git-core-foo -> git-foo, except the core package
[git/vmiklos.git] / apply.c
blob590adc6afafa98e389dcafc2f8c432db09412d4e
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"
13 // --check turns on checking that the working tree matches the
14 // files that are being modified, but doesn't apply the patch
15 // --stat does just a diffstat, and doesn't actually apply
16 // --numstat does numeric diffstat, and doesn't actually apply
17 // --index-info shows the old and new index info for paths if available.
19 static int check_index = 0;
20 static int write_index = 0;
21 static int diffstat = 0;
22 static int numstat = 0;
23 static int summary = 0;
24 static int check = 0;
25 static int apply = 1;
26 static int no_add = 0;
27 static int show_index_info = 0;
28 static int line_termination = '\n';
29 static const char apply_usage[] =
30 "git-apply [--stat] [--numstat] [--summary] [--check] [--index] [--apply] [--no-add] [--index-info] [-z] <patch>...";
33 * For "diff-stat" like behaviour, we keep track of the biggest change
34 * we've seen, and the longest filename. That allows us to do simple
35 * scaling.
37 static int max_change, max_len;
40 * Various "current state", notably line numbers and what
41 * file (and how) we're patching right now.. The "is_xxxx"
42 * things are flags, where -1 means "don't know yet".
44 static int linenr = 1;
46 struct fragment {
47 unsigned long oldpos, oldlines;
48 unsigned long newpos, newlines;
49 const char *patch;
50 int size;
51 struct fragment *next;
54 struct patch {
55 char *new_name, *old_name, *def_name;
56 unsigned int old_mode, new_mode;
57 int is_rename, is_copy, is_new, is_delete, is_binary;
58 int lines_added, lines_deleted;
59 int score;
60 struct fragment *fragments;
61 char *result;
62 unsigned long resultsize;
63 char old_sha1_prefix[41];
64 char new_sha1_prefix[41];
65 struct patch *next;
68 #define CHUNKSIZE (8192)
69 #define SLOP (16)
71 static void *read_patch_file(int fd, unsigned long *sizep)
73 unsigned long size = 0, alloc = CHUNKSIZE;
74 void *buffer = xmalloc(alloc);
76 for (;;) {
77 int nr = alloc - size;
78 if (nr < 1024) {
79 alloc += CHUNKSIZE;
80 buffer = xrealloc(buffer, alloc);
81 nr = alloc - size;
83 nr = read(fd, buffer + size, nr);
84 if (!nr)
85 break;
86 if (nr < 0) {
87 if (errno == EAGAIN)
88 continue;
89 die("git-apply: read returned %s", strerror(errno));
91 size += nr;
93 *sizep = size;
96 * Make sure that we have some slop in the buffer
97 * so that we can do speculative "memcmp" etc, and
98 * see to it that it is NUL-filled.
100 if (alloc < size + SLOP)
101 buffer = xrealloc(buffer, size + SLOP);
102 memset(buffer + size, 0, SLOP);
103 return buffer;
106 static unsigned long linelen(const char *buffer, unsigned long size)
108 unsigned long len = 0;
109 while (size--) {
110 len++;
111 if (*buffer++ == '\n')
112 break;
114 return len;
117 static int is_dev_null(const char *str)
119 return !memcmp("/dev/null", str, 9) && isspace(str[9]);
122 #define TERM_SPACE 1
123 #define TERM_TAB 2
125 static int name_terminate(const char *name, int namelen, int c, int terminate)
127 if (c == ' ' && !(terminate & TERM_SPACE))
128 return 0;
129 if (c == '\t' && !(terminate & TERM_TAB))
130 return 0;
132 return 1;
135 static char * find_name(const char *line, char *def, int p_value, int terminate)
137 int len;
138 const char *start = line;
139 char *name;
141 if (*line == '"') {
142 /* Proposed "new-style" GNU patch/diff format; see
143 * http://marc.theaimsgroup.com/?l=git&m=112927316408690&w=2
145 name = unquote_c_style(line, NULL);
146 if (name) {
147 char *cp = name;
148 while (p_value) {
149 cp = strchr(name, '/');
150 if (!cp)
151 break;
152 cp++;
153 p_value--;
155 if (cp) {
156 /* name can later be freed, so we need
157 * to memmove, not just return cp
159 memmove(name, cp, strlen(cp) + 1);
160 free(def);
161 return name;
163 else {
164 free(name);
165 name = NULL;
170 for (;;) {
171 char c = *line;
173 if (isspace(c)) {
174 if (c == '\n')
175 break;
176 if (name_terminate(start, line-start, c, terminate))
177 break;
179 line++;
180 if (c == '/' && !--p_value)
181 start = line;
183 if (!start)
184 return def;
185 len = line - start;
186 if (!len)
187 return def;
190 * Generally we prefer the shorter name, especially
191 * if the other one is just a variation of that with
192 * something else tacked on to the end (ie "file.orig"
193 * or "file~").
195 if (def) {
196 int deflen = strlen(def);
197 if (deflen < len && !strncmp(start, def, deflen))
198 return def;
201 name = xmalloc(len + 1);
202 memcpy(name, start, len);
203 name[len] = 0;
204 free(def);
205 return name;
209 * Get the name etc info from the --/+++ lines of a traditional patch header
211 * NOTE! This hardcodes "-p1" behaviour in filename detection.
213 * FIXME! The end-of-filename heuristics are kind of screwy. For existing
214 * files, we can happily check the index for a match, but for creating a
215 * new file we should try to match whatever "patch" does. I have no idea.
217 static void parse_traditional_patch(const char *first, const char *second, struct patch *patch)
219 int p_value = 1;
220 char *name;
222 first += 4; // skip "--- "
223 second += 4; // skip "+++ "
224 if (is_dev_null(first)) {
225 patch->is_new = 1;
226 patch->is_delete = 0;
227 name = find_name(second, NULL, p_value, TERM_SPACE | TERM_TAB);
228 patch->new_name = name;
229 } else if (is_dev_null(second)) {
230 patch->is_new = 0;
231 patch->is_delete = 1;
232 name = find_name(first, NULL, p_value, TERM_SPACE | TERM_TAB);
233 patch->old_name = name;
234 } else {
235 name = find_name(first, NULL, p_value, TERM_SPACE | TERM_TAB);
236 name = find_name(second, name, p_value, TERM_SPACE | TERM_TAB);
237 patch->old_name = patch->new_name = name;
239 if (!name)
240 die("unable to find filename in patch at line %d", linenr);
243 static int gitdiff_hdrend(const char *line, struct patch *patch)
245 return -1;
249 * We're anal about diff header consistency, to make
250 * sure that we don't end up having strange ambiguous
251 * patches floating around.
253 * As a result, gitdiff_{old|new}name() will check
254 * their names against any previous information, just
255 * to make sure..
257 static char *gitdiff_verify_name(const char *line, int isnull, char *orig_name, const char *oldnew)
259 if (!orig_name && !isnull)
260 return find_name(line, NULL, 1, 0);
262 if (orig_name) {
263 int len;
264 const char *name;
265 char *another;
266 name = orig_name;
267 len = strlen(name);
268 if (isnull)
269 die("git-apply: bad git-diff - expected /dev/null, got %s on line %d", name, linenr);
270 another = find_name(line, NULL, 1, 0);
271 if (!another || memcmp(another, name, len))
272 die("git-apply: bad git-diff - inconsistent %s filename on line %d", oldnew, linenr);
273 free(another);
274 return orig_name;
276 else {
277 /* expect "/dev/null" */
278 if (memcmp("/dev/null", line, 9) || line[9] != '\n')
279 die("git-apply: bad git-diff - expected /dev/null on line %d", linenr);
280 return NULL;
284 static int gitdiff_oldname(const char *line, struct patch *patch)
286 patch->old_name = gitdiff_verify_name(line, patch->is_new, patch->old_name, "old");
287 return 0;
290 static int gitdiff_newname(const char *line, struct patch *patch)
292 patch->new_name = gitdiff_verify_name(line, patch->is_delete, patch->new_name, "new");
293 return 0;
296 static int gitdiff_oldmode(const char *line, struct patch *patch)
298 patch->old_mode = strtoul(line, NULL, 8);
299 return 0;
302 static int gitdiff_newmode(const char *line, struct patch *patch)
304 patch->new_mode = strtoul(line, NULL, 8);
305 return 0;
308 static int gitdiff_delete(const char *line, struct patch *patch)
310 patch->is_delete = 1;
311 patch->old_name = patch->def_name;
312 return gitdiff_oldmode(line, patch);
315 static int gitdiff_newfile(const char *line, struct patch *patch)
317 patch->is_new = 1;
318 patch->new_name = patch->def_name;
319 return gitdiff_newmode(line, patch);
322 static int gitdiff_copysrc(const char *line, struct patch *patch)
324 patch->is_copy = 1;
325 patch->old_name = find_name(line, NULL, 0, 0);
326 return 0;
329 static int gitdiff_copydst(const char *line, struct patch *patch)
331 patch->is_copy = 1;
332 patch->new_name = find_name(line, NULL, 0, 0);
333 return 0;
336 static int gitdiff_renamesrc(const char *line, struct patch *patch)
338 patch->is_rename = 1;
339 patch->old_name = find_name(line, NULL, 0, 0);
340 return 0;
343 static int gitdiff_renamedst(const char *line, struct patch *patch)
345 patch->is_rename = 1;
346 patch->new_name = find_name(line, NULL, 0, 0);
347 return 0;
350 static int gitdiff_similarity(const char *line, struct patch *patch)
352 if ((patch->score = strtoul(line, NULL, 10)) == ULONG_MAX)
353 patch->score = 0;
354 return 0;
357 static int gitdiff_dissimilarity(const char *line, struct patch *patch)
359 if ((patch->score = strtoul(line, NULL, 10)) == ULONG_MAX)
360 patch->score = 0;
361 return 0;
364 static int gitdiff_index(const char *line, struct patch *patch)
366 /* index line is N hexadecimal, "..", N hexadecimal,
367 * and optional space with octal mode.
369 const char *ptr, *eol;
370 int len;
372 ptr = strchr(line, '.');
373 if (!ptr || ptr[1] != '.' || 40 < ptr - line)
374 return 0;
375 len = ptr - line;
376 memcpy(patch->old_sha1_prefix, line, len);
377 patch->old_sha1_prefix[len] = 0;
379 line = ptr + 2;
380 ptr = strchr(line, ' ');
381 eol = strchr(line, '\n');
383 if (!ptr || eol < ptr)
384 ptr = eol;
385 len = ptr - line;
387 if (40 < len)
388 return 0;
389 memcpy(patch->new_sha1_prefix, line, len);
390 patch->new_sha1_prefix[len] = 0;
391 if (*ptr == ' ')
392 patch->new_mode = patch->old_mode = strtoul(ptr+1, NULL, 8);
393 return 0;
397 * This is normal for a diff that doesn't change anything: we'll fall through
398 * into the next diff. Tell the parser to break out.
400 static int gitdiff_unrecognized(const char *line, struct patch *patch)
402 return -1;
405 static const char *stop_at_slash(const char *line, int llen)
407 int i;
409 for (i = 0; i < llen; i++) {
410 int ch = line[i];
411 if (ch == '/')
412 return line + i;
414 return NULL;
417 /* This is to extract the same name that appears on "diff --git"
418 * line. We do not find and return anything if it is a rename
419 * patch, and it is OK because we will find the name elsewhere.
420 * We need to reliably find name only when it is mode-change only,
421 * creation or deletion of an empty file. In any of these cases,
422 * both sides are the same name under a/ and b/ respectively.
424 static char *git_header_name(char *line, int llen)
426 int len;
427 const char *name;
428 const char *second = NULL;
430 line += strlen("diff --git ");
431 llen -= strlen("diff --git ");
433 if (*line == '"') {
434 const char *cp;
435 char *first = unquote_c_style(line, &second);
436 if (!first)
437 return NULL;
439 /* advance to the first slash */
440 cp = stop_at_slash(first, strlen(first));
441 if (!cp || cp == first) {
442 /* we do not accept absolute paths */
443 free_first_and_fail:
444 free(first);
445 return NULL;
447 len = strlen(cp+1);
448 memmove(first, cp+1, len+1); /* including NUL */
450 /* second points at one past closing dq of name.
451 * find the second name.
453 while ((second < line + llen) && isspace(*second))
454 second++;
456 if (line + llen <= second)
457 goto free_first_and_fail;
458 if (*second == '"') {
459 char *sp = unquote_c_style(second, NULL);
460 if (!sp)
461 goto free_first_and_fail;
462 cp = stop_at_slash(sp, strlen(sp));
463 if (!cp || cp == sp) {
464 free_both_and_fail:
465 free(sp);
466 goto free_first_and_fail;
468 /* They must match, otherwise ignore */
469 if (strcmp(cp+1, first))
470 goto free_both_and_fail;
471 free(sp);
472 return first;
475 /* unquoted second */
476 cp = stop_at_slash(second, line + llen - second);
477 if (!cp || cp == second)
478 goto free_first_and_fail;
479 cp++;
480 if (line + llen - cp != len + 1 ||
481 memcmp(first, cp, len))
482 goto free_first_and_fail;
483 return first;
486 /* unquoted first name */
487 name = stop_at_slash(line, llen);
488 if (!name || name == line)
489 return NULL;
491 name++;
493 /* since the first name is unquoted, a dq if exists must be
494 * the beginning of the second name.
496 for (second = name; second < line + llen; second++) {
497 if (*second == '"') {
498 const char *cp = second;
499 const char *np;
500 char *sp = unquote_c_style(second, NULL);
502 if (!sp)
503 return NULL;
504 np = stop_at_slash(sp, strlen(sp));
505 if (!np || np == sp) {
506 free_second_and_fail:
507 free(sp);
508 return NULL;
510 np++;
511 len = strlen(np);
512 if (len < cp - name &&
513 !strncmp(np, name, len) &&
514 isspace(name[len])) {
515 /* Good */
516 memmove(sp, np, len + 1);
517 return sp;
519 goto free_second_and_fail;
524 * Accept a name only if it shows up twice, exactly the same
525 * form.
527 for (len = 0 ; ; len++) {
528 char c = name[len];
530 switch (c) {
531 default:
532 continue;
533 case '\n':
534 return NULL;
535 case '\t': case ' ':
536 second = name+len;
537 for (;;) {
538 char c = *second++;
539 if (c == '\n')
540 return NULL;
541 if (c == '/')
542 break;
544 if (second[len] == '\n' && !memcmp(name, second, len)) {
545 char *ret = xmalloc(len + 1);
546 memcpy(ret, name, len);
547 ret[len] = 0;
548 return ret;
552 return NULL;
555 /* Verify that we recognize the lines following a git header */
556 static int parse_git_header(char *line, int len, unsigned int size, struct patch *patch)
558 unsigned long offset;
560 /* A git diff has explicit new/delete information, so we don't guess */
561 patch->is_new = 0;
562 patch->is_delete = 0;
565 * Some things may not have the old name in the
566 * rest of the headers anywhere (pure mode changes,
567 * or removing or adding empty files), so we get
568 * the default name from the header.
570 patch->def_name = git_header_name(line, len);
572 line += len;
573 size -= len;
574 linenr++;
575 for (offset = len ; size > 0 ; offset += len, size -= len, line += len, linenr++) {
576 static const struct opentry {
577 const char *str;
578 int (*fn)(const char *, struct patch *);
579 } optable[] = {
580 { "@@ -", gitdiff_hdrend },
581 { "--- ", gitdiff_oldname },
582 { "+++ ", gitdiff_newname },
583 { "old mode ", gitdiff_oldmode },
584 { "new mode ", gitdiff_newmode },
585 { "deleted file mode ", gitdiff_delete },
586 { "new file mode ", gitdiff_newfile },
587 { "copy from ", gitdiff_copysrc },
588 { "copy to ", gitdiff_copydst },
589 { "rename old ", gitdiff_renamesrc },
590 { "rename new ", gitdiff_renamedst },
591 { "rename from ", gitdiff_renamesrc },
592 { "rename to ", gitdiff_renamedst },
593 { "similarity index ", gitdiff_similarity },
594 { "dissimilarity index ", gitdiff_dissimilarity },
595 { "index ", gitdiff_index },
596 { "", gitdiff_unrecognized },
598 int i;
600 len = linelen(line, size);
601 if (!len || line[len-1] != '\n')
602 break;
603 for (i = 0; i < sizeof(optable) / sizeof(optable[0]); i++) {
604 const struct opentry *p = optable + i;
605 int oplen = strlen(p->str);
606 if (len < oplen || memcmp(p->str, line, oplen))
607 continue;
608 if (p->fn(line + oplen, patch) < 0)
609 return offset;
610 break;
614 return offset;
617 static int parse_num(const char *line, unsigned long *p)
619 char *ptr;
621 if (!isdigit(*line))
622 return 0;
623 *p = strtoul(line, &ptr, 10);
624 return ptr - line;
627 static int parse_range(const char *line, int len, int offset, const char *expect,
628 unsigned long *p1, unsigned long *p2)
630 int digits, ex;
632 if (offset < 0 || offset >= len)
633 return -1;
634 line += offset;
635 len -= offset;
637 digits = parse_num(line, p1);
638 if (!digits)
639 return -1;
641 offset += digits;
642 line += digits;
643 len -= digits;
645 *p2 = *p1;
646 if (*line == ',') {
647 digits = parse_num(line+1, p2);
648 if (!digits)
649 return -1;
651 offset += digits+1;
652 line += digits+1;
653 len -= digits+1;
656 ex = strlen(expect);
657 if (ex > len)
658 return -1;
659 if (memcmp(line, expect, ex))
660 return -1;
662 return offset + ex;
666 * Parse a unified diff fragment header of the
667 * form "@@ -a,b +c,d @@"
669 static int parse_fragment_header(char *line, int len, struct fragment *fragment)
671 int offset;
673 if (!len || line[len-1] != '\n')
674 return -1;
676 /* Figure out the number of lines in a fragment */
677 offset = parse_range(line, len, 4, " +", &fragment->oldpos, &fragment->oldlines);
678 offset = parse_range(line, len, offset, " @@", &fragment->newpos, &fragment->newlines);
680 return offset;
683 static int find_header(char *line, unsigned long size, int *hdrsize, struct patch *patch)
685 unsigned long offset, len;
687 patch->is_rename = patch->is_copy = 0;
688 patch->is_new = patch->is_delete = -1;
689 patch->old_mode = patch->new_mode = 0;
690 patch->old_name = patch->new_name = NULL;
691 for (offset = 0; size > 0; offset += len, size -= len, line += len, linenr++) {
692 unsigned long nextlen;
694 len = linelen(line, size);
695 if (!len)
696 break;
698 /* Testing this early allows us to take a few shortcuts.. */
699 if (len < 6)
700 continue;
703 * Make sure we don't find any unconnected patch fragmants.
704 * That's a sign that we didn't find a header, and that a
705 * patch has become corrupted/broken up.
707 if (!memcmp("@@ -", line, 4)) {
708 struct fragment dummy;
709 if (parse_fragment_header(line, len, &dummy) < 0)
710 continue;
711 error("patch fragment without header at line %d: %.*s", linenr, (int)len-1, line);
714 if (size < len + 6)
715 break;
718 * Git patch? It might not have a real patch, just a rename
719 * or mode change, so we handle that specially
721 if (!memcmp("diff --git ", line, 11)) {
722 int git_hdr_len = parse_git_header(line, len, size, patch);
723 if (git_hdr_len <= len)
724 continue;
725 if (!patch->old_name && !patch->new_name) {
726 if (!patch->def_name)
727 die("git diff header lacks filename information (line %d)", linenr);
728 patch->old_name = patch->new_name = patch->def_name;
730 *hdrsize = git_hdr_len;
731 return offset;
734 /** --- followed by +++ ? */
735 if (memcmp("--- ", line, 4) || memcmp("+++ ", line + len, 4))
736 continue;
739 * We only accept unified patches, so we want it to
740 * at least have "@@ -a,b +c,d @@\n", which is 14 chars
741 * minimum
743 nextlen = linelen(line + len, size - len);
744 if (size < nextlen + 14 || memcmp("@@ -", line + len + nextlen, 4))
745 continue;
747 /* Ok, we'll consider it a patch */
748 parse_traditional_patch(line, line+len, patch);
749 *hdrsize = len + nextlen;
750 linenr += 2;
751 return offset;
753 return -1;
757 * Parse a unified diff. Note that this really needs
758 * to parse each fragment separately, since the only
759 * way to know the difference between a "---" that is
760 * part of a patch, and a "---" that starts the next
761 * patch is to look at the line counts..
763 static int parse_fragment(char *line, unsigned long size, struct patch *patch, struct fragment *fragment)
765 int added, deleted;
766 int len = linelen(line, size), offset;
767 unsigned long oldlines, newlines;
769 offset = parse_fragment_header(line, len, fragment);
770 if (offset < 0)
771 return -1;
772 oldlines = fragment->oldlines;
773 newlines = fragment->newlines;
775 if (patch->is_new < 0) {
776 patch->is_new = !oldlines;
777 if (!oldlines)
778 patch->old_name = NULL;
780 if (patch->is_delete < 0) {
781 patch->is_delete = !newlines;
782 if (!newlines)
783 patch->new_name = NULL;
786 if (patch->is_new != !oldlines)
787 return error("new file depends on old contents");
788 if (patch->is_delete != !newlines) {
789 if (newlines)
790 return error("deleted file still has contents");
791 fprintf(stderr, "** warning: file %s becomes empty but is not deleted\n", patch->new_name);
794 /* Parse the thing.. */
795 line += len;
796 size -= len;
797 linenr++;
798 added = deleted = 0;
799 for (offset = len; size > 0; offset += len, size -= len, line += len, linenr++) {
800 if (!oldlines && !newlines)
801 break;
802 len = linelen(line, size);
803 if (!len || line[len-1] != '\n')
804 return -1;
805 switch (*line) {
806 default:
807 return -1;
808 case ' ':
809 oldlines--;
810 newlines--;
811 break;
812 case '-':
813 deleted++;
814 oldlines--;
815 break;
816 case '+':
817 added++;
818 newlines--;
819 break;
821 /* We allow "\ No newline at end of file". Depending
822 * on locale settings when the patch was produced we
823 * don't know what this line looks like. The only
824 * thing we do know is that it begins with "\ ".
825 * Checking for 12 is just for sanity check -- any
826 * l10n of "\ No newline..." is at least that long.
828 case '\\':
829 if (len < 12 || memcmp(line, "\\ ", 2))
830 return -1;
831 break;
834 /* If a fragment ends with an incomplete line, we failed to include
835 * it in the above loop because we hit oldlines == newlines == 0
836 * before seeing it.
838 if (12 < size && !memcmp(line, "\\ ", 2))
839 offset += linelen(line, size);
841 patch->lines_added += added;
842 patch->lines_deleted += deleted;
843 return offset;
846 static int parse_single_patch(char *line, unsigned long size, struct patch *patch)
848 unsigned long offset = 0;
849 struct fragment **fragp = &patch->fragments;
851 while (size > 4 && !memcmp(line, "@@ -", 4)) {
852 struct fragment *fragment;
853 int len;
855 fragment = xmalloc(sizeof(*fragment));
856 memset(fragment, 0, sizeof(*fragment));
857 len = parse_fragment(line, size, patch, fragment);
858 if (len <= 0)
859 die("corrupt patch at line %d", linenr);
861 fragment->patch = line;
862 fragment->size = len;
864 *fragp = fragment;
865 fragp = &fragment->next;
867 offset += len;
868 line += len;
869 size -= len;
871 return offset;
874 static inline int metadata_changes(struct patch *patch)
876 return patch->is_rename > 0 ||
877 patch->is_copy > 0 ||
878 patch->is_new > 0 ||
879 patch->is_delete ||
880 (patch->old_mode && patch->new_mode &&
881 patch->old_mode != patch->new_mode);
884 static int parse_chunk(char *buffer, unsigned long size, struct patch *patch)
886 int hdrsize, patchsize;
887 int offset = find_header(buffer, size, &hdrsize, patch);
889 if (offset < 0)
890 return offset;
892 patchsize = parse_single_patch(buffer + offset + hdrsize, size - offset - hdrsize, patch);
894 if (!patchsize && !metadata_changes(patch)) {
895 static const char binhdr[] = "Binary files ";
897 if (sizeof(binhdr) - 1 < size - offset - hdrsize &&
898 !memcmp(binhdr, buffer + hdrsize + offset,
899 sizeof(binhdr)-1))
900 patch->is_binary = 1;
902 if (patch->is_binary && !apply && !check)
904 else
905 die("patch with only garbage at line %d", linenr);
908 return offset + hdrsize + patchsize;
911 static const char pluses[] = "++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++";
912 static const char minuses[]= "----------------------------------------------------------------------";
914 static void show_stats(struct patch *patch)
916 const char *prefix = "";
917 char *name = patch->new_name;
918 char *qname = NULL;
919 int len, max, add, del, total;
921 if (!name)
922 name = patch->old_name;
924 if (0 < (len = quote_c_style(name, NULL, NULL, 0))) {
925 qname = xmalloc(len + 1);
926 quote_c_style(name, qname, NULL, 0);
927 name = qname;
931 * "scale" the filename
933 len = strlen(name);
934 max = max_len;
935 if (max > 50)
936 max = 50;
937 if (len > max) {
938 char *slash;
939 prefix = "...";
940 max -= 3;
941 name += len - max;
942 slash = strchr(name, '/');
943 if (slash)
944 name = slash;
946 len = max;
949 * scale the add/delete
951 max = max_change;
952 if (max + len > 70)
953 max = 70 - len;
955 add = patch->lines_added;
956 del = patch->lines_deleted;
957 total = add + del;
959 if (max_change > 0) {
960 total = (total * max + max_change / 2) / max_change;
961 add = (add * max + max_change / 2) / max_change;
962 del = total - add;
964 if (patch->is_binary)
965 printf(" %s%-*s | Bin\n", prefix, len, name);
966 else
967 printf(" %s%-*s |%5d %.*s%.*s\n", prefix,
968 len, name, patch->lines_added + patch->lines_deleted,
969 add, pluses, del, minuses);
970 if (qname)
971 free(qname);
974 static int read_old_data(struct stat *st, const char *path, void *buf, unsigned long size)
976 int fd;
977 unsigned long got;
979 switch (st->st_mode & S_IFMT) {
980 case S_IFLNK:
981 return readlink(path, buf, size);
982 case S_IFREG:
983 fd = open(path, O_RDONLY);
984 if (fd < 0)
985 return error("unable to open %s", path);
986 got = 0;
987 for (;;) {
988 int ret = read(fd, buf + got, size - got);
989 if (ret < 0) {
990 if (errno == EAGAIN)
991 continue;
992 break;
994 if (!ret)
995 break;
996 got += ret;
998 close(fd);
999 return got;
1001 default:
1002 return -1;
1006 static int find_offset(const char *buf, unsigned long size, const char *fragment, unsigned long fragsize, int line)
1008 int i;
1009 unsigned long start, backwards, forwards;
1011 if (fragsize > size)
1012 return -1;
1014 start = 0;
1015 if (line > 1) {
1016 unsigned long offset = 0;
1017 i = line-1;
1018 while (offset + fragsize <= size) {
1019 if (buf[offset++] == '\n') {
1020 start = offset;
1021 if (!--i)
1022 break;
1027 /* Exact line number? */
1028 if (!memcmp(buf + start, fragment, fragsize))
1029 return start;
1032 * There's probably some smart way to do this, but I'll leave
1033 * that to the smart and beautiful people. I'm simple and stupid.
1035 backwards = start;
1036 forwards = start;
1037 for (i = 0; ; i++) {
1038 unsigned long try;
1039 int n;
1041 /* "backward" */
1042 if (i & 1) {
1043 if (!backwards) {
1044 if (forwards + fragsize > size)
1045 break;
1046 continue;
1048 do {
1049 --backwards;
1050 } while (backwards && buf[backwards-1] != '\n');
1051 try = backwards;
1052 } else {
1053 while (forwards + fragsize <= size) {
1054 if (buf[forwards++] == '\n')
1055 break;
1057 try = forwards;
1060 if (try + fragsize > size)
1061 continue;
1062 if (memcmp(buf + try, fragment, fragsize))
1063 continue;
1064 n = (i >> 1)+1;
1065 if (i & 1)
1066 n = -n;
1067 return try;
1071 * We should start searching forward and backward.
1073 return -1;
1076 struct buffer_desc {
1077 char *buffer;
1078 unsigned long size;
1079 unsigned long alloc;
1082 static int apply_one_fragment(struct buffer_desc *desc, struct fragment *frag)
1084 char *buf = desc->buffer;
1085 const char *patch = frag->patch;
1086 int offset, size = frag->size;
1087 char *old = xmalloc(size);
1088 char *new = xmalloc(size);
1089 int oldsize = 0, newsize = 0;
1091 while (size > 0) {
1092 int len = linelen(patch, size);
1093 int plen;
1095 if (!len)
1096 break;
1099 * "plen" is how much of the line we should use for
1100 * the actual patch data. Normally we just remove the
1101 * first character on the line, but if the line is
1102 * followed by "\ No newline", then we also remove the
1103 * last one (which is the newline, of course).
1105 plen = len-1;
1106 if (len < size && patch[len] == '\\')
1107 plen--;
1108 switch (*patch) {
1109 case ' ':
1110 case '-':
1111 memcpy(old + oldsize, patch + 1, plen);
1112 oldsize += plen;
1113 if (*patch == '-')
1114 break;
1115 /* Fall-through for ' ' */
1116 case '+':
1117 if (*patch != '+' || !no_add) {
1118 memcpy(new + newsize, patch + 1, plen);
1119 newsize += plen;
1121 break;
1122 case '@': case '\\':
1123 /* Ignore it, we already handled it */
1124 break;
1125 default:
1126 return -1;
1128 patch += len;
1129 size -= len;
1132 offset = find_offset(buf, desc->size, old, oldsize, frag->newpos);
1133 if (offset >= 0) {
1134 int diff = newsize - oldsize;
1135 unsigned long size = desc->size + diff;
1136 unsigned long alloc = desc->alloc;
1138 if (size > alloc) {
1139 alloc = size + 8192;
1140 desc->alloc = alloc;
1141 buf = xrealloc(buf, alloc);
1142 desc->buffer = buf;
1144 desc->size = size;
1145 memmove(buf + offset + newsize, buf + offset + oldsize, size - offset - newsize);
1146 memcpy(buf + offset, new, newsize);
1147 offset = 0;
1150 free(old);
1151 free(new);
1152 return offset;
1155 static int apply_fragments(struct buffer_desc *desc, struct patch *patch)
1157 struct fragment *frag = patch->fragments;
1159 while (frag) {
1160 if (apply_one_fragment(desc, frag) < 0)
1161 return error("patch failed: %s:%ld", patch->old_name, frag->oldpos);
1162 frag = frag->next;
1164 return 0;
1167 static int apply_data(struct patch *patch, struct stat *st)
1169 char *buf;
1170 unsigned long size, alloc;
1171 struct buffer_desc desc;
1173 size = 0;
1174 alloc = 0;
1175 buf = NULL;
1176 if (patch->old_name) {
1177 size = st->st_size;
1178 alloc = size + 8192;
1179 buf = xmalloc(alloc);
1180 if (read_old_data(st, patch->old_name, buf, alloc) != size)
1181 return error("read of %s failed", patch->old_name);
1184 desc.size = size;
1185 desc.alloc = alloc;
1186 desc.buffer = buf;
1187 if (apply_fragments(&desc, patch) < 0)
1188 return -1;
1189 patch->result = desc.buffer;
1190 patch->resultsize = desc.size;
1192 if (patch->is_delete && patch->resultsize)
1193 return error("removal patch leaves file contents");
1195 return 0;
1198 static int check_patch(struct patch *patch)
1200 struct stat st;
1201 const char *old_name = patch->old_name;
1202 const char *new_name = patch->new_name;
1204 if (old_name) {
1205 int changed;
1206 int stat_ret = lstat(old_name, &st);
1208 if (check_index) {
1209 int pos = cache_name_pos(old_name, strlen(old_name));
1210 if (pos < 0)
1211 return error("%s: does not exist in index",
1212 old_name);
1213 if (stat_ret < 0) {
1214 struct checkout costate;
1215 if (errno != ENOENT)
1216 return error("%s: %s", old_name,
1217 strerror(errno));
1218 /* checkout */
1219 costate.base_dir = "";
1220 costate.base_dir_len = 0;
1221 costate.force = 0;
1222 costate.quiet = 0;
1223 costate.not_new = 0;
1224 costate.refresh_cache = 1;
1225 if (checkout_entry(active_cache[pos],
1226 &costate) ||
1227 lstat(old_name, &st))
1228 return -1;
1231 changed = ce_match_stat(active_cache[pos], &st);
1232 if (changed)
1233 return error("%s: does not match index",
1234 old_name);
1236 else if (stat_ret < 0)
1237 return error("%s: %s", old_name, strerror(errno));
1239 if (patch->is_new < 0)
1240 patch->is_new = 0;
1241 st.st_mode = ntohl(create_ce_mode(st.st_mode));
1242 if (!patch->old_mode)
1243 patch->old_mode = st.st_mode;
1244 if ((st.st_mode ^ patch->old_mode) & S_IFMT)
1245 return error("%s: wrong type", old_name);
1246 if (st.st_mode != patch->old_mode)
1247 fprintf(stderr, "warning: %s has type %o, expected %o\n",
1248 old_name, st.st_mode, patch->old_mode);
1251 if (new_name && (patch->is_new | patch->is_rename | patch->is_copy)) {
1252 if (check_index && cache_name_pos(new_name, strlen(new_name)) >= 0)
1253 return error("%s: already exists in index", new_name);
1254 if (!lstat(new_name, &st))
1255 return error("%s: already exists in working directory", new_name);
1256 if (errno != ENOENT)
1257 return error("%s: %s", new_name, strerror(errno));
1258 if (!patch->new_mode) {
1259 if (patch->is_new)
1260 patch->new_mode = S_IFREG | 0644;
1261 else
1262 patch->new_mode = patch->old_mode;
1266 if (new_name && old_name) {
1267 int same = !strcmp(old_name, new_name);
1268 if (!patch->new_mode)
1269 patch->new_mode = patch->old_mode;
1270 if ((patch->old_mode ^ patch->new_mode) & S_IFMT)
1271 return error("new mode (%o) of %s does not match old mode (%o)%s%s",
1272 patch->new_mode, new_name, patch->old_mode,
1273 same ? "" : " of ", same ? "" : old_name);
1276 if (apply_data(patch, &st) < 0)
1277 return error("%s: patch does not apply", old_name);
1278 return 0;
1281 static int check_patch_list(struct patch *patch)
1283 int error = 0;
1285 for (;patch ; patch = patch->next)
1286 error |= check_patch(patch);
1287 return error;
1290 static inline int is_null_sha1(const unsigned char *sha1)
1292 return !memcmp(sha1, null_sha1, 20);
1295 static void show_index_list(struct patch *list)
1297 struct patch *patch;
1299 /* Once we start supporting the reverse patch, it may be
1300 * worth showing the new sha1 prefix, but until then...
1302 for (patch = list; patch; patch = patch->next) {
1303 const unsigned char *sha1_ptr;
1304 unsigned char sha1[20];
1305 const char *name;
1307 name = patch->old_name ? patch->old_name : patch->new_name;
1308 if (patch->is_new)
1309 sha1_ptr = null_sha1;
1310 else if (get_sha1(patch->old_sha1_prefix, sha1))
1311 die("sha1 information is lacking or useless (%s).",
1312 name);
1313 else
1314 sha1_ptr = sha1;
1316 printf("%06o %s ",patch->old_mode, sha1_to_hex(sha1_ptr));
1317 if (line_termination && quote_c_style(name, NULL, NULL, 0))
1318 quote_c_style(name, NULL, stdout, 0);
1319 else
1320 fputs(name, stdout);
1321 putchar(line_termination);
1325 static void stat_patch_list(struct patch *patch)
1327 int files, adds, dels;
1329 for (files = adds = dels = 0 ; patch ; patch = patch->next) {
1330 files++;
1331 adds += patch->lines_added;
1332 dels += patch->lines_deleted;
1333 show_stats(patch);
1336 printf(" %d files changed, %d insertions(+), %d deletions(-)\n", files, adds, dels);
1339 static void numstat_patch_list(struct patch *patch)
1341 for ( ; patch; patch = patch->next) {
1342 const char *name;
1343 name = patch->old_name ? patch->old_name : patch->new_name;
1344 printf("%d\t%d\t", patch->lines_added, patch->lines_deleted);
1345 if (line_termination && quote_c_style(name, NULL, NULL, 0))
1346 quote_c_style(name, NULL, stdout, 0);
1347 else
1348 fputs(name, stdout);
1349 putchar('\n');
1353 static void show_file_mode_name(const char *newdelete, unsigned int mode, const char *name)
1355 if (mode)
1356 printf(" %s mode %06o %s\n", newdelete, mode, name);
1357 else
1358 printf(" %s %s\n", newdelete, name);
1361 static void show_mode_change(struct patch *p, int show_name)
1363 if (p->old_mode && p->new_mode && p->old_mode != p->new_mode) {
1364 if (show_name)
1365 printf(" mode change %06o => %06o %s\n",
1366 p->old_mode, p->new_mode, p->new_name);
1367 else
1368 printf(" mode change %06o => %06o\n",
1369 p->old_mode, p->new_mode);
1373 static void show_rename_copy(struct patch *p)
1375 const char *renamecopy = p->is_rename ? "rename" : "copy";
1376 const char *old, *new;
1378 /* Find common prefix */
1379 old = p->old_name;
1380 new = p->new_name;
1381 while (1) {
1382 const char *slash_old, *slash_new;
1383 slash_old = strchr(old, '/');
1384 slash_new = strchr(new, '/');
1385 if (!slash_old ||
1386 !slash_new ||
1387 slash_old - old != slash_new - new ||
1388 memcmp(old, new, slash_new - new))
1389 break;
1390 old = slash_old + 1;
1391 new = slash_new + 1;
1393 /* p->old_name thru old is the common prefix, and old and new
1394 * through the end of names are renames
1396 if (old != p->old_name)
1397 printf(" %s %.*s{%s => %s} (%d%%)\n", renamecopy,
1398 (int)(old - p->old_name), p->old_name,
1399 old, new, p->score);
1400 else
1401 printf(" %s %s => %s (%d%%)\n", renamecopy,
1402 p->old_name, p->new_name, p->score);
1403 show_mode_change(p, 0);
1406 static void summary_patch_list(struct patch *patch)
1408 struct patch *p;
1410 for (p = patch; p; p = p->next) {
1411 if (p->is_new)
1412 show_file_mode_name("create", p->new_mode, p->new_name);
1413 else if (p->is_delete)
1414 show_file_mode_name("delete", p->old_mode, p->old_name);
1415 else {
1416 if (p->is_rename || p->is_copy)
1417 show_rename_copy(p);
1418 else {
1419 if (p->score) {
1420 printf(" rewrite %s (%d%%)\n",
1421 p->new_name, p->score);
1422 show_mode_change(p, 0);
1424 else
1425 show_mode_change(p, 1);
1431 static void patch_stats(struct patch *patch)
1433 int lines = patch->lines_added + patch->lines_deleted;
1435 if (lines > max_change)
1436 max_change = lines;
1437 if (patch->old_name) {
1438 int len = quote_c_style(patch->old_name, NULL, NULL, 0);
1439 if (!len)
1440 len = strlen(patch->old_name);
1441 if (len > max_len)
1442 max_len = len;
1444 if (patch->new_name) {
1445 int len = quote_c_style(patch->new_name, NULL, NULL, 0);
1446 if (!len)
1447 len = strlen(patch->new_name);
1448 if (len > max_len)
1449 max_len = len;
1453 static void remove_file(struct patch *patch)
1455 if (write_index) {
1456 if (remove_file_from_cache(patch->old_name) < 0)
1457 die("unable to remove %s from index", patch->old_name);
1459 unlink(patch->old_name);
1462 static void add_index_file(const char *path, unsigned mode, void *buf, unsigned long size)
1464 struct stat st;
1465 struct cache_entry *ce;
1466 int namelen = strlen(path);
1467 unsigned ce_size = cache_entry_size(namelen);
1469 if (!write_index)
1470 return;
1472 ce = xmalloc(ce_size);
1473 memset(ce, 0, ce_size);
1474 memcpy(ce->name, path, namelen);
1475 ce->ce_mode = create_ce_mode(mode);
1476 ce->ce_flags = htons(namelen);
1477 if (lstat(path, &st) < 0)
1478 die("unable to stat newly created file %s", path);
1479 fill_stat_cache_info(ce, &st);
1480 if (write_sha1_file(buf, size, "blob", ce->sha1) < 0)
1481 die("unable to create backing store for newly created file %s", path);
1482 if (add_cache_entry(ce, ADD_CACHE_OK_TO_ADD) < 0)
1483 die("unable to add cache entry for %s", path);
1486 static void create_subdirectories(const char *path)
1488 int len = strlen(path);
1489 char *buf = xmalloc(len + 1);
1490 const char *slash = path;
1492 while ((slash = strchr(slash+1, '/')) != NULL) {
1493 len = slash - path;
1494 memcpy(buf, path, len);
1495 buf[len] = 0;
1496 if (mkdir(buf, 0777) < 0) {
1497 if (errno != EEXIST)
1498 break;
1501 free(buf);
1504 static int try_create_file(const char *path, unsigned int mode, const char *buf, unsigned long size)
1506 int fd;
1508 if (S_ISLNK(mode))
1509 return symlink(buf, path);
1510 fd = open(path, O_CREAT | O_EXCL | O_WRONLY | O_TRUNC, (mode & 0100) ? 0777 : 0666);
1511 if (fd < 0)
1512 return -1;
1513 while (size) {
1514 int written = write(fd, buf, size);
1515 if (written < 0) {
1516 if (errno == EINTR || errno == EAGAIN)
1517 continue;
1518 die("writing file %s: %s", path, strerror(errno));
1520 if (!written)
1521 die("out of space writing file %s", path);
1522 buf += written;
1523 size -= written;
1525 if (close(fd) < 0)
1526 die("closing file %s: %s", path, strerror(errno));
1527 return 0;
1531 * We optimistically assume that the directories exist,
1532 * which is true 99% of the time anyway. If they don't,
1533 * we create them and try again.
1535 static void create_one_file(const char *path, unsigned mode, const char *buf, unsigned long size)
1537 if (!try_create_file(path, mode, buf, size))
1538 return;
1540 if (errno == ENOENT) {
1541 create_subdirectories(path);
1542 if (!try_create_file(path, mode, buf, size))
1543 return;
1546 if (errno == EEXIST) {
1547 unsigned int nr = getpid();
1549 for (;;) {
1550 const char *newpath;
1551 newpath = mkpath("%s~%u", path, nr);
1552 if (!try_create_file(newpath, mode, buf, size)) {
1553 if (!rename(newpath, path))
1554 return;
1555 unlink(newpath);
1556 break;
1558 if (errno != EEXIST)
1559 break;
1562 die("unable to write file %s mode %o", path, mode);
1565 static void create_file(struct patch *patch)
1567 const char *path = patch->new_name;
1568 unsigned mode = patch->new_mode;
1569 unsigned long size = patch->resultsize;
1570 char *buf = patch->result;
1572 if (!mode)
1573 mode = S_IFREG | 0644;
1574 create_one_file(path, mode, buf, size);
1575 add_index_file(path, mode, buf, size);
1578 static void write_out_one_result(struct patch *patch)
1580 if (patch->is_delete > 0) {
1581 remove_file(patch);
1582 return;
1584 if (patch->is_new > 0 || patch->is_copy) {
1585 create_file(patch);
1586 return;
1589 * Rename or modification boils down to the same
1590 * thing: remove the old, write the new
1592 remove_file(patch);
1593 create_file(patch);
1596 static void write_out_results(struct patch *list, int skipped_patch)
1598 if (!list && !skipped_patch)
1599 die("No changes");
1601 while (list) {
1602 write_out_one_result(list);
1603 list = list->next;
1607 static struct cache_file cache_file;
1609 static struct excludes {
1610 struct excludes *next;
1611 const char *path;
1612 } *excludes;
1614 static int use_patch(struct patch *p)
1616 const char *pathname = p->new_name ? p->new_name : p->old_name;
1617 struct excludes *x = excludes;
1618 while (x) {
1619 if (fnmatch(x->path, pathname, 0) == 0)
1620 return 0;
1621 x = x->next;
1623 return 1;
1626 static int apply_patch(int fd)
1628 int newfd;
1629 unsigned long offset, size;
1630 char *buffer = read_patch_file(fd, &size);
1631 struct patch *list = NULL, **listp = &list;
1632 int skipped_patch = 0;
1634 if (!buffer)
1635 return -1;
1636 offset = 0;
1637 while (size > 0) {
1638 struct patch *patch;
1639 int nr;
1641 patch = xmalloc(sizeof(*patch));
1642 memset(patch, 0, sizeof(*patch));
1643 nr = parse_chunk(buffer + offset, size, patch);
1644 if (nr < 0)
1645 break;
1646 if (use_patch(patch)) {
1647 patch_stats(patch);
1648 *listp = patch;
1649 listp = &patch->next;
1650 } else {
1651 /* perhaps free it a bit better? */
1652 free(patch);
1653 skipped_patch++;
1655 offset += nr;
1656 size -= nr;
1659 newfd = -1;
1660 write_index = check_index && apply;
1661 if (write_index)
1662 newfd = hold_index_file_for_update(&cache_file, get_index_file());
1663 if (check_index) {
1664 if (read_cache() < 0)
1665 die("unable to read index file");
1668 if ((check || apply) && check_patch_list(list) < 0)
1669 exit(1);
1671 if (apply)
1672 write_out_results(list, skipped_patch);
1674 if (write_index) {
1675 if (write_cache(newfd, active_cache, active_nr) ||
1676 commit_index_file(&cache_file))
1677 die("Unable to write new cachefile");
1680 if (show_index_info)
1681 show_index_list(list);
1683 if (diffstat)
1684 stat_patch_list(list);
1686 if (numstat)
1687 numstat_patch_list(list);
1689 if (summary)
1690 summary_patch_list(list);
1692 free(buffer);
1693 return 0;
1696 int main(int argc, char **argv)
1698 int i;
1699 int read_stdin = 1;
1701 for (i = 1; i < argc; i++) {
1702 const char *arg = argv[i];
1703 int fd;
1705 if (!strcmp(arg, "-")) {
1706 apply_patch(0);
1707 read_stdin = 0;
1708 continue;
1710 if (!strncmp(arg, "--exclude=", 10)) {
1711 struct excludes *x = xmalloc(sizeof(*x));
1712 x->path = arg + 10;
1713 x->next = excludes;
1714 excludes = x;
1715 continue;
1717 if (!strcmp(arg, "--no-add")) {
1718 no_add = 1;
1719 continue;
1721 if (!strcmp(arg, "--stat")) {
1722 apply = 0;
1723 diffstat = 1;
1724 continue;
1726 if (!strcmp(arg, "--numstat")) {
1727 apply = 0;
1728 numstat = 1;
1729 continue;
1731 if (!strcmp(arg, "--summary")) {
1732 apply = 0;
1733 summary = 1;
1734 continue;
1736 if (!strcmp(arg, "--check")) {
1737 apply = 0;
1738 check = 1;
1739 continue;
1741 if (!strcmp(arg, "--index")) {
1742 check_index = 1;
1743 continue;
1745 if (!strcmp(arg, "--apply")) {
1746 apply = 1;
1747 continue;
1749 if (!strcmp(arg, "--index-info")) {
1750 apply = 0;
1751 show_index_info = 1;
1752 continue;
1754 if (!strcmp(arg, "-z")) {
1755 line_termination = 0;
1756 continue;
1758 fd = open(arg, O_RDONLY);
1759 if (fd < 0)
1760 usage(apply_usage);
1761 read_stdin = 0;
1762 apply_patch(fd);
1763 close(fd);
1765 if (read_stdin)
1766 apply_patch(0);
1767 return 0;