diff options
Diffstat (limited to 'marshal.c')
| -rw-r--r-- | marshal.c | 3069 |
1 files changed, 2112 insertions, 957 deletions
@@ -3,17 +3,13 @@ marshal.c - $Author$ - $Date$ created at: Thu Apr 27 16:30:01 JST 1995 - Copyright (C) 1993-2003 Yukihiro Matsumoto + Copyright (C) 1993-2007 Yukihiro Matsumoto **********************************************************************/ -#include "ruby.h" -#include "rubyio.h" -#include "st.h" -#include "util.h" +#include "ruby/internal/config.h" #include <math.h> #ifdef HAVE_FLOAT_H @@ -23,27 +19,49 @@ #include <ieeefp.h> #endif +#include "encindex.h" +#include "id_table.h" +#include "internal.h" +#include "internal/array.h" +#include "internal/bignum.h" +#include "internal/class.h" +#include "internal/encoding.h" +#include "internal/error.h" +#include "internal/hash.h" +#include "internal/numeric.h" +#include "internal/object.h" +#include "internal/re.h" +#include "internal/struct.h" +#include "internal/symbol.h" +#include "internal/util.h" +#include "internal/vm.h" +#include "ruby/io.h" +#include "ruby/ruby.h" +#include "ruby/st.h" +#include "ruby/util.h" +#include "builtin.h" +#include "shape.h" +#include "ruby/internal/attr/nonstring.h" + #define BITSPERSHORT (2*CHAR_BIT) #define SHORTMASK ((1<<BITSPERSHORT)-1) -#define SHORTDN(x) RSHIFT(x,BITSPERSHORT) +#define SHORTDN(x) RSHIFT((x),BITSPERSHORT) -#if SIZEOF_SHORT == SIZEOF_BDIGITS +#if SIZEOF_SHORT == SIZEOF_BDIGIT #define SHORTLEN(x) (x) #else -static int -shortlen(len, ds) - long len; - BDIGIT *ds; +static size_t +shortlen(size_t len, BDIGIT *ds) { BDIGIT num; int offset = 0; num = ds[len-1]; while (num) { - num = SHORTDN(num); - offset++; + num = SHORTDN(num); + offset++; } - return (len - 1)*sizeof(BDIGIT)/2 + offset; + return (len - 1)*SIZEOF_BDIGIT/2 + offset; } #define SHORTLEN(x) shortlen((x),d) #endif @@ -81,15 +99,71 @@ shortlen(len, ds) #define TYPE_LINK '@' static ID s_dump, s_load, s_mdump, s_mload; -static ID s_dump_data, s_load_data, s_alloc; -static ID s_getc, s_read, s_write, s_binmode; +static ID s_dump_data, s_load_data, s_alloc, s_call; +static ID s_getbyte, s_read, s_write, s_binmode; +static ID s_encoding_short, s_ruby2_keywords_flag; +#define s_encoding_long rb_id_encoding() + +#define name_s_dump "_dump" +#define name_s_load "_load" +#define name_s_mdump "marshal_dump" +#define name_s_mload "marshal_load" +#define name_s_dump_data "_dump_data" +#define name_s_load_data "_load_data" +#define name_s_alloc "_alloc" +#define name_s_call "call" +#define name_s_getbyte "getbyte" +#define name_s_read "read" +#define name_s_write "write" +#define name_s_binmode "binmode" +#define name_s_encoding_short "E" +#define name_s_encoding_long "encoding" +#define name_s_ruby2_keywords_flag "K" + +typedef struct { + VALUE newclass; + VALUE oldclass; + VALUE (*dumper)(VALUE); + VALUE (*loader)(VALUE, VALUE); +} marshal_compat_t; + +static st_table *compat_allocator_tbl; +static VALUE compat_allocator_tbl_wrapper; +static VALUE rb_marshal_dump_limited(VALUE obj, VALUE port, int limit); +static VALUE rb_marshal_load_with_proc(VALUE port, VALUE proc, bool freeze); + +static st_table *compat_allocator_table(void); + +void +rb_marshal_define_compat(VALUE newclass, VALUE oldclass, VALUE (*dumper)(VALUE), VALUE (*loader)(VALUE, VALUE)) +{ + marshal_compat_t *compat; + rb_alloc_func_t allocator = rb_get_alloc_func(newclass); + + if (!allocator) { + rb_raise(rb_eTypeError, "no allocator"); + } + + compat_allocator_table(); + compat = ALLOC(marshal_compat_t); + compat->newclass = newclass; + compat->oldclass = oldclass; + compat->dumper = dumper; + compat->loader = loader; + + st_insert(compat_allocator_table(), (st_data_t)allocator, (st_data_t)compat); + RB_OBJ_WRITTEN(compat_allocator_tbl_wrapper, Qundef, newclass); + RB_OBJ_WRITTEN(compat_allocator_tbl_wrapper, Qundef, oldclass); +} struct dump_arg { - VALUE obj; VALUE str, dest; st_table *symbols; st_table *data; - int taint; + st_table *compat_tbl; + st_table *encodings; + st_table *userdefs; + st_index_t num_entries; }; struct dump_call_arg { @@ -99,110 +173,188 @@ struct dump_call_arg { }; static VALUE -class2path(klass) - VALUE klass; +check_dump_arg(VALUE ret, struct dump_arg *arg, const char *name) { - VALUE path = rb_class_path(klass); - char *n = RSTRING(path)->ptr; + if (!arg->symbols) { + rb_raise(rb_eRuntimeError, "Marshal.dump reentered at %s", + name); + } + return ret; +} + +static VALUE +check_userdump_arg(VALUE obj, ID sym, int argc, const VALUE *argv, + struct dump_arg *arg, const char *name) +{ + VALUE ret = rb_funcallv(obj, sym, argc, argv); + VALUE klass = CLASS_OF(obj); + if (CLASS_OF(ret) == klass) { + rb_raise(rb_eRuntimeError, "%"PRIsVALUE"#%s returned same class instance", + klass, name); + } + return check_dump_arg(ret, arg, name); +} + +#define dump_funcall(arg, obj, sym, argc, argv) \ + check_userdump_arg(obj, sym, argc, argv, arg, name_##sym) +#define dump_check_funcall(arg, obj, sym, argc, argv) \ + check_dump_arg(rb_check_funcall(obj, sym, argc, argv), arg, name_##sym) + +static void clear_dump_arg(struct dump_arg *arg); +static void +mark_dump_arg(void *ptr) +{ + struct dump_arg *p = ptr; + if (!p->symbols) + return; + rb_mark_set(p->symbols); + rb_mark_set(p->data); + rb_mark_hash(p->compat_tbl); + rb_mark_set(p->userdefs); + rb_gc_mark(p->str); +} + +static void +free_dump_arg(void *ptr) +{ + clear_dump_arg(ptr); +} + +static size_t +memsize_dump_arg(const void *ptr) +{ + const struct dump_arg *p = (struct dump_arg *)ptr; + size_t memsize = 0; + if (p->symbols) memsize += rb_st_memsize(p->symbols); + if (p->data) memsize += rb_st_memsize(p->data); + if (p->compat_tbl) memsize += rb_st_memsize(p->compat_tbl); + if (p->userdefs) memsize += rb_st_memsize(p->userdefs); + if (p->encodings) memsize += rb_st_memsize(p->encodings); + return memsize; +} + +static const rb_data_type_t dump_arg_data = { + "dump_arg", + {mark_dump_arg, free_dump_arg, memsize_dump_arg,}, + 0, 0, RUBY_TYPED_FREE_IMMEDIATELY | RUBY_TYPED_EMBEDDABLE +}; + +static VALUE +must_not_be_anonymous(const char *type, VALUE path) +{ + char *n = RSTRING_PTR(path); + + if (!rb_enc_asciicompat(rb_enc_get(path))) { + /* cannot occur? */ + rb_raise(rb_eTypeError, "can't dump non-ascii %s name % "PRIsVALUE, + type, path); + } if (n[0] == '#') { - rb_raise(rb_eTypeError, "can't dump anonymous %s %s", - (TYPE(klass) == T_CLASS ? "class" : "module"), - n); + rb_raise(rb_eTypeError, "can't dump anonymous %s % "PRIsVALUE, + type, path); } - if (rb_path2class(n) != rb_class_real(klass)) { - rb_raise(rb_eTypeError, "%s can't be referred", n); + return path; +} + +static VALUE +class2path(VALUE klass) +{ + VALUE path = rb_class_path(klass); + + must_not_be_anonymous((RB_TYPE_P(klass, T_CLASS) ? "class" : "module"), path); + if (rb_path_to_class(path) != rb_class_real(klass)) { + rb_raise(rb_eTypeError, "% "PRIsVALUE" can't be referred to", path); } return path; } -static void w_long _((long, struct dump_arg*)); +int ruby_marshal_write_long(long x, char *buf); +static void w_long(long, struct dump_arg*); +static int w_encoding(VALUE encname, struct dump_call_arg *arg); +static VALUE encoding_name(VALUE obj, struct dump_arg *arg); static void -w_nbyte(s, n, arg) - char *s; - int n; - struct dump_arg *arg; +w_nbyte(const char *s, long n, struct dump_arg *arg) { VALUE buf = arg->str; rb_str_buf_cat(buf, s, n); - if (arg->dest && RSTRING(buf)->len >= BUFSIZ) { - if (arg->taint) OBJ_TAINT(buf); - rb_io_write(arg->dest, buf); - rb_str_resize(buf, 0); + if (arg->dest && RSTRING_LEN(buf) >= BUFSIZ) { + rb_io_write(arg->dest, buf); + rb_str_resize(buf, 0); } } static void -w_byte(c, arg) - char c; - struct dump_arg *arg; +w_byte(char c, struct dump_arg *arg) { w_nbyte(&c, 1, arg); } static void -w_bytes(s, n, arg) - char *s; - int n; - struct dump_arg *arg; +w_bytes(const char *s, long n, struct dump_arg *arg) { w_long(n, arg); w_nbyte(s, n, arg); } +#define w_cstr(s, arg) w_bytes((s), strlen(s), (arg)) + static void -w_short(x, arg) - int x; - struct dump_arg *arg; +w_short(int x, struct dump_arg *arg) { - w_byte((x >> 0) & 0xff, arg); - w_byte((x >> 8) & 0xff, arg); + w_byte((char)((x >> 0) & 0xff), arg); + w_byte((char)((x >> 8) & 0xff), arg); } static void -w_long(x, arg) - long x; - struct dump_arg *arg; +w_long(long x, struct dump_arg *arg) { char buf[sizeof(long)+1]; - int i, len = 0; + int i = ruby_marshal_write_long(x, buf); + if (i < 0) { + rb_raise(rb_eTypeError, "long too big to dump"); + } + w_nbyte(buf, i, arg); +} + +int +ruby_marshal_write_long(long x, char *buf) +{ + int i; #if SIZEOF_LONG > 4 if (!(RSHIFT(x, 31) == 0 || RSHIFT(x, 31) == -1)) { - /* big long does not fit in 4 bytes */ - rb_raise(rb_eTypeError, "long too big to dump"); + /* big long does not fit in 4 bytes */ + return -1; } #endif if (x == 0) { - w_byte(0, arg); - return; + buf[0] = 0; + return 1; } if (0 < x && x < 123) { - w_byte(x + 5, arg); - return; + buf[0] = (char)(x + 5); + return 1; } if (-124 < x && x < 0) { - w_byte((x - 5)&0xff, arg); - return; - } - for (i=1;i<sizeof(long)+1;i++) { - buf[i] = x & 0xff; - x = RSHIFT(x,8); - if (x == 0) { - buf[0] = i; - break; - } - if (x == -1) { - buf[0] = -i; - break; - } + buf[0] = (char)((x - 5)&0xff); + return 1; } - len = i; - for (i=0;i<=len;i++) { - w_byte(buf[i], arg); + for (i=1;i<(int)sizeof(long)+1;i++) { + buf[i] = (char)(x & 0xff); + x = RSHIFT(x,8); + if (x == 0) { + buf[0] = i; + break; + } + if (x == -1) { + buf[0] = -i; + break; + } } + return i+1; } #ifdef DBL_MANT_DIG @@ -218,73 +370,39 @@ w_long(x, arg) #define MANT_BITS 8 #endif -static int -save_mantissa(d, buf) - double d; - char *buf; -{ - int e, i = 0; - unsigned long m; - double n; - - d = modf(ldexp(frexp(fabs(d), &e), DECIMAL_MANT), &d); - if (d > 0) { - buf[i++] = 0; - do { - d = modf(ldexp(d, MANT_BITS), &n); - m = (unsigned long)n; -#if MANT_BITS > 24 - buf[i++] = m >> 24; -#endif -#if MANT_BITS > 16 - buf[i++] = m >> 16; -#endif -#if MANT_BITS > 8 - buf[i++] = m >> 8; -#endif - buf[i++] = m; - } while (d > 0); - while (!buf[i - 1]) --i; - } - return i; -} - static double -load_mantissa(d, buf, len) - double d; - const char *buf; - int len; +load_mantissa(double d, const char *buf, long len) { + if (!len) return d; if (--len > 0 && !*buf++) { /* binary mantissa mark */ - int e, s = d < 0, dig = 0; - unsigned long m; - - modf(ldexp(frexp(fabs(d), &e), DECIMAL_MANT), &d); - do { - m = 0; - switch (len) { - default: m = *buf++ & 0xff; + int e, s = d < 0, dig = 0; + unsigned long m; + + modf(ldexp(frexp(fabs(d), &e), DECIMAL_MANT), &d); + do { + m = 0; + switch (len) { + default: m = *buf++ & 0xff; /* fall through */ #if MANT_BITS > 24 - case 3: m = (m << 8) | (*buf++ & 0xff); + case 3: m = (m << 8) | (*buf++ & 0xff); /* fall through */ #endif #if MANT_BITS > 16 - case 2: m = (m << 8) | (*buf++ & 0xff); + case 2: m = (m << 8) | (*buf++ & 0xff); /* fall through */ #endif #if MANT_BITS > 8 - case 1: m = (m << 8) | (*buf++ & 0xff); + case 1: m = (m << 8) | (*buf++ & 0xff); #endif - } - dig -= len < MANT_BITS / 8 ? 8 * (unsigned)len : MANT_BITS; - d += ldexp((double)m, dig); - } while ((len -= MANT_BITS / 8) > 0); - d = ldexp(d, e - DECIMAL_MANT); - if (s) d = -d; + } + dig -= len < MANT_BITS / 8 ? 8 * (unsigned)len : MANT_BITS; + d += ldexp((double)m, dig); + } while ((len -= MANT_BITS / 8) > 0); + d = ldexp(d, e - DECIMAL_MANT); + if (s) d = -d; } return d; } #else #define load_mantissa(d, buf, len) (d) -#define save_mantissa(d, buf) 0 #endif #ifdef DBL_DIG @@ -294,415 +412,743 @@ load_mantissa(d, buf, len) #endif static void -w_float(d, arg) - double d; - struct dump_arg *arg; +w_float(double d, struct dump_arg *arg) { - char buf[100]; + char buf[FLOAT_DIG + (DECIMAL_MANT + 7) / 8 + 10]; if (isinf(d)) { - if (d < 0) strcpy(buf, "-inf"); - else strcpy(buf, "inf"); + if (d < 0) w_cstr("-inf", arg); + else w_cstr("inf", arg); } else if (isnan(d)) { - strcpy(buf, "nan"); + w_cstr("nan", arg); } else if (d == 0.0) { - if (1.0/d < 0) strcpy(buf, "-0"); - else strcpy(buf, "0"); + if (signbit(d)) w_cstr("-0", arg); + else w_cstr("0", arg); } else { - int len; + int decpt, sign, digs, len = 0; + char *e, *p = ruby_dtoa(d, 0, 0, &decpt, &sign, &e); + if (sign) buf[len++] = '-'; + digs = (int)(e - p); + if (decpt < -3 || decpt > digs) { + buf[len++] = p[0]; + if (--digs > 0) buf[len++] = '.'; + memcpy(buf + len, p + 1, digs); + len += digs; + len += snprintf(buf + len, sizeof(buf) - len, "e%d", decpt - 1); + } + else if (decpt > 0) { + memcpy(buf + len, p, decpt); + len += decpt; + if ((digs -= decpt) > 0) { + buf[len++] = '.'; + memcpy(buf + len, p + decpt, digs); + len += digs; + } + } + else { + buf[len++] = '0'; + buf[len++] = '.'; + if (decpt) { + memset(buf + len, '0', -decpt); + len -= decpt; + } + memcpy(buf + len, p, digs); + len += digs; + } + free(p); + w_bytes(buf, len, arg); + } +} + - /* xxx: should not use system's sprintf(3) */ - sprintf(buf, "%.*g", FLOAT_DIG, d); - len = strlen(buf); - w_bytes(buf, len + save_mantissa(d, buf + len), arg); - return; +static VALUE +w_encivar(VALUE str, struct dump_arg *arg) +{ + VALUE encname = encoding_name(str, arg); + if (NIL_P(encname) || + is_ascii_string(str)) { + return Qnil; } - w_bytes(buf, strlen(buf), arg); + w_byte(TYPE_IVAR, arg); + return encname; } static void -w_symbol(id, arg) - ID id; - struct dump_arg *arg; +w_encname(VALUE encname, struct dump_arg *arg) +{ + if (!NIL_P(encname)) { + struct dump_call_arg c_arg; + c_arg.limit = 1; + c_arg.arg = arg; + w_long(1L, arg); + w_encoding(encname, &c_arg); + } +} + +static void +w_symbol(VALUE sym, struct dump_arg *arg) { - char *sym = rb_id2name(id); st_data_t num; + VALUE encname; - if (st_lookup(arg->symbols, id, &num)) { - w_byte(TYPE_SYMLINK, arg); - w_long((long)num, arg); + if (st_lookup(arg->symbols, sym, &num)) { + w_byte(TYPE_SYMLINK, arg); + w_long((long)num, arg); } else { - w_byte(TYPE_SYMBOL, arg); - w_bytes(sym, strlen(sym), arg); - st_add_direct(arg->symbols, id, arg->symbols->num_entries); + const VALUE orig_sym = sym; + sym = rb_sym2str(sym); + if (!sym) { + rb_raise(rb_eTypeError, "can't dump anonymous ID %"PRIdVALUE, sym); + } + encname = w_encivar(sym, arg); + w_byte(TYPE_SYMBOL, arg); + w_bytes(RSTRING_PTR(sym), RSTRING_LEN(sym), arg); + st_add_direct(arg->symbols, orig_sym, arg->symbols->num_entries); + w_encname(encname, arg); } } static void -w_unique(s, arg) - char *s; - struct dump_arg *arg; +w_unique(VALUE s, struct dump_arg *arg) { - if (s[0] == '#') { - rb_raise(rb_eTypeError, "can't dump anonymous class %s", s); - } - w_symbol(rb_intern(s), arg); + must_not_be_anonymous("class", s); + w_symbol(rb_str_intern(s), arg); } -static void w_object _((VALUE,struct dump_arg*,int)); +static void w_object(VALUE,struct dump_arg*,int); static int -hash_each(key, value, arg) - VALUE key, value; - struct dump_call_arg *arg; +hash_each(VALUE key, VALUE value, VALUE v) { + struct dump_call_arg *arg = (void *)v; w_object(key, arg->arg, arg->limit); w_object(value, arg->arg, arg->limit); return ST_CONTINUE; } +#define SINGLETON_DUMP_UNABLE_P(klass) \ + (rb_id_table_size(RCLASS_M_TBL(klass)) > 0 || \ + rb_ivar_count(klass) > 0) + static void -w_extended(klass, arg, check) - VALUE klass; - struct dump_arg *arg; - int check; +w_extended(VALUE klass, struct dump_arg *arg, int check) { - char *path; - - if (check && FL_TEST(klass, FL_SINGLETON)) { - if (RCLASS(klass)->m_tbl->num_entries || - (RCLASS(klass)->iv_tbl && RCLASS(klass)->iv_tbl->num_entries > 1)) { - rb_raise(rb_eTypeError, "singleton can't be dumped"); - } - klass = RCLASS(klass)->super; + if (check && RCLASS_SINGLETON_P(klass)) { + VALUE origin = RCLASS_ORIGIN(klass); + if (SINGLETON_DUMP_UNABLE_P(klass) || + (origin != klass && SINGLETON_DUMP_UNABLE_P(origin))) { + rb_raise(rb_eTypeError, "singleton can't be dumped"); + } + klass = RCLASS_SUPER(klass); } while (BUILTIN_TYPE(klass) == T_ICLASS) { - path = rb_class2name(RBASIC(klass)->klass); - w_byte(TYPE_EXTENDED, arg); - w_unique(path, arg); - klass = RCLASS(klass)->super; + if (!RICLASS_IS_ORIGIN_P(klass) || + BUILTIN_TYPE(RBASIC(klass)->klass) != T_MODULE) { + VALUE path = rb_class_name(RBASIC(klass)->klass); + w_byte(TYPE_EXTENDED, arg); + w_unique(path, arg); + } + klass = RCLASS_SUPER(klass); } } static void -w_class(type, obj, arg, check) - int type; - VALUE obj; - struct dump_arg *arg; - int check; +w_class(char type, VALUE obj, struct dump_arg *arg, int check) { - char *path; + VALUE path; + st_data_t real_obj; + VALUE klass; - VALUE klass = CLASS_OF(obj); + if (arg->compat_tbl && + st_lookup(arg->compat_tbl, (st_data_t)obj, &real_obj)) { + obj = (VALUE)real_obj; + } + klass = CLASS_OF(obj); w_extended(klass, arg, check); w_byte(type, arg); - path = RSTRING(class2path(rb_class_real(klass)))->ptr; + path = class2path(rb_class_real(klass)); w_unique(path, arg); } static void -w_uclass(obj, base_klass, arg) - VALUE obj, base_klass; - struct dump_arg *arg; +w_uclass(VALUE obj, VALUE super, struct dump_arg *arg) { VALUE klass = CLASS_OF(obj); - w_extended(klass, arg, Qtrue); + w_extended(klass, arg, TRUE); klass = rb_class_real(klass); - if (klass != base_klass) { - w_byte(TYPE_UCLASS, arg); - w_unique(RSTRING(class2path(klass))->ptr, arg); + if (klass != super) { + w_byte(TYPE_UCLASS, arg); + w_unique(class2path(klass), arg); } } +static bool +rb_hash_ruby2_keywords_p(VALUE obj) +{ + return (RHASH(obj)->basic.flags & RHASH_PASS_AS_KEYWORDS) != 0; +} + +static void +rb_hash_ruby2_keywords(VALUE obj) +{ + RHASH(obj)->basic.flags |= RHASH_PASS_AS_KEYWORDS; +} + +/* + * if instance variable name `id` is a special name to be skipped, + * returns the name of it. otherwise it cannot be dumped (unnamed), + * returns `name` as-is. returns NULL for ID that can be dumped. + */ +static inline const char * +skipping_ivar_name(const ID id, const char *name) +{ +#define IS_SKIPPED_IVAR(idname) \ + ((id == idname) && (name = name_##idname, true)) + if (IS_SKIPPED_IVAR(s_encoding_short)) return name; + if (IS_SKIPPED_IVAR(s_ruby2_keywords_flag)) return name; + if (IS_SKIPPED_IVAR(s_encoding_long)) return name; + if (!rb_id2str(id)) return name; + return NULL; +} + +struct w_ivar_arg { + struct dump_call_arg *dump; + st_data_t num_ivar; +}; + static int -w_obj_each(id, value, arg) - ID id; - VALUE value; - struct dump_call_arg *arg; +w_obj_each(ID id, VALUE value, st_data_t a) { - w_symbol(id, arg->arg); + struct w_ivar_arg *ivarg = (struct w_ivar_arg *)a; + struct dump_call_arg *arg = ivarg->dump; + const char unnamed[] = "", *ivname = skipping_ivar_name(id, unnamed); + + if (ivname) { + if (ivname != unnamed) { + rb_warn("instance variable '%s' on class %"PRIsVALUE" is not dumped", + ivname, CLASS_OF(arg->obj)); + } + return ST_CONTINUE; + } + --ivarg->num_ivar; + w_symbol(ID2SYM(id), arg->arg); w_object(value, arg->arg, arg->limit); return ST_CONTINUE; } -static void -w_ivar(tbl, arg) - st_table *tbl; - struct dump_call_arg *arg; +static int +obj_count_ivars(ID id, VALUE val, st_data_t a) { - if (tbl) { - w_long(tbl->num_entries, arg->arg); - st_foreach_safe(tbl, w_obj_each, (st_data_t)arg); + if (!skipping_ivar_name(id, "") && UNLIKELY(!++*(st_index_t *)a)) { + rb_raise(rb_eRuntimeError, "too many instance variables"); + } + return ST_CONTINUE; +} + +static VALUE +encoding_name(VALUE obj, struct dump_arg *arg) +{ + if (rb_enc_capable(obj)) { + int encidx = rb_enc_get_index(obj); + rb_encoding *enc = 0; + st_data_t name; + + if (encidx <= 0 || !(enc = rb_enc_from_index(encidx))) { + return Qnil; + } + + /* special treatment for US-ASCII and UTF-8 */ + if (encidx == rb_usascii_encindex()) { + return Qfalse; + } + else if (encidx == rb_utf8_encindex()) { + return Qtrue; + } + + if (arg->encodings ? + !st_lookup(arg->encodings, (st_data_t)rb_enc_name(enc), &name) : + (arg->encodings = st_init_strcasetable(), 1)) { + name = (st_data_t)rb_str_new_cstr(rb_enc_name(enc)); + st_insert(arg->encodings, (st_data_t)rb_enc_name(enc), name); + } + return (VALUE)name; } else { - w_long(0, arg->arg); + return Qnil; } } +static int +w_encoding(VALUE encname, struct dump_call_arg *arg) +{ + int limit = arg->limit; + if (limit >= 0) ++limit; + switch (encname) { + case Qfalse: + case Qtrue: + w_symbol(ID2SYM(s_encoding_short), arg->arg); + w_object(encname, arg->arg, limit); + return 1; + case Qnil: + return 0; + } + w_symbol(ID2SYM(rb_id_encoding()), arg->arg); + w_object(encname, arg->arg, limit); + return 1; +} + +static st_index_t +has_ivars(VALUE obj, VALUE encname, VALUE *ivobj) +{ + st_index_t num = !NIL_P(encname); + + if (SPECIAL_CONST_P(obj)) goto generic; + switch (BUILTIN_TYPE(obj)) { + case T_OBJECT: + case T_CLASS: + case T_MODULE: + break; /* counted elsewhere */ + case T_HASH: + if (rb_hash_ruby2_keywords_p(obj)) ++num; + /* fall through */ + default: + generic: + rb_ivar_foreach(obj, obj_count_ivars, (st_data_t)&num); + if (num) *ivobj = obj; + } + + return num; +} + static void -w_object(obj, arg, limit) - VALUE obj; - struct dump_arg *arg; - int limit; +w_ivar_each(VALUE obj, st_index_t num, struct dump_call_arg *arg) { - struct dump_call_arg c_arg; - st_table *ivtbl = 0; - st_data_t num; + struct w_ivar_arg ivarg = {arg, num}; + if (!num) return; + rb_ivar_foreach_buffered(obj, w_obj_each, (st_data_t)&ivarg); +} - if (limit == 0) { - rb_raise(rb_eArgError, "exceed depth limit"); +static void +w_ivar(st_index_t num, VALUE ivobj, VALUE encname, struct dump_call_arg *arg) +{ + w_long(num, arg->arg); + num -= w_encoding(encname, arg); + if (RB_TYPE_P(ivobj, T_HASH) && rb_hash_ruby2_keywords_p(ivobj)) { + int limit = arg->limit; + if (limit >= 0) ++limit; + w_symbol(ID2SYM(s_ruby2_keywords_flag), arg->arg); + w_object(Qtrue, arg->arg, limit); + num--; + } + if (!UNDEF_P(ivobj) && num) { + w_ivar_each(ivobj, num, arg); + } +} + +static void +w_objivar(VALUE obj, struct dump_call_arg *arg) +{ + st_data_t num = 0; + + rb_ivar_foreach(obj, obj_count_ivars, (st_data_t)&num); + w_long(num, arg->arg); + w_ivar_each(obj, num, arg); +} + +#if SIZEOF_LONG > 4 +// Optimized dump for fixnum larger than 31-bits +static void +w_bigfixnum(VALUE obj, struct dump_arg *arg) +{ + RUBY_ASSERT(FIXNUM_P(obj)); + + w_byte(TYPE_BIGNUM, arg); + +#if SIZEOF_LONG == SIZEOF_VALUE + long num, slen_num; + num = FIX2LONG(obj); +#else + long long num, slen_num; + num = NUM2LL(obj); +#endif + + char sign = num < 0 ? '-' : '+'; + w_byte(sign, arg); + + // Guaranteed not to overflow, as FIXNUM is 1-bit less than long + if (num < 0) num = -num; + + // calculate the size in shorts + int slen = 0; + { + slen_num = num; + while (slen_num) { + slen++; + slen_num = SHORTDN(slen_num); + } } - limit--; - c_arg.limit = limit; - c_arg.arg = arg; + RUBY_ASSERT(slen > 0 && slen <= SIZEOF_LONG / 2); - if (st_lookup(arg->data, obj, &num)) { - w_byte(TYPE_LINK, arg); - w_long((long)num, arg); - return; + w_long((long)slen, arg); + + for (int i = 0; i < slen; i++) { + w_short(num & SHORTMASK, arg); + num = SHORTDN(num); } - if (ivtbl = rb_generic_ivar_table(obj)) { - w_byte(TYPE_IVAR, arg); + // We aren't adding this object to the link table, but we need to increment + // the index. + arg->num_entries++; + + RUBY_ASSERT(num == 0); +} +#endif + +static void +w_remember(VALUE obj, struct dump_arg *arg) +{ + st_add_direct(arg->data, obj, arg->num_entries++); +} + +static void +w_object(VALUE obj, struct dump_arg *arg, int limit) +{ + struct dump_call_arg c_arg; + VALUE ivobj = Qundef; + st_data_t num; + st_index_t hasiv = 0; + VALUE encname = Qnil; + + if (limit == 0) { + rb_raise(rb_eArgError, "exceed depth limit"); } - if (obj == Qnil) { - w_byte(TYPE_NIL, arg); + + if (NIL_P(obj)) { + w_byte(TYPE_NIL, arg); } else if (obj == Qtrue) { - w_byte(TYPE_TRUE, arg); + w_byte(TYPE_TRUE, arg); } else if (obj == Qfalse) { - w_byte(TYPE_FALSE, arg); + w_byte(TYPE_FALSE, arg); } else if (FIXNUM_P(obj)) { #if SIZEOF_LONG <= 4 - w_byte(TYPE_FIXNUM, arg); - w_long(FIX2INT(obj), arg); + w_byte(TYPE_FIXNUM, arg); + w_long(FIX2INT(obj), arg); #else - if (RSHIFT((long)obj, 31) == 0 || RSHIFT((long)obj, 31) == -1) { - w_byte(TYPE_FIXNUM, arg); - w_long(FIX2LONG(obj), arg); - } - else { - w_object(rb_int2big(FIX2LONG(obj)), arg, limit); - } + if (RSHIFT((long)obj, 31) == 0 || RSHIFT((long)obj, 31) == -1) { + w_byte(TYPE_FIXNUM, arg); + w_long(FIX2LONG(obj), arg); + } + else { + w_bigfixnum(obj, arg); + } #endif } else if (SYMBOL_P(obj)) { - w_symbol(SYM2ID(obj), arg); + w_symbol(obj, arg); } else { - if (OBJ_TAINTED(obj)) arg->taint = Qtrue; - - st_add_direct(arg->data, obj, arg->data->num_entries); - if (rb_respond_to(obj, s_mdump)) { - VALUE v; - - v = rb_funcall(obj, s_mdump, 0, 0); - w_class(TYPE_USRMARSHAL, obj, arg, Qfalse); - w_object(v, arg, limit); - if (ivtbl) w_ivar(0, &c_arg); - return; - } - if (rb_respond_to(obj, s_dump)) { - VALUE v; - - v = rb_funcall(obj, s_dump, 1, INT2NUM(limit)); - if (TYPE(v) != T_STRING) { - rb_raise(rb_eTypeError, "_dump() must return string"); - } - if (!ivtbl && (ivtbl = rb_generic_ivar_table(v))) { - w_byte(TYPE_IVAR, arg); - } - w_class(TYPE_USERDEF, obj, arg, Qfalse); - w_bytes(RSTRING(v)->ptr, RSTRING(v)->len, arg); - if (ivtbl) { - w_ivar(ivtbl, &c_arg); - } - return; - } - - switch (BUILTIN_TYPE(obj)) { - case T_CLASS: - if (FL_TEST(obj, FL_SINGLETON)) { - rb_raise(rb_eTypeError, "singleton class can't be dumped"); - } - w_byte(TYPE_CLASS, arg); - { - VALUE path = class2path(obj); - w_bytes(RSTRING(path)->ptr, RSTRING(path)->len, arg); - } - break; - - case T_MODULE: - w_byte(TYPE_MODULE, arg); - { - VALUE path = class2path(obj); - w_bytes(RSTRING(path)->ptr, RSTRING(path)->len, arg); - } - break; - - case T_FLOAT: - w_byte(TYPE_FLOAT, arg); - w_float(RFLOAT(obj)->value, arg); - break; - - case T_BIGNUM: - w_byte(TYPE_BIGNUM, arg); - { - char sign = RBIGNUM(obj)->sign ? '+' : '-'; - long len = RBIGNUM(obj)->len; - BDIGIT *d = RBIGNUM(obj)->digits; - - w_byte(sign, arg); - w_long(SHORTLEN(len), arg); /* w_short? */ - while (len--) { -#if SIZEOF_BDIGITS > SIZEOF_SHORT - BDIGIT num = *d; - int i; - - for (i=0; i<SIZEOF_BDIGITS; i+=SIZEOF_SHORT) { - w_short(num & SHORTMASK, arg); - num = SHORTDN(num); - if (len == 0 && num == 0) break; - } + if (st_lookup(arg->data, obj, &num)) { + w_byte(TYPE_LINK, arg); + w_long((long)num, arg); + return; + } + + if (limit > 0) limit--; + c_arg.limit = limit; + c_arg.arg = arg; + c_arg.obj = obj; + + if (FLONUM_P(obj)) { + w_remember(obj, arg); + w_byte(TYPE_FLOAT, arg); + w_float(RFLOAT_VALUE(obj), arg); + return; + } + + VALUE v; + + if (!RBASIC_CLASS(obj)) { + rb_raise(rb_eTypeError, "can't dump internal %s", + rb_builtin_type_name(BUILTIN_TYPE(obj))); + } + + if (rb_obj_respond_to(obj, s_mdump, TRUE)) { + w_remember(obj, arg); + + v = dump_funcall(arg, obj, s_mdump, 0, 0); + w_class(TYPE_USRMARSHAL, obj, arg, FALSE); + w_object(v, arg, limit); + return; + } + if (rb_obj_respond_to(obj, s_dump, TRUE)) { + VALUE ivobj2 = Qundef; + st_index_t hasiv2; + VALUE encname2; + + if (arg->userdefs && st_is_member(arg->userdefs, (st_data_t)obj)) { + rb_raise(rb_eRuntimeError, "can't dump recursive object using _dump()"); + } + v = INT2NUM(limit); + v = dump_funcall(arg, obj, s_dump, 1, &v); + if (!RB_TYPE_P(v, T_STRING)) { + rb_raise(rb_eTypeError, "_dump() must return string"); + } + hasiv = has_ivars(obj, (encname = encoding_name(obj, arg)), &ivobj); + hasiv2 = has_ivars(v, (encname2 = encoding_name(v, arg)), &ivobj2); + if (hasiv2) { + hasiv = hasiv2; + ivobj = ivobj2; + encname = encname2; + } + if (hasiv) w_byte(TYPE_IVAR, arg); + w_class(TYPE_USERDEF, obj, arg, FALSE); + w_bytes(RSTRING_PTR(v), RSTRING_LEN(v), arg); + if (hasiv) { + st_data_t userdefs = (st_data_t)obj; + if (!arg->userdefs) { + arg->userdefs = rb_init_identtable(); + } + st_add_direct(arg->userdefs, userdefs, 0); + w_ivar(hasiv, ivobj, encname, &c_arg); + st_delete(arg->userdefs, &userdefs, NULL); + } + w_remember(obj, arg); + return; + } + + w_remember(obj, arg); + + hasiv = has_ivars(obj, (encname = encoding_name(obj, arg)), &ivobj); + { + st_data_t compat_data; + VALUE klass = CLASS_OF(obj); + rb_alloc_func_t allocator = RCLASS_SINGLETON_P(klass) ? 0 : rb_get_alloc_func(klass); + if (allocator && st_lookup(compat_allocator_tbl, + (st_data_t)allocator, + &compat_data)) { + marshal_compat_t *compat = (marshal_compat_t*)compat_data; + VALUE real_obj = obj; + obj = compat->dumper(real_obj); + if (!arg->compat_tbl) { + arg->compat_tbl = rb_init_identtable(); + } + st_insert(arg->compat_tbl, (st_data_t)obj, (st_data_t)real_obj); + if (obj != real_obj && UNDEF_P(ivobj)) hasiv = 0; + } + } + if (hasiv) w_byte(TYPE_IVAR, arg); + + switch (BUILTIN_TYPE(obj)) { + case T_CLASS: + if (FL_TEST(obj, FL_SINGLETON)) { + rb_raise(rb_eTypeError, "singleton class can't be dumped"); + } + { + VALUE path = class2path(obj); + VALUE encname = w_encivar(path, arg); + w_byte(TYPE_CLASS, arg); + w_bytes(RSTRING_PTR(path), RSTRING_LEN(path), arg); + w_encname(encname, arg); + RB_GC_GUARD(path); + } + break; + + case T_MODULE: + { + VALUE path = class2path(obj); + VALUE encname = w_encivar(path, arg); + w_byte(TYPE_MODULE, arg); + w_bytes(RSTRING_PTR(path), RSTRING_LEN(path), arg); + w_encname(encname, arg); + RB_GC_GUARD(path); + } + break; + + case T_FLOAT: + w_byte(TYPE_FLOAT, arg); + w_float(RFLOAT_VALUE(obj), arg); + break; + + case T_BIGNUM: + w_byte(TYPE_BIGNUM, arg); + { + char sign = BIGNUM_SIGN(obj) ? '+' : '-'; + size_t len = BIGNUM_LEN(obj); + size_t slen; + size_t j; + BDIGIT *d = BIGNUM_DIGITS(obj); + + slen = SHORTLEN(len); + if (LONG_MAX < slen) { + rb_raise(rb_eTypeError, "too big Bignum can't be dumped"); + } + + w_byte(sign, arg); + w_long((long)slen, arg); + for (j = 0; j < len; j++) { +#if SIZEOF_BDIGIT > SIZEOF_SHORT + BDIGIT num = *d; + int i; + + for (i=0; i<SIZEOF_BDIGIT; i+=SIZEOF_SHORT) { + w_short(num & SHORTMASK, arg); + num = SHORTDN(num); + if (j == len - 1 && num == 0) break; + } #else - w_short(*d, arg); + w_short(*d, arg); #endif - d++; - } - } - break; - - case T_STRING: - w_uclass(obj, rb_cString, arg); - w_byte(TYPE_STRING, arg); - w_bytes(RSTRING(obj)->ptr, RSTRING(obj)->len, arg); - break; - - case T_REGEXP: - w_uclass(obj, rb_cRegexp, arg); - w_byte(TYPE_REGEXP, arg); - w_bytes(RREGEXP(obj)->str, RREGEXP(obj)->len, arg); - w_byte(rb_reg_options(obj), arg); - break; - - case T_ARRAY: - w_uclass(obj, rb_cArray, arg); - w_byte(TYPE_ARRAY, arg); - { - long len = RARRAY(obj)->len; - VALUE *ptr = RARRAY(obj)->ptr; - - w_long(len, arg); - while (len--) { - w_object(*ptr, arg, limit); - ptr++; - } - } - break; - - case T_HASH: - w_uclass(obj, rb_cHash, arg); - if (NIL_P(RHASH(obj)->ifnone)) { - w_byte(TYPE_HASH, arg); - } - else if (FL_TEST(obj, FL_USER2)) { - /* FL_USER2 means HASH_PROC_DEFAULT (see hash.c) */ - rb_raise(rb_eTypeError, "can't dump hash with default proc"); - } - else { - w_byte(TYPE_HASH_DEF, arg); - } - w_long(RHASH(obj)->tbl->num_entries, arg); - rb_hash_foreach(obj, hash_each, (st_data_t)&c_arg); - if (!NIL_P(RHASH(obj)->ifnone)) { - w_object(RHASH(obj)->ifnone, arg, limit); - } - break; - - case T_STRUCT: - w_class(TYPE_STRUCT, obj, arg, Qtrue); - { - long len = RSTRUCT(obj)->len; - VALUE mem; - long i; - - w_long(len, arg); - mem = rb_struct_members(obj); - for (i=0; i<len; i++) { - w_symbol(SYM2ID(RARRAY(mem)->ptr[i]), arg); - w_object(RSTRUCT(obj)->ptr[i], arg, limit); - } - } - break; - - case T_OBJECT: - w_class(TYPE_OBJECT, obj, arg, Qtrue); - w_ivar(ROBJECT(obj)->iv_tbl, &c_arg); - break; - - case T_DATA: - { - VALUE v; - - if (!rb_respond_to(obj, s_dump_data)) { - rb_raise(rb_eTypeError, - "no marshal_dump is defined for class %s", - rb_obj_classname(obj)); - } - v = rb_funcall(obj, s_dump_data, 0); - w_class(TYPE_DATA, obj, arg, Qtrue); - w_object(v, arg, limit); - } - break; - - default: - rb_raise(rb_eTypeError, "can't dump %s", - rb_obj_classname(obj)); - break; - } - } - if (ivtbl) { - w_ivar(ivtbl, &c_arg); - } -} + d++; + } + } + break; + + case T_STRING: + w_uclass(obj, rb_cString, arg); + w_byte(TYPE_STRING, arg); + w_bytes(RSTRING_PTR(obj), RSTRING_LEN(obj), arg); + break; + + case T_REGEXP: + w_uclass(obj, rb_cRegexp, arg); + w_byte(TYPE_REGEXP, arg); + { + int opts = rb_reg_options(obj); + w_bytes(RREGEXP_SRC_PTR(obj), RREGEXP_SRC_LEN(obj), arg); + w_byte((char)opts, arg); + } + break; + + case T_ARRAY: + w_uclass(obj, rb_cArray, arg); + w_byte(TYPE_ARRAY, arg); + { + long i, len = RARRAY_LEN(obj); + + w_long(len, arg); + for (i=0; i<RARRAY_LEN(obj); i++) { + w_object(RARRAY_AREF(obj, i), arg, limit); + if (len != RARRAY_LEN(obj)) { + rb_raise(rb_eRuntimeError, "array modified during dump"); + } + } + } + break; -static VALUE -dump(arg) - struct dump_call_arg *arg; -{ - w_object(arg->obj, arg->arg, arg->limit); - if (arg->arg->dest) { - rb_io_write(arg->arg->dest, arg->arg->str); - rb_str_resize(arg->arg->str, 0); + case T_HASH: + w_uclass(obj, rb_cHash, arg); + if (rb_hash_compare_by_id_p(obj)) { + w_byte(TYPE_UCLASS, arg); + w_symbol(rb_sym_intern_ascii_cstr("Hash"), arg); + } + if (NIL_P(RHASH_IFNONE(obj))) { + w_byte(TYPE_HASH, arg); + } + else if (FL_TEST(obj, RHASH_PROC_DEFAULT)) { + rb_raise(rb_eTypeError, "can't dump hash with default proc"); + } + else { + w_byte(TYPE_HASH_DEF, arg); + } + w_long(rb_hash_size_num(obj), arg); + rb_hash_foreach(obj, hash_each, (st_data_t)&c_arg); + if (!NIL_P(RHASH_IFNONE(obj))) { + w_object(RHASH_IFNONE(obj), arg, limit); + } + break; + + case T_STRUCT: + w_class(TYPE_STRUCT, obj, arg, TRUE); + { + long len = RSTRUCT_LEN_RAW(obj); + VALUE mem; + long i; + + w_long(len, arg); + mem = rb_struct_members(obj); + for (i=0; i<len; i++) { + w_symbol(RARRAY_AREF(mem, i), arg); + w_object(RSTRUCT_GET_RAW(obj, i), arg, limit); + } + } + break; + + case T_OBJECT: + w_class(TYPE_OBJECT, obj, arg, TRUE); + w_objivar(obj, &c_arg); + break; + + case T_DATA: + { + VALUE v; + + if (!rb_obj_respond_to(obj, s_dump_data, TRUE)) { + rb_raise(rb_eTypeError, + "no _dump_data is defined for class %"PRIsVALUE, + rb_obj_class(obj)); + } + v = dump_funcall(arg, obj, s_dump_data, 0, 0); + w_class(TYPE_DATA, obj, arg, TRUE); + w_object(v, arg, limit); + } + break; + + default: + rb_raise(rb_eTypeError, "can't dump %"PRIsVALUE, + rb_obj_class(obj)); + break; + } + RB_GC_GUARD(obj); + } + if (hasiv) { + w_ivar(hasiv, ivobj, encname, &c_arg); } - return 0; } -static VALUE -dump_ensure(arg) - struct dump_arg *arg; +static void +clear_dump_arg(struct dump_arg *arg) { + if (!arg->symbols) return; st_free_table(arg->symbols); + arg->symbols = 0; st_free_table(arg->data); - if (arg->taint) { - OBJ_TAINT(arg->str); + arg->data = 0; + arg->num_entries = 0; + if (arg->compat_tbl) { + st_free_table(arg->compat_tbl); + arg->compat_tbl = 0; + } + if (arg->encodings) { + st_free_table(arg->encodings); + arg->encodings = 0; } - return 0; + if (arg->userdefs) { + st_free_table(arg->userdefs); + arg->userdefs = 0; + } +} + +NORETURN(static inline void io_needed(void)); +static inline void +io_needed(void) +{ + rb_raise(rb_eTypeError, "instance of IO needed"); } /* * call-seq: - * dump( obj [, anIO] , limit=--1 ) => anIO + * dump( obj [, anIO] , limit=-1 ) -> anIO * - * Serializes obj and all descendent objects. If anIO is + * Serializes obj and all descendant objects. If anIO is * specified, the serialized data will be written to it, otherwise the * data will be returned as a String. If limit is specified, the * traversal of subobjects will be limited to that depth. If limit is @@ -712,7 +1158,7 @@ dump_ensure(arg) * def initialize(str) * @str = str * end - * def sayHello + * def say_hello * @str * end * end @@ -722,720 +1168,1299 @@ dump_ensure(arg) * o = Klass.new("hello\n") * data = Marshal.dump(o) * obj = Marshal.load(data) - * obj.sayHello #=> "hello\n" + * obj.say_hello #=> "hello\n" + * + * Marshal can't dump following objects: + * * anonymous Class/Module. + * * objects which are related to system (ex: Dir, File::Stat, IO, File, Socket + * and so on) + * * an instance of MatchData, Method, UnboundMethod, Proc, Thread, + * ThreadGroup, Continuation + * * objects which define singleton methods */ static VALUE -marshal_dump(argc, argv) - int argc; - VALUE* argv; +marshal_dump(int argc, VALUE *argv, VALUE _) { VALUE obj, port, a1, a2; int limit = -1; - struct dump_arg arg; - struct dump_call_arg c_arg; port = Qnil; rb_scan_args(argc, argv, "12", &obj, &a1, &a2); if (argc == 3) { - if (!NIL_P(a2)) limit = NUM2INT(a2); - if (NIL_P(a1)) goto type_error; - port = a1; + if (!NIL_P(a2)) limit = NUM2INT(a2); + if (NIL_P(a1)) io_needed(); + port = a1; } else if (argc == 2) { - if (FIXNUM_P(a1)) limit = FIX2INT(a1); - else if (NIL_P(a1)) goto type_error; - else port = a1; + if (FIXNUM_P(a1)) limit = FIX2INT(a1); + else if (NIL_P(a1)) io_needed(); + else port = a1; } - arg.dest = 0; + return rb_marshal_dump_limited(obj, port, limit); +} + +VALUE +rb_marshal_dump_limited(VALUE obj, VALUE port, int limit) +{ + struct dump_arg *arg; + VALUE wrapper; /* used to avoid memory leak in case of exception */ + + wrapper = TypedData_Make_Struct(0, struct dump_arg, &dump_arg_data, arg); + arg->dest = 0; + arg->symbols = st_init_numtable(); + arg->data = rb_init_identtable(); + arg->num_entries = 0; + arg->compat_tbl = 0; + arg->encodings = 0; + arg->userdefs = 0; + arg->str = rb_str_buf_new(0); if (!NIL_P(port)) { - if (!rb_respond_to(port, s_write)) { - type_error: - rb_raise(rb_eTypeError, "instance of IO needed"); - } - arg.str = rb_str_buf_new(0); - arg.dest = port; - if (rb_respond_to(port, s_binmode)) { - rb_funcall2(port, s_binmode, 0, 0); - } + if (!rb_respond_to(port, s_write)) { + io_needed(); + } + arg->dest = port; + dump_check_funcall(arg, port, s_binmode, 0, 0); } else { - port = rb_str_buf_new(0); - arg.str = port; + port = arg->str; } - arg.symbols = st_init_numtable(); - arg.data = st_init_numtable(); - arg.taint = Qfalse; - c_arg.obj = obj; - c_arg.arg = &arg; - c_arg.limit = limit; - - w_byte(MARSHAL_MAJOR, &arg); - w_byte(MARSHAL_MINOR, &arg); + w_byte(MARSHAL_MAJOR, arg); + w_byte(MARSHAL_MINOR, arg); - rb_ensure(dump, (VALUE)&c_arg, dump_ensure, (VALUE)&arg); + w_object(obj, arg, limit); + if (arg->dest) { + rb_io_write(arg->dest, arg->str); + rb_str_resize(arg->str, 0); + } + clear_dump_arg(arg); + RB_GC_GUARD(wrapper); return port; } struct load_arg { VALUE src; + char *buf; + long bufsize; + long buflen; + long readable; long offset; st_table *symbols; - VALUE data; + st_table *data; + st_table *partial_objects; VALUE proc; - int taint; + st_table *compat_tbl; + bool freeze; }; -static VALUE r_object _((struct load_arg *arg)); +static VALUE +check_load_arg(VALUE ret, struct load_arg *arg, const char *name) +{ + if (!arg->symbols) { + rb_raise(rb_eRuntimeError, "Marshal.load reentered at %s", + name); + } + return ret; +} +#define load_funcall(arg, obj, sym, argc, argv) \ + check_load_arg(rb_funcallv(obj, sym, argc, argv), arg, name_##sym) + +static void clear_load_arg(struct load_arg *arg); + +static void +mark_load_arg(void *ptr) +{ + struct load_arg *p = ptr; + if (!p->symbols) + return; + rb_mark_tbl(p->symbols); + rb_mark_tbl(p->data); + rb_mark_tbl(p->partial_objects); + rb_mark_hash(p->compat_tbl); +} + +static void +free_load_arg(void *ptr) +{ + clear_load_arg(ptr); +} + +static size_t +memsize_load_arg(const void *ptr) +{ + const struct load_arg *p = (struct load_arg *)ptr; + size_t memsize = 0; + if (p->symbols) memsize += rb_st_memsize(p->symbols); + if (p->data) memsize += rb_st_memsize(p->data); + if (p->partial_objects) memsize += rb_st_memsize(p->partial_objects); + if (p->compat_tbl) memsize += rb_st_memsize(p->compat_tbl); + return memsize; +} + +static const rb_data_type_t load_arg_data = { + "load_arg", + {mark_load_arg, free_load_arg, memsize_load_arg,}, + 0, 0, RUBY_TYPED_FREE_IMMEDIATELY | RUBY_TYPED_EMBEDDABLE +}; + +#define r_entry(v, arg) r_entry0((v), (arg)->data->num_entries, (arg)) +static VALUE r_object(struct load_arg *arg); +static VALUE r_symbol(struct load_arg *arg); + +NORETURN(static void too_short(void)); +static void +too_short(void) +{ + rb_raise(rb_eArgError, "marshal data too short"); +} + +static st_index_t +r_prepare(struct load_arg *arg) +{ + st_index_t idx = arg->data->num_entries; + + st_insert(arg->data, (st_data_t)idx, (st_data_t)Qundef); + return idx; +} + +static unsigned char +r_byte1_buffered(struct load_arg *arg) +{ + if (arg->buflen == 0) { + long readable = arg->readable < arg->bufsize ? arg->readable : arg->bufsize; + long read_len; + VALUE str, n = LONG2NUM(readable); + + str = load_funcall(arg, arg->src, s_read, 1, &n); + if (NIL_P(str)) too_short(); + StringValue(str); + read_len = RSTRING_LEN(str); + if (UNLIKELY(read_len < readable)) too_short(); + if (UNLIKELY(read_len > arg->bufsize)) { + arg->buf = ruby_sized_realloc_n(arg->buf, read_len, 1, arg->bufsize); + arg->bufsize = read_len; + } + memcpy(arg->buf, RSTRING_PTR(str), read_len); + arg->offset = 0; + arg->buflen = read_len; + RB_GC_GUARD(str); + } + arg->buflen--; + return arg->buf[arg->offset++]; +} static int -r_byte(arg) - struct load_arg *arg; +r_byte(struct load_arg *arg) { int c; - if (TYPE(arg->src) == T_STRING) { - if (RSTRING(arg->src)->len > arg->offset) { - c = (unsigned char)RSTRING(arg->src)->ptr[arg->offset++]; - } - else { - rb_raise(rb_eArgError, "marshal data too short"); - } + if (RB_TYPE_P(arg->src, T_STRING)) { + if (RSTRING_LEN(arg->src) > arg->offset) { + c = (unsigned char)RSTRING_PTR(arg->src)[arg->offset++]; + } + else { + too_short(); + } } else { - VALUE src = arg->src; - VALUE v = rb_funcall2(src, s_getc, 0, 0); - if (NIL_P(v)) rb_eof_error(); - c = (unsigned char)FIX2INT(v); + if (arg->readable >0 || arg->buflen > 0) { + c = r_byte1_buffered(arg); + } + else { + VALUE v = load_funcall(arg, arg->src, s_getbyte, 0, 0); + if (NIL_P(v)) rb_eof_error(); + c = (unsigned char)NUM2CHR(v); + } } return c; } +NORETURN(static void long_toobig(int size)); + static void -long_toobig(size) - int size; +long_toobig(int size) { - rb_raise(rb_eTypeError, "long too big for this architecture (size %d, given %d)", - sizeof(long), size); + rb_raise(rb_eTypeError, "long too big for this architecture (size " + STRINGIZE(SIZEOF_LONG)", given %d)", size); } -#undef SIGN_EXTEND_CHAR -#if __STDC__ -# define SIGN_EXTEND_CHAR(c) ((signed char)(c)) -#else /* not __STDC__ */ -/* As in Harbison and Steele. */ -# define SIGN_EXTEND_CHAR(c) ((((unsigned char)(c)) ^ 128) - 128) -#endif - static long -r_long(arg) - struct load_arg *arg; +r_long(struct load_arg *arg) { register long x; - int c = SIGN_EXTEND_CHAR(r_byte(arg)); + int c = (signed char)r_byte(arg); long i; if (c == 0) return 0; if (c > 0) { - if (4 < c && c < 128) { - return c - 5; - } - if (c > sizeof(long)) long_toobig(c); - x = 0; - for (i=0;i<c;i++) { - x |= (long)r_byte(arg) << (8*i); - } + if (4 < c && c < 128) { + return c - 5; + } + if (c > (int)sizeof(long)) long_toobig(c); + x = 0; + for (i=0;i<c;i++) { + x |= (long)r_byte(arg) << (8*i); + } } else { - if (-129 < c && c < -4) { - return c + 5; - } - c = -c; - if (c > sizeof(long)) long_toobig(c); - x = -1; - for (i=0;i<c;i++) { - x &= ~((long)0xff << (8*i)); - x |= (long)r_byte(arg) << (8*i); - } + if (-129 < c && c < -4) { + return c + 5; + } + c = -c; + if (c > (int)sizeof(long)) long_toobig(c); + x = -1; + for (i=0;i<c;i++) { + x &= ~((long)0xff << (8*i)); + x |= (long)r_byte(arg) << (8*i); + } } return x; } +long +ruby_marshal_read_long(const char **buf, long len) +{ + long x; + struct RString src = {RBASIC_INIT}; + struct load_arg arg; + memset(&arg, 0, sizeof(arg)); + arg.src = rb_setup_fake_str(&src, *buf, len, 0); + x = r_long(&arg); + *buf += arg.offset; + return x; +} + +static long +r_keep_readable(struct load_arg *arg, long len, size_t size) +{ + if (UNLIKELY(len < 0)) { + rb_raise(rb_eArgError, "negative length"); + } + if (UNLIKELY((unsigned long)len > SIZE_MAX / size || arg->readable >= LONG_MAX - len)) { + rb_raise(rb_eArgError, "marshaled data too big"); + } + return len; +} + +static VALUE +r_bytes1(long len, struct load_arg *arg) +{ + VALUE str, n = LONG2NUM(len); + + str = load_funcall(arg, arg->src, s_read, 1, &n); + if (NIL_P(str)) too_short(); + StringValue(str); + if (RSTRING_LEN(str) != len) too_short(); + + return str; +} + +static VALUE +r_bytes1_buffered(long len, struct load_arg *arg) +{ + VALUE str; + + if (len <= arg->buflen) { + str = rb_str_new(arg->buf+arg->offset, len); + arg->offset += len; + arg->buflen -= len; + } + else { + long buflen = arg->buflen; + long readable = arg->readable + 1; + long tmp_len, read_len, need_len = len - buflen; + VALUE tmp, n; + + readable = readable < arg->bufsize ? readable : arg->bufsize; + read_len = need_len > readable ? need_len : readable; + n = LONG2NUM(read_len); + tmp = load_funcall(arg, arg->src, s_read, 1, &n); + if (NIL_P(tmp)) too_short(); + StringValue(tmp); + + tmp_len = RSTRING_LEN(tmp); + + if (tmp_len < need_len) too_short(); + + str = rb_str_new(arg->buf+arg->offset, buflen); + rb_str_cat(str, RSTRING_PTR(tmp), need_len); + + if (tmp_len > need_len) { + buflen = tmp_len - need_len; + memcpy(arg->buf, RSTRING_PTR(tmp)+need_len, buflen); + arg->buflen = buflen; + } + else { + arg->buflen = 0; + } + arg->offset = 0; + } + + return str; +} + #define r_bytes(arg) r_bytes0(r_long(arg), (arg)) static VALUE -r_bytes0(len, arg) - long len; - struct load_arg *arg; +r_bytes0(long len, struct load_arg *arg) { VALUE str; if (len == 0) return rb_str_new(0, 0); - if (TYPE(arg->src) == T_STRING) { - if (RSTRING(arg->src)->len > arg->offset) { - str = rb_str_new(RSTRING(arg->src)->ptr+arg->offset, len); - arg->offset += len; - } - else { - too_short: - rb_raise(rb_eArgError, "marshal data too short"); - } + if (RB_TYPE_P(arg->src, T_STRING)) { + if (RSTRING_LEN(arg->src) - arg->offset >= len) { + str = rb_str_new(RSTRING_PTR(arg->src)+arg->offset, len); + arg->offset += len; + } + else { + too_short(); + } } else { - VALUE src = arg->src; - VALUE n = LONG2NUM(len); - str = rb_funcall2(src, s_read, 1, &n); - if (NIL_P(str)) goto too_short; - StringValue(str); - if (RSTRING(str)->len != len) goto too_short; - if (OBJ_TAINTED(str)) arg->taint = Qtrue; + if (arg->readable > 0 || arg->buflen > 0) { + str = r_bytes1_buffered(len, arg); + } + else { + str = r_bytes1(len, arg); + } } return str; } -static ID -r_symlink(arg) - struct load_arg *arg; +static inline int +name_equal(const char *name, size_t nlen, const char *p, long l) { - ID id; - long num = r_long(arg); + if ((size_t)l != nlen || *p != *name) return 0; + return nlen == 1 || memcmp(p+1, name+1, nlen-1) == 0; +} - if (st_lookup(arg->symbols, num, &id)) { - return id; +static int +sym2encidx(VALUE sym, VALUE val) +{ + RBIMPL_ATTR_NONSTRING() static const char name_encoding[8] = "encoding"; + const char *p; + long l; + if (rb_enc_get_index(sym) != ENCINDEX_US_ASCII) return -1; + RSTRING_GETMEM(sym, p, l); + if (l <= 0) return -1; + if (name_equal(name_encoding, sizeof(name_encoding), p, l)) { + int idx = rb_enc_find_index(StringValueCStr(val)); + return idx; } - rb_raise(rb_eArgError, "bad symbol"); + if (name_equal(name_s_encoding_short, rb_strlen_lit(name_s_encoding_short), p, l)) { + if (val == Qfalse) return rb_usascii_encindex(); + else if (val == Qtrue) return rb_utf8_encindex(); + /* bogus ignore */ + } + return -1; } -static ID -r_symreal(arg) - struct load_arg *arg; +static int +symname_equal(VALUE sym, const char *name, size_t nlen) { - ID id; + const char *p; + long l; + if (rb_enc_get_index(sym) != ENCINDEX_US_ASCII) return 0; + RSTRING_GETMEM(sym, p, l); + return name_equal(name, nlen, p, l); +} + +#define BUILD_ASSERT_POSITIVE(n) \ + /* make 0 negative to workaround the "zero size array" GCC extension, */ \ + ((sizeof(char [2*(ssize_t)(n)-1])+1)/2) /* assuming no overflow */ +#define symname_equal_lit(sym, sym_name) \ + symname_equal(sym, sym_name, BUILD_ASSERT_POSITIVE(rb_strlen_lit(sym_name))) + +static VALUE +r_symlink(struct load_arg *arg) +{ + st_data_t sym; + long num = r_long(arg); - id = rb_intern(RSTRING(r_bytes(arg))->ptr); - st_insert(arg->symbols, arg->symbols->num_entries, id); + if (!st_lookup(arg->symbols, num, &sym)) { + rb_raise(rb_eArgError, "bad symbol"); + } + return (VALUE)sym; +} + +static VALUE +r_symreal(struct load_arg *arg, int ivar) +{ + VALUE s = r_bytes(arg); + VALUE sym; + int idx = -1; + st_index_t n = arg->symbols->num_entries; + + if (rb_enc_str_asciionly_p(s)) rb_enc_associate_index(s, ENCINDEX_US_ASCII); + st_insert(arg->symbols, (st_data_t)n, (st_data_t)s); + if (ivar) { + long num = r_long(arg); + while (num-- > 0) { + sym = r_symbol(arg); + idx = sym2encidx(sym, r_object(arg)); + } + } + if (idx > 0) { + rb_enc_associate_index(s, idx); + if (is_broken_string(s)) { + rb_raise(rb_eArgError, "invalid byte sequence in %s: %+"PRIsVALUE, + rb_enc_name(rb_enc_from_index(idx)), s); + } + } - return id; + return s; } -static ID -r_symbol(arg) - struct load_arg *arg; +static VALUE +r_symbol(struct load_arg *arg) { - if (r_byte(arg) == TYPE_SYMLINK) { - return r_symlink(arg); + int type, ivar = 0; + + again: + switch ((type = r_byte(arg))) { + default: + rb_raise(rb_eArgError, "dump format error for symbol(0x%x)", type); + case TYPE_IVAR: + ivar = 1; + goto again; + case TYPE_SYMBOL: + return r_symreal(arg, ivar); + case TYPE_SYMLINK: + if (ivar) { + rb_raise(rb_eArgError, "dump format error (symlink with encoding)"); + } + return r_symlink(arg); } - return r_symreal(arg); } -static char* -r_unique(arg) - struct load_arg *arg; +static VALUE +r_unique(struct load_arg *arg) { - return rb_id2name(r_symbol(arg)); + return r_symbol(arg); } static VALUE -r_string(arg) - struct load_arg *arg; +r_string(struct load_arg *arg) { return r_bytes(arg); } static VALUE -r_entry(v, arg) - VALUE v; - struct load_arg *arg; +r_entry0(VALUE v, st_index_t num, struct load_arg *arg) { - rb_hash_aset(arg->data, INT2FIX(RHASH(arg->data)->tbl->num_entries), v); - if (arg->taint) OBJ_TAINT(v); + st_data_t real_obj = (st_data_t)v; + if (arg->compat_tbl) { + /* real_obj is kept if not found */ + st_lookup(arg->compat_tbl, v, &real_obj); + } + st_insert(arg->data, num, real_obj); + st_insert(arg->partial_objects, (st_data_t)real_obj, Qtrue); + return v; +} + +static VALUE +r_fixup_compat(VALUE v, struct load_arg *arg) +{ + st_data_t data; + st_data_t key = (st_data_t)v; + if (arg->compat_tbl && st_delete(arg->compat_tbl, &key, &data)) { + VALUE real_obj = (VALUE)data; + rb_alloc_func_t allocator = rb_get_alloc_func(CLASS_OF(real_obj)); + if (st_lookup(compat_allocator_tbl, (st_data_t)allocator, &data)) { + marshal_compat_t *compat = (marshal_compat_t*)data; + compat->loader(real_obj, v); + } + v = real_obj; + } return v; } +static VALUE +r_post_proc(VALUE v, struct load_arg *arg) +{ + if (arg->proc) { + v = load_funcall(arg, arg->proc, s_call, 1, &v); + } + return v; +} + +static VALUE +r_leave(VALUE v, struct load_arg *arg, bool partial) +{ + v = r_fixup_compat(v, arg); + if (!partial) { + st_data_t data; + st_data_t key = (st_data_t)v; + st_delete(arg->partial_objects, &key, &data); + if (arg->freeze) { + if (RB_TYPE_P(v, T_MODULE) || RB_TYPE_P(v, T_CLASS)) { + // noop + } + else if (RB_TYPE_P(v, T_STRING)) { + v = rb_str_to_interned_str(v); + } + else { + OBJ_FREEZE(v); + } + } + v = r_post_proc(v, arg); + } + return v; +} + +static int +copy_ivar_i(ID vid, VALUE value, st_data_t arg) +{ + VALUE obj = (VALUE)arg; + + if (!rb_ivar_defined(obj, vid)) + rb_ivar_set(obj, vid, value); + return ST_CONTINUE; +} + +static VALUE +r_copy_ivar(VALUE v, VALUE data) +{ + rb_ivar_foreach(data, copy_ivar_i, (st_data_t)v); + return v; +} + +#define override_ivar_error(type, str) \ + rb_raise(rb_eTypeError, \ + "can't override instance variable of "type" '%"PRIsVALUE"'", \ + (str)) + +static int +r_ivar_encoding(VALUE obj, struct load_arg *arg, VALUE sym, VALUE val) +{ + int idx = sym2encidx(sym, val); + if (idx >= 0) { + if (rb_enc_capable(obj)) { + // Check if needed to avoid rb_check_frozen() check for Regexps + if (rb_enc_get_index(obj) != idx) { + rb_enc_associate_index(obj, idx); + } + } + else { + rb_raise(rb_eArgError, "%"PRIsVALUE" is not enc_capable", obj); + } + return TRUE; + } + return FALSE; +} + +static long +r_encname(VALUE obj, struct load_arg *arg) +{ + long len = r_long(arg); + if (len > 0) { + VALUE sym = r_symbol(arg); + VALUE val = r_object(arg); + len -= r_ivar_encoding(obj, arg, sym, val); + } + return len; +} + static void -r_ivar(obj, arg) - VALUE obj; - struct load_arg *arg; +r_ivar(VALUE obj, int *has_encoding, struct load_arg *arg) { long len; len = r_long(arg); if (len > 0) { - while (len--) { - ID id = r_symbol(arg); - VALUE val = r_object(arg); - rb_ivar_set(obj, id, val); - } + if (RB_TYPE_P(obj, T_MODULE)) { + override_ivar_error("module", rb_mod_name(obj)); + } + else if (RB_TYPE_P(obj, T_CLASS)) { + override_ivar_error("class", rb_class_name(obj)); + } + do { + VALUE sym = r_symbol(arg); + VALUE val = r_object(arg); + if (r_ivar_encoding(obj, arg, sym, val)) { + if (has_encoding) *has_encoding = TRUE; + } + else if (symname_equal_lit(sym, name_s_ruby2_keywords_flag)) { + if (RB_TYPE_P(obj, T_HASH)) { + rb_hash_ruby2_keywords(obj); + } + else { + rb_raise(rb_eArgError, "ruby2_keywords flag is given but %"PRIsVALUE" is not a Hash", obj); + } + } + else { + rb_ivar_set(obj, rb_intern_str(sym), val); + } + } while (--len > 0); } } static VALUE -path2class(path) - char *path; +path2class(VALUE path) { - VALUE v = rb_path2class(path); + VALUE v = rb_path_to_class(path); - if (TYPE(v) != T_CLASS) { - rb_raise(rb_eArgError, "%s does not refer class", path); + if (!RB_TYPE_P(v, T_CLASS)) { + rb_raise(rb_eArgError, "%"PRIsVALUE" does not refer to class", path); } return v; } +#define path2module(path) must_be_module(rb_path_to_class(path), path) + static VALUE -path2module(path) - char *path; +must_be_module(VALUE v, VALUE path) { - VALUE v = rb_path2class(path); - - if (TYPE(v) != T_MODULE) { - rb_raise(rb_eArgError, "%s does not refer module", path); + if (!RB_TYPE_P(v, T_MODULE)) { + rb_raise(rb_eArgError, "%"PRIsVALUE" does not refer to module", path); } return v; } static VALUE -r_object0(arg, proc, ivp, extmod) - struct load_arg *arg; - VALUE proc; - int *ivp; - VALUE extmod; +obj_alloc_by_klass(VALUE klass, struct load_arg *arg, VALUE *oldclass) +{ + st_data_t data; + rb_alloc_func_t allocator; + + allocator = rb_get_alloc_func(klass); + if (st_lookup(compat_allocator_tbl, (st_data_t)allocator, &data)) { + marshal_compat_t *compat = (marshal_compat_t*)data; + VALUE real_obj = rb_obj_alloc(klass); + VALUE obj = rb_obj_alloc(compat->oldclass); + if (oldclass) *oldclass = compat->oldclass; + + if (!arg->compat_tbl) { + arg->compat_tbl = rb_init_identtable(); + } + st_insert(arg->compat_tbl, (st_data_t)obj, (st_data_t)real_obj); + return obj; + } + + return rb_obj_alloc(klass); +} + +static VALUE +obj_alloc_by_path(VALUE path, struct load_arg *arg) +{ + return obj_alloc_by_klass(path2class(path), arg, 0); +} + +static VALUE +append_extmod(VALUE obj, VALUE extmod) +{ + long i = RARRAY_LEN(extmod); + while (i > 0) { + VALUE m = RARRAY_AREF(extmod, --i); + rb_extend_object(obj, m); + } + return obj; +} + +#define prohibit_ivar(type, str) do { \ + if (!ivp || !*ivp) break; \ + override_ivar_error(type, str); \ + } while (0) + +static VALUE r_object_for(struct load_arg *arg, bool partial, int *ivp, VALUE klass, VALUE extmod, int type); + +static VALUE +r_object0(struct load_arg *arg, bool partial, int *ivp, VALUE extmod) { - VALUE v = Qnil; int type = r_byte(arg); + return r_object_for(arg, partial, ivp, 0, extmod, type); +} + +static VALUE +r_object_for(struct load_arg *arg, bool partial, int *ivp, VALUE klass, VALUE extmod, int type) +{ + VALUE (*hash_new_with_size)(st_index_t) = rb_hash_new_with_size; + VALUE v = Qnil; long id; + st_data_t link; switch (type) { case TYPE_LINK: - id = r_long(arg); - v = rb_hash_aref(arg->data, LONG2FIX(id)); - if (NIL_P(v)) { - rb_raise(rb_eArgError, "dump format error (unlinked)"); - } - return v; + id = r_long(arg); + if (!st_lookup(arg->data, (st_data_t)id, &link)) { + rb_raise(rb_eArgError, "dump format error (unlinked)"); + } + v = (VALUE)link; + if (!st_lookup(arg->partial_objects, (st_data_t)v, &link)) { + if (arg->freeze && RB_TYPE_P(v, T_STRING)) { + v = rb_str_to_interned_str(v); + } + v = r_post_proc(v, arg); + } + break; case TYPE_IVAR: { - int ivar = Qtrue; - - v = r_object0(arg, 0, &ivar, extmod); - if (ivar) r_ivar(v, arg); - } - break; + int ivar = TRUE; + v = r_object0(arg, true, &ivar, extmod); + if (ivar) r_ivar(v, NULL, arg); + v = r_leave(v, arg, partial); + } + break; case TYPE_EXTENDED: - { - VALUE m = path2module(r_unique(arg)); - - if (NIL_P(extmod)) extmod = rb_ary_new2(0); - rb_ary_push(extmod, m); + { + VALUE path = r_unique(arg); + VALUE m = rb_path_to_class(path); + if (NIL_P(extmod)) extmod = rb_ary_hidden_new(0); + + if (RB_TYPE_P(m, T_CLASS)) { /* prepended */ + VALUE c; + + v = r_object0(arg, true, 0, Qnil); + c = CLASS_OF(v); + if (c != m || FL_TEST(c, FL_SINGLETON)) { + rb_raise(rb_eArgError, + "prepended class %"PRIsVALUE" differs from class %"PRIsVALUE, + path, rb_class_name(c)); + } + c = rb_singleton_class(v); + while (RARRAY_LEN(extmod) > 0) { + m = rb_ary_pop(extmod); + rb_prepend_module(c, m); + } + } + else { + must_be_module(m, path); + rb_ary_push(extmod, m); - v = r_object0(arg, 0, 0, extmod); - while (RARRAY(extmod)->len > 0) { - m = rb_ary_pop(extmod); - rb_extend_object(v, m); + v = r_object0(arg, true, 0, extmod); + while (RARRAY_LEN(extmod) > 0) { + m = rb_ary_pop(extmod); + rb_extend_object(v, m); + } } - } - break; + v = r_leave(v, arg, partial); + } + break; case TYPE_UCLASS: - { - VALUE c = path2class(r_unique(arg)); - - if (FL_TEST(c, FL_SINGLETON)) { - rb_raise(rb_eTypeError, "singleton can't be loaded"); - } - v = r_object0(arg, 0, 0, extmod); - if (rb_special_const_p(v) || TYPE(v) == T_OBJECT || TYPE(v) == T_CLASS) { - format_error: - rb_raise(rb_eArgError, "dump format error (user class)"); - } - if (TYPE(v) == T_MODULE || !RTEST(rb_class_inherited_p(c, RBASIC(v)->klass))) { - VALUE tmp = rb_obj_alloc(c); - - if (TYPE(v) != TYPE(tmp)) goto format_error; - } - RBASIC(v)->klass = c; - } - break; + { + VALUE c = path2class(r_unique(arg)); + + if (FL_TEST(c, FL_SINGLETON)) { + rb_raise(rb_eTypeError, "singleton can't be loaded"); + } + type = r_byte(arg); + if ((c == rb_cHash) && + /* Hack for compare_by_identity */ + (type == TYPE_HASH || type == TYPE_HASH_DEF)) { + hash_new_with_size = rb_ident_hash_new_with_size; + goto type_hash; + } + v = r_object_for(arg, partial, 0, c, extmod, type); + if (RB_SPECIAL_CONST_P(v) || RB_TYPE_P(v, T_OBJECT) || RB_TYPE_P(v, T_CLASS)) { + goto format_error; + } + if (RB_TYPE_P(v, T_MODULE) || !RTEST(rb_class_inherited_p(c, RBASIC(v)->klass))) { + VALUE tmp = rb_obj_alloc(c); + + if (TYPE(v) != TYPE(tmp)) goto format_error; + } + RBASIC_SET_CLASS(v, c); + } + break; + + format_error: + rb_raise(rb_eArgError, "dump format error (user class)"); case TYPE_NIL: - v = Qnil; - break; + v = Qnil; + v = r_leave(v, arg, false); + break; case TYPE_TRUE: - v = Qtrue; - break; + v = Qtrue; + v = r_leave(v, arg, false); + break; case TYPE_FALSE: - v = Qfalse; - break; + v = Qfalse; + v = r_leave(v, arg, false); + break; case TYPE_FIXNUM: - { - long i = r_long(arg); - v = LONG2FIX(i); - } - break; + { + long i = r_long(arg); + v = LONG2FIX(i); + } + v = r_leave(v, arg, false); + break; case TYPE_FLOAT: - { - double d, t = 0.0; - VALUE str = r_bytes(arg); - const char *ptr = RSTRING(str)->ptr; - - if (strcmp(ptr, "nan") == 0) { - d = t / t; - } - else if (strcmp(ptr, "inf") == 0) { - d = 1.0 / t; - } - else if (strcmp(ptr, "-inf") == 0) { - d = -1.0 / t; - } - else { - char *e; - d = strtod(ptr, &e); - d = load_mantissa(d, e, RSTRING(str)->len - (e - ptr)); - } - v = rb_float_new(d); - r_entry(v, arg); - } - break; + { + double d; + VALUE str = r_bytes(arg); + const char *ptr = RSTRING_PTR(str); + + if (strcmp(ptr, "nan") == 0) { + d = nan(""); + } + else if (strcmp(ptr, "inf") == 0) { + d = HUGE_VAL; + } + else if (strcmp(ptr, "-inf") == 0) { + d = -HUGE_VAL; + } + else { + char *e; + d = strtod(ptr, &e); + d = load_mantissa(d, e, RSTRING_LEN(str) - (e - ptr)); + } + v = DBL2NUM(d); + v = r_entry(v, arg); + v = r_leave(v, arg, false); + } + break; case TYPE_BIGNUM: - { - long len; - BDIGIT *digits; - volatile VALUE data; - - NEWOBJ(big, struct RBignum); - OBJSETUP(big, rb_cBignum, T_BIGNUM); - big->sign = (r_byte(arg) == '+'); - len = r_long(arg); - data = r_bytes0(len * 2, arg); -#if SIZEOF_BDIGITS == SIZEOF_SHORT - big->len = len; -#else - big->len = (len + 1) * 2 / sizeof(BDIGIT); -#endif - big->digits = digits = ALLOC_N(BDIGIT, big->len); - MEMCPY(digits, RSTRING(data)->ptr, char, len * 2); -#if SIZEOF_BDIGITS > SIZEOF_SHORT - MEMZERO((char *)digits + len * 2, char, - big->len * sizeof(BDIGIT) - len * 2); -#endif - len = big->len; - while (len > 0) { - unsigned char *p = (unsigned char *)digits; - BDIGIT num = 0; -#if SIZEOF_BDIGITS > SIZEOF_SHORT - int shift = 0; - int i; - - for (i=0; i<SIZEOF_BDIGITS; i++) { - num |= (int)p[i] << shift; - shift += 8; - } + { + long len; + VALUE data; + int sign; + + sign = r_byte(arg); + if (sign != '+' && sign != '-') { + rb_raise(rb_eArgError, "invalid Bignum sign"); + } + len = r_keep_readable(arg, r_long(arg), 2); + + if (SIZEOF_VALUE >= 8 && len <= 4) { + // Representable within uintptr, likely FIXNUM + VALUE num = 0; + for (int i = 0; i < len; i++) { + num |= (VALUE)r_byte(arg) << (i * 16); + num |= (VALUE)r_byte(arg) << (i * 16 + 8); + } +#if SIZEOF_VALUE == SIZEOF_LONG + v = ULONG2NUM(num); #else - num = p[0] | (p[1] << 8); + v = ULL2NUM(num); #endif - *digits++ = num; - len--; - } - v = rb_big_norm((VALUE)big); - r_entry(v, arg); - } - break; + if (sign == '-') { + v = rb_int_uminus(v); + } + } + else { + data = r_bytes0(len * 2, arg); + v = rb_integer_unpack(RSTRING_PTR(data), len, 2, 0, + INTEGER_PACK_LITTLE_ENDIAN | (sign == '-' ? INTEGER_PACK_NEGATIVE : 0)); + rb_str_resize(data, 0L); + } + v = r_entry(v, arg); + v = r_leave(v, arg, false); + } + break; case TYPE_STRING: - v = r_entry(r_string(arg), arg); - break; + v = r_entry(r_string(arg), arg); + v = r_leave(v, arg, partial); + break; case TYPE_REGEXP: - { - volatile VALUE str = r_bytes(arg); - int options = r_byte(arg); - v = r_entry(rb_reg_new(RSTRING(str)->ptr, RSTRING(str)->len, options), arg); - } - break; + { + VALUE str = r_bytes(arg); + int options = r_byte(arg); + int has_encoding = FALSE; + st_index_t idx = r_prepare(arg); + + if (ivp) { + r_ivar(str, &has_encoding, arg); + *ivp = FALSE; + } + if (!has_encoding) { + /* 1.8 compatibility; remove escapes undefined in 1.8 */ + char *ptr = RSTRING_PTR(str), *dst = ptr, *src = ptr; + long len = RSTRING_LEN(str); + long bs = 0; + for (; len-- > 0; *dst++ = *src++) { + switch (*src) { + case '\\': bs++; break; + case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': + case 'm': case 'o': case 'p': case 'q': case 'u': case 'y': + case 'E': case 'F': case 'H': case 'I': case 'J': case 'K': + case 'L': case 'N': case 'O': case 'P': case 'Q': case 'R': + case 'S': case 'T': case 'U': case 'V': case 'X': case 'Y': + if (bs & 1) --dst; + /* fall through */ + default: bs = 0; break; + } + } + rb_str_set_len(str, dst - ptr); + } + if (!klass) { + klass = rb_cRegexp; + } + VALUE regexp = rb_reg_init_str(rb_reg_s_alloc(klass), str, options); + r_copy_ivar(regexp, str); - case TYPE_ARRAY: - { - volatile long len = r_long(arg); /* gcc 2.7.2.3 -O2 bug?? */ + v = r_entry0(regexp, idx, arg); + v = r_leave(v, arg, partial); + } + break; - v = rb_ary_new2(len); - r_entry(v, arg); - while (len--) { - rb_ary_push(v, r_object(arg)); - } - } - break; + case TYPE_ARRAY: + { + long len = r_keep_readable(arg, r_long(arg), 1); + + v = rb_ary_new2(len); + v = r_entry(v, arg); + arg->readable += len - 1; + while (len--) { + rb_ary_push(v, r_object(arg)); + arg->readable--; + } + v = r_leave(v, arg, partial); + arg->readable++; + } + break; case TYPE_HASH: case TYPE_HASH_DEF: - { - long len = r_long(arg); - - v = rb_hash_new(); - r_entry(v, arg); - while (len--) { - VALUE key = r_object(arg); - VALUE value = r_object(arg); - rb_hash_aset(v, key, value); - } - if (type == TYPE_HASH_DEF) { - RHASH(v)->ifnone = r_object(arg); - } - } - break; + type_hash: + { + long len = r_keep_readable(arg, r_long(arg), 2); + + v = hash_new_with_size(len); + v = r_entry(v, arg); + arg->readable += (len - 1) * 2; + while (len--) { + VALUE key = r_object(arg); + VALUE value = r_object(arg); + rb_hash_aset(v, key, value); + arg->readable -= 2; + } + arg->readable += 2; + if (type == TYPE_HASH_DEF) { + RHASH_SET_IFNONE(v, r_object(arg)); + } + v = r_leave(v, arg, partial); + } + break; case TYPE_STRUCT: - { - VALUE klass, mem, values; - volatile long i; /* gcc 2.7.2.3 -O2 bug?? */ - long len; - ID slot; - - klass = path2class(r_unique(arg)); - mem = rb_struct_s_members(klass); - if (mem == Qnil) { - rb_raise(rb_eTypeError, "uninitialized struct"); - } - len = r_long(arg); - - values = rb_ary_new2(len); - for (i=0; i<len; i++) { - rb_ary_push(values, Qnil); - } - v = rb_struct_alloc(klass, values); - r_entry(v, arg); - for (i=0; i<len; i++) { - slot = r_symbol(arg); - - if (RARRAY(mem)->ptr[i] != ID2SYM(slot)) { - rb_raise(rb_eTypeError, "struct %s not compatible (:%s for :%s)", - rb_class2name(klass), - rb_id2name(slot), - rb_id2name(SYM2ID(RARRAY(mem)->ptr[i]))); - } - rb_struct_aset(v, LONG2FIX(i), r_object(arg)); - } - } - break; + { + VALUE mem, values; + long i; + VALUE slot; + st_index_t idx = r_prepare(arg); + VALUE klass = path2class(r_unique(arg)); + long len = r_keep_readable(arg, r_long(arg), 2); + + v = rb_obj_alloc(klass); + if (!RB_TYPE_P(v, T_STRUCT)) { + rb_raise(rb_eTypeError, "class %"PRIsVALUE" not a struct", rb_class_name(klass)); + } + mem = rb_struct_s_members(klass); + if (RARRAY_LEN(mem) != len) { + rb_raise(rb_eTypeError, "struct %"PRIsVALUE" not compatible (struct size differs)", + rb_class_name(klass)); + } + + arg->readable += (len - 1) * 2; + v = r_entry0(v, idx, arg); + values = rb_ary_new2(len); + { + VALUE keywords = Qfalse; + if (RTEST(rb_struct_s_keyword_init(klass))) { + keywords = rb_hash_new(); + rb_ary_push(values, keywords); + } + + for (i=0; i<len; i++) { + VALUE n = rb_sym2str(RARRAY_AREF(mem, i)); + slot = r_symbol(arg); + + if (!rb_str_equal(n, slot)) { + rb_raise(rb_eTypeError, "struct %"PRIsVALUE" not compatible (:%"PRIsVALUE" for :%"PRIsVALUE")", + rb_class_name(klass), + slot, n); + } + if (keywords) { + rb_hash_aset(keywords, RARRAY_AREF(mem, i), r_object(arg)); + } + else { + rb_ary_push(values, r_object(arg)); + } + arg->readable -= 2; + } + } + rb_struct_initialize(v, values); + v = r_leave(v, arg, partial); + arg->readable += 2; + } + break; case TYPE_USERDEF: { - VALUE klass = path2class(r_unique(arg)); - VALUE data; - - if (!rb_respond_to(klass, s_load)) { - rb_raise(rb_eTypeError, "class %s needs to have method `_load'", - rb_class2name(klass)); - } - data = r_string(arg); - if (ivp) { - r_ivar(data, arg); - *ivp = Qfalse; - } - v = rb_funcall(klass, s_load, 1, data); - r_entry(v, arg); - } + VALUE name = r_unique(arg); + VALUE klass = path2class(name); + VALUE data; + st_data_t d; + + if (!rb_obj_respond_to(klass, s_load, TRUE)) { + rb_raise(rb_eTypeError, "class %"PRIsVALUE" needs to have method '_load'", + name); + } + data = r_string(arg); + if (ivp) { + r_ivar(data, NULL, arg); + *ivp = FALSE; + } + v = load_funcall(arg, klass, s_load, 1, &data); + v = r_entry(v, arg); + if (st_lookup(compat_allocator_tbl, (st_data_t)rb_get_alloc_func(klass), &d)) { + marshal_compat_t *compat = (marshal_compat_t*)d; + v = compat->loader(klass, v); + } + if (!partial) { + if (arg->freeze) { + OBJ_FREEZE(v); + } + v = r_post_proc(v, arg); + } + } break; case TYPE_USRMARSHAL: { - VALUE klass = path2class(r_unique(arg)); - VALUE data; - - v = rb_obj_alloc(klass); - if (! NIL_P(extmod)) { - while (RARRAY(extmod)->len > 0) { - VALUE m = rb_ary_pop(extmod); - rb_extend_object(v, m); - } + VALUE name = r_unique(arg); + VALUE klass = path2class(name); + VALUE oldclass = 0; + VALUE data; + + v = obj_alloc_by_klass(klass, arg, &oldclass); + if (!NIL_P(extmod)) { + /* for the case marshal_load is overridden */ + append_extmod(v, extmod); + } + if (!rb_obj_respond_to(v, s_mload, TRUE)) { + rb_raise(rb_eTypeError, "instance of %"PRIsVALUE" needs to have method 'marshal_load'", + name); + } + v = r_entry(v, arg); + data = r_object(arg); + load_funcall(arg, v, s_mload, 1, &data); + v = r_fixup_compat(v, arg); + v = r_copy_ivar(v, data); + if (arg->freeze) { + OBJ_FREEZE(v); } - if (!rb_respond_to(v, s_mload)) { - rb_raise(rb_eTypeError, "instance of %s needs to have method `marshal_load'", - rb_class2name(klass)); - } - r_entry(v, arg); - data = r_object(arg); - rb_funcall(v, s_mload, 1, data); - } + v = r_post_proc(v, arg); + if (!NIL_P(extmod)) { + if (oldclass) append_extmod(v, extmod); + rb_ary_clear(extmod); + } + } break; case TYPE_OBJECT: - { - VALUE klass = path2class(r_unique(arg)); - - v = rb_obj_alloc(klass); - if (TYPE(v) != T_OBJECT) { - rb_raise(rb_eArgError, "dump format error"); - } - r_entry(v, arg); - r_ivar(v, arg); - } - break; + { + st_index_t idx = r_prepare(arg); + v = obj_alloc_by_path(r_unique(arg), arg); + if (!RB_TYPE_P(v, T_OBJECT)) { + rb_raise(rb_eArgError, "dump format error"); + } + v = r_entry0(v, idx, arg); + r_ivar(v, NULL, arg); + v = r_leave(v, arg, partial); + } + break; case TYPE_DATA: - { - VALUE klass = path2class(r_unique(arg)); - if (rb_respond_to(klass, s_alloc)) { - static int warn = Qtrue; - if (warn) { - rb_warn("define `allocate' instead of `_alloc'"); - warn = Qfalse; - } - v = rb_funcall(klass, s_alloc, 0); - } - else { - v = rb_obj_alloc(klass); - } - if (TYPE(v) != T_DATA) { - rb_raise(rb_eArgError, "dump format error"); - } - r_entry(v, arg); - if (!rb_respond_to(v, s_load_data)) { - rb_raise(rb_eTypeError, - "class %s needs to have instance method `_load_data'", - rb_class2name(klass)); - } - rb_funcall(v, s_load_data, 1, r_object0(arg, 0, 0, extmod)); - } - break; + { + VALUE name = r_unique(arg); + VALUE klass = path2class(name); + VALUE oldclass = 0; + VALUE r; + + v = obj_alloc_by_klass(klass, arg, &oldclass); + if (!RB_TYPE_P(v, T_DATA)) { + rb_raise(rb_eArgError, "dump format error"); + } + v = r_entry(v, arg); + if (!rb_obj_respond_to(v, s_load_data, TRUE)) { + rb_raise(rb_eTypeError, + "class %"PRIsVALUE" needs to have instance method '_load_data'", + name); + } + r = r_object0(arg, partial, 0, extmod); + load_funcall(arg, v, s_load_data, 1, &r); + v = r_leave(v, arg, partial); + } + break; case TYPE_MODULE_OLD: { - volatile VALUE str = r_bytes(arg); + VALUE str = r_bytes(arg); - v = rb_path2class(RSTRING(str)->ptr); - r_entry(v, arg); - } - break; + v = rb_path_to_class(str); + prohibit_ivar("class/module", str); + v = r_entry(v, arg); + v = r_leave(v, arg, partial); + } + break; case TYPE_CLASS: { - volatile VALUE str = r_bytes(arg); - - v = path2class(RSTRING(str)->ptr); - r_entry(v, arg); - } - break; + VALUE str = r_bytes(arg); + + if (ivp && *ivp > 0) *ivp = r_encname(str, arg) > 0; + v = path2class(str); + prohibit_ivar("class", str); + v = r_entry(v, arg); + v = r_leave(v, arg, partial); + } + break; case TYPE_MODULE: { - volatile VALUE str = r_bytes(arg); - - v = path2module(RSTRING(str)->ptr); - r_entry(v, arg); - } - break; + VALUE str = r_bytes(arg); + + if (ivp && *ivp > 0) *ivp = r_encname(str, arg) > 0; + v = path2module(str); + prohibit_ivar("module", str); + v = r_entry(v, arg); + v = r_leave(v, arg, partial); + } + break; case TYPE_SYMBOL: - v = ID2SYM(r_symreal(arg)); - break; + if (ivp) { + v = r_symreal(arg, *ivp); + *ivp = FALSE; + } + else { + v = r_symreal(arg, 0); + } + v = rb_str_intern(v); + v = r_leave(v, arg, partial); + break; case TYPE_SYMLINK: - return ID2SYM(r_symlink(arg)); + v = rb_str_intern(r_symlink(arg)); + break; default: - rb_raise(rb_eArgError, "dump format error(0x%x)", type); - break; + rb_raise(rb_eArgError, "dump format error(0x%x)", type); + break; } - if (proc) { - rb_funcall(proc, rb_intern("call"), 1, v); + + if (UNDEF_P(v)) { + rb_raise(rb_eArgError, "dump format error (bad link)"); } - return v; -} -static VALUE -r_object(arg) - struct load_arg *arg; -{ - return r_object0(arg, arg->proc, 0, Qnil); + return v; } static VALUE -load(arg) - struct load_arg *arg; +r_object(struct load_arg *arg) { - return r_object(arg); + return r_object0(arg, false, 0, Qnil); } -static VALUE -load_ensure(arg) - struct load_arg *arg; +static void +clear_load_arg(struct load_arg *arg) { + ruby_xfree_sized(arg->buf, arg->bufsize); + arg->buf = NULL; + arg->bufsize = 0; + arg->buflen = 0; + arg->offset = 0; + arg->readable = 0; + if (!arg->symbols) return; st_free_table(arg->symbols); - return 0; + arg->symbols = 0; + st_free_table(arg->data); + arg->data = 0; + st_free_table(arg->partial_objects); + arg->partial_objects = 0; + if (arg->compat_tbl) { + st_free_table(arg->compat_tbl); + arg->compat_tbl = 0; + } } -/* - * call-seq: - * load( source [, proc] ) => obj - * restore( source [, proc] ) => obj - * - * Returns the result of converting the serialized data in source into a - * Ruby object (possibly with associated subordinate objects). source - * may be either an instance of IO or an object that responds to - * to_str. If proc is specified, it will be passed each object as it - * is deserialized. - */ -static VALUE -marshal_load(argc, argv) - int argc; - VALUE *argv; +VALUE +rb_marshal_load_with_proc(VALUE port, VALUE proc, bool freeze) { - VALUE port, proc; int major, minor; VALUE v; - struct load_arg arg; + VALUE wrapper; /* used to avoid memory leak in case of exception */ + struct load_arg *arg; - rb_scan_args(argc, argv, "11", &port, &proc); - if (rb_respond_to(port, rb_intern("to_str"))) { - arg.taint = OBJ_TAINTED(port); /* original taintedness */ - StringValue(port); /* possible conversion */ + v = rb_check_string_type(port); + if (!NIL_P(v)) { + port = v; } - else if (rb_respond_to(port, s_getc) && rb_respond_to(port, s_read)) { - if (rb_respond_to(port, s_binmode)) { - rb_funcall2(port, s_binmode, 0, 0); - } - arg.taint = Qtrue; + else if (rb_respond_to(port, s_getbyte) && rb_respond_to(port, s_read)) { + rb_check_funcall(port, s_binmode, 0, 0); } else { - rb_raise(rb_eTypeError, "instance of IO needed"); + io_needed(); + } + wrapper = TypedData_Make_Struct(0, struct load_arg, &load_arg_data, arg); + arg->src = port; + arg->offset = 0; + arg->symbols = st_init_numtable(); + arg->data = rb_init_identtable(); + arg->partial_objects = rb_init_identtable(); + arg->compat_tbl = 0; + arg->proc = 0; + arg->readable = 0; + arg->freeze = freeze; + + if (NIL_P(v)) { + arg->bufsize = BUFSIZ; + arg->buf = xmalloc(BUFSIZ); + } + else { + arg->bufsize = 0; + arg->buf = 0; } - arg.src = port; - arg.offset = 0; - major = r_byte(&arg); - minor = r_byte(&arg); + major = r_byte(arg); + minor = r_byte(arg); if (major != MARSHAL_MAJOR || minor > MARSHAL_MINOR) { - rb_raise(rb_eTypeError, "incompatible marshal file format (can't be read)\n\ + clear_load_arg(arg); + rb_raise(rb_eTypeError, "incompatible marshal file format (can't be read)\n\ \tformat version %d.%d required; %d.%d given", - MARSHAL_MAJOR, MARSHAL_MINOR, major, minor); + MARSHAL_MAJOR, MARSHAL_MINOR, major, minor); } if (RTEST(ruby_verbose) && minor != MARSHAL_MINOR) { - rb_warn("incompatible marshal file format (can be read)\n\ + rb_warn("incompatible marshal file format (can be read)\n\ \tformat version %d.%d required; %d.%d given", - MARSHAL_MAJOR, MARSHAL_MINOR, major, minor); + MARSHAL_MAJOR, MARSHAL_MINOR, major, minor); } - arg.symbols = st_init_numtable(); - arg.data = rb_hash_new(); - if (NIL_P(proc)) arg.proc = 0; - else arg.proc = proc; - v = rb_ensure(load, (VALUE)&arg, load_ensure, (VALUE)&arg); + if (!NIL_P(proc)) arg->proc = proc; + v = r_object(arg); + clear_load_arg(arg); + RB_GC_GUARD(wrapper); return v; } +static VALUE +marshal_load(rb_execution_context_t *ec, VALUE mod, VALUE source, VALUE proc, VALUE freeze) +{ + return rb_marshal_load_with_proc(source, proc, RTEST(freeze)); +} + +#include "marshal.rbinc" + /* * The marshaling library converts collections of Ruby objects into a * byte stream, allowing them to be stored outside the currently * active script. This data may subsequently be read and the original * objects reconstituted. + * * Marshaled data has major and minor version numbers stored along * with the object information. In normal use, marshaling can only * load data written with the same major version number and an equal @@ -1446,65 +2471,195 @@ marshal_load(argc, argv) * first two bytes of marshaled data. * * str = Marshal.dump("thing") - * RUBY_VERSION #=> "1.8.0" - * str[0] #=> 4 - * str[1] #=> 8 + * RUBY_VERSION #=> "1.9.0" + * str[0].ord #=> 4 + * str[1].ord #=> 8 * * Some objects cannot be dumped: if the objects to be dumped include * bindings, procedure or method objects, instances of class IO, or * singleton objects, a TypeError will be raised. + * * If your class has special serialization needs (for example, if you * want to serialize in some specific format), or if it contains * objects that would otherwise not be serializable, you can implement - * your own serialization strategy by defining two methods, _dump and - * _load: - * The instance method _dump should return a String object containing - * all the information necessary to reconstitute objects of this class - * and all referenced objects up to a maximum depth given as an integer - * parameter (a value of -1 implies that you should disable depth checking). - * The class method _load should take a String and return an object of this class. + * your own serialization strategy. + * + * There are two methods of doing this, your object can define either + * marshal_dump and marshal_load or _dump and _load. marshal_dump will take + * precedence over _dump if both are defined. marshal_dump may result in + * smaller Marshal strings. + * + * == Security considerations + * + * By design, Marshal.load can deserialize almost any class loaded into the + * Ruby process. In many cases this can lead to remote code execution if the + * Marshal data is loaded from an untrusted source. + * + * As a result, Marshal.load is not suitable as a general purpose serialization + * format and you should never unmarshal user supplied input or other untrusted + * data. + * + * If you need to deserialize untrusted data, use JSON or another serialization + * format that is only able to load simple, 'primitive' types such as String, + * Array, Hash, etc. Never allow user input to specify arbitrary types to + * deserialize into. + * + * == marshal_dump and marshal_load + * + * When dumping an object the method marshal_dump will be called. + * marshal_dump must return a result containing the information necessary for + * marshal_load to reconstitute the object. The result can be any object. + * + * When loading an object dumped using marshal_dump the object is first + * allocated then marshal_load is called with the result from marshal_dump. + * marshal_load must recreate the object from the information in the result. + * + * Example: + * + * class MyObj + * def initialize name, version, data + * @name = name + * @version = version + * @data = data + * end + * + * def marshal_dump + * [@name, @version] + * end + * + * def marshal_load array + * @name, @version = array + * end + * end + * + * == _dump and _load + * + * Use _dump and _load when you need to allocate the object you're restoring + * yourself. + * + * When dumping an object the instance method _dump is called with an Integer + * which indicates the maximum depth of objects to dump (a value of -1 implies + * that you should disable depth checking). _dump must return a String + * containing the information necessary to reconstitute the object. + * + * The class method _load should take a String and use it to return an object + * of the same class. + * + * Example: + * + * class MyObj + * def initialize name, version, data + * @name = name + * @version = version + * @data = data + * end + * + * def _dump level + * [@name, @version].join ':' + * end + * + * def self._load args + * new(*args.split(':')) + * end + * end + * + * Since Marshal.dump outputs a string you can have _dump return a Marshal + * string which is Marshal.loaded in _load for complex objects. */ void -Init_marshal() +Init_marshal(void) { VALUE rb_mMarshal = rb_define_module("Marshal"); - - s_dump = rb_intern("_dump"); - s_load = rb_intern("_load"); - s_mdump = rb_intern("marshal_dump"); - s_mload = rb_intern("marshal_load"); - s_dump_data = rb_intern("_dump_data"); - s_load_data = rb_intern("_load_data"); - s_alloc = rb_intern("_alloc"); - s_getc = rb_intern("getc"); - s_read = rb_intern("read"); - s_write = rb_intern("write"); - s_binmode = rb_intern("binmode"); +#define set_id(sym) sym = rb_intern_const(name_##sym) + set_id(s_dump); + set_id(s_load); + set_id(s_mdump); + set_id(s_mload); + set_id(s_dump_data); + set_id(s_load_data); + set_id(s_alloc); + set_id(s_call); + set_id(s_getbyte); + set_id(s_read); + set_id(s_write); + set_id(s_binmode); + set_id(s_encoding_short); + set_id(s_ruby2_keywords_flag); rb_define_module_function(rb_mMarshal, "dump", marshal_dump, -1); - rb_define_module_function(rb_mMarshal, "load", marshal_load, -1); - rb_define_module_function(rb_mMarshal, "restore", marshal_load, -1); + /* major version */ rb_define_const(rb_mMarshal, "MAJOR_VERSION", INT2FIX(MARSHAL_MAJOR)); + /* minor version */ rb_define_const(rb_mMarshal, "MINOR_VERSION", INT2FIX(MARSHAL_MINOR)); } -VALUE -rb_marshal_dump(obj, port) - VALUE obj, port; +static int +marshal_compat_table_mark_and_move_i(st_data_t key, st_data_t value, st_data_t _) { - int argc = 1; - VALUE argv[2]; + marshal_compat_t *p = (marshal_compat_t *)value; + rb_gc_mark_and_move(&p->newclass); + rb_gc_mark_and_move(&p->oldclass); + return ST_CONTINUE; +} - argv[0] = obj; - argv[1] = port; - if (!NIL_P(port)) argc = 2; - return marshal_dump(argc, argv); +static void +marshal_compat_table_mark_and_move(void *tbl) +{ + if (!tbl) return; + st_foreach(tbl, marshal_compat_table_mark_and_move_i, 0); +} + +static int +marshal_compat_table_free_i(st_data_t key, st_data_t value, st_data_t _) +{ + SIZED_FREE((marshal_compat_t *)value); + return ST_CONTINUE; +} + +static void +marshal_compat_table_free(void *data) +{ + st_foreach(data, marshal_compat_table_free_i, 0); + st_free_table(data); +} + +static size_t +marshal_compat_table_memsize(const void *data) +{ + return st_memsize(data) + sizeof(marshal_compat_t) * st_table_size(data); +} + +static const rb_data_type_t marshal_compat_type = { + .wrap_struct_name = "marshal_compat_table", + .function = { + .dmark = marshal_compat_table_mark_and_move, + .dfree = marshal_compat_table_free, + .dsize = marshal_compat_table_memsize, + .dcompact = marshal_compat_table_mark_and_move, + }, + .flags = RUBY_TYPED_WB_PROTECTED | RUBY_TYPED_FREE_IMMEDIATELY, +}; + +static st_table * +compat_allocator_table(void) +{ + if (compat_allocator_tbl) return compat_allocator_tbl; + compat_allocator_tbl = st_init_numtable(); + compat_allocator_tbl_wrapper = + TypedData_Wrap_Struct(0, &marshal_compat_type, compat_allocator_tbl); + rb_vm_register_global_object(compat_allocator_tbl_wrapper); + return compat_allocator_tbl; +} + +VALUE +rb_marshal_dump(VALUE obj, VALUE port) +{ + return rb_marshal_dump_limited(obj, port, -1); } VALUE -rb_marshal_load(port) - VALUE port; +rb_marshal_load(VALUE port) { - return marshal_load(1, &port); + return rb_marshal_load_with_proc(port, Qnil, false); } |