Problem

The racedetector reports false positive data races on stack-local variables that cannot actually have data races because each goroutine has its own stack.

Affected Patterns

Fixed in v0.7.2 (Quick Fix)

• ✅ Named return values: func foo() (result int, err error)
• ✅ Function parameters: func process(x int)

Still Affected (Phase 2)

• ❌ Local variables: var x int inside function
• ❌ Loop variables: for i, v := range items
• ❌ Short-lived variables in expressions

Root Cause

Our AST-based instrumentor lacks escape analysis that the standard Go race detector has (SSA-based with full type information). We instrument ALL variable accesses without knowing if they're stack-local or heap-allocated.

See internal analysis document: FALSE_POSITIVES_ANALYSIS.md

Solution Roadmap

• Phase 1 (v0.7.2): Skip named returns and parameters
• Phase 2: Add heuristics for loop variables, short-lived locals
• Phase 3: Integrate go/types for escape analysis

Related

• Issue [BUG] false positive races found, only half of a race is reported (previous access stack trace not available) #17 (stack trace display)
• Go race detector uses SSA-level instrumentation