Add an environment variable that will cause md5_check.py to write .stamp file that is debuggable

build/android/gyp/util/md5_check.py

Index: build/android/gyp/util/md5_check.py
diff --git a/build/android/gyp/util/md5_check.py b/build/android/gyp/util/md5_check.py
index 9f365aa08162c83b22192b675a1a53a5e9acb87f..4ac25462e4f40233bf2690e386a4a9ee7b370732 100644
--- a/build/android/gyp/util/md5_check.py
+++ b/build/android/gyp/util/md5_check.py
@@ -2,8 +2,17 @@
 # Use of this source code is governed by a BSD-style license that can be
 # found in the LICENSE file.
 
+import difflib
 import hashlib
 import os
+import re
+
+
+# When set and a difference is detected, a diff of what changed is printed.
+_PRINT_MD5_DIFFS = int(os.environ.get('PRINT_MD5_DIFFS', 0))

[jbudorick, 2015/08/26 16:48:49]

Environment variables tend to come back to bite us where we use them -- it's
very hard to see that they're altering program behavior.

This is debug only, so perhaps it's ok, but typically I'd want to see this
handled via command-line flag (or, in this case, a function argument controlled
by command-line flags elsewhere).

[agrieve, 2015/08/26 17:59:13]

Because a command-line flag would require re-gypping / re-GN'ing, it then causes
a rules that use md5_check.py to be marked as dirty, which kind of eliminates
the point of the flag (need to be able to turn it on without affecting what's
dirty).

The only side-effect it enables is extra printing (does not change md5
calculation at all), so I think it's best as a variable.

[jbudorick, 2015/08/26 18:02:51]

It's dirty on the switch, but not afterwards. Are you getting into states that
aren't (easily) reproducible?

[jbudorick, 2015/09/01 14:59:43]

Hrm. I guess I can live with this for now.

+
+# Used to strip off temp dir prefix.
+_TEMP_DIR_PATTERN = re.compile(r'^/tmp/.*?/')
 
 
 def CallAndRecordIfStale(
@@ -26,7 +35,12 @@ def CallAndRecordIfStale(
       record_path=record_path,
       input_paths=input_paths,
       input_strings=input_strings)
-  if force or md5_checker.IsStale():
+
+  is_stale = md5_checker.old_digest != md5_checker.new_digest
+  if force or is_stale:
+    if is_stale and _PRINT_MD5_DIFFS:
+      print '\033[93mDifference found in %s:\033[0m' % record_path

[jbudorick, 2015/08/26 16:48:49]

It looks like some of the other gyp utilities, notably javac, user colorama for
this.

[agrieve, 2015/08/26 17:59:13]

I looked into this, but IMO makes it more ugly than it's worth:

  if force or is_stale:
      import sys
      from util import build_utils
      if build_utils.COLORAMA_ROOT not in sys.path:
        sys.path.append(build_utils.COLORAMA_ROOT)
      import colorama

      print '%sDifference found in %s:%s' % (
          colorama.Fore.YELLOW, record_path, colorama.Fore.RESET)
      print md5_checker.DescribeDifference()

[jbudorick, 2015/08/26 18:02:51]

I'm not really interested in doing this in two different ways in this directory,
one with hand-rolled symbols and one with a library.

[jbudorick, 2015/09/01 14:59:43]

This, on the other hand...

[agrieve, 2015/09/01 17:32:58]

Colorama'ed.

+      print md5_checker.DescribeDifference()
     function()
     md5_checker.Write()
 
@@ -53,6 +67,14 @@ def _UpdateMd5ForPath(md5, path):
     _UpdateMd5ForFile(md5, path)
 
 
+def _TrimPathPrefix(path):
+  """Attempts to remove temp dir prefix from the path.
+
+  Use this only for extended_info (not for the actual md5).
+  """
+  return _TEMP_DIR_PATTERN.sub('{TMP}', path)
+
+
 class _Md5Checker(object):
   def __init__(self, record_path=None, input_paths=None, input_strings=None):
     if not input_paths:
@@ -66,21 +88,41 @@ class _Md5Checker(object):
 
     self.record_path = record_path
 
-    md5 = hashlib.md5()
+    extended_info = []
+    outer_md5 = hashlib.md5()
     for i in sorted(input_paths):
-      _UpdateMd5ForPath(md5, i)
+      inner_md5 = hashlib.md5()
+      _UpdateMd5ForPath(inner_md5, i)
+      i = _TrimPathPrefix(i)
+      extended_info.append(i + '=' + inner_md5.hexdigest())
+      # Include the digest in the overall diff, but not the path
+      outer_md5.update(inner_md5.hexdigest())
+
     for s in input_strings:
-      md5.update(s)
-    self.new_digest = md5.hexdigest()
+      outer_md5.update(s)
+      extended_info.append(s)
+
+    self.new_digest = outer_md5.hexdigest()
+    self.new_extended_info = extended_info
 
     self.old_digest = ''
+    self.old_extended_info = []
     if os.path.exists(self.record_path):
       with open(self.record_path, 'r') as old_record:
-        self.old_digest = old_record.read()
-
-  def IsStale(self):
-    return self.old_digest != self.new_digest
+        self.old_extended_info = [line.strip() for line in old_record]
+        self.old_digest = self.old_extended_info.pop(0)
 
   def Write(self):
     with open(self.record_path, 'w') as new_record:
       new_record.write(self.new_digest)
+      new_record.write('\n' + '\n'.join(self.new_extended_info) + '\n')

[jbudorick, 2015/08/26 16:48:49]

so now we're always writing the extended info?

[agrieve, 2015/08/26 17:59:13]

Yes. I measured if this slowed anything down, and it doesn't.

+
+  def DescribeDifference(self):
+    if self.old_digest == self.new_digest:
+      return 'There\'s no difference'

[jbudorick, 2015/08/26 16:48:49]

nit: double quotes when you've got an apostrophe.

[agrieve, 2015/08/26 17:59:13]

Done.

+    if not self.old_digest:
+      return 'Previous stamp file not found.'
+    if not self.old_extended_info:
+      return 'Previous stamp file lacks extended info.'
+    diff = difflib.unified_diff(self.old_extended_info, self.new_extended_info)
+    return '\n'.join(diff)