[Bug fortran/61831] [4.9/ 5 Regression] runtime error: pointer being freed was not allocated

["dominiq at lps dot ens.fr" <gcc-bugzilla at gcc dot gnu dot org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61831

--- Comment #35 from Dominique d'Humieres <dominiq at lps dot ens.fr> ---
I see three possibilities for this PR:

(1) Revert r211405 for 5.0 and r212329 for 4.9.
(2) Understand why elemental procedures expose the problem when expr->expr_type
== EXPR_ARRAY.
(3) Apply the conservative patch

--- ../4.9_clean/gcc/fortran/trans-expr.c    2014-10-04 12:59:45.000000000
+0200
+++ ../4.9_work/gcc/fortran/trans-expr.c    2014-10-04 13:11:36.000000000 +0200
@@ -6475,7 +6475,7 @@ gfc_conv_expr_reference (gfc_se * se, gf
   if (expr->ts.type == BT_DERIVED && expr->rank
       && !gfc_is_finalizable (expr->ts.u.derived, NULL)
       && expr->ts.u.derived->attr.alloc_comp
-      && expr->expr_type != EXPR_VARIABLE)
+      && expr->expr_type == EXPR_FUNCTION)
     {
       tree tmp;

--- ../4.9_clean/gcc/testsuite/gfortran.dg/class_array_15.f03    2014-07-07
14:33:14.000000000 +0200
+++ ../4.9_work/gcc/testsuite/gfortran.dg/class_array_15.f03    2014-11-08
14:26:21.000000000 +0100
@@ -115,5 +115,5 @@ subroutine pr54992  ! This test remains 
   bh => bhGet(b,instance=2)
   if (loc (b) .ne. loc(bh%hostNode)) call abort
 end
-! { dg-final { scan-tree-dump-times "builtin_free" 12 "original" } }
+! { dg-final { scan-tree-dump-times "builtin_free" 12 "original" { xfail *-*-*
} } }
 ! { dg-final { cleanup-tree-dump "original" } }

I am not fond of (1).

For (2) I need some help. So far the problem has been looked from the 'if'
block point of view. It may be easier to look at the alloc/free machinery
handling elemental.

I think (3) is probably the best solution for 4.9. What I have not done so far
is a test to check this PR, e.g., the test in comment 31, a test for pr41936,
and open a new PR for the memory leak in gfortran.dg/class_array_15.f03.

Feedback welcomed.