To: J3                                                     J3/20-145
From: Bill Long
Subject:  Response to UTI 010
Date: 2020-October-08
References: 20-007

Discussion:
-----------

UTI 10 from page 90 of 20-007.pdf:

----------------
Why can BOZ be used to init a PARAMETER but not a variable?
Seriously, that
       INTEGER,PARAMETER :: x = z'2000'
is valid but
       INTEGER :: y = z'2000'
is invalid, is obviously inconsistent.
---------------

We agree that this is inconsistent and variable initialization should
be allowed.

A simple test:

> cat uti10.f90
program uti10
  implicit none
  integer,parameter :: x = z'2000'
  integer :: y = z'2001'

  print *, x, y
end program uti10

Four compilers allow the code as is:

> ftn uti10.f90
> ./a.out
 8192,  8193
> ftn -VV
Cray Fortran : Version 10.0.3

> ifort uti10.f90
> ./a.out
        8192        8193
> ifort --version
ifort (IFORT) 19.1.2.254 20200623

> pgf90 uti10.f90
> ./a.out
         8192         8193
> pgf90 --version

pgf90 20.1-0 LLVM 64-bit target on x86-64 Linux -tp sandybridge

> nvfortran test.f90
> ./a.out
         8192         8193
> nvfortran --version

nvfortran 20.7-0 LLVM 64-bit target on x86-64 Linux -tp sandybridge


A fifth compiler gives an error by default for standard violation,
but provides an option to lower the error to a warning:

> gfortran -fallow-invalid-boz uti10.f90
uti10.f90:3:26:

    3 |   integer,parameter :: x = z'2000'
      |                          1
Warning: BOZ literal constant at (1) is neither a data-stmt-constant
nor an actual argument to INT, REAL, DBLE, or CMPLX intrinsic function
uti10.f90:4:16:

    4 |   integer :: y = z'2001'
      |                1
Warning: BOZ literal constant at (1) is neither a data-stmt-constant
nor an actual argument to INT, REAL, DBLE, or CMPLX intrinsic function
> ./a.out
        8192        8193
> gfortran --version
GNU Fortran (GCC) 10.1.0

Therefore, it appears to be a minimal implementation change to allow a
boz constant as the initialization in the declaration of a variable.

Edits are provided to modify this feature.

Edits to 20-007:
----------------

[90:26-30] In 7.7 Binary, octal, and hexadecimal literal constants, in
the last Constraint in the subclause, change "initialization for a
named constant" to "initialization for a named constant or variable"

{making the full constraint read


C7113 (R764) A <boz-literal-constant> shall appear only as a
data-stmt-constant in a DATA statement, as the initialization for a
named constant or variable of type integer or real, as the <expr> in
an intrinsic assignment whose variable is of type integer or real, as
an ac-value in an array constructor with a type-spec that specifies
type integer or real, as an output-item in an output statement, or
where explicitly allowed in 16.9 as an actual argument of an intrinsic
procedure.}

[90:30+] Remove Unresolved Technical Issue 010