[comp.lang.c++] AT&T C++ 2.1 bug fixes

billri@arte.sos.ivy.isc.com (Bill Rizzi) (04/04/91)

	My apologies if this has has been thrashed to death already,
but I've been off the net for months.

	I have recently finished porting AT&T C++ 2.1 to the following
platforms:

	Box		CPU	OS		C compiler

	IBM PC		80x86	MS-DOS		Microsoft C
	Unisys B2x/B3x	80x86	CTOS		Microsoft C, BTOS C
	Unisys U6000	80386	UNIX System V.3	"standard" C

	I have a list of the bugs I encountered and am wondering if
there is a similar list on the net.  Has AT&T expressed any interest
in receiving bug reports?

	Most of the bugs had to do with the bogus assumption that
sizeof(int) == sizeof(char *).  There were some other relatively minor
ones, but the mother-in-law of all bugs live(d) in the "#ifdef DENSE"
code.  The latter is a compile-time option that I used since all but the
UNIX C compilers had identifier name limits of 31.  This option truncates 
mangled names to a maximum of 31 characters.

	The cfront program itself has an annoying tendency to drop core
when fed buggy source code, particularly when a class is used before being
declared.  Unfortunately, on systems other than UNIX, dropping core is
not always a pleasant or acceptable means of termination.  In this case,
it is necessary to grab segmentation fault interrupts or whatever, if
possible.  I guess it's always something.

	Please reply to me privately if this subject has been adequately 
covered.  Thanks...

	Bill

-- 
Bill Rizzi		INTERACTIVE Systems Corporation
billri@ivy.isc.com	26635 W. Agoura Rd, Suite 200
818-880-1200 x2110	Calabasas, CA  91302