cue: add support for zero values in Value.Float64

Fix the error when attempting to use strconv.FormatFloat with a zero value:

	cannot use 0.0 (type float) as float64 in argument 0 to strconv.FormatFloat: value was rounded down

The tests for maximum and minimum in ValueFloat64 do not allow for zero.
The below check incorrectly sees zero as smaller, because the Sign of
zero (0) is less than the Sign of smallestPosFloat64 (1). So, it was
clear this logic was not accounting for zero.

		if n.X.Cmp(smallestPosFloat64) == -1 {

Fixes #1669.
Closes #1670 as merged as of commit 51fb5b8.

Signed-off-by: Kevin Burge <kcburge@pm.me>
Change-Id: I16a9fde39c36c519501c6cb3f87470a7f0cbab36
Reviewed-on: https://review.gerrithub.io/c/cue-lang/cue/+/552142
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
Unity-Result: CUEcueckoo <cueckoo@cuelang.org>
TryBot-Result: CUEcueckoo <cueckoo@cuelang.org>
Reviewed-by: Roger Peppe <rogpeppe@gmail.com>

cue/types.go

@@ -528,6 +528,9 @@ func (v Value) Float64() (float64, error) {
 	if err != nil {
 		return 0, err
 	}
+	if n.X.IsZero() {
+		return 0.0, nil
+	}
 	if n.X.Negative {
 		if n.X.Cmp(smallestNegFloat64) == 1 {
 			return -0, ErrAbove

cue/types_test.go

@@ -459,6 +459,14 @@ func TestFloat(t *testing.T) {
 		float64: -1,
 		fmt:     'g',
 		kind:    IntKind,
+	}, {
+		value:   "0.0",
+		float:   "0.0",
+		mant:    "0",
+		exp:     -1,
+		float64: 0.0,
+		fmt:     'g',
+		kind:    FloatKind,
 	}, {
 		value:   "1.0",
 		float:   "1.0",

pkg/strconv/testdata/gen.txtar

@@ -12,6 +12,7 @@ t6: strconv.FormatInt(170_141_183_460_469_231_731_687_303_715_884_105_728, 10)
 t7: strconv.FormatInt(64, 16)
 t8: strconv.FormatUint(170_141_183_460_469_231_731_687_303_715_884_105_728, 10)
 t9: strconv.FormatUint(61, 62)
+t10: strconv.FormatFloat(0.0, 102, -1, 64)
 -- out/strconv --
 Errors:
 t2: int 300 overflows byte in argument 1 in call to strconv.FormatFloat:
@@ -23,13 +24,14 @@ t4: cannot use 1.0 (type float) as int in argument 2 to strconv.FormatFloat:
     ./in.cue:6:31
 
 Result:
-t1: "40"
-t2: _|_ // t2: int 300 overflows byte in argument 1 in call to strconv.FormatFloat
-t3: _|_ // t3: cannot use -1 (type int) as byte in argument 1 to strconv.FormatFloat
-t4: _|_ // t4: cannot use 1.0 (type float) as int in argument 2 to strconv.FormatFloat
-t5: "true"
-t6: "170141183460469231731687303715884105728"
-t7: "40"
-t8: "170141183460469231731687303715884105728"
-t9: "Z"
+t1:  "40"
+t2:  _|_ // t2: int 300 overflows byte in argument 1 in call to strconv.FormatFloat
+t3:  _|_ // t3: cannot use -1 (type int) as byte in argument 1 to strconv.FormatFloat
+t4:  _|_ // t4: cannot use 1.0 (type float) as int in argument 2 to strconv.FormatFloat
+t5:  "true"
+t6:  "170141183460469231731687303715884105728"
+t7:  "40"
+t8:  "170141183460469231731687303715884105728"
+t9:  "Z"
+t10: "0"