[Bug fortran/40440] [4.4/4.5 Regression] Garbage or segmentation fault in allocatable array derived type structures

["burnus at gcc dot gnu dot org" <gcc-bugzilla at gcc dot gnu dot org>]

------- Comment #4 from burnus at gcc dot gnu dot org  2009-06-15 12:16 -------
Note: syntax_get_rule_ptr is defined as:

  function syntax_get_rule_ptr (syntax, key) result (rule)
    type(syntax_rule_t), pointer :: rule
    type(syntax_t), intent(in), target :: syntax
    type(string_t), intent(in) :: key

and "syntax_rule_set_sub" is defined as follows where "sub" is the relevant
argument:

  subroutine syntax_rule_set_sub (rule, i, sub)
    type(syntax_rule_t), intent(inout) :: rule
    integer, intent(in) :: i
    type(syntax_rule_t), intent(in), target :: sub

If one looks at the dump for the following line in set_rule_contents

            call syntax_rule_set_sub (rule, i, syntax_get_rule_ptr (syntax, &
                  lexeme_get_contents (lexeme(i+3))))

one finds:

  integer(kind=8) S.297;
  struct varying_string D.6550;
  D.6550 = lexeme_get_contents (&(*lexeme)[(integer(kind=8)) (i + 3) + -1]);
  syntax_rule_set_sub ((struct syntax_rule_t *) rule, &i,
         syntax_get_rule_ptr ((struct syntax_t *) syntax, &D.6550));

which looks OK. But then it continues:

  if (syntax_get_rule_ptr ((struct syntax_t *) syntax, 
                           &D.6550)->keyword.chars.data != 0B)
    __builtin_free (syntax_get_rule_ptr ((struct syntax_t *) syntax,
                                         &D.6550)->keyword.chars.data);
  syntax_get_rule_ptr ((struct syntax_t *) syntax,
                       &D.6550)->keyword.chars.data = 0B;
  if (syntax_get_rule_ptr ((struct syntax_t *) syntax,
                           &D.6550)->separator.chars.data != 0B)
     __builtin_free (syntax_get_rule_ptr ((struct syntax_t *) syntax,
                                          &D.6550)->separator.chars.data);
  syntax_get_rule_ptr ((struct syntax_t *) syntax,
                       &D.6550)->separator.chars.data = 0B;
  S.297 = 0;
  while (1)
    {
       if (S.297 > 1) goto L.27;
       if (syntax_get_rule_ptr ((struct syntax_t *) syntax, 
                                &D.6550)->delimiter[S.297].chars.data != 0B)
         __builtin_free (syntax_get_rule_ptr ((struct syntax_t *) syntax,
                                     &D.6550)->delimiter[S.297].chars.data);
       syntax_get_rule_ptr ((struct syntax_t *) syntax,
                            &D.6550)->delimiter[S.297].chars.data = 0B;
       S.297 = S.297 + 1;
    }
  L.27:;
  if (syntax_get_rule_ptr ((struct syntax_t *) syntax,
                           &D.6550)->child.data != 0B)
     __builtin_free (syntax_get_rule_ptr ((struct syntax_t *) syntax,
                                          &D.6550)->child.data);
  syntax_get_rule_ptr ((struct syntax_t *) syntax, &D.6550)->child.data = 0B;
  if (D.6550.chars.data != 0B)
    __builtin_free (D.6550.chars.data);
  D.6550.chars.data = 0B;


Thus the problem seems to be that "syntax_get_rule_ptr" is treated as variable
and not as function call.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40440