6a06be3025d0455259d7a94a884433b8998dbb99
[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 unidiff_zero;
31 static int p_value = 1;
32 static int check_index;
33 static int write_index;
34 static int cached;
35 static int diffstat;
36 static int numstat;
37 static int summary;
38 static int check;
39 static int apply = 1;
40 static int apply_in_reverse;
41 static int apply_with_reject;
42 static int apply_verbosely;
43 static int no_add;
44 static int show_index_info;
45 static int line_termination = '\n';
46 static unsigned long p_context = ULONG_MAX;
47 static const char apply_usage[] =
48 "git-apply [--stat] [--numstat] [--summary] [--check] [--index] [--cached] [--apply] [--no-add] [--index-info] [--allow-binary-replacement] [--reverse] [--reject] [--verbose] [-z] [-pNUM] [-CNUM] [--whitespace=<nowarn|warn|error|error-all|strip>] <patch>...";
49
50 static enum whitespace_eol {
51 nowarn_whitespace,
52 warn_on_whitespace,
53 error_on_whitespace,
54 strip_whitespace,
55 } new_whitespace = warn_on_whitespace;
56 static int whitespace_error;
57 static int squelch_whitespace_errors = 5;
58 static int applied_after_stripping;
59 static const char *patch_input_file;
60
61 static void parse_whitespace_option(const char *option)
62 {
63 if (!option) {
64 new_whitespace = warn_on_whitespace;
65 return;
66 }
67 if (!strcmp(option, "warn")) {
68 new_whitespace = warn_on_whitespace;
69 return;
70 }
71 if (!strcmp(option, "nowarn")) {
72 new_whitespace = nowarn_whitespace;
73 return;
74 }
75 if (!strcmp(option, "error")) {
76 new_whitespace = error_on_whitespace;
77 return;
78 }
79 if (!strcmp(option, "error-all")) {
80 new_whitespace = error_on_whitespace;
81 squelch_whitespace_errors = 0;
82 return;
83 }
84 if (!strcmp(option, "strip")) {
85 new_whitespace = strip_whitespace;
86 return;
87 }
88 die("unrecognized whitespace option '%s'", option);
89 }
90
91 static void set_default_whitespace_mode(const char *whitespace_option)
92 {
93 if (!whitespace_option && !apply_default_whitespace) {
94 new_whitespace = (apply
95 ? warn_on_whitespace
96 : nowarn_whitespace);
97 }
98 }
99
100 /*
101 * For "diff-stat" like behaviour, we keep track of the biggest change
102 * we've seen, and the longest filename. That allows us to do simple
103 * scaling.
104 */
105 static int max_change, max_len;
106
107 /*
108 * Various "current state", notably line numbers and what
109 * file (and how) we're patching right now.. The "is_xxxx"
110 * things are flags, where -1 means "don't know yet".
111 */
112 static int linenr = 1;
113
114 /*
115 * This represents one "hunk" from a patch, starting with
116 * "@@ -oldpos,oldlines +newpos,newlines @@" marker. The
117 * patch text is pointed at by patch, and its byte length
118 * is stored in size. leading and trailing are the number
119 * of context lines.
120 */
121 struct fragment {
122 unsigned long leading, trailing;
123 unsigned long oldpos, oldlines;
124 unsigned long newpos, newlines;
125 const char *patch;
126 int size;
127 int rejected;
128 struct fragment *next;
129 };
130
131 /*
132 * When dealing with a binary patch, we reuse "leading" field
133 * to store the type of the binary hunk, either deflated "delta"
134 * or deflated "literal".
135 */
136 #define binary_patch_method leading
137 #define BINARY_DELTA_DEFLATED 1
138 #define BINARY_LITERAL_DEFLATED 2
139
140 struct patch {
141 char *new_name, *old_name, *def_name;
142 unsigned int old_mode, new_mode;
143 int is_new, is_delete; /* -1 = unknown, 0 = false, 1 = true */
144 int rejected;
145 unsigned long deflate_origlen;
146 int lines_added, lines_deleted;
147 int score;
148 unsigned int inaccurate_eof:1;
149 unsigned int is_binary:1;
150 unsigned int is_copy:1;
151 unsigned int is_rename:1;
152 struct fragment *fragments;
153 char *result;
154 unsigned long resultsize;
155 char old_sha1_prefix[41];
156 char new_sha1_prefix[41];
157 struct patch *next;
158 };
159
160 static void say_patch_name(FILE *output, const char *pre, struct patch *patch, const char *post)
161 {
162 fputs(pre, output);
163 if (patch->old_name && patch->new_name &&
164 strcmp(patch->old_name, patch->new_name)) {
165 write_name_quoted(NULL, 0, patch->old_name, 1, output);
166 fputs(" => ", output);
167 write_name_quoted(NULL, 0, patch->new_name, 1, output);
168 }
169 else {
170 const char *n = patch->new_name;
171 if (!n)
172 n = patch->old_name;
173 write_name_quoted(NULL, 0, n, 1, output);
174 }
175 fputs(post, output);
176 }
177
178 #define CHUNKSIZE (8192)
179 #define SLOP (16)
180
181 static void *read_patch_file(int fd, unsigned long *sizep)
182 {
183 unsigned long size = 0, alloc = CHUNKSIZE;
184 void *buffer = xmalloc(alloc);
185
186 for (;;) {
187 int nr = alloc - size;
188 if (nr < 1024) {
189 alloc += CHUNKSIZE;
190 buffer = xrealloc(buffer, alloc);
191 nr = alloc - size;
192 }
193 nr = xread(fd, (char *) buffer + size, nr);
194 if (!nr)
195 break;
196 if (nr < 0)
197 die("git-apply: read returned %s", strerror(errno));
198 size += nr;
199 }
200 *sizep = size;
201
202 /*
203 * Make sure that we have some slop in the buffer
204 * so that we can do speculative "memcmp" etc, and
205 * see to it that it is NUL-filled.
206 */
207 if (alloc < size + SLOP)
208 buffer = xrealloc(buffer, size + SLOP);
209 memset((char *) buffer + size, 0, SLOP);
210 return buffer;
211 }
212
213 static unsigned long linelen(const char *buffer, unsigned long size)
214 {
215 unsigned long len = 0;
216 while (size--) {
217 len++;
218 if (*buffer++ == '\n')
219 break;
220 }
221 return len;
222 }
223
224 static int is_dev_null(const char *str)
225 {
226 return !memcmp("/dev/null", str, 9) && isspace(str[9]);
227 }
228
229 #define TERM_SPACE 1
230 #define TERM_TAB 2
231
232 static int name_terminate(const char *name, int namelen, int c, int terminate)
233 {
234 if (c == ' ' && !(terminate & TERM_SPACE))
235 return 0;
236 if (c == '\t' && !(terminate & TERM_TAB))
237 return 0;
238
239 return 1;
240 }
241
242 static char * find_name(const char *line, char *def, int p_value, int terminate)
243 {
244 int len;
245 const char *start = line;
246 char *name;
247
248 if (*line == '"') {
249 /* Proposed "new-style" GNU patch/diff format; see
250 * http://marc.theaimsgroup.com/?l=git&m=112927316408690&w=2
251 */
252 name = unquote_c_style(line, NULL);
253 if (name) {
254 char *cp = name;
255 while (p_value) {
256 cp = strchr(name, '/');
257 if (!cp)
258 break;
259 cp++;
260 p_value--;
261 }
262 if (cp) {
263 /* name can later be freed, so we need
264 * to memmove, not just return cp
265 */
266 memmove(name, cp, strlen(cp) + 1);
267 free(def);
268 return name;
269 }
270 else {
271 free(name);
272 name = NULL;
273 }
274 }
275 }
276
277 for (;;) {
278 char c = *line;
279
280 if (isspace(c)) {
281 if (c == '\n')
282 break;
283 if (name_terminate(start, line-start, c, terminate))
284 break;
285 }
286 line++;
287 if (c == '/' && !--p_value)
288 start = line;
289 }
290 if (!start)
291 return def;
292 len = line - start;
293 if (!len)
294 return def;
295
296 /*
297 * Generally we prefer the shorter name, especially
298 * if the other one is just a variation of that with
299 * something else tacked on to the end (ie "file.orig"
300 * or "file~").
301 */
302 if (def) {
303 int deflen = strlen(def);
304 if (deflen < len && !strncmp(start, def, deflen))
305 return def;
306 }
307
308 name = xmalloc(len + 1);
309 memcpy(name, start, len);
310 name[len] = 0;
311 free(def);
312 return name;
313 }
314
315 /*
316 * Get the name etc info from the --/+++ lines of a traditional patch header
317 *
318 * NOTE! This hardcodes "-p1" behaviour in filename detection.
319 *
320 * FIXME! The end-of-filename heuristics are kind of screwy. For existing
321 * files, we can happily check the index for a match, but for creating a
322 * new file we should try to match whatever "patch" does. I have no idea.
323 */
324 static void parse_traditional_patch(const char *first, const char *second, struct patch *patch)
325 {
326 char *name;
327
328 first += 4; /* skip "--- " */
329 second += 4; /* skip "+++ " */
330 if (is_dev_null(first)) {
331 patch->is_new = 1;
332 patch->is_delete = 0;
333 name = find_name(second, NULL, p_value, TERM_SPACE | TERM_TAB);
334 patch->new_name = name;
335 } else if (is_dev_null(second)) {
336 patch->is_new = 0;
337 patch->is_delete = 1;
338 name = find_name(first, NULL, p_value, TERM_SPACE | TERM_TAB);
339 patch->old_name = name;
340 } else {
341 name = find_name(first, NULL, p_value, TERM_SPACE | TERM_TAB);
342 name = find_name(second, name, p_value, TERM_SPACE | TERM_TAB);
343 patch->old_name = patch->new_name = name;
344 }
345 if (!name)
346 die("unable to find filename in patch at line %d", linenr);
347 }
348
349 static int gitdiff_hdrend(const char *line, struct patch *patch)
350 {
351 return -1;
352 }
353
354 /*
355 * We're anal about diff header consistency, to make
356 * sure that we don't end up having strange ambiguous
357 * patches floating around.
358 *
359 * As a result, gitdiff_{old|new}name() will check
360 * their names against any previous information, just
361 * to make sure..
362 */
363 static char *gitdiff_verify_name(const char *line, int isnull, char *orig_name, const char *oldnew)
364 {
365 if (!orig_name && !isnull)
366 return find_name(line, NULL, 1, TERM_TAB);
367
368 if (orig_name) {
369 int len;
370 const char *name;
371 char *another;
372 name = orig_name;
373 len = strlen(name);
374 if (isnull)
375 die("git-apply: bad git-diff - expected /dev/null, got %s on line %d", name, linenr);
376 another = find_name(line, NULL, 1, TERM_TAB);
377 if (!another || memcmp(another, name, len))
378 die("git-apply: bad git-diff - inconsistent %s filename on line %d", oldnew, linenr);
379 free(another);
380 return orig_name;
381 }
382 else {
383 /* expect "/dev/null" */
384 if (memcmp("/dev/null", line, 9) || line[9] != '\n')
385 die("git-apply: bad git-diff - expected /dev/null on line %d", linenr);
386 return NULL;
387 }
388 }
389
390 static int gitdiff_oldname(const char *line, struct patch *patch)
391 {
392 patch->old_name = gitdiff_verify_name(line, patch->is_new, patch->old_name, "old");
393 return 0;
394 }
395
396 static int gitdiff_newname(const char *line, struct patch *patch)
397 {
398 patch->new_name = gitdiff_verify_name(line, patch->is_delete, patch->new_name, "new");
399 return 0;
400 }
401
402 static int gitdiff_oldmode(const char *line, struct patch *patch)
403 {
404 patch->old_mode = strtoul(line, NULL, 8);
405 return 0;
406 }
407
408 static int gitdiff_newmode(const char *line, struct patch *patch)
409 {
410 patch->new_mode = strtoul(line, NULL, 8);
411 return 0;
412 }
413
414 static int gitdiff_delete(const char *line, struct patch *patch)
415 {
416 patch->is_delete = 1;
417 patch->old_name = patch->def_name;
418 return gitdiff_oldmode(line, patch);
419 }
420
421 static int gitdiff_newfile(const char *line, struct patch *patch)
422 {
423 patch->is_new = 1;
424 patch->new_name = patch->def_name;
425 return gitdiff_newmode(line, patch);
426 }
427
428 static int gitdiff_copysrc(const char *line, struct patch *patch)
429 {
430 patch->is_copy = 1;
431 patch->old_name = find_name(line, NULL, 0, 0);
432 return 0;
433 }
434
435 static int gitdiff_copydst(const char *line, struct patch *patch)
436 {
437 patch->is_copy = 1;
438 patch->new_name = find_name(line, NULL, 0, 0);
439 return 0;
440 }
441
442 static int gitdiff_renamesrc(const char *line, struct patch *patch)
443 {
444 patch->is_rename = 1;
445 patch->old_name = find_name(line, NULL, 0, 0);
446 return 0;
447 }
448
449 static int gitdiff_renamedst(const char *line, struct patch *patch)
450 {
451 patch->is_rename = 1;
452 patch->new_name = find_name(line, NULL, 0, 0);
453 return 0;
454 }
455
456 static int gitdiff_similarity(const char *line, struct patch *patch)
457 {
458 if ((patch->score = strtoul(line, NULL, 10)) == ULONG_MAX)
459 patch->score = 0;
460 return 0;
461 }
462
463 static int gitdiff_dissimilarity(const char *line, struct patch *patch)
464 {
465 if ((patch->score = strtoul(line, NULL, 10)) == ULONG_MAX)
466 patch->score = 0;
467 return 0;
468 }
469
470 static int gitdiff_index(const char *line, struct patch *patch)
471 {
472 /* index line is N hexadecimal, "..", N hexadecimal,
473 * and optional space with octal mode.
474 */
475 const char *ptr, *eol;
476 int len;
477
478 ptr = strchr(line, '.');
479 if (!ptr || ptr[1] != '.' || 40 < ptr - line)
480 return 0;
481 len = ptr - line;
482 memcpy(patch->old_sha1_prefix, line, len);
483 patch->old_sha1_prefix[len] = 0;
484
485 line = ptr + 2;
486 ptr = strchr(line, ' ');
487 eol = strchr(line, '\n');
488
489 if (!ptr || eol < ptr)
490 ptr = eol;
491 len = ptr - line;
492
493 if (40 < len)
494 return 0;
495 memcpy(patch->new_sha1_prefix, line, len);
496 patch->new_sha1_prefix[len] = 0;
497 if (*ptr == ' ')
498 patch->new_mode = patch->old_mode = strtoul(ptr+1, NULL, 8);
499 return 0;
500 }
501
502 /*
503 * This is normal for a diff that doesn't change anything: we'll fall through
504 * into the next diff. Tell the parser to break out.
505 */
506 static int gitdiff_unrecognized(const char *line, struct patch *patch)
507 {
508 return -1;
509 }
510
511 static const char *stop_at_slash(const char *line, int llen)
512 {
513 int i;
514
515 for (i = 0; i < llen; i++) {
516 int ch = line[i];
517 if (ch == '/')
518 return line + i;
519 }
520 return NULL;
521 }
522
523 /* This is to extract the same name that appears on "diff --git"
524 * line. We do not find and return anything if it is a rename
525 * patch, and it is OK because we will find the name elsewhere.
526 * We need to reliably find name only when it is mode-change only,
527 * creation or deletion of an empty file. In any of these cases,
528 * both sides are the same name under a/ and b/ respectively.
529 */
530 static char *git_header_name(char *line, int llen)
531 {
532 int len;
533 const char *name;
534 const char *second = NULL;
535
536 line += strlen("diff --git ");
537 llen -= strlen("diff --git ");
538
539 if (*line == '"') {
540 const char *cp;
541 char *first = unquote_c_style(line, &second);
542 if (!first)
543 return NULL;
544
545 /* advance to the first slash */
546 cp = stop_at_slash(first, strlen(first));
547 if (!cp || cp == first) {
548 /* we do not accept absolute paths */
549 free_first_and_fail:
550 free(first);
551 return NULL;
552 }
553 len = strlen(cp+1);
554 memmove(first, cp+1, len+1); /* including NUL */
555
556 /* second points at one past closing dq of name.
557 * find the second name.
558 */
559 while ((second < line + llen) && isspace(*second))
560 second++;
561
562 if (line + llen <= second)
563 goto free_first_and_fail;
564 if (*second == '"') {
565 char *sp = unquote_c_style(second, NULL);
566 if (!sp)
567 goto free_first_and_fail;
568 cp = stop_at_slash(sp, strlen(sp));
569 if (!cp || cp == sp) {
570 free_both_and_fail:
571 free(sp);
572 goto free_first_and_fail;
573 }
574 /* They must match, otherwise ignore */
575 if (strcmp(cp+1, first))
576 goto free_both_and_fail;
577 free(sp);
578 return first;
579 }
580
581 /* unquoted second */
582 cp = stop_at_slash(second, line + llen - second);
583 if (!cp || cp == second)
584 goto free_first_and_fail;
585 cp++;
586 if (line + llen - cp != len + 1 ||
587 memcmp(first, cp, len))
588 goto free_first_and_fail;
589 return first;
590 }
591
592 /* unquoted first name */
593 name = stop_at_slash(line, llen);
594 if (!name || name == line)
595 return NULL;
596
597 name++;
598
599 /* since the first name is unquoted, a dq if exists must be
600 * the beginning of the second name.
601 */
602 for (second = name; second < line + llen; second++) {
603 if (*second == '"') {
604 const char *cp = second;
605 const char *np;
606 char *sp = unquote_c_style(second, NULL);
607
608 if (!sp)
609 return NULL;
610 np = stop_at_slash(sp, strlen(sp));
611 if (!np || np == sp) {
612 free_second_and_fail:
613 free(sp);
614 return NULL;
615 }
616 np++;
617 len = strlen(np);
618 if (len < cp - name &&
619 !strncmp(np, name, len) &&
620 isspace(name[len])) {
621 /* Good */
622 memmove(sp, np, len + 1);
623 return sp;
624 }
625 goto free_second_and_fail;
626 }
627 }
628
629 /*
630 * Accept a name only if it shows up twice, exactly the same
631 * form.
632 */
633 for (len = 0 ; ; len++) {
634 switch (name[len]) {
635 default:
636 continue;
637 case '\n':
638 return NULL;
639 case '\t': case ' ':
640 second = name+len;
641 for (;;) {
642 char c = *second++;
643 if (c == '\n')
644 return NULL;
645 if (c == '/')
646 break;
647 }
648 if (second[len] == '\n' && !memcmp(name, second, len)) {
649 char *ret = xmalloc(len + 1);
650 memcpy(ret, name, len);
651 ret[len] = 0;
652 return ret;
653 }
654 }
655 }
656 return NULL;
657 }
658
659 /* Verify that we recognize the lines following a git header */
660 static int parse_git_header(char *line, int len, unsigned int size, struct patch *patch)
661 {
662 unsigned long offset;
663
664 /* A git diff has explicit new/delete information, so we don't guess */
665 patch->is_new = 0;
666 patch->is_delete = 0;
667
668 /*
669 * Some things may not have the old name in the
670 * rest of the headers anywhere (pure mode changes,
671 * or removing or adding empty files), so we get
672 * the default name from the header.
673 */
674 patch->def_name = git_header_name(line, len);
675
676 line += len;
677 size -= len;
678 linenr++;
679 for (offset = len ; size > 0 ; offset += len, size -= len, line += len, linenr++) {
680 static const struct opentry {
681 const char *str;
682 int (*fn)(const char *, struct patch *);
683 } optable[] = {
684 { "@@ -", gitdiff_hdrend },
685 { "--- ", gitdiff_oldname },
686 { "+++ ", gitdiff_newname },
687 { "old mode ", gitdiff_oldmode },
688 { "new mode ", gitdiff_newmode },
689 { "deleted file mode ", gitdiff_delete },
690 { "new file mode ", gitdiff_newfile },
691 { "copy from ", gitdiff_copysrc },
692 { "copy to ", gitdiff_copydst },
693 { "rename old ", gitdiff_renamesrc },
694 { "rename new ", gitdiff_renamedst },
695 { "rename from ", gitdiff_renamesrc },
696 { "rename to ", gitdiff_renamedst },
697 { "similarity index ", gitdiff_similarity },
698 { "dissimilarity index ", gitdiff_dissimilarity },
699 { "index ", gitdiff_index },
700 { "", gitdiff_unrecognized },
701 };
702 int i;
703
704 len = linelen(line, size);
705 if (!len || line[len-1] != '\n')
706 break;
707 for (i = 0; i < ARRAY_SIZE(optable); i++) {
708 const struct opentry *p = optable + i;
709 int oplen = strlen(p->str);
710 if (len < oplen || memcmp(p->str, line, oplen))
711 continue;
712 if (p->fn(line + oplen, patch) < 0)
713 return offset;
714 break;
715 }
716 }
717
718 return offset;
719 }
720
721 static int parse_num(const char *line, unsigned long *p)
722 {
723 char *ptr;
724
725 if (!isdigit(*line))
726 return 0;
727 *p = strtoul(line, &ptr, 10);
728 return ptr - line;
729 }
730
731 static int parse_range(const char *line, int len, int offset, const char *expect,
732 unsigned long *p1, unsigned long *p2)
733 {
734 int digits, ex;
735
736 if (offset < 0 || offset >= len)
737 return -1;
738 line += offset;
739 len -= offset;
740
741 digits = parse_num(line, p1);
742 if (!digits)
743 return -1;
744
745 offset += digits;
746 line += digits;
747 len -= digits;
748
749 *p2 = 1;
750 if (*line == ',') {
751 digits = parse_num(line+1, p2);
752 if (!digits)
753 return -1;
754
755 offset += digits+1;
756 line += digits+1;
757 len -= digits+1;
758 }
759
760 ex = strlen(expect);
761 if (ex > len)
762 return -1;
763 if (memcmp(line, expect, ex))
764 return -1;
765
766 return offset + ex;
767 }
768
769 /*
770 * Parse a unified diff fragment header of the
771 * form "@@ -a,b +c,d @@"
772 */
773 static int parse_fragment_header(char *line, int len, struct fragment *fragment)
774 {
775 int offset;
776
777 if (!len || line[len-1] != '\n')
778 return -1;
779
780 /* Figure out the number of lines in a fragment */
781 offset = parse_range(line, len, 4, " +", &fragment->oldpos, &fragment->oldlines);
782 offset = parse_range(line, len, offset, " @@", &fragment->newpos, &fragment->newlines);
783
784 return offset;
785 }
786
787 static int find_header(char *line, unsigned long size, int *hdrsize, struct patch *patch)
788 {
789 unsigned long offset, len;
790
791 patch->is_rename = patch->is_copy = 0;
792 patch->is_new = patch->is_delete = -1;
793 patch->old_mode = patch->new_mode = 0;
794 patch->old_name = patch->new_name = NULL;
795 for (offset = 0; size > 0; offset += len, size -= len, line += len, linenr++) {
796 unsigned long nextlen;
797
798 len = linelen(line, size);
799 if (!len)
800 break;
801
802 /* Testing this early allows us to take a few shortcuts.. */
803 if (len < 6)
804 continue;
805
806 /*
807 * Make sure we don't find any unconnected patch fragments.
808 * That's a sign that we didn't find a header, and that a
809 * patch has become corrupted/broken up.
810 */
811 if (!memcmp("@@ -", line, 4)) {
812 struct fragment dummy;
813 if (parse_fragment_header(line, len, &dummy) < 0)
814 continue;
815 return error("patch fragment without header at line %d: %.*s", linenr, (int)len-1, line);
816 }
817
818 if (size < len + 6)
819 break;
820
821 /*
822 * Git patch? It might not have a real patch, just a rename
823 * or mode change, so we handle that specially
824 */
825 if (!memcmp("diff --git ", line, 11)) {
826 int git_hdr_len = parse_git_header(line, len, size, patch);
827 if (git_hdr_len <= len)
828 continue;
829 if (!patch->old_name && !patch->new_name) {
830 if (!patch->def_name)
831 die("git diff header lacks filename information (line %d)", linenr);
832 patch->old_name = patch->new_name = patch->def_name;
833 }
834 *hdrsize = git_hdr_len;
835 return offset;
836 }
837
838 /** --- followed by +++ ? */
839 if (memcmp("--- ", line, 4) || memcmp("+++ ", line + len, 4))
840 continue;
841
842 /*
843 * We only accept unified patches, so we want it to
844 * at least have "@@ -a,b +c,d @@\n", which is 14 chars
845 * minimum
846 */
847 nextlen = linelen(line + len, size - len);
848 if (size < nextlen + 14 || memcmp("@@ -", line + len + nextlen, 4))
849 continue;
850
851 /* Ok, we'll consider it a patch */
852 parse_traditional_patch(line, line+len, patch);
853 *hdrsize = len + nextlen;
854 linenr += 2;
855 return offset;
856 }
857 return -1;
858 }
859
860 static void check_whitespace(const char *line, int len)
861 {
862 const char *err = "Adds trailing whitespace";
863 int seen_space = 0;
864 int i;
865
866 /*
867 * We know len is at least two, since we have a '+' and we
868 * checked that the last character was a '\n' before calling
869 * this function. That is, an addition of an empty line would
870 * check the '+' here. Sneaky...
871 */
872 if (isspace(line[len-2]))
873 goto error;
874
875 /*
876 * Make sure that there is no space followed by a tab in
877 * indentation.
878 */
879 err = "Space in indent is followed by a tab";
880 for (i = 1; i < len; i++) {
881 if (line[i] == '\t') {
882 if (seen_space)
883 goto error;
884 }
885 else if (line[i] == ' ')
886 seen_space = 1;
887 else
888 break;
889 }
890 return;
891
892 error:
893 whitespace_error++;
894 if (squelch_whitespace_errors &&
895 squelch_whitespace_errors < whitespace_error)
896 ;
897 else
898 fprintf(stderr, "%s.\n%s:%d:%.*s\n",
899 err, patch_input_file, linenr, len-2, line+1);
900 }
901
902
903 /*
904 * Parse a unified diff. Note that this really needs to parse each
905 * fragment separately, since the only way to know the difference
906 * between a "---" that is part of a patch, and a "---" that starts
907 * the next patch is to look at the line counts..
908 */
909 static int parse_fragment(char *line, unsigned long size, struct patch *patch, struct fragment *fragment)
910 {
911 int added, deleted;
912 int len = linelen(line, size), offset;
913 unsigned long oldlines, newlines;
914 unsigned long leading, trailing;
915
916 offset = parse_fragment_header(line, len, fragment);
917 if (offset < 0)
918 return -1;
919 oldlines = fragment->oldlines;
920 newlines = fragment->newlines;
921 leading = 0;
922 trailing = 0;
923
924 /* Parse the thing.. */
925 line += len;
926 size -= len;
927 linenr++;
928 added = deleted = 0;
929 for (offset = len;
930 0 < size;
931 offset += len, size -= len, line += len, linenr++) {
932 if (!oldlines && !newlines)
933 break;
934 len = linelen(line, size);
935 if (!len || line[len-1] != '\n')
936 return -1;
937 switch (*line) {
938 default:
939 return -1;
940 case '\n': /* newer GNU diff, an empty context line */
941 case ' ':
942 oldlines--;
943 newlines--;
944 if (!deleted && !added)
945 leading++;
946 trailing++;
947 break;
948 case '-':
949 deleted++;
950 oldlines--;
951 trailing = 0;
952 break;
953 case '+':
954 if (new_whitespace != nowarn_whitespace)
955 check_whitespace(line, len);
956 added++;
957 newlines--;
958 trailing = 0;
959 break;
960
961 /* We allow "\ No newline at end of file". Depending
962 * on locale settings when the patch was produced we
963 * don't know what this line looks like. The only
964 * thing we do know is that it begins with "\ ".
965 * Checking for 12 is just for sanity check -- any
966 * l10n of "\ No newline..." is at least that long.
967 */
968 case '\\':
969 if (len < 12 || memcmp(line, "\\ ", 2))
970 return -1;
971 break;
972 }
973 }
974 if (oldlines || newlines)
975 return -1;
976 fragment->leading = leading;
977 fragment->trailing = trailing;
978
979 /* If a fragment ends with an incomplete line, we failed to include
980 * it in the above loop because we hit oldlines == newlines == 0
981 * before seeing it.
982 */
983 if (12 < size && !memcmp(line, "\\ ", 2))
984 offset += linelen(line, size);
985
986 patch->lines_added += added;
987 patch->lines_deleted += deleted;
988
989 if (0 < patch->is_new && oldlines)
990 return error("new file depends on old contents");
991 if (0 < patch->is_delete && newlines)
992 return error("deleted file still has contents");
993 return offset;
994 }
995
996 static int parse_single_patch(char *line, unsigned long size, struct patch *patch)
997 {
998 unsigned long offset = 0;
999 unsigned long oldlines = 0, newlines = 0, context = 0;
1000 struct fragment **fragp = &patch->fragments;
1001
1002 while (size > 4 && !memcmp(line, "@@ -", 4)) {
1003 struct fragment *fragment;
1004 int len;
1005
1006 fragment = xcalloc(1, sizeof(*fragment));
1007 len = parse_fragment(line, size, patch, fragment);
1008 if (len <= 0)
1009 die("corrupt patch at line %d", linenr);
1010 fragment->patch = line;
1011 fragment->size = len;
1012 oldlines += fragment->oldlines;
1013 newlines += fragment->newlines;
1014 context += fragment->leading + fragment->trailing;
1015
1016 *fragp = fragment;
1017 fragp = &fragment->next;
1018
1019 offset += len;
1020 line += len;
1021 size -= len;
1022 }
1023
1024 /*
1025 * If something was removed (i.e. we have old-lines) it cannot
1026 * be creation, and if something was added it cannot be
1027 * deletion. However, the reverse is not true; --unified=0
1028 * patches that only add are not necessarily creation even
1029 * though they do not have any old lines, and ones that only
1030 * delete are not necessarily deletion.
1031 *
1032 * Unfortunately, a real creation/deletion patch do _not_ have
1033 * any context line by definition, so we cannot safely tell it
1034 * apart with --unified=0 insanity. At least if the patch has
1035 * more than one hunk it is not creation or deletion.
1036 */
1037 if (patch->is_new < 0 &&
1038 (oldlines || (patch->fragments && patch->fragments->next)))
1039 patch->is_new = 0;
1040 if (patch->is_delete < 0 &&
1041 (newlines || (patch->fragments && patch->fragments->next)))
1042 patch->is_delete = 0;
1043 if (!unidiff_zero || context) {
1044 /* If the user says the patch is not generated with
1045 * --unified=0, or if we have seen context lines,
1046 * then not having oldlines means the patch is creation,
1047 * and not having newlines means the patch is deletion.
1048 */
1049 if (patch->is_new < 0 && !oldlines) {
1050 patch->is_new = 1;
1051 patch->old_name = NULL;
1052 }
1053 if (patch->is_delete < 0 && !newlines) {
1054 patch->is_delete = 1;
1055 patch->new_name = NULL;
1056 }
1057 }
1058
1059 if (0 < patch->is_new && oldlines)
1060 die("new file %s depends on old contents", patch->new_name);
1061 if (0 < patch->is_delete && newlines)
1062 die("deleted file %s still has contents", patch->old_name);
1063 if (!patch->is_delete && !newlines && context)
1064 fprintf(stderr, "** warning: file %s becomes empty but "
1065 "is not deleted\n", patch->new_name);
1066
1067 return offset;
1068 }
1069
1070 static inline int metadata_changes(struct patch *patch)
1071 {
1072 return patch->is_rename > 0 ||
1073 patch->is_copy > 0 ||
1074 patch->is_new > 0 ||
1075 patch->is_delete ||
1076 (patch->old_mode && patch->new_mode &&
1077 patch->old_mode != patch->new_mode);
1078 }
1079
1080 static char *inflate_it(const void *data, unsigned long size,
1081 unsigned long inflated_size)
1082 {
1083 z_stream stream;
1084 void *out;
1085 int st;
1086
1087 memset(&stream, 0, sizeof(stream));
1088
1089 stream.next_in = (unsigned char *)data;
1090 stream.avail_in = size;
1091 stream.next_out = out = xmalloc(inflated_size);
1092 stream.avail_out = inflated_size;
1093 inflateInit(&stream);
1094 st = inflate(&stream, Z_FINISH);
1095 if ((st != Z_STREAM_END) || stream.total_out != inflated_size) {
1096 free(out);
1097 return NULL;
1098 }
1099 return out;
1100 }
1101
1102 static struct fragment *parse_binary_hunk(char **buf_p,
1103 unsigned long *sz_p,
1104 int *status_p,
1105 int *used_p)
1106 {
1107 /* Expect a line that begins with binary patch method ("literal"
1108 * or "delta"), followed by the length of data before deflating.
1109 * a sequence of 'length-byte' followed by base-85 encoded data
1110 * should follow, terminated by a newline.
1111 *
1112 * Each 5-byte sequence of base-85 encodes up to 4 bytes,
1113 * and we would limit the patch line to 66 characters,
1114 * so one line can fit up to 13 groups that would decode
1115 * to 52 bytes max. The length byte 'A'-'Z' corresponds
1116 * to 1-26 bytes, and 'a'-'z' corresponds to 27-52 bytes.
1117 */
1118 int llen, used;
1119 unsigned long size = *sz_p;
1120 char *buffer = *buf_p;
1121 int patch_method;
1122 unsigned long origlen;
1123 char *data = NULL;
1124 int hunk_size = 0;
1125 struct fragment *frag;
1126
1127 llen = linelen(buffer, size);
1128 used = llen;
1129
1130 *status_p = 0;
1131
1132 if (!strncmp(buffer, "delta ", 6)) {
1133 patch_method = BINARY_DELTA_DEFLATED;
1134 origlen = strtoul(buffer + 6, NULL, 10);
1135 }
1136 else if (!strncmp(buffer, "literal ", 8)) {
1137 patch_method = BINARY_LITERAL_DEFLATED;
1138 origlen = strtoul(buffer + 8, NULL, 10);
1139 }
1140 else
1141 return NULL;
1142
1143 linenr++;
1144 buffer += llen;
1145 while (1) {
1146 int byte_length, max_byte_length, newsize;
1147 llen = linelen(buffer, size);
1148 used += llen;
1149 linenr++;
1150 if (llen == 1) {
1151 /* consume the blank line */
1152 buffer++;
1153 size--;
1154 break;
1155 }
1156 /* Minimum line is "A00000\n" which is 7-byte long,
1157 * and the line length must be multiple of 5 plus 2.
1158 */
1159 if ((llen < 7) || (llen-2) % 5)
1160 goto corrupt;
1161 max_byte_length = (llen - 2) / 5 * 4;
1162 byte_length = *buffer;
1163 if ('A' <= byte_length && byte_length <= 'Z')
1164 byte_length = byte_length - 'A' + 1;
1165 else if ('a' <= byte_length && byte_length <= 'z')
1166 byte_length = byte_length - 'a' + 27;
1167 else
1168 goto corrupt;
1169 /* if the input length was not multiple of 4, we would
1170 * have filler at the end but the filler should never
1171 * exceed 3 bytes
1172 */
1173 if (max_byte_length < byte_length ||
1174 byte_length <= max_byte_length - 4)
1175 goto corrupt;
1176 newsize = hunk_size + byte_length;
1177 data = xrealloc(data, newsize);
1178 if (decode_85(data + hunk_size, buffer + 1, byte_length))
1179 goto corrupt;
1180 hunk_size = newsize;
1181 buffer += llen;
1182 size -= llen;
1183 }
1184
1185 frag = xcalloc(1, sizeof(*frag));
1186 frag->patch = inflate_it(data, hunk_size, origlen);
1187 if (!frag->patch)
1188 goto corrupt;
1189 free(data);
1190 frag->size = origlen;
1191 *buf_p = buffer;
1192 *sz_p = size;
1193 *used_p = used;
1194 frag->binary_patch_method = patch_method;
1195 return frag;
1196
1197 corrupt:
1198 free(data);
1199 *status_p = -1;
1200 error("corrupt binary patch at line %d: %.*s",
1201 linenr-1, llen-1, buffer);
1202 return NULL;
1203 }
1204
1205 static int parse_binary(char *buffer, unsigned long size, struct patch *patch)
1206 {
1207 /* We have read "GIT binary patch\n"; what follows is a line
1208 * that says the patch method (currently, either "literal" or
1209 * "delta") and the length of data before deflating; a
1210 * sequence of 'length-byte' followed by base-85 encoded data
1211 * follows.
1212 *
1213 * When a binary patch is reversible, there is another binary
1214 * hunk in the same format, starting with patch method (either
1215 * "literal" or "delta") with the length of data, and a sequence
1216 * of length-byte + base-85 encoded data, terminated with another
1217 * empty line. This data, when applied to the postimage, produces
1218 * the preimage.
1219 */
1220 struct fragment *forward;
1221 struct fragment *reverse;
1222 int status;
1223 int used, used_1;
1224
1225 forward = parse_binary_hunk(&buffer, &size, &status, &used);
1226 if (!forward && !status)
1227 /* there has to be one hunk (forward hunk) */
1228 return error("unrecognized binary patch at line %d", linenr-1);
1229 if (status)
1230 /* otherwise we already gave an error message */
1231 return status;
1232
1233 reverse = parse_binary_hunk(&buffer, &size, &status, &used_1);
1234 if (reverse)
1235 used += used_1;
1236 else if (status) {
1237 /* not having reverse hunk is not an error, but having
1238 * a corrupt reverse hunk is.
1239 */
1240 free((void*) forward->patch);
1241 free(forward);
1242 return status;
1243 }
1244 forward->next = reverse;
1245 patch->fragments = forward;
1246 patch->is_binary = 1;
1247 return used;
1248 }
1249
1250 static int parse_chunk(char *buffer, unsigned long size, struct patch *patch)
1251 {
1252 int hdrsize, patchsize;
1253 int offset = find_header(buffer, size, &hdrsize, patch);
1254
1255 if (offset < 0)
1256 return offset;
1257
1258 patchsize = parse_single_patch(buffer + offset + hdrsize, size - offset - hdrsize, patch);
1259
1260 if (!patchsize) {
1261 static const char *binhdr[] = {
1262 "Binary files ",
1263 "Files ",
1264 NULL,
1265 };
1266 static const char git_binary[] = "GIT binary patch\n";
1267 int i;
1268 int hd = hdrsize + offset;
1269 unsigned long llen = linelen(buffer + hd, size - hd);
1270
1271 if (llen == sizeof(git_binary) - 1 &&
1272 !memcmp(git_binary, buffer + hd, llen)) {
1273 int used;
1274 linenr++;
1275 used = parse_binary(buffer + hd + llen,
1276 size - hd - llen, patch);
1277 if (used)
1278 patchsize = used + llen;
1279 else
1280 patchsize = 0;
1281 }
1282 else if (!memcmp(" differ\n", buffer + hd + llen - 8, 8)) {
1283 for (i = 0; binhdr[i]; i++) {
1284 int len = strlen(binhdr[i]);
1285 if (len < size - hd &&
1286 !memcmp(binhdr[i], buffer + hd, len)) {
1287 linenr++;
1288 patch->is_binary = 1;
1289 patchsize = llen;
1290 break;
1291 }
1292 }
1293 }
1294
1295 /* Empty patch cannot be applied if it is a text patch
1296 * without metadata change. A binary patch appears
1297 * empty to us here.
1298 */
1299 if ((apply || check) &&
1300 (!patch->is_binary && !metadata_changes(patch)))
1301 die("patch with only garbage at line %d", linenr);
1302 }
1303
1304 return offset + hdrsize + patchsize;
1305 }
1306
1307 #define swap(a,b) myswap((a),(b),sizeof(a))
1308
1309 #define myswap(a, b, size) do { \
1310 unsigned char mytmp[size]; \
1311 memcpy(mytmp, &a, size); \
1312 memcpy(&a, &b, size); \
1313 memcpy(&b, mytmp, size); \
1314 } while (0)
1315
1316 static void reverse_patches(struct patch *p)
1317 {
1318 for (; p; p = p->next) {
1319 struct fragment *frag = p->fragments;
1320
1321 swap(p->new_name, p->old_name);
1322 swap(p->new_mode, p->old_mode);
1323 swap(p->is_new, p->is_delete);
1324 swap(p->lines_added, p->lines_deleted);
1325 swap(p->old_sha1_prefix, p->new_sha1_prefix);
1326
1327 for (; frag; frag = frag->next) {
1328 swap(frag->newpos, frag->oldpos);
1329 swap(frag->newlines, frag->oldlines);
1330 }
1331 }
1332 }
1333
1334 static const char pluses[] = "++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++";
1335 static const char minuses[]= "----------------------------------------------------------------------";
1336
1337 static void show_stats(struct patch *patch)
1338 {
1339 const char *prefix = "";
1340 char *name = patch->new_name;
1341 char *qname = NULL;
1342 int len, max, add, del, total;
1343
1344 if (!name)
1345 name = patch->old_name;
1346
1347 if (0 < (len = quote_c_style(name, NULL, NULL, 0))) {
1348 qname = xmalloc(len + 1);
1349 quote_c_style(name, qname, NULL, 0);
1350 name = qname;
1351 }
1352
1353 /*
1354 * "scale" the filename
1355 */
1356 len = strlen(name);
1357 max = max_len;
1358 if (max > 50)
1359 max = 50;
1360 if (len > max) {
1361 char *slash;
1362 prefix = "...";
1363 max -= 3;
1364 name += len - max;
1365 slash = strchr(name, '/');
1366 if (slash)
1367 name = slash;
1368 }
1369 len = max;
1370
1371 /*
1372 * scale the add/delete
1373 */
1374 max = max_change;
1375 if (max + len > 70)
1376 max = 70 - len;
1377
1378 add = patch->lines_added;
1379 del = patch->lines_deleted;
1380 total = add + del;
1381
1382 if (max_change > 0) {
1383 total = (total * max + max_change / 2) / max_change;
1384 add = (add * max + max_change / 2) / max_change;
1385 del = total - add;
1386 }
1387 if (patch->is_binary)
1388 printf(" %s%-*s | Bin\n", prefix, len, name);
1389 else
1390 printf(" %s%-*s |%5d %.*s%.*s\n", prefix,
1391 len, name, patch->lines_added + patch->lines_deleted,
1392 add, pluses, del, minuses);
1393 free(qname);
1394 }
1395
1396 static int read_old_data(struct stat *st, const char *path, void *buf, unsigned long size)
1397 {
1398 int fd;
1399 unsigned long got;
1400
1401 switch (st->st_mode & S_IFMT) {
1402 case S_IFLNK:
1403 return readlink(path, buf, size);
1404 case S_IFREG:
1405 fd = open(path, O_RDONLY);
1406 if (fd < 0)
1407 return error("unable to open %s", path);
1408 got = 0;
1409 for (;;) {
1410 int ret = xread(fd, (char *) buf + got, size - got);
1411 if (ret <= 0)
1412 break;
1413 got += ret;
1414 }
1415 close(fd);
1416 return got;
1417
1418 default:
1419 return -1;
1420 }
1421 }
1422
1423 static int find_offset(const char *buf, unsigned long size, const char *fragment, unsigned long fragsize, int line, int *lines)
1424 {
1425 int i;
1426 unsigned long start, backwards, forwards;
1427
1428 if (fragsize > size)
1429 return -1;
1430
1431 start = 0;
1432 if (line > 1) {
1433 unsigned long offset = 0;
1434 i = line-1;
1435 while (offset + fragsize <= size) {
1436 if (buf[offset++] == '\n') {
1437 start = offset;
1438 if (!--i)
1439 break;
1440 }
1441 }
1442 }
1443
1444 /* Exact line number? */
1445 if (!memcmp(buf + start, fragment, fragsize))
1446 return start;
1447
1448 /*
1449 * There's probably some smart way to do this, but I'll leave
1450 * that to the smart and beautiful people. I'm simple and stupid.
1451 */
1452 backwards = start;
1453 forwards = start;
1454 for (i = 0; ; i++) {
1455 unsigned long try;
1456 int n;
1457
1458 /* "backward" */
1459 if (i & 1) {
1460 if (!backwards) {
1461 if (forwards + fragsize > size)
1462 break;
1463 continue;
1464 }
1465 do {
1466 --backwards;
1467 } while (backwards && buf[backwards-1] != '\n');
1468 try = backwards;
1469 } else {
1470 while (forwards + fragsize <= size) {
1471 if (buf[forwards++] == '\n')
1472 break;
1473 }
1474 try = forwards;
1475 }
1476
1477 if (try + fragsize > size)
1478 continue;
1479 if (memcmp(buf + try, fragment, fragsize))
1480 continue;
1481 n = (i >> 1)+1;
1482 if (i & 1)
1483 n = -n;
1484 *lines = n;
1485 return try;
1486 }
1487
1488 /*
1489 * We should start searching forward and backward.
1490 */
1491 return -1;
1492 }
1493
1494 static void remove_first_line(const char **rbuf, int *rsize)
1495 {
1496 const char *buf = *rbuf;
1497 int size = *rsize;
1498 unsigned long offset;
1499 offset = 0;
1500 while (offset <= size) {
1501 if (buf[offset++] == '\n')
1502 break;
1503 }
1504 *rsize = size - offset;
1505 *rbuf = buf + offset;
1506 }
1507
1508 static void remove_last_line(const char **rbuf, int *rsize)
1509 {
1510 const char *buf = *rbuf;
1511 int size = *rsize;
1512 unsigned long offset;
1513 offset = size - 1;
1514 while (offset > 0) {
1515 if (buf[--offset] == '\n')
1516 break;
1517 }
1518 *rsize = offset + 1;
1519 }
1520
1521 struct buffer_desc {
1522 char *buffer;
1523 unsigned long size;
1524 unsigned long alloc;
1525 };
1526
1527 static int apply_line(char *output, const char *patch, int plen)
1528 {
1529 /* plen is number of bytes to be copied from patch,
1530 * starting at patch+1 (patch[0] is '+'). Typically
1531 * patch[plen] is '\n', unless this is the incomplete
1532 * last line.
1533 */
1534 int i;
1535 int add_nl_to_tail = 0;
1536 int fixed = 0;
1537 int last_tab_in_indent = -1;
1538 int last_space_in_indent = -1;
1539 int need_fix_leading_space = 0;
1540 char *buf;
1541
1542 if ((new_whitespace != strip_whitespace) || !whitespace_error) {
1543 memcpy(output, patch + 1, plen);
1544 return plen;
1545 }
1546
1547 if (1 < plen && isspace(patch[plen-1])) {
1548 if (patch[plen] == '\n')
1549 add_nl_to_tail = 1;
1550 plen--;
1551 while (0 < plen && isspace(patch[plen]))
1552 plen--;
1553 fixed = 1;
1554 }
1555
1556 for (i = 1; i < plen; i++) {
1557 char ch = patch[i];
1558 if (ch == '\t') {
1559 last_tab_in_indent = i;
1560 if (0 <= last_space_in_indent)
1561 need_fix_leading_space = 1;
1562 }
1563 else if (ch == ' ')
1564 last_space_in_indent = i;
1565 else
1566 break;
1567 }
1568
1569 buf = output;
1570 if (need_fix_leading_space) {
1571 /* between patch[1..last_tab_in_indent] strip the
1572 * funny spaces, updating them to tab as needed.
1573 */
1574 for (i = 1; i < last_tab_in_indent; i++, plen--) {
1575 char ch = patch[i];
1576 if (ch != ' ')
1577 *output++ = ch;
1578 else if ((i % 8) == 0)
1579 *output++ = '\t';
1580 }
1581 fixed = 1;
1582 i = last_tab_in_indent;
1583 }
1584 else
1585 i = 1;
1586
1587 memcpy(output, patch + i, plen);
1588 if (add_nl_to_tail)
1589 output[plen++] = '\n';
1590 if (fixed)
1591 applied_after_stripping++;
1592 return output + plen - buf;
1593 }
1594
1595 static int apply_one_fragment(struct buffer_desc *desc, struct fragment *frag, int inaccurate_eof)
1596 {
1597 int match_beginning, match_end;
1598 char *buf = desc->buffer;
1599 const char *patch = frag->patch;
1600 int offset, size = frag->size;
1601 char *old = xmalloc(size);
1602 char *new = xmalloc(size);
1603 const char *oldlines, *newlines;
1604 int oldsize = 0, newsize = 0;
1605 unsigned long leading, trailing;
1606 int pos, lines;
1607
1608 while (size > 0) {
1609 char first;
1610 int len = linelen(patch, size);
1611 int plen;
1612
1613 if (!len)
1614 break;
1615
1616 /*
1617 * "plen" is how much of the line we should use for
1618 * the actual patch data. Normally we just remove the
1619 * first character on the line, but if the line is
1620 * followed by "\ No newline", then we also remove the
1621 * last one (which is the newline, of course).
1622 */
1623 plen = len-1;
1624 if (len < size && patch[len] == '\\')
1625 plen--;
1626 first = *patch;
1627 if (apply_in_reverse) {
1628 if (first == '-')
1629 first = '+';
1630 else if (first == '+')
1631 first = '-';
1632 }
1633 switch (first) {
1634 case '\n':
1635 /* Newer GNU diff, empty context line */
1636 if (plen < 0)
1637 /* ... followed by '\No newline'; nothing */
1638 break;
1639 old[oldsize++] = '\n';
1640 new[newsize++] = '\n';
1641 break;
1642 case ' ':
1643 case '-':
1644 memcpy(old + oldsize, patch + 1, plen);
1645 oldsize += plen;
1646 if (first == '-')
1647 break;
1648 /* Fall-through for ' ' */
1649 case '+':
1650 if (first != '+' || !no_add)
1651 newsize += apply_line(new + newsize, patch,
1652 plen);
1653 break;
1654 case '@': case '\\':
1655 /* Ignore it, we already handled it */
1656 break;
1657 default:
1658 return -1;
1659 }
1660 patch += len;
1661 size -= len;
1662 }
1663
1664 if (inaccurate_eof && oldsize > 0 && old[oldsize - 1] == '\n' &&
1665 newsize > 0 && new[newsize - 1] == '\n') {
1666 oldsize--;
1667 newsize--;
1668 }
1669
1670 oldlines = old;
1671 newlines = new;
1672 leading = frag->leading;
1673 trailing = frag->trailing;
1674
1675 /*
1676 * If we don't have any leading/trailing data in the patch,
1677 * we want it to match at the beginning/end of the file.
1678 *
1679 * But that would break if the patch is generated with
1680 * --unified=0; sane people wouldn't do that to cause us
1681 * trouble, but we try to please not so sane ones as well.
1682 */
1683 if (unidiff_zero) {
1684 match_beginning = (!leading && !frag->oldpos);
1685 match_end = 0;
1686 }
1687 else {
1688 match_beginning = !leading && (frag->oldpos == 1);
1689 match_end = !trailing;
1690 }
1691
1692 lines = 0;
1693 pos = frag->newpos;
1694 for (;;) {
1695 offset = find_offset(buf, desc->size,
1696 oldlines, oldsize, pos, &lines);
1697 if (match_end && offset + oldsize != desc->size)
1698 offset = -1;
1699 if (match_beginning && offset)
1700 offset = -1;
1701 if (offset >= 0) {
1702 int diff = newsize - oldsize;
1703 unsigned long size = desc->size + diff;
1704 unsigned long alloc = desc->alloc;
1705
1706 /* Warn if it was necessary to reduce the number
1707 * of context lines.
1708 */
1709 if ((leading != frag->leading) ||
1710 (trailing != frag->trailing))
1711 fprintf(stderr, "Context reduced to (%ld/%ld)"
1712 " to apply fragment at %d\n",
1713 leading, trailing, pos + lines);
1714
1715 if (size > alloc) {
1716 alloc = size + 8192;
1717 desc->alloc = alloc;
1718 buf = xrealloc(buf, alloc);
1719 desc->buffer = buf;
1720 }
1721 desc->size = size;
1722 memmove(buf + offset + newsize,
1723 buf + offset + oldsize,
1724 size - offset - newsize);
1725 memcpy(buf + offset, newlines, newsize);
1726 offset = 0;
1727
1728 break;
1729 }
1730
1731 /* Am I at my context limits? */
1732 if ((leading <= p_context) && (trailing <= p_context))
1733 break;
1734 if (match_beginning || match_end) {
1735 match_beginning = match_end = 0;
1736 continue;
1737 }
1738 /* Reduce the number of context lines
1739 * Reduce both leading and trailing if they are equal
1740 * otherwise just reduce the larger context.
1741 */
1742 if (leading >= trailing) {
1743 remove_first_line(&oldlines, &oldsize);
1744 remove_first_line(&newlines, &newsize);
1745 pos--;
1746 leading--;
1747 }
1748 if (trailing > leading) {
1749 remove_last_line(&oldlines, &oldsize);
1750 remove_last_line(&newlines, &newsize);
1751 trailing--;
1752 }
1753 }
1754
1755 free(old);
1756 free(new);
1757 return offset;
1758 }
1759
1760 static int apply_binary_fragment(struct buffer_desc *desc, struct patch *patch)
1761 {
1762 unsigned long dst_size;
1763 struct fragment *fragment = patch->fragments;
1764 void *data;
1765 void *result;
1766
1767 /* Binary patch is irreversible without the optional second hunk */
1768 if (apply_in_reverse) {
1769 if (!fragment->next)
1770 return error("cannot reverse-apply a binary patch "
1771 "without the reverse hunk to '%s'",
1772 patch->new_name
1773 ? patch->new_name : patch->old_name);
1774 fragment = fragment->next;
1775 }
1776 data = (void*) fragment->patch;
1777 switch (fragment->binary_patch_method) {
1778 case BINARY_DELTA_DEFLATED:
1779 result = patch_delta(desc->buffer, desc->size,
1780 data,
1781 fragment->size,
1782 &dst_size);
1783 free(desc->buffer);
1784 desc->buffer = result;
1785 break;
1786 case BINARY_LITERAL_DEFLATED:
1787 free(desc->buffer);
1788 desc->buffer = data;
1789 dst_size = fragment->size;
1790 break;
1791 }
1792 if (!desc->buffer)
1793 return -1;
1794 desc->size = desc->alloc = dst_size;
1795 return 0;
1796 }
1797
1798 static int apply_binary(struct buffer_desc *desc, struct patch *patch)
1799 {
1800 const char *name = patch->old_name ? patch->old_name : patch->new_name;
1801 unsigned char sha1[20];
1802
1803 /* For safety, we require patch index line to contain
1804 * full 40-byte textual SHA1 for old and new, at least for now.
1805 */
1806 if (strlen(patch->old_sha1_prefix) != 40 ||
1807 strlen(patch->new_sha1_prefix) != 40 ||
1808 get_sha1_hex(patch->old_sha1_prefix, sha1) ||
1809 get_sha1_hex(patch->new_sha1_prefix, sha1))
1810 return error("cannot apply binary patch to '%s' "
1811 "without full index line", name);
1812
1813 if (patch->old_name) {
1814 /* See if the old one matches what the patch
1815 * applies to.
1816 */
1817 hash_sha1_file(desc->buffer, desc->size, blob_type, sha1);
1818 if (strcmp(sha1_to_hex(sha1), patch->old_sha1_prefix))
1819 return error("the patch applies to '%s' (%s), "
1820 "which does not match the "
1821 "current contents.",
1822 name, sha1_to_hex(sha1));
1823 }
1824 else {
1825 /* Otherwise, the old one must be empty. */
1826 if (desc->size)
1827 return error("the patch applies to an empty "
1828 "'%s' but it is not empty", name);
1829 }
1830
1831 get_sha1_hex(patch->new_sha1_prefix, sha1);
1832 if (is_null_sha1(sha1)) {
1833 free(desc->buffer);
1834 desc->alloc = desc->size = 0;
1835 desc->buffer = NULL;
1836 return 0; /* deletion patch */
1837 }
1838
1839 if (has_sha1_file(sha1)) {
1840 /* We already have the postimage */
1841 char type[10];
1842 unsigned long size;
1843
1844 free(desc->buffer);
1845 desc->buffer = read_sha1_file(sha1, type, &size);
1846 if (!desc->buffer)
1847 return error("the necessary postimage %s for "
1848 "'%s' cannot be read",
1849 patch->new_sha1_prefix, name);
1850 desc->alloc = desc->size = size;
1851 }
1852 else {
1853 /* We have verified desc matches the preimage;
1854 * apply the patch data to it, which is stored
1855 * in the patch->fragments->{patch,size}.
1856 */
1857 if (apply_binary_fragment(desc, patch))
1858 return error("binary patch does not apply to '%s'",
1859 name);
1860
1861 /* verify that the result matches */
1862 hash_sha1_file(desc->buffer, desc->size, blob_type, sha1);
1863 if (strcmp(sha1_to_hex(sha1), patch->new_sha1_prefix))
1864 return error("binary patch to '%s' creates incorrect result (expecting %s, got %s)", name, patch->new_sha1_prefix, sha1_to_hex(sha1));
1865 }
1866
1867 return 0;
1868 }
1869
1870 static int apply_fragments(struct buffer_desc *desc, struct patch *patch)
1871 {
1872 struct fragment *frag = patch->fragments;
1873 const char *name = patch->old_name ? patch->old_name : patch->new_name;
1874
1875 if (patch->is_binary)
1876 return apply_binary(desc, patch);
1877
1878 while (frag) {
1879 if (apply_one_fragment(desc, frag, patch->inaccurate_eof)) {
1880 error("patch failed: %s:%ld", name, frag->oldpos);
1881 if (!apply_with_reject)
1882 return -1;
1883 frag->rejected = 1;
1884 }
1885 frag = frag->next;
1886 }
1887 return 0;
1888 }
1889
1890 static int apply_data(struct patch *patch, struct stat *st, struct cache_entry *ce)
1891 {
1892 char *buf;
1893 unsigned long size, alloc;
1894 struct buffer_desc desc;
1895
1896 size = 0;
1897 alloc = 0;
1898 buf = NULL;
1899 if (cached) {
1900 if (ce) {
1901 char type[20];
1902 buf = read_sha1_file(ce->sha1, type, &size);
1903 if (!buf)
1904 return error("read of %s failed",
1905 patch->old_name);
1906 alloc = size;
1907 }
1908 }
1909 else if (patch->old_name) {
1910 size = st->st_size;
1911 alloc = size + 8192;
1912 buf = xmalloc(alloc);
1913 if (read_old_data(st, patch->old_name, buf, alloc) != size)
1914 return error("read of %s failed", patch->old_name);
1915 }
1916
1917 desc.size = size;
1918 desc.alloc = alloc;
1919 desc.buffer = buf;
1920
1921 if (apply_fragments(&desc, patch) < 0)
1922 return -1; /* note with --reject this succeeds. */
1923
1924 /* NUL terminate the result */
1925 if (desc.alloc <= desc.size)
1926 desc.buffer = xrealloc(desc.buffer, desc.size + 1);
1927 desc.buffer[desc.size] = 0;
1928
1929 patch->result = desc.buffer;
1930 patch->resultsize = desc.size;
1931
1932 if (0 < patch->is_delete && patch->resultsize)
1933 return error("removal patch leaves file contents");
1934
1935 return 0;
1936 }
1937
1938 static int check_patch(struct patch *patch, struct patch *prev_patch)
1939 {
1940 struct stat st;
1941 const char *old_name = patch->old_name;
1942 const char *new_name = patch->new_name;
1943 const char *name = old_name ? old_name : new_name;
1944 struct cache_entry *ce = NULL;
1945 int ok_if_exists;
1946
1947 patch->rejected = 1; /* we will drop this after we succeed */
1948 if (old_name) {
1949 int changed = 0;
1950 int stat_ret = 0;
1951 unsigned st_mode = 0;
1952
1953 if (!cached)
1954 stat_ret = lstat(old_name, &st);
1955 if (check_index) {
1956 int pos = cache_name_pos(old_name, strlen(old_name));
1957 if (pos < 0)
1958 return error("%s: does not exist in index",
1959 old_name);
1960 ce = active_cache[pos];
1961 if (stat_ret < 0) {
1962 struct checkout costate;
1963 if (errno != ENOENT)
1964 return error("%s: %s", old_name,
1965 strerror(errno));
1966 /* checkout */
1967 costate.base_dir = "";
1968 costate.base_dir_len = 0;
1969 costate.force = 0;
1970 costate.quiet = 0;
1971 costate.not_new = 0;
1972 costate.refresh_cache = 1;
1973 if (checkout_entry(ce,
1974 &costate,
1975 NULL) ||
1976 lstat(old_name, &st))
1977 return -1;
1978 }
1979 if (!cached)
1980 changed = ce_match_stat(ce, &st, 1);
1981 if (changed)
1982 return error("%s: does not match index",
1983 old_name);
1984 if (cached)
1985 st_mode = ntohl(ce->ce_mode);
1986 }
1987 else if (stat_ret < 0)
1988 return error("%s: %s", old_name, strerror(errno));
1989
1990 if (!cached)
1991 st_mode = ntohl(create_ce_mode(st.st_mode));
1992
1993 if (patch->is_new < 0)
1994 patch->is_new = 0;
1995 if (!patch->old_mode)
1996 patch->old_mode = st_mode;
1997 if ((st_mode ^ patch->old_mode) & S_IFMT)
1998 return error("%s: wrong type", old_name);
1999 if (st_mode != patch->old_mode)
2000 fprintf(stderr, "warning: %s has type %o, expected %o\n",
2001 old_name, st_mode, patch->old_mode);
2002 }
2003
2004 if (new_name && prev_patch && 0 < prev_patch->is_delete &&
2005 !strcmp(prev_patch->old_name, new_name))
2006 /* A type-change diff is always split into a patch to
2007 * delete old, immediately followed by a patch to
2008 * create new (see diff.c::run_diff()); in such a case
2009 * it is Ok that the entry to be deleted by the
2010 * previous patch is still in the working tree and in
2011 * the index.
2012 */
2013 ok_if_exists = 1;
2014 else
2015 ok_if_exists = 0;
2016
2017 if (new_name &&
2018 ((0 < patch->is_new) | (0 < patch->is_rename) | patch->is_copy)) {
2019 if (check_index &&
2020 cache_name_pos(new_name, strlen(new_name)) >= 0 &&
2021 !ok_if_exists)
2022 return error("%s: already exists in index", new_name);
2023 if (!cached) {
2024 struct stat nst;
2025 if (!lstat(new_name, &nst)) {
2026 if (S_ISDIR(nst.st_mode) || ok_if_exists)
2027 ; /* ok */
2028 else
2029 return error("%s: already exists in working directory", new_name);
2030 }
2031 else if ((errno != ENOENT) && (errno != ENOTDIR))
2032 return error("%s: %s", new_name, strerror(errno));
2033 }
2034 if (!patch->new_mode) {
2035 if (0 < patch->is_new)
2036 patch->new_mode = S_IFREG | 0644;
2037 else
2038 patch->new_mode = patch->old_mode;
2039 }
2040 }
2041
2042 if (new_name && old_name) {
2043 int same = !strcmp(old_name, new_name);
2044 if (!patch->new_mode)
2045 patch->new_mode = patch->old_mode;
2046 if ((patch->old_mode ^ patch->new_mode) & S_IFMT)
2047 return error("new mode (%o) of %s does not match old mode (%o)%s%s",
2048 patch->new_mode, new_name, patch->old_mode,
2049 same ? "" : " of ", same ? "" : old_name);
2050 }
2051
2052 if (apply_data(patch, &st, ce) < 0)
2053 return error("%s: patch does not apply", name);
2054 patch->rejected = 0;
2055 return 0;
2056 }
2057
2058 static int check_patch_list(struct patch *patch)
2059 {
2060 struct patch *prev_patch = NULL;
2061 int err = 0;
2062
2063 for (prev_patch = NULL; patch ; patch = patch->next) {
2064 if (apply_verbosely)
2065 say_patch_name(stderr,
2066 "Checking patch ", patch, "...\n");
2067 err |= check_patch(patch, prev_patch);
2068 prev_patch = patch;
2069 }
2070 return err;
2071 }
2072
2073 static void show_index_list(struct patch *list)
2074 {
2075 struct patch *patch;
2076
2077 /* Once we start supporting the reverse patch, it may be
2078 * worth showing the new sha1 prefix, but until then...
2079 */
2080 for (patch = list; patch; patch = patch->next) {
2081 const unsigned char *sha1_ptr;
2082 unsigned char sha1[20];
2083 const char *name;
2084
2085 name = patch->old_name ? patch->old_name : patch->new_name;
2086 if (0 < patch->is_new)
2087 sha1_ptr = null_sha1;
2088 else if (get_sha1(patch->old_sha1_prefix, sha1))
2089 die("sha1 information is lacking or useless (%s).",
2090 name);
2091 else
2092 sha1_ptr = sha1;
2093
2094 printf("%06o %s ",patch->old_mode, sha1_to_hex(sha1_ptr));
2095 if (line_termination && quote_c_style(name, NULL, NULL, 0))
2096 quote_c_style(name, NULL, stdout, 0);
2097 else
2098 fputs(name, stdout);
2099 putchar(line_termination);
2100 }
2101 }
2102
2103 static void stat_patch_list(struct patch *patch)
2104 {
2105 int files, adds, dels;
2106
2107 for (files = adds = dels = 0 ; patch ; patch = patch->next) {
2108 files++;
2109 adds += patch->lines_added;
2110 dels += patch->lines_deleted;
2111 show_stats(patch);
2112 }
2113
2114 printf(" %d files changed, %d insertions(+), %d deletions(-)\n", files, adds, dels);
2115 }
2116
2117 static void numstat_patch_list(struct patch *patch)
2118 {
2119 for ( ; patch; patch = patch->next) {
2120 const char *name;
2121 name = patch->new_name ? patch->new_name : patch->old_name;
2122 printf("%d\t%d\t", patch->lines_added, patch->lines_deleted);
2123 if (line_termination && quote_c_style(name, NULL, NULL, 0))
2124 quote_c_style(name, NULL, stdout, 0);
2125 else
2126 fputs(name, stdout);
2127 putchar(line_termination);
2128 }
2129 }
2130
2131 static void show_file_mode_name(const char *newdelete, unsigned int mode, const char *name)
2132 {
2133 if (mode)
2134 printf(" %s mode %06o %s\n", newdelete, mode, name);
2135 else
2136 printf(" %s %s\n", newdelete, name);
2137 }
2138
2139 static void show_mode_change(struct patch *p, int show_name)
2140 {
2141 if (p->old_mode && p->new_mode && p->old_mode != p->new_mode) {
2142 if (show_name)
2143 printf(" mode change %06o => %06o %s\n",
2144 p->old_mode, p->new_mode, p->new_name);
2145 else
2146 printf(" mode change %06o => %06o\n",
2147 p->old_mode, p->new_mode);
2148 }
2149 }
2150
2151 static void show_rename_copy(struct patch *p)
2152 {
2153 const char *renamecopy = p->is_rename ? "rename" : "copy";
2154 const char *old, *new;
2155
2156 /* Find common prefix */
2157 old = p->old_name;
2158 new = p->new_name;
2159 while (1) {
2160 const char *slash_old, *slash_new;
2161 slash_old = strchr(old, '/');
2162 slash_new = strchr(new, '/');
2163 if (!slash_old ||
2164 !slash_new ||
2165 slash_old - old != slash_new - new ||
2166 memcmp(old, new, slash_new - new))
2167 break;
2168 old = slash_old + 1;
2169 new = slash_new + 1;
2170 }
2171 /* p->old_name thru old is the common prefix, and old and new
2172 * through the end of names are renames
2173 */
2174 if (old != p->old_name)
2175 printf(" %s %.*s{%s => %s} (%d%%)\n", renamecopy,
2176 (int)(old - p->old_name), p->old_name,
2177 old, new, p->score);
2178 else
2179 printf(" %s %s => %s (%d%%)\n", renamecopy,
2180 p->old_name, p->new_name, p->score);
2181 show_mode_change(p, 0);
2182 }
2183
2184 static void summary_patch_list(struct patch *patch)
2185 {
2186 struct patch *p;
2187
2188 for (p = patch; p; p = p->next) {
2189 if (p->is_new)
2190 show_file_mode_name("create", p->new_mode, p->new_name);
2191 else if (p->is_delete)
2192 show_file_mode_name("delete", p->old_mode, p->old_name);
2193 else {
2194 if (p->is_rename || p->is_copy)
2195 show_rename_copy(p);
2196 else {
2197 if (p->score) {
2198 printf(" rewrite %s (%d%%)\n",
2199 p->new_name, p->score);
2200 show_mode_change(p, 0);
2201 }
2202 else
2203 show_mode_change(p, 1);
2204 }
2205 }
2206 }
2207 }
2208
2209 static void patch_stats(struct patch *patch)
2210 {
2211 int lines = patch->lines_added + patch->lines_deleted;
2212
2213 if (lines > max_change)
2214 max_change = lines;
2215 if (patch->old_name) {
2216 int len = quote_c_style(patch->old_name, NULL, NULL, 0);
2217 if (!len)
2218 len = strlen(patch->old_name);
2219 if (len > max_len)
2220 max_len = len;
2221 }
2222 if (patch->new_name) {
2223 int len = quote_c_style(patch->new_name, NULL, NULL, 0);
2224 if (!len)
2225 len = strlen(patch->new_name);
2226 if (len > max_len)
2227 max_len = len;
2228 }
2229 }
2230
2231 static void remove_file(struct patch *patch)
2232 {
2233 if (write_index) {
2234 if (remove_file_from_cache(patch->old_name) < 0)
2235 die("unable to remove %s from index", patch->old_name);
2236 cache_tree_invalidate_path(active_cache_tree, patch->old_name);
2237 }
2238 if (!cached)
2239 unlink(patch->old_name);
2240 }
2241
2242 static void add_index_file(const char *path, unsigned mode, void *buf, unsigned long size)
2243 {
2244 struct stat st;
2245 struct cache_entry *ce;
2246 int namelen = strlen(path);
2247 unsigned ce_size = cache_entry_size(namelen);
2248
2249 if (!write_index)
2250 return;
2251
2252 ce = xcalloc(1, ce_size);
2253 memcpy(ce->name, path, namelen);
2254 ce->ce_mode = create_ce_mode(mode);
2255 ce->ce_flags = htons(namelen);
2256 if (!cached) {
2257 if (lstat(path, &st) < 0)
2258 die("unable to stat newly created file %s", path);
2259 fill_stat_cache_info(ce, &st);
2260 }
2261 if (write_sha1_file(buf, size, blob_type, ce->sha1) < 0)
2262 die("unable to create backing store for newly created file %s", path);
2263 if (add_cache_entry(ce, ADD_CACHE_OK_TO_ADD) < 0)
2264 die("unable to add cache entry for %s", path);
2265 }
2266
2267 static int try_create_file(const char *path, unsigned int mode, const char *buf, unsigned long size)
2268 {
2269 int fd;
2270
2271 if (S_ISLNK(mode))
2272 /* Although buf:size is counted string, it also is NUL
2273 * terminated.
2274 */
2275 return symlink(buf, path);
2276 fd = open(path, O_CREAT | O_EXCL | O_WRONLY, (mode & 0100) ? 0777 : 0666);
2277 if (fd < 0)
2278 return -1;
2279 while (size) {
2280 int written = xwrite(fd, buf, size);
2281 if (written < 0)
2282 die("writing file %s: %s", path, strerror(errno));
2283 if (!written)
2284 die("out of space writing file %s", path);
2285 buf += written;
2286 size -= written;
2287 }
2288 if (close(fd) < 0)
2289 die("closing file %s: %s", path, strerror(errno));
2290 return 0;
2291 }
2292
2293 /*
2294 * We optimistically assume that the directories exist,
2295 * which is true 99% of the time anyway. If they don't,
2296 * we create them and try again.
2297 */
2298 static void create_one_file(char *path, unsigned mode, const char *buf, unsigned long size)
2299 {
2300 if (cached)
2301 return;
2302 if (!try_create_file(path, mode, buf, size))
2303 return;
2304
2305 if (errno == ENOENT) {
2306 if (safe_create_leading_directories(path))
2307 return;
2308 if (!try_create_file(path, mode, buf, size))
2309 return;
2310 }
2311
2312 if (errno == EEXIST || errno == EACCES) {
2313 /* We may be trying to create a file where a directory
2314 * used to be.
2315 */
2316 struct stat st;
2317 errno = 0;
2318 if (!lstat(path, &st) && S_ISDIR(st.st_mode) && !rmdir(path))
2319 errno = EEXIST;
2320 }
2321
2322 if (errno == EEXIST) {
2323 unsigned int nr = getpid();
2324
2325 for (;;) {
2326 const char *newpath;
2327 newpath = mkpath("%s~%u", path, nr);
2328 if (!try_create_file(newpath, mode, buf, size)) {
2329 if (!rename(newpath, path))
2330 return;
2331 unlink(newpath);
2332 break;
2333 }
2334 if (errno != EEXIST)
2335 break;
2336 ++nr;
2337 }
2338 }
2339 die("unable to write file %s mode %o", path, mode);
2340 }
2341
2342 static void create_file(struct patch *patch)
2343 {
2344 char *path = patch->new_name;
2345 unsigned mode = patch->new_mode;
2346 unsigned long size = patch->resultsize;
2347 char *buf = patch->result;
2348
2349 if (!mode)
2350 mode = S_IFREG | 0644;
2351 create_one_file(path, mode, buf, size);
2352 add_index_file(path, mode, buf, size);
2353 cache_tree_invalidate_path(active_cache_tree, path);
2354 }
2355
2356 /* phase zero is to remove, phase one is to create */
2357 static void write_out_one_result(struct patch *patch, int phase)
2358 {
2359 if (patch->is_delete > 0) {
2360 if (phase == 0)
2361 remove_file(patch);
2362 return;
2363 }
2364 if (patch->is_new > 0 || patch->is_copy) {
2365 if (phase == 1)
2366 create_file(patch);
2367 return;
2368 }
2369 /*
2370 * Rename or modification boils down to the same
2371 * thing: remove the old, write the new
2372 */
2373 if (phase == 0)
2374 remove_file(patch);
2375 if (phase == 1)
2376 create_file(patch);
2377 }
2378
2379 static int write_out_one_reject(struct patch *patch)
2380 {
2381 FILE *rej;
2382 char namebuf[PATH_MAX];
2383 struct fragment *frag;
2384 int cnt = 0;
2385
2386 for (cnt = 0, frag = patch->fragments; frag; frag = frag->next) {
2387 if (!frag->rejected)
2388 continue;
2389 cnt++;
2390 }
2391
2392 if (!cnt) {
2393 if (apply_verbosely)
2394 say_patch_name(stderr,
2395 "Applied patch ", patch, " cleanly.\n");
2396 return 0;
2397 }
2398
2399 /* This should not happen, because a removal patch that leaves
2400 * contents are marked "rejected" at the patch level.
2401 */
2402 if (!patch->new_name)
2403 die("internal error");
2404
2405 /* Say this even without --verbose */
2406 say_patch_name(stderr, "Applying patch ", patch, " with");
2407 fprintf(stderr, " %d rejects...\n", cnt);
2408
2409 cnt = strlen(patch->new_name);
2410 if (ARRAY_SIZE(namebuf) <= cnt + 5) {
2411 cnt = ARRAY_SIZE(namebuf) - 5;
2412 fprintf(stderr,
2413 "warning: truncating .rej filename to %.*s.rej",
2414 cnt - 1, patch->new_name);
2415 }
2416 memcpy(namebuf, patch->new_name, cnt);
2417 memcpy(namebuf + cnt, ".rej", 5);
2418
2419 rej = fopen(namebuf, "w");
2420 if (!rej)
2421 return error("cannot open %s: %s", namebuf, strerror(errno));
2422
2423 /* Normal git tools never deal with .rej, so do not pretend
2424 * this is a git patch by saying --git nor give extended
2425 * headers. While at it, maybe please "kompare" that wants
2426 * the trailing TAB and some garbage at the end of line ;-).
2427 */
2428 fprintf(rej, "diff a/%s b/%s\t(rejected hunks)\n",
2429 patch->new_name, patch->new_name);
2430 for (cnt = 1, frag = patch->fragments;
2431 frag;
2432 cnt++, frag = frag->next) {
2433 if (!frag->rejected) {
2434 fprintf(stderr, "Hunk #%d applied cleanly.\n", cnt);
2435 continue;
2436 }
2437 fprintf(stderr, "Rejected hunk #%d.\n", cnt);
2438 fprintf(rej, "%.*s", frag->size, frag->patch);
2439 if (frag->patch[frag->size-1] != '\n')
2440 fputc('\n', rej);
2441 }
2442 fclose(rej);
2443 return -1;
2444 }
2445
2446 static int write_out_results(struct patch *list, int skipped_patch)
2447 {
2448 int phase;
2449 int errs = 0;
2450 struct patch *l;
2451
2452 if (!list && !skipped_patch)
2453 return error("No changes");
2454
2455 for (phase = 0; phase < 2; phase++) {
2456 l = list;
2457 while (l) {
2458 if (l->rejected)
2459 errs = 1;
2460 else {
2461 write_out_one_result(l, phase);
2462 if (phase == 1 && write_out_one_reject(l))
2463 errs = 1;
2464 }
2465 l = l->next;
2466 }
2467 }
2468 return errs;
2469 }
2470
2471 static struct lock_file lock_file;
2472
2473 static struct excludes {
2474 struct excludes *next;
2475 const char *path;
2476 } *excludes;
2477
2478 static int use_patch(struct patch *p)
2479 {
2480 const char *pathname = p->new_name ? p->new_name : p->old_name;
2481 struct excludes *x = excludes;
2482 while (x) {
2483 if (fnmatch(x->path, pathname, 0) == 0)
2484 return 0;
2485 x = x->next;
2486 }
2487 if (0 < prefix_length) {
2488 int pathlen = strlen(pathname);
2489 if (pathlen <= prefix_length ||
2490 memcmp(prefix, pathname, prefix_length))
2491 return 0;
2492 }
2493 return 1;
2494 }
2495
2496 static int apply_patch(int fd, const char *filename, int inaccurate_eof)
2497 {
2498 unsigned long offset, size;
2499 char *buffer = read_patch_file(fd, &size);
2500 struct patch *list = NULL, **listp = &list;
2501 int skipped_patch = 0;
2502
2503 patch_input_file = filename;
2504 if (!buffer)
2505 return -1;
2506 offset = 0;
2507 while (size > 0) {
2508 struct patch *patch;
2509 int nr;
2510
2511 patch = xcalloc(1, sizeof(*patch));
2512 patch->inaccurate_eof = inaccurate_eof;
2513 nr = parse_chunk(buffer + offset, size, patch);
2514 if (nr < 0)
2515 break;
2516 if (apply_in_reverse)
2517 reverse_patches(patch);
2518 if (use_patch(patch)) {
2519 patch_stats(patch);
2520 *listp = patch;
2521 listp = &patch->next;
2522 } else {
2523 /* perhaps free it a bit better? */
2524 free(patch);
2525 skipped_patch++;
2526 }
2527 offset += nr;
2528 size -= nr;
2529 }
2530
2531 if (whitespace_error && (new_whitespace == error_on_whitespace))
2532 apply = 0;
2533
2534 write_index = check_index && apply;
2535 if (write_index && newfd < 0)
2536 newfd = hold_lock_file_for_update(&lock_file,
2537 get_index_file(), 1);
2538 if (check_index) {
2539 if (read_cache() < 0)
2540 die("unable to read index file");
2541 }
2542
2543 if ((check || apply) &&
2544 check_patch_list(list) < 0 &&
2545 !apply_with_reject)
2546 exit(1);
2547
2548 if (apply && write_out_results(list, skipped_patch))
2549 exit(1);
2550
2551 if (show_index_info)
2552 show_index_list(list);
2553
2554 if (diffstat)
2555 stat_patch_list(list);
2556
2557 if (numstat)
2558 numstat_patch_list(list);
2559
2560 if (summary)
2561 summary_patch_list(list);
2562
2563 free(buffer);
2564 return 0;
2565 }
2566
2567 static int git_apply_config(const char *var, const char *value)
2568 {
2569 if (!strcmp(var, "apply.whitespace")) {
2570 apply_default_whitespace = xstrdup(value);
2571 return 0;
2572 }
2573 return git_default_config(var, value);
2574 }
2575
2576
2577 int cmd_apply(int argc, const char **argv, const char *prefix)
2578 {
2579 int i;
2580 int read_stdin = 1;
2581 int inaccurate_eof = 0;
2582 int errs = 0;
2583
2584 const char *whitespace_option = NULL;
2585
2586 for (i = 1; i < argc; i++) {
2587 const char *arg = argv[i];
2588 char *end;
2589 int fd;
2590
2591 if (!strcmp(arg, "-")) {
2592 errs |= apply_patch(0, "<stdin>", inaccurate_eof);
2593 read_stdin = 0;
2594 continue;
2595 }
2596 if (!strncmp(arg, "--exclude=", 10)) {
2597 struct excludes *x = xmalloc(sizeof(*x));
2598 x->path = arg + 10;
2599 x->next = excludes;
2600 excludes = x;
2601 continue;
2602 }
2603 if (!strncmp(arg, "-p", 2)) {
2604 p_value = atoi(arg + 2);
2605 continue;
2606 }
2607 if (!strcmp(arg, "--no-add")) {
2608 no_add = 1;
2609 continue;
2610 }
2611 if (!strcmp(arg, "--stat")) {
2612 apply = 0;
2613 diffstat = 1;
2614 continue;
2615 }
2616 if (!strcmp(arg, "--allow-binary-replacement") ||
2617 !strcmp(arg, "--binary")) {
2618 continue; /* now no-op */
2619 }
2620 if (!strcmp(arg, "--numstat")) {
2621 apply = 0;
2622 numstat = 1;
2623 continue;
2624 }
2625 if (!strcmp(arg, "--summary")) {
2626 apply = 0;
2627 summary = 1;
2628 continue;
2629 }
2630 if (!strcmp(arg, "--check")) {
2631 apply = 0;
2632 check = 1;
2633 continue;
2634 }
2635 if (!strcmp(arg, "--index")) {
2636 check_index = 1;
2637 continue;
2638 }
2639 if (!strcmp(arg, "--cached")) {
2640 check_index = 1;
2641 cached = 1;
2642 continue;
2643 }
2644 if (!strcmp(arg, "--apply")) {
2645 apply = 1;
2646 continue;
2647 }
2648 if (!strcmp(arg, "--index-info")) {
2649 apply = 0;
2650 show_index_info = 1;
2651 continue;
2652 }
2653 if (!strcmp(arg, "-z")) {
2654 line_termination = 0;
2655 continue;
2656 }
2657 if (!strncmp(arg, "-C", 2)) {
2658 p_context = strtoul(arg + 2, &end, 0);
2659 if (*end != '\0')
2660 die("unrecognized context count '%s'", arg + 2);
2661 continue;
2662 }
2663 if (!strncmp(arg, "--whitespace=", 13)) {
2664 whitespace_option = arg + 13;
2665 parse_whitespace_option(arg + 13);
2666 continue;
2667 }
2668 if (!strcmp(arg, "-R") || !strcmp(arg, "--reverse")) {
2669 apply_in_reverse = 1;
2670 continue;
2671 }
2672 if (!strcmp(arg, "--unidiff-zero")) {
2673 unidiff_zero = 1;
2674 continue;
2675 }
2676 if (!strcmp(arg, "--reject")) {
2677 apply = apply_with_reject = apply_verbosely = 1;
2678 continue;
2679 }
2680 if (!strcmp(arg, "--verbose")) {
2681 apply_verbosely = 1;
2682 continue;
2683 }
2684 if (!strcmp(arg, "--inaccurate-eof")) {
2685 inaccurate_eof = 1;
2686 continue;
2687 }
2688
2689 if (check_index && prefix_length < 0) {
2690 prefix = setup_git_directory();
2691 prefix_length = prefix ? strlen(prefix) : 0;
2692 git_config(git_apply_config);
2693 if (!whitespace_option && apply_default_whitespace)
2694 parse_whitespace_option(apply_default_whitespace);
2695 }
2696 if (0 < prefix_length)
2697 arg = prefix_filename(prefix, prefix_length, arg);
2698
2699 fd = open(arg, O_RDONLY);
2700 if (fd < 0)
2701 usage(apply_usage);
2702 read_stdin = 0;
2703 set_default_whitespace_mode(whitespace_option);
2704 errs |= apply_patch(fd, arg, inaccurate_eof);
2705 close(fd);
2706 }
2707 set_default_whitespace_mode(whitespace_option);
2708 if (read_stdin)
2709 errs |= apply_patch(0, "<stdin>", inaccurate_eof);
2710 if (whitespace_error) {
2711 if (squelch_whitespace_errors &&
2712 squelch_whitespace_errors < whitespace_error) {
2713 int squelched =
2714 whitespace_error - squelch_whitespace_errors;
2715 fprintf(stderr, "warning: squelched %d "
2716 "whitespace error%s\n",
2717 squelched,
2718 squelched == 1 ? "" : "s");
2719 }
2720 if (new_whitespace == error_on_whitespace)
2721 die("%d line%s add%s trailing whitespaces.",
2722 whitespace_error,
2723 whitespace_error == 1 ? "" : "s",
2724 whitespace_error == 1 ? "s" : "");
2725 if (applied_after_stripping)
2726 fprintf(stderr, "warning: %d line%s applied after"
2727 " stripping trailing whitespaces.\n",
2728 applied_after_stripping,
2729 applied_after_stripping == 1 ? "" : "s");
2730 else if (whitespace_error)
2731 fprintf(stderr, "warning: %d line%s add%s trailing"
2732 " whitespaces.\n",
2733 whitespace_error,
2734 whitespace_error == 1 ? "" : "s",
2735 whitespace_error == 1 ? "s" : "");
2736 }
2737
2738 if (write_index) {
2739 if (write_cache(newfd, active_cache, active_nr) ||
2740 close(newfd) || commit_lock_file(&lock_file))
2741 die("Unable to write new index file");
2742 }
2743
2744 return !!errs;
2745 }