**** BEGIN LOGGING AT Tue Jun 07 23:59:56 2005 Jun 08 03:45:55 <[g2]> dyoung, ping Jun 08 03:48:29 hi [g2] were you working on a way to package native openslug builds? Jun 08 03:49:02 <[g2]> jacques, I'm working on several different things Jun 08 03:49:13 <[g2]> native builds is one of them Jun 08 03:50:20 I've got some native builds going on openslug now, and we need a way to package up perl, emacs, etc Jun 08 03:50:34 I'd rather not re-invent the wheel Jun 08 03:51:37 <[g2]> jacques, with the current openslug-packages build I'm working on making all the steps needed to have it work as a simple script Jun 08 03:52:52 <[g2]> basically it should be "bb openslug-image openslug-packages; bb openslug-image;" then on the slug target run a script that does the ipkg update and all the funky steps to make things right Jun 08 03:53:24 are you talking about getting a native devel env? Jun 08 03:53:38 <[g2]> jacques do you run nfs as a server from the slug by chance ? Jun 08 03:53:50 <[g2]> I've got a native devel envir Jun 08 03:53:57 no, I have a thing against nfs and avaoid it like the plague Jun 08 03:54:07 <[g2]> I'm just going to make that easily avialable for everyone Jun 08 03:54:18 I have a native devel env Jun 08 03:54:27 I just want to make ipkg's of perl and emacs etc Jun 08 03:55:43 <[g2]> I had most of perl built Jun 08 03:55:45 are you saying your method of making the packages of native-built apps is to use nfs and do the packaging on the server ? Jun 08 03:56:01 <[g2]> jacques, nfs is totally unrelated Jun 08 03:56:08 ah ok, that confused me Jun 08 03:56:34 <[g2]> beewoolie had asked about nfs server and I ran the client but the server didn't seem to get setup with the nfsd Jun 08 03:57:50 ah, I know there are some nfs experts in our group Jun 08 03:58:04 but I usually tune them out when they are talking about it :-) Jun 08 03:58:50 BTW, I built perl 5.8.7 native on openslug and it tested out Jun 08 03:59:09 building python now (more as a test since it's also in oe) Jun 08 03:59:26 <[g2]> I built python natively before Jun 08 03:59:40 <[g2]> I ran a zope3 app on the slug -- schoolbell Jun 08 03:59:52 <[g2]> it worked but was really slow due to the memory issue Jun 08 04:00:05 <[g2]> I'm gonna run it again on the 64MB slug Jun 08 04:01:00 yeah that's gotta be sweet - 64MB Jun 08 04:02:36 <[g2]> On the nfs client tests of big files last night it pulled 60-65Mb across the net Jun 08 04:02:54 <[g2]> I think it could pull more in polled mode Jun 08 04:03:56 I'm confused about if we're currently in interrupt mode, why is the load still 1.0 at idle? Jun 08 04:04:33 <[g2]> someone was saying that diskwait counts as busy Jun 08 04:04:44 <[g2]> and a couple of the tasks are in diskwait Jun 08 04:05:32 are these tasks CSR-related ? Jun 08 04:05:52 because in the past when I booted without the CSR modules loaded, the load was 0.0 as expected Jun 08 04:06:11 (I dunno if the modules are rmmod-able - they didn't used to be) Jun 08 04:08:11 <[g2]> jacques, they exit and ugly death but I think it works Jun 08 04:10:59 OK I'll try it sometime soon Jun 08 09:42:17 new openslug build finished :) Jun 08 10:34:20 Hi [g2] Jun 08 10:34:32 <[g2]> jbowler, howdy! Jun 08 10:35:05 I got the 'update ffs2' stuff to work yesterday (i.e. I did an in-place overwrite of /dev/mtdblock4 and it all worked) Jun 08 10:35:18 <[g2]> AWESOME! Jun 08 10:35:27 <[g2]> it's with devio ? Jun 08 10:35:29 Here's the command: devio '<>/dev/mtdblock4;cp$;fb#$-255' (all seems to work fine - jffs2 is the name of the deploy/image file) Jun 08 10:35:50 So just replace 'jffs2' with the input file. Jun 08 10:36:10 There's the start of a manual on http://devio.sourceforge.net Jun 08 10:36:17 * [g2] types away on Fatslug Jun 08 10:36:27 <[g2]> Congrats on splitting that out too Jun 08 10:36:36 <[g2]> I think that will be useful to many projects Jun 08 10:37:01 <[g2]> does it matter which version of devio I've got ? Jun 08 10:37:21 I haven't changed the code for weeks ;-) Jun 08 10:38:23 I'm guessing that the fill with 0xff is important for some reason. Jun 08 10:39:00 <[g2]> Ok the command completed Jun 08 10:39:31 [g2]: You're working on making a native dev env easy? Getting any further? :D Jun 08 10:40:50 <[g2]> NAiL, yes I've had a running native dev environment working Jun 08 10:41:03 <[g2]> I just need to cleanup the proceedure Jun 08 10:41:18 <[g2]> and relearn some stuff that I did last time :) Jun 08 10:42:32 [g2]: Brilliant. I've been missing that since I switched away from my uClibc openslug build. Jun 08 10:42:43 <[g2]> jbowler, Congrats! That looked like it worked Jun 08 10:43:20 the devio stuff is basically for reflashing without redboot? Jun 08 10:43:43 <[g2]> devio is just a general tool Jun 08 10:43:56 <[g2]> it can be used to reflash the jffs2 partition Jun 08 10:44:01 It's for reading and writing from the flash (or any block device), it's used to extract the sysconf info too. Jun 08 10:44:07 aha Jun 08 10:44:22 <[g2]> jbowler, could you explain that full command to me ? Jun 08 10:44:41 Documentation is currently limited by the total absence of a non-useless Linux web page editor, but: Jun 08 10:45:13 << jffs2 --- open file 'jffs2' for input Jun 08 10:45:27 <[g2]> ; ends a command Jun 08 10:45:33 >> /dev/mtdblock4 -- open file (device) /dev/mtdblock4 for output Jun 08 10:46:12 (yes: lexical syntax is command arguments end-of-line. command is two characters, arguments are the rest of the line, eol is ; \n or \0) Jun 08 10:46:50 cp $ --- '$' is the length (total size) of the input, so copy bytes to output Jun 08 10:47:29 <[g2]> I guess ; is a statement separator Jun 08 10:47:32 fb $#- 255 -- fill (#$-) bytes with the value 255. # is the total size of the output, RPN so #$- evaluates (#-$) Jun 08 10:48:45 <[g2]> that's a little clearer Jun 08 10:49:46 <[g2]> so devio '<>/dev/mtdblock2;cp$;fb#$-255' would update the kernel correct ? Jun 08 10:50:03 <[g2]> actually Jun 08 10:50:26 <[g2]> so devio '<>/dev/mtdblock2;cp$;fb#$-255' would update the kernel correct ? Jun 08 10:50:42 <[g2]> devio follows symlinks right ? Jun 08 10:52:27 <[g2]> jbowler, I'm guessing devio doesn't have a binary compare operator Jun 08 10:54:02 The kernel needs the header, so the command line which works is: devio '<>/dev/mtdblock2;wb$16+4;fb12,0;cp$' Jun 08 10:54:30 <[g2]> Oh... yeah... I was too excited and forgot about the header :) Jun 08 10:54:36 That command doesn't do the fill. fb #t- 255 should work ('t' is the current output pointer) Jun 08 10:54:47 <[g2]> although..... the would work for me if the APEX Jun 08 10:55:27 devio just uses open(2) (it's actually next to impossible not to follow a symlink - you have to parse the path yourself and do readlink(2) on every element...) Jun 08 10:55:46 <[g2]> cool Jun 08 10:55:50 <[g2]> just checking Jun 08 10:56:40 Binary compare is '=' - compares the two top elements of the stack. You can open the output file for input after the open for output to read stuff from it. Jun 08 10:56:40 <[g2]> Ok so I'm getting extremely close to a Linksys Free slug Jun 08 10:57:34 <[g2]> So what exactly would be the command to verify the mtdblock up to the lenght of the file is identical with the file ? Jun 08 10:57:45 To verify the write just repeat it with '-v' - that causes devio to error out if the write would change anything. Jun 08 10:57:52 <[g2]> my RPN is quite rusty :) Jun 08 10:58:43 So in the jffs2 case, if you don't care about the fill (but I think you probably do ;-) just use '-v' and leave the fb#$-255 off Jun 08 10:59:36 <[g2]> no the fill is important, It's just that verifying the fill isn't as important as the non-fill bytes Jun 08 10:59:58 Or, if you don't trust the kernel, repeat the write with -v but change /dev/mtdblock4 to /dev/mtd4... Jun 08 11:00:12 <[g2]> I think a fb#$-000 would "work" but would generate some errors from the jffs2 mount Jun 08 11:01:10 It's a simple test (since it's worked once with 255) - I suspect it won't boot. Jun 08 11:04:41 <[g2]> the other MAN has just arrived :) Jun 08 11:05:12 <[g2]> wb beewoolie Jun 08 11:05:43 <[g2]> hey... prpplagu in openjtag missed you :) Jun 08 11:10:09 <[g2]> jbowler-away, thx for the update! Jun 08 11:10:27 <[g2]> and pls keep me posted on the preserve/restore Jun 08 11:10:56 <[g2]> I'm looking at the reboot/shutdown issue now and I'm sure beewoolie will have some great insights Jun 08 11:12:35 <[g2]> jbowler-away, one other question. sysconfsetup uses devio to search the SysConf mtdblock correct ? Jun 08 11:13:14 <[g2]> hmmm... Jun 08 11:13:40 <[g2]> *actually*.... I really like this update mechanism Jun 08 11:14:16 <[g2]> becuse one could mount the jffs2 after flashing it and copying the /etc/default/sysconf data and keys etc..... Jun 08 11:16:56 [g2]: yes, sysconfsetup uses devio to read SysConf. What I will do is make it read RedBoot if SysConf isn't there (for recovery) and, if SysConf is there, it will be possible to restore the configuration files from it. Jun 08 11:17:58 <[g2]> jbowler-away, The two thoughts I had are: Jun 08 11:18:01 Restoring the files immediately after a reflash is potentially dangerous and unnecessary - sysconfsetup needs to do the correct thing anyway. Jun 08 11:18:41 <[g2]> danerous how ? Jun 08 11:19:45 It uses the old intelligence - an upgrade to something in sysconfsetup won't be known about, so it may restore information incorrectly when the sysconfsetup on the new flash can get it right. Jun 08 11:20:39 E.g. suppose the name of the interface changes from eth0 for some reason... Jun 08 11:21:22 <[g2]> jbowler-away, That's OK I think. Jun 08 11:21:35 <[g2]> The person that just flashed the jffs2 should know that. Jun 08 11:22:25 <[g2]> I'm talking about going linksys free with APEX and eliminating the sysconf stuff or moving the mtdblock1 Jun 08 11:23:26 <[g2]> So it would look like Block 0 (APEX), (Block 1 -- APEX conf + Sysconf info + maybe partition table), Block 2-$ jffs2 Jun 08 11:23:57 <[g2]> or Block 2+1MB kernel and rest jffs2 Jun 08 11:27:49 If you know what you are doing it's possible to save/flash/restore, without the need to write the saved information anywhere other than /tmp. Jun 08 11:28:34 I will probably write save/restore as a separate script anyway - putting it inside turnup will make that script too big. Likewise for reflashing the partitions. Jun 08 11:31:16 <[g2]> jbowler-away, separate is fine. Right now I think externally updating the /etc/default/sysconf will work Jun 08 12:57:34 <[cc]smart> is there a docu available for the new switchbox replacement ? Jun 08 13:07:37 <[g2]> [cc]smart, for OpenSlug ? Jun 08 13:08:25 <[cc]smart> yes Jun 08 13:08:51 <[cc]smart> looking at the code i assume it's meant to replace /linuxrc with a two-liner shell script Jun 08 13:09:12 <[cc]smart> calling /boot/disk sda1 -o reiserfs (in my case that is) Jun 08 13:09:21 <[g2]> cool :) Jun 08 13:09:41 <[cc]smart> would be good to have a some info on the wiki Jun 08 13:09:47 <[g2]> the latest version of turnup is pretty powerful although I haven't used it other than memstick Jun 08 13:09:52 <[cc]smart> could do that if i knew my way is the one it's meant to Jun 08 13:10:14 <[g2]> I've been wanting to try turnup nfs Jun 08 13:10:21 <[g2]> I'll be trying turnup disk next Jun 08 13:10:49 <[cc]smart> well, seems what i did didn't work anyways :/ Jun 08 13:11:14 <[g2]> So turnup with memstick/disk/nfs seem to have much of the capabilities Jun 08 13:11:14 <[cc]smart> memstick is exactly what i'm trying Jun 08 13:11:24 <[g2]> I've used memstick a bit Jun 08 13:11:35 <[g2]> when was your last pull ? Jun 08 13:11:41 <[cc]smart> you don't use the disk script for that ? Jun 08 13:11:51 <[g2]> no Jun 08 13:11:55 <[g2]> I don't Jun 08 13:12:26 <[cc]smart> how do you go at it ? Jun 08 13:12:30 <[g2]> are you trying to run reiserfs on the memstick ? Jun 08 13:14:10 <[cc]smart> yes Jun 08 13:14:25 <[cc]smart> it's in the base modules set now Jun 08 13:14:48 <[g2]> I probably won't run reiser on a memstick Jun 08 13:15:01 <[cc]smart> for the wear ? Jun 08 13:15:10 <[g2]> yes the updates Jun 08 13:15:22 <[cc]smart> what fs are you comparing to ? Jun 08 13:15:25 <[g2]> I probably shouldn't be running ext3 but that's what I've used Jun 08 13:15:25 <[cc]smart> ext2 ? Jun 08 13:15:35 <[g2]> I do the following Jun 08 13:15:41 <[g2]> mke2fs -j /dev/sda1 Jun 08 13:15:44 <[cc]smart> for ext3 i wouldn't expect much of a difference. Jun 08 13:15:59 <[cc]smart> that should make it ext3 Jun 08 13:16:04 <[g2]> turnup -i memstick /dev/sda1 -t ext3 Jun 08 13:16:54 <[cc]smart> ha, now it works Jun 08 13:17:02 <[cc]smart> just had a typo Jun 08 13:17:06 <[g2]> or maybe turnup memstick -i /dev/sda1 -ext3 Jun 08 13:17:18 <[cc]smart> what's that turnup script, should i use it ? Jun 08 13:17:52 <[g2]> jbowler rewrote my simple turnup script and really exteneded the functionallity Jun 08 13:18:01 <[g2]> spelling not withstanding Jun 08 13:19:46 <[cc]smart> what the heck. help goes straight into detail, while i don't know what it's about upfront :) Jun 08 13:20:32 <[cc]smart> ah, so turnup does the configuration Jun 08 13:21:14 <[g2]> it copies the jffs2 files to the external device and sets stuff up for subsequent reboots Jun 08 13:22:20 <[cc]smart> seems pretty hughe for that :) Jun 08 13:22:47 <[g2]> well there are lots of comments in there :) Jun 08 13:22:48 <[cc]smart> did a >find / -mount | cpio -p /mnt Jun 08 13:23:45 <[cc]smart> i just missed the main comment: this tool is provided for convenience in setting up an external root filesystem Jun 08 13:24:06 <[cc]smart> and the knowledge of its existence certainly Jun 08 13:24:31 <[cc]smart> when you guys implement nifty things, it owuld be good to know it's there before :) Jun 08 13:25:50 <[cc]smart> seems i should add a little wiki on the weekend Jun 08 13:26:48 <[g2]> well jbowler has been very busy implementing Jun 08 13:27:00 <[cc]smart> anyways, need to go to bed now. had a little success i can dream about tonight. it's 10:30 P now here and i need to get up at 4A tomorrow, so g'night Jun 08 13:27:14 <[g2]> congrats! Jun 08 13:27:22 <[g2]> sweet dreams :) Jun 08 13:27:29 <[cc]smart> i'll try to gather what i find and get sth into the wiki Jun 08 13:27:33 <[cc]smart> on the weekend Jun 08 13:27:51 <[cc]smart> cu Jun 08 13:28:11 <[g2]> nod **** ENDING LOGGING AT Wed Jun 08 23:59:56 2005