**** BEGIN LOGGING AT Fri Aug 28 02:59:56 2009 Aug 28 03:05:29 hey can someone tell me how to get 'cc1' to show what options were passed to it during compile?' Aug 28 03:05:31 * m4t forgets Aug 28 03:07:25 oh --print Aug 28 03:07:55 er thats gcc-xml Aug 28 03:07:56 damn Aug 28 03:11:32 oh duh, i was calling cc1 instead of *gcc Aug 28 03:11:54 -v shows : -disable-__cxa_atexit --enable-tls=no Aug 28 03:12:02 thanks, zecke Aug 28 03:29:56 *mutters bad things about Apache's build system* Aug 28 03:37:00 m4t: I have fixed the powerpc uclibc gcc for you Aug 28 03:37:06 one patch is remaining Aug 28 03:37:14 which I have nt pushed yet Aug 28 03:38:50 has anyone ever seen this type of error message? ERROR: This autoconf log indicates errors, it looked at host includes. Aug 28 03:41:33 ah damn looks like apr-util is pulling in pg_config and odbc_config from the host system. Aug 28 03:41:44 yea i've seen those Aug 28 03:42:02 m4t: any advice? Aug 28 03:42:17 khem : word Aug 28 03:42:18 thanks Aug 28 03:42:53 kg4ysn i think i ended up passing some specific include directories to ./configure Aug 28 03:43:07 m4t: ok thanks. i might see about disabling pg and odbc for now. Aug 28 03:48:32 kg4ysn that comes from insane.bbclass btw Aug 28 03:48:52 m4t: is there any way to get it to spit out exactly what it's having issues with, or is it non-specific? Aug 28 03:49:50 grep 'CROSS COMPILE Badness: Aug 28 03:50:21 you might be able to look at config.log and see which specific headers it was looking for when it hit /usr/include or the like Aug 28 03:50:44 i think that configure.in, or one of the .m4 files might have default places to check in it somewhere Aug 28 03:51:22 m4t: wow that's too cool :-) yeah it found it. it's pulling in the postgresql libraries for some odd reason. Aug 28 03:53:11 how to fix it will be dependent on how ./configure ends up being run Aug 28 03:53:29 m4t: i'm searching for a --disable-feature flag now. Aug 28 03:54:03 its usually run by autoreconf i think Aug 28 03:54:09 kg4ysn yea that is how i go about it Aug 28 03:54:34 when i build things from source tarballs i usually just ./configure --with-options, rather than messing with autoreconf and that kind of thing Aug 28 03:54:38 really, its all new to me Aug 28 03:54:56 m4t: you're doing good, it seems. i appreciate the help! Aug 28 03:55:00 in most recipes, it seems that autoreconf is the preferred way Aug 28 03:56:21 try cd openembedded/recipes; ls -l */files/*host*inc* Aug 28 03:58:12 oh great, so the tls error is gone with the gcc recompile Aug 28 03:58:16 but this one still pops up: | /home/matt/devel/tmp/cross/ppc405/lib/gcc/powerpc-angstrom-linux-uclibc/4.4.1/../../../../powerpc-angstrom-linux-uclibc/lib/nof/libstdc++.so: undefined reference to `void std::__convert_to_v(char const*, long double&, std::_Ios_Iostate&, short* const&)' Aug 28 04:10:33 Configuring locale-base-en-gb Aug 28 04:10:33 localedef invoked oom-killer: gfp_mask=0x1201d2, order=0, oomkilladj=0 Aug 28 04:10:42 ... that's not good :( Aug 28 04:20:02 205-uclibc-locale-update.patch: __convert_to_v(const char* __s, long double& __v, ios_base::iostate& __err, Aug 28 04:20:05 that looks about right Aug 28 04:23:21 m4t: thats not needed Aug 28 04:23:33 m4t: I will push a patch shortly that I am testing Aug 28 04:23:37 then you should be fine Aug 28 04:23:44 oh, cool Aug 28 04:24:05 we have had a new locale called uclibc so far Aug 28 04:24:27 but in libstdc++ there is uclibc stuff that can be used with generic locales Aug 28 04:24:42 so I have coerced 4.4.1 to use it Aug 28 04:24:55 and so far it was only tested for arm Aug 28 04:25:19 the problem that we have is that uclibc does not support 128 bit long doubles Aug 28 04:25:20 there is this one too Aug 28 04:25:24 powerpc-link-with-math-lib.patch Aug 28 04:25:30 let me post it Aug 28 04:25:31 yes I added that Aug 28 04:25:34 already Aug 28 04:25:35 oh cool Aug 28 04:25:38 yea my memory sucks :/ Aug 28 04:25:41 update it Aug 28 04:26:15 gcc builds take forever I am waiting for it to finish before I push the patch Aug 28 04:26:38 after that I would suggest that you clean out your tmp and build from scratch Aug 28 04:26:52 and let me know if you get stuck in toolchain parts Aug 28 04:26:54 yea i can do that Aug 28 04:27:23 i have to chpax git before it can fetch uboot Aug 28 04:27:31 and uhm, python is tricky like that as well Aug 28 04:27:44 so it does require a bit of interaction to get a toolchain going Aug 28 04:29:18 03Khem Raj  07org.openembedded.dev * r192b086f07 10openembedded.git/recipes/gcc/gcc-4.4.1.inc: Aug 28 04:29:18 gcc-4.4.1.inc: Disable 128 bit long double for powerpc-*-uclibc Aug 28 04:29:18 Signed-off-by: Khem Raj Aug 28 04:29:40 m4t: there you go now pull the latest and try it out Aug 28 04:30:14 btw. you should be using uclibc 0.9.30.1 Aug 28 04:30:21 yup i am Aug 28 04:30:38 cool then try to rebuild and let me know how it goes Aug 28 04:30:50 will do, thanks a lot for spending some time on that Aug 28 04:31:31 no problem, I am encouraging people to use gcc 4.4.1 Aug 28 04:32:04 I have already fixed it for sh,arm,powerpc,mips Aug 28 04:32:28 someone should try it on x86 but that should work I expect Aug 28 04:32:51 once the configuration parts are settled Aug 28 04:32:58 I will start pulling patches Aug 28 04:33:07 bug fixes etc. Aug 28 04:34:07 * khem out for dinner Aug 28 04:48:13 cool i think i got it, i had to figure out git first Aug 28 05:01:38 03Koen Kooi  07org.openembedded.dev * r8c8ece9df6 10openembedded.git/: Merge branch 'org.openembedded.dev' of git@git.openembedded.org:openembedded into org.openembedded.dev Aug 28 05:01:49 03Brijesh Singh  07org.openembedded.dev * ra6e10835a1 10openembedded.git/ (4 files in 2 dirs): Aug 28 05:01:49 ti-lpm-module 2.24: update lpm src uri, it's in its own tarball now Aug 28 05:01:49 * fix DSPLINK path Aug 28 05:01:49 * keep versioning inline with CE till we sort it out properly Aug 28 07:02:53 Hello all, I'm a newbie, and i have synced up the openembedded and bitbake, and setup the environment as per the user manual. Aug 28 07:03:04 I'm currently facing the following issue Aug 28 07:03:30 NOTE: Running task 20 of 6920 (ID: 32, /home/harinarayanan/work/oe/openembedded/recipes/quilt/quilt-native_0.46.bb, do_fetch) NOTE: fetch http://download.savannah.gnu.org/releases/quilt/quilt-0.46.tar.gz --2009-08-28 11:53:14-- http://download.savannah.gnu.org/releases/quilt/quilt-0.46.tar.gz Resolving download.savannah.gnu.org... 199.232.41.75 Connecting to download.savannah.gnu.org|199.232.41.75|:80... failed: Connection timed out Aug 28 07:04:28 it is trying to fetch a particular quilt-0.46.tar.gz and it fails to find it Aug 28 07:04:28 hari___: it should continue to try another mirror? Aug 28 07:04:49 yes it does, but it fails even then and the build exits Aug 28 07:04:51 hari___: it dosn't fail to find it... downloads.savannah.gnu.org does not respond Aug 28 07:05:04 hari___: so your networking is not setup properly, firewalled or such Aug 28 07:05:14 it is firewalled. Aug 28 07:05:37 in previous steps it downloaded files aswell from http servers Aug 28 07:05:54 hari___: OE is executing "wget http://download.savannah.gnu.org/releases/quilt/quilt-0.46.tar.gz" and that seems to fail Aug 28 07:06:12 zecke : is there any specific setting to be done for this Aug 28 07:07:15 hari___: No idea, in more simple words "telnet download.savannah.gnu.org 80" will not open a connection because something in your setup forbids it Aug 28 07:07:22 do i need to set http_user and http_passwd somewhere Aug 28 07:07:54 hari___: I have no idea about your firewalled internet, I would quit my job and search a company were I don't have firewall nazis Aug 28 07:08:06 the *.tar.gz file can be downloaded from firefox. Aug 28 07:08:18 :) Aug 28 07:09:52 hari___: well, seriously... find out how to use your corporate proxy Aug 28 07:11:01 or is there some way to skip quilt build. (probably it might cause build issues for other recipes) Aug 28 07:11:40 hari___: right, it will cause issues... OE needs internet to download sourcecode, no internet no sourcecode Aug 28 07:12:17 hari___: you can execute bitbake -c fetchall at home, then put everything on floppy disks, then go to your office, and put that on a local server :) Aug 28 07:13:22 is there a posibility that i can get the built files online? Aug 28 07:13:41 hari___: sorry? Aug 28 07:13:47 hari___: you mean the to be used sourcecode? Aug 28 07:14:26 the files that are the result of the bitbake Aug 28 07:15:04 hari___: I don't get it. If you ask if you can just use a read built distribution... then the answer would be yes, you could use angstrom and then install already built packages Aug 28 07:16:27 sorry, I'm still not familiar with the flow yet. Aug 28 07:17:28 hari___: what problem do you attempt to solve with OE? Aug 28 07:17:54 I wuld want to use openembedded tree, and build the beagleboard-demo-image first, later on i would want to add my custom code/drivers to it Aug 28 07:18:47 what would be the best way to get started with it Aug 28 07:19:39 hari___: In your organisation, figure how to be able to download from the command line. :) Aug 28 07:20:08 hari___: and in general for just using the beagleboard, you can use a precompiled version of Angstrom, put that on the SD card, use opkg install to install more software Aug 28 07:20:50 hari___: if you want to compile drivers, sooner or later you will need git... which means you will need to figure out how internet works :) Aug 28 07:28:55 morning Aug 28 07:31:22 how do I reset a bitbake package? (I'm trying to write a .bb file for a new package and want bitbake to restart from scratch for that package) Aug 28 07:32:29 czr: depends on what scratch means. you can do bitbake -c listtasks -b your/file.bb... and it lists e.g. "clean" Aug 28 07:32:39 czr: this will clean the work area of this one recipe Aug 28 07:33:08 zecke, I think that's what I want, let me try Aug 28 07:37:28 hey zecke , that tls patch worked, but i had another error due to some 128bit long long thing that khem fixed Aug 28 07:37:34 good morning Aug 28 07:37:40 zecke, that worked, thanks Aug 28 07:37:46 * czr is now having other troubles :-) Aug 28 07:37:54 bloody autohell.. Aug 28 07:37:58 so, now with a fresh get pull, the other issues are fixed, but the tls problem with gnutls happened again Aug 28 07:38:13 i am going to let it run overnight, but with your powers combined, i think its fixed Aug 28 07:38:55 zecke, if a package already contains a configure script, does it make sense to "inherit autotools" in the bb recipe? Aug 28 07:39:34 seems that it will now recreate configure, which leads to failing build (I guess). Aug 28 08:07:30 czr: yes, it makes sense Aug 28 08:07:51 czr: we want to regenerate the libtool script, and some other builtin autoconf checks Aug 28 08:08:11 czr: e.g. the classic sizeof check used to compile an app, run it, print things, we use size/nm to determine this Aug 28 08:10:50 good morning Aug 28 08:23:12 florian: good morning Aug 28 08:23:15 zecke: good morning Aug 28 08:23:17 can anyone recommend an arm dev board? I'm looking for ~500Mhz, <$500, USB, but not the OMAP processor as on beagleboard Aug 28 08:23:21 or evening, I suppose Aug 28 08:23:38 and openembedded support Aug 28 08:23:59 mini2440? not quite 500MHz but I think it meets the other criteria Aug 28 08:24:53 out of interest, why do you not want omap? Aug 28 08:25:10 i'm having problems with usb performance Aug 28 08:25:24 ah, I see Aug 28 08:25:38 the Inventra USB controller is a POS as far as i can see Aug 28 08:28:04 cajun-rat: there are several cheap boards with usb based on some pxa or freescale soc. but usually they are modules that requite a more expensive baseboard. Aug 28 08:28:59 cajun-rat: the openrd might be a choice... very fast i/o but needs more power compared to an omap Aug 28 08:30:29 power isn't a problem openrd might be good, given marvel are behind it Aug 28 08:31:55 florian: I've been hit by USB poor performance since it came to me (Intel 8237-1AB/EB). 300 Aug 28 08:32:05 300 power, btw Aug 28 08:32:35 what has poor usb performance? Aug 28 08:32:45 ant_work: true... Aug 28 08:32:51 cajun-rat: joking ;) Aug 28 08:32:59 cajun-rat: ieee1394 ;) Aug 28 08:33:17 heh Aug 28 08:33:37 cajun-rat: it really depends on what you intend to achieve Aug 28 08:33:40 zecke, right. this is the problem that I'm having: a project (http://www.cleancode.org/projects/email/). It uses autotools. it has configure (and .in) in main dir and also a separate ones in a dir called "dlib" which contains some generic code that the project uses. Aug 28 08:34:00 zecke, now, the main level configure gets regenerated, but the one under dlib doesn't. infact, oe removes the old configure script. Aug 28 08:34:23 so, when the oe generated configure is run, and at the end it tries to execute dlib/configure, that will fail. since it doesn't exist. Aug 28 08:35:48 thanks for the suggestions. the marvel sheeva processor is a good call Aug 28 08:40:07 hmm. is it possible to run oe_runconfigure in multiple directories? Aug 28 08:40:24 once at the main level, and second time in dlib, before actually executing the configure-scripts? Aug 28 08:40:45 or hmm. I guess oe_runconfigure also executes the ./configure script.. Aug 28 08:48:46 hmm. autotools_do_configure() seems to remove recursively all configure scripts, but it will regen only the top level one. using stable though. Aug 28 08:57:22 czr: go to the other directories too... so this build system is calling another configure script? Aug 28 08:57:39 recursive make is considered evil, i think there is no word for recursive configure :) Aug 28 08:57:40 zecke, yes. there's one at the top-dir and one in ./dlib. Aug 28 08:57:47 double the evil? :-) Aug 28 08:58:05 tail -3 ./configure.in: Aug 28 08:58:07 cd dlib Aug 28 08:58:07 ./configure $args Aug 28 08:58:07 cd .. Aug 28 08:58:15 nice, ain't it. Aug 28 08:58:51 anyhow, the autotools bbclass function (autotools_do_configure) will remove recursively all generated configure-scripts Aug 28 08:58:57 but it will only regenerate the top-level one. Aug 28 08:59:12 czr: go to the dlib dir and call autotools_do_configure? Aug 28 09:00:18 so something like this: do_configure() { ( cd dlib; autotools_do_configure ); autotools_do_configure } ? Aug 28 09:04:33 zecke, cool, seems to do the trick Aug 28 09:04:43 lets see whether the universe will end when running the binary Aug 28 09:09:06 zecke, seems to work, thanks a bunch Aug 28 09:26:53 my OE generates all locales for some packages... I don't like this. is there a way to limit this? for example want only EN (english) locales. Aug 28 09:28:07 and how can I delete the other locales? Aug 28 09:28:46 sx: why? they don't harm... Aug 28 09:28:47 sx: LIBC_GENERATE_LOCALES = "en_US.UTF-8 en_GB.UTF-8 de_DE.UTF-8 it_IT.UTF-8 fr_FR.UTF-8" Aug 28 09:29:00 this is for the bigger suxxer: glibc Aug 28 09:29:07 *biggest Aug 28 09:30:12 when I write ls angstrom-dev/deploy/glibc/ipk/armv4t , I see many locales.. Aug 28 09:30:24 oh..I mangled the copy and paste.... is GLIBC_GENERATE_LOCALES fwiw Aug 28 09:30:42 * XorA wonders why L2 cache is broken on pxa320 Aug 28 09:30:46 ant_work, thanks Aug 28 09:31:09 sx: en_US.UTF-8 is a must Aug 28 09:31:13 (bug ?) Aug 28 09:31:54 one could just live with Posix C locales Aug 28 10:41:39 I've disabled all binary locale generation but I keep getting this error when building image: "Cannot find package glibc-localedata-i18n." Aug 28 10:41:51 any ideas? Aug 28 10:47:10 Here's the complete output http://pastebin.ca/1545671 Aug 28 11:16:09 morning Aug 28 11:16:12 no idea th1 Aug 28 11:18:40 its and OpenEmbedded channel? Aug 28 11:18:45 *its an? Aug 28 11:19:49 it is Aug 28 11:20:06 why there's no recipes for iptables-1.4.x in openembedded? Aug 28 11:20:16 only 1.3.8 Aug 28 11:20:50 because no one added them? send patch if you need them Aug 28 11:25:44 actually, xtables-addons is what I need but it seems that the whole xtables concept doesn't exist in 1.3 Aug 28 11:27:38 but yeah, I'll try to add Aug 28 11:49:44 I found it.. it does generate that package even with locales off, but it wasn't looking in the right place (PACKAGE_EXTRA_ARCHS had a missing entry) Aug 28 11:51:15 th1_: pls elaborate Aug 28 11:51:48 I had issues with opkg-cl -f -> * Cannot find package Aug 28 11:52:07 but only using packaged-staging Aug 28 11:54:27 http://tinderbox.openembedded.net/public/logs/task/2589035.txt Aug 28 11:54:59 fwiw s/I had/I have/ Aug 28 12:23:12 anyone here cre about erlang? Aug 28 12:24:30 http://cgit.openembedded.org/cgit.cgi/openembedded/tree/recipes/erlang/erlang_R13B01.bb Aug 28 12:24:49 There is a def get_erlan_libs python function in the recipe Aug 28 12:25:00 which was OK as long as there was only one recipe Aug 28 12:25:18 but I updated the version, so there are two copies of the function Aug 28 12:25:27 and this will not parse Aug 28 12:25:36 Any ideaa on dealiing with this? Aug 28 12:26:12 * Crofton|work is thinking of adding recipe version to function name is a hack Aug 28 12:27:17 Does moving it to erlang.inc work? Or are the two functions the same? Aug 28 12:29:23 they are the same :) Aug 28 12:29:52 but I suspect it will still appear "twice" if it is in the the inc file Aug 28 12:30:03 since it is "included" in each of the bb files Aug 28 12:30:42 I think this is why there are "classes" Aug 28 12:30:55 but this function is pretty recipe specific Aug 28 12:31:26 ok, but why the recipe-level python function are global then? Aug 28 12:31:33 *functions Aug 28 12:31:46 not sure Aug 28 12:32:01 I was hoping someone smart is around :) Aug 28 12:41:39 | /tmp/ccEqvIUs.s:9449: Error: selected processor does not support `ldrex r0,[r2]' Aug 28 12:41:48 urg Aug 28 12:41:50 glib-2.0 2.21.4 for armv5te ;( Aug 28 12:41:53 arnv5? Aug 28 12:41:59 crap Aug 28 12:42:09 ldrex is armv6 and up Aug 28 12:43:30 Crofton|work: I like the erlang concept of message passing vs. shared variables and locking Aug 28 12:43:52 1eb89aacbc6830fe242eac36ba8bcb61d98807f5 Aug 28 12:43:57 hrw, see that commit Aug 28 12:44:14 cbrake, a friend of mine is into erlang Aug 28 12:44:23 I'm just trying to get it to build Aug 28 12:44:43 it is on my list of things to learn Aug 28 12:45:29 cbrake, any thoughts on dealing with the python function in the recipe? Aug 28 12:45:34 I am about to hack :) Aug 28 12:46:02 ~curse glib02.0 Aug 28 12:46:04 May the fleas of a thousand camels infest your most sensitive regions, glib02.0 ! Aug 28 12:46:07 r4 is older then r1 Aug 28 12:46:12 ARGH!@!@!##%#$@^(%#&^@(%&!(#@$!!!!!!1 Aug 28 12:46:53 hrw, did you see the revision I referenced Aug 28 12:47:18 yes Aug 28 12:47:35 it is the "problem" I suspect Aug 28 12:47:51 yes Aug 28 12:48:20 I suspect it should only be applied to machines that support armv6 and up. not globally Aug 28 12:50:00 sent mail to ML Aug 28 12:51:59 hrw, did you see my question about python functions in bb files? Aug 28 12:52:23 saw, no idea now Aug 28 12:52:28 ok Aug 28 12:52:34 I hacked the function name for now Aug 28 13:09:44 Hey, How can I get the list of all ${} variables (those accessible via d.getvar()) Aug 28 13:10:14 bitbake -b recipe.bb -e Aug 28 13:12:47 hrw, thanks! Aug 28 13:14:28 03Philip Balister  07org.openembedded.dev * r8d169f4e30 10openembedded.git/recipes/erlang/ (8 files in 2 dirs): Aug 28 13:14:28 erlang : Add recipe for erlang-R13B01 Aug 28 13:14:28 * Move patches to a common directory as they are shared. Aug 28 13:14:28 * Better solution needed for pythin function name in recipe. Aug 28 13:16:17 * Crofton needs to learn to spell python Aug 28 13:17:47 Crofton|work: you added patches and did not bumped PR Aug 28 13:19:46 new version Aug 28 13:19:53 * Crofton|work forgot to look at PR Aug 28 13:20:00 I moved them for the old version only Aug 28 13:20:32 PR = O Aug 28 13:20:43 I do not see the problem? Aug 28 13:20:54 new recipe, no PR Aug 28 13:21:03 old recipe only moved them Aug 28 13:21:14 to avoid having two copies of common patches Aug 28 13:21:27 ok Aug 28 13:21:27 new recipe built with fewer patches :) Aug 28 13:21:37 thanks for checking though :) Aug 28 13:27:39 np Aug 28 13:51:16 03Koen Kooi  07org.openembedded.dev * r039ba48ab1 10openembedded.git/recipes/glib-2.0/ (glib-2.0_2.21.4.bb glib.inc): glib-2.0 2.21.4: fix pre armv6 builds Aug 28 14:08:54 crap... Aug 28 14:09:08 Xorg crashes like hell on at91sam9263ek ;( Aug 28 14:17:17 hi! Aug 28 14:17:22 some question Aug 28 14:18:34 there are some command in bitbake for query which versions ofa a package are available? Aug 28 14:18:40 bitbake -s Aug 28 14:24:19 004f88f03ecfceae06242fb3aa5f7a36f99a2573 should be backported to stable/2009 to fix qemu-native compilation on hosts with no zlib-dev package installed Aug 28 14:24:39 without it, I get: Error: zlib check failed Aug 28 14:28:27 check patchwork... Aug 28 14:28:43 booxter: and remember: ANYONE can send patch for stable/2009 Aug 28 14:28:46 anyone Aug 28 14:31:18 hrw: ah, I see it's already in patchwork for 2 monthes already... :-\ Aug 28 14:32:28 booxter: pester stable maintainers to Ack it Aug 28 14:33:04 I gaveup a bit but will send big set of updates soon for review Aug 28 14:33:23 so far they do not broke my set of tests Aug 28 14:37:31 what is our PR bump policy ? Aug 28 14:37:45 is it in wiki somehwere Aug 28 14:38:13 if result changes, bump Aug 28 14:40:23 ah Aug 28 14:41:02 hrw: so any change to configure ops. cflags or add/delete patches Aug 28 14:41:07 yep Aug 28 14:41:43 or even {un}packaging some file in a recipe Aug 28 14:41:59 yes Aug 28 14:43:22 do you guys have, by any chance, a recommendation regarding a well-built stylus/ballpoint pen? I only find hong kong crap on ebay Aug 28 14:43:37 PDAir Aug 28 14:43:51 I have their zaurus stylus with pen Aug 28 14:44:04 * CoreDump|home looks Aug 28 14:46:37 well, the search function is b0rked and I see no universal styluses in the menu. I'm really looking for a quality stylus with a clip like a ballpoint pen. Aug 28 14:46:53 ah.. I see Aug 28 14:47:00 like those openmoko ones Aug 28 14:47:13 they are surprisingly hard to come by Aug 28 14:47:28 morning Aug 28 14:47:34 hrw: I broke mine after a day or two... Aug 28 14:47:44 chouimat: morning Aug 28 14:47:47 similar here Aug 28 14:47:54 FUCK! Aug 28 14:47:55 hehe Aug 28 14:48:11 is there any target in OE which has working X.org???? other then beagleboard Aug 28 14:48:19 X.org not Xkdrive Aug 28 14:50:16 2.5 hours before the interview ... Aug 28 14:51:46 chouimat: new job? Aug 28 14:52:11 hrw: I hope Aug 28 14:52:37 hrw: got laid off in June Aug 28 14:52:59 03Koen Kooi  07org.openembedded.dev * r250a96582e 10openembedded.git/recipes/erlang/ (erlang-native_R13B01.bb erlang_R13B01.bb): erlang: also blacklist java 1.5 Aug 28 14:55:01 anybody knows USB port programming? Aug 28 14:55:09 chouimat: good luck man Aug 28 14:55:39 chouimat, good position? Aug 28 14:56:11 everything is better than no job =) Aug 28 14:56:40 Crofton: I think so ... I didn't expect them to call me back ... so I will see Aug 28 14:57:47 mun24: i'm just learning about it :) Aug 28 15:00:26 Crofton: http://www.transgaming.com/about/careers/linuxdeveloper/ Aug 28 15:00:48 hrw: Hi Aug 28 15:01:18 hrw: I have seen that you are interested by 2.6.30 patches for AT91... Aug 28 15:01:23 03Steffen Sledz  07org.openembedded.dev * re15d43a442 10openembedded.git/conf/checksums.ini: checksums.ini: add missing entry for diffstat 1.47 Aug 28 15:01:51 hi Aug 28 15:02:06 hi woglinde Aug 28 15:04:17 I have cooked a big patch series in preview for 2.6.30 and stored here: ftp://ftp.linux4sam.org/devel/linux/ Aug 28 15:05:36 noglitch: you have linux4sam access? Aug 28 15:05:46 hi florian and hi hrw Aug 28 15:05:55 hrw: I am Nico from Atmel ;-) Aug 28 15:05:56 ah.. Nico Ferre. there was no question then Aug 28 15:06:09 hehe Aug 28 15:06:27 noglitch: give me some kind of access and I will update your OE docs a bit. they are obsolete/deprecated etc now Aug 28 15:06:32 and of course we plan a big OE update Aug 28 15:06:49 2.6.30 updates... Ulf has one too Aug 28 15:07:01 my at91sam9263ek runs 2.6.30.5 now Aug 28 15:07:04 indeet I work with him Aug 28 15:08:57 47 patches.. nice set Aug 28 15:09:16 noglitch: how much of it will be usable on avr32 too? Aug 28 15:09:19 morning Aug 28 15:09:20 I hope that we can synchronize in a way.. Aug 28 15:09:25 kergoth: morning Aug 28 15:09:41 hrw: well in fact they are most from 9g45 inclusion... Aug 28 15:09:48 I see Aug 28 15:10:05 just noticed avr32 few times Aug 28 15:10:33 I hope that Ulf's branch is not too far from .dev one... Aug 28 15:10:34 hi kergoth Aug 28 15:12:32 Moreover, I do not know if it is not better for a stable offer like the one we propose on linux4sam.org to build on top of stable/2009 Aug 28 15:13:17 hmmm Aug 28 15:13:23 noglitch: stable/2009 is good base Aug 28 15:14:08 hrw: ok good Aug 28 15:15:06 noglitch: but you have to push to .dev first and then prepare patch for stable/2009 - thats policy Aug 28 15:16:11 is there a way i can make a kernel recipe satisfy rdepends for kernel modules that it provides as built-in? Aug 28 15:16:42 mike_cw: rdepends for kernel modules are error Aug 28 15:16:59 use rrecommends Aug 28 15:17:09 hrw: openvpn has one in .dev Aug 28 15:17:12 hrw: that was the next question: ok understood Aug 28 15:17:20 mike_cw: then it is the bug Aug 28 15:18:03 ok. i'll file a report Aug 28 15:18:09 and patch :) Aug 28 15:18:31 shopping now Aug 28 15:19:23 I hate busybox sh Aug 28 15:21:29 who doesn't? Aug 28 15:22:14 I also hate vortex86sx core Aug 28 15:22:41 300MHz 486sx... no fpu, no power... only gfx chip is nice in box Aug 28 15:23:01 hrw: what do you need more? :-P Aug 28 15:23:07 ;D Aug 28 15:24:32 noglitch: 9g45? good to know :-D Aug 28 15:25:12 mckoan: plan to get one? Aug 28 15:25:33 hehe Aug 28 15:25:47 uf.. mb desktop etc booted Aug 28 15:26:33 hrw: yes, is in design phase Aug 28 15:27:14 noglitch: any plans to get CAN on at91sam9263ek working? Aug 28 15:27:45 hrw: touchè :-D Aug 28 15:28:26 mckoan: Hi Marco ! Aug 28 15:28:46 noglitch: hi Nico Aug 28 15:29:38 I wonder how to ship /etc/X11/Xsession.d/90matchbox-session in image Aug 28 15:30:06 In fact I am collecting all patches that we sent to several mailing-lists to several subsystems maintainers. So yes, 9g45 patches are there to wait 2.6.32 Aug 28 15:30:40 has anyone looked into patching opkg so it won't hang on its prompts (i.e. conffiles) when stdin isn't a tty? Aug 28 15:31:02 hrw: I recall some socket CAN patches floating for at91sam9263... Aug 28 15:31:16 anyone know if a doxygen-native recipe has ever been attempted? Aug 28 15:31:32 noglitch: I recall them too but usually with big 'do not use it unless you are insane' Aug 28 15:32:01 hrw: outh ! Aug 28 15:32:26 hrw: it was refering to driver implementation or hardware ? Aug 28 15:32:54 driver Aug 28 15:32:59 too many versions of it Aug 28 15:33:52 florian: I am trying to locate USB port on my board, iT has 5 pin and 2 pin header without pin. I already located serial and Jtag and using them Aug 28 15:33:52 ok, good to know Aug 28 15:34:03 khem , that patch along with zecke's tls gcc tls patch let minimal-image build through just fine Aug 28 15:37:07 florian: Moin. Did I misunderstand you yesterday that you had already deleted my account? I'm still receiving mail from it. Aug 28 15:41:33 florian: alive? Aug 28 15:43:00 mun24: oh.. ok - the lines from the cpu only or are the common resistors and protection stuff supposed to be there too? Aug 28 15:43:06 Laibsch: you did Aug 28 15:43:11 hrw: yes Aug 28 15:44:05 florian: OK. Your "np" was probably directed at something else, then. I'd appreciate it if you deleted the account as soon as possible. Aug 28 15:44:25 florian: gpe-login launches /etc/X11/Xsession after login? Aug 28 15:44:55 Laibsch: like i said, no chanc for me to learn about oe mail soon Aug 28 15:44:58 I can't do the tracing because the board is many layered so don't know where the connection is going. I can only write to some register in the CPU and then check if the signal at the pin is low or high Aug 28 15:45:09 florian ^ Aug 28 15:45:39 hrw: i think so yes Aug 28 15:46:15 mun24: oh... this might be nasty. are you able to control the usb pins in this way? Aug 28 15:46:17 florian: but USB is wierd it doesn't work unless it is connected to something Aug 28 15:46:17 ah.. gpe-session-scripts is what I needed Aug 28 15:46:45 * kergoth tries running an lsb compliance check on an OpenEmbedded rootfs, and cringes in anticipation Aug 28 15:47:38 kergoth: heh Aug 28 15:59:41 welp, looks like a bitbake modified to pass each task its own fresh COW datastore, so its changes are lost on exit, doesn't break builds Aug 28 15:59:44 so thats good news Aug 28 16:10:31 RP: opinion on this? mind if i push it to master? Aug 28 16:11:34 * kergoth doesn't feel comfortable pushing things without at least one other pair of eyes to say it isn't completely insane Aug 28 16:12:02 wonder if reviewboard would be worth looking at as an alternative to patchwork/email Aug 28 16:12:51 kergoth: What did you do exactly? Aug 28 16:13:06 I thought we already did have a new data dict for each task? Aug 28 16:13:22 hmm, didn't look like it, maybe i missed it Aug 28 16:14:25 I'm not seeing it, do you know where it's done? Aug 28 16:14:56 trybuild does a loaddatafull and passes that directly to buildpackage Aug 28 16:15:06 cache doesn't copy it in the load data, so.. /shrug Aug 28 16:17:06 if it is, great, i'll cross that off the list :) Aug 28 16:17:20 oh, i see, i am blind Aug 28 16:17:25 nevermind, ignore me Aug 28 16:17:28 have a nice rest of the day Aug 28 16:17:29 * kergoth moves on to the next task Aug 28 16:17:32 later mckoan Aug 28 16:18:12 RP: clearly i hadn't had enough caffeine yet ;) Aug 28 16:19:15 * kergoth tries to convert his giant mess of ideas in his omnioutliner into a concrete task list Aug 28 16:19:22 * rkirti passes kergoth some a cofffe mug Aug 28 16:19:27 yay Aug 28 16:19:27 s/some// Aug 28 16:19:40 err..s/cofffe/coffee Aug 28 16:19:59 christ, this is a long list Aug 28 16:20:01 * kergoth sighs Aug 28 16:20:17 kergoth: Sorry, I got distracted. Its in build.py Aug 28 16:20:24 kergoth: but you've noticed now :) Aug 28 16:20:59 :) Aug 28 16:21:27 well thats good, one less thing to be concerned with. making our builds as deterministic as possible is, i think, one of the key areas to work on Aug 28 16:21:34 if annoying to try to do anything about.. Aug 28 16:25:47 ldconfig in hal.postinst suxx Aug 28 16:26:03 or whatever in hal.postinst ;D Aug 28 16:41:27 opinion on this task? autotools.bbclass: If a configure script defines AM_MAINTAINER_MODE, we should pass --disable-maintainer-mode, to disable the makefile rules that cause autoconf, etc to be rerun. Our run of those tools is explicit, not implicit. Aug 28 16:46:23 03Marcin Juszkiewicz  07org.openembedded.dev * redc60166ad 10openembedded.git/recipes/matchbox-session/matchbox-session.bb: Aug 28 16:46:23 matchbox-session: simple provider of Matchbox session for X11 startup Aug 28 16:46:23 In GPE images we use gpe-session-scripts to provide Xsession init Aug 28 16:46:23 scripts for launching Matchbox components. But what if someone do not Aug 28 16:46:23 want GPE but want Matchbox? Aug 28 16:46:27 Poky has it cleaned in other way. Aug 28 16:46:29 03Marcin Juszkiewicz  07org.openembedded.dev * r862583337c 10openembedded.git/recipes/xorg-driver/xf86-video-vesa_2.2.0.bb: xf86-video-vesa: needs int10 X.org module to work Aug 28 16:46:32 03Marcin Juszkiewicz  07org.openembedded.dev * ra19c9664a8 10openembedded.git/conf/machine/vortex86sx.conf: vortex86sx: use X.org Aug 28 16:46:35 03Marcin Juszkiewicz  07org.openembedded.dev * rb921667e5d 10openembedded.git/recipes/xorg-xserver/ (2 files in 2 dirs): xserver-xorg-conf: provide one for Vortex86sx Aug 28 16:48:38 hrw: nice (re: matchbox-session) Aug 28 16:49:01 kergoth: Poky has it much nicer Aug 28 16:49:10 I cleaned Poky one long time ago Aug 28 16:49:50 this OE one is ugly but works Aug 28 16:52:27 03Marcin Juszkiewicz  07org.openembedded.dev * r02a1ff4e8f 10openembedded.git/recipes/xserver-common/ (2 files in 2 dirs): xserver-common 1.30: -screen option is only for Kdrive Aug 28 16:52:29 03Marcin Juszkiewicz  07org.openembedded.dev * r58367f3a79 10openembedded.git/recipes/xserver-common/xserver-common_1.30.bb: xserver-common 1.30: -screen option is only for Kdrive Aug 28 16:52:47 hrw: ah Aug 28 16:52:58 really need to try to get more of the poky stuff upstream Aug 28 16:53:02 somehow Aug 28 16:53:49 ;) Aug 28 16:54:03 I will probably merge Poky Matchbox stuff in September Aug 28 16:58:22 again anybody know USB port programming here? Aug 28 17:02:16 bye Aug 28 17:30:32 kergoth: re AM_MAINTAINER_MODE yes thats a good idea but sometimes we have to force the regeneration Aug 28 17:31:34 m4t: good to know Aug 28 17:32:05 yea, base-image just finished building Aug 28 17:32:29 m4t: which patch from zecke did you use --disable-tls ? Aug 28 17:32:34 let me find it Aug 28 17:32:39 btw not base-image, minimal-image Aug 28 17:32:51 was it for gcc ? Aug 28 17:34:28 gcc-common.inc and gcc-configure-common.inc Aug 28 17:34:44 yeah the one which disables tls I see Aug 28 17:34:47 hi Aug 28 17:34:54 http://patchwork.openembedded.org/patch/907/ Aug 28 17:35:23 Question: The variables defined in distro are superseded by my variables in local.conf? Aug 28 17:35:24 #uclibc was saying that really, the individual broken packages tls detection should be fixed Aug 28 17:35:35 though with a quick look, i couldn't find it in the *.m4 of gnutls Aug 28 17:37:47 yeah tls.m4 did partial checks Aug 28 17:37:59 so it should be fixed Aug 28 17:38:15 but if we disable tls in gcc iself then evil is nipped in bug Aug 28 17:38:20 s/bug/bud Aug 28 17:39:11 I think I will apply zecke's patch for now Aug 28 17:39:40 yea it seems to do the trick Aug 28 17:40:25 Actually best is to fix the tls.m4 for all affected packages Aug 28 17:40:38 that way we can build uclibc with and without tls/nptl Aug 28 17:44:38 * sushisan feel like at home, nobody listen to me!! sniiif Aug 28 17:45:22 khem: that's the point. we already force the regeneration in autotools.bbclass, with an autoreconf -f Aug 28 17:45:41 khem: AM_MAINTAINER_MOE with --enable-maintainer mode is what sets up the makefile rules so a make might re-run configure if timestamps change Aug 28 17:45:50 implicit, rather than explicit execution of the maintainer tools Aug 28 17:45:57 which is what i want to disable Aug 28 17:46:46 i doubt its an issue in practice, but given our configure/compile split, i don't think its a good idea to let it happen, just in case :) Aug 28 17:47:08 that is true Aug 28 17:47:42 what I'd really like to see happen is to avoid the autoreconf where possible. only rebuild if we've patched configure.ac, or we dep on .m4 files that have changed since our last aclocal run, or whatever.. Aug 28 17:47:49 but that will be a pain in the ass to make happen Aug 28 17:49:07 yeah agree Aug 28 17:50:47 the problem is aclocal. autoreconf is good enough to handle figuring out what needs to be redone (though it might go by timestamps, in which case we'd want to change it to look at an md5sum or something), but aclocal isn't smart enough to check whether a new aclocal.m4 is needed Aug 28 17:51:36 please! Aug 28 17:51:36 hmm Aug 28 17:52:32 sushisan: local.conf should be preferred if there are conditional assignment in other conf files Aug 28 17:52:46 sushisan: kergoth also mentioned _local override Aug 28 17:52:54 I have not used it you may try it Aug 28 17:53:17 yeah, distro and machine loading after local is somewhat annoying, if necessary Aug 28 17:53:35 i think with a smarter parser and an intermediate format we may be able to change that, grab the distro/machine, then move the .conf files up the stack Aug 28 17:53:56 change the .conf layering to not just be inclusions, but true layers Aug 28 17:54:21 kergoth: I think distro and machine should be preferred over local.conf imo Aug 28 17:54:25 can I make a recipe require gcc 3.3? Aug 28 17:54:46 khem: ahh..then if are a definition in distro (with = , not ?=) has precedence the distro definition? Aug 28 17:55:05 sushisan: yes Aug 28 17:55:16 txs! Aug 28 17:55:20 khem: it was never designed that way. local was always intended to let the user override anything, which is why local is where it is in OVERRIDES Aug 28 17:55:31 the original intent was that our metadata would be layered Aug 28 17:55:36 from least specific to most specific Aug 28 17:55:38 hmm Aug 28 17:55:44 with the more specific ones overriding the lesser Aug 28 17:55:57 its conceivable we could want to change that, but that was the idea behind the current implementation Aug 28 17:56:16 * kergoth thinks some of this reasoning should be documented somewhere Aug 28 17:56:18 oneshel: we prefer one compiler I think it wont go well if you compile with different compilers and as old as 3.3 Aug 28 17:56:57 khem: thanks, yeah, that was my understanding, I'm trying to build http://www.gnu.org/software/cgicc/ and it doesn't like gcc 4.3 Aug 28 17:56:59 kergoth: I believe a distro and machine combination defines the right profile Aug 28 17:57:22 it does, but the user should be able to exert control over those decisions for their local build if they choose to Aug 28 17:57:29 otherwise you lose flexibility Aug 28 17:57:29 kergoth: then? I dont understand. local.conf overrides the variables? (or should do that?) Aug 28 17:57:50 sushisan: no, local.conf is parsed before machine and distro, because it *has to be*, because its local.conf that usually sets DISTRO and MACHINE Aug 28 17:58:03 ahhh Aug 28 17:58:05 ok Aug 28 17:58:12 sushisan: most machine/distro use ?= to let you override them, but not all, in which case you can use the _local override to override them Aug 28 17:58:25 theare are some conf that have priority in this sense? Aug 28 17:58:37 kergoth: then local.conf should be parsed last wiping out prior set variable that it wants to override Aug 28 17:58:50 _local is a conf archive? Aug 28 17:59:09 oneshel: I would say fix it beat it or do whatever to compile with current gcc versions Aug 28 17:59:15 conf? Aug 28 17:59:29 what is (or where is) _local Aug 28 17:59:34 variable_ will collapse down to just variable if is in OVERIDES Aug 28 17:59:37 sushisan: no its a override you define Aug 28 17:59:42 'local' is in OVERRIDES, so variable_local -> variable Aug 28 17:59:45 khem: yeah I think that's what I'll end up doing, thanks! Aug 28 17:59:52 this collapsing happens at the end of the parse time, after the distro/machine .conf files are loaded Aug 28 18:00:02 so it overwrites theirs Aug 28 18:01:52 kergoth: oh I was assuming like a single pass parser Aug 28 18:02:17 override expansions, and _append/_prepend are lazy, and occur in the 'finalize' step at the end of the parse Aug 28 18:02:28 kergoth: I think if I have var = value then val_local = value1 seems not to work Aug 28 18:02:36 it does here. Aug 28 18:02:38 every time Aug 28 18:02:56 I must have not looked hard Aug 28 18:03:01 ?= were ok Aug 28 18:03:30 in local.conf: Aug 28 18:03:32 foo = "bar" Aug 28 18:03:35 foo_local = "baz" Aug 28 18:03:40 bitbake -e | grep \^foo= Aug 28 18:03:42 foo="baz" Aug 28 18:03:51 kergoth: If I have PREFERRED_VERSION_uclibc or PREFERRED_VERSION_uclibc_local in local.conf Aug 28 18:04:07 the latter will overwrite the former Aug 28 18:04:31 I think we should update out sample local.conf Aug 28 18:04:40 to emphesize the use of _local override Aug 28 18:05:02 _local is only needed if the distro doesn't use ?=, though. *but*, that said, there are cases where ?= can't be used in that way Aug 28 18:05:22 if bitbake.conf defines a default, and the distro defines a better default, the only way to override the distro is with an _override Aug 28 18:05:32 ?= wont work in distro, cause it is set, in bitbake.conf Aug 28 18:05:58 yes then distro has to override it with = assignment Aug 28 18:06:27 yup Aug 28 18:07:01 most users look at conf/local.conf.sample Aug 28 18:07:17 so if we can add a use case of _local overrride it will be good Aug 28 18:08:10 kergoth: I want to define a variable which will have its value depend upon a give recipe PV Aug 28 18:08:15 how should I do it Aug 28 18:08:22 pn-${PN} is in overrides Aug 28 18:08:28 foo_pn-busybox = "bar" Aug 28 18:08:41 i use that pretty regularly to append to IMAGE_INSTALL for just one image, or what have you Aug 28 18:10:35 I want something if uclibc_ver > 0.9.30 then set var = 1; else set var = 0 and then use var in another recipe Aug 28 18:10:39 is that doable Aug 28 18:11:54 I'd create a .inc, def uclibcver(d): ... check the PV, return the right one, then set PV_pn-uclibc = "${@uclibcver(d)}", and include that .inc in youer local.conf Aug 28 18:11:57 s/youer/your/ Aug 28 18:22:01 ok... I've changed my local conf for myy PREFERRED_VERSION_glib-2 and added _local and works! Aug 28 18:22:43 great Aug 28 18:26:16 there are not explicit reference to this in the OE manual Aug 28 18:28:26 the bitbake manual covers OVERRIDES, i guess the OE one should get more info on it Aug 28 18:31:58 should I discard patches to configure Aug 28 18:32:15 an rely upon reconfigure :) Aug 28 18:39:09 it's never been advised to patch configure directly in autotools builds, always configure.in/configure.ac Aug 28 18:39:19 unless someone changed that recently Aug 28 18:48:44 if I use IMAGE_FSTYPES = "squashfs-lzma" in the krnl are automatically added the support? Aug 28 18:51:49 sushisan: i don't think so. that's an interesting idea though :) Aug 28 19:00:20 http://dl.getdropbox.com/u/112715/Documents/OpenEmbedded%20Tasks.html/index.html Aug 28 19:00:42 extracted all the tasks / action items from my big pile of random OE notes Aug 28 19:17:04 * * OE Bug 5314 has been created by jim.sheldon(AT)gmail.com Aug 28 19:17:06 * * libcgicc 3.2.7 recipe Aug 28 19:17:09 * * http://bugs.openembedded.org/show_bug.cgi?id=5314 Aug 28 19:35:32 re Aug 28 20:29:49 hi woglinde Aug 28 20:29:59 hey woglinde, pb__ Aug 28 20:30:09 g'day kergoth Aug 28 20:31:07 florian: ah, I see that Nils lives! Aug 28 20:32:06 hi pb Aug 28 20:32:07 hms Aug 28 20:32:38 hm python and thread/threading gives me some headaches Aug 28 20:33:24 anyone knows a sane way to stop a thread which has a loop-method which only breaks at keyboardInterrupt Aug 28 20:33:35 pb___: heh yes :) Aug 28 20:34:08 pb___: ah i see... the meeting next week :) Aug 28 20:36:04 pb___: any opinion on http://dl.getdropbox.com/u/112715/Files/0001-Make-BitBake-not-bother-running-tasks-which-are-empt.patch ? Aug 28 20:36:37 kergoth you have an idea? Aug 28 20:36:50 haven't messed with python threads Aug 28 20:37:03 florian: right, he sent me an email to say he was riding into town. Aug 28 20:37:09 *sigh* Aug 28 20:37:20 why always me Aug 28 20:38:10 * florian hopes he offered beer ;) Aug 28 20:38:18 ah beer Aug 28 20:38:20 good hint Aug 28 20:38:28 heh Aug 28 20:38:30 kergoth: hm, dunno. is that really a win? seems harmless enough but it's not obvious to me what it really buys you. Aug 28 20:38:33 I will get mine now from the fridge Aug 28 20:39:03 florian: heh, of course he did. Aug 28 20:39:21 now I wonder how to find the variable names for the site files to avoid the cross compile tests Aug 28 20:39:28 :) Aug 28 20:40:48 pb___: i doubt the processing time is significant, but it seems silly to execute nothing. also, right now, to make a task do nothing, the user has to create a shell task with a single ":" line or a python task with a single "pass" line, afaik, which isn't very intuitive Aug 28 20:41:29 * kergoth isn't sure if it's a win either, just saw it on his list and made a proof of concept Aug 28 20:41:53 kergoth: ah, I see. I didn't know that latter thing. Aug 28 20:42:01 http://dl.getdropbox.com/u/112715/Documents/OpenEmbedded%20Tasks.html/index.html :) Aug 28 20:42:37 if you can't already write "do_configure() { }" without a : line then I agree, something should be done about that and your patch seems as good as anything. Aug 28 20:43:08 kergoth: heh, that's quite a list Aug 28 20:43:13 aye Aug 28 20:44:40 what does "Can probably scrap the whole layout_* split and all" (in the staging section) mean? Aug 28 20:46:56 in that section i was playing with the private staging areas, staging was populated by the installed files from ${D}, so the layout_bindir vs bindir split seemed a bit pointless, when its always STAGING_DIR + ${bindir} Aug 28 20:46:59 iirc, anyway Aug 28 20:47:19 that came from an outline that's been collecting random ideas for a year or 2 now Aug 28 21:06:35 hello. i'm having some trouble with git behind a proxy. are there any pointers? Aug 28 21:07:22 my GIT_PROXY_COMMAND var is ignored if i put it in site.conf, local.conf, or even if i white list it Aug 28 21:08:33 mrc3_: it's a bitbake bug/limitation. there's a hardcoded list of vars to pass along in the runfetchcmd method that the git (and others) fetcher uses to run the git commands Aug 28 21:08:44 right now it includes GIT_PROXY_HOST and GIT_PROXY_PORT, but not GIT_PROXY_COMMAND Aug 28 21:09:00 around line 278 of lib/bb/fetch/__init__.py, if you want to test out adding it to the list manually Aug 28 21:09:03 kergoth, i noticed those two. it also has GIT_CONFIG correctly pointing to my .gitconfig Aug 28 21:09:18 * kergoth nods Aug 28 21:09:30 * kergoth shrugs, haven't tested it personally Aug 28 21:12:00 kergoth, the usefulness rate of the last 5 minutes have topped my week Aug 28 21:12:13 that's sad Aug 28 21:12:19 that worked! i now noticed i had set the wrong GIT_PROXY_COMMAND and GIT_CONFIG Aug 28 21:12:30 hah Aug 28 21:15:00 wow, i never noticed, git shows checking out 20,354 files in a clone of the OpenEmbedded git repo Aug 28 21:15:06 20,354 files... yow Aug 28 21:25:10 evening Aug 28 21:26:03 Hello th1_ Aug 28 21:26:07 for some reason when I do a ./bb myrecipe after the original image it fails to find the cross compiler. it's set the PATH to include /home/xcbuild/builds/active/build/cross/i686core2/bin but the compiler is in /home/xcbuild/builds/active/build/cross/bin Aug 28 21:26:10 i'm using pstaging Aug 28 21:26:26 Dunno about pstaging Aug 28 21:26:42 i'm not sure if it's related or not Aug 28 21:26:53 th1_: yep, thats a pstage bug. Aug 28 21:26:59 there's a patch for it floating around Aug 28 21:27:06 one of which is on the patchwork, i think Aug 28 21:27:23 kergoth: yes, about this...I think I discovered what's happened Aug 28 21:27:26 th1_: http://patchwork.openembedded.org/patch/979/ Aug 28 21:27:38 er, wrong patch Aug 28 21:27:39 hrmph Aug 28 21:27:58 hmm Aug 28 21:27:59 kergoth: http://cgit.openembedded.org/cgit.cgi/openembedded/commit/?id=f876463527bd9ae92a09159fdb4899b4a4d19121 Aug 28 21:28:22 -CROSS_DIR = "${TMPDIR}/cross" Aug 28 21:28:22 +CROSS_DIR = "${TMPDIR}/cross/${BASE_PACKAGE_ARCH} Aug 28 21:28:34 kergoth: Interesting list. Lots of things on there I've thought about too Aug 28 21:28:37 but not in all .bbclasses Aug 28 21:28:42 ant__: ahh, yep Aug 28 21:28:46 ant__: good catch Aug 28 21:29:32 kergoth: Kill all .la files? Aug 28 21:29:45 e.g. external and packaged-staging defines /cross/bin Aug 28 21:30:03 RP: they're not necessary, and then we end up having to mangle them up Aug 28 21:30:07 builds work fine without em Aug 28 21:30:20 kergoth: not on darwin Aug 28 21:30:24 ant__, I just want the CROSS_DIR change from that patch? Aug 28 21:30:40 or I want the whole patch? Aug 28 21:30:48 darwin have a crappy linker? Aug 28 21:31:24 kergoth: yes Aug 28 21:31:36 * kergoth sighs Aug 28 21:32:07 kergoth: Sad but true from my meta-darwin experiments :/ Aug 28 21:32:29 will have to give it more thought, i still think its worth looking into, I'd hate to have to keep all that cruft around for one platform :( Aug 28 21:32:35 ah well Aug 28 21:32:41 * kergoth takes note of the darwin issues Aug 28 21:33:01 th1_: this fixes the immediate issue it seems...but then I had the opkg cannot find package problem :) Aug 28 21:33:02 kergoth: Teach libtool sysroot would be a nice one... Aug 28 21:33:09 i was trying to avoid that Aug 28 21:33:10 :) Aug 28 21:33:13 and I'm not sure it would be that hard Aug 28 21:33:14 ant__, that's almost worse ;) Aug 28 21:33:22 kergoth: understandable Aug 28 21:33:27 I just manually created a symlink in cross for now Aug 28 21:34:16 th1: I'm confident the core-gurus here will fix it soon Aug 28 21:34:20 ;) Aug 28 21:34:40 kergoth: We have --as-needed in poky FWIW Aug 28 21:34:53 RP: we could always start patching everything to use pkg-config ;) Aug 28 21:34:55 ah, nice Aug 28 21:34:59 never had time to test a build with it Aug 28 21:35:14 kergoth: There is a ,inc with a blacklist in Poky Aug 28 21:35:48 kergoth: I keep meaning to pinch moblin's binutils patch for it too Aug 28 21:36:03 whats the patch do? Aug 28 21:36:22 kergoth: Gives you a global environment variable to control it Aug 28 21:36:30 Since libtool messes with the commandline order Aug 28 21:36:30 ah Aug 28 21:36:56 We set the variable in Poky but how much effect it has is debatable Aug 28 21:40:48 A question: I am compiling a program using the oe cross compiler, but not bitbake. I am not using a prebuilt toolchain. On the target it tells me I am missing libstdc++.so.6 . For previous versions of OE I have a note indicating I manually installed the package libstdc++6 and that worked. What should I try? Aug 28 21:41:30 hi Aug 28 21:41:51 there is an error in xterm_207.bb Aug 28 21:41:58 in fedora don't works! Aug 28 21:42:08 archae0pteryx1: uh, exactly what you did before? it isn't going to magically install the shared libs you need for you, unless you built your images to have them already installed, or used OE to install your program into the image with a recipe.. Aug 28 21:42:20 because install see the install mode with a dot Aug 28 21:42:26 archae0pteryx1: Are you using gcc or g++ to build this binary? Aug 28 21:42:40 /home/koan/devel/build/kaeilos/tmp/staging/x86_64-linux/usr/bin/install: invalid mode `755.' Aug 28 21:43:01 repair this is trivial Aug 28 21:43:08 how can report this? Aug 28 21:43:14 good call, forgot about the automatic link with g++, vs gcc (+ -lsupc++)... been ages since I've had to use c++ & qt :) Aug 28 21:43:32 kergoth: I've been there more recently :/ Aug 28 21:43:43 sushisan: the bugzilla or the mailing list? Aug 28 21:45:25 sushisan: Looks like the package itself is broken and needs patching as I can't see that in the recipe Aug 28 21:46:20 RP: apparently both gcc / g++ based on the build file which was given to me Aug 28 21:46:36 kergoth : I manually installed the libraries before Aug 28 21:46:43 archae0pteryx1: If you try and link a c++ lib with gcc you get that error Aug 28 21:47:30 We never used to hit the error due to the way the toolchain worked but we do now Aug 28 21:47:31 RP: I linked with g++ Aug 28 21:48:17 archae0pteryx1: Check your toolchain has libstdc then... Aug 28 21:49:23 he said he got the error on the target, not at link time, though. Aug 28 21:49:32 i think, anyway Aug 28 21:49:34 she Aug 28 21:49:38 RP: cross/armv7a/arm-angstrom-linux-gnueabi/lib/libstdc++.so.6 Aug 28 21:49:40 right, sorry :) Aug 28 21:49:58 RP: tried a urllib2 fetcher once, it too breaks on darwin :\ Aug 28 21:50:00 I can't find it in staging, and I think it should be there Aug 28 21:51:45 archae0pteryx1: As long as its in cross it should work. You're 100% sure its using g++ to link? Aug 28 21:52:05 Well, actually I have something staging/armv7a-angstrom-linux-gnueabi/shlibs/libstdc++.list which contains the line "ibstdc++.so.6" Aug 28 21:52:08 but yes, it should also be on the target and if not, something went badly wrong Aug 28 21:53:05 I'm linking using opt/beagleboard-oe/tmp/cross/armv7a/bin/arm-angstrom-linux-gnueabi-g++ Aug 28 21:53:25 So yeah :P I'm using bb Aug 28 21:53:37 Lucky me Aug 28 21:54:21 I can't recall, are there any command line options to g++ that would screw it up, it's been a while, though this built fine against prev. revs Aug 28 21:55:13 kergoth: grep -R '/cross' -> only ./packaged-staging.bbclass Aug 28 21:55:49 It's just the documentation of external toolchains referring to /cross/bin Aug 28 21:56:05 but I don't see anything in /classes Aug 28 21:56:15 (perhaps my eyes are tired :) Aug 28 21:56:55 I might be linking incorrectly, let me review that Aug 28 21:57:03 ant__: all we have in mvl6 for it is the change in the call to stage-manager in packaged-staging to install to /cross/${BASE_PACKAGE_ARCH} Aug 28 21:57:04 afaict Aug 28 21:57:21 indeed Aug 28 21:57:37 I could not find other fishy occurrences Aug 28 21:58:08 so please, ack between sirs with RP and push ;) Aug 28 21:58:33 i'll ack, patchwork got the patch wrong, but the fix is correct Aug 28 21:58:50 ah, yes, carnivore got it wrong Aug 28 22:00:20 there seems to be parsing issues with patches pasted in the e-mail body Aug 28 22:00:36 and not sent with git Aug 28 22:14:37 libstdc++ falsely detects TLS support for powerpc-uclibc but does ok for arm-uclibc Aug 28 22:14:46 hmm I think I know what to backport Aug 28 22:31:50 undefined reference to `_dl_tls_get_addr_soft@GLIBC_PRIVATE' ... huh. x86 MACHINE must not be quite up to par :) Aug 28 22:34:46 khem the tls macro is false Aug 28 22:35:34 khem consider to read this thread http://www.nabble.com/Improve-TLS-link-test-for-cross-compiling-td24312975.html Aug 28 22:40:18 yes thats one patch I have backported Aug 28 22:40:30 and actually this is not what got committed Aug 28 22:40:38 its slightly modified Aug 28 22:40:45 and then there was a followup to it Aug 28 22:45:23 kergoth: I modified a m4 macro but it seems configure is not regenerated during gcc compile Aug 28 22:45:32 so it does not pick up the changes Aug 28 22:45:40 do_configure regenerates everything if you inherit autotools Aug 28 22:45:44 not compile Aug 28 22:46:47 s/compile/build/ Aug 28 22:47:00 and it inherits autotools Aug 28 22:47:13 in gcc-common.inc Aug 28 22:47:49 but it overrides do_configure(), right? Aug 28 22:48:31 pb___: yes it does Aug 28 22:48:42 ah Aug 28 22:48:59 hi Aug 28 22:49:03 there are sombody? Aug 28 22:50:41 pb___: it calls gnu-configize and oe_runconf manually inside do_configure Aug 28 22:50:59 now I wonder if I start regenerating configure it will be bad Aug 28 22:50:59 khem: can I make a question? Aug 28 22:51:16 sushisan: sure Aug 28 22:51:27 txs Aug 28 22:51:46 I'm trying to compile xf86-video-vesa Aug 28 22:51:51 and I have this error Aug 28 22:51:54 openembedded/recipes/xorg-driver/xf86-video-vesa_2.2.0.bb']' RDEPENDS/RRECOMMENDS or otherwise requires the runtime entity 'xserver-xorg-module-int10' but it wasn't found in any PACKAGE or RPROVIDES variables Aug 28 22:52:00 khem: I suspect it probably will be bad for some versions of gcc, particularly 3.x. Aug 28 22:52:03 xserver-xorg-module-int10 don't exist Aug 28 22:52:11 with recent 4.x I imagine you will probably be ok Aug 28 22:52:22 pb___: yeah thats was on back of my mind too Aug 28 22:52:50 at some point we should stop supporting say 3.3 on .dev Aug 28 22:53:13 I think there are still some kernels that will only build with 3.3.x. Aug 28 22:53:22 khem: do you know where is this dependency? Aug 28 22:53:24 hmm thats right Aug 28 22:53:37 and, of course, there are existing systems which are abi-incompatible with g++ 4.x Aug 28 22:54:03 uff I think I will just generate configure manually and shove it into the patch Aug 28 22:54:06 so I don't think it's likely to be practical to just ditch gcc 3.x any time soon Aug 28 22:54:16 yeah, that is probably the easiest solution Aug 28 22:54:17 kehm yes Aug 28 22:54:29 or may be in gcc-4.x recipe I can override do_configure yet again Aug 28 22:54:38 *g* Aug 28 22:55:04 and twist already twisted gcc recipes a bit more Aug 28 22:55:40 no? Aug 28 22:55:41 hm what dump idea to say threads are bad and your way to fix it use observer Aug 28 23:00:11 sushisan: which branch are u using Aug 28 23:00:46 openembedded.dev Aug 28 23:01:28 one problem with patching configure is that the patch order becomes important and applying this patch to slightly newer version will require re-generating these configure scripts (most likely) Aug 28 23:02:55 sushisan: virtual/xserver should provide it Aug 28 23:05:03 preferred_provider_virtual/xserver = "xserver-xorg" Aug 28 23:05:05 ? Aug 28 23:06:21 yes Aug 28 23:08:39 I have BBMASK = "xserver-kdrive" Aug 28 23:08:52 then compile by default xserver-xorg Aug 28 23:09:04 no? Aug 28 23:10:48 the bb have :RDEPENDS += "xserver-xorg-module-int10 " Aug 28 23:10:49 no idea Aug 28 23:11:28 prolly look at output of bitbake -g and sort out whats getting pulled in Aug 28 23:11:32 this dont force that must to exist the bb? Aug 28 23:13:58 hmm, does createCopy not behave when more than one level deep? Aug 28 23:14:00 would think it would Aug 28 23:14:01 hmm Aug 28 23:14:34 ok Aug 28 23:14:36 txs Aug 28 23:14:41 again! Aug 28 23:15:33 I know I annoying but I'm new in oe! Aug 28 23:15:55 sushisan: np Aug 28 23:16:13 woglinde: how do we create pages in wiki Aug 28 23:16:27 Do I need perms Aug 28 23:18:48 khem hm dont think so Aug 28 23:19:01 never tried it in long time Aug 28 23:22:15 ugh, ast statements changing itself during iteration makes me sad Aug 28 23:22:19 hmm Aug 28 23:23:12 kergoth: terminals ? Aug 28 23:24:16 kergoth: I was wondering if you could do a braindump about bitbake internals on somewhere blog or whereever .. Aug 28 23:24:29 hmm, okay, i'll add it to the todo Aug 28 23:24:43 I am getting interested Aug 28 23:24:53 once upon a time I wrote C/C++ parsers Aug 28 23:25:55 yess now tls detection works correctly for libstdc++ Aug 28 23:26:09 config.log:gcc_cv_have_tls=no Aug 28 23:34:02 03Khem Raj  07org.openembedded.dev * rb88a3344fc 10openembedded.git/recipes/gcc/ (gcc-4.4.1.inc gcc-4.4.1/gcc-improve-tls-macro.patch): (log message trimmed) Aug 28 23:34:02 gcc-4.4.1: Backport improved TLS detection patch in cross environment Aug 28 23:34:02 This patch is required for libstdc++ mainly as far as gcc is Aug 28 23:34:02 concerned. Without this patch configure detects TLS falsely Aug 28 23:34:03 and enables tls when compiling libstdc++ on uclibc based systems Aug 28 23:34:05 uclibc lacks TLS support as of now. With NPTL it will have TLS support Aug 28 23:34:07 but thats a distant future. Aug 28 23:34:12 03Khem Raj  07org.openembedded.dev * rd24436424f 10openembedded.git/: Merge branch 'org.openembedded.dev' of git@git.openembedded.org:openembedded into org.openembedded.dev Aug 29 00:24:17 khem|away: -rw-r--r-- 1 matt matt 7014400 2009-08-28 15:56 Angstrom-base-image-uclibc-ipk-2009.X-test-20090828-dht-walnut.rootfs.tar Aug 29 00:24:23 :-) Aug 29 00:35:26 This may be a silly question, but I am actually supposed to have a package with a name resembling "libstdc++* I also don't remember how to list available recipes Aug 29 00:37:15 it sounds like you're mixing up recipes and packages. Aug 29 00:37:23 a recipe can emit any number of binary packages Aug 29 00:37:29 there is no libstdc++ recipe, there is a libstdc++ package Aug 29 00:37:50 Ok.. how do I look up how to make that Aug 29 00:38:10 grep FILES_.*libstdc++ recipes/*/*.bb Aug 29 00:38:21 or rather, grep PACKAGES.*libstdc++ recipes/*/*.bb Aug 29 00:38:25 either way Aug 29 00:43:27 Ok.. I find that in my ancient version of OE but not from stable/2009 or devel Aug 29 00:47:17 I have an entry in armv7a-angstrom-linux-gnueabi/shlibs but no corresponding library in staging, is that unexpected? Aug 29 00:48:46 Oh well... thanks if anyone gets back to me Aug 29 01:02:38 hi I am not able to bitbake on the latest git release Aug 29 01:02:47 i am getting glib HASH¡¡ error Aug 29 01:02:50 could someone please help me? Aug 29 01:26:48 arg, that didn't work at ALL Aug 29 01:26:51 * kergoth grumbles Aug 29 01:30:31 hmm, zecke must be asleep. i need to keep track of time zones Aug 29 01:33:54 why does everyone but me that hacks on bitbake have to be in europe? Aug 29 01:33:58 it's kind of annoying at times :) Aug 29 02:29:06 \ Aug 29 02:29:11 :/ **** ENDING LOGGING AT Sat Aug 29 02:59:56 2009