So that the generated code, and more short

5 files changed, 572 insertions, 39 deletions

diff --git a/lib/Makefile b/lib/Makefile
index 300fa46..e9e12f1 100644
--- a/lib/Makefile
+++ b/lib/Makefile
@@ -6,7 +6,7 @@ TOP = ..
 include $(TOP)/Makefile
 VPATH = $(top_srcdir)/lib $(top_srcdir)/win32/lib
 
-ifndef TARGET
+ifndef TARGET # native library
  ifdef CONFIG_WIN64
   TARGET = x86_64-win32
  else
@@ -24,6 +24,11 @@ ifndef TARGET
   ifneq ($(TARGETOS),Darwin)
    XCC = $(CC)
   endif
+ else
+ ifeq ($(ARCH),arm)
+  TARGET = arm
+  XCC = $(CC)
+ endif
  endif
  endif
  endif
@@ -41,20 +46,27 @@ cross : TCC = $(TOP)/$(TARGET)-tcc$(EXESUF)
 
 I386_O = libtcc1.o alloca86.o alloca86-bt.o $(BCHECK_O)
 X86_64_O = libtcc1.o alloca86_64.o
+ARM_O = libtcc1.o armeabi.o alloca-arm.o
 WIN32_O = $(I386_O) crt1.o wincrt1.o dllcrt1.o dllmain.o chkstk.o
 WIN64_O = $(X86_64_O) crt1.o wincrt1.o dllcrt1.o dllmain.o chkstk.o
 
+# build TCC runtime library to contain PIC code, so it can be linked
+# into shared libraries
+PICFLAGS = -fPIC
+
 ifeq "$(TARGET)" "i386-win32"
  OBJ = $(addprefix $(DIR)/,$(WIN32_O))
  TGT = -DTCC_TARGET_I386 -DTCC_TARGET_PE
  XCC = $(TCC) -B$(top_srcdir)/win32 -I$(top_srcdir)/include
  XAR = $(DIR)/tiny_libmaker$(EXESUF)
+ PICFLAGS =
 else
 ifeq "$(TARGET)" "x86_64-win32"
  OBJ = $(addprefix $(DIR)/,$(WIN64_O))
  TGT = -DTCC_TARGET_X86_64 -DTCC_TARGET_PE
  XCC = $(TCC) -B$(top_srcdir)/win32 -I$(top_srcdir)/include
  XAR = $(DIR)/tiny_libmaker$(EXESUF)
+ PICFLAGS =
 else
 ifeq "$(TARGET)" "i386"
  OBJ = $(addprefix $(DIR)/,$(I386_O))
@@ -66,13 +78,19 @@ ifeq "$(TARGET)" "x86_64"
  TGT = -DTCC_TARGET_X86_64
  XCC ?= $(TCC) -B$(TOP)
 else
+ifeq "$(TARGET)" "arm"
+ OBJ = $(addprefix $(DIR)/,$(ARM_O))
+ TGT = -DTCC_TARGET_ARM
+ XCC ?= $(TCC) -B$(TOP)
+else
  $(error libtcc1.a not supported on target '$(TARGET)')
 endif
 endif
 endif
 endif
+endif
 
-XFLAGS = $(CPPFLAGS) $(CFLAGS) $(TGT)
+XFLAGS = $(CPPFLAGS) $(CFLAGS) $(PICFLAGS) $(TGT)
 
 ifeq ($(TARGETOS),Darwin)
  XAR = $(DIR)/tiny_libmaker$(EXESUF)
@@ -100,6 +118,3 @@ $(DIR)/exists :
 
 clean :
 	rm -rfv i386-win32 x86_64-win32 i386 x86_64
-
-Makefile: $(top_srcdir)/lib/Makefile
-	cp $< $@
diff --git a/lib/alloca-arm.S b/lib/alloca-arm.S
new file mode 100644
index 0000000..9deae63
--- /dev/null
+++ b/lib/alloca-arm.S
@@ -0,0 +1,11 @@
+	.text
+	.align	2
+	.global	alloca
+	.type	alloca, %function
+alloca:
+	rsb	sp, r0, sp
+	bic	sp, sp, #7
+	mov	r0, sp
+	mov	pc, lr
+	.size	alloca, .-alloca
+	.section	.note.GNU-stack,"",%progbits
diff --git a/lib/armeabi.c b/lib/armeabi.c
new file mode 100644
index 0000000..ce5e232
--- /dev/null
+++ b/lib/armeabi.c
@@ -0,0 +1,468 @@
+#include <limits.h>
+
+/* We rely on the little endianness and EABI calling convention for this to
+   work */
+
+typedef struct double_unsigned_struct {
+    unsigned low;
+    unsigned high;
+} double_unsigned_struct;
+
+typedef struct unsigned_int_struct {
+    unsigned low;
+    int high;
+} unsigned_int_struct;
+
+#define REGS_RETURN(name, type) \
+    void name ## _return(type ret) {}
+
+
+/* Float helper functions */
+
+#define FLOAT_EXP_BITS 8
+#define FLOAT_FRAC_BITS 23
+
+#define DOUBLE_EXP_BITS 11
+#define DOUBLE_FRAC_BITS 52
+
+#define ONE_EXP(type) ((1 << (type ## _EXP_BITS - 1)) - 1)
+
+REGS_RETURN(unsigned_int_struct, unsigned_int_struct)
+REGS_RETURN(double_unsigned_struct, double_unsigned_struct)
+
+/* float -> integer: (sign) 1.fraction x 2^(exponent - exp_for_one) */
+
+
+/* float to [unsigned] long long conversion */
+#define DEFINE__AEABI_F2XLZ(name, with_sign)                                 \
+void __aeabi_ ## name(unsigned val)                                          \
+{                                                                            \
+    int exp, high_shift, sign;                                               \
+    double_unsigned_struct ret;                                              \
+                                                                             \
+    /* compute sign */                                                       \
+    sign = val >> 31;                                                        \
+                                                                             \
+    /* compute real exponent */                                              \
+    exp = val >> FLOAT_FRAC_BITS;                                            \
+    exp &= (1 << FLOAT_EXP_BITS) - 1;                                        \
+    exp -= ONE_EXP(FLOAT);                                                   \
+                                                                             \
+    /* undefined behavior if truncated value cannot be represented */        \
+    if (with_sign) {                                                         \
+        if (exp > 62) /* |val| too big, double cannot represent LLONG_MAX */ \
+            return;                                                          \
+    } else {                                                                 \
+        if ((sign && exp >= 0) || exp > 63) /* if val < 0 || val too big */  \
+            return;                                                          \
+    }                                                                        \
+                                                                             \
+    val &= (1 << FLOAT_FRAC_BITS) - 1;                                       \
+    if (exp >= 32) {                                                         \
+        ret.high = 1 << (exp - 32);                                          \
+        if (exp - 32 >= FLOAT_FRAC_BITS) {                                   \
+            ret.high |= val << (exp - 32 - FLOAT_FRAC_BITS);                 \
+            ret.low = 0;                                                     \
+        } else {                                                             \
+            high_shift = FLOAT_FRAC_BITS - (exp - 32);                       \
+            ret.high |= val >> high_shift;                                   \
+            ret.low = val << (32 - high_shift);                              \
+        }                                                                    \
+    } else {                                                                 \
+        ret.high = 0;                                                        \
+        ret.low = 1 << exp;                                                  \
+        if (exp > FLOAT_FRAC_BITS)                                           \
+            ret.low |= val << (exp - FLOAT_FRAC_BITS);                       \
+        else                                                                 \
+            ret.low |= val >> (FLOAT_FRAC_BITS - exp);                       \
+    }                                                                        \
+                                                                             \
+    /* encode negative integer using 2's complement */                       \
+    if (with_sign && sign) {                                                 \
+        ret.low = ~ret.low;                                                  \
+        ret.high = ~ret.high;                                                \
+        if (ret.low == UINT_MAX) {                                           \
+            ret.low = 0;                                                     \
+            ret.high++;                                                      \
+        } else                                                               \
+            ret.low++;                                                       \
+    }                                                                        \
+                                                                             \
+    double_unsigned_struct_return(ret);                                      \
+}
+
+/* float to unsigned long long conversion */
+DEFINE__AEABI_F2XLZ(f2ulz, 0)
+
+/* float to long long conversion */
+DEFINE__AEABI_F2XLZ(f2lz, 1)
+
+/* double to [unsigned] long long conversion */
+#define DEFINE__AEABI_D2XLZ(name, with_sign)                                 \
+void __aeabi_ ## name(double_unsigned_struct val)                            \
+{                                                                            \
+    int exp, high_shift, sign;                                               \
+    double_unsigned_struct ret;                                              \
+                                                                             \
+    /* compute sign */                                                       \
+    sign = val.high >> 31;                                                   \
+                                                                             \
+    /* compute real exponent */                                              \
+    exp = (val.high >> (DOUBLE_FRAC_BITS - 32));                             \
+    exp &= (1 << DOUBLE_EXP_BITS) - 1;                                       \
+    exp -= ONE_EXP(DOUBLE);                                                  \
+                                                                             \
+    /* undefined behavior if truncated value cannot be represented */        \
+    if (with_sign) {                                                         \
+        if (exp > 62) /* |val| too big, double cannot represent LLONG_MAX */ \
+            return;                                                          \
+    } else {                                                                 \
+        if ((sign && exp >= 0) || exp > 63) /* if val < 0 || val too big */  \
+            return;                                                          \
+    }                                                                        \
+                                                                             \
+    val.high &= (1 << (DOUBLE_FRAC_BITS - 32)) - 1;                          \
+    if (exp >= 32) {                                                         \
+        ret.high = 1 << (exp - 32);                                          \
+        if (exp >= DOUBLE_FRAC_BITS) {                                       \
+            high_shift = exp - DOUBLE_FRAC_BITS;                             \
+            ret.high |= val.high << high_shift;                              \
+            ret.high |= val.low >> (32 - high_shift);                        \
+            ret.low = val.low << high_shift;                                 \
+        } else {                                                             \
+            high_shift = DOUBLE_FRAC_BITS - exp;                             \
+            ret.high |= val.high >> high_shift;                              \
+            ret.low = val.high << (32 - high_shift);                         \
+            ret.low |= val.low >> high_shift;                                \
+        }                                                                    \
+    } else {                                                                 \
+        ret.high = 0;                                                        \
+        ret.low = 1 << exp;                                                  \
+        if (exp > DOUBLE_FRAC_BITS - 32) {                                   \
+            high_shift = exp - DOUBLE_FRAC_BITS - 32;                        \
+            ret.low |= val.high << high_shift;                               \
+            ret.low |= val.low >> (32 - high_shift);                         \
+        } else                                                               \
+            ret.low |= val.high >> (DOUBLE_FRAC_BITS - 32 - exp);            \
+    }                                                                        \
+                                                                             \
+    /* encode negative integer using 2's complement */                       \
+    if (with_sign && sign) {                                                 \
+        ret.low = ~ret.low;                                                  \
+        ret.high = ~ret.high;                                                \
+        if (ret.low == UINT_MAX) {                                           \
+            ret.low = 0;                                                     \
+            ret.high++;                                                      \
+        } else                                                               \
+            ret.low++;                                                       \
+    }                                                                        \
+                                                                             \
+    double_unsigned_struct_return(ret);                                      \
+}
+
+/* double to unsigned long long conversion */
+DEFINE__AEABI_D2XLZ(d2ulz, 0)
+
+/* double to long long conversion */
+DEFINE__AEABI_D2XLZ(d2lz, 1)
+
+/* long long to float conversion */
+#define DEFINE__AEABI_XL2F(name, with_sign)                             \
+unsigned __aeabi_ ## name(unsigned long long v)                         \
+{                                                                       \
+    int s /* shift */, flb /* first lost bit */, sign = 0;              \
+    unsigned p = 0 /* power */, ret;                                    \
+    double_unsigned_struct val;                                         \
+                                                                        \
+    /* fraction in negative float is encoded in 1's complement */       \
+    if (with_sign && (v & (1ULL << 63))) {                              \
+        sign = 1;                                                       \
+        v = ~v + 1;                                                     \
+    }                                                                   \
+    val.low = v;                                                        \
+    val.high = v >> 32;                                                 \
+    /* fill fraction bits */                                            \
+    for (s = 31, p = 1 << 31; p && !(val.high & p); s--, p >>= 1);      \
+    if (p) {                                                            \
+        ret = val.high & (p - 1);                                       \
+        if (s < FLOAT_FRAC_BITS) {                                      \
+            ret <<= FLOAT_FRAC_BITS - s;                                \
+            ret |= val.low >> (32 - (FLOAT_FRAC_BITS - s));             \
+            flb = (val.low >> (32 - (FLOAT_FRAC_BITS - s - 1))) & 1;    \
+        } else {                                                        \
+            flb = (ret >> (s - FLOAT_FRAC_BITS - 1)) & 1;               \
+            ret >>= s - FLOAT_FRAC_BITS;                                \
+        }                                                               \
+        s += 32;                                                        \
+    } else {                                                            \
+        for (s = 31, p = 1 << 31; p && !(val.low & p); s--, p >>= 1);   \
+        if (p) {                                                        \
+            ret = val.low & (p - 1);                                    \
+            if (s <= FLOAT_FRAC_BITS) {                                 \
+                ret <<= FLOAT_FRAC_BITS - s;                            \
+                flb = 0;                                                \
+	    } else {                                                    \
+                flb = (ret >> (s - FLOAT_FRAC_BITS - 1)) & 1;           \
+                ret >>= s - FLOAT_FRAC_BITS;                            \
+	    }                                                           \
+        } else                                                          \
+            return 0;                                                   \
+    }                                                                   \
+    if (flb)                                                            \
+        ret++;                                                          \
+                                                                        \
+    /* fill exponent bits */                                            \
+    ret |= (s + ONE_EXP(FLOAT)) << FLOAT_FRAC_BITS;                     \
+                                                                        \
+    /* fill sign bit */                                                 \
+    ret |= sign << 31;                                                  \
+                                                                        \
+    return ret;                                                         \
+}
+
+/* unsigned long long to float conversion */
+DEFINE__AEABI_XL2F(ul2f, 0)
+
+/* long long to float conversion */
+DEFINE__AEABI_XL2F(l2f, 1)
+
+/* long long to double conversion */
+#define __AEABI_XL2D(name, with_sign)                                   \
+void __aeabi_ ## name(unsigned long long v)                             \
+{                                                                       \
+    int s /* shift */, high_shift, sign = 0;                            \
+    unsigned tmp, p = 0;                                                \
+    double_unsigned_struct val, ret;                                    \
+                                                                        \
+    /* fraction in negative float is encoded in 1's complement */       \
+    if (with_sign && (v & (1ULL << 63))) {                              \
+        sign = 1;                                                       \
+        v = ~v + 1;                                                     \
+    }                                                                   \
+    val.low = v;                                                        \
+    val.high = v >> 32;                                                 \
+                                                                        \
+    /* fill fraction bits */                                            \
+    for (s = 31, p = 1 << 31; p && !(val.high & p); s--, p >>= 1);      \
+    if (p) {                                                            \
+        tmp = val.high & (p - 1);                                       \
+        if (s < DOUBLE_FRAC_BITS - 32) {                                \
+            high_shift = DOUBLE_FRAC_BITS - 32 - s;                     \
+            ret.high = tmp << high_shift;                               \
+            ret.high |= val.low >> (32 - high_shift);                   \
+            ret.low = val.low << high_shift;                            \
+        } else {                                                        \
+            high_shift = s - (DOUBLE_FRAC_BITS - 32);                   \
+            ret.high = tmp >> high_shift;                               \
+            ret.low = tmp << (32 - high_shift);                         \
+            ret.low |= val.low >> high_shift;                           \
+            if ((val.low >> (high_shift - 1)) & 1) {                    \
+                if (ret.low == UINT_MAX) {                              \
+                    ret.high++;                                         \
+                    ret.low = 0;                                        \
+		} else                                                  \
+                    ret.low++;                                          \
+            }                                                           \
+        }                                                               \
+        s += 32;                                                        \
+    } else {                                                            \
+        for (s = 31, p = 1 << 31; p && !(val.low & p); s--, p >>= 1);   \
+        if (p) {                                                        \
+            tmp = val.low & (p - 1);                                    \
+            if (s <= DOUBLE_FRAC_BITS - 32) {                           \
+                high_shift = DOUBLE_FRAC_BITS - 32 - s;                 \
+                ret.high = tmp << high_shift;                           \
+                ret.low = 0;                                            \
+	    } else {                                                    \
+                high_shift = s - (DOUBLE_FRAC_BITS - 32);               \
+                ret.high = tmp >> high_shift;                           \
+                ret.low = tmp << (32 - high_shift);                     \
+            }                                                           \
+        } else {                                                        \
+            ret.high = ret.low = 0;                                     \
+            double_unsigned_struct_return(ret);                         \
+        }                                                               \
+    }                                                                   \
+                                                                        \
+    /* fill exponent bits */                                            \
+    ret.high |= (s + ONE_EXP(DOUBLE)) << (DOUBLE_FRAC_BITS - 32);       \
+                                                                        \
+    /* fill sign bit */                                                 \
+    ret.high |= sign << 31;                                             \
+                                                                        \
+    double_unsigned_struct_return(ret);                                 \
+}
+
+/* unsigned long long to double conversion */
+__AEABI_XL2D(ul2d, 0)
+
+/* long long to double conversion */
+__AEABI_XL2D(l2d, 1)
+
+
+/* Long long helper functions */
+
+/* TODO: add error in case of den == 0 (see §4.3.1 and §4.3.2) */
+
+#define define_aeabi_xdivmod_signed_type(basetype, type) \
+typedef struct type {                                    \
+    basetype quot;                                       \
+    unsigned basetype rem;                               \
+} type
+
+#define define_aeabi_xdivmod_unsigned_type(basetype, type) \
+typedef struct type {                                      \
+    basetype quot;                                         \
+    basetype rem;                                          \
+} type
+
+#define AEABI_UXDIVMOD(name,type, rettype, typemacro)                     \
+static inline rettype aeabi_ ## name (type num, type den)                 \
+{                                                                         \
+    rettype ret;                                                          \
+    type quot = 0;                                                        \
+                                                                          \
+    /* Increase quotient while it is less than numerator */               \
+    while (num >= den) {                                                  \
+        type q = 1;                                                       \
+                                                                          \
+        /* Find closest power of two */                                   \
+        while ((q << 1) * den <= num && q * den <= typemacro ## _MAX / 2) \
+            q <<= 1;                                                      \
+                                                                          \
+        /* Compute difference between current quotient and numerator */   \
+        num -= q * den;                                                   \
+        quot += q;                                                        \
+    }                                                                     \
+    ret.quot = quot;                                                      \
+    ret.rem = num;                                                        \
+    return ret;                                                           \
+}
+
+#define __AEABI_XDIVMOD(name, type, uiname, rettype, urettype, typemacro)     \
+void __aeabi_ ## name(type numerator, type denominator)                       \
+{                                                                             \
+    unsigned type num, den;                                                   \
+    urettype uxdiv_ret;                                                       \
+    rettype ret;                                                              \
+                                                                              \
+    if (numerator >= 0)                                                       \
+      num = numerator;                                                        \
+    else                                                                      \
+      num = 0 - numerator;                                                    \
+    if (denominator >= 0)                                                     \
+      den = denominator;                                                      \
+    else                                                                      \
+      den = 0 - denominator;                                                  \
+    uxdiv_ret = aeabi_ ## uiname(num, den);                                   \
+    /* signs differ */                                                        \
+    if ((numerator & typemacro ## _MIN) != (denominator & typemacro ## _MIN)) \
+        ret.quot = 0 - uxdiv_ret.quot;                                        \
+    else                                                                      \
+        ret.quot = uxdiv_ret.quot;                                            \
+    if (numerator < 0)                                                        \
+        ret.rem = 0 - uxdiv_ret.rem;                                          \
+    else                                                                      \
+        ret.rem = uxdiv_ret.rem;                                              \
+                                                                              \
+    rettype ## _return(ret);                                                  \
+}
+
+define_aeabi_xdivmod_signed_type(long long, lldiv_t);
+define_aeabi_xdivmod_unsigned_type(unsigned long long, ulldiv_t);
+define_aeabi_xdivmod_signed_type(int, idiv_t);
+define_aeabi_xdivmod_unsigned_type(unsigned, uidiv_t);
+
+REGS_RETURN(lldiv_t, lldiv_t)
+REGS_RETURN(ulldiv_t, ulldiv_t)
+REGS_RETURN(idiv_t, idiv_t)
+REGS_RETURN(uidiv_t, uidiv_t)
+
+AEABI_UXDIVMOD(uldivmod, unsigned long long, ulldiv_t, ULONG)
+
+__AEABI_XDIVMOD(ldivmod, long long, uldivmod, lldiv_t, ulldiv_t, LLONG)
+
+void __aeabi_uldivmod(unsigned long long num, unsigned long long den)
+{
+    ulldiv_t_return(aeabi_uldivmod(num, den));
+}
+
+void __aeabi_llsl(double_unsigned_struct val, int shift)
+{
+    double_unsigned_struct ret;
+
+    if (shift >= 32) {
+        val.high = val.low;
+        val.low = 0;
+        shift -= 32;
+    }
+    if (shift > 0) {
+        ret.low = val.low << shift;
+        ret.high = (val.high << shift) | (val.low >> (32 - shift));
+        double_unsigned_struct_return(ret);
+	return;
+    }
+    double_unsigned_struct_return(val);
+}
+
+#define aeabi_lsr(val, shift, fill, type)                          \
+    type ## _struct ret;                                           \
+                                                                   \
+    if (shift >= 32) {                                             \
+        val.low = val.high;                                        \
+        val.high = fill;                                           \
+        shift -= 32;                                               \
+    }                                                              \
+    if (shift > 0) {                                               \
+        ret.high = val.high >> shift;                              \
+        ret.low = (val.high << (32 - shift)) | (val.low >> shift); \
+        type ## _struct_return(ret);                               \
+	return;                                                    \
+    }                                                              \
+    type ## _struct_return(val);
+
+void __aeabi_llsr(double_unsigned_struct val, int shift)
+{
+    aeabi_lsr(val, shift, 0, double_unsigned);
+}
+
+void __aeabi_lasr(unsigned_int_struct val, int shift)
+{
+    aeabi_lsr(val, shift, val.high >> 31, unsigned_int);
+}
+
+
+/* Integer division functions */
+
+AEABI_UXDIVMOD(uidivmod, unsigned, uidiv_t, UINT)
+
+int __aeabi_idiv(int numerator, int denominator)
+{
+    unsigned num, den;
+    uidiv_t ret;
+
+    if (numerator >= 0)
+        num = numerator;
+    else
+        num = 0 - numerator;
+    if (denominator >= 0)
+        den = denominator;
+    else
+        den = 0 - denominator;
+    ret = aeabi_uidivmod(num, den);
+    if ((numerator & INT_MIN) != (denominator & INT_MIN)) /* signs differ */
+        ret.quot *= -1;
+    return ret.quot;
+}
+
+unsigned __aeabi_uidiv(unsigned num, unsigned den)
+{
+    return aeabi_uidivmod(num, den).quot;
+}
+
+__AEABI_XDIVMOD(idivmod, int, uidivmod, idiv_t, uidiv_t, INT)
+
+void __aeabi_uidivmod(unsigned num, unsigned den)
+{
+    uidiv_t_return(aeabi_uidivmod(num, den));
+}
diff --git a/lib/bcheck.c b/lib/bcheck.c
index 064fb5d..8ac7d4e 100644
--- a/lib/bcheck.c
+++ b/lib/bcheck.c
@@ -22,7 +22,7 @@
 #include <stdarg.h>
 #include <string.h>
 #if !defined(__FreeBSD__) && !defined(__FreeBSD_kernel__) \
-    && !defined(__DragonFly__) && !defined(__OpenBSD__)
+    && !defined(__DragonFly__) && !defined(__OpenBSD__) && !defined(__NetBSD__)
 #include <malloc.h>
 #endif
 #if !defined(_WIN32)
@@ -41,7 +41,7 @@
 
 #if defined(__FreeBSD__) || defined(__FreeBSD_kernel__) \
     || defined(__DragonFly__) || defined(__dietlibc__) \
-    || defined(__UCLIBC__) || defined(__OpenBSD__) \
+    || defined(__UCLIBC__) || defined(__OpenBSD__) || defined(__NetBSD__) \
     || defined(_WIN32) || defined(TCC_UCLIBC)
 #warning Bound checking does not support malloc (etc.) in this environment.
 #undef CONFIG_TCC_MALLOC_HOOKS
@@ -418,6 +418,13 @@ void __bound_init(void)
     }
 }
 
+void __bound_main_arg(void **p)
+{
+    void *start = p;
+    while (*p++);
+    __bound_new_region(start, (void *) p - start);
+}
+
 void __bound_exit(void)
 {
     restore_malloc_hooks();
@@ -628,7 +635,7 @@ int __bound_delete_region(void *p)
 }
 
 /* return the size of the region starting at p, or EMPTY_SIZE if non
-   existant region. */
+   existent region. */
 static unsigned long get_region_size(void *p)
 {
     unsigned long addr = (unsigned long)p;
diff --git a/lib/libtcc1.c b/lib/libtcc1.c
index 53dbec4..284965e 100644
--- a/lib/libtcc1.c
+++ b/lib/libtcc1.c
@@ -28,6 +28,8 @@ the Free Software Foundation, 59 Temple Place - Suite 330,
 Boston, MA 02111-1307, USA.  
 */
 
+#include <stdint.h>
+
 #define W_TYPE_SIZE   32
 #define BITS_PER_UNIT 8
 
@@ -107,10 +109,10 @@ union float_long {
 };
 
 /* XXX: we don't support several builtin supports for now */
-#ifndef __x86_64__
+#if !defined(TCC_TARGET_X86_64) && !defined(TCC_TARGET_ARM)
 
 /* XXX: use gcc/tcc intrinsic ? */
-#if defined(__i386__)
+#if defined(TCC_TARGET_I386)
 #define sub_ddmmss(sh, sl, ah, al, bh, bl) \
   __asm__ ("subl %5,%1\n\tsbbl %3,%0"					\
 	   : "=r" ((USItype) (sh)),					\
@@ -478,13 +480,6 @@ long long __ashldi3(long long a, int b)
 #endif
 }
 
-#if defined(__i386__)
-/* FPU control word for rounding to nearest mode */
-unsigned short __tcc_fpu_control = 0x137f;
-/* FPU control word for round to zero mode for int conversion */
-unsigned short __tcc_int_fpu_control = 0x137f | 0x0c00;
-#endif
-
 #endif /* !__x86_64__ */
 
 /* XXX: fix tcc's code generator to do this instead */
@@ -605,15 +600,38 @@ unsigned long long __fixunsxfdi (long double a1)
         return 0;
 }
 
-#if defined(__x86_64__) && !defined(_WIN64)
+long long __fixsfdi (float a1)
+{
+    long long ret; int s;
+    ret = __fixunssfdi((s = a1 >= 0) ? a1 : -a1);
+    return s ? ret : -ret;
+}
+
+long long __fixdfdi (double a1)
+{
+    long long ret; int s;
+    ret = __fixunsdfdi((s = a1 >= 0) ? a1 : -a1);
+    return s ? ret : -ret;
+}
+
+long long __fixxfdi (long double a1)
+{
+    long long ret; int s;
+    ret = __fixunsxfdi((s = a1 >= 0) ? a1 : -a1);
+    return s ? ret : -ret;
+}
+
+#if defined(TCC_TARGET_X86_64) && !defined(_WIN64)
 
 #ifndef __TINYC__
 #include <stdlib.h>
 #include <stdio.h>
+#include <string.h>
 #else
 /* Avoid including stdlib.h because it is not easily available when
    cross compiling */
 extern void *malloc(unsigned long long);
+void *memset(void *s, int c, size_t n);
 extern void free(void*);
 extern void abort(void);
 #endif
@@ -622,8 +640,9 @@ enum __va_arg_type {
     __va_gen_reg, __va_float_reg, __va_stack
 };
 
+//This should be in sync with the declaration on our include/stdarg.h
 /* GCC compatible definition of va_list. */
-struct __va_list_struct {
+typedef struct {
     unsigned int gp_offset;
     unsigned int fp_offset;
     union {
@@ -631,19 +650,22 @@ struct __va_list_struct {
         char *overflow_arg_area;
     };
     char *reg_save_area;
-};
+} __va_list_struct;
 
-void *__va_start(void *fp)
+#undef __va_start
+#undef __va_arg
+#undef __va_copy
+#undef __va_end
+
+void __va_start(__va_list_struct *ap, void *fp)
 {
-    struct __va_list_struct *ap =
-        (struct __va_list_struct *)malloc(sizeof(struct __va_list_struct));
-    *ap = *(struct __va_list_struct *)((char *)fp - 16);
+    memset(ap, 0, sizeof(__va_list_struct));
+    *ap = *(__va_list_struct *)((char *)fp - 16);
     ap->overflow_arg_area = (char *)fp + ap->overflow_offset;
     ap->reg_save_area = (char *)fp - 176 - 16;
-    return ap;
 }
 
-void *__va_arg(struct __va_list_struct *ap,
+void *__va_arg(__va_list_struct *ap,
                enum __va_arg_type arg_type,
                int size, int align)
 {
@@ -669,7 +691,7 @@ void *__va_arg(struct __va_list_struct *ap,
     case __va_stack:
     use_overflow_area:
         ap->overflow_arg_area += size;
-        ap->overflow_arg_area = (char*)((long long)(ap->overflow_arg_area + align - 1) & -(long long)align);
+        ap->overflow_arg_area = (char*)((intptr_t)(ap->overflow_arg_area + align - 1) & -(intptr_t)align);
         return ap->overflow_arg_area - size;
 
     default:
@@ -680,26 +702,36 @@ void *__va_arg(struct __va_list_struct *ap,
     }
 }
 
-void *__va_copy(struct __va_list_struct *src)
-{
-    struct __va_list_struct *dest =
-        (struct __va_list_struct *)malloc(sizeof(struct __va_list_struct));
-    *dest = *src;
-    return dest;
-}
+#endif /* __x86_64__ */
 
-void __va_end(struct __va_list_struct *ap)
+/* Flushing for tccrun */
+#if defined(TCC_TARGET_X86_64) || defined(TCC_TARGET_I386)
+
+void __clear_cache(char *beginning, char *end)
 {
-    free(ap);
 }
 
-#endif /* __x86_64__ */
+#elif defined(TCC_TARGET_ARM)
 
-/* Flushing for tccrun */
-#if defined(__x86_64__) || defined(__i386__)
+#define _GNU_SOURCE
+#include <unistd.h>
+#include <sys/syscall.h>
+#include <stdio.h>
 
 void __clear_cache(char *beginning, char *end)
 {
+/* __ARM_NR_cacheflush is kernel private and should not be used in user space.
+ * However, there is no ARM asm parser in tcc so we use it for now */
+#if 1
+    syscall(__ARM_NR_cacheflush, beginning, end, 0);
+#else
+    __asm__ ("push {r7}\n\t"
+             "mov r7, #0xf0002\n\t"
+             "mov r2, #0\n\t"
+             "swi 0\n\t"
+             "pop {r7}\n\t"
+             "ret");
+#endif
 }
 
 #else