Is IntPtr.Zero equivalent to null?
I am trying to setup ReadFile
to run asynchronously and according to MSDN, I need to set lpNumberOfBytesRead
to null
:
"Use NULL for this parameter if this is an asynchronous operation to avoid potentially erroneous results."
For example, if I have the following:
[DllImport("kernel32.dll", SetLastError = true, CharSet = CharSet.Auto)]
public static extern bool ReadFile(
IntPtr hFile,
out byte[] aBuffer,
int cbToRead,
IntPtr cbThatWereRead,
ref OVERLAPPED pOverlapped
);
and I call it like this (with the intention of having the 4th parameter being null):
Win32API.ReadFile(readHandle, out data_read, Win32API.BUFFER_SIZE, IntPtr.Zero, ref over_lapped);
will that be the same as calling it with null? If not, what should I change in the declaration or in the function call itself?
I was also curious if I should be using SafeHandle
or HandleRef
instead of IntPtr
for the hFile
reference? I know to make sure that I close the handle with CloseHandle(IntPtr)
when I'm done with it, just not sure if there is any othe reason to use the other two options over IntPtr
. I am also tryingn to avoid using unsafe code.
EDIT: As it turns out, I shouldnt be setting the fourth parameter to IntPtr.Zero
anyway, because even though I am running asynchronously, it could still return right away. See Asynchronous Disk I/O. Ahh, I love contradicting stories.
For P/Invoke purposes like you've listed, you should use IntPtr.Zero
in place of NULL
. Note that this is not equivalent to the C# null
keyword, however.
You cannot assign null to a value-type. A reference-type can be null, as in, not referring to an object instance, but a value-type always has a value.
IntPtr.Zero is just a constant value that represents a null pointer.
Be aware that there is a bug (feature??) in C# >= 2.0, where
if (IntPtr.Zero == null)
{
// Won't enter here
}
will compile correctly, but it won't ever enter in the if
.
I opened an issue on the github of roslyn and they replied that they won't fix it because there are projects that are built with warnings-as-errors. Still there is a partial fix for this: there is a strict
compilation mode that generates this warning:
<Features>strict</Features>