**** BEGIN LOGGING AT Wed May 30 02:59:58 2012 May 30 03:26:55 anyone tried json-rpc under OpenWrt? May 30 03:26:59 http://luci.subsignal.org/trac/browser/luci/trunk/modules/rpc/luasrc/controller/rpc.lua <--- May 30 09:16:26 <_trine> nbd, are you here? May 30 10:14:04 hi, about the new netifd infrastructure: what happens when the ifup command is called? I've put a new script in /lib/netifd/proto which define the proto_wing_init_config and proto_wing_setup. May 30 10:16:37 the problem is that none of the two functions are called when i run ifup May 30 10:27:44 jogo * r32003 /trunk/target/linux/brcm63xx/base-files/etc/uci-defaults/network: bcm63xx: readd accidentially dropped network config for SPW303V May 30 10:27:47 jogo * r32004 /trunk/target/linux/brcm63xx/ (12 files in 4 dirs): May 30 10:27:48 bcm63xx: build images for 6328 reference boards with various flash sizes May 30 10:27:48 Build images for 96328avng and 963281TAN (untested) for three different May 30 10:27:48 flash sizes: 4MB, 8MB, and 16MB. May 30 10:27:48 Known issues: May 30 10:27:48 * Button setup unknown, thus no fail-safe available. May 30 10:27:49 * All other BCM6328 and BCM63XX issues May 30 11:53:45 Ticket #11552 is spam; you might want to kill it. May 30 11:57:20 mystica555: done May 30 11:57:46 cool; just reviewing some recent changes with the whole netifd/qos scripts to see if I am to expect bugs on my current build May 30 11:57:55 that stuck out like a sore thumb ;) May 30 12:10:06 * mystica555 is curious about the regdomain differences now that its using mac80211 internal regulatory stuff from Changeset 31954; does this affect the requirement for ATH_USER_REGD if I want to set the router to anything but the EEPROM regdomain? May 30 12:11:08 <_trine> you can still use crda and regulatory.bin for any other regdomain May 30 12:11:44 well, if you look at that changeset, it removes crda due to race conditions May 30 12:12:13 <_trine> it does not remove it unless you remove it in make menuconfig May 30 12:13:45 it removes it from the dependencies of mac80211, which would to me, by default, remove it unless manually selected otherwise.. May 30 12:13:53 maybe I'm understanding it wrong May 30 12:14:05 would +'mean to' me May 30 12:14:39 <_trine> do you mean manually selecting it in make menuconfig May 30 12:15:42 i mean if it is not manually selected, its removal as a mac80211 dependency (i believe) removes it from the build entirely May 30 12:16:01 <_trine> manually selected where ? May 30 12:16:05 in menuconfig May 30 12:16:08 sorry May 30 12:16:18 <_trine> well that what I said in the first place May 30 12:19:00 <_trine> there does seem to be a problem in trunk at the moment as it does not appear to give out dhcp ip's when you use udhcpc -i eth0 May 30 12:19:25 you said it does not remove it.. i'm saying that it would appear to me to do just that. May 30 12:19:29 as default May 30 12:19:43 ie, a clean 'make menuconfig' without any old config lying around May 30 12:19:54 because as of now, crda = n May 30 12:19:55 <_trine> well that might be right May 30 12:19:57 is default May 30 12:20:19 <_trine> but if you select crda you can use it May 30 12:20:24 in other news, am i missing something when "block-mount" does not select kmod-usb-storage ? May 30 12:21:06 or is there supposed to be a different dependency or is that just something i'm to get manually? May 30 13:56:12 anyone around with commit access? May 30 13:59:41 mystica555: block-mount could also be for the spi-mmc module, not just USB. so would not make sense i think for it to require USB on devices that may not even have it May 30 14:00:00 good point May 30 14:00:11 i'm horrible about remembering exactly everything i put into every build ive done.. May 30 14:00:22 so im likely remembering wrong or forgetting that i did a step, etc May 30 14:01:59 yea, there's a lot there May 30 14:02:53 indeed May 30 15:04:45 this channel is soooooo dead now... May 30 15:30:11 jow * r32005 /trunk/package/uhttpd/ (Makefile src/uhttpd.c): [package] uhttpd: fix wrongly applied sizeof() leading to writing beyound end of buffer and subsequent data corruption (#11557) May 30 15:36:51 <_trine> jow the current trunk does not give out dhcp ip address's when you use udhcpc -i eth0 May 30 15:37:08 <_trine> jow_laptop, May 30 15:38:04 yeah, might be May 30 15:38:18 <_trine> can ya fix it May 30 15:38:23 <_trine> ;) May 30 15:39:28 <_trine> its only just started happening recently May 30 16:01:02 jow_laptop: http://www.gimpelevich.san-francisco.ca.us/danielg4/fakeidentd_pkg.tgz ready for commit May 30 16:09:43 DonkeyHotei: post it as a patch to the mailinglist, or trac. see https://dev.openwrt.org/wiki/SubmittingPatches May 30 16:10:25 loswillios: ok, i'll open a new ticket; should i just attach the file? May 30 16:10:39 no, a patch please May 30 16:19:42 ok May 30 16:20:06 loswillios: btw, do you know anything about the current status of ar7? May 30 16:22:37 i tried flashing a gt704wg with current trunk and got no ethernet May 30 16:23:06 near as i can tell, the marvell switch driver is not working May 30 16:23:50 seems all support for fixed phy was removed at some point May 30 16:25:04 (that would be a problem too, i think, for devices with no switch, like the speedport w501v) May 30 16:30:26 @DonkeyHotei: recently tried the ar7 port on a device with no switch and static phys on both ethernet ports, and it worked May 30 16:36:01 dangole: ok, must be the mvswitch driver then May 30 16:55:58 <[florian]> DonkeyHotei: yes it's not a perfect state right now wrt fixed/switch phys May 30 16:58:11 Warning! Found recursive dependency: PACKAGE_kmod-ipt-core PACKAGE_ipset PACKAGE_kmod-ipt-ipset PACKAGE_kmod-nfnetlink PACKAGE_kmod-ipt-core May 30 16:58:29 if thats of consequence May 30 17:41:30 blogic * r32006 /trunk/target/linux/ (3 files in 3 dirs): [kernel] adds patch to make Tantos (psb6970) switch buildable again May 30 17:46:41 _trine: I need to find access to a test device first, that can take a while May 30 17:46:48 [florian]: mvswitch doesn't seem to work with it at all; are there any clues to where the problem is? May 30 17:53:08 <[florian]> DonkeyHotei: so far, not really May 30 17:53:31 <[florian]> DonkeyHotei: there has been quite some rework of the mvswitch hack May 30 17:53:38 <[florian]> and I think we dropped some cpmac patches at the same time May 30 17:53:45 i noticed May 30 17:54:10 but dangole reported trunk cpmac works with no switch May 30 17:54:25 <[florian]> it certainly does, but mvswitch is not no switch :) May 30 17:54:37 <[florian]> especially since they either they need the trailer or header fix May 30 17:54:57 <[florian]> to remove 2 bytes from the head or the tail of the frame May 30 17:55:01 any knowledge of whether it works with the adm9669 series? May 30 17:55:22 <[florian]> you mean adm6996 right? May 30 17:55:30 <[florian]> depends on the wiring of the switch May 30 17:55:49 <[florian]> that's actually where the problem comes from, some adm6996 switch have their MDC/MDIO connected to cpmac May 30 17:55:51 <[florian]> some don't May 30 17:58:49 i might be able to test with adm6996 later May 30 17:59:37 i can never remember whether it's 6996 or 9669, and figured i'd say 9669 because there is a cpu target 8668 May 30 18:00:17 but for the moment, i want to get mvswitch to work May 30 18:00:44 and i would be able to test with no switch as well May 30 18:01:49 [florian]: i don't see any other target that uses mvswitch except atheros. which devices? May 30 18:04:17 blogic * r32007 /trunk/target/imagebuilder/Makefile: [imagebuilder] add $(SUBTARGET) to the filename if available May 30 18:05:44 │ Symbol: PACKAGE_freifunk-gwcheck [=n] May 30 18:05:46 why is this compiling? May 30 18:06:04 make[3] -C feeds/luci/freifunk-gwcheck compile May 30 18:06:33 every single freifunk in .config "is not set" May 30 18:08:53 same with 6in4/6to4 ; both 'is not set' in the .config; both built May 30 18:09:14 am I missing something, or is this broken? May 30 18:09:25 they're built because luci references them as possible dependency May 30 18:09:35 luci is a monolythic makefile providing several dozen packages May 30 18:09:51 the buildroot aggregates all potential depends of a source makefile as build depends May 30 18:10:20 see tmp/.packagedeps May 30 18:11:07 I could make them all conditionl but that will introduce circular deps May 30 18:11:09 fun; these won't be added into the squashfs, will they? May 30 18:11:24 I could rework luci to be modular but that will take days May 30 18:11:27 no May 30 18:11:30 ok, good. May 30 18:11:41 i don't mind if my build time is slightly longer as long as they aren't getting added when they shouldnt be May 30 18:11:42 build depend != get included May 30 18:12:29 i had presumed that luci was already modular, being able to select many different packages and not others in the menuconfig May 30 18:12:45 such was my misunderstanding May 30 18:19:27 * mystica555 now really wishes he knew why net-tools, seemingly only on my system, calls the wrong ranlib (or none at all) May 30 18:19:54 its been doing that for almost 2 years now..through 2 installs of funtoo... May 30 18:29:21 make V=s output on the net-tools build; doesn't appear that ranlib is being called at all May 30 18:29:23 http://ompldr.org/vZTBwZg May 30 18:40:01 * mystica555 has to chdir, run the proper ranlib on the .a file, and it continues compiling fine: ~/openwrt/trunk/build_dir/target-mips_r2_uClibc-0.9.33.2/net-tools-1.60/lib $ ../../../toolchain-mips_r2_gcc-4.7-linaro_uClibc-0.9.33.2/binutils/binutils/ranlib libnet-tools.a May 30 18:51:41 mystica555: net-tools-1.60/lib/Makefile executes $(AR) to build the library May 30 18:51:49 mystica555: the commandline is masked with @ May 30 18:52:03 try to remove the @ through a patch or similar, rerun the build and paste the log May 30 18:54:36 I suspect the *.o files are not even built at the time the archive is created May 30 18:54:45 therfore the .a ends up being empt May 30 18:54:46 y May 30 18:54:58 which later leads to the index error May 30 18:55:20 its not really about ranlib, "ar rcs" would usually do ranlib's job as well May 30 18:59:10 Omploaded 'ranlib2.txt' to http://ompldr.org/vZTBwdA May 30 19:02:24 ok, it calls the wrong ar May 30 19:02:33 openwrt passes AR=... May 30 19:02:40 the Makefile uses $(AR) May 30 19:02:48 yet it ends up being just "ar" May 30 19:02:54 lovely May 30 19:03:44 * mystica555 wonders why the output of that command is masked in the first place.. :\ i love debug output.. May 30 19:04:14 <[florian]> DonkeyHotei: apart from atheros, I also do not know about any device using mvswitch May 30 19:04:31 which atheros devices, i mean May 30 19:05:40 mystica555: found the reason May 30 19:05:47 <[florian]> oh, well no idea, nbd might tell May 30 19:06:03 jow_laptop: oh? what was it? May 30 19:06:46 eventually i want to get rid of the mvswitch driver entirely May 30 19:07:03 <[florian]> and use the dsa driver instead? May 30 19:07:04 nbd: and do what? May 30 19:07:06 yes May 30 19:07:08 oh, right May 30 19:07:08 use dsa May 30 19:07:16 but that won't work for the atheros target as it is now May 30 19:07:26 mystica555: the net-tools OpenWrt Makefile overrides the default Build/Compile, thus the AR="..." flag gets lost May 30 19:07:28 dsa needs either mips-machine or device tree May 30 19:07:31 mystica555: try this http://pastebin.com/eNdCwXSh May 30 19:07:34 i now remember you mentioned the dsa driver three years ago May 30 19:07:42 <[florian]> nbd: can't you register it using a phy-fixup? May 30 19:07:54 [florian]: i'd like to avoid such hacks May 30 19:08:18 nbd: well, mvswitch as it is now doesn't seem to work for at7 May 30 19:08:21 *ar7 May 30 19:08:32 what's the problem? May 30 19:08:37 not sure May 30 19:08:45 i was hoping for ideas May 30 19:08:47 <[florian]> nbd: do we still need the hacks to manipulate the skbs? May 30 19:08:54 in earlier kernel versions it didn't work, because mvswitch needed hacks in the ethernet driver May 30 19:09:00 however i made those unnecessary with some rework that i did May 30 19:09:23 well, the rework doesn't appear to work for ar7 May 30 19:09:33 <[florian]> DonkeyHotei: I think the mvswitch driver just does is not bound by the mdio driver May 30 19:09:40 <[florian]> DonkeyHotei: that's why you don't have it working May 30 19:09:44 DonkeyHotei: did you set up vlan 1/2? May 30 19:09:50 well, dmesg reports it bound May 30 19:10:02 nbd: how? May 30 19:10:11 lan should be on eth0.1, wan on eth0.2 May 30 19:10:18 that doesn't work May 30 19:10:30 no rx or tx traffic on eth0? May 30 19:10:34 did the driver attach at all? May 30 19:10:43 setting up vlans errors out May 30 19:10:58 huh? May 30 19:11:11 that doesn't make any sense May 30 19:11:13 <[florian]> sounds like the driver is not attached May 30 19:11:15 the driver simply does not respond May 30 19:11:33 drop the vague descriptions, give us some raw data instead May 30 19:11:42 log messages, error messages, etc. May 30 19:11:57 the thing that is normally known as a "bugreport" May 30 19:11:59 :-) May 30 19:13:23 jow_laptop: I seem to be unable to find the right place to put that patch so it applies; i'm probably doing it wrong.. May 30 19:14:19 mystica555: yes, I based it against the feeds dir, try to apply it with -p1 within feeds/packages/ May 30 19:14:22 putting the patch in feeds/packages/ then patch -p1 < patch it rejects it.. May 30 19:14:30 then do it manually May 30 19:14:40 easy enough to do that May 30 19:14:43 maybe pastebin borked the patch May 30 19:14:44 or I did May 30 19:14:47 *shrug* May 30 19:14:52 <[florian]> DonkeyHotei: did we loose you? May 30 19:14:53 it was copy-paste out of the temrinal (whitespace-fu) May 30 19:14:57 yea May 30 19:15:05 that will do it 9 times out of 10 ;) May 30 19:15:33 [florian]: sorry, i fail at teh multitask May 30 19:15:57 and the error message is itself quite vague May 30 19:16:21 <[florian]> like "Failed to connect to switch"? May 30 19:16:48 DonkeyHotei: vague error messages are better than vague interpretations of vague error messages May 30 19:17:26 one less layer of abstractions and information omission May 30 19:17:58 [florian]: yep May 30 19:18:08 <[florian]> then the driver is not bound as you think it is May 30 19:18:09 that's exactly it May 30 19:18:17 dmesg reports it is May 30 19:19:47 blogic * r32008 /packages/utils/mdadm/Makefile: May 30 19:19:47 [packages] bump mdadm to 3.2.3 May 30 19:19:47 Ticket #11231 mdadm segmentation fault on trunk May 30 19:20:22 mips-openwrt-linux-uclibc-ar rcs libnet-tools.a activate.o af.o arcnet.o ash.o ax25.o ax25_gr.o ddp.o ddp_gr.o ec_hw.o econet.o ether.o fddi.o frame.o getargs.o getroute.o hdlclapb.o hippi.o hw.o inet.o inet6.o inet6_gr.o inet6_sr.o inet_gr.o inet_sr.o interface.o ipx.o ipx_gr.o ipx_sr.o irda.o loopback.o masq_info.o netrom.o netrom_gr.o netrom_sr.o nstrcmp.o ppp.o ppp_ac.o proc.o rose.o rose_g May 30 19:20:22 r.o setroute.o sit.o slip.o slip_ac.o sockets.o strip.o tr.o tunnel.o unix.o util-ank.o util.o x25.o x25_gr.o x25_sr.o May 30 19:20:29 ding! May 30 19:20:40 and no failures May 30 19:20:54 great May 30 19:20:57 thanks May 30 19:22:00 jow * r32009 /packages/net/net-tools/Makefile: [packages] net-tools: ensure that the cross-ar gets used, fixes linker issues on certain systems May 30 19:22:42 now I wonder how other distro ar's are able to indulge the foreign object files May 30 19:22:49 * nbd waits for DonkeyHotei to provide some usable data May 30 19:22:58 no idea May 30 19:23:00 blogic * r32010 /packages/libs/libnfc/Makefile: [packages] fix libnfc build breakage May 30 19:23:18 ive been doing this with funtoo for the entire time, and ive come across a few seemingly random issues with ar creating the archives May 30 19:23:31 nbd: in the middle of something atm; gimme 10 min or so May 30 19:23:36 ok May 30 19:24:03 i have to leave soon May 30 19:24:19 mystica555: well guess what matters is that it is finnaly fixed, and the next time something like this pops up there's a clue what to look for May 30 19:24:50 indeed May 30 19:25:50 https://dev.openwrt.org/ticket/7620 was initially made for that basic issue.. but, in that time, ive not tried rebuilding pgsql or cyrus-sasl (that early-on compile i did was way too bloated..heh) and samba36 has built fine for me recently May 30 19:41:55 root@OpenWrt:/# swconfig dev eth0 vlan 0 show May 30 19:41:55 Failed to connect to the switch May 30 19:42:06 [florian], nbd: ^ May 30 19:42:31 the driver has no swconfig support May 30 19:42:47 that would explain it May 30 19:43:27 so, let's check the kernel messages if the driver attached properly May 30 19:43:39 look for things related to mvswitch init and the ethernet driver May 30 19:44:58 [ 1.410000] eth%d: Marvell 88E6060 PHY driver attached. May 30 19:44:58 [ 1.470000] cpmac: device eth0 (regs: 08612800, irq: 41, phy: cpmac-1:10, mac: 00:18:01:a9:b9:3c) May 30 19:44:58 [ 1.570000] cpmac: device eth1 (regs: 08610000, irq: 27, phy: cpmac-1:1f, mac: 00:20:e0:02:07:e1) May 30 19:46:43 ok May 30 19:46:47 <[florian]> that looks right so far May 30 19:49:50 i wonder if the driver attached to the right interface May 30 19:50:01 no way to know May 30 19:50:03 since there's eth0 and eth1 May 30 19:50:14 <[florian]> try swconfig dev eth1 and see which one it picked up May 30 19:50:23 no swconfig support, remember May 30 19:50:26 <[florian]> oh right May 30 19:50:33 but i would assume it would simply not bind to the wrong one May 30 19:52:39 how do i talk to it without swconfig? May 30 19:53:03 jow * r32011 /trunk/package/libubox/Makefile: [package] libubox: update to 2012-05-30, fixes bad non-empty return values in case of already cached variables May 30 19:54:07 _trine: the above should fix your dnsmasq issue May 30 19:54:11 I need to leave, bbl May 30 19:54:19 <_trine> thankyou May 30 19:57:05 anyone want to fix xburst's (SENSORS_JZ4740) & au1000's (ARC_CONSOLE) missing 3.3 symbols? May 30 20:16:30 blogic * r32012 /trunk/package/util-linux/Makefile: May 30 20:16:30 [util-linux] fixes a descriptions mismatch May 30 20:16:30 fixes #11381 **** ENDING LOGGING AT Thu May 31 02:59:58 2012