GNAT & RTEMS-Ada?

GNAT & RTEMS-Ada?

[David M. Gabrius]

Has anyone gotten the two of these to work together?  I know that the
RTEMS documentation says that RTEMS has worked with Tartan Ada, but I
want to go as cheaply as possible.

If GNAT hasn't/can't work with RTEMS, is there some sort of free-ish
Ada compiler that can target 680x0 family processors available that
will?  Thanks...
-- 
 David Gabrius -- Motorola Cellular \\ gabrius@cig.mot.com \\ 708-632-5944
 Software Engineer \\"You miss too much these days if you stop to think" -U2 
 "And you can find/Your own way out/You can build/And I can will..." -U2 
 "Some days take less but most days take more" \\ #include<stddisclaimer.h>

Re: GNAT & RTEMS-Ada?

[Phillip R. Acuff]

In article <3goh9e$n0e@delphinium.cig.mot.com>, gabrius@riker.cig.mot.com (David M. Gabrius) writes:
>Has anyone gotten the two of these to work together?  I know that the
>RTEMS documentation says that RTEMS has worked with Tartan Ada, but I
>want to go as cheaply as possible.
>
>If GNAT hasn't/can't work with RTEMS, is there some sort of free-ish
>Ada compiler that can target 680x0 family processors available that
>will?  Thanks...
>-- 
> David Gabrius -- Motorola Cellular \\ gabrius@cig.mot.com \\ 708-632-5944
> Software Engineer \\"You miss too much these days if you stop to think" -U2 
> "And you can find/Your own way out/You can build/And I can will..." -U2 
> "Some days take less but most days take more" \\ #include<stddisclaimer.h>

The current release of the Ada version of RTEMS (ver. 3.1.0) depends upon the
STORAGE_SIZE attribute and Representation clauses.  In GNAT 2.0, the STORAGE_SIZE
attribute is unimplemented and there is a bug is Rep. clauses.  According to
the GNAT developers, both of these will be fixed in GNAT release 2.0.2, due
out 'any day'!

Also, GNAT does not yet work well in a cross environment, but we have promises
that they are getting there.  We have compiled RTEMS using GNAT 2.0, placing
stubs where GNAT is broken, and manually invoked ld-cross.  We are actively
working on getting RTEMS working with GNAT, in fact it is our primary goal
over the next 2 months.  

We are planning one more release of RTEMS/Ada for Tartan ada680x0, release
3.2 due out 2/20/95.  This will be followed immediately (2-4 weeks) by RTEMS/Ada
release 3.2.1 which should be completely compatible with GNAT-cross 680x0.

As a heads up, by 9/30/95, barring anything catastrophic, RTEMS will be fully
integrated with GNAT.  In this product, calls to the kernel will be made thru
the Ada 95 real-time extensions defined in the Ada 9x annexes.

For further information and to obtain the current status of RTEMS, periodically
watch  http://lancelot.gcs.redstone.army.mil/rg4/rtems.html or send e-mail
to rtems@redstone.army.mil or to the undersigned.  

We know of no other 'free-ish' Ada 9x compiler.  We do not endorse or recommend
Ada compilers, however, GNAT 2.0 has done a better job of detecting conflicting
types than some of the commercial (== expensive) compilers we have in house. 

----------
Phil Acuff
RTEMS Project Manager
pacuff@redstone.army.mil

Re: GNAT & RTEMS-Ada?

[Richard Kenner]

In article <3gtp1c$ih4@michp1.redstone.army.mil> acuff@redstone.army.mil writes:
>The problem we've encountered is related more to gnatbl.  The gnatbl build
>procedure is forcing a native gnatbl even in a cross build (i.e., gnatbl
>referenced the native toolsets rather than the cross).  This is a recognized
>problem in GNAT 2.00 which did not occur in GNAT 1.85.  Our solution was to
>invoke gnatbind and mk68k-coff-ld manually (or via. makefile).

Peculiar.  I don't see what changed in 2.00 in this area.  I would have
expected what you described to always have been the case.

I've long argued that gnatbl should be viewed as a temporary
mechanism, to be eliminated when the "gcc" driver program gets more
intelligent.  Others in the group have disagreed with me on this goal,
but you've given me more ammunition since the driver program knows how
to find cross-tools.

If you're doing sophisticated processing of object files, such as
having a cross-environment, it's probably best not to use gnatbl but
to do as you do anyway since gnatbl is a relatively simplistic tool.