**** BEGIN LOGGING AT Thu Aug 23 03:00:00 2012 Aug 23 07:02:04 I will post subversion 1.7.6 to oe-core later today Aug 23 07:24:38 sent Aug 23 07:27:57 hrw: why do you name that file subversion-1.7.6_mod_dontdothat_svnserve_only.patch ? Aug 23 07:28:16 hrw: without subversion-1.7.6_ it would be better for future subversion upgrades Aug 23 07:28:38 or put it in subversion-1.7.6 subdirectory if you find it so version specific Aug 23 07:29:30 hrw: and it's also missing Upstream-Status field Aug 23 07:30:29 JaMa: its upstream name Aug 23 07:30:45 I mean: author named it like that Aug 23 07:54:33 i deleted a recipe workdir, and now bitbake -c clean fails, and when i try to rebuild this recipe the workdir is empty and it fails on the LIC_FILES_CHKSUM, is it fixable or do i need to rebuild everything ? :/ Aug 23 07:54:43 it looks like the tar.gz is not unpacked Aug 23 07:58:17 let's bump the PR... maybe it will fix it Aug 23 08:13:28 afournier: should be totally fixable Aug 23 08:14:02 afournier: when you say -c clean fails what actually happens? Aug 23 08:14:38 afournier: also, try -c fetch or something Aug 23 08:15:23 ok Aug 23 08:15:32 just building the recipe as normal will already do that, that won't help Aug 23 08:15:56 it cannot rm a directory Aug 23 08:17:07 afournier: could you just pastebin the actual error? Aug 23 08:17:25 yes in a moment, parsing the recipies takes time..; Aug 23 08:18:52 http://pastebin.com/4vBujRp4 Aug 23 08:19:46 OSError: [Errno 39] Directory not empty: '/home/alex/NAS/oe/tmp-eglibc/work/i586-oe-linux/asterisk-1.6.2.14-r1/temp' Aug 23 08:23:31 when i try to -c fetch is does not do anything Aug 23 08:24:05 it looks like bitbake has a state somewhere saying it's already there, but as i deleted it manually, it's not Aug 23 08:24:12 NOTE: Tasks Summary: Attempted 1 tasks of which 1 didn't need to be rerun and all succeeded. Aug 23 08:45:56 afournier: which version are you using? Aug 23 08:46:04 version of bitbake/oe ? Aug 23 08:46:09 oe-core Aug 23 08:46:15 oe-core denzil Aug 23 08:46:31 ok, that issue has been fixed in master; perhaps we need to backport the fix Aug 23 08:47:03 i don't know if using denzil is a good idea anyway... what do you think ? Aug 23 08:47:11 that would be nice :) Aug 23 08:52:12 it's the stable branch, it should be reasonable to use and receives bugfixes where necessary Aug 23 08:52:27 yes Aug 23 08:52:32 if you're comfortable on the bleeding edge you can use master, but be prepared for the occasional breakage... Aug 23 08:52:40 :/ Aug 23 08:52:47 i like the stable version concept :) Aug 23 08:53:07 i am trying to fix it manually but i cannot find where oe.path.remove is defined Aug 23 08:53:39 afournier: just cherry-pick 61ed74a72524b7f18fe7e48d5ae7f1f6b5756919 Aug 23 08:53:45 er wait, wrong rev Aug 23 08:53:54 git cherry-pick eef9d6c2d52f5264a6e7a9d882f8323f9793fd7f Aug 23 08:54:00 ok Aug 23 08:56:57 it worked Aug 23 08:57:09 you can backport it ;) Aug 23 09:00:01 goodmorning people! Aug 23 09:02:41 good morning silvio Aug 23 09:40:41 is pppd a native feature? Aug 23 10:01:33 anyone here good at remote controlling an openembedded session? Aug 23 11:49:21 RP: hi Aug 23 11:49:40 RP: I would like to patch svn fetcher to run 'svn cleanup' before fetching. Aug 23 11:49:53 hrw: hi Aug 23 11:49:55 svn: E155004: '/mnt/ci_build/workspace/openembedded-armhf-linaro-sdk/builddir/build/downloads/svn/www.eglibc.org/svn/branches/eglibc-2_16' is already locked. Aug 23 11:49:59 svn: run 'svn cleanup' to remove locks (type 'svn help cleanup' for details) Aug 23 11:50:13 hrw: isn't that potentially dangerous? Aug 23 11:51:06 RP: iirc user is not expected to play with DL_DIR contents Aug 23 11:51:36 hrw: no, but how do we tell the difference between some kind of real lock issue and the user having interrupted a download? Aug 23 11:51:44 RP: and my situation is a result of failed checkout due to disk error. but I do not have a way to login there to fix it Aug 23 11:52:02 RP: hm, right Aug 23 11:52:23 will add command to get rid of directory before build instead Aug 23 11:53:36 hrw: I think that is safer Aug 23 11:53:55 hrw: a bitbake eglibc -c cleanall will wipe it out FWIW Aug 23 11:55:33 thanks Aug 23 12:41:46 anyone know if there's a reason the gst-omapfb-recipe hasn't made it over to oe-core? Aug 23 12:49:19 tasslehoff: cause it is omap related? Aug 23 12:50:52 hrw: I guess he means in the OE-Core based layer landscape... Aug 23 12:51:56 tasslehoff: maybe worth asking the meta-ti folks Aug 23 12:56:06 denix: ^ do you know about this? Aug 23 13:03:56 RP: is it possible to get gnu-config 2012-08-14 in oe-core? I have local copy for aarch64 work Aug 23 13:21:36 someone have any suggest about guideline to make how recipes? Aug 23 13:21:58 someone have any suggest about guideline to make my hown* recipes? Aug 23 13:25:43 hrw: send a patch Aug 23 13:26:56 bluelightning: yeah, need to get my terminology straight :) Aug 23 13:28:25 silvio_: have a look at the yocto project development manual section on "adding a package" Aug 23 13:28:33 bluelightning: I made it compile in my OE-Core based layer landscape, so I'll send a patch if it works and is wanted Aug 23 13:30:52 bluelightning: thanks Aug 23 13:52:40 the interactive option was removed from bitbake version 1.12.0? Aug 23 14:11:34 silvio_: you mean the bitbake shell? Aug 23 14:11:53 I think so, yes Aug 23 14:12:11 RP: ok Aug 23 14:23:36 silvio_: that sounds about right. it was when the ui/server got split up Aug 23 14:28:43 i mean the -i option for the interactive bitbake, now remain only devshell? Aug 23 14:30:12 -i does not exist anymore in any form. Aug 23 14:30:18 devshell is something else entirely Aug 23 14:49:45 why it was removed? Aug 23 14:50:01 as i just said, it happened when we split the ui from the server Aug 23 14:50:20 internally it was restructured, and no one ever reimplemented the interactive mode as a proper bitbake UI Aug 23 14:51:35 so is not for technical reason, thanks for the answer Aug 23 14:51:52 I guess it could be done, but one of the reasons for having it was the extremely long parse times which has been greatly improved since those days Aug 23 14:58:55 i still like hte idea, but i want it more as a commandline than an interactive mode Aug 23 14:59:01 that is, i want a subcommand based commandline interface Aug 23 14:59:07 bitbake show that, bitbake whatdepends this, .. Aug 23 14:59:09 heh Aug 23 14:59:26 that sort of thing is basically interactive mode, just with a differen tinterface Aug 23 15:02:58 I never saw -i so I'm courious, i found it on the OE user manual Aug 23 15:03:29 how do you visualize .dot dependency trees (generated with bitbake -g) as they are the kind of huge files graphviz/dotty does not like... Aug 23 15:03:58 kergoth, you mean that same things can be done with -c option Aug 23 15:04:24 hm? Aug 23 15:04:51 afournier: *shrug*, i've always used graphviz in the past, but it's so massive the visualization is of limited use Aug 23 15:05:04 afournier: if you haven't seen bitbake's depexp ui, that may be worth looking at, depending on your needs Aug 23 15:05:07 bitbake -g -u depexp foo Aug 23 15:05:20 it's a little gtk ui for exploring the dependency graph Aug 23 15:05:24 not visually, though :) Aug 23 15:05:41 cool Aug 23 15:05:47 i did not know that Aug 23 15:05:57 i try one time to use graphviz, but the graph resulting is too complex and big to be manageded in useful way Aug 23 15:06:18 bah yes Aug 23 15:06:21 that was my point Aug 23 15:07:17 afournier: xdot is very good I've found Aug 23 15:07:23 ah, i thought you menat graphviz itself had a problem with it, not us making sense of its output :) Aug 23 15:07:34 http://code.google.com/p/jrfonseca/wiki/XDot Aug 23 15:07:54 graphiz itself sucks a lot of memory when graph are as big as bitbake generated ones Aug 23 15:08:03 i'll check XDot tonight Aug 23 15:08:05 right, xdot helps a little with that Aug 23 15:09:08 I tried a bunch of graph viewer applications, xdot seemed the best of all of them for larger graphs; unfortunately some of our graphs are still too large even for that Aug 23 15:09:20 i love : bitbake -g -u depexp foo Aug 23 15:09:21 :) Aug 23 15:09:31 yes, depexp is good Aug 23 15:09:36 we need to find a etter way to show *how* something got pulled in Aug 23 15:09:47 the graph shows that this sucks in that, but not if it was DEPENDS, RDEPENDS, whateer Aug 23 15:09:55 and if a PROVIDES was involved, it misses theindirection Aug 23 15:10:16 cool: bitbake -g -u depexp foo Aug 23 15:10:29 right, yes, that's the reason people usually want to look at the graphs and it's not as useful for that as it could be Aug 23 15:10:30 figuring out why bitbake has decided that both providers of something need to be built is a giant pain in the ass :) Aug 23 15:23:17 holy hell, i forgot how long it takes to build qt4 Aug 23 15:23:23 this is insane Aug 23 15:24:13 no worse than gtk or x11 or any other GUI thing is seems. Aug 23 15:24:24 i think it's worse Aug 23 15:24:26 damn c++ :) Aug 23 15:24:48 ya, forgot about that. Aug 23 15:26:57 I guess I just kick it off and walk away... Aug 23 15:27:08 indeed Aug 23 15:27:14 watched pot and all Aug 23 15:27:17 heh Aug 23 15:28:00 sstate +1 Aug 23 15:28:16 indeed Aug 23 15:28:23 it's still pretty brittle though Aug 23 15:30:22 bluelightning: any news about rootfs pollution for cpio's? I'd send an RFC about that. It looks like 2 oe-core classes are part of the problem... Aug 23 15:30:49 ant_work: I haven't looked any further into it yet I'm afraid Aug 23 15:37:09 QT on the BeagleBoard but you have to "sit it and forget it" for ~12-18hrs," Aug 23 15:38:24 haha Aug 23 15:40:08 kergoth, try to use amazon ec2 or similar to make it faster ;) Aug 23 15:40:43 you'd think this 8 core vm would be enough, but.. Aug 23 15:40:47 i'm in the midst of a world build Aug 23 15:40:51 cause i'm crazy like that Aug 23 15:40:59 * kergoth hasn't done one of these in years Aug 23 15:42:22 I hope Ec2 use more tha 8 core, but i courios of know in how many time need to built qt4 Aug 23 15:42:52 i'd hope so too, never used it Aug 23 15:46:41 playing with AUFS (overlaying SSD or tmpFS) to speed up builds Aug 23 15:47:43 ka6sox, ? Aug 23 15:47:56 Solid State Disk? Aug 23 15:47:59 yes Aug 23 15:48:28 i have no ssd Aug 23 15:48:48 the best is tmpFS in RAM (if you have it) Aug 23 15:49:20 but i need 60-80gb of ram? Aug 23 15:49:35 depends on the build Aug 23 15:49:46 i mean to build an image Aug 23 15:50:22 60-80gb to build an image? must be a big image :) Aug 23 15:51:33 usually is this angstrom-lxde-image Aug 23 15:51:37 ah Aug 23 15:52:07 is big enough :) Aug 23 15:52:10 ? Aug 23 15:52:25 silvio_, do you use rm_work? Aug 23 15:52:46 isn't rm_work still dangerous due to the interactions with sstate? or is it stable now? Aug 23 15:52:54 i don't use it and just wipe tmpdir regularly :) Aug 23 15:53:16 yup. Aug 23 15:53:39 with tmpFS I just reinit. Aug 23 15:53:46 ka6sox: no, I think no, also because the work dir is full Aug 23 15:54:11 heh, nice Aug 23 15:54:12 what is size of work dir after a build? Aug 23 15:54:25 i should put together a nice fast high-ram build machine at some point, for home Aug 23 15:54:37 wait... i check but i need some minutes :) Aug 23 15:54:51 hehe Aug 23 15:54:55 du on a build dir does take a long time Aug 23 15:55:07 pidge did experiments and now all I'm spec'ing is 256GB RAM minimum builders. Aug 23 15:55:28 nice Aug 23 15:55:51 wow how u put all this ram in a pc? Aug 23 15:56:02 silvio_, server class hardware. Aug 23 15:56:23 ka6sox: I'd like to see your buildstats if you get a chance, in order to compare Aug 23 15:56:24 not ec2 but quite near Aug 23 15:57:01 pidge, will do, Aug 23 15:57:15 i know that tar.bz2 of output is about 17-18gb Aug 23 15:57:51 so to improve build time, some (not crazy suggest) ? Aug 23 15:58:20 buy RAM, if you can't afford it...buy SSD...if you can't afford it, build a RAID10 Aug 23 15:59:28 RAM is aproblem i need special hardware, probably buy a 256 Gb ssd is the better solution or I m wrong? Aug 23 15:59:54 or raid 0 of 2 ssd :P Aug 23 16:00:02 for serious builders HD replacement every 6 months, SSD in a year (although a developer machine would be longer) Aug 23 16:00:42 You found decrease and fault on hdd over the 6 month? Aug 23 16:01:08 silvio_, I currently have a RAID0 of 2 SSD's that have been building for a year...and no faults yet... Aug 23 16:01:14 nice Aug 23 16:01:41 silvio_, for the autobuilders every 6 months I was replacing 25~50% of the HDs Aug 23 16:02:19 since I used SSD haven't replaced a HD Aug 23 16:02:26 (over a year) Aug 23 16:02:37 good Aug 23 16:03:10 for backups what you suggest to use? I mean something that is practise and fast Aug 23 16:03:51 I just use a eSATA or USB3 external drive (for dev machine) Aug 23 16:04:13 Autobuilders the stuff I care about fits on a thumb drive. Aug 23 16:06:15 nothing of special so, with thumb drive u mean a usb key?? Aug 23 16:06:25 yes Aug 23 16:06:41 ok thanks Aug 23 16:07:00 it can all be rebuilt/redownloaded. Aug 23 16:07:17 you will lose sstate (unless you have a big key) Aug 23 16:08:21 now, first thing before next(s) build I want to do is create a local (i mean in LAN) mirror of the OE git Aug 23 16:09:19 * ka6sox thought every up-to-date git repo (local) was by definition a mirror... Aug 23 16:09:23 because i notice that many waste time is in the fetching of packages Aug 23 16:09:40 you mean downloads dir? Aug 23 16:10:02 yeah sorry for my bad english Aug 23 16:10:20 I waste time in fetch of packages Aug 23 16:10:44 last time I looked source.oe.org was about 40GB.... Aug 23 16:10:56 sources.oe.org Aug 23 16:11:24 please limit the transfer to 10Mb/sec Aug 23 16:11:27 now the domain is in sold Aug 23 16:12:48 10Mb/sec lol idf i reach the 180kb/sec I was lucky, sometimes my line works at 5kb/s Aug 23 16:13:03 media is about 40-50 kb/s Aug 23 16:13:35 okay...let me know what week it is done... Aug 23 16:14:24 silvio@ubuntu-EP400:~/oe/output$ du -hs work/ Aug 23 16:14:24 57G work/ Aug 23 16:14:51 so 64GB of RAM should do nicely Aug 23 16:15:04 I have to go, see you tomorrow bb people Aug 23 16:15:35 yes but on EP400 is not feasible Aug 23 17:38:40 hello Aug 23 18:22:52 can you pass -std=c++0x to bitbake for building boost? or should it still be abi compatible? Aug 23 18:49:05 it is possible to execute a command thru bitbake ? Aug 23 18:49:25 like bitbake -exec ls Aug 23 18:49:36 yes a shell command Aug 23 19:03:27 tworkin: i've added that to a couple of packages... are you talking global? Aug 23 19:04:48 mr_science: i'm getting undefined reference for boost::thread::thread(). I Believe it is because I'm compiling my sources with -std=c++0x and that bitbake compiled boost without that. Aug 23 19:05:11 mr_science: -std differences breaks ABI compatibility, right? Aug 23 19:05:43 it changes the way g++ interprets the code, yes Aug 23 19:06:11 i usually add that in extra_oeconf, something like CFLAGS="${CFLAGS} -std=gnu99" Aug 23 19:08:16 you need them to match, ie, either change your compile flags or rebuild the affected packages Aug 23 19:08:37 i'd like to rebuild. i dont see an extra_oeconf though. Aug 23 19:09:00 arago/conf/local.conf, arago-bitbake/bitbake.conf Aug 23 19:09:10 I guess this is TI's not quite vanilla version of OE Aug 23 19:10:56 no, the extra_oeconf goes in the package recipe Aug 23 19:11:32 or extra_oemake i guess Aug 23 19:12:00 eg, EXTRA_OEMAKE = "CFLAGS="${CFLAGS} -std=gnu99" Aug 23 19:12:18 I have arago-oe-dev/recipes/boost/boost_1.45.0.bb. next to this file? Aug 23 19:13:37 oh IN the recipe file Aug 23 19:13:41 add the EXTRA... line to the boost recipe, if that's the one you want to rebuild with a new -std= option Aug 23 19:14:27 with the nested quotes? Aug 23 21:32:39 exit Aug 23 21:32:46 time to go... Aug 23 21:33:00 stupid window focus Aug 24 02:15:33 is there a way to run bitbake from the command line? Aug 24 02:16:21 yes Aug 24 02:16:28 how do you do it otherwise? Aug 24 02:16:36 Is there a gui? Aug 24 02:16:44 I'M SO ALONE AND CONFUSED! Aug 24 02:17:14 lol Aug 24 02:17:35 since when is there a gui? Aug 24 02:17:43 MOMMA SAID YOU PEOPLE WERE BAD BAD MEN (women? OHMAGERDYERSH!) ! Aug 24 02:17:51 ccssnet I had no idea one existed Aug 24 02:17:55 mIKEjONES yes Aug 24 02:17:58 mIKEjONES there is Aug 24 02:18:07 mIKEjONES it's called 'bitbake' Aug 24 02:18:19 you use 'bitbake' to bitbake things Aug 24 02:18:29 alexhairyman: your nick name on irc is humourus Aug 24 02:18:33 :) Aug 24 02:18:37 by way of science (which is really thinly disguised magic) Aug 24 02:19:08 ccssnet I'm not really that hairy... in the places /not/ covered by swimwear Aug 24 02:19:08 completely off topic and random... Aug 24 02:19:16 but it sounds like my name Aug 24 02:19:20 Alex Herrmann Aug 24 02:19:20 so i have finally upgraded to hdmi display. it sucks Aug 24 02:19:29 ccssnet lol Aug 24 02:19:33 yes... DVI! Aug 24 02:19:37 yup Aug 24 02:19:44 WHOOP WHOOP WHOOPI! Aug 24 02:21:08 mIKEjONES still need helper pelper? Aug 24 02:21:17 why the hell did I say that? Aug 24 02:21:21 that was stupid Aug 24 02:22:33 well, technically i am using DVI to hdmi cable to my lcd screen (32" hdmi tv) from a DVI 4770 vid card. Aug 24 02:22:40 either this tv sucks. or hdmi does Aug 24 02:22:42 lol Aug 24 02:23:22 lol Aug 24 02:23:32 FUCK VIDEO! Aug 24 02:23:37 I CODE BY SOUND ONLY! Aug 24 02:23:41 lol Aug 24 02:23:58 THROUGH A MIX OF BADASSERY AND HELLEN KELLER SMARTNESS! Aug 24 02:24:04 except she was deaf Aug 24 02:24:05 right? Aug 24 02:24:07 oop Aug 24 02:24:10 *oops Aug 24 02:24:17 hmmm Aug 24 02:24:24 *oops = "ISCREWEDUPP"; Aug 24 02:26:48 pff Aug 24 02:27:02 alexhairyman: I can't seem to run bitbake from the command line though :( Aug 24 02:27:19 a lot of the things I run for some reason want X to be running as well Aug 24 02:27:54 and because I'm running this on a server I don't have X so bitbake crashes loading gnome libraries Aug 24 02:28:11 I remember there being a way of running bitbake from the command line, has this changed? Aug 24 02:28:17 * ccssnet downloads a new copy of bitbake and open embedded to see what hes missed Aug 24 02:48:43 needing screen, gnome-terminal, or xterm is such a fucking stupid requirement for bitbake Aug 24 02:48:53 keep up the shitty work guys Aug 24 02:49:59 i run our builds on a "server" with minimal xlibs/vnc Aug 24 02:50:26 huh? Aug 24 02:50:30 some of the doc-book stuff and some X headers are required Aug 24 02:50:34 a terminal is only needed to run devshell Aug 24 02:50:39 if you don't want to install one, don't use devshell Aug 24 02:51:21 i'm just talking bitbake at a bash prompt Aug 24 02:51:27 so am I. Aug 24 02:51:36 mIKEjONES: set PATCHRESOLVE = "noop" Aug 24 02:51:51 mIKEjONES: this will make sure it never tries to automatically spawn a terminal to resolve patch rejects if a patch fails to apply Aug 24 02:51:53 i have some shell scripts to get jenkins going... Aug 24 02:51:59 mIKEjONES: taht way spawning a terminal will only ever be manual Aug 24 02:52:05 mIKEjONES: (this should really be default, imo) Aug 24 02:53:27 there's no X server running in this case... Aug 24 02:55:36 mIKEjONES: bitbake runs from the commandline, always, unless you run 'hob' explicitly, or some other non-standard UI. the only time it tries to run a terminal automatically is for patch rejects, otherwise it's explicit (bitbake -c devshell foo) **** ENDING LOGGING AT Fri Aug 24 03:00:04 2012