Re: mesa-users] Missing a tmp.txt file in the kap module?

[Bill P. <pa...@ki...>]

On Jun 25, 2010, at 1:46 PM, Eric Blais wrote:

> I'm using ifort.  Version 11.1.038.

well this is looking like a compiler problem, so I'll ask you to update to the most recent ifort 11.1 which I believe is 088.

the output that makes me think this is the compiler comes from these lines in load_co_kap

            allocate(co_tables(num_tables), STAT=status)
            if (status .ne. 0) then
               ierr = -1
               call alert(ierr, 'InsufficientMemory for Prepare_Kap_CO_X_Table')
               if (CO_dbg) write(*,*) 'InsufficientMemory for Prepare_Kap_CO_X_Table', iz, ix
               return
            end if
            x_tables(ix)% co_tables => co_tables
            if (CO_dbg) write(*,*) 'Setup_Kap_CO_X_Table: allocate co_tables', iz, ix, num_tables, status
            if (CO_dbg) write(*,*) 'Setup_Kap_CO_X_Table: associated(x_tables(ix)% co_tables)', &
                           associated(x_tables(ix)% co_tables), associated(co_tables), ix


Are you familiar with pointers in fortran95?   

allocate(co_tables(num_tables), STAT=status) allocates storage and sets the pointer co_tables
the status == 0 means the allocation was okay
x_tables(ix)% co_tables => co_tables copies the pointer so that x_tables(ix)% co_tables points to the same thing as co_tables
associated(p) is true if the pointer p points to an allocated structure.

in this case x_tables(ix)% co_tables and co_tables should both give true for associated, 
and that is in fact what I get when I run it 

but on your machine we get this:
 Setup_Kap_CO_X_Table: associated(x_tables(ix)% co_tables) F T           4

so associated(x_tables(ix)% co_tables) is false -- that's bad and very bogus.


hopefully a new ifort will do better.    let me know.

Thanks,
Bill