8:01:56_nir_karlosz: I really enjoy writing useful documentation and I'll definitely work on restoring knowledge about SBCL internals but it's a long way to go since it's my first experience of porting software to another architecture.
16:08:08dougk_pkhuong: I'm pretty sure the bug in your card marking code was that it assumed that 'pinning' applies to objects in _any_ generation, whereas pinning actually only applies to the objects in the condemned set. So if some code was touching the mark of an object in generation 1 in preparation to write, then we did a gen0 collection, the mark would or could get unset on the card for the about-to-be-touched gen1 object because it was _not_ in t
16:08:08dougk_he pinned set. So after that GC the remembered set is bogus