ICE seg fault
Florian Ladstaedter
flad@gmx.at
Tue Jul 31 12:18:00 GMT 2007
Daniel Franke wrote:
> On Tuesday 31 July 2007 10:22:03 Florian Ladstaedter wrote:
>> I could reduce the test case quite a lot:
>> ------------------------------------------------
>> module AtmoIonoSphere
>> use AtmoIono
>> use EGOPS_Utilities, ONLY: dirname
>> end module AtmoIonoSphere
>> ------------------------------------------------
>
> It now depends on what these modules contain. Could you send me either the
> corresponding source or the module files (.mod) privately, please?
>
ok, finally I could hunt it down to a short complete testcase.
First, it seems to be some kind of regression since it compiles with 4.2.1
Second, it does not occur when I don't use separate modules.
Here is the test case:
file egopsutil_red.f90:
-------------------------------------------------
MODULE EGOPS_Utilities
IMPLICIT NONE
PUBLIC
CONTAINS
FUNCTION dirname(fullfilename)
Character(LEN=*), Intent(In) :: fullfilename
Character(LEN=LEN(fullfilename)) :: dirname
dirname = ''
END FUNCTION dirname
END MODULE EGOPS_Utilities
-------------------------------------------------
file AtmoIono_red.f90:
-------------------------------------------------
MODULE AtmoIono
IMPLICIT NONE
PUBLIC
CHARACTER(LEN=10), PARAMETER :: ComputeDryAtmModel = 'Dry Atm. '
type AtmModel
character (len=len(ComputeDryAtmModel)) :: moistDryStr
end type AtmModel
END MODULE AtmoIono
-------------------------------------------------
file AtmoIonoSphere_red.f90:
-------------------------------------------------
module AtmoIonoSphere
use AtmoIono
use EGOPS_Utilities
end module AtmoIonoSphere
-------------------------------------------------
Compiling with:
gfortran -c -J../include -I../include AtmoIono_red.f90
gfortran -c -J../include -I../include egopsutil_red.f90
gfortran -c -J../include AtmoIonoSphere_red.f90
gives:
AtmoIonoSphere_red.f90:0: internal compiler error: Segmentation fault
The "Character(LEN=*)" seems to be part of the problem, since using a
fixed value here it works.
thanks Florian
More information about the Fortran
mailing list