You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Research: ProveNFix, CTL-guided repair and OO incorrectness donors (bounded) #462
#272 owns finite barrier obligations; #446 investigates unbounded temporal inclusion; #448 owns requirements→code conformance; #460 owns Frame. Here: only missing correctness and repair verification capabilities supported by actual code evidence.
Gate 0: pinned read-only gap analysis
Check commits, licenses, supported languages, reproducible paper commands, solver assumptions and whether any component even supports C#. The published tool results on C do not transfer to C#.
Compare existing OwnIR CFG/dataflow + mutation + explicit finite FSM baseline on one concrete finite protocol bug, including correct negative control.
Look for distinct OO runtime-dispatch/alias/incorrectness case with reliable reachable-bug witness; handle virtual/override/cast/reflection uncertainty honestly.
For a candidate repair check patch quality, behavior preservation, independent regression, source anchor and reproducible red→green witness. Passing tests alone do not prove safety.
CTL branch-path vs LTL single-path vs simple safety: require specific extra property and measured value before Datalog/CTL experiments; never integrate just because algorithm sounds impressive.
Deliverable
Matrix of baseline vs donor (property, semantics, assumptions, source witness, unknowns, repair validity, cost); verdict KEEP-REFERENCE / TINY-SPIKE-PROPOSAL / KILL. No production code, solver, automatic patching or CI until separate explicit GO.
Activity