Found while benchmarking BigInt division (#9141); reproduces on pristine main 653e886699, unrelated to that change.
Repro
The division identity (a / d) * d + (a % d) === a evaluates to (a / d) * d — the remainder term vanishes — when a and d are bindings from a for...of over a BigInt array:
for (const a of [123456789012345678901234567890n]) {
for (const d of [1000000007n]) {
console.log((a / d) * d + (a % d)); // node: a perry: (a/d)*d
}
}Correct when the same operands are compile-time constants or plain locals — so this is about how for...of bindings type/lower in a compound BigInt expression, not about the division itself (division and remainder each produce correct values individually).
Repro files: secret-tests/scratchpad/probe_q5.ts, probe_q6.ts.
Why it matters
Silent wrong arithmetic in the shape most real code uses to iterate big-number collections, and the dropped term makes results look plausible rather than obviously broken.
Found while benchmarking BigInt division (#9141); reproduces on pristine main
653e886699, unrelated to that change.Repro
The division identity
(a / d) * d + (a % d) === aevaluates to(a / d) * d— the remainder term vanishes — whenaanddare bindings from afor...ofover a BigInt array:Correct when the same operands are compile-time constants or plain locals — so this is about how
for...ofbindings type/lower in a compound BigInt expression, not about the division itself (division and remainder each produce correct values individually).Repro files:
secret-tests/scratchpad/probe_q5.ts,probe_q6.ts.Why it matters
Silent wrong arithmetic in the shape most real code uses to iterate big-number collections, and the dropped term makes results look plausible rather than obviously broken.