Future of 32 bit time_t?

[Hal Murray]

> I think we can hang with Plan A for now.

Sounds good to me.

I think I have all the leap stuff converted.

I had troubles with one test until I figured out that it was testing a 
feature that I didn't understand and had removed.  There was an option to 
skip loading leap seconds that were older than the build date.  I can't see 
any need for that.  (or even a use)

There was only one place that printed a time_t.  Some systems with 32 bit 
time_t didn't complain.  Some did.  It was easy to fix.  (There may be others 
in a build option I haven't tested yet.)

It will take me a day or two to clean things up and test some more.

I haven't undone any other time64_t usage, just the stuff I bumped into.

---------

We need to figure out how to test leap seconds.

In particular, we need to test the smearing stuff.  It's currently using cos. 
 I think google and friends switched to simple linear.  We should try to find 
out if they have converged on the time-spam to smear over.



-- 
These are my opinions.  I hate spam.