Reland "[gclient] Read submodule status information

This reverts commit 58ff1badf9.

This change also restricts diff only to changes that are submodule related by
utilizing patch diff search.

Original change's description:
> [gclient] Read submodule status information
>
> This allow us to skip sync if we know the state is correct.
>
> R=gavinmak@google.com
>
> Bug: 40283612, 40942309
> Change-Id: I30fd5bfb9ca8ab0f7dcce567e2a5cb4aebdc7b2f
> Reviewed-on: https://chromium-review.googlesource.com/c/chromium/tools/depot_tools/+/5480172
> Commit-Queue: Josip Sokcevic <sokcevic@chromium.org>
> Reviewed-by: Gavin Mak <gavinmak@google.com>

Bug: 40283612, 40942309
Change-Id: Iac7ed8c927de1a03a3d60dd50108ddb0b6d96c7e
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/tools/depot_tools/+/5513190
Reviewed-by: Gavin Mak <gavinmak@google.com>
Commit-Queue: Josip Sokcevic <sokcevic@chromium.org>

gclient_scm.py

@@ -229,6 +229,7 @@ class GitWrapper(SCMWrapper):
             filter_kwargs['predicate'] = self.out_cb
         self.filter = gclient_utils.GitFilter(**filter_kwargs)
         self._running_under_rosetta = None
+        self.current_revision = None
 
     def GetCheckoutRoot(self):
         return scm.GIT.GetCheckoutRoot(self.checkout_path)
@@ -250,6 +251,75 @@ class GitWrapper(SCMWrapper):
                     ['-c', 'core.quotePath=false', 'diff', '--name-only', base])
             )).split()
 
+    def GetSubmoduleStateFromIndex(self):
+        """Returns a map where keys are submodule names and values are commit
+        hashes. It reads data from the Git index, so only committed values are
+        present."""
+        out = self._Capture(['ls-files', '-s'])
+        result = {}
+        for l in out.split('\n'):
+            if not l.startswith('160000'):
+                # Not a submodule
+                continue
+            (_, commit, _, filepath) = l.split(maxsplit=3)
+            result[filepath] = commit
+        return result
+
+    def GetSubmoduleDiff(self):
+        """Returns a map where keys are submodule names and values are tuples of
+        (old_commit_hash, new_commit_hash). old_commit_hash matches the Git
+        index, whereas new_commit_hash matches currently checked out commit
+        hash."""
+        out = self._Capture([
+            'diff',
+            '--no-prefix',
+            '--no-ext-diff',
+            '--no-color',
+            '--ignore-submodules=dirty',
+            '--submodule=short',
+            '-G',
+            'Subproject commit',
+        ])
+        NO_COMMIT = 40 * '0'
+        committed_submodule = None
+        checked_submodule = None
+        filepath = None
+        state = 0
+        diff = {}
+        # Parsing git diff uses simple state machine. States:
+        # 0 - start state
+        # 1 - diff file/line detected, ready to process content
+        # 2 - gitlink detected, ready to process gitlink past and current
+        # content.
+        # 3 - past gitlink content detected. It contains a commit hash that's in
+        # git index.
+        # 4 - new gitlink content detected. It contains currently checked
+        # commit. At this point, we have all information needed, and we can
+        # reset state to 0.
+        for l in out.split('\n'):
+            if l.startswith('diff --git'):
+                # New file detected, reset state.
+                state = 1
+            elif state == 1 and l.startswith('index') and l.endswith('160000'):
+                # We detected gitlink
+                state = 2
+            elif state == 2 and l.startswith('+++ '):
+                # This line contains filename
+                filepath = l[4:]
+                state = 3
+            elif state == 3 and l.startswith('-Subproject commit '):
+                # This line contains what commit hash Git index expects
+                # (ls-files).
+                committed_submodule = l.split(' ')[-1]
+                state = 4
+            elif state == 4 and l.startswith('+Subproject commit '):
+                # This line contains currently checked out commit for this submodule.
+                checked_submodule = l.split(' ')[-1]
+                if NO_COMMIT not in (committed_submodule, checked_submodule):
+                    diff[filepath] = (committed_submodule, checked_submodule)
+                state = 0
+        return diff
+
     def diff(self, options, _args, _file_list):
         _, revision = gclient_utils.SplitUrlRevision(self.url)
         if not revision:
@@ -638,7 +708,6 @@ class GitWrapper(SCMWrapper):
             raise gclient_utils.Error("Unsupported argument(s): %s" %
                                       ",".join(args))
 
-        current_revision = None
         url, deps_revision = gclient_utils.SplitUrlRevision(self.url)
         revision = deps_revision
         managed = True
@@ -714,11 +783,11 @@ class GitWrapper(SCMWrapper):
                 self._UpdateMirrorIfNotContains(mirror, options, rev_type,
                                                 revision)
             try:
-                current_revision = self._Clone(revision, url, options)
+                self.current_revision = self._Clone(revision, url, options)
             except subprocess2.CalledProcessError as e:
                 logging.warning('Clone failed due to: %s', e)
                 self._DeleteOrMove(options.force)
-                current_revision = self._Clone(revision, url, options)
+                self.current_revision = self._Clone(revision, url, options)
             if file_list is not None:
                 files = self._Capture(
                     ['-c', 'core.quotePath=false', 'ls-files']).splitlines()
@@ -742,6 +811,17 @@ class GitWrapper(SCMWrapper):
                        self.relpath)
             return self._Capture(['rev-parse', '--verify', 'HEAD'])
 
+        # Special case for rev_type = hash. If we use submodules, we can check
+        # information already.
+        if rev_type == 'hash':
+            if self.current_revision == revision:
+                if verbose:
+                    self.Print('Using submodule information to skip check')
+                if options.reset or options.force:
+                    self._Scrub('HEAD', options)
+
+                return revision
+
         self._maybe_break_locks(options)
 
         if mirror: