Question

I have a byte array which I would like to access by Int32 pointer (unsafe context). I am doing this

byte[] bgImageBytes = new byte[1000];
unsafe
{
   fixed (byte* bgImgPtr = bgImageBytes)
   {
      // I have a byte pointer ... How can I get an Int32 pointer?
   }
}

I’m already accessing a pointer returned from kernel32.dll as both Byte and Int32 pointer without any problem. But when I try to make an Int32 pointer on the managed byte array (example above) it seems to complain about it being managed code so it won’t work.

Simply doing UInt32* bgImgIntPtr = (UInt32*)bgImgPtr; results in MDA FatalExecutionEngineError: The CLR has been fatally corrupted. This is most often caused by data corruption, which can be caused by a number of problems, such as calls to malformed platform invoke functions and passing invalid data to the CLR.

My goal: Have both UInt32 and Byte pointers to a single bytearray so I can read the Kinect “heatmap” both as integer and as individual colors. I know I can easily convert between the types, but since I’m working with multiple arrays in different formats it would be much better if I could access them directly without converting between them all the time. There is a lot of plain copying going on so it will just add overhead to keep converting.

Answer 1

Ok, funny story. Turns out it is not only possible to reference an null array, but it also points to somewhere. This really messed up my debugging.

The “UInt32* bgImgIntPtr = (UInt32*)bgImgPtr;” that leads to the MDA exception is because the array was uninitialized. Making a pointer to the bytepointer that goes to the bytearray is the correct way to go.

The answer:

byte[] bgImageBytes = new byte[1000];
unsafe
{   
   // Make a byte pointer to the byte array
   fixed (byte* bgImgPtr = bgImageBytes)   {
      // Make a UInt32 pointer to the byte pointer
      UInt32* bgImgIntPtr = (UInt32*)bgImgPtr;
   }
}