xm_periodic/ym_periodic: loneeqn halo copy uses loop variable after the loop (out-of-range plane)

[mvreeuwijk]

Found by the #330 audit, still present on master.

In xm_periodic (src/modboundary.f90; same pattern in ym_periodic), the one-equation-model halo copy sits AFTER the do m = 1, ih loop ends but still indexes with m:

do m = 1, ih
   u0(ib - m, :, :) = u0(ie + 1 - m, :, :)
   ...
end do

if (loneeqn) then
   e120(ib - m, :, :) = e120(ie + 1 - m, :, :)   ! m == ih + 1 here
   ...
end if

After the loop Fortran leaves m = ih + 1, so e120/e12m copy a single plane one past the intended halo range instead of planes 1..ih — wrong periodic BCs for the one-equation SGS model and an out-of-bounds access when the halo width is ih (with ih = 1 it touches ib-2/ie+2). Fix: move the loneeqn block inside the loop.

🤖 Generated with Claude Code