Refuse a contour XFoil's panel code cannot take

[1-Bort-1]

TL;DR

analyze_sweep(::XFoilSolver, ...) checks the contour before handing it to XFoil and throws ArgumentError, because a self-crossing contour can reach a bare Fortran STOP in there and end the Julia process with exit code 0 and no exception. The same check catches a contour with more nodes than XFoil's panel arrays hold, which XFoil refused while leaving the previously loaded airfoil in place, so every angle came back solved on the wrong shape.

What was wrong

Found from 1-Bart-1/BeyondTheSim.jl#53: an example script ended its Julia process partway through, silently, exit code 0, no stacktrace — in a REPL it takes the session away, in a script it looks like a successful run that printed half its output. Bisecting the call, Xfoil.set_coordinates returns normally and Xfoil.solve_alpha(0.0, Re; reinit=true) never does. Exit 0 with nothing on stdout or stderr is a gfortran STOP with no message.

The contours that do it come out of deform_section, whose re-wrap of an already-wrapped thin section produces a curve whose upper and lower surfaces cross. That is not an airfoil, and a panel method has no solution for it. Of nine shapes sliced from one mesh section at nine rolling-ball radii, four returned and five took the process with them, while every geometric statistic I measured (shortest panel, neighbouring-panel ratio, maximum turn, crossing count) put survivors and killers in the same range:

wrap radius	shortest panel	neighbour ratio	max turn	crossings	XFoil
0.05	8.03e-4	2.20	45.0°	34	killed the process
0.10	7.38e-4	4.14	48.3°	34	returned
0.15	2.08e-4	16.65	180.0°	60	killed the process
0.20	2.66e-4	14.45	180.0°	58	killed the process
0.30	3.09e-4	13.54	180.0°	48	killed the process
0.40	4.07e-4	3.54	180.0°	29	returned
0.60	3.82e-4	6.29	180.0°	29	returned
0.80	3.90e-4	5.53	180.0°	35	returned
the stored .dat, one wrap of the same cloud	5.01e-4	7.50	32.3°	0	returned

The rows at 0.05 and 0.10 agree on every column and disagree on life and death, because the STOP is not geometric. It is the silent one in XFoil's IBLPAN (xfoil_light src/xpanel.f:1432, its message lines commented out), which fires when the longer boundary-layer side plus the wake exceeds IVX = 229 stations. On a self-crossing contour the stagnation point lands beside the trailing edge, so one side takes nearly all N nodes; it cannot fire below 203 nodes. Predicting that from the inviscid solution matched 9 of 9 synthetic runs (a figure-eight and a rear-swapped NACA 0012 die at 220-250 nodes and return NaN at 160-200). The guard enforces the precondition a panel method actually has, a simple closed curve, rather than predicting that overflow.

A contour that does not already end on its first node is closed by a panel back to it before the test, so a trailing-edge gap does not hide a fold there. That case is not hypothetical: compare_live_polar builds its section straight from kulfan_to_coordinates, which at TE_thickness=0.02 returns first (1.0, 0.01) and last (1.0, -0.01).

I do not claim this makes analyze_sweep proof against every contour XFoil might STOP on: a simple closed contour over 202 nodes whose stagnation point moves far from mid-contour could still overflow IBLPAN. It ends the class a shrink wrap produces, and turns a process death into a failure the caller can catch. The cause is fixed upstream in byuflowlab/xfoil_light, which this branch does not wait on. The upstream branch IBLPAN checks the bound before writing and returns the point unconverged instead of stopping (1-Bort-1/xfoil_light@63a3065). Once it reaches a xfoil_light_jll release, such a contour gives converged = false rather than ending the process. The guard here still earns its place then: a self-crossing contour's "solution" is NaN or nonsense (cm = -19 on one of them), and ArgumentError names the crossing panels.

The second defect in the same function

While bisecting I found that a contour with more than Xfoil.IQX - 5 = 281 nodes is refused by XFoil's ABCOPY, which prints to stdout and returns, leaving the previously loaded airfoil in the panel arrays. analyze_sweep never looked, so it solved the sweep on whatever shape was loaded before and returned it as this one. Feeding a NACA 0012 at 299, 339, 399, 499 and 569 nodes gave cl = 0.5604994052186286 every time — the 279-node answer, to the last digit. Same function, same idea (say when XFoil did not take the contour), so it rides here.

Where I would push back

The real bug is upstream of this: deform_section's re-wrap should not produce a self-crossing contour in the first place, and until it stops doing so these sections have no XFoil polar at all rather than a poor one. That is the shrink wrap's own concern and would change what every generated polar means, so it is #321 rather than this diff.

The contour test is O(n²) in the node count — 239 nodes is ~28k segment pairs, run once per sweep against a viscous march of many angles. I did not benchmark it; it is not on any hot path I can find, and compare_live_polar's per-panel call is still one XFoil sweep per check.

The throw does not take a batch down with it where one is running: generate_airfoils already wraps generate_airfoil_aero in try/catch and, with the default reuse_valid_airfoils=true, warns and reuses the nearest valid airfoil, so one self-crossing delta costs that airfoil rather than every column of the grid. compare_live_polar has no such catch and a panel whose shape is not a simple closed curve now stops a sweep over panels there — before this branch that same panel ended the process, so nothing graceful is being given up, but that is the caller where somebody will want a catch first.

The three predicates side_of_line, segments_cross and crossing_panels sit in airfoil_solvers/common.jl although only XFoil calls them today. They say nothing about XFoil and belong beside DeformedSection and deform_section, where the shared section geometry already is; what is XFoil's own policy, validate_xfoil_contour, is in xfoil_solver.jl. The end-to-end check that a valid contour still sweeps is test/solver/test_backend_comparison.jl, the suite's only XFoil march — the new testset pins the guard itself, that it accepts a closed contour and one with a trailing-edge gap alike.

Verification

• Reproduced first: Xfoil.set_coordinates ok, then solve_alpha(0.0, Re; reinit=true) never returns — process gone, exit 0, no Julia error and no Fortran message
• test/airfoil_aero/test_airfoil_aero.jl red before, green after (8/8 new assertions, exit 0). Red was run with the source change stashed: the folded contour returned instead of throwing, and the 399-node contour printed Maximum number of panel nodes : 281 / Current airfoil cannot be set. and still returned a solution. The trailing-edge panel was red on the unpatched function too — crossing_panels([0,1,1,2],[0,2,-2,1]) gave nothing, and gives (2, 4) now
• The nine real contours above now raise ArgumentError naming the crossing panels, all nine in one process where each killer previously needed its own; the valid one still runs
• test/solver/test_backend_comparison.jl PASS 15/15, the suite's only XFoil march · test/airfoil_aero/test_live_polar.jl PASS 93/93
• Docs build clean (exit 0, no missing docstring) · up to date with origin/main at 9d0f44e
• Full local suite on the head commit d1ffa0c: PASS 6721/6722 (1 pre-existing broken), 8m34s, exit 0, Julia 1.13.0 — the new testset 8/8 and ForwardDiff linearize 10/10, the latter being the check that was red on Windows before the branch merged main (Flaky POLAR_MATRICES forwarddiff test on Windows / Julia 1.12 (knot-proximity in piecewise-linear polar interpolation) #360, fixed there). The box's earlier red run on this commit ended in KILLED by SIGTERM after 4m while compiling test_plotting.jl, sent from outside the run, and not in a test; the Project.toml … name and a UUID error it reported comes from Pkg.test after that kill.
• GitHub CI on the head commit d1ffa0c: all 7 checks green, Julia 1.12 windows included
• REUSE lint: n/a, this repo has no bin/reuse_lint
• Benchmark: n/a
• Risk: the guard rejects contours that today happen to return — four of the nine above. Their answers were a panel solution on a curve that crosses itself, so I judge that no loss, but it is a behaviour change for any caller that was using one.

Scope

+101 / -1 across 5 files: the check and its three geometry helpers in airfoil_aero/, one regression testset, one changelog entry under ## Unreleased, four @docs entries. The regression test uses a synthetic folded Kulfan section rather than the mesh section that found this — that geometry is in a private repo and this one is public. Closes #322.

Task VortexStepMethod.jl-320

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[1-Bart-1]

Merge main into this branch

[1-Bart-1]

Fix the failing test and conflicts @1-Bort-1

[1-Bort-1]

Picked up as VortexStepMethod.jl-320, on the branch this pr already has.

[1-Bort-1]

The description above is this branch as it stands.

Commits

• 2637859 Test and support Julia 1.12 and 1.13, drop 1.11
• 0576422 Changelog: Julia 1.12 and 1.13
• 2b312ea Merge pull request Test and support Julia 1.12 and 1.13, drop 1.11 #327 from OpenSourceAWE/agent/326-julia-1-13-vortexstepmethod-jl
• f63f7b6 Test solve! on a body built from two wings
• dcaf811 Zero the damping correction when smooth_circulation! skips smoothing
• d3cf02f Pin lei_poly_coeffs to the Python Breukels regression output
• deac49a Test against the Python verification cases' references
• 4078a92 Test that moments scale as N·m when the geometry scales
• 78bba14 Compare the solve! solutions directly and changelog the unit labels
• e971c74 Give untyped empty containers their element type
• 9ef9f10 Build a Solver from VSMSettings or from panel and section counts
• 47d8f19 Rename the *_array locals to _dist, after the per-panel fields
• be131ba Give the renamed AIC and results docstrings their full signatures
• 7f2d15f Build every flat inviscid test wing with one helper
• 1fac603 Merge pull request Test the solver against the references in the Python verification cases #334 from OpenSourceAWE/agent/93-add-test-cases-from-the-python-verificat
• 226fe70 Merge pull request Test that moments come out in N·m by scaling the wing's geometry #335 from OpenSourceAWE/agent/209-add-tests-that-check-if-moment-have-righ
• a95bfac Merge pull request Give every untyped empty container its element type #338 from OpenSourceAWE/agent/147-refactoring-part-iii
• b83d4c1 Merge pull request Test solve! on a body built from two wings #339 from OpenSourceAWE/agent/143-add-test-for-multiple-wings
• a932ee1 Merge remote-tracking branch 'origin/main' into agent/153-remove-the-parameter-body-aero-from-the-
• 246237a Merge pull request Rename the *_array locals to _dist, matching the per-panel fields they hold #341 from OpenSourceAWE/agent/336-rename-the-array-locals-to-say-whether-t
• d165c7f Deprecate the Solver constructors that take a body_aero
• fc3daab Name the apparent wind va, va_vec and va_dist in src locals and arguments
• 245ab96 Truncate the SVD in the least-squares Kulfan fit
• 30b0756 Share one straight-segment kernel between bound and trailing vortices
• 90afa39 Name the apparent wind va, va_vec and va_dist in tests, examples and docs
• 4965a5f Add the Gaunaa et al. 2024 spanwise-flow viscous drag correction as a solver option
• 701adef Wrap the renamed lines at 92 columns
• 524103b Finish the va rename in error messages, test helpers and alignment
• 9af8d69 Turn the body about a reference point in set_va!, and apply yaw_rate
• e185eec Plot cl, cd and cm against alpha per panel for every aero model
• de35935 Share one panel coefficient path between the polar plots
• 580cd2f linearize and make_dual_shadow take a body with several wings
• 1d76212 Name linearize's locals and share the two-wing test setup
• 1c493cf Merge pull request Add the Gaunaa et al. 2024 spanwise-flow viscous drag correction as an opt-in solver setting #352 from OpenSourceAWE/agent/332-add-the-optional-spanwise-flow-viscous-d
• 0aefbe8 Merge branch 'main' into agent/295-the-kulfan-least-squares-fit-has-no-rank
• bf7246f Merge origin/main into agent/329-set-va-rotates-the-body-about-the-origin
• 4ccdea6 Merge origin/main into agent/344-linearize-and-make-dual-shadow-accept-a-
• 597339d Merge origin/main into agent/153-remove-the-parameter-body-aero-from-the-
• ff2900d Merge origin/main: fold the va names into the viscous drag correction
• 0cc53b2 Merge remote-tracking branch 'origin/main' into agent/331-plot-section-polars-covers-only-polar-ve
• 9ae2ab2 List plot_section_polars under Added in the changelog
• 574b15b Give each wing's unrefined-section range one source and test the angle split
• 465d199 Test both vortex functions for azimuthal velocity in one testset
• 3bef8ce Merge pull request Zero the damping correction when smooth_circulation! skips smoothing, and cover it and lei_poly_coeffs with unit tests #328 from OpenSourceAWE/agent/53-add-unit-tests-for-uncovered-functions
• 83867a3 Merge pull request Name the apparent wind va, va_vec and va_dist everywhere except the public API #349 from OpenSourceAWE/agent/337-one-name-per-apparent-wind-quantity-va-v
• 28d4e11 Merge origin/main into agent/153-remove-the-parameter-body-aero-from-the-
• 6ecd77a Merge origin/main: carry the va_vec names into the reference_point change
• 0b28330 Merge origin/main into agent/344-linearize-and-make-dual-shadow-accept-a-
• c02a2ab Merge origin/main: carry the va rename into the shared vortex kernel
• 3dc74fc Fold the second Unreleased "Fixed" changelog heading into the first
• f8761aa Fix tests
• 8a2b61d Merge branch 'main' into agent/329-set-va-rotates-the-body-about-the-origin
• e24e1e4 Merge pull request linearize and make_dual_shadow take a BodyAerodynamics with several wings #355 from OpenSourceAWE/agent/344-linearize-and-make-dual-shadow-accept-a-
• 086ee32 Merge pull request Bound and trailing vortices share one segment kernel, so the trailing core induces azimuthal velocity #351 from OpenSourceAWE/agent/333-velocity-3d-trailing-vortex-projects-ont
• 7103702 Merge remote-tracking branch 'origin/main' into agent/331-plot-section-polars-covers-only-polar-ve
• a035f39 Merge origin/main: make_dual_shadow builds every wing and keeps the reference point
• 1a80247 Merge remote-tracking branch 'origin/main' into agent/295-the-kulfan-least-squares-fit-has-no-rank
• 12c11f4 Merge origin/main: build the two-wing linearize test solver from counts
• da666bf Build the billowing example's solvers with Solver(settings)
• b725371 Warn when the least-squares Kulfan fit drops singular values
• 2665240 Merge pull request plot_section_polars draws cl, cd and cm against alpha per panel, for every aero model #354 from OpenSourceAWE/agent/331-plot-section-polars-covers-only-polar-ve
• 7724c0d Merge pull request set_va! turns the body about a stored reference_point, and set_va!(body_aero, settings) applies yaw_rate #353 from OpenSourceAWE/agent/329-set-va-rotates-the-body-about-the-origin
• 4cd7a7e Merge origin/main: keep the count constructors under set_va!'s reference point
• 0c94c6d Merge pull request Build a Solver from VSMSettings or from panel and section counts, deprecate the body_aero constructors, and check the sizes in solve! #340 from OpenSourceAWE/agent/153-remove-the-parameter-body-aero-from-the-
• abd1176 Merge remote-tracking branch 'origin/main' into agent/BeyondTheSim.jl-53
• 9d0f44e Merge pull request Truncate the SVD in the least-squares Kulfan fit, and warn when it does, so crowded stations give bounded weights #342 from OpenSourceAWE/agent/295-the-kulfan-least-squares-fit-has-no-rank
• fefa7a6 Merge remote-tracking branch 'origin/main' into agent/BeyondTheSim.jl-53
• d1ffa0c Test the panel that closes the contour back to node 1

Review just these changes

[1-Bort-1]

Merged in fefa7a6, with the follow-up merge of #342 in the same push; CHANGELOG.md was the only conflict. The PR reads MERGEABLE and all 7 checks are green.

[1-Bort-1]

Merged origin/main into the branch in fefa7a6; the only conflict was CHANGELOG.md, where this branch's Fixed bullet moved under main's ## Unreleased. No test change was needed: the red check was AutoForwardDiff matches AutoFiniteDiff (LOOP, POLAR_MATRICES) on Julia 1.12 / windows at 0.0421, which is #360 — fixed on main on 2026-09-17, and the branch was 60 commits behind it. All 7 checks are green now, windows included, and the local full suite passes 6718/6719 (the 1 is the pre-existing broken).

[1-Bort-1]

Independent review (advisory)

Verdict: APPROVE WITH COMMENTS · 2 inline, 0 off the diff

Good

• Single chokepoint: grep shows no Xfoil.* call site outside xfoil_solver.jl, and analyze_section delegates to analyze_sweep, so the guard cannot be bypassed.
• Xfoil.IQX = 286 confirmed in the installed package (Xfoil/n7wnc/src/Xfoil.jl:12), so the 281-node bound matches XFoil's ABCOPY refusal the card describes.
• Order is right: validating before set_coordinates is the only place it works, since a refused contour leaves the previous airfoil loaded and pane runs after set_coordinates.
• No duplication: grepped intersect|cross|orient|signed|det over src — the repo had no segment-intersection or side-of-line helper, only plane crossings in obj_slice.jl.
• Regression test throws before any Fortran call, so it is fast and cannot itself kill the process; it covers both classes plus a clean control contour.
• Docstrings state only what, no why; CHANGELOG entry sits under Unreleased/Fixed at the neighbours' length; all four new private names are added to private_functions.md so docs stay complete.
• Scope matches the card (5 files, guard + 3 helpers + 1 testset + changelog + docs), tree clean, .agent/plan.md untracked.

Not good

• src/airfoil_aero/airfoil_solvers/common.jl:148 — crossing_panels assumes a wrap-around closed contour: it never tests the closing segment node n → node 1 and it skips the pair (1, last_panel). That holds for deform_section output (shrink_wrap ends at px[1], so first == last), but compare_live_polar builds its section straight from kulfan_to_coordinates, which leaves first ≠ last for any non-zero TE_thickness — there panels 1 and n-1 are genuinely non-neighbouring and the trailing-edge segment is untested, so a fold at the TE passes the guard and can still reach the STOP. For a truly closed contour the skip is dead anyway, since the shared node makes the strict test false.
• src/airfoil_aero/airfoil_solvers/xfoil_solver.jl:62 — Throwing here aborts batches that were built to degrade instead: generate_airfoil_aero loops deltas and takes the modal node count specifically so "a single degenerate wrap cannot drop every other column", and compare_live_polar's docstring promises a bad reference "comes back NaN rather than throwing, so a sweep over panels does not stop at the first one that fails". One self-crossing delta or panel now fails the whole grid or panel sweep; the card notes the behaviour change but not that these two callers document the opposite contract.
• The three geometry helpers sit in airfoil_solvers/common.jl (shared-backend file) although their only caller is the XFoil backend.
• Testset name says "XFoil refuses" while XFoil is never reached — the refusal is entirely ours.
• side_of_line(a, b, p) -> Float64 in the docstring, but the function is generic in its argument types.
• No end-to-end assertion here that a valid contour still sweeps; that is left to test/solver/test_backend_comparison.jl.
• The card's disclosed behaviour change (four of nine contours that used to return now throw) gets no caller-side handling anywhere.

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

[1-Bart-1]

Solving the cause rather than the effect

[1-Bort-1]

Question from VortexStepMethod.jl-320

byuflowlab/xfoil_light is not a repo this box may push to.
Task VortexStepMethod.jl-320 (#320 — Refuse a contour XFoil's panel code cannot take) wants to fork it, open a pull request and file an issue there.
Thumbs up, or reply yes, and it goes ahead. Anything else and it does not.

Asked on Slack, and answered there.

[1-Bort-1]

Opened in the repository this belongs to:

• Pull request: Return unconverged instead of STOP when the BL arrays overflow byuflowlab/xfoil_light#1

[1-Bort-1]

Fixed at the cause upstream in 1-Bort-1/xfoil_light@63a3065, with the pull request against byuflowlab/xfoil_light opened from it. IBLPAN now checks the bound before writing and returns the point unconverged instead of the silent STOP. The three contours that ended the process now return converged=false (real and complex-step library); the other six are bit-identical. Xfoil.jl needs no change; this needs a Yggdrasil rebuild of xfoil_light_jll after merge. #320 is unchanged.

[1-Bort-1]

Local full suite: FAIL (0 min, Julia 1.13.0, one cell of the matrix)

unknown function (ip: (nil)) at (unknown file)
unknown function (ip: 0x7fb3ae75e9ee) at /lib/x86_64-linux-gnu/libc.so.6
unknown function (ip: 0x7fb3ae753667) at /lib/x86_64-linux-gnu/libc.so.6
unknown function (ip: 0x7fb3ae7536ac) at /lib/x86_64-linux-gnu/libc.so.6
epoll_pwait at /lib/x86_64-linux-gnu/libc.so.6 (unknown line)
uv__io_poll at /workspace/srcdir/libuv/src/unix/linux.c:1404
uv_run at /workspace/srcdir/libuv/src/unix/core.c:430
ijl_task_get_next at /cache/build/builder-amdci4-4/julialang/julia-ci/src/scheduler.c:533
wait at ./task.jl:1296
wait_forever at ./task.jl:1196
jfptr_wait_forever_65680.1 at /home/agent/.julia/juliaup/julia-1.13.0+0.x64.linux.gnu/lib/julia/sys.so (unknown line)
jl_apply at /cache/build/builder-amdci4-4/julialang/julia-ci/src/julia.h:2318 [inlined]
start_task at /cache/build/builder-amdci4-4/julialang/julia-ci/src/task.c:1285
unknown function (ip: (nil)) at (unknown file)
Allocations: 7078566 (Pool: 7076302; Big: 2264); GC: 8
2026-09-21T12:25:22Z  KILLED by SIGTERM after 4m
ERROR: The Project.toml of the package being tested must have a name and a UUID entry
Stacktrace:
  [1] pkgerror(msg::String)
    @ Pkg.Types ~/.julia/juliaup/julia-1.13.0+0.x64.linux.gnu/share/julia/stdlib/v1.13/Pkg/src/Types.jl:68
  [2] test(ctx::Pkg.Types.Context, pkgs::Vector{PackageSpec}; coverage::Bool, test_fn::Nothing, julia_args::Cmd, test_args::Cmd, force_latest_compatible_version::Bool, allow_earlier_backwards_compatible_versions::Bool, allow_reresolve::Bool, kwargs::@Kwargs{io::IOContext{IO}})
    @ Pkg.API ~/.julia/juliaup/julia-1.13.0+0.x64.linux.gnu/share/julia/stdlib/v1.13/Pkg/src/API.jl:578
  [3] kwcall(::@NamedTuple{io::IOContext{IO}}, ::typeof(Pkg.API.test), ctx::Pkg.Types.Context, pkgs::Vector{PackageSpec})
    @ Pkg.API ~/.julia/juliaup/julia-1.13.0+0.x64.linux.gnu/share/julia/stdlib/v1.13/Pkg/src/API.jl:562
  [4] test(pkgs::Vector{PackageSpec}; io::IOContext{IO}, kwargs::@Kwargs{})
    @ Pkg.API ~/.julia/juliaup/julia-1.13.0+0.x64.linux.gnu/share/julia/stdlib/v1.13/Pkg/src/API.jl:172
  [5] test(pkgs::Vector{PackageSpec})
    @ Pkg.API ~/.julia/juliaup/julia-1.13.0+0.x64.linux.gnu/share/julia/stdlib/v1.13/Pkg/src/API.jl:161
  [6] test(; name::Nothing, uuid::Nothing, version::Nothing, url::Nothing, rev::Nothing, path::Nothing, mode::PackageMode, subdir::Nothing, kwargs::@Kwargs{})
    @ Pkg.API ~/.julia/juliaup/julia-1.13.0+0.x64.linux.gnu/share/julia/stdlib/v1.13/Pkg/src/API.jl:189
  [7] test()
    @ Pkg.API ~/.julia/juliaup/julia-1.13.0+0.x64.linux.gnu/share/julia/stdlib/v1.13/Pkg/src/API.jl:178
  [8] top-level scope
    @ none:1
  [9] eval(m::Module, e::Any)
    @ Core ./boot.jl:489
 [10] exec_options(opts::Base.JLOptions)
    @ Base ./client.jl:310
 [11] _start()
    @ Base ./client.jl:577