malloc(): Not enough space in some MiniZinc subsolvers

solver: MiniZinc’s int, cp, lcg and chuffed subsolvers CPMpy: version v0.9.9 commit e79b3afedc934a9437c2ddb3a9f54d7e2d7bd3ee minizinc-python release version 0.7.0 commit {https://github.com/MiniZinc/minizinc-python/commit/a195cf63fcfbc98665d70ab64efb5424db25bd7e)

While using CPMpy to model a problem and some Minzinc’s subsolvers to solve the problem, I found a malloc() error. This error is probably for https://github.com/MiniZinc/minizinc-python, but after advise from Ignace I’m posting it here just to be sure. The amount of constraints are minimized, but the remaining constraints can probably still smaller.

almostMinimimal.zip

from cpmpy import *

file = "almostMinimimal" # don't forget to unzip first
m = Model().from_file(file)
m.solve(solver="minizinc:lcg") # or int, cp, lcg, chuffed

bug found while working on my master thesis.

[tias]

No, OK, then it is on our side, e.g.

x1 = intvar(0,5, name="x1")
c = cpm_array([1,2,5,7,13,21])[x1]
cc = flatten_constraint(abs(c) > 5)
print_variables(cc)

should have IV1: 1…21 but it is -int…int and that is what you carry along in your mutations. That is the bug we should fix

[rubenkindt]

The file above got created by taking a random (sub)solver and a random existing (solvable) problem executing 4 of the 40 created sat-preserving transformations on some (or all) the constraints,then test if the problem is still solve-able and extracting it if it failed via to_file. I did find some places where I take a random argument’s “.lb” without first checking if that argument had a “.lb”. So it could be that I set those values to gibberish assuming Python allowed it (normally Python would crash I think). Tomorrow I will search further to see if this is what happened.