**** BEGIN LOGGING AT Sun May 04 02:59:57 2008 May 04 03:10:05 seems like python 2.5 is required for building slugos? May 04 03:12:08 Yes. SlugOS is built using OpenEmbedded which requires python May 04 03:12:49 I have Python 2.4.3 but error message says: ERROR: Importing sqlite3 and pysqlite2 failed, please install one of them. Python 2.5 or a 'python-pysqlite2' like package is likely to be what you need. May 04 03:15:46 I have 2.4.4 May 04 03:16:26 I'm running centos 5, trying to find pysqlite2 as it's not in my extended yum repository May 04 03:17:47 pysqlite-dev? May 04 03:17:57 I'm running Gentoo so can't help with that. May 04 03:19:01 yeah seems like make file is stale ("make setup-host-centos") but perhaps not enough developers use CentOS... May 04 03:19:37 did you try a "yum install python-sqlite"? May 04 03:20:26 famous words of failure (after command run): "Nothing to do" May 04 03:21:07 Well, I'd tell you to get a better distro but that would probably just annoy you more. :p May 04 03:21:24 Ah wait, this works... yum install python-sqlite2 (did that via asterisk at end of sqlite) May 04 03:21:32 heh May 04 03:21:46 Thanks, lemme press on :-) May 04 03:21:53 :) Good luck! May 04 03:22:42 Thanks! May 04 03:25:01 Glad I could be of assistance. May 04 03:25:24 Sure, I'm surely going to have more errors and issues (I'm a hardware developer not software hacker) May 04 03:26:39 Dunno what this is - some error message: ERROR: Openembedded's config sanity checker detected a potential misconfiguration. Either fix the cause of this error or at your own risk disable the checker (see sanity.conf). Please install following missing utilities: help2man May 04 03:26:52 did yum install help2man* but of course nothing grrr... May 04 03:28:50 Think I found it on rpm.pbone.net May 04 03:34:48 Sweet. seems to be working... whew what a mess... "handling BitBake files"... it's in the oven! :-) May 04 03:43:03 Yay, be careful not to burn it. May 04 03:43:45 Thanks, will see how it goes. May 04 03:56:54 CentOS is like pushing a rope. It can be done (I've got one working), but it's really not worth the effort if you can just put a more cutting-edge distro on something. May 04 04:01:54 mwester: agreed, but unfortunately, we use RHEL5 at work and CentOS (for us) is the easiest learning curve. Fedora changes too fast. I tried debian once (and ubuntu - which hides a lot of stuff from you - it seems) but at the time I chose CentOS. Perhaps next OS install I'll try etch or latest Debian 5. May 04 04:02:50 good thing about CentOS is long development cycles. That also hurts it as well when trying bleeding-edge tools and such. May 04 04:03:57 I use Fedora. I also use RHEL and some CentOS for our customers, but I find being behind several years to be frustrating. I use here only because it is sooo bloody old that it fails to recognize the RAID controller on my quad-core motherboard, which crashes the boot on all other newer distros. May 04 04:04:42 ah yeah this is true. Fortunately for me (and my work) we don't have bleeding-edge stuff. That's reserved for the management. *GRIN* May 04 04:04:47 That's a nice OE system (quad-core and 4GB RAM) but CentOS is such a pain that I usually just stick with the old dual-core running Fedora. May 04 04:05:09 yeah with dual-core you want to go fedora no question. May 04 04:05:27 (The quad-core is my wife's system, not provided by work) May 04 04:05:47 Dumb question but with 4GB do you have to use the memory hole technique and 64-bit processor to get around I/O overlap at higher memory space? May 04 04:06:25 I stuck with a 32-bit install, it doesn't see all the RAM (about 3.25GB or so). May 04 04:06:38 You can mess with the BIOS to hide the extra. May 04 04:07:02 (I though CentOS and OE would be difficult enough, much less 64-bit + CentOS + OE!) May 04 04:07:08 Gotcha, I thought it was a windows issue but it's PCI (32-bit) unfortunately. May 04 04:07:23 Heh yeah that's asking for trouble. and don't forget quad core to the brew. May 04 04:08:06 :) it's a Q6600 which is supposed to overclock very very nicely... so it will get more interesting over time. :D May 04 04:09:18 Yeah true, hopefully it doesn't get too hot. Wonder if AMD Phenom is faster than Intel Qxxxx for compiling clock-for-clock. May 04 04:10:35 Clock for clock, I would expect so. May 04 04:12:37 In regards to wiki sandbox not being editable, is it possible that it's related to permissions? You used to be able to "save" the sandbox and view it, and I can't recall if it self-cleared after so many hours or if each user had their own sandbox. May 04 04:15:52 Thanks for your help guys and good chatting. Gotta zzz May 04 15:03:05 03bzhou * r8280 10optware/trunk/make/py-simplejson.mk: py-simplejson: 1.9 -> 1.9.1 May 04 15:04:51 03bzhou * r8281 10optware/trunk/make/pen.mk: pen: 0.17.2 -> 0.17.3 May 04 15:27:42 Question for the software developers here... NPE 2.4 with crypto. If I can get approved for downloading that, will that allow IXP420 crypto access in hardware (for such things as dm-crypt and LUKS)? May 04 15:49:50 RobNC: There is no crypto hardware in the IXP in the NSLU2 May 04 15:54:33 mwester: ok thanks... must be 425? I saw this in the release notes (Intel): The packages required depend primarily on the inclusion of cryptographic acceleration support. (this is with or without crypto) May 04 16:15:06 NM: seems 420 doesn't have the hardware. 422 and 425 do (or can do AES/DES/3DES in h/w). May 04 16:31:06 Thanks mwester - updated wiki with that info. :-) May 04 16:31:38 RobNC, did you mod memory? May 04 16:33:31 ka6sox do you mean obeseslug? May 04 16:42:05 ooyeah May 04 17:01:31 More transfats for slugs! May 04 17:03:19 joshin: right! May 04 17:03:32 thats what I'm talking with him about now. May 04 17:04:09 only problem now is to try and help the linux-arm kernel folks debug this >64MB DMA issue May 04 17:04:55 no kidding...I've read the code and its NOT trivial May 04 17:08:37 Rod suggested something a while back (perhaps last year) regarding patching the kernel to use the 64MB routines when using more than 64MB, unless I'm mistaken May 04 19:31:06 can someone help me with syntax... May 04 19:31:08 svn co [=http://svn.nslu2-linux.org/svnroot/kernel/trunk=] /home/slug/kernel May 04 19:31:14 what's the correct syntax? May 04 19:31:40 says: "svn: URL '[=http://svn.nslu2-linux.org/svnroot/kernel/trunk=]' is not properly URI-encoded" May 04 19:32:40 Where are you getting the [= from? May 04 19:32:53 nevermind, gotta use non-wiki-edited mode (big "d'oh") May 04 19:41:29 lolz May 04 19:43:17 * RobNC shows his multitasking skillz May 04 19:55:43 with Ada.Text_IO; May 04 19:55:44 procedure Hello is May 04 19:55:44 begin May 04 19:55:44 Ada.Text_IO.Put_Line("Hello, world!"); May 04 19:55:44 end Hello; May 04 19:56:03 whoops May 04 19:56:21 just verifying that gnat works. May 04 20:05:27 dumb question but if I've just made "make slugosle" and it has a 16MB variant (for 16MB flash geeks), does that mean it recognizes the higher flash and uses it as JFFS? May 04 20:08:07 another dumb question: if RedBoot is modified to support recognition of 16MB flash, is apex still needed (2nd stage boot loader) to recognize 256MB for slugos? May 04 20:08:46 apex seems a lot more flexible. May 04 20:11:05 that's true, only 50k lost so no biggie May 04 20:11:27 just would be nice to have integrated ethernet with apex, then replace redboot and never look back May 04 20:48:49 Anyone have any experience with indirect PCI memory access when building the kernel? May 04 21:09:00 anyone know where I can get a good working config file for kernel 2.6.24.4 build? looks like the default one has lots of crap in it (i.e., IPSec, IPv6, etc.) that most people don't need or use for SLUGs. May 04 22:05:53 with the build process, is it still necessary to upslug2 an image? I got "not have eRcOmM" which seems to indicate the trailer isn't found? May 04 22:23:17 i use apex 1.5.13 as first stage bootloader on my 128MB fatslug. Now i want change the environment vars, but the env region is unwritable. May 04 22:24:23 After writing the etch-modified-mtdblock2.bin the env region is still unwritable May 04 22:24:27 any ideas? May 04 22:26:54 isn't there some tool called apex-env that helps with that? May 04 22:30:48 i cant run apex-env when booting only from flash (own build image from may 2007 with bootstrap-method) May 04 22:31:07 and booting from hdd crashes May 04 22:31:29 yeah I saw that some time back when I built my own apex. Dunno how to fix that, rwhitby helped with that. May 04 22:33:53 go to wiki page and search for "apex env" - it says to pad with 0xff at end so apex env can be changed May 04 22:34:08 look for compile apex May 04 22:34:58 apex wiki or nslu2-linux.org wiki? May 04 22:35:54 sorry nslu2-linux.org May 04 22:41:12 hmmm... i use apex as FIRST bootloader and this is not on mtd2 May 04 22:42:01 from wiki it seems to be that if you have non-0xFF's apex gets confused and thinks the info is allocated? May 04 22:42:02 and used this apex binary from wiki.buici.com May 04 22:43:37 yeah me too, some time ago, but I ran into same problem as you. Strange! :-) May 04 22:43:47 you May 04 22:43:48 you May 04 22:44:12 you'd think that the apex default compile would zero-pad, or perhaps that's an issue with slugimage program? May 04 22:44:34 ok, i try the Python script May 04 22:45:23 but only what add the 0xFF's and not swap May 04 22:45:50 yeah that is true, perhaps it's not quite as easy as drop/replace apex May 04 22:46:19 oh, file not found May 04 22:46:45 for these i what now a writable env May 04 22:47:10 my old apex cannot do this May 04 22:48:09 prolly easiest way is to use slugimage, I think you can take your old image, disassemble, then replace apex, and re-assemble and upload the image via upslug2 (but I use tftp via console) May 04 22:50:16 think i have not apex in my image-file May 04 22:50:43 the process of changing my whole network around is not trivial so I use console redboot method. Oh whoops... are you replacing redboot (NOT recommended)? May 04 22:51:10 i have replaced 2 years ago May 04 22:51:47 this means "FIRST stage bootloader" = APEX !!! May 04 22:53:31 whoops, that's scarey, b/c if failure, JTAG is required. May 04 22:53:40 i know May 04 22:54:27 i have access to a JTAG programmer May 04 22:54:46 yeah I modded mine as well. Used it many times *GRIN* May 04 22:54:48 but, i dont want use it if possible May 04 22:55:27 Can you simply figure out where the end of Apex is (in memory) then fill the rest of that sector with 0xff's ? May 04 22:55:50 thats my idea now May 04 22:56:42 or is it better to add the FF's to the apex-bin an then reflash this? May 04 22:56:48 and May 04 22:57:14 should be the same, but main issue is to make sure you're running apex out of RAM and not flash. May 04 22:57:58 every time befor flashing i try apex from ram May 04 22:58:06 once I didn't pay attention - running out of flash - and whoops JTAG to the rescue May 04 22:58:33 i know... May 04 22:59:25 how can i add a number of 0xFF's to a file or create a new file with a number of FF's from linux? May 04 22:59:52 isn't there some fill command in apex? May 04 22:59:58 or is that redboot I can't recall May 04 23:00:16 fill command is in apex May 04 23:23:59 dump reports that after apex all is filled with FF's up to 0x1FFFF May 04 23:24:43 sounds right, you had to unprotect, etc. right? May 04 23:24:44 but the first mtd is : 0x00000000-0x00040000 : "RedBoot" May 04 23:25:22 i did no fill command May 04 23:25:46 only reflash the apex.bin an then use dump May 04 23:26:18 weird, that doesn't make sense May 04 23:28:17 and 0x20000-0x3FFFF ? May 04 23:28:38 that's in RAM May 04 23:28:38 this is not FF May 04 23:28:55 no, all "nor:" May 04 23:29:00 thought flash is 0x50000000 May 04 23:34:38 ah nm, I remember apex sees the flash as NOR May 04 23:35:25 is flash, apex is in the beginning and some old parts from redboot are still im flash, i thing because of apex is much smaller than redboot May 04 23:35:48 yeah but apex is in another location than redboot right? May 04 23:37:12 can someone tell me where to find these made during "make slugosbe": ramdisk, fis, kernel, microcode May 04 23:39:49 this is what i want to know too May 04 23:40:15 I need to know this for "slugimage" creation. I have 16MB flash so I can't use old ramdisk, for instance. May 04 23:40:43 Kernel I think I know where that is, and microcode I could get I believe from searching slugos directory during build. May 04 23:47:38 Sinclair73de: it was the debian scripts that didn't 0xFF pad - slugimage has always 0xFF padded May 04 23:48:24 hi rwhitby May 04 23:48:57 i only looked for a script to create a file with 0xFF's May 04 23:49:09 RobNC: Yes, all the 16mb images are for obeseslugs, and depend on having a patched redboot that sees the extra flash. I have such a redboot if you need one. May 04 23:49:40 RobNC: apex is still needed to see extra ram, even with 16mb flash-aware redboot. I did a binary hack, not a recompile. May 04 23:49:49 i found, the env-region is nor:128k+64k and this is not filled with 0xFF May 04 23:50:07 RobNC: I'm also looking at u-boot now for second stage instead of apex, cause u-boot already works with the ixp ethernet. May 04 23:52:44 RobNC: you can get everything by unpacking an existing image, and repacking with slugimage -F option for 16mb flash and -L option for second-stage bootloader May 04 23:53:09 rwhitby: I patched my RedBoot, yours should be the same as mine (per ObeseSlug site). Now I realize the "make slugosbe" didn't actually do what I thought - it isn't built to just drop into flash May 04 23:53:20 RobNC: yes it is. May 04 23:53:33 it will create a 16mb image, which you can upslug16 to the slug May 04 23:53:34 rwhitby: we use uboot at work. May 04 23:53:48 (note that upslug16 is a custom build of upslug2 - the patch is in the repo) May 04 23:53:59 rwhitby: I tried that, but then got the "not find eRcOmM" message May 04 23:54:13 RobNC: Did you flash a 16mb image? May 04 23:54:48 I used instructions for ObeseSlug to tftp via RedBoot the file "slugosbe-4.10-alpha-nslu2-16mb.bin" May 04 23:56:53 hmm - that should work. I always use the upslug16 custom build myself, so I haven't actually tried tftping it May 04 23:57:10 that's the file after "make slugosbe" is done (and passes w/o errors). I updated CentOS wiki (lengthy process!). But shouldn't "slugimage -d -i file" output all the disassembled parts of the 16MB flash, including ramdisk, kernel, loader, trailer, etc.? May 05 00:01:39 yes May 05 00:01:50 make sure you give the -F switch as well May 05 00:02:13 realistically, how long should it take to "make slugosbe"? It took my CentOS 2GHz 1GB RAM machine about 6 hours. That's crazy! May 05 00:02:43 that's about right. May 05 00:02:52 (for the first run). May 05 00:03:14 subsequent runs should be less than 10 minutes depending on what's changed. May 05 00:03:41 rwhitby: thanks yeah that's once I have all the pieces to put together (with -F). Without knowing, does slugimage automatically expand ramdisk.gz to available space (with or without -F)? May 05 00:04:37 yep you're right - just re-ran "make slugosbe" and it took about 5 minutes. May 05 00:07:24 rwhitby: any idea how to make my env-region writeable? Apex reports: env => nor:128k+64k (no-write) May 05 00:08:00 Sinclair73de: depends how you built apex May 05 00:08:20 RobNC: slugimage auto expands the final partition to full space May 05 00:08:57 Sinclair73de: I build mine using the nslu2-linux svn kernel repo makefile, and the config files in there. May 05 00:09:07 Apex 1.5.13 binary from ftp://ftp.buici.com/pub/apex/apex-1.5.13 May 05 00:09:11 rwhitby: thanks, just have to find where the makes puts all the components so I can put that into slugimage May 05 00:09:16 Binaries are at http://ipkg.nslu2-linux.org/feeds/kernel/ May 05 00:09:29 RobNC: look in classes/nslu2_image.bbclass May 05 00:10:29 thanks rwhitby I will update wiki instructions! :-) May 05 00:11:01 RobNC: but you should never need that location, cause you can just unpack a 16mb image May 05 00:11:12 (using slugimage -F -d -i ...) May 05 00:11:34 (or is that '-u' for unpack?) May 05 00:11:43 rwhitby: agreed! However, the image that was built for me seems like it didn't have that. May 05 00:11:53 it's -u for unpack you're right May 05 00:11:57 built by who? May 05 00:12:06 didn't have what? May 05 00:14:09 built by me by make files, didn't have anything other than RedBoot, sysconf and trailer. That's using "slugimage -d -i slugosbe-4.10-alpha-nslu2-16mb.bin" found in slugos/tmp/deploy/images May 05 00:18:42 RobNC: you forgot the -F switch May 05 00:19:35 oh I thought that was only in output but input it's needed too? Hmm. Wonder why I couldn't use that file directly and the ObeseSlug instructions as well, saw that eRcOmM message. May 05 00:20:04 BTW, I just did "make update" and had some issue with certificate from openwrt https host, I chose to accept permanently. May 05 00:20:12 yep May 05 00:22:42 I'll disassemble it again with -F - see what it shows. I should replace with my own config if necessary, but RedBoot was not replaced (as offset starts at 0x600000 I think) May 05 00:23:31 ok, gotta go work now :-) May 05 00:23:44 You are right about -F. Thanks for your help, have a good day! May 05 00:53:40 This is weird - anyone ever seen slugimage run out of flash space? May 05 00:55:19 ok, env is now writable May 05 00:58:47 good job, make sure u document that in a wiki! :-) **** ENDING LOGGING AT Mon May 05 02:59:57 2008