Kernel.hash should not be propagated to set_trace_func callback

Pasting from http://archive.jruby.codehaus.org/dev/45D4038E.804%40sun.com:

There seems to be superfluous calls to set_trace_func callback. Every
time a function call, which was not called before, triggers call to
Kernel.hash which is propagated to the callback. Probably the callback
should receive only source-code-related events. But not sure since I'm
not aware of any set_trace_func formal specification.

Probably best to take a look directly at an example. Content of t.rb:

====
printf "%8s %20s:%-2d %10s %8s\n", 'event', 'file', 0, 'id', 'classname'
$stdout.puts "-" * 80
set_trace_func proc { |event, file, line, id, binding, classname|
printf "%8s %20s:%-2d %10s %-8s\n", event, file, line, id ? id : 'nil', classname
}

sleep(0.1)
sleep(0.1)
self.class
hash
self.class
====

Then try to run:

jruby t.rb > t-j.txt && ruby t.rb > t-r.txt && vim d t?.txt

or similar. Output is the same except those Kernel.hash events.
Presumably bug. But asking before filling. If you consider it as "not
a bug" we would probably need to tweak some tests in debug-commons and
might be also debugger itself.