**** BEGIN LOGGING AT Mon Sep 02 02:59:58 2013 Sep 02 04:58:19 hello. does anybody may test in master, on board imx6qsabresd build image qt4e-demo-image? i'v got next error, during build : http://pastebin.com/HrvudDfn Sep 02 05:00:40 its last master Sep 02 06:04:27 the python 3 patches on the mailing list means also bitbake, or only for recipes? Sep 02 06:04:45 i.e. will we be able to use python 3 on the host afterwards? Sep 02 08:05:55 good morning Sep 02 08:33:00 morning all Sep 02 08:33:39 bluelightning: good morning Sep 02 08:38:32 bluelightning: why is openflow depending on virtual/libc, what is this virtual/foo stuff? Sep 02 08:39:07 lpapp_: adding that to DEPENDS is bogus, libc is already a dependency by default Sep 02 08:42:39 bluelightning: yeah, I was about to remove... that was just a left-over from the old package, I believe. Sep 02 09:46:58 morning all Sep 02 09:56:03 g 30 Sep 02 10:13:49 bluelightning: net-snmp/net-snmp_5.7.2.bb:24: Could not inherit file classes/systemd.bbclass Sep 02 10:14:16 lpapp: are you back on denzil? Sep 02 10:14:16 I am trying to backport that package to denzil. Should I remove that inheritance with all its belonging in the net-snmp recipe, or I should copy that class to denzil? Sep 02 10:14:20 yes Sep 02 10:14:35 if you don't intend on using systemd just delete that inherit Sep 02 10:14:49 and the corresponding lines, right? Sep 02 10:14:59 yep you could also delete anything else relating to systemd Sep 02 10:15:45 thanks Sep 02 10:16:04 NOTE: Error expanding variable autotools_do_configure | ETA: --:--:-- Sep 02 10:16:09 NOTE: Error during finalise of net-snmp-recipe Sep 02 10:16:20 ERROR: Failure expanding variable EXTRA_OECONF, expression was --disable-embedded-perl --with-perl-modules=no --enable-shared --disable-manuals --with-defaults ${@base_conditional('SITEINFO_ENDIANNESS', 'le', '--with-endianness=little', '--with-endianness=big', d)} ${@base_contains('DISTRO_FEATURES', d)} which triggered exception TypeError: base_contains() takes exactly 5 arguments (2 gi Sep 02 10:16:26 ERROR: Command execution failed: Traceback (most recent call last): Sep 02 10:16:31 "UnboundLocalError: local variable 'result' referenced before assignment" Sep 02 10:17:35 EXTRA_OECONF += "${@base_contains('DISTRO_FEATURES', 'systemd', '--with-systemd', '--without-systemd', d)}" -> EXTRA_OECONF += "${@base_contains('DISTRO_FEATURES', d)}" Sep 02 10:17:39 might this be related? Sep 02 10:17:45 I can probably remove the whole line? Sep 02 10:18:40 yes, you've changed that to something invalid Sep 02 10:18:41 yeah, that seems to have made it built. Sep 02 10:18:43 build* Sep 02 10:22:47 bluelightning: where should I push the uthash recipe to, oe? Sep 02 10:23:20 sounds like a meta-oe candidate yes Sep 02 10:23:27 not meta? Sep 02 10:23:41 it is very lightweight, just four header files. Sep 02 10:26:10 bluelightning: the good practice is to create a package group for my own needs in my layer when having a custom distro image? Sep 02 10:27:35 any mods in local.conf will be taken into account at the next bitbake run even though most of the stuff is "cached"? Sep 02 10:29:10 lpapp: if it's not needed by most people and not needed by anything else currently in OE-Core, then it probably belongs elsewhere Sep 02 10:29:32 ok Sep 02 10:29:47 you don't have to create a packagegroup, you can if it helps you when creating a number of different images or adding and removing functionality at runtime Sep 02 10:29:55 to me it seems oe is more like a catch-all where it does not fit elsewhere... Sep 02 10:30:13 bluelightning: well, how would else would do it? Sep 02 10:30:23 I mean nicely. Sep 02 10:30:38 IMAGE_INSTALL is less nicer IMHO Sep 02 10:31:12 and yes, changing local.conf will invalidate the cache Sep 02 10:31:41 meta-oe is more the catch-all Sep 02 10:32:09 it's up to you, create a packagegroup if you feel it helps you Sep 02 10:32:17 well, I am seeking best practices. Sep 02 10:32:23 for the general cases. Sep 02 10:35:36 bluelightning: I do not even know an alternative nice way. Got a clue? Sep 02 10:37:14 lpapp: there isn't really an alternative other than packagegroups or IMAGE_INSTALL Sep 02 10:37:32 well, other than implementing your own IMAGE_FEATURES Sep 02 10:37:52 see meta/classes/core-image.bbclass if you want to look into the latter Sep 02 10:38:51 bluelightning: so you recommend IMAGE_INSTALL or packagegroup for a slightly extended core-image-minimal? Sep 02 10:38:57 say, 10-15 more packages maximum Sep 02 10:42:13 pretty much yep Sep 02 10:51:03 bluelightning: I mean... Sep 02 10:51:07 which one do you recommend Sep 02 10:51:44 depends upon the situation Sep 02 10:51:52 frankly for a single image I'd not bother with the packagegroup Sep 02 10:52:03 multiple images with the same block of functionality, different story Sep 02 10:52:54 bluelightning: then IMAGE_INSTALL, right? Sep 02 10:53:15 is that present properly for denzil though? Sep 02 10:54:22 of course, IMAGE_INSTALL has been used since the very earliest OE days AFAIK Sep 02 10:54:26 ok, there is the image.bbclass Sep 02 10:57:08 bluelightning: IMAGE_INSTALL_append = " package-name" -> so I should use this ... Sep 02 10:57:20 but will it grab the dependencies of that package-name into the image as well? Sep 02 10:57:26 I guess "yes", but just in case... Sep 02 10:57:32 yes Sep 02 10:57:40 iow, it is enough to specify the highest layer for that tree. Sep 02 10:57:58 if you mean the top-level dependency, sure Sep 02 10:58:00 Hey, I'm still having trouble with timezones. I've made a bbappend for tzdata that adds ${sysconfdir}/localtime to FILES_${PN}, and I can see /etc/localtime as in package/, image/ and packages-split/ for the package - And yet, it's not included in the image. Any advice? Sep 02 10:58:11 bluelightning: yes, thanks. Sep 02 10:58:56 My entire bbappend is FILES_${PN} += '${sysconfdir}/localtime', but since /etc/localtime definitively exists, I don't think I'd need any more? I want the one from the package itself. Sep 02 11:03:28 Stygia: given that you said before that /etc/localtime appears under packages-split then FILES is not the issue Sep 02 11:04:02 bluelightning, Hmm. So packages-split indicates the files that are supposed to be part of the image? Sep 02 11:04:20 Stygia: if the package in which the file appears is installed, yes Sep 02 11:04:40 bluelightning, Hmm alright. It's probably a bug somewhere in our configuration, then. The package is definitively installed. Sep 02 11:04:44 bluelightning, I'll look into it, thanks. :) Sep 02 11:06:33 bluelightning: interesting that nothin really uses this IMAGE_INSTALL in the layers.... Sep 02 11:06:39 hard to find an example how to do it right ... Sep 02 11:07:41 lpapp: there are loads of references to IMAGE_INSTALL in the layers I have here Sep 02 11:07:52 it's just a list of packages to install Sep 02 11:07:57 not much to it Sep 02 11:08:57 I do not see any to be honest ... Sep 02 11:09:05 grep -rn IMAGE_INSTALL_append ../../../poky-dylan-9.0.1/meta* Sep 02 11:09:07 gives no results. Sep 02 11:09:28 hmm, IMAGE_INSTALL is better. Sep 02 11:09:30 probably because it's not common to use _append to extend it in an image recipe Sep 02 11:09:40 but why on earth gives the documentation examples for the former then?! Sep 02 11:10:00 because AFAIK that talks about extending it from local.conf, not an image recipe Sep 02 11:10:19 it should document the common use case Sep 02 11:10:23 and give example for that one. Sep 02 11:10:24 it does Sep 02 11:10:27 no, it does not. Sep 02 11:10:37 it gives an example only for the rare use case. Sep 02 11:10:39 I'm sorry but it does Sep 02 11:11:13 bluelightning, Hey, "QA issue", does this stand for "quality assurance"? Sep 02 11:11:19 the very first subsection talks about providing your own custom image file Sep 02 11:11:22 bluelightning: it does not. Sep 02 11:11:26 https://www.yoctoproject.org/docs/current/poky-ref-manual/poky-ref-manual.html#var-IMAGE_INSTALL Sep 02 11:11:26 QED Sep 02 11:11:37 lpapp: http://www.yoctoproject.org/docs/current/dev-manual/dev-manual.html#usingpoky-extend-customimage Sep 02 11:11:52 Stygia: correct Sep 02 11:12:13 bluelightning: when I care about IMAGE_INSTALL, I go to its variable documentation. Sep 02 11:12:23 bluelightning: and I am expecting a good docs available from there. Sep 02 11:12:31 not at some random place. Sep 02 11:12:37 bluelightning, Alright, makes sense. The only other option would be "Question and Answer" but that makes little sense in context. Sep 02 11:12:38 it's far from random Sep 02 11:12:53 Stygia: that would probably be "Q&A" ;) Sep 02 11:13:29 it is random Sep 02 11:13:39 that section is not about defining IMAGE_INSTALL Sep 02 11:13:41 bluelightning, Ah fair enough. I'm not all that savvy vis a vi English business lingo. Sep 02 11:13:59 when I need to understand IMAGE_INSTALL, I do not wanna get a rare use case example right at its definition. Sep 02 11:14:05 *vis-a-visl Sep 02 11:14:11 *vis-a-vis... Sep 02 11:14:14 Stygia: fair enough Sep 02 11:14:33 lpapp: that's exactly what that section covers Sep 02 11:17:42 bluelightning: yes, that is the problem. Rare use case should not be covered in there. Sep 02 11:17:49 where it is defined. Sep 02 11:20:07 also, what is the reason why denzil is building hell slow? Sep 02 11:20:17 even with -j8 stuff, and dylan is significantly faster? Sep 02 11:20:20 by order of magnitudes? Sep 02 11:23:37 an order of magnitude is unexpected, but dylan is faster because it does less disk IO when packaging Sep 02 11:32:11 rburton: you are back? :) Sep 02 11:33:27 Hmm. Alright, I got /etc/localtime to be part of my image, and ntpd is running. However, the time is still approximately 15 minutes behind what it "should" be, although NTPD has clearly done something, as it at least now has the right hour set... Any idea how NTP could be running, with a correct timezone set, and still be behind roughly 15 minutes+ Sep 02 11:33:28 ? Sep 02 11:33:49 running rdate to some NIST server works just fine, but rdate AFAIK is obsolete in favour of NTP. Sep 02 11:34:11 lpapp: yes, with 437 email remaining to read (!) Sep 02 11:34:27 clearly "delete all" should be part of my monday morning workflow Sep 02 11:35:22 Aaah, never mind. The NTP recipe apparently only contains "dummy" NTP servers... I'll find some appropriate ones. :) Sep 02 11:37:09 Stygia: the ntp pool is likely the best choice, but we can't just enable it out of the box Sep 02 11:37:19 And I remember someone here telling me to request a NTP pool, I'll be nice. :P Sep 02 11:37:29 rburton: :-) Sep 02 11:37:40 Once we actually have it working we'll do that, for now we don't have any significant amount of devices anyway, so no significant load on their servers. Sep 02 11:38:19 bluelightning: hmm, net-snmp does not build with denzil. :( Sep 02 11:59:51 And when I say we don't have a significant amount, I mean we have less than 10. So I don't think we're violating too much of the spirit of public NTP pools. Sep 02 12:34:54 if i want to customise a config file in my image, e.g. /etc/fstab, is the correct way to do it to find out which recipe installs said file, and bbappend it? Sep 02 12:36:34 BCMM, Not sure about correct, but I move to the build directory and do find tmp/ -name file Sep 02 12:36:54 BCMM, Then you can see which packages puts it into packages-split/ Sep 02 12:37:16 /etc/fstab is base-files IIRC Sep 02 12:37:30 yep Sep 02 12:37:43 Stygia, bluelightning: thanks. and if i want to modify it, the correct way would be to bbappend base-files? Sep 02 12:38:06 BCMM: correct Sep 02 12:38:09 thanks Sep 02 12:38:28 BCMM, Heh, of course nothing beats just knowing which package provides something. Sep 02 12:38:41 bluelightning, Is there an easier way than to find tmp/ ? Something build-in perhaps? Sep 02 12:38:47 I suppose you can grep on the source directories... Sep 02 12:39:00 Stygia: well, in this instance I just did "git grep fstab" to confirm Sep 02 12:39:23 bluelightning, I suppose that works, too. Sep 02 12:39:30 on the device you can use your package management tools Sep 02 12:39:31 Stygia: it's not always that easy though, since it can be that there is no mention of the file if it comes from an upstream source and we're not patching it or otherwise needing to specify its name Sep 02 12:39:58 Stygia: the trouble is, we often can't know exactly what a package contains until after do_package Sep 02 12:40:00 maybe we should produce a file->package index at rootfs time Sep 02 12:40:13 rburton: we basically have that, in pkgdata Sep 02 12:40:39 rburton: i think somebody was telling me to use buildhistory to get that data Sep 02 12:40:46 rburton: we could do with an extra option to oe-pkgdata-util that would let you easily search that data for target file paths though Sep 02 12:41:17 i haven't had to actually do it yet, but when my project is basically finished and i'm just trying to make it smaller, identifying the package responsible for each file will be useful Sep 02 12:41:30 BCMM: buildhistory also includes a per-package list of files; the advantage there is you can see changes between builds as well Sep 02 12:41:57 oe-pkgdata-util? Sep 02 12:42:02 neat! Sep 02 12:42:02 bluelightning: if i'm not bothered about changes between builds, what's the simplest way to find out which package installs a file? Sep 02 12:42:15 BCMM: FYI, we're intending to provide some of this deeper image content analysis in Web Hob Sep 02 12:42:22 bluelightning: oh, looking at logs, it was you telling me to use buildhistory Sep 02 12:42:46 BCMM: grep pkgdata or buildhistory basically, assuming you've built the package containing it that is Sep 02 12:43:02 bluelightning: so what is pkgdata and how do i use it? Sep 02 12:43:26 BCMM: when packaging we output some metadata into a "pkgdata" directory in TMPDIR Sep 02 12:43:47 cd buildhistory, git grep FILELIST |grep /etc/fstab Sep 02 12:44:02 BCMM: this metadata gets used during packaging, image construction and also by some external tools Sep 02 12:44:05 -> packages/atom_pc-poky-linux/base-files/base-files/latest:FILELIST = /etc/default/usbd /etc/filesystems /etc/fstab /etc/host.conf /etc/hostname /etc/inputrc /etc/issue /etc/issue.net /etc/motd /etc/mtab /etc/nsswitch.conf /etc/profile /etc/shells /etc/skel/.bashrc /etc/skel/.profile /var/lock Sep 02 12:46:53 bluelightning: i haven't got buildhistory enabled at the moment. did you say pkgdata contains file lists? i can't seem to find them Sep 02 12:47:48 oh i see, under runtime/ Sep 02 12:59:38 hmm, it may be that the data under pkgdata isn't easily greppable Sep 02 12:59:48 for target file paths that is Sep 02 13:00:25 I'm having a look to see how hard it would be to whip up an option for oe-pkgdata-util to do it Sep 02 13:07:54 Hi guys, in poky there is recipe for mtd-utils. The recipe says: PACKAGES =+ "mtd-utils-jffs2 mtd-utils-ubifs mtd-utils-misc". I need only jffs2 support. Can I strip ubifs and misc? Is that the way of interaction with recipe I should be doing? Sep 02 13:14:27 Krz: you can just install mtd-utils-jffs2 and not the others Sep 02 13:15:18 rburton: Ok, understand now the 'PACKAGES' variable, thanks Sep 02 13:16:15 PACKAGES is the list of packages that will be created Sep 02 14:18:50 Hey, I have an issue making my boxes have the proper time set, and remembering the time when it has already been set. NTPD is now running on the box, and it seems like the box got the proper time when flashed. However, if I keep it unplugged for a while, the time gradually lags behind. Any debugging steps I could take? Sep 02 14:20:49 sounds like your RTC is messed up... Sep 02 14:21:20 is it a predicable skew? Sep 02 14:21:33 rburton, Possible, yea. But the weird thing is, it used to completely reset to UTC and be off by several minutes, but after adding NTP, it's not skewed from the "correct" basis, at least. Sep 02 14:21:52 rburton, I'm not entirely sure. But it does seem like it doesn't increment time when turned off. Sep 02 14:22:05 rburton, Although it jumped backwards in time about a minute on of the times I rebooted it... Sep 02 14:22:19 *one Sep 02 14:24:41 rburton But yes. More or less predictable. Sep 02 14:24:57 rburton, The weird thing, I suppose, is that on reboot, although NTPd is running, it doesn't correct it again. Sep 02 14:25:42 odd Sep 02 14:25:47 hardware sucks Sep 02 14:25:58 rburton, Hmm, yup. I'll try and ask over at #hardware if I can somehow verify that the RTC is broken. Sep 02 14:27:52 stygia: ISC's ntpd will refuse to sync time if your local time is too far off. can't remember the limits off the top of my head Sep 02 14:28:26 zibri, Hmm alright, it didn't when it was just a few minutes, though. But I suppose that is pretty far off for hardware RTC. Sep 02 14:28:30 you can do ntpd -qg to force it Sep 02 14:28:46 stygia: hum, i think it should be able to handle a few minutes though Sep 02 14:29:06 Seems like hardware is preoccupied talking about music. Sep 02 14:29:08 #hardware Sep 02 14:29:17 "Normally, ntpd exits with a message to the system log if the offset exceeds the panic threshold, which is 1000 s by default." Sep 02 14:30:23 zibri, It says "ntpd: time slew +0.000000s", and I suppose it means I'm beyond the treshold. Sep 02 14:31:09 hum, i don't know what that means. it look strange. as if it didn't feel any need to adjust your clock Sep 02 14:33:28 Hi gang ; can anyone tell me if systemd is formally part of Yocto at the mo or is it only partly hacked in? It seems a bit unclear from what I can find... Sep 02 14:35:26 pev: its a supported option. just not default Sep 02 14:36:22 Ah cool, do you know if there's any documentation for enabling or is it genuinely just enable and it works? Sep 02 14:37:22 pev: the docs say how to enable it Sep 02 14:37:29 it should just work when enabled Sep 02 14:37:42 zibri, Hmm. Weirder yet, hwclock clearly returns a skewed date... \ Sep 02 14:40:23 stygia: right, there's at least one problem :) Sep 02 14:52:03 ops Sep 02 14:52:05 | install: cannot stat `/home/eren/sourcebox/oe/poky.git/build/tmp/sysroots/alix3d3/usr/lib/syslinux/ldlinux.sys': No such file or directory Sep 02 14:52:33 sholdn't it be present on syslinux-native package? Sep 02 14:53:37 Stygia: is it posible the RTC battery is knackered? Sep 02 15:09:28 rburton: Do you know where I should look? I've had a dig via google and can't see anything formal?? Sep 02 15:11:08 pev: http://www.yoctoproject.org/docs/current/mega-manual/mega-manual.html#selecting-an-initialization-manager Sep 02 15:15:16 i want to use a .bbappend to change the contents of the DEPENDS variable. is there a way i can modify the existing one, rather than overwriting it? Sep 02 15:16:31 i just want to remove one entry (i'm patching wpa_supplicant's config so it doesn't require dbus), and i don't like repeating the entire DEPENDS var because my bbappend will break if the upstream layer adds other dependencies Sep 02 15:17:47 BCMM: if you're using master you can do DEPENDS_remove = "dbus" Sep 02 15:17:57 rburton: Brilliant, thanks! Sep 02 15:18:00 rburton: what do you mean by "master"? Sep 02 15:18:12 BCMM: git master, not a release. this was added last week. Sep 02 15:18:28 oh, master like "trunk". yeah i'm on git Sep 02 15:18:30 bleeding edge, hehe Sep 02 15:19:42 rburton: thanks Sep 02 15:45:52 there are 2 similar functions for inserting opkg feeds Sep 02 15:46:01 image.bbclass: insert_feed_uris() Sep 02 15:46:11 rootfs_ipk.bbclass: ipk_insert_feed_uris() Sep 02 15:46:42 rootfs_ipk.bbclass: ipk_insert_feed_uris() creates opkg.conf in image temp directory (not rootfs) Sep 02 15:46:51 or the distro-feed-configs recipe in meta-oe Sep 02 15:46:59 (which is the only one i've made work) Sep 02 15:47:06 the other one creates the seperate feed configuration in the actual rootfs Sep 02 15:47:18 I have IPK_FEED_URIS in distro.conf Sep 02 15:47:29 and these uris are not in rootfs Sep 02 15:47:32 I have only base file Sep 02 15:47:38 now I am confused Sep 02 15:48:41 rburton: which one should I use? Sep 02 15:48:54 IPK_FEED_URIS never did to image Sep 02 15:49:21 eren: i use a fork of distro-feed-config Sep 02 15:49:48 i made a mental note that this all needs to be sorted out, the broken stuff in the image generation deleted/fixed Sep 02 15:49:57 and/or distro-feed-config made to handle more than ipkg Sep 02 15:50:08 if you want a nice task... ;) Sep 02 15:51:43 rburton: well, I will make my image depend on distro-feed-config and put the variables in distro.conf Sep 02 15:52:34 tf, Oh sorry, didn't see your reply to me. Sep 02 15:52:59 tf, That is exactly what was up. I ended up taking it apart and using a multimeter, my 3V battery only gave me ~1V. So that seems to have been it. Sep 02 17:12:21 Hello all! Sep 02 17:12:43 Anybody with some deep knowledge of linux/perf/kallsyms? Sep 02 17:13:11 Ihave an issues with a couple of symbols that i don't find in proc/kallsyms Sep 02 17:13:28 issue * Sep 02 18:20:06 Another fun one - I've implemented a package_postinst_${PN} rule in my recipe. I'm anticipating it should be both run during build and on the device itself, at what point should I be able to see it though, and can I easily provoke it with a specific bitbake command or command on the device to debug during development Sep 02 18:22:26 (the background being that I don't see the net result of my postinst occuring so I'm pretty sure it isn't running for some reason!) Sep 02 21:29:23 pev: if it succeeds at rootfs time it won't run on target Sep 02 21:29:41 pev: you can tell the difference by seeing if $D is set (not ${D}) Sep 02 21:29:56 pev: if you want to run on target, exit1 if $D is set Sep 02 21:39:47 is there a vim config file somewhere to make it syntax-highlight .bbs? Sep 02 21:51:28 nm, found kergoth's vim syntax on github. thanks. Sep 02 21:51:50 if i have a do_install_append() in my bbappend, does that replace the do_install_append() in the original recipe, or run as well as it? Sep 02 22:49:18 RP: I copied that from other recipe examples, but I cant seeing it actually being run in either case? Sep 02 22:49:42 RP: Just wondering how I could force it to be run explicitly to see why it's not being called... **** ENDING LOGGING AT Tue Sep 03 02:59:58 2013