Originally posted by: T_O

I am trying to solve a strange problem: A program that I just wrote sometimes crashes. Here is the backtrace:

[New Thread 0x7fffec549700 (LWP 23320)]
[New Thread 0x7fffebd48700 (LWP 23321)]
[New Thread 0x7fffeb547700 (LWP 23322)]

Program received signal SIGSEGV, Segmentation fault.
__memcpy_sse2 () at ../sysdeps/x86_64/multiarch/../memcpy.S:273
273     ../sysdeps/x86_64/multiarch/../memcpy.S: No such file or directory.
(gdb) where
#0  __memcpy_sse2 () at ../sysdeps/x86_64/multiarch/../memcpy.S:273
#1  0x00000000005ff7d6 in _intel_fast_memcpy.A ()
#2  0x00000000004bf741 in _7990eedf5561d55902eafb45196ffa3a ()
#3  0x00000000004bcf35 in _661e48ed2bc34363b45ff2f59b03c257 ()
#4  0x000000000055d68c in _5802e8c70437348dc0dbfb43e4ff1349 ()
#5  0x00000000004d7329 in _33647440a26d640acca1886f9152d44a ()
#6  0x00000000004f35a0 in _100e05e6bb6c894bc6fa2bddd586c4b3 ()
#7  0x000000000051dcce in _6b5a3526294c9e72ffb429fc6db9deff ()
#8  0x000000000051c22b in _4485b3e1cd9144232540150354ec3820 ()
#9  0x00000000008d6fd4 in _8727296eaaa73edeafa0b13f8264cf6b ()
#10 0x000000000083cfad in _5ce57952ba3c58d45c4ff1caf38ccdb0 ()
#11 0x000000000083c79e in _2aa011ff11483daf73a86fadae02922d ()
#12 0x0000000000866b4e in _c69cadd8f44da6e42b444e6f0807896a ()
#13 0x000000000086a671 in _6874c3b6b6be3f68ecba8390ec5eef57 ()
#14 0x00000000008695b8 in _0ff8510b91f0240cafe918f6ab601dab ()
#15 0x0000000000866c6b in _cac20a5c347b91b75f3929b6cf82c567 ()
#16 0x000000000083cf18 in _5ce57952ba3c58d45c4ff1caf38ccdb0 ()
#17 0x00000000004817f9 in CPXLmipopt ()
#18 0x000000000044ba5c in CPXXmipopt (env=0x18feeb0, lp=0x1908420)
    at /opt/cplex/cplex/include/ilcplex/cplexx.h:3035
#19 0x000000000044ee1e in main (argc=2, argv=0x7fffffffe0f8) at ../main.cpp:472
(gdb)

Could someone tell me what is segfaulting here? This is CPLEX 12.5.0.0 on 64-bit Linux
It does not happen if I write a sav-file and read it back before calling CPXXmipopt.

Best regards,
Thomas
 

Originally posted by: VivekPeriaraj

I don't know but sometimes with g++/gcc on linux, I have noticed that the memory corruption happen much before the actual CPLEX segfaults. That could be the case here too.

Regards,
Vivek.

Originally posted by: T_O

Interestingly, the code works fine very often on Linux and always on Windows (with CPLEX 12.5.0.1). So, it is very difficult to find the memory corruption.

My above post was mainly addressed to Daniel or Tobias. Maybe they see what is going wrong.

Best regards,
Thomas

Originally posted by: VivekPeriaraj

I agree. You could try Intel's C++ compiler :)

OK, I can decrypt the obfuscated functions names but the backtrace does not ring a bell :-( Could you try to additional things:

1. Run with valgrind to see if that segfault is the actual problem or may be just a deferred issue.

2. Enable the CPX_PARAM_DATACHECK parameter to make sure no invalid input is submitted to CPLEX.

Originally posted by: T_O

Thanks, Daniel!

CPX_PARAM_DATACHECK is on. Running valgrind is a little difficult in this case, but at least, I compiled my C++-vectors as debug vectors.

I'll do some more tests during the next days. Maybe the problem vanishes. I'll contact you again if other problems appear.

Best regards,
Thomas