**** BEGIN LOGGING AT Tue May 03 02:59:58 2016 May 03 06:00:04 kergoth: agreed that helps May 03 12:35:17 * Crofton|work curses boost again May 03 12:35:37 cortexa9hf-neon-oe-linux-gnueabi/boost/latest:PACKAGES = boost-dbg boost-atomic May 03 12:35:38 boost-chrono boost-date-time boost-filesystem boost-graph boost-iostreams boost- May 03 12:35:41 log boost-program-options boost-random boost-regex boost-serialization boost-sig May 03 12:35:43 nals boost-system boost-timer boost-test boost-thread boost-dev boost-staticdev May 03 12:35:47 boost May 03 12:35:49 install boost package, get -dev package May 03 12:36:34 bug or feature? May 03 12:55:56 hm, bug May 03 12:56:28 are you sure boost is pulling in boost-dev? May 03 12:56:46 maybe there's a bad library dependency May 03 13:03:22 Crofton|work: testing here says that nothhing depends on boost-dev, and boost correctly depends on the library packages only May 03 13:03:51 hmm May 03 13:04:05 see th eline from buildhistory May 03 13:04:08 let me update May 03 13:04:37 that's just the list of packages May 03 13:04:44 der May 03 13:04:59 early here May 03 13:05:05 i'll invoice you this evening May 03 13:06:07 was reading recipe May 03 13:06:12 still something funny May 03 13:09:44 really need to write a bb tool to dump dependencies from a recipe May 03 13:57:47 hi RP: few months back someone had sent gclient fetcher support for bitbake, but that didn't get much attention/review. did you see it? do you have any concerns about it? we can update/resend if you want. May 03 14:02:23 ndec: the fetcher looked horrible, e.g. the hardcoded need_update() == True May 03 14:02:50 RP: ok, good. i mean at least there is feedback ;-) May 03 14:03:44 if you have more feedback let me know. i haven't looked at it closely, but i will too. May 03 14:03:55 ndec: it just looked like a pretty quick/dirty hack to me :/. Equally I've never used it so I know little about it May 03 14:04:27 ndec: no mirror tarball support, no revision/srcrev support, hardcoded to always hit the network etc May 03 14:04:33 my main concern was the lack of feedback. May 03 14:05:11 ndec: I guess I didn't really feel in a position to review it knowing nothing about gclient May 03 14:05:39 ok, thanks May 03 14:06:37 ndec: it basically looks like a wrapper which abstracts out our own SRC_URI, the quesiton is where would we expect the SCM metadata to be maintained May 03 14:06:52 and it does it badly with no mirror support etc May 03 14:59:01 ndec: perhaps its something that should live in a separate layer until they get it polished off, teh way the bitbake-npm one was May 03 14:59:02 heh May 03 14:59:36 i wasn't aware we had fetchers outside of bitbake. where is that layer? May 03 15:00:04 i've done it before, you basically just use python to inject the module into the fetchers list after importing it from your layer path May 03 15:00:10 did one like that at montavista May 03 15:00:15 here, i'll get you a link to the npm one May 03 15:00:31 ndec: https://github.com/imyller/bitbake-npm May 03 15:01:05 ehh. nice hack ;-) May 03 15:01:25 you can see the usage is pretty clean, so it wouldn't be too terrible to deal with until the merge May 03 15:01:50 yeah. indeed. May 03 15:02:29 in fact the main use case for it right now would be the chromium recipe.. so that can be isolated even in meta-browser, potentially. May 03 15:02:51 * kergoth nods May 03 15:43:57 Keyboard Interrupt, closing down... May 03 15:43:58 WARNING: Timeout while attempting to communicate with bitbake server WARNING: Timeout while attempting to communicate with bitbake server May 03 15:44:02 this still happens to me left and right May 03 15:44:09 RP: ^ known issue? May 03 16:04:04 kergoth: not really :/ May 03 16:04:12 k May 03 16:04:24 haven't had a chance to nail it down, i'll add to the todo May 03 16:04:56 ndec: I would really like fetchers to get cleaned up and into the core ideally May 03 16:05:15 ok. May 03 18:02:08 kergoth: we used to see the timeout a few times per day since we push out build farm pretty hard May 03 18:02:33 this timeout isn't due to lag, it's due to bitbake not shutting down correctly when i hit ^C May 03 18:02:38 note the KeyboardInterrupt May 03 18:03:08 I decided to stop using the PR server for the builds running in docker and now I only see the timeout to bitbake server on one overloaded machine. May 03 18:03:35 kergoth: ah.. ok ignore my comments then! :) May 03 18:04:33 I do find it frustrating that the build tool has a timeout... I'd prefer to move the timouts to the user since build times and system load can vary significantly. May 03 18:08:58 some sort of timeout is important in case there really is a loss of communication between the ui and server May 03 19:43:47 Hmm, should enhance the newappend/appendsrc commands to be able to put the append in the layer-specific area, for layers supporting that common pattern May 03 20:27:26 cbrake: kicked off a bitbake build on my mac earlier with docker for mac beta + cbrake/oe-build, with some small tweaks :) May 03 20:33:28 cbrake: i wanted to avoid hardcoding the source of the setup script inside of the container, so instead i mount the volume at COREBASE/.. rather than $PWD, and explicitly pass in the key variables into the container environment. so now i can source the setup script outside of the container and transparently run bitbake from there, so doesn't make the assumption about how the setup is done May 03 20:33:34 not perfect, though May 03 20:33:55 https://gist.github.com/kergoth/f5f8f5361f2c1ca7817c105cfb58931e May 03 20:56:47 kergoth: nice :-) -- I'll check it out ... May 03 20:57:37 http://termbin.com/u1j3 May 03 20:57:41 This is what I use May 03 20:57:43 cbrake: kergoth May 03 20:57:47 just a slightly different way of handling it. i want to make the env var passing a bit more intelligent, like i'd want MACHINE/DISTRO/etc to propogate, but id on't want all of BB_ENV_EXTRAWHITE, because some of that isn't going to funciton across the container boundary, i.e. SSH_AUTH_SOCK May 03 20:57:57 ah, nice May 03 20:58:19 i take it the otavio user exists in the container too? May 03 20:58:29 it'd be nice to remap the uids of outside vs inside May 03 20:58:54 kergoth: yes May 03 20:59:06 kergoth: I fixed it when generating my image and commited May 03 20:59:10 * kergoth nods May 03 20:59:18 i thought about doing that, i'm hitting a uid mismatch issue atm May 03 20:59:23 hmmmmmm... http://git.midipix.org/cgit.cgi/slibtool/tree/README May 03 20:59:30 kergoth: so I use it as a VM May 03 20:59:33 and keep logged in May 03 21:00:01 kergoth: yeah, I ended up wrapping bitbake as well, but not nearly as sophisticated as your example May 03 21:00:27 i thought about writing the env to a file and then reading it inside the container, might be safer than passing through variable parameter expansion in the commandline May 03 21:00:29 haven't decided May 03 21:00:31 also fighting UID mismatch at a customer install -- no ideas how to work around that one yet May 03 21:00:48 i know some container mechanisms support uid remapping for user containers May 03 21:00:51 no idea about docker, though May 03 21:00:57 but like nspawn, pflask, etc May 03 21:01:04 yeah, I've been digging but have not figured it out yet May 03 21:01:21 ended up creating another container tag with a user that matched their build machine May 03 21:01:30 as I use Arch, Docker saves me May 03 21:01:39 otavio: same here :-) May 03 21:02:17 * cbrake really likes Arch for workstation -- been upgrading machines here for 3 or so years -- can't think of any major issues I've had with hit May 03 21:02:28 https://github.com/docker/docker/pull/4572 seems interesting May 03 21:02:31 also https://github.com/docker/docker/issues/7198 has lots of hacks May 03 21:02:52 cbrake: i'm a big fan of arch too, though its been a while since i installed it, so my setup scripts for it are all out of date now, going to have to revert back to manual :) May 03 21:10:03 kergoth: slibtool is interesting although for linux you can surely go even more minimal (thinking of dolt here) May 03 21:10:10 indeed May 03 21:11:28 you should test slibtool May 03 21:39:33 armpit: just read over the notes from today's technical team meeting May 03 21:40:16 armpit: could i get more details on the layerindex issue? the email is a bit vague May 03 21:58:09 cbrake: rather ugly, but .. docker run --rm -it -v "$PWD:$PWD" -v "$COREBASE/..:$COREBASE/.." -v ~/.ssh:/home/build/.ssh -v "$XDG_CONFIG_HOME/git:/home/build/.config/git" -e "BUILDDIR=$BUILDDIR" -e "BB_ENV_EXTRAWHITE=$BB_ENV_EXTRAWHITE" -e "EXT_PATH=$PATH" -e "EXT_PWD=$PWD" -e "USER_ID=$(id -u)" -u root cbrake/oe-build /bin/bash -c 'usermod -u $USER_ID build && sudo -E -u build bash -c "cd \"$EXT_PWD\" && env \"PATH=$EXT_PATH\" bitbake \"$@\"" - "$@"' - -p May 03 21:58:15 need to look into docker's userns support, though **** ENDING LOGGING AT Wed May 04 02:59:58 2016