**** BEGIN LOGGING AT Tue May 31 23:59:56 2005 Jun 01 01:00:55 03jeanfabrice * 10unslung/ (9 files in 4 dirs): Promote wpa-supplicant exclusively for Wiley Jun 01 01:19:12 03jeanfabrice * 10unslung/ (4 files in 2 dirs): Add template conf file, postinst and prerm Jun 01 02:02:06 jbowler-zzz: dunno - I've never used upslug Jun 01 02:03:00 if you are confident in your improvements, then feel free to contact Christopher, and to check your improvements into CVS (we can always go back to the current version if Christopher doesn't agree with the changes). Jun 01 02:06:31 03jeanfabrice * 10unslung/ (make/wpa-supplicant.mk sources/wpa-supplicant/defconfig): Update config file to support Xsupplicant Jun 01 04:10:11 re: upslug. I've had some issues too, but if it fails, I just do again. Jun 01 05:38:16 Anyone seen ka6sox lately? Jun 01 05:40:48 anyone: on the wiki it says that the "Unslung packages build system now builds its own toolchain". What does this mean? I've been looking for an equivalent of http://www.nslu2-linux.org/wiki/HowTo/NativelyCompileUnslungPackages for setting up a CrossCompileUnslungPackages environment. Jun 01 05:41:15 Perhaps it's so obvious it doesn;t need documenting - but I'm feeling a little puzzled. Jun 01 05:42:05 (disclaimer: I've searched the wiki lots already) Is there a HowTo which covers setting up the cross build environment for package development for Unslung? Jun 01 05:42:33 jbot, seen ka6sox Jun 01 05:42:52 ka6sox was last seen on IRC in channel #openjtag, 17h 37m 33s ago, saying: 'I need to get on the road...long trip ahead.'. Jun 01 05:43:25 make toolchain ? Jun 01 05:44:05 dyoung: is that after getting the CVS cvs/unslung ? Jun 01 05:44:12 Yes Jun 01 05:44:40 It might be make crosstool, I dont remember. Jun 01 05:44:57 for native builds it may be differnt, I just dont remember. Jun 01 05:45:01 ok - thanks - I'll have to write wiki page when I get it figured out for people like me who haven't spent a lot of time round linux development. Jun 01 05:45:26 Yeah it is. I've got native builds working fine... I just need to get myself sorted with a crossbuild environment Jun 01 05:45:54 Right. toolchain depends on crosstool, so if you make toolchain it builds it. Jun 01 05:46:32 native building is for folks more heroic than I. Jun 01 05:47:12 *laughs* I have found it easier simply because I get faced down every time I consider setting up a linux environment on my laptop. Jun 01 05:47:20 Although I led the push for native building... I just dont have the patience to actually do it. :-) Jun 01 05:47:28 native building just takes ages that's all :-) Jun 01 05:48:17 I had it set up witha nohup build going and piping the output to a file in my webspace so I could just pull a URL to see the prgress when away from home. Jun 01 05:48:19 My one purpose for getting native stuff going in the first place was to make perl work. Jun 01 05:49:01 well it's a great tool for learning the basics of linux on the slug... Jun 01 05:49:08 I've learned loads with it. Jun 01 05:49:15 Cool. learning is great! Jun 01 05:49:41 it is. I find it limited by time. Jun 01 05:49:42 (and should be near the top of the list for wanting to do slug stuff anyways) Jun 01 05:50:21 the problem I'mm finding is the "learning dependencies" Jun 01 05:51:04 were you able to get your colinux env going okay? Jun 01 05:52:39 I wanted to have a secure are on my website... but appweb has some dependencies missing from the ipkg... so I had to find out how to build appweb... which led my to the native development env... which I need to test on croo-env... which means I need to set up a linux env on my laptop... which means I need to learn about co-linux TAP... which broke my XP box... which meant I started looking... Jun 01 05:52:41 ...at knoppix... which leads to setting up local hdd space etc etc etc... Jun 01 05:52:55 copperbeech, if youre starting from scratch, you may need to make directories too. Jun 01 05:53:40 I didnt have any issues making colinux work.... the TAP "just works". Jun 01 05:53:43 for me at least. Jun 01 05:54:00 I have actually solved all my problems with the native env, but to give back to the community I need a cross env to test with. Jun 01 05:54:04 but I linked it to my wireless connection, so that has been a source of some issues... Jun 01 05:54:33 I think you'll be a lot happier developing cross. Jun 01 05:54:40 I was setting up co-linux networking on my laptop which has a wireless network card, and needs to be resident on both my work network and also my home network (router is 192.168.0.1). Jun 01 05:54:41 it just saves a lot of time. Jun 01 05:55:15 Microsoft bridging causes lots of pain with 192.168.0.1 gateway... Jun 01 05:55:41 I know I can get it sorted, but it just takes so much time... Jun 01 05:56:12 My internal network is something other than 192.168.0.x, so thats probably why it "just worked" for me. Jun 01 05:56:26 The WAF keep me very busy and don't make it easy for long stints on the console :-) Jun 01 05:56:31 I know ms ICS likes to use 192.168.0.x.... Jun 01 05:56:40 WAF? Jun 01 05:56:45 Wife and Family Jun 01 05:56:49 :-) Jun 01 05:56:58 Thats a pretty important component. Jun 01 05:57:17 Thats why I was so interested when [g2] was speaking about "sluggix" Jun 01 05:57:27 I dont have that busy-ness. Jun 01 05:57:34 Sluggix is a reality. Jun 01 05:57:41 I've had knoppix working on a previous laptop Jun 01 05:57:57 and it sounds like such a great option... Jun 01 05:58:09 I have yet to test it out; but it is a great idea. Jun 01 05:58:21 I'd be up for testing it and giving it a go! Can we download it from anywhere? Jun 01 05:58:30 * dyoung-zzzz dunnos Jun 01 05:58:50 * copperbeech grins Jun 01 05:58:53 I think [g2] wanted to make it perfect before letting it loose. Jun 01 05:59:05 so maybe in "final testing stage" now. Jun 01 05:59:14 yes his work seems to be high quality Jun 01 05:59:45 I think we have a pretty good track record of not releasing broken stuff. :-) Jun 01 06:00:24 and it should be kept that way... I'm just eager to have a look at it :-) Jun 01 06:02:21 for dev stuff I ended up using a external USB drive. Then I can mount it with Knoppix or Colinux or ubuntu, or whatever. Jun 01 06:02:28 (on the laptop) Jun 01 06:02:42 Has ubuntu got a "live" distribution? Jun 01 06:02:48 theres a live cd. Jun 01 06:03:14 5.04 Jun 01 06:03:17 I had a look on their website this morning but didn't see it. Jun 01 06:03:24 I'll have anopther look. Jun 01 06:04:56 http://us.releases.ubuntu.com/releases/5.04/ubuntu-5.04-live-i386.iso Jun 01 06:05:00 I suspected there was a live ubuntu, because I ran a live Mono preview and I was sure that was running on ubuntu. My laptop has a USB CD drive and knoppix barfs on it wheras ubuntu (mono live) worked fine. Jun 01 06:05:39 I used the PPC livecd too on my mini. But man on the mini, its kinda sluggish. Jun 01 06:05:55 (no pun intended) Jun 01 06:06:53 * copperbeech grins Jun 01 06:07:29 downloading now - will have to try and find a CD writer to play around with it Jun 01 06:08:22 I like the idea of using an external HDD... Jun 01 06:09:08 it helps immensely for having the downloads on-hand. Jun 01 06:09:26 I was toying with an idea of mounting a network share on the slug via NFS so you could just sling in a CD and go... but [g2] pointed out you would take a huge performance hit on file access. Jun 01 06:09:40 My laptop spanks the heck out of my desktop; so it was kind of a no-brainer. Jun 01 06:09:50 I don't even had a desktop Jun 01 06:09:54 had=have Jun 01 06:10:03 The slug is my "server" Jun 01 06:10:37 see: http://www.copperbeech.pwp.blueyonder.co.uk/html/network.htm Jun 01 06:11:04 I have a blank: Is openslug using glibc by default or uclibc? Jun 01 06:11:28 "about the size of a DAT tape" ??! Jun 01 06:11:56 VoodooZ_Work, deafult is glibc Jun 01 06:12:16 yeah I can't remember where I saw that description - but I liked it :-) Jun 01 06:12:29 it's not a bad comparison. Jun 01 06:12:35 Maybe the size of a Gen1 MD recorder.... Jun 01 06:12:48 * copperbeech grins Jun 01 06:12:49 its like maybe the size of a Beta tape. Jun 01 06:13:00 betacam. Jun 01 06:13:18 "a small paperback book" ? Jun 01 06:13:20 heh Jun 01 06:13:27 dyoung-zzzz: thanks. Jun 01 06:13:48 quite a think paperback book. Jun 01 06:13:54 think=thick Jun 01 06:14:20 VoodooZ_Work, I think jbowler wrote up a howto for uclibc openslugging. Jun 01 06:16:31 yeah, I might have to think about it later as my flash is full with all the stuff included in openslug by default (plus what I need!) Jun 01 06:17:12 uclibc might save me a little space but then again, might cause more trouble... Jun 01 06:17:33 I use pthreads too. I don't know if uclibc has it. Jun 01 06:17:39 Sounds like you need the [g2] 16MB Flash Enabled slug. Jun 01 06:17:45 hehehe Jun 01 06:17:54 Yeah, that would be cool Jun 01 06:18:10 is it an easy hack? (what am I thinking!?) Jun 01 06:18:13 that is close to reality too. Jun 01 06:18:23 oh if you can solder good, you can do it today. Jun 01 06:18:26 Can you just piggy-back a second FLASH chip on? Jun 01 06:18:34 I'm not bad with the iron Jun 01 06:18:40 But no expert Jun 01 06:18:43 its a complete chip replacement. Jun 01 06:18:47 ouch Jun 01 06:18:56 the pin pitch on the part is insane Jun 01 06:18:58 You just drop-in a bigger chip then? Jun 01 06:19:02 I dunno if I want to attempt it myself. Jun 01 06:19:14 Yeah, the 128mbit part fits. Jun 01 06:19:36 you may have to jumper a open resistor pad too, but it should be do-able. Jun 01 06:19:43 yeah, the last time I did a 64TQFP (AVR) on a little daughterboard I ended up doing a poor man's reflow machine with my oven! :) Jun 01 06:20:08 64TQ makes me cringe. Jun 01 06:20:14 Nah. I'll have to pass as there's lots of junk install in openslug by default I could remove. Jun 01 06:20:26 Like all those useless filesystems! Jun 01 06:20:42 For a robot I only need ext2/3 really. Jun 01 06:20:56 yeah. Jun 01 06:21:00 In fact, I had to remove that reiserfs stuff yesterday to make it fit! Jun 01 06:21:26 That's the problem. everybody has different needs. ANd mine are the opposite of most people so... Jun 01 06:21:27 you should almost make the robotslug distro. Jun 01 06:21:39 too much work. Jun 01 06:21:46 hehe Jun 01 06:22:00 I've been wasting my time lately at making that USB2 cam of mine work with the OV511 drivers Jun 01 06:22:03 my needs are pretty minimalistic too. Jun 01 06:22:41 They only support RGB out so I have to figure out how to get YUV422 and there's lots of REgisters! Jun 01 06:22:53 * dyoung-zzzz ughs Jun 01 06:23:03 And that is assuming the Cypress FX2 bridge chip doesn't mess with the data in between! Jun 01 06:23:36 I guess It's a good thing as I've always been too lazy to learn how to program with USB. Jun 01 06:23:40 I was just reading an article about FPGA's and machine vision. Jun 01 06:23:48 where/ Jun 01 06:24:05 Definitely faster than s/w but not as flexible. Jun 01 06:24:15 it started in Xcell (from xilinx), but I started following links.... Jun 01 06:24:28 A good idea for the lower level processing like segmentation, filtering, thresholding, etc.. Jun 01 06:24:57 I was initially interested because of the signal processing involved. Jun 01 06:25:06 There's a guy from a robotics club in the states (Andrew Maxon???) that did a whole FPGA vision design. Jun 01 06:25:10 but started getting distracted with the possibilities. Jun 01 06:25:19 yeah. Damn possibilities! Jun 01 06:25:58 and now its 0330 (AM) and I'm damn tired. ;-) Jun 01 06:26:27 Here's the URL: Jun 01 06:26:48 http://www.users.uswest.net/~kmaxon/page/index.html Jun 01 06:26:55 Lots of goodies on his site Jun 01 06:26:58 cool, thanks! Jun 01 06:27:28 If you google elphel, you'll probably run accross some of the stuff iwas reading earlier. Jun 01 06:27:38 cool. thanks. Jun 01 06:28:19 i still have dreams of making a multi-phase walking robot. Jun 01 06:28:31 the mechanics of it hurt my head though. Jun 01 06:29:08 Not just the mechanics though! The software to time all this! The PAIN!!!! Make it stop! Jun 01 06:29:14 and I'm sure my machine shop friends are tired of listening to my endless bantor by now. Jun 01 06:29:58 hehe Jun 01 06:30:05 the solution is to make my own machine shop. ;-) Jun 01 06:30:32 another project... Perhaps for another lifetime! :) Jun 01 06:31:34 If you can get your hands on the Q2 Xcell mag, theres a lot of interesting stuff in it regarding machine vision. Jun 01 06:31:51 nice. Jun 01 06:32:40 nature call... Jun 01 06:33:32 One of my projects is to make a thing that knows how to detect and process the movements of ocean waves. so this is of interest to me. Jun 01 06:33:57 but I'm still at the point where its just plain hurting my head. Jun 01 06:34:18 * my feeble head Jun 01 06:38:22 waves? sounds complex Jun 01 06:39:44 wow! the elphel search gave good stuff. That company has nice articles Jun 01 06:40:20 nod Jun 01 08:14:46 VoodooZ_Work: remove /boot/zImage*, that saves 800k, then remove all the adds in openslug.conf and just add back in what you know you need. openslug will boot with everything in openslug.conf removed. Jun 01 08:15:26 <[g2]> jbowler, nice update to the main home page Jun 01 08:15:27 <[g2]> thx Jun 01 08:15:42 <[g2]> main Openslug home page that is Jun 01 08:16:28 <[g2]> I'd like to start a punchlist for the next OpenSlug release Jun 01 08:17:13 Yes - I can't see anything which is a real showstopper. Jun 01 08:18:03 <[g2]> the "nobody cared" is still an open issue Jun 01 08:18:04 Ah, it would be good to have upslug built automagically in the source tarball. Jun 01 08:18:18 <[g2]> updating is an isse Jun 01 08:18:25 <[g2]> I'd like to fix that Jun 01 08:18:34 updating? Jun 01 08:18:47 <[g2]> reflashing Jun 01 08:19:21 <[g2]> I think we are at the point that we can do it with turnup, ipkg, and kern_header and a little script no ? Jun 01 08:19:26 Do you mean reflashing from within OpenSlug or reflashing without losing system config (e.g. from upslug?)) Jun 01 08:19:33 Ok, question answered. Jun 01 08:19:50 <[g2]> that needs to be address also Jun 01 08:20:03 <[g2]> along with roll-back Jun 01 08:20:21 Yes. Reflash of the kernel is basically a single command (kern_header isn't necessary). Reflash of the jffs requires going to RAM. Jun 01 08:21:07 <[g2]> so how do you reflash the kernel without using kern_header ? Jun 01 08:21:14 Still, I don't think it can be done safely - something has to happen first. When you have a new kernel first reboot may crash without the modules, when you have a new jffs2 the same applies with the old kernel. Jun 01 08:22:38 This command: devio '<>/dev/mtdblock2;wb$16+4;fb12,0;cp$' Jun 01 08:23:12 Plus an 'fb#t-255' at the end to fill the flash with \0377 Jun 01 08:23:14 <[g2]> I was looking at devio yesterday Jun 01 08:23:24 Ah, I got carried away. Jun 01 08:23:24 <[g2]> you've done a bunch of work there Jun 01 08:23:46 <[g2]> we should put together a little "cheat sheet" Jun 01 08:24:06 Yes, definately, since the commands are so crytic. Jun 01 08:24:16 s/crytic/cryptic/ Jun 01 08:24:28 <[g2]> and so extensive :) Jun 01 08:24:46 It seemed to end up Turing equivalent... Jun 01 08:25:25 <[g2]> are you planning on releasing that as a stand-alone project ? Jun 01 08:26:30 (Hum, actually it might not be Turing equivalent). Stand-alone: it's possible, it seems to be generally useful for hacking files in place. Jun 01 08:26:51 * [g2] is honored to have it in openslug-init now, but it could easily be it's own package in OE (devio) Jun 01 08:27:10 <[g2]> and openslug could just use it as many other distros could also Jun 01 08:28:27 Do you use upslug? Jun 01 08:28:39 <[g2]> kern header was written to address a simple binary editing (without using sed and even more cryptic stuff) Jun 01 08:28:48 <[g2]> I've used it once Jun 01 08:29:05 How do you reflash normally? Jun 01 08:29:12 <[g2]> I have a tftp server with the r25 release split into the kernel and rootfs Jun 01 08:30:01 <[g2]> so I ctr-c; ip_addr ....; load ramsdisk; load kernel; boot; and flash from the UI Jun 01 08:30:27 I tried that but had problems setting up a satisfactory tftp server, so I just use upslug. Jun 01 08:30:40 <[g2]> I've done is *SO* many times I've been too lazy to change Jun 01 08:30:51 <[g2]> which tftp server ? Jun 01 08:30:59 <[g2]> Mine's been flawless Jun 01 08:31:03 So far as I can see a typical user will either end up using upslug (Linux) or Sercomm (Windows). Jun 01 08:31:26 <[g2]> actully I'm looking for the third option now :) Jun 01 08:31:39 tftp server: admin problems, and I like to upslug from the deploy/images directory, so I don't want to have to copy the image (lazy) Jun 01 08:32:10 Third option: reflash from within OpenSlug? Jun 01 08:32:16 <[g2]> nod. Jun 01 08:32:28 <[g2]> with APEX I'll be easy soon Jun 01 08:32:50 <[g2]> two full rootfs systems with booting out of the kernel in the root fs Jun 01 08:33:07 How about just being able to 'ipkg install openslug-image' and have it do everything? Jun 01 08:33:49 <[g2]> to start with that could work given the two conditions Jun 01 08:34:01 <[g2]> 1) The user has turnup to RAM Jun 01 08:34:10 <[g2]> 2) It's a known tested kernel Jun 01 08:34:24 (1): turnup anywhere not-flash will work Jun 01 08:34:49 <[g2]> that's true :) Jun 01 08:35:24 (2): yes, it's a release process, and it has the potential to render one or more old turnup rootfs's unusable. Jun 01 08:35:39 <[g2]> I'm just thinking in the degenerate case of running out of jffs2 the user will have to turnup to RAM first Jun 01 08:36:16 (1): I think it's easier to make it work from disk or nfs, because otherwise the user has to first do a 'turnup ram' and reboot. Jun 01 08:37:05 <[g2]> an important issue to me is the following: Jun 01 08:37:22 <[g2]> how persistent is the rootfs on the non-jffs2 partition ? Jun 01 08:38:20 <[g2]> I'm at the point where I'd like the option of using the /etc/default/sysconfig versus always reloading from the SysConf partition Jun 01 08:38:57 To my way of thinking the turnup target is the persistent store. flash is there as a recovery (it's a boot disk). Jun 01 08:39:39 Reloading SysConf each time is easy, but the files sysconfsetup edits are config files and the user may further edit them. Jun 01 08:40:22 <[g2]> right. and that's the point, on subsequent turnups the user will probably want to keep that data Jun 01 08:40:23 It's very difficult to roll user edits to (e.g.) /etc/network/interfaces back into SysConf, other than by tar'ing up the files and storing them in there. Jun 01 08:41:03 Yes, so 'ipkg install openslug-image' just upgrades - it respects config files. A complete reflash of the jffs2 doesn't (can't). Jun 01 08:41:10 <[g2]> that's where TLV's are nice Jun 01 08:41:17 <[g2]> Type, Length, Values Jun 01 08:41:44 Applied to what? The files in rootfs? Jun 01 08:41:53 <[g2]> no the SysConf info Jun 01 08:42:41 <[g2]> except for the length at the beginng they could all be considered strings Jun 01 08:43:36 <[g2]> maybe the solution is this Jun 01 08:44:06 This is like the NT profile (i.e. the database used inside NT to control system and user level configuration). Jun 01 08:44:07 <[g2]> add an option to "save off SysConf and the last 144 bytes of Redboot" Jun 01 08:44:33 I believe it's the last 80 bytes - the previous 64 are all zero. Jun 01 08:44:35 * [g2] is *blissfully* ignorant of NT :) Jun 01 08:45:12 <[g2]> nod on the last 80 but it turned out that was the truncation on 256... Jun 01 08:45:29 <[g2]> what is it 262144 = 256k ? Jun 01 08:46:03 <[g2]> yeah... that what the calc says Jun 01 08:46:07 Ah, so 144 is an error? It really is 80? Jun 01 08:46:47 <[g2]> I think there are only about 20 unique bytes but I haven't really verified it Jun 01 08:46:47 NT is a useful model because it's not UNIX but it faces all the same issues (I guess VMS would be another model) Jun 01 08:47:32 The last 80 contains the MAC, four unknown bytes (0,0,0,'A' on my slug) then a sercomm trailier Jun 01 08:47:51 <[g2]> there's also the firmware version in there Jun 01 08:47:54 The sercomm trailer contains a HW ID (which, I believe, gets places in sysconf) and various revision info. Jun 01 08:48:56 It's described in detail here: http://www.nslu2-linux.org/wiki/Info/SercommRedBootTrailer Jun 01 08:49:26 <[g2]> nod. Jun 01 08:49:56 <[g2]> I'm just saying that the last 80/144 bytes is enough to capture the unique Redboot portion Jun 01 08:50:21 <[g2]> it could be made smaller, but it's probably not worth it Jun 01 08:51:03 Yes, I think it is. It's easy to dump and restore, but that's only an issue for APEX - for RedBoot systems it's enough just to get the MAC and hw id out of those 80 bytes. Jun 01 08:51:43 <[g2]> so lets talk the Redboot path for now Jun 01 08:51:50 That could be done every time, but I think it's enough to check for an erased SysConf and load it then. Jun 01 08:52:38 <[g2]> Currently we don't modify sysconf Jun 01 08:52:53 <[g2]> we just use it as an input Jun 01 08:53:07 Yes, but if the user messes up /etc/network/interfaces (easy to do) there needs to be a recovery method... Jun 01 08:53:38 <[g2]> currently there are 2 main recovery mechanisms Jun 01 08:53:51 <[g2]> 1) pull the flash stick/ NFS mount / HD Jun 01 08:53:55 <[g2]> 2) Redboot telnet Jun 01 08:54:33 <[g2]> If somebody borks a kernel upgrade they still have #2 Jun 01 08:55:39 <[g2]> ALL opensluger's are supposed to be familiar with #2 so it really isn't that inconvient Jun 01 08:55:43 <[g2]> sp Jun 01 08:56:50 When there are binary-only users they won't be familiar with #2, at least not to the extent of setting up tftp to reflash the kernel. Jun 01 08:56:55 <[g2]> can we make a 128K ext2 fs ? Jun 01 08:57:18 <[g2]> there's an http option too Jun 01 08:57:22 It may be possible to load a kernel from upslug (without changing the ffs2) Jun 01 08:58:36 <[g2]> If we can make a 128K ext2 fs (which we probably can) then I'd propose this option Jun 01 08:58:45 128k - are you thinking of SysConf? cpio is probably a better option. Jun 01 08:59:15 <[g2]> on turnup convert the SysConf partition into a tiny mount point Jun 01 08:59:33 <[g2]> have it store both the original SysConf info and our file Jun 01 08:59:40 You can't read/write into SysConf like a file system, so there's no point making it a file system - better to use it as an archive (tar, cpio) Jun 01 09:00:25 It's trivial to append data after the current sysconf, and there's about 110K available. Jun 01 09:00:56 03jp30 * 10unslung/ (make/openldap.mk sources/openldap/install.patch): openldap: remove static libs, strip Jun 01 09:01:53 <[g2]> jbowler, I'd like the /etc/default/sysconf to be persistent across jffs2 flashes Jun 01 09:02:32 <[g2]> I think if we convert the SysConf in to a 1 block FS then we can store both the original and our converted data in there Jun 01 09:02:41 <[g2]> and be able to convert it back on demand Jun 01 09:03:13 Yes, but it doesn't have to be a file system, indeed it's a lot of work to make it a file system. It's 128k of data which is written in one shot. Jun 01 09:03:41 <[g2]> I think it's trival to create it as a file system Jun 01 09:04:39 * [g2] wonders whether mke2fs /dev/mntblock1 will work or not Jun 01 09:04:42 It would have to be a flash file system - it isn't possible to erase bits, only append! Jun 01 09:05:47 mke2fs will work, block access is fine. There are two blocks - the device has 64k blocks. Try it, you may have to buy new flash after a couple of days ;-) Jun 01 09:05:47 <[g2]> that's a good point Jun 01 09:06:15 <[g2]> not if it's mounted noatime Jun 01 09:06:25 Hello there Jun 01 09:06:28 <[g2]> and I'm not planning on updating it a bunch Jun 01 09:06:40 <[g2]> hey CB Jun 01 09:06:42 So the alternative is just to treat it as a sequence of , or maybe as you suggested Jun 01 09:07:05 Im making progress on my slug-carputer Jun 01 09:07:05 <[g2]> does jffs2 fit in 1 block ? Jun 01 09:07:25 Still looking for someone to compile usbsound though. :P Jun 01 09:07:36 <[g2]> CB cool, jbowler and I are discussing OpenSlug upgrade paths right now Jun 01 09:07:51 ah, good Jun 01 09:08:26 * CompBrain reads scrollback Jun 01 09:08:37 <[g2]> or ext2 ro and memory backed with an option to update the flash Jun 01 09:08:43 jffs2: I don't know, probably not because it can't safely erase the block when the journal fills it. Jun 01 09:09:32 I figured 'turnup preserve' would save the system settings and 'turnup restore' would load them back. Jun 01 09:09:45 'turnup restore' would happen automatically on first boot. Jun 01 09:10:02 <[g2]> and if it failed load from sysconf Jun 01 09:10:17 Yes. Jun 01 09:10:37 <[g2]> OK... I"m cool with that Jun 01 09:11:02 ... i.e. if it failed load from /etc/default/sysconf - since the data would be in SysConf if that got erased it's back to the initial settings. Jun 01 09:11:04 <[g2]> we're only flashing when the user says turnup preserve Jun 01 09:11:37 Yes - and 'turnup flash-new-kernel' would say something like 'have you run turnup preserve?' Jun 01 09:12:13 I still like the idea of ipkg upgrade just working if the base packages are upgraded. Jun 01 09:12:57 However it happens, there needs to be a list of config files to preserve, and it probably needs to be possible for the user to add to it. Jun 01 09:12:59 <[g2]> well that would work fine with the turnup preserve/restore concept Jun 01 09:13:42 <[g2]> I think the users need to manage their own files Jun 01 09:14:02 <[g2]> I think we have our hands full enough just managing this stuff :) Jun 01 09:14:33 <[g2]> without dealing with stuff like you're trying to save 160K and we only have 110K Jun 01 09:15:28 Right: it should be sufficient to provide a way of preserving system settings across a reflash. Individual user configurations may require information from the user about what to preserve. Jun 01 09:15:48 03jp30 * 10unslung/make/cyrus-sasl.mk: cyrus-sasl: strip Jun 01 09:15:49 <[g2]> we'll fry those bigger fish later Jun 01 09:16:05 <[g2]> jp is a *mad man*' Jun 01 09:16:24 Do you want to have basic preserve/restore commands before the next release? Jun 01 09:16:42 <[g2]> I'd like to see that Jun 01 09:16:56 <[g2]> along with a kernel upgrade Jun 01 09:17:02 <[g2]> that works Jun 01 09:17:52 <[g2]> however I think we need to ponder this for a few minutes Jun 01 09:18:10 Ah. Well, upgrading mtdblock2 is easy, but getting /lib/modules and all the tools right may be more difficult. Jun 01 09:18:12 <[g2]> I think there are two conflicting paths Jun 01 09:18:52 <[g2]> The easy path for now is Jun 01 09:19:05 <[g2]> turnup preserve Jun 01 09:19:12 <[g2]> and reflash new rootfs Jun 01 09:19:25 <[g2]> then turnup restore .... device .... Jun 01 09:19:41 <[g2]> and ipkg re-install stuff Jun 01 09:20:07 03jp30 * 10unslung/make/libdb.mk: libdb: fix strip Jun 01 09:20:14 <[g2]> The tough path would be Jun 01 09:20:20 <[g2]> ipkg upgrade Jun 01 09:20:29 <[g2]> and then allow Jun 01 09:20:38 <[g2]> 1) warm reboot to test a kernel from memory Jun 01 09:21:04 <[g2]> 2) be able to build the jffs2 partition and flash it to match the running system Jun 01 09:21:29 <[g2]> 3) update the kernel which is not to run with the externel upgraded device Jun 01 09:21:41 <[g2]> s/not/known/ Jun 01 09:22:05 Do you know how to do (1)? Jun 01 09:22:45 (Without doing it from redboot) Jun 01 09:32:14 I think for the next release preserve/restore of known config files (user specifiable), under user control, is achievable and being able to reflash just the kernel (user must ensure valid modules/tools) is also possible, but anything which generate a new jffs2 (other than complete reflash) probably isn't. Jun 01 09:50:31 jacmet, ping Jun 01 10:02:50 <[g2]> jbowler-away, sorry I was away :( Jun 01 10:03:53 I'm still here... Jun 01 10:04:17 <[g2]> jbowler-away, sorry I went for a drink of water and was distracted Jun 01 10:04:43 The conclusion I came to is that I know how to do preserve/restore and even reflash a kernel, but all the other stuff is, well, scary... Jun 01 10:05:12 <[g2]> scary ? scary how ? Jun 01 10:06:12 On RedBoot it requires writing two separate pieces of information - kernel and jffs2 - which must be in sync. Jun 01 10:07:08 There's no advantage to doing that (over writing a single flash image) unless it's possible to preserve a whole lot of user config (more than can be preserved with 'turnup preserve' like stuff.) Jun 01 10:07:47 <[g2]> I've heard stories of other ppl replacing the kernel live from memory Jun 01 10:07:55 So some part of the system is 'live' - altered in ways which can't be predicted - and it's scary trying to guess how to change the other parts. Jun 01 10:08:36 Yes - replace the kernel *temporarily*, upgrade the jffs2 temporarily, test, write it back to flash if it works. That's not scary, just difficult! Jun 01 10:08:54 <[g2]> I've reflashed Redboot maybe 50+ times on my slug Jun 01 10:09:08 <[g2]> I switch between Redboot and APEX often Jun 01 10:09:29 <[g2]> and I don't even have JTAG on the slug yet Jun 01 10:09:50 <[g2]> I did have JTAG on the aviala board and tested there first Jun 01 10:10:05 <[g2]> avila Jun 01 10:10:38 <[g2]> after is was stable enough and working pretty well I flash the slug and it worked enough to boot the kernel and rootfs Jun 01 10:10:50 <[g2]> now it fully works Jun 01 10:10:52 Yes, but you probably don't have much config there. I anticipate an OpenSlug user will install a lot of packages and not be happy trying to reinstall them. Jun 01 10:11:21 <[g2]> well this is the upgrade path we need to consider Jun 01 10:11:39 Which one? I.e. which upgrade path? Jun 01 10:12:03 <[g2]> I mean we should try to consider supporting the users that have stuff already installed Jun 01 10:12:17 <[g2]> I think right now those users just do an ipkg upgrade Jun 01 10:12:24 Yes - I agree fully... but maybe not in the next release. Jun 01 10:12:32 <[g2]> right. Jun 01 10:12:47 <[g2]> for now... it's ipkg upgrade works for non-kernel packages Jun 01 10:13:06 Right... so it should be safe ;-) Jun 01 10:13:29 <[g2]> and this next pass will allow the preservastion of config data cross jffs2 reflashs (via the sysconf partition) Jun 01 10:13:43 <[g2]> but not support saving the install package state Jun 01 10:14:00 Yes, so then attempts to do something with more preservation always have good fallback (if they fail). Jun 01 10:14:09 <[g2]> however, it would allow manual kernel upgrading for the brave of heart and trail blazers Jun 01 10:14:14 ... reflash the whole thing... Jun 01 10:14:22 <[g2]> right Jun 01 10:14:57 The kernel reflash is good for source tarball users - it allows experimenting with different modules built in. Jun 01 10:15:31 <[g2]> well the other part of the release is all the openslug-packages build for me Jun 01 10:15:39 If it doesn't boot they either use redboot to put the old one back or reflash the whole thing... Jun 01 10:15:54 <[g2]> I'd like to straighten out the recipe for all the native compile / building Jun 01 10:16:04 openslug-packages: all download and build for me, but I had to may a local mysql_4.1.12.bb to get that. Jun 01 10:16:12 <[g2]> I've done it before (mostly) but need to productize it Jun 01 10:16:41 <[g2]> there's about 20-40 steps to make that stuff work after just being built and installed Jun 01 10:17:15 I'd also like to add upslug to the build, so that I can write a 'binary download' page which works easily. Jun 01 10:17:50 <[g2]> I think upslug is a user optional package Jun 01 10:18:10 <[g2]> I do not like the way upslug works Jun 01 10:18:14 <[g2]> or is written Jun 01 10:18:38 I meant an upslug-native - so that Windows users can use sercomm and linux users upslug in otherwise identical ways. Jun 01 10:18:52 <[g2]> right Jun 01 10:18:58 <[g2]> that's what I'm talking about Jun 01 10:19:17 [g2]: written? You mean the way it manages the Sercomm protocol, or the lack of any error checking? Jun 01 10:20:06 <[g2]> the C app is not clean at all. There are lots of host-to-network conversions for no real reason it's a mess Jun 01 10:20:36 <[g2]> besides the protocol works on an wierd IP protocol type which is non-standard Jun 01 10:21:09 <[g2]> or it might even be ethertype I forget Jun 01 10:22:08 The ntohs is in the VCI_TABLE stuff, it relies on the structure layout matching the stream byte layout, but it's correct apart from that. Jun 01 10:22:23 <[g2]> I think we could corrupt the rootfs and Redboot might wind up in upslug mode automagically Jun 01 10:22:59 The socket mode is PF_PACKET/SOCK_DGRAM, that's standard - same as DHCP? Jun 01 10:23:23 The slug itself gets into upgrade mode by holding the reset button for 10s during power up (hold it until the ready/status goes red). Jun 01 10:23:49 <[g2]> that's hard to do remotely :) Jun 01 10:24:01 That's good ;-) Jun 01 10:24:24 <[g2]> well with APEX I'm planning to have slugs out on the big I Jun 01 10:24:52 <[g2]> so the only upgrade mechanism will probably be one via SCP Jun 01 10:25:04 <[g2]> and there should be no back door Jun 01 10:25:23 <[g2]> I'm even considering only serial grade only Jun 01 10:25:34 That's a lot better. At present an NSLU2 is extremely non-secure... Jun 01 10:25:40 <[g2]> exactly Jun 01 10:26:09 <[g2]> you can upslug any non-openslug that you have net access to Jun 01 10:26:17 <[g2]> convienent for upgrades Jun 01 10:26:30 <[g2]> not so good for security Jun 01 10:27:18 With APEX will the last 80 bytes match the RedBoot ones (apart, maybe, from the sercomm trailer)? Jun 01 10:27:45 <[g2]> not likely Jun 01 10:27:57 <[g2]> they could possibly but I wouldn't count on it Jun 01 10:28:23 * [g2] still wonders if there some nv memory on chip somewhere Jun 01 10:28:36 <[g2]> hmmm.... Jun 01 10:28:54 <[g2]> I wonder what the x1206 costs and where it's located Jun 01 10:29:16 <[g2]> the bigger brother parts have like 4Kbits/bytes EEPROM Jun 01 10:29:32 * [g2] searches digikey Jun 01 10:31:01 Ok, so my plan is to work on preserve/restore, the 'how to use the binary bits' and 'source tarball' web pages and fix upslug. Jun 01 10:31:51 <[g2]> the 'how to use the binary bits' is for devio right ? Jun 01 10:32:18 The two non-existent pages on the openslug home page. Jun 01 10:32:36 <[g2]> ok if you want Jun 01 10:32:39 So I'll add 'how to use devio' somewhere. Jun 01 10:33:00 <[g2]> I'll probably have option 4 that I was going to add Jun 01 10:33:25 ? Jun 01 10:33:30 <[g2]> the liveDVD Jun 01 10:34:20 Oh yes - probably option 3, or maybe 2 (it's equivalent to build-from-source isn't it, but more reliable?) Jun 01 10:35:05 I.e. people who have systems which, for some reason, can't build from the source tarball will be able to use the LiveDVD. Jun 01 10:35:06 <[g2]> it's like 3, but known to produce good images Jun 01 10:35:30 <[g2]> the two issues I have with it right now are Jun 01 10:35:46 <[g2]> 1) where to create the partitions for swap and tmp Jun 01 10:36:05 <[g2]> 2) whether to try creating files on NTFS Jun 01 10:37:09 <[g2]> but I've built post beta images (openslug-image, unslung-image and openslug-packages) on a hd partition Jun 01 10:37:39 This sounds good. Jun 01 10:37:58 <[g2]> it is good and it just the beginning Jun 01 10:38:35 <[g2]> I'm planning on timing the next snapshot with the release Jun 01 10:39:02 <[g2]> so one would be able to build the binaries (openslug-image and unslung-image) to match the release Jun 01 10:39:20 <[g2]> and openslug-packages if we ant Jun 01 10:39:22 <[g2]> want Jun 01 10:39:42 <[g2]> So I'm gonna build for both glibc and uclibc Jun 01 10:40:09 <[g2]> This release I'm currently calling the (BigEndian) release Jun 01 10:40:55 <[g2]> also the reboot issue needs to be fixed Jun 01 10:41:18 That one is much harder I think. Jun 01 10:41:19 <[g2]> even something like priming the dog to bite Jun 01 10:42:08 * [g2] thinks we are looping in our code which causes the "lockup" Jun 01 10:42:32 There's the reboot hang, the behavior of halt (doesn't seem to power down, may be the same thing but it doesn't hang) and the irq problem. Jun 01 10:42:46 <[g2]> right Jun 01 10:42:57 <[g2]> the power-down is two issue Jun 01 10:43:11 <[g2]> a) we keep executing code when we shouldn't Jun 01 10:43:23 <[g2]> b) we don't drive the GPIO Jun 01 10:43:37 <[g2]> I want to fix a) because it should be fixed Jun 01 10:43:42 I know of no way of making progress on any of these - if I was to do anything it would be to try v2.0 of the Intel code, but that's a big change. Jun 01 10:43:48 <[g2]> and b) assures of a power down Jun 01 10:44:17 Well, halt keeps flashing the LED, so fixing (b) may just work. Jun 01 10:44:18 <[g2]> is the V2.0 really out ? Jun 01 10:44:30 v2.0: yes, May 9 iirc Jun 01 10:44:41 <[g2]> b) will definitly work Jun 01 10:45:11 The problem is that 2.0 pulls the NPE microcode from a separate file, which has to be fed to a character device. Jun 01 10:45:21 I suspect they did this to reduce memory size. Jun 01 10:45:34 <[g2]> what's the license on the 2.0 code ? Jun 01 10:45:56 license: it seems to be the same, but I didn't run a diff. Jun 01 10:46:49 <[g2]> well currently, I'm not playing much with the IAL/csr stuff Jun 01 10:48:05 <[g2]> I think moving to 1.5 for the LE would be next based on Jacmet's progress Jun 01 10:49:31 Ok, gotta go now - probably back in about 18 hours... Jun 01 10:49:40 <[g2]> cheers.... Jun 01 10:49:49 <[g2]> I think we're all clear on what's next Jun 01 10:51:53 <[g2]> hmm... the X1226 with 4K eeprom is only 2.85 in Q 25 :) Jun 01 11:25:30 llagendijk, i checked in the changes needed for apr and apr-util to build for wiley. now apache and svn will build, but fail at runtime Jun 01 11:27:46 (fail at runtime on wiley that is) Jun 01 11:40:17 jp30-work, will have a look when I have time (too much work :-( ....) Jun 01 11:41:35 jp30-work, does wiley have enough memory to run apache? Jun 01 11:43:53 wl500gx does Jun 01 12:29:14 jp30-work, are you saying ...about an hour ago...that you are actively working on making apache run on wiley ? Jun 01 12:29:28 What about sendmail / postfix - does anyone do any work in that direction ? Jun 01 12:30:03 all I really "need" for wiley to be a complete server-solution for my needs, is some kind of mailserver, and mysql (or postgresql) Jun 01 12:34:27 macsat, yes jp30 is working on apache, it compiles, but crashes Jun 01 12:35:29 macsat, for a good server, we first need bind now I think (but I will work on that) Jun 01 12:52:09 03jbowler * 10upslug/ (upslug.c upslug.h Makefile): Jun 01 12:52:09 Changes to make the upload more reliable. It should no longer fail on Jun 01 12:52:09 busy networks. More extensive error messages for debugging any problems, Jun 01 12:52:09 all system calls are now error checked and only the known-to-be-ok error Jun 01 12:52:09 returns are accepted. Jun 01 12:58:49 jp30-work, did you make changes to expat to get apr-util to compile? Jun 01 13:00:54 not that i remember Jun 01 13:02:16 the only recent change i've made to expat was to remove unnecessary material from the ipk Jun 01 13:02:48 funny, configure for apr-util can not find expat... It looks for a STATIC library, not a shared object Jun 01 13:05:02 hmm... i'll look into that. back in 15 min Jun 01 13:05:47 jp30-work: for the logs, got your mail - not problem at all, feel free o do changes that seems usefill Jun 01 13:05:58 s/o/to Jun 01 13:11:46 llagendijk, bind ? Jun 01 13:12:09 wht would you want to include a DNS server? Jun 01 13:12:32 most people are using it on private lines, where their ISP's have perfectly working DNS servers, dont you think ? Jun 01 13:13:32 macsat, true, but I run an own domains (external as well as internal) Jun 01 13:13:56 oki :) Jun 01 13:14:23 a mailserver would really be something I could use....but I dont have the skills to get the damn thing to cross compile I fear... Jun 01 13:14:44 smtp/pop/imap would be a nice combo....sounds like qmail dosnt it ?....but that one is kindda hard to get running. Jun 01 13:14:54 I guess sendmail/postfix could do the job as well Jun 01 13:15:16 macsat, DHCP with dynamic updates of DNS works very nice for internal traffic Jun 01 13:16:13 ok Jun 01 13:16:32 macsat, I am still running sendmail here. I am anyhow not sure whether that type of thing will run on wiley. Not enough memory? Jun 01 13:17:00 I am not even sure if that is a problem for apache + friends Jun 01 13:18:44 I believe there is actuately something called "mini-sendmail" or something...a small-scale version. Jun 01 13:19:03 the gx version of the wiley has 32 megs Jun 01 13:19:14 and I have some 200 megs swap here as well ;-) Jun 01 13:19:44 I know, but even then it gets very tight if you want to run something like spamassassin too Jun 01 13:20:34 Even my Athlon 650 gets on its knees when I get a load of mail in Jun 01 13:20:55 Or do you want to run without spam killer Jun 01 13:20:58 I dont wanna run spamd or any antivirus soft :) Jun 01 13:21:21 I know what that stuff does to a mailserver, I have been running one before on a "real" linux box as well ;-) Jun 01 13:21:33 I do spam-filtering client-side Jun 01 13:21:55 the mail-addys I wanna host there, are only very few, with few mails in and out. Jun 01 13:22:36 the thing just is, that I do host a couple of domains, and it would be nice to have the possibility to recieve mail to thoose domains as well :) Jun 01 13:24:35 jp30-work, Jun 01 13:25:37 jp30-work, I commented the rm $(STAGING_LIB_DIR)/libexpat.la in the expat makefile for now, and now apr-util compiles Jun 01 13:31:24 macsat mini-sendmail is in the wiley firmware Jun 01 13:32:13 llagendijk - when you said that apr-util looks for a static expat, did you mean that it looks for a .la? .la files are libtool archives, not static libs Jun 01 13:32:38 oops. sorry Jun 01 13:33:23 I am still somewhat confused when it comes to all the different extensions used Jun 01 13:33:52 ok, can you tell me more about what goes wrong when building apr-util - i can't reproduce the problem Jun 01 13:35:08 jp30-work, Apache compile fails Jun 01 13:35:45 with what error? Jun 01 13:36:06 jp30-work, apr-util compilation fails as it can not find apr. It looks for libexpat.la and then says that it cannot find apr Jun 01 13:36:35 apache fails with #error mod_ldap requires APR-util to have LDAP support built in Jun 01 13:36:36 this is during compilation or during configure? Jun 01 13:36:45 configure Jun 01 13:37:00 ah, apache needs tweaking to disable ldap on wiley Jun 01 13:37:47 i think i have an appropriate patch here... will check it in Jun 01 13:38:03 do you need ldap, otherwise I an simply disable it in apache makefile Jun 01 13:38:33 that's all that's needed, why don't you do that locally until we can build a working binary from it Jun 01 13:38:57 just remove --enable-ldap and --enable-auth-ldap Jun 01 13:39:13 ok, that was what I planned to do Jun 01 13:39:42 can you quote the exact error you get when apr-util fails? Jun 01 13:40:22 ok, need to try it again, just a few minutes Jun 01 13:41:51 ps, I really like the build system, it is so easy to work with it Jun 01 13:42:38 it is mostly rwhitby's work i believe Jun 01 13:43:36 <[g2]> jp30-work, will you be around in a couple hours ? Jun 01 13:44:08 unfortunately not, i will have to go to a meeting in 30 min Jun 01 13:44:25 <[g2]> nod. I'll try and ping you tomorrow Jun 01 13:44:44 ok, you could also email me: jbparsons@ucdavis.edu Jun 01 13:44:54 <[g2]> ok will do Jun 01 13:44:56 <[g2]> thx Jun 01 13:45:24 llagendijk, i think i have reproduced the apr-util problem Jun 01 13:49:01 i wonder why apr-util used to compile? Jun 01 13:49:32 maybe because you had an old libexpat.a hanging around? Jun 01 13:51:26 if so, it used that. I had problems reproducing the problem due to the fact that I hacked the makefile to build the static lib Jun 01 13:52:18 i see - i added --disable-static to expat's configure invocation recently Jun 01 13:52:25 apr-util appears to use libexpat.a or libexpat.la (and that is why I got confused) Jun 01 13:53:17 Did you change expat version too? If so there may be a mismatch between headers and lib in your compilation explaining the crashes Jun 01 13:54:20 no i didn't change expat in any other way. well, why don't we go with your solution of allowing expat to install libexpat.la in staging, that's less harmful than building a whole static lib Jun 01 13:56:14 ok, will use that for now (I am testing in a different tree, using anonymous cvs, so I can not get your changes in realtime as the anonymous cvs lags a few hours) Jun 01 13:57:38 ok. i'll check in the change to expat as it was broken for slug too. the change to apache can wait until we can build a working apache for wiley Jun 01 13:57:53 fine Jun 01 13:59:10 jp30-work, what did you use to test apache? Jun 01 13:59:27 03jp30 * 10unslung/make/expat.mk: apr-util requires staging_lib_dir to contain libexpat.la or libexpat.a Jun 01 13:59:53 i just started it up - it segfaulted before getting very far Jun 01 14:00:06 ok, that makes it easier Jun 01 14:00:11 it is probably easier to test subversion, though Jun 01 14:00:42 you need to build apache before subversion because of the way the unslung svn package is set up, but the svn binary doesn't depend on apache itself Jun 01 14:00:51 ...just on apr-util Jun 01 14:01:33 subversion didn't segfault, but didn't work either. Jun 01 14:02:16 never user subversion, so that may be the problem Jun 01 14:02:51 well, neither of them work, the problem is likely the same in both cases Jun 01 14:03:00 we'll see Jun 01 14:04:04 ok, I'm off to bed (this stuff is too facinating, I keep going on till late) Jun 01 14:04:16 g'dnight Jun 01 14:04:40 or better "have a nice day" Jun 01 14:04:45 'nite Jun 01 14:16:22 info: am currently attempting to install linux - very slo-o-o-wly... Jun 01 14:28:18 jp30-work, here is the reason why apache appears to fail: Jun 01 14:28:37 open("/home/home1/louis/unslung/wiley/staging/opt/lib/libc.so.0", O_RDONLY) = -1 ENOENT (No such file or directory) Jun 01 14:28:37 open("/opt/lib/libc.so.0", O_RDONLY) = -1 ENOENT (No such file or directory) Jun 01 14:28:37 open("/home/home1/louis/unslung/wiley/staging/opt/lib/libc.so.0", O_RDONLY) = -1 ENOENT (No such file or directory) Jun 01 14:28:37 open("/opt/lib/libc.so.0", O_RDONLY) = -1 ENOENT (No such file or directory) Jun 01 14:28:38 open("/home/home1/louis/unslung/wiley/staging/opt/lib/libc.so.0", O_RDONLY) = -1 ENOENT (No such file or directory) Jun 01 14:28:41 open("/opt/lib/libc.so.0", O_RDONLY) = -1 ENOENT (No such file or directory) Jun 01 14:28:43 open("/home/home1/louis/unslung/wiley/staging/opt/lib/libc.so.0", O_RDONLY) = -1 ENOENT (No such file or directory) Jun 01 14:28:46 open("/opt/lib/libc.so.0", O_RDONLY) = -1 ENOENT (No such file or directory) Jun 01 14:28:48 open("/home/home1/louis/unslung/wiley/staging/opt/lib/libc.so.0", O_RDONLY) = -1 ENOENT (No such file or directory) Jun 01 14:28:51 open("/opt/lib/libc.so.0", O_RDONLY) = -1 ENOENT (No such file or directory) Jun 01 14:28:53 open("/home/home1/louis/unslung/wiley/staging/opt/lib/libc.so.0", O_RDONLY) = -1 ENOENT (No such file or directory) Jun 01 14:29:00 --- SIGSEGV (Segmentation fault) @ 0 (0) --- Jun 01 14:29:18 apache is attempting to load libraries from the path it found when compiling Jun 01 14:29:47 strace is your friend.... Jun 01 14:53:26 03sjmelhuish * 10unslung/make/rrdtool.mk: Increment to rrdtool 1.2.9 Jun 01 14:59:35 03sjmelhuish * 10unslung/Makefile: libart and rrdtool ready for (cross) testing Jun 01 16:35:05 [g2] hi Jun 01 16:35:44 <[g2]> hey Christopher ! Jun 01 16:35:55 I have a couple of questions for you Jun 01 16:36:02 <[g2]> absolutely! Jun 01 16:36:13 First of all, did you ever try the EALL in upslug? Jun 01 16:36:41 * [g2] hangs his head in shame Jun 01 16:36:48 <[g2]> :( Jun 01 16:37:03 and secondly, when I asked about memory chips you said that you had the digikey page to find them, I'm getting another slug and thinking about making it a 64mB version so that would be helpful Jun 01 16:37:33 <[g2]> I'm getting my 1st article of the rework back tomorrow Jun 01 16:37:50 <[g2]> with 64MB, serial header and JTAG lines brought out Jun 01 16:37:52 cool Jun 01 16:38:04 <[g2]> APEX is ready to go at 64MB Jun 01 16:38:04 Christopher, hi :-) Jun 01 16:38:09 hi jacques Jun 01 16:38:42 will APEX work with unslung or just openslug? Jun 01 16:38:47 Christopher, did you hear about jbowler's modifications to upslug? I think he wanted to talk to you before committing anything Jun 01 16:39:05 yeah, I've been emailing him, he committed the changes a few hours ago Jun 01 16:39:32 excellent Jun 01 16:40:16 <[g2]> check the logs from today... jbowler has got an upslug.c iirc Jun 01 16:40:34 <[g2]> we had a long talk about upgrades for openslug Jun 01 16:41:28 yeah, his code does a lot more error checking and resends packets if they don't make it Jun 01 16:41:31 oh I see that now - I need to read that Jun 01 16:42:01 i'm going to put openslug on my new slug when I get it, then I can be running both unslung and openslug at the same time Jun 01 16:42:40 <[g2]> Christopher, are you interested in one of those new fangles fat slugs ? Jun 01 16:42:46 yes Jun 01 16:43:12 I want apache to run faster :-) Jun 01 16:43:31 <[g2]> are you planning on running apache with unslung or openslug ? Jun 01 16:43:35 going from 32MB to 64MB is about a 4x increase in usable RAM Jun 01 16:43:45 i haven't decided that yet Jun 01 16:43:56 I think i will run mysql on one box and apache on the other Jun 01 16:44:02 but I don't know which way yet Jun 01 16:44:22 I think I may put apache on the openslug box and have that box just run off of a memory stick Jun 01 16:44:57 <[g2]> I been wanting to setup a benchmark and a configs for apache Jun 01 16:45:15 <[g2]> and/or thttpd to do some performance testing Jun 01 16:45:28 <[g2]> I'm doing that on my GigE NAS server box Jun 01 16:45:47 yeah, that would be good, the real performance issues come with php pages esp. phpmyadmin Jun 01 16:46:41 <[g2]> One of the next couple spins will be attempting to put 128MB on the slug Jun 01 16:46:55 <[g2]> 32M16 times 2 Jun 01 16:47:28 what would we call that? Jun 01 16:47:34 the super heavyweight slug? Jun 01 16:47:36 :-) Jun 01 16:48:42 well, a slug *is* a unit of mass Jun 01 16:49:21 true Jun 01 16:51:51 <[g2]> uber slug :) Jun 01 16:52:15 <[g2]> or uber fat slug Jun 01 16:53:45 slug X2, uh, X4.. Jun 01 17:00:49 [g2] how expensive where the RAM chips for upgrading your slug? Jun 01 17:01:17 <[g2]> $8.40 per chip in Q 25 iirc Jun 01 17:01:35 k Jun 01 17:02:55 <[g2]> to keep stuff simple at first I'm gonna do a bulk buy of new units and burn them in for several days, then have them re-worked, and burn them in for another couple days Jun 01 17:03:23 <[g2]> that way I'd weed out the infant mortalities from Linksys and the rework Jun 01 17:03:56 <[g2]> I may even build an enclosure for temp testing Jun 01 17:04:01 <[g2]> dunno yet Jun 01 17:05:03 MT48LC16M16A2TG-75:D TR Jun 01 17:05:16 is that the right chip? Jun 01 17:05:37 <[g2]> I think that looks close Jun 01 17:05:46 <[g2]> but I think it's the 7E Jun 01 17:06:14 ok Jun 01 17:06:16 <[g2]> I think that may just be the lead free version (the 7E that is) Jun 01 17:06:28 yeah, I think so Jun 01 17:06:37 <[g2]> I've got those chips on my Avila board which runs with 64MB Jun 01 17:06:40 <[g2]> :) Jun 01 17:06:45 <[g2]> APEX and all Jun 01 17:10:03 Avila? Jun 01 17:12:45 I like to get the lead free versions, because then I can suck on the chips with impunity Jun 01 17:15:36 It may be gum sized but the legs would probably hurt quite a bit :) Jun 01 17:15:54 hmm, damn I thought they were talking about a BGA Jun 01 17:16:17 The RAM isn't Jun 01 17:17:25 yeah, ok thanks for pointing that out, I guess the lead ones are fine then :-) Jun 01 17:17:35 heh Jun 01 17:18:10 damn it's a nice day, I guess I have no excuse to not mow the lawn Jun 01 17:18:14 BIAW Jun 01 17:18:22 Speaking of BGA, somebody I was talking to yesterday said they had fairly good success at mounting them using a toaster oven Jun 01 17:18:26 Have fun... Jun 01 17:18:38 that's ingenious Jun 01 17:21:56 <[g2]> Christopher, www.gateworks.com Jun 01 17:23:45 looks cool, like a slug but with mini-pci and more RAM :-) Jun 01 17:31:03 the NSLU2 project should join up Jun 01 17:31:08 sounds cool Jun 01 17:39:06 <[g2]> Christopher, I've had an avila board for many months Jun 01 17:39:24 <[g2]> It only took about a day to port all the OpenSlug stuff over to it Jun 01 17:53:37 Christopher: I hard-disabled EALL (an fprintf then an exit(1) in the correct place). When I have a slug with JTAG then I'll test it ;-) Jun 01 17:54:04 Tiersten, http://www.sparkfun.com/tutorial/ReflowToaster/reflow-hotplate.htm Jun 01 17:54:21 <[g2]> that's a funny article :) Jun 01 17:54:23 also of interest, http://www.sparkfun.com/tutorial/SMD_Printing/SMD_Printing.htm Jun 01 19:17:15 [g2]-away: jbowler: all this talk about preservation I see in the logs: doesn't the conffiles facility in ipkg do that for us? Jun 01 19:18:00 e.g. when I do ipkg upgrade on OpenSlug, it tells me that I have a modified /etc/network/interfaces, and allows me to diff it against the new one, and even edit it in place if I need to. Jun 01 19:18:45 so what config data are you talking about which can't be handled by the conffiles mechanism? Jun 01 19:55:38 i am looking for a kerberos package - does anyone know if one exist? Jun 01 21:51:31 anyone alive in here? Jun 01 21:59:20 rwhitby-away: CONFFILE: works in the "ipkg upgrade" case, i.e. if a new openslug-image were to be released (RDEPENDing on anything else new it needs) the upgrade should preserve all the user configuration. Jun 01 22:00:23 The problem is that getting this right for a new kernel (i.e. openslug-kernel plus all required kernel-module-*) requires a successful upgrade to the flash as well as the active rootfs. Jun 01 22:01:15 That's because the module required to mount the active rootfs (it may not be built in) must be upgraded on the jffs2 partition so that it matches the upgraded kernel (in /dev/mtdblock2). Jun 01 22:02:30 This starts to sound tricky (hence my 'scary' comment). So everything is fine with ipkg upgrade up to but not including the kernel ;-) Jun 01 22:03:15 hey jbowler you mind if I ask you a couple of simple noob questions Jun 01 22:04:31 jbowler re: EALL, sounds good Jun 01 22:05:00 Lunch[BoX] hi do you need help? Jun 01 22:05:01 So I think we reached the conclusion that, for the next release, we need to get the flash upgrade to preserve/restore the core user config semi-automagically (where the user overwrites the whole jffs2). Jun 01 22:05:18 Yeah - it's something simple I'm missing =/ Jun 01 22:05:20 Lunch[BoX]: go ahead... Thus talking to a wall Jun 01 22:05:37 s/thus/just/ Jun 01 22:07:06 made sure I could telnet into redboot - fine - upgraded firmware to unslung 4.20 beta - fine - unslung - fine - opened up the telnet port on nslu2 - fine - tried to telnet in to upgrade ipkg - can connect to nslu2 via telnet but can't log in Jun 01 22:07:36 Everything else works I'm just missing something simple =/ Jun 01 22:08:36 Do you know the unslung password? Jun 01 22:09:00 uNSLUng? Jun 01 22:09:53 Yep, that's the one. Does telnet give 'permission denied' or some other message? Jun 01 22:10:09 I have the drive connected is that the problem? Jun 01 22:10:47 try without drive attached Jun 01 22:10:58 Yes, that used to be a problem, I believe it still is with 4.x Jun 01 22:11:14 k Jun 01 22:26:41 that worked thanks - now I just need to figure out the whole telnet/ipkg stuff =) thanks Jun 01 23:38:03 jacques have you looked at the google summer of code program? Jun 01 23:40:43 Christopher, nope Jun 01 23:41:07 i thought it was just for students Jun 01 23:41:26 yeah it is, I was thinking about trying to find something to work on Jun 01 23:41:59 something slug-related? Jun 01 23:42:14 that would be cool Jun 01 23:42:37 the nslu2 group would have to become a mentoring organization though Jun 01 23:42:50 i'm still a little confused as to how the whole thing works Jun 01 23:43:45 i haven't read about it so I don't know the requirements or how flexible it is Jun 01 23:43:54 I'm sure most of qualify for being totally mental. Jun 01 23:44:06 well that goes without saying Jun 01 23:44:11 oh mentOR... :-) Jun 01 23:44:19 LOL Jun 01 23:44:33 the mentoring organization receives $500 and the student receives $4500 on completion of the project Jun 01 23:45:46 organization probably has to do reports and stuff Jun 01 23:45:52 not a big deal if so Jun 01 23:46:56 hmmmmm Jun 01 23:47:08 I tried googling "google summer of code" and retuned nothing that looks related.. is there a link? Jun 01 23:47:23 i know it was on /. a couple of days ago Jun 01 23:47:36 Ah, found it Jun 01 23:47:55 http://code.google.com/summerofcode.html Jun 01 23:48:25 I guess you need to a "prominent Open Source Organization". Jun 01 23:48:31 oir we would need to be.... Jun 01 23:48:40 "established prominent open source organizations" bah Jun 01 23:49:31 " Do you want to be a mentoring organization (it isn't too late!)" Jun 01 23:49:38 the faq says it is Jun 01 23:49:39 Are you accepting any more Organizations into the program? Jun 01 23:49:39 No. Jun 01 23:49:44 WTF Jun 01 23:50:03 they need a consistancy checker Jun 01 23:50:03 non-self-consistent POS bastages Jun 01 23:50:21 didn't see that part, guess it rules the slug out Jun 01 23:50:31 thought I might see about working with the ubuntu project Jun 01 23:51:01 May 31st: Announce on code.google.com and open application process Jun 01 23:51:01 June 1st: Last Day new Organizations will be listed on Code.Google.Com Jun 01 23:51:04 uh Jun 01 23:51:09 1 fscking day Jun 01 23:51:10 ? Jun 01 23:52:48 hmmm, should have got kergoth to let me work with openembedded Jun 01 23:53:09 hmmm, OSCON is in Portland Aug 1-5 Jun 01 23:53:19 closer Jun 01 23:53:22 maybe we should try to send a delagation Jun 01 23:53:29 dammit Jun 01 23:53:35 i know how to spell but not type Jun 01 23:53:38 delegation Jun 01 23:53:56 hmmm Jun 01 23:54:08 wait, when's LWE/SF? that sounds almost overlapping Jun 01 23:54:15 they wouldn't dare Jun 01 23:54:27 8-11 Jun 01 23:54:43 yeah ok Jun 01 23:54:56 going to be a bust first couple of weeks in august for me Jun 01 23:54:59 busy Jun 01 23:55:14 I may be in Guam. Jun 01 23:55:22 dyoung-zzzz, :-\ Jun 01 23:55:30 Chasing Hurricaines. Jun 01 23:55:40 Typhoons... whatever they call um down there. Jun 01 23:55:50 i have it on good authority there won't be any hurricanes at that time Jun 01 23:56:09 you might as well have fun with ka6sox-zzzz and me at these conferences Jun 01 23:56:14 heh Jun 01 23:56:21 Good Plan. Jun 01 23:56:35 I like that plan better. Jun 01 23:58:01 aha Jun 01 23:58:01 ! Jun 01 23:58:16 those silly ppl finally got the info up Jun 01 23:58:50 hey, handhelds.org is part of the summer of code, cool **** ENDING LOGGING AT Wed Jun 01 23:59:56 2005