Re: patch attached this time for *-rtems Ada on 3.2 branch

[Joel Sherrill <joel dot sherrill at OARcorp dot com>]

Geert Bosch wrote:
> 
> On Wednesday, Feb 5, 2003, at 10:16 America/New_York, Gabriel Dos Reis
> wrote:
> > | The important thing (which may be too late) is to revert what I did.
> >
> > It is too late now.
> 
> I have no idea why this patch was approved in such a hurry, just days
> before a release. There was not even a mention of any testing that has
> been
> done. Also, especially on a release branch, the patch should have listed
> *why* it should be applied, and what regressions it fixes. It is very
> unfortunate that other targets have been affected by this.

First I apologize.  

It had been in the RTEMS patch set since the first incorporation
of Ada into the gcc tree (3.1) and had been discussed on the gcc
mailing lists and privately with Laurent.  It's not worth 
arguing about.  I broke other targets.
 
> I understand that the patch is supposed to be RTEMS specific, but it
> gratuitously eliminated the two utilities gnatpsta and gnatpsys for
> other cross-targets. Especially on cross platforms it is useful to
> be able to see what the definition is of language-defined types.

I won't argue with your presumption that they are useful but they
were initially disabled because they did not build and can not
reliably be built on all cross targets to produce a useful program.

If these programs are intended to execute on the cross target, then
I contend they need to be optional per target.  For deeply embedded
systems, they may have to be linked with hardware specific code like
a board support package that is not available at this time.  For sure,
that's the case on RTEMS.  So they are can not be built in a useful 
manner for any RTEMS target during the gcc/gnat build.

>    -Geert

--joel