Please help with segmentation fault issue

[DevShark]

Hello,

I use bayesopt with the python bindings. It works nicely most of the time.
However, I have a bug that comes up.

It computes the first starting_points, and then runs nlopt. Here's the backtrace.

Anyone knows what the issue could be?

I tried looking into the code to determine what's going on, but I'm a bit lost.

Thanks a lot in advance for any pointer!

Program received signal SIGSEGV, Segmentation fault. 0x00007fffcd75c024 in altmov_.isra.0 () from installs/bayesopt/lib/bayesopt.so Missing separate debuginfos, use: debuginfo-install glibc-2.17-78.el7.x86_64 libX11-1.6.0-2.1.el7.x86_64 libXau-1.0.8-2.1.el7.x86_64 libpng-1.5.13-5.el7.x86_64 libxcb-1.9-5.el7.x86_64 qhull-2003.1-20.el7.x86_64 (gdb) traceback Undefined command: "traceback".  Try "help". (gdb) backtrace
#0  0x00007fffcd75c024 in altmov_.isra.0 () from installs/bayesopt/lib/bayesopt.so
#1  0x00007fffcd7606e1 in bobyqa () from installs/bayesopt/lib/bayesopt.so
#2  0x00007fffcd729ace in nlopt_optimize () from installs/bayesopt/lib/bayesopt.so
#3  0x00007fffcd676916 in bayesopt::run_nlopt(nlopt::algorithm, double (*)(unsigned int, double const*, double*, void*), boost::numeric::ublas::vector<double, boost::numeric::ublas::unbounded_array<double, std::allocator<double> &, int, std::vector<double, std::allocator<double> > const&, std::vector<double, std::allocator<double> > const&, void*) () from installs/bayesopt/lib/bayesopt.so
#4  0x00007fffcd678244 in bayesopt::NLOPT_Optimization::run(boost::numeric::ublas::vector<double, boost::numeric::ublas::unbounded_array<double, std::allocator<double> &) () from installs/bayesopt/lib/bayesopt.so
#5  0x00007fffcd64bf35 in bayesopt::ContinuousModel::findOptimal(boost::numeric::ublas::vector<double, boost::numeric::ublas::unbounded_array<double, std::allocator<double> &) () from installs/bayesopt/lib/bayesopt.so
#6  0x00007fffcd6566fc in bayesopt::BayesOptBase::nextPoint() () from installs/bayesopt/lib/bayesopt.so
#7  0x00007fffcd65a0e4 in bayesopt::BayesOptBase::stepOptimization() () from installs/bayesopt/lib/bayesopt.so
#8  0x00007fffcd65c6cc in bayesopt::BayesOptBase::optimize(boost::numeric::ublas::vector<double, boost::numeric::ublas::unbounded_array<double, std::allocator<double> &) () from installs/bayesopt/lib/bayesopt.so
#9  0x00007fffcd6bcf28 in bayes_optimization () from installs/bayesopt/lib/bayesopt.so
#10 0x00007fffcd722954 in __pyx_pf_8bayesopt_2optimize(_object*, _object*, int, tagPyArrayObject_fields*, tagPyArrayObject_fields*, _object*) [clone .isra.17] () from installs/bayesopt/lib/bayesopt.so
#11 0x00007fffcd723dfa in __pyx_pw_8bayesopt_3optimize(_object*, _object*, _object*) () from installs/bayesopt/lib/bayesopt.so

[Ruben Martinez-Cantin]

Hi,

NLOPT is an external library. Maybe you should contact them
http://ab-initio.mit.edu/wiki/index.php/NLopt

But that problem have not happened before within BayesOpt. Which
parameters are you using? What are the inputs of your function? Which
bounds?

Best,

Ruben

> --
> You received this message because you are subscribed to the Google Groups
> "BayesOpt discussion" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to bayesopt-discus...@googlegroups.com.
> To post to this group, send email to bayesopt-...@googlegroups.com.
> Visit this group at https://groups.google.com/group/bayesopt-discussion.

[Sebastien Mouton]

Hi Ruben,

Apologies for the very slow response.

Thanks for your email back. Unfortunately, the function is very complex (it's the result of a full simulation, ran by a complex framework), so too complicated to share easily.

I'll follow up with the nlopt developers.

Thanks!

[Ruben Martinez-Cantin]

Hi Sebastian,

Please, replace src/inneroptimization.cpp with the attached file and see if the problem persists. Also, if the problem persists, please set verbose_level = 5 and send me the log file.

Best,

Ruben

[Sebastien Mouton]

Hi Ruben,

FYI, to solve this problem I modified this file as well. I changed nlopt.BOBYQA to nlopt.COBYLA. It was then working. So this points indeed to an nlopt problem.

Thanks,

Sebastien

<inneroptimization.cpp>

[DevShark]

To clarify my previous message, that's how I fixed my problem, and I unfortunately did not keep the parameters set to reproduce the problem. Thanks for your help.