GitLab

pickthe interfaces. This allows the user create links without having
to know what native bw is supported.

I had disallowed this in the parser, but I hadn't thought about
LAN nodes, which typically get assigned to switches. So, handle
this case correctly.

We still do *not* handle links between two switches (eg. directly
between two LANs). However, what the semantics of this should be
is totally unclear - do we use the *whole* interswtich bandwidth?
Since we are not likely to hit this in our regular usage, we punt
for now with an error if it happens.

Also, remove the parser check, since we now handle correctly the
case it was trying to disallow.

Fixed several XML bugs. Fixed two bugs where raw URNs were being passed to Emulab lookup functions. I have no idea how those were working before.

Made the obvious change and it seems to work.  I'm sure Keith must have sent
us this change before!

if someone has already done so

Conflicts:

	protogeni/lib/GeniAggregate.pm.in

since it has to be aware of the CM version. Add a Version() call to
GeniAuthority with goes asks the CM what version it is exporting.
Based on that, we know how to do a resolve of a component. Refactored
the code that was used in GeniAggregate when creating tunnels, since
that is where we have to Resolve components. This also turns up in
cooked mode.

Continuine moving towards a urn-only world. If a GeniAuthority or a
GeniComponent does not have the URN set locally in the DB, go back to
the clearinghouse and get it. Error if it is not known, and go bang on
the remote site to update and rerun register_resources.

Checkpoint. All the conversion is done, but I need to refactor the xpaths to account for brittleness in the xpath specification.

sitevariables.

'preamble'.

Mostly for our latest round of nodes since new switches are not yet part of
the main stack.  However, this is probably a useful feature in general so
that you don't bother to probe old switches where you know nothing is being
added.

Handle forced updates, including rebases.  git-cherry is used to
determine if the commits pushed are present upstream (with a different
hash) or not.  Ones already present are excluded from the commit mail
since they haven't changed (i.e., they point to the same blobs the old
ones did).  Ones not present are either new commits or modified commits.
It's non-trivial to tell which is which, and we probably don't care
anyway.  Added text to 'forced-update' email messages to state that the
commits shown are only the new/modified ones and not all that were
pushed.

point; it took me a while to figure this out.

Add 'excluderepo' option to filter out commits that may exist in another
repository, such as commits to emulab public repositories that exist in
emulab-devel already.

Use '*' as the bandwidth on the link line in the .top file to have
assign auto-select the speed for the link - it uses the slower of the
two interfaces selected as endpoints for the link.

Tested, but not heavily.

Known limitations:

* Doesn't work with trivial or emulated links, it's not clear what the
semantics would be.

* Doesn't work on links that go directly to switches, for complicated
reasons of code strcutre. We never use these in practice anyway.

* Not yet supported in XML parser; it should be a special tag or
attribute in that format.

* Might not find valid solutions in a very unlikely case: if a single
node has interfaces of different speeds on the same physical switch,
this code will always select the faster interface that's not already in
use. If the faster interface is faster than the inter-switch bandwidth,
and the slower interface is slower than it, assign *could* find a
feasible solution by using the slower interface.  However, it will never
try this solution.

* Assign doesn't print the speed it picks; its up to the wrapper to know
how fast the interface is. However, this shouldn't be an undue burden,
as assign *never* prints out link speeds.

vlink speed to native interface speed.

Nearly, but not quite completely, functional

is given, it skips bandwidth checks.

Not well tested yet.

The only reason it didn't before was because we didn't have a "portable"
version of disklabel.h.  We have had that for quite some time now...

This reverts commit dfaa3bc2.

I didn't modify the code to grab the password from the correct place,
since this code is for an old version of PLC that isn't used anyway.
The password that was in this file has been changed.

It was not ready to be committed to the emulab-devel repo yet. Oops!

This reverts commit 5771f06b.

page.

Also, remove the 'NEW!' icon from the 5.0 release, which is now about
18 months old.

From branch 'srikanth' of repository:
    git-public.flux.utah.edu:/flux/git/users/srikanth/emulab-devel

I cannot believe we didn't do this earlier..

the right place to set it probably is.