Add Clang static analyzer to Clang toolchain defs in GN

build/toolchain/clang_static_analyzer_wrapper.py

Index: build/toolchain/clang_static_analyzer_wrapper.py
diff --git a/build/toolchain/clang_static_analyzer_wrapper.py b/build/toolchain/clang_static_analyzer_wrapper.py
new file mode 100755
index 0000000000000000000000000000000000000000..6f70e9a0c6240c95d3ee79aa3397f1cbf6675dd4
--- /dev/null
+++ b/build/toolchain/clang_static_analyzer_wrapper.py
@@ -0,0 +1,61 @@
+#!/usr/bin/env python
+# Copyright 2016 The Chromium Authors. All rights reserved.
+# Use of this source code is governed by a BSD-style license that can be
+# found in the LICENSE file.
+
+"""Invokes the Clang static analysis command using arguments provided on the
+command line.
+"""
+
+import argparse
+import fnmatch
+import os
+import shutil
+import sys
+import tempfile
+
+import wrapper_utils
+
+
+def main():
+  parser = argparse.ArgumentParser(description=__doc__)
+  parser.add_argument('--clang-cc-path',

[Nico, 2017/01/11 20:20:05]

can't the script hardcode that to
script_dir/../../third_party/llvm-build/Release+Asserts/bin/clang? But after
reading the rest of the CL, I actually think what you have is better, so please
ignore.

[Kevin M, 2017/01/18 23:04:27]

Acknowledged.

+                      help='Path to the clang compiler.',
+                      metavar='PATH')
+  parser.add_argument('--clang-cxx-path',
+                      help='Path to the clang++ compiler',
+                      metavar='PATH')
+  parser.add_argument('--analyzer',
+                      help='Path to the language-specific Clang analysis tool.',
+                      required=True,
+                      metavar='PATH')
+  args, compile_args = parser.parse_known_args()
+
+  # Check that only one of --clang-cc-path or --clang-cxx-path are set.
+  assert ((args.clang_cc_path != None) != (args.clang_cxx_path != None))
+  assert ((args.clang_cc_path != None) or (args.clang_cxx_path != None))

[Wez, 2017/01/14 00:04:58]

Isn't this implicit from the != case?  If they were both None then != check
would fail.

[Kevin M, 2017/01/18 23:04:27]

Both of them being set is also a failure case.

[Wez, 2017/01/21 01:10:15]

But if both of them are set (or not-set) then (a != None) == (b != None), which
will fail the first assert, won't it?  For a Boolean value, != is effectively
xor.

[Kevin M, 2017/01/24 01:20:49]

Done.

+
+  is_cxx = args.clang_cxx_path != None
+  env = os.environ
+  env['CCC_ANALYZER_FORCE_ANALYZE_DEBUG_CODE'] = '0'
+  env['CCC_ANALYZER_OUTPUT_FORMAT'] = 'text'
+  clang_path = args.clang_cxx_path or args.clang_cc_path
+  if is_cxx:
+    env['CCC_CXX'] = clang_path
+    env['CLANG_CXX'] = clang_path

[Wez, 2017/01/14 00:04:58]

This differs from the previous version, where you set CLANG - is CLANG_CXX the
correct form?

[Kevin M, 2017/01/18 23:04:28]

Yes.

+  else:
+    env['CCC_CC'] = clang_path
+    env['CLANG'] = clang_path
+
+  temp_dir = tempfile.mkdtemp()

[Wez, 2017/01/14 00:04:58]

Do we want to be dumping stuff in temp files, rather than in some part of the
output directory? Or is this purely for intermediate files that the command
needs to create?

[Kevin M, 2017/01/18 23:04:25]

This is for unsummarized output files. We have some reworking to do to replicate
the summarization behavior of scan-build in a GN-friendly way. It's typically
run as a post-build step after the Makefile has finished its evaluation, which
doesn't map to the granularity of how GN executes its rules.

[Wez, 2017/01/21 01:10:15]

Acknowledged.

+  try:
+    env['CCC_ANALYZER_HTML'] = temp_dir
+    returncode, stderr = wrapper_utils.CaptureCommandStderr(
+        wrapper_utils.CommandToRun([args.analyzer] + compile_args), env)
+    sys.stderr.write(stderr)

[Wez, 2017/01/14 00:04:58]

nit: It seems strange to "capture command stderr", capturing the return code,
and stderr output, only to return the return code and write the stderr output
to... stderr. :P Is this doing something else valuable?

[Kevin M, 2017/01/18 23:04:27]

I think this is the only way to get output to stderr with subprocess.popen - the
only other options are to redirect it to stdout, or to suppress it altogether.

[Wez, 2017/01/21 01:10:15]

According to the documentation the default behavious of subprocess.popen() is
for the child to inherit the stdio handles of the parent.  Failing that, the
call does allow the files to output to to be specified explicitly as parameters.

[Kevin M, 2017/01/24 01:20:49]

Done.

+    return returncode
+  finally:
+    shutil.rmtree(temp_dir)
+
+if __name__ == "__main__":
+  sys.exit(main())