* gc.c (rb_objspace_free): With our "lazy-sweep" GC engine, it is

  possible for an object to survive until its surrounding object
  space is about to be freed.  Those objects, if any, remains not
  leaked for the rest of a process life.  This is problematic
  because for instance a T_DATA object may have its own destructor
  to terminate something.

* vm.c (ruby_vm_destruct): ruby_current_vm termination should be
  somewhere after rb_objspace_free for above reason.



git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@30065 b2dd03c8-39d4-4d8f-98ff-823fe69b080e

1 files changed, 6 insertions, 0 deletions

diff --git a/gc.c b/gc.c
index cf9107ab8e..d7dfa017e5 100644
--- a/gc.c
+++ b/gc.c
@@ -400,9 +400,15 @@ rb_objspace_alloc(void)
     return objspace;
 }
 
+static void gc_sweep(rb_objspace_t *);
+static void slot_sweep(rb_objspace_t *, struct heaps_slot *);
+static void gc_clear_mark_on_sweep_slots(rb_objspace_t *);
+
 void
 rb_objspace_free(rb_objspace_t *objspace)
 {
+    gc_clear_mark_on_sweep_slots(objspace);
+    gc_sweep(objspace);
     if (objspace->profile.record) {
 	free(objspace->profile.record);
 	objspace->profile.record = 0;