May 7-10, 2017 Asilomar, California
Automated program repair techniques typically aim to compute a syntactically
minimal or small fix to an incorrect program. However, programs that are
synactically close may differ significantly in their behaviors. We propose the
use of program distances that can quantify changes not only to the program
syntax but also to the program semantics to drive program repair. I will
present a solution strategy and tool - Qlose (Quantitatively close) - to this
"quantitative program repair" problem.
I will also describe our current efforts to extend Qlose and discuss how
program distances can be used to drive applications beyond repair.