Fix %g in fmt, and make %v default to %g for floats

core/fmt/fmt.odin

@@ -1232,8 +1232,12 @@ _pad :: proc(fi: ^Info, s: string) {
 //
 // NOTE: Can return "NaN", "+Inf", "-Inf", "+<value>", or "-<value>".
 //
-_fmt_float_as :: proc(fi: ^Info, v: f64, bit_size: int, verb: rune, float_fmt: byte) {
+_fmt_float_as :: proc(fi: ^Info, v: f64, bit_size: int, verb: rune, float_fmt: byte, prec: int) {
-	prec := fi.prec if fi.prec_set else 3
+	prec := prec
+	if fi.prec_set {
+		prec = fi.prec
+	}
+
 	buf: [386]byte
 
 	// Can return "NaN", "+Inf", "-Inf", "+<value>", "-<value>".
@@ -1242,7 +1246,7 @@ _fmt_float_as :: proc(fi: ^Info, v: f64, bit_size: int, verb: rune, float_fmt: b
 	if !fi.plus {
 		// Strip sign from "+<value>" but not "+Inf".
 		if str[0] == '+' && str[1] != 'I' {
-			str = str[1:] 
+			str = str[1:]
 		}
 	}
 
@@ -1258,11 +1262,13 @@ _fmt_float_as :: proc(fi: ^Info, v: f64, bit_size: int, verb: rune, float_fmt: b
 //
 fmt_float :: proc(fi: ^Info, v: f64, bit_size: int, verb: rune) {
 	switch verb {
-	case 'f', 'F', 'g', 'G', 'v':
+	case 'g', 'G', 'v':
-		_fmt_float_as(fi, v, bit_size, verb, 'f')
+		_fmt_float_as(fi, v, bit_size, verb, 'g', -1)
+	case 'f', 'F':
+		_fmt_float_as(fi, v, bit_size, verb, 'f', 3)
 	case 'e', 'E':
 		// BUG(): "%.3e" returns "3.000e+00"
-		_fmt_float_as(fi, v, bit_size, verb, 'e')
+		_fmt_float_as(fi, v, bit_size, verb, 'e', 6)
 
 	case 'h', 'H':
 		prev_fi := fi^

core/strconv/generic_float.odin

@@ -2,6 +2,8 @@ package strconv
 
 import "decimal"
 
+import "core:runtime"
+
 Decimal_Slice :: struct {
 	digits:        []byte,
 	count:         int,
@@ -103,8 +105,11 @@ generic_ftoa :: proc(buf: []byte, val: f64, fmt: byte, precision, bit_size: int)
 		}
 	} else {
 		switch fmt {
-		case 'e', 'E': decimal.round(d, prec+1)
+		case 'e', 'E':
-		case 'f', 'F': decimal.round(d, d.decimal_point+prec)
+			prec += 1
+			decimal.round(d, prec)
+		case 'f', 'F':
+			decimal.round(d, d.decimal_point+prec)
 		case 'g', 'G':
 			if prec == 0 {
 				prec = 1