ref(seer grouping): Add ability to ignore useless filenames (#85277)

A customer recently switched the build tool for their iOS app, and it changed their app's stacktraces such that new groups are now being created when ideally they wouldn't be. There were multiple differences in the stacktraces, not all easily mitigated, but one straightforward one was that some frames went from having no filename property to having a filename property of `<compiler-generated>`. 

To make it so that at least Seer will ignore this particular difference, this PR adds a list of filenames for us to ignore when constructing the stacktrace string we send it. We should also consider marking filename components with such filenames as non-contributing, but that will have to wait for a separate PR.

Author: lobsterkatie

src/sentry/seer/similarity/utils.py

@@ -43,6 +43,8 @@
     "javascript;base64",
 ]
 
+IGNORED_FILENAMES = ["<compiler-generated>"]
+
 
 class ReferrerOptions(StrEnum):
     INGEST = "ingest"
@@ -239,6 +241,8 @@ def extract_filename(frame_dict: Mapping[str, Any]) -> str:
     Extract the filename from the frame dictionary. Fallback to module if filename is not present.
     """
     filename = frame_dict["filename"]
+    if filename in IGNORED_FILENAMES:
+        filename = ""
     if filename == "" and frame_dict["module"] != "":
         filename = frame_dict["module"]
     return filename

tests/sentry/seer/similarity/test_utils.py

@@ -8,6 +8,7 @@
 from sentry.grouping.variants import CustomFingerprintVariant
 from sentry.seer.similarity.utils import (
     BASE64_ENCODED_PREFIXES,
+    IGNORED_FILENAMES,
     MAX_FRAME_COUNT,
     ReferrerOptions,
     _is_snipped_context_line,
@@ -811,6 +812,21 @@ def test_no_filename_or_module(self):
             == 'ZeroDivisionError: division by zero\n  File "None", function divide_by_zero\n    divide = 1/0'
         )
 
+    def test_ignores_meaningless_filenames(self):
+        for ignored_filename in IGNORED_FILENAMES:
+            exception = copy.deepcopy(self.BASE_APP_DATA)
+            # delete module from the exception so we don't fall back to that
+            del exception["app"]["component"]["values"][0]["values"][0]["values"][0]["values"][0]
+            # replace filename with ignored value
+            exception["app"]["component"]["values"][0]["values"][0]["values"][0]["values"][0][
+                "values"
+            ][0] = ignored_filename
+            stacktrace_string = get_stacktrace_string(exception)
+            assert (
+                stacktrace_string
+                == 'ZeroDivisionError: division by zero\n  File "None", function divide_by_zero\n    divide = 1/0'
+            )
+
     @patch("sentry.seer.similarity.utils.metrics")
     def test_no_header_one_frame_no_filename(self, mock_metrics):
         exception = copy.deepcopy(self.MOBILE_THREAD_DATA)