Thus fixes bug 358940: fifteen puzzle often appears unsolveable.
There were two issues.
- Floating point division when determining the row with blank tile (blankRow = 2.75 seems weird).
- Even after fixing the first issue, applet didn't want to confirm that I've solved the puzzle. Seems like corresponding check was wrong: indeed, if one has solved 4x4 board (size=16, so pieces[i] = {1,2,...,15,0}), comparison with i=15 corresponds to 15>0, which is always true.
Clearly, to check if 15 numbers are ordered, one needs 14 (size-2) comparison operations.