OK, so what you are seeing is that the profiler keeps existing encapsulations around for later reuse. You need to empty the hashtables in profile.lisp if you change encapsulation code.
We should probably drop that optimization, it is pretty pointless on a multi-GHz Ninja-Macho pentium and it clutters up code that needs some work done. Martin rif wrote on Thu, Oct 17, 2002 at 10:20:02AM -0400: > > Yes. If I start up CMUCL, I immediately define (redefine) the > total-consing macro to 0, and THEN profile functions, I don't observe > this error. > > Cheers, > > rif > > > rif wrote on Thu, Oct 17, 2002 at 09:19:45AM -0400: > > > > > > I'm confused. > > > > Me too :-) > > > > So does it work when you define the new macro directly after startup > > of the Lisp, before doing an profiling? > > > > Martin >
