Beginnings of flag validator to compare GN and gyp build flags

tools/gn/bin/gyp_flag_compare.py

Index: tools/gn/bin/gyp_flag_compare.py
diff --git a/tools/gn/bin/gyp_flag_compare.py b/tools/gn/bin/gyp_flag_compare.py
new file mode 100755
index 0000000000000000000000000000000000000000..790685af0eef0a9ad5e6f1615f69cf8cf8e13ebc
--- /dev/null
+++ b/tools/gn/bin/gyp_flag_compare.py
@@ -0,0 +1,127 @@
+#!/usr/bin/env python

[awong, 2014/09/03 20:25:35]

Add a newline?

[scottmg, 2014/09/03 20:42:29]

Done. M-A always wants it removed, so now I'm all over the place. :)

+# Copyright 2014 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.
+
+"""Given the output of -t commands from a ninja build for a gyp and GN generated
+build, report on differences between the command lines."""
+
+
+import os
+import sys
+
+
+def FindAndRemoveArgWithValue(command_line, argname):
+  """Given a command line as a list, remove and return the value of an option
+  that takes a value as a separate entry. Modifies |command_line| in place."""

[awong, 2014/09/03 20:25:35]

Follow docstring style?

https://google-styleguide.googlecode.com/svn/trunk/pyguide.html

[scottmg, 2014/09/03 20:42:29]

Done.

+  location = command_line.index(argname)
+  value = command_line[location + 1]
+  command_line[location:location + 2] = []
+  return value
+
+
+def MergeSpacedArgs(command_line, argname):
+  """Combine an argument |argname| with its value separated by a space.
+  Modifies |command_line| in place."""
+  i = 0
+  while i < len(command_line):
+    if command_line[i] == argname:
+      command_line[i:i + 2] = [command_line[i] + ' ' + command_line[i + 1]]

[awong, 2014/09/03 20:25:35]

Why not just create a new array and return? This looks n^2 due to shuffling and
it's really confusing when you mutate an array you're iterating over.

[scottmg, 2014/09/03 20:42:29]

Done.

+    i += 1
+
+
+def GetFlags(lines):
+  """Turn a list of command lines into a semi-structured dict."""
+  flags_by_output = {}
+  for line in lines:
+    # TODO(scottmg): Hacky way of getting only cc for now.
+    if 'clang' not in line:
+      continue
+
+    # TODO(scottmg): Proper escapes.
+    command_line = line.strip().split()[1:]
+
+    output_name = FindAndRemoveArgWithValue(command_line, '-o')
+    dep_name = FindAndRemoveArgWithValue(command_line, '-MF')
+
+    MergeSpacedArgs(command_line, '-Xclang')
+
+    defines = [x for x in command_line if x.startswith('-D')]
+    include_dirs = [x for x in command_line if x.startswith('-I')]
+    dash_f = [x for x in command_line if x.startswith('-f')]
+    warnings = [x for x in command_line if x.startswith('-W')]
+    cc_file = [x for x in command_line if x.endswith('.cc') or x.endswith('.c')]
+    assert len(cc_file) == 1
+    others = [x for x in command_line if x not in defines and \
+                                         x not in include_dirs and \
+                                         x not in dash_f and \
+                                         x not in warnings and \
+                                         x not in cc_file]
+    flags_by_output[cc_file[0]] = {
+      'output': output_name,
+      'depname': dep_name,
+      'defines': sorted(defines),
+      'include_dirs': sorted(include_dirs),  # TODO(scottmg): This is wrong.
+      'dash_f': sorted(dash_f),
+      'warnings': sorted(warnings),
+      'other': sorted(others),
+    }
+  return flags_by_output
+
+
+def CompareLists(gyp, gn, name, dont_care=[]):
+  """Output any differences between to lists, ignoring anything in
+  |dont_care|."""
+  if gyp[name] != gn[name]:
+    print '  %s differ:' % name
+    gyp_set = set(gyp[name]) - set(dont_care)
+    gn_set = set(gn[name]) - set(dont_care)
+    print '    In gyp, but not in GN:\n      %s' % '\n      '.join(
+        sorted(gyp_set - gn_set))
+    print '    In GN, but not in gyp:\n      %s' % '\n      '.join(
+        sorted(gn_set - gyp_set))
+    print
+    return 1

[awong, 2014/09/03 20:25:35]

Use True and False?

[scottmg, 2014/09/03 20:42:29]

Done. (was just because it's used as a main() return code)

+  return 0
+
+
+def main():
+  if len(sys.argv) != 3:
+    print 'usage: %s gyp_commands.txt gn_commands.txt' % __file__
+    return 1
+  gyp = open(sys.argv[1], 'rb').readlines()

[awong, 2014/09/03 20:25:35]

Nit..with syntax is a little cleaner.

with open(...) as gyp:
   blah

[scottmg, 2014/09/03 20:42:29]

Done.

+  gn = open(sys.argv[2], 'rb').readlines()
+  all_gyp_flags = GetFlags(gyp)
+  all_gn_flags = GetFlags(gn)
+  gyp_files = set(all_gyp_flags.keys())
+  gn_files = set(all_gn_flags.keys())
+  diffs = 0

[awong, 2014/09/03 20:25:35]

True and False?

[scottmg, 2014/09/03 20:42:29]

Done.

+  if gyp_files != gn_files:
+    print 'Different set of sources files:'
+    print '  In gyp, not in GN:\n    %s' % '\n    '.join(
+        sorted(gyp_files - gn_files))
+    print '  In GN, not in gyp:\n    %s' % '\n    '.join(
+        sorted(gn_files - gyp_files))
+    print '\nNote that flags will only be compared for files in both sets.\n'
+    diffs |= 1
+  file_list = gyp_files & gn_files
+  for filename in sorted(file_list):
+    gyp_flags = all_gyp_flags[filename]
+    gn_flags = all_gn_flags[filename]
+    print filename
+    diffs |= CompareLists(gyp_flags, gn_flags, 'dash_f')
+    diffs |= CompareLists(gyp_flags, gn_flags, 'defines', dont_care=[
+        '-DENABLE_PRE_SYNC_BACKUP',
+        '-DENABLE_WEBRTC=1',
+        '-DUSE_LIBJPEG_TURBO=1',
+        '-DUSE_PANGO=1',
+        '-DUSE_SYMBOLIZE',
+        ])
+    diffs |= CompareLists(gyp_flags, gn_flags, 'include_dirs')
+    diffs |= CompareLists(gyp_flags, gn_flags, 'warnings')
+    diffs |= CompareLists(gyp_flags, gn_flags, 'other')
+  return diffs
+
+
+if __name__ == '__main__':
+  sys.exit(main())