Re: ACT announces first 100% validations of Ada 95 compilers

[dewar@merv.cs.nyu.edu (Robert Dewar)]

Mats asks

<<Did optimisation cause problems for the validation ? I'm asking this
because we have had many optimisation related problems with non-GNAT
compilers, to the point that we cannot deliver any code compiled with
optimisation.>>

No, optimization did not cause toruble for validation, because, as Richard
noted we chose to do the validation at -O0. You have to choose one
specific set of switches for validation, and we chose -O0 because it
makes debugging easier.

Obviously it is impractical to run the suite with every possible combination
of compile switches, but it is useful to run the suite at different
optimziation levels, and we are in fact in the process of doing that
now, and at some point will do it systematically (we run the ACVC every
night on several machines, so our plan was to use different options
each night :-)

In practice, we have known cases of bugs appearing only at -O2, and
also cases of bugs appearing only at -O0, but both events are rare,
and lots of our customers are routinely using -O2 for delivered code.
The GNAT compiler itself is tyically compiled with -O1.

Robert Dewar
Ada Core Technologies