**** BEGIN LOGGING AT Wed Jan 06 02:59:57 2010 Jan 06 06:05:17 nbd * r19051 /trunk/ (2 files in 2 dirs): Jan 06 06:05:17 ixp4xx: allocate all skbs using GFP_DMA Jan 06 06:05:17 this ensures that dma bounce buffers will not be used frequently when bridging or routing frames between ethernet and wifi. Jan 06 06:05:17 significantly improves ath9k performance and obsoletes an ath specific patch Jan 06 09:36:00 hcg * r19052 /trunk/package/wpa_supplicant/patches/150-Remove-PS3-gelic-legacy-wpa-support.patch: wpa_supplicant: remove legacy PS3 gelic wpa driver support Jan 06 12:36:39 what is the recommend way to automate the building process? Jan 06 12:37:06 saving a configuration and patching the the filesystem before creating the image Jan 06 12:37:34 patching the filesystem? Jan 06 12:37:39 why? Jan 06 12:38:13 I don't know if it is the right word:) Jan 06 12:38:22 making changes to the config files, for example Jan 06 12:38:44 you can create a files/ directory in the top-level build system directory Jan 06 12:38:53 but that will replace? Jan 06 12:38:55 anything you put there will be copied to the filesystem after all packages have been installed Jan 06 12:39:03 so it can override anything Jan 06 12:39:25 ok, but there is no mechanism to only change some lines, and keep the rest unchanged? Jan 06 12:39:34 no Jan 06 12:39:39 well, not on the host Jan 06 12:39:54 ok, I think I can survive Jan 06 12:39:54 you could add a /etc/uci-defaults/ file Jan 06 12:40:05 which contains a few uci set and uci commit commands Jan 06 12:40:12 it will only be run on the first boot Jan 06 12:40:19 and then automatically be deleted Jan 06 12:41:01 owh, ok Jan 06 12:41:16 but anyway, I can just replace the files Jan 06 12:41:36 a hotplug event for firstboot && restart will always put the system back at factory settings? Jan 06 12:44:27 I am planning to do a package for that, with config section for Button and Delay Jan 06 12:45:20 yes Jan 06 12:48:35 Does something like that already exist? is that a stupid idea? I don't get why there isn't such thing already done:S Jan 06 12:49:17 there is a button-hotplug package Jan 06 12:49:30 which is configurable Jan 06 12:52:23 sorry, kmod-button-hotplug? Jan 06 12:53:16 no Jan 06 12:53:27 oh, wait Jan 06 12:53:43 yes, for some reason there's a separate kernel package for that Jan 06 12:54:27 i forgot where i saw the configurable scripting stuff, though Jan 06 12:54:40 must have been in a different branch or something Jan 06 12:56:47 well, I'm doing a very simple ash script on top of kmod-button-hotplug Jan 06 12:57:00 and it will take the config section from /etc/config/system Jan 06 12:57:14 button/action/timeout Jan 06 12:57:44 I feel that this would help a lot beginners Jan 06 12:59:21 ok Jan 06 13:31:22 the first action on a button on the wrt160nl doesn't give a hotplug event Jan 06 13:31:26 also they are inverted Jan 06 13:32:01 so usual a push is a ACTION=released, and the release is ACTION=pressed Jan 06 13:32:27 but the first time I press and release, I only get a ACTION=pressed Jan 06 14:58:16 so the /etc/uci-defaults/ is a folder that will on boot run all the scripts and then delete them? Jan 06 14:58:40 and usually used to had uci sections to config files that already exist? Jan 06 15:03:58 but how does that handle a package install via opkg? Jan 06 15:04:43 do I need to restart to get the settings apllied? Jan 06 15:23:59 juhosg * r19053 /trunk/target/linux/generic-2.6/ (3 files in 3 dirs): Jan 06 15:23:59 kernel: use generic thread_info allocator on MIPS Jan 06 15:23:59 * Fixes kernel panic if SLUB_DEBUG is enabled. Jan 06 15:24:02 juhosg * r19054 /trunk/target/linux/ar71xx/files/drivers/net/ag71xx/ (ag71xx.h ag71xx_debugfs.c ag71xx_main.c): ar71xx: add detailed interrupt statistics for the ag71xx driver Jan 06 15:24:05 juhosg * r19055 /trunk/target/linux/ar71xx/files/arch/mips/ar71xx/mach-wrt400n.c: ar71xx: don't include nvram.h for the WRT400N Jan 06 16:04:52 ping xMff: using opkg on r18953, I get this error when trying to upgrade my custom scripts package: http://openwrt.pastebin.ca/1740276 Jan 06 16:05:16 xMff: works ok if I use --force-overwrite Jan 06 16:05:48 netprince: looks okay to me Jan 06 16:06:07 I mean, it acts as intended Jan 06 16:06:56 maybe it shouldn't check for clashes once the package is installed but updated versions could introduce new files that would clash with other packages so I think the behaviour is okay Jan 06 16:07:36 ok, sounds fine to me too Jan 06 16:08:05 as long as --force-overwrite actually works :) Jan 06 16:08:51 although it would be nice to not have to force-overwrite just because its an upgrade, right? what if there are actually other conflicts? Jan 06 16:09:54 I think opkg should never overwrite files belonging to other packages unless forced to do so Jan 06 16:10:23 and tracking which overwrites were already approved and which not is fairly complex# Jan 06 16:10:27 that file doesn't belong to other packages though, it was included in my package, first install didn't report it either Jan 06 16:10:45 is the firewall package installed? Jan 06 16:11:00 I use fully custom firewall scripts Jan 06 16:11:15 okay, so the uci firewall is not installed` Jan 06 16:11:19 ? Jan 06 16:11:23 leme check Jan 06 16:12:36 yes it is installed Jan 06 16:12:49 then the behaviour is 100% okay Jan 06 16:12:50 it just gets overwritten when I install my package Jan 06 16:12:54 yep Jan 06 16:13:06 remove "firewall" and the problem will go away Jan 06 16:13:07 but I dont get a warning when doing a first install ?? Jan 06 16:13:16 this is indeed unusual Jan 06 16:13:52 I can uninstall the uci firewall package and see what happens... Jan 06 16:14:01 was opkg this particular version when you first installed "customizer" ? Jan 06 16:14:42 not sure what you mean... I did a fresh install yesterday, installed customizer then, then tried to do an upgrade today Jan 06 16:14:51 ok Jan 06 16:15:09 did you update the package lists before installing "customzier" ? Jan 06 16:15:16 yeah Jan 06 16:15:20 ok Jan 06 16:15:30 then I'll report this as bug to upstream Jan 06 16:15:48 ok, it still works, I just thought it was a bit strange Jan 06 16:16:56 you can work-around that by renaming your init script and disable the original one in postinstall Jan 06 16:17:08 if you want Jan 06 16:17:58 I should probably do that anyway just to simplify things Jan 06 16:18:22 good idea, thanks Jan 06 16:19:57 np Jan 06 17:35:53 hi all! a question about mac80211 (b43 or ath9k) APs: Jan 06 17:36:22 when 'iw dev wlan0 station dump' for every station there is an 'inactive time' Jan 06 17:36:58 I have APs with inactive times of 83003536 which is like 23hrs Jan 06 17:37:31 is the inactive time hardcoded somewhere and what is the current limit? Jan 06 18:07:37 doesnt changeset 18957 break the broadcom 2.4 build? since the modules dont have the _generic suffix there? Jan 06 18:42:52 juhosg is around? Jan 06 18:43:03 nbd * r19056 /trunk/package/mac80211/patches/550-queue_stop_fix.patch: mac80211: merge some further fixes to the queue stop patch Jan 06 18:43:04 latest patch just broke ar71xx Jan 06 18:44:07 https://dev.openwrt.org/changeset/19054/trunk Jan 06 18:46:45 nico * r19057 /trunk/package/kernel/modules/ (other.mk video.mk wireless.mk): [package] kernel: remove even more calls to CompareKernelPatchVer Jan 06 18:50:39 nico * r19058 /trunk/package/kernel/modules/crypto.mk: [package] kernel: cleanup crypto stuff and fix module names on 2.4 Jan 06 18:51:11 <{Nico}> netprince: thanks for pointing ^^^ Jan 06 18:51:32 np Jan 06 19:12:45 nunojpg: a small typo actually Jan 06 19:14:05 nunojpg: http://openwrt.pastebin.com/md5ef541 Jan 06 19:19:12 yes, I also looked Jan 06 19:19:33 but I didn't tought It would be usefull to submit a patch:) Jan 06 19:42:13 [ -n "$${IPKG_INSTROOT}" ] Jan 06 19:42:16 what is this for? Jan 06 19:43:48 to check whether running within the sdk or on the target Jan 06 19:44:52 why does it "run on the sdk"? Jan 06 19:46:15 when the buildroot preinstall packages into the rootfilesystem during build it will run each packages reinst/postinst scripts as well Jan 06 20:16:08 jow * r19059 /trunk/target/linux/ar71xx/files/drivers/net/ag71xx/ag71xx.h: [ar71xx] fix typo after r19054 Jan 06 21:13:19 xMff: might there possibly be a gwak ipkg ? Jan 06 21:14:30 xMff: ermmm gawk Jan 06 21:41:48 trunk failed to compile ar71xx http://pastebin.com/m4b531435 Jan 06 21:43:49 xxiao: svn up, I belive your issue is fixed in r19059 Jan 06 21:43:55 believe* Jan 06 21:46:39 i svn-ed up already, maybe i need build from scratch, wait Jan 06 21:46:55 compiled fine here, maybe you need make clean idd Jan 06 21:47:07 make target/linux/clean && make Jan 06 21:48:39 looks good now. thanks Jan 06 22:18:14 could someone be so kind to submit my patch https://dev.openwrt.org/ticket/6081? :D Jan 07 00:00:43 hey guys Jan 07 00:00:53 what is supposed to create /etc/hosts? Jan 07 00:01:03 for some reason dnsmasq doesn't and just complains about it missing Jan 07 01:22:38 gah Jan 07 01:22:43 how is this supposed to expand: "--dhcp-host=$mac,${name:+net:$name,}$ip" ? Jan 07 01:23:42 is ${name:+net:$name} right? Jan 07 01:24:04 +net means if name has a value net name=net Jan 07 01:24:08 I don't think so because of this: https://dev.openwrt.org/ticket/6459 Jan 07 01:26:11 what's is *supposed* to exand to? Jan 07 01:27:19 according to http://www.thekelleys.org.uk/dnsmasq/docs/dnsmasq-man.html it should not have "net:" Jan 07 01:27:44 it should just be --dhcp-host=00:22:15:64:A0:3D,yeti,10.2.2.10 Jan 07 01:28:01 as written, if name is blah it will be net:blah Jan 07 01:28:30 ok, thanks Jan 07 01:28:33 so net:yeti ? Jan 07 01:28:33 I don't know how ash works Jan 07 01:28:50 yes, right now, it's --dhcp-host=00:22:15:64:A0:3D,net:yeti,10.2.2.10 Jan 07 01:29:04 I just ran it in the shell...the second : fooled me...the first is a token but the second is literal text Jan 07 01:29:47 on what condition should the net: be applied, and what not? Jan 07 01:30:00 I don't think it ever should be applied Jan 07 01:30:34 dhcp_host_add() doesn't have enough info (network name), as far as I can tell Jan 07 01:30:40 hmm....you sure about that? it was put in as net for a reason I think Jan 07 01:30:59 I'm trying to find the log for that file to understand that right now Jan 07 01:31:31 I'm really just guessing about how dnsmasq works, so take what I say witha grain of salt Jan 07 01:31:42 what does dnsmasq use the net: for? Jan 07 01:31:43 I don't even have a manual for the same version as on the router Jan 07 01:32:09 check $ROOTDIR/build_dir/target-blah/dnsmasq Jan 07 01:32:11 well, whatever this says, I don't understand it: http://www.thekelleys.org.uk/dnsmasq/docs/dnsmasq-man.html Jan 07 01:32:22 "The net: sets the network-id tag whenever this dhcp-host directive is in use" Jan 07 01:32:26 I don't know what a network-id is Jan 07 01:33:16 how about $SVN_CHECKOUT/kamikaze/package/dnsmasq? Jan 07 01:33:22 oops, nvm Jan 07 01:35:15 network-id is arbitrary tag you assign for use by in other options to limit options to a particular network-id Jan 07 01:35:49 e.g. a dhcp-range that's different for different interfaces Jan 07 01:35:55 ok Jan 07 01:36:20 well, currently dhcp_host_add() doesn't seem to gather that kind of info as it's arguments Jan 07 01:36:33 if you see a network-id and you don't also have a hostname you could have a problem Jan 07 01:36:50 well the interface name in UCI is what is probably used as the network-id Jan 07 01:37:00 right now, the init.d script is setting the network-id to what the hostname should be Jan 07 01:37:08 which is wrong Jan 07 01:37:24 which would explain your bug Jan 07 01:37:58 check out what the other dhcp-options use net: for and expect it to be Jan 07 01:38:18 e.g. dhcp-range Jan 07 01:38:42 I think someone just stupidly copy-pasted that bit Jan 07 01:39:08 from append args "--dhcp-boot=${name:+net:$name,}$filename,$servername,$serveraddress" Jan 07 01:39:28 or maybe not Jan 07 01:39:44 no, I thought it about it...I think that parts right Jan 07 01:40:23 just changing to append args "--dhcp-host=$mac,$name,$ip" Jan 07 01:40:25 fixes my bug Jan 07 01:40:36 per UCI config ($name) will have a net:value Jan 07 01:40:49 humm? Jan 07 01:40:53 maybe you should add a repeated value (leave net) Jan 07 01:40:55 where is this UCI config? Jan 07 01:41:07 /etc/config/dhcp Jan 07 01:41:41 what file are you hacking /etc/init.d/dnsmasq? Jan 07 01:42:02 --dhcp-host=$mac,$name,${name:+net:$name,}$ip ? Jan 07 01:42:16 yeah, I'm hacking /etc/init.d/dnsmasq Jan 07 01:42:18 directly on the router Jan 07 01:42:40 --dhcp-host=00:22:15:64:A0:3D,yeti,net:yeti,10.2.2.10 also works Jan 07 01:43:35 if someone really needs to specify the network-id Jan 07 01:43:43 there should be another directive in /etc/config/dhcp Jan 07 01:43:46 is yeti the hostname Jan 07 01:43:48 like option network Jan 07 01:43:53 yes, yeti is the hostname Jan 07 01:43:59 at least for me! Jan 07 01:44:05 erm...$name isn't hostname Jan 07 01:44:07 I dont' know about dnsmasq :) Jan 07 01:44:42 well Jan 07 01:44:45 that's debatable Jan 07 01:44:55 you might be right according to; http://wiki.openwrt.org/doc/uci/dhcp#static.leases Jan 07 01:45:22 but option name is a really ambiguous option Jan 07 01:45:44 plus, there is this: https://dev.openwrt.org/ticket/6426 Jan 07 01:45:44 you need to add a hostname option Jan 07 01:46:21 if I have "config host ", can I just read the from $1 or something like that? Jan 07 01:47:13 local hostname="$2 Jan 07 01:47:33 in dnsmasw config get varname "$cfg" uciname Jan 07 01:47:44 config_get rather Jan 07 01:47:58 I don't understand what you just said Jan 07 01:48:46 if in /etc/config/dhcp you have Jan 07 01:48:46 config host Jan 07 01:48:46 option hostname yeti Jan 07 01:48:46 option ip '6.6.6.6' Jan 07 01:49:11 you can use Jan 07 01:49:11 config_get hostname "$cfg" hostname Jan 07 01:49:20 in the dnsmasq init.d Jan 07 01:49:29 oh Jan 07 01:49:31 and the variable would be $hostname Jan 07 01:49:58 why not something like: Jan 07 01:50:02 config host yeti Jan 07 01:50:07 but what are you trying to do Jan 07 01:50:10 option name Jan 07 01:50:19 that would assign a name to a UCI section Jan 07 01:50:21 I'm trying to make my dhcp+dns work :) Jan 07 01:50:22 not what you want Jan 07 01:50:49 I personally don't care about network-id Jan 07 01:50:57 I mean are you trying to assign a specific address to a specific host without a /etc/hosts? Jan 07 01:50:59 but I'm just trying to fix the broken /etc/init.d/dnsmasq Jan 07 01:51:05 yes Jan 07 01:51:21 I want to replace /etc/ethers+/etc/hosts with just /etc/config/dhcp Jan 07 01:51:39 I'll do what you said Jan 07 01:51:53 actually there is a better way Jan 07 01:52:13 https://dev.openwrt.org/ticket/5017 Jan 07 01:52:25 use the -A option of dnsmasq Jan 07 01:52:47 it wasn't supported previously, but is now Jan 07 01:52:58 or rather when the ticket is incorporated it will be Jan 07 01:53:16 I don't see how that is different from --dhcp-host= Jan 07 01:53:42 and the problem is still in /etc/init.d/dnsmasq :) Jan 07 01:55:45 I also wonder what is the purpose of: [ -n "$name" ] && dhcp_option_add "$cfg" "$name" Jan 07 01:55:55 and if I should copy-paste that for $hostname Jan 07 02:01:18 bah Jan 07 02:01:34 /home/plouj/work/kamikaze/build_dir/target-mipsel_uClibc-0.9.30.1/dnsmasq-2.51/ipkg-brcm47xx/dnsmasq/etc/init.d/dnsmasq isn't version controled Jan 07 02:01:45 I don't see any 'dnsmasq' file which is Jan 07 02:02:04 I mean, I don't see what in the repository actually becomes /etc/init.d/dnsmasq Jan 07 02:02:44 $ROOTDIR/packages/dnsmasq/files Jan 07 02:02:53 dnsmasq.init Jan 07 02:02:56 ah Jan 07 02:04:02 the advantage of -A is that you're a) not messing around with the network-id which is useful for other things and b) if you apply the patch it just works (though you might want to not do the bit about eliminating local). That patch will soon be applied by Florian. Jan 07 02:04:56 also it allows to you assign names for entire lists of hosts, not just a single one Jan 07 02:05:22 (more easily that is) Jan 07 02:06:07 oh, wait Jan 07 02:06:14 I thought this was a patch for dnsmasq binary Jan 07 02:06:24 I didn't know it had an -A option Jan 07 02:07:08 umm Jan 07 02:07:17 so, how would I apply that patch to my router? Jan 07 02:07:27 which of the files do I download, first of all? Jan 07 02:07:53 nico * r19060 /trunk/tools/mpfr/Makefile: [tools] mpfr: add missing MD5SUM Jan 07 02:10:10 oh, if you want it right to the router simply use vi on the router (ssh in and edit) Jan 07 02:10:27 rather than applying the patch Jan 07 02:10:37 well, I want to apply it to my source Jan 07 02:10:39 first Jan 07 02:11:03 but I'm not sure which patch to use: https://dev.openwrt.org/attachment/ticket/5017/remove_local_and_add_address.patch or https://dev.openwrt.org/attachment/ticket/5017/remove_local_and_add_address2.patch Jan 07 02:11:10 the second one Jan 07 02:11:57 in $ROOTDIR Jan 07 02:12:08 patch -p1 < file.patch Jan 07 02:12:32 with a clean tree of course Jan 07 02:12:59 (or at least no edits on the same files) Jan 07 02:14:14 umm Jan 07 02:14:17 weird Jan 07 02:14:31 I'm too tired to tell why it's not applying Jan 07 02:15:06 just a sec Jan 07 02:15:56 probably edits on dnsmasq.init Jan 07 02:16:06 nope Jan 07 02:16:09 I svn reverted it Jan 07 02:16:35 http://fpaste.org/w1Qw/ Jan 07 02:16:44 what about package/dnsmasq/files/dhcp.conf Jan 07 02:16:50 reverted it too Jan 07 02:18:02 oh well, I'm toooo tired, so I'll just keep my fix Jan 07 02:18:12 they've been edited in svn since the patch was made Jan 07 02:18:21 now my printing works! Jan 07 02:18:25 :) Jan 07 02:18:29 that's always good Jan 07 02:18:47 yeah, it's never good when you can't ping your network printer Jan 07 02:19:00 thanks Jan 07 02:19:05 m[ Jan 07 02:19:07 np **** ENDING LOGGING AT Thu Jan 07 02:59:56 2010