**** BEGIN LOGGING AT Tue Aug 25 02:59:59 2015 Aug 25 05:23:30 Hmm, what was Darren Hart's nick again? Aug 25 08:26:27 When I do a bitbake -c fetchall, it runs a bunch of compile jobs too. Is that normal? Aug 25 08:29:36 or is there some better way to do only the source download Aug 25 08:32:47 gunnarx: without taking a closer look, it probably builds the stuff needed for getting. (curl, git, whatever) Aug 25 08:34:03 LetoThe2nd: Oh, ok, I thought those tools were taken from the host system. They are built first? Aug 25 08:34:29 gunnarx: as little as possible is taken from the host system, so i'm relatively sure these are built first. Aug 25 08:34:37 gunnarx: or at least some of them Aug 25 08:43:46 LetoThe2nd: You're right, they are all -native components. And since the tools have lots of dependencies themselves it was more than I had expected. Aug 25 08:44:13 \o/ Aug 25 09:44:25 morning all Aug 25 09:44:55 gm Aug 25 09:45:12 bluelightning, any word on atlas/csipy? Aug 25 10:04:01 Crofton: I think Alejandro had it basically building after bashing on it for some time, but he only got it to build for x86 so far unfortunately Aug 25 10:04:21 I suggested he talk to you about getting some help to fix the remaining issues Aug 25 10:04:36 has he been in touch? Aug 25 11:25:55 hey guys Aug 25 11:27:26 I've written a script to fully automate building process for jenkins. It is executed on my local host perfectly, but if I run it on a jenkins host it fails with the following message: http://dpaste.com/37JK0MX Aug 25 11:27:34 how to resolve this? Aug 25 11:28:15 Use a umask which allows a+rx and u+rwx Aug 25 11:28:34 e.g. 0022 Aug 25 11:30:18 JaMa: for bitbake script? Aug 25 11:32:46 no, in your operating system Aug 25 11:38:31 I agree the message is a bit cryptic Aug 25 11:55:36 It looks like a bunch of perl modules exist in poky/OE: http://cgit.openembedded.org/cgit.cgi/meta-openembedded/tree/meta-perl/recipes-perl/libtest/libtest-harness-perl_3.30.bb?h=master Aug 25 11:55:44 how do I access them from perl scripts? Aug 25 11:55:57 the usual 'use foo::bar' doesn't seem to work Aug 25 12:00:53 are you trying to use that from the target or the build environment? Aug 25 12:03:26 On the target machine Aug 25 12:04:14 I'd suggest checking if the package is installed to the rootfs and where Aug 25 12:04:37 fredcadete: It don't see it anywhere on a 'find' of the rootfs Aug 25 12:05:00 in your build system, do you see it in the work directory? Aug 25 12:05:14 ls build/tmp/worl//libtest-harness-perl Aug 25 12:05:21 or something like that Aug 25 12:05:50 then you can see it in the "package" directory Aug 25 12:07:17 I'll check Aug 25 12:07:18 ./perl/5.20.0-r1/image/usr/lib/perl/5.20.0/TAP/Parser.pm Aug 25 12:07:37 which leads me to believe it ought to be on the "image" under /usr/lib/ Aug 25 12:07:44 Or am I misunderstanding things? Aug 25 12:08:21 I'm not sure, that "image" is tripping me up Aug 25 12:08:28 I was expecting "package" Aug 25 12:08:39 I think I have things in 'package' as well Aug 25 12:09:00 yep Aug 25 12:09:01 ./perl/5.20.0-r1/packages-split/perl-module-tap-parser-result/usr/lib/perl/5.20.0/TAP/Parser/Result.pm Aug 25 12:09:58 under packages-split you don't have a package that includes TAP/Parser.pm? Aug 25 12:13:02 on the image? Aug 25 12:13:17 Or is it just on the build machine? Aug 25 12:15:07 Holy crap, there's a ton of perl modules there. Aug 25 12:15:13 thanks fredcadete Aug 25 12:15:34 I guess I can just move'm over Aug 25 12:16:00 the right thing would be to add the corresponding package to your image Aug 25 12:17:05 Okay, I'll do that. Aug 25 12:29:25 fredcadete: It appears that meta-perl was not listed in my bb.layers file Aug 25 12:29:31 I've added that now. Aug 25 12:29:40 do I need to bitbake clean? Aug 25 12:30:35 bitbake -c clean rarely works for me to get a failed state back into working mode. Just saying. Aug 25 12:30:58 It's a small step to bitbake clean only the offending packages, but as I said it's not guaranteed to work (in my experience) Aug 25 12:31:37 you know that -c cleanall will remove also the downloaded source right? Rarely needed but worth trying if all else fails. Aug 25 12:32:40 jeremiah: no, no need to clean Aug 25 12:33:11 bluelightning: thanks, and thanks gunnarx Aug 25 12:33:55 gunnarx: -c clean isn't meant to be an unbreak-everything type of tool though, in fact almost always there's some underlying problem that cleaning and rebuilding is not going to help fix Aug 25 12:36:20 bluelightning: Yeah I don't know what it's meant to do, but I am just noting that it rarely helps in the type of problems I run into, despite that the build itself is in some sense OK because if I wipe all state it is typically successful Aug 25 12:36:40 but that means some hours of repeated build time. Aug 25 12:36:58 gunnarx: I would strongly urge you to try to get those problems fixed by reporting them, or we'll never be able to fix them Aug 25 12:37:20 if you wipe your TMPDIR any information that might be used to track down the issue is gone Aug 25 12:38:02 bluelightning: Of course I would if I can identify the yocto problem. But if a system definition builds find from clean I suspect that yocto/bitbake is considered OK. Aug 25 12:38:16 For example, I've had this computer that shuts down due to heat in the middle of the build Aug 25 12:38:31 gunnarx: ah well, that's a problem we cannot fix, no ;) Aug 25 12:38:33 Usually the build will not continue cleanly, there are errors. Aug 25 12:39:12 I guess it's fair to say we do assume the hardware on which we're running can be relied upon Aug 25 12:39:16 (after rebooting). So all I am looking to do is to understand how to make it all a bit less brittle if there are any problems along the way. Aug 25 12:40:13 there could be a lot of factors at work under those circumstances... did the writes that the system thought succeeded actually succeed? Aug 25 12:40:20 Theoretically it would be possible to recognize that the previous step was not cleanly done, and backtrack to a stable state, I'm just assuming... Aug 25 12:40:31 Many tasks running in parallel is of course a factor Aug 25 12:40:50 well, we kind of already do that - we only write a stamp for a task (i.e. mark it as done) if it really did run successfully to completion Aug 25 12:41:23 bluelightning: I know and that's why I'm often surprised at the level of difficulty to restart things cleanly when it is done half way like this. Aug 25 12:41:46 Of course the build scripts could be unclean too. They only work when run in the right order, undisturbed and so on. Aug 25 12:41:54 now, if internally that task writes some state that is bad and it's unable to recover from that, then there might be a problem - but cleaning that particular recipe should solve that kind of problem Aug 25 12:42:11 "should" which brings us full circle :-) Aug 25 12:42:26 I can't see how it couldn't though Aug 25 12:42:33 I'm not here only to complain, just sharing and discussing... Aug 25 12:42:45 if it doesn't, there is something else happening - e.g., writes that you thought got back to disk actually did not Aug 25 12:43:16 or, files that were written OK got corrupted after the fact Aug 25 12:43:22 makes sense what you're saying. I've had this idea to dump out all tasks and run them one by one, storing the state that is created by each one. Aug 25 12:43:34 But that means serializing the build so I'm changing the parameters then. Aug 25 12:43:52 storing state = in git/git-annex for example Aug 25 12:44:27 I'm certainly all for trying to mitigate failures if we practically can, at the moment without actual details I'm not sure as to the actual cause of the problem Aug 25 12:44:40 Hi, everyone! Is there simple method to add some tree of files and directories to image? I mean some configuration files, soft links, etc… Aug 25 12:44:48 np, I'm not expecting anything more from you atm Aug 25 12:44:52 thanks for the chat Aug 25 12:49:47 pohmelie: Seems like you could just follow any recipe definition tutorial for this. Your recipe is just to copy files that you can either keep in your layer, or alternatively unpack a tarball with them. Aug 25 12:50:03 I'm about to ask a question on this file: http://tinyurl.com/qb89s5q So why don't you look there. You can see weston.ini stored in the layer, and you can see it being installed with install. Of course "cp" is an alternative. Aug 25 12:51:18 Specific question: http://tinyurl.com/qb89s5q What's an idiomatic way to select one or another file (weston.ini in this case) based on the value of $MACHINE? Aug 25 12:51:48 Could the value of MACHINE just be appended to the filename maybe... Aug 25 12:52:10 I have no idea if that's considered a clean way to do it or not though... Aug 25 12:52:38 (i'm talking about line 33) Aug 25 12:53:13 gunnarx: yeah, that's what I'm talking about. If I have two files, recipe is ok, but when I need to copy hundreds? I think about archive, but maybe there is more elegant solution? Aug 25 12:53:35 or not… Aug 25 12:53:52 gunnarx: you should just be able to put the file into a MACHINE-named subdirectory/subdirectories and it should just work Aug 25 12:54:02 I would think about that as source code and provide a tarball that you fetch from somewhere. Aug 25 12:54:35 That $MACHINE dir goes in same dir as the bb file? Aug 25 12:54:39 gunnarx: hm, interesting, thanks Aug 25 12:55:12 gunnarx: no, under a directory next to it that's the same name as the name part of the recipe, or "files" Aug 25 12:56:00 normally you'd just put the files in that directory itself, but to make them specific to a machine or architecture you can put them under a subdirectory under that with the machine/arch name Aug 25 12:56:51 Like wayland/ in this example? Aug 25 12:57:21 No actually wayland/weston/ I suppose... Aug 25 12:59:28 bluelightning,I think I got it. thanks. Aug 25 13:03:31 having trouble setting up authenication to push to poky contributions - can anyone help? Aug 25 13:06:10 raykinsella78: I can try - what seems to be the problem? Aug 25 13:12:31 well I asked to send a public key to Michael Halstead  Aug 25 13:12:43 which I did ... Aug 25 13:13:36 so now I setting up ssh to clone from poky-contrib ... and something is going wrong. Aug 25 13:15:04 Cloning into 'poky-contrib'... Aug 25 13:15:04 Permission denied (publickey). Aug 25 13:15:04 fatal: Could not read from remote repository. Aug 25 13:15:04 Please make sure you have the correct access rights Aug 25 13:15:04 and the repository exists. Aug 25 13:15:10 Does any body know if 'backtrace' is supported with Fido ARM builds? I getting some bogust stack traces in my libray, not sure if it could be related to ARM limitations Aug 25 13:15:46 in my .ssh I specifically set the private key for the public key I sent Michael. Aug 25 13:15:55 So I am definitely doing something wrong. Aug 25 13:19:36 raykinsella78. can you pm the line you are using for the clone ? Aug 25 13:22:01 k Aug 25 13:22:19 strange Aug 25 13:23:35 I've executed script from jenkins workspace and it is executed successfully. But if I trigger a job from the jenkins web interface I receive sanity check misconfiguration :-( Aug 25 13:29:07 Ox4: is this the same umask issue you pasted above? Aug 25 13:29:20 bluelightning: yes, it is Aug 25 13:29:28 Ox4: did you follow JaMa's advice? Aug 25 13:30:02 I executed chmod a+rw -R in the jenkins workspace but it didn't help Aug 25 13:31:24 erm Aug 25 13:31:27 a+rw* Aug 25 13:31:27 that's not umask Aug 25 13:31:42 umask is about files you create from that point on Aug 25 13:31:53 u+rwx? Aug 25 13:31:54 umask is set with the "umask" command Aug 25 13:32:03 hold on Aug 25 13:33:41 the same Aug 25 13:34:10 it's also about what umask is set to within the jenkins environment, not outside of it Aug 25 13:36:41 bluelightning: do I have to do umask all jenkins workspace? Aug 25 13:36:47 to all* Aug 25 13:37:53 Ox4: I'd imagine you could set it in the build script just before running bitbake Aug 25 13:38:28 Ox4: a general note, Jenkins is not a build system. bitbake and yocto are a build system. Jenkins needs a lot of hand holding to execute jobs in a sane way, e.g. builds should be isolated from each other, start with clean workspace (or git clean -f -f -x -v && git reset --hard), stop properly without leaking processes... I suggest running these in all of your jenkins job shell scripts: env && set -euxo pipefail && bash. It might be the jenkins is not running thing Aug 25 13:39:37 * mcfrisk has a lov.. mostly hate relationship with jenkins Aug 25 13:41:20 bluelightning: for poky directory? Aug 25 13:41:25 ok I will try Aug 25 13:41:58 Ox4: it's not for a directory - umask is associated with your shell session Aug 25 13:42:17 hm Aug 25 13:45:00 umask 022, right? Aug 25 13:46:16 Ox4: 022 should be fine I think yes Aug 25 13:47:07 it works Aug 25 13:47:17 thank you bluelightning and mcfrisk Aug 25 14:19:36 I am stuck in "Waiting for root device /dev/mmcblkp2" when starting yocto Aug 25 14:19:43 does anyone know what this is ? Aug 25 14:22:55 someone changed some gnome pkgs from bz2 to xz and it seems not all mirrors have the xz (yet?) :( Aug 25 14:23:06 causing an obvious fetcher failure. Aug 25 14:23:56 patches! there's a patch on the list for libnotify Aug 25 14:24:17 the fetcher needs a way of doing a HEAD to sanity check URLs without downloading the entire file Aug 25 14:27:39 the fetcher can and does check urls before it downloads the entire file, that's what the checkstatus() method in the fetchers is Aug 25 14:27:47 not entirely sure where all it's used offhand, though, i haven't had my coffee Aug 25 14:27:49 * kergoth yawns Aug 25 14:28:21 though, maybe you meant checking more than whether its a 404 :) Aug 25 14:32:14 * kergoth mulls over what tests need writing for the git fetcher Aug 25 14:40:04 paulg_: even if they have it, it will fail because different checksums Aug 25 14:44:09 kergoth: its used in the recipe reporting system to figure out various things about recipe upgrades amongst other places Aug 25 14:44:18 ah, cool Aug 25 14:44:26 kergoth: think there is a checkstatus task like do_fetch too, or there was... Aug 25 14:44:43 Hi, how can I install all python-XXX packages in my image? Aug 25 14:44:57 ljose: install python-modules ? Aug 25 14:45:32 yep, checkuri/checkuriall from utility-tasks.bbclass. cool, that sounds useful Aug 25 14:47:05 kergoth: checking for 404 would help Aug 25 14:47:07 ljose: I think this might be what you're looking for http://www.yoctoproject.org/docs/1.8/ref-manual/ref-manual.html#var-CORE_IMAGE_EXTRA_INSTALL Aug 25 14:47:14 ah, neat Aug 25 14:47:42 ljose: plus, check the python-manifest file to see what modules you want in specific Aug 25 14:48:55 yeah, I'm getting 404s by the handful. Aug 25 14:49:53 (well, at least four...) Aug 25 15:02:32 guys, I have one more question Aug 25 15:03:02 | cc1: error: unrecognized command line option "-fuse-ld=bfd" Aug 25 15:03:44 in the meta/classes/kernel-arch.bbclass I see this option: KERNEL_CC = "${CCACHE}${HOST_PREFIX}gcc ${HOST_CC_KERNEL_ARCH} -fuse-ld=bfd" Aug 25 15:03:57 can I remove this? Aug 25 15:14:47 hm Aug 25 15:14:49 seebs: around? Aug 25 15:14:51 http://git.yoctoproject.org/cgit/cgit.cgi/poky/commit/meta/classes/kernel-arch.bbclass?id=aced11f4dc8fb4f217360198e8c4cda2cdf620e0 Aug 25 15:16:04 seebs: we're seeing oddness like https://autobuilder.yoctoproject.org/main/builders/nightly-ipk/builds/451/steps/BuildImages/logs/stdio, the pseudo upgrade was in there and I'm wondering if its related Aug 25 15:16:13 seebs: no proof as yet mind :/ Aug 25 15:17:09 0x4 - you using a non-default toolchain? Otherwise I'd expect others to be seeing that error too.... Aug 25 15:18:10 paulg_: yes, I am using sourcery gcc-4.5.1 Aug 25 15:19:04 0x4 probably easiest to just bbappend that option out of existence in your custom layer then. Aug 25 15:20:25 that'll be a pain, you'd have to do it for every kernel recipe for every bsp. I'd use KERNEL_CC_remove in config metadata Aug 25 15:20:32 i'll probably add that to meta-sourcery Aug 25 15:22:35 kergoth: I just set KERNEL_CC in my linux-yocto-custom.bb file Aug 25 15:22:46 * kergoth nods Aug 25 15:52:41 Is there a way to better manage the SDK installer generated by -c populate_sdk? Specifically we need to ship that out to various developers and it would be nice if the contents could be broken down so we can have packages for the various pieces. Aug 25 15:52:47 0x4, kergoth: -fuse-ld was added in gcc 4.8, so yeah, a 4.5-based toolchain won't know it Aug 25 16:14:25 I'm running a command (. ./oe-init-env bitbake -e | grep ...) and when it returns my console is all messed up Aug 25 16:14:39 is there a way to get bitbake stop stop screwing with the console? Aug 25 16:14:52 I just need the value of a specific variable so I can do some setup Aug 25 16:17:34 (it's odd, I can't always replicate the failure mode with the console either.. :P Aug 25 16:17:41 RP: can you explain why the fetcher removes the entire clone when an update of it fails? I have a commit changing this, but figured I'd run it by you to see what the purpose behind that behavior was. was it to deal with repository corruption? if it's a transitory network failure, ir eally don't want to have to re-clone my entire kernel repo Aug 25 16:17:58 fray: heh, yeah, seem slike bitbake -e is pretty useless when piped nowadays. not sure when that changed Aug 25 16:18:12 kergoth, I thought it was due to non-fast forward git trees Aug 25 16:18:18 ok.. good so it's NOT just me.. Aug 25 16:18:29 git can fetch non-fast-forward branches just fine.. Aug 25 16:18:39 it's the repo in DL_DIR i'm talking about, not the clone in WORKDIR Aug 25 16:18:50 the latter would be more of a factor, since it has checkouts of the branches Aug 25 16:19:06 is there a simple way to get a variable out of bitbake then (better then bitbake -e | grep ...?) (I need the value of bblayers Aug 25 16:19:20 dunno Aug 25 16:20:44 bb would work for now, as a workaround. https://github.com/kergoth/bb - bb show BBLAYERS Aug 25 16:21:11 * kergoth hasn't touched bb in ages, but still uses it from time to time Aug 25 16:22:03 kergoth: I think it was just being paranoid Aug 25 16:22:06 * nerdboy wakes up with buddy guy and a stray kitty... Aug 25 16:22:27 so bb-show is actually using bitbake itself/ Aug 25 16:22:52 fray: it uses tinfoil and other bitbake modules, yes Aug 25 16:22:56 ok Aug 25 16:25:48 I wish I could figure out what is triggering the console going nuts.. sometimes it does, sometimes it doesn't.. Aug 25 16:26:49 fray: any idea when it started doing that? Aug 25 16:27:18 sometime between 1.7 and master.... (it definitely did NOT go it in 1.7.. I didn't use 1.8 enough to notice.. and it's been doing it for some time in 1.9... Aug 25 16:27:23 RP: I'm wondering about the case where we already have a git mirror tarball, e.g. from the original fetch, the repo was successfully updated in the past, but there's a transitory network failure causing the update to fail, then try_mirrors() is run on MIRRORS, which runs try_mirror_uri, which eventually re-runs the origud.download() again, but in this case both the tarball and the clone exist, so that second call of the original fetcher download method Aug 25 16:27:24 won't actually do the expected unpack of the mirror tarball, it'll try again to update the clone, and it could keep doing this through all the applicable MIRRORS Aug 25 16:27:33 I THOUGHT it went away, but it came back.. and like I said, I'm not getting consistent 'corruption' Aug 25 16:27:35 RP: unless I'm missing something, this is a theoretical case in my head from reading the code Aug 25 16:27:49 I have to do a 'reset' to get it back.. (basically it hides the cursor and never restores it) Aug 25 16:27:57 * nerdboy enjoys occasionally harassing the uni IT people on the SIG-UCCS mailing list... Aug 25 16:28:02 blind typing sucks.. ;) Aug 25 16:28:16 RP: although, previously the clone would have been cleaned, then the first mirror would have re-unpacked the tarball, and then it would have tried to update the newly unpacked clone and failed, anyway, so i guess it's no worse Aug 25 16:28:19 * kergoth scratches head Aug 25 16:28:27 fray: it sounds like there is an uncaught exception path in knotty Aug 25 16:28:44 thats what i was wondering if somehow it's dropping out without resetting the cursor mode(s) Aug 25 16:28:55 kergoth: I think I've looked at this before and decided not to worry too much about it. I agree it could do with being improved Aug 25 16:28:58 any suggestions on what I can do to try to nail it down? Aug 25 16:29:20 fray: try and spot the pattern to start with? Aug 25 16:29:23 (it wouldn't surprise me at all if it was an artifact of bitbake -e and something failing to expand) Aug 25 16:29:36 RP, I rarely have it happen with a poky checkout.. Aug 25 16:29:42 I almost always have it happen with a complex set of layers Aug 25 16:29:55 RP: Long term, I'd like to see the case where the mirror fetcher is different than the main fetcher be handled more explicitly, rather than re-running the origud.download(). e.g. have a new method for it, or at least pass the mirror ud into the download method so we can tell in what case we're being called Aug 25 16:30:01 * nerdboy would tie a string to kergoth and send him down the rabbit-hole... Aug 25 16:30:05 * kergoth ponders Aug 25 16:30:18 kergoth: that would seem reasonable Aug 25 16:30:43 RP: these crazy possible cases are hurting my head, we need more fetch unit tests :) Aug 25 16:30:59 fray: I wonder why that would be? Unless its extra parse time which opens up some kind of window Aug 25 16:31:01 it's odd if I run it from the command line it works.. but if I run it from a script it's not.. (still trying for a more reliable reproducer) Aug 25 16:31:09 kergoth: this is why I started them last time... Aug 25 16:31:16 wait, i thought you were immune to that stuff... Aug 25 16:31:25 * kergoth nods Aug 25 16:31:38 nerdboy: have you looked at the fetcher code? ;-) Aug 25 16:31:56 yeah, but then i sat back down and forgot about it Aug 25 16:32:05 every time i touch this code i cringe worrying that i might have broken some corner case i didnt' think about Aug 25 16:32:23 nerdboy: probably good from the nightmare perspective ;-) Aug 25 16:32:28 kergoth: I know the feeling Aug 25 16:32:31 it was probably at least a year ago... Aug 25 16:32:42 kergoth: I dread patches on the list for it too Aug 25 16:33:20 are we okay with python3 constructs? Aug 25 16:33:43 Hi all, I am get some weird error messages, when my bitbake build fails: "log_check: Matched keyword: [E:] ERROR: E: //; p }', as for example on libexpat-dev when running bitbake for the meta-raspberrypi. Does anyone have a clue? Thanks. Aug 25 16:33:48 or does it need to conform with 2.7 Aug 25 16:33:55 * kergoth is working on simplifying and splitting up the git shallow mirror tarball support to try to make it less painful to pull in Aug 25 16:34:07 nerdboy: the latter, we're not ready to move to 3 yet Aug 25 16:34:26 too bad, could use that kind of refactoring... Aug 25 16:34:41 3.4 is stable now i think? Aug 25 16:34:44 picolcd: hmm, that looks ugly :( Aug 25 16:35:00 whatever that means on gentoo... Aug 25 16:35:16 nerdboy: i think it's mainly the unicode bits that are holding us up, i don't think anyone has even started on that Aug 25 16:35:21 * nerdboy is not yet thinking in python3 Aug 25 16:35:36 much Aug 25 16:35:57 picolcd: log_check only runs when building an image though - so is it the image that's failing to build? Aug 25 16:36:23 kergoth: don't suppose anyone is jumping up and down to hack at it? Aug 25 16:36:34 there has been some work on it already, moving to more portable constructs and whatnot Aug 25 16:36:41 i did some ages ago, and others more recently Aug 25 16:36:47 but there's still work to be done Aug 25 16:36:51 is that in a branch somewhere? Aug 25 16:36:53 that is correct, actually it always fails on do_rootfs Aug 25 16:37:37 nerdboy: thus far all the work being done works in both 2 and 3, so hasn't needed a long lived branch Aug 25 16:37:39 afaik anyway Aug 25 16:37:40 bluelightning: that is correct, actually it always fails on do_rootfs Aug 25 16:37:47 k Aug 25 16:38:10 picolcd: ok, can you pastebin your log.do_rootfs ? Aug 25 16:39:07 RP: https://github.com/openembedded/bitbake/commit/76d57e1 is one that worries me a little. It *seems* like it should work fine, but I worry I'm missing something Aug 25 16:39:44 https://github.com/kergoth/bitbake/commit/76d57e1 rather Aug 25 16:41:11 kergoth: I agree it should in principle but I know what you mean... Aug 25 16:49:18 bluelightning: http://pastebin.com/wThud9Ti ... had to trim it a little, please tell me if you need more. Aug 25 16:54:08 picolcd: it looks like set -x is on, which is rather ugly... Aug 25 17:00:18 bluelightning: do you think that it would help if I do "set +x" before running bitbake again? Aug 25 17:10:58 Has anyone looked into adding a general 'if this package is in the sdk, then add these other specific packages to it' logic to populate_sdk? Somewhat similar to complementary packages, but a one-off with specific package names, not extending with a suffix. I think that's what we need for qt5 to be handled cleanly, but not sure if anyone has started on adding that to the core classes or if they're just punting and using separate sdk recipes or manual Aug 25 17:10:58 TOOLCHAIN_HOST_TASK Aug 25 17:12:00 we've always setup dynamic packagegroup packages based on whatever logic is reasonable.. Aug 25 17:12:03 then included them.. Aug 25 17:12:10 but we've avoided trying to look at what packages are 'installed' Aug 25 17:12:27 Hmm, fair enough Aug 25 17:12:38 but doing a post-process "install more based on what we did" isn't that hard.. preprocess is nearly impossible to get right Aug 25 17:13:19 we've also done this with the RRECOMMENDS and enabling/disabling them in variable installs.. Aug 25 18:00:45 okay, cleaned up and split -- https://github.com/kergoth/bitbake/compare/master...kergoth:fetch-bugfixes and https://github.com/kergoth/bitbake/compare/fetch-bugfixes...kergoth:simplified-shallow-git .. will re-submit as PATCH & RFC, respectively, later today Aug 25 19:19:24 Hi trying to install python pip, installed a bunch of python packages, but is failing because plistlib is not found, any ideas if there is a package for this? Aug 25 19:23:20 just throw in python-modules Aug 25 19:24:09 off the bat i would guess it's included in the python-setuptools (if i remember the packagename coreectly) Aug 25 19:25:01 i'd like to create an SDK for a qt5 based image. i don't want to build the meta-qt5-toolchain target but rather go the populate_sdk route so my SDK contains all things i need based on what i already have in the image. the problem is that the resulting SDK can't be used for qt5 development Aug 25 19:25:29 i was told that my image should contain this line: "inherit core-image populate_sdk_qt5" Aug 25 19:26:25 when i now do a "bitbake myimage -c populate_sdk" i get an SDK that is missing important bits and pieces Aug 25 19:26:33 like the mkspecs or qt.conf Aug 25 19:27:52 kergoth: we chatted about that topic a while ago. based on that i've added TOOLCHAIN_HOST_TASK_append = " nativesdk-packagegroup-qt5-toolchain-host" to my local.conf Aug 25 19:27:59 with some limited success Aug 25 19:28:09 now i have at least qmake and friends in my sdk Aug 25 19:28:22 but none of the mkspecs Aug 25 19:30:03 pretty sure the example i gave you included the specs. Aug 25 19:30:06 * kergoth shrugs Aug 25 19:39:21 kergoth: you are right Aug 25 19:39:36 hmm Aug 25 19:40:33 kergoth: but the names weren't exect matches. for the tools i managed to find the right name Aug 25 19:40:42 nativesdk-packagegroup-qt5-toolchain-host Aug 25 19:41:00 but for the mkspecs i am not sure Aug 25 19:41:09 i have to re-test Aug 25 19:44:40 kergoth: can I add packagegroup-qt5-toolchain-target to my TOOLCHAIN_HOST_TASK_append (in addition to nativesdk-packagegroup-qt5-toolchain-host)? it contains a dependency to qtbase-mkspecs Aug 25 19:44:41 RP: That does look like a bug that would typically be pseudo-related, except that it's also the sort of bug pseudo has been (historically) pretty good at avoiding. Were you using the xattrdb feature? Aug 25 19:48:53 fmeerkoetter: no, that wouldn't make much sense. target and host are completely different. host is stuff that runs on your machine, target ist arget. the sdk has two sysroots, they're indpenedent, with independent package install lists (TOOLCHAIN_HOST_TASK, TOOLCHAIN_TARGET_TASK), you can't install one in the other Aug 25 19:49:54 kergoth: should that approach of having an image inherit from populate_sdk_qt5 work? Aug 25 19:50:14 no idea, not familiar with populate_sdk_qt5. presumably it's supposed to handle this for ou Aug 25 19:50:39 yeah. it also does TOOLCHAIN_HOST_TASK_append = " nativesdk-packagegroup-qt5-toolchain-host" Aug 25 20:28:10 seebs: no, we wouldn't be using that Aug 25 20:29:50 Okay. In that case, I can't immediately think of anything that ought to be able to cause that. If you can get a reproducer, or point me at a failed build, I can have a look? Aug 25 20:30:25 seebs: well, we backed out the pseudo change and those kinds of failures stopped happening Aug 25 20:30:50 seebs: so I'm pretty sure it is pseudo. Have you run builds where its built an rpm package feed and constructed a rootfs from it? Aug 25 20:31:31 topik: I add python-modules but still get the same failure no module name plistlib Aug 25 20:31:53 plistlib sounds like something mac-specific, or a third party module Aug 25 20:33:19 I don't think so as install pip requires it Aug 25 20:35:12 *** lamego (lamego@nat/intel/x-qmimbglsztevbvee) has joined channel #yocto Aug 25 20:35:12 seebs: no, we wouldn't be using that [22:28] Aug 25 20:35:12 Okay. In that case, I can't immediately think of anything that ought Aug 25 20:35:15 to be able to cause that. If you can get a reproducer, or point me at Aug 25 20:35:18 a failed build, I can have a look? [22:29] Aug 25 20:35:21 seebs: well, we backed out the pseudo change and those kinds of failures Aug 25 20:35:25 stopped happening [22:30] Aug 25 20:35:28 seebs: so I'm pretty sure it is pseudo. Have you run builds where its Aug 25 20:35:31 built an rpm package feed and constructed a rootfs from it? Aug 25 20:35:34 topik: I add python-modules but still get the same failure no module Aug 25 20:35:37 name plistlib [22:31] Aug 25 20:35:41 plistlib sounds like something mac-specific, or a third party module Aug 25 20:35:41 *** jmpdelos (~polk@delos.delosent.com) has quit: Quit: Leaving [22:32] Aug 25 20:35:44 I don't think so as install pip requires it [22:33] Aug 25 20:35:48 *** jmpdelos (~polk@delos.delosent.com) has joined channel #yocto Aug 25 20:35:52 ERC> Changed in version 2.6: This module was previously only available in the Mac-specific library, it is now available for all platforms. Aug 25 20:35:55 Aug 25 20:35:58 Aug 25 20:36:04 opps Aug 25 20:48:29 I believe that's how the build I did worked, but I may have missed some step. Aug 25 20:50:16 ast module seems is also missing https://docs.python.org/2/library/ast.html Aug 25 21:05:36 halstead: have you touched the SendErrorReport part of the autobuilder at all? Aug 25 21:06:03 halstead: anything that might have resulted in https://autobuilder.yoctoproject.org/main/builders/poky-tiny/builds/470 ? Aug 25 21:06:42 RP, No but I have been troubleshooting the yocto-builds list not receiving e-mail. Aug 25 21:07:04 RP, No code changes on my part. Could this be related? Aug 25 21:08:19 halstead: not sure what is going on right now. Its unlikely I guess Aug 25 21:08:46 halstead: I did merge a couple of emergency fixes into the autobuilder code but I didn't touch this area :/ Aug 25 21:08:57 halstead: (top two commits of yocto-autobuilder master) Aug 25 21:09:13 RP, What I'm fixing stopped working in July. :( Aug 25 21:09:34 halstead: we appear to have multiple issues :( Aug 25 21:09:53 halstead: just wanted to check it wasn't something related to what you were looking at, thanks Aug 25 21:10:07 halstead: will try a master build and see where we are Aug 25 21:10:20 Looking at my python2.7 build I see tmp/sysroots/beaglebone/usr/lib/python2.7/plistlib.py and tmp/sysroots/beaglebone/usr/lib/python2.7/ast.py but those files are not being installed by python-modules Aug 25 21:10:28 could this be a problem with python recipe? Aug 25 21:10:34 RP, Thank you. I'll message if I think any changes will help. Aug 25 21:37:16 So any ideas on getting the other modules installed, should I look into patch python2.7-manifest? Aug 25 23:50:41 im getting to add a group into the a proyect, but im getting a awful error "ERROR: useradd not found in the base feeds (qemux86 i586 x86 noarch any all). Aug 25 23:50:42 acl-2.2.52-r0@" using as template /meta-skeleton/recipes-skeleton/useradd/useradd-example.bb Aug 25 23:51:53 if i used the full template it works fine but, as soon y remove useraddPARAM and do_install it sent that error Aug 25 23:52:13 i just want to create some new groups for other time use Aug 26 00:01:37 best bet is to have a look at meta/classes/useradd.bbclass and see what it does with the params. Aug 26 00:37:42 I am having some issues getting a base yocto image to boot on the beagle bone black, u-boot appears to boot, but I get errors such as "No valid environment area found", "unrecognised filesystem type" etc. I am following the instructions in the README.hardware, but it doesn't specify how to partition the sdcard. I tried having yocto generate a ".sdcard" image to avoid this, but that failed.. I did this trivially for an imx6 wandboard, but I have had no luc Aug 26 00:37:42 k with the Beagle Bone Black.. Any help would be much appreciated Aug 26 01:01:59 dhirc: maybe https://www.yoctoproject.org/downloads/bsps/fido18/beaglebone helps? there's partitioning info at least Aug 26 01:09:17 ntl: thanks for the response. the thing that is missing for me are the steps to partition the sdcard like this. I found a blogpost somewhere and followed it, but still no luck. I find it strange that all the docs refer to cylinders vs sectors as this is deprecated by fdisk. Adds some confusion. Aug 26 01:49:21 hm, i see distro-feed-configs but nothing for apt feeds Aug 26 01:49:41 and google is amazingly unhelpful on this topic Aug 26 01:50:04 anybody got an example of apt package feeds? Aug 26 01:50:16 * nerdboy has always used opkg until now Aug 26 01:50:47 the generic debian apt feed stuff isn't much more help either... Aug 26 02:00:15 nerdboy: did you read package_deb.bbclass and the python modules? could be of use, iirc we generate and use feeds for the rootfs population the way we do for the other package manager Aug 26 02:22:03 is that really the best/only place to look for setting up apache/sources.list? Aug 26 02:23:25 something like apt-feed-configs.bb and a blurb for apache .conf would be a little more obvious... Aug 26 02:24:20 * nerdboy on a short schedule with no time for extended "research" task Aug 26 02:24:41 fallback to opkg for now... Aug 26 02:26:10 i really doubt many people use debian packaging / apt with yocto at the moment Aug 26 02:26:16 * kergoth shrugs Aug 26 02:26:48 i would just use ipk but somebody requested it... Aug 26 02:28:17 on another topic, should i not use TUNE_FEATURES_append? Aug 26 02:28:34 and set DEFAULT_TUNE instead? Aug 26 02:32:43 i think that would be preferred, generally, yes Aug 26 02:33:13 do you ever TUNE_CCARGS_append ? Aug 26 02:36:24 the only time i've done so is when using custom / external toolchains with custom tuning arguments, if i don't want to override the entire tune files Aug 26 02:37:44 thanks Aug 26 02:38:09 * nerdboy wondering about that for the last couple of weeks... Aug 26 02:55:42 and otavio, i got another dev board, rebuilt the imx28evk stuff on master with -fslc and everything boots up fine Aug 26 02:56:14 once you find the magic boot sdcard jumper info... Aug 26 02:58:23 e.g. https://github.com/MentorEmbedded/meta-sourcery/blob/master/conf/distro/include/sourcery-tuning.inc#L14 (re TUNE_CCARGS) Aug 26 02:58:27 except for the slight interrupt deadlock on boot, it looks good Aug 26 02:59:09 [ 4.409925] *** DEADLOCK *** <= whee! Aug 26 02:59:37 i think we can fix that... **** ENDING LOGGING AT Wed Aug 26 02:59:59 2015