Tests on main are failing

[1-Bort-1]

TL;DR

Tests on main are red because the POLAR_MATRICES forwarddiff check in test/solver/test_forwarddiff.jl flakes when a finite-difference step spans a knot of the polar's 5° alpha grid. #292 carries the fix; this PR sits on top of it and clears the advisory review's one comment, the misleading step-width comment.

What is failing and why

main CI failed the "AutoForwardDiff matches AutoFiniteDiff (LOOP, POLAR_MATRICES)" testset twice with the identical assertion: Expression: rel_err < 0.001 / Evaluated: 0.04001429189872729 < 0.001 at test/solver/test_forwarddiff.jl:88 — run 34446797250 (the merge of #285) and again on release PR #302 (run 34628824540). Nothing else on main failed.

The check differentiates a piecewise-linear polar whose alpha grid has knots every 5°, and a 1e-5 forward-difference step straddling a knot measures the average of two slopes while forward AD takes one. Whether a panel's converged alpha lands inside that window is decided by the NeuralFoil fixture, which is redrawn every CI run (#291) — that is the #287 flake, and the recorded failure value matches #292's knot-bisection exactly.

What changed

#292 (the base of this stack) fixes the flake in test/solver/test_forwarddiff.jl only: rtol 1e-7→1e-11 so the converged panel alpha stops wobbling, the finite-difference step 1e-5→1e-8 so the secant stays inside one polar cell, and the bound 1e-3→1e-4 — a tightening, not a loosening. Its CI is green.

This PR is the one commit on top: #292's advisory review flagged the added comment at line 83, which described the old wide step next to the new narrow one and could invert the intent for the next reader. Reworded to say what is true: # 1e-8 keeps the secant inside one cell of the polar's 5° alpha grid. The test result is untouched.

Left alone: the fixture-reproducibility root cause is #291, and the release is #302 — neither belongs to this stack.

Verification

• Reproduced first: main CI run 34446797250 and release Release VortexStepMethod v5.1.0 #302 run 34628824540 both fail rel_err < 0.001 / Evaluated: 0.04001429189872729 < 0.001 at test/solver/test_forwarddiff.jl:88
• test/solver/test_forwarddiff.jl green on the stacked branch (juliaserver, test env, 7/7) and green on unpatched main locally (7/7 — the flake only bites on a fresh fixture draw, so the recorded CI value is the reproduction)
• Local full suite: agent ci-local running · GitHub CI: pending PR open
• Docs build: n/a — no public symbol touched · REUSE lint: repo carries no REUSE · up to date with the stack base
• Benchmark: n/a — test-only change
• Risk: none for the comment reword; the flake fix's residual risk is stated in test_forwarddiff.jl's POLAR_MATRICES check flakes at ~4%, and the oper #292 (forward AD and a forward difference must agree on the side of the corner when the panel lands exactly on a knot)

Scope

+1 / −1 in test/solver/test_forwarddiff.jl and nothing else. Stack: 2/2 → #292.

Closes #303 · task VortexStepMethod.jl-303

[1-Bort-1]

Independent review (advisory)

Verdict: APPROVE · 0 inline, 0 off the diff

Good

• The diff is exactly one line and matches the PR card: it rewords the misleading comment to state the truth about the 1e-8 step.
• The new comment follows §3/§5: it is one line, states a non-obvious fact at the spot it bites, and contains no history, rationale, or negation of a previous version.
• Scope is correct: +1/-1 in the test file the base PR already touched; no public symbols, no docstrings, no dependencies, no generated files.

_{opencode, rubric CLEAN_CODE.md. A different lab from the implementer
on purpose: a reviewer sharing its blind spots would not flag its mistakes.}

[1-Bort-1]

Local full suite: FAIL (7 min, Julia 1.12.7, one cell of the matrix)

  YAML Geometry Tests                                            | �[32m 178  �[39m�[91m       �[39m�[33m        �[39m�[36m  178  �[39m�[0m   3.5s
  Kulfan fit and NeuralFoil                                      | �[32m  19  �[39m�[91m       �[39m�[33m        �[39m�[36m   19  �[39m�[0m   0.5s
  Polar matrix generation (create_2d_polars)                     | �[32m   8  �[39m�[91m       �[39m�[33m        �[39m�[36m    8  �[39m�[0m   0.1s
  SectionAero table round-trip and interpolation                 | �[32m  12  �[39m�[91m       �[39m�[33m        �[39m�[36m   12  �[39m�[0m   8.6s
  generate_section_aero builds a surface table                   | �[32m   7  �[39m�[91m       �[39m�[33m        �[39m�[36m    7  �[39m�[0m   0.6s
  NeuralFoil physical invariants                                 | �[32m  11  �[39m�[91m       �[39m�[33m        �[39m�[36m   11  �[39m�[0m   0.2s
  deform_section node count is delta-independent                 | �[32m   7  �[39m�[91m       �[39m�[33m        �[39m�[36m    7  �[39m�[0m   0.0s
  generate_polar_from_coordinates POLAR_VECTORS sweep            | �[32m   4  �[39m�[91m       �[39m�[33m        �[39m�[36m    4  �[39m�[0m   0.1s
  turn_trailing_edge! legacy crease cleanup                      | �[32m   8  �[39m�[91m       �[39m�[33m        �[39m�[36m    8  �[39m�[0m   0.2s
  load_neuralfoil_model missing weights errors                   | �[32m   1  �[39m�[91m    1  �[39m�[33m        �[39m�[36m    2  �[39m�[0m   1.1s
RNG of the outermost testset: Random.Xoshiro(0x5a86d6c2c2dea6ae, 0x95bbcbb32f949e9c, 0x098947565a9c2782, 0xfc5974c2d4abb8cc, 0x31dacfdfe75b55a0)
�[91m�[1mERROR: �[22m�[39mLoadError: �[91mSome tests did not pass: 6164 passed, 0 failed, 1 errored, 1 broken.�[39m
in expression starting at /home/agent/worktrees/VortexStepMethod.jl-303/test/runtests.jl:71
�[91m�[1mERROR: �[22m�[39mPackage VortexStepMethod errored during testing
Stacktrace:
  [1] �[0m�[1mpkgerror�[22m�[0m�[1m(�[22m�[90mmsg�[39m::�[0mString�[0m�[1m)�[22m
�[90m    @�[39m �[35mPkg.Types�[39m �[90m~/.julia/juliaup/julia-1.12.7+0.x64.linux.gnu/share/julia/stdlib/v1.12/Pkg/src/�[39m�[90m�[4mTypes.jl:68�[24m�[39m
  [2] �[0m�[1mtest�[22m�[0m�[1m(�[22m�[90mctx�[39m::�[0mPkg.Types.Context, �[90mpkgs�[39m::�[0mVector�[90m{PackageSpec}�[39m; �[90mcoverage�[39m::�[0mBool, �[90mjulia_args�[39m::�[0mCmd, �[90mtest_args�[39m::�[0mCmd, �[90mtest_fn�[39m::�[0mNothing, �[90mforce_latest_compatible_version�[39m::�[0mBool, �[90mallow_earlier_backwards_compatible_versions�[39m::�[0mBool, �[90mallow_reresolve�[39m::�[0mBool�[0m�[1m)�[22m
�[90m    @�[39m �[35mPkg.Operations�[39m �[90m~/.julia/juliaup/julia-1.12.7+0.x64.linux.gnu/share/julia/stdlib/v1.12/Pkg/src/�[39m�[90m�[4mOperations.jl:2642�[24m�[39m
  [3] �[0m�[1mtest�[22m
�[90m    @�[39m �[90m~/.julia/juliaup/julia-1.12.7+0.x64.linux.gnu/share/julia/stdlib/v1.12/Pkg/src/�[39m�[90m�[4mOperations.jl:2487�[24m�[39m�[90m [inlined]�[39m
  [4] �[0m�[1mtest�[22m�[0m�[1m(�[22m�[90mctx�[39m::�[0mPkg.Types.Context, �[90mpkgs�[39m::�[0mVector�[90m{PackageSpec}�[39m; �[90mcoverage�[39m::�[0mBool, �[90mtest_fn�[39m::�[0mNothing, �[90mjulia_args�[39m::�[0mCmd, �[90mtest_args�[39m::�[0mCmd, �[90mforce_latest_compatible_version�[39m::�[0mBool, �[90mallow_earlier_backwards_compatible_versions�[39m::�[0mBool, �[90mallow_reresolve�[39m::�[0mBool, �[90mkwargs�[39m::�[0m@Kwargs�[90m{io::IOContext{IO}}�[39m�[0m�[1m)�[22m
�[90m    @�[39m �[35mPkg.API�[39m �[90m~/.julia/juliaup/julia-1.12.7+0.x64.linux.gnu/share/julia/stdlib/v1.12/Pkg/src/�[39m�[90m�[4mAPI.jl:552�[24m�[39m
  [5] �[0m�[1mtest�[22m�[0m�[1m(�[22m�[90mpkgs�[39m::�[0mVector�[90m{PackageSpec}�[39m; �[90mio�[39m::�[0mIOContext�[90m{IO}�[39m, �[90mkwargs�[39m::�[0m@Kwargs�[90m{}�[39m�[0m�[1m)�[22m
�[90m    @�[39m �[35mPkg.API�[39m �[90m~/.julia/juliaup/julia-1.12.7+0.x64.linux.gnu/share/julia/stdlib/v1.12/Pkg/src/�[39m�[90m�[4mAPI.jl:169�[24m�[39m
  [6] �[0m�[1mtest�[22m�[0m�[1m(�[22m�[90mpkgs�[39m::�[0mVector�[90m{PackageSpec}�[39m�[0m�[1m)�[22m
�[90m    @�[39m �[35mPkg.API�[39m �[90m~/.julia/juliaup/julia-1.12.7+0.x64.linux.gnu/share/julia/stdlib/v1.12/Pkg/src/�[39m�[90m�[4mAPI.jl:158�[24m�[39m
  [7] �[0m�[1mtest�[22m�[0m�[1m(�[22m; �[90mname�[39m::�[0mNothing, �[90muuid�[39m::�[0mNothing, �[90mversion�[39m::�[0mNothing, �[90murl�[39m::�[0mNothing, �[90mrev�[39m::�[0mNothing, �[90mpath�[39m::�[0mNothing, �[90mmode�[39m::�[0mPackageMode, �[90msubdir�[39m::�[0mNothing, �[90mkwargs�[39m::�[0m@Kwargs�[90m{}�[39m�[0m�[1m)�[22m
�[90m    @�[39m �[35mPkg.API�[39m �[90m~/.julia/juliaup/julia-1.12.7+0.x64.linux.gnu/share/julia/stdlib/v1.12/Pkg/src/�[39m�[90m�[4mAPI.jl:186�[24m�[39m
  [8] �[0m�[1mtest�[22m�[0m�[1m(�[22m�[0m�[1m)�[22m
�[90m    @�[39m �[35mPkg.API�[39m �[90m~/.julia/juliaup/julia-1.12.7+0.x64.linux.gnu/share/julia/stdlib/v1.12/Pkg/src/�[39m�[90m�[4mAPI.jl:175�[24m�[39m
  [9] top-level scope
�[90m    @�[39m �[90m�[4mnone:1�[24m�[39m
 [10] �[0m�[1meval�[22m�[0m�[1m(�[22m�[90mm�[39m::�[0mModule, �[90me�[39m::�[0mAny�[0m�[1m)�[22m
�[90m    @�[39m �[90mCore�[39m �[90m./�[39m�[90m�[4mboot.jl:489�[24m�[39m
 [11] �[0m�[1mexec_options�[22m�[0m�[1m(�[22m�[90mopts�[39m::�[0mBase.JLOptions�[0m�[1m)�[22m
�[90m    @�[39m �[90mBase�[39m �[90m./�[39m�[90m�[4mclient.jl:283�[24m�[39m
 [12] �[0m�[1m_start�[22m�[0m�[1m(�[22m�[0m�[1m)�[22m
�[90m    @�[39m �[90mBase�[39m �[90m./�[39m�[90m�[4mclient.jl:550�[24m�[39m

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!