**** BEGIN LOGGING AT Thu Mar 15 02:59:58 2012 Mar 15 07:01:22 build #132 of rb532 is complete: Failure [failed compile_10] Build details are at http://buildbot.openwrt.org:8010/builders/rb532/builds/132 Mar 15 09:56:17 juhosg * r30944 /trunk/target/linux/generic/patches-3.2/ (8 files): linux/3.2: build mdio_register_board_info into the kernel if PHYLIB is selected Mar 15 09:56:18 juhosg * r30945 /trunk/target/linux/generic/patches-3.3/ (8 files): linux/3.3: build mdio_register_board_info into the kernel if PHYLIB is selected Mar 15 09:56:21 juhosg * r30946 /trunk/target/linux/ (6 files in 6 dirs): ar71xx: move micrel PHY driver to the generic linux target Mar 15 10:19:52 nbd * r30947 /trunk/package/netifd/files/lib/network/config.sh: netifd: implement find_config() Mar 15 10:19:56 nbd * r30948 /trunk/package/netifd/Makefile: netifd: fix ipv6 issues, add sit tunnel support Mar 15 10:19:59 nbd * r30949 /trunk/package/6to4/ (6 files in 3 dirs): 6to4: add netifd support Mar 15 10:24:47 ah, nice. Mar 15 10:25:10 guess I'll be retrying netifd again tonight. Mar 15 10:26:48 cool, let me know if you run into any issues with it Mar 15 10:27:23 will do, I'll have to port my 6rd.sh though Mar 15 10:27:46 but it's based on 6to4.sh, shouldn't be too hard now I have an example for netifd too ;) Mar 15 10:27:51 right Mar 15 10:28:02 and the new protocol handler script format is much easier than the old one Mar 15 10:28:20 since netifd handles all of the core stuff Mar 15 10:28:25 I had looked at it before but couldn't find how to add new proto Mar 15 10:28:36 guess I missed the 6to4.hotplug part Mar 15 10:28:56 the hotplug part hasn't changed much Mar 15 10:29:11 and it's not necessary for protocols that directly attach to a device Mar 15 10:29:49 at some point i want to get rid of those hotplug scripts entirely, when i've implemented a proper replacement Mar 15 10:30:41 right now I also have config 'interface' 'vpn' for tun0 (openvpn) in /etc/config/network, will that still work with netifd ? Mar 15 10:31:08 it's with proto none, so I guess yes Mar 15 10:31:13 yes Mar 15 10:31:33 ok, looking forward to testing this again Mar 15 10:32:04 I tried before but as my whole network is v6 enabled, not having v6 uplink breaks part of the internet :) Mar 15 10:32:13 makes sense :) Mar 15 10:32:24 yeah Mar 15 10:32:36 i haven't really done anything with ipv6 yet Mar 15 10:32:56 ah, ic :) Mar 15 10:33:16 i'm hoping i can implement the missing functionality for netifd this week Mar 15 10:33:20 so i can enable it by default soon Mar 15 10:33:36 my first tunnel was in 2001 or 2002. had to use usagi kernels since the vanilla kernel stopped routing v6 after a short time Mar 15 10:33:37 i'm really looking forward to getting rid of the old scripts Mar 15 10:33:49 where are the days :) Mar 15 10:34:01 ok, I'll see what gives tonight and let you know Mar 15 10:34:22 I can test both 6to4, and 6rd Mar 15 10:34:30 ok Mar 15 10:34:45 could even cleanup the 6rd.sh script a bit and send it to ML, dunno if it would be interesting Mar 15 10:34:47 i don't have a 6to4 endpoint yet, so i could only verify that the interface and the route is there Mar 15 10:34:56 and i see the tunneled packets in tcpdump Mar 15 10:35:22 when you're done with 6rd.sh, i'll review it Mar 15 10:35:26 and commit it when it's done Mar 15 10:35:38 should it go in a different package? or just with netifd ? Mar 15 10:35:44 coz 6to4 is a diff package Mar 15 10:35:59 a different package at first Mar 15 10:36:16 at some point i'll probably merge all the ipv6 tunnel related scripts into one package Mar 15 10:40:48 this is what I have currently, be aware, it might be ugly Mar 15 10:40:49 http://stewie.be.tintel.eu/6rd.sh Mar 15 10:40:52 quick and dirty Mar 15 10:41:37 would be nice to integrate it with radvd or quagga, but that could require some nasty hacking Mar 15 10:42:03 as the prefix changes when your external v4 IP changes Mar 15 10:42:05 6to4 already has hackery like that Mar 15 10:42:14 and it is quite nasty Mar 15 10:42:32 yeah, but does it also *replace* the prefix in radvd.conf ? Mar 15 10:42:40 that would be even uglier :/ Mar 15 10:42:43 it uses uci state vars for that Mar 15 10:42:47 in a very ugly way Mar 15 10:42:48 ah Mar 15 10:42:55 but I guess it works :) Mar 15 10:43:01 with quagga that might be different Mar 15 10:43:06 fixing this will have to wait until netifd is enabled by default Mar 15 10:43:16 ah, interesting. Mar 15 10:43:18 but i have a plan for fixing it Mar 15 10:43:30 i intend to allow protocol handlers to attach bits of information to the interface Mar 15 10:43:34 which can be queried via ubus Mar 15 10:43:35 with quagga maybe one could use vtysh or something Mar 15 10:43:51 to reconfigure the advertised prefix Mar 15 10:44:25 i have a different plan for an initial fix Mar 15 10:44:35 it depends on having config reload support in init scripts Mar 15 10:44:44 of which most of it can be implemented in the core Mar 15 10:45:06 the idea is to share the code between reload() and start() Mar 15 10:45:07 so /etc/init.d/something reload that does something like kill -HUP ? Mar 15 10:45:18 but have reload check if the new command line and the new generated config files are identical to the old ones Mar 15 10:45:24 and restart the service if they aren't Mar 15 10:45:33 ah Mar 15 10:45:49 that way it can be properly implemented with any service, not just the ones that have -HUP or similar stuff Mar 15 10:46:06 once netifd is made default, it also pulls in ubus Mar 15 10:46:19 so i'm planning to implement a 'process manager' daemon Mar 15 10:46:28 which will simplify reload handling Mar 15 10:46:35 by keeping the state and tracking the processes Mar 15 10:46:43 yeah ubusd is already in my builds Mar 15 10:46:56 it'll have no configuration, all commands are sent over from init scripts Mar 15 10:46:57 didn't bother disabling it when disabled netifd by default Mar 15 10:48:42 eventually i want to be able to make any change to any config file and simply tell the system to reload its configuration Mar 15 10:49:05 and it should just work, regardless of whether i used a cli or hand-edited the configuration Mar 15 10:49:41 and i want it to be fast too ;) Mar 15 10:52:24 sounds cool. can't wait till it's finished :) Mar 15 10:52:46 so if I can help with testing the 6to4 and maybe other stuff, I'd be happy to Mar 15 11:14:40 ah great, too fast. rebuilt without enabling netifd Mar 15 11:17:35 nbd: I was trying to port host-sflowd(.sf.net) to openwrt, had to enable libubacktrace in uclibc. is that something that could be done via menuconfig, or some dependency maybe ? Mar 15 11:24:40 stintel: why does it depend on ubacktrace? Sounds like something that can be easily patched out Mar 15 11:30:11 jow_laptop: well, it doesn't compile without. I'm not really a programming hero like you guys though, so I just looked for bactrace support in uclibc and found libubacktrace. when I enabled it, compiling hsflowd succeeded Mar 15 11:32:16 nbd * r30950 /trunk/package/6to4/files/6to4.sh: 6to4: set the tunnel remote endpoint to any, put the gateway in the route instead Mar 15 11:33:58 ah, crap. I mixed up, again Mar 15 11:34:07 I use 6in4, not 6to4. lol Mar 15 12:05:45 stintel: do you have a log of the compile failure? Mar 15 12:12:18 jow_laptop: not anymore Mar 15 12:12:25 ok Mar 15 12:12:40 can you paste the makefile? Mar 15 12:12:50 I see if the backtrace stuff can be removed Mar 15 12:14:40 jow_laptop: it's ugly. I did some ugly hacks too Mar 15 12:14:42 just to make it work Mar 15 12:15:15 I'm not worried about its quality, just want something to build the flow thing without starting from scratch Mar 15 12:15:22 currently idling at work Mar 15 12:15:48 k Mar 15 12:16:48 argh Mar 15 12:16:54 * stintel facedesks Mar 15 12:17:46 the makefile is gone :) Mar 15 12:18:48 argh, and I never locally committed it, so it's really gone. ffs Mar 15 12:19:29 ok I'll try myself then Mar 15 12:19:47 got an url for the source tarball? Mar 15 12:19:54 and how is it called exactly? Mar 15 12:20:22 jow_laptop: http://downloads.sourceforge.net/project/host-sflow/REL-1_20/hsflowd-1.20.tar.gz?r=http%3A%2F%2Fsourceforge.net%2Fprojects%2Fhost-sflow%2Ffiles%2FREL-1_20%2F&ts=1331814015&use_mirror=netcologne Mar 15 12:20:34 project page is http://host-sflow.sourceforge.net/ Mar 15 12:21:06 it hardcodes cc and stuff Mar 15 12:21:23 damnit. must find a way to restore those files Mar 15 12:22:23 let's give extundelete a try Mar 15 12:24:57 build #128 of x86 is complete: Success [build successful] Build details are at http://buildbot.openwrt.org:8010/builders/x86/builds/128 Mar 15 12:28:43 lame Mar 15 12:36:24 heh and seems enabling libubacktrace messes up other stuff: Mar 15 12:36:25 Sep 8 17:44:33 wrt1 user.info sysinit: quagga.init: Starting zebra ... /usr/sbin/zebra: can't load library 'libubacktrace.so.0' Mar 15 12:43:59 yes Mar 15 12:44:05 I'd rather avoid to enable it Mar 15 12:44:29 so throwing it in here was the right thing to do after all :) Mar 15 13:04:50 http://luci.subsignal.org/~jow/hsflowd.patch Mar 15 13:06:27 jow_laptop: neat. Mar 15 13:08:26 let me know if it works, I'll commit it then Mar 15 13:08:56 jow_laptop: will it go in package/ or in feeds/packages ? Mar 15 13:09:03 guess that doesn't matter to test it though Mar 15 13:10:27 feeds/packages/net/ Mar 15 13:10:41 for testing oyu can put it in package/ Mar 15 13:11:02 ah, I already put it in feeds/packages/net (where I initially put it myself) Mar 15 13:11:14 then git reset --hard, without local committing Mar 15 13:11:28 guess that's what I get for not using git often enough :) Mar 15 13:12:11 I am actually curious how hsflowd will integrate with opensvwitch Mar 15 13:12:20 currently playing around with both, on a gentoo box @ home Mar 15 13:59:29 hi Mar 15 13:59:39 does someone of you know a good c++ profiler? Mar 15 13:59:47 does someone of you know a good and simple c++ profiler? :) Mar 15 14:00:00 any recommendations Mar 15 16:14:18 jow_laptop: the hsflowd seems to be running. I have an init script for it, and /etc/config/hsflowd too. but very basic Mar 15 16:18:47 the init script for rhel has various ways to get an uuid from somewhere (dmidecode, /dev/disk/by-uuid, blkid), but those aren't usable on most embedded devices I guess Mar 15 16:35:30 Hello ! I start to add support for asus wl 330n/330n3g device. I now have a working kernel for it. Mar 15 16:35:48 Which way should I go to publish my patch for review ? Mar 15 16:39:11 starox: mail it to the -devel mailinglist. but hold on a sec Mar 15 16:39:42 http://wiki.openwrt.org/doc/devel/add.new.platform Mar 15 16:39:49 stintel: thanks :) Mar 15 16:40:01 or http://wiki.openwrt.org/doc/devel/add.new.device Mar 15 16:40:44 starox: you might want to read one of those pages and see if you followed the "guidelines" Mar 15 16:41:06 This is what I was looking for : guidelines. Mar 15 16:45:02 stintel: You make my day, I was looking for modifying base-files Mar 15 16:47:30 starox: glad I could help :) Mar 15 16:47:35 :) Mar 15 16:48:56 need to look into these docs myself Mar 15 16:50:08 I did the wrong way code first, then read :) Mar 15 16:50:20 a friend bought this 99 euro nas, and it's based on Marvell 6281 Mar 15 16:50:28 so it shouldn't be too hard to get openwrt running on it Mar 15 16:51:46 Yep ! linux kernel has greatly improved for adding new hardware, and openwrt is easy to understand. Mar 15 16:51:59 but last time I tried I didn't even understand how to hook up the ttl to serial converter Mar 15 16:52:44 WL330N is on ramips, you just have to define some leds and flash, and it works. Howerver I found something wrong for the wifi eeprom in the driver. Mar 15 16:53:21 stintel: The wiki on openwrt documents well the way of cabling serial port. Mar 15 16:54:07 http://wiki.openwrt.org/doc/hardware/port.serial Mar 15 16:54:35 was just looking at it :) Mar 15 16:55:35 parallel port hmmm Mar 15 16:55:42 (reading on jtag too) Mar 15 17:09:10 hm Mar 15 17:09:12 eglibc 2.15 Mar 15 17:09:20 I'm seeing flock bugs in 2.13 Mar 15 17:09:36 stintel: what NAS did your friend get? Mar 15 17:09:51 question; anyone know of non kernel module packages that get built out of the kernel tree? Mar 15 17:10:10 compat-wireless ? Mar 15 17:10:33 the new usbip userspace code is being developed in the kernel. looking for examples of how to build packages that depend on kernel source. https://dev.openwrt.org/ticket/10992 Mar 15 17:10:40 stintel: i'll check it out thanks Mar 15 17:11:07 things like tcpdump depend upon kernel headers and such Mar 15 17:13:22 rektide: a verbatim powerbay 4 Mar 15 17:13:53 rektide: and tabfail ;) Mar 15 17:27:54 nbd * r30951 /trunk/target/linux/generic/files/drivers/net/phy/ar8216.c: Mar 15 17:27:54 ar8216: fix a MTU related regression Mar 15 17:27:54 Switch reset on AR8316 appears to clobber the MTU configuration register Mar 15 17:27:54 and possibly other global config registers. Move global configuration Mar 15 17:27:54 register init writes back to the reset callback. Mar 15 17:30:55 stintel: there is /proc/sys/kernel/random/uuid - that one is also used by miniupnpd Mar 15 17:31:09 root@uplink:~# cat /proc/sys/kernel/random/uuid Mar 15 17:31:09 c0aef476-d666-4621-b8de-ac7b0e33999b Mar 15 17:33:57 stintel: what are the cpu MHz, DDR and flash size? Mar 15 17:34:40 jow_laptop: hsflowd writes files to /etc .. might not be good for the flash. the paths are "hardcoded" though (#define) Mar 15 17:35:00 jow_laptop: and uuid -> nice. is it consistent across reboot ? Mar 15 17:35:08 lol, probably not Mar 15 17:35:13 just noticed the random :) Mar 15 17:35:23 you can use it to seed a persistant id Mar 15 17:35:49 e.g. /etc/hsflowd/uuid Mar 15 17:36:14 this way users also have the ability to deploy specific ids Mar 15 17:36:38 ok. since upnp also uses this, can we have a system-wide uuid ? seed it during firstboot ? Mar 15 17:37:04 would probably make sense Mar 15 17:37:15 in /etc/config/system ? Mar 15 17:37:42 nbd: what do you think? Mar 15 17:38:41 putting an "option uuid '...'" from /proc/sys/kernel/random/uuid into /etc/config/system, config system on firstboot Mar 15 17:39:39 note to self: git stash! Mar 15 17:39:46 should use it more often Mar 15 17:44:35 maybe it could also be used by hostapd for wps ? Mar 15 17:44:53 the hostapd config example suggests that when also running upnp on the same host, the uuid's are the samre Mar 15 17:44:59 same* Mar 15 17:46:57 jow_laptop: i think may be better to generate a uuid from things that do not change, even over reflashes Mar 15 17:47:21 i.e. mac addresses, maybe hash of a board config partition Mar 15 17:49:58 makes sense. that's how hostapd does it when not configured at all Mar 15 17:50:36 but a wifi mac address can change ;) so better do it on the lowest (or highest) onboard mac ? Mar 15 17:58:17 stintel: probably only the boardconfig partition Mar 15 17:58:36 but since that stuff is wildly inconsistent I don't see a good way to do that yet Mar 15 17:58:53 on x86 it probably needs dmidecode or similar Mar 15 17:59:23 stintel: I Mar 15 17:59:50 'd say do the uuid handling within the package first, its a bigger endeavour to implement a system wide uuid Mar 15 17:59:51 yes, the hsflowd init script for rhel tries to get it from dmidecode first Mar 15 18:00:37 ok Mar 15 18:02:10 when finished I just send it to the ml, including your patch ? Mar 15 18:02:44 well make a new patch including everything and send that to the list Mar 15 18:02:59 ok Mar 15 20:02:58 nbd * r30952 /trunk/target/linux/generic/files/ (drivers/net/phy/swconfig.c include/linux/switch.h): swconfig: use a mutex instead of a spinlock, many swconfig calls sleep Mar 15 21:02:05 nbd: pong Mar 15 21:02:19 Chocks: peng Mar 15 21:02:34 why is the ag300h LED mapping so minimal? Mar 15 21:02:43 no idea Mar 15 21:02:58 * Chocks peers at the copyright notice Mar 15 21:03:04 it's clearly your fault! Mar 15 21:03:06 um Mar 15 21:03:21 is it PCI/ath wired like the G300NH2? Mar 15 21:03:30 no idea Mar 15 21:03:50 who knows? Mar 15 21:04:11 what leds are missing? Mar 15 21:04:19 maybe you can find them in the dd-wrt source Mar 15 21:04:35 hm Mar 15 21:04:36 i don't have time to look into that stuff at the moment Mar 15 21:05:23 the LEDs that are there now are diag and the phy0/phy1 Mar 15 21:05:38 none of the others on G300NH/2 Mar 15 21:05:49 * Chocks will peer at dd-wrt Mar 15 21:16:34 * Chocks is going to feel dirty, but here goes Mar 15 21:17:27 "Building DD-WRT from source is difficult and according to the text here definitly not working on first try. You will see lots of strange errors and many confusing install-scripts" Mar 15 21:17:36 that's from the dd-wrt wiki. seriously? Mar 15 21:17:46 yeah, not many people succeed in building it Mar 15 21:18:03 and i wouldn't recommend trying it, if you like to keep your sanity ;) Mar 15 21:18:13 I happen to be a master of build systems, but this just looks like bad news Mar 15 21:20:56 my main problem with OpenWrt's build system is that it's hella slow. that's largely make's fault, AFAICT Mar 15 21:22:12 i find it to be fast enough for all of my purposes, but i use shortcuts when i only want to update specific things Mar 15 21:22:29 it gets slow when you install all packages from the feed though Mar 15 21:22:35 so i only install what i need Mar 15 21:22:58 * Chocks runs comedy dd-wrt check out script Mar 15 21:23:18 i'd recommend checking out only parts of it Mar 15 21:23:23 aye Mar 15 21:23:37 otherwise you're looking at grabbing gigabytes of crap from that svn server Mar 15 21:23:41 ;) Mar 15 21:24:04 it's full of kernel trees in various versions for various platforms Mar 15 21:24:09 seems so Mar 15 21:24:16 I have no idea where to look Mar 15 21:24:23 must be in src/router somewhere Mar 15 21:24:41 maybe services, libutils, rc, sysinit or something like that Mar 15 21:24:47 no idea where exactly Mar 15 21:28:34 ugh Mar 15 21:30:06 SVN not browseable Mar 15 21:32:09 it is browseable through trac Mar 15 21:32:18 http://svn.dd-wrt.com/ Mar 15 21:39:12 I'm crying Mar 15 21:40:13 Chocks: man up Mar 15 21:40:25 good call Mar 15 21:40:39 any recomendations for a cheap vps? Mar 15 21:40:48 yes Mar 15 21:40:51 linode Mar 15 21:41:02 best customer service ever. right here on freenode. Mar 15 21:41:56 does not seem cheap... Mar 15 21:42:15 $240/year I think I'm paying Mar 15 21:42:25 they don't come much cheaper Mar 15 21:44:06 there are cheaper, you can find for 10$ per month Mar 15 21:44:18 but i dont know if they are any good Mar 15 21:44:29 yes, I suppose it depends what features you want Mar 15 21:44:50 but I don't know anyone else who offers instant help on IRC Mar 15 21:45:02 not that I've used that more than once in 4 years Mar 15 21:46:56 yes Mar 15 21:47:21 what happens if you take down ssh or lock your self out with iptables Mar 15 21:47:32 how can you get to the machine? Mar 15 21:47:37 there's an ajax console Mar 15 21:47:54 or you can make another system on another partition to access it Mar 15 21:47:59 do you know is that a "standard" fature for other vps providers as well Mar 15 21:48:09 the ayax console? Mar 15 21:48:09 various, but most of them these days have consoles Mar 15 21:48:15 various = varies Mar 15 21:48:22 ok Mar 15 21:48:41 I went through a couple of bad ones. one of them didn't even tell me about it until I complain about problems with their VM Mar 15 21:48:56 ./src/router/services/sysinit/sysinit-wrt400.c FWIW Mar 15 21:49:08 actually, that's not kerenl Mar 15 22:02:54 * Chocks has no idea Mar 15 22:05:47 * Chocks mans up and checks out everything Mar 15 22:11:27 im fine with paying 30 a month for a vps, wouldnt be caught dead paying 10/mo ever Mar 15 22:13:28 Chocks: you won't find the gpio stuff in the kernel Mar 15 22:13:35 Chocks: dd-wrt has some user space hackery for that Mar 15 22:14:47 ah, right Mar 15 22:15:04 perhaps I was looking at the right file after all then Mar 16 02:31:04 jow * r30953 /packages/libs/uclibc++/ (Makefile patches/010-cstring_conflicts.patch): [packages] uclibc++: remove conflicting inline declarations Mar 16 02:33:16 jow_laptop: thank you! Mar 16 02:33:33 anyone seen thepeople lately? Mar 16 02:33:44 briefly. hes busy Mar 16 02:34:11 ah. was hoping he could add alix2 to the list of builds... Mar 16 02:35:43 so far he only builds default subtargets due to lack of buildslaves Mar 16 02:43:23 and none of them can be retired? I have to figure that at least one of them can be deprecated... Mar 16 02:43:58 plus I offered to make cycles available on the 4-core (Phenom/FX Quad), RAID, 8GB build machine... he never followed up with me. Mar 16 02:45:00 yeah, soe organization bottleneck. but he runs this by himself so I cannot really comment on it Mar 16 02:45:22 I need sleep, bbl Mar 16 02:53:34 ok, night! **** ENDING LOGGING AT Fri Mar 16 02:59:58 2012