WHIZARD 1.93 compilation error with g95

[Juergen Reuter]

WHIZARD does not compile with g95, neither in the PURE nor in the IMPURE environment. In the 1st case it chokes over make[1]: Entering directory `/localscratch/reuter/version_tests/1.93_g95/processes-src' g95 -I../include -I/localscratch/reuter/version_tests/1.93_g95/omega-src/bundle/lib -c ee_o.f90 In file ee_o.f90:367

call omega_scatter (amplitude_2, k, rho_in, rho_out, &

1

Error: Host-associated variable 'amplitude_2' is used in a defining context in a PURE procedure at (1) In the 2nd case it chokes over: /afs/physik.uni-freiburg.de/home/reuter/bin/g95 -O -I../lib -c -o testbed.o ../src/testbed.f95 In file ../src/testbed.f95:210

omega_spin_sum_sqme_1 (omega1, p, 1, size (spin_states, dim=2)) &

1

Error: Dummy procedure 'omega1' at (1) must be PURE In file ../src/testbed.f95:213

omega_spin_sum_sqme_1 (omega2, p, 1, size (spin_states, dim=2)) &

1

Error: Dummy procedure 'omega2' at (1) must be PURE make[3]: * [testbed.o] Error 1

[Juergen Reuter]

It seems that gfortran (4.5.0) is crashing for the IMPURE environment for the same reason as g95.

[Christian Speckner]

I never found the patience to properly track this one down, but this is due to the O'Mega build system not properly stripping the all pure statements. Manually stripping the sources via "sed" allows for successfull compilation.

[Juergen Reuter]

This has been settled. It arose from the fact that O'Mega was connecting the cat/cpif transferral from .f95 to .f90 files (for compilers which do not understand .f95 files) with the sed command for removing pure (and possibly elemental) statements from functions and subroutines prior to compilation. As g95 and NAGFOR do understand .f95, the program crashed during impure compilations. I moved everything to .f90 convention (as in WHIZARD 2). Also, a typo in whizard.nw has been corrected which triggered a bug with NAGFOR.

[Juergen Reuter]

Milestone v1.94 deleted