Re: How to exit an Ada program with (unix shell) error code?

[Martin <martin.dowie@btopenworld.com>]

On May 6, 9:41 am, a...@anon.org (anon) wrote:
> By your logic, who said procedures are portable!

Perhaps you could rephrase that into a meaning sentence?


>  Using RM 10.2 (29), a
> vendor could easily restrict the "main subprograms" to functions only.

No they can't - read the sentence begining "However". In fact, here it
is:

  "However, an implementation is required to support all main
   subprograms that are public parameterless library procedures."

Read it slowly. Read it repeatedly.

Here's a link to it:

http://www.adaic.org/standards/1zrm/html/RM-10-2.html


> And there is no definition in the RM that states a "main subprograms"
> subroutine must be a parameterless library procedure, either.

See above. Especially the link.


> So, its better to have and state that both subprogram types are portable. Then
> one can point out that the "main subprograms" that are procedures are more
> widely used in Ada.

That's would be lying, so probably better not to state this at all.



> In <948d6a6b-d603-4e38-add2-50016c2dc...@s20g2000vbp.googlegroups.com>, Martin <martin.do...@btopenworld.com> writes:
>
> >On May 5, 9:31=A0pm, a...@anon.org (anon) wrote:
> >> Martin.
>
> >> =A0 You misread RM 10.2 (29), its quoted below.
>
> >> Both of the subprograms types are valid and portable and even though
> >> RM 10.2 (29) allows a vendor to restrict the main subprogram. Adacore
> >> (GNAT) and IBM which follow the RM to the letter have decided to allow
> >> and support both main subprograms types in Ada 83/95/2005 specs.
>
> >> Note: At this time Adacore and IBM are the only ones supporting Ada 2005.
>
> >> Now, in the GNAT design the "GNATBIND" process actually wraps the main
> >> subprogram ( "procedure main" or a "function main return integer" ) withi=
> >n
> >> a function called "main" within a package called "ada_main". =A0And if th=
> >e
> >> "main" subprogram is a procedure, the binder function "main" calls this
> >> procedure and returns the exit status which is normally set to zero unles=
> >s it
> >> is altered by "Ada.Command_Line.Set_Exit_Status". But if main subprogram
> >> is a function type of partition, then the binder function "main" returns =
> >the
> >> value returned from the partition "function main" to the OS as the exit
> >> status. To see this, just compile any program and then bind it. Check the
> >> resulting output files "b~<name>.adb" and its spec file before linking.
>
> >> Note: normally during linking the linker "GNATLINK" deletes the
> >> binder source/object files.
>
> >> So both of these are valid in the RM. =A0It just that most programmers pr=
> >efer
> >> the "Procedure" type of partitions design.
>
> >> --
> >> -- main subprograms as a procedure
> >> --
> >> procedure main is
> >> =A0 begin -- main
> >> =A0 =A0 =A0null ;
> >> =A0 end main ;
> >> --
> >> -- or main subprograms as a public parameterless function
> >> --
> >> function main return integer is
> >> =A0 begin -- main
> >> =A0 =A0 =A0return ( 0 ) ; -- 0 :=3D normal exit status.
> >> =A0 end main ;
>
> >> Ada RM -- quoted
>
> >> =A0 RM 10.2 Program Execution
> >> =A0 =A0...
>
> >> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0Implementation Permiss=
> >ions
> >> =A0 =A0...
>
> >> 29 =A0 =A0An implementation may restrict the kinds of subprograms it supp=
> >orts as
> >> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 ------------
> >> main subprograms. However, an implementation is required to support all m=
> >ain
> >> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =
> >=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 ----------------
> >> subprograms that are public parameterless library procedures.
> >> -----------
>
> >> =A0 RM 10.1 Separate Compilation
>
> >> 1 =A0 =A0 A program unit is either a package, a task unit, a protected un=
> >it, a
> >> protected entry, a generic unit, or an explicitly declared subprogram oth=
> >er
> >> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =
> >=A0 =A0------------------------------
> >> than an enumeration literal. Certain kinds of program units can be separa=
> >tely
> >> compiled. Alternatively, they can appear physically nested within other
> >> program units.
>
> >You forgot to underline the important bit - "procedures". Not
> >"subprograms", not "functions or procedures" - just "procedures". The
> >_only_ subprogram that _must_ be supported is:
>
> >   procedure <Main_Subprogram_Name> is
> >   ...
>
> >Yes an implementation is free to support other forms (including
> >functions) but they are not portable.
>
> >Lord only know why your quoting 10.1 - it's nothing to do with main
> >procedures.