[pull] master from ruby:master

ast.c

@@ -866,6 +866,11 @@ node_locations(VALUE ast_value, const NODE *node)
                                     location_new(&RNODE_IF(node)->if_keyword_loc),
                                     location_new(&RNODE_IF(node)->then_keyword_loc),
                                     location_new(&RNODE_IF(node)->end_keyword_loc));
+      case NODE_MODULE:
+        return rb_ary_new_from_args(3,
+                                    location_new(nd_code_loc(node)),
+                                    location_new(&RNODE_MODULE(node)->module_keyword_loc),
+                                    location_new(&RNODE_MODULE(node)->end_keyword_loc));
       case NODE_NEXT:
         return rb_ary_new_from_args(2,
                                     location_new(nd_code_loc(node)),

autogen.sh

@@ -10,7 +10,7 @@ case "$0" in
     *   )  srcdir="";; # Otherwise
 esac
 
-# If install-only is explicitly requested, disbale symlink flags
+# If install-only is explicitly requested, disable symlink flags
 case " $* " in
     *" -i "* | *" --install"* ) symlink_flags="" ;;
     *                         ) symlink_flags="--install --symlink" ;;

benchmark/struct_accessor.yml

@@ -1,5 +1,12 @@
 prelude: |
   C = Struct.new(:x) do
+    def initialize(...)
+      super
+      @ivar = 42
+    end
+
+    attr_accessor :ivar
+
     class_eval <<-END
       def r
         #{'x;'*256}
@@ -15,11 +22,16 @@ prelude: |
         m = method(:x=)
         #{'m.call(nil);'*256}
       end
+      def r_ivar
+        #{'ivar;'*256}
+      end
     END
   end
+  C.new(nil) # ensure common shape is known
   obj = C.new(nil)
 benchmark:
   member_reader: "obj.r"
   member_writer: "obj.w"
   member_reader_method: "obj.rm"
   member_writer_method: "obj.wm"
+  ivar_reader: "obj.r_ivar"

depend

@@ -6065,6 +6065,7 @@ hash.$(OBJEXT): $(top_srcdir)/internal/set_table.h
 hash.$(OBJEXT): $(top_srcdir)/internal/st.h
 hash.$(OBJEXT): $(top_srcdir)/internal/static_assert.h
 hash.$(OBJEXT): $(top_srcdir)/internal/string.h
+hash.$(OBJEXT): $(top_srcdir)/internal/struct.h
 hash.$(OBJEXT): $(top_srcdir)/internal/symbol.h
 hash.$(OBJEXT): $(top_srcdir)/internal/thread.h
 hash.$(OBJEXT): $(top_srcdir)/internal/time.h
@@ -6288,6 +6289,7 @@ hash.$(OBJEXT): {$(VPATH)}symbol.h
 hash.$(OBJEXT): {$(VPATH)}thread_$(THREAD_MODEL).h
 hash.$(OBJEXT): {$(VPATH)}thread_native.h
 hash.$(OBJEXT): {$(VPATH)}util.h
+hash.$(OBJEXT): {$(VPATH)}variable.h
 hash.$(OBJEXT): {$(VPATH)}vm_core.h
 hash.$(OBJEXT): {$(VPATH)}vm_debug.h
 hash.$(OBJEXT): {$(VPATH)}vm_opts.h
@@ -12926,6 +12928,7 @@ range.$(OBJEXT): $(top_srcdir)/internal/enumerator.h
 range.$(OBJEXT): $(top_srcdir)/internal/error.h
 range.$(OBJEXT): $(top_srcdir)/internal/fixnum.h
 range.$(OBJEXT): $(top_srcdir)/internal/gc.h
+range.$(OBJEXT): $(top_srcdir)/internal/imemo.h
 range.$(OBJEXT): $(top_srcdir)/internal/numeric.h
 range.$(OBJEXT): $(top_srcdir)/internal/range.h
 range.$(OBJEXT): $(top_srcdir)/internal/serial.h
@@ -12948,6 +12951,7 @@ range.$(OBJEXT): {$(VPATH)}config.h
 range.$(OBJEXT): {$(VPATH)}defines.h
 range.$(OBJEXT): {$(VPATH)}encoding.h
 range.$(OBJEXT): {$(VPATH)}id.h
+range.$(OBJEXT): {$(VPATH)}id_table.h
 range.$(OBJEXT): {$(VPATH)}intern.h
 range.$(OBJEXT): {$(VPATH)}internal.h
 range.$(OBJEXT): {$(VPATH)}internal/abi.h
@@ -15580,6 +15584,7 @@ shape.$(OBJEXT): $(top_srcdir)/internal/serial.h
 shape.$(OBJEXT): $(top_srcdir)/internal/set_table.h
 shape.$(OBJEXT): $(top_srcdir)/internal/static_assert.h
 shape.$(OBJEXT): $(top_srcdir)/internal/string.h
+shape.$(OBJEXT): $(top_srcdir)/internal/struct.h
 shape.$(OBJEXT): $(top_srcdir)/internal/symbol.h
 shape.$(OBJEXT): $(top_srcdir)/internal/variable.h
 shape.$(OBJEXT): $(top_srcdir)/internal/vm.h
@@ -16569,6 +16574,7 @@ string.$(OBJEXT): $(top_srcdir)/internal/serial.h
 string.$(OBJEXT): $(top_srcdir)/internal/set_table.h
 string.$(OBJEXT): $(top_srcdir)/internal/static_assert.h
 string.$(OBJEXT): $(top_srcdir)/internal/string.h
+string.$(OBJEXT): $(top_srcdir)/internal/struct.h
 string.$(OBJEXT): $(top_srcdir)/internal/transcode.h
 string.$(OBJEXT): $(top_srcdir)/internal/variable.h
 string.$(OBJEXT): $(top_srcdir)/internal/vm.h
@@ -16766,6 +16772,7 @@ string.$(OBJEXT): {$(VPATH)}thread.h
 string.$(OBJEXT): {$(VPATH)}thread_$(THREAD_MODEL).h
 string.$(OBJEXT): {$(VPATH)}thread_native.h
 string.$(OBJEXT): {$(VPATH)}util.h
+string.$(OBJEXT): {$(VPATH)}variable.h
 string.$(OBJEXT): {$(VPATH)}vm_core.h
 string.$(OBJEXT): {$(VPATH)}vm_debug.h
 string.$(OBJEXT): {$(VPATH)}vm_opts.h
@@ -18103,6 +18110,7 @@ variable.$(OBJEXT): $(top_srcdir)/internal/serial.h
 variable.$(OBJEXT): $(top_srcdir)/internal/set_table.h
 variable.$(OBJEXT): $(top_srcdir)/internal/static_assert.h
 variable.$(OBJEXT): $(top_srcdir)/internal/string.h
+variable.$(OBJEXT): $(top_srcdir)/internal/struct.h
 variable.$(OBJEXT): $(top_srcdir)/internal/symbol.h
 variable.$(OBJEXT): $(top_srcdir)/internal/thread.h
 variable.$(OBJEXT): $(top_srcdir)/internal/variable.h

gc.c

@@ -3260,6 +3260,10 @@ rb_gc_mark_children(void *objspace, VALUE obj)
             gc_mark_internal(ptr[i]);
         }
 
+        if (!FL_TEST_RAW(obj, RSTRUCT_GEN_FIELDS)) {
+            gc_mark_internal(RSTRUCT_FIELDS_OBJ(obj));
+        }
+
         break;
       }
 
@@ -4188,6 +4192,15 @@ rb_gc_update_object_references(void *objspace, VALUE obj)
             for (i = 0; i < len; i++) {
                 UPDATE_IF_MOVED(objspace, ptr[i]);
             }
+
+            if (RSTRUCT_EMBED_LEN(obj)) {
+                if (!FL_TEST_RAW(obj, RSTRUCT_GEN_FIELDS)) {
+                    UPDATE_IF_MOVED(objspace, ptr[len]);
+                }
+            }
+            else {
+                UPDATE_IF_MOVED(objspace, RSTRUCT(obj)->as.heap.fields_obj);
+            }
         }
         break;
       default:

internal/struct.h

@@ -11,10 +11,23 @@
 #include "ruby/internal/stdbool.h"     /* for bool */
 #include "ruby/ruby.h"          /* for struct RBasic */
 
+/* Flags of RStruct
+ *
+ * 1-7: RSTRUCT_EMBED_LEN
+ *          If non-zero, the struct is embedded (its contents follow the
+ *          header, rather than being on a separately allocated buffer) and
+ *          these bits are the length of the Struct.
+ * 8:   RSTRUCT_GEN_FIELDS
+ *          The struct is embedded and has no space left to store the
+ *          IMEMO/fields reference. Any ivar this struct may have will be in
+ *          the generic_fields_tbl. This flag doesn't imply the struct has
+ *          ivars.
+ */
 enum {
     RSTRUCT_EMBED_LEN_MASK = RUBY_FL_USER7 | RUBY_FL_USER6 | RUBY_FL_USER5 | RUBY_FL_USER4 |
                                  RUBY_FL_USER3 | RUBY_FL_USER2 | RUBY_FL_USER1,
     RSTRUCT_EMBED_LEN_SHIFT = (RUBY_FL_USHIFT+1),
+    RSTRUCT_GEN_FIELDS = RUBY_FL_USER8,
 };
 
 struct RStruct {
@@ -23,6 +36,7 @@ struct RStruct {
         struct {
             long len;
             const VALUE *ptr;
+            VALUE fields_obj;
         } heap;
         /* This is a length 1 array because:
          *   1. GCC has a bug that does not optimize C flexible array members
@@ -116,4 +130,31 @@ RSTRUCT_GET(VALUE st, long k)
     return RSTRUCT_CONST_PTR(st)[k];
 }
 
+static inline VALUE
+RSTRUCT_FIELDS_OBJ(VALUE st)
+{
+    const long embed_len = RSTRUCT_EMBED_LEN(st);
+    VALUE fields_obj;
+    if (embed_len) {
+        RUBY_ASSERT(!FL_TEST_RAW(st, RSTRUCT_GEN_FIELDS));
+        fields_obj = RSTRUCT_GET(st, embed_len);
+    }
+    else {
+        fields_obj = RSTRUCT(st)->as.heap.fields_obj;
+    }
+    return fields_obj;
+}
+
+static inline void
+RSTRUCT_SET_FIELDS_OBJ(VALUE st, VALUE fields_obj)
+{
+    const long embed_len = RSTRUCT_EMBED_LEN(st);
+    if (embed_len) {
+        RUBY_ASSERT(!FL_TEST_RAW(st, RSTRUCT_GEN_FIELDS));
+        RSTRUCT_SET(st, embed_len, fields_obj);
+    }
+    else {
+        RB_OBJ_WRITE(st, &RSTRUCT(st)->as.heap.fields_obj, fields_obj);
+    }
+}
 #endif /* INTERNAL_STRUCT_H */

node_dump.c

@@ -1009,8 +1009,10 @@ dump_node(VALUE buf, VALUE indent, int comment, const NODE * node)
         ANN("format: module [nd_cpath]; [nd_body]; end");
         ANN("example: module M; ..; end");
         F_NODE(nd_cpath, RNODE_MODULE, "module path");
-        LAST_NODE;
         F_NODE(nd_body, RNODE_MODULE, "module definition");
+        F_LOC(module_keyword_loc, RNODE_MODULE);
+        LAST_NODE;
+        F_LOC(end_keyword_loc, RNODE_MODULE);
         return;
 
       case NODE_SCLASS:

parse.y

@@ -1145,7 +1145,7 @@ static rb_node_alias_t *rb_node_alias_new(struct parser_params *p, NODE *nd_1st,
 static rb_node_valias_t *rb_node_valias_new(struct parser_params *p, ID nd_alias, ID nd_orig, const YYLTYPE *loc, const YYLTYPE *keyword_loc);
 static rb_node_undef_t *rb_node_undef_new(struct parser_params *p, NODE *nd_undef, const YYLTYPE *loc);
 static rb_node_class_t *rb_node_class_new(struct parser_params *p, NODE *nd_cpath, NODE *nd_body, NODE *nd_super, const YYLTYPE *loc, const YYLTYPE *class_keyword_loc, const YYLTYPE *inheritance_operator_loc, const YYLTYPE *end_keyword_loc);
-static rb_node_module_t *rb_node_module_new(struct parser_params *p, NODE *nd_cpath, NODE *nd_body, const YYLTYPE *loc);
+static rb_node_module_t *rb_node_module_new(struct parser_params *p, NODE *nd_cpath, NODE *nd_body, const YYLTYPE *loc, const YYLTYPE *module_keyword_loc, const YYLTYPE *end_keyword_loc);
 static rb_node_sclass_t *rb_node_sclass_new(struct parser_params *p, NODE *nd_recv, NODE *nd_body, const YYLTYPE *loc);
 static rb_node_colon2_t *rb_node_colon2_new(struct parser_params *p, NODE *nd_head, ID nd_mid, const YYLTYPE *loc, const YYLTYPE *delimiter_loc, const YYLTYPE *name_loc);
 static rb_node_colon3_t *rb_node_colon3_new(struct parser_params *p, ID nd_mid, const YYLTYPE *loc, const YYLTYPE *delimiter_loc, const YYLTYPE *name_loc);
@@ -1253,7 +1253,7 @@ static rb_node_error_t *rb_node_error_new(struct parser_params *p, const YYLTYPE
 #define NEW_VALIAS(n,o,loc,k_loc) (NODE *)rb_node_valias_new(p,n,o,loc,k_loc)
 #define NEW_UNDEF(i,loc) (NODE *)rb_node_undef_new(p,i,loc)
 #define NEW_CLASS(n,b,s,loc,ck_loc,io_loc,ek_loc) (NODE *)rb_node_class_new(p,n,b,s,loc,ck_loc,io_loc,ek_loc)
-#define NEW_MODULE(n,b,loc) (NODE *)rb_node_module_new(p,n,b,loc)
+#define NEW_MODULE(n,b,loc,mk_loc,ek_loc) (NODE *)rb_node_module_new(p,n,b,loc,mk_loc,ek_loc)
 #define NEW_SCLASS(r,b,loc) (NODE *)rb_node_sclass_new(p,r,b,loc)
 #define NEW_COLON2(c,i,loc,d_loc,n_loc) (NODE *)rb_node_colon2_new(p,c,i,loc,d_loc,n_loc)
 #define NEW_COLON3(i,loc,d_loc,n_loc) (NODE *)rb_node_colon3_new(p,i,loc,d_loc,n_loc)
@@ -4621,7 +4621,7 @@ primary		: inline_primary
               bodystmt
               k_end
                 {
-                    $$ = NEW_MODULE($cpath, $bodystmt, &@$);
+                    $$ = NEW_MODULE($cpath, $bodystmt, &@$, &@k_module, &@k_end);
                     nd_set_line(RNODE_MODULE($$)->nd_body, @k_end.end_pos.lineno);
                     set_line_body($bodystmt, @cpath.end_pos.lineno);
                     nd_set_line($$, @cpath.end_pos.lineno);
@@ -11438,13 +11438,15 @@ rb_node_sclass_new(struct parser_params *p, NODE *nd_recv, NODE *nd_body, const
 }
 
 static rb_node_module_t *
-rb_node_module_new(struct parser_params *p, NODE *nd_cpath, NODE *nd_body, const YYLTYPE *loc)
+rb_node_module_new(struct parser_params *p, NODE *nd_cpath, NODE *nd_body, const YYLTYPE *loc, const YYLTYPE *module_keyword_loc, const YYLTYPE *end_keyword_loc)
 {
     /* Keep the order of node creation */
     NODE *scope = NEW_SCOPE(0, nd_body, loc);
     rb_node_module_t *n = NODE_NEWNODE(NODE_MODULE, rb_node_module_t, loc);
     n->nd_cpath = nd_cpath;
     n->nd_body = scope;
+    n->module_keyword_loc = *module_keyword_loc;
+    n->end_keyword_loc = *end_keyword_loc;
 
     return n;
 }

ractor.c

@@ -2302,7 +2302,7 @@ static const rb_data_type_t cross_ractor_require_data_type = {
         NULL, // memsize
         NULL, // compact
     },
-    0, 0, RUBY_TYPED_FREE_IMMEDIATELY | RUBY_TYPED_WB_PROTECTED | RUBY_TYPED_DECL_MARKING
+    0, 0, RUBY_TYPED_FREE_IMMEDIATELY | RUBY_TYPED_WB_PROTECTED | RUBY_TYPED_DECL_MARKING | RUBY_TYPED_EMBEDDABLE
 };
 
 static VALUE

ractor_sync.c

@@ -1273,7 +1273,7 @@ static const rb_data_type_t ractor_selector_data_type = {
         ractor_selector_memsize,
         NULL, // update
     },
-    0, 0, RUBY_TYPED_FREE_IMMEDIATELY,
+    0, 0, RUBY_TYPED_FREE_IMMEDIATELY | RUBY_TYPED_WB_PROTECTED,
 };
 
 static struct ractor_selector *
@@ -1318,6 +1318,8 @@ ractor_selector_add(VALUE selv, VALUE rpv)
     }
 
     st_insert(s->ports, (st_data_t)rpv, (st_data_t)rp);
+    RB_OBJ_WRITTEN(selv, Qundef, rpv);
+
     return selv;
 }
 

rubyparser.h

@@ -901,6 +901,8 @@ typedef struct RNode_MODULE {
 
     struct RNode *nd_cpath;
     struct RNode *nd_body;
+    rb_code_location_t module_keyword_loc;
+    rb_code_location_t end_keyword_loc;
 } rb_node_module_t;
 
 typedef struct RNode_SCLASS {

struct.c

@@ -811,13 +811,22 @@ struct_alloc(VALUE klass)
 {
     long n = num_members(klass);
     size_t embedded_size = offsetof(struct RStruct, as.ary) + (sizeof(VALUE) * n);
+    if (RCLASS_MAX_IV_COUNT(klass) > 0) {
+        embedded_size += sizeof(VALUE);
+    }
+
     VALUE flags = T_STRUCT | (RGENGC_WB_PROTECTED_STRUCT ? FL_WB_PROTECTED : 0);
 
     if (n > 0 && rb_gc_size_allocatable_p(embedded_size)) {
         flags |= n << RSTRUCT_EMBED_LEN_SHIFT;
 
         NEWOBJ_OF(st, struct RStruct, klass, flags, embedded_size, 0);
-
+        if (RCLASS_MAX_IV_COUNT(klass) == 0 && embedded_size == rb_gc_obj_slot_size((VALUE)st)) {
+            FL_SET_RAW((VALUE)st, RSTRUCT_GEN_FIELDS);
+        }
+        else {
+            RSTRUCT_SET_FIELDS_OBJ((VALUE)st, 0);
+        }
         rb_mem_clear((VALUE *)st->as.ary, n);
 
         return (VALUE)st;

symbol.c

@@ -99,7 +99,9 @@ typedef struct {
     VALUE ids;
 } rb_symbols_t;
 
-rb_symbols_t ruby_global_symbols = {tNEXT_ID-1};
+rb_symbols_t ruby_global_symbols = {
+    .next_id = tNEXT_ID,
+};
 
 struct sym_set_static_sym_entry {
     VALUE sym;

test/ruby/test_ast.rb

@@ -1491,6 +1491,11 @@ def test_lambda_locations
       assert_locations(node.children[-1].locations, [[1, 0, 1, 20], [1, 0, 1, 2], [1, 10, 1, 12], [1, 17, 1, 20]])
     end
 
+    def test_module_locations
+      node = ast_parse('module A end')
+      assert_locations(node.children[-1].locations, [[1, 0, 1, 12], [1, 0, 1, 6], [1, 9, 1, 12]])
+    end
+
     def test_if_locations
       node = ast_parse("if cond then 1 else 2 end")
       assert_locations(node.children[-1].locations, [[1, 0, 1, 25], [1, 0, 1, 2], [1, 8, 1, 12], [1, 22, 1, 25]])

test/ruby/test_object_id.rb

@@ -252,3 +252,52 @@ def initialize
     end;
   end
 end
+
+class TestObjectIdStruct < TestObjectId
+  EmbeddedStruct = Struct.new(:embedded_field)
+
+  def setup
+    @obj = EmbeddedStruct.new
+  end
+end
+
+class TestObjectIdStructGenIvar < TestObjectId
+  GenIvarStruct = Struct.new(:a, :b, :c)
+
+  def setup
+    @obj = GenIvarStruct.new
+  end
+end
+
+class TestObjectIdStructNotEmbed < TestObjectId
+  MANY_IVS = 80
+
+  StructNotEmbed = Struct.new(*MANY_IVS.times.map { |i| :"field_#{i}" })
+
+  def setup
+    @obj = StructNotEmbed.new
+  end
+end
+
+class TestObjectIdStructTooComplex < TestObjectId
+  StructTooComplex = Struct.new(:a) do
+    def initialize
+      @too_complex_obj_id_test = 1
+    end
+  end
+
+  def setup
+    if defined?(RubyVM::Shape::SHAPE_MAX_VARIATIONS)
+      assert_equal 8, RubyVM::Shape::SHAPE_MAX_VARIATIONS
+    end
+    8.times do |i|
+      StructTooComplex.new.instance_variable_set("@TestObjectIdStructTooComplex#{i}", 1)
+    end
+    @obj = StructTooComplex.new
+    @obj.instance_variable_set("@a#{rand(10_000)}", 1)
+
+    if defined?(RubyVM::Shape)
+      assert_predicate(RubyVM::Shape.of(@obj), :too_complex?)
+    end
+  end
+end