[MITgcm-devel] [MITgcm-cvs] MITgcm/doc CVS Commit

[Jean-Michel Campin]

Hi Martin (and Ralf),

As pointed by Ralf, the problem is in the modified code (from yesterday)
and not how TAF handle the code (in fact seems that the OpenAD test are failing too).

The forward test that I run everyday with intel compiler (and -devel) do 
check for arguments consistency between S/R declaration and S/R call ;
And these test are now failing to compile (.e.g, on engaging, 5/95 exp
pass compared to 86/95 yesterday, with only few exp that don't compile
pkg/mom_vecinv that are left unaffected).

Martin, can you take a look at this ?

Thanks,
Jean-Michel

On Sat, Apr 29, 2017 at 02:37:27PM +0200, Martin Losch wrote:
> Hi Ralf,
> 
> after a small modification yesterday, taf cannot handle part of the MITgcm vector invariant code anymore and returns a segmentation fault, e.g. here: <http://mitgcm.org/testing/results/2017_04/tr_baudelaire-b_20170429_0/global_ocean.cs32x15/make.tr_log_tail>
> 
> I am citing from the log file:
> > Segmentation fault
> > 
> > -------------------------------------------
> >  something went wrong
> >  Please send email to support at fastopt.com
> > ???????????????????????????????????????????????????????????????
> 
> which I am doing now. In the new version of the code a few fortran integer and logical flags (namelist runtime parameters) are passed as formal parameters of 4 different subroutines; in the old version they were part of a common block, which is no longer needed (and consequently removed from the files: PARAMS.h). The files in question are (all in pkg/mom_vecinv):
> mom_vi_u_coriolis.F
> mom_vi_u_coriolis_c4.F
> mom_vi_v_coriolis.F
> mom_vi_v_coriolis_c4.F
> and mom_vecinv.F (where they are called from).
> A typical difference can be seen here (as you know):
> <http://wwwcvs.mitgcm.org/viewvc/MITgcm/MITgcm/pkg/mom_vecinv/mom_vi_u_coriolis_c4.F?r1=1.14&r2=1.15> 
> 
> You should be able to reproduce this problem with current code like this:
> cvs co MITgcm
> cd MITgcm/verification
> ./testreport -t global_ocean.cs32x15 -devel -adm 
> or 
> ./testreport -t global_ocean.cs32x15 -devel -adm -ncad
> (the former sends one file to taf, the latter multiple files)
> 
> What can I do to fix this?
> 
> Martin
> 
> 
> > On 28 Apr 2017, at 19:23, Martin Losch <mlosch at forge.csail.mit.edu> wrote:
> > 
> > Update of /u/gcmpack/MITgcm/doc
> > In directory forge:/tmp/cvs-serv24070/doc
> > 
> > Modified Files:
> > 	tag-index 
> > Log Message:
> > add todays changes:
> > o pkg/seaice: add momemtum advection, off by default, still in test phase
> >  most of the momentum advection code is adapted and reused from mom_vecinv
> > o pkg/mom_vecinv: pass these runtime flags as formal parameters to 
> >  s/r mom_vi_u/v_coriolis, mom_vi_u/v_coriolis_c4, so that these routines
> >  can also be used in pkg/seaice:
> >  selectVortScheme, highOrderVorticity, upwindVorticity, useJamartMomAdv
> > 
> > 
> > _______________________________________________
> > MITgcm-cvs mailing list
> > MITgcm-cvs at mitgcm.org
> > http://mitgcm.org/mailman/listinfo/mitgcm-cvs
> 
> 
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel