I am compiling abinit and looking at the output I found a couple of warnings (which I agree with
).Here is an outline of an example code that has a bug (starting from line 573 in 68_dmft/m_hu.F90):
Code: Select all
subroutine rotatevee_hu(cryst_struc,hu,nspinor,nsppol,pawprtvol,rot_mat,udens_atoms)
... lots of code ...
type(coeff2_type),intent(out) :: udens_atoms(cryst_struc%natom)
... start of code ...
write(std_out,*)"udens before vee2udensaomt", udens_atoms(1)%value
... down to end of code ...As you can see the udens_atoms is defined with intent(out). However, it is only used as a print-out statement.
I regard this as a bug as the udens_atoms *could* be undefined on return.
Looking at the remaining code it seems like it previously was set, hence I would recommend that you change the interface to intent(inout).
This is also the case for these files:
src/70_gw/m_screen.F90 (line 1651, routine screen_times_ket), variable "sigcme" * this is unimportant as it has to do with an unimplemented feature *, should be "intent(inout)"
src/69_wfdesc/m_oscillators.F90 (line 1288, routine calc_pw_oscillator), variable "Osc", should be "intent(in)" or "intent(inout)", it gets deferred further down in other calls
src/65_lotf_base/m_glue_lotf.F90 (line 192, routine glue_pair_devs), variable "r_au", should be "intent(in)"
Kind regards Nick
