Tracing a False Positive in FBInfer’s Purity Analyzer

The client sought a technical deep dive into a false positive bug triggered by Facebook’s static analysis tool, FBInfer —specifically its Purity Analyzer module. The goal was to unpack the behavior behind a misclassification and provide a rigorous write-up that could be used to guide future development or onboarding efforts.

This required understanding and explaining complex interactions between two memory modeling systems used by FBInfer: InferBO and Pulse, across a large, sparsely documented OCaml codebase.

I was brought on as a solo contributor to analyze the problem from the ground up, trace the underlying logic across multiple components, and synthesize findings into a technical document.

Investigating the source of the false positive by analyzing logs, memory models, and internal assumptions

Mapping the flow of memory modeling and purity checks across InferBO and Pulse

Identifying and documenting the breaking point where the misclassification occurs

Writing a structured, developer-facing report with annotated findings, error reproduction details, and commentary on model differences

Revising the document based on client feedback to emphasize onboarding and storytelling value

Producing a screencast to walk through the structure and logic for easier comprehension

FBInfer is a large OCaml-based tool with minimal inline documentation and niche internal terminology. I worked without prior familiarity with OCaml, quickly ramping up to follow data flow across modules and interpret log behavior line-by-line. This required reverse-engineering the system’s intent based solely on output patterns and internal naming conventions.

The final deliverables included:

A multi-section technical report (Notion + Word)

Inline code references and analysis of key logic

A follow-up screencast video walking through the document and reasoning

A client-tailored rewrite focused on onboarding readability and logical flow

Despite the project’s technical depth and time constraints, I delivered a comprehensive analysis that aligned with the client’s goals. The final materials were recognized for their clarity and depth, and the engagement concluded with positive feedback and further interest in potential collaboration.