Key points are not available for this paper at this time.
Dynamic logic and its variations, because of their good expressive forms capturing program specifications clearly by isolating programs from logical formulas, have been used as a formalism in program reasoning for decades and have many applications in different areas. The program models of traditional dynamic logics are in explicit forms. With a clearly-defined syntactic structure, compositional verification is made possible, in which a deduction step transfers proving a program into proving its sub-programs. This structure-based reasoning forms the basis of many dynamic logics and popular Hoare-style logics. However, structural rules induce a major drawback that for different target programs, different rules have to be proposed to adapt different program structures. Moreover, there exist programs that does not support (or not entirely support) a structure-based reasoning. In this paper, we propose a parameterized `dynamic-logic-like' logic called DLp with general forms of program models and formulas, and propose a cyclic proof system for this logic. Program reasoning in DLp is directly based on symbolic executions of programs according to their operational semantics. This reduces the burden of designing a large set of rules when specializing a logic theory to a specific domain, and facilitates verifying programs without a suitable structure for direct reasoning. Without reasoning by dissolving program structures, DLp can cause an infinite proof structure. To solve this, we build a cyclic preproof structure for the proof system of DLp and prove its soundness. Case studies are analyzed to show how DLp works for reasoning about different types of programs.
Yuanrui Zhang (Sun,) studied this question.