**** BEGIN LOGGING AT Fri Jun 10 23:59:57 2005 Jun 11 00:13:18 hey, why is openslug back to gcc 3.4.3 ??? Jun 11 00:14:14 gcc-cross-initial is 3.4.4 and gcc-cross is 3.4.3 Jun 11 00:14:27 hmm is that something in my local.conf? let me check Jun 11 00:14:52 oops, sorry, false alarm Jun 11 00:15:09 from when I was testing flex breakage Jun 11 00:42:29 flashing maintenance-mode-free unslung alpha now .... Jun 11 00:58:06 YES! freshly built V2.3R29-uNSLUng-5.2-alpha booted on my box. Jun 11 00:58:21 one single problem during build: ERROR: dependency devio (for unslung-image) not satisfied Jun 11 00:58:36 but the second attempt worked Jun 11 00:59:00 ilg: did you bk pull from nslu2-package-symlinks? Jun 11 00:59:18 you will need devio from 5.3 onwards for upgrade mode entry Jun 11 00:59:21 yes, that was the second attempt :-) Jun 11 01:02:24 5.3-alpha will be pushed very soon Jun 11 01:02:55 next comes the hard part: I have to patch the Linksys web interface (page Administration->System) to accept '.' in the workgroup name, since my workgroup is named livius.net and it is rejected Jun 11 01:03:52 03rwhitby 07 * r1.3480 10openembedded/ (5 files in 3 dirs): Unslung 5.3-alpha: Replacement maintenance mode with code to force entry into RedBoot Upgrade Mode Jun 11 01:04:07 s/Replacement/Replaced/ Jun 11 02:06:12 03rwhitby 07 * r1.3482 10openembedded/packages/nslu2-binary-only/ (4 files in 2 dirs): Updated the Unslung README and NOTES for 5.x Jun 11 02:06:48 next step is to add the USB IDs patch from R29, and then add dropbear to the image. Jun 11 02:07:19 5.3 running on my box too Jun 11 02:08:07 bah, dropbear can bite me. Jun 11 02:08:31 since switching to openssh, no more disconnects Jun 11 02:12:00 jacques: we'll see if we have room for openssh ... Jun 11 02:12:55 rwhitby-away, I doubt we will, it's rather huge. the dropbear issue will probably get fixed at some point - I checked their list and saw no mention of it but it's very low traffic Jun 11 02:13:23 but for me, openssh is the way to go, because it's also much faster than dropbear at scp Jun 11 02:14:05 yeah, for me too Jun 11 02:14:31 we have 1.2MB available ... Jun 11 02:14:59 glc: upgrade mode pushed, please test Jun 11 02:15:41 also please everyone review the latest README flashing instructions Jun 11 02:16:12 probably doable as long as the packages are properly done - as in no .a files except in -dev Jun 11 02:16:28 some of the openssl and libcrypto .a files are huge Jun 11 02:16:33 checking Jun 11 02:17:36 do we need the openssl binary ? Jun 11 02:17:56 317096 Jun 8 00:42 openssl* Jun 11 02:18:14 need libcrypto, libssl, ... Jun 11 02:19:15 openssh, openssh-sshd Jun 11 02:19:22 we could probably fit all those in Jun 11 02:19:43 don't "really" need openssh-scp and openssh-ssh not for the flash anyway Jun 11 02:25:06 I think it's all or nothing, or it gets too confusing mixing oe packages in the image and optware after that Jun 11 02:27:25 define all Jun 11 02:27:32 -dev ? Jun 11 02:27:46 do we need several megabytes of static libs ? Jun 11 02:28:06 or just above + ssh and scp ? Jun 11 02:28:23 and yeah, I momentarily forgot this is for unslung Jun 11 02:29:11 in that case I guess you mean just all the clients, not -dev Jun 11 02:30:36 I'm still not sure if we need the openssl package tho - it just includes the openssl binary Jun 11 02:34:59 I mean any package that you install in the image can't really have an optware alternative, or else it gets really confusing Jun 11 02:36:12 oh... yeah that's true Jun 11 02:37:45 maybe I'll just leave that out for the first 5.x release Jun 11 02:41:38 yah, probably should Jun 11 03:14:43 in which case, the current repo contents is the release candidate, and the release version will be 5.4-beta Jun 11 03:15:02 alpha testing of the release candidate begins Jun 11 03:16:51 so should I start a fresh unslung build ? Jun 11 03:19:25 yep, no outstanding changes from me at the moment Jun 11 03:19:58 if we get no problems in the next 24 hours of testing this one, then we will bump it to 5.4-beta and hand it over to the binary crew Jun 11 03:20:17 `k Jun 11 03:20:41 clearing tmp and .ccache Jun 11 03:20:56 wow - VGA on the slug just announced on the mailing list Jun 11 03:21:56 ooh gotta go look Jun 11 03:22:16 http://webuser.fh-furtwangen.de/~dersch/ - end of the page Jun 11 03:23:02 neat Jun 11 03:24:07 Helmut is a animal! Jun 11 03:24:30 no pictures. Bummers. Jun 11 03:25:06 we need to get him to visit here .... Jun 11 03:25:23 dyoung: can you do your torture tests on 5.3-alpha ? Jun 11 03:25:49 started fresh unslung-image build Jun 11 03:28:37 we also need to try and get helmut to contribute packages rather than tar files Jun 11 03:33:17 ok I gotta pull and build. And find my slug. Jun 11 03:43:49 rwhitby-away, so you did pin binutils back for unslung ? Jun 11 03:45:35 yep Jun 11 03:54:06 I cant build. Jun 11 03:54:22 ? Jun 11 03:54:23 it barfs after cvs co-ing devio with some ugly python messages. Jun 11 03:54:57 either I didn't get that or haven't gotten that far yet Jun 11 03:55:10 maybe its time for a blowaway. Jun 11 03:55:27 I'm on glibc Jun 11 03:55:39 I blew away tmp. Jun 11 03:56:01 trying again... Jun 11 03:56:40 I think I'll blow away .ccache for good measure. Jun 11 03:56:51 I would :-) Jun 11 04:10:57 where is the glibc version set for openslug? Jun 11 04:11:38 Its not. Jun 11 04:11:52 it would be in conf/distro/openslug.conf though. Jun 11 04:12:06 hmm, then why does it not try to use the latest? Jun 11 04:12:41 You got me. Jun 11 04:13:10 Alls I know is thats where I have been setting preferred versions for ixp drivers and kernels and stuff. Jun 11 04:13:31 sure, and that's correct AFAIK Jun 11 04:13:46 ooh, maybe glibc is pinned in an arm or armeb arch config Jun 11 04:13:49 checking Jun 11 04:14:32 I just passed binutils, now on gcc. Jun 11 04:14:53 hmm, I'm not finding it in the machine confs Jun 11 04:15:30 morning sunshine. Jun 11 04:15:47 I found my slug. Jun 11 04:15:53 <[g2]> cool Jun 11 04:15:59 <[g2]> morning Jun 11 04:16:01 it was on the floor, buried under a pile of '244's. :-) Jun 11 04:16:07 <[g2]> :) Jun 11 04:16:07 dyoung: did you bk pull nslu2-package-symlinks for devio? Jun 11 04:16:34 Yeah, I always pull the symlinks. Jun 11 04:17:04 its part of the bk-pull-everything-I-need script. Jun 11 04:17:04 <[g2]> looks like we've got a new upgrade mechanism for OpenSlug thx to jbowler Jun 11 04:17:10 unslung build completed Jun 11 04:17:17 does anyone know if Tiersten pushed the USB IDs patch for unslung kernel? Jun 11 04:17:21 (from R29)? Jun 11 04:17:28 <[g2]> dunno Jun 11 04:17:34 * dyoung dunnos Jun 11 04:18:44 [g2], do you know where the openslug glibc version is set ? Jun 11 04:21:12 <[g2]> jacques, not off-hand Jun 11 04:21:43 <[g2]> I think we're tied to a cvs version Jun 11 04:22:51 we are but I can't find where Jun 11 04:23:00 <[g2]> NOTE: package glibc-2.3.2+cvs20040726: started Jun 11 04:23:56 [g2], sure, but where is that set? Jun 11 04:24:00 I've looked in the places I know Jun 11 04:24:06 distro conf, machine conf Jun 11 04:24:21 there are newer versions of glibc in oe Jun 11 04:29:42 Does anyone have a good link for CAN interface/software? Jun 11 04:30:11 <[g2]> no. but CAN would be fun Jun 11 04:30:16 been having fun in the new car, but want to be sure not to blow something up.... Jun 11 04:30:32 so semi-realtime monitoring might be nice.... Jun 11 04:30:43 <[g2]> to a slug right ? Jun 11 04:30:44 <[g2]> :) Jun 11 04:30:55 <[g2]> so you're looking for USB-CAN :) Jun 11 04:31:04 Yeah, i figure can use a I2C LCD Display.... Jun 11 04:31:06 If anyone can, usb-can. Jun 11 04:31:53 <[g2]> don't you just love to do the "can can, everybody can can ....." Jun 11 04:32:50 <[g2]> I'd use a laptop connect with ethernet for big colorful graphics :) Jun 11 04:33:04 one should be able to hack a PL2303 for can right? Jun 11 04:33:18 and thats "hack" as a pretty loose term. Jun 11 04:34:34 <[g2]> I thought CAN was higher speed Jun 11 04:34:39 Well okay I guess the driver would need hacking too... Jun 11 04:34:57 1Mbit or so, which the PL2303 does I thought? Jun 11 04:35:00 03rwhitby 07 * r1.3485 10openembedded/packages/linux/ (2 files in 2 dirs): Added the patch from R25 to R29 to the Unslung kernel (or at least as much of it as we want to add). Jun 11 04:35:06 or am I thinking of the PL2303X ? Jun 11 04:35:45 all that was left in that patch was the unusual usb id's to make onetouch_detect work on more maxtor drives Jun 11 04:37:51 ok, make sure you add your entry to http://groups.yahoo.com/group/nslu2-linux/database?method=reportRows&tbl=10 as you test out Unslung 5.x Jun 11 04:45:07 I'm still on GCC, so I have a feeling this is gonna take a while... Jun 11 04:53:50 I like upgrade mode so much better than both maintenance mode and even reboot loading. Jun 11 04:55:15 WTF. Jun 11 04:55:31 ? Jun 11 04:55:45 after a cvs co of devio it croaks with the same python traceback message Jun 11 04:56:18 I updated bitbake and did a full blowaway build. Jun 11 04:57:01 I have a checkout from 4 hours and 20 minutes ago Jun 11 04:57:18 do you oguys have CVS_TARBALL_STASH set in your local.conf 's ? Jun 11 04:58:33 I'll check Jun 11 04:58:48 I unset mine it didnt make a differnce. Jun 11 04:58:52 not mine Jun 11 04:58:58 not me Jun 11 04:59:00 it does the CVS check out, then it spipts this out. Jun 11 04:59:02 Traceback (most recent call last): Jun 11 04:59:02 File "/usr/local/bin/bitbake", line 715, in ? Jun 11 04:59:17 then a whole slew of other ugly messages. Jun 11 04:59:49 dyoung: do you still install bitbake. bad plan. Jun 11 05:00:03 I just updated it like an hour ago. Jun 11 05:00:12 how do I run it in-situ? Jun 11 05:00:26 use the local.conf from svn Jun 11 05:00:30 (and the makefile) Jun 11 05:03:35 * rwhitby is watching the new Dr Who ... Jun 11 05:03:49 That one has the CVS_TARBALL_STASH set to http://www.oesources.org/source/current . Jun 11 05:04:03 thats the only appreciable difference AFAICT. Jun 11 05:04:42 my local.conf for unslung has five lines: Jun 11 05:05:23 BBROOT = /home/unslung Jun 11 05:05:24 DL_DIR = /home/downloads Jun 11 05:05:24 BBFILES = ${BBROOT}/nslu2-package-symlinks/packages/*/*.bb Jun 11 05:05:24 MACHINE = nslu2 Jun 11 05:05:24 DISTRO = unslung Jun 11 05:11:27 what version of python are y'all running? Jun 11 05:11:58 Aha, that did it Jun 11 05:12:09 2.4.1 Jun 11 05:12:10 I had a spurious CVSDATE set. Jun 11 05:12:15 that broke it. Jun 11 05:12:16 dev-lang/python-2.3.5 Jun 11 05:13:08 its done now Jun 11 05:13:58 any comments on the upgrade mode warning dialog box from anyone? Jun 11 05:15:00 I'll let you know tommrow. Jun 11 05:15:10 I seem to have lost the power supply for the slug. Jun 11 05:22:31 oh yeah, now I remember, I recycled the power supply to use with the Wiley. Jun 11 05:26:05 upslugging Jun 11 05:33:42 okay. Jun 11 05:33:50 Thats a pretty comprehensive message. Jun 11 05:34:22 And it even works. :-) Jun 11 05:34:38 of course, it uses devio :-) Jun 11 05:41:32 rwhitby, the webpage is clean. Jun 11 05:42:02 It looks pretty idiot resistant. Jun 11 05:42:25 I wonder if we should insert stronger language into the popup though. Jun 11 05:42:35 http://www.nslu2-linux.org/wiki/Unslung/AboutUpgradeMode - thx to caplink811 Jun 11 05:42:54 make SURE you have the sercomm tool or upslug tool available before doing this. Jun 11 05:43:21 dyoung: since we know that upgrade mode is benign, and anyone *should* be able to get and use an appropriate tool, then I'm not too fussed. Jun 11 05:43:56 "If you dont have a Windows or Linux box to run the appropriate tool, dont do this" Jun 11 05:44:10 "ie: if you have a Mac, youre SOL" Jun 11 05:44:28 ah, Mac's will be running windows soon enough anyway .... Jun 11 05:44:58 Well, I think Tiersten was thinking about how to adapt upslug to use the pcap libs. Jun 11 05:45:14 yeah, necessity will be the mother of invention Jun 11 05:45:23 I spent a bit of time looking at how redboot does it. Jun 11 05:45:56 I dont think we can (rightly) directly pilfer that code. Jun 11 05:46:19 the protocol is quite straight forward Jun 11 05:46:45 It is, and having a generic raw ethernet frame doo-dad should be enough to make it go. Jun 11 05:48:42 well, someone who doesn't have a windows or linux box will figure it out .... Jun 11 05:48:53 our job is simply to make it available when they do Jun 11 05:51:34 Hmm - we still don't have a logo ... Jun 11 05:51:47 I thought ka6sox was getting one done professionally ? Jun 11 05:52:57 i think that was byront. Jun 11 05:53:09 the serial cable I sent him was payment for that service. Jun 11 05:53:59 did we pay you for that? Jun 11 05:54:37 The project didnt. Byront paypal-ed me though. Jun 11 05:55:10 ok, so the project owes ByronT then Jun 11 05:55:13 or rather, he paypal-ed ka6sox in my name. Jun 11 05:55:19 but you get the idea. Jun 11 05:55:38 we can ask him when he gets back online what he wants to do. Jun 11 06:02:56 dyoung: so does 5.3-alpha get your seal of torture approval? Jun 11 06:05:21 I'm not done yet. Jun 11 06:05:28 Actually I didnt start yet. Jun 11 06:05:45 Need to find a drive thats empty. Jun 11 06:06:01 My testing 256 USB Stick is on the Wiley now. Jun 11 06:06:01 I've been able to upgrade with disks attached with no problem (as expected, cause we're in redboot when it happens) Jun 11 06:06:09 I've also been doing it across wireless. Jun 11 06:06:19 I flashed it 3 times. Jun 11 06:06:31 it failed on verify once; but it worked anyways. Jun 11 06:06:37 next test will use the wl530g as a dumb switch between the wireless router and the devslug Jun 11 06:07:59 Thats a good point, between the slug and linux box theres a GigE switch then a crappy netgear 100BT switch. Jun 11 06:09:06 does UpSlug do a verify step like SerComm does? Jun 11 06:09:17 (and give a satisfying beep at the end?) Jun 11 06:09:21 I like that if you try to run ipkg on a pre-unsling-ed slug it says "permission denied". Jun 11 06:09:34 upslug verifies; but it doesnt beep. Jun 11 06:09:58 it just says "Requesting Reset... Confirmed" Jun 11 06:10:37 we could have upslug do the equivilent of "echo ^G" at the end when it resets. Jun 11 06:11:19 wow: http://www.nslu2-linux.org/wiki/FriendlyDocumentation/HomePage Jun 11 06:12:33 I like it when someone has a plan and actually does it. Jun 11 06:12:45 so unlike myself. Jun 11 06:15:58 I'm not sure if I said it or only thought it. Jun 11 06:16:56 but I would almost vote byront out of the "for dummies" docomentation; because that level of documentation is really best written by someone who *is* a new user. Jun 11 06:17:49 I think Byront is at a level of understanding now that it may be difficult to write at that level. Jun 11 06:18:39 Its ultimately his decision of course. Jun 11 06:19:02 (my secret objective is to have him write some new-user doco for openslug...) Jun 11 06:21:00 Looks like Craig and Philip are heading up the friendly doco effort anyway Jun 11 06:25:00 dyoung: should we remove the Sluggo link from the wiki sidebar? Jun 11 06:25:52 I dont want to arbitrarily exclude byront from that effort, if he wants to do it. He just doesnt have access right now. Jun 11 06:26:45 Its probably a good idea to remove that link. After I get a fatslug working I'll probably spend the couple hours to make a redboot work with it using those scripts. Jun 11 06:26:57 but at this point it may be misleading. Jun 11 06:29:02 03rwhitby 07 * r1.3486.1.1 10openembedded/BitKeeper/deleted/.del-ftpd-topfield_cvs.bb~ef7a682d37d9a3a9: Removed ftpd-topfield_cvs Jun 11 06:31:58 is there a preferred fs type for NAND flash devices? Jun 11 06:32:22 DiskOnChip? Jun 11 06:32:50 So whatever happened to SlugNews ..... Jun 11 06:33:25 It was lacking in content because everyone was busy making unslung 3.x happen? Jun 11 06:33:37 and openslug? Jun 11 06:33:42 * dyoung dunnos Jun 11 06:33:53 night all. Jun 11 06:34:23 night dyoung-zzzz Jun 11 09:59:08 upslug verify: it doesn't, there's something wrong in how the protocol is handled by upslug, it always suceeds (even if the verification data passed in is deliberately trashed). Jun 11 09:59:37 Is the protocol documented anywhere? I reverse engineered bits (black box) when I made the changes... Jun 11 10:01:24 devio CVS: it's the only thing in the builds which does not have a CVSDATE, i.e. it's following the head. It should probably be pinned for a release, any date is fine (20050609 != 20050611 because I fixed a help error last night). Jun 11 10:19:56 rwhitby: what's the protocol for adding to meta/openslug-packages? I'd like nfs-utils and samba to be in there, and I just checked the stuff in nslu2-package-symlinks. The following packages build fine: python, samba, strace, wget(the standalone), nfs-utils (only on glibc). I think it would be good to build all of these... Jun 11 10:20:50 rwhitby: I also added packages/meta/package-index.bb - adding this to the end of openslug-packages DEPENDS causes the package index to be rebuilt (works fine in my local build). Jun 11 10:39:21 nfs-utils is, however, incomprehensibly broken... A simple exportfs -a with a valid /etc/exports produces 'other.hostname:/usr/share/dir: No such device' Jun 11 10:40:10 Ah, incomprehensible error messages I think: I need kernel-module-nfsd and friends. Jun 11 10:40:15 <[g2]> jbowler, ran on glibc just a couple days ago Jun 11 10:40:33 <[g2]> nod on the kernel modules :) Jun 11 10:40:53 It's missing the RDEPENDS for the kenel modules (and it's missing a startup script and so on.) Jun 11 10:43:09 <[g2]> the RDPENDS is interesting... I'm not sure how kernel module versus built-in kernel moduiles dependancies are handled in OE Jun 11 10:45:31 They have to be given explicitly - OE/bitbake has no way of determining the dependencies. Jun 11 10:46:18 Hum, now the remote machine is says 'can't read superblock'... Jun 11 10:47:36 <[g2]> I've been meaning to write-up the how-to for a nfs server using nfs-utils and nfsd Jun 11 10:47:55 <[g2]> dyoung gave me the proper chants the other day Jun 11 10:49:01 So I have a valid /etc/exports and I have started mountd and statd, run exportfs -a, but I get the 'superblock' error - the mount request is authenticated, there's nothing in the local log. Jun 11 10:49:37 The remote log has 'server lochalsh(aka the slug) not responding, timed out' so something is missing. Jun 11 10:50:37 Strange, I thought there was a 'statd' kernel module, but I can't find it... Jun 11 10:51:34 <[g2]> Ok the first thing that happens is the /var/lib/nfs directory is missing Jun 11 10:51:50 <[g2]> I create that and then "nfsd" runs Jun 11 10:52:51 <[g2]> then "exportfs -a" Jun 11 10:53:03 In the kernel config 'Provide NFSv3 server support' is missing, but 'Provide NFS server over TCP support' is present. That doesn't make sense, the default is UDP. Jun 11 10:53:09 <[g2]> mountd Jun 11 10:53:30 <[g2]> both tcp and udp are provided I've run both Jun 11 10:55:03 I did all those steps, plus creating /var/lib/nfs/rmtab and installing kernel-module-nfsd, however I still get the timeout after the mount request. Jun 11 10:56:40 How do you force the client to use TCP? Jun 11 10:56:49 <[g2]> -o tcp Jun 11 10:57:29 <[g2]> Ok it's running for me Jun 11 10:57:38 <[g2]> the mountd seems to have died the first time Jun 11 10:57:45 <[g2]> but ran the second Jun 11 10:57:55 <[g2]> ps informed me it wasn't running Jun 11 10:58:03 <[g2]> or I mistyped it Jun 11 10:59:01 <[g2]> ok maybe I didn't type it in :) Jun 11 10:59:21 <[g2]> it's not in my command history (probably different window) Jun 11 10:59:24 With -o tcp 'nfs server reported service unavailable: Address already in use' Jun 11 11:00:05 Time to shutdown and reboot (and pull the power plug...) Jun 11 11:00:17 <[g2]> I"m sure you've got portmap running on you client host right ? Jun 11 11:00:39 portmap: yes, it's running Jun 11 11:01:06 A clue: my shutdown -r now worked and it hasn't been before - the only change was to do the NFS stuff... Jun 11 11:01:29 <[g2]> it's actually worked some for me Jun 11 11:01:33 <[g2]> the -r that is Jun 11 11:02:04 <[g2]> it worked 3 times when I first started, but then a day later it didn't Jun 11 11:02:40 <[g2]> beewoolie has a theory that the jffs2 leaves the flash in a bad state where it can't read on the next reboot Jun 11 11:02:58 Oh, I'd umounted /initrd too, a while back - that may be it. Jun 11 11:03:02 <[g2]> he's testing on different target and having the same result with a nfs mounted root Jun 11 11:03:25 <[g2]> no.. I've umount'ed /initrd a bunch and it fails Jun 11 11:03:52 Could just be random then... Jun 11 11:04:10 <[g2]> well not totally random Jun 11 11:04:23 <[g2]> I'm sure we'll nail this sooner or later Jun 11 11:05:36 Ok, here's what I did after a reboot (/var/lib/nfs exist and the kernel modules are now installed): Jun 11 11:05:37 exportfs -a Jun 11 11:05:41 mountd Jun 11 11:05:43 statd Jun 11 11:05:45 lockd Jun 11 11:05:58 Then mount lochalsh:/exports/mt-daapd /mnt Jun 11 11:06:04 And I get 'can't read superblock' Jun 11 11:06:25 And mount -o tcp lochalsh:/exports/mt-daapd /mnt Jun 11 11:06:32 <[g2]> where's the running of nfsd ? Jun 11 11:06:42 And I get 'nfs server reported service unavailable: Address already in use' Jun 11 11:06:48 lsmod Jun 11 11:07:03 It's in the kernel isn't it? Jun 11 11:07:15 <[g2]> type nfsd Jun 11 11:08:08 Yes, that fixes it. That pulls in kernel threads for nfsd, lockd and rpciod/0 Jun 11 11:08:17 <[g2]> nod. Jun 11 11:08:47 <[g2]> I guess there needs to be a daemon to service the connections Jun 11 11:09:26 <[g2]> I was going re-test with the latest 2.16 build I've got Jun 11 11:09:37 There's no nfs daemon - just a kernel thread, the daemon which handles the connections is mountd. Jun 11 11:09:39 <[g2]> I'll post the how-to on the wiki tonight Jun 11 11:10:00 Better to get the startup script - installing nfs-utils should just make everything work. Jun 11 11:10:05 <[g2]> I don't need to do statd or lockd Jun 11 11:10:28 Now I can't rmmod nfsd... Jun 11 11:10:29 <[g2]> are you going to work on the script ? Jun 11 11:10:57 No: I could do if you want (since the other stuff is done.) Jun 11 11:11:33 <[g2]> I'm not asking you to Jun 11 11:11:53 <[g2]> I'd like to learn a little more about the nfs stuff Jun 11 11:12:12 <[g2]> I've run both client and server now from OpenSlug Jun 11 11:12:38 Last time I looked it was pretty confusing. I tried to bb the latest nfs-utils but didn't get anywhere. Jun 11 11:12:43 <[g2]> I' was pulling 60-65Mbs with out going to disk on large nfs reads with my 64MB slug Jun 11 11:13:12 <[g2]> it works with a couple commands right now Jun 11 11:13:28 <[g2]> you are correct, the startup script should be fixed Jun 11 11:13:31 It's certainly worth getting it right because it's a major piece of functionality for a file server (and it's missing in LinkSys) Jun 11 11:14:09 <[g2]> I've also built the samba server Jun 11 11:14:16 Take care where you put the startup - it'll be somewhere in rcS.d, there's a lot of documentation in packages/initscripts/initscripts-openslug about where to fit things. Jun 11 11:14:16 <[g2]> but haven't tried running it yet Jun 11 11:14:36 <[g2]> ok. Jun 11 11:14:45 What about that question I asked rwhitby - should these things go in the build? (I think yes, then they don't get broken.) Jun 11 11:15:58 <[g2]> I missed the question you asked rwhitby Jun 11 11:16:48 I've got builds with samba, nfs-utils, python, wget and strace all building fine - I asked about the protocol for updating openslug-packages to include all these (nfs-utils doesn't build on uclibc) Jun 11 11:17:20 shutdown -r now hung on me this time, and I didn't umount /initrd Jun 11 11:17:32 <[g2]> samba, nfs-utils, and python all built fine for me Jun 11 11:17:46 <[g2]> wget used to and strace was broken Jun 11 11:17:58 <[g2]> I'd *LOVE* a fixed strace Jun 11 11:18:13 jacques said strace wasn't working when installed, but it builds ok. Seems more likely to get fixed if it gets built ;-) Jun 11 11:18:44 <[g2]> You shoud add all of those to openslug-packages Jun 11 11:19:15 <[g2]> also... Thx for adding the packages-index which should also be added to the end of openslug-package dependencies Jun 11 11:20:58 Ok, here's the minimal NFSD setup (this sequence worked for me): Jun 11 11:21:02 nfsd Jun 11 11:21:04 exportfs -a Jun 11 11:21:06 mountd Jun 11 11:21:25 <[g2]> plus the mkdir /var/lib/nfs Jun 11 11:21:34 <[g2]> that's my chant Jun 11 11:21:55 <[g2]> and also the ipkg installs to get kernel module and nfs-utils Jun 11 11:21:56 Yes, I think :>/var/lib/nfs/rmtab too, but I'm not sure about that - it maybe that nfsd will create it. Jun 11 11:22:28 <[g2]> it does create it Jun 11 11:22:30 The ipkg install is just RDEPENDS += "kernel-module-nfsd" Jun 11 11:22:57 And the directory creation is just a mkdir in do_install Jun 11 11:23:34 (well, I think that works, but it's a tar archive so it may be necessary to create a file to, like rmtab/xtab/etab) Jun 11 11:24:02 Yep, the -o tcp works fine too. Jun 11 11:24:15 <[g2]> cool Jun 11 11:24:23 <[g2]> I'm happy for you Jun 11 11:24:59 <[g2]> it's nice that you run everything on uClibc Jun 11 11:26:20 I don't - this is glibc - nfs-utils doesn't build on uClibc because it tries to call rpc_getnumber (or something like that) in the code of exportfs Jun 11 11:26:34 <[g2]> AHHH.... Jun 11 11:26:41 Still samba and openldap (libldap-2.2) do build. Jun 11 11:26:43 <[g2]> you switch-hitter! Jun 11 11:27:17 I build three separate builds - glibc, uclibc and glibc with reiser4. The latter isn't currently working... Jun 11 11:27:28 <[g2]> btw you did a nice job with the openslug-packages and the glibc/uclibc specific builds Jun 11 11:27:41 Pity about tar ;-) Jun 11 11:28:04 The missing API is get_rpcbynumber. Jun 11 11:28:25 Are you looking at samba? Jun 11 11:28:32 <[g2]> I'll try to run it Jun 11 11:28:54 <[g2]> I'll be looking at apache, mysql and php first probably Jun 11 11:29:07 <[g2]> I may try a quick test with samba though Jun 11 11:30:32 <[g2]> jbowler, THX for all the help.. Time for me to cut the grass :) Jun 11 11:30:45 <[g2]> have fun with nfs :) Jun 11 11:30:59 Have fun with the grass... Jun 11 11:31:14 <[g2]> time to go "accomplish something" :) Jun 11 11:31:37 Yeah, I've got to fix the garden path :) Jun 11 11:31:47 <[g2]-away> If I set that to -grass ppl would probably get the wrong idea Jun 11 13:39:59 hi all, got a quick question on ipkg-utils Jun 11 13:40:31 in the makefile there is an option IPKG_BUILD := PATH=$(TARGET_PATH) TAR_OPTIONS=--format=ustar ipkg-build -c -o root -g root Jun 11 13:41:13 does anybody know why the --format=ustar is there? My gnu tar does not know that option under Linux Jun 11 13:49:58 hi jf-work: do you know why the ipkg-utils gives option TAR_OPTIONS=--format=ustar, my tar does not understand that option? Jun 11 14:09:36 jbowler-away, [g2]-away: I'm listed as the maintainer for nfs-utils, as far as I know /var/lib/nfs gets created in do_install. Jun 11 14:10:13 And yes, I never got around to writing a startup script for it, feel free. Jun 11 14:15:38 <[g2]-away> dyoung-zzzz, thx, I'll check the latest, but it didn't seem to be happening Jun 11 14:48:27 <[g2]-away> dyoung-zzzz, jbowler-away there's a bunch of stuff in do_install but it doesn't seem to be happening Jun 11 14:59:00 Yes, the /var/lib/nfs directory isn't created - take a look at work/nfs-utils-1.0.6-r0/image Jun 11 15:01:04 The only way to make the files owned by something other than root is to make a postinst script - because fakeroot can only fake root. Also the statduser would have to be in the passwd file, and that depends on the individual system config. Jun 11 15:01:30 hehe, thats why theres a comment in there thats a question. Jun 11 15:01:50 I dunno why /var/lib/nfs isnt made. thats kind of annoying. do I need a extra / in there or something? Jun 11 15:02:38 You need a "/" after ${D} Jun 11 15:02:51 Look at temp/run.do_install* Jun 11 15:03:31 yep... work/nfs-utils-1.0.6-r0/imagevar has all of the stuff in it ;-) Jun 11 15:05:35 Use ${localstatedir} which is set in bitbake.conf to "/var" Jun 11 15:06:06 And it's in FILES_${PN} so it will go into the base package. Jun 11 15:06:10 jbowler-away, if you are in a position to apply tht fix; please do. Jun 11 15:06:24 The grass is calling me to be cut... Jun 11 15:06:29 Ok. Jun 11 15:06:33 Bunch of stoners. Jun 11 15:07:04 ;-) Jun 11 15:07:23 This is a differnt kind of grass. :-) Jun 11 15:08:20 Same problem, I think I need to invest in some sheep. Jun 11 15:10:13 You've got --with-statduser=nobody in EXTRA_OECONF, nobody tends to always be 32767 so a postinst fixup should be possible. Jun 11 15:10:49 (Hum, it should work anyway, even if someone is using NIS for passwd, or openldap, just so long as it gets run late enough.) Jun 11 15:12:53 I found a way round the forgotten-defines problem before, but I can't remember how I did it... Jun 11 15:14:44 [g2]-away: I just changed every 'var' to ${localstatedir}, and bumped the PR. Jun 11 15:20:54 [g2]-away: another clue, I've shutdown -r'ed several times now. Every time I umount /initrd well before the shutdown it's fine, every time I don't (so it gets umounted in the shutdown) it isn't. Jun 11 15:22:54 jbowler-away: linksys do something special on reboot to enable flash reads Jun 11 15:24:59 From what beewoolie said it seems likely that the jffs2 file system is in an unstable state so the reboot hangs - I wonder if it takes a while for the kernel to put the file system into a safe state after the umount. Jun 11 15:25:32 <[g2]-away> well that was his theory. I don't know if it's actually the case Jun 11 15:25:44 <[g2]-away> it very well may be but that's to be proven yet Jun 11 15:25:48 I did see the original LinkSys software have similar problems too (hang on reboot) though, and LinkSys doesn't use jffs2... Jun 11 15:26:44 <[g2]-away> whatever the case is we should be able to cleanup the hw state before the reboot or initialize it properly after Jun 11 15:27:13 <[g2]-away> I don't think we are actually rebooting because I don't see the + character Jun 11 15:27:46 <[g2]-away> jbowler-away, did you push that nfs-utils change ? Jun 11 15:27:50 I'm tempted to simply rmdir /initrd and see whether I ever see the problem again ;-) Jun 11 15:27:58 nfs-utils: just testing it. Jun 11 15:28:11 <[g2]-away> ok Jun 11 15:28:35 <[g2]-away> iirc it's there without /initrd Jun 11 15:29:25 <[g2]-away> personally I think it's an un-initialized value or race condition Jun 11 15:29:27 Looks like rmtab needs to be added to - I removed nfs-utils and rm -rf'ed /var/lib/nfs but 'exportfs' now says "can't open /var/lib/nfs/rmtab for reading" Jun 11 15:29:47 AdamBaker: welcome Jun 11 15:29:49 <[g2]-away> it'll create that Jun 11 15:30:01 <[g2]-away> but you are correct about the fix Jun 11 15:31:44 anyone try 5.3-alpha? Jun 11 15:32:11 <[g2]-away> sorry, but no :( I've been too busy with Openslug Jun 11 15:35:13 Hello, I've got everything ready for the whois package - is now a good time to hit commit? Jun 11 15:35:24 jbowler-away: can you put a cvs tag on devio andduse that in a versioned .bb? Jun 11 15:36:05 see ftpd-topfield for an example of the oe syntax Jun 11 15:36:53 Yes, I'll do that. Jun 11 15:37:52 It's annoying CVS doesn't support tags like "1.0", oh well... Jun 11 15:38:35 yeah Jun 11 15:38:57 1_0 is what I do ... Jun 11 15:39:11 <[g2]-away> speaking of version control... are we just waiting to see what #oe does with the bk issue ? Jun 11 15:39:26 jbowler-away: I've always found that product_1_0 works well so you'd have separate 1.0 tags for Unslung and OE Jun 11 15:40:22 [g2]-away: yeah. we can switch to svn in an instant if needed Jun 11 15:40:23 The problem is that '_' is reserved in OE Jun 11 15:40:54 And '-' is significant too... Jun 11 15:41:42 hm? the cvs tag only needs to be in the SRC_URI, where oe doesnt give a shit about - and _ Jun 11 15:42:03 <[g2]-away> kergoth, welcome ! Jun 11 15:42:06 hey Jun 11 15:42:12 <[g2]-away> I missed you man! Jun 11 15:42:53 kergoth: I find 'package_version' and use ${P} whereever possible to be preferable, sure I can work round the two different sets of rules, but... Jun 11 15:43:09 I.e. I want to say ;tag=${PV}; Jun 11 15:43:31 ${P} use in the .bb isnt what should ever be done, in case that .bb gets pulled into another via the include directive, which causes PN to be different. Jun 11 15:43:39 thats why the package name is usually hardcoded. Jun 11 15:44:13 ;tag=${@(bb.data.getVar('PV', d, 1) or '').replace('_','.')}; Jun 11 15:44:30 err, that should be '.','_' Jun 11 15:44:31 but you get the point Jun 11 15:45:26 oh yes, I can do it... I take the point about hardcoding ${PN} (since devio-native does exist ;-) Jun 11 15:45:47 then dont act like its a problem. they're only _reserved_ in the _filename_ Jun 11 15:45:50 kergoth: any progress on bk replacement for OE? Jun 11 15:45:51 they arent "reserved in oe" Jun 11 15:46:15 rwhitby: not really, i've been out of open source for a bit Jun 11 15:49:15 [g2]-away: I've pushed nfs-utils_1.0.6.bb - it's in changeset 1.3500 Jun 11 15:49:22 <[g2]> jbowler-away, thx Jun 11 15:49:58 I didn't attempt anything other than fixing the /var installation (no added RDEPENDS for the kernel module etc) Jun 11 15:50:55 <[g2]> jbowler-away, every little bit counts Jun 11 15:51:23 yeah, don't know how we do kernel module depends without having to use pre-knowledge about whether it is a module or compiled into the kernel Jun 11 15:56:01 We have to use pre-knowledge I think. The only other way I can think of involves looking in the deploy/ipk directory. Jun 11 15:57:17 In this case it's only compiled in if someone makes a special NFS server openslug - even then they don't really need to compile it in. Jun 11 15:59:50 [g2]: I'm not sure about the protections on rmtab, it's readable and it shows the remote mounting hosts. That's probably ok, but I'm not sure about security issues. Jun 11 16:00:34 <[g2]> jbowler-away, nod. If you want to lock it down a little more that's ok with me Jun 11 16:01:41 <[g2]> the kernel-module / built-in kernel I think is a good question for us to ask in #oe Jun 11 16:02:10 <[g2]> s/I think/RDPENEDS I think/ Jun 11 16:02:20 biab... Jun 11 16:03:18 As no-one said no when I asked earlier I'm going to go ahead and commit the whois package for unslung Jun 11 16:04:08 <[g2]> Go Adam, Go Adam, do the check-in, check it in, Ohh Yeah.... Go Adam, Be bold, Check it in .... Ohh yeah Jun 11 16:04:54 03aabaker * 10unslung/ (3 files in 2 dirs): Jun 11 16:04:55 Add a package containing a newer version of whois than is in inetutils. Jun 11 16:04:55 This shouldn't be added to the feed until whois is removed from inetutils. Jun 11 16:04:58 <[g2]> Just kidding, don't do it, you'll get in trouble OH YEAH.... Jun 11 16:05:23 <[g2]> actually checking it in was fine :) Jun 11 16:05:32 I guessed Jun 11 16:06:10 Due to the conflict with inetutils I haven't marked it as ready for testing Jun 11 16:08:40 <[g2]> sounds good to me Jun 11 16:08:56 AdamBaker: good plan Jun 11 16:09:50 let inge know it's there, and he'll update inetutils Jun 11 16:10:39 (bob_tm is Inge) Jun 11 16:16:02 Upslug worked great with 5.3 here. Hooking up a drive right now. Jun 11 16:32:12 glc: what does upslug do not when unslung 5.3 is running (i.e. not in redboot upgrade mode) ? Jun 11 16:32:34 can we get rid of the download process now? Jun 11 16:32:35 it still picks it up because download is running. Jun 11 16:34:16 does it try to upgrade, or does it do the hw mismatch? Jun 11 16:41:10 For me it tried to upgrade. Jun 11 16:42:19 It's erasing flash right now. Jun 11 16:43:06 so should we keep the download process, or nuke it? I guess individuals can nuke it if they are tight on RAM Jun 11 16:43:38 Flash fails. looks like we should remove upgrade. Jun 11 16:44:09 or download, or whatever it is called. Jun 11 16:45:06 Yeah, I never had it actually work. Jun 11 16:45:18 I vote nuking. Jun 11 16:45:36 It is very unhappy. No data nodes found. Jun 11 16:46:50 Reset button works just fine though. Jun 11 16:48:03 Make space for something mor useful and less insecure. Jun 11 16:48:21 what about when rootfs is external and initrd is still mounted? Jun 11 16:48:42 I'll try that in a minute. Jun 11 16:53:31 should we nuke the new onetouch_detect processes too? Do they do anything? Jun 11 16:54:31 I agree we should nuke download for security reasons Jun 11 16:55:35 I'll be out and about today, so if anyone else wants to implement the nuking, feel free. Just comment things out in the rc scripts (using a patch), so that someone can re-enable them easily if desired. Jun 11 16:55:50 (by editing the jffs2 copy) Jun 11 16:56:48 Also upslug should be changed so -U requires a -V(erify) option to prevent accidents. Jun 11 16:57:34 would be good to change -U to a different letter, or require you to type the whole word Jun 11 16:57:50 Flash failed. I'll try to reboot anyway. Jun 11 16:59:30 No eRcOmM and right into upgrade mode. Jun 11 17:01:02 Also nuking download will prevent accidental upgrade of wrong slug. Jun 11 17:13:38 yeah Jun 11 17:21:16 jbowler-away: your updated openslug-package are now in the feed Jun 11 17:21:18 back later Jun 11 17:42:26 <[g2]> jbowler-away, strace is still busted for me Jun 11 17:42:33 <[g2]> fyi Jun 11 18:08:14 dunno if I said this here or only in #oe, Jun 11 18:08:31 strace broke for us between 4.5.9 and 4.5.10 Jun 11 18:08:54 I posted to strace-devel, but since it's 98% spam my post might be missed Jun 11 18:14:49 <[g2]> jacques, yeah... jbowler said it worked for him Jun 11 18:15:27 [g2], worked in uclibc ? Jun 11 18:15:36 <[g2]> jacques, the native development environ is shaping up Jun 11 18:15:47 I read the backlog and missed where he said that... Jun 11 18:15:58 <[g2]> he said he switched to glibc for nfs-utils Jun 11 18:16:17 <[g2]> couple hours ago Jun 11 18:16:43 <[g2]> ok... 7 hours ago Jun 11 18:16:50 <[g2]> 6:58 actually Jun 11 18:17:34 jacques said strace wasn't working when installed, but it builds ok. Seems more likely to get fixed if it gets built ;-) Jun 11 18:17:39 sure, it builds for me too Jun 11 18:17:44 it doesn't work tho Jun 11 18:17:54 <[g2]> id doesn't build for me Jun 11 18:18:00 <[g2]> it Jun 11 18:18:01 ok that's strange Jun 11 18:18:38 trying strace build Jun 11 18:19:01 <[g2]> speaking of strange.... Do you remember the fix for the native cpp issue ? Jun 11 18:19:28 hmmm remind me what it is... Jun 11 18:19:50 <[g2]> Ok... maybe I don't have it installed that might be my first problem :) Jun 11 18:20:09 lol, I just checked and neither do I! Jun 11 18:20:28 apparently everything I built decided to use gcc -E (or whatever it is) Jun 11 18:21:13 native devel env - yes it's better than it was a few months ago, meaning I have less things on my manual checklist Jun 11 18:21:33 <[g2]> we should compare notes Jun 11 18:21:40 <[g2]> I'm restoring my native environ Jun 11 18:21:49 <[g2]> native build environ that is Jun 11 18:21:53 I still have: Jun 11 18:21:58 1) fix /usr/lib/libc.so and /usr/lib/libpthread.so Jun 11 18:22:03 2) install kernel includes: asm->asm-arm asm-generic linux Jun 11 18:22:38 and then fixing the fact that I don't like to remove busybox - I install coreutils over it - but that's my own issue Jun 11 18:23:32 NOTE: package strace-4.5.11: completed Jun 11 18:23:37 still builds for me Jun 11 18:23:40 <[g2]> hmmm Jun 11 18:23:43 <[g2]> strange Jun 11 18:24:56 [g2]: you alive ? Jun 11 18:25:06 <[g2]> alive and well thx Jun 11 18:25:20 cool :) just reading http://source.schooltool.org/irclogs/%23schooltool.2005-04-07.log.html Jun 11 18:25:29 oe-built flex should work for us again, but pb I think wants to add the env patch again (which breaks it for us) Jun 11 18:25:33 as I am having the same error building moin moin on openslug Jun 11 18:25:49 AFAIK bison is still broken - builds but does not work Jun 11 18:25:52 did you end up finding a solution other then building all of python on on the device? Jun 11 18:26:41 <[g2]> yes I built python 2.4.1 natively back then Jun 11 18:26:58 <[g2]> I've just cross-built python 2.4.0 Jun 11 18:27:10 <[g2]> I was thinkg about trying to install that Jun 11 18:27:20 <[g2]> I'm actually going to try running schoolbell agani Jun 11 18:27:22 <[g2]> I'm actually going to try running schoolbell again Jun 11 18:27:38 <[g2]> I've got my updated 64MB nslu2 Jun 11 18:27:53 <[g2]> and I'm sure the Zope3 will run much happier with that Jun 11 18:28:04 indeed it will Jun 11 18:28:49 <[g2]> I've been wanting to benchmark apache/mysql/php/zope3/etc... with the 64MB Jun 11 18:28:58 <[g2]> and maybe 128MB soon Jun 11 18:29:37 <[g2]> is moin moin in OE ? Jun 11 18:29:41 nup Jun 11 18:29:48 i dont think so Jun 11 18:29:59 <[g2]> is it just a python script ? Jun 11 18:30:06 <[g2]> script(s) Jun 11 18:30:13 oh god it is i think Jun 11 18:31:00 how stupid am I Jun 11 18:31:12 <[g2]> what's that ? Jun 11 18:31:38 Moin is in openembedded so I can probably just simply install it Jun 11 18:31:49 without building it on the nslu2 Jun 11 18:33:44 <[g2]> well there's a) compiles, b)runs kinda, and c) works Jun 11 18:37:28 Anyone know what controls which of the FEED_URIS gets used preferentially? (In openslug) Jun 11 18:38:55 <[g2]> there listed in the .bb's somewhere Jun 11 18:39:14 <[g2]> I ran across the list a while ago but I forget where Jun 11 18:39:59 top of openslug.conf isn't it? Jun 11 18:40:01 <[g2]> looks like gun-config is broke on openslug due to /usr/lib/libperl.so.5: undefined symbol: __udivsi3 Jun 11 18:40:02 I added the list to openslug.conf - but it's a list of one at present. I also have my local ftp feed in local.conf, but ipkg seems to use the remote in preference... Jun 11 18:40:54 <[g2]> jbowler-away, yeah that's the setup I'll be going to soon Jun 11 18:41:08 <[g2]> I've got may real NAS 90% ready to go Jun 11 18:41:17 <[g2]> actually server Jun 11 18:41:55 python installs file, but if I import unittest then help(unittest) I get 'No module named tempfile', likewise for python-tests... Jun 11 18:41:56 <[g2]> I'll probably have a slug setup too for openslug stuff Jun 11 18:42:30 The __udivsi3 is, I believe, a dynmic loading issue - the function is in libgcc Jun 11 18:42:57 <[g2]> there was a *bunch* of path setup stuff and config for Python that just seems missing / broken Jun 11 18:43:40 <[g2]> when I did the schoolbell stuff a couple months ago the natively build python was quite different than the installed python Jun 11 18:44:01 I have never been able to build a oe-built perl that does dynamic loading properly Jun 11 18:44:02 <[g2]> there's a bunch of python stuff like configuration which doesn't get installed Jun 11 18:44:15 What I have looks fine apart from the tempfile issue - and that just looks like a missing dependency. Jun 11 18:45:28 Ok, it's in python-io - missing dependency I guess Jun 11 18:46:10 <[g2]> did you install all the python ipkgs ? Jun 11 18:46:17 <[g2]> or just the core ? Jun 11 18:48:36 I installed python-core, then python-pydoc then python-tests and python-unittest, then python-io Jun 11 18:48:49 <[g2]> I had everything installed Jun 11 18:50:04 I think I would have to actually read some documentation to work out how to run the tests ;-) Jun 11 18:50:18 <[g2]> make test Jun 11 18:50:52 This is the cross compiled python... Jun 11 18:50:57 <[g2]> I think there's like 6-7 hundered or over 1K I forget Jun 11 18:51:14 <[g2]> I ran 'em when I build native and most passed Jun 11 18:51:26 <[g2]> there were some xml issues and some other stuff Jun 11 18:52:32 Ok, looks like it is "python -E -tt /Lib/test/regrtest.py -l" Jun 11 18:53:02 And regrtest.py is installed in /usr/lib/python2.4 Jun 11 18:53:55 Looks like this is something for lochalsh to do while I make some supper ;-) Jun 11 18:54:13 <[g2]> so "python -E -tt /usr/lib/python2.4/Lib/test/regrtest.py -l" Jun 11 18:55:21 I think so - that's just by looking at the Makefile. Jun 11 18:55:49 I'm installing everything except the 'old and deprecated' libs first. Jun 11 18:56:51 <[g2]> Ahh... that's nice..... Jun 11 18:56:59 * [g2] just installed tiny-vim Jun 11 18:57:18 <[g2]> s/ .*$// Jun 11 18:57:20 Yes, that sounds good, better than broken-vi Jun 11 18:57:39 vi is broken? Jun 11 18:57:46 <[g2]> it seemed *really* fast too Jun 11 18:57:50 python-curses doesn't install because it can't find libcurses5 Jun 11 18:58:10 <[g2]> I've already got ncurses installed Jun 11 19:06:50 [g2]: without /initrd my shutdown -r seems to be working fine. Jun 11 19:07:20 <[g2]> jbowler, mine did two the other day, but after 12 hours it didn't work Jun 11 19:07:24 <[g2]> maybe 20 Jun 11 19:10:02 The weird thing is that it always works fine when I do it from the jffs2 partition (i.e. if /dev/mtdblock4 is root). Jun 11 19:10:35 <[g2]> yes... that usually works for me to, but not 100% Jun 11 19:11:52 Well, now I get to test the /.recovery stuff, looks like I've just damaged by turnup disk /dev... Jun 11 19:12:46 <[g2]> really ? I just turnup disk just a couple hours ago and I'm running of the disk right now Jun 11 19:13:25 <[g2]> btw I still get the libcurses error too Jun 11 19:13:35 Yes, but I just tried to make it so that the disk boot would populate /dev without mounting ramfs (I have a deliberately hacked root). Jun 11 19:13:53 <[g2]> ah. Jun 11 19:14:17 python: I got a power cut when I was about to run the testsuite, so I haven't tried that yet. Jun 11 19:14:44 <[g2]> my install of all the python modules just finished Jun 11 19:14:47 I got three letters for you: UPS Jun 11 19:15:57 I've got two, one for the telephone system one for the computer closet (in the garage), but lochalsh is sitting on my desk for rapid power button access ;-) Jun 11 19:17:22 <[g2]> python seems much happier with 64MB Jun 11 19:18:14 <[g2]> python -E -tt /usr/lib/python2.4/test/regrtest.py -l Jun 11 19:18:19 <[g2]> is the magic chant Jun 11 19:18:24 <[g2]> is a magic chant Jun 11 19:18:52 I'd think lots of things would be happier with 64MB Jun 11 19:20:16 <[g2]> jacques, I hope they'll be even happier with 128MB :) Jun 11 19:20:26 [g2]: that path is wrong Jun 11 19:20:38 <[g2]> well the tests are running Jun 11 19:20:59 <[g2]> test_grammar Jun 11 19:20:59 <[g2]> test_opcodes Jun 11 19:20:59 <[g2]> test_operations Jun 11 19:20:59 <[g2]> test_builtin Jun 11 19:21:22 <[g2]> ... Jun 11 19:21:26 <[g2]> test_decimal skipped -- No module named decimal Jun 11 19:21:26 <[g2]> test_decorators Jun 11 19:21:26 <[g2]> test_deque Jun 11 19:22:30 [g2], you have a 128MB slug being created? nice Jun 11 19:22:49 <[g2]> jacques, not yet, but that'll be one of the next tests Jun 11 19:23:03 <[g2]> I'd like to verify the traces are all there Jun 11 19:23:36 <[g2]> I'm really pretty happy with the 64MB upgrade Jun 11 19:24:04 <[g2]> schoolbell will really be an interesting test Jun 11 19:24:17 [g2], with about 4x the usable RAM, 64 should be quite nice Jun 11 19:24:20 <[g2]> because it just *crawled* with Zope3 and python last time Jun 11 19:25:46 <[g2]> I really need to get the RV082 back Jun 11 19:26:02 <[g2]> cause that's got the 533 in it Jun 11 19:26:11 <[g2]> 425 @ 533 that is Jun 11 19:26:42 Ah, a bug. /etc/init.d/banner just writes into /dev/tty0 without checking if it exists. Subsequently makedevs attempts to create /dev/tty0, if the original /dev hasn't been overwritten makedevs gives up, at that point - consequently /dev ends up half empty. Jun 11 19:27:14 <[g2]> sounds like an important find Jun 11 19:27:19 <[g2]> good work! Jun 11 19:30:26 jbowler, d`oh :-) Jun 11 19:31:11 Not enough debugging though to make it work... Something is still going wrong with /etc/init.d/devices Jun 11 19:31:30 Oh well, something to do later, bbiab Jun 11 19:32:33 who is pushing unslung packages while jp30 is away? Jun 11 19:46:37 jf-work: can you cover for jp30 while he is away? Jun 11 19:48:46 rwhitby-away, you see my private msg ? Jun 11 19:56:11 rwhitby-away, I'm copying over the Source files for all the Unslung releases to Berlios.de and was wondering if the release notes are files or just cut and paste. Jun 11 19:57:53 cut and paste Jun 11 19:58:29 we should be ready for a 5.x release later today my time Jun 11 19:58:35 okay we have moved all the unslung releases to berlios and have left the sources on both places. Jun 11 20:03:28 cool - we need alpha testers for 5.x to register in the database Jun 11 20:04:07 we'll aim for 10 Jun 11 20:04:27 (since there is no real functionality change apart from the maintenance mode removal Jun 11 20:05:23 I should probably rebuild my unslung 5.x since rwhitby-away added a kernel patch after I built it last night Jun 11 20:09:48 yeah, probably. It's pretty benign unless you have a maxtor onetouch II Jun 11 20:10:09 speaking of which, should we disable that onetouch_detect process? Jun 11 20:13:03 pro: saves resources and probably not too many onetouch users, con: it removes some linksys compatibility Jun 11 20:14:51 does linksys actually use it? Jun 11 20:15:17 no idea, I would assume they do since they put it in Jun 11 20:15:29 interesting assumption :-) Jun 11 20:15:53 I couldn't immediately see how it is used. Jun 11 20:16:02 I guess we leave it in Jun 11 20:17:10 <[g2]> jacques, for the kernel includes what do you do ? Jun 11 20:18:10 [g2], tar up the dirs from the kernel build dir in oe Jun 11 20:18:17 that way they're properly configured Jun 11 20:18:48 install kernel includes: asm->asm-arm asm-generic linux Jun 11 20:19:00 the dirs asm-arm, asm-generic, and linux Jun 11 20:19:18 and you have to make the asm->asm-arm symlink Jun 11 20:19:19 <[g2]> under /usr/include Jun 11 20:20:00 tmp/work/openslug-kernel-2.6.11.2-r10/linux-2.6.11.2/include Jun 11 20:20:09 yeah and install them under /usr/include on the slug Jun 11 20:20:36 <[g2]> ok. Jun 11 20:21:04 <[g2]> so /usr/include/asm points to arm-asm Jun 11 20:21:09 yeah Jun 11 20:21:27 <[g2]> and the asm-generic and linux are there own directories Jun 11 20:21:37 yeah Jun 11 20:21:41 <[g2]> OK THX Jun 11 20:22:02 how do you deal with the /usr/lib/libc.so and /usr/lib/libpthread.so issue? manually edit like me? Jun 11 20:22:18 <[g2]> I edited the /usr/lib/libc.so Jun 11 20:22:33 <[g2]> I didn't run into the libpthread.so issue yet Jun 11 20:22:56 they are the same Jun 11 20:23:30 <[g2]> you mean libpthread points to libc.so Jun 11 20:23:41 no, they have the same problem and are fixed the same way Jun 11 20:23:52 <[g2]> yeah I thought so Jun 11 20:24:00 do you ipkg remove busybox ? Jun 11 20:24:09 <[g2]> not yet Jun 11 20:24:29 I consider coreutils necessary for devel env Jun 11 20:25:50 <[g2]> I ipkg installed coreutils Jun 11 20:26:08 <[g2]> maybe I got an error ? Jun 11 20:26:11 didn't it conflict with busybox? Jun 11 20:26:22 maybe that does not conflict - I have another note: Jun 11 20:26:29 3) not all busybox binaries are replaced Jun 11 20:26:29 in /bin: Jun 11 20:26:29 egrep fgrep grep cat chgrp chmod chown cp date dd echo false hostname kill ln Jun 11 20:27:02 so you end up with lots of symlinks in /bin to busybox while the same coreutils apps are in /usr/bin Jun 11 20:27:10 did somebody fix top? Jun 11 20:27:24 procps package you mean? Jun 11 20:27:29 I don't think so Jun 11 20:27:33 :( Jun 11 20:27:35 I had to build it manually Jun 11 20:27:40 k Jun 11 20:27:40 natively Jun 11 20:27:42 ugh Jun 11 20:28:13 I reported it on #oe but it might have gotten lost Jun 11 20:28:26 or nobody cared Jun 11 20:28:27 :-) Jun 11 20:30:35 <[g2]> jacques, COOL... now the native kernel config completed Jun 11 20:30:46 <[g2]> make oldonfig with the real .config Jun 11 20:41:01 just checked, procps is still broken - trying to do install --strip Jun 11 20:44:31 jacques, thanks. Jun 11 20:44:40 back in a bit. Jun 11 20:46:16 apart from disabling download, anything else which needs to go in the 5.4-beta release candidate? Jun 11 20:46:41 If so, speak in the next 8 hours, after that the release snapshot will be transferred to svn Jun 11 20:51:11 back later Jun 11 22:32:09 top seams to work ok for me Jun 11 22:34:16 having problems with perl however Jun 11 22:34:18 root@minoc:/var/www/www.kared.net/wiki# perl Jun 11 22:34:18 perl: symbol lookup error: /usr/lib/libperl.so.5: undefined symbol: __udivsi3 Jun 11 22:34:18 root@minoc:/var/www/www.kared.net/wiki# Jun 11 23:31:55 has anybody seen this problem before Jun 11 23:31:57 gcc: installation problem, cannot exec `as': No such file or directory Jun 11 23:31:57 root@minoc:~# Jun 11 23:33:51 looks like it can't exec as Jun 11 23:34:10 yeah :( Jun 11 23:34:26 any oe package that provides as? Jun 11 23:36:22 do you know what as is ? Jun 11 23:36:39 the assmbler Jun 11 23:37:02 right, then you should know where it comes from Jun 11 23:37:19 but, so as to not be a bastard I'll tell you - binutils Jun 11 23:37:34 you also need binutils-symlinks Jun 11 23:38:33 thanks alot :) Jun 11 23:38:48 :-P Jun 11 23:39:11 are you trying to make a native openslug devel environment? Jun 11 23:39:56 yep Jun 11 23:40:26 well, let me know if you have any more questions Jun 11 23:40:32 ok thanks Jun 11 23:43:32 I dont suppose you have gotten perl to install from oe? Jun 11 23:43:52 it wont work Jun 11 23:44:00 i build perl matively Jun 11 23:44:04 natively Jun 11 23:44:14 ahh ok, thats where im headed now Jun 11 23:44:15 thanks Jun 11 23:44:23 np, good luck Jun 11 23:44:49 soon we hope to have official openslug native-built packages feed Jun 11 23:44:59 for perl, emacs, and maybea few others Jun 11 23:45:59 ahh that would be good, i really should be writing down all the things im doing so when if i upgrade to a new version of openslug in the future i can recover it Jun 11 23:46:30 yeah, I keep notes like that Jun 11 23:51:04 BTW, the top we were talking about before is the procps one, not the busybox one Jun 11 23:51:48 ahh ok npbs Jun 11 23:52:05 root@minoc:~# ipkg search /usr/lib/libc.so Jun 11 23:52:05 libc6-dev - 2.3.2+cvs20040726-r17 - /usr/lib/libc.so Jun 11 23:52:16 /usr/lib/gcc/armeb-linux/3.4.3/../../../../armeb-linux/bin/ld:/usr/lib/gcc/armeb-linux/3.4.3/../../../libc.so: file format not recognized; treating as linker script Jun 11 23:52:31 sorry that paste is confusing Jun 11 23:52:57 the last line is the error im getting returned from ld, the top line is the package providing that file Jun 11 23:53:18 heh heh Jun 11 23:53:31 cat /usr/lib/libc.so Jun 11 23:54:05 Use the shared library, but some functions are only in Jun 11 23:54:05 the static library, so try that secondarily. */ Jun 11 23:54:05 *** BUG in libc/scripts/output-format.sed *** elf32-bigarm,elf32-littlearm Jun 11 23:54:05 GROUP ( /lib/libc.so.6 /usr/lib/libc_nonshared.a ) Jun 11 23:54:26 there's your problem Jun 11 23:54:45 replace the *** line with OUTPUT_FORMAT(elf32-bigarm) Jun 11 23:54:53 do the same for /usr/lib/libpthread.so Jun 11 23:58:07 still getting this error /usr/lib/gcc/armeb-linux/3.4.3/../../../../armeb-linux/bin/ld:/usr/lib/gcc/armeb-linux/3.4.3/../../../libc.so:5: syntax error Jun 11 23:58:10 line 5 is : Jun 11 23:58:20 GROUP ( /lib/libc.so.6 /usr/lib/libc_nonshared.a ) Jun 11 23:58:54 here is my entire file: Jun 11 23:58:57 Use the shared library, but some functions are only in Jun 11 23:58:57 the static library, so try that secondarily. */ Jun 11 23:58:57 OUTPUT_FORMAT(elf32-bigarm) Jun 11 23:58:57 GROUP ( /lib/libc.so.6 /usr/lib/libc_nonshared.a ) Jun 11 23:59:02 damn Jun 11 23:59:05 just a sec Jun 11 23:59:15 /* GNU ld script Jun 11 23:59:15 Use the shared library, but some functions are only in Jun 11 23:59:15 the static library, so try that secondarily. */ Jun 11 23:59:15 OUTPUT_FORMAT(elf32-bigarm) Jun 11 23:59:15 GROUP ( /lib/libc.so.6 /usr/lib/libc_nonshared.a ) Jun 11 23:59:18 there Jun 11 23:59:45 root@minoc:~# ./a.out Jun 11 23:59:45 Hello World Jun 11 23:59:46 :) **** ENDING LOGGING AT Sat Jun 11 23:59:56 2005