Re: Verdix Corp ADA-language development systems

[baur@spp2.UUCP (Steven L. Baur)]

in article <2410@elrond.CalComp.COM>, bobb@elrond.CalComp.COM (Robert J. Boulanger) says:
> I'm looking for some information about ADA development systems from
> Verdix Corporation.
> ... I'm especially interested in their documentation.  How good
> is it, and recommnedations do you make for improving it?

The documentation on the whole is fair.  I would like to see more details on
how the implementation is done (because of so many potentially expensive
features), and more details on available libraries would be nice too.

All run time libraries are delivered in source code, and that is basically all
the documentation you get.

> How is
> Verdix's customer support and training?

Customer support is weak at best.  They answer the phone rarely, and are
not too terribly helpful.  There is hope though.  The last time I dealt with
customer support, they spent a good deal of time looking at my problem, and
responded formally with written up bug reports.  What will happen next, I
don't know.

Customer training?  If there exists such a beast I have never heard of it.



Currently there exist many problems with verdix ada.  First and foremost is
the release of VADS 5.5 has been suspended because generics do not work,
and the optimizer produces incorrect code.  We bought 5.5 many months ago
(at least January I believe), and still have not received the real 5.5.
There exist many situations where use of NFS file systems completely
confuses the compiler such that it thinks that nothing is up to date
(this can be fixed by touch'ing the objects, but is ugly).
A.make is easily confused into thinking there are circular dependencies
when there aren't.  (Since VADS does not produce .o's in any kind of simple
fashion, the use of make(1) is largely precluded).

A.pr crashes when it sees a form feed.  Verdix ada prints an error message
when it sees a 'pragma page'.
Earlier versions are worse.  Verdix 5.4.1 (the last major release before
5.5) handled generics better, but the run time system had no concept of
critical regions of code.  The interrupt handler could and did interrupt
itself.

The code generator sometimes produces bogus code for arrays that will
generate an exception (bus error, segmentation violation ...) at run time.
This has not occurred in any kind of a fashion that I can repeat, but the
symptoms are such that the array name if treated as pointer, will result in
correct code generated.  The bug always occurs when compiling output from
the Intermetrics ada-yacc which is used in all the Intermetrics tools.

Summary:
If you do not use optimization, and use generics in limited fashion,
Verdix works fine.

Steve  -- ucbvax!trwrb!trwspp!spp2!baur