[flang-commits] [flang] [flang][OpenMP] Diagnose privatized array section reductions (PR #215997)

[via flang-commits]

================
@@ -5,14 +5,27 @@
 ! RUN: %not_todo_cmd %flang_fc1 -emit-hlfir -fopenmp -fopenmp-version=50 -o - %t/taskloop-in.f90 2>&1 | FileCheck %s --check-prefix=TASKLOOP-IN
 ! RUN: %not_todo_cmd bbc -emit-hlfir -fopenmp -fopenmp-version=50 -o - %t/taskloop-reduction.f90 2>&1 | FileCheck %s --check-prefix=TASKLOOP-REDUCTION
 ! RUN: %not_todo_cmd %flang_fc1 -emit-hlfir -fopenmp -fopenmp-version=50 -o - %t/taskloop-reduction.f90 2>&1 | FileCheck %s --check-prefix=TASKLOOP-REDUCTION
+! RUN: %not_todo_cmd bbc -emit-hlfir -fopenmp -fopenmp-version=50 -o - %t/task-section.f90 2>&1 | FileCheck %s --check-prefix=TASK-SECTION
+! RUN: %not_todo_cmd %flang_fc1 -emit-hlfir -fopenmp -fopenmp-version=50 -o - %t/task-section.f90 2>&1 | FileCheck %s --check-prefix=TASK-SECTION
+! RUN: %not_todo_cmd bbc -emit-hlfir -fopenmp -fopenmp-version=50 -o - %t/taskloop-in-section.f90 2>&1 | FileCheck %s --check-prefix=TASKLOOP-IN-SECTION
+! RUN: %not_todo_cmd %flang_fc1 -emit-hlfir -fopenmp -fopenmp-version=50 -o - %t/taskloop-in-section.f90 2>&1 | FileCheck %s --check-prefix=TASKLOOP-IN-SECTION
+! RUN: %not_todo_cmd bbc -emit-hlfir -fopenmp -fopenmp-version=50 -o - %t/taskloop-reduction-section.f90 2>&1 | FileCheck %s --check-prefix=TASKLOOP-REDUCTION-SECTION
+! RUN: %not_todo_cmd %flang_fc1 -emit-hlfir -fopenmp -fopenmp-version=50 -o - %t/taskloop-reduction-section.f90 2>&1 | FileCheck %s --check-prefix=TASKLOOP-REDUCTION-SECTION
+! RUN: bbc -emit-hlfir -fopenmp -fopenmp-version=50 --enable-delayed-privatization=false -o %t/eager-in-bbc.mlir %t/taskloop-in.f90
----------------
MattPD wrote:

The eager array-element tests now cover the original regression.

The eager array-section tests still assert only successful HLFIR generation. A `reduction(+: a(2:3))` case that also reads `a(1)` passes even though the firstprivate clone is dead, and `a(1)` comes from the zero-initialized reduction box.

Could the eager array-section cases either expect the TODO diagnostic or verify storage selection with an unlisted element?

https://github.com/llvm/llvm-project/pull/215997