**** BEGIN LOGGING AT Tue May 22 02:59:58 2012 May 22 06:03:14 hi khem May 22 06:03:18 are you still there? May 22 06:03:48 I'm digging into _start () _init () etc.... May 22 06:23:37 can anyone tell me what happend to the wireless in 3.0.1, i miss acx100 for hx4700 ? May 22 06:37:29 do you use mainline? May 22 06:41:00 GNUtoo-desktop, i pulled from kernel-org and applied patches from OE May 22 06:41:17 ok May 22 06:41:30 acx100 is in mainline? May 22 06:42:09 GNUtoo-desktop, afair yes but neither oe-core nor mine has it build May 22 06:42:34 it's not in mainline May 22 06:42:38 that's the problem May 22 06:42:49 http://en.wikipedia.org/wiki/Comparison_of_open_source_wireless_drivers says no to integration in mainline May 22 06:43:24 GNUtoo-desktop, the acx is from texas instruments, in 2.6.21 there was something special for hx4700 because it was obv. attached via another bus. May 22 06:44:21 GNUtoo-desktop, koen from angstrom told about jernel3 long time ago but i never got proper sources in my hands. May 22 06:44:32 *ker May 22 06:44:43 basically the patches you pulled have to contain acx100 May 22 06:45:14 yes details are here: May 22 06:45:15 http://www.angstrom-distribution.org/kernel-30-hx4700 May 22 06:45:25 GNUtoo-desktop, they don't just a bit around the asic3 and timing May 22 06:46:07 GNUtoo-desktop, that's the link i mentioned but........sources ? May 22 06:46:28 I guess it's mainline May 22 06:46:51 else look at the kernel recipes May 22 06:48:24 GNUtoo-desktop, did all that, building angstrom with oe does produce a kernel 3.0.1 which won't boot et all. i prepared a qemu-arm for pseudo-native compilation and it works but no wifi. May 22 06:49:04 maybe it doesn't boot because of problems with alignements May 22 06:49:47 GNUtoo-desktop, MACHINE=hx4700 should align proper, shouldn't it ? May 22 06:52:00 http://git.freesmartphone.org/?p=linux-2.6.git;a=blobdiff;f=Makefile;h=9f6e3cd38548e64e2f610cd8aef5405c1caf8862;hp=2f4a72e2021608cf7c9b281b17a9b500a743ceb1;hb=bef69c533748bdc3ad0ccae56f1f5440ebc3aa0f;hpb=7ae1282f5b3b80bda87ef9aa1d8bf582875ffb67 May 22 06:53:46 ok, that'S 2.6 and if that would be a reason, i see no reason why my qemu-arm build works. May 22 06:55:10 can't also be related to the toolchain because rootfs works. May 22 06:58:12 yes my fix is toolchain-related May 22 06:58:34 also note that my fix is a bad fix May 22 06:58:45 the correct fix was to import a kernel patches that fixes that May 22 06:58:48 in code May 22 06:58:51 not in Makefile May 22 07:00:07 ok, that would be about alignment, will give it a shot (if not already in oe-core) but then, i still miss acx100 May 22 07:00:58 yes, look if there is a recipe for an external driver May 22 07:12:55 good morning May 22 08:15:30 hi all May 22 08:20:27 silvio_: hi May 22 09:11:57 i've a problem with gobject-introspection-native: compilation fails with symbol lookup error: /usr/lib/libgirepository-1.0.so.1: undefined symbol: g_mutex_lock May 22 09:12:43 g_mutex_lock came from glib-2.0-native but it seems that it tries to be taken from system instead of native recipes May 22 09:13:37 searching on google can't help, as this particular error (and similar) seem to affect just me. Is it possible I make something wrong before this point, but at present moment, I've no clue over it. May 22 09:33:57 morning all May 22 09:38:55 hi bluelightning, all May 22 09:44:08 hi eFfeM_work May 22 09:44:27 hi bluelightning .........got it managed to boot hx4700 with 3.0.1, just lacks on certain kernel-modules which appear to be nowhere ;) May 22 09:44:54 lautriv: FYI I managed to build a 3.4 kernel last night with the latest patches from Paul... haven't tested it yet though May 22 09:45:46 bluelightning, what does those patches include ? May 22 09:46:13 lautriv: whole bunch of things... if you like I can push a branch for you to try out May 22 09:46:24 sure May 22 09:47:32 a mail with patches would be even better ;) May 22 09:51:34 lautriv: well you really need the recipe if you want to be able to build it May 22 09:52:18 lautriv: https://github.com/bluelightning/meta-handheld/tree/paule/hx4700-update May 22 09:54:01 bluelightning, OE got it never managed to build anything for my hx4700, narcissus from angstrom fails too so i produce this kernel in a qemu-arm. May 22 09:56:29 hi all May 22 09:57:38 lautriv: I have tested this, it does at least build properly... May 22 09:57:42 hi pb_ May 22 09:58:09 hi evryone May 22 09:58:13 hi mohtadi May 22 09:58:37 bluelightning, did that earlier, builds with success and booting the kernel freezing on kexecboot-menu. May 22 09:59:01 i've a question please, anyone knows what's the best embedded hardware pc that has an internal 3G module May 22 09:59:01 ? May 22 09:59:13 lautriv: so what is the difference when you build your own? May 22 09:59:50 other than not cross-compiling, of course May 22 10:00:34 in term of quality / price May 22 10:01:19 mohtadi: define "embedded hardware pc" May 22 10:01:32 embedded pc May 22 10:01:43 bluelightning, running in qemu-arm i have a pseudo-native compiler which does the job while the OE toolchain seems to have a problem with alignment like GNUtoo-desktop formerly mentioned. May 22 10:02:15 lautriv: well, we ought to fix that then... May 22 10:02:56 lautriv: is this using latest gcc 4.7 (as in current master)? May 22 10:04:07 bluelightning, but then, whenever i search something in OE, regardless if it's a config or file or whatever grep and find running hot without success and the whole repo looks like something on my FS explode. May 22 10:05:14 lautriv: well, building a full image is complicated, we need all those files... they are organised May 22 10:05:55 lautriv: use 'git grep' and when you use 'find' ensure it's not searching in your build directory May 22 10:06:50 building a kernel by hand is all well and good, but then reproducibly building an entire image without a proper system to do it isn't really practical May 22 10:07:32 bluelightning, organised, maybe......functional, unfortunately not. how comes if i bitbake console-image i get firmware for acx wifi but no modules and if i bitbake systemd-gnome-image i get neither firmware nor module ? May 22 10:08:20 lautriv: I can't vouch for the contents of either of those images, they are angstrom-specific May 22 10:10:02 bluelightning, i was on all ways, oe-classic, oe-core, angstroms's setup-scripts . anything related to MACHINE=hx4700 sucks. even their online-builder wont produce anything. May 22 10:10:51 lautriv: well, this machine is old and there are few who care about it; we have to pull together and work on the same thing or we'll get nowhere May 22 10:12:15 bluelightning, i'm on the road to find something working but wth are those missing pathes if they ever existed ? May 22 10:12:24 lautriv: provided the image includes task-machine-base (which will be pulled in by task-base), the appropriate modules will be pulled in via MACHINE_EXTRA_RRECOMMENDS May 22 10:12:42 lautriv: that's assuming the kernel defconfig is set to build those modules, of course May 22 10:13:56 lautriv: the latest ones are here: https://github.com/bluelightning/meta-handheld/tree/paule/hx4700-update/recipes-kernel/linux/linux-3.4/hx4700 May 22 10:14:35 bluelightning, kernel-module-acx is in RRECOMENTS and tiinit + acx-firmware is in RDEPENDS May 22 10:14:37 lautriv: you can see MACHINE_EXTRA_RRECOMMENDS is set correctly in the machine configuration for hx4700 which can be found in conf/machine/hx4700.conf in meta-oe May 22 10:14:44 er, not meta-oe, I mean meta-handheld May 22 10:15:09 see above May 22 10:15:14 lautriv: yes, that's deliberate - we always use RRECOMMENDS for kernel modules in case the kernel config specifies to build them into the kernel May 22 10:15:50 RRECOMMENDS will result in the package being pulled in if it exists May 22 10:16:08 bluelightning, in the case of acx it can't be build in due to init. May 22 10:16:23 lautriv: it's not important though May 22 10:16:39 lautriv: if it exists, it will be installed - provided task-machine-base is installed May 22 10:17:02 if it's not installed, then task-machine-base is not being installed or something else is going wrong May 22 10:18:26 neither console-image nor systemd-gnome-image would make any sense without task-machine-base, wouldn't it ? May 22 10:20:00 lautriv: depends on what you want to do with them, but especially in the case of systemd-gnome-image I would say you would expect it to include task-machine-base May 22 10:21:04 bluelightning, i expect to have task-machine-base to be a foundation for anything, it makes no sense to have a machine up nad can't work with it because unreachable. May 22 10:21:25 the idea is task-core-boot pulls in packages that absolutely have to be installed in order to boot (MACHINE_ESSENTIAL_*) and task-machine-base adds everything else May 22 10:21:36 like I said I can't vouch for the contents of those images May 22 10:22:16 they should include task-machine-base, no idea if they actually do May 22 10:23:31 bluelightning, regardless of those images, the hx4700 has no keyboard and as such it needs ANY way to get in, console must have net or onscreen-kbd or both, none of them is useless. May 22 10:24:06 lautriv: what about serial or usbnet? May 22 10:24:34 I'm not saying those are necessarily helpful to you, but they are ways to interact with the device in the absence of some other network connection May 22 10:25:50 sure but that's more a backdoor than a solution, "wire this portable thing if you need it" May 22 10:28:55 lautriv: so according to 'git grep' in meta-angstrom, console-image does include task-machine-base (via task-base-extended then task-base) May 22 10:29:13 lautriv: so I'm not entirely sure why you would not have the module... May 22 10:29:27 bluelightning, that is what i suspect May 22 10:30:21 lautriv: can you verify that a package for that module exists? see tmp/deploy/(rpm|ipk)/hx4700/kernel-module-* May 22 10:33:52 bluelightning, 14 ipk's no acx May 22 10:34:31 lunchtime and work, back in approx 6h May 22 10:34:40 lautriv: ok, there's your problem... does the kernel defconfig definitely enable the acx module? May 22 10:43:38 lautriv: looking at the acx100 driver it sounds like it's not even in mainline May 22 10:43:55 lautriv: so we don't even have it in our current kernels, that's the real problem May 22 10:44:32 was not acx100 in staging and then removed due to no one working on it? May 22 10:45:50 hmm, looks like it never got there even May 22 11:06:40 hrw: there's been some recent activity looking at the mailing list, but I'm not sure what the current state of it is May 22 12:55:49 Hi Guys, I try to use oe and bitbake for the first time. I tried 'bitbake nano' and the process is staying half a day on 'NOTE: package quilt-native-0.48-r7.1: task do_fetch: Started'. Is this normal? May 22 13:03:30 sounds like its having issues fetching the sources. kill it and try again May 22 13:47:00 trygvis: thank you, can I diagnose this issue? May 22 13:48:57 ctr+c it stais on Waiting for 1 active tasks to finish: .. I had to prez ctr+c again May 22 13:50:02 right, that's normal (although taking forever to fetch isn't) May 22 13:50:27 BTW this is centos 5.2 and I've installed python2.6 I have modified few bitbake scripts so they use python26 May 22 13:50:53 dpenev: have a look in TMPDIR/work/*/quilt-native-0.48-r7.1/temp/log.do_fetch May 22 13:52:20 dpenev: does your internet connection require you to use any kind of proxy? is there any chance you are being blocked by a firewall? May 22 13:55:27 bluelightning: I don't have this log file right now. But before I interrupt the process I had it and it was empty May 22 13:56:47 dpenev: are you using OE-Classic or OE-Core btw? May 22 13:57:40 iptables is not active at the moment May 22 13:58:02 classic/core oh let me check what git repository I have cloned May 22 13:59:07 bluelightning: or can I check which version I use from the files I have? May 22 14:00:54 bluelightning: I have two files log.base_do_fetch.17851 and log.do_fetch.17851 in the directory you pointed me, both are empty May 22 14:03:03 dpenev: that's strange... I would expect some kind of output... May 22 14:03:24 dpenev: the easiest way to check is which repo did you clone? May 22 14:03:44 "git remote -v" will list that FYI May 22 14:05:24 bluelightning: I just followed the OE getting started and I did 'git clone git://github.com/openembedded/openembedded.git' May 22 14:06:10 bluelightning: Probably this is an instalation problem ? is bitbake ok with python2.6 ? May 22 14:06:25 dpenev: yes python 2.6 is fine May 22 14:06:31 dpenev: that is OE-Classic though May 22 14:06:43 dpenev: if you're starting from scratch I would recommend OE-Core instead May 22 14:06:51 dpenev: or the Yocto Project May 22 14:07:04 (more or less the same thing, at least in terms of the metadata) May 22 14:07:12 I want to test image on beagleboard May 22 14:08:57 right, then you probably want to add the meta-ti layer on top of OE-Core / the Yocto Project May 22 14:09:07 (for full support of that board) May 22 14:10:55 bluelightning: Hmm I tought I will be able to compile the kernel/rootfs without having to study the meta description/recepies. What I want right now is a gentle introduction to the openembeded and the Angstrom distribution available for beagleboard May 22 14:11:42 dpenev: I'm not sure if there is such a thing; but getting started building angstrom isn't too hard, check out their website May 22 14:12:43 bluelightning: I've checked their site and they say please confirm that 'bitbake nano' is working before proceed further. Thats how I came here May 22 14:18:54 dpenev: are you using the latest version of the angstrom setup scripts? May 22 14:19:10 (they are based on OE-Core) May 22 14:24:06 bluelightning: I still have not taken the angstrom scripts, I just installed the OE folowing the geting started and test with bitbake nano to confirm that Ihave installed OE properly May 22 14:27:03 dpenev: honestly I'd recommend using the angstrom setup scripts, especially if you're going to use them anyway May 22 14:33:13 bluelightning: ok I will try that. WHat I am not sure is the following. On centos I have python 2.4. I got bitbake issue so I ahd to upgread. I wanted to instal python2.6 as extra package and still keeping 2.4 on my system. Now I have them both. I have modified all *.py in bitbake/ and openembedded/ so now I have #!/usr/bin/env python26 and #!/usr/bin/python26, Is this the way to go or I better put a symlink python to my python26 ? May 22 14:33:49 dpenev: probably a symlink would be less painful May 22 14:34:02 dpenev: I think that's what I used when I was testing on centos 5.2 May 22 14:34:06 as I am ne wto python I ma asking May 22 14:34:18 ah ok May 22 14:34:56 bluelightning: I have to go for a while an dwill test later on again . Thank you for your help!!! May 22 14:35:03 dpenev: np May 22 15:36:47 BTW we just found a bug w/ the prelinker and the default baselib of '/lib' for all architectures.. see bugzilla.yoctoproject.org bug 2484 May 22 17:16:16 Is there a stable branch of oe-core yet? I see a branch called denzil, but I'm not sure what that is. May 22 17:17:19 pingswept, yes it's a release May 22 17:17:26 denzil corresponds to the most recent yocto/poky release, as it's bound to a release, yes, it should be somewhat stable May 22 17:18:57 Does Yocto/Poky add a substantial amount of stuff to oe-core, or is it just a branch without more risky updates? May 22 17:20:19 What I'm really getting at is: if I want to build oe-core, am I better off using denzil, or does that have weird customizations in it that make it unsuitable for other use? May 22 17:30:52 no, yocto uses oe-core as is May 22 17:31:17 that branch is just a snapshot as of the point in time of the release, plus any bugfixes that occurred after that May 22 17:32:35 Thanks, kergoth` May 22 17:32:52 I'll look at denzil-- sounds like what I want. May 22 19:17:52 hi khem May 22 19:18:25 GNUtoo-desktop: hi May 22 19:19:30 I've tried to upgrade chromium and now I've a problem with _init() May 22 19:19:44 00206768 T _init May 22 19:20:08 bt shows that: May 22 19:20:18 http://pastie.org/private/xctnd7vvplh9oz5qzk1gla May 22 19:20:30 and bt in C shows _init() May 22 19:20:41 in frame 1 May 22 19:20:57 hmm May 22 19:21:03 is it dying there May 22 19:21:08 yes May 22 19:21:18 exactly in 00206768 =>segfault May 22 19:21:25 basically nm shows 00206768 as init May 22 19:21:28 but when I disass init May 22 19:21:33 the address is different May 22 19:22:21 hmm I dont see where its failing May 22 19:22:57 need to get some food May 22 19:23:01 ok May 22 19:23:11 in the meantime I reinstall all the debug environment May 22 19:23:18 because I got out of space May 22 19:23:23 so it didn't boot anymore May 22 19:27:09 the update-alternatives refactoring was just sent to the list..... I'm looking forward to some reviews of the changes May 22 22:34:31 khem, ping May 22 22:47:13 GNUtoo-desktop: hi, any luck wrt ubi/subpages verify ? May 22 22:49:51 ant__, with the htcdream? May 22 23:04:14 GNUtoo-desktop: whats up May 22 23:04:42 khem, hi I'll setup the debug environment now May 22 23:05:06 GNUtoo-desktop: ok. May 22 23:05:16 and email me the access info May 22 23:05:25 btw. are you using gold with my pathc ? May 22 23:10:19 can't I instead type the command you ask for? May 22 23:10:31 because that would be faster than preparing a machine May 22 23:11:26 for now it's reading the symbols.... May 22 23:12:48 GNUtoo-desktop: I will need time at the same time May 22 23:13:06 GNUtoo-desktop: its harder to intruct May 22 23:13:09 ok May 22 23:13:17 I already know why it fails May 22 23:13:17 since there are multiple things that needs to be done it will get slow May 22 23:13:22 ok May 22 23:13:23 tell me May 22 23:13:26 but I've no idea how to fix it May 22 23:13:48 nm shows the following address: May 22 23:13:55 00206768 T _init May 22 23:14:10 hmm so what did you try May 22 23:14:16 its angstrom/gold linker May 22 23:14:22 and do you have the patch I provided on top May 22 23:14:26 no May 22 23:14:29 and its armv7 I suppose May 22 23:14:30 I've no patches May 22 23:14:40 ok so plain OE-Core ? May 22 23:14:44 plain SHR May 22 23:14:55 ok which uses OE-Core May 22 23:14:59 I've removed the patch you gave me for firefox May 22 23:15:02 ok May 22 23:15:13 so what you did is you updarted chromium May 22 23:15:16 and thats it May 22 23:15:19 yes May 22 23:15:28 I suspect chromium linking scripts May 22 23:15:35 they do nasty stuff with _start() May 22 23:16:08 let me explain the problem first tough May 22 23:16:09 ok May 22 23:16:13 plz May 22 23:16:27 the debug symbols are nearly loaded May 22 23:16:29 ok loaded May 22 23:18:12 http://pastie.org/private/8gc2am3ekm5uwfecffxk9g May 22 23:18:37 first we see that #0 0x00206768 in ?? () is the same than _init in nm (00206768 T _init) May 22 23:19:06 hmmm May 22 23:19:07 http://pastie.org/private/nofo6bgc8pwdysefckwoq May 22 23:19:09 dont go by that May 22 23:19:15 set a bp on __libc_csu_init May 22 23:19:17 and then step May 22 23:19:19 then we see clearly here that it says _init() May 22 23:19:27 but.... May 22 23:19:57 http://pastie.org/private/tqmmpi7nluvqqok44r6wa May 22 23:20:09 here it shows clearly that _init is at another address May 22 23:20:16 I'll restart the debugging with the bp May 22 23:20:42 yes this is the real address of init May 22 23:20:54 what you are seeing is probably non virtual May 22 23:21:02 ok May 22 23:21:08 ah right May 22 23:21:10 sorry May 22 23:21:22 so go by rules May 22 23:21:23 but why does it want to jump to it then May 22 23:21:24 ok May 22 23:21:26 I've the bp May 22 23:21:33 are you stopped there May 22 23:21:35 Breakpoint 1, __libc_csu_init (argc=1, argc@entry=-1239158784, argv=0xbefffd24, argv@entry=0xb612db20, envp=0xbefffd2c) at elf-init.c:107 May 22 23:21:40 ok May 22 23:21:45 then do si May 22 23:21:58 and keep going until it jumps to that weird address May 22 23:22:02 ok May 22 23:22:13 it's not a problem if it's in C? May 22 23:22:18 and not in assembly? May 22 23:22:19 it seems to me some PLT relocation might have gone wrong May 22 23:22:44 and its on armv7 right May 22 23:22:47 http://pastie.org/private/rq5dkiijqmc9wkni9adcjg May 22 23:22:49 yes May 22 23:22:50 not on armv4 device May 22 23:22:52 ok May 22 23:23:07 and a very recent armv7 May 22 23:23:19 DM3730 May 22 23:23:41 (not n900 with some erratas) May 22 23:27:38 ok May 22 23:27:53 hmm, can you disable thumb2 in chromium May 22 23:28:25 there should be a include.gypi file May 22 23:28:30 in build tree May 22 23:28:34 yes I can May 22 23:28:35 where it should mention it May 22 23:28:42 'variables': { ... May 22 23:28:45 I'll try May 22 23:28:49 'arm_thumb': 1, May 22 23:29:02 make that 'arm_thumb': 0, May 22 23:29:19 what should I grep for in log.do_compile for thumb2 ? May 22 23:29:26 no May 22 23:29:37 search for that include fine in build tree of chromium May 22 23:29:41 and edit it May 22 23:29:42 ok May 22 23:29:48 yes I'll do that too May 22 23:29:55 but then I must be sure that it works May 22 23:30:10 because I had problem with defines for armv6 on another machine May 22 23:30:20 or add export GYP_DEFINES="armv7=1 arm_thumb=0" to recipe May 22 23:30:26 yes that worked May 22 23:30:34 the include.gypi didn't work May 22 23:30:38 for armv6 May 22 23:30:48 ok try same for armv7 too May 22 23:30:52 ok May 22 23:30:59 and if that works I would stay with that for a while May 22 23:31:03 until we fix gold May 22 23:31:05 ok May 22 23:31:14 to resolve thumb PLT entries properly May 22 23:31:16 let me look if it uses gold in the first place May 22 23:31:32 I think it does probably May 22 23:31:38 it should not fail with GNU ld May 22 23:31:57 yes it does May 22 23:32:18 does what ? May 22 23:32:28 it seem to use gold May 22 23:32:28 fail with gnu ld ? or uses gold May 22 23:32:31 ok May 22 23:32:36 I expected that May 22 23:32:43 altough I've something strange May 22 23:32:46 lrwxrwxrwx 1 gnutoo gnutoo 16 2012-05-21 21:38 ld -> /usr/bin/ld.gold May 22 23:33:02 in the /home/gnutoo/embedded/oe/oe-core/shr/oetmps/build/work/armv7a-vfp-neon-oe-linux-gnueabi/chromium/chromium-21.0.1145.0-r0 dir May 22 23:33:10 maybe it's for native stuff May 22 23:33:40 ok May 22 23:33:42 let me know how May 22 23:33:44 it goes May 22 23:33:47 I am off now May 22 23:33:47 yes May 22 23:33:52 ok thanks a lot!!!! May 22 23:33:52 need to hit the road May 22 23:33:54 np May 22 23:33:55 ok May 22 23:34:07 I should really learn the stuff you know.... May 22 23:34:19 like with stuff like that: http://www.linuxjournal.com/node/1060/print May 22 23:34:26 to begin with May 22 23:34:59 yes thats good stuff :) May 23 01:14:27 khem, it still segfault after disabling thumb May 23 01:19:48 khem, -mthumb-interwork was in the buiid tough May 23 01:21:01 and at some place there is -marm May 23 01:21:06 and at some other -mthumb May 23 02:32:59 GNUtoo-desktop: hmm. So it seems it really did not disable thumb code **** ENDING LOGGING AT Wed May 23 02:59:59 2012