Re: Help Me Please :)

[Robert Dewar <robert_dewar@my-deja.com>]

In article <2000Mar3.183321.69279@ludens>,
  gazso@ludens.elte.hu (Erdelyi Gaspar) wrote:

> I think that GNAT is not comparable to DEC Ada (for OpenVMS)
> which provides very useful error messages. DEC Ada is Ada83
> only and the prices are not comoparable, too...

Well the prices are not that far from comparable, but that's
another issue anyway. With regard to error messages there are
four cases of interest:

1. Cases where DEC Ada 83 simply gets confused, and GNAT does
not. There are quite a few of these, one notable example is
the GNAT circuit for dealing with the semicolon/IS confusion.
All other compilers I know can get confused by this, including
DEC Ada 83.

2. Cases where GNAT gets confused, and DEC Ada 83 does not. I
don't currently know of any such cases, but that does not mean
they do not exist. If you know of such cases, it is always
useful to submit them to report@gnat.com, we are always working
to tune the error message circuit (indeed giving good messages
is a hobby of mine :-)

3. Cases where neither compiler gets confused and the error
messages are comparable. Having worked through the entire
DEC test suite, which contains thousands of examples of
error situations, I would say this is probably the most
common case.

4. Cases where neither compiler gets confused and the
error messages are substantially different. These are
interesting cases to look at. Sometimes there is simply
a difference in style. GNAT aims at trying to keep error
messages short and informative, whereas DEC Ada 83 often
gives rather voluminous messages. In GNAT, you can more
nearly approximate this verbose mode by using the -gnatf
switch. But there will still be differences, some of these
might just be a matter of taste, some may be cases where
real improvements are possible (there are certainly cases
in the DEC test suite where the GNAT message is more useful,
I don't know of any cases left where I consider the DEC
Ada 83 message significantly better in this test suite,
because we used the suite to tune up such cases in GNAT :-)

Here again, in case 4, it is very useful if you send in
cases where you think an error message can be improved.
This actually is a much more general request than just
comparing these two compilers. Any time you get a compile
time message that you think could be clearer, send it along.
We may well agree, and we may well be able to do something
about it.

Robert Dewar
Ada Core Technologies


Sent via Deja.com http://www.deja.com/
Before you buy.