Ada Language Support in sparc-rtems-gdb

[Joel Sherrill <joel@OARcorp.com>]

Ralf Corsepius wrote:
> I received several comments on PM, which I think are of interest to be
> answered to a wider audience.
> 
> On Wed, 2004-09-22 at 13:59, <an undisclosed poster> wrote: 
> 
>>Ralf Corsepius wrote:
>>
>>
>>>FSF/GNAT in general is in poor shape (newer versions aren't even
>>>buildable) and so is Ada support in RTEMS. 
>>>
>>>So if you should have managed to *run* any RTEMS Ada code you're
>>>probably further than anybody else on this planet ;)
> 
> 
>>That doesn't sound too reassuring. 
> 
> Your impression is right. IMO, neither GNAT nor RTEMS/Ada are in a
> usable shape. Something urgently needs to be done about both of them,
> otherwise I fear, there will be no future for both of them.
> 
> [Try building an ada-enabled cross-compiler and you probably will
> experience the reasons why I am saying this yourself.]
> 
> 
>> I haven't done my RTEMS project yet, 
>>however, this makes me wonder if I am going the wrong direction.
> 
> 
> Well, I don't want to discourage anybody from using RTEMS. However I
> doubt the usability of the FSF's GNAT *in general*. RTEMS/Ada's shape
> only is an indirect victim.
> 
> I for one had been fighting with the technical impact of GNAT on RTEMS
> and the "friendliness" of the GNAT developers for several years.
> Meanwhile, I therefore have decided to join the "passive resistance
> faction" amongst the GCC developers. Ie. I will not to touch Ada in
> RTEMS nor GNAT until GNAT is in a shape to be usable "out of the box".

I started working on GNAT/RTEMS back in the 1994-95 timeframe.  To say
that I have been frustrated at times is an understatement.  I try to
persevere and keep Ada on RTEMS working as best I can. But it is
difficult when the ACT 3.1x versions are based upon a GCC that
is so old, we don't want to make them work with RTEMS anymore. The
new gcc 3.x based GNATs seem to be broken in new and interesting
ways each release.  GNAT on RTEMS can never be any better than
GNAT natively and GCC 3.x GNAT native is often not in tip-top shape.

It does look like there is starting to be some GNAT developer
activity on the list but Ralf is right.  A portion of the GCC community
is frustrated.  Supporting GNAT in GCC 3.x requires specialized
knowledge.  Those with the specialized knowledge don't seem to
contribute fixes in timely manner.

> Check the GCC mailing list archive and you will probably notice what I
> am referring to. A group of the GCC developers currently is performing
> "passive resistance" against the FSF/GNAT maintainers, others
> (comprising leading GCC developers) are actively bashing them.
> 
> So, if you think you can change something about this situation, I'd
> strongly ask those people who want GNAT+RTEMS/Ada, to take care about
> it, to carefully test future GCC/GNAT versions and to contribute bug
> fixes.

This is good advice and the best long-term solution.  In general, if you
can actually (1) build a native GNAT and (2) build the cross CPU-rtems
GNAT, you usually have a fully working Ada compiler.  But getting
both those built on an arbitrary CPU-rtems is iffy on newer gcc
versions.

With all that said, if you have specific problems we won't hesitate to
help.  We can try to address RTEMS specific issues and submit PRs for
those which are more general.

> Ralf
> 
> 

--joel
-- 
Joel Sherrill, Ph.D.             Director of Research & Development
joel at OARcorp.com                 On-Line Applications Research
Ask me about RTEMS: a free RTOS  Huntsville AL 35805
    Support Available             (256) 722-9985