**** BEGIN LOGGING AT Wed Jul 07 02:59:56 2010 Jul 07 03:07:03 dcordes: svn: Failed to add directory 'applets': an unversioned directory of the same name already exists Jul 07 03:07:22 dcordes: so somehow your local tree already contains the 'applet' directory and svn does not like that Jul 07 04:17:50 03Graham Gower 07org.openembedded.dev * r9247871d8e 10openembedded.git/recipes/efl1/imlib2_svn.bb: (log message trimmed) Jul 07 04:17:50 imlib2_svn.bb: Don't use host /usr/X11R6/include and /usr/X11R6/lib. Jul 07 04:17:50 Fixes the following problem, where host asm cannot build for target. Jul 07 04:17:50 ccache mipsel-angstrom-linux-gcc -march=mips32 -DHAVE_CONFIG_H -I. -I../.. Jul 07 04:17:50 -I../lib -I../../src/lib -DPACKAGE_DATA_DIR=\"/usr/share/imlib2\" Jul 07 06:58:44 hi there. has anybody tried building oe in a chroot environment? Jul 07 07:01:48 geo_gcju: I'm using minimalistic chroot Jul 07 07:03:13 good. How did you setup it? I'm using deboostrap but after chroot into it and bitbaking I get errors I don't get when build outside chroot Jul 07 07:03:59 show me the errors and I described the setup a bit on ML Jul 07 07:04:23 thanks for your kind help Jul 07 07:04:32 http://www.mail-archive.com/openembedded-devel@lists.openembedded.org/msg03765.html Jul 07 07:04:35 O get an error while building libxml2_2.7.7.bb Jul 07 07:06:43 this is my debug output http://pastebin.com/yg3tR8sy Jul 07 07:07:36 so I suspect sth. is wrong with the generated python modules Jul 07 07:09:48 geo_gcju: are you sure it's chroot issue? Jul 07 07:10:41 well, I've copied the oe setup from my host, there it works w/o a problem, built several images there Jul 07 07:17:22 I didn't mount /proc or any other, just use schroot -c to enter Jul 07 07:19:00 at least /proc is needed in few places (there was report on ML few days ago) Jul 07 07:23:03 hm, ok, that would explain it a bit, nevertheless a hard to debug error Jul 07 07:27:44 so I'm gonna see if I can't work around this. The reason I'm trying to do this Jul 07 07:28:10 morning Jul 07 07:29:49 The reason I'm trying to do this: I have a build server where I just have normal user privileges. So I need to have sth. to be isolated from the host itself, because I can't rely on what packages are already on the machine. Jul 07 07:30:09 hi Jay7 Jul 07 07:33:41 morning Jul 07 07:34:01 * Jay7 agreed that linux' IO scheduler sucks Jul 07 07:48:36 * Jay7 wonders about some letters to angstrom-distro-users ML.. Jul 07 07:48:49 not sure that I've subscribed.. Jul 07 07:53:38 ah.. that was wrong email filters.. Jul 07 08:04:27 please any one help, when install .bb file, then one message is shown Jul 07 08:04:57 message: "the following files were installed but not shipped in any package", what is meaning of this. Jul 07 08:06:00 akumar: "make install" created files, which doesn't belong to any specified package in recipe Jul 07 08:06:22 akumar: so they will be lost, and cannot be installed on target device Jul 07 08:07:31 akumar: in case they are usefull then FILES_${PN} or some other FILES_${PN}-sth should be adjusted or added new subpackage shipping those files Jul 07 08:09:56 JaMa: thanks for reply. Jul 07 08:20:43 Hi guys, I'm getting my head wrapped around what the TI DVSDK is and how it can be installed and deployed on the beagleboard . Any links would be appreciated. Jul 07 08:21:40 I've started by bitbaking the ti recipes like ti-dvsdk-examples.bb Jul 07 08:54:22 khem still have the /ld.so.conf issue (see also my mail to the list), using both oe and bitbake head Jul 07 09:08:38 eFfeM_work : maybe this will help http://permalink.gmane.org/gmane.comp.handhelds.openembedded/34100 Jul 07 09:09:20 zenob: ah ok Jul 07 09:09:42 hi effem Jul 07 09:10:00 hi woglinde Jul 07 09:10:11 zenob: trying that right now Jul 07 09:14:10 zenob, khem changing // to /// helped me, didn't search for the root cause yet Jul 07 09:20:59 03Martin Jansa 07org.openembedded.dev * red60e3c819 10openembedded.git/recipes/xorg-xserver/xserver-xorg_git.bb: Jul 07 09:20:59 xserver-xorg_git: bump SRCREV Jul 07 09:20:59 Signed-off-by: Martin Jansa Jul 07 09:21:00 03Martin Jansa 07org.openembedded.dev * r934edd1e58 10openembedded.git/conf/distro/include/preferred-shr-versions.inc: Jul 07 09:21:00 shr: prefer dbus 1.3.1 Jul 07 09:22:12 eFfeM_work: that will only work for the glibc thing, i got problems with libpam-base-files.bb and there the patch from Khem was the only way out Jul 07 09:22:42 janp, I have that patch applied! verified it by looking at the code Jul 07 09:22:54 fortunately my project does not need libpam Jul 07 09:25:16 janp just saw your latest mail Jul 07 09:25:53 Hi, when I tried to build the ti-dvsdk-demo recipe, it was doing a ti-cgt6x_6.1.9.bb recipe which failed to download from the src Jul 07 09:26:16 screwgoth hm Jul 07 09:26:28 screwgoth thats normaly a case for #beagle Jul 07 09:26:35 but I will answer you Jul 07 09:27:02 you have to download cgt6x manually from the ti side Jul 07 09:27:15 not all dsp is downloadable for free Jul 07 09:27:23 dsp-stuff Jul 07 09:28:06 woglinde: thanks for answering. ok so, I have to manually download... and then change the recipe to point to the local dir ? Jul 07 09:30:03 no Jul 07 09:30:13 put it under your archive-dir Jul 07 09:30:29 and make md5sum cgt.zip > cgt.zip.md5 Jul 07 09:30:41 where cgt.zip is the whole file name Jul 07 09:32:32 got it .. nice ... thanks .... Jul 07 09:35:09 hi folks, Jul 07 09:35:34 i can't seem to clone the oe git repo (git.openembedded.org[0: 140.211.169.165]: errno=Connection refused fatal: unable to connect a socket (Connection refused)) Jul 07 09:35:49 i can access other repos from this machine just fine Jul 07 09:36:09 + there seems to be no response from the http mirror Jul 07 09:38:41 khem: [oe] [PATCH][v2] base.bbclass: re-add backward compatibility with bitbake w.r.t decodeurl fix in bitbake works for me, I've send an ack Jul 07 09:39:28 hm Jul 07 09:45:11 damiannz: both are working here... Jul 07 09:50:41 hms the gcc-dir is as bad as the glibc-dir Jul 07 10:32:13 hi mickey (how do you get that patio into the room ;-) ) Jul 07 10:36:39 btw would love to see some more opinions on why pinning things like gcc in a machine include is good/bad: http://lists.linuxtogo.org/pipermail/openembedded-devel/2010-July/021377.html and descendants Jul 07 10:37:25 my opinion is that machine/arch related toolchain versions are distro thing Jul 07 10:37:52 look how avr32 is handled Jul 07 10:38:22 hrw, why Jul 07 10:38:32 that is why distro not why look at avr Jul 07 10:39:12 yes, agreed, toolchain versions are always a distro policy decision. Jul 07 10:39:15 eFfeM_work: what if there will be two gcc versions for yourarch? who will select proper? maybe something will work better with other then your one? Jul 07 10:39:23 Error in bitbaking ti-dsplink .... in the do_compile() task, I see that when compiling a file called failure.c "/OE/build/tmp/sysroots/beagleboard-angstrom-linux-gnueabi/usr/share/ti/ti-cgt6x-tree/bin/cl6x: No such file or directory" Jul 07 10:39:39 eFfeM_work: I know that distro can override machine config things but still apply Jul 07 10:39:48 hrw, called away will get back to you in 30 mins Jul 07 10:39:51 ok Jul 07 10:40:02 I see that my path only exists till : "/OE/build/tmp/sysroots/beagleboard-angstrom-linux-gnueabi/usr/share/ti/ti-cgt6x-tree/" Jul 07 10:40:29 screwgoth hm Jul 07 10:42:25 woglinde: I am actually trying to build gstreamer-ti , but since it has a dependancy for ti-dsplink , I'm stuck .... Jul 07 10:42:51 eFfeM_work: hi :) Jul 07 10:43:01 mickey|patio: elo Jul 07 10:43:15 screwgoth I know Jul 07 10:43:26 I build gstreamer-ti serval times Jul 07 10:43:50 but there some patches latetly Jul 07 10:44:08 maybee koen didnt update his sane-file for it Jul 07 10:44:22 woglinde: Hmm... are you suggesting that I do a git pull Jul 07 10:51:57 18:32 <@prpplague> Magdalena: that will get you the basic 1024x768 mode that is common to most displays that support vesa/dvi Jul 07 10:52:27 ops Jul 07 10:58:35 woglinde: any hints on where I can find the patches to make ti-dsplink ? Jul 07 10:58:48 morning all Jul 07 10:59:43 pb_: If you have OVERRIDES = "a:b:c" and VAR_a , VAR_b and VAR_c, which should getVar(VAR) return? Jul 07 11:00:18 It turns out I'm able to bitbake almost all the ti recipes(ti-linuxutils, ti-edma3lld , etc), except ti-dsplink. Damn !!! Jul 07 11:01:37 RP: VAR_c, I think. It's documented in the manual somewhere if I remember right. Jul 07 11:02:07 ah, no, it's in bitbake.conf. ${OVERRIDES} is processed left to right, the later ones take precedence. Jul 07 11:02:17 screwgoth search through the latest patches in oe.dev Jul 07 11:03:30 screwgoth1 and look at the arago-project Jul 07 11:03:54 wokay ... check it out now Jul 07 11:03:57 pb_: Poky's documentation in bitbake.conf does not quite read the same :/ Jul 07 11:05:02 ah. what does poky say? Jul 07 11:06:04 pb_: "And finally '_local' overrides anything" Jul 07 11:06:16 which implied to me right to left Jul 07 11:07:18 hmm, bye pb_ :/ Jul 07 11:09:35 re (almost forgot lunch) Jul 07 11:09:56 pb_: if it were the other way it would also make "_local" all powerful which is kind of what you'd expect Jul 07 11:10:03 pb_: welcome back btw Jul 07 11:10:25 * pb_ stabs freenode Jul 07 11:10:34 hrw, pb_, all: issue is that i see no way to avoid that a distro selects a version of say gcc that is not supported by that machine Jul 07 11:11:00 eFfeM_work: well, yes, that's just too bad. it's up to the distro to pick a version or versions that works for the machines it wants to support. Jul 07 11:11:08 would be if DEFAULT_PREFERENCE would win and you could simply say that way a certain recipe is not supported Jul 07 11:11:08 woglinde: The arago-project seems to have similar .bb files. The issues seems to be that there is no cl6x compiler in my sysroot path .. or anywhere else for that matter Jul 07 11:11:10 pb_: I think someone synced up the documentation with the reality at some point Jul 07 11:11:14 eFfeM_work: there is conf/distro/include/sane-toolchain.inc which is used by some distros + there is angstrom.inc Jul 07 11:11:27 eFfeM_work: submit patches against both and distros will be covered Jul 07 11:11:29 pb_: which means I should go and work out why bitbake is doing the opposite to what I'd expect :/ Jul 07 11:11:50 screwgoth1 hm rebuild the recpie manually Jul 07 11:11:54 hrw, i know these exist. not sure if everyone uses them Jul 07 11:12:06 maybee do_stage is going wrong Jul 07 11:12:17 hrw and wrt the two toolchains: as long as the machine sticks to a certain toolchain I'd say it is ok Jul 07 11:12:24 RP, pb_: I've synced it after IIRC pb's explanation and confirmation that it works as described Jul 07 11:12:31 RP: yeah, I guess the issue of whether "local" should be highest or lowest priority is sort of an open question. you can imagine having valid use cases for both ways round. Jul 07 11:12:43 eFfeM_work: I would rather say "as long as the distro sticks to a certain toolchain machine is fine" Jul 07 11:12:51 http://git.openembedded.org/cgit.cgi/openembedded/commit/?id=70af4a882915cb51302734ca0de3b810f05df072 Jul 07 11:13:01 woglinde: The question is , does the ti-dsplink do_stage() take care of installing the cl6x compiler or should it be done by some other package which was not included as a dependancy in ti-dsplink Jul 07 11:14:11 hrw: understand what you are saying but for me toolchain is much more coupled to machine as to distro Jul 07 11:14:34 RP: as to why bitbake does what it does, I haven't looked at the overrides logic for a while but it used to literally work by stepping across ${OVERRIDES} left to right, folding VAR_xx into VAR for each member. so, if you had multiple matching overrides, the value you would be left with would be the last matching one. Jul 07 11:15:05 it could well be different now but I hope it still behaves the same. if the ordering has gotten changed then that would be a little bit unfortunate. Jul 07 11:15:20 screwgoth *sigh* because cl6x-compiler has its own recipe, guess which do_stage is maybee corruppted Jul 07 11:16:37 pb I am working at the arm-egabi atomic patches for gcc-4.3.x Jul 07 11:16:52 woglinde: very good Jul 07 11:17:10 but first I am fixig denys gcc-breakage Jul 07 11:17:16 righto Jul 07 11:17:34 *sigh* whole gcc-dirs needs cleanup Jul 07 11:17:53 yes Jul 07 11:19:48 pb_: I just looked at the code and it does that Jul 07 11:20:16 pb_: It means something odd is going on in my case... Jul 07 11:20:28 pb_: Possibly an unfortunate combination of variables Jul 07 11:20:44 It does mean my RDEPENDS for a package disappear which is very annoying Jul 07 11:24:49 jo raster Jul 07 11:24:56 boop Jul 07 11:25:38 woglinde: I see the cgt6x recipes look innocent enough. And even if it has screwed up the staging phase, the cl6x should have been installed _somewhere_ ...... _anywhere_ ..... Jul 07 11:26:16 I see that instead of share/ti/ti-cgt6x-tree/bin/cl6x , everywhere I see share/ti/ti-cgt6x-tree/patches , which , btw, is an empty dir Jul 07 11:37:41 pb_: Imagine you have a variable called VAR_c_b and a VAR_c. Which one would you expect to see? Jul 07 11:38:07 screwgoth run bitbbake -c build footarget Jul 07 11:38:16 and look under the work-dir Jul 07 11:38:29 there it shows where all stuff will be installed Jul 07 11:38:49 RP: VAR_c_b Jul 07 11:39:22 pb_: right. and in this case the code wants VAR_c :/ Jul 07 11:39:30 I wouldn't be at all surprised if bitbake does something unexpected in that case, though. Jul 07 11:40:00 pb_: You get VAR_c_b but VAR_c is what I actually care about :/ Jul 07 11:40:13 ah. Jul 07 11:40:29 pb_: Think of the pkg override in package.bbclass Jul 07 11:40:50 pb_: We dump load of data into variables and we really want those values back... Jul 07 11:41:10 * RP wonders if that code should just do "OVERRIDES = pkg" Jul 07 11:41:42 pb_: I mean package_*.bbclass Jul 07 11:41:51 The actual package writing functions Jul 07 11:42:06 yeah, very likely. or stop leaning on overrides at all and do all the subpackage handling explicitly. Jul 07 11:44:30 the trouble is we'll have to check X_pkg and X in each case Jul 07 11:44:39 or rewrite the insane output code... Jul 07 11:46:14 yeah Jul 07 11:46:52 I think it was done like this to make the files more readable - only write data that exists and isn't duplicate Jul 07 11:47:33 yeah, I think that's right. I suspect nowadays though that it might be having the opposite effect. Jul 07 11:49:14 All the other overrides would have already been expanded anyway, Just using pkg in OVERRIDES should therefore be safe Jul 07 11:59:14 woglinde: HEre's what I did : I manually ran the downloaded cgt6x binary which installs the compiler and pointed it to the path required in the error. Jul 07 11:59:54 * Jay7 have sent proposal for talk about OE and kexecboot on Chaos Construction 2010 in St.Petersburg, Russia Jul 07 12:00:10 will try to meet lumag there :) Jul 07 12:00:13 woglinde: I know, it's a hack ... I'll try to fix the reicpe if possible.... but alteast the ti-dsplink recipe was successfully built after that Jul 07 12:05:22 Hi! As I see in OE wiki (http://wiki.openembedded.net/index.php/Branching_Policy) it is possible to request personal user repository. Whom can I ask about this issue? Jul 07 12:05:59 gim hm Jul 07 12:06:43 Hm, I noticed words "Any OE Developer" and I'm not one of them Jul 07 12:07:20 wasnt decided yet Jul 07 12:07:35 there was plans for doing something like this Jul 07 12:07:39 afaik, you should have write access to .dev atm Jul 07 12:07:57 but seems it was a little bit forgotten Jul 07 12:08:21 from tech-side your ssh public key should be on git server Jul 07 12:08:52 gim you could use github Jul 07 12:08:56 for now Jul 07 12:09:02 +1 Jul 07 12:09:14 woglinde: I think TSC should consider this :) Jul 07 12:09:24 jay7 no Jul 07 12:09:33 this is more a task for the e.V. Jul 07 12:09:50 its a strcutual problem Jul 07 12:10:06 someone should note this in ML imho Jul 07 12:10:20 gim: can you ask this on oe-devel ML? Jul 07 12:10:46 jay7 look at the minutes from last oedem Jul 07 12:10:50 there it was discussed Jul 07 12:12:16 Jay7: ok, I'll ask about it Jul 07 12:17:00 By the way, where can I find history of this IRC channel? Jul 07 12:18:27 ~logs Jul 07 12:18:28 All conversations are logged to http://ibot.rikers.org/channel, where "channel" is replaced by the URL-encoded channel name, such as %23freenode for #freenode. Lines starting with spaces are not logged. logs are updated daily Jul 07 12:18:46 ok, thanks Jul 07 12:18:48 what's the way to get commit access to oe? I got 3 votes in May already (http://lists.linuxtogo.org/pipermail/openembedded-devel/2010-May/020250.html) but there did not happen anything since then Jul 07 12:20:05 ensc|w send your key to someone who can add it. not sure who can, guess mickey Jul 07 12:20:35 ensc|w: http://lists.linuxtogo.org/pipermail/openembedded-devel/2010-May/019949.html Jul 07 12:20:39 ensc|w didnt you answer mickeyl to send your ssh-key? Jul 07 12:20:58 thanks jama Jul 07 12:21:09 ensc|w: this is not from "your" thread, but the same apply for you Jul 07 12:21:51 ah ok... Jul 07 12:33:55 What does it mean when I get the following error while installing a bitbakes ipk package : "opkg_install_pkg: Package ti-dspbios md5sum mismatch. Either the opkg or the package index are corrupt. Try 'opkg update'." Jul 07 12:34:40 the checksum does not match, you might want to remove ti-dspbios* from your downloads dir Jul 07 12:34:54 or update the package-index Jul 07 12:36:05 ofc, did overlook this was about installing not building Jul 07 12:41:20 well, I did opkg update, but the error persists ... Jul 07 12:41:46 and this is not specific to ti-dspbios .... I have encountered this before for other applications as well ... Jul 07 12:42:41 and the same package , say ti-dspbios in this case, when I try to install it from the repos on the net via opkg, "opkg install ti-dspbios" , it installs fine Jul 07 12:44:28 screwgoth1: are you trying to install a local package? if so I guess you did not bump PR while you changed the package (so the checksum of your ipk file does not match what is in the info you get from the feed. Jul 07 12:46:17 hi, i'm still have problem with Permission denied: '/arm' in libart-lgpl, any idea how to fix this one ? Jul 07 12:47:45 btw has there been a TSC meeting in july ? Jul 07 12:47:54 eFfeM_work: do you also have such a problem ? Jul 07 12:48:34 zenob, i had a problem with /ld.so.conf, this was fixed by applying the patch from khem (the one mentioned by janp) Jul 07 12:48:48 i'm at head for both bitbake and oe Jul 07 12:53:36 Hi all. I'm totally new to OE and i'm trying to add support for colibri in it. Untill now i succeed to build an image using : "bitbake helloworld-image" and now i'm trying to build the kernel "bitbake virtual/kernel" But it fail (see log http://pastebin.com/tDJFHeJW) Jul 07 12:55:22 I'm using DISTRO = "angstrom-2008.1" and i've added a colibri.conf file and edited recipes/linux/linux_2.6.34.bb to define the default associated kernel version with colibri machine. Jul 07 12:55:52 But now i have now idea where to look to fix this error Jul 07 12:56:19 kerloi: not sure what you did in your conf file, but I am very surprised you are buidling the linux-nios2 kernel Jul 07 12:56:36 ah, here it is: Jul 07 12:56:53 NOTE: multiple providers are available for virtual/kernel (linux-nios2, linux, linux-bast, linux-palmpre, linux-rt, linux-omap-2.6, gumstix-kernel); Jul 07 12:56:53 NOTE: consider defining PREFERRED_PROVIDER_virtual/kernel Jul 07 12:56:53 Jul 07 12:56:53 Jul 07 12:56:53 Jul 07 12:56:54 Jul 07 12:56:55 oops Jul 07 12:57:10 eFfeM_work: well basically i want to use linux linux_2.6.34 kernel with a patch Jul 07 12:57:28 eFfeM_work: Thanks .. I'll check that out .... But as far as I remember ... I have not changed any recipe. Would it count if I just saved it with no extra changes (maybe a newline or two) Jul 07 12:57:36 in your colibri.conf file add a PREFERRED_PROVIDER_virtual/kernel line specifying which of the kernels you want Jul 07 12:58:12 screwgoth1: if you compiled by yourself there might be differences due to timestamps etc Jul 07 12:58:33 kerloi you probably want the linux kernel (and definitely not the linux-nios2 one) Jul 07 12:58:52 eFfeM_work: Yeah... I compiled myself .... but then how do I fix this ...clean the packages and bitbake them again ? Jul 07 12:59:25 eFfeM_work: Like PREFERRED_PROVIDER_virtual/kernel="linux_2.6.34" Jul 07 12:59:26 screwgoth1: how do you install? opkg install xx.ipk ? Jul 07 12:59:31 yup Jul 07 12:59:55 screwgoth1: never had problems with that as far as I can tell, so can't help you with this (apart from suggesting to bump PR) Jul 07 13:00:09 kerloi probably without the _2.6.34 part Jul 07 13:00:23 eFfeM_work: ok Jul 07 13:00:56 kerloi, have a peek at some of the other machine files Jul 07 13:02:07 kerloi you haven't really picked the simplest task to start with oe Jul 07 13:02:24 eFfeM_work: that's what i did but i've deleted this line (stupid me :) ) Jul 07 13:02:32 np Jul 07 13:02:46 eFfeM_work: I known but i don't have the choice ... Jul 07 13:02:59 :-) Jul 07 13:07:31 pb_, hrw, all: wrt our earlier discussion: do we consider it as ok, that a distro can select a version of a recipe that is explicitly given a DEFAULT_PREFERENCE_arch = "-1". E.g. if say nios2 does not support gcc 4.5.0 and has DEFAULT_PREFERENCE_nios2 = "-1" in the recipe should the distro then still be able to override this ? Jul 07 13:09:29 eFfeM_work: yes. if it then doesn't work, this is the distro's own problem. Jul 07 13:09:47 agree with pb Jul 07 13:10:35 ok, understood (although I fear that people will start complaining to me, not to the distro) Jul 07 13:14:16 that's a bit of a risk, but you can tell them where to go. or forestall the problem by telling them in advance which distros will work with your machine. Jul 07 13:14:56 personally I'd say that a generic pinning of a distro should not override a DEFAULT_PREFERENCE_arch but a distro pinning like PREFERRED_VERSION_package_arch should override a DEFAULT_PREFERENCE_arch Jul 07 13:15:21 pb_ there is no real way to tell them, apart from saying so every time someone complains Jul 07 13:15:26 DEFAULT_PREFERENCE is just that, the default preference. it has always been the intent that it should only set an actual preference if nothing else does. Jul 07 13:15:58 if DEFAULT_PREFERENCE is going to start trumping an explicitly set PREFERRED_VERSION then that would be quite a significant change (and not, in my view, a desirable one) Jul 07 13:16:47 * eFfeM_work would like to be able to output some message to the user when he builds for his machine Jul 07 13:17:46 pb_ I agree with that, but there is simply no way at the moment by which a package can say "i do not support this machine/board Jul 07 13:19:04 other than COMPATIBLE_MACHINE, you mean? Jul 07 13:20:02 pb_ I did not consider COMPATIBLE_MACHINE, haven't really used it, and was mostly unaware of it (but I must say I've seen it somewhere) Jul 07 13:20:16 Hum i have another question. To indicate that the default kernel version for colibri machine i've added DEFAULT_PREFERENCE_colibri = "1" in the linux_2.6.34.bb file. But it have built a 2.6.33 ... Jul 07 13:20:42 Should i add something else elswhere to tell OE to build a 2.6.34 ? Jul 07 13:21:51 pb_ actually COMPATIBLE_MACHINE might be what I really wanted, but as I see it, I cannot invert it (e.g. say COMPATIBLE_MACHINE = !nios2 (or ~nios2), or am I missing something? Jul 07 13:22:19 I'd be happy to add such a line to e.g. the gcc versions I cannot support Jul 07 13:24:13 alternative could also be create recipes like gcc-nios2 and override virtual/gcc, but gcc is already quite crowded, and that would obfuscate thigns even more Jul 07 13:31:05 eFfeM_work: COMPATIBLE_MACHINE = "(?!nios2)"? Jul 07 13:31:23 not that I've actually tried, but I think that ought to work. it's just a regex. Jul 07 13:31:56 pb_ does that work? brilliant! was just digging into the way a regex is specified Jul 07 13:32:28 pb_ do you feel it is a good solution to add such a line (provided it works) to the gcc recipes that do not work for nios2 ? Jul 07 13:33:08 I think that would be fine, yes. I do still feel that you are making rather heavy weather of an issue that doesn't seem all that important to me, but I certainly don't have an objection to you adding those lines. Jul 07 13:34:28 03Sebastian Krzyszkowiak 07org.openembedded.dev * r3c4955f19a 10openembedded.git/recipes/freesmartphone/ (2 files in 2 dirs): Jul 07 13:34:28 frameworkd: add patch for SHR to workaround bug in kernel resulting in weak vibrator strength Jul 07 13:34:28 Signed-off-by: Martin Jansa Jul 07 13:34:46 well I've no control over what a distro tries to do with my machine, but I fear that if it does not work because the distro didn't do its thing properly I'll be getting the heat Jul 07 13:35:17 eFfeM_work: and you can fix all OE official distros with one changeset to conf/distro/ Jul 07 13:35:49 true, but that is generically true for machine maintainers and I haven't heard of it actually being a real problem in practice. Jul 07 13:35:56 eFfeM_work: out-of-OE ones can complain but you can say them "look at commit HASHID and apply logic" Jul 07 13:36:13 unless you are anticipating a sudden flood of nios2 users running random distros then I would not worry too much. Jul 07 13:36:32 actually I don't understand why koen is making such a fuzz about me pinning a few recipes in my machine conf for a machine that he does not even have and that he has not really shown interest in adding to his distro Jul 07 13:36:49 * mwester looks at pb_'s COMPATIBLE_MACHINE assignment, and his brain boggles. Jul 07 13:36:54 eFfeM_work: angstrom tries to support all machines in OE Jul 07 13:37:07 eFfeM_work: and toolchain selection is distro thing Jul 07 13:37:29 eFfeM_work: those 2 rules are quite old and we got used to them Jul 07 13:38:13 eFfeM_work: when angstrom will get *_nios2 overrides then you can ask koen to add nios2 machines to list of officially built ones so your nios2 users will get packages for free Jul 07 13:38:26 well I can't guarantee if a distro works with nios2 so I feel reluctant to provide a patch for them Jul 07 13:38:56 eFfeM_work: so instead you prefer to waaste hours on complaining why your way got rejected? Jul 07 13:38:59 eFfeM_work: what distro are you planning to use, yourself? Jul 07 13:39:10 hrw as nios2 is a softcore it is quite possible users will have dedicated machines anyway Jul 07 13:39:20 doing angstrom test build takes few hours and happen during night. same for other distros Jul 07 13:39:59 hi hrw, pb_ Jul 07 13:40:03 hi kgilmer Jul 07 13:40:03 "for distro in angstrom micro minimal;do DISTRO=$distro MACHINE=mynios2 TMPDIR=$PWD/tmp-$distro bitbake console-image;done Jul 07 13:40:07 hi Ken Jul 07 13:40:19 pb_: we've used different distro's in the past, currently we are using a stripped down angstrom image with only those packages validated that we need and with some patched versions Jul 07 13:40:39 console-image is probably already more than we need and I guess some things in it won't work Jul 07 13:40:55 eFfeM_work: s/console-image/my-own-nios2-compatible-image/g Jul 07 13:40:55 but just for fun will give it a try Jul 07 13:41:29 eFfeM_work: first patch distro configs and drop machine part - will give you extra bonus points if things will work Jul 07 13:41:34 hrw, my own image works Jul 07 13:42:01 eFfeM_work: with your rejected machine config changes... Jul 07 13:43:15 actually just tried console-image and hit ^c after a few packages, this drags in things like xproto, and libxua and xtrans, etc etc Jul 07 13:43:22 * eFfeM_work does not understand why these should be in console-image Jul 07 13:43:30 dbus? Jul 07 13:43:36 bluez? Jul 07 13:43:44 * mwester has been fighting dbus and bluez pulling in x11 Jul 07 13:43:49 eFfeM_work: build deps != image contents Jul 07 13:44:14 hrw, true Jul 07 13:44:21 I send the gcc-arm-atomic patches now out Jul 07 13:44:37 But it still feels wrong to build x11 even if it never gets installed. Jul 07 13:44:52 mwester: I fully agree Jul 07 13:46:30 mwester *g* koen's answer is disks and cpu's are cheap Jul 07 13:46:44 btw, I do not consider my machine config changes as rejected, but if they are I'll happily withdraw all my nios2 changes as they are pointless without a machine for them Jul 07 13:47:22 woglinde, that is true, but developer's time is not cheap! Jul 07 13:47:38 mwester depends Jul 07 13:47:43 eFfeM_work: do you have a problem working with us, a lot of your statements boil down to "do it my way or I take my toys home" Jul 07 13:48:10 re raster Jul 07 13:48:27 mwester: did you have any luck with a non-x11 DISTRO_FEATURE? Jul 07 13:48:46 XorA: i have no problem with working with most of you, but I have some problems with the attitude and responses of some of you Jul 07 13:49:02 ~logs Jul 07 13:49:03 All conversations are logged to http://ibot.rikers.org/channel, where "channel" is replaced by the URL-encoded channel name, such as %23freenode for #freenode. Lines starting with spaces are not logged. logs are updated daily Jul 07 13:49:25 XorA: I just want to contribute my work, but if it is not appreciated, I can spent my time otherwise Jul 07 13:49:27 pb_, I have limited success -- dbus without x11 is easy, but I am not sure how to approach the bluez stuff yet. Jul 07 13:49:31 and yes, I know the channel is logged Jul 07 13:49:44 eFfeM_work: thats the sort of dump and run attitude google and apple use for opensource Jul 07 13:50:11 mwester: I think bluez might be a slightly different problem. what extraneous dependencies are you seeing there, exactly? Jul 07 13:50:24 gstreamer and maybe pulseaudio Jul 07 13:50:45 eFfeM_work: you got suggestions Jul 07 13:50:53 hrw: in bluez-libs, or just in the apps? Jul 07 13:50:56 eFfeM_work: that wasnt aimed at you I needed the logs Jul 07 13:51:10 bluez-sound libs needs gstreamer stuff Jul 07 13:51:29 thats the root Jul 07 13:51:35 yep Jul 07 13:51:39 XorA: no, I am not dumping it here, I am fully prepared to support what I made, but I need no problems on my plate that are outside my control Jul 07 13:51:49 ah right, and that's in bluez-libs? in that case I guess it should probably be split into a separate recipe Jul 07 13:53:00 Bluez-libs is only missing a dependency on dbus - it's ok. But what I am struggling with is my ignorance on bluez -- is bluez-libs useful without any of the rest of bluez? Jul 07 13:53:35 eFfeM_work: then I suggest you listen to some of OEs top developers like hrw Jul 07 13:53:38 mwester: maybe the solution is to have an audio-less bluez variant Jul 07 13:53:48 sort of. the reason bluez-libs is "special" is that it's the piece which other packages need to link against and hence build-depend on. Jul 07 13:54:21 the other apps (the "rest" of bluez) are more of a pick-and-mix thing: you can install some, none or all of them and no other packages really need to care. Jul 07 13:55:19 it probably is true that a system which only included bluez-libs and none of the rest of bluez would not have a huge amount of useful bluetooth functionality as it stood, but it would then be in a "bluetooth ready" state where you could just install a few more packages and everything would be good to go. Jul 07 13:58:33 bluez-nox11-*.bb :) Jul 07 13:58:49 bletch Jul 07 13:59:33 woglinde: have you actually tested all four of the gcc versions that you patched? Jul 07 14:00:38 pb as robert stated out the patch is used by ubuntu for a long time Jul 07 14:00:50 sure, that's not quite what I meant. Jul 07 14:01:02 the second helps at linking time Jul 07 14:01:04 did you verify that the patch does apply cleanly and all four recipes still compile afterwards? Jul 07 14:01:17 yes I did patch test Jul 07 14:01:30 I am not denys Jul 07 14:01:36 ok, cool, just checking Jul 07 14:01:43 as you noticed, not everybody is so diligent :-} Jul 07 14:01:52 I fixed denys stuff any Jul 07 14:02:04 but robert didnt checked it in yet Jul 07 14:02:19 didnt have my ssh-key at work Jul 07 14:03:34 very good Jul 07 14:07:07 Jay7, that's actually what I have created for dbus, and have been working on for bluez. But I agree that a collection of duplicate recipes is "blech". :) I just don't have the "foo" to create FEATURES stuff (yet). Jul 07 14:07:13 hi pb Jul 07 14:07:25 hi blindman Jul 07 14:07:35 hi woglinde, all set for the evening? :) Jul 07 14:07:52 blindman sure Jul 07 14:08:11 I hope spain is playing like agains swiss Jul 07 14:08:11 haha Jul 07 14:08:35 haha, we will see, i hope loew's tactic play out as nicely as against england and argentina :) Jul 07 14:08:38 mwester: yeah.. that is about USE flags RFC.. Jul 07 14:09:07 blindman sure Jul 07 14:09:21 let me fetch this very intressting article for your Jul 07 14:10:32 sorry german only Jul 07 14:10:36 http://www.heise.de/tr/artikel/Elf-Dateien-muesst-Ihr-sein-1017763.html Jul 07 14:12:47 woglinde: i had an interesting conversation on #debian.de with yath who works for a corpy which records all the football games to the end, that customers come and want certain archive footage of certain player, like very special things, lahm with ever ypass that ends in a goal or whatever Jul 07 14:12:56 woglinde: used by scouts, etc. Jul 07 14:13:05 uh Jul 07 14:13:09 cool Jul 07 14:13:12 woglinde: and they also get lots of statistical data, more than fifa.com itself can provide Jul 07 14:13:21 then its maybee the firm mentioned there Jul 07 14:13:28 woglinde: put together by some sweatshop somehwere Jul 07 14:13:44 woglinde: but where you have very exrtensive statistical data about every single player Jul 07 14:14:00 woglinde: haven't gone through the whole article yet, just wanted to note that :) Jul 07 14:14:05 sure Jul 07 14:14:19 they analyize the whole match Jul 07 14:14:26 each tackle Jul 07 14:14:31 each shot Jul 07 14:14:41 even postions from the referee Jul 07 14:16:32 quite fascination, actually Jul 07 14:16:50 but the algorithm arnet perfect yet Jul 07 14:17:26 so they have to outsource monkeywork to spain and otherlands for tacking players Jul 07 14:19:00 hi blindman Jul 07 14:20:15 woglinde: yeah sure, but that's prolly just a matter of time Jul 07 14:20:21 yes Jul 07 14:20:34 woglinde: like everything in IT ... getting rid of oneself :) Jul 07 14:25:20 for the record, restarted building console-image for nios2, it fails in libX11: | os2Stubs.c:48:17: error: os2.h: No such file or directory :-( Jul 07 14:27:31 khem: I see you were ahead of me with the base.bbclass fix, thanks Jul 07 14:30:50 woglinde: nice article Jul 07 14:33:44 indeed Jul 07 14:57:42 hi stefan Jul 07 14:58:24 hi woglinde Jul 07 14:58:36 woglinde: thanks for takign ove rthe gcc patches, very much apreciated Jul 07 14:58:47 woglinde: I'm testing them now Jul 07 14:59:05 normal update and then build, build from scratch as well as llvm and pulseaudio Jul 07 15:00:39