Discard high nibble when decoding Unit Exponent (#17)

HidSharp assumes the high nibble in a unit exponent field is meaningful, while the HID Usage Tables 1.3 defines the unit exponent as a 4-bit value (Exponent 0 through F, page 302 / 303). Report Descriptors only support values in multitudes of 8 bits, so the first 4 bits are padding / meaningless. This means we should ignore the high nibble when decoding the report descriptor and not make any assumptions about its meaning.

There are some devices which define the unit exponent as 0xF0 through 0xFF. Reconstructing the report descriptor fails on these devices.

This is how the Linux kernel does it as well: https://github.com/torvalds/linux/blob/master/drivers/hid/hid-core.c#L427

Author: Joop-Schilder

HidSharp/Reports/Units/Unit.cs

@@ -63,12 +63,14 @@ public int GetExponent(UnitKind kind)
 
         /// <summary>
         /// Decodes an encoded HID unit exponent.
+        /// The HID Usage Table 1.3 states that valid values for the unit 
+        /// exponent field are 0x_0 - 0x_F so we ignore the high nibble.
         /// </summary>
         /// <param name="value">The encoded exponent.</param>
         /// <returns>The exponent.</returns>
         public static int DecodeExponent(uint value)
         {
-            if (value > 15) { throw new ArgumentOutOfRangeException("value", "Value range is [0, 15]."); }
+            value &= 0x0f;
             return value >= 8 ? (int)value - 16 : (int)value;
         }