I hope this is the right place...

A long time back, I complained about some limitations in the number
of hard links being obeyed correctly.  The problem looked like a
SCO emulation one, and disappeared of its own accord with an upgrade
from NetBSD 1.3.x to 1.4.x.

We're still running on 1.4.x and we picked up the same anomaly
(attempting to open a file and reaching another) only when the link
count was exactly 128.  Once the number was exceeded, the problem
disappeared.

Should I send-PR this?  It is difficult to reproduce, as one has
to open that particular file a number of times, and possibly in
the context of many other opens (Cobol application, under SCO
emulation) before tickling the flaw.  At a guess, it is not the
number of open files, more likely the number of times open (or
something else) has been invoked in the process group, because
we're running the same application repeatedly over a large number
of filesets in different directories, which is why we use hard
links, that particular file has to appear in each directory, but
with a few versions to pick from.

Does anyone see the relevance of 128 in this picture?  I'm almost
certain the SCO emulation is not involved, but I guess one can't
entirely discount the possibility.

++L