[Patch, Fortran, OOP] PR 42769: ICE in resolve_typebound_procedure

[Janus Weil <janus at gcc dot gnu dot org>]

Hi all,

here is another small OOP wrong-code patch (fixing comment #27 of the
aforementioned PR). All it does is to make sure a vtab is generated
for modules like this:

module mod1
  type :: t1
  contains
    procedure, nopass :: get => my_get
  end type
contains
  integer function my_get()
    my_get = 1
  end function
end module


Usually we only build a vtab once we find a CLASS variable. In this
module we have no CLASS variable (due to NOPASS), but still we need a
vtab. Therefore we generate one in 'resolve_typebound_procedures',
which in turn guarantees that we initialize the TBP pointer with the
right procedure. The second hunk was needed fix a testsuite
regression.

The only testsuite failure I see with this patch is 'float128_1.f90',
which surely has a different origin.

Ok for trunk?

Cheers,
Janus


2010-08-29  Janus Weil  <janus@gcc.gnu.org>

	PR fortran/42769
	* resolve.c (resolve_structure_cons): For derived types, make sure the
	type has been resolved.
	(resolve_typebound_procedures): Make sure the vtab has been generated.


2010-08-29  Janus Weil  <janus@gcc.gnu.org>

	PR fortran/42769
	* gfortran.dg/dynamic_dispatch_11.f03: New.

Attachment: pr42769.diff
Description: Binary data

Attachment: dynamic_dispatch_11.f03
Description: Binary data