Custom Query (210 matches)

When trying to compile the release-10 branch with Intel ifort, I get the error

ifort -c -I/opt/software/eb/software/ecCodes/2.12.5-iimpi-2019a/include -I/opt/software/eb/software/netCDF-Fortran/4.4.5-iimpi-2019a/include -O3  -g -fpp -fp-model source -mcmodel=medium -unroll -ipo -inline -heap-arrays 32 -xHost -traceback assignland.f90
concoutput_inversion.f90(357): error #6503: This keyword is invalid for this intrinsic procedure reference.   [UNIT]
            call flush(unit=unitrelnames)
-----------------------^
compilation aborted for concoutput_inversion.f90 (code 1)
make: *** [concoutput_inversion.o] Fehler 1

Compilation succeeds when I remove the call from that line, as shown in this patch:

diff --git a/src/concoutput_inversion.f90 b/src/concoutput_inversion.f90
index fb327cf..9562bd2 100644
--- a/src/concoutput_inversion.f90
+++ b/src/concoutput_inversion.f90
@@ -354,7 +354,7 @@ subroutine concoutput_inversion(itime,outnum,gridtotalunc,wetgridtotalunc, &
             open(unit=unitrelnames, file=path(2)(1:length(2))//'releases_out',form='formatted', &
                  & access='APPEND', iostat=ierr)
             write(unitrelnames,'(a)') areldate//areltime//'_'//anspec
-            call flush(unit=unitrelnames)
+            flush(unit=unitrelnames)
             close(unitrelnames)
           endif
         endif

After checking ​https://gcc.gnu.org/onlinedocs/gcc-9.2.0/gfortran/FLUSH.html, it seems that recent Fortran compilers don't need the call any more. And indeed, on our system (gfortran 6.3.0) the modified code compiles as well.

So I suggest to apply the above patch.

I have built Flexpart from source on a couple of Linux machines and I always get a compile error due to a problem in erf.f90. The error message is:

gfortran -c -O2 -m64 -mcmodel=medium -fconvert=little-endian -frecord-marker=4 -I/usr/lib64/gfortran/modules erf.f90 erf.f90:106.26:

real, external

gammln 1

Error: Return type mismatch of function 'gammln' at (1) (REAL(4)/REAL(8)) erf.f90:111.13:

gln=gammln(a)

1

Warning: Type mismatch in argument 'xx' at (1); passed REAL(4) to REAL(8)

The same error happens at line 140. The following patch allows Flexpart to compile although it may not be the best way of achieving this:

aurora.hcp>diff baderf.f90 erf.f90 106c106 < real, external :: gammln ---

double precision, external

gammln

111c111 < gln=gammln(a) ---

gln=gammln(DBLE(a))

140c140 < real, external :: gammln ---

double precision, external

gammln

145c145 < gln=gammln(a) ---

gln=gammln(DBLE(a))

The version of gfortran used does not appear to be important; the same thing happens with gfortran 4.8.5 on Scientific Linux 7 and with gfortran 6.3.0 on Debian 9.2.

I got run from Flexpart-Wrf test case and I got the message that I have successfully completed a Flexpart-Wrf model run. However, when I see the netcdf result, all of the concentration values are equal to zero. I don't know why I get this result. Thanks in advance for your help.

P.S: I changed line 233 of convmix_kfeta.f90 file as it was mentioned in ticket #107, but I still get zero as the result of model.