The first small patch adds VOLATILE and the internal-file namelist
support to the supported Fortran 2003 features.
The second corrects the CHMOD entry. I claimed that CHMOD as function
does not work. It does, but subroutine and function don't work in the
same program.
Question to the latter: How to handle intrinsics which are function and
subroutine in one?
One entry? Two entries? How distinguish argument and return value?
My patch puts them in one entry, is this ok?
Tested by "make info" and "make pdf".