Skip to content

fix(Euler): upstream dev/harry changes and sync Codex skills - #23

Merged
harryzhou2000 merged 5 commits into
CFDLAB-THU:mainfrom
harryzhou2000:harry/fix/dev-harry-main-20260825
Aug 25, 2026
Merged

fix(Euler): upstream dev/harry changes and sync Codex skills#23
harryzhou2000 merged 5 commits into
CFDLAB-THU:mainfrom
harryzhou2000:harry/fix/dev-harry-main-20260825

Conversation

@harryzhou2000

Copy link
Copy Markdown
Collaborator

Summary

  • Bring the three current dev/harry commits onto upstream main: the wave CGNS generator, recorded solver configurations, and configurable Euler/RANS hard limits with schemas and tests.
  • Expose repository-local OpenCode skills to Codex through a relative symlink, normalize Codex skill frontmatter, and link the PyVista helper to the canonical detonation2d renderer.
  • Remove a trailing-whitespace defect from the CRM run configuration already carried by the branch.

Validation

  • All three repository-local skills passed Codex quick_validate.
  • Bundled Python helpers compiled and their help entry points loaded.
  • Per request, C++ builds and unit tests were skipped; no build or UT result is claimed.

Add a parameterized CGNS generator for periodic wave meshes used by the scalar/vector Fourier analysis. The script supports configurable cell count, domain length, and origin so h/2h/4h meshes can share an aligned physical core while retaining periodic boundary face records.

Verification: py_compile scripts/generate_8x8_wave_cgns.py; generated and read-validated the 8x8, 6x6 h2, and 5x5 h4 meshes during Fourier analysis work; regenerated MGTestFourier integrated plots.

committed by opencode with model internal_openai/openai/openai/gpt-5.5
Move model-specific RANS configuration types next to the inline RANS kernels and thread concrete config objects through all RANS primitive call paths. Raise the default SA production cap from 100 to 1e5 to recover ordinary-SA convergence while preserving Orion behavior with an explicit limit of 100.

Expose existing k-omega, SST, and realizable-k-epsilon hard limits without changing their historical defaults, regenerate Euler schemas, and add focused config and SA source-limit tests. Document that formatting is owned by the pre-commit hook and the repository formatting script.

Verification: built all eight non-EulerP Euler executables with -j8; euler_test_rans passed 29 cases and 66 assertions.

committed by codex with model gpt-5.6-sol
@harryzhou2000
harryzhou2000 marked this pull request as ready for review August 25, 2026 14:35
@harryzhou2000 harryzhou2000 self-assigned this Aug 25, 2026
@harryzhou2000

Copy link
Copy Markdown
Collaborator Author

/ci run

@harryzhou2000

Copy link
Copy Markdown
Collaborator Author

/ci-run

Repair reactive cell-mean species after initialization and restart loading, while validating that density and temperature remain physical. Share the species repair logic with fixed-increment updates so initialization follows the same conservative correction policy.

Use a temporary simplex-repaired composition for chemical source evaluation at reconstructed quadrature points. Keep convective and diffusive transported species values unprojected, and document that only thermodynamic and transport properties use repaired mass fractions.

Add MPI evaluator coverage for repaired reactive cell means. Verified the focused test at 1, 2, 4, and 8 ranks and ran the large 16-rank non-Strang detonation case through two coupled RHS iterations past the former strict mass-fraction failure.

committed by codex with model gpt-5
Expose the existing OpenCode skill tree through a relative .codex/skills symlink so both agent environments share one maintained source. Normalize the skill frontmatter for Codex validation and link the PyVista helper to the canonical detonation2d renderer instead of retaining a stale duplicate. Clean the pre-existing trailing whitespace in the CRM run config included by the branch.

Validation performed before the final no-check directive: all three skills passed Codex quick_validate; bundled Python helpers compiled and their help entry points loaded. No C++ build or unit tests were run to completion, per request.

committed by codex with model deepseek-v4-flash
@harryzhou2000
harryzhou2000 force-pushed the harry/fix/dev-harry-main-20260825 branch from a55a0e5 to 34a57be Compare August 25, 2026 15:30
@harryzhou2000
harryzhou2000 merged commit c3d9f6a into CFDLAB-THU:main Aug 25, 2026
@harryzhou2000
harryzhou2000 deleted the harry/fix/dev-harry-main-20260825 branch August 25, 2026 15:37
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant