Ludovic Courtès wrote:
The patch fixes `scm_i_mark_weak_vector_non_weaks ()'. This way, a
value attached to a weak key will only become unmarked during the mark
phase _after_ the mark phase during which its weak key became unmarked.
You might want to re-read this sentence carefully. ;-)
Does your patch solve the problem that cyclical structures (values that
point back to keys) should also be GC-ed?
/* Attach TO_BE_ATTACHED to OBJ. */
static SCM
attach_object (SCM obj, SCM to_be_attached)
{
my_object_t *c_obj, *c_to_be_attached;
assert (SCM_SMOB_PREDICATE (my_object_type, obj));
assert (SCM_SMOB_PREDICATE (my_object_type, to_be_attached));
/* TO_BE_ATTACHED is attached as a property of OBJ. As such, OBJ will get
GC'd _before_ TO_BE_ATTACHED. */
scm_primitive_property_set_x (some_property, obj, to_be_attached);
Why are you storing SCM references as properties? It's more efficient
both in time and space to use a SCM member in my_object_t.
--
Han-Wen Nienhuys - [EMAIL PROTECTED] - http://www.xs4all.nl/~hanwen
_______________________________________________
Guile-devel mailing list
Guile-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-devel
