Floating point numbers
Conversions involving floating-point numbers, where moving to or from a real type drops information without warning. The characteristic case is an implicit float-to-integer conversion that truncates the fractional part toward zero — and produces undefined behaviour when the value is too large for the target — turning a deliberate-looking computation into a silent loss of precision.