**** BEGIN LOGGING AT Mon Oct 31 02:59:56 2011 Oct 31 03:20:58 build #85 of ar7 is complete: Failure [failed compile_4 shell_14] Build details are at http://buildbot.openwrt.org:8010/builders/ar7/builds/85 Oct 31 03:48:33 build #107 of brcm63xx is complete: Failure [failed compile_6] Build details are at http://buildbot.openwrt.org:8010/builders/brcm63xx/builds/107 Oct 31 04:17:33 build #84 of sibyte is complete: Exception [exception shell_2 compile_1 shell_14 compile_12] Build details are at http://buildbot.openwrt.org:8010/builders/sibyte/builds/84 Oct 31 04:17:33 build #104 of s3c24xx is complete: Exception [exception compile_6 interrupted] Build details are at http://buildbot.openwrt.org:8010/builders/s3c24xx/builds/104 Oct 31 04:17:36 build #83 of avr32 is complete: Exception [exception shell_1 compile_1 shell_14 compile_12] Build details are at http://buildbot.openwrt.org:8010/builders/avr32/builds/83 Oct 31 04:17:38 build #99 of atheros is complete: Exception [exception compile_6 interrupted] Build details are at http://buildbot.openwrt.org:8010/builders/atheros/builds/99 Oct 31 04:42:09 build #88 of x86 is complete: Failure [failed compile_6 compile_12] Build details are at http://buildbot.openwrt.org:8010/builders/x86/builds/88 Oct 31 05:30:38 build #69 of adm5120 is complete: Failure [failed compile_6 shell_14 compile_12] Build details are at http://buildbot.openwrt.org:8010/builders/adm5120/builds/69 Oct 31 06:22:34 menuconfig is segfaulting on me when i try to /-search Oct 31 06:38:27 russell--: hey yeah - i saw that too... Oct 31 06:40:15 oh, and btw, freeswitch is the package that seems to be looking for /usr/bin/bunzip2 Oct 31 06:46:44 build #66 of octeon is complete: Failure [failed compile_6 compile_12] Build details are at http://buildbot.openwrt.org:8010/builders/octeon/builds/66 Oct 31 07:35:04 build #76 of iop32x is complete: Failure [failed compile_3 compile_12] Build details are at http://buildbot.openwrt.org:8010/builders/iop32x/builds/76 Oct 31 08:44:46 build #83 of xburst is complete: Failure [failed compile_6] Build details are at http://buildbot.openwrt.org:8010/builders/xburst/builds/83 Oct 31 09:29:39 acinonyx * r28689 /trunk/target/linux/ramips/image/Makefile: [ramips] rt305x: Create factory image for Fonera 2.0n Oct 31 09:43:54 acinonyx * r28690 /trunk/.gitignore: .gitignore: Ignore Emacs editor generated files Oct 31 12:54:43 what's a good example of an app using libuci for config? Oct 31 12:55:22 I had a look at ntpclient for instance, thinking it was nice and small, but it seems taht UCI is only used in the init script, to read from /e/c and pass them as command line params Oct 31 12:55:52 I'm looking at moving my app to use libuci to make it easier (hopefully) to hook up my config file in the luci gui Oct 31 12:58:17 or is that the general way, don't integrate libuci directly, just provide command line options? Oct 31 13:09:36 build #73 of ep93xx is complete: Failure [failed compile_6] Build details are at http://buildbot.openwrt.org:8010/builders/ep93xx/builds/73 Oct 31 13:14:30 cool, freeswitch is forkbombing me now :-/ Oct 31 13:16:44 Makefile:2238: build/openwrt_rules.mk: No such file or directory Oct 31 13:16:44 make[2560]: Entering directory `/aux/src/openwrt/build_dir/target-mipsel_uClibc-0.9.32/freeswitch-1.0.7' Oct 31 13:16:52 etc, etc, etc Oct 31 13:18:37 mazilo-- ;-) Oct 31 13:20:41 russell--: Whatzzzzzup? Oct 31 13:21:05 karlp: there is new way to use it, for example check out netifd Oct 31 13:21:47 karlp: for now i use "old" way with owsip: http://dev.phrozen.org/gitweb/?p=owsip.git;a=summary Oct 31 13:22:03 see config.h and config.c for examples Oct 31 13:22:56 russell--: Which Makefile bombed out? Oct 31 13:23:00 mazilo: your patch to freeswitch points at a build/openwrt_rules.mk that doesn't seem to exist Oct 31 13:24:12 i did a make -j17 BUILD_LOG=1 IGNORE_ERRORS=m V=99 package/freeswitch/{clean,compile} and it was doing its best to kill my machine ;-) Oct 31 13:24:53 afaict Oct 31 13:25:07 russell--: The latest patch I committed will create the build/openwrt_rules.mk file. If you build FS from scratch, this file will created from the included patches/000-build-openwrt_rules.mk.patch file. Oct 31 13:25:50 okay, i was building r28686 Oct 31 13:25:57 When you execute 'make package/freeswitch/clean, make sure the freeswitch_git directory is completely deleted. Oct 31 13:26:22 luka12345|wiik: thanks, looking now... Oct 31 13:27:33 mazillo, where does that directory live? Oct 31 13:28:16 russell--: feeds/packages/net/freeswitch/patches/000-build-openwrt_rules.mk.patch Oct 31 13:28:36 i mean freeswitch_git Oct 31 13:30:14 i don't have one of those feeds/packages/net/freeswitch/patches/000-build-openwrt_rules.mk.patch in my tree Oct 31 13:31:08 russell-->: When you do a 'make package/freeswitch/compile', it will create a 'build_dir/target-/freeswitch_git' directory and then extract the 'dl/freeswitch_git-50328a663.tar.bz2' (with a --strip-components=1 to remove its own root directory) Oct 31 13:34:10 000-build-openwrt_rules.mk.patch isn't in svn either Oct 31 13:34:24 you sure you commited that? Oct 31 13:34:40 russell--: Hold on, I probably forgot to upload it. Oct 31 13:35:27 mazilo * r28691 /packages/net/freeswitch/patches/000-build-openwrt_rules.mk.patch: added missing patches/000-build-openwrt_rules.mk.patch Oct 31 13:35:52 that should help ;-) Oct 31 13:35:53 russell--: Canyou please confirm if the latest commit has the 'patches/000-build-openwrt_rules.mk.patch" file? Oct 31 13:37:39 it's there now ;-) Oct 31 13:38:39 russell--: Thanks. Oct 31 13:38:42 retrying the build Oct 31 13:39:23 mazilo: there was a problem before also with bunzip2 Oct 31 13:39:50 russell-->: I belive you were talking about flite-1.5.x? Oct 31 13:39:54 the package wanted to find it in /usr/bin/bunzip2 whereas my debian box has it in /bin/bunzip2 Oct 31 13:40:46 russell--: I think that was addressed by FS with the latest commit. Oct 31 13:41:36 ok Oct 31 13:42:20 russell--: Take a look at "http://fisheye.freeswitch.org" and you will find the commit of 50328a663452006b78e79e2004c20d111ffb263f (latest ATM) has that fixed. Oct 31 13:43:54 russell--: Or this one (http://fisheye.freeswitch.org/qsearch?q=50328a663452006b78e79e2004c20d111ffb263f) Oct 31 13:45:21 russell-: Than you for letting me know the missing 'patches/000-build-openwrt_rules.mk.patch" file. Cheers. Oct 31 13:45:46 [864729.049631] Out of memory: Kill process 1507 (sshd) score 1 or sacrifice child Oct 31 13:45:49 [864729.050300] Killed process 1509 (sshd) total-vm:79488kB, anon-rss:116kB, file-rss:0kB Oct 31 13:47:20 russell--: Try to compile with a smaller "X" value for -jX switch? Oct 31 13:49:27 russell--: I am using an AMD64 PhenomII X3 with 4GB RAM with an 'ionice -c3 nice -n20 make package/freeswitch/compile QUILT=1 BUILD_LOG=1 V=99 -j4' to cross compile FS in background without a problem for BRCM47XX platform. Oct 31 13:49:49 s/am using/used/ Oct 31 13:53:53 mazilo, the freeswitch compile was looping on the missing openwrt_rules.mk before ... the paste was from then, the fix is still working Oct 31 13:54:37 russell-: ??? Oct 31 13:55:28 my out-of-memory message was pre-your-latest-fix Oct 31 13:55:36 -j17 is working okay now Oct 31 13:56:05 at least, so far, still building Oct 31 13:56:17 russell-: OK. On my system, if I do a '-j20', the compilation behaves erratic. Oct 31 13:56:36 russell--: What hardware platform is your host? Oct 31 13:57:41 russell--" On mine, it is an AMD64 PhenomII X3 with 3 CPUs. So, I just use a '-j4' switch. Oct 31 14:06:22 mazilo: it's an core i7 quadcore Oct 31 14:06:38 compile failed with this: Oct 31 14:06:42 Compiling gsmopen_protocol.cpp... Oct 31 14:06:42 gsmopen_protocol.cpp:5:28: fatal error: gsmlib/gsm_sms.h: No such file or directory Oct 31 14:06:45 compilation terminated. Oct 31 14:08:18 russell--: can you please PasteBin the failing part? Oct 31 14:09:54 russell--: Has gsmlib been built? Oct 31 14:15:43 mazilo: here's the whole http://www.personaltelco.net/~russell/freeswitch-compile.txt Oct 31 14:16:12 * russell-- is doing (or trying) a Select-All-Packages build Oct 31 14:16:17 for brcm47xx Oct 31 14:18:36 russell--: 'gsmlib/gsm_sms.h' file is part of gsmlib header files. Make sure 'gsmlib' package has been compiled and installed in 'staging_dir/target-mipsel_uClibc-0.9.32/root-brcm47xx/usr/include/gsmlib/gsm_sms.h' Oct 31 14:19:27 shouldn't dependencies take care of that? Oct 31 14:20:05 russell--: I would think so. But, I encountered problems with dependencies build when compiling FS and I don't know why. Oct 31 14:20:06 it isn't there Oct 31 14:20:40 * russell-- will see if i can build it manually Oct 31 14:20:53 russell--: If you manually build gsmlib first, i.e. 'make package/gsmlig/compile', I believe this will resolve the issue. Oct 31 14:25:55 i don't see that in packages Oct 31 14:26:01 gsmlib, that is Oct 31 14:29:57 it's called gsm Oct 31 14:31:46 russell--: OK. Perhaps, I need to change the FS Makefile for OpenWRT to reflect that. Oct 31 14:32:44 russell--: It is called libgsm (feeds/packages/libs/gsm/Makefile). Oct 31 14:32:50 juhosg * r28692 /trunk/target/linux/ar71xx/base-files/etc/defconfig/wnr2000/network: Oct 31 14:32:50 ar71xx: remove default network config of the wnr2000 board Oct 31 14:32:50 It is the same as the generic configuration. Oct 31 14:32:52 juhosg * r28693 /trunk/target/linux/ar71xx/base-files/etc/uci-defaults/rb750: ar71xx: remove LED setup of nonexistent interfaces on RB750 Oct 31 14:34:56 * russell-- has to crash, back in 8 hrs or so Oct 31 14:42:44 juhosg * r28694 /trunk/target/linux/ar71xx/base-files/etc/defconfig/wnr2000/: ar71xx: remove empty dir after r28692 Oct 31 14:57:49 luka12345|wiik: I've got a simple uci file, /etc/config/mine, with just config 'mysection' and two options Oct 31 14:58:22 if I use uci_foreach_element(&pkg->sections, e) it can see the option values, but not if I use, uci_lookup_section() Oct 31 14:58:52 it seems to have turned the "config 'mysection'" into an element with the name cfg027713, some random anonymous name I presume Oct 31 14:59:34 do you know how uci_lookup_section is meant to work? Oct 31 15:00:04 uci_lookup_section looks up a section by name Oct 31 15:00:16 not by type Oct 31 15:01:09 so a line in the config file is actually "config [type] [name]" ? Oct 31 15:01:18 yes Oct 31 15:01:24 and the name is optional Oct 31 15:01:25 right, that's what I was missing :) Oct 31 15:03:45 is there anyway to get more than just a string back from uci_lookup_option? can I get magic parsing of integers/booleans? Oct 31 15:04:01 like, parsing hex, or yes/no/true/false for booleans? Oct 31 15:04:28 it only supports strings and lists Oct 31 15:21:57 ok, that was relatively painless, Oct 31 15:22:06 now to try and make it show up in luci :) Oct 31 15:49:59 build #101 of brcm47xx is complete: Exception [exception interrupted] Build details are at http://buildbot.openwrt.org:8010/builders/brcm47xx/builds/101 Oct 31 15:50:00 build #75 of etrax is complete: Exception [exception interrupted] Build details are at http://buildbot.openwrt.org:8010/builders/etrax/builds/75 Oct 31 15:50:03 build #71 of gemini is complete: Exception [exception interrupted] Build details are at http://buildbot.openwrt.org:8010/builders/gemini/builds/71 Oct 31 15:50:05 build #70 of rdc is complete: Exception [exception interrupted] Build details are at http://buildbot.openwrt.org:8010/builders/rdc/builds/70 Oct 31 15:53:48 nbd: how far behind is the version of libuci in backfire? Oct 31 15:54:01 it seems to not have the uci_lookup_option and lookup_option_string? Oct 31 15:55:33 i guess i could update it there Oct 31 15:55:56 no! Oct 31 15:56:02 ok Oct 31 15:56:03 at least not now Oct 31 15:56:06 hehe Oct 31 15:56:11 i'll wait until your builds are done Oct 31 16:00:38 nbd, I've tried just inlining those methods, but the backfire version doesn't have uci_lookup_next either, is there any rough way of getting similar functionality, or should I not even try? Oct 31 16:01:41 you can probably just copy it from the uci source into your own source Oct 31 16:01:44 or make a compat wrapper or something Oct 31 16:01:55 it does not depend on any data structure changes Oct 31 16:02:04 well, that's what I've started, but it's getting to be a bit bigger than I was expecting :) Oct 31 16:03:16 there's no version info anywhere that I can use to do this at compile time is there? Oct 31 16:10:55 you could just copy package/uci from trunk for your own build Oct 31 16:11:01 and save yourself the trouble of making a compat layer Oct 31 16:13:10 probably a better idea. Oct 31 16:13:45 currently, just added 70 lins copied and pasted from trunk, it's enough for my configs for now. Oct 31 16:23:00 don't forget to rebuild libuci-lua and swconfig as well if you replace your local libuci Oct 31 16:23:17 the buildroot will not do that Oct 31 16:25:08 is this: http://luci.subsignal.org/trac/wiki/Documentation/ModulesHowTo Oct 31 16:25:21 particualrly the CBI section, the current best reference on how to hook a /e/c file up to luci? Oct 31 16:26:06 the best reference is still the code of other applications Oct 31 16:26:51 however the howto will lead you to a working page so yes, it a good reference Oct 31 16:28:22 I guess I need to pick a "good" application to look at then :) Oct 31 16:28:57 cbi models can get very complex, depending on your requirements Oct 31 17:16:42 jow_laptop maybe you dev guys can enable by default CONFIG_IKCONFIG and CONFIG_IKCONFIG_PROC so people can inspect a running kernel's .config ? i wonder how much would add to the kernel, i can gzip my 3.1.0 51k .config to 13k.. Oct 31 17:16:58 no way Oct 31 17:17:04 its a waste of space Oct 31 17:17:15 it kindof is Oct 31 17:17:35 yes it is. especially as regular openwrt builds do not deal with kernel configs Oct 31 17:17:54 so its kind of pointless to embed a kernel .config which merely is a result of defaults Oct 31 17:18:08 ah,hmm Oct 31 17:18:40 some architectures like rdc are very space contrained when it comes to kernel image size Oct 31 17:19:30 hmm, a crazy idea: could be cool if kernel devs could real time compile a certain selected option in make kernel_menuconfig so you can see in real time how much size to the kernek would add if you select that :) Oct 31 17:19:57 but i guess there are too man deps to see that info in the menu Oct 31 17:20:05 yes Oct 31 17:20:14 it also varies wildly among architectures Oct 31 17:20:26 and compression does the rest to make the size changes impossible to know in advance Oct 31 17:20:37 indeed Oct 31 17:37:35 jow_laptop: I get "Unable to read UCI data: rme_mrelay" when I access my page via make runhttpd. but "uci show rme_mrelay" dumps the config ok from the commanbd line. Oct 31 17:37:39 is this going to be permissions? Oct 31 17:37:52 I take it even my local luci will need write access to the file? Oct 31 17:38:51 or, is runhttpd looking in [luci]/host/etc/config, rather than the machine's /etc/config? Oct 31 17:38:52 could be that it indeed looks in /etc/config/ on your host system Oct 31 17:39:24 but double check permissions Oct 31 17:39:54 another common cause is a parse error Oct 31 17:41:03 ok, it was looking in host/etc/config, and the file didn't exist there. Oct 31 17:41:18 kk Oct 31 17:41:19 can't luci create the file if it doesn't exist? or is that out of scope Oct 31 17:41:40 its probably doable Oct 31 17:41:49 we just never had a reason to Oct 31 17:41:57 yeah, fair enough Oct 31 17:43:48 in my controller, I can get a new tab to show up with entry({'admin', 'network', 'myname'}, cbi(blahblah)) Oct 31 17:44:03 but I can't make a new top level tab, (to sit at a level alongside network) Oct 31 17:44:13 do I need to register that tab somewhere else? Oct 31 17:44:19 why not? Oct 31 17:44:27 it just doesn't show up Oct 31 17:44:35 entry({'admin', 'foo'}) should spawn a new top level tab foo Oct 31 17:46:38 also make sure that it has a title Oct 31 17:46:44 otherwise it becomes hidden Oct 31 17:47:06 also "runhttpd" spawns lucid afair which does in-memory caching of the menu tree Oct 31 17:47:13 so you need to restart it Oct 31 17:47:25 make runuhttpd should work better in this case Oct 31 17:47:46 make run... MODULES="" helps to speed thing up (do not rebuild everything on each invocation) Oct 31 17:53:11 should I be able to edit the lua files and press f5, or do I need to restart the server each time? Oct 31 17:53:41 if at all the webserver restart is only needed if you change the index() function in a controller Oct 31 17:53:50 since that is cached Oct 31 17:53:56 or may be cached Oct 31 17:54:07 the rest should be effective with a simple browser reload Oct 31 17:55:01 do you think http://pastebin.com/EkxYfYdt is getting in the way? Oct 31 17:55:15 I've got lots of errors about freifunk, and missing passwords, and so on, Oct 31 17:55:56 does it happen on every request? Oct 31 17:57:56 only if I click the "administration" link in the rop right, where it as "essentials, administration, NIU, freifunk" Oct 31 17:58:04 or on the freifunk tab Oct 31 17:58:33 the path information in the 500 causes html validation problems too, Oct 31 17:59:02 the <> in the stack traceback needs to be html escaped Oct 31 18:00:11 cannot be cleanly solved anyway as no footer is sent afterwards Oct 31 18:01:02 ok, that's not my concern anyway really :) Oct 31 18:02:34 ok, got the top level tab, what's the trick to have it be "empty" just having the other tabs? Oct 31 18:03:02 or should I just make the top level entry poitn to whatever CBI I want to be default? Oct 31 18:03:32 let's just try that instead of asking .) Oct 31 18:04:29 whee Oct 31 18:04:47 if you have a very recent luci you can use "firstchild()" as action Oct 31 18:04:58 it will simply redirect to the first submenu entry then Oct 31 18:05:13 alternatively simply point to your initial cbi, yes Oct 31 18:07:14 so, can [luci]/applications/myapplication just be installed like any other package? if I want to put this in feed? Oct 31 18:07:37 can I just use this two line Makefile including config and module? Oct 31 18:08:06 or, is it better to add this luasrc directory to the package that this is manage? Oct 31 18:08:14 moment Oct 31 18:08:33 http://luci.subsignal.org/~jow/luci-app-standalone.tar.gz Oct 31 18:08:50 use that as template Oct 31 18:10:35 thank you Oct 31 18:11:34 are there any notes on what goes in the ipkg folder? http://wiki.openwrt.org/doc/devel/packages makes no mention of it. Oct 31 18:12:05 that is a luci specific anomaly to define preint and postinst actions Oct 31 18:12:19 usually you would simply use define Package/.../preinst ... endef Oct 31 18:12:58 somewhat related, if you build a package into the image, when do the postinst sections get run? Oct 31 18:13:03 first boot? never? Oct 31 18:13:15 on the host system, during build Oct 31 18:13:37 is there any common way of doing something once, on the target? Oct 31 18:13:59 yes, place a script in .../etc/uci-defaults/ Oct 31 18:14:15 it gets run once on first boot and is removed then Oct 31 18:14:30 so that can be in the files/ of my package right? Oct 31 18:14:50 for luci packages I usually add a postinstall script too which simply sources and deletes /etc/uci-defaults/... if -z $PKG_INSTROOT Oct 31 18:15:16 so it gets executed immediately after opkg install Oct 31 18:17:00 alright, well, lots to experiment with tomorrow then I think :) Oct 31 20:15:24 Has anyone here tried to compile ubus package (required by 'netifd' package)? I tried to build it after a 'make dirclean' and the compilation of 'ubus' package spits out some error messages as shown here (http://pastebin.com/cCUHHzwg). It looks like related to 'cmake' package which is under tools/cmake. Oct 31 20:15:57 gcc 4.6 is not default Oct 31 20:16:13 jow_laptop: OK. Thanks. Oct 31 23:12:38 mazilo: nbd fixed the unused var upstream Nov 01 00:56:50 swalker: Thanks. **** ENDING LOGGING AT Tue Nov 01 02:59:57 2011