[Mitgcm-support] possible problem in gmredi

[mitgcm-support at dev.mitgcm.org]

Alistair, there may be a potential problem in c29 if GM_VISBECK_VARIABLE_K is
defined, GM_taper_scheme.EQ.'ldd97' and dSigmaDrReal.gt.0 (unstable).
dRdSigmaLtd from gmredi_slope_limit will then be positive, in which case

         N2=(-Gravity*recip_Rhonil)/dRdSigmaLtd(i,j)
         SN=sqrt(Ssq*N2)

will blow up (actually it did blow up in the JPL implementation after removing
slope clipping).  What's the right way to deal with this?  I reintroduced
clipping with stratLimit=-Small_Number to keep dRdSigmaLtd negative.

Dimitris