Standard violation or Absoft bug? 
Author Message
 Standard violation or Absoft bug?

[bcc'd to Absoft support]

We compile and test with a bunch of compilers and platforms
and many compiler vendors have a copy of our code for testing,
but so far we haven't used Absoft routinely.  I just tried it
on an x86 platform running RedHat 7.3.

The following seems to be a bug in Absoft, but it may also be
a standard violation?  (Having a leading period in an operator
name?) We have not encountered this error with Lahey, SGI, NAG,
g95, Intel, Compaq, PGI, or PathScale.

  $ abf95 -g -c -o linearsolve_module.o linearsolve_module.f90

  module linearsolve_module
         ^
  cf90-855 f90fe: ERROR LINEARSOLVE_MODULE, File = linearsolve_module.f90, Line = 1, Column = 8
    The compiler has detected errors in module "LINEARSOLVE_MODULE".  No module information file will be created for this module.

    use complex_functions, only: operator(>),  operator(.cge.), &
                                                       ^
  cf90-732 f90fe: ERROR LINEARSOLVE_MODULE, File = linearsolve_module.f90, Line = 3, Column = 55
    "cge." is not in module "COMPLEX_FUNCTIONS".

  f90: Copyright Absoft Corporation 1994-2005; Absoft fortran Compiler Version 9.0e

Note the missing period at the front of .cge. in the error report.

The complex_functions module has already been compiled
and looks like

  module complex_functions
    [..]
    public :: operator(.cge.)

  ! There is/was a bug in the SGI, Absoft, and Cray compilers that
  ! prevents the overloading of >= for complex expressions so
  ! we will define our own for all combinations

    interface operator(.cge.)
      module procedure c_c_ge
      module procedure c_r_ge
      [..]
      module procedure i_r_ge
    end interface

Thanks,
--
Bil
http://www.*-*-*.com/



Wed, 26 Nov 2008 01:26:09 GMT  
 Standard violation or Absoft bug?

Quote:

> The following seems to be a bug in Absoft, but it may also be
> a standard violation?  (Having a leading period in an operator
> name?)

Not only is there no problem with having a period, it is *REQUIRED* to
have a period (both leading and trailing). That's part of the syntax.

Technically, these are not "names", by the way. Names have a different
syntax rule that operators don't fit under. (Operators can't have digits
or underscored - just letters, plus, of course, the two periods). The
standard uses the term "identifier" for the broader class of of which
includes both names and other forms such as this one that are used to
identify things.

--
Richard Maine                    | Good judgement comes from experience;
email: last name at domain . net | experience comes from bad judgement.
domain: summertriangle           |  -- Mark Twain



Wed, 26 Nov 2008 02:45:15 GMT  
 Standard violation or Absoft bug?

Quote:
>  ! There is/was a bug in the SGI, Absoft, and Cray compilers that
>  ! prevents the overloading of >= for complex expressions

Hmm, I thought I'd learnt in Analysis 101 that the complex numbers do
not admit to a complete ordering - so what does you comparison function
actually do?

        Jan



Fri, 28 Nov 2008 17:31:43 GMT  
 Standard violation or Absoft bug?

Quote:

>>  ! There is/was a bug in the SGI, Absoft, and Cray compilers that
>>  ! prevents the overloading of >= for complex expressions

> Hmm, I thought I'd learnt in Analysis 101 that the complex numbers do
> not admit to a complete ordering - so what does you comparison function
> actually do?

It ignores the imaginary parts, and compares the real parts.

We use the complex extensions to form adjoints, viz,

  http://fun3d.larc.nasa.gov/papers/nielsen2h.pdf

Regards,
--
Bil
http://fun3d.larc.nasa.gov



Fri, 28 Nov 2008 17:55:22 GMT  
 
 [ 4 post ] 

 Relevant Pages 

1. Absoft MacFortran/020 & Standard File

2. Violation of F77 standard?

3. Violation of F90 Standard?

4. CVF, Absoft bug involving ELEMENTAL character functions

5. Cray, SGI, and Absoft compiler bugs?

6. IBM/Absoft bug on Mac OS-X

7. Absoft Fortran 7.0 bug?

8. deallocate bug in absoft f90 linux ?

9. No bug in "r := i", just a Report violation

10. Bug in gadfly: Violation of Unique key

11. BUG: ET (mktclapp) segmentation violation (patch included)

12. BUGS, BUGS, BUGS, BUGS, C4 BUGS

 

 
Powered by phpBB® Forum Software