Merge branch 'jc/show-merge'
[git/git.git] / builtin-apply.c
1 /*
2 * apply.c
3 *
4 * Copyright (C) Linus Torvalds, 2005
5 *
6 * This applies patches on top of some (arbitrary) version of the SCM.
7 *
8 */
9 #include <fnmatch.h>
10 #include "cache.h"
11 #include "cache-tree.h"
12 #include "quote.h"
13 #include "blob.h"
14 #include "delta.h"
15 #include "builtin.h"
16
17 /*
18 * --check turns on checking that the working tree matches the
19 * files that are being modified, but doesn't apply the patch
20 * --stat does just a diffstat, and doesn't actually apply
21 * --numstat does numeric diffstat, and doesn't actually apply
22 * --index-info shows the old and new index info for paths if available.
23 * --index updates the cache as well.
24 * --cached updates only the cache without ever touching the working tree.
25 */
26 static const char *prefix;
27 static int prefix_length = -1;
28 static int newfd = -1;
29
30 static int p_value = 1;
31 static int allow_binary_replacement = 0;
32 static int check_index = 0;
33 static int write_index = 0;
34 static int cached = 0;
35 static int diffstat = 0;
36 static int numstat = 0;
37 static int summary = 0;
38 static int check = 0;
39 static int apply = 1;
40 static int no_add = 0;
41 static int show_index_info = 0;
42 static int line_termination = '\n';
43 static unsigned long p_context = -1;
44 static const char apply_usage[] =
45 "git-apply [--stat] [--numstat] [--summary] [--check] [--index] [--cached] [--apply] [--no-add] [--index-info] [--allow-binary-replacement] [-z] [-pNUM] [-CNUM] [--whitespace=<nowarn|warn|error|error-all|strip>] <patch>...";
46
47 static enum whitespace_eol {
48 nowarn_whitespace,
49 warn_on_whitespace,
50 error_on_whitespace,
51 strip_whitespace,
52 } new_whitespace = warn_on_whitespace;
53 static int whitespace_error = 0;
54 static int squelch_whitespace_errors = 5;
55 static int applied_after_stripping = 0;
56 static const char *patch_input_file = NULL;
57
58 static void parse_whitespace_option(const char *option)
59 {
60 if (!option) {
61 new_whitespace = warn_on_whitespace;
62 return;
63 }
64 if (!strcmp(option, "warn")) {
65 new_whitespace = warn_on_whitespace;
66 return;
67 }
68 if (!strcmp(option, "nowarn")) {
69 new_whitespace = nowarn_whitespace;
70 return;
71 }
72 if (!strcmp(option, "error")) {
73 new_whitespace = error_on_whitespace;
74 return;
75 }
76 if (!strcmp(option, "error-all")) {
77 new_whitespace = error_on_whitespace;
78 squelch_whitespace_errors = 0;
79 return;
80 }
81 if (!strcmp(option, "strip")) {
82 new_whitespace = strip_whitespace;
83 return;
84 }
85 die("unrecognized whitespace option '%s'", option);
86 }
87
88 static void set_default_whitespace_mode(const char *whitespace_option)
89 {
90 if (!whitespace_option && !apply_default_whitespace) {
91 new_whitespace = (apply
92 ? warn_on_whitespace
93 : nowarn_whitespace);
94 }
95 }
96
97 /*
98 * For "diff-stat" like behaviour, we keep track of the biggest change
99 * we've seen, and the longest filename. That allows us to do simple
100 * scaling.
101 */
102 static int max_change, max_len;
103
104 /*
105 * Various "current state", notably line numbers and what
106 * file (and how) we're patching right now.. The "is_xxxx"
107 * things are flags, where -1 means "don't know yet".
108 */
109 static int linenr = 1;
110
111 struct fragment {
112 unsigned long leading, trailing;
113 unsigned long oldpos, oldlines;
114 unsigned long newpos, newlines;
115 const char *patch;
116 int size;
117 struct fragment *next;
118 };
119
120 struct patch {
121 char *new_name, *old_name, *def_name;
122 unsigned int old_mode, new_mode;
123 int is_rename, is_copy, is_new, is_delete, is_binary;
124 #define BINARY_DELTA_DEFLATED 1
125 #define BINARY_LITERAL_DEFLATED 2
126 unsigned long deflate_origlen;
127 int lines_added, lines_deleted;
128 int score;
129 int inaccurate_eof:1;
130 struct fragment *fragments;
131 char *result;
132 unsigned long resultsize;
133 char old_sha1_prefix[41];
134 char new_sha1_prefix[41];
135 struct patch *next;
136 };
137
138 #define CHUNKSIZE (8192)
139 #define SLOP (16)
140
141 static void *read_patch_file(int fd, unsigned long *sizep)
142 {
143 unsigned long size = 0, alloc = CHUNKSIZE;
144 void *buffer = xmalloc(alloc);
145
146 for (;;) {
147 int nr = alloc - size;
148 if (nr < 1024) {
149 alloc += CHUNKSIZE;
150 buffer = xrealloc(buffer, alloc);
151 nr = alloc - size;
152 }
153 nr = xread(fd, (char *) buffer + size, nr);
154 if (!nr)
155 break;
156 if (nr < 0)
157 die("git-apply: read returned %s", strerror(errno));
158 size += nr;
159 }
160 *sizep = size;
161
162 /*
163 * Make sure that we have some slop in the buffer
164 * so that we can do speculative "memcmp" etc, and
165 * see to it that it is NUL-filled.
166 */
167 if (alloc < size + SLOP)
168 buffer = xrealloc(buffer, size + SLOP);
169 memset((char *) buffer + size, 0, SLOP);
170 return buffer;
171 }
172
173 static unsigned long linelen(const char *buffer, unsigned long size)
174 {
175 unsigned long len = 0;
176 while (size--) {
177 len++;
178 if (*buffer++ == '\n')
179 break;
180 }
181 return len;
182 }
183
184 static int is_dev_null(const char *str)
185 {
186 return !memcmp("/dev/null", str, 9) && isspace(str[9]);
187 }
188
189 #define TERM_SPACE 1
190 #define TERM_TAB 2
191
192 static int name_terminate(const char *name, int namelen, int c, int terminate)
193 {
194 if (c == ' ' && !(terminate & TERM_SPACE))
195 return 0;
196 if (c == '\t' && !(terminate & TERM_TAB))
197 return 0;
198
199 return 1;
200 }
201
202 static char * find_name(const char *line, char *def, int p_value, int terminate)
203 {
204 int len;
205 const char *start = line;
206 char *name;
207
208 if (*line == '"') {
209 /* Proposed "new-style" GNU patch/diff format; see
210 * http://marc.theaimsgroup.com/?l=git&m=112927316408690&w=2
211 */
212 name = unquote_c_style(line, NULL);
213 if (name) {
214 char *cp = name;
215 while (p_value) {
216 cp = strchr(name, '/');
217 if (!cp)
218 break;
219 cp++;
220 p_value--;
221 }
222 if (cp) {
223 /* name can later be freed, so we need
224 * to memmove, not just return cp
225 */
226 memmove(name, cp, strlen(cp) + 1);
227 free(def);
228 return name;
229 }
230 else {
231 free(name);
232 name = NULL;
233 }
234 }
235 }
236
237 for (;;) {
238 char c = *line;
239
240 if (isspace(c)) {
241 if (c == '\n')
242 break;
243 if (name_terminate(start, line-start, c, terminate))
244 break;
245 }
246 line++;
247 if (c == '/' && !--p_value)
248 start = line;
249 }
250 if (!start)
251 return def;
252 len = line - start;
253 if (!len)
254 return def;
255
256 /*
257 * Generally we prefer the shorter name, especially
258 * if the other one is just a variation of that with
259 * something else tacked on to the end (ie "file.orig"
260 * or "file~").
261 */
262 if (def) {
263 int deflen = strlen(def);
264 if (deflen < len && !strncmp(start, def, deflen))
265 return def;
266 }
267
268 name = xmalloc(len + 1);
269 memcpy(name, start, len);
270 name[len] = 0;
271 free(def);
272 return name;
273 }
274
275 /*
276 * Get the name etc info from the --/+++ lines of a traditional patch header
277 *
278 * NOTE! This hardcodes "-p1" behaviour in filename detection.
279 *
280 * FIXME! The end-of-filename heuristics are kind of screwy. For existing
281 * files, we can happily check the index for a match, but for creating a
282 * new file we should try to match whatever "patch" does. I have no idea.
283 */
284 static void parse_traditional_patch(const char *first, const char *second, struct patch *patch)
285 {
286 char *name;
287
288 first += 4; /* skip "--- " */
289 second += 4; /* skip "+++ " */
290 if (is_dev_null(first)) {
291 patch->is_new = 1;
292 patch->is_delete = 0;
293 name = find_name(second, NULL, p_value, TERM_SPACE | TERM_TAB);
294 patch->new_name = name;
295 } else if (is_dev_null(second)) {
296 patch->is_new = 0;
297 patch->is_delete = 1;
298 name = find_name(first, NULL, p_value, TERM_SPACE | TERM_TAB);
299 patch->old_name = name;
300 } else {
301 name = find_name(first, NULL, p_value, TERM_SPACE | TERM_TAB);
302 name = find_name(second, name, p_value, TERM_SPACE | TERM_TAB);
303 patch->old_name = patch->new_name = name;
304 }
305 if (!name)
306 die("unable to find filename in patch at line %d", linenr);
307 }
308
309 static int gitdiff_hdrend(const char *line, struct patch *patch)
310 {
311 return -1;
312 }
313
314 /*
315 * We're anal about diff header consistency, to make
316 * sure that we don't end up having strange ambiguous
317 * patches floating around.
318 *
319 * As a result, gitdiff_{old|new}name() will check
320 * their names against any previous information, just
321 * to make sure..
322 */
323 static char *gitdiff_verify_name(const char *line, int isnull, char *orig_name, const char *oldnew)
324 {
325 if (!orig_name && !isnull)
326 return find_name(line, NULL, 1, 0);
327
328 if (orig_name) {
329 int len;
330 const char *name;
331 char *another;
332 name = orig_name;
333 len = strlen(name);
334 if (isnull)
335 die("git-apply: bad git-diff - expected /dev/null, got %s on line %d", name, linenr);
336 another = find_name(line, NULL, 1, 0);
337 if (!another || memcmp(another, name, len))
338 die("git-apply: bad git-diff - inconsistent %s filename on line %d", oldnew, linenr);
339 free(another);
340 return orig_name;
341 }
342 else {
343 /* expect "/dev/null" */
344 if (memcmp("/dev/null", line, 9) || line[9] != '\n')
345 die("git-apply: bad git-diff - expected /dev/null on line %d", linenr);
346 return NULL;
347 }
348 }
349
350 static int gitdiff_oldname(const char *line, struct patch *patch)
351 {
352 patch->old_name = gitdiff_verify_name(line, patch->is_new, patch->old_name, "old");
353 return 0;
354 }
355
356 static int gitdiff_newname(const char *line, struct patch *patch)
357 {
358 patch->new_name = gitdiff_verify_name(line, patch->is_delete, patch->new_name, "new");
359 return 0;
360 }
361
362 static int gitdiff_oldmode(const char *line, struct patch *patch)
363 {
364 patch->old_mode = strtoul(line, NULL, 8);
365 return 0;
366 }
367
368 static int gitdiff_newmode(const char *line, struct patch *patch)
369 {
370 patch->new_mode = strtoul(line, NULL, 8);
371 return 0;
372 }
373
374 static int gitdiff_delete(const char *line, struct patch *patch)
375 {
376 patch->is_delete = 1;
377 patch->old_name = patch->def_name;
378 return gitdiff_oldmode(line, patch);
379 }
380
381 static int gitdiff_newfile(const char *line, struct patch *patch)
382 {
383 patch->is_new = 1;
384 patch->new_name = patch->def_name;
385 return gitdiff_newmode(line, patch);
386 }
387
388 static int gitdiff_copysrc(const char *line, struct patch *patch)
389 {
390 patch->is_copy = 1;
391 patch->old_name = find_name(line, NULL, 0, 0);
392 return 0;
393 }
394
395 static int gitdiff_copydst(const char *line, struct patch *patch)
396 {
397 patch->is_copy = 1;
398 patch->new_name = find_name(line, NULL, 0, 0);
399 return 0;
400 }
401
402 static int gitdiff_renamesrc(const char *line, struct patch *patch)
403 {
404 patch->is_rename = 1;
405 patch->old_name = find_name(line, NULL, 0, 0);
406 return 0;
407 }
408
409 static int gitdiff_renamedst(const char *line, struct patch *patch)
410 {
411 patch->is_rename = 1;
412 patch->new_name = find_name(line, NULL, 0, 0);
413 return 0;
414 }
415
416 static int gitdiff_similarity(const char *line, struct patch *patch)
417 {
418 if ((patch->score = strtoul(line, NULL, 10)) == ULONG_MAX)
419 patch->score = 0;
420 return 0;
421 }
422
423 static int gitdiff_dissimilarity(const char *line, struct patch *patch)
424 {
425 if ((patch->score = strtoul(line, NULL, 10)) == ULONG_MAX)
426 patch->score = 0;
427 return 0;
428 }
429
430 static int gitdiff_index(const char *line, struct patch *patch)
431 {
432 /* index line is N hexadecimal, "..", N hexadecimal,
433 * and optional space with octal mode.
434 */
435 const char *ptr, *eol;
436 int len;
437
438 ptr = strchr(line, '.');
439 if (!ptr || ptr[1] != '.' || 40 < ptr - line)
440 return 0;
441 len = ptr - line;
442 memcpy(patch->old_sha1_prefix, line, len);
443 patch->old_sha1_prefix[len] = 0;
444
445 line = ptr + 2;
446 ptr = strchr(line, ' ');
447 eol = strchr(line, '\n');
448
449 if (!ptr || eol < ptr)
450 ptr = eol;
451 len = ptr - line;
452
453 if (40 < len)
454 return 0;
455 memcpy(patch->new_sha1_prefix, line, len);
456 patch->new_sha1_prefix[len] = 0;
457 if (*ptr == ' ')
458 patch->new_mode = patch->old_mode = strtoul(ptr+1, NULL, 8);
459 return 0;
460 }
461
462 /*
463 * This is normal for a diff that doesn't change anything: we'll fall through
464 * into the next diff. Tell the parser to break out.
465 */
466 static int gitdiff_unrecognized(const char *line, struct patch *patch)
467 {
468 return -1;
469 }
470
471 static const char *stop_at_slash(const char *line, int llen)
472 {
473 int i;
474
475 for (i = 0; i < llen; i++) {
476 int ch = line[i];
477 if (ch == '/')
478 return line + i;
479 }
480 return NULL;
481 }
482
483 /* This is to extract the same name that appears on "diff --git"
484 * line. We do not find and return anything if it is a rename
485 * patch, and it is OK because we will find the name elsewhere.
486 * We need to reliably find name only when it is mode-change only,
487 * creation or deletion of an empty file. In any of these cases,
488 * both sides are the same name under a/ and b/ respectively.
489 */
490 static char *git_header_name(char *line, int llen)
491 {
492 int len;
493 const char *name;
494 const char *second = NULL;
495
496 line += strlen("diff --git ");
497 llen -= strlen("diff --git ");
498
499 if (*line == '"') {
500 const char *cp;
501 char *first = unquote_c_style(line, &second);
502 if (!first)
503 return NULL;
504
505 /* advance to the first slash */
506 cp = stop_at_slash(first, strlen(first));
507 if (!cp || cp == first) {
508 /* we do not accept absolute paths */
509 free_first_and_fail:
510 free(first);
511 return NULL;
512 }
513 len = strlen(cp+1);
514 memmove(first, cp+1, len+1); /* including NUL */
515
516 /* second points at one past closing dq of name.
517 * find the second name.
518 */
519 while ((second < line + llen) && isspace(*second))
520 second++;
521
522 if (line + llen <= second)
523 goto free_first_and_fail;
524 if (*second == '"') {
525 char *sp = unquote_c_style(second, NULL);
526 if (!sp)
527 goto free_first_and_fail;
528 cp = stop_at_slash(sp, strlen(sp));
529 if (!cp || cp == sp) {
530 free_both_and_fail:
531 free(sp);
532 goto free_first_and_fail;
533 }
534 /* They must match, otherwise ignore */
535 if (strcmp(cp+1, first))
536 goto free_both_and_fail;
537 free(sp);
538 return first;
539 }
540
541 /* unquoted second */
542 cp = stop_at_slash(second, line + llen - second);
543 if (!cp || cp == second)
544 goto free_first_and_fail;
545 cp++;
546 if (line + llen - cp != len + 1 ||
547 memcmp(first, cp, len))
548 goto free_first_and_fail;
549 return first;
550 }
551
552 /* unquoted first name */
553 name = stop_at_slash(line, llen);
554 if (!name || name == line)
555 return NULL;
556
557 name++;
558
559 /* since the first name is unquoted, a dq if exists must be
560 * the beginning of the second name.
561 */
562 for (second = name; second < line + llen; second++) {
563 if (*second == '"') {
564 const char *cp = second;
565 const char *np;
566 char *sp = unquote_c_style(second, NULL);
567
568 if (!sp)
569 return NULL;
570 np = stop_at_slash(sp, strlen(sp));
571 if (!np || np == sp) {
572 free_second_and_fail:
573 free(sp);
574 return NULL;
575 }
576 np++;
577 len = strlen(np);
578 if (len < cp - name &&
579 !strncmp(np, name, len) &&
580 isspace(name[len])) {
581 /* Good */
582 memmove(sp, np, len + 1);
583 return sp;
584 }
585 goto free_second_and_fail;
586 }
587 }
588
589 /*
590 * Accept a name only if it shows up twice, exactly the same
591 * form.
592 */
593 for (len = 0 ; ; len++) {
594 char c = name[len];
595
596 switch (c) {
597 default:
598 continue;
599 case '\n':
600 return NULL;
601 case '\t': case ' ':
602 second = name+len;
603 for (;;) {
604 char c = *second++;
605 if (c == '\n')
606 return NULL;
607 if (c == '/')
608 break;
609 }
610 if (second[len] == '\n' && !memcmp(name, second, len)) {
611 char *ret = xmalloc(len + 1);
612 memcpy(ret, name, len);
613 ret[len] = 0;
614 return ret;
615 }
616 }
617 }
618 return NULL;
619 }
620
621 /* Verify that we recognize the lines following a git header */
622 static int parse_git_header(char *line, int len, unsigned int size, struct patch *patch)
623 {
624 unsigned long offset;
625
626 /* A git diff has explicit new/delete information, so we don't guess */
627 patch->is_new = 0;
628 patch->is_delete = 0;
629
630 /*
631 * Some things may not have the old name in the
632 * rest of the headers anywhere (pure mode changes,
633 * or removing or adding empty files), so we get
634 * the default name from the header.
635 */
636 patch->def_name = git_header_name(line, len);
637
638 line += len;
639 size -= len;
640 linenr++;
641 for (offset = len ; size > 0 ; offset += len, size -= len, line += len, linenr++) {
642 static const struct opentry {
643 const char *str;
644 int (*fn)(const char *, struct patch *);
645 } optable[] = {
646 { "@@ -", gitdiff_hdrend },
647 { "--- ", gitdiff_oldname },
648 { "+++ ", gitdiff_newname },
649 { "old mode ", gitdiff_oldmode },
650 { "new mode ", gitdiff_newmode },
651 { "deleted file mode ", gitdiff_delete },
652 { "new file mode ", gitdiff_newfile },
653 { "copy from ", gitdiff_copysrc },
654 { "copy to ", gitdiff_copydst },
655 { "rename old ", gitdiff_renamesrc },
656 { "rename new ", gitdiff_renamedst },
657 { "rename from ", gitdiff_renamesrc },
658 { "rename to ", gitdiff_renamedst },
659 { "similarity index ", gitdiff_similarity },
660 { "dissimilarity index ", gitdiff_dissimilarity },
661 { "index ", gitdiff_index },
662 { "", gitdiff_unrecognized },
663 };
664 int i;
665
666 len = linelen(line, size);
667 if (!len || line[len-1] != '\n')
668 break;
669 for (i = 0; i < ARRAY_SIZE(optable); i++) {
670 const struct opentry *p = optable + i;
671 int oplen = strlen(p->str);
672 if (len < oplen || memcmp(p->str, line, oplen))
673 continue;
674 if (p->fn(line + oplen, patch) < 0)
675 return offset;
676 break;
677 }
678 }
679
680 return offset;
681 }
682
683 static int parse_num(const char *line, unsigned long *p)
684 {
685 char *ptr;
686
687 if (!isdigit(*line))
688 return 0;
689 *p = strtoul(line, &ptr, 10);
690 return ptr - line;
691 }
692
693 static int parse_range(const char *line, int len, int offset, const char *expect,
694 unsigned long *p1, unsigned long *p2)
695 {
696 int digits, ex;
697
698 if (offset < 0 || offset >= len)
699 return -1;
700 line += offset;
701 len -= offset;
702
703 digits = parse_num(line, p1);
704 if (!digits)
705 return -1;
706
707 offset += digits;
708 line += digits;
709 len -= digits;
710
711 *p2 = 1;
712 if (*line == ',') {
713 digits = parse_num(line+1, p2);
714 if (!digits)
715 return -1;
716
717 offset += digits+1;
718 line += digits+1;
719 len -= digits+1;
720 }
721
722 ex = strlen(expect);
723 if (ex > len)
724 return -1;
725 if (memcmp(line, expect, ex))
726 return -1;
727
728 return offset + ex;
729 }
730
731 /*
732 * Parse a unified diff fragment header of the
733 * form "@@ -a,b +c,d @@"
734 */
735 static int parse_fragment_header(char *line, int len, struct fragment *fragment)
736 {
737 int offset;
738
739 if (!len || line[len-1] != '\n')
740 return -1;
741
742 /* Figure out the number of lines in a fragment */
743 offset = parse_range(line, len, 4, " +", &fragment->oldpos, &fragment->oldlines);
744 offset = parse_range(line, len, offset, " @@", &fragment->newpos, &fragment->newlines);
745
746 return offset;
747 }
748
749 static int find_header(char *line, unsigned long size, int *hdrsize, struct patch *patch)
750 {
751 unsigned long offset, len;
752
753 patch->is_rename = patch->is_copy = 0;
754 patch->is_new = patch->is_delete = -1;
755 patch->old_mode = patch->new_mode = 0;
756 patch->old_name = patch->new_name = NULL;
757 for (offset = 0; size > 0; offset += len, size -= len, line += len, linenr++) {
758 unsigned long nextlen;
759
760 len = linelen(line, size);
761 if (!len)
762 break;
763
764 /* Testing this early allows us to take a few shortcuts.. */
765 if (len < 6)
766 continue;
767
768 /*
769 * Make sure we don't find any unconnected patch fragments.
770 * That's a sign that we didn't find a header, and that a
771 * patch has become corrupted/broken up.
772 */
773 if (!memcmp("@@ -", line, 4)) {
774 struct fragment dummy;
775 if (parse_fragment_header(line, len, &dummy) < 0)
776 continue;
777 error("patch fragment without header at line %d: %.*s", linenr, (int)len-1, line);
778 }
779
780 if (size < len + 6)
781 break;
782
783 /*
784 * Git patch? It might not have a real patch, just a rename
785 * or mode change, so we handle that specially
786 */
787 if (!memcmp("diff --git ", line, 11)) {
788 int git_hdr_len = parse_git_header(line, len, size, patch);
789 if (git_hdr_len <= len)
790 continue;
791 if (!patch->old_name && !patch->new_name) {
792 if (!patch->def_name)
793 die("git diff header lacks filename information (line %d)", linenr);
794 patch->old_name = patch->new_name = patch->def_name;
795 }
796 *hdrsize = git_hdr_len;
797 return offset;
798 }
799
800 /** --- followed by +++ ? */
801 if (memcmp("--- ", line, 4) || memcmp("+++ ", line + len, 4))
802 continue;
803
804 /*
805 * We only accept unified patches, so we want it to
806 * at least have "@@ -a,b +c,d @@\n", which is 14 chars
807 * minimum
808 */
809 nextlen = linelen(line + len, size - len);
810 if (size < nextlen + 14 || memcmp("@@ -", line + len + nextlen, 4))
811 continue;
812
813 /* Ok, we'll consider it a patch */
814 parse_traditional_patch(line, line+len, patch);
815 *hdrsize = len + nextlen;
816 linenr += 2;
817 return offset;
818 }
819 return -1;
820 }
821
822 /*
823 * Parse a unified diff. Note that this really needs
824 * to parse each fragment separately, since the only
825 * way to know the difference between a "---" that is
826 * part of a patch, and a "---" that starts the next
827 * patch is to look at the line counts..
828 */
829 static int parse_fragment(char *line, unsigned long size, struct patch *patch, struct fragment *fragment)
830 {
831 int added, deleted;
832 int len = linelen(line, size), offset;
833 unsigned long oldlines, newlines;
834 unsigned long leading, trailing;
835
836 offset = parse_fragment_header(line, len, fragment);
837 if (offset < 0)
838 return -1;
839 oldlines = fragment->oldlines;
840 newlines = fragment->newlines;
841 leading = 0;
842 trailing = 0;
843
844 if (patch->is_new < 0) {
845 patch->is_new = !oldlines;
846 if (!oldlines)
847 patch->old_name = NULL;
848 }
849 if (patch->is_delete < 0) {
850 patch->is_delete = !newlines;
851 if (!newlines)
852 patch->new_name = NULL;
853 }
854
855 if (patch->is_new && oldlines)
856 return error("new file depends on old contents");
857 if (patch->is_delete != !newlines) {
858 if (newlines)
859 return error("deleted file still has contents");
860 fprintf(stderr, "** warning: file %s becomes empty but is not deleted\n", patch->new_name);
861 }
862
863 /* Parse the thing.. */
864 line += len;
865 size -= len;
866 linenr++;
867 added = deleted = 0;
868 for (offset = len; size > 0; offset += len, size -= len, line += len, linenr++) {
869 if (!oldlines && !newlines)
870 break;
871 len = linelen(line, size);
872 if (!len || line[len-1] != '\n')
873 return -1;
874 switch (*line) {
875 default:
876 return -1;
877 case ' ':
878 oldlines--;
879 newlines--;
880 if (!deleted && !added)
881 leading++;
882 trailing++;
883 break;
884 case '-':
885 deleted++;
886 oldlines--;
887 trailing = 0;
888 break;
889 case '+':
890 /*
891 * We know len is at least two, since we have a '+' and
892 * we checked that the last character was a '\n' above.
893 * That is, an addition of an empty line would check
894 * the '+' here. Sneaky...
895 */
896 if ((new_whitespace != nowarn_whitespace) &&
897 isspace(line[len-2])) {
898 whitespace_error++;
899 if (squelch_whitespace_errors &&
900 squelch_whitespace_errors <
901 whitespace_error)
902 ;
903 else {
904 fprintf(stderr, "Adds trailing whitespace.\n%s:%d:%.*s\n",
905 patch_input_file,
906 linenr, len-2, line+1);
907 }
908 }
909 added++;
910 newlines--;
911 trailing = 0;
912 break;
913
914 /* We allow "\ No newline at end of file". Depending
915 * on locale settings when the patch was produced we
916 * don't know what this line looks like. The only
917 * thing we do know is that it begins with "\ ".
918 * Checking for 12 is just for sanity check -- any
919 * l10n of "\ No newline..." is at least that long.
920 */
921 case '\\':
922 if (len < 12 || memcmp(line, "\\ ", 2))
923 return -1;
924 break;
925 }
926 }
927 if (oldlines || newlines)
928 return -1;
929 fragment->leading = leading;
930 fragment->trailing = trailing;
931
932 /* If a fragment ends with an incomplete line, we failed to include
933 * it in the above loop because we hit oldlines == newlines == 0
934 * before seeing it.
935 */
936 if (12 < size && !memcmp(line, "\\ ", 2))
937 offset += linelen(line, size);
938
939 patch->lines_added += added;
940 patch->lines_deleted += deleted;
941 return offset;
942 }
943
944 static int parse_single_patch(char *line, unsigned long size, struct patch *patch)
945 {
946 unsigned long offset = 0;
947 struct fragment **fragp = &patch->fragments;
948
949 while (size > 4 && !memcmp(line, "@@ -", 4)) {
950 struct fragment *fragment;
951 int len;
952
953 fragment = xcalloc(1, sizeof(*fragment));
954 len = parse_fragment(line, size, patch, fragment);
955 if (len <= 0)
956 die("corrupt patch at line %d", linenr);
957
958 fragment->patch = line;
959 fragment->size = len;
960
961 *fragp = fragment;
962 fragp = &fragment->next;
963
964 offset += len;
965 line += len;
966 size -= len;
967 }
968 return offset;
969 }
970
971 static inline int metadata_changes(struct patch *patch)
972 {
973 return patch->is_rename > 0 ||
974 patch->is_copy > 0 ||
975 patch->is_new > 0 ||
976 patch->is_delete ||
977 (patch->old_mode && patch->new_mode &&
978 patch->old_mode != patch->new_mode);
979 }
980
981 static int parse_binary(char *buffer, unsigned long size, struct patch *patch)
982 {
983 /* We have read "GIT binary patch\n"; what follows is a line
984 * that says the patch method (currently, either "deflated
985 * literal" or "deflated delta") and the length of data before
986 * deflating; a sequence of 'length-byte' followed by base-85
987 * encoded data follows.
988 *
989 * Each 5-byte sequence of base-85 encodes up to 4 bytes,
990 * and we would limit the patch line to 66 characters,
991 * so one line can fit up to 13 groups that would decode
992 * to 52 bytes max. The length byte 'A'-'Z' corresponds
993 * to 1-26 bytes, and 'a'-'z' corresponds to 27-52 bytes.
994 * The end of binary is signaled with an empty line.
995 */
996 int llen, used;
997 struct fragment *fragment;
998 char *data = NULL;
999
1000 patch->fragments = fragment = xcalloc(1, sizeof(*fragment));
1001
1002 /* Grab the type of patch */
1003 llen = linelen(buffer, size);
1004 used = llen;
1005 linenr++;
1006
1007 if (!strncmp(buffer, "delta ", 6)) {
1008 patch->is_binary = BINARY_DELTA_DEFLATED;
1009 patch->deflate_origlen = strtoul(buffer + 6, NULL, 10);
1010 }
1011 else if (!strncmp(buffer, "literal ", 8)) {
1012 patch->is_binary = BINARY_LITERAL_DEFLATED;
1013 patch->deflate_origlen = strtoul(buffer + 8, NULL, 10);
1014 }
1015 else
1016 return error("unrecognized binary patch at line %d: %.*s",
1017 linenr-1, llen-1, buffer);
1018 buffer += llen;
1019 while (1) {
1020 int byte_length, max_byte_length, newsize;
1021 llen = linelen(buffer, size);
1022 used += llen;
1023 linenr++;
1024 if (llen == 1)
1025 break;
1026 /* Minimum line is "A00000\n" which is 7-byte long,
1027 * and the line length must be multiple of 5 plus 2.
1028 */
1029 if ((llen < 7) || (llen-2) % 5)
1030 goto corrupt;
1031 max_byte_length = (llen - 2) / 5 * 4;
1032 byte_length = *buffer;
1033 if ('A' <= byte_length && byte_length <= 'Z')
1034 byte_length = byte_length - 'A' + 1;
1035 else if ('a' <= byte_length && byte_length <= 'z')
1036 byte_length = byte_length - 'a' + 27;
1037 else
1038 goto corrupt;
1039 /* if the input length was not multiple of 4, we would
1040 * have filler at the end but the filler should never
1041 * exceed 3 bytes
1042 */
1043 if (max_byte_length < byte_length ||
1044 byte_length <= max_byte_length - 4)
1045 goto corrupt;
1046 newsize = fragment->size + byte_length;
1047 data = xrealloc(data, newsize);
1048 if (decode_85(data + fragment->size,
1049 buffer + 1,
1050 byte_length))
1051 goto corrupt;
1052 fragment->size = newsize;
1053 buffer += llen;
1054 size -= llen;
1055 }
1056 fragment->patch = data;
1057 return used;
1058 corrupt:
1059 return error("corrupt binary patch at line %d: %.*s",
1060 linenr-1, llen-1, buffer);
1061 }
1062
1063 static int parse_chunk(char *buffer, unsigned long size, struct patch *patch)
1064 {
1065 int hdrsize, patchsize;
1066 int offset = find_header(buffer, size, &hdrsize, patch);
1067
1068 if (offset < 0)
1069 return offset;
1070
1071 patchsize = parse_single_patch(buffer + offset + hdrsize, size - offset - hdrsize, patch);
1072
1073 if (!patchsize) {
1074 static const char *binhdr[] = {
1075 "Binary files ",
1076 "Files ",
1077 NULL,
1078 };
1079 static const char git_binary[] = "GIT binary patch\n";
1080 int i;
1081 int hd = hdrsize + offset;
1082 unsigned long llen = linelen(buffer + hd, size - hd);
1083
1084 if (llen == sizeof(git_binary) - 1 &&
1085 !memcmp(git_binary, buffer + hd, llen)) {
1086 int used;
1087 linenr++;
1088 used = parse_binary(buffer + hd + llen,
1089 size - hd - llen, patch);
1090 if (used)
1091 patchsize = used + llen;
1092 else
1093 patchsize = 0;
1094 }
1095 else if (!memcmp(" differ\n", buffer + hd + llen - 8, 8)) {
1096 for (i = 0; binhdr[i]; i++) {
1097 int len = strlen(binhdr[i]);
1098 if (len < size - hd &&
1099 !memcmp(binhdr[i], buffer + hd, len)) {
1100 linenr++;
1101 patch->is_binary = 1;
1102 patchsize = llen;
1103 break;
1104 }
1105 }
1106 }
1107
1108 /* Empty patch cannot be applied if:
1109 * - it is a binary patch and we do not do binary_replace, or
1110 * - text patch without metadata change
1111 */
1112 if ((apply || check) &&
1113 (patch->is_binary
1114 ? !allow_binary_replacement
1115 : !metadata_changes(patch)))
1116 die("patch with only garbage at line %d", linenr);
1117 }
1118
1119 return offset + hdrsize + patchsize;
1120 }
1121
1122 static const char pluses[] = "++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++";
1123 static const char minuses[]= "----------------------------------------------------------------------";
1124
1125 static void show_stats(struct patch *patch)
1126 {
1127 const char *prefix = "";
1128 char *name = patch->new_name;
1129 char *qname = NULL;
1130 int len, max, add, del, total;
1131
1132 if (!name)
1133 name = patch->old_name;
1134
1135 if (0 < (len = quote_c_style(name, NULL, NULL, 0))) {
1136 qname = xmalloc(len + 1);
1137 quote_c_style(name, qname, NULL, 0);
1138 name = qname;
1139 }
1140
1141 /*
1142 * "scale" the filename
1143 */
1144 len = strlen(name);
1145 max = max_len;
1146 if (max > 50)
1147 max = 50;
1148 if (len > max) {
1149 char *slash;
1150 prefix = "...";
1151 max -= 3;
1152 name += len - max;
1153 slash = strchr(name, '/');
1154 if (slash)
1155 name = slash;
1156 }
1157 len = max;
1158
1159 /*
1160 * scale the add/delete
1161 */
1162 max = max_change;
1163 if (max + len > 70)
1164 max = 70 - len;
1165
1166 add = patch->lines_added;
1167 del = patch->lines_deleted;
1168 total = add + del;
1169
1170 if (max_change > 0) {
1171 total = (total * max + max_change / 2) / max_change;
1172 add = (add * max + max_change / 2) / max_change;
1173 del = total - add;
1174 }
1175 if (patch->is_binary)
1176 printf(" %s%-*s | Bin\n", prefix, len, name);
1177 else
1178 printf(" %s%-*s |%5d %.*s%.*s\n", prefix,
1179 len, name, patch->lines_added + patch->lines_deleted,
1180 add, pluses, del, minuses);
1181 if (qname)
1182 free(qname);
1183 }
1184
1185 static int read_old_data(struct stat *st, const char *path, void *buf, unsigned long size)
1186 {
1187 int fd;
1188 unsigned long got;
1189
1190 switch (st->st_mode & S_IFMT) {
1191 case S_IFLNK:
1192 return readlink(path, buf, size);
1193 case S_IFREG:
1194 fd = open(path, O_RDONLY);
1195 if (fd < 0)
1196 return error("unable to open %s", path);
1197 got = 0;
1198 for (;;) {
1199 int ret = xread(fd, (char *) buf + got, size - got);
1200 if (ret <= 0)
1201 break;
1202 got += ret;
1203 }
1204 close(fd);
1205 return got;
1206
1207 default:
1208 return -1;
1209 }
1210 }
1211
1212 static int find_offset(const char *buf, unsigned long size, const char *fragment, unsigned long fragsize, int line, int *lines)
1213 {
1214 int i;
1215 unsigned long start, backwards, forwards;
1216
1217 if (fragsize > size)
1218 return -1;
1219
1220 start = 0;
1221 if (line > 1) {
1222 unsigned long offset = 0;
1223 i = line-1;
1224 while (offset + fragsize <= size) {
1225 if (buf[offset++] == '\n') {
1226 start = offset;
1227 if (!--i)
1228 break;
1229 }
1230 }
1231 }
1232
1233 /* Exact line number? */
1234 if (!memcmp(buf + start, fragment, fragsize))
1235 return start;
1236
1237 /*
1238 * There's probably some smart way to do this, but I'll leave
1239 * that to the smart and beautiful people. I'm simple and stupid.
1240 */
1241 backwards = start;
1242 forwards = start;
1243 for (i = 0; ; i++) {
1244 unsigned long try;
1245 int n;
1246
1247 /* "backward" */
1248 if (i & 1) {
1249 if (!backwards) {
1250 if (forwards + fragsize > size)
1251 break;
1252 continue;
1253 }
1254 do {
1255 --backwards;
1256 } while (backwards && buf[backwards-1] != '\n');
1257 try = backwards;
1258 } else {
1259 while (forwards + fragsize <= size) {
1260 if (buf[forwards++] == '\n')
1261 break;
1262 }
1263 try = forwards;
1264 }
1265
1266 if (try + fragsize > size)
1267 continue;
1268 if (memcmp(buf + try, fragment, fragsize))
1269 continue;
1270 n = (i >> 1)+1;
1271 if (i & 1)
1272 n = -n;
1273 *lines = n;
1274 return try;
1275 }
1276
1277 /*
1278 * We should start searching forward and backward.
1279 */
1280 return -1;
1281 }
1282
1283 static void remove_first_line(const char **rbuf, int *rsize)
1284 {
1285 const char *buf = *rbuf;
1286 int size = *rsize;
1287 unsigned long offset;
1288 offset = 0;
1289 while (offset <= size) {
1290 if (buf[offset++] == '\n')
1291 break;
1292 }
1293 *rsize = size - offset;
1294 *rbuf = buf + offset;
1295 }
1296
1297 static void remove_last_line(const char **rbuf, int *rsize)
1298 {
1299 const char *buf = *rbuf;
1300 int size = *rsize;
1301 unsigned long offset;
1302 offset = size - 1;
1303 while (offset > 0) {
1304 if (buf[--offset] == '\n')
1305 break;
1306 }
1307 *rsize = offset + 1;
1308 }
1309
1310 struct buffer_desc {
1311 char *buffer;
1312 unsigned long size;
1313 unsigned long alloc;
1314 };
1315
1316 static int apply_line(char *output, const char *patch, int plen)
1317 {
1318 /* plen is number of bytes to be copied from patch,
1319 * starting at patch+1 (patch[0] is '+'). Typically
1320 * patch[plen] is '\n'.
1321 */
1322 int add_nl_to_tail = 0;
1323 if ((new_whitespace == strip_whitespace) &&
1324 1 < plen && isspace(patch[plen-1])) {
1325 if (patch[plen] == '\n')
1326 add_nl_to_tail = 1;
1327 plen--;
1328 while (0 < plen && isspace(patch[plen]))
1329 plen--;
1330 applied_after_stripping++;
1331 }
1332 memcpy(output, patch + 1, plen);
1333 if (add_nl_to_tail)
1334 output[plen++] = '\n';
1335 return plen;
1336 }
1337
1338 static int apply_one_fragment(struct buffer_desc *desc, struct fragment *frag,
1339 int inaccurate_eof)
1340 {
1341 int match_beginning, match_end;
1342 char *buf = desc->buffer;
1343 const char *patch = frag->patch;
1344 int offset, size = frag->size;
1345 char *old = xmalloc(size);
1346 char *new = xmalloc(size);
1347 const char *oldlines, *newlines;
1348 int oldsize = 0, newsize = 0;
1349 unsigned long leading, trailing;
1350 int pos, lines;
1351
1352 while (size > 0) {
1353 int len = linelen(patch, size);
1354 int plen;
1355
1356 if (!len)
1357 break;
1358
1359 /*
1360 * "plen" is how much of the line we should use for
1361 * the actual patch data. Normally we just remove the
1362 * first character on the line, but if the line is
1363 * followed by "\ No newline", then we also remove the
1364 * last one (which is the newline, of course).
1365 */
1366 plen = len-1;
1367 if (len < size && patch[len] == '\\')
1368 plen--;
1369 switch (*patch) {
1370 case ' ':
1371 case '-':
1372 memcpy(old + oldsize, patch + 1, plen);
1373 oldsize += plen;
1374 if (*patch == '-')
1375 break;
1376 /* Fall-through for ' ' */
1377 case '+':
1378 if (*patch != '+' || !no_add)
1379 newsize += apply_line(new + newsize, patch,
1380 plen);
1381 break;
1382 case '@': case '\\':
1383 /* Ignore it, we already handled it */
1384 break;
1385 default:
1386 return -1;
1387 }
1388 patch += len;
1389 size -= len;
1390 }
1391
1392 if (inaccurate_eof && oldsize > 0 && old[oldsize - 1] == '\n' &&
1393 newsize > 0 && new[newsize - 1] == '\n') {
1394 oldsize--;
1395 newsize--;
1396 }
1397
1398 oldlines = old;
1399 newlines = new;
1400 leading = frag->leading;
1401 trailing = frag->trailing;
1402
1403 /*
1404 * If we don't have any leading/trailing data in the patch,
1405 * we want it to match at the beginning/end of the file.
1406 */
1407 match_beginning = !leading && (frag->oldpos == 1);
1408 match_end = !trailing;
1409
1410 lines = 0;
1411 pos = frag->newpos;
1412 for (;;) {
1413 offset = find_offset(buf, desc->size, oldlines, oldsize, pos, &lines);
1414 if (match_end && offset + oldsize != desc->size)
1415 offset = -1;
1416 if (match_beginning && offset)
1417 offset = -1;
1418 if (offset >= 0) {
1419 int diff = newsize - oldsize;
1420 unsigned long size = desc->size + diff;
1421 unsigned long alloc = desc->alloc;
1422
1423 /* Warn if it was necessary to reduce the number
1424 * of context lines.
1425 */
1426 if ((leading != frag->leading) || (trailing != frag->trailing))
1427 fprintf(stderr, "Context reduced to (%ld/%ld) to apply fragment at %d\n",
1428 leading, trailing, pos + lines);
1429
1430 if (size > alloc) {
1431 alloc = size + 8192;
1432 desc->alloc = alloc;
1433 buf = xrealloc(buf, alloc);
1434 desc->buffer = buf;
1435 }
1436 desc->size = size;
1437 memmove(buf + offset + newsize, buf + offset + oldsize, size - offset - newsize);
1438 memcpy(buf + offset, newlines, newsize);
1439 offset = 0;
1440
1441 break;
1442 }
1443
1444 /* Am I at my context limits? */
1445 if ((leading <= p_context) && (trailing <= p_context))
1446 break;
1447 if (match_beginning || match_end) {
1448 match_beginning = match_end = 0;
1449 continue;
1450 }
1451 /* Reduce the number of context lines
1452 * Reduce both leading and trailing if they are equal
1453 * otherwise just reduce the larger context.
1454 */
1455 if (leading >= trailing) {
1456 remove_first_line(&oldlines, &oldsize);
1457 remove_first_line(&newlines, &newsize);
1458 pos--;
1459 leading--;
1460 }
1461 if (trailing > leading) {
1462 remove_last_line(&oldlines, &oldsize);
1463 remove_last_line(&newlines, &newsize);
1464 trailing--;
1465 }
1466 }
1467
1468 free(old);
1469 free(new);
1470 return offset;
1471 }
1472
1473 static char *inflate_it(const void *data, unsigned long size,
1474 unsigned long inflated_size)
1475 {
1476 z_stream stream;
1477 void *out;
1478 int st;
1479
1480 memset(&stream, 0, sizeof(stream));
1481
1482 stream.next_in = (unsigned char *)data;
1483 stream.avail_in = size;
1484 stream.next_out = out = xmalloc(inflated_size);
1485 stream.avail_out = inflated_size;
1486 inflateInit(&stream);
1487 st = inflate(&stream, Z_FINISH);
1488 if ((st != Z_STREAM_END) || stream.total_out != inflated_size) {
1489 free(out);
1490 return NULL;
1491 }
1492 return out;
1493 }
1494
1495 static int apply_binary_fragment(struct buffer_desc *desc, struct patch *patch)
1496 {
1497 unsigned long dst_size;
1498 struct fragment *fragment = patch->fragments;
1499 void *data;
1500 void *result;
1501
1502 data = inflate_it(fragment->patch, fragment->size,
1503 patch->deflate_origlen);
1504 if (!data)
1505 return error("corrupt patch data");
1506 switch (patch->is_binary) {
1507 case BINARY_DELTA_DEFLATED:
1508 result = patch_delta(desc->buffer, desc->size,
1509 data,
1510 patch->deflate_origlen,
1511 &dst_size);
1512 free(desc->buffer);
1513 desc->buffer = result;
1514 free(data);
1515 break;
1516 case BINARY_LITERAL_DEFLATED:
1517 free(desc->buffer);
1518 desc->buffer = data;
1519 dst_size = patch->deflate_origlen;
1520 break;
1521 }
1522 if (!desc->buffer)
1523 return -1;
1524 desc->size = desc->alloc = dst_size;
1525 return 0;
1526 }
1527
1528 static int apply_binary(struct buffer_desc *desc, struct patch *patch)
1529 {
1530 const char *name = patch->old_name ? patch->old_name : patch->new_name;
1531 unsigned char sha1[20];
1532 unsigned char hdr[50];
1533 int hdrlen;
1534
1535 if (!allow_binary_replacement)
1536 return error("cannot apply binary patch to '%s' "
1537 "without --allow-binary-replacement",
1538 name);
1539
1540 /* For safety, we require patch index line to contain
1541 * full 40-byte textual SHA1 for old and new, at least for now.
1542 */
1543 if (strlen(patch->old_sha1_prefix) != 40 ||
1544 strlen(patch->new_sha1_prefix) != 40 ||
1545 get_sha1_hex(patch->old_sha1_prefix, sha1) ||
1546 get_sha1_hex(patch->new_sha1_prefix, sha1))
1547 return error("cannot apply binary patch to '%s' "
1548 "without full index line", name);
1549
1550 if (patch->old_name) {
1551 /* See if the old one matches what the patch
1552 * applies to.
1553 */
1554 write_sha1_file_prepare(desc->buffer, desc->size,
1555 blob_type, sha1, hdr, &hdrlen);
1556 if (strcmp(sha1_to_hex(sha1), patch->old_sha1_prefix))
1557 return error("the patch applies to '%s' (%s), "
1558 "which does not match the "
1559 "current contents.",
1560 name, sha1_to_hex(sha1));
1561 }
1562 else {
1563 /* Otherwise, the old one must be empty. */
1564 if (desc->size)
1565 return error("the patch applies to an empty "
1566 "'%s' but it is not empty", name);
1567 }
1568
1569 get_sha1_hex(patch->new_sha1_prefix, sha1);
1570 if (!memcmp(sha1, null_sha1, 20)) {
1571 free(desc->buffer);
1572 desc->alloc = desc->size = 0;
1573 desc->buffer = NULL;
1574 return 0; /* deletion patch */
1575 }
1576
1577 if (has_sha1_file(sha1)) {
1578 /* We already have the postimage */
1579 char type[10];
1580 unsigned long size;
1581
1582 free(desc->buffer);
1583 desc->buffer = read_sha1_file(sha1, type, &size);
1584 if (!desc->buffer)
1585 return error("the necessary postimage %s for "
1586 "'%s' cannot be read",
1587 patch->new_sha1_prefix, name);
1588 desc->alloc = desc->size = size;
1589 }
1590 else {
1591 /* We have verified desc matches the preimage;
1592 * apply the patch data to it, which is stored
1593 * in the patch->fragments->{patch,size}.
1594 */
1595 if (apply_binary_fragment(desc, patch))
1596 return error("binary patch does not apply to '%s'",
1597 name);
1598
1599 /* verify that the result matches */
1600 write_sha1_file_prepare(desc->buffer, desc->size, blob_type,
1601 sha1, hdr, &hdrlen);
1602 if (strcmp(sha1_to_hex(sha1), patch->new_sha1_prefix))
1603 return error("binary patch to '%s' creates incorrect result", name);
1604 }
1605
1606 return 0;
1607 }
1608
1609 static int apply_fragments(struct buffer_desc *desc, struct patch *patch)
1610 {
1611 struct fragment *frag = patch->fragments;
1612 const char *name = patch->old_name ? patch->old_name : patch->new_name;
1613
1614 if (patch->is_binary)
1615 return apply_binary(desc, patch);
1616
1617 while (frag) {
1618 if (apply_one_fragment(desc, frag, patch->inaccurate_eof) < 0)
1619 return error("patch failed: %s:%ld",
1620 name, frag->oldpos);
1621 frag = frag->next;
1622 }
1623 return 0;
1624 }
1625
1626 static int apply_data(struct patch *patch, struct stat *st, struct cache_entry *ce)
1627 {
1628 char *buf;
1629 unsigned long size, alloc;
1630 struct buffer_desc desc;
1631
1632 size = 0;
1633 alloc = 0;
1634 buf = NULL;
1635 if (cached) {
1636 if (ce) {
1637 char type[20];
1638 buf = read_sha1_file(ce->sha1, type, &size);
1639 if (!buf)
1640 return error("read of %s failed",
1641 patch->old_name);
1642 alloc = size;
1643 }
1644 }
1645 else if (patch->old_name) {
1646 size = st->st_size;
1647 alloc = size + 8192;
1648 buf = xmalloc(alloc);
1649 if (read_old_data(st, patch->old_name, buf, alloc) != size)
1650 return error("read of %s failed", patch->old_name);
1651 }
1652
1653 desc.size = size;
1654 desc.alloc = alloc;
1655 desc.buffer = buf;
1656 if (apply_fragments(&desc, patch) < 0)
1657 return -1;
1658 patch->result = desc.buffer;
1659 patch->resultsize = desc.size;
1660
1661 if (patch->is_delete && patch->resultsize)
1662 return error("removal patch leaves file contents");
1663
1664 return 0;
1665 }
1666
1667 static int check_patch(struct patch *patch)
1668 {
1669 struct stat st;
1670 const char *old_name = patch->old_name;
1671 const char *new_name = patch->new_name;
1672 const char *name = old_name ? old_name : new_name;
1673 struct cache_entry *ce = NULL;
1674
1675 if (old_name) {
1676 int changed = 0;
1677 int stat_ret = 0;
1678 unsigned st_mode = 0;
1679
1680 if (!cached)
1681 stat_ret = lstat(old_name, &st);
1682 if (check_index) {
1683 int pos = cache_name_pos(old_name, strlen(old_name));
1684 if (pos < 0)
1685 return error("%s: does not exist in index",
1686 old_name);
1687 ce = active_cache[pos];
1688 if (stat_ret < 0) {
1689 struct checkout costate;
1690 if (errno != ENOENT)
1691 return error("%s: %s", old_name,
1692 strerror(errno));
1693 /* checkout */
1694 costate.base_dir = "";
1695 costate.base_dir_len = 0;
1696 costate.force = 0;
1697 costate.quiet = 0;
1698 costate.not_new = 0;
1699 costate.refresh_cache = 1;
1700 if (checkout_entry(ce,
1701 &costate,
1702 NULL) ||
1703 lstat(old_name, &st))
1704 return -1;
1705 }
1706 if (!cached)
1707 changed = ce_match_stat(ce, &st, 1);
1708 if (changed)
1709 return error("%s: does not match index",
1710 old_name);
1711 if (cached)
1712 st_mode = ntohl(ce->ce_mode);
1713 }
1714 else if (stat_ret < 0)
1715 return error("%s: %s", old_name, strerror(errno));
1716
1717 if (!cached)
1718 st_mode = ntohl(create_ce_mode(st.st_mode));
1719
1720 if (patch->is_new < 0)
1721 patch->is_new = 0;
1722 if (!patch->old_mode)
1723 patch->old_mode = st_mode;
1724 if ((st_mode ^ patch->old_mode) & S_IFMT)
1725 return error("%s: wrong type", old_name);
1726 if (st_mode != patch->old_mode)
1727 fprintf(stderr, "warning: %s has type %o, expected %o\n",
1728 old_name, st_mode, patch->old_mode);
1729 }
1730
1731 if (new_name && (patch->is_new | patch->is_rename | patch->is_copy)) {
1732 if (check_index && cache_name_pos(new_name, strlen(new_name)) >= 0)
1733 return error("%s: already exists in index", new_name);
1734 if (!cached) {
1735 if (!lstat(new_name, &st))
1736 return error("%s: already exists in working directory", new_name);
1737 if (errno != ENOENT)
1738 return error("%s: %s", new_name, strerror(errno));
1739 }
1740 if (!patch->new_mode) {
1741 if (patch->is_new)
1742 patch->new_mode = S_IFREG | 0644;
1743 else
1744 patch->new_mode = patch->old_mode;
1745 }
1746 }
1747
1748 if (new_name && old_name) {
1749 int same = !strcmp(old_name, new_name);
1750 if (!patch->new_mode)
1751 patch->new_mode = patch->old_mode;
1752 if ((patch->old_mode ^ patch->new_mode) & S_IFMT)
1753 return error("new mode (%o) of %s does not match old mode (%o)%s%s",
1754 patch->new_mode, new_name, patch->old_mode,
1755 same ? "" : " of ", same ? "" : old_name);
1756 }
1757
1758 if (apply_data(patch, &st, ce) < 0)
1759 return error("%s: patch does not apply", name);
1760 return 0;
1761 }
1762
1763 static int check_patch_list(struct patch *patch)
1764 {
1765 int error = 0;
1766
1767 for (;patch ; patch = patch->next)
1768 error |= check_patch(patch);
1769 return error;
1770 }
1771
1772 static inline int is_null_sha1(const unsigned char *sha1)
1773 {
1774 return !memcmp(sha1, null_sha1, 20);
1775 }
1776
1777 static void show_index_list(struct patch *list)
1778 {
1779 struct patch *patch;
1780
1781 /* Once we start supporting the reverse patch, it may be
1782 * worth showing the new sha1 prefix, but until then...
1783 */
1784 for (patch = list; patch; patch = patch->next) {
1785 const unsigned char *sha1_ptr;
1786 unsigned char sha1[20];
1787 const char *name;
1788
1789 name = patch->old_name ? patch->old_name : patch->new_name;
1790 if (patch->is_new)
1791 sha1_ptr = null_sha1;
1792 else if (get_sha1(patch->old_sha1_prefix, sha1))
1793 die("sha1 information is lacking or useless (%s).",
1794 name);
1795 else
1796 sha1_ptr = sha1;
1797
1798 printf("%06o %s ",patch->old_mode, sha1_to_hex(sha1_ptr));
1799 if (line_termination && quote_c_style(name, NULL, NULL, 0))
1800 quote_c_style(name, NULL, stdout, 0);
1801 else
1802 fputs(name, stdout);
1803 putchar(line_termination);
1804 }
1805 }
1806
1807 static void stat_patch_list(struct patch *patch)
1808 {
1809 int files, adds, dels;
1810
1811 for (files = adds = dels = 0 ; patch ; patch = patch->next) {
1812 files++;
1813 adds += patch->lines_added;
1814 dels += patch->lines_deleted;
1815 show_stats(patch);
1816 }
1817
1818 printf(" %d files changed, %d insertions(+), %d deletions(-)\n", files, adds, dels);
1819 }
1820
1821 static void numstat_patch_list(struct patch *patch)
1822 {
1823 for ( ; patch; patch = patch->next) {
1824 const char *name;
1825 name = patch->new_name ? patch->new_name : patch->old_name;
1826 printf("%d\t%d\t", patch->lines_added, patch->lines_deleted);
1827 if (line_termination && quote_c_style(name, NULL, NULL, 0))
1828 quote_c_style(name, NULL, stdout, 0);
1829 else
1830 fputs(name, stdout);
1831 putchar('\n');
1832 }
1833 }
1834
1835 static void show_file_mode_name(const char *newdelete, unsigned int mode, const char *name)
1836 {
1837 if (mode)
1838 printf(" %s mode %06o %s\n", newdelete, mode, name);
1839 else
1840 printf(" %s %s\n", newdelete, name);
1841 }
1842
1843 static void show_mode_change(struct patch *p, int show_name)
1844 {
1845 if (p->old_mode && p->new_mode && p->old_mode != p->new_mode) {
1846 if (show_name)
1847 printf(" mode change %06o => %06o %s\n",
1848 p->old_mode, p->new_mode, p->new_name);
1849 else
1850 printf(" mode change %06o => %06o\n",
1851 p->old_mode, p->new_mode);
1852 }
1853 }
1854
1855 static void show_rename_copy(struct patch *p)
1856 {
1857 const char *renamecopy = p->is_rename ? "rename" : "copy";
1858 const char *old, *new;
1859
1860 /* Find common prefix */
1861 old = p->old_name;
1862 new = p->new_name;
1863 while (1) {
1864 const char *slash_old, *slash_new;
1865 slash_old = strchr(old, '/');
1866 slash_new = strchr(new, '/');
1867 if (!slash_old ||
1868 !slash_new ||
1869 slash_old - old != slash_new - new ||
1870 memcmp(old, new, slash_new - new))
1871 break;
1872 old = slash_old + 1;
1873 new = slash_new + 1;
1874 }
1875 /* p->old_name thru old is the common prefix, and old and new
1876 * through the end of names are renames
1877 */
1878 if (old != p->old_name)
1879 printf(" %s %.*s{%s => %s} (%d%%)\n", renamecopy,
1880 (int)(old - p->old_name), p->old_name,
1881 old, new, p->score);
1882 else
1883 printf(" %s %s => %s (%d%%)\n", renamecopy,
1884 p->old_name, p->new_name, p->score);
1885 show_mode_change(p, 0);
1886 }
1887
1888 static void summary_patch_list(struct patch *patch)
1889 {
1890 struct patch *p;
1891
1892 for (p = patch; p; p = p->next) {
1893 if (p->is_new)
1894 show_file_mode_name("create", p->new_mode, p->new_name);
1895 else if (p->is_delete)
1896 show_file_mode_name("delete", p->old_mode, p->old_name);
1897 else {
1898 if (p->is_rename || p->is_copy)
1899 show_rename_copy(p);
1900 else {
1901 if (p->score) {
1902 printf(" rewrite %s (%d%%)\n",
1903 p->new_name, p->score);
1904 show_mode_change(p, 0);
1905 }
1906 else
1907 show_mode_change(p, 1);
1908 }
1909 }
1910 }
1911 }
1912
1913 static void patch_stats(struct patch *patch)
1914 {
1915 int lines = patch->lines_added + patch->lines_deleted;
1916
1917 if (lines > max_change)
1918 max_change = lines;
1919 if (patch->old_name) {
1920 int len = quote_c_style(patch->old_name, NULL, NULL, 0);
1921 if (!len)
1922 len = strlen(patch->old_name);
1923 if (len > max_len)
1924 max_len = len;
1925 }
1926 if (patch->new_name) {
1927 int len = quote_c_style(patch->new_name, NULL, NULL, 0);
1928 if (!len)
1929 len = strlen(patch->new_name);
1930 if (len > max_len)
1931 max_len = len;
1932 }
1933 }
1934
1935 static void remove_file(struct patch *patch)
1936 {
1937 if (write_index) {
1938 if (remove_file_from_cache(patch->old_name) < 0)
1939 die("unable to remove %s from index", patch->old_name);
1940 cache_tree_invalidate_path(active_cache_tree, patch->old_name);
1941 }
1942 if (!cached)
1943 unlink(patch->old_name);
1944 }
1945
1946 static void add_index_file(const char *path, unsigned mode, void *buf, unsigned long size)
1947 {
1948 struct stat st;
1949 struct cache_entry *ce;
1950 int namelen = strlen(path);
1951 unsigned ce_size = cache_entry_size(namelen);
1952
1953 if (!write_index)
1954 return;
1955
1956 ce = xcalloc(1, ce_size);
1957 memcpy(ce->name, path, namelen);
1958 ce->ce_mode = create_ce_mode(mode);
1959 ce->ce_flags = htons(namelen);
1960 if (!cached) {
1961 if (lstat(path, &st) < 0)
1962 die("unable to stat newly created file %s", path);
1963 fill_stat_cache_info(ce, &st);
1964 }
1965 if (write_sha1_file(buf, size, blob_type, ce->sha1) < 0)
1966 die("unable to create backing store for newly created file %s", path);
1967 if (add_cache_entry(ce, ADD_CACHE_OK_TO_ADD) < 0)
1968 die("unable to add cache entry for %s", path);
1969 }
1970
1971 static int try_create_file(const char *path, unsigned int mode, const char *buf, unsigned long size)
1972 {
1973 int fd;
1974
1975 if (S_ISLNK(mode))
1976 return symlink(buf, path);
1977 fd = open(path, O_CREAT | O_EXCL | O_WRONLY, (mode & 0100) ? 0777 : 0666);
1978 if (fd < 0)
1979 return -1;
1980 while (size) {
1981 int written = xwrite(fd, buf, size);
1982 if (written < 0)
1983 die("writing file %s: %s", path, strerror(errno));
1984 if (!written)
1985 die("out of space writing file %s", path);
1986 buf += written;
1987 size -= written;
1988 }
1989 if (close(fd) < 0)
1990 die("closing file %s: %s", path, strerror(errno));
1991 return 0;
1992 }
1993
1994 /*
1995 * We optimistically assume that the directories exist,
1996 * which is true 99% of the time anyway. If they don't,
1997 * we create them and try again.
1998 */
1999 static void create_one_file(char *path, unsigned mode, const char *buf, unsigned long size)
2000 {
2001 if (cached)
2002 return;
2003 if (!try_create_file(path, mode, buf, size))
2004 return;
2005
2006 if (errno == ENOENT) {
2007 if (safe_create_leading_directories(path))
2008 return;
2009 if (!try_create_file(path, mode, buf, size))
2010 return;
2011 }
2012
2013 if (errno == EEXIST) {
2014 unsigned int nr = getpid();
2015
2016 for (;;) {
2017 const char *newpath;
2018 newpath = mkpath("%s~%u", path, nr);
2019 if (!try_create_file(newpath, mode, buf, size)) {
2020 if (!rename(newpath, path))
2021 return;
2022 unlink(newpath);
2023 break;
2024 }
2025 if (errno != EEXIST)
2026 break;
2027 ++nr;
2028 }
2029 }
2030 die("unable to write file %s mode %o", path, mode);
2031 }
2032
2033 static void create_file(struct patch *patch)
2034 {
2035 char *path = patch->new_name;
2036 unsigned mode = patch->new_mode;
2037 unsigned long size = patch->resultsize;
2038 char *buf = patch->result;
2039
2040 if (!mode)
2041 mode = S_IFREG | 0644;
2042 create_one_file(path, mode, buf, size);
2043 add_index_file(path, mode, buf, size);
2044 cache_tree_invalidate_path(active_cache_tree, path);
2045 }
2046
2047 static void write_out_one_result(struct patch *patch)
2048 {
2049 if (patch->is_delete > 0) {
2050 remove_file(patch);
2051 return;
2052 }
2053 if (patch->is_new > 0 || patch->is_copy) {
2054 create_file(patch);
2055 return;
2056 }
2057 /*
2058 * Rename or modification boils down to the same
2059 * thing: remove the old, write the new
2060 */
2061 remove_file(patch);
2062 create_file(patch);
2063 }
2064
2065 static void write_out_results(struct patch *list, int skipped_patch)
2066 {
2067 if (!list && !skipped_patch)
2068 die("No changes");
2069
2070 while (list) {
2071 write_out_one_result(list);
2072 list = list->next;
2073 }
2074 }
2075
2076 static struct lock_file lock_file;
2077
2078 static struct excludes {
2079 struct excludes *next;
2080 const char *path;
2081 } *excludes;
2082
2083 static int use_patch(struct patch *p)
2084 {
2085 const char *pathname = p->new_name ? p->new_name : p->old_name;
2086 struct excludes *x = excludes;
2087 while (x) {
2088 if (fnmatch(x->path, pathname, 0) == 0)
2089 return 0;
2090 x = x->next;
2091 }
2092 if (0 < prefix_length) {
2093 int pathlen = strlen(pathname);
2094 if (pathlen <= prefix_length ||
2095 memcmp(prefix, pathname, prefix_length))
2096 return 0;
2097 }
2098 return 1;
2099 }
2100
2101 static int apply_patch(int fd, const char *filename, int inaccurate_eof)
2102 {
2103 unsigned long offset, size;
2104 char *buffer = read_patch_file(fd, &size);
2105 struct patch *list = NULL, **listp = &list;
2106 int skipped_patch = 0;
2107
2108 patch_input_file = filename;
2109 if (!buffer)
2110 return -1;
2111 offset = 0;
2112 while (size > 0) {
2113 struct patch *patch;
2114 int nr;
2115
2116 patch = xcalloc(1, sizeof(*patch));
2117 patch->inaccurate_eof = inaccurate_eof;
2118 nr = parse_chunk(buffer + offset, size, patch);
2119 if (nr < 0)
2120 break;
2121 if (use_patch(patch)) {
2122 patch_stats(patch);
2123 *listp = patch;
2124 listp = &patch->next;
2125 } else {
2126 /* perhaps free it a bit better? */
2127 free(patch);
2128 skipped_patch++;
2129 }
2130 offset += nr;
2131 size -= nr;
2132 }
2133
2134 if (whitespace_error && (new_whitespace == error_on_whitespace))
2135 apply = 0;
2136
2137 write_index = check_index && apply;
2138 if (write_index && newfd < 0) {
2139 newfd = hold_lock_file_for_update(&lock_file,
2140 get_index_file());
2141 if (newfd < 0)
2142 die("unable to create new index file");
2143 }
2144 if (check_index) {
2145 if (read_cache() < 0)
2146 die("unable to read index file");
2147 }
2148
2149 if ((check || apply) && check_patch_list(list) < 0)
2150 exit(1);
2151
2152 if (apply)
2153 write_out_results(list, skipped_patch);
2154
2155 if (show_index_info)
2156 show_index_list(list);
2157
2158 if (diffstat)
2159 stat_patch_list(list);
2160
2161 if (numstat)
2162 numstat_patch_list(list);
2163
2164 if (summary)
2165 summary_patch_list(list);
2166
2167 free(buffer);
2168 return 0;
2169 }
2170
2171 static int git_apply_config(const char *var, const char *value)
2172 {
2173 if (!strcmp(var, "apply.whitespace")) {
2174 apply_default_whitespace = strdup(value);
2175 return 0;
2176 }
2177 return git_default_config(var, value);
2178 }
2179
2180
2181 int cmd_apply(int argc, const char **argv, char **envp)
2182 {
2183 int i;
2184 int read_stdin = 1;
2185 int inaccurate_eof = 0;
2186
2187 const char *whitespace_option = NULL;
2188
2189 for (i = 1; i < argc; i++) {
2190 const char *arg = argv[i];
2191 char *end;
2192 int fd;
2193
2194 if (!strcmp(arg, "-")) {
2195 apply_patch(0, "<stdin>", inaccurate_eof);
2196 read_stdin = 0;
2197 continue;
2198 }
2199 if (!strncmp(arg, "--exclude=", 10)) {
2200 struct excludes *x = xmalloc(sizeof(*x));
2201 x->path = arg + 10;
2202 x->next = excludes;
2203 excludes = x;
2204 continue;
2205 }
2206 if (!strncmp(arg, "-p", 2)) {
2207 p_value = atoi(arg + 2);
2208 continue;
2209 }
2210 if (!strcmp(arg, "--no-add")) {
2211 no_add = 1;
2212 continue;
2213 }
2214 if (!strcmp(arg, "--stat")) {
2215 apply = 0;
2216 diffstat = 1;
2217 continue;
2218 }
2219 if (!strcmp(arg, "--allow-binary-replacement") ||
2220 !strcmp(arg, "--binary")) {
2221 allow_binary_replacement = 1;
2222 continue;
2223 }
2224 if (!strcmp(arg, "--numstat")) {
2225 apply = 0;
2226 numstat = 1;
2227 continue;
2228 }
2229 if (!strcmp(arg, "--summary")) {
2230 apply = 0;
2231 summary = 1;
2232 continue;
2233 }
2234 if (!strcmp(arg, "--check")) {
2235 apply = 0;
2236 check = 1;
2237 continue;
2238 }
2239 if (!strcmp(arg, "--index")) {
2240 check_index = 1;
2241 continue;
2242 }
2243 if (!strcmp(arg, "--cached")) {
2244 check_index = 1;
2245 cached = 1;
2246 continue;
2247 }
2248 if (!strcmp(arg, "--apply")) {
2249 apply = 1;
2250 continue;
2251 }
2252 if (!strcmp(arg, "--index-info")) {
2253 apply = 0;
2254 show_index_info = 1;
2255 continue;
2256 }
2257 if (!strcmp(arg, "-z")) {
2258 line_termination = 0;
2259 continue;
2260 }
2261 if (!strncmp(arg, "-C", 2)) {
2262 p_context = strtoul(arg + 2, &end, 0);
2263 if (*end != '\0')
2264 die("unrecognized context count '%s'", arg + 2);
2265 continue;
2266 }
2267 if (!strncmp(arg, "--whitespace=", 13)) {
2268 whitespace_option = arg + 13;
2269 parse_whitespace_option(arg + 13);
2270 continue;
2271 }
2272 if (!strcmp(arg, "--inaccurate-eof")) {
2273 inaccurate_eof = 1;
2274 continue;
2275 }
2276
2277 if (check_index && prefix_length < 0) {
2278 prefix = setup_git_directory();
2279 prefix_length = prefix ? strlen(prefix) : 0;
2280 git_config(git_apply_config);
2281 if (!whitespace_option && apply_default_whitespace)
2282 parse_whitespace_option(apply_default_whitespace);
2283 }
2284 if (0 < prefix_length)
2285 arg = prefix_filename(prefix, prefix_length, arg);
2286
2287 fd = open(arg, O_RDONLY);
2288 if (fd < 0)
2289 usage(apply_usage);
2290 read_stdin = 0;
2291 set_default_whitespace_mode(whitespace_option);
2292 apply_patch(fd, arg, inaccurate_eof);
2293 close(fd);
2294 }
2295 set_default_whitespace_mode(whitespace_option);
2296 if (read_stdin)
2297 apply_patch(0, "<stdin>", inaccurate_eof);
2298 if (whitespace_error) {
2299 if (squelch_whitespace_errors &&
2300 squelch_whitespace_errors < whitespace_error) {
2301 int squelched =
2302 whitespace_error - squelch_whitespace_errors;
2303 fprintf(stderr, "warning: squelched %d whitespace error%s\n",
2304 squelched,
2305 squelched == 1 ? "" : "s");
2306 }
2307 if (new_whitespace == error_on_whitespace)
2308 die("%d line%s add%s trailing whitespaces.",
2309 whitespace_error,
2310 whitespace_error == 1 ? "" : "s",
2311 whitespace_error == 1 ? "s" : "");
2312 if (applied_after_stripping)
2313 fprintf(stderr, "warning: %d line%s applied after"
2314 " stripping trailing whitespaces.\n",
2315 applied_after_stripping,
2316 applied_after_stripping == 1 ? "" : "s");
2317 else if (whitespace_error)
2318 fprintf(stderr, "warning: %d line%s add%s trailing"
2319 " whitespaces.\n",
2320 whitespace_error,
2321 whitespace_error == 1 ? "" : "s",
2322 whitespace_error == 1 ? "s" : "");
2323 }
2324
2325 if (write_index) {
2326 if (write_cache(newfd, active_cache, active_nr) ||
2327 close(newfd) || commit_lock_file(&lock_file))
2328 die("Unable to write new index file");
2329 }
2330
2331 return 0;
2332 }