Erroneous use of surface geopotential in co2condens?

[emillour]

In co2condens (phymars/co2condens_mod.F, cuurent revision r2685) there is what looks like a computation of a heating term due to latent heat for CO2 snowfall sublimation in the first atmospheric layer which involves pphi(ig,1) the geopotential of the first layer and the surface geopotential phisfi(ig) :

            IF (zfallice(ig,1).gt.0 .and. .not. co2clouds) then
              zfallheat = zfallice(ig,1) * (pphi(ig,1)- phisfi(ig) +
     &                    cpice*(ztcond(ig,1)-ztcondsol(ig)))/latcond
            ELSE

This doesn't look right: pphi(ig,1) is the geopotential relative to the surface whereas phisfi(ig) is the surface geopotential (i.e. relative to aeroid)... should'nt only pphi(ig,1) be used? But then what about the (ztcond(ig,1)-ztcondsol(ig)) term?

The use of pphi(ig,1)- phisfi(ig) dates back to at least r38 (in 2011, at the time the routine was called newcondens.F)

Fixed with r2739. Tests moreover showed that (fortunately) this bug had little impact on the simulations.