[android] Stop using isolate.py for data dependency management.

build/android/pylib/utils/device_dependencies.py

+# Copyright 2016 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.
+
+import ast
+import os
+
+from pylib import constants
+
+
+def DevicePathComponentsFor(host_path, output_directory):
+  """Returns the device path components for a given host path.
+
+  This returns the device path as a list of joinable path components,
+  with None as the first element to indicate that the path should be
+  rooted at $EXTERNAL_STORAGE.
+
+  e.g., given
+
+    '$CHROMIUM_SRC/foo/bar/baz.txt'
+
+  this would return
+
+    [None, 'foo', 'bar', 'baz.txt']
+
+  This handles a couple classes of paths differently than it otherwise would:
+    - All .pak files get mapped to top-level paks/
+    - Anything in the output directory gets mapped relative to the output
+      directory rather than the source directory.
+
+  e.g. given
+
+    '$CHROMIUM_SRC/out/Release/icu_fake_dir/icudtl.dat'
+
+  this would return
+
+    [None, 'icu_fake_dir', 'icudtl.dat']

[Dirk Pranke, 2016/11/15 22:33:06]

I don't know any of the surrounding code, so don't take this comment as a
blocker, but this seems kind of strange that you can't distinguish between
something relative to //out and something relative to //out/Release. Why not map
everything to //out/Release?

[jbudorick, 2016/11/16 03:47:51]

I'm not sure what you mean here, which leads me to believe that my comment may
be written confusingly. The distinction isn't between things in //out and
//out/Release; it's between things in //out/Release and things not in //out at
all.

[Dirk Pranke, 2016/11/17 03:05:45]

Ah, it doesn't help that I wrote "//out" when I meant to write "//". So I
contributed a fair
amount to the confusion.

That said, I think I just didn't understand what was going on. I now see that
this is specifying
the place to write *to*. I was thinking it was the place to read *from*.

+
+  Args:
+    host_path: The absolute path to the host file.
+  Returns:
+    A list of device path components.
+  """
+  if host_path.startswith(output_directory):
+    if os.path.splitext(host_path)[1] == '.pak':
+      return [None, 'paks', os.path.basename(host_path)]
+    rel_host_path = os.path.relpath(host_path, output_directory)
+  else:
+    rel_host_path = os.path.relpath(host_path, constants.DIR_SOURCE_ROOT)
+
+  device_path_components = [None]
+  p = rel_host_path
+  while p:
+    p, d = os.path.split(p)
+    if d:
+      device_path_components.insert(1, d)
+  return device_path_components
+
+
+def GetDataDependencies(isolate_file_path):
+  """Returns a list of device data dependencies.
+
+  Args:
+    isolate_file_path: A str path to the isolate file.
+  Returns:
+    A list of (host_path, device_path) tuples.
+  """
+  if not isolate_file_path:
+    return []
+
+  with open(isolate_file_path, 'r') as isolate_file:
+    isolate_contents = ast.literal_eval(isolate_file.read())

[Dirk Pranke, 2016/11/15 22:33:05]

Where does the isolate_file_path come from? Can we change this to just read the
.runtime_deps file instead, so that there's no dependency on isolates
whatsoever?

[jbudorick, 2016/11/16 03:47:51]

Rewritten.

+
+  isolate_file_dir = os.path.dirname(isolate_file_path)
+  rel_host_files = isolate_contents.get('variables', {}).get('files', [])
+  abs_host_files = [
+      os.path.abspath(os.path.join(isolate_file_dir, r))
+      for r in rel_host_files]
+
+  return [(a, DevicePathComponentsFor(a, constants.GetOutDirectory()))
+          for a in abs_host_files]
+