Multiply/Divide wrap the exponent on int overflow: 1e2000000000 squared returns 1e-294967296 instead of throwing

[matt-edmondson]

What's wrong

Multiply and Divide compute the result exponent with unchecked int arithmetic:

• Multiply (PreciseNumber/PreciseNumber.cs ~L1429): new PreciseNumber(left.Exponent + right.Exponent, ...)
• Divide (~L1487): int exponent = left.Exponent - right.Exponent;
• The same pattern appears in TryDivideExactly and DivideToPrecision (~L1546, L1565, L1586)

Other parts of the library already guard against this overflow. The constructor uses checked(exponent + trailingZeros), Parse uses checked(exponent - decimalDigits), and ToString widens to long. Pow also documents OverflowException "when the result needs an exponent outside the range of an int". But integer Pow goes through Multiply, so it wraps too, and that documented exception never happens.

Failure scenario (reproduced)

With big = Parse("1e2000000000") and small = Parse("1e-2000000000"):

Expression	Result
big * big	1e-294967296
small * small	1e+294967296
big / small	1e-294967296
(big * big) > big	false
PreciseNumber.Pow(big, 2)	1e-294967296 (the documented OverflowException is not thrown)

The result is a silently wrong value, with its magnitude off by billions of orders. An arbitrary-precision type exists to avoid exactly this kind of error.

Suggested fix

Compute the exponent in checked(...), or in long with an explicit range check. Either way, throw OverflowException, the same way the constructor and Parse already do.

Acceptance criteria

• big * big, small * small, big / small, and Pow(big, 2) all throw OverflowException.
• Tests cover Multiply, Divide, and the precision-division paths at the int exponent boundary.

[matt-edmondson]

Triage

• Category: Bug
• Priority: Medium. The result is silently wrong by billions of orders of magnitude, and Pow's documented OverflowException can never be thrown. It is only reachable with exponents near the int limits, though, so few real callers will hit it.
• Area / suggested assignee: exponent arithmetic in Multiply, Divide, TryDivideExactly and DivideToPrecision, @matt-edmondson. The fix is mechanical: wrap it in checked(...), matching the constructor and Parse.
• Duplicates: None. Related to Pow(0, y) returns 0 for negative y, but 1/0 and RootN(0, -n) throw DivideByZeroException #110, which is another Pow edge case where the expected exception isn't thrown. Both could go in one PR.
• In progress: No open PR covers this.

---

Generated by Claude Code