ea6fb4c8806203a0268261a4aaa4a07dcdefdca2
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"
16 // --check turns on checking that the working tree matches the
17 // files that are being modified, but doesn't apply the patch
18 // --stat does just a diffstat, and doesn't actually apply
19 // --numstat does numeric diffstat, and doesn't actually apply
20 // --index-info shows the old and new index info for paths if available.
21 //
22 static const char *prefix;
23 static int prefix_length = -1;
24 static int newfd = -1;
26 static int p_value = 1;
27 static int allow_binary_replacement = 0;
28 static int check_index = 0;
29 static int write_index = 0;
30 static int diffstat = 0;
31 static int numstat = 0;
32 static int summary = 0;
33 static int check = 0;
34 static int apply = 1;
35 static int no_add = 0;
36 static int show_index_info = 0;
37 static int line_termination = '\n';
38 static unsigned long p_context = -1;
39 static const char apply_usage[] =
40 "git-apply [--stat] [--numstat] [--summary] [--check] [--index] [--apply] [--no-add] [--index-info] [--allow-binary-replacement] [-z] [-pNUM] [-CNUM] [--whitespace=<nowarn|warn|error|error-all|strip>] <patch>...";
42 static enum whitespace_eol {
43 nowarn_whitespace,
44 warn_on_whitespace,
45 error_on_whitespace,
46 strip_whitespace,
47 } new_whitespace = warn_on_whitespace;
48 static int whitespace_error = 0;
49 static int squelch_whitespace_errors = 5;
50 static int applied_after_stripping = 0;
51 static const char *patch_input_file = NULL;
53 static void parse_whitespace_option(const char *option)
54 {
55 if (!option) {
56 new_whitespace = warn_on_whitespace;
57 return;
58 }
59 if (!strcmp(option, "warn")) {
60 new_whitespace = warn_on_whitespace;
61 return;
62 }
63 if (!strcmp(option, "nowarn")) {
64 new_whitespace = nowarn_whitespace;
65 return;
66 }
67 if (!strcmp(option, "error")) {
68 new_whitespace = error_on_whitespace;
69 return;
70 }
71 if (!strcmp(option, "error-all")) {
72 new_whitespace = error_on_whitespace;
73 squelch_whitespace_errors = 0;
74 return;
75 }
76 if (!strcmp(option, "strip")) {
77 new_whitespace = strip_whitespace;
78 return;
79 }
80 die("unrecognized whitespace option '%s'", option);
81 }
83 static void set_default_whitespace_mode(const char *whitespace_option)
84 {
85 if (!whitespace_option && !apply_default_whitespace) {
86 new_whitespace = (apply
87 ? warn_on_whitespace
88 : nowarn_whitespace);
89 }
90 }
92 /*
93 * For "diff-stat" like behaviour, we keep track of the biggest change
94 * we've seen, and the longest filename. That allows us to do simple
95 * scaling.
96 */
97 static int max_change, max_len;
99 /*
100 * Various "current state", notably line numbers and what
101 * file (and how) we're patching right now.. The "is_xxxx"
102 * things are flags, where -1 means "don't know yet".
103 */
104 static int linenr = 1;
106 struct fragment {
107 unsigned long leading, trailing;
108 unsigned long oldpos, oldlines;
109 unsigned long newpos, newlines;
110 const char *patch;
111 int size;
112 struct fragment *next;
113 };
115 struct patch {
116 char *new_name, *old_name, *def_name;
117 unsigned int old_mode, new_mode;
118 int is_rename, is_copy, is_new, is_delete, is_binary;
119 #define BINARY_DELTA_DEFLATED 1
120 #define BINARY_LITERAL_DEFLATED 2
121 unsigned long deflate_origlen;
122 int lines_added, lines_deleted;
123 int score;
124 struct fragment *fragments;
125 char *result;
126 unsigned long resultsize;
127 char old_sha1_prefix[41];
128 char new_sha1_prefix[41];
129 struct patch *next;
130 };
132 #define CHUNKSIZE (8192)
133 #define SLOP (16)
135 static void *read_patch_file(int fd, unsigned long *sizep)
136 {
137 unsigned long size = 0, alloc = CHUNKSIZE;
138 void *buffer = xmalloc(alloc);
140 for (;;) {
141 int nr = alloc - size;
142 if (nr < 1024) {
143 alloc += CHUNKSIZE;
144 buffer = xrealloc(buffer, alloc);
145 nr = alloc - size;
146 }
147 nr = xread(fd, buffer + size, nr);
148 if (!nr)
149 break;
150 if (nr < 0)
151 die("git-apply: read returned %s", strerror(errno));
152 size += nr;
153 }
154 *sizep = size;
156 /*
157 * Make sure that we have some slop in the buffer
158 * so that we can do speculative "memcmp" etc, and
159 * see to it that it is NUL-filled.
160 */
161 if (alloc < size + SLOP)
162 buffer = xrealloc(buffer, size + SLOP);
163 memset(buffer + size, 0, SLOP);
164 return buffer;
165 }
167 static unsigned long linelen(const char *buffer, unsigned long size)
168 {
169 unsigned long len = 0;
170 while (size--) {
171 len++;
172 if (*buffer++ == '\n')
173 break;
174 }
175 return len;
176 }
178 static int is_dev_null(const char *str)
179 {
180 return !memcmp("/dev/null", str, 9) && isspace(str[9]);
181 }
183 #define TERM_SPACE 1
184 #define TERM_TAB 2
186 static int name_terminate(const char *name, int namelen, int c, int terminate)
187 {
188 if (c == ' ' && !(terminate & TERM_SPACE))
189 return 0;
190 if (c == '\t' && !(terminate & TERM_TAB))
191 return 0;
193 return 1;
194 }
196 static char * find_name(const char *line, char *def, int p_value, int terminate)
197 {
198 int len;
199 const char *start = line;
200 char *name;
202 if (*line == '"') {
203 /* Proposed "new-style" GNU patch/diff format; see
204 * http://marc.theaimsgroup.com/?l=git&m=112927316408690&w=2
205 */
206 name = unquote_c_style(line, NULL);
207 if (name) {
208 char *cp = name;
209 while (p_value) {
210 cp = strchr(name, '/');
211 if (!cp)
212 break;
213 cp++;
214 p_value--;
215 }
216 if (cp) {
217 /* name can later be freed, so we need
218 * to memmove, not just return cp
219 */
220 memmove(name, cp, strlen(cp) + 1);
221 free(def);
222 return name;
223 }
224 else {
225 free(name);
226 name = NULL;
227 }
228 }
229 }
231 for (;;) {
232 char c = *line;
234 if (isspace(c)) {
235 if (c == '\n')
236 break;
237 if (name_terminate(start, line-start, c, terminate))
238 break;
239 }
240 line++;
241 if (c == '/' && !--p_value)
242 start = line;
243 }
244 if (!start)
245 return def;
246 len = line - start;
247 if (!len)
248 return def;
250 /*
251 * Generally we prefer the shorter name, especially
252 * if the other one is just a variation of that with
253 * something else tacked on to the end (ie "file.orig"
254 * or "file~").
255 */
256 if (def) {
257 int deflen = strlen(def);
258 if (deflen < len && !strncmp(start, def, deflen))
259 return def;
260 }
262 name = xmalloc(len + 1);
263 memcpy(name, start, len);
264 name[len] = 0;
265 free(def);
266 return name;
267 }
269 /*
270 * Get the name etc info from the --/+++ lines of a traditional patch header
271 *
272 * NOTE! This hardcodes "-p1" behaviour in filename detection.
273 *
274 * FIXME! The end-of-filename heuristics are kind of screwy. For existing
275 * files, we can happily check the index for a match, but for creating a
276 * new file we should try to match whatever "patch" does. I have no idea.
277 */
278 static void parse_traditional_patch(const char *first, const char *second, struct patch *patch)
279 {
280 char *name;
282 first += 4; // skip "--- "
283 second += 4; // skip "+++ "
284 if (is_dev_null(first)) {
285 patch->is_new = 1;
286 patch->is_delete = 0;
287 name = find_name(second, NULL, p_value, TERM_SPACE | TERM_TAB);
288 patch->new_name = name;
289 } else if (is_dev_null(second)) {
290 patch->is_new = 0;
291 patch->is_delete = 1;
292 name = find_name(first, NULL, p_value, TERM_SPACE | TERM_TAB);
293 patch->old_name = name;
294 } else {
295 name = find_name(first, NULL, p_value, TERM_SPACE | TERM_TAB);
296 name = find_name(second, name, p_value, TERM_SPACE | TERM_TAB);
297 patch->old_name = patch->new_name = name;
298 }
299 if (!name)
300 die("unable to find filename in patch at line %d", linenr);
301 }
303 static int gitdiff_hdrend(const char *line, struct patch *patch)
304 {
305 return -1;
306 }
308 /*
309 * We're anal about diff header consistency, to make
310 * sure that we don't end up having strange ambiguous
311 * patches floating around.
312 *
313 * As a result, gitdiff_{old|new}name() will check
314 * their names against any previous information, just
315 * to make sure..
316 */
317 static char *gitdiff_verify_name(const char *line, int isnull, char *orig_name, const char *oldnew)
318 {
319 if (!orig_name && !isnull)
320 return find_name(line, NULL, 1, 0);
322 if (orig_name) {
323 int len;
324 const char *name;
325 char *another;
326 name = orig_name;
327 len = strlen(name);
328 if (isnull)
329 die("git-apply: bad git-diff - expected /dev/null, got %s on line %d", name, linenr);
330 another = find_name(line, NULL, 1, 0);
331 if (!another || memcmp(another, name, len))
332 die("git-apply: bad git-diff - inconsistent %s filename on line %d", oldnew, linenr);
333 free(another);
334 return orig_name;
335 }
336 else {
337 /* expect "/dev/null" */
338 if (memcmp("/dev/null", line, 9) || line[9] != '\n')
339 die("git-apply: bad git-diff - expected /dev/null on line %d", linenr);
340 return NULL;
341 }
342 }
344 static int gitdiff_oldname(const char *line, struct patch *patch)
345 {
346 patch->old_name = gitdiff_verify_name(line, patch->is_new, patch->old_name, "old");
347 return 0;
348 }
350 static int gitdiff_newname(const char *line, struct patch *patch)
351 {
352 patch->new_name = gitdiff_verify_name(line, patch->is_delete, patch->new_name, "new");
353 return 0;
354 }
356 static int gitdiff_oldmode(const char *line, struct patch *patch)
357 {
358 patch->old_mode = strtoul(line, NULL, 8);
359 return 0;
360 }
362 static int gitdiff_newmode(const char *line, struct patch *patch)
363 {
364 patch->new_mode = strtoul(line, NULL, 8);
365 return 0;
366 }
368 static int gitdiff_delete(const char *line, struct patch *patch)
369 {
370 patch->is_delete = 1;
371 patch->old_name = patch->def_name;
372 return gitdiff_oldmode(line, patch);
373 }
375 static int gitdiff_newfile(const char *line, struct patch *patch)
376 {
377 patch->is_new = 1;
378 patch->new_name = patch->def_name;
379 return gitdiff_newmode(line, patch);
380 }
382 static int gitdiff_copysrc(const char *line, struct patch *patch)
383 {
384 patch->is_copy = 1;
385 patch->old_name = find_name(line, NULL, 0, 0);
386 return 0;
387 }
389 static int gitdiff_copydst(const char *line, struct patch *patch)
390 {
391 patch->is_copy = 1;
392 patch->new_name = find_name(line, NULL, 0, 0);
393 return 0;
394 }
396 static int gitdiff_renamesrc(const char *line, struct patch *patch)
397 {
398 patch->is_rename = 1;
399 patch->old_name = find_name(line, NULL, 0, 0);
400 return 0;
401 }
403 static int gitdiff_renamedst(const char *line, struct patch *patch)
404 {
405 patch->is_rename = 1;
406 patch->new_name = find_name(line, NULL, 0, 0);
407 return 0;
408 }
410 static int gitdiff_similarity(const char *line, struct patch *patch)
411 {
412 if ((patch->score = strtoul(line, NULL, 10)) == ULONG_MAX)
413 patch->score = 0;
414 return 0;
415 }
417 static int gitdiff_dissimilarity(const char *line, struct patch *patch)
418 {
419 if ((patch->score = strtoul(line, NULL, 10)) == ULONG_MAX)
420 patch->score = 0;
421 return 0;
422 }
424 static int gitdiff_index(const char *line, struct patch *patch)
425 {
426 /* index line is N hexadecimal, "..", N hexadecimal,
427 * and optional space with octal mode.
428 */
429 const char *ptr, *eol;
430 int len;
432 ptr = strchr(line, '.');
433 if (!ptr || ptr[1] != '.' || 40 < ptr - line)
434 return 0;
435 len = ptr - line;
436 memcpy(patch->old_sha1_prefix, line, len);
437 patch->old_sha1_prefix[len] = 0;
439 line = ptr + 2;
440 ptr = strchr(line, ' ');
441 eol = strchr(line, '\n');
443 if (!ptr || eol < ptr)
444 ptr = eol;
445 len = ptr - line;
447 if (40 < len)
448 return 0;
449 memcpy(patch->new_sha1_prefix, line, len);
450 patch->new_sha1_prefix[len] = 0;
451 if (*ptr == ' ')
452 patch->new_mode = patch->old_mode = strtoul(ptr+1, NULL, 8);
453 return 0;
454 }
456 /*
457 * This is normal for a diff that doesn't change anything: we'll fall through
458 * into the next diff. Tell the parser to break out.
459 */
460 static int gitdiff_unrecognized(const char *line, struct patch *patch)
461 {
462 return -1;
463 }
465 static const char *stop_at_slash(const char *line, int llen)
466 {
467 int i;
469 for (i = 0; i < llen; i++) {
470 int ch = line[i];
471 if (ch == '/')
472 return line + i;
473 }
474 return NULL;
475 }
477 /* This is to extract the same name that appears on "diff --git"
478 * line. We do not find and return anything if it is a rename
479 * patch, and it is OK because we will find the name elsewhere.
480 * We need to reliably find name only when it is mode-change only,
481 * creation or deletion of an empty file. In any of these cases,
482 * both sides are the same name under a/ and b/ respectively.
483 */
484 static char *git_header_name(char *line, int llen)
485 {
486 int len;
487 const char *name;
488 const char *second = NULL;
490 line += strlen("diff --git ");
491 llen -= strlen("diff --git ");
493 if (*line == '"') {
494 const char *cp;
495 char *first = unquote_c_style(line, &second);
496 if (!first)
497 return NULL;
499 /* advance to the first slash */
500 cp = stop_at_slash(first, strlen(first));
501 if (!cp || cp == first) {
502 /* we do not accept absolute paths */
503 free_first_and_fail:
504 free(first);
505 return NULL;
506 }
507 len = strlen(cp+1);
508 memmove(first, cp+1, len+1); /* including NUL */
510 /* second points at one past closing dq of name.
511 * find the second name.
512 */
513 while ((second < line + llen) && isspace(*second))
514 second++;
516 if (line + llen <= second)
517 goto free_first_and_fail;
518 if (*second == '"') {
519 char *sp = unquote_c_style(second, NULL);
520 if (!sp)
521 goto free_first_and_fail;
522 cp = stop_at_slash(sp, strlen(sp));
523 if (!cp || cp == sp) {
524 free_both_and_fail:
525 free(sp);
526 goto free_first_and_fail;
527 }
528 /* They must match, otherwise ignore */
529 if (strcmp(cp+1, first))
530 goto free_both_and_fail;
531 free(sp);
532 return first;
533 }
535 /* unquoted second */
536 cp = stop_at_slash(second, line + llen - second);
537 if (!cp || cp == second)
538 goto free_first_and_fail;
539 cp++;
540 if (line + llen - cp != len + 1 ||
541 memcmp(first, cp, len))
542 goto free_first_and_fail;
543 return first;
544 }
546 /* unquoted first name */
547 name = stop_at_slash(line, llen);
548 if (!name || name == line)
549 return NULL;
551 name++;
553 /* since the first name is unquoted, a dq if exists must be
554 * the beginning of the second name.
555 */
556 for (second = name; second < line + llen; second++) {
557 if (*second == '"') {
558 const char *cp = second;
559 const char *np;
560 char *sp = unquote_c_style(second, NULL);
562 if (!sp)
563 return NULL;
564 np = stop_at_slash(sp, strlen(sp));
565 if (!np || np == sp) {
566 free_second_and_fail:
567 free(sp);
568 return NULL;
569 }
570 np++;
571 len = strlen(np);
572 if (len < cp - name &&
573 !strncmp(np, name, len) &&
574 isspace(name[len])) {
575 /* Good */
576 memmove(sp, np, len + 1);
577 return sp;
578 }
579 goto free_second_and_fail;
580 }
581 }
583 /*
584 * Accept a name only if it shows up twice, exactly the same
585 * form.
586 */
587 for (len = 0 ; ; len++) {
588 char c = name[len];
590 switch (c) {
591 default:
592 continue;
593 case '\n':
594 return NULL;
595 case '\t': case ' ':
596 second = name+len;
597 for (;;) {
598 char c = *second++;
599 if (c == '\n')
600 return NULL;
601 if (c == '/')
602 break;
603 }
604 if (second[len] == '\n' && !memcmp(name, second, len)) {
605 char *ret = xmalloc(len + 1);
606 memcpy(ret, name, len);
607 ret[len] = 0;
608 return ret;
609 }
610 }
611 }
612 return NULL;
613 }
615 /* Verify that we recognize the lines following a git header */
616 static int parse_git_header(char *line, int len, unsigned int size, struct patch *patch)
617 {
618 unsigned long offset;
620 /* A git diff has explicit new/delete information, so we don't guess */
621 patch->is_new = 0;
622 patch->is_delete = 0;
624 /*
625 * Some things may not have the old name in the
626 * rest of the headers anywhere (pure mode changes,
627 * or removing or adding empty files), so we get
628 * the default name from the header.
629 */
630 patch->def_name = git_header_name(line, len);
632 line += len;
633 size -= len;
634 linenr++;
635 for (offset = len ; size > 0 ; offset += len, size -= len, line += len, linenr++) {
636 static const struct opentry {
637 const char *str;
638 int (*fn)(const char *, struct patch *);
639 } optable[] = {
640 { "@@ -", gitdiff_hdrend },
641 { "--- ", gitdiff_oldname },
642 { "+++ ", gitdiff_newname },
643 { "old mode ", gitdiff_oldmode },
644 { "new mode ", gitdiff_newmode },
645 { "deleted file mode ", gitdiff_delete },
646 { "new file mode ", gitdiff_newfile },
647 { "copy from ", gitdiff_copysrc },
648 { "copy to ", gitdiff_copydst },
649 { "rename old ", gitdiff_renamesrc },
650 { "rename new ", gitdiff_renamedst },
651 { "rename from ", gitdiff_renamesrc },
652 { "rename to ", gitdiff_renamedst },
653 { "similarity index ", gitdiff_similarity },
654 { "dissimilarity index ", gitdiff_dissimilarity },
655 { "index ", gitdiff_index },
656 { "", gitdiff_unrecognized },
657 };
658 int i;
660 len = linelen(line, size);
661 if (!len || line[len-1] != '\n')
662 break;
663 for (i = 0; i < ARRAY_SIZE(optable); i++) {
664 const struct opentry *p = optable + i;
665 int oplen = strlen(p->str);
666 if (len < oplen || memcmp(p->str, line, oplen))
667 continue;
668 if (p->fn(line + oplen, patch) < 0)
669 return offset;
670 break;
671 }
672 }
674 return offset;
675 }
677 static int parse_num(const char *line, unsigned long *p)
678 {
679 char *ptr;
681 if (!isdigit(*line))
682 return 0;
683 *p = strtoul(line, &ptr, 10);
684 return ptr - line;
685 }
687 static int parse_range(const char *line, int len, int offset, const char *expect,
688 unsigned long *p1, unsigned long *p2)
689 {
690 int digits, ex;
692 if (offset < 0 || offset >= len)
693 return -1;
694 line += offset;
695 len -= offset;
697 digits = parse_num(line, p1);
698 if (!digits)
699 return -1;
701 offset += digits;
702 line += digits;
703 len -= digits;
705 *p2 = 1;
706 if (*line == ',') {
707 digits = parse_num(line+1, p2);
708 if (!digits)
709 return -1;
711 offset += digits+1;
712 line += digits+1;
713 len -= digits+1;
714 }
716 ex = strlen(expect);
717 if (ex > len)
718 return -1;
719 if (memcmp(line, expect, ex))
720 return -1;
722 return offset + ex;
723 }
725 /*
726 * Parse a unified diff fragment header of the
727 * form "@@ -a,b +c,d @@"
728 */
729 static int parse_fragment_header(char *line, int len, struct fragment *fragment)
730 {
731 int offset;
733 if (!len || line[len-1] != '\n')
734 return -1;
736 /* Figure out the number of lines in a fragment */
737 offset = parse_range(line, len, 4, " +", &fragment->oldpos, &fragment->oldlines);
738 offset = parse_range(line, len, offset, " @@", &fragment->newpos, &fragment->newlines);
740 return offset;
741 }
743 static int find_header(char *line, unsigned long size, int *hdrsize, struct patch *patch)
744 {
745 unsigned long offset, len;
747 patch->is_rename = patch->is_copy = 0;
748 patch->is_new = patch->is_delete = -1;
749 patch->old_mode = patch->new_mode = 0;
750 patch->old_name = patch->new_name = NULL;
751 for (offset = 0; size > 0; offset += len, size -= len, line += len, linenr++) {
752 unsigned long nextlen;
754 len = linelen(line, size);
755 if (!len)
756 break;
758 /* Testing this early allows us to take a few shortcuts.. */
759 if (len < 6)
760 continue;
762 /*
763 * Make sure we don't find any unconnected patch fragmants.
764 * That's a sign that we didn't find a header, and that a
765 * patch has become corrupted/broken up.
766 */
767 if (!memcmp("@@ -", line, 4)) {
768 struct fragment dummy;
769 if (parse_fragment_header(line, len, &dummy) < 0)
770 continue;
771 error("patch fragment without header at line %d: %.*s", linenr, (int)len-1, line);
772 }
774 if (size < len + 6)
775 break;
777 /*
778 * Git patch? It might not have a real patch, just a rename
779 * or mode change, so we handle that specially
780 */
781 if (!memcmp("diff --git ", line, 11)) {
782 int git_hdr_len = parse_git_header(line, len, size, patch);
783 if (git_hdr_len <= len)
784 continue;
785 if (!patch->old_name && !patch->new_name) {
786 if (!patch->def_name)
787 die("git diff header lacks filename information (line %d)", linenr);
788 patch->old_name = patch->new_name = patch->def_name;
789 }
790 *hdrsize = git_hdr_len;
791 return offset;
792 }
794 /** --- followed by +++ ? */
795 if (memcmp("--- ", line, 4) || memcmp("+++ ", line + len, 4))
796 continue;
798 /*
799 * We only accept unified patches, so we want it to
800 * at least have "@@ -a,b +c,d @@\n", which is 14 chars
801 * minimum
802 */
803 nextlen = linelen(line + len, size - len);
804 if (size < nextlen + 14 || memcmp("@@ -", line + len + nextlen, 4))
805 continue;
807 /* Ok, we'll consider it a patch */
808 parse_traditional_patch(line, line+len, patch);
809 *hdrsize = len + nextlen;
810 linenr += 2;
811 return offset;
812 }
813 return -1;
814 }
816 /*
817 * Parse a unified diff. Note that this really needs
818 * to parse each fragment separately, since the only
819 * way to know the difference between a "---" that is
820 * part of a patch, and a "---" that starts the next
821 * patch is to look at the line counts..
822 */
823 static int parse_fragment(char *line, unsigned long size, struct patch *patch, struct fragment *fragment)
824 {
825 int added, deleted;
826 int len = linelen(line, size), offset;
827 unsigned long oldlines, newlines;
828 unsigned long leading, trailing;
830 offset = parse_fragment_header(line, len, fragment);
831 if (offset < 0)
832 return -1;
833 oldlines = fragment->oldlines;
834 newlines = fragment->newlines;
835 leading = 0;
836 trailing = 0;
838 if (patch->is_new < 0) {
839 patch->is_new = !oldlines;
840 if (!oldlines)
841 patch->old_name = NULL;
842 }
843 if (patch->is_delete < 0) {
844 patch->is_delete = !newlines;
845 if (!newlines)
846 patch->new_name = NULL;
847 }
849 if (patch->is_new && oldlines)
850 return error("new file depends on old contents");
851 if (patch->is_delete != !newlines) {
852 if (newlines)
853 return error("deleted file still has contents");
854 fprintf(stderr, "** warning: file %s becomes empty but is not deleted\n", patch->new_name);
855 }
857 /* Parse the thing.. */
858 line += len;
859 size -= len;
860 linenr++;
861 added = deleted = 0;
862 for (offset = len; size > 0; offset += len, size -= len, line += len, linenr++) {
863 if (!oldlines && !newlines)
864 break;
865 len = linelen(line, size);
866 if (!len || line[len-1] != '\n')
867 return -1;
868 switch (*line) {
869 default:
870 return -1;
871 case ' ':
872 oldlines--;
873 newlines--;
874 if (!deleted && !added)
875 leading++;
876 trailing++;
877 break;
878 case '-':
879 deleted++;
880 oldlines--;
881 trailing = 0;
882 break;
883 case '+':
884 /*
885 * We know len is at least two, since we have a '+' and
886 * we checked that the last character was a '\n' above.
887 * That is, an addition of an empty line would check
888 * the '+' here. Sneaky...
889 */
890 if ((new_whitespace != nowarn_whitespace) &&
891 isspace(line[len-2])) {
892 whitespace_error++;
893 if (squelch_whitespace_errors &&
894 squelch_whitespace_errors <
895 whitespace_error)
896 ;
897 else {
898 fprintf(stderr, "Adds trailing whitespace.\n%s:%d:%.*s\n",
899 patch_input_file,
900 linenr, len-2, line+1);
901 }
902 }
903 added++;
904 newlines--;
905 trailing = 0;
906 break;
908 /* We allow "\ No newline at end of file". Depending
909 * on locale settings when the patch was produced we
910 * don't know what this line looks like. The only
911 * thing we do know is that it begins with "\ ".
912 * Checking for 12 is just for sanity check -- any
913 * l10n of "\ No newline..." is at least that long.
914 */
915 case '\\':
916 if (len < 12 || memcmp(line, "\\ ", 2))
917 return -1;
918 break;
919 }
920 }
921 if (oldlines || newlines)
922 return -1;
923 fragment->leading = leading;
924 fragment->trailing = trailing;
926 /* If a fragment ends with an incomplete line, we failed to include
927 * it in the above loop because we hit oldlines == newlines == 0
928 * before seeing it.
929 */
930 if (12 < size && !memcmp(line, "\\ ", 2))
931 offset += linelen(line, size);
933 patch->lines_added += added;
934 patch->lines_deleted += deleted;
935 return offset;
936 }
938 static int parse_single_patch(char *line, unsigned long size, struct patch *patch)
939 {
940 unsigned long offset = 0;
941 struct fragment **fragp = &patch->fragments;
943 while (size > 4 && !memcmp(line, "@@ -", 4)) {
944 struct fragment *fragment;
945 int len;
947 fragment = xcalloc(1, sizeof(*fragment));
948 len = parse_fragment(line, size, patch, fragment);
949 if (len <= 0)
950 die("corrupt patch at line %d", linenr);
952 fragment->patch = line;
953 fragment->size = len;
955 *fragp = fragment;
956 fragp = &fragment->next;
958 offset += len;
959 line += len;
960 size -= len;
961 }
962 return offset;
963 }
965 static inline int metadata_changes(struct patch *patch)
966 {
967 return patch->is_rename > 0 ||
968 patch->is_copy > 0 ||
969 patch->is_new > 0 ||
970 patch->is_delete ||
971 (patch->old_mode && patch->new_mode &&
972 patch->old_mode != patch->new_mode);
973 }
975 static int parse_binary(char *buffer, unsigned long size, struct patch *patch)
976 {
977 /* We have read "GIT binary patch\n"; what follows is a line
978 * that says the patch method (currently, either "deflated
979 * literal" or "deflated delta") and the length of data before
980 * deflating; a sequence of 'length-byte' followed by base-85
981 * encoded data follows.
982 *
983 * Each 5-byte sequence of base-85 encodes up to 4 bytes,
984 * and we would limit the patch line to 66 characters,
985 * so one line can fit up to 13 groups that would decode
986 * to 52 bytes max. The length byte 'A'-'Z' corresponds
987 * to 1-26 bytes, and 'a'-'z' corresponds to 27-52 bytes.
988 * The end of binary is signalled with an empty line.
989 */
990 int llen, used;
991 struct fragment *fragment;
992 char *data = NULL;
994 patch->fragments = fragment = xcalloc(1, sizeof(*fragment));
996 /* Grab the type of patch */
997 llen = linelen(buffer, size);
998 used = llen;
999 linenr++;
1001 if (!strncmp(buffer, "delta ", 6)) {
1002 patch->is_binary = BINARY_DELTA_DEFLATED;
1003 patch->deflate_origlen = strtoul(buffer + 6, NULL, 10);
1004 }
1005 else if (!strncmp(buffer, "literal ", 8)) {
1006 patch->is_binary = BINARY_LITERAL_DEFLATED;
1007 patch->deflate_origlen = strtoul(buffer + 8, NULL, 10);
1008 }
1009 else
1010 return error("unrecognized binary patch at line %d: %.*s",
1011 linenr-1, llen-1, buffer);
1012 buffer += llen;
1013 while (1) {
1014 int byte_length, max_byte_length, newsize;
1015 llen = linelen(buffer, size);
1016 used += llen;
1017 linenr++;
1018 if (llen == 1)
1019 break;
1020 /* Minimum line is "A00000\n" which is 7-byte long,
1021 * and the line length must be multiple of 5 plus 2.
1022 */
1023 if ((llen < 7) || (llen-2) % 5)
1024 goto corrupt;
1025 max_byte_length = (llen - 2) / 5 * 4;
1026 byte_length = *buffer;
1027 if ('A' <= byte_length && byte_length <= 'Z')
1028 byte_length = byte_length - 'A' + 1;
1029 else if ('a' <= byte_length && byte_length <= 'z')
1030 byte_length = byte_length - 'a' + 27;
1031 else
1032 goto corrupt;
1033 /* if the input length was not multiple of 4, we would
1034 * have filler at the end but the filler should never
1035 * exceed 3 bytes
1036 */
1037 if (max_byte_length < byte_length ||
1038 byte_length <= max_byte_length - 4)
1039 goto corrupt;
1040 newsize = fragment->size + byte_length;
1041 data = xrealloc(data, newsize);
1042 if (decode_85(data + fragment->size,
1043 buffer + 1,
1044 byte_length))
1045 goto corrupt;
1046 fragment->size = newsize;
1047 buffer += llen;
1048 size -= llen;
1049 }
1050 fragment->patch = data;
1051 return used;
1052 corrupt:
1053 return error("corrupt binary patch at line %d: %.*s",
1054 linenr-1, llen-1, buffer);
1055 }
1057 static int parse_chunk(char *buffer, unsigned long size, struct patch *patch)
1058 {
1059 int hdrsize, patchsize;
1060 int offset = find_header(buffer, size, &hdrsize, patch);
1062 if (offset < 0)
1063 return offset;
1065 patchsize = parse_single_patch(buffer + offset + hdrsize, size - offset - hdrsize, patch);
1067 if (!patchsize) {
1068 static const char *binhdr[] = {
1069 "Binary files ",
1070 "Files ",
1071 NULL,
1072 };
1073 static const char git_binary[] = "GIT binary patch\n";
1074 int i;
1075 int hd = hdrsize + offset;
1076 unsigned long llen = linelen(buffer + hd, size - hd);
1078 if (llen == sizeof(git_binary) - 1 &&
1079 !memcmp(git_binary, buffer + hd, llen)) {
1080 int used;
1081 linenr++;
1082 used = parse_binary(buffer + hd + llen,
1083 size - hd - llen, patch);
1084 if (used)
1085 patchsize = used + llen;
1086 else
1087 patchsize = 0;
1088 }
1089 else if (!memcmp(" differ\n", buffer + hd + llen - 8, 8)) {
1090 for (i = 0; binhdr[i]; i++) {
1091 int len = strlen(binhdr[i]);
1092 if (len < size - hd &&
1093 !memcmp(binhdr[i], buffer + hd, len)) {
1094 linenr++;
1095 patch->is_binary = 1;
1096 patchsize = llen;
1097 break;
1098 }
1099 }
1100 }
1102 /* Empty patch cannot be applied if:
1103 * - it is a binary patch and we do not do binary_replace, or
1104 * - text patch without metadata change
1105 */
1106 if ((apply || check) &&
1107 (patch->is_binary
1108 ? !allow_binary_replacement
1109 : !metadata_changes(patch)))
1110 die("patch with only garbage at line %d", linenr);
1111 }
1113 return offset + hdrsize + patchsize;
1114 }
1116 static const char pluses[] = "++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++";
1117 static const char minuses[]= "----------------------------------------------------------------------";
1119 static void show_stats(struct patch *patch)
1120 {
1121 const char *prefix = "";
1122 char *name = patch->new_name;
1123 char *qname = NULL;
1124 int len, max, add, del, total;
1126 if (!name)
1127 name = patch->old_name;
1129 if (0 < (len = quote_c_style(name, NULL, NULL, 0))) {
1130 qname = xmalloc(len + 1);
1131 quote_c_style(name, qname, NULL, 0);
1132 name = qname;
1133 }
1135 /*
1136 * "scale" the filename
1137 */
1138 len = strlen(name);
1139 max = max_len;
1140 if (max > 50)
1141 max = 50;
1142 if (len > max) {
1143 char *slash;
1144 prefix = "...";
1145 max -= 3;
1146 name += len - max;
1147 slash = strchr(name, '/');
1148 if (slash)
1149 name = slash;
1150 }
1151 len = max;
1153 /*
1154 * scale the add/delete
1155 */
1156 max = max_change;
1157 if (max + len > 70)
1158 max = 70 - len;
1160 add = patch->lines_added;
1161 del = patch->lines_deleted;
1162 total = add + del;
1164 if (max_change > 0) {
1165 total = (total * max + max_change / 2) / max_change;
1166 add = (add * max + max_change / 2) / max_change;
1167 del = total - add;
1168 }
1169 if (patch->is_binary)
1170 printf(" %s%-*s | Bin\n", prefix, len, name);
1171 else
1172 printf(" %s%-*s |%5d %.*s%.*s\n", prefix,
1173 len, name, patch->lines_added + patch->lines_deleted,
1174 add, pluses, del, minuses);
1175 if (qname)
1176 free(qname);
1177 }
1179 static int read_old_data(struct stat *st, const char *path, void *buf, unsigned long size)
1180 {
1181 int fd;
1182 unsigned long got;
1184 switch (st->st_mode & S_IFMT) {
1185 case S_IFLNK:
1186 return readlink(path, buf, size);
1187 case S_IFREG:
1188 fd = open(path, O_RDONLY);
1189 if (fd < 0)
1190 return error("unable to open %s", path);
1191 got = 0;
1192 for (;;) {
1193 int ret = xread(fd, buf + got, size - got);
1194 if (ret <= 0)
1195 break;
1196 got += ret;
1197 }
1198 close(fd);
1199 return got;
1201 default:
1202 return -1;
1203 }
1204 }
1206 static int find_offset(const char *buf, unsigned long size, const char *fragment, unsigned long fragsize, int line, int *lines)
1207 {
1208 int i;
1209 unsigned long start, backwards, forwards;
1211 if (fragsize > size)
1212 return -1;
1214 start = 0;
1215 if (line > 1) {
1216 unsigned long offset = 0;
1217 i = line-1;
1218 while (offset + fragsize <= size) {
1219 if (buf[offset++] == '\n') {
1220 start = offset;
1221 if (!--i)
1222 break;
1223 }
1224 }
1225 }
1227 /* Exact line number? */
1228 if (!memcmp(buf + start, fragment, fragsize))
1229 return start;
1231 /*
1232 * There's probably some smart way to do this, but I'll leave
1233 * that to the smart and beautiful people. I'm simple and stupid.
1234 */
1235 backwards = start;
1236 forwards = start;
1237 for (i = 0; ; i++) {
1238 unsigned long try;
1239 int n;
1241 /* "backward" */
1242 if (i & 1) {
1243 if (!backwards) {
1244 if (forwards + fragsize > size)
1245 break;
1246 continue;
1247 }
1248 do {
1249 --backwards;
1250 } while (backwards && buf[backwards-1] != '\n');
1251 try = backwards;
1252 } else {
1253 while (forwards + fragsize <= size) {
1254 if (buf[forwards++] == '\n')
1255 break;
1256 }
1257 try = forwards;
1258 }
1260 if (try + fragsize > size)
1261 continue;
1262 if (memcmp(buf + try, fragment, fragsize))
1263 continue;
1264 n = (i >> 1)+1;
1265 if (i & 1)
1266 n = -n;
1267 *lines = n;
1268 return try;
1269 }
1271 /*
1272 * We should start searching forward and backward.
1273 */
1274 return -1;
1275 }
1277 static void remove_first_line(const char **rbuf, int *rsize)
1278 {
1279 const char *buf = *rbuf;
1280 int size = *rsize;
1281 unsigned long offset;
1282 offset = 0;
1283 while (offset <= size) {
1284 if (buf[offset++] == '\n')
1285 break;
1286 }
1287 *rsize = size - offset;
1288 *rbuf = buf + offset;
1289 }
1291 static void remove_last_line(const char **rbuf, int *rsize)
1292 {
1293 const char *buf = *rbuf;
1294 int size = *rsize;
1295 unsigned long offset;
1296 offset = size - 1;
1297 while (offset > 0) {
1298 if (buf[--offset] == '\n')
1299 break;
1300 }
1301 *rsize = offset + 1;
1302 }
1304 struct buffer_desc {
1305 char *buffer;
1306 unsigned long size;
1307 unsigned long alloc;
1308 };
1310 static int apply_line(char *output, const char *patch, int plen)
1311 {
1312 /* plen is number of bytes to be copied from patch,
1313 * starting at patch+1 (patch[0] is '+'). Typically
1314 * patch[plen] is '\n'.
1315 */
1316 int add_nl_to_tail = 0;
1317 if ((new_whitespace == strip_whitespace) &&
1318 1 < plen && isspace(patch[plen-1])) {
1319 if (patch[plen] == '\n')
1320 add_nl_to_tail = 1;
1321 plen--;
1322 while (0 < plen && isspace(patch[plen]))
1323 plen--;
1324 applied_after_stripping++;
1325 }
1326 memcpy(output, patch + 1, plen);
1327 if (add_nl_to_tail)
1328 output[plen++] = '\n';
1329 return plen;
1330 }
1332 static int apply_one_fragment(struct buffer_desc *desc, struct fragment *frag)
1333 {
1334 char *buf = desc->buffer;
1335 const char *patch = frag->patch;
1336 int offset, size = frag->size;
1337 char *old = xmalloc(size);
1338 char *new = xmalloc(size);
1339 const char *oldlines, *newlines;
1340 int oldsize = 0, newsize = 0;
1341 unsigned long leading, trailing;
1342 int pos, lines;
1344 while (size > 0) {
1345 int len = linelen(patch, size);
1346 int plen;
1348 if (!len)
1349 break;
1351 /*
1352 * "plen" is how much of the line we should use for
1353 * the actual patch data. Normally we just remove the
1354 * first character on the line, but if the line is
1355 * followed by "\ No newline", then we also remove the
1356 * last one (which is the newline, of course).
1357 */
1358 plen = len-1;
1359 if (len < size && patch[len] == '\\')
1360 plen--;
1361 switch (*patch) {
1362 case ' ':
1363 case '-':
1364 memcpy(old + oldsize, patch + 1, plen);
1365 oldsize += plen;
1366 if (*patch == '-')
1367 break;
1368 /* Fall-through for ' ' */
1369 case '+':
1370 if (*patch != '+' || !no_add)
1371 newsize += apply_line(new + newsize, patch,
1372 plen);
1373 break;
1374 case '@': case '\\':
1375 /* Ignore it, we already handled it */
1376 break;
1377 default:
1378 return -1;
1379 }
1380 patch += len;
1381 size -= len;
1382 }
1384 #ifdef NO_ACCURATE_DIFF
1385 if (oldsize > 0 && old[oldsize - 1] == '\n' &&
1386 newsize > 0 && new[newsize - 1] == '\n') {
1387 oldsize--;
1388 newsize--;
1389 }
1390 #endif
1392 oldlines = old;
1393 newlines = new;
1394 leading = frag->leading;
1395 trailing = frag->trailing;
1396 lines = 0;
1397 pos = frag->newpos;
1398 for (;;) {
1399 offset = find_offset(buf, desc->size, oldlines, oldsize, pos, &lines);
1400 if (offset >= 0) {
1401 int diff = newsize - oldsize;
1402 unsigned long size = desc->size + diff;
1403 unsigned long alloc = desc->alloc;
1405 /* Warn if it was necessary to reduce the number
1406 * of context lines.
1407 */
1408 if ((leading != frag->leading) || (trailing != frag->trailing))
1409 fprintf(stderr, "Context reduced to (%ld/%ld) to apply fragment at %d\n",
1410 leading, trailing, pos + lines);
1412 if (size > alloc) {
1413 alloc = size + 8192;
1414 desc->alloc = alloc;
1415 buf = xrealloc(buf, alloc);
1416 desc->buffer = buf;
1417 }
1418 desc->size = size;
1419 memmove(buf + offset + newsize, buf + offset + oldsize, size - offset - newsize);
1420 memcpy(buf + offset, newlines, newsize);
1421 offset = 0;
1423 break;
1424 }
1426 /* Am I at my context limits? */
1427 if ((leading <= p_context) && (trailing <= p_context))
1428 break;
1429 /* Reduce the number of context lines
1430 * Reduce both leading and trailing if they are equal
1431 * otherwise just reduce the larger context.
1432 */
1433 if (leading >= trailing) {
1434 remove_first_line(&oldlines, &oldsize);
1435 remove_first_line(&newlines, &newsize);
1436 pos--;
1437 leading--;
1438 }
1439 if (trailing > leading) {
1440 remove_last_line(&oldlines, &oldsize);
1441 remove_last_line(&newlines, &newsize);
1442 trailing--;
1443 }
1444 }
1446 free(old);
1447 free(new);
1448 return offset;
1449 }
1451 static char *inflate_it(const void *data, unsigned long size,
1452 unsigned long inflated_size)
1453 {
1454 z_stream stream;
1455 void *out;
1456 int st;
1458 memset(&stream, 0, sizeof(stream));
1460 stream.next_in = (unsigned char *)data;
1461 stream.avail_in = size;
1462 stream.next_out = out = xmalloc(inflated_size);
1463 stream.avail_out = inflated_size;
1464 inflateInit(&stream);
1465 st = inflate(&stream, Z_FINISH);
1466 if ((st != Z_STREAM_END) || stream.total_out != inflated_size) {
1467 free(out);
1468 return NULL;
1469 }
1470 return out;
1471 }
1473 static int apply_binary_fragment(struct buffer_desc *desc, struct patch *patch)
1474 {
1475 unsigned long dst_size;
1476 struct fragment *fragment = patch->fragments;
1477 void *data;
1478 void *result;
1480 data = inflate_it(fragment->patch, fragment->size,
1481 patch->deflate_origlen);
1482 if (!data)
1483 return error("corrupt patch data");
1484 switch (patch->is_binary) {
1485 case BINARY_DELTA_DEFLATED:
1486 result = patch_delta(desc->buffer, desc->size,
1487 data,
1488 patch->deflate_origlen,
1489 &dst_size);
1490 free(desc->buffer);
1491 desc->buffer = result;
1492 free(data);
1493 break;
1494 case BINARY_LITERAL_DEFLATED:
1495 free(desc->buffer);
1496 desc->buffer = data;
1497 dst_size = patch->deflate_origlen;
1498 break;
1499 }
1500 if (!desc->buffer)
1501 return -1;
1502 desc->size = desc->alloc = dst_size;
1503 return 0;
1504 }
1506 static int apply_binary(struct buffer_desc *desc, struct patch *patch)
1507 {
1508 const char *name = patch->old_name ? patch->old_name : patch->new_name;
1509 unsigned char sha1[20];
1510 unsigned char hdr[50];
1511 int hdrlen;
1513 if (!allow_binary_replacement)
1514 return error("cannot apply binary patch to '%s' "
1515 "without --allow-binary-replacement",
1516 name);
1518 /* For safety, we require patch index line to contain
1519 * full 40-byte textual SHA1 for old and new, at least for now.
1520 */
1521 if (strlen(patch->old_sha1_prefix) != 40 ||
1522 strlen(patch->new_sha1_prefix) != 40 ||
1523 get_sha1_hex(patch->old_sha1_prefix, sha1) ||
1524 get_sha1_hex(patch->new_sha1_prefix, sha1))
1525 return error("cannot apply binary patch to '%s' "
1526 "without full index line", name);
1528 if (patch->old_name) {
1529 /* See if the old one matches what the patch
1530 * applies to.
1531 */
1532 write_sha1_file_prepare(desc->buffer, desc->size,
1533 blob_type, sha1, hdr, &hdrlen);
1534 if (strcmp(sha1_to_hex(sha1), patch->old_sha1_prefix))
1535 return error("the patch applies to '%s' (%s), "
1536 "which does not match the "
1537 "current contents.",
1538 name, sha1_to_hex(sha1));
1539 }
1540 else {
1541 /* Otherwise, the old one must be empty. */
1542 if (desc->size)
1543 return error("the patch applies to an empty "
1544 "'%s' but it is not empty", name);
1545 }
1547 get_sha1_hex(patch->new_sha1_prefix, sha1);
1548 if (!memcmp(sha1, null_sha1, 20)) {
1549 free(desc->buffer);
1550 desc->alloc = desc->size = 0;
1551 desc->buffer = NULL;
1552 return 0; /* deletion patch */
1553 }
1555 if (has_sha1_file(sha1)) {
1556 /* We already have the postimage */
1557 char type[10];
1558 unsigned long size;
1560 free(desc->buffer);
1561 desc->buffer = read_sha1_file(sha1, type, &size);
1562 if (!desc->buffer)
1563 return error("the necessary postimage %s for "
1564 "'%s' cannot be read",
1565 patch->new_sha1_prefix, name);
1566 desc->alloc = desc->size = size;
1567 }
1568 else {
1569 /* We have verified desc matches the preimage;
1570 * apply the patch data to it, which is stored
1571 * in the patch->fragments->{patch,size}.
1572 */
1573 if (apply_binary_fragment(desc, patch))
1574 return error("binary patch does not apply to '%s'",
1575 name);
1577 /* verify that the result matches */
1578 write_sha1_file_prepare(desc->buffer, desc->size, blob_type,
1579 sha1, hdr, &hdrlen);
1580 if (strcmp(sha1_to_hex(sha1), patch->new_sha1_prefix))
1581 return error("binary patch to '%s' creates incorrect result", name);
1582 }
1584 return 0;
1585 }
1587 static int apply_fragments(struct buffer_desc *desc, struct patch *patch)
1588 {
1589 struct fragment *frag = patch->fragments;
1590 const char *name = patch->old_name ? patch->old_name : patch->new_name;
1592 if (patch->is_binary)
1593 return apply_binary(desc, patch);
1595 while (frag) {
1596 if (apply_one_fragment(desc, frag) < 0)
1597 return error("patch failed: %s:%ld",
1598 name, frag->oldpos);
1599 frag = frag->next;
1600 }
1601 return 0;
1602 }
1604 static int apply_data(struct patch *patch, struct stat *st)
1605 {
1606 char *buf;
1607 unsigned long size, alloc;
1608 struct buffer_desc desc;
1610 size = 0;
1611 alloc = 0;
1612 buf = NULL;
1613 if (patch->old_name) {
1614 size = st->st_size;
1615 alloc = size + 8192;
1616 buf = xmalloc(alloc);
1617 if (read_old_data(st, patch->old_name, buf, alloc) != size)
1618 return error("read of %s failed", patch->old_name);
1619 }
1621 desc.size = size;
1622 desc.alloc = alloc;
1623 desc.buffer = buf;
1624 if (apply_fragments(&desc, patch) < 0)
1625 return -1;
1626 patch->result = desc.buffer;
1627 patch->resultsize = desc.size;
1629 if (patch->is_delete && patch->resultsize)
1630 return error("removal patch leaves file contents");
1632 return 0;
1633 }
1635 static int check_patch(struct patch *patch)
1636 {
1637 struct stat st;
1638 const char *old_name = patch->old_name;
1639 const char *new_name = patch->new_name;
1640 const char *name = old_name ? old_name : new_name;
1642 if (old_name) {
1643 int changed;
1644 int stat_ret = lstat(old_name, &st);
1646 if (check_index) {
1647 int pos = cache_name_pos(old_name, strlen(old_name));
1648 if (pos < 0)
1649 return error("%s: does not exist in index",
1650 old_name);
1651 if (stat_ret < 0) {
1652 struct checkout costate;
1653 if (errno != ENOENT)
1654 return error("%s: %s", old_name,
1655 strerror(errno));
1656 /* checkout */
1657 costate.base_dir = "";
1658 costate.base_dir_len = 0;
1659 costate.force = 0;
1660 costate.quiet = 0;
1661 costate.not_new = 0;
1662 costate.refresh_cache = 1;
1663 if (checkout_entry(active_cache[pos],
1664 &costate,
1665 NULL) ||
1666 lstat(old_name, &st))
1667 return -1;
1668 }
1670 changed = ce_match_stat(active_cache[pos], &st, 1);
1671 if (changed)
1672 return error("%s: does not match index",
1673 old_name);
1674 }
1675 else if (stat_ret < 0)
1676 return error("%s: %s", old_name, strerror(errno));
1678 if (patch->is_new < 0)
1679 patch->is_new = 0;
1680 st.st_mode = ntohl(create_ce_mode(st.st_mode));
1681 if (!patch->old_mode)
1682 patch->old_mode = st.st_mode;
1683 if ((st.st_mode ^ patch->old_mode) & S_IFMT)
1684 return error("%s: wrong type", old_name);
1685 if (st.st_mode != patch->old_mode)
1686 fprintf(stderr, "warning: %s has type %o, expected %o\n",
1687 old_name, st.st_mode, patch->old_mode);
1688 }
1690 if (new_name && (patch->is_new | patch->is_rename | patch->is_copy)) {
1691 if (check_index && cache_name_pos(new_name, strlen(new_name)) >= 0)
1692 return error("%s: already exists in index", new_name);
1693 if (!lstat(new_name, &st))
1694 return error("%s: already exists in working directory", new_name);
1695 if (errno != ENOENT)
1696 return error("%s: %s", new_name, strerror(errno));
1697 if (!patch->new_mode) {
1698 if (patch->is_new)
1699 patch->new_mode = S_IFREG | 0644;
1700 else
1701 patch->new_mode = patch->old_mode;
1702 }
1703 }
1705 if (new_name && old_name) {
1706 int same = !strcmp(old_name, new_name);
1707 if (!patch->new_mode)
1708 patch->new_mode = patch->old_mode;
1709 if ((patch->old_mode ^ patch->new_mode) & S_IFMT)
1710 return error("new mode (%o) of %s does not match old mode (%o)%s%s",
1711 patch->new_mode, new_name, patch->old_mode,
1712 same ? "" : " of ", same ? "" : old_name);
1713 }
1715 if (apply_data(patch, &st) < 0)
1716 return error("%s: patch does not apply", name);
1717 return 0;
1718 }
1720 static int check_patch_list(struct patch *patch)
1721 {
1722 int error = 0;
1724 for (;patch ; patch = patch->next)
1725 error |= check_patch(patch);
1726 return error;
1727 }
1729 static inline int is_null_sha1(const unsigned char *sha1)
1730 {
1731 return !memcmp(sha1, null_sha1, 20);
1732 }
1734 static void show_index_list(struct patch *list)
1735 {
1736 struct patch *patch;
1738 /* Once we start supporting the reverse patch, it may be
1739 * worth showing the new sha1 prefix, but until then...
1740 */
1741 for (patch = list; patch; patch = patch->next) {
1742 const unsigned char *sha1_ptr;
1743 unsigned char sha1[20];
1744 const char *name;
1746 name = patch->old_name ? patch->old_name : patch->new_name;
1747 if (patch->is_new)
1748 sha1_ptr = null_sha1;
1749 else if (get_sha1(patch->old_sha1_prefix, sha1))
1750 die("sha1 information is lacking or useless (%s).",
1751 name);
1752 else
1753 sha1_ptr = sha1;
1755 printf("%06o %s ",patch->old_mode, sha1_to_hex(sha1_ptr));
1756 if (line_termination && quote_c_style(name, NULL, NULL, 0))
1757 quote_c_style(name, NULL, stdout, 0);
1758 else
1759 fputs(name, stdout);
1760 putchar(line_termination);
1761 }
1762 }
1764 static void stat_patch_list(struct patch *patch)
1765 {
1766 int files, adds, dels;
1768 for (files = adds = dels = 0 ; patch ; patch = patch->next) {
1769 files++;
1770 adds += patch->lines_added;
1771 dels += patch->lines_deleted;
1772 show_stats(patch);
1773 }
1775 printf(" %d files changed, %d insertions(+), %d deletions(-)\n", files, adds, dels);
1776 }
1778 static void numstat_patch_list(struct patch *patch)
1779 {
1780 for ( ; patch; patch = patch->next) {
1781 const char *name;
1782 name = patch->old_name ? patch->old_name : patch->new_name;
1783 printf("%d\t%d\t", patch->lines_added, patch->lines_deleted);
1784 if (line_termination && quote_c_style(name, NULL, NULL, 0))
1785 quote_c_style(name, NULL, stdout, 0);
1786 else
1787 fputs(name, stdout);
1788 putchar('\n');
1789 }
1790 }
1792 static void show_file_mode_name(const char *newdelete, unsigned int mode, const char *name)
1793 {
1794 if (mode)
1795 printf(" %s mode %06o %s\n", newdelete, mode, name);
1796 else
1797 printf(" %s %s\n", newdelete, name);
1798 }
1800 static void show_mode_change(struct patch *p, int show_name)
1801 {
1802 if (p->old_mode && p->new_mode && p->old_mode != p->new_mode) {
1803 if (show_name)
1804 printf(" mode change %06o => %06o %s\n",
1805 p->old_mode, p->new_mode, p->new_name);
1806 else
1807 printf(" mode change %06o => %06o\n",
1808 p->old_mode, p->new_mode);
1809 }
1810 }
1812 static void show_rename_copy(struct patch *p)
1813 {
1814 const char *renamecopy = p->is_rename ? "rename" : "copy";
1815 const char *old, *new;
1817 /* Find common prefix */
1818 old = p->old_name;
1819 new = p->new_name;
1820 while (1) {
1821 const char *slash_old, *slash_new;
1822 slash_old = strchr(old, '/');
1823 slash_new = strchr(new, '/');
1824 if (!slash_old ||
1825 !slash_new ||
1826 slash_old - old != slash_new - new ||
1827 memcmp(old, new, slash_new - new))
1828 break;
1829 old = slash_old + 1;
1830 new = slash_new + 1;
1831 }
1832 /* p->old_name thru old is the common prefix, and old and new
1833 * through the end of names are renames
1834 */
1835 if (old != p->old_name)
1836 printf(" %s %.*s{%s => %s} (%d%%)\n", renamecopy,
1837 (int)(old - p->old_name), p->old_name,
1838 old, new, p->score);
1839 else
1840 printf(" %s %s => %s (%d%%)\n", renamecopy,
1841 p->old_name, p->new_name, p->score);
1842 show_mode_change(p, 0);
1843 }
1845 static void summary_patch_list(struct patch *patch)
1846 {
1847 struct patch *p;
1849 for (p = patch; p; p = p->next) {
1850 if (p->is_new)
1851 show_file_mode_name("create", p->new_mode, p->new_name);
1852 else if (p->is_delete)
1853 show_file_mode_name("delete", p->old_mode, p->old_name);
1854 else {
1855 if (p->is_rename || p->is_copy)
1856 show_rename_copy(p);
1857 else {
1858 if (p->score) {
1859 printf(" rewrite %s (%d%%)\n",
1860 p->new_name, p->score);
1861 show_mode_change(p, 0);
1862 }
1863 else
1864 show_mode_change(p, 1);
1865 }
1866 }
1867 }
1868 }
1870 static void patch_stats(struct patch *patch)
1871 {
1872 int lines = patch->lines_added + patch->lines_deleted;
1874 if (lines > max_change)
1875 max_change = lines;
1876 if (patch->old_name) {
1877 int len = quote_c_style(patch->old_name, NULL, NULL, 0);
1878 if (!len)
1879 len = strlen(patch->old_name);
1880 if (len > max_len)
1881 max_len = len;
1882 }
1883 if (patch->new_name) {
1884 int len = quote_c_style(patch->new_name, NULL, NULL, 0);
1885 if (!len)
1886 len = strlen(patch->new_name);
1887 if (len > max_len)
1888 max_len = len;
1889 }
1890 }
1892 static void remove_file(struct patch *patch)
1893 {
1894 if (write_index) {
1895 if (remove_file_from_cache(patch->old_name) < 0)
1896 die("unable to remove %s from index", patch->old_name);
1897 cache_tree_invalidate_path(active_cache_tree, patch->old_name);
1898 }
1899 unlink(patch->old_name);
1900 }
1902 static void add_index_file(const char *path, unsigned mode, void *buf, unsigned long size)
1903 {
1904 struct stat st;
1905 struct cache_entry *ce;
1906 int namelen = strlen(path);
1907 unsigned ce_size = cache_entry_size(namelen);
1909 if (!write_index)
1910 return;
1912 ce = xcalloc(1, ce_size);
1913 memcpy(ce->name, path, namelen);
1914 ce->ce_mode = create_ce_mode(mode);
1915 ce->ce_flags = htons(namelen);
1916 if (lstat(path, &st) < 0)
1917 die("unable to stat newly created file %s", path);
1918 fill_stat_cache_info(ce, &st);
1919 if (write_sha1_file(buf, size, blob_type, ce->sha1) < 0)
1920 die("unable to create backing store for newly created file %s", path);
1921 if (add_cache_entry(ce, ADD_CACHE_OK_TO_ADD) < 0)
1922 die("unable to add cache entry for %s", path);
1923 }
1925 static int try_create_file(const char *path, unsigned int mode, const char *buf, unsigned long size)
1926 {
1927 int fd;
1929 if (S_ISLNK(mode))
1930 return symlink(buf, path);
1931 fd = open(path, O_CREAT | O_EXCL | O_WRONLY, (mode & 0100) ? 0777 : 0666);
1932 if (fd < 0)
1933 return -1;
1934 while (size) {
1935 int written = xwrite(fd, buf, size);
1936 if (written < 0)
1937 die("writing file %s: %s", path, strerror(errno));
1938 if (!written)
1939 die("out of space writing file %s", path);
1940 buf += written;
1941 size -= written;
1942 }
1943 if (close(fd) < 0)
1944 die("closing file %s: %s", path, strerror(errno));
1945 return 0;
1946 }
1948 /*
1949 * We optimistically assume that the directories exist,
1950 * which is true 99% of the time anyway. If they don't,
1951 * we create them and try again.
1952 */
1953 static void create_one_file(char *path, unsigned mode, const char *buf, unsigned long size)
1954 {
1955 if (!try_create_file(path, mode, buf, size))
1956 return;
1958 if (errno == ENOENT) {
1959 if (safe_create_leading_directories(path))
1960 return;
1961 if (!try_create_file(path, mode, buf, size))
1962 return;
1963 }
1965 if (errno == EEXIST) {
1966 unsigned int nr = getpid();
1968 for (;;) {
1969 const char *newpath;
1970 newpath = mkpath("%s~%u", path, nr);
1971 if (!try_create_file(newpath, mode, buf, size)) {
1972 if (!rename(newpath, path))
1973 return;
1974 unlink(newpath);
1975 break;
1976 }
1977 if (errno != EEXIST)
1978 break;
1979 ++nr;
1980 }
1981 }
1982 die("unable to write file %s mode %o", path, mode);
1983 }
1985 static void create_file(struct patch *patch)
1986 {
1987 char *path = patch->new_name;
1988 unsigned mode = patch->new_mode;
1989 unsigned long size = patch->resultsize;
1990 char *buf = patch->result;
1992 if (!mode)
1993 mode = S_IFREG | 0644;
1994 create_one_file(path, mode, buf, size);
1995 add_index_file(path, mode, buf, size);
1996 cache_tree_invalidate_path(active_cache_tree, path);
1997 }
1999 static void write_out_one_result(struct patch *patch)
2000 {
2001 if (patch->is_delete > 0) {
2002 remove_file(patch);
2003 return;
2004 }
2005 if (patch->is_new > 0 || patch->is_copy) {
2006 create_file(patch);
2007 return;
2008 }
2009 /*
2010 * Rename or modification boils down to the same
2011 * thing: remove the old, write the new
2012 */
2013 remove_file(patch);
2014 create_file(patch);
2015 }
2017 static void write_out_results(struct patch *list, int skipped_patch)
2018 {
2019 if (!list && !skipped_patch)
2020 die("No changes");
2022 while (list) {
2023 write_out_one_result(list);
2024 list = list->next;
2025 }
2026 }
2028 static struct cache_file cache_file;
2030 static struct excludes {
2031 struct excludes *next;
2032 const char *path;
2033 } *excludes;
2035 static int use_patch(struct patch *p)
2036 {
2037 const char *pathname = p->new_name ? p->new_name : p->old_name;
2038 struct excludes *x = excludes;
2039 while (x) {
2040 if (fnmatch(x->path, pathname, 0) == 0)
2041 return 0;
2042 x = x->next;
2043 }
2044 if (0 < prefix_length) {
2045 int pathlen = strlen(pathname);
2046 if (pathlen <= prefix_length ||
2047 memcmp(prefix, pathname, prefix_length))
2048 return 0;
2049 }
2050 return 1;
2051 }
2053 static int apply_patch(int fd, const char *filename)
2054 {
2055 unsigned long offset, size;
2056 char *buffer = read_patch_file(fd, &size);
2057 struct patch *list = NULL, **listp = &list;
2058 int skipped_patch = 0;
2060 patch_input_file = filename;
2061 if (!buffer)
2062 return -1;
2063 offset = 0;
2064 while (size > 0) {
2065 struct patch *patch;
2066 int nr;
2068 patch = xcalloc(1, sizeof(*patch));
2069 nr = parse_chunk(buffer + offset, size, patch);
2070 if (nr < 0)
2071 break;
2072 if (use_patch(patch)) {
2073 patch_stats(patch);
2074 *listp = patch;
2075 listp = &patch->next;
2076 } else {
2077 /* perhaps free it a bit better? */
2078 free(patch);
2079 skipped_patch++;
2080 }
2081 offset += nr;
2082 size -= nr;
2083 }
2085 if (whitespace_error && (new_whitespace == error_on_whitespace))
2086 apply = 0;
2088 write_index = check_index && apply;
2089 if (write_index && newfd < 0)
2090 newfd = hold_index_file_for_update(&cache_file, get_index_file());
2091 if (check_index) {
2092 if (read_cache() < 0)
2093 die("unable to read index file");
2094 }
2096 if ((check || apply) && check_patch_list(list) < 0)
2097 exit(1);
2099 if (apply)
2100 write_out_results(list, skipped_patch);
2102 if (show_index_info)
2103 show_index_list(list);
2105 if (diffstat)
2106 stat_patch_list(list);
2108 if (numstat)
2109 numstat_patch_list(list);
2111 if (summary)
2112 summary_patch_list(list);
2114 free(buffer);
2115 return 0;
2116 }
2118 static int git_apply_config(const char *var, const char *value)
2119 {
2120 if (!strcmp(var, "apply.whitespace")) {
2121 apply_default_whitespace = strdup(value);
2122 return 0;
2123 }
2124 return git_default_config(var, value);
2125 }
2128 int main(int argc, char **argv)
2129 {
2130 int i;
2131 int read_stdin = 1;
2132 const char *whitespace_option = NULL;
2134 for (i = 1; i < argc; i++) {
2135 const char *arg = argv[i];
2136 char *end;
2137 int fd;
2139 if (!strcmp(arg, "-")) {
2140 apply_patch(0, "<stdin>");
2141 read_stdin = 0;
2142 continue;
2143 }
2144 if (!strncmp(arg, "--exclude=", 10)) {
2145 struct excludes *x = xmalloc(sizeof(*x));
2146 x->path = arg + 10;
2147 x->next = excludes;
2148 excludes = x;
2149 continue;
2150 }
2151 if (!strncmp(arg, "-p", 2)) {
2152 p_value = atoi(arg + 2);
2153 continue;
2154 }
2155 if (!strcmp(arg, "--no-add")) {
2156 no_add = 1;
2157 continue;
2158 }
2159 if (!strcmp(arg, "--stat")) {
2160 apply = 0;
2161 diffstat = 1;
2162 continue;
2163 }
2164 if (!strcmp(arg, "--allow-binary-replacement") ||
2165 !strcmp(arg, "--binary")) {
2166 allow_binary_replacement = 1;
2167 continue;
2168 }
2169 if (!strcmp(arg, "--numstat")) {
2170 apply = 0;
2171 numstat = 1;
2172 continue;
2173 }
2174 if (!strcmp(arg, "--summary")) {
2175 apply = 0;
2176 summary = 1;
2177 continue;
2178 }
2179 if (!strcmp(arg, "--check")) {
2180 apply = 0;
2181 check = 1;
2182 continue;
2183 }
2184 if (!strcmp(arg, "--index")) {
2185 check_index = 1;
2186 continue;
2187 }
2188 if (!strcmp(arg, "--apply")) {
2189 apply = 1;
2190 continue;
2191 }
2192 if (!strcmp(arg, "--index-info")) {
2193 apply = 0;
2194 show_index_info = 1;
2195 continue;
2196 }
2197 if (!strcmp(arg, "-z")) {
2198 line_termination = 0;
2199 continue;
2200 }
2201 if (!strncmp(arg, "-C", 2)) {
2202 p_context = strtoul(arg + 2, &end, 0);
2203 if (*end != '\0')
2204 die("unrecognized context count '%s'", arg + 2);
2205 continue;
2206 }
2207 if (!strncmp(arg, "--whitespace=", 13)) {
2208 whitespace_option = arg + 13;
2209 parse_whitespace_option(arg + 13);
2210 continue;
2211 }
2213 if (check_index && prefix_length < 0) {
2214 prefix = setup_git_directory();
2215 prefix_length = prefix ? strlen(prefix) : 0;
2216 git_config(git_apply_config);
2217 if (!whitespace_option && apply_default_whitespace)
2218 parse_whitespace_option(apply_default_whitespace);
2219 }
2220 if (0 < prefix_length)
2221 arg = prefix_filename(prefix, prefix_length, arg);
2223 fd = open(arg, O_RDONLY);
2224 if (fd < 0)
2225 usage(apply_usage);
2226 read_stdin = 0;
2227 set_default_whitespace_mode(whitespace_option);
2228 apply_patch(fd, arg);
2229 close(fd);
2230 }
2231 set_default_whitespace_mode(whitespace_option);
2232 if (read_stdin)
2233 apply_patch(0, "<stdin>");
2234 if (whitespace_error) {
2235 if (squelch_whitespace_errors &&
2236 squelch_whitespace_errors < whitespace_error) {
2237 int squelched =
2238 whitespace_error - squelch_whitespace_errors;
2239 fprintf(stderr, "warning: squelched %d whitespace error%s\n",
2240 squelched,
2241 squelched == 1 ? "" : "s");
2242 }
2243 if (new_whitespace == error_on_whitespace)
2244 die("%d line%s add%s trailing whitespaces.",
2245 whitespace_error,
2246 whitespace_error == 1 ? "" : "s",
2247 whitespace_error == 1 ? "s" : "");
2248 if (applied_after_stripping)
2249 fprintf(stderr, "warning: %d line%s applied after"
2250 " stripping trailing whitespaces.\n",
2251 applied_after_stripping,
2252 applied_after_stripping == 1 ? "" : "s");
2253 else if (whitespace_error)
2254 fprintf(stderr, "warning: %d line%s add%s trailing"
2255 " whitespaces.\n",
2256 whitespace_error,
2257 whitespace_error == 1 ? "" : "s",
2258 whitespace_error == 1 ? "s" : "");
2259 }
2261 if (write_index) {
2262 if (write_cache(newfd, active_cache, active_nr) ||
2263 commit_index_file(&cache_file))
2264 die("Unable to write new cachefile");
2265 }
2267 return 0;
2268 }