IRC logs for #baserock for Friday, 2015-05-08

*** sambishop has quit IRC02:11
*** zoli__ has joined #baserock03:26
*** zoli__ has quit IRC05:28
*** zoli__ has joined #baserock05:28
*** zoli__ has quit IRC05:44
*** zoli__ has joined #baserock06:21
*** a1exhughe5 has joined #baserock07:21
*** mariaderidder has joined #baserock07:23
*** Albert_ has joined #baserock07:24
*** Albert_ has joined #baserock07:25
paulsherwoodfwiw ybd /win 2407:46
paulsherwoodi started to say, fwiw ybd built all of the release x86_64 definitions in approx 3 hours07:47
*** sambishop has joined #baserock08:00
straycatdistbuild-list-jobs, could we maybe change it to distbuild-list, or distbuild-ls, or something a little less long winded08:01
straycator… aliases, yes one day, one day…08:02
*** bashrc_ has joined #baserock08:03
straycatmaybe even distls, and then we could have diststat08:03
paulsherwoodsorry if this is a dumb question, but why is this functionality needed, straycat?08:03
paulsherwood(i am trying to answer your question, btw)08:04
straycatbecause you can now detach from your running distbuilds, like screen but not screen08:04
*** edcragg has joined #baserock08:11
*** Albert_ has quit IRC08:24
pedroalvarezradiofree: looks like finally they decided to do somthing -;a=commitdiff;h=5fac47c728183c4dfee12524348240a07e5ef05c08:30
pedroalvarezI was proud of my patch :/08:31
pedroalvarezdid anyone integrate Mariadb in baserock?08:36
paulsherwoodi think i gave up08:37
paulsherwoodpedroalvarez: does that mean that Ingo Huerner re-implemented your solution?08:38
pedroalvarezbut I believe someone continued08:38
pedroalvarezpaulsherwood: worse, my solution was to add support to --disable-werror, he dropped my patch because "GENIVI requests to treat warnings as errors", and now he disabled -werror always08:40
pedroalvarezwell, IMO is worse08:40
paulsherwoodis he right?08:40
paulsherwoodthis is a philosophical point. better to force the build to be clean?08:41
rjekI prefer pedroalvarez's approach.08:42
KinnisonIt's better if the from-git variant always puts -Werror on by default08:42
Kinnisontraditionally you should have -Werror for developers and -Wno-error for integration/distribution08:42
*** ssam2 has joined #baserock08:42
*** ChanServ sets mode: +v ssam208:42
rjekI like -Werror for development, but it's important to let people opt to disable it when building for production or end-user08:42
*** gary_perkins has joined #baserock08:43
pedroalvarezhe contradicted himself with his patch, I think08:43
* paulsherwood wonders whether it's worth raising this on the genivi-projects list08:43
pedroalvarezI don't know.. I'm very merciful08:47
* paulsherwood is not08:50
straycatwe know :)08:50
pedroalvarezCTtpollard:  was it you who was playing with mariadb?08:58
*** tiagogomes has joined #baserock09:00
CTtpollardpedroalvarez: our project was, ctgriffiths did the work with it09:00
pedroalvarezyay, I found him! thanks CTtpollard09:01
* ctgriffiths was never any good at hide and seek09:02
ctgriffithspedroalvarez: it's on my github but it's jumbled with the storyboard stuff, it will need cleaning up:
KinnisonWhy do we want mariadb?  We already have pgsql09:04
pedroalvarezKinnison: just in case openstack stops working with pgsql09:04
KinnisonCan we just murder whoever proposes that?09:05
KinnisonI remember back when I tried OpenStack a few years ago and it used MySQL and it was awfully slow at the DB layer and massively inefficient and broken09:06
KinnisonBecause they had to layer proper transactionality on top09:06
rjekISTR that quite a few parts of the OpenStack project already don't work with PostgreSQL.  StoryBoard IIRC, for example09:06
pedroalvarezKinnison: I think we are late to murder people09:07
KinnisonOpenStack Infra. projects are very different from OpenStack itself09:08
Kinnisontheir Infra. projects are often just "enough" to work09:08
CTtpollardStroyboard did drop pgsql support09:08
*** Krin has joined #baserock09:10
pedroalvarezI can't find any info about this, but I believe that OpenStack is not going to drop patches if they break pgsql09:11
pedroalvarezI think is good to bear that in mind if someday we plan to upgrade to kilo09:12
paulsherwoodKinnison: on this i agree, pgsql is much simpler. trouble with software is other people's choices may not align with what's *best*09:14
Kinnisonpaulsherwood: in a lot of ways, MySQL/MariaDB is simpler because people find pgsql's adherence to a properly ACID model hard to work with09:15
Kinnisonpaulsherwood: also, less relevant these days, but libpq-dev was much harder to bind cleanly to09:15
paulsherwoodpaulsherwood: i've just had better experiences with pgsql - it was trivial to add it to baserock, for example09:16
paulsherwoods/paulsherwood/Kinnison/ obviously ;)09:16
straycatmariadb would be better for working with ceilometer and apparently has a bette license09:16
straycatslackware removed support for mysql in favour of mariadb in 201309:17
Kinnisonstraycat: mariadb's licence is better than mysql's -- pgsql's is best09:17
rjekOracle screwed over MySQL so I think basically everybody went for MariaDB09:18
* bashrc_ uses MariaDB09:18
pedroalvarezbaserock is currently using: redis, mariadb, pgsql, sqlite...09:20
* rjek has an extremely strong preference for predicability, reliability, and data integrity, so uses PostgreSQL09:20
*** lachlanmackenzie has joined #baserock09:23
*** CTtpollard has quit IRC10:03
*** CTtpollard has joined #baserock10:14
*** Albert_ has joined #baserock10:14
edcraggi'm getting this error building binutils on armv7lhf: any ideas?10:38
straycatthe morph test suite seems to be failing its deploy scenarios, i can't look into this right now but just thought i'd say10:41
SotKstraycat: it seems ok to me, what error are you seing?10:42
rdalewhy is gobject-introspection in the core stratum - it isn't a core build tool?10:43
pedroalvarezedcragg: I think you are having problems with your filesystem10:48
pedroalvareza release is being done right now, and we are not hitting any problems when building10:48
ssam2that is when the change was made10:48
ssam2straycat: oh, I guess that's my fault, thanks for pointing it out10:49
SotKstraycat: oh, its failing for me too... it must be a recent change10:49
edcraggpedroalvarez: thanks... weirdly, it builds if i unmount /tmp before running morph11:02
edcraggprobably because i also symlinked /tmp/morph_tmp to the external sdd11:04
edcraggi'm guessing11:04
*** Albert_ has quit IRC11:36
ssam2fix for sysroot.write issue:
ssam2can anyone give it a quick review so I can merge it for the 15.19 release/11:39
pedroalvareztiagogomes ^ ?11:42
edcraggstage1 gcc seems to be failing as far as i can tell with 'Switch "--with-arch" may not be used with switch "--with-cpu"' on armv711:54
radiofreei just compiled gcc on armv7 fine11:56
edcragggood to know, dunno what i'm doing wrong then11:57
radiofreepastebin the full build log11:59
radiofreewell, the log from gcc, though at state1 there's not much more11:59
edcraggradiofree: thanks, the whole thing is on fs1... fs1:/tmp/rb-out but the end is something like
ssam2I realised that we need to mark baserock:baserock/definitions as using format version 3, before the 15.19 release12:12
ssam2as a way of noting that it requires the install-essential-files.configure extension12:13
paulsherwoodssam2: this is too complex, isn't it? too easy to miss stuff, notice later12:14
ssam2I've sent to update definitions.git12:14
ssam2paulsherwood: we didn't miss this. it caused build failures on the Masons, until the version of Morph was updated on them12:15
ssam2paulsherwood: I think in future we should only update the Masons immediately after a release12:15
ssam2so they will spot if someone commits something to master that doesn't build with the previous release12:15
pedroalvarezand whenever they break, revert the cause?12:15
ssam2yes, until we make another release12:16
ssam2but I agree it's too complex!12:16
ssam2I also updated the description of V3 in to mention install-essential-files.configure12:16
paulsherwoodssam2: have you thought any further about the ideas we discussed the other day? i tried them on persia, but i crashed and burned12:17
ssam2they're in my mind. I do think having a bootstrapping method for definitions that was committed to definitions would mean we could stop worring about whether one release could build the next one12:18
ssam2am going to lunch now, will respond to any other comments when i'm back!12:18
*** zoli__ has quit IRC12:20
*** zoli__ has joined #baserock12:22
*** zoli__ has quit IRC12:26
Zarahalp... What am I missing?12:32
*** fay__ has quit IRC12:32
mwilliams_ctZara: are you using the isc-dhcp tarball or git repo?12:33
*** fay_ has joined #baserock12:34
pedroalvarezmwilliams_ct: looks like a git lorry in git.baserock.org12:36
pedroalvarez(it has history)12:36
Zarait's lorried upstream so I assume a git repo. (the download link is a tar.gz but I think it's been turned into a git repo)12:36
mwilliams_ctZara: pedroalvarez: ok, so the trouble is that the git repository doesn't include all the files the tarball does and those files are needed to compile12:36
mwilliams_ctI think jjardon mentioned it might make sense to lorry the tarball as well/instead12:36
pedroalvarezoh, indeed, there isn't a bind/ folder12:37
pedroalvarezutil/ seems to generate it12:38
pedroalvarezhm.. you might be able to use part of that script to generate what you need, but yes, looks like a tarball import might be easier12:41
*** zoli__ has joined #baserock12:44
ZaraI see. It's weird that their git is different from their tarball.12:47
Zarahow would I do a tarball import?12:47
KinnisonNot uncommon for release tarballs to be materially different from tags in revision control12:47
persiaEven common: several projects have a tarball release procedure that sets up the code for builds in ways that would be annoying to commit to version control, but are essential to be confident of consistent support.12:50
ZaraI was just surprised that they were so different.12:52
Zarathe tarball is about 8 times the size, btw, if that matters12:52
persiaSometimes upstream build tools encode the mechaism to generate the tarball: `make tarball` or similar.12:53
persiaIF your environment isn't carefully controlled, you won't end up with the same file content they get when they make a tarball, so upstream may not wish to support you using the results, but you can work around the odditieis.12:54
edcraggradiofree: the branch i'm building from was 20 or so commits behind master, rebased and gcc seems to be building13:16
Zarahad a look around, am still in the dark about how to do a tarball import. sorry if this is documented and I've missed it! (alternatively, if running util/ will work equally well, how do I put that in the morph? I'm getting 'permission denied' so I assume I need more than just ./util/ on a line!)13:22
paulsherwoodsh ./util/bind.sh13:26
paulsherwoodor chmod a+x ./util/bind.sh13:26
paulsherwoodZara: ^^ (maybe, or i may be wrong)13:27
mwilliams_ctZara: I think what pedroalvarez means is to lorry the tarball. if you have a look in the lorries repo, try grepping for "tarball" and you should find some examples13:28
*** Albert_ has joined #baserock13:29
pedroalvarezI always prefer to go with git whenever possible13:32
Zaraoh, found it in the trove reference manual. I'd searched 'tarball' and that hadn't come up, so I thought it might not be on there, but it is.13:35
Zarapaulsherwood: I tried the former and it didn't work. :( will try the latter now! otherwise, will lorry.13:36
SotK may help with any other lorry questions13:37
ratmice__wish there was a standard dist-like target that didn't generate the tarball, but the sources that would go into the tarball...13:39
persiaratmice__: Unfortunately of limited use to most folk.  Even Debian, which frequently regenerates tarballs appreciates ending up with a tarball afterwards.13:41
paulsherwoodratmice__: can you expand on that use-case, please?13:42
* paulsherwood doesn't fully understand, but knows that this needs work13:42
ratmice__paulsherwood: in general something to make it simpler to create a branch which includes the files that normally go into the EXTRA_DIST automake variable13:43
persiaZara: provides some validation that upstream doesn't provide an importable VCS: you are not alone in your frustration.13:44
ratmice__I should dig through automake :)13:44
paulsherwoodratmice__: rather you than me :-)13:45
pedroalvarez`mkdir bind; cp util/ bind/` as a preconfigure command seems to "fix" the configure failure, but other things might break13:51
ratmice__seems like 'distdir' target may work13:51
pedroalvarezit starts complaining for other things when running make13:51
persiaOne of the reasons that tarball generation is separate is that sometimes the set of things required to build something from VCS are much larger than the set of things required to build from the tarball.13:52
pedroalvarezZara: tarball import13:53
Zarahehe, was about to say that chmoda+x worked but apparently running that script wasn't enough to create the directory, something was weird, I didn't know what, and I was going to do a tarball import. glad that's the consensus :P13:54
Zaranew lorry coming soon.13:54
pedroalvarezI believe that script tries to fetch repository13:56
pedroalvarezI may be wrong13:56
ratmice__persia: indeed, binutils,gdb are a rather extreme case of this there (was/is) a lot of stuff in the repo, simulators, newlib, binutils, gdb, things for generating opcodes, some of which have no release tarball, and some of which have their own tarball thus, there's many tarballs coming from one repo, and a few only ever used by a small subset of developers13:57
persiaEclipse belongs on that list :)13:58
Zarawhat should I call this lorry, given that we already have isc-dhcp? I was going to go with isc-dhcp-tarball , but might as well check before I send it out into the world.14:05
persia+1 for that nomenclature14:05
*** fay_ has quit IRC14:06
ratmice__fwiw distdir is even documented :)14:08
persiaZara: Does `make distdir` work?14:08
* ratmice__ notes it doesn't seem to help in this case since the release process seems to go outside of the Makefiles14:08
ratmice__persia: nope14:08
persiaratmice__: Thanks14:09
ratmice__SUBDIRS=bind ... (where generates bind/) so it requires to be run before you go through any of that..14:10
*** fay_ has joined #baserock14:20
Zarapersia: I don't know, didn't try it, made the lorry now instead. (
persiaZara: No worries: ratmice__ tried it, and it didn't help.14:28
ratmice__I actually didn't try it, but i'm confident it won't work14:30
ssam2radiofree: VT switching works in the x86_32 15.19 release image. (no vga=xx in the kernel commandline)14:42
ssam2also, it'd be really good if we could fix Baserock so VirtIO root disks work! we even have an initramfs now!14:43
ssam2ah, this is because the x86_32 systems don't have an initramfs still14:45
radiofreessam2: yes, it works without vga=15:10
radiofreeso devel images are fine, however weston/genivi images don't work15:10
radiofreeas a monstrous hack i was thinking about adding a getty@tty2.service15:10
radiofreeas long as noone does ctrl-alt-f3 (+) it should work15:11
* persia fears that level of monstrosity15:13
Zara(oops, didn't realise gerrit updates the patch set every time you edit the commit message. sorry if that just spammed anyone, in future I'll draft changes first!)15:14
*** mariaderidder has quit IRC15:27
Albert_Question from an old git: I have a patch for AudioManager which is going into a branch from 6.2, so I have created branch baserock/6.2 as advised. How do I then push this to the right place? Or what do I do next?15:38
persiaAlbert_: Upstream is not likely to take your patches in a timely manner?15:39
Albert_The last tag upstream was a while agao, it like like a bit hase changed sinse then. My patch probably isn't valid any more, but in the meantime...15:40
ssam2I've reworked the 15.19 release notes a bit, here is another darft:
ssam2darft? draft.15:40
Albert_* it looks like a bit ...15:41
SotKssam2: s/cancel the running/cancel the running build/ in the ui changes section?15:42
radiofreelots of changes, maybe mention there's now a new way to deploy a jetson?15:42
Albert_Meaning my patch would perhaps not be valid upstream because of changes, but in the meantime, it doesn't build15:42
ssam2radiofree: oh, is the baserock-flash script new for this release? I'll put it in then15:43
ssam2SotK: good spot15:43
Albert_So the question remains: how to get this patch in. I've not done it on BR before15:45
SotKssam2: looks fine other than that15:48
pedroalvarezradiofree: regarding the flash script, I just fixed a couple of things on the wiki15:51
pedroalvarezalso, is taking ages to dd the image15:51
persiaAlbert_: The best method is to land it upstream, and then submit a definitions patch to use it.15:53
tiagogomespedroalvarez did you set a block size?15:54
persiaAlbert_: If that doesn't work, for one reason or another, the backup is to find a merger who is willing to add a local branch of the project on git.baserock.org15:54
persiaThen you can submit a change to definitions to use the local branch.15:54
pedroalvareztiagogomes: 8M15:54
pedroalvarez( I didn't, the script has that set)15:54
tiagogomespedroalvarez it is not much if is an SSD. You can use iotop to check the transfer rate15:55
Albert_The latter is the way that has been suggested to me. As for changing def'ns, that's my next job.15:55
Albert_Perhaps jonathanmaw or pedroalvarez who've been involved with AudioManager?15:56
ratmice__Albert_: from their blog post 'Release tag set', it seems like they definitely anticipate supporting 2 branches, one of them being stable, so I doubt they would discourage patches.15:57
radiofreessam2: maybe worth pointing out that mesa 10.5.4 is only used on a jetson, vm/x86 systems are still using the older one15:57
pedroalvareztiagogomes: thanks for the suggestion15:57
*** zoli__ has quit IRC15:58
Albert_ratmice__, That's helpful. Where was this blog post?15:58
tiagogomesiotop would fit nice in devtools15:58
*** CTtpollard has quit IRC15:58
pedroalvareztiagogomes: looks like my limit is the usb wire15:59
ratmice__hopefully thats the write AudioManager :)15:59
Albert_thanks ratmice__15:59
persiaAlbert_: Even if you can't land the patch, most Baserock mergers are more likely to approve a local branch if the patch is at least under review upstream.16:05
Albert_Noted. I'm also looking to see if I can get an existing commit from which to branch locally so that there's no need to make upstream changes16:06
* persia fails to parse that16:07
*** a1exhughe5 has quit IRC16:10
Albert_What I mean is, if there is a commit which I can rely on, and in which the build error I have experienced has been fixed, then I may be able to void having to patch. But which commit, when even tagged commits have broken the build?16:14
Albert_For it is only a build-time error.16:14
rdaleisn't it because you are using a newer cmake that the audiomanger maintainers?16:15
Albert_That is a factor, but there's more: there is an additional error16:16
persiaAlbert_: Does the tip of the current development branch work?16:17
Albert_For the smake it's just involves a change to a morph file16:17
persiaIf you can fix it in a morph file, you don't need to patch AudioManager, making your life easier.16:18
Albert_It builds but doesn't run, not on AudioManager16:18
persia(although submitting things upsteam is always good anyway)16:18
Albert_persia, one fix can be done n a morph file, but not the other.16:19
Albert_It's REALLY simple - just a couple of spelling errors16:19
Albert_I don't know why it wasn't spotted. It kills the build.16:20
persiasimplicity isn't important, as annoying as that can be.16:20
*** CTtpollard has joined #baserock16:58
*** zoli__ has joined #baserock16:58
*** Albert_ has quit IRC17:01
*** Albert_ has joined #baserock17:01
*** zoli__ has quit IRC17:03
*** bashrc_ has quit IRC17:06
ssam2radiofree: good point on mesa. thanks17:06
*** Krin has quit IRC17:07
*** sambishop has quit IRC17:10
*** CTtpollard has quit IRC17:12
ssam2jjardon, radiofree: what's the deal with mesa-common-vm ?17:13
ssam2it's something to do with 3D drivers in VMs but I don't understand it enough to write about it in the release notes17:13
ssam2ok, so:17:16
ssam2Jetson systems now use Mesa 10.5.4. x86 systems still use a patched17:16
ssam2  version of 10.3.7, so that gallium-egl is available: that driver is17:16
ssam2  needed to use Weston in a VM, and it is removed in 10.4.17:16
pdarI started to deploy a base-system-x86_64-generic.morph at 16:25 and its hanging at the "installing extlinux" stage. How could Igo about finding out whats wrong?17:17
radiofreessam2: looks good17:17
ssam2hmm, except the 10.4 release notes don't mention removal of gallium-egl17:17
ssam2i'll leave it at 'so that gallium-egl is available: that driver is needed to use Weston in a VM.'17:19
pedroalvarezpdar: hard to say17:22
pedroalvarezpdar: that should take few seconds17:22
pdarYep, I noticed the problem when testing something else so I thought I'd check a simpler system, then forgot about it..17:24
*** sambishop has joined #baserock17:35
*** gary_perkins has quit IRC17:35
*** gary_perkins has joined #baserock17:35
*** gary_perkins has quit IRC17:39
pdaroh, just rebooted my baserock and it done worked. Hoorah o/17:39
*** tiagogomes has quit IRC17:48
*** edcragg has quit IRC17:52
ssam2Baserock 15.19 is released!18:15
ssam2The word 'baserock' appears 39 times in the release notes18:15
*** ssam2 has quit IRC18:23
*** Albert_ has quit IRC18:29
*** zoli__ has joined #baserock18:29
*** zoli__ has quit IRC18:34
*** lachlanmackenzie has quit IRC18:35
*** rdale has quit IRC18:57
*** zoli__ has joined #baserock19:30
*** zoli__ has quit IRC19:35
paulsherwoodw00t! :)19:43
*** zoli__ has joined #baserock21:01
*** zoli__ has joined #baserock21:01
*** zoli__ has joined #baserock21:02
*** zoli__ has joined #baserock21:03
*** zoli__ has joined #baserock21:04
*** sambishop has quit IRC21:19
*** zoli__ has joined #baserock22:20
*** zoli__ has quit IRC22:24
*** zoli__ has joined #baserock22:48
*** zoli__ has quit IRC23:36

Generated by 2.14.0 by Marius Gedminas - find it at!