COMPILER VALIDATION

COMPILER VALIDATION

[Malcolm Petcher - 466-4146]

I can see the benefit of having the British MoD allowed to validate compilers, 
but I see no reason for them to establish their own validation suite.  That's 
courting disaster.  If there is to be only one specification for the language 
there should be only one validation suite.  If MoD has the need for some 
different or additional tests, it seems they could work together with the US 
DoD ACVC to establish a common validation suite acceptable to both.

Malcolm Petcher

compiler validation

[Kenneth Ng]

I've got two questions on Ada compilers.  First, an associate of mine
recently said that Ada compilers are so big that every compiler
produced has been certified and decertified at least half a dozen
times.  Is that an exageration or an understatement?  I've always
been under the impression that once certification has been issued
that it was permanent until the ANSI standard comes up for review.

Second, is there something akin to an implementers guide available
for Ada?  (Please don't tell me I'm crazy, I already know that)
If it is available electronically, is it available to someone
without FTP capabilities?

compiler validation

[Kenneth Ng]

I've got two questions on Ada compilers.  First, an associate of mine
recently said that Ada compilers are so big that every compiler
produced has been certified and decertified at least half a dozen
times.  Is that an exageration or an understatement?  I've always
been under the impression that once certification has been issued
that it was permanent until the ANSI standard comes up for review.

Second, is there something akin to an implementers guide available
for Ada?  (Please don't tell me I'm crazy, I already know that)
If it is available electronically, is it available to someone
without FTP capabilities?

Re: compiler validation

[stt]

Ada compilers must be revalidated periodically (it used to be
every year; now it is every two years).  This is because the
validation suite becomes more stringent as time goes on,
and the language review board "refines" the official
interpretation of the language reference manual.  This
is not directly due to the size of the compilers (though they
are admittedly big), but rather both are due to the
size of the language and the extraordinary concern for precision
in the language definition.

Yes, there is an implementor's guide (available on the ArpaNet
somewhere), but it is not exactly a compiler-writer's cookbook.
Rather it is a series of tips about what to watch out for,
as well as motivation and explanation for parts of the validation
suite.