**** BEGIN LOGGING AT Wed Jun 08 02:59:56 2011 Jun 08 04:50:58 swalker * r27134 /packages/utils/ap51-flash/Makefile: [packages] ap51-flash: open-mesh -> cloudtrax Jun 08 08:00:12 hi, I have a problem from yesterday, I cant compile the compiler anymore on my debian testing: It always fails on error: Jun 08 08:00:14 /usr/bin/ld: libbackend.a(graphite.o): undefined reference to symbol 'ppl_finalize' Jun 08 08:00:14 /usr/bin/ld: note: 'ppl_finalize' is defined in DSO /usr/lib64/libppl_c.so.4 so try adding it to the linker command line Jun 08 08:00:14 /usr/lib64/libppl_c.so.4: could not read symbols: Invalid operation Jun 08 08:00:35 (sometimes its on cloog also). Jun 08 08:01:28 I've tried with gcc 4.4.1-codesourcery and 4.5-linaro (target=bcm63xx) Jun 08 08:01:42 both show the same error, after a make dirclean Jun 08 08:02:34 I've also tried to disable or enable "Graphite" in the config, and/or using the system version on pll and cloog Jun 08 08:03:30 whatever I do, it's always pll and/or cloog which failed, with different errors given the compiler and the config options. Jun 08 08:23:31 build #38 of ixp4xx is complete: Failure [failed shell_4] Build details are at http://buildbot.openwrt.org:8010/builders/ixp4xx/builds/38 Jun 08 08:33:35 build #37 of kirkwood is complete: Failure [failed shell_4] Build details are at http://buildbot.openwrt.org:8010/builders/kirkwood/builds/37 Jun 08 08:57:49 build #45 of ps3 is complete: Failure [failed compile_3] Build details are at http://buildbot.openwrt.org:8010/builders/ps3/builds/45 Jun 08 09:31:59 build #45 of pxcab is complete: Failure [failed shell_4 compile_3] Build details are at http://buildbot.openwrt.org:8010/builders/pxcab/builds/45 Jun 08 09:43:54 any ideas why I get NO-CARRIER on the eth0 on RB493G when I munge things so it operates as an ar8316 switch? Jun 08 12:48:58 jow * r27135 /packages/net/tinyproxy/ (2 files in 2 dirs): [packages] tinyproxy: import fix for upstream bug #95 - fix url filters in transparent mode Jun 08 13:03:38 xMff: starting a build now... so it was an unclean buffer? Jun 08 13:03:48 netprince: yes Jun 08 13:04:21 netprince: I don't pretend to undertstand the details, but now it seems solid Jun 08 13:05:00 that is great, nice work Jun 08 13:05:25 eh, no wonder I had so much more free memory with 3.0; insmod bails out when it isn't a 2.x kernel Jun 08 13:05:35 haha Jun 08 13:05:49 its big prother the modutils-ng had the same issue Jun 08 13:06:13 but it looks like the insmod search patch from openwrt is at fault here Jun 08 13:06:48 vanilla busybox doesn't do the check as far as I can tell Jun 08 13:07:13 ah Jun 08 13:08:08 but the search will break anyway, since linux-3.0 claims to be 3.0.0 (for compatibility reasons), but the modules are in ../3.0, and insmod searches in /`uname -r` Jun 08 13:09:33 so I'll do the obvious thing and let linux report to be 3.0 and see what will break ;) Jun 08 13:10:36 bah this sounds so crappy Jun 08 13:11:10 I really wonder why this isn't a (temporary) kernel option Jun 08 13:11:28 doing three digits for compat reasons in the version but breaking the dir = uname assumption with the module install palce Jun 08 13:11:32 thats stupid Jun 08 13:11:39 they should do one clean cut and be done with it Jun 08 13:12:32 I think of patching the version number you should patch the module installation to use the version as dirname Jun 08 13:12:37 I haven't looked at what's responsible for copying the modules into /lib/modules; chances are this is done by openwrt itself Jun 08 13:12:44 hm maybe Jun 08 13:14:54 yes, it is: $(CP) -L $$(FILES) $$(1)/lib/modules/$(LINUX_VERSION)/ Jun 08 13:16:47 yeah but then its still the same, the tarball and release is 3.x and it reports itself as 3.x.y Jun 08 13:16:58 according to the -rc1 release mail, stable is supposed to call itself only 3.0 - I don't see any reason for waiting for stable to make the switch; gives us more time to fix everything Jun 08 13:19:23 (I would laugh if something in the kernel breaks after dropping the subversion ;) Jun 08 13:19:25 jow_laptop: Hello,how is official Chinese board going on? Jun 08 13:22:20 * ^Willie^ sends his last email to ubicom... give them 2 weeks answering howto revert u-boot back to make it use the serial uart again.. Jun 08 13:22:30 <^Willie^> else trash this crap.. its rediculous.. Jun 08 13:28:35 My mr3420 is running on 450mhz now,Bogomips rises up to 300 mips. Jun 08 13:28:42 http://pastebin.com/7BzPQFBn Jun 08 13:32:05 ping jow_laptop Jun 08 13:56:00 build #31 of octeon is complete: Failure [failed shell_4] Build details are at http://buildbot.openwrt.org:8010/builders/octeon/builds/31 Jun 08 14:10:04 hi Jun 08 14:10:21 how long is the arp value cached after the host goes down? Jun 08 14:10:33 the/a Jun 08 14:10:44 in /proc/net/arp Jun 08 14:49:45 u8 *eeprom = (u8 *) KSEG1ADDR(0x1fff1000); (for ar71xx platform) Jun 08 14:49:51 how is that address calculated? Jun 08 14:50:49 it's not calculated Jun 08 14:51:10 it's simply where the eeprom data on the flash is (within the memory window for flash access) Jun 08 14:51:37 build #42 of uml is complete: Failure [failed compile_4] Build details are at http://buildbot.openwrt.org:8010/builders/uml/builds/42 Jun 08 14:54:31 nbd: flash as in the contents of /dev/mtdX or something different entirely? Jun 08 14:55:29 How to compile an individual kenel module after select it in make menu? Jun 08 15:28:07 fullstop: typically the eeprom shows up as the 'art' or 'eeprom' partition in mtd Jun 08 15:28:19 in this case there's probably an offset of 0x1000 Jun 08 15:28:29 nbd: thanks. This is useful. Jun 08 15:38:03 Flashed a custom OpenWrt image to a WRT54G which is working grand, but if the Network interface on the Windows machine goes down, it stays down until I reboot. Am I missing some cable detect/switch package etc? Jun 08 15:38:37 Reboot the Windows machine btw. Jun 08 15:38:42 b43 ? Jun 08 15:39:03 jow_laptop: No. ethernet. Jun 08 15:39:13 ah Jun 08 15:39:21 never hear of such an issue with openwrt Jun 08 15:39:27 I would invstigate the windows box Jun 08 15:40:06 on the openwrt side you could use "robocfg" to see what the wrt's idea of the link state is Jun 08 15:40:21 if the windows box claims to be down/disconnected Jun 08 15:40:55 Link LED stays off, so I the ethernet driver might be turning the chip off or into powersave mode or something Jun 08 15:41:14 jmccrohan: are you running on battery? Jun 08 15:41:29 fullstop: Nope. Desktop system Jun 08 15:42:25 jmccrohan: okay. Some laptops will turn off ethernet on a disconnect, and won't turn it on until you go into the network manager and re-enable or you connect the PS. Jun 08 15:43:39 fullstop: thanks for the idea anyway. I'll have a poke around the Windows driver settings. Jun 08 16:14:40 jmccrohan: sounds like power safe going wrong Jun 08 17:39:44 mirko * r27136 /trunk/package/block-mount/ (Makefile files/fstab.config): Jun 08 17:39:44 [package/block-mount] remove example config file Jun 08 17:39:44 Having this file prevents the possibility of having target-specific Jun 08 17:39:44 fstab config files (as target/linux/${target}/base-files/etc/config/fstab). Jun 08 17:39:44 There is no need of a non-working example file on the target anyway. Jun 08 17:40:30 I disagree Jun 08 17:40:54 now when I install this I have to create the entire file myself? Jun 08 17:40:58 it will also break the ui Jun 08 17:41:44 xMff: targets aren't able to override this file Jun 08 17:42:00 which breaks targets as well which rely on mounting other partitions Jun 08 17:42:13 xMff: sorry, didn't know this is gonna break anything Jun 08 17:42:24 thats the problem with simple solutions Jun 08 17:42:41 mirko: see target/linux/ar71xx/base-files.mk for a solution Jun 08 17:43:21 KanjiMonster: hmm, i'm not sure if i wanna clal that a solution.. - but it would do the job for now indeed.. Jun 08 17:44:24 what's the issue with letting target-specific base-files override prior to that existing files at all? Jun 08 17:44:38 I wonder why per-target bas-efiles aren't copied during the target/linux/install phase Jun 08 17:44:47 then it would work as expected Jun 08 17:45:00 er can't remove stuff from packages everytime a target wants to override it Jun 08 17:45:04 xMff: ok, so that might not even be intentional Jun 08 17:45:05 s/er/we/ Jun 08 17:46:12 xMff: but deleting files before copy over target-specific files isn't a solution either - that would mean adjusting targets everytime a package changes (in regard of copying a file which the target has as well) Jun 08 17:46:45 therfore I wonder why target-base files are copied before packages Jun 08 17:46:53 and not after Jun 08 17:56:10 xMff: hmmm... stuff seems to depend that base-files/install is called quite early.. so we can't just postpone it to the end Jun 08 17:56:30 we could call it a second time in the end, however that's quite hackish Jun 08 17:58:23 blogic * r27137 /trunk/target/linux/ (7 files in 5 dirs): Jun 08 17:58:23 [lantiq] Jun 08 17:58:23 * fixes spi flash for ar9 Jun 08 17:58:23 * adds limited support for netgear dgn3500 (enough for others to start working with) Jun 08 18:00:01 xMff: ah, and we can't just call it a second time, since it got stamp'ed before and won't do anything Jun 08 18:03:39 *sigh* Jun 08 18:10:28 I still search the correct place to hook something between package install and image build Jun 08 18:14:15 xMff: the place just before where env/files/* are copied over should work out Jun 08 18:16:46 dan't find that... Jun 08 18:16:49 *cant Jun 08 18:23:07 me neither.. weird.. Jun 08 18:23:56 I hate it Jun 08 18:24:03 I always miss pacakge/Makefile :) Jun 08 18:24:08 its in there Jun 08 18:24:33 argh Jun 08 18:25:04 maybe we can just move the whole base-files/install section into this place? Jun 08 18:25:23 not the whole Jun 08 18:25:27 oh boy, this is gonna break something for sure.. Jun 08 18:25:30 but the target base-files Jun 08 18:25:41 okay Jun 08 18:25:44 and instead of move I'd rather copy it Jun 08 18:25:58 otherwise the base-files *package* is incomplete Jun 08 18:26:23 hihi ;) Jun 08 18:26:28 if someone decides to upgrade it he'd loose the target specific overrides Jun 08 18:27:32 he won't since even the target specific base-files are marked as conffiles Jun 08 18:28:16 ehrm, at least i _thought_ this is gonna happen.. Jun 08 18:28:45 xMff: '' for conffile in $(1)/etc/config/*; do '' inside base-files/Makefile Jun 08 18:29:09 I think we should move the base-files/install target specigfic base-file-overlay-handling into a macro in include/target.mk and then call that from base-files/install and rootfs-prepare Jun 08 18:36:21 build #45 of orion is complete: Failure [failed shell_4 compile_7] Build details are at http://buildbot.openwrt.org:8010/builders/orion/builds/45 Jun 08 18:40:27 mirko: http://luci.subsignal.org/~jow/target-base-files.patch Jun 08 18:54:04 xMff: looks good to me - however it should be documented why/that CopyTargetBaseFiles gets called twice Jun 08 18:54:24 like the rest of the system? ;) Jun 08 18:54:51 it doesn't work btw, package/Makefile does not include target.mk Jun 08 18:54:53 xMff: i was thinking about a hint in the commit msg ;) Jun 08 18:55:07 need to find a better way Jun 08 18:55:11 maybe in rules.mk Jun 08 18:56:16 xMff: hmm.. including target.mk shouldn't hurt - it doesn't do anything.. Jun 08 18:56:25 apparently it does hurt Jun 08 18:56:33 afterwards I get no target package/compile Jun 08 18:57:11 rules.mk however feels wrong.. Jun 08 18:57:21 it's just toolchain stuff Jun 08 19:00:33 what about a dedicated include/package-basefiles.mk ? Jun 08 19:01:00 xMff: was thinking about the same - agree.. Jun 08 19:01:20 MarkV: ping Jun 08 19:04:14 mirko: I updated the patch Jun 08 19:11:04 xMff: the patch doesn't remove the line which calls base-files/install - does it? Jun 08 19:11:22 in package/base-files/Makefile Jun 08 19:11:30 no it doesnt Jun 08 19:11:31 ah, ehrm Jun 08 19:11:33 i'm confused Jun 08 19:11:35 :) Jun 08 19:11:55 it should still call it.. Jun 08 19:12:04 if it works - fine by me.. Jun 08 19:12:39 reminds me of my pending patchsets.. Jun 08 19:12:45 I found out that I have to include rules.mk in package/Makefile to populate$(TARGET_DIR) Jun 08 19:12:55 I hope it has no further side effects Jun 08 19:14:05 let's figure them out Jun 08 19:14:10 "let's break trunk again.. lalala" Jun 08 19:17:28 hm does not work either Jun 08 19:22:36 I can add a quick hack to install base-files last Jun 08 19:22:48 thats the best solution I see atm Jun 08 19:26:31 okay, updated the patch again Jun 08 19:26:33 its way simpler now Jun 08 19:27:45 I have a package here, in my local tree, which overlays a few files in base-files.. I'll have to keep this in mind the next time I refresh my tree. Jun 08 19:28:06 yeah, that will not work anymore Jun 08 19:28:22 I try not to modify those files, but I have changed at least 2. Jun 08 19:29:54 I think the whole situation of two packages providing the same files is undefined Jun 08 19:30:09 if base files would be called zzz-base-files it would not happen Jun 08 19:30:31 if some other package cals itself zzzz it would override base-files again Jun 08 19:30:45 Yes, I figured out the order that the build process was doing and used that to my advantage. Jun 08 19:30:51 according to the ordering defined right now Jun 08 19:30:56 correct Jun 08 19:31:31 I tried to play nice but, in the end, I needed to change a little bit. Jun 08 19:31:54 well theoretically you could make a profile now Jun 08 19:32:01 and stuff your changed files in there Jun 08 19:32:17 in target/linux/foo/base-files-Profilename/ Jun 08 19:32:43 I modified lib/network/3g.sh ..and lib/wifi/mac80211.sh Jun 08 19:32:50 xMff: ok, if it works that way i would go for it - base-files won't be obsolete / not a dependency for an image anymore soon Jun 08 19:34:17 and those aren't part of base-files.. \o/ Jun 08 19:34:20 mirko: I will commit the patch that forces base-files to be installed last Jun 08 19:34:51 mirko: you could revert the extroot stuff then Jun 08 19:36:23 xMff: will do Jun 08 19:38:54 jow * r27138 /trunk/package/Makefile: [package] make sure that base-files is installed last, this allows per-target base-file overlays to work correctly Jun 08 19:39:50 blogic: ping Jun 08 19:44:40 blogic: or rather, why did you modify trunk/target/linux/*atheros*/config-2.6.37 in your *lantiq* commit? ;) Jun 08 20:45:49 mirko * r27139 /trunk/package/block-mount/ (Makefile files/fstab.config): Jun 08 20:45:49 Revert "[package/block-mount] remove example config file" Jun 08 20:45:49 This isn't needed anymore due to commit 27138 Jun 08 23:52:58 jow * r27140 /packages/ipv6/wide-dhcpv6/ (Makefile files/dhcp6c.conf files/dhcp6c.init): Jun 08 23:52:58 [package] wide-dhcpv6: Jun 08 23:52:58 - fix DUID calculation for virtual interfaces like pppoe-wan (#9503) Jun 08 23:52:58 - add a config option "duid" to allow users to override the DUID Jun 08 23:52:58 - use service_kill to cleanly terminate dhcp6c Jun 08 23:52:59 - bump package revision Jun 09 00:04:08 build #44 of rb532 is complete: Failure [failed compile_7] Build details are at http://buildbot.openwrt.org:8010/builders/rb532/builds/44 Jun 09 00:05:08 * xMff wonder who did the stund package Jun 09 00:31:53 well, if it hasn't been noticed in ~2 years... Jun 09 00:35:15 jow * r27141 /packages/net/stun/ (Makefile files/stund.init): [packages] stun: fix init script (#9528) Jun 09 00:37:06 I really dislike badly written initscripts Jun 09 00:37:22 if its just doing "killall foo" I do not need it Jun 09 00:44:35 ~94 killall instances Jun 09 00:47:10 best are those initscripts that manage to kill themselves :) Jun 09 00:49:41 okay, with patched busybox, linux reporting itself as 3.0 and therefore insmod working again I'm back at the used mem of .39 Jun 09 00:50:46 sadly, the majority at a glance **** ENDING LOGGING AT Thu Jun 09 02:59:56 2011