Back to proposals-2013
Title: The true cost of software bugs and what to do about it
Proposer: greg_law
Proposer: julian_smith
Type: Tutorial (with case-study elements)
Duration: 90 minutes
Description:
Bugs have been estimated to cost the global economy $600bn annually, making debugging an endeavour of similar scale and impact to solving the Euro crisis. Yet startlingly little attention is paid to the problem by wider society or by the industry itself. In this talk we present the results of our own work with Cambridge University to better estimate the economic costs of software bugs, and go on to examine the state of the art of techniques and technologies to address the burden.
We split the problem of debugging into three categories, and discuss tools and techniques to address each:
- Preventing bugs in the first place: programming languages and techniques to reduce the number of bugs created.
- Finding the bugs lurking in your software: static and dynamic analysis tools and testing techniques to uncover the bugs before your customers see them.
- Panic debugging: tools and techniques to help find and fix the bugs found during development, testing, or (worst of all) reported by end users.
We show some of the more interesting work to address each of the above, including an overview of free (as in speech) software tools as well as proprietary ones. The talk focuses on practical use of tools and advice, and explicitly does not cover the more social aspects (methodologies etc). In particular we look at use of languages and libraries in order to prevent bugs, preventative tools such as Clokwork, CoVerity Prevent and Valgrind, advanced debuggers such as UndoDB and the tried and trusted, and much maligned, printf.
We also examine the economic and psychological barriers to preventing broader adoption of the tools and techniques we cover, and present the results of recent research quantifying the benefits that can be obtained by using more advanced tools and techniques to deal with bugs.
Roger: Yes: much better than Julian's other proposal
Ewan: (very) Probably YES. I agree, clearly this is the one to pick from Julian.