**** BEGIN LOGGING AT Fri Aug 16 02:59:58 2013 Aug 16 03:03:29 the results are much more interesting than the reactions Aug 16 08:22:38 morning all Aug 16 08:23:46 morning bluelightning and #oe Aug 16 09:02:16 florian: ping Aug 16 09:02:43 XorA: pong Aug 16 09:03:03 florian: is it possible to find out why my emails to angstrom-distro-devel seem to drop into a black hole? Aug 16 09:05:57 XorA: hmm... I can check for obvious problems (full disk e.g.) right now, but if its more complicated it will have to wait till this evening Aug 16 09:06:26 florian: thats cool, nothing urgent, I see the standard 451 greylisting then acceptance in my logs Aug 16 09:06:40 florian: but I do not see them in archive Aug 16 09:19:15 XorA: might be better now... looks like mailman crashed Aug 16 09:19:28 florian: thanks, your the man Aug 16 09:21:31 bluelightning: hi. just sent RFC for the run.do_xxx symlink.. we'll see.. Aug 16 09:26:07 ndec: how does this work for python functions? Aug 16 09:26:59 bluelightning: can you name one? i can check. Aug 16 09:27:08 ndec: do_patch I think Aug 16 09:27:25 yep Aug 16 09:27:29 lrwxrwxrwx 1 ndec ndec 18 août 16 10:13 run.do_patch -> run.do_patch.26002 Aug 16 09:28:06 i am not making any difference between python and shell, since i create the link in exec_task(). Aug 16 09:28:24 right, and there is a run file for those tasks Aug 16 09:28:32 but as i said yesterday, i am touching things i am very unfamiliar with ;-) Aug 16 09:28:33 that's not an issue then :) Aug 16 09:28:52 also, i only do it for task, not functions. Aug 16 09:28:58 right Aug 16 09:29:17 I don't think I'm familiar with every inch of the bitbake code FWIW :) Aug 16 09:30:16 anyone is, really? Aug 16 09:30:17 ;-) Aug 16 09:51:03 bitbake make my brain hurt :-( Aug 16 09:51:40 you can tell from the classes I wrote that python is WAY down my list of languages Aug 16 10:51:21 hi all Aug 16 10:57:42 hi pb Aug 16 12:04:31 woglinde: I'm getting fetch errors when trying to install openjdk-7-jre: Aug 16 12:04:33 ERROR: Function failed: Fetcher failure for URL: 'http://icedtea.classpath.org/hg/release/icedtea7-forest-2.1/archive/c1c649636704.tar.bz2;name=openjdk;unpack=false'. Unable to fetch URL from any source. Aug 16 12:06:03 It seems that the web app which provides tarballs is not handling the query string variables/values as it used to do (guessing), since http://icedtea.classpath.org/hg/release/icedtea7-forest-2.1/archive/c1c649636704.tar.bz2 seems to work. Aug 16 12:17:53 mario-goulart hm Aug 16 12:18:03 whats in the logs? Aug 16 12:18:37 mario-goulart 2.1 is deprecated anyway I am on the way for 2.3 Aug 16 12:20:30 woglinde: here's the foll do_fetch log: http://paste.debian.net/25870/ Aug 16 12:21:45 mario-goulart did you tried the wget from console? Aug 16 12:24:23 hm latest jamvm makes some problems with mem allocation Aug 16 12:27:11 anyone seen Eric Benard lately? Aug 16 12:28:59 ~seen ericben Aug 16 12:29:00 bluelightning: I have not seen ericben. Aug 16 12:29:00 ericben <~ebenard@pac33-2-82-240-38-71.fbx.proxad.net> was last seen on IRC in channel #oe, 72d 3m 25s ago, saying: 'yes and if that's the case they may need to update /etc/security/limits.conf'. Aug 16 12:29:23 72 days... did he change his handle? I'm sure I saw him around more recently than that... Aug 16 12:29:42 unless infobot went MIA, I did see it rejoin the otherday Aug 16 12:30:26 bluelightning: I've asked because I wanted to include seeing him in real life (e.g. by some eukrea employee) Aug 16 12:30:36 fair enough Aug 16 12:31:21 maybe he has just very long vacation as I have pending dylan backport since May 28 Aug 16 12:45:28 no fallback? Aug 16 12:57:50 woglinde: yeah, I get a 404 error (wget -t 2 -T 30 -nv --passive-ftp --no-check-certificate -P /home/mario/src/ossystems-yocto-platform/build/downloads 'http://sources.openembedded.org/c1c649636704.tar.bz2' Aug 16 12:58:36 hm intressting Aug 16 12:58:44 some proxy in the way? Aug 16 12:58:50 oh Aug 16 12:58:56 right sources.openembedded.org Aug 16 12:59:01 that will not work Aug 16 13:26:20 Oh, damn. Pasted the wrong one. Aug 16 13:29:10 mario-goulart I am away for some time Aug 16 13:30:40 woglinde: ok. Thanks for your attention. I'll try to sort it out. Aug 16 14:11:46 re Aug 16 14:15:23 when i build a qt5 app and try to run it in an image which uses meta-qt5 i get an error saying: Aug 16 14:15:42 Failed to load platform plugin "xcb" Aug 16 14:15:48 hm Aug 16 14:16:02 did you check if there is an extra package for it? Aug 16 14:16:13 i'm using OE to build the application, and as far as i can tell the app builds okay Aug 16 14:16:16 opkg list | grep qt | xcb Aug 16 14:16:30 building != runtime Aug 16 14:18:17 $ tmp-qemux86-64/sysroots/x86_64-linux/usr/bin/opkg-cl -o tmp-qemux86-64/work/qemux86_64-poky-linux/core-image-x11/1.0-r0/rootfs list Aug 16 14:18:38 ? Aug 16 14:18:41 $ tmp-qemux86-64/sysroots/x86_64-linux/usr/bin/opkg-cl -o tmp-qemux86-64/work/qemux86_64-poky-linux/core-image-x11/1.0-r0/rootfs list | grep qt Aug 16 14:18:45 pastebin Aug 16 14:18:46 only gives Aug 16 14:18:47 please Aug 16 14:18:52 qtbase - 5.0.2-r3 Aug 16 14:18:53 ok Aug 16 14:19:02 hm okay Aug 16 14:19:20 i can use the native opkg to provide me the list of installed packages, as though i was on the target image Aug 16 14:19:27 by using the "-o" option Aug 16 14:19:40 opkg files pkgname Aug 16 14:19:46 should print out the file list Aug 16 14:19:53 (but it only works on master, it doesn't seem to work on dylan) Aug 16 14:21:02 http://pastebin.com/eZAAEseM Aug 16 14:22:07 woglinde: it seems i don't have anything xcb-related Aug 16 14:23:14 tlwoerner hm did you check the compile log if xcb support is build? Aug 16 14:23:30 looks like you have to explicit enable it Aug 16 14:24:05 I did qt5 build here myself and it it dont enable xcb Aug 16 14:24:12 non-oe Aug 16 14:24:35 DISTRO_FEATURES? Aug 16 14:27:21 hm let have a look at bb files Aug 16 14:27:52 the do_configure for qtbase looks for xcb and then concludes: Aug 16 14:28:01 Xcb support ............ system Aug 16 14:30:33 woglinde: I seems that it was a temporary problem on my network. I could build and install the java stuff. Sorry for the noise. Aug 16 14:34:24 tlwoerner okay Aug 16 14:34:33 tlwoerner and the compile log? Aug 16 14:36:40 woglinde: i see a number of places where it is compiling xcb files for a plugin Aug 16 14:36:48 e.g. compiling /home/trevor/build/yocto/tmp/fullbuild/qemux86-64/work/x86_64-poky-linux/qtbase/5.0.2-r3/qtbase-opensource-src-5.0.2/src/plugins/platforms/xcb/qxcbmime.cpp Aug 16 14:37:56 linking ../../../../plugins/platforms/libqxcb.so Aug 16 14:37:56 1944 mv -f libqxcb.so ../../../../plugins/platforms/ Aug 16 14:38:14 okay Aug 16 14:38:26 so it must be somewhere Aug 16 14:40:42 i wonder if i need to look more at x11? Aug 16 14:40:54 tlwoerner hm can you go to the ipk dir Aug 16 14:41:07 and search there for the package Aug 16 14:57:39 JaMa: it used to be that meta-qt5 dylan and master were quite similar. now master has got some big rework (and PACKAGECONFIG stuff). are you planning to have dylan upgraded similarly? Aug 16 14:58:09 and more generally, are you planning to 'maintain' qt5 for dylan? Aug 16 14:58:23 bugfix branches don't usually get big reworks on general principle, though :) Aug 16 15:06:58 kergoth: right. but meta-qt5 is not quite stable branch yet, still in dev. so rework is 'more possible' ;-) Aug 16 15:07:46 ndec make some merge request Aug 16 15:07:49 *g* Aug 16 15:08:13 well, at that point, i am not saying i need them, but just wanted to know if that will happen ;-) Aug 16 15:08:38 we are building our distro using dylan, and QT5, and we won't be able to switch to master anytime soon. Aug 16 15:08:53 ndec: ah, good point Aug 16 15:09:17 so, i need to understand what's the state/future of meta-qt5/dylan (hoping we won't have to maintain that ourselves..) Aug 16 16:05:46 I'm trying to solve a problem.. for some PowerPC systems we're working on, we need a 64-bit kernel, but want 32-bit userspace.. Aug 16 16:06:00 this is causing problems with the out-of-tree kernel modules failing the arch QA check.. Aug 16 16:06:13 they are properly building as 64-bit, but the QA checker says wait, you asked to build 32-bit.. Aug 16 16:06:47 I was hoping I could use overrides or something to check for kernel modules and machine arch and reconcile this in some way.. but I'm not sure I can even do that.. does any one have any suggestions (other then disabling the QA check..) Aug 16 16:15:51 fray: kind of sounds like PACKAGE_ARCH for the recipe is incorrect Aug 16 16:17:19 actually no, that QA check doesn't look at PACKAGE_ARCH Aug 16 16:17:31 it checks against TARGET_ARCH Aug 16 16:17:40 the issue is that the tune for the -kernel- is NOT the same tune for the userspace Aug 16 16:17:45 and AFAIK we have no real way to indicate that.. Aug 16 16:17:48 right Aug 16 16:18:16 I can't find an override that is only set when building kernel bits.. I'm tempted to add one.. but like I said, I want a generic fix -- not something specific to one bsp or one machine Aug 16 16:19:03 should be easy enough to define an override when building kernel or module(s)...... Aug 16 16:19:37 the other issue is what about packages that build both userspace software and modules together.. I've told my developers that it's invalid.. but I think in the general oe-core case, it's not.. (or am I wrong?) Aug 16 16:19:50 we always split the out of tree modules and userspace software... AFAIK Aug 16 16:21:26 fray: I think trying to do both in one recipe is ill-advised because it'll lead to problems specifically in the kind of situation you're dealing with Aug 16 16:21:48 so split recipes seems like the best approach Aug 16 16:22:18 fray: so how are you telling the system to use a different architecture for the kernel and userspace? Aug 16 16:28:30 the BSP does it by setting the karch to power64 Aug 16 16:28:34 I think.. Aug 16 16:28:59 ya Aug 16 16:29:06 require conf/machine/include/tune-ppce5500.inc Aug 16 16:29:07 DEFAULTTUNE ?= "ppce5500" Aug 16 16:29:07 MULTILIBS = "multilib:lib64" Aug 16 16:29:07 DEFAULTTUNE_virtclass-multilib-lib64 = "ppc64e5500" Aug 16 16:29:07 KARCH = "powerpc64" Aug 16 16:30:05 maybe the right approach is to check if karch is defined, if it is.. use that instead of the tune arch for comparison of modules (when the module bbclass is loaded) Aug 16 16:34:12 fray: there's no reference to a variable KARCH in poky (only in the kernel config itself...) Aug 16 16:34:39 fray: kind of seems to me that DEFAULTTUNE needs to be actually set for the kernel/modules and it isn't being Aug 16 16:35:05 since DEFAULTTUNE will influence TARGET_ARCH Aug 16 16:43:35 bluelightning the problem is that we don't want 64-bit PowerPC binaries.. they just waste space, unless you REALLY need more then 32-bit memory access.. Aug 16 16:43:52 but we do need a 64-bit kernel, so we can access devices (and potentially memory) beyond 32-bit Aug 16 16:44:23 on a lot of PowerPC systems that is very common, similarly on MIPS64.. it's not uncommon to use a 32-bit usersapce, and 64-bit kernel Aug 16 16:46:15 looking at our BSPs, the KARCH thing appears to be a variable we added to work around that problem already.. Aug 16 16:46:26 if KARCH is not set, then the kernel uses the tune arch.. otherwise it uses the KARCH.. Aug 16 16:46:39 fray: yes I understand the problem Aug 16 16:47:03 fray: however what the fact that you're getting these errors means that TARGET_ARCH *is* wrong Aug 16 16:47:04 KERNEL_EXTRA_CC_OPTIONS ?= "${@bb.utils.contains("KARCH", "powerpc64", "-m64", "", d)}" Aug 16 16:47:10 yes.. Aug 16 16:47:16 the question is how do we fix it though.. Aug 16 16:47:38 building lib64-linux-yocto -might-, but thats not a reasonable answer by itself Aug 16 16:47:50 what if you set DEFAULTTUNE for kernel/modules to the right value, instead of setting KARCH ? Aug 16 16:48:09 That is what I was going to try to do, but I can't find a way to do it Aug 16 16:48:37 I can't find anything to trigger that I'm running a kernel/module build... other then perhaps adding an override when I am.. Aug 16 16:49:02 something like: Aug 16 16:49:20 DEFAULTTUNE_kernel = "ppce5500_64" Aug 16 16:49:31 then in kernel/modules bbclass OVERRIDES_prepend = "kernel:"? Aug 16 16:49:47 but I'm not sure that is right either.. (could be append.. and is kernel safe to use.. probably not) Aug 16 16:52:48 fray: I'm not sure but it's possible we need an explicit KERNEL_DEFAULTTUNE that both modules and the kernel can use Aug 16 16:53:18 possibly.. I don't have any problem with making the BSPs define an alternative defaulttune for the kernel items if needed.. Aug 16 17:18:38 Can't you just set HOST_CC_KERNEL_ARCH to something appropriate? Aug 16 17:21:02 somehow the QA test needs to know what it's set to though... Aug 16 17:21:21 doing it via 'CC' args or similar isn't generic enough.. it really should be based on tune, KARCH, or similar.. Aug 16 17:22:10 pb_: will that change TARGET_ARCH for the kernel though? Aug 16 17:23:43 bluelightning: no, it won't, but I guess you could set that (or PACKAGE_ARCH perhaps) directly as well. Aug 16 17:24:20 hmm Aug 16 17:24:25 though, if this is just about keeping insane.bbclass happy, I would be fairly sanguine about just setting INSANE_SKIP for the kernel. Aug 16 17:24:39 this seems like something to bring up on the OE-Core list, there ought to be one way to configure this Aug 16 17:33:08 bluelightning I tried playing with the defaulttune and that doesn't work.. Aug 16 17:33:34 it doesn't work because it changes the toolchain being used from 'powerpc-wrs-linux-gnu' to 'powerpc64-wrs-linux-gnu'.. the problem is we never built (nor do we need to) a powerpc64 toolchain Aug 16 17:34:23 RP suggested reading the karch from the kernel configuration and using that when modules are being compared... Aug 16 17:34:29 that may turn out to be the easier way to do this afterall Aug 16 17:42:29 hmm OK, I guess I'd go with that then Aug 16 19:34:35 ndec: as kergot said, I don't plan to backport "bigger" changes like PACKAGECONFIG which requires changes in distro config Aug 16 19:34:49 ok Aug 16 19:34:52 thanks Aug 16 19:35:04 ndec: master branch is still compatible with dylan release so people who need newer stuff should use some tested revision from master Aug 16 19:46:30 * mr_science has been living off master for a couple of months Aug 16 19:47:06 no real issues until i pull and integrate changes... Aug 16 19:55:15 is there a quick guide to usin g devshell to fix a patch Aug 16 19:56:16 that's not an option, not directly Aug 16 19:56:21 the devshell task runs after do_patch Aug 16 19:57:08 If you set PATCHRESOLVE=user, it *should* automatically open a terminal if a patch application fails Aug 16 19:57:14 otherwise you can drill down into workdir manually Aug 16 20:17:37 i have a couple of drill bits for that... Aug 16 20:56:56 hms latest bitbake needs more than 3 gig Aug 16 20:57:25 or the virtual containers here are a bit broken **** ENDING LOGGING AT Sat Aug 17 02:59:58 2013