**** BEGIN LOGGING AT Fri Dec 14 03:00:02 2018 Dec 14 06:57:34 khem, hi Dec 14 06:57:41 I tried to build with sumo and it failed ( https://pastebin.com/gUCviDwH ). Dec 14 06:58:01 with glibc is ok: https://pastebin.com/buMUaeZE Dec 14 06:58:43 Error: opcode not supported on this processor: mips3 (mips3) `sdc1 ... Dec 14 07:00:30 this nite I built with oe-master and GCCVERSION ?= "7.%", also ok with default TCLIBC Dec 14 07:01:18 (fwiw https://pastebin.com/daEiJbKi) Dec 14 07:02:45 afais can only be miscompiled assembler/linker Dec 14 07:02:55 I don't think gc itself Dec 14 07:17:04 strange bug Dec 14 07:17:16 * ant_home heading to office Dec 14 07:17:19 bbl Dec 14 13:39:38 khem, about kernel 3.13.5 vs musl Dec 14 13:39:40 so CC=mipsel-oe-linux-gcc is good but CC=mipsel-oe-linux-musl-gcc is bad Dec 14 13:40:02 arnd> ant_home: as I said, no idea why it would depend on the libc you use, but it explains why you get that error when you use an unmatched gcc configuration Dec 14 13:40:03 i.e. a gcc targetting a CPU that lacks instructions from the .S file Dec 14 13:41:15 ant_home: assuming this is a BMIPS based CPU, the kernel passes "cflags-$(CONFIG_CPU_BMIPS) += -march=mips32 -Wa,-mips32 -Wa,--trap" Dec 14 13:41:15 but the assembler complains that it's targetting mips3, so that option did not reach it Dec 14 14:30:25 kergoth: https://github.com/rossburton/meta-ross/blob/master/classes/bbin.bbclass <-- well it's almost production ready Dec 14 17:15:30 i just added meta-selinux to my morty build, but morty branch looks pretty stale Dec 14 17:16:15 since there are no branches newer than rocko, should i be using master with newer build config? Dec 14 18:11:23 rburton: good start. i'd suggest processing after patch rather than unpac, though, in case there's ever a need to patch a bbin before processing it. and consider a parse time read of the file if and only if it's local, to add the vardeps, but that's optional.. Dec 14 18:15:12 sure are a lot of meta-selinux forks on github... Dec 14 18:18:12 rburton: i also see no reason we couldn't fetch and then process a bbin, so you don't really need to limit to file, other than for vardep determination Dec 14 18:27:01 hmm Dec 14 19:30:06 kergoth: its the vardeps bit thats tricky Dec 14 19:33:16 kergoth: i guess one evil thing would be to use a dummy object that logs what items are fetched using __getitem__ Dec 14 19:34:56 yeah that was my thought too, use your exsitin dataexpander to capture it by actually doing the substitution, basically Dec 14 19:35:21 easy route, anyway. otherwise examine the Template class definition to skip that part Dec 14 22:43:27 khem, maybe I found the reason Dec 14 22:43:35 https://github.com/gcc-mirror/gcc/blob/master/gcc/config/mips/linux.h Dec 14 23:06:22 I remember a patch forcing soft-float...dunno why Dec 14 23:06:34 TUNE_FEATURES = "o32 fpu-hard mips32" Dec 14 23:53:16 ant_home: what is your hardware ? Dec 14 23:53:38 some mips are sf but qemu we use in OE for mips is configured with hf Dec 15 00:24:05 this is a vu+ vuduo2 bcm7425 Dec 15 00:24:58 khem, I meant this old https://patchwork.openembedded.org/patch/97629/ Dec 15 00:33:26 what is funny, default TCLIBC does not expose the issue Dec 15 00:54:59 ant_home: yes that patch should help your case Dec 15 00:55:49 I still dont understand why it shows up with musl alone Dec 15 00:55:59 I wonder if tunes are different ? Dec 15 00:59:01 no no Dec 15 01:04:12 that patch was merged in 2.24 or 2.25 I guess Dec 15 01:04:59 I still can't point to gcc or gas Dec 15 01:10:28 khem, only the kernel, I did build userland with mipsel/musl Dec 15 01:10:53 so maybe there is also a mips-kernel bug in that old 3.x Dec 15 01:11:34 fwiw I compiled kexecboot-klibc-static and it runs Dec 15 01:12:06 now, th ekernel had not cconfigured kexec...see why I must recompile it :) Dec 15 01:30:45 thats a kernel patch Dec 15 01:31:01 which you need **** ENDING LOGGING AT Sat Dec 15 02:59:58 2018