**** BEGIN LOGGING AT Mon Nov 17 03:00:00 2014 Nov 17 07:50:21 Good morning every body. I have got a little problem with yocto daisy. I have a proper image which compil. A day, I needed a sdk for a collegue and used populate_sdk function but now yocto populate a new sdk for each new image compilation. It make time, and it is not needed each time. Every body now how to disabling this function ? Nov 17 08:01:38 Nobody can help me ? Nov 17 08:04:56 good morning Nov 17 08:05:18 adbelhimoun: how did you used populate_sdk? Nov 17 08:06:12 I used 'bitbake myimage -c populate_sdk' only one time but now, sdk is populating automaticaly each times Nov 17 08:13:28 adbelhimoun: I don't think it works that way Nov 17 08:51:25 morning all Nov 17 08:52:58 hi bluelightning Nov 17 08:53:21 hi mckoan Nov 17 09:37:49 Hi! I'm trying to build chromium on yocto for i.MX6 on a VM running 32bit Ubuntu. Although I set up my VM to have 10G of RAM and 10G of swap I get "Memory exhausted" during link of libblink_web. Nov 17 09:38:05 Can someone help? Or is it even possible on a 32bit OS? Nov 17 09:38:52 I even tried the "sharedbuild" PCKCONFIG, but without success. Nov 17 09:51:40 hi guys Nov 17 09:51:49 anybody had any luck with alsa in qemu? Nov 17 10:19:39 hi ppl, I have the following question: I have a recipe which I build from an external source tree using the INHERIT += "externalsrc" and EXTERNALSRC_pn-myrecipe = "path-to-your-source-tree". When I make a change to a file in this source tree and try to rebuild it using `bitbake recipe` it says that does not need to rerun, which I suppose it's because it doesn't detect the change. I have to make `bitbake -ccleanall recipe` in order to build it which rebuild Nov 17 11:08:59 stavrosv1: I guess you can try to cheat a little bit and use sth like 'do_fetch[nostamp] = "1"' - thanks to that sources will always be fetched, even when no change happened Nov 17 11:45:21 frsc: new chromium cannot be linked on a 32bit system Nov 17 11:46:07 frsc: for their own build infra, they change the 32bit builder to use a 64bit toolchain+kernel Nov 17 11:46:18 bloatware :) Nov 17 11:47:48 frsc: https://groups.google.com/a/chromium.org/forum/#!topic/chromium-packagers/mxPyPeoMwxk Nov 17 11:48:36 see the answer from Lei Zhang Nov 17 12:03:10 Xz: thanks for the suggestion. it worked using `do_compile[nostamp] = "1"`, though I still getting the whole source tree to build. At least I don't need to make `bitbake -ccleanall recipe`. The rebuild of the whole source tree it might be related to the build files of the recipe. The recipe is for webkit which every build it takes 15 minutes in my machine in every change that I make. I will investigate more on this. Thanks Nov 17 12:28:35 kimrhh: Thank you for the info! Is there anything I have to watch out for when I change my OS to a 64bit Ubuntu, copy my current Yocto environment and do a rebuild of my image? Nov 17 12:29:07 don't ask me about that :) Nov 17 13:31:24 frsc: I'm pretty sure I've done that successfully. However I keep my workspacve drive separate to the OS so effectively reinstalled ubuntu 14.04 64bit with all packages for yocto then rebuilt my working code... Nov 17 13:57:41 pev_: Thanks! I'll try. Nov 17 13:57:55 how i can patch a kernel in a custom image? Nov 17 13:59:39 add the patch file name into the SRC_URI list Nov 17 14:00:38 e.g. SRC_URI += "file://halt.c.patch" Nov 17 14:00:47 Nov 17 14:02:07 T0mW: I already put in a meta-custom/linux/linux-yocto_3.14.bbappend Nov 17 14:02:17 tyno: what I do is not bother with patching the kernel as I find the Yocto approach to handling the build of the kernel to be a bit clumsy, I use my own git repository for my customized kernel. Nov 17 14:02:52 s/clumsy/arcane/ Nov 17 14:04:05 T0mW when I run devshell I can see patch apply but when I build my custom image it is not applyed Nov 17 14:08:04 T0mW: maybe there is a way to select kernel? Nov 17 14:08:40 tyno: If you can see it applied in devshell then it is... Have you done some basic tests like putting Nov 17 14:08:53 #error's in files that should be compiled to test if they are? Nov 17 14:10:18 pev_: Ok I'll do now Nov 17 14:12:04 * T0mW wishes he can run a devshell Nov 17 14:13:11 pev_: but patch changes headers and I didn't find headers changed Nov 17 14:17:08 If wishes were horses, we'd be eating steak! Nov 17 14:18:20 lart Crofton Nov 17 14:18:39 tyno: Ok, lets start again :-) You said that you could see the patch applied but the headers hadnt changed. Can you say how you knew it had applied if it wasn't applied? Nov 17 14:18:40 pev_:adding a error in kernel source by patch build fails, so why I don't find header changed? Nov 17 14:19:48 T0mW: How come you can't devshell? Nov 17 14:20:01 oe.terminal.spawn(terminal, command, title, None, d) Nov 17 14:20:29 'command' is a bash shell function, my opensuse has the shellshock patches in it. Nov 17 14:22:04 tyno, Wich headers are unchanged? The installed kernel headers (libc), or the one used to build the kernel? Nov 17 14:22:19 T0mW: hm, that sucks... Nov 17 14:23:42 pev_: patch changes kernel include/uapi/linux/can.h Nov 17 14:33:44 AndersD: patch changes include/uapi/linux/can.h but when I look into sysroot /usr/include/linux/can.h Nov 17 14:34:09 AndersD: there isn't my changes Nov 17 14:34:52 tyno: you'd need to manually export the changes I think Nov 17 14:35:08 sysroot is the libc headers IIRC Nov 17 14:35:40 tyno: Have a look at the discussion https://lists.yoctoproject.org/pipermail/yocto/2014-April/019260.html Nov 17 14:36:17 This script, mc-wrapper.sh, is that even necessary for the devshell ? Nov 17 14:36:26 tyno, just like pev_ says, the headers installed under sysroot/usr/include/linux/* comes from the linux-libc-headers package. Nov 17 14:36:39 Or, is that some "maybe someone will need this" type of thing? Nov 17 14:36:58 ok many thaks I'll ready now Nov 17 14:43:34 If I edit 'meta/classes/terminal.bbclass and comment out "command = emit_terminal_func(command, envdata, d)" from emit_terminal_func(), I can get a devshell. Nov 17 14:45:55 The shell seems to work just fine without the BASH_FUNC_mc%% Nov 17 15:10:59 hey.. bitbake can not find a *.pc file, so how to set PKG_CONFIG_PATH additionally? Nov 17 15:18:01 anyone with nice one-liner to generate "installed-packages" report from package-depends.dot? Nov 17 15:18:46 I'm thinking about using networkx to find reach-able packages, but there should be easier way Nov 17 15:50:17 is there a way to disable sstate cache for a package (or certain tasks of packages)? Nov 17 15:59:41 pev_: so better solution is to add an extra directory with changed linux header Nov 17 16:00:37 pev_: and not change /usr/include/linux header Nov 17 16:03:42 erbo: afaik, nope. might be possible with some fiddling, but there's no provided mechanism to do so. why would you want to? Nov 17 16:18:52 tyno: exactly. I've had the same problems here... Nov 17 19:29:11 hi Nov 17 19:34:25 Should I report a bug to request an upgrade to latest syslinux ? Nov 17 19:34:58 can anyone mentor me , or is it a regular process I can figure out / Nov 17 19:47:46 and eventually provide patches ;) Nov 17 19:51:03 I've just inadvertently overwritten /lib/libc.so Nov 17 19:51:09 How can I recover ? Nov 17 19:56:39 interested Nov 17 19:57:02 on your host I presume ? Nov 17 19:57:39 do you have a static busybox ? Nov 17 19:58:25 No. It would seem to be dynamically linked too :( Nov 17 19:58:51 you have only < > shell redirectors ? Nov 17 19:58:57 yep! Nov 17 19:59:03 that's better than nothing Nov 17 19:59:04 echo > > < Nov 17 19:59:36 And I have only one shell running. Once I logout I'm gone forever! Nov 17 19:59:45 reboot and mount ? Nov 17 20:00:32 But if I reboot, nothing will run. It'll fall over on the first thing that needs libc Nov 17 20:00:50 * RzR sudo apt-get install busybox-static Nov 17 20:01:04 i mean poweroff , mount Nov 17 20:01:12 cp fie Nov 17 20:01:15 and boot again Nov 17 20:01:43 Now you've lost me. How can I mount something when powered off? Nov 17 20:01:59 boot an other os ? Nov 17 20:02:04 or pull the hdd out Nov 17 20:02:33 is your host next to you or down in the cloud ? Nov 17 20:02:48 In the cupboard in the next room. Nov 17 20:03:20 so you're safe Nov 17 20:04:11 what distro is it running ? Nov 17 20:04:26 Actually I'm not sure. Nov 17 20:07:17 jmd, cat you type /lib/libc ? Nov 17 20:07:41 libc-2.18.so libcidn-2.20.so libcom_err.so.2 libcrypt-2.20.so libcrypto.so.1.0.0 Nov 17 20:07:41 libc-2.20.so libcidn.so libcom_err.so.2.1 libcrypt.so Nov 17 20:07:41 libc.so.6 libcidn.so.1 libcrypt-2.18.so libcrypt.so.1 Nov 17 20:07:42 jmd, maybe you have other files to link to redirect ? Nov 17 20:07:55 jmd, ok good Nov 17 20:08:20 jmd, /lib/libc.so < /lib/libc.so.6 Nov 17 20:08:45 jmd, no ln ? Nov 17 20:08:51 no Nov 17 20:09:06 lib# ln Nov 17 20:09:06 ln: error while loading shared libraries: libc.so.6: cannot open shared object file: No such file or directory Nov 17 20:09:28 jmd, cat ? Nov 17 20:09:33 same Nov 17 20:10:12 jmd, mv ? Nov 17 20:10:18 no Nov 17 20:10:19 cp ? Nov 17 20:10:21 :) Nov 17 20:10:24 no Nov 17 20:10:42 bash ? Nov 17 20:11:02 no Nov 17 20:11:18 i mean what is your shell ? Nov 17 20:11:22 echo $SHELL Nov 17 20:11:31 bash has printf Nov 17 20:11:40 Yes. Its bash Nov 17 20:12:32 # rzr@lap:tmp/ # [1] # date > input Nov 17 20:12:32 # rzr@lap:tmp/ # [0] # printf "%s" < input > output Nov 17 20:12:51 should be possible to deal with printf Nov 17 20:13:01 date requires libc Nov 17 20:13:40 we dont care of date Nov 17 20:13:45 that was an example Nov 17 20:14:37 What will printf "%s" < input > output achieve ? Nov 17 20:14:57 it copy the file but in text Nov 17 20:15:03 you want a binary one Nov 17 20:16:37 When I try it on my laptop, it creates an empty file. Nov 17 20:16:57 which is what I would expect. Nov 17 20:17:48 jmd: it's really going to be easiest to copy a file in externally (from another system or a recovery environment) Nov 17 20:18:11 fishey: Yeah I was afraid that was going to be the case. Nov 17 20:18:33 you can mess with pasting hex encoded data (which is the direction I think RzR is headed), but it's probably not worth it. Nov 17 20:18:33 fishey, this is less fun Nov 17 20:20:11 fishey, do you have eval ? Nov 17 20:20:16 jmd, ^ Nov 17 20:20:24 yes Nov 17 20:21:18 I think theres a way to open an arbitrary file with read, but I don't remember how. Nov 17 20:23:18 you want to read a binary stream right ? Nov 17 20:23:39 exec 3 IFS= Nov 17 20:23:52 read -N1 -u3 char Nov 17 20:24:52 while read foo ; do process_foo ; done < /tmp/foo Nov 17 20:25:09 Yeah thats the way. Nov 17 20:25:37 you want to avoid cat /tmp/foo | while read .... because in some shells the pipe opens a new subshell, and any variable changes you make may get lost Nov 17 20:26:41 Anyway, I'm not sure where that gets me. Nov 17 20:31:29 Hm, with busybox are there no options for setting mode/uid/gid on mountpoints? That seems like a bit of an omission?! Nov 17 20:32:31 fray: oh i hate that pipe subshell behavior, so irritating, if understandable. always makes the shell code slightly more complex Nov 17 20:32:50 * kergoth grumbles at bitbake Nov 17 20:44:19 jmd, no chmod ? Nov 17 20:44:34 no Nov 17 20:44:51 jmd, I what thinkgint about serializing busybox-static Nov 17 20:44:59 but this is no use for you Nov 17 20:49:15 I think fishey is probably right, I will have to get it reinstalled by someone who knows how. Nov 17 20:51:30 giving up ? Nov 17 20:51:53 I think I've exhaughsted most avenues. Nov 17 21:06:37 jmd: on a working box with the libc in question: `hexdump -v --format '1 1 "%02x\n"' libc.so.6 | xclip -i` Nov 17 21:07:07 jmd: on the dead one: `while read byte; do printf "\x$byte"; done >/lib/libc.so.6` Nov 17 21:07:12 and then paste :) Nov 17 21:07:29 and press ctrl-d after pasting. Nov 17 21:07:47 I'll try it Nov 17 21:10:36 No idea how to paste binary Nov 17 21:10:59 jmd: the first command puts the hex encoded file into the clipboard Nov 17 21:11:05 or rather, the primary selection Nov 17 21:11:23 you can tweak the xclip flags to put it in the clipboard if that's more convenient for you Nov 17 21:11:51 pasting binary won't work because it probably has shell escape sequences in it. Nov 17 21:12:52 [which is why the hex encoding is needed] Nov 17 21:15:16 My hexdump must be different to yours Nov 17 21:16:33 use od Nov 17 21:17:08 /lib/libc.so Nov 17 21:17:16 is the one you want to write Nov 17 21:17:20 not /lib/libc.so.6 Nov 17 21:19:58 Anyone know how a python function can fail without throwing an exception? (bitbake) Nov 17 21:20:08 then I'll have to find a way to do chmod Nov 17 21:20:24 Or maybe I should set umask first Nov 17 21:20:47 RzR: `cat /lib/libc.so | head -n1` => '/* GNU ld script' Nov 17 21:20:52 depends on the system Nov 17 21:21:25 jmd: libc doesn't need to be executable, you're thinking of /lib/ld-linux.so.2 Nov 17 21:22:45 Nah I need to do it to some things in bin I think. Nov 17 21:23:54 as long as something still has +x, you can `cat < chmod >thing_to_trash` (or some equiv with a shell builtin) and then run `thing_to_trash` Nov 17 21:24:19 Yeah trying that. Nov 17 21:24:39 I get the error 'Text file busy' Nov 17 21:25:14 choose something else as your sacrifice? Nov 17 21:25:51 Same with all I've found so far Nov 17 21:26:05 fishey, he said he rm libc.so Nov 17 21:26:10 ok found one Nov 17 21:26:17 he has libc.so.6 Nov 17 21:27:02 RzR: ah, I missed that Nov 17 21:41:18 #jmd, while read -n 1 byte ; do printf "%b" "$byte" ; done < /lib/libc.so.6 > /lib/libc.so Nov 17 21:41:38 unsure it will help Nov 17 21:42:00 since libc.so is not supposed to be a lib as fishey suggested Nov 17 21:43:16 else just replace content Nov 17 21:43:21 I have one like Nov 17 21:43:42 OUTPUT_FORMAT(elf64-x86-64) Nov 17 21:43:42 GROUP ( /lib64/libc.so.6 /usr/lib64/libc_nonshared.a AS_NEEDED ( /lib64/ld-linux-x86-64.so.2 ) ) Nov 17 21:48:04 RzR: libc.so shouldn't be needed except for development Nov 17 21:48:30 Which Is why I assumed he had the runtime one missing Nov 17 21:48:31 yea rpath use versionned libs Nov 17 21:48:51 i suspect he lost more than that Nov 17 21:49:03 since he had the libc.so.6 Nov 17 21:49:12 jmd, how did you lost that file ? Nov 17 21:49:29 fishey, btw i am new to yocto , is there a process to request a package upgrade ? Nov 17 21:49:44 fishey, would a wish bug help ? Nov 17 21:50:19 RzR: Simple - I fucked up. Nov 17 21:50:35 RzR: sending a patch to the appropriate repo to the appropriate mailing list with the appropriate header is probably the best way :) Nov 17 21:51:16 ok I prefer trackers to mailing list , so I'll may do both Nov 17 21:52:29 jmd, just like rm /lib/libc.so ? Nov 17 21:52:40 jmd, or doing crazy other things ? Nov 17 21:53:38 No I untared a tarball which overwrote a lot of stuff in /lib Nov 17 21:54:17 ahh Nov 17 21:54:35 so you probably cant rely on existing files /lib Nov 17 21:54:40 No. Nov 17 21:54:41 in ^ Nov 17 21:54:48 forget my suggestions then Nov 17 21:55:01 But using your method I have copied some statically linked binaries to /bin Nov 17 21:55:02 it is just spearding the nightmare :) Nov 17 21:55:19 maybe I'll be able to see what it goind on Nov 17 21:55:41 which binaries ? Nov 17 21:56:08 let me share my static bustybox Nov 17 21:56:16 this way you can have a decent shell Nov 17 21:56:30 Oh I have a static bash already Nov 17 22:02:36 busy have more command like like netcat Nov 17 22:18:32 I thought I restored my libraries now. Nov 17 22:18:59 But anything non statically linked returns: error while loading shared libraries: libc.so.6: cannot open shared object file: No such file or directory Nov 17 22:19:15 I have run ldconfig Nov 17 22:19:22 What else do I need to do? **** ENDING LOGGING AT Tue Nov 18 02:59:59 2014