GitLab

eventsys changes in previous revision.

knows that some tests failed. Otherwise we say all test completed
okay, and then print a bunch of errors!

so that different types can use different MFSes as necessary.  This
ultimately replaces the quadruple-X libos hack which I removed quite some
time ago.

Some event system changes for linktest and any future things we want to
run with the event system in the swapin path:

	* event/linktest/linktest_control.in: Let linktest be run while
	the experiment is activating.

	* event/sched/event-sched.c, event/sched/rpc.h,
	event/sched/rpc.cc: Don't wait for the experiment to become active
	before loading the eventlist so any system defined agents are
	available to use.  Don't start time unless the experiment is
	already active, let boss do it otherwise.  Send out COMPLETE
	events so 'tevc' can listen for them.

	* event/sched/timeline-agent.c: Send a complete event even if the
	timeline is empty.

	* event/tbgen/tevc.c: Add a '-w' option so that tevc can wait for
	an event that sends back a COMPLETE.

	* tbsetup/tbswap.in: Explicitly send an event to start event time.

to include the date and the experiment index.

      page a bit.

  * *.jpg: images for mobilewireless.php3

interface, and so the textarea on the web page can be updated in
real time as it runs.

expected, then use the current test. If the latency is higher then
expected, then allow for some fudge. Using 3ms fudge, since as Jay
points out, its either really close, or its totally broken. This will
reduce false positives, especially when using endnodeshaping.

groklilo when setting root device

with ugly "undefined variable" messages in the log.  Futureproof the
initialization.

in a cross platform (browser) manner. This is handy when updating
elements from javascript.

the web interface.

Add the ability to set the BIOS version, serial number, and service tag from
the nodecontrol page and command line util.  Must be admin.

a link cause the BW or Loss is out of range; post the messages back
through the event system so that it ends up in the log or on the web
page.

Oops - change length of varchar to 32 (was 64).

Add serial and service_tag columns to the database for tracking machines.

 * Add a "showclass" param to shownodes so that the
   node listing will include only nodes of the indicated class.
   Prepend "no-" to exclude nodes of a particular class.  So:
        showclass=pc,mote
   would show just pcs and motes.

 * Now the real reason I added the above: change the hwdown experiment
   node list to include "showclass=no-pcplabphys" by default so I don't have
   to flip through 600+ dead plab nodes.  Change the URL to "showclass=all"
   if you want to see plab nodes in hwdown.

addresses in the list, use real email addresses. Why? Well, cause I'm
a dope.  Oh, the real reason is that people cannot post to the lists
if we use their @emulab.net addresses cause we close the lists (to
avoid spammers). I did it this way originally cause it was easier;
there is a lot more bookkeeping to do if using real addresses, and I
never consider problem of not being able to post.

lists.  This (harmless) bug has been around a while; needed to move
the call to genelists up a level from setgroups to modgroups, since
that is where we have the information about what groups a user has
been deleted from, and so can tell genelists to regen those project
mailing lists.