[flang-commits] [flang] [Flang][Semantics] Allow EVENT_TYPE, LOCK_TYPE and NOTIFY TYPE to be deallocate (PR #192940)

[Jean-Didier PAILLEUX via flang-commits]

================
@@ -0,0 +1,55 @@
+! RUN: %python %S/test_errors.py %s %flang_fc1
+! Check there are no semantic errors in DEALLOCATE statements with EVENT_TYPE and LOCK_TYPE.
+
+! CHECK-NOT: Object in DEALLOCATE statement is not deallocatable
+
+module not_iso_fortran_env
+  type event_type
+  end type
+  type lock_type
+  end type
+end module
+
+subroutine deallocate_lock_event_type()
+  use iso_fortran_env
+
+  type oktype1
+    type(event_type), pointer :: event
+    type(lock_type), pointer :: lock
+  end type
+
+  type oktype5
+    type(event_type), allocatable :: event
+  end type
+
+  type oktype2
+    type(event_type) :: event
+  end type
+
+  type oktype3
+    type(lock_type), allocatable :: lock
+  end type
+
+  type oktype4
+    type(lock_type) :: lock
+  end type
+
+  ! Variable with event_type or lock_type have to be coarrays
+  ! see C1604 and 1608.
+  type(oktype1), allocatable :: okt1[:]
+  type(oktype2), allocatable :: okt2[:]
+  type(oktype3), allocatable :: okt3[:]
+  type(oktype5), allocatable :: okt5[:]
+  class(oktype4), allocatable :: okt4[:]
+  type(event_type), allocatable :: event[:]
+  type(lock_type), allocatable :: lock(:)[:]
+
+
+  deallocate(okt1)
+  deallocate(okt2)
+  deallocate(okt3)
+  deallocate(okt4)
+  deallocate(okt5)
+  deallocate(lock)
+  deallocate(event)
----------------
JDPailleux wrote:

This test is just "Semantics" validation. I don't think it's really necessary to add the `allocate` calls here. In that case, we can just add them in the test of `allocate` for these types. (ex: allocate13.f90)

https://github.com/llvm/llvm-project/pull/192940