Update double-conversion to 7630f84a10f9428b041d0471e71a562141e9684b.

Change-Id: I37d176a938798cb74eae0a3a445182beb55a117d
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/434900
Commit-Queue: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Alexander Aprelev <aam@google.com>

PRESUBMIT.py

@@ -340,7 +340,8 @@ def _CheckClangFormat(input_api, output_api):
                     f.ChangedContents())):
             is_deps = True
             break
-        if is_cpp_file(path) and os.path.isfile(path):
+        if is_cpp_file(path) and os.path.isfile(
+                path) and not path.startswith('third_party/'):
             files.append(path)
 
     if is_deps:

third_party/double-conversion/README.dart

@@ -1,5 +1,5 @@
 URL: https://github.com/google/double-conversion
-Version: 032fa6a7d2c319b20d3928f5d762648fa4029acf
+Version: 7630f84a10f9428b041d0471e71a562141e9684b
 License: BSD
 License File: LICENSE
 

third_party/double-conversion/README.md

@@ -19,7 +19,7 @@ There is extensive documentation in `double-conversion/string-to-double.h` and
 Building
 ========
 
-This library can be built with [scons][0] or [cmake][1].
+This library can be built with [scons][0], [cmake][1] or [bazel][2].
 The checked-in Makefile simply forwards to scons, and provides a
 shortcut to run all tests:
 
@@ -55,5 +55,23 @@ Use `-DBUILD_TESTING=ON` to build the test executable.
     make
     test/cctest/cctest
 
+Bazel
+---
+
+The simplest way to adopt this library is through the [Bazel Central Registry](https://registry.bazel.build/modules/double-conversion).
+
+To build the library from the latest repository, run:
+
+```
+bazel build //:double-conversion
+```
+
+To run the unit test, run:
+
+```
+bazel test //:cctest
+```
+
 [0]: http://www.scons.org/
 [1]: https://cmake.org/
+[2]: https://bazel.build/

third_party/double-conversion/src/double-to-string.cc

@@ -180,7 +180,7 @@ bool DoubleToStringConverter::ToShortestIeeeNumber(
     return HandleSpecialValues(value, result_builder);
   }
 
-  int decimal_point;
+  int decimal_point = 0;
   bool sign;
   const int kDecimalRepCapacity = kBase10MaximalLength + 1;
   char decimal_rep[kDecimalRepCapacity];
@@ -405,6 +405,7 @@ void DoubleToStringConverter::DoubleToAscii(double v,
   if (mode == PRECISION && requested_digits == 0) {
     vector[0] = '\0';
     *length = 0;
+    *point = 0;
     return;
   }
 

third_party/double-conversion/src/double-to-string.h

@@ -237,11 +237,24 @@ class DoubleToStringConverter {
   //   kBase10MaximalLength significant digits).
   //      "-1.7976931348623157e+308", "-1.7976931348623157E308"
   // In addition, the buffer must be able to hold the trailing '\0' character.
+  //
+  // Since the algorithm finds the shortest number of significant digits, it
+  // can produce an output that isn't the shortest possible if the
+  // decimal_in_shortest_high is high enough. For example, the number
+  // 1e23 could be written as 99999999999999991611392 with 23
+  // digits, however, it only needs one significant digit 1, and thus the
+  // result is 100000000000000000000000, which has 24 digits.
   bool ToShortest(double value, StringBuilder* result_builder) const {
     return ToShortestIeeeNumber(value, result_builder, SHORTEST);
   }
 
   // Same as ToShortest, but for single-precision floats.
+  //
+  // Since the algorithm finds the shortest number of significant digits, it
+  // can, in very rare cases, produce an output that isn't the shortest possible.
+  // For example, the number 1e11f could be written as 99999997952 with 11
+  // digits, however, it only needs one significant digit 1, and thus the
+  // result is 100000000000, which has 12 digits.
   bool ToShortestSingle(float value, StringBuilder* result_builder) const {
     return ToShortestIeeeNumber(value, result_builder, SHORTEST_SINGLE);
   }