diff --git a/src/lib_package.c b/src/lib_package.c
index 552db305..40bafed5 100644
--- a/src/lib_package.c
+++ b/src/lib_package.c
@@ -572,8 +572,7 @@ LUALIB_API int luaopen_package(lua_State *L)
   lj_lib_pushcf(L, lj_cf_package_unloadlib, 1);
   lua_setfield(L, -2, "__gc");
   luaL_register(L, LUA_LOADLIBNAME, package_lib);
-  lua_pushvalue(L, -1);
-  lua_replace(L, LUA_ENVIRONINDEX);
+  lua_copy(L, -1, LUA_ENVIRONINDEX);
   lua_createtable(L, sizeof(package_loaders)/sizeof(package_loaders[0])-1, 0);
   for (i = 0; package_loaders[i] != NULL; i++) {
     lj_lib_pushcf(L, package_loaders[i], 1);
diff --git a/src/lj_api.c b/src/lj_api.c
index f0f41766..3c1a07bc 100644
--- a/src/lj_api.c
+++ b/src/lj_api.c
@@ -159,30 +159,40 @@ LUA_API void lua_insert(lua_State *L, int idx)
   copyTV(L, p, L->top);
 }
 
-LUA_API void lua_replace(lua_State *L, int idx)
+static void copy_slot(lua_State *L, TValue *f, int idx)
 {
-  api_checknelems(L, 1);
   if (idx == LUA_GLOBALSINDEX) {
-    api_check(L, tvistab(L->top-1));
+    api_check(L, tvistab(f));
     /* NOBARRIER: A thread (i.e. L) is never black. */
-    setgcref(L->env, obj2gco(tabV(L->top-1)));
+    setgcref(L->env, obj2gco(tabV(f)));
   } else if (idx == LUA_ENVIRONINDEX) {
     GCfunc *fn = curr_func(L);
     if (fn->c.gct != ~LJ_TFUNC)
       lj_err_msg(L, LJ_ERR_NOENV);
-    api_check(L, tvistab(L->top-1));
-    setgcref(fn->c.env, obj2gco(tabV(L->top-1)));
-    lj_gc_barrier(L, fn, L->top-1);
+    api_check(L, tvistab(f));
+    setgcref(fn->c.env, obj2gco(tabV(f)));
+    lj_gc_barrier(L, fn, f);
   } else {
     TValue *o = index2adr(L, idx);
     api_checkvalidindex(L, o);
-    copyTV(L, o, L->top-1);
+    copyTV(L, o, f);
     if (idx < LUA_GLOBALSINDEX)  /* Need a barrier for upvalues. */
-      lj_gc_barrier(L, curr_func(L), L->top-1);
+      lj_gc_barrier(L, curr_func(L), f);
   }
+}
+
+LUA_API void lua_replace(lua_State *L, int idx)
+{
+  api_checknelems(L, 1);
+  copy_slot(L, L->top - 1, idx);
   L->top--;
 }
 
+LUA_API void lua_copy(lua_State *L, int fromidx, int toidx)
+{
+  copy_slot(L, index2adr(L, fromidx), toidx);
+}
+
 LUA_API void lua_pushvalue(lua_State *L, int idx)
 {
   copyTV(L, L->top, index2adr(L, idx));
diff --git a/src/lua.h b/src/lua.h
index f067483a..0c653f56 100644
--- a/src/lua.h
+++ b/src/lua.h
@@ -349,6 +349,7 @@ LUA_API void lua_upvaluejoin (lua_State *L, int idx1, int n1, int idx2, int n2);
 LUA_API int lua_loadx (lua_State *L, lua_Reader reader, void *dt,
 		       const char *chunkname, const char *mode);
 LUA_API const lua_Number *lua_version (lua_State *L);
+LUA_API void lua_copy (lua_State *L, int fromidx, int toidx);
 
 
 struct lua_Debug {