The example directories and master templates were reviewed for stale configuration, and the channel and duct campaign cases were prepared for the cluster verification campaign. plane_spectrum and line_spectrum are now supported, on a measurement that matched a direct DFT of a known checkpoint to round-off. Several fixes change results or configuration.
- Every shipped channel and duct grid, and the driven-periodic smoke fixture, used an odd cell count, so its multigrid coarsening was misaligned. They now use 16, 32, 64, 128 or 256 cells and declare mg_levels, which makes grid.gen refuse a count that cannot coarsen.
- The msd_final study metric read the last column of the MSD table (frac_3sigma_pct) instead of MSD_total. Explicit log_regex metrics now pass validation and must capture their value in a group.
- The wall spectral seeds (channel_spectral_velocity, duct_spectral_velocity) now grow linearly from the wall instead of as the fourth power, which puts perturbation energy into the buffer layer. The same seed gives a different field.
- Generated initial conditions and inlet profiles refuse output_file, summary_json and spectrum_csv instead of ignoring them. validate warns that source_data.directory is ignored instead of checking that it exists; omitting source_data keeps the recipe ID.
- Spectra on a programmatic_c case are refused at validation instead of failing after the solve, and picurv init no longer drops a post recipe that has no pipeline.
- The restart guides no longer tell users to set eulerian_field_source: load to continue a run; that mode replays stored fields and never solves.
- New les_wallmodel_retau1000 driven-channel case and humphrey_bend_wm.cfg bend grid; the driven channel and duct campaign cases seed with the wall spectral providers.