https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115632
Bug ID: 115632
Summary: resolve.cc: FORALL check function is has wrong-code
bugs
Product: gcc
Version: 15.0
Status: UNCONFIRMED
Keywords: wrong-code
Severity: normal
Priority: P3
Component: fortran
Assignee: unassigned at gcc dot gnu.org
Reporter: burnus at gcc dot gnu.org
Target Milestone: ---
The following doesn't make sense on multiple ways:
(A)
-----------------
find_forall_index (gfc_expr *expr, gfc_symbol *sym, int f)
{
if (gfc_traverse_expr (expr, sym, forall_index, f))
-----------------
The function takes as last argument a POINTER; if the return value is not
needed, passing '&f' would do – otherwise, 'int f' should be 'int *f' in the
parameter list.
(B)
-----------------
static bool
forall_index (gfc_expr *expr, gfc_symbol *sym, int *f)
{
-----------------
The cleaner & more explicit solution is:
forall_index (gfc_expr *expr, gfc_symbol *sym, void *f2)
{
int *f = (int *) f2;
(C)
/* A scalar assignment */
if (!expr->ref || *f == 1)
I don't understand the comment about 'assignment'.
(D)
/* .... f == 2 signals that
that variable itself is not to be checked - only the references. */
...
if (*f == 2)
*f = 1;
This one looks completely bogus.
Callers:
resolve.cc - those seem to be fine at a glance:
if (find_forall_index (iter2->start, iter->var->symtree->n.sym, 0)
|| find_forall_index (iter2->end, iter->var->symtree->n.sym, 0)
|| find_forall_index (iter2->stride, iter->var->symtree->n.sym, 0))
if (!find_forall_index (code->expr1, forall_index, 0))
trans-stmt.cc - but this one is really bogus:
new_symtree = NULL;
if (find_forall_index (c->expr1, lsym, 2))
{
forall_make_variable_temp (c, pre, post);
need_temp = 0;
}
The '2' definitely doesn't do what was advertised here …
* * *
Note: the forall index is a 'scalar integer name', i.e. there are not many
possibilies for 'ref->', ignoring, e.g., 'idx%kind' which should be fine.
