**** BEGIN LOGGING AT Wed Nov 02 02:59:57 2011 Nov 02 05:25:45 build #88 of uml is complete: Failure [failed compile_6] Build details are at http://buildbot.openwrt.org:8010/builders/uml/builds/88 Nov 02 06:48:57 build #83 of ixp4xx is complete: Failure [failed compile_6] Build details are at http://buildbot.openwrt.org:8010/builders/ixp4xx/builds/83 Nov 02 06:49:26 build #81 of kirkwood is complete: Failure [failed compile_6] Build details are at http://buildbot.openwrt.org:8010/builders/kirkwood/builds/81 Nov 02 07:19:49 build #83 of ar71xx is complete: Failure [failed compile_6] Build details are at http://buildbot.openwrt.org:8010/builders/ar71xx/builds/83 Nov 02 11:23:15 jow * r28714 /packages/multimedia/ushare/ (Makefile files/ushare.init): [packages] ushare: quote server name (#10322) Nov 02 11:27:05 jow * r28715 /branches/packages_10.03.1/multimedia/ushare/ (Makefile files/ushare.init): [backfire/packages] ushare: merge r28714 Nov 02 11:38:10 i'm looking for some documentation on atheros ar8216 switch Nov 02 11:38:10 anyone has access to it? Nov 02 11:38:11 nbd has, but it's likely under NDA Nov 02 11:41:44 hmm, luci really needs those ts_*.ko and seems dnsmasq too Nov 02 11:41:44 ? Nov 02 11:41:45 it neither depends on specific kmods nor on dnsmasq Nov 02 11:41:45 in the search of holy grail of space efficiency i blacklisted the ts_* modules, textsearch maybe? and dnsmasq wont start, luci complained and even crashed, dmesg showed kernel hexa errors and so on Nov 02 11:41:45 well most stuff is there for a reason Nov 02 11:45:20 it is, anyways, i will correct and sysupgrade (by the way does sysupgrade run even if my terminal crashed? it killed my pppd, wireless, everything..) Nov 02 11:45:20 i was connected via wireless :P Nov 02 11:52:52 by the way, how can i make a service not start after flashing? i included multiwan in the firmware but didnt configured it yet so i would like not to start until i have it ready.. Nov 02 11:56:19 like instruct the build process to do a /etc/init.d/multiwan disable Nov 02 12:25:05 anyway, very strange, seems like scripts/env doesnt handle kernel_menuconfig too Nov 02 12:25:22 i changed to another environment and i got the last kernel config on Nov 02 12:25:33 not cloned, ofc Nov 02 14:06:05 jow_laptop: I'm trying that luci-app-standalone makefile you sent me, but it fails to build. shows up nicely in menuconfig though: http://pastebin.com/shcXibxx Nov 02 14:09:03 I think you just need some ... || true in the install section Nov 02 14:09:14 it uses tests like [ -d ... ] && cp ... Nov 02 14:09:24 which may return with a non-zero exit code Nov 02 14:09:44 so change them to [ -d ... ] && cp ... || true Nov 02 14:10:41 yeah, was just trying that, Nov 02 14:11:03 looks like it was getting upset with the root and htdocs directories not existing and having contents. Nov 02 14:17:04 schweet Nov 02 14:17:19 htdocs can be used to provide any custom images or skins or things right? Nov 02 14:17:26 htdocs maps to /www Nov 02 14:17:38 so yeah, you would stuff css or images in there Nov 02 14:52:43 I've got my package installing a /etc/init.d script, but how do the links in /etc/rc.d get setup? are the etc/init.d scripts scanned looking for the START=XX STOP=XX lines? Nov 02 14:53:24 I was having a look at the uhttpd package for inspiration, as it has S50uhttp links, but they don't seem to be created in the package Makefile at least Nov 02 14:53:42 kind of, actually all init scripts execute rc.common which does it Nov 02 14:54:00 symlinks are set up during build, the build system will call /etc/init.d/foo enable for all found init scripts Nov 02 14:54:13 therfore it is also important to write the script in such a way that it is runnable on the host Nov 02 14:54:34 so if you check some uci variables for example you must do that within the start() procedure and not on the outside Nov 02 14:55:10 same for sourcing files that only exist on the target Nov 02 14:55:32 hmm, ok. good to know Nov 02 14:55:37 currently it's just: http://pastebin.com/6DYVzDNT Nov 02 14:55:41 so I guess it should be good Nov 02 14:56:01 yes Nov 02 14:56:10 will the symlinks still be installed even if you just opkg install, or only if it's built into a new image? Nov 02 14:56:21 only if builtin Nov 02 14:56:47 ok then, looks like it's coming up time to build a new image :) Nov 02 14:57:16 you can make packages "enable themselves" if you add a postinstall script which runs /etc/init.d/foo enable if $IPKG_INSTROOT is empty Nov 02 14:57:59 when IPKG_INSTROOT is defined, you run within the buildroot, if it is empty you run on the target Nov 02 14:58:17 sometimes it is needed to differentiate, e.g. in a postinstall Nov 02 15:10:57 hmm, if i run it with a proper init.d script, where does the console output go? Nov 02 15:11:49 of what? Nov 02 15:11:50 I was running them with & int he rc.local file earlier, and the console output went nicely into the "status->system log" Nov 02 15:12:13 well, nowhere Nov 02 15:12:33 s-s-d dups stdout, stdin, stderr to /dev/null Nov 02 15:12:44 as it is supposed to be if you daemonize properly Nov 02 15:13:32 what methods are available for me to log into the system log then? Nov 02 15:13:44 depends on the language Nov 02 15:14:04 for C there is syslog() Nov 02 15:14:10 for shell you have logger Nov 02 15:14:36 for Lua there is nixio.syslog() Nov 02 15:19:04 hmm, or, I could be a "bad" daemon, and use & and killall instead of s-s-d :| Nov 02 15:19:14 this is what I get for writing it as a console app Nov 02 15:20:03 don't you have some utility function to log output? Nov 02 15:20:25 just make it switch between printf and syslog, depending on some switch Nov 02 15:20:38 yeah, that's what i'm looking at now. Nov 02 15:20:58 I hate how init scripts and things end up taking so long to get right :) Nov 02 15:22:22 you could make it depend on parent_pid == 1 Nov 02 15:22:36 in this case you're most likely a daemon Nov 02 15:30:18 jow_laptop: I've added some of this to http://wiki.openwrt.org/doc/devel/packages Nov 02 16:15:12 jow_laptop: I'm trying this post inst step, to add the symlink, and it works, but it still reports errors? http://pastebin.com/xuzFLgWx Nov 02 16:15:19 the symlink is properly set up... Nov 02 16:16:22 karlp: exit 0 Nov 02 16:16:35 if it is already enabled it will return with non-zero exit code Nov 02 16:16:46 simply put exit 0 after the fi Nov 02 16:20:44 bah, of course. Nov 02 16:20:58 I was trying to test this with running the script again myself and inspecting $? Nov 02 16:21:02 but it was always 0 for me Nov 02 16:21:21 who knows Nov 02 16:21:33 maybe depends on whether it runs in an interactive shell or whatever Nov 02 16:22:07 also maybe add an echo indicating that the init script is getting enabled Nov 02 16:25:37 good idea. I've added a prerm step to call disable, removing the symlink too. Nov 02 16:25:45 just rebuilding the package and testing now Nov 02 16:40:01 yep, that's nicely working, I've added the relevant snippets to http://wiki.openwrt.org/doc/devel/packages#packaging.a.service Nov 02 17:41:55 build #70 of adm5120 is complete: Failure [failed compile_6] Build details are at http://buildbot.openwrt.org:8010/builders/adm5120/builds/70 Nov 02 17:58:01 build #108 of brcm63xx is complete: Failure [failed compile_6] Build details are at http://buildbot.openwrt.org:8010/builders/brcm63xx/builds/108 Nov 02 18:12:28 build #104 of at91 is complete: Failure [failed compile_4] Build details are at http://buildbot.openwrt.org:8010/builders/at91/builds/104 Nov 02 18:14:13 build #102 of ubicom32 is complete: Failure [failed compile_3] Build details are at http://buildbot.openwrt.org:8010/builders/ubicom32/builds/102 Nov 02 18:42:07 cshore * r28716 /trunk/package/block-extroot/: [package] block-extroot: Finally remove empty dir Nov 02 18:44:51 cshore * r28717 /trunk/package/block-mount/files/extmount.sh: [package] block-mount: Fixed module loading (double cat is wrong) Nov 02 20:15:47 what does WARNING: kmod-gpio-dev is not available in the kernel config mean? Nov 02 20:16:19 I selected that module by hand, shouldn't it have pulled in whatever it needed? Nov 02 20:16:42 it means that kmod-gpio-dev is dependant on kconfig symbols which are not present or disabled in the kernel config Nov 02 20:17:04 ahh, make kernel_menuconfig, Nov 02 20:17:37 is there anyway for a package to specify that it depends on a kernel config like that? Nov 02 20:17:46 nope Nov 02 20:18:24 kernel config has GPIO device support, under character devices, Nov 02 20:18:48 well maybe it got renamed Nov 02 20:19:09 I only upgraded from backfire branch, sometime after rc5 to rc6 :| Nov 02 20:19:39 GPIO_DEVICE Nov 02 20:19:43 is that enabled? Nov 02 20:19:54 yep Nov 02 20:19:58 y or m ? Nov 02 20:20:01 y Nov 02 20:20:08 well it wants m Nov 02 20:20:22 hmm, oh yeah. Nov 02 20:20:39 or, I take out the kmod package, as I odn't need it then Nov 02 20:20:45 right Nov 02 20:20:53 somewhere I'm missing the gpiodev.h file Nov 02 20:21:15 I know it's a legacy interface, but until my changes for sysfs on atheros get merged in, it's all that exists Nov 02 20:29:14 do you know what package I need to select to get this https://dev.openwrt.org/browser/trunk/target/linux/generic-2.6/files/include/linux/gpio_dev.h?rev=21787 Nov 02 20:29:18 well, any rev, Nov 02 20:30:06 or is my package just being built too early, before it's been copied into the staging_dir yet? Nov 02 20:32:39 hm nope, kernel tree should be prepared at the time normal packages are built Nov 02 20:42:25 hmm, ls staging_dir/toolchain-mips_gcc-4.3.3+cs_uClibc-0.9.30.1/usr/include/linux/ is missing gpio_dev.h Nov 02 20:43:07 can't you just ship a copy? Nov 02 20:43:25 sure, I'm just curious how it worked last week Nov 02 20:43:36 heh Nov 02 20:43:56 I mean, ls target/linux/generic-2.6/files/include/linux/ has the files Nov 02 20:44:05 s/files/file Nov 02 20:44:23 I don't _remember_ ever copying it into staging's include directories myself Nov 02 21:07:51 build_dir/linux_atheros/linux-2.6.30.10/include has the file, as does build_dir/toolchain/linux-xxx/include Nov 02 21:08:00 shouldn't both those be available to package builds? Nov 02 22:03:08 bah, I just went and checked out the old version of backfire from svn too, still no gpio_dev.h Nov 02 22:08:43 ok, gpioctl builds, and it uses it, but it was using -I$(LINUX_DIR)/include" Nov 02 22:08:49 I must have copied this file around at some point :( Nov 02 22:22:53 hmm, adding -I$(LINUX_DIR)/include to TARGET_CFLAGS isn't helping :| Nov 02 22:23:32 you need include $(INCLUDE_DIR)/kernel.mk Nov 02 22:23:50 to populate LINUX_DIR and friends Nov 02 22:25:07 I have no idea how I had this working in the past Nov 02 22:25:17 :) Nov 02 22:25:39 sometimes its a big surprise if you try to use your own stuff outside of the own development environment Nov 02 22:26:02 or the initial one in this case Nov 02 22:26:12 well, I just checked out backfire branch a month or two ago, Nov 02 22:26:18 and it all seemed nice and fine Nov 02 22:26:18 anyway, I'm gone for a while, good luck Nov 02 22:26:25 made dist clean and svn updated and rebuilt Nov 02 22:26:27 and ran into this Nov 02 22:26:32 yeah, hopefully that will fix things Nov 02 22:26:34 yep! Nov 02 22:26:35 working Nov 02 22:26:37 excellent Nov 02 22:26:47 I'll go fix up my makefiles and get the hell home **** ENDING LOGGING AT Thu Nov 03 02:59:59 2011