**** BEGIN LOGGING AT Thu Sep 09 02:59:58 2010 Sep 09 03:37:30 i'm getting an error building libpam-base-files, install : cannot stat `/home/richard/oe/tmp/work/armv5te-angstrom-linux-gnueabi/libpam-base-files-1.0-r6/pam.d/*': No such file or directory. It looks like the pam files are not in pam.d but instead being copied one directory up. the bb file has SRC_URI="file://pam.d/*" so is the problem caused by bitbake not copying to the right place? Sep 09 03:43:21 htns: libpam-base-files unpacks fine with current oe. Sep 09 03:44:01 kergoth, is there something in my environment that could be causing this problem? Sep 09 03:45:43 i assume it is bitbake that is copying the SRC_URI pam.d/* to workdir/... instead of workdir/...pam.d/ Sep 09 03:50:58 kergoth, in your build, do the files from libpam-base-files/pam.d/* show up in a pam.d directory in the workdir/libpam-base-files-1.0-r6/pam.d/? Sep 09 04:00:30 i'm using bitbake top of tree 992e460f24d4da707c76d6e6d74d3684c9646279 . is that a problem? Sep 09 04:09:58 bitbake is irrelevent. unpacking is controlled by code in base.bbclass in oe Sep 09 04:10:06 bibake just runs tasks, thats all, it doesn't care what the tasks do Sep 09 05:05:58 yeah, i can't make sense of it. seems consistent, i had the same problem with base-files Sep 09 06:13:47 hi Sep 09 06:27:57 morning Sep 09 06:39:41 ka6sox any progress on the freenx stuff? didn't have time to peek in it myself, too much other things at hand Sep 09 06:40:26 khem, btw was reading the backlog, I'll see if I can kick off building world with dev head on our build server; let it use some of those idle cycles; machine calamari, distro minimal Sep 09 06:41:46 btw 1.8.18 with oe testing does not seem to work any more: Sep 09 06:41:57 OTE: :EOL while scanning string literal (, line 1) while evaluating: ${@oe.utils.ifelse(bool(d.getVar('MACHINE', True)), '${MACHINE}', 'BASE_PACKAGE_ARCH')} ERROR: Error in executing: /home/hudson/jobs/FM_test/workspace/openembedded/recipes/ncftp/ncftp_3.2.0.bb Sep 09 06:42:10 get a lot of these Sep 09 06:59:43 khem, did you get world parsing without unmet dependency errors ? Sep 09 06:59:48 first thing I get is: Sep 09 06:59:49 ERROR: '['/home/hudson/jobs/FM_world/workspace/openembedded/recipes/powervr-drivers/bc-cube_0.2.0.bb']' RDEPENDS/RRECOMMENDS or otherwise requires the runtime entity 'libgles-omap3-x11wsegl' but it wasn't found in any PACKAGE or RPROVIDES variables Sep 09 07:00:45 this package does not exist :-( Sep 09 07:03:16 c Sep 09 07:04:03 other Q: Sep 09 07:04:13 can recipe names have two underscores in them? Sep 09 07:17:48 eFfeM_work, are you using hudson to manage builds? Sep 09 07:18:00 yes Sep 09 07:18:09 how do you like it? Sep 09 07:18:26 actually only recently started to use it, I have not too much experience with it, but initial impression is ok Sep 09 07:18:57 good to hear... Sep 09 07:19:19 you're using it too? or maybe involved in development of it? Sep 09 07:19:54 no, I'm interested in using it for one of the projects I mangle. Sep 09 07:20:58 well it is quite nice as an autobuilder, will provide regression info etc, we use it also for QA on our archive (to verify that things checked in actually build) Sep 09 07:21:23 and it had some cycles left Sep 09 07:21:42 ya, this project I hear a lot of complaints about the autobuilder... Sep 09 07:22:09 it feels like the devs are standing in front of a microwave oven and yelling HURRY! Sep 09 07:22:28 for now I can use the spare resources to do some tests for the testing branch Sep 09 07:22:44 ah... Sep 09 07:23:19 trying to set up some tests while my project is compiling Sep 09 07:39:10 morning Sep 09 07:39:11 http://www.youtube.com/watch?v=Eq3CuMDXaPs Sep 09 07:52:01 hi hrw Sep 09 07:53:00 hrw, guess you will know, can recipes have two _' s in the name ? Sep 09 07:58:48 should not have Sep 09 07:59:04 as name is parsed as PN_PV.bb Sep 09 07:59:22 so it you use two _ in recipe filename you will have to set PN/PV in recipe Sep 09 08:00:45 hi all Sep 09 08:17:46 hrw was afk, but that was what I thougth as well Sep 09 08:17:47 frans@frans-desktop:~/workspace/openembedded.git/recipes$ ls */*_*_*bb Sep 09 08:17:47 clutter/clutter-box2d_0.10.0_git.bb ntpclient/ntpclient_2007_365.bb Sep 09 08:17:47 ntpclient/ntpclient_2003_194.bb rpm2cpio/rpm2cpio-native_1.2_2.bb Sep 09 08:18:43 ntpclient sets PV, not PN (but still uses PN) Sep 09 08:19:54 rpm2cpio does not touch PV and PN Sep 09 08:21:42 clutter does not seem to do anything either Sep 09 08:28:51 good morning Sep 09 08:29:12 gm Sep 09 08:29:23 good morning Sep 09 08:29:48 eFfeM_work: hi Frans, are you using hudson for your autobuilder at work ? Sep 09 08:29:56 yes Sep 09 08:30:11 just started to use it Sep 09 08:30:32 ok, how to you handle the fact that bitbake returns 1 even on success which makes hudson believe the build failed ? Sep 09 08:30:43 not yet Sep 09 08:30:51 ok :) Sep 09 08:30:59 we have just recently started using it, we're still on the learning curve Sep 09 08:32:30 ok, I launched one build this night and I'm first doing a bitbake -c fetchall to check all sources are ok, then a bitbake myimage. And hudson stopped everything after the fatchall (which was a success) because bitbake returne 1 Sep 09 08:34:44 ericben: if you want to you can ignore the error code I guess Sep 09 08:34:54 if hudson can parse the log for x failed and return error only if x > 0 this would be great Sep 09 08:35:08 do something like (bitbake fetchall ; bitbake recipe) Sep 09 08:35:12 eFfeM_work: yes but in that case you don't know it has failed unless you check tinderbox Sep 09 08:35:14 wiht () so it is in a subshell Sep 09 08:35:16 true Sep 09 08:35:34 then fix bitbake :-) Sep 09 08:36:12 well, considering I'm using python as an evolved shell I'm far from being abel to understand bitbake's code ;-) Sep 09 08:37:49 http://wiki.hudson-ci.org/display/HUDSON/Log+Parser+Plugin may be the solution Sep 09 08:41:18 ericben: that log plugin seems nice, didn't get to looking at plugins Sep 09 08:41:28 (well, i did, but only a few) Sep 09 09:03:45 Hi all Sep 09 09:04:40 is there an replacement for STAGING_BINDIR_CROSS variable just like we have bindir for STAGING_BINDIR Sep 09 09:06:06 eFfeM_work: in fact bitbake return 1 because there was something marked as an error (ERROR: Multiple .bb files are due to be built which each provide virtual/psplash) but this error is not critical so the build is successful Sep 09 09:06:27 ah ok, got some of these too Sep 09 09:06:55 so bitbake is doing his job properly and the problem is that I didn't put a prefered provider for psplash :) Sep 09 09:07:31 this kind of tool helps to improve quality, only to have the green light on the build status page :-) Sep 09 09:07:46 well I tried bitbake world triggered by what khem said before, but I still get errors on unmet dependencies :-( (using dev head) Sep 09 09:08:09 ericben: we have not yet much experience with it, but it looks like quite a good and useful tool Sep 09 09:25:42 Noor: no idea Sep 09 09:27:18 eFfeM_work: OK Sep 09 09:28:00 there is a variable bindir_cross but the value is different from STADING_BINDIR_CROSS Sep 09 09:29:14 hi Sep 09 09:29:20 i have one question Sep 09 09:29:33 i am merging native and non-native recipes Sep 09 09:30:14 the non-native recipe has its native recipe in its depends Sep 09 09:30:38 and the native recipe has another recipe in its depends with += Sep 09 09:31:11 how should i tackle this in the combined recipe? Sep 09 09:34:33 any idea on how to fix this : ERROR: '['/home/hudson/test-oe/workspace/cpuimx25/openembedded/recipes/qt4/qt4-embedded-gles_4.6.3.bb']' RDEPENDS/RRECOMMENDS or otherwise requires the runtime entity 'qt4-embedded-gles' but it wasn't found in any PACKAGE or RPROVIDES variables of any buildable targets. Sep 09 09:35:32 fahad: I wrote a mail to you on the mailinglist Sep 09 09:36:05 Basically you use DEPENDS_pn-glib-1.2 += "glib-1.2-native" Sep 09 09:36:57 So you may modify the dependencies seperately Sep 09 09:50:50 03Enrico Scholz 07org.openembedded.dev * r10a92f2524 10openembedded.git/recipes/ncurses/ (ncurses-5.7/libtermcap.so ncurses_5.7.bb): (log message trimmed) Sep 09 09:50:50 ncurses: use linker scripts for libncurses(w) Sep 09 09:50:50 Some software (e.g. util-linux-ng) assumes that symbols from -ltinfo Sep 09 09:50:50 will be added when it is linked against -lncurses. This breaks when Sep 09 09:50:50 linkerflags are containing --no-copy-dt-needed-entries which is the case Sep 09 09:50:50 e.g. in Fedora 13+. Sep 09 09:50:51 This patch replaces the libncurses.so symlink with a linkerscript which Sep 09 09:52:16 03Enrico Scholz 07org.openembedded.dev * r7fce8ef127 10openembedded.git/recipes/ncurses/ncurses_5.7.bb: Sep 09 09:52:17 ncurses: fixed PR Sep 09 09:52:17 Signed-off-by: Enrico Scholz Sep 09 09:52:24 vps: thanks Vitus, i didn't know that you can use this override in this way Sep 09 09:52:51 ericben: rm the recipe? this is apparently work in progress Sep 09 09:52:56 Just found it the other day in base.bbclass :-) Sep 09 09:52:57 i will update the recipes and see how it works Sep 09 09:53:34 ti/koen seems to be adding quite some stuff that requires recipes that are not committed :-( Sep 09 09:53:42 eg: ERROR: '['/home/hudson/jobs/FM_world/workspace/openembedded/recipes/powervr-drivers/bc-cube_0.2.0.bb']' RDEPENDS/RRECOMMENDS or otherwise requires the runtime entity 'libgles-omap3-x11wsegl' but it wasn't found in any PACKAGE or RPROVIDES variables Sep 09 09:53:50 eFfeM_work: I was thinking of a MACHINE_FEATURES egl ? Sep 09 09:55:01 and then tuning DEFAULT_PREFERENCE depending on MACHIN_FEATURE. Do you think this is possible ? Sep 09 09:55:13 koen seems to be your big firend ;-) Sep 09 09:55:13 fahad: sorry, it's bitbake.conf. (OVERRIDES = "local:${MACHINE}:${DISTRO}:${TARGET_OS}:${TARGET_ARCH}:build-${BUILD_OS}:fail-fast:pn-${PN}") Sep 09 09:56:19 vps: thanks Sep 09 09:57:24 ericben: i think you got some errors before (so did I) and these are the followup Sep 09 09:57:32 indeed virtual/egl might play a role Sep 09 10:01:41 eFfeM_work: is there a way to enable a recipe based on a MACHINE_FEATURES ? Sep 09 10:01:51 I have no idea Sep 09 10:02:52 helu Sep 09 10:05:02 simple solution : only libgles-omap3.inc provides virtual/egl, but it also have a COMPATIBLE_MACHINE , so we could copy this COMPATIBLE_MACHINE to each recipe which has a DEPEND on virtual/egl ? Sep 09 10:08:56 no Sep 09 10:09:14 ericben: there could be out-of-tree providers of virtual/egl for other machines Sep 09 10:10:12 hrw: yes. Do you know if there is a way to enalbe a recipe based on a MACHINE_FEATURE ? Sep 09 10:11:03 or is there a clever solution for this problem ? Sep 09 10:12:07 hrw I feel that if there is an out-of-tree provider it should be mentioned somewhere Sep 09 10:12:42 i do not find it desirable to have a tree with broken dependencies which are only there for people having stuff out-of-tree that they apparently do not want to share Sep 09 10:15:05 hmm Sep 09 10:16:50 ericben: define enable Sep 09 10:17:34 hrw: equivalent of COMPATIBLE_MACHINE Sep 09 10:17:41 but based on a FEATURE Sep 09 10:18:24 gm Sep 09 10:18:34 SOC_FAMILY discussion? :-) Sep 09 10:18:40 hi likewise Sep 09 10:18:50 in fact in qt4-embedded-gles I think the problems comes from the fact that this recipe adds PROVIDES += "qt4-embedded" which means it will trigger the problem for each person who do bitbake qt4-embedded but hasn't an Omap3 Sep 09 10:18:54 ericben: I do not see a usecase Sep 09 10:18:57 hello eFfeM_work Sep 09 10:19:17 hello hrw, long time no see. How's Ubuntu work? Sep 09 10:19:24 hrw: ok in that case how to prevent this : ERROR: '['/home/hudson/test-oe/workspace/cpuimx25/openembedded/recipes/qt4/qt4-embedded-gles_4.6.3.bb']' RDEPENDS/RRECOMMENDS or otherwise requires the runtime entity 'qt4-embedded-gles' but it wasn't found in any PACKAGE or RPROVIDES variables of any buildable targets. Sep 09 10:19:47 ERROR: '[]' RDEPENDS/RRECOMMENDS or otherwise requires the runtime entity 'virtual/egl' but it wasn't found in any PACKAGE or RPROVIDES variables Sep 09 10:21:11 likewise: fine, bootstrapped their toolchain finally Sep 09 10:21:45 let me pull Sep 09 10:22:15 ericben: I guess that should be fixed in bitbake. It ought to just ignore the unbuildable PROVIDEr. Sep 09 10:22:56 likewise: i see a use for SOC_FAMILY but COMPATIBLE_MACHINE could be used there as well afaik Sep 09 10:23:13 and the way it was added is not an example on how things should be done Sep 09 10:23:19 but it should work for all, and not only for TI Sep 09 11:11:05 pb_: ok I'll submit a patch for a temporary fix for OE as I'm not enough skilled for hacking bitbake. Sep 09 11:11:28 righto Sep 09 11:12:06 ok, let's see in ncurses behaves better today Sep 09 11:15:32 what's the procedure of submitting patches? I looked through the OE wiki but I guess I missed it Sep 09 11:15:50 generate a patch, send it to the mailing list Sep 09 11:16:19 Jin^eLD: git format-patch then git send-email --to openembedded-devel@lists.openembedded.org Sep 09 11:16:19 pb_: any special topic [bla] tags or guidelines or something? Sep 09 11:17:20 http://wiki.openembedded.net/index.php/Commit_log_example http://wiki.openembedded.net/index.php/Styleguide Sep 09 11:17:33 and more generaly : http://wiki.openembedded.net/index.php/Category:Policy Sep 09 11:17:33 Jin^eLD: git send-email will use the first line of the patch as subject Sep 09 11:17:45 ok thanks Sep 09 11:18:09 especially http://wiki.openembedded.net/index.php/Commit_Policy Sep 09 11:20:02 Why does kernel.bbclass set CC and LD in the compile command instead of setting CROSS_COMPILE? Sep 09 11:20:07 Jin^eLD: you're welcome Sep 09 11:20:41 florian: mostly so that KERNEL_CCSUFFIX works, I think Sep 09 11:20:54 (It does set CROSS_COMPILE as well, iirc.) Sep 09 11:23:08 I just ran into a kernel that fails because it fails to find the correct ar. Sep 09 11:24:51 damn util-linux-ng is still failing Sep 09 11:26:40 trying to clean ncurses now ... Sep 09 11:30:42 trying to clean ncurses now ... Sep 09 11:35:32 that doesn't work either Sep 09 12:24:16 is it possible in new style staging that something is being staged but nt deplyed for non-native packages Sep 09 12:33:43 is enrico around/ Sep 09 12:43:39 anyone has an idea on how to use base_conditional to return a value containing a variable? like @base_conditional("FOO", "a", "VAR", d) where VAR='b' Sep 09 13:00:45 Noor: good question Sep 09 13:01:34 pb_: you are right, on top of kernel.bbclass CROSS_COMPILE gets set... Sep 09 13:02:02 But I do not see why it doesn't seem to be defined any more later. Sep 09 13:02:48 hi Jin^eLD Sep 09 13:03:09 hi florian Sep 09 13:09:31 likewise: I was converting llvm2.7 recipe to new style staging Sep 09 13:10:51 I observed this behavior in that recipe it is staging some file which are in deply folder Sep 09 13:11:53 and when I add it in image filder using ${D} I get some errors "ERROR: QA Issue with llvm2.7-dev: package llvm2.7-dev contains bad RPATH /home/nahsan/SB/open_comm/tmp/work/armv7a-angstrom-linux-gnueabi/llvm2.7-2.7-r1.r8/llvm-2.7/build/lib in file /home/nahsan/SB/open_comm/tmp/work/armv7a-angstrom-linux-gnueabi/llvm2.7-2.7-r1.r8/packages-split/llvm2.7-dev/usr/bin/llvm2.7/tblgen" Sep 09 13:12:09 in do_package task Sep 09 13:34:03 Noor: An RPATH may not refer to the tmp/work directory AFAIK, so you are packaging files from the wrong location. (I am not the packaging expert here, this is semi-guessing.) Sep 09 13:53:57 * Tartarus kicks openjade-native Sep 09 13:54:23 khem: ping Sep 09 14:30:45 ensc, ping Sep 09 14:55:57 eFfeM_work: what all are unmet dependencies you get Sep 09 15:00:49 Crofton: pong Sep 09 15:01:14 ensc, the problem went away :) Sep 09 15:01:19 ok, good Sep 09 15:01:33 hi there Sep 09 15:01:39 we thought the fix might break an ubuntu build Sep 09 15:01:47 I'm getting gettext-native build failure, here's build log: http://pastebin.com/AAQLtAR3 Sep 09 15:01:52 but cleaning all ncurses fixed it Sep 09 15:02:03 is it known issue or something new? Sep 09 15:02:27 Crofton: yes; staging fails to convert the symlink into a regular file Sep 09 15:02:29 hi Tartarus Sep 09 15:03:26 hey khem Sep 09 15:03:27 ensc|w, thanks for figuring out the solution Sep 09 15:03:33 Do you have openjade-native in your ASSUME_PROVIDED? Sep 09 15:04:45 I dont think so Sep 09 15:04:53 ok Sep 09 15:05:03 are you seeing xalan-j fail ? Sep 09 15:05:17 no, i got a crazy mangle in one of the openjade-native's Makefiles Sep 09 15:05:25 I think it's just bad luck... Sep 09 15:05:33 yeah could be Sep 09 15:06:10 actually xalan-j when building from scratch does not build Sep 09 15:06:18 it has some goody DEPENDS problem Sep 09 15:06:41 it builds in a bunch becuase someone elese stages it before xalan-j needs it Sep 09 15:07:02 anarsoul: its a problem on your build machine Sep 09 15:08:01 khem: hm, how to solve it? Sep 09 15:08:43 * anarsoul will run revdep-rebuild, but doubts that it will found something Sep 09 15:08:54 anarsoul: Dont know see your distro mailing lists Sep 09 15:09:02 seems buggy zlib Sep 09 15:09:25 it's zlib-1.2.5 Sep 09 15:09:25 anarsoul: if you have ASSUME_PROVIDED then dont use them Sep 09 15:09:39 I don't use ASSUME_PROVIDED Sep 09 15:09:40 1.2.5 should have fixed it Sep 09 15:13:48 khem: libxml and zlib are OK on my build machine Sep 09 15:14:54 and revdep-rebuild (gentoo tool to find binaries and libraries broken by a package update) found no problems Sep 09 15:15:55 pb_: found it... OE isn't guilty. Its an evil modification of the Makefile Sep 09 15:47:42 khem: so, should I file a bug about this build failure? Sep 09 15:50:30 florian: some stuff gpephone related is not fetchable anymore Sep 09 15:51:56 /recipes/gpephone/libiac_svn.bb, do_fetch) failed with 256 Sep 09 15:53:12 anarsoul: you can add -lz to LDFLAGS of gettext-native as workaround Sep 09 15:53:23 may be that will help Sep 09 15:54:02 oh its already there Sep 09 15:55:37 florian: also this one recipes/gpephone/libgpephone_svn.bb, do_fetch) failed with 256 Sep 09 15:56:22 it seems all gpephone svn recipes dont fetch anymore. is the svn server decommissioned ? Sep 09 16:00:01 khem: I suspect it happens because oe tries to link with some system libraries, not built by oe Sep 09 16:00:15 khem: run-qemu.sh, what does that assume on the host, anything? Sep 09 16:00:38 Going to quick test some of these native-sdk-images once it's done Sep 09 16:00:56 anyone can confirm that usr/lib/python2.6/config can be removed from python-distutils ? Sep 09 16:02:16 hmm Sep 09 16:02:25 i don't think so Sep 09 16:02:42 iirc distutils needs the config otherwise can no longer compile extension modules Sep 09 16:03:10 i could be wrong though, it's been ages since i looked at that stuff Sep 09 16:03:14 on ubuntu there is only : Makefile and a link to libpython2.6.so.1 Sep 09 16:03:45 Tartarus: it needs bridge-utils Sep 09 16:03:52 installed and setup Sep 09 16:04:02 if you want qemu to boot with networking Sep 09 16:04:04 ok in doubt, I let it, I just remove the win* files from distutils and the big libpython2.6.a from the coinfig directoy. In the end I get a ~400kB packages which is far better than before Sep 09 16:04:11 svn ls "svn://projects.linuxtogo.org/svn/gpephone/trunk/" Sep 09 16:04:12 svn: Can't open file '/svn/gpephone/trunk/format': Permission denied Sep 09 16:04:34 thanks mickey|office updated patch will follow for ack on the ml Sep 09 16:04:38 khem: hrm nasty... no the server is still there but maybe some things changed moving it to fusionforge Sep 09 16:04:40 khem, k, thanks Sep 09 16:04:40 ericben: cool, thanks Sep 09 16:05:05 Tartarus: read thru the file I have tried to document what one would need Sep 09 16:05:13 khem: since noone is working on it i'm tempted to move the whole gpephone stuff to obsolete Sep 09 16:05:14 btw, to remove the files, I added a do_install_append which does rm the files. Is this the right way to do it ? Sep 09 16:05:15 Tartarus: it still might have some cruft Sep 09 16:05:22 florian: most welcome Sep 09 16:06:06 you being gpephone developer have the say :) Sep 09 16:07:05 just being realistic... would be fine with me. if you don't do it i'll do it in a free minute Sep 09 16:07:32 ericben: well, it's probably more maintainable than patching configure.ac, so i'd say yes Sep 09 16:07:50 florian: should the whole directory be moved to obsolete Sep 09 16:09:33 khem: yes Sep 09 16:09:35 can someone do 'git rm -r obsolete nonworking' instead? Sep 09 16:10:04 hmm no, I wouldn't do this Sep 09 16:10:47 florian: ok. will do Sep 09 16:10:58 and settle it in Sep 09 16:59:40 hmmm, what's that: Sep 09 16:59:42 *** %n in writable segment detected *** Sep 09 17:02:37 GNUtoo|laptop: printf has a writable msg string which contains %n Sep 09 17:02:58 e.g. 'char buf[] = "%n"; printf(buf, &n);' Sep 09 17:03:02 ah? Sep 09 17:03:10 ok Sep 09 17:03:29 is it problematic? Sep 09 17:03:38 it's a security issue Sep 09 17:03:45 ah ok Sep 09 17:04:15 you can workaround it (when buffer must be rw) by remapping it read-only after filling it with content Sep 09 17:04:40 ok but I dont' even know which package emited that during the build Sep 09 17:04:50 the target is armv7 Sep 09 17:06:52 it might be emitted by a native tool; I am not sure whether -D_FORTIFY has such a check or whether it is done at runtime only Sep 09 17:07:36 ok Sep 09 17:07:50 so the security issue is in the native tool and not on target? Sep 09 17:08:24 yes Sep 09 17:08:32 03Peter Klassen 07org.openembedded.dev * r944bbfa83f 10openembedded.git/recipes/xorg-xserver/xserver-xorg-conf_0.1.bb: Sep 09 17:08:32 xserver-xorg-conf: add xorg.conf for palmpre machine Sep 09 17:08:32 Signed-off-by: Peter Klassen Sep 09 17:08:32 Acked-by: Simon Busch Sep 09 17:09:00 ok thanks a lot Sep 09 17:09:03 03Simon Busch 07org.openembedded.dev * rfef2ddd336 10openembedded.git/recipes/xorg-xserver/xserver-xorg-conf_0.1.bb: Sep 09 17:09:04 Revert "xserver-xorg-conf: add xorg.conf for palmpre machine" Sep 09 17:09:04 This reverts commit 944bbfa83fedd417eb87daefd5f94ac8eb524a0a. Sep 09 17:09:56 ouch Sep 09 17:10:00 I'll talk to morphis Sep 09 17:13:39 Hello all, I am new at openembedded and I want to create my own image with my selected packages only. Is there any howto which defines creating a new distro and building a new image? Sep 09 17:13:52 2 solutions: Sep 09 17:13:58 look at existing images and modify it Sep 09 17:14:03 or look at the manual Sep 09 17:15:35 manual says just copy the existing distro conf and do changes, then use it. but what kind of changes? Sep 09 17:16:13 I think there should be a howto which shows the relationships between configs directories. Sep 09 17:16:35 I couldnt see the big picture, just travelling in files and directories. Sep 09 17:16:50 What is going on when I enter "bitbake anything".. Sep 09 17:17:22 People says openembedded is the best but nobody direct me to see the best features :) Sep 09 17:19:14 03Peter Klassen 07org.openembedded.dev * rf641c14f2d 10openembedded.git/recipes/xorg-xserver/ (xserver-xorg-conf/palmpre/xorg.conf xserver-xorg-conf_0.1.bb): Sep 09 17:19:14 xserver-xorg-conf: add xorg.conf for palmpre machine Sep 09 17:19:14 Signed-off-by: Peter Klassen Sep 09 17:19:14 Acked-by: Simon Busch Sep 09 17:26:36 kays: see under images Sep 09 17:26:46 kays: those are final images it can build Sep 09 17:26:54 then see under conf/distro Sep 09 17:27:00 those are distros it supports Sep 09 17:27:05 then see under conf/machine Sep 09 17:27:15 those are machines which can be used Sep 09 17:36:42 images, distros and machines are can be called by bitbake directly? In wiki examples just machines are called directly. Sep 09 17:37:45 what do you mean by "called"? Sep 09 17:38:01 a machine is just a variable. the MACHINE variable is set to a value, and bitbake.conf includes conf/machine/${MACHINE}.conf Sep 09 17:40:20 ok I got it Sep 09 17:40:48 debug messages show some usefull messages Sep 09 17:40:50 if you mean what you specify on the bitbake commandline, you specify a provider -- a recipe, basically, though recipes can PROVIDES multiple things Sep 09 17:41:06 MACHINE and DISTRO are just ways to break up the metadata in a logical way Sep 09 17:41:10 so that things are orthagonal Sep 09 17:41:31 distro controls policy, machine controls the hardware you're targeting, and both are generally independent of the recipes themselves Sep 09 17:42:01 this is why just about any DISTRO can support any MACHINE, and the recipes generally support any combination of those Sep 09 17:42:05 (within limits, of course) Sep 09 17:42:22 hm yes, thanks Sep 09 17:42:42 It is working now Sep 09 17:42:50 I realize we're really lacking some good high level conceptual documentation Sep 09 17:43:07 the user manual is well and good, but its more reference material with a bit of tutorialish stuff, not much architectural Sep 09 17:43:25 I agree Sep 09 17:43:51 i'd love to see that change, but my writing is pretty weak personally. ideall we'd have some technical writers contributing to the project Sep 09 17:43:58 open source projects tend to emphasize the code too much Sep 09 17:44:06 not enough support is given to those wanting to contribute in other ways Sep 09 17:44:36 what kind of policy does the distro control? Sep 09 17:44:42 http://free-electrons.com/pub/video/2010/elc/elc2010-osier-mixon-documentation.ogv is apt Sep 09 17:45:33 koobe_, well, it's convention. from a technical standpoint its just an other variable, associated config file, and entry in OVERRIDES (which allows conditional variable definitions or modifications based on the distro, and distro specific config files) Sep 09 17:46:18 in reality it tends to control things like target filesystem layout, packaging method used, as well as high level features (see the DISTRO_FEATURES variable) like ipv6 Sep 09 17:47:30 if you think about it like traditional linux distros, the distribution controls layout, packaging, as well as "distributionisms" -- things like, say, /etc/network/interfaces vs /etc/sysconfig/, chkconfig vs update-rc.d, etc Sep 09 17:47:45 of course, today, pretty much every oe based distro uses the same debian-like distributionisms, but that doesn't have to be the case Sep 09 17:48:00 does it pin down any versions of the packages/recipes? Sep 09 17:48:12 yeah, distro usually does that too Sep 09 17:48:21 as well as preferences for providers Sep 09 17:48:26 e.g. uclibc vs glibc vs eglibc Sep 09 17:49:07 is it feasible to fork or inherit a distribution, to customize some of the things you described, without having to specify them all? Sep 09 17:49:14 yeah, absolutely Sep 09 17:49:17 well, sort of Sep 09 17:49:27 you could create your own distro config file that includes angstroms, for example Sep 09 17:49:37 but the difficulty lies in the previously mentioned conditional variable definitions in the metadata Sep 09 17:49:45 if some recipe does FOO_angstrom = "bar" Sep 09 17:49:49 how about the overrides, they'd be for angstrom only? Sep 09 17:49:51 and you arenj't named angstrom, you won't pick that up Sep 09 17:49:52 right Sep 09 17:49:53 exactly Sep 09 17:50:07 downside to having machine and distro specific things in recipes Sep 09 17:50:36 can you arbitrarily specify any overrides, could I force the angstrom overrides to be used in addition to those of my 'fork'? Sep 09 17:50:47 sort of, yeah Sep 09 17:50:58 what you could do is add/inject "angstrom" directly into OVERRIDES Sep 09 17:51:03 even though your distro is still there. Sep 09 17:51:09 ideally in the correct order so yours are preferred to theirs Sep 09 17:51:26 i find it helps to visualize this stuff if you think of it as a layering mechanism for the metadata Sep 09 17:51:36 implemented in two ways: order of config file inclusion, and overrides application Sep 09 17:51:47 its designed such that more specific information always overrides less specific information Sep 09 17:52:04 so the machine version of a varaible always overrides the target architecture version, which in turn overrides the global default Sep 09 17:53:27 good explanations, thanks kergoth_ Sep 09 17:53:32 np Sep 09 17:53:45 ok, it definitely helps to understand that even the distro doesn't do any magic but builts on top of the same mechanism as the other settings. Sep 09 17:53:56 thanks for the clarifications Sep 09 17:54:00 np Sep 09 17:54:35 i keep meaning to try writing up some architectural documentation, based upon the original design we put together when the project was started Sep 09 17:54:42 so people can see why it went in the direction it did Sep 09 17:57:11 It will be really helpful, all configs and bb files scare new users :) Sep 09 17:58:51 it was one of the primary goals to create the collaborative repository of metadata that we have today -- ensuring that people with entirely different target hardware and goals could collaborate and share improvements, rather than just having a common starting point (as you would if you, say, copied buildroot as a starting point and ran with it) Sep 09 18:01:31 the thing I like most are the orthogonality of recipes and targets, and the fact that the external dependencies are minimal, resulting in repeatability of builds. Sep 09 18:02:25 that's actually one of the biggest problems with oe quality today. external dependencies are minimal, but it can still bite us in the ass. nature of crosscompilation and autoconf. if we miss something in one of them, it'll just happily run off and do somethin gbased on the existance of a file on the build machine Sep 09 18:02:28 pain in the ass Sep 09 18:04:19 if it was easy, OE wouldn't exist Sep 09 18:04:33 well said Sep 09 18:09:50 03Chris Larson 07org.openembedded.dev * r8cfa7531a0 10openembedded.git/ (8 files in 3 dirs): Sep 09 18:09:51 Per the TSC decision, make packaged-staging default Sep 09 18:09:51 For now, just ensures its inherited. In the future, we can merge / simplify Sep 09 18:09:51 staging.bbclass with packaged-staging.bbclass as appropriate. Sep 09 18:09:51 Signed-off-by: Chris Larson Sep 09 18:10:14 kergoth_: I have issue with packages staging Sep 09 18:10:21 it does not go well with staging_qa Sep 09 18:10:29 when using rm_work Sep 09 18:10:34 khem: how so? Sep 09 18:10:38 i'm using all 3 of those right now Sep 09 18:10:41 when I restart unfinished job Sep 09 18:10:51 it tries to run and exits randomly Sep 09 18:10:56 with error 256 Sep 09 18:11:11 I use rm_work Sep 09 18:11:24 staging_qa Sep 09 18:11:42 and then it starts using pstaging Sep 09 18:11:48 fine Sep 09 18:12:03 but dies on one of staging_qa on any recipe Sep 09 18:12:19 then I run qa_staging for that recipe individually works Sep 09 18:12:32 Hmm Sep 09 18:12:35