Verifying Programs with Arrays and Lists

Automatically verifying safety properties of programs is a tough problem that has been tackled using many different approaches: rewriting systems, abstract interpretation, SMT solving, . . . Most techniques restrict themselves to programs operating on boolean and integer values and transposing them to infinite data structures such as arrays has not yet been satisfyingly achieved. Recent work in Monniaux and Gonnord [2016] suggests the use of abstract interpretation to transpose programs containing arrays into Horn clauses that do not contain arrays. The major innovation of their work is that they use Horn clauses which are more general than programs, to obtain better results. In this work, we first set the work of Monniaux and Gonnord in a more general framework that allows us to extend their abstractions, simplify the expressions they generate, and analyze the precision of their abstraction. Finally we extend their abstractions so that we can the analyze lists and experiments show that we succeed to analyze several classical examples, including sorting algorithms.