[clang] Drop generated files in compilation database.

Generated files (e.g. bind_unittest_nc.cc) are less interesting than regular
human-written files, e.g. it doesn't make sense to apply clang-tidy FixIts to
them; clangd indexer are not interested in indexing them.

And these files are usually not generated when running clang-based tools.

size of compilation databse: 204M => 176M
number of entries : 44450 => 38198

Change-Id: I32c639562187f814a9286f945b9dfcfbc4ab21d3
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/1526264
Commit-Queue: Daniel Cheng <dcheng@chromium.org>
Auto-Submit: Haojian Wu <hokein.wu@gmail.com>
Reviewed-by: Daniel Cheng <dcheng@chromium.org>
Cr-Commit-Position: refs/heads/master@{#644005}

tools/clang/pylib/clang/compile_db.py

@@ -102,6 +102,17 @@ def GetNinjaPath():
         '..', '..', '..', '..', 'third_party', 'depot_tools', ninja_executable)
 
 
+def IsGeneratedFile(entry):
+  """Check whether an entry in a compilation database is for generated file.
+  """
+  build_dir = entry['directory']
+  # According to https://www.chromium.org/developers/generated-files, Chromium's
+  # generated files are output in <build_dir>/gen directory.
+  gen_dir = os.path.join(build_dir, 'gen', '') # with trailing path separator
+  absolute_file_path = os.path.normpath(os.path.join(build_dir, entry['file']))
+  return absolute_file_path.startswith(gen_dir)
+
+
 # FIXME: This really should be a build target, rather than generated at runtime.
 def GenerateWithNinja(path, targets=[]):
   """Generates a compile database using ninja.
@@ -116,10 +127,13 @@ def GenerateWithNinja(path, targets=[]):
   # TODO(dcheng): Ensure that clang is enabled somehow.
 
   # First, generate the compile database.
-  json_compile_db = subprocess.check_output(
+  json_compile_db = json.loads(subprocess.check_output(
       [GetNinjaPath(), '-C', path] + targets +
-      ['-t', 'compdb', 'cc', 'cxx', 'objc', 'objcxx'])
-  return json.loads(json_compile_db)
+      ['-t', 'compdb', 'cc', 'cxx', 'objc', 'objcxx']))
+  # Drop generated files.
+  # The directory of the compliation database entry is expected to be the
+  # absolute path of the given 'path'.
+  return [entry for entry in json_compile_db if not IsGeneratedFile(entry) ]
 
 
 def Read(path):

tools/clang/pylib/clang/compile_db_test.py

@@ -73,6 +73,26 @@ class CompileDbTest(unittest.TestCase):
     for actual, expected in zip(processed_compile_db, _EXPECTED_COMPILE_DB):
       self.assertDictEqual(actual, expected)
 
+  def testIsGeneratedFiles(self):
+    if sys.platform != 'win32':
+      return
+
+    test_cases = [
+        ({
+            'directory': 'chromium/src/out/Release',
+            'command': 'clang foo.cpp',
+            'file': '../../chrome/foo.cpp'
+        }, False),
+        ({
+            'directory': 'chromium/src/out/Release',
+            'command': 'clang gen.cpp',
+            'file': 'gen/gen.cpp'
+        }, True),
+    ]
+
+    for entry, expected in test_cases:
+      self.assertEqual(compile_db.IsGeneratedFile(entry), expected)
+
 
 if __name__ == '__main__':
   unittest.main()