**** BEGIN LOGGING AT Thu Aug 04 23:59:56 2005 Aug 05 00:30:39 jbot, test Aug 05 00:30:40 Test Failed! Aug 05 01:15:48 anyone notice any problems with the ipkg feeds transition? Aug 05 01:23:45 not any longer Aug 05 01:31:47 what are the pre-requisites for building wl500g packages? Aug 05 01:32:00 I've got /opt/brcm Aug 05 01:32:08 morning Aug 05 01:32:25 discovered a dependency of the Makefile which isn't in a default gentoo Aug 05 01:32:47 and isn't psyco, ccache, monotone, subversion or cvs Aug 05 01:33:11 blaster8: feel free to update the wiki Aug 05 01:33:37 I'm curious as to what it is though Aug 05 01:34:48 unzip ;) Aug 05 01:38:17 haha Aug 05 01:38:26 ok, easy to forget ;) Aug 05 01:38:43 what package is .zip? Aug 05 01:38:53 ah. Aug 05 01:38:56 the linksys stuff Aug 05 01:39:17 http://www.intel.com/design/network/swsup/ixp400AccessLibrary-1_4.zip Aug 05 01:39:19 for me Aug 05 01:39:29 and the intel stuff, yes Aug 05 01:39:33 (building ucslugc) Aug 05 01:39:46 NSLU2_V23R29.zip ixp400AccessLibrary-1_4.zip ixp400LinuxEthernetDriverPatch-1_1.zip Aug 05 01:40:35 what do I run to restart a half-way through build? Aug 05 01:40:43 just rerun build ucslugc? Aug 05 01:41:08 yeeah Aug 05 01:41:14 it'll continue where it left off Aug 05 01:41:57 just got a 10GB Laptop hard drive through the post Aug 05 01:42:13 I'm taking apart the NSLU2 and remounting it in an Aluminium 5.25" Drive Case Aug 05 01:42:23 neat :) Aug 05 01:42:39 Adding a few bits and pieces like a USB-Serial Adaptor and leaving it as a Web/Email server Aug 05 01:42:50 do take pictures :) Aug 05 01:43:13 What USB to RS232 adaptors are known to work well in 2.4 kernels? Aug 05 01:43:18 the only thing is that I just found a USB2 - 10/100 Ethernet adaptor and am wondering if I need another firewall Aug 05 01:43:39 peter: I think the Prolific adaptor is the most common Aug 05 01:43:51 uh Aug 05 01:44:04 pl2303 isn't always working without a patch in 2.3 Aug 05 01:44:06 2.4 Aug 05 01:44:23 mm, thought so Aug 05 01:44:25 yeah, I recall prolific mentioned a few times, but usually with complaints. Aug 05 01:44:39 hehe Aug 05 01:44:43 those are the ones Aug 05 01:44:52 Theres like 2 revisions of the pl2303 Aug 05 01:45:01 we have patched the dirver towork with both. Aug 05 01:45:06 but NOT the PL2303X Aug 05 01:45:39 I've got two pl2303's and both work like a charm. But I'm running 2.6 Aug 05 01:45:52 I will be running 2.6 Aug 05 01:46:13 Don't mind losing the default Linksys firmware as I have no need for Samba Aug 05 01:46:19 peteru-home: the other one that usually "just works" isi the Keyspan US19HS Aug 05 01:46:39 blaster8: openslug's samba works great now ;-) Aug 05 01:46:40 the added benefit to the 19HS is you can use it as a cheap Dev Platform for the part inside. :-) Aug 05 01:47:03 (and ucslugs samba as well) Aug 05 01:47:13 I'll actually need to use it with a Netgear wgt634u and that's stuck in 2.4.20 land. Aug 05 01:47:16 no windows machines here Aug 05 01:48:05 blaster8: USB-ethernet? great, then you can run Debian on it ;) Aug 05 01:48:27 the debian install was a great one-man job Aug 05 01:48:37 figuring out the byteswapping was impressive Aug 05 01:48:52 but still, I'm not that attached to Debian Aug 05 01:49:12 blaster8: well, it's quite obvious once you think about it (most things are ;) Aug 05 01:49:38 True - to be fair, if I could run OpenBSD on this thing I would though Aug 05 01:49:57 blaster8: well, just get started porting then ;) Aug 05 01:50:29 you'll probably have more luck getting netbsd running though Aug 05 01:51:04 well, openbsd has a good record on running on embedded x86 systems without patches Aug 05 01:51:11 but arm is probably a step too far Aug 05 01:51:27 netbsd runs on anything including that darn kitchen sink :P Aug 05 01:51:29 blaster8: there is support for a strongarm board though Aug 05 01:51:30 dyoung: the keyspan looks neat. Now I just need a DB9 null modem converter to put on the dongle. Aug 05 01:51:33 http://www.openbsd.org/cats.html Aug 05 01:51:46 blaster8: note the setup-debian-host target in the makefile - feel free to suggest a setup-host-gentoo target ... Aug 05 01:52:01 and the zaurus Aug 05 01:52:14 peteru-home: you need to setup the toolchain from wl500g.dyndns.org Aug 05 01:52:42 Jacmet: but of course look at the note: OpenBSD/cats was started based on NetBSD-current Aug 05 01:52:50 blaster8: yeah Aug 05 01:53:28 for the BSDs, it's all about borrowing security code from OpenBSD, borrowing portability code from NetBSD and borrowing well-optimised code from FreeBSD Aug 05 01:53:34 but it seems to work for them Aug 05 01:53:58 NAiL: rather than suggesting to update the wiki for the host requirements, we need to add it to the makefile Aug 05 01:54:00 sharing is good Aug 05 01:54:27 and add setup-host-foo targets for other distros as people come work it out Aug 05 01:54:31 rwhitby: I did tweak the wiki - feel free to change it back if you need to Aug 05 01:54:37 rwhitby-away: oh... Aug 05 01:54:48 blaster8: that's fine - we just need to take that info into the makefile too Aug 05 01:55:02 and that will require a gentoo developer to do it :-) Aug 05 01:55:06 rwhitby-away: the setup-debian-host works for ubuntu, if you u se the universe feed. Aug 05 01:55:30 cool Aug 05 01:55:39 bbiab Aug 05 01:55:44 the setup-debian-host stuff will be easy to 'port' to gentoo Aug 05 01:56:01 all you have to do (practically) is change apt-get to emerge Aug 05 01:56:05 and wait much longer :) Aug 05 01:56:09 and remove "install" Aug 05 01:56:29 remove the -dev packages, I think they're installed automatically Aug 05 01:56:47 dunno if the python names are the same on getnoo Aug 05 01:56:48 yeah - and you shouldn't really need to fetch python - it's already there in any gentoo (Portage needs Python) Aug 05 01:56:55 yeah Aug 05 01:57:09 do you need glib and libc6 for building ucslugc? Aug 05 01:57:51 does anyone here use 'ddclient' for dyndns updating? Aug 05 01:59:49 hmmm, so oe can not build a working uclibc mipsel toolchain? Aug 05 02:00:39 or to phrase it another way, should you grab uClibc to setup debian-host? Aug 05 02:04:01 rwhitby-away: any reason the setup-debian target doesn't install monotone/svn? Aug 05 02:05:23 I guess it might install the wrong version of monotone Aug 05 02:06:20 gentoo apparently still has 0.19 Aug 05 02:06:59 there, target made Aug 05 02:08:41 I'm running monotone 0.19 on gentoo now Aug 05 02:08:44 no real problems Aug 05 02:09:05 NAiL: Sudo isn't really suitable for gentoo Aug 05 02:09:24 as it isn't always installed or configured - use 'su -' instead Aug 05 02:09:38 peteru-home: it probably can, but no-ones verified that it works. Mainly cause the version of uClibc in OE is newer than the one on the wl500g Aug 05 02:10:12 blaster8: you would not be able to pull from monotone.nslu2-linux.org if you were runnin 0.91 Aug 05 02:10:14 03nail 07org.nslu2-linux.dev * r2bc14800... 10/Makefile: Add setup-host-gentoo target Aug 05 02:10:14 0.19 Aug 05 02:11:02 I have some warnings but I didn't see it not working Aug 05 02:11:15 'monotone: warning: ignoring unknown signature by 'nslu2-linux@nslu2-linux.org' on '[branch@a4031fea83dade173313d0b051be89e17141bdb7:b3JnLm5zbHUyLWxpbnV4LmRldg==]'' Aug 05 02:11:52 well, anyway, the build continued Aug 05 02:12:06 what is the minimum monotone version required? Aug 05 02:12:25 0.20, 0.21 if you want it to be a bit faster and not have as many bugs Aug 05 02:12:33 the server is running 0.21 Aug 05 02:12:46 bb in about 5 minutes Aug 05 02:12:56 Its not possible to pull a 0.20 database using 0.19 because the version of netsync is differnt. Aug 05 02:13:41 this might be a newbie question, but which script do i edit so that a daemon process is called when the NSLU2 boots up? Aug 05 02:16:54 is it /etc/inittab ? Aug 05 02:18:28 Airw0lf: meet you in #nslu2-general Aug 05 02:19:32 03nail 07org.nslu2-linux.dev * rd95cdcc1... 10/Makefile: Switch to su instead of sudo on Gentoo, correct svn -> subversion Aug 05 02:22:37 macsat: have you used the master makefile structure yet? Aug 05 02:25:10 nope - I havent gotten around to looking at it yet Aug 05 02:25:32 03nail 07org.nslu2-linux.dev * rdeacaab5... 10/Makefile: Make setup-host-gentoo skip already installed packages. emerge is slow enough already ;-) Aug 05 02:26:19 NAiL - have you specified monotone as testing in package.keywords? Aug 05 02:26:34 otherwise emerge will install monotone 0.19 Aug 05 02:27:46 I haven't set up monotone in the Makefile at all actually Aug 05 02:27:54 ? Aug 05 02:28:02 rwhitby, to be honest - all I have done witht he cross compiling stuff, is what is described in my Cross Compile Tutorial.....hence I have more or less only setup the environment, never really got around to making any packages myselves :-( Aug 05 02:28:05 I was considering the best way to do it Aug 05 02:28:22 I can tell you the gentoo recommended way Aug 05 02:30:19 mkdir /etc/portage; touch /etc/portage/package.keywords; echo "dev-util/monotone-0.21 ~x86" >> /etc/portage/package.keywords Aug 05 02:30:26 then emerge monotone Aug 05 02:30:38 oh yeah, su to root b4 that stuff Aug 05 02:30:48 and exit afterwards Aug 05 02:31:03 hmmm Aug 05 02:31:37 that way, when monotone 0.22 (hypothetical) becomes stable emerge will automagically prefer that version Aug 05 02:32:06 hmm, never search for the man page for touch on google Aug 05 02:32:12 note to self Aug 05 02:32:54 oops, one mistake Aug 05 02:32:59 haha Aug 05 02:33:05 su -; mkdir /etc/portage; touch /etc/portage/package.keywords; echo "=dev-util/monotone-0.21 ~x86" >> /etc/portage/package.keywords Aug 05 02:33:10 su - -c "mkdir -p /etc/portage ; echo >> /etc/portage/package.keywords ; echo dev-util/monotone-0.21 ~x86 >> /etc/portage/package.keywords" Aug 05 02:33:15 there we are, the definitive version Aug 05 02:33:16 ah, = Aug 05 02:33:38 no, yours will not add a newline before, if the file already exists (like it did on my testbox Aug 05 02:33:46 which screwed up the file :P Aug 05 02:34:01 ok - I'm not that great at Linux Aug 05 02:34:34 thanks though Aug 05 02:35:42 ugh Aug 05 02:35:43 All ebuilds that could satisfy ">=dev-libs/boost-1.32" have been masked. Aug 05 02:36:39 but not hard masked Aug 05 02:36:48 so we can override them in the same way, I guess Aug 05 02:37:11 but we probably should put in a warning somewhere Aug 05 02:37:38 /etc/portage/package.keywords Aug 05 02:37:59 macsat: have a look at www.nslu2-linux.org/Makefile, as it may make your tutorial steps easier now Aug 05 02:38:33 ohh ok - so you have changed the ways ? Aug 05 02:38:51 nothing too important depends on boost - at least as far as portage knows Aug 05 02:38:51 http://www.gentoo-portage.com/dev-libs/boost/rdep Aug 05 02:39:42 hmm Aug 05 02:39:59 is there anything I can do to switch off vi's autoindent stuff? Aug 05 02:42:46 monotone depends on boost. Aug 05 02:43:55 no, so upgrading it won't kill anything important ;) Aug 05 02:44:18 got it NAiL Aug 05 02:44:23 blaster8: No need to use su or to poke things into /etc/portage Aug 05 02:44:42 use ~/dev-libs/boost-1.32.0 ~x86 Aug 05 02:44:53 peter: ? Aug 05 02:45:13 why not just prefix emerge with USE="~x86" ? Aug 05 02:45:25 because that behaviour is not recommended Aug 05 02:45:36 this is much more controlled - only enables very specific versions Aug 05 02:45:42 and so is less likely to break in future Aug 05 02:46:27 the ~/dev-libs/boost-1.32.0 ~x86 command will unmask all 1.32.0 versions - so if they come out with 1.32.0-r5 then that will be emerged instead automatically Aug 05 02:47:12 su - -c "mkdir -p /etc/portage ; echo >> /etc/portage/package.keywords ; echo ~dev-libs/boost-1.32.0 ~x86 >> /etc/portage/package.keywords ; echo dev-util/monotone-0.21 ~x86 >> /etc/portage/package.keywords" Aug 05 02:47:17 ~x86 is not really right either, it could be ~mips ;) Aug 05 02:47:17 ...but x86 is already something else... Aug 05 02:47:30 jbot: good point Aug 05 02:47:52 ~x86 Aug 05 02:47:52 x86 is probably this frankinsteins monster of a system compaired to arm. Aug 05 02:47:58 but to be fair - anyone doing this on ppc or others would be slightly mad Aug 05 02:47:59 ~arm Aug 05 02:48:00 somebody said arm was both big-endian and little-endian. hh-linux is little-endian Aug 05 02:48:23 blaster8: people are known for trying ;-) Aug 05 02:48:41 NAiL: I know - I'm compiling the whole thing inside Virtual PC for Mac Aug 05 02:51:01 ok Aug 05 02:51:04 does ~* work instead Aug 05 02:52:12 no idea Aug 05 02:52:42 apparently it does Aug 05 02:54:42 03nail 07org.nslu2-linux.dev * r4e7f053e... 10/Makefile: Add monotone and boost in package.keywords if they're not there already Aug 05 02:55:56 ok, changed x86 to * Aug 05 02:57:21 I am tempted to install a fresh gentoo on this box and see if it works from a clean slate Aug 05 02:57:49 but that means .... no development for like... two hours Aug 05 02:58:16 Hey nail... Aug 05 02:58:40 Do you happen to know offhand what your mirror's uplink speed to the Net is? Aug 05 02:58:57 NAiL: Set up a qemu vm Aug 05 02:59:20 emerge qemu-softmmu with USE=kqemu Aug 05 02:59:38 03nail 07org.nslu2-linux.dev * r03d1e61f... 10/Makefile: Change ~x86 to ~*, so that it'll work on other ARCHs Aug 05 03:00:18 dyoung: 8mbit, why? Aug 05 03:00:34 apparently using ~1mbit right now Aug 05 03:01:23 Just wondering how fast our mirrors really are. Aug 05 03:06:02 argh Aug 05 03:07:01 the offshore world championships are in Trondheim today Aug 05 03:07:34 the trucks with the boats are driving past Aug 05 03:07:43 and they're honking their horns continously Aug 05 03:07:49 Scantily Clad Women are about? Aug 05 03:07:55 always Aug 05 03:08:05 As I said, you should be here in the summer ;-) Aug 05 03:13:20 I look forward to moving this autumn Aug 05 03:13:34 Then I'll be able to setup a fast mirror for you ;) Aug 05 03:15:44 * NAiL sneaks in a server @ kolla's work Aug 05 03:17:10 It's annoying to know that less than 100 meters away, there's a 2.5gbit link. And here I'm sitting on my ~4mbit line Aug 05 03:19:48 I wonder if they notice it if I dig a tunnel and latch on to their pipe Aug 05 03:46:52 NAiL: error in Makefile, line 224 Aug 05 03:48:53 hmm Aug 05 03:48:57 what line is 224? Aug 05 03:49:56 Makefile:224: *** commands commence before first target. Stop. Aug 05 03:50:26 the flex line has whitespace at the end Aug 05 03:50:42 you can't have whitespace after the continuation backslash in a makefile Aug 05 03:51:01 I know, wonder where it came from Aug 05 03:51:44 fixed Aug 05 03:54:29 03nail 07org.nslu2-linux.dev * r029a1ca5... 10/Makefile: Remove erroneous whitespace Aug 05 03:56:48 anyone got ascii logos for openslug, openembedded, unslung.. ? :) Aug 05 03:57:25 NAiL: how did you run "make push-master" without a broken Makefile? Aug 05 03:59:43 cool - ccache is at work speeding up by build of boost & monotone Aug 05 03:59:47 v handy Aug 05 04:16:16 can someone look at the Master Makefile and tell me what I'm doing wrong in the autobuild target with the $errors stuff? Aug 05 04:17:28 rwhitby: The blank line ? Aug 05 04:17:54 no, the autobuild target Aug 05 04:27:50 rwhitby: not that I know what the problem is, but wouldn't it be nice set errors += buildtarget ? Aug 05 04:28:31 yeah, good idea. BTW, how did you run "make push-master" with a broken Makefile? Aug 05 04:29:08 *cough* I didn't, because I don't have the sshkey on my local build system Aug 05 04:29:14 * NAiL wanders off to fix Aug 05 04:29:58 I wonder what OE is going to say about my perl packages RDEPENDS hack.. Aug 05 04:30:15 They're gonna say great! ;-) Aug 05 04:30:27 rwhitby: was it dsa or rsa I needed? Aug 05 04:30:27 Im not so sure :-) Aug 05 04:31:26 NAiL: don't think it matters Aug 05 04:32:19 Master Makefile pushed Aug 05 04:32:53 I guess I can set up so my local box pushes the stable feed as well now Aug 05 04:33:23 The >250ms ping to nudi annoys me ;-) Aug 05 04:37:22 what's the easiest way of mounting a JFFS2 filesystem on a dev host? Aug 05 04:37:57 I thought mtdram would do the trick, but I can't figure out how to use it :-/ Aug 05 04:39:44 NAiL: One last thing with the gentoo setup-host file - I would personally use ~dev-libs/boost-1.32.0 and ~dev-util/monotone-0.21 - that way if there is a dev-libs/boost-1.32.0-r5 or a dev-util/monotone -0.21-r1 it will automatically use that instead without going to boost 1.33 or monotone 0.22 Aug 05 04:40:39 aha Aug 05 04:40:41 yes Aug 05 04:40:45 good idea Aug 05 04:41:56 I'm mucking around with the makefile again, so I can fix soon Aug 05 04:50:12 never mind, I just need to load mtdblock on top of mtdram Aug 05 04:51:16 how the.. I just managed to cut myself on the foil in a FTP cat5 cable... one less finger to type with.. Aug 05 04:55:30 congratulations :) Aug 05 04:55:55 thanks.. I think... Aug 05 04:58:13 you have a sepeate cable for ftp? :) Aug 05 04:58:30 s/sepeate/dedicated Aug 05 04:58:40 STP? :) Aug 05 04:58:49 scp -i $SSH_KEY Makefile slug@nugabe.nslu2-linux.org <-- what is wrong with this command? Aug 05 04:58:53 :) ftp = foil twisted pair Aug 05 04:59:05 stp = sheilded twisted pair Aug 05 04:59:05 NAiL: lacking : ? Aug 05 04:59:19 kolla: there's a : and a path afterwards Aug 05 04:59:32 it's just when I do that, scp complains it can't find identity file SH_KEY Aug 05 04:59:36 03nail 07org.nslu2-linux.dev * reec9f0df... 10/Makefile: Added SSH_KEY variable, fixed gentoo package masking Aug 05 04:59:41 strange Aug 05 05:04:31 {} Aug 05 05:07:04 now Aug 05 05:07:08 it works Aug 05 05:07:10 yaya Aug 05 05:07:37 Make has single char variables, unless you () or {} them Aug 05 05:07:49 make is a very fussy beast. Aug 05 05:08:18 I noticed that ;) Aug 05 05:08:33 but the Master Makefile is a work of genious :D Aug 05 05:08:56 peteru-home: have you got any idea on how to make the autobuild target in the master Makefile report the targets that fail (instead of the poor attempt at incrementing a counter which doesn't work) ? Aug 05 05:09:22 won't errors += "buildtarget" work? Aug 05 05:09:30 let me have a look at it... Aug 05 05:09:48 03nail 07org.nslu2-linux.dev * rec5603c5... 10/Makefile: Add nice curlybrackets around those SSH_KEY variables of mine Aug 05 05:09:51 03nail 07org.nslu2-linux.dev * r51508965... 10/Makefile: Ugh, unquote the contents of SSH_KEY variable so it's not looking for ~/.ssh/id_dsa in the current dir Aug 05 05:09:54 got my free mcp120 reset controller through the post Aug 05 05:11:28 http://www.oreilly.com/catalog/make3/book/index.csp Aug 05 05:11:36 Chapters 4 or 5 might have some hints. Aug 05 05:14:05 Got my new bag of 25 MCP120's. I used my free one(s) on devslug; now to add them to the others... Aug 05 05:14:14 I guess the other option is to use $(warning target blah blew up) Aug 05 05:14:49 peteru-home: I want it to be printed at the end, or I'll miss it. Aug 05 05:15:45 Oh, I see, you want to accumulate all the error messages Aug 05 05:16:03 "17 errors were detect during autobuild" Aug 05 05:16:43 which would then warrant furthur investigation. Aug 05 05:16:53 Look at those nice three green nslu2-linux columns: http://ewi546.ewi.utwente.nl/OE_qa/OpenEmbeddedBuild/status.html Aug 05 05:17:33 beauty! Aug 05 05:17:36 dyoung: actually, something like "Failed: openslug optware/nslu2 optware/wl500g" Aug 05 05:18:03 well done Aug 05 05:18:23 rwhitby: how are those generated? Ie... how does the tinderbox stuff really work? Aug 05 05:18:47 theres a oe hook for that. Aug 05 05:19:07 NAiL: the autobuild runs on nudi do a HTTP post to a machine in .nl, which does the rest. Aug 05 05:19:15 cool Aug 05 05:19:39 03nail 07org.nslu2-linux.dev * rab8a2feb... 10/Makefile: Create a separate openslug-2.3-beta convenience target that does all my work Aug 05 05:20:07 hmm Aug 05 05:20:07 * peteru-home isn't enough of a GNU make expert to be able to give an answer :( Aug 05 05:20:25 peteru-home: I was thinking shell stuff rather than make stuff Aug 05 05:25:27 NAiL: rwhitby@linux:/home/slug$ set foo += "bar" ; echo $foo Aug 05 05:25:30 doesn't work Aug 05 05:26:09 how about redirecting stderr to a tmp file and then showing that at the end? Aug 05 05:27:04 it'll be too long. I need something that tail will pick up easily. It might even end up being echo'd here in the channel to announce errors in the autobuild. So it needs to be a one line summary. Aug 05 05:29:21 argh Aug 05 05:29:26 keyboard mucked up Aug 05 05:29:35 could someone type a dollar sign and an at sign? Aug 05 05:30:06 $@ Aug 05 05:30:16 TY Aug 05 05:30:59 abcdefghijklmnopqrtsuvwxy1234567890!@#$%^&*()-_+={[}]:;"'?/>.<,~`|\ Aug 05 05:31:16 cut and paste away ... Aug 05 05:31:26 how about... (rm /tmp/builderrs ; cmd1 || echo >> cmd1 ; ... ; [ -e /tmp/builderrs ] && echo "*** Errors in $(cat /tmp/builderrs) ) Aug 05 05:31:28 it looks like it's normal again Aug 05 05:32:20 basically, the error log won't get created if all subcommands succeed, so you can use the presence of it as an indicator of failure. Aug 05 05:32:36 good idea Aug 05 05:33:21 are there any size constraints on variables in bash? Aug 05 05:33:37 ie, could I put an entire buildlog in a variable and just parse that? :) Aug 05 05:34:33 yes, the maximum size is imposed by the underlying OS Aug 05 05:34:46 oh well Aug 05 05:34:50 I'll just use a file Aug 05 05:35:27 Um, I don't want the buildlog in a file. Just a simple pass/fail summary at the end. Aug 05 05:35:40 the buildlog is already in autobuild.log Aug 05 05:35:54 rwhitby: I'm working on a completely different thing right now ;) Aug 05 05:35:59 oh,ok :-) Aug 05 05:37:14 basically a simple makefile that'll enable me to sit at home and "ssh repvik@nudi make" to update the feed, and let it send me a mail to say "Hey, the build borked" or s/bo/wo/, and possibly which packages were pushed. Aug 05 05:38:30 like now, it borked Aug 05 05:38:31 :P Aug 05 05:42:14 ok, generic regexp... ".*" is any number of unknown characters, am I right? Aug 05 05:43:08 yes Aug 05 05:43:47 NAiL: the Makefile is now broken for me. scp -i ~/.ssh/id_dsa Makefile ... is asking for a password. Aug 05 05:43:49 thats 0-inifinite unknown characters (or ".\*"), 1-infinite is ".\+" Aug 05 05:44:29 rwhitby: where is your slug ssh key? Aug 05 05:44:54 probably rsa Aug 05 05:45:01 03nail 07org.nslu2-linux.dev * rcfcc95d7... 10/Makefile: Move a ; from the beginning of a line Aug 05 05:45:05 I've never managed to just do "make" and have everything complete OK. Not once in the last month :( Aug 05 05:45:11 hmm Aug 05 05:45:20 but that's not something I should have to put in the makefile. ssh sorts that out. Aug 05 05:45:45 I'd like to remove that line and ask you to find another way to solve your individual problem. Aug 05 05:45:45 not on my box, I already have an id_dsa and _rsa there Aug 05 05:46:56 you shouldn't specify the location of your identity like that. Look in ~/.ssh/rc and ~/ssh/environment to fix up your installation. Aug 05 05:47:21 peteru-home: problem is that I have three identities in .ssh/ Aug 05 05:47:44 peteru-home: Me neither. I'm shooting for a hat trick right now. Aug 05 05:47:59 ah, I'll just move the "-i" inside the variable Aug 05 05:48:15 urgh Aug 05 05:48:16 no Aug 05 05:48:24 But rwhitby is still using a rsa key.... Aug 05 05:48:30 yes :\ Aug 05 05:48:34 so that fix wont really solve his problem.... Aug 05 05:48:52 dyoung: I don't have a problem :-) Aug 05 05:49:07 issue? Aug 05 05:49:12 nope, not here. Aug 05 05:49:23 * rwhitby waits for CIA ... Aug 05 05:49:38 03rwhitby 07org.nslu2-linux.dev * r443f6abe... 10/Makefile: Removed the SSH_USER stuff Aug 05 05:49:46 hahaha Aug 05 05:49:49 cool. Aug 05 05:50:04 heh Aug 05 05:50:27 NAiL: I have a different idea, but am checking on peteru-home's lead first. Aug 05 05:51:56 fair enough Aug 05 05:52:34 NAiL: you need an ssh_config file, in which you specify different hosts, and the keys which should be used for each host Aug 05 05:52:40 do man ssh_config Aug 05 05:54:41 03nail 07org.nslu2-linux.dev * rf86d6680... 10/Makefile: Fix typo in openslug-2.3-beta target Aug 05 05:59:25 /home/slug/optware/nslu2/builds/mysql dies with libtool.m4 and ltmain.sh have a version mismatch! Aug 05 06:00:06 dyoung: How far do you get? Aug 05 06:01:50 peteru-home: right now its still going, earlier today it croaked in the middle of openslug. some perl module.... Aug 05 06:01:59 its in ucslugc now Aug 05 06:12:27 there.. now I have a beta of perl with rdepends for all the module packages Aug 05 06:12:42 its not pretty, but it works.. Aug 05 06:12:53 ~ube Aug 05 06:12:54 it has been said that ube is Ugly But Effective. Or sometimes Unorthrodox But Effective Aug 05 06:13:07 do I dare just upload, or should I let someone look at the mess first? :-) Aug 05 06:13:22 DaKa2: are you saying that installing "perl" will also bring all the modules with it? Aug 05 06:13:28 rwhitby: no Aug 05 06:13:40 installing one perl-module package brings the ones it needs to work Aug 05 06:14:02 rdepends betwwen the individual module-packages Aug 05 06:14:35 oh, ok. That's cool! Is it worked out automatically somehow? Aug 05 06:15:24 rwhitby: a "small" script is run by me in an old install dir, and that spits out the RDEPENDS lines to put in the bb-file Aug 05 06:15:49 me == the one packaging perl Aug 05 06:16:06 the script is included as a comment in the bb.. Aug 05 06:16:26 only problem is the bb turn out 74Kb Aug 05 06:16:42 DaKa2: have you discussed this in #oe ? Aug 05 06:16:54 no.. this is just an experiment.. Aug 05 06:17:15 jbowler files a slugbug on it Aug 05 06:17:16 actually, jbowler filed a slugbug on perl module rdepends Aug 05 06:17:18 ah Aug 05 06:17:18 filed* Aug 05 06:18:03 this is one way to solve it, only other way is lots and lots of python, and probably redesign of package*.bbclass Aug 05 06:18:05 http://slugbug.nslu2-linux.org/bug.php?op=show&bugid=240&pos=1 Aug 05 06:18:38 Can jbot query slugbug? Aug 05 06:22:17 how can you get the build system to redo the early part of the (ucslugc) build that requres monotone? Aug 05 06:22:37 blaster8: you can run "make setup" at any time Aug 05 06:22:56 right Aug 05 06:25:05 rwhitby: btw, what do you think about splitting up ntp so tickadj is separate and can be set to run by default in a firmware image? It takes <4k. Aug 05 06:25:50 NAiL: is it an OE package, or one we added? Aug 05 06:26:57 it's an OE package. I'm just asking about the firmware bit here actually Aug 05 06:27:36 (or, it appears to be since it has no maintainer/author) Aug 05 06:29:16 I could add RDEPENDS to ntp, so that it automatically installs the tickadj package, and nobody would notice ;) Aug 05 06:34:37 NAiL: it sounds like a reasonable idea to me. It would have to be discussed in #oe too ... Aug 05 06:34:49 of course Aug 05 06:35:06 and yes, we could add the tickadj bit to the firmware. I presume your referring to openslug derivatives Aug 05 06:35:16 um, yes Aug 05 06:35:35 It's one of the alternatives I've come up with for http://slugbug.nslu2-linux.org/bug.php?op=show&bugid=239&pos=0 Aug 05 06:36:13 I'd like to see it fixed in the kernel, but I've got no idea how to do that. Aug 05 06:41:02 03purbanec * 10unslung/make/mysql.mk: Fix up a potential conflict between shipped and platform specific autotools files. Aug 05 06:41:43 that should fix the failing build of mysql - at least it did on my Gentoo system Aug 05 06:41:54 03rwhitby * 10unslung/make/ipkg-utils.mk: Fixed the ipkg-utils download location. Aug 05 06:44:52 does eaccelerator build for everyone? Aug 05 06:45:13 I'm still in ucslugc ... Aug 05 06:45:26 but that package was problematic for me around 3 weeks ago... Aug 05 06:45:49 03rwhitby * 10unslung/make/asterisk.mk: Fixed the anoncvs password. Aug 05 06:46:52 peteru-home: make autobuild completes on nudi (Debian Etch, with make setup-host-debian run on it) Aug 05 06:47:29 I moved to ubuntu a couple days ago, so this build is a complete blow-away build for me. Aug 05 06:57:03 I'm on a very fresh Gentoo system. Aug 05 06:58:45 ./configure: line 546: PHP_INIT_BUILD_SYSTEM: command not found Aug 05 06:59:45 about the timing issue - how about using adjtimex Aug 05 07:00:02 guide here: http://www.togaware.com/linux/survivor/Using_adjtimex.shtml Aug 05 07:00:32 03nail 07org.openembedded.nslu2-linux * r746a5402... 10/packages/ntp/ntp_4.2.0.bb: Split tickadj to separate package, add RDEPENDS in ntp. Noone should notice ;) Aug 05 07:01:08 blaster8: how large is adjtimex? Aug 05 07:01:32 and, I can't access that link Aug 05 07:01:34 just times out Aug 05 07:03:22 03nail 07org.openembedded.nslu2-linux * r5ee09924... 10/packages/ (23 files in 12 dirs): Aug 05 07:03:22 propagate from branch 'org.openembedded.dev' (head 354126a72c55235d0954846f3a20d7176a0d0823) Aug 05 07:03:22 to branch 'org.openembedded.nslu2-linux' (head 746a5402012c5e160469061339fc8ebe2f339340) Aug 05 07:07:36 NAiL - I can access it?? Aug 05 07:08:13 The adjtimex command can be used to counter systematic drift in the system clock. It is tricky to fine tune, and it is recommended that you use ntp instead if you can. <-- Aug 05 07:10:34 apparently, its only dependency is sed (good start) Aug 05 07:12:07 apart from /etc/issue, is there some file that contains info about what OpenSlug version on is running? Aug 05 07:12:13 one Aug 05 07:12:29 adjtimex.c is about 82KB Aug 05 07:12:39 sorry, 52KB Aug 05 07:13:17 which is ~48k larger than tickadj Aug 05 07:13:57 and tickadj doesn't depend on squat ;) Aug 05 07:14:12 k Aug 05 07:14:29 rwhitby: what versions of autoconf, automake and libtool are you using? Aug 05 07:15:47 rwhitby@nudi:~$ dpkg -l |grep auto Aug 05 07:15:47 ii autoconf 2.59a-3 automatic configure script builder Aug 05 07:15:47 ii automake1.4 1.4-p6-9 A tool for generating GNU Standards-complian Aug 05 07:15:47 ii automake1.9 1.9.4-1 A tool for generating GNU Standards-complian Aug 05 07:15:48 ii autotools-dev 20041130.2 Update infrastructure for config.{guess,sub} Aug 05 07:15:48 rwhitby@nudi:~$ dpkg -l |grep libtool Aug 05 07:15:50 ii libtool 1.5.6-4 Generic library support script Aug 05 07:15:52 rwhitby@nudi:~$ Aug 05 07:17:02 i'll try and use spotlight to find the default tick value in the kernel source Aug 05 07:17:05 GNU Standard-complains? :) Aug 05 07:20:51 I have sys-devel/libtool-1.5.18-r1, sys-devel/automake-1.9.6 and sys-devel/autoconf-2.59-r7 Aug 05 07:21:42 Looks like I am at same or newer versions for everything. Aug 05 07:22:15 note that we actually do need to have both automakes installed (< 1.5 and > 1.5) Aug 05 07:23:06 that's why setup-host-debian explicitly says "automake" and "automake1.9" Aug 05 07:24:57 I have automake 1.7.9-r1 1.8.5-r3 1.5 1.6.3 1.9.6 and 1.4_p6 Aug 05 07:25:48 it's a mess : Aug 05 07:25:49 :) Aug 05 07:31:38 spotlight is indexing the entire linux kernel source (still looking for the default kernel tick value) Aug 05 07:33:15 blaster8: it won't find it thataway, I think. It's not stored as a value, but as a formula Aug 05 07:41:11 blaster8: default value? the startup value of jiffies? Aug 05 07:42:30 Jacmet: basically I'm wondering how the kernel derives its 10000 value for the tick Aug 05 07:42:51 blaster8: you mean the adjtimex value? Aug 05 07:42:52 sorry, number of ticks per day Aug 05 07:42:58 basically, yes Aug 05 07:43:11 only a novice at these things, mind Aug 05 07:43:27 blaster8: ticks per day? Aug 05 07:45:08 http://www.togaware.com/linux/survivor/Using_adjtimex.shtml - default ticks per day is 10000 according to this Aug 05 07:47:11 btw - has everyone looking at the nslu2's timing seen this: http://www.nslu2-linux.org/wiki/HowTo/SetUpAnNtpServer Aug 05 07:48:05 several times Aug 05 07:48:34 but running tickadj allows the slug to minimize drift even if it's not connected to the 'net Aug 05 07:48:50 heh, NAiL wrote some of that page if I remember correctly Aug 05 07:49:00 yes, I did ;) Aug 05 07:49:24 it's very odd the unslung has 9903 and needs 10000 whereas openslug needs 11011 and has 10000 Aug 05 07:49:45 given that the time source code hasn't changed since about 2000 Aug 05 07:50:01 where can you get the linksys kernel source Aug 05 07:50:54 because that suggests the Linksys people fixed the problem twice, somehow - there may be a timer kernel patch somewhere in it Aug 05 07:50:59 sorry if you lot have been here before Aug 05 07:51:25 they didn't fix the problem twice. They "fixed" the problem, but got the wrong number, the inverse result Aug 05 07:51:50 blaster8: the 10000 is the value to tick_usec (in kernel/timer.c) Aug 05 07:52:16 blaster8: which is initially set to TICK_USEC Aug 05 07:52:37 NAiL: aah, incompetence Aug 05 07:52:39 I get it Aug 05 07:52:44 blaster8: which is defined in include/linux/jiffies.h Aug 05 07:53:04 thanks Jacmet - I'll be quiet for a while and have a dig around Aug 05 07:54:14 blaster8: TICK_USEC is defined to 1000000/USER_HZ (which is normally 100) - so 10000 Aug 05 07:54:42 so if we patched the kernel with a different value then this would fix the problem? Aug 05 07:54:49 where different=correct Aug 05 07:56:33 I'll hold off committing any further ntp changes until you figure out something =) Aug 05 07:56:44 ;) Aug 05 08:04:47 btw, the current plan is to see what Linksys did wrong, and from that try and do the same thing, but right Aug 05 08:04:52 simple and effective :) Aug 05 08:05:02 ~ube Aug 05 08:05:02 [ube] Ugly But Effective. Or sometimes Unorthrodox But Effective Aug 05 08:06:00 If we can fix it in the kernel, we should do that. Aug 05 08:06:05 night all Aug 05 08:06:21 rwhitby-asleep: agreed Aug 05 08:19:00 is it possible that eaccelerator will not build on a host that has no php installed? The log seems to indicate that files are being pulled from /usr/share/aclocal/ rather than the staging dir. Aug 05 08:22:39 Change USER_HZ to 99 Aug 05 08:22:58 But I'm not sure what else that changes. Aug 05 08:23:50 (99 is exactly the right value, unlike 10101 - should be 10101.*01* Aug 05 08:24:25 jbowler-zzz: Im working on a fix for #240, should be done kindof soon Aug 05 08:25:42 DaKa2: notice that some of the module names have ${TARGET_SYS} in them (linux and linux-uclibc are different). Aug 05 08:26:03 That's fine if the script runs at build time I think. Aug 05 08:26:07 jbowler-zzz: not anymore :-) so that needs to be changed in netpbm, though I do RPROVIDE the new names Aug 05 08:26:25 s/new/old Aug 05 08:26:52 That's a better approach. netpbm should just list the things it uses directly. Aug 05 08:27:09 yep Aug 05 08:27:54 but the script to generate the RDEPENDS is not run at build time, but #oe didn't seem to care until someone invents do_repackage Aug 05 08:30:50 Means it has to be done on each new perl release, at least. Aug 05 08:31:11 yes, but the script is included as a comment in the perl bb atleast Aug 05 08:31:23 but its ugly.. Aug 05 08:31:26 really ugly Aug 05 08:31:44 ~ube Aug 05 08:31:44 methinks ube is Ugly But Effective. Or sometimes Unorthrodox But Effective Aug 05 08:33:39 DaKa2: ah well, some people say that perl is a great way to do stuff like this, but I always just use awk... Aug 05 08:33:57 I used sed and grep ;-) Aug 05 08:34:32 and uniq I assume, to avoid multiple occurs if the script contains multiple use of the same module? Aug 05 08:34:37 yep Aug 05 08:34:43 sort | uniq Aug 05 08:35:15 I don't know if I dare paste it here... Aug 05 08:35:33 maybe I should just commit, and run away hoping noone will look at it.. Aug 05 08:37:27 maybe I missed something important in the way perl works, but it atleast works now.. Aug 05 08:38:35 Sounds good to me. Aug 05 08:41:22 Looks like USER_HZ is the correct thing to hack, from kernel/time.c in do_adjtimex: Aug 05 08:41:27 /* if the quartz is off by more than 10% something is VERY wrong ! */ Aug 05 08:41:27 if (txc->modes & ADJ_TICK) Aug 05 08:41:27 if (txc->tick < 900000/USER_HZ || Aug 05 08:41:28 txc->tick > 1100000/USER_HZ) Aug 05 08:41:28 return -EINVAL; Aug 05 08:42:14 (The actual deviation from 100 is only 1%, so that test should never be a problem even though USER_HZ is wrong). Aug 05 08:43:01 ok Aug 05 08:43:18 USER_HZ is set in /include/asm-arm/param.h Aug 05 08:43:35 Correct - what I don't understand is where this got screwed up. Aug 05 08:43:53 I'm looking at the linksys kernel source now (just downloaded) Aug 05 08:44:02 Either they got the wrong crystal (and, let's face it, they put a resistor in the wrong place), Aug 05 08:44:15 Or the hardware drops 1 interrupt per second? Aug 05 08:44:21 well, that could easily be possible (bad design) Aug 05 08:44:40 they mucked up the ID of the IXP board too, if I remember correctly Aug 05 08:44:43 The linksys kernel source doesn't help - that's after the event, they just tried to fix the hardware error and got it wrong. Aug 05 08:45:17 No, they got the IXP board id correct - just they shouldn't have used it... Aug 05 08:45:58 oops Aug 05 08:46:13 It would really help if Linux kernel writers would document their code. Aug 05 08:46:41 haha - reminds of me of a comment in one of the source files: 'Does this go here?' Aug 05 08:46:53 makes you glad sometimes that they don't Aug 05 08:47:18 No, I like comments like that - it's the lines of source which just do something with no comment which annoy me. Aug 05 08:47:56 Like "oh, I know how to fix this", "but I'm not so sure that I will actually explain what I'm trying to do" Aug 05 08:49:25 jbowler-zzz: what's the fun in easy to understand code anyway? ;) Aug 05 08:49:33 Like /* FIXME: This is broken. */ Aug 05 08:50:54 Yes, tha'ts useful - saves hours looking at a time of code saying "hey, this can't possibly work" Aug 05 08:51:31 ... hours of time lokking at ... Aug 05 08:51:51 looking (the sun is shining into my eyes :) Aug 05 08:52:08 jbowler: do you have a spare slug to try the timing patch out on? Aug 05 08:52:12 or anyone, in fact Aug 05 08:52:20 03purbanec * 10unslung/make/eaccelerator.mk: Explicitly specify automake version to prevent breakage on some systems. Aug 05 08:53:16 Yes, I will try the param.h patch - I'm pretty sure it will work, problem is that it doesn't make sense that "USER_HZ" means "number of hardware clock interrupts per second". Aug 05 08:53:32 jbowler-zzz: do you think anybody wants to use perl 5.8.4 with netpbm? Aug 05 08:56:43 jbowler-zzz: we coudl quite easily check if the interrupts are really running at 100Hz by blinking one of the leds in the interrupt handler and hook up a scope Aug 05 08:57:02 gpio is too slow, innit? Aug 05 08:57:30 NAiL: for 100Hz? no Aug 05 08:58:13 03purbanec * 10unslung/make/file.mk: Update the source URL to a site that actually hosts the required file. Aug 05 08:58:58 there.. I hope I didn't do anything wrong/bad.. Aug 05 08:59:37 perl is committed? Aug 05 08:59:45 yup.. Aug 05 08:59:57 * NAiL waits for cries of despair Aug 05 09:00:02 * DaKa2 too Aug 05 09:00:40 DaKa2: if anyone does want netpbm+perl_5.8.4 then they can fix it themselves I think. Aug 05 09:00:46 03daka 07org.openembedded.nslu2-linux * r03fafb11... 10/packages/perl/ (perl-rdepends_5.8.7.inc perl.inc perl_5.8.7.bb): Aug 05 09:00:46 Modify the way packages get split so no packages get ${TARGET_SYS} in the names, unfortunately makes some packages change n Aug 05 09:00:46 ames, but this is they way it should be. Aug 05 09:00:46 Add RDEPENDS in every perl-module- package to its dependencys. Aug 05 09:01:00 dammit... oh well.. linebraked changelog.. Aug 05 09:01:09 03daka 07org.openembedded.nslu2-linux * r46d57b1a... 10/packages/netpbm/netpbm_10.28.bb: Change RDEPENDS from perl-module-${TARGET_SYS}-xsloader to perl-module-xsloader to match perl packaging change Aug 05 09:01:10 jbowler-zzz: good :) Aug 05 09:02:00 jacmet: If I knew where the interrupts were happening I could work out whether USER_HZ is the right thing to change. Aug 05 09:02:36 This should really be fixed by a kernel expert, but kernel experts seem to get scared if they have to look at kernel code. Aug 05 09:03:19 jbowler-zzz: I'm pretty sure it isn't Aug 05 09:03:33 jbowler-zzz: I might have a look this weekend Aug 05 09:03:40 I know 2.6.13 apparently allows USER_HZ to be changed (and defaults to 250). I don't see how that can work if it is all in the hardware, so maybe there is a bug in the hardware interrupt handler. Aug 05 09:03:51 this is when someone says 'hey, you know you could have done that perl thing with a single command?'.. Aug 05 09:04:28 jbowler-zzz: that's HZ not USER_HZ Aug 05 09:04:30 famous last words^H^H^H^H^Honliner Aug 05 09:05:51 HZ is unlikely to be greater than the fastest hardware interrupt - so if it can be set to 250 soomething is dividing an internal interrupt down and has probably got it wrong... Aug 05 09:11:05 Ok, it looks like USER_HZ is the right thing to change - just Linus gave it a dumb name Aug 05 09:11:23 exactly! Aug 05 09:11:29 (And the sensible name, CLOCKS_PER_SEC, from someone else was ignored) Aug 05 09:11:36 exactly Aug 05 09:11:37 NAiL: is it possible to remove all the old perl packages from the feeds? there are some r5 in there, and when r14 hits it there will be some r13 left over Aug 05 09:12:20 he is a man of true subtlety: 'I chose 1000 originally partly as a way to make sure that people that assumed HZ was 100 would get a swift kick in the pants.' Aug 05 09:12:24 Still, if the x86 stuff is done by programming the timer with a divisor it is quite likely that the NSLU2 is as well, and it is difficult to get integer division right wrt errors ;-) Aug 05 09:13:41 That comment is reasonable, but the real reason may lie in the other part - along the lines of forcing other people to debug the code. Aug 05 09:17:27 DaKa2: the old perl packages should be removed automagically from the unstable feed Aug 05 09:17:52 NAiL: no, since there are no new ones with the same names anymore Aug 05 09:18:06 oh Aug 05 09:18:09 hmm Aug 05 09:18:24 well, yes, it's possible to do Aug 05 09:21:09 03purbanec * 10unslung/make/freeradius.mk: Update the source URL to a site that actually hosts the required file. Aug 05 09:23:30 03nail 07org.openembedded.nslu2-linux * re4d3d3ab... 10/packages/ (6 files in 4 dirs): Aug 05 09:23:30 propagate from branch 'org.openembedded.dev' (head 15f551f6c01721f6885a48d2227e2211eaea3a0c) Aug 05 09:23:30 to branch 'org.openembedded.nslu2-linux' (head 46d57b1afb6a0c0939707ca827bf4abc67945f1a) Aug 05 09:24:17 * DaKa2 hides Aug 05 09:24:50 arch/arm/mach-ixp425/ixp425-time.c refers to LATCH, which I pressume is the countdown value used to fire the clock interrupt. Aug 05 09:25:10 haha Aug 05 09:25:29 DaKa2: The correct perl should be in the feeds in a while, with all the old ones dropped Aug 05 09:25:41 NAiL: great :) Aug 05 09:25:48 when it's been properly tested, I'll even propagate it to the stable feed ;) Aug 05 09:25:58 :) Aug 05 09:26:15 I guess I should start learning perl Aug 05 09:26:18 and python Aug 05 09:26:22 oh, there the build should be done Aug 05 09:26:24 yeah.. me too Aug 05 09:27:10 the only perl I know is because of the similarities in php :) Aug 05 09:27:13 arch/arm/kernel/xscale-time.c may also be of interest Aug 05 09:27:30 well, I know a bit more than that.. but.. Aug 05 09:27:38 NOTE: package perl-5.8.7: started Aug 05 09:27:38 NOTE: package perl-5.8.7-r14: task do_fetch: started Aug 05 09:27:45 That's mah local build Aug 05 09:27:51 :) Aug 05 09:28:25 and the build on nudi started *now* Aug 05 09:28:31 tinderbox is really strange... Aug 05 09:28:36 so, give it 15 minutes to get pushed to the feed Aug 05 09:38:16 interesting. It seems that the LATCH value is derived from IXP425_PERIPHERAL_BUS_CLOCK, which *should* be 66MHz Aug 05 09:44:44 werd Aug 05 09:44:53 * DaKa2 broke something? Aug 05 09:44:54 the perl change doesn't seem to have passed on to oe Aug 05 09:45:06 yeah, I was wondering about that.. Aug 05 09:45:27 nudi is a little bit stressed now, but it should've propagated anyhow Aug 05 09:52:04 poor nudi Aug 05 09:52:32 not that bad now Aug 05 09:52:35 2.3 load Aug 05 10:16:06 hmm Aug 05 10:16:17 anyone still looking at the timing issue Aug 05 10:16:25 I have an interesting comment for you Aug 05 10:16:37 "/* change this if you have some constant time drift */" Aug 05 10:18:04 Yes, that's interesting - where is it? Aug 05 10:18:16 arch/arm/kernel/time.c Aug 05 10:18:21 or arch/arm26/kernel/time.c Aug 05 10:18:41 NSLU2 Linux 2.4 kernel Aug 05 10:18:57 its in openslug-kernel-2.6.12.2-r6 too Aug 05 10:18:58 ~/arch/arm/kernel/time.c Aug 05 10:19:15 where it defines USECS_PER_JIFFY Aug 05 10:19:30 and in 2.6 Aug 05 10:19:31 http://www.kernel.org/git/gitweb.cgi?p=linux/kernel/git/torvalds/linux-2.6.git;a=blob;h=1b7fcd50c3e25c8a208bc3347dc55dc7d0605e1e;hb=403fe5ae57c831968c3dbbaba291ae825a1c5aaa;f=arch/arm/kernel/time.c Aug 05 10:19:39 sorry for the ugly link Aug 05 10:19:40 LATCH is used in arch/arm/mach-ixp4xx/common.c and is used as an increment on 'last_jiffy_time'. Aug 05 10:20:28 *IXP4XX_OSTS is incremented in 'jiffy's, so what matters is the definition of a jiffy (not LATCH). Aug 05 10:20:52 That's consistent with changing USECS_PER_JIFFY - i.e. I think LATCH is irrelevant. Aug 05 10:20:56 which suggests changing USECS_PER_JIFFY makes sense Aug 05 10:21:53 IIRC I got to that point before, then discovered some problem with chaning USECS_PER_JIFFY (like it was used elsewhere for something different). Aug 05 10:22:15 hmm, well we always have changing USER_HZ Aug 05 10:22:58 I think it's time to try and compile a kernel with one of these values adjusted and see what happens Aug 05 10:22:59 What seems totally wrong is that there are multiple definitions with overlapping uses. Aug 05 10:23:45 There is also 'CLOCK_TICK_RATE' Aug 05 10:24:25 I'm pretty sure that the code in the kernel is fundamentally broken. Aug 05 10:24:47 well, as long as it can be hacked, people obviously don't care Aug 05 10:28:46 It can't necessarily be hacked - it might not work for different timer interrupt rates or if the timer interrupt rate is not divisible by a given number. Aug 05 10:29:18 There's a lot of rounding going on - "m := (n + d/2) / 2" - that's very scary. Aug 05 10:30:23 I think it can be fixed in common.c - because IXP4XX_OSTS delivers known values and those values can be used to increment a pseudo timer. Aug 05 10:35:02 * We use IXP425 General purpose timer for our timer needs, it runs at Aug 05 10:35:02 * 66.66... MHz Aug 05 10:35:02 */ Aug 05 10:35:03 #define CLOCK_TICK_RATE (66666666) Aug 05 10:35:32 But I think that is not being used everywhere where it should be used... Aug 05 10:37:27 /* LATCH is used in the interval timer and ftape setup. */ Aug 05 10:37:27 #define LATCH ((CLOCK_TICK_RATE + HZ/2) / HZ) /* For divider */ Aug 05 10:37:37 Ha ha. Aug 05 10:38:34 That's a very optimistic definition. Let me see.... error in integer arithmetic = +/-0.5, divisor is 100, error is... oops 1/100 Aug 05 10:39:09 heh Aug 05 10:39:18 No, not quite that bad, but use the number elsewhere and sooner or later that will happen. Aug 05 10:39:29 yeah Aug 05 10:40:57 All the numbers need to be exact - USER_HZ, HZ, CLOCK_TICK_RATE, LATCH, USECS_PER_JIFFY - anything which is used in multiple places. Aug 05 10:41:35 So CLOCK_TICK_RATE should be 66666700 probably. Aug 05 10:42:02 common.c just has to fix it up by propagating the error. Aug 05 10:43:11 I guess the safe approach is to make all Hz values divisible by 250. Aug 05 11:15:22 Dammit. OSC_IN is 33.33MHz by the intel spec - not 33.333333 as implied by the source code! Aug 05 11:19:49 Does anyone out there know what the actual crystal frequency is on the NSLU2 board? Aug 05 11:22:46 jbowler-zzz: doesn't it say? I'll have a look at the board? Aug 05 11:24:45 Just checked the wiki - it is 30.000 MHz on the BB photo Aug 05 11:25:00 Unreadable on the BL Photo Aug 05 11:25:05 jbowler-zzz: there's an 33.000 (Y5) on the back Aug 05 11:25:36 and a 30.000 (Y6) at the front Aug 05 11:25:44 33.00 vs 33.33 is a 1% error! Aug 05 11:26:24 <[cc]smart> hm, are you sure about this ? Aug 05 11:26:26 <[cc]smart> :) Aug 05 11:35:44 That would explain it, by I can only fine three crystals and the numbers go up to Y6 Aug 05 11:37:03 yeah, one at the back, one for the phy and one close to the USB Aug 05 11:37:04 In fact if that is the source of the problem the error is 66/66.666666, which is a tickadj of exactly 10101.01 Aug 05 11:37:37 hello Aug 05 11:37:42 jacmet: Y5? Aug 05 11:37:55 jbowler-zzz: it at the back Aug 05 11:38:09 jbowler-zzz: Y3 is left of the CPU (not mounted) Aug 05 11:38:31 And there should be a real small 32Khz one near the RTC too Aug 05 11:38:45 jbowler-zzz: Y4 is beneath the CPU Aug 05 11:39:04 jbowler-zzz: Y4 is the 32KHz one Aug 05 11:39:25 Well, my best ntp long term measurement of the error is very close indeed to 10101.01 Aug 05 11:41:01 Ok, I'm going to try the hack to CLOCK_TICK_RATE, if the crystal really is that one it may be that the twisty code inside the kernel is actually working... Aug 05 11:41:37 jbowler-zzz: so that only leaves Y2 to be found Aug 05 11:44:05 i have a small problem, someone can help mee plz ? I have flashed my nslu with unslung 5.5 beta but i can't boot from my harddisk :( i have unsling my disk by telnet ( unsling disk1 and reboot ) ... the slug seems boot but no response from 192.168.1.77. If the hardisk is not plug no problem. I don't find the solution :( can you help mee plz ? Aug 05 11:44:57 mcfly587: try on #nslu2-general Aug 05 11:45:13 oh, ok sorry :) Aug 05 11:48:14 jbowler-zzz: Do you know the ramifications of correcting openslug's default runlevel (from 5 to 3)? *Technically*, 3 is correct ;-) Aug 05 11:48:57 rc3.d and rc5.d are identical with all the packages I've got installed Aug 05 11:49:22 Yes - I just checked too. Therefore there are no ramifications. Aug 05 11:49:27 <[cc]smart> verify with rc4.d, too. going to 5 will pass by 4 Aug 05 11:49:42 huh? It will? noooo? Aug 05 11:49:46 <[cc]smart> should Aug 05 11:49:48 It doesn't matter - the init script just runs rc.d directly Aug 05 11:49:52 <[cc]smart> afaik Aug 05 11:49:58 <[cc]smart> hmm Aug 05 11:50:06 <[cc]smart> surprises me somewhat Aug 05 11:50:58 <[cc]smart> i'll have to go an check .) Aug 05 11:51:29 I'll commit that change then Aug 05 11:51:37 Seems fine. Aug 05 11:51:37 I've been running it for a while Aug 05 11:52:22 So I'm working on the assumption that Y5 is the source for OSC_IN and that the '033.000' printed on it really means 33MHz (exactly). Aug 05 11:53:17 If that's right changing CLOCK_TICK_RATE to 66000000 should be the correct solution and, theoretically, no other changes in the kernel are required. Aug 05 11:55:58 neat Aug 05 11:58:23 03nail 07org.openembedded.nslu2-linux * r1b933f7c... 10/packages/sysvinit/sysvinit/openslug/inittab: Change openslug default runlevel from 5 to 3 Aug 05 11:59:21 NAiL: if the CLOCK_TICK_RATE works that's another firmware change... Aug 05 11:59:30 yes, I know Aug 05 11:59:40 (And if it doesn't work something else inside the kernel will) Aug 05 11:59:53 jbowler-zzz: exactly Aug 05 12:00:12 <[cc]smart> you are correct. runlevel change is to be done directly. Aug 05 12:00:36 if the usb isdn changes the kernel as well, we've got "substancial" changes to the firmware ;) Aug 05 12:00:58 What is the usb isdn? Aug 05 12:01:16 malfi's done some work on getting some usb isdn-stuff to work Aug 05 12:01:21 03nail 07org.openembedded.nslu2-linux * r0558010e... 10/packages/sysvinit/sysvinit_2.86.bb: Forgot PR. *sigh* Aug 05 12:01:23 03nail 07org.openembedded.nslu2-linux * rdb631e19... 10/packages/ (4 files in 4 dirs): Aug 05 12:01:24 propagate from branch 'org.openembedded.dev' (head cd4b7abdfdbefabaf7aea80423ae2ff45feb7d1f) Aug 05 12:01:24 to branch 'org.openembedded.nslu2-linux' (head 0558010e1482ae76056505548d911ae74cd4efe2) Aug 05 12:01:36 It's all modules, but I dunno if it will affect the kernel or not Aug 05 12:02:32 There is also udev, which is a big firmware change (because it probably won't work unless both the flash and final rootfs do the same thing). Aug 05 12:03:01 yeah, I was going to start with that when dyoungs disk came Aug 05 12:03:15 jbowler-zzz: how do you easily measure if the clock is off? with ntp? Aug 05 12:03:15 it was supposed to come last week, but it hasn't arrived yet Aug 05 12:03:52 jacmet: yes, tickadj runs automatically and sets 10101, so that needs to be removed. Aug 05 12:04:36 NAiL: what worries me about 2.5 is that, at present, it has only one firmware change (turnup memstick), but then there are a whole bunch of changes just happening. Aug 05 12:05:14 Yeah, I know. 2.5 should've been later Aug 05 12:06:12 Even the tickadj change is a problem - the kernel change is incompatible with the current feed (because that will reset tickadj to 10101) Aug 05 12:07:16 So 2.5 == 2.3 (for the feed) and 2.6+ needs a new feed for NTP at least. Aug 05 12:07:46 hmm.. Aug 05 12:09:26 yes Aug 05 12:09:57 2.5 should just be symlinked to 2.3, and a new feed started for the next release version Aug 05 12:12:31 jbowler-zzz: did you recompile with CLOCK_TICK_RATE changed yet? Aug 05 12:16:04 jacmet: yes, I have, but I don't have an image yet. Aug 05 12:16:44 jbowler-zzz: image? you can just run it from ram in Redboot Aug 05 12:17:17 That would require a tftp server, which I don't have. Aug 05 12:19:28 jbowler-zzz: ohh Aug 05 12:19:59 * Jacmet mumbles something about apt-get install tftpd-hpa .. Aug 05 12:21:01 NAiL: in fact the fix to tickadj is just to delete the value from /etc/default/rcS (i.e. if TICKADJ isn't set tickadj won't be run). Aug 05 12:22:33 ok Aug 05 12:22:39 but that goes in after 2.5 ;) Aug 05 12:23:03 Well, that's the question I'm asking - should 2.5 go out with just one firmware change? Aug 05 12:23:34 Particularly as, unlike a kernel change, it can be done by hand (edit /etc/volatiles). Aug 05 12:24:00 no, that's not exactly... necessary I'd say Aug 05 12:24:15 but I don't think that udev should be in 2.5 though Aug 05 12:24:22 unless we can test it well first Aug 05 12:24:26 Yes, that's a big change. Aug 05 12:24:36 And it doesn't work yet... Aug 05 12:52:51 03jbowler 07org.openembedded.nslu2-linux * r16b882cb... 10/ (conf/distro/openslug.conf packages/meta/openslug-packages.bb): OpenSlug-2.5-beta Aug 05 12:52:54 03jbowler 07org.openembedded.nslu2-linux * r4f3b068b... 10/ (conf/distro/openslug.conf packages/meta/openslug-packages.bb): Return postfix, move to OpenSlug-2.6-beta for furture development. Aug 05 12:53:01 03jbowler 07org.openembedded.nslu2-linux * rd9996b78... 10/ (3 files in 3 dirs): Aug 05 12:53:01 propagate from branch 'org.openembedded.dev' (head 1d5079e0a2930e5f1c383e83c898c6459c81d3ec) Aug 05 12:53:01 to branch 'org.openembedded.nslu2-linux' (head 5111f0fb812d9046053e13b2e9a371075b2b2be1) Aug 05 13:02:07 Hum: looks like the ucslugc cross feed might be misnamed: Aug 05 13:02:08 Downloading http://ipkg.nslu2-linux.org/feeds/ucslugc/cross/1-beta/Packages.gz Aug 05 13:02:08 wget: server returned error 404: HTTP/1.1 404 Not Found Aug 05 13:02:27 (And similarly for native). Aug 05 13:12:25 wee, native kernel compilation worked fine Aug 05 13:13:43 :-D Aug 05 13:14:44 jacques: on DebianSlug with the default debian gcc Aug 05 13:14:59 Jacmet, excellent! Aug 05 13:23:16 slugbot`:/tmp# flashcp -v kernel.img /dev/mtd2 Aug 05 13:23:16 Erasing blocks: 8/8 (100%) Aug 05 13:23:16 Writing data: 899k/899k (100%) Aug 05 13:23:42 ooh, what's flashcp ? Aug 05 13:24:07 a tool from mtd-tools to write to flash (mtd partitions) from linux Aug 05 14:52:58 03repvik * r153 10/releases/OpenSlug-2.3-beta/openembedded/packages/ntp/ (files/tickadj ntp_4.2.0.bb): Split tickadj from ntp, add RDEPENDS, add openslug initscript for 2.3 Aug 05 15:06:05 03jbowler * r154 10/trunk/openslug/openembedded/ (614 files in 291 dirs): Initial sync of openembedded with the monotone database tag for OpenSlug-2.5-beta Aug 05 15:08:03 jbowler-away: looks good so far. The firmware image built, and it's eating through building the packages. Aug 05 15:20:56 NAiL: did all the downloads exist? Aug 05 15:21:20 rwhitby-asleep: so far it hasn't hit any non-existing downloads. I removed the downloads dir first to be sure Aug 05 15:21:38 currently, I'm at gphoto Aug 05 15:21:45 so it's still some way to go Aug 05 15:22:12 but it does look good Aug 05 15:39:09 03jbowler * r155 10/trunk/openslug/ (5 files in 3 dirs): Sync to org.openembedded.dev Aug 05 15:44:29 03jbowler * r156 10/trunk/openslug/ (. setup-env): Removed setup-env from ignores and added it to svn Aug 05 15:47:03 03rwhitby 07org.nslu2-linux.dev * r8536eccd... 10/Makefile: Renamed existing clobber targets to distclean, and added new clobber targets Aug 05 16:06:08 ~weather kcvo Aug 05 16:06:10 Corvallis, Corvallis Municipal Airport, OR, United States; (KCVO) 44-30N 123-17W; last updated: 2005.08.05 2155 UTC; Dew Point: 42 F (6 C); Pressure (altimeter): 30.02 in. Hg (1016 hPa); Relative Humidity: 15%; Sky conditions: clear; Temperature: 96 F (36 C); Visibility: 10 mile(s); Wind: from the NNE (030 degrees) at 8 MPH (7 KT) (direction variable) Aug 05 16:06:24 hmm Aug 05 16:06:28 does it do intl? Aug 05 16:06:32 yes Aug 05 16:06:46 ~weather TRD Aug 05 16:06:49 I can't find station code "TRD" (see http://www.nws.noaa.gov/oso/site.shtml or http://www.nws.noaa.gov/tg/siteloc.shtml for ICAO locations codes). Aug 05 16:07:02 ~weather YPAD Aug 05 16:07:05 Adelaide Airport, Australia; (YPAD) 34-56S 138-31E 4M; last updated: 2005.08.05 2100 UTC; Dew Point: 37 F (3 C); Pressure (altimeter): 30.30 in. Hg (1026 hPa); Relative Humidity: 86%; Temperature: 41 F (5 C); Visibility: greater than 7 mile(s); Wind: from the NE (050 degrees) at 8 MPH (7 KT) Aug 05 16:07:20 it's cold this morning Aug 05 16:08:00 but at least the sun is shining and it's a beautiful blue sky with absolutely no clouds. A nice Adelaide crisp winter morning. Aug 05 16:08:22 ~weather enva Aug 05 16:08:25 Trondheim / Vaernes, Norway; (ENVA) 63-28N 010-56E 17M; last updated: 2005.08.05 2020 UTC; Dew Point: 46 F (8 C); Pressure (altimeter): 29.77 in. Hg (1008 hPa); Relative Humidity: 87%; Sky conditions: partly cloudy; Temperature: 50 F (10 C); Visibility: greater than 7 mile(s); Wind: from the E (100 degrees) at 5 MPH (4 KT) Aug 05 16:08:25 you have winter in the summer? strange Aug 05 16:08:30 ;-) Aug 05 16:08:47 jacques: you have north at the top of your compass? strange .. Aug 05 16:08:53 lol Aug 05 16:09:12 rwhitby: "Don't you fall off?" Aug 05 16:09:17 heh Aug 05 16:09:51 * rwhitby should work out how to turn www.nslu2-linux.org/developers.php upside down Aug 05 16:10:00 hahah ;) Aug 05 16:10:08 so that it matches reality Aug 05 16:10:50 ~weather enva Aug 05 16:10:52 Trondheim / Vaernes, Norway; (ENVA) 63-28N 010-56E 17M; last updated: 2005.08.05 2020 UTC; Dew Point: 46 F (8 C); Pressure (altimeter): 29.77 in. Hg (1008 hPa); Relative Humidity: 87%; Sky conditions: partly cloudy; Temperature: 50 F (10 C); Visibility: greater than 7 mile(s); Wind: from the E (100 degrees) at 5 MPH (4 KT) Aug 05 16:11:21 Anyone else want to add some coords (rounded to the nearest .1 degree for privacy reasons) to that map? Aug 05 16:11:29 sigh, foreigners and their weird ideas Aug 05 16:12:58 what, like thinking that the "world cup" should include teams outside the USA ? Aug 05 16:14:20 the US doesn't count ;) Aug 05 16:17:48 No-one else wants to be listed on the nslu2-linux developers map? Aug 05 16:18:09 ~slugtime Aug 05 16:18:09 from memory, slugtime is 40 hour days 10 hours in 4 timezones with overlap Aug 05 16:18:33 I'd give them a little time to respond, the poor sleeping devs ;) Aug 05 16:41:50 03jbowler 07org.openembedded.nslu2-linux * r90381a67... 10/packages/microcom/microcom_1.02.bb: Aug 05 16:41:50 Fix the microcom source download by using oesources directly - the mirror Aug 05 16:41:50 will not cut in because the port5.com source path does not match any Aug 05 16:41:50 mirror pattern (it is missing a path). Aug 05 16:41:54 03jbowler 07org.openembedded.nslu2-linux * re40105ff... 10/ (3 files in 3 dirs): Aug 05 16:41:54 propagate from branch 'org.openembedded.dev' (head 5aab3b574105c74c9718f6213acf73847b9010b3) Aug 05 16:41:54 to branch 'org.openembedded.nslu2-linux' (head f04c0a8ef28f85aa17e6413d1c485338da112c98) Aug 05 16:52:44 03jbowler * r157 10/trunk/openslug/openembedded/packages/microcom/microcom_1.02.bb: Fix microcom download prior to 2.5-beta Aug 05 18:17:45 03jbowler * r158 10/releases/OpenSlug-2.5-beta/: OpenSlug-2.5-beta source Aug 05 19:14:52 03jbowler 07org.openembedded.nslu2-linux * reca2a27a... 10/packages/ntp/ntp_4.2.0.bb: Aug 05 19:14:52 Make NTP 4.2.0 available for the whole OE world by removing the Aug 05 19:14:52 DEFAULT_PREFERENCE Aug 05 19:14:55 03jbowler 07org.openembedded.nslu2-linux * r629c8f39... 10/packages/meta/ucslugc-packages.bb: Aug 05 19:14:55 Document disabling iperf (I tried it, it doesn't work) and explain why Aug 05 19:14:55 some things will never build with uclibc. Aug 05 20:04:42 03jbowler 07org.openembedded.nslu2-linux * r9c9c90ae... 10/packages/ (6 files in 5 dirs): Aug 05 20:04:42 This fixes the problem with the tick count on NSLU2 boards by correcting Aug 05 20:04:42 the CLOCKS_PER_TICK in the kernel. It also (necessarily) removes the Aug 05 20:04:42 TICKADJ setting in the sysvinit nslu2 rcS-default. It adds a commented Aug 05 20:04:42 out TICKADJ to rcS-default - because that is supported now by the ntp Aug 05 20:04:43 4.2.0 startup (ntpd) script in OE. Aug 05 21:45:44 03jbowler 07org.openembedded.nslu2-linux * rfb1746ed... 10/packages/microcom/microcom_1.02.bb: Aug 05 21:45:44 Change to the OpenWRT mirror so that oesources has something to pull from Aug 05 21:45:44 (perhaps openwrt is pulling from oesources too...) Aug 05 21:58:15 jbot ar Aug 05 21:58:16 rumour has it, ar is argon or a program to create, modify, and extract from archives (man ar), or in binutils Aug 05 22:00:32 does openslung busybox include 'ar'? If not I'm going to file a RFE to have it included Aug 05 22:15:50 03rwhitby 07org.nslu2-linux.dev * r19376ceb... 10/Makefile: Made the branches explicit in the initial monotone setup step. Aug 05 22:29:49 jumpkick, yes busybox in openslug does include 'ar'. Did you try typing 'busybox ar' ? Aug 05 22:30:11 k, it's just the unslung feed that needs it then Aug 05 23:26:52 migration from nudi.nslu2-linux.org to nugabe.nslu2-linux.org for monotone should now be complete. I will leave the server on nudi running overnight to make sure that no-one is still accessing it. Please continue to access monotone.nslu2-linux.org, and you should be talking to the new server on nugabe.nslu2-linux.org ... Aug 05 23:36:51 hey ByronT Aug 05 23:37:20 can I physically hurt an Alcatel representative for making such cheap equipment Aug 05 23:37:39 depends what country you are in, I think Aug 05 23:37:44 <- been fighting a "dumb" DSL ethernet modem for most of the night Aug 05 23:38:00 2nd bad one in as many days.... Aug 05 23:38:16 ./rant off... Aug 05 23:38:21 How can I help you? **** ENDING LOGGING AT Fri Aug 05 23:59:56 2005