If unwrap returns null, we need to fall through to load to try it again.
(cherry picked from commit 04c87d8d)

If a build record gets GCed, a load is attempted, and that fails, then RunMap can return null. Not sure if this alone would explain the entirety of JENKINS-15465, but this is a related fix.
(cherry picked from commit ca44a90e)

I can't really think of how null ends up in the resulting list, but adding a check in Iterator to detect that situation.

Also, improved the efficiency of the subList implementation.
(cherry picked from commit 807dc717)

Ensure that SortedList clones the underlying data list.
Without this changes to a clone also change the original list.
(cherry picked from commit 761c4379)

Thanks to newenigma for catching.

Clearly state that FingerprintAction, not UpstreamCause, is used for all upstream/downstream methods in this class (contrary to what you might expect).

Lots left untested—*fingerprint* and so on.

Conflicts:
	changelog.html

Fixed JENKINS-15293
Thanks

do not remove maven installation which does not have a home, like maven installation with automatic installer (bug associated with JENKINS-14510 fix ... sorry..)