What's the state of current Win32 PC Ada compilers?

What's the state of current Win32 PC Ada compilers?

[John *NOSPAM* Bunk]

Hello,
I was wondering if anyone in USENET land had any real experience with using
Aeonix or RR Software's Win32 capable Ada compilers for a large Windows NT
project?  If so, was the compiler fast/slow?  Stable/Buggy?  Code produced
had compiler induced bugs / clean?  Development environment decent / hard to
use?  Debugging facilities?  Would you use these compilers again?

Thanks for the input in advance,
John

Re: What's the state of current Win32 PC Ada compilers?

[Stephen Leake]

"John *NOSPAM* Bunk" <jkb@nospam.vcnet.com> writes:

> Hello,
> I was wondering if anyone in USENET land had any real experience with using
> Aeonix or RR Software's Win32 capable Ada compilers for a large Windows NT
> project?  If so, was the compiler fast/slow?  Stable/Buggy?  Code produced
> had compiler induced bugs / clean?  Development environment decent / hard to
> use?  Debugging facilities?  Would you use these compilers again?

I'm using Aonix ObjectAda 7.1.1 on a medium Windows NT project. The
user interface and a large part of the system is in Borland C++ 5.02;
some models are in Ada. I also use GNAT Ada; I have not used RR's compiler.

Most of the system is in Borland because ObjectAda wasn't out when we
started. As soon as ObjectAda 7.1.1 came out, I started writing new code
in Ada, because I just can't stand C++. I'm using ObjectAda instead of
GNAT because it's easier to write a DLL, and I can afford their
maintenance fees. 

I have found bugs in ObjectAda; none are showstoppers, except that the
debugger is practically useless. Fortunately, I've been able to use
GNAT and gdb to find any Ada bugs, then recompile with ObjectAda. All
compiler bugs were compile-time (ie, it won't compile legal Ada); I
have found no bugs due to bad code generation. Aonix just released
7.1.2, which should fix some of the bugs I reported.

The compiler is fast enough for me (Windows NT 133 Mhz Pentium 64 Meg
RAM). Not quite as fast as GNAT.

The IDE that comes with ObjectAda is a pain - use NT Emacs from
http://www.cs.washington.edu/homes/voelker/ntemacs.html . I use emacs
to develop the C++ code, too.

Aonix support has been adequate, once I figured out how to pay for it.
They have not sent me any patches (they don't promise too), but they
have responded to bug reports in a timely manner, and as I said above,
I didn't find any bugs that I really needed patches for.

Aonix is redesigning the debugger for version 7.1.3, and they are more
committed to the Windows platform than ACT (they sell supported Win32
and MFC bindings), so yes, I will continue to use ObjectAda.

-- Stephe

Re: What's the state of current Win32 PC Ada compilers?

[Steve Doiel]

>I was wondering if anyone in USENET land had any real experience with using
>Aeonix or RR Software's Win32 capable Ada compilers for a large Windows NT
>project?  If so, was the compiler fast/slow?  Stable/Buggy?  Code produced
>had compiler induced bugs / clean?  Development environment decent / hard
to
>use?  Debugging facilities?  Would you use these compilers again?
>
>Thanks for the input in advance,
>John
>

We have been using ObjectAda as the target for 250K SLOC of code that was
translated from a dialect of Pascal.  The majority of the code was tested
quite extensively in its Pascal form.

The compiler is quite acceptable.  It runs faster than GNAT but slower than
Delphi.  I have not seen any bugs introduced by the compiler.  The
development environment lacks  a number of bells and whistles, but it
appears that Aonix is being quite conservative on releasing new features
before they are ready.

I started with ObjectAda 7.0 and have experenced a few updates 7.1, 7.1.1,
7.1.2.

The single factor that I have seen the most lacking of all PC based Ada
development systems is the Debugger.

In ObjectAda 7.0 the debugger was basically unusable.  In version 7.1 the
debugger started to have some (though little) usability.  Version 7.1.1 saw
great improvement.  I just received 7.1.2 and have not had the occasion to
exercise the debugger to see any changes form 7.1.1.  A completely rewritten
debugger has been promised for ObjectAda 7.1.3 which is due to be released
before the end of the year.

The team at Aonix has been very responsive to requests for reasonable
changes to their system and to finding work-arounds or corrections to
problems.

Yes I would choose them again.

When I selected ObjectAda the three development systems I was considering at
the time were from RR Software, ACT (GNAT) and Aonix.  In my case I had to
sell Ada as well as the development environment to my departement.  This was
much easier to do with ObjectAda than with GNAT.  I ruled out RR Software at
the time since they did not have a debugger available for their system.

I am guessing that the next publically released version of GNAT will include
a more comprehensive debugger than the one included with ObjectAda, but I
won't know until they release that version.  If you have deep pockets, then
supported GNAT may very well be the best route to go.

Other observations:
  Comparing a compute intensive application between GNAT 3.10p and ObjectAda
7.1.1, the GNAT version runs about 2x faster.  I haven't done any
comparisons with ObjectAda 7.1.2 and I know that they have done some
optimizations.

  With regard to tasking and protected types, a simple test shows that
ObjectAda's 7.1.2 task switching and protected operations are about 2x
faster than GNAT.  Your mileage may vary.

I hope this helps,
SteveD

Re: What's the state of current Win32 PC Ada compilers?

[Larry Kilgallen]

In article <360c75d9.0@news.pacifier.com>, "Steve Doiel" <nospam_steved@pacifier.com> writes:

>   Comparing a compute intensive application between GNAT 3.10p and ObjectAda
> 7.1.1, the GNAT version runs about 2x faster.  I haven't done any
> comparisons with ObjectAda 7.1.2 and I know that they have done some
> optimizations.

One area where Aonix claims they have significantly improved their
performance for ObjectAda 7.1.2 is Text_IO.  Personally I have not
done enough Text_IO with ObjectAda to notice a problem but others
have complained, and Aonix seems to be considerably affected by the
feedback on their mailing list (open to all, regardless of their
degree of customer-ness).

Larry Kilgallen