Fixing remaining VC++ 2015 64-bit build breaks

content/common/pepper_file_util.cc

Index: content/common/pepper_file_util.cc
diff --git a/content/common/pepper_file_util.cc b/content/common/pepper_file_util.cc
index 6264faa5374422e65609de14913359318e0b66fa..e9e806d48da00e3e2266b954efbc435789301368 100644
--- a/content/common/pepper_file_util.cc
+++ b/content/common/pepper_file_util.cc
@@ -23,7 +23,9 @@ storage::FileSystemType PepperFileSystemTypeToFileSystemType(
 int IntegerFromSyncSocketHandle(
     const base::SyncSocket::Handle& socket_handle) {
 #if defined(OS_WIN)
-  return reinterpret_cast<int>(socket_handle);
+  // Cast through intptr_t and then int to make the truncation explicit.
+  // Handles are size-of-pointer but are always 32-bit values.
+  return static_cast<int>(reinterpret_cast<intptr_t>(socket_handle));

[Will Harris, 2015/11/10 18:05:14]

Can HANDLE values go above INT_MAX (signed int vs unsigned int)? Shouldn't this
function return unsigned int?

[brucedawson, 2015/11/10 18:43:26]

INVALID_HANDLE_VALUE (0xFFFFFFFF) is probably the only signed handle value,
although I don't know that for sure. MSDN says that sign extension should be
used when converting handles from 32-bit to 64-bit, but if comparing two 32-bit
handles it doesn't matter if they are signed or unsigned, so I'm inclined to
leave this alone. Although, having a truncation macro will force standardizing
which adds some temporary complication.

[Will Harris, 2015/11/10 19:45:45]

seems it's hard to change pepper api to have this return unsigned, sounds better
to leave this alone.

while looking at this code, I did find PlatformFileToInt in platform_file.h
which I think does what your helper function will do.

[brucedawson, 2015/11/10 19:54:33]

Good find. I'll use this in this context, and my other function elsewhere. Using
the base function elsewhere makes sense because the result sign is different and
because using ppapi code elsewhere would be bad.

 #elif defined(OS_POSIX)
   return socket_handle;
 #else