**** BEGIN LOGGING AT Wed May 11 02:59:58 2011 May 11 06:27:19 good morning May 11 07:34:59 hi all , how can i change the crontab file using my customized recipe May 11 08:52:23 Hi there! Does anybody know how to debug with gdb and using the symbol-files from .debug? May 11 08:53:24 they are loaded automatically May 11 08:53:59 if you're debuging something which loads modules in runtime (like Xorg) then you need to get starting address from /proc/pid/maps May 11 08:54:02 sure? does gdb searches for symbols in /usr/lib/.debug a.s.o May 11 08:54:10 and load-symbol-file url adress May 11 08:54:57 ok, you mean with add-symbol-file
May 11 08:55:26 yeah right, sorry my memory is not that good :) May 11 08:56:15 but that's needed only for libs loaded explicitly from runtime May 11 08:56:15 i hoped that gdb can load them automatically May 11 08:56:38 my lib is loaded by dlopen :-) May 11 08:58:40 i'm not shure, but when my lib contains the debug symbols in the shared object, i think it could be debugged without specifying any options... May 11 09:01:56 gm May 11 09:04:49 JaMa: when i got a executable in /usr/bin/myBin without debug symbols (as build from openembedded) and the symbols /usr/bin/.debug/myBin how could gdb find it? is there a way to set a search path? May 11 09:11:41 pnxs, you can always load a symbol file using "add-symbol-file" from the gdb prompt May 11 09:12:27 pnxs, in case of a shared-lib you do also "set stop-on-solib-events 1" which instructs gdb to stop whenever a shared lib is loaded May 11 09:12:35 after that you can load you symbols May 11 09:13:22 yes this is correct. But i want to let gdb do that for me. So gdb is searching for the symbols in the executable files (also in that, that where dynamically loaded) May 11 09:14:31 when there were a mechanism to let gdb search in another file for the symbols of the executable, gdb should be able to load it without any user input (i think) May 11 09:15:33 pnxs, http://sourceware.org/gdb/current/onlinedocs/gdb/Separate-Debug-Files.html May 11 09:15:35 when for example my binary with debug symbols is stripped and strip saves the stripped parts into another file, it should be possible to merge them again afterwards May 11 09:15:55 ...reading... May 11 09:17:05 thank you! i will try it May 11 09:25:22 hello May 11 09:26:04 prashant_: hello May 11 09:26:55 I had builded a filesystem for mini6410 using oe...... but now when I compiled fltk apps across oe .... I m getting an error :......error: Connections.o uses VFP register arguments, libSTPLCore.so does not May 11 09:27:22 can anyone help me out... May 11 09:28:59 this are the commands I m using ... >arm-angstrom-linux-gnueabi-g++ -shared -o libSTPLCore.so Connections.o -lcrypt May 11 09:30:37 can anyone help me what is VFP register arguments? May 11 09:37:28 prashant_, do you compile on your own or use bitbake and a recipe? May 11 09:39:05 I had created x11-image using bitbake.... now using that oe toolchain I m trying to compile my fltk apps across it... May 11 09:40:34 prashant_, then, you also need to pass -L to the linker since you try to link against -lcrypt May 11 09:42:38 /home/prashant/Downloads/oebase/build/tmp/sysroots/i686-linux/usr/armv6/lib/gcc/arm-angstrom-linux-gnueabi/4.5.3/../../../../arm-angstrom-linux-gnueabi/bin/ld: failed to merge target specific data of file Connections.o May 11 09:42:39 collect2: ld returned 1 exit status May 11 09:43:09 error: Connections.o uses VFP register arguments, libSTPLCore.so does not May 11 09:43:34 I m getting this error... May 11 09:43:51 its not a problem of linking i think so May 11 09:47:42 prashant_, you probably have to set the mfpu/fpu in some kind, but I am no expert on this May 11 09:51:47 "-march=armv7-a -mtune=cortex-a8 -mfpu=neon -mfloat-abi=softfp" may help though May 11 09:53:04 prashant_, if you have some knowledge on autotools I would encourage you to write an autotools based recipe May 11 09:53:22 JaMa: after some investigation i found out that gdb already loaded all my symbols but doesnt say so May 11 09:54:01 when a gnu-debuglink section is found in the executable gdb searches also in the .debug directory on that level for a filename set in the debuglink May 11 09:55:09 the only think is: the printouts when the gdb is connected to a process it prints out that no debugsymbols where found for every file, but load them from debuglink-files silently... May 11 09:55:22 so the problem was: no problem at all :-) May 11 09:56:14 so have a nice day and thank you for your help. cu May 11 10:08:21 thanks......mlip....my problem is solved........ May 11 10:46:40 I'll push wordwarvi May 11 10:46:44 very soon May 11 10:51:56 hello May 11 10:54:22 can any one just me a MIPS Development kit which support embedded Linux, Windows,... Having JTAG, USB, LAN, TV out, SDCARD , VGA, Serial, etc ... May 11 10:56:20 LCD, May 11 11:54:39 hello May 11 11:55:41 I'm trying to create a recipe, but I get an error during the configure step: "missing: GNU_PATCH_EXECUTABLE" May 11 11:56:11 I'm using cmake May 11 11:56:30 it is triggered by CMakeLists.txt:9 (find_package) (-> find_package(GSOAP REQUIRED)) May 11 11:56:58 I already build gsoap (bitbake gsoap) and that workerd fine May 11 12:02:41 kannerke, maybe you need to point it to where gsoap is May 11 12:02:53 s/it to// May 11 12:04:05 kannerke: most times the cmake scripts are not ready for cross compilation May 11 12:04:17 kannerke: try to find out where cmake looks for gsoap May 11 12:05:56 morphis: how can I make the script ready for cross compilation? May 11 12:06:38 patch must be availble because there must be a patch applied to gsoap for openssl and that worked May 11 12:14:44 kannerke: if you look at this http://git.openembedded.net/cgit.cgi/openembedded/commit/?id=aa4b7e83ed6f4a7b371c60a2a75575ba5bb3c787 May 11 12:14:49 how does openembedded-core-contrib/master relate to openembedded-core/master? May 11 12:14:54 I have there my own FindQt4.cmake script May 11 12:15:01 which is adjusted for cross compilation May 11 12:15:06 to search in the right directories May 11 12:15:28 there is maybe a FindGSOAP.cmake file somewhere May 11 12:15:34 which cmake uses to find GSOAP May 11 12:19:42 likewise: I always rebase my -contrib repos against openembedded-core/master not it's -contrib ekvivalent May 11 12:23:01 JaMa: so core-contrib should track core? who May 11 12:23:10 jama: who does that? May 11 12:23:47 RP is merging pull requests sent to oe-core ML May 11 12:24:06 I wonder why we have a core-contrib, where people have feature branches, which are intended for core. Why not have feature branches on core itself? May 11 12:26:46 don't know if gitolite supports not only per-branch permissions (it does), but also stuff like allowed -f for push and removing branches by uses (which is usefull for -contrib repo) May 11 12:27:38 JaMa: ok, so it's mostly permission based why -contrib is there May 11 12:28:11 JaMa: My concern is that core-contrib/master is going to divert from core/master slowly. May 11 12:28:28 likewise: core-contrib/master doesn't make sense to me May 11 12:29:01 likewise: rebasing against oe-core/master makes sense to be sure that it will merge cleanly when RP decides to pull it May 11 12:32:31 JaMa: exactly. May 11 12:33:26 but don't know who pushes oe-core-contrib/master or if it was really intentional May 11 12:39:55 JaMa: I have sent a message to verify if someone keeps -contrib/master in sync. May 11 12:42:17 morphis: there are both a FindGSOAP.cmake and a FINDGnuPatch.cmake file May 11 13:45:00 hi! can anyone help me in 'root' login without keyboard rootfs created using OE? May 11 13:45:28 i mean can i login as root using script?? May 11 14:37:46 varunb: yes that's possible.. you would have to change /etc/sysconfig/gpelogin May 11 15:13:58 Hello everyone. I created a new kernel object and installed under ${D}/...lib/modules..../dsp/new.ko. using the default "do_install" I expect will automatically add my new package to "PACKAGES" May 11 15:14:06 I get this error "following files were installed but not shipped in any package" May 11 15:14:11 with my kernel object May 11 15:14:32 but if I add my recipe to "PACKAGES" it says that my recipe is included twice. May 11 15:15:13 From reading the documentation, I think that the right way to do this is to use the default do_install(){} and populates "FILES_${PN}" May 11 15:22:00 mrmichael: did you have a question? May 11 15:22:59 hi May 11 15:23:24 yes. for some reason my new.ko kernel object isn't getting added to a package, even though I've added it to FILES_${PN} May 11 15:23:29 that's the only file I've created May 11 15:23:32 I've installed it May 11 15:23:36 I list it May 11 15:23:52 can you paste your .bb file somewhere? May 11 15:23:57 but when I run the default do_install() it says "no files included, not creating package" May 11 15:23:58 yes May 11 15:24:27 I'm surprised that do_install() would say that. Is that the exact message it prints? May 11 15:25:05 NOTE: package eric-0.1-r97i: task do_package_write_ipk: Started Packaged contents of eric into /home/erict/overo-oe/tmp/deploy/glibc/ipk/armv7a/ eric_0.1-r97i.6_armv7a.ipk NOTE: Not creating empty archive for eric-doc-0.1-r97i.6 Packaged contents of eric-dev into /home/erict/overo-oe/tmp/deploy/glibc/ipk/arm v7a/eric-dev_0.1-r97i.6_armv7a.ipk May 11 15:25:31 it's do_package() not do_install() that prints it. my bad May 11 15:25:33 okay, so that is the -doc package that it isn't creating. May 11 15:26:13 it apparently is creating eric_0.1-r97i.6_armv7a.ipk and eric-dev_0.1-r97i.6_armv7a.ipk so those must not be empty. May 11 15:26:20 it's the package which contains my new kernel object. I am not createing a doc package. I'll cut/paste the whole recipe May 11 15:26:34 DESCRIPTION = "ERIC's Personal Program" PR = "r0" LICENSE = "GPLv2" PR = "${MACHINE_KERNEL_PR}" PR_append = "i" PROVIDES = "eric" export LINUXKERNEL_INSTALL_DIR = "/home/erict/overo-oe/tmp/sysroots/armv7a-angst rom-linux-gnueabi/lib/modules/2.6.36/build" SRC_URI = "file://eric.c \ file://Makefile \ file://README.txt" inherit autotools do_compile() { unset CFLAGS CPPFLAGS CXXFLAGS LDFLAGS cd ${WORKDIR} May 11 15:26:40 args May 11 15:26:46 ~pastebin May 11 15:26:46 [~pastebin] A "pastebin" is a web-based service where you should paste anything over 3 lines so you don't flood the channel. Here are links to a few : http://www.pastebin.com , http://pastebin.ca , http://channels.debian.net/paste , http://paste.lisp.org , http://bin.cakephp.org/ , http://asterisk.pastey.net/ , or install pastebinit with yum or aptitude. May 11 15:27:03 ok May 11 15:27:07 I'll use one of those May 11 15:27:08 so, what exactly is in the eric_0.1-r97i.6_armv7a.ipk that does get created? May 11 15:27:17 just one kernel object May 11 15:27:18 eric.ko May 11 15:27:21 should be in it May 11 15:27:26 that's all I want May 11 15:27:28 right, but what actually _is_ in it? May 11 15:27:50 how can I print the contents of the package? I don't know the command May 11 15:28:19 or, to put it another way, what makes you think that this package doesn't contain the things you want it to? May 11 15:28:39 you can use dpkg-deb -c, or "ipkg files" if it's installed on your target, or bust it apart manually with ar and cpio. May 11 15:29:04 http://asterisk.pastey.net/149866 May 11 15:29:09 there's my pastey :) May 11 15:29:31 dpkg -X fooo.deb moo/ May 11 15:29:38 unpacks it May 11 15:40:00 I see! The DOCUMENT archive is empty.! May 11 15:47:52 eric_0.1-r97i.6_armv7a.ipk contains exactly what I want May 11 15:48:16 then there's a dbg and a dev version which are empty May 11 15:48:21 they contain only /. May 11 15:48:39 is that the right behavior of the default do_package() {} May 11 15:48:40 ? May 11 17:14:50 Hi everyone May 11 17:52:02 khem, ping May 11 17:53:21 khem, i don't have enough diskspace nowadays, could you 'git mv ./libpthread/nptl/sysdeps/generic/unwind.h include/' and see if that breaks ARM EABI compilation of libubacktrace? TIA May 11 18:01:16 khem, (see ML) May 11 18:58:21 hi ensc May 11 19:19:18 re May 11 19:26:09 hi frans May 11 19:26:22 ups effem May 11 19:30:39 ~lart eticket.linuxtag.org May 11 19:30:39 * ibot tries to shut eticket.linuxtag.org up May 11 19:31:38 We still can't send tickets... May 11 19:35:24 hi woglinde, all May 11 19:36:27 is some native english speaker around? May 11 19:37:50 i'd /but with an FLAGS parameter/s/an/a/ ;# no? I can parse 'an fd' but not 'an flag'.. May 11 19:39:47 eFfeM, i suppose your english is better than mine. May 11 19:51:09 I'd say it is a flag (or a FLAGS in your example) May 11 19:51:23 (sorry did not see the q, was typing an email) May 11 19:52:55 blindvt: and there is always the google test: search for "a flag" gives 9.600.000 hits) search for "an flag" gives 36.700 hits May 11 19:53:00 iirc, 'an' is used with words starting from vowels May 11 19:53:26 an apple e.g. May 11 19:54:01 heh, that's right. May 11 19:54:03 Jay7: that is more or less correct, afaik an fd would also be ok as you say "efde" so when saying it it is like it starts with a vowel May 11 19:54:18 and it's "an eff dee", thus 'i open an fd' is ok since it's pronounced as a vowel May 11 19:54:23 wait. what? May 11 19:55:04 then again the google test says "a fd" 1.470.000 hits, "an fd" 248.000 May 11 19:55:12 so not too sure about that May 11 19:55:23 i was never good at english ;-) May 11 19:55:51 me too :) May 11 19:56:11 oh well :P May 11 20:04:21 we need kergoth May 11 20:31:53 PaulePanter, there you are May 11 20:33:22 blindvt I will see him tomorrow May 11 20:33:26 again May 11 20:37:36 khem, kergoth you guys around today? want to discuss SDK some more May 11 20:39:39 I find that the issues I'm seeing with building autoconf projects with meta-toolchain largely stem from the fact that the various binconfig's (ie freetype-config) from the -dev ipks (ie freetype-dev.ipk) do 'not' have their paths fixed up as the ones in OE's STAGING_DIR so when autoconf uses them they spit out / based paths incorrectly May 11 20:40:17 I figure this is because the -dev packages are also meant to be installed on a target that has gcc, yet because they are also used to provide libs for sdk they conflict May 11 20:41:02 I'm still learning about autoconf but I'm curious why these various -config's exist at all instead of just using pkgconfig which looks to be a much cleaner/flexible way to do the same thing May 11 20:42:16 because they existed long before pkgconfig ever did May 11 20:43:36 ok, I buy that - so how do we get around the issue of them not providing the right paths when used with sdk? May 11 20:43:58 and why hasn't autoconf updated to use pkgconfig? May 11 20:44:46 tharvey: to use pkgconfig for what? May 11 20:45:25 just curious why autoconf hasn't evolved to use pkgconfig instead of the various -config apps to obtain cflags/libs etc May 11 20:45:39 autoconf itself doesn't get to choose May 11 20:46:03 generally, those checks are done by m4 macros which are shipped by the libraries in question May 11 20:46:04 look at freetype-config for example - its a stand-alone shell script that provides the same thing pkgconfig freetype2 provides - yet pkgconfig is more flexible and maintainable May 11 20:47:00 oh? I thought the m4 macros came as part of autoconf - or at least can be overridden by macro's for your system May 11 20:47:24 wouldn't you want macros tuned for your host (or sdk in my case) to be making those decisions? May 11 20:48:02 in the case of freetype, the macro in question is AC_CHECK_FT2, and it's distributed (primarily) as part of freetype May 11 20:48:32 I'm fairly sure you will not find that in the autoconf source tarball, though I haven't checked May 11 20:49:05 autoconf itself is quite a simple-minded tool, and in some sense it is a tool from a bygone era. May 11 20:49:36 it doesn't really have any semantic knowledge of the things it's testing for, and it certainly doesn't have any understanding that there is an equivalence between pkg-config and freetype-config. May 11 20:49:47 I would guess your right - its also not part of libfreetype-dev May 11 20:50:45 pb_, so that you understand the issue I'm trying to work out: I can't figure out how to use meta-toolchain sdk for example to build gd May 11 20:51:37 I'm supporting an sdk and trying to document how to use it appropriately and the docs that tell me I simply need to do a './configure --host=$TARGET_SYS' are incorrect in this case, and I assume for a lot of other cases as well May 11 20:51:59 ie, anything that looks for X headers also adds / to your CFLAGS which is not correct for an sdk May 11 20:53:31 yeah, that is an unfortunate piece of suckage May 11 20:54:00 I don't think there is any "silver bullet" fix for that; I can't think of any way of patching autoconf that would make it work better (and not introduce other obscure failure modes) May 11 20:54:22 would it help to include our own autoconf in sdk? May 11 20:54:38 no, not really. May 11 20:54:50 if you're just running ./configure, autoconf doesn't get invoked at that stage May 11 20:55:09 autoconf is used (at some prior time) to generate the configure script, but it isn't needed at the point where configure runs May 11 20:55:41 but you can run autoreconf - just not sure what can be done to make that work either.... it still ends up pulilng invalid macro's May 11 20:56:00 shipping your own h4x0red autoconf might work if you taught people to always rerun autoreconf, but that can lead to a world of pain of its own so it is probably not a good idea in general. May 11 20:56:10 unless perhaps the rule is to run autoreconf and macro's in the SDK are used that do the right thing... I think that's what khem was telling me 'should happen' May 11 20:56:42 yeah, in a perfect world that is exactly what would happen May 11 20:56:49 I'm ok with telling people they have to run autoreconf (oe for example does this to build gd w/o any patches) May 11 20:59:19 still trying to understand why khem keeps talking about libtool helping here and where that comes into play May 11 20:59:42 I don't understand why libtool would be involved either. May 11 21:00:20 unless he's talking about something to do with paths in .la files, but that is a separate issue (and probably irrelevant for you) May 11 21:02:39 hmm.. can't see to find via opkg/info/*.list where freetype2.m4 comes from in sdk May 11 21:06:31 pb_, how do you figure out that AC_CHECK_FT2 is responsible for a check? looking at configure scripts seems difficult - there is no reference there back to that macro, and I can't figure out what includes that macro either (its not in configure.ac) May 11 21:11:22 ah, so, gd is an unfortunate example of a library which doesn't use the macros, and instead has cut-and-pasted them into its own configure.ac script May 11 21:11:58 hm isnt gd older than freetype? May 11 21:12:17 probably, but it supports freetype as a build option May 11 21:12:35 pb sure and the only dev knowing about tools left the project? May 11 21:14:24 tharvey: so, in this case, even shipping your own macros won't help. I don't think there is any feasible way of fixing this other than shipping a freetype-config utility which does the right thing. May 11 21:14:47 pb fixing gd May 11 21:14:55 and sent the patch upstream May 11 21:15:14 well, sure, but that's not what we are talking about May 11 21:15:42 actually, I guess you could provide a custom AC_PATH_PROG which noticed when you were asking for freetype-config and instead handed you back "pkg-config freetype2". May 11 21:15:53 no, I don't think gd is 'wrong' May 11 21:15:56 that would probably be good enough to keep gd happy, at least May 11 21:16:09 pb_, not sure how you say my own macros wont work May 11 21:17:07 tharvey: well, as you said, gd's configure.ac doesn't use freetype's macro, it has its own hand-coded test for freetype May 11 21:17:33 ie if I can find out what pkg in the sdk provides freetype2.m4 I could re-write it to use pkgconfig for example, or provide arg to config-freetype which is another route that adds a prefix May 11 21:18:07 well, yeah, but that won't help with gd since it doesn't appear to actually use freetype2.m4 May 11 21:18:17 dnl FreeType configure tests snarfed from libwmf .. May 11 21:18:17 AC_ARG_WITH(freetype,dnl May 11 21:18:20 etc May 11 21:18:25 is what it seems to be doing May 11 21:18:26 ah... thats why I couldn't find it in configure.ac - so where is it handcoded? ie if I do an autoreconf isn't configure completely regenerated from configure.ac and others (what others?) - if I run autoreconf it still has the check for freetype in there May 11 21:18:37 yeah, it's in configure.ac May 11 21:21:11 ok so in this case you 'can' consider gd 'broken' in that it doesn't use the std macro's to do freetype detection - but even if it did that detection might also do the wrong thing (but could be updated/provided correctly in the sdk if autoreconfigure is used May 11 21:21:42 right May 11 21:24:13 any idea what provides freetype2.m4 in sdk? not clear why grep freetype $SDK_PATH/usr/lib/opkg/info/*.list and grep freetype $SDK_PATH/$TARGET_SYS/usr/lib/opkg/info/*.list come up empty May 11 21:24:43 libfreetype6-dev: /usr/share/aclocal/freetype2.m4 May 11 21:24:43 I would have expected it to be in libfreetype-dev. May 11 21:24:48 its on ubuntu May 11 21:24:53 doh nvr mind... it comes from libfreetype-dev May 11 21:24:55 if it's not in there then that sounds like a packaging error in the recipe May 11 21:33:46 I notice that the choice taken for OE internal build for gd and likely others is to hack all the *-config files to do the right thing (binconfig.bbclass) - wondering if that kind of general approach wouldn't be better for an sdk. Such as packaging up the various *-configs and installing them in the SDK path or something May 11 21:35:53 yeah, that might be a fine plan May 11 21:36:38 I'll post to the ml and discuss that May 11 21:37:46 still want to hear from khem about how libtool may affect this and how/when to use LIBTOOL_SYSROOT_PATH in an SDK - he seemed to indicate that if gd's configure was doing the wrong thing after running autoreconf that it must be libtools fault May 11 21:40:06 yes, I am still baffled as to what he can have meant by that May 11 22:00:48 Tartarus, are we suppose to ack pull requests for the release maintenance branch? May 11 22:01:19 You may. I haven't pulled stefan's yet just because my laptop has been all busy-melty today May 11 22:01:32 And Koen's hasn't yet had the 24h review period May 11 22:04:49 hmm... can't recall, can you not have symlinks in ipks? May 11 22:05:13 good nite May 11 22:05:19 Crofton|work, ping? May 11 22:05:27 piong May 11 22:05:34 Tartarus, ok May 11 22:05:48 I need to send one for system-tools-updates May 11 22:06:06 /msg Crofton|work would it be okay to have a gitadmin@openembedded.org user? May 11 22:06:10 whoops May 11 22:06:12 hee he May 11 22:06:45 * ka6sox-work wants the github system out of his name :D May 11 22:09:26 sre May 11 22:10:10 okay I'll work with philb on getting that done May 11 22:10:18 then we can use the http: method too. **** ENDING LOGGING AT Thu May 12 02:59:58 2011