IRC logs for #baserock for Thursday, 2014-08-28

*** flatmush [] has joined #baserock00:00
*** flatmush [] has quit [Ping timeout: 255 seconds]01:15
*** flatmush [] has joined #baserock01:17
*** JPohlmann [~jannis@xfce/core-developer/JPohlmann] has quit [Ping timeout: 240 seconds]02:12
*** JPohlmann [~jannis@xfce/core-developer/JPohlmann] has joined #baserock02:13
*** vmeson [~quassel@] has quit [Ping timeout: 250 seconds]05:23
*** vmeson [~quassel@] has joined #baserock05:25
paulsherwoodhas anyone here managed to self-upgrade a jetson?07:00
*** fay_ [] has joined #baserock07:39
*** franred [] has joined #baserock07:48
SotKpaulsherwood: I have07:50
SotKwhat problem are you having?07:50
paulsherwoodi feel like i'm in uncharted territory... looks like james' updates are pre-c.i.d07:52
SotKthat is strange, and I don't know enough about the inner workings of the upgrade process to help you I feel...07:55
SotKI upgraded with a system in the baserock/tegra branch (but rebased to use chunks in definitions) and didn't see anything like that07:56
paulsherwoodok let me check - i may ask you again :)07:56
pedroalvarezgood morning!08:10
pedroalvarezthe upgrade deployment is expecting a  "tegra124-jetson-tk1.dtb" in the rootfs of the system that you are deploying08:13
tlsa_ is now known as tlsa08:14
pedroalvarezThis is specific of a jetson board upgrade, so I don't have any idea of why it should be there, and why is not there08:14
*** jonathanmaw [] has joined #baserock08:19
persiaTo me, that looks like the device tree data file, although I don't know why it might be absent.08:19
robtaylo1 is now known as robtaylor08:23
SotKI want to build a Baserock system with Qt in it, but my devel system doesn't have enough RAM08:44
persiaAdd swap08:45
persiaIt will be slow and painful, but it should build.08:45
SotKslow and painful is better than not at all I guess08:46
* persia has built boost with 512M RAM and koffice with 256M RAM, both of which required patience08:46
pedroalvarezin some chunks I found that using "max-jobs: 1" helped with small RAM08:47
persiaIf you can cause the build to happen in an environment in which you don't need to do anything until tomorrow, that helps.08:47
liw-orcswap may save you, unless the linker needs more memory than your address space allows08:47
persialiw-orc: Do we have any chunks that can't build in 32-bit 2G/2G?08:47
robtaylor(only case i've ever heard of needing that is building chromium with full LTO)08:48
liw-orcI was told in 2011 by people who've tried it that 32-bit ARM didn't have the address space to link then-modern versions of OO.o and LibreOffice; I haven't investigated in detail08:48
persiaHmm.  I'd believe that.08:49
robtaylorI suspect again, linking them with LTO08:49
persiaMind you, that's a kernel option: one can use a 3G/1G split in 32-bit to get a bit more space.08:49
robtaylorif you really had to do it, you could farm it out to a cross-linker on a 64bit system08:50
persialet's avoid that: we don't have such tooling already in place.08:50
persia(and any upstream that needs more than 3G to link should be admonished)08:51
robtaylorI don't think anyone needs to do this at this moment in time08:51
robtaylorif you do hit a problem, just disable LTO08:51
*** jonathanmaw [] has quit [Ping timeout: 240 seconds]08:58
paulsherwoodSotK: maxjobs?09:02
*** jonathanmaw [] has joined #baserock09:16
*** ssam2 [] has joined #baserock09:23
paulsherwoodSotK: did you succeed?11:09
radiofreepaulsherwood: the patch to your device tree is wrong11:24
radiofreesolution: use the correct path (don't have a leading /)11:26
radiofreeso boot/tegra-foo.dtb not /boot/tegra-foo.dtb11:26
paulsherwoodthank you! i'll try that when i'm next near to the jetson11:27
radiofreemax-jobs:1 when building qt will take ages btw11:29
persiaTaking ages is better than not completing.11:39
pedroalvarezradiofree: define ages :)11:41
radiofreewhich version of qt?11:42
SotKpaulsherwood: no, gstreamer is failing to build :(11:42
SotKradiofree: qt511:43
radiofreeSotK: i sent patches for that ages ago11:43
radiofree"[PATCH 0/7] Fix gstreamer 0.10 build with newer glibs"11:43
pedroalvarezyeah, they were merged, and I built it last week to do a Genivi release11:43
* SotK investigates11:44
radiofreepedroalvarez: i don't see it in
radiofreea genivi release doesn't have gstreamer 0.10?11:45
SotKthere is only 1 +1 on the patch on the ML11:45
pedroalvarezradiofree: you are right, Genivi doesnt have that version of gstreamer11:47
pedroalvarezI assumed they were merged, because Genivi stop failing building11:47
SotKradiofree: using the SHA from your patch series it built fine, thanks11:53
radiofreei think you'll be looking at 3-5 hours to build qtbase with maxjobs 112:03
radiofreethen of course there's qtwebkit...12:03
persia3-5 hours isn't so bad.  One gets to work on the patches twice a day.12:05
liw-orcpersia, I don't know if we have anything that doesn't build in 32-bit 2G/2G; I assume we don't, or we would've noticed12:06
radiofreeqt5 is one of the things that would really benefit from using distcc12:08
persiaAs long as the address space split is identical for all 32-architectures, it's usually safe.  It's when folk change X86_32 to 3G/1G and not the other 32-bit architectures, and continue to depend on x86_32 as a canary that trouble occurs.12:08
*** thecorconian [] has joined #baserock12:17
*** fay_ [] has quit [Ping timeout: 260 seconds]13:02
*** fay_ [] has joined #baserock13:51
*** genii [~quassel@ubuntu/member/genii] has joined #baserock13:59
pedroalvarezhm... when configuring a distbuild worker we do this:14:20
pedroalvarezssh-keyscan -t dsa,ecdsa,rsa "$TROVE_HOST" >> "$1/root/.ssh/known_hosts"14:20
pedroalvarezI'm trying to figure out how can I do that command idempotent14:21
richard_mawif your known_hosts is encrypted, I'm not sure, but otherwise you could grep known_hosts for $TROVE_HOST14:21
liw-orcthat file shouldn't be going into root's home directory, btw; it should be going to /etc/ssh14:22
pedroalvarezliw-orc: I can change that14:23
pedroalvarezI found  `ssh-keygen -H -FHOSNTAME` useful to check if a host is defined in known_hosts14:24
persiaThe relevant user on the trove should probably have an ssh configuration to not encrypt known_hosts, but you can use `ssh-keygen -r` to find out if it is already there if it is encrypted14:24
liw-orcI think it should be /etc/ssh/ssh_known_hosts (under $1 obviously)14:24
persia(or even if it isn't)14:24
pedroalvarezbut, what about if the host has changed?14:24
persia-H doesn't do that according to my docs.14:24
persiaIf you want to replace unconditionally, use -R14:24
pedroalvarezpersia: -H -F HOSTNAME, sorry14:25
liw-orcpedroalvarez, btw, what do you mean by idempotent? this is being run at configuration extension time, and not multiple times14:25
pedroalvarezliw-orc: is because I'm moving it out from the configuration extension14:26
liw-orcpedroalvarez, and it might get run multiple times?14:26
pedroalvarezevery boot14:26
persiapedroalvarez: I'm unsure about -H-F: use -r to check if it is there, or use -R to replace it (and -R will add it if it isn't present)14:27
persia*every boot*  Why?14:27
liw-orcpedroalvarez, do you need to notice if the server host key has changed? if so, maybe it'd be better to just have the client not care about host key checking14:27
persiaThat's not safe.14:28
pedroalvarezI think we care14:28
persiaBut resetting it every boot is just as unsafe as not checking14:28
liw-orcpersia, yeah, that was my point14:28
pedroalvarezAnd I agree with you both14:28
persiaSo don't do this then :)14:28
pedroalvarezjust check if the host has already an entry in the known_hosts file, and not update it14:29
liw-orcpedroalvarez, will there ever be other hosts in that known hosts file?14:29
pedroalvarezliw-orc: the one that is in the /root/.ssh/, yes14:31
pedroalvarezI don't know about the /etc/ssh one14:31
persiaWhen is the user expected to provide TROVE_HOST?  Is this deploy time or instantiation time?14:31
liw-orcgo with the assumption that /etc/ssh/known_hosts only ever has the Trove's host key, and only generate if the file doesn't exist yet14:31
* persia is still worried with that solution that spoofed DNS can provide the wrong value.14:33
pedroalvarezJust to let you know what I'm trying to achieve: I'm trying to do a generic distbuild-node deployment. IMO this is needed to have Mason in OpenStack14:33
richard_mawis it feasible to generate host and user keys in advance, and pass them to cloud-init for the trove and the distbuild nodes?14:35
paulsherwoodquick question - if i want to track down the GENIVI I-0.1 release, what do i look for?14:35
persiarichard_maw: Yes, but that means passing private keys over the wire.14:35
paulsherwoodi see no tag14:35
pedroalvarezpaulsherwood: seems I missed that step from our release process14:36
pedroalvarezpaulsherwood: there is a branch, I will do the tag now14:37
pedroalvarezpaulsherwood: tag GENIVI-I0.114:42
paulsherwoodgreat, thank you14:43
ssam2franred: regarding your cpython patch, merge and fix however pleases you. once we have an actual policy we can clean up the refs and tags and unpetrify-refs to be consistent :)14:51
persiaWell, either that, or we remove the unpetrify-ref entirely, if it comes to that (which I've seen proposed before).14:52
ssam2if fran wants to remove it, he may14:55
persiaIs it safe to do so?  I thought morph still complained if it was absent.14:56
radiofreei usually don't use unpetrify-refs (just ref: branchname) and i've never had problems14:58
persiaOh, cool.15:00
richard_mawref: branchname means that if someone else moves those refs you have to build their changes though15:00
franredssam2, persia, Im going to merge the patch as it was reviewed. I leave the unpetrify-refs discussion for the mailing list15:00
persiaSeems reasonable15:00
geniiHm, 50 Commandments. Even God only had 1015:10
geniiIs there a recommended CAN adapter ?15:12
persiaBsaerock is mostly focused on the software side: anything with kernel support should work.15:14
geniipersia: OK, thanks.15:15
*** ssam2 [] has quit [Ping timeout: 240 seconds]15:28
paulsherwoodgenii: commandments seemed like a good name at the time... but that was in the 90s :)15:32
geniipaulsherwood: Hah, understandable.15:33
* richard_maw realises he'll have to start referring to the 90s as the generation before him referred to the 80s15:35
richard_mawthough 00s (pronounced noughties) sounds silly15:35
*** fay_ [] has quit [Remote host closed the connection]15:59
*** jonathanmaw [] has quit [Quit: Leaving]16:00
*** franred [] has quit [Quit: Leaving]17:00
paulsherwoodis stripping debug symbols as simple as running 'strip'?17:05
persiaYes, but generally speaking one wants to save them, so when the system breaks, one has the means to recover them and debug the issue.17:06
persiaBe aware that this may be dangerous: some sorts of object files rather like to keep at least some of their symbols (e.g. .so files).17:07
paulsherwoodi was just wondering where i might insert this in my workflow, to have smaller images and thus reduce deploy/upload times17:08
persiaIt's something morph should probably do when it splits the artifacts.17:09
paulsherwoodi'm willing to risk breakages while experimenting17:09
persiaThat seems a logical point to separate blobs into two pieces: the stripped executable and the debug symbols (which can later be used for debugging).17:09
persiaBut someone more familar with how morph manages builds may have a better opinion than I.17:10
*** ssam2 [] has joined #baserock17:32
robtaylorgenii: looking at drivers/net/can/usb looks like kernel has buit in support for:17:49
robtaylorEMS CPC-USB/ARM7 CAN/USB interface17:49
robtaylorESD USB/2 CAN/USB interface17:49
robtaylorKvaser CAN/USB interface17:49
robtaylorPEAK PCAN-USB/USB Pro interfaces17:49
robtaylor8 devices USB2CAN interface17:49
robtaylor EMS Dr. Thomas Wuensche (
robtaylor esd electronic system design gmbh (
robtaylor PEAK-System Technik (
robtaylor 8 devices (
robtaylorgenii: hth :)17:50
geniirobtaylor: Thanks! Yes, I also found most of those listed at ...I did manage to get some good assistance also in the #genivi channel17:51
robtaylorgenii: cool :)17:51
robtaylorgenii: course, if you're using an autmotive SoC, they usually have a can interface built in (imx6, jacinto, jetson, etc)17:52
geniirobtaylor: This is what we need to sort out on our end right now17:52
*** ssam2 [] has quit [Quit: Leaving]18:15
*** flatmush [] has quit [Ping timeout: 246 seconds]18:25
*** flatmush [] has joined #baserock18:28
*** flatmush [] has quit [Ping timeout: 246 seconds]18:32
*** flatmush [] has joined #baserock18:33
*** flatmush [] has quit [Ping timeout: 260 seconds]18:40
*** flatmush [] has joined #baserock18:44
*** flatmush [] has quit [Ping timeout: 264 seconds]18:49
*** flatmush [] has joined #baserock18:49
*** flatmush [] has quit [Ping timeout: 245 seconds]18:54
*** flatmush [] has joined #baserock18:55
*** flatmush [] has quit [Ping timeout: 246 seconds]19:00
*** flatmush [] has joined #baserock19:11
*** flatmush [] has quit [Ping timeout: 250 seconds]19:15
*** flatmush [] has joined #baserock19:26
*** De|ta [~arc@] has quit [Quit: leaving]19:28
*** flatmush [] has quit [Ping timeout: 240 seconds]19:32
*** De|ta [~arc@] has joined #baserock19:37
richard_mawpaulsherwood, persia: Kinnison, liw-orc and I discussed this previously. The rough idea we had was to add "strip-commands" to chunk morphologies which operate on "$DESTDIR" only, and change the build-system defaults to include something appropriate that splits the binaries' debug symbols out, replacing it with a debug-link entry, and putting the debug symbol object files somewhere that chunk splitting will notice19:39
richard_mawcan't stay to chat about this tonight though, I only logged in to see what irssi looks like with cool-old-term19:40
*** flatmush [] has joined #baserock19:40
*** flatmush [] has quit [Ping timeout: 260 seconds]19:45
*** flatmush [] has joined #baserock19:51
*** flatmush [] has quit [Ping timeout: 255 seconds]20:00
*** flatmush [] has joined #baserock20:05
*** flatmush [] has quit [Ping timeout: 245 seconds]20:09
*** flatmush [] has joined #baserock20:10
*** flatmush [] has quit [Ping timeout: 250 seconds]20:25
*** flatmush [] has joined #baserock20:28
*** flatmush [] has quit [Ping timeout: 244 seconds]20:32
*** flatmush [] has joined #baserock20:36
*** flatmush [] has quit [Ping timeout: 240 seconds]20:44
*** flatmush [] has joined #baserock20:46
*** flatmush [] has quit [Ping timeout: 244 seconds]20:51
*** flatmush [] has joined #baserock20:52
*** flatmush [] has quit [Ping timeout: 240 seconds]20:57
*** flatmush [] has joined #baserock21:04
*** flatmush [] has quit [Ping timeout: 260 seconds]21:12
*** flatmush [] has joined #baserock21:42
*** flatmush [] has quit [Ping timeout: 240 seconds]21:47
*** flatmush [] has joined #baserock21:50
*** flatmush [] has quit [Ping timeout: 240 seconds]21:55
*** flatmush [] has joined #baserock21:56
*** flatmush [] has quit [Ping timeout: 245 seconds]22:01
*** flatmush [] has joined #baserock22:07
*** genii [~quassel@ubuntu/member/genii] has quit [Remote host closed the connection]22:11
*** flatmush [] has quit [Ping timeout: 244 seconds]22:12
*** flatmush [] has joined #baserock22:15
*** flatmush [] has quit [Ping timeout: 240 seconds]22:27
*** flatmush [] has joined #baserock22:33
*** flatmush [] has quit [Ping timeout: 240 seconds]22:44
*** flatmush [] has joined #baserock22:44
*** flatmush [] has quit [Ping timeout: 260 seconds]22:59
*** flatmush [] has joined #baserock22:59
*** flatmush [] has quit [Ping timeout: 244 seconds]23:05
*** flatmush [] has joined #baserock23:05
*** flatmush [] has quit [Ping timeout: 260 seconds]23:11
*** flatmush [] has joined #baserock23:31
*** flatmush [] has quit [Ping timeout: 240 seconds]23:48
*** flatmush [] has joined #baserock23:53

Generated by 2.15.3 by Marius Gedminas - find it at!