On Fri, 31 Oct 2003 14:21:31 -0800
Stephen Hemminger [off-list ref] wrote:

Actually, the before the "optimization" went in I did testing between
old 2.4.x and 2.6.x as well as standalone comparisons.  The problem is
a compiler screwup, that probably isn't worth investigating further.

See Mark Huth's email, I think his analysis is correct and
that the two pieces of C code are really doing different
things.