[PATCH] diff: consolidate test helper script pieces.
[git/kirr.git] / apply.c
blobdba8292267abe48b6f671315a3366d28b33fe0e3
1 /*
2 * apply.c
4 * Copyright (C) Linus Torvalds, 2005
6 * This applies patches on top of some (arbitrary) version of the SCM.
8 * NOTE! It does all its work in the index file, and only cares about
9 * the files in the working directory if you tell it to "merge" the
10 * patch apply.
12 * Even when merging it always takes the source from the index, and
13 * uses the working tree as a "branch" for a 3-way merge.
15 #include <ctype.h>
17 #include "cache.h"
19 // We default to the merge behaviour, since that's what most people would
20 // expect.
22 // --check turns on checking that the working tree matches the
23 // files that are being modified, but doesn't apply the patch
24 // --stat does just a diffstat, and doesn't actually apply
25 // --show-files shows the directory changes
27 static int merge_patch = 1;
28 static int diffstat = 0;
29 static int check = 0;
30 static int apply = 1;
31 static int show_files = 0;
32 static const char apply_usage[] = "git-apply [--stat] [--check] [--show-files] <patch>";
35 * For "diff-stat" like behaviour, we keep track of the biggest change
36 * we've seen, and the longest filename. That allows us to do simple
37 * scaling.
39 static int max_change, max_len;
42 * Various "current state", notably line numbers and what
43 * file (and how) we're patching right now.. The "is_xxxx"
44 * things are flags, where -1 means "don't know yet".
46 static int linenr = 1;
48 struct fragment {
49 unsigned long oldpos, oldlines;
50 unsigned long newpos, newlines;
51 const char *patch;
52 int size;
53 struct fragment *next;
56 struct patch {
57 char *new_name, *old_name, *def_name;
58 unsigned int old_mode, new_mode;
59 int is_rename, is_copy, is_new, is_delete;
60 int lines_added, lines_deleted;
61 struct fragment *fragments;
62 struct patch *next;
65 #define CHUNKSIZE (8192)
66 #define SLOP (16)
68 static void *read_patch_file(int fd, unsigned long *sizep)
70 unsigned long size = 0, alloc = CHUNKSIZE;
71 void *buffer = xmalloc(alloc);
73 for (;;) {
74 int nr = alloc - size;
75 if (nr < 1024) {
76 alloc += CHUNKSIZE;
77 buffer = xrealloc(buffer, alloc);
78 nr = alloc - size;
80 nr = read(fd, buffer + size, nr);
81 if (!nr)
82 break;
83 if (nr < 0) {
84 if (errno == EAGAIN)
85 continue;
86 die("git-apply: read returned %s", strerror(errno));
88 size += nr;
90 *sizep = size;
93 * Make sure that we have some slop in the buffer
94 * so that we can do speculative "memcmp" etc, and
95 * see to it that it is NUL-filled.
97 if (alloc < size + SLOP)
98 buffer = xrealloc(buffer, size + SLOP);
99 memset(buffer + size, 0, SLOP);
100 return buffer;
103 static unsigned long linelen(char *buffer, unsigned long size)
105 unsigned long len = 0;
106 while (size--) {
107 len++;
108 if (*buffer++ == '\n')
109 break;
111 return len;
114 static int is_dev_null(const char *str)
116 return !memcmp("/dev/null", str, 9) && isspace(str[9]);
119 #define TERM_SPACE 1
120 #define TERM_TAB 2
122 static int name_terminate(const char *name, int namelen, int c, int terminate)
124 if (c == ' ' && !(terminate & TERM_SPACE))
125 return 0;
126 if (c == '\t' && !(terminate & TERM_TAB))
127 return 0;
129 return 1;
132 static char * find_name(const char *line, char *def, int p_value, int terminate)
134 int len;
135 const char *start = line;
136 char *name;
138 for (;;) {
139 char c = *line;
141 if (isspace(c)) {
142 if (c == '\n')
143 break;
144 if (name_terminate(start, line-start, c, terminate))
145 break;
147 line++;
148 if (c == '/' && !--p_value)
149 start = line;
151 if (!start)
152 return def;
153 len = line - start;
154 if (!len)
155 return def;
158 * Generally we prefer the shorter name, especially
159 * if the other one is just a variation of that with
160 * something else tacked on to the end (ie "file.orig"
161 * or "file~").
163 if (def) {
164 int deflen = strlen(def);
165 if (deflen < len && !strncmp(start, def, deflen))
166 return def;
169 name = xmalloc(len + 1);
170 memcpy(name, start, len);
171 name[len] = 0;
172 free(def);
173 return name;
177 * Get the name etc info from the --/+++ lines of a traditional patch header
179 * NOTE! This hardcodes "-p1" behaviour in filename detection.
181 * FIXME! The end-of-filename heuristics are kind of screwy. For existing
182 * files, we can happily check the index for a match, but for creating a
183 * new file we should try to match whatever "patch" does. I have no idea.
185 static void parse_traditional_patch(const char *first, const char *second, struct patch *patch)
187 int p_value = 1;
188 char *name;
190 first += 4; // skip "--- "
191 second += 4; // skip "+++ "
192 if (is_dev_null(first)) {
193 patch->is_new = 1;
194 patch->is_delete = 0;
195 name = find_name(second, NULL, p_value, TERM_SPACE | TERM_TAB);
196 patch->new_name = name;
197 } else if (is_dev_null(second)) {
198 patch->is_new = 0;
199 patch->is_delete = 1;
200 name = find_name(first, NULL, p_value, TERM_SPACE | TERM_TAB);
201 patch->old_name = name;
202 } else {
203 name = find_name(first, NULL, p_value, TERM_SPACE | TERM_TAB);
204 name = find_name(second, name, p_value, TERM_SPACE | TERM_TAB);
205 patch->old_name = patch->new_name = name;
207 if (!name)
208 die("unable to find filename in patch at line %d", linenr);
211 static int gitdiff_hdrend(const char *line, struct patch *patch)
213 return -1;
217 * We're anal about diff header consistency, to make
218 * sure that we don't end up having strange ambiguous
219 * patches floating around.
221 * As a result, gitdiff_{old|new}name() will check
222 * their names against any previous information, just
223 * to make sure..
225 static char *gitdiff_verify_name(const char *line, int isnull, char *orig_name, const char *oldnew)
227 int len;
228 const char *name;
230 if (!orig_name && !isnull)
231 return find_name(line, NULL, 1, 0);
233 name = "/dev/null";
234 len = 9;
235 if (orig_name) {
236 name = orig_name;
237 len = strlen(name);
238 if (isnull)
239 die("git-apply: bad git-diff - expected /dev/null, got %s on line %d", name, linenr);
242 if (*name == '/')
243 goto absolute_path;
245 for (;;) {
246 char c = *line++;
247 if (c == '\n')
248 break;
249 if (c != '/')
250 continue;
251 absolute_path:
252 if (memcmp(line, name, len) || line[len] != '\n')
253 break;
254 return orig_name;
256 die("git-apply: bad git-diff - inconsistent %s filename on line %d", oldnew, linenr);
257 return NULL;
260 static int gitdiff_oldname(const char *line, struct patch *patch)
262 patch->old_name = gitdiff_verify_name(line, patch->is_new, patch->old_name, "old");
263 return 0;
266 static int gitdiff_newname(const char *line, struct patch *patch)
268 patch->new_name = gitdiff_verify_name(line, patch->is_delete, patch->new_name, "new");
269 return 0;
272 static int gitdiff_oldmode(const char *line, struct patch *patch)
274 patch->old_mode = strtoul(line, NULL, 8);
275 return 0;
278 static int gitdiff_newmode(const char *line, struct patch *patch)
280 patch->new_mode = strtoul(line, NULL, 8);
281 return 0;
284 static int gitdiff_delete(const char *line, struct patch *patch)
286 patch->is_delete = 1;
287 patch->old_name = patch->def_name;
288 return gitdiff_oldmode(line, patch);
291 static int gitdiff_newfile(const char *line, struct patch *patch)
293 patch->is_new = 1;
294 patch->new_name = patch->def_name;
295 return gitdiff_newmode(line, patch);
298 static int gitdiff_copysrc(const char *line, struct patch *patch)
300 patch->is_copy = 1;
301 patch->old_name = find_name(line, NULL, 0, 0);
302 return 0;
305 static int gitdiff_copydst(const char *line, struct patch *patch)
307 patch->is_copy = 1;
308 patch->new_name = find_name(line, NULL, 0, 0);
309 return 0;
312 static int gitdiff_renamesrc(const char *line, struct patch *patch)
314 patch->is_rename = 1;
315 patch->old_name = find_name(line, NULL, 0, 0);
316 return 0;
319 static int gitdiff_renamedst(const char *line, struct patch *patch)
321 patch->is_rename = 1;
322 patch->new_name = find_name(line, NULL, 0, 0);
323 return 0;
326 static int gitdiff_similarity(const char *line, struct patch *patch)
328 return 0;
331 static int gitdiff_dissimilarity(const char *line, struct patch *patch)
333 return 0;
337 * This is normal for a diff that doesn't change anything: we'll fall through
338 * into the next diff. Tell the parser to break out.
340 static int gitdiff_unrecognized(const char *line, struct patch *patch)
342 return -1;
345 static char *git_header_name(char *line)
347 int len;
348 char *name, *second;
351 * Find the first '/'
353 name = line;
354 for (;;) {
355 char c = *name++;
356 if (c == '\n')
357 return NULL;
358 if (c == '/')
359 break;
363 * We don't accept absolute paths (/dev/null) as possibly valid
365 if (name == line+1)
366 return NULL;
369 * Accept a name only if it shows up twice, exactly the same
370 * form.
372 for (len = 0 ; ; len++) {
373 char c = name[len];
375 switch (c) {
376 default:
377 continue;
378 case '\n':
379 break;
380 case '\t': case ' ':
381 second = name+len;
382 for (;;) {
383 char c = *second++;
384 if (c == '\n')
385 return NULL;
386 if (c == '/')
387 break;
389 if (second[len] == '\n' && !memcmp(name, second, len)) {
390 char *ret = xmalloc(len + 1);
391 memcpy(ret, name, len);
392 ret[len] = 0;
393 return ret;
397 return NULL;
400 /* Verify that we recognize the lines following a git header */
401 static int parse_git_header(char *line, int len, unsigned int size, struct patch *patch)
403 unsigned long offset;
405 /* A git diff has explicit new/delete information, so we don't guess */
406 patch->is_new = 0;
407 patch->is_delete = 0;
410 * Some things may not have the old name in the
411 * rest of the headers anywhere (pure mode changes,
412 * or removing or adding empty files), so we get
413 * the default name from the header.
415 patch->def_name = git_header_name(line + strlen("diff --git "));
417 line += len;
418 size -= len;
419 linenr++;
420 for (offset = len ; size > 0 ; offset += len, size -= len, line += len, linenr++) {
421 static const struct opentry {
422 const char *str;
423 int (*fn)(const char *, struct patch *);
424 } optable[] = {
425 { "@@ -", gitdiff_hdrend },
426 { "--- ", gitdiff_oldname },
427 { "+++ ", gitdiff_newname },
428 { "old mode ", gitdiff_oldmode },
429 { "new mode ", gitdiff_newmode },
430 { "deleted file mode ", gitdiff_delete },
431 { "new file mode ", gitdiff_newfile },
432 { "copy from ", gitdiff_copysrc },
433 { "copy to ", gitdiff_copydst },
434 { "rename from ", gitdiff_renamesrc },
435 { "rename to ", gitdiff_renamedst },
436 { "similarity index ", gitdiff_similarity },
437 { "dissimilarity index ", gitdiff_dissimilarity },
438 { "", gitdiff_unrecognized },
440 int i;
442 len = linelen(line, size);
443 if (!len || line[len-1] != '\n')
444 break;
445 for (i = 0; i < sizeof(optable) / sizeof(optable[0]); i++) {
446 const struct opentry *p = optable + i;
447 int oplen = strlen(p->str);
448 if (len < oplen || memcmp(p->str, line, oplen))
449 continue;
450 if (p->fn(line + oplen, patch) < 0)
451 return offset;
452 break;
456 return offset;
459 static int parse_num(const char *line, unsigned long *p)
461 char *ptr;
463 if (!isdigit(*line))
464 return 0;
465 *p = strtoul(line, &ptr, 10);
466 return ptr - line;
469 static int parse_range(const char *line, int len, int offset, const char *expect,
470 unsigned long *p1, unsigned long *p2)
472 int digits, ex;
474 if (offset < 0 || offset >= len)
475 return -1;
476 line += offset;
477 len -= offset;
479 digits = parse_num(line, p1);
480 if (!digits)
481 return -1;
483 offset += digits;
484 line += digits;
485 len -= digits;
487 *p2 = *p1;
488 if (*line == ',') {
489 digits = parse_num(line+1, p2);
490 if (!digits)
491 return -1;
493 offset += digits+1;
494 line += digits+1;
495 len -= digits+1;
498 ex = strlen(expect);
499 if (ex > len)
500 return -1;
501 if (memcmp(line, expect, ex))
502 return -1;
504 return offset + ex;
508 * Parse a unified diff fragment header of the
509 * form "@@ -a,b +c,d @@"
511 static int parse_fragment_header(char *line, int len, struct fragment *fragment)
513 int offset;
515 if (!len || line[len-1] != '\n')
516 return -1;
518 /* Figure out the number of lines in a fragment */
519 offset = parse_range(line, len, 4, " +", &fragment->oldpos, &fragment->oldlines);
520 offset = parse_range(line, len, offset, " @@", &fragment->newpos, &fragment->newlines);
522 return offset;
525 static int find_header(char *line, unsigned long size, int *hdrsize, struct patch *patch)
527 unsigned long offset, len;
529 patch->is_rename = patch->is_copy = 0;
530 patch->is_new = patch->is_delete = -1;
531 patch->old_mode = patch->new_mode = 0;
532 patch->old_name = patch->new_name = NULL;
533 for (offset = 0; size > 0; offset += len, size -= len, line += len, linenr++) {
534 unsigned long nextlen;
536 len = linelen(line, size);
537 if (!len)
538 break;
540 /* Testing this early allows us to take a few shortcuts.. */
541 if (len < 6)
542 continue;
545 * Make sure we don't find any unconnected patch fragmants.
546 * That's a sign that we didn't find a header, and that a
547 * patch has become corrupted/broken up.
549 if (!memcmp("@@ -", line, 4)) {
550 struct fragment dummy;
551 if (parse_fragment_header(line, len, &dummy) < 0)
552 continue;
553 error("patch fragment without header at line %d: %.*s", linenr, len-1, line);
556 if (size < len + 6)
557 break;
560 * Git patch? It might not have a real patch, just a rename
561 * or mode change, so we handle that specially
563 if (!memcmp("diff --git ", line, 11)) {
564 int git_hdr_len = parse_git_header(line, len, size, patch);
565 if (git_hdr_len < 0)
566 continue;
567 if (!patch->old_name && !patch->new_name)
568 die("git diff header lacks filename information");
569 *hdrsize = git_hdr_len;
570 return offset;
573 /** --- followed by +++ ? */
574 if (memcmp("--- ", line, 4) || memcmp("+++ ", line + len, 4))
575 continue;
578 * We only accept unified patches, so we want it to
579 * at least have "@@ -a,b +c,d @@\n", which is 14 chars
580 * minimum
582 nextlen = linelen(line + len, size - len);
583 if (size < nextlen + 14 || memcmp("@@ -", line + len + nextlen, 4))
584 continue;
586 /* Ok, we'll consider it a patch */
587 parse_traditional_patch(line, line+len, patch);
588 *hdrsize = len + nextlen;
589 linenr += 2;
590 return offset;
592 return -1;
596 * Parse a unified diff. Note that this really needs
597 * to parse each fragment separately, since the only
598 * way to know the difference between a "---" that is
599 * part of a patch, and a "---" that starts the next
600 * patch is to look at the line counts..
602 static int parse_fragment(char *line, unsigned long size, struct patch *patch, struct fragment *fragment)
604 int added, deleted;
605 int len = linelen(line, size), offset;
606 unsigned long pos[4], oldlines, newlines;
608 offset = parse_fragment_header(line, len, fragment);
609 if (offset < 0)
610 return -1;
611 oldlines = fragment->oldlines;
612 newlines = fragment->newlines;
614 if (patch->is_new < 0 && (pos[0] || oldlines))
615 patch->is_new = 0;
616 if (patch->is_delete < 0 && (pos[1] || newlines))
617 patch->is_delete = 0;
619 /* Parse the thing.. */
620 line += len;
621 size -= len;
622 linenr++;
623 added = deleted = 0;
624 for (offset = len; size > 0; offset += len, size -= len, line += len, linenr++) {
625 if (!oldlines && !newlines)
626 break;
627 len = linelen(line, size);
628 if (!len || line[len-1] != '\n')
629 return -1;
630 switch (*line) {
631 default:
632 return -1;
633 case ' ':
634 oldlines--;
635 newlines--;
636 break;
637 case '-':
638 deleted++;
639 oldlines--;
640 break;
641 case '+':
642 added++;
643 newlines--;
644 break;
645 /* We allow "\ No newline at end of file" */
646 case '\\':
647 break;
650 patch->lines_added += added;
651 patch->lines_deleted += deleted;
652 return offset;
655 static int parse_single_patch(char *line, unsigned long size, struct patch *patch)
657 unsigned long offset = 0;
658 struct fragment **fragp = &patch->fragments;
660 while (size > 4 && !memcmp(line, "@@ -", 4)) {
661 struct fragment *fragment;
662 int len;
664 fragment = xmalloc(sizeof(*fragment));
665 memset(fragment, 0, sizeof(*fragment));
666 len = parse_fragment(line, size, patch, fragment);
667 if (len <= 0)
668 die("corrupt patch at line %d", linenr);
670 fragment->patch = line;
671 fragment->size = len;
673 *fragp = fragment;
674 fragp = &fragment->next;
676 offset += len;
677 line += len;
678 size -= len;
680 return offset;
683 static int parse_chunk(char *buffer, unsigned long size, struct patch *patch)
685 int hdrsize, patchsize;
686 int offset = find_header(buffer, size, &hdrsize, patch);
688 if (offset < 0)
689 return offset;
691 patchsize = parse_single_patch(buffer + offset + hdrsize, size - offset - hdrsize, patch);
693 return offset + hdrsize + patchsize;
696 const char pluses[] = "++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++";
697 const char minuses[]= "----------------------------------------------------------------------";
699 static void show_stats(struct patch *patch)
701 char *name = patch->old_name;
702 int len, max, add, del;
704 if (!name)
705 name = patch->new_name;
708 * "scale" the filename
710 len = strlen(name);
711 max = max_len;
712 if (max > 50)
713 max = 50;
714 if (len > max)
715 name += len - max;
716 len = max;
719 * scale the add/delete
721 max = max_change;
722 if (max + len > 70)
723 max = 70 - len;
725 add = (patch->lines_added * max + max_change/2) / max_change;
726 del = (patch->lines_deleted * max + max_change/2) / max_change;
727 printf(" %-*s |%5d %.*s%.*s\n",
728 len, name, patch->lines_added + patch->lines_deleted,
729 add, pluses, del, minuses);
732 static int check_patch(struct patch *patch)
734 struct stat st;
735 const char *old_name = patch->old_name;
736 const char *new_name = patch->new_name;
738 if (old_name) {
739 int pos = cache_name_pos(old_name, strlen(old_name));
740 int changed;
742 if (pos < 0)
743 return error("%s: does not exist in index", old_name);
744 if (patch->is_new < 0)
745 patch->is_new = 0;
746 if (lstat(old_name, &st) < 0)
747 return error("%s: %s\n", strerror(errno));
748 changed = ce_match_stat(active_cache[pos], &st);
749 if (changed)
750 return error("%s: does not match index", old_name);
751 if (!patch->old_mode)
752 patch->old_mode = st.st_mode;
755 if (new_name && (patch->is_new | patch->is_rename | patch->is_copy)) {
756 if (cache_name_pos(new_name, strlen(new_name)) >= 0)
757 return error("%s: already exists in index", new_name);
758 if (!lstat(new_name, &st))
759 return error("%s: already exists in working directory", new_name);
760 if (errno != ENOENT)
761 return error("%s: %s", new_name, strerror(errno));
763 return 0;
766 static int check_patch_list(struct patch *patch)
768 int error = 0;
770 for (;patch ; patch = patch->next)
771 error |= check_patch(patch);
772 return error;
775 static void show_file(int c, unsigned int mode, const char *name)
777 printf("%c %o %s\n", c, mode, name);
780 static void show_file_list(struct patch *patch)
782 for (;patch ; patch = patch->next) {
783 if (patch->is_rename) {
784 show_file('-', patch->old_mode, patch->old_name);
785 show_file('+', patch->new_mode, patch->new_name);
786 continue;
788 if (patch->is_copy || patch->is_new) {
789 show_file('+', patch->new_mode, patch->new_name);
790 continue;
792 if (patch->is_delete) {
793 show_file('-', patch->old_mode, patch->old_name);
794 continue;
796 if (patch->old_mode && patch->new_mode && patch->old_mode != patch->new_mode) {
797 printf("M %o:%o %s\n", patch->old_mode, patch->new_mode, patch->old_name);
798 continue;
800 printf("M %o %s\n", patch->old_mode, patch->old_name);
804 static void stat_patch_list(struct patch *patch)
806 int files, adds, dels;
808 for (files = adds = dels = 0 ; patch ; patch = patch->next) {
809 files++;
810 adds += patch->lines_added;
811 dels += patch->lines_deleted;
812 show_stats(patch);
815 printf(" %d files changed, %d insertions(+), %d deletions(-)\n", files, adds, dels);
818 static void patch_stats(struct patch *patch)
820 int lines = patch->lines_added + patch->lines_deleted;
822 if (lines > max_change)
823 max_change = lines;
824 if (patch->old_name) {
825 int len = strlen(patch->old_name);
826 if (len > max_len)
827 max_len = len;
829 if (patch->new_name) {
830 int len = strlen(patch->new_name);
831 if (len > max_len)
832 max_len = len;
836 static int apply_patch(int fd)
838 unsigned long offset, size;
839 char *buffer = read_patch_file(fd, &size);
840 struct patch *list = NULL, **listp = &list;
842 if (!buffer)
843 return -1;
844 offset = 0;
845 while (size > 0) {
846 struct patch *patch;
847 int nr;
849 patch = xmalloc(sizeof(*patch));
850 memset(patch, 0, sizeof(*patch));
851 nr = parse_chunk(buffer + offset, size, patch);
852 if (nr < 0)
853 break;
854 patch_stats(patch);
855 *listp = patch;
856 listp = &patch->next;
857 offset += nr;
858 size -= nr;
861 if ((check || apply) && check_patch_list(list) < 0)
862 exit(1);
864 if (show_files)
865 show_file_list(list);
867 if (diffstat)
868 stat_patch_list(list);
870 free(buffer);
871 return 0;
874 int main(int argc, char **argv)
876 int i;
877 int read_stdin = 1;
879 if (read_cache() < 0)
880 die("unable to read index file");
882 for (i = 1; i < argc; i++) {
883 const char *arg = argv[i];
884 int fd;
886 if (!strcmp(arg, "-")) {
887 apply_patch(0);
888 read_stdin = 0;
889 continue;
891 if (!strcmp(arg, "--no-merge")) {
892 merge_patch = 0;
893 continue;
895 if (!strcmp(arg, "--stat")) {
896 apply = 0;
897 diffstat = 1;
898 continue;
900 if (!strcmp(arg, "--check")) {
901 apply = 0;
902 check = 1;
903 continue;
905 if (!strcmp(arg, "--show-files")) {
906 show_files = 1;
907 continue;
909 fd = open(arg, O_RDONLY);
910 if (fd < 0)
911 usage(apply_usage);
912 read_stdin = 0;
913 apply_patch(fd);
914 close(fd);
916 if (read_stdin)
917 apply_patch(0);
918 return 0;