IRC logs for #baserock for Friday, 2015-08-28

*** edcragg has quit IRC00:51
*** straycat_ has joined #baserock02:42
*** straycat has quit IRC02:46
*** zoli__ has joined #baserock04:47
*** zoli__ has quit IRC05:27
*** zoli__ has joined #baserock05:28
*** paulw has joined #baserock06:18
*** zoli__ has quit IRC06:25
*** zoli__ has joined #baserock06:44
*** zoli__ has quit IRC06:52
*** zoli__ has joined #baserock07:34
*** bashrc has joined #baserock08:07
*** jonathanmaw has joined #baserock08:13
*** ssam2 has joined #baserock08:47
*** ChanServ sets mode: +v ssam208:47
ssam2I want to merge (migrate to definitions version 6 in definitions) today, unless anyone objects08:53
*** edcragg has joined #baserock09:00
*** flatmush has quit IRC09:14
*** flatmush has joined #baserock09:14
*** lmackenzie_2015 has joined #baserock09:22
*** franred has joined #baserock09:32
*** ssam2 has quit IRC09:53
*** mdunford has quit IRC10:05
*** ssam2 has joined #baserock10:08
*** ChanServ sets mode: +v ssam210:08
*** paulw has quit IRC10:27
*** lachlanmackenzie has quit IRC10:27
*** persia has quit IRC10:27
*** _longines has quit IRC10:27
*** paulw has joined #baserock10:29
*** lachlanmackenzie has joined #baserock10:29
*** persia has joined #baserock10:29
*** _longines has joined #baserock10:29
*** paulw has quit IRC10:30
*** lachlanmackenzie has quit IRC10:30
*** persia has quit IRC10:30
*** _longines has quit IRC10:30
*** mdunford has joined #baserock10:31
*** paulw has joined #baserock10:32
*** lachlanmackenzie has joined #baserock10:32
*** persia has joined #baserock10:32
*** _longines has joined #baserock10:32
*** ssam2 has quit IRC10:49
*** ssam2 has joined #baserock11:06
*** ChanServ sets mode: +v ssam211:06
*** paulw has quit IRC11:29
*** paulw has joined #baserock11:34
*** paulw has quit IRC12:14
*** paulw has joined #baserock12:26
*** paulw has quit IRC12:27
*** paulw has joined #baserock12:27
*** wschaller has joined #baserock13:22
*** zoli__ has quit IRC13:29
*** jonathanmaw has quit IRC13:35
rjekssam2: The old adage is that C++ compilers work by sending your source into the future to be compiled by Skynet.  The computations required for the transmission require lots of RAM.13:44
robtaylorrjek: indeed, its a little know fact that compiling C++ is AI-Complete13:48
*** paulw has quit IRC13:55
*** mdunford has quit IRC13:58
*** paulw has joined #baserock14:00
*** paulw has quit IRC14:02
*** mdunford has joined #baserock14:13
paulsherwoodi've asked this before, but i can't remember the answer... can i build armv7 natively on aarch64 (eg moonshot?)14:18
rjekIt's complex.14:18
radiofreei have a mustang on my desk if you'd like to try it14:18
robtaylorrjek: can you boot an armv7 rootfs on one?14:18
rjekYou can ask GCC to generate 32 bit ARM code, and it will link.  There is no support for interworking between 64 and 32 bit instruction sets like there is between 32 and 16.14:18
radiofreebooting armv7 rootfs would probably be the easiest thing14:19
rjekARMv8 does not /require/ a CPU to contain support for the 32 bit instruction set.14:19
robtaylorrjek: but they all do..14:19
robtaylor(so far)14:19
rjekI have not conducted a survey to confirm that, but it seems likely doesn't it? :)14:19
robtaylorpaulsherwood: so, just boot an armv7 rootfs :)14:20
paulsherwoodrewind... to be more precise... i want to run ybd on moonshot. moonshot runs ubuntu, little endian. would the armv7 definitions build or not?14:20
rjekI don't think morph would enjoy trying to build a 32 bit ARMv7 world while running on a 64 bit ARMv8 would, though, so easiest solution is probably to boot the ARMv7 world and build there14:20
* paulsherwood doens't have ability to boot his target machine into anything other than ubuntu, it seems14:20
rjekpaulsherwood: If you can control what parameters get passed to GCC (to set -march=armv7 and similar) it might14:21
rjekpaulsherwood: Unpack a 32 bit world, and chroot?14:21
paulsherwoodrjek:  you assume i understand more than i do14:21
radiofreei can test that out if you want14:21
paulsherwooddo you have access to a moonshot?14:22
rjekNo, but he has a Mustang.14:22
radiofreei have a mustang next to me14:22
paulsherwoodah! :)14:22
rjekpaulsherwood: You can use a tool called "debootstrap" to build a Debian or Ubuntu chroot system which you can then enter, or you can somehow arrange for GCC to always be invoked with -march=armv7.  I suspect compiling "natively" via a chroot would probably be easiest/most reliable14:23
rjekActually, I've never tried debootstrapping Ubuntu; I have no idea if that works.14:23
radiofreeif this works i suppose i could leave it populating armv7 caches on cache.br14:24
rjekdebootstrap --arch=armhf --include=build-essential jessie my-jessie-chroot && time-passes && chroot my-jessie-chroot14:24
radiofreei can imagine it's quite fast14:25
rjekOh you need --foreign in there too, and after chrooting run /debootstrap/debootstrap --second-stage to complete the install14:25
*** rdale has joined #baserock14:28
radiofreedoes baserock publish armv7 rootfs images?14:31
radiofreeerm.. archives14:31
rjekradiofree: Only x86 for chroot tarballs IIRC14:34
rjekBut you should be able to download the ARMv7 raw disk image and use kpartx to loopback mount it14:35
paulsherwoodthis all sounds lovely, but it might as well be in a foreign language to me :)14:36
radiofreerjek: that's exactly what i did14:37
radiofree"mount img.img foo/" is good enough14:37
rjekAh, no partition table?  Bonus.14:37
bwhsince I just did it today, this is *all* the steps for a foreign debootstrap:
paulsherwoodbwh: great!14:48
bwhvmdebootstrap is a wrapper that builds a disk image complete with kernel and boot loader, but I don't know if it does cross-building or supports boot loaders other than grub14:48
* rjek moans a bit at bwh's CSS.14:49
bwhrjek: Not me, I just read and (roughly) followed that14:49
rjekOh, right, sorry, misinterpreted what you said14:49
bwhunless you're talking about CSS on which is, ahem, minimal14:50
rjekFor some reason I read night-shade as decadent-place14:50
radiofreeok this mustang is building a jetson armv7 image in a chroot14:52
rjekDo you have a Jetson to try the result on? :)14:54
radiofreei do14:54
radiofreeif it works maybe we can setup an arm mason on it?14:54
rjekOr you could chroot into the chroot from inside your chroot and try building another chroot there.14:54
radiofreealthough i had to use ybd because i'd have to hack morph not to get the arch from uname14:54
paulsherwoodw00t! :)14:54
rjekradiofree: Does "linux32" do on ARMv8 what it does on x86?14:55
rjek(What does uname return on the Mustang?)14:55
radiofreeooh right, i could have done that14:55
paulsherwoodradiofree: any chance yo ucould paste the log somewhere, if/when it inishes?14:55
rjeksetarch seems to be the general thing14:55
radiofreerjek: ah no i couldn't do that14:56
radiofreelinux32 uname -a14:56
radiofreeLinux baserock-mustang 4.1.0-rc7 #7 SMP PREEMPT Mon Jun 8 17:32:42 BST 2015 armv8l GNU/Linux14:56
rjeksetarch armv7l uname -a ?14:56
radiofreeuname -a14:56
radiofreeLinux baserock-mustang 4.1.0-rc7 #7 SMP PREEMPT Mon Jun 8 17:32:42 BST 2015 aarch64 GNU/Linux14:56
radiofreeapparently armv7l is an "Unrecognized architecture"14:57
radiofreeanyway, you can just pass the arch to ybd14:57
* rjek nods14:57
rjekarmv7l was a guess for what it returns on a real ARMv7 box; it may be wrong.14:58
radiofreepaulsherwood: the log is stdout right?14:58
paulsherwoodradiofree: yup. i tend to tee it14:58
rjekAnyway, uname is full of lies at the best of times, perhaps morph shouldn't be using it14:58
radiofreepaulsherwood: can you give me the command please14:58
rjekybd ... | tee logfile.txt14:58
paulsherwoodyup. for example ../ybd/ clusters/ci.morph | tee some-memorable-name.log14:59
* radiofree adds sys.stdout.flush() to the log() function15:01
paulsherwoodpatch please? :)15:02
radiofreeybd fetches everything before building?15:11
* radiofree doesn't think he's actually built anything yet15:11
paulsherwoodradiofree: do you have gits already on that box?15:12
radiofreenot in the chroot15:16
paulsherwoodah, ok15:16
paulsherwooddo you have enough storage in the chroot for this activity?15:16
radiofreei have 28G free15:17
radiofree36G free now15:17
paulsherwoodnot enough15:18
paulsherwoodfor one system maybe15:18
radiofreethe ssd is only 120G15:19
radiofreei have built a devel system on the mustang before, so 44G is probably enough15:20
radiofreeno wait, it's 64G15:20
radiofree(the ssd)15:20
paulsherwoodwell it will probably run out of space15:20
radiofreereally? it didn't run out space when i built a devel system previously15:21
paulsherwoodoh? maybe not then :)15:21
paulsherwoodradiofree: thanks for the patch!15:24
ssam2I just submitted my first ever patch to :-)15:46
ssam2only as a "proof of concept", though, it's still a really ugly hack15:46
*** rdale has quit IRC15:49
paulsherwoodssam2: cool! is there any demo or something to see?15:50
radiofreeit failed! mount(/, MS_PRIVATE | MS_REC): Invalid argument15:55
radiofreei guess it's because i'm chrooting in a chroot?15:55
radiofreei'll use schroot? i know that works15:55
radiofreeanyone know where to get schroot source?16:02
radiofreeconfigure: error: Boost.Format (Boost C++ Libraries) is not installed, but is required by schroot16:04
radiofreewhat's the correct chroot foo for launching a baserock chroot16:04
radiofreei'd rather do this without baserock-chroot16:10
radiofreesince that needs schroot which would involve me building boost16:10
radiofreeknown issue!16:11
* radiofree tries the fix16:11
paulsherwoodradiofree: does your system have linux-user-chroot?16:11
paulsherwoodif not, adding it might be a soln16:12
paulsherwoodybd defaults to luc if it finds it, falls back to chroot16:12
radiofreeyes it has it16:13
paulsherwoodis it using it?16:13
paulsherwood(ie does your log show any message about rogue build?16:14
paulsherwoodif it mentions rogue build, it's not finding luc16:14
radiofreerestarting usually fixes things, i'll try that16:14
radiofree+ mount bind /dev16:17
radiofreeand proc16:18
paulsherwoodwhat was the fix?16:21
radiofree[17:17:41] <radiofree>
radiofree[17:17:47] <radiofree> + mount bind /dev16:21
radiofree[17:18:45] <radiofree> and proc16:21
paulsherwoodah, so should ybd do that?16:22
radiofreeno you do this outside of the chroot16:22
radiofreepaulsherwood: basically
paulsherwoodradiofree: aha, ok. thanks!16:25
radiofreeit's using all 8 cores to build16:25
radiofreewhich is nice16:25
radiofreefor my v7 chroot i just downloaded a jetson devel image16:27
radiofreemount -o jetson.img /mnt16:27
radiofreemkdir v7chroot16:27
paulsherwoodack, thanks16:28
radiofreecp -r /mnt/systems/factory/run/* /src/v7chroot/16:28
radiofreegcc fails16:30
radiofreegcc.c:(.rodata+0x1b0): undefined reference to `host_detect_local_cpu(int, char const**)'16:30
*** zoli__ has joined #baserock16:35
radiofreei added the commands to a script and ran it with /usr/bin/linux32 -B ./launch.sh16:54
*** bashrc has quit IRC17:17
*** zoli__ has quit IRC17:28
*** franred has quit IRC17:32
radiofreethis might actually work17:32
radiofree15-08-28 17:30:03 [14/280/282] [glibc] Elapsed time for assembly 00:24:17 ??17:32
*** lmackenzie_2015 has quit IRC17:51
*** ssam2 has quit IRC17:54
*** mdunford has quit IRC17:58
*** wschaller has quit IRC18:15
*** zoli__ has joined #baserock19:12
*** zoli__ has quit IRC19:23
*** zoli__ has joined #baserock19:25
*** edcragg has quit IRC19:30
*** zoli__ has quit IRC19:38
paulsherwoodradiofree: assembly is a bit of a movable feast... may or may not include build time of dependencies too21:14
paulsherwoodElapsed time for build is more interesting in general i think21:14
*** zoli__ has joined #baserock21:18
*** zoli__ has joined #baserock21:19
*** zoli__ has quit IRC21:23
radiofreepaulsherwood: that time is totally inaccurate23:38
* radiofree will check on the progress tomorrow 23:39
radiofreeA fast arm build server for caches will be pretty neat to have23:39

Generated by 2.15.3 by Marius Gedminas - find it at!