Prev: The Fortran 2003 Handbook
Next: WatFor77 -- Re: Command-line argument capture
From: mrestelli on 15 Dec 2008 06:16
Dear group,
I get an error when compiling the attached code with the NAG
compiler. My impression is that this is a compiler bug... but I am
maybe missing something. Am I right?

Thank you!
Marco

nagf95 -c test_maxloc.f90
test_maxloc.f90: In function ‘mod_l_MP_pv’:
test_maxloc.f90:15: error: invalid operands to binary +

nagf95 -V
NAGWare Fortran 95 compiler Release 5.1(347,355,357-364,365)
Product NPLUX51NA for x86-32 Linux
Copyright 1990-2007 The Numerical Algorithms Group Ltd., Oxford, U.K.


module mod_l

integer, parameter :: wp = selected_real_kind(12,307)

contains

pure subroutine pv(a,k)
integer, intent(in) :: k
real(wp), intent(in) :: a(:,:)

integer :: n, ijpiv(2)

n = size(a,1)
! this does not work
ijpiv = maxloc(abs(a(k:n,k:n)))+k-1
! this works
! ijpiv = maxloc(abs(a(k:n,k:n)))
! ijpiv = ijpiv+k-1

end subroutine pv

end module mod_l
From: Tobias Burnus on 15 Dec 2008 07:13
Hi Marco,

mrestelli(a)gmail.com wrote:
> I get an error when compiling the attached code with the NAG
> compiler. My impression is that this is a compiler bug... but I am
> maybe missing something. Am I right?

I think it is a compiler bug. Note: The error message does not come from
NAG itself but from the C compiler invoked by the NAG f95 compiler.

By the way, the program compiles with gfortran, g95, ifort and openf95.

> nagf95 -c test_maxloc.f90
> test_maxloc.f90: In function ‘mod_l_MP_pv’:
> test_maxloc.f90:15: error: invalid operands to binary +

Tobias
From: George on 15 Dec 2008 14:31
On Mon, 15 Dec 2008 13:13:37 +0100, Tobias Burnus wrote:

> Hi Marco,
>
> mrestelli(a)gmail.com wrote:
>> I get an error when compiling the attached code with the NAG
>> compiler. My impression is that this is a compiler bug... but I am
>> maybe missing something. Am I right?
>
> I think it is a compiler bug. Note: The error message does not come from
> NAG itself but from the C compiler invoked by the NAG f95 compiler.
>
> By the way, the program compiles with gfortran, g95, ifort and openf95.
>
>> nagf95 -c test_maxloc.f90
>> test_maxloc.f90: In function �mod_l_MP_pv�:
>> test_maxloc.f90:15: error: invalid operands to binary +
>
> Tobias

Tobias,

If this were gfortran, how would the process go for fixing it?
--
George

I will never relent in defending America - whatever it takes.
George W. Bush

Picture of the Day http://apod.nasa.gov/apod/
From: paul.richard.thomas on 15 Dec 2008 15:38

> Tobias,
>
> If this were gfortran, how would the process go for fixing it?  
> --

We would hope that the reporter would register the bug in our bug
database gcc at gnu at org slash bugzilla slash. It's pretty easy to
do and leaves bugs on the maintainers' consciences for as long as it
takes. The bugs are subject to triage by the maintainers -
particularly to bring out the bugs that are either regressions, wrong
code or ICEs. Please go to our wiki for up-to-date information on the
different categories of bug. These are treated pretty quickly, given
that all the gfortran maintainers are part-time volunteers. The
majority of gfortran entries in Bugzilla are for enhancements of one
kind or another or for things like unclear error messages.

'twixt us all, we have been known to register bugs that come up on clf
ourselves:-) Tobias is probably the champion for doing this and
making sure that we miss nothing.

With best regards

Paul
From: mrestelli on 16 Dec 2008 06:11
On Dec 15, 1:13 pm, Tobias Burnus <bur...(a)net-b.de> wrote:
> Hi Marco,
>
> mreste...(a)gmail.com wrote:
> > I get an error when compiling the attached code with the NAG
> > compiler. My impression is that this is a compiler bug... but I am
> > maybe missing something. Am I right?
>
> I think it is a compiler bug. Note: The error message does not come from
> NAG itself but from the C compiler invoked by the NAG f95 compiler.
>
> By the way, the program compiles with gfortran, g95, ifort and openf95.
>
> > nagf95 -c test_maxloc.f90
> > test_maxloc.f90: In function ‘mod_l_MP_pv’:
> > test_maxloc.f90:15: error: invalid operands to binary +
>

Thank you all, it is now reported.

Marco
 |  Next  |  Last
Pages: 1 2
Prev: The Fortran 2003 Handbook
Next: WatFor77 -- Re: Command-line argument capture