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