https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84523
Paul Thomas <pault at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Assignee|unassigned at gcc dot gnu.org |pault at gcc dot gnu.org
--- Comment #5 from Paul Thomas <pault at gcc dot gnu.org> ---
The if statement is generating the code:
struct array01_t_spot D.3815;
if ((real(kind=4)[0:] * restrict) ((struct t_spot *)
(rc->spots.data + (sizetype) ((rc->spots.offset + SAVE_EXPR <(integer(kind=8))
k> * rc->spots.dim[0].stride) * D.3815.span)))->vm.data != 0B)
You will see that D.3815's span is being used but is never set.
This temporary is being generated at trans-array.c:3664. It should produce an
assignment of the array descriptor to the temporary, as well as the
declaration.
Somewhere, the se->pre is not being retained so that the assignment disappears.
I am onto it.
Paul
PS
Index: gcc/fortran/trans.c
===================================================================
*** gcc/fortran/trans.c (revision 257937)
--- gcc/fortran/trans.c (working copy)
*************** get_array_span (tree type, tree decl)
*** 358,364 ****
{
if (TREE_CODE (decl) == PARM_DECL)
decl = build_fold_indirect_ref_loc (input_location, decl);
! span = gfc_conv_descriptor_span_get (decl);
}
else
span = NULL_TREE;
--- 358,365 ----
{
if (TREE_CODE (decl) == PARM_DECL)
decl = build_fold_indirect_ref_loc (input_location, decl);
! span = TYPE_SIZE_UNIT (gfc_get_element_type (TREE_TYPE (decl)));
! span = fold_convert (gfc_array_index_type, span);
}
else
span = NULL_TREE;
fixes this bug but produces a number of regressions.
