Update double-conversion and reapply NO_LINTs.

Fixes a -Werror=strict-overflow in gcc 5.2.1.

google/double-conversion@56a0457

BUG=http://dartbug.com/25630
[email protected]

Review URL: https://codereview.chromium.org/1658343002 .

Author: rmacnak-google

runtime/third_party/double-conversion/src/bignum.cc

@@ -104,7 +104,7 @@ void Bignum::AssignDecimalString(Vector<const char> value) {
   const int kMaxUint64DecimalDigits = 19;
   Zero();
   int length = value.length();
-  int pos = 0;
+  unsigned int pos = 0;
   // Let's just say that each digit needs 4 bits.
   while (length >= kMaxUint64DecimalDigits) {
     uint64_t digits = ReadUInt64(value, pos, kMaxUint64DecimalDigits);

runtime/third_party/double-conversion/src/bignum.h

@@ -49,7 +49,6 @@ class Bignum {
 
   void AssignPowerUInt16(uint16_t base, int exponent);
 
-  void AddUInt16(uint16_t operand);
   void AddUInt64(uint64_t operand);
   void AddBignum(const Bignum& other);
   // Precondition: this >= other.

runtime/third_party/double-conversion/src/cached-powers.cc

@@ -131,7 +131,6 @@ static const CachedPower kCachedPowers[] = {
   {UINT64_2PART_C(0xaf87023b, 9bf0ee6b), 1066, 340},
 };
 
-static const int kCachedPowersLength = ARRAY_SIZE(kCachedPowers);
 static const int kCachedPowersOffset = 348;  // -1 * the first decimal_exponent.
 static const double kD_1_LOG2_10 = 0.30102999566398114;  //  1 / lg(10)
 // Difference between the decimal exponents in the table above.
@@ -149,7 +148,7 @@ void PowersOfTenCache::GetCachedPowerForBinaryExponentRange(
   int foo = kCachedPowersOffset;
   int index =
       (foo + static_cast<int>(k) - 1) / kDecimalExponentDistance + 1;
-  ASSERT(0 <= index && index < kCachedPowersLength);
+  ASSERT(0 <= index && index < static_cast<int>(ARRAY_SIZE(kCachedPowers)));
   CachedPower cached_power = kCachedPowers[index];
   ASSERT(min_exponent <= cached_power.binary_exponent);
   (void) max_exponent;  // Mark variable as used.

runtime/third_party/double-conversion/src/diy-fp.h

@@ -42,7 +42,7 @@ class DiyFp {
   static const int kSignificandSize = 64;
 
   DiyFp() : f_(0), e_(0) {}
-  DiyFp(uint64_t f, int e) : f_(f), e_(e) {}
+  DiyFp(uint64_t significand, int exponent) : f_(significand), e_(exponent) {}
 
   // this = this - other.
   // The exponents of both numbers must be the same and the significand of this
@@ -76,22 +76,22 @@ class DiyFp {
 
   void Normalize() {
     ASSERT(f_ != 0);
-    uint64_t f = f_;
-    int e = e_;
+    uint64_t significand = f_;
+    int exponent = e_;
 
     // This method is mainly called for normalizing boundaries. In general
     // boundaries need to be shifted by 10 bits. We thus optimize for this case.
     const uint64_t k10MSBits = UINT64_2PART_C(0xFFC00000, 00000000);
-    while ((f & k10MSBits) == 0) {
-      f <<= 10;
-      e -= 10;
+    while ((significand & k10MSBits) == 0) {
+      significand <<= 10;
+      exponent -= 10;
     }
-    while ((f & kUint64MSB) == 0) {
-      f <<= 1;
-      e--;
+    while ((significand & kUint64MSB) == 0) {
+      significand <<= 1;
+      exponent--;
     }
-    f_ = f;
-    e_ = e;
+    f_ = significand;
+    e_ = exponent;
   }
 
   static DiyFp Normalize(const DiyFp& a) {