Good pointer-function handling.
It seems obvious that a function returning a pointer should be valid anywhere that a pointer of the same value is valid. Sun's Fortran is the first one I have tried that gets this right. There are several standards clarifications about pointer functions because the idea seems to be confusing to traditional Fortran programmers.
The sort of problems I am referring to are using pointer functions as arguments, such as:
call sub(get_data_ptr('data name'))
This should be exactly equivalent to:
p_data => get_data_ptr('data name')
call sub(p_data)
Many Fortran programmers discourage pointer functions, so the compilers seem to be poorly tested with such use. It seems that the Sun compiler developers see pointers as useful, and have done a good job supporting them.
