[flang][OpenMP] Avoid analyzing assumed-size array bases

[kparzysz]

A check for character substrings masquerading as array sections was using expression analyzer on the array base. When this array happened to be an assumed-size array, the analyzer emitted a semantic error that did not correspond to any issue with the source code.

To avoid that, check whether the object is an assumed-size array before using the expression analyzer on it.

While at it, replace the call to GetShape with a simple check for rank, since that's the only information needed.

Fixes #150297

[llvmbot]

@llvm/pr-subscribers-flang-openmp

@llvm/pr-subscribers-flang-semantics

Author: Krzysztof Parzyszek (kparzysz)

Changes

A check for character substrings masquerading as array sections was using expression analyzer on the array base. When this array happened to be an assumed-size array, the analyzer emitted a semantic error that did not correspond to any issue with the source code.

To avoid that, check whether the object is an assumed-size array before using the expression analyzer on it.

While at it, replace the call to GetShape with a simple check for rank, since that's the only information needed.

Fixes #150297

---

Full diff: https://github.com/llvm/llvm-project/pull/150324.diff

2 Files Affected:

• (modified) flang/lib/Semantics/check-omp-structure.cpp (+20-16)
• (added) flang/test/Semantics/OpenMP/assumed-size-array.f90 (+25)

diff --git a/flang/lib/Semantics/check-omp-structure.cpp b/flang/lib/Semantics/check-omp-structure.cpp
index 8264e1d5e8fd9..3d22f0a5b9d0d 100644
--- a/flang/lib/Semantics/check-omp-structure.cpp
+++ b/flang/lib/Semantics/check-omp-structure.cpp
@@ -16,7 +16,6 @@
 #include "flang/Common/idioms.h"
 #include "flang/Common/indirection.h"
 #include "flang/Common/visit.h"
-#include "flang/Evaluate/shape.h"
 #include "flang/Evaluate/tools.h"
 #include "flang/Evaluate/type.h"
 #include "flang/Parser/char-block.h"
@@ -4151,21 +4150,26 @@ void OmpStructureChecker::CheckArraySection(
   // Detect this by looking for array accesses on character variables which are
   // not arrays.
   bool isSubstring{false};
-  evaluate::ExpressionAnalyzer ea{context_};
-  if (MaybeExpr expr = ea.Analyze(arrayElement.base)) {
-    std::optional<evaluate::Shape> shape = evaluate::GetShape(expr);
-    // Not an array: rank 0
-    if (shape && shape->size() == 0) {
-      if (std::optional<evaluate::DynamicType> type = expr->GetType()) {
-        if (type->category() == evaluate::TypeCategory::Character) {
-          // Substrings are explicitly denied by the standard [6.0:163:9-11].
-          // This is supported as an extension. This restriction was added in
-          // OpenMP 5.2.
-          isSubstring = true;
-          context_.Say(GetContext().clauseSource,
-              "The use of substrings in OpenMP argument lists has been disallowed since OpenMP 5.2."_port_en_US);
-        } else {
-          llvm_unreachable("Array indexing on a variable that isn't an array");
+  // Cannot analyze a base of an assumed-size array on its own. If we know
+  // this is an array (assumed-size or not) we can ignore it, since we're
+  // looking for strings.
+  if (!IsAssumedSizeArray(*name.symbol)) {
+    evaluate::ExpressionAnalyzer ea{context_};
+    if (MaybeExpr expr = ea.Analyze(arrayElement.base)) {
+      if (expr->Rank() == 0) {
+        // Not an array: rank 0
+        if (std::optional<evaluate::DynamicType> type = expr->GetType()) {
+          if (type->category() == evaluate::TypeCategory::Character) {
+            // Substrings are explicitly denied by the standard [6.0:163:9-11].
+            // This is supported as an extension. This restriction was added in
+            // OpenMP 5.2.
+            isSubstring = true;
+            context_.Say(GetContext().clauseSource,
+                "The use of substrings in OpenMP argument lists has been disallowed since OpenMP 5.2."_port_en_US);
+          } else {
+            llvm_unreachable(
+                "Array indexing on a variable that isn't an array");
+          }
         }
       }
     }
diff --git a/flang/test/Semantics/OpenMP/assumed-size-array.f90 b/flang/test/Semantics/OpenMP/assumed-size-array.f90
new file mode 100644
index 0000000000000..6e36db178dc8e
--- /dev/null
+++ b/flang/test/Semantics/OpenMP/assumed-size-array.f90
@@ -0,0 +1,25 @@
+!RUN: %flang_fc1 -emit-hlfir -fopenmp %s -o - | FileCheck %s
+
+! This should compile without errors. Check for a symptom of a reasonable
+! output.
+
+!CHECK: omp.task depend
+
+subroutine omp_task_depend_reproducer(work, myid, shift)
+  implicit none
+  integer, intent(in) :: myid, shift
+  real, intent(inout) :: work(*)
+
+!$omp parallel shared(work, myid, shift)
+  !$omp single
+    !$omp task depend(in:work(myid+shift-1)) depend(in:work(myid-1)) depend(out:work(myid))
+      call dummy_kernel(work(myid))
+    !$omp end task
+  !$omp end single
+!$omp end parallel
+contains
+  subroutine dummy_kernel(x)
+    real :: x
+    x = x + 1.0
+  end subroutine dummy_kernel
+end subroutine omp_task_depend_reproducer

[tblah]

Thank you for the quick fix!

Please could you backport this to LLVM 21. There is a guide here

[kparzysz]

/cherry-pick 8fcbd06

[llvmbot]

/pull-request #150411