**** BEGIN LOGGING AT Thu May 26 23:59:56 2005 May 27 05:56:09 jbowler-zzz: I don't know if you saw this or not but busybox still fails: May 27 05:57:22 test May 27 05:57:44 ERROR: /local/openslug/nslu2-package-symlinks/packages/busybox/busybox_1.00.bb:99 May 27 05:57:53 <[g2]> fails for me to I'm looking at it right now May 27 05:57:56 unparsed line: 'pkg_postinst_${PN} () {' while parsing /local/openslug/nslu2-package-symlinks/packages/busybox/busybox_1.00.bb May 27 05:57:59 Ok. May 27 05:58:17 <[g2]> pkg_postinst_${PN} () { May 27 05:58:17 <[g2]> update-alternatives --install /bin/mount mount /bin/mount.busybox 50 May 27 05:58:17 <[g2]> update-alternatives --install /bin/umount umount /bin/umount.busybox 50 May 27 05:58:17 <[g2]> } May 27 05:58:17 <[g2]> pkg_prerm_${PN} () { May 27 05:58:18 <[g2]> update-alternatives --remove mount /bin/mount.busybox May 27 05:58:19 I noticed the OE guys pushed in 2-3 busybox last night but doesn't seem to be related. May 27 05:58:22 <[g2]> update-alternatives --remove umount /bin/umount.busybox May 27 05:58:24 <[g2]> } May 27 05:58:26 <[g2]> May 27 05:58:28 <[g2]> nod. May 27 05:58:58 I don't know the syntax enough to say if there's a problem or not unfortunately May 27 05:59:08 <[g2]> same boat. May 27 05:59:12 :) May 27 05:59:23 <[g2]> I don't know if postinst went away May 27 05:59:28 perhaps. May 27 05:59:36 Where's kergoth when you need him! ;) May 27 05:59:49 <[g2]> I'll ask in OE May 27 05:59:52 How about replacing the $PN part with a static name? May 27 05:59:58 thanks May 27 06:01:18 <[g2]> I don't think that's it May 27 06:01:25 no? May 27 06:01:41 jbowler would know as I think he made the last change. May 27 06:15:47 <[g2]> VoodooZ_Work, I think jbowler was trying to do the right thing May 27 06:16:00 <[g2]> I don't think OE is setup quite that way May 27 06:16:43 <[g2]> change both _${PN} to _busybox and it'll not error and build May 27 06:27:02 cool May 27 06:27:30 I thought it look odd. (as in, I've never seen it before) May 27 07:33:38 You need a new bitbake! May 27 07:34:23 You need a minium 212. If you get the later versions you will get mickeyl's interactive bitbake changes, I haven't tried these. May 27 07:39:41 <[g2]> jbowler-zzz, I svn updated a couple days ago May 27 07:43:46 Then you haven't got 212. pb made that in the same few hours as he made all the changes to busybox. May 27 07:43:59 i.e. 'made that change' May 27 07:44:15 <[g2]> is there a way to see the version I've got ? May 27 07:44:23 Run bk revtool packages/busybox_1.00.bb and examine the log for CoreDump's changes. May 27 07:44:29 svn info May 27 07:44:50 <[g2]> THX 202 May 27 07:45:45 No chance of success ;-) May 27 07:46:14 <[g2]> actually I changed _${PN} to _busybox and it built fine :) May 27 07:46:15 svn -r 212 update - this is safe, it works May 27 07:46:45 [g2]: like I said, go and examine the log of CoreDumps last three changes to packages/busybox_1.00.bb May 27 07:47:00 <[g2]> I want straight to 223 any idea if that'll work ? May 27 07:47:04 <[g2]> went May 27 07:47:26 Wow, mickeyl has been busy. If I was a betting man I'd say evens. May 27 07:48:07 You can go backwards with the "svn -r 212 up" command as well. May 27 07:48:40 <[g2]> nod, [g2] is a betting man and going with evens :) May 27 07:49:34 <[g2]> busybox built May 27 07:50:00 Well you can try 'bitbake -i' ;-) May 27 07:50:49 <[g2]> bitbake -i May 27 07:50:49 <[g2]> BitBake Shell Version 0.1 (C) 2005 Michael 'Mickey' Lauer May 27 07:50:49 <[g2]> Type 'help' for more information, press CTRL-D to exit. May 27 07:50:49 <[g2]> BB>> May 27 07:50:58 [g2]: I've got some changes to openslug.conf which seem to be fine. I made ext2/ext3 into a module (no switchbox, doesn't need to be built in) and added reiserfs May 27 07:51:31 [g2]: bitbake -i allows multiple bitbakes without reparsing May 27 07:52:13 <[g2]> neat can you incrementally parse just a couple files ? May 27 07:52:30 <[g2]> parse :) May 27 07:52:35 [g2]: The other things I did were to change the DISTRO_VERSION to 1.13-pre (to avoid confusion with 1.12) and to move 'lrzsz' out of openslug-image into the RDEPENDS addition in conf. May 27 07:54:00 Oh, and I did the same move with the ext2 fs progs and added reiserfsprogs. So someone who wants a flash only build can now avoid all the file system stuff just by overriding OPENSLUG_EXTRA_RDEPENDS, but the image is unchanged (apart from ext2 in a module) May 27 07:54:05 <[g2]> jbowler, I'd like to remove the dependency on the linksys stuff for Openslug May 27 07:54:42 <[g2]> jbowler, what's the delay on boot with the module getting loaded ? May 27 07:54:47 Well, I don't dare try this, but what if you replace the RedBoot with a bunch of 0's? May 27 07:56:01 <[g2]> I"m not running Redboot :) May 27 07:56:05 <[g2]> I'm running APEX :) May 27 07:56:14 [g2]: if you specify '-t' to turnup a fraction of a second, i.e. the first fs message is in the same second as the previous kernel one. May 27 07:56:58 Hum. Ok. So if I zap RedBoot and SysConf in the pack-image stuff do you believe I will have a brick? May 27 07:57:36 <[g2]> As a betting man I wouldn't try that without JTAG May 27 07:57:37 Here's the answer to your timing question with reiserfs: May 27 07:57:39 May 26 22:53:28 (none) user.debug kernel: timer: 0 May 27 07:57:40 May 26 22:53:28 (none) user.notice kernel: ReiserFS: sda2: found reiserfs format "3.6" with standard May 27 07:59:35 <[g2]> jbowler, ok.. I'm guessing in /linuxrc that the ./sdxx causes a mount from either /linuxrc or more likely the /boot/xxxx file May 27 07:59:59 <[g2]> that mount is fully synchronous May 27 08:00:16 <[g2]> so there should be no timing issues correct ? May 27 08:00:31 In /boot/disk there is a mount -t ext2 /dev/sda2 (or -t reiserfs in my case) May 27 08:01:00 <[g2]> right mount I'm talking about May 27 08:01:11 <[g2]> it my case I think it's a -t ext3 May 27 08:01:12 The only timing issue I'm aware of is the '10 second delay', which I don't have (and it works) May 27 08:01:48 <[g2]> OK... May 27 08:01:51 [g2]: yes, that's a big user problem. 'mount /dev/sda2 /mnt' without an option mounts an ext3 partition as ext2! May 27 08:02:15 <[g2]> that's not really a big problem May 27 08:02:25 <[g2]> you just lose journaling May 27 08:02:41 <[g2]> it's a big problem on a big disk :) May 27 08:02:43 What I intend to do to turnup is to get it to read the /proc/mounts entry and get the -t and -o exactly right - not rely on the string the user passed in. May 27 08:03:14 <[g2]> hmm... I'd rely on the string May 27 08:03:20 <[g2]> but warn the user May 27 08:03:39 <[g2]> but I'm sure will get it sorted out May 27 08:03:57 <[g2]> jbowler, THX for the heads up about bb -i May 27 08:04:03 <[g2]> and versioning info May 27 08:04:04 What I mean is that the user options are used to do the mount, but then I read the /proc/mounts to generate /linuxrc. May 27 08:04:29 <[g2]> ah.. ok May 27 08:05:03 Then if a user doesn't specify a -t, but it works, it will still work on boot and it won't go loading the FAT modules. May 27 08:05:05 <[g2]> so I just do a "turnup ram" to reboot to the ram ? May 27 08:05:20 Yes. May 27 08:05:25 <[g2]> and reboot May 27 08:05:41 ;-) yes. May 27 08:05:52 BTW have you seen the hang-on-shutdown problem? May 27 08:06:03 <[g2]> do you recommend "reboot", "shutdown -r now", yes on the problem May 27 08:06:12 <[g2]> but not lately in the last couple reboots May 27 08:06:23 <[g2]> I haven't tried the new stuff yet May 27 08:06:28 Mine hung yesterday when I rebooted. But I'm still running the 1.12 release May 27 08:06:41 <[g2]> I'm post 1.12 May 27 08:06:43 They should be identical. On the problem I saw a [fatal] from the ethernet driver complaining about a bad state machine setting. May 27 08:06:52 * NAiL crawls out of hibernation May 27 08:07:21 I'm fully up-to-date and I see it pretty much 100% of the time on reboot after turnup disk, and very rarely on reboot from a flash root. May 27 08:07:28 <[g2]> jbowler, so what process are you using to update the jffs2 now ? May 27 08:07:32 Yeah May 27 08:08:09 upslug. I've been thinking about turnup ram and update in place, but if it fails the user still has to go back to upslug (etc) May 27 08:08:31 <[g2]> jbowler, that's what I'm trying to move to May 27 08:09:51 I've successfully done a kernel update, but the real priority here is to fix the boot sequence - the initscripts stuff is baroque and broken. May 27 08:09:55 <[g2]> Ok I just hung on reboot May 27 08:10:10 jbowler: When does it output the error? When it tries to take down the network interface? May 27 08:10:46 [g2]: did you get a [fatal] from the IAL stuff? May 27 08:10:55 NAiL: yes May 27 08:10:58 <[g2]> I did 4 May 27 08:11:15 <[g2]> Deconfiguring network interfaces... [fatal] IXETHACC:ixEthAccPortDisable: ixEthAccPortDisable failed port 0 (state = 4) May 27 08:11:15 <[g2]> ixp425_eth: eth0: BUG: ixEthAccPortDisable(0) failed May 27 08:11:15 <[g2]> done. May 27 08:11:15 <[g2]> Sending all processes the TERM signal... May 27 08:11:15 <[g2]> Sending all processes the KILL signal... May 27 08:11:17 Is it necessary to take down the iface at all? It doesn't get taken down if there's a smb or nfs mount anyways? May 27 08:11:17 <[g2]> Unmounting remote filesystems... May 27 08:11:18 I.e. the error is in the driver shutdown (I believe). I updated the bug report with the one I encountered. May 27 08:11:19 <[g2]> Rebooting... Restarting system. May 27 08:11:41 <[g2]> I think it's in the "Restarting System" that isn't correct May 27 08:12:30 Ah! It's something before S20sendsigs. May 27 08:12:54 [g2]: because the ixp driver is hung? May 27 08:13:12 <[g2]> dunno May 27 08:13:42 <[g2]> I'd think the sending all .... means S20Sendsigs is running from the text May 27 08:13:52 Duh. K40networking is immediately before S20sendsigs. NAiL: you may be correct. Try removing /etc/rc6.d/K40networking ;-) May 27 08:14:25 <[g2]> so you think K40networking is hung ? May 27 08:14:27 rebooting May 27 08:15:02 <[g2]> I'd probably try a set -x in S20Sendsigs May 27 08:15:17 hmm... doesn't look too good so far. But I only removed the ifdown -a. May 27 08:15:18 'Deconfiguring network interfaces...' line 84 of K40networking May 27 08:15:55 ok, still hung. Gonna remove the symlink May 27 08:16:12 <[g2]> put a set -x in S20Sendsigs May 27 08:16:18 mmkay May 27 08:16:28 NAiL: probably won't help. I suspect the answer to your question is that the networking must be down for the reboot to work. May 27 08:17:21 I'm going to try the force-reload option... May 27 08:17:51 That's fine. May 27 08:17:56 [g2]: Uh... set -x won't work now, will it? I haven't got serial yet ;) May 27 08:17:59 [g2]: how about upgrading to 1.5? May 27 08:18:15 <[g2]> no not for you :) May 27 08:18:53 NAiL: no, it won't work. What I do is "exec 2>>/var/log/shutdown.log >&2" at the head of each script. May 27 08:19:51 <[g2]> Ok it's in exit somewhere May 27 08:19:55 jbowler: That takes the output and puts it in the log? May 27 08:20:17 Never seen that construct before :) May 27 08:20:22 NAiL: yes, appends it to a log, which must be on a hard disk (but turnup makes /var on the disk) May 27 08:20:54 NAiL: it means 'take my (the shell's) streams and permanently change them', May 27 08:22:06 NAiL: see the line in /etc/init.d/checkroot.sh for a really obscure use. May 27 08:22:52 hrrm.. ok. May 27 08:23:16 <[g2]> I'll need to check, but I think there was some warm boot code that kas had oringially added that looped and needed a FIXME May 27 08:23:21 Is there an easy way to do a "for a in *;" kind of loop to add that to all the useful scripts? May 27 08:23:30 [g2]: rmmod might work, but there's clearly something strange in the shutdown which is why I want to clean it up so it does things in the write order (e.g. umount NFS before bringing down the network!) May 27 08:23:49 <[g2]> jbowler, ABSOLUTELY May 27 08:24:02 * [g2] hugs jbowler for using the proper approach May 27 08:24:09 <[g2]> and all his hard work May 27 08:24:10 NAiL: there aren't that many ;-) May 27 08:25:05 jbowler: Thanks. I updated bitbake and everything works fine now. May 27 08:25:21 NAiL: edit /etc/rc, the case statement where it runs the script, or just put the line at the top. You can also 'set -x' and pass -x to the script. May 27 08:26:45 Since that would leave the file descriptor open when S90reboot runs I just did the redirection on the script command line(s) then did an 'exec 2>&- >&-' in S90reboot. May 27 08:27:27 After I'd made humungous numbers of such changes it started working for me, but I couldn't work out what had fixed it. I think just doing things in a different order with different delays... May 27 08:27:58 VoodooZ_Work: what version of bitbake do you have now ( May 27 08:28:02 "svn info")? May 27 08:29:14 223 May 27 08:29:46 seems to work fine. May 27 08:30:07 Ah thanks, same as [g2] - bitbake -i capable. Sounds like that is a good version too (I'm still at 212). May 27 08:30:24 what's the -i for? May 27 08:30:47 I see interactive mode. May 27 08:31:07 Does it wait for confirmation on each step? May 27 08:36:39 I don't know - I haven't used it. I believe you get to enter a command, it does it, then you do another command. It doesn't need to reparse the bb files each time. May 27 08:37:44 <[g2]> Ok jbowler can you peek at http://www.nslu2-linux.org/wiki/OpenSlug/RebootIssue May 27 08:40:05 <[g2]> jbowler, nice post to the ml about bb btw May 27 08:40:17 [g2]: I can't see anything unexpected in there. May 27 08:40:41 <[g2]> Ok... I think we need to look at exit May 27 08:40:55 ml: should have done it yesterday, I put the info on nslu2-linux then forgot about it. May 27 08:41:08 <[g2]> because it's clearly after Rebooting.... Restarting sytem May 27 08:41:19 <[g2]> np. May 27 08:41:30 <[g2]> you've been busy May 27 11:02:19 <[g2]> jbowler-away, I'm testing 1.20-pre1 and mke2fs will need to be added back in :) May 27 11:03:54 It's in there already: e2fsprogs-mke2fs May 27 11:05:41 <[g2]> jbowler-away, it's not on the image :) May 27 11:05:48 <[g2]> the jffs2 image that's been flashed May 27 11:07:12 Does your work directory have a work/e2fsprogs* subdirectory? May 27 11:07:39 <[g2]> nope May 27 11:07:42 That would be e2fsprogs-1.36-WIP-0408-r3 May 27 11:07:56 <[g2]> I deleted "tmp" and did a full rebuilt May 27 11:07:57 Have you overriden OPENSLUG_EXTRA_RDEPENDS? May 27 11:07:57 <[g2]> I deleted "tmp" and did a full rebuild May 27 11:08:30 <[g2]> OPENSLUG_EXTRA_DEPENDS = "" May 27 11:08:30 <[g2]> OPENSLUG_EXTRA_RDEPENDS = "" May 27 11:09:09 Yes, that's what I intended to happen - so that builders can remove packages that aren't required to boot. May 27 11:10:52 Ah, it's in the SVN local.conf in the trunk. Duh. May 27 11:12:10 <[g2]> and we'll need to add reiserfsprogs to nslu2-symlinks May 27 11:13:23 <[g2]> Ok gotta run May 27 11:13:36 <[g2]> bb in a couple hours May 27 11:14:30 I'll add the symlinks... Not sure about the DEPENDS/RDEPENDS. I think bk should reflect what the SVN build does (i.e. drop the extra kernel modules). May 27 11:18:38 Pushed the symlinks May 27 11:42:42 [g2]: I made an update to trunk/openslug/conf/local.conf.template which tests out ok with my bitbake build (i.e. the SVN trunk should now generate a good build which includes the mkfs and lrzsz programs) May 27 11:58:48 <[g2]-away> jbowler-away, you did push that did you ? It doesn't seem to be there when I pull May 27 11:59:30 <[g2]-away> you diddn't push that did you .. May 27 11:59:43 It's in the SVN tree, I pushed it - see the CIA-1 messages on nslu2-linx May 27 11:59:54 <[g2]-away> no. just seconds ago :) May 27 11:59:58 <[g2]-away> nod. May 27 12:00:31 The r44 revision was several minutes ago- you should have got it. Are you pulling the svn trunk, not the release? May 27 12:01:56 Ok, flex is fixed, I'm going to remove the downloads/flex hack from the trunk. May 27 12:02:07 <[g2]-away> Parent repository is bk://nslu2-linux.bkbits.net/openembedded May 27 12:02:09 BTW jacques found a whole load of other missing downloads... May 27 12:02:57 <[g2]-away> cool.. way to go jacques May 27 12:03:02 <[g2]-away> and jbowler-away May 27 12:03:05 [g2]-away - right, I changed the SVN trunk, not the BK. I can't change local.conf in BK, but SVN generates one automagically. (So I'm saying you need to alter your own local.conf) May 27 12:03:57 <[g2]-away> ok May 27 12:04:46 [g2]-away - can you tell me which version of flex you built? (I'm still building the old one, but I think that is my error.) May 27 12:05:26 <[g2]-away> OTE: Unpacking /home/tom/dev/berlios/openslug/downloads/flex-2.5.31.tar.bz2 to /home/tom/dev/bb_1x/tmp/work/flex-native-2.5.31-r0/ May 27 12:06:54 Weird, I wonder if bitbake regards '4a' as earlier than '31' I've just checked and I don't pick up the 4a release. May 27 12:11:39 <[g2]-away> jbowler-away, what the svn trunk ? May 27 12:11:42 <[g2]-away> I mean the url for the openembedded May 27 12:11:42 [g2]-away - so the svn trunk should be correct, but it hasn't been updated with the latest openembedded or nslu2-linux-symlinks. The flex hack is still there, I'll try to find out what is going on later. May 27 12:12:47 <[g2]-away> can you e-mail me you local.conf ? May 27 12:12:53 URL: svn+ssh://svn.berlios.de/svnroot/repos/openslug May 27 12:13:25 <[g2]-away> yeah. I still haven't signed up for a dev account on berlio May 27 12:22:56 <[g2]-away> THX May 27 12:25:43 So I just changed the SVN local.conf template to add the mkfs programs. I didn't change openslug.conf, though the obvious thing is for it to match the SVN local.conf (so if the local.conf settings aren't made the result is the same._ May 27 12:26:08 There's a whole load of weird kernel modules (pegasus etc) in there, which are not in OpenSlug-1.12 because of the lines in local.conf. May 27 12:27:15 <[g2]-away> ok May 27 12:27:21 Also, once a user has a flash boot it's straight forward (now) to 'ipkg update' 'ipkg install e2fsprogs-mke2fs', but the problem is knowing what packages to ask for ;-) May 27 12:27:58 <[g2]-away> nod. :) May 27 12:39:24 <[cc]smart> hi May 27 12:40:47 [g2]-away - I have a modified SVN trunk (and, boy, is this going to be difficult...) and I'm trying an SVN build now - i.e. I have the trunk with my mods plus a known-good bk pull. May 27 12:41:17 <[g2]-away> bb is acting funny for me now :( May 27 12:41:54 [g2]-away this is going to be difficult because the trunk has to be updated with all the adds/deletes in bk. May 27 12:41:54 <[g2]-away> bb May 27 12:41:54 <[g2]-away> ERROR: Unable to parse conf/bitbake.conf May 27 12:42:31 Um, you have 223? May 27 12:42:38 <[g2]-away> jbowler-away, I wouldn't worry about the SVN trunk right now May 27 12:42:40 <[g2]-away> yes May 27 12:43:06 <[g2]-away> it worked before but now it's unhappy May 27 12:43:42 So only that local.conf change - is the OEROOT at the top set correctly? May 27 12:52:16 Using the SVN Makefile with that local.conf.template I have a working build in progress (i.e. bb parsed everything fine - no errors). The only change the Makefile makes is to change the %%%OEROOT%%% to the directory containing openembedded. May 27 12:53:21 I'm going to leave this building, it's a good test because it doesn't use anything special in my environment. biab. May 27 13:28:31 <[g2]-away> jbowler-away, The reboot hanging has something to do with the sdaXX root May 27 13:28:49 <[g2]-away> It's may be the pivot root May 27 13:29:12 <[g2]-away> because rebooting works without the flashstick plugged in at least a couple times May 27 16:19:53 [g2]-away I also consistently got the reboot hang with an NFS root. I checked the /initrd - there is no problem umounting it (so no process is still using it after the pivot_root). May 27 16:20:49 Still, I'm sure you are correct - something is different. It might be something to do with having /dev on the disk, not in ramfs... **** ENDING LOGGING AT Fri May 27 23:59:56 2005