Merge branch 'jc/cache-tree'
[git/git.git] / builtin-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 "cache-tree.h"
12 #include "quote.h"
13 #include "blob.h"
14 #include "delta.h"
15 #include "builtin.h"
16
17 // --check turns on checking that the working tree matches the
18 // files that are being modified, but doesn't apply the patch
19 // --stat does just a diffstat, and doesn't actually apply
20 // --numstat does numeric diffstat, and doesn't actually apply
21 // --index-info shows the old and new index info for paths if available.
22 // --index updates the cache as well.
23 // --cached updates only the cache without ever touching the working tree.
24 //
25 static const char *prefix;
26 static int prefix_length = -1;
27 static int newfd = -1;
28
29 static int p_value = 1;
30 static int allow_binary_replacement = 0;
31 static int check_index = 0;
32 static int write_index = 0;
33 static int cached = 0;
34 static int diffstat = 0;
35 static int numstat = 0;
36 static int summary = 0;
37 static int check = 0;
38 static int apply = 1;
39 static int no_add = 0;
40 static int show_index_info = 0;
41 static int line_termination = '\n';
42 static unsigned long p_context = -1;
43 static const char apply_usage[] =
44 "git-apply [--stat] [--numstat] [--summary] [--check] [--index] [--cached] [--apply] [--no-add] [--index-info] [--allow-binary-replacement] [-z] [-pNUM] [-CNUM] [--whitespace=<nowarn|warn|error|error-all|strip>] <patch>...";
45
46 static enum whitespace_eol {
47 nowarn_whitespace,
48 warn_on_whitespace,
49 error_on_whitespace,
50 strip_whitespace,
51 } new_whitespace = warn_on_whitespace;
52 static int whitespace_error = 0;
53 static int squelch_whitespace_errors = 5;
54 static int applied_after_stripping = 0;
55 static const char *patch_input_file = NULL;
56
57 static void parse_whitespace_option(const char *option)
58 {
59 if (!option) {
60 new_whitespace = warn_on_whitespace;
61 return;
62 }
63 if (!strcmp(option, "warn")) {
64 new_whitespace = warn_on_whitespace;
65 return;
66 }
67 if (!strcmp(option, "nowarn")) {
68 new_whitespace = nowarn_whitespace;
69 return;
70 }
71 if (!strcmp(option, "error")) {
72 new_whitespace = error_on_whitespace;
73 return;
74 }
75 if (!strcmp(option, "error-all")) {
76 new_whitespace = error_on_whitespace;
77 squelch_whitespace_errors = 0;
78 return;
79 }
80 if (!strcmp(option, "strip")) {
81 new_whitespace = strip_whitespace;
82 return;
83 }
84 die("unrecognized whitespace option '%s'", option);
85 }
86
87 static void set_default_whitespace_mode(const char *whitespace_option)
88 {
89 if (!whitespace_option && !apply_default_whitespace) {
90 new_whitespace = (apply
91 ? warn_on_whitespace
92 : nowarn_whitespace);
93 }
94 }
95
96 /*
97 * For "diff-stat" like behaviour, we keep track of the biggest change
98 * we've seen, and the longest filename. That allows us to do simple
99 * scaling.
100 */
101 static int max_change, max_len;
102
103 /*
104 * Various "current state", notably line numbers and what
105 * file (and how) we're patching right now.. The "is_xxxx"
106 * things are flags, where -1 means "don't know yet".
107 */
108 static int linenr = 1;
109
110 struct fragment {
111 unsigned long leading, trailing;
112 unsigned long oldpos, oldlines;
113 unsigned long newpos, newlines;
114 const char *patch;
115 int size;
116 struct fragment *next;
117 };
118
119 struct patch {
120 char *new_name, *old_name, *def_name;
121 unsigned int old_mode, new_mode;
122 int is_rename, is_copy, is_new, is_delete, is_binary;
123 #define BINARY_DELTA_DEFLATED 1
124 #define BINARY_LITERAL_DEFLATED 2
125 unsigned long deflate_origlen;
126 int lines_added, lines_deleted;
127 int score;
128 struct fragment *fragments;
129 char *result;
130 unsigned long resultsize;
131 char old_sha1_prefix[41];
132 char new_sha1_prefix[41];
133 struct patch *next;
134 };
135
136 #define CHUNKSIZE (8192)
137 #define SLOP (16)
138
139 static void *read_patch_file(int fd, unsigned long *sizep)
140 {
141 unsigned long size = 0, alloc = CHUNKSIZE;
142 void *buffer = xmalloc(alloc);
143
144 for (;;) {
145 int nr = alloc - size;
146 if (nr < 1024) {
147 alloc += CHUNKSIZE;
148 buffer = xrealloc(buffer, alloc);
149 nr = alloc - size;
150 }
151 nr = xread(fd, buffer + size, nr);
152 if (!nr)
153 break;
154 if (nr < 0)
155 die("git-apply: read returned %s", strerror(errno));
156 size += nr;
157 }
158 *sizep = size;
159
160 /*
161 * Make sure that we have some slop in the buffer
162 * so that we can do speculative "memcmp" etc, and
163 * see to it that it is NUL-filled.
164 */
165 if (alloc < size + SLOP)
166 buffer = xrealloc(buffer, size + SLOP);
167 memset(buffer + size, 0, SLOP);
168 return buffer;
169 }
170
171 static unsigned long linelen(const char *buffer, unsigned long size)
172 {
173 unsigned long len = 0;
174 while (size--) {
175 len++;
176 if (*buffer++ == '\n')
177 break;
178 }
179 return len;
180 }
181
182 static int is_dev_null(const char *str)
183 {
184 return !memcmp("/dev/null", str, 9) && isspace(str[9]);
185 }
186
187 #define TERM_SPACE 1
188 #define TERM_TAB 2
189
190 static int name_terminate(const char *name, int namelen, int c, int terminate)
191 {
192 if (c == ' ' && !(terminate & TERM_SPACE))
193 return 0;
194 if (c == '\t' && !(terminate & TERM_TAB))
195 return 0;
196
197 return 1;
198 }
199
200 static char * find_name(const char *line, char *def, int p_value, int terminate)
201 {
202 int len;
203 const char *start = line;
204 char *name;
205
206 if (*line == '"') {
207 /* Proposed "new-style" GNU patch/diff format; see
208 * http://marc.theaimsgroup.com/?l=git&m=112927316408690&w=2
209 */
210 name = unquote_c_style(line, NULL);
211 if (name) {
212 char *cp = name;
213 while (p_value) {
214 cp = strchr(name, '/');
215 if (!cp)
216 break;
217 cp++;
218 p_value--;
219 }
220 if (cp) {
221 /* name can later be freed, so we need
222 * to memmove, not just return cp
223 */
224 memmove(name, cp, strlen(cp) + 1);
225 free(def);
226 return name;
227 }
228 else {
229 free(name);
230 name = NULL;
231 }
232 }
233 }
234
235 for (;;) {
236 char c = *line;
237
238 if (isspace(c)) {
239 if (c == '\n')
240 break;
241 if (name_terminate(start, line-start, c, terminate))
242 break;
243 }
244 line++;
245 if (c == '/' && !--p_value)
246 start = line;
247 }
248 if (!start)
249 return def;
250 len = line - start;
251 if (!len)
252 return def;
253
254 /*
255 * Generally we prefer the shorter name, especially
256 * if the other one is just a variation of that with
257 * something else tacked on to the end (ie "file.orig"
258 * or "file~").
259 */
260 if (def) {
261 int deflen = strlen(def);
262 if (deflen < len && !strncmp(start, def, deflen))
263 return def;
264 }
265
266 name = xmalloc(len + 1);
267 memcpy(name, start, len);
268 name[len] = 0;
269 free(def);
270 return name;
271 }
272
273 /*
274 * Get the name etc info from the --/+++ lines of a traditional patch header
275 *
276 * NOTE! This hardcodes "-p1" behaviour in filename detection.
277 *
278 * FIXME! The end-of-filename heuristics are kind of screwy. For existing
279 * files, we can happily check the index for a match, but for creating a
280 * new file we should try to match whatever "patch" does. I have no idea.
281 */
282 static void parse_traditional_patch(const char *first, const char *second, struct patch *patch)
283 {
284 char *name;
285
286 first += 4; // skip "--- "
287 second += 4; // skip "+++ "
288 if (is_dev_null(first)) {
289 patch->is_new = 1;
290 patch->is_delete = 0;
291 name = find_name(second, NULL, p_value, TERM_SPACE | TERM_TAB);
292 patch->new_name = name;
293 } else if (is_dev_null(second)) {
294 patch->is_new = 0;
295 patch->is_delete = 1;
296 name = find_name(first, NULL, p_value, TERM_SPACE | TERM_TAB);
297 patch->old_name = name;
298 } else {
299 name = find_name(first, NULL, p_value, TERM_SPACE | TERM_TAB);
300 name = find_name(second, name, p_value, TERM_SPACE | TERM_TAB);
301 patch->old_name = patch->new_name = name;
302 }
303 if (!name)
304 die("unable to find filename in patch at line %d", linenr);
305 }
306
307 static int gitdiff_hdrend(const char *line, struct patch *patch)
308 {
309 return -1;
310 }
311
312 /*
313 * We're anal about diff header consistency, to make
314 * sure that we don't end up having strange ambiguous
315 * patches floating around.
316 *
317 * As a result, gitdiff_{old|new}name() will check
318 * their names against any previous information, just
319 * to make sure..
320 */
321 static char *gitdiff_verify_name(const char *line, int isnull, char *orig_name, const char *oldnew)
322 {
323 if (!orig_name && !isnull)
324 return find_name(line, NULL, 1, 0);
325
326 if (orig_name) {
327 int len;
328 const char *name;
329 char *another;
330 name = orig_name;
331 len = strlen(name);
332 if (isnull)
333 die("git-apply: bad git-diff - expected /dev/null, got %s on line %d", name, linenr);
334 another = find_name(line, NULL, 1, 0);
335 if (!another || memcmp(another, name, len))
336 die("git-apply: bad git-diff - inconsistent %s filename on line %d", oldnew, linenr);
337 free(another);
338 return orig_name;
339 }
340 else {
341 /* expect "/dev/null" */
342 if (memcmp("/dev/null", line, 9) || line[9] != '\n')
343 die("git-apply: bad git-diff - expected /dev/null on line %d", linenr);
344 return NULL;
345 }
346 }
347
348 static int gitdiff_oldname(const char *line, struct patch *patch)
349 {
350 patch->old_name = gitdiff_verify_name(line, patch->is_new, patch->old_name, "old");
351 return 0;
352 }
353
354 static int gitdiff_newname(const char *line, struct patch *patch)
355 {
356 patch->new_name = gitdiff_verify_name(line, patch->is_delete, patch->new_name, "new");
357 return 0;
358 }
359
360 static int gitdiff_oldmode(const char *line, struct patch *patch)
361 {
362 patch->old_mode = strtoul(line, NULL, 8);
363 return 0;
364 }
365
366 static int gitdiff_newmode(const char *line, struct patch *patch)
367 {
368 patch->new_mode = strtoul(line, NULL, 8);
369 return 0;
370 }
371
372 static int gitdiff_delete(const char *line, struct patch *patch)
373 {
374 patch->is_delete = 1;
375 patch->old_name = patch->def_name;
376 return gitdiff_oldmode(line, patch);
377 }
378
379 static int gitdiff_newfile(const char *line, struct patch *patch)
380 {
381 patch->is_new = 1;
382 patch->new_name = patch->def_name;
383 return gitdiff_newmode(line, patch);
384 }
385
386 static int gitdiff_copysrc(const char *line, struct patch *patch)
387 {
388 patch->is_copy = 1;
389 patch->old_name = find_name(line, NULL, 0, 0);
390 return 0;
391 }
392
393 static int gitdiff_copydst(const char *line, struct patch *patch)
394 {
395 patch->is_copy = 1;
396 patch->new_name = find_name(line, NULL, 0, 0);
397 return 0;
398 }
399
400 static int gitdiff_renamesrc(const char *line, struct patch *patch)
401 {
402 patch->is_rename = 1;
403 patch->old_name = find_name(line, NULL, 0, 0);
404 return 0;
405 }
406
407 static int gitdiff_renamedst(const char *line, struct patch *patch)
408 {
409 patch->is_rename = 1;
410 patch->new_name = find_name(line, NULL, 0, 0);
411 return 0;
412 }
413
414 static int gitdiff_similarity(const char *line, struct patch *patch)
415 {
416 if ((patch->score = strtoul(line, NULL, 10)) == ULONG_MAX)
417 patch->score = 0;
418 return 0;
419 }
420
421 static int gitdiff_dissimilarity(const char *line, struct patch *patch)
422 {
423 if ((patch->score = strtoul(line, NULL, 10)) == ULONG_MAX)
424 patch->score = 0;
425 return 0;
426 }
427
428 static int gitdiff_index(const char *line, struct patch *patch)
429 {
430 /* index line is N hexadecimal, "..", N hexadecimal,
431 * and optional space with octal mode.
432 */
433 const char *ptr, *eol;
434 int len;
435
436 ptr = strchr(line, '.');
437 if (!ptr || ptr[1] != '.' || 40 < ptr - line)
438 return 0;
439 len = ptr - line;
440 memcpy(patch->old_sha1_prefix, line, len);
441 patch->old_sha1_prefix[len] = 0;
442
443 line = ptr + 2;
444 ptr = strchr(line, ' ');
445 eol = strchr(line, '\n');
446
447 if (!ptr || eol < ptr)
448 ptr = eol;
449 len = ptr - line;
450
451 if (40 < len)
452 return 0;
453 memcpy(patch->new_sha1_prefix, line, len);
454 patch->new_sha1_prefix[len] = 0;
455 if (*ptr == ' ')
456 patch->new_mode = patch->old_mode = strtoul(ptr+1, NULL, 8);
457 return 0;
458 }
459
460 /*
461 * This is normal for a diff that doesn't change anything: we'll fall through
462 * into the next diff. Tell the parser to break out.
463 */
464 static int gitdiff_unrecognized(const char *line, struct patch *patch)
465 {
466 return -1;
467 }
468
469 static const char *stop_at_slash(const char *line, int llen)
470 {
471 int i;
472
473 for (i = 0; i < llen; i++) {
474 int ch = line[i];
475 if (ch == '/')
476 return line + i;
477 }
478 return NULL;
479 }
480
481 /* This is to extract the same name that appears on "diff --git"
482 * line. We do not find and return anything if it is a rename
483 * patch, and it is OK because we will find the name elsewhere.
484 * We need to reliably find name only when it is mode-change only,
485 * creation or deletion of an empty file. In any of these cases,
486 * both sides are the same name under a/ and b/ respectively.
487 */
488 static char *git_header_name(char *line, int llen)
489 {
490 int len;
491 const char *name;
492 const char *second = NULL;
493
494 line += strlen("diff --git ");
495 llen -= strlen("diff --git ");
496
497 if (*line == '"') {
498 const char *cp;
499 char *first = unquote_c_style(line, &second);
500 if (!first)
501 return NULL;
502
503 /* advance to the first slash */
504 cp = stop_at_slash(first, strlen(first));
505 if (!cp || cp == first) {
506 /* we do not accept absolute paths */
507 free_first_and_fail:
508 free(first);
509 return NULL;
510 }
511 len = strlen(cp+1);
512 memmove(first, cp+1, len+1); /* including NUL */
513
514 /* second points at one past closing dq of name.
515 * find the second name.
516 */
517 while ((second < line + llen) && isspace(*second))
518 second++;
519
520 if (line + llen <= second)
521 goto free_first_and_fail;
522 if (*second == '"') {
523 char *sp = unquote_c_style(second, NULL);
524 if (!sp)
525 goto free_first_and_fail;
526 cp = stop_at_slash(sp, strlen(sp));
527 if (!cp || cp == sp) {
528 free_both_and_fail:
529 free(sp);
530 goto free_first_and_fail;
531 }
532 /* They must match, otherwise ignore */
533 if (strcmp(cp+1, first))
534 goto free_both_and_fail;
535 free(sp);
536 return first;
537 }
538
539 /* unquoted second */
540 cp = stop_at_slash(second, line + llen - second);
541 if (!cp || cp == second)
542 goto free_first_and_fail;
543 cp++;
544 if (line + llen - cp != len + 1 ||
545 memcmp(first, cp, len))
546 goto free_first_and_fail;
547 return first;
548 }
549
550 /* unquoted first name */
551 name = stop_at_slash(line, llen);
552 if (!name || name == line)
553 return NULL;
554
555 name++;
556
557 /* since the first name is unquoted, a dq if exists must be
558 * the beginning of the second name.
559 */
560 for (second = name; second < line + llen; second++) {
561 if (*second == '"') {
562 const char *cp = second;
563 const char *np;
564 char *sp = unquote_c_style(second, NULL);
565
566 if (!sp)
567 return NULL;
568 np = stop_at_slash(sp, strlen(sp));
569 if (!np || np == sp) {
570 free_second_and_fail:
571 free(sp);
572 return NULL;
573 }
574 np++;
575 len = strlen(np);
576 if (len < cp - name &&
577 !strncmp(np, name, len) &&
578 isspace(name[len])) {
579 /* Good */
580 memmove(sp, np, len + 1);
581 return sp;
582 }
583 goto free_second_and_fail;
584 }
585 }
586
587 /*
588 * Accept a name only if it shows up twice, exactly the same
589 * form.
590 */
591 for (len = 0 ; ; len++) {
592 char c = name[len];
593
594 switch (c) {
595 default:
596 continue;
597 case '\n':
598 return NULL;
599 case '\t': case ' ':
600 second = name+len;
601 for (;;) {
602 char c = *second++;
603 if (c == '\n')
604 return NULL;
605 if (c == '/')
606 break;
607 }
608 if (second[len] == '\n' && !memcmp(name, second, len)) {
609 char *ret = xmalloc(len + 1);
610 memcpy(ret, name, len);
611 ret[len] = 0;
612 return ret;
613 }
614 }
615 }
616 return NULL;
617 }
618
619 /* Verify that we recognize the lines following a git header */
620 static int parse_git_header(char *line, int len, unsigned int size, struct patch *patch)
621 {
622 unsigned long offset;
623
624 /* A git diff has explicit new/delete information, so we don't guess */
625 patch->is_new = 0;
626 patch->is_delete = 0;
627
628 /*
629 * Some things may not have the old name in the
630 * rest of the headers anywhere (pure mode changes,
631 * or removing or adding empty files), so we get
632 * the default name from the header.
633 */
634 patch->def_name = git_header_name(line, len);
635
636 line += len;
637 size -= len;
638 linenr++;
639 for (offset = len ; size > 0 ; offset += len, size -= len, line += len, linenr++) {
640 static const struct opentry {
641 const char *str;
642 int (*fn)(const char *, struct patch *);
643 } optable[] = {
644 { "@@ -", gitdiff_hdrend },
645 { "--- ", gitdiff_oldname },
646 { "+++ ", gitdiff_newname },
647 { "old mode ", gitdiff_oldmode },
648 { "new mode ", gitdiff_newmode },
649 { "deleted file mode ", gitdiff_delete },
650 { "new file mode ", gitdiff_newfile },
651 { "copy from ", gitdiff_copysrc },
652 { "copy to ", gitdiff_copydst },
653 { "rename old ", gitdiff_renamesrc },
654 { "rename new ", gitdiff_renamedst },
655 { "rename from ", gitdiff_renamesrc },
656 { "rename to ", gitdiff_renamedst },
657 { "similarity index ", gitdiff_similarity },
658 { "dissimilarity index ", gitdiff_dissimilarity },
659 { "index ", gitdiff_index },
660 { "", gitdiff_unrecognized },
661 };
662 int i;
663
664 len = linelen(line, size);
665 if (!len || line[len-1] != '\n')
666 break;
667 for (i = 0; i < ARRAY_SIZE(optable); i++) {
668 const struct opentry *p = optable + i;
669 int oplen = strlen(p->str);
670 if (len < oplen || memcmp(p->str, line, oplen))
671 continue;
672 if (p->fn(line + oplen, patch) < 0)
673 return offset;
674 break;
675 }
676 }
677
678 return offset;
679 }
680
681 static int parse_num(const char *line, unsigned long *p)
682 {
683 char *ptr;
684
685 if (!isdigit(*line))
686 return 0;
687 *p = strtoul(line, &ptr, 10);
688 return ptr - line;
689 }
690
691 static int parse_range(const char *line, int len, int offset, const char *expect,
692 unsigned long *p1, unsigned long *p2)
693 {
694 int digits, ex;
695
696 if (offset < 0 || offset >= len)
697 return -1;
698 line += offset;
699 len -= offset;
700
701 digits = parse_num(line, p1);
702 if (!digits)
703 return -1;
704
705 offset += digits;
706 line += digits;
707 len -= digits;
708
709 *p2 = 1;
710 if (*line == ',') {
711 digits = parse_num(line+1, p2);
712 if (!digits)
713 return -1;
714
715 offset += digits+1;
716 line += digits+1;
717 len -= digits+1;
718 }
719
720 ex = strlen(expect);
721 if (ex > len)
722 return -1;
723 if (memcmp(line, expect, ex))
724 return -1;
725
726 return offset + ex;
727 }
728
729 /*
730 * Parse a unified diff fragment header of the
731 * form "@@ -a,b +c,d @@"
732 */
733 static int parse_fragment_header(char *line, int len, struct fragment *fragment)
734 {
735 int offset;
736
737 if (!len || line[len-1] != '\n')
738 return -1;
739
740 /* Figure out the number of lines in a fragment */
741 offset = parse_range(line, len, 4, " +", &fragment->oldpos, &fragment->oldlines);
742 offset = parse_range(line, len, offset, " @@", &fragment->newpos, &fragment->newlines);
743
744 return offset;
745 }
746
747 static int find_header(char *line, unsigned long size, int *hdrsize, struct patch *patch)
748 {
749 unsigned long offset, len;
750
751 patch->is_rename = patch->is_copy = 0;
752 patch->is_new = patch->is_delete = -1;
753 patch->old_mode = patch->new_mode = 0;
754 patch->old_name = patch->new_name = NULL;
755 for (offset = 0; size > 0; offset += len, size -= len, line += len, linenr++) {
756 unsigned long nextlen;
757
758 len = linelen(line, size);
759 if (!len)
760 break;
761
762 /* Testing this early allows us to take a few shortcuts.. */
763 if (len < 6)
764 continue;
765
766 /*
767 * Make sure we don't find any unconnected patch fragmants.
768 * That's a sign that we didn't find a header, and that a
769 * patch has become corrupted/broken up.
770 */
771 if (!memcmp("@@ -", line, 4)) {
772 struct fragment dummy;
773 if (parse_fragment_header(line, len, &dummy) < 0)
774 continue;
775 error("patch fragment without header at line %d: %.*s", linenr, (int)len-1, line);
776 }
777
778 if (size < len + 6)
779 break;
780
781 /*
782 * Git patch? It might not have a real patch, just a rename
783 * or mode change, so we handle that specially
784 */
785 if (!memcmp("diff --git ", line, 11)) {
786 int git_hdr_len = parse_git_header(line, len, size, patch);
787 if (git_hdr_len <= len)
788 continue;
789 if (!patch->old_name && !patch->new_name) {
790 if (!patch->def_name)
791 die("git diff header lacks filename information (line %d)", linenr);
792 patch->old_name = patch->new_name = patch->def_name;
793 }
794 *hdrsize = git_hdr_len;
795 return offset;
796 }
797
798 /** --- followed by +++ ? */
799 if (memcmp("--- ", line, 4) || memcmp("+++ ", line + len, 4))
800 continue;
801
802 /*
803 * We only accept unified patches, so we want it to
804 * at least have "@@ -a,b +c,d @@\n", which is 14 chars
805 * minimum
806 */
807 nextlen = linelen(line + len, size - len);
808 if (size < nextlen + 14 || memcmp("@@ -", line + len + nextlen, 4))
809 continue;
810
811 /* Ok, we'll consider it a patch */
812 parse_traditional_patch(line, line+len, patch);
813 *hdrsize = len + nextlen;
814 linenr += 2;
815 return offset;
816 }
817 return -1;
818 }
819
820 /*
821 * Parse a unified diff. Note that this really needs
822 * to parse each fragment separately, since the only
823 * way to know the difference between a "---" that is
824 * part of a patch, and a "---" that starts the next
825 * patch is to look at the line counts..
826 */
827 static int parse_fragment(char *line, unsigned long size, struct patch *patch, struct fragment *fragment)
828 {
829 int added, deleted;
830 int len = linelen(line, size), offset;
831 unsigned long oldlines, newlines;
832 unsigned long leading, trailing;
833
834 offset = parse_fragment_header(line, len, fragment);
835 if (offset < 0)
836 return -1;
837 oldlines = fragment->oldlines;
838 newlines = fragment->newlines;
839 leading = 0;
840 trailing = 0;
841
842 if (patch->is_new < 0) {
843 patch->is_new = !oldlines;
844 if (!oldlines)
845 patch->old_name = NULL;
846 }
847 if (patch->is_delete < 0) {
848 patch->is_delete = !newlines;
849 if (!newlines)
850 patch->new_name = NULL;
851 }
852
853 if (patch->is_new && oldlines)
854 return error("new file depends on old contents");
855 if (patch->is_delete != !newlines) {
856 if (newlines)
857 return error("deleted file still has contents");
858 fprintf(stderr, "** warning: file %s becomes empty but is not deleted\n", patch->new_name);
859 }
860
861 /* Parse the thing.. */
862 line += len;
863 size -= len;
864 linenr++;
865 added = deleted = 0;
866 for (offset = len; size > 0; offset += len, size -= len, line += len, linenr++) {
867 if (!oldlines && !newlines)
868 break;
869 len = linelen(line, size);
870 if (!len || line[len-1] != '\n')
871 return -1;
872 switch (*line) {
873 default:
874 return -1;
875 case ' ':
876 oldlines--;
877 newlines--;
878 if (!deleted && !added)
879 leading++;
880 trailing++;
881 break;
882 case '-':
883 deleted++;
884 oldlines--;
885 trailing = 0;
886 break;
887 case '+':
888 /*
889 * We know len is at least two, since we have a '+' and
890 * we checked that the last character was a '\n' above.
891 * That is, an addition of an empty line would check
892 * the '+' here. Sneaky...
893 */
894 if ((new_whitespace != nowarn_whitespace) &&
895 isspace(line[len-2])) {
896 whitespace_error++;
897 if (squelch_whitespace_errors &&
898 squelch_whitespace_errors <
899 whitespace_error)
900 ;
901 else {
902 fprintf(stderr, "Adds trailing whitespace.\n%s:%d:%.*s\n",
903 patch_input_file,
904 linenr, len-2, line+1);
905 }
906 }
907 added++;
908 newlines--;
909 trailing = 0;
910 break;
911
912 /* We allow "\ No newline at end of file". Depending
913 * on locale settings when the patch was produced we
914 * don't know what this line looks like. The only
915 * thing we do know is that it begins with "\ ".
916 * Checking for 12 is just for sanity check -- any
917 * l10n of "\ No newline..." is at least that long.
918 */
919 case '\\':
920 if (len < 12 || memcmp(line, "\\ ", 2))
921 return -1;
922 break;
923 }
924 }
925 if (oldlines || newlines)
926 return -1;
927 fragment->leading = leading;
928 fragment->trailing = trailing;
929
930 /* If a fragment ends with an incomplete line, we failed to include
931 * it in the above loop because we hit oldlines == newlines == 0
932 * before seeing it.
933 */
934 if (12 < size && !memcmp(line, "\\ ", 2))
935 offset += linelen(line, size);
936
937 patch->lines_added += added;
938 patch->lines_deleted += deleted;
939 return offset;
940 }
941
942 static int parse_single_patch(char *line, unsigned long size, struct patch *patch)
943 {
944 unsigned long offset = 0;
945 struct fragment **fragp = &patch->fragments;
946
947 while (size > 4 && !memcmp(line, "@@ -", 4)) {
948 struct fragment *fragment;
949 int len;
950
951 fragment = xcalloc(1, sizeof(*fragment));
952 len = parse_fragment(line, size, patch, fragment);
953 if (len <= 0)
954 die("corrupt patch at line %d", linenr);
955
956 fragment->patch = line;
957 fragment->size = len;
958
959 *fragp = fragment;
960 fragp = &fragment->next;
961
962 offset += len;
963 line += len;
964 size -= len;
965 }
966 return offset;
967 }
968
969 static inline int metadata_changes(struct patch *patch)
970 {
971 return patch->is_rename > 0 ||
972 patch->is_copy > 0 ||
973 patch->is_new > 0 ||
974 patch->is_delete ||
975 (patch->old_mode && patch->new_mode &&
976 patch->old_mode != patch->new_mode);
977 }
978
979 static int parse_binary(char *buffer, unsigned long size, struct patch *patch)
980 {
981 /* We have read "GIT binary patch\n"; what follows is a line
982 * that says the patch method (currently, either "deflated
983 * literal" or "deflated delta") and the length of data before
984 * deflating; a sequence of 'length-byte' followed by base-85
985 * encoded data follows.
986 *
987 * Each 5-byte sequence of base-85 encodes up to 4 bytes,
988 * and we would limit the patch line to 66 characters,
989 * so one line can fit up to 13 groups that would decode
990 * to 52 bytes max. The length byte 'A'-'Z' corresponds
991 * to 1-26 bytes, and 'a'-'z' corresponds to 27-52 bytes.
992 * The end of binary is signalled with an empty line.
993 */
994 int llen, used;
995 struct fragment *fragment;
996 char *data = NULL;
997
998 patch->fragments = fragment = xcalloc(1, sizeof(*fragment));
999
1000 /* Grab the type of patch */
1001 llen = linelen(buffer, size);
1002 used = llen;
1003 linenr++;
1004
1005 if (!strncmp(buffer, "delta ", 6)) {
1006 patch->is_binary = BINARY_DELTA_DEFLATED;
1007 patch->deflate_origlen = strtoul(buffer + 6, NULL, 10);
1008 }
1009 else if (!strncmp(buffer, "literal ", 8)) {
1010 patch->is_binary = BINARY_LITERAL_DEFLATED;
1011 patch->deflate_origlen = strtoul(buffer + 8, NULL, 10);
1012 }
1013 else
1014 return error("unrecognized binary patch at line %d: %.*s",
1015 linenr-1, llen-1, buffer);
1016 buffer += llen;
1017 while (1) {
1018 int byte_length, max_byte_length, newsize;
1019 llen = linelen(buffer, size);
1020 used += llen;
1021 linenr++;
1022 if (llen == 1)
1023 break;
1024 /* Minimum line is "A00000\n" which is 7-byte long,
1025 * and the line length must be multiple of 5 plus 2.
1026 */
1027 if ((llen < 7) || (llen-2) % 5)
1028 goto corrupt;
1029 max_byte_length = (llen - 2) / 5 * 4;
1030 byte_length = *buffer;
1031 if ('A' <= byte_length && byte_length <= 'Z')
1032 byte_length = byte_length - 'A' + 1;
1033 else if ('a' <= byte_length && byte_length <= 'z')
1034 byte_length = byte_length - 'a' + 27;
1035 else
1036 goto corrupt;
1037 /* if the input length was not multiple of 4, we would
1038 * have filler at the end but the filler should never
1039 * exceed 3 bytes
1040 */
1041 if (max_byte_length < byte_length ||
1042 byte_length <= max_byte_length - 4)
1043 goto corrupt;
1044 newsize = fragment->size + byte_length;
1045 data = xrealloc(data, newsize);
1046 if (decode_85(data + fragment->size,
1047 buffer + 1,
1048 byte_length))
1049 goto corrupt;
1050 fragment->size = newsize;
1051 buffer += llen;
1052 size -= llen;
1053 }
1054 fragment->patch = data;
1055 return used;
1056 corrupt:
1057 return error("corrupt binary patch at line %d: %.*s",
1058 linenr-1, llen-1, buffer);
1059 }
1060
1061 static int parse_chunk(char *buffer, unsigned long size, struct patch *patch)
1062 {
1063 int hdrsize, patchsize;
1064 int offset = find_header(buffer, size, &hdrsize, patch);
1065
1066 if (offset < 0)
1067 return offset;
1068
1069 patchsize = parse_single_patch(buffer + offset + hdrsize, size - offset - hdrsize, patch);
1070
1071 if (!patchsize) {
1072 static const char *binhdr[] = {
1073 "Binary files ",
1074 "Files ",
1075 NULL,
1076 };
1077 static const char git_binary[] = "GIT binary patch\n";
1078 int i;
1079 int hd = hdrsize + offset;
1080 unsigned long llen = linelen(buffer + hd, size - hd);
1081
1082 if (llen == sizeof(git_binary) - 1 &&
1083 !memcmp(git_binary, buffer + hd, llen)) {
1084 int used;
1085 linenr++;
1086 used = parse_binary(buffer + hd + llen,
1087 size - hd - llen, patch);
1088 if (used)
1089 patchsize = used + llen;
1090 else
1091 patchsize = 0;
1092 }
1093 else if (!memcmp(" differ\n", buffer + hd + llen - 8, 8)) {
1094 for (i = 0; binhdr[i]; i++) {
1095 int len = strlen(binhdr[i]);
1096 if (len < size - hd &&
1097 !memcmp(binhdr[i], buffer + hd, len)) {
1098 linenr++;
1099 patch->is_binary = 1;
1100 patchsize = llen;
1101 break;
1102 }
1103 }
1104 }
1105
1106 /* Empty patch cannot be applied if:
1107 * - it is a binary patch and we do not do binary_replace, or
1108 * - text patch without metadata change
1109 */
1110 if ((apply || check) &&
1111 (patch->is_binary
1112 ? !allow_binary_replacement
1113 : !metadata_changes(patch)))
1114 die("patch with only garbage at line %d", linenr);
1115 }
1116
1117 return offset + hdrsize + patchsize;
1118 }
1119
1120 static const char pluses[] = "++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++";
1121 static const char minuses[]= "----------------------------------------------------------------------";
1122
1123 static void show_stats(struct patch *patch)
1124 {
1125 const char *prefix = "";
1126 char *name = patch->new_name;
1127 char *qname = NULL;
1128 int len, max, add, del, total;
1129
1130 if (!name)
1131 name = patch->old_name;
1132
1133 if (0 < (len = quote_c_style(name, NULL, NULL, 0))) {
1134 qname = xmalloc(len + 1);
1135 quote_c_style(name, qname, NULL, 0);
1136 name = qname;
1137 }
1138
1139 /*
1140 * "scale" the filename
1141 */
1142 len = strlen(name);
1143 max = max_len;
1144 if (max > 50)
1145 max = 50;
1146 if (len > max) {
1147 char *slash;
1148 prefix = "...";
1149 max -= 3;
1150 name += len - max;
1151 slash = strchr(name, '/');
1152 if (slash)
1153 name = slash;
1154 }
1155 len = max;
1156
1157 /*
1158 * scale the add/delete
1159 */
1160 max = max_change;
1161 if (max + len > 70)
1162 max = 70 - len;
1163
1164 add = patch->lines_added;
1165 del = patch->lines_deleted;
1166 total = add + del;
1167
1168 if (max_change > 0) {
1169 total = (total * max + max_change / 2) / max_change;
1170 add = (add * max + max_change / 2) / max_change;
1171 del = total - add;
1172 }
1173 if (patch->is_binary)
1174 printf(" %s%-*s | Bin\n", prefix, len, name);
1175 else
1176 printf(" %s%-*s |%5d %.*s%.*s\n", prefix,
1177 len, name, patch->lines_added + patch->lines_deleted,
1178 add, pluses, del, minuses);
1179 if (qname)
1180 free(qname);
1181 }
1182
1183 static int read_old_data(struct stat *st, const char *path, void *buf, unsigned long size)
1184 {
1185 int fd;
1186 unsigned long got;
1187
1188 switch (st->st_mode & S_IFMT) {
1189 case S_IFLNK:
1190 return readlink(path, buf, size);
1191 case S_IFREG:
1192 fd = open(path, O_RDONLY);
1193 if (fd < 0)
1194 return error("unable to open %s", path);
1195 got = 0;
1196 for (;;) {
1197 int ret = xread(fd, buf + got, size - got);
1198 if (ret <= 0)
1199 break;
1200 got += ret;
1201 }
1202 close(fd);
1203 return got;
1204
1205 default:
1206 return -1;
1207 }
1208 }
1209
1210 static int find_offset(const char *buf, unsigned long size, const char *fragment, unsigned long fragsize, int line, int *lines)
1211 {
1212 int i;
1213 unsigned long start, backwards, forwards;
1214
1215 if (fragsize > size)
1216 return -1;
1217
1218 start = 0;
1219 if (line > 1) {
1220 unsigned long offset = 0;
1221 i = line-1;
1222 while (offset + fragsize <= size) {
1223 if (buf[offset++] == '\n') {
1224 start = offset;
1225 if (!--i)
1226 break;
1227 }
1228 }
1229 }
1230
1231 /* Exact line number? */
1232 if (!memcmp(buf + start, fragment, fragsize))
1233 return start;
1234
1235 /*
1236 * There's probably some smart way to do this, but I'll leave
1237 * that to the smart and beautiful people. I'm simple and stupid.
1238 */
1239 backwards = start;
1240 forwards = start;
1241 for (i = 0; ; i++) {
1242 unsigned long try;
1243 int n;
1244
1245 /* "backward" */
1246 if (i & 1) {
1247 if (!backwards) {
1248 if (forwards + fragsize > size)
1249 break;
1250 continue;
1251 }
1252 do {
1253 --backwards;
1254 } while (backwards && buf[backwards-1] != '\n');
1255 try = backwards;
1256 } else {
1257 while (forwards + fragsize <= size) {
1258 if (buf[forwards++] == '\n')
1259 break;
1260 }
1261 try = forwards;
1262 }
1263
1264 if (try + fragsize > size)
1265 continue;
1266 if (memcmp(buf + try, fragment, fragsize))
1267 continue;
1268 n = (i >> 1)+1;
1269 if (i & 1)
1270 n = -n;
1271 *lines = n;
1272 return try;
1273 }
1274
1275 /*
1276 * We should start searching forward and backward.
1277 */
1278 return -1;
1279 }
1280
1281 static void remove_first_line(const char **rbuf, int *rsize)
1282 {
1283 const char *buf = *rbuf;
1284 int size = *rsize;
1285 unsigned long offset;
1286 offset = 0;
1287 while (offset <= size) {
1288 if (buf[offset++] == '\n')
1289 break;
1290 }
1291 *rsize = size - offset;
1292 *rbuf = buf + offset;
1293 }
1294
1295 static void remove_last_line(const char **rbuf, int *rsize)
1296 {
1297 const char *buf = *rbuf;
1298 int size = *rsize;
1299 unsigned long offset;
1300 offset = size - 1;
1301 while (offset > 0) {
1302 if (buf[--offset] == '\n')
1303 break;
1304 }
1305 *rsize = offset + 1;
1306 }
1307
1308 struct buffer_desc {
1309 char *buffer;
1310 unsigned long size;
1311 unsigned long alloc;
1312 };
1313
1314 static int apply_line(char *output, const char *patch, int plen)
1315 {
1316 /* plen is number of bytes to be copied from patch,
1317 * starting at patch+1 (patch[0] is '+'). Typically
1318 * patch[plen] is '\n'.
1319 */
1320 int add_nl_to_tail = 0;
1321 if ((new_whitespace == strip_whitespace) &&
1322 1 < plen && isspace(patch[plen-1])) {
1323 if (patch[plen] == '\n')
1324 add_nl_to_tail = 1;
1325 plen--;
1326 while (0 < plen && isspace(patch[plen]))
1327 plen--;
1328 applied_after_stripping++;
1329 }
1330 memcpy(output, patch + 1, plen);
1331 if (add_nl_to_tail)
1332 output[plen++] = '\n';
1333 return plen;
1334 }
1335
1336 static int apply_one_fragment(struct buffer_desc *desc, struct fragment *frag)
1337 {
1338 char *buf = desc->buffer;
1339 const char *patch = frag->patch;
1340 int offset, size = frag->size;
1341 char *old = xmalloc(size);
1342 char *new = xmalloc(size);
1343 const char *oldlines, *newlines;
1344 int oldsize = 0, newsize = 0;
1345 unsigned long leading, trailing;
1346 int pos, lines;
1347
1348 while (size > 0) {
1349 int len = linelen(patch, size);
1350 int plen;
1351
1352 if (!len)
1353 break;
1354
1355 /*
1356 * "plen" is how much of the line we should use for
1357 * the actual patch data. Normally we just remove the
1358 * first character on the line, but if the line is
1359 * followed by "\ No newline", then we also remove the
1360 * last one (which is the newline, of course).
1361 */
1362 plen = len-1;
1363 if (len < size && patch[len] == '\\')
1364 plen--;
1365 switch (*patch) {
1366 case ' ':
1367 case '-':
1368 memcpy(old + oldsize, patch + 1, plen);
1369 oldsize += plen;
1370 if (*patch == '-')
1371 break;
1372 /* Fall-through for ' ' */
1373 case '+':
1374 if (*patch != '+' || !no_add)
1375 newsize += apply_line(new + newsize, patch,
1376 plen);
1377 break;
1378 case '@': case '\\':
1379 /* Ignore it, we already handled it */
1380 break;
1381 default:
1382 return -1;
1383 }
1384 patch += len;
1385 size -= len;
1386 }
1387
1388 #ifdef NO_ACCURATE_DIFF
1389 if (oldsize > 0 && old[oldsize - 1] == '\n' &&
1390 newsize > 0 && new[newsize - 1] == '\n') {
1391 oldsize--;
1392 newsize--;
1393 }
1394 #endif
1395
1396 oldlines = old;
1397 newlines = new;
1398 leading = frag->leading;
1399 trailing = frag->trailing;
1400 lines = 0;
1401 pos = frag->newpos;
1402 for (;;) {
1403 offset = find_offset(buf, desc->size, oldlines, oldsize, pos, &lines);
1404 if (offset >= 0) {
1405 int diff = newsize - oldsize;
1406 unsigned long size = desc->size + diff;
1407 unsigned long alloc = desc->alloc;
1408
1409 /* Warn if it was necessary to reduce the number
1410 * of context lines.
1411 */
1412 if ((leading != frag->leading) || (trailing != frag->trailing))
1413 fprintf(stderr, "Context reduced to (%ld/%ld) to apply fragment at %d\n",
1414 leading, trailing, pos + lines);
1415
1416 if (size > alloc) {
1417 alloc = size + 8192;
1418 desc->alloc = alloc;
1419 buf = xrealloc(buf, alloc);
1420 desc->buffer = buf;
1421 }
1422 desc->size = size;
1423 memmove(buf + offset + newsize, buf + offset + oldsize, size - offset - newsize);
1424 memcpy(buf + offset, newlines, newsize);
1425 offset = 0;
1426
1427 break;
1428 }
1429
1430 /* Am I at my context limits? */
1431 if ((leading <= p_context) && (trailing <= p_context))
1432 break;
1433 /* Reduce the number of context lines
1434 * Reduce both leading and trailing if they are equal
1435 * otherwise just reduce the larger context.
1436 */
1437 if (leading >= trailing) {
1438 remove_first_line(&oldlines, &oldsize);
1439 remove_first_line(&newlines, &newsize);
1440 pos--;
1441 leading--;
1442 }
1443 if (trailing > leading) {
1444 remove_last_line(&oldlines, &oldsize);
1445 remove_last_line(&newlines, &newsize);
1446 trailing--;
1447 }
1448 }
1449
1450 free(old);
1451 free(new);
1452 return offset;
1453 }
1454
1455 static char *inflate_it(const void *data, unsigned long size,
1456 unsigned long inflated_size)
1457 {
1458 z_stream stream;
1459 void *out;
1460 int st;
1461
1462 memset(&stream, 0, sizeof(stream));
1463
1464 stream.next_in = (unsigned char *)data;
1465 stream.avail_in = size;
1466 stream.next_out = out = xmalloc(inflated_size);
1467 stream.avail_out = inflated_size;
1468 inflateInit(&stream);
1469 st = inflate(&stream, Z_FINISH);
1470 if ((st != Z_STREAM_END) || stream.total_out != inflated_size) {
1471 free(out);
1472 return NULL;
1473 }
1474 return out;
1475 }
1476
1477 static int apply_binary_fragment(struct buffer_desc *desc, struct patch *patch)
1478 {
1479 unsigned long dst_size;
1480 struct fragment *fragment = patch->fragments;
1481 void *data;
1482 void *result;
1483
1484 data = inflate_it(fragment->patch, fragment->size,
1485 patch->deflate_origlen);
1486 if (!data)
1487 return error("corrupt patch data");
1488 switch (patch->is_binary) {
1489 case BINARY_DELTA_DEFLATED:
1490 result = patch_delta(desc->buffer, desc->size,
1491 data,
1492 patch->deflate_origlen,
1493 &dst_size);
1494 free(desc->buffer);
1495 desc->buffer = result;
1496 free(data);
1497 break;
1498 case BINARY_LITERAL_DEFLATED:
1499 free(desc->buffer);
1500 desc->buffer = data;
1501 dst_size = patch->deflate_origlen;
1502 break;
1503 }
1504 if (!desc->buffer)
1505 return -1;
1506 desc->size = desc->alloc = dst_size;
1507 return 0;
1508 }
1509
1510 static int apply_binary(struct buffer_desc *desc, struct patch *patch)
1511 {
1512 const char *name = patch->old_name ? patch->old_name : patch->new_name;
1513 unsigned char sha1[20];
1514 unsigned char hdr[50];
1515 int hdrlen;
1516
1517 if (!allow_binary_replacement)
1518 return error("cannot apply binary patch to '%s' "
1519 "without --allow-binary-replacement",
1520 name);
1521
1522 /* For safety, we require patch index line to contain
1523 * full 40-byte textual SHA1 for old and new, at least for now.
1524 */
1525 if (strlen(patch->old_sha1_prefix) != 40 ||
1526 strlen(patch->new_sha1_prefix) != 40 ||
1527 get_sha1_hex(patch->old_sha1_prefix, sha1) ||
1528 get_sha1_hex(patch->new_sha1_prefix, sha1))
1529 return error("cannot apply binary patch to '%s' "
1530 "without full index line", name);
1531
1532 if (patch->old_name) {
1533 /* See if the old one matches what the patch
1534 * applies to.
1535 */
1536 write_sha1_file_prepare(desc->buffer, desc->size,
1537 blob_type, sha1, hdr, &hdrlen);
1538 if (strcmp(sha1_to_hex(sha1), patch->old_sha1_prefix))
1539 return error("the patch applies to '%s' (%s), "
1540 "which does not match the "
1541 "current contents.",
1542 name, sha1_to_hex(sha1));
1543 }
1544 else {
1545 /* Otherwise, the old one must be empty. */
1546 if (desc->size)
1547 return error("the patch applies to an empty "
1548 "'%s' but it is not empty", name);
1549 }
1550
1551 get_sha1_hex(patch->new_sha1_prefix, sha1);
1552 if (!memcmp(sha1, null_sha1, 20)) {
1553 free(desc->buffer);
1554 desc->alloc = desc->size = 0;
1555 desc->buffer = NULL;
1556 return 0; /* deletion patch */
1557 }
1558
1559 if (has_sha1_file(sha1)) {
1560 /* We already have the postimage */
1561 char type[10];
1562 unsigned long size;
1563
1564 free(desc->buffer);
1565 desc->buffer = read_sha1_file(sha1, type, &size);
1566 if (!desc->buffer)
1567 return error("the necessary postimage %s for "
1568 "'%s' cannot be read",
1569 patch->new_sha1_prefix, name);
1570 desc->alloc = desc->size = size;
1571 }
1572 else {
1573 /* We have verified desc matches the preimage;
1574 * apply the patch data to it, which is stored
1575 * in the patch->fragments->{patch,size}.
1576 */
1577 if (apply_binary_fragment(desc, patch))
1578 return error("binary patch does not apply to '%s'",
1579 name);
1580
1581 /* verify that the result matches */
1582 write_sha1_file_prepare(desc->buffer, desc->size, blob_type,
1583 sha1, hdr, &hdrlen);
1584 if (strcmp(sha1_to_hex(sha1), patch->new_sha1_prefix))
1585 return error("binary patch to '%s' creates incorrect result", name);
1586 }
1587
1588 return 0;
1589 }
1590
1591 static int apply_fragments(struct buffer_desc *desc, struct patch *patch)
1592 {
1593 struct fragment *frag = patch->fragments;
1594 const char *name = patch->old_name ? patch->old_name : patch->new_name;
1595
1596 if (patch->is_binary)
1597 return apply_binary(desc, patch);
1598
1599 while (frag) {
1600 if (apply_one_fragment(desc, frag) < 0)
1601 return error("patch failed: %s:%ld",
1602 name, frag->oldpos);
1603 frag = frag->next;
1604 }
1605 return 0;
1606 }
1607
1608 static int apply_data(struct patch *patch, struct stat *st, struct cache_entry *ce)
1609 {
1610 char *buf;
1611 unsigned long size, alloc;
1612 struct buffer_desc desc;
1613
1614 size = 0;
1615 alloc = 0;
1616 buf = NULL;
1617 if (cached) {
1618 if (ce) {
1619 char type[20];
1620 buf = read_sha1_file(ce->sha1, type, &size);
1621 if (!buf)
1622 return error("read of %s failed",
1623 patch->old_name);
1624 alloc = size;
1625 }
1626 }
1627 else if (patch->old_name) {
1628 size = st->st_size;
1629 alloc = size + 8192;
1630 buf = xmalloc(alloc);
1631 if (read_old_data(st, patch->old_name, buf, alloc) != size)
1632 return error("read of %s failed", patch->old_name);
1633 }
1634
1635 desc.size = size;
1636 desc.alloc = alloc;
1637 desc.buffer = buf;
1638 if (apply_fragments(&desc, patch) < 0)
1639 return -1;
1640 patch->result = desc.buffer;
1641 patch->resultsize = desc.size;
1642
1643 if (patch->is_delete && patch->resultsize)
1644 return error("removal patch leaves file contents");
1645
1646 return 0;
1647 }
1648
1649 static int check_patch(struct patch *patch)
1650 {
1651 struct stat st;
1652 const char *old_name = patch->old_name;
1653 const char *new_name = patch->new_name;
1654 const char *name = old_name ? old_name : new_name;
1655 struct cache_entry *ce = NULL;
1656
1657 if (old_name) {
1658 int changed = 0;
1659 int stat_ret = 0;
1660 unsigned st_mode = 0;
1661
1662 if (!cached)
1663 stat_ret = lstat(old_name, &st);
1664 if (check_index) {
1665 int pos = cache_name_pos(old_name, strlen(old_name));
1666 if (pos < 0)
1667 return error("%s: does not exist in index",
1668 old_name);
1669 ce = active_cache[pos];
1670 if (stat_ret < 0) {
1671 struct checkout costate;
1672 if (errno != ENOENT)
1673 return error("%s: %s", old_name,
1674 strerror(errno));
1675 /* checkout */
1676 costate.base_dir = "";
1677 costate.base_dir_len = 0;
1678 costate.force = 0;
1679 costate.quiet = 0;
1680 costate.not_new = 0;
1681 costate.refresh_cache = 1;
1682 if (checkout_entry(ce,
1683 &costate,
1684 NULL) ||
1685 lstat(old_name, &st))
1686 return -1;
1687 }
1688 if (!cached)
1689 changed = ce_match_stat(ce, &st, 1);
1690 if (changed)
1691 return error("%s: does not match index",
1692 old_name);
1693 if (cached)
1694 st_mode = ntohl(ce->ce_mode);
1695 }
1696 else if (stat_ret < 0)
1697 return error("%s: %s", old_name, strerror(errno));
1698
1699 if (!cached)
1700 st_mode = ntohl(create_ce_mode(st.st_mode));
1701
1702 if (patch->is_new < 0)
1703 patch->is_new = 0;
1704 if (!patch->old_mode)
1705 patch->old_mode = st_mode;
1706 if ((st_mode ^ patch->old_mode) & S_IFMT)
1707 return error("%s: wrong type", old_name);
1708 if (st_mode != patch->old_mode)
1709 fprintf(stderr, "warning: %s has type %o, expected %o\n",
1710 old_name, st_mode, patch->old_mode);
1711 }
1712
1713 if (new_name && (patch->is_new | patch->is_rename | patch->is_copy)) {
1714 if (check_index && cache_name_pos(new_name, strlen(new_name)) >= 0)
1715 return error("%s: already exists in index", new_name);
1716 if (!cached) {
1717 if (!lstat(new_name, &st))
1718 return error("%s: already exists in working directory", new_name);
1719 if (errno != ENOENT)
1720 return error("%s: %s", new_name, strerror(errno));
1721 }
1722 if (!patch->new_mode) {
1723 if (patch->is_new)
1724 patch->new_mode = S_IFREG | 0644;
1725 else
1726 patch->new_mode = patch->old_mode;
1727 }
1728 }
1729
1730 if (new_name && old_name) {
1731 int same = !strcmp(old_name, new_name);
1732 if (!patch->new_mode)
1733 patch->new_mode = patch->old_mode;
1734 if ((patch->old_mode ^ patch->new_mode) & S_IFMT)
1735 return error("new mode (%o) of %s does not match old mode (%o)%s%s",
1736 patch->new_mode, new_name, patch->old_mode,
1737 same ? "" : " of ", same ? "" : old_name);
1738 }
1739
1740 if (apply_data(patch, &st, ce) < 0)
1741 return error("%s: patch does not apply", name);
1742 return 0;
1743 }
1744
1745 static int check_patch_list(struct patch *patch)
1746 {
1747 int error = 0;
1748
1749 for (;patch ; patch = patch->next)
1750 error |= check_patch(patch);
1751 return error;
1752 }
1753
1754 static inline int is_null_sha1(const unsigned char *sha1)
1755 {
1756 return !memcmp(sha1, null_sha1, 20);
1757 }
1758
1759 static void show_index_list(struct patch *list)
1760 {
1761 struct patch *patch;
1762
1763 /* Once we start supporting the reverse patch, it may be
1764 * worth showing the new sha1 prefix, but until then...
1765 */
1766 for (patch = list; patch; patch = patch->next) {
1767 const unsigned char *sha1_ptr;
1768 unsigned char sha1[20];
1769 const char *name;
1770
1771 name = patch->old_name ? patch->old_name : patch->new_name;
1772 if (patch->is_new)
1773 sha1_ptr = null_sha1;
1774 else if (get_sha1(patch->old_sha1_prefix, sha1))
1775 die("sha1 information is lacking or useless (%s).",
1776 name);
1777 else
1778 sha1_ptr = sha1;
1779
1780 printf("%06o %s ",patch->old_mode, sha1_to_hex(sha1_ptr));
1781 if (line_termination && quote_c_style(name, NULL, NULL, 0))
1782 quote_c_style(name, NULL, stdout, 0);
1783 else
1784 fputs(name, stdout);
1785 putchar(line_termination);
1786 }
1787 }
1788
1789 static void stat_patch_list(struct patch *patch)
1790 {
1791 int files, adds, dels;
1792
1793 for (files = adds = dels = 0 ; patch ; patch = patch->next) {
1794 files++;
1795 adds += patch->lines_added;
1796 dels += patch->lines_deleted;
1797 show_stats(patch);
1798 }
1799
1800 printf(" %d files changed, %d insertions(+), %d deletions(-)\n", files, adds, dels);
1801 }
1802
1803 static void numstat_patch_list(struct patch *patch)
1804 {
1805 for ( ; patch; patch = patch->next) {
1806 const char *name;
1807 name = patch->new_name ? patch->new_name : patch->old_name;
1808 printf("%d\t%d\t", patch->lines_added, patch->lines_deleted);
1809 if (line_termination && quote_c_style(name, NULL, NULL, 0))
1810 quote_c_style(name, NULL, stdout, 0);
1811 else
1812 fputs(name, stdout);
1813 putchar('\n');
1814 }
1815 }
1816
1817 static void show_file_mode_name(const char *newdelete, unsigned int mode, const char *name)
1818 {
1819 if (mode)
1820 printf(" %s mode %06o %s\n", newdelete, mode, name);
1821 else
1822 printf(" %s %s\n", newdelete, name);
1823 }
1824
1825 static void show_mode_change(struct patch *p, int show_name)
1826 {
1827 if (p->old_mode && p->new_mode && p->old_mode != p->new_mode) {
1828 if (show_name)
1829 printf(" mode change %06o => %06o %s\n",
1830 p->old_mode, p->new_mode, p->new_name);
1831 else
1832 printf(" mode change %06o => %06o\n",
1833 p->old_mode, p->new_mode);
1834 }
1835 }
1836
1837 static void show_rename_copy(struct patch *p)
1838 {
1839 const char *renamecopy = p->is_rename ? "rename" : "copy";
1840 const char *old, *new;
1841
1842 /* Find common prefix */
1843 old = p->old_name;
1844 new = p->new_name;
1845 while (1) {
1846 const char *slash_old, *slash_new;
1847 slash_old = strchr(old, '/');
1848 slash_new = strchr(new, '/');
1849 if (!slash_old ||
1850 !slash_new ||
1851 slash_old - old != slash_new - new ||
1852 memcmp(old, new, slash_new - new))
1853 break;
1854 old = slash_old + 1;
1855 new = slash_new + 1;
1856 }
1857 /* p->old_name thru old is the common prefix, and old and new
1858 * through the end of names are renames
1859 */
1860 if (old != p->old_name)
1861 printf(" %s %.*s{%s => %s} (%d%%)\n", renamecopy,
1862 (int)(old - p->old_name), p->old_name,
1863 old, new, p->score);
1864 else
1865 printf(" %s %s => %s (%d%%)\n", renamecopy,
1866 p->old_name, p->new_name, p->score);
1867 show_mode_change(p, 0);
1868 }
1869
1870 static void summary_patch_list(struct patch *patch)
1871 {
1872 struct patch *p;
1873
1874 for (p = patch; p; p = p->next) {
1875 if (p->is_new)
1876 show_file_mode_name("create", p->new_mode, p->new_name);
1877 else if (p->is_delete)
1878 show_file_mode_name("delete", p->old_mode, p->old_name);
1879 else {
1880 if (p->is_rename || p->is_copy)
1881 show_rename_copy(p);
1882 else {
1883 if (p->score) {
1884 printf(" rewrite %s (%d%%)\n",
1885 p->new_name, p->score);
1886 show_mode_change(p, 0);
1887 }
1888 else
1889 show_mode_change(p, 1);
1890 }
1891 }
1892 }
1893 }
1894
1895 static void patch_stats(struct patch *patch)
1896 {
1897 int lines = patch->lines_added + patch->lines_deleted;
1898
1899 if (lines > max_change)
1900 max_change = lines;
1901 if (patch->old_name) {
1902 int len = quote_c_style(patch->old_name, NULL, NULL, 0);
1903 if (!len)
1904 len = strlen(patch->old_name);
1905 if (len > max_len)
1906 max_len = len;
1907 }
1908 if (patch->new_name) {
1909 int len = quote_c_style(patch->new_name, NULL, NULL, 0);
1910 if (!len)
1911 len = strlen(patch->new_name);
1912 if (len > max_len)
1913 max_len = len;
1914 }
1915 }
1916
1917 static void remove_file(struct patch *patch)
1918 {
1919 if (write_index) {
1920 if (remove_file_from_cache(patch->old_name) < 0)
1921 die("unable to remove %s from index", patch->old_name);
1922 cache_tree_invalidate_path(active_cache_tree, patch->old_name);
1923 }
1924 if (!cached)
1925 unlink(patch->old_name);
1926 }
1927
1928 static void add_index_file(const char *path, unsigned mode, void *buf, unsigned long size)
1929 {
1930 struct stat st;
1931 struct cache_entry *ce;
1932 int namelen = strlen(path);
1933 unsigned ce_size = cache_entry_size(namelen);
1934
1935 if (!write_index)
1936 return;
1937
1938 ce = xcalloc(1, ce_size);
1939 memcpy(ce->name, path, namelen);
1940 ce->ce_mode = create_ce_mode(mode);
1941 ce->ce_flags = htons(namelen);
1942 if (!cached) {
1943 if (lstat(path, &st) < 0)
1944 die("unable to stat newly created file %s", path);
1945 fill_stat_cache_info(ce, &st);
1946 }
1947 if (write_sha1_file(buf, size, blob_type, ce->sha1) < 0)
1948 die("unable to create backing store for newly created file %s", path);
1949 if (add_cache_entry(ce, ADD_CACHE_OK_TO_ADD) < 0)
1950 die("unable to add cache entry for %s", path);
1951 }
1952
1953 static int try_create_file(const char *path, unsigned int mode, const char *buf, unsigned long size)
1954 {
1955 int fd;
1956
1957 if (S_ISLNK(mode))
1958 return symlink(buf, path);
1959 fd = open(path, O_CREAT | O_EXCL | O_WRONLY, (mode & 0100) ? 0777 : 0666);
1960 if (fd < 0)
1961 return -1;
1962 while (size) {
1963 int written = xwrite(fd, buf, size);
1964 if (written < 0)
1965 die("writing file %s: %s", path, strerror(errno));
1966 if (!written)
1967 die("out of space writing file %s", path);
1968 buf += written;
1969 size -= written;
1970 }
1971 if (close(fd) < 0)
1972 die("closing file %s: %s", path, strerror(errno));
1973 return 0;
1974 }
1975
1976 /*
1977 * We optimistically assume that the directories exist,
1978 * which is true 99% of the time anyway. If they don't,
1979 * we create them and try again.
1980 */
1981 static void create_one_file(char *path, unsigned mode, const char *buf, unsigned long size)
1982 {
1983 if (cached)
1984 return;
1985 if (!try_create_file(path, mode, buf, size))
1986 return;
1987
1988 if (errno == ENOENT) {
1989 if (safe_create_leading_directories(path))
1990 return;
1991 if (!try_create_file(path, mode, buf, size))
1992 return;
1993 }
1994
1995 if (errno == EEXIST) {
1996 unsigned int nr = getpid();
1997
1998 for (;;) {
1999 const char *newpath;
2000 newpath = mkpath("%s~%u", path, nr);
2001 if (!try_create_file(newpath, mode, buf, size)) {
2002 if (!rename(newpath, path))
2003 return;
2004 unlink(newpath);
2005 break;
2006 }
2007 if (errno != EEXIST)
2008 break;
2009 ++nr;
2010 }
2011 }
2012 die("unable to write file %s mode %o", path, mode);
2013 }
2014
2015 static void create_file(struct patch *patch)
2016 {
2017 char *path = patch->new_name;
2018 unsigned mode = patch->new_mode;
2019 unsigned long size = patch->resultsize;
2020 char *buf = patch->result;
2021
2022 if (!mode)
2023 mode = S_IFREG | 0644;
2024 create_one_file(path, mode, buf, size);
2025 add_index_file(path, mode, buf, size);
2026 cache_tree_invalidate_path(active_cache_tree, path);
2027 }
2028
2029 static void write_out_one_result(struct patch *patch)
2030 {
2031 if (patch->is_delete > 0) {
2032 remove_file(patch);
2033 return;
2034 }
2035 if (patch->is_new > 0 || patch->is_copy) {
2036 create_file(patch);
2037 return;
2038 }
2039 /*
2040 * Rename or modification boils down to the same
2041 * thing: remove the old, write the new
2042 */
2043 remove_file(patch);
2044 create_file(patch);
2045 }
2046
2047 static void write_out_results(struct patch *list, int skipped_patch)
2048 {
2049 if (!list && !skipped_patch)
2050 die("No changes");
2051
2052 while (list) {
2053 write_out_one_result(list);
2054 list = list->next;
2055 }
2056 }
2057
2058 static struct cache_file cache_file;
2059
2060 static struct excludes {
2061 struct excludes *next;
2062 const char *path;
2063 } *excludes;
2064
2065 static int use_patch(struct patch *p)
2066 {
2067 const char *pathname = p->new_name ? p->new_name : p->old_name;
2068 struct excludes *x = excludes;
2069 while (x) {
2070 if (fnmatch(x->path, pathname, 0) == 0)
2071 return 0;
2072 x = x->next;
2073 }
2074 if (0 < prefix_length) {
2075 int pathlen = strlen(pathname);
2076 if (pathlen <= prefix_length ||
2077 memcmp(prefix, pathname, prefix_length))
2078 return 0;
2079 }
2080 return 1;
2081 }
2082
2083 static int apply_patch(int fd, const char *filename)
2084 {
2085 unsigned long offset, size;
2086 char *buffer = read_patch_file(fd, &size);
2087 struct patch *list = NULL, **listp = &list;
2088 int skipped_patch = 0;
2089
2090 patch_input_file = filename;
2091 if (!buffer)
2092 return -1;
2093 offset = 0;
2094 while (size > 0) {
2095 struct patch *patch;
2096 int nr;
2097
2098 patch = xcalloc(1, sizeof(*patch));
2099 nr = parse_chunk(buffer + offset, size, patch);
2100 if (nr < 0)
2101 break;
2102 if (use_patch(patch)) {
2103 patch_stats(patch);
2104 *listp = patch;
2105 listp = &patch->next;
2106 } else {
2107 /* perhaps free it a bit better? */
2108 free(patch);
2109 skipped_patch++;
2110 }
2111 offset += nr;
2112 size -= nr;
2113 }
2114
2115 if (whitespace_error && (new_whitespace == error_on_whitespace))
2116 apply = 0;
2117
2118 write_index = check_index && apply;
2119 if (write_index && newfd < 0)
2120 newfd = hold_index_file_for_update(&cache_file, get_index_file());
2121 if (check_index) {
2122 if (read_cache() < 0)
2123 die("unable to read index file");
2124 }
2125
2126 if ((check || apply) && check_patch_list(list) < 0)
2127 exit(1);
2128
2129 if (apply)
2130 write_out_results(list, skipped_patch);
2131
2132 if (show_index_info)
2133 show_index_list(list);
2134
2135 if (diffstat)
2136 stat_patch_list(list);
2137
2138 if (numstat)
2139 numstat_patch_list(list);
2140
2141 if (summary)
2142 summary_patch_list(list);
2143
2144 free(buffer);
2145 return 0;
2146 }
2147
2148 static int git_apply_config(const char *var, const char *value)
2149 {
2150 if (!strcmp(var, "apply.whitespace")) {
2151 apply_default_whitespace = strdup(value);
2152 return 0;
2153 }
2154 return git_default_config(var, value);
2155 }
2156
2157
2158 int cmd_apply(int argc, const char **argv, char **envp)
2159 {
2160 int i;
2161 int read_stdin = 1;
2162 const char *whitespace_option = NULL;
2163
2164 for (i = 1; i < argc; i++) {
2165 const char *arg = argv[i];
2166 char *end;
2167 int fd;
2168
2169 if (!strcmp(arg, "-")) {
2170 apply_patch(0, "<stdin>");
2171 read_stdin = 0;
2172 continue;
2173 }
2174 if (!strncmp(arg, "--exclude=", 10)) {
2175 struct excludes *x = xmalloc(sizeof(*x));
2176 x->path = arg + 10;
2177 x->next = excludes;
2178 excludes = x;
2179 continue;
2180 }
2181 if (!strncmp(arg, "-p", 2)) {
2182 p_value = atoi(arg + 2);
2183 continue;
2184 }
2185 if (!strcmp(arg, "--no-add")) {
2186 no_add = 1;
2187 continue;
2188 }
2189 if (!strcmp(arg, "--stat")) {
2190 apply = 0;
2191 diffstat = 1;
2192 continue;
2193 }
2194 if (!strcmp(arg, "--allow-binary-replacement") ||
2195 !strcmp(arg, "--binary")) {
2196 allow_binary_replacement = 1;
2197 continue;
2198 }
2199 if (!strcmp(arg, "--numstat")) {
2200 apply = 0;
2201 numstat = 1;
2202 continue;
2203 }
2204 if (!strcmp(arg, "--summary")) {
2205 apply = 0;
2206 summary = 1;
2207 continue;
2208 }
2209 if (!strcmp(arg, "--check")) {
2210 apply = 0;
2211 check = 1;
2212 continue;
2213 }
2214 if (!strcmp(arg, "--index")) {
2215 check_index = 1;
2216 continue;
2217 }
2218 if (!strcmp(arg, "--cached")) {
2219 check_index = 1;
2220 cached = 1;
2221 continue;
2222 }
2223 if (!strcmp(arg, "--apply")) {
2224 apply = 1;
2225 continue;
2226 }
2227 if (!strcmp(arg, "--index-info")) {
2228 apply = 0;
2229 show_index_info = 1;
2230 continue;
2231 }
2232 if (!strcmp(arg, "-z")) {
2233 line_termination = 0;
2234 continue;
2235 }
2236 if (!strncmp(arg, "-C", 2)) {
2237 p_context = strtoul(arg + 2, &end, 0);
2238 if (*end != '\0')
2239 die("unrecognized context count '%s'", arg + 2);
2240 continue;
2241 }
2242 if (!strncmp(arg, "--whitespace=", 13)) {
2243 whitespace_option = arg + 13;
2244 parse_whitespace_option(arg + 13);
2245 continue;
2246 }
2247
2248 if (check_index && prefix_length < 0) {
2249 prefix = setup_git_directory();
2250 prefix_length = prefix ? strlen(prefix) : 0;
2251 git_config(git_apply_config);
2252 if (!whitespace_option && apply_default_whitespace)
2253 parse_whitespace_option(apply_default_whitespace);
2254 }
2255 if (0 < prefix_length)
2256 arg = prefix_filename(prefix, prefix_length, arg);
2257
2258 fd = open(arg, O_RDONLY);
2259 if (fd < 0)
2260 usage(apply_usage);
2261 read_stdin = 0;
2262 set_default_whitespace_mode(whitespace_option);
2263 apply_patch(fd, arg);
2264 close(fd);
2265 }
2266 set_default_whitespace_mode(whitespace_option);
2267 if (read_stdin)
2268 apply_patch(0, "<stdin>");
2269 if (whitespace_error) {
2270 if (squelch_whitespace_errors &&
2271 squelch_whitespace_errors < whitespace_error) {
2272 int squelched =
2273 whitespace_error - squelch_whitespace_errors;
2274 fprintf(stderr, "warning: squelched %d whitespace error%s\n",
2275 squelched,
2276 squelched == 1 ? "" : "s");
2277 }
2278 if (new_whitespace == error_on_whitespace)
2279 die("%d line%s add%s trailing whitespaces.",
2280 whitespace_error,
2281 whitespace_error == 1 ? "" : "s",
2282 whitespace_error == 1 ? "s" : "");
2283 if (applied_after_stripping)
2284 fprintf(stderr, "warning: %d line%s applied after"
2285 " stripping trailing whitespaces.\n",
2286 applied_after_stripping,
2287 applied_after_stripping == 1 ? "" : "s");
2288 else if (whitespace_error)
2289 fprintf(stderr, "warning: %d line%s add%s trailing"
2290 " whitespaces.\n",
2291 whitespace_error,
2292 whitespace_error == 1 ? "" : "s",
2293 whitespace_error == 1 ? "s" : "");
2294 }
2295
2296 if (write_index) {
2297 if (write_cache(newfd, active_cache, active_nr) ||
2298 commit_index_file(&cache_file))
2299 die("Unable to write new cachefile");
2300 }
2301
2302 return 0;
2303 }