When using PInvoke, I noticed that we need to use IntPtr
to refer to Windows handles. I am wondering why not just use int
for the handle? My understanding of a handle is that it is just an integer value.

- 111,048
- 26
- 262
- 307

- 20,472
- 39
- 112
- 155
-
7Your making the most common mistake that causes x64 incompatibilities. – SLaks Feb 07 '12 at 00:24
-
5Sure, they're integer values. But no one said they have to be 32 bits... – user541686 Feb 07 '12 at 00:25
4 Answers
A windows handle is defined as an integer of the native machine pointer size. That's so that they can secretly be a pointer if they need to be. (A handle probably is not a pointer, but it is permitted to be one if the operating system implementers deem it necessary. They typically are not actually pointers for security reasons; it makes it too easy for people to abuse the system if they are actually pointers.)
An int in C# defined as a 32 bit integer, which will be too small on a 64 bit machine. An IntPtr is defined as an integer that can hold a pointer of the machine size. That's why you always use IntPtr when interoperating with handles.

- 647,829
- 179
- 1,238
- 2,067
Handles are pointer-sized values.
They're 8 bytes wide on 64-bit platforms.

- 868,454
- 176
- 1,908
- 1,964
-
1I'd say "pointer-*sized*" instead of "pointer" (even though I guess in the strict C/C++ sense, that isn't too accurate, since handles are `typedef`'d as a pointer all right, and `size_t` isn't necessarily the size of a pointer)... but if they're indices into a table, chances are they're used as integers. But yeah, point taken. – user541686 Feb 07 '12 at 00:26
The size of a handle is dependant on the machine architecture (32/64 bit). IntPtr
will take care of that.

- 64,175
- 10
- 70
- 92
A handle is essentially a system level pointer, and a pointer does not implicitly cast to an int. It is a type by itself. So that's why you have IntPtr
in .NET to represent the Handle type.

- 61,704
- 67
- 242
- 415