binary diff and apply: testsuite.
[git/dscho.git] / apply.c
blob1b93aab8af1a5498e69fe3cffe7cdfb527fa4317
1 /*
2 * apply.c
4 * Copyright (C) Linus Torvalds, 2005
6 * This applies patches on top of some (arbitrary) version of the SCM.
8 */
9 #include <fnmatch.h>
10 #include "cache.h"
11 #include "quote.h"
12 #include "blob.h"
13 #include "delta.h"
15 // --check turns on checking that the working tree matches the
16 // files that are being modified, but doesn't apply the patch
17 // --stat does just a diffstat, and doesn't actually apply
18 // --numstat does numeric diffstat, and doesn't actually apply
19 // --index-info shows the old and new index info for paths if available.
21 static const char *prefix;
22 static int prefix_length = -1;
24 static int p_value = 1;
25 static int allow_binary_replacement = 0;
26 static int check_index = 0;
27 static int write_index = 0;
28 static int diffstat = 0;
29 static int numstat = 0;
30 static int summary = 0;
31 static int check = 0;
32 static int apply = 1;
33 static int no_add = 0;
34 static int show_index_info = 0;
35 static int line_termination = '\n';
36 static unsigned long p_context = -1;
37 static const char apply_usage[] =
38 "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>...";
40 static enum whitespace_eol {
41 nowarn_whitespace,
42 warn_on_whitespace,
43 error_on_whitespace,
44 strip_whitespace,
45 } new_whitespace = warn_on_whitespace;
46 static int whitespace_error = 0;
47 static int squelch_whitespace_errors = 5;
48 static int applied_after_stripping = 0;
49 static const char *patch_input_file = NULL;
51 static void parse_whitespace_option(const char *option)
53 if (!option) {
54 new_whitespace = warn_on_whitespace;
55 return;
57 if (!strcmp(option, "warn")) {
58 new_whitespace = warn_on_whitespace;
59 return;
61 if (!strcmp(option, "nowarn")) {
62 new_whitespace = nowarn_whitespace;
63 return;
65 if (!strcmp(option, "error")) {
66 new_whitespace = error_on_whitespace;
67 return;
69 if (!strcmp(option, "error-all")) {
70 new_whitespace = error_on_whitespace;
71 squelch_whitespace_errors = 0;
72 return;
74 if (!strcmp(option, "strip")) {
75 new_whitespace = strip_whitespace;
76 return;
78 die("unrecognized whitespace option '%s'", option);
81 static void set_default_whitespace_mode(const char *whitespace_option)
83 if (!whitespace_option && !apply_default_whitespace) {
84 new_whitespace = (apply
85 ? warn_on_whitespace
86 : nowarn_whitespace);
91 * For "diff-stat" like behaviour, we keep track of the biggest change
92 * we've seen, and the longest filename. That allows us to do simple
93 * scaling.
95 static int max_change, max_len;
98 * Various "current state", notably line numbers and what
99 * file (and how) we're patching right now.. The "is_xxxx"
100 * things are flags, where -1 means "don't know yet".
102 static int linenr = 1;
104 struct fragment {
105 unsigned long leading, trailing;
106 unsigned long oldpos, oldlines;
107 unsigned long newpos, newlines;
108 const char *patch;
109 int size;
110 struct fragment *next;
113 struct patch {
114 char *new_name, *old_name, *def_name;
115 unsigned int old_mode, new_mode;
116 int is_rename, is_copy, is_new, is_delete, is_binary;
117 #define BINARY_DELTA_DEFLATED 1
118 #define BINARY_LITERAL_DEFLATED 2
119 unsigned long deflate_origlen;
120 int lines_added, lines_deleted;
121 int score;
122 struct fragment *fragments;
123 char *result;
124 unsigned long resultsize;
125 char old_sha1_prefix[41];
126 char new_sha1_prefix[41];
127 struct patch *next;
130 #define CHUNKSIZE (8192)
131 #define SLOP (16)
133 static void *read_patch_file(int fd, unsigned long *sizep)
135 unsigned long size = 0, alloc = CHUNKSIZE;
136 void *buffer = xmalloc(alloc);
138 for (;;) {
139 int nr = alloc - size;
140 if (nr < 1024) {
141 alloc += CHUNKSIZE;
142 buffer = xrealloc(buffer, alloc);
143 nr = alloc - size;
145 nr = xread(fd, buffer + size, nr);
146 if (!nr)
147 break;
148 if (nr < 0)
149 die("git-apply: read returned %s", strerror(errno));
150 size += nr;
152 *sizep = size;
155 * Make sure that we have some slop in the buffer
156 * so that we can do speculative "memcmp" etc, and
157 * see to it that it is NUL-filled.
159 if (alloc < size + SLOP)
160 buffer = xrealloc(buffer, size + SLOP);
161 memset(buffer + size, 0, SLOP);
162 return buffer;
165 static unsigned long linelen(const char *buffer, unsigned long size)
167 unsigned long len = 0;
168 while (size--) {
169 len++;
170 if (*buffer++ == '\n')
171 break;
173 return len;
176 static int is_dev_null(const char *str)
178 return !memcmp("/dev/null", str, 9) && isspace(str[9]);
181 #define TERM_SPACE 1
182 #define TERM_TAB 2
184 static int name_terminate(const char *name, int namelen, int c, int terminate)
186 if (c == ' ' && !(terminate & TERM_SPACE))
187 return 0;
188 if (c == '\t' && !(terminate & TERM_TAB))
189 return 0;
191 return 1;
194 static char * find_name(const char *line, char *def, int p_value, int terminate)
196 int len;
197 const char *start = line;
198 char *name;
200 if (*line == '"') {
201 /* Proposed "new-style" GNU patch/diff format; see
202 * http://marc.theaimsgroup.com/?l=git&m=112927316408690&w=2
204 name = unquote_c_style(line, NULL);
205 if (name) {
206 char *cp = name;
207 while (p_value) {
208 cp = strchr(name, '/');
209 if (!cp)
210 break;
211 cp++;
212 p_value--;
214 if (cp) {
215 /* name can later be freed, so we need
216 * to memmove, not just return cp
218 memmove(name, cp, strlen(cp) + 1);
219 free(def);
220 return name;
222 else {
223 free(name);
224 name = NULL;
229 for (;;) {
230 char c = *line;
232 if (isspace(c)) {
233 if (c == '\n')
234 break;
235 if (name_terminate(start, line-start, c, terminate))
236 break;
238 line++;
239 if (c == '/' && !--p_value)
240 start = line;
242 if (!start)
243 return def;
244 len = line - start;
245 if (!len)
246 return def;
249 * Generally we prefer the shorter name, especially
250 * if the other one is just a variation of that with
251 * something else tacked on to the end (ie "file.orig"
252 * or "file~").
254 if (def) {
255 int deflen = strlen(def);
256 if (deflen < len && !strncmp(start, def, deflen))
257 return def;
260 name = xmalloc(len + 1);
261 memcpy(name, start, len);
262 name[len] = 0;
263 free(def);
264 return name;
268 * Get the name etc info from the --/+++ lines of a traditional patch header
270 * NOTE! This hardcodes "-p1" behaviour in filename detection.
272 * FIXME! The end-of-filename heuristics are kind of screwy. For existing
273 * files, we can happily check the index for a match, but for creating a
274 * new file we should try to match whatever "patch" does. I have no idea.
276 static void parse_traditional_patch(const char *first, const char *second, struct patch *patch)
278 char *name;
280 first += 4; // skip "--- "
281 second += 4; // skip "+++ "
282 if (is_dev_null(first)) {
283 patch->is_new = 1;
284 patch->is_delete = 0;
285 name = find_name(second, NULL, p_value, TERM_SPACE | TERM_TAB);
286 patch->new_name = name;
287 } else if (is_dev_null(second)) {
288 patch->is_new = 0;
289 patch->is_delete = 1;
290 name = find_name(first, NULL, p_value, TERM_SPACE | TERM_TAB);
291 patch->old_name = name;
292 } else {
293 name = find_name(first, NULL, p_value, TERM_SPACE | TERM_TAB);
294 name = find_name(second, name, p_value, TERM_SPACE | TERM_TAB);
295 patch->old_name = patch->new_name = name;
297 if (!name)
298 die("unable to find filename in patch at line %d", linenr);
301 static int gitdiff_hdrend(const char *line, struct patch *patch)
303 return -1;
307 * We're anal about diff header consistency, to make
308 * sure that we don't end up having strange ambiguous
309 * patches floating around.
311 * As a result, gitdiff_{old|new}name() will check
312 * their names against any previous information, just
313 * to make sure..
315 static char *gitdiff_verify_name(const char *line, int isnull, char *orig_name, const char *oldnew)
317 if (!orig_name && !isnull)
318 return find_name(line, NULL, 1, 0);
320 if (orig_name) {
321 int len;
322 const char *name;
323 char *another;
324 name = orig_name;
325 len = strlen(name);
326 if (isnull)
327 die("git-apply: bad git-diff - expected /dev/null, got %s on line %d", name, linenr);
328 another = find_name(line, NULL, 1, 0);
329 if (!another || memcmp(another, name, len))
330 die("git-apply: bad git-diff - inconsistent %s filename on line %d", oldnew, linenr);
331 free(another);
332 return orig_name;
334 else {
335 /* expect "/dev/null" */
336 if (memcmp("/dev/null", line, 9) || line[9] != '\n')
337 die("git-apply: bad git-diff - expected /dev/null on line %d", linenr);
338 return NULL;
342 static int gitdiff_oldname(const char *line, struct patch *patch)
344 patch->old_name = gitdiff_verify_name(line, patch->is_new, patch->old_name, "old");
345 return 0;
348 static int gitdiff_newname(const char *line, struct patch *patch)
350 patch->new_name = gitdiff_verify_name(line, patch->is_delete, patch->new_name, "new");
351 return 0;
354 static int gitdiff_oldmode(const char *line, struct patch *patch)
356 patch->old_mode = strtoul(line, NULL, 8);
357 return 0;
360 static int gitdiff_newmode(const char *line, struct patch *patch)
362 patch->new_mode = strtoul(line, NULL, 8);
363 return 0;
366 static int gitdiff_delete(const char *line, struct patch *patch)
368 patch->is_delete = 1;
369 patch->old_name = patch->def_name;
370 return gitdiff_oldmode(line, patch);
373 static int gitdiff_newfile(const char *line, struct patch *patch)
375 patch->is_new = 1;
376 patch->new_name = patch->def_name;
377 return gitdiff_newmode(line, patch);
380 static int gitdiff_copysrc(const char *line, struct patch *patch)
382 patch->is_copy = 1;
383 patch->old_name = find_name(line, NULL, 0, 0);
384 return 0;
387 static int gitdiff_copydst(const char *line, struct patch *patch)
389 patch->is_copy = 1;
390 patch->new_name = find_name(line, NULL, 0, 0);
391 return 0;
394 static int gitdiff_renamesrc(const char *line, struct patch *patch)
396 patch->is_rename = 1;
397 patch->old_name = find_name(line, NULL, 0, 0);
398 return 0;
401 static int gitdiff_renamedst(const char *line, struct patch *patch)
403 patch->is_rename = 1;
404 patch->new_name = find_name(line, NULL, 0, 0);
405 return 0;
408 static int gitdiff_similarity(const char *line, struct patch *patch)
410 if ((patch->score = strtoul(line, NULL, 10)) == ULONG_MAX)
411 patch->score = 0;
412 return 0;
415 static int gitdiff_dissimilarity(const char *line, struct patch *patch)
417 if ((patch->score = strtoul(line, NULL, 10)) == ULONG_MAX)
418 patch->score = 0;
419 return 0;
422 static int gitdiff_index(const char *line, struct patch *patch)
424 /* index line is N hexadecimal, "..", N hexadecimal,
425 * and optional space with octal mode.
427 const char *ptr, *eol;
428 int len;
430 ptr = strchr(line, '.');
431 if (!ptr || ptr[1] != '.' || 40 < ptr - line)
432 return 0;
433 len = ptr - line;
434 memcpy(patch->old_sha1_prefix, line, len);
435 patch->old_sha1_prefix[len] = 0;
437 line = ptr + 2;
438 ptr = strchr(line, ' ');
439 eol = strchr(line, '\n');
441 if (!ptr || eol < ptr)
442 ptr = eol;
443 len = ptr - line;
445 if (40 < len)
446 return 0;
447 memcpy(patch->new_sha1_prefix, line, len);
448 patch->new_sha1_prefix[len] = 0;
449 if (*ptr == ' ')
450 patch->new_mode = patch->old_mode = strtoul(ptr+1, NULL, 8);
451 return 0;
455 * This is normal for a diff that doesn't change anything: we'll fall through
456 * into the next diff. Tell the parser to break out.
458 static int gitdiff_unrecognized(const char *line, struct patch *patch)
460 return -1;
463 static const char *stop_at_slash(const char *line, int llen)
465 int i;
467 for (i = 0; i < llen; i++) {
468 int ch = line[i];
469 if (ch == '/')
470 return line + i;
472 return NULL;
475 /* This is to extract the same name that appears on "diff --git"
476 * line. We do not find and return anything if it is a rename
477 * patch, and it is OK because we will find the name elsewhere.
478 * We need to reliably find name only when it is mode-change only,
479 * creation or deletion of an empty file. In any of these cases,
480 * both sides are the same name under a/ and b/ respectively.
482 static char *git_header_name(char *line, int llen)
484 int len;
485 const char *name;
486 const char *second = NULL;
488 line += strlen("diff --git ");
489 llen -= strlen("diff --git ");
491 if (*line == '"') {
492 const char *cp;
493 char *first = unquote_c_style(line, &second);
494 if (!first)
495 return NULL;
497 /* advance to the first slash */
498 cp = stop_at_slash(first, strlen(first));
499 if (!cp || cp == first) {
500 /* we do not accept absolute paths */
501 free_first_and_fail:
502 free(first);
503 return NULL;
505 len = strlen(cp+1);
506 memmove(first, cp+1, len+1); /* including NUL */
508 /* second points at one past closing dq of name.
509 * find the second name.
511 while ((second < line + llen) && isspace(*second))
512 second++;
514 if (line + llen <= second)
515 goto free_first_and_fail;
516 if (*second == '"') {
517 char *sp = unquote_c_style(second, NULL);
518 if (!sp)
519 goto free_first_and_fail;
520 cp = stop_at_slash(sp, strlen(sp));
521 if (!cp || cp == sp) {
522 free_both_and_fail:
523 free(sp);
524 goto free_first_and_fail;
526 /* They must match, otherwise ignore */
527 if (strcmp(cp+1, first))
528 goto free_both_and_fail;
529 free(sp);
530 return first;
533 /* unquoted second */
534 cp = stop_at_slash(second, line + llen - second);
535 if (!cp || cp == second)
536 goto free_first_and_fail;
537 cp++;
538 if (line + llen - cp != len + 1 ||
539 memcmp(first, cp, len))
540 goto free_first_and_fail;
541 return first;
544 /* unquoted first name */
545 name = stop_at_slash(line, llen);
546 if (!name || name == line)
547 return NULL;
549 name++;
551 /* since the first name is unquoted, a dq if exists must be
552 * the beginning of the second name.
554 for (second = name; second < line + llen; second++) {
555 if (*second == '"') {
556 const char *cp = second;
557 const char *np;
558 char *sp = unquote_c_style(second, NULL);
560 if (!sp)
561 return NULL;
562 np = stop_at_slash(sp, strlen(sp));
563 if (!np || np == sp) {
564 free_second_and_fail:
565 free(sp);
566 return NULL;
568 np++;
569 len = strlen(np);
570 if (len < cp - name &&
571 !strncmp(np, name, len) &&
572 isspace(name[len])) {
573 /* Good */
574 memmove(sp, np, len + 1);
575 return sp;
577 goto free_second_and_fail;
582 * Accept a name only if it shows up twice, exactly the same
583 * form.
585 for (len = 0 ; ; len++) {
586 char c = name[len];
588 switch (c) {
589 default:
590 continue;
591 case '\n':
592 return NULL;
593 case '\t': case ' ':
594 second = name+len;
595 for (;;) {
596 char c = *second++;
597 if (c == '\n')
598 return NULL;
599 if (c == '/')
600 break;
602 if (second[len] == '\n' && !memcmp(name, second, len)) {
603 char *ret = xmalloc(len + 1);
604 memcpy(ret, name, len);
605 ret[len] = 0;
606 return ret;
610 return NULL;
613 /* Verify that we recognize the lines following a git header */
614 static int parse_git_header(char *line, int len, unsigned int size, struct patch *patch)
616 unsigned long offset;
618 /* A git diff has explicit new/delete information, so we don't guess */
619 patch->is_new = 0;
620 patch->is_delete = 0;
623 * Some things may not have the old name in the
624 * rest of the headers anywhere (pure mode changes,
625 * or removing or adding empty files), so we get
626 * the default name from the header.
628 patch->def_name = git_header_name(line, len);
630 line += len;
631 size -= len;
632 linenr++;
633 for (offset = len ; size > 0 ; offset += len, size -= len, line += len, linenr++) {
634 static const struct opentry {
635 const char *str;
636 int (*fn)(const char *, struct patch *);
637 } optable[] = {
638 { "@@ -", gitdiff_hdrend },
639 { "--- ", gitdiff_oldname },
640 { "+++ ", gitdiff_newname },
641 { "old mode ", gitdiff_oldmode },
642 { "new mode ", gitdiff_newmode },
643 { "deleted file mode ", gitdiff_delete },
644 { "new file mode ", gitdiff_newfile },
645 { "copy from ", gitdiff_copysrc },
646 { "copy to ", gitdiff_copydst },
647 { "rename old ", gitdiff_renamesrc },
648 { "rename new ", gitdiff_renamedst },
649 { "rename from ", gitdiff_renamesrc },
650 { "rename to ", gitdiff_renamedst },
651 { "similarity index ", gitdiff_similarity },
652 { "dissimilarity index ", gitdiff_dissimilarity },
653 { "index ", gitdiff_index },
654 { "", gitdiff_unrecognized },
656 int i;
658 len = linelen(line, size);
659 if (!len || line[len-1] != '\n')
660 break;
661 for (i = 0; i < ARRAY_SIZE(optable); i++) {
662 const struct opentry *p = optable + i;
663 int oplen = strlen(p->str);
664 if (len < oplen || memcmp(p->str, line, oplen))
665 continue;
666 if (p->fn(line + oplen, patch) < 0)
667 return offset;
668 break;
672 return offset;
675 static int parse_num(const char *line, unsigned long *p)
677 char *ptr;
679 if (!isdigit(*line))
680 return 0;
681 *p = strtoul(line, &ptr, 10);
682 return ptr - line;
685 static int parse_range(const char *line, int len, int offset, const char *expect,
686 unsigned long *p1, unsigned long *p2)
688 int digits, ex;
690 if (offset < 0 || offset >= len)
691 return -1;
692 line += offset;
693 len -= offset;
695 digits = parse_num(line, p1);
696 if (!digits)
697 return -1;
699 offset += digits;
700 line += digits;
701 len -= digits;
703 *p2 = 1;
704 if (*line == ',') {
705 digits = parse_num(line+1, p2);
706 if (!digits)
707 return -1;
709 offset += digits+1;
710 line += digits+1;
711 len -= digits+1;
714 ex = strlen(expect);
715 if (ex > len)
716 return -1;
717 if (memcmp(line, expect, ex))
718 return -1;
720 return offset + ex;
724 * Parse a unified diff fragment header of the
725 * form "@@ -a,b +c,d @@"
727 static int parse_fragment_header(char *line, int len, struct fragment *fragment)
729 int offset;
731 if (!len || line[len-1] != '\n')
732 return -1;
734 /* Figure out the number of lines in a fragment */
735 offset = parse_range(line, len, 4, " +", &fragment->oldpos, &fragment->oldlines);
736 offset = parse_range(line, len, offset, " @@", &fragment->newpos, &fragment->newlines);
738 return offset;
741 static int find_header(char *line, unsigned long size, int *hdrsize, struct patch *patch)
743 unsigned long offset, len;
745 patch->is_rename = patch->is_copy = 0;
746 patch->is_new = patch->is_delete = -1;
747 patch->old_mode = patch->new_mode = 0;
748 patch->old_name = patch->new_name = NULL;
749 for (offset = 0; size > 0; offset += len, size -= len, line += len, linenr++) {
750 unsigned long nextlen;
752 len = linelen(line, size);
753 if (!len)
754 break;
756 /* Testing this early allows us to take a few shortcuts.. */
757 if (len < 6)
758 continue;
761 * Make sure we don't find any unconnected patch fragmants.
762 * That's a sign that we didn't find a header, and that a
763 * patch has become corrupted/broken up.
765 if (!memcmp("@@ -", line, 4)) {
766 struct fragment dummy;
767 if (parse_fragment_header(line, len, &dummy) < 0)
768 continue;
769 error("patch fragment without header at line %d: %.*s", linenr, (int)len-1, line);
772 if (size < len + 6)
773 break;
776 * Git patch? It might not have a real patch, just a rename
777 * or mode change, so we handle that specially
779 if (!memcmp("diff --git ", line, 11)) {
780 int git_hdr_len = parse_git_header(line, len, size, patch);
781 if (git_hdr_len <= len)
782 continue;
783 if (!patch->old_name && !patch->new_name) {
784 if (!patch->def_name)
785 die("git diff header lacks filename information (line %d)", linenr);
786 patch->old_name = patch->new_name = patch->def_name;
788 *hdrsize = git_hdr_len;
789 return offset;
792 /** --- followed by +++ ? */
793 if (memcmp("--- ", line, 4) || memcmp("+++ ", line + len, 4))
794 continue;
797 * We only accept unified patches, so we want it to
798 * at least have "@@ -a,b +c,d @@\n", which is 14 chars
799 * minimum
801 nextlen = linelen(line + len, size - len);
802 if (size < nextlen + 14 || memcmp("@@ -", line + len + nextlen, 4))
803 continue;
805 /* Ok, we'll consider it a patch */
806 parse_traditional_patch(line, line+len, patch);
807 *hdrsize = len + nextlen;
808 linenr += 2;
809 return offset;
811 return -1;
815 * Parse a unified diff. Note that this really needs
816 * to parse each fragment separately, since the only
817 * way to know the difference between a "---" that is
818 * part of a patch, and a "---" that starts the next
819 * patch is to look at the line counts..
821 static int parse_fragment(char *line, unsigned long size, struct patch *patch, struct fragment *fragment)
823 int added, deleted;
824 int len = linelen(line, size), offset;
825 unsigned long oldlines, newlines;
826 unsigned long leading, trailing;
828 offset = parse_fragment_header(line, len, fragment);
829 if (offset < 0)
830 return -1;
831 oldlines = fragment->oldlines;
832 newlines = fragment->newlines;
833 leading = 0;
834 trailing = 0;
836 if (patch->is_new < 0) {
837 patch->is_new = !oldlines;
838 if (!oldlines)
839 patch->old_name = NULL;
841 if (patch->is_delete < 0) {
842 patch->is_delete = !newlines;
843 if (!newlines)
844 patch->new_name = NULL;
847 if (patch->is_new && oldlines)
848 return error("new file depends on old contents");
849 if (patch->is_delete != !newlines) {
850 if (newlines)
851 return error("deleted file still has contents");
852 fprintf(stderr, "** warning: file %s becomes empty but is not deleted\n", patch->new_name);
855 /* Parse the thing.. */
856 line += len;
857 size -= len;
858 linenr++;
859 added = deleted = 0;
860 for (offset = len; size > 0; offset += len, size -= len, line += len, linenr++) {
861 if (!oldlines && !newlines)
862 break;
863 len = linelen(line, size);
864 if (!len || line[len-1] != '\n')
865 return -1;
866 switch (*line) {
867 default:
868 return -1;
869 case ' ':
870 oldlines--;
871 newlines--;
872 if (!deleted && !added)
873 leading++;
874 trailing++;
875 break;
876 case '-':
877 deleted++;
878 oldlines--;
879 trailing = 0;
880 break;
881 case '+':
883 * We know len is at least two, since we have a '+' and
884 * we checked that the last character was a '\n' above.
885 * That is, an addition of an empty line would check
886 * the '+' here. Sneaky...
888 if ((new_whitespace != nowarn_whitespace) &&
889 isspace(line[len-2])) {
890 whitespace_error++;
891 if (squelch_whitespace_errors &&
892 squelch_whitespace_errors <
893 whitespace_error)
895 else {
896 fprintf(stderr, "Adds trailing whitespace.\n%s:%d:%.*s\n",
897 patch_input_file,
898 linenr, len-2, line+1);
901 added++;
902 newlines--;
903 trailing = 0;
904 break;
906 /* We allow "\ No newline at end of file". Depending
907 * on locale settings when the patch was produced we
908 * don't know what this line looks like. The only
909 * thing we do know is that it begins with "\ ".
910 * Checking for 12 is just for sanity check -- any
911 * l10n of "\ No newline..." is at least that long.
913 case '\\':
914 if (len < 12 || memcmp(line, "\\ ", 2))
915 return -1;
916 break;
919 if (oldlines || newlines)
920 return -1;
921 fragment->leading = leading;
922 fragment->trailing = trailing;
924 /* If a fragment ends with an incomplete line, we failed to include
925 * it in the above loop because we hit oldlines == newlines == 0
926 * before seeing it.
928 if (12 < size && !memcmp(line, "\\ ", 2))
929 offset += linelen(line, size);
931 patch->lines_added += added;
932 patch->lines_deleted += deleted;
933 return offset;
936 static int parse_single_patch(char *line, unsigned long size, struct patch *patch)
938 unsigned long offset = 0;
939 struct fragment **fragp = &patch->fragments;
941 while (size > 4 && !memcmp(line, "@@ -", 4)) {
942 struct fragment *fragment;
943 int len;
945 fragment = xcalloc(1, sizeof(*fragment));
946 len = parse_fragment(line, size, patch, fragment);
947 if (len <= 0)
948 die("corrupt patch at line %d", linenr);
950 fragment->patch = line;
951 fragment->size = len;
953 *fragp = fragment;
954 fragp = &fragment->next;
956 offset += len;
957 line += len;
958 size -= len;
960 return offset;
963 static inline int metadata_changes(struct patch *patch)
965 return patch->is_rename > 0 ||
966 patch->is_copy > 0 ||
967 patch->is_new > 0 ||
968 patch->is_delete ||
969 (patch->old_mode && patch->new_mode &&
970 patch->old_mode != patch->new_mode);
973 static int parse_binary(char *buffer, unsigned long size, struct patch *patch)
975 /* We have read "GIT binary patch\n"; what follows is a line
976 * that says the patch method (currently, either "deflated
977 * literal" or "deflated delta") and the length of data before
978 * deflating; a sequence of 'length-byte' followed by base-85
979 * encoded data follows.
981 * Each 5-byte sequence of base-85 encodes up to 4 bytes,
982 * and we would limit the patch line to 66 characters,
983 * so one line can fit up to 13 groups that would decode
984 * to 52 bytes max. The length byte 'A'-'Z' corresponds
985 * to 1-26 bytes, and 'a'-'z' corresponds to 27-52 bytes.
986 * The end of binary is signalled with an empty line.
988 int llen, used;
989 struct fragment *fragment;
990 char *data = NULL;
992 patch->fragments = fragment = xcalloc(1, sizeof(*fragment));
994 /* Grab the type of patch */
995 llen = linelen(buffer, size);
996 used = llen;
997 linenr++;
999 if (!strncmp(buffer, "delta ", 6)) {
1000 patch->is_binary = BINARY_DELTA_DEFLATED;
1001 patch->deflate_origlen = strtoul(buffer + 6, NULL, 10);
1003 else if (!strncmp(buffer, "literal ", 8)) {
1004 patch->is_binary = BINARY_LITERAL_DEFLATED;
1005 patch->deflate_origlen = strtoul(buffer + 8, NULL, 10);
1007 else
1008 return error("unrecognized binary patch at line %d: %.*s",
1009 linenr-1, llen-1, buffer);
1010 buffer += llen;
1011 while (1) {
1012 int byte_length, max_byte_length, newsize;
1013 llen = linelen(buffer, size);
1014 used += llen;
1015 linenr++;
1016 if (llen == 1)
1017 break;
1018 /* Minimum line is "A00000\n" which is 7-byte long,
1019 * and the line length must be multiple of 5 plus 2.
1021 if ((llen < 7) || (llen-2) % 5)
1022 goto corrupt;
1023 max_byte_length = (llen - 2) / 5 * 4;
1024 byte_length = *buffer;
1025 if ('A' <= byte_length && byte_length <= 'Z')
1026 byte_length = byte_length - 'A' + 1;
1027 else if ('a' <= byte_length && byte_length <= 'z')
1028 byte_length = byte_length - 'a' + 27;
1029 else
1030 goto corrupt;
1031 /* if the input length was not multiple of 4, we would
1032 * have filler at the end but the filler should never
1033 * exceed 3 bytes
1035 if (max_byte_length < byte_length ||
1036 byte_length <= max_byte_length - 4)
1037 goto corrupt;
1038 newsize = fragment->size + byte_length;
1039 data = xrealloc(data, newsize);
1040 if (decode_85(data + fragment->size,
1041 buffer + 1,
1042 byte_length))
1043 goto corrupt;
1044 fragment->size = newsize;
1045 buffer += llen;
1046 size -= llen;
1048 fragment->patch = data;
1049 return used;
1050 corrupt:
1051 return error("corrupt binary patch at line %d: %.*s",
1052 linenr-1, llen-1, buffer);
1055 static int parse_chunk(char *buffer, unsigned long size, struct patch *patch)
1057 int hdrsize, patchsize;
1058 int offset = find_header(buffer, size, &hdrsize, patch);
1060 if (offset < 0)
1061 return offset;
1063 patchsize = parse_single_patch(buffer + offset + hdrsize, size - offset - hdrsize, patch);
1065 if (!patchsize) {
1066 static const char *binhdr[] = {
1067 "Binary files ",
1068 "Files ",
1069 NULL,
1071 static const char git_binary[] = "GIT binary patch\n";
1072 int i;
1073 int hd = hdrsize + offset;
1074 unsigned long llen = linelen(buffer + hd, size - hd);
1076 if (llen == sizeof(git_binary) - 1 &&
1077 !memcmp(git_binary, buffer + hd, llen)) {
1078 int used;
1079 linenr++;
1080 used = parse_binary(buffer + hd + llen,
1081 size - hd - llen, patch);
1082 if (used)
1083 patchsize = used + llen;
1084 else
1085 patchsize = 0;
1087 else if (!memcmp(" differ\n", buffer + hd + llen - 8, 8)) {
1088 for (i = 0; binhdr[i]; i++) {
1089 int len = strlen(binhdr[i]);
1090 if (len < size - hd &&
1091 !memcmp(binhdr[i], buffer + hd, len)) {
1092 linenr++;
1093 patch->is_binary = 1;
1094 patchsize = llen;
1095 break;
1100 /* Empty patch cannot be applied if:
1101 * - it is a binary patch and we do not do binary_replace, or
1102 * - text patch without metadata change
1104 if ((apply || check) &&
1105 (patch->is_binary
1106 ? !allow_binary_replacement
1107 : !metadata_changes(patch)))
1108 die("patch with only garbage at line %d", linenr);
1111 return offset + hdrsize + patchsize;
1114 static const char pluses[] = "++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++";
1115 static const char minuses[]= "----------------------------------------------------------------------";
1117 static void show_stats(struct patch *patch)
1119 const char *prefix = "";
1120 char *name = patch->new_name;
1121 char *qname = NULL;
1122 int len, max, add, del, total;
1124 if (!name)
1125 name = patch->old_name;
1127 if (0 < (len = quote_c_style(name, NULL, NULL, 0))) {
1128 qname = xmalloc(len + 1);
1129 quote_c_style(name, qname, NULL, 0);
1130 name = qname;
1134 * "scale" the filename
1136 len = strlen(name);
1137 max = max_len;
1138 if (max > 50)
1139 max = 50;
1140 if (len > max) {
1141 char *slash;
1142 prefix = "...";
1143 max -= 3;
1144 name += len - max;
1145 slash = strchr(name, '/');
1146 if (slash)
1147 name = slash;
1149 len = max;
1152 * scale the add/delete
1154 max = max_change;
1155 if (max + len > 70)
1156 max = 70 - len;
1158 add = patch->lines_added;
1159 del = patch->lines_deleted;
1160 total = add + del;
1162 if (max_change > 0) {
1163 total = (total * max + max_change / 2) / max_change;
1164 add = (add * max + max_change / 2) / max_change;
1165 del = total - add;
1167 if (patch->is_binary)
1168 printf(" %s%-*s | Bin\n", prefix, len, name);
1169 else
1170 printf(" %s%-*s |%5d %.*s%.*s\n", prefix,
1171 len, name, patch->lines_added + patch->lines_deleted,
1172 add, pluses, del, minuses);
1173 if (qname)
1174 free(qname);
1177 static int read_old_data(struct stat *st, const char *path, void *buf, unsigned long size)
1179 int fd;
1180 unsigned long got;
1182 switch (st->st_mode & S_IFMT) {
1183 case S_IFLNK:
1184 return readlink(path, buf, size);
1185 case S_IFREG:
1186 fd = open(path, O_RDONLY);
1187 if (fd < 0)
1188 return error("unable to open %s", path);
1189 got = 0;
1190 for (;;) {
1191 int ret = xread(fd, buf + got, size - got);
1192 if (ret <= 0)
1193 break;
1194 got += ret;
1196 close(fd);
1197 return got;
1199 default:
1200 return -1;
1204 static int find_offset(const char *buf, unsigned long size, const char *fragment, unsigned long fragsize, int line, int *lines)
1206 int i;
1207 unsigned long start, backwards, forwards;
1209 if (fragsize > size)
1210 return -1;
1212 start = 0;
1213 if (line > 1) {
1214 unsigned long offset = 0;
1215 i = line-1;
1216 while (offset + fragsize <= size) {
1217 if (buf[offset++] == '\n') {
1218 start = offset;
1219 if (!--i)
1220 break;
1225 /* Exact line number? */
1226 if (!memcmp(buf + start, fragment, fragsize))
1227 return start;
1230 * There's probably some smart way to do this, but I'll leave
1231 * that to the smart and beautiful people. I'm simple and stupid.
1233 backwards = start;
1234 forwards = start;
1235 for (i = 0; ; i++) {
1236 unsigned long try;
1237 int n;
1239 /* "backward" */
1240 if (i & 1) {
1241 if (!backwards) {
1242 if (forwards + fragsize > size)
1243 break;
1244 continue;
1246 do {
1247 --backwards;
1248 } while (backwards && buf[backwards-1] != '\n');
1249 try = backwards;
1250 } else {
1251 while (forwards + fragsize <= size) {
1252 if (buf[forwards++] == '\n')
1253 break;
1255 try = forwards;
1258 if (try + fragsize > size)
1259 continue;
1260 if (memcmp(buf + try, fragment, fragsize))
1261 continue;
1262 n = (i >> 1)+1;
1263 if (i & 1)
1264 n = -n;
1265 *lines = n;
1266 return try;
1270 * We should start searching forward and backward.
1272 return -1;
1275 static void remove_first_line(const char **rbuf, int *rsize)
1277 const char *buf = *rbuf;
1278 int size = *rsize;
1279 unsigned long offset;
1280 offset = 0;
1281 while (offset <= size) {
1282 if (buf[offset++] == '\n')
1283 break;
1285 *rsize = size - offset;
1286 *rbuf = buf + offset;
1289 static void remove_last_line(const char **rbuf, int *rsize)
1291 const char *buf = *rbuf;
1292 int size = *rsize;
1293 unsigned long offset;
1294 offset = size - 1;
1295 while (offset > 0) {
1296 if (buf[--offset] == '\n')
1297 break;
1299 *rsize = offset + 1;
1302 struct buffer_desc {
1303 char *buffer;
1304 unsigned long size;
1305 unsigned long alloc;
1308 static int apply_line(char *output, const char *patch, int plen)
1310 /* plen is number of bytes to be copied from patch,
1311 * starting at patch+1 (patch[0] is '+'). Typically
1312 * patch[plen] is '\n'.
1314 int add_nl_to_tail = 0;
1315 if ((new_whitespace == strip_whitespace) &&
1316 1 < plen && isspace(patch[plen-1])) {
1317 if (patch[plen] == '\n')
1318 add_nl_to_tail = 1;
1319 plen--;
1320 while (0 < plen && isspace(patch[plen]))
1321 plen--;
1322 applied_after_stripping++;
1324 memcpy(output, patch + 1, plen);
1325 if (add_nl_to_tail)
1326 output[plen++] = '\n';
1327 return plen;
1330 static int apply_one_fragment(struct buffer_desc *desc, struct fragment *frag)
1332 char *buf = desc->buffer;
1333 const char *patch = frag->patch;
1334 int offset, size = frag->size;
1335 char *old = xmalloc(size);
1336 char *new = xmalloc(size);
1337 const char *oldlines, *newlines;
1338 int oldsize = 0, newsize = 0;
1339 unsigned long leading, trailing;
1340 int pos, lines;
1342 while (size > 0) {
1343 int len = linelen(patch, size);
1344 int plen;
1346 if (!len)
1347 break;
1350 * "plen" is how much of the line we should use for
1351 * the actual patch data. Normally we just remove the
1352 * first character on the line, but if the line is
1353 * followed by "\ No newline", then we also remove the
1354 * last one (which is the newline, of course).
1356 plen = len-1;
1357 if (len < size && patch[len] == '\\')
1358 plen--;
1359 switch (*patch) {
1360 case ' ':
1361 case '-':
1362 memcpy(old + oldsize, patch + 1, plen);
1363 oldsize += plen;
1364 if (*patch == '-')
1365 break;
1366 /* Fall-through for ' ' */
1367 case '+':
1368 if (*patch != '+' || !no_add)
1369 newsize += apply_line(new + newsize, patch,
1370 plen);
1371 break;
1372 case '@': case '\\':
1373 /* Ignore it, we already handled it */
1374 break;
1375 default:
1376 return -1;
1378 patch += len;
1379 size -= len;
1382 #ifdef NO_ACCURATE_DIFF
1383 if (oldsize > 0 && old[oldsize - 1] == '\n' &&
1384 newsize > 0 && new[newsize - 1] == '\n') {
1385 oldsize--;
1386 newsize--;
1388 #endif
1390 oldlines = old;
1391 newlines = new;
1392 leading = frag->leading;
1393 trailing = frag->trailing;
1394 lines = 0;
1395 pos = frag->newpos;
1396 for (;;) {
1397 offset = find_offset(buf, desc->size, oldlines, oldsize, pos, &lines);
1398 if (offset >= 0) {
1399 int diff = newsize - oldsize;
1400 unsigned long size = desc->size + diff;
1401 unsigned long alloc = desc->alloc;
1403 /* Warn if it was necessary to reduce the number
1404 * of context lines.
1406 if ((leading != frag->leading) || (trailing != frag->trailing))
1407 fprintf(stderr, "Context reduced to (%ld/%ld) to apply fragment at %d\n",
1408 leading, trailing, pos + lines);
1410 if (size > alloc) {
1411 alloc = size + 8192;
1412 desc->alloc = alloc;
1413 buf = xrealloc(buf, alloc);
1414 desc->buffer = buf;
1416 desc->size = size;
1417 memmove(buf + offset + newsize, buf + offset + oldsize, size - offset - newsize);
1418 memcpy(buf + offset, newlines, newsize);
1419 offset = 0;
1421 break;
1424 /* Am I at my context limits? */
1425 if ((leading <= p_context) && (trailing <= p_context))
1426 break;
1427 /* Reduce the number of context lines
1428 * Reduce both leading and trailing if they are equal
1429 * otherwise just reduce the larger context.
1431 if (leading >= trailing) {
1432 remove_first_line(&oldlines, &oldsize);
1433 remove_first_line(&newlines, &newsize);
1434 pos--;
1435 leading--;
1437 if (trailing > leading) {
1438 remove_last_line(&oldlines, &oldsize);
1439 remove_last_line(&newlines, &newsize);
1440 trailing--;
1444 free(old);
1445 free(new);
1446 return offset;
1449 static char *inflate_it(const void *data, unsigned long size,
1450 unsigned long inflated_size)
1452 z_stream stream;
1453 void *out;
1454 int st;
1456 memset(&stream, 0, sizeof(stream));
1458 stream.next_in = (unsigned char *)data;
1459 stream.avail_in = size;
1460 stream.next_out = out = xmalloc(inflated_size);
1461 stream.avail_out = inflated_size;
1462 inflateInit(&stream);
1463 st = inflate(&stream, Z_FINISH);
1464 if ((st != Z_STREAM_END) || stream.total_out != inflated_size) {
1465 free(out);
1466 return NULL;
1468 return out;
1471 static int apply_binary_fragment(struct buffer_desc *desc, struct patch *patch)
1473 unsigned long dst_size;
1474 struct fragment *fragment = patch->fragments;
1475 void *data;
1476 void *result;
1478 data = inflate_it(fragment->patch, fragment->size,
1479 patch->deflate_origlen);
1480 if (!data)
1481 return error("corrupt patch data");
1482 switch (patch->is_binary) {
1483 case BINARY_DELTA_DEFLATED:
1484 result = patch_delta(desc->buffer, desc->size,
1485 data,
1486 patch->deflate_origlen,
1487 &dst_size);
1488 free(desc->buffer);
1489 desc->buffer = result;
1490 free(data);
1491 break;
1492 case BINARY_LITERAL_DEFLATED:
1493 free(desc->buffer);
1494 desc->buffer = data;
1495 dst_size = patch->deflate_origlen;
1496 break;
1498 if (!desc->buffer)
1499 return -1;
1500 desc->size = desc->alloc = dst_size;
1501 return 0;
1504 static int apply_binary(struct buffer_desc *desc, struct patch *patch)
1506 const char *name = patch->old_name ? patch->old_name : patch->new_name;
1507 unsigned char sha1[20];
1508 unsigned char hdr[50];
1509 int hdrlen;
1511 if (!allow_binary_replacement)
1512 return error("cannot apply binary patch to '%s' "
1513 "without --allow-binary-replacement",
1514 name);
1516 /* For safety, we require patch index line to contain
1517 * full 40-byte textual SHA1 for old and new, at least for now.
1519 if (strlen(patch->old_sha1_prefix) != 40 ||
1520 strlen(patch->new_sha1_prefix) != 40 ||
1521 get_sha1_hex(patch->old_sha1_prefix, sha1) ||
1522 get_sha1_hex(patch->new_sha1_prefix, sha1))
1523 return error("cannot apply binary patch to '%s' "
1524 "without full index line", name);
1526 if (patch->old_name) {
1527 /* See if the old one matches what the patch
1528 * applies to.
1530 write_sha1_file_prepare(desc->buffer, desc->size,
1531 blob_type, sha1, hdr, &hdrlen);
1532 if (strcmp(sha1_to_hex(sha1), patch->old_sha1_prefix))
1533 return error("the patch applies to '%s' (%s), "
1534 "which does not match the "
1535 "current contents.",
1536 name, sha1_to_hex(sha1));
1538 else {
1539 /* Otherwise, the old one must be empty. */
1540 if (desc->size)
1541 return error("the patch applies to an empty "
1542 "'%s' but it is not empty", name);
1545 get_sha1_hex(patch->new_sha1_prefix, sha1);
1546 if (!memcmp(sha1, null_sha1, 20)) {
1547 free(desc->buffer);
1548 desc->alloc = desc->size = 0;
1549 desc->buffer = NULL;
1550 return 0; /* deletion patch */
1553 if (has_sha1_file(sha1)) {
1554 /* We already have the postimage */
1555 char type[10];
1556 unsigned long size;
1558 free(desc->buffer);
1559 desc->buffer = read_sha1_file(sha1, type, &size);
1560 if (!desc->buffer)
1561 return error("the necessary postimage %s for "
1562 "'%s' cannot be read",
1563 patch->new_sha1_prefix, name);
1564 desc->alloc = desc->size = size;
1566 else {
1567 /* We have verified desc matches the preimage;
1568 * apply the patch data to it, which is stored
1569 * in the patch->fragments->{patch,size}.
1571 if (apply_binary_fragment(desc, patch))
1572 return error("binary patch does not apply to '%s'",
1573 name);
1575 /* verify that the result matches */
1576 write_sha1_file_prepare(desc->buffer, desc->size, blob_type,
1577 sha1, hdr, &hdrlen);
1578 if (strcmp(sha1_to_hex(sha1), patch->new_sha1_prefix))
1579 return error("binary patch to '%s' creates incorrect result", name);
1582 return 0;
1585 static int apply_fragments(struct buffer_desc *desc, struct patch *patch)
1587 struct fragment *frag = patch->fragments;
1588 const char *name = patch->old_name ? patch->old_name : patch->new_name;
1590 if (patch->is_binary)
1591 return apply_binary(desc, patch);
1593 while (frag) {
1594 if (apply_one_fragment(desc, frag) < 0)
1595 return error("patch failed: %s:%ld",
1596 name, frag->oldpos);
1597 frag = frag->next;
1599 return 0;
1602 static int apply_data(struct patch *patch, struct stat *st)
1604 char *buf;
1605 unsigned long size, alloc;
1606 struct buffer_desc desc;
1608 size = 0;
1609 alloc = 0;
1610 buf = NULL;
1611 if (patch->old_name) {
1612 size = st->st_size;
1613 alloc = size + 8192;
1614 buf = xmalloc(alloc);
1615 if (read_old_data(st, patch->old_name, buf, alloc) != size)
1616 return error("read of %s failed", patch->old_name);
1619 desc.size = size;
1620 desc.alloc = alloc;
1621 desc.buffer = buf;
1622 if (apply_fragments(&desc, patch) < 0)
1623 return -1;
1624 patch->result = desc.buffer;
1625 patch->resultsize = desc.size;
1627 if (patch->is_delete && patch->resultsize)
1628 return error("removal patch leaves file contents");
1630 return 0;
1633 static int check_patch(struct patch *patch)
1635 struct stat st;
1636 const char *old_name = patch->old_name;
1637 const char *new_name = patch->new_name;
1638 const char *name = old_name ? old_name : new_name;
1640 if (old_name) {
1641 int changed;
1642 int stat_ret = lstat(old_name, &st);
1644 if (check_index) {
1645 int pos = cache_name_pos(old_name, strlen(old_name));
1646 if (pos < 0)
1647 return error("%s: does not exist in index",
1648 old_name);
1649 if (stat_ret < 0) {
1650 struct checkout costate;
1651 if (errno != ENOENT)
1652 return error("%s: %s", old_name,
1653 strerror(errno));
1654 /* checkout */
1655 costate.base_dir = "";
1656 costate.base_dir_len = 0;
1657 costate.force = 0;
1658 costate.quiet = 0;
1659 costate.not_new = 0;
1660 costate.refresh_cache = 1;
1661 if (checkout_entry(active_cache[pos],
1662 &costate,
1663 NULL) ||
1664 lstat(old_name, &st))
1665 return -1;
1668 changed = ce_match_stat(active_cache[pos], &st, 1);
1669 if (changed)
1670 return error("%s: does not match index",
1671 old_name);
1673 else if (stat_ret < 0)
1674 return error("%s: %s", old_name, strerror(errno));
1676 if (patch->is_new < 0)
1677 patch->is_new = 0;
1678 st.st_mode = ntohl(create_ce_mode(st.st_mode));
1679 if (!patch->old_mode)
1680 patch->old_mode = st.st_mode;
1681 if ((st.st_mode ^ patch->old_mode) & S_IFMT)
1682 return error("%s: wrong type", old_name);
1683 if (st.st_mode != patch->old_mode)
1684 fprintf(stderr, "warning: %s has type %o, expected %o\n",
1685 old_name, st.st_mode, patch->old_mode);
1688 if (new_name && (patch->is_new | patch->is_rename | patch->is_copy)) {
1689 if (check_index && cache_name_pos(new_name, strlen(new_name)) >= 0)
1690 return error("%s: already exists in index", new_name);
1691 if (!lstat(new_name, &st))
1692 return error("%s: already exists in working directory", new_name);
1693 if (errno != ENOENT)
1694 return error("%s: %s", new_name, strerror(errno));
1695 if (!patch->new_mode) {
1696 if (patch->is_new)
1697 patch->new_mode = S_IFREG | 0644;
1698 else
1699 patch->new_mode = patch->old_mode;
1703 if (new_name && old_name) {
1704 int same = !strcmp(old_name, new_name);
1705 if (!patch->new_mode)
1706 patch->new_mode = patch->old_mode;
1707 if ((patch->old_mode ^ patch->new_mode) & S_IFMT)
1708 return error("new mode (%o) of %s does not match old mode (%o)%s%s",
1709 patch->new_mode, new_name, patch->old_mode,
1710 same ? "" : " of ", same ? "" : old_name);
1713 if (apply_data(patch, &st) < 0)
1714 return error("%s: patch does not apply", name);
1715 return 0;
1718 static int check_patch_list(struct patch *patch)
1720 int error = 0;
1722 for (;patch ; patch = patch->next)
1723 error |= check_patch(patch);
1724 return error;
1727 static inline int is_null_sha1(const unsigned char *sha1)
1729 return !memcmp(sha1, null_sha1, 20);
1732 static void show_index_list(struct patch *list)
1734 struct patch *patch;
1736 /* Once we start supporting the reverse patch, it may be
1737 * worth showing the new sha1 prefix, but until then...
1739 for (patch = list; patch; patch = patch->next) {
1740 const unsigned char *sha1_ptr;
1741 unsigned char sha1[20];
1742 const char *name;
1744 name = patch->old_name ? patch->old_name : patch->new_name;
1745 if (patch->is_new)
1746 sha1_ptr = null_sha1;
1747 else if (get_sha1(patch->old_sha1_prefix, sha1))
1748 die("sha1 information is lacking or useless (%s).",
1749 name);
1750 else
1751 sha1_ptr = sha1;
1753 printf("%06o %s ",patch->old_mode, sha1_to_hex(sha1_ptr));
1754 if (line_termination && quote_c_style(name, NULL, NULL, 0))
1755 quote_c_style(name, NULL, stdout, 0);
1756 else
1757 fputs(name, stdout);
1758 putchar(line_termination);
1762 static void stat_patch_list(struct patch *patch)
1764 int files, adds, dels;
1766 for (files = adds = dels = 0 ; patch ; patch = patch->next) {
1767 files++;
1768 adds += patch->lines_added;
1769 dels += patch->lines_deleted;
1770 show_stats(patch);
1773 printf(" %d files changed, %d insertions(+), %d deletions(-)\n", files, adds, dels);
1776 static void numstat_patch_list(struct patch *patch)
1778 for ( ; patch; patch = patch->next) {
1779 const char *name;
1780 name = patch->old_name ? patch->old_name : patch->new_name;
1781 printf("%d\t%d\t", patch->lines_added, patch->lines_deleted);
1782 if (line_termination && quote_c_style(name, NULL, NULL, 0))
1783 quote_c_style(name, NULL, stdout, 0);
1784 else
1785 fputs(name, stdout);
1786 putchar('\n');
1790 static void show_file_mode_name(const char *newdelete, unsigned int mode, const char *name)
1792 if (mode)
1793 printf(" %s mode %06o %s\n", newdelete, mode, name);
1794 else
1795 printf(" %s %s\n", newdelete, name);
1798 static void show_mode_change(struct patch *p, int show_name)
1800 if (p->old_mode && p->new_mode && p->old_mode != p->new_mode) {
1801 if (show_name)
1802 printf(" mode change %06o => %06o %s\n",
1803 p->old_mode, p->new_mode, p->new_name);
1804 else
1805 printf(" mode change %06o => %06o\n",
1806 p->old_mode, p->new_mode);
1810 static void show_rename_copy(struct patch *p)
1812 const char *renamecopy = p->is_rename ? "rename" : "copy";
1813 const char *old, *new;
1815 /* Find common prefix */
1816 old = p->old_name;
1817 new = p->new_name;
1818 while (1) {
1819 const char *slash_old, *slash_new;
1820 slash_old = strchr(old, '/');
1821 slash_new = strchr(new, '/');
1822 if (!slash_old ||
1823 !slash_new ||
1824 slash_old - old != slash_new - new ||
1825 memcmp(old, new, slash_new - new))
1826 break;
1827 old = slash_old + 1;
1828 new = slash_new + 1;
1830 /* p->old_name thru old is the common prefix, and old and new
1831 * through the end of names are renames
1833 if (old != p->old_name)
1834 printf(" %s %.*s{%s => %s} (%d%%)\n", renamecopy,
1835 (int)(old - p->old_name), p->old_name,
1836 old, new, p->score);
1837 else
1838 printf(" %s %s => %s (%d%%)\n", renamecopy,
1839 p->old_name, p->new_name, p->score);
1840 show_mode_change(p, 0);
1843 static void summary_patch_list(struct patch *patch)
1845 struct patch *p;
1847 for (p = patch; p; p = p->next) {
1848 if (p->is_new)
1849 show_file_mode_name("create", p->new_mode, p->new_name);
1850 else if (p->is_delete)
1851 show_file_mode_name("delete", p->old_mode, p->old_name);
1852 else {
1853 if (p->is_rename || p->is_copy)
1854 show_rename_copy(p);
1855 else {
1856 if (p->score) {
1857 printf(" rewrite %s (%d%%)\n",
1858 p->new_name, p->score);
1859 show_mode_change(p, 0);
1861 else
1862 show_mode_change(p, 1);
1868 static void patch_stats(struct patch *patch)
1870 int lines = patch->lines_added + patch->lines_deleted;
1872 if (lines > max_change)
1873 max_change = lines;
1874 if (patch->old_name) {
1875 int len = quote_c_style(patch->old_name, NULL, NULL, 0);
1876 if (!len)
1877 len = strlen(patch->old_name);
1878 if (len > max_len)
1879 max_len = len;
1881 if (patch->new_name) {
1882 int len = quote_c_style(patch->new_name, NULL, NULL, 0);
1883 if (!len)
1884 len = strlen(patch->new_name);
1885 if (len > max_len)
1886 max_len = len;
1890 static void remove_file(struct patch *patch)
1892 if (write_index) {
1893 if (remove_file_from_cache(patch->old_name) < 0)
1894 die("unable to remove %s from index", patch->old_name);
1896 unlink(patch->old_name);
1899 static void add_index_file(const char *path, unsigned mode, void *buf, unsigned long size)
1901 struct stat st;
1902 struct cache_entry *ce;
1903 int namelen = strlen(path);
1904 unsigned ce_size = cache_entry_size(namelen);
1906 if (!write_index)
1907 return;
1909 ce = xcalloc(1, ce_size);
1910 memcpy(ce->name, path, namelen);
1911 ce->ce_mode = create_ce_mode(mode);
1912 ce->ce_flags = htons(namelen);
1913 if (lstat(path, &st) < 0)
1914 die("unable to stat newly created file %s", path);
1915 fill_stat_cache_info(ce, &st);
1916 if (write_sha1_file(buf, size, blob_type, ce->sha1) < 0)
1917 die("unable to create backing store for newly created file %s", path);
1918 if (add_cache_entry(ce, ADD_CACHE_OK_TO_ADD) < 0)
1919 die("unable to add cache entry for %s", path);
1922 static int try_create_file(const char *path, unsigned int mode, const char *buf, unsigned long size)
1924 int fd;
1926 if (S_ISLNK(mode))
1927 return symlink(buf, path);
1928 fd = open(path, O_CREAT | O_EXCL | O_WRONLY, (mode & 0100) ? 0777 : 0666);
1929 if (fd < 0)
1930 return -1;
1931 while (size) {
1932 int written = xwrite(fd, buf, size);
1933 if (written < 0)
1934 die("writing file %s: %s", path, strerror(errno));
1935 if (!written)
1936 die("out of space writing file %s", path);
1937 buf += written;
1938 size -= written;
1940 if (close(fd) < 0)
1941 die("closing file %s: %s", path, strerror(errno));
1942 return 0;
1946 * We optimistically assume that the directories exist,
1947 * which is true 99% of the time anyway. If they don't,
1948 * we create them and try again.
1950 static void create_one_file(char *path, unsigned mode, const char *buf, unsigned long size)
1952 if (!try_create_file(path, mode, buf, size))
1953 return;
1955 if (errno == ENOENT) {
1956 if (safe_create_leading_directories(path))
1957 return;
1958 if (!try_create_file(path, mode, buf, size))
1959 return;
1962 if (errno == EEXIST) {
1963 unsigned int nr = getpid();
1965 for (;;) {
1966 const char *newpath;
1967 newpath = mkpath("%s~%u", path, nr);
1968 if (!try_create_file(newpath, mode, buf, size)) {
1969 if (!rename(newpath, path))
1970 return;
1971 unlink(newpath);
1972 break;
1974 if (errno != EEXIST)
1975 break;
1976 ++nr;
1979 die("unable to write file %s mode %o", path, mode);
1982 static void create_file(struct patch *patch)
1984 char *path = patch->new_name;
1985 unsigned mode = patch->new_mode;
1986 unsigned long size = patch->resultsize;
1987 char *buf = patch->result;
1989 if (!mode)
1990 mode = S_IFREG | 0644;
1991 create_one_file(path, mode, buf, size);
1992 add_index_file(path, mode, buf, size);
1995 static void write_out_one_result(struct patch *patch)
1997 if (patch->is_delete > 0) {
1998 remove_file(patch);
1999 return;
2001 if (patch->is_new > 0 || patch->is_copy) {
2002 create_file(patch);
2003 return;
2006 * Rename or modification boils down to the same
2007 * thing: remove the old, write the new
2009 remove_file(patch);
2010 create_file(patch);
2013 static void write_out_results(struct patch *list, int skipped_patch)
2015 if (!list && !skipped_patch)
2016 die("No changes");
2018 while (list) {
2019 write_out_one_result(list);
2020 list = list->next;
2024 static struct cache_file cache_file;
2026 static struct excludes {
2027 struct excludes *next;
2028 const char *path;
2029 } *excludes;
2031 static int use_patch(struct patch *p)
2033 const char *pathname = p->new_name ? p->new_name : p->old_name;
2034 struct excludes *x = excludes;
2035 while (x) {
2036 if (fnmatch(x->path, pathname, 0) == 0)
2037 return 0;
2038 x = x->next;
2040 if (0 < prefix_length) {
2041 int pathlen = strlen(pathname);
2042 if (pathlen <= prefix_length ||
2043 memcmp(prefix, pathname, prefix_length))
2044 return 0;
2046 return 1;
2049 static int apply_patch(int fd, const char *filename)
2051 int newfd;
2052 unsigned long offset, size;
2053 char *buffer = read_patch_file(fd, &size);
2054 struct patch *list = NULL, **listp = &list;
2055 int skipped_patch = 0;
2057 patch_input_file = filename;
2058 if (!buffer)
2059 return -1;
2060 offset = 0;
2061 while (size > 0) {
2062 struct patch *patch;
2063 int nr;
2065 patch = xcalloc(1, sizeof(*patch));
2066 nr = parse_chunk(buffer + offset, size, patch);
2067 if (nr < 0)
2068 break;
2069 if (use_patch(patch)) {
2070 patch_stats(patch);
2071 *listp = patch;
2072 listp = &patch->next;
2073 } else {
2074 /* perhaps free it a bit better? */
2075 free(patch);
2076 skipped_patch++;
2078 offset += nr;
2079 size -= nr;
2082 newfd = -1;
2083 if (whitespace_error && (new_whitespace == error_on_whitespace))
2084 apply = 0;
2086 write_index = check_index && apply;
2087 if (write_index)
2088 newfd = hold_index_file_for_update(&cache_file, get_index_file());
2089 if (check_index) {
2090 if (read_cache() < 0)
2091 die("unable to read index file");
2094 if ((check || apply) && check_patch_list(list) < 0)
2095 exit(1);
2097 if (apply)
2098 write_out_results(list, skipped_patch);
2100 if (write_index) {
2101 if (write_cache(newfd, active_cache, active_nr) ||
2102 commit_index_file(&cache_file))
2103 die("Unable to write new cachefile");
2106 if (show_index_info)
2107 show_index_list(list);
2109 if (diffstat)
2110 stat_patch_list(list);
2112 if (numstat)
2113 numstat_patch_list(list);
2115 if (summary)
2116 summary_patch_list(list);
2118 free(buffer);
2119 return 0;
2122 static int git_apply_config(const char *var, const char *value)
2124 if (!strcmp(var, "apply.whitespace")) {
2125 apply_default_whitespace = strdup(value);
2126 return 0;
2128 return git_default_config(var, value);
2132 int main(int argc, char **argv)
2134 int i;
2135 int read_stdin = 1;
2136 const char *whitespace_option = NULL;
2138 for (i = 1; i < argc; i++) {
2139 const char *arg = argv[i];
2140 char *end;
2141 int fd;
2143 if (!strcmp(arg, "-")) {
2144 apply_patch(0, "<stdin>");
2145 read_stdin = 0;
2146 continue;
2148 if (!strncmp(arg, "--exclude=", 10)) {
2149 struct excludes *x = xmalloc(sizeof(*x));
2150 x->path = arg + 10;
2151 x->next = excludes;
2152 excludes = x;
2153 continue;
2155 if (!strncmp(arg, "-p", 2)) {
2156 p_value = atoi(arg + 2);
2157 continue;
2159 if (!strcmp(arg, "--no-add")) {
2160 no_add = 1;
2161 continue;
2163 if (!strcmp(arg, "--stat")) {
2164 apply = 0;
2165 diffstat = 1;
2166 continue;
2168 if (!strcmp(arg, "--allow-binary-replacement") ||
2169 !strcmp(arg, "--binary")) {
2170 allow_binary_replacement = 1;
2171 continue;
2173 if (!strcmp(arg, "--numstat")) {
2174 apply = 0;
2175 numstat = 1;
2176 continue;
2178 if (!strcmp(arg, "--summary")) {
2179 apply = 0;
2180 summary = 1;
2181 continue;
2183 if (!strcmp(arg, "--check")) {
2184 apply = 0;
2185 check = 1;
2186 continue;
2188 if (!strcmp(arg, "--index")) {
2189 check_index = 1;
2190 continue;
2192 if (!strcmp(arg, "--apply")) {
2193 apply = 1;
2194 continue;
2196 if (!strcmp(arg, "--index-info")) {
2197 apply = 0;
2198 show_index_info = 1;
2199 continue;
2201 if (!strcmp(arg, "-z")) {
2202 line_termination = 0;
2203 continue;
2205 if (!strncmp(arg, "-C", 2)) {
2206 p_context = strtoul(arg + 2, &end, 0);
2207 if (*end != '\0')
2208 die("unrecognized context count '%s'", arg + 2);
2209 continue;
2211 if (!strncmp(arg, "--whitespace=", 13)) {
2212 whitespace_option = arg + 13;
2213 parse_whitespace_option(arg + 13);
2214 continue;
2217 if (check_index && prefix_length < 0) {
2218 prefix = setup_git_directory();
2219 prefix_length = prefix ? strlen(prefix) : 0;
2220 git_config(git_apply_config);
2221 if (!whitespace_option && apply_default_whitespace)
2222 parse_whitespace_option(apply_default_whitespace);
2224 if (0 < prefix_length)
2225 arg = prefix_filename(prefix, prefix_length, arg);
2227 fd = open(arg, O_RDONLY);
2228 if (fd < 0)
2229 usage(apply_usage);
2230 read_stdin = 0;
2231 set_default_whitespace_mode(whitespace_option);
2232 apply_patch(fd, arg);
2233 close(fd);
2235 set_default_whitespace_mode(whitespace_option);
2236 if (read_stdin)
2237 apply_patch(0, "<stdin>");
2238 if (whitespace_error) {
2239 if (squelch_whitespace_errors &&
2240 squelch_whitespace_errors < whitespace_error) {
2241 int squelched =
2242 whitespace_error - squelch_whitespace_errors;
2243 fprintf(stderr, "warning: squelched %d whitespace error%s\n",
2244 squelched,
2245 squelched == 1 ? "" : "s");
2247 if (new_whitespace == error_on_whitespace)
2248 die("%d line%s add%s trailing whitespaces.",
2249 whitespace_error,
2250 whitespace_error == 1 ? "" : "s",
2251 whitespace_error == 1 ? "s" : "");
2252 if (applied_after_stripping)
2253 fprintf(stderr, "warning: %d line%s applied after"
2254 " stripping trailing whitespaces.\n",
2255 applied_after_stripping,
2256 applied_after_stripping == 1 ? "" : "s");
2257 else if (whitespace_error)
2258 fprintf(stderr, "warning: %d line%s add%s trailing"
2259 " whitespaces.\n",
2260 whitespace_error,
2261 whitespace_error == 1 ? "" : "s",
2262 whitespace_error == 1 ? "s" : "");
2264 return 0;