[Patch, fortran] PR122109 - gfortran 16 rejects component declaration in composite derived type with kind parameter

[Paul Richard Thomas]

Hi All,

The comment in the patch and the ChangeLog explains this one and a
half liner adequately. The irony is that the error test was removed
because an error return was being called, which caused failures in
valid cases.

Regtests on FC42/x86_64. Variants of the testcase, mentioned in the
PR, have been tested. OK for mainline?

Paul

Change.Logs
Description: Binary data

diff --git a/gcc/fortran/decl.cc b/gcc/fortran/decl.cc
index f00f0e11378..3761b6589e8 100644
--- a/gcc/fortran/decl.cc
+++ b/gcc/fortran/decl.cc
@@ -4038,7 +4038,15 @@ gfc_get_pdt_instance (gfc_actual_arglist *param_list, gfc_symbol **sym,
 	}
 
       kind_value = 0;
-      gfc_extract_int (kind_expr, &kind_value);
+      /* This can come about during the parsing of nested pdt_templates. An
+	 error arises because the KIND parameter expression has not been
+	 provided. Use the template instead of an incorrect instance.  */
+      if (gfc_extract_int (kind_expr, &kind_value))
+	{
+	  gfc_free_actual_arglist (type_param_spec_list);
+	  return MATCH_YES;
+	}
+
       sprintf (name + strlen (name), "_%d", kind_value);
 
       if (!name_seen && actual_param)
diff --git a/gcc/testsuite/gfortran.dg/pdt_52.f03 b/gcc/testsuite/gfortran.dg/pdt_52.f03
new file mode 100644
index 00000000000..40591bffc48
--- /dev/null
+++ b/gcc/testsuite/gfortran.dg/pdt_52.f03
@@ -0,0 +1,33 @@
+! { dg-do compile }
+!
+! Test the fix for PR122089 in which an error occured in compiling the module
+! because a spurios REAL(KIND=0) was being produced for 'values_'.
+!
+! Contributed by Damian Rouson  <damian@archaeologic.codes>
+!
+module tensor_m
+  implicit none
+
+  type tensor_t(k)
+    integer, kind :: k = kind(1.)
+    real(k), allocatable :: values_ 
+  end type
+
+  type input_output_pair_t(k)
+    integer, kind :: k
+    type(tensor_t(k)) inputs_, expected_outputs_
+  end type
+
+  type mini_batch_t(k)
+    integer, kind :: k
+    type(input_output_pair_t(k)) input_output_pairs_
+  end type
+
+end module tensor_m
+
+  use tensor_m
+  type (mini_batch_t(k = kind(1d0))) :: x
+  allocate (x%input_output_pairs_%inputs_%values_, source = 42d0)
+  print *, kind (x%input_output_pairs_%inputs_%values_), x%input_output_pairs_%inputs_%values_
+  deallocate (x%input_output_pairs_%inputs_%values_)
+end