**** BEGIN LOGGING AT Sat Jun 11 23:59:56 2005 Jun 12 04:59:36 Hmmm... is it possible to "turnup" to a non-empty disk file system? Hard do see why a drive has to be empty as described in the help text. Jun 12 04:59:50 s/drive/file system/ Jun 12 05:00:00 I think it is possible Jun 12 05:00:05 maybe without the -i Jun 12 05:00:17 jbowler-away and I were discussing this recently Jun 12 05:01:36 I thought that without the -i, the file system must already be populated with the rootfs Jun 12 05:01:47 So it only sets the linuxrc link. Jun 12 05:02:21 Tried looking at the script, but shell code quickly becomes unreadable. Jun 12 05:05:21 Ah, the "-f" option. Jun 12 05:05:23 bob_tm, I'm not sure :-\ Jun 12 05:05:40 what's that do? Jun 12 05:05:58 Force install of rootfs even if file system is not completely empty Jun 12 05:06:06 sounds promising :-) Jun 12 05:06:39 Yes, will try. Risky business - no backup :-o Jun 12 05:07:15 I love those undocumented options. Jun 12 05:13:52 i love having the source to things :-) Jun 12 05:15:44 Without it, it would be hard to figure out the undocumented options :-) Jun 12 05:16:03 yep Jun 12 05:16:30 everything flows from the source Jun 12 05:16:59 crap, my build is being held up because it can't connect to linksys Jun 12 05:17:12 (on a machine I hadn't built on since before R29 was added) Jun 12 05:17:41 ftp.linksys.com appears to be down Jun 12 05:27:06 jacques: Yes - it connects but no login. Jun 12 05:27:52 luckily I had another machine to copy from, but oe bit me again because it wants to connect to the site even tho it has the file, but not the .md5 Jun 12 05:28:08 hitting ctrl-c causes it to realize it doesn't need the file - bug I think Jun 12 05:52:58 bob_tm: so you've moved over to openslug now ? Jun 12 05:53:44 rwhitby: Better - I borrowed a second slug to which I installed OS. Jun 12 05:55:44 I still the need, though. The USB limitations on Unslung are to big and at some point I must migrate. With only a prod-slug it is hard. Jun 12 08:42:37 bob_tm: I deliberately didn't mention '-f' because the results are unpredictable. in particular the cpio just overwrites everything, however: Jun 12 08:42:38 root@lochalsh:/etc/rcS.d# turnup disk -i /dev/sda2 -t reiserfs -o noatime Jun 12 08:42:38 turnup: /tmp/rootfs.1058: partition contains existing files, specify -f to overwrite Jun 12 08:43:59 jbowler-away: I guess it will only overwrite existing files/directories with new files with same path name? That would be predictable enough for me. Jun 12 08:44:18 It's fine if /etc and /usr do not exist (/lib, /bin and /sbin can be overwritten safely), Jun 12 08:44:53 The problem arises if somthing like /etc/rcS.d/S??foo already exists, the system may well become unbootable in that case. Jun 12 08:45:50 Yes - that's what I thought. Perfectly fine. It was just some files that I could not store elsewhere in the meantime, safely tucked away under /music :-) Jun 12 08:47:12 Yes, the script could be more exact in its "find" - e.g "find etc" would probably be enough (i.e. if etc doesn't exist it's almost certainly safe!) Jun 12 08:48:37 Luxury patches :-) Jun 12 08:50:33 jbowler-away, how does one pass multiple monut options? Jun 12 08:50:35 mount Jun 12 08:50:48 -o "noatime,notail" ? Jun 12 08:51:47 It's the same as mount - noatime,notail, but in fact it will also combine -o noatime -o notail Jun 12 08:52:16 Because the NFS mount stuff has to prepend nolock etc. Jun 12 08:52:58 ok Jun 12 08:54:37 I'm tempted to add an option to turnup to a pre-mounted system, I'm using separate mounts for /usr and /var at present - seems to be fine except my syslog isn't starting on first boot. Jun 12 08:55:07 <[g2]> jacques, native compile env keeps rolling along... I built a kernel (less IAL) last night, Python 2.4.1 and now I"m building libmxml2 for php5 and apache Jun 12 08:55:27 [g2], did you run the python tests? Jun 12 08:55:55 <[g2]> I've run them before but I didn't run time last night Jun 12 08:56:05 <[g2]> I'll probably run time tonight Jun 12 08:56:18 jacques: I ran them. No crashes, a number of non-found modules (missing diffutils caused the re test to fail), but the basic program looks good to me. Jun 12 08:56:21 <[g2]> I'll setup a script and let it run while I sleep Jun 12 08:56:22 I want to complare notes on them, tho we should get the same results Jun 12 08:56:35 <[g2]> are you running 2.4.1 ? Jun 12 08:56:36 I get a crash on dbm Jun 12 08:56:54 if oe-build bdb is installed at config-time Jun 12 08:57:17 <[g2]> jacques, btw Zope3 and schoolbell ran Jun 12 08:58:58 cool Jun 12 09:00:28 jacques: maybe I should try swapping the whole build over to using db4.3 Jun 12 09:03:21 pb got round to fixing the patch which breaks busybox ifup/ifdown, but I think we're safer without it (his fix didn't break my removal of the patch on openslug) Jun 12 09:04:09 it's still removed on openslug? I thought I built openslug this morning to test the patch for him and it was applied Jun 12 09:06:53 Yes, it has been, weird - have OE pulled back all the nslu2-linux.bkbits.net changes? Jun 12 09:07:39 I'm not sure Jun 12 09:08:52 No, they haven't - there are 612 ChangeSets in nslu2-linux.bkbits.net which are not in oe-devel.bkbits.net Jun 12 09:09:08 Still, apparently an automerge removed my hack to busybox_1.00.bb... Jun 12 09:11:34 Ok, I understand. I'd previously hand-merged to the OE version, so despite the apparently parallel development we match them. Jun 12 15:35:41 <[g2]> morning sunshine ! rwhitby Jun 12 15:38:37 morning Jun 12 16:56:51 Log'd on to brag a bit. Jun 12 16:57:13 go for it, we're listening :-) Jun 12 16:57:17 APEX can read from JFFS2. I'm going to do a bit more testing and implement symbolic links. Jun 12 16:57:30 nice Jun 12 16:57:46 Should be good either today or maybe wednesday. I've got a coupe of days of paying work. Jun 12 16:58:23 * beewoolie-afk celebrates Jun 12 16:58:49 ~celebrate Jun 12 16:59:13 [g2]: hey dood Jun 12 16:59:21 * [g2] does the happy dance Jun 12 16:59:55 BTW, it's kinda big. Turns out to be about 16k of object code. I think it's mostly due to the zlib decompressor. Jun 12 17:00:58 The driver portion is only 3k. Jun 12 17:01:11 <[g2]> well really we've got 128 spent either way Jun 12 17:01:21 I'M COUNTING! Jun 12 17:01:31 <[g2]> that's good Jun 12 17:01:46 We don't want to get into the MS conundrum of requring a memory upgrade to run notepad. Jun 12 17:01:49 :-) Jun 12 17:01:50 <[g2]> but we save 1MB and a partition by booting from jffs2 Jun 12 17:01:57 Indeed. Jun 12 17:02:06 Also, I added a command to show the contents of a directory. Jun 12 17:02:15 [g2]: 1MB? Jun 12 17:02:17 <[g2]> and we can go to dual-ing rootfs's Jun 12 17:02:23 And how. Jun 12 17:02:44 rwhitby: We can put the kernel in the rootfs. The data moves, I suppose it isn't purely a savings. Jun 12 17:02:44 <[g2]> rwhitby, right now with OpenSlug there's a copy of the kernel in the jffs2 Jun 12 17:02:57 Oh. Didn't know there was two copies. Jun 12 17:03:01 <[g2]> yes Jun 12 17:03:03 <[g2]> in /boot Jun 12 17:03:07 [g2]: yeah, there's only a copy if you leave it there :-) Jun 12 17:03:55 <[g2]> rwhitby, correct but it's wasted because you have to boot with it first then delete then use the space Jun 12 17:04:07 <[g2]> it's not terrible, just not optimal Jun 12 17:04:38 <[g2]> Plus will pick up 4 or 5 other block Jun 12 17:04:45 [g2]: yeah, not wishing to detract at all from beewoolie's achievement :-) Jun 12 17:04:46 <[g2]> so that's another .5MB Jun 12 17:05:13 yeah, 1 for ramdisk, 1 for sysconf, 1 for last block .... Jun 12 17:05:21 what are the other 2? Jun 12 17:05:49 (oh, one for smaller bootloader) Jun 12 17:06:00 is APEX still a block smaller than redboot? Jun 12 17:06:13 <[g2]> Isn't the ramdisk currently size at 2 blocks ? Jun 12 17:06:20 nope, just one since 4.x Jun 12 17:06:32 rwhitby: Well, I've been gunning for that, but we still need a place in flash to write user settings. Jun 12 17:07:04 yeah, 128K is a lot of user settings :-) Jun 12 17:07:09 rwhitby: We've currently got one block for the loader and one for settings. Jun 12 17:07:30 rwhitby: I'm planning to implement a scheme to put all of it in one block. It's kinda low priority. Jun 12 17:07:57 <[g2]> so we'll save sysconf, ramdisk fragmentation and trailer Jun 12 17:08:07 Hmm - so changing settings will rewrite the bootloader block? Jun 12 17:08:15 rwhitby: Not realy. Jun 12 17:08:18 <[g2]> no that's the second block Jun 12 17:08:30 rwhitby: The way it works now, the APEX has a block of flash for settings. Jun 12 17:08:53 yeah, but wen you put it all in one block .... Jun 12 17:08:58 it is smart and can append to the existing data in a way that marks the most recent version. Jun 12 17:09:15 rwhitby: The only time it needs to rewrite the block is if the space fills up. Jun 12 17:09:19 <[g2]> right just walk a chain Jun 12 17:09:20 ah, so no erase, just an append - very nice Jun 12 17:09:42 Right. With 128K, I'll be in my grave before I fill that up. Jun 12 17:10:02 <[g2]> beewoolie-afk, any luck with the jffs2 and the reset issue ? You know that theory about the flash being left in a bad state ? Jun 12 17:10:07 I think we could make it 32k and have a special command to perform an environment erase. Jun 12 17:10:19 [g2]: I haven't seen it since. Jun 12 17:10:56 <[g2]> what did you change on the shutdown ? Jun 12 17:11:00 Nothing. Jun 12 17:11:07 I just haven't seen the problem again. Jun 12 17:11:16 <[g2]> ok... Jun 12 17:11:30 <[g2]> it's gotten better on OpenSlug but it's still there Jun 12 17:11:50 It may be worth a shot to just write the 0xff to flash before letting the watchdog take over. Jun 12 17:12:16 [g2]: BTW, thanks for the schematics. Jun 12 17:12:28 <[g2]> np. thanks ep1220 Jun 12 17:12:34 <[g2]> he wanted you to have his e-mail Jun 12 17:15:16 <[g2]> beewoolie-afk, so the reading is working and it's just the layered driver that's holding you back from booting from jffs2 ? Jun 12 17:15:39 [g2]: Not exactly. Jun 12 17:15:51 It boots now. What I'm addng is symlinks. Jun 12 17:15:59 er, support for symlinks. Jun 12 17:16:24 <[g2]> is that up on the FTP site ? Jun 12 17:16:50 <[g2]> symlinks are nice, but certainly no show stopper Jun 12 17:16:50 it is true that the lack of layering means that it is hard to work with since the jffs2 source region has to be ompiled in. Jun 12 17:17:13 Understood. Still, I want to add them since it will break if someone attempts to use them. Jun 12 17:17:31 I'm also thinking of adding a simple command, like mount, to bind a source region to a filesystem driver. Jun 12 17:17:46 And no, it isn't up yet. Jun 12 17:17:59 <[g2]> ok... Thx for all the great work. Jun 12 17:18:02 <[g2]> just checking Jun 12 17:18:22 I'll release it as soon as I get the symlink support in. Beta quality. Jun 12 17:21:52 <[g2]> beewoolie-afk, here's a crazy idea Jun 12 17:22:00 go crazy. Jun 12 17:22:19 <[g2]> If you can read jffs2, what about the possibility of reading the boot loader config data from a file in the jffs2 fs ? Jun 12 17:22:44 <[g2]> then there'd be little need for the config block Jun 12 17:22:47 Well, there will still be something hard-coded somewhere. Jun 12 17:23:00 Also, that means that the loader cannot write the config data. Jun 12 17:24:14 <[g2]> beewoolie-afk, nod. until I teach APEX to write jffs2 :) Jun 12 17:24:15 I can imaging reading a script from jffs2, tho. Jun 12 17:24:42 <[g2]> all kidding aside, I think reading would be great Jun 12 17:24:51 <[g2]> writing is an issue Jun 12 17:24:57 Actually, it wouldn't be hard, but I don't see exactly the benefit. To save a block? In that case, I'd rather do the block sharing. Jun 12 17:25:43 [g2]: you still need a setting somewhere to say which of the two rootfs's to boot from ... Jun 12 17:25:52 that can't really be in the jffs2 Jun 12 17:25:58 <[g2]> THE FINAL ARBITOR! Jun 12 17:26:11 tee hee Jun 12 17:26:12 <[g2]> actually the date stamp or fallback could be the issue Jun 12 17:26:31 <[g2]> just ping-pong between if one fails Jun 12 17:26:38 <[g2]> we had talked about setting the dog Jun 12 17:26:38 As I wrote, it could be useful for other things. For example, a splash screen source file. Jun 12 17:26:56 does APEX have a logo? Jun 12 17:26:57 Boot scripts. Jun 12 17:27:15 rwhitby: Yeah. One of my friends helped me create one. Jun 12 17:27:16 <[g2]> also stuff like finding file in the rootfs Jun 12 17:27:23 <[g2]> that's kinda handy in grub Jun 12 17:27:33 How does grub do that? Jun 12 17:27:47 <[g2]> well grub can read many fs's Jun 12 17:27:50 Can it have a shell-like script? Jun 12 17:27:56 <[g2]> Oh yeah Jun 12 17:28:06 <[g2]> and tab complete on the naming :) Jun 12 17:28:11 [ -f jffs2:/zImage ] && ... Jun 12 17:28:21 yeah, we could. Jun 12 17:28:26 <[g2]> /boot/bz -> Jun 12 17:28:41 Wouldn't really be to difficult. Jun 12 17:28:58 <[g2]> Every 3-4 months I blow the name in the grub.conf file of my kernel Jun 12 17:29:27 <[g2]> then when the boot fails I tab complete or search for the real name to boot the kernel Jun 12 17:30:04 Daiquiri mixing next. Jun 12 17:30:19 <[g2]> beewoolie-afk, THAT's great news about APEX Jun 12 17:30:36 :=p Jun 12 17:31:32 <[g2]> btw, I haven't run samba yet, but I natively built the kernel (less IAL), Python, Zope3, Apache and PHP5 on the FatSlug today Jun 12 17:32:02 Cool. Jun 12 17:32:03 <[g2]> Apache seemed to run pretty quickly for the default pages Jun 12 17:34:03 <[g2]> jacques, ping Jun 12 17:34:17 [g2]: you going to upload those packages to a native feed? Jun 12 17:34:34 share the wealth around? Jun 12 17:34:56 <[g2]> I'll release the recepie for building the native environment first Jun 12 17:35:06 <[g2]> that's what I've been working on Jun 12 17:35:19 <[g2]> it's all pretty straight forward now Jun 12 17:35:47 <[g2]> it's the 2nd or 3rd time I've done it and its much cleaner than the last couple times Jun 12 17:35:59 well, if the recipe works for those, why isn't it up somewhere already to get help with automating it? Jun 12 17:36:17 <[g2]> that's what I'm working on Jun 12 17:36:25 * rwhitby thinks [g2] is going commercial on us :-) Jun 12 17:36:48 (just kidding) Jun 12 17:36:50 <[g2]> I am, but I'm not holding back on you guys Jun 12 17:37:06 <[g2]> I'm gonna sell the DVD Jun 12 17:37:12 <[g2]> $19.99 Jun 12 17:37:21 <[g2]> $9.99 for any devs Jun 12 17:37:25 <[g2]> plus shipping Jun 12 17:37:43 I just want to be invited to the parties. Jun 12 17:37:51 <[g2]> I'll probably torrent a CD that doesn't have the metadata and tarballs Jun 12 17:38:49 <[g2]> It's really an OE build system because all the OE packages probably work Jun 12 17:39:03 <[g2]> I'd toyed with a world build Jun 12 17:39:21 <[g2]> but right now it's OpenSlug and Unslug in OE Jun 12 17:39:27 cool Jun 12 17:39:29 <[g2]> plus the OpenSlug packages Jun 12 17:39:50 <[g2]> I haven't had the time to test the Unslung cross stuff Jun 12 17:39:59 <[g2]> or I'd have all that on there too Jun 12 17:40:01 you should add Unslung packages too Jun 12 17:40:07 (from the CVS) Jun 12 17:40:17 and UpSlug source, etc Jun 12 17:40:30 make it a full project source archive Jun 12 17:40:50 <[g2]> well I'm gonna pull the IAL source before I ship Jun 12 17:41:20 <[g2]> and I'd like to sync up with the binary releases of Unslung and Openslug Jun 12 17:41:54 <[g2]> but really, it's just a kick-butt embedded linux development environ on a DVD Jun 12 17:42:08 <[g2]> I'd like to get APEX on there first Jun 12 17:43:08 <[g2]> but I might wait for APEX until the next release Jun 12 17:43:52 <[g2]> it's also missing the BK tools Jun 12 17:44:02 <[g2]> but that's really the only thing missing Jun 12 17:44:12 <[g2]> other than the IAL source of course Jun 12 18:16:54 FYI: svn repo for openslug is now up to date with BK, due to the Unslung release preparations .... Jun 12 18:21:15 <[g2]> rwhitby, THX is that tagged with 5.5 ? Jun 12 18:32:18 [g2]-away: it's tagged in BK, but not yet in svn Jun 12 18:32:45 <[g2]-away> rwhitby, OK. Jun 12 18:32:55 <[g2]-away> I can do a pull and an export from BK then Jun 12 18:33:20 you can co from svn - it's all there and up to date, just not tagged yet Jun 12 18:33:29 the trunk is Unslung 5.5-beta Jun 12 18:33:48 as soon as I confirm it builds, I will svn copy it to the release branch Jun 12 18:34:33 <[g2]-away> I haven't built unslung since the last test of the DVD Jun 12 18:35:49 <[g2]-away> bbl Jun 12 18:51:09 beewoolie-afk, rwhitby: the way OpenSlug works now it only needs the ethernet id on boot, and it will find it from the last 80 bytes of 'RedBoot' if necessary. Jun 12 18:51:35 SysConf is totally ignored after first boot, never written - there's no point writing it... Jun 12 18:51:47 ok. Jun 12 18:52:00 I didn't think anyone was suggesting that we do. Jun 12 18:52:30 do? Jun 12 18:53:06 Or should. Jun 12 18:53:31 Oh, I see - no I meant it isn't written from OpenSlug. It's only useful if the system has been inited by the LinkSys stuff. Jun 12 18:54:34 I.e. if an unbooted slug is upgraded (reset button/RedBoot on first power on), so the SysConf is still all zero the only interesting thing is the ethernet id, which is in the last 80 bytes. Jun 12 18:55:19 If I can read the ethernet id out of APEX somewhere that's all that is necessary, assuming the kernel boots with /proc/mtd working. Jun 12 19:08:48 jbowler-away: At present, APEX doesn't do anything with the ethernet ID. If we decide to clobber the second block as part of an upgrade, there will have to be an accomodation for the MAC address. Jun 12 19:34:11 I released the version of APEX with jffs2 support. apex-1.2.13 Jun 12 19:34:17 ttfn Jun 12 19:41:53 <[g2]> jbowler-away, ping Jun 12 22:43:33 ka6sox: all good? **** ENDING LOGGING AT Sun Jun 12 23:59:57 2005