Old mail on gc leak?

I have a vague memory that many years ago, someone sent mail to one of
the mailing lists demonstrating a gencgc leak.  The example code would
cause gc to run and the gc messages would show increasing memory usage
with bytes collected being negative.  When the routine finished, a
subsequent gc would clean up everything.  The same code on a Sparc
with the old stop-and-copy gc didn't have this problem.

Does this sound familiar to anyone?  I'm just looking for some test
code to exercise the sparc gencgc and thought this would be a good
test to see if the sparc gencgc leaked memory or not.

It could have been from the old seamew archives, but I don't remember
exactly when this message was sent.  I've looked through all the
archives I could find, but don't have any seamew archives.

Thanks,

Ray
P.S.  I've finally tested sparc gencgc on something other than cmucl.
It compiles and runs maxima fine, and also runs my own simulation code
just fine.  Whew!