Potential bug in `sedshi.F90` about blom HOT 5 CLOSED

Hi @JorgSchwinger an @TomasTorsvik , ok, yes, I see what you mean. Being explicit about it wouldn't harm, though. I guess, in this case, I will add the change simply to my nitrogen cycle branch which will enter the beyone-CMIP6 branch at some point (or the master, if is has been merged).

from blom.

Hi @jmaerz , I agree that this looks suspicious. Looking at the code history:

https://github.com/NorESMhub/BLOM/blame/master/hamocc/sedshi.F90

most of the sedshi.F90 code has not been changed for a long time, except l. 143 that was included in PR #189.

As it happens, k should take the value ks in this case, since the previous loop terminates at ks-1, but I don't think there is any good reason to rely on this implied behavior.

See discussion in https://stackoverflow.com/questions/62229922/value-of-index-variable-after-loop-exit-in-fortran
and example:

  program main
  implicit none
  integer :: i, n
  n = 10

  do i = 1, n - 1
     print *, "inside: ", i
  enddo

  print *, "after exit: ", i   !! guaranteed to be n?
  end

from blom.

Also, it would be a good idea to update beyond-CMIP6 from master before including the fix.

from blom.

Hi @jmaerz and @TomasTorsvik

It looks like k=ks at this point of the code, right?

In this case, I don't think this classifies as a bug (since it relies on behavior that is part of the Fortran standard). But it is definitely not nice and I agree k should be changed to ks here. There should be no effect on the results (hopefully).

from blom.

Hi @JorgSchwinger an @TomasTorsvik , ok, yes, I see what you mean. Being explicit about it wouldn't harm, though. I guess, in this case, I will add the change simply to my nitrogen cycle branch which will enter the beyone-CMIP6 branch at some point (or the master, if is has been merged).

Seem like a good plan to me :)

from blom.