X3J3/96-078

Date:    April 29, 1996
To:      X3J3
From:    Kurt W. Hirchert
Subject: JOR Item No. 112

  <FORTREQ>
  <NUMBER> 112
  <TITLE> "Clean up" conformance rules
  <KEYWORDS> CONFORMANCE
  <STATUS> Registered
  <TARGET>
  <SUBGROUP>
  <VERSION> 1

  <REQUIREMENT>
  Review conformance rules with an eye towards making them more regular,
  easier to understand, more useful, etc.

  <JUSTIFICATION>
  The standard would be easier to understand, more useful, etc.

  <SUGGESTED IMPLEMENTATION>
  Specific issues I would suggest looking at include the following:

  o  If a Fortran program uses a procedure defined by means other than
     Fortran, it can still be a standard-conforming program.  However,
     if the vendor changes that procedure to an intrinsic procedure to
     make it more efficient, the program becomes nonconforming.  We
     should try to treat these cases similarly.

  o  If a format is written as a FORMAT statement, the processor is required
     to be able to check it for correctness, but if it is written as
     a CHARACTER constant, the processor is not.

  o  If we should separate the concepts of a program which is standard
     conforming in form (i.e., on a static basis) from whether or not
     the execution of that program conforms.

  <ESTIMATED IMPACT>
  Much of this could be done with only localized changes to the document,
  but identifying the difference between static and dynamic requirements
  in the standard might be more widespread.

  Most of this should have very limited impact on existing processors.  The
  obvious exception would be a requirement that processors check the
  correctness of formats expressed as character constant expressions.

  <SUBMITTED BY> Kurt W. Hirchert, hirchert@ncsa.uiuc.edu

  <HISTORY>
  <EVENT> 29 April 96: submitted
  </HISTORY>

  </FORTREQ>
--
Kurt W. Hirchert     hirchert@ncsa.uiuc.edu
National Center for Supercomputing Applications