CHANGELOG Summary

📋 CHANGELOG Summary

GRANITE v0.6.8 | ← Benchmarks & Validation | Developer Guide →

High-level version history. For granular bug fix details see the full CHANGELOG and Known Fixed Bugs.

---

v0.6.7.2 — Package Overhaul, CLI Hardening & Documentation Integrity (April 30, 2026)

Focus: Complete architectural overhaul of the Python analysis layer and post-release hardening.

• granite_analysis package overhaul: Monolithic scripts/sim_tracker.py and scripts/dev_benchmark.py decomposed into a 4-layer typed Python package: core/models.py (13 frozen dataclasses), core/parsers.py (12 compiled regex patterns), core/runners.py (shared telemetry loop), cli/ (thin argparse entry points). PEP 517 installable via pip install -e .[dev].
• Old scripts permanently removed: scripts/sim_tracker.py, scripts/dev_benchmark.py, scripts/dev_stability_test.py
• C++ engine version banner corrected from "0.6.5" to "0.6.7" in src/core/cli_parser.cpp
• CLI exit hardening: raw FileNotFoundError tracebacks replaced with clean sys.exit() messages
• run_granite.py docs subcommand: Sphinx HTML build with --open flag
• python/requirements.txt: flat dependency manifest for CI/Docker environments
• 25 broken relative links fixed across README.md and all docs/ files
• Quick Start Guide rewritten for granite_analysis package, new Step 7 (HPC/SLURM)
• Wiki Documentation Index synced to v0.6.8

v0.6.7 — The Repository Seal (April 27, 2026)

Focus: CI stabilization, identity audit, VORTEX Gold Master, 107-test milestone.

• C++ smoke tests: 4 new suites (92 → 107 tests) — AMR, horizon finder, M1 radiation, HDF5 roundtrip
• VORTEX Gold Master: GW audio sonification, Zen mode, cinematic autopilot, volume controls
• Identity purge: "GRANITE Collaboration" → "Liran M. Schwartz" across 37 files
• Documentation voice: "We/Our" → "I/The" across 24 passages
• Git history recovery after accidental --force push (214 commits restored)
• .gitattributes Linguist override for C++ language dominance
• include/README.md: C++ header architecture map

v0.6.5 — The Stability Update (April 7, 2026)

Focus: Forensic rescue operation. Master reset of CCZ4 and AMR to the v0.6.0 verified baseline, preserving all P0 memory safety fixes.

• CCZ4 core reset: removed over-engineered mathematical constraints that destroyed interpolation balance
• AMR reset: restored std::round in child cell snapping; restored full-domain prolongation
• Memory safety: levels_.reserve() prevents pointer invalidation during refinement cascade
• MICRO_OFFSET: domain phase-shift prevents r=0 division-by-zero for all centered-domain runs
• sim_tracker.py: procedural architecture restored; 8-layer stability summary guard
• Test count: 92/92 (100% pass rate)
• Validated: single_puncture (t=500M, ×84.8 ‖H‖₂ reduction), B2_eq (t=500M, ×61.3 reduction)

v0.6.0 — The Puncture Tracking Update (April 4, 2026)

Focus: Binary black hole capability. AMR with subcycling. Python dashboard.

• TwoPuncturesBBH spectral initial data with Bowen-York extrinsic curvature
• Full Berger-Oliger recursive subcycling implemented
• Block-merging algorithm for close-binary refinement (eliminates MPI deadlocks)
• CFL synchronization: level timesteps cascade from L0 master
• sim_tracker.py: append-only live dashboard with NaN forensics and Matplotlib summaries
• gauge_wave and B2_eq benchmark configurations added
• validation_tests.yaml: machine-readable pass/fail criteria for 3 physics sectors
• 90 → 92 tests (+2 selective advection unit tests)

v0.5.0 — Repository Professionalization (April 2, 2026)

Focus: Repository structure, security, CI/CD, community standards.

• Root directory cleaned to 8 essential files
• scripts/, docs/, .github/ directories created and populated
• GitHub Actions CI workflow (.github/workflows/ci.yml)
• CodeQL static analysis (.github/workflows/codeql.yml)
• SECURITY.md, CONTRIBUTING.md, CODE_OF_CONDUCT.md

Phase 7 — Selective Upwinding & Sommerfeld BCs (April 2, 2026)

• Chi-weighted hybrid advection (d1up far-field, d1 near puncture)
• Sommerfeld radiative BCs for Two-Punctures data
• Algebraic constraint enforcement (det(γ̃)=1, tr(Ã)=0)
• Adaptive CFL guardian (emergency dt reduction at CFL > 0.95)
• Numerical floors (χ, α) with IEEE 754-safe NaN guards

Phase 6 — Single Puncture Stability Diagnostics (March 31, 2026)

• Root cause of t≈6.25M crash confirmed: domain-size limitation (not code bug)
• Evidence matrix from 6 controlled experiments (CFL, η, Sommerfeld, upwinding)
• dev_benchmark.py forensic diagnostic runner
• Gauge wave travel time formula: t_crash ≈ domain_size / √2

Phase 5 — PPM, GridBlock Flat Layout, GW Analysis (March 30, 2026)

• H2 fix: GridBlock single flat allocation (22 → 1 heap allocation)
• H3 fix: RHS zero-out loop spatial-outer, var-inner
• PPM reconstruction (Colella & Woodward 1984) implemented
• Full spin-weighted spherical harmonics (all ℓ,m modes via Wigner d-matrices)
• Recoil kick (Ruiz et al. 2008) implemented
• 81 → 90 tests

Phase 4D — GR Metric Coupling, MP5 (March 30, 2026)

• C1 fix: EOS-exact sound speed (not hardcoded Γ=1)
• C3 fix: HLLE uses actual GRMetric3 (not flat dummy)
• H1 fix: KO dissipation in single OpenMP region (not 22 separate spawns)
• MP5 5th-order reconstruction (Suresh & Huynh 1997)
• 74 → 81 tests

Phases 4A/4B/4C — HLLD, CT, Tabulated EOS (March 30, 2026)

• HLLD Riemann solver (Miyoshi & Kusano 2005) — all 7 wave families
• Constrained Transport ∇·B=0 to machine precision (Evans & Hawley 1988)
• TabulatedEOS with tri-linear interpolation, NR temperature inversion, beta-equilibrium
• TOV fix: 1.0e5 cm/km (not RSUN_CGS)
• 47 → 74 tests (+20 tabulated EOS tests)

---

This summary covers all phases through v0.6.7.2. For complete technical details including all bug IDs, exact line references, and physics derivations: Full CHANGELOG on GitHub

---