Most defect data is usually found in the form of defects per 1000 lines of code, where such lines can be executable, source or otherwise. For proprietary systems such as Windows and embedded control systems, these are nearly always quoted in bytes of object code so it is very difficult to relate them to conventional measures of quality.
This paper describes a predictive system for doing this, calibrated on real systems allowing easy movement between source, pre-processed and executable lines of code and bytes of object code for systems using the programming language C.
LOC2005.pdf (244KB)
