[Ecls-list] Bug report (?) and two questions
Juan Jose Garcia-Ripoll
Juan.Ripoll at mpq.mpg.de
Wed Dec 12 00:13:01 UTC 2001
"Dr. Edmund Weitz" wrote:
> worm <worm at arrakis.es> writes:
> > Then this is a problem with the Boehm-Weiser garbage collector. This
> > collector keeps a list of memory that it believes to be
> > unreclaimable. If this list gets too large, then the amount of
> > allocatable memory drops down to zero. This surely deserves
> > investigation, but it would help me if you could try the same
> > program, on the same platform but building ECLS with
> > --disable-boehm. That reverts to the original garbage collector,
> > which may or may not work on your version of Unix.
> Sorry for the delay - I had the flu and was in bed.
> I rebuilt ecls with --disable-boehm and rerun the tests with the
> faculty function. I got basically the same results - same error
> messages when compiled, same unresponsive behaviour when interpreted.
I hope you are better by now. The memory allocation problem was due to a
bug in src/c/alloc_2.d The problem is that, as H. Koivisto pointed out,
the memory for bignums was not tagged as pointerless memory, and thus
the Boehm collector got confused when inspecting the contents of these
This is now fixed in CVS. What worries me now is that the calculation of
factorials is not very fast. I lack a lisp to compare with, but could
anybody compare CLISP, a new copy of ECL and CMUCL on the same platform?
More information about the ecl-devel