[jkhav]

Closes #85.

Exit status (F39)

milq now exits non-zero when synthesis fails:

Status	Meaning
0	a circuit was synthesized
1	error: missing input, unparsable QASM, solver failure, Gurobi license limit, ...
2	invalid options (argparse usage message)
3	no circuit found: unsat, or none within --max-depth / -pt

synthesis failed and errors go to stderr. The exit codes are documented in docs/reference.md.

Option defaults and errors (F40)

• -a defaults to the backend's solver: yices2 for -s smt, cbc for -s milp, gurobi for -s gurobi. milq x.qasm -s smt now works. A solver that doesn't fit the backend (e.g. -s smt -a cbc) is a usage error that lists the valid choices.
• ValueError, OSError, RuntimeError, SolverError and GurobiError are caught and printed as milq: error: ... instead of a traceback.
• Numeric options are checked by argparse: -d ≥ 0, --max-depth ≥ 1, -T ≥ 1, -A ≥ 0, -f in [0, 1], -pt > 0. synthesis() also rejects a negative d.
• Fixed the -f help text (it is for -ap, not dreal).

Options outside their mode (F79 and similar)

These options used to be accepted and silently ignored. They are now rejected outside the mode they apply to:

• -d: only basic/binary/topdown/bottomup (the message points to --max-depth)
• --max-depth: only incremental/pareto-incremental (the message points to -d)
• -f: only with -ap
• -pt: only with -m pareto-incremental
• -T, -A: only with -v rus

-d stays separate from --max-depth. In the incremental modes d sizes the encoding (preallocated layers, value bounds, max_k) and is not a limit, so each CLI flag still maps to exactly one synthesis() argument.

cli/simulator.py

• Same treatment: simulator: error: ... instead of tracebacks, -m limited to 0/1, -T/-A checked and only with -v rus, works through a symlink (realpath), corrected help text, unused imports removed.
• Unparsable QASM raised a bare Exception. It now raises ValueError and names the line.
• -v rus with no targets, or with fewer targets + ancillas than the circuit's qubits, crashed with IndexError. It now raises a ValueError that explains why.

Behaviour changes

• Commands that passed -d, --max-depth, -f, -pt, -T or -A where they had no effect now stop with exit 2. Results of valid commands are unchanged.
• A failed synthesis now exits 3 instead of 0.

Tests

• New tests/test_cli.py runs both scripts and checks the exit codes and that no traceback is printed.
• Unit tests for the sim.py errors and for a negative d.
• Full suite: 501 passed.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!