[patch,gfortran] PR 19929: Deallocation of pointer components of derived type arrays

[Paul Thomas]

Erik,

:REVIEWMAIL:

>we try to generate code for deallocating an array.  This happens
>because in gfc_trans_deallocate()/trans-stmt.c, we check if the
>object to be deallocated is an array by looking at
>expr->symtree->n.sym->attr.dimension.  This is wrong, because
>expr->symtree->n.sym is the variable 'a' (which is an array),
>not it's component 'p' (which is not an array).  I changed
>the code to look at expr->rank instead.
>  
>
The logic is impeccable and the patch is fine.  I tested it to death and 
it held up fine.

I will commit tomorrow morning, if nobody has any objections.

Paul T