Tihana Galinac Grbac

Theory on the distributions and predictive capability of verification faults

Software engineering is a relatively young discipline with just few general theories. General problem is in lack of industrial data, and/or inconsistently reported data and studies. Theory on the predictive capability of verification faults has been grounded by systematic approach for empirical research on fault distributions suggested by Fenton and Ohlsson followed by replications. Fault distributions are interesting because of their seemingly similar behavior across different environments. This talk summarizes our recent findings on fault distributions and the predictive capability of early verification on late detected faults. We will also discuss some novel findings related to the predictive capability and distributions of unit verification faults. The results are of particular importance for large scale complex systems that are developed in evolutionary fashion with majority of reused software.


Zoltan Horvath

More than types!

Types in strongly typed languages can help avoid many programming errors. However, no matter how rigorous your type system is, many low quality or even faulty code fragments may remain unrevealed. In order to find and eliminate them, we need more sophisticated tools than static type checkers. Can these sophisticated, and henceforth rarther complex tools be used in practice? This talk will give an affirmative answer. We illustrate the advantages and the practical applicability of RefactorErl, a static program analysis and transformation tool for Erlang. RefactorErl can help the programmer better understand the static and dynamic structure of the code base (a.k.a. grokking), find structures (e.g. call chains) that violate some existing requirements or standards, point out bad smells, and identify duplicated or overly complex code fragments. The tool was succesfully used in analyzing large industrial telecommunication software.

Go to timetable