builtin-for-each-ref.c::copy_name() - do not overstep the buffer.
[git/dscho.git] / builtin-apply.c
blobdb337f1aca48d7f7ec8b5e3c4c9230621d346d81
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 "cache.h"
10 #include "cache-tree.h"
11 #include "quote.h"
12 #include "blob.h"
13 #include "delta.h"
14 #include "builtin.h"
17 * --check turns on checking that the working tree matches the
18 * files that are being modified, but doesn't apply the patch
19 * --stat does just a diffstat, and doesn't actually apply
20 * --numstat does numeric diffstat, and doesn't actually apply
21 * --index-info shows the old and new index info for paths if available.
22 * --index updates the cache as well.
23 * --cached updates only the cache without ever touching the working tree.
25 static const char *prefix;
26 static int prefix_length = -1;
27 static int newfd = -1;
29 static int unidiff_zero;
30 static int p_value = 1;
31 static int p_value_known;
32 static int check_index;
33 static int update_index;
34 static int cached;
35 static int diffstat;
36 static int numstat;
37 static int summary;
38 static int check;
39 static int apply = 1;
40 static int apply_in_reverse;
41 static int apply_with_reject;
42 static int apply_verbosely;
43 static int no_add;
44 static int show_index_info;
45 static int line_termination = '\n';
46 static unsigned long p_context = ULONG_MAX;
47 static const char apply_usage[] =
48 "git-apply [--stat] [--numstat] [--summary] [--check] [--index] [--cached] [--apply] [--no-add] [--index-info] [--allow-binary-replacement] [--reverse] [--reject] [--verbose] [-z] [-pNUM] [-CNUM] [--whitespace=<nowarn|warn|error|error-all|strip>] <patch>...";
50 static enum whitespace_eol {
51 nowarn_whitespace,
52 warn_on_whitespace,
53 error_on_whitespace,
54 strip_whitespace,
55 } new_whitespace = warn_on_whitespace;
56 static int whitespace_error;
57 static int squelch_whitespace_errors = 5;
58 static int applied_after_fixing_ws;
59 static const char *patch_input_file;
61 static void parse_whitespace_option(const char *option)
63 if (!option) {
64 new_whitespace = warn_on_whitespace;
65 return;
67 if (!strcmp(option, "warn")) {
68 new_whitespace = warn_on_whitespace;
69 return;
71 if (!strcmp(option, "nowarn")) {
72 new_whitespace = nowarn_whitespace;
73 return;
75 if (!strcmp(option, "error")) {
76 new_whitespace = error_on_whitespace;
77 return;
79 if (!strcmp(option, "error-all")) {
80 new_whitespace = error_on_whitespace;
81 squelch_whitespace_errors = 0;
82 return;
84 if (!strcmp(option, "strip")) {
85 new_whitespace = strip_whitespace;
86 return;
88 die("unrecognized whitespace option '%s'", option);
91 static void set_default_whitespace_mode(const char *whitespace_option)
93 if (!whitespace_option && !apply_default_whitespace) {
94 new_whitespace = (apply
95 ? warn_on_whitespace
96 : nowarn_whitespace);
101 * For "diff-stat" like behaviour, we keep track of the biggest change
102 * we've seen, and the longest filename. That allows us to do simple
103 * scaling.
105 static int max_change, max_len;
108 * Various "current state", notably line numbers and what
109 * file (and how) we're patching right now.. The "is_xxxx"
110 * things are flags, where -1 means "don't know yet".
112 static int linenr = 1;
115 * This represents one "hunk" from a patch, starting with
116 * "@@ -oldpos,oldlines +newpos,newlines @@" marker. The
117 * patch text is pointed at by patch, and its byte length
118 * is stored in size. leading and trailing are the number
119 * of context lines.
121 struct fragment {
122 unsigned long leading, trailing;
123 unsigned long oldpos, oldlines;
124 unsigned long newpos, newlines;
125 const char *patch;
126 int size;
127 int rejected;
128 struct fragment *next;
132 * When dealing with a binary patch, we reuse "leading" field
133 * to store the type of the binary hunk, either deflated "delta"
134 * or deflated "literal".
136 #define binary_patch_method leading
137 #define BINARY_DELTA_DEFLATED 1
138 #define BINARY_LITERAL_DEFLATED 2
140 struct patch {
141 char *new_name, *old_name, *def_name;
142 unsigned int old_mode, new_mode;
143 int is_new, is_delete; /* -1 = unknown, 0 = false, 1 = true */
144 int rejected;
145 unsigned long deflate_origlen;
146 int lines_added, lines_deleted;
147 int score;
148 unsigned int is_toplevel_relative:1;
149 unsigned int inaccurate_eof:1;
150 unsigned int is_binary:1;
151 unsigned int is_copy:1;
152 unsigned int is_rename:1;
153 struct fragment *fragments;
154 char *result;
155 unsigned long resultsize;
156 char old_sha1_prefix[41];
157 char new_sha1_prefix[41];
158 struct patch *next;
161 static void say_patch_name(FILE *output, const char *pre, struct patch *patch, const char *post)
163 fputs(pre, output);
164 if (patch->old_name && patch->new_name &&
165 strcmp(patch->old_name, patch->new_name)) {
166 write_name_quoted(NULL, 0, patch->old_name, 1, output);
167 fputs(" => ", output);
168 write_name_quoted(NULL, 0, patch->new_name, 1, output);
170 else {
171 const char *n = patch->new_name;
172 if (!n)
173 n = patch->old_name;
174 write_name_quoted(NULL, 0, n, 1, output);
176 fputs(post, output);
179 #define CHUNKSIZE (8192)
180 #define SLOP (16)
182 static void *read_patch_file(int fd, unsigned long *sizep)
184 struct strbuf buf;
186 strbuf_init(&buf, 0);
187 if (strbuf_read(&buf, fd, 0) < 0)
188 die("git-apply: read returned %s", strerror(errno));
189 *sizep = buf.len;
192 * Make sure that we have some slop in the buffer
193 * so that we can do speculative "memcmp" etc, and
194 * see to it that it is NUL-filled.
196 strbuf_grow(&buf, SLOP);
197 memset(buf.buf + buf.len, 0, SLOP);
198 return strbuf_detach(&buf);
201 static unsigned long linelen(const char *buffer, unsigned long size)
203 unsigned long len = 0;
204 while (size--) {
205 len++;
206 if (*buffer++ == '\n')
207 break;
209 return len;
212 static int is_dev_null(const char *str)
214 return !memcmp("/dev/null", str, 9) && isspace(str[9]);
217 #define TERM_SPACE 1
218 #define TERM_TAB 2
220 static int name_terminate(const char *name, int namelen, int c, int terminate)
222 if (c == ' ' && !(terminate & TERM_SPACE))
223 return 0;
224 if (c == '\t' && !(terminate & TERM_TAB))
225 return 0;
227 return 1;
230 static char *find_name(const char *line, char *def, int p_value, int terminate)
232 int len;
233 const char *start = line;
234 char *name;
236 if (*line == '"') {
237 /* Proposed "new-style" GNU patch/diff format; see
238 * http://marc.theaimsgroup.com/?l=git&m=112927316408690&w=2
240 name = unquote_c_style(line, NULL);
241 if (name) {
242 char *cp = name;
243 while (p_value) {
244 cp = strchr(cp, '/');
245 if (!cp)
246 break;
247 cp++;
248 p_value--;
250 if (cp) {
251 /* name can later be freed, so we need
252 * to memmove, not just return cp
254 memmove(name, cp, strlen(cp) + 1);
255 free(def);
256 return name;
258 else {
259 free(name);
260 name = NULL;
265 for (;;) {
266 char c = *line;
268 if (isspace(c)) {
269 if (c == '\n')
270 break;
271 if (name_terminate(start, line-start, c, terminate))
272 break;
274 line++;
275 if (c == '/' && !--p_value)
276 start = line;
278 if (!start)
279 return def;
280 len = line - start;
281 if (!len)
282 return def;
285 * Generally we prefer the shorter name, especially
286 * if the other one is just a variation of that with
287 * something else tacked on to the end (ie "file.orig"
288 * or "file~").
290 if (def) {
291 int deflen = strlen(def);
292 if (deflen < len && !strncmp(start, def, deflen))
293 return def;
294 free(def);
297 return xmemdupz(start, len);
300 static int count_slashes(const char *cp)
302 int cnt = 0;
303 char ch;
305 while ((ch = *cp++))
306 if (ch == '/')
307 cnt++;
308 return cnt;
312 * Given the string after "--- " or "+++ ", guess the appropriate
313 * p_value for the given patch.
315 static int guess_p_value(const char *nameline)
317 char *name, *cp;
318 int val = -1;
320 if (is_dev_null(nameline))
321 return -1;
322 name = find_name(nameline, NULL, 0, TERM_SPACE | TERM_TAB);
323 if (!name)
324 return -1;
325 cp = strchr(name, '/');
326 if (!cp)
327 val = 0;
328 else if (prefix) {
330 * Does it begin with "a/$our-prefix" and such? Then this is
331 * very likely to apply to our directory.
333 if (!strncmp(name, prefix, prefix_length))
334 val = count_slashes(prefix);
335 else {
336 cp++;
337 if (!strncmp(cp, prefix, prefix_length))
338 val = count_slashes(prefix) + 1;
341 free(name);
342 return val;
346 * Get the name etc info from the --/+++ lines of a traditional patch header
348 * FIXME! The end-of-filename heuristics are kind of screwy. For existing
349 * files, we can happily check the index for a match, but for creating a
350 * new file we should try to match whatever "patch" does. I have no idea.
352 static void parse_traditional_patch(const char *first, const char *second, struct patch *patch)
354 char *name;
356 first += 4; /* skip "--- " */
357 second += 4; /* skip "+++ " */
358 if (!p_value_known) {
359 int p, q;
360 p = guess_p_value(first);
361 q = guess_p_value(second);
362 if (p < 0) p = q;
363 if (0 <= p && p == q) {
364 p_value = p;
365 p_value_known = 1;
368 if (is_dev_null(first)) {
369 patch->is_new = 1;
370 patch->is_delete = 0;
371 name = find_name(second, NULL, p_value, TERM_SPACE | TERM_TAB);
372 patch->new_name = name;
373 } else if (is_dev_null(second)) {
374 patch->is_new = 0;
375 patch->is_delete = 1;
376 name = find_name(first, NULL, p_value, TERM_SPACE | TERM_TAB);
377 patch->old_name = name;
378 } else {
379 name = find_name(first, NULL, p_value, TERM_SPACE | TERM_TAB);
380 name = find_name(second, name, p_value, TERM_SPACE | TERM_TAB);
381 patch->old_name = patch->new_name = name;
383 if (!name)
384 die("unable to find filename in patch at line %d", linenr);
387 static int gitdiff_hdrend(const char *line, struct patch *patch)
389 return -1;
393 * We're anal about diff header consistency, to make
394 * sure that we don't end up having strange ambiguous
395 * patches floating around.
397 * As a result, gitdiff_{old|new}name() will check
398 * their names against any previous information, just
399 * to make sure..
401 static char *gitdiff_verify_name(const char *line, int isnull, char *orig_name, const char *oldnew)
403 if (!orig_name && !isnull)
404 return find_name(line, NULL, p_value, TERM_TAB);
406 if (orig_name) {
407 int len;
408 const char *name;
409 char *another;
410 name = orig_name;
411 len = strlen(name);
412 if (isnull)
413 die("git-apply: bad git-diff - expected /dev/null, got %s on line %d", name, linenr);
414 another = find_name(line, NULL, p_value, TERM_TAB);
415 if (!another || memcmp(another, name, len))
416 die("git-apply: bad git-diff - inconsistent %s filename on line %d", oldnew, linenr);
417 free(another);
418 return orig_name;
420 else {
421 /* expect "/dev/null" */
422 if (memcmp("/dev/null", line, 9) || line[9] != '\n')
423 die("git-apply: bad git-diff - expected /dev/null on line %d", linenr);
424 return NULL;
428 static int gitdiff_oldname(const char *line, struct patch *patch)
430 patch->old_name = gitdiff_verify_name(line, patch->is_new, patch->old_name, "old");
431 return 0;
434 static int gitdiff_newname(const char *line, struct patch *patch)
436 patch->new_name = gitdiff_verify_name(line, patch->is_delete, patch->new_name, "new");
437 return 0;
440 static int gitdiff_oldmode(const char *line, struct patch *patch)
442 patch->old_mode = strtoul(line, NULL, 8);
443 return 0;
446 static int gitdiff_newmode(const char *line, struct patch *patch)
448 patch->new_mode = strtoul(line, NULL, 8);
449 return 0;
452 static int gitdiff_delete(const char *line, struct patch *patch)
454 patch->is_delete = 1;
455 patch->old_name = patch->def_name;
456 return gitdiff_oldmode(line, patch);
459 static int gitdiff_newfile(const char *line, struct patch *patch)
461 patch->is_new = 1;
462 patch->new_name = patch->def_name;
463 return gitdiff_newmode(line, patch);
466 static int gitdiff_copysrc(const char *line, struct patch *patch)
468 patch->is_copy = 1;
469 patch->old_name = find_name(line, NULL, 0, 0);
470 return 0;
473 static int gitdiff_copydst(const char *line, struct patch *patch)
475 patch->is_copy = 1;
476 patch->new_name = find_name(line, NULL, 0, 0);
477 return 0;
480 static int gitdiff_renamesrc(const char *line, struct patch *patch)
482 patch->is_rename = 1;
483 patch->old_name = find_name(line, NULL, 0, 0);
484 return 0;
487 static int gitdiff_renamedst(const char *line, struct patch *patch)
489 patch->is_rename = 1;
490 patch->new_name = find_name(line, NULL, 0, 0);
491 return 0;
494 static int gitdiff_similarity(const char *line, struct patch *patch)
496 if ((patch->score = strtoul(line, NULL, 10)) == ULONG_MAX)
497 patch->score = 0;
498 return 0;
501 static int gitdiff_dissimilarity(const char *line, struct patch *patch)
503 if ((patch->score = strtoul(line, NULL, 10)) == ULONG_MAX)
504 patch->score = 0;
505 return 0;
508 static int gitdiff_index(const char *line, struct patch *patch)
510 /* index line is N hexadecimal, "..", N hexadecimal,
511 * and optional space with octal mode.
513 const char *ptr, *eol;
514 int len;
516 ptr = strchr(line, '.');
517 if (!ptr || ptr[1] != '.' || 40 < ptr - line)
518 return 0;
519 len = ptr - line;
520 memcpy(patch->old_sha1_prefix, line, len);
521 patch->old_sha1_prefix[len] = 0;
523 line = ptr + 2;
524 ptr = strchr(line, ' ');
525 eol = strchr(line, '\n');
527 if (!ptr || eol < ptr)
528 ptr = eol;
529 len = ptr - line;
531 if (40 < len)
532 return 0;
533 memcpy(patch->new_sha1_prefix, line, len);
534 patch->new_sha1_prefix[len] = 0;
535 if (*ptr == ' ')
536 patch->new_mode = patch->old_mode = strtoul(ptr+1, NULL, 8);
537 return 0;
541 * This is normal for a diff that doesn't change anything: we'll fall through
542 * into the next diff. Tell the parser to break out.
544 static int gitdiff_unrecognized(const char *line, struct patch *patch)
546 return -1;
549 static const char *stop_at_slash(const char *line, int llen)
551 int i;
553 for (i = 0; i < llen; i++) {
554 int ch = line[i];
555 if (ch == '/')
556 return line + i;
558 return NULL;
561 /* This is to extract the same name that appears on "diff --git"
562 * line. We do not find and return anything if it is a rename
563 * patch, and it is OK because we will find the name elsewhere.
564 * We need to reliably find name only when it is mode-change only,
565 * creation or deletion of an empty file. In any of these cases,
566 * both sides are the same name under a/ and b/ respectively.
568 static char *git_header_name(char *line, int llen)
570 int len;
571 const char *name;
572 const char *second = NULL;
574 line += strlen("diff --git ");
575 llen -= strlen("diff --git ");
577 if (*line == '"') {
578 const char *cp;
579 char *first = unquote_c_style(line, &second);
580 if (!first)
581 return NULL;
583 /* advance to the first slash */
584 cp = stop_at_slash(first, strlen(first));
585 if (!cp || cp == first) {
586 /* we do not accept absolute paths */
587 free_first_and_fail:
588 free(first);
589 return NULL;
591 len = strlen(cp+1);
592 memmove(first, cp+1, len+1); /* including NUL */
594 /* second points at one past closing dq of name.
595 * find the second name.
597 while ((second < line + llen) && isspace(*second))
598 second++;
600 if (line + llen <= second)
601 goto free_first_and_fail;
602 if (*second == '"') {
603 char *sp = unquote_c_style(second, NULL);
604 if (!sp)
605 goto free_first_and_fail;
606 cp = stop_at_slash(sp, strlen(sp));
607 if (!cp || cp == sp) {
608 free_both_and_fail:
609 free(sp);
610 goto free_first_and_fail;
612 /* They must match, otherwise ignore */
613 if (strcmp(cp+1, first))
614 goto free_both_and_fail;
615 free(sp);
616 return first;
619 /* unquoted second */
620 cp = stop_at_slash(second, line + llen - second);
621 if (!cp || cp == second)
622 goto free_first_and_fail;
623 cp++;
624 if (line + llen - cp != len + 1 ||
625 memcmp(first, cp, len))
626 goto free_first_and_fail;
627 return first;
630 /* unquoted first name */
631 name = stop_at_slash(line, llen);
632 if (!name || name == line)
633 return NULL;
635 name++;
637 /* since the first name is unquoted, a dq if exists must be
638 * the beginning of the second name.
640 for (second = name; second < line + llen; second++) {
641 if (*second == '"') {
642 const char *cp = second;
643 const char *np;
644 char *sp = unquote_c_style(second, NULL);
646 if (!sp)
647 return NULL;
648 np = stop_at_slash(sp, strlen(sp));
649 if (!np || np == sp) {
650 free_second_and_fail:
651 free(sp);
652 return NULL;
654 np++;
655 len = strlen(np);
656 if (len < cp - name &&
657 !strncmp(np, name, len) &&
658 isspace(name[len])) {
659 /* Good */
660 memmove(sp, np, len + 1);
661 return sp;
663 goto free_second_and_fail;
668 * Accept a name only if it shows up twice, exactly the same
669 * form.
671 for (len = 0 ; ; len++) {
672 switch (name[len]) {
673 default:
674 continue;
675 case '\n':
676 return NULL;
677 case '\t': case ' ':
678 second = name+len;
679 for (;;) {
680 char c = *second++;
681 if (c == '\n')
682 return NULL;
683 if (c == '/')
684 break;
686 if (second[len] == '\n' && !memcmp(name, second, len)) {
687 return xmemdupz(name, len);
691 return NULL;
694 /* Verify that we recognize the lines following a git header */
695 static int parse_git_header(char *line, int len, unsigned int size, struct patch *patch)
697 unsigned long offset;
699 /* A git diff has explicit new/delete information, so we don't guess */
700 patch->is_new = 0;
701 patch->is_delete = 0;
704 * Some things may not have the old name in the
705 * rest of the headers anywhere (pure mode changes,
706 * or removing or adding empty files), so we get
707 * the default name from the header.
709 patch->def_name = git_header_name(line, len);
711 line += len;
712 size -= len;
713 linenr++;
714 for (offset = len ; size > 0 ; offset += len, size -= len, line += len, linenr++) {
715 static const struct opentry {
716 const char *str;
717 int (*fn)(const char *, struct patch *);
718 } optable[] = {
719 { "@@ -", gitdiff_hdrend },
720 { "--- ", gitdiff_oldname },
721 { "+++ ", gitdiff_newname },
722 { "old mode ", gitdiff_oldmode },
723 { "new mode ", gitdiff_newmode },
724 { "deleted file mode ", gitdiff_delete },
725 { "new file mode ", gitdiff_newfile },
726 { "copy from ", gitdiff_copysrc },
727 { "copy to ", gitdiff_copydst },
728 { "rename old ", gitdiff_renamesrc },
729 { "rename new ", gitdiff_renamedst },
730 { "rename from ", gitdiff_renamesrc },
731 { "rename to ", gitdiff_renamedst },
732 { "similarity index ", gitdiff_similarity },
733 { "dissimilarity index ", gitdiff_dissimilarity },
734 { "index ", gitdiff_index },
735 { "", gitdiff_unrecognized },
737 int i;
739 len = linelen(line, size);
740 if (!len || line[len-1] != '\n')
741 break;
742 for (i = 0; i < ARRAY_SIZE(optable); i++) {
743 const struct opentry *p = optable + i;
744 int oplen = strlen(p->str);
745 if (len < oplen || memcmp(p->str, line, oplen))
746 continue;
747 if (p->fn(line + oplen, patch) < 0)
748 return offset;
749 break;
753 return offset;
756 static int parse_num(const char *line, unsigned long *p)
758 char *ptr;
760 if (!isdigit(*line))
761 return 0;
762 *p = strtoul(line, &ptr, 10);
763 return ptr - line;
766 static int parse_range(const char *line, int len, int offset, const char *expect,
767 unsigned long *p1, unsigned long *p2)
769 int digits, ex;
771 if (offset < 0 || offset >= len)
772 return -1;
773 line += offset;
774 len -= offset;
776 digits = parse_num(line, p1);
777 if (!digits)
778 return -1;
780 offset += digits;
781 line += digits;
782 len -= digits;
784 *p2 = 1;
785 if (*line == ',') {
786 digits = parse_num(line+1, p2);
787 if (!digits)
788 return -1;
790 offset += digits+1;
791 line += digits+1;
792 len -= digits+1;
795 ex = strlen(expect);
796 if (ex > len)
797 return -1;
798 if (memcmp(line, expect, ex))
799 return -1;
801 return offset + ex;
805 * Parse a unified diff fragment header of the
806 * form "@@ -a,b +c,d @@"
808 static int parse_fragment_header(char *line, int len, struct fragment *fragment)
810 int offset;
812 if (!len || line[len-1] != '\n')
813 return -1;
815 /* Figure out the number of lines in a fragment */
816 offset = parse_range(line, len, 4, " +", &fragment->oldpos, &fragment->oldlines);
817 offset = parse_range(line, len, offset, " @@", &fragment->newpos, &fragment->newlines);
819 return offset;
822 static int find_header(char *line, unsigned long size, int *hdrsize, struct patch *patch)
824 unsigned long offset, len;
826 patch->is_toplevel_relative = 0;
827 patch->is_rename = patch->is_copy = 0;
828 patch->is_new = patch->is_delete = -1;
829 patch->old_mode = patch->new_mode = 0;
830 patch->old_name = patch->new_name = NULL;
831 for (offset = 0; size > 0; offset += len, size -= len, line += len, linenr++) {
832 unsigned long nextlen;
834 len = linelen(line, size);
835 if (!len)
836 break;
838 /* Testing this early allows us to take a few shortcuts.. */
839 if (len < 6)
840 continue;
843 * Make sure we don't find any unconnected patch fragments.
844 * That's a sign that we didn't find a header, and that a
845 * patch has become corrupted/broken up.
847 if (!memcmp("@@ -", line, 4)) {
848 struct fragment dummy;
849 if (parse_fragment_header(line, len, &dummy) < 0)
850 continue;
851 die("patch fragment without header at line %d: %.*s",
852 linenr, (int)len-1, line);
855 if (size < len + 6)
856 break;
859 * Git patch? It might not have a real patch, just a rename
860 * or mode change, so we handle that specially
862 if (!memcmp("diff --git ", line, 11)) {
863 int git_hdr_len = parse_git_header(line, len, size, patch);
864 if (git_hdr_len <= len)
865 continue;
866 if (!patch->old_name && !patch->new_name) {
867 if (!patch->def_name)
868 die("git diff header lacks filename information (line %d)", linenr);
869 patch->old_name = patch->new_name = patch->def_name;
871 patch->is_toplevel_relative = 1;
872 *hdrsize = git_hdr_len;
873 return offset;
876 /** --- followed by +++ ? */
877 if (memcmp("--- ", line, 4) || memcmp("+++ ", line + len, 4))
878 continue;
881 * We only accept unified patches, so we want it to
882 * at least have "@@ -a,b +c,d @@\n", which is 14 chars
883 * minimum
885 nextlen = linelen(line + len, size - len);
886 if (size < nextlen + 14 || memcmp("@@ -", line + len + nextlen, 4))
887 continue;
889 /* Ok, we'll consider it a patch */
890 parse_traditional_patch(line, line+len, patch);
891 *hdrsize = len + nextlen;
892 linenr += 2;
893 return offset;
895 return -1;
898 static void check_whitespace(const char *line, int len)
900 const char *err = "Adds trailing whitespace";
901 int seen_space = 0;
902 int i;
905 * We know len is at least two, since we have a '+' and we
906 * checked that the last character was a '\n' before calling
907 * this function. That is, an addition of an empty line would
908 * check the '+' here. Sneaky...
910 if (isspace(line[len-2]))
911 goto error;
914 * Make sure that there is no space followed by a tab in
915 * indentation.
917 err = "Space in indent is followed by a tab";
918 for (i = 1; i < len; i++) {
919 if (line[i] == '\t') {
920 if (seen_space)
921 goto error;
923 else if (line[i] == ' ')
924 seen_space = 1;
925 else
926 break;
928 return;
930 error:
931 whitespace_error++;
932 if (squelch_whitespace_errors &&
933 squelch_whitespace_errors < whitespace_error)
935 else
936 fprintf(stderr, "%s.\n%s:%d:%.*s\n",
937 err, patch_input_file, linenr, len-2, line+1);
942 * Parse a unified diff. Note that this really needs to parse each
943 * fragment separately, since the only way to know the difference
944 * between a "---" that is part of a patch, and a "---" that starts
945 * the next patch is to look at the line counts..
947 static int parse_fragment(char *line, unsigned long size, struct patch *patch, struct fragment *fragment)
949 int added, deleted;
950 int len = linelen(line, size), offset;
951 unsigned long oldlines, newlines;
952 unsigned long leading, trailing;
954 offset = parse_fragment_header(line, len, fragment);
955 if (offset < 0)
956 return -1;
957 oldlines = fragment->oldlines;
958 newlines = fragment->newlines;
959 leading = 0;
960 trailing = 0;
962 /* Parse the thing.. */
963 line += len;
964 size -= len;
965 linenr++;
966 added = deleted = 0;
967 for (offset = len;
968 0 < size;
969 offset += len, size -= len, line += len, linenr++) {
970 if (!oldlines && !newlines)
971 break;
972 len = linelen(line, size);
973 if (!len || line[len-1] != '\n')
974 return -1;
975 switch (*line) {
976 default:
977 return -1;
978 case '\n': /* newer GNU diff, an empty context line */
979 case ' ':
980 oldlines--;
981 newlines--;
982 if (!deleted && !added)
983 leading++;
984 trailing++;
985 break;
986 case '-':
987 if (apply_in_reverse &&
988 new_whitespace != nowarn_whitespace)
989 check_whitespace(line, len);
990 deleted++;
991 oldlines--;
992 trailing = 0;
993 break;
994 case '+':
995 if (!apply_in_reverse &&
996 new_whitespace != nowarn_whitespace)
997 check_whitespace(line, len);
998 added++;
999 newlines--;
1000 trailing = 0;
1001 break;
1003 /* We allow "\ No newline at end of file". Depending
1004 * on locale settings when the patch was produced we
1005 * don't know what this line looks like. The only
1006 * thing we do know is that it begins with "\ ".
1007 * Checking for 12 is just for sanity check -- any
1008 * l10n of "\ No newline..." is at least that long.
1010 case '\\':
1011 if (len < 12 || memcmp(line, "\\ ", 2))
1012 return -1;
1013 break;
1016 if (oldlines || newlines)
1017 return -1;
1018 fragment->leading = leading;
1019 fragment->trailing = trailing;
1021 /* If a fragment ends with an incomplete line, we failed to include
1022 * it in the above loop because we hit oldlines == newlines == 0
1023 * before seeing it.
1025 if (12 < size && !memcmp(line, "\\ ", 2))
1026 offset += linelen(line, size);
1028 patch->lines_added += added;
1029 patch->lines_deleted += deleted;
1031 if (0 < patch->is_new && oldlines)
1032 return error("new file depends on old contents");
1033 if (0 < patch->is_delete && newlines)
1034 return error("deleted file still has contents");
1035 return offset;
1038 static int parse_single_patch(char *line, unsigned long size, struct patch *patch)
1040 unsigned long offset = 0;
1041 unsigned long oldlines = 0, newlines = 0, context = 0;
1042 struct fragment **fragp = &patch->fragments;
1044 while (size > 4 && !memcmp(line, "@@ -", 4)) {
1045 struct fragment *fragment;
1046 int len;
1048 fragment = xcalloc(1, sizeof(*fragment));
1049 len = parse_fragment(line, size, patch, fragment);
1050 if (len <= 0)
1051 die("corrupt patch at line %d", linenr);
1052 fragment->patch = line;
1053 fragment->size = len;
1054 oldlines += fragment->oldlines;
1055 newlines += fragment->newlines;
1056 context += fragment->leading + fragment->trailing;
1058 *fragp = fragment;
1059 fragp = &fragment->next;
1061 offset += len;
1062 line += len;
1063 size -= len;
1067 * If something was removed (i.e. we have old-lines) it cannot
1068 * be creation, and if something was added it cannot be
1069 * deletion. However, the reverse is not true; --unified=0
1070 * patches that only add are not necessarily creation even
1071 * though they do not have any old lines, and ones that only
1072 * delete are not necessarily deletion.
1074 * Unfortunately, a real creation/deletion patch do _not_ have
1075 * any context line by definition, so we cannot safely tell it
1076 * apart with --unified=0 insanity. At least if the patch has
1077 * more than one hunk it is not creation or deletion.
1079 if (patch->is_new < 0 &&
1080 (oldlines || (patch->fragments && patch->fragments->next)))
1081 patch->is_new = 0;
1082 if (patch->is_delete < 0 &&
1083 (newlines || (patch->fragments && patch->fragments->next)))
1084 patch->is_delete = 0;
1085 if (!unidiff_zero || context) {
1086 /* If the user says the patch is not generated with
1087 * --unified=0, or if we have seen context lines,
1088 * then not having oldlines means the patch is creation,
1089 * and not having newlines means the patch is deletion.
1091 if (patch->is_new < 0 && !oldlines) {
1092 patch->is_new = 1;
1093 patch->old_name = NULL;
1095 if (patch->is_delete < 0 && !newlines) {
1096 patch->is_delete = 1;
1097 patch->new_name = NULL;
1101 if (0 < patch->is_new && oldlines)
1102 die("new file %s depends on old contents", patch->new_name);
1103 if (0 < patch->is_delete && newlines)
1104 die("deleted file %s still has contents", patch->old_name);
1105 if (!patch->is_delete && !newlines && context)
1106 fprintf(stderr, "** warning: file %s becomes empty but "
1107 "is not deleted\n", patch->new_name);
1109 return offset;
1112 static inline int metadata_changes(struct patch *patch)
1114 return patch->is_rename > 0 ||
1115 patch->is_copy > 0 ||
1116 patch->is_new > 0 ||
1117 patch->is_delete ||
1118 (patch->old_mode && patch->new_mode &&
1119 patch->old_mode != patch->new_mode);
1122 static char *inflate_it(const void *data, unsigned long size,
1123 unsigned long inflated_size)
1125 z_stream stream;
1126 void *out;
1127 int st;
1129 memset(&stream, 0, sizeof(stream));
1131 stream.next_in = (unsigned char *)data;
1132 stream.avail_in = size;
1133 stream.next_out = out = xmalloc(inflated_size);
1134 stream.avail_out = inflated_size;
1135 inflateInit(&stream);
1136 st = inflate(&stream, Z_FINISH);
1137 if ((st != Z_STREAM_END) || stream.total_out != inflated_size) {
1138 free(out);
1139 return NULL;
1141 return out;
1144 static struct fragment *parse_binary_hunk(char **buf_p,
1145 unsigned long *sz_p,
1146 int *status_p,
1147 int *used_p)
1149 /* Expect a line that begins with binary patch method ("literal"
1150 * or "delta"), followed by the length of data before deflating.
1151 * a sequence of 'length-byte' followed by base-85 encoded data
1152 * should follow, terminated by a newline.
1154 * Each 5-byte sequence of base-85 encodes up to 4 bytes,
1155 * and we would limit the patch line to 66 characters,
1156 * so one line can fit up to 13 groups that would decode
1157 * to 52 bytes max. The length byte 'A'-'Z' corresponds
1158 * to 1-26 bytes, and 'a'-'z' corresponds to 27-52 bytes.
1160 int llen, used;
1161 unsigned long size = *sz_p;
1162 char *buffer = *buf_p;
1163 int patch_method;
1164 unsigned long origlen;
1165 char *data = NULL;
1166 int hunk_size = 0;
1167 struct fragment *frag;
1169 llen = linelen(buffer, size);
1170 used = llen;
1172 *status_p = 0;
1174 if (!prefixcmp(buffer, "delta ")) {
1175 patch_method = BINARY_DELTA_DEFLATED;
1176 origlen = strtoul(buffer + 6, NULL, 10);
1178 else if (!prefixcmp(buffer, "literal ")) {
1179 patch_method = BINARY_LITERAL_DEFLATED;
1180 origlen = strtoul(buffer + 8, NULL, 10);
1182 else
1183 return NULL;
1185 linenr++;
1186 buffer += llen;
1187 while (1) {
1188 int byte_length, max_byte_length, newsize;
1189 llen = linelen(buffer, size);
1190 used += llen;
1191 linenr++;
1192 if (llen == 1) {
1193 /* consume the blank line */
1194 buffer++;
1195 size--;
1196 break;
1198 /* Minimum line is "A00000\n" which is 7-byte long,
1199 * and the line length must be multiple of 5 plus 2.
1201 if ((llen < 7) || (llen-2) % 5)
1202 goto corrupt;
1203 max_byte_length = (llen - 2) / 5 * 4;
1204 byte_length = *buffer;
1205 if ('A' <= byte_length && byte_length <= 'Z')
1206 byte_length = byte_length - 'A' + 1;
1207 else if ('a' <= byte_length && byte_length <= 'z')
1208 byte_length = byte_length - 'a' + 27;
1209 else
1210 goto corrupt;
1211 /* if the input length was not multiple of 4, we would
1212 * have filler at the end but the filler should never
1213 * exceed 3 bytes
1215 if (max_byte_length < byte_length ||
1216 byte_length <= max_byte_length - 4)
1217 goto corrupt;
1218 newsize = hunk_size + byte_length;
1219 data = xrealloc(data, newsize);
1220 if (decode_85(data + hunk_size, buffer + 1, byte_length))
1221 goto corrupt;
1222 hunk_size = newsize;
1223 buffer += llen;
1224 size -= llen;
1227 frag = xcalloc(1, sizeof(*frag));
1228 frag->patch = inflate_it(data, hunk_size, origlen);
1229 if (!frag->patch)
1230 goto corrupt;
1231 free(data);
1232 frag->size = origlen;
1233 *buf_p = buffer;
1234 *sz_p = size;
1235 *used_p = used;
1236 frag->binary_patch_method = patch_method;
1237 return frag;
1239 corrupt:
1240 free(data);
1241 *status_p = -1;
1242 error("corrupt binary patch at line %d: %.*s",
1243 linenr-1, llen-1, buffer);
1244 return NULL;
1247 static int parse_binary(char *buffer, unsigned long size, struct patch *patch)
1249 /* We have read "GIT binary patch\n"; what follows is a line
1250 * that says the patch method (currently, either "literal" or
1251 * "delta") and the length of data before deflating; a
1252 * sequence of 'length-byte' followed by base-85 encoded data
1253 * follows.
1255 * When a binary patch is reversible, there is another binary
1256 * hunk in the same format, starting with patch method (either
1257 * "literal" or "delta") with the length of data, and a sequence
1258 * of length-byte + base-85 encoded data, terminated with another
1259 * empty line. This data, when applied to the postimage, produces
1260 * the preimage.
1262 struct fragment *forward;
1263 struct fragment *reverse;
1264 int status;
1265 int used, used_1;
1267 forward = parse_binary_hunk(&buffer, &size, &status, &used);
1268 if (!forward && !status)
1269 /* there has to be one hunk (forward hunk) */
1270 return error("unrecognized binary patch at line %d", linenr-1);
1271 if (status)
1272 /* otherwise we already gave an error message */
1273 return status;
1275 reverse = parse_binary_hunk(&buffer, &size, &status, &used_1);
1276 if (reverse)
1277 used += used_1;
1278 else if (status) {
1279 /* not having reverse hunk is not an error, but having
1280 * a corrupt reverse hunk is.
1282 free((void*) forward->patch);
1283 free(forward);
1284 return status;
1286 forward->next = reverse;
1287 patch->fragments = forward;
1288 patch->is_binary = 1;
1289 return used;
1292 static int parse_chunk(char *buffer, unsigned long size, struct patch *patch)
1294 int hdrsize, patchsize;
1295 int offset = find_header(buffer, size, &hdrsize, patch);
1297 if (offset < 0)
1298 return offset;
1300 patchsize = parse_single_patch(buffer + offset + hdrsize, size - offset - hdrsize, patch);
1302 if (!patchsize) {
1303 static const char *binhdr[] = {
1304 "Binary files ",
1305 "Files ",
1306 NULL,
1308 static const char git_binary[] = "GIT binary patch\n";
1309 int i;
1310 int hd = hdrsize + offset;
1311 unsigned long llen = linelen(buffer + hd, size - hd);
1313 if (llen == sizeof(git_binary) - 1 &&
1314 !memcmp(git_binary, buffer + hd, llen)) {
1315 int used;
1316 linenr++;
1317 used = parse_binary(buffer + hd + llen,
1318 size - hd - llen, patch);
1319 if (used)
1320 patchsize = used + llen;
1321 else
1322 patchsize = 0;
1324 else if (!memcmp(" differ\n", buffer + hd + llen - 8, 8)) {
1325 for (i = 0; binhdr[i]; i++) {
1326 int len = strlen(binhdr[i]);
1327 if (len < size - hd &&
1328 !memcmp(binhdr[i], buffer + hd, len)) {
1329 linenr++;
1330 patch->is_binary = 1;
1331 patchsize = llen;
1332 break;
1337 /* Empty patch cannot be applied if it is a text patch
1338 * without metadata change. A binary patch appears
1339 * empty to us here.
1341 if ((apply || check) &&
1342 (!patch->is_binary && !metadata_changes(patch)))
1343 die("patch with only garbage at line %d", linenr);
1346 return offset + hdrsize + patchsize;
1349 #define swap(a,b) myswap((a),(b),sizeof(a))
1351 #define myswap(a, b, size) do { \
1352 unsigned char mytmp[size]; \
1353 memcpy(mytmp, &a, size); \
1354 memcpy(&a, &b, size); \
1355 memcpy(&b, mytmp, size); \
1356 } while (0)
1358 static void reverse_patches(struct patch *p)
1360 for (; p; p = p->next) {
1361 struct fragment *frag = p->fragments;
1363 swap(p->new_name, p->old_name);
1364 swap(p->new_mode, p->old_mode);
1365 swap(p->is_new, p->is_delete);
1366 swap(p->lines_added, p->lines_deleted);
1367 swap(p->old_sha1_prefix, p->new_sha1_prefix);
1369 for (; frag; frag = frag->next) {
1370 swap(frag->newpos, frag->oldpos);
1371 swap(frag->newlines, frag->oldlines);
1376 static const char pluses[] = "++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++";
1377 static const char minuses[]= "----------------------------------------------------------------------";
1379 static void show_stats(struct patch *patch)
1381 const char *prefix = "";
1382 char *name = patch->new_name;
1383 char *qname = NULL;
1384 int len, max, add, del, total;
1386 if (!name)
1387 name = patch->old_name;
1389 if (0 < (len = quote_c_style(name, NULL, NULL, 0))) {
1390 qname = xmalloc(len + 1);
1391 quote_c_style(name, qname, NULL, 0);
1392 name = qname;
1396 * "scale" the filename
1398 len = strlen(name);
1399 max = max_len;
1400 if (max > 50)
1401 max = 50;
1402 if (len > max) {
1403 char *slash;
1404 prefix = "...";
1405 max -= 3;
1406 name += len - max;
1407 slash = strchr(name, '/');
1408 if (slash)
1409 name = slash;
1411 len = max;
1414 * scale the add/delete
1416 max = max_change;
1417 if (max + len > 70)
1418 max = 70 - len;
1420 add = patch->lines_added;
1421 del = patch->lines_deleted;
1422 total = add + del;
1424 if (max_change > 0) {
1425 total = (total * max + max_change / 2) / max_change;
1426 add = (add * max + max_change / 2) / max_change;
1427 del = total - add;
1429 if (patch->is_binary)
1430 printf(" %s%-*s | Bin\n", prefix, len, name);
1431 else
1432 printf(" %s%-*s |%5d %.*s%.*s\n", prefix,
1433 len, name, patch->lines_added + patch->lines_deleted,
1434 add, pluses, del, minuses);
1435 free(qname);
1438 static int read_old_data(struct stat *st, const char *path, struct strbuf *buf)
1440 int fd;
1442 switch (st->st_mode & S_IFMT) {
1443 case S_IFLNK:
1444 strbuf_grow(buf, st->st_size);
1445 if (readlink(path, buf->buf, st->st_size) != st->st_size)
1446 return -1;
1447 strbuf_setlen(buf, st->st_size);
1448 return 0;
1449 case S_IFREG:
1450 fd = open(path, O_RDONLY);
1451 if (fd < 0)
1452 return error("unable to open %s", path);
1453 if (strbuf_read(buf, fd, st->st_size) < 0) {
1454 close(fd);
1455 return -1;
1457 close(fd);
1458 convert_to_git(path, buf->buf, buf->len, buf);
1459 return 0;
1460 default:
1461 return -1;
1465 static int find_offset(const char *buf, unsigned long size, const char *fragment, unsigned long fragsize, int line, int *lines)
1467 int i;
1468 unsigned long start, backwards, forwards;
1470 if (fragsize > size)
1471 return -1;
1473 start = 0;
1474 if (line > 1) {
1475 unsigned long offset = 0;
1476 i = line-1;
1477 while (offset + fragsize <= size) {
1478 if (buf[offset++] == '\n') {
1479 start = offset;
1480 if (!--i)
1481 break;
1486 /* Exact line number? */
1487 if ((start + fragsize <= size) &&
1488 !memcmp(buf + start, fragment, fragsize))
1489 return start;
1492 * There's probably some smart way to do this, but I'll leave
1493 * that to the smart and beautiful people. I'm simple and stupid.
1495 backwards = start;
1496 forwards = start;
1497 for (i = 0; ; i++) {
1498 unsigned long try;
1499 int n;
1501 /* "backward" */
1502 if (i & 1) {
1503 if (!backwards) {
1504 if (forwards + fragsize > size)
1505 break;
1506 continue;
1508 do {
1509 --backwards;
1510 } while (backwards && buf[backwards-1] != '\n');
1511 try = backwards;
1512 } else {
1513 while (forwards + fragsize <= size) {
1514 if (buf[forwards++] == '\n')
1515 break;
1517 try = forwards;
1520 if (try + fragsize > size)
1521 continue;
1522 if (memcmp(buf + try, fragment, fragsize))
1523 continue;
1524 n = (i >> 1)+1;
1525 if (i & 1)
1526 n = -n;
1527 *lines = n;
1528 return try;
1532 * We should start searching forward and backward.
1534 return -1;
1537 static void remove_first_line(const char **rbuf, int *rsize)
1539 const char *buf = *rbuf;
1540 int size = *rsize;
1541 unsigned long offset;
1542 offset = 0;
1543 while (offset <= size) {
1544 if (buf[offset++] == '\n')
1545 break;
1547 *rsize = size - offset;
1548 *rbuf = buf + offset;
1551 static void remove_last_line(const char **rbuf, int *rsize)
1553 const char *buf = *rbuf;
1554 int size = *rsize;
1555 unsigned long offset;
1556 offset = size - 1;
1557 while (offset > 0) {
1558 if (buf[--offset] == '\n')
1559 break;
1561 *rsize = offset + 1;
1564 static int apply_line(char *output, const char *patch, int plen)
1566 /* plen is number of bytes to be copied from patch,
1567 * starting at patch+1 (patch[0] is '+'). Typically
1568 * patch[plen] is '\n', unless this is the incomplete
1569 * last line.
1571 int i;
1572 int add_nl_to_tail = 0;
1573 int fixed = 0;
1574 int last_tab_in_indent = -1;
1575 int last_space_in_indent = -1;
1576 int need_fix_leading_space = 0;
1577 char *buf;
1579 if ((new_whitespace != strip_whitespace) || !whitespace_error ||
1580 *patch != '+') {
1581 memcpy(output, patch + 1, plen);
1582 return plen;
1585 if (1 < plen && isspace(patch[plen-1])) {
1586 if (patch[plen] == '\n')
1587 add_nl_to_tail = 1;
1588 plen--;
1589 while (0 < plen && isspace(patch[plen]))
1590 plen--;
1591 fixed = 1;
1594 for (i = 1; i < plen; i++) {
1595 char ch = patch[i];
1596 if (ch == '\t') {
1597 last_tab_in_indent = i;
1598 if (0 <= last_space_in_indent)
1599 need_fix_leading_space = 1;
1601 else if (ch == ' ')
1602 last_space_in_indent = i;
1603 else
1604 break;
1607 buf = output;
1608 if (need_fix_leading_space) {
1609 int consecutive_spaces = 0;
1610 /* between patch[1..last_tab_in_indent] strip the
1611 * funny spaces, updating them to tab as needed.
1613 for (i = 1; i < last_tab_in_indent; i++, plen--) {
1614 char ch = patch[i];
1615 if (ch != ' ') {
1616 consecutive_spaces = 0;
1617 *output++ = ch;
1618 } else {
1619 consecutive_spaces++;
1620 if (consecutive_spaces == 8) {
1621 *output++ = '\t';
1622 consecutive_spaces = 0;
1626 fixed = 1;
1627 i = last_tab_in_indent;
1629 else
1630 i = 1;
1632 memcpy(output, patch + i, plen);
1633 if (add_nl_to_tail)
1634 output[plen++] = '\n';
1635 if (fixed)
1636 applied_after_fixing_ws++;
1637 return output + plen - buf;
1640 static int apply_one_fragment(struct strbuf *buf, struct fragment *frag, int inaccurate_eof)
1642 int match_beginning, match_end;
1643 const char *patch = frag->patch;
1644 int offset, size = frag->size;
1645 char *old = xmalloc(size);
1646 char *new = xmalloc(size);
1647 const char *oldlines, *newlines;
1648 int oldsize = 0, newsize = 0;
1649 int new_blank_lines_at_end = 0;
1650 unsigned long leading, trailing;
1651 int pos, lines;
1653 while (size > 0) {
1654 char first;
1655 int len = linelen(patch, size);
1656 int plen;
1657 int added_blank_line = 0;
1659 if (!len)
1660 break;
1663 * "plen" is how much of the line we should use for
1664 * the actual patch data. Normally we just remove the
1665 * first character on the line, but if the line is
1666 * followed by "\ No newline", then we also remove the
1667 * last one (which is the newline, of course).
1669 plen = len-1;
1670 if (len < size && patch[len] == '\\')
1671 plen--;
1672 first = *patch;
1673 if (apply_in_reverse) {
1674 if (first == '-')
1675 first = '+';
1676 else if (first == '+')
1677 first = '-';
1680 switch (first) {
1681 case '\n':
1682 /* Newer GNU diff, empty context line */
1683 if (plen < 0)
1684 /* ... followed by '\No newline'; nothing */
1685 break;
1686 old[oldsize++] = '\n';
1687 new[newsize++] = '\n';
1688 break;
1689 case ' ':
1690 case '-':
1691 memcpy(old + oldsize, patch + 1, plen);
1692 oldsize += plen;
1693 if (first == '-')
1694 break;
1695 /* Fall-through for ' ' */
1696 case '+':
1697 if (first != '+' || !no_add) {
1698 int added = apply_line(new + newsize, patch,
1699 plen);
1700 newsize += added;
1701 if (first == '+' &&
1702 added == 1 && new[newsize-1] == '\n')
1703 added_blank_line = 1;
1705 break;
1706 case '@': case '\\':
1707 /* Ignore it, we already handled it */
1708 break;
1709 default:
1710 if (apply_verbosely)
1711 error("invalid start of line: '%c'", first);
1712 return -1;
1714 if (added_blank_line)
1715 new_blank_lines_at_end++;
1716 else
1717 new_blank_lines_at_end = 0;
1718 patch += len;
1719 size -= len;
1722 if (inaccurate_eof && oldsize > 0 && old[oldsize - 1] == '\n' &&
1723 newsize > 0 && new[newsize - 1] == '\n') {
1724 oldsize--;
1725 newsize--;
1728 oldlines = old;
1729 newlines = new;
1730 leading = frag->leading;
1731 trailing = frag->trailing;
1734 * If we don't have any leading/trailing data in the patch,
1735 * we want it to match at the beginning/end of the file.
1737 * But that would break if the patch is generated with
1738 * --unified=0; sane people wouldn't do that to cause us
1739 * trouble, but we try to please not so sane ones as well.
1741 if (unidiff_zero) {
1742 match_beginning = (!leading && !frag->oldpos);
1743 match_end = 0;
1745 else {
1746 match_beginning = !leading && (frag->oldpos == 1);
1747 match_end = !trailing;
1750 lines = 0;
1751 pos = frag->newpos;
1752 for (;;) {
1753 offset = find_offset(buf->buf, buf->len,
1754 oldlines, oldsize, pos, &lines);
1755 if (match_end && offset + oldsize != buf->len)
1756 offset = -1;
1757 if (match_beginning && offset)
1758 offset = -1;
1759 if (offset >= 0) {
1760 if (new_whitespace == strip_whitespace &&
1761 (buf->len - oldsize - offset == 0)) /* end of file? */
1762 newsize -= new_blank_lines_at_end;
1764 /* Warn if it was necessary to reduce the number
1765 * of context lines.
1767 if ((leading != frag->leading) ||
1768 (trailing != frag->trailing))
1769 fprintf(stderr, "Context reduced to (%ld/%ld)"
1770 " to apply fragment at %d\n",
1771 leading, trailing, pos + lines);
1773 strbuf_splice(buf, offset, oldsize, newlines, newsize);
1774 offset = 0;
1775 break;
1778 /* Am I at my context limits? */
1779 if ((leading <= p_context) && (trailing <= p_context))
1780 break;
1781 if (match_beginning || match_end) {
1782 match_beginning = match_end = 0;
1783 continue;
1785 /* Reduce the number of context lines
1786 * Reduce both leading and trailing if they are equal
1787 * otherwise just reduce the larger context.
1789 if (leading >= trailing) {
1790 remove_first_line(&oldlines, &oldsize);
1791 remove_first_line(&newlines, &newsize);
1792 pos--;
1793 leading--;
1795 if (trailing > leading) {
1796 remove_last_line(&oldlines, &oldsize);
1797 remove_last_line(&newlines, &newsize);
1798 trailing--;
1802 if (offset && apply_verbosely)
1803 error("while searching for:\n%.*s", oldsize, oldlines);
1805 free(old);
1806 free(new);
1807 return offset;
1810 static int apply_binary_fragment(struct strbuf *buf, struct patch *patch)
1812 struct fragment *fragment = patch->fragments;
1813 unsigned long len;
1814 void *dst;
1816 /* Binary patch is irreversible without the optional second hunk */
1817 if (apply_in_reverse) {
1818 if (!fragment->next)
1819 return error("cannot reverse-apply a binary patch "
1820 "without the reverse hunk to '%s'",
1821 patch->new_name
1822 ? patch->new_name : patch->old_name);
1823 fragment = fragment->next;
1825 switch (fragment->binary_patch_method) {
1826 case BINARY_DELTA_DEFLATED:
1827 dst = patch_delta(buf->buf, buf->len, fragment->patch,
1828 fragment->size, &len);
1829 if (!dst)
1830 return -1;
1831 /* XXX patch_delta NUL-terminates */
1832 strbuf_attach(buf, dst, len, len + 1);
1833 return 0;
1834 case BINARY_LITERAL_DEFLATED:
1835 strbuf_reset(buf);
1836 strbuf_add(buf, fragment->patch, fragment->size);
1837 return 0;
1839 return -1;
1842 static int apply_binary(struct strbuf *buf, struct patch *patch)
1844 const char *name = patch->old_name ? patch->old_name : patch->new_name;
1845 unsigned char sha1[20];
1847 /* For safety, we require patch index line to contain
1848 * full 40-byte textual SHA1 for old and new, at least for now.
1850 if (strlen(patch->old_sha1_prefix) != 40 ||
1851 strlen(patch->new_sha1_prefix) != 40 ||
1852 get_sha1_hex(patch->old_sha1_prefix, sha1) ||
1853 get_sha1_hex(patch->new_sha1_prefix, sha1))
1854 return error("cannot apply binary patch to '%s' "
1855 "without full index line", name);
1857 if (patch->old_name) {
1858 /* See if the old one matches what the patch
1859 * applies to.
1861 hash_sha1_file(buf->buf, buf->len, blob_type, sha1);
1862 if (strcmp(sha1_to_hex(sha1), patch->old_sha1_prefix))
1863 return error("the patch applies to '%s' (%s), "
1864 "which does not match the "
1865 "current contents.",
1866 name, sha1_to_hex(sha1));
1868 else {
1869 /* Otherwise, the old one must be empty. */
1870 if (buf->len)
1871 return error("the patch applies to an empty "
1872 "'%s' but it is not empty", name);
1875 get_sha1_hex(patch->new_sha1_prefix, sha1);
1876 if (is_null_sha1(sha1)) {
1877 strbuf_release(buf);
1878 return 0; /* deletion patch */
1881 if (has_sha1_file(sha1)) {
1882 /* We already have the postimage */
1883 enum object_type type;
1884 unsigned long size;
1885 char *result;
1887 result = read_sha1_file(sha1, &type, &size);
1888 if (!result)
1889 return error("the necessary postimage %s for "
1890 "'%s' cannot be read",
1891 patch->new_sha1_prefix, name);
1892 /* XXX read_sha1_file NUL-terminates */
1893 strbuf_attach(buf, result, size, size + 1);
1894 } else {
1895 /* We have verified buf matches the preimage;
1896 * apply the patch data to it, which is stored
1897 * in the patch->fragments->{patch,size}.
1899 if (apply_binary_fragment(buf, patch))
1900 return error("binary patch does not apply to '%s'",
1901 name);
1903 /* verify that the result matches */
1904 hash_sha1_file(buf->buf, buf->len, blob_type, sha1);
1905 if (strcmp(sha1_to_hex(sha1), patch->new_sha1_prefix))
1906 return error("binary patch to '%s' creates incorrect result (expecting %s, got %s)",
1907 name, patch->new_sha1_prefix, sha1_to_hex(sha1));
1910 return 0;
1913 static int apply_fragments(struct strbuf *buf, struct patch *patch)
1915 struct fragment *frag = patch->fragments;
1916 const char *name = patch->old_name ? patch->old_name : patch->new_name;
1918 if (patch->is_binary)
1919 return apply_binary(buf, patch);
1921 while (frag) {
1922 if (apply_one_fragment(buf, frag, patch->inaccurate_eof)) {
1923 error("patch failed: %s:%ld", name, frag->oldpos);
1924 if (!apply_with_reject)
1925 return -1;
1926 frag->rejected = 1;
1928 frag = frag->next;
1930 return 0;
1933 static int read_file_or_gitlink(struct cache_entry *ce, struct strbuf *buf)
1935 if (!ce)
1936 return 0;
1938 if (S_ISGITLINK(ntohl(ce->ce_mode))) {
1939 strbuf_grow(buf, 100);
1940 strbuf_addf(buf, "Subproject commit %s\n", sha1_to_hex(ce->sha1));
1941 } else {
1942 enum object_type type;
1943 unsigned long sz;
1944 char *result;
1946 result = read_sha1_file(ce->sha1, &type, &sz);
1947 if (!result)
1948 return -1;
1949 /* XXX read_sha1_file NUL-terminates */
1950 strbuf_attach(buf, result, sz, sz + 1);
1952 return 0;
1955 static int apply_data(struct patch *patch, struct stat *st, struct cache_entry *ce)
1957 struct strbuf buf;
1959 strbuf_init(&buf, 0);
1960 if (cached) {
1961 if (read_file_or_gitlink(ce, &buf))
1962 return error("read of %s failed", patch->old_name);
1963 } else if (patch->old_name) {
1964 if (S_ISGITLINK(patch->old_mode)) {
1965 if (ce) {
1966 read_file_or_gitlink(ce, &buf);
1967 } else {
1969 * There is no way to apply subproject
1970 * patch without looking at the index.
1972 patch->fragments = NULL;
1974 } else {
1975 if (read_old_data(st, patch->old_name, &buf))
1976 return error("read of %s failed", patch->old_name);
1980 if (apply_fragments(&buf, patch) < 0)
1981 return -1; /* note with --reject this succeeds. */
1982 patch->result = buf.buf;
1983 patch->resultsize = buf.len;
1985 if (0 < patch->is_delete && patch->resultsize)
1986 return error("removal patch leaves file contents");
1988 return 0;
1991 static int check_to_create_blob(const char *new_name, int ok_if_exists)
1993 struct stat nst;
1994 if (!lstat(new_name, &nst)) {
1995 if (S_ISDIR(nst.st_mode) || ok_if_exists)
1996 return 0;
1998 * A leading component of new_name might be a symlink
1999 * that is going to be removed with this patch, but
2000 * still pointing at somewhere that has the path.
2001 * In such a case, path "new_name" does not exist as
2002 * far as git is concerned.
2004 if (has_symlink_leading_path(new_name, NULL))
2005 return 0;
2007 return error("%s: already exists in working directory", new_name);
2009 else if ((errno != ENOENT) && (errno != ENOTDIR))
2010 return error("%s: %s", new_name, strerror(errno));
2011 return 0;
2014 static int verify_index_match(struct cache_entry *ce, struct stat *st)
2016 if (S_ISGITLINK(ntohl(ce->ce_mode))) {
2017 if (!S_ISDIR(st->st_mode))
2018 return -1;
2019 return 0;
2021 return ce_match_stat(ce, st, 1);
2024 static int check_patch(struct patch *patch, struct patch *prev_patch)
2026 struct stat st;
2027 const char *old_name = patch->old_name;
2028 const char *new_name = patch->new_name;
2029 const char *name = old_name ? old_name : new_name;
2030 struct cache_entry *ce = NULL;
2031 int ok_if_exists;
2033 patch->rejected = 1; /* we will drop this after we succeed */
2036 * Make sure that we do not have local modifications from the
2037 * index when we are looking at the index. Also make sure
2038 * we have the preimage file to be patched in the work tree,
2039 * unless --cached, which tells git to apply only in the index.
2041 if (old_name) {
2042 int stat_ret = 0;
2043 unsigned st_mode = 0;
2045 if (!cached)
2046 stat_ret = lstat(old_name, &st);
2047 if (check_index) {
2048 int pos = cache_name_pos(old_name, strlen(old_name));
2049 if (pos < 0)
2050 return error("%s: does not exist in index",
2051 old_name);
2052 ce = active_cache[pos];
2053 if (stat_ret < 0) {
2054 struct checkout costate;
2055 if (errno != ENOENT)
2056 return error("%s: %s", old_name,
2057 strerror(errno));
2058 /* checkout */
2059 costate.base_dir = "";
2060 costate.base_dir_len = 0;
2061 costate.force = 0;
2062 costate.quiet = 0;
2063 costate.not_new = 0;
2064 costate.refresh_cache = 1;
2065 if (checkout_entry(ce,
2066 &costate,
2067 NULL) ||
2068 lstat(old_name, &st))
2069 return -1;
2071 if (!cached && verify_index_match(ce, &st))
2072 return error("%s: does not match index",
2073 old_name);
2074 if (cached)
2075 st_mode = ntohl(ce->ce_mode);
2076 } else if (stat_ret < 0)
2077 return error("%s: %s", old_name, strerror(errno));
2079 if (!cached)
2080 st_mode = ntohl(ce_mode_from_stat(ce, st.st_mode));
2082 if (patch->is_new < 0)
2083 patch->is_new = 0;
2084 if (!patch->old_mode)
2085 patch->old_mode = st_mode;
2086 if ((st_mode ^ patch->old_mode) & S_IFMT)
2087 return error("%s: wrong type", old_name);
2088 if (st_mode != patch->old_mode)
2089 fprintf(stderr, "warning: %s has type %o, expected %o\n",
2090 old_name, st_mode, patch->old_mode);
2093 if (new_name && prev_patch && 0 < prev_patch->is_delete &&
2094 !strcmp(prev_patch->old_name, new_name))
2095 /* A type-change diff is always split into a patch to
2096 * delete old, immediately followed by a patch to
2097 * create new (see diff.c::run_diff()); in such a case
2098 * it is Ok that the entry to be deleted by the
2099 * previous patch is still in the working tree and in
2100 * the index.
2102 ok_if_exists = 1;
2103 else
2104 ok_if_exists = 0;
2106 if (new_name &&
2107 ((0 < patch->is_new) | (0 < patch->is_rename) | patch->is_copy)) {
2108 if (check_index &&
2109 cache_name_pos(new_name, strlen(new_name)) >= 0 &&
2110 !ok_if_exists)
2111 return error("%s: already exists in index", new_name);
2112 if (!cached) {
2113 int err = check_to_create_blob(new_name, ok_if_exists);
2114 if (err)
2115 return err;
2117 if (!patch->new_mode) {
2118 if (0 < patch->is_new)
2119 patch->new_mode = S_IFREG | 0644;
2120 else
2121 patch->new_mode = patch->old_mode;
2125 if (new_name && old_name) {
2126 int same = !strcmp(old_name, new_name);
2127 if (!patch->new_mode)
2128 patch->new_mode = patch->old_mode;
2129 if ((patch->old_mode ^ patch->new_mode) & S_IFMT)
2130 return error("new mode (%o) of %s does not match old mode (%o)%s%s",
2131 patch->new_mode, new_name, patch->old_mode,
2132 same ? "" : " of ", same ? "" : old_name);
2135 if (apply_data(patch, &st, ce) < 0)
2136 return error("%s: patch does not apply", name);
2137 patch->rejected = 0;
2138 return 0;
2141 static int check_patch_list(struct patch *patch)
2143 struct patch *prev_patch = NULL;
2144 int err = 0;
2146 for (prev_patch = NULL; patch ; patch = patch->next) {
2147 if (apply_verbosely)
2148 say_patch_name(stderr,
2149 "Checking patch ", patch, "...\n");
2150 err |= check_patch(patch, prev_patch);
2151 prev_patch = patch;
2153 return err;
2156 /* This function tries to read the sha1 from the current index */
2157 static int get_current_sha1(const char *path, unsigned char *sha1)
2159 int pos;
2161 if (read_cache() < 0)
2162 return -1;
2163 pos = cache_name_pos(path, strlen(path));
2164 if (pos < 0)
2165 return -1;
2166 hashcpy(sha1, active_cache[pos]->sha1);
2167 return 0;
2170 static void show_index_list(struct patch *list)
2172 struct patch *patch;
2174 /* Once we start supporting the reverse patch, it may be
2175 * worth showing the new sha1 prefix, but until then...
2177 for (patch = list; patch; patch = patch->next) {
2178 const unsigned char *sha1_ptr;
2179 unsigned char sha1[20];
2180 const char *name;
2182 name = patch->old_name ? patch->old_name : patch->new_name;
2183 if (0 < patch->is_new)
2184 sha1_ptr = null_sha1;
2185 else if (get_sha1(patch->old_sha1_prefix, sha1))
2186 /* git diff has no index line for mode/type changes */
2187 if (!patch->lines_added && !patch->lines_deleted) {
2188 if (get_current_sha1(patch->new_name, sha1) ||
2189 get_current_sha1(patch->old_name, sha1))
2190 die("mode change for %s, which is not "
2191 "in current HEAD", name);
2192 sha1_ptr = sha1;
2193 } else
2194 die("sha1 information is lacking or useless "
2195 "(%s).", name);
2196 else
2197 sha1_ptr = sha1;
2199 printf("%06o %s ",patch->old_mode, sha1_to_hex(sha1_ptr));
2200 if (line_termination && quote_c_style(name, NULL, NULL, 0))
2201 quote_c_style(name, NULL, stdout, 0);
2202 else
2203 fputs(name, stdout);
2204 putchar(line_termination);
2208 static void stat_patch_list(struct patch *patch)
2210 int files, adds, dels;
2212 for (files = adds = dels = 0 ; patch ; patch = patch->next) {
2213 files++;
2214 adds += patch->lines_added;
2215 dels += patch->lines_deleted;
2216 show_stats(patch);
2219 printf(" %d files changed, %d insertions(+), %d deletions(-)\n", files, adds, dels);
2222 static void numstat_patch_list(struct patch *patch)
2224 for ( ; patch; patch = patch->next) {
2225 const char *name;
2226 name = patch->new_name ? patch->new_name : patch->old_name;
2227 if (patch->is_binary)
2228 printf("-\t-\t");
2229 else
2230 printf("%d\t%d\t",
2231 patch->lines_added, patch->lines_deleted);
2232 if (line_termination && quote_c_style(name, NULL, NULL, 0))
2233 quote_c_style(name, NULL, stdout, 0);
2234 else
2235 fputs(name, stdout);
2236 putchar(line_termination);
2240 static void show_file_mode_name(const char *newdelete, unsigned int mode, const char *name)
2242 if (mode)
2243 printf(" %s mode %06o %s\n", newdelete, mode, name);
2244 else
2245 printf(" %s %s\n", newdelete, name);
2248 static void show_mode_change(struct patch *p, int show_name)
2250 if (p->old_mode && p->new_mode && p->old_mode != p->new_mode) {
2251 if (show_name)
2252 printf(" mode change %06o => %06o %s\n",
2253 p->old_mode, p->new_mode, p->new_name);
2254 else
2255 printf(" mode change %06o => %06o\n",
2256 p->old_mode, p->new_mode);
2260 static void show_rename_copy(struct patch *p)
2262 const char *renamecopy = p->is_rename ? "rename" : "copy";
2263 const char *old, *new;
2265 /* Find common prefix */
2266 old = p->old_name;
2267 new = p->new_name;
2268 while (1) {
2269 const char *slash_old, *slash_new;
2270 slash_old = strchr(old, '/');
2271 slash_new = strchr(new, '/');
2272 if (!slash_old ||
2273 !slash_new ||
2274 slash_old - old != slash_new - new ||
2275 memcmp(old, new, slash_new - new))
2276 break;
2277 old = slash_old + 1;
2278 new = slash_new + 1;
2280 /* p->old_name thru old is the common prefix, and old and new
2281 * through the end of names are renames
2283 if (old != p->old_name)
2284 printf(" %s %.*s{%s => %s} (%d%%)\n", renamecopy,
2285 (int)(old - p->old_name), p->old_name,
2286 old, new, p->score);
2287 else
2288 printf(" %s %s => %s (%d%%)\n", renamecopy,
2289 p->old_name, p->new_name, p->score);
2290 show_mode_change(p, 0);
2293 static void summary_patch_list(struct patch *patch)
2295 struct patch *p;
2297 for (p = patch; p; p = p->next) {
2298 if (p->is_new)
2299 show_file_mode_name("create", p->new_mode, p->new_name);
2300 else if (p->is_delete)
2301 show_file_mode_name("delete", p->old_mode, p->old_name);
2302 else {
2303 if (p->is_rename || p->is_copy)
2304 show_rename_copy(p);
2305 else {
2306 if (p->score) {
2307 printf(" rewrite %s (%d%%)\n",
2308 p->new_name, p->score);
2309 show_mode_change(p, 0);
2311 else
2312 show_mode_change(p, 1);
2318 static void patch_stats(struct patch *patch)
2320 int lines = patch->lines_added + patch->lines_deleted;
2322 if (lines > max_change)
2323 max_change = lines;
2324 if (patch->old_name) {
2325 int len = quote_c_style(patch->old_name, NULL, NULL, 0);
2326 if (!len)
2327 len = strlen(patch->old_name);
2328 if (len > max_len)
2329 max_len = len;
2331 if (patch->new_name) {
2332 int len = quote_c_style(patch->new_name, NULL, NULL, 0);
2333 if (!len)
2334 len = strlen(patch->new_name);
2335 if (len > max_len)
2336 max_len = len;
2340 static void remove_file(struct patch *patch, int rmdir_empty)
2342 if (update_index) {
2343 if (remove_file_from_cache(patch->old_name) < 0)
2344 die("unable to remove %s from index", patch->old_name);
2346 if (!cached) {
2347 if (S_ISGITLINK(patch->old_mode)) {
2348 if (rmdir(patch->old_name))
2349 warning("unable to remove submodule %s",
2350 patch->old_name);
2351 } else if (!unlink(patch->old_name) && rmdir_empty) {
2352 char *name = xstrdup(patch->old_name);
2353 char *end = strrchr(name, '/');
2354 while (end) {
2355 *end = 0;
2356 if (rmdir(name))
2357 break;
2358 end = strrchr(name, '/');
2360 free(name);
2365 static void add_index_file(const char *path, unsigned mode, void *buf, unsigned long size)
2367 struct stat st;
2368 struct cache_entry *ce;
2369 int namelen = strlen(path);
2370 unsigned ce_size = cache_entry_size(namelen);
2372 if (!update_index)
2373 return;
2375 ce = xcalloc(1, ce_size);
2376 memcpy(ce->name, path, namelen);
2377 ce->ce_mode = create_ce_mode(mode);
2378 ce->ce_flags = htons(namelen);
2379 if (S_ISGITLINK(mode)) {
2380 const char *s = buf;
2382 if (get_sha1_hex(s + strlen("Subproject commit "), ce->sha1))
2383 die("corrupt patch for subproject %s", path);
2384 } else {
2385 if (!cached) {
2386 if (lstat(path, &st) < 0)
2387 die("unable to stat newly created file %s",
2388 path);
2389 fill_stat_cache_info(ce, &st);
2391 if (write_sha1_file(buf, size, blob_type, ce->sha1) < 0)
2392 die("unable to create backing store for newly created file %s", path);
2394 if (add_cache_entry(ce, ADD_CACHE_OK_TO_ADD) < 0)
2395 die("unable to add cache entry for %s", path);
2398 static int try_create_file(const char *path, unsigned int mode, const char *buf, unsigned long size)
2400 int fd;
2401 struct strbuf nbuf;
2403 if (S_ISGITLINK(mode)) {
2404 struct stat st;
2405 if (!lstat(path, &st) && S_ISDIR(st.st_mode))
2406 return 0;
2407 return mkdir(path, 0777);
2410 if (has_symlinks && S_ISLNK(mode))
2411 /* Although buf:size is counted string, it also is NUL
2412 * terminated.
2414 return symlink(buf, path);
2416 fd = open(path, O_CREAT | O_EXCL | O_WRONLY, (mode & 0100) ? 0777 : 0666);
2417 if (fd < 0)
2418 return -1;
2420 strbuf_init(&nbuf, 0);
2421 if (convert_to_working_tree(path, buf, size, &nbuf)) {
2422 size = nbuf.len;
2423 buf = nbuf.buf;
2425 write_or_die(fd, buf, size);
2426 strbuf_release(&nbuf);
2428 if (close(fd) < 0)
2429 die("closing file %s: %s", path, strerror(errno));
2430 return 0;
2434 * We optimistically assume that the directories exist,
2435 * which is true 99% of the time anyway. If they don't,
2436 * we create them and try again.
2438 static void create_one_file(char *path, unsigned mode, const char *buf, unsigned long size)
2440 if (cached)
2441 return;
2442 if (!try_create_file(path, mode, buf, size))
2443 return;
2445 if (errno == ENOENT) {
2446 if (safe_create_leading_directories(path))
2447 return;
2448 if (!try_create_file(path, mode, buf, size))
2449 return;
2452 if (errno == EEXIST || errno == EACCES) {
2453 /* We may be trying to create a file where a directory
2454 * used to be.
2456 struct stat st;
2457 if (!lstat(path, &st) && (!S_ISDIR(st.st_mode) || !rmdir(path)))
2458 errno = EEXIST;
2461 if (errno == EEXIST) {
2462 unsigned int nr = getpid();
2464 for (;;) {
2465 const char *newpath;
2466 newpath = mkpath("%s~%u", path, nr);
2467 if (!try_create_file(newpath, mode, buf, size)) {
2468 if (!rename(newpath, path))
2469 return;
2470 unlink(newpath);
2471 break;
2473 if (errno != EEXIST)
2474 break;
2475 ++nr;
2478 die("unable to write file %s mode %o", path, mode);
2481 static void create_file(struct patch *patch)
2483 char *path = patch->new_name;
2484 unsigned mode = patch->new_mode;
2485 unsigned long size = patch->resultsize;
2486 char *buf = patch->result;
2488 if (!mode)
2489 mode = S_IFREG | 0644;
2490 create_one_file(path, mode, buf, size);
2491 add_index_file(path, mode, buf, size);
2494 /* phase zero is to remove, phase one is to create */
2495 static void write_out_one_result(struct patch *patch, int phase)
2497 if (patch->is_delete > 0) {
2498 if (phase == 0)
2499 remove_file(patch, 1);
2500 return;
2502 if (patch->is_new > 0 || patch->is_copy) {
2503 if (phase == 1)
2504 create_file(patch);
2505 return;
2508 * Rename or modification boils down to the same
2509 * thing: remove the old, write the new
2511 if (phase == 0)
2512 remove_file(patch, patch->is_rename);
2513 if (phase == 1)
2514 create_file(patch);
2517 static int write_out_one_reject(struct patch *patch)
2519 FILE *rej;
2520 char namebuf[PATH_MAX];
2521 struct fragment *frag;
2522 int cnt = 0;
2524 for (cnt = 0, frag = patch->fragments; frag; frag = frag->next) {
2525 if (!frag->rejected)
2526 continue;
2527 cnt++;
2530 if (!cnt) {
2531 if (apply_verbosely)
2532 say_patch_name(stderr,
2533 "Applied patch ", patch, " cleanly.\n");
2534 return 0;
2537 /* This should not happen, because a removal patch that leaves
2538 * contents are marked "rejected" at the patch level.
2540 if (!patch->new_name)
2541 die("internal error");
2543 /* Say this even without --verbose */
2544 say_patch_name(stderr, "Applying patch ", patch, " with");
2545 fprintf(stderr, " %d rejects...\n", cnt);
2547 cnt = strlen(patch->new_name);
2548 if (ARRAY_SIZE(namebuf) <= cnt + 5) {
2549 cnt = ARRAY_SIZE(namebuf) - 5;
2550 fprintf(stderr,
2551 "warning: truncating .rej filename to %.*s.rej",
2552 cnt - 1, patch->new_name);
2554 memcpy(namebuf, patch->new_name, cnt);
2555 memcpy(namebuf + cnt, ".rej", 5);
2557 rej = fopen(namebuf, "w");
2558 if (!rej)
2559 return error("cannot open %s: %s", namebuf, strerror(errno));
2561 /* Normal git tools never deal with .rej, so do not pretend
2562 * this is a git patch by saying --git nor give extended
2563 * headers. While at it, maybe please "kompare" that wants
2564 * the trailing TAB and some garbage at the end of line ;-).
2566 fprintf(rej, "diff a/%s b/%s\t(rejected hunks)\n",
2567 patch->new_name, patch->new_name);
2568 for (cnt = 1, frag = patch->fragments;
2569 frag;
2570 cnt++, frag = frag->next) {
2571 if (!frag->rejected) {
2572 fprintf(stderr, "Hunk #%d applied cleanly.\n", cnt);
2573 continue;
2575 fprintf(stderr, "Rejected hunk #%d.\n", cnt);
2576 fprintf(rej, "%.*s", frag->size, frag->patch);
2577 if (frag->patch[frag->size-1] != '\n')
2578 fputc('\n', rej);
2580 fclose(rej);
2581 return -1;
2584 static int write_out_results(struct patch *list, int skipped_patch)
2586 int phase;
2587 int errs = 0;
2588 struct patch *l;
2590 if (!list && !skipped_patch)
2591 return error("No changes");
2593 for (phase = 0; phase < 2; phase++) {
2594 l = list;
2595 while (l) {
2596 if (l->rejected)
2597 errs = 1;
2598 else {
2599 write_out_one_result(l, phase);
2600 if (phase == 1 && write_out_one_reject(l))
2601 errs = 1;
2603 l = l->next;
2606 return errs;
2609 static struct lock_file lock_file;
2611 static struct excludes {
2612 struct excludes *next;
2613 const char *path;
2614 } *excludes;
2616 static int use_patch(struct patch *p)
2618 const char *pathname = p->new_name ? p->new_name : p->old_name;
2619 struct excludes *x = excludes;
2620 while (x) {
2621 if (fnmatch(x->path, pathname, 0) == 0)
2622 return 0;
2623 x = x->next;
2625 if (0 < prefix_length) {
2626 int pathlen = strlen(pathname);
2627 if (pathlen <= prefix_length ||
2628 memcmp(prefix, pathname, prefix_length))
2629 return 0;
2631 return 1;
2634 static void prefix_one(char **name)
2636 char *old_name = *name;
2637 if (!old_name)
2638 return;
2639 *name = xstrdup(prefix_filename(prefix, prefix_length, *name));
2640 free(old_name);
2643 static void prefix_patches(struct patch *p)
2645 if (!prefix || p->is_toplevel_relative)
2646 return;
2647 for ( ; p; p = p->next) {
2648 if (p->new_name == p->old_name) {
2649 char *prefixed = p->new_name;
2650 prefix_one(&prefixed);
2651 p->new_name = p->old_name = prefixed;
2653 else {
2654 prefix_one(&p->new_name);
2655 prefix_one(&p->old_name);
2660 static int apply_patch(int fd, const char *filename, int inaccurate_eof)
2662 unsigned long offset, size;
2663 char *buffer = read_patch_file(fd, &size);
2664 struct patch *list = NULL, **listp = &list;
2665 int skipped_patch = 0;
2667 patch_input_file = filename;
2668 if (!buffer)
2669 return -1;
2670 offset = 0;
2671 while (size > 0) {
2672 struct patch *patch;
2673 int nr;
2675 patch = xcalloc(1, sizeof(*patch));
2676 patch->inaccurate_eof = inaccurate_eof;
2677 nr = parse_chunk(buffer + offset, size, patch);
2678 if (nr < 0)
2679 break;
2680 if (apply_in_reverse)
2681 reverse_patches(patch);
2682 if (prefix)
2683 prefix_patches(patch);
2684 if (use_patch(patch)) {
2685 patch_stats(patch);
2686 *listp = patch;
2687 listp = &patch->next;
2689 else {
2690 /* perhaps free it a bit better? */
2691 free(patch);
2692 skipped_patch++;
2694 offset += nr;
2695 size -= nr;
2698 if (whitespace_error && (new_whitespace == error_on_whitespace))
2699 apply = 0;
2701 update_index = check_index && apply;
2702 if (update_index && newfd < 0)
2703 newfd = hold_locked_index(&lock_file, 1);
2705 if (check_index) {
2706 if (read_cache() < 0)
2707 die("unable to read index file");
2710 if ((check || apply) &&
2711 check_patch_list(list) < 0 &&
2712 !apply_with_reject)
2713 exit(1);
2715 if (apply && write_out_results(list, skipped_patch))
2716 exit(1);
2718 if (show_index_info)
2719 show_index_list(list);
2721 if (diffstat)
2722 stat_patch_list(list);
2724 if (numstat)
2725 numstat_patch_list(list);
2727 if (summary)
2728 summary_patch_list(list);
2730 free(buffer);
2731 return 0;
2734 static int git_apply_config(const char *var, const char *value)
2736 if (!strcmp(var, "apply.whitespace")) {
2737 apply_default_whitespace = xstrdup(value);
2738 return 0;
2740 return git_default_config(var, value);
2744 int cmd_apply(int argc, const char **argv, const char *unused_prefix)
2746 int i;
2747 int read_stdin = 1;
2748 int inaccurate_eof = 0;
2749 int errs = 0;
2750 int is_not_gitdir = 0;
2752 const char *whitespace_option = NULL;
2754 prefix = setup_git_directory_gently(&is_not_gitdir);
2755 prefix_length = prefix ? strlen(prefix) : 0;
2756 git_config(git_apply_config);
2757 if (apply_default_whitespace)
2758 parse_whitespace_option(apply_default_whitespace);
2760 for (i = 1; i < argc; i++) {
2761 const char *arg = argv[i];
2762 char *end;
2763 int fd;
2765 if (!strcmp(arg, "-")) {
2766 errs |= apply_patch(0, "<stdin>", inaccurate_eof);
2767 read_stdin = 0;
2768 continue;
2770 if (!prefixcmp(arg, "--exclude=")) {
2771 struct excludes *x = xmalloc(sizeof(*x));
2772 x->path = arg + 10;
2773 x->next = excludes;
2774 excludes = x;
2775 continue;
2777 if (!prefixcmp(arg, "-p")) {
2778 p_value = atoi(arg + 2);
2779 p_value_known = 1;
2780 continue;
2782 if (!strcmp(arg, "--no-add")) {
2783 no_add = 1;
2784 continue;
2786 if (!strcmp(arg, "--stat")) {
2787 apply = 0;
2788 diffstat = 1;
2789 continue;
2791 if (!strcmp(arg, "--allow-binary-replacement") ||
2792 !strcmp(arg, "--binary")) {
2793 continue; /* now no-op */
2795 if (!strcmp(arg, "--numstat")) {
2796 apply = 0;
2797 numstat = 1;
2798 continue;
2800 if (!strcmp(arg, "--summary")) {
2801 apply = 0;
2802 summary = 1;
2803 continue;
2805 if (!strcmp(arg, "--check")) {
2806 apply = 0;
2807 check = 1;
2808 continue;
2810 if (!strcmp(arg, "--index")) {
2811 if (is_not_gitdir)
2812 die("--index outside a repository");
2813 check_index = 1;
2814 continue;
2816 if (!strcmp(arg, "--cached")) {
2817 if (is_not_gitdir)
2818 die("--cached outside a repository");
2819 check_index = 1;
2820 cached = 1;
2821 continue;
2823 if (!strcmp(arg, "--apply")) {
2824 apply = 1;
2825 continue;
2827 if (!strcmp(arg, "--index-info")) {
2828 apply = 0;
2829 show_index_info = 1;
2830 continue;
2832 if (!strcmp(arg, "-z")) {
2833 line_termination = 0;
2834 continue;
2836 if (!prefixcmp(arg, "-C")) {
2837 p_context = strtoul(arg + 2, &end, 0);
2838 if (*end != '\0')
2839 die("unrecognized context count '%s'", arg + 2);
2840 continue;
2842 if (!prefixcmp(arg, "--whitespace=")) {
2843 whitespace_option = arg + 13;
2844 parse_whitespace_option(arg + 13);
2845 continue;
2847 if (!strcmp(arg, "-R") || !strcmp(arg, "--reverse")) {
2848 apply_in_reverse = 1;
2849 continue;
2851 if (!strcmp(arg, "--unidiff-zero")) {
2852 unidiff_zero = 1;
2853 continue;
2855 if (!strcmp(arg, "--reject")) {
2856 apply = apply_with_reject = apply_verbosely = 1;
2857 continue;
2859 if (!strcmp(arg, "-v") || !strcmp(arg, "--verbose")) {
2860 apply_verbosely = 1;
2861 continue;
2863 if (!strcmp(arg, "--inaccurate-eof")) {
2864 inaccurate_eof = 1;
2865 continue;
2867 if (0 < prefix_length)
2868 arg = prefix_filename(prefix, prefix_length, arg);
2870 fd = open(arg, O_RDONLY);
2871 if (fd < 0)
2872 usage(apply_usage);
2873 read_stdin = 0;
2874 set_default_whitespace_mode(whitespace_option);
2875 errs |= apply_patch(fd, arg, inaccurate_eof);
2876 close(fd);
2878 set_default_whitespace_mode(whitespace_option);
2879 if (read_stdin)
2880 errs |= apply_patch(0, "<stdin>", inaccurate_eof);
2881 if (whitespace_error) {
2882 if (squelch_whitespace_errors &&
2883 squelch_whitespace_errors < whitespace_error) {
2884 int squelched =
2885 whitespace_error - squelch_whitespace_errors;
2886 fprintf(stderr, "warning: squelched %d "
2887 "whitespace error%s\n",
2888 squelched,
2889 squelched == 1 ? "" : "s");
2891 if (new_whitespace == error_on_whitespace)
2892 die("%d line%s add%s whitespace errors.",
2893 whitespace_error,
2894 whitespace_error == 1 ? "" : "s",
2895 whitespace_error == 1 ? "s" : "");
2896 if (applied_after_fixing_ws)
2897 fprintf(stderr, "warning: %d line%s applied after"
2898 " fixing whitespace errors.\n",
2899 applied_after_fixing_ws,
2900 applied_after_fixing_ws == 1 ? "" : "s");
2901 else if (whitespace_error)
2902 fprintf(stderr, "warning: %d line%s add%s whitespace errors.\n",
2903 whitespace_error,
2904 whitespace_error == 1 ? "" : "s",
2905 whitespace_error == 1 ? "s" : "");
2908 if (update_index) {
2909 if (write_cache(newfd, active_cache, active_nr) ||
2910 close(newfd) || commit_locked_index(&lock_file))
2911 die("Unable to write new index file");
2914 return !!errs;