**** BEGIN LOGGING AT Thu Dec 15 02:59:57 2011 Dec 15 06:33:36 anyone know how to get bitbake to not complain about chrpath on OSX? Dec 15 07:58:03 Hey friends, a question here: How can I write a bbclass-function which is executet after do_install but before do_install_append. I have tried it with "addtask blablub before do_install_append after do_install". But it doesn't work... Dec 15 08:27:45 silviof: do_install_append isn't a function by itself, it gets appended to the end of do_install Dec 15 08:28:06 silviof: can you be more specific about where you find a need to do this? Dec 15 08:41:30 bluelightning: ahhh - OK your "it gets appended to the end of do_install" helps - it is only concatenated. Then a "before do_install_append after do_install" is not able. Thanks Dec 15 08:42:54 silviof: no... but if we've got this going on in public layers and it's causing problems we ought to try to fix it Dec 15 08:49:49 bluelightning: I will rethink my logic. I have the requirement to write a layer between openembedded and some projects here in my firm. To conclude these severals projects i write some classes to provide some (but not too much :-) ) flexibility in setup of a basis image. ... Dec 15 08:51:20 bluelightning: Now I have a bbclass. It will take two variables one too setup a directory-layout and another variable to copy files to the image. But I need the flexibility after copy of files to start some work on image Dec 15 08:51:33 bluelightning: (sorry for my poor english) Dec 15 08:54:18 hm.. don't we have usbip recipe in any layer? Dec 15 09:18:35 morni Dec 15 09:18:36 n Dec 15 09:24:11 Hi all Dec 15 09:42:06 I'm moving from dev to 2011.03-maintenance. Is deleting my TMPDIR and rebuilding the safest approach? Dec 15 10:30:50 hi could anyone link me to a page that explains me how to use bitbake to build a self written program? Dec 15 10:31:14 such that it works for an image i built Dec 15 10:36:13 hmm, how can I force a package to get automatically rebuilt each time it is bitbake? I tried setting PV="${DATETIME}, I bitbake the package - works, bitbake it again - not being rebuilt Dec 15 10:36:21 any hints on what to hack? Dec 15 10:37:39 Jin^eLD: see PRINC Dec 15 10:38:39 JaMa: in what sence? you mean setting PRINC to $DATETIME? Dec 15 10:39:27 1. submitted MR for user feedback after update (use new opkg retcodes), started implementing image version feature Dec 15 10:39:30 oops wrong window Dec 15 10:39:52 yup something like PRINC := "${@int(PRINC) + @int(DATETIME)}" maybe Dec 15 10:40:58 hmm, why summing them up? date time will increase anyway? Dec 15 10:41:03 but let me try, thanks Dec 15 10:47:05 what defines if a package needs a LIC_FILES_CHKSUM or not? I see there are a few packages that do not have that; how can I avoid having to fake it if I have just a metapackage? Dec 15 10:48:15 JaMa: the PRINC thing doesn't do the trick either; same as setting PV to $DATETIME; bitbake mypkg -> gets built, then again: bitbake mypkg -> nothing happens anymore Dec 15 10:48:50 Jin^eLD: even metapackages have LIC_FILES_CHKSUM defined Dec 15 10:50:17 I guess I missed it then Dec 15 10:50:39 but anyway, that's the least of my problems right now Dec 15 10:50:41 any other hints on this auto-rebuild thing? Dec 15 10:50:48 arguably they shouldn't need to; it does seem a bit silly having a license on a package that doesn't contain any files. Dec 15 10:51:13 if it's just for internal use then you can set LICENSE = "CLOSED" in which case a checksum isn't required Dec 15 10:52:11 pb_: I know about the closed thing, but it's not really internal.. well I get around this by a hack: LIC_FILES_CHKSUM = "file://${COMMON_LICENSE_DIR}/GPL-3;md5=ab735e923e58a6e33297f8a617c45937" Dec 15 10:52:45 which is kind of not nice, but I really don't want to ship just a copying file for a package that has no sources Dec 15 10:52:46 pb_: with all oe metadata having some license specified in layer/LICENSE I don't mind repeating that ie task recipe has the same license Dec 15 10:53:23 JaMa: I did not know that, I guess I'll fix my stuff then Dec 15 10:53:31 LIC_FILES_CHKSUM = "file://${COREBASE}/LICENSE;md5=3f40d7994397109285ec7b81fdeb3b58 \ file://${COREBASE}/meta/COPYING.MIT;md5=3da9cfbcb788c80a0384361b4de20420" Dec 15 10:53:42 is imho better see ie recipes-core/tasks/task-base.bb Dec 15 10:54:18 yeah, I guess. it seems a bit inconsistent that the metadata license goes into LICENSE for recipes with no files, but doesn't for the majority of "normal" recipes Dec 15 10:54:44 morning Dec 15 10:55:32 by the way, what do you mean by "ie"? Dec 15 10:56:26 RP__, kergoth: Should bitbake inform from which recipe a require failed? Dec 15 10:57:09 with u-boot vs. uboot some recipes have not been updated and bitabke just failed parsing with exit 1, because it did not find the required file Dec 15 10:57:21 ...soo.. guys, any idea on how to force bitbake to rebuild a particular package each time? setting PV/PR to $DATETIME did not do the trick Dec 15 10:57:22 pb_: actually e.g. I guess, sorry Dec 15 10:57:23 sadly it did not tell me from which recipe he had this problem Dec 15 10:57:54 grepping it was easy, but it would be handy to have the recipe name/path right away imho Dec 15 10:57:55 stefan_schmidt: it should, yes Dec 15 10:58:16 stefan_schmidt: The parallel parsing regressed the error handling badly :( Dec 15 10:58:42 RP__: ah, its missing context then I guess Dec 15 10:59:06 RP__: nothing to problematic but worth keeping in mind for later improvement Dec 15 11:00:02 stefan_schmidt: is it in some internal layer? Dec 15 11:00:16 stefan_schmidt: I've git greped few and haven' Dec 15 11:00:21 JaMa: nope, openmoko and nokia :) Dec 15 11:00:23 t found any uboot users Dec 15 11:00:35 JaMa: hmm, have you pushed your changes? Dec 15 11:00:38 stefan_schmidt: they were fixed even before that patch was in oe-core.. Dec 15 11:00:40 let me re-check here Dec 15 11:00:57 meta-xilinx had the same problem Dec 15 11:01:17 stefan_schmidt: http://git.shr-project.org/git/?p=meta-smartphone.git;a=commit;h=c13c0159b3d831af76ef29b9921e74e8f02ae951 Dec 15 11:01:53 JaMa: aaargh Dec 15 11:02:18 JaMa: it did not update as I changed the mplayer stuff manually some time back (as it was breaking the build at that time) Dec 15 11:02:33 and now it did not update further when running oebb update Dec 15 11:02:39 JaMa: totally my bad on this one Dec 15 11:02:54 * stefan_schmidt goes and checks if meta-xilinx is up-to-date Dec 15 11:03:35 ok, that one is real. Will send a patch Dec 15 11:05:57 done Dec 15 11:06:43 JaMa: grr, it really annoys me that I did not notice my problem with meta-smartphone before. Dec 15 11:08:25 stefan_schmidt: you can also use without_shr branch from meta-smartphone if you''re not using oe-core-contrib/shr and meta-oe-contrib/shr branches.. Dec 15 11:09:17 u-boot-xilinx_git.bb is meta-micro? your patch is missing layer in subject Dec 15 11:09:58 JaMa: no, its an own bsp layer. Dec 15 11:10:03 ah and it's in oe-core ML Dec 15 11:10:10 (Subject) sigh, not my day today Dec 15 11:10:24 JaMa: There is no README in the layer Dec 15 11:10:50 JaMa: I had to guess who is the mainatiner from the log. No ml for patches so I choose oe-core Dec 15 11:10:51 * JaMa just wanted to say that he has no idea what's in meta-xilinix/README :) Dec 15 11:12:31 :) Dec 15 11:19:06 Who reads README's :) Dec 15 11:21:21 Crofton|work: looser's like me :) Dec 15 11:21:28 hhe Dec 15 11:21:44 years of training taught me they all contain the same generix text :) Dec 15 11:50:41 otavio: :) I have same patch for xorg-app/* in my queue.. I'll send only xf86-input-void then Dec 15 12:35:07 can i read back the contents of the fb Dec 15 12:35:26 sure, go ahead Dec 15 12:36:42 :) Dec 15 12:36:47 but i guess the frames read back will not be continous Dec 15 12:36:59 since one application is writing and another is reading Dec 15 12:37:16 right Dec 15 12:37:32 are there any commonly used utlities for this Dec 15 12:37:59 none I can think of. "cat" maybe? Dec 15 12:38:18 rather depends what exactly you are trying to accomplish. Dec 15 12:39:19 okay, i just want to capture whatever is displayed on the fb Dec 15 12:39:50 pb_: some experimental thing i am trying to do Dec 15 12:42:15 JaMa: oh ok :) thx Dec 15 13:04:33 doh I am still fighting, how can I cause a package to get autorebuild each time it is bitbaked or something that depends on it is bitbake? basically I am trying to get the same behaviour that angstrom-version package had in oe-classic Dec 15 13:04:51 ideas, anyone? so far tried setting PV to $DATETIME, tried setting PRINC to $DATETIME Dec 15 13:04:58 did not help Dec 15 13:33:37 does it work if you set __BB_DONT_CACHE=1 in your recipe? Dec 15 13:34:39 I can try, but I figured something else out now, after some more hacking; I use a variable that I set in my local.conf, this variable is echoed into a version file; if I do indeed change the variable, then the package gets rebuilt; that does make sense, so I guess I am fine after all; I am still testing this Dec 15 13:34:52 but I will remember that dont cache setting just in case Dec 15 13:54:36 ah, qt 4.8.0 is out Dec 15 13:54:50 * bluelightning touches up the recipes Dec 15 14:10:27 bluelightning: thanks ! Dec 15 15:18:47 while buulding armv5: armv6 and armv7 targets in the same tmp, I notice a problem with meta-toolchain-qte generation Dec 15 15:19:12 in the armv7a-toolchain, I find ./usr/local/angstrom-eglibc-x86_64-armv5te/sysroot ... Dec 15 15:21:11 and in fact sysroots/x86_64-angstromsdk-linux/usr/bin/ under angstrom-eglibc-x86_64-armv7a is empty Dec 15 15:21:33 so it seems it takes all the native packages for the 1st arch which was built Dec 15 15:27:08 Hi everybody Dec 15 15:27:24 I have three directories in /tmp/staging Dec 15 15:27:45 armv6-oe-linux-gnueabi, i686-linux and pc7302-oe-linux-gnueabi Dec 15 15:28:12 first one is HOST_SYS and second one is BUILD_SYS Dec 15 15:28:16 what about third one Dec 15 15:28:29 my machine name is pc7302 Dec 15 15:28:55 is there any special environment specify pc7302-oe-linux-gnueabi directory Dec 15 15:29:01 as /tmp/staging/pc7302-oe-linux-gnueabi Dec 15 15:29:26 I want to use linux headers and I need to define kernel include directory path Dec 15 15:29:49 path is tmp/staging/pc7302-oe-linux-gnueabi/kernel/include/ Dec 15 15:30:17 I tried to examine from environments "bitbake -e" Dec 15 15:30:28 but as usual I could not :) Dec 15 16:54:09 when i bitbake gumstix-basic-image I get error saying Please set the cache varible....can any one help me what is wrong Dec 15 17:04:50 ERROR: Please set the 'PERSISTENT_DIR' or 'CACHE' variable Dec 15 17:56:32 I'm using oe-core and some meta- stuff from other places. I keep getting failed fetches from kernel.org. Is there a way to specify mirror sites or something? Fetching them manually from other places is getting tedious. Dec 15 17:59:56 yes, you can set KERNELORG_MIRROR in your local.conf Dec 15 18:00:16 we should really just change the default in oe, it is fairly lame pointing to a site that has been broken for more than 3 months Dec 15 18:01:06 like this? : KERNELORG_MIRROR = "http://ftp.osuosl.org/pub/kernel/pub" Dec 15 18:02:05 IMHO, fairly lame kernel.org has been broken for more than three months. =) Dec 15 18:02:48 we should really change the default mirrors.bbclass to have some sane kernel.org fallbacks Dec 15 18:03:59 yeah, exactly like that Dec 15 18:04:37 kergoth: yeah, that too, but after this length of time it seems as though we should just change the primary url. I think that's started happening for a few recipes already. Dec 15 18:04:52 if the kernel h4x0rs were going to mend their site then you've got to think they would have done it by now. Dec 15 18:05:49 hm, yes, pseudo does seem to take up rather a lot of cpu time. I wonder what it's doing. Dec 15 18:06:09 also a bit surprising how long it seems to take to gzip the ipks. maybe we should consider using lzo as an option. Dec 15 18:08:18 pb_:thank you Dec 15 19:34:21 Has anyone tried compiling gdb with python support (--with-python)? Dec 15 19:37:06 fcooper: it would be default if not disabled Dec 15 19:37:19 and python dev files are around Dec 15 19:50:31 khem: I tried adding python to RDEPENDS and DEPENDS to gdb-cross-sdk which failed and I haven't found a recipe that supplies python_dev. Only reference I found was within python-2.7-manifest.inc which it doesn't seem like any python recipe includes that file. Dec 15 20:03:16 where's the best place to get help with meta-gumstix ? Dec 15 20:45:23 awozniak, what dos the readme say? Dec 15 21:16:22 khem: I finally got it working. By default the gdb-cross-sdk couldn't find the python libraries. I had to hardcode this to finally get it to work --with-python="/home/franklin/git_projects/bitbake/arago-tmp/sysroots/i686-linux/usr/lib/python2.6" and several other changes to export env variables (BUILD_SYS HOST_SYS etc...). Is this expected. By looking at other recipes that use --width-python these changes werent needed. Dec 15 21:22:56 hi fcooper did you maange to solve your issues with qt ? Dec 15 21:28:13 ecricben: Yes. For the arago distrib, the qt related files are moved in a folder called linux-devkit within the ti sdk. So the way qmake is compiled it was looking for the bins, lib, includes at /linux-devkit/ (absolute paths). I had to create the file qt.conf in the same directory as qmake which modifies the paths hard coded within qmake and points to the proper directories. After that Qt Creator was able to find everything it needed. I was going Dec 15 21:28:50 ok. I just tested qtcreator 2.4.0 and meta-toolchain-qte produced by oe-core and that works out of the box Dec 15 21:30:14 RP: how well tested / how safe is that opkg fix? I wonder if I should take it in for my image... Dec 15 21:31:03 Yes because it extracts to the folder /arago-2011.09/armv7a/ and qmake is looking for files within that folder. For the ti sdk qt is located within /home/franklin/ti-sdk-am335x-evm-05.03.01.00/linux-devkit while qmake is just looking within /linux-devkit/ which doesn't exist which is why it was failing. Dec 15 21:33:32 ericben: Current issue is getting gdb Qt Creator to do remote debugging with the embedded platform. It seems that Qt Creator requires gdb to include python support. However gdb-cross-sdk failed to properly locate python Dec 15 21:33:44 well I'm not using arago Dec 15 21:33:52 yes, that was working with oe-classic Dec 15 21:34:26 I'll test tomorrow with oe-core generated sdk Dec 15 21:39:09 fcooper: I confirm gdb in oe-classic generated sdk had libpython Dec 15 21:41:43 fcooper: gdb readme says : `--with-python[=PATH]' Build GDB with Python scripting support. (Done by default if libpython is present and found at configure time.) Python makes Dec 15 21:51:16 ericben: You didn't make any modifications to fix this error? /usr/lib/python2.6/distutils/sysconfig.py", line 22, in PREFIX = os.path.normpath(sys.prefix).replace( os.getenv("BUILD_SYS"), os.getenv("HOST_SYS") ). I had to export several variables within the recipe to get it to work. Dec 15 21:52:27 fcooper: in oe-classic this was working out of the box Dec 15 21:53:39 ericben: Ok thanks. Dec 15 21:55:12 fcooper: btw I was testing with qtsdk-2010.05 Dec 15 22:15:03 Jin^eLD: A bit late but you could set one of the tasks as nostamp Dec 15 22:17:36 RP: aah, regarding the "always rebuild" thing? indeed, I did not think about that at all, although I remember seeing it in some classes Dec 15 22:17:52 thanks Dec 15 22:17:52 :) Dec 15 22:55:17 I'm trying to get started with BitBake but everything I try to build fails to patch Makefile.in Dec 15 22:57:11 Is there something thats maybe out of date? Like the Makefile patch Dec 15 22:58:54 Or..... everything is dependent on quilt-native and its failing to patch Dec 16 00:02:44 I'm going insane. In oe-classic I would commonly -c devshell virtual/kernel, then ( . ../temp/run.do_compile.* ) to build it. Does this not work in oe-classic? There's no ../temp/run.do_compile.*, even after a bitbake -c clean and bitbake virtual/kernel Dec 16 00:04:34 erm, I mean does this not work in oe-core ? Dec 16 00:06:09 If you want devshell. inherit devshell. Add INHERIT += devshell to your local.conf, *then* try to run the task. Dec 16 00:08:53 devshell starts ok, there's just no ../temp/run.do_compile.* there Dec 16 00:09:22 I've bitbake -c clean virtual/kernel ; bitbake virtual/kernel like a dozen times now Dec 16 00:09:44 Why is ../temp/run.do_compile.* not there? Dec 16 00:10:54 I've changed the defconfig and I want to rebuild the bloody thing. Dec 16 00:21:52 bitbake -c compile virtual/kernel # forgot about that! Dec 16 01:31:04 the run scripts are only emitted when the task is run **** ENDING LOGGING AT Fri Dec 16 02:59:56 2011