fix: modify error tracking exceptions [backport 3.18] (#15168)

Backport 84db0e6 from #15148 to 3.18.

## Description

<!-- Provide an overview of the change and motivation for the change -->

For [APMS-17683](https://datadoghq.atlassian.net/browse/APMS-17683)
customer was having an issue where their custom exception object was
unhashable, resulting in a `TypeError` when trying to use the object as
a key. This fix changes so that we store the exception object id, not
the exception object.

## Testing

<!-- Describe your testing strategy or note what tests are included -->

## Risks

<!-- Note any risks associated with this change, or "None" if no risks
-->

## Additional Notes

<!-- Any other information that would be helpful for reviewers -->


[APMS-17683]:
https://datadoghq.atlassian.net/browse/APMS-17683?atlOrigin=eyJpIjoiNWRkNTljNzYxNjVmNDY3MDlhMDU5Y2ZhYzA5YTRkZjUiLCJwIjoiZ2l0aHViLWNvbS1KU1cifQ

Co-authored-by: Quinna Halim <quinna.halim@datadoghq.com>

Author: dd-octo-sts[bot]

ddtrace/errortracking/_handled_exceptions/collector.py

@@ -21,7 +21,8 @@ def _add_span_events(span: Span) -> None:
     a span event for every handled exceptions, we store them in the span
     and add them when the span finishes.
     """
-    span_exc_events = list(HandledExceptionCollector.get_exception_events(span.span_id).values())
+    exception_data = HandledExceptionCollector.get_exception_events(span.span_id).values()
+    span_exc_events = [event for _exc, event in exception_data]
     if span_exc_events:
         span._set_tag_str(SPAN_EVENTS_HAS_EXCEPTION, "true")
         span._events.extend(span_exc_events)
@@ -30,13 +31,14 @@ def _add_span_events(span: Span) -> None:
 
 def _on_span_exception(span, _exc_msg, exc_val, _exc_tb):
     exception_events = HandledExceptionCollector.get_exception_events(span.span_id)
-    if exception_events and exc_val in exception_events:
-        del exception_events[exc_val]
+    exc_id = id(exc_val)
+    if exception_events and exc_id in exception_events:
+        del exception_events[exc_id]
 
 
 class HandledExceptionCollector(Service):
     _instance: t.Optional["HandledExceptionCollector"] = None
-    _span_exception_events: t.Dict[int, t.Dict[Exception, SpanEvent]] = {}
+    _span_exception_events: t.Dict[int, t.Dict[int, t.Tuple[Exception, SpanEvent]]] = {}
 
     def __init__(self) -> None:
         super(HandledExceptionCollector, self).__init__()
@@ -111,8 +113,10 @@ def capture_exception_event(cls, span: Span, exc: Exception, event: SpanEvent):
         events_dict = cls._span_exception_events.setdefault(span_id, {})
         if not events_dict:
             span._add_on_finish_exception_callback(_add_span_events)
-        if exc in events_dict or len(events_dict) < COLLECTOR_MAX_SIZE_PER_SPAN:
-            events_dict[exc] = event
+        exc_id = id(exc)
+        if exc_id in events_dict or len(events_dict) < COLLECTOR_MAX_SIZE_PER_SPAN:
+            # Store both exception and event to keep exception alive and prevent ID reuse
+            events_dict[exc_id] = (exc, event)
 
     @classmethod
     def get_exception_events(cls, span_id: int):

releasenotes/notes/modify-error-tracking-exceptions-6be5aa66cb4cd2ef.yaml

@@ -0,0 +1,4 @@
+---
+fixes:
+  - |
+    Error Tracking: Modifies the way exception events are stored such that the exception id is stored instead of the exception object, to prevent TypeErrors with custom exception objects.

tests/errortracking/_test_functions.py

@@ -1,6 +1,24 @@
 import asyncio
 
 
+class UnhashableException(Exception):
+    def __init__(self, message, mutable_data):
+        super().__init__(message)
+        self.mutable_data = mutable_data
+
+    def __eq__(self, other):
+        # This makes the exception unhashable if __hash__ is not defined
+        return isinstance(other, UnhashableException) and str(self) == str(other)
+
+
+def test_unhashable_exception_f(value):
+    try:
+        raise UnhashableException("unhashable error", {"key": "value"})
+    except UnhashableException:
+        value = 10
+    return value
+
+
 def test_basic_try_except_f(value):
     try:
         raise ValueError("auto caught error")

tests/errortracking/test_handled_exceptions.py

@@ -91,7 +91,9 @@ def test_handle_same_error_multiple_times(self):
         )
         # This asserts that the reported span event contained the info
         # of the last time the error is handled
-        assert "line 30" in self.spans[0]._events[0].attributes["exception.stacktrace"]
+        assert "line 48" in self.spans[0]._events[0].attributes["exception.stacktrace"], (
+            self.spans[0]._events[0].attributes["exception.stacktrace"]
+        )
 
     @run_in_subprocess(env_overrides=dict(DD_ERROR_TRACKING_HANDLED_ERRORS="all"))
     def test_handled_same_error_different_type(self):
@@ -159,11 +161,30 @@ def test_handled_in_parent_span(self):
         self.spans[0].assert_span_event_attributes(
             0, {"exception.type": "builtins.ValueError", "exception.message": "auto caught error"}
         )
-        assert "line 72" in self.spans[0]._events[0].attributes["exception.stacktrace"]
+        assert "line 100" in self.spans[0]._events[0].attributes["exception.stacktrace"], (
+            self.spans[0]._events[0].attributes["exception.stacktrace"]
+        )
 
         assert self.spans[1].name == "child_span"
         assert len(self.spans[1]._events) == 0
 
+    @run_in_subprocess(env_overrides=dict(DD_ERROR_TRACKING_HANDLED_ERRORS="all"))
+    def test_unhashable_exception(self):
+        """Test that unhashable exceptions (e.g., with mutable attributes) are handled correctly."""
+        self._run_error_test(
+            "test_unhashable_exception_f",
+            initial_value=0,
+            expected_value=10,
+            expected_events=[
+                [
+                    {
+                        "exception.type": "tests.errortracking._test_functions.UnhashableException",
+                        "exception.message": "unhashable error",
+                    }
+                ]
+            ],
+        )
+
 
 @skipif_errortracking_not_supported
 class UserCodeErrorTestCases(TracerTestCase):