Logged In: YES 
user_id=31435

Neal, as I said in the checkin comment, I expect it's no
less likely that we'll get screwed by goofy platform values
for LONG_MIN and LONG_MAX than that we /got/ screwed by an
"over ambitious" compiler optimizing away "result ==
-result", so I'm not sure backporting is a good idea.  I
stuck in an assert that might fail if a platform is insane;
if there are no reports of that assert triggering, then I'd
feel better about backporting the change.