**** BEGIN LOGGING AT Wed Dec 15 02:59:57 2010 Dec 15 05:51:10 hey, if i have a register on a cpu at 0x80832440 how do i access it from user space, i tried the stuff i saw in the kernel sources, but it all segfaults when i run it Dec 15 05:58:58 gandhijee, you can't access it in userspace directly Dec 15 05:59:30 i tried using /dev/iomem and mapping it Dec 15 05:59:42 but i apparently am not doing it properly Dec 15 06:00:03 i'm unfamiliar with /dev/iomem Dec 15 06:00:11 i mean /dev/mem Dec 15 06:01:25 generally if you want to access hardware, you write a kernel driver Dec 15 06:01:40 and provide an interface to userspace from your driver Dec 15 06:01:57 direct hardware access from userspace is frowned upon Dec 15 06:02:04 i don't think i can write a driver Dec 15 06:02:57 presumably, 0x80832440 is a physical address. Physical addresses don't mean anything to userspace processes anyhow Dec 15 06:03:27 yeah, it is a physical address Dec 15 06:03:48 a process has some memory mapped into a virtual address space. When you use the memory, the OS and the mmu convert it to a physical address Dec 15 06:03:52 i know if its below a certain area, i can use inb/inl etc to get to it Dec 15 06:04:09 maybe if you're running DOS Dec 15 06:04:33 i can do it with serial ports and stuff Dec 15 06:05:00 yes, there are special hacks for that to work. and you have to set it up with ioperm() etc Dec 15 06:05:38 yeah Dec 15 06:05:53 there is no way to do that with addresses above a certain range? Dec 15 06:06:10 what's the machine and what's the hardware controller you're trying to access? Dec 15 06:06:47 its a TS-7260 Dec 15 06:06:56 arm9 cpu/EP9302 Dec 15 06:07:14 what's the register for? Dec 15 06:07:20 CPU serial number Dec 15 06:07:38 it shows up in /proc/cpuinfo, i want to figure out how to read it myself Dec 15 06:08:01 so i don't have to read all of /proc/cpuinfo and parse it all the time Dec 15 06:08:22 this is something that must be done in kernel space Dec 15 06:08:53 o Dec 15 06:09:05 how hard is something like this to do? Dec 15 06:09:16 you could write a very simple driver which created, e.g. /proc/7260-serial and just printed out the value of that register Dec 15 06:09:21 that would be rather simple Dec 15 06:09:30 how would i go about doing something like that? Dec 15 06:09:37 you should read up in "linux device drivers" Dec 15 06:10:30 would it be something that i could figure out fairly quickly from there? or would it require me to read it cover to cover? Dec 15 06:10:57 give me a moment, i'll find an in kernel driver you can look at instead... Dec 15 06:11:22 thank you Dec 15 06:11:53 Hi everybody Dec 15 06:12:13 I was using external-toolchain-csl.bb file placed in meta dir Dec 15 06:12:58 gandhijee, here's a small and simple example. arch/arm/mach-msm/last_radio_log.c Dec 15 06:13:16 I was jus wondring why this line is there in do_install function rm -rf ${D}${bindir}/gdbserver Dec 15 06:13:36 grg: thank you Dec 15 06:13:50 so this is not installing gdbserver .... any specific reason for that Dec 15 06:13:58 oh one more thing i guess, how would i compile that in my OE env? Dec 15 06:15:11 gandhijee, read the first two chapters or so of linux device drivers to learn how to build kernel modules... then the rest should be on the OE wiki or manual Dec 15 06:15:25 k Dec 15 06:19:52 Noor, is gdbserver provided by another package maybe? Dec 15 06:19:53 Can anybody explain to me why we are removing gdbserver from in the do_install function of external-toolchain-csl.bb Dec 15 06:20:23 grg gdbserver has its own recipe Dec 15 06:21:16 so you are saying that gdbserver will be bild through other recipe Dec 15 06:22:03 grg: what if we have a prebuild binary and want to use that ... in that case I probably have to remove this line from external-toolchain recupe Dec 15 06:24:51 Noor, yeah i suppose so Dec 15 06:25:30 and then ensure that the gdbserver recipe doesn't get used Dec 15 06:25:53 yeah Dec 15 06:26:01 i'll add it in assume provided Dec 15 06:32:59 Hi, I've fumbled together a small recipe for QextSerialPort and was wondering if someone could look it over and check that I'm on the right path. http://pastebin.com/cV6ZGJEX Dec 15 06:34:39 It seems to work, but I'm not sure if the PV and SRCREV are done right and I guess it needs to handle both x11 and qte Dec 15 07:43:06 kergoth_: another ping about scmdata=keep patch :) Dec 15 07:43:19 gm Dec 15 07:43:27 wiki seems to have a problem, will mail to list Dec 15 07:43:40 gm eFfeM Dec 15 07:49:12 gm eFfeM_work Dec 15 07:49:26 hi Noor haven't seen you around for a while Dec 15 07:49:56 yeah ... I was busy in some other stuff :) ... now back on OE :) Dec 15 07:50:17 dealing with gdbserver and external toolchain recipe Dec 15 07:50:28 ah cool Dec 15 08:18:10 good morning Dec 15 08:20:51 morning all ! Dec 15 08:29:57 gremlin[it]: hi Dec 15 08:33:04 good morning Dec 15 08:33:20 ericben: hi Dec 15 08:33:43 kergoth: concerning perl-5.8.8, I'm building using distro angstrom 2008.1 and for an armv5 machine Dec 15 08:34:47 kergoth: but now I have more log and before ./miniperl command not found, I have make no rule to build target nedded for "miniperlmain.o" Dec 15 08:34:59 kergoth: so the problem may be somewhere else Dec 15 08:35:09 hi mckoan Dec 15 08:54:22 hm.. interesting Dec 15 08:54:24 morning Dec 15 08:54:33 ERROR: Command execution failed: Exited with 1 Dec 15 08:54:42 this is message from bitbake Dec 15 09:03:12 kergoth: forget it, the problem is during generation of its makefile, perl has a sed script to remove some unwanted paterns which can leak from gcc output ... and my system is in french on this machine which means the world "interne" that I translated as 'internal' is not catched by the script (and in fact is "built-in" in english). What is strange is that using bitbake 4d081a0ed759bd526ab01849d650bd9e8d80ddd1 doesn't exhibit the problem s Dec 15 09:06:20 kergoth: I confirm the LANG problem, using bitbake master and doing a export LANG=en_US.UTF-8 Dec 15 09:07:41 kergoth: before bitbake fixe the problem on my machine. Dec 15 09:13:52 hm.. Dec 15 09:14:17 seems good that testbuilder have LANG=C inside :) Dec 15 09:14:35 good for builds, bad for testing Dec 15 09:15:04 I should move it to config file Dec 15 09:15:12 Jay7: :) Dec 15 09:17:25 very interesting.. kexec-tools-klibc was built fine for angstrom-2010.x but doesn't run compile for angstrom-2008.1 Dec 15 09:17:32 bluelightning: morning Dec 15 09:18:01 bluelightning: at least tosa and akita opie-images for angstrom-2010.x was built ok Dec 15 09:22:17 Jay7: awesome :) Dec 15 09:22:33 perhaps angstrom 2010 should be switched to 1.2.5 now... Dec 15 09:23:42 morning all Dec 15 09:24:05 03Koen Kooi  07org.openembedded.dev * r9a4fe6fc17 10openembedded.git/recipes/ti/libdce_git.bb: Dec 15 09:24:05 libdce: fix DEPENDS Dec 15 09:24:05 Signed-off-by: Koen Kooi Dec 15 09:36:28 Is bitbake broken for oe atm? I always get this error while executing bitbake after 'handling bitbake files' - http://pastebin.com/4KvBE1ns Dec 15 09:37:10 Jordinar: which bitbake are you using ? Dec 15 09:37:23 1.8.18 Dec 15 09:38:14 Jordinar: can you try 1.10.1 ? Dec 15 09:38:30 it should work out of the box Dec 15 09:39:02 i'll try that, thanks. Dec 15 09:54:09 03Koen Kooi  07org.openembedded.dev * r582a20d46f 10openembedded.git/ (5 files in 4 dirs): Dec 15 09:54:09 linux-omap4: switch to L24.11 based tree to match with L24.11 syslink userspace Dec 15 09:54:09 Signed-off-by: Koen Kooi Dec 15 09:55:10 this new bitbake just gives an ImportError (1.10.1 release from berlios) ..... ImportError: cannot import name ui .. hmm. Dec 15 10:03:48 Jordinar: strange Dec 15 10:04:05 1.10.0 seems to be ok Dec 15 10:14:25 * Jay7 sent all issues from testbuilder to ML.. Dec 15 10:23:58 I'm using busybox, and i try to change the date using the command 'date' but i can't do it only as root user. I'd like to be able to set it using www-data user. Permissons on busybox are rwx for all and permissions on date symbolic link too. But i get permission denied. What could be the issue ? Dec 15 10:24:27 i can only do it as root Dec 15 10:25:03 melchior: install sudo and use it Dec 15 10:25:23 date is using syscall that is available only for root imho Dec 15 10:25:40 I'd like to change date from a php script so www-data user Dec 15 10:26:05 Can i set permission for those syscall ? Dec 15 10:26:32 no, this is kernel thing Dec 15 10:26:44 only root can change time Dec 15 10:27:28 so you need some suid binary between www-data and root Dec 15 10:28:01 sudo is easiest way Dec 15 10:29:55 thx Dec 15 10:45:30 anybody knows where is this setting? Dec 15 10:45:32 Starting GPE display manager: gpe-dm Dec 15 10:45:33 Error: Can't identify this machine Dec 15 10:45:58 where must be defined in target fs image? Dec 15 11:00:45 mckoan: isn't that a message from X ? Dec 15 11:00:51 mckoan: interesting indeed Dec 15 11:01:24 03Koen Kooi  07org.openembedded.dev * r006748fc7a 10openembedded.git/recipes/xorg-driver/xf86-video-v4l2_git.bb: Dec 15 11:01:24 xf86-video-v4l2: bump SRCREV Dec 15 11:01:24 Signed-off-by: Koen Kooi Dec 15 11:02:22 mckoan: I'm pretty sure its not one of the GPE bits that creates this message. Dec 15 11:03:33 It might be some xserver startup script - these usually try to identify the machine. Dec 15 11:05:17 florian, ericben: looks like is a memory problem Dec 15 11:05:54 the same fs works on a PM9263 (SAM9), but fils on my PXA270 Dec 15 11:06:09 the diff is 32M vs. 64M sdram Dec 15 11:07:06 mckoan: the difference is also the ... machine ;-) Dec 15 11:07:26 mckoan: hum well... the machin id should be different too Dec 15 11:07:32 ericben: of course, but fb0 works on both Dec 15 11:07:47 and Angstrom splash works on both as well Dec 15 11:07:51 mckoan: but the message you get is not about fb0 but about the machine Dec 15 11:08:12 mckoan: are you using kdrive ? Dec 15 11:08:21 in last days I have been focusing on xinit failure Dec 15 11:09:11 ericben: angstrom-2008 or kaeilos, so both xorg Dec 15 11:09:39 http://pastebin.com/cJU14Zrq Dec 15 11:10:06 I could give it a try here... I have a Mainstone II which should be a good reference for the PXA270 :-) Dec 15 11:10:32 mckoan: what about "Xorg -verbose 6 -keeptty" ? Dec 15 11:10:43 mckoan: here Xorg works on spitz Dec 15 11:10:56 JaMa: Xorg is not present, there is Xfbdev Dec 15 11:11:39 mckoan: from memory you need to have your mode included into hw/kdrive/src/kmode.c Dec 15 11:12:02 florian: I also have a Mainstone Rev.E, but ethernet fails Dec 15 11:12:14 check into xserver-kdrive/xserver-kdrive for this Dec 15 11:12:29 ericben: not sure I am using kdrive Dec 15 11:12:55 mckoan: should be easy to work around using a pcmcia ethernet card Dec 15 11:13:46 florian: you're right! thanx Dec 15 11:14:00 * mckoan is testing on mainstone with USB-eth adapter Dec 15 11:15:00 let me guess - xserver-common 1.25? Dec 15 11:15:15 this should work as well, but since pcmcia isn't used that much any more and if you have an old non-cardbuc network adapter laying around that's a good choice Dec 15 11:15:26 * mckoan after lunch (and going to take daughter at school) Dec 15 11:15:48 florian: which xserver-common is now used in OE? 1.25 or 1.3x? Dec 15 11:16:07 let me check Dec 15 11:16:20 * florian always uses 1.3x Dec 15 11:16:37 if it's xserver-common, the machine as to be added as in xserver-common/files/Xserver-at91.patch Dec 15 11:17:07 1.33 seems to be the most common one Dec 15 11:17:19 hrw: angstrom-2008-preferred-versions.inc says xserver-common = "1.33" Dec 15 11:17:52 uf Dec 15 11:18:35 main symptom is the fs generated for Akita works smoothly with AT91 Dec 15 11:19:02 so diffs are: kernel and hw (ram?) Dec 15 11:19:37 splash and direct fb0 write work on both Dec 15 11:19:54 so I suspect the heavy of xorg Dec 15 11:20:05 s/heavy/wheight Dec 15 11:20:49 if you have Xfbdev rather than Xorg, you are not using xorg Dec 15 11:21:14 and, in any case, lack of ram wouldn't typically cause the error you're seeing. that does look very much like the machine value from /proc/cpuinfo just isn't recognised by the Xserver startup script. Dec 15 11:21:20 mckoan: I think the answer is a few lines upper ... Dec 15 11:21:22 mckoan: awk 'BEGIN { FS=": " } /Hardware/ { gsub(" ", "_", $2); print tolower($2) } ' mckoan: looks like some part of zaurusd or something like Dec 15 11:25:15 zauruses have special scripts to detect model and setup keymap/do other things Dec 15 11:25:38 JaMa: marvell/intel_pxa270 Dec 15 11:26:03 JaMa: ok, so I won't mix xorg and gpm problems Dec 15 11:26:23 JaMa: I focuso on xinit (Xfbdev) Dec 15 11:26:30 that's what is in /etc/init.d/functions as machine_id laster used in startup script Dec 15 11:26:51 thank you all Dec 15 11:27:00 * mckoan -> bb in 1,5 h Dec 15 11:27:16 AFAIK xinit starts /etc/X11/Xserver which checks machine_id Dec 15 11:28:19 so if you want to focus on Xfbdev then start it with Xfbdev directly not xinit/Xserver script Dec 15 11:28:26 JaMa: but mckoan runs akita image which probably still have zaurusd which is argh Dec 15 11:29:38 patches for xserver-common are funny Dec 15 11:30:50 when I did 1.30 I added /etc/default/xserver support which removes need for patching /etc/X11/xserver-common file Dec 15 11:32:08 * florian makes use of this feature quite frequently Dec 15 11:34:34 hrw: what's "argh" about zaurusd Dec 15 11:34:36 ? Dec 15 11:39:58 bluelightning: tries to do too much at once. alsa profiles, touchscreen keys, rotations Dec 15 11:40:21 hrw: ah yes Dec 15 12:44:03 hello Dec 15 12:44:36 Where I can find 2.6.34 config for Davinci ? Dec 15 12:44:51 Hello ... Dec 15 12:45:04 kergoth_: can you take a look on the class I just send to ml? Dec 15 12:45:09 kergoth_: gitpv Dec 15 12:45:47 stanv first you need the kernel which you are compiling Dec 15 12:46:01 There is only 2.6.32-rc2 config at http://arago-project.org/git / arago.git / linux / linux-davinci-staging / dm365-evm / configs / Dec 15 12:46:12 I have kernel sources Dec 15 12:46:13 stanv or a find in recipes/linux Dec 15 12:46:43 I need config for 2.6.34 Dec 15 12:47:46 ? Dec 15 12:47:59 so you have the sources without a recipe Dec 15 12:48:05 yes Dec 15 12:48:46 I don't use arago/official SDK.... Dec 15 12:49:18 okay Dec 15 12:49:29 mckoan, not sure if this is related to what you are doing, but a friend noticed a path to x86 sysroot in startx the other night Dec 15 12:49:30 frist you are using a overlay of oe Dec 15 12:49:31 I need only kernel config :) Dec 15 12:49:46 second you are using not even official one Dec 15 12:49:58 so why you come here and ask fior help Dec 15 12:50:25 :))) I need official config :) Dec 15 12:50:46 what official config? Dec 15 12:50:53 there is no official config Dec 15 12:50:55 make your own Dec 15 12:50:58 with menuconfig Dec 15 12:51:05 At http://processors.wiki.ti.com/index.php/FAQ_for_DaVinci_Linux noticed config Linux-2.6.23-rc3-davinci1.... Dec 15 12:51:26 2.6.23 is stoneage Dec 15 12:51:38 maybee you mean 2.6.32 Dec 15 12:51:52 ahh... Dec 15 12:51:54 :(( Dec 15 12:53:32 as I said make a find in recipes/linux Dec 15 12:55:37 Git repo has v2.6.34-davinci1 .... someone should have config for this version.. Dec 15 12:55:52 okay Dec 15 12:55:58 so seem to trolling more Dec 15 12:56:03 please read the topic Dec 15 12:56:12 no more help from me Dec 15 12:57:04 :)) okey,,, sorry Dec 15 12:57:06 :)) Dec 15 12:57:25 recipes has only config for 2.6.32 Dec 15 12:58:34 re Dec 15 12:58:42 jo mckoan Dec 15 13:01:10 Crofton|work: yes mcookie=`/home/koan/devel/build/angstrom/tmp/sysroots/x86_64-linux/usr/bin/mcookie` Dec 15 13:02:07 thanks for confirming that Dec 15 14:02:27 03Koen Kooi  07org.openembedded.dev * r99e5d05f9d 10openembedded.git/conf/distro/include/angstrom-2010-preferred-versions.inc: Dec 15 14:02:27 angstrom next: don't set preferred version for xserver-xorg, but trust the includes Dec 15 14:02:27 Signed-off-by: Koen Kooi Dec 15 14:03:00 florian: if I use mainstone with USB-eth adapter it is eth1 and won't bott NFS Dec 15 14:03:10 03Koen Kooi  07org.openembedded.dev * r1f24b55ee2 10openembedded.git/recipes/xorg-driver/xorg-driver-common.inc: Dec 15 14:03:10 xorg-drivers: bump INC_PR for API/ABI change in previous commit Dec 15 14:03:10 Signed-off-by: Koen Kooi Dec 15 14:04:31 mckoan: hmm... maybe it takes its time to appear. i wonder if rootwait would have an effect on nfs Dec 15 14:26:12 florian: NFS takes the first eth available (eth0), do you know how to specify eth1 ? Dec 15 14:26:32 mckoan: not offhand Dec 15 14:27:26 mckoan cmdline or directly in the initramfs Dec 15 14:30:28 maybe NFSROOT=w.x.y.z:/nfs ETHERNET=eth1 Dec 15 14:45:11 fidencio: ping Dec 15 14:48:57 florian: Documentation/filesystems/nfsroot.txt ;-) Dec 15 15:00:09 morning Dec 15 15:00:38 otavio: will do Dec 15 15:01:04 ericben: interesting, will look into it Dec 15 15:14:02 okay, github needs a way for me to say 'yes, i'm interested in this repository, but no, i don't want to see every push/ticket/etc, do a digest or filter out tickets or something' Dec 15 15:14:21 heh Dec 15 15:14:35 in the feed? Dec 15 15:14:40 yeah, the dashboard display Dec 15 15:16:51 kergoth: I think I can put pastebin support into goggle fairly easily (as soon as I manage to wrap my head around interacting with the TreeView some more) Dec 15 15:17:01 so for logs, you can right click -> pastbin it Dec 15 15:17:05 and it'll give you back the url Dec 15 15:19:39 nice, i'm sure people would find that handy Dec 15 15:20:02 foerster: good feature :-) Dec 15 15:20:22 ericben: yea, thanks for the suggestion. Dec 15 15:20:54 just need to figure out how the hell to get the treemodel item from a click event Dec 15 15:21:45 * kergoth_ decided to start over on the event thing, integrate it slowly -- slowly replace bb.event usage with self.event usage, and redirect self.event to bb.event temporarily, making sure nothing vanishes along the way Dec 15 15:22:13 kergoth: you had a commit which got rid of logs which were dups of events - where did you push that? Dec 15 15:22:43 * kergoth_ scratches head Dec 15 15:22:47 i have too many branches Dec 15 15:22:58 eh, I can search for it, no biggie. Dec 15 15:23:01 hmm Dec 15 15:23:03 simply cut the tree :-) Dec 15 15:23:31 but as it stands, right now, I get the event which gets put into the right task, then the log message shows at the top of goggle b/c I can't associate it to anything. Dec 15 15:23:39 ah Dec 15 15:23:40 having only the event would be nicer from the UI standpoint. Dec 15 15:24:12 part of it is in master Dec 15 15:24:14 ab831e867f09b47001cb8da2f8f060e04febf237 Dec 15 15:24:21 shifts the runqueue task start/failed messages to the ui Dec 15 15:24:26 there are probably other instances of duplication though Dec 15 15:24:51 ok, well that's the most noticeable (in my contrived cases where I'm making things fail on purpose) Dec 15 15:25:03 i'll just rebase on master again at some point Dec 15 15:25:28 * kergoth_ nods Dec 15 15:29:56 [18:16] and it'll give you back the url Dec 15 15:30:09 Jay7: yes Dec 15 15:30:09 foerster: and copy url to clipboard :) Dec 15 15:30:14 if pygtk can do that Dec 15 15:30:27 * foerster doesn't know enough pygtk to say definitively, but I'd imagine so Dec 15 15:30:48 foerster: nice idea! You'll probably want listen for gtk.TreeViewColumn's clicked signal and use get_tree_view() and then get_model() on that Dec 15 15:31:17 incandescant: Perhaps that's what I need. I was capturing the button press event on the treeview itself. Dec 15 15:31:21 perhaps I'm a level too far up. Dec 15 15:31:44 foerster: if that works, go for it ;-) Dec 15 15:31:51 * incandescant is a treeview noob Dec 15 15:32:01 me too Dec 15 15:40:50 kergoth_: ok; nice. Dec 15 15:41:49 ericben: got it. Dec 15 15:41:58 erm, maybe not Dec 15 15:41:59 hmm Dec 15 15:42:19 ericben: oh, yes, i found it Dec 15 15:43:34 03Chris Larson  07master * rbebdb554b5 10bitbake.git/lib/bb/build.py: Dec 15 15:43:35 build: fix copy/paste thinko regarding LC_ALL Dec 15 15:43:35 Signed-off-by: Chris Larson Dec 15 15:43:41 ericben: there, that'll fix your perl issue -- sorry about that, and thanks for the report and investigation, i doubt i would have spotted that anytime soon Dec 15 15:44:35 somehow i managed to screw up the new environment by putting LANG instead of LC_ALL Dec 15 15:44:39 no idea how i managed that one Dec 15 15:51:30 i wonder how much https://github.com/kergoth/bitbake/compare/runqueue-cleanup slows down the runqueue generation for completion, due to doing multiple sorts -- i know sorted is stable, but still. Dec 15 15:52:21 * kergoth_ tests Dec 15 16:11:02 Anyone have any opinions on https://github.com/kergoth/openembedded/commit/c278cd7 ? Dec 15 16:18:37 https://github.com/kergoth/openembedded/compare/master...typing shows more examples of usage Dec 15 16:19:45 so the typing is done in the class files Dec 15 16:20:11 well, its just like any other piece of metadata, just an extra flag or flags marking up the variable Dec 15 16:20:23 where you put it is up to you Dec 15 16:20:38 but it seem slike the best erturns are in the class files Dec 15 16:21:19 as long as its done somewhere definitive, so it doesn't conflict, that's the key. ideally next to the default value of the variable Dec 15 16:21:55 * kergoth_ composes mail to list about it Dec 15 16:22:57 hi, little question here - is it possible to override TOOLCHAIN_TYPE for a single recipe without changing it for the distro as a whole? Dec 15 16:23:17 FOO_pn-somerecipe = "bar" Dec 15 16:23:23 will set FOO = "bar" in somerecipe Dec 15 16:23:54 so that would be done in the conf file, not the recipe itself? Dec 15 16:24:16 that's always an option, i assumed you weren't the maintainer of the recipe Dec 15 16:24:24 * kergoth_ also doesn't know what TOOLCHAIN_TYPE does, so doesn't know the implications :) Dec 15 16:26:33 hrm Dec 15 16:26:41 aha, makes sense. So really just an organizational decision for conf vs recipe since it is our own recipe. Thanks. TOOLCHAIN_TYPE is internal/external toolchain related to using cmake in the build Dec 15 16:26:53 ah Dec 15 16:28:25 damnit. Dec 15 16:29:00 right click pastebin of log works -- except the newlines are displayed as \n (b/c of liburl.urlencode i assume) Dec 15 16:29:58 s/urllib/liburl/ Dec 15 16:30:28 * kergoth_ chuckles Dec 15 16:31:06 oh well. will have to come back to it later. Gotta get some work done. Dec 15 16:33:29 damnit, i keep forgetting that i changed my filters to auto-archive [PATCH] mails, so i need to review them explicitly Dec 15 16:35:04 hi kergoth Dec 15 16:35:30 kergoth should ask tom about how mentor stand to yocto? Dec 15 16:36:14 tom is a coworker, and i was the one at the summit -- if anything, we'd have to ask our boss what can be discussed publically Dec 15 16:36:35 okay Dec 15 16:36:50 its my only problem at the moment Dec 15 16:36:57 in the whole discussion Dec 15 16:37:12 would be nice to hear something about from mentor Dec 15 16:37:34 i can say that the responses all around were positive at the summit, including a variety of companies there Dec 15 16:37:47 seems ti hasnt problems at all Dec 15 16:37:49 okay Dec 15 16:37:49 beyond that, i'm sure yocto or any companies joining it will want to do some sort of press release or statement of support or something Dec 15 16:39:02 yeah my concern was/is that wind river is some competitor of yours Dec 15 16:40:17 yocto is vendor neutral. currently the main members are intel and wind river, but that's incidental Dec 15 16:40:46 yes Dec 15 16:40:54 damnit Dec 15 16:40:59 kergoth_: do you know how is the process of joining yocto? Dec 15 16:41:01 okay time to play with my son Dec 15 16:41:01 i hate when i send a mail to the list from the wrong address Dec 15 16:41:17 till later Dec 15 16:41:44 otavio: as mentioned in the threads, members need to contribute one developer to the effort, and there might be a financial piece, not sure, and of course they then get to put someone in the SC Dec 15 16:47:08 http://www.linuxfoundation.org/news-media/announcements/2010/12/mentor-graphics-joins-linux-foundation Dec 15 16:47:28 :) Dec 15 16:48:02 have a nice rest of the day Dec 15 16:48:07 gn Dec 15 16:52:10 it will be interesting to see what solutions are proposed for small companies to join Yocto (when we are only 3 peoples it's hard to delegate one ;-) Dec 15 16:53:10 They want you to be a member of the Linux Foundation Dec 15 16:53:16 * Crofton|work has been reading Dec 15 16:53:27 that is fee based it would appear Dec 15 16:53:34 i GUESS SOMEONE HAS TO PAY rp Dec 15 16:53:37 ooops Dec 15 16:53:45 stupid caps lock :) Dec 15 16:54:27 in its email he said "and can be for others (e.g. small business representative?)" Dec 15 16:55:01 which one? Dec 15 16:55:05 email? Dec 15 16:55:45 emails of Richard Purdie, sent @ 10:52 GMT Dec 15 16:56:07 Silver membership to Linux foundation is 5kUSD to 20k Dec 15 16:56:22 gold 100k and platinium 500k Dec 15 16:56:48 the minimal ticket of 5k is quite high for small companies Dec 15 16:56:56 it is :) Dec 15 16:56:58 http://www.linuxfoundation.org/about/join Dec 15 16:57:10 yeah Dec 15 16:57:16 let's create some linux punk foundation Dec 15 16:57:31 I should join an industry group, but it is 2K for me Dec 15 16:57:39 maybe LF has a brass membership? Dec 15 16:57:51 maybe small companies can make a sort of partnership to have a collective membership Dec 15 16:57:57 what is a "brass" ? Dec 15 16:58:20 ok coper Dec 15 16:58:37 sort of Dec 15 16:58:54 :) Dec 15 16:58:57 yeah, it's an alloy of copper. Dec 15 16:59:05 I was under the impression the fee was negotiable for smaller companies? Dec 15 16:59:14 translation says yellow copper the one for the coins, thanks Dec 15 16:59:15 ok, look s like the SC can waive LF requirements Dec 15 16:59:17 usually the term for the level below silver is "bronze", which is a slightly different alloy. Dec 15 16:59:34 * Crofton|work is being funny (and failing as usual) Dec 15 16:59:51 yes a bronzz membership @ 1 to 2k would be great for small companies Dec 15 17:00:05 bronze not bronzz ;) Dec 15 17:00:31 or small companies sponsor OE which is the member of the fundation and representative of them Dec 15 17:03:09 while we can use the tool and provide overlays or even patches for upstream integration this is not bad Dec 15 17:09:45 have a nice rest of day Dec 15 17:18:59 woot, right click pastebin is working in goggle. Dec 15 17:19:59 kergoth_: something i've noticed - if I clean a package, and then build - when it fails, I get a logfile. If I just run bitbake mypackage again, it still fails (as it should), but I don't get a log. Dec 15 17:20:03 any idea why that happens? Dec 15 17:20:34 hmm, no, the only time it should not show the log file is if the log file was empty -- try running with -DD and it'll show a message if it removes the logfile for that reason Dec 15 17:20:54 * kergoth_ thinks we should probably make it clearer Dec 15 17:21:06 just wondering why it would be empty? Dec 15 17:21:28 that i don't know. i assume the task its running really does output something? Dec 15 17:21:34 i purposely have a package failing b/c i removed a dependency from the recipe (and cleaned the depencency). Dec 15 17:21:40 It fails in do_compile Dec 15 17:21:46 b/c a dependent library cannot be linked Dec 15 17:22:06 can always bisect and see if the bb.build changes broke the behavior Dec 15 17:22:14 (which is likely :\) Dec 15 17:22:18 :) Dec 15 17:22:34 * kergoth_ dreams of bitbake unit tests Dec 15 17:22:55 much rewriting to get there Dec 15 17:23:05 re Dec 15 17:24:00 even in relatively isolated areas like bb.build itd be difficult. can't easily mock the Popen, would have to do all tests with real executed scripts.. necessary, of course, but itd be nice to confirm the sane behavior of the api functions first Dec 15 17:24:32 oh, i should do some unit tests for bb.event, though, when i finish it Dec 15 17:24:37 that should be easy to test Dec 15 17:25:19 hm. I see the message with logfile in knotty, so goggle must be doing something funny Dec 15 17:26:09 huh Dec 15 17:26:14 strange Dec 15 17:26:37 hmm Dec 15 17:33:06 holy hell, -DDD -u goggle :) Dec 15 17:34:43 kergoth_: ok, wtf. -DDD makes the log appear. Without it, no log. The test for os.file.exists fails. Dec 15 17:34:48 for knotty and goggle, it's the same Dec 15 17:39:37 foerster: wtf Dec 15 17:39:38 damnit Dec 15 17:39:42 well Dec 15 17:39:50 i see you fixed some things like logTee Dec 15 17:39:53 could that affect it? Dec 15 17:40:01 i haven't rebased on master since last week some time Dec 15 17:40:10 probably Dec 15 17:40:28 turns out if you inherit from 'file', some python things bypass the write() method and write directly to the fd Dec 15 17:40:37 which is clearly problematic for things like tee Dec 15 17:40:52 ok. I'll rebase later before I go any further. Dec 15 17:49:42 from within a devshell, there's no way to call do_* in a recipe, right? Dec 15 17:49:44 oh hai stefan_schmidt Dec 15 17:49:50 I'd have asked you ;D Dec 15 17:50:16 hi jconnolly Dec 15 17:50:25 hi there woglinde Dec 15 17:56:17 jconnolly: http://wiki.openembedded.org/index.php/Devshell#Background Dec 15 17:56:19 hi all Dec 15 17:56:43 * jconnolly just got rtfm'd. nice! ;D Dec 15 17:56:58 an now read it! Dec 15 17:57:00 and Dec 15 17:57:06 s/an/and/ Dec 15 18:02:04 hi stefan_schmidt Dec 15 18:02:21 hi woglinde Dec 15 18:09:59 okay, bitbake *really* needs something like -e but which accepts specific variables to show Dec 15 18:10:06 i'm sick and tired of screwing around with -e + grep Dec 15 18:15:44 03Chris Larson  07master * rb8d240902f 10openembedded.git/lib/oe/utils.py: Dec 15 18:15:44 oe.utils: add 'inherits' function Dec 15 18:15:44 This is simply a more convenient form of bb.data.inherits_class, as it can Dec 15 18:15:44 accept any number of classes, and will return True if it inherits any of them. Dec 15 18:15:44 This is similar to how isinstance() behaves. Dec 15 18:15:45 Signed-off-by: Chris Larson Dec 15 18:19:42 gm Dec 15 18:20:04 hi eFfeM Dec 15 18:20:10 hi effem Dec 15 18:30:45 kergoth_, seems that bitbake -k foo never returns (after what looks like a possible exception loop) if nothing provides foo (master) Dec 15 18:35:39 So, should libfsobasics-1_0.9.10+gitr0+09dd3ede231106dc2535145291c076b085db3018-r2.0 work on uclibc? Dec 15 18:35:54 | CC utilities.lo Dec 15 18:35:55 | utilities.c:35:22: fatal error: execinfo.h: No such file or directory Dec 15 18:35:57 is how it dies Dec 15 18:36:34 no Dec 15 18:36:49 you have to workaround execinfo.h stuff Dec 15 18:36:52 but its easy Dec 15 18:36:54 anything that needs stack backtrace from libc it wont work on uclibc Dec 15 18:37:12 hi khem Dec 15 18:37:17 Well, workarounds in autobuilders aren't easy Dec 15 18:37:23 hello Dec 15 18:37:24 So can we get a patch applied? Dec 15 18:37:24 ? Dec 15 18:37:34 you have to patch it Dec 15 18:37:43 Let me rephrase :) Dec 15 18:37:51 Can someone that knows what to do please patch it and push it? Dec 15 18:41:12 Tartarus: just saw your 2nd mail on testing 2010-12-10, I also used minimal, console-image Dec 15 18:41:29 targets calamari, mpc8313e, sheevaplug and neek Dec 15 18:42:09 that's fairly crazy Dec 15 18:42:31 * Tartarus pokes tinderbox Dec 15 18:44:46 tartarus check http://tinderbox.openembedded.net/builders/eFfeM/ Dec 15 18:44:47 OK, guess it was a different part that failed there Dec 15 18:45:01 http://tinderbox.openembedded.net/builders/mentor-graphics-autobuilder/ Dec 15 18:45:13 it are the builds from dec 14 and the last (top) 2 from dec 13 Dec 15 18:45:29 Tartarus frist wee need to know what is used from execinfo.h Dec 15 18:45:52 woglinde, good question. I just autobuild that stuff ;) Dec 15 18:47:22 whats the git repo? Dec 15 18:47:26 fso has to much Dec 15 18:48:51 tharvey: interesting Dec 15 18:50:44 hm ugh Dec 15 18:50:47 its vala code Dec 15 18:50:59 sorry dont know how to patch it Dec 15 18:51:00 03Stefan Schmidt  07master * r9a6abcdca5 10openembedded.git/recipes/linux/linux-omap-hah_2.6.31.bb: linux-omap-hah_2.6.31: Bump SRCREV for machid changes Dec 15 18:51:12 Tartarus try to ask mickeyl about it Dec 15 18:51:57 k Dec 15 18:53:57 bah, how do you get the failed logs out of tinderbox? Dec 15 18:54:40 paste the link? Dec 15 18:55:02 http://tinderbox.openembedded.net/packages/1286556/ Dec 15 18:56:44 reconfigure tinderbox Dec 15 18:57:32 on OE tinderbox I can keep klicking on failed and then eventually I end up with http://tinderbox.openembedded.net/packages/1282258/ Dec 15 18:57:36 thre is a link log Dec 15 18:57:42 not too easy to see Dec 15 18:58:02 I don't see it on your box and I cannot click on failed Dec 15 18:59:00 ok, is there something else I should have done outside of the normal stuff on the Testing wiki pages? Dec 15 19:01:53 for tinderbox ? no idea. for testing, no, I started clean, so did a git pull of the testing branch with an empty tmp dir etec Dec 15 19:03:25 right, ok, I don't get it then :) Dec 15 19:03:43 * Tartarus will enable local logging of the log files next time Dec 15 19:03:56 stupid Q, but why are you building autoconf, afaik it is not needed, I do not have the testing build handy but for my own image on git head I only have autoconf-native, no autoconf Dec 15 19:04:39 autoconf is only needed if you want to run compiler and make etc on target, not for the host, host is autoconf-native and that one seems to buid fine for you Dec 15 19:04:42 afk for 20- mins Dec 15 19:05:28 native-sdk-image Dec 15 19:21:36 khem: 4ba7aa07c0dcd28f94515ff9927e2a04403fcf15 isn't enough in all cases, minimal+qemuarm still fails Dec 15 19:31:29 Tartarus: hmmm is it using latest gcc too ? Dec 15 19:31:40 Tartarus: I tested it using qemuarm/minimal Dec 15 19:31:46 so its strange Dec 15 19:31:58 it depends on the gcc fix as well Dec 15 19:33:57 khem: top of tree as of a few hours ago Dec 15 19:34:08 I can see in the log it's still doing -mthumb Dec 15 19:34:17 hmm Dec 15 19:34:22 weird Dec 15 19:34:36 btw. is it using same samba Dec 15 19:34:38 version Dec 15 19:34:45 I did not fix it for all sambas Dec 15 19:35:07 Ah, could be the problem then Dec 15 19:35:16 samba_3.2.15.bb is the one I fixed Dec 15 19:35:25 And that's what I tried Dec 15 19:35:26 so, odd Dec 15 19:35:37 k Dec 15 19:35:41 I can retry Dec 15 19:35:57 if I see it I can try to fix it Dec 15 19:37:42 k Dec 15 19:54:35 JaMa: hi Dec 15 19:55:00 hi Dec 15 19:55:20 JaMa: I am sorry but I fail to understand all this for SRCPV, PKGV, PV, LOCALCOUNT and like Dec 15 19:55:44 JaMa: is it documented ? I read the thread you pointed to me but it is confusing Dec 15 19:55:45 have you read that thread about gitpkgv? Dec 15 19:55:52 ah :/ Dec 15 19:56:34 ok, I'll try to explain in short (but as I'm not native speaker) it could end long and confusing again :/ Dec 15 19:57:22 SRCPV for git is like LOCALCOUNT+SRCREV Dec 15 19:57:33 what is localcount? heh Dec 15 19:57:48 so it's similar to GITVER Dec 15 19:58:00 from gitver.bbclass Dec 15 19:58:30 Bitbake says: "ImportError: cannot import name msg" Dec 15 19:58:58 I know this was discussed some time ago (weeks? months?), and I thought it was fixed, but a quick google finds no hints. Dec 15 19:59:12 1.8.18 fails this way now, 1.10.1 works fine. Dec 15 19:59:17 JaMa: like an auto-changed PR? Dec 15 19:59:39 imho good time to say that 1.8 is not supported anymore Dec 15 19:59:45 I can switch to 1.10, but AFAIK 1.8 is still the supported bitbake version according to the OE web site. Dec 15 20:00:01 and python 2.5 as well.. Dec 15 20:00:03 otavio: if you have BB_GIT_CLONE_FOR_SRCREV then bitbake fetcher will count how many commits are before SRCREV (with git rev-list like you do) Dec 15 20:00:23 I've enough challenges to keep up with OE; not sure I want to go bleeding edge with bitbake as well! Dec 15 20:00:28 1.8 should be working today, yes, if it doesn't, its a bug -- but yes, we should bump the bitbake to 1.10 Dec 15 20:00:32 1.10 is nowhere near bleeding edge Dec 15 20:00:39 Ah, ok. Dec 15 20:00:40 we're releasing 1.11/1.12 in the next 2 weeks Dec 15 20:00:56 otavio: but without BB_GIT_CLONE_FOR_SRCREV you have to keep last revision and last LOCALCOUNT in local persitent cache like tmp//cache/om-gta02/bb_persist_data.sqlite3 Dec 15 20:01:16 What would you recommend for 1.10 -- 1.10.1 or is there a better version? I'll just switch. Dec 15 20:01:32 otavio: and LOCALCOUNT is auto incremented every time requested SRCREV is different than last one in that cache Dec 15 20:01:32 khem, btw samba buidls fine here with the patches you pushed, thanks alot Dec 15 20:01:53 mwester: git master have parallel parsing feature Dec 15 20:01:58 this is very useful thing.. Dec 15 20:02:20 otavio: but this behavior is by default disabled in bitbake.conf by BB_LOCALCOUNT_OVERRIDE = "1" Dec 15 20:03:09 Hmmm... that would be helpful for the autobuilder especially, since it often spends far more time parsing than actually doing any builds. But that would be beyond 1.10.x, no? Dec 15 20:03:36 1.11 Dec 15 20:03:41 or 1.12 Dec 15 20:03:58 as kergoth_ will decide :) Dec 15 20:04:09 it shouldn't be merged back to 1.10 Dec 15 20:04:59 mwester: do you have 1.8.18 from tarball or latest from 1.8 branch? Dec 15 20:05:01 JaMa: right Dec 15 20:05:11 JaMa: i just fail to see why to use it Dec 15 20:05:18 otavio: SRCPV? Dec 15 20:06:00 otavio: because that's the sortable revision you want for PV Dec 15 20:06:29 JaMa: LOCALCOUNT Dec 15 20:06:29 otavio: but provided by bitbake fetcher (not by clever .bbclass) Dec 15 20:06:56 JaMa: but what SRCPV is? is it the LOCALCOUNT used? Dec 15 20:07:25 otavio: ah, the problem is that with BB_GIT_CLONE_FOR_SRCREV you need to clone whole repo first, right? Dec 15 20:08:21 otavio: so to have sortable revision (based on git rev-list result) you have to clone every repo before evaluating SRCPV Dec 15 20:09:24 only open issue that I seem to have building is the gettext one, see http://thread.gmane.org/gmane.comp.handhelds.openembedded/40487 Dec 15 20:09:27 otavio: so LOCALCOUNT is work arround to prepare sortable revision without cloning all repositories to call "git rev-list" Dec 15 20:09:31 anyone an idea on this ???? Dec 15 20:11:29 JaMa: or you can do it on src dir Dec 15 20:11:39 JaMa: as I do on the gitpv Dec 15 20:12:13 otavio: but how can you expand PV before you even have that src dir? Dec 15 20:12:40 otavio: ie on new host Dec 15 20:15:17 JaMa: since you ask the srcrev for bitbake it clone it at first time only Dec 15 20:15:37 JaMa: but this is required since my SRC_URI uses git:// so it uses the git fetcher Dec 15 20:15:48 JaMa, sorry for the delay -- had to check. The build with the errors is from the repo, not a tarball. Dec 15 20:16:11 otavio: yes but while parsing _all_ recipes it will expand PV Dec 15 20:16:18 It's not the latest either, it's the 1.8.18 tag. Dec 15 20:16:33 otavio: so it will need to know PV not only for recipe you're going to build now, but all in BBPATH Dec 15 20:17:35 otavio: that's why people with BB_GIT_CLONE_FOR_SRCREV were angry when git recipes with SRCPV (which were never git cloned before) started to require working git clone to parse correctly Dec 15 20:18:00 JaMa: so you mean it would need to clone all them Dec 15 20:18:11 otavio: yes Dec 15 20:18:47 JaMa: I sort of see now Dec 15 20:18:59 JaMa: I am still lost at the difference between PKGV and PV Dec 15 20:19:07 JaMa: mind to clear it out for me? Dec 15 20:19:07 eFfeM: I've affected too Dec 15 20:19:20 otavio: what's great on gitpkgv.bbclass is that it defers this git rev-list for later (only for PKGV) while PV (for parsing) is known without git clone Dec 15 20:20:00 otavio: PV is version for bitbake (to decide which recipe is newest for you) Dec 15 20:20:29 otavio: PKGV is usually the same as PV, but it's used in resulting package (.ipk, .dep) so for package manager on target Dec 15 20:20:35 JaMa: BUT the revision for GIT is not sortable Dec 15 20:20:40 eFfeM: isn't this help? http://article.gmane.org/gmane.comp.handhelds.openembedded/40557 Dec 15 20:20:46 JaMa: so bitbake will think it is older, no? Dec 15 20:21:37 otavio: if you have abc_1.0+gitr${SRCPV} and abc_2.0+gitr${SRCPV} then the second is always higer version, right? Dec 15 20:21:57 JaMa: yes BUT Dec 15 20:22:12 JaMa: SRCREV can be a1 or 1a Dec 15 20:22:15 ... Dec 15 20:22:22 JaMa: since the hash is not sorted Dec 15 20:22:37 otavio: and clever hack (imho first proposed by pb) is that even with the same abc_2.0+gitr${SRCPV}, bitbake will rebuild this recipe when PKGV is changed Dec 15 20:23:29 JaMa: but this does not ensure it will be "newer" on opkg partsingo Dec 15 20:23:33 otavio: "1.0" and "2.0" should be enough to choose with recipe is newer for bitbake Dec 15 20:23:37 JaMa: since it can generate and older version Dec 15 20:23:58 that's why you need that GITPKGV used in PKGV Dec 15 20:24:18 otavio: because GITPKGV provides sortable revision Dec 15 20:24:32 JaMa: but PKGV is the one used on the ipk filename? Dec 15 20:24:41 otavio: similar like your gitpv but without leading tag_name Dec 15 20:25:03 JaMa: if that is the case I could fix gitpkgv to provide the tag Dec 15 20:25:37 Jay7: will dig into it, got sidetracked in another issue Dec 15 20:25:43 otavio: yes Dec 15 20:25:56 Jay7: actually I was using libtool 2.4 (if that makes a difference) Dec 15 20:26:20 eFfeM: when you say samba worked what was your MACHINE and DISTRO Dec 15 20:26:47 JaMa: and how GITPKGV will "discover" new revisions if used with AUTOREV? Dec 15 20:28:01 mwester: sorry, I haven't used 1.8 for long time but noticed some commit about bb.msg lately in 1.8 branch, but after checking it, it won't probably help in your case Dec 15 20:28:33 Ok, no worries. I shall switch to 1.10 or the next release. Dec 15 20:29:06 otavio: yes it should according to pb (and IIRC I've tried it and it really worked) Dec 15 20:30:33 khem: minimal/sheevaplug, note that I just verified that it builds, I have not yet run the package Dec 15 20:34:59 eFfeM: sheevaplug ok Dec 15 20:36:22 which is armv5te but qemuarm is armv4te Dec 15 20:36:24 hmmm Dec 15 20:36:26 anyway Dec 15 20:37:27 well both build for thumb so it should have fixed qemuarm too Dec 15 20:37:34 * khem is bewildered Dec 15 20:51:46 khem, haven't done an armv4te build of samba recently, I can try to kick off one Dec 15 20:51:57 ka6sox: are you aware of our wiki write problems ? Dec 15 20:52:31 eFfeM ? Dec 15 20:52:56 ka6sox: if I try to update a wiki page I get an error Dec 15 20:53:04 confirmed by others Dec 15 20:53:30 ka6sox: http://thread.gmane.org/gmane.comp.handhelds.openembedded/40559 Dec 15 20:54:23 kk...I'll look @ it now :( Dec 15 20:54:48 eFfeM: I've built samba on armv4t (om-gta02) OK after khem's patches (gcc and arm mode in samba) Dec 15 20:55:28 JaMa: ah ok Dec 15 20:56:40 khem I've started a build for machine minimal, target nslu2le recipe samba, but it will take a while, will probably get back to you around midnight your time (feel free to ping me if I sign in tomorrow morning) Dec 15 20:57:29 khem, hope the build succeeds, some of the things on the autobuilder fail on gettext, haven't really found the cause yet (although I see libtool adding -L/usr/lib to the path) Dec 15 20:57:41 * eFfeM is an auto n00b and a lib n00b too Dec 15 20:57:47 libtool n00b Dec 15 21:15:51 eFfeM semi-fixed...I know what is causing it but its going to take a bit more to "fix" it permanently Dec 15 21:15:55 but for now we can edit agian. Dec 15 21:16:43 ka6sox: thanks Dec 15 21:19:28 JaMa: please look at the patch proposal now Dec 15 21:19:53 JaMa: since there's no users of gitpkgv in recipes I think we don't need to bother too much about compatibility Dec 15 21:21:09 otavio: actually i'm using it in our distro's layers Dec 15 21:21:40 setting PV in gitpkgv is not a good idea Dec 15 21:21:45 otavio: I still think that PV/PKGV should be defined outside .bbclass Dec 15 21:21:46 obi: can you take a look if it fits your needs? Dec 15 21:22:01 otavio: and still think that with bad tags you won't get sortable revision Dec 15 21:22:21 JaMa: so you can use GITPKGV_USE_TAGS = 0 Dec 15 21:23:28 instead of GITPKGV_USE_TAGS, i'd add a parameter to get_git_pkgv() Dec 15 21:23:37 any specific reason to not define PV on it? it can be overrided if need Dec 15 21:23:40 and introduce a new variable, like JaMa proposed Dec 15 21:24:06 if you remove those PV and PGKV from bbclass then you can easily decide if you want PKGV="1.0+${GITPKGV}" or PKGV="${GITPKGVTAG}" Dec 15 21:24:28 otavio: you're already overriding PV from the recipe's filename Dec 15 21:27:08 otavio: and then both those variables can be added to gitver.bbclass and make gitpkgv obsolete (AFAIK GITPKGV is now the same as GITVER in gitver.bbclass and GITPKGVTAG is also available in gitver.bbclass (just before applying that tag filter there) Dec 15 21:28:48 then there will be just 1 bbclass instead of 2-3 with 3 vars Dec 15 21:28:53 GITSHA = just hash Dec 15 21:29:08 GITVER = sortable (count+hash) Dec 15 21:29:36 GITVERTAG = "sortable" tag_count+hash Dec 15 21:29:51 or better GITTAGVER.. Dec 15 21:29:52 OK Dec 15 21:29:53 doing it Dec 15 21:30:32 03Paul Eggleton  07master * rf2c64c91ae 10bitbake.git/lib/bb/data_smart.py: Dec 15 21:30:32 data_smart: add optional expansion to getVarFlag() Dec 15 21:30:33 Add a parameter to getVarFlag() to auto-expand the value of the flag. This Dec 15 21:30:33 makes getVarFlag() more consistent with getVar(), and allows expansion of Dec 15 21:30:33 vardeps and vardepsexclude (which has been done in this commit). Dec 15 21:30:33 Signed-off-by: Paul Eggleton Dec 15 21:31:28 khem, pity the nslu2 console-image build failed on gettext not sure how to fix that libtool issue Dec 15 21:32:33 Jay7: peeked into the patches of the post you gave but they do not seem appropriate for me Dec 15 21:32:37 i think with gitver you currently can't create a PV which doesn't start with 0.0, unless your git tree is tagged in a special way Dec 15 21:35:30 bluelightning, did you have a chance to test and verify things are good with the repo? Dec 15 21:39:04 JaMa: in case of gitpkgv, do you see any reason for not using GITPKGV for PV directly? (i.e. PV = "...${GITPKGV}" instead of PV = "...${SRCPV}" and PKGV = "...${GITPKGV}") Dec 15 21:39:04 JaMa: it looks like gitver will impose the need of PV Dec 15 21:39:14 JaMa: and then make it dependent Dec 15 21:39:20 JaMa: not a good option IMO Dec 15 21:39:55 obi: yes PV is expanded for all recipes during parsing IIRC Dec 15 21:40:19 obi: so you would need git clone all first Dec 15 21:40:30 you need it for SRCPV, too, right? Dec 15 21:40:58 obi: but PKGV only for those which are in dependency tree for your task Dec 15 21:41:15 obi: no unless you use BB_GIT_CLONE_FOR_SRCREV Dec 15 21:42:11 ok Dec 15 21:48:10 re Dec 15 21:50:07 JaMa: did you see my point about gitver Dec 15 21:51:28 otavio: I see it but don't understand it much :/ Dec 15 21:51:52 JaMa: I mean it Dec 15 21:51:53 python () { mark_recipe_dependencies(d.getVar("S", True), d) Dec 15 21:51:58 } Dec 15 21:52:49 otavio: imho it should check if ${S} is available Dec 15 21:53:21 and if not then repodir Dec 15 21:59:35 khem, as you did a lot of work on libtool 2.4: I see 2.2.6 versions remove these two lines in cross_compile.patch Dec 15 21:59:35 - # We cannot seem to hardcode it, guess we'll fake it. Dec 15 21:59:35 - add_dir="-L$libdir" Dec 15 21:59:55 these two lines are still present in libtool 2.4 /libltdl/config/ltmain.sh Dec 15 21:59:58 is that ok???? Dec 15 22:00:04 JaMa: http://pastebin.com/3h3SPrCg Dec 15 22:00:40 i feel that is causing my gettext problem, but I would appreciate some wiz feedback. Dec 15 22:01:00 Feel free to leave a msg to eFfeM_work as I'm going to Zzz also in a minute Dec 15 22:02:37 eFfeM: not sure, but i guess that libtool 2.4 translates absolute paths to be relative to the sysroot, so this should be fine Dec 15 22:02:55 at least that's how i understand the sysroot feature Dec 15 22:03:16 otavio: s/ver = "0.0-%s-g%s" % commits, rev[:7]/ver = "%s+%s" % commits, rev[:7]/ ? Dec 15 22:05:04 otavio: imho you should start with if use_tags and keep this format for GITPKGVTAG and keep old format (same like SRCPV) for GITPKGV Dec 15 22:05:39 eFfeM, khem: I'm affected by gettext problem too, so keep me informed :) Dec 15 22:05:50 and I can provide testing feedback Dec 15 22:06:20 kergoth_: have you seem python backtrace in my letter about testbuilder failures? Dec 15 22:07:40 nope, have a link? Dec 15 22:07:51 otavio: http://pastebin.com/S3CDUjWD or even http://pastebin.com/uV4cpA0Y Dec 15 22:08:24 kergoth: http://article.gmane.org/gmane.comp.handhelds.openembedded/40566 Dec 15 22:08:32 6. Error executing python function in 'do_distribute_sources' for angstrom-2010.x/collie/opie-image Dec 15 22:09:07 error before (5) is may be interesting for you too Dec 15 22:10:03 obi I see a -L/usr/lib in the generated output of libtool Dec 15 22:10:27 Jay7: you might try to remove those two lines I mentioned above Dec 15 22:10:44 please leave me a msg if you succeed Dec 15 22:10:48 now really bedtime Dec 15 22:11:00 Jay7: weird. Dec 15 22:11:03 eFfeM: ok, I'll try Dec 15 22:11:10 permission denied trying to run the source distribute function? Dec 15 22:11:35 eFfeM: in which output exactly? Dec 15 22:11:38 kergoth: I can give link to full build log Dec 15 22:11:58 eFfeM: for 2.4 this might be ok, for 2.2.6 probably not Dec 15 22:12:26 kergoth: my bitbake have revision 0ebb46e25261cfc85aaef2790cba7c1ec057c306 Dec 15 22:12:27 because libtool link does voodoo Dec 15 22:13:09 obi: see http://thread.gmane.org/gmane.comp.handhelds.openembedded/40487 Dec 15 22:13:57 somewhere 30 lines or so from the bottom of the log is a Dec 15 22:13:58 -L/usr/lib -lgettextlib Dec 15 22:14:20 actually it reads Dec 15 22:14:21 -L/usr/lib -lgettextlib -L/home/frans/oe/tmp_minimal/sysroots/armv7a-oe-linux-gnueabi/usr/lib Dec 15 22:14:28 with a newline in it Dec 15 22:14:36 kergoth: full build log: http://jay-home.ath.cx/testbuilder/logs/B.opie.angstrom-2010.x.collie.log Dec 15 22:14:38 but that means /usr/lib is searched first -L is processed in order Dec 15 22:14:41 nite Dec 15 22:14:50 i see Dec 15 22:15:11 Jay7: okay, i have a fix for the traceback. the root cause i have no idea, its nothing to do with bitbake itself Dec 15 22:15:18 * kergoth pushes Dec 15 22:15:32 03Chris Larson  07master * rbeadff2eca 10bitbake.git/lib/bb/build.py: Dec 15 22:15:32 build: write logfiles per task, not per function Dec 15 22:15:32 Based on d14f9bf6 from poky, reworked for master and other cleanup. Dec 15 22:15:32 Signed-off-by: Chris Larson Dec 15 22:15:43 03Chris Larson  07master * r7d80a5355c 10bitbake.git/lib/bb/process.py: Dec 15 22:15:43 process: handle OSErrors other than file not found Dec 15 22:15:43 Signed-off-by: Chris Larson Dec 15 22:16:22 kergoth: I fear process was tried to copy something into / Dec 15 22:16:31 process doesn't copy anything. Dec 15 22:17:17 hm.. may be out of process limits.. Dec 15 22:17:23 I should check it Dec 15 22:18:41 gnn Dec 15 22:18:43 er, hmm Dec 15 22:19:24 yes, I've limit for max 1024 processes per user.. Dec 15 22:19:38 will increase it Dec 15 22:20:25 25 commits left to consider merging from poky Dec 15 22:20:26 whew Dec 15 22:21:58 kergoth: then whats the plan for bb Dec 15 22:22:11 kergoth: will bb be hosted alongside yocto ? Dec 15 22:22:28 i don't see a need for that, personally. more likely yocto will use upstream bitbake Dec 15 22:23:43 trying to add bb recipie for building pure data and getting issues with the sanity checker calling out the .debug directories in the release build. I don't see any debug switches in the -D -D -D output. Are there any tips for working around the failure? Dec 15 22:24:09 kergoth: scmdata=keep ping? Dec 15 22:24:12 it doesn't have to do with debug switches, it has to do the packaging in the recipe Dec 15 22:24:53 oh. I didn't specify any packaging in my bb file. Dec 15 22:25:16 is there an existing bb file I could mimic? Dec 15 22:26:13 your'e going to need to stop summarizing and show us the error you're getting. Dec 15 22:26:18 if you expect us to be very helpful Dec 15 22:26:46 hey, is there a way to control the respawn time for inittab or do i need to modify something? Dec 15 22:27:10 gandhijee: what do you mean? Dec 15 22:27:16 thanks. my work is on my home box and I'll need to do that tomorrow or later tonight. Dec 15 22:28:02 kergoth: i have a pppd running through respawn, every once in a while it glitches and reconnects Dec 15 22:28:29 and then init complains that ppp is respawning to fast because pppd can't find /dev/ttyAMC0 Dec 15 22:29:01 i guess i could maybe put a check for /dev/ttyAMC0 in the ppp script Dec 15 22:29:43 yes, the problem isnt init, its pppd. init will respawn it when it dies, there's no way to make it delay that, you'd ahve to do that on the script end Dec 15 22:33:41 can the pppd connect scripts run shell commands and stuff? Dec 15 22:36:31 Jay7, you can fix your link now on the wiki. Dec 15 22:36:58 ka6sox: 10x Dec 15 22:37:00 03Chris Larson  07master * r3a4b5eb20e 10bitbake.git/lib/bb/build.py: Dec 15 22:37:00 build: fix FuncFailed for the no logfile case Dec 15 22:37:00 Signed-off-by: Chris Larson Dec 15 22:38:12 JaMa|Zzz: thanks for the reminder, looking at it again now Dec 15 22:38:51 gandhijee: just call script from init which will check tty presence and then run pppd or just sleep forever Dec 15 22:39:17 kergoth: thanks for looking at it Dec 15 22:39:21 if you are calling pppd from inittab Dec 15 22:39:39 Jay7: ok Dec 15 22:39:42 thanks Dec 15 22:42:30 03Andreas Oberritter  07master * rf42e6d33b9 10bitbake.git/doc/manual/usermanual.xml: Dec 15 22:42:30 manual: document scmdata=keep for CVS, SVN and Git Dec 15 22:42:30 Signed-off-by: Andreas Oberritter Dec 15 22:42:30 Signed-off-by: Chris Larson Dec 15 22:42:33 03Andreas Oberritter  07master * r05cbc1d1a0 10bitbake.git/lib/bb/fetch/ (bzr.py cvs.py git.py hg.py repo.py svn.py): Dec 15 22:42:33 fetchers: Add parameter scmdata=keep to include .git/ and others in generated tarballs. Dec 15 22:42:33 * Allows generating version information from SCMs during build. Dec 15 22:42:33 * Note that tar doesn't need to use --exclude '.git', because Dec 15 22:42:33 git checkout-index doesn't clone the repository. Dec 15 22:42:33 Signed-off-by: Andreas Oberritter Dec 15 22:42:34 Signed-off-by: Chris Larson Dec 15 22:45:53 kergoth: i understand that 1.8 and 1.10 only accept fixes, but i consider having no access to the git meta data a bug, if all other version control systems do have access to their meta data Dec 15 22:46:12 so would you accept a patch which includes .git unconditionally, i.e. without a new parameter? Dec 15 22:51:13 ka6sox: yep all verified OK, thanks Dec 15 22:51:21 Jay7: thanks, that worked perfectly Dec 15 22:51:31 cool Dec 15 22:56:25 what is smallest image with gettext? Dec 15 22:56:35 console-image? Dec 15 22:56:48 hm.. just gettext Dec 15 23:16:40 kergoth, bitbake is behaving funny.... Dec 15 23:16:44 i get e.g. ERROR: Task 4 (/home/grg/oe2/openembedded/recipes/kexec-tools/kexec-tools-klibc_2.0.2.bb, do_configure) failed with exit code '256' Dec 15 23:16:51 what is exit code 256? Dec 15 23:17:10 exit code 256 should be 0, which should be success Dec 15 23:17:12 right? Dec 15 23:21:52 grg: +1 Dec 15 23:21:59 have this problem too Dec 15 23:22:18 ah. no Dec 15 23:22:27 I have problem with do_compile Dec 15 23:24:16 but you're still getting exit 256 Dec 15 23:24:28 which aint right Dec 15 23:24:35 ERROR: Task 3383 (/storage/oe/testbuilder/openembedded/recipes/kexec-tools/kexec-tools-klibc_2.0.2.bb, do_compile) failed with exit code '256' Dec 15 23:24:38 exit codes are modulo 256 Dec 15 23:29:39 may it be -1? Dec 15 23:30:21 fork may return -1 Dec 15 23:30:39 so, it may be some fork+exec error Dec 15 23:34:57 if i put some logger.warn statements in bitbake, i actually get pipe.errorcode 77 Dec 15 23:35:13 pipe.returncode i mean Dec 15 23:35:39 can you get errno there? Dec 15 23:35:59 hm.. this is errno Dec 15 23:36:20 grg: is 77 decimal or hex? Dec 15 23:36:25 decimal Dec 15 23:37:02 hmm how does exception handling work in python? Dec 15 23:37:21 as it works on other languages i.e. lamely Dec 15 23:37:31 process.run is raising an ExecutionError, but nothing is catching it Dec 15 23:37:36 #define ENOLCK 77 /* No locks available */ Dec 15 23:37:41 maybe a catch all handler somewhere up the chain Dec 15 23:38:18 not sure that it is right error code Dec 15 23:38:21 i'm not too concerned with why kexec-tools-klibc is failing right now... want to fix bitbake first Dec 15 23:39:16 grg: have you tried with bb 1.10? Dec 15 23:39:25 nah Dec 15 23:39:40 i suppose i could Dec 15 23:39:44 and then bisect Dec 15 23:40:08 it was working ok when I've fixed kexec-tools-klibc Dec 15 23:40:32 you can look git log to look where it was to know date and revision Dec 15 23:44:38 2be5169a01e57bcfcb1f75cbd6bac3bf0f230edd was a pretty solid revision for me, and it gives a valid configure error log instead of exit 256 Dec 15 23:44:46 * grg bisects Dec 15 23:44:54 grg, was gonna say, sounds like a case for git bisect run Dec 15 23:45:29 annoyingly though, the failure mode changes a bit in newer revisions... Dec 16 00:01:35 * Jay7 is dancing aroung gettext Dec 16 00:07:37 Jay7: is it pole dance :) Dec 16 00:09:00 kinda Dec 16 00:09:21 sh@t.. it takes -L/usr/lib again.. Dec 16 00:09:39 why does it want to relink on install at all? Dec 16 00:14:28 well.. seems my experience is not enough to fix this.. Dec 16 00:22:20 whats the problem Dec 16 00:22:47 gettext fails to build on minimal Dec 16 00:23:09 http://tinderbox.openembedded.net/packages/1280010/ Dec 16 00:23:12 to install even Dec 16 00:23:26 eFfeM and I are affected Dec 16 00:34:49 grg: did you intend to use & 0xff in "[PATCH] runqueue: Fix error code."? Dec 16 00:35:33 obi, hmm... yes. oops Dec 16 00:36:00 i can't seem to do anything right wtih bitbake... Dec 16 00:36:28 :) Dec 16 00:39:43 * Jay7 -> sleep Dec 16 01:12:59 grg: strange. i *know* i fixed that error code issue at one point Dec 16 01:13:05 * kergoth wonders if it never got merged Dec 16 01:13:35 kergoth, btw, what happened to the log messages being printed out on the terminal? Dec 16 01:13:42 what do you mean? Dec 16 01:14:09 when do_configure failed in the past, id get the log.do_configure printed Dec 16 01:14:36 seems that has disappeared in the last week sometime Dec 16 01:14:43 that'd be a bug, then Dec 16 01:15:40 i think it stopped with this commit f99ee4680c9f67b7ed13fc06044ba2382f9a782c Dec 16 01:16:33 k, thanks Dec 16 01:19:18 hmmm Dec 16 01:21:19 hmm Dec 16 01:32:07 grg: fixed, sorry about that -- it was trying to get the size of the file before we closed it Dec 16 01:32:12 03Chris Larson  07master * red8d503e67 10bitbake.git/lib/bb/build.py: Dec 16 01:32:12 build: fix log file display Dec 16 01:32:12 Signed-off-by: Chris Larson Dec 16 01:32:20 03Chris Larson  07master * r4e25db1f82 10bitbake.git/lib/bb/runqueue.py: Dec 16 01:32:20 runqueue: resurrect the fix in 90c2b6c Dec 16 01:32:20 Signed-off-by: Chris Larson Dec 16 01:41:29 kergoth, one more quick Q... I have kexec-tools-klibc_2.0.2.bb failing in do_compile. I get the log.do_compile on the terminal now, and the log file is useful. But the tinderbox txt file is empty: http://tinderbox.openembedded.org/packages/1309745/ Dec 16 01:41:46 hmm Dec 16 01:42:12 * grg looks for lunch Dec 16 01:42:51 the oestats client looks for log..*, the filenames are still in that form Dec 16 01:43:21 before ed8d503e it could very well have been empty due to that Dec 16 01:43:26 so it might be fixed going forward Dec 16 02:08:51 Before I wrote it, any tpm-emulator (http://tpm-emulator.berlios.de/download.html) OE recipe anywhere? Nothing on Google. Dec 16 02:10:39 gregoiregentil1: if its not in metadata its wont be Dec 16 02:18:07 03irfan sadiq  07master * ra60a251a3e 10openembedded.git/recipes/ltp/ (ltp-20100831/runalltests.patch ltp_20100831.bb): (log message trimmed) Dec 16 02:18:07 To make runalltests.sh configurable for default tests Dec 16 02:18:07 This patch applied to ltp runalltests.sh script to make default set of test Dec 16 02:18:07 suits configurable which is not the default behaviour. Dec 16 02:18:07 By default rullalltests.sh invokes runltp wihout specifying any set of tests Dec 16 02:18:07 to run. This patch will give the user option to chose from defaults suits Dec 16 02:18:08 setting their respective variables named RUN_DEFAULT_SUITENAME to '0' or '1'. Dec 16 02:18:10 03Melissa Watkins  07master * r58d988d780 10openembedded.git/recipes/ti/ti-pru-sw-examples_1.00.00.bb: (log message trimmed) Dec 16 02:18:10 ti-pru-sw-examples: fix package run-time dependencies Dec 16 02:18:10 * This package depends on the ti-pru-sw-edma-driver and Dec 16 02:18:10 kernel-module-uio-pru modules at run-time. Added these packages Dec 16 02:18:10 to the RDEPENDS. Dec 16 02:18:11 * Bump PR Dec 16 02:18:11 * Update SRC_REV Dec 16 02:18:14 03Andreas Oberritter  07master * r5316c5b141 10openembedded.git/recipes/ffmpeg/ffmpeg_0.6.1.bb: Dec 16 02:18:14 ffmpeg: add recipe for 0.6.1 (LGPLv2.1+) Dec 16 02:18:14 Signed-off-by: Andreas Oberritter Dec 16 02:18:29 Signed-off-by: Andreas Oberritter Dec 16 02:18:29 Signed-off-by: Khem Raj Dec 16 02:18:29 03Andreas Oberritter  07master * r046452a456 10openembedded.git/docs/usermanual/reference/var_src_uri.xml: Dec 16 02:18:29 usermanual: remove unused variable CVS_TARBALL_STASH Dec 16 02:18:29 Signed-off-by: Andreas Oberritter Dec 16 02:18:29 Signed-off-by: Khem Raj **** ENDING LOGGING AT Thu Dec 16 02:59:57 2010