Squashed 'luajit-2.0/' changes from 8b6af89..8715ae6

8715ae6 Properly handle OOM in trace_save().
424940f Fix NYICF error message.
a5b1c4d FFI: Don't propagate qualifiers into subtypes of complex.
575bfe5 Limit number of arguments given to io.lines() and fp:lines().
7f454ae Fix narrowing of TOBIT.
7234789 x64: Allow building without external unwinder.
69e5342 RELEASE LuaJIT-2.0.4
cd4c59f Update changelog.
9622d6b x86/x64: Fix argument check for bit shifts.
11106aa Fix stack check in narrowing optimization.
a9fd686 Doc cleanup.

git-subtree-dir: luajit-2.0
git-subtree-split: 8715ae65c1f70f2402b89a004b88ab04ca103625

24 files changed, 153 insertions, 55 deletions

diff --git a/Makefile b/Makefile
index 0891b71..0cbe741 100644
--- a/Makefile
+++ b/Makefile
@@ -15,7 +15,7 @@
 
 MAJVER=  2
 MINVER=  0
-RELVER=  3
+RELVER=  4
 VERSION= $(MAJVER).$(MINVER).$(RELVER)
 ABIVER=  5.1
 
diff --git a/README b/README
index d151bb7..44366af 100644
--- a/README
+++ b/README
@@ -1,4 +1,4 @@
-README for LuaJIT 2.0.3
+README for LuaJIT 2.0.4
 -----------------------
 
 LuaJIT is a Just-In-Time (JIT) compiler for the Lua programming language.
diff --git a/doc/changes.html b/doc/changes.html
index 3d8f5db..d7b83ce 100644
--- a/doc/changes.html
+++ b/doc/changes.html
@@ -63,7 +63,7 @@ div.major { max-width: 600px; padding: 1em; margin: 1em 0 1em 0; }
 <div id="main">
 <p>
 This is a list of changes between the released versions of LuaJIT.<br>
-The current <span style="color: #0000c0;">stable version</span> is <strong>LuaJIT&nbsp;2.0.3</strong>.<br>
+The current <span style="color: #0000c0;">stable version</span> is <strong>LuaJIT&nbsp;2.0.4</strong>.<br>
 </p>
 <p>
 Please check the
@@ -72,6 +72,56 @@ to see whether newer versions are available.
 </p>
 
 <div class="major" style="background: #d0d0ff;">
+<h2 id="LuaJIT-2.0.4">LuaJIT 2.0.4 &mdash; 2015-05-14</h2>
+<ul>
+<li>Fix stack check in narrowing optimization.</li>
+<li>Fix Lua/C API typecheck error for special indexes.</li>
+<li>Fix string to number conversion.</li>
+<li>Fix lexer error for chunks without tokens.</li>
+<li>Don't compile <tt>IR_RETF</tt> after <tt>CALLT</tt> to ff with-side effects.</li>
+<li>Fix <tt>BC_UCLO</tt>/<tt>BC_JMP</tt> join optimization in Lua parser.</li>
+<li>Fix corner case in string to number conversion.</li>
+<li>Gracefully handle <tt>lua_error()</tt> for a suspended coroutine.</li>
+<li>Avoid error messages when building with Clang.</li>
+<li>Fix snapshot #0 handling for traces with a stack check on entry.</li>
+<li>Fix fused constant loads under high register pressure.</li>
+<li>Invalidate backpropagation cache after DCE.</li>
+<li>Fix ABC elimination.</li>
+<li>Fix debug info for main chunk of stripped bytecode.</li>
+<li>Fix FOLD rule for <tt>string.sub(s, ...) == k</tt>.</li>
+<li>Fix FOLD rule for <tt>STRREF</tt> of <tt>SNEW</tt>.</li>
+<li>Fix frame traversal while searching for error function.</li>
+<li>Prevent GC estimate miscalculation due to buffer growth.</li>
+<li>Prevent adding side traces for stack checks.</li>
+<li>Fix top slot calculation for snapshots with continuations.</li>
+<li>Fix check for reuse of SCEV results in <tt>FORL</tt>.</li>
+<li>Add PS Vita port.</li>
+<li>Fix compatibility issues with Illumos.</li>
+<li>Fix DragonFly build (unsupported).</li>
+<li>OpenBSD/x86: Better executable memory allocation for W^X mode.</li>
+<li>x86: Fix argument checks for <tt>ipairs()</tt> iterator.</li>
+<li>x86: <tt>lj_math_random_step()</tt> clobbers XMM regs on OSX Clang.</li>
+<li>x86: Fix code generation for unused result of <tt>math.random()</tt>.</li>
+<li>x64: Allow building with <tt>LUAJIT_USE_SYSMALLOC</tt> and <tt>LUAJIT_USE_VALGRIND</tt>.</li>
+<li>x86/x64: Fix argument check for bit shifts.</li>
+<li>x86/x64: Fix code generation for fused test/arith ops.</li>
+<li>ARM: Fix write barrier check in <tt>BC_USETS</tt>.</li>
+<li>PPC: Fix red zone overflow in machine code generation.</li>
+<li>PPC: Don't use <tt>mcrxr</tt> on PPE.</li>
+<li>Various archs: Fix excess stack growth in interpreter.</li>
+<li>FFI: Fix FOLD rule for <tt>TOBIT</tt> + <tt>CONV num.u32</tt>.</li>
+<li>FFI: Prevent DSE across <tt>ffi.string()</tt>.</li>
+<li>FFI: No meta fallback when indexing pointer to incomplete struct.</li>
+<li>FFI: Fix initialization of unions of subtypes.</li>
+<li>FFI: Fix cdata vs. non-cdata arithmetic and comparisons.</li>
+<li>FFI: Fix <tt>__index</tt>/<tt>__newindex</tt> metamethod resolution for ctypes.</li>
+<li>FFI: Fix compilation of reference field access.</li>
+<li>FFI: Fix frame traversal for backtraces with FFI callbacks.</li>
+<li>FFI: Fix recording of indexing a struct pointer ctype object itself.</li>
+<li>FFI: Allow non-scalar cdata to be compared for equality by address.</li>
+<li>FFI: Fix pseudo type conversions for type punning.</li>
+</ul>
+
 <h2 id="LuaJIT-2.0.3">LuaJIT 2.0.3 &mdash; 2014-03-12</h2>
 <ul>
 <li>Add PS4 port.</li>
diff --git a/doc/install.html b/doc/install.html
index 7369aeb..7a878b1 100644
--- a/doc/install.html
+++ b/doc/install.html
@@ -188,8 +188,8 @@ open a terminal window and change to this directory. Now unpack the archive
 and change to the newly created directory:
 </p>
 <pre class="code">
-tar zxf LuaJIT-2.0.3.tar.gz
-cd LuaJIT-2.0.3</pre>
+tar zxf LuaJIT-2.0.4.tar.gz
+cd LuaJIT-2.0.4</pre>
 <h3>Building LuaJIT</h3>
 <p>
 The supplied Makefiles try to auto-detect the settings needed for your
diff --git a/doc/running.html b/doc/running.html
index c2ec511..c6e1c29 100644
--- a/doc/running.html
+++ b/doc/running.html
@@ -186,7 +186,7 @@ itself. For a description of their options and output format, please
 read the comment block at the start of their source.
 They can be found in the <tt>lib</tt> directory of the source
 distribution or installed under the <tt>jit</tt> directory. By default
-this is <tt>/usr/local/share/luajit-2.0.3/jit</tt> on POSIX
+this is <tt>/usr/local/share/luajit-2.0.4/jit</tt> on POSIX
 systems.
 </p>
 
diff --git a/doc/status.html b/doc/status.html
index e3655d4..b4bbec7 100644
--- a/doc/status.html
+++ b/doc/status.html
@@ -1,7 +1,7 @@
 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN" "http://www.w3.org/TR/html4/strict.dtd">
 <html>
 <head>
-<title>Status &amp; Roadmap</title>
+<title>Status</title>
 <meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">
 <meta name="Author" content="Mike Pall">
 <meta name="Copyright" content="Copyright (C) 2005-2015, Mike Pall">
@@ -17,7 +17,7 @@ ul li { padding-bottom: 0.3em; }
 <a href="http://luajit.org"><span>Lua<span id="logo">JIT</span></span></a>
 </div>
 <div id="head">
-<h1>Status &amp; Roadmap</h1>
+<h1>Status</h1>
 </div>
 <div id="nav">
 <ul><li>
@@ -102,15 +102,6 @@ handled correctly. The error may fall through an on-trace
 garbage collector.
 </li>
 </ul>
-
-<h2>Roadmap</h2>
-<p>
-Please refer to the
-<a href="http://www.freelists.org/post/luajit/LuaJIT-Roadmap-20122013"><span class="ext">&raquo;</span>&nbsp;LuaJIT Roadmap 2012/2013</a> and an
-<a href="http://www.freelists.org/post/luajit/LuaJIT-Roadmap-20122013-UPDATE"><span class="ext">&raquo;</span>&nbsp;update on release planning</a> for details.
-</p>
-<p>
-</p>
 <br class="flush">
 </div>
 <div id="foot">
diff --git a/etc/luajit.pc b/etc/luajit.pc
index d3f0cf3..a652b40 100644
--- a/etc/luajit.pc
+++ b/etc/luajit.pc
@@ -1,7 +1,7 @@
 # Package information for LuaJIT to be used by pkg-config.
 majver=2
 minver=0
-relver=3
+relver=4
 version=${majver}.${minver}.${relver}
 abiver=5.1
 
diff --git a/src/Makefile b/src/Makefile
index 33b0a43..5021e47 100644
--- a/src/Makefile
+++ b/src/Makefile
@@ -12,7 +12,7 @@
 
 MAJVER=  2
 MINVER=  0
-RELVER=  3
+RELVER=  4
 ABIVER=  5.1
 NODOTABIVER= 51
 
@@ -243,9 +243,6 @@ ifneq (,$(findstring LJ_TARGET_PS3 1,$(TARGET_TESTARCH)))
   TARGET_ARCH+= -D__CELLOS_LV2__
   TARGET_XCFLAGS+= -DLUAJIT_USE_SYSMALLOC
 endif
-ifneq (,$(findstring LJ_NO_UNWIND 1,$(TARGET_TESTARCH)))
-  TARGET_ARCH+= -DLUAJIT_NO_UNWIND
-endif
 
 TARGET_XCFLAGS+= $(CCOPT_$(TARGET_LJARCH))
 TARGET_ARCH+= $(patsubst %,-DLUAJIT_TARGET=LUAJIT_ARCH_%,$(TARGET_LJARCH))
@@ -397,6 +394,10 @@ ifeq (,$(findstring LJ_ABI_SOFTFP 1,$(TARGET_TESTARCH)))
 else
   TARGET_ARCH+= -DLJ_ABI_SOFTFP=1
 endif
+ifneq (,$(findstring LJ_NO_UNWIND 1,$(TARGET_TESTARCH)))
+  DASM_AFLAGS+= -D NO_UNWIND
+  TARGET_ARCH+= -DLUAJIT_NO_UNWIND
+endif
 DASM_AFLAGS+= -D VER=$(subst LJ_ARCH_VERSION_,,$(filter LJ_ARCH_VERSION_%,$(subst LJ_ARCH_VERSION ,LJ_ARCH_VERSION_,$(TARGET_TESTARCH))))
 ifeq (Windows,$(TARGET_SYS))
   DASM_AFLAGS+= -D WIN
diff --git a/src/jit/bc.lua b/src/jit/bc.lua
index a179d50..46a4089 100644
--- a/src/jit/bc.lua
+++ b/src/jit/bc.lua
@@ -41,7 +41,7 @@
 
 -- Cache some library functions and objects.
 local jit = require("jit")
-assert(jit.version_num == 20003, "LuaJIT core/library version mismatch")
+assert(jit.version_num == 20004, "LuaJIT core/library version mismatch")
 local jutil = require("jit.util")
 local vmdef = require("jit.vmdef")
 local bit = require("bit")
diff --git a/src/jit/bcsave.lua b/src/jit/bcsave.lua
index 2ba234d..0319b3d 100644
--- a/src/jit/bcsave.lua
+++ b/src/jit/bcsave.lua
@@ -11,7 +11,7 @@
 ------------------------------------------------------------------------------
 
 local jit = require("jit")
-assert(jit.version_num == 20003, "LuaJIT core/library version mismatch")
+assert(jit.version_num == 20004, "LuaJIT core/library version mismatch")
 local bit = require("bit")
 
 -- Symbol name prefix for LuaJIT bytecode.
diff --git a/src/jit/dump.lua b/src/jit/dump.lua
index 589543f..d15c528 100644
--- a/src/jit/dump.lua
+++ b/src/jit/dump.lua
@@ -55,7 +55,7 @@
 
 -- Cache some library functions and objects.
 local jit = require("jit")
-assert(jit.version_num == 20003, "LuaJIT core/library version mismatch")
+assert(jit.version_num == 20004, "LuaJIT core/library version mismatch")
 local jutil = require("jit.util")
 local vmdef = require("jit.vmdef")
 local funcinfo, funcbc = jutil.funcinfo, jutil.funcbc
diff --git a/src/jit/v.lua b/src/jit/v.lua
index c622443..32666fd 100644
--- a/src/jit/v.lua
+++ b/src/jit/v.lua
@@ -59,7 +59,7 @@
 
 -- Cache some library functions and objects.
 local jit = require("jit")
-assert(jit.version_num == 20003, "LuaJIT core/library version mismatch")
+assert(jit.version_num == 20004, "LuaJIT core/library version mismatch")
 local jutil = require("jit.util")
 local vmdef = require("jit.vmdef")
 local funcinfo, traceinfo = jutil.funcinfo, jutil.traceinfo
diff --git a/src/lib_io.c b/src/lib_io.c
index 037aa28..a502dd5 100644
--- a/src/lib_io.c
+++ b/src/lib_io.c
@@ -278,6 +278,15 @@ static int io_file_iter(lua_State *L)
   return n;
 }
 
+static int io_file_lines(lua_State *L)
+{
+  int n = (int)(L->top - L->base);
+  if (n > LJ_MAX_UPVAL)
+    lj_err_caller(L, LJ_ERR_UNPACK);
+  lua_pushcclosure(L, io_file_iter, n);
+  return 1;
+}
+
 /* -- I/O file methods ---------------------------------------------------- */
 
 #define LJLIB_MODULE_io_method
@@ -361,8 +370,7 @@ LJLIB_CF(io_method_setvbuf)
 LJLIB_CF(io_method_lines)
 {
   io_tofile(L);
-  lua_pushcclosure(L, io_file_iter, (int)(L->top - L->base));
-  return 1;
+  return io_file_lines(L);
 }
 
 LJLIB_CF(io_method___gc)
@@ -492,8 +500,7 @@ LJLIB_CF(io_lines)
   } else {  /* io.lines() iterates over stdin. */
     setudataV(L, L->base, IOSTDF_UD(L, GCROOT_IO_INPUT));
   }
-  lua_pushcclosure(L, io_file_iter, (int)(L->top - L->base));
-  return 1;
+  return io_file_lines(L);
 }
 
 LJLIB_CF(io_type)
diff --git a/src/lj_arch.h b/src/lj_arch.h
index d3a9d57..d89d116 100644
--- a/src/lj_arch.h
+++ b/src/lj_arch.h
@@ -423,7 +423,7 @@
 #define LUAJIT_NO_EXP2
 #endif
 
-#if defined(LUAJIT_NO_UNWIND) || defined(__symbian__) || LJ_TARGET_IOS || LJ_TARGET_PS3
+#if defined(LUAJIT_NO_UNWIND) || defined(__symbian__) || LJ_TARGET_IOS || LJ_TARGET_PS3 || LJ_TARGET_PS4
 #define LJ_NO_UNWIND		1
 #endif
 
diff --git a/src/lj_cparse.c b/src/lj_cparse.c
index b9df88d..9e21361 100644
--- a/src/lj_cparse.c
+++ b/src/lj_cparse.c
@@ -798,6 +798,10 @@ static void cp_push_type(CPDecl *decl, CTypeID id)
     cp_push(decl, info & ~CTMASK_CID, size);  /* Copy type. */
     break;
   case CT_ARRAY:
+    if ((ct->info & (CTF_VECTOR|CTF_COMPLEX))) {
+      info |= (decl->attr & CTF_QUAL);
+      decl->attr &= ~CTF_QUAL;
+    }
     cp_push_type(decl, ctype_cid(info));  /* Unroll. */
     cp_push(decl, info & ~CTMASK_CID, size);  /* Copy type. */
     decl->stack[decl->pos].sib = 1;  /* Mark as already checked and sized. */
diff --git a/src/lj_err.c b/src/lj_err.c
index 081bfde..11b07b5 100644
--- a/src/lj_err.c
+++ b/src/lj_err.c
@@ -57,10 +57,10 @@
 ** EXT cannot be enabled on WIN32 since system exceptions use code-driven SEH.
 ** EXT is mandatory on WIN64 since the calling convention has an abundance
 ** of callee-saved registers (rbx, rbp, rsi, rdi, r12-r15, xmm6-xmm15).
-** EXT is mandatory on POSIX/x64 since the interpreter doesn't save r12/r13.
+** The POSIX/x64 interpreter only saves r12/r13 for INT (e.g. PS4).
 */
 
-#if defined(__GNUC__) && (LJ_TARGET_X64 || defined(LUAJIT_UNWIND_EXTERNAL))
+#if defined(__GNUC__) && (LJ_TARGET_X64 || defined(LUAJIT_UNWIND_EXTERNAL)) && !LJ_NO_UNWIND
 #define LJ_UNWIND_EXT	1
 #elif LJ_TARGET_X64 && LJ_TARGET_WINDOWS
 #define LJ_UNWIND_EXT	1
diff --git a/src/lj_frame.h b/src/lj_frame.h
index cd57be2..f981098 100644
--- a/src/lj_frame.h
+++ b/src/lj_frame.h
@@ -86,7 +86,11 @@ enum {
 #define CFRAME_OFS_ERRF		(5*4)
 #define CFRAME_OFS_NRES		(4*4)
 #define CFRAME_OFS_MULTRES	(1*4)
+#if LJ_NO_UNWIND
+#define CFRAME_SIZE		(12*8)
+#else
 #define CFRAME_SIZE		(10*8)
+#endif
 #define CFRAME_SIZE_JIT		(CFRAME_SIZE + 16)
 #define CFRAME_SHIFT_MULTRES	0
 #endif
diff --git a/src/lj_opt_narrow.c b/src/lj_opt_narrow.c
index b7bd323..fb64718 100644
--- a/src/lj_opt_narrow.c
+++ b/src/lj_opt_narrow.c
@@ -205,7 +205,6 @@ typedef struct NarrowConv {
   jit_State *J;		/* JIT compiler state. */
   NarrowIns *sp;	/* Current stack pointer. */
   NarrowIns *maxsp;	/* Maximum stack pointer minus redzone. */
-  int lim;		/* Limit on the number of emitted conversions. */
   IRRef mode;		/* Conversion mode (IRCONV_*). */
   IRType t;		/* Destination type: IRT_INT or IRT_I64. */
   NarrowIns stack[NARROW_MAX_STACK];  /* Stack holding stack-machine code. */
@@ -247,10 +246,16 @@ static void narrow_stripov_backprop(NarrowConv *nc, IRRef ref, int depth)
     if (bp) {
       ref = bp->val;
     } else if (++depth < NARROW_MAX_BACKPROP && nc->sp < nc->maxsp) {
+      NarrowIns *savesp = nc->sp;
       narrow_stripov_backprop(nc, ir->op1, depth);
-      narrow_stripov_backprop(nc, ir->op2, depth);
-      *nc->sp++ = NARROWINS(IRT(ir->o - IR_ADDOV + IR_ADD, IRT_INT), ref);
-      return;
+      if (nc->sp < nc->maxsp) {
+	narrow_stripov_backprop(nc, ir->op2, depth);
+	if (nc->sp < nc->maxsp) {
+	  *nc->sp++ = NARROWINS(IRT(ir->o - IR_ADDOV + IR_ADD, IRT_INT), ref);
+	  return;
+	}
+      }
+      nc->sp = savesp;  /* Path too deep, need to backtrack. */
     }
   }
   *nc->sp++ = NARROWINS(NARROW_REF, ref);
@@ -263,6 +268,8 @@ static int narrow_conv_backprop(NarrowConv *nc, IRRef ref, int depth)
   IRIns *ir = IR(ref);
   IRRef cref;
 
+  if (nc->sp >= nc->maxsp) return 10;  /* Path too deep. */
+
   /* Check the easy cases first. */
   if (ir->o == IR_CONV && (ir->op2 & IRCONV_SRCMASK) == IRT_INT) {
     if ((nc->mode & IRCONV_CONVMASK) <= IRCONV_ANY)
@@ -334,7 +341,7 @@ static int narrow_conv_backprop(NarrowConv *nc, IRRef ref, int depth)
       NarrowIns *savesp = nc->sp;
       int count = narrow_conv_backprop(nc, ir->op1, depth);
       count += narrow_conv_backprop(nc, ir->op2, depth);
-      if (count <= nc->lim) {  /* Limit total number of conversions. */
+      if (count <= 1) {  /* Limit total number of conversions. */
 	*nc->sp++ = NARROWINS(IRT(ir->o, nc->t), ref);
 	return count;
       }
@@ -406,12 +413,10 @@ TRef LJ_FASTCALL lj_opt_narrow_convert(jit_State *J)
     nc.t = irt_type(fins->t);
     if (fins->o == IR_TOBIT) {
       nc.mode = IRCONV_TOBIT;  /* Used only in the backpropagation cache. */
-      nc.lim = 2;  /* TOBIT can use a more optimistic rule. */
     } else {
       nc.mode = fins->op2;
-      nc.lim = 1;
     }
-    if (narrow_conv_backprop(&nc, fins->op1, 0) <= nc.lim)
+    if (narrow_conv_backprop(&nc, fins->op1, 0) <= 1)
       return narrow_conv_emit(J, &nc);
   }
   return NEXTFOLD;
diff --git a/src/lj_trace.c b/src/lj_trace.c
index e51ec54..7d1b093 100644
--- a/src/lj_trace.c
+++ b/src/lj_trace.c
@@ -117,15 +117,22 @@ static void perftools_addtrace(GCtrace *T)
 }
 #endif
 
-/* Save current trace by copying and compacting it. */
-static void trace_save(jit_State *J)
+/* Allocate space for copy of trace. */
+static GCtrace *trace_save_alloc(jit_State *J)
 {
   size_t sztr = ((sizeof(GCtrace)+7)&~7);
   size_t szins = (J->cur.nins-J->cur.nk)*sizeof(IRIns);
   size_t sz = sztr + szins +
 	      J->cur.nsnap*sizeof(SnapShot) +
 	      J->cur.nsnapmap*sizeof(SnapEntry);
-  GCtrace *T = lj_mem_newt(J->L, (MSize)sz, GCtrace);
+  return lj_mem_newt(J->L, (MSize)sz, GCtrace);
+}
+
+/* Save current trace by copying and compacting it. */
+static void trace_save(jit_State *J, GCtrace *T)
+{
+  size_t sztr = ((sizeof(GCtrace)+7)&~7);
+  size_t szins = (J->cur.nins-J->cur.nk)*sizeof(IRIns);
   char *p = (char *)T + sztr;
   memcpy(T, &J->cur, sizeof(GCtrace));
   setgcrefr(T->nextgc, J2G(J)->gc.root);
@@ -417,6 +424,7 @@ static void trace_stop(jit_State *J)
   BCOp op = bc_op(J->cur.startins);
   GCproto *pt = &gcref(J->cur.startpt)->pt;
   TraceNo traceno = J->cur.traceno;
+  GCtrace *T = trace_save_alloc(J);  /* Do this first. May throw OOM. */
   lua_State *L;
 
   switch (op) {
@@ -461,7 +469,7 @@ static void trace_stop(jit_State *J)
   /* Commit new mcode only after all patching is done. */
   lj_mcode_commit(J, J->cur.mcode);
   J->postproc = LJ_POST_NONE;
-  trace_save(J);
+  trace_save(J, T);
 
   L = J->L;
   lj_vmevent_send(L, TRACE,
diff --git a/src/lj_traceerr.h b/src/lj_traceerr.h
index 9bef117..150ac2d 100644
--- a/src/lj_traceerr.h
+++ b/src/lj_traceerr.h
@@ -23,7 +23,7 @@ TREDEF(BADTYPE,	"bad argument type")
 TREDEF(CJITOFF,	"JIT compilation disabled for function")
 TREDEF(CUNROLL,	"call unroll limit reached")
 TREDEF(DOWNREC,	"down-recursion, restarting")
-TREDEF(NYICF,	"NYI: C function %p")
+TREDEF(NYICF,	"NYI: C function %s")
 TREDEF(NYIFF,	"NYI: FastFunc %s")
 TREDEF(NYIFFU,	"NYI: unsupported variant of FastFunc %s")
 TREDEF(NYIRETL,	"NYI: return to lower frame")
diff --git a/src/luaconf.h b/src/luaconf.h
index 7f57bae..84fa641 100644
--- a/src/luaconf.h
+++ b/src/luaconf.h
@@ -37,7 +37,7 @@
 #endif
 #define LUA_LROOT	"/usr/local"
 #define LUA_LUADIR	"/lua/5.1/"
-#define LUA_LJDIR	"/luajit-2.0.3/"
+#define LUA_LJDIR	"/luajit-2.0.4/"
 
 #ifdef LUA_ROOT
 #define LUA_JROOT	LUA_ROOT
diff --git a/src/luajit.h b/src/luajit.h
index 901807a..9ced18e 100644
--- a/src/luajit.h
+++ b/src/luajit.h
@@ -30,9 +30,9 @@
 
 #include "lua.h"
 
-#define LUAJIT_VERSION		"LuaJIT 2.0.3"
-#define LUAJIT_VERSION_NUM	20003  /* Version 2.0.3 = 02.00.03. */
-#define LUAJIT_VERSION_SYM	luaJIT_version_2_0_3
+#define LUAJIT_VERSION		"LuaJIT 2.0.4"
+#define LUAJIT_VERSION_NUM	20004  /* Version 2.0.4 = 02.00.04. */
+#define LUAJIT_VERSION_SYM	luaJIT_version_2_0_4
 #define LUAJIT_COPYRIGHT	"Copyright (C) 2005-2015 Mike Pall"
 #define LUAJIT_URL		"http://luajit.org/"
 
diff --git a/src/ps4build.bat b/src/ps4build.bat
index 42fc9a6..337a44f 100644
--- a/src/ps4build.bat
+++ b/src/ps4build.bat
@@ -27,11 +27,11 @@ if exist minilua.exe.manifest^
 @minilua

 @if not errorlevel 8 goto :FAIL

 

-@set DASMFLAGS=-D P64

+@set DASMFLAGS=-D P64 -D NO_UNWIND

 minilua %DASM% -LN %DASMFLAGS% -o host\buildvm_arch.h vm_x86.dasc

 @if errorlevel 1 goto :BAD

 

-%LJCOMPILE% /I "." /I %DASMDIR% -DLUAJIT_TARGET=LUAJIT_ARCH_X64 -DLUAJIT_OS=LUAJIT_OS_OTHER -DLUAJIT_DISABLE_JIT -DLUAJIT_DISABLE_FFI host\buildvm*.c

+%LJCOMPILE% /I "." /I %DASMDIR% -DLUAJIT_TARGET=LUAJIT_ARCH_X64 -DLUAJIT_OS=LUAJIT_OS_OTHER -DLUAJIT_DISABLE_JIT -DLUAJIT_DISABLE_FFI -DLUAJIT_NO_UNWIND host\buildvm*.c

 @if errorlevel 1 goto :BAD

 %LJLINK% /out:buildvm.exe buildvm*.obj

 @if errorlevel 1 goto :BAD

diff --git a/src/vm_x86.dasc b/src/vm_x86.dasc
index 4544a3b..129ab00 100644
--- a/src/vm_x86.dasc
+++ b/src/vm_x86.dasc
@@ -222,6 +222,9 @@
 |.define CFRAME_SPACE,	aword*5			// Delta for rsp (see <--).
 |.macro saveregs_
 |  push rbx; push r15; push r14
+|.if NO_UNWIND
+|  push r13; push r12
+|.endif
 |  sub rsp, CFRAME_SPACE
 |.endmacro
 |.macro saveregs
@@ -229,15 +232,28 @@
 |.endmacro
 |.macro restoreregs
 |  add rsp, CFRAME_SPACE
+|.if NO_UNWIND
+|  pop r12; pop r13
+|.endif
 |  pop r14; pop r15; pop rbx; pop rbp
 |.endmacro
 |
 |//----- 16 byte aligned,
+|.if NO_UNWIND
+|.define SAVE_RET,	aword [rsp+aword*11]	//<-- rsp entering interpreter.
+|.define SAVE_R4,	aword [rsp+aword*10]
+|.define SAVE_R3,	aword [rsp+aword*9]
+|.define SAVE_R2,	aword [rsp+aword*8]
+|.define SAVE_R1,	aword [rsp+aword*7]
+|.define SAVE_RU2,	aword [rsp+aword*6]
+|.define SAVE_RU1,	aword [rsp+aword*5]	//<-- rsp after register saves.
+|.else
 |.define SAVE_RET,	aword [rsp+aword*9]	//<-- rsp entering interpreter.
 |.define SAVE_R4,	aword [rsp+aword*8]
 |.define SAVE_R3,	aword [rsp+aword*7]
 |.define SAVE_R2,	aword [rsp+aword*6]
 |.define SAVE_R1,	aword [rsp+aword*5]	//<-- rsp after register saves.
+|.endif
 |.define SAVE_CFRAME,	aword [rsp+aword*4]
 |.define SAVE_PC,	dword [rsp+dword*7]
 |.define SAVE_L,	dword [rsp+dword*6]
@@ -2564,8 +2580,8 @@ static void build_subroutines(BuildCtx *ctx)
   |
   |.define TOBIT_BIAS, 0x59c00000	// 2^52 + 2^51 (float, not double!).
   |
-  |.macro .ffunc_bit, name, kind
-  |  .ffunc_1 name
+  |.macro .ffunc_bit, name, kind, fdef
+  |  fdef name
   |.if kind == 2
   |.if SSE
   |  sseconst_tobit xmm1, RBa
@@ -2608,6 +2624,10 @@ static void build_subroutines(BuildCtx *ctx)
   |2:
   |.endmacro
   |
+  |.macro .ffunc_bit, name, kind
+  |  .ffunc_bit name, kind, .ffunc_1
+  |.endmacro
+  |
   |.ffunc_bit bit_tobit, 0
   |.if DUALNUM or SSE
   |.if not SSE
@@ -2681,7 +2701,7 @@ static void build_subroutines(BuildCtx *ctx)
   |
   |.macro .ffunc_bit_sh, name, ins
   |.if DUALNUM
-  |  .ffunc_bit name, 1
+  |  .ffunc_bit name, 1, .ffunc_2
   |  // Note: no inline conversion from number for 2nd argument!
   |  cmp dword [BASE+12], LJ_TISNUM; jne ->fff_fallback
   |  mov RA, dword [BASE+8]
@@ -6115,6 +6135,10 @@ static void emit_asm_debug(BuildCtx *ctx)
 	"\t.byte 0x83\n\t.uleb128 0x3\n"	/* offset rbx */
 	"\t.byte 0x8f\n\t.uleb128 0x4\n"	/* offset r15 */
 	"\t.byte 0x8e\n\t.uleb128 0x5\n"	/* offset r14 */
+#if LJ_NO_UNWIND
+	"\t.byte 0x8d\n\t.uleb128 0x6\n"	/* offset r13 */
+	"\t.byte 0x8c\n\t.uleb128 0x7\n"	/* offset r12 */
+#endif
 #else
 	"\t.long .Lbegin\n"
 	"\t.long %d\n"
@@ -6150,6 +6174,7 @@ static void emit_asm_debug(BuildCtx *ctx)
 	"\t.align " SZPTR "\n"
 	".LEFDE1:\n\n", (int)ctx->codesz - fcofs);
 #endif
+#if !LJ_NO_UNWIND
 #if (defined(__sun__) && defined(__svr4__))
 #if LJ_64
     fprintf(ctx->fp, "\t.section .eh_frame,\"a\",@unwind\n");
@@ -6238,7 +6263,9 @@ static void emit_asm_debug(BuildCtx *ctx)
 	"\t.align " SZPTR "\n"
 	".LEFDE3:\n\n", (int)ctx->codesz - fcofs);
 #endif
+#endif
     break;
+#if !LJ_NO_UNWIND
   /* Mental note: never let Apple design an assembler.
   ** Or a linker. Or a plastic case. But I digress.
   */
@@ -6366,6 +6393,7 @@ static void emit_asm_debug(BuildCtx *ctx)
 #endif
     }
     break;
+#endif
   default:  /* Difficult for other modes. */
     break;
   }