Always examine the stack, and take optional file name / line number args.
The stack information is useful even when we have the file/line number,
since it's interesting to see where the calls came from.  (Consider the
case of __os_malloc wrappers in db2.)

Interesting question is, how to display the stack information nicely when
dumping the chunk data to /tmp/mtrack.out..  Two-file output perhaps, one
containing concise information and the other has full stack traces?