Merge 1dab08b1ab into e3c70a7d81

Reported by Christian Clason. #1334

Makefile

@@ -110,7 +110,7 @@ else
 endif
 TARGET_SYS?= $(HOST_SYS)
 
-ifeq (Darwin,$(TARGET_SYS))
+ifneq (,$(filter $(TARGET_SYS),Darwin iOS))
   INSTALL_SONAME= $(INSTALL_DYLIBNAME)
   INSTALL_SOSHORT1= $(INSTALL_DYLIBSHORT1)
   INSTALL_SOSHORT2= $(INSTALL_DYLIBSHORT2)

doc/install.html

@@ -117,7 +117,7 @@ hold all user-configurable settings:
 <li><tt>Makefile</tt> has settings for <b>installing</b> LuaJIT (POSIX
 only).</li>
 <li><tt>src/Makefile</tt> has settings for <b>compiling</b> LuaJIT
-under POSIX, MinGW or Cygwin.</li>
+under POSIX or MinGW.</li>
 <li><tt>src/msvcbuild.bat</tt> has settings for compiling LuaJIT with
 MSVC (Visual Studio).</li>
 </ul>
@@ -195,10 +195,8 @@ Obviously the prefixes given during build and installation need to be the same.
 <h2 id="windows">Windows Systems</h2>
 <h3>Prerequisites</h3>
 <p>
-Either install one of the open source SDKs
+Either install the open source SDK <a href="http://mingw.org/"><span class="ext">&raquo;</span>&nbsp;MinGW</a>,
-(<a href="http://mingw.org/"><span class="ext">&raquo;</span>&nbsp;MinGW</a> or
+which comes with a modified GCC plus the required development headers.
-<a href="https://www.cygwin.com/"><span class="ext">&raquo;</span>&nbsp;Cygwin</a>), which come with a modified
-GCC plus the required development headers.
 Or install Microsoft's Visual Studio (MSVC).
 </p>
 <h3>Building with MSVC</h3>
@@ -217,9 +215,9 @@ Then follow the installation instructions below.
 <p>
 For an x64 to ARM64 cross-build run this first: <tt>vcvarsall.bat x64_arm64</tt>
 </p>
-<h3>Building with MinGW or Cygwin</h3>
+<h3>Building with MinGW</h3>
 <p>
-Open a command prompt window and make sure the MinGW or Cygwin programs
+Open a command prompt window and make sure the MinGW programs
 are in your path. Then <tt>cd</tt> to the directory of the git repository.
 Then run this command for MinGW:
 </p>
@@ -227,12 +225,6 @@ Then run this command for MinGW:
 mingw32-make
 </pre>
 <p>
-Or this command for Cygwin:
-</p>
-<pre class="code">
-make
-</pre>
-<p>
 Then follow the installation instructions below.
 </p>
 <h3>Installing LuaJIT</h3>
@@ -249,6 +241,19 @@ absolute path names &mdash; all modules are loaded relative to the
 directory where <tt>luajit.exe</tt> is installed
 (see <tt>src/luaconf.h</tt>).
 </p>
+<p>
+The final directory layout should look like this:
+</p>
+<pre class="code">
+├── luajit.exe
+├── lua51.dll
+├── <- put your own classic Lua/C API modules (*.dll) here
+└── lua
+    ├── <- put your own Lua modules (*.lua) here
+    └── jit
+        ├── bc.lua
+        └── (etc …)
+</pre>
 
 <h2 id="cross">Cross-compiling LuaJIT</h2>
 <p>

src/Makefile.dep

@@ -25,7 +25,7 @@ lib_ffi.o: lib_ffi.c lua.h luaconf.h lauxlib.h lualib.h lj_obj.h lj_def.h \
 lib_init.o: lib_init.c lua.h luaconf.h lauxlib.h lualib.h lj_arch.h
 lib_io.o: lib_io.c lua.h luaconf.h lauxlib.h lualib.h lj_obj.h lj_def.h \
  lj_arch.h lj_gc.h lj_err.h lj_errmsg.h lj_buf.h lj_str.h lj_state.h \
- lj_strfmt.h lj_ff.h lj_ffdef.h lj_lib.h lj_libdef.h
+ lj_strfmt.h lj_ff.h lj_ffdef.h lj_lib.h lj_strscan.h lj_libdef.h
 lib_jit.o: lib_jit.c lua.h luaconf.h lauxlib.h lualib.h lj_obj.h lj_def.h \
  lj_arch.h lj_gc.h lj_err.h lj_errmsg.h lj_debug.h lj_str.h lj_tab.h \
  lj_state.h lj_bc.h lj_ctype.h lj_ir.h lj_jit.h lj_ircall.h lj_iropt.h \

src/lib_io.c

@@ -25,6 +25,7 @@
 #include "lj_strfmt.h"
 #include "lj_ff.h"
 #include "lj_lib.h"
+#include "lj_strscan.h"
 
 /* Userdata payload for I/O file. */
 typedef struct IOFileUD {
@@ -323,13 +324,14 @@ LJLIB_CF(io_method_seek)
   FILE *fp = io_tofile(L)->fp;
   int opt = lj_lib_checkopt(L, 2, 1, "\3set\3cur\3end");
   int64_t ofs = 0;
-  cTValue *o;
+  TValue *o;
   int res;
   if (opt == 0) opt = SEEK_SET;
   else if (opt == 1) opt = SEEK_CUR;
   else if (opt == 2) opt = SEEK_END;
   o = L->base+2;
   if (o < L->top) {
+    if (tvisstr(o)) lj_strscan_num(strV(o), o);
     if (tvisint(o))
       ofs = (int64_t)intV(o);
     else if (tvisnum(o))

src/lj_asm_x86.h

@@ -886,32 +886,38 @@ static void asm_conv(ASMState *as, IRIns *ir)
       asm_tointg(as, ir, ra_alloc1(as, lref, RSET_FPR));
     } else {
       Reg dest = ra_dest(as, ir, RSET_GPR);
+      Reg tmp = ra_noreg(IR(lref)->r) ? ra_alloc1(as, lref, RSET_FPR) :
+					ra_scratch(as, RSET_FPR);
       x86Op op = st == IRT_NUM ? XO_CVTTSD2SI : XO_CVTTSS2SI;
-      if (LJ_64 ? irt_isu64(ir->t) : irt_isu32(ir->t)) {
+      Reg r64 = (LJ_64 && irt_is64 (ir->t)) ? REX_64 : 0;
-	/* LJ_64: For inputs >= 2^63 add -2^64, convert again. */
+      if (LJ_64 && (irt_isu32(ir->t) || irt_isint(ir->t)))
-	/* LJ_32: For inputs >= 2^31 add -2^31, convert again and add 2^31. */
+	emit_rr(as, XO_MOV, dest, dest);  /* Zero hiword. */
-	Reg tmp = ra_noreg(IR(lref)->r) ? ra_alloc1(as, lref, RSET_FPR) :
+      if (irt_isu64(ir->t) || irt_isu32(ir->t)) {
-					  ra_scratch(as, RSET_FPR);
+        /* The cvtsd2si family of instructions operates on the signed integers,
-	MCLabel l_end = emit_label(as);
+           producing INT_MIN on error.  However we're converting to an unsigned
-	if (LJ_32)
+           integer, so we want to accept the whole unsigned integer range.
-	  emit_gri(as, XG_ARITHi(XOg_ADD), dest, (int32_t)0x80000000);
+           Convert both the number and the number minus INT_MIN, choosing the
-	emit_rr(as, op, dest|REX_64, tmp);
+           first result if successful and the second otherwise.  */
-	if (st == IRT_NUM)
+        Reg dest2 = ra_scratch(as, rset_exclude(RSET_GPR, dest));
-	  emit_rma(as, XO_ADDSD, tmp, &as->J->k64[LJ_K64_M2P64_31]);
+        Reg tmp2 = ra_scratch(as, rset_exclude(RSET_FPR, tmp));
-	else
+        x86Op sub_op;
-	  emit_rma(as, XO_ADDSS, tmp, &as->J->k32[LJ_K32_M2P64_31]);
+        void *krange;
-	emit_sjcc(as, CC_NS, l_end);
+        if (st == IRT_NUM) {
-	emit_rr(as, XO_TEST, dest|REX_64, dest);  /* Check if dest negative. */
+          sub_op = XO_SUBSD;
-	emit_rr(as, op, dest|REX_64, tmp);
+          krange = &as->J->k64[irt_isu64(ir->t) ? LJ_K64_2P64 : LJ_K64_2P32];
-	ra_left(as, tmp, lref);
+        } else {
-      } else {
+          sub_op = XO_SUBSS;
-	if (LJ_64 && irt_isu32(ir->t))
+          krange = &as->J->k32[irt_isu64(ir->t) ? LJ_K32_2P64 : LJ_K32_2P32];
-	  emit_rr(as, XO_MOV, dest, dest);  /* Zero hiword. */
+        }
-	emit_mrm(as, op,
+        emit_rr(as, XO_CMOV + (CC_O<<24), dest|r64, dest2|r64);
-		 dest|((LJ_64 &&
+        emit_i8(as, 1);
-			(irt_is64(ir->t) || irt_isu32(ir->t))) ? REX_64 : 0),
+        emit_rr(as, XO_ARITHi8, XOg_CMP|r64, dest);
-		 asm_fuseload(as, lref, RSET_FPR));
+        emit_rr(as, op, dest2|r64, tmp2);
+        emit_rma(as, sub_op, tmp2, krange);
+        emit_rr(as, XO_MOVAPS, tmp2, tmp);
       }
+      emit_rr(as, op, dest|r64, tmp);
+      ra_left(as, tmp, lref);
     }
   } else if (st >= IRT_I8 && st <= IRT_U16) {  /* Extend to 32 bit integer. */
     Reg left, dest = ra_dest(as, ir, RSET_GPR);

src/lj_bcread.c

@@ -179,7 +179,7 @@ static const void *bcread_varinfo(GCproto *pt)
 }
 
 /* Read a single constant key/value of a template table. */
-static void bcread_ktabk(LexState *ls, TValue *o)
+static void bcread_ktabk(LexState *ls, TValue *o, GCtab *t)
 {
   MSize tp = bcread_uleb128(ls);
   if (tp >= BCDUMP_KTAB_STR) {
@@ -191,6 +191,8 @@ static void bcread_ktabk(LexState *ls, TValue *o)
   } else if (tp == BCDUMP_KTAB_NUM) {
     o->u32.lo = bcread_uleb128(ls);
     o->u32.hi = bcread_uleb128(ls);
+  } else if (tp == BCDUMP_KTAB_NIL) { /* Restore nil value marker. */
+    settabV(ls->L, o, t);
   } else {
     lj_assertLS(tp <= BCDUMP_KTAB_TRUE, "bad constant type %d", tp);
     setpriV(o, ~tp);
@@ -207,15 +209,15 @@ static GCtab *bcread_ktab(LexState *ls)
     MSize i;
     TValue *o = tvref(t->array);
     for (i = 0; i < narray; i++, o++)
-      bcread_ktabk(ls, o);
+      bcread_ktabk(ls, o, t);
   }
   if (nhash) {  /* Read hash entries. */
     MSize i;
     for (i = 0; i < nhash; i++) {
       TValue key;
-      bcread_ktabk(ls, &key);
+      bcread_ktabk(ls, &key, t);
       lj_assertLS(!tvisnil(&key), "nil key");
-      bcread_ktabk(ls, lj_tab_set(ls->L, t, &key));
+      bcread_ktabk(ls, lj_tab_set(ls->L, t, &key), t);
     }
   }
   return t;

src/lj_bcwrite.c

@@ -71,6 +71,8 @@ static void bcwrite_ktabk(BCWriteCtx *ctx, cTValue *o, int narrow)
     *p++ = BCDUMP_KTAB_NUM;
     p = lj_strfmt_wuleb128(p, o->u32.lo);
     p = lj_strfmt_wuleb128(p, o->u32.hi);
+  } else if (tvistab(o)) { /* Write the nil value marker as a nil. */
+    *p++ = BCDUMP_KTAB_NIL;
   } else {
     lj_assertBCW(tvispri(o), "unhandled type %d", itype(o));
     *p++ = BCDUMP_KTAB_NIL+~itype(o);
@@ -133,7 +135,7 @@ static void bcwrite_ktab_sorted_hash(BCWriteCtx *ctx, Node *node, MSize nhash)
   TValue **heap = ctx->heap;
   MSize i = nhash;
   for (;; node--) {  /* Build heap. */
-    if (!tvisnil(&node->key)) {
+    if (!tvisnil(&node->val)) {
       bcwrite_ktabk_heap_insert(heap, --i, nhash, &node->key);
       if (i == 0) break;
     }
@@ -163,7 +165,7 @@ static void bcwrite_ktab(BCWriteCtx *ctx, char *p, const GCtab *t)
     MSize i, hmask = t->hmask;
     Node *node = noderef(t->node);
     for (i = 0; i <= hmask; i++)
-      nhash += !tvisnil(&node[i].key);
+      nhash += !tvisnil(&node[i].val);
   }
   /* Write number of array slots and hash slots. */
   p = lj_strfmt_wuleb128(p, narray);
@@ -184,7 +186,7 @@ static void bcwrite_ktab(BCWriteCtx *ctx, char *p, const GCtab *t)
     } else {
       MSize i = nhash;
       for (;; node--)
-	if (!tvisnil(&node->key)) {
+	if (!tvisnil(&node->val)) {
 	  bcwrite_ktabk(ctx, &node->key, 0);
 	  bcwrite_ktabk(ctx, &node->val, 1);
 	  if (--i == 0) break;

src/lj_ccallback.c

@@ -262,6 +262,14 @@ static void *callback_mcode_init(global_State *g, uint32_t *page)
 #define CCPROT_CREATE	0
 #endif
 
+/* Check for macOS hardened runtime. */
+#if LUAJIT_SECURITY_MCODE != 0 && defined(MAP_JIT) && __ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__ >= 110000
+#include <pthread.h>
+#define CCMAP_CREATE	MAP_JIT
+#else
+#define CCMAP_CREATE	0
+#endif
+
 #endif
 
 /* Allocate and initialize area for callback function pointers. */
@@ -276,10 +284,13 @@ static void callback_mcode_new(CTState *cts)
   if (!p)
     lj_err_caller(cts->L, LJ_ERR_FFI_CBACKOV);
 #elif LJ_TARGET_POSIX
-  p = mmap(NULL, sz, (PROT_READ|PROT_WRITE|CCPROT_CREATE), MAP_PRIVATE|MAP_ANONYMOUS,
+  p = mmap(NULL, sz, PROT_READ|PROT_WRITE|CCPROT_CREATE,
-	   -1, 0);
+	   MAP_PRIVATE|MAP_ANONYMOUS|CCMAP_CREATE, -1, 0);
   if (p == MAP_FAILED)
     lj_err_caller(cts->L, LJ_ERR_FFI_CBACKOV);
+#if CCMAP_CREATE
+  pthread_jit_write_protect_np(0);
+#endif
 #else
   /* Fallback allocator. Fails if memory is not executable by default. */
   p = lj_mem_new(cts->L, sz);
@@ -296,8 +307,12 @@ static void callback_mcode_new(CTState *cts)
     LJ_WIN_VPROTECT(p, sz, PAGE_EXECUTE_READ, &oprot);
   }
 #elif LJ_TARGET_POSIX
+#if CCMAP_CREATE
+  pthread_jit_write_protect_np(1);
+#else
   mprotect(p, sz, (PROT_READ|PROT_EXEC));
 #endif
+#endif
 }
 
 /* Free area for callback function pointers. */

src/lj_ctype.c

@@ -33,10 +33,12 @@
   _("int16_t",			INT16) \
   _("int32_t",			INT32) \
   _("int64_t",			INT64) \
+  _("int128_t",			INT128) \
   _("uint8_t",			UINT8) \
   _("uint16_t",			UINT16) \
   _("uint32_t",			UINT32) \
   _("uint64_t",			UINT64) \
+  _("uint128_t",		UINT128) \
   _("intptr_t",			INT_PSZ) \
   _("uintptr_t",		UINT_PSZ) \
   /* From POSIX. */ \
@@ -55,6 +57,7 @@
   _("__int16",		2,	CTOK_INT) \
   _("__int32",		4,	CTOK_INT) \
   _("__int64",		8,	CTOK_INT) \
+  _("__int128",		16,	CTOK_INT) \
   _("float",		4,	CTOK_FP) \
   _("double",		8,	CTOK_FP) \
   _("long",		0,	CTOK_LONG) \

src/lj_ctype.h

@@ -292,6 +292,8 @@ typedef struct CTState {
   _(UINT32,		4,	CT_NUM, CTF_UNSIGNED|CTALIGN(2)) \
   _(INT64,		8,	CT_NUM, CTF_LONG_IF8|CTALIGN(3)) \
   _(UINT64,		8,	CT_NUM, CTF_UNSIGNED|CTF_LONG_IF8|CTALIGN(3)) \
+  _(INT128,		16,	CT_NUM, CTALIGN(4)) \
+  _(UINT128,		16,	CT_NUM, CTF_UNSIGNED|CTALIGN(4)) \
   _(FLOAT,		4,	CT_NUM, CTF_FP|CTALIGN(2)) \
   _(DOUBLE,		8,	CT_NUM, CTF_FP|CTALIGN(3)) \
   _(COMPLEX_FLOAT,	8,	CT_ARRAY, CTF_COMPLEX|CTALIGN(2)|CTID_FLOAT) \

src/lj_jit.h

@@ -359,14 +359,10 @@ enum {
 
 enum {
 #if LJ_TARGET_X86ORX64
+  LJ_K64_2P32,		/* 2^32 */
   LJ_K64_TOBIT,		/* 2^52 + 2^51 */
   LJ_K64_2P64,		/* 2^64 */
   LJ_K64_M2P64,		/* -2^64 */
-#if LJ_32
-  LJ_K64_M2P64_31,	/* -2^64 or -2^31 */
-#else
-  LJ_K64_M2P64_31 = LJ_K64_M2P64,
-#endif
 #endif
 #if LJ_TARGET_MIPS
   LJ_K64_2P31,		/* 2^31 */
@@ -381,7 +377,8 @@ enum {
 
 enum {
 #if LJ_TARGET_X86ORX64
-  LJ_K32_M2P64_31,	/* -2^64 or -2^31 */
+  LJ_K32_2P32,		/* 2^32 */
+  LJ_K32_2P64,		/* 2^64 */
 #endif
 #if LJ_TARGET_PPC
   LJ_K32_2P52_2P31,	/* 2^52 + 2^31 */

src/lj_mcode.c

@@ -98,6 +98,14 @@ static int mcode_setprot(void *p, size_t sz, DWORD prot)
 #define MAP_ANONYMOUS	MAP_ANON
 #endif
 
+/* Check for macOS hardened runtime. */
+#if LUAJIT_SECURITY_MCODE != 0 && defined(MAP_JIT) && __ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__ >= 110000
+#include <pthread.h>
+#define MCMAP_CREATE	MAP_JIT
+#else
+#define MCMAP_CREATE	0
+#endif
+
 #define MCPROT_RW	(PROT_READ|PROT_WRITE)
 #define MCPROT_RX	(PROT_READ|PROT_EXEC)
 #define MCPROT_RWX	(PROT_READ|PROT_WRITE|PROT_EXEC)
@@ -109,10 +117,14 @@ static int mcode_setprot(void *p, size_t sz, DWORD prot)
 
 static void *mcode_alloc_at(jit_State *J, uintptr_t hint, size_t sz, int prot)
 {
-  void *p = mmap((void *)hint, sz, prot|MCPROT_CREATE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0);
+  void *p = mmap((void *)hint, sz, prot|MCPROT_CREATE, MAP_PRIVATE|MAP_ANONYMOUS|MCMAP_CREATE, -1, 0);
   if (p == MAP_FAILED) {
     if (!hint) lj_trace_err(J, LJ_TRERR_MCODEAL);
     p = NULL;
+#if MCMAP_CREATE
+  } else {
+    pthread_jit_write_protect_np(0);
+#endif
   }
   return p;
 }
@@ -125,7 +137,12 @@ static void mcode_free(jit_State *J, void *p, size_t sz)
 
 static int mcode_setprot(void *p, size_t sz, int prot)
 {
+#if MCMAP_CREATE
+  pthread_jit_write_protect_np((prot & PROT_EXEC));
+  return 0;
+#else
   return mprotect(p, sz, prot);
+#endif
 }
 
 #else

src/lj_opt_fold.c

@@ -2217,9 +2217,11 @@ LJFOLD(HREF TDUP KNUM)
 LJFOLDF(fwd_href_tdup)
 {
   TValue keyv;
+  cTValue *val;
   lj_ir_kvalue(J->L, &keyv, fright);
-  if (lj_tab_get(J->L, ir_ktab(IR(fleft->op1)), &keyv) == niltvg(J2G(J)) &&
+  val = lj_tab_get(J->L, ir_ktab(IR(fleft->op1)), &keyv);
-      lj_opt_fwd_href_nokey(J))
+  /* Check for either nil or the nil value marker in the template table. */
+  if ((tvisnil(val) || tvistab(val)) && lj_opt_fwd_href_nokey(J))
     return lj_ir_kkptr(J, niltvg(J2G(J)));
   return NEXTFOLD;
 }

src/lj_opt_mem.c

@@ -233,7 +233,9 @@ static TRef fwd_ahload(jit_State *J, IRRef xref)
 	  return lj_ir_knum_u64(J, tv->u64);
 	else if (tvisint(tv))
 	  return lj_ir_kint(J, intV(tv));
-	else if (tvisgcv(tv))
+	else if (tvistab(tv)) /* Template table nil value marker. */
+	  return TREF_NIL;
+	else if (tvisstr(tv))
 	  return lj_ir_kstr(J, strV(tv));
       }
       /* Othwerwise: don't intern as a constant. */

src/lj_parse.c

@@ -1725,7 +1725,7 @@ static void expr_table(LexState *ls, ExpDesc *e)
   FuncState *fs = ls->fs;
   BCLine line = ls->linenumber;
   GCtab *t = NULL;
-  int vcall = 0, needarr = 0, fixt = 0;
+  int vcall = 0, needarr = 0;
   uint32_t narr = 1;  /* First array index. */
   uint32_t nhash = 0;  /* Number of hash entries. */
   BCReg freg = fs->freereg;
@@ -1769,9 +1769,10 @@ static void expr_table(LexState *ls, ExpDesc *e)
       lj_gc_anybarriert(fs->L, t);
       if (expr_isk_nojump(&val)) {  /* Add const key/value to template table. */
 	expr_kvalue(fs, v, &val);
-      } else {  /* Otherwise create dummy string key (avoids lj_tab_newkey). */
+	/* Mark nil value with table value itself to preserve the key. */
-	settabV(fs->L, v, t);  /* Preserve key with table itself as value. */
+	if (key.k == VKSTR && tvisnil(v)) settabV(fs->L, v, t);
-	fixt = 1;   /* Fix this later, after all resizes. */
+      } else {  /* Preserve the key for the following non-const store.  */
+	settabV(fs->L, v, t);
 	goto nonconst;
       }
     } else {
@@ -1813,17 +1814,6 @@ static void expr_table(LexState *ls, ExpDesc *e)
   } else {
     if (needarr && t->asize < narr)
       lj_tab_reasize(fs->L, t, narr-1);
-    if (fixt) {  /* Fix value for dummy keys in template table. */
-      Node *node = noderef(t->node);
-      uint32_t i, hmask = t->hmask;
-      for (i = 0; i <= hmask; i++) {
-	Node *n = &node[i];
-	if (tvistab(&n->val)) {
-	  lj_assertFS(tabV(&n->val) == t, "bad dummy key in template table");
-	  setnilV(&n->val);  /* Turn value into nil. */
-	}
-      }
-    }
     lj_gc_check(fs->L);
   }
 }

src/lj_record.c

@@ -2079,6 +2079,7 @@ static TRef rec_tnew(jit_State *J, uint32_t ah)
 /* -- Concatenation ------------------------------------------------------- */
 
 typedef struct RecCatDataCP {
+  TValue savetv[5+LJ_FR2];
   jit_State *J;
   BCReg baseslot, topslot;
   TRef tr;
@@ -2119,7 +2120,9 @@ static TValue *rec_mm_concat_cp(lua_State *L, lua_CFunction dummy, void *ud)
       return NULL;
     }
     /* Pass partial result. */
-    topslot = J->maxslot--;
+    rcd->topslot = topslot = J->maxslot--;
+    /* Save updated range of slots. */
+    memcpy(rcd->savetv, &L->base[topslot-1], sizeof(rcd->savetv));
     *xbase = tr;
     top = xbase;
     setstrV(J->L, &ix.keyv, &J2G(J)->strempty);  /* Simulate string result. */
@@ -2139,16 +2142,18 @@ static TRef rec_cat(jit_State *J, BCReg baseslot, BCReg topslot)
 {
   lua_State *L = J->L;
   ptrdiff_t delta = L->top - L->base;
-  TValue savetv[5+LJ_FR2], errobj;
+  TValue errobj;
   RecCatDataCP rcd;
   int errcode;
   rcd.J = J;
   rcd.baseslot = baseslot;
   rcd.topslot = topslot;
-  memcpy(savetv, &L->base[topslot-1], sizeof(savetv));  /* Save slots. */
+  /* Save slots. */
+  memcpy(rcd.savetv, &L->base[topslot-1], sizeof(rcd.savetv));
   errcode = lj_vm_cpcall(L, NULL, &rcd, rec_mm_concat_cp);
   if (errcode) copyTV(L, &errobj, L->top-1);
-  memcpy(&L->base[topslot-1], savetv, sizeof(savetv));  /* Restore slots. */
+  /* Restore slots. */
+  memcpy(&L->base[rcd.topslot-1], rcd.savetv, sizeof(rcd.savetv));
   if (errcode) {
     L->top = L->base + delta;
     copyTV(L, L->top++, &errobj);

src/lj_tab.c

@@ -194,6 +194,7 @@ GCtab * LJ_FASTCALL lj_tab_dup(lua_State *L, const GCtab *kt)
       Node *next = nextnode(kn);
       /* Don't use copyTV here, since it asserts on a copy of a dead key. */
       n->val = kn->val; n->key = kn->key;
+      if (tvistab(&n->val)) setnilV(&n->val); /* Replace nil value marker. */
       setmref(n->next, next == NULL? next : (Node *)((char *)next + d));
     }
   }

src/lj_target_x86.h

@@ -300,6 +300,7 @@ typedef enum {
   XO_CVTSS2SD =	XO_f30f(5a),
   XO_CVTSD2SS =	XO_f20f(5a),
   XO_ADDSS =	XO_f30f(58),
+  XO_SUBSS =	XO_f30f(5c),
   XO_MOVD =	XO_660f(6e),
   XO_MOVDto =	XO_660f(7e),
 

src/lj_trace.c

@@ -222,14 +222,6 @@ static void trace_unpatch(jit_State *J, GCtrace *T)
 	       bc_isret(op), "bad original bytecode %d", op);
     *pc = T->startins;
     break;
-  case BC_JMP:
-    lj_assertJ(op == BC_ITERL, "bad original bytecode %d", op);
-    pc += bc_j(*pc)+2;
-    if (bc_op(*pc) == BC_JITERL) {
-      lj_assertJ(traceref(J, bc_d(*pc)) == T, "JITERL references other trace");
-      *pc = T->startins;
-    }
-    break;
   case BC_JFUNCF:
     lj_assertJ(op == BC_FUNCF, "bad original bytecode %d", op);
     *pc = T->startins;
@@ -245,18 +237,19 @@ static void trace_flushroot(jit_State *J, GCtrace *T)
   GCproto *pt = &gcref(T->startpt)->pt;
   lj_assertJ(T->root == 0, "not a root trace");
   lj_assertJ(pt != NULL, "trace has no prototype");
-  /* First unpatch any modified bytecode. */
-  trace_unpatch(J, T);
   /* Unlink root trace from chain anchored in prototype. */
   if (pt->trace == T->traceno) {  /* Trace is first in chain. Easy. */
     pt->trace = T->nextroot;
+unpatch:
+    /* Unpatch modified bytecode only if the trace has not been flushed. */
+    trace_unpatch(J, T);
   } else if (pt->trace) {  /* Otherwise search in chain of root traces. */
     GCtrace *T2 = traceref(J, pt->trace);
     if (T2) {
       for (; T2->nextroot; T2 = traceref(J, T2->nextroot))
 	if (T2->nextroot == T->traceno) {
 	  T2->nextroot = T->nextroot;  /* Unlink from chain. */
-	  break;
+	  goto unpatch;
 	}
     }
   }
@@ -325,12 +318,11 @@ void lj_trace_initstate(global_State *g)
 
   /* Initialize 32/64 bit constants. */
 #if LJ_TARGET_X86ORX64
+  J->k64[LJ_K64_2P32].u64 = U64x(41f00000,00000000);
   J->k64[LJ_K64_TOBIT].u64 = U64x(43380000,00000000);
-#if LJ_32
-  J->k64[LJ_K64_M2P64_31].u64 = U64x(c1e00000,00000000);
-#endif
   J->k64[LJ_K64_2P64].u64 = U64x(43f00000,00000000);
-  J->k32[LJ_K32_M2P64_31] = LJ_64 ? 0xdf800000 : 0xcf000000;
+  J->k32[LJ_K32_2P64] = 0x5f800000;
+  J->k32[LJ_K32_2P32] = 0x4f800000;
 #endif
 #if LJ_TARGET_X86ORX64 || LJ_TARGET_MIPS64
   J->k64[LJ_K64_M2P64].u64 = U64x(c3f00000,00000000);

src/luajit.c

@@ -35,6 +35,21 @@
 
 #if !LJ_TARGET_CONSOLE
 #include <signal.h>
+
+#if LJ_TARGET_POSIX
+/* Improve signal handling on POSIX. Try CTRL-C on: luajit -e 'io.read()' */
+static void signal_set(int sig, void (*h)(int))
+{
+  struct sigaction sa;
+  memset(&sa, 0, sizeof(sa));
+  sa.sa_handler = h;
+  sigemptyset(&sa.sa_mask);
+  sigaction(sig, &sa, NULL);
+}
+#else
+#define signal_set	signal
+#endif
+
 #endif
 
 static lua_State *globalL = NULL;
@@ -54,8 +69,8 @@ static void lstop(lua_State *L, lua_Debug *ar)
 
 static void laction(int i)
 {
-  signal(i, SIG_DFL); /* if another SIGINT happens before lstop,
+  /* Terminate process if another SIGINT happens (double CTRL-C). */
-			 terminate process (default action) */
+  signal_set(i, SIG_DFL);
   lua_sethook(globalL, lstop, LUA_MASKCALL | LUA_MASKRET | LUA_MASKCOUNT, 1);
 }
 #endif
@@ -117,11 +132,11 @@ static int docall(lua_State *L, int narg, int clear)
   lua_pushcfunction(L, traceback);  /* push traceback function */
   lua_insert(L, base);  /* put it under chunk and args */
 #if !LJ_TARGET_CONSOLE
-  signal(SIGINT, laction);
+  signal_set(SIGINT, laction);
 #endif
   status = lua_pcall(L, narg, (clear ? 0 : LUA_MULTRET), base);
 #if !LJ_TARGET_CONSOLE
-  signal(SIGINT, SIG_DFL);
+  signal_set(SIGINT, SIG_DFL);
 #endif
   lua_remove(L, base);  /* remove traceback function */
   /* force a complete garbage collection in case of errors */

src/msvcbuild.bat

@@ -8,7 +8,8 @@
 @rem   nogc64   disable LJ_GC64 mode for x64
 @rem   debug    emit debug symbols
 @rem   amalg    amalgamated build
-@rem   static   static linkage
+@rem   static   create static lib to statically link into your project
+@rem   mixed    create static lib to build a DLL in your project
 
 @if not defined INCLUDE goto :FAIL
 
@@ -106,12 +107,14 @@ buildvm -m folddef -o lj_folddef.h lj_opt_fold.c
 @if "%1"=="static" goto :STATIC
 %LJCOMPILE% %LJDYNBUILD% lj_*.c lib_*.c
 @if errorlevel 1 goto :BAD
+@if "%1"=="mixed" goto :STATICLIB
 %LJLINK% /DLL /OUT:%LJDLLNAME% lj_*.obj lib_*.obj
 @if errorlevel 1 goto :BAD
 @goto :MTDLL
 :STATIC
 %LJCOMPILE% lj_*.c lib_*.c
 @if errorlevel 1 goto :BAD
+:STATICLIB
 %LJLIB% /OUT:%LJLIBNAME% lj_*.obj lib_*.obj
 @if errorlevel 1 goto :BAD
 @goto :MTDLL
@@ -119,13 +122,15 @@ buildvm -m folddef -o lj_folddef.h lj_opt_fold.c
 @if "%2"=="static" goto :AMALGSTATIC
 %LJCOMPILE% %LJDYNBUILD% ljamalg.c
 @if errorlevel 1 goto :BAD
+@if "%2"=="mixed" goto :AMALGSTATICLIB
 %LJLINK% /DLL /OUT:%LJDLLNAME% ljamalg.obj lj_vm.obj
 @if errorlevel 1 goto :BAD
 @goto :MTDLL
 :AMALGSTATIC
 %LJCOMPILE% ljamalg.c
 @if errorlevel 1 goto :BAD
-%LJLINK% /OUT:%LJDLLNAME% ljamalg.obj lj_vm.obj
+:AMALGSTATICLIB
+%LJLIB% /OUT:%LJLIBNAME% ljamalg.obj lj_vm.obj
 @if errorlevel 1 goto :BAD
 :MTDLL
 if exist %LJDLLNAME%.manifest^