Convert utility process extension Unpacker IPC to mojo

extensions/common/extension_unpacker.mojom

Index: extensions/common/extension_unpacker.mojom
diff --git a/extensions/common/extension_unpacker.mojom b/extensions/common/extension_unpacker.mojom
new file mode 100644
index 0000000000000000000000000000000000000000..2ccf08240071f7875f42a6b782988dd94d2c2f38
--- /dev/null
+++ b/extensions/common/extension_unpacker.mojom
@@ -0,0 +1,29 @@
+// Copyright 2017 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.
+
+// Secure chrome extension unpacker service provided by the utility process
+// and exposed by mojo policy to the chrome browser process.
+
+module extensions.mojom;
+
+import "mojo/common/file_path.mojom";
+import "mojo/common/string16.mojom";
+import "mojo/common/values.mojom";
+
+interface ExtensionUnpacker {
+  // Unzip |file| into the directory |path|.
+  Unzip(mojo.common.mojom.FilePath file,
+        mojo.common.mojom.FilePath path) => (bool success);
+
+  // Unpack, validate, and sanitize the extension in directory |path|, and
+  // return the parsed manifest.json file in |manifest|.
+  // If unpacking fails, |error| is a user-displayable explanation of what
+  // went wrong.
+  Unpack(mojo.common.mojom.FilePath path,
+         string extension_id,
+         int32 location,

[Devlin, 2017/02/14 17:24:59]

Mojo supports enums - can we do that instead?

[Noel Gordon, 2017/02/15 17:28:09]

Yeah I cargo-culted the existing API here, there were CHECK's in the existing
IPC handler to test if location is sain.  I retained those.

"Can we do that instead?"  yeah we could, but all clients of SandboxedUnpacker
would need to be changed to use the mojo enum defines (unless you want to
maintain two enums).

My feeling is there are lots of clients, and updating them might be a patch in
itself.  Maybe a job for crbug.com/692120 ?

[Devlin, 2017/02/17 15:53:56]

My understanding from rockot@ is that mojo can reuse the existing
IPC_ENUM_TRAITS declarations, which would mean that we wouldn't need to update
callers.  Is that not the case?

[Noel Gordon, 2017/02/27 11:47:50]

Yes, but that would also restrict the code to C++ callers only.  This CL allows
any caller, be they C++, Java, JavaScript etc.

[Devlin, 2017/02/27 19:30:15]

I'd rather we optimize for the C++ case, since we don't currently have any plans
to call this from any other language. :)

+         int32 creation_flags)
+      => (mojo.common.mojom.String16 error,
+          mojo.common.mojom.DictionaryValue manifest);
+};