**** BEGIN LOGGING AT Wed Oct 06 02:59:57 2010 Oct 06 03:08:00 kergoth_, what's a snippet? Oct 06 03:08:05 ${@} Oct 06 03:08:10 inline python Oct 06 03:08:13 is what i'm referring to Oct 06 03:08:16 ah. cool Oct 06 03:08:24 so you owuldn't have to use getVar() in them Oct 06 03:08:25 or ${} Oct 06 03:08:30 but just directly access the vars Oct 06 03:08:48 PF = "${@PN + "-" + PV}" Oct 06 03:08:52 or what have you Oct 06 03:09:21 that would certainly make things more readable Oct 06 03:09:26 yeah, thats my thought Oct 06 03:09:30 reduce the ugliness Oct 06 03:09:36 bit more straightforward Oct 06 03:09:48 only works in those, not python tasks, which i think is best, those can use getVar() easily enough Oct 06 03:10:47 http://github.com/kergoth/OE-Signatures/commit/49cf79bb545c731147ae552967484aa133070d6f shows the implementation in the signature code, the implementation in bitbake itself would be no more difficult Oct 06 03:43:44 huh Oct 06 03:43:51 implemented it in bitbake and started using it in bitbake.conf Oct 06 03:43:52 works great Oct 06 03:45:58 the only downside i see is that it will make the dev branch depend on bitbake head Oct 06 03:46:38 but everyone syncs their oe and bitbake trees simultaneously, right? Oct 06 03:56:30 oh, i wouldn't change OE to use it until it bumps its minimum bitbake version to one that includes the feature Oct 06 03:56:44 we haven't even bumped the minimum to 1.10 yet, and this would be going into whatever comes after 1.10 (1.12, likely) Oct 06 03:56:48 :) Oct 06 03:57:05 but thats true for every change to bitbake Oct 06 04:00:47 i think we need to have a new stable branch, so that this sort of thing can be merged into the dev branch sooner Oct 06 04:01:09 agreed Oct 06 04:01:17 i also think bitbake needs a proper release cycle Oct 06 04:01:30 right now its whenever we feel like it Oct 06 04:01:39 things are currently looking quite stable, so it might be a good time to propose such a thing Oct 06 04:01:54 * kergoth_ really likes cbrake's testing stuff Oct 06 04:02:07 yes, its worked marvelously Oct 06 04:03:10 03Chris Larson  07master * r8d661ce0c3 10bitbake.git/lib/bb/data_smart.py: Oct 06 04:03:10 Fix __getitem__ for DataSmart Oct 06 04:03:10 Ensure it raises KeyError for a missing key, this is required to use this as a Oct 06 04:03:10 mapping in various places, e.g. as locals in an eval. Oct 06 04:03:10 Signed-off-by: Chris Larson Oct 06 04:03:15 03Khem Raj  07master * rf264cb6d43 10bitbake.git/lib/bb/fetch/ (bzr.py cvs.py git.py hg.py svn.py): Oct 06 04:03:15 fetchers: Use tar --exclude pattern to remove SCM files Oct 06 04:03:15 This option will exclude the SCM metadata from tar files. Oct 06 04:03:15 Tested with gcc where svn tar which used to be 156M for gcc 4.5 Oct 06 04:03:16 is now 77M Oct 06 04:03:16 Signed-off-by: Khem Raj Oct 06 04:03:17 03Chris Larson  07master * r606fa1fd97 10bitbake.git/lib/bb/data_smart.py: Oct 06 04:03:17 Access metadata vars as locals in python snippets Oct 06 04:03:18 Example: Oct 06 04:03:18 FOO = "bar" Oct 06 04:03:19 BAR = "${@FOO + '/baz'}" Oct 06 04:03:19 ${BAR} == "bar/baz" Oct 06 04:03:20 Signed-off-by: Chris Larson Oct 06 04:05:24 kergoth_: Cool Oct 06 04:05:32 sorry for the delay :) Oct 06 04:05:41 kergoth_: how would I use this new feature for my purpose Oct 06 04:06:11 we can't use it until the version minimum is bumped, of course, but in your case you'd just use the var name directly Oct 06 04:06:25 ${@["foo", "bar"][LIBTOOL_USES_SYSROOT == 'foo'] Oct 06 04:06:26 } Oct 06 04:06:50 * kergoth_ thinks this will clean up some of our config files nicely Oct 06 04:06:52 ah right Oct 06 04:07:03 anyway I have more ugliness already Oct 06 04:07:06 working Oct 06 04:07:09 as usual :) Oct 06 04:07:17 hard to avoid in some cases, nature of the beast Oct 06 04:07:34 so now I guess I should be able to switch libtools with one var Oct 06 04:07:36 lets see Oct 06 04:07:40 nice Oct 06 04:07:59 ideally I would have kicked 2.2 out Oct 06 04:08:15 just for the sheer fugliness it brings in now Oct 06 04:09:06 yeah, its unfortunate that we're limited in that way right now Oct 06 04:09:45 yeah OE is less agine Oct 06 04:09:49 agile Oct 06 04:09:55 we take too much crap on us Oct 06 04:10:14 ok so someone broke images Oct 06 04:10:31 now I dont see /lib/modules being packaged into the image files Oct 06 04:10:37 wtf Oct 06 04:11:11 and obviously udev cries for mama Oct 06 04:11:40 kergoth: where is this code for embedding modules into image Oct 06 04:12:33 hmmm I am using 2.6.34 and some one applied a patch and the modules are now installed in lib/module/2.6.34.7 Oct 06 04:12:50 instead of lib/modules/2.6.34 Oct 06 04:12:52 when it worked Oct 06 04:13:16 kergoth: the recipe say linux_2.6.34.bb still could that be it Oct 06 04:17:24 so PV is 2.6.34 but the kernel version isn't that Oct 06 04:17:31 yep Oct 06 04:17:41 dunno if thats an issue or not without looking at the packaging bits, i assume it is since its broken for you :) Oct 06 04:17:42 its 2.6.34.7 Oct 06 04:18:00 yeah I reverting a patch Oct 06 04:18:18 5786f08ba6066e9af60104c5d5606c106e8df7d2 Oct 06 04:18:22 this one Oct 06 04:19:27 kergoth_: I was afraid libtool broke it o.0 Oct 06 04:19:38 * khem is paranoid Oct 06 06:34:57 gm Oct 06 06:48:41 03Martin Jansa  07master * r5b3a4eb994 10openembedded.git/recipes/mesa/ (5 files): Oct 06 06:48:41 mesa: move mesa-progs from mesa-common.inc to mesa-version.inc Oct 06 06:48:41 * for mesa-7.9 and newer there are mesa-demos in separate repository Oct 06 06:48:41 * I'll add recipe for latest mesa-demos later Oct 06 06:48:41 Signed-off-by: Martin Jansa Oct 06 06:55:25 hi, someone already working on talloc recipe? Oct 06 06:57:56 hmm someone tried outside OE http://samba.2283325.n4.nabble.com/cross-compiling-talloc-with-waf-td2482920.html Oct 06 07:31:09 good morning Oct 06 08:05:35 khem: not sure if it's result of latest gcc patch, but today I got ICE while building samba_3.2.15 http://tinderbox.openembedded.org/packages/816608/ Oct 06 08:06:24 khem: I lost my link yesterday. Do you think I should add http://pastebin.com/KiMhk4hx to (e)glibc recipes? Oct 06 08:13:14 03Marco Cavallini  07org.openembedded.dev * r9febf6a6bf 10openembedded.git/recipes/linux/linux_2.6.28.bb: linux_2.6.28.bb: changed link to Ronetix' patch for machine ronetix-pm9263 Oct 06 08:21:35 morning Oct 06 08:45:54 good morning Oct 06 08:56:30 hi florian Oct 06 08:56:56 florian, i saw you created the dockstar machine, are you actively working on it ? Oct 06 08:59:49 eFfeM_work: hi Oct 06 08:59:53 i'm the maintainer of sheevaplug and openrd-client; did spent too much time on other things, but the linux kirkwood tree seems not active any more, saw you used 2.6.35, i want to move sheeva and openrd also to 2.6.35; then come with a patch to update conf/machine/include/kirkwood.inc to make linux the default kernel Oct 06 09:00:39 eFfeM_work: "actively" depends on free time but yes I'm trying to do useful things with this device. Oct 06 09:00:39 i can alsoimage we move EXTRA_IMAGECMD_jffs2 to kirkwood.inc Oct 06 09:01:11 yeah, same here with sheeva, actually dockstar is a nice device which can be obtained cheaply nowadays Oct 06 09:02:30 I don't know offhand if this would work for all kirkwood, but for the "default" ones used in Sheevaplug, Dockstar and TK71 2.6.35 seems to sopport very well. The dockstar patch is not more then the machine file. Oct 06 09:03:30 yeah saw that Oct 06 09:03:41 The jffs2 I would leave in the device files. This makes it more clear that it depends on the actual flash used in the device and not the CPU. Oct 06 09:04:04 These devices are really cheap now... got mine for EUR 20 in a shop. Oct 06 09:04:05 sheeva had one or two additional patches, actually I already did those and booted it, didn't get to consolidatingthat Oct 06 09:04:13 wrt jffs2: you're right Oct 06 09:04:16 eur 20 is nice! Oct 06 09:04:33 guess wd is dropping the product Oct 06 09:04:54 at some point in time marvell suggested a $50 plug Oct 06 09:04:59 this is better Oct 06 09:05:44 There seem to be several similar devices around - Seagate has one for 3.5" disks now and one where two 2.5" SATA drives fit into. Oct 06 09:06:17 that would be cool, only thing missing in my sheeva is sata, taht's also why i got the openrd client Oct 06 09:06:30 which btw needs some more work as e.g. the video driver has not been mainlined Oct 06 09:06:41 Hardware hackers seem to like it... there are PCB designs around for adding another Ethernet, SD and such things :) Oct 06 09:06:53 cool, do you have a url ? Oct 06 09:07:50 The best collection I'm aware of is here: http://www.mikrocontroller.net/articles/Dockstar but German only :-( Oct 06 09:09:00 The OpenRD is nice... I wonder if we have one here at the office :-) Oct 06 09:11:23 hm, that openrd looks really nice Oct 06 09:11:28 and it's nice price Oct 06 09:11:56 wifi version would be nice Oct 06 09:12:05 considering I'm using this http://www.embeddedarm.com/products/board-detail.php?product=ts-7800 Oct 06 09:13:32 hallo Oct 06 09:13:41 filip: you can add wifi with usb stick easily Oct 06 09:14:27 ynezz: well I could use an old pc as well ;). This thing is all about completeness in a small box Oct 06 09:14:34 florian: thanks for the link, I can read german (& speak a tiny bit) Oct 06 09:14:43 filip: and power consumption! Oct 06 09:14:52 florian btw did you get it in an internet shop? Oct 06 09:15:14 and fancyness! Oct 06 09:15:27 I am using an old compaq deskpro for stuff like this Oct 06 09:15:32 it pulls around 30W Oct 06 09:15:48 has room for two 3.5" drives, has ethernet, pci, usb and everything Oct 06 09:15:56 and costs like... 5EUR now Oct 06 09:16:02 ifind this ts7800 thing too expensive when comparing with other things (beagle, hawk, sheeva, openrd) Oct 06 09:16:13 disadvantage of openrd is size Oct 06 09:16:59 30w, a sheeva with usb hd will probably use less than 10 w Oct 06 09:17:00 but the advantage of ts7800 board is their availability Oct 06 09:18:30 30W is quite a lot I think for 365/24/7 Oct 06 09:18:57 hi , i m having problem on toolchain sdk C++ hello.c , i tried to compile code : http://pastebin.com/wvnpg2Gr with this command , but when i add clutter.h i got huge errors as shown here http://pastebin.com/HqdtJmDR , whats wrong ? someone can help ? Oct 06 09:19:22 JDuke127: you really need to improve your reading skills Oct 06 09:19:31 JDuke127: why do you think it should find clutter.h like this? Oct 06 09:19:40 florian: found teh atelco link for dockstar, they noticed the thing is hot so now it is eur 29.90 Oct 06 09:19:52 JDuke127: i told you yesterday to use devshell, and then pkg-config --cflags NameOfClutterModule to get the right include paths Oct 06 09:20:34 ynezz: well the disks need power Oct 06 09:20:41 zecke , it works fine with devshell yes but , someone told me to install bitbake toolchain Oct 06 09:20:46 ynezz: and with ext3 it's not easy to spin them down Oct 06 09:21:10 JDuke127: well, please explain to me why should work? Oct 06 09:21:26 zecke: and i installed toolchain sdk "bitbake meta-toolchain" Oct 06 09:21:38 i m not using standard OE arm eabi cc Oct 06 09:21:45 its meta toolchain Oct 06 09:21:50 JDuke127: that is not answering the question Oct 06 09:21:51 i set PATH Oct 06 09:22:10 i donno , yes it works on devshell fine Oct 06 09:22:44 JDuke127: why do you think should work? Oct 06 09:23:09 i did "opkg-target libclutter" Oct 06 09:24:17 JDuke127: seriously, before asking others for help you should do something yourself. Do you even have a clutter.h in the unpacked toolchain? find /path/to/toolchain -name clutter.h Oct 06 09:25:24 i looked inside include , it has clutter.h , yes Oct 06 09:25:58 zecke , did you used meta-toolchain before ? Oct 06 09:27:31 zecke , i got clutter.h /home/kadirbasol/OE/sdk/usr/local/angstrom/arm/arm-angstrom-linux-gnueabi/usr/include/clutter-1.0/clutter/clutter.h Oct 06 09:29:10 JDuke127: yes, but that is not what you include... you should compile with foo-bar-g++ `pkg-config --cflags clutter-1.0` -o myapp main.cc Oct 06 09:29:40 hmm g++ , not gcc Oct 06 09:29:58 JDuke127: the compiler only searches in /home/kadirbasol/OE/sdk/usr/local/angstrom/arm/arm-angstrom-linux-gnueabi/usr/include/... it will not magically go into clutter-1.0/clutter Oct 06 09:30:33 JDuke127: well, you used C++ above, but for including it does not matter... on top of that the toolchain is unlikely to be relocatable... it was built to be in /usr/local/angstrom Oct 06 09:32:20 but my compiler recognizes , clutter.h , there is no error like , Eg: "File not found 'clutter.h' " , Oct 06 09:32:40 i think there r some conflicts or C or C++ err Oct 06 09:33:32 JDuke127: it is a string, how should it not recognize it? Oct 06 09:34:14 when i do this Oct 06 09:34:17 kadirbasol@kadirlinux:~/OE/sdk/usr/local/angstrom/arm$ arm-angstrom-linux-gnueabi-g++ ~/Desktop/hello.cc /home/kadirbasol/Desktop/hello.cc:4:21: error: clutter.h: No such file or directory Oct 06 09:34:24 JDuke127: use strace to see where your compiler is searching for clutter.h... it is certainly not in /home/kadirbasol/OE/sdk/usr/local/angstrom/arm/arm-angstrom-linux-gnueabi/usr/include/clutter.h Oct 06 09:34:24 but i add Oct 06 09:34:39 arm-angstrom-linux-gnueabi-g++ ~/Desktop/hello.cc -I /home/kadirbasol/OE/sdk/usr/local/angstrom/arm/arm-angstrom-linux-gnueabi/usr/include/clutter-1.0/clutter Oct 06 09:35:00 then , i got Oct 06 09:35:03 home/kadirbasol/OE/sdk/usr/local/angstrom/arm/arm-angstrom-linux-gnueabi/usr/include/clutter-1.0/clutter/clutter-stage.h:255: error: ‘height’ was not declared in this scope /home/kadirbasol/OE/sdk/usr/local/angstrom/arm/arm-angstrom-linux-gnueabi/usr/include/clutter-1.0/clutter/clutter-stage.h:266: error: ‘G_END_DECLS’ does not name a type Oct 06 09:35:07 and many more errors... Oct 06 09:35:32 maybe clutter library broken Oct 06 09:35:36 JDuke127: well, in that case the clutter authors messed up their G_BEGIN_DECLS/G_END_DECLS. Oct 06 09:36:08 ok Oct 06 09:36:14 i try devshell back now Oct 06 09:36:15 JDuke127: What is your goal? You do not seem to have a good understanding of C/C++, compiling, toolchains. Maybe you should refocus on the problem you want to solve? Oct 06 09:36:38 now i do Oct 06 09:36:38 bitbake -c devshell clutter Oct 06 09:36:46 JDuke127: even for the cross toolchain... you need to use `pkg-config --cflags --libs clutter-1.0` Oct 06 09:37:16 JDuke127: we use pkg-config to 'find' the right include paths... and probably the preferred way to include clutter.h then is to use clutter/clutter.h Oct 06 09:39:39 zecke , it compiles on devshell its ok Oct 06 09:39:59 i just get 1 warning Oct 06 09:40:00 http://pastebin.com/6i2wdmMm Oct 06 09:40:06 this warning makes problem ? Oct 06 09:40:26 warning: libc.so, needed by /home/kadirbasol/OE/build/tmp-igep0020/sysroots/x86_64-linux/usr/armv7a/lib/gcc/arm-angstrom-linux-gnueabi/4.3.3/../../../../arm-angstrom-linux-gnueabi/lib/libgcc_s.so, not found (try using -rpath or -rpath-link) Oct 06 09:41:51 JDuke127: warning, most likely an issue with the binutils 2.18.. I have sent an email to the oe-dev list but got little response Oct 06 09:42:31 hmm when we update binutils , problem will be solved Oct 06 09:43:46 JDuke127: well, it is a warning... nothing to worry about Oct 06 09:45:39 ok Oct 06 09:45:51 now i ve uploaded my binary to my card Oct 06 09:45:54 i ve ran app Oct 06 09:46:18 this time clutter seems broken : ClutterEGL-WARNING **: eglChooseConfig failed , ClutterEGL-CRITICAL **: Unable to find suitable GL visual. Oct 06 09:46:41 i got this errors in loop Oct 06 09:46:49 no display shown Oct 06 09:47:32 JDuke127: well, you are in an engineering channel. You will need to think before you post questions Oct 06 09:48:12 JDuke127: did you at least google gfor eglChooseConfig? Oct 06 09:49:59 yes Oct 06 09:51:12 * eFfeM_work admires zecke's patience :P Oct 06 09:51:39 JDuke127: from where do you start it? do you have an xserver running or such? Oct 06 09:55:11 no i dont like X11 Oct 06 09:55:14 i hate X11 Oct 06 09:55:22 i m using pure framebuffer on console Oct 06 09:55:31 and i can run simple OGL Apps Oct 06 09:55:37 but not clutter running Oct 06 09:55:40 EGL Oct 06 09:57:07 JDuke127: and you are sure that clutter has an EGL backend != ClutterEGLX that does not require X11? Oct 06 09:58:22 how can i do this config ? Oct 06 09:58:35 read the source luke Oct 06 09:58:45 or ask in a clutter forum Oct 06 09:59:12 eFfeM_work: that is mean to ask him to pester the clutter devs Oct 06 09:59:36 JDuke127: well, read the source.. search for ClutterEGL and calls to eglChooseConfig Oct 06 10:24:03 03Koen Kooi  07org.openembedded.dev * rf1822d1110 10openembedded.git/recipes/opkg/opkg-collateral.bb: opkg-collateral: make angstrom use SHR tmpdir fix and change package arch to all Oct 06 10:50:48 * XorA guesses the ltg planet got stuck sometime in September and stopped updating Oct 06 10:51:02 florian: is the ltg planet stuck? Oct 06 10:54:12 i'll take a look Oct 06 11:00:06 * XorA thinks OE devs are normally more verbose than a months gap :-) Oct 06 11:01:53 ;D Oct 06 11:05:11 florian: that looks better :-) Oct 06 11:06:02 indeed... looks liek there is a problem with the cronjob Oct 06 11:08:24 I have added qt-embedded to my image like this: http://pastebin.com/Lve9QqfE, cherry picking what I need. Now want to use qt4-embedded-gles, but I don't know how to tell it to compile that one instead of qt4-embedded. Is this something I do in DEPENDS? Oct 06 11:27:13 03Koen Kooi  07org.openembedded.dev * r77869f4261 10openembedded.git/recipes/powervr-drivers/omap3-sgx-modules_1.4.14.2616.bb: Oct 06 11:27:13 omap3-sgx-modules: update to latest SDK release and add ti816x support Oct 06 11:27:13 Signed-off-by: Koen Kooi Oct 06 11:27:27 03Koen Kooi  07org.openembedded.dev * rd50346e98e 10openembedded.git/recipes/powervr-drivers/libgles-omap3.inc: Oct 06 11:27:27 libgles-omap3: add ti816x support and fix packaging Oct 06 11:27:27 Signed-off-by: Koen Kooi Oct 06 11:28:34 Is there any documentation about setting up a ipk server for custom packages (conf file formats, http server requirements, etc)? Oct 06 11:38:44 mawillia: it is just a regular http server Oct 06 11:42:34 OK, so other than the .ipk file, what needs to be there to get picked up in the "opkg update" call? Looks like a compressed list get's pulled down. What needs to be in it? (I'm a noob, so if there's an FM, I'll be glad to go read it). Oct 06 11:48:06 mawillia: you also need package files, if you do a bitbake package-index (or something like that; don't to this very often) everything you need ends up in tmp/deploy/ipk/* Oct 06 11:48:12 if I am correct Oct 06 11:49:57 mawillia: and read this http://docs.openembedded.org/usermanual/html/image_class.html Oct 06 11:50:02 especially the method 2 secion Oct 06 11:50:26 eFfeM_work: hi Oct 06 11:50:35 hi ant_work Oct 06 11:50:43 how are you doing? Oct 06 11:50:54 I launched a first attack to /linux for pruning ;) Oct 06 11:51:05 great :-) Oct 06 11:51:15 very busy with biz, not enough time to go in depth Oct 06 11:51:48 pity, but sounds familiar Oct 06 11:52:00 did you plan to come to elc/oedem? Oct 06 11:52:01 next stepl would be remove the duplicate patches shared between similar linux recipes Oct 06 11:52:16 i.e. extending FILESPATHPKG Oct 06 11:52:30 same could be done for linux boot logo Oct 06 11:53:13 i have some mixed feelings on removing dup patches, it is nice filewise, but it becomes harder to find if a file is unused Oct 06 11:53:38 and I really dislike FILESPATHPKG that refer to dirs for other versions Oct 06 11:53:39 true, best would be move the recipe+patches to obsolete Oct 06 11:53:52 common patches could go to linux/ dir or so Oct 06 11:53:59 but the case linux-kexecboot -> linux is not the only one Oct 06 11:54:19 actually is just different defconfig Oct 06 11:54:36 (and extra .inc file) Oct 06 11:54:37 but i also saw that some glibc recipes extended the path to add glibc-2.4 or so which is a recipe for errors Oct 06 11:54:45 :/ Oct 06 11:55:17 it might also be interesting to merge linux and linux-libc-headers Oct 06 11:55:40 well, in the case there are many recipes, best would be to insulate the patch of each in a subdir Oct 06 11:55:52 general case being putting the shared files in /files Oct 06 11:56:03 If I clean "qt4-embedded" and then put "qt4-embedded-gles" into DEPENDS, will that make sure the libs I later add to IMAGE_INSTALL are -gles'y? Oct 06 11:56:40 eFfeM_work: s/many recipes/many versions of same recipe/ Oct 06 11:56:43 btw i also have some concerns with inc files, i feel people sometimes change the inc file but at best see if all versions build and at worst only check if it fixes their version and do not care about the other versions Oct 06 11:57:04 ant_work: agree Oct 06 11:57:21 and actually u-boot is also somewhat of a mess Oct 06 11:57:28 atm there is a bit of 'anarchy' :) Oct 06 11:57:50 yes Oct 06 11:58:19 heh /linux and /u-boot will be nightmares... Oct 06 11:59:13 btw pruning 2.4 kernels I saw a 'triton' machine. Seems almost unmaintained. Oct 06 11:59:31 I did resist and did not move it's recipes to obsolete Oct 06 11:59:33 but Oct 06 11:59:34 (actually checked u-boot, it is not that bad 9apart from omap3, omap3beagleboard, omap3pandora versions (that i would expect to be mergeable) and a very big u-boot_git recipe Oct 06 12:00:15 one of the things I would like to get on the agenda is clearly identifying owners/maintainers for every machine and distro Oct 06 12:00:25 well, u-boot git is 'refugium peccatorum' :) Oct 06 12:00:33 and preferably move the orphaned ones to obsolete or unsupported or so Oct 06 12:00:55 i know, i fear I added my share to u-boot git Oct 06 12:01:36 I for myelf added some lines to it too Oct 06 12:02:21 he.. the Zaurus had a single git revision working Oct 06 12:04:29 anyway, the flood of omap3 recipes will not end up quickly Oct 06 12:04:52 I see in the kernel ML there are still 'fixes' for 2.6.37 Oct 06 12:05:18 nope, omap people are good in creating recipes, but not too god at removing the old ones Oct 06 12:05:25 let's hope by then someone will remove all after 2.6.32 Oct 06 12:05:41 ^_^ Oct 06 12:06:06 actually every time I see the ti dir I am also wondering who knows what the difference is between all those variants Oct 06 13:00:17 eFfeM_work: probably few TI guys do Oct 06 13:02:12 hrw guess so, but imho the idea of recipes in oe is that they are usable for others; guess it would be nice to know what is the purpose of different versions of a recipe (in general, this does not only apply to ti) Oct 06 13:02:53 sometimes this is due to license changes, sometimes footprint, sometimes quality, but in most cases it is unknown Oct 06 13:08:13 * XorA plays hunt the TI guy Oct 06 13:09:17 XorA: score? Oct 06 13:09:33 200 grots to the person who nails koen :-) Oct 06 13:11:14 eFfeM_work: can you give me some "ti" example, say from the recipe/linux/ ? I'm just currious Oct 06 13:20:51 ynezz: there are 14 recipes of ti-xdctools in the ti dir, no idea what the difference is between those Oct 06 13:21:28 also kernel has some omap variants but these are for different boards or maybe they have use (never peeked into it) Oct 06 13:22:08 ah lol at that ti dir Oct 06 13:22:09 XorA: maybe ibot will nail him if you lart him Oct 06 13:22:54 * XorA misses his Nine Inch Nails gun from quake :-) Oct 06 13:25:00 XorA: tried to nail koen, ended up with a trauma surgeon :p : http://www.lmgtfy.com/?q=nail+koen&l=1 Oct 06 14:02:28 good morning .... i managed to get openjdk build for an armv4 ... and the biggest thing i can suggest is use gcc-4.4.4 because 4.4.2 is only the first version of gcc to support eabi on armv4 but in my experience building a lot of packages for oe and gentoo with each version 4.4.4 is much more stable and bug free ..... so In angstrom it defaults to 4.4.2 for armv4 ... bump that to 4.4.4 and things will go a lot easier Oct 06 14:08:38 maybe could appear a silly question but what is the purpose of qt4-tools-sdk package ? Oct 06 14:09:06 and how it works ? Oct 06 14:10:25 mckoan: it is a SDK package. So it is used inside the SDK (e.g. meta-toolchain-qt) Oct 06 14:11:07 mckoan: it contains the Qt tools used when building. E.g. qmake (to create a makefile), rcc (to convert files into a .c file with pathnames), moc (to preprocess files and create c++ code)... Oct 06 14:11:23 the *sdk classes are just regular classes with altered prefixes to the sdk root, iirc. crosssdk == cross + sdk prefix Oct 06 14:11:37 zecke: do you mean it is a meta-toolchain-qt dependency? Oct 06 14:11:48 * mckoan looking into meta-toolchain-qt.bb Oct 06 14:12:10 zecke: http://git.openembedded.org/cgit.cgi/bitbake/commit/?id=606fa1fd97cbd47a6a7ebdc7a2e6aa93a8f65cf5 - this seem sane to you? Oct 06 14:12:27 mckoan: indirectly yes, you will find it in task-*qt*-host.bb Oct 06 14:13:04 zecke: meta-toolchain-qt doesn't exist Oct 06 14:13:16 zecke: do you mean meta-toolchain-qte ? Oct 06 14:13:46 mckoan: qte for master, -qt exists in a branch Oct 06 14:15:06 what do I get with meta-toolchain-qt ? an environment to install everywhere ? Oct 06 14:16:34 its an sdk. use it to build things outside of OE Oct 06 14:16:50 mckoan: please check the usermanual Oct 06 14:17:17 mckoan: there is a section on meta-toolchain-qte, how it is build, how the result is used Oct 06 14:21:41 hello , i tried to compile a sample clutter app but i got this error on pkg-config : http://pastebin.com/urnXvEGA , someone know how to fix that problem ? Oct 06 14:22:30 JDuke127: mesa should provide gl Oct 06 14:22:55 but i m using omap3 Oct 06 14:22:59 omap3 has mesa ? Oct 06 14:23:06 igepv2 card Oct 06 14:23:17 zecke: I missed it, thx Oct 06 14:24:11 hmm Oct 06 14:24:17 seems like it has mesa Oct 06 14:24:20 very interesting Oct 06 14:28:00 zecke: I missed it because my paper printed version is too old :-D Oct 06 14:28:11 JaMa , i did opkg-target install mesa Oct 06 14:28:20 JaMa , but it still didnt success and failed Oct 06 14:29:00 i got same error Oct 06 14:29:01 Package gl was not found in the pkg-config search path. Oct 06 14:29:01 no idea what's opkg-target Oct 06 14:29:15 its meta-toolchain Oct 06 14:30:15 JDuke127: can you type which pkg-config? Oct 06 14:32:19 zecke , it gaves "/home/kadirbasol/OE/sdk/usr/local/angstrom/arm/bin/pkg-config" , correct path of meta-toolchain Oct 06 14:32:27 maybe i ve some lack of libraries ? Oct 06 14:32:50 without meta-toolchain i got this Oct 06 14:32:54 from devshell Oct 06 14:33:26 http://pastebin.com/n5PTXsFt Oct 06 14:33:30 this is from devshell Oct 06 14:33:36 from meta-toolchain when i do it Oct 06 14:33:41 i got this Oct 06 14:34:27 http://pastebin.com/KRM0GSiK Oct 06 14:34:36 this one is from meta-toolchain Oct 06 14:34:43 so why meta-toolchain suck Oct 06 14:34:49 i need more libraries ? Oct 06 14:45:10 wow, making referencing nonexistent variables an error is showing all sorts of bugs :) Oct 06 14:45:17 * kergoth_ fixes a copule Oct 06 14:46:29 its exposing many variable renames from classes which weren't done in all recipes Oct 06 14:46:31 heh Oct 06 14:46:35 well, not many, but some Oct 06 15:02:32 03Marco Cavallini  07org.openembedded.dev * r1d91da33cd 10openembedded.git/conf/distro/include/kaeilos-2010.inc: kaeilos-2010.inc: unset DISTRO_FEED_CONFIGS for kaeilos Oct 06 15:14:23 03Paul Menzel  07master * rf58a5d59b7 10openembedded.git/recipes/openmoko-3rdparty/mokosuite2_git.bb: Oct 06 15:14:23 mokosuite2: Add `libfakekey` and `vala-native` to DEPENDS. Oct 06 15:14:23 During the latest bumps of `SRCREV` I guess it was forgotten to add some dependencies. Oct 06 15:14:23 `libfakekey` is needed for a successful configure and without `vala-native` `valac` is not found on a clean and minimal build system. Oct 06 15:14:23 Signed-off-by: Paul Menzel Oct 06 15:14:23 Acked-by: Martin Jansa Oct 06 15:14:33 03Paul Menzel  07master * r0f83f5f36e 10openembedded.git/recipes/freesmartphone/msmcommd_git.bb: Oct 06 15:14:33 msmcommd_git: Add `libfsotransport` to `DEPENDS`. Oct 06 15:14:33 Signed-off-by: Paul Menzel Oct 06 15:14:33 Acked-by: Martin Jansa Oct 06 17:11:51 hm llvm 2.8 released Oct 06 17:21:11 Is there a way to spead up the bitbake command? What's with the 0 to 100% bitbake files? Oct 06 17:21:17 speed* Oct 06 17:21:26 ????? Oct 06 17:21:47 you can help improve bitbake so its parsing stuff parallel Oct 06 17:24:28 woglinde, what is that 0 to 100% file stuff is does everytime you run bitbake? Seems like it takes a while. Oct 06 17:25:29 only the first parsing Oct 06 17:25:33 the second is faster Oct 06 17:26:26 Also where are my SVN source files placed? I notice the S variable that most _svn.bb files have. I'm just doing a quick do_compile, but I'm not sure where main.cpp is placed. I tried ${WORKDIR} and ${S} but neither work. Kind of confused. Oct 06 17:26:31 http://pastebin.com/MMn9icdh Oct 06 17:27:06 03Frans Meulenbroeks  07org.openembedded.dev * r26bba2c633 10openembedded.git/recipes/matchbox-keyboard/files/ (2 files): Oct 06 17:27:06 matchbox-keyboard : moved unused files to obsolete dir Oct 06 17:27:06 Signed-off-by: Frans Meulenbroeks Oct 06 17:27:06 03Frans Meulenbroeks  07org.openembedded.dev * re85268535b 10openembedded.git/recipes/tar/files/ (configure.patch m4.patch): Oct 06 17:27:07 tar : moved unused files to obsolete dir Oct 06 17:27:07 Signed-off-by: Frans Meulenbroeks Oct 06 17:27:08 03Frans Meulenbroeks  07org.openembedded.dev * rf12a78922c 10openembedded.git/recipes/tcpdump/files/fix-paths.patch: Oct 06 17:27:08 tcpdump : moved unused files to obsolete dir Oct 06 17:27:09 Signed-off-by: Frans Meulenbroeks Oct 06 17:27:09 03Frans Meulenbroeks  07org.openembedded.dev * re2d2f6959c 10openembedded.git/recipes/tor/tor-0.1.1.26/openssl.patch: Oct 06 17:27:10 tor : moved unused files to obsolete dir Oct 06 17:27:11 Signed-off-by: Frans Meulenbroeks Oct 06 17:27:11 03Frans Meulenbroeks  07org.openembedded.dev * r9aedaeda16 10openembedded.git/recipes/tslib/tslib/multievent.patch: Oct 06 17:27:12 tslib : moved unused files to obsolete dir Oct 06 17:27:12 Signed-off-by: Frans Meulenbroeks Oct 06 17:27:13 03Frans Meulenbroeks  07org.openembedded.dev * rc8a76d4962 10openembedded.git/recipes/uclibc/files/ (armeb-kernel-stat.h.patch kernel-key-t-ipc.h.patch): Oct 06 17:27:14 uclibc : moved unused files to obsolete dir Oct 06 17:27:14 Signed-off-by: Frans Meulenbroeks Oct 06 17:27:17 svn checksout the stuff under archivedir/svn/projectname Oct 06 17:27:28 and makes a tar.gz-file of it Oct 06 17:27:30 03Frans Meulenbroeks  07org.openembedded.dev * r05ac7e7ead 10openembedded.git/recipes/matchbox2/files/glib-2.8-backport.patch: Oct 06 17:27:30 matchbox2 : moved unused files to obsolete dir Oct 06 17:27:30 Signed-off-by: Frans Meulenbroeks Oct 06 17:27:37 which it than extract to S Oct 06 17:27:38 03Frans Meulenbroeks  07org.openembedded.dev * ra83a9bf8a4 10openembedded.git/recipes/matchbox-wm/matchbox-wm/kbdconfig_keylaunch_simpad.patch: Oct 06 17:27:38 matchbox-wm : moved unused files to obsolete dir Oct 06 17:27:38 Signed-off-by: Frans Meulenbroeks Oct 06 17:28:14 How would that look in my do_compile stage then if I want main.cpp (It's in trunk by itself in the repository). Oct 06 17:29:03 foobaz, guess this will speed things up drastically: http://www.dell.com/us/business/p/poweredge-r910/pd?~ck=anav Oct 06 17:30:16 woglinde: also I turned off the tarball checkout. Oct 06 17:30:27 unless you mean something else Oct 06 17:32:11 oh I see Oct 06 17:32:25 So do I have to do something in like do_fetch to uncompress them someplace? Oct 06 17:33:31 foobaz: the archive is uncompressed in the unpack task.. that's what its for. Oct 06 17:33:54 its likely a dir in ${S} of the name of the last component of your svn path, e.g. projectname or trunk Oct 06 17:34:01 er, a dir in WORKDIR Oct 06 17:34:18 just bitbake -e yourrecipe |grep \^WORKDIR= Oct 06 17:34:20 and look in there. Oct 06 17:49:01 this may seem like a really stupid question once you done compiling for angstrom ... how to you do install the opkgs to a mounted sdcard .. i have all my ipkg just sitting here ready to go Oct 06 17:51:40 cargoudel: make an image (e.g. console-image) and extract teh tar file on your sd card Oct 06 18:01:41 kergoth_: odd. Yeah I see '...build/tmp-angstrom_2008_1/work/armv7a-angstrom-linux-gnueabi/walker-svn-r1/' as the working directory. In it is temp trunk and walker but there's no main.cpp file. In in the SVN though (I can see if I do a fresh checkout). Oct 06 18:01:53 like i just said Oct 06 18:01:57 what's the final component ofthe svn path? Oct 06 18:02:00 clearly its trunk Oct 06 18:02:06 so ${WORKDIR}/trunk is where the sources are Oct 06 18:02:14 so set S = "${WORKDIR}/trunk" int he recipe Oct 06 18:02:35 oh. I'll try that. Thanks Oct 06 18:02:44 np Oct 06 18:09:37 kergoth: should kernel-modules appear in RDEPENDS to get included in the final image ? Oct 06 18:10:11 if you want *all* modules, but its unlikely you do, more likely task-base will intelligently pull in the modules you need based on the capabilities in the features variables Oct 06 18:10:13 * kergoth shrugs Oct 06 18:11:00 kergoth: hmmm I end up with 0 modules that seems wrong Oct 06 18:11:16 kergoth: it worked few days ago Oct 06 18:11:27 dunno Oct 06 18:11:35 * kergoth doesn't watch individual oe recipe commits very closely Oct 06 18:11:59 kergoth: http://pastebin.com/fdU1Caqf still can't find it. I wonder why it's not unpacking main.cpp Oct 06 18:12:25 well, where is your main.cpp in your svn? Oct 06 18:12:27 this isn't an oe problem Oct 06 18:12:34 it checks out the svn path you gave it Oct 06 18:12:39 and you're looking in that checkout Oct 06 18:13:06 It's in trunk. Oct 06 18:13:35 I deleted my checkout and if I do a fresh checkout I get a folder with trunk,branches,tags and in trunk is main.cpp Oct 06 18:15:01 kergoth: hmmm I am using minimal-image which RDEPENDS on task-boot Oct 06 18:15:06 but not on task-base Oct 06 18:17:25 that'd explain it Oct 06 18:17:52 foobaz: i'd suggest removing the tarball it created from your svn in DL_DIR, then bitbake -c clean yourrecipe, then try baking it again. Oct 06 18:18:09 rm downloads/*wmich* Oct 06 18:18:10 or whatever Oct 06 18:19:15 kergoth: task-base DEPENDS on task-boot Oct 06 18:19:21 yep, it does Oct 06 18:19:51 kergoth: so I think the minimal-image bypasses the module inclusion logic Oct 06 18:19:56 it does, yes Oct 06 18:20:01 hmmm Oct 06 18:20:15 were you using that image a few days ago when it worked? :) Oct 06 18:20:27 yes but I shove in modules Oct 06 18:20:33 now that I remember Oct 06 18:20:35 ah Oct 06 18:21:02 I keep improving my local.conf for clutter Oct 06 18:21:27 if one uses udev then modules are needed I think Oct 06 18:21:37 few of them Oct 06 18:22:05 so it seems console-image is the next smallest image Oct 06 18:22:24 thats like 4000 tasks Oct 06 18:22:44 Why does bitbake base-image build X11 and gnome stuff? Does it have to do with my distron settings? Oct 06 18:23:28 khem: its that many even for a distro other than angstrom? its largely driven by the DISTRO_FEATURES, so if it has less features, it would build less Oct 06 18:23:42 course we still have some leakage, things like bluez pulling in x and crap, or whatever Oct 06 18:23:48 those need to be fixed period Oct 06 18:24:06 we need to take the time to sit down and review the .dot files from bitbake -g in some of these Oct 06 18:24:17 figure out exactly where things we don't want are coming from, and fix them, on a case by case basis Oct 06 18:24:24 this is how i found the gtk+-native / cross thing Oct 06 18:24:49 re Oct 06 18:24:56 * khem nods Oct 06 18:25:31 OT, but if anyone else is using osx and is tired of colloquy being buggy as hell, Textual is lovely Oct 06 18:25:57 kergoth: I use colloquy Oct 06 18:26:04 never tried textual Oct 06 18:26:22 colloquy has this bug where it randomly stops updating the channel displays, and you ahve to reload the style to fix it Oct 06 18:26:26 the bug is 4 years old in their bts Oct 06 18:26:29 :( Oct 06 18:26:59 kergoth: looks promising Oct 06 18:27:51 seems very simple, yet has the nice features that are above text ones like irssi (e.g. growl notification) Oct 06 18:28:10 kergoth: yeah I was about to ask about growl integration Oct 06 18:28:42 kergoth: nah didn't fix it. heh if I uncompress the tar.gz myself after removing it and calling bitbake again the main.cpp isn't even in there. Thanks for your help. I'll have to look around for the answer later. Oct 06 18:29:17 foobaz: weird. Oct 06 18:29:34 foobaz: check if the repo was checked out correctly Oct 06 18:29:42 how? Oct 06 18:29:54 if I run a checkout from the command line it works fine Oct 06 18:30:03 see under your download dir Oct 06 18:30:23 foobaz: what commandline and how do u put it in .vv Oct 06 18:30:25 .bb Oct 06 18:30:29 pastebin it Oct 06 18:30:58 kergoth: you can use quassel on mac too Oct 06 18:31:02 if u use it Oct 06 18:31:09 it works on mac and windoze Oct 06 18:31:14 along with ln Oct 06 18:31:16 lnx Oct 06 18:31:28 I did up there. I gtg. Oct 06 18:31:42 ok bye Oct 06 18:32:03 03Martin Jansa  07master * r3408a82d7f 10openembedded.git/recipes/freesmartphone/cornucopia.inc: cornucopia: bump SRCREV again for new vala really compatible versions Oct 06 18:32:09 kergoth: its default IRC client on kubuntu thats how I got introduced to it Oct 06 18:33:06 huh, thanks for the pointer, might try that on windows rather than the silverex xchat builds Oct 06 18:37:36 hmm xchat on windows is not feee Oct 06 18:37:46 google for silverex xchat Oct 06 18:37:53 old builds unfortunately Oct 06 18:43:23 whats the quickest way to disable all X from from angstrom Oct 06 18:44:09 This is not a distro or machine support channe Oct 06 18:44:21 seems to stand in the topic Oct 06 18:47:58 kergoth_: kvirc have builds for windows and linux at least :) Oct 06 18:51:16 sorry about that ... forgot about the separation between oe and the distros Oct 06 19:02:55 ~lart 150mb gcc-svn.tar Oct 06 19:02:56 * ibot does a little 'renice 20 -u 150mb gcc-svn.tar' Oct 06 19:03:28 why the hell its downloaidng the 150mb instead of makeing the tar from the svn dir Oct 06 19:09:23 kergoth is there some way to tell bitbake to use the svn instead of downloading the tar ball? Oct 06 19:09:39 empty CVS_TARBALL_STASH or whatever, probably Oct 06 19:09:43 * kergoth shrugs Oct 06 19:10:40 ah thanks Oct 06 19:13:11 damnit, i forgot to write up a proposal to split up classes/ and email the list Oct 06 19:13:13 * kergoth mutters Oct 06 19:15:15 hrms Oct 06 19:15:57 woglinde: use bitbake master Oct 06 19:16:14 khem I do Oct 06 19:16:19 and it dont works Oct 06 19:16:24 woglinde: it now strips all SCM crap before making tar Oct 06 19:16:36 woglinde: update like today ? Oct 06 19:16:39 khem cool Oct 06 19:16:41 khem yes Oct 06 19:16:45 course teh filename is the same, so it'll still download the massive one from teh mirrors.. Oct 06 19:16:54 still tries to download Oct 06 19:16:55 ah yes Oct 06 19:17:01 mirrors need updates Oct 06 19:17:14 so what now? Oct 06 19:17:36 can someone remove it? Oct 06 19:17:46 woglinde: which mirror are you hitting Oct 06 19:17:56 http://www.angstrom-distribution.org Oct 06 19:18:10 cbrake can help removing it on source.openembedded.org Oct 06 19:18:11 03Martin Jansa  07master * r1288cce0db 10openembedded.git/recipes/openmoko-3rdparty/mokowm_git.bb: Oct 06 19:18:11 mokowm: add libfakekey to DEPENDS Oct 06 19:18:11 Signed-off-by: Martin Jansa Oct 06 19:18:15 woglinde: ask koen Oct 06 19:18:40 hm I will fool with /etc/hosts Oct 06 19:18:48 ok so it builds fine with libtool 2.2.6b yippie Oct 06 19:18:56 now lets flip the switch Oct 06 19:19:39 woglinde: you can disable mirrors too Oct 06 19:19:42 hm which switch? Oct 06 19:19:57 woglinde: I am working on libtool 2.4 upgrade Oct 06 19:20:02 khem yes Oct 06 19:20:05 so added a variable to select Oct 06 19:20:11 I know Oct 06 19:20:11 which libtool to use Oct 06 19:20:12 khem, any changes over the past week that would impact sysv shared mem Oct 06 19:20:14 gr_vmcircbuf_sysv_shm: shmat (2): Invalid argument Oct 06 19:20:27 I just started seeing this in gnuradio ... Oct 06 19:20:47 Crofton: hmmm compiler changed Oct 06 19:20:58 Crofton: were you using gcc 4.5 Oct 06 19:21:11 * woglinde is testing llvm2.8 Oct 06 19:21:21 woglinde: llvm2.8 is impressive Oct 06 19:21:23 yes Oct 06 19:21:29 angstrom 2010 Oct 06 19:22:02 Crofton: ok, did you rebuild everything Oct 06 19:22:33 let see if openjdk runs in qemu Oct 06 19:22:46 I think so .... Oct 06 19:23:26 I am thinking I shoud try again Oct 06 19:23:57 Crofton: you might run it under strace Oct 06 19:24:05 and see whats being passed to shmat Oct 06 19:24:06 k Oct 06 19:24:15 Crofton: whats your glibc ver Oct 06 19:24:34 eglibc Oct 06 19:24:42 see angstrom-2010.x Oct 06 19:25:23 2.10.1 Oct 06 19:25:34 Crofton: I would ask you to use eglibc Oct 06 19:25:44 it does Oct 06 19:26:27 eglilbc-2.11-r14.7 Oct 06 19:26:47 ok thats what you are using ? Oct 06 19:33:33 yes Oct 06 19:34:16 How have other people dealt with kernel-arch not dealing with converting ARCH=i386 to ARCH=x86 for more recent kernels? Oct 06 19:34:36 Or does this not get tested often? It's still broken in kernel-arch.bbclass Oct 06 19:35:06 kelvie_: how old is your oe snapshot Oct 06 19:35:26 Crofton|work: hmm I think we need to debug the issue Oct 06 19:35:35 khem: Freshly pulled Oct 06 19:36:15 khem: Or has this been fixed in another branch? I pulled from the main repo, and am on master Oct 06 19:36:34 kelvie_: if its not on master other might not have it Oct 06 19:36:46 kelvie_: what are u building Oct 06 19:36:50 khem: Yeah, the most recent commit to that file is from you Oct 06 19:36:58 khem: An external kernel module Oct 06 19:37:00 kelvie_: if there is a bug, please author a patch and send it to the mailing list Oct 06 19:37:13 eFfeM: yeah, I'm trying to find out the best way to do this Oct 06 19:37:26 the arch/ directory in the kernel has changed from 2.6.22 to 2.6.23 Oct 06 19:37:30 Should I do a version check? Oct 06 19:37:46 Or just ls arch/ inside kernel-arch? Oct 06 19:38:18 Anyways, kernel.bbclass tries to look in arch/$ARCH/include/ to copy to staging Oct 06 19:38:25 which is wrong, because arch/i386/ does not exist anymore Oct 06 19:38:33 And thus no include files for my kernel module Oct 06 19:38:47 At least for include/asm Oct 06 19:39:13 kelvie_: if you specify ARCH=x86 then map_kernel_arch should return same back Oct 06 19:39:27 khem: Yeah, but this is i386 Oct 06 19:39:33 kelvie search for x86 and see your failure Oct 06 19:39:56 So I should not use ARCH=i386? Oct 06 19:40:24 TARGET_ARCH, tha tis Oct 06 19:40:53 kelvie_: well we support kernels where both i386 and x86 have to be consideref Oct 06 19:41:11 hm the old kernel recipe should die Oct 06 19:41:12 *g* Oct 06 19:41:15 kelvie_: pass correct ARCH in the external build Oct 06 19:41:20 khem: We have to add some type of check to kernel-arch, yeah Oct 06 19:41:21 for module Oct 06 19:41:33 And map i386 to x86 for kernels later than 2.6.22 Oct 06 19:41:40 As well as x86_64; that would break as well Oct 06 19:42:07 woglinde: Is there a newer one? Oct 06 19:42:13 o.O Oct 06 19:42:24 kelvie_: yes Oct 06 19:43:26 thas in my kernel.bbclass Oct 06 19:43:29 http://pastebin.com/jj2X6NtZ Oct 06 19:43:35 and deals with x86 Oct 06 19:44:05 woglinde: yeah, I'm doing that now as well Oct 06 19:44:13 It's just a workaround, though Oct 06 19:44:25 kelvie_: try to use get_kernelversion Oct 06 19:44:30 and set it accordingly Oct 06 19:44:47 khem: We have to require linux-kernel-base first though Oct 06 19:44:48 Is that fine? Oct 06 19:44:57 yes Oct 06 19:45:14 khem: Wouldn't a more flexible solution be to just ls the arch/ directory? Oct 06 19:45:24 Or is the unpacked sources not available at that time? Oct 06 19:45:46 look inside get_kernelversion Oct 06 19:45:52 it peeks into sources Oct 06 19:45:58 khem: Ah, okay. Oct 06 19:46:04 I shall peek in myself as well, then :( Oct 06 19:46:06 :) Oct 06 19:46:41 try to use that and if it works post a patch for kernel-arch.bbclass Oct 06 19:46:47 khem: Yep, will do that. Oct 06 19:55:13 khem, I need to remember my sysv shared mem Oct 06 19:59:16 hm, there's something fishy with the bitbake, removed all references to initscripts in tmp, removed even the tmp/cache, but after that if I run bitbake image, the initscripts package isn't build Oct 06 19:59:32 even if I do bitbake initscripts it's not build Oct 06 19:59:55 use -v and -DDD to see whats going on Oct 06 20:01:54 khem, this is the call that fails Oct 06 20:02:02 if (shmat (shmid_guard, first_copy, SHM_RDONLY) == (void *) -1){ Oct 06 20:07:36 woglinde: http://pastebin.com/R8fmDZL6 Oct 06 20:07:44 woglinde: I'm not clever from it either Oct 06 20:07:46 :) Oct 06 20:07:52 is there an easy way to weed out old directories from a work tree (apart from nuking the whole tree); with old directories I mean e.g. a r10 version if there is also an r11 work dir Oct 06 20:08:03 woglinde: it just grep of initscripts debug_log Oct 06 20:08:39 s/it/its/ Oct 06 20:09:35 Unknown package 'initscripts'. Oct 06 20:09:50 I see that also Oct 06 20:10:00 but it finds before Oct 06 20:10:08 if I bump PR, it will work Oct 06 20:10:57 DEBUG: sorted providers for initscripts are: ['/media/data/devel/oe/ts72xx.git/openembedded/recipes/initscripts/initscripts_1.0.bb'] Oct 06 20:11:15 so what? :) Oct 06 20:11:29 ynezz@ntbk:/media/data/devel/oe/ts72xx.git/tmp$ find . -name initscripts -print Oct 06 20:11:30 ./deploy/glibc/sources/GPL/initscripts Oct 06 20:12:09 should be nothing there, I've purged everything Oct 06 20:13:30 as I said before, purged tmp/cache also, so I wonder what could I do more to trigger building of that package Oct 06 20:14:24 ynezz: your find catches files named initscripts, not files with initscripts in it. did you remove the stamps in tmp/stamps? Oct 06 20:15:02 running grep now Oct 06 20:15:20 ynezz@ntbk:/media/data/devel/oe/ts72xx.git/tmp$ grep initscripts stamps/ -R Oct 06 20:15:21 ynezz@ntbk:/media/data/devel/oe/ts72xx.git/tmp$ Oct 06 20:15:24 i don't see why you wouldn't just bitbake -c clean initscripts Oct 06 20:15:28 grep is going to do absolutely nothing Oct 06 20:15:30 stamps are empty files Oct 06 20:15:35 find tmp/stamps -name \*initscripts\* Oct 06 20:15:37 oh Oct 06 20:15:49 if a stamp exists for a given task, bitbake knows its completed it, so it wont run it again Oct 06 20:15:50 I've tried -c clean first Oct 06 20:15:54 then -c distclean Oct 06 20:15:59 then remove ipk Oct 06 20:16:04 then remove tmp/cache Oct 06 20:16:08 then asked here :) Oct 06 20:16:44 ah stamps is full of it Oct 06 20:16:53 clean removes stamps Oct 06 20:16:57 so don't know what to tell you Oct 06 20:17:05 also the purpose of stamps is in the manuals.. Oct 06 20:18:46 * kergoth grumbles Oct 06 20:19:34 sorry and thanks Oct 06 20:19:48 np Oct 06 20:20:02 If I remember it correctly, stamps used to be in the different place Oct 06 20:20:07 nope Oct 06 20:20:21 well, its always been in tmp/stamps/ Oct 06 20:20:25 used to be at toplevel though Oct 06 20:20:33 rather than deeper there by arch Oct 06 20:20:37 or whatever Oct 06 20:20:58 hm, fixed, easy peasy :p Oct 06 20:21:25 jesus, not Oct 06 20:21:52 it finaly build the package, but image can't find it Oct 06 20:23:00 hm seems they didnt build llvm2.8 for arm Oct 06 20:33:29 03Frans Meulenbroeks  07org.openembedded.dev * rc2a9e77a93 10openembedded.git/recipes/mythtv/ (4 files): Oct 06 20:33:29 mythtv: fixed LICENSE added a few other headers for mythplugins and myththemes Oct 06 20:33:29 Signed-off-by: Frans Meulenbroeks Oct 06 20:34:42 hm damn they dont support shared libs anymore in a right way Oct 06 20:38:28 03Marco Cavallini  07org.openembedded.dev * rc53d253363 10openembedded.git/conf/distro/include/kaeilos-2010.inc: kaeilos-2010.inc: replaced typo, sorry Oct 06 20:43:14 When does python code get evaluated in bb/bbclass files? Oct 06 20:43:19 On evaluation of the variable? Oct 06 20:43:34 that depends on what you mean by python code Oct 06 20:43:39 inline snippets of the ${@} sort? Oct 06 20:43:43 kergoth: yeah Oct 06 20:43:48 all variables are expanded when used Oct 06 20:43:57 epxansion is when all ${} are evaluated, including the python ones Oct 06 20:44:02 so yes, on evaluation Oct 06 20:44:09 unless someone uses := in an assignment, this is immediate eval Oct 06 20:44:12 same behavior as gmake Oct 06 20:44:20 khem: you there by any chance? Oct 06 20:44:22 khem: [10:04] < filip> khem: I lost my link yesterday. Do you think I should add http://pastebin.com/KiMhk4hx to (e)glibc recipes? Oct 06 20:44:23 kergoth: gotcha, thanks Oct 06 20:44:25 np Oct 06 20:47:24 um, how should I prepare a list of packages to be build for a distro's online package repository? Oct 06 20:47:34 just some kind of a metapackage with many dependencies? Oct 06 20:48:13 03Frans Meulenbroeks  07org.openembedded.dev * rd0e2850d99 10openembedded.git/recipes/streamripper/streamripper_1.64.6.bb: Oct 06 20:48:13 Fix RDEPENDS Oct 06 20:48:13 Signed-off-by: Frans Meulenbroeks Oct 06 20:48:16 03Frans Meulenbroeks  07org.openembedded.dev * rae44179e69 10openembedded.git/recipes/streamripper/streamripper_1.61.3.bb: Oct 06 20:48:16 streamripper_1.61.3.bb: removed Oct 06 20:48:16 had cve's and was not properly configured and missing depends; Oct 06 20:48:16 latest version is already in OE for quite a while. Oct 06 20:48:17 Signed-off-by: Frans Meulenbroeks Oct 06 20:48:17 03Frans Meulenbroeks  07org.openembedded.dev * r1be6794639 10openembedded.git/recipes/perl/libtree-simple-perl_1.18.bb: Oct 06 20:48:18 Fix RDEPENDS Oct 06 20:48:18 Signed-off-by: Frans Meulenbroeks Oct 06 20:53:35 filip: I am currently testing this patch will push it shortl Oct 06 20:54:11 khem: ah thanks :) Oct 06 20:54:25 khem: do you know perhaps the answer to my question above? Oct 06 20:55:08 filip: look into recipes/tasks and recipes/images Oct 06 20:55:31 I can handle images, I am just wondering about feeds Oct 06 20:56:23 for example there is this ugly script: contrib/angstrom/build-feeds.sh, should I do the equivalent? Oct 06 20:56:53 as in specify a list of packages in the shell file executing bitbake? Oct 06 20:56:55 yes thats a starting point Oct 06 20:57:38 on the other side, ./recipes/tasks/task-openmoko-feed.bb looks like a waaay nicer method Oct 06 20:58:02 could I create recipes/tasks/task-jlime-feed.bb? Oct 06 21:03:59 can't believe i never tried this out before.. these helpers for datasmart are handy, and rather pleasant to look at Oct 06 21:04:12 d.setVars(foo=5, bar=4) Oct 06 21:04:21 d.setVarFlags("foo", func=True, task=True) Oct 06 21:04:32 (these don't exist yet, just playing with them) Oct 06 21:06:29 made my signature unit tests much much more readable Oct 06 21:08:28 03Khem Raj  07master * r5bcd2b8a09 10openembedded.git/recipes/eglibc/ (3 files in 2 dirs): Oct 06 21:08:28 eglibc_2.12.bb/eglibc_svn.bb: Fix build on ARMv4 with EABI Oct 06 21:08:28 Signed-off-by: Khem Raj Oct 06 21:11:24 calling it a day, cya Oct 06 21:11:35 eFfeM: bye Oct 06 21:12:23 hi all , i cant do make on clutter , i ve found the problems some of them but Oct 06 21:13:41 when i try to run "make" i got error as shown on pastebin : http://pastebin.com/dhbDQgSV , why compiling error occurs ? Oct 06 21:14:42 jduke128 it is using pixman from your host Oct 06 21:14:46 which is wrong Oct 06 21:15:00 i know but , there seems it has bug inside Oct 06 21:15:12 my configs seems fine ? how can i change it to use my target Oct 06 21:15:15 pixman Oct 06 21:15:18 not host Oct 06 21:15:22 how to fix ? Oct 06 21:15:49 /usr/lib/libpixman-1.la Oct 06 21:16:09 look at the m4 macros inside clutter Oct 06 21:16:17 maybee they are using some hardcoded stuff Oct 06 21:16:27 or clutter is missing a dep on pixman Oct 06 21:17:01 aclocal.m4 Oct 06 21:17:05 this Oct 06 21:17:38 there is no libpixman there Oct 06 21:18:10 florian: ping Oct 06 21:19:33 ant_: pong Oct 06 21:19:39 hello Oct 06 21:19:56 we have an issue with gpe-image Oct 06 21:20:06 in particular, gpe-package Oct 06 21:20:08 what ? Oct 06 21:20:54 still depends on 'libipkg' Oct 06 21:21:21 hum,IIRC i had the same issue,but i forget say... Oct 06 21:21:31 just changing with opkg won't work. there are hardcoded calls to ipkg iirc Oct 06 21:21:47 yes,i tried it too =d Oct 06 21:21:49 yes,i tried it too =D* Oct 06 21:21:50 angelox_123, hi, yes, I've seen you had same issue Oct 06 21:23:01 khem: I believe that the ARMv4 problem applies to glibc as well Oct 06 21:23:45 filip: it does Oct 06 21:23:53 filip: use eglibc Oct 06 21:24:18 khem: I am, but just saying... ;) Oct 06 21:24:28 eglibc is a superset of glibc + a lot more for embedded systems Oct 06 21:25:10 filip: we dont have recipes for glibc 2.12 yet Oct 06 21:25:17 so its not a problem in OE right now Oct 06 21:25:57 ah :) Oct 06 21:27:58 florian: this is compile log with deps on opkg: http://pastebin.ca/1955875 Oct 06 21:28:09 I am hopin that in future users will only use eglibc Oct 06 21:28:15 as it fits the OE bill better Oct 06 21:30:13 hm Oct 06 21:30:36 ant_: i'll take a look Oct 06 21:30:41 btw, is there a way I could use dpkg instead of ipkg as a package manager on the target system? Oct 06 21:31:55 hm, it seems like mamona does that Oct 06 21:32:16 khem, today Gentoo got libtool-2.2.10 stable for x86 Oct 06 21:36:02 and for HPPA Oct 06 21:47:54 filip: you can should not be that hard Oct 06 21:48:11 ant_: 2.2 is cakewalk Oct 06 21:48:20 ant_: 2.4 is a different beast Oct 06 22:35:59 03Andreas Oberritter  07master * r0741f11dcb 10openembedded.git/recipes/matroska/ (5 files): Oct 06 22:35:59 lib{ebml,matroska}-1.0.0: intitial recipes Oct 06 22:35:59 Signed-off-by: Andreas Oberritter Oct 06 22:35:59 Signed-off-by: Khem Raj Oct 06 22:35:59 03Petr Štetiar  07master * r4085252bf8 10openembedded.git/recipes/initscripts/ (files/save-rtc-uclibc.sh initscripts_1.0.bb): Oct 06 22:36:00 init-scripts: fix save-rtc error with date formatting on uclibc Oct 06 22:36:00 uClibc's date does support only SUSv3 strptime[1], so the date command in Oct 06 22:36:01 save-rtc.sh fails with 'date: invalid date '%2m%2d%2H%2M2010'. Oct 06 22:36:02 1. http://www.opengroup.org/onlinepubs/009695399/functions/strptime.html Oct 06 22:36:02 Signed-off-by: Petr Štetiar Oct 06 22:36:02 Signed-off-by: Khem Raj Oct 06 22:36:03 03Petr Štetiar  07master * rf2e95d221a 10openembedded.git/recipes/busybox/ (busybox.inc files/hwclock.sh): (log message trimmed) Oct 06 22:36:22 busybox: add posibility to specify correct rtc device for hwclock init script Oct 06 22:36:23 With this patch it's now possible to specify rtc device using HWCLOCKDEVICE Oct 06 22:36:23 variable in /etc/default/hwclock and the system time is then set using correct Oct 06 22:36:23 device and vice versa. It's wise to have such functionality, beacuse there're Oct 06 22:36:24 SBCs with more then one rtc device, for example ts72xx. Oct 06 22:36:24 Signed-off-by: Petr Štetiar Oct 06 22:36:25 03Paul Menzel  07master * r3b5ff5ac90 10openembedded.git/recipes/python/ (4 files in 2 dirs): (log message trimmed) Oct 06 22:36:26 python-{native}-2.6.5: Fix parallel build. Oct 06 22:36:39 Randomly the build failed for me with the following error message. Oct 06 22:36:39 libpython2.6.so: undefined reference to `_PyParser_Grammar' Oct 06 22:36:39 The applied patch in [5] tried to fix this problem, but it turns out that this also was due to a possible race. This is already known in OpenOffice too [3]. Oct 06 22:36:41 R. David Murray (dmalcolm) fixed this in Fedora [1] and I applied his patch. Oct 06 22:36:41 I reported this issue upstream [4] and this will only be fixed in Python 2.7, so we have to keep this patch for Python 2.6. Oct 06 22:36:41 03Andreas Oberritter  07master * r0f4f8f3e76 10openembedded.git/recipes/libsigc++-1.2/ (3 files in 2 dirs): Oct 06 22:36:42 libsigc++-1.2: fix installation Oct 06 22:36:42 * install complains about method_slot.h, which is mentioned twice Oct 06 22:36:42 inside Makefile.am. Oct 06 22:36:43 * v2: Added header to patch. Oct 06 22:36:44 Signed-off-by: Andreas Oberritter Oct 06 22:36:44 Signed-off-by: Khem Raj Oct 06 22:36:45 03Andreas Oberritter  07master * rf65559a539 10openembedded.git/recipes/mumudvb/ (mumudvb.inc mumudvb_1.6.bb): Oct 06 22:36:45 (56 lines omitted) Oct 06 22:37:31 03Andrea Adami  07org.openembedded.dev * r912a51db1b 10openembedded.git/conf/machine/include/zaurus-2.6.inc: zaurus-2.6.inc: evaluate here kernel bootlogo size Oct 06 22:37:33 03Andrea Adami  07org.openembedded.dev * r99c8541b71 10openembedded.git/recipes/linux/linux.inc: Oct 06 22:37:33 linux.inc: allow for more bootlogo customization. Oct 06 22:37:33 * kernel recipes can add a custom-sized bootlogo in SRC_URI Oct 06 22:37:33 * e.g. SRC_URI = "file://${LOGO_SIZE}/logo_linux_clut224.ppm.bz2" Oct 06 22:37:34 * before including/requiring linux.inc. Oct 06 22:37:34 03Andrea Adami  07org.openembedded.dev * r08cd165298 10openembedded.git/recipes/linux/linux-kexecboot.inc: linux-kexecboot.inc: don't define LOGO_SIZE here (machine-specific) Oct 06 22:43:01 kergoth: what do u think of this patch http://git.pokylinux.org/cgit/cgit.cgi/poky/commit/?id=0ee7a9e6bdce1f3094e831b9db83087fff1dad17 Oct 06 22:43:41 khem looks good Oct 06 22:43:52 we should merge it Oct 06 22:43:53 that'll get rid of the odd behavior where the local file doesn't exist and it doesnt' know why Oct 06 22:44:04 yep Oct 06 22:44:44 poky's bitbake is quite a ways apart from upstream now due to the work RP is doing on the signatures and per task archiving and all Oct 06 22:44:47 but that one we can grab just fine Oct 06 22:45:00 righto Oct 06 22:45:36 man, testing to catch pstage relocation issues takes a looooong time Oct 06 22:47:19 http://git.pokylinux.org/cgit/cgit.cgi/poky/commit/?id=e3d7890cace71b0a57d2530bf615a58dcb46d96f Oct 06 22:47:23 is interesting too Oct 06 22:47:28 we should consider that one Oct 06 22:47:33 indeed Oct 06 22:48:53 once I am thru libtool 2.4 Oct 06 22:49:11 I will try to separate gcc libraries/runtime Oct 06 22:49:22 * kergoth adds poky as a remote to bitbake to see if its possible to apply the commits directly with a -p path strip Oct 06 22:49:48 kergoth: can you do that Oct 06 22:49:53 I always thought Oct 06 22:50:07 aha Oct 06 22:50:08 i got it Oct 06 22:50:10 poky still uses packages/ instead of recipes/ Oct 06 22:50:12 check this out Oct 06 22:50:18 cool Oct 06 22:50:28 git remote add poky git://git.pokylinux.org/poky Oct 06 22:50:30 git fetch poky Oct 06 22:50:36 git format-patch --stdout -1 0ee7a9e6bdce1f3094e831b9db83087fff1dad17 | git am -p 2 Oct 06 22:50:42 the -p 2 strips off the extra bitbake/ path Oct 06 22:50:45 so it applies perfectly Oct 06 22:50:51 hmmm Oct 06 22:50:54 and then git am ? Oct 06 22:51:03 I always tried cherry-pick Oct 06 22:51:03 * kergoth points at the above | git am -p 2 Oct 06 22:51:11 i dont think cherry pick can do it Oct 06 22:51:12 no -p thing Oct 06 22:51:13 ugh linewrap Oct 06 22:51:15 right Oct 06 22:51:17 but format-patch + am will Oct 06 22:51:24 this should be handy.. Oct 06 22:51:29 cool stuff Oct 06 22:51:36 indeed Oct 06 22:51:50 then add -s to the am to sign off since i'm applying it, and adjust the commit message to kill bitbake/, and ready to push Oct 06 22:52:07 may be we should add a wiki page "How to add patches from poky" Oct 06 22:52:17 yeah, that could be useful Oct 06 22:52:39 I still need to buy a plane ticket for ELCE and OEDEM Oct 06 22:52:41 git am --sign-off ? Oct 06 22:52:41 hmm, the one concern i have with that patch Oct 06 22:52:56 it doesn't re-raise the fetcherror it got, it raises its own Oct 06 22:53:04 which loses any details the fetcher might have provided Oct 06 22:53:11 yeah, -s is --sign-off Oct 06 22:53:42 well, this is an improvement, can revisit it Oct 06 22:54:02 03Joshua Lock  07master * r12066bf508 10bitbake.git/lib/bb/fetch/__init__.py: (log message trimmed) Oct 06 22:54:02 fetch: if mirror fetching fails, ensure exception is raised Oct 06 22:54:02 We catch any exception raised by the fetchers go() method and attempt to work Oct 06 22:54:02 around it by trying any (post) mirrors which are configured. However, should Oct 06 22:54:02 the mirrors fail the exception is lost and the fetch is assumed to have Oct 06 22:54:03 completed successfully. Oct 06 22:54:04 Instead, save the exception and if the local file does not exist after trying Oct 06 22:54:11 heh. Oct 06 22:54:35 i don't know why i didn't think about using am to adjust paths before Oct 06 22:54:52 that was much easier than any other method i've used Oct 06 22:55:34 yeah I thought it preserved the fetcher exception Oct 06 22:56:34 NOTE: Gettext required but not in DEPENDS for file /home/balister/oe/tmp/work/armv7a-angstrom-linux-gnueabi/gcc-4.5-r12.1+svnr164562/gcc-4.5/intl/configure.ac. Oct 06 22:56:35 Missing inherit gettext? Oct 06 22:56:39 just went by ... Oct 06 22:57:19 Crofton|work: for now ignore those Oct 06 22:57:43 gcc uses its own version of gettext Oct 06 22:57:48 ick Oct 06 22:57:59 which is fine Oct 06 22:58:27 hmmm thats target gcc Oct 06 22:58:34 k Oct 06 22:58:39 I just saw it fklash by Oct 06 22:58:47 k Oct 06 23:00:26 kergoth: hmm I reckon same strip level wont work for oe metadata Oct 06 23:00:50 not directly, no, due to the difference in layout Oct 06 23:01:01 right Oct 06 23:01:01 hmmm. Oct 06 23:01:16 * kergoth wonders if there'd be some way to apply into a subdir, or combine a subtree merge with an am or something.. Oct 06 23:01:40 wait, this is promising Oct 06 23:01:50 yep, thatll do it Oct 06 23:01:58 khem: -p + --directory= Oct 06 23:02:13 e.g. -p 5 --directory=recipes/foo Oct 06 23:02:19 strips off, then prepends Oct 06 23:02:21 then applies Oct 06 23:02:23 to am Oct 06 23:02:26 perfect Oct 06 23:02:32 this is assuming of course the recipe filenames are the same Oct 06 23:02:35 not much we could do about that :) Oct 06 23:02:40 other than manually applying Oct 06 23:02:44 but its something Oct 06 23:03:05 yeah Oct 06 23:04:00 * kergoth <3 git Oct 06 23:05:41 khem: note that you could also manually apply the changes, like whent eh recipe files are different Oct 06 23:05:46 and then commit using ci -s -c Oct 06 23:05:53 to use the log, author, etc from the original commit Oct 06 23:06:40 correct Oct 06 23:06:50 I do that many times on ill formed patches sent to OE Oct 06 23:07:00 03Andrea Adami  07org.openembedded.dev * r57c890552b 10openembedded.git/conf/machine/include/zaurus-2.6.inc: zaurus-2.6.inc: remove superfluous MACHINE_KERNEL_VERSION = "2.6" Oct 06 23:07:04 03Andrea Adami  07org.openembedded.dev * rac99aeee69 10openembedded.git/conf/machine/ (include/simpad-2.6.inc simpad.conf): simpad: merge simpad-2.6.inc in simpad.conf Oct 06 23:07:04 03Andrea Adami  07org.openembedded.dev * ra934499cee 10openembedded.git/conf/machine/include/ (collie-2.4.inc simpad-2.4.inc): collie+simpad: remove unused -2.4 inc files Oct 06 23:07:04 03Andrea Adami  07org.openembedded.dev * raa376b7ded 10openembedded.git/conf/machine/simpad.conf: simpad.conf: remove superfluous MACHINE_KERNEL_VERSION ?= "2.6" Oct 06 23:08:08 ah, good :) Oct 06 23:08:15 maybe we need a git tips and tricks section of the oe wiki Oct 06 23:08:21 with both general and specific to oe Oct 06 23:08:43 kergoth, yes please. My git-fu is poor. Oct 06 23:09:11 kergoth: we do have one Oct 06 23:09:18 we should enhance it, then :) Oct 06 23:09:21 * kergoth adds to todo Oct 06 23:09:25 right Oct 06 23:09:48 http://wiki.openembedded.net/index.php/GitPhraseBook Oct 06 23:15:25 git am -s -p 3 --directory=recipes is what we need for recipes Oct 06 23:16:49 kergoth: LIC_FILES_CHKSUM is this something pokyness Oct 06 23:16:57 think so Oct 06 23:17:03 never heard of it Oct 06 23:17:06 hmmm Oct 06 23:17:57 http://git.pokylinux.org/cgit/cgit.cgi/poky/commit/?id=2ccd2e67c5768e59316b086fa70fef837eecb53a Oct 06 23:18:38 kergoth: seems a good think though Oct 06 23:19:42 anyone using x86_64 host? seems that python-psyco does not support this - looking for alternatives to help speed up build time Oct 06 23:20:04 correct, nope Oct 06 23:20:08 yes psyco is only for x86 Oct 06 23:20:45 read that psyco is no longer in devel and there is pypy perhaps to replace it? also read that python built for 32bit compat mode would support psyco - not sure what that entails and if its worth it Oct 06 23:21:19 i'd certainly be interested to see how pypy or unladen swallow or something would do Oct 06 23:21:20 how much do we gain with psyco in 32bit Oct 06 23:21:48 feels like a lot Oct 06 23:22:06 khem, thats a good question - all I know is I moved to a 8GB ram 64bit system with SSD and my build times are about as bad as they were on my old 2GB 32bit laptop Oct 06 23:22:34 tharvey: use ramfs Oct 06 23:22:39 seems like I'm sitting often waiting for cache Oct 06 23:22:55 khem, ramfs for what though? TMPDIR is way too big Oct 06 23:23:09 tharvey: with rm_work Oct 06 23:23:15 it peeks around 8G Oct 06 23:23:19 rm_work? Oct 06 23:23:24 and then you can spit deploy to disk Oct 06 23:23:49 * tharvey reads rm_work.bbclass Oct 06 23:23:50 btw, http://grab.by/6JX6 is Textual. don't miss colloquy at all Oct 06 23:23:57 tharvey: it wipes WORKDIR when the build is done for the recipe Oct 06 23:24:05 tharvey: rm_work is class if you inherit it then it deleted the build dir after generatin packages Oct 06 23:24:12 bitbake world with rm_work for mipsel is about 50GB Oct 06 23:25:10 interesting... so ramfs + rm_work gives you quite a speedup? Oct 06 23:25:17 kergoth: I am gonna try it tonight Oct 06 23:25:21 when I reach hoe Oct 06 23:25:24 home Oct 06 23:25:30 *thick fingers* Oct 06 23:26:01 tharvey: yes it should Oct 06 23:26:19 tharvey: 1:30 mins for a x11-image from scratch Oct 06 23:26:31 that was 1:30 hrs or 90mins Oct 06 23:26:47 tharvey: thats second hand info though Oct 06 23:27:11 my machines have 2G of RAM and a poor core2duo Oct 06 23:27:58 do you perhaps have this config written up anywhere? So you set TMDIR to a ramfs filesystem, and DEPLOY_DIR to a disk filesystem, then inherit rm_work? Oct 06 23:28:38 tharvey: add INHERIT += "rm_work" in your local.conf Oct 06 23:28:45 that will enable it for your builds Oct 06 23:29:44 right, but TMPDIR and DEPLOY_DIR point to ramfs and disk based fs respectively and thats all, or other dir tweaks too? Oct 06 23:34:13 tharvey: mkdir -p /mnt/tmp Oct 06 23:34:34 mount -t ramfs -o size=6000m ramfs /mnt/ram Oct 06 23:35:01 I mean mount -t ramfs -o size=6000m ramfs /mnt/tmp Oct 06 23:36:04 ya, going to give it a spin - thanks for the info! Oct 06 23:36:23 tharvey: but if your system crashes or reboots then it will be gone Oct 06 23:37:02 understood Oct 06 23:37:05 tharvey: cool let up know what crazy speedup you get Oct 06 23:37:18 ya, I'll do a comparison over the next day Oct 06 23:37:30 cool Oct 06 23:37:44 almost one year ago I had three images(console, opie, x11) in 2,5GB after build Oct 06 23:37:59 btw. you can emit PSTAGE_DIR into disk too Oct 06 23:38:48 well, this can be dangerous if you build recipes using package_stagefile_shell iirc Oct 06 23:42:00 yes, is still there Oct 06 23:42:01 http://blog.denix.org/2008/09/getting-even-more-dangerous-with-pstage.html Oct 06 23:42:34 destfile=`echo $srcfile | sed s#${TMPDIR}#${PSTAGE_TMPDIR_STAGE}#` Oct 06 23:44:54 hm Oct 06 23:46:41 I'm not sure what that's supposed to do now-a-days Oct 06 23:46:47 yeah, that can't be safely moved out Oct 06 23:46:49 pstaging is safe with kernels and uboot Oct 06 23:46:51 Tartarus: nobody knows :) Oct 06 23:46:51 it does what it always did Oct 06 23:47:09 guess I'll find out tonight Oct 06 23:47:10 pstage packages are installed relative to TMPDIR Oct 06 23:47:15 so it needs a path relative from TMPDIR to store in pstage Oct 06 23:47:28 stagefile puts it into place so that will happen Oct 06 23:47:39 still there - can't move deploy outside of TMPDIR w/o breaking pstage... Oct 06 23:47:40 if where it needs to be is outside of TMPDIR, you're hosed Oct 06 23:47:50 pstage can't install files outside of its offline root (TMPDIR) Oct 06 23:48:19 Oh that's right, that's where I get confused :) Oct 06 23:48:24 We do pstaging stored outside of tmpdir Oct 06 23:48:30 But deploy inside of tmpdir Oct 06 23:48:32 it's like you're populating an image, and want busybox installed outside of / Oct 06 23:48:34 not gonna happen Oct 06 23:48:36 right Oct 06 23:48:49 can deploy be moved out of TMPDIR, but pstage moved back to TMPDIR from deploy? Oct 06 23:48:54 And yeah, breaking up the structure of tmpdir is unsafe Oct 06 23:49:05 iirc DEPLOY_DIR_PSTAGE is fine, just not PSTAGE_DIR Oct 06 23:49:08 TMPDIR=/mnt/ram; PSTAGE_DIR=${TMPDIR}/../../scratch/oe/pstage Oct 06 23:49:11 pstage can be stored anywhere Oct 06 23:49:11 so moving deploy dir should be fine as of the reorg Oct 06 23:49:11 will that work Oct 06 23:49:14 since it's relocatable Oct 06 23:49:15 what if you built your tmpdir using binds? started off with ramfs with binds for pstage/deploy subdirs to disk fs? Oct 06 23:49:33 again, as far as i know its safe to move DEPLOY_DIR_PSTAGE since the two locations were split Oct 06 23:49:38 it's the whole of deploy that's the problem Oct 06 23:49:53 right, its the files deployed there that need to go into pstage packages that hose you Oct 06 23:50:05 DEPLOY_DIR_PSTAGE is dead :) It's just PSTAGE_DIR Oct 06 23:50:57 oh, right, it uses an internal location for the other files Oct 06 23:51:05 tharvey, tricks like that might be OK Oct 06 23:51:13 since the expected structure exists in some form Oct 06 23:51:19 DEPLOY_DIR_PSTAGE got renamed to PSTAGE_TMPDIR_STAGE, and PSTAGE_DIR has only the packages Oct 06 23:51:25 * kergoth always forgets the variable names Oct 06 23:51:36 kergoth, yeah Oct 06 23:51:39 confusing Oct 06 23:51:57 a bind mount for deploy would be fine, i expect Oct 06 23:52:09 not sure about a symlink, might confuse the classes, not sure Oct 06 23:52:36 PSTAGE_TMPDIR_STAGE and PSTAGE_DIR whats the difference in simple terms Oct 06 23:52:38 Tartarus: the confusion is that we have one variable for two purposes Oct 06 23:52:47 files go both into PSTAGE_DIR and come from PSTAGE_DIR Oct 06 23:52:50 along the lines of DL_DIR Oct 06 23:52:53 so DEPLOY is a misnomer Oct 06 23:53:02 which makes sense, but only if you think about it that way :) Oct 06 23:53:49 Errr Oct 06 23:53:55 Unless I'm way reading the code wrong Oct 06 23:54:20 PSTAGE_TMPDIR_STAGE is "temporary space for mangling a pstage package" Oct 06 23:54:59 PSTAGE_TMPDIR_STAGE is when you are building one recipe Oct 06 23:55:00 I see Oct 06 23:55:13 and the final stage ipks go into PSTAGE_DIR Oct 06 23:55:17 yeah Oct 06 23:55:25 Right Oct 06 23:55:29 ok, we want to keep those Oct 06 23:55:31 so one needs to move PSTAGE_DIR out of ramfs Oct 06 23:55:35 PSTAGE_DIR is what you re-use Oct 06 23:55:44 PSTAGE_DIR is safe to store /where/ever/ Oct 06 23:55:52 that was the whole point Oct 06 23:55:54 :) Oct 06 23:55:56 thanks Oct 06 23:56:14 keep TMPDIR in ramfs Oct 06 23:56:29 point PSTAGE_DIR and DEPLOY_DIR to somewher on disk Oct 06 23:56:38 Almost Oct 06 23:56:52 DEPLOY_DIR isn't safe to set outside of TMPDIR Oct 06 23:56:57 So you have to do bind mount games Oct 06 23:57:13 why Oct 06 23:57:27 That's what the post above is about Oct 06 23:57:48 Stuff expects DEPLOY_DIR as a subdir of TMPDIR Oct 07 00:01:25 why cant we change that sed'ing Oct 07 00:02:19 ya, what he sed ;P Oct 07 00:03:56 Well.. Oct 07 00:04:03 denix, have you confirmed it's still not working recently? Oct 07 00:04:38 I think this should be applied if it still exist Oct 07 00:05:26 I'm not sure Oct 07 00:07:46 Part of the reason I say that is that the underlying code got some changes made about what the variables mean and so on Oct 07 00:07:59 i see Oct 07 00:08:11 ack... can't use amend.inc to amend COMPATIBLE_MACHINE - recipe will be ignored prior to amend Oct 07 00:09:41 Maybe it would be ok, hm Oct 07 00:11:01 khem: pstage is an ipk file that's installed with TMPDIR as its rootfs Oct 07 00:11:13 tell me how you would include a file in it and have it end up outside of TMPDIR Oct 07 00:11:39 it would have to store the path, or relative path, inside the pstage package to supplement the file itself Oct 07 00:11:41 and then move it back Oct 07 00:11:46 and even that would be iffy Oct 07 00:16:50 So yeah, there we go Oct 07 00:16:55 That makes pstaging not relocatable Oct 07 00:19:56 now, deploy is rather a special case, since its files are only outputted, not used anywhere.. you could always deploy them into multiple places, even Oct 07 00:27:16 now, what i'd rather see if deploy dir handled differently Oct 07 00:27:28 have the recipe install deploy files into a subdir of WORKDIR, not a global area Oct 07 00:27:32 then store *that* in pstage Oct 07 00:27:47 and part of installing pstage is restoring files into DEPLOY_DIR specially, rather than using a generic mechanism the way we do now Oct 07 00:28:04 alternatively, the generic mechanism obeys a configuration variable with mappings Oct 07 00:28:23 e.g. deploy_dir:DEPLOY_DIR, and install files in ${WORKDIR}/deploy_dir Oct 07 00:28:26 and itd automatically handle it Oct 07 00:28:50 i think that would be best, a step toward recipes no longer touching areas outside of WORKDIR directly, other than for packaging Oct 07 00:30:10 ideally, they wouldn't even for packaging, and something would suck them out as a part of a rootfs creation or feed population step, imo, but thats way way out, and more controversial Oct 07 00:38:14 * kergoth thinks about how to implement this.. Oct 07 01:53:05 kergoth: have you applied http://patchwork.openembedded.org/patch/2549/ yet ? Oct 07 01:53:29 kergoth: why cant we populate pstage from ipks Oct 07 01:53:54 ideally it would be better if we constructed pstage from them Oct 07 01:54:09 but I might be overstating Oct 07 01:58:57 pstaging is made of ipkgs Oct 07 01:59:19 If we populated sysroot only from ipks then yes, that would mean we could make pstaging easier Oct 07 01:59:30 And no one says doing that is a bad idea :) Oct 07 01:59:34 just a complicated one Oct 07 01:59:44 (less so once legacy staging is dead dead dead really dead) Oct 07 02:19:04 do we have a way to list what uses legacy staging? Oct 07 02:29:00 not easily? :( Oct 07 02:29:36 git grep will do it Oct 07 02:35:41 khem: pb has already proposed populating sysroot entirely from emitted binary packages from do_package* before, but I'm not sure how that would fit into the new model being worked on by RP **** BEGIN LOGGING AT Thu Oct 07 02:53:27 2010 **** ENDING LOGGING AT Thu Oct 07 02:59:56 2010