`release/6.3.0` branch: Minizinc 2.8.5 test `tests/spec/unit/test-globals-float.mzn` causes OOM

[LebedevRI]

I've built gecode from release/6.3.0 branch (specifically, f7f0d7c),
plus libminizinc 2.8.5 release, using clang-18 -O2 -UNDEBUG.

At least one minizinc test has issues, the last output is: (only one test is run at once)

spec/unit/regression/test_parout.mzn::default.1.cbc
[gw0] [ 64%] SKIPPED spec/unit/regression/test_parout.mzn::default.1.cbc
spec/unit/regression/var_self_assign_bug.mzn::default.0.gecode
[gw0] [ 64%] PASSED spec/unit/regression/var_self_assign_bug.mzn::default.0.gecode
spec/unit/regression/var_self_assign_bug.mzn::default.0.chuffed
[gw0] [ 64%] PASSED spec/unit/regression/var_self_assign_bug.mzn::default.0.chuffed
spec/unit/regression/var_self_assign_bug.mzn::default.0.cbc
[gw0] [ 64%] SKIPPED spec/unit/regression/var_self_assign_bug.mzn::default.0.cbc
<this is the last output. after a while, OOM happens>

While waiting for OOM killer to trigger (after fzn-gecode consumes 100+Gigs of RAM!),
this is the relevant output of ps:

 145434 ?        S      0:00 /usr/bin/make -f debian/rules test_minizinc_late
 145451 ?        S      0:00 /bin/sh -c cd /build/minizinc-meta-27hHCQ/minizinc-meta-1.0.0/libminizinc/tests/; \ PATH="/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/build/minizinc-meta-27hHCQ/minizinc-meta-1.0.0-build/build/stage-2/minizinc-meta-install/usr/bin" PYTHONPATH=":/build/minizinc-meta-27hHCQ/minizinc-meta-1.0.0-build/build/stage-2/minizinc-meta-install/usr/lib/python3.12/dist-packages/" pytest --suite default --solvers gecode,chuffed -n 1
 145452 ?        Sl     0:00 /usr/bin/python3 /usr/bin/pytest --suite default --solvers gecode,chuffed -n 1
 145454 ?        Sl     0:02 /usr/bin/python3 -u -c import sys;exec(eval(sys.stdin.readline()))
 147367 pts/3    Ss     0:00 /bin/bash
 148088 ?        I      0:00 [kworker/26:0]
 148360 ?        S      0:00 /build/minizinc-meta-27hHCQ/minizinc-meta-1.0.0-build/build/stage-2/minizinc-meta-install/usr/bin/minizinc --solver org.gecode.gecode@6.3.0 --allow-multiple-assignments --output-mode json --output-time --output-objective --output-output-item --statistics --intermediate-solutions /build/minizinc-meta-27hHCQ/minizinc-meta-1.0.0/libminizinc/tests/spec/unit/test-globals-float.mzn --json-stream
 148362 ?        R      0:55 /build/minizinc-meta-27hHCQ/minizinc-meta-1.0.0-build/build/stage-2/minizinc-meta-install/usr/bin/fzn-gecode -s /tmp/mznfileyzWV5u.fzn
 148412 pts/3    R+     0:00 ps ax
 148413 pts/3    D+     0:00 /bin/bash

So it is indeed tests/spec/unit/test-globals-float.mzn.

This raises a bigger issue: looking at fzn-gecode, is there really no memory limit handling?

X-Ref: MiniZinc/libminizinc#835

[zayenz]

Verified against main using the model from MiniZinc 2.8.5, compiled with MiniZinc 2.9.7. fzn-gecode completes in about 2.2 seconds with 1,076 nodes, depth 1,075, and roughly 5.4 MB maximum RSS. The reported memory growth no longer reproduces on the branch used for the next release, so I am closing this as fixed or obsolete. A resident-memory limit would be a separate feature request.