TY - GEN
T1 - Condition Synthesis Realizability via Constrained Horn Clauses
AU - Rothenberg, Bat Chen
AU - Grumberg, Orna
AU - Vizel, Yakir
AU - Singher, Eytan
N1 - Publisher Copyright: © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.
PY - 2023
Y1 - 2023
N2 - Condition synthesis takes a program in which some of the conditions in conditional branches are missing, and a specification, and automatically infers conditions to fill-in the holes such that the program meets the specification. In this paper, we propose CoSyn, an algorithm for determining the realizability of a condition synthesis problem, with an emphasis on proving unrealizability efficiently. We use the novel concept of a doomed initial state, which is an initial state that can reach an error state along every run of the program. For a doomed initial state σ, there is no way to make the program safe by forcing σ (via conditions) to follow one computation or another. CoSyn checks for the existence of a doomed initial state via a reduction to Constrained Horn Clauses (CHC). We implemented CoSyn in SeaHorn using Spacer as the CHC solver and evaluated it on multiple examples. Our evaluation shows that CoSyn outperforms the state-of-the-art syntax-guided tool Cvc5 in proving both realizability and unrealizability. We also show that joining forces of CoSyn and Cvc5 outperforms Cvc5 alone, allowing to solve more instances, faster.
AB - Condition synthesis takes a program in which some of the conditions in conditional branches are missing, and a specification, and automatically infers conditions to fill-in the holes such that the program meets the specification. In this paper, we propose CoSyn, an algorithm for determining the realizability of a condition synthesis problem, with an emphasis on proving unrealizability efficiently. We use the novel concept of a doomed initial state, which is an initial state that can reach an error state along every run of the program. For a doomed initial state σ, there is no way to make the program safe by forcing σ (via conditions) to follow one computation or another. CoSyn checks for the existence of a doomed initial state via a reduction to Constrained Horn Clauses (CHC). We implemented CoSyn in SeaHorn using Spacer as the CHC solver and evaluated it on multiple examples. Our evaluation shows that CoSyn outperforms the state-of-the-art syntax-guided tool Cvc5 in proving both realizability and unrealizability. We also show that joining forces of CoSyn and Cvc5 outperforms Cvc5 alone, allowing to solve more instances, faster.
UR - https://www.scopus.com/pages/publications/85163987751
U2 - 10.1007/978-3-031-33170-1_23
DO - 10.1007/978-3-031-33170-1_23
M3 - منشور من مؤتمر
SN - 9783031331695
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 380
EP - 396
BT - NASA Formal Methods - 15th International Symposium, NFM 2023, Proceedings
A2 - Rozier, Kristin Yvonne
A2 - Chaudhuri, Swarat
PB - Springer Science and Business Media Deutschland GmbH
T2 - 15th International Symposium on NASA Formal Methods, NFM 2023
Y2 - 16 May 2023 through 18 May 2023
ER -