[gclist] Experience with conservative GC sought

Jim Blandy jimb@red-bean.com
Sat, 24 Jan 1998 17:59:11 -0500

I'm sorry, Carl, I didn't mean to imply that you were criticizing the
Emacs authors.  I think the question you raise is quite reasonable.
It sounds as if you have more experience with non-trivial GC than I.

The approach I think would be most interesting (David Love's idea
again) would be to leave Emacs's current (precise) stack tracing
mechanism in place, do a conservative collection, and then do a
precise collection to see how much more you pick up.  You could even
do it in the same marking pass, so the additional cost would be
proportional to the amount of leakage: either small, or you've got an
interesting result to announce. :)