reverse_qubit_order() KeyError on aliased operand nodes (same root cause as #331) #333

Summary

QasmModule.reverse_qubit_order() raises KeyError on any program whose unrolled AST contains aliased operand nodes — the same root cause as #331, surfacing one method below the fix in #332.

Found during adversarial review of #332: the visited-set guard added there fixes _remap_qubits, but reverse_qubit_order (src/pyqasm/modules/base.py:533-555) walks the same statements and mutates bit.indices[0][0].value in place with a negative-marker trick. On a node shared by k statements, the second visit reads the already-negated value and looks up new_qubit_mappings[reg][-2] → KeyError: -1.

Reproduction

from pyqasm import loads

m = loads('''OPENQASM 3;
include "stdgates.inc";
qubit[3] q;
crz(0.5) q[1], q[2];
''')
m.unroll()
m.reverse_qubit_order()   # KeyError: -1

Reproduces on main (pre-dates #332; that PR did not touch this method).

Fix options

Same two options as #331, not mutually exclusive:

1. Root cause: have unroll() / the decomposer emit fresh (deep-copied) operand nodes so no AST node is shared across statements. This retires the whole class of bug — every in-place index mutation in modules/base.py is currently exposed to aliasing.
2. Point fix: apply the same visited-node-id guard from fix: remove_idle_qubits KeyError on shared operand nodes #332 to both passes in reverse_qubit_order. Note the marker pass and the unmark pass each need their own guard (or the method should be rewritten to build the mapping without the negative-marker trick, which only exists to survive double visits it doesn't actually survive).

Option 1 is the better long-term fix; if we take it, the guard added in #332 becomes harmless belt-and-braces.

🤖 Filed with Claude Code