WASM engine: intermittent V8 crash / segfault after build completes

[carlos-alm]

Description

The WASM engine intermittently crashes with a V8 fatal error or segfault (exit code 139) after the build successfully completes. The graph data is written correctly, but the process crashes during cleanup/shutdown.

Reproduction

# Install dev build 3.9.4-dev.8
# Run WASM full build on codegraph's own src/ (~309 TS files)
codegraph build src --engine wasm --no-incremental

Crash rate: 2 out of 5 runs (40%) on Windows 11, Node 22.18.0.

Crash signatures

V8 Fatal Error:

Fatal error in , line 0
# unreachable code

Segfault:

Exit code 139 (Segmentation fault)

Stack trace points to tree-sitter WASM:

function get type() {
    return this.tree.language.types[this.typeId] || "ERROR";
}

The crash occurs during walkWithVisitors → walk recursion in the complexity/CFG/dataflow analysis phase, or during process shutdown GC cleanup of WASM objects.

Environment

• Version: 3.9.4-dev.8
• OS: Windows 11 Pro 10.0.26200
• Node: v22.18.0
• Engine: WASM only (native engine unaffected)

Impact

• Build data is written correctly before the crash — the graph DB is valid
• Exit code is non-zero, which would break CI pipelines that check exit status
• The crash is non-deterministic, making it hard to debug

Possible causes

• Tree-sitter WASM parser objects not being properly disposed before Node GC runs
• Race condition between WASM memory cleanup and V8 garbage collector
• Deep recursion in walkWithVisitors exhausting WASM stack space on larger codebases