Add optional ICC-only support for using long doubles on Win32

[p5pRT]

Migrated from rt.perl.org#132478 (status was 'new')

Searchable as RT132478$

[p5pRT]

From @steve-m-hay

Created by @steve-m-hay

The attached patch by bulk88 taken from
https://rt-archive.perl.org/perl5/Ticket/Display.html?id=123113
is a work in progress to add optional ICC (Intel C++ Compiler) support
for "long doubles"
on Windows.

See the following entries on that ticket for further details​:

https://rt-archive.perl.org/perl5/Ticket/Display.html?id=123113#txn-1316983
https://rt-archive.perl.org/perl5/Ticket/Display.html?id=123113#txn-1316988

Perl Info

Flags:
    category=core
    severity=low

Site configuration information for perl 5.27.7:

Configured by shay at Tue Nov 21 13:25:42 2017.

Summary of my perl5 (revision 5 version 27 subversion 7) configuration:

  Platform:
    osname=MSWin32
    osvers=6.3.9600
    archname=MSWin32-x86-multi-thread-ld
    uname=''
    config_args='undef'
    hint=recommended
    useposix=true
    d_sigaction=undef
    useithreads=define
    usemultiplicity=define
    use64bitint=undef
    use64bitall=undef
    uselongdouble=define
    usemymalloc=n
    default_inc_excludes_dot=define
    bincompat5005=undef
  Compiler:
    cc='gcc'
    ccflags =' -s -O2 -DWIN32  -DPERL_TEXTMODE_SCRIPTS
-DPERL_IMPLICIT_CONTEXT -DPERL_IMPLICIT_SYS -D__USE_MINGW_ANSI_STDIO
-fwrapv -fno-strict-aliasing -mms-bitfields'
    optimize='-s -O2'
    cppflags='-DWIN32'
    ccversion=''
    gccversion='7.1.0'
    gccosandvers=''
    intsize=4
    longsize=4
    ptrsize=4
    doublesize=8
    byteorder=1234
    doublekind=3
    d_longlong=define
    longlongsize=8
    d_longdbl=define
    longdblsize=12
    longdblkind=3
    ivtype='long'
    ivsize=4
    nvtype='long double'
    nvsize=12
    Off_t='long long'
    lseeksize=8
    alignbytes=8
    prototype=define
  Linker and Libraries:
    ld='g++'
    ldflags ='-s -L"c:\perl\lib\CORE"
-L"C:\Dev\Software\MinGW-w64\x86\7.1.0\lib"'
    libpth=C:\Dev\Software\MinGW-w64\x86\7.1.0\lib
    libs=-lmoldname -lkernel32 -luser32 -lgdi32 -lwinspool -lcomdlg32
-ladvapi32 -lshell32 -lole32 -loleaut32 -lnetapi32 -luuid -lws2_32
-lmpr -lwinmm -lversion -lodbc32 -lodbccp32 -lcomctl32
    perllibs=-lmoldname -lkernel32 -luser32 -lgdi32 -lwinspool
-lcomdlg32 -ladvapi32 -lshell32 -lole32 -loleaut32 -lnetapi32 -luuid
-lws2_32 -lmpr -lwinmm -lversion -lodbc32 -lodbccp32 -lcomctl32
    libc=
    so=dll
    useshrplib=true
    libperl=libperl527.a
    gnulibc_version=''
  Dynamic Linking:
    dlsrc=dl_win32.xs
    dlext=dll
    d_dlsymun=undef
    ccdlflags=' '
    cccdlflags=' '
    lddlflags='-mdll -s -L"c:\perl\lib\CORE"
-L"C:\Dev\Software\MinGW-w64\x86\7.1.0\lib"'



@INC for perl 5.27.7:
    C:/perl/site/lib
    C:/perl/lib


Environment for perl 5.27.7:
    HOME (unset)
    LANG (unset)
    LANGUAGE (unset)
    LD_LIBRARY_PATH (unset)
    LOGDIR (unset)
    PATH=C:\Dev\Software\MinGW-w64\x86\7.1.0\bin;C:\Windows\system32;C:\Windows;C:\Windows\system32\wbem;C:\perl\bin
    PERL_BADLANG (unset)
    SHELL (unset)

[p5pRT]

From @steve-m-hay

0001-icc-longdouble-wip.patch

From 6d149aa0e5feff84216d4da035c7f1b2278e81bc Mon Sep 17 00:00:00 2001
From: Daniel Dragan <bulk88@hotmail.com>
Date: Thu, 9 Jan 2014 03:10:40 -0500
Subject: [PATCH] icc longdouble wip

---
 generate_uudmap.c |  2 ++
 perl.h            | 12 ++++++++++++
 uconfig.h         |  8 ++++----
 util.c            |  9 +++++++--
 win32/Makefile    | 10 +++++-----
 win32/config.vc   | 18 +++++++++---------
 win32/config_H.vc | 30 +++++++++++++++---------------
 7 files changed, 54 insertions(+), 35 deletions(-)

diff --git a/generate_uudmap.c b/generate_uudmap.c
index b6307c0..b72f374 100644
--- a/generate_uudmap.c
+++ b/generate_uudmap.c
@@ -108,6 +108,8 @@ int main(int argc, char **argv) {
   int bits;
   struct mg_data_raw_t *p = mg_data_raw;
 
+    fprintf(stderr, "sizeof ld %d\n", sizeof(long double));
+
   if (argc < 4 || argv[1][0] == '\0' || argv[2][0] == '\0'
       || argv[3][0] == '\0') {
     fprintf(stderr, "Usage: %s uudemap.h bitcount.h mg_data.h\n", argv[0]);
diff --git a/perl.h b/perl.h
index b6e0c3e..6c7d5bd 100644
--- a/perl.h
+++ b/perl.h
@@ -1671,6 +1671,7 @@ typedef UVTYPE UV;
 #  endif
 #  if !(defined(HAS_LONG_DOUBLE) && (LONG_DOUBLESIZE > DOUBLESIZE))
 #     undef USE_LONG_DOUBLE /* Ouch! */
+      #error undefed ULD
 #  endif
 #endif
 
@@ -1763,10 +1764,15 @@ typedef NVTYPE NV;
 #   include <ieeefp.h>
 #endif
 
+#ifndef USE_LONG_DOUBLE
+#error no LD
+#endif
+
 #ifdef USE_LONG_DOUBLE
 #   ifdef I_SUNMATH
 #       include <sunmath.h>
 #   endif
+#   include <mathimf.h>
 #   define NV_DIG LDBL_DIG
 #   ifdef LDBL_MANT_DIG
 #       define NV_MANT_DIG LDBL_MANT_DIG
@@ -1798,6 +1804,7 @@ typedef NVTYPE NV;
 #           endif
 #       endif
 #   endif
+
 #   ifdef HAS_SQRTL
 #       define Perl_cos cosl
 #       define Perl_sin sinl
@@ -1843,6 +1850,7 @@ EXTERN_C long double modfl(long double, long double *);
 #       endif
 #   endif
 #else
+#error die
 #   define NV_DIG DBL_DIG
 #   ifdef DBL_MANT_DIG
 #       define NV_MANT_DIG DBL_MANT_DIG
@@ -5833,6 +5841,10 @@ extern void moncontrol(int);
 
 */
 
+#ifndef HAS_SQRTL
+#error no long
+#endif
+
 #endif /* Include guard */
 
 /*
diff --git a/uconfig.h b/uconfig.h
index 8e728ab..e29fabd 100644
--- a/uconfig.h
+++ b/uconfig.h
@@ -4361,10 +4361,10 @@
  *	This symbol, if defined, contains the string used by stdio to
  *	format long doubles (format 'f') for input.
  */
-/*#define PERL_PRIfldbl	"llf"	/ **/
-/*#define PERL_PRIgldbl	"llg"	/ **/
-/*#define PERL_PRIeldbl	"lle"	/ **/
-/*#define PERL_SCNfldbl	"llf"	/ **/
+#define PERL_PRIfldbl	"llf"	/ **/
+#define PERL_PRIgldbl	"llg"	/ **/
+#define PERL_PRIeldbl	"lle"	/ **/
+#define PERL_SCNfldbl	"llf"	/ **/
 
 /* PERL_MAD:
  *	This symbol, if defined, indicates that the Misc Attribution
diff --git a/util.c b/util.c
index f308e93..455b7bb 100644
--- a/util.c
+++ b/util.c
@@ -4769,10 +4769,12 @@ Perl_my_snprintf(char *buffer, const Size_t len, const char *format, ...)
 {
     int retval;
     va_list ap;
+    restart:
     PERL_ARGS_ASSERT_MY_SNPRINTF;
     va_start(ap, format);
+    memset(buffer, 0, len);
 #ifdef HAS_VSNPRINTF
-    retval = vsnprintf(buffer, len, format, ap);
+    retval = ___mingw_vsnprintf(buffer, len, format, ap);
 #else
     retval = vsprintf(buffer, format, ap);
 #endif
@@ -4784,8 +4786,11 @@ Perl_my_snprintf(char *buffer, const Size_t len, const char *format, ...)
         ||
         (len > 0 && (Size_t)retval >= len) 
 #endif
-    )
+    ) {
+	DebugBreak();
+	goto restart;
 	Perl_croak_nocontext("panic: my_snprintf buffer overflow");
+    }
     return retval;
 }
 
diff --git a/win32/Makefile b/win32/Makefile
index 885f8ce..2c90bef 100644
--- a/win32/Makefile
+++ b/win32/Makefile
@@ -134,14 +134,14 @@ CCTYPE		= MSVC90
 #
 # If you are using Intel C++ Compiler uncomment this
 #
-#__ICC		= define
+__ICC		= define
 
 #
 # uncomment next line if you want debug version of perl (big,slow)
 # If not enabled, we automatically try to use maximum optimization
 # with all compilers that are known to have a working optimizer.
 #
-#CFG		= Debug
+CFG		= Debug
 
 #
 # uncomment to enable linking with setargv.obj under the Visual C
@@ -443,10 +443,10 @@ LOCDEFS		= -DPERLDLL -DPERL_CORE
 SUBSYS		= console
 CXX_FLAG	= -TP -EHsc
 
-LIBC	= msvcrt.lib
+LIBC	= msvcrtd.lib
 
 !IF  "$(CFG)" == "Debug"
-OPTIMIZE	= -Od -MD -Zi -DDEBUGGING
+OPTIMIZE	= -Od -MDd -Zi -DDEBUGGING
 LINK_DBG	= -debug
 !ELSE
 # -O1 yields smaller code, which turns out to be faster than -O2 on x86 and x64
@@ -517,7 +517,7 @@ LIBBASEFILES    = $(LIBBASEFILES) bufferoverflowU.lib
 LIBFILES	= $(LIBBASEFILES) $(LIBC)
 
 #EXTRACFLAGS	= -nologo -GF -W4 -wd4127 -wd4706
-EXTRACFLAGS	= -nologo -GF -W3
+EXTRACFLAGS	= -nologo -GF -W3 /Qlong-double /Qpc80 -DUSE_LONG_DOUBLE -DHAS_LONG_DOUBLE
 CFLAGS		= $(EXTRACFLAGS) $(INCLUDES) $(DEFINES) $(LOCDEFS) \
 		$(PCHFLAGS) $(OPTIMIZE)
 LINK_FLAGS	= -nologo -nodefaultlib $(LINK_DBG) \
diff --git a/win32/config.vc b/win32/config.vc
index a15b926..bc90fd8 100644
--- a/win32/config.vc
+++ b/win32/config.vc
@@ -449,7 +449,7 @@ d_socklen_t='undef'
 d_sockpair='undef'
 d_socks5_init='undef'
 d_sprintf_returns_strlen='define'
-d_sqrtl='undef'
+d_sqrtl='define'
 d_srand48_r='undef'
 d_srandom_r='undef'
 d_sresgproto='undef'
@@ -828,16 +828,16 @@ nm_opt=''
 nm_so_opt=''
 nonxs_ext='Errno'
 nroff=''
-nvEUformat='"E"'
-nvFUformat='"F"'
-nvGUformat='"G"'
+nvEUformat='"LE"'
+nvFUformat='"LF"'
+nvGUformat='"LG"'
 nv_overflows_integers_at='256.0*256.0*256.0*256.0*256.0*256.0*2.0*2.0*2.0*2.0*2.0'
 nv_preserves_uv_bits='32'
-nveformat='"e"'
-nvfformat='"f"'
-nvgformat='"g"'
-nvsize='8'
-nvtype='double'
+nveformat='"Le"'
+nvfformat='"Lf"'
+nvgformat='"Lg"'
+nvsize='12'
+nvtype='long double'
 o_nonblock='O_NONBLOCK'
 obj_ext='.obj'
 old_pthread_create_joinable=''
diff --git a/win32/config_H.vc b/win32/config_H.vc
index f55a6d2..7f612f3 100644
--- a/win32/config_H.vc
+++ b/win32/config_H.vc
@@ -1930,9 +1930,9 @@
  *	C preprocessor can make decisions based on it.  It is only
  *	defined if the system supports long doubles.
  */
-#define HAS_LONG_DOUBLE		/**/
+/*#define HAS_LONG_DOUBLE		/**/
 #ifdef HAS_LONG_DOUBLE
-#define LONG_DOUBLESIZE 8		/**/
+#define LONG_DOUBLESIZE 16		/**/
 #endif
 
 /* HAS_LONG_LONG:
@@ -3558,7 +3558,7 @@
  *	available to break a long double floating-point number into
  *	a normalized fraction and an integral power of 2.
  */
-/*#define HAS_FREXPL		/ **/
+#define HAS_FREXPL
 
 /* HAS_STRUCT_FS_DATA:
  *	This symbol, if defined, indicates that the struct fs_data
@@ -3780,8 +3780,8 @@
  *	and 1.150000.  The bug has been seen in certain versions of glibc,
  *	release 2.2.2 is known to be okay.
  */
-/*#define HAS_MODFL		/ **/
-/*#define HAS_MODFL_PROTO		/ **/
+#define HAS_MODFL
+#define HAS_MODFL_PROTO
 /*#define HAS_MODFL_POW32_BUG		/ **/
 
 /* HAS_MPROTECT:
@@ -3954,7 +3954,7 @@
  *	This symbol, if defined, indicates that the sqrtl routine is
  *	available to do long double square roots.
  */
-/*#define HAS_SQRTL		/ **/
+#define HAS_SQRTL		/ **/
 
 /* HAS_SETRESGID_PROTO:
  *	This symbol, if defined, indicates that the system provides
@@ -4361,10 +4361,10 @@
  *	This symbol, if defined, contains the string used by stdio to
  *	format long doubles (format 'f') for input.
  */
-/*#define PERL_PRIfldbl	"f"	/ **/
-/*#define PERL_PRIgldbl	"g"	/ **/
-/*#define PERL_PRIeldbl	"e"	/ **/
-/*#define PERL_SCNfldbl	"f"	/ **/
+#define PERL_PRIfldbl	"f"
+#define PERL_PRIgldbl	"g"
+#define PERL_PRIeldbl	"e"
+#define PERL_SCNfldbl	"f"
 
 /* PERL_MAD:
  *	This symbol, if defined, indicates that the Misc Attribution
@@ -4479,7 +4479,7 @@
 #define	I64TYPE		__int64	/**/
 #define	U64TYPE		unsigned __int64	/**/
 #endif
-#define	NVTYPE		double		/**/
+#define	NVTYPE		long double		/**/
 #define	IVSIZE		4		/**/
 #define	UVSIZE		4		/**/
 #define	I8SIZE		1		/**/
@@ -4492,7 +4492,7 @@
 #define	I64SIZE		8	/**/
 #define	U64SIZE		8	/**/
 #endif
-#define	NVSIZE		8		/**/
+#define	NVSIZE		12		/**/
 #define	NV_PRESERVES_UV
 #define	NV_PRESERVES_UV_BITS	32
 #define	NV_OVERFLOWS_INTEGERS_AT	256.0*256.0*256.0*256.0*256.0*256.0*2.0*2.0*2.0*2.0*2.0
@@ -4548,9 +4548,9 @@
 #define	UVof		"lo"		/**/
 #define	UVxf		"lx"		/**/
 #define	UVXf		"lX"		/**/
-#define	NVef		"e"		/**/
-#define	NVff		"f"		/**/
-#define	NVgf		"g"		/**/
+#define	NVef		"Le"		/**/
+#define	NVff		"Lf"		/**/
+#define	NVgf		"Lg"		/**/
 
 /* SELECT_MIN_BITS:
  *	This symbol holds the minimum number of bits operated by select.
-- 
1.8.0.msysgit.0