Cleanup: Get rid of OS_WIN/OS_POSIX ifdefs from MojoApplication

base/files/platform_file.h

Index: base/files/platform_file.h
diff --git a/base/files/platform_file.h b/base/files/platform_file.h
new file mode 100644
index 0000000000000000000000000000000000000000..bb69551dc5c7bcef26a94a53b266e3f0bdf4176f
--- /dev/null
+++ b/base/files/platform_file.h
@@ -0,0 +1,30 @@
+// Copyright 2014 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.
+
+#ifndef BASE_FILES_PLATFORM_FILE_H_
+#define BASE_FILES_PLATFORM_FILE_H_
+
+#include "build/build_config.h"
+
+#if defined(OS_WIN)
+#include <windows.h>
+#endif
+
+namespace base {
+
+#if defined(OS_WIN)
+
+typedef HANDLE PlatformFile;
+const PlatformFile kInvalidPlatformFile = INVALID_HANDLE_VALUE;

[rvargas (doing something else), 2014/09/17 04:51:46]

I'm not really happy reviving PlatformFile and kInvalidPlatformFile as first
class citizens of base.

The main reason behind moving PlatformFile behind base::File was to make it
extremely obvious that File is what should be used in almost all cases and
HANDLE/fd should be avoided... to the point that code that really wants to use
raw handles has to redefine kInvalidPlatformFile (because really, there should
be an object somewhere with an IsValid method to check)

In fact I was planning removing PlatformFile from the definition of
PlatforFileForTransit and make users pass a const File& when they want to share
a file with another process. The only reason I haven't is that IPC is being
replaced by mojo ipc.

Improper sharing of handles (that ends up in double close) is still very high on
the reasons for Chrome crashing, and it is almost impossible to debug.

+
+#elif defined(OS_POSIX)
+
+typedef int PlatformFile;
+const PlatformFile kInvalidPlatformFile = -1;
+
+#endif  // defined(OS_POSIX)
+
+} // namespace base
+
+#endif // BASE_FILES_PLATFORM_FILE_H_