(1)
/**

• When an exception occurs, registers are saved temporarily. If
• the stack being scanned is in this state, we need to scan those
• registers for code pointers. If the codeLocations deque is null,
• then scanning for code pointers is not required, so we don't need
• to do anything. (SB: Why only code pointers?)
  *
• @param verbosity The level of verbosity to be used when
• performing the scan.
  */
  private void getHWExceptionRegisters(int verbosity) {
  ...

(2)
/**

• Identify all pointers into code pointers associated with a frame.
• There are two cases to be considered: a) the instruction pointer
• associated with each frame (stored in the thread's metadata for
• the top frame and as a return address for all subsequent frames),
• and b) local variables on the stack which happen to be pointers
• to code.<p>
  *
• FIXME: SB: Why is it that JNI frames are skipped when considering
• top of stack frames, while boot image frames are skipped when
• considering other frames. Shouldn't they both be considered in
• both cases?
  *
• @param verbosity The level of verbosity to be used when
• performing the scan.
  */
  private void processFrameForCode(int verbosity) {
  ...

(3)
/**

• AIX-specific code.<p>
  *
• If we are scanning the stack of a thread that entered the VM via
• a createVM or attachVM then the "bottom" of the stack had native
• C frames instead of the usual java frames. The JNIEnv for the
• thread may still contain jniRefs that have been returned to the
• native C code, but have not been reported for GC. calling
• getNextReferenceAddress without first calling setup... will
• report the remaining jniRefs in the current "frame" of the
• jniRefs stack. (this should be the bottom frame)
  *
• FIXME: SB: Why is this AIX specific? Why depend on the
• preprocessor?
  *
• @param verbosity The level of verbosity to be used when
• performing the scan.
  */
  private void checkJNIBase(int verbosity) {
  ...

These issues should be addressed, fixing the possible problems and the comments removed.