ATMET.com

[TROICA]

Hi Craig,

We used RAMS 6.0/HYPACT 1.5 in the project TROICA. The program stops in an error (segmentation fault), in forum we read that other person have same problem.

1. The problem appears in line
…..
hmet_update(npts,metc(ng)%v(1,1,1),metp(ng)%v(1,1,1),metf(ng)%v(1,1,1),wtp)
(hyp_gridavg () in hrdanal.f90)
The variable metp(ng)%v(k,i,j) give the error.

If cross error back, easy to find that this subroutine call from
…..
if (mod(hytime,freqavg) <dtpart>1) ,
you several time call (in section ‘future in past’)
……
call hyp_rdanal (metf(ng),ifutfl(ng),ng)
(hrdriv.f90)
……..
which several time call deallocate memory
……
call dealloc_met(met)


In this situation the program simply clean memory, so local variable metp(ng) is out of visible.

I think, that line (in section ‘future in past’)
…..
metp(ng)=metf(ng)
(hrdriv.f90)
give only pointer to array metp, not allocate real place in memory. It’s true, or isn’t?

I found in hrdanal.f90 allocated for metc.
The allocate memory for metf give subroutine hyp_rdanal (metf(ng),ifutfl(ng),ng).
But I don’t found any allocated for metp.
How fix this problem?

Thank you in advance,
Project TROICA

[tremback]

Sorry, I'm not quite following the explanation. metp is first allocated in hyp_setup, then it should never be deallocated. In hdriv.f90 (about lines 120-140), metf is assigned to metp (that is a copy of the entire structure). hyp_rdanal is called with metf, which will deallocate it, then allocate it again.

Is this crashing after a long period of simulation? It's possible that the system is not reusing the deallocated memory, so the allocations might stop working after a while.
_________________
- Craig