double a = computeA(); // returns double float b = computeB(); // returns float double mix = a + b; // implicit cast, different rounding paths possible return finalize(mix); After:
double normalize(double x) { // explicit, documented rounding to the desired precision return explicitRound(x); } fpre005 patched
If you want, I can expand this into a longer post with code snippets in your project's language, a timeline of discovery, or a short slide deck for engineering reviews. Which would you prefer? double a = computeA(); // returns double float
double a = computeA(); float b = computeB(); double mix = normalize(a) + normalize(b); return finalize(normalize(mix)); fpre005 patched is a reminder that in numeric code, “small” differences matter. Deterministic rounding and a single source of truth for conversions prevent elusive bugs that evade common testing strategies. This patch is a tidy, low-risk change that improves correctness, reproducibility, and developer clarity — a good example of the principle that robustness often comes from enforcing simple, consistent invariants. Deterministic rounding and a single source of truth