**** BEGIN LOGGING AT Wed Jun 01 23:59:56 2005 Jun 02 00:04:18 ah yeah, I think I saw joshua talking about that Jun 02 02:02:45 jacques, too bad its too late to submit a paper for oscon. Jun 02 02:03:21 someone could present "Generic application porting for commodity embedded devices" Jun 02 02:03:29 or something along those lines. Jun 02 02:04:33 just as someone should eventually present "A FOSS model for open source hardware" Jun 02 05:33:28 jbowler-zzz: re config files and upgrading kernels Jun 02 05:33:52 isn't it the case that as long as any modules you require to mount the rootfs are built in to the kernel, then you're ok? Jun 02 05:34:22 (this was one of the reasons why I pushed very hard for ext3 to continue being compiled into the kernel in OpenSlug) Jun 02 05:37:30 <[g2]> morning sunshine Jun 02 05:37:45 jbowler-zzz: but putting some config in the sysconf is a good idea too. You could even put it in the same format as the linksys sysconf, and remain compatible with stock linksys :-) Jun 02 05:37:48 hey [g2] Jun 02 05:38:25 <[g2]> rwhitby, jbowler-zzz was going to put the info in sysconf after the linksys info to remain compatible Jun 02 05:38:59 as part of the linksys info format (i.e. updating the length header to include it) ? Jun 02 05:39:17 <[g2]> I think it would be separate Jun 02 05:39:26 I think dyoung had ascertained that unslung linksys apps preserve anything new that you put in there Jun 02 05:39:33 <[g2]> as in start at some other offset Jun 02 05:40:22 <[g2]> this is primarily for OpenSlug Jun 02 05:40:27 Why not just overwrite it completely if you're not going to put it in the linksys format? Jun 02 05:41:17 <[g2]> well the 128K is used up and the last 110K are zeros Jun 02 05:41:37 <[g2]> jbowler-zzz, wrote the devio program which is quiet powerful Jun 02 05:42:06 yes, it's 128k, and yes, most of it is unused. Jun 02 05:42:14 <[g2]> so stuffing it in at another location is really trivial and removing for reversion is equally easy Jun 02 05:42:59 ok, I guess you can assume that the linksys stuff will never grow past a certain point. Jun 02 05:43:26 so the intention is that you could switch between openslug and unslung by flashing everything except redboot and sysconf? Jun 02 05:43:27 <[g2]> we were talking about saving off to be able to revert and also on subsequent flashes to use the data modified by the openslugger Jun 02 05:43:47 (and retain your important config settings, like ip addresses and network stuff) Jun 02 05:43:50 <[g2]> rwhitby, I think that would be possible Jun 02 05:44:15 <[g2]> if unslung started using devio to initialize its data too Jun 02 05:44:48 <[g2]> unslung could also then retain info across kernel and rootfs flashes also Jun 02 05:44:57 unslung would continue to use the linksys stuff. it would be up to the user to make sure that the linksys ip address and the openslug ip address was the same (assuming they want that). Jun 02 05:45:19 but yeah, unslung could also store other data in the second half of sysconf Jun 02 05:45:53 <[g2]> I'd guess that jbowler-zzz would just put a magic number in there and search for it Jun 02 05:46:02 however, one problem with not putting it as part of the linksys format is that the linksys apps may rewrite the whole 128k block and blow away the openslug stuff when someone temporarily flashes unslung Jun 02 05:46:29 <[g2]> that may be a problem Jun 02 05:46:41 if it was put in as part of the linksys sysconf format (which is versatile enough for anything you want to do), then it would be always preserved. Jun 02 05:48:21 <[g2]> I agree that would have been one clean implementation, however jbowler went the devio route and /etc/default/sysconf route. I'm certainly not complaining about all the work he's done Jun 02 05:48:27 so you could go from stock to openslug to unslung back to openslug without loosing any config data Jun 02 05:49:05 can devio do the linksys sysconf format? Jun 02 05:49:18 <[g2]> not quite Jun 02 05:49:28 can it be extended? Jun 02 05:58:56 <[g2]> ok... that was a little odd... The outlet went bad and the PC went down Jun 02 05:59:10 <[g2]> A quick check didn't seem to have any of the breakers flipped Jun 02 06:00:20 devio seems to be turing-complete, so it should be able to handle the linksys sysconf format Jun 02 06:00:57 <[g2]> it looked like it to me, but jb said he tried and id didn't seem to work Jun 02 06:01:02 <[g2]> dunno why Jun 02 06:01:23 <[g2]> I'd think it'd be pretty easy to get correct Jun 02 06:01:43 <[g2]> just dd mtdblock1 and produce something identical Jun 02 06:02:13 <[g2]> could be he enhanced devio and hasn't tried again Jun 02 06:26:52 just compiles smarttools on directly on the slug, but it seems that its not possible to get access to the IEC-mode via USB, so no point in having smartctl on the slug. Jun 02 06:27:14 s/compiles/compiled Jun 02 06:28:54 03rwhitby 07 * r1.3398 10openembedded/packages/nslu2-binary-only/ (3 files in 2 dirs): Added upgrade-inhibit.htm, which prevents entering maintenance mode when a rootfs is loaded from an external disk Jun 02 06:29:00 that needs testing ... Jun 02 06:29:26 http://pastebin.ca/13177 Jun 02 06:30:15 dzlabing: could be an endianness issue? Jun 02 06:32:13 might be, but i believe that also on an x86-plattform smartctl fails when the disk is attached via USB, i just had forgotten that there is a basic problem (or maybe it works, but at the moment i dont have physical access to any USB-drive (nor the slug itself). Jun 02 07:49:37 rwhitby-asleep: the problem with putting rootfs support into the kernel is that it doesn't fit. Jun 02 07:50:00 it used to fit Jun 02 07:50:08 what's new? Jun 02 07:50:10 It's possible to squeeze ext2,jbd,ext3 and reiserfs in there, but not the ethernet support Jun 02 07:50:24 ah - reiser Jun 02 07:50:53 so if something goes in and the upgrade mechanism relies on it if effectively precludes normal users from using any other rootfs Jun 02 07:51:57 certainly NFS rootfs is impossible, even with a reconfigured kernel, because base-kernel+Intel-code doesn't fit... Jun 02 07:52:35 It doesn't seem correct to do all the work for NFS then come up with an upgrade path which doesn't support it. Jun 02 07:52:48 agreed Jun 02 07:53:20 so what is the failure mechanism for a bad ipkg upgrade? Jun 02 07:54:24 Go back to the flash - i.e. normally upgrade the rootfs. For a kernel upgrade find some way to upgrade the module tree too. Jun 02 07:54:28 it is simply that a jffs2 kernel upgrade has to always have a /lib/modules/ upgrade kept in sync in the jffs2 and the external rootfs? Jun 02 07:54:39 Right Jun 02 07:55:14 03jp30 * 10unslung/make/apr-util.mk: correct typo introduced with wiley compatibility Jun 02 07:55:16 The programs in /boot/foo have to work, but they really are a minimal set (ash, ifconfig for NFS) Jun 02 07:55:38 so can we do a postinst script for kernel modules which copies the module to flash too? Jun 02 07:55:56 or make the upgrade-kernel script do the modules too? Jun 02 07:56:04 Yes. Jun 02 07:56:28 But kernel-modules RDEPENDS openslug-kernel, not the other way round. Jun 02 07:57:00 Still, that's a generic problem - I believe at present it is possible to upgrade an OE kernel using ipkg without the modules being updated (not sure though) Jun 02 07:57:23 ok, at the moment, if you do an "ipkg upgrade" and there is a kernel upgrade, then you have to run /sbin/update-kernel afterwards to write it to the flash Jun 02 07:57:38 so why can't the update-kernel script just do the kernel module copying too? Jun 02 07:57:56 it will be run after both the kernel and all the modules have already been updated in the current rootfs Jun 02 07:58:19 We have to ensure that a rev of openslug-kernel installs all the modules too, but that's the only problem - the postinst should be easy. Jun 02 07:58:52 update-kernel isn't a postinst - it's a command that the user has to run Jun 02 07:59:18 and is anyone really going to get an update of openslug-kernel other than by typing "ipkg uprade" ? Jun 02 07:59:18 Ok, that works. Jun 02 07:59:34 (in which case you're guaranteed that both kernel and modules are updated) Jun 02 08:00:05 Yes. update-kernel can check for the new modules. Jun 02 08:00:54 My assertion is that update-kernel and conffiles is everything you require to never had to reflash the whole jffs2 again. Jun 02 08:01:00 devio: it's just a block based dd. It can write into the middle of a device and it handles the blocking, that's all. Jun 02 08:01:05 s/had/have/ Jun 02 08:01:24 yeah, I had a look at the source. nice. Jun 02 08:01:27 conffiles: I agree Jun 02 08:02:05 my goal is that you flash openslug once, and then do "ipkg upgrade" and "update-kernel" from there to enternity Jun 02 08:02:56 I agree, but what about jffs2 - if it isn't the rootfs my assumption is that updates should be kept to a minimum (for safety). Jun 02 08:03:31 right, only kernel and kernel modules should be updated in jffs2 Jun 02 08:03:51 (unless the user chooses not to turnup, in which case caveat emptor) Jun 02 08:04:33 the assumption is that all the binaries in the jffs2 rootfs will still run on a new version of the kernel Jun 02 08:04:49 so that if you remove the disk, you still get a bootable system Jun 02 08:05:09 (albeit with a very old rootfs, except for the kernel and kernel modules) Jun 02 08:05:41 I can't think of anything which will break ABI compatibility, except that modprobe/depmod/insmod may need an upgrade at some time. Jun 02 08:05:44 but you could then even ipkg upgrade the jffs2 with the disk removed Jun 02 08:06:07 so we can have update-kernel handle those special cases too Jun 02 08:06:13 That won't work reliably because it's very likely to fill the partition Jun 02 08:06:24 (I.e. jffs2 upgrades add /lib/module trees) Jun 02 08:06:42 yeah, that's true. Jun 02 08:06:59 Hum. One thing to be sure of is that upgrade kernel-module-foo doesn't remove the old one! Jun 02 08:07:20 That's because until the kernel is flashed the old one is required. Jun 02 08:07:42 no, we've ascertained that kernel modules that are already loaded can be removed from the disk Jun 02 08:08:17 they are locked into mem Jun 02 08:08:35 If I ipkg upgrade and get a new kernel and new modules, but then I don't run update-kernel my system will not reboot. Jun 02 08:08:47 (If the old modules have been removed). Jun 02 08:09:22 correct. you would need to remove the disk, reboot, hotplug, and then remember to run update-kernel Jun 02 08:09:57 upgrading kernel and modules in jffs2 would be an (almost) atomic procedure Jun 02 08:12:22 so either the jffs2 is consistent with old kernel and old modules (and the external disk is consistent with new kernel and new modules), or both jffs2 and external disk are consistent with new kernel and new modules) Jun 02 08:12:28 I guess it's safe when there are too rootfs's (jffs2 and something else). One should always match the current kernel. Jun 02 08:13:38 feel free to modify update-kernel to implement this stuff - I won't have time in the next few months to do it Jun 02 08:13:42 SysConf can have extra registry entries added (following the LinkSys format), but putting whole files in seems likely to fail. Jun 02 08:13:52 agreed Jun 02 08:14:11 It would be good to be able to preserve basic network settings - that means /etc/network/interfaces (probably nothing else). Jun 02 08:15:06 Still, that's only an issue for people who need to use different (USB) networking devices. Jun 02 08:15:49 ok, I gotta sleep Jun 02 08:16:01 good discussion - thx jbowler Jun 02 08:16:16 g'nite rwhitby Jun 02 08:30:51 03jp30 * 10unslung/make/apr.mk: prevent apr's libtool from hardcoding host paths into apache Jun 02 08:54:54 03jp30 * 10unslung/make/apr-util.mk: apr-util's configure doesn't understand --without-ldap Jun 02 08:55:38 03jp30 * 10unslung/make/apache.mk: apache: wiley tweaks; bump ipk version because of changes to apr Jun 02 09:23:29 03jp30 * 10unslung/Makefile: add PATCH_LIBTOOL Jun 02 09:24:35 03jp30 * 10unslung/make/svn.mk: svn: use PATCH_LIBTOOL Jun 02 09:25:24 03jp30 * 10unslung/make/apr.mk: apr: use PATCH_LIBTOOL Jun 02 10:16:09 hi there Jun 02 10:17:00 howdy Jun 02 10:28:19 03jp30 * 10unslung/make/apr.mk: tweak apr libtool once more Jun 02 11:07:30 ~seen byront Jun 02 11:07:41 byront <~byron-por@ByronT.nslu2-linux> was last seen on IRC in channel #nslu2-linux, 5d 15h 44m 25s ago, saying: 'np'. Jun 02 11:08:43 hmm, that was a while ago Jun 02 11:10:22 ~seen /me Jun 02 11:10:24 bofh_omega: i haven't seen '/me' Jun 02 11:10:26 ~seen bofh_omega Jun 02 11:10:27 bofh_omega is currently on #nslu2-linux (54m 21s). Has said a total of 3 messages. Is idling for 1s Jun 02 11:10:49 /eot Jun 02 11:21:13 today I bought an NSLU2. I've installed Openslug and stuck in a flashstick which I formatted as ext3 and mounted on /usr/local with noatime. are these correct steps towards setting it up as a development box, and if so where should I be looking next for installing some interesting software? Jun 02 11:32:19 ext2 should reduce the wearout quite a bit. Jun 02 11:32:37 (no journal) Jun 02 11:33:35 for software you have every choice you like ;) download source packages, get a copy of openembedded... Jun 02 11:33:55 what 'src' lines should I put in my ipkg.conf? Jun 02 11:34:47 hmm, try src nslu2-oe http://ipkg.nslu2-linux.org/feeds/openslug/oe (from the wiki) Jun 02 11:36:49 is there a wiki page explaining basic of ipkg? I didn't find one and am not familiar. wanting to install this stuff in /usr/local, for now I added 'dest local /usr/local' Jun 02 11:38:08 is mounting external storage on /usr/local the normal way to "expand" the installation, or does one typically replace the whole root partition with external storage? Jun 02 11:39:04 thanks for the ipkg tip. I have rather more packages to choose from now :) Jun 02 11:41:47 one option you have is to chroot into your flash building a truly separate system Jun 02 11:42:27 another is you "unslung" your box. means put nearly everything on your flash drive and add symlinks Jun 02 11:43:17 I don't know a ipkg howto other than ipkg --help :} Jun 02 11:43:36 thanks, i'll get myself into some trouble :) Jun 02 11:44:04 hehe thats quite common ;) Jun 02 11:46:31 I already knocked the little fella off the network by breaking /etc/networking. was quite happy to discover this "redboot" thing :) Jun 02 11:48:46 hehe, a serial port is also quite usefull for debugging failed boots and get the option of fixing it without redboot/network access Jun 02 11:49:50 take care not to fill your space internal flash space too much. This will lead to problems also. Jun 02 11:49:56 -space Jun 02 11:50:28 I'll try not to add anything to it now Jun 02 11:51:35 * lukego wrestles with his concience. is it a crime to install emacs on such a tiny winy device? :) Jun 02 11:52:51 hehe vi has some advantages in that environment ;) but I've also seen a windowmanager in the ipkg list Jun 02 11:53:07 so overkill _is_ possible ;) Jun 02 11:53:21 as at AddAPackageToUnslung described - more information about ipkg http://www.handhelds.org/moin/moin.cgi/Ipkg Jun 02 11:53:47 thanks cap Jun 02 11:53:54 np Jun 02 11:54:12 hi all. any nslu2-bittorent-users around? when i start some torrents i get a "torrent not registered with this tracker". know what that means? torrent works on a pc in the same network.. thx for your help! Jun 02 11:58:39 anyone know if eno-away's erlang package is available as a binary? Jun 02 11:59:08 lukego, it should be available from the native feed Jun 02 11:59:44 Wipmac: http://www.google.de/search?hl=de&ie=ISO-8859-1&q=torrent+not+registered+with+this+tracker&btnG=Google-Suche&meta= Jun 02 12:00:00 * bofh_omega uses no torrent sry Jun 02 12:00:14 is the native feed http://ipkg.nslu2-linux.org/feeds/openslug/oe ? Jun 02 12:00:30 eno's erlang work is for unslung Jun 02 12:00:38 compatible with openslug? Jun 02 12:02:39 i don't know what the status of using unslung packages with openslug is. you could try it Jun 02 12:18:44 no binutils in this package dir? Jun 02 12:52:47 can anyone recommend a USB wireless lan dongle to go with the nslu2? Jun 02 13:02:57 not wireless but rj45 USB200M Linksys works like a charm (if you get the ax8557x modul ;) Jun 02 13:03:07 have a nice day ... bye Jun 02 13:04:25 hm I got a big ugly netgear dongle instead :) Jun 02 15:16:02 03jp30 * 10unslung/make/bash.mk: made /bin/bash -> /opt/bin/bash link persistent over firmware upgrades Jun 02 15:42:40 03jp30 * 10unslung/sources/bash/rc.bash: missing bit of recent bash change Jun 02 18:03:02 03jp30 * 10unslung/ (7 files in 2 dirs): upstream upgrade php to 5.0.4 Jun 02 18:28:32 <[g2]> jp30 ping Jun 02 18:46:32 hi, [g2] Jun 02 18:47:11 guess what, unslung exposed a bug in php Jun 02 18:47:56 03jp30 * 10unslung/ (make/php.mk sources/php/zend_strtod.patch): workaround for zend_strtod bug Jun 02 18:50:56 jp30-work, :-) Jun 02 18:53:32 the php guys had forgotten about big endian arm. hah! Jun 02 18:58:07 [g2], i must have missed you - may catch you later from home Jun 02 19:03:39 how dare they! :-) Jun 02 19:10:39 <[g2]> jacques, did you connect the VCC for the JTAG, is it required ? Jun 02 19:15:53 [g2], not on the usb one Jun 02 19:16:24 but yes on the MAX based one Jun 02 19:16:32 sorry was in other room Jun 02 19:16:40 oops Jun 02 19:16:43 sorry misread Jun 02 19:16:47 thought you were talking about serial Jun 02 19:16:53 <[g2]> no JTAG Jun 02 19:17:22 I connected my JTAG just as it has on the wiki - I dunno if it's requireed, but I got VCC from bottom of board where serial port is (and GND) Jun 02 19:17:46 hmm I guess that's not exactly on the wiki Jun 02 19:17:52 but that's what I did and it worked :-) Jun 02 19:18:06 <[g2]> hmmmm maybe I need to connect that over Jun 02 19:18:13 <[g2]> you used a digilent cable right ? Jun 02 19:19:34 yep, the 2.8V one Jun 02 19:19:51 <[g2]> yeah... that's the one I've got Jun 02 19:20:02 <[g2]> I think all the pins are connected but the VCC Jun 02 19:20:25 <[g2]> I could jumper that over from the serial line Jun 02 19:25:03 yep, that's what I did Jun 02 19:25:31 <[g2]> OK thx Jun 02 19:37:55 <[g2]> jacques, did you add the 10K pulldown ? Jun 02 19:39:45 <[g2]> Hey Tiersten Jun 02 19:46:54 [g2], nope, I added no parts Jun 02 19:53:21 BIAB Jun 02 19:53:28 <[g2]> hmmm Jun 02 19:53:30 <[g2]> thx Jun 02 20:21:18 [g2] ping Jun 02 20:21:41 <[g2]> dyoung-web, morning :) Jun 02 20:22:28 <[g2]> dyoung-web, when you hooked up the digilent, did you use the 10K pull-down ? Jun 02 20:22:32 I need to couble check, but I think the vcc is required, because it provides the vccio voltage refernce for the chip inside the JTAG cable. Jun 02 20:22:51 so for a slug, use 3.3V. Jun 02 20:23:00 its all documented on my JTAG pictures page. Jun 02 20:23:07 I didnt use any additional parts. Jun 02 20:23:24 <[g2]> maybe I just need to hookup VCC better Jun 02 20:24:22 <[g2]> dyoung-web, I've got Good News / Bad News on the FAT slug Jun 02 20:25:05 just checked, the NL17SZ125 is powered by the pin1 VCC, so you need to connect it. Jun 02 20:25:28 <[g2]> THX Jun 02 20:25:42 or is it pin6.... well you get the idea. Jun 02 20:26:08 so whats the good news? Jun 02 20:26:42 <[g2]> serial, 64MB, APEX work GREAT Jun 02 20:26:50 okay, whats the bad news? Jun 02 20:27:18 <[g2]> The rework looks good to me, but it's 2X or 3X the cost I had hoped for :( Jun 02 20:27:38 oh, like $50 ? Jun 02 20:27:41 Plus Jun 02 20:27:48 Jun 02 20:27:55 stupid cgiirc. Jun 02 20:27:55 <[g2]> Plus :( and that's just the labor Jun 02 20:28:57 Makes sense to me, minimum .5hr charge right? Jun 02 20:29:29 maybe if you do 20 of them the per unit price will be less? Jun 02 20:29:38 <[g2]> well I'll be following up but right now it looks like it takes 2 hours Jun 02 20:29:50 <[g2]> I got 10, 25, and 50 piece pricing Jun 02 20:29:52 HANH? Jun 02 20:30:00 2hrs per unit?! Jun 02 20:30:08 what does that include? Jun 02 20:30:08 <[g2]> nod. Jun 02 20:30:30 <[g2]> serial, JTAG, and memory upgrade Jun 02 20:30:36 hmmm Jun 02 20:30:40 how well do you know these ppl? Jun 02 20:30:57 I'm rather suprised at the time estimate. Jun 02 20:31:01 <[g2]> brand new, just met them Jun 02 20:31:12 methinks they may be taking you for a ride Jun 02 20:31:30 <[g2]> I doubt that Jun 02 20:31:41 <[g2]> the did the first unit for free :) Jun 02 20:31:57 what takes so long? Jun 02 20:31:59 Its free the first time. Jun 02 20:32:06 are they stacking chips? Jun 02 20:32:14 <[g2]> no replace chips Jun 02 20:32:18 <[g2]> no replaced chips Jun 02 20:32:22 that's faster Jun 02 20:32:24 after that, its $60 for a bag. ;-) Jun 02 20:32:26 <[g2]> 8M16s with 16M16s Jun 02 20:32:33 well, I say ask around Jun 02 20:32:42 see what other ppl think about that time Jun 02 20:32:48 Ask ka6sox how long it took him to replace the chips. Jun 02 20:32:52 ppl who should know Jun 02 20:32:57 he just did one for me. Jun 02 20:33:07 well he's not a professional board reworker is he? Jun 02 20:33:12 (or is he?) Jun 02 20:33:30 I dunno, he has a rework station. Jun 02 20:33:40 but he may not be an expert at it. Jun 02 20:33:52 for ppl that rework boards all day long, I think 0.5 hours sounds about right Jun 02 20:33:55 <[g2]> ka6sox has always been very busy for me Jun 02 20:34:22 <[g2]> the SLUG I sent him in Jan/Feb is still AWOL Jun 02 20:34:30 I think 30min is not out of line for the chip replacement. Jun 02 20:34:51 <[g2]> I was thinking in the .5-1 hour range also Jun 02 20:35:00 <[g2]> but hell I'm SW guy Jun 02 20:35:15 <[g2]> it takes me like 3 hours to do the 4-pin header Jun 02 20:35:17 Well, someone was teaching how to stack. Jun 02 20:35:40 with a rework station removing the old parts should take around 30 seconds. Jun 02 20:36:01 <[g2]> then you should clean the holes Jun 02 20:36:02 pasting /heating the new one on shouldnt take more than 15min each. Jun 02 20:36:07 what holes? Jun 02 20:36:11 <[g2]> pads Jun 02 20:36:23 <[g2]> then inspection Jun 02 20:37:15 okay 10minutes for inspection. so maybe an hour. Jun 02 20:38:46 I think if I were spending that much, I'd spend a little more and get it with 128MB. Jun 02 20:39:10 at those prices I'd do the serial and jtag myself Jun 02 20:39:12 <[g2]> I think if we're spending that much we should roll our own boards Jun 02 20:39:45 <[g2]> I think the avila boards would be cheaper Jun 02 20:40:13 around the same price right? Jun 02 20:40:20 except the avila has a minipci Jun 02 20:40:29 <[g2]> right Jun 02 20:40:33 insetad of USB2 Jun 02 20:40:36 <[g2]> and no usb2 Jun 02 20:41:14 <[g2]> to me $150 is a hard limit for developers Jun 02 20:42:00 <[g2]> I mean I spend $400 for a AMD64, Mobo and 512MB of 400Mhz DDR Jun 02 20:42:14 <[g2]> with a fast disk that does nearly 400Mbs Jun 02 20:42:55 <[g2]> especially with the new C7's and all the stuff around the corner Jun 02 20:47:30 wow, was just reading about it. Jun 02 20:47:38 the low power mode isnt too bad. Jun 02 20:48:28 <[g2]> my take is that this market is going to be limited on the upper-end by mini/nano/... -itx and even the PC market Jun 02 20:49:04 and power. Jun 02 20:49:22 unless someone starts making low power PC perpheral devices Jun 02 20:49:25 <[g2]> so except for custom applications or fanless low-power it's gonna get squeezed Jun 02 20:50:01 So, The Man wins again. Jun 02 20:50:19 <[g2]> nah.. Jun 02 20:52:34 <[g2]> dyoung-web, so what's your fatslug look like memory config wise ? Jun 02 20:52:51 <[g2]> are you running stacked 8M16 with an extra CS ? Jun 02 20:54:26 I just sent him 16M16's since stacking didnt seem to work so well. Jun 02 20:54:53 so it will be the single bank of 16M16. Jun 02 21:02:36 <[g2]> hey dyoung-web and jacques can you run the following for me ? Jun 02 21:02:40 <[g2]> dd if=/dev/zero of=/dev/null bs=8k count=125000 Jun 02 21:02:47 on what target? Jun 02 21:02:51 <[g2]> slug Jun 02 21:03:01 <[g2]> 125000+0 records in Jun 02 21:03:01 <[g2]> 125000+0 records out Jun 02 21:03:01 <[g2]> real 0m 16.07s Jun 02 21:03:01 <[g2]> user 0m 0.23s Jun 02 21:03:01 <[g2]> sys 0m 15.85s Jun 02 21:03:39 <[g2]> it runs about 2.5 seconds faster on the avila Jun 02 21:04:30 Unslung: real 5.007s, 3.330s, sys=5.730s Jun 02 21:04:53 4.730 even Jun 02 21:04:55 <[g2]> Huh ? Jun 02 21:05:13 time dd if=/dev/zero of=/dev/null bs=8k count=125000 Jun 02 21:05:13 125000 0 records in Jun 02 21:05:13 125000 0 records out Jun 02 21:05:13 real 0m5.007s Jun 02 21:05:13 user 0m0.330s Jun 02 21:05:15 sys 0m4.730s Jun 02 21:05:32 <[g2]> you've got the bs=8k in there right ? Jun 02 21:05:39 see above. Jun 02 21:06:55 <[g2]> Hmm..... Jun 02 21:07:25 <[g2]> Sounds like something isn't setup right Jun 02 21:07:44 wheres your rootfs? Jun 02 21:07:56 <[g2]> memstick Jun 02 21:08:02 <[g2]> /dev/sda1 Jun 02 21:08:09 hard drive for me. wonder if that makes a differnce. Jun 02 21:08:30 <[g2]> try swapoff Jun 02 21:09:18 The wiley does this: Jun 02 21:09:21 real 0m 14.72s Jun 02 21:09:21 user 0m 0.68s Jun 02 21:09:21 sys 0m 14.04s Jun 02 21:10:11 <[g2]> P4 does this :) Jun 02 21:10:15 <[g2]> real 0m0.267s Jun 02 21:10:15 <[g2]> user 0m0.047s Jun 02 21:10:15 <[g2]> sys 0m0.219s Jun 02 21:10:50 <[g2]> I think the amd64 will be faster too Jun 02 21:13:08 after turn off swap I get similar results. Jun 02 21:13:40 <[g2]> hm... Interesting Jun 02 21:13:54 I dont have a openslug box to try this on right now. Jun 02 21:14:01 not one I can get to anways. Jun 02 21:14:07 <[g2]> ok Jun 02 21:16:31 rwhitby-web: Can you re-forward me that zd1211 stuff please? Jun 02 21:16:35 I never received it. Jun 02 21:16:49 but you should recieve a serial cable in around 5 days. Jun 02 21:21:12 <[g2]-amd64> real 0m0.191s Jun 02 21:21:12 <[g2]-amd64> user 0m0.032s Jun 02 21:21:12 <[g2]-amd64> sys 0m0.156s Jun 02 21:23:03 dyoung: msg me the email address to use Jun 02 21:23:22 dyoung-web: how much do I owe you? msg me the paypal address too Jun 02 21:28:10 dyoung-web: resent to your rr.com address Jun 02 21:28:19 (previous one was to gmail.com) Jun 02 21:40:10 [g2], ping? Jun 02 21:40:18 <[g2]> jp30, pong Jun 02 21:40:30 you wanted to talk to me earlier? Jun 02 21:40:42 <[g2]> I was wondering about running some apache/thttp/php benchmarks on a 64MB slug Jun 02 21:41:29 hmm... apache, mysql, and php could probably benefit from more real ram Jun 02 21:42:11 the only testing i've down on them has been pretty informal though Jun 02 21:44:03 what would you like to know? i can suggest some applications that might benefit from the extra ram if that would help Jun 02 21:44:41 <[g2]> well you've worked with it a bunch more than have Jun 02 21:45:20 <[g2]> just chatting about an approach and usnlung versus openslug testing I think would be a good start Jun 02 21:46:18 Ah, okay I havnt check the gmail for a while.. Jun 02 21:46:34 <[g2]> It's a big area IMHO, maybe we could start a wiki page and ppl could contribute ideas and test scenarios ? Jun 02 21:46:50 seems like a good idea. Jun 02 21:47:17 <[g2]> I'd also like to benchmark the avila IXP422 with 64MB Jun 02 21:47:49 one thing i've found is that mediawiki's mmcache is useless because of lack on memory on unslung / slug. would be interesting to know if it would work better on a fat slug Jun 02 21:48:22 <[g2]> nod. I'd think we'd want to start with some basic disk benchmarks Jun 02 21:49:04 <[g2]> surely flashstick, cornice drive, 7200 RPM drives are going to perform differently Jun 02 21:49:06 Was there interseted in a fatslugged unslung? Jun 02 21:49:46 <[g2]> dyoung-web, a funny thing happened with my fatslug testing today Jun 02 21:50:03 It sucked too much power and the brick cooked? Jun 02 21:50:24 <[g2]> I booted with APEX and the kernel wasn't passed a memory parameter and it knew to use 64MB Jun 02 21:50:30 [g2] agree it would be nice to know about disk vs flash stick performance. slug as diskless webserver is a great application Jun 02 21:50:47 that sounds about right.... Jun 02 21:51:04 its inline with what dsaxena said earlier. Jun 02 21:51:08 <[g2]> which means with Unslung and no paramter unslung would run with 64MB right and APEX Jun 02 21:51:24 supposedly. Jun 02 21:51:45 also supposedly for the dyoung modifed redboot. Jun 02 21:52:04 if people wish to keep redboot recovery as an option. Jun 02 21:52:20 <[g2]> nod Jun 02 21:52:48 <[g2]> it's kinda funny to me because I've flashed APEX 50+ times and I STILL don't have JTAG on the slug Jun 02 21:53:26 <[g2]> I did have RedBoot replacment on the avila and it was all tested out there first Jun 02 21:53:30 [g2] you must be not your own friend. "friends don't let friends" and all that Jun 02 21:53:58 <[g2]> jp30 I had it on the avila and all was tested there first, and then in memory on the slug Jun 02 21:54:52 <[g2]> I may be just a VCC pin away from it right now.. there's some more checking needed Jun 02 21:55:22 dyoung-web: how much for the cable and shipping? Was it $15 or so ? Jun 02 21:57:20 jp30, heads up: although not ready for testing, i got erlang cross built Jun 02 21:57:46 eno, good news. that will save 16 hours of processor time on the official build machine Jun 02 21:58:03 let me know when it's ready to test Jun 02 21:58:05 heh Jun 02 21:58:22 it's not as difficult as i thought Jun 02 21:59:12 basically build a host version first. I'm wondering this general technique can be used for other hard to cross-build packages Jun 02 21:59:56 if you can make it work for emacs and xemacs, i will be very thankful Jun 02 22:00:54 so the next long weekend is july 4th, i may take a shot at these then Jun 02 22:00:56 eno, that technique is widely used, for example by oe Jun 02 22:01:08 emacs however is another matter entirely Jun 02 22:01:25 it uses dump Jun 02 22:01:31 yeah, the whole dumping thing Jun 02 22:01:36 ulch Jun 02 22:01:58 thank rms Jun 02 22:02:06 Anyone have good or bad reports about Powerware UPS's ? Jun 02 22:02:09 im sure it seemed like a good idea at the time Jun 02 22:02:29 not an emacs user myself, what is the "dump" Jun 02 22:02:33 ? Jun 02 22:02:58 I'm looking at this one: http://www.powerware.com/UPS/5110_UPS.asp Jun 02 22:03:22 to sit near the floor of slug-central and have all the power packs plugged into it. Jun 02 22:03:37 <[g2]> jp30 so if you get a minute can you jot down and note or two about the web server benchmarking and configs ? Jun 02 22:04:16 <[g2]> rwhitby-web, winner-take-all they all stand or all fall Jun 02 22:04:20 <[g2]> :) Jun 02 22:04:42 rwhitby-web, I'm not familiar with that brand Jun 02 22:05:01 [g2], i'm not totally sure what you want. advice about where the bottlenecks are i can give; mechanical ways of testing for speed i know little about Jun 02 22:05:32 <[g2]> jp30, ok. Jun 02 22:05:47 jacques: the form factor is the big thing. I need something with normal outlet sockets that can handle plug packs, not the standard IEC sockets that you get on most UPS's. Jun 02 22:05:48 eno, i think what emacs is doing is storing an image of its internal state in full machine dependent glory to disk Jun 02 22:05:51 <[g2]> I'm sure will get it worked out over time Jun 02 22:06:24 [g2], would it help if i wrote up a bit on the wiki about how i configured apache, php, eaccelerator and why? Jun 02 22:06:42 <[g2]> jp30, *absolutley* Jun 02 22:06:46 rwhitby-web, it looks nice in the photo - and yes, mine have 4 regular and 2 "wall wart" sockets Jun 02 22:06:48 ok, will do Jun 02 22:06:51 <[g2]> THX Jun 02 22:08:01 jp30, thx for the explanation Jun 02 22:09:15 it does it to reduce list file parsing at startup Jun 02 22:09:18 lisP Jun 02 22:10:14 jacques, i take it there's no way to get an emacs to write out a dump in a foreign byte-order / word-size / alignment? Jun 02 22:10:40 I've never heard of one Jun 02 22:10:51 there is an explanation here: http://www.xemacs.org/Documentation/21.5/html/lispref_66.html Jun 02 22:11:02 hmm. it is like nethack then. "great minds think alike" Jun 02 22:11:36 hmm wan't that one of stallman's too? Jun 02 22:11:46 here's a thing Jun 02 22:12:04 too much trouble probably, but you could cross-build temacs, and then only do the dump stel on the slug Jun 02 22:12:10 dump steP Jun 02 22:12:14 i was just going to say Jun 02 22:12:24 "great minds think alike" :-D Jun 02 22:12:29 that's the technique i used for pear, too Jun 02 22:12:46 pear? Jun 02 22:13:44 PHP Extension and Application Repository Jun 02 22:14:32 emacs takes a long time to compile and it's x11 support needs work, so it would be useful to have it cross-compiling Jun 02 22:16:15 jacques: what UPS do you have? Jun 02 22:24:17 rwhitby-web, I have three APC Back-UPS XS 1500 Jun 02 22:25:19 I've always bought that brand (APC) not like thei'r the best in the worls or anything, but they've worked fine - I've probably have about 7 of their devices - I bought these because I wore the batteries out on my last (smaller) ones Jun 02 22:25:49 damn I can't type tonight Jun 02 22:28:10 the batteries wore out after about 5 years Jun 02 22:28:15 of constant use Jun 02 22:28:29 the 1500's are new as of last year Jun 02 22:28:32 (to me) Jun 02 22:29:42 yeah, I have a height restriction, so form factor is the main thing Jun 02 22:29:45 ok, back later Jun 02 22:30:26 <[g2]> nite all **** ENDING LOGGING AT Thu Jun 02 23:59:57 2005