🚀 go-pugleaf

In article AA02010@jc3b23.UUCP, gwl@jc3b23.UUCP (George W. Leach) writes:
>	I've got a quick question.  The latest GNU Bulletin (6/92) 
>includes a description of G++, which indicates that AT&T's cfront is 
>diverging from the ANSI efforts.  This is presented as the reason why 
>G++ and cfront are not in sync.  Furthermore, there is an indication 
>that G++ is tracking ANSI.  The version of G++ listed is 1.40.3.  The
>version of cfront spoken of is not listed (3.0?).

Diverging would be a incorrect statement.  There are significant
differences between the behaviour of cfront 2.1 and the ARM (apart from
the fact the cfront 2.1 does not implement templates and exceptions).
The most obvious difference is that nested classes definitions are
given global scope in cfront 2.1.  This and most other (if not all)
discrepencies are fixed in cfront 3.0.  USL are actually moving closer
to conformance with the ANSI draft.

GNU C++ should, however, be a little more nimble in tracking the draft
standard than USL's cfront since releases of GNU C++ are rather more
frequent than releases and don't have to percolate out through
resellers (although cfront 3.0 was officially announced in October 1991
most resellers still appear to be at version 2.1).  Given the length of
time between USL releases it is unlikely that USL will attempt to track
the ANSI draft standard in detail; they may stick with their interpretation
of the ARM if there is any conflict with the ANSI interpretation until
the ANSI draft stabilizes.  Any differences that would result from this
are likely to be minor (so far).

In any case, it would be unwise to expect total source compatibility between
different implementations of C++ until the ANSI standard gets the official
stamp of approval.  

	Chris Flatters
	cflatter@nrao.edu