* parse.y: support Symbol GC. [ruby-trunk Feature #9634]

See this ticket about Symbol GC.

* include/ruby/ruby.h:
  Declare few functions.
  * rb_sym2id: almost same as old SYM2ID but support dynamic symbols.
  * rb_id2sym: almost same as old ID2SYM but support dynamic symbols.
  * rb_sym2str: almost same as `rb_id2str(SYM2ID(sym))` but not
    pin down a dynamic symbol.
  Declare a new struct.
  * struct RSymbol: represents a dynamic symbol as object in
    Ruby's heaps.
  Add few macros.
  * STATIC_SYM_P: check a static symbol.
  * DYNAMIC_SYM_P: check a dynamic symbol.
  * RSYMBOL: cast to RSymbol

* gc.c: declare RSymbol. support T_SYMBOL.

* internal.h: Declare few functions.
  * rb_gc_free_dsymbol: free up a dynamic symbol. GC call this
    function at a sweep phase.
  * rb_str_dynamic_intern: convert a string to a dynamic symbol.
  * rb_check_id_without_pindown: not pinning function.
  * rb_sym2id_without_pindown: ditto.
  * rb_check_id_cstr_without_pindown: ditto.

* string.c (Init_String): String#intern and String#to_sym use
  rb_str_dynamic_intern.

* template/id.h.tmpl: use LSB of ID as a flag for determining a
  static symbol, so we shift left other ruby_id_types.

* string.c: use rb_sym2str instead `rb_id2str(SYM2ID(sym))` to
  avoid pinning.

* load.c: use xx_without_pindown function at creating temporary ID
  to avoid pinning.

* object.c: ditto.

* sprintf.c: ditto.

* struct.c: ditto.

* thread.c: ditto.

* variable.c: ditto.

* vm_method.c: ditto.

git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@45426 b2dd03c8-39d4-4d8f-98ff-823fe69b080e

ChangeLog

@@ -1,3 +1,56 @@
+Tue Mar 25 22:57:11 2014  Narihiro Nakamura  <authornari@gmail.com>
+
+	* parse.y: support Symbol GC. [ruby-trunk Feature #9634]
+	  See this ticket about Symbol GC.
+
+	* include/ruby/ruby.h:
+	  Declare few functions.
+	  * rb_sym2id: almost same as old SYM2ID but support dynamic symbols.
+	  * rb_id2sym: almost same as old ID2SYM but support dynamic symbols.
+	  * rb_sym2str: almost same as `rb_id2str(SYM2ID(sym))` but not
+	    pin down a dynamic symbol.
+	  Declare a new struct.
+	  * struct RSymbol: represents a dynamic symbol as object in
+	    Ruby's heaps.
+	  Add few macros.
+	  * STATIC_SYM_P: check a static symbol.
+	  * DYNAMIC_SYM_P: check a dynamic symbol.
+	  * RSYMBOL: cast to RSymbol
+
+	* gc.c: declare RSymbol. support T_SYMBOL.
+
+	* internal.h: Declare few functions.
+	  * rb_gc_free_dsymbol: free up a dynamic symbol. GC call this
+	    function at a sweep phase.
+	  * rb_str_dynamic_intern: convert a string to a dynamic symbol.
+	  * rb_check_id_without_pindown: not pinning function.
+	  * rb_sym2id_without_pindown: ditto.
+	  * rb_check_id_cstr_without_pindown: ditto.
+
+	* string.c (Init_String): String#intern and String#to_sym use
+	  rb_str_dynamic_intern.
+
+	* template/id.h.tmpl: use LSB of ID as a flag for determining a
+	  static symbol, so we shift left other ruby_id_types.
+
+	* string.c: use rb_sym2str instead `rb_id2str(SYM2ID(sym))` to
+	  avoid pinning.
+
+	* load.c: use xx_without_pindown function at creating temporary ID
+	  to avoid pinning.
+
+	* object.c: ditto.
+
+	* sprintf.c: ditto.
+
+	* struct.c: ditto.
+
+	* thread.c: ditto.
+
+	* variable.c: ditto.
+
+	* vm_method.c: ditto.
+
 Wed Mar 26 13:25:54 2014  NARUSE, Yui  <naruse@ruby-lang.org>
 
 	* addr2line.c (fill_lines): loop reverse order not to overwrite

gc.c

@@ -354,6 +354,7 @@ typedef struct RVALUE {
 	struct RMatch  match;
 	struct RRational rational;
 	struct RComplex complex;
+	struct RSymbol symbol;
 	struct {
 	    struct RBasic basic;
 	    VALUE v1;
@@ -1652,6 +1653,12 @@ obj_free(rb_objspace_t *objspace, VALUE obj)
 	}
 	break;
 
+      case T_SYMBOL:
+	{
+            rb_gc_free_dsymbol(obj);
+	}
+	break;
+
       default:
 	rb_bug("gc_sweep(): unknown data type 0x%x(%p) 0x%"PRIxVALUE,
 	       BUILTIN_TYPE(obj), (void*)obj, RBASIC(obj)->flags);
@@ -2393,7 +2400,7 @@ rb_obj_id(VALUE obj)
      *  24 if 32-bit, double is 8-byte aligned
      *  40 if 64-bit
      */
-    if (SYMBOL_P(obj)) {
+    if (STATIC_SYM_P(obj)) {
         return (SYM2ID(obj) * sizeof(RVALUE) + (4 << 2)) | FIXNUM_FLAG;
     }
     else if (FLONUM_P(obj)) {
@@ -2499,6 +2506,7 @@ obj_memsize_of(VALUE obj, int use_tdata)
 	break;
 
       case T_FLOAT:
+      case T_SYMBOL:
 	break;
 
       case T_BIGNUM:
@@ -3918,6 +3926,7 @@ gc_mark_children(rb_objspace_t *objspace, VALUE ptr)
 
       case T_FLOAT:
       case T_BIGNUM:
+      case T_SYMBOL:
 	break;
 
       case T_MATCH:

include/ruby/ruby.h

@@ -351,9 +351,13 @@ rb_long2int_inline(long n)
 
 #define IMMEDIATE_P(x) ((VALUE)(x) & IMMEDIATE_MASK)
 
-#define SYMBOL_P(x) (((VALUE)(x)&~((~(VALUE)0)<<RUBY_SPECIAL_SHIFT))==SYMBOL_FLAG)
+ID rb_sym2id(VALUE);
-#define ID2SYM(x) (((VALUE)(x)<<RUBY_SPECIAL_SHIFT)|SYMBOL_FLAG)
+VALUE rb_id2sym(ID);
-#define SYM2ID(x) RSHIFT((unsigned long)(x),RUBY_SPECIAL_SHIFT)
+#define STATIC_SYM_P(x) (((VALUE)(x)&~((~(VALUE)0)<<RUBY_SPECIAL_SHIFT))==SYMBOL_FLAG)
+#define DYNAMIC_SYM_P(x) (!SPECIAL_CONST_P(x) && BUILTIN_TYPE(x) == (T_SYMBOL))
+#define SYMBOL_P(x) (STATIC_SYM_P(x)||DYNAMIC_SYM_P(x))
+#define ID2SYM(x) (rb_id2sym(x))
+#define SYM2ID(x) (rb_sym2id(x))
 
 #ifndef USE_FLONUM
 #if SIZEOF_VALUE >= SIZEOF_DOUBLE
@@ -957,6 +961,12 @@ struct RComplex {
 #define RCOMPLEX_SET_REAL(cmp, r) RB_OBJ_WRITE((cmp), &((struct RComplex *)(cmp))->real,(r))
 #define RCOMPLEX_SET_IMAG(cmp, i) RB_OBJ_WRITE((cmp), &((struct RComplex *)(cmp))->imag,(i))
 
+struct RSymbol {
+    struct RBasic basic;
+    VALUE fstr;
+    ID type;
+};
+
 struct RData {
     struct RBasic basic;
     void (*dmark)(void*);
@@ -1093,6 +1103,7 @@ struct RStruct {
 #define RFILE(obj)   (R_CAST(RFile)(obj))
 #define RRATIONAL(obj) (R_CAST(RRational)(obj))
 #define RCOMPLEX(obj) (R_CAST(RComplex)(obj))
+#define RSYMBOL(obj) (R_CAST(RSymbol)(obj))
 
 #define FL_SINGLETON FL_USER0
 #define FL_WB_PROTECTED (((VALUE)1)<<5)
@@ -1146,7 +1157,7 @@ struct RStruct {
     (OBJ_TAINTABLE(x) && FL_ABLE(s)) ? \
     RBASIC(x)->flags |= RBASIC(s)->flags & FL_TAINT : 0)
 
-#define OBJ_FROZEN(x) (!!(FL_ABLE(x)?(RBASIC(x)->flags&(FL_FREEZE)):(FIXNUM_P(x)||FLONUM_P(x)||SYMBOL_P(x))))
+#define OBJ_FROZEN(x) (!!(FL_ABLE(x)?(RBASIC(x)->flags&(FL_FREEZE)):(FIXNUM_P(x)||FLONUM_P(x)||STATIC_SYM_P(x))))
 #define OBJ_FREEZE(x) FL_SET((x), FL_FREEZE)
 
 #if USE_RGENGC
@@ -1381,6 +1392,7 @@ const char *rb_id2name(ID);
 ID rb_check_id(volatile VALUE *);
 ID rb_to_id(VALUE);
 VALUE rb_id2str(ID);
+VALUE rb_sym2str(VALUE);
 
 #define CONST_ID_CACHE(result, str)			\
     {							\
@@ -1597,7 +1609,7 @@ rb_class_of(VALUE obj)
 	if (FIXNUM_P(obj)) return rb_cFixnum;
 	if (FLONUM_P(obj)) return rb_cFloat;
 	if (obj == Qtrue)  return rb_cTrueClass;
-	if (SYMBOL_P(obj)) return rb_cSymbol;
+	if (STATIC_SYM_P(obj)) return rb_cSymbol;
     }
     else if (!RTEST(obj)) {
 	if (obj == Qnil)   return rb_cNilClass;
@@ -1613,7 +1625,7 @@ rb_type(VALUE obj)
 	if (FIXNUM_P(obj)) return T_FIXNUM;
         if (FLONUM_P(obj)) return T_FLOAT;
         if (obj == Qtrue)  return T_TRUE;
-	if (SYMBOL_P(obj)) return T_SYMBOL;
+	if (STATIC_SYM_P(obj)) return T_SYMBOL;
 	if (obj == Qundef) return T_UNDEF;
     }
     else if (!RTEST(obj)) {

internal.h

@@ -742,6 +742,13 @@ int rb_is_junk_name(VALUE name);
 void rb_gc_mark_parser(void);
 void rb_gc_mark_symbols(int full_mark);
 ID rb_make_internal_id(void);
+void rb_gc_free_dsymbol(VALUE);
+VALUE rb_str_dynamic_intern(VALUE);
+ID rb_check_id_without_pindown(VALUE *);
+ID rb_sym2id_without_pindown(VALUE);
+#ifdef RUBY_ENCODING_H
+ID rb_check_id_cstr_without_pindown(const char *, long, rb_encoding *);
+#endif
 
 /* proc.c */
 VALUE rb_proc_location(VALUE self);

load.c

@@ -1108,7 +1108,7 @@ rb_mod_autoload(VALUE mod, VALUE sym, VALUE file)
 static VALUE
 rb_mod_autoload_p(VALUE mod, VALUE sym)
 {
-    ID id = rb_check_id(&sym);
+    ID id = rb_check_id_without_pindown(&sym);
     if (!id) {
 	return Qnil;
     }

object.c

@@ -2125,7 +2125,7 @@ rb_mod_const_get(int argc, VALUE *argv, VALUE mod)
 
 	if (pbeg == p) goto wrong_name;
 
-	id = rb_check_id_cstr(pbeg, len = p-pbeg, enc);
+	id = rb_check_id_cstr_without_pindown(pbeg, len = p-pbeg, enc);
 	beglen = pbeg-path;
 
 	if (p < pend && p[0] == ':') {
@@ -2267,7 +2267,7 @@ rb_mod_const_defined(int argc, VALUE *argv, VALUE mod)
 
 	if (pbeg == p) goto wrong_name;
 
-	id = rb_check_id_cstr(pbeg, len = p-pbeg, enc);
+	id = rb_check_id_cstr_without_pindown(pbeg, len = p-pbeg, enc);
 	beglen = pbeg-path;
 
 	if (p < pend && p[0] == ':') {
@@ -2338,7 +2338,7 @@ rb_mod_const_defined(int argc, VALUE *argv, VALUE mod)
 static VALUE
 rb_obj_ivar_get(VALUE obj, VALUE iv)
 {
-    ID id = rb_check_id(&iv);
+    ID id = rb_check_id_without_pindown(&iv);
 
     if (!id) {
 	if (rb_is_instance_name(iv)) {
@@ -2409,7 +2409,7 @@ rb_obj_ivar_set(VALUE obj, VALUE iv, VALUE val)
 static VALUE
 rb_obj_ivar_defined(VALUE obj, VALUE iv)
 {
-    ID id = rb_check_id(&iv);
+    ID id = rb_check_id_without_pindown(&iv);
 
     if (!id) {
 	if (rb_is_instance_name(iv)) {
@@ -2446,7 +2446,7 @@ rb_obj_ivar_defined(VALUE obj, VALUE iv)
 static VALUE
 rb_mod_cvar_get(VALUE obj, VALUE iv)
 {
-    ID id = rb_check_id(&iv);
+    ID id = rb_check_id_without_pindown(&iv);
 
     if (!id) {
 	if (rb_is_class_name(iv)) {
@@ -2512,7 +2512,7 @@ rb_mod_cvar_set(VALUE obj, VALUE iv, VALUE val)
 static VALUE
 rb_mod_cvar_defined(VALUE obj, VALUE iv)
 {
-    ID id = rb_check_id(&iv);
+    ID id = rb_check_id_without_pindown(&iv);
 
     if (!id) {
 	if (rb_is_class_name(iv)) {

parse.y

@@ -41,26 +41,26 @@
 #define free	YYFREE
 
 #ifndef RIPPER
-static ID register_symid(ID, const char *, long, rb_encoding *);
+static ID register_static_symid(ID, const char *, long, rb_encoding *);
-static ID register_symid_str(ID, VALUE);
+static ID register_static_symid_str(ID, VALUE);
-#define REGISTER_SYMID(id, name) register_symid((id), (name), strlen(name), enc)
+#define REGISTER_SYMID(id, name) register_static_symid((id), (name), strlen(name), enc)
 #include "id.c"
 #endif
 
+static inline int id_type(ID);
 #define is_notop_id(id) ((id)>tLAST_OP_ID)
-#define is_local_id(id) (is_notop_id(id)&&((id)&ID_SCOPE_MASK)==ID_LOCAL)
+#define is_local_id(id) (id_type(id)==ID_LOCAL)
-#define is_global_id(id) (is_notop_id(id)&&((id)&ID_SCOPE_MASK)==ID_GLOBAL)
+#define is_global_id(id) (id_type(id)==ID_GLOBAL)
-#define is_instance_id(id) (is_notop_id(id)&&((id)&ID_SCOPE_MASK)==ID_INSTANCE)
+#define is_instance_id(id) (id_type(id)==ID_INSTANCE)
-#define is_attrset_id(id) (is_notop_id(id)&&((id)&ID_SCOPE_MASK)==ID_ATTRSET)
+#define is_attrset_id(id) (id_type(id)==ID_ATTRSET)
-#define is_const_id(id) (is_notop_id(id)&&((id)&ID_SCOPE_MASK)==ID_CONST)
+#define is_const_id(id) (id_type(id)==ID_CONST)
-#define is_class_id(id) (is_notop_id(id)&&((id)&ID_SCOPE_MASK)==ID_CLASS)
+#define is_class_id(id) (id_type(id)==ID_CLASS)
-#define is_junk_id(id) (is_notop_id(id)&&((id)&ID_SCOPE_MASK)==ID_JUNK)
+#define is_junk_id(id) (id_type(id)==ID_JUNK)
-#define id_type(id) (is_notop_id(id) ? (int)((id)&ID_SCOPE_MASK) : -1)
 
 #define is_asgn_or_id(id) ((is_notop_id(id)) && \
-	(((id)&ID_SCOPE_MASK) == ID_GLOBAL || \
+	((id_type(id)) == ID_GLOBAL || \
-	 ((id)&ID_SCOPE_MASK) == ID_INSTANCE || \
+	 (id_type(id)) == ID_INSTANCE || \
-	 ((id)&ID_SCOPE_MASK) == ID_CLASS))
+	 (id_type(id)) == ID_CLASS))
 
 enum lex_state_bits {
     EXPR_BEG_bit,		/* ignore newline, +/- is a sign. */
@@ -300,11 +300,17 @@ struct parser_params {
 #endif
 };
 
+#ifdef RIPPER
+#define intern_cstr_without_pindown(n,l,en) rb_intern3(n,l,en)
+#else
+static ID intern_cstr_without_pindown(const char *, long, rb_encoding *);
+#endif
+
 #define STR_NEW(p,n) rb_enc_str_new((p),(n),current_enc)
 #define STR_NEW0() rb_enc_str_new(0,0,current_enc)
 #define STR_NEW2(p) rb_enc_str_new((p),strlen(p),current_enc)
 #define STR_NEW3(p,n,e,func) parser_str_new((p),(n),(e),(func),current_enc)
-#define TOK_INTERN() rb_intern3(tok(), toklen(), current_enc)
+#define TOK_INTERN() intern_cstr_without_pindown(tok(), toklen(), current_enc)
 
 static int parser_yyerror(struct parser_params*, const char*);
 #define yyerror(msg) parser_yyerror(parser, (msg))
@@ -8010,7 +8016,7 @@ parser_yylex(struct parser_params *parser)
 		return '$';
 	    }
 	  gvar:
-	    set_yylval_name(rb_intern3(tok(), tokidx, current_enc));
+	    set_yylval_name(intern_cstr_without_pindown(tok(), tokidx, current_enc));
 	    return tGVAR;
 
 	  case '&':		/* $&: last match */
@@ -8813,6 +8819,8 @@ static const char id_type_names[][9] = {
     "JUNK",
 };
 
+static ID rb_pin_dynamic_symbol(VALUE);
+
 ID
 rb_id_attrset(ID id)
 {
@@ -8824,7 +8832,7 @@ rb_id_attrset(ID id)
 	rb_name_error(id, "cannot make operator ID :%s attrset", rb_id2name(id));
     }
     else {
-	int scope = (int)(id & ID_SCOPE_MASK);
+	int scope = id_type(id);
 	switch (scope) {
 	  case ID_LOCAL: case ID_INSTANCE: case ID_GLOBAL:
 	  case ID_CONST: case ID_CLASS: case ID_JUNK:
@@ -8837,8 +8845,19 @@ rb_id_attrset(ID id)
 
 	}
     }
-    id &= ~ID_SCOPE_MASK;
+    if (id&ID_STATIC_SYM) {
-    id |= ID_ATTRSET;
+        id &= ~ID_SCOPE_MASK;
+        id |= ID_ATTRSET;
+    }
+    else {
+	VALUE str;
+
+        /* make new dynamic symbol */
+	str = rb_str_dup(RSYMBOL((VALUE)id)->fstr);
+	rb_str_cat(str, "=", 1);
+	id = (ID)rb_str_dynamic_intern(str);
+	rb_pin_dynamic_symbol((VALUE)id);
+    }
     return id;
 }
 
@@ -9930,7 +9949,7 @@ reg_named_capture_assign_iter(const OnigUChar *name, const OnigUChar *name_end,
 	!rb_enc_symname2_p(s, len, enc)) {
         return ST_CONTINUE;
     }
-    var = rb_intern3(s, len, enc);
+    var = intern_cstr_without_pindown(s, len, enc);
     if (dvar_defined(var) || local_id(var)) {
         rb_warningS("named capture conflicts a local variable - %s",
                     rb_id2name(var));
@@ -10116,12 +10135,14 @@ static struct symbols {
     ID last_id;
     st_table *sym_id;
     st_table *id_str;
+    st_table *pinned_dsym;
 #if ENABLE_SELECTOR_NAMESPACE
     st_table *ivar2_id;
     st_table *id_ivar2;
 #endif
     VALUE op_sym[tLAST_OP_ID];
     int minor_marked;
+    int pinned_dsym_minor_marked;
 } global_symbols = {tLAST_TOKEN};
 
 static const struct st_hash_type symhash = {
@@ -10161,6 +10182,7 @@ Init_sym(void)
 {
     global_symbols.sym_id = st_init_table_with_size(&symhash, 1000);
     global_symbols.id_str = st_init_numtable_with_size(1000);
+    global_symbols.pinned_dsym = st_init_numtable_with_size(1000);
 #if ENABLE_SELECTOR_NAMESPACE
     global_symbols.ivar2_id = st_init_table_with_size(&ivar2_hash_type, 1000);
     global_symbols.id_ivar2 = st_init_numtable_with_size(1000);
@@ -10185,6 +10207,11 @@ rb_gc_mark_symbols(int full_mark)
 
 	if (!full_mark) global_symbols.minor_marked = 1;
     }
+
+    if (full_mark || global_symbols.pinned_dsym_minor_marked == 0) {
+	rb_mark_tbl(global_symbols.pinned_dsym);
+	if (!full_mark) global_symbols.pinned_dsym_minor_marked = 1;
+    }
 }
 #endif /* !RIPPER */
 
@@ -10193,7 +10220,22 @@ internal_id_gen(struct parser_params *parser)
 {
     ID id = (ID)vtable_size(lvtbl->args) + (ID)vtable_size(lvtbl->vars);
     id += ((tLAST_TOKEN - ID_INTERNAL) >> ID_SCOPE_SHIFT) + 1;
-    return ID_INTERNAL | (id << ID_SCOPE_SHIFT);
+    return ID_STATIC_SYM | ID_INTERNAL | (id << ID_SCOPE_SHIFT);
+}
+
+static inline int
+id_type(ID id)
+{
+    if (id<=tLAST_OP_ID) {
+	return -1;
+    }
+    if (id&ID_STATIC_SYM) {
+	return (int)((id)&ID_SCOPE_MASK);
+    }
+    else {
+	VALUE dsym = (VALUE)id;
+	return RSYMBOL(dsym)->type;
+    }
 }
 
 #ifndef RIPPER
@@ -10353,14 +10395,14 @@ rb_str_symname_type(VALUE name, unsigned int allowed_attrset)
 }
 
 static ID
-register_symid(ID id, const char *name, long len, rb_encoding *enc)
+register_static_symid(ID id, const char *name, long len, rb_encoding *enc)
 {
     VALUE str = rb_enc_str_new(name, len, enc);
-    return register_symid_str(id, str);
+    return register_static_symid_str(id, str);
 }
 
 static ID
-register_symid_str(ID id, VALUE str)
+register_static_symid_str(ID id, VALUE str)
 {
     OBJ_FREEZE(str);
     str = rb_fstring(str);
@@ -10406,8 +10448,35 @@ setup_fake_str(struct RString *fake_str, const char *name, long len)
     return (VALUE)fake_str;
 }
 
+#define ID_DYNAMIC_SYM_P(id) (!(id&ID_STATIC_SYM)&&id>tLAST_TOKEN)
+
 ID
-rb_intern3(const char *name, long len, rb_encoding *enc)
+rb_pin_dynamic_symbol(VALUE sym)
+{
+    /* stick dynamic symbol */
+    if (!st_insert(global_symbols.pinned_dsym, sym, (st_data_t)sym)) {
+	global_symbols.pinned_dsym_minor_marked = 0;
+    }
+    return (ID)sym;
+}
+
+static int
+lookup_sym_id(st_data_t str, st_data_t *data)
+{
+    ID id;
+
+    if (!st_lookup(global_symbols.sym_id, str, data)) {
+	return FALSE;
+    }
+    id = (ID)*data;
+    if (ID_DYNAMIC_SYM_P(id)) {
+	rb_pin_dynamic_symbol((VALUE)id);
+    }
+    return TRUE;
+}
+
+static ID
+intern_cstr_without_pindown(const char *name, long len, rb_encoding *enc)
 {
     st_data_t data;
     struct RString fake_str;
@@ -10422,17 +10491,31 @@ rb_intern3(const char *name, long len, rb_encoding *enc)
     return intern_str(str);
 }
 
+ID
+rb_intern3(const char *name, long len, rb_encoding *enc)
+{
+    ID id;
+
+    id = intern_cstr_without_pindown(name, len, enc);
+    if (ID_DYNAMIC_SYM_P(id)) {
+	rb_pin_dynamic_symbol((VALUE)id);
+    }
+
+    return id;
+}
+
 static ID
 next_id_base(void)
 {
     if (global_symbols.last_id >= ~(ID)0 >> (ID_SCOPE_SHIFT+RUBY_SPECIAL_SHIFT)) {
 	return (ID)-1;
     }
-    return ++global_symbols.last_id << ID_SCOPE_SHIFT;
+    ++global_symbols.last_id;
+    return global_symbols.last_id << ID_SCOPE_SHIFT;
 }
 
 static ID
-intern_str(VALUE str)
+next_id(VALUE str)
 {
     const char *name, *m, *e;
     long len, last;
@@ -10484,13 +10567,13 @@ intern_str(VALUE str)
 
 	    if (len == 1) {
 		id = c;
-		goto id_register;
+                return id;
 	    }
 	    for (i = 0; i < op_tbl_count; i++) {
 		if (*op_tbl[i].name == *m &&
 		    strcmp(op_tbl[i].name, m) == 0) {
 		    id = op_tbl[i].token;
-		    goto id_register;
+                    return id;
 		}
 	    }
 	}
@@ -10501,10 +10584,11 @@ intern_str(VALUE str)
 	if (last > 1 && name[last-1] == '=')
 	    goto junk;
 	id = rb_intern3(name, last, enc);
+	id |= ID_STATIC_SYM;
 	if (id > tLAST_OP_ID && !is_attrset_id(id)) {
 	    enc = rb_enc_get(rb_id2str(id));
 	    id = rb_id_attrset(id);
-	    goto id_register;
+	    return id;
 	}
 	id = ID_ATTRSET;
     }
@@ -10541,8 +10625,14 @@ intern_str(VALUE str)
 	}
     }
     id |= nid;
-  id_register:
+    id |= ID_STATIC_SYM;
-    return register_symid_str(id, str);
+    return id;
+}
+
+static ID
+intern_str(VALUE str)
+{
+    return register_static_symid_str(next_id(str), str);
 }
 
 ID
@@ -10563,11 +10653,113 @@ rb_intern_str(VALUE str)
 {
     st_data_t id;
 
-    if (st_lookup(global_symbols.sym_id, str, &id))
+    if (lookup_sym_id(str, &id))
 	return (ID)id;
     return intern_str(rb_str_dup(str));
 }
 
+void
+rb_gc_free_dsymbol(VALUE ptr)
+{
+    st_delete(global_symbols.sym_id, (st_data_t *)&RSYMBOL(ptr)->fstr, 0);
+    st_delete(global_symbols.id_str, (st_data_t *)&ptr, 0);
+    RSYMBOL(ptr)->fstr = (VALUE)NULL;
+}
+
+VALUE
+rb_str_dynamic_intern(VALUE s)
+{
+    VALUE str = RB_GC_GUARD(s);
+    VALUE dup;
+    rb_encoding *enc;
+    VALUE dsym;
+    ID id, type;
+
+    if (st_lookup(global_symbols.sym_id, str, &id)) {
+	return ID2SYM(id);
+    }
+
+    dup = rb_str_dup(str);
+    enc = rb_enc_get(str);
+    if (rb_enc_asciicompat(enc)) {
+	if (sym_check_asciionly(str)) {
+	    rb_enc_associate(dup, rb_usascii_encoding());
+	}
+    }
+
+    type = rb_str_symname_type(str, TRUE);
+    OBJ_FREEZE(dup);
+    str = rb_fstring(dup);
+    dsym = rb_newobj_of(rb_cSymbol, T_SYMBOL);
+    OBJ_FREEZE(dsym);
+    RSYMBOL(dsym)->fstr = str;
+    RSYMBOL(dsym)->type = type;
+
+    st_add_direct(global_symbols.sym_id, (st_data_t)str, (ID)dsym);
+    st_add_direct(global_symbols.id_str, (ID)dsym, (st_data_t)str);
+    global_symbols.minor_marked = 0;
+
+    if (RUBY_DTRACE_SYMBOL_CREATE_ENABLED()) {
+	RUBY_DTRACE_SYMBOL_CREATE(RSTRING_PTR(str), rb_sourcefile(), rb_sourceline());
+    }
+
+    return dsym;
+}
+
+static int
+lookup_id_str(ID id, st_data_t *data)
+{
+    if (ID_DYNAMIC_SYM_P(id)) {
+	*data = RSYMBOL(id)->fstr;
+	return TRUE;
+    }
+    if (st_lookup(global_symbols.id_str, id, data)) {
+	return TRUE;
+    }
+    return FALSE;
+}
+
+inline ID
+rb_sym2id(VALUE x)
+{
+    if (STATIC_SYM_P(x)) {
+	return RSHIFT((unsigned long)(x),RUBY_SPECIAL_SHIFT);
+    }
+    else {
+	return rb_pin_dynamic_symbol(x);
+    }
+}
+
+inline ID
+rb_sym2id_without_pindown(VALUE x)
+{
+    if (STATIC_SYM_P(x)) {
+	return RSHIFT((unsigned long)(x),RUBY_SPECIAL_SHIFT);
+    }
+    else {
+	return (ID)x;
+    }
+}
+
+inline VALUE
+rb_id2sym(ID x)
+{
+    if (!ID_DYNAMIC_SYM_P(x)) {
+	return ((VALUE)(x)<<RUBY_SPECIAL_SHIFT)|SYMBOL_FLAG;
+    }
+    else {
+	return (VALUE)x;
+    }
+}
+
+
+VALUE
+rb_sym2str(VALUE sym)
+{
+    return rb_id2str(rb_sym2id_without_pindown(sym));
+}
+
+
 VALUE
 rb_id2str(ID id)
 {
@@ -10605,7 +10797,7 @@ rb_id2str(ID id)
 	}
     }
 
-    if (st_lookup(global_symbols.id_str, id, &data)) {
+    if (lookup_id_str(id, &data)) {
         VALUE str = (VALUE)data;
         if (RBASIC(str)->klass == 0)
             RBASIC_SET_CLASS_RAW(str, rb_cString);
@@ -10613,7 +10805,7 @@ rb_id2str(ID id)
     }
 
     if (is_attrset_id(id)) {
-	ID id_stem = (id & ~ID_SCOPE_MASK);
+	ID id_stem = (id & ~ID_SCOPE_MASK) | ID_STATIC_SYM;
 	VALUE str;
 
 	do {
@@ -10627,7 +10819,7 @@ rb_id2str(ID id)
 	} while (0);
 	str = rb_str_dup(str);
 	rb_str_cat(str, "=", 1);
-	register_symid_str(id, str);
+	register_static_symid_str(id, str);
 	if (st_lookup(global_symbols.id_str, id, &data)) {
             VALUE str = (VALUE)data;
             if (RBASIC(str)->klass == 0)
@@ -10650,7 +10842,7 @@ rb_id2name(ID id)
 ID
 rb_make_internal_id(void)
 {
-    return next_id_base() | ID_INTERNAL;
+    return next_id_base() | ID_INTERNAL | ID_STATIC_SYM;
 }
 
 static int
@@ -10740,13 +10932,39 @@ rb_is_junk_id(ID id)
  */
 ID
 rb_check_id(volatile VALUE *namep)
+{
+    ID id;
+
+    id = rb_check_id_without_pindown((VALUE *)namep);
+    if (ID_DYNAMIC_SYM_P(id)) {
+        rb_pin_dynamic_symbol((VALUE)id);
+    }
+
+    return id;
+}
+
+ID
+rb_check_id_cstr(const char *ptr, long len, rb_encoding *enc)
+{
+    ID id;
+
+    id = rb_check_id_cstr_without_pindown(ptr, len, enc);
+    if (ID_DYNAMIC_SYM_P(id)) {
+        rb_pin_dynamic_symbol((VALUE)id);
+    }
+
+    return id;
+}
+
+ID
+rb_check_id_without_pindown(VALUE *namep)
 {
     st_data_t id;
     VALUE tmp;
     VALUE name = *namep;
 
     if (SYMBOL_P(name)) {
-	return SYM2ID(name);
+	return rb_sym2id_without_pindown(name);
     }
     else if (!RB_TYPE_P(name, T_STRING)) {
 	tmp = rb_check_string_type(name);
@@ -10781,7 +10999,7 @@ rb_check_id(volatile VALUE *namep)
 }
 
 ID
-rb_check_id_cstr(const char *ptr, long len, rb_encoding *enc)
+rb_check_id_cstr_without_pindown(const char *ptr, long len, rb_encoding *enc)
 {
     st_data_t id;
     struct RString fake_str;

sprintf.c

@@ -567,7 +567,7 @@ rb_str_format(int argc, const VALUE *argv, VALUE fmt)
 		    rb_enc_raise(enc, rb_eArgError, "named%.*s after <%s>",
 				 len, start, rb_id2name(id));
 		}
-		nextvalue = GETNAMEARG((id = rb_check_id_cstr(start + 1,
+		nextvalue = GETNAMEARG((id = rb_check_id_cstr_without_pindown(start + 1,
 							      len - 2 /* without parenthesis */,
 							      enc),
 					ID2SYM(id)),

string.c

@@ -8345,10 +8345,9 @@ sym_inspect(VALUE sym)
     VALUE str;
     const char *ptr;
     long len;
-    ID id = SYM2ID(sym);
     char *dest;
 
-    sym = rb_id2str(id);
+    sym = rb_sym2str(sym);
     if (!rb_str_symname_p(sym)) {
 	str = rb_str_inspect(sym);
 	len = RSTRING_LEN(str);
@@ -8384,9 +8383,7 @@ sym_inspect(VALUE sym)
 VALUE
 rb_sym_to_s(VALUE sym)
 {
-    ID id = SYM2ID(sym);
+    return str_new_shared(rb_cString, rb_sym2str(sym));
-
-    return str_new_shared(rb_cString, rb_id2str(id));
 }
 
 
@@ -8468,7 +8465,7 @@ sym_to_proc(VALUE sym)
 static VALUE
 sym_succ(VALUE sym)
 {
-    return rb_str_intern(rb_str_succ(rb_sym_to_s(sym)));
+    return rb_str_dynamic_intern(rb_str_succ(rb_sym_to_s(sym)));
 }
 
 /*
@@ -8552,7 +8549,7 @@ sym_aref(int argc, VALUE *argv, VALUE sym)
 static VALUE
 sym_length(VALUE sym)
 {
-    return rb_str_length(rb_id2str(SYM2ID(sym)));
+    return rb_str_length(rb_sym2str(sym));
 }
 
 /*
@@ -8565,7 +8562,7 @@ sym_length(VALUE sym)
 static VALUE
 sym_empty(VALUE sym)
 {
-    return rb_str_empty(rb_id2str(SYM2ID(sym)));
+    return rb_str_empty(rb_sym2str(sym));
 }
 
 /*
@@ -8578,7 +8575,7 @@ sym_empty(VALUE sym)
 static VALUE
 sym_upcase(VALUE sym)
 {
-    return rb_str_intern(rb_str_upcase(rb_id2str(SYM2ID(sym))));
+    return rb_str_dynamic_intern(rb_str_upcase(rb_sym2str(sym)));
 }
 
 /*
@@ -8591,7 +8588,7 @@ sym_upcase(VALUE sym)
 static VALUE
 sym_downcase(VALUE sym)
 {
-    return rb_str_intern(rb_str_downcase(rb_id2str(SYM2ID(sym))));
+    return rb_str_dynamic_intern(rb_str_downcase(rb_sym2str(sym)));
 }
 
 /*
@@ -8604,7 +8601,7 @@ sym_downcase(VALUE sym)
 static VALUE
 sym_capitalize(VALUE sym)
 {
-    return rb_str_intern(rb_str_capitalize(rb_id2str(SYM2ID(sym))));
+    return rb_str_dynamic_intern(rb_str_capitalize(rb_sym2str(sym)));
 }
 
 /*
@@ -8617,7 +8614,7 @@ sym_capitalize(VALUE sym)
 static VALUE
 sym_swapcase(VALUE sym)
 {
-    return rb_str_intern(rb_str_swapcase(rb_id2str(SYM2ID(sym))));
+    return rb_str_dynamic_intern(rb_str_swapcase(rb_sym2str(sym)));
 }
 
 /*
@@ -8630,7 +8627,7 @@ sym_swapcase(VALUE sym)
 static VALUE
 sym_encoding(VALUE sym)
 {
-    return rb_obj_encoding(rb_id2str(SYM2ID(sym)));
+    return rb_obj_encoding(rb_sym2str(sym));
 }
 
 ID
@@ -8742,8 +8739,8 @@ Init_String(void)
     rb_define_method(rb_cString, "<<", rb_str_concat, 1);
     rb_define_method(rb_cString, "prepend", rb_str_prepend, 1);
     rb_define_method(rb_cString, "crypt", rb_str_crypt, 1);
-    rb_define_method(rb_cString, "intern", rb_str_intern, 0);
+    rb_define_method(rb_cString, "intern", rb_str_dynamic_intern, 0); /* in parse.y */
-    rb_define_method(rb_cString, "to_sym", rb_str_intern, 0);
+    rb_define_method(rb_cString, "to_sym", rb_str_dynamic_intern, 0); /* in parse.y */
     rb_define_method(rb_cString, "ord", rb_str_ord, 0);
 
     rb_define_method(rb_cString, "include?", rb_str_include, 1);

struct.c

@@ -749,7 +749,7 @@ rb_struct_aref(VALUE s, VALUE idx)
 	return rb_struct_aref_id(s, SYM2ID(idx));
     }
     else if (RB_TYPE_P(idx, T_STRING)) {
-	ID id = rb_check_id(&idx);
+	ID id = rb_check_id_without_pindown(&idx);
 	if (!id) {
 	    rb_name_error_str(idx, "no member '%"PRIsVALUE"' in struct",
 			      QUOTE(idx));

template/id.h.tmpl

@@ -29,18 +29,20 @@ types = ids.keys.grep(/^[A-Z]/)
 #define RUBY_ID_H
 
 enum ruby_id_types {
+    RUBY_ID_STATIC_SYM  = 0x01,
     RUBY_ID_LOCAL       = 0x00,
-    RUBY_ID_INSTANCE    = 0x01,
+    RUBY_ID_INSTANCE    = (0x01<<1),
-    RUBY_ID_GLOBAL      = 0x03,
+    RUBY_ID_GLOBAL      = (0x03<<1),
-    RUBY_ID_ATTRSET     = 0x04,
+    RUBY_ID_ATTRSET     = (0x04<<1),
-    RUBY_ID_CONST       = 0x05,
+    RUBY_ID_CONST       = (0x05<<1),
-    RUBY_ID_CLASS       = 0x06,
+    RUBY_ID_CLASS       = (0x06<<1),
-    RUBY_ID_JUNK        = 0x07,
+    RUBY_ID_JUNK        = (0x07<<1),
     RUBY_ID_INTERNAL    = RUBY_ID_JUNK,
-    RUBY_ID_SCOPE_SHIFT = 3,
+    RUBY_ID_SCOPE_SHIFT = 4,
-    RUBY_ID_SCOPE_MASK  = ~(~0U<<RUBY_ID_SCOPE_SHIFT)
+    RUBY_ID_SCOPE_MASK  = (~(~0U<<(RUBY_ID_SCOPE_SHIFT-1))<<1)
 };
 
+#define ID_STATIC_SYM  RUBY_ID_STATIC_SYM
 #define ID_SCOPE_SHIFT RUBY_ID_SCOPE_SHIFT
 #define ID_SCOPE_MASK  RUBY_ID_SCOPE_MASK
 #define ID_LOCAL       RUBY_ID_LOCAL
@@ -99,7 +101,7 @@ enum ruby_method_ids {
 % types.each do |type|
 %   types = ids[type] or next
 %   types.empty? and next
-#define TOKEN2<%=type%>ID(n) id##n = ((t##n<<ID_SCOPE_SHIFT)|ID_<%=type%>)
+#define TOKEN2<%=type%>ID(n) id##n = ((t##n<<ID_SCOPE_SHIFT)|ID_<%=type%>|ID_STATIC_SYM)
 %   types.each do |token|
     TOKEN2<%=type%>ID(<%=token%>),
 %   end

thread.c

@@ -2776,7 +2776,7 @@ rb_thread_local_aref(VALUE thread, ID id)
 static VALUE
 rb_thread_aref(VALUE thread, VALUE key)
 {
-    ID id = rb_check_id(&key);
+    ID id = rb_check_id_without_pindown(&key);
     if (!id) return Qnil;
     return rb_thread_local_aref(thread, id);
 }
@@ -2853,7 +2853,7 @@ static VALUE
 rb_thread_variable_get(VALUE thread, VALUE key)
 {
     VALUE locals;
-    ID id = rb_check_id(&key);
+    ID id = rb_check_id_without_pindown(&key);
 
     if (!id) return Qnil;
     locals = rb_ivar_get(thread, id_locals);
@@ -2899,7 +2899,7 @@ static VALUE
 rb_thread_key_p(VALUE self, VALUE key)
 {
     rb_thread_t *th;
-    ID id = rb_check_id(&key);
+    ID id = rb_check_id_without_pindown(&key);
 
     GetThreadPtr(self, th);
 
@@ -3020,7 +3020,7 @@ static VALUE
 rb_thread_variable_p(VALUE thread, VALUE key)
 {
     VALUE locals;
-    ID id = rb_check_id(&key);
+    ID id = rb_check_id_without_pindown(&key);
 
     if (!id) return Qfalse;
 

variable.c

@@ -353,7 +353,7 @@ rb_path_to_class(VALUE pathname)
     }
     while (*p) {
 	while (*p && *p != ':') p++;
-	id = rb_check_id_cstr(pbeg, p-pbeg, enc);
+	id = rb_check_id_cstr_without_pindown(pbeg, p-pbeg, enc);
 	if (p[0] == ':') {
 	    if (p[1] != ':') goto undefined_class;
 	    p += 2;
@@ -1403,7 +1403,7 @@ VALUE
 rb_obj_remove_instance_variable(VALUE obj, VALUE name)
 {
     VALUE val = Qnil;
-    const ID id = rb_check_id(&name);
+    const ID id = rb_check_id_without_pindown(&name);
     st_data_t n, v;
     struct st_table *iv_index_tbl;
     st_data_t index;
@@ -1919,7 +1919,7 @@ rb_public_const_get_at(VALUE klass, ID id)
 VALUE
 rb_mod_remove_const(VALUE mod, VALUE name)
 {
-    const ID id = rb_check_id(&name);
+    const ID id = rb_check_id_without_pindown(&name);
 
     if (!id) {
 	if (rb_is_const_name(name)) {
@@ -2568,7 +2568,7 @@ rb_mod_class_variables(int argc, VALUE *argv, VALUE mod)
 VALUE
 rb_mod_remove_cvar(VALUE mod, VALUE name)
 {
-    const ID id = rb_check_id(&name);
+    const ID id = rb_check_id_without_pindown(&name);
     st_data_t val, n = id;
 
     if (!id) {

vm_method.c

@@ -791,7 +791,7 @@ rb_mod_remove_method(int argc, VALUE *argv, VALUE mod)
 
     for (i = 0; i < argc; i++) {
 	VALUE v = argv[i];
-	ID id = rb_check_id(&v);
+	ID id = rb_check_id_without_pindown(&v);
 	if (!id) {
 	    rb_name_error_str(v, "method `%s' not defined in %s",
 			      RSTRING_PTR(v), rb_class2name(mod));
@@ -1002,7 +1002,7 @@ rb_mod_undef_method(int argc, VALUE *argv, VALUE mod)
     int i;
     for (i = 0; i < argc; i++) {
 	VALUE v = argv[i];
-	ID id = rb_check_id(&v);
+	ID id = rb_check_id_without_pindown(&v);
 	if (!id) {
 	    rb_method_name_error(mod, v);
 	}
@@ -1042,7 +1042,7 @@ rb_mod_undef_method(int argc, VALUE *argv, VALUE mod)
 static VALUE
 rb_mod_method_defined(VALUE mod, VALUE mid)
 {
-    ID id = rb_check_id(&mid);
+    ID id = rb_check_id_without_pindown(&mid);
     if (!id || !rb_method_boundp(mod, id, 1)) {
 	return Qfalse;
     }
@@ -1056,7 +1056,7 @@ static VALUE
 check_definition(VALUE mod, VALUE mid, rb_method_flag_t noex)
 {
     const rb_method_entry_t *me;
-    ID id = rb_check_id(&mid);
+    ID id = rb_check_id_without_pindown(&mid);
     if (!id) return Qfalse;
     me = rb_method_entry(mod, id, 0);
     if (me) {
@@ -1685,7 +1685,7 @@ obj_respond_to(int argc, VALUE *argv, VALUE obj)
     ID id;
 
     rb_scan_args(argc, argv, "11", &mid, &priv);
-    if (!(id = rb_check_id(&mid))) {
+    if (!(id = rb_check_id_without_pindown(&mid))) {
 	if (!rb_method_basic_definition_p(CLASS_OF(obj), idRespond_to_missing)) {
 	    VALUE args[2];
 	    args[0] = ID2SYM(rb_to_id(mid));