+/*
+Negative numbers still screw it up... Need to think about what we're
+trying to do here. The fact that it worked with 72 seems to have been pure luck.
+It seems the problem is negative numbers: We can't let that happen.
+When taking away the zero, it pops over 1 px at zero, then goes about 1/2 a
+grid at x<0.
+
+The bitmap looks like this:
+
++---+---+---+---+---
+| | | | |
+| | | | |
++---+---+---+---+---
+| | | | |
+| | | | |
+| | | | |
+
+@ x = 1, we want it to look like:
+
+-+---+---+---+---+---
+ | | | | |
+ | | | | |
+-+---+---+---+---+---
+ | | | | |
+ | | | | |
+ | | | | |
+
+Which means we need to grab the sample from x = 3. @ x = -1:
+
+---+---+---+---+---
+ | | | |
+ | | | |
+---+---+---+---+---
+ | | | |
+ | | | |
+ | | | |
+
+Which means we need to grab the sample from x = 1. Which means we have to take
+the mirror of the modulus of gridPixels.
+
+Doing a mod of a negative number is problematic: 1st, the compiler converts the
+negative number to an unsigned int, then it does the mod. Gets you wrong answers
+most of the time, unless you use a power of 2. :-P So what we do here is just
+take the modulus of the negation, which means we don't have to worry about
+mirroring it later.
+
+The positive case looks gruesome (and it is) but it boils down to this: We take
+the modulus of the X coordinate, then mirror it by subtraction from the
+maximum (in this case, gridPixels). This gives us a number in the range of 1 to
+gridPixels. But we need the case where the result equalling gridPixels to be
+zero; so we do another modulus operation on the result to achieve this.
+*/
+ if (x < 0)
+ x = -x % gridPixels;
+ else
+ x = (gridPixels - (x % gridPixels)) % gridPixels;
+
+ if (y < 0)
+ y = -y % gridPixels;
+ else
+ y = (gridPixels - (y % gridPixels)) % gridPixels;