fix(dedupe): Fall back to a type/args fingerprint for unweakrefable exceptions

sentry_sdk/integrations/aws_lambda.py

@@ -17,6 +17,7 @@
     CLOUD_PLATFORM,
     CLOUD_PROVIDER,
 )
+from sentry_sdk.integrations.dedupe import DedupeIntegration
 from sentry_sdk.scope import Scope, should_send_default_pii
 from sentry_sdk.traces import SegmentNameSource
 from sentry_sdk.utils import (
@@ -134,6 +135,7 @@ def sentry_handler(
             timeout_thread = None
             with capture_internal_exceptions():
                 scope.clear_breadcrumbs()
+                DedupeIntegration.reset_last_seen()
                 scope.add_event_processor(
                     _make_request_event_processor(
                         request_data, aws_context, configured_time

sentry_sdk/integrations/dedupe.py

@@ -8,11 +8,16 @@
 from sentry_sdk.utils import logger
 
 if TYPE_CHECKING:
-    from typing import Any, Optional
+    from typing import Any, Optional, Tuple, Type
 
     from sentry_sdk._types import Event, Hint
 
 
+def _fingerprint(
+    exc: BaseException,
+) -> "Tuple[Type[BaseException], int, str]":
+    return (type(exc), hash(exc.args), hex(id(exc)))
+
 class DedupeIntegration(Integration):
     identifier = "dedupe"
 
@@ -35,22 +40,23 @@ def processor(event: "Event", hint: "Optional[Hint]") -> "Optional[Event]":
                 return event
 
             last_seen = integration._last_seen.get(None)
-            if last_seen is not None:
-                # last_seen is either a weakref or the original instance
-                last_seen = (
-                    last_seen() if isinstance(last_seen, weakref.ref) else last_seen
-                )
-
             exc = exc_info[1]
-            if last_seen is exc:
+
+            is_duplicate = False
+            if isinstance(last_seen, weakref.ref):
+                is_duplicate = last_seen() is exc
+            elif isinstance(last_seen, tuple):
+                is_duplicate = last_seen == _fingerprint(exc)
+
+            if is_duplicate:
                 logger.info("DedupeIntegration dropped duplicated error event %s", exc)
                 return None
 
             # we can only weakref non builtin types
             try:
                 integration._last_seen.set(weakref.ref(exc))
             except TypeError:
-                integration._last_seen.set(exc)
+                integration._last_seen.set(_fingerprint(exc))
 
             return event
 

sentry_sdk/integrations/gcp.py

@@ -11,6 +11,7 @@
 from sentry_sdk.integrations import Integration
 from sentry_sdk.integrations._wsgi_common import _filter_headers
 from sentry_sdk.integrations.cloud_resource_context import CLOUD_PROVIDER
+from sentry_sdk.integrations.dedupe import DedupeIntegration
 from sentry_sdk.scope import Scope, should_send_default_pii
 from sentry_sdk.traces import SegmentNameSource
 from sentry_sdk.utils import (
@@ -60,6 +61,7 @@ def sentry_func(
         with sentry_sdk.isolation_scope() as scope:
             with capture_internal_exceptions():
                 scope.clear_breadcrumbs()
+                DedupeIntegration.reset_last_seen()
                 scope.add_event_processor(
                     _make_request_event_processor(
                         gcp_event, configured_time, initial_time

sentry_sdk/integrations/serverless.py

@@ -3,6 +3,7 @@
 from typing import TYPE_CHECKING
 
 import sentry_sdk
+from sentry_sdk.integrations.dedupe import DedupeIntegration
 from sentry_sdk.utils import event_from_exception, reraise
 
 if TYPE_CHECKING:
@@ -34,6 +35,7 @@ def wrapper(f: "F") -> "F":
         def inner(*args: "Any", **kwargs: "Any") -> "Any":
             with sentry_sdk.isolation_scope() as scope:
                 scope.clear_breadcrumbs()
+                DedupeIntegration.reset_last_seen()
 
                 try:
                     return f(*args, **kwargs)

tests/integrations/launchdarkly/test_launchdarkly.py

@@ -69,7 +69,7 @@ def task(flag_key):
             client.variation(flag_key, context, False)
             # use a tag to identify to identify events later on
             sentry_sdk.set_tag("task_id", flag_key)
-            sentry_sdk.capture_exception(Exception("something wrong!"))
+            sentry_sdk.capture_exception(Exception(f"{flag_key}: something wrong!"))
 
     # Capture an eval before we split isolation scopes.
     client.variation("hello", context, False)

tests/test_basics.py

@@ -122,6 +122,46 @@ def test_generic_mechanism(sentry_init, capture_events):
     assert event["exception"]["values"][0]["mechanism"]["handled"]
 
 
+def test_dedupe_builtin_exceptions(sentry_init, capture_events):
+    def do_this():
+        try:
+            raise ValueError("hello world!")
+        except Exception:
+            capture_exception()
+
+    sentry_init()
+    events = capture_events()
+
+    do_this()
+    do_this()
+
+    assert len(events) == 1, "Built-in exceptions are not deduplicated"
+
+
+def test_dedupe_builtin_exceptions_different_raise_sites(sentry_init, capture_events):
+    def raise_here():
+        try:
+            raise ValueError("hello world!")
+        except Exception:
+            capture_exception()
+
+    def raise_there():
+        try:
+            raise ValueError("hello world!")
+        except Exception:
+            capture_exception()
+
+    sentry_init()
+    events = capture_events()
+
+    raise_here()
+    raise_there()
+
+    assert len(events) == 2, (
+        "Built-in exceptions with the exact same type and arguments raised from different code paths are not deduplicated"
+    )
+
+
 def test_option_before_send(sentry_init, capture_events):
     def before_send(event, hint):
         event["extra"] = {"before_send_called": True}
@@ -187,19 +227,19 @@ def before_breadcrumb(crumb, hint):
         sentry_sdk.get_client().transport, "record_lost_event", record_lost_event
     )
 
-    def do_this():
+    def do_this(msg):
         add_breadcrumb(message="Hello", hint={"foo": 42})
         try:
-            raise ValueError("aha!")
+            raise ValueError(msg)
         except Exception:
             capture_exception()
 
-    do_this()
+    do_this("aha!")
     drop_breadcrumbs = True
-    do_this()
+    do_this("another aha!")
     assert not reports
     drop_events = True
-    do_this()
+    do_this("why not one more aha!")
     assert reports == [("before_send", "error")]
 
     normal, no_crumbs = events
@@ -284,7 +324,7 @@ def test_breadcrumbs(sentry_init, capture_events):
 
     sentry_sdk.get_isolation_scope().clear()
 
-    capture_exception(ValueError())
+    capture_exception(ValueError("another one!"))
     (event,) = events
     assert len(event["breadcrumbs"]["values"]) == 0
 

tests/test_scope.py

@@ -11,6 +11,7 @@
     new_scope,
 )
 from sentry_sdk.client import Client, NonRecordingClient
+from sentry_sdk.integrations.dedupe import DedupeIntegration
 from sentry_sdk.scope import (
     Scope,
     ScopeType,
@@ -69,7 +70,7 @@ def test_scope_flags_copy():
 
 
 def test_set_user(sentry_init, capture_events):
-    sentry_init()
+    sentry_init(disabled_integrations=[DedupeIntegration])
     events = capture_events()
 
     sentry_sdk.get_isolation_scope().set_user({"id": "42", "email": "bob@example.com"})