xref: /freebsd/contrib/libdiff/lib/diff_patience.c (revision 5ca8e32633c4ffbbcd6762e5888b6a4ba0708c6c)
1 /* Implementation of the Patience Diff algorithm invented by Bram Cohen:
2  * Divide a diff problem into smaller chunks by an LCS (Longest Common Sequence)
3  * of common-unique lines. */
4 /*
5  * Copyright (c) 2020 Neels Hofmeyr <neels@hofmeyr.de>
6  *
7  * Permission to use, copy, modify, and distribute this software for any
8  * purpose with or without fee is hereby granted, provided that the above
9  * copyright notice and this permission notice appear in all copies.
10  *
11  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
12  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
13  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
14  * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
15  * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
16  * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
17  * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
18  */
19 
20 #include <assert.h>
21 #include <errno.h>
22 #include <stdbool.h>
23 #include <stdint.h>
24 #include <stdio.h>
25 #include <stdlib.h>
26 
27 #include <arraylist.h>
28 #include <diff_main.h>
29 
30 #include "diff_internal.h"
31 #include "diff_debug.h"
32 
33 /* Algorithm to find unique lines:
34  * 0: stupidly iterate atoms
35  * 1: qsort
36  * 2: mergesort
37  */
38 #define UNIQUE_STRATEGY 1
39 
40 /* Per-atom state for the Patience Diff algorithm */
41 struct atom_patience {
42 #if UNIQUE_STRATEGY == 0
43 	bool unique_here;
44 #endif
45 	bool unique_in_both;
46 	struct diff_atom *pos_in_other;
47 	struct diff_atom *prev_stack;
48 	struct diff_range identical_lines;
49 };
50 
51 /* A diff_atom has a backpointer to the root diff_data. That points to the
52  * current diff_data, a possibly smaller section of the root. That current
53  * diff_data->algo_data is a pointer to an array of struct atom_patience. The
54  * atom's index in current diff_data gives the index in the atom_patience array.
55  */
56 #define PATIENCE(ATOM) \
57 	(((struct atom_patience*)((ATOM)->root->current->algo_data))\
58 	 [diff_atom_idx((ATOM)->root->current, ATOM)])
59 
60 #if UNIQUE_STRATEGY == 0
61 
62 /* Stupid iteration and comparison of all atoms */
63 static int
64 diff_atoms_mark_unique(struct diff_data *d, unsigned int *unique_count)
65 {
66 	struct diff_atom *i;
67 	unsigned int count = 0;
68 	diff_data_foreach_atom(i, d) {
69 		PATIENCE(i).unique_here = true;
70 		PATIENCE(i).unique_in_both = true;
71 		count++;
72 	}
73 	diff_data_foreach_atom(i, d) {
74 		struct diff_atom *j;
75 
76 		if (!PATIENCE(i).unique_here)
77 			continue;
78 
79 		diff_data_foreach_atom_from(i + 1, j, d) {
80 			bool same;
81 			int r = diff_atom_same(&same, i, j);
82 			if (r)
83 				return r;
84 			if (!same)
85 				continue;
86 			if (PATIENCE(i).unique_here) {
87 				PATIENCE(i).unique_here = false;
88 				PATIENCE(i).unique_in_both = false;
89 				count--;
90 			}
91 			PATIENCE(j).unique_here = false;
92 			PATIENCE(j).unique_in_both = false;
93 			count--;
94 		}
95 	}
96 	if (unique_count)
97 		*unique_count = count;
98 	return 0;
99 }
100 
101 /* Mark those lines as PATIENCE(atom).unique_in_both = true that appear exactly
102  * once in each side. */
103 static int
104 diff_atoms_mark_unique_in_both(struct diff_data *left, struct diff_data *right,
105 			       unsigned int *unique_in_both_count)
106 {
107 	/* Derive the final unique_in_both count without needing an explicit
108 	 * iteration. So this is just some optimiziation to save one iteration
109 	 * in the end. */
110 	unsigned int unique_in_both;
111 	int r;
112 
113 	r = diff_atoms_mark_unique(left, &unique_in_both);
114 	if (r)
115 		return r;
116 	r = diff_atoms_mark_unique(right, NULL);
117 	if (r)
118 		return r;
119 
120 	debug("unique_in_both %u\n", unique_in_both);
121 
122 	struct diff_atom *i;
123 	diff_data_foreach_atom(i, left) {
124 		if (!PATIENCE(i).unique_here)
125 			continue;
126 		struct diff_atom *j;
127 		int found_in_b = 0;
128 		diff_data_foreach_atom(j, right) {
129 			bool same;
130 			int r = diff_atom_same(&same, i, j);
131 			if (r)
132 				return r;
133 			if (!same)
134 				continue;
135 			if (!PATIENCE(j).unique_here) {
136 				found_in_b = 2; /* or more */
137 				break;
138 			} else {
139 				found_in_b = 1;
140 				PATIENCE(j).pos_in_other = i;
141 				PATIENCE(i).pos_in_other = j;
142 			}
143 		}
144 
145 		if (found_in_b == 0 || found_in_b > 1) {
146 			PATIENCE(i).unique_in_both = false;
147 			unique_in_both--;
148 			debug("unique_in_both %u  (%d) ", unique_in_both,
149 			      found_in_b);
150 			debug_dump_atom(left, NULL, i);
151 		}
152 	}
153 
154 	/* Still need to unmark right[*]->patience.unique_in_both for atoms that
155 	 * don't exist in left */
156 	diff_data_foreach_atom(i, right) {
157 		if (!PATIENCE(i).unique_here
158 		    || !PATIENCE(i).unique_in_both)
159 			continue;
160 		struct diff_atom *j;
161 		bool found_in_a = false;
162 		diff_data_foreach_atom(j, left) {
163 			bool same;
164 			int r;
165 			if (!PATIENCE(j).unique_in_both)
166 				continue;
167 			r = diff_atom_same(&same, i, j);
168 			if (r)
169 				return r;
170 			if (!same)
171 				continue;
172 			found_in_a = true;
173 			break;
174 		}
175 
176 		if (!found_in_a)
177 			PATIENCE(i).unique_in_both = false;
178 	}
179 
180 	if (unique_in_both_count)
181 		*unique_in_both_count = unique_in_both;
182 	return 0;
183 }
184 
185 #else /* UNIQUE_STRATEGY != 0 */
186 
187 /* Use an optimized sorting algorithm (qsort, mergesort) to find unique lines */
188 
189 static int diff_atoms_compar(const void *_a, const void *_b)
190 {
191 	const struct diff_atom *a = *(struct diff_atom**)_a;
192 	const struct diff_atom *b = *(struct diff_atom**)_b;
193 	int cmp;
194 	int rc = 0;
195 
196 	/* If there's been an error (e.g. I/O error) in a previous compar, we
197 	 * have no way to abort the sort but just report the rc and stop
198 	 * comparing. Make sure to catch errors on either side. If atoms are
199 	 * from more than one diff_data, make sure the error, if any, spreads
200 	 * to all of them, so we can cut short all future comparisons. */
201 	if (a->root->err)
202 		rc = a->root->err;
203 	if (b->root->err)
204 		rc = b->root->err;
205 	if (rc) {
206 		a->root->err = rc;
207 		b->root->err = rc;
208 		/* just return 'equal' to not swap more positions */
209 		return 0;
210 	}
211 
212 	/* Sort by the simplistic hash */
213 	if (a->hash < b->hash)
214 		return -1;
215 	if (a->hash > b->hash)
216 		return 1;
217 
218 	/* If hashes are the same, the lines may still differ. Do a full cmp. */
219 	rc = diff_atom_cmp(&cmp, a, b);
220 
221 	if (rc) {
222 		/* Mark the I/O error so that the caller can find out about it.
223 		 * For the case atoms are from more than one diff_data, mark in
224 		 * both. */
225 		a->root->err = rc;
226 		if (a->root != b->root)
227 			b->root->err = rc;
228 		return 0;
229 	}
230 
231 	return cmp;
232 }
233 
234 /* Sort an array of struct diff_atom* in-place. */
235 static int diff_atoms_sort(struct diff_atom *atoms[],
236 			   size_t atoms_count)
237 {
238 #if UNIQUE_STRATEGY == 1
239 	qsort(atoms, atoms_count, sizeof(struct diff_atom*), diff_atoms_compar);
240 #else
241 	mergesort(atoms, atoms_count, sizeof(struct diff_atom*),
242 		  diff_atoms_compar);
243 #endif
244 	return atoms[0]->root->err;
245 }
246 
247 static int
248 diff_atoms_mark_unique_in_both(struct diff_data *left, struct diff_data *right,
249 			       unsigned int *unique_in_both_count_p)
250 {
251 	struct diff_atom *a;
252 	struct diff_atom *b;
253 	struct diff_atom **all_atoms;
254 	unsigned int len = 0;
255 	unsigned int i;
256 	unsigned int unique_in_both_count = 0;
257 	int rc;
258 
259 	all_atoms = calloc(left->atoms.len + right->atoms.len,
260 	    sizeof(struct diff_atom *));
261 	if (all_atoms == NULL)
262 		return ENOMEM;
263 
264 	left->err = 0;
265 	right->err = 0;
266 	left->root->err = 0;
267 	right->root->err = 0;
268 	diff_data_foreach_atom(a, left) {
269 		all_atoms[len++] = a;
270 	}
271 	diff_data_foreach_atom(b, right) {
272 		all_atoms[len++] = b;
273 	}
274 
275 	rc = diff_atoms_sort(all_atoms, len);
276 	if (rc)
277 		goto free_and_exit;
278 
279 	/* Now we have a sorted array of atom pointers. All similar lines are
280 	 * adjacent. Walk through the array and mark those that are unique on
281 	 * each side, but exist once in both sources. */
282 	for (i = 0; i < len; i++) {
283 		bool same;
284 		unsigned int next_differing_i;
285 		unsigned int last_identical_i;
286 		unsigned int j;
287 		unsigned int count_first_side = 1;
288 		unsigned int count_other_side = 0;
289 		a = all_atoms[i];
290 		debug("a: ");
291 		debug_dump_atom(a->root, NULL, a);
292 
293 		/* Do as few diff_atom_cmp() as possible: first walk forward
294 		 * only using the cheap hash as indicator for differing atoms;
295 		 * then walk backwards until hitting an identical atom. */
296 		for (next_differing_i = i + 1; next_differing_i < len;
297 		     next_differing_i++) {
298 			b = all_atoms[next_differing_i];
299 			if (a->hash != b->hash)
300 				break;
301 		}
302 		for (last_identical_i = next_differing_i - 1;
303 		     last_identical_i > i;
304 		     last_identical_i--) {
305 			b = all_atoms[last_identical_i];
306 			rc = diff_atom_same(&same, a, b);
307 			if (rc)
308 				goto free_and_exit;
309 			if (same)
310 				break;
311 		}
312 		next_differing_i = last_identical_i + 1;
313 
314 		for (j = i+1; j < next_differing_i; j++) {
315 			b = all_atoms[j];
316 			/* A following atom is the same. See on which side the
317 			 * repetition counts. */
318 			if (a->root == b->root)
319 				count_first_side ++;
320 			else
321 				count_other_side ++;
322 			debug("b: ");
323 			debug_dump_atom(b->root, NULL, b);
324 			debug("   count_first_side=%d count_other_side=%d\n",
325 			      count_first_side, count_other_side);
326 		}
327 
328 		/* Counted a section of similar atoms, put the results back to
329 		 * the atoms. */
330 		if ((count_first_side == 1)
331 		    && (count_other_side == 1)) {
332 			b = all_atoms[i+1];
333 			PATIENCE(a).unique_in_both = true;
334 			PATIENCE(a).pos_in_other = b;
335 			PATIENCE(b).unique_in_both = true;
336 			PATIENCE(b).pos_in_other = a;
337 			unique_in_both_count++;
338 		}
339 
340 		/* j now points at the first atom after 'a' that is not
341 		 * identical to 'a'. j is always > i. */
342 		i = j - 1;
343 	}
344 	*unique_in_both_count_p = unique_in_both_count;
345 	rc = 0;
346 free_and_exit:
347 	free(all_atoms);
348 	return rc;
349 }
350 #endif /* UNIQUE_STRATEGY != 0 */
351 
352 /* binary search to find the stack to put this atom "card" on. */
353 static int
354 find_target_stack(struct diff_atom *atom,
355 		  struct diff_atom **patience_stacks,
356 		  unsigned int patience_stacks_count)
357 {
358 	unsigned int lo = 0;
359 	unsigned int hi = patience_stacks_count;
360 	while (lo < hi) {
361 		unsigned int mid = (lo + hi) >> 1;
362 
363 		if (PATIENCE(patience_stacks[mid]).pos_in_other
364 		    < PATIENCE(atom).pos_in_other)
365 			lo = mid + 1;
366 		else
367 			hi = mid;
368 	}
369 	return lo;
370 }
371 
372 /* Among the lines that appear exactly once in each side, find the longest
373  * streak that appear in both files in the same order (with other stuff allowed
374  * to interleave). Use patience sort for that, as in the Patience Diff
375  * algorithm.
376  * See https://bramcohen.livejournal.com/73318.html and, for a much more
377  * detailed explanation,
378  * https://blog.jcoglan.com/2017/09/19/the-patience-diff-algorithm/ */
379 int
380 diff_algo_patience(const struct diff_algo_config *algo_config,
381 		   struct diff_state *state)
382 {
383 	int rc;
384 	struct diff_data *left = &state->left;
385 	struct diff_data *right = &state->right;
386 	struct atom_patience *atom_patience_left =
387 		calloc(left->atoms.len, sizeof(struct atom_patience));
388 	struct atom_patience *atom_patience_right =
389 		calloc(right->atoms.len, sizeof(struct atom_patience));
390 	unsigned int unique_in_both_count;
391 	struct diff_atom **lcs = NULL;
392 
393 	debug("\n** %s\n", __func__);
394 
395 	left->root->current = left;
396 	right->root->current = right;
397 	left->algo_data = atom_patience_left;
398 	right->algo_data = atom_patience_right;
399 
400 	/* Find those lines that appear exactly once in 'left' and exactly once
401 	 * in 'right'. */
402 	rc = diff_atoms_mark_unique_in_both(left, right, &unique_in_both_count);
403 	if (rc)
404 		goto free_and_exit;
405 
406 	debug("unique_in_both_count %u\n", unique_in_both_count);
407 	debug("left:\n");
408 	debug_dump(left);
409 	debug("right:\n");
410 	debug_dump(right);
411 
412 	if (!unique_in_both_count) {
413 		/* Cannot apply Patience, tell the caller to use fallback_algo
414 		 * instead. */
415 		rc = DIFF_RC_USE_DIFF_ALGO_FALLBACK;
416 		goto free_and_exit;
417 	}
418 
419 	rc = ENOMEM;
420 
421 	/* An array of Longest Common Sequence is the result of the below
422 	 * subscope: */
423 	unsigned int lcs_count = 0;
424 	struct diff_atom *lcs_tail = NULL;
425 
426 	{
427 		/* This subscope marks the lifetime of the atom_pointers
428 		 * allocation */
429 
430 		/* One chunk of storage for atom pointers */
431 		struct diff_atom **atom_pointers;
432 		atom_pointers = recallocarray(NULL, 0, unique_in_both_count * 2,
433 					      sizeof(struct diff_atom*));
434 		if (atom_pointers == NULL)
435 			return ENOMEM;
436 		/* Half for the list of atoms that still need to be put on
437 		 * stacks */
438 		struct diff_atom **uniques = atom_pointers;
439 
440 		/* Half for the patience sort state's "card stacks" -- we
441 		 * remember only each stack's topmost "card" */
442 		struct diff_atom **patience_stacks;
443 		patience_stacks = atom_pointers + unique_in_both_count;
444 		unsigned int patience_stacks_count = 0;
445 
446 		/* Take all common, unique items from 'left' ... */
447 
448 		struct diff_atom *atom;
449 		struct diff_atom **uniques_end = uniques;
450 		diff_data_foreach_atom(atom, left) {
451 			if (!PATIENCE(atom).unique_in_both)
452 				continue;
453 			*uniques_end = atom;
454 			uniques_end++;
455 		}
456 
457 		/* ...and sort them to the order found in 'right'.
458 		 * The idea is to find the leftmost stack that has a higher line
459 		 * number and add it to the stack's top.
460 		 * If there is no such stack, open a new one on the right. The
461 		 * line number is derived from the atom*, which are array items
462 		 * and hence reflect the relative position in the source file.
463 		 * So we got the common-uniques from 'left' and sort them
464 		 * according to PATIENCE(atom).pos_in_other. */
465 		unsigned int i;
466 		for (i = 0; i < unique_in_both_count; i++) {
467 			atom = uniques[i];
468 			unsigned int target_stack;
469 			target_stack = find_target_stack(atom, patience_stacks,
470 							 patience_stacks_count);
471 			assert(target_stack <= patience_stacks_count);
472 			patience_stacks[target_stack] = atom;
473 			if (target_stack == patience_stacks_count)
474 				patience_stacks_count++;
475 
476 			/* Record a back reference to the next stack on the
477 			 * left, which will form the final longest sequence
478 			 * later. */
479 			PATIENCE(atom).prev_stack = target_stack ?
480 				patience_stacks[target_stack - 1] : NULL;
481 
482 			{
483 				int xx;
484 				for (xx = 0; xx < patience_stacks_count; xx++) {
485 					debug(" %s%d",
486 					      (xx == target_stack) ? ">" : "",
487 					      diff_atom_idx(right,
488 							    PATIENCE(patience_stacks[xx]).pos_in_other));
489 				}
490 				debug("\n");
491 			}
492 		}
493 
494 		/* backtrace through prev_stack references to form the final
495 		 * longest common sequence */
496 		lcs_tail = patience_stacks[patience_stacks_count - 1];
497 		lcs_count = patience_stacks_count;
498 
499 		/* uniques and patience_stacks are no longer needed.
500 		 * Backpointers are in PATIENCE(atom).prev_stack */
501 		free(atom_pointers);
502 	}
503 
504 	lcs = recallocarray(NULL, 0, lcs_count, sizeof(struct diff_atom*));
505 	struct diff_atom **lcs_backtrace_pos = &lcs[lcs_count - 1];
506 	struct diff_atom *atom;
507 	for (atom = lcs_tail; atom; atom = PATIENCE(atom).prev_stack, lcs_backtrace_pos--) {
508 		assert(lcs_backtrace_pos >= lcs);
509 		*lcs_backtrace_pos = atom;
510 	}
511 
512 	unsigned int i;
513 	if (DEBUG) {
514 		debug("\npatience LCS:\n");
515 		for (i = 0; i < lcs_count; i++) {
516 			debug("\n L "); debug_dump_atom(left, right, lcs[i]);
517 			debug(" R "); debug_dump_atom(right, left,
518 						      PATIENCE(lcs[i]).pos_in_other);
519 		}
520 	}
521 
522 
523 	/* TODO: For each common-unique line found (now listed in lcs), swallow
524 	 * lines upwards and downwards that are identical on each side. Requires
525 	 * a way to represent atoms being glued to adjacent atoms. */
526 
527 	debug("\ntraverse LCS, possibly recursing:\n");
528 
529 	/* Now we have pinned positions in both files at which it makes sense to
530 	 * divide the diff problem into smaller chunks. Go into the next round:
531 	 * look at each section in turn, trying to again find common-unique
532 	 * lines in those smaller sections. As soon as no more are found, the
533 	 * remaining smaller sections are solved by Myers. */
534 	/* left_pos and right_pos are indexes in left/right->atoms.head until
535 	 * which the atoms are already handled (added to result chunks). */
536 	unsigned int left_pos = 0;
537 	unsigned int right_pos = 0;
538 	for (i = 0; i <= lcs_count; i++) {
539 		struct diff_atom *atom;
540 		struct diff_atom *atom_r;
541 		/* left_idx and right_idx are indexes of the start of this
542 		 * section of identical lines on both sides.
543 		 * left_pos marks the index of the first still unhandled line,
544 		 * left_idx is the start of an identical section some way
545 		 * further down, and this loop adds an unsolved chunk of
546 		 * [left_pos..left_idx[ and a solved chunk of
547 		 * [left_idx..identical_lines.end[. */
548 		unsigned int left_idx;
549 		unsigned int right_idx;
550 
551 		debug("iteration %u of %u  left_pos %u  right_pos %u\n",
552 		      i, lcs_count, left_pos, right_pos);
553 
554 		if (i < lcs_count) {
555 			atom = lcs[i];
556 			atom_r = PATIENCE(atom).pos_in_other;
557 			debug("lcs[%u] = left[%u] = right[%u]\n", i,
558 			      diff_atom_idx(left, atom), diff_atom_idx(right, atom_r));
559 			left_idx = diff_atom_idx(left, atom);
560 			right_idx = diff_atom_idx(right, atom_r);
561 		} else {
562 			/* There are no more identical lines until the end of
563 			 * left and right. */
564 			atom = NULL;
565 			atom_r = NULL;
566 			left_idx = left->atoms.len;
567 			right_idx = right->atoms.len;
568 		}
569 
570 		/* 'atom' (if not NULL) now marks an atom that matches on both
571 		 * sides according to patience-diff (a common-unique identical
572 		 * atom in both files).
573 		 * Handle the section before and the atom itself; the section
574 		 * after will be handled by the next loop iteration -- note that
575 		 * i loops to last element + 1 ("i <= lcs_count"), so that there
576 		 * will be another final iteration to pick up the last remaining
577 		 * items after the last LCS atom.
578 		 */
579 
580 		debug("iteration %u  left_pos %u  left_idx %u"
581 		      "  right_pos %u  right_idx %u\n",
582 		      i, left_pos, left_idx, right_pos, right_idx);
583 
584 		/* Section before the matching atom */
585 		struct diff_atom *left_atom = &left->atoms.head[left_pos];
586 		unsigned int left_section_len = left_idx - left_pos;
587 
588 		struct diff_atom *right_atom = &(right->atoms.head[right_pos]);
589 		unsigned int right_section_len = right_idx - right_pos;
590 
591 		if (left_section_len && right_section_len) {
592 			/* Record an unsolved chunk, the caller will apply
593 			 * inner_algo() on this chunk. */
594 			if (!diff_state_add_chunk(state, false,
595 						  left_atom, left_section_len,
596 						  right_atom,
597 						  right_section_len))
598 				goto free_and_exit;
599 		} else if (left_section_len && !right_section_len) {
600 			/* Only left atoms and none on the right, they form a
601 			 * "minus" chunk, then. */
602 			if (!diff_state_add_chunk(state, true,
603 						  left_atom, left_section_len,
604 						  right_atom, 0))
605 				goto free_and_exit;
606 		} else if (!left_section_len && right_section_len) {
607 			/* No left atoms, only atoms on the right, they form a
608 			 * "plus" chunk, then. */
609 			if (!diff_state_add_chunk(state, true,
610 						  left_atom, 0,
611 						  right_atom, right_section_len))
612 				goto free_and_exit;
613 		}
614 		/* else: left_section_len == 0 and right_section_len == 0, i.e.
615 		 * nothing here. */
616 
617 		/* The atom found to match on both sides forms a chunk of equals
618 		 * on each side. In the very last iteration of this loop, there
619 		 * is no matching atom, we were just cleaning out the remaining
620 		 * lines. */
621 		if (atom) {
622 			void *ok;
623 			ok = diff_state_add_chunk(state, true,
624 						  atom, 1,
625 						  PATIENCE(atom).pos_in_other, 1);
626 			if (!ok)
627 				goto free_and_exit;
628 		}
629 		left_pos = left_idx + 1;
630 		right_pos = right_idx + 1;
631 		debug("end of iteration %u  left_pos %u  left_idx %u"
632 		      "  right_pos %u  right_idx %u\n",
633 		      i, left_pos, left_idx, right_pos, right_idx);
634 	}
635 	debug("** END %s\n", __func__);
636 
637 	rc = DIFF_RC_OK;
638 
639 free_and_exit:
640 	left->root->current = NULL;
641 	right->root->current = NULL;
642 	free(atom_patience_left);
643 	free(atom_patience_right);
644 	if (lcs)
645 		free(lcs);
646 	return rc;
647 }
648