**** BEGIN LOGGING AT Thu Apr 11 02:59:59 2013 Apr 11 07:10:15 hi everybody Apr 11 07:10:46 hi Apr 11 07:10:49 Is immediate expansion ":=" allowed on RDEPENDS_${PN} Apr 11 07:11:11 ??? Apr 11 07:11:45 I try RDEPENDS_${PN} := "${@oe_filter_out('abcd', '${RDEPENDS_${PN}}', d)}" Apr 11 07:12:08 and get Failure expanding variable RDEPENDS_${PN}[:=], expression was ${@oe_filter_.... Apr 11 07:13:48 why you need such a construct? Apr 11 07:14:04 maybee there is something wrong in other recipes for RDEPENDS Apr 11 07:14:29 Well as it shows i am trying to filter out a rdepend Apr 11 07:14:39 which resides in another layer Apr 11 07:16:39 I mean the recipe that has this rdepend resides in another layer Apr 11 07:17:00 So I can't simply '#' it :) Apr 11 07:17:44 I've used this successfully before with SRC_URI etc but unable to understand why this is failing in the current scenario Apr 11 07:24:34 hmm; with the meta-fsl-arm tree, I get build errors with perf - anyone know how I can exclude it (I don't need it) ? Apr 11 07:27:37 rink_: you can run hob to see what includes a specific package Apr 11 07:29:48 good morning Apr 11 07:37:37 hi mckoan Apr 11 07:37:52 abelal use bbappend in your layer and overwrite the RDEPENDS Apr 11 07:40:07 woglinde: like RDEPENDS = "" Apr 11 07:40:09 ?? Apr 11 07:40:27 I only want to filter out one of the rdepended packages :S Apr 11 07:40:47 I am using a bbappend to do it already Apr 11 07:43:35 good morning Apr 11 07:43:59 gm Apr 11 07:44:07 hi ant Apr 11 07:44:32 hi ant_work, woglinde Apr 11 08:01:55 rink_: https://lists.yoctoproject.org/pipermail/meta-freescale/2013-April/002129.html Apr 11 08:21:05 morning all Apr 11 08:25:21 hello Apr 11 08:28:25 morning all, hi bluelightning , hi hrw Apr 11 08:30:36 hi all Apr 11 08:49:31 hi mckoan Apr 11 09:21:03 hi all Apr 11 09:21:25 hi pb_ Apr 11 09:21:44 hi bluelightning Apr 11 09:59:28 gm pb Apr 11 10:06:35 hi woglinde Apr 11 10:14:45 woglinde: ping Apr 11 10:14:53 any thoughts? Apr 11 10:14:58 abelal yes? Apr 11 10:15:38 as I said make your own version of the recipe with bbapends and change RDEPENDS Apr 11 10:18:46 well, not so much making your own version as customising the existing one via the bbappend Apr 11 10:20:18 bluelightning hm with bbapend you need a another version then orginial package otherwise it might be difficult for opkg to decide which version to install Apr 11 10:20:42 or does opkg support please load package foo only from this archive Apr 11 10:21:03 woglinde: that's what PRINC was for earlier, and now we have the PR server to handle the incrementing on signature change automatically Apr 11 10:22:37 bluelightning you still have the problem when the version in the original is changed and you did no rebuild your own Apr 11 10:22:59 hm hm Apr 11 10:23:25 woglinde: sorry, I'm not following... ? Apr 11 10:24:46 lets say you are based on angstroem Apr 11 10:25:02 angstroem bumps number up and put it in repo Apr 11 10:25:15 you are on vacation and can not rebuild Apr 11 10:25:21 whats opkg now installing Apr 11 10:25:46 what do you mean by "based on" ? Apr 11 10:29:05 bluelightning if you do not want maintain a whole distribution Apr 11 10:29:29 well, then you take your chances then Apr 11 10:30:38 it's no different than PPAs in ubuntu; if the PPA maintainer lags behind for whatever reason, the main distro packages can overtake the PPA Apr 11 10:31:36 you could change the PR value in some way that meant it was always above rXX Apr 11 10:31:52 ;) Apr 11 10:32:10 okay lunch now Apr 11 11:24:25 bluelightning: and not everything can be overriden in bbappend Apr 11 11:24:37 do_install_append() from original recipe for example Apr 11 11:25:06 hrw: true; you can use a big hammer and do _forcevariable I guess Apr 11 11:25:13 most of the time that's not an issue Apr 11 11:25:44 most of time you do not have such _append() :D Apr 11 11:25:47 and generally when that sort of thing happens it may be time to step back and look at whether the structure in the original recipe or the appends between needs to be changed Apr 11 11:35:43 bluelightning: could you tell me the difference between PN and BPN variables ? (I know that PN holds the name of the package, according to oe-core documentation.conf) Apr 11 11:36:09 I'm having an error with variable ALLOW_EMPTY not being package specific Apr 11 11:36:15 http://paste.debian.net/248867/ Apr 11 11:37:11 looked at the definition of that variable in some oe-core recipes, and found out that sometimes BPN is used, and sometimes it's PN Apr 11 11:38:56 examples: 'ALLOW_EMPTY_${PN} = "1"' and 'ALLOW_EMPTY_${BPN}-binaries = "1"' Apr 11 11:40:25 (and what is ALLOW_EMPTY used for btw ?) Apr 11 11:40:33 apelete: PN and BPN are the same except when using a different variant via BBCLASSEXTENDS or multilib Apr 11 11:41:07 apelete: e.g. if you have an xyz recipe with BBCLASSEXTENDS = "native", xyz would have PN and BPN = "xyz" Apr 11 11:41:27 but for the xyz-native variant, PN would be xyz-native and BPN would be xyz Apr 11 11:41:39 ALLOW_EMPTY_${BPN} is probably bogus Apr 11 11:41:57 since the package should almost always be named using the full PN Apr 11 11:42:06 ok, I see Apr 11 11:43:58 bluelightning: and just out of curiosity, do you know what ALLOW_EMPTY is used for ? Apr 11 11:47:16 apelete: if set to 1 it tells do_package to create the package even if it's empty Apr 11 11:47:29 apelete: the default behaviour is to not create the package in that case Apr 11 12:26:02 ok, thanks Apr 11 13:04:59 Heads up, don't upgrade your host gcc to 4.8.0 if plan to build a gcc 4.7.2 based cross toolchain and want it to compile. Apr 11 13:07:37 khem: ^ Apr 11 13:09:06 georgem: what's the error? Apr 11 13:09:19 build-cross/gcc-4.7.2/libgcc/libgcc2.c: In function '__absvdi2': Apr 11 13:09:23 build-cross/gcc-4.7.2/libgcc/libgcc2.c:273:1: internal compiler error: Segmentation fault Apr 11 13:09:30 georgem: does gcc-4.8 (patches from khem on oe-core ML) build with gcc-4.8 on host? Apr 11 13:10:08 georgem: I can confirm your issues with host gcc 4.8 Apr 11 13:10:40 bluelightning, what was my bug wrt to libudev and dev packages? Apr 11 13:10:45 I think it is fixed :) Apr 11 13:11:01 Crofton|work: it is fixed yes, I have marked it as such already I think Apr 11 13:11:11 JaMa: Haven't tried, I just downgraded to host 4.7.2 so I could build my toolchain. WebKitGTK 2.0 requires gcc >= 4.7 so I need 4.7.2 based toolchain in a hurry to continue my work. Apr 11 13:11:17 I can doubel check it Apr 11 13:11:48 erbo: can you test khem's patches for gcc-4.8 on your host, if you still have gcc-4.8 there? Apr 11 13:12:05 Crofton|work: https://bugzilla.yoctoproject.org/show_bug.cgi?id=4136 Apr 11 13:12:20 I think that ricers with gcc-4.8 on host should use gcc-4.8 in oe-core too :) Apr 11 13:12:36 ok, my sdk is good now Apr 11 13:12:43 JaMa: lol Apr 11 13:12:45 JaMa: I haven't, but a colleague might have.. I'll ping him Apr 11 13:13:18 there is a watcher watching the watcher? Apr 11 13:13:20 erbo: ok, thanks Apr 11 13:13:23 JaMa: lol. Didn't really want gcc 4.8 anyway but I updated arch linux yesterday to grab the newest versions of several other packages. Apr 11 13:13:48 Crofton|work: I'm not entirely sure what that's for to be honest Apr 11 13:14:11 georgem: it's default in arch? lol, gentoo unmasked 4.7.* recently and 4.8 will stay masked for a while (with good reasons) Apr 11 13:14:13 so donot use arch linux Apr 11 13:14:14 arch seems to be the bleeding-edge distro of choice... Apr 11 13:14:40 I like arch linux Apr 11 13:14:52 your collegues use it too? Apr 11 13:15:12 one does Apr 11 13:15:24 * JaMa usually switches gcc-native version (in gentoo) together with GCCVERSION change in OE Apr 11 13:15:28 we actually try our builds on quite a few distros to make sure problems are flushed out Apr 11 13:15:32 you just have to be aware you'll hit breakages like these if you do use arch... Apr 11 13:15:45 at least ICE from host's gcc is not OE fault for sure Apr 11 13:17:22 georgem: http://lists.linuxtogo.org/pipermail/openembedded-core/2013-April/037763.html maybe it's the same cause as ICE in elfutils Apr 11 13:17:25 yeah, if you use bleeding edge you're going to run into problems. but downgrading packages is so easy Apr 11 13:18:11 erbo: ^^ Apr 11 13:18:33 JaMa: Someone on #gcc pointed me here: https://bugs.gentoo.org/show_bug.cgi?id=465268 Apr 11 13:21:24 Building a stonewall around you and creating a "development time capsule" leads to eventual problems too xD Apr 11 13:21:50 ? Apr 11 13:22:23 at least I would setup a buildserver for testing diffrent distros Apr 11 13:22:23 georgem: we do add patches to support new host-side upgrades/changes on a regular basis Apr 11 13:22:43 georgem: I also ran into that problem, I used khems GCC 3.8 patches to fix it Apr 11 13:23:01 I have a bug open in bugzilla about it Apr 11 13:23:09 georgem: but we can't fix ICEs in the host compiler if that is the problem you're having... Apr 11 13:23:13 bluelightning: Sorry, I wasn't implying that you didn't. Apr 11 13:29:40 I was just trying to say that I find the benefits of being aware of the newest developments in open source outweigh the occasional but inevitable problems experienced when running arch linux. To each his own though. Apr 11 13:30:23 georgem: sure, that's why we have different distro choices :) Apr 11 13:30:32 georgem: I'm not saying arch is necessarily bad either Apr 11 13:31:09 as long as you know what you're doing, or at least the way to figure it out, you'll be fine :) Apr 11 13:31:16 :) Apr 11 13:32:51 hmm something is wrong :/ doing clean rebuild is now trying to fetch many packages again even with files already in downloads Apr 11 13:33:13 bluelightning: is it possible side-effect of that srcrev postfunct? Apr 11 13:33:35 bluelightning: if do_fetch was "covered" by .done lock file before and now it's not enough? Apr 11 13:33:36 JaMa: the signature will have changed back if this is the first time you've built after updating Apr 11 13:33:50 JaMa: do_fetch will still run in the presence of .done, it just shouldn't do anything Apr 11 13:34:14 see my e-mail about checksums, sent few minutes ago Apr 11 13:34:48 I've noticed it when building evas, but I'm sure the same version of evas was built before -> shouldn't try to refetch it Apr 11 13:34:59 JaMa: not sure it's in my inbox yet Apr 11 13:35:00 WARNING: Failed to fetch URL http://kernel.org/pub/linux/utils/kbd/kbd-1.15.2.tar.bz2, attempting MIRRORS if available Apr 11 13:35:08 kbd-1.15.2.tar.bz2 was also there for sure Apr 11 13:35:10 JaMa: I saw that you created a new testing branch in meta-qt5, is it in a useful state? :) Apr 11 13:35:31 -rw-r--r-- 1 bitbake bitbake 1040140 Apr 10 2010 downloads/kbd-1.15.2.tar.bz2 Apr 11 13:35:31 -rw-r--r-- 1 bitbake bitbake 0 Apr 11 09:13 downloads/kbd-1.15.2.tar.bz2.done Apr 11 13:35:51 erbo: it's better, but not yet ready Apr 11 13:36:38 erbo: now I'm working with upstream to add -external-host-bindir param, to help OE builds, such change would be applied upstream (unlike hacks we were using before) Apr 11 13:36:50 why is there CCLD and not CXXLD? Apr 11 13:36:50 khem: any plans to fork systemd? Apr 11 13:37:03 heh Apr 11 13:37:15 JaMa: I tried the thing you mentioned earlier, I still dont get the empty --sysroot= Apr 11 13:37:32 note though that I am using -no-gcc-sysroot Apr 11 13:40:58 hmm, chromebook just took the plunge of doom Apr 11 13:41:11 xora what? Apr 11 13:41:30 I just ouright dropped it on the floor for no apparent reason Apr 11 13:41:38 ugh Apr 11 13:41:47 seems to work still luckilly Apr 11 13:42:03 good Apr 11 13:43:12 no idea WTF that was about!!!!! Apr 11 15:04:57 bluelightning: missing downloads files explained, silly me (accidentaly umounted mount --bind under running build) Apr 11 15:05:12 JaMa: ah, phew :) Apr 11 15:05:14 thanks Apr 11 17:15:21 re Apr 11 17:18:36 re Apr 11 17:20:08 re Apr 11 17:20:31 re? Apr 11 17:23:33 so i am coming from oe-stable-2009.03 and in that version there were post install tasks, is there a correct way of handling that in newer metadata with the newer bitbake? Apr 11 17:24:03 waynr: postinst tasks are almost the same as in 2009.03 Apr 11 17:24:31 waynr: re - short for "hi again" i.e. "re hi" Apr 11 17:28:17 okay well i noticed in the rootfs_ipk_do_rootfs function in run.do_rootfs.### that there is a block of code that disables run-postinsts and then runs a function called "remove_packaging_data_files" Apr 11 17:30:46 i think the latter can be disabled in configuration file but i am not sure how to prevent the disabling of run-postinsts since it is based on whether or not var/lib/opkg/status has the text "Status:.install.ok.unpacked" Apr 11 17:31:26 do you have read-only images? Apr 11 17:32:31 yeah we generally produce rw images and use a postinst task on first boot to modify /etc/fstab such that the rootfs for subsequent boots is ro Apr 11 17:44:34 i think this has to do with OPKG_ARGS including "--force_postinstall" Apr 11 17:46:19 which is set in meta/classes/package_ipk.bbclass Apr 11 17:46:46 ah sorry i am speaking relative to yocto project directory in oe channel Apr 11 17:46:46 waynr: that's about ensuring the postinstalls happen during do_rootfs Apr 11 17:47:22 which I would assume you'd want Apr 11 17:47:31 bluelightning: in oe-stable-2009.03 postinstalls happen during first boot Apr 11 17:48:20 waynr: ok; now as you may have seen we try to run them and only run them on first boot if they've failed to run during do_rootfs Apr 11 17:48:43 yeah Apr 11 17:48:54 I have to admit I thought that functionality was quite old... Apr 11 17:49:06 esp. for ipk Apr 11 17:49:10 so i should make sure my postinstall fails during do_rootfs? Apr 11 17:49:43 only if it absolutely can't run on the host Apr 11 17:50:40 well since we want to boot read-write on first boot for things like ssh key generation and touchscreen calibration, and read-only on subsequent boots yeah we need a post install Apr 11 17:51:11 well we could use a temporary init script...but that's what /etc/init.d/postinst is for :D Apr 11 17:51:17 ah right, yes that makes sense Apr 11 17:51:45 in which case yes, just check if $D is assigned in the postinst script and exit 1 if it is Apr 11 17:56:20 thanks bluelightning, it looks like my main problem was an incorrectly-named "pkg_postinst_" script Apr 11 18:07:34 ah ok, np Apr 11 20:51:49 I'm looking for a clue, hopefully someone here can help me.. Apr 11 20:51:59 I have two difference classes adjusting variables.. Apr 11 20:52:23 one of the classes is doing it in a "python () { ... }" function, the other in a python __anonymous () { ... } function Apr 11 20:52:33 What is the difference between the two functions? Apr 11 20:53:02 I'm seeing the 'python ()' run multiple times and -before- the __anonymous function... Apr 11 20:59:27 essentially no difference afaik. __anonymous i think is an old form, no point using it Apr 11 21:01:03 ok.. so they are essentially the same.. and order then is however it happens to be Apr 11 21:01:13 ok.. that makes a lot more sense to me then.. Apr 11 21:03:17 hms with latest kernel qemu-native builds are broken on ubuntu 12.10 Apr 11 21:05:18 target qemu does not build on qemuarm too **** ENDING LOGGING AT Fri Apr 12 02:59:58 2013