so. I need to run ./buildconf to create configure and ./buildconf requires the source code17:16
richard_mawnot just the headers‽17:17
franredrichard_maw, it checks for a .m4 file17:17
franredif [ -f "$apr_src_dir/build/apr_common.m4" ]; then17:18
franredummm, let me try other way17:18
pedroalvarezfranred: can't you do --with-apr=/usr?17:19
franredpedroalvarez, /usr does not contain that file ;-)17:19
* pedroalvarez shuts up17:19
pedroalvarezLFS uses --with-apr=/usr/bin/apr-1-config17:20
* pedroalvarez stops saying random hints17:20
richard_mawfranred: the debian package has a /usr/share/apr-1.0/build directory which contains apr_common.m417:21
franredpedroalvarez, but in the configure command17:21
richard_mawfranred: so it may be worth looking at how debian makes libapr1-dev17:21
franredrichard_maw, that means that I need to add the files need it -->
franredsee the rm / cp commands17:22
richard_mawwhat do we get in our chunk artifacts when we build apr?17:23
*** sambishop [] has quit [Ping timeout: 256 seconds]17:26
franredrichard_maw, it is still not enough17:37
*** mariaderidder [] has quit [Quit: Ex-Chat]17:38
*** sambishop [] has joined #baserock17:39
*** zoli_ [~zoli_@linaro/zoli] has quit [Remote host closed the connection]17:40
*** zoli_ [~zoli_@linaro/zoli] has joined #baserock17:44
*** zoli_ [~zoli_@linaro/zoli] has quit [Remote host closed the connection]17:49
*** cosm [] has quit [Ping timeout: 244 seconds]17:49
*** wdutch [] has quit [Quit: Quit]17:58
*** sambishop [] has quit [Remote host closed the connection]18:06
*** bashrc [] has quit [Quit: Lost terminal]18:08
*** tiagogomes [] has quit [Quit: Leaving]18:09
ssam2I got really excited about because I thought we could get away from having to maintain our own OpenID server code18:23
ssam2but then I realised that FedOAuth currently only has backends for LDAP or Fedora accounts18:23
ssam2probably either setting up LDAP for Baserock, or writing a backend for FedOAuth that uses a MySQL database or django.contrib.accounts, will be as much work as fixing the remaining bugs in the baserock-openid-provider I wrote :(18:24
*** cosm [Unknown@gateway/vpn/mullvad/x-cvyihuglkgcseoud] has joined #baserock18:33
*** zoli_ [~zoli_@linaro/zoli] has joined #baserock18:35
*** ssam2 [] has quit [Remote host closed the connection]18:39
*** locallycompact [] has quit [Ping timeout: 264 seconds]18:54
*** rdale [] has joined #baserock19:41
*** rdale_ [] has quit [Ping timeout: 264 seconds]19:44
*** rdale_ [] has joined #baserock19:54
*** rdale [] has quit [Ping timeout: 244 seconds]19:57
*** zoli_ [~zoli_@linaro/zoli] has quit [Remote host closed the connection]20:01
*** rdale_ [] has quit [Ping timeout: 264 seconds]20:01
*** rdale [] has joined #baserock20:03
*** rdale_ [] has joined #baserock20:10
*** rdale [] has quit [Ping timeout: 255 seconds]20:13
*** cosm [Unknown@gateway/vpn/mullvad/x-cvyihuglkgcseoud] has quit [Quit: Leaving]20:14
*** rdale [~quassel@] has joined #baserock20:28
*** rdale__ [~quassel@] has joined #baserock20:31
*** rdale_ [] has quit [Ping timeout: 255 seconds]20:31
*** rdale [~quassel@] has quit [Ping timeout: 252 seconds]20:34
*** rdale__ [~quassel@] has quit [Read error: Connection reset by peer]20:34
*** rdale [] has joined #baserock20:36
*** rdale_ [] has joined #baserock21:09
*** rdale [] has quit [Ping timeout: 264 seconds]21:10
*** rdale [] has joined #baserock21:14
*** rdale_ [] has quit [Ping timeout: 240 seconds]21:15
*** rdale [] has quit [Read error: Connection reset by peer]21:18
*** rdale [] has joined #baserock21:23
*** zoli_ [~zoli_@linaro/zoli] has joined #baserock21:25
*** zoli_ [~zoli_@linaro/zoli] has quit [Remote host closed the connection]21:27
*** zoli_ [] has joined #baserock21:41
*** zoli_ [] has quit [Changing host]21:41
*** zoli_ [~zoli_@linaro/zoli] has joined #baserock21:41
paulsherwoodnot mysql, please :)22:09
persiaWhy not?  We're using it behind other baserock infrastructure systems.22:13
paulsherwoodbecause i favour less code, with less dependencies, that's easier to install22:14
paulsherwood(i'm being as polite as i can)22:14
persiaI prefer using the same software for as many purposes as possible, rather than different software for different things.  Since we know we can't use postgresql for some stuff, and have to carry mysql anyway, it seems extra work to do both.22:15
persiapoit is single-user22:15
paulsherwoodnot hard to fix that, i would think?22:16
paulsherwood(says smartass, after alcohol)22:16
persiaAfter looking at the code, I suspect the alcohol of having impaired judgement22:17
paulsherwoodi didn't look at the code :)22:19
paulsherwood'moin moin wiki includes an openid provider' - chance to move wiki to python too? :-)22:20
persiaI played with that a bit: it seemed to work OK, but it was a bit annoying.22:21
persiaAnyway, I've come to like git-backed wikis, with moinmoin most certainly isn't :)22:21
paulsherwoodyes. why is nothing ever perfect? :)22:23
persiaEveryone has different requirements, so when we look for open code, we either get a baseline that works for everyone but doesn't do enough or a specialised system.22:24
persiaI'd like everything to be modular and flexible, but people always tell me that it's too hard to do it that way.  I suspect that it's just a matter of time, and that everything will be perfect later, but it may be a very long time.22:25
paulsherwoodmaybe we could persuade everyone we know to ambush him in moin-dev and persuade him22:26
persiaProbably not, because
* paulsherwood gives up for the day...22:38
*** zoli_ [~zoli_@linaro/zoli] has quit [Remote host closed the connection]22:54
*** zoli_ [~zoli_@linaro/zoli] has joined #baserock22:55
*** zoli_ [~zoli_@linaro/zoli] has quit [Ping timeout: 240 seconds]23:06
rjekMySQL makes me very sad.23:11
bwhrjek: I imagine you especially like MySQL security patches23:12

Generated by 2.15.3 by Marius Gedminas - find it at!