Re: GDB Woes Continued...

[rracine@draper.com (Roger Racine)]

> <<Sounds like a fundamental problem in your beta testing program.>>
> 
> Not really ... we certainly could take the attitude of not publicly
> releasing the NT version of GDB until we were really sure it worked
> on *all* Win95 systems (there is a simple method of ensuring that
> we meet this requirement :-) :-) 
> 
> In fact the current version seems quite reliable on NT, and also seems
> to work on many (most?) Win95 systems (note for example Martin Carlisle's
> reported success at Air Force academy with a variety of machines and
> users). It is at this stage hard to know if the reported problems are
> oddities in particular versions of Win95, or installation/configuration
> problems (we always have the experience that many problems come from not
> reading all the documentation and installation instructions -- no way of
> knowing that this is the case here, but also no way of knowing that it is
> NOT the case).
> 
> Our strategy for this release was to release it as soon as we heard from
> some significant users (including Martin) that the fundamental Win95
> problem had been solved. 
> 
> Anway, you should consider that the status of the public release of
> GDB for GNAT/NT is that it works OK on NT, and may well work on Win95,
> but there are no guarantees (actually there are never any guarantees
> in using the public versions of GNAT technology in any case!)
> 
> We haven't been able to duplicate these problems so far, and
> we have run GDB on a variety of Win95 machines. One observation is
> that we are running the latest version of Win95 on pretty much all
> machines, not the broken version you can buy in stores, so perhaps
> that is one variable.

This is my first experience trying to use a compilation system on a
Windows 95 PC (I have used Ada on pretty much every other OS), so I have
no idea if these types of problems are usual with other commercial
compiler products (I picked my words carefully.  GNAT and GDB are
commercial products, right?  One happens to pay for the support instead of
the license, but that is, supposedly, the only difference).  

Perhaps ACT (and every other vendor; this advice is universal; the actual
suggestions are specific) should make sure that their testing is
realistic.  

1) Most people do not have access to the beta versions of commercial
OSes.  They have the "broken version you can buy in stores".  And those
computers have -lots- of other programs installed.

Do you have any idea on what configurations of Windows platforms your
testers are testing?  Are they all using some pre-release version, or a
"clean" machine used only for Ada development?

2) If you care about the unsupported folks out here (which it seems you
do, or you would not be responding) have relative novices (with the
platform and the download procedure) go through the installation process
and the documentation available on the server.  Many people are going to
simply download the .EXE files and install.  That is what they are used to
for commercial products, shareware and freeware.  Do not suggest that they
build the product from source code.  They have better things to do with
their time (and disk space).

3) Do not assume that the users of your unsupported version are not
important to your economic success.  Some are evaluating the language, the
product, or the company.  If they find a robust product, at least as good
as what they are used to, the language will get more converts, and you
will get more supported customers.  Lots of winners!

Roger Racine