[krbdev.mit.edu #7522] krb5-1.11, patch 1 of 4

[Greg Hudson via RT]

All four patches were mangled by your email client.  If you send them as 
text/plain attachments rather than pasting them inline, they might come 
through better.  Alternatively, we actually prefer getting code 
contributions as github pull requests these days, but that's not 
mandatory.

For this patch:

* What is the purpose of the ulog_resize() change?

* What is the purpose of the recsize parameter to ulog_init_header()?

* In krb5_db_put_policy(), if we need code to check whether the policy 
has changed, it should be factored out into a predicate function, which 
should handle policy tl-data.  But I'd much rather eliminate the need to 
check.

I believe non-substantive policy changes happen frequently because of 
the policy refcount.  Nico and I have discussed getting rid of the 
refcount, and dealing with the problem of dangling policy references at 
delete time (or just letting dangling policy references be equivalent to 
no policy).  It is already easy for policy refcounts to get out of sync 
between master and slaves for a variety of reasons, which is a problem 
when a slave is promoted to the master.