**** BEGIN LOGGING AT Mon Jun 22 02:59:57 2009 Jun 22 03:20:06 update on the perl/sigchld problem. i just managed to natively build perl using a native mipsel compiler from a uclibc buildroot image and it does *not* have the sigchld problem. Jun 22 10:33:32 i solved the mipsel perl signal handling problem. narrowed down and tested. someone please apply this asap: http://pastebin.ca/1469712 Jun 22 10:35:00 russell_: so the signal specifications were just wrong? Jun 22 10:35:22 i guess so Jun 22 10:35:38 i got the new values from a native toolchain autoconfiguration Jun 22 10:35:48 I see Jun 22 10:37:27 there are a bunch of other differences evident in a diff, but that's the one that seemed most relevant to my problem and that patch alone solved it Jun 22 10:37:45 * russell_ took the minimalist approach Jun 22 10:38:18 jow * r16539 /packages/lang/perl/ (Makefile files/config.sh-mipsel.in): [packages] perl: fix signal specifications in mipsel config, thanks Russell Senior! Jun 22 10:38:52 thanks! Jun 22 10:38:55 np Jun 22 11:42:44 phaidros: i'm gonna start working on some templates tonight. i'm thinking: router description (will consider links/redundancy problems w/ TOH), package description, bootloader description/flashing instructions Jun 22 11:44:47 danage: sounds good to me! Jun 22 11:46:53 i figured flashing procedure is best to be described for each boot loader and not router. this can be linked to from the router description page. if there is router specific flashing instructions, then that can remain on the router description page. this solves a redundancy problem Jun 22 11:47:19 smart! Jun 22 13:03:24 nbd: ping Jun 22 13:04:29 fofware: still trying to figure out the segfault? Jun 22 13:05:56 hello xMff, yes, I did find the issue is with strsep a function of string.h Jun 22 13:07:14 did you run your code under gdb? Jun 22 13:07:50 sometimes an error happens way earlier and results in a segfault dozens of instructions later Jun 22 13:07:54 maybe I need include somthing more, all that i did compile and linked fine, but this function give me error when the string have the separator Jun 22 13:11:10 a gdb backtrace would help I think Jun 22 13:13:25 I understand, but i guess nbd can know what happend because util.h of libuci call this function and in programs come from uci package run Jun 22 13:14:38 I think it's a user error :) Jun 22 13:15:20 xMff: realy I don't know if code run under gdb, i did a simple c program where set an string and try use strsep and givme the same message Jun 22 13:15:52 Yes, I think is something that i'm doing wrong too Jun 22 13:17:11 could you pastebin your code? Jun 22 13:17:40 I did read in string.h string to strsep must be end with "\0" I did set that but nothing changed Jun 22 13:17:44 yes Jun 22 13:17:54 giveme a minute Jun 22 13:23:53 xMff: http://openwrt.pastebin.com/d7ac8643e in lines, 21 I set string, if I set string without "." work fine, but if I set "adfad.asdfaf" Segmentation fault Jun 22 13:24:55 maybe the string have a wrong format, I don't know Jun 22 13:38:44 mirko * r16540 /feeds/phone/fso-gpsd/Makefile: make init script executable on target, fix category name Jun 22 13:40:43 fofware: iirc strsep() modifies it's first pointer argument which is the only reference to your source string Jun 22 13:41:15 fofware: so best is to define your string as const char foo[] = "..." and then make a separate pointer to it which you pass as first arg to strsep() Jun 22 13:43:21 Ok I will try Jun 22 13:43:50 fofware: http://www.ia.pw.edu.pl/~wujek/dokumentacja/gnu/libc/libc_5.html#SEC65 see the end of this page for an example Jun 22 13:55:19 thanks xMff, I will try like you say Jun 22 13:56:46 only for confirm, const char foo[]=".", should be the same that "." right? Jun 22 14:02:05 no, it should hold your key, like "network.wan.ipaddr" or watever Jun 22 14:04:34 nope, It's something more wrong Jun 22 14:10:20 fofware: compile it on x86, with the normal gcc, run it under gdb until it segfaults and execute the command "bt" to get a backtrace Jun 22 14:12:21 Ok I will build x86 staff and after I will ask to how to do Jun 22 14:16:37 well basicly you just need to download uci form openwrt.org, compile it, install it locally on your system Jun 22 14:16:45 then you can build and link against it locally Jun 22 14:24:04 xMff: I did copy exactly the example of strsep, compiled, tryed, and It's working Jun 22 14:26:16 so.... now I will try to call uci functions... obviusly the issue is in my strings initialization Jun 22 15:08:20 xMff: yes the issues was my string definition and initialization, to good use need char foo[]="......." and char *bar and after bar=foo, with this all work perfect Jun 22 15:08:50 thanks xMff Jun 22 15:10:15 now I can continue learning about use of uci inside some program and maybe I can write a chillispot.patch to read and write uci.files directly Jun 22 15:10:23 thanks very much Jun 22 15:12:30 If I can done that after I will try other for coovachilli and I will send to david, developer of coova maybe he include it in next version if he think is a good idea Jun 22 16:41:16 should ndyndns get svn rm'ed in favor of using ddns-scripts instead? Jun 22 17:24:13 gah. cgilib compile is screwing me up. doesn't build, is trying to build even though i've got it turned off in menuconfig Jun 22 17:36:31 Hello, the issue that i did have was by wron initialization of string... thanks by your help too Jun 22 17:36:59 i am always willing to give uninformed guesses Jun 22 19:04:23 hmm. more image non-creation Jun 22 19:05:04 weird. i tried to add some stuff to the image, now no image Jun 22 19:05:18 only adding stuff that built, afaik Jun 22 19:48:38 any LuCI guys here? I was wondering about CBI for the list UCI datatype Jun 22 19:48:43 if there's an example Jun 22 19:56:02 aport`: the type is DynamicList Jun 22 19:56:38 and an example is the additonal "DHCP-Options" field in Network -> DHCP Jun 22 20:00:28 aport`: http://luci.subsignal.org/trac/browser/luci/trunk/modules/admin-full/luasrc/model/cbi/admin_network/dhcp.lua#L63 Jun 22 20:00:31 pretty boring Jun 22 20:06:39 xMff, thanks :) Jun 22 20:35:30 [florian]: cgilib is build failing for me on wgt634u Jun 22 20:50:47 <[florian]> russell_: ok, will have a look at it Jun 22 20:57:11 xMff do you have time now to look at the dependency problem with the perl package? Jun 22 21:00:24 cgilib doesn't even build on x86 natively Jun 22 21:00:35 something eff'd up about libtool Jun 22 21:02:11 russell_ with your patch for perl it also compiles for i486. just the dependency problem for libdb/libgdbm is still there Jun 22 21:02:41 signals seemed to work on x86 builds for me Jun 22 21:30:15 [florian]: cgilib builds for me on native x86 (on my debian build machine), but only after: aclocal ; autoconf ; libtoolize --force ; automake --add-missing --foreign (from INSTALL). after that './configure ; make' works fine. Jun 22 21:30:37 not sure how that translates to buildroot Jun 22 21:50:12 russell_: You can add a Build/Configure target to the Makefile Jun 22 21:50:29 (cf. include/package-defaults.mk) Jun 22 21:51:00 s/target/define/ Jun 22 21:51:20 phaidros: sorry no progress tonight i got distracted playing with ubuntu one Jun 22 22:16:39 xMff ? Jun 22 22:19:03 damn. it worked (or at least, built). [florian] http://pastebin.ca/1470412 Less Broken Than Before(tm) **** ENDING LOGGING AT Tue Jun 23 02:59:57 2009