2018-09-09:
  Fix bug where large hex literals would lose their minus sign.
  Added support for separator characters (which adds a new optional
  argument). Thus increasing the version number to 3.1.0
  Added support for hexadecimal float literals.
  Support for more architectures.

2017-12-06:
  Renamed `DISALLOW_COPY_AND_ASSIGN` and `DISALLOW_IMPLICIT_CONSTRUCTORS`
  macros to `DC_DISALLOW_COPY_AND_ASSIGN` and
  `DC_DISALLOW_IMPLICIT_CONSTRUCTORS` to make it easier to integrate the
  library with other libraries that have similar macros.

2017-08-05:
  Tagged v3.0.0.
  Due to the directory rename switching to a new version number.
  The API for the library itself hasn't changed.

2017-03-04:
  Avoid negative shift. Fixes #41.

2016-11-17:
  Support RISC-V.


2016-09-10:
  Add fPIC flag on x86_64 if the compiler supports it. Fixes #34.

2015 and 2016:
  Lots of improvements to the build system.

2015:
  Warning fixes.

2015-05-19:
  Rename 'src' directory to 'double-conversion'.

2014-03-08:
  Update version number for cmake.
  Support shared libraries with cmake.
  Add build instructions to the README.

2014-01-12:
  Tagged v2.0.1.
  Fix compilation for ARMv8 64bit (used wrong define).
  Improved SConstruct file. Thanks to Milan Bouchet-Valat and Elan Ruusamäe.
  Fixed lots of warnings (especially on Windows). Thanks to Greg Smith.

2013-11-09:
  Tagged v2.0.0.
  String-to-Double|Float: ALLOW_LEADING_SPACES and similar flags now include
  new-lines, tabs and all Unicode whitespace characters.

2013-11-09:
  Tagged v1.1.2.
  Add support for ARM 64 and OsX ppc.
  Rewrite tests so they pass under Visual Studio.
  Add CMake build system support.
  Fix warnings.

2012-06-10:
  Tagged v1.1.1.
  Null terminate exponent buffer (only an issue when asserts are enabled).
  Support more architectures.

2012-02-05:
  Merged in Single-branch with single-precision support.
  Tagged v1.1 (based on b28450f33e1db493948a535d8f84e88fa211bd10).

2012-02-05:
  Tagged v1.0 (based on eda0196e9ac8fcdf59e92cb62885ee0af5391969).