GitLab

typedef TPM_NONCE is already defined by TrouSers.  I renamed it ETPM_NONCE.

Conflicts:
	tmcd/tmcd.c

they unregister expired slices, but we can't really rely on that.

This code includes various useful TPM tools along with doquote.c and
tpm-signoff.c which we use in the MFS to proceed down the secure
bootchain.

This allows swap modify to work once more.

In python 2.6, the implementation of serve_forever changed and it invokes
a new method, _handle_request_noblock, instead of handle_request.  The
former is not overridden by m2crypto resulting in m2crypto specific
exceptions (SSLError) not being caught.

the vinterface instead of the interface.

Mostly this means a new package set emulab-*-3.1 and some corresponding
tweaks to the install scripts and elabinelab scripts.

PLEASE NOTE: a 7.3 install uses PHP5, Perl 5.10.1 and Python 2.6.

It is deprecated as of python 2.6 and we don't appear to actually use it
anyway!

/etc/rc.early is no longer supported so we set it in /boot/loader.conf instead.

Also, get group/passwd files back in sync with 7.3 release.

Note to self: change DB so that the default sshdport for nodes is null
instead of some silly value like 11000.

port to a vnode. This can lead to confusion in the experiment if the
user does not reboot all the nodes. This fix turned out to be more
complicated then I thought, and I have not figured out how to apply it
to trafgens port numbers.

experiments, even on other testbeds. This is because snmpit -o does
not untrunk ports, and I do not have time to fix this now. Without
this change, the shared nodes holding experiment cannot be changed.

This is part of cleaning up the name space of pubsub.  Previously, this
program was getting the "info" call from libpubsub and that version is going
to become private.  So use the one in libtb instead.

Includes hack to "fix" upstart dependency problems:

A) eth* interfaces are probed by ifupdown serially instead of in
   parallel.
B) Legacy initscript support waits for the the control net probe
   to finish (with either success or failure) before running legacy
   initscripts.

The stand-alone desktop client is no longer supported - the GUI applet
is nearly as good (just doesn't allow user to ssh to the nodes), and
actually works.

Contributed by IBBT/UGent

  As you can see, most of those are in use even during our
  periods of "light" usage.
At Mike's request since "it isn't true now.  It will probably be true
again, but it will take awhile..."

That is when more than a fixed threshold of nodes (0.05 for now) are
in an invalid state for any particular class (ie "pc3000s", "d710s",
etc) only invalid that class for the time period, instead of all the
plot data.

Along the way fix various bugs with handling NaN (how invalid states
are flagged).

to avoid running out of memory.