Re: Internal procedure not allowed as actual argument

[Steve Kargl <sgk at troutmask dot apl dot washington dot edu>]

On Sat, Jun 30, 2007 at 09:01:50PM +0200, Tobias Schl?ter wrote:
> Janus Weil wrote:
> >Hi all,
> >I have a question: why doesn't gfortran allow passing an internal
> >procedure as an argument to another procedure, as in this example:
> >
> >subroutine sub
> >
> > call proc(f1)
> > call proc(f2)
> >
> > contains
> >
> > real function f1(x)
> > ...
> > end function f1
> >
> >end subroutine sub
> >
> >real function f2(x)
> >...
> >end function f2
> >
> >If we have a subroutine "proc" that expects an external function as
> >its argument, "call proc(f2)" is allowed by gfortran, while "call
> >proc(f1)" triggers an error like this:
> >
> >Error: Internal procedure 'f1' is not allowed as an actual argument at (1)
> >
> >g95 shows the same behaviour, but ifort doesn't complain about this at all.
> >Can anyone tell me what the reason for this restriction is?
> 
> Quoting 12.4.1.3. from the F2K draft standard:
> 
> "If a dummy argument is a dummy procedure without the POINTER attribute, 
> the associated actual argument shall be the specific name of an 
> external, module, dummy, or intrinsic procedure, an associated procedure 
> pointer, or a reference to a function that returns an associated 
> procedure pointer. The only intrinsic procedures permitted are those 
> listed in 13.6 and not marked with a bullet (?). If the specific name is 
> also a generic name, only the specific procedure is associated with the 
> dummy argument."
> 
> Note that internal procedures are not mentioned.  The idea behind this 
> is that it is hard to make the local variables of the containing 
> procedure visible to the contained procedure when it's called from 
> somewhere else.
> 

Along with Tobi's observation, if the contained function is fairly
short, gfortran/gcc will in-line the function so the function may
disappear.

-- 
Steve