Make javac invocations incremental when possible

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 054caa491460de28f056f767dfd605d85fa54ad1..c738606872f9da187427cc8bc7a2eb76de431151 100644
--- a/build/android/gyp/util/md5_check.py
+++ b/build/android/gyp/util/md5_check.py
@@ -4,21 +4,20 @@
 
 import difflib
 import hashlib
+import itertools
+import json
 import os
-import re
 import sys
+import zipfile
 
 
 # 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))
 
-# Used to strip off temp dir prefix.
-_TEMP_DIR_PATTERN = re.compile(r'^/tmp/.*?/')
-
 
 def CallAndRecordIfStale(
     function, record_path=None, input_paths=None, input_strings=None,
-    output_paths=None, force=False):
+    output_paths=None, force=False, pass_changes=False):
   """Calls function if outputs are stale.
 
   Outputs are considered stale if:
@@ -36,33 +35,315 @@ def CallAndRecordIfStale(
     input_paths: List of paths to calcualte an md5 sum on.
     input_strings: List of strings to record verbatim.
     output_paths: List of output paths.
-    force: When True, function is always called.
+    force: Whether to treat outputs as missing regardless of whether they
+      actually are.
+    pass_changes: Whether to pass a Changes instance to |function|.
   """
   input_paths = input_paths or []
   input_strings = input_strings or []
   output_paths = output_paths or []
   record_path = record_path or output_paths[0] + '.md5.stamp'
-  md5_checker = _Md5Checker(
-      record_path=record_path,
-      input_paths=input_paths,
-      input_strings=input_strings)
-
-  missing_outputs = [x for x in output_paths if not os.path.exists(x)]
-  is_stale = md5_checker.old_digest != md5_checker.new_digest
-
-  if force or missing_outputs or is_stale:
-    if _PRINT_MD5_DIFFS:
-      print '=' * 80
-      print 'Difference found in %s:' % record_path
-      if missing_outputs:
-        print 'Outputs do not exist:\n' + '\n'.join(missing_outputs)
-      elif force:
-        print 'force=True'
-      else:
-        print md5_checker.DescribeDifference()
-      print '=' * 80
-    function()
-    md5_checker.Write()
+
+  assert record_path.endswith('.stamp'), (
+      'record paths must end in \'.stamp\' so that they are easy to find '
+      'and delete')
+
+  new_metadata = _Metadata()
+  new_metadata.AddStrings(input_strings)
+
+  for path in input_paths:
+    if _IsZipFile(path):
+      entries = _ExtractZipEntries(path)
+      new_metadata.AddZipFile(path, entries)
+    else:
+      new_metadata.AddFile(path, _Md5ForPath(path))
+
+  old_metadata = None
+  missing_outputs = [x for x in output_paths if force or not os.path.exists(x)]
+  # When outputs are missing, don't bother gathering change information.
+  if not missing_outputs and os.path.exists(record_path):
+    with open(record_path, 'r') as jsonfile:
+      try:
+        old_metadata = _Metadata.FromFile(jsonfile)
+      except:  # pylint: disable=bare-except
+        pass  # Not yet using new file format.
+
+  changes = Changes(old_metadata, new_metadata, force, missing_outputs)
+  if not changes.HasChanges():
+    return
+
+  if _PRINT_MD5_DIFFS:
+    print '=' * 80
+    print 'Target is stale: %s' % record_path
+    print changes.DescribeDifference()
+    print '=' * 80
+
+  # Delete the old metdata beforehand since failures leave it in an
+  # inderterminate state.
+  if old_metadata:
+    os.unlink(record_path)
+
+  args = (changes,) if pass_changes else ()
+  function(*args)
+
+  with open(record_path, 'w') as f:
+    new_metadata.Write(f)
+
+
+class Changes(object):
+  """Provides and API for querying what changed between runs."""
+
+  def __init__(self, old_metadata, new_metadata, force, missing_outputs):
+    self.old_metadata = old_metadata
+    self.new_metadata = new_metadata
+    self.force = force
+    self.missing_outputs = missing_outputs
+
+  def _GetOldTag(self, path, subpath=None):
+    return self.old_metadata and self.old_metadata.GetTag(path, subpath)
+
+  def HasChanges(self):
+    """Returns whether any changes exist."""
+    return (self.force or
+            self.old_metadata is None or

[jbudorick, 2015/09/23 00:26:20]

not self.old_metadata?

[agrieve, 2015/09/23 02:07:56]

Done.

+            self.old_metadata.StringsMd5() != self.new_metadata.StringsMd5() or
+            self.old_metadata.FilesMd5() != self.new_metadata.FilesMd5())
+
+  def AddedOrModifiedOnly(self):
+    """Returns whether the only changes were from added or modified (sub)files.
+
+    No missing outputs, no removed paths/subpaths.
+    """
+    if (self.force or
+        not self.old_metadata or
+        self.old_metadata.StringsMd5() != self.new_metadata.StringsMd5()):
+      return False
+    if any(self.IterRemovedPaths()):
+      return False
+    for path in self.IterModifiedPaths():
+      if any(self.IterRemovedSubpaths(path)):
+        return False
+    return True
+
+  def IterAddedPaths(self):
+    """Generator for paths that were added."""
+    for path in self.new_metadata.IterPaths():
+      if self._GetOldTag(path) is None:
+        yield path
+
+  def IterAddedSubpaths(self, path):
+    """Generator for paths that were added within the given zip file."""
+    for subpath in self.new_metadata.IterSubpaths(path):
+      if self._GetOldTag(path, subpath) is None:
+        yield subpath
+
+  def IterRemovedPaths(self):
+    """Generator for paths that were removed."""
+    if self.old_metadata:
+      for path in self.old_metadata.IterPaths():
+        if self.new_metadata.GetTag(path) is None:
+          yield path
+
+  def IterRemovedSubpaths(self, path):
+    """Generator for paths that were removed within the given zip file."""
+    if self.old_metadata:
+      for subpath in self.old_metadata.IterSubpaths(path):
+        if self.new_metadata.GetTag(path, subpath) is None:
+          yield subpath
+
+  def IterModifiedPaths(self):
+    """Generator for paths whose contents have changed."""
+    for path in self.new_metadata.IterPaths():
+      old_tag = self._GetOldTag(path)
+      new_tag = self.new_metadata.GetTag(path)
+      if old_tag is not None and old_tag != new_tag:
+        yield path
+
+  def IterModifiedSubpaths(self, path):
+    """Generator for paths within a zip file whose contents have changed."""
+    for subpath in self.new_metadata.IterSubpaths(path):
+      old_tag = self._GetOldTag(path, subpath)
+      new_tag = self.new_metadata.GetTag(path, subpath)
+      if old_tag is not None and old_tag != new_tag:
+        yield subpath
+
+  def IterChangedPaths(self, path):
+    """Generator for all changed paths (added/removed/modified)."""
+    return itertools.chain(self.IterRemovedPaths(path),
+                           self.IterModifiedPaths(path),
+                           self.IterAddedPaths(path))
+
+  def IterChangedSubpaths(self, path):
+    """Generator for paths within a zip that were added/removed/modified."""
+    return itertools.chain(self.IterRemovedSubpaths(path),
+                           self.IterModifiedSubpaths(path),
+                           self.IterAddedSubpaths(path))
+
+  def DescribeDifference(self):
+    """Returns a human-readable description of what changed."""
+    if self.force:
+      return 'force=True'
+    elif self.missing_outputs:
+      return 'Outputs do not exist:\n  ' + '\n  '.join(self.missing_outputs)
+    elif self.old_metadata is None:
+      return 'Previous stamp file not found.'
+
+    if self.old_metadata.StringsMd5() != self.new_metadata.StringsMd5():
+      ndiff = difflib.ndiff(self.old_metadata.GetStrings(),
+                            self.new_metadata.GetStrings())
+      changed = [s for s in ndiff if not s.startswith(' ')]
+      return 'Input strings changed:\n  ' + '\n  '.join(changed)
+
+    if self.old_metadata.FilesMd5() == self.new_metadata.FilesMd5():
+      return "There's no difference."
+
+    lines = []
+    lines.extend('Added: ' + p for p in self.IterAddedPaths())
+    lines.extend('Removed: ' + p for p in self.IterRemovedPaths())
+    for path in self.IterModifiedPaths():
+      lines.append('Modified: ' + path)
+      lines.extend('  -> Subpath added: ' + p
+                   for p in self.IterAddedSubpaths(path))
+      lines.extend('  -> Subpath removed: ' + p
+                   for p in self.IterRemovedSubpaths(path))
+      lines.extend('  -> Subpath modified: ' + p
+                   for p in self.IterModifiedSubpaths(path))
+    if lines:
+      return 'Input files changed:\n  ' + '\n  '.join(lines)
+    return 'I have no idea what changed (there is a bug).'

[jbudorick, 2015/09/23 00:26:20]

hahaha

+
+
+class _Metadata(object):
+  """Data model for tracking change metadata."""
+  # Schema:
+  # {
+  #   "files-md5": "VALUE",
+  #   "strings-md5": "VALUE",
+  #   "input-files": [
+  #     {
+  #       "path": "path.jar",
+  #       "tag": "{MD5 of entries}",
+  #       "entries": [
+  #         { "path": "org/chromium/base/Foo.class", "tag": "{CRC32}" }, ...
+  #       ]
+  #     }, {
+  #       "path": "path.txt",
+  #       "tag": "{MD5}",

[jbudorick, 2015/09/23 00:26:20]

nit: missing closing }

[agrieve, 2015/09/23 02:07:56]

Done.

+  #   ],
+  #   "input-strings": ["a", "b", ...],
+  # }
+  def __init__(self):
+    self._files_md5 = None
+    self._strings_md5 = None
+    self._files = []
+    self._strings = []
+    # Map of (path, subpath) -> entry. Created upon first call to _GetEntry().
+    self._file_map = None
+
+  @classmethod
+  def FromFile(cls, fileobj):

[jbudorick, 2015/09/23 00:26:20]

not Read/Write or FromFile/ToFile? :(

[agrieve, 2015/09/23 02:07:56]

Done.

+    """Returns a _Metadata initialized from a file object."""
+    ret = cls()
+    obj = json.load(fileobj)
+    ret._files_md5 = obj['files-md5']
+    ret._strings_md5 = obj['strings-md5']
+    ret._files = obj['input-files']
+    ret._strings = obj['input-strings']
+    return ret
+
+  def Write(self, fileobj):
+    """Serializes metadata to the given file object."""
+    obj = {
+        "files-md5": self.FilesMd5(),
+        "strings-md5": self.StringsMd5(),
+        "input-files": self._files,
+        "input-strings": self._strings,
+    }
+    json.dump(obj, fileobj, indent=2)
+
+  def _AssertNotQueried(self):
+    assert self._files_md5 is None
+    assert self._strings_md5 is None
+    assert self._file_map is None
+
+  def AddStrings(self, values):
+    self._AssertNotQueried()
+    self._strings.extend(str(v) for v in values)
+
+  def AddFile(self, path, tag):
+    """Adds metadata for a non-zip file.
+
+    Args:
+      path: Path to the file.
+      tag: A short string representative of the file contents.
+    """
+    self._AssertNotQueried()
+    self._files.append({
+        'path': path,
+        'tag': tag,
+    })
+
+  def AddZipFile(self, path, entries):
+    """Adds metadata for a zip file.
+
+    Args:
+      path: Path to the file.
+      entries: List of (subpath, tag) tuples for entries within the zip.
+    """
+    self._AssertNotQueried()
+    tag = _ComputeInlineMd5(itertools.chain((e[0] for e in entries),
+                                            (e[1] for e in entries)))
+    self._files.append({
+        'path': path,
+        'tag': tag,
+        'entries': [{"path": e[0], "tag": e[1]} for e in entries],
+    })
+
+  def GetStrings(self):
+    """Returns the list of input strings."""
+    return self._strings
+
+  def FilesMd5(self):
+    """Lazily computes and returns the aggregate md5 of input files."""
+    if self._files_md5 is None:
+      # Omit paths from md5 since temporary files have random names.
+      self._files_md5 = _ComputeInlineMd5(
+          self.GetTag(p) for p in sorted(self.IterPaths()))
+    return self._files_md5
+
+  def StringsMd5(self):
+    """Lazily computes and returns the aggregate md5 of input strings."""
+    if self._strings_md5 is None:
+      self._strings_md5 = _ComputeInlineMd5(self._strings)
+    return self._strings_md5
+
+  def _GetEntry(self, path, subpath=None):
+    """Returns the JSON entry for the given path / subpath."""
+    if self._file_map is None:
+      self._file_map = {}
+      for entry in self._files:
+        self._file_map[(entry['path'], None)] = entry
+        for subentry in entry.get('entries', ()):
+          self._file_map[(entry['path'], subentry['path'])] = subentry
+    return self._file_map.get((path, subpath))
+
+  def GetTag(self, path, subpath=None):
+    """Returns the tag for the given path / subpath."""
+    ret = self._GetEntry(path, subpath)
+    return ret and ret['tag']
+
+  def IterPaths(self):
+    """Returns a generator for all top-level paths."""
+    return (e['path'] for e in self._files)
+
+  def IterSubpaths(self, path):
+    """Returns a generator for all subpaths in the given zip.
+
+    If the given path is not a zip file, returns an empty generator.
+    """
+    outer_entry = self._GetEntry(path)
+    subentries = outer_entry.get('entries', [])
+    return (entry['path'] for entry in subentries)
 
 
 def _UpdateMd5ForFile(md5, path, block_size=2**16):
@@ -80,70 +361,37 @@ def _UpdateMd5ForDirectory(md5, dir_path):
       _UpdateMd5ForFile(md5, os.path.join(root, f))
 
 
-def _UpdateMd5ForPath(md5, path):
+def _Md5ForPath(path):
+  md5 = hashlib.md5()
   if os.path.isdir(path):
     _UpdateMd5ForDirectory(md5, path)
   else:
     _UpdateMd5ForFile(md5, path)
+  return md5.hexdigest()
 
 
-def _TrimPathPrefix(path):
-  """Attempts to remove temp dir prefix from the path.
+def _ComputeInlineMd5(iterable):
+  """Computes the md5 of the concatenated parameters."""
+  md5 = hashlib.md5()
+  for item in iterable:
+    md5.update(str(item))
+  return md5.hexdigest()
 
-  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:
-      input_paths = []
-    if not input_strings:
-      input_strings = []
-
-    assert record_path.endswith('.stamp'), (
-        'record paths must end in \'.stamp\' so that they are easy to find '
-        'and delete')
-
-    self.record_path = record_path
-
-    extended_info = []
-    outer_md5 = hashlib.md5()
-    for i in sorted(input_paths):
-      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 map(str, input_strings):
-      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_extended_info = [line.strip() for line in old_record]
-        if self.old_extended_info:
-          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')
 
-  def DescribeDifference(self):
-    if self.old_digest == self.new_digest:
-      return "There's no difference."
-    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)
+def _IsZipFile(path):
+  """Returns whether to treat the given file as a zip file."""
+  # ijar doesn't set the CRC32 field.
+  if path.endswith('.interface.jar'):
+    return False
+  return path.endswith('.zip') or path.endswith('.apk') or path.endswith('.jar')
+
+
+def _ExtractZipEntries(path):
+  """Returns a list of (path, CRC32) of all files within |path|."""
+  entries = []
+  with zipfile.ZipFile(path) as zip_file:
+    for zip_info in zip_file.infolist():
+      # Skip directories and empty files.
+      if zip_info.CRC:
+        entries.append((zip_info.filename, zip_info.CRC))
+  return entries