**** BEGIN LOGGING AT Thu Nov 23 02:59:57 2006 Nov 23 03:36:31 any1 awake? i got something fun running on the slug Nov 23 03:36:57 the original unix v7 running under simh Nov 23 03:37:36 ipk and a wiki coming ... Nov 23 03:38:29 simh emulates pdp11 on a slug Nov 23 03:53:27 Now that would be fun! Nov 23 03:53:46 * mwester dates himself with that sentiment... Nov 23 03:55:20 ed hello.c ...; cc hello.c; ./a.out => "Hello World!" Nov 23 03:57:02 eno: Can we add the nbd userspace tools to optware? (http://nbd.sourceforge.net/) I've built an older version natively, seems to work. Nov 23 03:57:17 03bzhou * r4541 10optware/trunk/ (4 files in 3 dirs): add and promoted simh Nov 23 03:57:35 sure Nov 23 03:57:52 Ideally, it would be nice to have separate ipkgs for the server and the client portions -- I'm considering adding the client to the standard rootfs if we have space. Nov 23 03:58:09 As simple as it looks, I've not ever built an optware package... Nov 23 03:59:35 THe version I built successfully was 2.7.8 (http://sourceforge.net/project/showfiles.php?group_id=13229&package_id=173734); perhaps the newer one would work but it complained about not finding the header file nbd.h... Nov 23 04:00:49 does it need any kernel support? Nov 23 04:02:14 I'm not sure; that's the problem. It needs a kernel module (nbd), and to build it includes linux/nbd.h -- I have no idea if nbd.h is in the right spot in the build tree for cross, or if we have to "patch" to add a copy of the 2.4 version in the build directory. Nov 23 04:02:47 (that's not very clear: in order to run, it requires the kernel module. In order to build, it requires the header file. Nov 23 04:03:27 i'll build and let's see Nov 23 04:03:30 I just commited the change to add the kernel module to the feeds; all that's needed is the userspace client tool. Nov 23 04:03:43 k Nov 23 04:04:40 checking where to find a working nbd.h... configure: error: Could not find an nbd.h from 2.6 or above. Nov 23 04:05:23 That's the most recent version, or the older one? Nov 23 04:05:39 most recent Nov 23 04:06:24 ok - I'm not sure if it really needs a 2.6 kenel header file 'cause I'm not sure there's any difference. Let me go check a moment... Nov 23 04:06:48 same thing for 2.8.7 Nov 23 04:07:51 builds/nbd/cliserv.h:#include "nbd.h" Nov 23 04:07:51 builds/nbd/cliserv.h:#include Nov 23 04:08:47 Well, there is a diff in the header file for 2.6 and 2.4 kernels, but nothing that looks like it should make a difference in configure finding it... Nov 23 04:09:06 toolchain/crosstool/build/armv5b-softfloat-linux/gcc-3.3.5-glibc-2.2.5/linux-2.4.22/include/linux/nbd.h Nov 23 04:09:09 toolchain/crosstool/build/armv5b-softfloat-linux/gcc-3.3.5-glibc-2.2.5/gcc-core-prefix/armv5b-softfloat-linux/include/linux/nbd.h Nov 23 04:09:14 toolchain/armv5b-softfloat-linux/gcc-3.3.5-glibc-2.2.5/armv5b-softfloat-linux/include/linux/nbd.h Nov 23 04:09:15 toolchain/armv5b-softfloat-linux/gcc-3.3.5-glibc-2.2.5/armv5b-softfloat-linux/sys-include/linux/nbd.h Nov 23 04:09:33 try copying the header into the local directory before running the configure (that's what might have finally make it work for me) Nov 23 04:10:26 * mwester hopes all those copys of nbd.h in the build tree are identical... Nov 23 04:10:34 they are Nov 23 04:13:36 Hehe - looks like the newest version of the toosl fails with exactly the same error on my Fedora Core 6 system (2.6.18 kernel), so there's a problem with the configure script there. Nov 23 04:14:43 lemme see how to patch the configure script Nov 23 04:30:38 eno: looks like my 2.6 kernel is still using the old header file -- the configure script is correct, and the new source will not build with the 2.4 kernel header file. the older one (2.7.8) configures and builds. Nov 23 04:32:12 k Nov 23 04:47:06 hmm, it is checking NBD_CMD_DISC, an enum only exists in 2.6 nbd.h Nov 23 04:55:50 Yep. I've checked the sources more closely; looks like a specific 2.6.18.xx version switched kernel interfaces. So we need to go with the most recent version on the 2.7 line, which is for the older interface. Nov 23 04:58:51 Gack - it's distro-specific, too. Fedora 5, kernel 2.6.18.1 uses the old kernel interface, Fedora 6, kernel 2.6.18.1 builds with the new interface. That can't be right... but regardless of Fedora problems, we do need the older 2.7.x version for the 2.4 kernel on Unslung. Nov 23 05:01:07 (It *IS* right - Fedora Core 5 & 6, claim the same kernel release number 2.6.18.1, and have two completely different versions of the nbd.h header file!) Nov 23 05:06:32 for compilation, if i copy nbd.h from ixp4xx-kernel-2.6.18-r1.413/linux-2.6.18/include/linux/nbd.h Nov 23 05:07:03 then both 2.8.7 and 2.9.0 build fine Nov 23 05:12:55 Hmmm... but the question is, will the resultant exe work with the older kernel module? Nov 23 05:19:29 Actually, now that I look at it closely, it might well work regardless. I'll go see if I can beat it into submission and build the 2.9.0 version on my native unslung machine. Nov 23 05:36:03 do you think it's ok to commit a 2.6 nbd.h into sources/nbd/ ? Nov 23 05:45:47 I'm not sure. I cannot get 2.9.0 to build natively; no glibconf support. I'll keep hacking away on it, but I can't test to see the more recent client can talk to the old kernel module. Nov 23 05:48:54 Frankly, at this point, I would suggest that we go with the most recent 2.7.x version; the date stamp is in October, so it's fairly up-to-date in terms of bugfixes. The 2.9 adds some support for new features, but mainly on the server side, and I think the use cases would be for the slug to be on the client side. Nov 23 05:51:23 how about i check in a make/nbd.mk and sources/nbd/linux-2.6.18-nbd.h Nov 23 05:51:32 which should cross build fine Nov 23 05:51:56 and you can adjust the nbd.h version, and nbd version Nov 23 05:52:01 and runtime test it Nov 23 05:52:22 Cool; I'll do that. Nov 23 05:53:22 give me a few more minutes for some cleanup Nov 23 05:53:43 will have nbd, nbd-client and nbd-server 3 ipks Nov 23 05:54:06 i would like just two, but there's some limitation in autoclean Nov 23 05:54:36 No hurry at all; I'm bulding unslung right now so it'll be tied up for another 30 minutes building or so... nbd includes both client and server, then? Nov 23 05:55:00 * rwhitby watches the unslung/optware team fire on all cylinders ... Nov 23 05:55:34 nbd is just a virtual package that depends on nbd-client, nbd-server Nov 23 05:56:39 no, i'm not :) Nov 23 05:56:55 sorry, i just got a few highlights :) Nov 23 05:58:58 * mwester thinks that a name clash is inevitable, and thinks we may need to fully-qualify... i.e. "nbd.software.package" vs "nbd.irc.nick" :D Nov 23 05:59:21 lol Nov 23 06:00:39 of course, the downside of that is that shortly the conversations with fully-qualified ids and names would start to look like Java code. Nov 23 06:06:47 03bzhou * r4542 10optware/trunk/ (make/nbd.mk sources/nbd/ sources/nbd/linux-2.6.18-nbd.h): added nbd, cross buildable Nov 23 06:17:09 03bzhou * r4543 10optware/trunk/make/nbd.mk: nbd: removed nbd virtual package, nbd-client/nbd-server works fine Nov 23 06:25:24 * eno start writing the simh wiki page Nov 23 06:26:07 * rwhitby installs debian etch rc1 on devslug ... Nov 23 07:25:19 g'nite Nov 23 07:25:33 happy thanksgiving Nov 23 07:27:00 happy turkey eating day to all the US folk. Nov 23 07:29:03 for me it will be happy try to find a restaurant that's open day Nov 23 07:29:32 or not so happy starving day Nov 23 07:33:13 hmm Nov 23 07:33:20 some special holiday? Nov 23 07:37:44 http://en.wikipedia.org/wiki/Thanksgiving Nov 23 07:38:06 ah Nov 23 16:16:57 03bzhou * r4544 10optware/trunk/make/py-genshi.mk: py-genshi: 0.3.4 -> 0.3.5; added py25 Nov 23 19:18:31 03bzhou * r4545 10optware/trunk/make/whois.mk: whois: update to 4.7.19 and latest template Nov 23 22:36:25 morning Nov 23 23:06:27 rwhitby: Are you contemplating moving slugos over to angstrom for real? Nov 23 23:10:25 I want someone to look at how hard it will be to do so. Nov 23 23:10:48 If angstrom can be configured to do what we need, then there is no point in maintaining a separate distro. Nov 23 23:11:07 I fully agree. Nov 23 23:11:36 it would be a SlugOS 5 timeframe, not a SlugOS 4 timeframe. Nov 23 23:11:41 Too many people start their own distro instead of working towards a common goal with an existing one. Nov 23 23:11:48 understandably. Nov 23 23:12:05 that would give us glibc2.5 which is good for me. Nov 23 23:12:14 what's angstrom? Nov 23 23:12:33 How's the apex/open NPE stuff coming along? Nov 23 23:13:27 I got a student asking me about the slug for his robot project so I'm actively "advertising" slugos. :) Nov 23 23:21:05 03bzhou * r4546 10optware/trunk/make/coreutils.mk: coreutils: 5.97 -> 6.6, latest template Nov 23 23:22:33 nbd: Another openembedded distro similar to slugos Nov 23 23:23:42 VoodooZ: apex/npe is in slugos head right now Nov 23 23:24:57 yeah i read that but wasn't sure if it was completely functional yet. good work. Nov 24 00:13:50 03bzhou * r4547 10optware/trunk/make/py-bittorrent.mk: py-bittorrent: new py-bittorrent-common for py- and py25- coexistence Nov 24 00:23:47 any britons awake? Nov 24 00:24:06 Or anyone who would happen to know if the U.K. uses Annex A or B (xDSL) Nov 24 01:16:04 angstrom builds with nslu2 as the MACHINE type Nov 24 01:16:21 only thing is its bootstrap-image contains everything Nov 24 01:16:31 so it's way too big - over 7MB Nov 24 01:17:02 we might need to modify the kernel config to include EABI calling conventions Nov 24 01:17:55 (this is LE BTW, I haven't tried it BE - that very well might not work - even codesourcery's toolchain doesn't support BE) Nov 24 01:22:07 jacques: so there's some FEATURES thing to reduce what goes into an Angstrom build, right? Nov 24 01:22:44 there's gotta be, haven't checked that yet Nov 24 01:23:03 it includes so much stuff, I'm sure we can get it down to size Nov 24 01:24:19 I'm trying to remember everything I need to do to try it (angstrom) on the lude) Nov 24 01:24:51 my lude us partitioned with 2MB for kernel and like 13MB available for flash disk Nov 24 01:26:03 redboot makes u-boot look simple Nov 24 01:26:37 i like boot loaders that don't do their own partitioning stuff Nov 24 01:27:17 imho a boot loader should just take up the beginning or the end of the flash and launch the kernel from a fixed location at the beginning of the available flash area Nov 24 01:33:43 jacques: my lude is configured roughly the same Nov 24 01:34:22 currently it's like the second layout in http://www.nslu2-linux.org/wiki/NAS100d/FisList Nov 24 01:34:46 but npe_ucode might eventually go into sysconfig Nov 24 01:35:11 (with sysconfig expanded out to 256KB) Nov 24 01:38:40 unfortunately my macbook is booted to OSX (long story) Nov 24 01:38:53 and its linux side has the images I need Nov 24 01:39:43 ah what the heck, I'll reboot it Nov 24 02:02:11 well, that was ... interesting Nov 24 02:03:50 looks like a few issues, first of which is: Nov 24 02:03:58 Node at 0x0001fb9c with length 0x00000b23 would run over the end of the erase block Nov 24 02:03:58 Perhaps the file system was created with the wrong erase size? Nov 24 02:04:05 which I suppose is somewhat likely Nov 24 02:04:41 it ends with this: Nov 24 02:04:46 VFS: Mounted root (jffs2 filesystem). Nov 24 02:04:46 Freeing init memory: 100K Nov 24 02:04:46 Failed to execute /linuxrc. Attempting defaults... Nov 24 02:04:46 /sbin/init: relocation error: /sbin/initKernel panic - not syncing: Attempted to kill init! Nov 24 02:04:46 : symbol __environ, version GLIBC_2.4 not defined in file libc.s o.6 with link time reference Nov 24 02:12:31 I suppose second error could easily be a result of the first Nov 24 02:12:52 back later **** ENDING LOGGING AT Fri Nov 24 03:00:00 2006