How can I tell if cbSize should be int or uint in MonitorInfo?

Question

I'm trying to do some multi-monitor stuff using Platform Invoke. I've been using http://pinvoke.net to get me started but I've hit a problem where definitions don't agree.

In MONITORINFO (user32) (and MONITORINFOEX (user32)), size is defined as:

public int Size;

But in EnumDisplayMonitors (user32), in the sample code which uses MonitorInfo, we see:

mi.size = (uint)Marshal.SizeOf(mi);

Clearly, one of these is less than correct.

In the MSDN docs, MONITORINFO is declared as:

typedef struct tagMONITORINFO {
  DWORD cbSize;
  RECT  rcMonitor;
  RECT  rcWork;
  DWORD dwFlags;
} MONITORINFO, *LPMONITORINFO;

with cbSize defined as:

The size of the structure, in bytes.

Set this member to sizeof ( MONITORINFO ) before calling the GetMonitorInfo function. Doing so lets the function determine the type of structure you are passing to it.

Any idea how I can work out which it should be, int or uint?

Note: I know some of this stuff is available in System.Windows.Forms but I'm trying to do this using P/Invoke in Silverlight 5.

score 1 · Accepted Answer · answered Mar 13 '12 at 10:54

DWORD is an unsigned 32 bit integer, so uint is technically correct.

The version that uses int was presumably done that way through mild laze. The author presumably wanted not to have to cast the return value of Marshal.SizeOf() to uint.

In practice it simply does not matter here because the cbSize will never get to the point where the conversion from signed to unsigned makes any difference. However, in some cases it can matter. For example, a variable that held a combination of bit flags.

Hans Passant · Answer 2 · 2012-03-13T11:09:15.630

The native type is DWORD, an alias for unsigned int in the Windows api. Which makes sense, the size of a struct can never be negative. But a pinvoke declaration for such a struct often uses int instead of uint. It does in the declaration provided by pinvoke.net.

This is often for convenience, Marshal.SizeOf() returns an int, not an uint. This was done to make the method CLSCompliant, there are many languages that don't support unsigned types. Java and the original Visual Basic being good examples.

This is not a problem, an int and an uint have the same size and the same bit pattern for values between 0 and int.MaxValue. Going beyond int.MaxValue so that uint would matter is not a practical problem. Structs are never that large. And if you had one that was, it would be unusable anyway, the CLR does not support managed types larger than 2GB. At least up to .NET 4.5.

So, getting around to answering the question, the code snippet was written by somebody that declared the structure member as uint, thus requiring the cast. Either way is fine.

How can I tell if cbSize should be int or uint in MonitorInfo?

2 Answers2