[gfortran] Make all arguments to ENTRY master optional

[Paul Brook]

On Tuesday 24 August 2004 17:35, Tobias Schlüter wrote:
> Paul Brook wrote:
> > No. The special handling is there for a reason.
> > I've added the following to the testsuite.
> >
> > Because the master arguments are shared with the individual entry points,
> > I think you'd also end up (incorrectly) allowing:
> > call foo()
>
> I missed that. This problem exists because we call resolve_entries in the
> _beginning_ of gfc_resolve. I admit that introducing an order-dependency is
> not a good thing.

We already have an ordering dependency. We must resolve all contained 
functions (including their entry points argument lists) before resolve their 
bodies.

Paul