**** BEGIN LOGGING AT Sun Oct 31 02:59:58 2010 Oct 31 03:21:08 xMff: ping Oct 31 07:23:20 hlu: nice. did you add tagging stuff to the switch driver already? Currently it's just one big ethernet card Oct 31 10:10:55 good morning everybody - hope the Europeans here noticed the switch to winter time (standard time)? Oct 31 10:16:10 woke up pleasantly surprised Oct 31 10:21:47 heh Oct 31 10:31:43 sunday morning, i'm waking up. can't even focus on a coffee cup... Oct 31 10:43:15 :) Oct 31 10:45:09 I noticed in Luci rdate seems to be still a default setting. Oct 31 10:45:20 Isn't rdate hopelessly obsolete? Oct 31 10:45:20 http://en.wikipedia.org/wiki/Time_Protocol Oct 31 10:45:34 is there a reason to keep rdate? Oct 31 10:45:52 (size of implementation vs. full ntp client)? Oct 31 10:46:30 yes, it's included in busybox Oct 31 10:52:46 hlu: pin Oct 31 10:52:52 g Oct 31 10:53:11 hlu: let me know when u are online i can explain what needs to be done to make pci work on amazon Oct 31 10:54:58 blogic: morning, you up early Oct 31 10:55:13 hmm Oct 31 10:55:21 i am normally awake around 8 Oct 31 10:55:58 blogic: it's sunday :-) Oct 31 10:56:16 i am a freelancer, i have no concept of week days Oct 31 10:56:18 :D Oct 31 11:00:46 loswillios: ok... so it is a busy box thing. thx Oct 31 11:42:54 [florian]: ping Oct 31 11:55:03 <[florian]> KanjiMonster: pong Oct 31 12:33:29 nbd * r23730 /trunk/package/madwifi/patches/472-remove_11n_devids.patch: madwifi: remove device ids for AR5416 Oct 31 12:40:52 nbd * r23731 /branches/backfire/package/madwifi/patches/464-remove_11n_devids.patch: madwifi: merge ar5416 devid removal from r23730 Oct 31 12:41:49 hauke * r23732 /branches/backfire/package/crda/Makefile: Oct 31 12:41:49 crda: fix circular dependency Oct 31 12:41:49 Backport from r22304 Oct 31 12:51:46 nbd * r23733 /branches/backfire/package/ (90 files in 8 dirs): merge mac80211, iw, hostapd from trunk to backfire Oct 31 12:57:34 in the default dnsmasq settings there is an entry "option resolvfile '/tmp/resolv.conf.auto'" . Why not /etc/resolv.conf? Oct 31 12:57:47 does the startup script for dnsmasq copy it over? Oct 31 12:58:24 /etc/resolv.conf points at 127.0.0.1 Oct 31 12:58:34 so that local dns lookups go through dnsmasq Oct 31 12:58:48 pointing dnsmasq at /etc/resolv.conf would create a loop ;) Oct 31 13:00:26 doh, of course :) thx. I could have thought of this hehe Oct 31 13:00:42 same trick as with bind Oct 31 13:01:04 well, bind doesn't need an upstream server Oct 31 13:01:11 true Oct 31 13:01:13 re Oct 31 13:01:19 it can take one as forwarder Oct 31 13:13:33 nbd * r23734 /packages/devel/oprofile/Makefile: oprofile: add a separate package for extra utilities Oct 31 13:15:13 florian * r23735 /trunk/target/linux/generic/patches-2.6.32/ (21 files): [kernel] refresh 2.6.32 patches Oct 31 13:15:27 florian * r23736 /trunk/target/linux/ar7/patches-2.6.32/ (7 files): [ar7] refresh 2.6.32 patches Oct 31 13:15:32 florian * r23737 /trunk/target/linux/ar7/patches-2.6.32/160-vlynq_try_remote_first.patch: [ar7] vlynq: try remote clock first, then external (#7324) Oct 31 13:54:11 nbd * r23738 /trunk/package/hostapd/patches/ (330-madwifi_merge.patch 460-oper_state_fix.patch): hostapd: move the madwifi segfault fix to the right place Oct 31 13:54:19 nbd * r23739 /trunk/package/ (ifxmips-dsl-api/Makefile uboot-lantiq/Makefile): add missing SH_FUNC for remaining md5s calls (#7888) Oct 31 13:58:25 nbd * r23740 /branches/backfire/package/hostapd/patches/ (330-madwifi_merge.patch 460-oper_state_fix.patch): hostapd: merge operstate cleanup from r23738 Oct 31 14:36:53 nbd * r23741 /trunk/package/mac80211/patches/300-nl80211_dump_crash_fix.patch: nl80211: fix a station dump crash triggered by sending an invalid ifindex Oct 31 14:37:46 nbd * r23742 /branches/backfire/package/mac80211/patches/300-nl80211_dump_crash_fix.patch: mac80211: backport the nl80211 crash fix from r23741 Oct 31 15:21:03 greetings. I've written a small program and an openwrt Makefile which I'd like to see integrated in the distribution Oct 31 15:22:27 tommie: submit a patch to the list Oct 31 15:22:50 I've checked out the openwrt repository as well as the packages one, I guess I have to create the patch relative to the packages one? Oct 31 15:22:56 How do I do a testcompile of that? Oct 31 15:23:51 tommie: https://dev.openwrt.org/wiki/SubmittingPatches Oct 31 15:35:38 ok, does that one look suitable? http://wertarbyte.de/temp/0001-package-add-triggerhappy-hotkey-daemon.patch Oct 31 15:38:27 nbd * r23743 /trunk/rules.mk: fix usage of flock - use the command mode instead of locking a file descriptor to fix build issues on mac os x Oct 31 15:41:25 nbd * r23744 /trunk/rules.mk: remove an obsolete comment for the flock template Oct 31 15:43:01 loswillios: no, at first we want to setup wireless, because we want to use it as a wireless community network Oct 31 15:43:16 blogic: pong Oct 31 15:54:16 [florian]: ping Oct 31 15:59:00 [florian]: unping, found it myself :) Oct 31 16:28:44 updated openwrt/upstream, https://home.comcast.net/~sdwalker/uscan/uscan.shtml Oct 31 17:03:47 Hierarchical RCU implementation. Oct 31 17:03:47 RCU-based detection of stalled CPUs is disabled. Oct 31 17:03:47 Verbose stalled-CPUs detection is disabled. Oct 31 17:03:59 could this be the origin of my low performance Oct 31 17:09:16 danitool: unlikely, I get normal performance on my 6348 (e.g. ~1 MB/s when scp'ing from my device), and it has this too Oct 31 17:09:39 at least I assume 1 MB/s with scp is normal ;) Oct 31 17:10:20 :| Oct 31 17:11:31 my higher transfer is 700K but I know with the original firmware is at least 1M or 2M Oct 31 17:11:58 what does your /proc/interrups say? perhaps something is generating too many interrupts causing the cpu to be busy with acking them? Oct 31 17:13:27 http://pastebin.com/TstdtRTT Oct 31 17:14:38 well, now I don't have powered on the router, so this is when I started the router Oct 31 17:14:54 I see Oct 31 17:20:28 looks normal Oct 31 17:25:17 this is after transferring 120 M through ethernet http://pastebin.com/u4wsFE2T Oct 31 17:44:03 hm, I downloaded the packages repository, added the appropiate directory to feeds.conf and I can see the packages in menuconfig, but how do I compile a package from that feed? "make package/$FOO-compile V=99" did not work (no rule to make target...) Oct 31 17:44:36 slash instead of hyphen Oct 31 17:44:57 where? $FOO/compile? Oct 31 17:45:11 make package/foo/compile Oct 31 17:45:17 doesn't work, either Oct 31 17:45:17 yes, there is only one in your command ;) Oct 31 17:45:32 then foo is not called foo Oct 31 17:45:48 often one source package provides several virtual ones Oct 31 17:46:01 what package do you try to compile? Oct 31 17:46:31 anyone Oct 31 17:46:55 well make package/foo/compile definitely works Oct 31 17:47:04 at first of course my own one, which worked fine when I placed it in the openwrt/package directory Oct 31 17:47:25 if it does not in your case, the package is either not present, or not spelled correctly or the makefile is faulty Oct 31 17:47:34 hm, just a second Oct 31 17:49:18 I have no idea what changed, but it worked Oct 31 17:50:40 BTW, is there a reason that there is no handler directory for the input subsystem in /etc/hotplug2.rules? Oct 31 17:50:54 danitool: looks normal. Have you tried pinpointing the source of the low performance, by e.g. comparing cpu speed, memory speed, etc? You could also try to use oprofile and see if you can find something taking all cpu when transferring Oct 31 17:51:03 I have a program that I want to launch on input device hotplugging (USB keypad, bluetooth keyboard etc.) Oct 31 17:52:07 KanjiMonster: i tried to build openwrt with oprofile, but images are too large to fit in the flash, btw the problem is not only with transfer rate, it is with all applications Oct 31 17:52:21 SUBSYSTEM ~~ (net|button|input|usb|platform|ieee1394|block|atm) { Oct 31 17:52:38 I added "input" there myself, any reason why it was excluded? Oct 31 18:02:14 tommie: most likely because there was no demand for it yet Oct 31 18:02:25 tommie: did you actually test the input handler dir? Oct 31 18:02:43 xMff: yes, it works after adding it Oct 31 18:02:50 ok Oct 31 18:03:09 xMff: I placed a handler for my hotkey daemon there, and after connecting a usb keypad I was able to use it at conce Oct 31 18:04:26 http://github.com/wertarbyte/openwrt-packages/commit/7f1ba828252c622ebe827568b70e50b41854c948#diff-2 Oct 31 18:04:38 works :) Oct 31 18:07:01 btw, you should install the hotplug handler and config with INSTALL_DATA Oct 31 18:07:10 that will give them 0644 instead of 0755 Oct 31 18:07:25 ah ok Oct 31 18:08:45 xMff: but the hotplug handler with 0644 would be fatal Oct 31 18:08:49 xMff: since it lacks +x Oct 31 18:09:04 not it does not need to be executable Oct 31 18:09:09 why not? Oct 31 18:09:16 is it sourced? Oct 31 18:09:20 yes Oct 31 18:09:44 I see Oct 31 18:10:04 cat /sbin/hotplug-call Oct 31 18:11:43 xMff: http://github.com/wertarbyte/openwrt-packages/blob/master/triggerhappy/Makefile that should do it Oct 31 18:11:54 yes Oct 31 18:12:05 does anyone know if it is possible to let github create tar-files without those SHA-prefixes? Oct 31 18:12:40 no, but you could override PKG_UNPACK to have more control over the resulting build dir Oct 31 18:12:57 well, it doesn't really matter Oct 31 18:13:16 example: http://luci.subsignal.org/trac/browser/luci/trunk/contrib/package/olsrd-luci/Makefile Oct 31 18:13:26 but its kind of annoying to adjust this build-dir variable everytime Oct 31 18:13:43 ah, nice Oct 31 18:14:19 didn't know about that option Oct 31 18:14:28 how does tar handle name collisions? Oct 31 18:14:39 name collisions? Oct 31 18:15:06 xMff: you have a tar file with "foo/file" and "bar/file" - now I strip the first path component during extraction Oct 31 18:15:35 ah, no idea I suppose the later one overwrites the previous one Oct 31 18:25:48 xMff: thx, works great Oct 31 18:30:54 xMff: when submitting a new package Makefile as a patch, what would be the most useful reference point? as a standalone feed? relative to the packages-Repository? Oct 31 18:33:24 -p0 against any dir is fine Oct 31 18:33:38 the committer has to place it in the right directory anyway Oct 31 18:34:05 git running git-format-patch is probably sufficient? Oct 31 18:34:11 so Oct 31 18:34:12 probably Oct 31 18:34:18 as long as it applies with patch Oct 31 18:34:29 I personally do not use git Oct 31 18:34:55 xMff: any idea why my build is breaking? Oct 31 18:35:06 philipp64|laptop: eglibc ? no, sorry. Oct 31 18:35:14 not sure if the "svn update" or the "yum update" caused it. Oct 31 18:35:39 philipp64|laptop: http://geekandpoke.typepad.com/geekandpoke/2010/10/simply-explained-svn-guinea-pig.html Oct 31 18:38:45 hmmmm.... my gcc and binutils didn't change, just my glibc. Oct 31 18:38:57 so I'm going to guess it wasn't yum. Oct 31 18:45:58 I'll try building against uClibc-0.9.32 instead. Oct 31 18:46:36 keep in mind that eglibc and glibc targets are still experimental, they receive very little attention Oct 31 18:47:41 who's been committing to it? Oct 31 18:47:57 or who owns it? Oct 31 18:47:58 I think {Nico} cared a bit after it Oct 31 18:50:22 is there some kind of batch command to compile a package for multiple (all?) architectures? Oct 31 18:52:23 nope Oct 31 19:05:39 would be nice if there were.... for people making top-level changes it would simplify finding unintended damage. Oct 31 19:31:54 hm, anyone willing to test my new package? Oct 31 19:38:55 yup, it's eglibc that's broken. Oct 31 20:15:22 build #18 of kirkwood is complete: Success [build successful] Build details are at http://tksite.gotdns.org:8010/builders/kirkwood/builds/18 Oct 31 21:50:10 florian * r23745 /trunk/target/linux/ar7/patches-2.6.32/ (160-vlynq_try_remote_first.patch 930-titan-platform.patch): [ar7] refresh 2.6.32 patches Oct 31 21:50:31 florian * r23746 /trunk/target/linux/ar7/patches-2.6.32/170-clocks_init_earlier.patch: [ar7] initialize clocks earlier for lpj to be correct (#7548) Oct 31 22:03:56 [florian]: brcm63xx's wl-exports patch seems to need refreshing, too (applies with some fuzzyness for me) Oct 31 22:04:45 <[florian]> KanjiMonster: ok Oct 31 22:22:23 florian * r23747 /branches/backfire/target/linux/ar7/patches-2.6.32/140-uart_port_ar7.patch: backport r21457 Oct 31 22:22:55 florian * r23748 /branches/backfire/target/linux/ar7/patches-2.6.32/130-clocks_devices_init.patch: backport r21404, r21409 Oct 31 22:23:29 florian * r23749 /branches/backfire/target/linux/ar7/patches-2.6.32/160-vlynq_try_remote_first.patch: backport r23737 Oct 31 22:23:51 florian * r23750 /branches/backfire/target/linux/ar7/patches-2.6.32/170-clocks_init_earlier.patch: backport r23746 Oct 31 22:40:48 what is the status on parallel builds at the moment? I noticed the paralle build option. What could break actually if you just did a make j $num_cpus Oct 31 22:41:39 ? Oct 31 22:41:49 You need to do both to get optimal performance Oct 31 22:42:07 Set the config option and call the master make with -j Oct 31 22:42:35 The config option parallelizes the packages build and the main make's -j parallelizes the packages Oct 31 22:43:10 If you have enough RAM, you can set both to the number of CPUs. Oct 31 22:44:14 [florian]: prepare for inbound patches ;) Oct 31 22:45:39 mb_: ok i see Oct 31 22:46:40 YOu may also enable the options to parallelize the default rule and toolchain. They should be pretty safe now. Oct 31 22:50:05 ok Oct 31 22:50:09 i 'll try Oct 31 22:54:55 <[florian]> KanjiMonster: ok, got them :) Oct 31 22:56:56 florian * r23751 /trunk/target/linux/brcm63xx/ (patches-2.6.32/ patches-2.6.33/): [brcm63xx] remove empty directories Oct 31 23:06:51 is there a list of make targets kicking around with some brief description of use? Oct 31 23:08:37 <[florian]> make help should do that Oct 31 23:08:43 <[florian]> but it is out-of-sync imho Oct 31 23:15:51 ok thanks Oct 31 23:34:24 florian * r23752 /packages/net/openl2tp/ (Makefile patches/001-makefile.patch): [package] update openl2tp to 1.7 (#8141) Oct 31 23:34:40 florian * r23753 /packages/utils/hd-idle/Makefile: [package] update hd-idle to 1.01 (#8152) Oct 31 23:34:42 florian * r23754 /packages/net/mdnsresponder/ (3 files in 2 dirs): [package] update mdnsresponder to 214.3.2 (#8083) Oct 31 23:34:44 florian * r23755 /packages/libs/postgresql/ (3 files in 3 dirs): [package] update postgresql to 9.0.0 (#7999) Oct 31 23:34:47 florian * r23756 /packages/utils/lvm2/Makefile: [package] lvm depends on libncurses (#7953) Nov 01 00:24:42 jow * r23757 /trunk/package/uci/ (Makefile patches/130-lua_fix_nested_foreach_delete.patch): Nov 01 00:24:42 [package] uci: use a delete-safe iterator in the Lua :foreach() function Nov 01 00:24:42 The following test case exposed a NULL pointer dereference in uci_lua_foreach(): Nov 01 00:24:42 lua -luci -e 'x=uci.cursor(); x:foreach("firewall", "forwarding", function(s) x:delete("firewall", s[".name"]) end)' **** ENDING LOGGING AT Mon Nov 01 02:59:57 2010