valgrind/asan: update asan_symbolize.py

We need to update this for https://reviews.llvm.org/D60529

Bug: 1288169
Change-Id: I3c8f5b3f173be0ce5771e279f6a402deaafb9501
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/3301851
Auto-Submit: Takuto Ikuta <tikuta@chromium.org>
Reviewed-by: Nico Weber <thakis@chromium.org>
Commit-Queue: Nico Weber <thakis@chromium.org>
Cr-Commit-Position: refs/heads/main@{#964541}

.yapfignore

@@ -4,6 +4,7 @@
 # Third-party code.
 third_party/blink/tools/blinkpy/third_party/*
 third_party/blink/web_tests/external/wpt/*
+tools/valgrind/asan/third_party/asan_symbolize.py
 
 # TODO(crbug.com/1116155): Enable this for formatting by yapf.
 tools/json_schema_compiler/*

tools/valgrind/asan/asan_symbolize.py

@@ -100,30 +100,6 @@ def find_inode_at_path(inode, path):
   return ret
 
 
-# Create a binary name filter that works around https://crbug.com/444835.
-# When running tests on OSX swarming servers, ASan sometimes prints paths to
-# files in cache (ending with SHA1 filenames) instead of paths to hardlinks to
-# those files in the product dir.
-# For a given |binary_path| chrome_osx_binary_name_filter() returns one of the
-# hardlinks to the same inode in |product_dir_path|.
-def make_chrome_osx_binary_name_filter(product_dir_path=''):
-  def chrome_osx_binary_name_filter(binary_path):
-    basename = os.path.basename(binary_path)
-    if is_hash_name(basename) and product_dir_path:
-      inode = os.stat(binary_path).st_ino
-      new_binary_path = find_inode_at_path(inode, product_dir_path)
-      if new_binary_path:
-        return new_binary_path
-    return binary_path
-  return chrome_osx_binary_name_filter
-
-
-def make_sysroot_filter(sysroot):
-  """Creates a binary name filter for a symbol tree of a remote system."""
-
-  return lambda binary_path: sysroot + binary_path
-
-
 # Construct a path to the .dSYM bundle for the given binary.
 # There are three possible cases for binary location in Chromium:
 # 1. The binary is a standalone executable or dynamic library in the product
@@ -182,6 +158,11 @@ def symbolize_snippet(self, snippet):
   def symbolize(self, test_run):
     original_snippet = base64.b64decode(
         test_run['output_snippet_base64']).decode('utf-8', 'replace')
+
+    # replace non-ascii character with '?'.
+    original_snippet = ''.join(i if i <= u'~' else u'?'
+                               for i in original_snippet)
+
     symbolized_snippet = self.symbolize_snippet(original_snippet)
     if symbolized_snippet == original_snippet:
       # No sanitizer reports in snippet.
@@ -211,6 +192,26 @@ def symbolize_snippets_in_json(filename, symbolization_loop):
     json.dump(json_data, f, indent=3, sort_keys=True)
 
 
+class macOSBinaryNameFilterPlugin(asan_symbolize.AsanSymbolizerPlugIn):
+  def __init__(self):
+    self.product_dir_path = ''
+
+  def filter_binary_path(self, binary_path):
+    # Create a binary name filter that works around https://crbug.com/444835.
+    # When running tests on OSX swarming servers, ASan sometimes prints paths to
+    # files in cache (ending with SHA1 filenames) instead of paths to hardlinks
+    # to those files in the product dir.
+    # For a given |binary_path| macOSBinaryNameFilterPlugin returns one of the
+    # hardlinks to the same inode in |product_dir_path|.
+    basename = os.path.basename(binary_path)
+    if is_hash_name(basename) and self.product_dir_path:
+      inode = os.stat(binary_path).st_ino
+      new_binary_path = find_inode_at_path(inode, self.product_dir_path)
+      if new_binary_path:
+        return new_binary_path
+    return binary_path
+
+
 def main():
   parser = argparse.ArgumentParser(description='Symbolize sanitizer reports.')
   parser.add_argument('--test-summary-json-file',
@@ -234,22 +235,27 @@ def main():
   # Most source paths for Chromium binaries start with
   # /path/to/src/out/Release/../../
   asan_symbolize.fix_filename_patterns.append('Release/../../')
-  binary_name_filter = None
-  if args.sysroot:
-    binary_name_filter = make_sysroot_filter(args.sysroot)
-  elif platform.uname()[0] == 'Darwin':
-    binary_name_filter = make_chrome_osx_binary_name_filter(
-        chrome_product_dir_path(args.executable_path))
-  loop = asan_symbolize.SymbolizationLoop(
-      binary_name_filter=binary_name_filter,
-      dsym_hint_producer=chrome_dsym_hints)
-
-  if args.test_summary_json_file:
-    symbolize_snippets_in_json(args.test_summary_json_file, loop)
-  else:
-    # Process stdin.
-    asan_symbolize.logfile = sys.stdin
-    loop.process_logfile()
+
+  with asan_symbolize.AsanSymbolizerPlugInProxy() as plugin_proxy:
+    if args.sysroot:
+      sysroot_filter = asan_symbolize.SysRootFilterPlugIn()
+      sysroot_filter.sysroot_path = args.sysroot
+      plugin_proxy.add_plugin(sysroot_filter)
+    elif platform.uname()[0] == 'Darwin':
+      macos_filter = macOSBinaryNameFilterPlugin()
+      macos_filter.product_dir_path = args.executable_path
+      plugin_proxy.add_plugin(macos_filter)
+
+    loop = asan_symbolize.SymbolizationLoop(
+        plugin_proxy=plugin_proxy, dsym_hint_producer=chrome_dsym_hints)
+
+    if args.test_summary_json_file:
+      symbolize_snippets_in_json(args.test_summary_json_file, loop)
+    else:
+      # Process stdin.
+      asan_symbolize.logfile = sys.stdin
+      loop.process_logfile()
+
 
 if __name__ == '__main__':
   main()

tools/valgrind/asan/third_party/README.chromium

@@ -1,7 +1,7 @@
 Name: asan_symbolize.py
-License: University of Illinois Open Source License.
-Version: r227327
-URL: http://llvm.org/viewvc/llvm-project/compiler-rt/trunk/lib/asan/scripts/asan_symbolize.py?view=co&content-type=text%2Fplain
+License: Apache-2.0 WITH LLVM-exception
+Version: a81e45b8bcb8eb274ad73357e10e2cdf8a314a8c
+URL: https://github.com/llvm/llvm-project/blob/a81e45b8bcb8eb274ad73357e10e2cdf8a314a8c/compiler-rt/lib/asan/scripts/asan_symbolize.py
 Security Critical: no
 
 asan_symbolize.py is a verbatim copy of asan_symbolize.py in the LLVM trunk.