Say I have a project with modules A and B, built under hudson using the 'clean
deploy' as the goals. Further, say B has a dependency on A. When there's a
change in the project, hudson will therefore build A followed by B.

The problem is that if A's tests fail, A is not deployed. This is maven's
behavior (and, IMO, correct). Hudson will still trigger B, which is a problem.

Since A wasn't deployed, B will be looking at a previous locally installed
version of A. If both A & B were changed at the same time (e.g., because the
API A provides and B uses was changed), B will no longer compile against the old
version of A. This means that B will have a build failure. This means the
whole build will be marked as "failed" rather than "unstable," which obscures
the real source of the problem.

Since hudson knows that A's tests failed, would it be possible to have it not
trigger B? (I.e., just like if A completely failed.)