static analysis

Bilingual

OpenAI's work on Git for large repositories
2 months ago
- Fixed a reused loop index that caused Coccinelle's static analysis to time out after six hours; isolated loop indexes drop the analysis time.
- Improved bitmap selection to consider reachable history instead of just path length, reducing MIDX bitmap write time from 984.73s to 200.06s.
- Added per-worker counters (item count, bytes, failures, elapsed time, slowest item) to parallel checkout for better diagnosis of long checkouts.
- Fixed a race condition where files changed between bulk content check and filesystem-monitor query; kept scan data available until final confirmation.
- Documented an optional .csts cache file for git status --porcelain=v2 that skips status only when index, HEAD, config, attributes, ignore rules, and filesystem-monitor history are unchanged.
Go Analysis Framework: modular static analysis by go team
2 months ago
- Package analysis defines an interface for modular static analysis of Go code, where an Analyzer inspects one package at a time and can save information from lower-level packages for use in higher-level ones.
- The core type Analyzer includes fields like Name, Doc, Run function, Requires (dependencies on other analyzers), ResultType, and FactTypes, enabling composition and reuse.
- A Pass represents a single analysis unit, providing type info, syntax trees, and methods to report diagnostics (via Report/Reportf) and manage facts (export/import).
- Facts are serializable intermediate results that allow cross-package analysis, similar to compiler export data, and must be encoded using gob for deterministic caching.
- The framework supports validation (Validate function), testing utilities (analysistest), and driver programs like singlechecker and multichecker to run analyzers as standalone commands.
I scanned my AI agent framework for destructive/consequential actions, and wow
2 months ago
- A static scan of 25 AI agent frameworks found 30 instances where model-controlled parameters could reach consequential actions without authorization checks.
- The scanner distinguishes between validation (checking form) and authorization (checking permission for the specific action), catching patterns like network egress, file writes, shell execution, and SQL injection.
- The study had 81% precision on its first run; 12 false positives were found and fixed, with regression tests added to prevent recurrence.
- The scanner covers frameworks like MCP, LangChain, and OpenAI but not custom agent loops or action/observation dispatchers.
- The scanner is publicly available with no dependencies or network calls, and can be run on local code with a simple command.
Editing React components that never rendered
2 months ago
- CrossUI Studio uses static AST analysis instead of runtime introspection to understand React code structure, enabling tool functionality even when the app fails to render.
- Runtime tools like React DevTools rely on a live app, becoming useless during crashes, whereas static tools work on code that never rendered cleanly.
- Static analysis allows editing components that aren't rendering, including those inside .map() loops or imported from other files, by using mock data and AST nodes.
- The approach supports visualizing the dependency graph and code structure during app crashes, making broken pages entry points for debugging.
- Limitations include unresolvable dynamic imports, inability to see runtime values, and dependency on configuration for exotic resolution.
- Features like the canvas, prop inspector, and dependency graph are unified views over the AST, ensuring tool availability when runtime-based tools fail.

static analysis

OpenAI's work on Git for large repositories

Go Analysis Framework: modular static analysis by go team

I scanned my AI agent framework for destructive/consequential actions, and wow

Editing React components that never rendered