In article <YkSwSr302QU6AVKRAs@highland.com.au>, Andrew Cagney <cagney@highland.com.au> writes:
[ ... quad arithmetic problem ... ]
> Is any one aware of other problems similar to this one (before I spend
> another day chasing a red herring ...).

See also problem report lib/1005, where I reported a problem with quad
multiplication.  Unfortunately the comments preceding the program
example is slightly misleading as they were meant for another version
of the test program.  It also turns out that this is an architecure
dependent or maybe an endian-problem, as the test program runs fine on
an hp300-series machine, but produces the wrong result on my i486.

The results from the program are:
i486:
(quad) x=80000000, y=2, z=ffffffff00000000
(little) x=0:80000000, y=0:2, z=ffffffff:0

hp300:
(quad) x=80000000, y=2, z=100000000
(big) x=0:80000000, y=0:2, z=1:0

Maybe this problem and your problem are related?

					-jarle
PS.
-- 
"This isn't right.  This isn't even wrong."
			-- Wolfgang Pauli, on a paper submitted by a physicist
			   colleague.