Re: SEQUENTIA_IO and stdin/stdout (Standard_Input, Standard_Output)

[Marius Amado Alves <amado.alves@netcabo.pt>]

> Which of the three steps is not guaranteed ?
> 
> 	Finding the value of File_Type
> 	Finding the name of the corresponding file
> 	Opening the file
> 
> If it is the last, I would expect a failure to open would signal an exception.

I would to. But GNAT does not complaint. At least with a sequential file
of characters. It simply does not output. Maybe it's because I'm not
closing Standard_Output prior to opening it as a sequential file. But I
wouldn't know how to do that. Procedure Close requires a variable, and
you cannot set a file type variable to Standard_Output (because
File_Type is limited).

I agree some kind of clarification on this issue is desirable. Maybe it
has been done already. Or a binding interpretation. Someone should check
it out ;-) Note Robert Eachus already said "it may or may not work".
Maybe he's willing to elaborate ;-)

Personally I don't feel comfortable with two things:

1. The (external) name of a standard channel. Even in C this does not
exist. GNAT says "*stdout", "*stdin". Note the asterisk. This reeks of implementation dependency. I
seem to remember MS-DOS having "CON" for this. I think the ARM is silent
on this anyway.

2. Crossing information between Text_IO and (an instance of)
Sequential_IO. Even if this information is just the name of the external
file.

In sum: there is doubtful solution (and erroneous with GNAT, or so it
seems), and there is an alternate, legal solution. The choice should be
clear.

As I indicated before, I suspect Ada has a weak point here. Namely
binding the standard channels exclusively to Text_IO. Surely the
standard channel concept is less dependent of file format than that.