**** BEGIN LOGGING AT Sun Aug 10 02:59:57 2008 Aug 10 03:57:06 hmm Aug 10 03:57:23 who update vala without checking whether programs in vala still work with the new release? Aug 10 03:57:28 *sigh* Aug 10 05:37:21 !logs Aug 10 05:37:21 Channel logs for #oe are archived at: Aug 10 05:37:22 http://hentges.net/tmp/logs/irc/%23oe Aug 10 05:37:23 Live-logs are available at Aug 10 05:37:25 http://hentges.net/tmp/logs/irc/livelogs/%23oe.livelog Aug 10 05:37:27 See ?? help-logs for usage instructions Aug 10 05:47:15 * hvontres|home drops a pin to see what happens Aug 10 05:48:00 * mickey|tv catches the pin and throws it back to hvontres|home Aug 10 05:48:39 ouch Aug 10 06:12:01 * mwester looks about, and notices a pin that someone seems to have dropped. Aug 10 06:38:23 * hvontres|home picks the pesky pin back up and wonders if it came from the network cable for amethyst Aug 10 07:34:18 What's up with the oe server/services? Or rather down.... Aug 10 07:38:01 I have been wondering if it would be better to concentrate on the "core business" of OpenEmbedded development and maintenance, and use one of the free hosting partners for keeping some of the online services in the air. Aug 10 07:38:57 I am not thinking of the one-size-fits-all gforge like solution, but rather something like dotsrc.org where projects like openembedded.org can have their special requirements implemented. Aug 10 08:14:09 hi Aug 10 08:14:17 anybody alive in here? Aug 10 08:14:44 zoff99: yes Aug 10 08:14:54 ok :) Aug 10 08:15:04 nobody alive on #poky i think Aug 10 10:04:11 Hello Aug 10 10:05:53 I'm trying to produce kernel-modules for my 2.6.24 - Kernel on armv4 platform. Aug 10 10:05:58  Some modules work, some don't. I get "unknown relocation: 40". Where can I start to find out what's going on? Aug 10 10:14:26 fx5: look to see what relocation #40 is, I suppose Aug 10 10:19:39 pb: thank you. I did not find it in my linux-tree but i found a document saying it is R_ARM_V4BX. This seems to encapsulate a special ARM-Instruction. But this does not help me. I don't know how to fix that. Aug 10 10:20:10 It says: R_ARM_V4BX records the location of an ARMv4t BX instruction. This enables a static linker to generate ARMv4 compatible images from ARMv4t objects that contain only ARM code by converting the instruction to MOV PC, r, where r is the register used in the BX instruction. See [AAPCS] for details. The symbol is unused and may even be unnamed. Aug 10 10:20:40 fx5: so, you need to either teach your insmod how to apply those relocations, or teach your toolchain to stop emitting them. Aug 10 10:20:59 if your target cpu is armv4t or higher, "applying" those relocations would just mean ignoring the, Aug 10 10:21:42 if your cpu is actually armv4, you will have to apply the reloc by patching the instruction in question back to "mov pc, rn" Aug 10 10:22:16 thank you. I have a v4t-CPU, so i will try to ignore them Aug 10 10:22:36 ok, great Aug 10 10:23:37 Maybe you have a hint where to start? Aug 10 10:24:05 well, find the place in the insmod source where the error is being generated, and figure out how to make it just ignore the reloc instead. Aug 10 10:24:32 * pb_ afk now Aug 10 10:57:37 morning all Aug 10 11:06:43 pb: Hooray! It works. Thank you! Aug 10 12:12:51 anybody that knows about status for the git.openembedded.net server? Aug 10 12:13:18 $ git pull Aug 10 12:13:18 git.openembedded.net[0: 80.81.242.146]: errno=Connection timed out Aug 10 12:13:18 fatal: unable to connect a socket (Connection timed out) Aug 10 12:13:18 esben@esben:~/oe.git/openembedded$ less .git/config Aug 10 12:33:51 it's offline for the same reason as the monotone server, i.e. that the machine hosting it is unreachable Aug 10 12:57:34 re Aug 10 13:02:41 RP: ping Aug 10 13:04:32 ant_: pong Aug 10 13:04:43 hello, no bike today? Aug 10 13:05:09 ant_: No, I did that on Thursday/Friday and am worn out ;-) Aug 10 13:05:34 RP: I got some oddities moving pstage to /oe/build/pstage (out of /build/tmp) Aug 10 13:05:45 now pastebin is out of order...let see Aug 10 13:06:00 yes, pastebin too! Aug 10 13:07:33 RP: manual copy & paste: http://rafb.net/p/3ul5UQ65.html Aug 10 13:08:38 ant_: How exactly did you move pstage? Aug 10 13:08:58 It looks like you left some trace of the old location behind and its confusing bitbake/opkg :( Aug 10 13:09:13 DEPLOY_DIR_PSTAGE = "/oe/build/pstage" in local.conf Aug 10 13:11:03 RP: interesting is also this: http://lists.linuxtogo.org/pipermail/angstrom-distro-devel/2008-July/002416.html Aug 10 13:14:01 it seems there are issues moving directories out of /oe/build/tmp Aug 10 13:14:05 Changing it like that should work in theory Aug 10 13:31:09 is there a public available git clone/mirror of git.openembedded.net somewhere? Aug 10 14:49:55 Is there any ETA on oe.org coming back up? Aug 10 14:59:10 NAiL, I am hopeful things will get worked out Monday Aug 10 14:59:22 but I have no real basis for this optimism :( Aug 10 15:03:22 morning all Aug 10 15:04:31 * Laibsch wonders about the attitude of a hoster that apparently does not give a sh1t for a couple of days Aug 10 15:14:12 Laibsch: I previously spent a lot of my time working with the FOSS hosting site that were previously sunsite.dk, and is now dotsrc.org. Aug 10 15:14:59 I believe it could be of great benefit for openembedded.org to team up with them, as they are specialized exactly at providing special hosting setups for special FOSS projects (like xemacs.org, wxwidgets and a lot of others). Aug 10 15:20:53 the server is in Berlin Aug 10 15:21:02 it belongs to mickeyl Aug 10 15:21:25 I guess after this is over, he'll sit down and think if there is a need for action or not Aug 10 15:22:07 Esben: does dotsrc.org host complete servers? Aug 10 15:22:36 we just need a hot-backup, I guess Aug 10 15:23:07 not only for git/monontone, but for the website and other essential services Aug 10 15:23:29 bugzilla comes to mind ;-) Aug 10 15:23:54 And we need a reliable hoster Aug 10 15:29:00 dotsrc.org has a bunch of servers, and they are dedicated to services and not to projects. Aug 10 15:29:34 so one of them is webserver, another is holding FOSS mirrors, one is carrying usenet news, another one is hosting mail, and so on... Aug 10 15:29:58 Implementing hot standby for things like bugzilla is pretty hard. I'm not sure this is a realistic aspiration for openembedded.org; there are probably better things for mickeyl to spend his time on. Aug 10 15:32:00 the servers are hosted in proper server room at the danish research network, with a really big fat pipe to the internet, UPS, airconditioning and all it takes :-) Aug 10 15:32:10 pb_: I think so also. Aug 10 15:32:42 it would be easier to just relocate services to servers that have better uptime, and not waste openembedded.org developers time maintaining servers. Aug 10 15:35:04 pb_: actually, I think bugzilla already has some kind of mirroring code built in. The database can be replicated. Aug 10 15:35:31 bugzilla is mainly my responsibility, I think it's been ages since mickeyl touched it as admin Aug 10 15:36:01 with a replicated database, one could have a hotspare working rather quickly Aug 10 15:36:22 But I agree, it might be better to invest the energy into getting reliable hosting Aug 10 15:36:40 Laibsch: it does support database replication, but I don't think it has any HA features. The replication is (or was, last time I looked at it) just there to allow you to load-share across multiple servers. Aug 10 15:36:46 I think the main thing is to shorten the chain of people involved in hosting so "we" can quickly find the people who need to solve the problem Aug 10 15:36:50 Esben: I am pretty sure OE.org wants its own servers Aug 10 15:36:59 in particular, so that you could run your complicated and expensive read queries on a different server to the one that was handling all the write traffic Aug 10 15:37:17 pb_: OK Aug 10 15:39:04 Crofton|work: Not sure the chain is too complicated. The way I understand it, the hoster is cut off. And we knew about that fairly quickly. And I assume the hoster did, too. Even if they did not, not reestablishing connectivity for days tells me volumes about that hoster. Aug 10 15:39:23 My understanding is that it is not only OE that is affected. Aug 10 15:42:31 well, adjacent IPs are still pingable. It does look to me like it's just that one server that is down. Aug 10 15:42:41 But what can you do if the hoster turns out to be a lazy bum? My understanding is this problem was not about not being able to pinpoint quickly the person in charge of fixing things. The problem is the person in charge sits on his a** Aug 10 15:43:23 OK, just regurgitating info I got. If that info is not true, then it would become another story. Aug 10 15:43:29 afaik, the problem is simply that the data center will only deal with harald, because he is their customer, and he hasn't had time to take it up with them. Aug 10 15:44:08 OK, then indeed that would change things. I was told he dealt with them on Thursday and Friday. Aug 10 15:45:03 well, not that I have any first hand knowledge of the situation either. mickeyl would be the man who knows. Aug 10 15:49:04 That was my recollection of things I believe I heard from him (being careful here, my recollection is notoriously bad). I looked through my IRC logs, but did not find the passage. Aug 10 16:18:39 pb_ is right. Harald could not act before saturday and I did not hear back from him yet. I will not stress him on that. He mentioned he considers moving it to another computing centre if these outages don't stop. Aug 10 17:19:39 OK, thanks for this update Aug 10 19:20:45 anyone has problem compiling openssl-native on AMD64? Aug 10 19:30:47 do I recall someone talking about udev problems on spitz the other day? Aug 10 19:52:33 ant_: ping Aug 10 19:53:12 bluelightning: yes. There is something in the bugtracker about this. Aug 10 19:53:53 thesing: hi, is your framebuffer menu in OE tree somewhere? Aug 10 19:53:53 thesing: presumably this is why I don't get /dev/input/touchscreen0 anymore? Aug 10 19:54:12 hehe.. seems like udev trouble? Aug 10 19:54:32 bluelightning: yes. If you run udevtrigger it should come back. Aug 10 19:55:03 hmm ok, any idea what the proper fix for this is? Aug 10 19:55:17 Jay7: it is. Its the kexecboot package. If you want the source code: git.linuxtogo.org Aug 10 19:55:53 bluelightning: no idea. I heard that hrw fixed it by upgrading to 124 in poky. Aug 10 19:56:09 bluelightning: downgrade to udev 0.92 Aug 10 19:56:32 Jay7: ok, I'll try that, thanks Aug 10 19:56:46 It should work as I remember Aug 10 20:09:07 re Aug 10 20:09:22 Jay7: pong Aug 10 20:10:15 ant_: hi! I thinking now about installing u-boot + bootmenu image Aug 10 20:10:40 hi, I tried, read this: Aug 10 20:10:44 thesing: ping Aug 10 20:10:47 we should provide something like pdaxrom Aug 10 20:11:15 but there was emergency image that installs rootfs Aug 10 20:12:04 we (should) have no special emergency image Aug 10 20:12:06 Jay7: I think we could create a new emergency image using kexec-bootmenu as base and adding mtd-utils and nandlogical Aug 10 20:12:16 yep Aug 10 20:12:31 or even busybox / uclibc Aug 10 20:12:54 thesing: I suppose the kernel cmdline u-boot sends somehow blocks the bootmenu-init... Aug 10 20:13:02 I think exactly the same :) with busybox is better for clamshells at least Aug 10 20:13:13 e.g cat /proc/cmdline -> console=ttyS0,115200 console=tty1 root=/dev/mtdblock2 rootfstype=jffs2 mtdparts=sharpsl-nand:7168k(smf),98304k(root),-(home) Aug 10 20:13:27 that's from u-boot Aug 10 20:13:53 Jay7: well, good news...the ATAG seems ok too... Aug 10 20:14:01 ant_: wow Aug 10 20:14:07 Jay7: cat /proc/atags -> AT ...garbled... contains cmdline as above Aug 10 20:14:23 but we definitely need some qualified help Aug 10 20:14:32 ah.. no, I have seen this while experiments Aug 10 20:15:05 but kernel still complains about atags Aug 10 20:15:17 seen my logs ? Aug 10 20:15:23 not sure Aug 10 20:15:29 4 times 'unknown' Aug 10 20:15:35 yes Aug 10 20:15:52 I've seen that 'unknown' lines Aug 10 20:15:56 who knows... Aug 10 20:16:15 we needed kernel developer :) Aug 10 20:16:43 Jay7: I'm trying to get in touch with InSearchOf, who seems to be developing pdaxrom together with sash Aug 10 20:16:49 or may be we should look at new u-boot versions Aug 10 20:17:08 good news Aug 10 20:17:45 he's sometimes on OESF or on #pdaxrom Aug 10 20:17:58 s/sometimes/rarely/g Aug 10 20:18:08 s/rarely/almost never/g Aug 10 20:18:13 ;-] Aug 10 20:18:59 but in order to test thesing's work, I suppose we should 'empty' the bootargs cmdline Aug 10 20:19:19 passed by u-boot Aug 10 20:19:43 at least root=/dev/mtdblock2 ... Aug 10 20:19:50 :) Aug 10 20:20:41 hm.. can kexec pass args to newly started kernel? Aug 10 20:20:50 Jay7: could you find out why the uImage is marked XIP ? Aug 10 20:21:24 IIRC yes, by the means of ATAGS Aug 10 20:22:25 well, the cmdline for collie is even a file in /boot Aug 10 20:23:33 ant_: then we can safely clean u-boot cmdargs and set args from something calling kexec Aug 10 20:25:31 it looks like... Aug 10 20:26:49 ant_: it likes very good Aug 10 20:27:26 we can flash u-boot with default args Aug 10 20:27:53 Jay7: could work with 2.6.24 / 2.6.26 Aug 10 20:28:05 did not when I tried with 2.6.23 Aug 10 20:28:07 and set args via some interface Aug 10 20:28:30 Jay7: the best would be edit the u-boot environment Aug 10 20:28:41 there is already a tool for this in OE Aug 10 20:29:19 but then we should restore the old partitioning like this http://www.pdaxrom.org/?q=node/89 Aug 10 20:29:43 it seems the majority of boards are using custom partitioning Aug 10 20:30:03 (when using u-boot) Aug 10 20:30:34 but why passing args to kexec'ed kernel from u-boot args? Aug 10 20:30:47 s/args/env/ Aug 10 20:31:03 s/args/env\?$/ :) Aug 10 20:31:39 well, the issue could be mtdparts Aug 10 20:31:47 I like grub boot loader - there is command line :) Aug 10 20:32:05 we now can create something like via bootmenu Aug 10 20:32:31 Jay7: dunno if grub has been ported to arm Aug 10 20:32:47 ant_: I think it can't be ported Aug 10 20:33:35 I've saying about behaviour Aug 10 20:34:39 let u-boot just to boot emergency/bootmenu kernel Aug 10 20:35:06 hm.. I have understood why we needed partitioning args from u-boot Aug 10 20:36:01 Jay7: remember the kernel in pdaxrom emergency is old and doesn't recognize the partitioning. It is just hardcoded (no /home) Aug 10 20:36:17 try cat /proc/mtd Aug 10 20:37:12 I means our new super-kernel with super-initramfs :) Aug 10 20:37:18 yes, with a recent kernel, u-boot environment could be the only place where mtdparts are defined Aug 10 20:37:56 Jay7: you badly need that serial cable...! Aug 10 20:38:06 yeah.. Aug 10 20:39:05 anyway, I think we could extend the RFC and post a mtn diff of u-boot_git, kernel.bbclass and c7x0,conf + akita.conf Aug 10 20:39:18 ups..not kernel.bbclass Aug 10 20:39:27 zaurus-2.6.inc Aug 10 20:39:33 If cable does not contain active items I cat solder some by self :) Aug 10 20:39:34 (your fixes) Aug 10 20:39:50 well...don't risk to fry the Z! Aug 10 20:40:01 zaurus-2.6.inc can be safely merged Aug 10 20:40:10 +1 from me Aug 10 20:40:19 It should not break anything Aug 10 20:41:18 about u-boot, mtd parts and kexecboot - it seems like chicken-and-egg problem Aug 10 20:41:56 we should provide info that we can't provide.. Aug 10 20:43:36 may be just forget about root and home parts? Aug 10 20:44:13 and provide 'logical' partition and single tail Aug 10 20:45:00 well, with kexecboot you don't need necessarly to boot mtd Aug 10 20:45:03 then we can re-set new partitions info for kexec'ed kernel Aug 10 20:45:12 you can bott sd, cf, nfs... Aug 10 20:45:37 each one with a different /root and/or /home Aug 10 20:47:20 Jay7: the problem is there is/was some specific code in zaurus kernel to skip bad informations passed by 'broken' botloaders (Sharp) Aug 10 20:47:49 oh.. Aug 10 20:47:53 now that infos are passed and have precedence over kernel cmdline Aug 10 20:48:07 I had a nice link discussing this... Aug 10 20:48:12 a mom Aug 10 20:48:24 too much points where booting can be broken Aug 10 20:48:57 should be this thread http://fixunix.com/kernel/243253-re-bugme-new-bug-9217-new-config_cmdline-doesnt-pass-kernel.html Aug 10 20:52:47 well, at first we should boot kernel from mtd ('logical') Aug 10 20:53:36 so we should provide some info about partitioning Aug 10 20:54:57 Jay7: at the end of the story, the truth is we only need kexecboot because u-boot is unfinished atm (no screen-menus) Aug 10 20:56:18 if it is for the first kernel booted by u-boot, well, that works almost perfectly... Aug 10 20:56:55 and remeber u-boot env can be edited from userspace Aug 10 20:58:32 (u-boot-utils DESCRIPTION = "U-boot bootloader OS env. access tools") Aug 10 20:59:00 Jay7: BTW the mkimage built by u-boot is 'native' Aug 10 20:59:35 hm.. Aug 10 20:59:37 I tried ldd against it and against the binary built by u-boot-utils-native Aug 10 20:59:49 try by yourself... Aug 10 20:59:58 is 686 in my case Aug 10 21:00:01 ant_: I know it Aug 10 21:00:39 strange..why the native utils as separate package then? Aug 10 21:01:16 (btw there are two of them...) Aug 10 21:01:17 just remove it from u-boot FILES_ Aug 10 21:01:31 and let it be in -native Aug 10 21:01:51 Jay7: I only patched u-boot_git Aug 10 21:02:10 and the other machines in it then? Aug 10 21:02:24 hm.. Aug 10 21:02:46 or, better question, is only that particular makefile building the tools too? Aug 10 21:04:43 don't know.. Aug 10 21:04:52 I've thinking about kexec :) Aug 10 21:05:09 right, is not too important atm Aug 10 21:05:28 just remember the uImage OE creates is XIP enabled Aug 10 21:05:33 well.. how kexec passess args to kernel? Aug 10 21:06:07 I'll try to find why it creates XIP enabled uImage Aug 10 21:06:34 Jay7: It builds the atag argument list in memory when it loads the new kernel, then when it passes control to the new kernel it sets the R2 register to point to the atags list that it wrote. Aug 10 21:06:51 but I'll do it tomorrow. Now is kexec :) Aug 10 21:07:04 Jay7: probably mkimage in OE has a -x switch... Aug 10 21:07:09 So if your kernel does not honor, or overrides, the atag list passed in, then it will not see the atags from kexec -- there are some kernels that do that apparently. Aug 10 21:07:32 ant_: I've grep'ed agaings mkimage and did not found any... Aug 10 21:08:01 if you inherits rm_work is only packaged Aug 10 21:08:13 mwester: 10x! now It's cleaner :) Aug 10 21:08:50 mwester: probably the old version of u-boot we are using uses the old addressing Aug 10 21:09:26 oops Aug 10 21:09:31 s/cleaner/plainly/ :) Aug 10 21:09:50 but the fact that cmdline appears in /proc/atags makes me think it works... Aug 10 21:09:55 I really need to upgrade my english :) Aug 10 21:10:44 ant_: kernel peoples says that atags have precedence over cmdline parameters Aug 10 21:10:58 and atags is not ignored Aug 10 21:11:04 yes..but not the unknown ones Aug 10 21:11:08 :-] Aug 10 21:11:24 so it seems like u-boot makes wrong atags :) Aug 10 21:11:40 he, only one expert could confirm Aug 10 21:12:10 there are some declarations in u-boot machine code Aug 10 21:12:20 for akita / corgi Aug 10 21:13:20 but it was 2006 and at the time our zaurus kernel was probably ignoring atags Aug 10 21:13:38 surely Aug 10 21:13:48 so nobody noticed it Aug 10 21:14:51 someone should compare old and current u-boot code.. Aug 10 21:15:43 Jay7: All, Aug 10 21:15:44 So after attempting to port the code from and earlier uboot "nightly Aug 10 21:15:44 build" from 2006 to the latest 1.3.3 from 2008, I'm running in to Aug 10 21:15:44 config issue. "stuff isn't matching up" So it will be a little bit Aug 10 21:15:44 more of a headache. I will be continuing to work on this for the Aug 10 21:15:45 remainder of the day (prior to the preseason football game :-)) Aug 10 21:16:07 ^^^ pdaxrom-everyone@lists.sourceforge.net Aug 10 21:17:48 did he continued? :) Aug 10 21:18:09 was a couple of days ago Aug 10 21:18:32 who is this? Aug 10 21:18:33 Aug 4 Aug 10 21:18:41 Late Aug 10 21:18:41 InSearchOf Aug 10 21:19:10 http://www.pdaxrom.org/?q=node/240 Aug 10 21:19:12 hmm.. It is good :) Aug 10 21:20:11 I told him here we have testers for the untested Zaurus Aug 10 21:20:29 and we already have kexecboot Aug 10 21:20:30 *Building uboot menu based on OpenMokos boot menu Aug 10 21:20:40 that's strange, isn't? Aug 10 21:20:56 may be it's solution? :) Aug 10 21:21:29 dunno, hrw was pessimistic (c7x0 needs a video driver for ati imageon) Aug 10 21:21:39 pxa27x can use fb Aug 10 21:22:06 yeah.. bootmenu initramfs seems better Aug 10 21:22:37 (not for boot time...) Aug 10 21:22:41 so.. Aug 10 21:22:56 (for menu system) :) Aug 10 21:23:01 repost again: http://blog.haerwu.biz/2007/12/26/zaurus-bootloader-starts-to-be-a-real-pita/ Aug 10 21:23:23 I've read this already :) Aug 10 21:23:55 is not for you, is to show this has been already long discussed Aug 10 21:24:22 btw, tosa/poodle is out of a game? Aug 10 21:24:50 no, HvonTres is working on poodle (kexec) Aug 10 21:25:08 but with sharp loader only? Aug 10 21:25:29 yes Aug 10 21:25:40 tosa /u-boot seems to be work in progress @pdaxrom Aug 10 21:27:15 Jay7: here some OM exemples http://wiki.openmoko.org/wiki/U-boot Aug 10 21:30:17 hi all Aug 10 21:33:09 ant_: It seems like right solution is to edit u-boot env.. Aug 10 21:34:15 I'll retry to save the env..perhaps I did some mistakes with mtd naming last time Aug 10 21:34:16 then we can provide right mtd partitioning for nand* utils Aug 10 21:34:43 hi florian Aug 10 21:34:50 and for kexec'ing kernel from mtd Aug 10 21:38:17 btw, usbtty on OM is good thing Aug 10 21:43:44 well, now I'll go sleep :) Aug 10 21:44:10 bye Aug 10 21:44:15 me too Aug 10 21:44:19 cu Aug 10 21:44:46 see (read) you tomorrow :) Aug 10 21:44:52 bb **** ENDING LOGGING AT Mon Aug 11 02:59:57 2008