Change FP default precison from 53 to 64-bit mantissa.

	* Makefile.in (CRT0S): Add CRT_fp8.o.
	(MINGW_OBJS): Replace CRT_fp8.o with CRT_fp10.o.
	* include/float.h: Replace standard float.h defines with
	#include_next<float.h> to use GCC's defines.

3 files changed, 27 insertions, 81 deletions

diff --git a/winsup/mingw/ChangeLog b/winsup/mingw/ChangeLog
index 3e8410e09..017c8c742 100644
--- a/winsup/mingw/ChangeLog
+++ b/winsup/mingw/ChangeLog
@@ -1,3 +1,12 @@
+2002-04-29  Danny Smith  <dannysmith@users.sourceforge.net>
+
+	Change FP default precison from 53 to 64-bit mantissa.
+	* Makefile.in (CRT0S): Add CRT_fp8.o.
+	(MINGW_OBJS): Replace CRT_fp8.o with CRT_fp10.o.
+	* include/float.h: Replace standard float.h defines with
+	#include_next<float.h> to use GCC's defines. Adjust comments
+	to reflect change.	
+
 2002-04-26  Danny Smith  <dannysmith@users.sourceforge.net>
 
 	* include/dos.h: Change  prefix "__imp_"  to "_imp__" for
diff --git a/winsup/mingw/Makefile.in b/winsup/mingw/Makefile.in
index 6c81ba8f7..a4fe3e282 100644
--- a/winsup/mingw/Makefile.in
+++ b/winsup/mingw/Makefile.in
@@ -147,9 +147,9 @@ FLAGS_TO_PASS:=\
 	TARFILEEXT="$(TARFILEEXT)"
 
 CRT0S = crt1.o dllcrt1.o crt2.o dllcrt2.o CRT_noglob.o crtmt.o crtst.o \
-	CRT_fp10.o
+	CRT_fp8.o CRT_fp10.o
 MINGW_OBJS = CRTglob.o CRTfmode.o CRTinit.o  dllmain.o gccmain.o \
-	     main.o crtst.o mthr_stub.o CRT_fp8.o
+	     main.o crtst.o mthr_stub.o CRT_fp10.o
 MOLD_OBJS = ctype_old.o string_old.o
 
 LIBS = libcrtdll.a libmsvcrt.a libmsvcrt20.a libmsvcrt40.a libmingw32.a \
diff --git a/winsup/mingw/include/float.h b/winsup/mingw/include/float.h
index f24ecb459..dbd044155 100644
--- a/winsup/mingw/include/float.h
+++ b/winsup/mingw/include/float.h
@@ -6,11 +6,10 @@
  * Also included here are some non-ANSI bits for accessing the floating
  * point controller.
  *
- * NOTE: GCC provides float.h, and it is probably more accurate than this,
- *       but it doesn't include the non-standard stuff for accessing the
- *       fp controller. (TODO: Move those bits elsewhere?) Thus it is
- *       probably not a good idea to use the GCC supplied version instead
- *       of this header.
+ * NOTE: GCC provides float.h, but it doesn't include the non-standard
+ *       stuff for accessing the fp controller. We include_next the
+ *       GCC-supplied header and just define the MS-specific extensions
+ *       here.     
  *
  * This file is part of the Mingw32 package.
  *
@@ -33,79 +32,13 @@
  *
  */
 
-#ifndef _FLOAT_H_
-#define _FLOAT_H_
+#ifndef _MINGW_FLOAT_H_
+#define _MINGW_FLOAT_H_
 
 /* All the headers include this file. */
 #include <_mingw.h>
 
-#define FLT_ROUNDS	1
-#define FLT_GUARD	1
-#define FLT_NORMALIZE	1
-
-/*
- * The characteristics of float.
- */
-
-/* The radix for floating point representation. */
-#define FLT_RADIX	2
-
-/* Decimal digits of precision. */
-#define FLT_DIG		6
-
-/* Smallest number such that 1+x != 1 */
-#define FLT_EPSILON	1.19209290e-07F
-
-/* The number of base FLT_RADIX digits in the mantissa. */
-#define FLT_MANT_DIG	24
-
-/* The maximum floating point number. */
-#define FLT_MAX		3.40282347e+38F
-
-/* Maximum n such that FLT_RADIX^n - 1 is representable. */
-#define FLT_MAX_EXP	128
-
-/* Maximum n such that 10^n is representable. */
-#define FLT_MAX_10_EXP	38
-
-/* Minimum normalized floating-point number. */
-#define FLT_MIN		1.17549435e-38F
-
-/* Minimum n such that FLT_RADIX^n is a normalized number. */
-#define FLT_MIN_EXP	(-125)
-
-/* Minimum n such that 10^n is a normalized number. */
-#define FLT_MIN_10_EXP	(-37)
-
-
-/*
- * The characteristics of double.
- */
-#define DBL_DIG		15
-#define DBL_EPSILON	1.1102230246251568e-16
-#define DBL_MANT_DIG	53
-#define DBL_MAX		1.7976931348623157e+308
-#define DBL_MAX_EXP	1024
-#define DBL_MAX_10_EXP	308
-#define DBL_MIN		2.2250738585072014e-308
-#define DBL_MIN_EXP	(-1021)
-#define DBL_MIN_10_EXP	(-307)
-
-
-/*
- * The characteristics of long double.
- * NOTE: long double is the same as double.
- */
-#define LDBL_DIG	15
-#define LDBL_EPSILON	1.1102230246251568e-16L
-#define LDBL_MANT_DIG	53
-#define LDBL_MAX	1.7976931348623157e+308L
-#define LDBL_MAX_EXP	1024
-#define LDBL_MAX_10_EXP	308
-#define LDBL_MIN	2.2250738585072014e-308L
-#define LDBL_MIN_EXP	(-1021)
-#define LDBL_MIN_10_EXP	(-307)
-
+#include_next<float.h>
 
 /*
  * Functions and definitions for controlling the FPU.
@@ -193,11 +126,14 @@ unsigned int	_statusfp (void);	/* Report the FPU status word */
 
 
 /*
-   _fpreset initializes the control register to 0x27f,
+   MSVCRT.dll _fpreset initializes the control register to 0x27f,
    the status register to zero and the tag word to 0FFFFh.
-   This differs from asm instruction fninit which sets control
-   word to 0x37f (64 bit mantissa precison rather than 53 bit)
-*/  
+   This differs from asm instruction finit/fninit which set control
+   word to 0x37f (64 bit mantissa precison rather than 53 bit).
+   By default, the mingw version of _fpreset sets fp control as
+   per fninit. To use the MSVCRT.dll _fpreset, include CRT_fp8.o when
+   building your application.	 
+*/
 void		_fpreset (void);
 void		fpreset (void);
 
@@ -206,7 +142,8 @@ int *	__fpecode(void);
 #define	_fpecode	(*(__fpecode()))
 
 /*
- * IEEE recommended functions
+ * IEEE recommended functions.  MS puts them in float.h
+ * but they really belong in math.h.
  */
 
 double	_chgsign	(double);