feat(symbolication): Rewrite Electron minidump modules (#87176)

Electron minidumps frequently contain root modules with nonsensical
debug file names, which results from packagers like electron-forge.
Consequently, the module won't be found on the official Electron symbol
source.

To fix this, we rewrite the debug file name of the first module
according to these rules for Electron minidumps before doing the lookup.
The rules are tried in order and only the first that matches is applied.

The rewrites we perform are as follows:

* /PATH/TO/FILE Framework -> /PATH/TO/Electron Framework
* /PATH/TO/FILE Helper (FOOBAR) -> /PATH/TO/Electron Helper (FOOBAR)
* /PATH/TO/FILE.exe.pdb -> /PATH/TO/electron.exe.pdb
* /PATH/TO/FILE -> /PATH/TO/electron

These were determined by examining logs of successful downloads from the
Electron symbol source.

The rewriting itself is performed by Symbolicator. We can't do it
earlier because we only gain access to the modules contained in the
minidump during stackwalking.

The Symbolicator end of this change is
getsentry/symbolicator#1650.

Author: loewenheim

src/sentry/lang/native/processing.py

@@ -42,6 +42,39 @@
 # Attachment type used for Apple Crash Reports
 APPLECRASHREPORT_ATTACHMENT_TYPE = "event.applecrashreport"
 
+# Rules for rewriting the debug file of the first module
+# in an Electron minidump.
+#
+# Such minidumps frequently contain root modules
+# with nonsensical debug file names, which results from packagers like
+# electron-forge. Consequently, the module won't be found on the official
+# Electron symbol source.
+#
+# To fix this, we rewrite the debug file name of the first module
+# according to these rules for Electron minidumps before doing the lookup.
+# The rules are tried in order and only the first that matches is applied.
+#
+# The rewrites we perform are as follows:
+#
+# * /PATH/TO/FILE Framework -> /PATH/TO/Electron Framework
+# * /PATH/TO/FILE Helper (FOOBAR) -> /PATH/TO/Electron Helper (FOOBAR)
+# * /PATH/TO/FILE.exe.pdb -> /PATH/TO/electron.exe.pdb
+# * /PATH/TO/FILE -> /PATH/TO/electron
+#
+# These were determined by examining logs of successful downloads from the Electron symbol
+# source.
+#
+# The rewriting itself is performed by Symbolicator. We can't do it before
+# because we only gain access to the modules contained in the minidump
+# during stackwalking.
+#
+# NOTE: These regexes and replacement strings are written in the syntax the Rust regex crate accepts!
+ELECTRON_FIRST_MODULE_REWRITE_RULES = [
+    {"from": "[^/\\\\]+ (?<suffix>Framework|Helper( \\(.+\\))?)$", "to": "Electron $suffix"},
+    {"from": "[^/\\\\]+\\.exe\\.pdb$", "to": "electron.exe.pdb"},
+    {"from": "[^/\\\\]+$", "to": "electron"},
+]
+
 
 def _merge_frame(new_frame, symbolicated, platform="native"):
     # il2cpp events which have the "csharp" platform have good (C#) names
@@ -288,8 +321,17 @@ def process_minidump(symbolicator: Symbolicator, data: Any) -> Any:
         logger.error("Missing minidump for minidump event")
         return
 
+    # We do module rewriting only for Electron minidumps.
+    # See the documentation of `ELECTRON_FIRST_MODULE_REWRITE_RULES`.
+    if get_path(data, "sdk", "name") == "sentry.javascript.electron":
+        rewrite_first_module = ELECTRON_FIRST_MODULE_REWRITE_RULES
+    else:
+        rewrite_first_module = []
+
     metrics.incr("process.native.symbolicate.request")
-    response = symbolicator.process_minidump(data.get("platform"), minidump.data)
+    response = symbolicator.process_minidump(
+        data.get("platform"), minidump.data, rewrite_first_module
+    )
 
     if _handle_response_status(data, response):
         _merge_full_response(data, response)

src/sentry/lang/native/symbolicator.py

@@ -166,13 +166,14 @@ def _process(self, task_name: str, path: str, **kwargs):
                 # Otherwise, we are done processing, yay
                 return json_response
 
-    def process_minidump(self, platform, minidump):
+    def process_minidump(self, platform, minidump, rewrite_first_module):
         (sources, process_response) = sources_for_symbolication(self.project)
         scraping_config = get_scraping_config(self.project)
         data = {
             "platform": orjson.dumps(platform).decode(),
             "sources": orjson.dumps(sources).decode(),
             "scraping": orjson.dumps(scraping_config).decode(),
+            "rewrite_first_module": orjson.dumps(rewrite_first_module).decode(),
             "options": '{"dif_candidates": true}',
         }
 

tests/sentry/lang/native/test_processing.py

@@ -5,12 +5,14 @@
 
 from __future__ import annotations
 
+import re
 from typing import Any
 from unittest import mock
 
 import pytest
 
 from sentry.lang.native.processing import (
+    ELECTRON_FIRST_MODULE_REWRITE_RULES,
     _merge_image,
     get_frames_for_symbolication,
     process_native_stacktraces,
@@ -229,6 +231,44 @@ def test_instruction_addr_adjustment_none():
     assert not processed_frames[1]["adjust_instruction_addr"]
 
 
+def test_rewrite_electron_debug_file():
+    def rewrite(debug_file):
+        for rule in ELECTRON_FIRST_MODULE_REWRITE_RULES:
+            # Need to patch the regexes and replacement strings here
+            # from Rust to Python syntax.
+            # In regex: ?<group> -> ?P<group>
+            # In replacement: $group -> \g<group>
+            from_patched = re.sub("\\?<", "?P<", rule["from"])
+            to_patched = re.sub("\\$(\\w+)", "\\\\g<\\1>", rule["to"])
+            replaced = re.sub(from_patched, to_patched, debug_file)
+            if replaced != debug_file:
+                return replaced
+
+        return debug_file
+
+    assert rewrite("/home/My Awesome Crasher") == "/home/electron"
+    assert (
+        rewrite("/home/My Awesome Crasher Helper (Renderer)") == "/home/Electron Helper (Renderer)"
+    )
+    assert rewrite("/home/My Awesome Crasher Helper") == "/home/Electron Helper"
+    assert (
+        rewrite("C:/projects/src/out/Default/myapp.exe.pdb")
+        == "C:/projects/src/out/Default/electron.exe.pdb"
+    )
+    assert (
+        rewrite("C:\\projects\\src\\out\\Default\\myapp.exe.pdb")
+        == "C:\\projects\\src\\out\\Default\\electron.exe.pdb"
+    )
+    assert (
+        rewrite("C:\\projects\\src\\out\\Default\\myapp-exe-pdb")
+        == "C:\\projects\\src\\out\\Default\\electron"
+    )
+    assert (
+        rewrite("/home/************/usr/lib/slack/slack")
+        == "/home/************/usr/lib/slack/electron"
+    )
+
+
 @django_db_all
 @mock.patch("sentry.lang.native.processing.Symbolicator")
 def test_il2cpp_symbolication(mock_symbolicator, default_project):