Re: Aonix Ada 95

["Keith Allan Shillington" <keith@sd.aonix.com>]

> Robert Eachus said
> ...  (For those who have never dealt with the ACVC, the
> B-tests are--in 1.x--about 1000 test which not only require you to
> find every error at compile time, but to diagnose it correctly.  Some
> tests have/had hundreds of errors. Every compiler vendor I know spent
> most of their effort during validation getting the B-tests right.
> Most of the rest was spent on C-tests, and A,D,E and L collectively
> accounted for at most 1-2% of the effort.)

Me thinks your numbers are overly pessimized, but, that is all in the past
anyway.  Anyone who bothers to look at the 2.X ACVC suite will find a
stronger emphasis on C and L tests.  It was our expectation in writing them
that the vendors would find a little more need to focus on the implementation
of the executable bits.  And, let us not discount the value of a good error
message from a compiler.

Robert Dewar deftly points out:

<<Given that every line has a comment on it that says ERROR, I have no doubt
that the amazing Kruchten compiler would be able to read through a B test
and announce ERROR on all appropriate lines :-)>>

Takes me back to the days of the Basic interpreter that was possessed of the
single error message: "Huh?"