**** BEGIN LOGGING AT Fri Jan 25 02:59:59 2013 Jan 25 08:11:54 how is the defconfig for the kernel, PN/MACHINE/defconfig, found by bitbake? I'm trying to bbappend another kernel recipe changing the machine and adding my own defconfig, but bit bake doesn't see it. If I put it in the layer with the kernel recipe, it finds it. Jan 25 08:12:48 http://pastebin.com/C5W2MNfJ Jan 25 08:39:29 Dear All Jan 25 08:39:54 I want to add EXTERNAL_TOOLCHAIN to compile AArch64 platform to OE Jan 25 08:40:32 i copied the metadata from linaro toolchain and modify Jan 25 08:40:33 uvan why? Jan 25 08:40:48 hrw is pushing more and more aarch64 support into oe Jan 25 08:40:59 but it always try to compile gcc-cross-intial Jan 25 08:41:30 how can i ignore the gcc-cross-initial compilation? i want to used glibc from external toolchain Jan 25 08:43:29 are you listening what I am saying? Jan 25 08:43:52 woglinde, yes, but i'm still not much understand what you mean Jan 25 08:44:11 did you mean that aarch64 still not much supported? Jan 25 08:44:28 it is supported but I think not in all recipes Jan 25 08:44:35 just trry it out Jan 25 08:44:44 sure, that's i want to try Jan 25 08:44:46 no need for external toolchain Jan 25 08:45:44 linaro patches are integrated in oe-core Jan 25 08:46:08 woglinde, actually i can do compile with linaro already Jan 25 08:46:26 but i want to used it with my own company toolchain Jan 25 08:47:11 and for that toolchain, i want to ignore all gcc-cross-initial compilation Jan 25 08:49:07 do you know what recipies need to modify if i don't want gcc-cross-initial compiled? Jan 25 08:56:13 hm Jan 25 08:56:16 sorry Jan 25 08:56:22 cannt help you Jan 25 08:59:23 I gurss you can set ASSUME_PRVODED_gcc-cross-initial Jan 25 08:59:31 eh guess Jan 25 08:59:48 We used to habe something like this... haven't tried this in ages. Jan 25 09:01:17 hi shoragan Jan 25 09:01:28 ah right yes Jan 25 09:01:32 hi woglinde :) Jan 25 09:01:39 ASSUME_PROVIDED Jan 25 09:02:21 yep... can't type this morning :-/ Jan 25 09:06:15 ASSUME_PROVIED? let me see if i can find something Jan 25 09:06:20 i'm also new to OE Jan 25 09:07:18 sure Jan 25 09:07:23 put it in local.conf Jan 25 09:08:16 ASSUME_PROVIDED += "gcc-cross-initial" should work too Jan 25 09:08:26 but there is maybee more to packages to add Jan 25 09:11:28 I'm trying Jan 25 09:11:40 gm bluelightning Jan 25 09:15:30 morning woglinde, all Jan 25 09:17:01 bluelightning does bitbake find out what I changed in a recipe Jan 25 09:17:21 its a hell to wait 20 minutes of rebuild when only changing oneline in post-inst Jan 25 09:18:06 woglinde: we produce a signature for each task based on its inputs yes Jan 25 09:19:59 so for changing postinst stuff using the produced sstate should working? Jan 25 09:20:15 woglinde: sorry I don't understand the question... ? Jan 25 09:20:19 again Jan 25 09:20:26 I change one line in postinst Jan 25 09:20:36 but bitbake recompiles the whole recipe Jan 25 09:20:47 thats okay for small packages Jan 25 09:20:57 but not for one like openjdk which needs 20 minutes Jan 25 09:23:18 woglinde: clearly if you change a postinstall it's going to have to run do_package, and as a consequence it figures it needs to run any task that depends upon that Jan 25 09:24:12 but couldd that not shortcutend? Jan 25 09:24:19 whith sstate Jan 25 09:24:45 sstate relies on those checksums Jan 25 09:25:41 if a task's checksum (i.e. the checksum of all inputs to the task, which includes all of the checksums of the other tasks the task depends on) has changed it assumes the output could be different and re-runs the task Jan 25 09:26:31 sure but it would be clever to look what changed Jan 25 09:26:41 and use help of sstate Jan 25 09:33:27 uvan: you have to take a look at external-toolchain stuff Jan 25 09:34:10 woglinde: it's not that simple unfortunately Jan 25 09:35:11 it already does look at what changed Jan 25 09:35:39 it's all done at a task level, and it uses variable dependencies to determine what the inputs are to a task Jan 25 09:36:49 uvan: meta-linaro: conf/distro/include and recipes-devtools/external-*/* Jan 25 09:37:25 gm hrw Jan 25 09:37:25 woglinde: see here for further details: http://www.yoctoproject.org/docs/latest/poky-ref-manual/poky-ref-manual.html#shared-state-cache Jan 25 09:38:35 good morning Jan 25 09:39:59 hi mckoan, hrw Jan 25 09:40:46 hi mckoan Jan 25 09:43:06 uvan: can we talk? Jan 25 10:09:16 hrw yes Jan 25 10:09:50 hi hrw Jan 25 10:11:33 hrw, actually i copied all from meta-linaro Jan 25 10:12:33 and also compiled meta-aarch64 from linaro too Jan 25 10:13:08 uvan: ok Jan 25 10:13:24 uvan: can I ask you what you are working on? Jan 25 10:13:49 uvan: cause at linaro we try to group all aarch64 porting teams Jan 25 10:15:02 i'm trying to build OE for our chip will come out soon that based on aarch64 Jan 25 10:15:54 and i want to used my own toolchain for that to test Jan 25 10:16:02 ok Jan 25 10:16:20 uvan: so take a look at recipes for external linaro toolchain and adapt them to your toolchain Jan 25 10:16:54 but i always stuck at compile gcc-cross-intial, because aarch64-oe-linux not supported Jan 25 10:17:45 uvan: you probably have different TARGET_PREFIX Jan 25 10:17:56 but i though when i used external toolchain, i don't need to compile gcc-cross, gcc-cross-intial, ....because already have Jan 25 10:18:19 uvan: yes, but how much time it took? Jan 25 10:18:44 uvan: I think that usually it is faster to just add own gcc recipes then play with external toolchain integration Jan 25 10:19:42 i took 1 week already, :d Jan 25 10:19:53 uvan: I made meta-aarch64/recipes-devtools/gcc/ in ~1h including builds Jan 25 10:20:03 still not understand how can i play with external toolchain Jan 25 10:20:29 using linaro-external toolchain as defined in meta-linaro also failed at there Jan 25 10:20:50 uvan: it was configured for armv7a Jan 25 10:21:05 I just pushed change to allow to override it Jan 25 10:22:01 hrw: so right now, can you compile meta-aarch64 with external linaro toolchain? Jan 25 10:22:16 uvan: not tried Jan 25 10:22:46 i set TCMODE="external-linaro", EXTERNAL_TOOLCHAIN="" but build also failed at gcc-cross-initial Jan 25 10:23:57 letme try Jan 25 10:26:52 uvan: if you have problems with aarch64 support in OE then please report bugs at http://bugs.launchpad.net/linaro-oe Jan 25 10:27:10 uvan: aarch64 porting issues at http://bugs.launchpad.net/linaro-aarch64 Jan 25 10:27:55 ok hrw Jan 25 10:29:37 hrw; i'm just a new jump to OE. and aarch64 also. lot of thing still not much understand Jan 25 10:30:09 uvan: for start I would suggest using gcc-linaro 4.7 setup and get used to OE. Jan 25 10:34:48 yes, i'm also use now Jan 25 10:40:42 uvan: looks like unbroke external toolchain linaro. but it needs one patch to OECore ;( Jan 25 10:42:25 hrw: so can you tell me how that patch? Jan 25 10:49:06 uvan: first I have to take a look at armv7a cross toolchain to find out how to merge my changes. I do not want to unbreak armv8 and break armv7a Jan 25 10:51:36 does =+ in the recipes mean the same thing as += ? Jan 25 10:51:43 or are they typos? Jan 25 10:52:01 Zagor: no, =+ means prepend with a space rather than append with a space Jan 25 10:52:11 aha Jan 25 14:24:46 RP, you saw my PACKAGECONFIG update for cmake recipes? Jan 25 14:27:43 wasn't it merged already? Jan 25 14:28:08 http://git.openembedded.org/openembedded-core/commit/?id=d8c0ce5a4a27c8aa1d07fc15d6e000af725a51e6 Jan 25 14:29:36 Crofton|work: has been in master-next for a few days and then merged to master. All without me seeing it :) Jan 25 14:37:41 argh. who invented that crazy scheme of indents in bbshellparts... Jan 25 14:39:11 eglibc-package.inc has whole mix of indent styles... Jan 25 14:41:44 RP: I'll send patch for automake and gcc (info/dir removal) a bit later, testing them now Jan 25 14:41:59 JaMa: great, thanks Jan 25 14:43:19 WARNING: The recipe linux-libc-headers is trying to install files into a shared area when those files already exist. Those files and their manifest location a Jan 25 14:43:22 re: /OE/shr-core/tmp-eglibc/pkgdata/cortexa8-vfp-neon-oe-linux-gnueabi/linux-libc-headers Matched in manifest-cortexa8-vfp-neon-linux-libc-headers.package Jan 25 14:43:34 and first who will reply with "please do not mix spaces and tabs in commit": check email source! Jan 25 14:44:46 hrw: you should support my proposal to change that crazy styleguide.. Jan 25 14:45:05 hrw: and mutt does not mangle email source for me Jan 25 14:45:11 JaMa: 4/8 spaces to rule the world? Jan 25 14:45:12 afaik Jan 25 14:45:29 JaMa: thunderbird 'in theory' also does not Jan 25 14:45:30 hrw: 4 spaces for both types of tasks Jan 25 14:45:50 and no tabs except .patch files Jan 25 14:46:42 and vim syntax plugin highlighting every tab in .bb .bbappend .bbclass (.inc) Jan 25 14:47:38 anyway I got both linaro external toolchain working today. probably useless work but someone had to do it Jan 25 14:49:34 ugh, yes, i hate the tabs in shell spaces in python thing, partly because we don't have a vim bitbake indent plugin to automatically handle it Jan 25 14:53:38 for meta-smartphone/meta-efl/meta-webos* we're not using official styleguide because of that.. Jan 25 14:56:05 rp. oops I missed that Jan 25 14:56:45 + configure |10672 ++++++++++++++++++++++++++++++++++++++++------------------ Jan 25 14:56:52 heh, just to disable usb Jan 25 14:57:45 JaMa: I nearly suggested just dropping the configure bit since we don't need it Jan 25 15:29:28 RP: is it expected that transition to do_packagedata does not work correctly when rm_work (or OEBasic) is used? http://paste.debian.net/228595/ Jan 25 15:29:45 RP: in other builddir with OEBasicHash it seems to work Jan 25 15:38:22 JaMa: hmm, I hadn't really thought about OEBasic :/ Jan 25 15:39:02 JaMa: it may need a clean build directory, or at least remove the do_package stamps Jan 25 15:41:13 ok, I'll clean tmpdir Jan 25 15:41:33 removing do_package stamps would rebuild everything too I think Jan 25 17:24:35 moin Jan 25 17:41:42 kergoth: i did some searching through the list archive, but didn't really find the thread(s) you pointed out Jan 25 17:42:27 i did, however, find a set of patches on pokey-contrib that adds read-only rootfs support, including some new bbclasses and whatnot Jan 25 17:42:41 is that at all related to what you were referring to? Jan 25 17:44:36 which threads are you looking for, again? Jan 25 17:45:28 the latest read only rootfs threads for oe-core are [OE-core] [PATCH 00/12] read-only rootfs support and [OE-core] [PATCH V8 0/6] read-only rootfs support, as far as i can see Jan 25 17:45:43 building/configuring a read-only root Jan 25 17:46:35 thanks Jan 25 18:19:24 I'll have to check out those read only threads. I've been doing it by using an initramfs to create an overlay file system before mounting my root as r/o Jan 25 18:27:06 any reason why knotty is the default bit bake ui vs knotty2? Jan 25 18:27:07 * ensc|w runs oe for ages on ro nfs rootfs... Jan 25 18:27:36 afaik, knotty *is* knotty2 is recent bitbake Jan 25 18:28:07 ah, I must be between. Mine defaults to knotty and has both in the ui/ folder. Jan 25 18:28:24 1.15.2 Jan 25 18:28:55 ensc|w: how did you handle config files in your ro rootfs Jan 25 18:28:56 ? Jan 25 18:30:07 sr105: they are part of machine specific recipes or moved to /var Jan 25 18:30:45 So you'd symlink or patch the offending program to use /var? Jan 25 18:30:49 or, sometimes I bind-mount them to /etc Jan 25 18:31:24 and if you need persistence in one of those files? like wifi config, for example. Jan 25 18:31:55 persistence? Jan 25 18:32:17 sometimes you need a config file to survive a reboot Jan 25 18:33:24 /var is on an mmc card and surives a reboot (usually ;)) Jan 25 18:33:33 for our device, we use the overlay to allow saving of settings and such. But if the unit gets hosed for some reason, we can just wipe the overlay and it's effectively a factory reset Jan 25 18:33:41 ah Jan 25 18:33:42 bind-mount is for stuff like /etc/resolv.conf Jan 25 18:34:37 once we had it setup, it seemed to be an easier solution than tracking down offending apps Jan 25 18:34:39 I use a similar setup for a vdr machine; /var/lib/vdr/etc is aufs3 mounted there Jan 25 19:55:41 I curious is there a recipe out there for omapl138 running a 3.x kernel? I see there is 03.22 PSP but recipes I have found are using a 2.6 kernel. Jan 25 19:58:27 better aks on beagle Jan 25 20:10:18 will do thanks. Jan 25 20:38:57 In my custom image, what would be the syntax to append a package to IMAGE_INSTALL if and only if I was building for a certain MACHINE ? Jan 25 20:43:14 IMAGE_INSTALL_append_mymachine = " foo" Jan 25 20:43:20 standard use of OVERRIDES Jan 25 20:49:51 jkroon: or add MACHINE_ARCH packagegroup which will contain only extra packages you want for each MACHINE Jan 25 20:50:40 that will allow you to add more extra packages to all installed MACHINEs while upgrading that packagegroup Jan 25 21:05:12 thanks Jan 25 21:32:00 hm coreutils needs acl now Jan 25 21:32:32 * kergoth sighs at the creeping dep issues and how much gets built by a core-image-base Jan 25 21:32:37 we need to spend some quality time whittling at that Jan 25 21:33:44 be lucky not to compile opnejdk Jan 25 21:33:49 openjdk Jan 25 21:33:51 hehe Jan 25 22:59:50 kergoth: I do keep trying and complaining when people add stuff. Doesn't get me much appreciation though ;-) Jan 25 23:06:05 heh, such is life as a maintainer Jan 25 23:06:36 gn **** ENDING LOGGING AT Sat Jan 26 02:59:59 2013