apply: convert root string to strbuf
[git/git.git] / builtin / apply.c
CommitLineData
c1bb9350
LT
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 *
c1bb9350 8 */
c1bb9350 9#include "cache.h"
697cc8ef 10#include "lockfile.h"
03ac6e64 11#include "cache-tree.h"
22943f1a 12#include "quote.h"
8e440259 13#include "blob.h"
051308f6 14#include "delta.h"
ac6245e3 15#include "builtin.h"
c455c87c 16#include "string-list.h"
175a4948 17#include "dir.h"
7f814632 18#include "diff.h"
f26c4940 19#include "parse-options.h"
28ff0512
JH
20#include "xdiff-interface.h"
21#include "ll-merge.h"
f2633ebd 22#include "rerere.h"
c1bb9350 23
a9486b02
PR
24/*
25 * --check turns on checking that the working tree matches the
26 * files that are being modified, but doesn't apply the patch
27 * --stat does just a diffstat, and doesn't actually apply
28 * --numstat does numeric diffstat, and doesn't actually apply
29 * --index-info shows the old and new index info for paths if available.
30 * --index updates the cache as well.
31 * --cached updates only the cache without ever touching the working tree.
32 */
edf2e370
JH
33static const char *prefix;
34static int prefix_length = -1;
dbd0f7d3 35static int newfd = -1;
edf2e370 36
4be60962 37static int unidiff_zero;
e36f8b60 38static int p_value = 1;
3e8a5db9 39static int p_value_known;
96f1e58f 40static int check_index;
7da3bf37 41static int update_index;
96f1e58f
DR
42static int cached;
43static int diffstat;
44static int numstat;
45static int summary;
46static int check;
a577284a 47static int apply = 1;
96f1e58f 48static int apply_in_reverse;
57dc397c 49static int apply_with_reject;
a2bf404e 50static int apply_verbosely;
933e44d3 51static int allow_overlap;
96f1e58f 52static int no_add;
cfb6f9ac 53static int threeway;
c536c075 54static int unsafe_paths;
7a988699 55static const char *fake_ancestor;
22943f1a 56static int line_termination = '\n';
f26c4940
MV
57static unsigned int p_context = UINT_MAX;
58static const char * const apply_usage[] = {
9c9b4f2f 59 N_("git apply [<options>] [<patch>...]"),
f26c4940
MV
60 NULL
61};
c1bb9350 62
81bf96bb
JH
63static enum ws_error_action {
64 nowarn_ws_error,
65 warn_on_ws_error,
66 die_on_ws_error,
4b05548f 67 correct_ws_error
81bf96bb 68} ws_error_action = warn_on_ws_error;
96f1e58f 69static int whitespace_error;
fc96b7c9 70static int squelch_whitespace_errors = 5;
c94bf41c 71static int applied_after_fixing_ws;
86c91f91
GB
72
73static enum ws_ignore {
74 ignore_ws_none,
4b05548f 75 ignore_ws_change
86c91f91
GB
76} ws_ignore_action = ignore_ws_none;
77
78
96f1e58f 79static const char *patch_input_file;
6c31c22c 80static struct strbuf root = STRBUF_INIT;
f26c4940
MV
81static int read_stdin = 1;
82static int options;
19bfcd5a 83
2ae1c53b
JH
84static void parse_whitespace_option(const char *option)
85{
86 if (!option) {
81bf96bb 87 ws_error_action = warn_on_ws_error;
2ae1c53b
JH
88 return;
89 }
90 if (!strcmp(option, "warn")) {
81bf96bb 91 ws_error_action = warn_on_ws_error;
2ae1c53b
JH
92 return;
93 }
621603b7 94 if (!strcmp(option, "nowarn")) {
81bf96bb 95 ws_error_action = nowarn_ws_error;
621603b7
JH
96 return;
97 }
2ae1c53b 98 if (!strcmp(option, "error")) {
81bf96bb 99 ws_error_action = die_on_ws_error;
2ae1c53b
JH
100 return;
101 }
102 if (!strcmp(option, "error-all")) {
81bf96bb 103 ws_error_action = die_on_ws_error;
2ae1c53b
JH
104 squelch_whitespace_errors = 0;
105 return;
106 }
81bf96bb
JH
107 if (!strcmp(option, "strip") || !strcmp(option, "fix")) {
108 ws_error_action = correct_ws_error;
2ae1c53b
JH
109 return;
110 }
3638eb43 111 die(_("unrecognized whitespace option '%s'"), option);
2ae1c53b
JH
112}
113
86c91f91
GB
114static void parse_ignorewhitespace_option(const char *option)
115{
116 if (!option || !strcmp(option, "no") ||
117 !strcmp(option, "false") || !strcmp(option, "never") ||
118 !strcmp(option, "none")) {
119 ws_ignore_action = ignore_ws_none;
120 return;
121 }
122 if (!strcmp(option, "change")) {
123 ws_ignore_action = ignore_ws_change;
124 return;
125 }
3638eb43 126 die(_("unrecognized whitespace ignore option '%s'"), option);
86c91f91
GB
127}
128
f21d6726
JH
129static void set_default_whitespace_mode(const char *whitespace_option)
130{
81bf96bb
JH
131 if (!whitespace_option && !apply_default_whitespace)
132 ws_error_action = (apply ? warn_on_ws_error : nowarn_ws_error);
f21d6726
JH
133}
134
3f40315a
LT
135/*
136 * For "diff-stat" like behaviour, we keep track of the biggest change
137 * we've seen, and the longest filename. That allows us to do simple
138 * scaling.
139 */
140static int max_change, max_len;
141
a4acb0eb
LT
142/*
143 * Various "current state", notably line numbers and what
144 * file (and how) we're patching right now.. The "is_xxxx"
145 * things are flags, where -1 means "don't know yet".
146 */
46979f56 147static int linenr = 1;
19c58fb8 148
3cd4f5e8
JH
149/*
150 * This represents one "hunk" from a patch, starting with
151 * "@@ -oldpos,oldlines +newpos,newlines @@" marker. The
152 * patch text is pointed at by patch, and its byte length
153 * is stored in size. leading and trailing are the number
154 * of context lines.
155 */
19c58fb8 156struct fragment {
47495887 157 unsigned long leading, trailing;
19c58fb8
LT
158 unsigned long oldpos, oldlines;
159 unsigned long newpos, newlines;
92737a22
JH
160 /*
161 * 'patch' is usually borrowed from buf in apply_patch(),
162 * but some codepaths store an allocated buffer.
163 */
19c58fb8 164 const char *patch;
6fe53908
JH
165 unsigned free_patch:1,
166 rejected:1;
19c58fb8 167 int size;
77b15bbd 168 int linenr;
19c58fb8
LT
169 struct fragment *next;
170};
171
3cd4f5e8
JH
172/*
173 * When dealing with a binary patch, we reuse "leading" field
174 * to store the type of the binary hunk, either deflated "delta"
175 * or deflated "literal".
176 */
177#define binary_patch_method leading
178#define BINARY_DELTA_DEFLATED 1
179#define BINARY_LITERAL_DEFLATED 2
180
81bf96bb
JH
181/*
182 * This represents a "patch" to a file, both metainfo changes
183 * such as creation/deletion, filemode and content changes represented
184 * as a series of fragments.
185 */
19c58fb8 186struct patch {
5041aa70 187 char *new_name, *old_name, *def_name;
19c58fb8 188 unsigned int old_mode, new_mode;
3dad11bf 189 int is_new, is_delete; /* -1 = unknown, 0 = false, 1 = true */
57dc397c 190 int rejected;
cf1b7869 191 unsigned ws_rule;
3f40315a 192 int lines_added, lines_deleted;
96c912a4 193 int score;
9987d7c5 194 unsigned int is_toplevel_relative:1;
3dad11bf
RS
195 unsigned int inaccurate_eof:1;
196 unsigned int is_binary:1;
197 unsigned int is_copy:1;
198 unsigned int is_rename:1;
c14b9d1e 199 unsigned int recount:1;
28ff0512 200 unsigned int conflicted_threeway:1;
099f3c42 201 unsigned int direct_to_threeway:1;
19c58fb8 202 struct fragment *fragments;
5aa7d94c 203 char *result;
c32f749f 204 size_t resultsize;
2cf67f1e
JH
205 char old_sha1_prefix[41];
206 char new_sha1_prefix[41];
19c58fb8 207 struct patch *next;
28ff0512
JH
208
209 /* three-way fallback result */
d07d4ab4 210 struct object_id threeway_stage[3];
19c58fb8 211};
46979f56 212
9d16c2d5 213static void free_fragment_list(struct fragment *list)
6fe53908 214{
9d16c2d5
JH
215 while (list) {
216 struct fragment *next = list->next;
217 if (list->free_patch)
218 free((char *)list->patch);
219 free(list);
220 list = next;
a604ddef 221 }
9d16c2d5
JH
222}
223
224static void free_patch(struct patch *patch)
225{
226 free_fragment_list(patch->fragments);
2901bbe5
JH
227 free(patch->def_name);
228 free(patch->old_name);
229 free(patch->new_name);
8192a2fa 230 free(patch->result);
a604ddef
JH
231 free(patch);
232}
233
234static void free_patch_list(struct patch *list)
235{
236 while (list) {
237 struct patch *next = list->next;
238 free_patch(list);
239 list = next;
6fe53908
JH
240 }
241}
242
b94f2eda
JH
243/*
244 * A line in a file, len-bytes long (includes the terminating LF,
245 * except for an incomplete line at the end if the file ends with
246 * one), and its contents hashes to 'hash'.
247 */
248struct line {
249 size_t len;
250 unsigned hash : 24;
251 unsigned flag : 8;
c330fdd4 252#define LINE_COMMON 1
9d158601 253#define LINE_PATCHED 2
b94f2eda
JH
254};
255
256/*
257 * This represents a "file", which is an array of "lines".
258 */
259struct image {
260 char *buf;
261 size_t len;
262 size_t nr;
c330fdd4 263 size_t alloc;
b94f2eda
JH
264 struct line *line_allocated;
265 struct line *line;
266};
267
7a07841c
DZ
268/*
269 * Records filenames that have been touched, in order to handle
270 * the case where more than one patches touch the same file.
271 */
272
c455c87c 273static struct string_list fn_table;
7a07841c 274
b94f2eda
JH
275static uint32_t hash_line(const char *cp, size_t len)
276{
277 size_t i;
278 uint32_t h;
279 for (i = 0, h = 0; i < len; i++) {
280 if (!isspace(cp[i])) {
281 h = h * 3 + (cp[i] & 0xff);
282 }
283 }
284 return h;
285}
286
86c91f91
GB
287/*
288 * Compare lines s1 of length n1 and s2 of length n2, ignoring
289 * whitespace difference. Returns 1 if they match, 0 otherwise
290 */
291static int fuzzy_matchlines(const char *s1, size_t n1,
292 const char *s2, size_t n2)
293{
294 const char *last1 = s1 + n1 - 1;
295 const char *last2 = s2 + n2 - 1;
296 int result = 0;
297
86c91f91
GB
298 /* ignore line endings */
299 while ((*last1 == '\r') || (*last1 == '\n'))
300 last1--;
301 while ((*last2 == '\r') || (*last2 == '\n'))
302 last2--;
303
14d3bb49
JH
304 /* skip leading whitespaces, if both begin with whitespace */
305 if (s1 <= last1 && s2 <= last2 && isspace(*s1) && isspace(*s2)) {
306 while (isspace(*s1) && (s1 <= last1))
307 s1++;
308 while (isspace(*s2) && (s2 <= last2))
309 s2++;
310 }
86c91f91
GB
311 /* early return if both lines are empty */
312 if ((s1 > last1) && (s2 > last2))
313 return 1;
314 while (!result) {
315 result = *s1++ - *s2++;
316 /*
317 * Skip whitespace inside. We check for whitespace on
318 * both buffers because we don't want "a b" to match
319 * "ab"
320 */
321 if (isspace(*s1) && isspace(*s2)) {
322 while (isspace(*s1) && s1 <= last1)
323 s1++;
324 while (isspace(*s2) && s2 <= last2)
325 s2++;
326 }
327 /*
328 * If we reached the end on one side only,
329 * lines don't match
330 */
331 if (
332 ((s2 > last2) && (s1 <= last1)) ||
333 ((s1 > last1) && (s2 <= last2)))
334 return 0;
335 if ((s1 > last1) && (s2 > last2))
336 break;
337 }
338
339 return !result;
340}
341
c330fdd4
JH
342static void add_line_info(struct image *img, const char *bol, size_t len, unsigned flag)
343{
344 ALLOC_GROW(img->line_allocated, img->nr + 1, img->alloc);
345 img->line_allocated[img->nr].len = len;
346 img->line_allocated[img->nr].hash = hash_line(bol, len);
347 img->line_allocated[img->nr].flag = flag;
348 img->nr++;
349}
350
92737a22
JH
351/*
352 * "buf" has the file contents to be patched (read from various sources).
353 * attach it to "image" and add line-based index to it.
354 * "image" now owns the "buf".
355 */
b94f2eda
JH
356static void prepare_image(struct image *image, char *buf, size_t len,
357 int prepare_linetable)
358{
359 const char *cp, *ep;
b94f2eda 360
c330fdd4 361 memset(image, 0, sizeof(*image));
b94f2eda
JH
362 image->buf = buf;
363 image->len = len;
364
c330fdd4 365 if (!prepare_linetable)
b94f2eda 366 return;
b94f2eda
JH
367
368 ep = image->buf + image->len;
b94f2eda 369 cp = image->buf;
b94f2eda
JH
370 while (cp < ep) {
371 const char *next;
372 for (next = cp; next < ep && *next != '\n'; next++)
373 ;
374 if (next < ep)
375 next++;
c330fdd4 376 add_line_info(image, cp, next - cp, 0);
b94f2eda 377 cp = next;
b94f2eda 378 }
c330fdd4 379 image->line = image->line_allocated;
b94f2eda
JH
380}
381
382static void clear_image(struct image *image)
383{
384 free(image->buf);
798b9ce8
JH
385 free(image->line_allocated);
386 memset(image, 0, sizeof(*image));
b94f2eda
JH
387}
388
5613e811
NTND
389/* fmt must contain _one_ %s and no other substitution */
390static void say_patch_name(FILE *output, const char *fmt, struct patch *patch)
a2bf404e 391{
5613e811
NTND
392 struct strbuf sb = STRBUF_INIT;
393
a2bf404e
JH
394 if (patch->old_name && patch->new_name &&
395 strcmp(patch->old_name, patch->new_name)) {
5613e811
NTND
396 quote_c_style(patch->old_name, &sb, NULL, 0);
397 strbuf_addstr(&sb, " => ");
398 quote_c_style(patch->new_name, &sb, NULL, 0);
663af342 399 } else {
a2bf404e
JH
400 const char *n = patch->new_name;
401 if (!n)
402 n = patch->old_name;
5613e811 403 quote_c_style(n, &sb, NULL, 0);
a2bf404e 404 }
5613e811
NTND
405 fprintf(output, fmt, sb.buf);
406 fputc('\n', output);
407 strbuf_release(&sb);
a2bf404e
JH
408}
409
a4acb0eb 410#define SLOP (16)
c1bb9350 411
9a76adeb 412static void read_patch_file(struct strbuf *sb, int fd)
c1bb9350 413{
9a76adeb 414 if (strbuf_read(sb, fd, 0) < 0)
d824cbba 415 die_errno("git apply: failed to read");
a4acb0eb
LT
416
417 /*
418 * Make sure that we have some slop in the buffer
419 * so that we can do speculative "memcmp" etc, and
420 * see to it that it is NUL-filled.
421 */
9a76adeb
PH
422 strbuf_grow(sb, SLOP);
423 memset(sb->buf + sb->len, 0, SLOP);
c1bb9350
LT
424}
425
3cca928d 426static unsigned long linelen(const char *buffer, unsigned long size)
c1bb9350
LT
427{
428 unsigned long len = 0;
429 while (size--) {
430 len++;
431 if (*buffer++ == '\n')
432 break;
433 }
434 return len;
435}
436
a4acb0eb
LT
437static int is_dev_null(const char *str)
438{
e3f1da98 439 return skip_prefix(str, "/dev/null", &str) && isspace(*str);
a4acb0eb
LT
440}
441
381ca9a3
LT
442#define TERM_SPACE 1
443#define TERM_TAB 2
9a4a100e
LT
444
445static int name_terminate(const char *name, int namelen, int c, int terminate)
446{
447 if (c == ' ' && !(terminate & TERM_SPACE))
448 return 0;
449 if (c == '\t' && !(terminate & TERM_TAB))
450 return 0;
451
9a4a100e
LT
452 return 1;
453}
454
33eb4dd9
MM
455/* remove double slashes to make --index work with such filenames */
456static char *squash_slash(char *name)
457{
458 int i = 0, j = 0;
459
15862087
AG
460 if (!name)
461 return NULL;
462
33eb4dd9
MM
463 while (name[i]) {
464 if ((name[j++] = name[i++]) == '/')
465 while (name[i] == '/')
466 i++;
467 }
468 name[j] = '\0';
469 return name;
470}
471
2901bbe5 472static char *find_name_gnu(const char *line, const char *def, int p_value)
c1bb9350 473{
bb7306b5
JN
474 struct strbuf name = STRBUF_INIT;
475 char *cp;
15862087 476
bb7306b5
JN
477 /*
478 * Proposed "new-style" GNU patch/diff format; see
efe6de6e 479 * http://marc.info/?l=git&m=112927316408690&w=2
bb7306b5
JN
480 */
481 if (unquote_c_style(&name, line, NULL)) {
482 strbuf_release(&name);
483 return NULL;
484 }
7fb1011e 485
bb7306b5
JN
486 for (cp = name.buf; p_value; p_value--) {
487 cp = strchr(cp, '/');
488 if (!cp) {
489 strbuf_release(&name);
490 return NULL;
22943f1a 491 }
bb7306b5
JN
492 cp++;
493 }
494
bb7306b5 495 strbuf_remove(&name, 0, cp - name.buf);
6c31c22c
JK
496 if (root.len)
497 strbuf_insert(&name, 0, root.buf, root.len);
bb7306b5
JN
498 return squash_slash(strbuf_detach(&name, NULL));
499}
500
2d502e1f 501static size_t sane_tz_len(const char *line, size_t len)
c1bb9350 502{
5a12c886 503 const char *tz, *p;
15862087 504
5a12c886
JN
505 if (len < strlen(" +0500") || line[len-strlen(" +0500")] != ' ')
506 return 0;
507 tz = line + len - strlen(" +0500");
508
509 if (tz[1] != '+' && tz[1] != '-')
510 return 0;
511
512 for (p = tz + 2; p != line + len; p++)
513 if (!isdigit(*p))
514 return 0;
515
516 return line + len - tz;
517}
518
2d502e1f
JN
519static size_t tz_with_colon_len(const char *line, size_t len)
520{
521 const char *tz, *p;
522
523 if (len < strlen(" +08:00") || line[len - strlen(":00")] != ':')
524 return 0;
525 tz = line + len - strlen(" +08:00");
526
527 if (tz[0] != ' ' || (tz[1] != '+' && tz[1] != '-'))
528 return 0;
529 p = tz + 2;
530 if (!isdigit(*p++) || !isdigit(*p++) || *p++ != ':' ||
531 !isdigit(*p++) || !isdigit(*p++))
532 return 0;
533
534 return line + len - tz;
535}
536
5a12c886
JN
537static size_t date_len(const char *line, size_t len)
538{
539 const char *date, *p;
540
541 if (len < strlen("72-02-05") || line[len-strlen("-05")] != '-')
542 return 0;
543 p = date = line + len - strlen("72-02-05");
544
545 if (!isdigit(*p++) || !isdigit(*p++) || *p++ != '-' ||
546 !isdigit(*p++) || !isdigit(*p++) || *p++ != '-' ||
547 !isdigit(*p++) || !isdigit(*p++)) /* Not a date. */
548 return 0;
549
550 if (date - line >= strlen("19") &&
551 isdigit(date[-1]) && isdigit(date[-2])) /* 4-digit year */
552 date -= strlen("19");
553
554 return line + len - date;
555}
556
557static size_t short_time_len(const char *line, size_t len)
558{
559 const char *time, *p;
560
561 if (len < strlen(" 07:01:32") || line[len-strlen(":32")] != ':')
562 return 0;
563 p = time = line + len - strlen(" 07:01:32");
564
565 /* Permit 1-digit hours? */
566 if (*p++ != ' ' ||
567 !isdigit(*p++) || !isdigit(*p++) || *p++ != ':' ||
568 !isdigit(*p++) || !isdigit(*p++) || *p++ != ':' ||
569 !isdigit(*p++) || !isdigit(*p++)) /* Not a time. */
570 return 0;
571
572 return line + len - time;
573}
574
575static size_t fractional_time_len(const char *line, size_t len)
576{
577 const char *p;
578 size_t n;
579
580 /* Expected format: 19:41:17.620000023 */
581 if (!len || !isdigit(line[len - 1]))
582 return 0;
583 p = line + len - 1;
584
585 /* Fractional seconds. */
586 while (p > line && isdigit(*p))
587 p--;
588 if (*p != '.')
589 return 0;
590
591 /* Hours, minutes, and whole seconds. */
592 n = short_time_len(line, p - line);
593 if (!n)
594 return 0;
595
596 return line + len - p + n;
597}
598
599static size_t trailing_spaces_len(const char *line, size_t len)
600{
601 const char *p;
602
603 /* Expected format: ' ' x (1 or more) */
604 if (!len || line[len - 1] != ' ')
605 return 0;
606
607 p = line + len;
608 while (p != line) {
609 p--;
610 if (*p != ' ')
611 return line + len - (p + 1);
22943f1a
JH
612 }
613
5a12c886
JN
614 /* All spaces! */
615 return len;
616}
617
618static size_t diff_timestamp_len(const char *line, size_t len)
619{
620 const char *end = line + len;
621 size_t n;
622
623 /*
624 * Posix: 2010-07-05 19:41:17
625 * GNU: 2010-07-05 19:41:17.620000023 -0500
626 */
627
628 if (!isdigit(end[-1]))
629 return 0;
630
2d502e1f
JN
631 n = sane_tz_len(line, end - line);
632 if (!n)
633 n = tz_with_colon_len(line, end - line);
5a12c886
JN
634 end -= n;
635
636 n = short_time_len(line, end - line);
637 if (!n)
638 n = fractional_time_len(line, end - line);
639 end -= n;
640
641 n = date_len(line, end - line);
642 if (!n) /* No date. Too bad. */
643 return 0;
644 end -= n;
645
646 if (end == line) /* No space before date. */
647 return 0;
648 if (end[-1] == '\t') { /* Success! */
649 end--;
650 return line + len - end;
651 }
652 if (end[-1] != ' ') /* No space before date. */
653 return 0;
654
655 /* Whitespace damage. */
656 end -= trailing_spaces_len(line, end - line);
657 return line + len - end;
658}
659
2901bbe5
JH
660static char *find_name_common(const char *line, const char *def,
661 int p_value, const char *end, int terminate)
5a12c886
JN
662{
663 int len;
664 const char *start = NULL;
665
bb7306b5
JN
666 if (p_value == 0)
667 start = line;
5a12c886 668 while (line != end) {
a4acb0eb 669 char c = *line;
9a4a100e 670
5a12c886 671 if (!end && isspace(c)) {
9a4a100e
LT
672 if (c == '\n')
673 break;
674 if (name_terminate(start, line-start, c, terminate))
675 break;
676 }
a4acb0eb
LT
677 line++;
678 if (c == '/' && !--p_value)
679 start = line;
680 }
681 if (!start)
44406907 682 return squash_slash(xstrdup_or_null(def));
a4acb0eb
LT
683 len = line - start;
684 if (!len)
44406907 685 return squash_slash(xstrdup_or_null(def));
a4acb0eb
LT
686
687 /*
688 * Generally we prefer the shorter name, especially
689 * if the other one is just a variation of that with
690 * something else tacked on to the end (ie "file.orig"
691 * or "file~").
692 */
693 if (def) {
694 int deflen = strlen(def);
695 if (deflen < len && !strncmp(start, def, deflen))
2901bbe5 696 return squash_slash(xstrdup(def));
c1bb9350 697 }
a4acb0eb 698
6c31c22c
JK
699 if (root.len) {
700 char *ret = xstrfmt("%s%.*s", root.buf, len, start);
33eb4dd9 701 return squash_slash(ret);
c4730f35
JS
702 }
703
33eb4dd9 704 return squash_slash(xmemdupz(start, len));
a4acb0eb
LT
705}
706
5a12c886
JN
707static char *find_name(const char *line, char *def, int p_value, int terminate)
708{
709 if (*line == '"') {
710 char *name = find_name_gnu(line, def, p_value);
711 if (name)
712 return name;
713 }
714
715 return find_name_common(line, def, p_value, NULL, terminate);
716}
717
718static char *find_name_traditional(const char *line, char *def, int p_value)
719{
1f976bd0 720 size_t len;
5a12c886
JN
721 size_t date_len;
722
723 if (*line == '"') {
724 char *name = find_name_gnu(line, def, p_value);
725 if (name)
726 return name;
727 }
728
729 len = strchrnul(line, '\n') - line;
730 date_len = diff_timestamp_len(line, len);
731 if (!date_len)
732 return find_name_common(line, def, p_value, NULL, TERM_TAB);
733 len -= date_len;
734
735 return find_name_common(line, def, p_value, line + len, 0);
736}
737
3e8a5db9
JH
738static int count_slashes(const char *cp)
739{
740 int cnt = 0;
741 char ch;
742
743 while ((ch = *cp++))
744 if (ch == '/')
745 cnt++;
746 return cnt;
747}
748
749/*
750 * Given the string after "--- " or "+++ ", guess the appropriate
751 * p_value for the given patch.
752 */
753static int guess_p_value(const char *nameline)
754{
755 char *name, *cp;
756 int val = -1;
757
758 if (is_dev_null(nameline))
759 return -1;
5a12c886 760 name = find_name_traditional(nameline, NULL, 0);
3e8a5db9
JH
761 if (!name)
762 return -1;
763 cp = strchr(name, '/');
764 if (!cp)
765 val = 0;
766 else if (prefix) {
767 /*
768 * Does it begin with "a/$our-prefix" and such? Then this is
769 * very likely to apply to our directory.
770 */
771 if (!strncmp(name, prefix, prefix_length))
772 val = count_slashes(prefix);
773 else {
774 cp++;
775 if (!strncmp(cp, prefix, prefix_length))
776 val = count_slashes(prefix) + 1;
777 }
778 }
779 free(name);
780 return val;
781}
782
c4593faf 783/*
5fcadc3b 784 * Does the ---/+++ line have the POSIX timestamp after the last HT?
c4593faf
JH
785 * GNU diff puts epoch there to signal a creation/deletion event. Is
786 * this such a timestamp?
787 */
788static int has_epoch_timestamp(const char *nameline)
789{
790 /*
791 * We are only interested in epoch timestamp; any non-zero
792 * fraction cannot be one, hence "(\.0+)?" in the regexp below.
793 * For the same reason, the date must be either 1969-12-31 or
794 * 1970-01-01, and the seconds part must be "00".
795 */
796 const char stamp_regexp[] =
797 "^(1969-12-31|1970-01-01)"
798 " "
799 "[0-2][0-9]:[0-5][0-9]:00(\\.0+)?"
800 " "
a1980c4e
AK
801 "([-+][0-2][0-9]:?[0-5][0-9])\n";
802 const char *timestamp = NULL, *cp, *colon;
c4593faf
JH
803 static regex_t *stamp;
804 regmatch_t m[10];
805 int zoneoffset;
806 int hourminute;
807 int status;
808
809 for (cp = nameline; *cp != '\n'; cp++) {
810 if (*cp == '\t')
811 timestamp = cp + 1;
812 }
813 if (!timestamp)
814 return 0;
815 if (!stamp) {
816 stamp = xmalloc(sizeof(*stamp));
817 if (regcomp(stamp, stamp_regexp, REG_EXTENDED)) {
3638eb43 818 warning(_("Cannot prepare timestamp regexp %s"),
c4593faf
JH
819 stamp_regexp);
820 return 0;
821 }
822 }
823
824 status = regexec(stamp, timestamp, ARRAY_SIZE(m), m, 0);
825 if (status) {
826 if (status != REG_NOMATCH)
3638eb43 827 warning(_("regexec returned %d for input: %s"),
c4593faf
JH
828 status, timestamp);
829 return 0;
830 }
831
a1980c4e
AK
832 zoneoffset = strtol(timestamp + m[3].rm_so + 1, (char **) &colon, 10);
833 if (*colon == ':')
834 zoneoffset = zoneoffset * 60 + strtol(colon + 1, NULL, 10);
835 else
836 zoneoffset = (zoneoffset / 100) * 60 + (zoneoffset % 100);
c4593faf
JH
837 if (timestamp[m[3].rm_so] == '-')
838 zoneoffset = -zoneoffset;
839
840 /*
841 * YYYY-MM-DD hh:mm:ss must be from either 1969-12-31
842 * (west of GMT) or 1970-01-01 (east of GMT)
843 */
844 if ((zoneoffset < 0 && memcmp(timestamp, "1969-12-31", 10)) ||
845 (0 <= zoneoffset && memcmp(timestamp, "1970-01-01", 10)))
846 return 0;
847
848 hourminute = (strtol(timestamp + 11, NULL, 10) * 60 +
849 strtol(timestamp + 14, NULL, 10) -
850 zoneoffset);
851
852 return ((zoneoffset < 0 && hourminute == 1440) ||
853 (0 <= zoneoffset && !hourminute));
854}
855
a4acb0eb 856/*
88f6dbaf 857 * Get the name etc info from the ---/+++ lines of a traditional patch header
a4acb0eb 858 *
9a4a100e
LT
859 * FIXME! The end-of-filename heuristics are kind of screwy. For existing
860 * files, we can happily check the index for a match, but for creating a
861 * new file we should try to match whatever "patch" does. I have no idea.
a4acb0eb 862 */
19c58fb8 863static void parse_traditional_patch(const char *first, const char *second, struct patch *patch)
a4acb0eb 864{
a4acb0eb
LT
865 char *name;
866
a9486b02
PR
867 first += 4; /* skip "--- " */
868 second += 4; /* skip "+++ " */
3e8a5db9
JH
869 if (!p_value_known) {
870 int p, q;
871 p = guess_p_value(first);
872 q = guess_p_value(second);
873 if (p < 0) p = q;
874 if (0 <= p && p == q) {
875 p_value = p;
876 p_value_known = 1;
877 }
878 }
a4acb0eb 879 if (is_dev_null(first)) {
19c58fb8
LT
880 patch->is_new = 1;
881 patch->is_delete = 0;
5a12c886 882 name = find_name_traditional(second, NULL, p_value);
19c58fb8 883 patch->new_name = name;
a4acb0eb 884 } else if (is_dev_null(second)) {
19c58fb8
LT
885 patch->is_new = 0;
886 patch->is_delete = 1;
5a12c886 887 name = find_name_traditional(first, NULL, p_value);
19c58fb8 888 patch->old_name = name;
a4acb0eb 889 } else {
2901bbe5
JH
890 char *first_name;
891 first_name = find_name_traditional(first, NULL, p_value);
892 name = find_name_traditional(second, first_name, p_value);
893 free(first_name);
c4593faf
JH
894 if (has_epoch_timestamp(first)) {
895 patch->is_new = 1;
896 patch->is_delete = 0;
897 patch->new_name = name;
898 } else if (has_epoch_timestamp(second)) {
899 patch->is_new = 0;
900 patch->is_delete = 1;
901 patch->old_name = name;
902 } else {
2901bbe5 903 patch->old_name = name;
44406907 904 patch->new_name = xstrdup_or_null(name);
c4593faf 905 }
a4acb0eb
LT
906 }
907 if (!name)
3638eb43 908 die(_("unable to find filename in patch at line %d"), linenr);
a4acb0eb
LT
909}
910
19c58fb8 911static int gitdiff_hdrend(const char *line, struct patch *patch)
a4acb0eb
LT
912{
913 return -1;
914}
915
1e3f6b6e
LT
916/*
917 * We're anal about diff header consistency, to make
918 * sure that we don't end up having strange ambiguous
919 * patches floating around.
920 *
921 * As a result, gitdiff_{old|new}name() will check
922 * their names against any previous information, just
923 * to make sure..
924 */
4c5197d1
NTND
925#define DIFF_OLD_NAME 0
926#define DIFF_NEW_NAME 1
927
928static char *gitdiff_verify_name(const char *line, int isnull, char *orig_name, int side)
1e3f6b6e 929{
1e3f6b6e 930 if (!orig_name && !isnull)
79ee194e 931 return find_name(line, NULL, p_value, TERM_TAB);
1e3f6b6e 932
1e3f6b6e 933 if (orig_name) {
22943f1a
JH
934 int len;
935 const char *name;
936 char *another;
1e3f6b6e
LT
937 name = orig_name;
938 len = strlen(name);
939 if (isnull)
3638eb43 940 die(_("git apply: bad git-diff - expected /dev/null, got %s on line %d"), name, linenr);
79ee194e 941 another = find_name(line, NULL, p_value, TERM_TAB);
da915939 942 if (!another || memcmp(another, name, len + 1))
4c5197d1
NTND
943 die((side == DIFF_NEW_NAME) ?
944 _("git apply: bad git-diff - inconsistent new filename on line %d") :
945 _("git apply: bad git-diff - inconsistent old filename on line %d"), linenr);
22943f1a 946 free(another);
1e3f6b6e
LT
947 return orig_name;
948 }
22943f1a
JH
949 else {
950 /* expect "/dev/null" */
951 if (memcmp("/dev/null", line, 9) || line[9] != '\n')
3638eb43 952 die(_("git apply: bad git-diff - expected /dev/null on line %d"), linenr);
22943f1a
JH
953 return NULL;
954 }
1e3f6b6e
LT
955}
956
19c58fb8 957static int gitdiff_oldname(const char *line, struct patch *patch)
a4acb0eb 958{
2901bbe5 959 char *orig = patch->old_name;
4c5197d1
NTND
960 patch->old_name = gitdiff_verify_name(line, patch->is_new, patch->old_name,
961 DIFF_OLD_NAME);
2901bbe5
JH
962 if (orig != patch->old_name)
963 free(orig);
a4acb0eb
LT
964 return 0;
965}
966
19c58fb8 967static int gitdiff_newname(const char *line, struct patch *patch)
a4acb0eb 968{
2901bbe5 969 char *orig = patch->new_name;
4c5197d1
NTND
970 patch->new_name = gitdiff_verify_name(line, patch->is_delete, patch->new_name,
971 DIFF_NEW_NAME);
2901bbe5
JH
972 if (orig != patch->new_name)
973 free(orig);
a4acb0eb
LT
974 return 0;
975}
976
19c58fb8 977static int gitdiff_oldmode(const char *line, struct patch *patch)
a4acb0eb 978{
19c58fb8 979 patch->old_mode = strtoul(line, NULL, 8);
a4acb0eb
LT
980 return 0;
981}
982
19c58fb8 983static int gitdiff_newmode(const char *line, struct patch *patch)
a4acb0eb 984{
19c58fb8 985 patch->new_mode = strtoul(line, NULL, 8);
a4acb0eb
LT
986 return 0;
987}
988
19c58fb8 989static int gitdiff_delete(const char *line, struct patch *patch)
a4acb0eb 990{
19c58fb8 991 patch->is_delete = 1;
2901bbe5 992 free(patch->old_name);
44406907 993 patch->old_name = xstrdup_or_null(patch->def_name);
19c58fb8 994 return gitdiff_oldmode(line, patch);
a4acb0eb
LT
995}
996
19c58fb8 997static int gitdiff_newfile(const char *line, struct patch *patch)
a4acb0eb 998{
19c58fb8 999 patch->is_new = 1;
2901bbe5 1000 free(patch->new_name);
44406907 1001 patch->new_name = xstrdup_or_null(patch->def_name);
19c58fb8 1002 return gitdiff_newmode(line, patch);
a4acb0eb
LT
1003}
1004
19c58fb8 1005static int gitdiff_copysrc(const char *line, struct patch *patch)
a4acb0eb 1006{
19c58fb8 1007 patch->is_copy = 1;
2901bbe5 1008 free(patch->old_name);
cefd43b7 1009 patch->old_name = find_name(line, NULL, p_value ? p_value - 1 : 0, 0);
a4acb0eb
LT
1010 return 0;
1011}
1012
19c58fb8 1013static int gitdiff_copydst(const char *line, struct patch *patch)
a4acb0eb 1014{
19c58fb8 1015 patch->is_copy = 1;
2901bbe5 1016 free(patch->new_name);
cefd43b7 1017 patch->new_name = find_name(line, NULL, p_value ? p_value - 1 : 0, 0);
a4acb0eb
LT
1018 return 0;
1019}
1020
19c58fb8 1021static int gitdiff_renamesrc(const char *line, struct patch *patch)
a4acb0eb 1022{
19c58fb8 1023 patch->is_rename = 1;
2901bbe5 1024 free(patch->old_name);
cefd43b7 1025 patch->old_name = find_name(line, NULL, p_value ? p_value - 1 : 0, 0);
a4acb0eb
LT
1026 return 0;
1027}
1028
19c58fb8 1029static int gitdiff_renamedst(const char *line, struct patch *patch)
a4acb0eb 1030{
19c58fb8 1031 patch->is_rename = 1;
2901bbe5 1032 free(patch->new_name);
cefd43b7 1033 patch->new_name = find_name(line, NULL, p_value ? p_value - 1 : 0, 0);
a4acb0eb
LT
1034 return 0;
1035}
1036
19c58fb8 1037static int gitdiff_similarity(const char *line, struct patch *patch)
a4acb0eb 1038{
afcb6ac8
JK
1039 unsigned long val = strtoul(line, NULL, 10);
1040 if (val <= 100)
1041 patch->score = val;
a4acb0eb 1042 return 0;
c1bb9350
LT
1043}
1044
70aadac0
JH
1045static int gitdiff_dissimilarity(const char *line, struct patch *patch)
1046{
afcb6ac8
JK
1047 unsigned long val = strtoul(line, NULL, 10);
1048 if (val <= 100)
1049 patch->score = val;
70aadac0
JH
1050 return 0;
1051}
1052
2cf67f1e
JH
1053static int gitdiff_index(const char *line, struct patch *patch)
1054{
81bf96bb
JH
1055 /*
1056 * index line is N hexadecimal, "..", N hexadecimal,
2cf67f1e
JH
1057 * and optional space with octal mode.
1058 */
1059 const char *ptr, *eol;
1060 int len;
1061
1062 ptr = strchr(line, '.');
9add69b1 1063 if (!ptr || ptr[1] != '.' || 40 < ptr - line)
2cf67f1e
JH
1064 return 0;
1065 len = ptr - line;
1066 memcpy(patch->old_sha1_prefix, line, len);
1067 patch->old_sha1_prefix[len] = 0;
1068
1069 line = ptr + 2;
1070 ptr = strchr(line, ' ');
31bb6d37 1071 eol = strchrnul(line, '\n');
2cf67f1e
JH
1072
1073 if (!ptr || eol < ptr)
1074 ptr = eol;
1075 len = ptr - line;
1076
9add69b1 1077 if (40 < len)
2cf67f1e
JH
1078 return 0;
1079 memcpy(patch->new_sha1_prefix, line, len);
1080 patch->new_sha1_prefix[len] = 0;
1081 if (*ptr == ' ')
1f7903a3 1082 patch->old_mode = strtoul(ptr+1, NULL, 8);
2cf67f1e
JH
1083 return 0;
1084}
1085
9a4a100e
LT
1086/*
1087 * This is normal for a diff that doesn't change anything: we'll fall through
1088 * into the next diff. Tell the parser to break out.
1089 */
19c58fb8 1090static int gitdiff_unrecognized(const char *line, struct patch *patch)
9a4a100e
LT
1091{
1092 return -1;
1093}
1094
6a2abdc1
JH
1095/*
1096 * Skip p_value leading components from "line"; as we do not accept
1097 * absolute paths, return NULL in that case.
1098 */
1099static const char *skip_tree_prefix(const char *line, int llen)
22943f1a 1100{
6a2abdc1 1101 int nslash;
22943f1a
JH
1102 int i;
1103
6a2abdc1
JH
1104 if (!p_value)
1105 return (llen && line[0] == '/') ? NULL : line;
1106
1107 nslash = p_value;
22943f1a
JH
1108 for (i = 0; i < llen; i++) {
1109 int ch = line[i];
ec7fc0b1 1110 if (ch == '/' && --nslash <= 0)
6a2abdc1 1111 return (i == 0) ? NULL : &line[i + 1];
22943f1a
JH
1112 }
1113 return NULL;
1114}
1115
81bf96bb
JH
1116/*
1117 * This is to extract the same name that appears on "diff --git"
22943f1a
JH
1118 * line. We do not find and return anything if it is a rename
1119 * patch, and it is OK because we will find the name elsewhere.
1120 * We need to reliably find name only when it is mode-change only,
1121 * creation or deletion of an empty file. In any of these cases,
1122 * both sides are the same name under a/ and b/ respectively.
1123 */
26693ba8 1124static char *git_header_name(const char *line, int llen)
5041aa70 1125{
22943f1a
JH
1126 const char *name;
1127 const char *second = NULL;
cefd43b7 1128 size_t len, line_len;
5041aa70 1129
22943f1a
JH
1130 line += strlen("diff --git ");
1131 llen -= strlen("diff --git ");
1132
1133 if (*line == '"') {
1134 const char *cp;
f285a2d7
BC
1135 struct strbuf first = STRBUF_INIT;
1136 struct strbuf sp = STRBUF_INIT;
7fb1011e
PH
1137
1138 if (unquote_c_style(&first, line, &second))
1139 goto free_and_fail1;
22943f1a 1140
6a2abdc1
JH
1141 /* strip the a/b prefix including trailing slash */
1142 cp = skip_tree_prefix(first.buf, first.len);
1143 if (!cp)
7fb1011e 1144 goto free_and_fail1;
6a2abdc1 1145 strbuf_remove(&first, 0, cp - first.buf);
22943f1a 1146
81bf96bb
JH
1147 /*
1148 * second points at one past closing dq of name.
22943f1a
JH
1149 * find the second name.
1150 */
1151 while ((second < line + llen) && isspace(*second))
1152 second++;
1153
1154 if (line + llen <= second)
7fb1011e 1155 goto free_and_fail1;
22943f1a 1156 if (*second == '"') {
7fb1011e
PH
1157 if (unquote_c_style(&sp, second, NULL))
1158 goto free_and_fail1;
6a2abdc1
JH
1159 cp = skip_tree_prefix(sp.buf, sp.len);
1160 if (!cp)
7fb1011e 1161 goto free_and_fail1;
22943f1a 1162 /* They must match, otherwise ignore */
6a2abdc1 1163 if (strcmp(cp, first.buf))
7fb1011e
PH
1164 goto free_and_fail1;
1165 strbuf_release(&sp);
b315c5c0 1166 return strbuf_detach(&first, NULL);
22943f1a
JH
1167 }
1168
1169 /* unquoted second */
6a2abdc1
JH
1170 cp = skip_tree_prefix(second, line + llen - second);
1171 if (!cp)
7fb1011e 1172 goto free_and_fail1;
6a2abdc1 1173 if (line + llen - cp != first.len ||
7fb1011e
PH
1174 memcmp(first.buf, cp, first.len))
1175 goto free_and_fail1;
b315c5c0 1176 return strbuf_detach(&first, NULL);
7fb1011e
PH
1177
1178 free_and_fail1:
1179 strbuf_release(&first);
1180 strbuf_release(&sp);
1181 return NULL;
5041aa70
LT
1182 }
1183
22943f1a 1184 /* unquoted first name */
6a2abdc1
JH
1185 name = skip_tree_prefix(line, llen);
1186 if (!name)
5041aa70 1187 return NULL;
22943f1a 1188
81bf96bb
JH
1189 /*
1190 * since the first name is unquoted, a dq if exists must be
22943f1a
JH
1191 * the beginning of the second name.
1192 */
1193 for (second = name; second < line + llen; second++) {
1194 if (*second == '"') {
f285a2d7 1195 struct strbuf sp = STRBUF_INIT;
22943f1a 1196 const char *np;
7fb1011e 1197
7fb1011e
PH
1198 if (unquote_c_style(&sp, second, NULL))
1199 goto free_and_fail2;
1200
6a2abdc1
JH
1201 np = skip_tree_prefix(sp.buf, sp.len);
1202 if (!np)
7fb1011e 1203 goto free_and_fail2;
7fb1011e
PH
1204
1205 len = sp.buf + sp.len - np;
1206 if (len < second - name &&
22943f1a
JH
1207 !strncmp(np, name, len) &&
1208 isspace(name[len])) {
1209 /* Good */
7fb1011e 1210 strbuf_remove(&sp, 0, np - sp.buf);
b315c5c0 1211 return strbuf_detach(&sp, NULL);
22943f1a 1212 }
7fb1011e
PH
1213
1214 free_and_fail2:
1215 strbuf_release(&sp);
1216 return NULL;
22943f1a
JH
1217 }
1218 }
1219
5041aa70
LT
1220 /*
1221 * Accept a name only if it shows up twice, exactly the same
1222 * form.
1223 */
cefd43b7
FC
1224 second = strchr(name, '\n');
1225 if (!second)
1226 return NULL;
1227 line_len = second - name;
5041aa70 1228 for (len = 0 ; ; len++) {
dd305c84 1229 switch (name[len]) {
5041aa70
LT
1230 default:
1231 continue;
1232 case '\n':
e70a165d 1233 return NULL;
5041aa70 1234 case '\t': case ' ':
6a2abdc1
JH
1235 /*
1236 * Is this the separator between the preimage
1237 * and the postimage pathname? Again, we are
1238 * only interested in the case where there is
1239 * no rename, as this is only to set def_name
1240 * and a rename patch has the names elsewhere
1241 * in an unambiguous form.
1242 */
1243 if (!name[len + 1])
1244 return NULL; /* no postimage name */
1245 second = skip_tree_prefix(name + len + 1,
1246 line_len - (len + 1));
cefd43b7
FC
1247 if (!second)
1248 return NULL;
6a2abdc1
JH
1249 /*
1250 * Does len bytes starting at "name" and "second"
1251 * (that are separated by one HT or SP we just
1252 * found) exactly match?
1253 */
1254 if (second[len] == '\n' && !strncmp(name, second, len))
182af834 1255 return xmemdupz(name, len);
5041aa70
LT
1256 }
1257 }
5041aa70
LT
1258}
1259
c1bb9350 1260/* Verify that we recognize the lines following a git header */
26693ba8 1261static int parse_git_header(const char *line, int len, unsigned int size, struct patch *patch)
c1bb9350 1262{
a4acb0eb
LT
1263 unsigned long offset;
1264
1265 /* A git diff has explicit new/delete information, so we don't guess */
19c58fb8
LT
1266 patch->is_new = 0;
1267 patch->is_delete = 0;
a4acb0eb 1268
5041aa70
LT
1269 /*
1270 * Some things may not have the old name in the
1271 * rest of the headers anywhere (pure mode changes,
1272 * or removing or adding empty files), so we get
1273 * the default name from the header.
1274 */
22943f1a 1275 patch->def_name = git_header_name(line, len);
6c31c22c
JK
1276 if (patch->def_name && root.len) {
1277 char *s = xstrfmt("%s%s", root.buf, patch->def_name);
969c8775
JK
1278 free(patch->def_name);
1279 patch->def_name = s;
1280 }
5041aa70 1281
a4acb0eb
LT
1282 line += len;
1283 size -= len;
1284 linenr++;
1285 for (offset = len ; size > 0 ; offset += len, size -= len, line += len, linenr++) {
1286 static const struct opentry {
1287 const char *str;
19c58fb8 1288 int (*fn)(const char *, struct patch *);
a4acb0eb
LT
1289 } optable[] = {
1290 { "@@ -", gitdiff_hdrend },
1291 { "--- ", gitdiff_oldname },
1292 { "+++ ", gitdiff_newname },
1293 { "old mode ", gitdiff_oldmode },
1294 { "new mode ", gitdiff_newmode },
1295 { "deleted file mode ", gitdiff_delete },
1296 { "new file mode ", gitdiff_newfile },
1297 { "copy from ", gitdiff_copysrc },
1298 { "copy to ", gitdiff_copydst },
33f4d087
LT
1299 { "rename old ", gitdiff_renamesrc },
1300 { "rename new ", gitdiff_renamedst },
dc938417
LT
1301 { "rename from ", gitdiff_renamesrc },
1302 { "rename to ", gitdiff_renamedst },
a4acb0eb 1303 { "similarity index ", gitdiff_similarity },
70aadac0 1304 { "dissimilarity index ", gitdiff_dissimilarity },
2cf67f1e 1305 { "index ", gitdiff_index },
9a4a100e 1306 { "", gitdiff_unrecognized },
a4acb0eb
LT
1307 };
1308 int i;
c1bb9350 1309
c1bb9350 1310 len = linelen(line, size);
a4acb0eb 1311 if (!len || line[len-1] != '\n')
c1bb9350 1312 break;
b4f2a6ac 1313 for (i = 0; i < ARRAY_SIZE(optable); i++) {
a4acb0eb
LT
1314 const struct opentry *p = optable + i;
1315 int oplen = strlen(p->str);
1316 if (len < oplen || memcmp(p->str, line, oplen))
1317 continue;
19c58fb8 1318 if (p->fn(line + oplen, patch) < 0)
a4acb0eb 1319 return offset;
9a4a100e 1320 break;
a4acb0eb 1321 }
c1bb9350
LT
1322 }
1323
a4acb0eb 1324 return offset;
c1bb9350
LT
1325}
1326
fab2c257 1327static int parse_num(const char *line, unsigned long *p)
46979f56
LT
1328{
1329 char *ptr;
fab2c257
LT
1330
1331 if (!isdigit(*line))
1332 return 0;
1333 *p = strtoul(line, &ptr, 10);
1334 return ptr - line;
1335}
1336
1337static int parse_range(const char *line, int len, int offset, const char *expect,
81bf96bb 1338 unsigned long *p1, unsigned long *p2)
fab2c257 1339{
46979f56
LT
1340 int digits, ex;
1341
1342 if (offset < 0 || offset >= len)
1343 return -1;
1344 line += offset;
1345 len -= offset;
1346
fab2c257
LT
1347 digits = parse_num(line, p1);
1348 if (!digits)
46979f56 1349 return -1;
46979f56
LT
1350
1351 offset += digits;
1352 line += digits;
1353 len -= digits;
1354
c1504628 1355 *p2 = 1;
fab2c257
LT
1356 if (*line == ',') {
1357 digits = parse_num(line+1, p2);
1358 if (!digits)
1359 return -1;
1360
1361 offset += digits+1;
1362 line += digits+1;
1363 len -= digits+1;
1364 }
1365
46979f56
LT
1366 ex = strlen(expect);
1367 if (ex > len)
1368 return -1;
1369 if (memcmp(line, expect, ex))
1370 return -1;
1371
1372 return offset + ex;
1373}
1374
26693ba8 1375static void recount_diff(const char *line, int size, struct fragment *fragment)
c14b9d1e
JS
1376{
1377 int oldlines = 0, newlines = 0, ret = 0;
1378
1379 if (size < 1) {
1380 warning("recount: ignore empty hunk");
1381 return;
1382 }
1383
1384 for (;;) {
1385 int len = linelen(line, size);
1386 size -= len;
1387 line += len;
1388
1389 if (size < 1)
1390 break;
1391
1392 switch (*line) {
1393 case ' ': case '\n':
1394 newlines++;
1395 /* fall through */
1396 case '-':
1397 oldlines++;
1398 continue;
1399 case '+':
1400 newlines++;
1401 continue;
1402 case '\\':
6cf91492 1403 continue;
c14b9d1e 1404 case '@':
59556548 1405 ret = size < 3 || !starts_with(line, "@@ ");
c14b9d1e
JS
1406 break;
1407 case 'd':
59556548 1408 ret = size < 5 || !starts_with(line, "diff ");
c14b9d1e
JS
1409 break;
1410 default:
1411 ret = -1;
1412 break;
1413 }
1414 if (ret) {
3638eb43 1415 warning(_("recount: unexpected line: %.*s"),
c14b9d1e
JS
1416 (int)linelen(line, size), line);
1417 return;
1418 }
1419 break;
1420 }
1421 fragment->oldlines = oldlines;
1422 fragment->newlines = newlines;
1423}
1424
46979f56
LT
1425/*
1426 * Parse a unified diff fragment header of the
1427 * form "@@ -a,b +c,d @@"
1428 */
26693ba8 1429static int parse_fragment_header(const char *line, int len, struct fragment *fragment)
46979f56
LT
1430{
1431 int offset;
1432
1433 if (!len || line[len-1] != '\n')
1434 return -1;
1435
1436 /* Figure out the number of lines in a fragment */
fab2c257
LT
1437 offset = parse_range(line, len, 4, " +", &fragment->oldpos, &fragment->oldlines);
1438 offset = parse_range(line, len, offset, " @@", &fragment->newpos, &fragment->newlines);
46979f56
LT
1439
1440 return offset;
1441}
1442
26693ba8 1443static int find_header(const char *line, unsigned long size, int *hdrsize, struct patch *patch)
c1bb9350
LT
1444{
1445 unsigned long offset, len;
1446
9987d7c5 1447 patch->is_toplevel_relative = 0;
19c58fb8
LT
1448 patch->is_rename = patch->is_copy = 0;
1449 patch->is_new = patch->is_delete = -1;
1450 patch->old_mode = patch->new_mode = 0;
1451 patch->old_name = patch->new_name = NULL;
46979f56 1452 for (offset = 0; size > 0; offset += len, size -= len, line += len, linenr++) {
c1bb9350
LT
1453 unsigned long nextlen;
1454
1455 len = linelen(line, size);
1456 if (!len)
1457 break;
1458
1459 /* Testing this early allows us to take a few shortcuts.. */
1460 if (len < 6)
1461 continue;
46979f56
LT
1462
1463 /*
82e5a82f 1464 * Make sure we don't find any unconnected patch fragments.
46979f56
LT
1465 * That's a sign that we didn't find a header, and that a
1466 * patch has become corrupted/broken up.
1467 */
1468 if (!memcmp("@@ -", line, 4)) {
19c58fb8
LT
1469 struct fragment dummy;
1470 if (parse_fragment_header(line, len, &dummy) < 0)
46979f56 1471 continue;
3638eb43 1472 die(_("patch fragment without header at line %d: %.*s"),
65341411 1473 linenr, (int)len-1, line);
46979f56
LT
1474 }
1475
c1bb9350
LT
1476 if (size < len + 6)
1477 break;
1478
1479 /*
1480 * Git patch? It might not have a real patch, just a rename
1481 * or mode change, so we handle that specially
1482 */
1483 if (!memcmp("diff --git ", line, 11)) {
19c58fb8 1484 int git_hdr_len = parse_git_header(line, len, size, patch);
206de27e 1485 if (git_hdr_len <= len)
c1bb9350 1486 continue;
b7e8039a
LT
1487 if (!patch->old_name && !patch->new_name) {
1488 if (!patch->def_name)
3638eb43
NTND
1489 die(Q_("git diff header lacks filename information when removing "
1490 "%d leading pathname component (line %d)",
1491 "git diff header lacks filename information when removing "
1492 "%d leading pathname components (line %d)",
1493 p_value),
1494 p_value, linenr);
2901bbe5
JH
1495 patch->old_name = xstrdup(patch->def_name);
1496 patch->new_name = xstrdup(patch->def_name);
b7e8039a 1497 }
2c93286a
JM
1498 if (!patch->is_delete && !patch->new_name)
1499 die("git diff header lacks filename information "
1500 "(line %d)", linenr);
9987d7c5 1501 patch->is_toplevel_relative = 1;
a4acb0eb 1502 *hdrsize = git_hdr_len;
c1bb9350
LT
1503 return offset;
1504 }
1505
81bf96bb 1506 /* --- followed by +++ ? */
c1bb9350
LT
1507 if (memcmp("--- ", line, 4) || memcmp("+++ ", line + len, 4))
1508 continue;
1509
1510 /*
1511 * We only accept unified patches, so we want it to
1512 * at least have "@@ -a,b +c,d @@\n", which is 14 chars
81bf96bb 1513 * minimum ("@@ -0,0 +1 @@\n" is the shortest).
c1bb9350
LT
1514 */
1515 nextlen = linelen(line + len, size - len);
1516 if (size < nextlen + 14 || memcmp("@@ -", line + len + nextlen, 4))
1517 continue;
1518
1519 /* Ok, we'll consider it a patch */
19c58fb8 1520 parse_traditional_patch(line, line+len, patch);
c1bb9350 1521 *hdrsize = len + nextlen;
46979f56 1522 linenr += 2;
c1bb9350
LT
1523 return offset;
1524 }
1525 return -1;
1526}
1527
92a1747e 1528static void record_ws_error(unsigned result, const char *line, int len, int linenr)
d0c25035 1529{
c1795bb0 1530 char *err;
92a1747e 1531
c1795bb0
WC
1532 if (!result)
1533 return;
d0c25035 1534
d0c25035
JH
1535 whitespace_error++;
1536 if (squelch_whitespace_errors &&
1537 squelch_whitespace_errors < whitespace_error)
92a1747e
JH
1538 return;
1539
1540 err = whitespace_error_string(result);
1541 fprintf(stderr, "%s:%d: %s.\n%.*s\n",
1542 patch_input_file, linenr, err, len, line);
1543 free(err);
1544}
1545
1546static void check_whitespace(const char *line, int len, unsigned ws_rule)
1547{
1548 unsigned result = ws_check(line + 1, len - 1, ws_rule);
1549
1550 record_ws_error(result, line + 1, len - 2, linenr);
d0c25035
JH
1551}
1552
c1bb9350 1553/*
4be60962
JH
1554 * Parse a unified diff. Note that this really needs to parse each
1555 * fragment separately, since the only way to know the difference
1556 * between a "---" that is part of a patch, and a "---" that starts
1557 * the next patch is to look at the line counts..
c1bb9350 1558 */
26693ba8 1559static int parse_fragment(const char *line, unsigned long size,
81bf96bb 1560 struct patch *patch, struct fragment *fragment)
c1bb9350 1561{
3f40315a 1562 int added, deleted;
c1bb9350 1563 int len = linelen(line, size), offset;
30996652 1564 unsigned long oldlines, newlines;
47495887 1565 unsigned long leading, trailing;
c1bb9350 1566
19c58fb8 1567 offset = parse_fragment_header(line, len, fragment);
c1bb9350
LT
1568 if (offset < 0)
1569 return -1;
c14b9d1e
JS
1570 if (offset > 0 && patch->recount)
1571 recount_diff(line + offset, size - offset, fragment);
19c58fb8
LT
1572 oldlines = fragment->oldlines;
1573 newlines = fragment->newlines;
47495887
EB
1574 leading = 0;
1575 trailing = 0;
c1bb9350
LT
1576
1577 /* Parse the thing.. */
1578 line += len;
1579 size -= len;
46979f56 1580 linenr++;
3f40315a 1581 added = deleted = 0;
4be60962
JH
1582 for (offset = len;
1583 0 < size;
1584 offset += len, size -= len, line += len, linenr++) {
c1bb9350
LT
1585 if (!oldlines && !newlines)
1586 break;
1587 len = linelen(line, size);
1588 if (!len || line[len-1] != '\n')
1589 return -1;
1590 switch (*line) {
1591 default:
1592 return -1;
b507b465 1593 case '\n': /* newer GNU diff, an empty context line */
c1bb9350
LT
1594 case ' ':
1595 oldlines--;
1596 newlines--;
47495887
EB
1597 if (!deleted && !added)
1598 leading++;
1599 trailing++;
0a80bc9f
JH
1600 if (!apply_in_reverse &&
1601 ws_error_action == correct_ws_error)
1602 check_whitespace(line, len, patch->ws_rule);
c1bb9350
LT
1603 break;
1604 case '-':
5fda48d6 1605 if (apply_in_reverse &&
81bf96bb 1606 ws_error_action != nowarn_ws_error)
cf1b7869 1607 check_whitespace(line, len, patch->ws_rule);
3f40315a 1608 deleted++;
c1bb9350 1609 oldlines--;
47495887 1610 trailing = 0;
c1bb9350
LT
1611 break;
1612 case '+':
5fda48d6 1613 if (!apply_in_reverse &&
81bf96bb 1614 ws_error_action != nowarn_ws_error)
cf1b7869 1615 check_whitespace(line, len, patch->ws_rule);
3f40315a 1616 added++;
c1bb9350 1617 newlines--;
47495887 1618 trailing = 0;
c1bb9350 1619 break;
433ef8a2 1620
81bf96bb
JH
1621 /*
1622 * We allow "\ No newline at end of file". Depending
433ef8a2
FK
1623 * on locale settings when the patch was produced we
1624 * don't know what this line looks like. The only
56d33b11
JH
1625 * thing we do know is that it begins with "\ ".
1626 * Checking for 12 is just for sanity check -- any
1627 * l10n of "\ No newline..." is at least that long.
1628 */
fab2c257 1629 case '\\':
433ef8a2 1630 if (len < 12 || memcmp(line, "\\ ", 2))
3cca928d 1631 return -1;
fab2c257 1632 break;
c1bb9350
LT
1633 }
1634 }
c1504628
LT
1635 if (oldlines || newlines)
1636 return -1;
ad6e8ed3
JH
1637 if (!deleted && !added)
1638 return -1;
1639
47495887
EB
1640 fragment->leading = leading;
1641 fragment->trailing = trailing;
1642
81bf96bb
JH
1643 /*
1644 * If a fragment ends with an incomplete line, we failed to include
8b64647d
JH
1645 * it in the above loop because we hit oldlines == newlines == 0
1646 * before seeing it.
1647 */
433ef8a2 1648 if (12 < size && !memcmp(line, "\\ ", 2))
8b64647d
JH
1649 offset += linelen(line, size);
1650
3f40315a
LT
1651 patch->lines_added += added;
1652 patch->lines_deleted += deleted;
4be60962
JH
1653
1654 if (0 < patch->is_new && oldlines)
3638eb43 1655 return error(_("new file depends on old contents"));
4be60962 1656 if (0 < patch->is_delete && newlines)
3638eb43 1657 return error(_("deleted file still has contents"));
c1bb9350
LT
1658 return offset;
1659}
1660
92737a22
JH
1661/*
1662 * We have seen "diff --git a/... b/..." header (or a traditional patch
1663 * header). Read hunks that belong to this patch into fragments and hang
1664 * them to the given patch structure.
1665 *
1666 * The (fragment->patch, fragment->size) pair points into the memory given
1667 * by the caller, not a copy, when we return.
1668 */
26693ba8 1669static int parse_single_patch(const char *line, unsigned long size, struct patch *patch)
c1bb9350
LT
1670{
1671 unsigned long offset = 0;
4be60962 1672 unsigned long oldlines = 0, newlines = 0, context = 0;
19c58fb8 1673 struct fragment **fragp = &patch->fragments;
c1bb9350
LT
1674
1675 while (size > 4 && !memcmp(line, "@@ -", 4)) {
19c58fb8
LT
1676 struct fragment *fragment;
1677 int len;
1678
90321c10 1679 fragment = xcalloc(1, sizeof(*fragment));
77b15bbd 1680 fragment->linenr = linenr;
19c58fb8 1681 len = parse_fragment(line, size, patch, fragment);
c1bb9350 1682 if (len <= 0)
3638eb43 1683 die(_("corrupt patch at line %d"), linenr);
19c58fb8
LT
1684 fragment->patch = line;
1685 fragment->size = len;
4be60962
JH
1686 oldlines += fragment->oldlines;
1687 newlines += fragment->newlines;
1688 context += fragment->leading + fragment->trailing;
19c58fb8
LT
1689
1690 *fragp = fragment;
1691 fragp = &fragment->next;
c1bb9350
LT
1692
1693 offset += len;
1694 line += len;
1695 size -= len;
1696 }
4be60962
JH
1697
1698 /*
1699 * If something was removed (i.e. we have old-lines) it cannot
1700 * be creation, and if something was added it cannot be
1701 * deletion. However, the reverse is not true; --unified=0
1702 * patches that only add are not necessarily creation even
1703 * though they do not have any old lines, and ones that only
1704 * delete are not necessarily deletion.
1705 *
1706 * Unfortunately, a real creation/deletion patch do _not_ have
1707 * any context line by definition, so we cannot safely tell it
1708 * apart with --unified=0 insanity. At least if the patch has
1709 * more than one hunk it is not creation or deletion.
1710 */
1711 if (patch->is_new < 0 &&
1712 (oldlines || (patch->fragments && patch->fragments->next)))
1713 patch->is_new = 0;
1714 if (patch->is_delete < 0 &&
1715 (newlines || (patch->fragments && patch->fragments->next)))
1716 patch->is_delete = 0;
4be60962
JH
1717
1718 if (0 < patch->is_new && oldlines)
3638eb43 1719 die(_("new file %s depends on old contents"), patch->new_name);
4be60962 1720 if (0 < patch->is_delete && newlines)
3638eb43 1721 die(_("deleted file %s still has contents"), patch->old_name);
4be60962 1722 if (!patch->is_delete && !newlines && context)
3638eb43
NTND
1723 fprintf_ln(stderr,
1724 _("** warning: "
1725 "file %s becomes empty but is not deleted"),
1726 patch->new_name);
4be60962 1727
c1bb9350
LT
1728 return offset;
1729}
1730
1fea629f
LT
1731static inline int metadata_changes(struct patch *patch)
1732{
1733 return patch->is_rename > 0 ||
1734 patch->is_copy > 0 ||
1735 patch->is_new > 0 ||
1736 patch->is_delete ||
1737 (patch->old_mode && patch->new_mode &&
1738 patch->old_mode != patch->new_mode);
1739}
1740
3cd4f5e8
JH
1741static char *inflate_it(const void *data, unsigned long size,
1742 unsigned long inflated_size)
051308f6 1743{
ef49a7a0 1744 git_zstream stream;
3cd4f5e8
JH
1745 void *out;
1746 int st;
1747
1748 memset(&stream, 0, sizeof(stream));
1749
1750 stream.next_in = (unsigned char *)data;
1751 stream.avail_in = size;
1752 stream.next_out = out = xmalloc(inflated_size);
1753 stream.avail_out = inflated_size;
39c68542
LT
1754 git_inflate_init(&stream);
1755 st = git_inflate(&stream, Z_FINISH);
1756 git_inflate_end(&stream);
3cd4f5e8
JH
1757 if ((st != Z_STREAM_END) || stream.total_out != inflated_size) {
1758 free(out);
1759 return NULL;
1760 }
1761 return out;
1762}
1763
92737a22
JH
1764/*
1765 * Read a binary hunk and return a new fragment; fragment->patch
1766 * points at an allocated memory that the caller must free, so
1767 * it is marked as "->free_patch = 1".
1768 */
3cd4f5e8
JH
1769static struct fragment *parse_binary_hunk(char **buf_p,
1770 unsigned long *sz_p,
1771 int *status_p,
1772 int *used_p)
1773{
81bf96bb
JH
1774 /*
1775 * Expect a line that begins with binary patch method ("literal"
3cd4f5e8
JH
1776 * or "delta"), followed by the length of data before deflating.
1777 * a sequence of 'length-byte' followed by base-85 encoded data
1778 * should follow, terminated by a newline.
051308f6
JH
1779 *
1780 * Each 5-byte sequence of base-85 encodes up to 4 bytes,
1781 * and we would limit the patch line to 66 characters,
1782 * so one line can fit up to 13 groups that would decode
1783 * to 52 bytes max. The length byte 'A'-'Z' corresponds
1784 * to 1-26 bytes, and 'a'-'z' corresponds to 27-52 bytes.
051308f6
JH
1785 */
1786 int llen, used;
3cd4f5e8
JH
1787 unsigned long size = *sz_p;
1788 char *buffer = *buf_p;
1789 int patch_method;
1790 unsigned long origlen;
0660626c 1791 char *data = NULL;
3cd4f5e8
JH
1792 int hunk_size = 0;
1793 struct fragment *frag;
051308f6 1794
0660626c
JH
1795 llen = linelen(buffer, size);
1796 used = llen;
3cd4f5e8
JH
1797
1798 *status_p = 0;
0660626c 1799
59556548 1800 if (starts_with(buffer, "delta ")) {
3cd4f5e8
JH
1801 patch_method = BINARY_DELTA_DEFLATED;
1802 origlen = strtoul(buffer + 6, NULL, 10);
0660626c 1803 }
59556548 1804 else if (starts_with(buffer, "literal ")) {
3cd4f5e8
JH
1805 patch_method = BINARY_LITERAL_DEFLATED;
1806 origlen = strtoul(buffer + 8, NULL, 10);
0660626c
JH
1807 }
1808 else
3cd4f5e8
JH
1809 return NULL;
1810
1811 linenr++;
0660626c 1812 buffer += llen;
051308f6
JH
1813 while (1) {
1814 int byte_length, max_byte_length, newsize;
1815 llen = linelen(buffer, size);
1816 used += llen;
1817 linenr++;
03eb8f8a
JH
1818 if (llen == 1) {
1819 /* consume the blank line */
1820 buffer++;
1821 size--;
051308f6 1822 break;
03eb8f8a 1823 }
81bf96bb
JH
1824 /*
1825 * Minimum line is "A00000\n" which is 7-byte long,
051308f6
JH
1826 * and the line length must be multiple of 5 plus 2.
1827 */
1828 if ((llen < 7) || (llen-2) % 5)
1829 goto corrupt;
1830 max_byte_length = (llen - 2) / 5 * 4;
1831 byte_length = *buffer;
1832 if ('A' <= byte_length && byte_length <= 'Z')
1833 byte_length = byte_length - 'A' + 1;
1834 else if ('a' <= byte_length && byte_length <= 'z')
1835 byte_length = byte_length - 'a' + 27;
1836 else
1837 goto corrupt;
1838 /* if the input length was not multiple of 4, we would
1839 * have filler at the end but the filler should never
1840 * exceed 3 bytes
1841 */
1842 if (max_byte_length < byte_length ||
1843 byte_length <= max_byte_length - 4)
1844 goto corrupt;
3cd4f5e8 1845 newsize = hunk_size + byte_length;
0660626c 1846 data = xrealloc(data, newsize);
3cd4f5e8 1847 if (decode_85(data + hunk_size, buffer + 1, byte_length))
051308f6 1848 goto corrupt;
3cd4f5e8 1849 hunk_size = newsize;
051308f6
JH
1850 buffer += llen;
1851 size -= llen;
1852 }
3cd4f5e8
JH
1853
1854 frag = xcalloc(1, sizeof(*frag));
1855 frag->patch = inflate_it(data, hunk_size, origlen);
6fe53908 1856 frag->free_patch = 1;
3cd4f5e8
JH
1857 if (!frag->patch)
1858 goto corrupt;
1859 free(data);
1860 frag->size = origlen;
1861 *buf_p = buffer;
1862 *sz_p = size;
1863 *used_p = used;
1864 frag->binary_patch_method = patch_method;
1865 return frag;
1866
051308f6 1867 corrupt:
4cac42b1 1868 free(data);
3cd4f5e8 1869 *status_p = -1;
3638eb43 1870 error(_("corrupt binary patch at line %d: %.*s"),
3cd4f5e8
JH
1871 linenr-1, llen-1, buffer);
1872 return NULL;
1873}
1874
1875static int parse_binary(char *buffer, unsigned long size, struct patch *patch)
1876{
81bf96bb
JH
1877 /*
1878 * We have read "GIT binary patch\n"; what follows is a line
3cd4f5e8
JH
1879 * that says the patch method (currently, either "literal" or
1880 * "delta") and the length of data before deflating; a
1881 * sequence of 'length-byte' followed by base-85 encoded data
1882 * follows.
1883 *
1884 * When a binary patch is reversible, there is another binary
1885 * hunk in the same format, starting with patch method (either
1886 * "literal" or "delta") with the length of data, and a sequence
1887 * of length-byte + base-85 encoded data, terminated with another
1888 * empty line. This data, when applied to the postimage, produces
1889 * the preimage.
1890 */
1891 struct fragment *forward;
1892 struct fragment *reverse;
1893 int status;
1894 int used, used_1;
1895
1896 forward = parse_binary_hunk(&buffer, &size, &status, &used);
1897 if (!forward && !status)
1898 /* there has to be one hunk (forward hunk) */
3638eb43 1899 return error(_("unrecognized binary patch at line %d"), linenr-1);
3cd4f5e8
JH
1900 if (status)
1901 /* otherwise we already gave an error message */
1902 return status;
1903
1904 reverse = parse_binary_hunk(&buffer, &size, &status, &used_1);
1905 if (reverse)
1906 used += used_1;
1907 else if (status) {
81bf96bb
JH
1908 /*
1909 * Not having reverse hunk is not an error, but having
3cd4f5e8
JH
1910 * a corrupt reverse hunk is.
1911 */
1912 free((void*) forward->patch);
1913 free(forward);
1914 return status;
1915 }
1916 forward->next = reverse;
1917 patch->fragments = forward;
1918 patch->is_binary = 1;
1919 return used;
051308f6
JH
1920}
1921
d487b0ba
JH
1922static void prefix_one(char **name)
1923{
1924 char *old_name = *name;
1925 if (!old_name)
1926 return;
1927 *name = xstrdup(prefix_filename(prefix, prefix_length, *name));
1928 free(old_name);
1929}
1930
1931static void prefix_patch(struct patch *p)
1932{
1933 if (!prefix || p->is_toplevel_relative)
1934 return;
1935 prefix_one(&p->new_name);
1936 prefix_one(&p->old_name);
1937}
1938
3ee2ad14
JH
1939/*
1940 * include/exclude
1941 */
1942
1943static struct string_list limit_by_name;
1944static int has_include;
1945static void add_name_limit(const char *name, int exclude)
1946{
1947 struct string_list_item *it;
1948
1949 it = string_list_append(&limit_by_name, name);
1950 it->util = exclude ? NULL : (void *) 1;
1951}
1952
1953static int use_patch(struct patch *p)
1954{
1955 const char *pathname = p->new_name ? p->new_name : p->old_name;
1956 int i;
1957
1958 /* Paths outside are not touched regardless of "--include" */
1959 if (0 < prefix_length) {
1960 int pathlen = strlen(pathname);
1961 if (pathlen <= prefix_length ||
1962 memcmp(prefix, pathname, prefix_length))
1963 return 0;
1964 }
1965
1966 /* See if it matches any of exclude/include rule */
1967 for (i = 0; i < limit_by_name.nr; i++) {
1968 struct string_list_item *it = &limit_by_name.items[i];
ead51a75 1969 if (!wildmatch(it->string, pathname, 0, NULL))
3ee2ad14
JH
1970 return (it->util != NULL);
1971 }
1972
1973 /*
1974 * If we had any include, a path that does not match any rule is
1975 * not used. Otherwise, we saw bunch of exclude rules (or none)
1976 * and such a path is used.
1977 */
1978 return !has_include;
1979}
1980
1981
92737a22 1982/*
41ccfdd9 1983 * Read the patch text in "buffer" that extends for "size" bytes; stop
92737a22
JH
1984 * reading after seeing a single patch (i.e. changes to a single file).
1985 * Create fragments (i.e. patch hunks) and hang them to the given patch.
1986 * Return the number of bytes consumed, so that the caller can call us
1987 * again for the next patch.
1988 */
19c58fb8 1989static int parse_chunk(char *buffer, unsigned long size, struct patch *patch)
c1bb9350
LT
1990{
1991 int hdrsize, patchsize;
19c58fb8 1992 int offset = find_header(buffer, size, &hdrsize, patch);
c1bb9350
LT
1993
1994 if (offset < 0)
1995 return offset;
c1bb9350 1996
d487b0ba
JH
1997 prefix_patch(patch);
1998
477a08af
JH
1999 if (!use_patch(patch))
2000 patch->ws_rule = 0;
2001 else
2002 patch->ws_rule = whitespace_rule(patch->new_name
2003 ? patch->new_name
2004 : patch->old_name);
cf1b7869 2005
81bf96bb
JH
2006 patchsize = parse_single_patch(buffer + offset + hdrsize,
2007 size - offset - hdrsize, patch);
c1bb9350 2008
92927ed0 2009 if (!patchsize) {
051308f6 2010 static const char git_binary[] = "GIT binary patch\n";
3200d1ae
JH
2011 int hd = hdrsize + offset;
2012 unsigned long llen = linelen(buffer + hd, size - hd);
2013
051308f6
JH
2014 if (llen == sizeof(git_binary) - 1 &&
2015 !memcmp(git_binary, buffer + hd, llen)) {
2016 int used;
2017 linenr++;
2018 used = parse_binary(buffer + hd + llen,
2019 size - hd - llen, patch);
2020 if (used)
2021 patchsize = used + llen;
2022 else
2023 patchsize = 0;
2024 }
2025 else if (!memcmp(" differ\n", buffer + hd + llen - 8, 8)) {
e36f3a8a
EP
2026 static const char *binhdr[] = {
2027 "Binary files ",
2028 "Files ",
2029 NULL,
2030 };
2031 int i;
3200d1ae
JH
2032 for (i = 0; binhdr[i]; i++) {
2033 int len = strlen(binhdr[i]);
2034 if (len < size - hd &&
2035 !memcmp(binhdr[i], buffer + hd, len)) {
051308f6 2036 linenr++;
3200d1ae 2037 patch->is_binary = 1;
051308f6 2038 patchsize = llen;
3200d1ae
JH
2039 break;
2040 }
2041 }
051308f6 2042 }
ff36de08 2043
2b6eef94
JH
2044 /* Empty patch cannot be applied if it is a text patch
2045 * without metadata change. A binary patch appears
2046 * empty to us here.
92927ed0
JH
2047 */
2048 if ((apply || check) &&
2b6eef94 2049 (!patch->is_binary && !metadata_changes(patch)))
3638eb43 2050 die(_("patch with only garbage at line %d"), linenr);
ff36de08 2051 }
1fea629f 2052
c1bb9350
LT
2053 return offset + hdrsize + patchsize;
2054}
2055
e5a94313
JS
2056#define swap(a,b) myswap((a),(b),sizeof(a))
2057
2058#define myswap(a, b, size) do { \
2059 unsigned char mytmp[size]; \
2060 memcpy(mytmp, &a, size); \
2061 memcpy(&a, &b, size); \
2062 memcpy(&b, mytmp, size); \
2063} while (0)
2064
2065static void reverse_patches(struct patch *p)
2066{
2067 for (; p; p = p->next) {
2068 struct fragment *frag = p->fragments;
2069
2070 swap(p->new_name, p->old_name);
2071 swap(p->new_mode, p->old_mode);
2072 swap(p->is_new, p->is_delete);
2073 swap(p->lines_added, p->lines_deleted);
2074 swap(p->old_sha1_prefix, p->new_sha1_prefix);
2075
2076 for (; frag; frag = frag->next) {
2077 swap(frag->newpos, frag->oldpos);
2078 swap(frag->newlines, frag->oldlines);
2079 }
e5a94313
JS
2080 }
2081}
2082
81bf96bb
JH
2083static const char pluses[] =
2084"++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++";
2085static const char minuses[]=
2086"----------------------------------------------------------------------";
3f40315a
LT
2087
2088static void show_stats(struct patch *patch)
2089{
f285a2d7 2090 struct strbuf qname = STRBUF_INIT;
663af342
PH
2091 char *cp = patch->new_name ? patch->new_name : patch->old_name;
2092 int max, add, del;
3f40315a 2093
663af342 2094 quote_c_style(cp, &qname, NULL, 0);
22943f1a 2095
3f40315a
LT
2096 /*
2097 * "scale" the filename
2098 */
3f40315a
LT
2099 max = max_len;
2100 if (max > 50)
2101 max = 50;
663af342
PH
2102
2103 if (qname.len > max) {
2104 cp = strchr(qname.buf + qname.len + 3 - max, '/');
2105 if (!cp)
2106 cp = qname.buf + qname.len + 3 - max;
2107 strbuf_splice(&qname, 0, cp - qname.buf, "...", 3);
2108 }
2109
2110 if (patch->is_binary) {
2111 printf(" %-*s | Bin\n", max, qname.buf);
2112 strbuf_release(&qname);
2113 return;
62917097 2114 }
663af342
PH
2115
2116 printf(" %-*s |", max, qname.buf);
2117 strbuf_release(&qname);
3f40315a
LT
2118
2119 /*
2120 * scale the add/delete
2121 */
663af342 2122 max = max + max_change > 70 ? 70 - max : max_change;
95bedc9e
LT
2123 add = patch->lines_added;
2124 del = patch->lines_deleted;
95bedc9e 2125
69f956e1 2126 if (max_change > 0) {
663af342 2127 int total = ((add + del) * max + max_change / 2) / max_change;
69f956e1
SV
2128 add = (add * max + max_change / 2) / max_change;
2129 del = total - add;
2130 }
663af342
PH
2131 printf("%5d %.*s%.*s\n", patch->lines_added + patch->lines_deleted,
2132 add, pluses, del, minuses);
3f40315a
LT
2133}
2134
c7f9cb14 2135static int read_old_data(struct stat *st, const char *path, struct strbuf *buf)
3cca928d 2136{
3cca928d
LT
2137 switch (st->st_mode & S_IFMT) {
2138 case S_IFLNK:
b11b7e13 2139 if (strbuf_readlink(buf, path, st->st_size) < 0)
3638eb43 2140 return error(_("unable to read symlink %s"), path);
c7f9cb14 2141 return 0;
3cca928d 2142 case S_IFREG:
387e7e19 2143 if (strbuf_read_file(buf, path, st->st_size) != st->st_size)
3638eb43 2144 return error(_("unable to open or read %s"), path);
21e5ad50 2145 convert_to_git(path, buf->buf, buf->len, buf, 0);
c7f9cb14 2146 return 0;
3cca928d
LT
2147 default:
2148 return -1;
2149 }
2150}
2151
86c91f91
GB
2152/*
2153 * Update the preimage, and the common lines in postimage,
2154 * from buffer buf of length len. If postlen is 0 the postimage
2155 * is updated in place, otherwise it's updated on a new buffer
2156 * of length postlen
2157 */
2158
c1beba5b
JH
2159static void update_pre_post_images(struct image *preimage,
2160 struct image *postimage,
2161 char *buf,
86c91f91 2162 size_t len, size_t postlen)
3cca928d 2163{
5de7166d 2164 int i, ctx, reduced;
c1beba5b
JH
2165 char *new, *old, *fixed;
2166 struct image fixed_preimage;
3cca928d 2167
c1beba5b
JH
2168 /*
2169 * Update the preimage with whitespace fixes. Note that we
2170 * are not losing preimage->buf -- apply_one_fragment() will
2171 * free "oldlines".
2172 */
2173 prepare_image(&fixed_preimage, buf, len, 1);
5de7166d
JH
2174 assert(postlen
2175 ? fixed_preimage.nr == preimage->nr
2176 : fixed_preimage.nr <= preimage->nr);
2177 for (i = 0; i < fixed_preimage.nr; i++)
c1beba5b
JH
2178 fixed_preimage.line[i].flag = preimage->line[i].flag;
2179 free(preimage->line_allocated);
2180 *preimage = fixed_preimage;
3cca928d 2181
c1beba5b 2182 /*
86c91f91 2183 * Adjust the common context lines in postimage. This can be
250b3c6c
JH
2184 * done in-place when we are shrinking it with whitespace
2185 * fixing, but needs a new buffer when ignoring whitespace or
2186 * expanding leading tabs to spaces.
2187 *
86c91f91
GB
2188 * We trust the caller to tell us if the update can be done
2189 * in place (postlen==0) or not.
c1beba5b 2190 */
86c91f91
GB
2191 old = postimage->buf;
2192 if (postlen)
2193 new = postimage->buf = xmalloc(postlen);
2194 else
2195 new = old;
c1beba5b 2196 fixed = preimage->buf;
5de7166d
JH
2197
2198 for (i = reduced = ctx = 0; i < postimage->nr; i++) {
c1beba5b
JH
2199 size_t len = postimage->line[i].len;
2200 if (!(postimage->line[i].flag & LINE_COMMON)) {
2201 /* an added line -- no counterparts in preimage */
2202 memmove(new, old, len);
2203 old += len;
2204 new += len;
2205 continue;
3cca928d 2206 }
c1beba5b
JH
2207
2208 /* a common context -- skip it in the original postimage */
2209 old += len;
2210
2211 /* and find the corresponding one in the fixed preimage */
2212 while (ctx < preimage->nr &&
2213 !(preimage->line[ctx].flag & LINE_COMMON)) {
2214 fixed += preimage->line[ctx].len;
2215 ctx++;
2216 }
5de7166d
JH
2217
2218 /*
2219 * preimage is expected to run out, if the caller
2220 * fixed addition of trailing blank lines.
2221 */
2222 if (preimage->nr <= ctx) {
2223 reduced++;
2224 continue;
2225 }
c1beba5b
JH
2226
2227 /* and copy it in, while fixing the line length */
2228 len = preimage->line[ctx].len;
2229 memcpy(new, fixed, len);
2230 new += len;
2231 fixed += len;
2232 postimage->line[i].len = len;
2233 ctx++;
2234 }
2235
2988289f
JH
2236 if (postlen
2237 ? postlen < new - postimage->buf
2238 : postimage->len < new - postimage->buf)
2239 die("BUG: caller miscounted postlen: asked %d, orig = %d, used = %d",
2240 (int)postlen, (int) postimage->len, (int)(new - postimage->buf));
2241
c1beba5b
JH
2242 /* Fix the length of the whole thing */
2243 postimage->len = new - postimage->buf;
5de7166d 2244 postimage->nr -= reduced;
c1beba5b
JH
2245}
2246
b94f2eda
JH
2247static int match_fragment(struct image *img,
2248 struct image *preimage,
2249 struct image *postimage,
c89fb6b1 2250 unsigned long try,
b94f2eda 2251 int try_lno,
c607aaa2 2252 unsigned ws_rule,
dc41976a 2253 int match_beginning, int match_end)
c89fb6b1 2254{
b94f2eda 2255 int i;
c1beba5b 2256 char *fixed_buf, *buf, *orig, *target;
d511bd33 2257 struct strbuf fixed;
250b3c6c 2258 size_t fixed_len, postlen;
51667147 2259 int preimage_limit;
b94f2eda 2260
51667147
BG
2261 if (preimage->nr + try_lno <= img->nr) {
2262 /*
2263 * The hunk falls within the boundaries of img.
2264 */
2265 preimage_limit = preimage->nr;
2266 if (match_end && (preimage->nr + try_lno != img->nr))
2267 return 0;
2268 } else if (ws_error_action == correct_ws_error &&
0c3ef984 2269 (ws_rule & WS_BLANK_AT_EOF)) {
51667147 2270 /*
0c3ef984
BG
2271 * This hunk extends beyond the end of img, and we are
2272 * removing blank lines at the end of the file. This
2273 * many lines from the beginning of the preimage must
2274 * match with img, and the remainder of the preimage
2275 * must be blank.
51667147
BG
2276 */
2277 preimage_limit = img->nr - try_lno;
2278 } else {
2279 /*
2280 * The hunk extends beyond the end of the img and
2281 * we are not removing blanks at the end, so we
2282 * should reject the hunk at this position.
2283 */
b94f2eda 2284 return 0;
51667147 2285 }
b94f2eda
JH
2286
2287 if (match_beginning && try_lno)
c89fb6b1 2288 return 0;
dc41976a 2289
b94f2eda 2290 /* Quick hash check */
51667147 2291 for (i = 0; i < preimage_limit; i++)
9d158601
JH
2292 if ((img->line[try_lno + i].flag & LINE_PATCHED) ||
2293 (preimage->line[i].hash != img->line[try_lno + i].hash))
b94f2eda
JH
2294 return 0;
2295
51667147
BG
2296 if (preimage_limit == preimage->nr) {
2297 /*
2298 * Do we have an exact match? If we were told to match
2299 * at the end, size must be exactly at try+fragsize,
2300 * otherwise try+fragsize must be still within the preimage,
2301 * and either case, the old piece should match the preimage
2302 * exactly.
2303 */
2304 if ((match_end
2305 ? (try + preimage->len == img->len)
2306 : (try + preimage->len <= img->len)) &&
2307 !memcmp(img->buf + try, preimage->buf, preimage->len))
2308 return 1;
2309 } else {
2310 /*
2311 * The preimage extends beyond the end of img, so
2312 * there cannot be an exact match.
2313 *
2314 * There must be one non-blank context line that match
2315 * a line before the end of img.
2316 */
2317 char *buf_end;
2318
2319 buf = preimage->buf;
2320 buf_end = buf;
2321 for (i = 0; i < preimage_limit; i++)
2322 buf_end += preimage->line[i].len;
2323
2324 for ( ; buf < buf_end; buf++)
2325 if (!isspace(*buf))
2326 break;
2327 if (buf == buf_end)
2328 return 0;
2329 }
dc41976a 2330
86c91f91
GB
2331 /*
2332 * No exact match. If we are ignoring whitespace, run a line-by-line
2333 * fuzzy matching. We collect all the line length information because
2334 * we need it to adjust whitespace if we match.
2335 */
2336 if (ws_ignore_action == ignore_ws_change) {
2337 size_t imgoff = 0;
2338 size_t preoff = 0;
2339 size_t postlen = postimage->len;
51667147
BG
2340 size_t extra_chars;
2341 char *preimage_eof;
2342 char *preimage_end;
2343 for (i = 0; i < preimage_limit; i++) {
86c91f91 2344 size_t prelen = preimage->line[i].len;
0b1fac32 2345 size_t imglen = img->line[try_lno+i].len;
86c91f91 2346
0b1fac32
JH
2347 if (!fuzzy_matchlines(img->buf + try + imgoff, imglen,
2348 preimage->buf + preoff, prelen))
86c91f91
GB
2349 return 0;
2350 if (preimage->line[i].flag & LINE_COMMON)
0b1fac32
JH
2351 postlen += imglen - prelen;
2352 imgoff += imglen;
86c91f91
GB
2353 preoff += prelen;
2354 }
2355
2356 /*
9b25949a
BG
2357 * Ok, the preimage matches with whitespace fuzz.
2358 *
2359 * imgoff now holds the true length of the target that
51667147
BG
2360 * matches the preimage before the end of the file.
2361 *
2362 * Count the number of characters in the preimage that fall
2363 * beyond the end of the file and make sure that all of them
2364 * are whitespace characters. (This can only happen if
2365 * we are removing blank lines at the end of the file.)
86c91f91 2366 */
51667147
BG
2367 buf = preimage_eof = preimage->buf + preoff;
2368 for ( ; i < preimage->nr; i++)
2369 preoff += preimage->line[i].len;
2370 preimage_end = preimage->buf + preoff;
2371 for ( ; buf < preimage_end; buf++)
2372 if (!isspace(*buf))
2373 return 0;
2374
2375 /*
2376 * Update the preimage and the common postimage context
2377 * lines to use the same whitespace as the target.
2378 * If whitespace is missing in the target (i.e.
2379 * if the preimage extends beyond the end of the file),
2380 * use the whitespace from the preimage.
2381 */
2382 extra_chars = preimage_end - preimage_eof;
d511bd33
CW
2383 strbuf_init(&fixed, imgoff + extra_chars);
2384 strbuf_add(&fixed, img->buf + try, imgoff);
2385 strbuf_add(&fixed, preimage_eof, extra_chars);
2386 fixed_buf = strbuf_detach(&fixed, &fixed_len);
86c91f91 2387 update_pre_post_images(preimage, postimage,
d511bd33 2388 fixed_buf, fixed_len, postlen);
86c91f91
GB
2389 return 1;
2390 }
2391
c1beba5b
JH
2392 if (ws_error_action != correct_ws_error)
2393 return 0;
2394
dc41976a 2395 /*
c1beba5b 2396 * The hunk does not apply byte-by-byte, but the hash says
923fc5ab 2397 * it might with whitespace fuzz. We weren't asked to
86c91f91
GB
2398 * ignore whitespace, we were asked to correct whitespace
2399 * errors, so let's try matching after whitespace correction.
51667147 2400 *
407a792e
JH
2401 * While checking the preimage against the target, whitespace
2402 * errors in both fixed, we count how large the corresponding
2403 * postimage needs to be. The postimage prepared by
2404 * apply_one_fragment() has whitespace errors fixed on added
2405 * lines already, but the common lines were propagated as-is,
2406 * which may become longer when their whitespace errors are
2407 * fixed.
2408 */
2409
2410 /* First count added lines in postimage */
2411 postlen = 0;
2412 for (i = 0; i < postimage->nr; i++) {
2413 if (!(postimage->line[i].flag & LINE_COMMON))
2414 postlen += postimage->line[i].len;
2415 }
2416
2417 /*
51667147
BG
2418 * The preimage may extend beyond the end of the file,
2419 * but in this loop we will only handle the part of the
2420 * preimage that falls within the file.
dc41976a 2421 */
d511bd33 2422 strbuf_init(&fixed, preimage->len + 1);
c1beba5b
JH
2423 orig = preimage->buf;
2424 target = img->buf + try;
51667147 2425 for (i = 0; i < preimage_limit; i++) {
c1beba5b
JH
2426 size_t oldlen = preimage->line[i].len;
2427 size_t tgtlen = img->line[try_lno + i].len;
d511bd33
CW
2428 size_t fixstart = fixed.len;
2429 struct strbuf tgtfix;
c1beba5b
JH
2430 int match;
2431
2432 /* Try fixing the line in the preimage */
d511bd33 2433 ws_fix_copy(&fixed, orig, oldlen, ws_rule, NULL);
c1beba5b
JH
2434
2435 /* Try fixing the line in the target */
d511bd33
CW
2436 strbuf_init(&tgtfix, tgtlen);
2437 ws_fix_copy(&tgtfix, target, tgtlen, ws_rule, NULL);
c1beba5b
JH
2438
2439 /*
2440 * If they match, either the preimage was based on
2441 * a version before our tree fixed whitespace breakage,
2442 * or we are lacking a whitespace-fix patch the tree
2443 * the preimage was based on already had (i.e. target
2444 * has whitespace breakage, the preimage doesn't).
2445 * In either case, we are fixing the whitespace breakages
2446 * so we might as well take the fix together with their
2447 * real change.
2448 */
d511bd33
CW
2449 match = (tgtfix.len == fixed.len - fixstart &&
2450 !memcmp(tgtfix.buf, fixed.buf + fixstart,
2451 fixed.len - fixstart));
407a792e
JH
2452
2453 /* Add the length if this is common with the postimage */
2454 if (preimage->line[i].flag & LINE_COMMON)
2455 postlen += tgtfix.len;
c1beba5b 2456
d511bd33 2457 strbuf_release(&tgtfix);
c1beba5b
JH
2458 if (!match)
2459 goto unmatch_exit;
2460
2461 orig += oldlen;
c1beba5b 2462 target += tgtlen;
3cca928d
LT
2463 }
2464
51667147
BG
2465
2466 /*
2467 * Now handle the lines in the preimage that falls beyond the
2468 * end of the file (if any). They will only match if they are
2469 * empty or only contain whitespace (if WS_BLANK_AT_EOL is
2470 * false).
2471 */
2472 for ( ; i < preimage->nr; i++) {
d511bd33 2473 size_t fixstart = fixed.len; /* start of the fixed preimage */
51667147
BG
2474 size_t oldlen = preimage->line[i].len;
2475 int j;
2476
2477 /* Try fixing the line in the preimage */
d511bd33 2478 ws_fix_copy(&fixed, orig, oldlen, ws_rule, NULL);
51667147 2479
d511bd33
CW
2480 for (j = fixstart; j < fixed.len; j++)
2481 if (!isspace(fixed.buf[j]))
51667147
BG
2482 goto unmatch_exit;
2483
2484 orig += oldlen;
51667147
BG
2485 }
2486
c1beba5b
JH
2487 /*
2488 * Yes, the preimage is based on an older version that still
2489 * has whitespace breakages unfixed, and fixing them makes the
2490 * hunk match. Update the context lines in the postimage.
2491 */
d511bd33 2492 fixed_buf = strbuf_detach(&fixed, &fixed_len);
250b3c6c
JH
2493 if (postlen < postimage->len)
2494 postlen = 0;
c1beba5b 2495 update_pre_post_images(preimage, postimage,
250b3c6c 2496 fixed_buf, fixed_len, postlen);
c1beba5b
JH
2497 return 1;
2498
2499 unmatch_exit:
d511bd33 2500 strbuf_release(&fixed);
dc41976a 2501 return 0;
c89fb6b1
JH
2502}
2503
b94f2eda
JH
2504static int find_pos(struct image *img,
2505 struct image *preimage,
2506 struct image *postimage,
2507 int line,
c607aaa2 2508 unsigned ws_rule,
b94f2eda 2509 int match_beginning, int match_end)
3cca928d 2510{
b94f2eda
JH
2511 int i;
2512 unsigned long backwards, forwards, try;
2513 int backwards_lno, forwards_lno, try_lno;
3cca928d 2514
ecf4c2ec 2515 /*
24ff4d56 2516 * If match_beginning or match_end is specified, there is no
ecf4c2ec
JH
2517 * point starting from a wrong line that will never match and
2518 * wander around and wait for a match at the specified end.
2519 */
2520 if (match_beginning)
2521 line = 0;
2522 else if (match_end)
2523 line = img->nr - preimage->nr;
2524
24ff4d56
BG
2525 /*
2526 * Because the comparison is unsigned, the following test
2527 * will also take care of a negative line number that can
2528 * result when match_end and preimage is larger than the target.
2529 */
2530 if ((size_t) line > img->nr)
52f3c81a
JH
2531 line = img->nr;
2532
b94f2eda
JH
2533 try = 0;
2534 for (i = 0; i < line; i++)
2535 try += img->line[i].len;
3cca928d 2536
6e7c92a9
LT
2537 /*
2538 * There's probably some smart way to do this, but I'll leave
2539 * that to the smart and beautiful people. I'm simple and stupid.
2540 */
b94f2eda
JH
2541 backwards = try;
2542 backwards_lno = line;
2543 forwards = try;
2544 forwards_lno = line;
2545 try_lno = line;
fcb77bc5 2546
6e7c92a9 2547 for (i = 0; ; i++) {
b94f2eda 2548 if (match_fragment(img, preimage, postimage,
c607aaa2 2549 try, try_lno, ws_rule,
b94f2eda
JH
2550 match_beginning, match_end))
2551 return try_lno;
fcb77bc5
JH
2552
2553 again:
b94f2eda 2554 if (backwards_lno == 0 && forwards_lno == img->nr)
fcb77bc5 2555 break;
6e7c92a9 2556
6e7c92a9 2557 if (i & 1) {
b94f2eda 2558 if (backwards_lno == 0) {
fcb77bc5
JH
2559 i++;
2560 goto again;
6e7c92a9 2561 }
b94f2eda
JH
2562 backwards_lno--;
2563 backwards -= img->line[backwards_lno].len;
6e7c92a9 2564 try = backwards;
b94f2eda 2565 try_lno = backwards_lno;
6e7c92a9 2566 } else {
b94f2eda 2567 if (forwards_lno == img->nr) {
fcb77bc5
JH
2568 i++;
2569 goto again;
6e7c92a9 2570 }
b94f2eda
JH
2571 forwards += img->line[forwards_lno].len;
2572 forwards_lno++;
6e7c92a9 2573 try = forwards;
b94f2eda 2574 try_lno = forwards_lno;
6e7c92a9
LT
2575 }
2576
6e7c92a9 2577 }
3cca928d
LT
2578 return -1;
2579}
2580
b94f2eda 2581static void remove_first_line(struct image *img)
47495887 2582{
b94f2eda
JH
2583 img->buf += img->line[0].len;
2584 img->len -= img->line[0].len;
2585 img->line++;
2586 img->nr--;
47495887
EB
2587}
2588
b94f2eda 2589static void remove_last_line(struct image *img)
47495887 2590{
b94f2eda 2591 img->len -= img->line[--img->nr].len;
47495887
EB
2592}
2593
92737a22
JH
2594/*
2595 * The change from "preimage" and "postimage" has been found to
2596 * apply at applied_pos (counts in line numbers) in "img".
2597 * Update "img" to remove "preimage" and replace it with "postimage".
2598 */
b94f2eda
JH
2599static void update_image(struct image *img,
2600 int applied_pos,
2601 struct image *preimage,
2602 struct image *postimage)
b5767dd6 2603{
81bf96bb 2604 /*
b94f2eda
JH
2605 * remove the copy of preimage at offset in img
2606 * and replace it with postimage
81bf96bb 2607 */
b94f2eda
JH
2608 int i, nr;
2609 size_t remove_count, insert_count, applied_at = 0;
2610 char *result;
51667147
BG
2611 int preimage_limit;
2612
2613 /*
2614 * If we are removing blank lines at the end of img,
2615 * the preimage may extend beyond the end.
2616 * If that is the case, we must be careful only to
2617 * remove the part of the preimage that falls within
2618 * the boundaries of img. Initialize preimage_limit
2619 * to the number of lines in the preimage that falls
2620 * within the boundaries.
2621 */
2622 preimage_limit = preimage->nr;
2623 if (preimage_limit > img->nr - applied_pos)
2624 preimage_limit = img->nr - applied_pos;
d5a41641 2625
b94f2eda
JH
2626 for (i = 0; i < applied_pos; i++)
2627 applied_at += img->line[i].len;
2628
2629 remove_count = 0;
51667147 2630 for (i = 0; i < preimage_limit; i++)
b94f2eda
JH
2631 remove_count += img->line[applied_pos + i].len;
2632 insert_count = postimage->len;
2633
2634 /* Adjust the contents */
2635 result = xmalloc(img->len + insert_count - remove_count + 1);
2636 memcpy(result, img->buf, applied_at);
2637 memcpy(result + applied_at, postimage->buf, postimage->len);
2638 memcpy(result + applied_at + postimage->len,
2639 img->buf + (applied_at + remove_count),
2640 img->len - (applied_at + remove_count));
2641 free(img->buf);
2642 img->buf = result;
2643 img->len += insert_count - remove_count;
2644 result[img->len] = '\0';
2645
2646 /* Adjust the line table */
51667147
BG
2647 nr = img->nr + postimage->nr - preimage_limit;
2648 if (preimage_limit < postimage->nr) {
81bf96bb 2649 /*
b94f2eda
JH
2650 * NOTE: this knows that we never call remove_first_line()
2651 * on anything other than pre/post image.
d0c25035 2652 */
2756ca43 2653 REALLOC_ARRAY(img->line, nr);
b94f2eda 2654 img->line_allocated = img->line;
d0c25035 2655 }
51667147 2656 if (preimage_limit != postimage->nr)
b94f2eda 2657 memmove(img->line + applied_pos + postimage->nr,
51667147
BG
2658 img->line + applied_pos + preimage_limit,
2659 (img->nr - (applied_pos + preimage_limit)) *
b94f2eda
JH
2660 sizeof(*img->line));
2661 memcpy(img->line + applied_pos,
2662 postimage->line,
2663 postimage->nr * sizeof(*img->line));
933e44d3
JH
2664 if (!allow_overlap)
2665 for (i = 0; i < postimage->nr; i++)
2666 img->line[applied_pos + i].flag |= LINE_PATCHED;
b94f2eda 2667 img->nr = nr;
b5767dd6
JH
2668}
2669
92737a22
JH
2670/*
2671 * Use the patch-hunk text in "frag" to prepare two images (preimage and
2672 * postimage) for the hunk. Find lines that match "preimage" in "img" and
2673 * replace the part of "img" with "postimage" text.
2674 */
b94f2eda 2675static int apply_one_fragment(struct image *img, struct fragment *frag,
334f8cb2
JH
2676 int inaccurate_eof, unsigned ws_rule,
2677 int nth_fragment)
3cca928d 2678{
65aadb92 2679 int match_beginning, match_end;
3cca928d 2680 const char *patch = frag->patch;
b94f2eda 2681 int size = frag->size;
d511bd33
CW
2682 char *old, *oldlines;
2683 struct strbuf newlines;
077e1af5 2684 int new_blank_lines_at_end = 0;
85572639
JH
2685 int found_new_blank_lines_at_end = 0;
2686 int hunk_linenr = frag->linenr;
47495887 2687 unsigned long leading, trailing;
b94f2eda
JH
2688 int pos, applied_pos;
2689 struct image preimage;
2690 struct image postimage;
3cca928d 2691
c330fdd4
JH
2692 memset(&preimage, 0, sizeof(preimage));
2693 memset(&postimage, 0, sizeof(postimage));
61e08cca 2694 oldlines = xmalloc(size);
d511bd33 2695 strbuf_init(&newlines, size);
c330fdd4 2696
61e08cca 2697 old = oldlines;
3cca928d 2698 while (size > 0) {
e5a94313 2699 char first;
3cca928d 2700 int len = linelen(patch, size);
d511bd33 2701 int plen;
077e1af5 2702 int added_blank_line = 0;
efa57443 2703 int is_blank_context = 0;
d511bd33 2704 size_t start;
3cca928d
LT
2705
2706 if (!len)
2707 break;
2708
2709 /*
2710 * "plen" is how much of the line we should use for
2711 * the actual patch data. Normally we just remove the
2712 * first character on the line, but if the line is
2713 * followed by "\ No newline", then we also remove the
2714 * last one (which is the newline, of course).
2715 */
61e08cca 2716 plen = len - 1;
8b64647d 2717 if (len < size && patch[len] == '\\')
3cca928d 2718 plen--;
e5a94313 2719 first = *patch;
f686d030 2720 if (apply_in_reverse) {
e5a94313
JS
2721 if (first == '-')
2722 first = '+';
2723 else if (first == '+')
2724 first = '-';
2725 }
efe7f358 2726
e5a94313 2727 switch (first) {
b507b465
LT
2728 case '\n':
2729 /* Newer GNU diff, empty context line */
2730 if (plen < 0)
2731 /* ... followed by '\No newline'; nothing */
2732 break;
61e08cca 2733 *old++ = '\n';
d511bd33 2734 strbuf_addch(&newlines, '\n');
c330fdd4
JH
2735 add_line_info(&preimage, "\n", 1, LINE_COMMON);
2736 add_line_info(&postimage, "\n", 1, LINE_COMMON);
efa57443 2737 is_blank_context = 1;
b507b465 2738 break;
3cca928d 2739 case ' ':
94ea026b
JH
2740 if (plen && (ws_rule & WS_BLANK_AT_EOF) &&
2741 ws_blank_line(patch + 1, plen, ws_rule))
efa57443 2742 is_blank_context = 1;
3cca928d 2743 case '-':
61e08cca
JH
2744 memcpy(old, patch + 1, plen);
2745 add_line_info(&preimage, old, plen,
c330fdd4 2746 (first == ' ' ? LINE_COMMON : 0));
61e08cca 2747 old += plen;
e5a94313 2748 if (first == '-')
3cca928d
LT
2749 break;
2750 /* Fall-through for ' ' */
2751 case '+':
8441a9a8
JH
2752 /* --no-add does not add new lines */
2753 if (first == '+' && no_add)
2754 break;
2755
d511bd33 2756 start = newlines.len;
8441a9a8
JH
2757 if (first != '+' ||
2758 !whitespace_error ||
2759 ws_error_action != correct_ws_error) {
d511bd33 2760 strbuf_add(&newlines, patch + 1, plen);
8441a9a8
JH
2761 }
2762 else {
d511bd33 2763 ws_fix_copy(&newlines, patch + 1, plen, ws_rule, &applied_after_fixing_ws);
077e1af5 2764 }
d511bd33 2765 add_line_info(&postimage, newlines.buf + start, newlines.len - start,
8441a9a8 2766 (first == '+' ? 0 : LINE_COMMON));
8441a9a8 2767 if (first == '+' &&
94ea026b
JH
2768 (ws_rule & WS_BLANK_AT_EOF) &&
2769 ws_blank_line(patch + 1, plen, ws_rule))
8441a9a8 2770 added_blank_line = 1;
3cca928d
LT
2771 break;
2772 case '@': case '\\':
2773 /* Ignore it, we already handled it */
2774 break;
2775 default:
aeabfa07 2776 if (apply_verbosely)
3638eb43 2777 error(_("invalid start of line: '%c'"), first);
f0b1f1ec
SB
2778 applied_pos = -1;
2779 goto out;
3cca928d 2780 }
85572639
JH
2781 if (added_blank_line) {
2782 if (!new_blank_lines_at_end)
2783 found_new_blank_lines_at_end = hunk_linenr;
077e1af5 2784 new_blank_lines_at_end++;
85572639 2785 }
efa57443
JH
2786 else if (is_blank_context)
2787 ;
077e1af5
JH
2788 else
2789 new_blank_lines_at_end = 0;
3cca928d
LT
2790 patch += len;
2791 size -= len;
85572639 2792 hunk_linenr++;
3cca928d 2793 }
81bf96bb 2794 if (inaccurate_eof &&
61e08cca 2795 old > oldlines && old[-1] == '\n' &&
d511bd33 2796 newlines.len > 0 && newlines.buf[newlines.len - 1] == '\n') {
61e08cca 2797 old--;
d511bd33 2798 strbuf_setlen(&newlines, newlines.len - 1);
5b5d4d9e 2799 }
47495887 2800
47495887
EB
2801 leading = frag->leading;
2802 trailing = frag->trailing;
1bf1a859
LT
2803
2804 /*
ee5a317e
JH
2805 * A hunk to change lines at the beginning would begin with
2806 * @@ -1,L +N,M @@
ed0f47a8
JH
2807 * but we need to be careful. -U0 that inserts before the second
2808 * line also has this pattern.
4be60962 2809 *
ee5a317e
JH
2810 * And a hunk to add to an empty file would begin with
2811 * @@ -0,0 +N,M @@
2812 *
2813 * In other words, a hunk that is (frag->oldpos <= 1) with or
2814 * without leading context must match at the beginning.
1bf1a859 2815 */
ed0f47a8
JH
2816 match_beginning = (!frag->oldpos ||
2817 (frag->oldpos == 1 && !unidiff_zero));
ee5a317e
JH
2818
2819 /*
2820 * A hunk without trailing lines must match at the end.
2821 * However, we simply cannot tell if a hunk must match end
2822 * from the lack of trailing lines if the patch was generated
2823 * with unidiff without any context.
2824 */
2825 match_end = !unidiff_zero && !trailing;
1bf1a859 2826
b94f2eda 2827 pos = frag->newpos ? (frag->newpos - 1) : 0;
61e08cca
JH
2828 preimage.buf = oldlines;
2829 preimage.len = old - oldlines;
d511bd33
CW
2830 postimage.buf = newlines.buf;
2831 postimage.len = newlines.len;
c330fdd4
JH
2832 preimage.line = preimage.line_allocated;
2833 postimage.line = postimage.line_allocated;
2834
47495887 2835 for (;;) {
efe7f358 2836
c607aaa2
JH
2837 applied_pos = find_pos(img, &preimage, &postimage, pos,
2838 ws_rule, match_beginning, match_end);
b94f2eda
JH
2839
2840 if (applied_pos >= 0)
47495887 2841 break;
47495887
EB
2842
2843 /* Am I at my context limits? */
2844 if ((leading <= p_context) && (trailing <= p_context))
2845 break;
65aadb92
JH
2846 if (match_beginning || match_end) {
2847 match_beginning = match_end = 0;
1bf1a859
LT
2848 continue;
2849 }
b94f2eda 2850
81bf96bb
JH
2851 /*
2852 * Reduce the number of context lines; reduce both
2853 * leading and trailing if they are equal otherwise
2854 * just reduce the larger context.
47495887
EB
2855 */
2856 if (leading >= trailing) {
b94f2eda
JH
2857 remove_first_line(&preimage);
2858 remove_first_line(&postimage);
47495887
EB
2859 pos--;
2860 leading--;
2861 }
2862 if (trailing > leading) {
b94f2eda
JH
2863 remove_last_line(&preimage);
2864 remove_last_line(&postimage);
47495887 2865 trailing--;
6e7c92a9 2866 }
3cca928d
LT
2867 }
2868
b94f2eda 2869 if (applied_pos >= 0) {
77b15bbd 2870 if (new_blank_lines_at_end &&
51667147 2871 preimage.nr + applied_pos >= img->nr &&
77b15bbd
JH
2872 (ws_rule & WS_BLANK_AT_EOF) &&
2873 ws_error_action != nowarn_ws_error) {
85572639
JH
2874 record_ws_error(WS_BLANK_AT_EOF, "+", 1,
2875 found_new_blank_lines_at_end);
77b15bbd
JH
2876 if (ws_error_action == correct_ws_error) {
2877 while (new_blank_lines_at_end--)
2878 remove_last_line(&postimage);
2879 }
b94f2eda 2880 /*
77b15bbd
JH
2881 * We would want to prevent write_out_results()
2882 * from taking place in apply_patch() that follows
2883 * the callchain led us here, which is:
2884 * apply_patch->check_patch_list->check_patch->
2885 * apply_data->apply_fragments->apply_one_fragment
b94f2eda 2886 */
77b15bbd
JH
2887 if (ws_error_action == die_on_ws_error)
2888 apply = 0;
b94f2eda 2889 }
aeabfa07 2890
334f8cb2
JH
2891 if (apply_verbosely && applied_pos != pos) {
2892 int offset = applied_pos - pos;
2893 if (apply_in_reverse)
2894 offset = 0 - offset;
3638eb43
NTND
2895 fprintf_ln(stderr,
2896 Q_("Hunk #%d succeeded at %d (offset %d line).",
2897 "Hunk #%d succeeded at %d (offset %d lines).",
2898 offset),
2899 nth_fragment, applied_pos + 1, offset);
334f8cb2
JH
2900 }
2901
b94f2eda
JH
2902 /*
2903 * Warn if it was necessary to reduce the number
2904 * of context lines.
2905 */
2906 if ((leading != frag->leading) ||
2907 (trailing != frag->trailing))
3638eb43
NTND
2908 fprintf_ln(stderr, _("Context reduced to (%ld/%ld)"
2909 " to apply fragment at %d"),
2910 leading, trailing, applied_pos+1);
b94f2eda
JH
2911 update_image(img, applied_pos, &preimage, &postimage);
2912 } else {
2913 if (apply_verbosely)
3638eb43 2914 error(_("while searching for:\n%.*s"),
61e08cca 2915 (int)(old - oldlines), oldlines);
b94f2eda 2916 }
aeabfa07 2917
f0b1f1ec 2918out:
61e08cca 2919 free(oldlines);
d511bd33 2920 strbuf_release(&newlines);
b94f2eda
JH
2921 free(preimage.line_allocated);
2922 free(postimage.line_allocated);
2923
2924 return (applied_pos < 0);
3cca928d
LT
2925}
2926
b94f2eda 2927static int apply_binary_fragment(struct image *img, struct patch *patch)
0660626c 2928{
0660626c 2929 struct fragment *fragment = patch->fragments;
c7f9cb14
PH
2930 unsigned long len;
2931 void *dst;
0660626c 2932
24305cd7 2933 if (!fragment)
3638eb43 2934 return error(_("missing binary patch data for '%s'"),
24305cd7
JK
2935 patch->new_name ?
2936 patch->new_name :
2937 patch->old_name);
2938
3cd4f5e8
JH
2939 /* Binary patch is irreversible without the optional second hunk */
2940 if (apply_in_reverse) {
2941 if (!fragment->next)
2942 return error("cannot reverse-apply a binary patch "
2943 "without the reverse hunk to '%s'",
2944 patch->new_name
2945 ? patch->new_name : patch->old_name);
03eb8f8a 2946 fragment = fragment->next;
3cd4f5e8 2947 }
3cd4f5e8 2948 switch (fragment->binary_patch_method) {
0660626c 2949 case BINARY_DELTA_DEFLATED:
b94f2eda 2950 dst = patch_delta(img->buf, img->len, fragment->patch,
c7f9cb14
PH
2951 fragment->size, &len);
2952 if (!dst)
2953 return -1;
b94f2eda
JH
2954 clear_image(img);
2955 img->buf = dst;
2956 img->len = len;
c7f9cb14 2957 return 0;
0660626c 2958 case BINARY_LITERAL_DEFLATED:
b94f2eda
JH
2959 clear_image(img);
2960 img->len = fragment->size;
5c0b13f8 2961 img->buf = xmemdupz(fragment->patch, img->len);
c7f9cb14 2962 return 0;
0660626c 2963 }
c7f9cb14 2964 return -1;
0660626c
JH
2965}
2966
92737a22
JH
2967/*
2968 * Replace "img" with the result of applying the binary patch.
2969 * The binary patch data itself in patch->fragment is still kept
2970 * but the preimage prepared by the caller in "img" is freed here
2971 * or in the helper function apply_binary_fragment() this calls.
2972 */
b94f2eda 2973static int apply_binary(struct image *img, struct patch *patch)
3cca928d 2974{
011f4274 2975 const char *name = patch->old_name ? patch->old_name : patch->new_name;
051308f6 2976 unsigned char sha1[20];
011f4274 2977
81bf96bb
JH
2978 /*
2979 * For safety, we require patch index line to contain
051308f6
JH
2980 * full 40-byte textual SHA1 for old and new, at least for now.
2981 */
2982 if (strlen(patch->old_sha1_prefix) != 40 ||
2983 strlen(patch->new_sha1_prefix) != 40 ||
2984 get_sha1_hex(patch->old_sha1_prefix, sha1) ||
2985 get_sha1_hex(patch->new_sha1_prefix, sha1))
2986 return error("cannot apply binary patch to '%s' "
2987 "without full index line", name);
011f4274 2988
051308f6 2989 if (patch->old_name) {
81bf96bb
JH
2990 /*
2991 * See if the old one matches what the patch
051308f6 2992 * applies to.
011f4274 2993 */
b94f2eda 2994 hash_sha1_file(img->buf, img->len, blob_type, sha1);
051308f6
JH
2995 if (strcmp(sha1_to_hex(sha1), patch->old_sha1_prefix))
2996 return error("the patch applies to '%s' (%s), "
2997 "which does not match the "
2998 "current contents.",
2999 name, sha1_to_hex(sha1));
3000 }
3001 else {
3002 /* Otherwise, the old one must be empty. */
b94f2eda 3003 if (img->len)
051308f6
JH
3004 return error("the patch applies to an empty "
3005 "'%s' but it is not empty", name);
3006 }
011f4274 3007
0660626c 3008 get_sha1_hex(patch->new_sha1_prefix, sha1);
0bef57ee 3009 if (is_null_sha1(sha1)) {
b94f2eda 3010 clear_image(img);
051308f6 3011 return 0; /* deletion patch */
0660626c 3012 }
011f4274 3013
051308f6 3014 if (has_sha1_file(sha1)) {
0660626c 3015 /* We already have the postimage */
21666f1a 3016 enum object_type type;
051308f6 3017 unsigned long size;
c7f9cb14 3018 char *result;
051308f6 3019
c7f9cb14
PH
3020 result = read_sha1_file(sha1, &type, &size);
3021 if (!result)
051308f6
JH
3022 return error("the necessary postimage %s for "
3023 "'%s' cannot be read",
3024 patch->new_sha1_prefix, name);
b94f2eda
JH
3025 clear_image(img);
3026 img->buf = result;
3027 img->len = size;
c7f9cb14 3028 } else {
81bf96bb
JH
3029 /*
3030 * We have verified buf matches the preimage;
0660626c
JH
3031 * apply the patch data to it, which is stored
3032 * in the patch->fragments->{patch,size}.
011f4274 3033 */
b94f2eda 3034 if (apply_binary_fragment(img, patch))
3638eb43 3035 return error(_("binary patch does not apply to '%s'"),
051308f6 3036 name);
011f4274 3037
051308f6 3038 /* verify that the result matches */
b94f2eda 3039 hash_sha1_file(img->buf, img->len, blob_type, sha1);
051308f6 3040 if (strcmp(sha1_to_hex(sha1), patch->new_sha1_prefix))
3638eb43 3041 return error(_("binary patch to '%s' creates incorrect result (expecting %s, got %s)"),
c7f9cb14 3042 name, patch->new_sha1_prefix, sha1_to_hex(sha1));
011f4274 3043 }
3cca928d 3044
051308f6
JH
3045 return 0;
3046}
3047
b94f2eda 3048static int apply_fragments(struct image *img, struct patch *patch)
051308f6
JH
3049{
3050 struct fragment *frag = patch->fragments;
3051 const char *name = patch->old_name ? patch->old_name : patch->new_name;
cf1b7869
JH
3052 unsigned ws_rule = patch->ws_rule;
3053 unsigned inaccurate_eof = patch->inaccurate_eof;
334f8cb2 3054 int nth = 0;
051308f6
JH
3055
3056 if (patch->is_binary)
b94f2eda 3057 return apply_binary(img, patch);
051308f6 3058
3cca928d 3059 while (frag) {
334f8cb2
JH
3060 nth++;
3061 if (apply_one_fragment(img, frag, inaccurate_eof, ws_rule, nth)) {
3638eb43 3062 error(_("patch failed: %s:%ld"), name, frag->oldpos);
57dc397c
JH
3063 if (!apply_with_reject)
3064 return -1;
3065 frag->rejected = 1;
3066 }
3cca928d
LT
3067 frag = frag->next;
3068 }
30996652 3069 return 0;
3cca928d
LT
3070}
3071
e42a96e7 3072static int read_blob_object(struct strbuf *buf, const unsigned char *sha1, unsigned mode)
e06c5a6c 3073{
e42a96e7 3074 if (S_ISGITLINK(mode)) {
c7f9cb14 3075 strbuf_grow(buf, 100);
e42a96e7 3076 strbuf_addf(buf, "Subproject commit %s\n", sha1_to_hex(sha1));
e06c5a6c
SV
3077 } else {
3078 enum object_type type;
c7f9cb14
PH
3079 unsigned long sz;
3080 char *result;
3081
e42a96e7 3082 result = read_sha1_file(sha1, &type, &sz);
c7f9cb14 3083 if (!result)
e06c5a6c 3084 return -1;
c7f9cb14
PH
3085 /* XXX read_sha1_file NUL-terminates */
3086 strbuf_attach(buf, result, sz, sz + 1);
e06c5a6c
SV
3087 }
3088 return 0;
3089}
3090
9c5e6c80 3091static int read_file_or_gitlink(const struct cache_entry *ce, struct strbuf *buf)
e42a96e7
JH
3092{
3093 if (!ce)
3094 return 0;
3095 return read_blob_object(buf, ce->sha1, ce->ce_mode);
3096}
3097
7a07841c
DZ
3098static struct patch *in_fn_table(const char *name)
3099{
c455c87c 3100 struct string_list_item *item;
7a07841c
DZ
3101
3102 if (name == NULL)
3103 return NULL;
3104
e8c8b713 3105 item = string_list_lookup(&fn_table, name);
7a07841c
DZ
3106 if (item != NULL)
3107 return (struct patch *)item->util;
3108
3109 return NULL;
3110}
3111
7fac0eef
MK
3112/*
3113 * item->util in the filename table records the status of the path.
3114 * Usually it points at a patch (whose result records the contents
3115 * of it after applying it), but it could be PATH_WAS_DELETED for a
f3b8f91a
JH
3116 * path that a previously applied patch has already removed, or
3117 * PATH_TO_BE_DELETED for a path that a later patch would remove.
3118 *
3119 * The latter is needed to deal with a case where two paths A and B
3120 * are swapped by first renaming A to B and then renaming B to A;
41ccfdd9 3121 * moving A to B should not be prevented due to presence of B as we
f3b8f91a 3122 * will remove it in a later patch.
7fac0eef 3123 */
f3b8f91a 3124#define PATH_TO_BE_DELETED ((struct patch *) -2)
7fac0eef
MK
3125#define PATH_WAS_DELETED ((struct patch *) -1)
3126
3127static int to_be_deleted(struct patch *patch)
3128{
3129 return patch == PATH_TO_BE_DELETED;
3130}
3131
3132static int was_deleted(struct patch *patch)
3133{
3134 return patch == PATH_WAS_DELETED;
3135}
3136
7a07841c
DZ
3137static void add_to_fn_table(struct patch *patch)
3138{
c455c87c 3139 struct string_list_item *item;
7a07841c
DZ
3140
3141 /*
3142 * Always add new_name unless patch is a deletion
3143 * This should cover the cases for normal diffs,
3144 * file creations and copies
3145 */
3146 if (patch->new_name != NULL) {
78a395d3 3147 item = string_list_insert(&fn_table, patch->new_name);
7a07841c
DZ
3148 item->util = patch;
3149 }
3150
3151 /*
3152 * store a failure on rename/deletion cases because
3153 * later chunks shouldn't patch old names
3154 */
3155 if ((patch->new_name == NULL) || (patch->is_rename)) {
78a395d3 3156 item = string_list_insert(&fn_table, patch->old_name);
7fac0eef
MK
3157 item->util = PATH_WAS_DELETED;
3158 }
3159}
3160
3161static void prepare_fn_table(struct patch *patch)
3162{
3163 /*
3164 * store information about incoming file deletion
3165 */
3166 while (patch) {
3167 if ((patch->new_name == NULL) || (patch->is_rename)) {
3168 struct string_list_item *item;
78a395d3 3169 item = string_list_insert(&fn_table, patch->old_name);
7fac0eef
MK
3170 item->util = PATH_TO_BE_DELETED;
3171 }
3172 patch = patch->next;
7a07841c
DZ
3173 }
3174}
3175
d4a2024a
NTND
3176static int checkout_target(struct index_state *istate,
3177 struct cache_entry *ce, struct stat *st)
f4c66eed
JH
3178{
3179 struct checkout costate;
3180
3181 memset(&costate, 0, sizeof(costate));
3182 costate.base_dir = "";
3183 costate.refresh_cache = 1;
d4a2024a 3184 costate.istate = istate;
f4c66eed
JH
3185 if (checkout_entry(ce, &costate, NULL) || lstat(ce->name, st))
3186 return error(_("cannot checkout %s"), ce->name);
3187 return 0;
3188}
3189
ccf998b2
JH
3190static struct patch *previous_patch(struct patch *patch, int *gone)
3191{
3192 struct patch *previous;
3193
3194 *gone = 0;
3195 if (patch->is_copy || patch->is_rename)
3196 return NULL; /* "git" patches do not depend on the order */
3197
3198 previous = in_fn_table(patch->old_name);
3199 if (!previous)
3200 return NULL;
3201
3202 if (to_be_deleted(previous))
3203 return NULL; /* the deletion hasn't happened yet */
3204
3205 if (was_deleted(previous))
3206 *gone = 1;
3207
3208 return previous;
3209}
3210
9c5e6c80 3211static int verify_index_match(const struct cache_entry *ce, struct stat *st)
e09837e2
JH
3212{
3213 if (S_ISGITLINK(ce->ce_mode)) {
3214 if (!S_ISDIR(st->st_mode))
3215 return -1;
3216 return 0;
3217 }
3218 return ce_match_stat(ce, st, CE_MATCH_IGNORE_VALID|CE_MATCH_IGNORE_SKIP_WORKTREE);
3219}
3220
5a812661
JH
3221#define SUBMODULE_PATCH_WITHOUT_INDEX 1
3222
3223static int load_patch_target(struct strbuf *buf,
9c5e6c80 3224 const struct cache_entry *ce,
5a812661
JH
3225 struct stat *st,
3226 const char *name,
3227 unsigned expected_mode)
3228{
3c37a2e3 3229 if (cached || check_index) {
5a812661
JH
3230 if (read_file_or_gitlink(ce, buf))
3231 return error(_("read of %s failed"), name);
3232 } else if (name) {
3233 if (S_ISGITLINK(expected_mode)) {
3234 if (ce)
3235 return read_file_or_gitlink(ce, buf);
3236 else
3237 return SUBMODULE_PATCH_WITHOUT_INDEX;
fdc2c3a9
JH
3238 } else if (has_symlink_leading_path(name, strlen(name))) {
3239 return error(_("reading from '%s' beyond a symbolic link"), name);
5a812661
JH
3240 } else {
3241 if (read_old_data(st, name, buf))
3242 return error(_("read of %s failed"), name);
3243 }
3244 }
3245 return 0;
3246}
3247
ccf998b2
JH
3248/*
3249 * We are about to apply "patch"; populate the "image" with the
3250 * current version we have, from the working tree or from the index,
3251 * depending on the situation e.g. --cached/--index. If we are
3252 * applying a non-git patch that incrementally updates the tree,
3253 * we read from the result of a previous diff.
3254 */
37b9c903 3255static int load_preimage(struct image *image,
9c5e6c80
NTND
3256 struct patch *patch, struct stat *st,
3257 const struct cache_entry *ce)
3cca928d 3258{
f285a2d7 3259 struct strbuf buf = STRBUF_INIT;
b94f2eda
JH
3260 size_t len;
3261 char *img;
ccf998b2
JH
3262 struct patch *previous;
3263 int status;
3cca928d 3264
ccf998b2
JH
3265 previous = previous_patch(patch, &status);
3266 if (status)
3267 return error(_("path %s has been renamed/deleted"),
3268 patch->old_name);
3269 if (previous) {
92737a22 3270 /* We have a patched copy in memory; use that. */
ccf998b2 3271 strbuf_add(&buf, previous->result, previous->resultsize);
5a812661
JH
3272 } else {
3273 status = load_patch_target(&buf, ce, st,
3274 patch->old_name, patch->old_mode);
3275 if (status < 0)
3276 return status;
3277 else if (status == SUBMODULE_PATCH_WITHOUT_INDEX) {
3278 /*
3279 * There is no way to apply subproject
3280 * patch without looking at the index.
3281 * NEEDSWORK: shouldn't this be flagged
3282 * as an error???
3283 */
3284 free_fragment_list(patch->fragments);
3285 patch->fragments = NULL;
3286 } else if (status) {
3638eb43 3287 return error(_("read of %s failed"), patch->old_name);
04e4888e
JH
3288 }
3289 }
6e7c92a9 3290
b94f2eda 3291 img = strbuf_detach(&buf, &len);
37b9c903
JH
3292 prepare_image(image, img, len, !patch->is_binary);
3293 return 0;
3294}
b94f2eda 3295
519d1a5b
JH
3296static int three_way_merge(struct image *image,
3297 char *path,
3298 const unsigned char *base,
3299 const unsigned char *ours,
3300 const unsigned char *theirs)
3301{
28ff0512
JH
3302 mmfile_t base_file, our_file, their_file;
3303 mmbuffer_t result = { NULL };
3304 int status;
5aa7d94c 3305
28ff0512
JH
3306 read_mmblob(&base_file, base);
3307 read_mmblob(&our_file, ours);
3308 read_mmblob(&their_file, theirs);
3309 status = ll_merge(&result, path,
3310 &base_file, "base",
3311 &our_file, "ours",
3312 &their_file, "theirs", NULL);
3313 free(base_file.ptr);
3314 free(our_file.ptr);
3315 free(their_file.ptr);
3316 if (status < 0 || !result.ptr) {
3317 free(result.ptr);
3318 return -1;
3319 }
3320 clear_image(image);
3321 image->buf = result.ptr;
3322 image->len = result.size;
5aa7d94c 3323
28ff0512 3324 return status;
519d1a5b
JH
3325}
3326
099f3c42
JH
3327/*
3328 * When directly falling back to add/add three-way merge, we read from
3329 * the current contents of the new_name. In no cases other than that
3330 * this function will be called.
3331 */
3332static int load_current(struct image *image, struct patch *patch)
3333{
3334 struct strbuf buf = STRBUF_INIT;
3335 int status, pos;
3336 size_t len;
3337 char *img;
3338 struct stat st;
3339 struct cache_entry *ce;
3340 char *name = patch->new_name;
3341 unsigned mode = patch->new_mode;
3342
3343 if (!patch->is_new)
3344 die("BUG: patch to %s is not a creation", patch->old_name);
3345
3346 pos = cache_name_pos(name, strlen(name));
3347 if (pos < 0)
3348 return error(_("%s: does not exist in index"), name);
3349 ce = active_cache[pos];
3350 if (lstat(name, &st)) {
3351 if (errno != ENOENT)
3352 return error(_("%s: %s"), name, strerror(errno));
d4a2024a 3353 if (checkout_target(&the_index, ce, &st))
099f3c42
JH
3354 return -1;
3355 }
3356 if (verify_index_match(ce, &st))
3357 return error(_("%s: does not match index"), name);
3358
3359 status = load_patch_target(&buf, ce, &st, name, mode);
3360 if (status < 0)
3361 return status;
3362 else if (status)
3363 return -1;
3364 img = strbuf_detach(&buf, &len);
3365 prepare_image(image, img, len, !patch->is_binary);
3cca928d
LT
3366 return 0;
3367}
3368
cfb6f9ac 3369static int try_threeway(struct image *image, struct patch *patch,
9c5e6c80 3370 struct stat *st, const struct cache_entry *ce)
64cab591 3371{
519d1a5b
JH
3372 unsigned char pre_sha1[20], post_sha1[20], our_sha1[20];
3373 struct strbuf buf = STRBUF_INIT;
3374 size_t len;
28ff0512 3375 int status;
519d1a5b
JH
3376 char *img;
3377 struct image tmp_image;
3378
3379 /* No point falling back to 3-way merge in these cases */
099f3c42 3380 if (patch->is_delete ||
519d1a5b
JH
3381 S_ISGITLINK(patch->old_mode) || S_ISGITLINK(patch->new_mode))
3382 return -1;
64cab591 3383
519d1a5b 3384 /* Preimage the patch was prepared for */
099f3c42
JH
3385 if (patch->is_new)
3386 write_sha1_file("", 0, blob_type, pre_sha1);
3387 else if (get_sha1(patch->old_sha1_prefix, pre_sha1) ||
3388 read_blob_object(&buf, pre_sha1, patch->old_mode))
519d1a5b 3389 return error("repository lacks the necessary blob to fall back on 3-way merge.");
28ff0512
JH
3390
3391 fprintf(stderr, "Falling back to three-way merge...\n");
3392
519d1a5b
JH
3393 img = strbuf_detach(&buf, &len);
3394 prepare_image(&tmp_image, img, len, 1);
3395 /* Apply the patch to get the post image */
3396 if (apply_fragments(&tmp_image, patch) < 0) {
3397 clear_image(&tmp_image);
3398 return -1;
3399 }
3400 /* post_sha1[] is theirs */
3401 write_sha1_file(tmp_image.buf, tmp_image.len, blob_type, post_sha1);
3402 clear_image(&tmp_image);
3403
3404 /* our_sha1[] is ours */
099f3c42
JH
3405 if (patch->is_new) {
3406 if (load_current(&tmp_image, patch))
3407 return error("cannot read the current contents of '%s'",
3408 patch->new_name);
3409 } else {
3410 if (load_preimage(&tmp_image, patch, st, ce))
3411 return error("cannot read the current contents of '%s'",
3412 patch->old_name);
3413 }
519d1a5b
JH
3414 write_sha1_file(tmp_image.buf, tmp_image.len, blob_type, our_sha1);
3415 clear_image(&tmp_image);
3416
3417 /* in-core three-way merge between post and our using pre as base */
28ff0512
JH
3418 status = three_way_merge(image, patch->new_name,
3419 pre_sha1, our_sha1, post_sha1);
3420 if (status < 0) {
3421 fprintf(stderr, "Failed to fall back on three-way merge...\n");
3422 return status;
3423 }
3424
3425 if (status) {
3426 patch->conflicted_threeway = 1;
4f4a6cb9 3427 if (patch->is_new)
d07d4ab4 3428 oidclr(&patch->threeway_stage[0]);
4f4a6cb9 3429 else
d07d4ab4 3430 hashcpy(patch->threeway_stage[0].hash, pre_sha1);
3431 hashcpy(patch->threeway_stage[1].hash, our_sha1);
3432 hashcpy(patch->threeway_stage[2].hash, post_sha1);
28ff0512
JH
3433 fprintf(stderr, "Applied patch to '%s' with conflicts.\n", patch->new_name);
3434 } else {
3435 fprintf(stderr, "Applied patch to '%s' cleanly.\n", patch->new_name);
64cab591 3436 }
64cab591
JH
3437 return 0;
3438}
3439
9c5e6c80 3440static int apply_data(struct patch *patch, struct stat *st, const struct cache_entry *ce)
e06c5a6c 3441{
37b9c903
JH
3442 struct image image;
3443
3444 if (load_preimage(&image, patch, st, ce) < 0)
3445 return -1;
b94f2eda 3446
099f3c42
JH
3447 if (patch->direct_to_threeway ||
3448 apply_fragments(&image, patch) < 0) {
cfb6f9ac
JH
3449 /* Note: with --reject, apply_fragments() returns 0 */
3450 if (!threeway || try_threeway(&image, patch, st, ce) < 0)
e06c5a6c 3451 return -1;
e06c5a6c 3452 }
b94f2eda
JH
3453 patch->result = image.buf;
3454 patch->resultsize = image.len;
7a07841c 3455 add_to_fn_table(patch);
b94f2eda 3456 free(image.line_allocated);
5aa7d94c 3457
4be60962 3458 if (0 < patch->is_delete && patch->resultsize)
3638eb43 3459 return error(_("removal patch leaves file contents"));
5aa7d94c 3460
3cca928d 3461 return 0;
e06c5a6c
SV
3462}
3463
ccf998b2
JH
3464/*
3465 * If "patch" that we are looking at modifies or deletes what we have,
3466 * we would want it not to lose any local modification we have, either
3467 * in the working tree or in the index.
3468 *
3469 * This also decides if a non-git patch is a creation patch or a
3470 * modification to an existing empty file. We do not check the state
3471 * of the current tree for a creation patch in this function; the caller
3472 * check_patch() separately makes sure (and errors out otherwise) that
3473 * the path the patch creates does not exist in the current tree.
3474 */
5c47f4c6 3475static int check_preimage(struct patch *patch, struct cache_entry **ce, struct stat *st)
fab2c257
LT
3476{
3477 const char *old_name = patch->old_name;
ccf998b2
JH
3478 struct patch *previous = NULL;
3479 int stat_ret = 0, status;
5c47f4c6 3480 unsigned st_mode = 0;
e06c5a6c 3481
5c47f4c6
JH
3482 if (!old_name)
3483 return 0;
04e4888e 3484
5c47f4c6 3485 assert(patch->is_new <= 0);
ccf998b2 3486 previous = previous_patch(patch, &status);
a9a3e82e 3487
ccf998b2
JH
3488 if (status)
3489 return error(_("path %s has been renamed/deleted"), old_name);
3490 if (previous) {
3491 st_mode = previous->new_mode;
7a07841c 3492 } else if (!cached) {