GitLab

In some cases systemd doesn't like our behavior of starting emulab services
from rc.local and killing them from the testbed service.  There's really no
reason to be doing it this way under systemd anyway, so make the testbed
service do both.

The IMQ driver has been patched to allow devices to be created
dynamically.  They will be automatically deleted when the container is
shut down, as will the macvlan devices.

control node. Added target genirack-install.

additional bridges. Add -b option to specify the name of the bridge to
create (which was defaulting to xenbr0). So, to create a bridge for
eth1 (say, attached to a switch): xenbridge-setup -b xenbr1 eth1

slice urn, slice uuid, or a date. 

Fix up the manifest popup windows. Been broken for a long time. New
approach is to stick all the xml into hidden div sections, and then
suck the XML out into a new popup window. The wonders of javascript.

go in before Monday. Once a CM updates to stable, it will start
sending aggregate/manifest history records to the ClearingHouse, which
will store all records for all CMs. The CM actually goes back and sends
all the records from the beginning of time; hopefully there will not
be any serious problems with the very early records at the early CMs.

Still to be done, is an interface at the CH to export the records to
someone with a proper credential. But that can wait till next week.

push to stable. These changes properly record URNs in the aggregate
and manifest history tables.

Previously a number of invalid transitions were allowed due to the way the
check was coded. Note that this change has no functional impact as detected
invalid transitions were allowed anyway. Essentially this just changes what
gets logged.

geniuser).

accept an rsa key.

At some point in time, the preceding call to /etc/rc.d/nfsd starting firing
up mountd as well. So the explict /etc/rc.d/mountd call just after may fail.

images to croak.

Also modify grub config to use the vga console only when VGAONLY is
set.

/etc/emulab. So stick the isvgaonly file in both places!

Use CP instead of MV when setting up pxeboot.

Fedora 15 "df" shows two lines for "/" mount; we ignore the "rootfs" line.

Well, isn't that special. We had the arguments to memcpy backward (probably
dating from the time when we use bcopy instead). So instead of copying bytes
of goodness into the zeroed nonce buffer, we were copying bytes out.
Net result: a nonce of zeros.