**** BEGIN LOGGING AT Sat Oct 20 03:00:03 2012 Oct 20 05:45:25 nbd: r33516 was the commit that broke x86-generic (net45xx was the one i bisect'd) when CONFIG_ALL=y Oct 20 05:55:23 broke kernel configuration, that is. now i'm off to try to find what causes the ones that build to kernel panic on boot Oct 20 12:35:06 russell--: show me the boot panic Oct 20 12:35:31 russell--: and on what device does it occur? Oct 20 16:13:17 Hello guys, I would love to say that I am very thankful for the existence OpenWrt and all the community around it. Without you, guys, devs, there would not be any of it. Thank you so much! //sry for my poor english Oct 20 22:01:02 nbd: just built r33867 with config patches for x86-generic-net45xx and it didn't panic ... /me suspects weirdness in my normal build tree (also built alix2 in the same source tree... wonder if there is some cross-contamination), it was pretty reliably panic'ing for me until this last distclean build Oct 20 22:15:22 the problem i was seeing was on a CONFIG_ALL=y build for x86-generic-net45xx for recent trunk, i'm trying again with a custom package set right now in my normal build tree Oct 20 22:27:02 hmm, there's the panic again Oct 20 22:27:15 * russell-- going to try reflashing the same image to confirm Oct 20 22:35:08 it didn't panic this time (maybe a flashing problem), sysupgrade is broken, but i got something else i've seen before Oct 20 22:39:15 here's the bootlog: http://sprunge.us/BMji Oct 20 22:39:29 i get that illegal instruction, then no console Oct 20 22:47:30 odd Oct 20 23:25:35 nbd: yes Oct 20 23:25:56 * russell-- trying again with a clean newly built .config Oct 20 23:31:53 nbd, yeah, with this build tree (with packages feed included, and some other targets built) does the same thing, that illegal instruction, then no console Oct 20 23:32:15 that is, i see output, carriage returns are echo'd, but no shell Oct 20 23:34:58 trying make target/linux/clean and a rebuild Oct 20 23:35:01 show me the output of ./scrpts/diffconfig.sh Oct 20 23:38:03 before you do a rebuild Oct 20 23:38:09 doh Oct 20 23:38:19 just cancel the build and show me the output Oct 20 23:40:10 http://sprunge.us/GhIO Oct 20 23:41:11 try a make dirclean Oct 20 23:41:33 do you know if your geode has mmx and 3dnow? Oct 20 23:41:58 because your alix2 build had -march=geode -Os -mmmx -m3dnow in the cflags Oct 20 23:42:08 ... which are used for the toolchain as well Oct 20 23:43:22 not sure Oct 20 23:43:38 that'll be in /proc/cpuinfo? Oct 20 23:43:44 i think so Oct 20 23:44:01 'flags' Oct 20 23:44:10 * russell-- looks another net45xx we've got deployed Oct 20 23:45:57 flags: fpu Oct 20 23:46:48 http://sprunge.us/FgZM Oct 20 23:46:58 that's the whole cpuinfo dump Oct 20 23:47:09 ok, then that's probably it Oct 20 23:47:18 yeah, that sounds likely Oct 20 23:47:20 after a make dirclean it should produce working images again Oct 20 23:47:47 okay, how do i keep alix2 from contaminating net45xx in future? Oct 20 23:59:56 not sure yet Oct 20 23:59:58 i'll think of something Oct 21 00:00:05 for now, use different build trees Oct 21 00:00:52 i will probably change something to add a arch-specific suffix to the toolchain/target dir name Oct 21 00:00:56 as was done on arm Oct 21 00:04:06 cool Oct 21 00:04:30 * russell-- will report back on results, to confirm the theory, full build will take 3 hours or so Oct 21 00:06:35 3 hours? what kind of machine is that? Oct 21 00:07:03 i7 quad core with 24 gig of ram Oct 21 00:07:36 at least, i vaguely remember the 3 hour figure Oct 21 00:07:39 hm, i'd expect it to be a bit faster, but i haven't done CONFIG_ALL builds in a while Oct 21 00:07:52 are you using make -j8? Oct 21 00:07:54 with a packages feed Oct 21 00:07:59 ah Oct 21 00:08:02 um, i think it's -j5 Oct 21 00:08:21 without the packages feed it is much faster Oct 21 00:08:26 on my i7 i tend to get the best performance with -j8 Oct 21 00:08:33 (hyperthreading enabled) Oct 21 00:08:40 i haven't done a thorough test Oct 21 00:09:04 i only looked at it because i did quite a bit of work on improving build system parallelization Oct 21 00:09:09 like comparing -jN's but a friend who works at intel suggested sticking to number of cores Oct 21 00:09:24 but i don't recall if he was just guessing too Oct 21 00:09:33 probably depends on the OS as well Oct 21 00:09:53 he was on linux too i think Oct 21 00:09:57 i've been doing my tests on mac os x Oct 21 00:10:00 ah Oct 21 00:10:05 (my main work machine is a mac) Oct 21 00:10:28 i've got an 8 core xeon available too Oct 21 00:10:48 that's where i did the bisecting stuff Oct 21 00:10:56 and where i was building working images ;-) Oct 21 00:11:15 all distclean'd every time Oct 21 02:50:56 any thoughts on taking http://patch-tracker.debian.org/patch/series/view/python2.7/2.7.3-5/sqlite-rpath.diff for #12239? i486-openwrt-linux-uclibc-gcc: error: unrecognized option '-R' **** ENDING LOGGING AT Sun Oct 21 02:59:58 2012