From e4fdd30787822bd0aa1eec3ebf0539c55b7cfa3c Mon Sep 17 00:00:00 2001
From: David Hewitt <mail@davidhewitt.dev>
Date: Thu, 21 Mar 2024 20:22:34 +0000
Subject: [PATCH] avoid reference count cycle in tuple extraction

---
 src/types/tuple.rs | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/src/types/tuple.rs b/src/types/tuple.rs
index d8053e1441a..d89830daa97 100644
--- a/src/types/tuple.rs
+++ b/src/types/tuple.rs
@@ -690,10 +690,10 @@ fn type_output() -> TypeInfo {
             let t = obj.downcast::<PyTuple>()?;
             if t.len() == $length {
                 #[cfg(any(Py_LIMITED_API, PyPy))]
-                return Ok(($(t.get_item($n)?.extract::<$T>()?,)+));
+                return Ok(($(t.get_borrowed_item($n)?.extract::<$T>()?,)+));
 
                 #[cfg(not(any(Py_LIMITED_API, PyPy)))]
-                unsafe {return Ok(($(t.get_item_unchecked($n).extract::<$T>()?,)+));}
+                unsafe {return Ok(($(t.get_borrowed_item_unchecked($n).extract::<$T>()?,)+));}
             } else {
                 Err(wrong_tuple_length(t, $length))
             }