J3/14-195
To:        J3
From:      Van Snyder
Subject:   Interp concerning undefinition
Date: 2014 June 18
Reference: 14-106, 14-125r1, Interp F03/0124, Corrigendum 1

----------------------------------------------------------------------

NUMBER: TBD
TITLE: Undefinition
KEYWORD: undefinition
DEFECT TYPE: Erratum
STATUS: J3 consideration in progress

QUESTIONS:

1. Does the following cause D to become undefined?  Ought it?

  integer :: I(2)
  double precision :: D
  equivalence ( D, I )
  d = 42.0
  i(1) = 666

2. Does the following cause C to become undefined?  Ought it?

  integer, parameter :: RK = kind(0.0d0)
  real(rk), pointer :: P
  complex(rk), target :: C = ( 1.0, 2.0 )
  c = ( 1.0d0, 2.0d0 )
  p => c%re
  p = 24.0d0

3. Does the following cause D1 to become undefined?  Does it cause C%IM
to become undefined?  Ought it?

  integer, parameter :: RK = kind(0.0d0)
  real(rk), pointer :: D1, D2
  complex(rk), target :: C
  c = ( 1.0d0, 2.0d0 )
  d1 => c%re
  d2 => c%im
  call s ( d2 )
  ...
  contains
    subroutine S ( D )
      real(rk), intent(out) :: D
  ....

4. Does the following cause Y to become undefined?  Ought it?

  type :: T1
    sequence
    integer :: C1 = 42
  end type T1
  type :: T2
    sequence
    real :: C1 = 42.0e0
  end type T2
  type(t1) :: X
  type(t2) :: Y
  equivalence ( X, Y )
  x = t1(21)

ANSWERS:

1. Item (1) in subclause 16.6.6 says

  "When a scalar variable of intrinsic type becomes defined, all totally
  associated variables of different type become undefined....  When a
  scalar variable becomes undefined, all partially associated double
  precision scalar variables become undefined."

D is not totally associated with I.  I does not become undefined.
Therefore, D does not become undefined.  D ought to become undefined.
Edits are provided to correct this mistake.

2. Item (1) in subclause 16.6.6 says

  "When a double precision scalar variable becomes defined, all
  partially associated scalar variables become undefined."

P is partly associated with C, and therefore C becomes undefined by the
assignment to the scalar variable P.  C ought not to become undefined.
Edits are provided to correct this mistake.

3. Item (1) in subclause 16.6.6 says

  "When a scalar variable becomes undefined, all partially associated
  double precision scalar variables become undefined."

  Subclause 16.6.1p5 says

  "A complex... scalar object is defined if and only if all of its
  subobjects are defined."

The association of the scalar variable D2 with an INTENT(OUT) dummy
argument causes it become undefined.  16.6.6p1(1) does not specify that
when a scalar real variable becomes undefined, a totally associated part
of a complex variable becomes undefined.  Assuming C%IM ought to become
undefined, then C ought to become undefined, even though C%RE is not
undefined.  D1 is associated with the C%RE, and therefore becomes
undefined when C%IM becomes undefined.  C%IM ought to become undefined,
and D1 ought not to become undefined.  Edits are provided to correct
this mistake.

4. Item (1) in subclause 16.6.6 says

  "When a scalar variable of intrinsic type becomes defined, all totally
  associated variables of different type become undefined.

X and Y are totally associated, and of diffent type, but neither one is
of intrinsic type.  Therefore, Y does not become undefined as a
consequence of X becoming defined.  Y ought to become undefined.  Edits
are provided to correct this mistake.

EDITS:

In Subclause 16.6.6p1, item(1), within the first sentence, remove "of
intrinsic type".  Before "different" insert "the same type with
different values of kind type parameters, or of"

In Subclause 16.6.6p1, replace item(1) with the following four sentences
(all in one paragraph)

  "When a scalar variable becomes defined, all totally associated
  variables of different type become undefined.

{Removing "of intrinsic type" from the existing first sentence of
16.6.6p1(1) addresses question 4.}

  When a scalar variable becomes defined, all partially associated
  variables of different type, or of the same type if corresponding kind
  type parameters have different values, become undefined, unless the
  scalar variable is totally associated with a subobject of the
  partially associated variable, that subobject is of the same type as
  the scalar variable, and corresponding kind type parameters have the
  same values.

{This addresses questions 1 and 2.  The references to kind type
parameter values address the possibility of partially associating a
double precision variable with a default real variable.  Default real
and double precision are now the same type (since 1990), but with a
different kind type parameter value.}

  When a scalar variable becomes undefined, all partially associated
  variables become undefined, unless the partially associated variable
  is totally associated with a subobject of the scalar variable that
  does not become undefined.

  When a scalar variable becomes undefined, all totally associated
  subobjects of other variables become undefined if those subobjects are
  of different type from the scalar variable, or corresponding kind type
  parameters have different values."

{The third and fourth sentences address question 3.}

SUBMITTED BY: Van Snyder

HISTORY: m204  14-xxx    Submitted

----------------------------------------------------------------------