Kano,
There might be version mismatch between core and executable
I will post correct info when it crashes again
I am not quite sure if that info is useful at all
Flowing your advise i got some gdb core info to share. cgminer 2.10.2 coredumps
[91932.126721] cgminer[2138]: segfault at 260 ip 000000000040be10 sp 00007fff4c6b48b0 error 4 in cgminer[400000+57000]
[New LWP 2680]
[New LWP 2731]
[New LWP 2618]
warning: Error reading shared library list entry at 0x780000003c
warning: Corrupted shared library list: 0x0 != 0x4830408b48d00148
Core was generated by `/usr/local/bin/cgminer xxxxxx.
Program terminated with signal 11, Segmentation fault.
#0 0x0000000000410ea1 in stage_thread (userdata=0x1c4a080) at cgminer.c:3576
3576 cgminer.c: No such file or directory.
line 3576 tq_freeze(mythr->q);
10X
At a guess, a corrupted shared library list is corrupted RAM.
So either a runaway pointer, or you need to run a memory check.
Although it's not impossible, it doesn't seem likely that the actual transfer address to the library function was corrupted.
Also note that when you run memtest86+ you should actually let it complete the full test (which takes quite a while) even it the first 4 tests return OK.
I've had one case in the past of RAM that only failed much later, I think it was test #7, but succeeded on all the tests before it.
If it's not RAM then it's gonna need a lot of info to work out what corrupted the library transfer address.