Fixes #84: Choice of float/double

* Can now choose between using `double` and `float` internally, for floating-point work, using the `PRINTF_USE_DOUBLE_INTERNALLY` definition and corresponding CMake option.
* Adjusted test suite to support the different choice of floating-point type:
	* Tests relevant both to `float` and `double`, but with different precisions, are adjusted with an `#if #else #endif`
	* Tests and test cases relevant only for `double` precision are not compiled at all when `float` is used.
* Lots of new explicit conversions :-(

Author: eyalroz

CMakeLists.txt

@@ -20,6 +20,7 @@ option(SUPPORT_EXPONENTIAL_SPECIFIERS         "Support exponential floating poin
 option(SUPPORT_MSVC_STYLE_INTEGER_SPECIFIERS  "Support the I + bit size integer specifiers (%I8, %I16, %I32, %I64) as in Microsoft Visual C++" ON)
 option(SUPPORT_WRITEBACK_SPECIFIER            "Support the length write-back specifier (%n)" ON)
 option(SUPPORT_LONG_LONG                      "Support long long integral types (allows for the ll length modifier and affects %p)" ON)
+option(USE_DOUBLE_INTERNALLY                  "Use the C `double` type - typically 64-bit in size - for internal floating-point arithmetic " ON)
 option(CHECK_FOR_NUL_IN_FORMAT_SPECIFIER      "Be defensive in the undefined-behavior case of a format specifier not ending before the string ends" ON)
 
 set(ALIASING_MODES NONE HARD SOFT)
@@ -38,6 +39,7 @@ foreach(opt
 	SUPPORT_MSVC_STYLE_INTEGER_SPECIFIERS
 	SUPPORT_WRITEBACK_SPECIFIER
 	SUPPORT_LONG_LONG
+	USE_DOUBLE_INTERNALLY
 	ALIAS_STANDARD_FUNCTION_NAMES_SOFT
 	ALIAS_STANDARD_FUNCTION_NAMES_HARD
 	CHECK_FOR_NUL_IN_FORMAT_SPECIFIER        

README.md

@@ -82,13 +82,14 @@ Options used both in CMake and in the library source code via a preprocessor def
 | PRINTF_ALIAS_STANDARD_FUNCTION_NAMES   | NONE    |  Alias the standard library function names (`printf()`, `sprintf()` etc.) to the library's functions.<br>The possible values are `NONE`, `SOFT` and `HARD`. With Soft aliasing, the library's object files contain symbols which do not clash with the standard library's: `printf_`, `sprintd_` etc; and a macro in `printf.h` replaces usages of `printf()`, `sprintf()` etc. with the underscored versions. With Hard aliasing, no such macro is used, and the library's object files contain `printf`, `sprintf` etc. - and thus cannot be linked together with a full-fledged standard library.  **Note:** The preprocessort definitions `#define PRINTF_ALIAS_STANDARD_FUNCTION_NAMES_SOFT` and `#define PRINTF_ALIAS_STANDARD_FUNCTION_NAMES_HARD` should be defined to have the same values when using the library as when having compiled the list. |
 | PRINTF_INTEGER_BUFFER_SIZE             | 32      |  ntoa (integer) conversion buffer size. This must be big enough to hold one converted numeric number _including_ leading zeros, normally 32 is a sufficient value. Created on the stack. |
 | PRINTF_DECIMAL_BUFFER_SIZE             | 32      |  ftoa (float) conversion buffer size. This must be big enough to hold one converted float number _including_ leading zeros, normally 32 is a sufficient value. Created on the stack. |
-| PRINTF_DEFAULT_FLOAT_PRECISION         | 6       |  Define the default floating point precision|
+| PRINTF_DEFAULT_FLOAT_PRECISION         | 6       |  Define the default floating point precision digits |
 | PRINTF_MAX_INTEGRAL_DIGITS_FOR_DECIMAL | 9       |  Maximum number of integral-part digits of a floating-point value for which printing with %f uses decimal (non-exponential) notation |
 | PRINTF_SUPPORT_DECIMAL_SPECIFIERS      | YES     |  Support decimal notation floating-point conversion specifiers (%f, %F) |
 | PRINTF_SUPPORT_EXPONENTIAL_SPECIFIERS  | YES     |  Support exponential floating point format conversion specifiers (%e, %E, %g, %G) |
 | SUPPORT_MSVC_STYLE_INTEGER_SPECIFIERS  | YES     |  Support the 'I' + bit size integer specifiers (%I8, %I16, %I32, %I64) as in Microsoft Visual C++ |
 | PRINTF_SUPPORT_WRITEBACK_SPECIFIER     | YES     |  Support the length write-back specifier (%n) |
 | PRINTF_SUPPORT_LONG_LONG               | YES     |  Support long long integral types (allows for the ll length modifier and affects %p) |
+| PRINTF_USE_DOUBLE_INTERNALLY           | YES     |  Use the `double` for internal floating-point calculations (rather than using the single-precision `float` type |
 
 Within CMake, these options lack the `PRINTF_` prefix.
 

printf_config.h.in

@@ -7,6 +7,7 @@
 #define PRINTF_SUPPORT_WRITEBACK_SPECIFIER      @PRINTF_SUPPORT_WRITEBACK_SPECIFIER@
 #define PRINTF_SUPPORT_MSVC_STYLE_INTEGER_SPECIFIERS @PRINTF_SUPPORT_MSVC_STYLE_INTEGER_SPECIFIERS@
 #define PRINTF_SUPPORT_LONG_LONG                @PRINTF_SUPPORT_LONG_LONG@
+#define PRINTF_USE_DOUBLE_INTERNALLY            @PRINTF_USE_DOUBLE_INTERNALLY@
 #define PRINTF_ALIAS_STANDARD_FUNCTION_NAMES_SOFT    @PRINTF_ALIAS_STANDARD_FUNCTION_NAMES_SOFT@
 #define PRINTF_ALIAS_STANDARD_FUNCTION_NAMES_HARD    @PRINTF_ALIAS_STANDARD_FUNCTION_NAMES_HARD@