[MITgcm-devel] offline_exf_seaice experiment

[Jean-Michel Campin]

Hi Patrick,

I am going to check-in a new version of
offline_exf_seaice/code/CPP_OPTIONS.h
without #include "PACKAGES_CONFIG.h" (but all the other options unchanged).
I don't want to see this CPP_OPTIONS.h being copied elsewhere
and with #include "PACKAGES_CONFIG.h" starting to multiply (was
hard enough to get rid of it).
Also presently, it's shows a lot of differences when doing a diff
with model/inc/CPP_OPTIONS.h (was also the case when originally checked in)
so will go for a more standard, up-to date version.

I would also prefer to have #undef EXF_VERBOSE
and to update offline_exf_seaice/results/output.txt & output.seaicetd.txt
since it's painful to search in those file with tons of
exf_GetFFieldsRec: messages.
It's also make those 2 output files more than 10 times bigger
than the average output.txt file from any other experiment.

And the last thing (but not going to change it now, until every one agree)
is the seaice_monitor.F and exf_monitor.F things:
#ifdef ALLOW_BULK_OFFLINE
c-- Trick monitor output for testreport
C-- by writing some statistics into dynstat fields
c-- which are tested by testreport

If someone thinks that something is not right in testreport & results
comparison, let's change testreport.
But here, writing the area as cg2d_init_res output, or the area again
written as dynstat_theta (which is part of the standard monitor output,
with meaningful values) is not the right way to go, I think.
And yes, I prefer not to see any "cg2d" output in the testreport
summary if the 2-D solver is not called (at least, it's telling that
it's an offline experiment), rather than something that pretend to
be a cg2d output but is not.

And finally, the reason I would vote for changing back seaice_monitor.F
and exf_monitor.F (even before changing testreport),
is that 1) it make the MONITOR output useless (e.g.:,
> grep dynstat_theta_mean offline_exf_seaice/results/output.seaicetd.txt )
2) there are other offline experiments (and passive tracer advection
test) which have been there for long time and do not have fake output,
and consequently, it's even more confusing (because unexpected).

Cheers,
Jean-Michel