segfault in MC

[loudesplat]

with 4at/uc, depending on version, will segfault randomly after doing a few temperatures, or right after setting hamiltonian.

# 2Fe/Mo(110)
simulation metropolis 

magnetic_J
1 1 1  41.6189d-3     # between atoms 1 and 1, shell 1, amplitude in eV  
1 1 2  14.4771d-3     # between atoms 1 and 1, shell 2, amplitude in eV,etc
1 1 3  8.77002d-3
1 1 4  6.20242d-3

magnetic_anisotropy
1  0.0 0.0 1.0 -0.4916d-3 

magnetic_D
1 1 2 1 0.72247d-3

gra_topo .F.
gra_fft .F.
gra_log .T.
gra_Beff .F.
gra_freq 10000
Efreq 10000
CalTheta .F.
Energy_Distrib .F.
energy_detail .F. 
print_Econt .T.
calc_topo .F.

############################################
# rectangular setup unit cell spindynamics #
############################################

# Lattice
Periodic_log .T. .T. .F.
alat 0.44629703113509506 0.3155796571390494 0.354
Nsize 20 20 1
lattice
1.0 0.0 0.0     # a_11 a_12 a_1
0.0 1.0 0.0      # a_21 a_22 a_23
0.0 0.0 1.0

# Atoms
atomtypes  2
Fe 2.5517 0.0 0.0 F 0 # atom type: (name, mag. moment, mass, charge, displacement, number TB-orb.)
Ir 0.0 0.0 0.0 F 0
atoms 5
Fe 0.0 0.0 0.0
Fe 0.5 0.5 0.0
Fe 0.5 0.0 0.5
Fe 0.0 0.5 0.5
Ir 0.0 0.5 -0.7

#minimize
min_Efreq 100
convergence_criteria 1.0d-8
#------------------------------#
#              MC              #
#------------------------------#

Tini 31.0d0 !loop on T: start with Tini, finish with Tfin and do n_Tsteps
Tfin 60.0d0
n_Tsteps 10 !number of MC simulations at different T, uses Tini and Tfin
Total_MC_Steps 2000 
T_relax 20 !number of thermalization steps (will be T_relax*N)
T_auto 5 !measure steps occur every T_auto autocorrelation steps
tracker .F. !tracks skyrmions
ising .F. !Ising model: only up and down orientations
do_fluct .T. !for fluctuations
fluct_direction 0.0 1.0 0.0

cone 3.1415926535897931     
restart .F.
underrelaxation .F.
overrelaxation .F.
sphere_samp .T.
equi_samp  .F.
print_relax .F.
Cor_log  .F.

n_sizerelax  1
n_relaxation 1000
#restart_MC_steps  0

[bertdupe]

code works fine. The error can not be reproduce with debugging flags and full optimization. Must be a local computer problem.

[loudesplat]

with version on develop: (this crashes right away and does not run)

[loudesplat]

error with older version #cb5a6a7de9caf4bbef4bd268cfa0725522b9ac45

[bertdupe]

Monte Carlo is working fine in parallel (without the fluctuations) n_relaxation 5000 50 temperatures between 1 and 60 Total_MC_Steps 10000