Ruby 3.5.0dev (2025-08-02 revision 30a20bc166bc37acd7dcb3788686df149c7f428a)
iseq.c (30a20bc166bc37acd7dcb3788686df149c7f428a)
1/**********************************************************************
2
3 iseq.c -
4
5 $Author$
6 created at: 2006-07-11(Tue) 09:00:03 +0900
7
8 Copyright (C) 2006 Koichi Sasada
9
10**********************************************************************/
11
12#define RUBY_VM_INSNS_INFO 1
13/* #define RUBY_MARK_FREE_DEBUG 1 */
14
15#include "ruby/internal/config.h"
16
17#ifdef HAVE_DLADDR
18# include <dlfcn.h>
19#endif
20
21#include "eval_intern.h"
22#include "id_table.h"
23#include "internal.h"
24#include "internal/bits.h"
25#include "internal/class.h"
26#include "internal/compile.h"
27#include "internal/error.h"
28#include "internal/file.h"
29#include "internal/gc.h"
30#include "internal/hash.h"
31#include "internal/io.h"
32#include "internal/ruby_parser.h"
33#include "internal/sanitizers.h"
34#include "internal/set_table.h"
35#include "internal/symbol.h"
36#include "internal/thread.h"
37#include "internal/variable.h"
38#include "iseq.h"
39#include "ruby/util.h"
40#include "vm_core.h"
41#include "vm_callinfo.h"
42#include "yjit.h"
43#include "ruby/ractor.h"
44#include "builtin.h"
45#include "insns.inc"
46#include "insns_info.inc"
47#include "zjit.h"
48
49VALUE rb_cISeq;
50static VALUE iseqw_new(const rb_iseq_t *iseq);
51static const rb_iseq_t *iseqw_check(VALUE iseqw);
52
53#if VM_INSN_INFO_TABLE_IMPL == 2
54static struct succ_index_table *succ_index_table_create(int max_pos, int *data, int size);
55static unsigned int *succ_index_table_invert(int max_pos, struct succ_index_table *sd, int size);
56static int succ_index_lookup(const struct succ_index_table *sd, int x);
57#endif
58
59#define hidden_obj_p(obj) (!SPECIAL_CONST_P(obj) && !RBASIC(obj)->klass)
60
61static inline VALUE
62obj_resurrect(VALUE obj)
63{
64 if (hidden_obj_p(obj)) {
65 switch (BUILTIN_TYPE(obj)) {
66 case T_STRING:
67 obj = rb_str_resurrect(obj);
68 break;
69 case T_ARRAY:
70 obj = rb_ary_resurrect(obj);
71 break;
72 case T_HASH:
73 obj = rb_hash_resurrect(obj);
74 break;
75 default:
76 break;
77 }
78 }
79 return obj;
80}
81
82static void
83free_arena(struct iseq_compile_data_storage *cur)
84{
85 struct iseq_compile_data_storage *next;
86
87 while (cur) {
88 next = cur->next;
89 ruby_xfree(cur);
90 cur = next;
91 }
92}
93
94static void
95compile_data_free(struct iseq_compile_data *compile_data)
96{
97 if (compile_data) {
98 free_arena(compile_data->node.storage_head);
99 free_arena(compile_data->insn.storage_head);
100 if (compile_data->ivar_cache_table) {
101 rb_id_table_free(compile_data->ivar_cache_table);
102 }
103 ruby_xfree(compile_data);
104 }
105}
106
107static void
108remove_from_constant_cache(ID id, IC ic)
109{
110 rb_vm_t *vm = GET_VM();
111 VALUE lookup_result;
112 st_data_t ic_data = (st_data_t)ic;
113
114 if (rb_id_table_lookup(vm->constant_cache, id, &lookup_result)) {
115 set_table *ics = (set_table *)lookup_result;
116 set_table_delete(ics, &ic_data);
117
118 if (ics->num_entries == 0 &&
119 // See comment in vm_track_constant_cache on why we need this check
120 id != vm->inserting_constant_cache_id) {
121 rb_id_table_delete(vm->constant_cache, id);
122 set_free_table(ics);
123 }
124 }
125}
126
127// When an ISEQ is being freed, all of its associated ICs are going to go away
128// as well. Because of this, we need to iterate over the ICs, and clear them
129// from the VM's constant cache.
130static void
131iseq_clear_ic_references(const rb_iseq_t *iseq)
132{
133 // In some cases (when there is a compilation error), we end up with
134 // ic_size greater than 0, but no allocated is_entries buffer.
135 // If there's no is_entries buffer to loop through, return early.
136 // [Bug #19173]
137 if (!ISEQ_BODY(iseq)->is_entries) {
138 return;
139 }
140
141 for (unsigned int ic_idx = 0; ic_idx < ISEQ_BODY(iseq)->ic_size; ic_idx++) {
142 IC ic = &ISEQ_IS_IC_ENTRY(ISEQ_BODY(iseq), ic_idx);
143
144 // Iterate over the IC's constant path's segments and clean any references to
145 // the ICs out of the VM's constant cache table.
146 const ID *segments = ic->segments;
147
148 // It's possible that segments is NULL if we overallocated an IC but
149 // optimizations removed the instruction using it
150 if (segments == NULL)
151 continue;
152
153 for (int i = 0; segments[i]; i++) {
154 ID id = segments[i];
155 if (id == idNULL) continue;
156 remove_from_constant_cache(id, ic);
157 }
158
159 ruby_xfree((void *)segments);
160 }
161}
162
163void
164rb_iseq_free(const rb_iseq_t *iseq)
165{
166 RUBY_FREE_ENTER("iseq");
167
168 if (iseq && ISEQ_BODY(iseq)) {
169 iseq_clear_ic_references(iseq);
170 struct rb_iseq_constant_body *const body = ISEQ_BODY(iseq);
171#if USE_YJIT
172 rb_yjit_iseq_free(iseq);
173 if (FL_TEST_RAW((VALUE)iseq, ISEQ_TRANSLATED)) {
174 RUBY_ASSERT(rb_yjit_live_iseq_count > 0);
175 rb_yjit_live_iseq_count--;
176 }
177#endif
178 ruby_xfree((void *)body->iseq_encoded);
179 ruby_xfree((void *)body->insns_info.body);
180 ruby_xfree((void *)body->insns_info.positions);
181#if VM_INSN_INFO_TABLE_IMPL == 2
182 ruby_xfree(body->insns_info.succ_index_table);
183#endif
184 ruby_xfree((void *)body->is_entries);
185 ruby_xfree(body->call_data);
186 ruby_xfree((void *)body->catch_table);
187 ruby_xfree((void *)body->param.opt_table);
188 if (ISEQ_MBITS_BUFLEN(body->iseq_size) > 1 && body->mark_bits.list) {
189 ruby_xfree((void *)body->mark_bits.list);
190 }
191
192 ruby_xfree(body->variable.original_iseq);
193
194 if (body->param.keyword != NULL) {
195 if (body->param.keyword->table != &body->local_table[body->param.keyword->bits_start - body->param.keyword->num])
196 ruby_xfree((void *)body->param.keyword->table);
197 if (body->param.keyword->default_values) {
198 ruby_xfree((void *)body->param.keyword->default_values);
199 }
200 ruby_xfree((void *)body->param.keyword);
201 }
202 if (LIKELY(body->local_table != rb_iseq_shared_exc_local_tbl))
203 ruby_xfree((void *)body->local_table);
204 compile_data_free(ISEQ_COMPILE_DATA(iseq));
205 if (body->outer_variables) rb_id_table_free(body->outer_variables);
206 ruby_xfree(body);
207 }
208
209 if (iseq && ISEQ_EXECUTABLE_P(iseq) && iseq->aux.exec.local_hooks) {
210 rb_hook_list_free(iseq->aux.exec.local_hooks);
211 }
212
213 RUBY_FREE_LEAVE("iseq");
214}
215
216typedef VALUE iseq_value_itr_t(void *ctx, VALUE obj);
217
218static inline void
219iseq_scan_bits(unsigned int page, iseq_bits_t bits, VALUE *code, VALUE *original_iseq)
220{
221 unsigned int offset;
222 unsigned int page_offset = (page * ISEQ_MBITS_BITLENGTH);
223
224 while (bits) {
225 offset = ntz_intptr(bits);
226 VALUE op = code[page_offset + offset];
227 rb_gc_mark_and_move(&code[page_offset + offset]);
228 VALUE newop = code[page_offset + offset];
229 if (original_iseq && newop != op) {
230 original_iseq[page_offset + offset] = newop;
231 }
232 bits &= bits - 1; // Reset Lowest Set Bit (BLSR)
233 }
234}
235
236static void
237rb_iseq_mark_and_move_each_compile_data_value(const rb_iseq_t *iseq, VALUE *original_iseq)
238{
239 unsigned int size;
240 VALUE *code;
241 const struct iseq_compile_data *const compile_data = ISEQ_COMPILE_DATA(iseq);
242
243 size = compile_data->iseq_size;
244 code = compile_data->iseq_encoded;
245
246 // Embedded VALUEs
247 if (compile_data->mark_bits.list) {
248 if(compile_data->is_single_mark_bit) {
249 iseq_scan_bits(0, compile_data->mark_bits.single, code, original_iseq);
250 }
251 else {
252 for (unsigned int i = 0; i < ISEQ_MBITS_BUFLEN(size); i++) {
253 iseq_bits_t bits = compile_data->mark_bits.list[i];
254 iseq_scan_bits(i, bits, code, original_iseq);
255 }
256 }
257 }
258}
259static void
260rb_iseq_mark_and_move_each_body_value(const rb_iseq_t *iseq, VALUE *original_iseq)
261{
262 unsigned int size;
263 VALUE *code;
264 const struct rb_iseq_constant_body *const body = ISEQ_BODY(iseq);
265
266 size = body->iseq_size;
267 code = body->iseq_encoded;
268
269 union iseq_inline_storage_entry *is_entries = body->is_entries;
270
271 if (body->is_entries) {
272 // Skip iterating over ivc caches
273 is_entries += body->ivc_size;
274
275 // ICVARC entries
276 for (unsigned int i = 0; i < body->icvarc_size; i++, is_entries++) {
277 ICVARC icvarc = (ICVARC)is_entries;
278 if (icvarc->entry) {
279 RUBY_ASSERT(!RB_TYPE_P(icvarc->entry->class_value, T_NONE));
280
281 rb_gc_mark_and_move(&icvarc->entry->class_value);
282 }
283 }
284
285 // ISE entries
286 for (unsigned int i = 0; i < body->ise_size; i++, is_entries++) {
287 union iseq_inline_storage_entry *const is = (union iseq_inline_storage_entry *)is_entries;
288 if (is->once.value) {
289 rb_gc_mark_and_move(&is->once.value);
290 }
291 }
292
293 // IC Entries
294 for (unsigned int i = 0; i < body->ic_size; i++, is_entries++) {
295 IC ic = (IC)is_entries;
296 if (ic->entry) {
297 rb_gc_mark_and_move_ptr(&ic->entry);
298 }
299 }
300 }
301
302 // Embedded VALUEs
303 if (body->mark_bits.list) {
304 if (ISEQ_MBITS_BUFLEN(size) == 1) {
305 iseq_scan_bits(0, body->mark_bits.single, code, original_iseq);
306 }
307 else {
308 for (unsigned int i = 0; i < ISEQ_MBITS_BUFLEN(size); i++) {
309 iseq_bits_t bits = body->mark_bits.list[i];
310 iseq_scan_bits(i, bits, code, original_iseq);
311 }
312 }
313 }
314}
315
316static bool
317cc_is_active(const struct rb_callcache *cc, bool reference_updating)
318{
319 if (cc) {
320 if (cc == rb_vm_empty_cc() || rb_vm_empty_cc_for_super()) {
321 return false;
322 }
323
324 if (reference_updating) {
325 cc = (const struct rb_callcache *)rb_gc_location((VALUE)cc);
326 }
327
328 if (vm_cc_markable(cc) && vm_cc_valid(cc)) {
329 const struct rb_callable_method_entry_struct *cme = vm_cc_cme(cc);
330 if (reference_updating) {
331 cme = (const struct rb_callable_method_entry_struct *)rb_gc_location((VALUE)cme);
332 }
333 if (!METHOD_ENTRY_INVALIDATED(cme)) {
334 return true;
335 }
336 }
337 }
338 return false;
339}
340
341void
342rb_iseq_mark_and_move(rb_iseq_t *iseq, bool reference_updating)
343{
344 RUBY_MARK_ENTER("iseq");
345
346 rb_gc_mark_and_move(&iseq->wrapper);
347
348 if (ISEQ_BODY(iseq)) {
349 struct rb_iseq_constant_body *body = ISEQ_BODY(iseq);
350
351 rb_iseq_mark_and_move_each_body_value(iseq, reference_updating ? ISEQ_ORIGINAL_ISEQ(iseq) : NULL);
352
353 rb_gc_mark_and_move(&body->variable.coverage);
354 rb_gc_mark_and_move(&body->variable.pc2branchindex);
355 rb_gc_mark_and_move(&body->variable.script_lines);
356 rb_gc_mark_and_move(&body->location.label);
357 rb_gc_mark_and_move(&body->location.base_label);
358 rb_gc_mark_and_move(&body->location.pathobj);
359 if (body->local_iseq) rb_gc_mark_and_move_ptr(&body->local_iseq);
360 if (body->parent_iseq) rb_gc_mark_and_move_ptr(&body->parent_iseq);
361 if (body->mandatory_only_iseq) rb_gc_mark_and_move_ptr(&body->mandatory_only_iseq);
362
363 if (body->call_data) {
364 for (unsigned int i = 0; i < body->ci_size; i++) {
365 struct rb_call_data *cds = body->call_data;
366
367 if (cds[i].ci) rb_gc_mark_and_move_ptr(&cds[i].ci);
368
369 if (cc_is_active(cds[i].cc, reference_updating)) {
370 rb_gc_mark_and_move_ptr(&cds[i].cc);
371 }
372 else if (cds[i].cc != rb_vm_empty_cc()) {
373 cds[i].cc = rb_vm_empty_cc();
374 }
375 }
376 }
377
378 if (body->param.flags.has_kw && body->param.keyword != NULL) {
379 const struct rb_iseq_param_keyword *const keyword = body->param.keyword;
380
381 if (keyword->default_values != NULL) {
382 for (int j = 0, i = keyword->required_num; i < keyword->num; i++, j++) {
383 rb_gc_mark_and_move(&keyword->default_values[j]);
384 }
385 }
386 }
387
388 if (body->catch_table) {
389 struct iseq_catch_table *table = body->catch_table;
390
391 for (unsigned int i = 0; i < table->size; i++) {
392 struct iseq_catch_table_entry *entry;
393 entry = UNALIGNED_MEMBER_PTR(table, entries[i]);
394 if (entry->iseq) {
395 rb_gc_mark_and_move_ptr(&entry->iseq);
396 }
397 }
398 }
399
400 if (reference_updating) {
401#if USE_YJIT
402 rb_yjit_iseq_update_references(iseq);
403#endif
404#if USE_ZJIT
405 rb_zjit_iseq_update_references(body->zjit_payload);
406#endif
407 }
408 else {
409#if USE_YJIT
410 rb_yjit_iseq_mark(body->yjit_payload);
411#endif
412#if USE_ZJIT
413 rb_zjit_iseq_mark(body->zjit_payload);
414#endif
415 }
416 }
417
418 if (FL_TEST_RAW((VALUE)iseq, ISEQ_NOT_LOADED_YET)) {
419 rb_gc_mark_and_move(&iseq->aux.loader.obj);
420 }
421 else if (FL_TEST_RAW((VALUE)iseq, ISEQ_USE_COMPILE_DATA)) {
422 const struct iseq_compile_data *const compile_data = ISEQ_COMPILE_DATA(iseq);
423
424 rb_iseq_mark_and_move_insn_storage(compile_data->insn.storage_head);
425 rb_iseq_mark_and_move_each_compile_data_value(iseq, reference_updating ? ISEQ_ORIGINAL_ISEQ(iseq) : NULL);
426
427 rb_gc_mark_and_move((VALUE *)&compile_data->err_info);
428 rb_gc_mark_and_move((VALUE *)&compile_data->catch_table_ary);
429 }
430 else {
431 /* executable */
432 VM_ASSERT(ISEQ_EXECUTABLE_P(iseq));
433
434 if (iseq->aux.exec.local_hooks) {
435 rb_hook_list_mark_and_update(iseq->aux.exec.local_hooks);
436 }
437 }
438
439 RUBY_MARK_LEAVE("iseq");
440}
441
442static size_t
443param_keyword_size(const struct rb_iseq_param_keyword *pkw)
444{
445 size_t size = 0;
446
447 if (!pkw) return size;
448
449 size += sizeof(struct rb_iseq_param_keyword);
450 size += sizeof(VALUE) * (pkw->num - pkw->required_num);
451
452 return size;
453}
454
455size_t
456rb_iseq_memsize(const rb_iseq_t *iseq)
457{
458 size_t size = 0; /* struct already counted as RVALUE size */
459 const struct rb_iseq_constant_body *body = ISEQ_BODY(iseq);
460 const struct iseq_compile_data *compile_data;
461
462 /* TODO: should we count original_iseq? */
463
464 if (ISEQ_EXECUTABLE_P(iseq) && body) {
465 size += sizeof(struct rb_iseq_constant_body);
466 size += body->iseq_size * sizeof(VALUE);
467 size += body->insns_info.size * (sizeof(struct iseq_insn_info_entry) + sizeof(unsigned int));
468 size += body->local_table_size * sizeof(ID);
469 size += ISEQ_MBITS_BUFLEN(body->iseq_size) * ISEQ_MBITS_SIZE;
470 if (body->catch_table) {
471 size += iseq_catch_table_bytes(body->catch_table->size);
472 }
473 size += (body->param.opt_num + 1) * sizeof(VALUE);
474 size += param_keyword_size(body->param.keyword);
475
476 /* body->is_entries */
477 size += ISEQ_IS_SIZE(body) * sizeof(union iseq_inline_storage_entry);
478
479 if (ISEQ_BODY(iseq)->is_entries) {
480 /* IC entries constant segments */
481 for (unsigned int ic_idx = 0; ic_idx < body->ic_size; ic_idx++) {
482 IC ic = &ISEQ_IS_IC_ENTRY(body, ic_idx);
483 const ID *ids = ic->segments;
484 if (!ids) continue;
485 while (*ids++) {
486 size += sizeof(ID);
487 }
488 size += sizeof(ID); // null terminator
489 }
490 }
491
492 /* body->call_data */
493 size += body->ci_size * sizeof(struct rb_call_data);
494 // TODO: should we count imemo_callinfo?
495 }
496
497 compile_data = ISEQ_COMPILE_DATA(iseq);
498 if (compile_data) {
499 struct iseq_compile_data_storage *cur;
500
501 size += sizeof(struct iseq_compile_data);
502
503 cur = compile_data->node.storage_head;
504 while (cur) {
505 size += cur->size + offsetof(struct iseq_compile_data_storage, buff);
506 cur = cur->next;
507 }
508 }
509
510 return size;
511}
512
514rb_iseq_constant_body_alloc(void)
515{
516 struct rb_iseq_constant_body *iseq_body;
517 iseq_body = ZALLOC(struct rb_iseq_constant_body);
518 return iseq_body;
519}
520
521static rb_iseq_t *
522iseq_alloc(void)
523{
524 rb_iseq_t *iseq = iseq_imemo_alloc();
525 ISEQ_BODY(iseq) = rb_iseq_constant_body_alloc();
526 return iseq;
527}
528
529VALUE
530rb_iseq_pathobj_new(VALUE path, VALUE realpath)
531{
532 VALUE pathobj;
533 VM_ASSERT(RB_TYPE_P(path, T_STRING));
534 VM_ASSERT(NIL_P(realpath) || RB_TYPE_P(realpath, T_STRING));
535
536 if (path == realpath ||
537 (!NIL_P(realpath) && rb_str_cmp(path, realpath) == 0)) {
538 pathobj = rb_fstring(path);
539 }
540 else {
541 if (!NIL_P(realpath)) realpath = rb_fstring(realpath);
542 pathobj = rb_ary_new_from_args(2, rb_fstring(path), realpath);
543 rb_ary_freeze(pathobj);
544 }
545 return pathobj;
546}
547
548void
549rb_iseq_pathobj_set(const rb_iseq_t *iseq, VALUE path, VALUE realpath)
550{
551 RB_OBJ_WRITE(iseq, &ISEQ_BODY(iseq)->location.pathobj,
552 rb_iseq_pathobj_new(path, realpath));
553}
554
555// Make a dummy iseq for a dummy frame that exposes a path for profilers to inspect
556rb_iseq_t *
557rb_iseq_alloc_with_dummy_path(VALUE fname)
558{
559 rb_iseq_t *dummy_iseq = iseq_alloc();
560
561 ISEQ_BODY(dummy_iseq)->type = ISEQ_TYPE_TOP;
562 RB_OBJ_WRITE(dummy_iseq, &ISEQ_BODY(dummy_iseq)->location.pathobj, fname);
563 RB_OBJ_WRITE(dummy_iseq, &ISEQ_BODY(dummy_iseq)->location.label, fname);
564
565 return dummy_iseq;
566}
567
568static rb_iseq_location_t *
569iseq_location_setup(rb_iseq_t *iseq, VALUE name, VALUE path, VALUE realpath, int first_lineno, const rb_code_location_t *code_location, const int node_id)
570{
571 rb_iseq_location_t *loc = &ISEQ_BODY(iseq)->location;
572
573 rb_iseq_pathobj_set(iseq, path, realpath);
574 RB_OBJ_WRITE(iseq, &loc->label, name);
575 RB_OBJ_WRITE(iseq, &loc->base_label, name);
576 loc->first_lineno = first_lineno;
577
578 if (ISEQ_BODY(iseq)->local_iseq == iseq && strcmp(RSTRING_PTR(name), "initialize") == 0) {
579 ISEQ_BODY(iseq)->param.flags.use_block = 1;
580 }
581
582 if (code_location) {
583 loc->node_id = node_id;
584 loc->code_location = *code_location;
585 }
586 else {
587 loc->code_location.beg_pos.lineno = 0;
588 loc->code_location.beg_pos.column = 0;
589 loc->code_location.end_pos.lineno = -1;
590 loc->code_location.end_pos.column = -1;
591 }
592
593 return loc;
594}
595
596static void
597set_relation(rb_iseq_t *iseq, const rb_iseq_t *piseq)
598{
599 struct rb_iseq_constant_body *const body = ISEQ_BODY(iseq);
600 const VALUE type = body->type;
601
602 /* set class nest stack */
603 if (type == ISEQ_TYPE_TOP) {
604 body->local_iseq = iseq;
605 }
606 else if (type == ISEQ_TYPE_METHOD || type == ISEQ_TYPE_CLASS) {
607 body->local_iseq = iseq;
608 }
609 else if (piseq) {
610 RB_OBJ_WRITE(iseq, &body->local_iseq, ISEQ_BODY(piseq)->local_iseq);
611 }
612
613 if (piseq) {
614 RB_OBJ_WRITE(iseq, &body->parent_iseq, piseq);
615 }
616
617 if (type == ISEQ_TYPE_MAIN) {
618 body->local_iseq = iseq;
619 }
620}
621
622static struct iseq_compile_data_storage *
623new_arena(void)
624{
625 struct iseq_compile_data_storage * new_arena =
627 ALLOC_N(char, INITIAL_ISEQ_COMPILE_DATA_STORAGE_BUFF_SIZE +
628 offsetof(struct iseq_compile_data_storage, buff));
629
630 new_arena->pos = 0;
631 new_arena->next = 0;
632 new_arena->size = INITIAL_ISEQ_COMPILE_DATA_STORAGE_BUFF_SIZE;
633
634 return new_arena;
635}
636
637static VALUE
638prepare_iseq_build(rb_iseq_t *iseq,
639 VALUE name, VALUE path, VALUE realpath, int first_lineno, const rb_code_location_t *code_location, const int node_id,
640 const rb_iseq_t *parent, int isolated_depth, enum rb_iseq_type type,
641 VALUE script_lines, const rb_compile_option_t *option)
642{
643 VALUE coverage = Qfalse;
644 VALUE err_info = Qnil;
645 struct rb_iseq_constant_body *const body = ISEQ_BODY(iseq);
646
647 if (parent && (type == ISEQ_TYPE_MAIN || type == ISEQ_TYPE_TOP))
648 err_info = Qfalse;
649
650 body->type = type;
651 set_relation(iseq, parent);
652
653 name = rb_fstring(name);
654 iseq_location_setup(iseq, name, path, realpath, first_lineno, code_location, node_id);
655 if (iseq != body->local_iseq) {
656 RB_OBJ_WRITE(iseq, &body->location.base_label, ISEQ_BODY(body->local_iseq)->location.label);
657 }
658 ISEQ_COVERAGE_SET(iseq, Qnil);
659 ISEQ_ORIGINAL_ISEQ_CLEAR(iseq);
660 body->variable.flip_count = 0;
661
662 if (NIL_P(script_lines)) {
663 RB_OBJ_WRITE(iseq, &body->variable.script_lines, Qnil);
664 }
665 else {
666 RB_OBJ_WRITE(iseq, &body->variable.script_lines, rb_ractor_make_shareable(script_lines));
667 }
668
669 ISEQ_COMPILE_DATA_ALLOC(iseq);
670 RB_OBJ_WRITE(iseq, &ISEQ_COMPILE_DATA(iseq)->err_info, err_info);
671 RB_OBJ_WRITE(iseq, &ISEQ_COMPILE_DATA(iseq)->catch_table_ary, Qnil);
672
673 ISEQ_COMPILE_DATA(iseq)->node.storage_head = ISEQ_COMPILE_DATA(iseq)->node.storage_current = new_arena();
674 ISEQ_COMPILE_DATA(iseq)->insn.storage_head = ISEQ_COMPILE_DATA(iseq)->insn.storage_current = new_arena();
675 ISEQ_COMPILE_DATA(iseq)->isolated_depth = isolated_depth;
676 ISEQ_COMPILE_DATA(iseq)->option = option;
677 ISEQ_COMPILE_DATA(iseq)->ivar_cache_table = NULL;
678 ISEQ_COMPILE_DATA(iseq)->builtin_function_table = GET_VM()->builtin_function_table;
679
680 if (option->coverage_enabled) {
681 VALUE coverages = rb_get_coverages();
682 if (RTEST(coverages)) {
683 coverage = rb_hash_lookup(coverages, rb_iseq_path(iseq));
684 if (NIL_P(coverage)) coverage = Qfalse;
685 }
686 }
687 ISEQ_COVERAGE_SET(iseq, coverage);
688 if (coverage && ISEQ_BRANCH_COVERAGE(iseq))
689 ISEQ_PC2BRANCHINDEX_SET(iseq, rb_ary_hidden_new(0));
690
691 return Qtrue;
692}
693
694#if VM_CHECK_MODE > 0 && VM_INSN_INFO_TABLE_IMPL > 0
695static void validate_get_insn_info(const rb_iseq_t *iseq);
696#endif
697
698void
699rb_iseq_insns_info_encode_positions(const rb_iseq_t *iseq)
700{
701#if VM_INSN_INFO_TABLE_IMPL == 2
702 /* create succ_index_table */
703 struct rb_iseq_constant_body *const body = ISEQ_BODY(iseq);
704 int size = body->insns_info.size;
705 int max_pos = body->iseq_size;
706 int *data = (int *)body->insns_info.positions;
707 if (body->insns_info.succ_index_table) ruby_xfree(body->insns_info.succ_index_table);
708 body->insns_info.succ_index_table = succ_index_table_create(max_pos, data, size);
709#if VM_CHECK_MODE == 0
710 ruby_xfree(body->insns_info.positions);
711 body->insns_info.positions = NULL;
712#endif
713#endif
714}
715
716#if VM_INSN_INFO_TABLE_IMPL == 2
717unsigned int *
718rb_iseq_insns_info_decode_positions(const struct rb_iseq_constant_body *body)
719{
720 int size = body->insns_info.size;
721 int max_pos = body->iseq_size;
722 struct succ_index_table *sd = body->insns_info.succ_index_table;
723 return succ_index_table_invert(max_pos, sd, size);
724}
725#endif
726
727void
728rb_iseq_init_trace(rb_iseq_t *iseq)
729{
730 iseq->aux.exec.global_trace_events = 0;
731 if (ruby_vm_event_enabled_global_flags & ISEQ_TRACE_EVENTS) {
732 rb_iseq_trace_set(iseq, ruby_vm_event_enabled_global_flags & ISEQ_TRACE_EVENTS);
733 }
734}
735
736static VALUE
737finish_iseq_build(rb_iseq_t *iseq)
738{
739 struct iseq_compile_data *data = ISEQ_COMPILE_DATA(iseq);
740 const struct rb_iseq_constant_body *const body = ISEQ_BODY(iseq);
741 VALUE err = data->err_info;
742 ISEQ_COMPILE_DATA_CLEAR(iseq);
743 compile_data_free(data);
744
745#if VM_CHECK_MODE > 0 && VM_INSN_INFO_TABLE_IMPL > 0
746 validate_get_insn_info(iseq);
747#endif
748
749 if (RTEST(err)) {
750 VALUE path = pathobj_path(body->location.pathobj);
751 if (err == Qtrue) err = rb_exc_new_cstr(rb_eSyntaxError, "compile error");
752 rb_funcallv(err, rb_intern("set_backtrace"), 1, &path);
753 rb_exc_raise(err);
754 }
755
756 RB_DEBUG_COUNTER_INC(iseq_num);
757 RB_DEBUG_COUNTER_ADD(iseq_cd_num, ISEQ_BODY(iseq)->ci_size);
758
759 rb_iseq_init_trace(iseq);
760 return Qtrue;
761}
762
763static rb_compile_option_t COMPILE_OPTION_DEFAULT = {
764 .inline_const_cache = OPT_INLINE_CONST_CACHE,
765 .peephole_optimization = OPT_PEEPHOLE_OPTIMIZATION,
766 .tailcall_optimization = OPT_TAILCALL_OPTIMIZATION,
767 .specialized_instruction = OPT_SPECIALISED_INSTRUCTION,
768 .operands_unification = OPT_OPERANDS_UNIFICATION,
769 .instructions_unification = OPT_INSTRUCTIONS_UNIFICATION,
770 .frozen_string_literal = OPT_FROZEN_STRING_LITERAL,
771 .debug_frozen_string_literal = OPT_DEBUG_FROZEN_STRING_LITERAL,
772 .coverage_enabled = TRUE,
773};
774
775static const rb_compile_option_t COMPILE_OPTION_FALSE = {
776 .frozen_string_literal = -1, // unspecified
777};
778
779int
780rb_iseq_opt_frozen_string_literal(void)
781{
782 return COMPILE_OPTION_DEFAULT.frozen_string_literal;
783}
784
785static void
786set_compile_option_from_hash(rb_compile_option_t *option, VALUE opt)
787{
788#define SET_COMPILE_OPTION(o, h, mem) \
789 { VALUE flag = rb_hash_aref((h), ID2SYM(rb_intern(#mem))); \
790 if (flag == Qtrue) { (o)->mem = 1; } \
791 else if (flag == Qfalse) { (o)->mem = 0; } \
792 }
793#define SET_COMPILE_OPTION_NUM(o, h, mem) \
794 { VALUE num = rb_hash_aref((h), ID2SYM(rb_intern(#mem))); \
795 if (!NIL_P(num)) (o)->mem = NUM2INT(num); \
796 }
797 SET_COMPILE_OPTION(option, opt, inline_const_cache);
798 SET_COMPILE_OPTION(option, opt, peephole_optimization);
799 SET_COMPILE_OPTION(option, opt, tailcall_optimization);
800 SET_COMPILE_OPTION(option, opt, specialized_instruction);
801 SET_COMPILE_OPTION(option, opt, operands_unification);
802 SET_COMPILE_OPTION(option, opt, instructions_unification);
803 SET_COMPILE_OPTION(option, opt, frozen_string_literal);
804 SET_COMPILE_OPTION(option, opt, debug_frozen_string_literal);
805 SET_COMPILE_OPTION(option, opt, coverage_enabled);
806 SET_COMPILE_OPTION_NUM(option, opt, debug_level);
807#undef SET_COMPILE_OPTION
808#undef SET_COMPILE_OPTION_NUM
809}
810
811static rb_compile_option_t *
812set_compile_option_from_ast(rb_compile_option_t *option, const rb_ast_body_t *ast)
813{
814#define SET_COMPILE_OPTION(o, a, mem) \
815 ((a)->mem < 0 ? 0 : ((o)->mem = (a)->mem > 0))
816 SET_COMPILE_OPTION(option, ast, coverage_enabled);
817#undef SET_COMPILE_OPTION
818 if (ast->frozen_string_literal >= 0) {
819 option->frozen_string_literal = ast->frozen_string_literal;
820 }
821 return option;
822}
823
824static void
825make_compile_option(rb_compile_option_t *option, VALUE opt)
826{
827 if (NIL_P(opt)) {
828 *option = COMPILE_OPTION_DEFAULT;
829 }
830 else if (opt == Qfalse) {
831 *option = COMPILE_OPTION_FALSE;
832 }
833 else if (opt == Qtrue) {
834 int i;
835 for (i = 0; i < (int)(sizeof(rb_compile_option_t) / sizeof(int)); ++i)
836 ((int *)option)[i] = 1;
837 }
838 else if (RB_TYPE_P(opt, T_HASH)) {
839 *option = COMPILE_OPTION_DEFAULT;
840 set_compile_option_from_hash(option, opt);
841 }
842 else {
843 rb_raise(rb_eTypeError, "Compile option must be Hash/true/false/nil");
844 }
845}
846
847static VALUE
848make_compile_option_value(rb_compile_option_t *option)
849{
850 VALUE opt = rb_hash_new_with_size(11);
851#define SET_COMPILE_OPTION(o, h, mem) \
852 rb_hash_aset((h), ID2SYM(rb_intern(#mem)), RBOOL((o)->mem))
853#define SET_COMPILE_OPTION_NUM(o, h, mem) \
854 rb_hash_aset((h), ID2SYM(rb_intern(#mem)), INT2NUM((o)->mem))
855 {
856 SET_COMPILE_OPTION(option, opt, inline_const_cache);
857 SET_COMPILE_OPTION(option, opt, peephole_optimization);
858 SET_COMPILE_OPTION(option, opt, tailcall_optimization);
859 SET_COMPILE_OPTION(option, opt, specialized_instruction);
860 SET_COMPILE_OPTION(option, opt, operands_unification);
861 SET_COMPILE_OPTION(option, opt, instructions_unification);
862 SET_COMPILE_OPTION(option, opt, debug_frozen_string_literal);
863 SET_COMPILE_OPTION(option, opt, coverage_enabled);
864 SET_COMPILE_OPTION_NUM(option, opt, debug_level);
865 }
866#undef SET_COMPILE_OPTION
867#undef SET_COMPILE_OPTION_NUM
868 VALUE frozen_string_literal = option->frozen_string_literal == -1 ? Qnil : RBOOL(option->frozen_string_literal);
869 rb_hash_aset(opt, ID2SYM(rb_intern("frozen_string_literal")), frozen_string_literal);
870 return opt;
871}
872
873rb_iseq_t *
874rb_iseq_new(const VALUE ast_value, VALUE name, VALUE path, VALUE realpath,
875 const rb_iseq_t *parent, enum rb_iseq_type type)
876{
877 return rb_iseq_new_with_opt(ast_value, name, path, realpath, 0, parent,
878 0, type, &COMPILE_OPTION_DEFAULT,
879 Qnil);
880}
881
882static int
883ast_line_count(const VALUE ast_value)
884{
885 rb_ast_t *ast = rb_ruby_ast_data_get(ast_value);
886 return ast->body.line_count;
887}
888
889static VALUE
890iseq_setup_coverage(VALUE coverages, VALUE path, int line_count)
891{
892 if (line_count >= 0) {
893 int len = (rb_get_coverage_mode() & COVERAGE_TARGET_ONESHOT_LINES) ? 0 : line_count;
894
895 VALUE coverage = rb_default_coverage(len);
896 rb_hash_aset(coverages, path, coverage);
897
898 return coverage;
899 }
900
901 return Qnil;
902}
903
904static inline void
905iseq_new_setup_coverage(VALUE path, int line_count)
906{
907 VALUE coverages = rb_get_coverages();
908
909 if (RTEST(coverages)) {
910 iseq_setup_coverage(coverages, path, line_count);
911 }
912}
913
914rb_iseq_t *
915rb_iseq_new_top(const VALUE ast_value, VALUE name, VALUE path, VALUE realpath, const rb_iseq_t *parent)
916{
917 iseq_new_setup_coverage(path, ast_line_count(ast_value));
918
919 return rb_iseq_new_with_opt(ast_value, name, path, realpath, 0, parent, 0,
920 ISEQ_TYPE_TOP, &COMPILE_OPTION_DEFAULT,
921 Qnil);
922}
923
927rb_iseq_t *
928pm_iseq_new_top(pm_scope_node_t *node, VALUE name, VALUE path, VALUE realpath, const rb_iseq_t *parent, int *error_state)
929{
930 iseq_new_setup_coverage(path, (int) (node->parser->newline_list.size - 1));
931
932 return pm_iseq_new_with_opt(node, name, path, realpath, 0, parent, 0,
933 ISEQ_TYPE_TOP, &COMPILE_OPTION_DEFAULT, error_state);
934}
935
936rb_iseq_t *
937rb_iseq_new_main(const VALUE ast_value, VALUE path, VALUE realpath, const rb_iseq_t *parent, int opt)
938{
939 iseq_new_setup_coverage(path, ast_line_count(ast_value));
940
941 return rb_iseq_new_with_opt(ast_value, rb_fstring_lit("<main>"),
942 path, realpath, 0,
943 parent, 0, ISEQ_TYPE_MAIN, opt ? &COMPILE_OPTION_DEFAULT : &COMPILE_OPTION_FALSE,
944 Qnil);
945}
946
951rb_iseq_t *
952pm_iseq_new_main(pm_scope_node_t *node, VALUE path, VALUE realpath, const rb_iseq_t *parent, int opt, int *error_state)
953{
954 iseq_new_setup_coverage(path, (int) (node->parser->newline_list.size - 1));
955
956 return pm_iseq_new_with_opt(node, rb_fstring_lit("<main>"),
957 path, realpath, 0,
958 parent, 0, ISEQ_TYPE_MAIN, opt ? &COMPILE_OPTION_DEFAULT : &COMPILE_OPTION_FALSE, error_state);
959}
960
961rb_iseq_t *
962rb_iseq_new_eval(const VALUE ast_value, VALUE name, VALUE path, VALUE realpath, int first_lineno, const rb_iseq_t *parent, int isolated_depth)
963{
964 if (rb_get_coverage_mode() & COVERAGE_TARGET_EVAL) {
965 VALUE coverages = rb_get_coverages();
966 if (RTEST(coverages) && RTEST(path) && !RTEST(rb_hash_has_key(coverages, path))) {
967 iseq_setup_coverage(coverages, path, ast_line_count(ast_value) + first_lineno - 1);
968 }
969 }
970
971 return rb_iseq_new_with_opt(ast_value, name, path, realpath, first_lineno,
972 parent, isolated_depth, ISEQ_TYPE_EVAL, &COMPILE_OPTION_DEFAULT,
973 Qnil);
974}
975
976rb_iseq_t *
977pm_iseq_new_eval(pm_scope_node_t *node, VALUE name, VALUE path, VALUE realpath,
978 int first_lineno, const rb_iseq_t *parent, int isolated_depth, int *error_state)
979{
980 if (rb_get_coverage_mode() & COVERAGE_TARGET_EVAL) {
981 VALUE coverages = rb_get_coverages();
982 if (RTEST(coverages) && RTEST(path) && !RTEST(rb_hash_has_key(coverages, path))) {
983 iseq_setup_coverage(coverages, path, ((int) (node->parser->newline_list.size - 1)) + first_lineno - 1);
984 }
985 }
986
987 return pm_iseq_new_with_opt(node, name, path, realpath, first_lineno,
988 parent, isolated_depth, ISEQ_TYPE_EVAL, &COMPILE_OPTION_DEFAULT, error_state);
989}
990
991static inline rb_iseq_t *
992iseq_translate(rb_iseq_t *iseq)
993{
994 if (rb_respond_to(rb_cISeq, rb_intern("translate"))) {
995 VALUE v1 = iseqw_new(iseq);
996 VALUE v2 = rb_funcall(rb_cISeq, rb_intern("translate"), 1, v1);
997 if (v1 != v2 && CLASS_OF(v2) == rb_cISeq) {
998 iseq = (rb_iseq_t *)iseqw_check(v2);
999 }
1000 }
1001
1002 return iseq;
1003}
1004
1005rb_iseq_t *
1006rb_iseq_new_with_opt(VALUE ast_value, VALUE name, VALUE path, VALUE realpath,
1007 int first_lineno, const rb_iseq_t *parent, int isolated_depth,
1008 enum rb_iseq_type type, const rb_compile_option_t *option,
1009 VALUE script_lines)
1010{
1011 rb_ast_t *ast = rb_ruby_ast_data_get(ast_value);
1012 rb_ast_body_t *body = ast ? &ast->body : NULL;
1013 const NODE *node = body ? body->root : 0;
1014 /* TODO: argument check */
1015 rb_iseq_t *iseq = iseq_alloc();
1016 rb_compile_option_t new_opt;
1017
1018 if (!option) option = &COMPILE_OPTION_DEFAULT;
1019 if (body) {
1020 new_opt = *option;
1021 option = set_compile_option_from_ast(&new_opt, body);
1022 }
1023
1024 if (!NIL_P(script_lines)) {
1025 // noop
1026 }
1027 else if (body && body->script_lines) {
1028 script_lines = rb_parser_build_script_lines_from(body->script_lines);
1029 }
1030 else if (parent) {
1031 script_lines = ISEQ_BODY(parent)->variable.script_lines;
1032 }
1033
1034 prepare_iseq_build(iseq, name, path, realpath, first_lineno, node ? &node->nd_loc : NULL, node ? nd_node_id(node) : -1,
1035 parent, isolated_depth, type, script_lines, option);
1036
1037 rb_iseq_compile_node(iseq, node);
1038 finish_iseq_build(iseq);
1039 RB_GC_GUARD(ast_value);
1040
1041 return iseq_translate(iseq);
1042}
1043
1045 rb_iseq_t *iseq;
1046 pm_scope_node_t *node;
1047};
1048
1049VALUE
1050pm_iseq_new_with_opt_try(VALUE d)
1051{
1052 struct pm_iseq_new_with_opt_data *data = (struct pm_iseq_new_with_opt_data *)d;
1053
1054 // This can compile child iseqs, which can raise syntax errors
1055 pm_iseq_compile_node(data->iseq, data->node);
1056
1057 // This raises an exception if there is a syntax error
1058 finish_iseq_build(data->iseq);
1059
1060 return Qundef;
1061}
1062
1075rb_iseq_t *
1076pm_iseq_new_with_opt(pm_scope_node_t *node, VALUE name, VALUE path, VALUE realpath,
1077 int first_lineno, const rb_iseq_t *parent, int isolated_depth,
1078 enum rb_iseq_type type, const rb_compile_option_t *option, int *error_state)
1079{
1080 rb_iseq_t *iseq = iseq_alloc();
1081 ISEQ_BODY(iseq)->prism = true;
1082
1083 rb_compile_option_t next_option;
1084 if (!option) option = &COMPILE_OPTION_DEFAULT;
1085
1086 next_option = *option;
1087 next_option.coverage_enabled = node->coverage_enabled < 0 ? 0 : node->coverage_enabled > 0;
1088 option = &next_option;
1089
1090 pm_location_t *location = &node->base.location;
1091 int32_t start_line = node->parser->start_line;
1092
1093 pm_line_column_t start = pm_newline_list_line_column(&node->parser->newline_list, location->start, start_line);
1094 pm_line_column_t end = pm_newline_list_line_column(&node->parser->newline_list, location->end, start_line);
1095
1096 rb_code_location_t code_location = (rb_code_location_t) {
1097 .beg_pos = { .lineno = (int) start.line, .column = (int) start.column },
1098 .end_pos = { .lineno = (int) end.line, .column = (int) end.column }
1099 };
1100
1101 prepare_iseq_build(iseq, name, path, realpath, first_lineno, &code_location, node->ast_node->node_id,
1102 parent, isolated_depth, type, node->script_lines == NULL ? Qnil : *node->script_lines, option);
1103
1104 struct pm_iseq_new_with_opt_data data = {
1105 .iseq = iseq,
1106 .node = node
1107 };
1108 rb_protect(pm_iseq_new_with_opt_try, (VALUE)&data, error_state);
1109
1110 if (*error_state) return NULL;
1111
1112 return iseq_translate(iseq);
1113}
1114
1115rb_iseq_t *
1116rb_iseq_new_with_callback(
1117 const struct rb_iseq_new_with_callback_callback_func * ifunc,
1118 VALUE name, VALUE path, VALUE realpath,
1119 int first_lineno, const rb_iseq_t *parent,
1120 enum rb_iseq_type type, const rb_compile_option_t *option)
1121{
1122 /* TODO: argument check */
1123 rb_iseq_t *iseq = iseq_alloc();
1124
1125 if (!option) option = &COMPILE_OPTION_DEFAULT;
1126 prepare_iseq_build(iseq, name, path, realpath, first_lineno, NULL, -1, parent, 0, type, Qnil, option);
1127
1128 rb_iseq_compile_callback(iseq, ifunc);
1129 finish_iseq_build(iseq);
1130
1131 return iseq;
1132}
1133
1134const rb_iseq_t *
1135rb_iseq_load_iseq(VALUE fname)
1136{
1137 VALUE iseqv = rb_check_funcall(rb_cISeq, rb_intern("load_iseq"), 1, &fname);
1138
1139 if (!SPECIAL_CONST_P(iseqv) && RBASIC_CLASS(iseqv) == rb_cISeq) {
1140 return iseqw_check(iseqv);
1141 }
1142
1143 return NULL;
1144}
1145
1146#define CHECK_ARRAY(v) rb_to_array_type(v)
1147#define CHECK_HASH(v) rb_to_hash_type(v)
1148#define CHECK_STRING(v) rb_str_to_str(v)
1149#define CHECK_SYMBOL(v) rb_to_symbol_type(v)
1150static inline VALUE CHECK_INTEGER(VALUE v) {(void)NUM2LONG(v); return v;}
1151
1152static enum rb_iseq_type
1153iseq_type_from_sym(VALUE type)
1154{
1155 const ID id_top = rb_intern("top");
1156 const ID id_method = rb_intern("method");
1157 const ID id_block = rb_intern("block");
1158 const ID id_class = rb_intern("class");
1159 const ID id_rescue = rb_intern("rescue");
1160 const ID id_ensure = rb_intern("ensure");
1161 const ID id_eval = rb_intern("eval");
1162 const ID id_main = rb_intern("main");
1163 const ID id_plain = rb_intern("plain");
1164 /* ensure all symbols are static or pinned down before
1165 * conversion */
1166 const ID typeid = rb_check_id(&type);
1167 if (typeid == id_top) return ISEQ_TYPE_TOP;
1168 if (typeid == id_method) return ISEQ_TYPE_METHOD;
1169 if (typeid == id_block) return ISEQ_TYPE_BLOCK;
1170 if (typeid == id_class) return ISEQ_TYPE_CLASS;
1171 if (typeid == id_rescue) return ISEQ_TYPE_RESCUE;
1172 if (typeid == id_ensure) return ISEQ_TYPE_ENSURE;
1173 if (typeid == id_eval) return ISEQ_TYPE_EVAL;
1174 if (typeid == id_main) return ISEQ_TYPE_MAIN;
1175 if (typeid == id_plain) return ISEQ_TYPE_PLAIN;
1176 return (enum rb_iseq_type)-1;
1177}
1178
1179static VALUE
1180iseq_load(VALUE data, const rb_iseq_t *parent, VALUE opt)
1181{
1182 rb_iseq_t *iseq = iseq_alloc();
1183
1184 VALUE magic, version1, version2, format_type, misc;
1185 VALUE name, path, realpath, code_location, node_id;
1186 VALUE type, body, locals, params, exception;
1187
1188 st_data_t iseq_type;
1189 rb_compile_option_t option;
1190 int i = 0;
1191 rb_code_location_t tmp_loc = { {0, 0}, {-1, -1} };
1192
1193 /* [magic, major_version, minor_version, format_type, misc,
1194 * label, path, first_lineno,
1195 * type, locals, args, exception_table, body]
1196 */
1197
1198 data = CHECK_ARRAY(data);
1199
1200 magic = CHECK_STRING(rb_ary_entry(data, i++));
1201 version1 = CHECK_INTEGER(rb_ary_entry(data, i++));
1202 version2 = CHECK_INTEGER(rb_ary_entry(data, i++));
1203 format_type = CHECK_INTEGER(rb_ary_entry(data, i++));
1204 misc = CHECK_HASH(rb_ary_entry(data, i++));
1205 ((void)magic, (void)version1, (void)version2, (void)format_type);
1206
1207 name = CHECK_STRING(rb_ary_entry(data, i++));
1208 path = CHECK_STRING(rb_ary_entry(data, i++));
1209 realpath = rb_ary_entry(data, i++);
1210 realpath = NIL_P(realpath) ? Qnil : CHECK_STRING(realpath);
1211 int first_lineno = RB_NUM2INT(rb_ary_entry(data, i++));
1212
1213 type = CHECK_SYMBOL(rb_ary_entry(data, i++));
1214 locals = CHECK_ARRAY(rb_ary_entry(data, i++));
1215 params = CHECK_HASH(rb_ary_entry(data, i++));
1216 exception = CHECK_ARRAY(rb_ary_entry(data, i++));
1217 body = CHECK_ARRAY(rb_ary_entry(data, i++));
1218
1219 ISEQ_BODY(iseq)->local_iseq = iseq;
1220
1221 iseq_type = iseq_type_from_sym(type);
1222 if (iseq_type == (enum rb_iseq_type)-1) {
1223 rb_raise(rb_eTypeError, "unsupported type: :%"PRIsVALUE, rb_sym2str(type));
1224 }
1225
1226 node_id = rb_hash_aref(misc, ID2SYM(rb_intern("node_id")));
1227
1228 code_location = rb_hash_aref(misc, ID2SYM(rb_intern("code_location")));
1229 if (RB_TYPE_P(code_location, T_ARRAY) && RARRAY_LEN(code_location) == 4) {
1230 tmp_loc.beg_pos.lineno = NUM2INT(rb_ary_entry(code_location, 0));
1231 tmp_loc.beg_pos.column = NUM2INT(rb_ary_entry(code_location, 1));
1232 tmp_loc.end_pos.lineno = NUM2INT(rb_ary_entry(code_location, 2));
1233 tmp_loc.end_pos.column = NUM2INT(rb_ary_entry(code_location, 3));
1234 }
1235
1236 if (SYM2ID(rb_hash_aref(misc, ID2SYM(rb_intern("parser")))) == rb_intern("prism")) {
1237 ISEQ_BODY(iseq)->prism = true;
1238 }
1239
1240 make_compile_option(&option, opt);
1241 option.peephole_optimization = FALSE; /* because peephole optimization can modify original iseq */
1242 prepare_iseq_build(iseq, name, path, realpath, first_lineno, &tmp_loc, NUM2INT(node_id),
1243 parent, 0, (enum rb_iseq_type)iseq_type, Qnil, &option);
1244
1245 rb_iseq_build_from_ary(iseq, misc, locals, params, exception, body);
1246
1247 finish_iseq_build(iseq);
1248
1249 return iseqw_new(iseq);
1250}
1251
1252/*
1253 * :nodoc:
1254 */
1255static VALUE
1256iseq_s_load(int argc, VALUE *argv, VALUE self)
1257{
1258 VALUE data, opt=Qnil;
1259 rb_scan_args(argc, argv, "11", &data, &opt);
1260 return iseq_load(data, NULL, opt);
1261}
1262
1263VALUE
1264rb_iseq_load(VALUE data, VALUE parent, VALUE opt)
1265{
1266 return iseq_load(data, RTEST(parent) ? (rb_iseq_t *)parent : NULL, opt);
1267}
1268
1269static rb_iseq_t *
1270rb_iseq_compile_with_option(VALUE src, VALUE file, VALUE realpath, VALUE line, VALUE opt)
1271{
1272 rb_iseq_t *iseq = NULL;
1273 rb_compile_option_t option;
1274#if !defined(__GNUC__) || (__GNUC__ == 4 && __GNUC_MINOR__ == 8)
1275# define INITIALIZED volatile /* suppress warnings by gcc 4.8 */
1276#else
1277# define INITIALIZED /* volatile */
1278#endif
1279 VALUE (*parse)(VALUE vparser, VALUE fname, VALUE file, int start);
1280 int ln;
1281 VALUE INITIALIZED ast_value;
1282 rb_ast_t *ast;
1283 VALUE name = rb_fstring_lit("<compiled>");
1284
1285 /* safe results first */
1286 make_compile_option(&option, opt);
1287 ln = NUM2INT(line);
1288 StringValueCStr(file);
1289 if (RB_TYPE_P(src, T_FILE)) {
1290 parse = rb_parser_compile_file_path;
1291 }
1292 else {
1293 parse = rb_parser_compile_string_path;
1294 StringValue(src);
1295 }
1296 {
1297 const VALUE parser = rb_parser_new();
1298 const rb_iseq_t *outer_scope = rb_iseq_new(Qnil, name, name, Qnil, 0, ISEQ_TYPE_TOP);
1299 VALUE outer_scope_v = (VALUE)outer_scope;
1300 rb_parser_set_context(parser, outer_scope, FALSE);
1301 if (ruby_vm_keep_script_lines) rb_parser_set_script_lines(parser);
1302 RB_GC_GUARD(outer_scope_v);
1303 ast_value = (*parse)(parser, file, src, ln);
1304 }
1305
1306 ast = rb_ruby_ast_data_get(ast_value);
1307
1308 if (!ast || !ast->body.root) {
1309 rb_ast_dispose(ast);
1310 rb_exc_raise(GET_EC()->errinfo);
1311 }
1312 else {
1313 iseq = rb_iseq_new_with_opt(ast_value, name, file, realpath, ln,
1314 NULL, 0, ISEQ_TYPE_TOP, &option,
1315 Qnil);
1316 rb_ast_dispose(ast);
1317 }
1318
1319 return iseq;
1320}
1321
1322static rb_iseq_t *
1323pm_iseq_compile_with_option(VALUE src, VALUE file, VALUE realpath, VALUE line, VALUE opt)
1324{
1325 rb_iseq_t *iseq = NULL;
1326 rb_compile_option_t option;
1327 int ln;
1328 VALUE name = rb_fstring_lit("<compiled>");
1329
1330 /* safe results first */
1331 make_compile_option(&option, opt);
1332 ln = NUM2INT(line);
1333 StringValueCStr(file);
1334
1335 bool parse_file = false;
1336 if (RB_TYPE_P(src, T_FILE)) {
1337 parse_file = true;
1338 src = rb_io_path(src);
1339 }
1340 else {
1341 src = StringValue(src);
1342 }
1343
1344 pm_parse_result_t result = { 0 };
1345 pm_options_line_set(&result.options, NUM2INT(line));
1346 pm_options_scopes_init(&result.options, 1);
1347 result.node.coverage_enabled = 1;
1348
1349 switch (option.frozen_string_literal) {
1350 case ISEQ_FROZEN_STRING_LITERAL_UNSET:
1351 break;
1352 case ISEQ_FROZEN_STRING_LITERAL_DISABLED:
1353 pm_options_frozen_string_literal_set(&result.options, false);
1354 break;
1355 case ISEQ_FROZEN_STRING_LITERAL_ENABLED:
1356 pm_options_frozen_string_literal_set(&result.options, true);
1357 break;
1358 default:
1359 rb_bug("pm_iseq_compile_with_option: invalid frozen_string_literal=%d", option.frozen_string_literal);
1360 break;
1361 }
1362
1363 VALUE script_lines;
1364 VALUE error;
1365
1366 if (parse_file) {
1367 error = pm_load_parse_file(&result, src, ruby_vm_keep_script_lines ? &script_lines : NULL);
1368 }
1369 else {
1370 error = pm_parse_string(&result, src, file, ruby_vm_keep_script_lines ? &script_lines : NULL);
1371 }
1372
1373 RB_GC_GUARD(src);
1374
1375 if (error == Qnil) {
1376 int error_state;
1377 iseq = pm_iseq_new_with_opt(&result.node, name, file, realpath, ln, NULL, 0, ISEQ_TYPE_TOP, &option, &error_state);
1378
1379 pm_parse_result_free(&result);
1380
1381 if (error_state) {
1382 RUBY_ASSERT(iseq == NULL);
1383 rb_jump_tag(error_state);
1384 }
1385 }
1386 else {
1387 pm_parse_result_free(&result);
1388 rb_exc_raise(error);
1389 }
1390
1391 return iseq;
1392}
1393
1394VALUE
1395rb_iseq_path(const rb_iseq_t *iseq)
1396{
1397 return pathobj_path(ISEQ_BODY(iseq)->location.pathobj);
1398}
1399
1400VALUE
1401rb_iseq_realpath(const rb_iseq_t *iseq)
1402{
1403 return pathobj_realpath(ISEQ_BODY(iseq)->location.pathobj);
1404}
1405
1406VALUE
1407rb_iseq_absolute_path(const rb_iseq_t *iseq)
1408{
1409 return rb_iseq_realpath(iseq);
1410}
1411
1412int
1413rb_iseq_from_eval_p(const rb_iseq_t *iseq)
1414{
1415 return NIL_P(rb_iseq_realpath(iseq));
1416}
1417
1418VALUE
1419rb_iseq_label(const rb_iseq_t *iseq)
1420{
1421 return ISEQ_BODY(iseq)->location.label;
1422}
1423
1424VALUE
1425rb_iseq_base_label(const rb_iseq_t *iseq)
1426{
1427 return ISEQ_BODY(iseq)->location.base_label;
1428}
1429
1430VALUE
1431rb_iseq_first_lineno(const rb_iseq_t *iseq)
1432{
1433 return RB_INT2NUM(ISEQ_BODY(iseq)->location.first_lineno);
1434}
1435
1436VALUE
1437rb_iseq_method_name(const rb_iseq_t *iseq)
1438{
1439 struct rb_iseq_constant_body *const body = ISEQ_BODY(ISEQ_BODY(iseq)->local_iseq);
1440
1441 if (body->type == ISEQ_TYPE_METHOD) {
1442 return body->location.base_label;
1443 }
1444 else {
1445 return Qnil;
1446 }
1447}
1448
1449void
1450rb_iseq_code_location(const rb_iseq_t *iseq, int *beg_pos_lineno, int *beg_pos_column, int *end_pos_lineno, int *end_pos_column)
1451{
1452 const rb_code_location_t *loc = &ISEQ_BODY(iseq)->location.code_location;
1453 if (beg_pos_lineno) *beg_pos_lineno = loc->beg_pos.lineno;
1454 if (beg_pos_column) *beg_pos_column = loc->beg_pos.column;
1455 if (end_pos_lineno) *end_pos_lineno = loc->end_pos.lineno;
1456 if (end_pos_column) *end_pos_column = loc->end_pos.column;
1457}
1458
1459static ID iseq_type_id(enum rb_iseq_type type);
1460
1461VALUE
1462rb_iseq_type(const rb_iseq_t *iseq)
1463{
1464 return ID2SYM(iseq_type_id(ISEQ_BODY(iseq)->type));
1465}
1466
1467VALUE
1468rb_iseq_coverage(const rb_iseq_t *iseq)
1469{
1470 return ISEQ_COVERAGE(iseq);
1471}
1472
1473static int
1474remove_coverage_i(void *vstart, void *vend, size_t stride, void *data)
1475{
1476 VALUE v = (VALUE)vstart;
1477 for (; v != (VALUE)vend; v += stride) {
1478 void *ptr = rb_asan_poisoned_object_p(v);
1479 rb_asan_unpoison_object(v, false);
1480
1481 if (rb_obj_is_iseq(v)) {
1482 rb_iseq_t *iseq = (rb_iseq_t *)v;
1483 ISEQ_COVERAGE_SET(iseq, Qnil);
1484 }
1485
1486 asan_poison_object_if(ptr, v);
1487 }
1488 return 0;
1489}
1490
1491void
1492rb_iseq_remove_coverage_all(void)
1493{
1494 rb_objspace_each_objects(remove_coverage_i, NULL);
1495}
1496
1497/* define wrapper class methods (RubyVM::InstructionSequence) */
1498
1499static void
1500iseqw_mark(void *ptr)
1501{
1502 rb_gc_mark_movable(*(VALUE *)ptr);
1503}
1504
1505static size_t
1506iseqw_memsize(const void *ptr)
1507{
1508 return rb_iseq_memsize(*(const rb_iseq_t **)ptr);
1509}
1510
1511static void
1512iseqw_ref_update(void *ptr)
1513{
1514 VALUE *vptr = ptr;
1515 *vptr = rb_gc_location(*vptr);
1516}
1517
1518static const rb_data_type_t iseqw_data_type = {
1519 "T_IMEMO/iseq",
1520 {
1521 iseqw_mark,
1523 iseqw_memsize,
1524 iseqw_ref_update,
1525 },
1526 0, 0, RUBY_TYPED_FREE_IMMEDIATELY|RUBY_TYPED_WB_PROTECTED
1527};
1528
1529static VALUE
1530iseqw_new(const rb_iseq_t *iseq)
1531{
1532 if (iseq->wrapper) {
1533 if (*(const rb_iseq_t **)rb_check_typeddata(iseq->wrapper, &iseqw_data_type) != iseq) {
1534 rb_raise(rb_eTypeError, "wrong iseq wrapper: %" PRIsVALUE " for %p",
1535 iseq->wrapper, (void *)iseq);
1536 }
1537 return iseq->wrapper;
1538 }
1539 else {
1540 rb_iseq_t **ptr;
1541 VALUE obj = TypedData_Make_Struct(rb_cISeq, rb_iseq_t *, &iseqw_data_type, ptr);
1542 RB_OBJ_WRITE(obj, ptr, iseq);
1543
1544 /* cache a wrapper object */
1545 RB_OBJ_WRITE((VALUE)iseq, &iseq->wrapper, obj);
1546
1547 return obj;
1548 }
1549}
1550
1551VALUE
1552rb_iseqw_new(const rb_iseq_t *iseq)
1553{
1554 return iseqw_new(iseq);
1555}
1556
1562static VALUE
1563iseqw_s_compile_parser(int argc, VALUE *argv, VALUE self, bool prism)
1564{
1565 VALUE src, file = Qnil, path = Qnil, line = Qnil, opt = Qnil;
1566 int i;
1567
1568 i = rb_scan_args(argc, argv, "1*:", &src, NULL, &opt);
1569 if (i > 4+NIL_P(opt)) rb_error_arity(argc, 1, 5);
1570 switch (i) {
1571 case 5: opt = argv[--i];
1572 case 4: line = argv[--i];
1573 case 3: path = argv[--i];
1574 case 2: file = argv[--i];
1575 }
1576
1577 if (NIL_P(file)) file = rb_fstring_lit("<compiled>");
1578 if (NIL_P(path)) path = file;
1579 if (NIL_P(line)) line = INT2FIX(1);
1580
1581 Check_Type(path, T_STRING);
1582 Check_Type(file, T_STRING);
1583
1584 rb_iseq_t *iseq;
1585 if (prism) {
1586 iseq = pm_iseq_compile_with_option(src, file, path, line, opt);
1587 }
1588 else {
1589 iseq = rb_iseq_compile_with_option(src, file, path, line, opt);
1590 }
1591
1592 return iseqw_new(iseq);
1593}
1594
1595/*
1596 * call-seq:
1597 * InstructionSequence.compile(source[, file[, path[, line[, options]]]]) -> iseq
1598 * InstructionSequence.new(source[, file[, path[, line[, options]]]]) -> iseq
1599 *
1600 * Takes +source+, which can be a string of Ruby code, or an open +File+ object.
1601 * that contains Ruby source code.
1602 *
1603 * Optionally takes +file+, +path+, and +line+ which describe the file path,
1604 * real path and first line number of the ruby code in +source+ which are
1605 * metadata attached to the returned +iseq+.
1606 *
1607 * +file+ is used for `__FILE__` and exception backtrace. +path+ is used for
1608 * +require_relative+ base. It is recommended these should be the same full
1609 * path.
1610 *
1611 * +options+, which can be +true+, +false+ or a +Hash+, is used to
1612 * modify the default behavior of the Ruby iseq compiler.
1613 *
1614 * For details regarding valid compile options see ::compile_option=.
1615 *
1616 * RubyVM::InstructionSequence.compile("a = 1 + 2")
1617 * #=> <RubyVM::InstructionSequence:<compiled>@<compiled>>
1618 *
1619 * path = "test.rb"
1620 * RubyVM::InstructionSequence.compile(File.read(path), path, File.expand_path(path))
1621 * #=> <RubyVM::InstructionSequence:<compiled>@test.rb:1>
1622 *
1623 * file = File.open("test.rb")
1624 * RubyVM::InstructionSequence.compile(file)
1625 * #=> <RubyVM::InstructionSequence:<compiled>@<compiled>:1>
1626 *
1627 * path = File.expand_path("test.rb")
1628 * RubyVM::InstructionSequence.compile(File.read(path), path, path)
1629 * #=> <RubyVM::InstructionSequence:<compiled>@/absolute/path/to/test.rb:1>
1630 *
1631 */
1632static VALUE
1633iseqw_s_compile(int argc, VALUE *argv, VALUE self)
1634{
1635 return iseqw_s_compile_parser(argc, argv, self, rb_ruby_prism_p());
1636}
1637
1638/*
1639 * call-seq:
1640 * InstructionSequence.compile_parsey(source[, file[, path[, line[, options]]]]) -> iseq
1641 *
1642 * Takes +source+, which can be a string of Ruby code, or an open +File+ object.
1643 * that contains Ruby source code. It parses and compiles using parse.y.
1644 *
1645 * Optionally takes +file+, +path+, and +line+ which describe the file path,
1646 * real path and first line number of the ruby code in +source+ which are
1647 * metadata attached to the returned +iseq+.
1648 *
1649 * +file+ is used for `__FILE__` and exception backtrace. +path+ is used for
1650 * +require_relative+ base. It is recommended these should be the same full
1651 * path.
1652 *
1653 * +options+, which can be +true+, +false+ or a +Hash+, is used to
1654 * modify the default behavior of the Ruby iseq compiler.
1655 *
1656 * For details regarding valid compile options see ::compile_option=.
1657 *
1658 * RubyVM::InstructionSequence.compile_parsey("a = 1 + 2")
1659 * #=> <RubyVM::InstructionSequence:<compiled>@<compiled>>
1660 *
1661 * path = "test.rb"
1662 * RubyVM::InstructionSequence.compile_parsey(File.read(path), path, File.expand_path(path))
1663 * #=> <RubyVM::InstructionSequence:<compiled>@test.rb:1>
1664 *
1665 * file = File.open("test.rb")
1666 * RubyVM::InstructionSequence.compile_parsey(file)
1667 * #=> <RubyVM::InstructionSequence:<compiled>@<compiled>:1>
1668 *
1669 * path = File.expand_path("test.rb")
1670 * RubyVM::InstructionSequence.compile_parsey(File.read(path), path, path)
1671 * #=> <RubyVM::InstructionSequence:<compiled>@/absolute/path/to/test.rb:1>
1672 *
1673 */
1674static VALUE
1675iseqw_s_compile_parsey(int argc, VALUE *argv, VALUE self)
1676{
1677 return iseqw_s_compile_parser(argc, argv, self, false);
1678}
1679
1680/*
1681 * call-seq:
1682 * InstructionSequence.compile_prism(source[, file[, path[, line[, options]]]]) -> iseq
1683 *
1684 * Takes +source+, which can be a string of Ruby code, or an open +File+ object.
1685 * that contains Ruby source code. It parses and compiles using prism.
1686 *
1687 * Optionally takes +file+, +path+, and +line+ which describe the file path,
1688 * real path and first line number of the ruby code in +source+ which are
1689 * metadata attached to the returned +iseq+.
1690 *
1691 * +file+ is used for `__FILE__` and exception backtrace. +path+ is used for
1692 * +require_relative+ base. It is recommended these should be the same full
1693 * path.
1694 *
1695 * +options+, which can be +true+, +false+ or a +Hash+, is used to
1696 * modify the default behavior of the Ruby iseq compiler.
1697 *
1698 * For details regarding valid compile options see ::compile_option=.
1699 *
1700 * RubyVM::InstructionSequence.compile_prism("a = 1 + 2")
1701 * #=> <RubyVM::InstructionSequence:<compiled>@<compiled>>
1702 *
1703 * path = "test.rb"
1704 * RubyVM::InstructionSequence.compile_prism(File.read(path), path, File.expand_path(path))
1705 * #=> <RubyVM::InstructionSequence:<compiled>@test.rb:1>
1706 *
1707 * file = File.open("test.rb")
1708 * RubyVM::InstructionSequence.compile_prism(file)
1709 * #=> <RubyVM::InstructionSequence:<compiled>@<compiled>:1>
1710 *
1711 * path = File.expand_path("test.rb")
1712 * RubyVM::InstructionSequence.compile_prism(File.read(path), path, path)
1713 * #=> <RubyVM::InstructionSequence:<compiled>@/absolute/path/to/test.rb:1>
1714 *
1715 */
1716static VALUE
1717iseqw_s_compile_prism(int argc, VALUE *argv, VALUE self)
1718{
1719 return iseqw_s_compile_parser(argc, argv, self, true);
1720}
1721
1722/*
1723 * call-seq:
1724 * InstructionSequence.compile_file(file[, options]) -> iseq
1725 *
1726 * Takes +file+, a String with the location of a Ruby source file, reads,
1727 * parses and compiles the file, and returns +iseq+, the compiled
1728 * InstructionSequence with source location metadata set.
1729 *
1730 * Optionally takes +options+, which can be +true+, +false+ or a +Hash+, to
1731 * modify the default behavior of the Ruby iseq compiler.
1732 *
1733 * For details regarding valid compile options see ::compile_option=.
1734 *
1735 * # /tmp/hello.rb
1736 * puts "Hello, world!"
1737 *
1738 * # elsewhere
1739 * RubyVM::InstructionSequence.compile_file("/tmp/hello.rb")
1740 * #=> <RubyVM::InstructionSequence:<main>@/tmp/hello.rb>
1741 */
1742static VALUE
1743iseqw_s_compile_file(int argc, VALUE *argv, VALUE self)
1744{
1745 VALUE file, opt = Qnil;
1746 VALUE parser, f, exc = Qnil, ret;
1747 rb_ast_t *ast;
1748 VALUE ast_value;
1749 rb_compile_option_t option;
1750 int i;
1751
1752 i = rb_scan_args(argc, argv, "1*:", &file, NULL, &opt);
1753 if (i > 1+NIL_P(opt)) rb_error_arity(argc, 1, 2);
1754 switch (i) {
1755 case 2: opt = argv[--i];
1756 }
1757 FilePathValue(file);
1758 file = rb_fstring(file); /* rb_io_t->pathv gets frozen anyways */
1759
1760 f = rb_file_open_str(file, "r");
1761
1762 rb_execution_context_t *ec = GET_EC();
1763 VALUE v = rb_vm_push_frame_fname(ec, file);
1764
1765 parser = rb_parser_new();
1766 rb_parser_set_context(parser, NULL, FALSE);
1767 ast_value = rb_parser_load_file(parser, file);
1768 ast = rb_ruby_ast_data_get(ast_value);
1769 if (!ast->body.root) exc = GET_EC()->errinfo;
1770
1771 rb_io_close(f);
1772 if (!ast->body.root) {
1773 rb_ast_dispose(ast);
1774 rb_exc_raise(exc);
1775 }
1776
1777 make_compile_option(&option, opt);
1778
1779 ret = iseqw_new(rb_iseq_new_with_opt(ast_value, rb_fstring_lit("<main>"),
1780 file,
1781 rb_realpath_internal(Qnil, file, 1),
1782 1, NULL, 0, ISEQ_TYPE_TOP, &option,
1783 Qnil));
1784 rb_ast_dispose(ast);
1785 RB_GC_GUARD(ast_value);
1786
1787 rb_vm_pop_frame(ec);
1788 RB_GC_GUARD(v);
1789 return ret;
1790}
1791
1792/*
1793 * call-seq:
1794 * InstructionSequence.compile_file_prism(file[, options]) -> iseq
1795 *
1796 * Takes +file+, a String with the location of a Ruby source file, reads,
1797 * parses and compiles the file, and returns +iseq+, the compiled
1798 * InstructionSequence with source location metadata set. It parses and
1799 * compiles using prism.
1800 *
1801 * Optionally takes +options+, which can be +true+, +false+ or a +Hash+, to
1802 * modify the default behavior of the Ruby iseq compiler.
1803 *
1804 * For details regarding valid compile options see ::compile_option=.
1805 *
1806 * # /tmp/hello.rb
1807 * puts "Hello, world!"
1808 *
1809 * # elsewhere
1810 * RubyVM::InstructionSequence.compile_file_prism("/tmp/hello.rb")
1811 * #=> <RubyVM::InstructionSequence:<main>@/tmp/hello.rb>
1812 */
1813static VALUE
1814iseqw_s_compile_file_prism(int argc, VALUE *argv, VALUE self)
1815{
1816 VALUE file, opt = Qnil, ret;
1817 rb_compile_option_t option;
1818 int i;
1819
1820 i = rb_scan_args(argc, argv, "1*:", &file, NULL, &opt);
1821 if (i > 1+NIL_P(opt)) rb_error_arity(argc, 1, 2);
1822 switch (i) {
1823 case 2: opt = argv[--i];
1824 }
1825 FilePathValue(file);
1826 file = rb_fstring(file); /* rb_io_t->pathv gets frozen anyways */
1827
1828 rb_execution_context_t *ec = GET_EC();
1829 VALUE v = rb_vm_push_frame_fname(ec, file);
1830
1831 pm_parse_result_t result = { 0 };
1832 result.options.line = 1;
1833 result.node.coverage_enabled = 1;
1834
1835 VALUE script_lines;
1836 VALUE error = pm_load_parse_file(&result, file, ruby_vm_keep_script_lines ? &script_lines : NULL);
1837
1838 if (error == Qnil) {
1839 make_compile_option(&option, opt);
1840
1841 int error_state;
1842 rb_iseq_t *iseq = pm_iseq_new_with_opt(&result.node, rb_fstring_lit("<main>"),
1843 file,
1844 rb_realpath_internal(Qnil, file, 1),
1845 1, NULL, 0, ISEQ_TYPE_TOP, &option, &error_state);
1846
1847 pm_parse_result_free(&result);
1848
1849 if (error_state) {
1850 RUBY_ASSERT(iseq == NULL);
1851 rb_jump_tag(error_state);
1852 }
1853
1854 ret = iseqw_new(iseq);
1855 rb_vm_pop_frame(ec);
1856 RB_GC_GUARD(v);
1857 return ret;
1858 }
1859 else {
1860 pm_parse_result_free(&result);
1861 rb_vm_pop_frame(ec);
1862 RB_GC_GUARD(v);
1863 rb_exc_raise(error);
1864 }
1865}
1866
1867/*
1868 * call-seq:
1869 * InstructionSequence.compile_option = options
1870 *
1871 * Sets the default values for various optimizations in the Ruby iseq
1872 * compiler.
1873 *
1874 * Possible values for +options+ include +true+, which enables all options,
1875 * +false+ which disables all options, and +nil+ which leaves all options
1876 * unchanged.
1877 *
1878 * You can also pass a +Hash+ of +options+ that you want to change, any
1879 * options not present in the hash will be left unchanged.
1880 *
1881 * Possible option names (which are keys in +options+) which can be set to
1882 * +true+ or +false+ include:
1883 *
1884 * * +:inline_const_cache+
1885 * * +:instructions_unification+
1886 * * +:operands_unification+
1887 * * +:peephole_optimization+
1888 * * +:specialized_instruction+
1889 * * +:tailcall_optimization+
1890 *
1891 * Additionally, +:debug_level+ can be set to an integer.
1892 *
1893 * These default options can be overwritten for a single run of the iseq
1894 * compiler by passing any of the above values as the +options+ parameter to
1895 * ::new, ::compile and ::compile_file.
1896 */
1897static VALUE
1898iseqw_s_compile_option_set(VALUE self, VALUE opt)
1899{
1900 rb_compile_option_t option;
1901 make_compile_option(&option, opt);
1902 COMPILE_OPTION_DEFAULT = option;
1903 return opt;
1904}
1905
1906/*
1907 * call-seq:
1908 * InstructionSequence.compile_option -> options
1909 *
1910 * Returns a hash of default options used by the Ruby iseq compiler.
1911 *
1912 * For details, see InstructionSequence.compile_option=.
1913 */
1914static VALUE
1915iseqw_s_compile_option_get(VALUE self)
1916{
1917 return make_compile_option_value(&COMPILE_OPTION_DEFAULT);
1918}
1919
1920static const rb_iseq_t *
1921iseqw_check(VALUE iseqw)
1922{
1923 rb_iseq_t **iseq_ptr;
1924 TypedData_Get_Struct(iseqw, rb_iseq_t *, &iseqw_data_type, iseq_ptr);
1925 rb_iseq_t *iseq = *iseq_ptr;
1926
1927 if (!ISEQ_BODY(iseq)) {
1928 rb_ibf_load_iseq_complete(iseq);
1929 }
1930
1931 if (!ISEQ_BODY(iseq)->location.label) {
1932 rb_raise(rb_eTypeError, "uninitialized InstructionSequence");
1933 }
1934 return iseq;
1935}
1936
1937const rb_iseq_t *
1938rb_iseqw_to_iseq(VALUE iseqw)
1939{
1940 return iseqw_check(iseqw);
1941}
1942
1943/*
1944 * call-seq:
1945 * iseq.eval -> obj
1946 *
1947 * Evaluates the instruction sequence and returns the result.
1948 *
1949 * RubyVM::InstructionSequence.compile("1 + 2").eval #=> 3
1950 */
1951static VALUE
1952iseqw_eval(VALUE self)
1953{
1954 const rb_iseq_t *iseq = iseqw_check(self);
1955 if (0 == ISEQ_BODY(iseq)->iseq_size) {
1956 rb_raise(rb_eTypeError, "attempt to evaluate dummy InstructionSequence");
1957 }
1958 return rb_iseq_eval(iseq);
1959}
1960
1961/*
1962 * Returns a human-readable string representation of this instruction
1963 * sequence, including the #label and #path.
1964 */
1965static VALUE
1966iseqw_inspect(VALUE self)
1967{
1968 const rb_iseq_t *iseq = iseqw_check(self);
1969 const struct rb_iseq_constant_body *const body = ISEQ_BODY(iseq);
1970 VALUE klass = rb_class_name(rb_obj_class(self));
1971
1972 if (!body->location.label) {
1973 return rb_sprintf("#<%"PRIsVALUE": uninitialized>", klass);
1974 }
1975 else {
1976 return rb_sprintf("<%"PRIsVALUE":%"PRIsVALUE"@%"PRIsVALUE":%d>",
1977 klass,
1978 body->location.label, rb_iseq_path(iseq),
1979 FIX2INT(rb_iseq_first_lineno(iseq)));
1980 }
1981}
1982
1983/*
1984 * Returns the path of this instruction sequence.
1985 *
1986 * <code><compiled></code> if the iseq was evaluated from a string.
1987 *
1988 * For example, using irb:
1989 *
1990 * iseq = RubyVM::InstructionSequence.compile('num = 1 + 2')
1991 * #=> <RubyVM::InstructionSequence:<compiled>@<compiled>>
1992 * iseq.path
1993 * #=> "<compiled>"
1994 *
1995 * Using ::compile_file:
1996 *
1997 * # /tmp/method.rb
1998 * def hello
1999 * puts "hello, world"
2000 * end
2001 *
2002 * # in irb
2003 * > iseq = RubyVM::InstructionSequence.compile_file('/tmp/method.rb')
2004 * > iseq.path #=> /tmp/method.rb
2005 */
2006static VALUE
2007iseqw_path(VALUE self)
2008{
2009 return rb_iseq_path(iseqw_check(self));
2010}
2011
2012/*
2013 * Returns the absolute path of this instruction sequence.
2014 *
2015 * +nil+ if the iseq was evaluated from a string.
2016 *
2017 * For example, using ::compile_file:
2018 *
2019 * # /tmp/method.rb
2020 * def hello
2021 * puts "hello, world"
2022 * end
2023 *
2024 * # in irb
2025 * > iseq = RubyVM::InstructionSequence.compile_file('/tmp/method.rb')
2026 * > iseq.absolute_path #=> /tmp/method.rb
2027 */
2028static VALUE
2029iseqw_absolute_path(VALUE self)
2030{
2031 return rb_iseq_realpath(iseqw_check(self));
2032}
2033
2034/* Returns the label of this instruction sequence.
2035 *
2036 * <code><main></code> if it's at the top level, <code><compiled></code> if it
2037 * was evaluated from a string.
2038 *
2039 * For example, using irb:
2040 *
2041 * iseq = RubyVM::InstructionSequence.compile('num = 1 + 2')
2042 * #=> <RubyVM::InstructionSequence:<compiled>@<compiled>>
2043 * iseq.label
2044 * #=> "<compiled>"
2045 *
2046 * Using ::compile_file:
2047 *
2048 * # /tmp/method.rb
2049 * def hello
2050 * puts "hello, world"
2051 * end
2052 *
2053 * # in irb
2054 * > iseq = RubyVM::InstructionSequence.compile_file('/tmp/method.rb')
2055 * > iseq.label #=> <main>
2056 */
2057static VALUE
2058iseqw_label(VALUE self)
2059{
2060 return rb_iseq_label(iseqw_check(self));
2061}
2062
2063/* Returns the base label of this instruction sequence.
2064 *
2065 * For example, using irb:
2066 *
2067 * iseq = RubyVM::InstructionSequence.compile('num = 1 + 2')
2068 * #=> <RubyVM::InstructionSequence:<compiled>@<compiled>>
2069 * iseq.base_label
2070 * #=> "<compiled>"
2071 *
2072 * Using ::compile_file:
2073 *
2074 * # /tmp/method.rb
2075 * def hello
2076 * puts "hello, world"
2077 * end
2078 *
2079 * # in irb
2080 * > iseq = RubyVM::InstructionSequence.compile_file('/tmp/method.rb')
2081 * > iseq.base_label #=> <main>
2082 */
2083static VALUE
2084iseqw_base_label(VALUE self)
2085{
2086 return rb_iseq_base_label(iseqw_check(self));
2087}
2088
2089/* Returns the number of the first source line where the instruction sequence
2090 * was loaded from.
2091 *
2092 * For example, using irb:
2093 *
2094 * iseq = RubyVM::InstructionSequence.compile('num = 1 + 2')
2095 * #=> <RubyVM::InstructionSequence:<compiled>@<compiled>>
2096 * iseq.first_lineno
2097 * #=> 1
2098 */
2099static VALUE
2100iseqw_first_lineno(VALUE self)
2101{
2102 return rb_iseq_first_lineno(iseqw_check(self));
2103}
2104
2105static VALUE iseq_data_to_ary(const rb_iseq_t *iseq);
2106
2107/*
2108 * call-seq:
2109 * iseq.to_a -> ary
2110 *
2111 * Returns an Array with 14 elements representing the instruction sequence
2112 * with the following data:
2113 *
2114 * [magic]
2115 * A string identifying the data format. <b>Always
2116 * +YARVInstructionSequence/SimpleDataFormat+.</b>
2117 *
2118 * [major_version]
2119 * The major version of the instruction sequence.
2120 *
2121 * [minor_version]
2122 * The minor version of the instruction sequence.
2123 *
2124 * [format_type]
2125 * A number identifying the data format. <b>Always 1</b>.
2126 *
2127 * [misc]
2128 * A hash containing:
2129 *
2130 * [+:arg_size+]
2131 * the total number of arguments taken by the method or the block (0 if
2132 * _iseq_ doesn't represent a method or block)
2133 * [+:local_size+]
2134 * the number of local variables + 1
2135 * [+:stack_max+]
2136 * used in calculating the stack depth at which a SystemStackError is
2137 * thrown.
2138 *
2139 * [#label]
2140 * The name of the context (block, method, class, module, etc.) that this
2141 * instruction sequence belongs to.
2142 *
2143 * <code><main></code> if it's at the top level, <code><compiled></code> if
2144 * it was evaluated from a string.
2145 *
2146 * [#path]
2147 * The relative path to the Ruby file where the instruction sequence was
2148 * loaded from.
2149 *
2150 * <code><compiled></code> if the iseq was evaluated from a string.
2151 *
2152 * [#absolute_path]
2153 * The absolute path to the Ruby file where the instruction sequence was
2154 * loaded from.
2155 *
2156 * +nil+ if the iseq was evaluated from a string.
2157 *
2158 * [#first_lineno]
2159 * The number of the first source line where the instruction sequence was
2160 * loaded from.
2161 *
2162 * [type]
2163 * The type of the instruction sequence.
2164 *
2165 * Valid values are +:top+, +:method+, +:block+, +:class+, +:rescue+,
2166 * +:ensure+, +:eval+, +:main+, and +plain+.
2167 *
2168 * [locals]
2169 * An array containing the names of all arguments and local variables as
2170 * symbols.
2171 *
2172 * [params]
2173 * An Hash object containing parameter information.
2174 *
2175 * More info about these values can be found in +vm_core.h+.
2176 *
2177 * [catch_table]
2178 * A list of exceptions and control flow operators (rescue, next, redo,
2179 * break, etc.).
2180 *
2181 * [bytecode]
2182 * An array of arrays containing the instruction names and operands that
2183 * make up the body of the instruction sequence.
2184 *
2185 * Note that this format is MRI specific and version dependent.
2186 *
2187 */
2188static VALUE
2189iseqw_to_a(VALUE self)
2190{
2191 const rb_iseq_t *iseq = iseqw_check(self);
2192 return iseq_data_to_ary(iseq);
2193}
2194
2195#if VM_INSN_INFO_TABLE_IMPL == 1 /* binary search */
2196static const struct iseq_insn_info_entry *
2197get_insn_info_binary_search(const rb_iseq_t *iseq, size_t pos)
2198{
2199 const struct rb_iseq_constant_body *const body = ISEQ_BODY(iseq);
2200 size_t size = body->insns_info.size;
2201 const struct iseq_insn_info_entry *insns_info = body->insns_info.body;
2202 const unsigned int *positions = body->insns_info.positions;
2203 const int debug = 0;
2204
2205 if (debug) {
2206 printf("size: %"PRIuSIZE"\n", size);
2207 printf("insns_info[%"PRIuSIZE"]: position: %d, line: %d, pos: %"PRIuSIZE"\n",
2208 (size_t)0, positions[0], insns_info[0].line_no, pos);
2209 }
2210
2211 if (size == 0) {
2212 return NULL;
2213 }
2214 else if (size == 1) {
2215 return &insns_info[0];
2216 }
2217 else {
2218 size_t l = 1, r = size - 1;
2219 while (l <= r) {
2220 size_t m = l + (r - l) / 2;
2221 if (positions[m] == pos) {
2222 return &insns_info[m];
2223 }
2224 if (positions[m] < pos) {
2225 l = m + 1;
2226 }
2227 else {
2228 r = m - 1;
2229 }
2230 }
2231 if (l >= size) {
2232 return &insns_info[size-1];
2233 }
2234 if (positions[l] > pos) {
2235 return &insns_info[l-1];
2236 }
2237 return &insns_info[l];
2238 }
2239}
2240
2241static const struct iseq_insn_info_entry *
2242get_insn_info(const rb_iseq_t *iseq, size_t pos)
2243{
2244 return get_insn_info_binary_search(iseq, pos);
2245}
2246#endif
2247
2248#if VM_INSN_INFO_TABLE_IMPL == 2 /* succinct bitvector */
2249static const struct iseq_insn_info_entry *
2250get_insn_info_succinct_bitvector(const rb_iseq_t *iseq, size_t pos)
2251{
2252 const struct rb_iseq_constant_body *const body = ISEQ_BODY(iseq);
2253 size_t size = body->insns_info.size;
2254 const struct iseq_insn_info_entry *insns_info = body->insns_info.body;
2255 const int debug = 0;
2256
2257 if (debug) {
2258#if VM_CHECK_MODE > 0
2259 const unsigned int *positions = body->insns_info.positions;
2260 printf("size: %"PRIuSIZE"\n", size);
2261 printf("insns_info[%"PRIuSIZE"]: position: %d, line: %d, pos: %"PRIuSIZE"\n",
2262 (size_t)0, positions[0], insns_info[0].line_no, pos);
2263#else
2264 printf("size: %"PRIuSIZE"\n", size);
2265 printf("insns_info[%"PRIuSIZE"]: line: %d, pos: %"PRIuSIZE"\n",
2266 (size_t)0, insns_info[0].line_no, pos);
2267#endif
2268 }
2269
2270 if (size == 0) {
2271 return NULL;
2272 }
2273 else if (size == 1) {
2274 return &insns_info[0];
2275 }
2276 else {
2277 int index;
2278 VM_ASSERT(body->insns_info.succ_index_table != NULL);
2279 index = succ_index_lookup(body->insns_info.succ_index_table, (int)pos);
2280 return &insns_info[index-1];
2281 }
2282}
2283
2284static const struct iseq_insn_info_entry *
2285get_insn_info(const rb_iseq_t *iseq, size_t pos)
2286{
2287 return get_insn_info_succinct_bitvector(iseq, pos);
2288}
2289#endif
2290
2291#if VM_CHECK_MODE > 0 || VM_INSN_INFO_TABLE_IMPL == 0
2292static const struct iseq_insn_info_entry *
2293get_insn_info_linear_search(const rb_iseq_t *iseq, size_t pos)
2294{
2295 const struct rb_iseq_constant_body *const body = ISEQ_BODY(iseq);
2296 size_t i = 0, size = body->insns_info.size;
2297 const struct iseq_insn_info_entry *insns_info = body->insns_info.body;
2298 const unsigned int *positions = body->insns_info.positions;
2299 const int debug = 0;
2300
2301 if (debug) {
2302 printf("size: %"PRIuSIZE"\n", size);
2303 printf("insns_info[%"PRIuSIZE"]: position: %d, line: %d, pos: %"PRIuSIZE"\n",
2304 i, positions[i], insns_info[i].line_no, pos);
2305 }
2306
2307 if (size == 0) {
2308 return NULL;
2309 }
2310 else if (size == 1) {
2311 return &insns_info[0];
2312 }
2313 else {
2314 for (i=1; i<size; i++) {
2315 if (debug) printf("insns_info[%"PRIuSIZE"]: position: %d, line: %d, pos: %"PRIuSIZE"\n",
2316 i, positions[i], insns_info[i].line_no, pos);
2317
2318 if (positions[i] == pos) {
2319 return &insns_info[i];
2320 }
2321 if (positions[i] > pos) {
2322 return &insns_info[i-1];
2323 }
2324 }
2325 }
2326 return &insns_info[i-1];
2327}
2328#endif
2329
2330#if VM_INSN_INFO_TABLE_IMPL == 0 /* linear search */
2331static const struct iseq_insn_info_entry *
2332get_insn_info(const rb_iseq_t *iseq, size_t pos)
2333{
2334 return get_insn_info_linear_search(iseq, pos);
2335}
2336#endif
2337
2338#if VM_CHECK_MODE > 0 && VM_INSN_INFO_TABLE_IMPL > 0
2339static void
2340validate_get_insn_info(const rb_iseq_t *iseq)
2341{
2342 const struct rb_iseq_constant_body *const body = ISEQ_BODY(iseq);
2343 size_t i;
2344 for (i = 0; i < body->iseq_size; i++) {
2345 if (get_insn_info_linear_search(iseq, i) != get_insn_info(iseq, i)) {
2346 rb_bug("validate_get_insn_info: get_insn_info_linear_search(iseq, %"PRIuSIZE") != get_insn_info(iseq, %"PRIuSIZE")", i, i);
2347 }
2348 }
2349}
2350#endif
2351
2352unsigned int
2353rb_iseq_line_no(const rb_iseq_t *iseq, size_t pos)
2354{
2355 const struct iseq_insn_info_entry *entry = get_insn_info(iseq, pos);
2356
2357 if (entry) {
2358 return entry->line_no;
2359 }
2360 else {
2361 return 0;
2362 }
2363}
2364
2365#ifdef USE_ISEQ_NODE_ID
2366int
2367rb_iseq_node_id(const rb_iseq_t *iseq, size_t pos)
2368{
2369 const struct iseq_insn_info_entry *entry = get_insn_info(iseq, pos);
2370
2371 if (entry) {
2372 return entry->node_id;
2373 }
2374 else {
2375 return 0;
2376 }
2377}
2378#endif
2379
2381rb_iseq_event_flags(const rb_iseq_t *iseq, size_t pos)
2382{
2383 const struct iseq_insn_info_entry *entry = get_insn_info(iseq, pos);
2384 if (entry) {
2385 return entry->events;
2386 }
2387 else {
2388 return 0;
2389 }
2390}
2391
2392// Clear tracing event flags and turn off tracing for a given instruction as needed.
2393// This is currently used after updating a one-shot line coverage for the current instruction.
2394void
2395rb_iseq_clear_event_flags(const rb_iseq_t *iseq, size_t pos, rb_event_flag_t reset)
2396{
2397 struct iseq_insn_info_entry *entry = (struct iseq_insn_info_entry *)get_insn_info(iseq, pos);
2398 if (entry) {
2399 entry->events &= ~reset;
2400 if (!(entry->events & iseq->aux.exec.global_trace_events)) {
2401 void rb_iseq_trace_flag_cleared(const rb_iseq_t *iseq, size_t pos);
2402 rb_iseq_trace_flag_cleared(iseq, pos);
2403 }
2404 }
2405}
2406
2407static VALUE
2408local_var_name(const rb_iseq_t *diseq, VALUE level, VALUE op)
2409{
2410 VALUE i;
2411 VALUE name;
2412 ID lid;
2413 int idx;
2414
2415 for (i = 0; i < level; i++) {
2416 diseq = ISEQ_BODY(diseq)->parent_iseq;
2417 }
2418 idx = ISEQ_BODY(diseq)->local_table_size - (int)op - 1;
2419 lid = ISEQ_BODY(diseq)->local_table[idx];
2420 name = rb_id2str(lid);
2421 if (!name) {
2422 name = rb_str_new_cstr("?");
2423 }
2424 else if (!rb_is_local_id(lid)) {
2425 name = rb_str_inspect(name);
2426 }
2427 else {
2428 name = rb_str_dup(name);
2429 }
2430 rb_str_catf(name, "@%d", idx);
2431 return name;
2432}
2433
2434int rb_insn_unified_local_var_level(VALUE);
2435VALUE rb_dump_literal(VALUE lit);
2436
2437VALUE
2438rb_insn_operand_intern(const rb_iseq_t *iseq,
2439 VALUE insn, int op_no, VALUE op,
2440 int len, size_t pos, const VALUE *pnop, VALUE child)
2441{
2442 const char *types = insn_op_types(insn);
2443 char type = types[op_no];
2444 VALUE ret = Qundef;
2445
2446 switch (type) {
2447 case TS_OFFSET: /* LONG */
2448 ret = rb_sprintf("%"PRIdVALUE, (VALUE)(pos + len + op));
2449 break;
2450
2451 case TS_NUM: /* ULONG */
2452 if (insn == BIN(defined) && op_no == 0) {
2453 enum defined_type deftype = (enum defined_type)op;
2454 switch (deftype) {
2455 case DEFINED_FUNC:
2456 ret = rb_fstring_lit("func");
2457 break;
2458 case DEFINED_REF:
2459 ret = rb_fstring_lit("ref");
2460 break;
2461 case DEFINED_CONST_FROM:
2462 ret = rb_fstring_lit("constant-from");
2463 break;
2464 default:
2465 ret = rb_iseq_defined_string(deftype);
2466 break;
2467 }
2468 if (ret) break;
2469 }
2470 else if (insn == BIN(checktype) && op_no == 0) {
2471 const char *type_str = rb_type_str((enum ruby_value_type)op);
2472 if (type_str) {
2473 ret = rb_str_new_cstr(type_str); break;
2474 }
2475 }
2476 ret = rb_sprintf("%"PRIuVALUE, op);
2477 break;
2478
2479 case TS_LINDEX:{
2480 int level;
2481 if (types[op_no+1] == TS_NUM && pnop) {
2482 ret = local_var_name(iseq, *pnop, op - VM_ENV_DATA_SIZE);
2483 }
2484 else if ((level = rb_insn_unified_local_var_level(insn)) >= 0) {
2485 ret = local_var_name(iseq, (VALUE)level, op - VM_ENV_DATA_SIZE);
2486 }
2487 else {
2488 ret = rb_inspect(INT2FIX(op));
2489 }
2490 break;
2491 }
2492 case TS_ID: /* ID (symbol) */
2493 ret = rb_inspect(ID2SYM(op));
2494 break;
2495
2496 case TS_VALUE: /* VALUE */
2497 op = obj_resurrect(op);
2498 if (insn == BIN(defined) && op_no == 1 && FIXNUM_P(op)) {
2499 /* should be DEFINED_REF */
2500 int type = NUM2INT(op);
2501 if (type) {
2502 if (type & 1) {
2503 ret = rb_sprintf(":$%c", (type >> 1));
2504 }
2505 else {
2506 ret = rb_sprintf(":$%d", (type >> 1));
2507 }
2508 break;
2509 }
2510 }
2511 ret = rb_dump_literal(op);
2512 if (CLASS_OF(op) == rb_cISeq) {
2513 if (child) {
2514 rb_ary_push(child, op);
2515 }
2516 }
2517 break;
2518
2519 case TS_ISEQ: /* iseq */
2520 {
2521 if (op) {
2522 const rb_iseq_t *iseq = rb_iseq_check((rb_iseq_t *)op);
2523 ret = ISEQ_BODY(iseq)->location.label;
2524 if (child) {
2525 rb_ary_push(child, (VALUE)iseq);
2526 }
2527 }
2528 else {
2529 ret = rb_str_new2("nil");
2530 }
2531 break;
2532 }
2533
2534 case TS_IC:
2535 {
2536 ret = rb_sprintf("<ic:%"PRIdPTRDIFF" ", (union iseq_inline_storage_entry *)op - ISEQ_BODY(iseq)->is_entries);
2537 const ID *segments = ((IC)op)->segments;
2538 rb_str_cat2(ret, rb_id2name(*segments++));
2539 while (*segments) {
2540 rb_str_catf(ret, "::%s", rb_id2name(*segments++));
2541 }
2542 rb_str_cat2(ret, ">");
2543 }
2544 break;
2545 case TS_IVC:
2546 case TS_ICVARC:
2547 case TS_ISE:
2548 ret = rb_sprintf("<is:%"PRIdPTRDIFF">", (union iseq_inline_storage_entry *)op - ISEQ_BODY(iseq)->is_entries);
2549 break;
2550
2551 case TS_CALLDATA:
2552 {
2553 struct rb_call_data *cd = (struct rb_call_data *)op;
2554 const struct rb_callinfo *ci = cd->ci;
2555 VALUE ary = rb_ary_new();
2556 ID mid = vm_ci_mid(ci);
2557
2558 if (mid) {
2559 rb_ary_push(ary, rb_sprintf("mid:%"PRIsVALUE, rb_id2str(mid)));
2560 }
2561
2562 rb_ary_push(ary, rb_sprintf("argc:%d", vm_ci_argc(ci)));
2563
2564 if (vm_ci_flag(ci) & VM_CALL_KWARG) {
2565 const struct rb_callinfo_kwarg *kw_args = vm_ci_kwarg(ci);
2566 VALUE kw_ary = rb_ary_new_from_values(kw_args->keyword_len, kw_args->keywords);
2567 rb_ary_push(ary, rb_sprintf("kw:[%"PRIsVALUE"]", rb_ary_join(kw_ary, rb_str_new2(","))));
2568 }
2569
2570 if (vm_ci_flag(ci)) {
2571 VALUE flags = rb_ary_new();
2572# define CALL_FLAG(n) if (vm_ci_flag(ci) & VM_CALL_##n) rb_ary_push(flags, rb_str_new2(#n))
2573 CALL_FLAG(ARGS_SPLAT);
2574 CALL_FLAG(ARGS_SPLAT_MUT);
2575 CALL_FLAG(ARGS_BLOCKARG);
2576 CALL_FLAG(FCALL);
2577 CALL_FLAG(VCALL);
2578 CALL_FLAG(ARGS_SIMPLE);
2579 CALL_FLAG(TAILCALL);
2580 CALL_FLAG(SUPER);
2581 CALL_FLAG(ZSUPER);
2582 CALL_FLAG(KWARG);
2583 CALL_FLAG(KW_SPLAT);
2584 CALL_FLAG(KW_SPLAT_MUT);
2585 CALL_FLAG(FORWARDING);
2586 CALL_FLAG(OPT_SEND); /* maybe not reachable */
2587 rb_ary_push(ary, rb_ary_join(flags, rb_str_new2("|")));
2588 }
2589
2590 ret = rb_sprintf("<calldata!%"PRIsVALUE">", rb_ary_join(ary, rb_str_new2(", ")));
2591 }
2592 break;
2593
2594 case TS_CDHASH:
2595 ret = rb_str_new2("<cdhash>");
2596 break;
2597
2598 case TS_FUNCPTR:
2599 {
2600#ifdef HAVE_DLADDR
2601 Dl_info info;
2602 if (dladdr((void *)op, &info) && info.dli_sname) {
2603 ret = rb_str_new_cstr(info.dli_sname);
2604 break;
2605 }
2606#endif
2607 ret = rb_str_new2("<funcptr>");
2608 }
2609 break;
2610
2611 case TS_BUILTIN:
2612 {
2613 const struct rb_builtin_function *bf = (const struct rb_builtin_function *)op;
2614 ret = rb_sprintf("<builtin!%s/%d>",
2615 bf->name, bf->argc);
2616 }
2617 break;
2618
2619 default:
2620 rb_bug("unknown operand type: %c", type);
2621 }
2622 return ret;
2623}
2624
2625static VALUE
2626right_strip(VALUE str)
2627{
2628 const char *beg = RSTRING_PTR(str), *end = RSTRING_END(str);
2629 while (end-- > beg && *end == ' ');
2630 rb_str_set_len(str, end - beg + 1);
2631 return str;
2632}
2633
2638int
2639rb_iseq_disasm_insn(VALUE ret, const VALUE *code, size_t pos,
2640 const rb_iseq_t *iseq, VALUE child)
2641{
2642 VALUE insn = code[pos];
2643 int len = insn_len(insn);
2644 int j;
2645 const char *types = insn_op_types(insn);
2646 VALUE str = rb_str_new(0, 0);
2647 const char *insn_name_buff;
2648
2649 insn_name_buff = insn_name(insn);
2650 if (1) {
2651 extern const int rb_vm_max_insn_name_size;
2652 rb_str_catf(str, "%04"PRIuSIZE" %-*s ", pos, rb_vm_max_insn_name_size, insn_name_buff);
2653 }
2654 else {
2655 rb_str_catf(str, "%04"PRIuSIZE" %-28.*s ", pos,
2656 (int)strcspn(insn_name_buff, "_"), insn_name_buff);
2657 }
2658
2659 for (j = 0; types[j]; j++) {
2660 VALUE opstr = rb_insn_operand_intern(iseq, insn, j, code[pos + j + 1],
2661 len, pos, &code[pos + j + 2],
2662 child);
2663 rb_str_concat(str, opstr);
2664
2665 if (types[j + 1]) {
2666 rb_str_cat2(str, ", ");
2667 }
2668 }
2669
2670 {
2671 unsigned int line_no = rb_iseq_line_no(iseq, pos);
2672 unsigned int prev = pos == 0 ? 0 : rb_iseq_line_no(iseq, pos - 1);
2673 if (line_no && line_no != prev) {
2674 long slen = RSTRING_LEN(str);
2675 slen = (slen > 70) ? 0 : (70 - slen);
2676 str = rb_str_catf(str, "%*s(%4d)", (int)slen, "", line_no);
2677 }
2678 }
2679
2680 {
2681 rb_event_flag_t events = rb_iseq_event_flags(iseq, pos);
2682 if (events) {
2683 str = rb_str_catf(str, "[%s%s%s%s%s%s%s%s%s%s%s%s]",
2684 events & RUBY_EVENT_LINE ? "Li" : "",
2685 events & RUBY_EVENT_CLASS ? "Cl" : "",
2686 events & RUBY_EVENT_END ? "En" : "",
2687 events & RUBY_EVENT_CALL ? "Ca" : "",
2688 events & RUBY_EVENT_RETURN ? "Re" : "",
2689 events & RUBY_EVENT_C_CALL ? "Cc" : "",
2690 events & RUBY_EVENT_C_RETURN ? "Cr" : "",
2691 events & RUBY_EVENT_B_CALL ? "Bc" : "",
2692 events & RUBY_EVENT_B_RETURN ? "Br" : "",
2693 events & RUBY_EVENT_RESCUE ? "Rs" : "",
2694 events & RUBY_EVENT_COVERAGE_LINE ? "Cli" : "",
2695 events & RUBY_EVENT_COVERAGE_BRANCH ? "Cbr" : "");
2696 }
2697 }
2698
2699 right_strip(str);
2700 if (ret) {
2701 rb_str_cat2(str, "\n");
2702 rb_str_concat(ret, str);
2703 }
2704 else {
2705 printf("%.*s\n", (int)RSTRING_LEN(str), RSTRING_PTR(str));
2706 }
2707 return len;
2708}
2709
2710static const char *
2711catch_type(int type)
2712{
2713 switch (type) {
2714 case CATCH_TYPE_RESCUE:
2715 return "rescue";
2716 case CATCH_TYPE_ENSURE:
2717 return "ensure";
2718 case CATCH_TYPE_RETRY:
2719 return "retry";
2720 case CATCH_TYPE_BREAK:
2721 return "break";
2722 case CATCH_TYPE_REDO:
2723 return "redo";
2724 case CATCH_TYPE_NEXT:
2725 return "next";
2726 default:
2727 rb_bug("unknown catch type: %d", type);
2728 return 0;
2729 }
2730}
2731
2732static VALUE
2733iseq_inspect(const rb_iseq_t *iseq)
2734{
2735 const struct rb_iseq_constant_body *const body = ISEQ_BODY(iseq);
2736 if (!body->location.label) {
2737 return rb_sprintf("#<ISeq: uninitialized>");
2738 }
2739 else {
2740 const rb_code_location_t *loc = &body->location.code_location;
2741 return rb_sprintf("#<ISeq:%"PRIsVALUE"@%"PRIsVALUE":%d (%d,%d)-(%d,%d)>",
2742 body->location.label, rb_iseq_path(iseq),
2743 loc->beg_pos.lineno,
2744 loc->beg_pos.lineno,
2745 loc->beg_pos.column,
2746 loc->end_pos.lineno,
2747 loc->end_pos.column);
2748 }
2749}
2750
2751static const rb_data_type_t tmp_set = {
2752 "tmpset",
2753 {(void (*)(void *))rb_mark_set, (void (*)(void *))st_free_table, 0, 0,},
2754 0, 0, RUBY_TYPED_FREE_IMMEDIATELY
2755};
2756
2757static VALUE
2758rb_iseq_disasm_recursive(const rb_iseq_t *iseq, VALUE indent)
2759{
2760 const struct rb_iseq_constant_body *const body = ISEQ_BODY(iseq);
2761 VALUE *code;
2762 VALUE str = rb_str_new(0, 0);
2763 VALUE child = rb_ary_hidden_new(3);
2764 unsigned int size;
2765 unsigned int i;
2766 long l;
2767 size_t n;
2768 enum {header_minlen = 72};
2769 st_table *done_iseq = 0;
2770 VALUE done_iseq_wrapper = Qnil;
2771 const char *indent_str;
2772 long indent_len;
2773
2774 size = body->iseq_size;
2775
2776 indent_len = RSTRING_LEN(indent);
2777 indent_str = RSTRING_PTR(indent);
2778
2779 rb_str_cat(str, indent_str, indent_len);
2780 rb_str_cat2(str, "== disasm: ");
2781
2782 rb_str_append(str, iseq_inspect(iseq));
2783 if ((l = RSTRING_LEN(str) - indent_len) < header_minlen) {
2784 rb_str_modify_expand(str, header_minlen - l);
2785 memset(RSTRING_END(str), '=', header_minlen - l);
2786 }
2787 if (iseq->body->builtin_attrs) {
2788#define disasm_builtin_attr(str, iseq, attr) \
2789 if (iseq->body->builtin_attrs & BUILTIN_ATTR_ ## attr) { \
2790 rb_str_cat2(str, " " #attr); \
2791 }
2792 disasm_builtin_attr(str, iseq, LEAF);
2793 disasm_builtin_attr(str, iseq, SINGLE_NOARG_LEAF);
2794 disasm_builtin_attr(str, iseq, INLINE_BLOCK);
2795 disasm_builtin_attr(str, iseq, C_TRACE);
2796 }
2797 rb_str_cat2(str, "\n");
2798
2799 /* show catch table information */
2800 if (body->catch_table) {
2801 rb_str_cat(str, indent_str, indent_len);
2802 rb_str_cat2(str, "== catch table\n");
2803 }
2804 if (body->catch_table) {
2805 rb_str_cat_cstr(indent, "| ");
2806 indent_str = RSTRING_PTR(indent);
2807 for (i = 0; i < body->catch_table->size; i++) {
2808 const struct iseq_catch_table_entry *entry =
2809 UNALIGNED_MEMBER_PTR(body->catch_table, entries[i]);
2810 rb_str_cat(str, indent_str, indent_len);
2811 rb_str_catf(str,
2812 "| catch type: %-6s st: %04d ed: %04d sp: %04d cont: %04d\n",
2813 catch_type((int)entry->type), (int)entry->start,
2814 (int)entry->end, (int)entry->sp, (int)entry->cont);
2815 if (entry->iseq && !(done_iseq && st_is_member(done_iseq, (st_data_t)entry->iseq))) {
2816 rb_str_concat(str, rb_iseq_disasm_recursive(rb_iseq_check(entry->iseq), indent));
2817 if (!done_iseq) {
2818 done_iseq = st_init_numtable();
2819 done_iseq_wrapper = TypedData_Wrap_Struct(0, &tmp_set, done_iseq);
2820 }
2821 st_insert(done_iseq, (st_data_t)entry->iseq, (st_data_t)0);
2822 indent_str = RSTRING_PTR(indent);
2823 }
2824 }
2825 rb_str_resize(indent, indent_len);
2826 indent_str = RSTRING_PTR(indent);
2827 }
2828 if (body->catch_table) {
2829 rb_str_cat(str, indent_str, indent_len);
2830 rb_str_cat2(str, "|-------------------------------------"
2831 "-----------------------------------\n");
2832 }
2833
2834 /* show local table information */
2835 if (body->local_table) {
2836 const struct rb_iseq_param_keyword *const keyword = body->param.keyword;
2837 rb_str_cat(str, indent_str, indent_len);
2838 rb_str_catf(str,
2839 "local table (size: %d, argc: %d "
2840 "[opts: %d, rest: %d, post: %d, block: %d, kw: %d@%d, kwrest: %d])\n",
2841 body->local_table_size,
2842 body->param.lead_num,
2843 body->param.opt_num,
2844 body->param.flags.has_rest ? body->param.rest_start : -1,
2845 body->param.post_num,
2846 body->param.flags.has_block ? body->param.block_start : -1,
2847 body->param.flags.has_kw ? keyword->num : -1,
2848 body->param.flags.has_kw ? keyword->required_num : -1,
2849 body->param.flags.has_kwrest ? keyword->rest_start : -1);
2850
2851 for (i = body->local_table_size; i > 0;) {
2852 int li = body->local_table_size - --i - 1;
2853 long width;
2854 VALUE name = local_var_name(iseq, 0, i);
2855 char argi[0x100];
2856 char opti[0x100];
2857
2858 opti[0] = '\0';
2859 if (body->param.flags.has_opt) {
2860 int argc = body->param.lead_num;
2861 int opts = body->param.opt_num;
2862 if (li >= argc && li < argc + opts) {
2863 snprintf(opti, sizeof(opti), "Opt=%"PRIdVALUE,
2864 body->param.opt_table[li - argc]);
2865 }
2866 }
2867
2868 snprintf(argi, sizeof(argi), "%s%s%s%s%s%s", /* arg, opts, rest, post, kwrest, block */
2869 (body->param.lead_num > li) ? (body->param.flags.ambiguous_param0 ? "AmbiguousArg" : "Arg") : "",
2870 opti,
2871 (body->param.flags.has_rest && body->param.rest_start == li) ? (body->param.flags.anon_rest ? "AnonRest" : "Rest") : "",
2872 (body->param.flags.has_post && body->param.post_start <= li && li < body->param.post_start + body->param.post_num) ? "Post" : "",
2873 (body->param.flags.has_kwrest && keyword->rest_start == li) ? (body->param.flags.anon_kwrest ? "AnonKwrest" : "Kwrest") : "",
2874 (body->param.flags.has_block && body->param.block_start == li) ? "Block" : "");
2875
2876 rb_str_cat(str, indent_str, indent_len);
2877 rb_str_catf(str, "[%2d] ", i + 1);
2878 width = RSTRING_LEN(str) + 11;
2879 rb_str_append(str, name);
2880 if (*argi) rb_str_catf(str, "<%s>", argi);
2881 if ((width -= RSTRING_LEN(str)) > 0) rb_str_catf(str, "%*s", (int)width, "");
2882 }
2883 rb_str_cat_cstr(right_strip(str), "\n");
2884 }
2885
2886 /* show each line */
2887 code = rb_iseq_original_iseq(iseq);
2888 for (n = 0; n < size;) {
2889 rb_str_cat(str, indent_str, indent_len);
2890 n += rb_iseq_disasm_insn(str, code, n, iseq, child);
2891 }
2892
2893 for (l = 0; l < RARRAY_LEN(child); l++) {
2894 VALUE isv = rb_ary_entry(child, l);
2895 if (done_iseq && st_is_member(done_iseq, (st_data_t)isv)) continue;
2896 rb_str_cat_cstr(str, "\n");
2897 rb_str_concat(str, rb_iseq_disasm_recursive(rb_iseq_check((rb_iseq_t *)isv), indent));
2898 indent_str = RSTRING_PTR(indent);
2899 }
2900 RB_GC_GUARD(done_iseq_wrapper);
2901
2902 return str;
2903}
2904
2905VALUE
2906rb_iseq_disasm(const rb_iseq_t *iseq)
2907{
2908 VALUE str = rb_iseq_disasm_recursive(iseq, rb_str_new(0, 0));
2909 rb_str_resize(str, RSTRING_LEN(str));
2910 return str;
2911}
2912
2913/*
2914 * Estimates the number of instance variables that will be set on
2915 * a given `class` with the initialize method defined in
2916 * `initialize_iseq`
2917 */
2918attr_index_t
2919rb_estimate_iv_count(VALUE klass, const rb_iseq_t * initialize_iseq)
2920{
2921 struct rb_id_table * iv_names = rb_id_table_create(0);
2922
2923 for (unsigned int i = 0; i < ISEQ_BODY(initialize_iseq)->ivc_size; i++) {
2924 IVC cache = (IVC)&ISEQ_BODY(initialize_iseq)->is_entries[i];
2925
2926 if (cache->iv_set_name) {
2927 rb_id_table_insert(iv_names, cache->iv_set_name, Qtrue);
2928 }
2929 }
2930
2931 attr_index_t count = (attr_index_t)rb_id_table_size(iv_names);
2932
2933 VALUE superclass = rb_class_superclass(klass);
2934 count += RCLASS_MAX_IV_COUNT(superclass);
2935
2936 rb_id_table_free(iv_names);
2937
2938 return count;
2939}
2940
2941/*
2942 * call-seq:
2943 * iseq.disasm -> str
2944 * iseq.disassemble -> str
2945 *
2946 * Returns the instruction sequence as a +String+ in human readable form.
2947 *
2948 * puts RubyVM::InstructionSequence.compile('1 + 2').disasm
2949 *
2950 * Produces:
2951 *
2952 * == disasm: <RubyVM::InstructionSequence:<compiled>@<compiled>>==========
2953 * 0000 trace 1 ( 1)
2954 * 0002 putobject 1
2955 * 0004 putobject 2
2956 * 0006 opt_plus <ic:1>
2957 * 0008 leave
2958 */
2959static VALUE
2960iseqw_disasm(VALUE self)
2961{
2962 return rb_iseq_disasm(iseqw_check(self));
2963}
2964
2965static int
2966iseq_iterate_children(const rb_iseq_t *iseq, void (*iter_func)(const rb_iseq_t *child_iseq, void *data), void *data)
2967{
2968 unsigned int i;
2969 VALUE *code = rb_iseq_original_iseq(iseq);
2970 const struct rb_iseq_constant_body *const body = ISEQ_BODY(iseq);
2971 const rb_iseq_t *child;
2972 VALUE all_children = rb_obj_hide(rb_ident_hash_new());
2973
2974 if (body->catch_table) {
2975 for (i = 0; i < body->catch_table->size; i++) {
2976 const struct iseq_catch_table_entry *entry =
2977 UNALIGNED_MEMBER_PTR(body->catch_table, entries[i]);
2978 child = entry->iseq;
2979 if (child) {
2980 if (NIL_P(rb_hash_aref(all_children, (VALUE)child))) {
2981 rb_hash_aset(all_children, (VALUE)child, Qtrue);
2982 (*iter_func)(child, data);
2983 }
2984 }
2985 }
2986 }
2987
2988 for (i=0; i<body->iseq_size;) {
2989 VALUE insn = code[i];
2990 int len = insn_len(insn);
2991 const char *types = insn_op_types(insn);
2992 int j;
2993
2994 for (j=0; types[j]; j++) {
2995 switch (types[j]) {
2996 case TS_ISEQ:
2997 child = (const rb_iseq_t *)code[i+j+1];
2998 if (child) {
2999 if (NIL_P(rb_hash_aref(all_children, (VALUE)child))) {
3000 rb_hash_aset(all_children, (VALUE)child, Qtrue);
3001 (*iter_func)(child, data);
3002 }
3003 }
3004 break;
3005 default:
3006 break;
3007 }
3008 }
3009 i += len;
3010 }
3011
3012 return (int)RHASH_SIZE(all_children);
3013}
3014
3015static void
3016yield_each_children(const rb_iseq_t *child_iseq, void *data)
3017{
3018 rb_yield(iseqw_new(child_iseq));
3019}
3020
3021/*
3022 * call-seq:
3023 * iseq.each_child{|child_iseq| ...} -> iseq
3024 *
3025 * Iterate all direct child instruction sequences.
3026 * Iteration order is implementation/version defined
3027 * so that people should not rely on the order.
3028 */
3029static VALUE
3030iseqw_each_child(VALUE self)
3031{
3032 const rb_iseq_t *iseq = iseqw_check(self);
3033 iseq_iterate_children(iseq, yield_each_children, NULL);
3034 return self;
3035}
3036
3037static void
3038push_event_info(const rb_iseq_t *iseq, rb_event_flag_t events, int line, VALUE ary)
3039{
3040#define C(ev, cstr, l) if (events & ev) rb_ary_push(ary, rb_ary_new_from_args(2, l, ID2SYM(rb_intern(cstr))));
3041 C(RUBY_EVENT_CLASS, "class", rb_iseq_first_lineno(iseq));
3042 C(RUBY_EVENT_CALL, "call", rb_iseq_first_lineno(iseq));
3043 C(RUBY_EVENT_B_CALL, "b_call", rb_iseq_first_lineno(iseq));
3044 C(RUBY_EVENT_LINE, "line", INT2FIX(line));
3045 C(RUBY_EVENT_END, "end", INT2FIX(line));
3046 C(RUBY_EVENT_RETURN, "return", INT2FIX(line));
3047 C(RUBY_EVENT_B_RETURN, "b_return", INT2FIX(line));
3048 C(RUBY_EVENT_RESCUE, "rescue", INT2FIX(line));
3049#undef C
3050}
3051
3052/*
3053 * call-seq:
3054 * iseq.trace_points -> ary
3055 *
3056 * Return trace points in the instruction sequence.
3057 * Return an array of [line, event_symbol] pair.
3058 */
3059static VALUE
3060iseqw_trace_points(VALUE self)
3061{
3062 const rb_iseq_t *iseq = iseqw_check(self);
3063 const struct rb_iseq_constant_body *const body = ISEQ_BODY(iseq);
3064 unsigned int i;
3065 VALUE ary = rb_ary_new();
3066
3067 for (i=0; i<body->insns_info.size; i++) {
3068 const struct iseq_insn_info_entry *entry = &body->insns_info.body[i];
3069 if (entry->events) {
3070 push_event_info(iseq, entry->events, entry->line_no, ary);
3071 }
3072 }
3073 return ary;
3074}
3075
3076/*
3077 * Returns the instruction sequence containing the given proc or method.
3078 *
3079 * For example, using irb:
3080 *
3081 * # a proc
3082 * > p = proc { num = 1 + 2 }
3083 * > RubyVM::InstructionSequence.of(p)
3084 * > #=> <RubyVM::InstructionSequence:block in irb_binding@(irb)>
3085 *
3086 * # for a method
3087 * > def foo(bar); puts bar; end
3088 * > RubyVM::InstructionSequence.of(method(:foo))
3089 * > #=> <RubyVM::InstructionSequence:foo@(irb)>
3090 *
3091 * Using ::compile_file:
3092 *
3093 * # /tmp/iseq_of.rb
3094 * def hello
3095 * puts "hello, world"
3096 * end
3097 *
3098 * $a_global_proc = proc { str = 'a' + 'b' }
3099 *
3100 * # in irb
3101 * > require '/tmp/iseq_of.rb'
3102 *
3103 * # first the method hello
3104 * > RubyVM::InstructionSequence.of(method(:hello))
3105 * > #=> #<RubyVM::InstructionSequence:0x007fb73d7cb1d0>
3106 *
3107 * # then the global proc
3108 * > RubyVM::InstructionSequence.of($a_global_proc)
3109 * > #=> #<RubyVM::InstructionSequence:0x007fb73d7caf78>
3110 */
3111static VALUE
3112iseqw_s_of(VALUE klass, VALUE body)
3113{
3114 const rb_iseq_t *iseq = NULL;
3115
3116 if (rb_frame_info_p(body)) {
3117 iseq = rb_get_iseq_from_frame_info(body);
3118 }
3119 else if (rb_obj_is_proc(body)) {
3120 iseq = vm_proc_iseq(body);
3121
3122 if (!rb_obj_is_iseq((VALUE)iseq)) {
3123 iseq = NULL;
3124 }
3125 }
3126 else if (rb_obj_is_method(body)) {
3127 iseq = rb_method_iseq(body);
3128 }
3129 else if (rb_typeddata_is_instance_of(body, &iseqw_data_type)) {
3130 return body;
3131 }
3132
3133 return iseq ? iseqw_new(iseq) : Qnil;
3134}
3135
3136/*
3137 * call-seq:
3138 * InstructionSequence.disasm(body) -> str
3139 * InstructionSequence.disassemble(body) -> str
3140 *
3141 * Takes +body+, a +Method+ or +Proc+ object, and returns a +String+
3142 * with the human readable instructions for +body+.
3143 *
3144 * For a +Method+ object:
3145 *
3146 * # /tmp/method.rb
3147 * def hello
3148 * puts "hello, world"
3149 * end
3150 *
3151 * puts RubyVM::InstructionSequence.disasm(method(:hello))
3152 *
3153 * Produces:
3154 *
3155 * == disasm: <RubyVM::InstructionSequence:hello@/tmp/method.rb>============
3156 * 0000 trace 8 ( 1)
3157 * 0002 trace 1 ( 2)
3158 * 0004 putself
3159 * 0005 putstring "hello, world"
3160 * 0007 send :puts, 1, nil, 8, <ic:0>
3161 * 0013 trace 16 ( 3)
3162 * 0015 leave ( 2)
3163 *
3164 * For a +Proc+ object:
3165 *
3166 * # /tmp/proc.rb
3167 * p = proc { num = 1 + 2 }
3168 * puts RubyVM::InstructionSequence.disasm(p)
3169 *
3170 * Produces:
3171 *
3172 * == disasm: <RubyVM::InstructionSequence:block in <main>@/tmp/proc.rb>===
3173 * == catch table
3174 * | catch type: redo st: 0000 ed: 0012 sp: 0000 cont: 0000
3175 * | catch type: next st: 0000 ed: 0012 sp: 0000 cont: 0012
3176 * |------------------------------------------------------------------------
3177 * local table (size: 2, argc: 0 [opts: 0, rest: -1, post: 0, block: -1] s1)
3178 * [ 2] num
3179 * 0000 trace 1 ( 1)
3180 * 0002 putobject 1
3181 * 0004 putobject 2
3182 * 0006 opt_plus <ic:1>
3183 * 0008 dup
3184 * 0009 setlocal num, 0
3185 * 0012 leave
3186 *
3187 */
3188static VALUE
3189iseqw_s_disasm(VALUE klass, VALUE body)
3190{
3191 VALUE iseqw = iseqw_s_of(klass, body);
3192 return NIL_P(iseqw) ? Qnil : rb_iseq_disasm(iseqw_check(iseqw));
3193}
3194
3195static VALUE
3196register_label(struct st_table *table, unsigned long idx)
3197{
3198 VALUE sym = rb_str_intern(rb_sprintf("label_%lu", idx));
3199 st_insert(table, idx, sym);
3200 return sym;
3201}
3202
3203static VALUE
3204exception_type2symbol(VALUE type)
3205{
3206 ID id;
3207 switch (type) {
3208 case CATCH_TYPE_RESCUE: CONST_ID(id, "rescue"); break;
3209 case CATCH_TYPE_ENSURE: CONST_ID(id, "ensure"); break;
3210 case CATCH_TYPE_RETRY: CONST_ID(id, "retry"); break;
3211 case CATCH_TYPE_BREAK: CONST_ID(id, "break"); break;
3212 case CATCH_TYPE_REDO: CONST_ID(id, "redo"); break;
3213 case CATCH_TYPE_NEXT: CONST_ID(id, "next"); break;
3214 default:
3215 rb_bug("unknown exception type: %d", (int)type);
3216 }
3217 return ID2SYM(id);
3218}
3219
3220static int
3221cdhash_each(VALUE key, VALUE value, VALUE ary)
3222{
3223 rb_ary_push(ary, obj_resurrect(key));
3224 rb_ary_push(ary, value);
3225 return ST_CONTINUE;
3226}
3227
3228static const rb_data_type_t label_wrapper = {
3229 "label_wrapper",
3230 {(void (*)(void *))rb_mark_tbl, (void (*)(void *))st_free_table, 0, 0,},
3231 0, 0, RUBY_TYPED_FREE_IMMEDIATELY
3232};
3233
3234#define DECL_ID(name) \
3235 static ID id_##name
3236
3237#define INIT_ID(name) \
3238 id_##name = rb_intern(#name)
3239
3240static VALUE
3241iseq_type_id(enum rb_iseq_type type)
3242{
3243 DECL_ID(top);
3244 DECL_ID(method);
3245 DECL_ID(block);
3246 DECL_ID(class);
3247 DECL_ID(rescue);
3248 DECL_ID(ensure);
3249 DECL_ID(eval);
3250 DECL_ID(main);
3251 DECL_ID(plain);
3252
3253 if (id_top == 0) {
3254 INIT_ID(top);
3255 INIT_ID(method);
3256 INIT_ID(block);
3257 INIT_ID(class);
3258 INIT_ID(rescue);
3259 INIT_ID(ensure);
3260 INIT_ID(eval);
3261 INIT_ID(main);
3262 INIT_ID(plain);
3263 }
3264
3265 switch (type) {
3266 case ISEQ_TYPE_TOP: return id_top;
3267 case ISEQ_TYPE_METHOD: return id_method;
3268 case ISEQ_TYPE_BLOCK: return id_block;
3269 case ISEQ_TYPE_CLASS: return id_class;
3270 case ISEQ_TYPE_RESCUE: return id_rescue;
3271 case ISEQ_TYPE_ENSURE: return id_ensure;
3272 case ISEQ_TYPE_EVAL: return id_eval;
3273 case ISEQ_TYPE_MAIN: return id_main;
3274 case ISEQ_TYPE_PLAIN: return id_plain;
3275 };
3276
3277 rb_bug("unsupported iseq type: %d", (int)type);
3278}
3279
3280static VALUE
3281iseq_data_to_ary(const rb_iseq_t *iseq)
3282{
3283 unsigned int i;
3284 long l;
3285 const struct rb_iseq_constant_body *const iseq_body = ISEQ_BODY(iseq);
3286 const struct iseq_insn_info_entry *prev_insn_info;
3287 unsigned int pos;
3288 int last_line = 0;
3289 VALUE *seq, *iseq_original;
3290
3291 VALUE val = rb_ary_new();
3292 ID type; /* Symbol */
3293 VALUE locals = rb_ary_new();
3294 VALUE params = rb_hash_new();
3295 VALUE body = rb_ary_new(); /* [[:insn1, ...], ...] */
3296 VALUE nbody;
3297 VALUE exception = rb_ary_new(); /* [[....]] */
3298 VALUE misc = rb_hash_new();
3299
3300 static ID insn_syms[VM_BARE_INSTRUCTION_SIZE]; /* w/o-trace only */
3301 struct st_table *labels_table = st_init_numtable();
3302 VALUE labels_wrapper = TypedData_Wrap_Struct(0, &label_wrapper, labels_table);
3303
3304 if (insn_syms[0] == 0) {
3305 int i;
3306 for (i=0; i<numberof(insn_syms); i++) {
3307 insn_syms[i] = rb_intern(insn_name(i));
3308 }
3309 }
3310
3311 /* type */
3312 type = iseq_type_id(iseq_body->type);
3313
3314 /* locals */
3315 for (i=0; i<iseq_body->local_table_size; i++) {
3316 ID lid = iseq_body->local_table[i];
3317 if (lid) {
3318 if (rb_id2str(lid)) {
3319 rb_ary_push(locals, ID2SYM(lid));
3320 }
3321 else { /* hidden variable from id_internal() */
3322 rb_ary_push(locals, ULONG2NUM(iseq_body->local_table_size-i+1));
3323 }
3324 }
3325 else {
3326 rb_ary_push(locals, ID2SYM(rb_intern("#arg_rest")));
3327 }
3328 }
3329
3330 /* params */
3331 {
3332 const struct rb_iseq_param_keyword *const keyword = iseq_body->param.keyword;
3333 int j;
3334
3335 if (iseq_body->param.flags.has_opt) {
3336 int len = iseq_body->param.opt_num + 1;
3337 VALUE arg_opt_labels = rb_ary_new2(len);
3338
3339 for (j = 0; j < len; j++) {
3340 VALUE l = register_label(labels_table, iseq_body->param.opt_table[j]);
3341 rb_ary_push(arg_opt_labels, l);
3342 }
3343 rb_hash_aset(params, ID2SYM(rb_intern("opt")), arg_opt_labels);
3344 }
3345
3346 /* commit */
3347 if (iseq_body->param.flags.has_lead) rb_hash_aset(params, ID2SYM(rb_intern("lead_num")), INT2FIX(iseq_body->param.lead_num));
3348 if (iseq_body->param.flags.has_post) rb_hash_aset(params, ID2SYM(rb_intern("post_num")), INT2FIX(iseq_body->param.post_num));
3349 if (iseq_body->param.flags.has_post) rb_hash_aset(params, ID2SYM(rb_intern("post_start")), INT2FIX(iseq_body->param.post_start));
3350 if (iseq_body->param.flags.has_rest) rb_hash_aset(params, ID2SYM(rb_intern("rest_start")), INT2FIX(iseq_body->param.rest_start));
3351 if (iseq_body->param.flags.has_block) rb_hash_aset(params, ID2SYM(rb_intern("block_start")), INT2FIX(iseq_body->param.block_start));
3352 if (iseq_body->param.flags.has_kw) {
3353 VALUE keywords = rb_ary_new();
3354 int i, j;
3355 for (i=0; i<keyword->required_num; i++) {
3356 rb_ary_push(keywords, ID2SYM(keyword->table[i]));
3357 }
3358 for (j=0; i<keyword->num; i++, j++) {
3359 VALUE key = rb_ary_new_from_args(1, ID2SYM(keyword->table[i]));
3360 if (!UNDEF_P(keyword->default_values[j])) {
3361 rb_ary_push(key, keyword->default_values[j]);
3362 }
3363 rb_ary_push(keywords, key);
3364 }
3365
3366 rb_hash_aset(params, ID2SYM(rb_intern("kwbits")),
3367 INT2FIX(keyword->bits_start));
3368 rb_hash_aset(params, ID2SYM(rb_intern("keyword")), keywords);
3369 }
3370 if (iseq_body->param.flags.has_kwrest) rb_hash_aset(params, ID2SYM(rb_intern("kwrest")), INT2FIX(keyword->rest_start));
3371 if (iseq_body->param.flags.ambiguous_param0) rb_hash_aset(params, ID2SYM(rb_intern("ambiguous_param0")), Qtrue);
3372 if (iseq_body->param.flags.use_block) rb_hash_aset(params, ID2SYM(rb_intern("use_block")), Qtrue);
3373 }
3374
3375 /* body */
3376 iseq_original = rb_iseq_original_iseq((rb_iseq_t *)iseq);
3377
3378 for (seq = iseq_original; seq < iseq_original + iseq_body->iseq_size; ) {
3379 VALUE insn = *seq++;
3380 int j, len = insn_len(insn);
3381 VALUE *nseq = seq + len - 1;
3382 VALUE ary = rb_ary_new2(len);
3383
3384 rb_ary_push(ary, ID2SYM(insn_syms[insn%numberof(insn_syms)]));
3385 for (j=0; j<len-1; j++, seq++) {
3386 enum ruby_insn_type_chars op_type = insn_op_type(insn, j);
3387
3388 switch (op_type) {
3389 case TS_OFFSET: {
3390 unsigned long idx = nseq - iseq_original + *seq;
3391 rb_ary_push(ary, register_label(labels_table, idx));
3392 break;
3393 }
3394 case TS_LINDEX:
3395 case TS_NUM:
3396 rb_ary_push(ary, INT2FIX(*seq));
3397 break;
3398 case TS_VALUE:
3399 rb_ary_push(ary, obj_resurrect(*seq));
3400 break;
3401 case TS_ISEQ:
3402 {
3403 const rb_iseq_t *iseq = (rb_iseq_t *)*seq;
3404 if (iseq) {
3405 VALUE val = iseq_data_to_ary(rb_iseq_check(iseq));
3406 rb_ary_push(ary, val);
3407 }
3408 else {
3409 rb_ary_push(ary, Qnil);
3410 }
3411 }
3412 break;
3413 case TS_IC:
3414 {
3415 VALUE list = rb_ary_new();
3416 const ID *ids = ((IC)*seq)->segments;
3417 while (*ids) {
3418 rb_ary_push(list, ID2SYM(*ids++));
3419 }
3420 rb_ary_push(ary, list);
3421 }
3422 break;
3423 case TS_IVC:
3424 case TS_ICVARC:
3425 case TS_ISE:
3426 {
3427 union iseq_inline_storage_entry *is = (union iseq_inline_storage_entry *)*seq;
3428 rb_ary_push(ary, INT2FIX(is - ISEQ_IS_ENTRY_START(ISEQ_BODY(iseq), op_type)));
3429 }
3430 break;
3431 case TS_CALLDATA:
3432 {
3433 struct rb_call_data *cd = (struct rb_call_data *)*seq;
3434 const struct rb_callinfo *ci = cd->ci;
3435 VALUE e = rb_hash_new();
3436 int argc = vm_ci_argc(ci);
3437
3438 ID mid = vm_ci_mid(ci);
3439 rb_hash_aset(e, ID2SYM(rb_intern("mid")), mid ? ID2SYM(mid) : Qnil);
3440 rb_hash_aset(e, ID2SYM(rb_intern("flag")), UINT2NUM(vm_ci_flag(ci)));
3441
3442 if (vm_ci_flag(ci) & VM_CALL_KWARG) {
3443 const struct rb_callinfo_kwarg *kwarg = vm_ci_kwarg(ci);
3444 int i;
3445 VALUE kw = rb_ary_new2((long)kwarg->keyword_len);
3446
3447 argc -= kwarg->keyword_len;
3448 for (i = 0; i < kwarg->keyword_len; i++) {
3449 rb_ary_push(kw, kwarg->keywords[i]);
3450 }
3451 rb_hash_aset(e, ID2SYM(rb_intern("kw_arg")), kw);
3452 }
3453
3454 rb_hash_aset(e, ID2SYM(rb_intern("orig_argc")),
3455 INT2FIX(argc));
3456 rb_ary_push(ary, e);
3457 }
3458 break;
3459 case TS_ID:
3460 rb_ary_push(ary, ID2SYM(*seq));
3461 break;
3462 case TS_CDHASH:
3463 {
3464 VALUE hash = *seq;
3465 VALUE val = rb_ary_new();
3466 int i;
3467
3468 rb_hash_foreach(hash, cdhash_each, val);
3469
3470 for (i=0; i<RARRAY_LEN(val); i+=2) {
3471 VALUE pos = FIX2INT(rb_ary_entry(val, i+1));
3472 unsigned long idx = nseq - iseq_original + pos;
3473
3474 rb_ary_store(val, i+1,
3475 register_label(labels_table, idx));
3476 }
3477 rb_ary_push(ary, val);
3478 }
3479 break;
3480 case TS_FUNCPTR:
3481 {
3482#if SIZEOF_VALUE <= SIZEOF_LONG
3483 VALUE val = LONG2NUM((SIGNED_VALUE)*seq);
3484#else
3485 VALUE val = LL2NUM((SIGNED_VALUE)*seq);
3486#endif
3487 rb_ary_push(ary, val);
3488 }
3489 break;
3490 case TS_BUILTIN:
3491 {
3492 VALUE val = rb_hash_new();
3493#if SIZEOF_VALUE <= SIZEOF_LONG
3494 VALUE func_ptr = LONG2NUM((SIGNED_VALUE)((RB_BUILTIN)*seq)->func_ptr);
3495#else
3496 VALUE func_ptr = LL2NUM((SIGNED_VALUE)((RB_BUILTIN)*seq)->func_ptr);
3497#endif
3498 rb_hash_aset(val, ID2SYM(rb_intern("func_ptr")), func_ptr);
3499 rb_hash_aset(val, ID2SYM(rb_intern("argc")), INT2NUM(((RB_BUILTIN)*seq)->argc));
3500 rb_hash_aset(val, ID2SYM(rb_intern("index")), INT2NUM(((RB_BUILTIN)*seq)->index));
3501 rb_hash_aset(val, ID2SYM(rb_intern("name")), rb_str_new_cstr(((RB_BUILTIN)*seq)->name));
3502 rb_ary_push(ary, val);
3503 }
3504 break;
3505 default:
3506 rb_bug("unknown operand: %c", insn_op_type(insn, j));
3507 }
3508 }
3509 rb_ary_push(body, ary);
3510 }
3511
3512 nbody = body;
3513
3514 /* exception */
3515 if (iseq_body->catch_table) for (i=0; i<iseq_body->catch_table->size; i++) {
3516 VALUE ary = rb_ary_new();
3517 const struct iseq_catch_table_entry *entry =
3518 UNALIGNED_MEMBER_PTR(iseq_body->catch_table, entries[i]);
3519 rb_ary_push(ary, exception_type2symbol(entry->type));
3520 if (entry->iseq) {
3521 rb_ary_push(ary, iseq_data_to_ary(rb_iseq_check(entry->iseq)));
3522 }
3523 else {
3524 rb_ary_push(ary, Qnil);
3525 }
3526 rb_ary_push(ary, register_label(labels_table, entry->start));
3527 rb_ary_push(ary, register_label(labels_table, entry->end));
3528 rb_ary_push(ary, register_label(labels_table, entry->cont));
3529 rb_ary_push(ary, UINT2NUM(entry->sp));
3530 rb_ary_push(exception, ary);
3531 }
3532
3533 /* make body with labels and insert line number */
3534 body = rb_ary_new();
3535 prev_insn_info = NULL;
3536#ifdef USE_ISEQ_NODE_ID
3537 VALUE node_ids = rb_ary_new();
3538#endif
3539
3540 for (l=0, pos=0; l<RARRAY_LEN(nbody); l++) {
3541 const struct iseq_insn_info_entry *info;
3542 VALUE ary = RARRAY_AREF(nbody, l);
3543 st_data_t label;
3544
3545 if (st_lookup(labels_table, pos, &label)) {
3546 rb_ary_push(body, (VALUE)label);
3547 }
3548
3549 info = get_insn_info(iseq, pos);
3550#ifdef USE_ISEQ_NODE_ID
3551 rb_ary_push(node_ids, INT2FIX(info->node_id));
3552#endif
3553
3554 if (prev_insn_info != info) {
3555 int line = info->line_no;
3556 rb_event_flag_t events = info->events;
3557
3558 if (line > 0 && last_line != line) {
3559 rb_ary_push(body, INT2FIX(line));
3560 last_line = line;
3561 }
3562#define CHECK_EVENT(ev) if (events & ev) rb_ary_push(body, ID2SYM(rb_intern(#ev)));
3563 CHECK_EVENT(RUBY_EVENT_LINE);
3564 CHECK_EVENT(RUBY_EVENT_CLASS);
3565 CHECK_EVENT(RUBY_EVENT_END);
3566 CHECK_EVENT(RUBY_EVENT_CALL);
3567 CHECK_EVENT(RUBY_EVENT_RETURN);
3568 CHECK_EVENT(RUBY_EVENT_B_CALL);
3569 CHECK_EVENT(RUBY_EVENT_B_RETURN);
3570 CHECK_EVENT(RUBY_EVENT_RESCUE);
3571#undef CHECK_EVENT
3572 prev_insn_info = info;
3573 }
3574
3575 rb_ary_push(body, ary);
3576 pos += RARRAY_LENINT(ary); /* reject too huge data */
3577 }
3578 RB_GC_GUARD(nbody);
3579 RB_GC_GUARD(labels_wrapper);
3580
3581 rb_hash_aset(misc, ID2SYM(rb_intern("arg_size")), INT2FIX(iseq_body->param.size));
3582 rb_hash_aset(misc, ID2SYM(rb_intern("local_size")), INT2FIX(iseq_body->local_table_size));
3583 rb_hash_aset(misc, ID2SYM(rb_intern("stack_max")), INT2FIX(iseq_body->stack_max));
3584 rb_hash_aset(misc, ID2SYM(rb_intern("node_id")), INT2FIX(iseq_body->location.node_id));
3585 rb_hash_aset(misc, ID2SYM(rb_intern("code_location")),
3586 rb_ary_new_from_args(4,
3587 INT2FIX(iseq_body->location.code_location.beg_pos.lineno),
3588 INT2FIX(iseq_body->location.code_location.beg_pos.column),
3589 INT2FIX(iseq_body->location.code_location.end_pos.lineno),
3590 INT2FIX(iseq_body->location.code_location.end_pos.column)));
3591#ifdef USE_ISEQ_NODE_ID
3592 rb_hash_aset(misc, ID2SYM(rb_intern("node_ids")), node_ids);
3593#endif
3594 rb_hash_aset(misc, ID2SYM(rb_intern("parser")), iseq_body->prism ? ID2SYM(rb_intern("prism")) : ID2SYM(rb_intern("parse.y")));
3595
3596 /*
3597 * [:magic, :major_version, :minor_version, :format_type, :misc,
3598 * :name, :path, :absolute_path, :start_lineno, :type, :locals, :args,
3599 * :catch_table, :bytecode]
3600 */
3601 rb_ary_push(val, rb_str_new2("YARVInstructionSequence/SimpleDataFormat"));
3602 rb_ary_push(val, INT2FIX(ISEQ_MAJOR_VERSION)); /* major */
3603 rb_ary_push(val, INT2FIX(ISEQ_MINOR_VERSION)); /* minor */
3604 rb_ary_push(val, INT2FIX(1));
3605 rb_ary_push(val, misc);
3606 rb_ary_push(val, iseq_body->location.label);
3607 rb_ary_push(val, rb_iseq_path(iseq));
3608 rb_ary_push(val, rb_iseq_realpath(iseq));
3609 rb_ary_push(val, RB_INT2NUM(iseq_body->location.first_lineno));
3610 rb_ary_push(val, ID2SYM(type));
3611 rb_ary_push(val, locals);
3612 rb_ary_push(val, params);
3613 rb_ary_push(val, exception);
3614 rb_ary_push(val, body);
3615 return val;
3616}
3617
3618VALUE
3619rb_iseq_parameters(const rb_iseq_t *iseq, int is_proc)
3620{
3621 int i, r;
3622 const struct rb_iseq_constant_body *const body = ISEQ_BODY(iseq);
3623 const struct rb_iseq_param_keyword *const keyword = body->param.keyword;
3624 VALUE a, args = rb_ary_new2(body->param.size);
3625 ID req, opt, rest, block, key, keyrest;
3626#define PARAM_TYPE(type) rb_ary_push(a = rb_ary_new2(2), ID2SYM(type))
3627#define PARAM_ID(i) body->local_table[(i)]
3628#define PARAM(i, type) ( \
3629 PARAM_TYPE(type), \
3630 rb_id2str(PARAM_ID(i)) ? \
3631 rb_ary_push(a, ID2SYM(PARAM_ID(i))) : \
3632 a)
3633
3634 CONST_ID(req, "req");
3635 CONST_ID(opt, "opt");
3636
3637 if (body->param.flags.forwardable) {
3638 // [[:rest, :*], [:keyrest, :**], [:block, :&]]
3639 CONST_ID(rest, "rest");
3640 CONST_ID(keyrest, "keyrest");
3641 CONST_ID(block, "block");
3642 rb_ary_push(args, rb_ary_new_from_args(2, ID2SYM(rest), ID2SYM(idMULT)));
3643 rb_ary_push(args, rb_ary_new_from_args(2, ID2SYM(keyrest), ID2SYM(idPow)));
3644 rb_ary_push(args, rb_ary_new_from_args(2, ID2SYM(block), ID2SYM(idAnd)));
3645 }
3646
3647 if (is_proc) {
3648 for (i = 0; i < body->param.lead_num; i++) {
3649 PARAM_TYPE(opt);
3650 if (rb_id2str(PARAM_ID(i))) {
3651 rb_ary_push(a, ID2SYM(PARAM_ID(i)));
3652 }
3653 rb_ary_push(args, a);
3654 }
3655 }
3656 else {
3657 for (i = 0; i < body->param.lead_num; i++) {
3658 rb_ary_push(args, PARAM(i, req));
3659 }
3660 }
3661 r = body->param.lead_num + body->param.opt_num;
3662 for (; i < r; i++) {
3663 PARAM_TYPE(opt);
3664 if (rb_id2str(PARAM_ID(i))) {
3665 rb_ary_push(a, ID2SYM(PARAM_ID(i)));
3666 }
3667 rb_ary_push(args, a);
3668 }
3669 if (body->param.flags.has_rest) {
3670 CONST_ID(rest, "rest");
3671 rb_ary_push(args, PARAM(body->param.rest_start, rest));
3672 }
3673 r = body->param.post_start + body->param.post_num;
3674 if (is_proc) {
3675 for (i = body->param.post_start; i < r; i++) {
3676 PARAM_TYPE(opt);
3677 if (rb_id2str(PARAM_ID(i))) {
3678 rb_ary_push(a, ID2SYM(PARAM_ID(i)));
3679 }
3680 rb_ary_push(args, a);
3681 }
3682 }
3683 else {
3684 for (i = body->param.post_start; i < r; i++) {
3685 rb_ary_push(args, PARAM(i, req));
3686 }
3687 }
3688 if (body->param.flags.accepts_no_kwarg) {
3689 ID nokey;
3690 CONST_ID(nokey, "nokey");
3691 PARAM_TYPE(nokey);
3692 rb_ary_push(args, a);
3693 }
3694 if (body->param.flags.has_kw) {
3695 i = 0;
3696 if (keyword->required_num > 0) {
3697 ID keyreq;
3698 CONST_ID(keyreq, "keyreq");
3699 for (; i < keyword->required_num; i++) {
3700 PARAM_TYPE(keyreq);
3701 if (rb_id2str(keyword->table[i])) {
3702 rb_ary_push(a, ID2SYM(keyword->table[i]));
3703 }
3704 rb_ary_push(args, a);
3705 }
3706 }
3707 CONST_ID(key, "key");
3708 for (; i < keyword->num; i++) {
3709 PARAM_TYPE(key);
3710 if (rb_id2str(keyword->table[i])) {
3711 rb_ary_push(a, ID2SYM(keyword->table[i]));
3712 }
3713 rb_ary_push(args, a);
3714 }
3715 }
3716 if (body->param.flags.has_kwrest || body->param.flags.ruby2_keywords) {
3717 ID param;
3718 CONST_ID(keyrest, "keyrest");
3719 PARAM_TYPE(keyrest);
3720 if (body->param.flags.has_kwrest &&
3721 rb_id2str(param = PARAM_ID(keyword->rest_start))) {
3722 rb_ary_push(a, ID2SYM(param));
3723 }
3724 else if (body->param.flags.ruby2_keywords) {
3725 rb_ary_push(a, ID2SYM(idPow));
3726 }
3727 rb_ary_push(args, a);
3728 }
3729 if (body->param.flags.has_block) {
3730 CONST_ID(block, "block");
3731 rb_ary_push(args, PARAM(body->param.block_start, block));
3732 }
3733 return args;
3734}
3735
3736VALUE
3737rb_iseq_defined_string(enum defined_type type)
3738{
3739 static const char expr_names[][18] = {
3740 "nil",
3741 "instance-variable",
3742 "local-variable",
3743 "global-variable",
3744 "class variable",
3745 "constant",
3746 "method",
3747 "yield",
3748 "super",
3749 "self",
3750 "true",
3751 "false",
3752 "assignment",
3753 "expression",
3754 };
3755 const char *estr;
3756
3757 if ((unsigned)(type - 1) >= (unsigned)numberof(expr_names)) rb_bug("unknown defined type %d", type);
3758 estr = expr_names[type - 1];
3759 return rb_fstring_cstr(estr);
3760}
3761
3762// A map from encoded_insn to insn_data: decoded insn number, its len,
3763// decoded ZJIT insn number, non-trace version of encoded insn,
3764// trace version, and zjit version.
3765static st_table *encoded_insn_data;
3766typedef struct insn_data_struct {
3767 int insn;
3768 int insn_len;
3769 void *notrace_encoded_insn;
3770 void *trace_encoded_insn;
3771#if USE_ZJIT
3772 int zjit_insn;
3773 void *zjit_encoded_insn;
3774#endif
3775} insn_data_t;
3776static insn_data_t insn_data[VM_BARE_INSTRUCTION_SIZE];
3777
3778void
3779rb_free_encoded_insn_data(void)
3780{
3781 st_free_table(encoded_insn_data);
3782}
3783
3784// Initialize a table to decode bare, trace, and zjit instructions.
3785// This function also determines which instructions are used when TracePoint is enabled.
3786void
3787rb_vm_encoded_insn_data_table_init(void)
3788{
3789#if OPT_DIRECT_THREADED_CODE || OPT_CALL_THREADED_CODE
3790 const void * const *table = rb_vm_get_insns_address_table();
3791#define INSN_CODE(insn) ((VALUE)table[insn])
3792#else
3793#define INSN_CODE(insn) ((VALUE)(insn))
3794#endif
3795 encoded_insn_data = st_init_numtable_with_size(VM_BARE_INSTRUCTION_SIZE);
3796
3797 for (int insn = 0; insn < VM_BARE_INSTRUCTION_SIZE; insn++) {
3798 insn_data[insn].insn = insn;
3799 insn_data[insn].insn_len = insn_len(insn);
3800
3801 // When tracing :return events, we convert opt_invokebuiltin_delegate_leave + leave into
3802 // opt_invokebuiltin_delegate + trace_leave, presumably because we don't want to fire
3803 // :return events before invokebuiltin. https://github.com/ruby/ruby/pull/3256
3804 int notrace_insn = (insn != BIN(opt_invokebuiltin_delegate_leave)) ? insn : BIN(opt_invokebuiltin_delegate);
3805 insn_data[insn].notrace_encoded_insn = (void *)INSN_CODE(notrace_insn);
3806 insn_data[insn].trace_encoded_insn = (void *)INSN_CODE(notrace_insn + VM_BARE_INSTRUCTION_SIZE);
3807
3808 st_data_t key1 = (st_data_t)INSN_CODE(insn);
3809 st_data_t key2 = (st_data_t)INSN_CODE(insn + VM_BARE_INSTRUCTION_SIZE);
3810 st_add_direct(encoded_insn_data, key1, (st_data_t)&insn_data[insn]);
3811 st_add_direct(encoded_insn_data, key2, (st_data_t)&insn_data[insn]);
3812
3813#if USE_ZJIT
3814 int zjit_insn = vm_bare_insn_to_zjit_insn(insn);
3815 insn_data[insn].zjit_insn = zjit_insn;
3816 insn_data[insn].zjit_encoded_insn = (insn != zjit_insn) ? (void *)INSN_CODE(zjit_insn) : 0;
3817
3818 if (insn != zjit_insn) {
3819 st_data_t key3 = (st_data_t)INSN_CODE(zjit_insn);
3820 st_add_direct(encoded_insn_data, key3, (st_data_t)&insn_data[insn]);
3821 }
3822#endif
3823 }
3824}
3825
3826// Decode an insn address to an insn. This returns bare instructions
3827// even if they're trace/zjit instructions. Use rb_vm_insn_addr2opcode
3828// to decode trace/zjit instructions as is.
3829int
3830rb_vm_insn_addr2insn(const void *addr)
3831{
3832 st_data_t key = (st_data_t)addr;
3833 st_data_t val;
3834
3835 if (st_lookup(encoded_insn_data, key, &val)) {
3836 insn_data_t *e = (insn_data_t *)val;
3837 return (int)e->insn;
3838 }
3839
3840 rb_bug("rb_vm_insn_addr2insn: invalid insn address: %p", addr);
3841}
3842
3843// Decode an insn address to an insn. Unlike rb_vm_insn_addr2insn,
3844// this function can return trace/zjit opcode variants.
3845int
3846rb_vm_insn_addr2opcode(const void *addr)
3847{
3848 st_data_t key = (st_data_t)addr;
3849 st_data_t val;
3850
3851 if (st_lookup(encoded_insn_data, key, &val)) {
3852 insn_data_t *e = (insn_data_t *)val;
3853 int opcode = e->insn;
3854 if (addr == e->trace_encoded_insn) {
3855 opcode += VM_BARE_INSTRUCTION_SIZE;
3856 }
3857#if USE_ZJIT
3858 else if (addr == e->zjit_encoded_insn) {
3859 opcode = e->zjit_insn;
3860 }
3861#endif
3862 return opcode;
3863 }
3864
3865 rb_bug("rb_vm_insn_addr2opcode: invalid insn address: %p", addr);
3866}
3867
3868// Decode `ISEQ_BODY(iseq)->iseq_encoded[i]` to an insn. This returns
3869// bare instructions even if they're trace/zjit instructions. Use
3870// rb_vm_insn_addr2opcode to decode trace/zjit instructions as is.
3871int
3872rb_vm_insn_decode(const VALUE encoded)
3873{
3874#if OPT_DIRECT_THREADED_CODE || OPT_CALL_THREADED_CODE
3875 int insn = rb_vm_insn_addr2insn((void *)encoded);
3876#else
3877 int insn = (int)encoded;
3878#endif
3879 return insn;
3880}
3881
3882// Turn on or off tracing for a given instruction address
3883static inline int
3884encoded_iseq_trace_instrument(VALUE *iseq_encoded_insn, rb_event_flag_t turnon, bool remain_current_trace)
3885{
3886 st_data_t key = (st_data_t)*iseq_encoded_insn;
3887 st_data_t val;
3888
3889 if (st_lookup(encoded_insn_data, key, &val)) {
3890 insn_data_t *e = (insn_data_t *)val;
3891 if (remain_current_trace && key == (st_data_t)e->trace_encoded_insn) {
3892 turnon = 1;
3893 }
3894 *iseq_encoded_insn = (VALUE) (turnon ? e->trace_encoded_insn : e->notrace_encoded_insn);
3895 return e->insn_len;
3896 }
3897
3898 rb_bug("trace_instrument: invalid insn address: %p", (void *)*iseq_encoded_insn);
3899}
3900
3901// Turn off tracing for an instruction at pos after tracing event flags are cleared
3902void
3903rb_iseq_trace_flag_cleared(const rb_iseq_t *iseq, size_t pos)
3904{
3905 const struct rb_iseq_constant_body *const body = ISEQ_BODY(iseq);
3906 VALUE *iseq_encoded = (VALUE *)body->iseq_encoded;
3907 encoded_iseq_trace_instrument(&iseq_encoded[pos], 0, false);
3908}
3909
3910// We need to fire call events on instructions with b_call events if the block
3911// is running as a method. So, if we are listening for call events, then
3912// instructions that have b_call events need to become trace variants.
3913// Use this function when making decisions about recompiling to trace variants.
3914static inline rb_event_flag_t
3915add_bmethod_events(rb_event_flag_t events)
3916{
3917 if (events & RUBY_EVENT_CALL) {
3918 events |= RUBY_EVENT_B_CALL;
3919 }
3920 if (events & RUBY_EVENT_RETURN) {
3921 events |= RUBY_EVENT_B_RETURN;
3922 }
3923 return events;
3924}
3925
3926// Note, to support call/return events for bmethods, turnon_event can have more events than tpval.
3927static int
3928iseq_add_local_tracepoint(const rb_iseq_t *iseq, rb_event_flag_t turnon_events, VALUE tpval, unsigned int target_line)
3929{
3930 unsigned int pc;
3931 int n = 0;
3932 const struct rb_iseq_constant_body *const body = ISEQ_BODY(iseq);
3933 VALUE *iseq_encoded = (VALUE *)body->iseq_encoded;
3934
3935 VM_ASSERT(ISEQ_EXECUTABLE_P(iseq));
3936
3937 for (pc=0; pc<body->iseq_size;) {
3938 const struct iseq_insn_info_entry *entry = get_insn_info(iseq, pc);
3939 rb_event_flag_t pc_events = entry->events;
3940 rb_event_flag_t target_events = turnon_events;
3941 unsigned int line = (int)entry->line_no;
3942
3943 if (target_line == 0 || target_line == line) {
3944 /* ok */
3945 }
3946 else {
3947 target_events &= ~RUBY_EVENT_LINE;
3948 }
3949
3950 if (pc_events & target_events) {
3951 n++;
3952 }
3953 pc += encoded_iseq_trace_instrument(&iseq_encoded[pc], pc_events & (target_events | iseq->aux.exec.global_trace_events), true);
3954 }
3955
3956 if (n > 0) {
3957 if (iseq->aux.exec.local_hooks == NULL) {
3958 ((rb_iseq_t *)iseq)->aux.exec.local_hooks = RB_ZALLOC(rb_hook_list_t);
3959 iseq->aux.exec.local_hooks->is_local = true;
3960 }
3961 rb_hook_list_connect_tracepoint((VALUE)iseq, iseq->aux.exec.local_hooks, tpval, target_line);
3962 }
3963
3964 return n;
3965}
3966
3968 rb_event_flag_t turnon_events;
3969 VALUE tpval;
3970 unsigned int target_line;
3971 int n;
3972};
3973
3974static void
3975iseq_add_local_tracepoint_i(const rb_iseq_t *iseq, void *p)
3976{
3978 data->n += iseq_add_local_tracepoint(iseq, data->turnon_events, data->tpval, data->target_line);
3979 iseq_iterate_children(iseq, iseq_add_local_tracepoint_i, p);
3980}
3981
3982int
3983rb_iseq_add_local_tracepoint_recursively(const rb_iseq_t *iseq, rb_event_flag_t turnon_events, VALUE tpval, unsigned int target_line, bool target_bmethod)
3984{
3986 if (target_bmethod) {
3987 turnon_events = add_bmethod_events(turnon_events);
3988 }
3989 data.turnon_events = turnon_events;
3990 data.tpval = tpval;
3991 data.target_line = target_line;
3992 data.n = 0;
3993
3994 iseq_add_local_tracepoint_i(iseq, (void *)&data);
3995 if (0) rb_funcall(Qnil, rb_intern("puts"), 1, rb_iseq_disasm(iseq)); /* for debug */
3996 return data.n;
3997}
3998
3999static int
4000iseq_remove_local_tracepoint(const rb_iseq_t *iseq, VALUE tpval)
4001{
4002 int n = 0;
4003
4004 if (iseq->aux.exec.local_hooks) {
4005 unsigned int pc;
4006 const struct rb_iseq_constant_body *const body = ISEQ_BODY(iseq);
4007 VALUE *iseq_encoded = (VALUE *)body->iseq_encoded;
4008 rb_event_flag_t local_events = 0;
4009
4010 rb_hook_list_remove_tracepoint(iseq->aux.exec.local_hooks, tpval);
4011 local_events = iseq->aux.exec.local_hooks->events;
4012
4013 if (local_events == 0) {
4014 rb_hook_list_free(iseq->aux.exec.local_hooks);
4015 ((rb_iseq_t *)iseq)->aux.exec.local_hooks = NULL;
4016 }
4017
4018 local_events = add_bmethod_events(local_events);
4019 for (pc = 0; pc<body->iseq_size;) {
4020 rb_event_flag_t pc_events = rb_iseq_event_flags(iseq, pc);
4021 pc += encoded_iseq_trace_instrument(&iseq_encoded[pc], pc_events & (local_events | iseq->aux.exec.global_trace_events), false);
4022 }
4023 }
4024 return n;
4025}
4026
4028 VALUE tpval;
4029 int n;
4030};
4031
4032static void
4033iseq_remove_local_tracepoint_i(const rb_iseq_t *iseq, void *p)
4034{
4036 data->n += iseq_remove_local_tracepoint(iseq, data->tpval);
4037 iseq_iterate_children(iseq, iseq_remove_local_tracepoint_i, p);
4038}
4039
4040int
4041rb_iseq_remove_local_tracepoint_recursively(const rb_iseq_t *iseq, VALUE tpval)
4042{
4044 data.tpval = tpval;
4045 data.n = 0;
4046
4047 iseq_remove_local_tracepoint_i(iseq, (void *)&data);
4048 return data.n;
4049}
4050
4051void
4052rb_iseq_trace_set(const rb_iseq_t *iseq, rb_event_flag_t turnon_events)
4053{
4054 if (iseq->aux.exec.global_trace_events == turnon_events) {
4055 return;
4056 }
4057
4058 if (!ISEQ_EXECUTABLE_P(iseq)) {
4059 /* this is building ISeq */
4060 return;
4061 }
4062 else {
4063 unsigned int pc;
4064 const struct rb_iseq_constant_body *const body = ISEQ_BODY(iseq);
4065 VALUE *iseq_encoded = (VALUE *)body->iseq_encoded;
4066 rb_event_flag_t enabled_events;
4067 rb_event_flag_t local_events = iseq->aux.exec.local_hooks ? iseq->aux.exec.local_hooks->events : 0;
4068 ((rb_iseq_t *)iseq)->aux.exec.global_trace_events = turnon_events;
4069 enabled_events = add_bmethod_events(turnon_events | local_events);
4070
4071 for (pc=0; pc<body->iseq_size;) {
4072 rb_event_flag_t pc_events = rb_iseq_event_flags(iseq, pc);
4073 pc += encoded_iseq_trace_instrument(&iseq_encoded[pc], pc_events & enabled_events, true);
4074 }
4075 }
4076}
4077
4078void rb_vm_cc_general(const struct rb_callcache *cc);
4079
4080static bool
4081clear_attr_cc(VALUE v)
4082{
4083 if (imemo_type_p(v, imemo_callcache) && vm_cc_ivar_p((const struct rb_callcache *)v)) {
4084 rb_vm_cc_general((struct rb_callcache *)v);
4085 return true;
4086 }
4087 else {
4088 return false;
4089 }
4090}
4091
4092static bool
4093clear_bf_cc(VALUE v)
4094{
4095 if (imemo_type_p(v, imemo_callcache) && vm_cc_bf_p((const struct rb_callcache *)v)) {
4096 rb_vm_cc_general((struct rb_callcache *)v);
4097 return true;
4098 }
4099 else {
4100 return false;
4101 }
4102}
4103
4104static int
4105clear_attr_ccs_i(void *vstart, void *vend, size_t stride, void *data)
4106{
4107 VALUE v = (VALUE)vstart;
4108 for (; v != (VALUE)vend; v += stride) {
4109 void *ptr = rb_asan_poisoned_object_p(v);
4110 rb_asan_unpoison_object(v, false);
4111 clear_attr_cc(v);
4112 asan_poison_object_if(ptr, v);
4113 }
4114 return 0;
4115}
4116
4117void
4118rb_clear_attr_ccs(void)
4119{
4120 rb_objspace_each_objects(clear_attr_ccs_i, NULL);
4121}
4122
4123static int
4124clear_bf_ccs_i(void *vstart, void *vend, size_t stride, void *data)
4125{
4126 VALUE v = (VALUE)vstart;
4127 for (; v != (VALUE)vend; v += stride) {
4128 void *ptr = rb_asan_poisoned_object_p(v);
4129 rb_asan_unpoison_object(v, false);
4130 clear_bf_cc(v);
4131 asan_poison_object_if(ptr, v);
4132 }
4133 return 0;
4134}
4135
4136void
4137rb_clear_bf_ccs(void)
4138{
4139 rb_objspace_each_objects(clear_bf_ccs_i, NULL);
4140}
4141
4142static int
4143trace_set_i(void *vstart, void *vend, size_t stride, void *data)
4144{
4145 rb_event_flag_t turnon_events = *(rb_event_flag_t *)data;
4146
4147 VALUE v = (VALUE)vstart;
4148 for (; v != (VALUE)vend; v += stride) {
4149 void *ptr = rb_asan_poisoned_object_p(v);
4150 rb_asan_unpoison_object(v, false);
4151
4152 if (rb_obj_is_iseq(v)) {
4153 rb_iseq_trace_set(rb_iseq_check((rb_iseq_t *)v), turnon_events);
4154 }
4155 else if (clear_attr_cc(v)) {
4156 }
4157 else if (clear_bf_cc(v)) {
4158 }
4159
4160 asan_poison_object_if(ptr, v);
4161 }
4162 return 0;
4163}
4164
4165void
4166rb_iseq_trace_set_all(rb_event_flag_t turnon_events)
4167{
4168 rb_objspace_each_objects(trace_set_i, &turnon_events);
4169}
4170
4171VALUE
4172rb_iseqw_local_variables(VALUE iseqval)
4173{
4174 return rb_iseq_local_variables(iseqw_check(iseqval));
4175}
4176
4177/*
4178 * call-seq:
4179 * iseq.to_binary(extra_data = nil) -> binary str
4180 *
4181 * Returns serialized iseq binary format data as a String object.
4182 * A corresponding iseq object is created by
4183 * RubyVM::InstructionSequence.load_from_binary() method.
4184 *
4185 * String extra_data will be saved with binary data.
4186 * You can access this data with
4187 * RubyVM::InstructionSequence.load_from_binary_extra_data(binary).
4188 *
4189 * Note that the translated binary data is not portable.
4190 * You can not move this binary data to another machine.
4191 * You can not use the binary data which is created by another
4192 * version/another architecture of Ruby.
4193 */
4194static VALUE
4195iseqw_to_binary(int argc, VALUE *argv, VALUE self)
4196{
4197 VALUE opt = !rb_check_arity(argc, 0, 1) ? Qnil : argv[0];
4198 return rb_iseq_ibf_dump(iseqw_check(self), opt);
4199}
4200
4201/*
4202 * call-seq:
4203 * RubyVM::InstructionSequence.load_from_binary(binary) -> iseq
4204 *
4205 * Load an iseq object from binary format String object
4206 * created by RubyVM::InstructionSequence.to_binary.
4207 *
4208 * This loader does not have a verifier, so that loading broken/modified
4209 * binary causes critical problem.
4210 *
4211 * You should not load binary data provided by others.
4212 * You should use binary data translated by yourself.
4213 */
4214static VALUE
4215iseqw_s_load_from_binary(VALUE self, VALUE str)
4216{
4217 return iseqw_new(rb_iseq_ibf_load(str));
4218}
4219
4220/*
4221 * call-seq:
4222 * RubyVM::InstructionSequence.load_from_binary_extra_data(binary) -> str
4223 *
4224 * Load extra data embed into binary format String object.
4225 */
4226static VALUE
4227iseqw_s_load_from_binary_extra_data(VALUE self, VALUE str)
4228{
4229 return rb_iseq_ibf_load_extra_data(str);
4230}
4231
4232#if VM_INSN_INFO_TABLE_IMPL == 2
4233
4234/* An implementation of succinct bit-vector for insn_info table.
4235 *
4236 * A succinct bit-vector is a small and efficient data structure that provides
4237 * a bit-vector augmented with an index for O(1) rank operation:
4238 *
4239 * rank(bv, n): the number of 1's within a range from index 0 to index n
4240 *
4241 * This can be used to lookup insn_info table from PC.
4242 * For example, consider the following iseq and insn_info_table:
4243 *
4244 * iseq insn_info_table
4245 * PC insn+operand position lineno event
4246 * 0: insn1 0: 1 [Li]
4247 * 2: insn2 2: 2 [Li] <= (A)
4248 * 5: insn3 8: 3 [Li] <= (B)
4249 * 8: insn4
4250 *
4251 * In this case, a succinct bit-vector whose indexes 0, 2, 8 is "1" and
4252 * other indexes is "0", i.e., "101000001", is created.
4253 * To lookup the lineno of insn2, calculate rank("10100001", 2) = 2, so
4254 * the line (A) is the entry in question.
4255 * To lookup the lineno of insn4, calculate rank("10100001", 8) = 3, so
4256 * the line (B) is the entry in question.
4257 *
4258 * A naive implementation of succinct bit-vector works really well
4259 * not only for large size but also for small size. However, it has
4260 * tiny overhead for very small size. So, this implementation consist
4261 * of two parts: one part is the "immediate" table that keeps rank result
4262 * as a raw table, and the other part is a normal succinct bit-vector.
4263 */
4264
4265#define IMMEDIATE_TABLE_SIZE 54 /* a multiple of 9, and < 128 */
4266
4267struct succ_index_table {
4268 uint64_t imm_part[IMMEDIATE_TABLE_SIZE / 9];
4269 struct succ_dict_block {
4270 unsigned int rank;
4271 uint64_t small_block_ranks; /* 9 bits * 7 = 63 bits */
4272 uint64_t bits[512/64];
4273 } succ_part[FLEX_ARY_LEN];
4274};
4275
4276#define imm_block_rank_set(v, i, r) (v) |= (uint64_t)(r) << (7 * (i))
4277#define imm_block_rank_get(v, i) (((int)((v) >> ((i) * 7))) & 0x7f)
4278#define small_block_rank_set(v, i, r) (v) |= (uint64_t)(r) << (9 * ((i) - 1))
4279#define small_block_rank_get(v, i) ((i) == 0 ? 0 : (((int)((v) >> (((i) - 1) * 9))) & 0x1ff))
4280
4281static struct succ_index_table *
4282succ_index_table_create(int max_pos, int *data, int size)
4283{
4284 const int imm_size = (max_pos < IMMEDIATE_TABLE_SIZE ? max_pos + 8 : IMMEDIATE_TABLE_SIZE) / 9;
4285 const int succ_size = (max_pos < IMMEDIATE_TABLE_SIZE ? 0 : (max_pos - IMMEDIATE_TABLE_SIZE + 511)) / 512;
4286 struct succ_index_table *sd =
4287 rb_xcalloc_mul_add_mul(
4288 imm_size, sizeof(uint64_t),
4289 succ_size, sizeof(struct succ_dict_block));
4290 int i, j, k, r;
4291
4292 r = 0;
4293 for (j = 0; j < imm_size; j++) {
4294 for (i = 0; i < 9; i++) {
4295 if (r < size && data[r] == j * 9 + i) r++;
4296 imm_block_rank_set(sd->imm_part[j], i, r);
4297 }
4298 }
4299 for (k = 0; k < succ_size; k++) {
4300 struct succ_dict_block *sd_block = &sd->succ_part[k];
4301 int small_rank = 0;
4302 sd_block->rank = r;
4303 for (j = 0; j < 8; j++) {
4304 uint64_t bits = 0;
4305 if (j) small_block_rank_set(sd_block->small_block_ranks, j, small_rank);
4306 for (i = 0; i < 64; i++) {
4307 if (r < size && data[r] == k * 512 + j * 64 + i + IMMEDIATE_TABLE_SIZE) {
4308 bits |= ((uint64_t)1) << i;
4309 r++;
4310 }
4311 }
4312 sd_block->bits[j] = bits;
4313 small_rank += rb_popcount64(bits);
4314 }
4315 }
4316 return sd;
4317}
4318
4319static unsigned int *
4320succ_index_table_invert(int max_pos, struct succ_index_table *sd, int size)
4321{
4322 const int imm_size = (max_pos < IMMEDIATE_TABLE_SIZE ? max_pos + 8 : IMMEDIATE_TABLE_SIZE) / 9;
4323 const int succ_size = (max_pos < IMMEDIATE_TABLE_SIZE ? 0 : (max_pos - IMMEDIATE_TABLE_SIZE + 511)) / 512;
4324 unsigned int *positions = ALLOC_N(unsigned int, size), *p;
4325 int i, j, k, r = -1;
4326 p = positions;
4327 for (j = 0; j < imm_size; j++) {
4328 for (i = 0; i < 9; i++) {
4329 int nr = imm_block_rank_get(sd->imm_part[j], i);
4330 if (r != nr) *p++ = j * 9 + i;
4331 r = nr;
4332 }
4333 }
4334 for (k = 0; k < succ_size; k++) {
4335 for (j = 0; j < 8; j++) {
4336 for (i = 0; i < 64; i++) {
4337 if (sd->succ_part[k].bits[j] & (((uint64_t)1) << i)) {
4338 *p++ = k * 512 + j * 64 + i + IMMEDIATE_TABLE_SIZE;
4339 }
4340 }
4341 }
4342 }
4343 return positions;
4344}
4345
4346static int
4347succ_index_lookup(const struct succ_index_table *sd, int x)
4348{
4349 if (x < IMMEDIATE_TABLE_SIZE) {
4350 const int i = x / 9;
4351 const int j = x % 9;
4352 return imm_block_rank_get(sd->imm_part[i], j);
4353 }
4354 else {
4355 const int block_index = (x - IMMEDIATE_TABLE_SIZE) / 512;
4356 const struct succ_dict_block *block = &sd->succ_part[block_index];
4357 const int block_bit_index = (x - IMMEDIATE_TABLE_SIZE) % 512;
4358 const int small_block_index = block_bit_index / 64;
4359 const int small_block_popcount = small_block_rank_get(block->small_block_ranks, small_block_index);
4360 const int popcnt = rb_popcount64(block->bits[small_block_index] << (63 - block_bit_index % 64));
4361
4362 return block->rank + small_block_popcount + popcnt;
4363 }
4364}
4365#endif
4366
4367
4368/*
4369 * call-seq:
4370 * iseq.script_lines -> array or nil
4371 *
4372 * It returns recorded script lines if it is available.
4373 * The script lines are not limited to the iseq range, but
4374 * are entire lines of the source file.
4375 *
4376 * Note that this is an API for ruby internal use, debugging,
4377 * and research. Do not use this for any other purpose.
4378 * The compatibility is not guaranteed.
4379 */
4380static VALUE
4381iseqw_script_lines(VALUE self)
4382{
4383 const rb_iseq_t *iseq = iseqw_check(self);
4384 return ISEQ_BODY(iseq)->variable.script_lines;
4385}
4386
4387/*
4388 * Document-class: RubyVM::InstructionSequence
4389 *
4390 * The InstructionSequence class represents a compiled sequence of
4391 * instructions for the Virtual Machine used in MRI. Not all implementations of Ruby
4392 * may implement this class, and for the implementations that implement it,
4393 * the methods defined and behavior of the methods can change in any version.
4394 *
4395 * With it, you can get a handle to the instructions that make up a method or
4396 * a proc, compile strings of Ruby code down to VM instructions, and
4397 * disassemble instruction sequences to strings for easy inspection. It is
4398 * mostly useful if you want to learn how YARV works, but it also lets
4399 * you control various settings for the Ruby iseq compiler.
4400 *
4401 * You can find the source for the VM instructions in +insns.def+ in the Ruby
4402 * source.
4403 *
4404 * The instruction sequence results will almost certainly change as Ruby
4405 * changes, so example output in this documentation may be different from what
4406 * you see.
4407 *
4408 * Of course, this class is MRI specific.
4409 */
4410
4411void
4412Init_ISeq(void)
4413{
4414 /* declare ::RubyVM::InstructionSequence */
4415 rb_cISeq = rb_define_class_under(rb_cRubyVM, "InstructionSequence", rb_cObject);
4416 rb_undef_alloc_func(rb_cISeq);
4417 rb_define_method(rb_cISeq, "inspect", iseqw_inspect, 0);
4418 rb_define_method(rb_cISeq, "disasm", iseqw_disasm, 0);
4419 rb_define_method(rb_cISeq, "disassemble", iseqw_disasm, 0);
4420 rb_define_method(rb_cISeq, "to_a", iseqw_to_a, 0);
4421 rb_define_method(rb_cISeq, "eval", iseqw_eval, 0);
4422
4423 rb_define_method(rb_cISeq, "to_binary", iseqw_to_binary, -1);
4424 rb_define_singleton_method(rb_cISeq, "load_from_binary", iseqw_s_load_from_binary, 1);
4425 rb_define_singleton_method(rb_cISeq, "load_from_binary_extra_data", iseqw_s_load_from_binary_extra_data, 1);
4426
4427 /* location APIs */
4428 rb_define_method(rb_cISeq, "path", iseqw_path, 0);
4429 rb_define_method(rb_cISeq, "absolute_path", iseqw_absolute_path, 0);
4430 rb_define_method(rb_cISeq, "label", iseqw_label, 0);
4431 rb_define_method(rb_cISeq, "base_label", iseqw_base_label, 0);
4432 rb_define_method(rb_cISeq, "first_lineno", iseqw_first_lineno, 0);
4433 rb_define_method(rb_cISeq, "trace_points", iseqw_trace_points, 0);
4434 rb_define_method(rb_cISeq, "each_child", iseqw_each_child, 0);
4435
4436#if 0 /* TBD */
4437 rb_define_private_method(rb_cISeq, "marshal_dump", iseqw_marshal_dump, 0);
4438 rb_define_private_method(rb_cISeq, "marshal_load", iseqw_marshal_load, 1);
4439 /* disable this feature because there is no verifier. */
4440 rb_define_singleton_method(rb_cISeq, "load", iseq_s_load, -1);
4441#endif
4442 (void)iseq_s_load;
4443
4444 rb_define_singleton_method(rb_cISeq, "compile", iseqw_s_compile, -1);
4445 rb_define_singleton_method(rb_cISeq, "compile_parsey", iseqw_s_compile_parsey, -1);
4446 rb_define_singleton_method(rb_cISeq, "compile_prism", iseqw_s_compile_prism, -1);
4447 rb_define_singleton_method(rb_cISeq, "compile_file_prism", iseqw_s_compile_file_prism, -1);
4448 rb_define_singleton_method(rb_cISeq, "new", iseqw_s_compile, -1);
4449 rb_define_singleton_method(rb_cISeq, "compile_file", iseqw_s_compile_file, -1);
4450 rb_define_singleton_method(rb_cISeq, "compile_option", iseqw_s_compile_option_get, 0);
4451 rb_define_singleton_method(rb_cISeq, "compile_option=", iseqw_s_compile_option_set, 1);
4452 rb_define_singleton_method(rb_cISeq, "disasm", iseqw_s_disasm, 1);
4453 rb_define_singleton_method(rb_cISeq, "disassemble", iseqw_s_disasm, 1);
4454 rb_define_singleton_method(rb_cISeq, "of", iseqw_s_of, 1);
4455
4456 // script lines
4457 rb_define_method(rb_cISeq, "script_lines", iseqw_script_lines, 0);
4458
4459 rb_undef_method(CLASS_OF(rb_cISeq), "translate");
4460 rb_undef_method(CLASS_OF(rb_cISeq), "load_iseq");
4461}
#define RUBY_ASSERT(...)
Asserts that the given expression is truthy if and only if RUBY_DEBUG is truthy.
Definition assert.h:219
#define rb_define_method(klass, mid, func, arity)
Defines klass#mid.
#define rb_define_singleton_method(klass, mid, func, arity)
Defines klass.mid.
#define rb_define_private_method(klass, mid, func, arity)
Defines klass#mid and makes it private.
#define RUBY_EVENT_END
Encountered an end of a class clause.
Definition event.h:40
#define RUBY_EVENT_C_CALL
A method, written in C, is called.
Definition event.h:43
#define RUBY_EVENT_B_RETURN
Encountered a next statement.
Definition event.h:56
#define RUBY_EVENT_CLASS
Encountered a new class.
Definition event.h:39
#define RUBY_EVENT_LINE
Encountered a new line.
Definition event.h:38
#define RUBY_EVENT_RETURN
Encountered a return statement.
Definition event.h:42
#define RUBY_EVENT_C_RETURN
Return from a method, written in C.
Definition event.h:44
#define RUBY_EVENT_B_CALL
Encountered an yield statement.
Definition event.h:55
uint32_t rb_event_flag_t
Represents event(s).
Definition event.h:108
#define RUBY_EVENT_CALL
A method, written in Ruby, is called.
Definition event.h:41
#define RUBY_EVENT_RESCUE
Encountered a rescue statement.
Definition event.h:61
VALUE rb_define_class_under(VALUE outer, const char *name, VALUE super)
Defines a class under the namespace of outer.
Definition class.c:1510
void rb_undef_method(VALUE klass, const char *name)
Defines an undef of a method.
Definition class.c:2663
int rb_scan_args(int argc, const VALUE *argv, const char *fmt,...)
Retrieves argument from argc and argv to given VALUE references according to the format string.
Definition class.c:3133
#define rb_str_new2
Old name of rb_str_new_cstr.
Definition string.h:1675
#define T_FILE
Old name of RUBY_T_FILE.
Definition value_type.h:62
#define T_STRING
Old name of RUBY_T_STRING.
Definition value_type.h:78
#define Qundef
Old name of RUBY_Qundef.
#define INT2FIX
Old name of RB_INT2FIX.
Definition long.h:48
#define rb_str_cat2
Old name of rb_str_cat_cstr.
Definition string.h:1683
#define ID2SYM
Old name of RB_ID2SYM.
Definition symbol.h:44
#define SPECIAL_CONST_P
Old name of RB_SPECIAL_CONST_P.
#define ULONG2NUM
Old name of RB_ULONG2NUM.
Definition long.h:60
#define SYM2ID
Old name of RB_SYM2ID.
Definition symbol.h:45
#define ZALLOC
Old name of RB_ZALLOC.
Definition memory.h:402
#define LL2NUM
Old name of RB_LL2NUM.
Definition long_long.h:30
#define CLASS_OF
Old name of rb_class_of.
Definition globals.h:206
#define T_NONE
Old name of RUBY_T_NONE.
Definition value_type.h:74
#define FIX2INT
Old name of RB_FIX2INT.
Definition int.h:41
#define T_HASH
Old name of RUBY_T_HASH.
Definition value_type.h:65
#define ALLOC_N
Old name of RB_ALLOC_N.
Definition memory.h:399
#define FL_TEST_RAW
Old name of RB_FL_TEST_RAW.
Definition fl_type.h:131
#define LONG2NUM
Old name of RB_LONG2NUM.
Definition long.h:50
#define Qtrue
Old name of RUBY_Qtrue.
#define NUM2INT
Old name of RB_NUM2INT.
Definition int.h:44
#define INT2NUM
Old name of RB_INT2NUM.
Definition int.h:43
#define Qnil
Old name of RUBY_Qnil.
#define Qfalse
Old name of RUBY_Qfalse.
#define T_ARRAY
Old name of RUBY_T_ARRAY.
Definition value_type.h:56
#define NIL_P
Old name of RB_NIL_P.
#define BUILTIN_TYPE
Old name of RB_BUILTIN_TYPE.
Definition value_type.h:85
#define NUM2LONG
Old name of RB_NUM2LONG.
Definition long.h:51
#define UINT2NUM
Old name of RB_UINT2NUM.
Definition int.h:46
#define FIXNUM_P
Old name of RB_FIXNUM_P.
#define CONST_ID
Old name of RUBY_CONST_ID.
Definition symbol.h:47
#define rb_ary_new2
Old name of rb_ary_new_capa.
Definition array.h:657
void rb_exc_raise(VALUE mesg)
Raises an exception in the current thread.
Definition eval.c:682
VALUE rb_eTypeError
TypeError exception.
Definition error.c:1430
void * rb_check_typeddata(VALUE obj, const rb_data_type_t *data_type)
Identical to rb_typeddata_is_kind_of(), except it raises exceptions instead of returning false.
Definition error.c:1397
VALUE rb_eSyntaxError
SyntaxError exception.
Definition error.c:1447
VALUE rb_class_superclass(VALUE klass)
Queries the parent of the given class.
Definition object.c:2179
VALUE rb_obj_hide(VALUE obj)
Make the object invisible from Ruby code.
Definition object.c:101
VALUE rb_obj_class(VALUE obj)
Queries the class of an object.
Definition object.c:243
VALUE rb_inspect(VALUE obj)
Generates a human-readable textual representation of the given object.
Definition object.c:655
#define RB_OBJ_WRITE(old, slot, young)
Declaration of a "back" pointer.
Definition gc.h:603
VALUE rb_funcall(VALUE recv, ID mid, int n,...)
Calls a method.
Definition vm_eval.c:1117
Defines RBIMPL_HAS_BUILTIN.
VALUE rb_ary_new_from_values(long n, const VALUE *elts)
Identical to rb_ary_new_from_args(), except how objects are passed.
VALUE rb_ary_resurrect(VALUE ary)
I guess there is no use case of this function in extension libraries, but this is a routine identical...
VALUE rb_ary_new(void)
Allocates a new, empty array.
VALUE rb_ary_hidden_new(long capa)
Allocates a hidden (no class) empty array.
VALUE rb_ary_push(VALUE ary, VALUE elem)
Special case of rb_ary_cat() that it adds only one element.
VALUE rb_ary_freeze(VALUE obj)
Freeze an array, preventing further modifications.
VALUE rb_ary_entry(VALUE ary, long off)
Queries an element of an array.
VALUE rb_ary_join(VALUE ary, VALUE sep)
Recursively stringises the elements of the passed array, flattens that result, then joins the sequenc...
void rb_ary_store(VALUE ary, long key, VALUE val)
Destructively stores the passed value to the passed array's passed index.
static int rb_check_arity(int argc, int min, int max)
Ensures that the passed integer is in the passed range.
Definition error.h:284
VALUE rb_file_open_str(VALUE fname, const char *fmode)
Identical to rb_file_open(), except it takes the pathname as a Ruby's string instead of C's.
Definition io.c:7264
VALUE rb_io_close(VALUE io)
Closes the IO.
Definition io.c:5753
int rb_is_local_id(ID id)
Classifies the given ID, then sees if it is a local variable.
Definition symbol.c:1114
VALUE rb_obj_is_method(VALUE recv)
Queries if the given object is a method.
Definition proc.c:1675
VALUE rb_obj_is_proc(VALUE recv)
Queries if the given object is a proc.
Definition proc.c:120
VALUE rb_str_append(VALUE dst, VALUE src)
Identical to rb_str_buf_append(), except it converts the right hand side before concatenating.
Definition string.c:3757
#define rb_str_new(str, len)
Allocates an instance of rb_cString.
Definition string.h:1498
#define rb_exc_new_cstr(exc, str)
Identical to rb_exc_new(), except it assumes the passed pointer is a pointer to a C string.
Definition string.h:1670
VALUE rb_str_dup(VALUE str)
Duplicates a string.
Definition string.c:1956
VALUE rb_str_cat(VALUE dst, const char *src, long srclen)
Destructively appends the passed contents to the string.
Definition string.c:3525
VALUE rb_str_resurrect(VALUE str)
I guess there is no use case of this function in extension libraries, but this is a routine identical...
Definition string.c:1974
void rb_str_set_len(VALUE str, long len)
Overwrites the length of the string.
Definition string.c:3347
VALUE rb_str_inspect(VALUE str)
Generates a "readable" version of the receiver.
Definition string.c:7301
int rb_str_cmp(VALUE lhs, VALUE rhs)
Compares two strings, as in strcmp(3).
Definition string.c:4177
VALUE rb_str_concat(VALUE dst, VALUE src)
Identical to rb_str_append(), except it also accepts an integer as a codepoint.
Definition string.c:3994
#define rb_str_cat_cstr(buf, str)
Identical to rb_str_cat(), except it assumes the passed pointer is a pointer to a C string.
Definition string.h:1656
void rb_str_modify_expand(VALUE str, long capa)
Identical to rb_str_modify(), except it additionally expands the capacity of the receiver.
Definition string.c:2704
#define rb_str_new_cstr(str)
Identical to rb_str_new, except it assumes the passed pointer is a pointer to a C string.
Definition string.h:1514
VALUE rb_str_intern(VALUE str)
Identical to rb_to_symbol(), except it assumes the receiver being an instance of RString.
Definition symbol.c:942
VALUE rb_class_name(VALUE obj)
Queries the name of the given object's class.
Definition variable.c:497
int rb_respond_to(VALUE obj, ID mid)
Queries if the object responds to the method.
Definition vm_method.c:3298
void rb_undef_alloc_func(VALUE klass)
Deletes the allocator function of a class.
Definition vm_method.c:1591
VALUE rb_check_funcall(VALUE recv, ID mid, int argc, const VALUE *argv)
Identical to rb_funcallv(), except it returns RUBY_Qundef instead of raising rb_eNoMethodError.
Definition vm_eval.c:686
ID rb_check_id(volatile VALUE *namep)
Detects if the given name is already interned or not.
Definition symbol.c:1138
VALUE rb_sym2str(VALUE symbol)
Obtain a frozen string representation of a symbol (not including the leading colon).
Definition symbol.c:998
VALUE rb_io_path(VALUE io)
Returns the path for the given IO.
Definition io.c:2973
int len
Length of the buffer.
Definition io.h:8
VALUE rb_ractor_make_shareable(VALUE obj)
Destructively transforms the passed object so that multiple Ractors can share it.
Definition ractor.c:1405
#define RB_NUM2INT
Just another name of rb_num2int_inline.
Definition int.h:38
#define RB_INT2NUM
Just another name of rb_int2num_inline.
Definition int.h:37
VALUE rb_yield(VALUE val)
Yields the block.
Definition vm_eval.c:1372
#define RB_GC_GUARD(v)
Prevents premature destruction of local objects.
Definition memory.h:167
#define RB_ZALLOC(type)
Shorthand of RB_ZALLOC_N with n=1.
Definition memory.h:249
VALUE type(ANYARGS)
ANYARGS-ed function type.
void rb_hash_foreach(VALUE q, int_type *w, VALUE e)
Iteration over the given hash.
#define RARRAY_LEN
Just another name of rb_array_len.
Definition rarray.h:51
static int RARRAY_LENINT(VALUE ary)
Identical to rb_array_len(), except it differs for the return type.
Definition rarray.h:281
#define RARRAY_AREF(a, i)
Definition rarray.h:403
static VALUE RBASIC_CLASS(VALUE obj)
Queries the class of an object.
Definition rbasic.h:163
#define RHASH_SIZE(h)
Queries the size of the hash.
Definition rhash.h:69
#define StringValue(v)
Ensures that the parameter object is a String.
Definition rstring.h:66
static char * RSTRING_END(VALUE str)
Queries the end of the contents pointer of the string.
Definition rstring.h:442
#define StringValueCStr(v)
Identical to StringValuePtr, except it additionally checks for the contents for viability as a C stri...
Definition rstring.h:89
#define RUBY_TYPED_DEFAULT_FREE
This is a value you can set to rb_data_type_struct::dfree.
Definition rtypeddata.h:79
#define TypedData_Get_Struct(obj, type, data_type, sval)
Obtains a C struct from inside of a wrapper Ruby object.
Definition rtypeddata.h:515
#define TypedData_Wrap_Struct(klass, data_type, sval)
Converts sval, a pointer to your struct, into a Ruby object.
Definition rtypeddata.h:450
#define TypedData_Make_Struct(klass, type, data_type, sval)
Identical to TypedData_Wrap_Struct, except it allocates a new data region internally instead of takin...
Definition rtypeddata.h:497
#define FilePathValue(v)
Ensures that the parameter object is a path.
Definition ruby.h:90
#define RTEST
This is an old name of RB_TEST.
Definition iseq.h:280
const ID * segments
A null-terminated list of ids, used to represent a constant's path idNULL is used to represent the ::...
Definition vm_core.h:287
Definition vm_core.h:295
Definition vm_core.h:290
Definition iseq.h:251
A line and column in a string.
uint32_t column
The column number.
int32_t line
The line number.
This represents a range of bytes in the source string to which a node or token corresponds.
Definition ast.h:544
const uint8_t * start
A pointer to the start location of the range in the source.
Definition ast.h:546
const uint8_t * end
A pointer to the end location of the range in the source.
Definition ast.h:549
size_t size
The number of offsets in the list.
uint32_t node_id
The unique identifier for this node, which is deterministic based on the source.
Definition ast.h:1085
pm_location_t location
This is the location of the node in the source.
Definition ast.h:1091
int32_t line
The line within the file that the parse starts on.
Definition options.h:124
pm_scope_node_t node
The resulting scope node that will hold the generated AST.
pm_options_t options
The options that will be passed to the parser.
int32_t start_line
The line number at the start of the parse.
Definition parser.h:809
pm_newline_list_t newline_list
This is the list of newline offsets in the source file.
Definition parser.h:789
VALUE * script_lines
This is a pointer to the list of script lines for the ISEQs that will be associated with this scope n...
Definition method.h:63
This is the struct that holds necessary info for a struct.
Definition rtypeddata.h:203
struct rb_iseq_constant_body::@154 param
parameter information
Definition st.h:79
Definition vm_core.h:299
intptr_t SIGNED_VALUE
A signed integer type that has the same width with VALUE.
Definition value.h:63
uintptr_t ID
Type that represents a Ruby identifier such as a variable name.
Definition value.h:52
uintptr_t VALUE
Type that represents a Ruby object.
Definition value.h:40
static void Check_Type(VALUE v, enum ruby_value_type t)
Identical to RB_TYPE_P(), except it raises exceptions on predication failure.
Definition value_type.h:433
static bool RB_TYPE_P(VALUE obj, enum ruby_value_type t)
Queries if the given object is of given type.
Definition value_type.h:376
ruby_value_type
C-level type of an object.
Definition value_type.h:113