Summary
The v3.9.0 benchmark run (PR #853) shows a 1-edge divergence between engines on the same 567-file codebase:
- Native: 15,483 nodes, 30,609 edges
- WASM: 15,483 nodes, 30,610 edges
Node counts are identical, so the extraction layer agrees on symbols. The difference is in edge building or import resolution.
Context
Per project policy, both engines must produce identical results. Any divergence means the less-accurate engine has a bug that must be fixed.
Previous edge parity fixes in #810 resolved larger divergences; this 1-edge gap is a residual.
Next Steps
- Identify which edge differs (run both engines with
--json and diff the edge sets)
- Determine which engine is correct
- Fix the extraction/resolution bug in the incorrect engine
Summary
The v3.9.0 benchmark run (PR #853) shows a 1-edge divergence between engines on the same 567-file codebase:
Node counts are identical, so the extraction layer agrees on symbols. The difference is in edge building or import resolution.
Context
Per project policy, both engines must produce identical results. Any divergence means the less-accurate engine has a bug that must be fixed.
Previous edge parity fixes in #810 resolved larger divergences; this 1-edge gap is a residual.
Next Steps
--jsonand diff the edge sets)