Re: GNAT 3.13p build for Solaris 8_x86

[dewar@gnat.com (Robert Dewar)]

Kilgallen@eisner.decus.org.nospam (Larry Kilgallen) wrote in message news:<pffjeWzAZDhz@eisner.encompasserve.org>...
> I was under the impression that the validation tests are available to
> all.  If you were to run the tests yourself, some people might not be
> satisfied they were not run by the official testing site ($10,000 a
> pop ?).  But presumable other people, including especially _you_ 
> would be satisfied.


And even if they were run by the official testing site, this would
not make the compiler validated, since to have an officially validated
compiler, you have to have some kind of formal guarantee that the
compiler you are using is the one that was validated, which typically
requires a formal commercial contract of some kind.

But indeed it is useful to run the tests informally, and that is
certainly possible, since the test suite is indeed freely available.
However, it is far from trivial to setup this test suite (we have
probably invested in aggregate 6-12 person months setting up our
testing environment, but that environment is much more complex than
is required for this kind of use.

I think it is still asking a lot for someone to try to set this up
on their own (certainly running the validation tests is a much harder
task than building GNAT from sources).

One positive thing here is that Laurent Guerby has volunteered to setup
a suitable test environment for GNAT version 5.00p (that will be the
designation of the GNAT sources that are placed at the GNU site that
are compatible with GCC 3.x). This test environment will include the
ACVC tests, as well as a set of tests that we at ACT have generated for
testing GNAT specific features (the main test suites we use, the
DEC test suite, and the much more extensive ACT suite, are not
distributable, since they contain proprietary customer code in both
cases).

Robert Dewar
Ada Core Technologies