goal's in obstacles: move towards the edge of the exclusion zone where
the path intersects and wait before trying again; if the obstacle is a
robot, move to the edge perpendicular to the robot to clear the way
for the other guy.

Make use of the stargate's onboard watchdog timer.

	* mote/sgtools/GNUmakefile.in: Standard makefile for sgtools.  Run
	configure as described in doc/cross-compiling.txt to make it
	compile for the stargate.

	* mote/sgtools/wagthedog.C: Enable and feed the stargate's onboard
	watchdog timer.

	* robots/primotion/pilot-wrapper.sh: Rewrite in perl.

	* tmcd/common/watchdog: Change the path of the battery log file.

	* tmcd/linux-sg/rc.stargate: Start the stargate watchdog tool and
	run the pilot-wrapper script instead of pilot directly.

replaced with a continuous nonlinear controller. The path planner is bypassed at the moment.

positions are on both sides of the 0,2PI boundary).  This fix should solve
that problem.  Also a bunch more debug output.

"isPopupTrigger" instead of checking the button number.  Certain
people with only one mouse button need this.

the vmc-clients when a robot is lost.  Bump merge tolerance up.

I fixed a couple of minor problems, but mostly this worked fine. Note that
I have tested this with the installed perl, *NOT* perl 5.8. I am just
making sure this stuff gets committed before too much more bitrot sets in.

might different, as reported by the newnode MFS. At present, I need
the ordering (and thus, the iface names) to agree with the outer
emulab since that is how I request vlans to be setup. This is not
very pleasing, and I will probably have to change it to use MACs
instead. For now, reorder the new_interfaces based on what switchmac
(the proxy) tells us.

what locpiper to connect to. Locpiper now binds a dynamic port, and
stores that into the DB, and the php script expects to get that pid,eid
back.

Changed the showexp page so that pid,eid are passed to the robotrack
top level page.

Changed the robotrack page so that if it does not get a pid,eid, it looks
for an experiment running (a locpiper). If none is running, let the applet
start of course, but there will be an error dialog when it starts. You can
still play with the applet of course. If there is more then one experiment
running, then just pick first one. At some point I might add a page to ask
which experiment, but its low priority.

adding the nodes, or else they check in and get a dynamic address
again. I'm kinda surprised this has not come up before?

in debugging.  The frames are dumped to disk on SIGHUP or if an
MTP_SNAPSHOT packet is received.  Fortunately, we can do this since
there is just barely enough CPU and RAM to do this for all four
cameras on a machine.

to the fault detection code.

obstacles that represent a robot.

thing; restart mysqld after it gets a signal other then TERM.

* Use the "role" slot of the interfaces table, instead of joining with
  node_type, to find the control interface. Must do this inside an
  inner Elab of course, since the node_types table does not reflect
  how assign selected the interface we use for the "control" network.

* Use !is_numeric() instead of is_string() to determine if the naming
  patter in alpha based instead of number based ("1" is a string).

  RobTransparentBase, just modified things to blink leds differently
(i.e.,
  toggle the yellow led on packet receive), and to
  dump original sender id, rssi, and AM type for any packet the mote
hears.
  There's an sflisten.c program that will decode and display these values.
  It requires several of the tinyos C sf header files and code.

RobTransparentBase, just modified things to blink leds differently (i.e.,
toggle the yellow led on packet receive), and to
dump original sender id, rssi, and AM type for any packet the mote hears.
There's an sflisten.c program that will decode and display these values.
It requires several of the tinyos C sf header files and code.

ISUP or else they leave off at SHUTDOWN, which later confuses swapmod
(tbswap) which waits for them to reach ISUP, but of course they never
do. This is arguably a misfeature of os_setup, which should probably
not wait for nodes it didn't reboot, but I want to ponder that change.
Besides, doing this causes the nodes not to say SHUTDOWN in the web
interface.

Another change to add there (web interface) is to not say "possibly
down" for elabinelab nodes.

their serial lines.  Useful for testing packet reception, power usage,
etc.

enough and seem to work.  Currently, the moving average window size is set
to 5 frames (in testbed/event/sched/event-sched.c); a window size of 10
produced too much position lag.

not been created yet. This happens during the bootstrap of the initial
groups. Not a big deal, it all works out in the end (the groups get
set later in a subsequent call to setgroups).

file if it does not exists, and then append to it. In our elabinelab
setup, there is always an existing root authkeys file, so we missed
this error.

negotiate:

# Divide the nodes up into used and unused based on the status of
# their WhOL experimental interface:
#
# enabled=yes     port in use
# enabled=no      not used
#
# If a port is "in use" we assume it has talked to the switch and
# no speed/duplex changes are needed.
#
# If the port is unused, we assume that the port will try to auto
# negotiate with the switch when enabled.  In this case we set
# auto-negotiation on the switch to maximize the chance that we will
# sucessfully communicate with the card.

Also, fix a small missing paren from Keith's contributed version.

need it.

trunks for convenient checking against.

generate month by month files, and then a little index to include into the
php script.

I know, this is really important stuff! But, I was reminded to do this by a
message that Christopher Johnson (the young one) sent on April 29th about
his project. His code was quaint, but I did not think we needed a new
python script when it was much easier just to roll it into the existing
perl script that imports the commitlog.