**** BEGIN LOGGING AT Mon Jun 06 23:59:56 2005
Jun 07 00:21:09 <dyoung> jacques, did you have an interest/obsession with Giant Robots?
Jun 07 00:21:40 <dyoung> This thing is kind of twisted...
Jun 07 00:21:42 <dyoung> http://www.sakakibara-kikai.co.jp/products/other/LW.htm
Jun 07 00:23:12 <glc> I unmounted /initrd in unslung 5.1
Jun 07 00:24:10 <glc> To do it, I had to delete /dev/console from jffs.
Jun 07 00:27:03 <jacques> dyoung is that thing real??
Jun 07 00:28:01 <dyoung> as far as I can tell.  This company normally makes farm equipment.
Jun 07 00:28:13 <jacques> amazing :-)
Jun 07 00:28:30 <jacques> hmm I need to setup a local ipkg repo
Jun 07 00:29:29 <dyoung> glc, Hmm, any suggestions as how to non-destructively umount /initrd ?
Jun 07 00:29:43 <jacques> bothers you does it?  :-)
Jun 07 00:29:48 <jacques> it kinda bothers me a bit too
Jun 07 00:30:59 <glc> What do you mean non-destructively?
Jun 07 00:32:48 <dyoung> removing the /dev/console from jffs2 means that you can never boot it without disks?
Jun 07 00:34:40 <jacques> is there a howto on setting up a local ipkg repo on the wiki?
Jun 07 00:37:22 <glc> I should have an answer tomorrow. Experiments look promising.
Jun 07 00:50:56 <dyoung> okay cool!
Jun 07 00:51:20 <dyoung> jacques, I remember it was discussed before, but I dunno if anyone bothered to write the howto.
Jun 07 00:51:28 <jacques> :-(
Jun 07 00:54:37 <dyoung> was it was easy as adding /etc/ipkg/local-repo.conf whihc contains, "src/gz local file://directory/of/feed" ?
Jun 07 01:06:04 <jacques> something screwed up
Jun 07 01:06:18 <jacques> ipkg-make-index buggy
Jun 07 01:06:27 <jacques> didn't include the bash package
Jun 07 01:06:33 <jacques> no trust
Jun 07 01:06:37 <jacques> POS
Jun 07 01:11:02 <jacques> it just ignores the bash package
Jun 07 01:11:10 <jacques> well it says "reading bash package info"
Jun 07 01:11:18 <jacques> but then does not include it in the Packages file
Jun 07 01:11:24 <jacques> who knows what else it's missing
Jun 07 01:15:02 <jacques> it also missed ncurses
Jun 07 01:15:17 <jacques> just the stuff I need is inexplicably left out
Jun 07 01:16:26 <jacques> totally useless
Jun 07 01:20:30 <jacques> oh I might know what the problem is
Jun 07 01:20:35 <jacques> but not how to solve it
Jun 07 01:21:06 <jacques> I just have no clue how to use ipkg-make-index and /usr/bin/ipkg-make-index [-h] [-s] [-m] [-l Packages.filelist] [-p Packages] [-r Packages.old] [-v] packagesdir doesn't help a whole lot
Jun 07 01:23:03 <dyoung> I guess you need to have a Packages{.gz} and Packages.filelist handy...
Jun 07 01:25:17 <glc> dyoung: make the 2nd line of linuxrc be:
Jun 07 01:25:36 <glc> exec </dev/cons >/dev/cons 2>/dev/cons
Jun 07 01:25:49 <glc> and rename /dev/console /dev/cons.
Jun 07 01:26:12 <glc> unsling will have to create /dev/console on the target.
Jun 07 01:26:16 <jacques> dyoung I found it on google:  ipkg-make-index . > Packages
Jun 07 01:26:24 <jacques> I was missing the > Packages part
Jun 07 01:26:39 <jacques> and it had an old Packages file that oe had generated and wasn't overwriting it
Jun 07 01:27:27 <glc> I will test further.
Jun 07 01:31:08 <dyoung> Ok, I'm still digesting that.
Jun 07 01:33:58 <jacques> OK this is really screwed up too - I mount -t ext3 /dev/sda1 /home
Jun 07 01:34:09 <jacques> and now /home is drwx------    6 root     root         4096 Jun  7 00:19 home
Jun 07 01:34:24 <jacques> umask ??
Jun 07 01:34:37 <jacques> no I don't think so
Jun 07 01:36:08 <jacques> hmm, chmoding it once while mounted seems to have fixed it - weird
Jun 07 01:39:36 <dyoung> I used to have that in the old days... one cause was the directory permission of the directory that you mount it on... I dunno if thats still the case..
Jun 07 01:40:46 <jacques> that's the weird thing, the mount point was 755, but after mounting it was 700
Jun 07 01:40:56 <jacques> then unmount and it's 755 again
Jun 07 01:51:49 <rwhitby-away> jacques: umask?
Jun 07 01:54:18 <jacques> rwhitby-away, apparently that was not it
Jun 07 01:54:29 <jacques> I get interesting errors when I do halt on openslug
Jun 07 01:54:37 <jacques> with a disk attached
Jun 07 01:54:53 <jacques> Unmounting local filesystems...
Jun 07 01:54:53 <jacques> Power down.
Jun 07 01:54:53 <jacques> usb 2-1: new full speed USB device using ohci_hcd and address 2
Jun 07 01:54:53 <jacques> usb 2-1: not running at top speed; connect to a high speed hub 
Jun 07 01:54:53 <jacques> SCSI error : <0 0 0 0> return code = 0x70000                  
Jun 07 01:54:54 <jacques> end_request: I/O error, dev sda, sector 77837032
Jun 07 01:54:56 <jacques> etc etc
Jun 07 02:03:30 <jacques>  Timing buffered disk reads:   68 MB in  3.00 seconds =  22.67 MB/sec - I wish the slug could do that (same drive hooked to my shuttle)
Jun 07 02:11:07 <rwhitby-away> both openslug and unslung finally build ...
Jun 07 02:15:57 <jacques> ok, my openslug drive is backed up, so I can repartition it and set it up optimally for turnuping :-)
Jun 07 03:10:13 <dyoung> Its amazing what one can build otu of aluminum using only a file, drill and hacksaw.
Jun 07 03:59:40 <jacques> weird - oe does not have groff
Jun 07 04:07:04 <jacques> dammit
Jun 07 04:07:11 <jacques> oe is so broken right now 
Jun 07 06:23:33 <NoRemorse> hi all
Jun 07 06:23:55 <NoRemorse>  does the backup function under the administration menu on the web interface of the nslu2 , when used to backup to a disk in port 2, does it make a disk that can be swapped to port 1 ready to boot from? or is it just a copy of the folders, no permissions,config, or anything? (standard shipped firmware)
Jun 07 08:58:28 <macsat> hi
Jun 07 08:59:23 <phedny> good day
Jun 07 08:59:45 <phedny> macsat: nice wlguide btw :)
Jun 07 09:01:05 <macsat> thanks :)
Jun 07 09:13:34 <phedny> i'm trying to make an ipkg package
Jun 07 09:13:48 <phedny> but the ./configure keeps stopping on a missing gettext
Jun 07 09:13:54 <phedny> checking for gettext in -lintl... no
Jun 07 09:13:54 <phedny> Have libintl.h but no usable gettext!
Jun 07 09:14:17 <phedny> (btw, it's poptop i'm working on)
Jun 07 09:25:10 <phedny> whow, it compiles :)
Jun 07 16:37:03 <[g2]> hey kergoth ! How goes ?
Jun 07 18:40:47 <jacques> Disconnecting: Protocol error: expected packet type 31, got 82
Jun 07 18:41:02 <jacques> I haven't seen that before - got it on my openslug when ssh'd in
Jun 07 18:44:48 <jacques> root@LKG0FB08F:~# uptime
Jun 07 18:44:48 <jacques> uptime: couldn't get boot time: No such file or directory
Jun 07 18:44:52 <jacques> also on openslug
Jun 07 18:46:04 <jbowler-away> jacques: on openslug (a very recent build):  01:45:18 up  3:59, load average: 1.00, 1.00, 0.91
Jun 07 18:46:48 <jacques> jbowler-away, oh crap I just realized the problem
Jun 07 18:47:14 <jacques> I installed coreutils - built an openslug native devel env
Jun 07 18:47:36 <jacques> root@LKG0FB08F:~# busybox uptime
Jun 07 18:47:36 <jacques>  18:39:56 up 15:14, load average: 1.00, 1.00, 1.00
Jun 07 18:47:53 <jacques> jbowler-away, did you write turnup?
Jun 07 18:48:24 <jbowler-away> yes, assuming you have the 800 line version not the 10 line version.
Jun 07 18:48:38 <NAiL> jacques: What do you need to get a native dev env? I tried to set that up earlier, but failed.
Jun 07 18:48:41 <jacques> jbowler-away, it's freakin' awesome
Jun 07 18:49:24 <NAiL> jacques: You wouldn't happen to have it nicely wrapped up in some ipkgs? :D
Jun 07 18:49:33 <jacques> NAiL, the answer to that question is long and complex, made more so by current oe deficiencies
Jun 07 18:49:47 <NAiL> urk
Jun 07 18:50:12 <jacques> oe-built bison doesn't work, and flex doesn't even build currently
Jun 07 18:50:15 <NAiL> I had a working dev env for uClibc running. Lord knows how I got that to work.
Jun 07 18:50:27 <jacques> they are getting too agressive with putting in new versions and removing the old tested ones IMHO
Jun 07 18:50:57 <jacques> there are still manual changes which need to be made too, not just installing ipks
Jun 07 18:51:26 <jacques> my checklist is only a bit shorter than it was a few months ago :-(
Jun 07 18:52:17 <NAiL> A shame actually. Compiling stuff on the slug is much, much easier.  Although samba and mysql took quite some time to compile :P
Jun 07 18:52:43 <jbowler-away> The latest (a few hours ago) checkin to flex seems to reveal a 3.4.4 ARM compiler bug - the build aborts with a 'report me' gcc errror.
Jun 07 18:53:15 <jacques> the bit test is later tonight when I try to build perl on openslug. it worked months ago, we'll see how it does with the latest stuff
Jun 07 18:54:04 <jacques> jbowler-away, ah, last I had heard from this morning was 
Jun 07 18:54:05 <jacques> <hrw|work> | tables.c: In function `yytbl_write16':
Jun 07 18:54:05 <jacques> <hrw|work> | tables.c:313: internal compiler error: in arm_print_operand, at config/arm/arm.c:11410
Jun 07 18:54:28 <jacques> and that they were going to mark it as broken on arm and be done with it :-(
Jun 07 18:54:42 <jbowler-away> Yes, that's it - I was trying to install KDE 3.4.1 so I had to /bye...
Jun 07 18:55:03 <jacques> after that, I natively built flex 2.5.4a and 2.5.31 and ran the testsuites - all passed
Jun 07 18:55:22 <jacques> so the compiler bug is only in cross
Jun 07 18:55:57 <jacques> I also built natively bison 1.875 and 2.0 (version oe has) and both tested out 100% (oe-built bison aborts)
Jun 07 18:56:41 <jacques> makes me wonder if we shouldn't go back to 3.4.3 for now
Jun 07 18:57:21 <jbowler-away> What happened, it worked yesterday (and I thought it worked with 3.4.4)
Jun 07 18:57:43 <jacques> which worked yesterday?
Jun 07 18:59:30 <jbowler-away> I've got a flex-2.5.31 build directory from June 6 15:21 (which is PDT), but, curiously, it has not install subdir, yet I got a complete openslug-packages.
Jun 07 19:00:10 <jacques> jbowler-away, two things going on here
Jun 07 19:00:13 <jbowler-away> tables.o is from June 6 15:12
Jun 07 19:00:59 <jacques> until I reported the issue this morning, there was a bug in the flex .bb which was causing it to build a host-native binary
Jun 07 19:01:16 <jacques> so if you look in your flex build dir you should find 80386 binaries
Jun 07 19:01:32 <jacques> I reported that and they fixed it, which exposed the arm cross-build error
Jun 07 19:02:37 <jbowler-away> Oh yes, totally bogus 'image' directory ;-)  Ok, so it needs to be moved to 'broken' in packages.
Jun 07 19:03:03 <jacques> <hrw|work> flex.inc inherit native...
Jun 07 19:03:03 <jacques> <hrw|work> Im fixing it now
Jun 07 19:03:20 <jacques> yeah, right now it is broken, tho I am tempted to try with 3,4,3
Jun 07 19:04:06 <jbowler-away> It was building on 27 May
Jun 07 19:04:51 <jacques> yeah I know it used to work - I have it installed in my chroot envs 
Jun 07 19:05:02 <jacques> but that was pre-3.4.4 IIRC
Jun 07 19:06:19 <jacques> anyone recall (or know where I can find) the PREFERRED_VERSION lines I need to temporarily change back to 3.4.3 ? I wanna try
Jun 07 19:06:28 <jbowler-away> Yes, and there are no significant changes.
Jun 07 19:07:12 <jbowler-away> PREFERRED_VERSION_gcc-cross = "3.4.3"
Jun 07 19:08:01 <jbowler-away> If you have a complete build and just set that in local.conf bb should immediately (and only) build gcc-cross.  The gcc-cross-initial will still be 3.4.4
Jun 07 19:08:35 <jacques> ok let me try that route
Jun 07 19:10:23 <jacques> bk is getting slower and slower...
Jun 07 19:10:59 <jacques> had to pull to get the flex fix
Jun 07 19:13:23 <jacques> build starte
Jun 07 19:13:23 <jacques> d
Jun 07 19:26:18 <jacques> damn, I get the same compile error with 3.4.3, which implies it's the new version of flex which exposes the arm cross-compiler bug
Jun 07 19:27:01 <jacques> why do they have to remove ALL the older versions of apps when they "upgrade" ?
Jun 07 19:27:06 <jbowler-away> Did it really build 3.4.3?
Jun 07 19:27:19 <jacques> yeah
Jun 07 19:27:49 <jacques> NOTE: package gcc-cross-3.4.3: completed
Jun 07 19:28:06 <jbowler-away> Hum.  What happened, I have an 0422 work/flex-2.5.31-r0, but that pre-dates the flex_2.5.31.bb
Jun 07 19:28:49 <jacques> maybe they removed some patches or something, changed compiler options...
Jun 07 19:30:03 <jbowler-away> Sh**t, in the clone I have from 0524 flex_2.5.31.bb is at rev 1.15 from 09Dec2004!
Jun 07 19:31:19 <jacques> configure:2374: ccache armeb-linux-gcc --version </dev/null >&5
Jun 07 19:31:19 <jacques> armeb-linux-gcc (GCC) 3.4.3
Jun 07 19:31:59 <jacques> jbowler-away, ok, well I may be wrong about this flex version being new to oe - I know it's been out since 2003
Jun 07 19:32:13 <jacques> flex is not exactly being actively developed
Jun 07 19:32:48 <jbowler-away> The old version of the .bb file included a patch, but this only seems to change AM_CPPFLAGS
Jun 07 19:33:52 <jacques> I'm checking svn openslug 1.12
Jun 07 19:33:57 <jacques> flex built and is arm binary
Jun 07 19:34:21 <jacques> and packaged correctly
Jun 07 19:34:51 <jbowler-away> I believe that when mickeyl changed the flex bb files on May 27 somehow the compiler bug was introduced and simultaneously hidden by the native bug.
Jun 07 19:35:08 <jacques> and they *did* used to have an older version:
Jun 07 19:35:11 <jacques> flex-2.5.31/  flex-2.5.4-native_2.5.4a.bb  flex-native_2.5.31.bb  flex-native_2.5.4a.bb  flex_2.5.31.bb
Jun 07 19:35:29 <jacques> tho the successful build I mention above was 2.5.31
Jun 07 19:36:11 <jbowler-away> The only obvious thing is the include.patch
Jun 07 19:37:15 <jacques> yeah, I'm rather confuzzled
Jun 07 19:37:52 <jbowler-away> ... copy the flex_2.5.31.bb from 1.12 along with flex-2.5.31/include.patch ;-)
Jun 07 19:38:21 <jacques> `course, they coulda changed something in the gcc patches :-\
Jun 07 19:40:12 <jbowler-away> No, because you just tried with 3.4.3.
Jun 07 19:40:25 <jacques> they counda changed some 3.4.3 patches I meant
Jun 07 19:40:38 <jacques> I wouldn't put it past them
Jun 07 19:40:39 <jbowler-away> Yes, you're right
Jun 07 19:40:58 <jacques> especially in light of recent events
Jun 07 19:41:15 <ka6sox-away> so are we on 3.4.3 or 3.4.4?
Jun 07 19:41:16 <jbowler-away> My best guess is that apparently innocuous 'include.patch' prevents the build pulling in a host file and this in turn hides the compiler bug.
Jun 07 19:41:31 <jbowler-away> ka6sox-away: 3.4.4
Jun 07 19:41:59 <jacques> jbowler-away, not a bad guess, I guess I should test
Jun 07 19:45:21 <jacques> testing with openslug-1.12 flex files
Jun 07 19:46:49 <jacques> NOTE: package flex-2.5.31: completed
Jun 07 19:46:51 <jacques> !
Jun 07 19:47:20 <jacques> jbowler-away wins a cigar
Jun 07 19:47:52 * jacques goes to report on #oe.
Jun 07 19:52:14 <dyoung-web> ka6sox-away: I got a fedex envelope today.
Jun 07 19:52:32 <dyoung-web> no packe with slug or board though
Jun 07 19:58:43 <glc> I now have an umountable /initrd on unslung 5.1.
Jun 07 20:00:08 <ka6sox-away> arrrrgh...lemme ask my wife now.
Jun 07 20:00:14 <dyoung-web> doing the console rediction trick?
Jun 07 20:00:34 <ka6sox-away> has anybody hacked the WMA11?
Jun 07 20:00:41 <dyoung-web> Yes.
Jun 07 20:00:52 <glc> I removed /dev/console.  If no drive is mounted, a ramfs is created to hold a console node, and a symbolic linke is created in /dev.  The exec is then redirected to the new /dev/console.
Jun 07 20:01:13 <dyoung-web> http://www.turtlehead.co.uk/
Jun 07 20:01:13 <ka6sox-away> dyoung-web, I have one now.
Jun 07 20:02:20 <dyoung-web> glc, cool, can you mail me your changed linuxrc?
Jun 07 20:02:47 <glc> Email address? Perhaps in other chat room.
Jun 07 20:08:20 <jbowler-away> glc: there's a shell function 'swivel' in packages/openslug-init/openslug/openslug-init-0.10/functions which closes the /dev/console before the pivot_root (the function does everything to all the umount, at least on openslug0
Jun 07 20:09:46 <dyoung-web> I like that word. 'swivel' .
Jun 07 20:09:59 <dyoung-web> its right up there with nefarious.
Jun 07 20:10:08 <jbowler-away> It was swizzle, but that seemed a little too silly.
Jun 07 20:12:16 <glc> What is the syntax?
Jun 07 20:12:31 <jbowler-away> Somehow busybox ash manages to keep something else open too while running /linuxrc - I think even after the chroot bin/sh that sh still has fd 10 open on /linuxrc...
Jun 07 20:15:09 <glc> Is it a 2.6 thing?
Jun 07 20:17:39 <jbowler-away> No, fd 10 is how ash holds the shell script it is executing open.  (The swivel function should be generic, and exec <&- works in any bash-like shell).
Jun 07 20:19:49 <jbowler-away> glc: yeah, that's it - take a look at /proc/1/fd and /proc/$$/fd.  init points 10 at /dev/initctl, so it kills the old ash one on fd 10, so only after the exec of init can the initrd be umounted...
Jun 07 20:20:21 <jbowler-away> exec 10>&- isn't valid ;-)
Jun 07 21:00:16 * jacques starts a native openslug perl build.
Jun 07 21:04:15 <ka6sox> on Gastro?
Jun 07 21:04:45 <jacques> gastro is unslung :-)
Jun 07 21:05:46 <ka6sox> ahj
Jun 07 21:08:39 <jacques> I wonder if perl 5.8.7 still needs the perlio patch
Jun 07 21:08:53 <jacques> or was that just for uclibc... never mind
Jun 07 21:10:12 <ka6sox> okay...forgotten
Jun 07 21:10:41 <jacques> :-)
Jun 07 21:10:46 <jacques> how's the wma11b look?
Jun 07 21:11:36 <ka6sox> looks nice..but the flash is only 2mB and uses cramfs
Jun 07 21:12:08 <ka6sox> I kinda wish I could throw out the gui stuffs and just have dropbear and then media stuffs.
Jun 07 21:12:20 <ka6sox> s/then/they
Jun 07 21:12:37 <ka6sox> (including rpedde's ivorbis.lib
Jun 07 21:16:57 <ka6sox> I'm examining the source tarball from Linksys and the wma stuff that was done about 9 mo ago.
Jun 07 21:17:58 <jacques> the newest devel for it I saw was a guy trying to write his own bootloader
Jun 07 21:18:10 <ka6sox> ah
Jun 07 21:18:24 <jacques> it was having some issues, but I bet beewoolie could make APEX work on it
Jun 07 21:18:46 <ka6sox> yeah
Jun 07 21:19:02 <ka6sox> he would have to add cramfs support
Jun 07 21:19:46 <jacques> I wonder how the jffs2 support is going
Jun 07 21:23:54 <ka6sox> in APEX?
Jun 07 21:25:21 <jacques> yeah
Jun 07 21:38:15 <ka6sox> jacques, how long for that perl build?
Jun 07 21:40:44 <jacques> it'll be going for a while
Jun 07 21:41:20 <jacques> after the build I'm going to run the testsuite
Jun 07 21:41:46 <jacques> it's building the dynamic modules now
Jun 07 22:53:59 <jacques> ka6sox openslug native perl 5.8.7 build took 93m31.182s
Jun 07 22:54:25 <ka6sox-away> jacques, faster than I thought.
Jun 07 22:54:34 <ka6sox-away> does that include the verification suite?
Jun 07 22:55:34 <jacques> nope
Jun 07 22:55:49 <jacques> doesn't include configuration time either
Jun 07 22:56:17 <jacques> running testsuite now
Jun 07 22:59:24 <jacques> t/op/lfs..................................skipping test on this platform
Jun 07 22:59:27 <jacques> hmmmmmmm
Jun 07 22:59:29 <jacques> not good
Jun 07 22:59:39 <jacques> i hope oe isn't building glibc without lfs support
**** ENDING LOGGING AT Tue Jun 07 23:59:56 2005