GNAT non-validation

GNAT non-validation

[tmoran]

I was surprised to hear the GNAT contract explicitly eschews
validation.  What exactly are the deliverables and acceptance
tests for GNAT?
-----------------
PGP:Thomas Welling Moran <tmoran@bix.com>

Re: GNAT non-validation

[Richard Kenner]

In article <3k5vcu$sgc@news1.delphi.com> tmoran@bix.com writes:
>I was surprised to hear the GNAT contract explicitly eschews
>validation.

There wasn't much choice since the validation suite was not expected
to be available by the end of the contract period.  There is also a
belief that validation is more appropriate in a commercial, rather
than university setting, and indeed ACT has a contract from SGI to
validate GNAT for SGI systems.

>What exactly are the deliverables and acceptance tests for GNAT?

No acceptance tests were specified in the contract.  The deliverables
were an Ada 95 compiler (minus some specified features, which the GNAT
team implmented on their own time) for SunOS on Sparc and OS/2 on PCs.

Re: GNAT non-validation

[tmoran]

>No doubt you are aware that the contract is written as it is for
>purely political reasons
No, I am innocent of any knowledge of the process that led to GNAT.

>GNAT will be validated, first on the SGI, because SGI will pay
>ACT to do it.
Sounds like a mighty inexpensive way for SGI to get a compiler!
I would think lots of hardware and OS vendors would see that.  And it
will certainly help portability when all machines have the same ACT
compiler, with the same design decisions, implemented/not-implemented
parts, etc.

>My understanding is that several other platforms will be
>validated at the same time, presumably SunOS4.x and OS/2.
Are Sun and IBM paying for the SunOS4.x and OS/2 validations then?
Maybe Microsoft will pay for the NT validation?