reset allocation counts in gc thread-local buffers

Author: kevmo314

Python/gc_free_threading.c

@@ -2207,11 +2207,10 @@ record_deallocation(PyThreadState *tstate)
 {
     struct _gc_thread_state *gc = &((_PyThreadStateImpl *)tstate)->gc;
 
-    gc->alloc_count--;
-    if (gc->alloc_count <= -LOCAL_ALLOC_COUNT_THRESHOLD) {
-        GCState *gcstate = &tstate->interp->gc;
-        _Py_atomic_add_int(&gcstate->young.count, (int)gc->alloc_count);
-        gc->alloc_count = 0;
+    // Only decrement if positive, matching gc.c behavior which prevents
+    // negative counts (see PyObject_GC_Del in gc.c).
+    if (gc->alloc_count > 0) {
+        gc->alloc_count--;
     }
 }
 

Python/pystate.c

@@ -1810,6 +1810,14 @@ tstate_delete_common(PyThreadState *tstate, int release_gil)
     assert(tstate_impl->refcounts.values == NULL);
 #endif
 
+#ifdef Py_GIL_DISABLED
+    // Flush the thread's local GC allocation count to the global count
+    // before the thread state is deleted, otherwise the count is lost.
+    _Py_atomic_add_int(&tstate->interp->gc.young.count,
+                       (int)((_PyThreadStateImpl *)tstate)->gc.alloc_count);
+    ((_PyThreadStateImpl *)tstate)->gc.alloc_count = 0;
+#endif
+
 #if _Py_TIER2
     _PyThreadStateImpl *_tstate = (_PyThreadStateImpl *)tstate;
     if (_tstate->jit_tracer_state.code_buffer != NULL) {