irclog2html for #nslu2-linux on 20041116

00:01.43VoodooZkergoth: I've read that you can copy devices in /etc/udev/devices and they will be copied to /dev/ automatically. I'll try it.
00:03.12VoodooZUnless it's a rh/fedora specific hack as I read about it on the fedora udev page.
00:08.00ka6sox-tiraneits amazing that I've held this connection for so long.
00:14.23KsileboUh oh
00:14.26VoodooZkergoth: coincidence? I just found out that under debian the file /etc/udev/links.conf does exactly that!!! Creates manual device nodes!!
00:14.35kergoth_i suspected
00:15.01*** join/#nslu2-linux KsiLaptop (
00:16.29VoodooZfunny that I wanted to get rid of it this morning! Were you going to add it?
00:18.04kergoth_yes, i was planning on pulling in links.conf and any other debian .conf files, along with debian's udev rules
00:18.33VoodooZok. I'll wait for yours then. Thanks.
00:40.14[g2]rwhitby-away, I'll have to straighten out the mail sometime. send mail is around, but I"m not sure it's working properly
00:40.44[g2]the merge failed on gentoo emerge (which almost never happens) and is quite odd in itself
00:40.44kergoth_jacques: i'm knee deep in i18n crap making glib build with uclibc anyway, i'll take a look at disabling its locale support
00:41.06[g2]my changeset # was k send -f -r 2554
00:41.37[g2]could you forward that changeset for the list ?
00:48.21jacqueskergoth_, I would really appreciate it - it would help save my sanity - I'm going through all the patches right now and being rather confused
00:48.44kergoth_i still dont know as much about this sort of thing as i'd like, but am getting there
00:52.51VoodooZkergoth: fyi, I found another bug in /etc/init.d/udev. It tries to call mknod with --mode which is again not supported under busybox.
00:57.48VoodooZactually, mknod --help says it takes options but anyways. I'm trying it now.
00:58.02kergoth_VoodooZ: it'd probably accept the short form, if there is one.
00:58.10kergoth_generally busybox commands dont accept the long forms of options
00:58.25kergoth_so -m [blah] would probably work
00:58.58VoodooZyep. that's it. it worked.
01:00.08kergoth_course the question is, should you send your oe changes that affect more than just nslu2 to the oe mailing list, or check them into the nslu2 repo
01:07.42glc has been upgraded to 1.2.2 causing oemake unslung-image not to build.
01:08.00kergoth_thats already fixed in upstream oe
01:08.43*** join/#nslu2-linux Christopher (
01:15.22*** join/#nslu2-linux ka6sox-tirane (
01:18.58rwhitby-websomeone with bk admin privs can do a pull from oe packages to nslu2-linux packages (kergoth, I assume by fixed upstream you meant oe packages, not packages-dev?)
01:19.10kergoth_i meant packages-devel. all development happens there
01:19.19kergoth_packages-devel syncs with packages whenever mickey feels like it, right now
01:19.23kergoth_need to get a cron job going
01:20.02rwhitby-webwould you suggest we reparent nslu2-linux packages to point to oe packages-dev ?
01:20.10rwhitby-webor not ...
01:20.27kergoth_that may be best, as long as you arent syncing constantly
01:20.38kergoth_packages-devel exists mainly becuase of the vast # of people doing clones from dialup
01:20.45kergoth_keeping packages locked for hours so development is stalled
01:20.57rwhitby-webwe sync whenever someone (usually me) feels like it
01:21.31rwhitby-webI can
01:21.36kergoth_you dont need admin to do a sync, actually
01:21.46rwhitby-webcan't do any bk stuff from here
01:21.48kergoth_bk pull from oe packages, BK_SILENT=1 bk push to nslu2 packages
01:29.04kergoth_there, synced.  pushing the latest nslu2 stuff into upstream now
01:32.38*** join/#nslu2-linux mickg10_ (
01:32.47*** part/#nslu2-linux mickg10_ (
01:38.29ka6sox-tiranesee ya in about 30hrs 1/2 way around the world.
01:41.01glcThanks Chris.
01:46.53*** join/#nslu2-linux _chris_ (
01:55.08Christopherrwhitby-web: I seem to be having trouble moving to unslung 3.x can I just load the image file, or are there other steps?
02:09.51glcrwhitby-web: Can you include nbd module? It might be useful as a root device.  Peteru wants NFS root, but the network block device might be smaller and easier.
02:10.38glcmd and raid1 are working quite nicely.
02:29.26*** join/#nslu2-linux mickg10 (
02:29.35*** part/#nslu2-linux mickg10 (
02:33.56peteru-workIt does not have to be NFS just on my account, as long as I can have a network shared rootfs that I can poke on the dev machine and get the slug to boot/run from it.
02:34.09peteru-workI'll be happy to take whatever works.
02:45.02rwhitby-webglc: go for it - add it to the -able kernel defconfig.  teach me what ndb is all about ...
02:47.12*** join/#nslu2-linux Christopher (
02:54.44glcThat is about it.
02:54.56jacquesthe thing I remember about nbd is that it works well but don't try to swap over it
02:55.31rwhitby-webOne other thing we need is the ability to store an ext3 image in a file mounted from vfat.  Anyone know how to do that?
02:56.12glcThe Sourceforge link states that swapping works.
02:56.21jacquessince when I wonder
02:56.37jacquesfine go ahead. swap over it - see if I care  :-P
02:56.55rwhitby-web[then you could have a vfat disk connected to your slug, and store your installed package tree in a file on the vfat  disk]
02:57.36rwhitby-webit'd be like a loopback mount, right?
02:57.47rwhitby-webcan we do loopback mounts on unslung at the moment?
02:57.50kergothloopback should work fine.
02:57.56kergothas long as its enabled, obviously
02:58.07peteru-worknbd filesystems can't be shared :(
02:58.48glcHow many slugs you plan on running? Slug farm?
02:58.51jacquesrwhitby-web, loop is enabled at least in able IIRC
02:58.52peteru-workand running on top of TCP... I suppose it's ok with modern networks
02:59.11jacquesmaybe it's not :-\
02:59.39jacquesthat must have been one of my custom configs that had it
03:00.26rwhitby-webjacques: can you add it to -able?
03:00.40rwhitby-webthat'll make the guy who wants loopback ISO mounts happy to ... ?
03:01.15peteru-workLimitations:It is impossible to use NBD as a root file system, as a user-land program is required to start
03:01.46jacquesrwhitby-web, yep I'll add it
03:01.53peteru-workin which case it may as well be a module, since there's nothing gained by having it in the kernle itself
03:02.13rwhitby-webpeteru-work: on another page there is a link to a package which lets you do it in an initrd
03:02.17rwhitby-webyeah, that's it
03:02.24rwhitby-webwe could do it in switchbox
03:02.29jacquespeteru-work,  well one way to look at things is, until we run out of the 1MB space - maybe we should build stuff into the kernel to save space ?
03:04.07rwhitby-webto go into the unslung-able kernel, it needs to be something that at least camp #2 has a chance of using ...
03:04.47rwhitby-webif it's so esoteric that you need to be an OpenSlug-type anyway, then there's no incentive to put it in the unslung kernel ...
03:05.15rwhitby-webspeaking of a CSR-enabled OpenSlug in OE .... ;-)
03:05.17jacquesheh what are we talking about?
03:05.26kergothCSR is in oe. the ethernet driver isnt yet.
03:05.29kergothi'll take care of it
03:05.35jacquesI'm confused are we talking about nbd or loop?
03:05.38rwhitby-web~praise kergoth
03:05.40jbotAll hail kergoth!
03:05.56rwhitby-webjacques: I was referring to nbd compiled into the kernel
03:06.23rwhitby-webloop is something that I think lots of camp #2 would use
03:06.27glcHow about as a module, like md and raid?
03:06.33KsileboAnyone going to be playing Half Life 2?
03:06.41rwhitby-webyeah, module is always good
03:07.06peteru-workI think the policy should be that if it is something that you may reasonably require at boot time, it should be in kernel.
03:07.15peteru-workmodules for rest.
03:07.18jacquesrwhitby-web, OK but do you want me to put loop builtin or as a module ?
03:07.25*** join/#nslu2-linux dyoung-web (
03:07.38peteru-workbut maybe that policy can take effect when we run out of room
03:07.46rwhitby-webagreed - if it's something that an unslung user (as opposed to an openslug user) may reasonably require at boot time, then it goes in the kernel
03:08.08rwhitby-webjacques: for loop, I think builting
03:08.11jacquesI want reiserfs!  (j/k)
03:08.21peteru-workme too
03:08.22jacquesthat would more than blow our budget
03:08.29jacques(builtin that is)
03:08.32dyoung-webrwhitby-web, I think I have the same problem that you were having last night.,
03:08.37rwhitby-webyeah, size is a factor too
03:08.45rwhitby-webI presume that loop is quite small
03:08.52jacquesyeah loop is small
03:08.56dyoung-webloop works if you enable it in ther kernel.
03:09.03rwhitby-webdyoung-web: what problem?
03:09.12peteru-workloop requires an underlying fs to mount, so in theory could be a module.
03:09.13dyoung-webthe booting of jffs2 as rootfs.
03:09.30rwhitby-webdyoung-web: check your linuxrc to make sure it is umounting
03:09.41dyoung-webI built loop as a module; didnt try build it in.
03:10.08dyoung-webits supposedly unmounting.  it seems to be ignoring the real-root-dev.
03:10.25jacquesoh BTW, here is static switchbox in (hacked) oe: 118220 Nov 15 18:42 busybox*
03:10.50dyoung-weboh yeah, is init enabled in switchbox?
03:11.13rwhitby-webpeteru-work: agreed - another item to the policy: if it's likely to be used by unslung users as part of switchbox (i.e. we make it an official option for booting from) then it should be in kernel, as long as there is enough room
03:11.50jacqueshmm that's using my config - I need to try it using the latest switchbox config (been hacking up oe so havent pulled on that box lately)
03:11.54rwhitby-webdyoung-web: it is at the moment, but it doesn't need to be - I'll probably take that test out of the linuxrc
03:12.24dyoung-webinit makes things weird because there isnt a inittab.
03:12.33dyoung-webor if there it its not optimized
03:12.49rwhitby-webyeah, I know - jacques can you remove init from the switchbox in OE please and push?
03:12.50dyoung-webI didnt check.  Either way it gives me loads of "bummer" messages when it cant find my rootfs.
03:13.22jacquesrwhitby-web, OK I am about to add loop builtin to able kernel and remove init from switchbox
03:14.43glcThe eternal - continuous compile . . .
03:15.08rwhitby-webok, the test in linuxrc stays as is - it's checking for /bin/init in the jffs2 partition, not the switchbox ramdisk.  That's correct.
03:16.14rwhitby-webwhat's the consensus on whether the optional running on /unslung/rc.linuxrc on a jffs2 partition stays or goes?
03:17.12dyoung-webthe 2nd to the last stanza of slungbox linuxrc seems funny.
03:17.16rwhitby-webmy idea is that we can have a single /linuxrc for unslung and openslug (with stick redboot), so I'm thinking that the /linuxrc should not reference any unslung-specific stuff in the jffs2 portion.
03:17.59rwhitby-webdyoung-web: that's for the case where the /linuxrc is actually inside a single big ramdisk, not in a switchbox
03:17.59dyoung-webIt claims to be wanting to mount from ram0, but it echoes 0x1f04 into real_root_dev .
03:18.14rwhitby-webok - that's a bug
03:18.26rwhitby-webit should echo 0100
03:18.35dyoung-webyeah, thats what I figgered. ;-)
03:18.55rwhitby-webjacques: can you fix that and push?
03:19.00CIA-703jacques 07 * r1.2556 10packages/ (2 files in 2 dirs): removed init from switchbox and added loop block dev builtin to -able kernel
03:19.28rwhitby-webline 57 of linuxrc in switchbox
03:19.51rwhitby-webAnd I'm thinking that line 45 should be deleted.
03:20.27jacqueschange this: /bin/echo "0x1f04" > /proc/sys/kernel/real-root-dev  to 0x0100 ?
03:20.54jacquesand delete line 45 ?
03:21.00rwhitby-webto switch it back to mounting from ram0 - yeah I know it's superfluous cause we've already set it to that at the top
03:21.23rwhitby-webshould be no references to unslung in the jffs2 stanza
03:21.50rwhitby-webHmm - what about line 42 ?
03:22.15rwhitby-webdelete that too
03:22.20dyoung-webappend -able to it?
03:22.51rwhitby-webno, cause the idea is that this /linuxrc and switchbox can be used for openslug (with stock redboot) too
03:23.20rwhitby-webwith openslug having the 128Kb switchbox and a big jffs2
03:23.27jacquesthis C-x C-q thing only half works - after I exit emacs, it I want to edit the file again, it asks if I want to seal the lock from myself - and if I say yes the changes are lost
03:23.58*** join/#nslu2-linux mickg10__ (
03:23.59rwhitby-webexit emacs ?  why would you ever do that ?
03:24.15rwhitby-webC-x C-q to check it back in again too, right?
03:24.17mickg10__to start vi ? :)
03:24.27jacquesrwhitby-web, oh I didn't know about that
03:24.29rwhitby-webmickg10__ speaks!!!!!!
03:24.38rwhitby-webHi mickg10__ :-)
03:24.43jacquesnothing like an editor holy war to get ppl talking
03:24.52mickg10__true that...
03:25.03jacquesrwhitby-web, what's the general comment for what I just did ?
03:25.15jacques"deleted some stuff rwhitby told me to"
03:25.29rwhitby-webRemoved unslung specific references from the switchbox linuxrc jffs2 stanza
03:25.43*** join/#nslu2-linux seadevil (
03:26.02rwhitby-webwho was it yesterday who was saying they had never seen mickg10__ speak ?
03:26.28[g2]Hi seadevil
03:26.36*** join/#nslu2-linux Christopher (
03:26.50rwhitby-web[g2]: are you diskless-unslung yet?
03:27.02seadevilnot yet
03:27.06[g2]I wish you could swing by "Taste of Saigon"and e-mail me some food
03:27.27[g2]I've got unsulung images
03:27.34[g2]can I just load them
03:28.02rwhitby-webyou have to flash to change the FIS directory
03:28.13rwhitby-webyou have to build an image using slugimage
03:28.23CIA-703jacques 07 * r1.2557 10packages/busybox/switchbox-1.00/linuxrc:
03:28.23CIA-7Removed unslung specific references from the switchbox linuxrc jffs2 stanza
03:28.23[g2]I've got slugimage
03:28.24CIA-7changed /bin/echo "0x1f04" to /bin/echo "0x0100" in ramfs stanza
03:29.21rwhitby-webcan someone put slugimage into OE ?  should be a simple copy and edit of the slugtool package (which it will eventually replace)
03:29.44rwhitby-webthen we can create a flashable jffs2 8mb image in OE
03:29.45[g2]rwhitby-web, slugimage needs a clean
03:29.51[g2]to clean all the images up
03:30.17[g2]from the unpack
03:30.27rwhitby-webbut you need them for your next pack
03:30.47rwhitby-web[most of them at least]
03:30.54[g2]not if your unpacking first and want to make sure there gone
03:31.02[g2]like now
03:31.06jacquesslugimage differs from slugtool in that it does not need to be built
03:31.21[g2]slugimage in oe will require perl-native I'd guess
03:31.30rwhitby-webif you don't want them written out, then set the filename for that to /dev/null on the unpack
03:31.33[g2]cause oe build everything it needs
03:31.53rwhitby-webyep -we'll need to add some nslu2-package-symlinks ...
03:32.00[g2]I'm just talking about creating a clean state
03:32.50[g2]anyway it was just a suggestion
03:32.54jacquesdo we really want to have oe build perl when it's 99.9% probably already installed ?
03:33.01jacquesthat sounds rather gratuitous to me
03:33.12[g2]I don't
03:33.15rwhitby-web[g2]: Hmm - how does it know what to delete - you might have used options on the first run to change filenames
03:33.31[g2]all the non .img files
03:33.45kergothjacques: i'm making all of OE obey a 'USE_NLS' variable.
03:33.54kergothso far i've got uclibc and every gettext based package obeying it
03:33.58kergothhavent tested any builds yet
03:34.09rwhitby-web[g2]: bit dangerous - I do my image building in a directory with lots of other files in it
03:34.34rwhitby-webkergoth: cool, so we can just set USE_NLS to 0 in unslung.conf ?
03:34.41jacqueskergoth cool I think that will be rather useful (to more than just us :-)
03:34.43kergoththats the idea
03:34.51rwhitby-web~praise kergoth
03:34.53jbotAll hail kergoth!
03:34.55[g2]rwhitby-web, I meant Redboot  SysConf  Trailer  ramdisk.gz  vmlinuz
03:35.27rwhitby-web[g2]: yeah, but you might have used the -b or -s or -t or -r or -k options to write them to other files
03:36.03*** join/#nslu2-linux mickg10__ (
03:36.27rwhitby-webso on the second run, slugimage can't be sure which you want deleted, unless you tell it explicitly, in which case you might as well have typed rm instead :-)
03:37.15rwhitby-webbut feel free to add something if I'm just not getting it ...
03:38.05mickg10__well, I just got the just desserts for attempting to start a holy war :)
03:38.13mickg10__My system just kernel-panicked.
03:38.19[g2]rwhitby-web back to diskless
03:38.24kergothwonder if i should create a gettext oeclass, rather than using whether or not 'gettext-native' is in DEPENDS to decide whether or not it supports it
03:38.47[g2]I've got slugimage and the unslung-able-nslu2-20041115231519.img in the directory
03:39.09[g2]I'll unpack it and copy the .jffs2 fs in there
03:39.18[g2]then I need the right slugimage command correct ?
03:40.11*** part/#nslu2-linux mickg10__ (
03:40.59rwhitby-lunchyou need to copy in the switchbox .ext2.gz too
03:41.12[g2]ok, it's built
03:41.34rwhitby-lunchthen you pack with -r Ramdisk:foo.ext2.gz:Flashdisk:foo.jffs2
03:42.24rwhitby-lunchwill jacques, dyoung or [g2] get to be Unslung 3.x #2 ???
03:42.39rwhitby-lunchthe race is on ....
03:43.12jacqueshmm, what local.conf changes are required ?
03:43.13rwhitby-lunch[g2] is wondering: "Which will upload faster, the web interface or upslug?"
03:43.21[g2]are you running standard or albe
03:43.30rwhitby-lunchneeds to be able
03:44.05rwhitby-lunchjacques: don't think I made any local.conf changes - it's just a normal -able unslung build
03:44.44jacquesthen where is jffs2 image type added ?
03:45.00jacquesis that committed ?
03:45.30rwhitby-lunchyep, in unslung-able-image
03:45.56jacquescool - so can just start an unslung build as my first step to getting 3.0-ified ?
03:45.58rwhitby-lunchoh - in unslung-standard-image actually - the jffs2 and ext2.gz are both built for both -standard and -able
03:46.18rwhitby-lunchunslung-image will get you everything you need (including switchbox)
03:46.44rwhitby-lunchthen you just need to pack it together using the new slugimage which pads the last variable partition
03:47.03[g2]rwhitby-lunch, like this? ./slugimage -p -o unslung-3.0.img -r Ramdisk:unslung-switchbox-nslu2-20041115231519.rootfs.ext2.gz:Flashdisk:unslung-able-nslu2-20041115231519.rootfs.jffs2  
03:47.05jacques`k sounds good I'll start the build
03:47.07rwhitby-lunchthat's the bit for which we need slugimage in OE
03:47.23rwhitby-lunch[g2]: that should do it
03:47.36[g2]Invalid syntax in --ramdisk
03:47.59[g2]not :
03:48.44[g2]that's a flashable image ? :)
03:48.46dyoung-webbkpulling, oemaking, so I can be #2 again.
03:49.07rwhitby-lunch[g2]: should be - did the directory look right?
03:49.08[g2]Um, I'm about do flash so you better hurry up
03:49.31dyoung-webI should be expidennt and just fix the ramdisk.gz linuxrc.
03:49.34[g2]Allocated <RedBoot> from 0x50000000 to 0x50040000
03:49.34[g2]Allocated <SysConf> from 0x50040000 to 0x50060000
03:49.34[g2]Allocated <Kernel> from 0x50060000 to 0x50160000
03:49.34[g2]Allocated <Ramdisk> from 0x50160000 to 0x50300000
03:49.34[g2]Allocated <Flashdisk> from 0x50300000 to 0x50760000
03:49.35[g2]Allocated <FIS directory> from 0x507E0000 to 0x50800000
03:49.35jacquesone of my machines is seriously screwed up with regard to bk
03:49.48jacqueswhen I bk pull it takes forever
03:49.56[g2]Wrote 0x00040000 bytes (0x00000000 to 0x00040000) from <RedBoot> into "unslung-3.0.img"
03:49.56[g2]Wrote 0x00020000 bytes (0x00040000 to 0x00060000) from <SysConf> into "unslung-3.0.img"
03:49.56[g2]Wrote 0x00100000 bytes (0x00060000 to 0x00160000) from <Kernel> into "unslung-3.0.img"
03:49.56[g2]Wrote 0x001A0000 bytes (0x00160000 to 0x00300000) from <Ramdisk> into "unslung-3.0.img"
03:49.56[g2]Wrote 0x00460000 bytes (0x00300000 to 0x00760000) from <Flashdisk> into "unslung-3.0.img"
03:49.57[g2]Padded 4 blocks before <FIS directory> in "unslung-3.0.img"
03:49.59[g2]Wrote 0x00020000 bytes (0x007E0000 to 0x00800000) from <FIS directory> into "unslung-3.0.img"
03:50.01[g2]Rewound 0x00000010 bytes before <Trailer> in "unslung-3.0.img"
03:50.05[g2]Wrote 0x00000010 bytes (0x007FFFF0 to 0x00800000) from <Trailer> into "unslung-3.0.img"
03:50.05rwhitby-lunch[g2]: is that latest slugimage?
03:50.07kergotheh, dood
03:50.09kergothdont flood
03:50.16dyoung-webWe'll see if [g2] has the bummer message now too.
03:50.22[g2]it's just a couple lines more than 4
03:50.26dyoung-webwhich means I still might have chance.
03:50.31rwhitby-lunchwhat's the roots of the don't flood thing anyway?
03:50.37kergoth15 is more than a couple more than 4.
03:50.45kergothi dont have time to teach you irc ettiquite. go read the freenode rules
03:50.49[g2]it was 9
03:51.00kergothno, it wasnt. look again.
03:51.16[g2]it was 6 then 2 from jacques then 9
03:51.27rwhitby-lunchI know the rules, I'm asking about what the background is behind why the rules were set that way ...
03:51.32kergothlearn to use irc.
03:51.43kergothrwhitby-lunch: flooding can desync the channel.  disconnect users. etc.
03:51.44rwhitby-lunchdid people used to read IRC on five-line screens or something?
03:51.53kergothi dont feel like explaining to the freenode ircops that'll joint he channel if that happens
03:51.57kergothdo you?
03:51.59[g2]it used to be async
03:52.04rwhitby-lunchah - it's a protocol thing?
03:52.27rwhitby-lunchis it historical, or will it still desync today ?
03:52.44kergothdunno on the desyncing. it *will* disconnect users
03:52.53kergothif you flood anyone offline, people will not be happy with you
03:53.18dyoung-webstill get chance....
03:53.33jacquesI wanna be #4 anyway
03:55.23glcWhere is slugimage?
03:55.25rwhitby-lunchjacques: if we're trying to replicate the good old days, then we need to get the nefarious mr buzbee in here to be #5.
03:55.35[g2]glc cvs
03:56.00peteru-workI remember the week irc came to Australia. An exchange student brought it back with him on a Mag tape :-) We had 4 users, two of them were on TTYs, actually watching the chat print on fanfold :-)
03:56.09[g2]and someone to test some else's work first
03:56.16kergothpeteru-work: you're showing your age :)
03:57.07jacquesI was addicted to the original Bitnet Relay in college
03:57.17peteru-workI was a young-un then, able to drink Black Sambuca by the pint ;)
03:57.34[g2]Bummer, can't open /dev/tty2
03:57.34[g2]Bummer, can't open /dev/tty3
03:57.34[g2]Bummer, can't open /dev/tty4
03:57.43rwhitby-lunchheh - read the description of flood here :
03:58.30kergothjeeze, that site is taking forever to load
03:58.51rwhitby-lunchit's 0wnzing your machine while you wait
03:58.56jacquesand the text is like 1 point
04:00.25rwhitby-lunch[g2]: no go?
04:00.26peteru-workserved over dialup from a boat near a Thail village.
04:00.34[g2]rwhitby-lunch, no go
04:00.58[g2]i get the bummer and it doesn't look like proc is mounted because that was df says
04:01.00glcIs there another CVS?  I can't find it in the unslung 1.x CVS.
04:01.39jacqueswhat broke since rwhitby-lunch made his 3.0 image ?
04:01.50[g2]Root filesystem cannot be found - rebooting ...
04:01.50[g2]+ /sbin/reboot
04:02.53[g2]+ /bin/mount -rt jffs2 /dev/mtdblock4 /mnt/tmpmnt
04:02.53[g2]mount: Mounting /dev/mtdblock4 on /mnt/tmpmnt failed: No such device
04:02.55kergoth| /home/kergoth/code/build-tosa/tmp/work/gcc-cross-3.4.2-r1/gcc-3.4.2/build.i686-linux.arm-linux-uclibc/arm-linux-uclibc/libstdc++-v3/include/arm-linux-uclibc/bits/codecvt_specializations.h:423: error: there are no arguments to `__iconv_adaptor' that depend on a template parameter, so a declaration of `__iconv_adaptor' must be available
04:02.56rwhitby-lunchglc: cvs module "slugimage"
04:03.01kergothalright, this looks familiar
04:03.02kergothhere we go
04:03.44[g2]isn't this an old device-txt file ?
04:04.08rwhitby-lunchdid you rebuild from scratch?
04:04.44[g2]no, I had a rebuild from a couple hours ago, and some recent pulls
04:05.15rwhitby-lunchHmm - looks like switchbox is missing /mnt/tmpmnt
04:05.18jacquesi thought the 8 mtdblock thing was committed
04:07.33rwhitby-lunchnope - it's there.
04:07.50rwhitby-lunch[g2]: did you rebuild switchbox?
04:08.07[g2]not in the last hour
04:08.33rwhitby-lunchsince I pushed this morning
04:08.47rwhitby-lunchafter making my son's breakfast
04:08.52[g2]it was 5 hours ago
04:10.00dyoung-webOkay, I'm not al;one.
04:10.02dyoung-webStill get chance.
04:10.16jacquesmy build is up to glibc
04:10.30rwhitby-lunch[g2]: loopback mount the switchbox image, and make sure it has /dev/mtdblock4 and /mnt/tmpmnt in it
04:11.29rwhitby-lunchthe one I built a while ago does
04:11.53rwhitby-lunchremember that switchbox is an ipk, and it gets loaded into unslung-switchbox-image
04:12.35rwhitby-lunchan early version of the switchbox ipk didn't have /mnt/tmpmnt
04:13.43[g2]um actually the appear to be in there twice
04:13.55[g2]which seems rather odd
04:14.04kergothcome on baby.. i think i fixed the gcc w/ uclibc w/ no locales bug
04:15.48kergothmoment of truth.. libstdc++ build
04:16.18kergothshit, still failed
04:16.27kergothno, those were warnings
04:16.32kergothi think it worked
04:16.35kergothhey, it did
04:16.39kergoththat was easy
04:16.57jacquesi suck
04:17.02jacquesi might as well give up now
04:17.10kergothfuck, false alarm
04:17.14kergothstill failed, just later
04:17.30kergothwhen mjn3 gets back, i'm going to throw something at him
04:18.24rwhitby-lunchSo this page ( says that FreeNode's server throttle floods anyway, and just automatically kick the person who is flooding ...
04:19.23rwhitby-lunchand I can't find an actual *rule* about flooding here: or here: or here:
04:19.33dyoung-webI remember IRC in the "early" days.  it was quit handy during the first gulf war.
04:19.50rwhitby-lunchso while I agree that it's bad form, and was historically malicious, the question still stands: is it a real problem today?
04:20.16kergothwell, i've seen floods knock more people off than just the flooder. on freenode.
04:20.21kergothso i wouldnt say its full proof.
04:21.05dyoung-webThis build is taking a while.  [g2] might beat me.
04:21.22[g2]I'm back to building gcc
04:21.24rwhitby-lunchok, kergoth has empirical evidence.  that answers the question
04:21.32jacquesI'm packaging glibc
04:21.43[g2]binutils finished
04:21.52[g2]how big was the channel ?
04:21.55[g2]100 users ?
04:22.12rwhitby-lunchdon't forget that jacques has the quad opteron 64 Titanium server farm backing him ...
04:22.41[g2]actually it mostly in ccache
04:23.03[g2]smp (HT) + 1G Mem is pretty speedy
04:23.05dyoung-webI need to un-scurvy-ify my build box.
04:23.17kergothremember that if you rebuild the toolchain (i.e. doing a full oe build), ccache is largely worthless
04:23.27kergothsince changing toolchains dirties the cached objects for everything built with it
04:23.58rwhitby-lunchif history repeats itself, it will be dyoung, then [g2], then jacques ...
04:24.54[g2]plus oe single threads everything because glibc install is hosed
04:25.45[g2]so it'll be tomorrow at that rate
04:25.56[g2]35 minutes here
04:26.16kergothoe isnt threaded regardless.  dont know what you mean by that
04:26.33[g2]the MAKE overrides
04:27.48kergothoverriding MAKE *will* break things in oe.
04:27.58kergothglibc wont do_install, just to start with
04:28.03kergothmuch less all the other packages that cant deal with it
04:28.10kergoththere's no such thing as an OEMAKE variable
04:28.27kergothif you put -j in oe's MAKE variable, it *will* break the build.
04:28.33kergothjust keep that in mind
04:28.43kergothoh those FUCKERS
04:28.51kergothjacques: look at that shit.
04:28.53kergothjacques: you cant explicitly disable it.
04:28.56kergothjacques: it drops back to auto
04:29.07kergoth6 lines is not 15. fuck off.
04:29.24[g2]hey mine was 6 then 9
04:29.42jacqueskergoth wow where is that?
04:29.53jacquesin gcc config ?
04:30.20kergothjacques: libstdc++'s configure.  i'm trying to disable its support for clocales completely.
04:30.24jacqueswhy why why
04:30.24kergothcant figure out how
04:30.27kergothwading through its macros
04:30.40[g2]it's in the log file that dyoung has a 6 + 2 jacques then 9
04:31.00kergoththe fact that jacques typed in the middle of your flood means nothing.
04:31.09[g2]I paused
04:31.11kergothi dont really give a shit. just dont do it in the future
04:31.24jacquesmaybe buildroot rm's some stuff to make libstdc++ not detect it
04:31.38kergothjacques: could be, i havent looked at buildroot at all.
04:31.38jacquesI didn't understand all the rm's in the makefile
04:36.09jacquesbuildroot actually has the rm's commented out:      #rm -rf $(STAGING_DIR)/info $(STAGING_DIR)/man $(STAGING_DIR)/share/doc $(STAGING_DIR)/share/locale      #rm -rf $(TARGET_DIR)/share/locale $(TARGET_DIR)/usr/info \
04:36.11rwhitby-lunchkegoth has been appointed #nslu2-linux ircop apparently :-)
04:36.48kergoththats just cleanup aftter the install
04:36.50kergothwouldnt affect this
04:37.27jacquesoh duh right
04:37.30kergothdamn you stdc++
04:37.31kergothchecking for C locale to use... uclibc    
04:39.24jacquesone of the patches does patch the uclibc locale selection config in gcc, but I tried that patch in oe (it was only slightly different from the one oe already has anyway)
04:42.28jacqueskergoth in the config.log of gcc-final in buildroot I see this: --enable-clocale=generic
04:42.45kergothaha, so its not using the uclibc C locale.
04:43.01kergoththatd explain it.
04:43.12kergothare they specifying that on their call to configure?
04:43.51jacques(I don't think so tho because I've looked at that like a million times)
04:44.31jacquesnope there's nothing about locale in the configure line
04:44.34kergoth/home/kergoth/code/build-tosa/tmp/work/gcc-cross-3.4.2-r1/gcc-3.4.2/build.i686-linux.arm-linux-uclibc/arm-linux-uclibc/libstdc++-v3/include/iosfwd:46:29: bits/c++locale.h: No such file or directory
04:44.39kergothdisabling it completely doesnt work
04:44.44kergothyou must specify one
04:45.04jacquesthere is however a --disable-nls
04:45.13kergothwe do that too, it doesnt affect the clocale
04:45.16kergoththey're independent
04:46.26jacquesok this is weird - later in the same file I see another configure line with: --enable-clocale=generic
04:46.53kergothin what file?
04:46.54jacquesoh, that's from the native compiler
04:46.59jacquesignore that  :-\
04:47.04jacquesthis is driving me insane
04:47.18kergothlook at the configure in arm-linux-uclibc/libstdc++ or whatever
04:47.22kergothin the compile output from buildroot
04:47.26kergoth(if you've got it handy)
04:49.17jacquesthis is weird - in the gcc-final I can't fing libstdc++ anywhere
04:49.30kergothwhat are they setting for languages?
04:49.42jacqueshmm sigh
04:49.50jacquesthis buildroot I might have built with no c++
04:49.56kergothah, heh
04:49.58kergoththatd do it
04:50.19jacquesbut I know that buildroot will build c++/libstdc++ without uclibc locales enabled
04:50.42[g2]hey how do I oemake just the zImages ?
04:50.49jacquesit doesn't take too long to do a buildroot build - I'm going to do one with c++
04:54.06jacques[g2], you should be able to do oemake unslung-able-kernel
04:55.01dyoung-webWell, I have my 7 partitio ns now.
04:55.04[g2]jacques, thx. I did an oemake -c clean unslung-standard-kernel then unslung-able-kernel
04:55.24jacquesok this does not look right:  73147 Nov 15 21:36 unslung-switchbox-nslu2-20041116045113.rootfs.ext2.gz
04:55.36*** join/#nslu2-linux kergoth (~kergoth@2001:5c0:8fff:fffe:0:0:0:c29)
04:55.59rwhitby-lunch[g2]: unslung-able-kernel
04:56.13dyoung-webmine is 875k
04:56.51rwhitby-lunchsame hre
04:57.28rwhitby-webjacques: if your's has everything in it, then that's good :-)
04:57.41rwhitby-webbut I'd say you're missing the dynamic libraries
04:58.27rwhitby-webwhich get installed into unslung-switchbox-image by switchbox requiring glibc which is provided by nslu2-linksys-libs
04:58.41jacquesrwhitby-web, what did I do wrong _ there are no shared libs
04:58.42kergothjacques: forcing to generic worked.
04:58.49kergothi'll make it obey USE_NLS and check it all in
04:58.57jacqueskergoth fantastic I can't wait to try it
04:59.21jacquesrwhitby-web, something not checked in ?
04:59.38rwhitby-webjacques: we probably should make nslu2-linksys-libs explicit  in UNSLUNG_BASE_PACKAGES in unslung-switchbox-image
04:59.50rwhitby-webyou're shlibdeps must be different from previous builds
05:00.22rwhitby-webmake it explicit in unslung-switchbox-image and it will work
05:00.32rwhitby-webdyoung-web: go or no go ?
05:00.57[g2]should I stay or should I go now ?
05:01.12[g2]If I stay there will be trouble
05:01.13rwhitby-webif I stay there will be trouble
05:01.24[g2]and I if flood it will be double
05:01.26rwhitby-webso I gotta let you know ....
05:01.52rwhitby-webs/I gotta let you/you gotta let me/
05:02.07rwhitby-weball together now ....
05:02.45[g2]dyoung must be updating the db
05:04.08dyoung-webI no go yet.
05:04.46dyoung-webBut I'll let you know.
05:05.14dyoung-webupslug kicks ass.
05:05.22dyoung-web~worship Christopher
05:05.24jbotACTION kneels before Christopher.  "I'm not worthy!"
05:05.35jbot:), dyoung-web
05:07.06jacquesthis looks more like it: 873901 Nov 15 22:05 unslung-switchbox-nslu2-20041116060405.rootfs.ext2.gz
05:09.11jacqueswhy are most of the devices doubled in switchbox-image ???
05:10.39jacquesyeah I saw your q [g2] but I didn't understand what you meant at the time
05:11.01jacquesthis is bizarre
05:11.15jacquesbrw-r--r--  1 root root  31,   0 Nov 15 22:05 mtdblock0
05:11.22jacquesalmost all devices are like that
05:11.53[g2]jacques, are you logged in ?
05:12.05jacquesno I just mounted the switchbox image to look at it
05:12.16jacquesto make sure the shared libs were in and I saw that
05:12.25[g2]ah, I'm flashing with an updated kernel
05:12.49jacqueshow can two files in the same dir have the same name???
05:13.24[g2]different inodes ?
05:13.30jacquessame inodes
05:13.56rwhitby-webmakedevs and also genext2fs
05:14.18rwhitby-webmkfs.jffs2 can't take a device_table.txt file, so I had to do makedevs.
05:14.34rwhitby-webif makedevs works for ext2.gz too, then we just just use that
05:14.55jacquesbut how can two files have the same name - I didn;t think that was possible
05:15.23[g2]# df
05:15.23[g2]Filesystem           1k-blocks      Used Available Use% Mounted on
05:15.23[g2]rootfs                   10070      9970         0 100% /
05:15.23[g2]/dev/root.old            10070      9970         0 100% /
05:15.32rwhitby-weband not pass the device_table.txt file to genext2fs - I seem to remember a good reason why we did that though - perhaps something to do with permissions - can makedevs create devices owned by root in ext2.gz if you are not running as root?
05:16.07rwhitby-web[g2]: you didn't use the latest slugimage which extends the last variable partition to fill the space ?
05:16.47rwhitby-webif makedevs can create what we need in the ext2.gz image, then we can just use that - but that must be verified by someone ...
05:16.54[g2]rwhitby-web, I think I did
05:17.09rwhitby-webDid slugimage report that it was padding anything?
05:17.15rwhitby-webif it did, then you didn't :-)
05:17.34rwhitby-web[or there's a bug in slugimage]
05:18.31[g2]no, not mention of padding
05:18.42[g2]I just pulled cvs and re-ran
05:19.48dyoung-webthis "bummer" stuff is making me cranky.
05:20.02[g2]I didn't get the bummer
05:20.08[g2]the second time around
05:20.23dyoung-webbut your rootfs is too big to be the jffs2 rootfs.
05:20.53[g2]no it's the initrd
05:21.31[g2]the linuxrc did quite work
05:21.41dyoung-webthe initrd is around 4M for Unslung 3.x.
05:21.48rwhitby-webrootfs                    5760      5024       736  87% /
05:21.54rwhitby-webthat was my jffs2 from last night
05:22.04[g2]the jffs2 isn't mounted
05:22.26[g2]Creating 6 MTD partitions on "IXP425 Flash":
05:22.26[g2]0x00000000-0x00040000 : "RedBoot"
05:22.26[g2]0x00040000-0x00060000 : "SysConf"
05:22.27[g2]0x00060000-0x00160000 : "Kernel"
05:22.27[g2]0x00160000-0x004e0000 : "Ramdisk"
05:22.27[g2]0x004e0000-0x007e0000 : "unallocated space"
05:22.29[g2]0x007e0000-0x00800000 : "FIS directory"
05:22.37jacquesI can't get my brain around this two files with same name thing
05:22.50rwhitby-web[g2]: you have no jffs2
05:23.19[g2]at least it's sized right :)
05:23.27jacquesthe unallocated space should mount right up tho ?
05:24.14jacquesif it's just a bunch of 0xffff it should be mountable as jffs2
05:24.42[g2]./slugimage -p -o unslung-3.0.img Ramdisk:unslung-switchbox-nslu2-20041116041613.rootfs.ext2.gz,Flashdisk:unslung-able-nslu2-20041116041613.rootfs.jffs2  
05:24.44jacques[g2], try mount -t jffs2 /dev/mtdblock4 /mnt/whatever
05:24.58rwhitby-webif it doesn't have /bin/init in it, then linuxrc will not mount it
05:25.04dyoung-webit kinda works if I ignore the bummers and pivot.
05:25.14dyoung-webbin or sbin?
05:25.15[g2]I've got the wrong jffs2 Flashdisk
05:25.25jacqueswhy did slugimage not even make a Flashdisk partition ?
05:25.41jacqueswhere's the -r  ???
05:25.49Christopherhmmm, wish I could work on this stuff but I have to work on my math homework
05:25.50rwhitby-webdyoung-web: linuxrc says /bin/init
05:25.52jacquesI don;t think it will parse that right rithough -r
05:26.06dyoung-web~praise Christopher
05:26.08jbotAll hail Christopher!
05:26.30dyoung-webupslug makes it possible for me to beat [g2] to the #2 spot remotely!
05:26.38[g2]-r is helpful
05:26.59rwhitby-webstill no #2 yet - you guys take longer to follow that path than I did to trailblaze it !
05:27.17jacquesI'm not going forward til we get this doubled device names thing taken care of
05:27.42dyoung-webI dunno why it dont mount.
05:28.02rwhitby-webjacques: remove the  EXTRA_IMAGECMD_ext2.gz = "-f ${UNSLUNG_DEVICE_TABLE}" line from unslung-switchbox-image
05:28.26rwhitby-webor remove the makedevs call
05:28.49rwhitby-webthat's why it's doubling up.  genext2fs will just duplicate them without checking probably
05:30.47rwhitby-webI keep forgetting that jacques is doing this on prodslug, so he needs to be very careful
05:32.04[g2]good point.
05:32.21[g2]~praise jacques
05:32.23jbotAll hail jacques!
05:32.30jacqueswell both of mine are pretty much devslugs :-)
05:33.40jacquesOK rwhitby-web that worked (I commented out the makedevs)
05:33.51jacquesnow to assemble an image
05:36.49[g2]I think dyoung-web and I are in the same space
05:37.12[g2]I've got the bummer again, but I think it mounted mtdblock4
05:37.26[g2]+ /bin/mount -rt jffs2 /dev/mtdblock4 /mnt/tmpmnt
05:37.26[g2]+ [ -x /mnt/tmpmnt/bin/init ]
05:37.26[g2]+ [ -d /mnt/tmpmnt/unslung ]
05:37.26[g2]+ /bin/umount /mnt/tmpmnt
05:37.26[g2]+ [ -z  ]
05:37.27[g2]+ [ -x /mnt/tmpmnt/bin/init ]
05:38.19[g2]which is followed up with
05:38.41[g2]+ /bin/umount /proc
05:38.41[g2]+ [ -z  ]
05:38.41[g2]+ /bin/echo Root filesystem cannot be found - rebooting ...
05:38.59[g2]Root filesystem cannot be found - rebooting ...
05:38.59[g2]+ /sbin/reboot
05:38.59[g2]+ exit 0
05:41.38*** join/#nslu2-linux CIA-10 (
05:45.31Christopherdyoung-web: glad to see that you like upslug
05:46.13dyoung-webdammit jim
05:46.13dyoung-webthis bummer is really bumming me out
05:46.17[g2]dyoung-web, does that mean I can upslug your device remotely too :)
05:46.39dyoung-webI cant kill it either.  I need slingbox with no init.
05:46.55Christopherdyoung, does upslug detect both of your slugs properly?
05:46.58dyoung-webto make a long story longer, I edited linuxrc and pivoted to the jffs2.
05:47.46kergothjacques: i pushed the uclibc and gcc changes.
05:47.52kergothyou should be able to do a build now
05:48.05dyoung-webthat gives
05:48.09dyoung-web/dev/mtdblock4            5760      5020       740  87% /                      
05:48.16rwhitby-web[g2]: if your jffs2 stanza is checking for unslung, then it is not up to date
05:48.31*** join/#nslu2-linux KsiLaptop (
05:48.44rwhitby-web[16:07] <[g2]>   [ -d /mnt/tmpmnt/unslung ]
05:49.09rwhitby-webthat line shows you've got an old /linuxrc - you need to rebuild switchbox and unslung-switchbox-image
05:49.19dyoung-web[g2], you probably wont be able to get yourself into my ethernet space.
05:49.40jacquesI'm booted to the jffs2  :-D
05:49.43dyoung-webBut since Tiersten 0WnZ my redboot.
05:49.51jacquesFilesystem           1k-blocks      Used Available Use% Mounted on
05:49.51jacquesrootfs                    5760      5024       736  87% /
05:49.51jacques/dev/root                 5760      5024       736  87% /
05:49.55rwhitby-webjacques is #2 ?
05:50.00jacqueswhat number does that make me?
05:50.03dyoung-webyou could work with him to 0wn2 my slug and run upslug on it.
05:50.29dyoung-webjacques, if you got it without doing something ugly, youre #2.
05:50.41rwhitby-web[g2] and dyoung were too busy 0wning each other's slugs, and jacques powered through on the inside ...
05:50.44jacquesI did nothing ugly - it booted to the jffs2 :-)
05:51.14dyoung-webI got the thing mounted as root.  But I had to do sometihng Very Ugly.
05:51.15[g2]jacques, can you push that patch for the no mknod's
05:51.38jacques[g2], yeah I will do it now
05:51.51[g2]no win, place or show for poor [g2]
05:52.01rwhitby-webwell, I'm happy it has been replicated now :-)
05:52.22[g2]oh you're just happy you were first
05:53.03Christopherso, what has to happen before I can build an image with jffs2 in it?
05:53.14rwhitby-web[g2]: yeah, but I still want my CSR-enabled OpenSlug in OE birthday present :-)
05:53.38[g2]I'm sure it'll happen for you're next birthday :)
05:53.47rwhitby-webChristopher: we just need slugimage in OE (which just about anyone here can do except me at the moment)
05:54.18[g2]Christopher, and jacques needs to push his changes
05:54.26rwhitby-webworst case, you'll be able to build an 8mb image in 5 hours
05:54.30[g2]for the dup dev devices
05:54.38rwhitby-webdup shmoop
05:54.54CIA-1003jacques 07 * r1.2558 10packages/meta/unslung-switchbox-image.oe:
05:54.54CIA-10added nslu2-linksys-libs to UNSLUNG_BASE_PACKAGES
05:54.54CIA-10removed makedevs
05:55.02Christopher~praise jacques
05:55.07jbotAll hail jacques!
05:55.14[g2]rwhitby-web, you know anyu2 builds CSR-enabled 2.6.9 out-of-the-box right
05:55.34[g2]and jacques had jffs2 running a week before you started playing with it
05:55.59rwhitby-web[g2]: I said jacques could have the #1 spot if he wanted it ...
05:56.22[g2]no, I just meant the openslug/CSR issue
05:56.31jacques#2 is fine with me - I really wanted lucky #4 but it just worked out this way  :-)
05:56.43[g2]me too
05:56.55dyoung-webI wanted #3.
05:57.00[g2]that's why I waited, wink, wink
05:57.29rwhitby-web[g2]: I will not be party to continuing the fragmentation - I won't use OpenSlug/CSR until it's in OE.  I thank kas11 for trail-blazing, but I'll wait for the highwaymen on that one.
05:57.30dyoung-webI thought jcques should be #1 or #2.  Since he did a lot of the heavy lifting.
05:58.07jacques#2 is fine with me  :-D I'll add myself now
05:58.24Christopherso, anyone volunteer to add slugimage to oe?
05:58.26dyoung-webverifying... moment of truth...
05:59.20jacquesOK I added myself - it feels good :-)
05:59.56rwhitby-webjacques: note that the unslung-able-image will still have duplicate devs in the ext2.gz image - a jffs2 image needs makedev run
06:00.37jacquesmkfs.jffs2 takes a device_table.txt arg you know
06:00.49rwhitby-webthat I didn't know
06:01.20kergothoe passes it by default, using the minimal table .txt
06:01.38rwhitby-webcool - let's use that for jffs2 as well then in unslung-standard-image
06:01.48rwhitby-webpassing in the device_table-unslung.txt file
06:01.55rwhitby-webfrom packages/files/...
06:05.40Christopherwhere can one download slugimage?
06:05.51[g2]sf cvs
06:05.57jacquesChristopher, from fs cvs module slugimage
06:06.05jacquessf cvs that is
06:06.35[g2]how long does it take for the web interface on sf to update ?
06:06.46[g2]it's just one perl file
06:06.58Christophermy upslug file took a couple of hours
06:08.39jacqueswhat's  ?
06:08.41rwhitby-webup to six hours, sometimes even longer apparently
06:09.15rwhitby-webjacques: something which should have been removed by switchbox_clean_image ... don't know why it wasn't
06:10.03rwhitby-webwe should remove line 29 from unslung-switchbox-image.oe
06:10.38rwhitby-webthat's something different, not related to libntpass
06:11.04jacquesyou want to remove this:  echo "unslung-switchbox" > ${IMAGE_ROOTFS}/.unlsung   ?
06:11.22rwhitby-webyeah.  I think we'll end up having separate unslung-switchbox-image and openslug-switchbox-image files anyway
06:11.42jacquesOK I will remove it
06:12.12rwhitby-weband switchbox_clean_image will go away altogether when we get the <128Kb switchbox
06:16.23jacquesI wonder how much space bash ipkg takes
06:17.57dyoung-webI give up
06:18.18dyoung-webi'll try again when I'm not hungry.
06:18.54jacquesrwhitby-web, is switchbox_clean_image supposed to remove things from the jffs2 image ? I wouldn't think so
06:21.06[g2]time for some sleep
06:21.18jacques`night [g2]
06:21.24[g2]the last changes make the slug continually reboot
06:21.35[g2]one messages was bad
06:21.52[g2]jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00000204: 0x656c instead
06:21.52[g2]Further such events for this erase block will not be printed
06:21.52[g2]JFFS2: Erase block at 0x00000000 is not formatted. It will be erased
06:21.52[g2]Cowardly refusing to erase blocks on filesystem with no valid JFFS2 nodes
06:22.42[g2]there were some more eraseblock before that
06:22.54jacqueslooks like maybe not a jffs2 image
06:22.56[g2]~praise TheCoward
06:22.58jbotAll hail TheCoward!
06:23.48rwhitby-webjacques: no switchbox_clean_image only works on the switchbox ext2.gz image
06:24.53rwhitby-weboh - were you looking at libntpass in the jffs2 root?
06:25.38[g2]ah... my bad there was a SPACE before the , on the cmd line
06:26.12jacquesrwhitby-web, yeah I was just looking at all the big files in the jffs2 root trying to figure out what I can delete :-D
06:28.12rwhitby-web[g2]: the sooner we get OE to type that command for you, the better :-)
06:28.50rwhitby-webjacques: we need an unslung-make-it-almost-openslug ipkg which installs nothing, but does a whole heap of removes in the postinst :-)
06:28.54[g2]rwhitby-web, it's late, I should have been sleeping hours ago, but it's been just one more build.....
06:29.29Christopherrwhitby-web: good idea
06:29.42rwhitby-webyou just want to beat Christopher to #3 :-)
06:30.03[g2]I thought dyoung was #3
06:30.12rwhitby-webnot yet - it's not in the database
06:30.14jacquesrwhitby-web, I think that would be quite useful - I would defnitely play with it
06:30.29jacques<dyoung-web> I give up
06:30.34jacquessounds like he didn't quite make it yet
06:31.18rwhitby-webyeah, you need to be able to boot, and run ipkg update, with no other commands inbetween :-)
06:31.58rwhitby-webthat's the litmus test for whether you are truly diskless unslung from jffs2 or not ...
06:32.41[g2]it's better, but not quite rite
06:32.46[g2]# df
06:32.46[g2]Filesystem           1k-blocks      Used Available Use% Mounted on
06:32.46[g2]rootfs                    4992      4992         0 100% /
06:32.46[g2]/dev/root                 4992      4992         0 100% /
06:33.09[g2]init: /sbin/syslogd respawning too fast
06:33.20jacquescat /proc/mounts|grep root
06:33.47[g2]# cat /proc/mounts|grep root
06:33.47[g2]rootfs / rootfs rw 0 0
06:33.47[g2]/dev/root / jffs2 rw 0 0
06:34.18jacquesthat's weird
06:34.29rwhitby-web[g2]: you still have a full jffs2
06:34.31jacquesyou have a full jffs2
06:34.37rwhitby-webdid slugimage report any padding?
06:35.00[g2]Wrote 39 blocks (0x00300000 to 0x007E0000) from <Flashdisk> into "unslung-3.0.img"
06:35.00[g2]Wrote 1 blocks (0x007E0000 to 0x00800000) from <FIS directory> into "unslung-3.0.img"
06:35.05jacqueswanna see my slugimage line? I'll show it anyway
06:35.09jacques./slugimage -d -p -r Ramdisk:unslung-switchbox-nslu2-20041116063005.rootfs.ext2.gz,Rootdisk:unslung-able-nslu2-20041116045113.rootfs.jffs2 -o able6
06:35.23jacquesWrote 45 blocks (0x00240000 to 0x007E0000) from <Rootdisk> into "able6"
06:35.25[g2]what's the -d
06:35.42jacquesyou got 39 blocks I got 45
06:35.46rwhitby-web[it shouldn't] - it should read the jffs2 file (and report it as N blocks) and then allocate it N X blocks in the table.
06:36.16rwhitby-web[that's N plus X - CGI::IRC doesn't pass plus signs]
06:36.40rwhitby-webin our case, N is 39 and N plus X is 45
06:37.17[g2]the -p for pad right
06:37.33[g2]boy it is late :)
06:37.47kergoth[g2]: heh, where you at?
06:37.53[g2]./slugimage -p -o unslung-3.0.img -r Ramdisk:unslung-switchbox-nslu2-20041116055907.rootfs.ext2.gz,Flashdisk:unslung-able-nslu2-20041116055907.rootfs.jffs2
06:38.21kergothbah, its not late :)
06:38.30rwhitby-web[g2]: grep "fill the rest of the space" slugimage
06:38.31[g2]it's early, early
06:39.06[g2]grep "fill the rest of the space" slugimage
06:39.18rwhitby-webok, you have the latest slugimage
06:39.51jacques[g2], how many blocks did it write for Ramdisk ?
06:39.57rwhitby-webignore kergoth, and paste in your whole slugimage stdout :-)
06:40.10jacquesWrote 7 blocks (0x00160000 to 0x00240000) from <Ramdisk> into "able6"
06:40.16jacquesjust Ramdisk might be enough
06:40.36rwhitby-weblet's not play 20 questions  - paste the whole thing
06:40.38[g2]ah I think it's the ramdisk size
06:40.53jacques20 questions is more fun/challenging
06:41.00[g2]./slugimage -p -o unslung-3.0.img -r Ramdisk:unslung-switchbox-nslu2-20041116055907.rootfs.ext2.gz,Flashdisk:unslung-able-nslu2-20041116055907.rootfs.jffs2
06:41.00[g2]Read 0x00040000 bytes from "Redboot" into <RedBoot> (2 blocks)
06:41.00[g2]Read 0x00020000 bytes from "SysConf" into <SysConf> (1 blocks)
06:41.00[g2]Read 0x000D0104 bytes from "vmlinuz" into <Kernel> (7 blocks)
06:41.00[g2]Read 0x0019CCEB bytes from "unslung-switchbox-nslu2-20041116055907.rootfs.ext2.gz" into <Ramdisk> (13 blocks)
06:41.07[g2]Read 0x00460000 bytes from "unslung-able-nslu2-20041116055907.rootfs.jffs2" into <Flashdisk> (35 blocks)
06:41.10[g2]Read 0x00000010 bytes from "Trailer" into <Trailer> (1 blocks)
06:41.16[g2]Allocated <RedBoot> from 0x50000000 to 0x50040000 (2 blocks)
06:41.16[g2]Allocated <SysConf> from 0x50040000 to 0x50060000 (1 blocks)
06:41.16[g2]Allocated <Kernel> from 0x50060000 to 0x50160000 (8 blocks)
06:41.18[g2]Allocated <Ramdisk> from 0x50160000 to 0x50300000 (13 blocks)
06:41.20[g2]Allocated <Flashdisk> from 0x50300000 to 0x507E0000 (39 blocks)
06:41.22[g2]Allocated <FIS directory> from 0x507E0000 to 0x50800000 (1 blocks)
06:41.36[g2]Wrote 2 blocks (0x00000000 to 0x00040000) from <RedBoot> into "unslung-3.0.img"
06:41.36[g2]Wrote 1 blocks (0x00040000 to 0x00060000) from <SysConf> into "unslung-3.0.img"
06:41.36[g2]Wrote 8 blocks (0x00060000 to 0x00160000) from <Kernel> into "unslung-3.0.img"
06:41.36[g2]Wrote 13 blocks (0x00160000 to 0x00300000) from <Ramdisk> into "unslung-3.0.img"
06:41.36[g2]Wrote 39 blocks (0x00300000 to 0x007E0000) from <Flashdisk> into "unslung-3.0.img"
06:41.37[g2]Wrote 1 blocks (0x007E0000 to 0x00800000) from <FIS directory> into "unslung-3.0.img"
06:41.39[g2]Rewound 16 bytes before <Trailer> in "unslung-3.0.img"
06:41.41[g2]Wrote 16 bytes (0x007FFFF0 to 0x00800000) from <Trailer> into "unslung-3.0.img"
06:41.51[g2]that was 3 chunks
06:41.53jacqueswhy is your ramdisk so large?
06:42.38jacques872529 Nov 15 22:35 unslung-switchbox-nslu2-20041116063005.rootfs.ext2.gz
06:42.50[g2]shouldn't it be using the switchbox, right
06:43.08[g2]-rw-r--r--  1 tom users 1690859 Nov 16 01:07 unslung-switchbox-nslu2-20041116055907.rootfs.ext2.gz
06:43.45[g2]that was 3 pastes
06:43.50rwhitby-web-rw-r--r--  1 rwhitby rwhitby 879430 Nov 15 21:06 ramdisk.gz
06:44.17[g2]where did you get your ramdisk from ?
06:44.27rwhitby-webI get 7 blocks for ramdisk., 35 (padded to 45) for flashdisk
06:44.38kergothjacques: okay, wiped tmp, and rebuilt the toolchain from scratch with USE_NLS=no.  confirmed it worked fine and uclibc was indeed built with no locale support.
06:44.57jacqueskergoth fantabulous!
06:45.09jacquesis it in is it in??
06:45.23kergothyep, its all checked in
06:45.39rwhitby-web-rw-r--r--  1 rwhitby rwhitby 883641 Nov 16 08:51 /home/openembedded/unslung/tmp/deploy/images/unslung-switchbox-nslu2-200411152117\
06:46.54jacqueskergoth COOL
06:47.00[g2]yes the common theme is that your switchboxes are old
06:47.07[g2]and smaller
06:47.34[g2]I did an oemake -c clean switchbox and unslung-switchbox-image
06:47.45[g2]then oemake unslung-image
06:48.00jacqueswell I can start a clean build
06:48.23[g2]it's getting close
06:48.33[g2]very close
06:49.01rwhitby-webmy switchbox was built before the pushes after I left for work this morning, so there may be a mismatch
06:49.35[g2]well I'll pull and try again in the later morning
06:49.41[g2]thx for all your help
06:49.42rwhitby-webI will do a full rm -rf tmp build tonight
06:50.09rwhitby-webnight [g2]
06:50.18jacques`night [g2]
06:50.21[g2]actually, we should be better about those PR values and have stuff automatically built
06:50.58[g2]unless there's a problem with big values like nnn
06:51.45jacqueskergoth - where is packages-devel ? I tried bk://  but no go
06:51.47[g2]the make clean is nice every now and then, but we should be able to oemake -c clean or just oemake and have thing build right
06:52.00[g2]sweet dreams all
06:52.09kergothjacques: its an OE repo, not nslu2.
06:52.15jacquessorry I just figured that out
06:52.26*** part/#nslu2-linux [g2] (
06:52.27kergothjacques: all development happens in packages-devel now, because users doing clones kept 'packages' locked all the time.  slowed us down
06:53.30jacquesyeah this is much better - bk://
06:53.59jacqueskergoth that makes sense - you gotta do that when you start having users
06:58.49jacquesstarting a uclibc USE_NLS=no build
07:05.08Christopherrwhitby-web: I think I added slugimage to oe, but I don't have push access
07:05.35rwhitby-webbk send the changeset to
07:06.00Christopheri have to check it in, then I'll do that
07:06.01rwhitby-weband send me your rsa1 key
07:06.33jacquesyeah Christopher why haven't you sent your key yet :-)
07:06.51rwhitby-webmake sure it's rsa[one]
07:08.18jacquesim hungry for some reason - biab
07:30.54jacqueshmm i think i screwed something up in unslung-image
07:31.01jacquesbuild failed
07:31.46jacques| Cannot find package nslu2-linksys-libs.
07:36.05rwhitby-webback later
07:36.08*** part/#nslu2-linux rwhitby-web (
07:56.35Christophernight all
08:00.10jacques`night Christopher_zzz
08:01.25dyoung-zzzzDid I beat [g2]?
08:02.09dyoungrootfs                    5760      5024       736  87% /                      
08:02.21dyoungas jffs2
08:03.29*** join/#nslu2-linux KsiLaptop (
08:03.31dyoungipkg update grumbles about wget though.
08:06.58jacquesdyoung yeah looks like you are #3  :-)
08:08.39jacques124596 Nov 16 00:00 busybox*
08:10.36rwhitby-awayjacques: how?
08:14.37rwhitby-awaypackages is now reparented to oe:/packages-devel
08:14.41rwhitby-awayand pulled
08:16.08jacquesrwhitby-away, set static in switchbox defconfig, uncommented static in switchbox .oe, used dist linux-uclibc USE_NLS=no build in local.conf
08:19.12jacquesnot sure what to make of this:
08:19.15jacques-rw-r--r--  1 fontenot users 81920 Nov 16 00:17 unslung-switchbox-nslu2-20041116081414.rootfs.cramfs
08:19.15jacques-rw-r--r--  1 fontenot users 78871 Nov 16 00:17 unslung-switchbox-nslu2-20041116081414.rootfs.ext2.gz
08:19.35jacquesi expected cramfs to be smaller
08:20.27kergothgzip compresses pretty well.
08:20.47rwhitby-awayso you're saying we've got lots of room in switchbox ?
08:20.51kergothbut since cramfs doesnt have to be decompressed, thats still likely a gain
08:21.03kergothbleh, i should sleep
08:21.08jacquesrwhitby-away, I guess I am saying something like that :-)
08:21.29jacquessizes are within expectations as busybox.gz is around 70kB
08:22.00rwhitby-awayso we can probably put some basic utilities in switchbox, like wget and mtd_debug :-)
08:22.31jacquesif you mean the bb wget then probably
08:22.35rwhitby-awayboot into switchbox, wget a new jffs2 image, mtd_debug write it, reboot
08:22.52jacqueswhy need mtd_debug to write it?
08:22.55jacquesjust dd it
08:23.03rwhitby-awayoh - of course :-)
08:23.04kergoth`zzzdd to flash is -bad-
08:23.12kergoth`zzzdd isnt tolerant of bad blocks.
08:23.25rwhitby-awayeven with a mtd device driver ?
08:23.26jacquesI've never had a problem with it
08:23.41jacquesif there's a bad block what do you want it to do? keep going?
08:23.58kergoth`zzzbad blocks are marked, so operations can skip that block.
08:24.09rwhitby-awayjacques: let's add dd, ls and wget to switchbox
08:24.11kergoth`zzzthey dont happen often, but on the off chance you have one, you want to be able to handle it
08:24.23kergoth`zzzdd doesnt know to skip it
08:24.23rwhitby-awaydoesn't the mtd driver handle that?
08:24.26kergoth`zzz(yes, even with mtd)
08:24.32rwhitby-awayoh, ok
08:24.39rwhitby-awayback later
08:24.57jacques`night kergoth`zzz
08:25.09jacquesthanks for doing the NLS thing!!!
08:27.26jacquesthe cramfs doesn't have the devices in it :-\   looks like everything else is there tho
08:30.43dyoungupslug would probably fit.
08:31.36dyoungls.  ls is so helpful.
08:31.42dyoungit sure beats echo *
08:36.55jacquesrwhitby-away, above I should have said: TARGET_OS = "linux-uclibc"
08:58.46rwhitby-awayhmm - I guess what we can do is build switchbox, and upload it to - then the image creation can get it from there by default.
08:59.08rwhitby-awayI'm starting an rm-rf tmp build
09:00.33rwhitby-awayI'll create a switchbox.conf for us to use as dist, which has all those things in it
09:01.17rwhitby-awayuntil oe-ng arrives which multi-toolchain capability
09:03.10jacquesI'm building images and running tests now
09:03.18jacquesgive me a few minutes and I might have some stuff to checkin
09:03.22jacquespush taht is
09:05.13jacques106496 Nov 16 01:03 unslung-switchbox-nslu2-20041116090134.rootfs.cramfs
09:05.26jacquesthat includes devices and featureful ls, dd and wget
09:11.04dyoungI'm so giddy about this jffs2 I dont know what to do with myself.
09:11.21jacquescrap crap
09:11.34dyoungWhats happened?
09:11.45dyoung(or NOT happened as the case may be...)
09:12.06jacquescrap crap carp a caprapcpas
09:12.48jacquesresolve: 1 unresolved conflicts, nothing is applied.
09:12.54jacquesi cant push
09:13.06rwhitby-awaypull first
09:13.31jacquesthe push doesnt work either
09:13.42jacquesConflicts during automerge of meta/unslung-switchbox-image.oe
09:13.42jacquesresolve: resolved 1 conflicts in pass 3
09:13.42jacquesresolve: 1 unresolved conflicts, nothing is applied.
09:13.47rwhitby-awaybk resolve
09:14.10jacquesthen what?
09:14.16jacques(content conflict) meta/unslung-switchbox-image.oe>>
09:14.24jacquesthat's really intuitive as to what to do next
09:14.34rwhitby-awayhit ?
09:15.19rwhitby-awayunslung-switchbox-image.oe is going to be renamed to switchbox-image.oe (part of the new switchbox distro)
09:16.35rwhitby-awayjacques: what IMAGE_ROOTFS_SIZE_ext2.gz do we use now ?
09:18.04jacqueswhy do we care?
09:18.18rwhitby-awayfor switchbox-image
09:18.27jacquesCheck FAILED, contact BitMover.
09:18.53jacquesrwhitby-away, are we going to be using ext2 ?
09:19.06rwhitby-awayfor switchbox, yes?
09:20.00rwhitby-awayhas that been proven yet?
09:20.16jacquescan you look at the changeset I send to the list?
09:20.24jacquesbecause this repos is now totally fscked
09:22.10jacquesgoing to \rm -rf it
09:23.55jacquescrap my hard drive is failing - that's why the errors
09:24.16jacquesstupid POS maxtor
09:28.08rwhitby-awayapplied your patch by hand
09:28.45jacquesrwhitby-away, thanks I didn't want it to get lost
09:32.00jacquesrunning reiserfsck now
09:32.15dyoungI think I also bk recieved it properly...
09:32.51jacquesi hope it doesn't cause problems for others since there is a conflict
09:32.53rwhitby-awayhmm - wonder what that will do when I've applied it by hand and will be pushing that
09:33.21jacquesthe conflict is only in unslung-switchbox-image.oe
09:33.55jacquesand the new version is what should be used
09:34.33rwhitby-awayI have renamed that anyway
09:34.47rwhitby-awaythere is now a new "switchbox" distro
09:35.23jacquesyeah that sounds like the right way to do it
09:35.38jacquesman reiserfsck takes a long time
09:36.08rwhitby-awaythen a few people can work on switchbox, and release a binary to be used in the image creation of unslung or openslug
09:37.07rwhitby-awayok, pushing the untested changes - take care
09:37.11CIA-1003rwhitby 07 * r1.2561 10packages/ (busybox/switchbox-1.00/defconfig meta/switchbox-image.oe): Merge
09:37.27CIA-1003rwhitby 07 * r1.2553.1.1 10packages/ (5 files in 4 dirs): Created switchbox distro
09:38.39rwhitby-awayshould be able to copy your local.conf, and change distro to switchbox
09:38.49rwhitby-awaythen oemake switchbox-image
09:39.50rwhitby-awaydyoung: you might want to bk undo jacques's patch, and then pull
09:40.06dyoungI just did.
09:40.12rwhitby-awayok, I'd doing two rm -rf builds now, so that will take a while before I get results
09:40.21rwhitby-awayremember, this stuff is not tested
09:40.50rwhitby-awaydid anyone see Christopher's changeset on the list ?
09:41.02dyoungI've started a switchbox-image build
09:41.49jacquesrwhitby-away, did you put USE_NLS=no in the distro file ?
09:43.19rwhitby-awayI copied from openslug, then added USE_NLS = "no" and removed all the kernel stuff which won't be used, and removed nslu2_flash_img stuff
09:43.58jacquesreiserfsck is fscking useless
09:44.20rwhitby-awaySo no changeset email from Christopher?
09:44.28jacqueshow do you write to a specific clock on a drive?
09:44.38rwhitby-awaythat's unfortunate, cause it means I will have to redo his work tonight :-(
09:45.22jacqueswhy does it give advice like that when there's no way to use it?
09:46.30rwhitby-awayno pending requests on nslu-changesets ...
09:46.43dyoungThis is ridiculous.  I'm going to get a faster CPU when I get back next month.
09:47.05dyoungThis 2Ghz Celery is .. uh... Slow.
09:52.27jacquesno CR at end of switchbox.conf
09:54.31jacquesrwhitby-away, I think we will get smaller binaries with hard float
09:55.39rwhitby-awayit has to match the kernel we've built, doesn't it?  Let's not change things like that mid-stream ...
09:56.25rwhitby-awaygoing from ext2.gz to cramfs in switchbox tonight is making me nervous enough already ,,,,
09:56.25jacquesI thought kernel had NWFPE
09:56.37jacqueswell go with ext2 then
09:56.55rwhitby-awayno, I can live with that much uncertainty :-)
09:57.07dyoungMade a new build directory for switchbox dev.
09:57.23rwhitby-awaycause that will either work or not.  whereas hard/soft sounds to me like it could have subtle errors later
09:57.38jacqueswe are only talking about switchbox right?
09:57.45rwhitby-awaybut I no nothing about hard/soft, so I'm not speaking with any authority
09:57.47dyoungisnt it going to be kind of binary for switchbox?
09:58.06dyoung(ie: its going to work, or its not) ?
09:58.13rwhitby-awayone binary thing at a time, please
09:58.41jacquesi forgot ext2.gz is smaller
09:58.48jacquesso maybe we should use that
09:58.52rwhitby-awaywhen I have OE building a 8mb image that boots, then we can talk about all these other things
09:58.56jacquesit's just such a pain to keep messing with the size
09:59.04dyoungWell, its probably going to take until tommrow to populate the tooling for switchbox dev anyways.
09:59.38jacquesI guess we don;t care how big the ext2 is - only the compressed size right?
09:59.46rwhitby-awayyou guys feel free to experiment, but please don't push any toolchain changes
10:00.27rwhitby-awayI like cramfs better anyway, cause then we can mount it from inside a jffs2 root to work on it
10:00.44rwhitby-away[read/copy, not write of course]
10:01.28rwhitby-awaynot that I can think of a reason to do that ....
10:01.59rwhitby-awaycramfs is better cause then we don't get images confused.  ext2.gz is stock ramdisk, cramfs is switchbox, jffs2 is rootfs
10:02.03jacqueswell any reason to not have it emit both image types ? I had that in and it worked fine
10:02.09dyoungyou can kinda do that with the ext2 too... mtd_debug read ... | gzip -dc > foo; mount -o loop foo /bar
10:02.22dyoungsince we have loop in there now too.
10:02.37rwhitby-awayfor the next 12 hours, it's cramfs
10:04.20jacquesrwhitby-away, how's your clean slitchbox-image build going?  I got a strange error on one of mine but that wasn't with a current packages
10:04.45jacquesERROR: log data follows (/home/packages/oe/nslu2/build_switchbox/tmp/work/unslung-switchbox-image-1.0-r1/temp/log.do_rootfs.3712)
10:04.52jacques| touch: cannot touch `/home/packages/oe/nslu2/build_switchbox/tmp/deploy/ipk/Packages': No such file or directory
10:12.24rwhitbymy builds haven't got that far yet - they are both still on gcc-cross-initial
10:12.40jacquesok I just started two clean ones on two different boxes
10:12.55jacquesthat error was right at the end and looks maybe easy to fix anyway
10:14.42rwhitbyI have to make some dinner - that's why I started clean builds
10:15.44jacquesok have a good dinner
10:16.21rwhitbyit's just some pasta - no big deal :-)
10:30.30dyoungI wonder how painful imagemagick tools would be on a Slug.
10:32.50jacquesdepends on which ones I guess
10:33.48dyoungconvert mostly.
10:34.28dyoungIt'll prolly run outta memory.
10:35.11jacquesi was thinking about that but it would probably depend on the data size
10:37.36rwhitby-rw-r--r--  1 rwhitby rwhitby 102400 Nov 16 21:08 switchbox-nslu2-20041116093947.rootfs.cramfs
10:37.43jacques-rw-r--r--  1 fontenot users 102400 Nov 16 03:36 switchbox-nslu2-20041116110222.rootfs.cramfs
10:38.17rwhitbyless than 128Kb :-)
10:39.07dyoungWait me me guys!
10:39.16dyoungmachiune is slow!
10:39.50jacquesrwhitby, I just mounted it on my pc - it looks good - devices are there, linuxrc, 153648 byte switchbox
10:39.54rwhitbyI'm peeved that Christopher didn't send that changeset ..
10:40.19rwhitbybut I'm guessing he had a good reason - must not have worked or something
10:40.32dyoungOr math homework got the best of him.
10:41.02rwhitbyhe sent me his key first - I wish he had sent the changeset first
10:43.55rwhitbythat 102400 is such a round number ...
10:47.25dyoungfuriously building!
10:48.07dyoungthats cheating though.
10:49.08jacqueswhat's cheating ?
10:49.24dyoungDownloading the cramfs switchbox that rwhitby built instead of building it myself.
10:49.35rwhitbydyoung: that's not for you, it's for slugimage
10:53.39*** join/#nslu2-linux Ksilebo|WARK (
10:54.29dyoungwhat the hell, I d/l and made an image, and upslugging it now
10:55.37rwhitbycheater :-)
10:56.27jacquesargh - no loop devices in able kernel
10:56.55jacqueshmm no strike that
10:57.36dyoungChristopher, I dont suggest upslugging with the slug running a jffs2.  It doesnt work. ;-)
10:58.10rwhitbyhmm - that cramfs switchbox won't boot
10:58.50jacqueswhat's the error?
10:59.02rwhitbyRAMDISK: Couldn't find valid RAM disk image starting at 0.  
10:59.22jacquesyeah it's not a ramdisk
10:59.37dyoungGotta add cramfs to linuxrc?
10:59.39jacquesmaybe it needs rootfstype=cramfs
10:59.57jacquesthe static switchbox binary runs :-)
11:00.22jacques/dev/loop0                 184       184         0 100% /mnt/cramfs
11:00.35jacques(on the slug)
11:00.56dyoungOh, I see the problem.
11:01.02dyoungits rather immediate isnt it.
11:01.31jacquessomething just occurred to me
11:02.22jacqueshow do we boot to a non-ramfs fs that's in the first ramfs block
11:03.00jacqueslooks like we gotta use ext2.gz ...  (unless I'm wrong about cramfs not being a ramfs)
11:03.11jacquesyeah that's tried and true
11:03.18dyoungand smaller.
11:03.28jacquesyeah that too
11:05.01KsiLaptopOh man this is awesome...
11:07.10rwhitbywhat size do you think for ext2.gz ?  256 ?
11:07.32dyoungwasnt it a bit smaller?
11:07.33jacquesrwhitby, that should be plenty for now
11:07.51rwhitbylets try 512
11:07.58dyoung512 should be luxury.
11:07.59jacqueslooks like there is about 184kB of data
11:08.24rwhitbyI have no idea how genext2fs uses that number
11:08.58jacquesit *should* make a fs with that total size
11:09.16rwhitbyyeah, but it never does ....
11:09.38jacquesi guess as long as the compressed size is < 128kB-16B we don;t care too much ?
11:10.00rwhitby512 didn't work either - removing the size from switchbox.conf to let the nslu2 default of 10MB rule
11:10.22jacquesdidn't work in what sense?
11:11.03rwhitbythe old genext2fs: couldn't allocate an inode (no free inode)
11:11.39jacquesah, you might have to explicitly allocate inodes
11:11.55jacquesthe auto-sized number fopr a 512kB fs must be like 10
11:12.02rwhitby-rw-r--r--  1 rwhitby rwhitby 107263 Nov 16 21:42 switchbox-nslu2-20041116111116.rootfs.ext2.gz
11:12.26jacquesit doesn't expect to have such a small average file size (most files are devices or symlinks)
11:12.35jacquesor directories
11:13.03jacquesrwhitby, you trying to boot to that ?
11:13.25rwhitbyjust about to
11:17.50rwhitbyFilesystem           1k-blocks      Used Available Use% Mounted on              
11:17.51rwhitbyrootfs                    6528      5024      1504  77% /                  
11:18.04jacqueslooking goooood babyyy!
11:18.15dyoungPost um.
11:19.44CIA-1003rwhitby 07 * r1.2562 10packages/conf/distro/switchbox.conf: Changed switchbox back to ext2.gz
11:22.05rwhitbyok, so how to we add 20k (compressed) more stuff which allows us to get a prompt if there is no jffs2 partition ...
11:23.04jacqueswhat we missing?
11:23.06rwhitby[I'm going to leave the slugimage package for Christopher to do]
11:23.07jacqueswe have ash
11:23.45rwhitbyso we just change /sbin/reboot to /bin/sh in linuxrc?
11:24.09jacquesyeah I think it's that easy or almost that easy
11:24.20jacquesprobably exactly that
11:25.29rwhitbyheh - the other good thing about changing unslung-switchbox-image to just switchbox-image is that it makes it easy to rm both switchbox and switchbox-image :-)
11:29.32dyoung"Do not flash this image if you are not serial enabled" ?
11:30.08rwhitbywhere's that ?
11:30.53dyoungIf you change the reboot to sh.
11:31.03rwhitbyah - right
11:31.19rwhitbyI'm actually going to flash an image now that intentionally will not boot :-)
11:32.15rwhitby./slugimage -p -r Ramdisk:ramdisk.gz,Flashdisk:0 -o empty.bin
11:36.13rwhitbygot a prompt :-)
11:36.22jacquesnice :-)
11:36.46jacquesso what else do we want in our 128kB?
11:37.06rwhitbya reboot that works :-)
11:37.23rwhitbyKernel panic: Attempted to kill init!                                            
11:39.08rwhitbytftp client :-)
11:39.47rwhitbythere must be a kernel thingo to reboot on panic, right?
11:41.07jacquesive seen that behavior before - Kernel panic: Attempted to kill init!  - I think it's fixable... somehow
11:41.30rwhitbythat's what happened when I ran /sbin/reboot
11:41.40jacquesmaybe you need to exec ash
11:42.00jacquesexec /bin/sh that is
11:42.11CIA-1003rwhitby 07 * r1.2563 10packages/busybox/switchbox-1.00/linuxrc: Switchbox now drops to shell if a valid rootfs cannot be mounted
11:48.11rwhitbysame result
11:48.49jacqueswhat's running as pid 1 ?
11:49.03rwhitbywe don't have ps :-)
11:49.09jacqueslol oops
11:49.25jacquesmaybe there's something clever you can do by looking in /proc
11:49.34rwhitbyself -> 25
11:49.39rwhitbycd 1
11:49.51jacquescat /proc/1/cmdline
11:50.39rwhitbycat status
11:50.46rwhitby2: keventd
11:51.01rwhitby3: ksoftirqd_CPU0
11:52.28rwhitbysys/kernel/panic ...
11:53.39rwhitbyecho "5" > /proc/sys/kernel/panic
11:56.42jacqueswhat's that do rwhitby?
11:56.53rwhitbyreboot after 5 seconds on panic
11:57.34rwhitbyturning off the /bin/sh -x in llinuxrc now - it's stable enough
12:07.32rwhitbyuploading flashdisk.jffs2 to nslu2-linux downloads, so I can then wget it onto devslug and dd it into /dev/mtdblock4
12:07.44jacquesmaybe of interest:
12:11.05rwhitbywe need to turn on command line editing in ash :-)
12:11.56jacquesah yeah
12:13.09rwhitbyand we need a /tmp on which to mount a ramdisk
12:13.40rwhitbyhmm - I'm already inside a ramdisk at this point, right?
12:14.53rwhitbyheh - wget is no good without CSR modules :-)
12:15.18rwhitbylrzsz ?
12:15.38jacquesI wonder waht this feature does: CONFIG_FEATURE_INITRD=y
12:15.57jacquesCSR would totally blow our budget several times over
12:16.25jacquesbb should have rz
12:17.05jacqueswell it has rz in misc utils section
12:17.20jacquesoops meant rx - receive xmodem
12:17.44rwhitbywell, without network access, this isn't a very interesting line of investigation any more ,,,
12:18.17rwhitbyok, how about pluggin in a flash key, and copying a file from that ....
12:18.53jacquesthat should work - if we wanted we could probably fit in support for some usb nic
12:19.51rwhitbyflash key is the way to go
12:21.06jacqueswe could do a CSR-enabled version in 256K
12:21.59rwhitbynah - then you need to go through the rigmarole of ifconfig and stuff - much easier just to plug in a usb flash disk and dd your new jffs2 partition from there
12:22.36jacquescompressed CSR modules add just over 100kB
12:22.49jacquesjust FYI
12:23.04rwhitbytempting ...
12:23.22jacquessomething to keep in mind
12:23.53rwhitbydd if=flashdisk.jffs2 of=/dev/mtdblock4
12:25.08jacquesheh I know
12:25.13rwhitbyshould probably have said bs=128k
12:25.50rwhitby# dd if=flashdisk.jffs2 of=/dev/mtdblock4                                      
12:25.51rwhitby8960+0 records in                                                              
12:25.51rwhitby8960+0 records out    
12:26.02jacqueswe store the compressed CSR modules in the last block after the partition table - and have a script that uses dd to extract them
12:26.14jacquesinto the ramdisk
12:26.23rwhitbyooooh - that's nefarious