Enhancing Infer Compositional Analysis with Summary Specialization
summary
Infer adds summary specialization to its compositional, on‑demand bottom‑up analyzer: compute per‑function summaries under lightweight assumptions (specialization keys, typically runtime types), use separation logic for summaries, and re‑analyze a function for different keys when needed. Deployed on Hack and Python code, the prototype analyzes ~7,000 files, resolves 90% of closure call sites (removing specialization drops resolution to ~50%), and incurs ~40% extra analysis time from reanalyzing functions; Java/Android results are weaker.
AI-written summary · about
chapters (9)
- 0:00Interprocedural analysis approaches
- 2:56Infer static analysis platform
- 4:40Bottom-up summary computation
- 7:11Non-static calls challenges
- 8:15Summary specialization for virtual calls
- 10:02Summary specialization example
- 12:50Experimental results and applications
- 13:58Conclusion and future work
- 15:04Q&A
related talks
50mState of the Infer Platform
25mStatic Resource Analysis at Scale
01h 06mOn-The-Fly Verification via Incremental, Interactive Abstract Interpretation wi...
51mAI for the People
45mInfer Internals for Integrations and Infrastructure
56mMy Journey to the Dark Side: Under-Approximation, Incorrectness, Proof
57mBuilding code analysis tools at Twitter
50mProgram Analysis