[logan-nc]

Not urgent. Julia work still has priority; review whenever convenient.

Patch release on top of v1.6.0, following the usual hotfix workflow (branch from master; after approval, merge to master, tag v1.6.1, then merge into develop).

Fixes

• Clebsch displacement units (Fixes Correct units for displacements in netcdf #243). xigradpsi, xigradpsi_dpsi and xigradalpha in gpec_profile_output_n*.nc were labeled m. They are ξ·∇ψ_N and ξ·∇α divided by χ₁, so they have no units. They are now labeled unitless, the spelling already used elsewhere in gpout.f. Only the attribute changes.
• Jbgradpsi_rmatch was empty in most runs. gpout_xbnormal filled it only when jac_out == jac_type and tmag_out /= 0. In every other run it was written without ever being set; this build happened to leave it zero. It is now converted to the output coordinates the same way Jbgradpsi is. I checked the units of Jbgradpsi, Jbgradpsi_pest and Jbgradpsi_rmatch: all three are area-normalized, so Tesla is correct and was left unchanged.
• Version bump: release = '1.6.1' in docs/conf.py, and a v1.6.1 section in docs/releases.rst.

Verification

regression_solovev_resistive_example (rdcon → rmatch → gpec). I built v1.6.0 and this branch with identical gfortran settings and compared every netCDF output variable by variable:

variant	branch taken	data diffs other than Jbgradpsi_rmatch	Jbgradpsi_rmatch in v1.6.0	Jbgradpsi_rmatch here
as shipped (jac_out="pest", tmag_out=1)	unchanged branch	none	‖rmatch − Jbgradpsi‖/‖Jbgradpsi‖ = 2.4e-14	bit-identical to v1.6.0
jac_out="hamada"	fixed branch	none	all zeros	finite; ‖rmatch − Jbgradpsi‖/‖Jbgradpsi‖ = 2.3e-14
tmag_out=0	fixed branch	none	all zeros	finite; ‖rmatch − Jbgradpsi‖/‖Jbgradpsi‖ = 2.2e-14

• In all three variants, the only differences in attributes are the three units changes. I did not compare global attributes, which include the version string.
• In this example the RMATCH solution matches the ideal one to round-off. So the new output reproducing Jbgradpsi to about 1e-14, the same agreement as on the unchanged branch, is the expected result. The arrays are not bit-identical, which shows they come from the RMATCH data, not from a copy of Jbgradpsi.
• CI (make_test.yml, started by hand because PRs to master don't trigger it): passed with gfortran and ifort, https://github.com/PrincetonUniversity/GPEC/actions/runs/42634301501

🤖 Generated with Claude Code

https://claude.ai/code/session_01VMDGyfQKRZQMdk3NJXsKQE

[logan-nc]

@matt-pharr this is just a few lines to check - lets get it in so we don't forget

[matt-pharr]

Looks good. Approving and releasing