Add Java support to checkdeps.py

tools/checkdeps/cpp_checker.py

Index: tools/checkdeps/cpp_checker.py
diff --git a/tools/checkdeps/cpp_checker.py b/tools/checkdeps/cpp_checker.py
new file mode 100644
index 0000000000000000000000000000000000000000..f1be2e2b79341f71e4cbade2070728423879a087
--- /dev/null
+++ b/tools/checkdeps/cpp_checker.py
@@ -0,0 +1,104 @@
+# Copyright (c) 2012 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.
+
+"""Checks C++ and Objective-C files for illegal includes."""
+
+import codecs
+import re
+import sys
+
+
+class CppChecker(object):
+
+  EXTENSIONS = [
+      '.h',
+      '.cc',
+      '.m',
+      '.mm',
+  ]
+
+  # The maximum number of non-include lines we can see before giving up.
+  _MAX_UNINTERESTING_LINES = 50
+
+  # The maximum line length, this is to be efficient in the case of very long
+  # lines (which can't be #includes).
+  _MAX_LINE_LENGTH = 128
+
+  # This regular expression will be used to extract filenames from include
+  # statements.
+  _EXTRACT_INCLUDE_PATH = re.compile(
+      '[ \t]*#[ \t]*(?:include|import)[ \t]+"(.*)"')
+
+  def __init__(self, verbose):
+    self._verbose = verbose
+
+  def _CheckLine(self, rules, line):
+    """Checks the given file with the given rule set.
+    Returns a tuple (is_include, illegal_description).
+    If the line is an #include directive the first value will be True.
+    If it is also an illegal include, the second value will be a string
+    describing the error.  Otherwise, it will be None."""
+    found_item = self._EXTRACT_INCLUDE_PATH.match(line)
+    if not found_item:
+      return False, None  # Not a match
+
+    include_path = found_item.group(1)
+
+    if '\\' in include_path:
+      return True, 'Include paths may not include backslashes'
+
+    if '/' not in include_path:
+      # Don't fail when no directory is specified. We may want to be more
+      # strict about this in the future.
+      if self._verbose:
+        print ' WARNING: directory specified with no path: ' + include_path
+      return True, None
+
+    (allowed, why_failed) = rules.DirAllowed(include_path)
+    if not allowed:
+      if self._verbose:
+        retval = '\nFor %s' % rules
+      else:
+        retval = ''
+      return True, retval + ('Illegal include: "%s"\n    Because of %s' %
+          (include_path, why_failed))
+
+    return True, None
+
+  def CheckFile(self, rules, filepath):
+    if self._verbose:
+      print 'Checking: ' + filepath
+
+    ret_val = ''  # We'll collect the error messages in here
+    last_include = 0
+    with codecs.open(filepath, encoding='utf-8') as f:
+      in_if0 = 0
+      for line_num in xrange(sys.maxint):

[M-A Ruel, 2012/07/18 14:19:37]

An alternative is:

for line_num, line in enumerate(f):

[Iain Merrick, 2012/07/19 12:44:34]

Good call, done. And I think that's the only reason for importing "sys" so I can
lose that too.

+        if line_num - last_include > self._MAX_UNINTERESTING_LINES:
+          break
+
+        cur_line = f.readline(self._MAX_LINE_LENGTH).strip()

[M-A Ruel, 2012/07/18 14:19:37]

  line = line.strip()

I don't think line 82-83 are a good idea. It would stop too early.

[Iain Merrick, 2012/07/19 12:44:34]

Seemed like a harmless optimization to me, so I left it in...?

Not sure what you mean about stopping early, but sure, happy to get rid of this.

+        if cur_line == '':
+          break
+
+        # Check to see if we're at / inside a #if 0 block
+        if cur_line == '#if 0':

[M-A Ruel, 2012/07/18 14:19:37]

if line.startswith('#if 0'):

would catch: #if 0  // This code is broken.

[Iain Merrick, 2012/07/19 12:44:34]

Good idea, done.

This is just the original algorithm from checkdeps.py, so I don't fully
understand the intentions. But it seems okay for it to catch only *some* of the
commented-out includes. The only way to be completely accurate would be to
preprocess all the files, which is clearly a bad idea.

Therefore, an imperfect heuristic seems fine; but your suggestion is essentially
free and strictly better, so I'll do that.

+          in_if0 += 1
+          continue
+        if in_if0 > 0:
+          if cur_line.startswith('#if'):
+            in_if0 += 1
+          elif cur_line == '#endif':
+            in_if0 -= 1
+          continue
+
+        is_include, line_status = self._CheckLine(rules, cur_line)
+        if is_include:
+          last_include = line_num
+        if line_status is not None:
+          if len(line_status) > 0:  # Add newline to separate messages.
+            line_status += '\n'
+          ret_val += line_status
+
+    return ret_val