Complete cache key for inference tip (#2158)

The cache key was lacking the `context` arg.

Co-authored-by: Sylvain Ackermann <sylvain.ackermann@gmail.com>

1 files changed, 22 insertions, 8 deletions

diff --git a/astroid/inference_tip.py b/astroid/inference_tip.py
index 5b855c9e..92cb6b4f 100644
--- a/astroid/inference_tip.py
+++ b/astroid/inference_tip.py
@@ -9,6 +9,7 @@ from __future__ import annotations
 import sys
 from collections.abc import Callable, Iterator
 
+from astroid.context import InferenceContext
 from astroid.exceptions import InferenceOverwriteError, UseInferenceDefault
 from astroid.nodes import NodeNG
 from astroid.typing import InferenceResult, InferFn
@@ -20,7 +21,11 @@ else:
 
 _P = ParamSpec("_P")
 
-_cache: dict[tuple[InferFn, NodeNG], list[InferenceResult] | None] = {}
+_cache: dict[
+    tuple[InferFn, NodeNG, InferenceContext | None], list[InferenceResult]
+] = {}
+
+_CURRENTLY_INFERRING: set[tuple[InferFn, NodeNG]] = set()
 
 
 def clear_inference_tip_cache() -> None:
@@ -35,16 +40,25 @@ def _inference_tip_cached(
 
     def inner(*args: _P.args, **kwargs: _P.kwargs) -> Iterator[InferenceResult]:
         node = args[0]
-        try:
-            result = _cache[func, node]
+        context = args[1]
+        partial_cache_key = (func, node)
+        if partial_cache_key in _CURRENTLY_INFERRING:
             # If through recursion we end up trying to infer the same
             # func + node we raise here.
-            if result is None:
-                raise UseInferenceDefault()
+            raise UseInferenceDefault
+        try:
+            return _cache[func, node, context]
         except KeyError:
-            _cache[func, node] = None
-            result = _cache[func, node] = list(func(*args, **kwargs))
-            assert result
+            # Recursion guard with a partial cache key.
+            # Using the full key causes a recursion error on PyPy.
+            # It's a pragmatic compromise to avoid so much recursive inference
+            # with slightly different contexts while still passing the simple
+            # test cases included with this commit.
+            _CURRENTLY_INFERRING.add(partial_cache_key)
+            result = _cache[func, node, context] = list(func(*args, **kwargs))
+            # Remove recursion guard.
+            _CURRENTLY_INFERRING.remove(partial_cache_key)
+
         return iter(result)
 
     return inner