Jörg Höhle : msvcrt: sizeof(long double) is 12 with MS, but not necessarily so on UNIX.

[Alexandre Julliard]

Module: wine
Branch: stable
Commit: 0dca7f478f1ca3d4f8dc82b5c0aed3668f740264
URL:    http://source.winehq.org/git/wine.git/?a=commit;h=0dca7f478f1ca3d4f8dc82b5c0...

Author: Jörg Höhle <hoehle(a)users.sourceforge.net>
Date:   Wed Dec  1 22:26:26 2010 +0100

msvcrt: sizeof(long double) is 12 with MS, but not necessarily so on UNIX.
(cherry picked from commit c6d67fad8f6d0ac199d0945314564e39aa64644f)

---

 dlls/msvcrt/msvcrt.h |    2 +-
 dlls/msvcrt/string.c |   20 +++++++++++++++-----
 2 files changed, 16 insertions(+), 6 deletions(-)

diff --git a/dlls/msvcrt/msvcrt.h b/dlls/msvcrt/msvcrt.h
index f68790b..9064fd0 100644
--- a/dlls/msvcrt/msvcrt.h
+++ b/dlls/msvcrt/msvcrt.h
@@ -80,7 +80,7 @@ typedef unsigned int (__stdcall *MSVCRT__beginthreadex_start_routine_t)(void *);
 typedef int (*__cdecl MSVCRT__onexit_t)(void);
 typedef void (__cdecl *MSVCRT_invalid_parameter_handler)(const MSVCRT_wchar_t*, const MSVCRT_wchar_t*, const MSVCRT_wchar_t*, unsigned, MSVCRT_uintptr_t);
 
-typedef struct {long double x;} MSVCRT__LDOUBLE;
+typedef struct {ULONG x80[3];} MSVCRT__LDOUBLE; /* Intel 80 bit FP format has sizeof() 12 */
 
 struct MSVCRT_tm {
     int tm_sec;
diff --git a/dlls/msvcrt/string.c b/dlls/msvcrt/string.c
index 8c99399..7086e12 100644
--- a/dlls/msvcrt/string.c
+++ b/dlls/msvcrt/string.c
@@ -381,8 +381,10 @@ int CDECL MSVCRT__atoldbl(MSVCRT__LDOUBLE *value, const char *str)
 {
   /* FIXME needs error checking for huge/small values */
 #ifdef HAVE_STRTOLD
+  long double ld;
   TRACE("str %s value %p\n",str,value);
-  value->x = strtold(str,0);
+  ld = strtold(str,0);
+  memcpy(value, &ld, 10);
 #else
   FIXME("stub, str %s value %p\n",str,value);
 #endif
@@ -395,8 +397,10 @@ int CDECL MSVCRT__atoldbl(MSVCRT__LDOUBLE *value, const char *str)
 int CDECL __STRINGTOLD( MSVCRT__LDOUBLE *value, char **endptr, const char *str, int flags )
 {
 #ifdef HAVE_STRTOLD
+    long double ld;
     FIXME("%p %p %s %x partial stub\n", value, endptr, str, flags );
-    value->x = strtold(str,endptr);
+    ld = strtold(str,0);
+    memcpy(value, &ld, 10);
 #else
     FIXME("%p %p %s %x stub\n", value, endptr, str, flags );
 #endif
@@ -682,7 +686,7 @@ struct _I10_OUTPUT_DATA {
 
 /*********************************************************************
  *              $I10_OUTPUT (MSVCRT.@)
- * ld - long double to be printed to data
+ * ld80 - long double (Intel 80 bit FP in 12 bytes) to be printed to data
  * prec - precision of part, we're interested in
  * flag - 0 for first prec digits, 1 for fractional part
  * data - data to be populated
@@ -695,16 +699,22 @@ struct _I10_OUTPUT_DATA {
  *      Native sets last byte of data->str to '0' or '9', I don't know what
  *      it means. Current implementation sets it always to '0'.
  */
-int CDECL MSVCRT_I10_OUTPUT(MSVCRT__LDOUBLE ld, int prec, int flag, struct _I10_OUTPUT_DATA *data)
+int CDECL MSVCRT_I10_OUTPUT(MSVCRT__LDOUBLE ld80, int prec, int flag, struct _I10_OUTPUT_DATA *data)
 {
     static const char inf_str[] = "1#INF";
     static const char nan_str[] = "1#QNAN";
 
-    double d = ld.x;
+    /* MS' long double type wants 12 bytes for Intel's 80 bit FP format.
+     * Some UNIX have sizeof(long double) == 16, yet only 80 bit are used.
+     * Assume long double uses 80 bit FP, never seen 128 bit FP. */
+    long double ld = 0;
+    double d;
     char format[8];
     char buf[I10_OUTPUT_MAX_PREC+9]; /* 9 = strlen("0.e+0000") + '\0' */
     char *p;
 
+    memcpy(&ld, &ld80, 10);
+    d = ld;
     TRACE("(%lf %d %x %p)\n", d, prec, flag, data);
 
     if(d<0) {