In rehashtab, do counthash before adding asize

Thanks alug!

src/Makefile

@@ -632,6 +632,10 @@ clean:
 	$(REMOVE) $(OBJDIR)/*.o
 ifndef NOBLUAJIT
 	$(REMOVE) $(OBJDIR)/buildvm* $(OBJDIR)/minilua* $(OBJDIR)/lj_vm* $(ALL_HDRGEN)
+
+cleanjit:
+	$(REMOVE) $(LJVMCORE_O) $(ALL_GEN)
+	$(REMOVE) $(OBJDIR)/buildvm* $(OBJDIR)/minilua* $(OBJDIR)/lj_vm* $(ALL_HDRGEN)
 endif
 
 ifdef MINGW

src/bluajit/jit/p.lua

@@ -227,9 +227,7 @@ local function prof_finish()
     local samples = prof_samples
     if samples == 0 then
       if prof_raw ~= true then out:write("[No samples collected]\n") end
-      return
+    elseif prof_ann then
-    end
-    if prof_ann then
       prof_annotate(prof_count1, samples)
     else
       prof_top(prof_count1, prof_count2, samples, "")

src/bluajit/lib_base.c

@@ -470,7 +470,11 @@ LJLIB_CF(collectgarbage)
     "\4stop\7restart\7collect\5count\1\377\4step\10setpause\12setstepmul\1\377\11isrunning");
   int32_t data = lj_lib_optint(L, 2, 0);
   if (opt == LUA_GCCOUNT) {
+#if LJ_INTONLY
+    setintV(L->top, G(L)->gc.total/1024);
+#else
     setnumV(L->top, (lua_Number)G(L)->gc.total/1024.0);
+#endif
   } else {
     int res = lua_gc(L, opt, data);
     if (opt == LUA_GCSTEP || opt == LUA_GCISRUNNING)

src/bluajit/lib_bit.c

@@ -98,7 +98,7 @@ LJLIB_ASM(bit_lshift)		LJLIB_REC(bit_shift IR_BSHL)
     x = lj_carith_shift64(x, sh, curr_func(L)->c.ffid - (int)FF_bit_lshift);
     return bit_result64(L, id, x);
   }
-  if (id2) setintV(L->base+1, sh);
+  setintV(L->base+1, sh);
   return FFH_RETRY;
 #else
   lj_lib_checknumber(L, 1);

src/bluajit/lib_table.c

@@ -72,7 +72,11 @@ LJLIB_CF(table_maxn)
       lua_Number n = numberVnum(&node[i].key);
       if (n > m) m = n;
     }
+#if LJ_INTONLY
+  setintV(L->top-1, lj_num2int(m));
+#else
   setnumV(L->top-1, m);
+#endif
   return 1;
 }
 

src/bluajit/lj_asm.c

@@ -949,11 +949,11 @@ static int asm_sunk_store(ASMState *as, IRIns *ira, IRIns *irs)
 static void asm_snap_alloc1(ASMState *as, IRRef ref)
 {
   IRIns *ir = IR(ref);
-  if (!irref_isk(ref) && ir->r != RID_SUNK) {
+  if (!irref_isk(ref)) {
     bloomset(as->snapfilt1, ref);
     bloomset(as->snapfilt2, hashrot(ref, ref + HASH_BIAS));
     if (ra_used(ir)) return;
-    if (ir->r == RID_SINK) {
+    if (ir->r == RID_SINK || ir->r == RID_SUNK) {
       ir->r = RID_SUNK;
 #if LJ_HASFFI
       if (ir->o == IR_CNEWI) {  /* Allocate CNEWI value. */

src/bluajit/lj_bcread.c

@@ -408,7 +408,7 @@ static int bcread_header(LexState *ls)
 #endif
   }
   if ((flags & BCDUMP_F_STRIP)) {
-    ls->chunkname = lj_str_newz(ls->L, ls->chunkarg);
+    ls->chunkname = lj_str_newz(ls->L, *ls->chunkarg == BCDUMP_HEAD1 ? "=?" : ls->chunkarg);
   } else {
     MSize len = bcread_uleb128(ls);
     bcread_need(ls, len);

src/bluajit/lj_parse.c

@@ -2970,10 +2970,11 @@ static int parse_stmt(LexState *ls)
   case TK_break:
     lj_lex_next(ls);
     #if BLUAJIT_MULTILEVEL_BREAK
-      int levels = numberVint(&ls->tokval);
+      if (ls->tok == TK_number) {
-      if (lex_opt(ls, TK_number))
+	   int levels = numberVint(&ls->tokval);
-        parse_break(ls, levels);
+	   lj_lex_next(ls);
-      else
+	   parse_break(ls, levels);
+	} else
     #endif
     parse_break(ls, 1);
     return !LJ_52;  /* Must be last in Lua 5.1. */

src/bluajit/lj_record.c

@@ -2084,25 +2084,19 @@ static TRef rec_tnew(jit_State *J, uint32_t ah)
 
 typedef struct RecCatDataCP {
   jit_State *J;
-  RecordIndex *ix;
+  BCReg baseslot, topslot;
+  TRef tr;
 } RecCatDataCP;
 
 static TValue *rec_mm_concat_cp(lua_State *L, lua_CFunction dummy, void *ud)
 {
   RecCatDataCP *rcd = (RecCatDataCP *)ud;
-  UNUSED(L); UNUSED(dummy);
+  jit_State *J = rcd->J;
-  rec_mm_arith(rcd->J, rcd->ix, MM_concat);  /* Call __concat metamethod. */
+  BCReg baseslot = rcd->baseslot, topslot = rcd->topslot;
-  return NULL;
-}
-
-static TRef rec_cat(jit_State *J, BCReg baseslot, BCReg topslot)
-{
   TRef *top = &J->base[topslot];
-  TValue savetv[5+LJ_FR2];
   BCReg s;
   RecordIndex ix;
-  RecCatDataCP rcd;
+  UNUSED(L); UNUSED(dummy);
-  int errcode;
   lj_assertJ(baseslot < topslot, "bad CAT arg");
   for (s = baseslot; s <= topslot; s++)
     (void)getslot(J, s);  /* Ensure all arguments have a reference. */
@@ -2124,7 +2118,10 @@ static TRef rec_cat(jit_State *J, BCReg baseslot, BCReg topslot)
     } while (trp <= top);
     tr = emitir(IRTG(IR_BUFSTR, IRT_STR), tr, hdr);
     J->maxslot = (BCReg)(xbase - J->base);
-    if (xbase == base) return tr;  /* Return simple concatenation result. */
+    if (xbase == base) {
+      rcd->tr = tr;  /* Return simple concatenation result. */
+      return NULL;
+    }
     /* Pass partial result. */
     topslot = J->maxslot--;
     *xbase = tr;
@@ -2137,13 +2134,31 @@ static TRef rec_cat(jit_State *J, BCReg baseslot, BCReg topslot)
   copyTV(J->L, &ix.tabv, &J->L->base[topslot-1]);
   ix.tab = top[-1];
   ix.key = top[0];
-  memcpy(savetv, &J->L->base[topslot-1], sizeof(savetv));  /* Save slots. */
+  rec_mm_arith(J, &ix, MM_concat);  /* Call __concat metamethod. */
+  rcd->tr = 0;  /* No result yet. */
+  return NULL;
+}
+
+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;
+  RecCatDataCP rcd;
+  int errcode;
   rcd.J = J;
-  rcd.ix = &ix;
+  rcd.baseslot = baseslot;
-  errcode = lj_vm_cpcall(J->L, NULL, &rcd, rec_mm_concat_cp);
+  rcd.topslot = topslot;
-  memcpy(&J->L->base[topslot-1], savetv, sizeof(savetv));  /* Restore slots. */
+  memcpy(savetv, &L->base[topslot-1], sizeof(savetv));  /* Save slots. */
-  if (errcode) return (TRef)(-errcode);
+  errcode = lj_vm_cpcall(L, NULL, &rcd, rec_mm_concat_cp);
-  return 0;  /* No result yet. */
+  if (errcode) copyTV(L, &errobj, L->top-1);
+  memcpy(&L->base[topslot-1], savetv, sizeof(savetv));  /* Restore slots. */
+  if (errcode) {
+    L->top = L->base + delta;
+    copyTV(L, L->top++, &errobj);
+    return (TRef)(-errcode);
+  }
+  return rcd.tr;
 }
 
 /* -- Record bytecode ops ------------------------------------------------- */

src/bluajit/lj_tab.c

@@ -359,8 +359,14 @@ static void rehashtab(lua_State *L, GCtab *t, cTValue *ek)
   uint32_t total, asize, na, i;
   for (i = 0; i < LJ_MAX_ABITS; i++) bins[i] = 0;
   asize = countarray(t, bins);
+#if 0
   total = 1 + asize;
   total += counthash(t, bins, &asize);
+#else
+  /* more closely matches vanilla lua when abusing undefined behavior (frictionfix.....) */
+  total = counthash(t, bins, &asize);
+  total += 1 + asize;
+#endif
   asize += countint(ek, bins);
   na = bestasize(bins, &asize);
   total -= na;
@@ -635,7 +641,7 @@ LJ_NOINLINE static MSize tab_len_slow(GCtab *t, size_t hi)
   while ((tv = lj_tab_getint(t, (int32_t)hi)) && !tvisnil(tv)) {
     lo = hi;
     hi += hi;
-    if (hi > (size_t)(INT_MAX-2)) {  /* Punt and do a linear search. */
+    if (hi > (size_t)(0x7fffffff - 2)) {  /* Punt and do a linear search. */
       lo = 1;
       while ((tv = lj_tab_getint(t, (int32_t)lo)) && !tvisnil(tv)) lo++;
       return (MSize)(lo - 1);

src/bluajit/lj_vmmath.c

@@ -78,7 +78,7 @@ int32_t lj_vm_foldarith_int(int32_t x, int32_t y, int op)
   case IR_MOD - IR_ADD: return x%y; break;
   case IR_POW - IR_ADD: return pow(x, y); break;
   case IR_NEG - IR_ADD: return -x; break;
-  case IR_ABS - IR_ADD: return fabs(x); break;
+  case IR_ABS - IR_ADD: return abs(x); break;
 #if LJ_HASJIT
   case IR_LDEXP - IR_ADD: return ldexp(x, (int)y); break;
   case IR_MIN - IR_ADD: return x < y ? x : y; break;

src/bluajit/luaconf.h

@@ -9,7 +9,6 @@
 #ifndef WINVER
 #define WINVER 0x0501
 #endif
-#include <limits.h>
 #include <stddef.h>
 
 /* Default path for loading Lua and C modules with require(). */

src/bluajit/vm_mips64.dasc

@@ -1420,9 +1420,10 @@ static void build_subroutines(BuildCtx *ctx)
   |  sltu AT, TMP1, TMP2
   |  bnez AT, ->fff_fallback
   |.  lbu TMP3, DISPATCH_GL(hookmask)(DISPATCH)
-  |  daddiu NARGS8:RC, NARGS8:RC, -8
+  |  daddiu NARGS8:TMP0, NARGS8:RC, -8
-  |  bltz NARGS8:RC, ->fff_fallback
+  |  bltz NARGS8:TMP0, ->fff_fallback
   |.   move TMP2, BASE
+  |  move NARGS8:RC, NARGS8:TMP0
   |   daddiu BASE, BASE, 16
   |  // Remember active hook before pcall.
   |  srl TMP3, TMP3, HOOK_ACTIVE_SHIFT

src/lua_baselib.c

@@ -40,6 +40,8 @@ boolean luaL_checkboolean(lua_State *L, int narg) {
 	return lua_toboolean(L, narg);
 }
 
+static int ref_open, ref_iowrite, ref_fwrite; // original versions of io functions
+
 #define FILELIMIT 1024*1024 // Size limit for reading/writing files
 
 static const char *whitelist[] = { // Allow scripters to write files of these types to SRB2's folder
@@ -102,7 +104,7 @@ static int lib_open(lua_State *L)
 	lua_pushstring(L, destFilename);
 	lua_replace(L, 1);
 	// now call the real io.open
-	lua_pushvalue(L, lua_upvalueindex(1));
+	lua_rawgeti(L, LUA_REGISTRYINDEX, ref_open);
 	lua_insert(L, 1);
 	lua_call(L, lua_gettop(L)-1, LUA_MULTRET);
 	return lua_gettop(L);
@@ -114,8 +116,10 @@ static int lib_write(lua_State *L, int arg)
 	int top = lua_gettop(L);
 	boolean error = true;
 	size_t len, size = 0;
+	int ogref = ref_iowrite;
 
 	if (arg != 1) {
+		ogref = ref_fwrite;
 		FILE *fp = *(FILE **)luaL_checkudata(L, 1, LUA_FILEHANDLE);
 		if (fp == NULL) goto call; // real file:write will throw
 		size = ftell(fp);
@@ -134,7 +138,7 @@ static int lib_write(lua_State *L, int arg)
 	error = false;
 
 call:
-	lua_pushvalue(L, lua_upvalueindex(1));
+	lua_rawgeti(L, LUA_REGISTRYINDEX, ogref);
 	lua_insert(L, 1);
 	lua_call(L, lua_gettop(L)-1, LUA_MULTRET);
 
@@ -3449,25 +3453,24 @@ int LUA_BaseLib(lua_State *L)
 	lua_pop(L, 2); // pop metatable and dummy string
 
 	// replaces an existing function with another,
-	// saving the original as an upvalue
+	// saving the original as a reference
-#define REPLACE(name, func) \
+#define REPLACE(name, func, ref) \
 	lua_pushliteral(L, name); \
 	lua_rawget(L, -2); \
-	lua_pushcclosure(L, &func, 1); \
+	ref = luaL_ref(L, LUA_REGISTRYINDEX); \
 	lua_pushliteral(L, name); \
-	lua_pushvalue(L, -2); \
+	lua_pushcclosure(L, &func, 0); \
-	lua_rawset(L, -4); \
+	lua_rawset(L, -3);
-	lua_pop(L, 1);
 
 	// replace io.open
 	lua_getglobal(L, "io");
-	REPLACE("open", lib_open)
+	REPLACE("open", lib_open, ref_open)
-	REPLACE("write", lib_write_io)
+	REPLACE("write", lib_write_io, ref_iowrite)
 	lua_pop(L, 1);
 
 	// replace file:write
 	luaL_getmetatable(L, LUA_FILEHANDLE);
-	REPLACE("write", lib_write_f)
+	REPLACE("write", lib_write_f, ref_fwrite)
 	lua_pop(L, 1);
 #undef REPLACE
 

src/lua_consolelib.c

@@ -329,6 +329,7 @@ static int lib_cvRegisterVar(lua_State *L)
 	const char* category = NULL;
 	const char* menu_name = NULL;
 	boolean hasfunc = false;
+	int funcref = 0;
 
 	luaL_checktype(L, 1, LUA_TTABLE);
 	lua_settop(L, 1); // Clear out all other possible arguments, leaving only the first one.
@@ -429,7 +430,9 @@ static int lib_cvRegisterVar(lua_State *L)
 		} else if (i == 5 || (k && fasticmp(k, "func"))) {
 			if (!lua_isfunction(L, 4))
 				TYPEERROR("func", LUA_TFUNCTION)
-			lua_setfield(L, LUA_REGISTRYINDEX, "CVFUNC");
+			if (hasfunc)
+				luaL_unref(L, LUA_REGISTRYINDEX, funcref); // real clever, huh?
+			funcref = luaL_ref(L, LUA_REGISTRYINDEX);
 			hasfunc = true;
 			continue;
 		}
@@ -450,10 +453,11 @@ static int lib_cvRegisterVar(lua_State *L)
 	if (hasfunc) {
 		lua_getfield(L, LUA_REGISTRYINDEX, "CV_OnChange");
 		I_Assert(lua_istable(L, 3));
-		lua_getfield(L, LUA_REGISTRYINDEX, "CVFUNC");
+		lua_rawgeti(L, LUA_REGISTRYINDEX, funcref);
 		lua_setfield(L, 3, cvar->name);
 		lua_pop(L, 1);
 		cvar->func = Lua_OnChange;
+		luaL_unref(L, LUA_REGISTRYINDEX, funcref);
 	}
 
 	lua_getfield(L, LUA_REGISTRYINDEX, "CV_Vars");