**** BEGIN LOGGING AT Tue Apr 06 02:59:56 2010 Apr 06 03:00:08 but it needs pretty new compilers Apr 06 03:00:38 but it is pretty crampled stunnel + a couple of other goodies + LuCI is the max without USB Apr 06 03:51:25 thepeople: http://home.comcast.net/~sdwalker/uscan.tar.gz has the shell scripts, the shtml and the mostly clean diff Apr 06 04:03:57 did the trac recently get an upgrade? /chrome/common/js/jquery.js went missing Apr 06 05:35:54 build #59 of ppc40x is complete: Failure [failed compile_6] Build details are at http://tksite.gotdns.org:8010/builders/ppc40x/builds/59 Apr 06 08:59:55 nico * r20729 /packages/net/sipp/Makefile: [packages] sipp: prevent spurious rebuilds (caused by PKG_BUILD_DIR containing .svn), link against libgcc_s when possible, remove trailing whitespaces Apr 06 09:14:53 nico * r20730 /packages/net/git/Makefile: [packages] git: prevent spurious rebuilds caused by $(MAKE) ... install being called in Package/git/install, use PKG_INSTALL instead Apr 06 09:20:45 build #56 of atheros is complete: Failure [failed compile_6] Build details are at http://tksite.gotdns.org:8010/builders/atheros/builds/56 Apr 06 09:26:44 nico * r20731 /packages/libs/file/Makefile: [packages] file: prevent spurious rebuilds, caused by touching files during Package/file/install, use Hooks/... instead, cleanup Apr 06 09:36:44 nico * r20732 /packages/net/aprx/Makefile: [packages] aprx: prevent spurious rebuilds caused by PKG_BUILD_DIR containing .svn, use svn upstream repository instead Apr 06 09:42:10 build #41 of ifxmips is complete: Failure [failed compile_6] Build details are at http://tksite.gotdns.org:8010/builders/ifxmips/builds/41 Apr 06 09:48:24 gmorning Apr 06 11:01:55 build #32 of x86 is complete: Failure [failed compile_6] Build details are at http://tksite.gotdns.org:8010/builders/x86/builds/32 Apr 06 12:52:04 [florian]: pin Apr 06 12:52:06 g Apr 06 12:52:33 <[florian]> rtz2: pong Apr 06 12:56:11 [florian]: I have a couple of remarks/questions Apr 06 12:56:20 [florian]: 1. sysupgrade Apr 06 12:57:04 [florian]: I didn't test it, but as far as I can see it, it's only necessary to modify the platfrom.sh script Apr 06 12:57:20 <[florian]> correct Apr 06 12:58:14 [florian]: problem is, every script has a different image header, so maybe it would be best to simply accept every file >1mb and smaller then the flash size Apr 06 12:58:31 [florian]: or do you have another idea? Apr 06 12:58:39 <[florian]> cannot we check for some magic in the first couple of bytes? Apr 06 12:58:44 <[florian]> that's how the other devices do it Apr 06 12:58:56 build #39 of iop32x is complete: Failure [failed compile_6] Build details are at http://tksite.gotdns.org:8010/builders/iop32x/builds/39 Apr 06 12:59:43 [florian]: well, problem is, every device has a different magic, so it could still be the wrong image Apr 06 13:00:18 [florian]: hmm, but maybe we could compare the first few bytes of the image with the first few bytes in flash Apr 06 13:00:22 <[florian]> rtz2: but we can know which device we are on Apr 06 13:00:26 <[florian]> rtz2: yes, that's the idea Apr 06 13:00:51 [florian]: ok, will do Apr 06 13:01:25 [florian]: 2. there was some guy, who managed to get pretty good NAS200 support working Apr 06 13:01:44 build #35 of xburst is complete: Failure [failed compile_6] Build details are at http://tksite.gotdns.org:8010/builders/xburst/builds/35 Apr 06 13:01:53 [florian]: he's currently not online, but he goes by dFence or something in #openwrt Apr 06 13:02:03 <[florian]> rtz2: oh ok I see Apr 06 13:02:12 <[florian]> rtz2: does he have some patches that go well with the current board model? Apr 06 13:02:25 [florian]: yes Apr 06 13:02:38 <[florian]> rtz2: great Apr 06 13:02:41 [florian]: I talked him through the stuff Apr 06 13:03:03 [florian]: he also has a working reset button and leds Apr 06 13:03:56 [florian]: but there were some odd problems with the buttons, they aren't directly connected but use some kind of circuit, I don't really understand it Apr 06 13:04:25 [florian]: he's getting spurious irq messages, but none of the external irq lines seemed to be used Apr 06 13:04:49 [florian]: well, we will see Apr 06 13:04:54 <[florian]> rtz2: what about some gpios triggering interrupts? Apr 06 13:05:05 <[florian]> rtz2: maybe the button and leds are behind some kind of expander device Apr 06 13:06:01 [florian]: it works like this: if he presses the reset button, a gpio (don't remember the number) changes Apr 06 13:06:36 [florian]: and then, after holding it a few seconds, he gets the irq Apr 06 13:07:09 <[florian]> looks like some kind of buffer which generates an interrupt Apr 06 13:07:17 <[florian]> rtz2: or a device with an integrated timer Apr 06 13:07:58 [florian]: my first idea was, that it's connected via a delay circuit to one of the external irqs, but if you enable the gpio mode for the irq pin, the irq shouldn't be triggered Apr 06 13:08:07 [florian]: but it's still there Apr 06 13:09:31 [florian]: I told him, to change to route the external irq lines to different irq numbers, to find out which one it is, but not sure if he did it Apr 06 13:09:42 [florian]: I will have to get back to him tomorrow Apr 06 13:09:55 [florian]: this evening Apr 06 13:10:05 <[florian]> allright Apr 06 16:07:48 build #57 of ppc44x is complete: Failure [failed compile_6] Build details are at http://tksite.gotdns.org:8010/builders/ppc44x/builds/57 Apr 06 22:22:47 ping cshore Apr 06 22:23:23 pong xMff Apr 06 22:23:33 where is preinit.arch sourced? Apr 06 22:24:29 xMff: it no longer is....anything that used preinit.arch should now have target/linux/.../base-files/lib/preinit/XX_script-name for the things preinit.arch was used for Apr 06 22:24:40 have an issue on the dir-300 here where the failsafe ip is not cleared from eth0 Apr 06 22:24:58 what arch? Apr 06 22:25:03 atheros Apr 06 22:25:13 with switch Apr 06 22:25:45 target/linux/atheros/base-files/lib/preinit/XX-something-about-interfaces IIRC Apr 06 22:25:59 got it Apr 06 22:26:24 xMff: but the deconfig should be happening after failsafe Apr 06 22:26:31 xMff: or rather before init Apr 06 22:26:44 ifname=eth0 is set in 15* Apr 06 22:26:57 but the pi_ifname=$ifname happens in 10* Apr 06 22:27:14 ifname takes precedence Apr 06 22:27:28 oh....I see Apr 06 22:27:41 yeah, that's wrong Apr 06 22:28:32 if it's using the switch though, you might need to do a special deconfig in 9X_before-init Apr 06 22:28:44 or maybe not... hmm Apr 06 22:28:48 could be a race Apr 06 22:30:08 hotplug doesn't do configuration during preinit Apr 06 22:30:23 preinit should be completely serial Apr 06 22:30:41 yeah, but it somehow looks like it uses eth0 when setup the failsafe ip and eth0.1 when trying to tear it down Apr 06 22:31:04 build #38 of rb532 is complete: Failure [failed compile_6] Build details are at http://tksite.gotdns.org:8010/builders/rb532/builds/38 Apr 06 22:31:24 xMff: is there another spot with a swith configuration....I remember atheros was strange that way Apr 06 22:31:55 no wait Apr 06 22:32:06 it unconditionally redefines preinit_ip_deconfig() Apr 06 22:32:22 grep -q 'Atheros AR231[567]' /proc/cpuinfo && { Apr 06 22:32:22 Apr 06 22:32:26 but the code in this procedure is only called if there is a switch Apr 06 22:32:29 is it one of those Apr 06 22:33:10 you have two exceptions for switches but this devices use none of them Apr 06 22:34:07 Hmmm.....pi_ifname is not set by ifname for non-switches Apr 06 22:34:18 or rather non Ifineon/Marvell Apr 06 22:34:50 its a logic flaw in preinit_ip() from 15_preinit_iface_atheros Apr 06 22:35:16 In fact it shouldn't be set at all, unless the user has defined it in preinit options Apr 06 22:35:17 preinit_ip() falls back to eth0 when neither of the both switches is detected Apr 06 22:35:42 but the deconfig precedure which is redefined in the same file only does something if there is a switch Apr 06 22:35:46 oh, I see Apr 06 22:36:08 but only for Atheros ART231[567] Apr 06 22:36:13 right Apr 06 22:36:29 unfortunately this device is AR2317 :) Apr 06 22:36:33 and it has a switch Apr 06 22:36:41 but neither adm nor marvell Apr 06 22:36:48 but ic+ instead Apr 06 22:37:00 right....the problem is that this device was not previously supported with a switch config Apr 06 22:37:07 in preinit Apr 06 22:37:08 it was in 8.09.2 Apr 06 22:37:12 ah yes Apr 06 22:37:17 however Apr 06 22:37:21 it is a generic issue Apr 06 22:37:42 it will break on any hypothetical device which has a ar231x cpu and none of the mentioned switches Apr 06 22:39:01 right. i don't recall the exact code from before, but it the specific switch tests were grabbed from preinit.arch Apr 06 22:39:11 there are ar231x devices without switch Apr 06 22:39:17 i have one at home Apr 06 22:39:38 nbd: that's why the test was added to preinit.arch, right? Apr 06 22:39:56 dunno why it was added Apr 06 22:40:18 correction: it will break on any device which has an ar231x cpu _and_ a switch _but_ neither adm nor marvell :) Apr 06 22:40:55 on single eth or non-switch devices it does not harm since the network config later will redefine the ip on eth0 anyway Apr 06 22:41:26 but on the dir-300 it is an issue because it has eth0.1 and eth0.2 at this point and eth0 is not touched anymore Apr 06 22:41:50 preinit has a ifconfig 0.0.0.0 down, but I think it's for ifname Apr 06 22:42:04 which doesn't deconfig the switch Apr 06 22:42:10 is not called since the deconfig procedure is redefined Apr 06 22:42:34 http://paste.tksite.gotdns.org/d793ee7c9 Apr 06 22:42:45 hmm, anybody on avr32 here? Apr 06 22:43:00 thepeople Apr 06 22:43:14 thepeople: ping Apr 06 22:43:31 xMff: ah, I see, the redefined ip_deconfig needs to deconfig for non-switch devices too (Marvell/Infineon switch) that is Apr 06 22:43:41 cshore: correct Apr 06 22:44:38 and somehow it works anyway if one sysupgrades from 8.09.2 Apr 06 22:44:44 no idea why Apr 06 22:44:45 <_trine> are you back in Germany now Felix? Apr 06 22:46:18 jow * r20733 /trunk/target/linux/atheros/base-files/lib/preinit/15_preinit_iface_atheros: [atheros] fix failsafe ip deconfig on AR23xx devices without ADM or Marvell switch Apr 06 22:47:10 jow * r20734 /branches/backfire/target/linux/atheros/base-files/lib/preinit/15_preinit_iface_atheros: [backfire] merge r20733 Apr 06 22:51:01 cshore: I think it worked after sysupgrading from 8.09, becase the eth0.1 had a different ip in my case, so it didn't clash with the 192.168.1.1 set on eth0 Apr 06 22:51:26 xMff: that makes sense Apr 06 22:52:48 for more robustness the switch setup should probably always deconfig the parent iface Apr 06 22:53:00 xMff: yes Apr 06 22:55:23 _trine: no, still in california Apr 06 22:55:29 i'll head back tomorrow Apr 06 22:57:18 jow * r20735 /trunk/package/block-extroot/Makefile: [package] block-extroot: unblock package if kmod-broadcom-sdhc has been selected Apr 06 22:57:56 jow * r20736 /branches/backfire/package/block-extroot/Makefile: [backfire] merge r20735 Apr 06 22:59:30 nbd: https://dev.openwrt.org/ticket/7102 Apr 06 22:59:47 we should probably reboot after applying the nvram fixes Apr 06 23:00:00 or remove /e/c/wireless in the mac fixup code Apr 06 23:00:52 since the nvram stuff is now userspace-only, the changes are not propagated into the in-kernel copy Apr 06 23:00:59 ... of nvram Apr 06 23:02:25 thus the wifi mac in sysfs is wrong during the initial wifi detection because the kernel did not re-read the nvram yet Apr 06 23:03:00 i don't want to add an unconditional reboot Apr 06 23:03:08 because if it goes wrong, it means having a reboot loop Apr 06 23:03:12 yes Apr 06 23:03:22 i'd much rather have a second fixup that fixes up the wireless mac from nvram Apr 06 23:03:51 even though that's still going to make it fail on the first boot Apr 06 23:08:50 I have an atheros ap96 based board, and I'm trying to make a mach-myboard.c for it... I've got everything up but the switch. (This stuff is very new to me) I was wondering where I can find all these PHY_MASK values... I have the u-boot source and LSDK for the board Apr 06 23:10:32 actually I have the WAN (eth1) working, but eth0 is dead. I found eth1 defined in athrs16_phy.h as 0x4 Apr 06 23:12:43 whoops, athrs16_phy.c Apr 06 23:13:21 aport: It should be enough to set the phy_mask of eth0 to 0x1 in your board setup code Apr 06 23:13:38 KanjiMonster, I'll try it out Apr 06 23:17:50 it's using an 8316, so I wanted to say thanks for the work Kanji :) Apr 06 23:19:01 you're welcome :) Apr 06 23:19:12 nbd: what about this? http://paste.tksite.gotdns.org/d3cf01074 Apr 06 23:19:26 KanjiMonster, for ar71xx_add_device_mdio, should 0x00 suffice? or what exactly is the value it's looking for? Apr 06 23:20:19 aport: tbh, I don't know - I never looked at the code behind this Apr 06 23:20:28 okie dokie, thanks though :D Apr 06 23:20:31 but 0x0 should be fine Apr 06 23:21:06 when we start shipping these boards, I'd like to submit a patch against trunk to add it. I'd like for customers to be able to checkout trunk and compile for this board Apr 06 23:21:43 there's a slight chance that the switch might not work - the register 0x8, which seems to control the hardware configuration, might need a different value than the one for rs pro/rb 450g Apr 06 23:22:19 xMff: no way. mac80211.sh will not get tainted by nvram crap ;) Apr 06 23:22:37 nbd: yeah, don't like it either Apr 06 23:22:53 if you run a fixup script after the network script has run, it's much better Apr 06 23:22:56 KanjiMonster, thanks for the heads up, I see in the LSDK that for ap96 based boards the value is Apr 06 23:23:01 athrs16_reg_write(0x8, 0x012e1bea); Apr 06 23:23:27 aport: you should probably talk to juhosg about this stuff Apr 06 23:24:00 nbd, I'll send him an email. I didn't want to be a nuisance but I noticed that he's the maintainer of ar71xx port Apr 06 23:24:30 if I wanted to edit /etc/init.d/rcS ... if I do that, what targets do I need to remake? also, how can a script send output to the console in preinit? Apr 06 23:25:14 philipp64: if it's after the console has been setup you can just echo Apr 06 23:25:18 aport: yeah, I've seen that line, too Apr 06 23:25:35 I'm not sure it is after the console has been set up. Apr 06 23:25:43 what about doing an echo to >/dev/ttyS0? Apr 06 23:25:58 not all boards have a /dev/ttyS0 Apr 06 23:26:08 no, but the one I'm trying to bring up does. Apr 06 23:26:21 KanjiMonster, in ar8216.c I changed value from 0x81461bea to 0x012e1bea, I'll let you know :D Apr 06 23:26:29 remind me again how I know if the console is present? Apr 06 23:26:51 xMff: pong Apr 06 23:26:54 wigyori: pong Apr 06 23:27:39 cshore, were you able to test that uci_firewall.sh patch with miniupnpd's chains? Apr 06 23:28:08 philipp64: it's not for the preinit_essential hook, but should be for all other hooks Apr 06 23:28:14 aport: no Apr 06 23:28:24 it worked beautifully for strongswan's custom rules... closed a long standing bug for me Apr 06 23:28:32 aport: I'll try it soon Apr 06 23:28:50 cshore, okay, I'd be interested in if it worked or if there was a better way to do it Apr 06 23:28:50 cshore: ok, and if I tweak package/base-files/files/etc/init.d/rcS ... then I have to rebuild package/base-files/clean, package/base-files/install, ... Apr 06 23:29:22 make package/base-files/{clean,compile.install} Apr 06 23:29:41 and target/linux/install Apr 06 23:30:03 thanks. much quicker than building from scratch. Apr 06 23:33:33 philipp64: the {clean,compile,install} is meant to type literally exactly like that Apr 06 23:34:03 except , not . Apr 06 23:35:42 thepeople: pong Apr 06 23:36:27 wigyori: I was ponging you :-) Apr 06 23:36:28 thepeople: umm, just a question about how avr32 goes on nowadays, should i still try the jffs2 approach as in the docs? Apr 06 23:36:42 I use squashfs Apr 06 23:36:56 cshore: moment of truth... Apr 06 23:37:04 drumroll please... Apr 06 23:37:19 thepeople: that needs repartitioning? Apr 06 23:37:48 it has to flashed from uboot Apr 06 23:37:56 cshore: Running /etc/rc.d/S99sysctl Apr 06 23:38:16 I do a 1 meg kernel partition and the rest as rootfs Apr 06 23:38:16 philipp64: hey that's a good sign Apr 06 23:38:25 then I can reflash from within openwrt Apr 06 23:38:30 but... where's my console login? Apr 06 23:38:45 tried pressing enter? Apr 06 23:39:17 thepeople: hmm, so you use 1meg as kernel, and the rest as squashfs Apr 06 23:39:22 philipp64: it should just got straight to a root shell when you press enter Apr 06 23:39:35 unless inittab is wrong Apr 06 23:39:49 wigyori: yea Apr 06 23:39:52 xMff: gah!!!! all this time I thought it was wedged. Apr 06 23:40:00 lol Apr 06 23:40:03 sry Apr 06 23:40:22 thepeople: ok, and is it the stock uboot from atmel, or is it a patched one? Apr 06 23:40:34 I still have the stock one on it Apr 06 23:40:53 thought about upgrading it, but meh Apr 06 23:40:57 it works Apr 06 23:41:16 need to turn on iproute2 ... Apr 06 23:42:13 hm Apr 06 23:42:53 aha, i screwed up something Apr 06 23:47:15 xMff: so where is the configuration state all kept (network address, mask, etc)? Apr 06 23:48:04 philipp64: /etc/config Apr 06 23:48:23 in the case of network stuff /etc/config/network Apr 06 23:48:43 yup, ok... Apr 06 23:52:06 what's required to turn -x on in bash? Apr 06 23:52:24 or busybox shell or whatever you all use. Apr 07 00:01:00 xMff: so what do I need to get commit rights to trunk? Apr 07 00:11:04 cshore: why are the start up scripts run in the background with & ? Apr 07 00:13:17 philipp64: you mean rcS? Apr 07 00:13:36 yes. Apr 07 00:14:24 philipp64: I think it's so you can get a terminal sooner, but I don't know for sure....I wasn't around for that design decision Apr 07 00:14:54 other question... I changed the DEFAULT_PACKAGES, did an "rm .config ; make menuconfig" but I'm not seeing the additional kmod's showing up. Apr 07 00:14:57 what did I not do? Apr 07 00:15:20 well, we're talking 5-15 seconds tops... Apr 07 00:16:06 you probably need to remove the $ROOTDIR/tmp Apr 07 00:16:20 wheere ROOTDIR is OpenWRT source root Apr 07 00:17:45 yup, that did it. Apr 07 00:18:44 cshore: could you do me a quick favour and check how big your liblua.so in /usr/lib/ is? Apr 07 00:19:31 215976 bytes Apr 07 00:20:20 ok, thanks Apr 07 00:21:06 and now the big question: why the heck can Visual Studio compile the lua dll at less then 80kb? Apr 07 00:22:54 is it a standalone dll? Apr 07 00:24:45 yes Apr 07 00:25:40 I did enable link time code generation, but I doubt it has that much influence Apr 07 00:28:31 cshore: well, the .config is correct, but the make target/linux/clean target/linux/compile target/linux/install still isn't rebuilding all of the kmods... Apr 07 00:28:35 hmmm.... Apr 07 00:32:34 philipp64: for kmods recompile you have to do make target/linux/clean Apr 07 00:33:06 did that... but it's still missing stuff... Apr 07 00:33:49 I'm afraid I have no idea...try make clean and rebuilding the whole thing Apr 07 00:34:06 (except toolchain) Apr 07 00:35:58 ok, how do you remove everything but the toolchain? "make dirclean" looks pretty indiscriminate... Apr 07 00:40:47 make clean Apr 07 00:41:46 yeah, ok, doing that now... Apr 07 01:02:43 xMff: say, how come some of the KCONFIG:= strings contain =y and some don't have any value at all? Apr 07 01:21:57 xMff: seems I have a weird problem with luci-app-statistics / ping module when I put more than one host: http://img406.imageshack.us/img406/1982/pinga.png Apr 07 01:22:43 first I added just www.dommel.com, worked fine. then added the stewie machine, after which it also showed the correct graph, but after a reboot the graph seems wrong Apr 07 02:19:19 xMff: who understands the kernel setup stuff? like what $(KCONFIG) does? is that a dependency, or what gets forced on? **** ENDING LOGGING AT Wed Apr 07 02:59:57 2010