Microsoft KB Archive/109874

From BetaArchive Wiki
< Microsoft KB Archive
Revision as of 16:06, 18 July 2020 by 3155ffGd (talk | contribs) (importing KB archive)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)

Article ID: 109874

Article Last Modified on 10/17/2003



APPLIES TO

  • Microsoft FORTRAN PowerStation 1.0 Standard Edition
  • Microsoft Fortran PowerStation 1.0a for MS-DOS
  • Microsoft FORTRAN PowerStation 32



This article was previously published under Q109874

SYMPTOMS

A program that contains a simple arithmetic expression including constant out of range may produce incorrect results. Using the -4Yb compiler option does not give a run-time error on constants that are out of range even though it should.

CAUSE

The constant folding algorithm (arithmetic performed at compile time) generates 0 (zero) for the variable out of range.

RESOLUTION

Avoid using a simple constant expression that might be out of range. Use the temporary variable for the intermediate results and then use this variable in the expression. Compile the program with -4Yb and the appropriate run-time error will be generated.

STATUS

Microsoft has confirmed this to be a problem in Microsoft FORTRAN PowerStation 32 for NT, version 1.0 and FORTRAN Powerstation for MS-DOS version 1.0 and 1.0a. This is not a problem in Microsoft FORTRAN for MS-DOS, version 5.1, or FORTRAN PowerStation 32, version 4.0.

MORE INFORMATION

Sample Code #1

c Compile options required: -4Yb does not affect the behavior
c The following code reproduces the problem.

       i=2**50
       write (*,*) i
       end
                

Sample Code #2

c Compile options required: -4Yb
c The following code solves the problem and generates the following
c run-time error: run-time  F6099: $DEBUG
c            -INTEGER arithmetic overflow

       j=50
       i=2**j
       write (*,*) i
       end
                


Additional query words: 1.00 runtime run time

Keywords: kbbug kbfix kbcompiler KB109874