**** BEGIN LOGGING AT Wed Mar 06 02:59:58 2013 Mar 06 03:25:07 wow, it's nice to see the 3.8.0 kernel. nice work! Mar 06 05:42:12 hi Mar 06 05:50:27 hello ther Mar 06 06:59:55 Hi Mar 06 07:00:19 with SRC_URI i am trying to copy my source to work directory Mar 06 07:00:26 but its not copying there Mar 06 07:00:37 please can any one tell me what might be the problem Mar 06 07:00:54 i have done similer for some modules which are working fine Mar 06 07:21:56 bhaskar_: are you making a recipe? Mar 06 07:22:03 you should pastebin.org it Mar 06 08:28:27 good morning Mar 06 09:08:09 morning all Mar 06 09:18:40 hi bluelightning, all Mar 06 09:36:24 is there a way to clean stuff in the sstate cache for a package? bitbake -c clean just cleans tmp/work Mar 06 09:36:33 I want to force a complete rebuild Mar 06 09:38:47 HI everyone Mar 06 09:38:59 i have an issue regarding adding external toolchain in build webos Mar 06 09:39:12 this is the error m getting "Failure expanding variable sysroot_stage_all: ShellSyntaxError: LexToken(LPARENS,'(',0,0)" Mar 06 09:39:14 oh, looks like it's as simple as bitbake -f -c configure and then bitbake Mar 06 09:41:29 erbo, are you replying for sysroot error i posted? Mar 06 09:43:48 Rao: no, I was answering myself (from before you joined the channel) Mar 06 09:46:26 i have written bb file to include external toolchain Mar 06 10:04:06 erbo: you can also do bitbake -C configure to do those two steps in one Mar 06 10:04:48 bluelightning: nice, thanks Mar 06 10:04:56 though FYI if you have actually changed anything in the configure step that bitbake knows about it will rebuild it automatically the next time anyway Mar 06 10:05:14 e.g. if you changed EXTRA_OECONF in the recipe Mar 06 10:06:44 bluelightning: in my case a new libpng had been built, but the package didn't got rebuilt and my do_rootfs failed due to the old libpng not being available Mar 06 10:07:38 erbo: usually that's a result of a missing dependency, because bitbake should automatically ensure rebuilds for that case... Mar 06 10:08:46 bluelightning: oh, and indeed it was :) Mar 06 11:14:38 bluelightning: its actually libtool being broken :( Mar 06 11:14:51 erbo: For reference there is bitbake -c cleansstate Mar 06 11:19:49 i'm going to hack up a class that wipes .la files from the sysroot Mar 06 11:19:55 as I HATE THIS SO MUCH Mar 06 11:44:49 rburton: hmm, what happened to pb_'s patch that was supposed to do that? I thought it added an option Mar 06 11:45:05 maybe it got stuck on the naming of the option, I think I objected to that ;) Mar 06 11:45:46 yeah i'll dig it out Mar 06 11:46:17 would have saved me 20 minutes this morning chasing a libsystemd-daemon dependency in libudev - removed it in systemd and had to cleansstate every package that linked to udev to get it out of the image Mar 06 11:46:22 grrr etc Mar 06 12:47:07 heh la files are annoying :) Mar 06 13:27:28 rburton: there was a patch from walters a while back Mar 06 13:27:45 hmm, maybe it was pb Mar 06 14:05:47 RP, the ADT manual should be editted for terminology sometime Mar 06 14:06:05 Crofton|work: can you file a bug with pointers to the problems? Mar 06 14:06:09 built by the OpenEmbedded build system, which uses Poky. Mar 06 14:06:21 I guess :) Mar 06 14:06:46 built by the OpenEmbedded build system Poky and Mar 06 14:06:54 I hope ADT is distro neutral Mar 06 14:07:43 Crofton|work: should be Mar 06 14:07:56 yeah Mar 06 14:08:10 I am going to start poking at it Mar 06 14:08:46 Crofton|work: if there are issues, please make a note of them in the bugzilla. We do want to sort this kind of thing out but need pointers Mar 06 14:08:46 calling the repo "eclipse-poky" does not give me that distro neutral feelilng Mar 06 14:08:51 yep Mar 06 14:23:56 RP, https://bugzilla.yoctoproject.org/show_bug.cgi?id=3992 Mar 06 14:23:57 Bug 3992: normal, Undecided, ---, jessica.zhang, NEW , ADT Manual needs to be clear about OpenEmbedded and Poky Mar 06 14:26:11 heh Mar 06 14:26:15 I beat the bot Mar 06 14:31:00 anyone know if ADT will work with java-1.7.0-openjdk-1.7.0.9-2.3.7.0.fc17.x86_64? Mar 06 14:33:31 Crofton|work: the bot responded to your link ;-) Mar 06 14:33:53 you just need to talk about bug 3991 iirc Mar 06 14:33:54 Bug https://bugzilla.yoctoproject.org/show_bug.cgi?id=3991 normal, Undecided, ---, jessica.zhang, NEW , eclipse-plugin 1.3 cannot reconfigure and build a clutter project for 1.3.1 Mar 06 14:35:28 there is a lot of jargon I do not understand yet Mar 06 14:55:40 RP: pb sent the patch, walters does the same in ostree (and says only wipe libdir/baselibdir, not recursively) Mar 06 14:55:55 well Mar 06 14:56:18 heh Mar 06 14:56:22 well, something like that Mar 06 14:56:42 one sec Mar 06 14:58:01 some discussion here https://bugzilla.gnome.org/show_bug.cgi?id=654013 Mar 06 14:58:02 Bug 654013: was not found. Mar 06 14:59:03 except that got reverted elsewhere... anyways the TL;DR is that we do delete all .la files Mar 06 14:59:59 the debian page: http://wiki.debian.org/ReleaseGoals/LAFileRemoval raises the libltld concern, but apparently recent libltdl doesn't actually need .la Mar 06 15:00:05 *libltdl Mar 06 15:01:26 ah right, found it Mar 06 15:01:27 https://git.gnome.org/browse/jhbuild/commit/?id=965c8d5ceda9d1c5d6021ef2c534e0a7f68ca976 Mar 06 15:01:42 so the problem with just cleaning toplevel ones is that subdirectory ones will reference the toplevel ones Mar 06 15:01:57 ah Mar 06 15:02:03 so yeah, this all argues for ensuring libtldl doesn't need .la files Mar 06 15:02:15 any idea what version doesn't need it? Mar 06 15:02:52 step one would be to post-process the sysroot population to wipe out all .la files in there Mar 06 15:04:49 walters: doesn't PA use ltdl? Mar 06 15:04:53 yeah Mar 06 15:05:51 oh incidentally, i don't know if the latest eglibc in oe-core master has the fix, but pulseaudio is totally broken without https://github.com/cgwalters/poky/commit/b1c2d2470b925f2d53c91bd01bf103991fab3859 Mar 06 15:06:37 applies to the elibc in danny cleanly, and fixed gnome-shell locking up when i hit tab in gnome-terminal, and that makes my life a lot better =) Mar 06 15:06:47 anyways let me look at the libltdl source Mar 06 15:06:58 walters: ouch! Mar 06 15:09:14 yeah at least in the danny libtool-2.4.2.tar.gz you can clearly see in lt_dlopenadvise() that it checks #if defined(LT_MODULE_EXT) and falls back to trying to open that Mar 06 15:18:29 walters: are you keeping ostree up to date with the danny branch? Mar 06 15:18:40 nope Mar 06 15:19:34 walters: we're close to a 1.3.1 release, so the danny branch currently should be good Mar 06 15:34:47 kergoth: I'm going to have to revert the cooker pool change :( Mar 06 15:34:59 doh, what's going on with it? Mar 06 15:35:26 kergoth: Just spend an hour trying to make it work properly and terminate is simply just broken even in python 2.7.3 :( Mar 06 15:35:33 ugh Mar 06 15:36:32 kergoth: https://bugzilla.yoctoproject.org/show_bug.cgi?id=3978 Mar 06 15:36:33 Bug 3978: normal, Undecided, ---, richard.purdie, NEW , Infinite hang in bitbake parse if you have certain types of parse errors Mar 06 15:36:34 I'm guessing it must use SIGTERM under the hood still, and that the child isn't catching it and shutting down the queues and threads used for interprocess communication cleanly, and without that, the main process could hang on the send Mar 06 15:37:04 i hit that today Mar 06 15:37:04 kergoth: right, something like that, yes Mar 06 15:37:17 * kergoth mutters Mar 06 15:38:40 kergoth: do you want to see if we can do something better or should I revert the problematic change? Mar 06 15:39:03 I'd say go ahead and revert it, we can always pursue something better in parallel without users risking hangs Mar 06 15:39:09 kergoth: only other option I could find is let the parse complete in case of errors and not call terminate() Mar 06 15:39:14 * kergoth nods Mar 06 15:39:25 that makes bitbake feel non-responsive too, though :) Mar 06 15:39:29 due to the latency after hitting ^C Mar 06 15:39:42 kergoth: I know, wasn't happy with it, just a useful data point Mar 06 15:39:46 * kergoth nods Mar 06 15:39:47 agreed Mar 06 15:40:19 means terminate is breaking rather than the exception handling Mar 06 15:40:26 that's disappointing, I was really hoping to be able to rely on the standard module. boo for stadnard libraries you can't rely on.. Mar 06 15:41:07 I'll look into it on the side if I find the time, in theory we should be able to catch the TERM on the child side and do a bit of cleanup before shutdown Mar 06 15:41:31 kergoth: I agree its disappointing and annoying we can't use the standard functions :( Mar 06 15:41:38 course i'm not sure if we have access to the bits multiprocessing is using under the hood Mar 06 15:41:54 perhaps we can subclass and extend Pool Mar 06 15:41:57 kergoth: I'll revert for now, ensure this doesn't bite people and we can look at it as we have time Mar 06 15:42:01 sounds good Mar 06 15:46:14 RP: I've recently discovered that you can install a sigterm handler and in the handler raise an exception, and it will be raised from the point in the code where the handler got fired, so you can basically implement a sigterm equivalent ot KeyboardInterrupt. opens up some interesting ways of handling things Mar 06 15:46:57 kergoth: I didn't know you could do that, interesting indeed... Mar 06 15:47:06 e.g. https://github.com/kergoth/oe-test-scripts/blob/master/bb_test.py#L29-L45 Mar 06 15:48:19 kergoth: interesting... Mar 06 15:49:00 kergoth: On a different note, I think we need to create some kind of cached path intercept library, during do_package the stat calls cripple us :( Mar 06 15:49:15 that reminds me, we should add signal handling to the bitbake wrapper. today if you kill the wrapper bitbake itself keeps running away :) Mar 06 15:49:20 kergoth: We know the filesystem doesn't change but python doesn't :( Mar 06 15:49:23 ugh, that's unfortunate Mar 06 15:50:10 hmm Mar 06 15:50:11 kergoth: personally, I want to kill the wrapper ;-) Mar 06 15:50:29 heh, even better Mar 06 15:50:57 kergoth: just need to figure out a way of nicely being able to exec and then fork worker contexts in runqueue Mar 06 15:51:04 does anyone know if there are an example on how to write a reciepe for building a Qt5 application on top of the meta-qt5 layer? Mar 06 15:51:13 so we could create a one with pseudo around Mar 06 15:51:30 fatal: otavio might Mar 06 15:53:02 * kergoth tests theoretical fix for the parsing pool issue Mar 06 15:57:23 RP: did you find a more reliable way to repro the hang than the patch on the bug? i haven't seen that hang yet Mar 06 15:57:41 kergoth: touch conf/local.conf too Mar 06 15:58:09 kergoth: it was reproducible as long as there were many recipes being parsed and the one using syslinux wasn't last Mar 06 15:58:19 ah Mar 06 17:57:11 gm all Mar 06 17:57:13 rburton: yes Mar 06 17:57:16 now I am Mar 06 18:20:01 is my understanding flawed or should I be able to build for multiple platforms into the same tmp directory (not at the same time of course). Mar 06 18:20:28 yes you can Mar 06 18:20:37 it will construct a new sysroot/MACHINE for you Mar 06 18:22:15 Ok, will look closer and try to figure why I get bad builds on my second machine when I build it after the first, but it seems to work when built first. Mar 06 18:23:01 you might be using the wrong sysroot for some things - ie putting stuff in the native sysroot that isn't actually native Mar 06 18:24:09 To my knowledge, none of my stuff puts things in the native sysroot. Although I was looking at having a -native package soon. Mar 06 18:26:06 blloyd: make sure if hte recipe is machine specific it sets PACKAGE_ARCH=MACHINE or whatever. the sysroots are broken up by package arch Mar 06 18:26:59 kernel module should do that for me, and that 'should' be the only thing machine specific. The rest are more generic (i586, i686, etc). Mar 06 18:27:00 blloyd: MACHINE_ARCH Mar 06 18:27:10 * kergoth nods at JaMa Mar 06 18:28:01 it's probably some auto configuration of a package biting me somewhere. Mar 06 18:28:12 I wasn't reading backlog, but s/-/_/g between MACHINE and MACHINE_ARCH is annoying so I wanted to mention it Mar 06 18:29:37 blloyd: use openembedded-core/scripts/sstate-diff-machines.sh to check that only MACHINE_ARCH recipes have different checksums Mar 06 18:29:50 blloyd: assuming those 2 MACHINES have the same architecture Mar 06 18:30:41 JaMa: ah, forgot about that, good call (machine vs machine_arch) Mar 06 19:41:39 yeah we have all > arch > machine-arch > machine Mar 06 19:41:54 machine-arch is more like sub-family Mar 06 19:42:11 thanks arm for being so divergent Mar 06 21:05:08 hello Mar 06 21:13:02 anyone familiar with adt here? Mar 06 21:19:23 Crofton|work: jzhang might be able to help if she is around Mar 06 21:20:10 the eclipse jargon is making me crazy Mar 06 21:20:37 I want to check it out and install on my machine Mar 06 21:20:48 but I want to talk about exact steps so I do not make a mess Mar 06 21:21:04 I already have a newer version of open-jdk etc Mar 06 21:27:43 Crofton|work: http://www.yoctoproject.org/docs/1.3/adt-manual/adt-manual.html Mar 06 21:27:48 worked for me Mar 06 21:28:36 only thing that can confuse you is it lists more than 1 method of doing things but I guess thats ok Mar 06 21:31:59 yeah, that is what is annoying me Mar 06 21:32:08 so I want to work from first principles Mar 06 21:32:27 and I want to install for oe-core, not poky Mar 06 21:32:40 and I want to install for oe-core based rootfs, not poky Mar 06 21:32:50 and use toolchain from populate_sdk Mar 06 21:58:24 is there a way to get authentication credentials into bitbake for connecting to say a subversion server that requires authentication? Mar 06 21:59:14 I think you can put them in the SRC_URI Mar 06 22:02:20 for http/https, i'm guessing subversion probably obeys netrc or ~/.subverison/config or whatever Mar 06 22:18:50 kergoth, thanks. I'll double check the $HOME/.svn/ directory passwords. That's a good idea Mar 06 22:19:29 (I know how to set a password for a checkout folder, just not from the repository itself) Mar 06 23:07:46 otavio: fsl danny kicker per request. Mar 06 23:13:07 Crofton|work: so were you able to setup eclipse now by following the adt manual. Just follow the path of download juno, go to windows to install the extra plug-in + yocto eclipse plugin Mar 06 23:13:28 well, I want to try from git Mar 06 23:13:36 will openjdk 1.7 be an issue? Mar 06 23:14:14 Crofton|work: should be doable Mar 06 23:14:15 Crofton|work: there're some API changes for Java so we normally suggest 1.6 Mar 06 23:14:28 well, that will be a nuisance :) Mar 06 23:14:36 jzhang: I have used 1.7 partially Mar 06 23:14:39 I wonder if F17 can even downgrade to that Mar 06 23:14:41 but mainly 1.6 Mar 06 23:15:24 so if I check out from git Mar 06 23:15:47 and run the appropriate script, it will download eclipse and install the yocto specific bits? Mar 06 23:16:06 Crofton|work: it should work but since it's your first time just for taking the safe route, but since your distro is F17, then no need to downgrade Mar 06 23:16:18 Crofton|work: usually plugins are meant for a given version of eclipse Mar 06 23:16:29 Crofton|work: check out from git, you mean the Yocto eclipse or eclipse upstream? Mar 06 23:16:37 so I assume that is why it gets the eclipse it wants Mar 06 23:16:44 yocto eclipse Mar 06 23:17:46 Crofton|work: no, you need to have a functioning eclipse first, Juno for our case, then Yocto eclipse since YP eclipse is only one plug-in and it relies on other plug-ins, just like (I know you Mar 06 23:18:25 also, I am assuming the ADT will install a toolchain from the output of populate_sdk Mar 06 23:18:33 ah Mar 06 23:18:36 yocto eclipse is just a plugin IIRC Mar 06 23:18:55 it expect the full bloat of eclipse to be in your dev box Mar 06 23:19:45 Crofton|work: yes populate_sdk will have a toolchain + the sysroot Mar 06 23:20:37 I've done cross compiles with the output of populate_sdk, I am assuming that ADT can take the sh file and instlll it, or I just tell ADK where the output is Mar 06 23:21:40 Crofton|work: you run the sh and specify the location you want things to be installed Mar 06 23:22:05 this page seems dated btw: https://wiki.yoctoproject.org/wiki/How_to_setup_environment_for_ADT_with_1.1_on_Fedora_16 Mar 06 23:22:33 ok, so the remaining part that confuses me, is do I install eclipse from yum and the fedora feeds? Mar 06 23:24:04 or from the eclipse website Mar 06 23:24:31 Crofton|work: you can, but I'd suggest just go to eclipse download side and download the eclipse matching your dev box, I think before there's incident the eclipse packages is not uptodate with the distro, etc. There, you just extract the tarball and run eclipse, and install extra plugins follow the ADT manual Mar 06 23:24:32 yes, I suck at blindly following the exact steps Mar 06 23:25:26 Crofton|work: what's the issue u ran into then? Mar 06 23:25:40 I do not understand what all the pieces are Mar 06 23:25:46 :) Mar 06 23:26:08 Crofton|work: ok, now have a better idea or even worse ;-) Mar 06 23:26:20 a little Mar 06 23:26:34 fedora has loads of eclipse packages Mar 06 23:27:02 Crofton|work: I'd suggest don't go down that route ... Mar 06 23:27:55 so the adt installer contains a copy of eclipse? Mar 06 23:37:34 so adt is built from a bitbake recipe, does it have any special knwoledage about machines and distros present when the tarball is created? Mar 06 23:42:56 hello Mar 06 23:43:19 i'm trying to learn how to use yocto. ( never used bitbake before ) Mar 06 23:43:42 and .. i'm trying to build an image based upon meta-raspberrypi layer Mar 06 23:44:13 and when i try to build ... i get this : WARNING: The free inode of /dev/sda6 is running low (0.000K left) Mar 06 23:44:27 i do have 49Gb of disk left Mar 06 23:44:34 that is weird Mar 06 23:45:07 feydrautha80: which file system are you using on /dev/sda6? Mar 06 23:45:14 btrfs Mar 06 23:45:32 ah ok, I think there was a fix to avoid that bug recently, one sec Mar 06 23:45:44 oohh i see :)) Mar 06 23:45:56 fixed in master, and in danny branch Mar 06 23:46:37 hmm, I run the adt_installer Mar 06 23:46:53 and it talks about suported rootfs types, but mine is custom Mar 06 23:47:42 so how can i get a newer yocto ? Mar 06 23:47:51 feydrautha80: http://git.yoctoproject.org/cgit/cgit.cgi/poky/commit/?id=b0587077a47b2af0d88e17613fb48787e4c11716 is the fix you need Mar 06 23:48:20 Iand the installer dies Mar 06 23:48:54 feydrautha80: if you're running from a tarball, just apply that patch Mar 06 23:49:08 i got it from git .. Mar 06 23:49:19 feydrautha80: what revision/branch? Mar 06 23:50:39 4a36a32567ecfbc7ce7b967803e6e23314953ef5 Mar 06 23:51:16 so it looks like its older Mar 06 23:51:16 ah, 1.4m1 Mar 06 23:51:19 http://pastebin.com/sK6SdBNM Mar 06 23:52:05 feydrautha80: as you're already running master branch, just fetch and checkout master. it mostly works. :) Mar 06 23:54:02 umm:) Mar 06 23:54:05 (there's a 1.4m4rc1 tag if you want something that's been tested) Mar 06 23:54:06 thanks :) Mar 06 23:54:13 need to read about git too :)) Mar 06 23:56:17 oooh Mar 06 23:56:24 thank you mr Burton Mar 06 23:56:31 it works Mar 06 23:56:41 too bad Liz is dead :) Mar 07 00:00:18 Crofton|work: adt_installer doesn't contain eclipse and it doesn't have knowledge about machine, etc. it use a config file to figure things out Mar 07 00:00:44 yeah, I see that from the file size :) Mar 07 00:01:01 I reied installing in my home dir and it failed, see pastebin Mar 07 00:07:18 afk Mar 07 00:07:31 OK, adt_installer using opkg and the first thing it does it download and build opkg on your host, the issue is somehow the opkg failed at autoconf on your machine... hmm, probably file a bug for it. So to make you moving forward, must you use adt_installer? Mar 07 00:09:28 no Mar 07 00:09:59 Crofton|work: and you want to do cross dev using eclipse? Mar 07 00:10:40 right Mar 07 00:11:04 basically, I have a toolchain from popoualte_sdk that I want to use to cross compile and debug on the target board Mar 07 00:11:19 Crofton|work: then you can generate the toolchain with sysroot using -c populate_sdk against an image. And setup your cross env that way... Mar 07 00:11:48 yes, I ahve done that Mar 07 00:11:53 Crofton|work: then you have everything you need as far as ADT goes, so do you have your eclipse side ready? Mar 07 00:11:59 no Mar 07 00:12:29 I need to go now, I will look at it some more in the morning Mar 07 00:12:47 Crofton|work: then focus on setting up eclipse IDE and I'll help u once that's done to configure using your ADT Mar 07 00:12:48 it is beer time here Mar 07 00:13:02 setting up eclipse is confusing me Mar 07 00:13:16 talk to u later... beer is above everything;-) Mar 07 00:13:21 it sems like adt installer is supposed to get the correct one? **** ENDING LOGGING AT Thu Mar 07 02:59:58 2013