[MITgcm-devel] bug with #undef SEAICE_ALLOW_DYNAMICS and JFNK

[Martin Losch]

Hi,

sorry, I was on vacation, and obviously screwed up my vacation message, so that it was only sent to addresses in my address-book (which appears to be empty).

I agree that the CPP-Flag business it not very nice and inconsistent with the JFNK solver. I will change and simplify the related CPP flags according to Jean-Michel's suggestion shortly, (but maybe not today).

Martin

On Mar 28, 2013, at 12:06 AM, Jean-Michel Campin <jmc at ocean.mit.edu> wrote:

> Hi Martin and Torge,
> 
> I think the simpler (and less prone to error/mistake)
> is to always compile seaice_fgmres.F as soon as 
> SEAICE_ALLOW_JFNK is defined.
> 
> If one wants to exclude the compilation of this file 
> when SEAICE_ALLOW_DYNAMCIS is #undef, then we can come up with
> something in the default SEAICE_OPTIONS.h
> 
> I also think that it's a general rule: if we can avoid complicated 
> logic with CPP options in the source code, it's safer.
> And moving these conditional setting into the corresponding 
> {PKG}_OPTIONS.h makes it more clear (instead of having to search 
> through all the files).
> 
> Cheers,
> Jean-Michel
> 
> On Wed, Mar 27, 2013 at 02:44:07PM -0700, Torge Martin wrote:
>> Hi,
>> 
>> while compiling my 1-D ITD test scenario on acesgrid I ran into the
>> following error:
>> 
>> in SEAICE_OPTIONS
>> I disabled SEAICE_ALLOW_DYNAMCIS (since I am only running a 1-D scenario)
>> but I kept SEAICE_CGRID enabled (because this is the default setting)
>> which then by default also enables SEAICE_ALLOW_JFNK.
>> 
>> However, all three of these must be enabled so that routine
>> SEAICE_MAP_RS2VEC is included from seaice_fgmres.F.
>> If I disable dynamics (but have JFNK enabled) SEAICE_MAP_RS2VEC is called
>> in SEAICE_INIT_VARIA although it was not included due to the preprocessor
>> option setting.
>> 
>> I guess I have unintentionally created an unlikely scenario but this should
>> be caught somewhere.
>> 
>> Thanks,
>> Torge
> 
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel