[Patch, fortran] PR83118 - [8/9/10/11 Regression] Bad intrinsic assignment of class(*) array component of derived type

[Thomas Koenig]

First of all, merry Christmas everybody!

Regarding the patch:

> Regtests on FC33/x86_64 and the testcase is fixed with -m32. OK for master?

It's OK by me.

There is just one piece of code that I don't understand:

+      if (expr2->expr_type == EXPR_FUNCTION
+	  && expr2->ts.type == BT_DERIVED
+	  && expr2->ts.u.derived->attr.alloc_comp)
+	{
+	  tree tmp2 = gfc_deallocate_alloc_comp (expr2->ts.u.derived,
+						 rse.expr, expr2->rank);
+	  if (lss == gfc_ss_terminator)
+	    gfc_add_expr_to_block (&rse.post, tmp2);
+	  else
+	    gfc_add_expr_to_block (&loop.post, tmp2);
+	}
+    }

This part appears to do the right thing, but I do have one request.

Could you maybe (for my personal education, and that of others who read
this code) add a comment what "lss = gfc_ss_terminator" tests, and why
it that one or the other post block is selected for adding the
deallocation?  I simply don't know what this idiom means.

Thanks for the patch!

Best regards

	Thomas



> Paul
> 
> Fix failures with -m32 and some memory leaks.
> 
> 2020-12-23  Paul Thomas  <pault@gcc.gnu.org <mailto:pault@gcc.gnu.org>>
> 
> gcc/fortran
> PR fortran/83118
> * trans-array.c (gfc_alloc_allocatable_for_assignment): Make
> sure that class expressions are captured for dummy arguments by
> use of gfc_get_class_from_gfc_expr otherwise the wrong vptr is
> used.
> * trans-expr.c (gfc_get_class_from_gfc_expr): New function.
> (gfc_get_class_from_expr): If a constant expression is
> encountered, return NULL_TREE;
> (gfc_trans_assignment_1): Deallocate rhs allocatable components
> after passing derived type function results to class lhs.
> * trans.h : Add prototype for gfc_get_class_from_gfc_expr.
>