Invariant Synthesis for Programs Manipulating Lists with Unbounded Data

Full version of the paper with the same title accepted at CAV'10.We address the issue of automatic invariant synthesis for sequential programs manipulating singly-linked lists carrying data over infinite data domains. We define for that a framework based on abstract interpretation which combines a specific finite-range abstraction on the shape of the heap with an abstract domain on sequences of data, considered as a parameter of the approach. We instantiate our framework by introducing different abstractions on data sequences allowing to reason about various aspects such as their sizes, the sums or the multisets of their elements, or relations on their data at different (linearly ordered or successive) positions. To express the latter relations we define a new domain whose elements correspond to an expressive class of first order universally quantified formulas. We have implemented our techniques in an efficient prototype tool and we have shown that our approach is powerful enough to generate non-trivial invariants for a significant class of programs