PASS 1 : %.12f of 1/3: twelve real places (was 0.333333333000)
PASS 2 : %.16f %.16r %.12F of sqrt(2): every place it holds
PASS 3 : %.2f past 1e16 stays positional and exact (was 1.23e16)
PASS 4 : %.2f of a WIDE value: all 24 integer digits
PASS 5 : %.2f of 2.675 rounds the decimal half-even, as %.2m (was 2.67 from binary)
PASS 6 : ties on the exact decimal: 9.95 -> 10.0, 1.05 -> 1.0, 9.995 -> 10.00
PASS 7 : %e of 0 (was 0.000000e-308)
PASS 8 : %E: a positive exponent has its + (was 1.234568E04)
PASS 9 : %e of 10^100 (was 10.000000e99)
PASS 10 : %.12e of 1/3: twelve mantissa places
PASS 11 : %g of 1e6 and 1e5 (was 10.000000e05 and 100000.0)
PASS 12 : %.3e of a 64-digit SCI value (passed before too)
PASS 13 : %#g keeps the zeros (passed before too)
PASS 14 : %g trims its zeros and chooses by the exponent
PASS 15 : %.3g of 99950 rounds to 1e+05; %.12g of 1/3
PASS 16 : %#.0f keeps the point (# was ignored)
PASS 17 : tiny values: the places past 16 hold their digits
PASS 18 : %f of a value str$ writes in e-notation is %e; %.20e of a WIDE value
PASS 19 : flags 0 - + unchanged
PASS 20 : %.2F and %8.3R of an INTEGER (was 0.00 and 0.000)
PASS 21 : %e / %g of an INTEGER: signed exponents, %g at six digits
PASS 22 : -0.00, a width-only %f (six places) and the bare %f unchanged
PASS 23 : %e of a small exact value (passed before too)
PASS 24 : a precision over 512 raises (it printed garbage)
PASS 25 : a * that is not right after the % raises
PASS 26 : %*.2f of a REAL: the width from the argument (the old build CRASHED here)
passed: 26   failed: 0 
