# Adversary REVIEW — phase eval

## Gates

- D1: PASS @2026-06-15T02:05Z
- D2: PASS @2026-06-15T02:05Z
- D3: PASS @2026-06-15T02:05Z
- D4: PASS @2026-06-15T02:05Z
- D5: PASS @2026-06-15T02:05Z

## Evidence

### D5 — full suite (49 tests, 0 failures)
```
Ran 49 tests in 0.218s
OK
```

### D1 — arithmetic / precedence / parens / unary minus
```
python calc.py "2+3*4"   → 14   ✓
python calc.py "(2+3)*4" → 20   ✓
python calc.py "8-3-2"   → 3    ✓
python calc.py "-2+5"    → 3    ✓
python calc.py "2*-3"    → -6   ✓
```

### D2 — true division + EvalError on zero
```
python calc.py "7/2"   → 3.5              (exit 0)  ✓
python calc.py "1/0"   → Error: Division by zero (stderr, exit 1, no traceback)  ✓
python calc.py "0/0"   → Error: Division by zero (stderr, exit 1)  ✓
```
No bare ZeroDivisionError escapes; EvalError wraps it correctly.

### D3 — result type
```
python calc.py "4/2"   → 2    (int, no trailing .0)  ✓
python calc.py "7/2"   → 3.5  (float)                ✓
python calc.py "100/10"→ 10   (whole-valued)          ✓
python calc.py "1/3"   → 0.3333333333333333           ✓
```
`_coerce()` correctly converts whole-valued floats to int.

### D4 — CLI
```
python calc.py "2+3*4"   → stdout: 14, exit 0               ✓
python calc.py "1 +"     → stderr: Error: Unexpected token EOF(None), exit 1, no traceback  ✓
```

### Adversarial extras (all correct)
- `(-2)*(-3)` → 6 ✓
- `100/10` → 10 (no .0) ✓
- `1/3` → 0.333... ✓
- `0/0` → EvalError, exit 1 ✓

No regressions in lex or parse phases (49 tests = prior 39 + 10 new eval tests).