Experiments with MIPLIB 2017 and address sanitizer

[h-g-s]

Hi,

Before doing more extensive benchmarking experiments, I executed CBC compiled with debug info and address sanitizer flags to check if there was some clear leak / memory related problem in master in any of the MIPLIB 2017 instances. In the following excel file there is summary of these results.

CBC experiments master December 2025.xlsx

In the Errors sheet of the attached spreadsheet all error messages are detailed. Some are assertion failures and some are related to address sanitizer (please note that you have to click twice on the cell for the error to appear fully). One of the errors, for example, is:

================================================================
==987354==error: leaksanitizer: detected memory leaks
direct leak of 27776 byte(s) in 1 object(s) allocated from:
    #0 0x7f7017ef58d8 in operator new[](unsigned long) ../../../../src/libsanitizer/asan/asan_new_delete.cpp:98
    #1 0x7f701720cb0a in idiot::crossover(int) /home/haroldo/dev/cbc/clp/src/idiot.cpp:1384
    #2 0x7f70172139f7 in idiot::crash(int, coinmessagehandler*, coinmessages const*, bool) /home/haroldo/dev/cbc/clp/src/idiot.cpp:392
    #3 0x7f70171f3f36 in clpsimplex::initialsolve(clpsolve&) /home/haroldo/dev/cbc/clp/src/clpsolve.cpp:2304
    #4 0x7f701775123b in osiclpsolverinterface::initialsolve() /home/haroldo/dev/cbc/clp/src/osiclp/osiclpsolverinterface.cpp:828
    #5 0x7f70179cfb3a in cbcmodel::initialsolve() /home/haroldo/dev/cbc/cbc/src/cbcmodel.cpp:6474
    #6 0x7f7017ae2f23 in cbcmain1(std::deque<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >, std::allocator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > > >, cbcmodel&, cbcparameters&, INT(*)(cbcmodel*, int), ampl_info*) /home/haroldo/dev/cbc/cbc/src/cbcsolver.cpp:3891
    #7 0x55844f6daca3 in main /home/haroldo/dev/cbc/cbc/src/coinsolve.cpp:285
    #8 0x7f7015833ca7 in __libc_start_call_main ../sysdeps/nptl/libc_start_call_main.h:58
direct leak of 27776 byte(s) in 1 object(s) allocated from:
    #0 0x7f7017ef58d8 in operator new[](unsigned long) ../../../../src/libsanitizer/asan/asan_new_delete.cpp:98
    #1 0x7f701720cafb in idiot::crossover(int) /home/haroldo/dev/cbc/clp/src/idiot.cpp:1383
    #2 0x7f70172139f7 in idiot::crash(int, coinmessagehandler*, coinmessages const*, bool) /home/haroldo/dev/cbc/clp/src/idiot.cpp:392
    #3 0x7f70171f3f36 in clpsimplex::initialsolve(clpsolve&) /home/haroldo/dev/cbc/clp/src/clpsolve.cpp:2304
    #4 0x7f701775123b in osiclpsolverinterface::initialsolve() /home/haroldo/dev/cbc/clp/src/osiclp/osiclpsolverinterface.cpp:828
    #5 0x7f70179cfb3a in cbcmodel::initialsolve() /home/haroldo/dev/cbc/cbc/src/cbcmodel.cpp:6474
    #6 0x7f7017ae2f23 in cbcmain1(std::deque<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >, std::allocator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > > >, cbcmodel&, cbcparameters&, INT(*)(cbcmodel*, int), ampl_info*) /home/haroldo/dev/cbc/cbc/src/cbcsolver.cpp:3891
    #7 0x55844f6daca3 in main /home/haroldo/dev/cbc/cbc/src/coinsolve.cpp:285
    #8 0x7f7015833ca7 in __libc_start_call_main ../sysdeps/nptl/libc_start_call_main.h:58
summary: addresssanitizer: 55552 byte(s) leaked in 2 allocation(s).

In the Timeouts sheet there is a list of instances which exceeded the time limit in 1000 seconds without stopping. This is probably due to a hard to solve LP relaxation as I don't think CBC checks for time limit at this phase.

[tkralphs]

Thanks for doing this! A couple of questions.

• There were some instances that are listed as having error, but there is no log. Do you know what happened there?
• For the instances that timed out, I guess that you mean they did not exit on time limit as they were supposed to?
• Some of the instances with timeouts are listed on the results page and some are not. What happened with the ones that are not? There was just no output? The two that are included in results are
  • chromaticindex512-7
  • cryptanalysiskb128n5obj16

[h-g-s]

Attaching the non-empty error files below so that it easier to check.
errors.zip

W.r.t. to the two instances

• chromaticindex512-7 did succeeded, timeout happened on chromaticindex1024-7
• cryptanalysiskb128n5obj16 finished gracefully also, but cryptanalysiskb128n5obj14 exceeded the time limit without exiting cbc

[jjhforrest]

Fixed one memory leak. Could not reproduce error in Cgl012Cut and do not know enough about that code.

Fixed (I think) two asserts - could not reproduce other errors.