**** BEGIN LOGGING AT Thu Nov 12 02:59:58 2015 Nov 12 06:03:19 Anybody home? Nov 12 07:25:51 hi all Nov 12 08:13:02 good morning Nov 12 08:42:11 I really want to read about wic and devtool, there should be a page on yp.o that points me to relevant info for each of the cool tools being developed Nov 12 08:52:12 aha, wic: https://www.yoctoproject.org/docs/2.0/dev-manual/dev-manual.html#creating-partitioned-images Nov 12 08:52:18 devtool: https://www.yoctoproject.org/docs/2.0/dev-manual/dev-manual.html#using-devtool-in-your-workflow Nov 12 08:53:58 I should check if wic now creates EFI images better Nov 12 09:05:28 i have a question regarding buildhistory, does is store the SRCREV only for the packages that have ${AUTOREV} and not for those which have a SRCREV explicitaly defined? Nov 12 09:19:59 hi, is it intended that recipes with 'do_configure[noexec] = "1"' don't get licence checked? Nov 12 09:23:29 (in terms of checksum mismatch via LIC_FILES_CHKSUM) Nov 12 09:36:51 tobiash: that's not the right way to disable that Nov 12 09:37:15 tobiash: is the code you're building open or closed source? Nov 12 09:40:18 bluelightning: both Nov 12 09:40:30 just grepped through the source Nov 12 09:40:55 found also something like this here: Nov 12 09:40:55 poky/meta/recipes-core/glibc/glibc-collateral.inc:do_configure[noexec] = "1" Nov 12 09:41:09 yeah, that's not why that's setting that though Nov 12 09:41:42 so the correct way to disable configure would be to define an empty do_configure? Nov 12 09:42:04 do_configure isn't just about checking the license Nov 12 09:42:22 does the software really not have anything to be configured prior to compilation? Nov 12 09:42:34 nope (java stuff) Nov 12 09:42:40 I see Nov 12 09:42:50 well in any case, I'd suggest if there is a license in there other than your own that you may need to comply with you should consider setting LICENSE and LIC_FILES_CHKSUM properly rather than trying to disable them Nov 12 09:43:15 you can set LICENSE = "CLOSED", but if you're incorporating open source components that's probably inaccurate Nov 12 09:43:53 probably best to talk to a lawyer if there's any doubt (I'm not one, naturally) Nov 12 09:45:26 I've set that properly and noticed that when changing the checksum, there was no complaint about that Nov 12 09:45:58 after some debugging I noticed that the disabled configure step was the reason of the skipped license checksum check Nov 12 09:46:46 ah right, I see Nov 12 09:47:11 I wonder if we ought to consider splitting that out Nov 12 09:48:03 I also just thought if it would be better to do the license check as an independent task> Nov 12 09:48:39 currently trying that out Nov 12 10:12:08 bluelightning: first build looks promising, it did check the licenses of my recipe now even when skipping do_configure :) Nov 12 10:22:38 where can i add an .inc file that stores all the SRCREV_pn_? i'd like to avoid including the whole list in my local.conf Nov 12 10:22:46 in the machine conf? Nov 12 10:25:29 drou: your custom distro config is the best place Nov 12 10:27:56 thanks, i'll take a look to the manual for that Nov 12 11:45:58 how do I debug why a .bbappend isn't picked up? Nov 12 11:48:11 via bitbake-layer show appends i can see that it is found Nov 12 11:48:21 but i don't see the effect Nov 12 11:50:15 my append file contains a single line CMDLINE = "dwc_otg.lpm_enable=0 console=ttyAMA0,115200 root=/dev/mmcblk0p2 rootfstype=btrfs rootwait" Nov 12 11:50:40 which tries to override CMDLINE which is set in the recipe i am overriding Nov 12 11:51:20 the recipe to be overriden has ... rootfstype=ext4.... i am trying to change this into btrfs Nov 12 11:51:34 but the resulting kernel has the wrong commandline Nov 12 11:51:43 i already did a clean Nov 12 11:51:50 any ideas? Nov 12 12:00:13 fmeerkoetter, you can run bitbake -e and see history of changes of the variable you're interested in, it may help as maybe it's altered somewhere else again. Nov 12 12:03:49 * joshuagl finds the easter eggs in the release announcement and chuckles Nov 12 12:21:55 when building a bitbake target i'd often would like to see the whole build output of the target on the console Nov 12 12:22:00 how do i do this? Nov 12 12:23:27 as in the full compile log, install log, and so on? Nov 12 12:24:00 | tee Nov 12 12:24:47 rburton: yes, full compiler output Nov 12 12:24:50 for example Nov 12 12:25:04 fmeerkoetter: read the logs after the event, they're in the work/machine/recipe/version/temp/log.* Nov 12 12:26:41 (symlink one is the most recent) Nov 12 12:31:32 rburton: thanks. i knew about that one. thought there is an (additional) way Nov 12 12:31:37 a more direct way Nov 12 12:31:53 especialy when debugging/developing a single package Nov 12 12:32:53 figiel: thanks. i figured it out. my .bbappend is fine. kernel config was b0rked Nov 12 14:11:12 hello guys, I seem to have a sicrular dependency for the base-files recipe, any idea on how to investigate why this error is caused Nov 12 14:58:55 JaMa: I'm trying out the suggested _append_linuxstdbase line for samba - but I can't get my mind wrapped around that ... Nov 12 14:59:56 it looks as if there is no opposite to that linuxstdbase - so if (linuxstdbase); then use lsb + lsb-init; else use sysv-plain-init; fi Nov 12 15:03:35 do you have a hint how to solve that? Nov 12 15:05:18 sno: I haven't read the whole discussion with khem, but if you need to implement "else" then you can use interim variable to do that Nov 12 15:05:25 FOO_linuxstdbase = "bar" Nov 12 15:05:31 FOO = "blah" Nov 12 15:05:44 PACKAGECONFIG_append = " ${FOO}" Nov 12 15:05:52 JaMa: I'm simply not convinced restricting the init to libuxstdbase override Nov 12 15:06:41 JaMa: but that's a good idea and before having a time consuming discussion ;) Nov 12 15:07:44 JaMa: what I finally want/need is a reliable start/stop for samba when not using systemd Nov 12 15:49:54 hi folks - trying to build an SDK with the kernel headers, have added IMAGE += " kernel-devsrc", but it did seem to do anything? Nov 12 15:50:06 what exactly is "IMAGE"? Nov 12 15:50:09 i'm not aware of such a variable Nov 12 15:50:41 you'd have to add it to IMAGE_INSTALL in the image recipe you're running populate_sdk against, or add it to TOOLCHAIN_TARGET_TASK Nov 12 15:52:59 tried adding it to TOOLCHAIN_TARGET_TASK and it complain against about RPROVIDES. Nov 12 15:55:10 put the exact error message in a pastebin Nov 12 15:55:12 don't summarize Nov 12 16:14:52 why does oe_runmake add '-e'? just to deal with makefiles that do 'CFLAGS = ...' (rather than ?=) and the like? Nov 12 16:15:15 seems a bit dangerous for other stuff, in case something in the environment just happens to override something internal to the makefile... Nov 12 16:18:02 Ulfalizer: yep, pretty much. and yep it is dangerous, which is why folks are looking into removing it as the default. it's not the case for autotools/etc, just pure make based systems, and it's not ideal. it makes many projects 'just work', but it encourages folks to not even read the makefiles, which is bad. folks should read the makefiles, and override EXTRA_OEMAKE to explicitly pass all the vars we care about on the make commandline Nov 12 16:18:55 it's because we pass -e that we then have to force makeflags, which makes it not carry the -e into sub-makes, otherwise a toplevel makefile couldn't append to our env var without that being overridden by the env in the submake Nov 12 16:19:01 heh Nov 12 16:19:21 * kergoth is to blame for this (among other things) :) Nov 12 16:22:39 kergoth: okay, thanks for the explanation Nov 12 16:24:19 the original idea was to minimize the recipes, make as many things Just Work as possible, but it's too implicit and risky, and encourages ignorance, so in retrospect, not the best of plans :) Nov 12 16:27:54 kergoth: the enthusiasm of youth eh Nov 12 17:00:05 JaMa: do you have a resource where the other missing samba rdepends are from? Nov 12 17:00:16 it's always good to have a tester Nov 12 17:00:34 JaMa: my patch fixes more (eg. the aio), but not all of them Nov 12 17:13:46 sno: "resource" ? like which recipe provides them? Nov 12 17:14:23 JaMa: like the people reported the issue(s) ;) Nov 12 17:14:56 I've reported it as you can see in the e-mail Nov 12 17:15:15 ok, so you're my tester when sending 2nd patch ;) Nov 12 17:16:00 JaMa: but for now I'd like to have the one tested / committed Nov 12 17:19:24 sno: nothing will get merged to meta-oe until all these issues are fixed Nov 12 17:20:06 JaMa: why not? they're currently in there and half of them are fixed with that commit Nov 12 17:20:35 see my e-mails from today Nov 12 17:21:27 but I'll include your change in next test Nov 12 17:21:32 JaMa: the reply to my patch or other ones? Nov 12 17:21:37 just won't merge it until we have final and complete fix Nov 12 17:22:02 JaMa: don't get me wrong, I'm happy to provide another patch for the remaining fixes Nov 12 17:22:27 but with the way how waf configure works, there might be dozen sleeping similar ones Nov 12 17:23:22 there shouldn't be (at least in layers included in my world build) Nov 12 17:24:02 I understand there aren't --disable options for waf, but at least listing them all in DEPENDS (or even using sed to disable them explicitly) will help us get deterministic builds Nov 12 17:25:24 that's not what I mean - you discovered now 3 more: cyrus-sasl, dmapi and libbsd plus figure out that --disable-pam does only half of the job Nov 12 17:26:19 but how many hidden dependencies sleep in the configure you never discover nowadays, because we don't have to exotic dependencies around? Nov 12 17:26:38 and tomorrow - meta-oe grows and more hidden dependencies pop up Nov 12 17:32:06 that's why I have to re-execute test-dependencie script from time to time and fix new issues (from new or upgraded components) Nov 12 17:32:27 but if you manage to fix already known issues then I'll happily merge such fix Nov 12 17:32:34 JaMa: can we (and is it sane?) add patches depend on PACKAGECONFIG? Nov 12 17:32:35 I think about something like SRC_URI+= "${@base_contains('PACKAGECONFIG', 'ldap', '', 'remove-ldap-check.patch', d)} Nov 12 17:32:41 we can worry about unknowns later Nov 12 17:32:55 yes we can Nov 12 17:33:08 but if they overlap, it's bad Nov 12 17:33:21 bacause you won't ever know which combination user will have Nov 12 17:33:30 that's why sed can work better here (if possible Nov 12 17:34:09 sed is suit, but for larger expressions it becomes evit :D Nov 12 17:34:18 for those who have to maintain the expressions Nov 12 17:34:26 agreed Nov 12 17:34:32 especially when qotes are in the expression Nov 12 17:34:32 "if possible" :) Nov 12 17:36:01 I try to schedule the samba deps tomorrow - I wanted to prepare releases for MIME::Base32 and List::MoreUtils - let's see how far I come ... Nov 12 18:22:05 adding kernel-devsrc to IMAGE_FEATURES doesn't seem to have affect ... confused on how to get Kernel Headers in the SDK Nov 12 18:29:46 [08:50:43] you'd have to add it to IMAGE_INSTALL in the image recipe you're running populate_sdk against, or add it to TOOLCHAIN_TARGET_TASK Nov 12 18:29:48 as i said earlier Nov 12 18:30:06 IMAGE_INSTALL != IMAGE_FEATURES Nov 12 18:31:28 wooops ... hang on Nov 12 18:31:56 IMAGE_INSTALL += " \ Nov 12 18:31:56 kernel-devsrc" Nov 12 18:32:09 IMAGE_INSTALL is what I had ... still nothing. Nov 12 18:34:53 never mind .. something is going wrong, its getting stripped. Nov 12 18:35:57 i have a problem modifying a kernel config. i am using meta-raspberrypi as my bsp. this layer provides a defconfig for the kernel. Nov 12 18:36:22 this defconfig contains CONFIG_BTRFS=m Nov 12 18:36:31 i would like to change this into =y Nov 12 18:36:42 todo this i created a kernel config fragment Nov 12 18:36:49 named btrfs.cfg Nov 12 18:37:08 which i placed in one of my layers Nov 12 18:37:43 FILESEXTRAPATHS_prepend := "${THISDIR}/files:" Nov 12 18:37:51 SRC_URI += "file://btrfs.cfg" Nov 12 18:37:57 is used to pull it in Nov 12 18:38:26 i know that the btrfs.cfg is picked up Nov 12 18:38:50 bc. i see it in .....linux-raspberrypi/3.18.11+gitd64fa8121fca9883d6fb14ca06d2abf66496195e-r0/ Nov 12 18:38:56 alongside the defconfig Nov 12 18:39:17 but the resulting .config file has CONFIG_BTRFS=m Nov 12 18:39:24 any ideas? Nov 12 18:39:55 http://starbaseatlanta.com/ebaypics/stmangab20495.jpg Nov 12 18:40:01 missing the bb ... Nov 12 18:40:35 got it Nov 12 18:40:40 thanks :-) Nov 12 18:43:02 kergoth: there is no kernel-devsrc.bb in Dizzy ... any suggestion other than the obvious (upgrade :-)) Nov 12 18:48:40 fmeerkoetter: don't know about meta-raspberrypi, but some kernel recipes don't support configuration fragments. Nov 12 18:49:16 mario-goulart: that is good to know Nov 12 18:49:37 for that case i just have to provide my custom config? Nov 12 18:50:35 Yeah Nov 12 18:55:15 mario-goulart: meta-raspberrypi supports fragments since mid-2014. jfyi Nov 12 18:57:17 fmeerkoetter: ah, ok. Thanks. Nov 12 20:00:38 knu Nov 12 20:40:29 Say I want to have /var/log not be symlinked to /var/volatile , is there an easy way to do that? I've found the fs-perms.txt file that appears to control it, but it isn't clear how to override that file. Nov 12 20:41:25 fishey1: fs-perms corrects things in binary packages, e.g. a link defined there will be followed if a binary package tries to install a file into a volatile path. that won't handle making sure everything is what oyu want on target Nov 12 20:41:39 for what you want, you'd have to modify volatiles in the case of sysvinit, and tmpfiles.d in the case of systemd Nov 12 20:41:45 sad to say we don't have a single consistent wayt o handle that Nov 12 20:41:55 that said, i'm pretty sure /var/log is already handled via volatile Nov 12 20:42:01 so you shouldn't have to do anything in that regard Nov 12 20:42:33 kergoth: the issue is that bitbake gives me an error when trying to build the image when I try to have a real directory named /var/log/journal (appended to base_files) Nov 12 20:42:48 I don't want it to be volatile, I want it to be persistent Nov 12 20:42:48 that makes sense, becuase you can't package files into a volatile path Nov 12 20:42:51 gotcha Nov 12 20:43:08 error " * extract_archive: Cannot create symlink from ./var/log to 'volatile/log': File exists." Nov 12 20:43:09 you'd have to remove it both from fs-perms and also from volatiles in the initscripts package for sysvinit and tmpfiles for systemd Nov 12 20:43:25 we really need to improve the way we do this Nov 12 20:44:02 also we should spot cases where we're trying to package files in a volatile path and just error out at packaging time rather than at rootfs construction time Nov 12 20:44:32 kergoth: can I do that from a layer? this is really a distro configuration choice, and I'd like to avoid maintaining a fork of meta. Nov 12 20:45:04 (the fs-perms part, the tempfiles stuff I'm sure is fixable) Nov 12 20:47:29 fairly sure it'll search the layers for it Nov 12 20:47:41 so just set your layer to a higher priority, move it first in BBLAYERS, and put it in your layer in the expected path Nov 12 20:48:02 (technically the bblayers order is a factor here, not the bbfile layer priority, just mentioend that for consistency) Nov 12 20:48:08 so create a copy of the fs-perms.txt with my modifications to it (removal of that line) Nov 12 20:48:30 and make sure it's earlier in the BBLAYERS list? Nov 12 20:48:41 yep Nov 12 20:48:49 afaik that should work Nov 12 20:48:54 ok, I'll try it. Thanks. Nov 12 20:49:10 np Nov 12 20:49:36 but again, that'll just avoid the redirection when building packages, not what ends up on target until you deal with volatiles/tmpfiles Nov 12 20:53:08 ack Nov 12 20:57:05 I'd like to see us split off the base tmpfiles configuration, ship a tmpfiles population script/tool for use in sysvinit images, and deprecate volatiles Nov 12 20:58:44 Has anyone figure out how to search worth a damn for info on path relocation fixes? searching for relocation invariably gives you results relating to linker and loader relocation stuff, not relocation on disk Nov 12 21:20:03 Is there any way I can use KERNEL_MODULE_AUTOLOAD from outside of a kernel bitbake file? I have one kernel that I am wanting to share between several rootfs images, but each of those have different module autoload requirements. If I try to put the KERNEL_MODULE_AUTOLOAD line inside the rootfs BB file, it doesn't get picked up. Nov 12 21:21:20 afaik KERNEL_MODULE_AUTOLOAD just writes a file in /etc. you could write your own via a ROOTFS_POSTPROCESS_COMMAND function or a custom binary package/recipe Nov 12 21:21:35 specifically, write a .conf file to /etc/modules-load.d/ Nov 12 21:21:44 one line per module nmae Nov 12 21:21:50 see the man pages in any linux system for details Nov 12 21:22:32 fortunately it's trivial to work around, but it's possible we should rethink how it's implemented Nov 12 21:22:41 OK, thanks. I was just making sure that there wasn't any more magic that had to happen. Nov 12 21:23:32 bluelightning: It would be nice to do it external to the kernel build. Nov 12 21:24:28 indeed Nov 12 21:42:22 YoctoAutoBuilder: DAMN YOU Nov 12 22:14:46 Is there a may to create a directory in the root of the rootfs to use as a mountpoint specified in fstab? I tried install -d ${D}/mymountpoint, but do_rootfs failed Nov 12 22:19:18 edtyler: you would probably do that by defining a shell function to do that (using mkdir -p ${IMAGE_ROOTFS}/mymountpoint) then adding a call to that function to ROOTFS_POSTPROCESS_COMMAND Nov 12 22:19:33 (both of those would go in your image recipe) Nov 12 22:21:25 i hate the new limitations of ROOTFS_POSTPROCESS_COMMAND Nov 12 22:21:35 it used to be able to be any shell code, now it's function names only Nov 12 22:21:55 i get the idea, be able to add python function hooks as well, but it's really irritating to have to include a .inc just to touch a file in the rootfs Nov 12 22:35:13 bluelightning: Thanks. doing that or having a "firstboot" script called from rc.local was my fallback. So it sounds like the rootfs staging does not allow placing arbitrary directories or files in the root directory. Correct? Nov 12 22:42:19 edtyler: I'm not quite sure what you mean by rootfs staging Nov 12 22:45:56 bluelightning: Sorry. I meant the process by which files are transferred from each package's area to the place where the rootfs is built (build/tmp/work/cgtqmx6-poky-linux-gnueabi/bit-linux-image/1.0-r0/rootfs) in my case. I can put files most any plave I desire, except the root directory. Nov 12 22:47:20 edtyler: FYI the only way things get from ${D} to the final image is through packaging Nov 12 22:48:05 a few things could have prevented that from working - maybe that path wasn't packaged (it wouldn't have been by default); or alternatively maybe empty directories are being ignored, I'm not sure what the semantics are there Nov 12 22:48:26 either way the package probably ended up empty and thus was never produced, leading to an error when do_rootfs expected it to still be there Nov 12 22:48:58 bluelightning: Thanks for your help. I need to learn more about the phases of Yocto recipe operation. Nov 12 22:49:11 in this instance creating such a directory via the method I described is probably the correct approach vs. trying to do it through a recipe Nov 12 22:49:30 unless it's required by a specific application though I guess Nov 12 22:49:49 time to head out, g'night all Nov 13 00:13:09 slowly getting there working through the upstream submission queue.. https://gist.github.com/kergoth/fc5355c87d9f7f6aae6c Nov 13 00:13:11 whew Nov 13 01:24:10 * nerdboy losing the battle against kitty "cuteness" **** ENDING LOGGING AT Fri Nov 13 02:59:59 2015