[MITgcm-devel] testing on stan1

[Jean-Michel Campin]

Hi Martin,

I will go back to previous genmake2 version, but keeping the addition 
of OAD_regular_cp.F90
And we can figure out later how to avoid (or limit to just OpenAD built)
this "export" instruction within the Makefile.

Cheers,
Jean-Michel

On Tue, Sep 01, 2015 at 04:16:36PM +0200, Martin Losch wrote:
> Hi,
> 
> I played a little with syntax. The problem is obviously that the OS on iblade (AIX) does not support ???export??? in a makefile in the same way as e.g. gnu-make.
> 
> SOMEVAR = ???bla??? (or SOMEVAR =)
> export SOMEVAR
> 
> does not work, and clearly just ???export??? does not work either (this caused the error my particular case), but 
> 
> export SOMEVAR:=???bla??? (or export SOMEVAR := )
> 
> is not a problem. Can the code be rewritten into that format?
> 
> Or: Since iblade is an old machine which will probably go offline by the end of the year, I can try to make my script use some gnu-make.
> What should we do?
> 
> Martin
> 
> 
> > On 01 Sep 2015, at 16:10, Jean-Michel Campin <jmc at ocean.mit.edu> wrote:
> > 
> > Hi Martin and Patrick,
> > 
> > This Jul 03 genmake2 modification does not seem very robust.
> > Should we revert to previous version of genmake2 (without OPENAD_DIVA
> > setting) until we come up with something better ?
> > 
> > also, I am concerned that this new (& undocumented) "-diva" option 
> > (specific to OPENAD) might be confused with TAF-diva way.
> > 
> > Cheers,
> > Jean-Michel
> > 
> > On Tue, Sep 01, 2015 at 10:15:49AM +0200, Martin Losch wrote:
> >> Hi Jean-Michel,
> >> 
> >> I am moving this to the devel list, because the reason for the fail is this error message:
> >> 
> >> "Makefile", line 63: make: 1254-055 Dependency line needs colon or double colon operator.
> >> make: 1254-058 Fatal errors encountered -- cannot continue.
> >> 
> >> the corresponding line is this
> >> 61 # Link editor
> >> 62 LINK = mpxlf95_r -q64
> >> 63 export
> >> 64
> >> 65 # Defines for CPP
> >> 
> >> and on July 03 this line has been added to genmake2 (together with others):
> >> 2897	 	 export $OPENAD_DIVA
> >> (see <http://mitgcm.org/viewvc/MITgcm/MITgcm/tools/genmake2?r1=1.271&r2=1.272> for more details)
> >> OPENAD_DIVA is empty in my case.
> >> 
> >> Not clear to me, how to fix this. I didn???t know you could use ???export??? in a Makefile. Is it really needed?
> >> 
> >> Martin
> >> 
> >>> On 30 Aug 2015, at 19:48, Jean-Michel Campin <jmc at ocean.mit.edu> wrote:
> >>> 
> >>> Hi Martin,
> >>> 
> >>> There are also some problems with testing on iblade that I should have reported
> >>> earlier: Since July 04, all the experiments on iblade are failing (to compile ?).
> >>> 
> >>> Cheers,
> >>> Jean-Michel
> >>> 
> >>> On Sat, Aug 15, 2015 at 12:00:22PM -0400, Jean-Michel Campin wrote:
> >>>> Hi Martin,
> >>>> 
> >>>> How are you ?
> >>>> 
> >>>> Regarding the testing on stan1, I've noticed that, after one day (Aug 12) with everal fails,
> >>>> we iare back to normal since AUg 13, except for the (standard) experiment: adjustment.cs-32x32x1
> >>>> (not the nlfs secondary test which always fails with MPI) which continue to fail.
> >>>> 
> >>>> I now have seen 3 - almost identical - testreport output (Aug 13,14 & 15) 
> >>>> where adjustment.cs-32x32x1 failed (at compilation stage ?) so may be there is
> >>>> something not cleaned up there ?
> >>>> 
> >>>> It's not urgent, but when you have time to take a look at this.
> >>>> 
> >>>> Cheers,
> >>>> Jean-Michel
> >> 
> >> 
> >> _______________________________________________
> >> MITgcm-devel mailing list
> >> MITgcm-devel at mitgcm.org
> >> http://mitgcm.org/mailman/listinfo/mitgcm-devel
> > 
> > _______________________________________________
> > MITgcm-devel mailing list
> > MITgcm-devel at mitgcm.org
> > http://mitgcm.org/mailman/listinfo/mitgcm-devel
> 
> 
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel