The standard formatter outputs the incorrect precision for zero and negative zero due to the specialStr function not taking the specified precision into account and always outputting 0.0 and -0.0.
I have added tests in #638 that find this bug and #636 fixes this bug.
