SetLocalObject says (well, it will say this after the typo of saying
'retrieve' instead of 'set' is fixed):

    This function can be used to set the value of a local variable whose type is Object or
    a subclass of Object.

In the return values, the spec specifies:
   JVMTI_ERROR_TYPE_MISMATCH The supplied value is not compatible with the variable type.

But it allows one to set a variable of any object type (eg. String) to a value of any object type,
(eg. Object).

Seems like it should return JVMTI_ERROR_TYPE_MISMATCH in this case.
The following test is affected by this bug:
nsk/jvmti/ForceEarlyReturn/ForceEarlyReturn002
The test also has another failure mode. Here is the entry from my
nightly analysis report:

New nsk.quick_jdwp failures (from 2006.12.04)
    nsk/jdwp/ThreadReference/ForceEarlyReturn/forceEarlyReturn001
        This test failed the following assertion:

            Internal Error (src/share/vm/runtime/sharedRuntime.cpp, 848)
            Error: assert(receiver_klass->is_subtype_of(static_receiver_klass()),
            "actual receiver must be subclass of static receiver klass")

        on Solaris SPARC Client VM (machine vm-ultra80-02). This
        assertion failure is covered by the following bug:

            6458173 3/3 nsk/jdwp/ThreadReference/ForceEarlyReturn/
                forceEarlyReturn001 asserts with fastdebug -Xcomp -client

        However, 6458173 was closed as a duplicate of:

            6444285 4/3 SetLocalObject and ForceEarlyReturnObject don't
                        check for type mismatches on Objects

        I will add this entry to 6444285 so the assertion failure mode
        is visible via that bug.