Merge branch 'jc/grep'
[git/git.git] / apply.c
1 /*
2 * apply.c
3 *
4 * Copyright (C) Linus Torvalds, 2005
5 *
6 * This applies patches on top of some (arbitrary) version of the SCM.
7 *
8 */
9 #include <fnmatch.h>
10 #include "cache.h"
11 #include "quote.h"
12 #include "blob.h"
13 #include "delta.h"
14
15 // --check turns on checking that the working tree matches the
16 // files that are being modified, but doesn't apply the patch
17 // --stat does just a diffstat, and doesn't actually apply
18 // --numstat does numeric diffstat, and doesn't actually apply
19 // --index-info shows the old and new index info for paths if available.
20 //
21 static const char *prefix;
22 static int prefix_length = -1;
23 static int newfd = -1;
24
25 static int p_value = 1;
26 static int allow_binary_replacement = 0;
27 static int check_index = 0;
28 static int write_index = 0;
29 static int diffstat = 0;
30 static int numstat = 0;
31 static int summary = 0;
32 static int check = 0;
33 static int apply = 1;
34 static int no_add = 0;
35 static int show_index_info = 0;
36 static int line_termination = '\n';
37 static unsigned long p_context = -1;
38 static const char apply_usage[] =
39 "git-apply [--stat] [--numstat] [--summary] [--check] [--index] [--apply] [--no-add] [--index-info] [--allow-binary-replacement] [-z] [-pNUM] [-CNUM] [--whitespace=<nowarn|warn|error|error-all|strip>] <patch>...";
40
41 static enum whitespace_eol {
42 nowarn_whitespace,
43 warn_on_whitespace,
44 error_on_whitespace,
45 strip_whitespace,
46 } new_whitespace = warn_on_whitespace;
47 static int whitespace_error = 0;
48 static int squelch_whitespace_errors = 5;
49 static int applied_after_stripping = 0;
50 static const char *patch_input_file = NULL;
51
52 static void parse_whitespace_option(const char *option)
53 {
54 if (!option) {
55 new_whitespace = warn_on_whitespace;
56 return;
57 }
58 if (!strcmp(option, "warn")) {
59 new_whitespace = warn_on_whitespace;
60 return;
61 }
62 if (!strcmp(option, "nowarn")) {
63 new_whitespace = nowarn_whitespace;
64 return;
65 }
66 if (!strcmp(option, "error")) {
67 new_whitespace = error_on_whitespace;
68 return;
69 }
70 if (!strcmp(option, "error-all")) {
71 new_whitespace = error_on_whitespace;
72 squelch_whitespace_errors = 0;
73 return;
74 }
75 if (!strcmp(option, "strip")) {
76 new_whitespace = strip_whitespace;
77 return;
78 }
79 die("unrecognized whitespace option '%s'", option);
80 }
81
82 static void set_default_whitespace_mode(const char *whitespace_option)
83 {
84 if (!whitespace_option && !apply_default_whitespace) {
85 new_whitespace = (apply
86 ? warn_on_whitespace
87 : nowarn_whitespace);
88 }
89 }
90
91 /*
92 * For "diff-stat" like behaviour, we keep track of the biggest change
93 * we've seen, and the longest filename. That allows us to do simple
94 * scaling.
95 */
96 static int max_change, max_len;
97
98 /*
99 * Various "current state", notably line numbers and what
100 * file (and how) we're patching right now.. The "is_xxxx"
101 * things are flags, where -1 means "don't know yet".
102 */
103 static int linenr = 1;
104
105 struct fragment {
106 unsigned long leading, trailing;
107 unsigned long oldpos, oldlines;
108 unsigned long newpos, newlines;
109 const char *patch;
110 int size;
111 struct fragment *next;
112 };
113
114 struct patch {
115 char *new_name, *old_name, *def_name;
116 unsigned int old_mode, new_mode;
117 int is_rename, is_copy, is_new, is_delete, is_binary;
118 #define BINARY_DELTA_DEFLATED 1
119 #define BINARY_LITERAL_DEFLATED 2
120 unsigned long deflate_origlen;
121 int lines_added, lines_deleted;
122 int score;
123 struct fragment *fragments;
124 char *result;
125 unsigned long resultsize;
126 char old_sha1_prefix[41];
127 char new_sha1_prefix[41];
128 struct patch *next;
129 };
130
131 #define CHUNKSIZE (8192)
132 #define SLOP (16)
133
134 static void *read_patch_file(int fd, unsigned long *sizep)
135 {
136 unsigned long size = 0, alloc = CHUNKSIZE;
137 void *buffer = xmalloc(alloc);
138
139 for (;;) {
140 int nr = alloc - size;
141 if (nr < 1024) {
142 alloc += CHUNKSIZE;
143 buffer = xrealloc(buffer, alloc);
144 nr = alloc - size;
145 }
146 nr = xread(fd, buffer + size, nr);
147 if (!nr)
148 break;
149 if (nr < 0)
150 die("git-apply: read returned %s", strerror(errno));
151 size += nr;
152 }
153 *sizep = size;
154
155 /*
156 * Make sure that we have some slop in the buffer
157 * so that we can do speculative "memcmp" etc, and
158 * see to it that it is NUL-filled.
159 */
160 if (alloc < size + SLOP)
161 buffer = xrealloc(buffer, size + SLOP);
162 memset(buffer + size, 0, SLOP);
163 return buffer;
164 }
165
166 static unsigned long linelen(const char *buffer, unsigned long size)
167 {
168 unsigned long len = 0;
169 while (size--) {
170 len++;
171 if (*buffer++ == '\n')
172 break;
173 }
174 return len;
175 }
176
177 static int is_dev_null(const char *str)
178 {
179 return !memcmp("/dev/null", str, 9) && isspace(str[9]);
180 }
181
182 #define TERM_SPACE 1
183 #define TERM_TAB 2
184
185 static int name_terminate(const char *name, int namelen, int c, int terminate)
186 {
187 if (c == ' ' && !(terminate & TERM_SPACE))
188 return 0;
189 if (c == '\t' && !(terminate & TERM_TAB))
190 return 0;
191
192 return 1;
193 }
194
195 static char * find_name(const char *line, char *def, int p_value, int terminate)
196 {
197 int len;
198 const char *start = line;
199 char *name;
200
201 if (*line == '"') {
202 /* Proposed "new-style" GNU patch/diff format; see
203 * http://marc.theaimsgroup.com/?l=git&m=112927316408690&w=2
204 */
205 name = unquote_c_style(line, NULL);
206 if (name) {
207 char *cp = name;
208 while (p_value) {
209 cp = strchr(name, '/');
210 if (!cp)
211 break;
212 cp++;
213 p_value--;
214 }
215 if (cp) {
216 /* name can later be freed, so we need
217 * to memmove, not just return cp
218 */
219 memmove(name, cp, strlen(cp) + 1);
220 free(def);
221 return name;
222 }
223 else {
224 free(name);
225 name = NULL;
226 }
227 }
228 }
229
230 for (;;) {
231 char c = *line;
232
233 if (isspace(c)) {
234 if (c == '\n')
235 break;
236 if (name_terminate(start, line-start, c, terminate))
237 break;
238 }
239 line++;
240 if (c == '/' && !--p_value)
241 start = line;
242 }
243 if (!start)
244 return def;
245 len = line - start;
246 if (!len)
247 return def;
248
249 /*
250 * Generally we prefer the shorter name, especially
251 * if the other one is just a variation of that with
252 * something else tacked on to the end (ie "file.orig"
253 * or "file~").
254 */
255 if (def) {
256 int deflen = strlen(def);
257 if (deflen < len && !strncmp(start, def, deflen))
258 return def;
259 }
260
261 name = xmalloc(len + 1);
262 memcpy(name, start, len);
263 name[len] = 0;
264 free(def);
265 return name;
266 }
267
268 /*
269 * Get the name etc info from the --/+++ lines of a traditional patch header
270 *
271 * NOTE! This hardcodes "-p1" behaviour in filename detection.
272 *
273 * FIXME! The end-of-filename heuristics are kind of screwy. For existing
274 * files, we can happily check the index for a match, but for creating a
275 * new file we should try to match whatever "patch" does. I have no idea.
276 */
277 static void parse_traditional_patch(const char *first, const char *second, struct patch *patch)
278 {
279 char *name;
280
281 first += 4; // skip "--- "
282 second += 4; // skip "+++ "
283 if (is_dev_null(first)) {
284 patch->is_new = 1;
285 patch->is_delete = 0;
286 name = find_name(second, NULL, p_value, TERM_SPACE | TERM_TAB);
287 patch->new_name = name;
288 } else if (is_dev_null(second)) {
289 patch->is_new = 0;
290 patch->is_delete = 1;
291 name = find_name(first, NULL, p_value, TERM_SPACE | TERM_TAB);
292 patch->old_name = name;
293 } else {
294 name = find_name(first, NULL, p_value, TERM_SPACE | TERM_TAB);
295 name = find_name(second, name, p_value, TERM_SPACE | TERM_TAB);
296 patch->old_name = patch->new_name = name;
297 }
298 if (!name)
299 die("unable to find filename in patch at line %d", linenr);
300 }
301
302 static int gitdiff_hdrend(const char *line, struct patch *patch)
303 {
304 return -1;
305 }
306
307 /*
308 * We're anal about diff header consistency, to make
309 * sure that we don't end up having strange ambiguous
310 * patches floating around.
311 *
312 * As a result, gitdiff_{old|new}name() will check
313 * their names against any previous information, just
314 * to make sure..
315 */
316 static char *gitdiff_verify_name(const char *line, int isnull, char *orig_name, const char *oldnew)
317 {
318 if (!orig_name && !isnull)
319 return find_name(line, NULL, 1, 0);
320
321 if (orig_name) {
322 int len;
323 const char *name;
324 char *another;
325 name = orig_name;
326 len = strlen(name);
327 if (isnull)
328 die("git-apply: bad git-diff - expected /dev/null, got %s on line %d", name, linenr);
329 another = find_name(line, NULL, 1, 0);
330 if (!another || memcmp(another, name, len))
331 die("git-apply: bad git-diff - inconsistent %s filename on line %d", oldnew, linenr);
332 free(another);
333 return orig_name;
334 }
335 else {
336 /* expect "/dev/null" */
337 if (memcmp("/dev/null", line, 9) || line[9] != '\n')
338 die("git-apply: bad git-diff - expected /dev/null on line %d", linenr);
339 return NULL;
340 }
341 }
342
343 static int gitdiff_oldname(const char *line, struct patch *patch)
344 {
345 patch->old_name = gitdiff_verify_name(line, patch->is_new, patch->old_name, "old");
346 return 0;
347 }
348
349 static int gitdiff_newname(const char *line, struct patch *patch)
350 {
351 patch->new_name = gitdiff_verify_name(line, patch->is_delete, patch->new_name, "new");
352 return 0;
353 }
354
355 static int gitdiff_oldmode(const char *line, struct patch *patch)
356 {
357 patch->old_mode = strtoul(line, NULL, 8);
358 return 0;
359 }
360
361 static int gitdiff_newmode(const char *line, struct patch *patch)
362 {
363 patch->new_mode = strtoul(line, NULL, 8);
364 return 0;
365 }
366
367 static int gitdiff_delete(const char *line, struct patch *patch)
368 {
369 patch->is_delete = 1;
370 patch->old_name = patch->def_name;
371 return gitdiff_oldmode(line, patch);
372 }
373
374 static int gitdiff_newfile(const char *line, struct patch *patch)
375 {
376 patch->is_new = 1;
377 patch->new_name = patch->def_name;
378 return gitdiff_newmode(line, patch);
379 }
380
381 static int gitdiff_copysrc(const char *line, struct patch *patch)
382 {
383 patch->is_copy = 1;
384 patch->old_name = find_name(line, NULL, 0, 0);
385 return 0;
386 }
387
388 static int gitdiff_copydst(const char *line, struct patch *patch)
389 {
390 patch->is_copy = 1;
391 patch->new_name = find_name(line, NULL, 0, 0);
392 return 0;
393 }
394
395 static int gitdiff_renamesrc(const char *line, struct patch *patch)
396 {
397 patch->is_rename = 1;
398 patch->old_name = find_name(line, NULL, 0, 0);
399 return 0;
400 }
401
402 static int gitdiff_renamedst(const char *line, struct patch *patch)
403 {
404 patch->is_rename = 1;
405 patch->new_name = find_name(line, NULL, 0, 0);
406 return 0;
407 }
408
409 static int gitdiff_similarity(const char *line, struct patch *patch)
410 {
411 if ((patch->score = strtoul(line, NULL, 10)) == ULONG_MAX)
412 patch->score = 0;
413 return 0;
414 }
415
416 static int gitdiff_dissimilarity(const char *line, struct patch *patch)
417 {
418 if ((patch->score = strtoul(line, NULL, 10)) == ULONG_MAX)
419 patch->score = 0;
420 return 0;
421 }
422
423 static int gitdiff_index(const char *line, struct patch *patch)
424 {
425 /* index line is N hexadecimal, "..", N hexadecimal,
426 * and optional space with octal mode.
427 */
428 const char *ptr, *eol;
429 int len;
430
431 ptr = strchr(line, '.');
432 if (!ptr || ptr[1] != '.' || 40 < ptr - line)
433 return 0;
434 len = ptr - line;
435 memcpy(patch->old_sha1_prefix, line, len);
436 patch->old_sha1_prefix[len] = 0;
437
438 line = ptr + 2;
439 ptr = strchr(line, ' ');
440 eol = strchr(line, '\n');
441
442 if (!ptr || eol < ptr)
443 ptr = eol;
444 len = ptr - line;
445
446 if (40 < len)
447 return 0;
448 memcpy(patch->new_sha1_prefix, line, len);
449 patch->new_sha1_prefix[len] = 0;
450 if (*ptr == ' ')
451 patch->new_mode = patch->old_mode = strtoul(ptr+1, NULL, 8);
452 return 0;
453 }
454
455 /*
456 * This is normal for a diff that doesn't change anything: we'll fall through
457 * into the next diff. Tell the parser to break out.
458 */
459 static int gitdiff_unrecognized(const char *line, struct patch *patch)
460 {
461 return -1;
462 }
463
464 static const char *stop_at_slash(const char *line, int llen)
465 {
466 int i;
467
468 for (i = 0; i < llen; i++) {
469 int ch = line[i];
470 if (ch == '/')
471 return line + i;
472 }
473 return NULL;
474 }
475
476 /* This is to extract the same name that appears on "diff --git"
477 * line. We do not find and return anything if it is a rename
478 * patch, and it is OK because we will find the name elsewhere.
479 * We need to reliably find name only when it is mode-change only,
480 * creation or deletion of an empty file. In any of these cases,
481 * both sides are the same name under a/ and b/ respectively.
482 */
483 static char *git_header_name(char *line, int llen)
484 {
485 int len;
486 const char *name;
487 const char *second = NULL;
488
489 line += strlen("diff --git ");
490 llen -= strlen("diff --git ");
491
492 if (*line == '"') {
493 const char *cp;
494 char *first = unquote_c_style(line, &second);
495 if (!first)
496 return NULL;
497
498 /* advance to the first slash */
499 cp = stop_at_slash(first, strlen(first));
500 if (!cp || cp == first) {
501 /* we do not accept absolute paths */
502 free_first_and_fail:
503 free(first);
504 return NULL;
505 }
506 len = strlen(cp+1);
507 memmove(first, cp+1, len+1); /* including NUL */
508
509 /* second points at one past closing dq of name.
510 * find the second name.
511 */
512 while ((second < line + llen) && isspace(*second))
513 second++;
514
515 if (line + llen <= second)
516 goto free_first_and_fail;
517 if (*second == '"') {
518 char *sp = unquote_c_style(second, NULL);
519 if (!sp)
520 goto free_first_and_fail;
521 cp = stop_at_slash(sp, strlen(sp));
522 if (!cp || cp == sp) {
523 free_both_and_fail:
524 free(sp);
525 goto free_first_and_fail;
526 }
527 /* They must match, otherwise ignore */
528 if (strcmp(cp+1, first))
529 goto free_both_and_fail;
530 free(sp);
531 return first;
532 }
533
534 /* unquoted second */
535 cp = stop_at_slash(second, line + llen - second);
536 if (!cp || cp == second)
537 goto free_first_and_fail;
538 cp++;
539 if (line + llen - cp != len + 1 ||
540 memcmp(first, cp, len))
541 goto free_first_and_fail;
542 return first;
543 }
544
545 /* unquoted first name */
546 name = stop_at_slash(line, llen);
547 if (!name || name == line)
548 return NULL;
549
550 name++;
551
552 /* since the first name is unquoted, a dq if exists must be
553 * the beginning of the second name.
554 */
555 for (second = name; second < line + llen; second++) {
556 if (*second == '"') {
557 const char *cp = second;
558 const char *np;
559 char *sp = unquote_c_style(second, NULL);
560
561 if (!sp)
562 return NULL;
563 np = stop_at_slash(sp, strlen(sp));
564 if (!np || np == sp) {
565 free_second_and_fail:
566 free(sp);
567 return NULL;
568 }
569 np++;
570 len = strlen(np);
571 if (len < cp - name &&
572 !strncmp(np, name, len) &&
573 isspace(name[len])) {
574 /* Good */
575 memmove(sp, np, len + 1);
576 return sp;
577 }
578 goto free_second_and_fail;
579 }
580 }
581
582 /*
583 * Accept a name only if it shows up twice, exactly the same
584 * form.
585 */
586 for (len = 0 ; ; len++) {
587 char c = name[len];
588
589 switch (c) {
590 default:
591 continue;
592 case '\n':
593 return NULL;
594 case '\t': case ' ':
595 second = name+len;
596 for (;;) {
597 char c = *second++;
598 if (c == '\n')
599 return NULL;
600 if (c == '/')
601 break;
602 }
603 if (second[len] == '\n' && !memcmp(name, second, len)) {
604 char *ret = xmalloc(len + 1);
605 memcpy(ret, name, len);
606 ret[len] = 0;
607 return ret;
608 }
609 }
610 }
611 return NULL;
612 }
613
614 /* Verify that we recognize the lines following a git header */
615 static int parse_git_header(char *line, int len, unsigned int size, struct patch *patch)
616 {
617 unsigned long offset;
618
619 /* A git diff has explicit new/delete information, so we don't guess */
620 patch->is_new = 0;
621 patch->is_delete = 0;
622
623 /*
624 * Some things may not have the old name in the
625 * rest of the headers anywhere (pure mode changes,
626 * or removing or adding empty files), so we get
627 * the default name from the header.
628 */
629 patch->def_name = git_header_name(line, len);
630
631 line += len;
632 size -= len;
633 linenr++;
634 for (offset = len ; size > 0 ; offset += len, size -= len, line += len, linenr++) {
635 static const struct opentry {
636 const char *str;
637 int (*fn)(const char *, struct patch *);
638 } optable[] = {
639 { "@@ -", gitdiff_hdrend },
640 { "--- ", gitdiff_oldname },
641 { "+++ ", gitdiff_newname },
642 { "old mode ", gitdiff_oldmode },
643 { "new mode ", gitdiff_newmode },
644 { "deleted file mode ", gitdiff_delete },
645 { "new file mode ", gitdiff_newfile },
646 { "copy from ", gitdiff_copysrc },
647 { "copy to ", gitdiff_copydst },
648 { "rename old ", gitdiff_renamesrc },
649 { "rename new ", gitdiff_renamedst },
650 { "rename from ", gitdiff_renamesrc },
651 { "rename to ", gitdiff_renamedst },
652 { "similarity index ", gitdiff_similarity },
653 { "dissimilarity index ", gitdiff_dissimilarity },
654 { "index ", gitdiff_index },
655 { "", gitdiff_unrecognized },
656 };
657 int i;
658
659 len = linelen(line, size);
660 if (!len || line[len-1] != '\n')
661 break;
662 for (i = 0; i < ARRAY_SIZE(optable); i++) {
663 const struct opentry *p = optable + i;
664 int oplen = strlen(p->str);
665 if (len < oplen || memcmp(p->str, line, oplen))
666 continue;
667 if (p->fn(line + oplen, patch) < 0)
668 return offset;
669 break;
670 }
671 }
672
673 return offset;
674 }
675
676 static int parse_num(const char *line, unsigned long *p)
677 {
678 char *ptr;
679
680 if (!isdigit(*line))
681 return 0;
682 *p = strtoul(line, &ptr, 10);
683 return ptr - line;
684 }
685
686 static int parse_range(const char *line, int len, int offset, const char *expect,
687 unsigned long *p1, unsigned long *p2)
688 {
689 int digits, ex;
690
691 if (offset < 0 || offset >= len)
692 return -1;
693 line += offset;
694 len -= offset;
695
696 digits = parse_num(line, p1);
697 if (!digits)
698 return -1;
699
700 offset += digits;
701 line += digits;
702 len -= digits;
703
704 *p2 = 1;
705 if (*line == ',') {
706 digits = parse_num(line+1, p2);
707 if (!digits)
708 return -1;
709
710 offset += digits+1;
711 line += digits+1;
712 len -= digits+1;
713 }
714
715 ex = strlen(expect);
716 if (ex > len)
717 return -1;
718 if (memcmp(line, expect, ex))
719 return -1;
720
721 return offset + ex;
722 }
723
724 /*
725 * Parse a unified diff fragment header of the
726 * form "@@ -a,b +c,d @@"
727 */
728 static int parse_fragment_header(char *line, int len, struct fragment *fragment)
729 {
730 int offset;
731
732 if (!len || line[len-1] != '\n')
733 return -1;
734
735 /* Figure out the number of lines in a fragment */
736 offset = parse_range(line, len, 4, " +", &fragment->oldpos, &fragment->oldlines);
737 offset = parse_range(line, len, offset, " @@", &fragment->newpos, &fragment->newlines);
738
739 return offset;
740 }
741
742 static int find_header(char *line, unsigned long size, int *hdrsize, struct patch *patch)
743 {
744 unsigned long offset, len;
745
746 patch->is_rename = patch->is_copy = 0;
747 patch->is_new = patch->is_delete = -1;
748 patch->old_mode = patch->new_mode = 0;
749 patch->old_name = patch->new_name = NULL;
750 for (offset = 0; size > 0; offset += len, size -= len, line += len, linenr++) {
751 unsigned long nextlen;
752
753 len = linelen(line, size);
754 if (!len)
755 break;
756
757 /* Testing this early allows us to take a few shortcuts.. */
758 if (len < 6)
759 continue;
760
761 /*
762 * Make sure we don't find any unconnected patch fragmants.
763 * That's a sign that we didn't find a header, and that a
764 * patch has become corrupted/broken up.
765 */
766 if (!memcmp("@@ -", line, 4)) {
767 struct fragment dummy;
768 if (parse_fragment_header(line, len, &dummy) < 0)
769 continue;
770 error("patch fragment without header at line %d: %.*s", linenr, (int)len-1, line);
771 }
772
773 if (size < len + 6)
774 break;
775
776 /*
777 * Git patch? It might not have a real patch, just a rename
778 * or mode change, so we handle that specially
779 */
780 if (!memcmp("diff --git ", line, 11)) {
781 int git_hdr_len = parse_git_header(line, len, size, patch);
782 if (git_hdr_len <= len)
783 continue;
784 if (!patch->old_name && !patch->new_name) {
785 if (!patch->def_name)
786 die("git diff header lacks filename information (line %d)", linenr);
787 patch->old_name = patch->new_name = patch->def_name;
788 }
789 *hdrsize = git_hdr_len;
790 return offset;
791 }
792
793 /** --- followed by +++ ? */
794 if (memcmp("--- ", line, 4) || memcmp("+++ ", line + len, 4))
795 continue;
796
797 /*
798 * We only accept unified patches, so we want it to
799 * at least have "@@ -a,b +c,d @@\n", which is 14 chars
800 * minimum
801 */
802 nextlen = linelen(line + len, size - len);
803 if (size < nextlen + 14 || memcmp("@@ -", line + len + nextlen, 4))
804 continue;
805
806 /* Ok, we'll consider it a patch */
807 parse_traditional_patch(line, line+len, patch);
808 *hdrsize = len + nextlen;
809 linenr += 2;
810 return offset;
811 }
812 return -1;
813 }
814
815 /*
816 * Parse a unified diff. Note that this really needs
817 * to parse each fragment separately, since the only
818 * way to know the difference between a "---" that is
819 * part of a patch, and a "---" that starts the next
820 * patch is to look at the line counts..
821 */
822 static int parse_fragment(char *line, unsigned long size, struct patch *patch, struct fragment *fragment)
823 {
824 int added, deleted;
825 int len = linelen(line, size), offset;
826 unsigned long oldlines, newlines;
827 unsigned long leading, trailing;
828
829 offset = parse_fragment_header(line, len, fragment);
830 if (offset < 0)
831 return -1;
832 oldlines = fragment->oldlines;
833 newlines = fragment->newlines;
834 leading = 0;
835 trailing = 0;
836
837 if (patch->is_new < 0) {
838 patch->is_new = !oldlines;
839 if (!oldlines)
840 patch->old_name = NULL;
841 }
842 if (patch->is_delete < 0) {
843 patch->is_delete = !newlines;
844 if (!newlines)
845 patch->new_name = NULL;
846 }
847
848 if (patch->is_new && oldlines)
849 return error("new file depends on old contents");
850 if (patch->is_delete != !newlines) {
851 if (newlines)
852 return error("deleted file still has contents");
853 fprintf(stderr, "** warning: file %s becomes empty but is not deleted\n", patch->new_name);
854 }
855
856 /* Parse the thing.. */
857 line += len;
858 size -= len;
859 linenr++;
860 added = deleted = 0;
861 for (offset = len; size > 0; offset += len, size -= len, line += len, linenr++) {
862 if (!oldlines && !newlines)
863 break;
864 len = linelen(line, size);
865 if (!len || line[len-1] != '\n')
866 return -1;
867 switch (*line) {
868 default:
869 return -1;
870 case ' ':
871 oldlines--;
872 newlines--;
873 if (!deleted && !added)
874 leading++;
875 trailing++;
876 break;
877 case '-':
878 deleted++;
879 oldlines--;
880 trailing = 0;
881 break;
882 case '+':
883 /*
884 * We know len is at least two, since we have a '+' and
885 * we checked that the last character was a '\n' above.
886 * That is, an addition of an empty line would check
887 * the '+' here. Sneaky...
888 */
889 if ((new_whitespace != nowarn_whitespace) &&
890 isspace(line[len-2])) {
891 whitespace_error++;
892 if (squelch_whitespace_errors &&
893 squelch_whitespace_errors <
894 whitespace_error)
895 ;
896 else {
897 fprintf(stderr, "Adds trailing whitespace.\n%s:%d:%.*s\n",
898 patch_input_file,
899 linenr, len-2, line+1);
900 }
901 }
902 added++;
903 newlines--;
904 trailing = 0;
905 break;
906
907 /* We allow "\ No newline at end of file". Depending
908 * on locale settings when the patch was produced we
909 * don't know what this line looks like. The only
910 * thing we do know is that it begins with "\ ".
911 * Checking for 12 is just for sanity check -- any
912 * l10n of "\ No newline..." is at least that long.
913 */
914 case '\\':
915 if (len < 12 || memcmp(line, "\\ ", 2))
916 return -1;
917 break;
918 }
919 }
920 if (oldlines || newlines)
921 return -1;
922 fragment->leading = leading;
923 fragment->trailing = trailing;
924
925 /* If a fragment ends with an incomplete line, we failed to include
926 * it in the above loop because we hit oldlines == newlines == 0
927 * before seeing it.
928 */
929 if (12 < size && !memcmp(line, "\\ ", 2))
930 offset += linelen(line, size);
931
932 patch->lines_added += added;
933 patch->lines_deleted += deleted;
934 return offset;
935 }
936
937 static int parse_single_patch(char *line, unsigned long size, struct patch *patch)
938 {
939 unsigned long offset = 0;
940 struct fragment **fragp = &patch->fragments;
941
942 while (size > 4 && !memcmp(line, "@@ -", 4)) {
943 struct fragment *fragment;
944 int len;
945
946 fragment = xcalloc(1, sizeof(*fragment));
947 len = parse_fragment(line, size, patch, fragment);
948 if (len <= 0)
949 die("corrupt patch at line %d", linenr);
950
951 fragment->patch = line;
952 fragment->size = len;
953
954 *fragp = fragment;
955 fragp = &fragment->next;
956
957 offset += len;
958 line += len;
959 size -= len;
960 }
961 return offset;
962 }
963
964 static inline int metadata_changes(struct patch *patch)
965 {
966 return patch->is_rename > 0 ||
967 patch->is_copy > 0 ||
968 patch->is_new > 0 ||
969 patch->is_delete ||
970 (patch->old_mode && patch->new_mode &&
971 patch->old_mode != patch->new_mode);
972 }
973
974 static int parse_binary(char *buffer, unsigned long size, struct patch *patch)
975 {
976 /* We have read "GIT binary patch\n"; what follows is a line
977 * that says the patch method (currently, either "deflated
978 * literal" or "deflated delta") and the length of data before
979 * deflating; a sequence of 'length-byte' followed by base-85
980 * encoded data follows.
981 *
982 * Each 5-byte sequence of base-85 encodes up to 4 bytes,
983 * and we would limit the patch line to 66 characters,
984 * so one line can fit up to 13 groups that would decode
985 * to 52 bytes max. The length byte 'A'-'Z' corresponds
986 * to 1-26 bytes, and 'a'-'z' corresponds to 27-52 bytes.
987 * The end of binary is signalled with an empty line.
988 */
989 int llen, used;
990 struct fragment *fragment;
991 char *data = NULL;
992
993 patch->fragments = fragment = xcalloc(1, sizeof(*fragment));
994
995 /* Grab the type of patch */
996 llen = linelen(buffer, size);
997 used = llen;
998 linenr++;
999
1000 if (!strncmp(buffer, "delta ", 6)) {
1001 patch->is_binary = BINARY_DELTA_DEFLATED;
1002 patch->deflate_origlen = strtoul(buffer + 6, NULL, 10);
1003 }
1004 else if (!strncmp(buffer, "literal ", 8)) {
1005 patch->is_binary = BINARY_LITERAL_DEFLATED;
1006 patch->deflate_origlen = strtoul(buffer + 8, NULL, 10);
1007 }
1008 else
1009 return error("unrecognized binary patch at line %d: %.*s",
1010 linenr-1, llen-1, buffer);
1011 buffer += llen;
1012 while (1) {
1013 int byte_length, max_byte_length, newsize;
1014 llen = linelen(buffer, size);
1015 used += llen;
1016 linenr++;
1017 if (llen == 1)
1018 break;
1019 /* Minimum line is "A00000\n" which is 7-byte long,
1020 * and the line length must be multiple of 5 plus 2.
1021 */
1022 if ((llen < 7) || (llen-2) % 5)
1023 goto corrupt;
1024 max_byte_length = (llen - 2) / 5 * 4;
1025 byte_length = *buffer;
1026 if ('A' <= byte_length && byte_length <= 'Z')
1027 byte_length = byte_length - 'A' + 1;
1028 else if ('a' <= byte_length && byte_length <= 'z')
1029 byte_length = byte_length - 'a' + 27;
1030 else
1031 goto corrupt;
1032 /* if the input length was not multiple of 4, we would
1033 * have filler at the end but the filler should never
1034 * exceed 3 bytes
1035 */
1036 if (max_byte_length < byte_length ||
1037 byte_length <= max_byte_length - 4)
1038 goto corrupt;
1039 newsize = fragment->size + byte_length;
1040 data = xrealloc(data, newsize);
1041 if (decode_85(data + fragment->size,
1042 buffer + 1,
1043 byte_length))
1044 goto corrupt;
1045 fragment->size = newsize;
1046 buffer += llen;
1047 size -= llen;
1048 }
1049 fragment->patch = data;
1050 return used;
1051 corrupt:
1052 return error("corrupt binary patch at line %d: %.*s",
1053 linenr-1, llen-1, buffer);
1054 }
1055
1056 static int parse_chunk(char *buffer, unsigned long size, struct patch *patch)
1057 {
1058 int hdrsize, patchsize;
1059 int offset = find_header(buffer, size, &hdrsize, patch);
1060
1061 if (offset < 0)
1062 return offset;
1063
1064 patchsize = parse_single_patch(buffer + offset + hdrsize, size - offset - hdrsize, patch);
1065
1066 if (!patchsize) {
1067 static const char *binhdr[] = {
1068 "Binary files ",
1069 "Files ",
1070 NULL,
1071 };
1072 static const char git_binary[] = "GIT binary patch\n";
1073 int i;
1074 int hd = hdrsize + offset;
1075 unsigned long llen = linelen(buffer + hd, size - hd);
1076
1077 if (llen == sizeof(git_binary) - 1 &&
1078 !memcmp(git_binary, buffer + hd, llen)) {
1079 int used;
1080 linenr++;
1081 used = parse_binary(buffer + hd + llen,
1082 size - hd - llen, patch);
1083 if (used)
1084 patchsize = used + llen;
1085 else
1086 patchsize = 0;
1087 }
1088 else if (!memcmp(" differ\n", buffer + hd + llen - 8, 8)) {
1089 for (i = 0; binhdr[i]; i++) {
1090 int len = strlen(binhdr[i]);
1091 if (len < size - hd &&
1092 !memcmp(binhdr[i], buffer + hd, len)) {
1093 linenr++;
1094 patch->is_binary = 1;
1095 patchsize = llen;
1096 break;
1097 }
1098 }
1099 }
1100
1101 /* Empty patch cannot be applied if:
1102 * - it is a binary patch and we do not do binary_replace, or
1103 * - text patch without metadata change
1104 */
1105 if ((apply || check) &&
1106 (patch->is_binary
1107 ? !allow_binary_replacement
1108 : !metadata_changes(patch)))
1109 die("patch with only garbage at line %d", linenr);
1110 }
1111
1112 return offset + hdrsize + patchsize;
1113 }
1114
1115 static const char pluses[] = "++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++";
1116 static const char minuses[]= "----------------------------------------------------------------------";
1117
1118 static void show_stats(struct patch *patch)
1119 {
1120 const char *prefix = "";
1121 char *name = patch->new_name;
1122 char *qname = NULL;
1123 int len, max, add, del, total;
1124
1125 if (!name)
1126 name = patch->old_name;
1127
1128 if (0 < (len = quote_c_style(name, NULL, NULL, 0))) {
1129 qname = xmalloc(len + 1);
1130 quote_c_style(name, qname, NULL, 0);
1131 name = qname;
1132 }
1133
1134 /*
1135 * "scale" the filename
1136 */
1137 len = strlen(name);
1138 max = max_len;
1139 if (max > 50)
1140 max = 50;
1141 if (len > max) {
1142 char *slash;
1143 prefix = "...";
1144 max -= 3;
1145 name += len - max;
1146 slash = strchr(name, '/');
1147 if (slash)
1148 name = slash;
1149 }
1150 len = max;
1151
1152 /*
1153 * scale the add/delete
1154 */
1155 max = max_change;
1156 if (max + len > 70)
1157 max = 70 - len;
1158
1159 add = patch->lines_added;
1160 del = patch->lines_deleted;
1161 total = add + del;
1162
1163 if (max_change > 0) {
1164 total = (total * max + max_change / 2) / max_change;
1165 add = (add * max + max_change / 2) / max_change;
1166 del = total - add;
1167 }
1168 if (patch->is_binary)
1169 printf(" %s%-*s | Bin\n", prefix, len, name);
1170 else
1171 printf(" %s%-*s |%5d %.*s%.*s\n", prefix,
1172 len, name, patch->lines_added + patch->lines_deleted,
1173 add, pluses, del, minuses);
1174 if (qname)
1175 free(qname);
1176 }
1177
1178 static int read_old_data(struct stat *st, const char *path, void *buf, unsigned long size)
1179 {
1180 int fd;
1181 unsigned long got;
1182
1183 switch (st->st_mode & S_IFMT) {
1184 case S_IFLNK:
1185 return readlink(path, buf, size);
1186 case S_IFREG:
1187 fd = open(path, O_RDONLY);
1188 if (fd < 0)
1189 return error("unable to open %s", path);
1190 got = 0;
1191 for (;;) {
1192 int ret = xread(fd, buf + got, size - got);
1193 if (ret <= 0)
1194 break;
1195 got += ret;
1196 }
1197 close(fd);
1198 return got;
1199
1200 default:
1201 return -1;
1202 }
1203 }
1204
1205 static int find_offset(const char *buf, unsigned long size, const char *fragment, unsigned long fragsize, int line, int *lines)
1206 {
1207 int i;
1208 unsigned long start, backwards, forwards;
1209
1210 if (fragsize > size)
1211 return -1;
1212
1213 start = 0;
1214 if (line > 1) {
1215 unsigned long offset = 0;
1216 i = line-1;
1217 while (offset + fragsize <= size) {
1218 if (buf[offset++] == '\n') {
1219 start = offset;
1220 if (!--i)
1221 break;
1222 }
1223 }
1224 }
1225
1226 /* Exact line number? */
1227 if (!memcmp(buf + start, fragment, fragsize))
1228 return start;
1229
1230 /*
1231 * There's probably some smart way to do this, but I'll leave
1232 * that to the smart and beautiful people. I'm simple and stupid.
1233 */
1234 backwards = start;
1235 forwards = start;
1236 for (i = 0; ; i++) {
1237 unsigned long try;
1238 int n;
1239
1240 /* "backward" */
1241 if (i & 1) {
1242 if (!backwards) {
1243 if (forwards + fragsize > size)
1244 break;
1245 continue;
1246 }
1247 do {
1248 --backwards;
1249 } while (backwards && buf[backwards-1] != '\n');
1250 try = backwards;
1251 } else {
1252 while (forwards + fragsize <= size) {
1253 if (buf[forwards++] == '\n')
1254 break;
1255 }
1256 try = forwards;
1257 }
1258
1259 if (try + fragsize > size)
1260 continue;
1261 if (memcmp(buf + try, fragment, fragsize))
1262 continue;
1263 n = (i >> 1)+1;
1264 if (i & 1)
1265 n = -n;
1266 *lines = n;
1267 return try;
1268 }
1269
1270 /*
1271 * We should start searching forward and backward.
1272 */
1273 return -1;
1274 }
1275
1276 static void remove_first_line(const char **rbuf, int *rsize)
1277 {
1278 const char *buf = *rbuf;
1279 int size = *rsize;
1280 unsigned long offset;
1281 offset = 0;
1282 while (offset <= size) {
1283 if (buf[offset++] == '\n')
1284 break;
1285 }
1286 *rsize = size - offset;
1287 *rbuf = buf + offset;
1288 }
1289
1290 static void remove_last_line(const char **rbuf, int *rsize)
1291 {
1292 const char *buf = *rbuf;
1293 int size = *rsize;
1294 unsigned long offset;
1295 offset = size - 1;
1296 while (offset > 0) {
1297 if (buf[--offset] == '\n')
1298 break;
1299 }
1300 *rsize = offset + 1;
1301 }
1302
1303 struct buffer_desc {
1304 char *buffer;
1305 unsigned long size;
1306 unsigned long alloc;
1307 };
1308
1309 static int apply_line(char *output, const char *patch, int plen)
1310 {
1311 /* plen is number of bytes to be copied from patch,
1312 * starting at patch+1 (patch[0] is '+'). Typically
1313 * patch[plen] is '\n'.
1314 */
1315 int add_nl_to_tail = 0;
1316 if ((new_whitespace == strip_whitespace) &&
1317 1 < plen && isspace(patch[plen-1])) {
1318 if (patch[plen] == '\n')
1319 add_nl_to_tail = 1;
1320 plen--;
1321 while (0 < plen && isspace(patch[plen]))
1322 plen--;
1323 applied_after_stripping++;
1324 }
1325 memcpy(output, patch + 1, plen);
1326 if (add_nl_to_tail)
1327 output[plen++] = '\n';
1328 return plen;
1329 }
1330
1331 static int apply_one_fragment(struct buffer_desc *desc, struct fragment *frag)
1332 {
1333 char *buf = desc->buffer;
1334 const char *patch = frag->patch;
1335 int offset, size = frag->size;
1336 char *old = xmalloc(size);
1337 char *new = xmalloc(size);
1338 const char *oldlines, *newlines;
1339 int oldsize = 0, newsize = 0;
1340 unsigned long leading, trailing;
1341 int pos, lines;
1342
1343 while (size > 0) {
1344 int len = linelen(patch, size);
1345 int plen;
1346
1347 if (!len)
1348 break;
1349
1350 /*
1351 * "plen" is how much of the line we should use for
1352 * the actual patch data. Normally we just remove the
1353 * first character on the line, but if the line is
1354 * followed by "\ No newline", then we also remove the
1355 * last one (which is the newline, of course).
1356 */
1357 plen = len-1;
1358 if (len < size && patch[len] == '\\')
1359 plen--;
1360 switch (*patch) {
1361 case ' ':
1362 case '-':
1363 memcpy(old + oldsize, patch + 1, plen);
1364 oldsize += plen;
1365 if (*patch == '-')
1366 break;
1367 /* Fall-through for ' ' */
1368 case '+':
1369 if (*patch != '+' || !no_add)
1370 newsize += apply_line(new + newsize, patch,
1371 plen);
1372 break;
1373 case '@': case '\\':
1374 /* Ignore it, we already handled it */
1375 break;
1376 default:
1377 return -1;
1378 }
1379 patch += len;
1380 size -= len;
1381 }
1382
1383 #ifdef NO_ACCURATE_DIFF
1384 if (oldsize > 0 && old[oldsize - 1] == '\n' &&
1385 newsize > 0 && new[newsize - 1] == '\n') {
1386 oldsize--;
1387 newsize--;
1388 }
1389 #endif
1390
1391 oldlines = old;
1392 newlines = new;
1393 leading = frag->leading;
1394 trailing = frag->trailing;
1395 lines = 0;
1396 pos = frag->newpos;
1397 for (;;) {
1398 offset = find_offset(buf, desc->size, oldlines, oldsize, pos, &lines);
1399 if (offset >= 0) {
1400 int diff = newsize - oldsize;
1401 unsigned long size = desc->size + diff;
1402 unsigned long alloc = desc->alloc;
1403
1404 /* Warn if it was necessary to reduce the number
1405 * of context lines.
1406 */
1407 if ((leading != frag->leading) || (trailing != frag->trailing))
1408 fprintf(stderr, "Context reduced to (%ld/%ld) to apply fragment at %d\n",
1409 leading, trailing, pos + lines);
1410
1411 if (size > alloc) {
1412 alloc = size + 8192;
1413 desc->alloc = alloc;
1414 buf = xrealloc(buf, alloc);
1415 desc->buffer = buf;
1416 }
1417 desc->size = size;
1418 memmove(buf + offset + newsize, buf + offset + oldsize, size - offset - newsize);
1419 memcpy(buf + offset, newlines, newsize);
1420 offset = 0;
1421
1422 break;
1423 }
1424
1425 /* Am I at my context limits? */
1426 if ((leading <= p_context) && (trailing <= p_context))
1427 break;
1428 /* Reduce the number of context lines
1429 * Reduce both leading and trailing if they are equal
1430 * otherwise just reduce the larger context.
1431 */
1432 if (leading >= trailing) {
1433 remove_first_line(&oldlines, &oldsize);
1434 remove_first_line(&newlines, &newsize);
1435 pos--;
1436 leading--;
1437 }
1438 if (trailing > leading) {
1439 remove_last_line(&oldlines, &oldsize);
1440 remove_last_line(&newlines, &newsize);
1441 trailing--;
1442 }
1443 }
1444
1445 free(old);
1446 free(new);
1447 return offset;
1448 }
1449
1450 static char *inflate_it(const void *data, unsigned long size,
1451 unsigned long inflated_size)
1452 {
1453 z_stream stream;
1454 void *out;
1455 int st;
1456
1457 memset(&stream, 0, sizeof(stream));
1458
1459 stream.next_in = (unsigned char *)data;
1460 stream.avail_in = size;
1461 stream.next_out = out = xmalloc(inflated_size);
1462 stream.avail_out = inflated_size;
1463 inflateInit(&stream);
1464 st = inflate(&stream, Z_FINISH);
1465 if ((st != Z_STREAM_END) || stream.total_out != inflated_size) {
1466 free(out);
1467 return NULL;
1468 }
1469 return out;
1470 }
1471
1472 static int apply_binary_fragment(struct buffer_desc *desc, struct patch *patch)
1473 {
1474 unsigned long dst_size;
1475 struct fragment *fragment = patch->fragments;
1476 void *data;
1477 void *result;
1478
1479 data = inflate_it(fragment->patch, fragment->size,
1480 patch->deflate_origlen);
1481 if (!data)
1482 return error("corrupt patch data");
1483 switch (patch->is_binary) {
1484 case BINARY_DELTA_DEFLATED:
1485 result = patch_delta(desc->buffer, desc->size,
1486 data,
1487 patch->deflate_origlen,
1488 &dst_size);
1489 free(desc->buffer);
1490 desc->buffer = result;
1491 free(data);
1492 break;
1493 case BINARY_LITERAL_DEFLATED:
1494 free(desc->buffer);
1495 desc->buffer = data;
1496 dst_size = patch->deflate_origlen;
1497 break;
1498 }
1499 if (!desc->buffer)
1500 return -1;
1501 desc->size = desc->alloc = dst_size;
1502 return 0;
1503 }
1504
1505 static int apply_binary(struct buffer_desc *desc, struct patch *patch)
1506 {
1507 const char *name = patch->old_name ? patch->old_name : patch->new_name;
1508 unsigned char sha1[20];
1509 unsigned char hdr[50];
1510 int hdrlen;
1511
1512 if (!allow_binary_replacement)
1513 return error("cannot apply binary patch to '%s' "
1514 "without --allow-binary-replacement",
1515 name);
1516
1517 /* For safety, we require patch index line to contain
1518 * full 40-byte textual SHA1 for old and new, at least for now.
1519 */
1520 if (strlen(patch->old_sha1_prefix) != 40 ||
1521 strlen(patch->new_sha1_prefix) != 40 ||
1522 get_sha1_hex(patch->old_sha1_prefix, sha1) ||
1523 get_sha1_hex(patch->new_sha1_prefix, sha1))
1524 return error("cannot apply binary patch to '%s' "
1525 "without full index line", name);
1526
1527 if (patch->old_name) {
1528 /* See if the old one matches what the patch
1529 * applies to.
1530 */
1531 write_sha1_file_prepare(desc->buffer, desc->size,
1532 blob_type, sha1, hdr, &hdrlen);
1533 if (strcmp(sha1_to_hex(sha1), patch->old_sha1_prefix))
1534 return error("the patch applies to '%s' (%s), "
1535 "which does not match the "
1536 "current contents.",
1537 name, sha1_to_hex(sha1));
1538 }
1539 else {
1540 /* Otherwise, the old one must be empty. */
1541 if (desc->size)
1542 return error("the patch applies to an empty "
1543 "'%s' but it is not empty", name);
1544 }
1545
1546 get_sha1_hex(patch->new_sha1_prefix, sha1);
1547 if (!memcmp(sha1, null_sha1, 20)) {
1548 free(desc->buffer);
1549 desc->alloc = desc->size = 0;
1550 desc->buffer = NULL;
1551 return 0; /* deletion patch */
1552 }
1553
1554 if (has_sha1_file(sha1)) {
1555 /* We already have the postimage */
1556 char type[10];
1557 unsigned long size;
1558
1559 free(desc->buffer);
1560 desc->buffer = read_sha1_file(sha1, type, &size);
1561 if (!desc->buffer)
1562 return error("the necessary postimage %s for "
1563 "'%s' cannot be read",
1564 patch->new_sha1_prefix, name);
1565 desc->alloc = desc->size = size;
1566 }
1567 else {
1568 /* We have verified desc matches the preimage;
1569 * apply the patch data to it, which is stored
1570 * in the patch->fragments->{patch,size}.
1571 */
1572 if (apply_binary_fragment(desc, patch))
1573 return error("binary patch does not apply to '%s'",
1574 name);
1575
1576 /* verify that the result matches */
1577 write_sha1_file_prepare(desc->buffer, desc->size, blob_type,
1578 sha1, hdr, &hdrlen);
1579 if (strcmp(sha1_to_hex(sha1), patch->new_sha1_prefix))
1580 return error("binary patch to '%s' creates incorrect result", name);
1581 }
1582
1583 return 0;
1584 }
1585
1586 static int apply_fragments(struct buffer_desc *desc, struct patch *patch)
1587 {
1588 struct fragment *frag = patch->fragments;
1589 const char *name = patch->old_name ? patch->old_name : patch->new_name;
1590
1591 if (patch->is_binary)
1592 return apply_binary(desc, patch);
1593
1594 while (frag) {
1595 if (apply_one_fragment(desc, frag) < 0)
1596 return error("patch failed: %s:%ld",
1597 name, frag->oldpos);
1598 frag = frag->next;
1599 }
1600 return 0;
1601 }
1602
1603 static int apply_data(struct patch *patch, struct stat *st)
1604 {
1605 char *buf;
1606 unsigned long size, alloc;
1607 struct buffer_desc desc;
1608
1609 size = 0;
1610 alloc = 0;
1611 buf = NULL;
1612 if (patch->old_name) {
1613 size = st->st_size;
1614 alloc = size + 8192;
1615 buf = xmalloc(alloc);
1616 if (read_old_data(st, patch->old_name, buf, alloc) != size)
1617 return error("read of %s failed", patch->old_name);
1618 }
1619
1620 desc.size = size;
1621 desc.alloc = alloc;
1622 desc.buffer = buf;
1623 if (apply_fragments(&desc, patch) < 0)
1624 return -1;
1625 patch->result = desc.buffer;
1626 patch->resultsize = desc.size;
1627
1628 if (patch->is_delete && patch->resultsize)
1629 return error("removal patch leaves file contents");
1630
1631 return 0;
1632 }
1633
1634 static int check_patch(struct patch *patch)
1635 {
1636 struct stat st;
1637 const char *old_name = patch->old_name;
1638 const char *new_name = patch->new_name;
1639 const char *name = old_name ? old_name : new_name;
1640
1641 if (old_name) {
1642 int changed;
1643 int stat_ret = lstat(old_name, &st);
1644
1645 if (check_index) {
1646 int pos = cache_name_pos(old_name, strlen(old_name));
1647 if (pos < 0)
1648 return error("%s: does not exist in index",
1649 old_name);
1650 if (stat_ret < 0) {
1651 struct checkout costate;
1652 if (errno != ENOENT)
1653 return error("%s: %s", old_name,
1654 strerror(errno));
1655 /* checkout */
1656 costate.base_dir = "";
1657 costate.base_dir_len = 0;
1658 costate.force = 0;
1659 costate.quiet = 0;
1660 costate.not_new = 0;
1661 costate.refresh_cache = 1;
1662 if (checkout_entry(active_cache[pos],
1663 &costate,
1664 NULL) ||
1665 lstat(old_name, &st))
1666 return -1;
1667 }
1668
1669 changed = ce_match_stat(active_cache[pos], &st, 1);
1670 if (changed)
1671 return error("%s: does not match index",
1672 old_name);
1673 }
1674 else if (stat_ret < 0)
1675 return error("%s: %s", old_name, strerror(errno));
1676
1677 if (patch->is_new < 0)
1678 patch->is_new = 0;
1679 st.st_mode = ntohl(create_ce_mode(st.st_mode));
1680 if (!patch->old_mode)
1681 patch->old_mode = st.st_mode;
1682 if ((st.st_mode ^ patch->old_mode) & S_IFMT)
1683 return error("%s: wrong type", old_name);
1684 if (st.st_mode != patch->old_mode)
1685 fprintf(stderr, "warning: %s has type %o, expected %o\n",
1686 old_name, st.st_mode, patch->old_mode);
1687 }
1688
1689 if (new_name && (patch->is_new | patch->is_rename | patch->is_copy)) {
1690 if (check_index && cache_name_pos(new_name, strlen(new_name)) >= 0)
1691 return error("%s: already exists in index", new_name);
1692 if (!lstat(new_name, &st))
1693 return error("%s: already exists in working directory", new_name);
1694 if (errno != ENOENT)
1695 return error("%s: %s", new_name, strerror(errno));
1696 if (!patch->new_mode) {
1697 if (patch->is_new)
1698 patch->new_mode = S_IFREG | 0644;
1699 else
1700 patch->new_mode = patch->old_mode;
1701 }
1702 }
1703
1704 if (new_name && old_name) {
1705 int same = !strcmp(old_name, new_name);
1706 if (!patch->new_mode)
1707 patch->new_mode = patch->old_mode;
1708 if ((patch->old_mode ^ patch->new_mode) & S_IFMT)
1709 return error("new mode (%o) of %s does not match old mode (%o)%s%s",
1710 patch->new_mode, new_name, patch->old_mode,
1711 same ? "" : " of ", same ? "" : old_name);
1712 }
1713
1714 if (apply_data(patch, &st) < 0)
1715 return error("%s: patch does not apply", name);
1716 return 0;
1717 }
1718
1719 static int check_patch_list(struct patch *patch)
1720 {
1721 int error = 0;
1722
1723 for (;patch ; patch = patch->next)
1724 error |= check_patch(patch);
1725 return error;
1726 }
1727
1728 static inline int is_null_sha1(const unsigned char *sha1)
1729 {
1730 return !memcmp(sha1, null_sha1, 20);
1731 }
1732
1733 static void show_index_list(struct patch *list)
1734 {
1735 struct patch *patch;
1736
1737 /* Once we start supporting the reverse patch, it may be
1738 * worth showing the new sha1 prefix, but until then...
1739 */
1740 for (patch = list; patch; patch = patch->next) {
1741 const unsigned char *sha1_ptr;
1742 unsigned char sha1[20];
1743 const char *name;
1744
1745 name = patch->old_name ? patch->old_name : patch->new_name;
1746 if (patch->is_new)
1747 sha1_ptr = null_sha1;
1748 else if (get_sha1(patch->old_sha1_prefix, sha1))
1749 die("sha1 information is lacking or useless (%s).",
1750 name);
1751 else
1752 sha1_ptr = sha1;
1753
1754 printf("%06o %s ",patch->old_mode, sha1_to_hex(sha1_ptr));
1755 if (line_termination && quote_c_style(name, NULL, NULL, 0))
1756 quote_c_style(name, NULL, stdout, 0);
1757 else
1758 fputs(name, stdout);
1759 putchar(line_termination);
1760 }
1761 }
1762
1763 static void stat_patch_list(struct patch *patch)
1764 {
1765 int files, adds, dels;
1766
1767 for (files = adds = dels = 0 ; patch ; patch = patch->next) {
1768 files++;
1769 adds += patch->lines_added;
1770 dels += patch->lines_deleted;
1771 show_stats(patch);
1772 }
1773
1774 printf(" %d files changed, %d insertions(+), %d deletions(-)\n", files, adds, dels);
1775 }
1776
1777 static void numstat_patch_list(struct patch *patch)
1778 {
1779 for ( ; patch; patch = patch->next) {
1780 const char *name;
1781 name = patch->old_name ? patch->old_name : patch->new_name;
1782 printf("%d\t%d\t", patch->lines_added, patch->lines_deleted);
1783 if (line_termination && quote_c_style(name, NULL, NULL, 0))
1784 quote_c_style(name, NULL, stdout, 0);
1785 else
1786 fputs(name, stdout);
1787 putchar('\n');
1788 }
1789 }
1790
1791 static void show_file_mode_name(const char *newdelete, unsigned int mode, const char *name)
1792 {
1793 if (mode)
1794 printf(" %s mode %06o %s\n", newdelete, mode, name);
1795 else
1796 printf(" %s %s\n", newdelete, name);
1797 }
1798
1799 static void show_mode_change(struct patch *p, int show_name)
1800 {
1801 if (p->old_mode && p->new_mode && p->old_mode != p->new_mode) {
1802 if (show_name)
1803 printf(" mode change %06o => %06o %s\n",
1804 p->old_mode, p->new_mode, p->new_name);
1805 else
1806 printf(" mode change %06o => %06o\n",
1807 p->old_mode, p->new_mode);
1808 }
1809 }
1810
1811 static void show_rename_copy(struct patch *p)
1812 {
1813 const char *renamecopy = p->is_rename ? "rename" : "copy";
1814 const char *old, *new;
1815
1816 /* Find common prefix */
1817 old = p->old_name;
1818 new = p->new_name;
1819 while (1) {
1820 const char *slash_old, *slash_new;
1821 slash_old = strchr(old, '/');
1822 slash_new = strchr(new, '/');
1823 if (!slash_old ||
1824 !slash_new ||
1825 slash_old - old != slash_new - new ||
1826 memcmp(old, new, slash_new - new))
1827 break;
1828 old = slash_old + 1;
1829 new = slash_new + 1;
1830 }
1831 /* p->old_name thru old is the common prefix, and old and new
1832 * through the end of names are renames
1833 */
1834 if (old != p->old_name)
1835 printf(" %s %.*s{%s => %s} (%d%%)\n", renamecopy,
1836 (int)(old - p->old_name), p->old_name,
1837 old, new, p->score);
1838 else
1839 printf(" %s %s => %s (%d%%)\n", renamecopy,
1840 p->old_name, p->new_name, p->score);
1841 show_mode_change(p, 0);
1842 }
1843
1844 static void summary_patch_list(struct patch *patch)
1845 {
1846 struct patch *p;
1847
1848 for (p = patch; p; p = p->next) {
1849 if (p->is_new)
1850 show_file_mode_name("create", p->new_mode, p->new_name);
1851 else if (p->is_delete)
1852 show_file_mode_name("delete", p->old_mode, p->old_name);
1853 else {
1854 if (p->is_rename || p->is_copy)
1855 show_rename_copy(p);
1856 else {
1857 if (p->score) {
1858 printf(" rewrite %s (%d%%)\n",
1859 p->new_name, p->score);
1860 show_mode_change(p, 0);
1861 }
1862 else
1863 show_mode_change(p, 1);
1864 }
1865 }
1866 }
1867 }
1868
1869 static void patch_stats(struct patch *patch)
1870 {
1871 int lines = patch->lines_added + patch->lines_deleted;
1872
1873 if (lines > max_change)
1874 max_change = lines;
1875 if (patch->old_name) {
1876 int len = quote_c_style(patch->old_name, NULL, NULL, 0);
1877 if (!len)
1878 len = strlen(patch->old_name);
1879 if (len > max_len)
1880 max_len = len;
1881 }
1882 if (patch->new_name) {
1883 int len = quote_c_style(patch->new_name, NULL, NULL, 0);
1884 if (!len)
1885 len = strlen(patch->new_name);
1886 if (len > max_len)
1887 max_len = len;
1888 }
1889 }
1890
1891 static void remove_file(struct patch *patch)
1892 {
1893 if (write_index) {
1894 if (remove_file_from_cache(patch->old_name) < 0)
1895 die("unable to remove %s from index", patch->old_name);
1896 }
1897 unlink(patch->old_name);
1898 }
1899
1900 static void add_index_file(const char *path, unsigned mode, void *buf, unsigned long size)
1901 {
1902 struct stat st;
1903 struct cache_entry *ce;
1904 int namelen = strlen(path);
1905 unsigned ce_size = cache_entry_size(namelen);
1906
1907 if (!write_index)
1908 return;
1909
1910 ce = xcalloc(1, ce_size);
1911 memcpy(ce->name, path, namelen);
1912 ce->ce_mode = create_ce_mode(mode);
1913 ce->ce_flags = htons(namelen);
1914 if (lstat(path, &st) < 0)
1915 die("unable to stat newly created file %s", path);
1916 fill_stat_cache_info(ce, &st);
1917 if (write_sha1_file(buf, size, blob_type, ce->sha1) < 0)
1918 die("unable to create backing store for newly created file %s", path);
1919 if (add_cache_entry(ce, ADD_CACHE_OK_TO_ADD) < 0)
1920 die("unable to add cache entry for %s", path);
1921 }
1922
1923 static int try_create_file(const char *path, unsigned int mode, const char *buf, unsigned long size)
1924 {
1925 int fd;
1926
1927 if (S_ISLNK(mode))
1928 return symlink(buf, path);
1929 fd = open(path, O_CREAT | O_EXCL | O_WRONLY, (mode & 0100) ? 0777 : 0666);
1930 if (fd < 0)
1931 return -1;
1932 while (size) {
1933 int written = xwrite(fd, buf, size);
1934 if (written < 0)
1935 die("writing file %s: %s", path, strerror(errno));
1936 if (!written)
1937 die("out of space writing file %s", path);
1938 buf += written;
1939 size -= written;
1940 }
1941 if (close(fd) < 0)
1942 die("closing file %s: %s", path, strerror(errno));
1943 return 0;
1944 }
1945
1946 /*
1947 * We optimistically assume that the directories exist,
1948 * which is true 99% of the time anyway. If they don't,
1949 * we create them and try again.
1950 */
1951 static void create_one_file(char *path, unsigned mode, const char *buf, unsigned long size)
1952 {
1953 if (!try_create_file(path, mode, buf, size))
1954 return;
1955
1956 if (errno == ENOENT) {
1957 if (safe_create_leading_directories(path))
1958 return;
1959 if (!try_create_file(path, mode, buf, size))
1960 return;
1961 }
1962
1963 if (errno == EEXIST) {
1964 unsigned int nr = getpid();
1965
1966 for (;;) {
1967 const char *newpath;
1968 newpath = mkpath("%s~%u", path, nr);
1969 if (!try_create_file(newpath, mode, buf, size)) {
1970 if (!rename(newpath, path))
1971 return;
1972 unlink(newpath);
1973 break;
1974 }
1975 if (errno != EEXIST)
1976 break;
1977 ++nr;
1978 }
1979 }
1980 die("unable to write file %s mode %o", path, mode);
1981 }
1982
1983 static void create_file(struct patch *patch)
1984 {
1985 char *path = patch->new_name;
1986 unsigned mode = patch->new_mode;
1987 unsigned long size = patch->resultsize;
1988 char *buf = patch->result;
1989
1990 if (!mode)
1991 mode = S_IFREG | 0644;
1992 create_one_file(path, mode, buf, size);
1993 add_index_file(path, mode, buf, size);
1994 }
1995
1996 static void write_out_one_result(struct patch *patch)
1997 {
1998 if (patch->is_delete > 0) {
1999 remove_file(patch);
2000 return;
2001 }
2002 if (patch->is_new > 0 || patch->is_copy) {
2003 create_file(patch);
2004 return;
2005 }
2006 /*
2007 * Rename or modification boils down to the same
2008 * thing: remove the old, write the new
2009 */
2010 remove_file(patch);
2011 create_file(patch);
2012 }
2013
2014 static void write_out_results(struct patch *list, int skipped_patch)
2015 {
2016 if (!list && !skipped_patch)
2017 die("No changes");
2018
2019 while (list) {
2020 write_out_one_result(list);
2021 list = list->next;
2022 }
2023 }
2024
2025 static struct cache_file cache_file;
2026
2027 static struct excludes {
2028 struct excludes *next;
2029 const char *path;
2030 } *excludes;
2031
2032 static int use_patch(struct patch *p)
2033 {
2034 const char *pathname = p->new_name ? p->new_name : p->old_name;
2035 struct excludes *x = excludes;
2036 while (x) {
2037 if (fnmatch(x->path, pathname, 0) == 0)
2038 return 0;
2039 x = x->next;
2040 }
2041 if (0 < prefix_length) {
2042 int pathlen = strlen(pathname);
2043 if (pathlen <= prefix_length ||
2044 memcmp(prefix, pathname, prefix_length))
2045 return 0;
2046 }
2047 return 1;
2048 }
2049
2050 static int apply_patch(int fd, const char *filename)
2051 {
2052 unsigned long offset, size;
2053 char *buffer = read_patch_file(fd, &size);
2054 struct patch *list = NULL, **listp = &list;
2055 int skipped_patch = 0;
2056
2057 patch_input_file = filename;
2058 if (!buffer)
2059 return -1;
2060 offset = 0;
2061 while (size > 0) {
2062 struct patch *patch;
2063 int nr;
2064
2065 patch = xcalloc(1, sizeof(*patch));
2066 nr = parse_chunk(buffer + offset, size, patch);
2067 if (nr < 0)
2068 break;
2069 if (use_patch(patch)) {
2070 patch_stats(patch);
2071 *listp = patch;
2072 listp = &patch->next;
2073 } else {
2074 /* perhaps free it a bit better? */
2075 free(patch);
2076 skipped_patch++;
2077 }
2078 offset += nr;
2079 size -= nr;
2080 }
2081
2082 if (whitespace_error && (new_whitespace == error_on_whitespace))
2083 apply = 0;
2084
2085 write_index = check_index && apply;
2086 if (write_index && newfd < 0)
2087 newfd = hold_index_file_for_update(&cache_file, get_index_file());
2088 if (check_index) {
2089 if (read_cache() < 0)
2090 die("unable to read index file");
2091 }
2092
2093 if ((check || apply) && check_patch_list(list) < 0)
2094 exit(1);
2095
2096 if (apply)
2097 write_out_results(list, skipped_patch);
2098
2099 if (show_index_info)
2100 show_index_list(list);
2101
2102 if (diffstat)
2103 stat_patch_list(list);
2104
2105 if (numstat)
2106 numstat_patch_list(list);
2107
2108 if (summary)
2109 summary_patch_list(list);
2110
2111 free(buffer);
2112 return 0;
2113 }
2114
2115 static int git_apply_config(const char *var, const char *value)
2116 {
2117 if (!strcmp(var, "apply.whitespace")) {
2118 apply_default_whitespace = strdup(value);
2119 return 0;
2120 }
2121 return git_default_config(var, value);
2122 }
2123
2124
2125 int main(int argc, char **argv)
2126 {
2127 int i;
2128 int read_stdin = 1;
2129 const char *whitespace_option = NULL;
2130
2131 for (i = 1; i < argc; i++) {
2132 const char *arg = argv[i];
2133 char *end;
2134 int fd;
2135
2136 if (!strcmp(arg, "-")) {
2137 apply_patch(0, "<stdin>");
2138 read_stdin = 0;
2139 continue;
2140 }
2141 if (!strncmp(arg, "--exclude=", 10)) {
2142 struct excludes *x = xmalloc(sizeof(*x));
2143 x->path = arg + 10;
2144 x->next = excludes;
2145 excludes = x;
2146 continue;
2147 }
2148 if (!strncmp(arg, "-p", 2)) {
2149 p_value = atoi(arg + 2);
2150 continue;
2151 }
2152 if (!strcmp(arg, "--no-add")) {
2153 no_add = 1;
2154 continue;
2155 }
2156 if (!strcmp(arg, "--stat")) {
2157 apply = 0;
2158 diffstat = 1;
2159 continue;
2160 }
2161 if (!strcmp(arg, "--allow-binary-replacement") ||
2162 !strcmp(arg, "--binary")) {
2163 allow_binary_replacement = 1;
2164 continue;
2165 }
2166 if (!strcmp(arg, "--numstat")) {
2167 apply = 0;
2168 numstat = 1;
2169 continue;
2170 }
2171 if (!strcmp(arg, "--summary")) {
2172 apply = 0;
2173 summary = 1;
2174 continue;
2175 }
2176 if (!strcmp(arg, "--check")) {
2177 apply = 0;
2178 check = 1;
2179 continue;
2180 }
2181 if (!strcmp(arg, "--index")) {
2182 check_index = 1;
2183 continue;
2184 }
2185 if (!strcmp(arg, "--apply")) {
2186 apply = 1;
2187 continue;
2188 }
2189 if (!strcmp(arg, "--index-info")) {
2190 apply = 0;
2191 show_index_info = 1;
2192 continue;
2193 }
2194 if (!strcmp(arg, "-z")) {
2195 line_termination = 0;
2196 continue;
2197 }
2198 if (!strncmp(arg, "-C", 2)) {
2199 p_context = strtoul(arg + 2, &end, 0);
2200 if (*end != '\0')
2201 die("unrecognized context count '%s'", arg + 2);
2202 continue;
2203 }
2204 if (!strncmp(arg, "--whitespace=", 13)) {
2205 whitespace_option = arg + 13;
2206 parse_whitespace_option(arg + 13);
2207 continue;
2208 }
2209
2210 if (check_index && prefix_length < 0) {
2211 prefix = setup_git_directory();
2212 prefix_length = prefix ? strlen(prefix) : 0;
2213 git_config(git_apply_config);
2214 if (!whitespace_option && apply_default_whitespace)
2215 parse_whitespace_option(apply_default_whitespace);
2216 }
2217 if (0 < prefix_length)
2218 arg = prefix_filename(prefix, prefix_length, arg);
2219
2220 fd = open(arg, O_RDONLY);
2221 if (fd < 0)
2222 usage(apply_usage);
2223 read_stdin = 0;
2224 set_default_whitespace_mode(whitespace_option);
2225 apply_patch(fd, arg);
2226 close(fd);
2227 }
2228 set_default_whitespace_mode(whitespace_option);
2229 if (read_stdin)
2230 apply_patch(0, "<stdin>");
2231 if (whitespace_error) {
2232 if (squelch_whitespace_errors &&
2233 squelch_whitespace_errors < whitespace_error) {
2234 int squelched =
2235 whitespace_error - squelch_whitespace_errors;
2236 fprintf(stderr, "warning: squelched %d whitespace error%s\n",
2237 squelched,
2238 squelched == 1 ? "" : "s");
2239 }
2240 if (new_whitespace == error_on_whitespace)
2241 die("%d line%s add%s trailing whitespaces.",
2242 whitespace_error,
2243 whitespace_error == 1 ? "" : "s",
2244 whitespace_error == 1 ? "s" : "");
2245 if (applied_after_stripping)
2246 fprintf(stderr, "warning: %d line%s applied after"
2247 " stripping trailing whitespaces.\n",
2248 applied_after_stripping,
2249 applied_after_stripping == 1 ? "" : "s");
2250 else if (whitespace_error)
2251 fprintf(stderr, "warning: %d line%s add%s trailing"
2252 " whitespaces.\n",
2253 whitespace_error,
2254 whitespace_error == 1 ? "" : "s",
2255 whitespace_error == 1 ? "s" : "");
2256 }
2257
2258 if (write_index) {
2259 if (write_cache(newfd, active_cache, active_nr) ||
2260 commit_index_file(&cache_file))
2261 die("Unable to write new cachefile");
2262 }
2263
2264 return 0;
2265 }