[flang] Duplicate names for ac-implied-do variables erroneously cause errors

[PeteSteinfeld]

Summary

According to section 19.4, paragraph 5, the scope of an ac-implied-do variable
is the enclosing ac-implied-do. But we were not creating new scopes upon
entry to an ac-implied-do. This was causing error messages to be erroneously
emitted.

I fixed, the code, added a test to array-constr-values.f90, and corrected the
test symbol05.f90.

[klausler]

flang/lib/Semantics/resolve-names.cpp

4999

enclosing

ac-implied-do

flang/test/Semantics/array-constr-values.f90

65

You should probably use the duplicated ac-implied-do names in the body and loop control expressions of the nested implied DO loop, and verify that a correct constant array is created.

[tskeith]

flang/lib/Semantics/resolve-names.cpp

4999

enclosing

[PeteSteinfeld]

flang/lib/Semantics/resolve-names.cpp

4999

Ay-yi-yi. I'll fix these.

flang/test/Semantics/array-constr-values.f90

65

Good idea. I'll create a constant folding test.

[PeteSteinfeld]

I fixed some spelling errors and added a test to verify that the duplicate
ac-implied-do variable names get evaluated correctly in the presence of
constant folding.

[klausler]

flang/test/Evaluate/folding15.f90

8

Can you test using the "outer" iDuplicate as part of an implied DO loop bound expression for the "inner" iDuplicate, please?

[PeteSteinfeld]

flang/test/Evaluate/folding15.f90

8

I'm not sure what you mean. In the two tests I added, the "iDuplicate" variables are defined in parallel scopes. It's not allowed to have an ac-implied-do variable with the same name as one in an outer scope.

Can you give me an example?

[klausler]

flang/test/Evaluate/folding15.f90

8

Actually, I can't, because I think that C7115 implies that the kind of scoping that you're implementing isn't necessary in the first place.

[PeteSteinfeld]

flang/test/Evaluate/folding15.f90

8

Right. My reading of the standard is that C7115 precludes nested scopes of ac-implied-do variables with the same name. But section 19.4, paragraph 5 implies that parallel regions with the same name are allowed. C7115 was already tested in tests/Semantics/array-constr-values.f90, and it still passes with these changes.

flang/lib/Semantics/resolve-names.cpp

 bool ConstructVisitor::Pre(const parser::AcSpec &x) {
   ProcessTypeSpec(x.type);
   PushScope(Scope::Kind::ImpliedDos, nullptr);
   Walk(x.values);
   PopScope();
   return false;
 }
 
+// Section 19.4, paragraph 5 says that each ac-do-variable has the scope of the
+// enclosing ac-implied-do
 bool ConstructVisitor::Pre(const parser::AcImpliedDo &x) {
   auto &values{std::get<std::list<parser::AcValue>>(x.t)};
   auto &control{std::get<parser::AcImpliedDoControl>(x.t)};
   auto &type{std::get<std::optional<parser::IntegerTypeSpec>>(control.t)};
   auto &bounds{std::get<parser::AcImpliedDoControl::Bounds>(control.t)};
+  PushScope(Scope::Kind::ImpliedDos, nullptr);
   DeclareStatementEntity(bounds.name.thing.thing, type);
   Walk(bounds);
   Walk(values);
+  PopScope();
   return false;
 }
 
 bool ConstructVisitor::Pre(const parser::DataImpliedDo &x) {
   auto &objects{std::get<std::list<parser::DataIDoObject>>(x.t)};
   auto &type{std::get<std::optional<parser::IntegerTypeSpec>>(x.t)};
   auto &bounds{std::get<parser::DataImpliedDo::Bounds>(x.t)};
   DeclareStatementEntity(bounds.name.thing.thing, type);

flang/test/Evaluate/folding15.f90

+! RUN: %S/test_folding.sh %s %t %f18
+! Test folding of array constructors with duplicate names for the implied
+! DO variables
+module m1
+  integer, parameter :: expected(12) = [1, 2, 4, 6, 1, 2, 4, 6, 1, 2, 3, 6]
+  integer, parameter :: dups(12) = &
+    [ ((iDuplicate, iDuplicate = 1,j), &
+       (2 * iDuplicate, iDuplicate = j,3 ), &
+        j = 1,3 ) ]
+  logical, parameter :: test_dups = all(dups == expected)
+end module

flang/test/Semantics/array-constr-values.f90

subroutine arrayconstructorvalues()
   !ERROR: ABSTRACT derived type 'base_type' may not be used in a structure constructor
   !ERROR: Values in array constructor must have the same declared type when no explicit type appears
   intarray = (/ base_type(10), 2, 3, 4, 5 /)
 end subroutine arrayconstructorvalues
 subroutine checkC7115()
   real, dimension(10), parameter :: good1 = [(99.9, i = 1, 10)]
   real, dimension(100), parameter :: good2 = [((88.8, i = 1, 10), j = 1, 10)]
   !ERROR: Implied DO index is active in surrounding implied DO loop and may not have the same name
-  !ERROR: 'i' is already declared in this scoping unit
   real, dimension(100), parameter :: bad = [((88.8, i = 1, 10), i = 1, 10)]
 
   !ERROR: The stride of an implied DO loop must not be zero
   integer, parameter :: bad2(*) = [(j, j=1,1,0)]
 end subroutine checkC7115
+subroutine checkOkDuplicates
+  real :: realArray(21) = &
+    [ ((1.0, iDuplicate = 1,j), &
+       (0.0, iDuplicate = j,3 ), &
+        j = 1,5 ) ]
+end subroutine

flang/test/Semantics/symbol05.f90

 !DEF: /s3 (Subroutine) Subprogram
 subroutine s3
  !DEF: /s3/j ObjectEntity INTEGER(8)
  integer(kind=8) j
  block
   !DEF: /s3/Block1/t DerivedType
   type :: t
    !DEF: /s3/Block1/t/x ObjectEntity REAL(4)
-   !DEF: /s3/Block1/t/ImpliedDos1/i (Implicit) ObjectEntity INTEGER(4)
+   !DEF: /s3/Block1/t/ImpliedDos1/ImpliedDos1/i (Implicit) ObjectEntity INTEGER(4)
    real :: x(10) = [(i, i=1,10)]
    !DEF: /s3/Block1/t/y ObjectEntity REAL(4)
-   !DEF: /s3/Block1/t/ImpliedDos2/j ObjectEntity INTEGER(8)
+   !DEF: /s3/Block1/t/ImpliedDos2/ImpliedDos1/j ObjectEntity INTEGER(8)
    real :: y(10) = [(j, j=1,10)]
   end type
  end block
 end subroutine
 
 !DEF: /s4 (Subroutine) Subprogram
 subroutine s4
  implicit integer(x)