Patch for Ticket 6344

[Greg Hudson]

On 04/29/2013 10:40 PM, bwleake wrote:
> I've pushed a second
> commit to that branch that doesn't have any issues with the checking
> script, as well as updated the kadmind.rst and krb5kdc.rst files.

Thanks, I'll try to take a more serious look when I can, although it may
be a while.  Feel free to rebase your branch and/or squash commits as
much as you like; it won't get in my way.

> Attempting to run `make man` in src/man resulted in several unrelated
> .man files being changed.

Yeah, don't worry about updating the man pages, just adding text to the
.rst files.

> Could you elaborate on your concerns for testing the changes?

The best automated tests are ones we can include in "make check", which
needs to be able to run as an unprivileged user without making any
changes to the system beyond creating files in the build directory.

127.0.0.1 is always a loopback interface, and we can explicitly bind to
that and test that it works.  127.0.0.2 is typically not a loopback
interface.  On some systems you can find a second local IP address by
resolving the local hostname or using 127.0.1.1, but there's no guarantee.

We also prefer tests which don't involve delays (or else the aggregate
runtime of "make check" becomes long enough that developers stop running
it as often), and testing lack of connectivity might be difficult
without incurring a substantial delay.