Cavity Ghia validation failing (L2=0.74, Linf=1.49 at Re=100)

[benzwick]

Problem

The lid-driven cavity validation against Ghia et al. (1982) at Re=100 is failing in CI with large errors:

L2 error:   0.742539
Linf error: 1.494235
L2 < 5%: FAIL
Linf < 10%: FAIL

The ASCII plot shows the computed solution is shifted significantly from the reference data, suggesting the BCs or the solver may not be converging to the correct steady state.

Affected CI

• Run Examples → Demo: cavity → Validate against Ghia
• Validation Tests workflow

Possible Causes

• Insufficient simulation time to reach steady state
• Grid resolution too coarse (40×40 with 25 neighbors)
• Numerical issues with the explicit Euler time stepping
• Incorrect boundary vorticity computation

Parameters

From CI: -nx 40 -ny 40 -Re 100 -dt 0.001 -ft 30.0

[benzwick]

This issue was observed in the Run Examples workflow, not the Validation Tests workflow. The Validation Tests (which use CLI mode) pass with L2=2%, Linf=4%.

The Run Examples failure was caused by two separate bugs, both now fixed:

1. MFEM boundary attributes never applied (commit 7546159) — SDL attribute numbers weren't set on the MFEM mesh, causing wrong node-to-BC mapping
2. Validation filename mismatch (commit 65cc2d3) — validation looked for _u_centerline_x0.5.dat but SDL probes produce _vertical-center.dat

The Run Examples cavity demo should pass on the next CI run. If it still fails, reopen.