lockfile.h: extract new header file for the functions in lockfile.c
[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;
7a988699 54static const char *fake_ancestor;
22943f1a 55static int line_termination = '\n';
f26c4940
MV
56static unsigned int p_context = UINT_MAX;
57static const char * const apply_usage[] = {
de373933 58 N_("git apply [options] [<patch>...]"),
f26c4940
MV
59 NULL
60};
c1bb9350 61
81bf96bb
JH
62static enum ws_error_action {
63 nowarn_ws_error,
64 warn_on_ws_error,
65 die_on_ws_error,
4b05548f 66 correct_ws_error
81bf96bb 67} ws_error_action = warn_on_ws_error;
96f1e58f 68static int whitespace_error;
fc96b7c9 69static int squelch_whitespace_errors = 5;
c94bf41c 70static int applied_after_fixing_ws;
86c91f91
GB
71
72static enum ws_ignore {
73 ignore_ws_none,
4b05548f 74 ignore_ws_change
86c91f91
GB
75} ws_ignore_action = ignore_ws_none;
76
77
96f1e58f 78static const char *patch_input_file;
c4730f35
JS
79static const char *root;
80static int root_len;
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 */
210 unsigned char threeway_stage[3][20];
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{
439 return !memcmp("/dev/null", str, 9) && isspace(str[9]);
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);
bb7306b5
JN
496 if (root)
497 strbuf_insert(&name, 0, root, root_len);
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 *null_strdup(const char *s)
661{
662 return s ? xstrdup(s) : NULL;
663}
664
665static char *find_name_common(const char *line, const char *def,
666 int p_value, const char *end, int terminate)
5a12c886
JN
667{
668 int len;
669 const char *start = NULL;
670
bb7306b5
JN
671 if (p_value == 0)
672 start = line;
5a12c886 673 while (line != end) {
a4acb0eb 674 char c = *line;
9a4a100e 675
5a12c886 676 if (!end && isspace(c)) {
9a4a100e
LT
677 if (c == '\n')
678 break;
679 if (name_terminate(start, line-start, c, terminate))
680 break;
681 }
a4acb0eb
LT
682 line++;
683 if (c == '/' && !--p_value)
684 start = line;
685 }
686 if (!start)
2901bbe5 687 return squash_slash(null_strdup(def));
a4acb0eb
LT
688 len = line - start;
689 if (!len)
2901bbe5 690 return squash_slash(null_strdup(def));
a4acb0eb
LT
691
692 /*
693 * Generally we prefer the shorter name, especially
694 * if the other one is just a variation of that with
695 * something else tacked on to the end (ie "file.orig"
696 * or "file~").
697 */
698 if (def) {
699 int deflen = strlen(def);
700 if (deflen < len && !strncmp(start, def, deflen))
2901bbe5 701 return squash_slash(xstrdup(def));
c1bb9350 702 }
a4acb0eb 703
c4730f35
JS
704 if (root) {
705 char *ret = xmalloc(root_len + len + 1);
706 strcpy(ret, root);
707 memcpy(ret + root_len, start, len);
708 ret[root_len + len] = '\0';
33eb4dd9 709 return squash_slash(ret);
c4730f35
JS
710 }
711
33eb4dd9 712 return squash_slash(xmemdupz(start, len));
a4acb0eb
LT
713}
714
5a12c886
JN
715static char *find_name(const char *line, char *def, int p_value, int terminate)
716{
717 if (*line == '"') {
718 char *name = find_name_gnu(line, def, p_value);
719 if (name)
720 return name;
721 }
722
723 return find_name_common(line, def, p_value, NULL, terminate);
724}
725
726static char *find_name_traditional(const char *line, char *def, int p_value)
727{
1f976bd0 728 size_t len;
5a12c886
JN
729 size_t date_len;
730
731 if (*line == '"') {
732 char *name = find_name_gnu(line, def, p_value);
733 if (name)
734 return name;
735 }
736
737 len = strchrnul(line, '\n') - line;
738 date_len = diff_timestamp_len(line, len);
739 if (!date_len)
740 return find_name_common(line, def, p_value, NULL, TERM_TAB);
741 len -= date_len;
742
743 return find_name_common(line, def, p_value, line + len, 0);
744}
745
3e8a5db9
JH
746static int count_slashes(const char *cp)
747{
748 int cnt = 0;
749 char ch;
750
751 while ((ch = *cp++))
752 if (ch == '/')
753 cnt++;
754 return cnt;
755}
756
757/*
758 * Given the string after "--- " or "+++ ", guess the appropriate
759 * p_value for the given patch.
760 */
761static int guess_p_value(const char *nameline)
762{
763 char *name, *cp;
764 int val = -1;
765
766 if (is_dev_null(nameline))
767 return -1;
5a12c886 768 name = find_name_traditional(nameline, NULL, 0);
3e8a5db9
JH
769 if (!name)
770 return -1;
771 cp = strchr(name, '/');
772 if (!cp)
773 val = 0;
774 else if (prefix) {
775 /*
776 * Does it begin with "a/$our-prefix" and such? Then this is
777 * very likely to apply to our directory.
778 */
779 if (!strncmp(name, prefix, prefix_length))
780 val = count_slashes(prefix);
781 else {
782 cp++;
783 if (!strncmp(cp, prefix, prefix_length))
784 val = count_slashes(prefix) + 1;
785 }
786 }
787 free(name);
788 return val;
789}
790
c4593faf
JH
791/*
792 * Does the ---/+++ line has the POSIX timestamp after the last HT?
793 * GNU diff puts epoch there to signal a creation/deletion event. Is
794 * this such a timestamp?
795 */
796static int has_epoch_timestamp(const char *nameline)
797{
798 /*
799 * We are only interested in epoch timestamp; any non-zero
800 * fraction cannot be one, hence "(\.0+)?" in the regexp below.
801 * For the same reason, the date must be either 1969-12-31 or
802 * 1970-01-01, and the seconds part must be "00".
803 */
804 const char stamp_regexp[] =
805 "^(1969-12-31|1970-01-01)"
806 " "
807 "[0-2][0-9]:[0-5][0-9]:00(\\.0+)?"
808 " "
a1980c4e
AK
809 "([-+][0-2][0-9]:?[0-5][0-9])\n";
810 const char *timestamp = NULL, *cp, *colon;
c4593faf
JH
811 static regex_t *stamp;
812 regmatch_t m[10];
813 int zoneoffset;
814 int hourminute;
815 int status;
816
817 for (cp = nameline; *cp != '\n'; cp++) {
818 if (*cp == '\t')
819 timestamp = cp + 1;
820 }
821 if (!timestamp)
822 return 0;
823 if (!stamp) {
824 stamp = xmalloc(sizeof(*stamp));
825 if (regcomp(stamp, stamp_regexp, REG_EXTENDED)) {
3638eb43 826 warning(_("Cannot prepare timestamp regexp %s"),
c4593faf
JH
827 stamp_regexp);
828 return 0;
829 }
830 }
831
832 status = regexec(stamp, timestamp, ARRAY_SIZE(m), m, 0);
833 if (status) {
834 if (status != REG_NOMATCH)
3638eb43 835 warning(_("regexec returned %d for input: %s"),
c4593faf
JH
836 status, timestamp);
837 return 0;
838 }
839
a1980c4e
AK
840 zoneoffset = strtol(timestamp + m[3].rm_so + 1, (char **) &colon, 10);
841 if (*colon == ':')
842 zoneoffset = zoneoffset * 60 + strtol(colon + 1, NULL, 10);
843 else
844 zoneoffset = (zoneoffset / 100) * 60 + (zoneoffset % 100);
c4593faf
JH
845 if (timestamp[m[3].rm_so] == '-')
846 zoneoffset = -zoneoffset;
847
848 /*
849 * YYYY-MM-DD hh:mm:ss must be from either 1969-12-31
850 * (west of GMT) or 1970-01-01 (east of GMT)
851 */
852 if ((zoneoffset < 0 && memcmp(timestamp, "1969-12-31", 10)) ||
853 (0 <= zoneoffset && memcmp(timestamp, "1970-01-01", 10)))
854 return 0;
855
856 hourminute = (strtol(timestamp + 11, NULL, 10) * 60 +
857 strtol(timestamp + 14, NULL, 10) -
858 zoneoffset);
859
860 return ((zoneoffset < 0 && hourminute == 1440) ||
861 (0 <= zoneoffset && !hourminute));
862}
863
a4acb0eb 864/*
88f6dbaf 865 * Get the name etc info from the ---/+++ lines of a traditional patch header
a4acb0eb 866 *
9a4a100e
LT
867 * FIXME! The end-of-filename heuristics are kind of screwy. For existing
868 * files, we can happily check the index for a match, but for creating a
869 * new file we should try to match whatever "patch" does. I have no idea.
a4acb0eb 870 */
19c58fb8 871static void parse_traditional_patch(const char *first, const char *second, struct patch *patch)
a4acb0eb 872{
a4acb0eb
LT
873 char *name;
874
a9486b02
PR
875 first += 4; /* skip "--- " */
876 second += 4; /* skip "+++ " */
3e8a5db9
JH
877 if (!p_value_known) {
878 int p, q;
879 p = guess_p_value(first);
880 q = guess_p_value(second);
881 if (p < 0) p = q;
882 if (0 <= p && p == q) {
883 p_value = p;
884 p_value_known = 1;
885 }
886 }
a4acb0eb 887 if (is_dev_null(first)) {
19c58fb8
LT
888 patch->is_new = 1;
889 patch->is_delete = 0;
5a12c886 890 name = find_name_traditional(second, NULL, p_value);
19c58fb8 891 patch->new_name = name;
a4acb0eb 892 } else if (is_dev_null(second)) {
19c58fb8
LT
893 patch->is_new = 0;
894 patch->is_delete = 1;
5a12c886 895 name = find_name_traditional(first, NULL, p_value);
19c58fb8 896 patch->old_name = name;
a4acb0eb 897 } else {
2901bbe5
JH
898 char *first_name;
899 first_name = find_name_traditional(first, NULL, p_value);
900 name = find_name_traditional(second, first_name, p_value);
901 free(first_name);
c4593faf
JH
902 if (has_epoch_timestamp(first)) {
903 patch->is_new = 1;
904 patch->is_delete = 0;
905 patch->new_name = name;
906 } else if (has_epoch_timestamp(second)) {
907 patch->is_new = 0;
908 patch->is_delete = 1;
909 patch->old_name = name;
910 } else {
2901bbe5 911 patch->old_name = name;
212eb96a 912 patch->new_name = null_strdup(name);
c4593faf 913 }
a4acb0eb
LT
914 }
915 if (!name)
3638eb43 916 die(_("unable to find filename in patch at line %d"), linenr);
a4acb0eb
LT
917}
918
19c58fb8 919static int gitdiff_hdrend(const char *line, struct patch *patch)
a4acb0eb
LT
920{
921 return -1;
922}
923
1e3f6b6e
LT
924/*
925 * We're anal about diff header consistency, to make
926 * sure that we don't end up having strange ambiguous
927 * patches floating around.
928 *
929 * As a result, gitdiff_{old|new}name() will check
930 * their names against any previous information, just
931 * to make sure..
932 */
4c5197d1
NTND
933#define DIFF_OLD_NAME 0
934#define DIFF_NEW_NAME 1
935
936static char *gitdiff_verify_name(const char *line, int isnull, char *orig_name, int side)
1e3f6b6e 937{
1e3f6b6e 938 if (!orig_name && !isnull)
79ee194e 939 return find_name(line, NULL, p_value, TERM_TAB);
1e3f6b6e 940
1e3f6b6e 941 if (orig_name) {
22943f1a
JH
942 int len;
943 const char *name;
944 char *another;
1e3f6b6e
LT
945 name = orig_name;
946 len = strlen(name);
947 if (isnull)
3638eb43 948 die(_("git apply: bad git-diff - expected /dev/null, got %s on line %d"), name, linenr);
79ee194e 949 another = find_name(line, NULL, p_value, TERM_TAB);
da915939 950 if (!another || memcmp(another, name, len + 1))
4c5197d1
NTND
951 die((side == DIFF_NEW_NAME) ?
952 _("git apply: bad git-diff - inconsistent new filename on line %d") :
953 _("git apply: bad git-diff - inconsistent old filename on line %d"), linenr);
22943f1a 954 free(another);
1e3f6b6e
LT
955 return orig_name;
956 }
22943f1a
JH
957 else {
958 /* expect "/dev/null" */
959 if (memcmp("/dev/null", line, 9) || line[9] != '\n')
3638eb43 960 die(_("git apply: bad git-diff - expected /dev/null on line %d"), linenr);
22943f1a
JH
961 return NULL;
962 }
1e3f6b6e
LT
963}
964
19c58fb8 965static int gitdiff_oldname(const char *line, struct patch *patch)
a4acb0eb 966{
2901bbe5 967 char *orig = patch->old_name;
4c5197d1
NTND
968 patch->old_name = gitdiff_verify_name(line, patch->is_new, patch->old_name,
969 DIFF_OLD_NAME);
2901bbe5
JH
970 if (orig != patch->old_name)
971 free(orig);
a4acb0eb
LT
972 return 0;
973}
974
19c58fb8 975static int gitdiff_newname(const char *line, struct patch *patch)
a4acb0eb 976{
2901bbe5 977 char *orig = patch->new_name;
4c5197d1
NTND
978 patch->new_name = gitdiff_verify_name(line, patch->is_delete, patch->new_name,
979 DIFF_NEW_NAME);
2901bbe5
JH
980 if (orig != patch->new_name)
981 free(orig);
a4acb0eb
LT
982 return 0;
983}
984
19c58fb8 985static int gitdiff_oldmode(const char *line, struct patch *patch)
a4acb0eb 986{
19c58fb8 987 patch->old_mode = strtoul(line, NULL, 8);
a4acb0eb
LT
988 return 0;
989}
990
19c58fb8 991static int gitdiff_newmode(const char *line, struct patch *patch)
a4acb0eb 992{
19c58fb8 993 patch->new_mode = strtoul(line, NULL, 8);
a4acb0eb
LT
994 return 0;
995}
996
19c58fb8 997static int gitdiff_delete(const char *line, struct patch *patch)
a4acb0eb 998{
19c58fb8 999 patch->is_delete = 1;
2901bbe5
JH
1000 free(patch->old_name);
1001 patch->old_name = null_strdup(patch->def_name);
19c58fb8 1002 return gitdiff_oldmode(line, patch);
a4acb0eb
LT
1003}
1004
19c58fb8 1005static int gitdiff_newfile(const char *line, struct patch *patch)
a4acb0eb 1006{
19c58fb8 1007 patch->is_new = 1;
2901bbe5
JH
1008 free(patch->new_name);
1009 patch->new_name = null_strdup(patch->def_name);
19c58fb8 1010 return gitdiff_newmode(line, patch);
a4acb0eb
LT
1011}
1012
19c58fb8 1013static int gitdiff_copysrc(const char *line, struct patch *patch)
a4acb0eb 1014{
19c58fb8 1015 patch->is_copy = 1;
2901bbe5 1016 free(patch->old_name);
cefd43b7 1017 patch->old_name = find_name(line, NULL, p_value ? p_value - 1 : 0, 0);
a4acb0eb
LT
1018 return 0;
1019}
1020
19c58fb8 1021static int gitdiff_copydst(const char *line, struct patch *patch)
a4acb0eb 1022{
19c58fb8 1023 patch->is_copy = 1;
2901bbe5 1024 free(patch->new_name);
cefd43b7 1025 patch->new_name = find_name(line, NULL, p_value ? p_value - 1 : 0, 0);
a4acb0eb
LT
1026 return 0;
1027}
1028
19c58fb8 1029static int gitdiff_renamesrc(const char *line, struct patch *patch)
a4acb0eb 1030{
19c58fb8 1031 patch->is_rename = 1;
2901bbe5 1032 free(patch->old_name);
cefd43b7 1033 patch->old_name = find_name(line, NULL, p_value ? p_value - 1 : 0, 0);
a4acb0eb
LT
1034 return 0;
1035}
1036
19c58fb8 1037static int gitdiff_renamedst(const char *line, struct patch *patch)
a4acb0eb 1038{
19c58fb8 1039 patch->is_rename = 1;
2901bbe5 1040 free(patch->new_name);
cefd43b7 1041 patch->new_name = find_name(line, NULL, p_value ? p_value - 1 : 0, 0);
a4acb0eb
LT
1042 return 0;
1043}
1044
19c58fb8 1045static int gitdiff_similarity(const char *line, struct patch *patch)
a4acb0eb 1046{
afcb6ac8
JK
1047 unsigned long val = strtoul(line, NULL, 10);
1048 if (val <= 100)
1049 patch->score = val;
a4acb0eb 1050 return 0;
c1bb9350
LT
1051}
1052
70aadac0
JH
1053static int gitdiff_dissimilarity(const char *line, struct patch *patch)
1054{
afcb6ac8
JK
1055 unsigned long val = strtoul(line, NULL, 10);
1056 if (val <= 100)
1057 patch->score = val;
70aadac0
JH
1058 return 0;
1059}
1060
2cf67f1e
JH
1061static int gitdiff_index(const char *line, struct patch *patch)
1062{
81bf96bb
JH
1063 /*
1064 * index line is N hexadecimal, "..", N hexadecimal,
2cf67f1e
JH
1065 * and optional space with octal mode.
1066 */
1067 const char *ptr, *eol;
1068 int len;
1069
1070 ptr = strchr(line, '.');
9add69b1 1071 if (!ptr || ptr[1] != '.' || 40 < ptr - line)
2cf67f1e
JH
1072 return 0;
1073 len = ptr - line;
1074 memcpy(patch->old_sha1_prefix, line, len);
1075 patch->old_sha1_prefix[len] = 0;
1076
1077 line = ptr + 2;
1078 ptr = strchr(line, ' ');
31bb6d37 1079 eol = strchrnul(line, '\n');
2cf67f1e
JH
1080
1081 if (!ptr || eol < ptr)
1082 ptr = eol;
1083 len = ptr - line;
1084
9add69b1 1085 if (40 < len)
2cf67f1e
JH
1086 return 0;
1087 memcpy(patch->new_sha1_prefix, line, len);
1088 patch->new_sha1_prefix[len] = 0;
1089 if (*ptr == ' ')
1f7903a3 1090 patch->old_mode = strtoul(ptr+1, NULL, 8);
2cf67f1e
JH
1091 return 0;
1092}
1093
9a4a100e
LT
1094/*
1095 * This is normal for a diff that doesn't change anything: we'll fall through
1096 * into the next diff. Tell the parser to break out.
1097 */
19c58fb8 1098static int gitdiff_unrecognized(const char *line, struct patch *patch)
9a4a100e
LT
1099{
1100 return -1;
1101}
1102
6a2abdc1
JH
1103/*
1104 * Skip p_value leading components from "line"; as we do not accept
1105 * absolute paths, return NULL in that case.
1106 */
1107static const char *skip_tree_prefix(const char *line, int llen)
22943f1a 1108{
6a2abdc1 1109 int nslash;
22943f1a
JH
1110 int i;
1111
6a2abdc1
JH
1112 if (!p_value)
1113 return (llen && line[0] == '/') ? NULL : line;
1114
1115 nslash = p_value;
22943f1a
JH
1116 for (i = 0; i < llen; i++) {
1117 int ch = line[i];
ec7fc0b1 1118 if (ch == '/' && --nslash <= 0)
6a2abdc1 1119 return (i == 0) ? NULL : &line[i + 1];
22943f1a
JH
1120 }
1121 return NULL;
1122}
1123
81bf96bb
JH
1124/*
1125 * This is to extract the same name that appears on "diff --git"
22943f1a
JH
1126 * line. We do not find and return anything if it is a rename
1127 * patch, and it is OK because we will find the name elsewhere.
1128 * We need to reliably find name only when it is mode-change only,
1129 * creation or deletion of an empty file. In any of these cases,
1130 * both sides are the same name under a/ and b/ respectively.
1131 */
26693ba8 1132static char *git_header_name(const char *line, int llen)
5041aa70 1133{
22943f1a
JH
1134 const char *name;
1135 const char *second = NULL;
cefd43b7 1136 size_t len, line_len;
5041aa70 1137
22943f1a
JH
1138 line += strlen("diff --git ");
1139 llen -= strlen("diff --git ");
1140
1141 if (*line == '"') {
1142 const char *cp;
f285a2d7
BC
1143 struct strbuf first = STRBUF_INIT;
1144 struct strbuf sp = STRBUF_INIT;
7fb1011e
PH
1145
1146 if (unquote_c_style(&first, line, &second))
1147 goto free_and_fail1;
22943f1a 1148
6a2abdc1
JH
1149 /* strip the a/b prefix including trailing slash */
1150 cp = skip_tree_prefix(first.buf, first.len);
1151 if (!cp)
7fb1011e 1152 goto free_and_fail1;
6a2abdc1 1153 strbuf_remove(&first, 0, cp - first.buf);
22943f1a 1154
81bf96bb
JH
1155 /*
1156 * second points at one past closing dq of name.
22943f1a
JH
1157 * find the second name.
1158 */
1159 while ((second < line + llen) && isspace(*second))
1160 second++;
1161
1162 if (line + llen <= second)
7fb1011e 1163 goto free_and_fail1;
22943f1a 1164 if (*second == '"') {
7fb1011e
PH
1165 if (unquote_c_style(&sp, second, NULL))
1166 goto free_and_fail1;
6a2abdc1
JH
1167 cp = skip_tree_prefix(sp.buf, sp.len);
1168 if (!cp)
7fb1011e 1169 goto free_and_fail1;
22943f1a 1170 /* They must match, otherwise ignore */
6a2abdc1 1171 if (strcmp(cp, first.buf))
7fb1011e
PH
1172 goto free_and_fail1;
1173 strbuf_release(&sp);
b315c5c0 1174 return strbuf_detach(&first, NULL);
22943f1a
JH
1175 }
1176
1177 /* unquoted second */
6a2abdc1
JH
1178 cp = skip_tree_prefix(second, line + llen - second);
1179 if (!cp)
7fb1011e 1180 goto free_and_fail1;
6a2abdc1 1181 if (line + llen - cp != first.len ||
7fb1011e
PH
1182 memcmp(first.buf, cp, first.len))
1183 goto free_and_fail1;
b315c5c0 1184 return strbuf_detach(&first, NULL);
7fb1011e
PH
1185
1186 free_and_fail1:
1187 strbuf_release(&first);
1188 strbuf_release(&sp);
1189 return NULL;
5041aa70
LT
1190 }
1191
22943f1a 1192 /* unquoted first name */
6a2abdc1
JH
1193 name = skip_tree_prefix(line, llen);
1194 if (!name)
5041aa70 1195 return NULL;
22943f1a 1196
81bf96bb
JH
1197 /*
1198 * since the first name is unquoted, a dq if exists must be
22943f1a
JH
1199 * the beginning of the second name.
1200 */
1201 for (second = name; second < line + llen; second++) {
1202 if (*second == '"') {
f285a2d7 1203 struct strbuf sp = STRBUF_INIT;
22943f1a 1204 const char *np;
7fb1011e 1205
7fb1011e
PH
1206 if (unquote_c_style(&sp, second, NULL))
1207 goto free_and_fail2;
1208
6a2abdc1
JH
1209 np = skip_tree_prefix(sp.buf, sp.len);
1210 if (!np)
7fb1011e 1211 goto free_and_fail2;
7fb1011e
PH
1212
1213 len = sp.buf + sp.len - np;
1214 if (len < second - name &&
22943f1a
JH
1215 !strncmp(np, name, len) &&
1216 isspace(name[len])) {
1217 /* Good */
7fb1011e 1218 strbuf_remove(&sp, 0, np - sp.buf);
b315c5c0 1219 return strbuf_detach(&sp, NULL);
22943f1a 1220 }
7fb1011e
PH
1221
1222 free_and_fail2:
1223 strbuf_release(&sp);
1224 return NULL;
22943f1a
JH
1225 }
1226 }
1227
5041aa70
LT
1228 /*
1229 * Accept a name only if it shows up twice, exactly the same
1230 * form.
1231 */
cefd43b7
FC
1232 second = strchr(name, '\n');
1233 if (!second)
1234 return NULL;
1235 line_len = second - name;
5041aa70 1236 for (len = 0 ; ; len++) {
dd305c84 1237 switch (name[len]) {
5041aa70
LT
1238 default:
1239 continue;
1240 case '\n':
e70a165d 1241 return NULL;
5041aa70 1242 case '\t': case ' ':
6a2abdc1
JH
1243 /*
1244 * Is this the separator between the preimage
1245 * and the postimage pathname? Again, we are
1246 * only interested in the case where there is
1247 * no rename, as this is only to set def_name
1248 * and a rename patch has the names elsewhere
1249 * in an unambiguous form.
1250 */
1251 if (!name[len + 1])
1252 return NULL; /* no postimage name */
1253 second = skip_tree_prefix(name + len + 1,
1254 line_len - (len + 1));
cefd43b7
FC
1255 if (!second)
1256 return NULL;
6a2abdc1
JH
1257 /*
1258 * Does len bytes starting at "name" and "second"
1259 * (that are separated by one HT or SP we just
1260 * found) exactly match?
1261 */
1262 if (second[len] == '\n' && !strncmp(name, second, len))
182af834 1263 return xmemdupz(name, len);
5041aa70
LT
1264 }
1265 }
5041aa70
LT
1266}
1267
c1bb9350 1268/* Verify that we recognize the lines following a git header */
26693ba8 1269static int parse_git_header(const char *line, int len, unsigned int size, struct patch *patch)
c1bb9350 1270{
a4acb0eb
LT
1271 unsigned long offset;
1272
1273 /* A git diff has explicit new/delete information, so we don't guess */
19c58fb8
LT
1274 patch->is_new = 0;
1275 patch->is_delete = 0;
a4acb0eb 1276
5041aa70
LT
1277 /*
1278 * Some things may not have the old name in the
1279 * rest of the headers anywhere (pure mode changes,
1280 * or removing or adding empty files), so we get
1281 * the default name from the header.
1282 */
22943f1a 1283 patch->def_name = git_header_name(line, len);
969c8775 1284 if (patch->def_name && root) {
b2724c87 1285 char *s = xstrfmt("%s%s", root, patch->def_name);
969c8775
JK
1286 free(patch->def_name);
1287 patch->def_name = s;
1288 }
5041aa70 1289
a4acb0eb
LT
1290 line += len;
1291 size -= len;
1292 linenr++;
1293 for (offset = len ; size > 0 ; offset += len, size -= len, line += len, linenr++) {
1294 static const struct opentry {
1295 const char *str;
19c58fb8 1296 int (*fn)(const char *, struct patch *);
a4acb0eb
LT
1297 } optable[] = {
1298 { "@@ -", gitdiff_hdrend },
1299 { "--- ", gitdiff_oldname },
1300 { "+++ ", gitdiff_newname },
1301 { "old mode ", gitdiff_oldmode },
1302 { "new mode ", gitdiff_newmode },
1303 { "deleted file mode ", gitdiff_delete },
1304 { "new file mode ", gitdiff_newfile },
1305 { "copy from ", gitdiff_copysrc },
1306 { "copy to ", gitdiff_copydst },
33f4d087
LT
1307 { "rename old ", gitdiff_renamesrc },
1308 { "rename new ", gitdiff_renamedst },
dc938417
LT
1309 { "rename from ", gitdiff_renamesrc },
1310 { "rename to ", gitdiff_renamedst },
a4acb0eb 1311 { "similarity index ", gitdiff_similarity },
70aadac0 1312 { "dissimilarity index ", gitdiff_dissimilarity },
2cf67f1e 1313 { "index ", gitdiff_index },
9a4a100e 1314 { "", gitdiff_unrecognized },
a4acb0eb
LT
1315 };
1316 int i;
c1bb9350 1317
c1bb9350 1318 len = linelen(line, size);
a4acb0eb 1319 if (!len || line[len-1] != '\n')
c1bb9350 1320 break;
b4f2a6ac 1321 for (i = 0; i < ARRAY_SIZE(optable); i++) {
a4acb0eb
LT
1322 const struct opentry *p = optable + i;
1323 int oplen = strlen(p->str);
1324 if (len < oplen || memcmp(p->str, line, oplen))
1325 continue;
19c58fb8 1326 if (p->fn(line + oplen, patch) < 0)
a4acb0eb 1327 return offset;
9a4a100e 1328 break;
a4acb0eb 1329 }
c1bb9350
LT
1330 }
1331
a4acb0eb 1332 return offset;
c1bb9350
LT
1333}
1334
fab2c257 1335static int parse_num(const char *line, unsigned long *p)
46979f56
LT
1336{
1337 char *ptr;
fab2c257
LT
1338
1339 if (!isdigit(*line))
1340 return 0;
1341 *p = strtoul(line, &ptr, 10);
1342 return ptr - line;
1343}
1344
1345static int parse_range(const char *line, int len, int offset, const char *expect,
81bf96bb 1346 unsigned long *p1, unsigned long *p2)
fab2c257 1347{
46979f56
LT
1348 int digits, ex;
1349
1350 if (offset < 0 || offset >= len)
1351 return -1;
1352 line += offset;
1353 len -= offset;
1354
fab2c257
LT
1355 digits = parse_num(line, p1);
1356 if (!digits)
46979f56 1357 return -1;
46979f56
LT
1358
1359 offset += digits;
1360 line += digits;
1361 len -= digits;
1362
c1504628 1363 *p2 = 1;
fab2c257
LT
1364 if (*line == ',') {
1365 digits = parse_num(line+1, p2);
1366 if (!digits)
1367 return -1;
1368
1369 offset += digits+1;
1370 line += digits+1;
1371 len -= digits+1;
1372 }
1373
46979f56
LT
1374 ex = strlen(expect);
1375 if (ex > len)
1376 return -1;
1377 if (memcmp(line, expect, ex))
1378 return -1;
1379
1380 return offset + ex;
1381}
1382
26693ba8 1383static void recount_diff(const char *line, int size, struct fragment *fragment)
c14b9d1e
JS
1384{
1385 int oldlines = 0, newlines = 0, ret = 0;
1386
1387 if (size < 1) {
1388 warning("recount: ignore empty hunk");
1389 return;
1390 }
1391
1392 for (;;) {
1393 int len = linelen(line, size);
1394 size -= len;
1395 line += len;
1396
1397 if (size < 1)
1398 break;
1399
1400 switch (*line) {
1401 case ' ': case '\n':
1402 newlines++;
1403 /* fall through */
1404 case '-':
1405 oldlines++;
1406 continue;
1407 case '+':
1408 newlines++;
1409 continue;
1410 case '\\':
6cf91492 1411 continue;
c14b9d1e 1412 case '@':
59556548 1413 ret = size < 3 || !starts_with(line, "@@ ");
c14b9d1e
JS
1414 break;
1415 case 'd':
59556548 1416 ret = size < 5 || !starts_with(line, "diff ");
c14b9d1e
JS
1417 break;
1418 default:
1419 ret = -1;
1420 break;
1421 }
1422 if (ret) {
3638eb43 1423 warning(_("recount: unexpected line: %.*s"),
c14b9d1e
JS
1424 (int)linelen(line, size), line);
1425 return;
1426 }
1427 break;
1428 }
1429 fragment->oldlines = oldlines;
1430 fragment->newlines = newlines;
1431}
1432
46979f56
LT
1433/*
1434 * Parse a unified diff fragment header of the
1435 * form "@@ -a,b +c,d @@"
1436 */
26693ba8 1437static int parse_fragment_header(const char *line, int len, struct fragment *fragment)
46979f56
LT
1438{
1439 int offset;
1440
1441 if (!len || line[len-1] != '\n')
1442 return -1;
1443
1444 /* Figure out the number of lines in a fragment */
fab2c257
LT
1445 offset = parse_range(line, len, 4, " +", &fragment->oldpos, &fragment->oldlines);
1446 offset = parse_range(line, len, offset, " @@", &fragment->newpos, &fragment->newlines);
46979f56
LT
1447
1448 return offset;
1449}
1450
26693ba8 1451static int find_header(const char *line, unsigned long size, int *hdrsize, struct patch *patch)
c1bb9350
LT
1452{
1453 unsigned long offset, len;
1454
9987d7c5 1455 patch->is_toplevel_relative = 0;
19c58fb8
LT
1456 patch->is_rename = patch->is_copy = 0;
1457 patch->is_new = patch->is_delete = -1;
1458 patch->old_mode = patch->new_mode = 0;
1459 patch->old_name = patch->new_name = NULL;
46979f56 1460 for (offset = 0; size > 0; offset += len, size -= len, line += len, linenr++) {
c1bb9350
LT
1461 unsigned long nextlen;
1462
1463 len = linelen(line, size);
1464 if (!len)
1465 break;
1466
1467 /* Testing this early allows us to take a few shortcuts.. */
1468 if (len < 6)
1469 continue;
46979f56
LT
1470
1471 /*
82e5a82f 1472 * Make sure we don't find any unconnected patch fragments.
46979f56
LT
1473 * That's a sign that we didn't find a header, and that a
1474 * patch has become corrupted/broken up.
1475 */
1476 if (!memcmp("@@ -", line, 4)) {
19c58fb8
LT
1477 struct fragment dummy;
1478 if (parse_fragment_header(line, len, &dummy) < 0)
46979f56 1479 continue;
3638eb43 1480 die(_("patch fragment without header at line %d: %.*s"),
65341411 1481 linenr, (int)len-1, line);
46979f56
LT
1482 }
1483
c1bb9350
LT
1484 if (size < len + 6)
1485 break;
1486
1487 /*
1488 * Git patch? It might not have a real patch, just a rename
1489 * or mode change, so we handle that specially
1490 */
1491 if (!memcmp("diff --git ", line, 11)) {
19c58fb8 1492 int git_hdr_len = parse_git_header(line, len, size, patch);
206de27e 1493 if (git_hdr_len <= len)
c1bb9350 1494 continue;
b7e8039a
LT
1495 if (!patch->old_name && !patch->new_name) {
1496 if (!patch->def_name)
3638eb43
NTND
1497 die(Q_("git diff header lacks filename information when removing "
1498 "%d leading pathname component (line %d)",
1499 "git diff header lacks filename information when removing "
1500 "%d leading pathname components (line %d)",
1501 p_value),
1502 p_value, linenr);
2901bbe5
JH
1503 patch->old_name = xstrdup(patch->def_name);
1504 patch->new_name = xstrdup(patch->def_name);
b7e8039a 1505 }
2c93286a
JM
1506 if (!patch->is_delete && !patch->new_name)
1507 die("git diff header lacks filename information "
1508 "(line %d)", linenr);
9987d7c5 1509 patch->is_toplevel_relative = 1;
a4acb0eb 1510 *hdrsize = git_hdr_len;
c1bb9350
LT
1511 return offset;
1512 }
1513
81bf96bb 1514 /* --- followed by +++ ? */
c1bb9350
LT
1515 if (memcmp("--- ", line, 4) || memcmp("+++ ", line + len, 4))
1516 continue;
1517
1518 /*
1519 * We only accept unified patches, so we want it to
1520 * at least have "@@ -a,b +c,d @@\n", which is 14 chars
81bf96bb 1521 * minimum ("@@ -0,0 +1 @@\n" is the shortest).
c1bb9350
LT
1522 */
1523 nextlen = linelen(line + len, size - len);
1524 if (size < nextlen + 14 || memcmp("@@ -", line + len + nextlen, 4))
1525 continue;
1526
1527 /* Ok, we'll consider it a patch */
19c58fb8 1528 parse_traditional_patch(line, line+len, patch);
c1bb9350 1529 *hdrsize = len + nextlen;
46979f56 1530 linenr += 2;
c1bb9350
LT
1531 return offset;
1532 }
1533 return -1;
1534}
1535
92a1747e 1536static void record_ws_error(unsigned result, const char *line, int len, int linenr)
d0c25035 1537{
c1795bb0 1538 char *err;
92a1747e 1539
c1795bb0
WC
1540 if (!result)
1541 return;
d0c25035 1542
d0c25035
JH
1543 whitespace_error++;
1544 if (squelch_whitespace_errors &&
1545 squelch_whitespace_errors < whitespace_error)
92a1747e
JH
1546 return;
1547
1548 err = whitespace_error_string(result);
1549 fprintf(stderr, "%s:%d: %s.\n%.*s\n",
1550 patch_input_file, linenr, err, len, line);
1551 free(err);
1552}
1553
1554static void check_whitespace(const char *line, int len, unsigned ws_rule)
1555{
1556 unsigned result = ws_check(line + 1, len - 1, ws_rule);
1557
1558 record_ws_error(result, line + 1, len - 2, linenr);
d0c25035
JH
1559}
1560
c1bb9350 1561/*
4be60962
JH
1562 * Parse a unified diff. Note that this really needs to parse each
1563 * fragment separately, since the only way to know the difference
1564 * between a "---" that is part of a patch, and a "---" that starts
1565 * the next patch is to look at the line counts..
c1bb9350 1566 */
26693ba8 1567static int parse_fragment(const char *line, unsigned long size,
81bf96bb 1568 struct patch *patch, struct fragment *fragment)
c1bb9350 1569{
3f40315a 1570 int added, deleted;
c1bb9350 1571 int len = linelen(line, size), offset;
30996652 1572 unsigned long oldlines, newlines;
47495887 1573 unsigned long leading, trailing;
c1bb9350 1574
19c58fb8 1575 offset = parse_fragment_header(line, len, fragment);
c1bb9350
LT
1576 if (offset < 0)
1577 return -1;
c14b9d1e
JS
1578 if (offset > 0 && patch->recount)
1579 recount_diff(line + offset, size - offset, fragment);
19c58fb8
LT
1580 oldlines = fragment->oldlines;
1581 newlines = fragment->newlines;
47495887
EB
1582 leading = 0;
1583 trailing = 0;
c1bb9350
LT
1584
1585 /* Parse the thing.. */
1586 line += len;
1587 size -= len;
46979f56 1588 linenr++;
3f40315a 1589 added = deleted = 0;
4be60962
JH
1590 for (offset = len;
1591 0 < size;
1592 offset += len, size -= len, line += len, linenr++) {
c1bb9350
LT
1593 if (!oldlines && !newlines)
1594 break;
1595 len = linelen(line, size);
1596 if (!len || line[len-1] != '\n')
1597 return -1;
1598 switch (*line) {
1599 default:
1600 return -1;
b507b465 1601 case '\n': /* newer GNU diff, an empty context line */
c1bb9350
LT
1602 case ' ':
1603 oldlines--;
1604 newlines--;
47495887
EB
1605 if (!deleted && !added)
1606 leading++;
1607 trailing++;
c1bb9350
LT
1608 break;
1609 case '-':
5fda48d6 1610 if (apply_in_reverse &&
81bf96bb 1611 ws_error_action != nowarn_ws_error)
cf1b7869 1612 check_whitespace(line, len, patch->ws_rule);
3f40315a 1613 deleted++;
c1bb9350 1614 oldlines--;
47495887 1615 trailing = 0;
c1bb9350
LT
1616 break;
1617 case '+':
5fda48d6 1618 if (!apply_in_reverse &&
81bf96bb 1619 ws_error_action != nowarn_ws_error)
cf1b7869 1620 check_whitespace(line, len, patch->ws_rule);
3f40315a 1621 added++;
c1bb9350 1622 newlines--;
47495887 1623 trailing = 0;
c1bb9350 1624 break;
433ef8a2 1625
81bf96bb
JH
1626 /*
1627 * We allow "\ No newline at end of file". Depending
433ef8a2
FK
1628 * on locale settings when the patch was produced we
1629 * don't know what this line looks like. The only
56d33b11
JH
1630 * thing we do know is that it begins with "\ ".
1631 * Checking for 12 is just for sanity check -- any
1632 * l10n of "\ No newline..." is at least that long.
1633 */
fab2c257 1634 case '\\':
433ef8a2 1635 if (len < 12 || memcmp(line, "\\ ", 2))
3cca928d 1636 return -1;
fab2c257 1637 break;
c1bb9350
LT
1638 }
1639 }
c1504628
LT
1640 if (oldlines || newlines)
1641 return -1;
47495887
EB
1642 fragment->leading = leading;
1643 fragment->trailing = trailing;
1644
81bf96bb
JH
1645 /*
1646 * If a fragment ends with an incomplete line, we failed to include
8b64647d
JH
1647 * it in the above loop because we hit oldlines == newlines == 0
1648 * before seeing it.
1649 */
433ef8a2 1650 if (12 < size && !memcmp(line, "\\ ", 2))
8b64647d
JH
1651 offset += linelen(line, size);
1652
3f40315a
LT
1653 patch->lines_added += added;
1654 patch->lines_deleted += deleted;
4be60962
JH
1655
1656 if (0 < patch->is_new && oldlines)
3638eb43 1657 return error(_("new file depends on old contents"));
4be60962 1658 if (0 < patch->is_delete && newlines)
3638eb43 1659 return error(_("deleted file still has contents"));
c1bb9350
LT
1660 return offset;
1661}
1662
92737a22
JH
1663/*
1664 * We have seen "diff --git a/... b/..." header (or a traditional patch
1665 * header). Read hunks that belong to this patch into fragments and hang
1666 * them to the given patch structure.
1667 *
1668 * The (fragment->patch, fragment->size) pair points into the memory given
1669 * by the caller, not a copy, when we return.
1670 */
26693ba8 1671static int parse_single_patch(const char *line, unsigned long size, struct patch *patch)
c1bb9350
LT
1672{
1673 unsigned long offset = 0;
4be60962 1674 unsigned long oldlines = 0, newlines = 0, context = 0;
19c58fb8 1675 struct fragment **fragp = &patch->fragments;
c1bb9350
LT
1676
1677 while (size > 4 && !memcmp(line, "@@ -", 4)) {
19c58fb8
LT
1678 struct fragment *fragment;
1679 int len;
1680
90321c10 1681 fragment = xcalloc(1, sizeof(*fragment));
77b15bbd 1682 fragment->linenr = linenr;
19c58fb8 1683 len = parse_fragment(line, size, patch, fragment);
c1bb9350 1684 if (len <= 0)
3638eb43 1685 die(_("corrupt patch at line %d"), linenr);
19c58fb8
LT
1686 fragment->patch = line;
1687 fragment->size = len;
4be60962
JH
1688 oldlines += fragment->oldlines;
1689 newlines += fragment->newlines;
1690 context += fragment->leading + fragment->trailing;
19c58fb8
LT
1691
1692 *fragp = fragment;
1693 fragp = &fragment->next;
c1bb9350
LT
1694
1695 offset += len;
1696 line += len;
1697 size -= len;
1698 }
4be60962
JH
1699
1700 /*
1701 * If something was removed (i.e. we have old-lines) it cannot
1702 * be creation, and if something was added it cannot be
1703 * deletion. However, the reverse is not true; --unified=0
1704 * patches that only add are not necessarily creation even
1705 * though they do not have any old lines, and ones that only
1706 * delete are not necessarily deletion.
1707 *
1708 * Unfortunately, a real creation/deletion patch do _not_ have
1709 * any context line by definition, so we cannot safely tell it
1710 * apart with --unified=0 insanity. At least if the patch has
1711 * more than one hunk it is not creation or deletion.
1712 */
1713 if (patch->is_new < 0 &&
1714 (oldlines || (patch->fragments && patch->fragments->next)))
1715 patch->is_new = 0;
1716 if (patch->is_delete < 0 &&
1717 (newlines || (patch->fragments && patch->fragments->next)))
1718 patch->is_delete = 0;
4be60962
JH
1719
1720 if (0 < patch->is_new && oldlines)
3638eb43 1721 die(_("new file %s depends on old contents"), patch->new_name);
4be60962 1722 if (0 < patch->is_delete && newlines)
3638eb43 1723 die(_("deleted file %s still has contents"), patch->old_name);
4be60962 1724 if (!patch->is_delete && !newlines && context)
3638eb43
NTND
1725 fprintf_ln(stderr,
1726 _("** warning: "
1727 "file %s becomes empty but is not deleted"),
1728 patch->new_name);
4be60962 1729
c1bb9350
LT
1730 return offset;
1731}
1732
1fea629f
LT
1733static inline int metadata_changes(struct patch *patch)
1734{
1735 return patch->is_rename > 0 ||
1736 patch->is_copy > 0 ||
1737 patch->is_new > 0 ||
1738 patch->is_delete ||
1739 (patch->old_mode && patch->new_mode &&
1740 patch->old_mode != patch->new_mode);
1741}
1742
3cd4f5e8
JH
1743static char *inflate_it(const void *data, unsigned long size,
1744 unsigned long inflated_size)
051308f6 1745{
ef49a7a0 1746 git_zstream stream;
3cd4f5e8
JH
1747 void *out;
1748 int st;
1749
1750 memset(&stream, 0, sizeof(stream));
1751
1752 stream.next_in = (unsigned char *)data;
1753 stream.avail_in = size;
1754 stream.next_out = out = xmalloc(inflated_size);
1755 stream.avail_out = inflated_size;
39c68542
LT
1756 git_inflate_init(&stream);
1757 st = git_inflate(&stream, Z_FINISH);
1758 git_inflate_end(&stream);
3cd4f5e8
JH
1759 if ((st != Z_STREAM_END) || stream.total_out != inflated_size) {
1760 free(out);
1761 return NULL;
1762 }
1763 return out;
1764}
1765
92737a22
JH
1766/*
1767 * Read a binary hunk and return a new fragment; fragment->patch
1768 * points at an allocated memory that the caller must free, so
1769 * it is marked as "->free_patch = 1".
1770 */
3cd4f5e8
JH
1771static struct fragment *parse_binary_hunk(char **buf_p,
1772 unsigned long *sz_p,
1773 int *status_p,
1774 int *used_p)
1775{
81bf96bb
JH
1776 /*
1777 * Expect a line that begins with binary patch method ("literal"
3cd4f5e8
JH
1778 * or "delta"), followed by the length of data before deflating.
1779 * a sequence of 'length-byte' followed by base-85 encoded data
1780 * should follow, terminated by a newline.
051308f6
JH
1781 *
1782 * Each 5-byte sequence of base-85 encodes up to 4 bytes,
1783 * and we would limit the patch line to 66 characters,
1784 * so one line can fit up to 13 groups that would decode
1785 * to 52 bytes max. The length byte 'A'-'Z' corresponds
1786 * to 1-26 bytes, and 'a'-'z' corresponds to 27-52 bytes.
051308f6
JH
1787 */
1788 int llen, used;
3cd4f5e8
JH
1789 unsigned long size = *sz_p;
1790 char *buffer = *buf_p;
1791 int patch_method;
1792 unsigned long origlen;
0660626c 1793 char *data = NULL;
3cd4f5e8
JH
1794 int hunk_size = 0;
1795 struct fragment *frag;
051308f6 1796
0660626c
JH
1797 llen = linelen(buffer, size);
1798 used = llen;
3cd4f5e8
JH
1799
1800 *status_p = 0;
0660626c 1801
59556548 1802 if (starts_with(buffer, "delta ")) {
3cd4f5e8
JH
1803 patch_method = BINARY_DELTA_DEFLATED;
1804 origlen = strtoul(buffer + 6, NULL, 10);
0660626c 1805 }
59556548 1806 else if (starts_with(buffer, "literal ")) {
3cd4f5e8
JH
1807 patch_method = BINARY_LITERAL_DEFLATED;
1808 origlen = strtoul(buffer + 8, NULL, 10);
0660626c
JH
1809 }
1810 else
3cd4f5e8
JH
1811 return NULL;
1812
1813 linenr++;
0660626c 1814 buffer += llen;
051308f6
JH
1815 while (1) {
1816 int byte_length, max_byte_length, newsize;
1817 llen = linelen(buffer, size);
1818 used += llen;
1819 linenr++;
03eb8f8a
JH
1820 if (llen == 1) {
1821 /* consume the blank line */
1822 buffer++;
1823 size--;
051308f6 1824 break;
03eb8f8a 1825 }
81bf96bb
JH
1826 /*
1827 * Minimum line is "A00000\n" which is 7-byte long,
051308f6
JH
1828 * and the line length must be multiple of 5 plus 2.
1829 */
1830 if ((llen < 7) || (llen-2) % 5)
1831 goto corrupt;
1832 max_byte_length = (llen - 2) / 5 * 4;
1833 byte_length = *buffer;
1834 if ('A' <= byte_length && byte_length <= 'Z')
1835 byte_length = byte_length - 'A' + 1;
1836 else if ('a' <= byte_length && byte_length <= 'z')
1837 byte_length = byte_length - 'a' + 27;
1838 else
1839 goto corrupt;
1840 /* if the input length was not multiple of 4, we would
1841 * have filler at the end but the filler should never
1842 * exceed 3 bytes
1843 */
1844 if (max_byte_length < byte_length ||
1845 byte_length <= max_byte_length - 4)
1846 goto corrupt;
3cd4f5e8 1847 newsize = hunk_size + byte_length;
0660626c 1848 data = xrealloc(data, newsize);
3cd4f5e8 1849 if (decode_85(data + hunk_size, buffer + 1, byte_length))
051308f6 1850 goto corrupt;
3cd4f5e8 1851 hunk_size = newsize;
051308f6
JH
1852 buffer += llen;
1853 size -= llen;
1854 }
3cd4f5e8
JH
1855
1856 frag = xcalloc(1, sizeof(*frag));
1857 frag->patch = inflate_it(data, hunk_size, origlen);
6fe53908 1858 frag->free_patch = 1;
3cd4f5e8
JH
1859 if (!frag->patch)
1860 goto corrupt;
1861 free(data);
1862 frag->size = origlen;
1863 *buf_p = buffer;
1864 *sz_p = size;
1865 *used_p = used;
1866 frag->binary_patch_method = patch_method;
1867 return frag;
1868
051308f6 1869 corrupt:
4cac42b1 1870 free(data);
3cd4f5e8 1871 *status_p = -1;
3638eb43 1872 error(_("corrupt binary patch at line %d: %.*s"),
3cd4f5e8
JH
1873 linenr-1, llen-1, buffer);
1874 return NULL;
1875}
1876
1877static int parse_binary(char *buffer, unsigned long size, struct patch *patch)
1878{
81bf96bb
JH
1879 /*
1880 * We have read "GIT binary patch\n"; what follows is a line
3cd4f5e8
JH
1881 * that says the patch method (currently, either "literal" or
1882 * "delta") and the length of data before deflating; a
1883 * sequence of 'length-byte' followed by base-85 encoded data
1884 * follows.
1885 *
1886 * When a binary patch is reversible, there is another binary
1887 * hunk in the same format, starting with patch method (either
1888 * "literal" or "delta") with the length of data, and a sequence
1889 * of length-byte + base-85 encoded data, terminated with another
1890 * empty line. This data, when applied to the postimage, produces
1891 * the preimage.
1892 */
1893 struct fragment *forward;
1894 struct fragment *reverse;
1895 int status;
1896 int used, used_1;
1897
1898 forward = parse_binary_hunk(&buffer, &size, &status, &used);
1899 if (!forward && !status)
1900 /* there has to be one hunk (forward hunk) */
3638eb43 1901 return error(_("unrecognized binary patch at line %d"), linenr-1);
3cd4f5e8
JH
1902 if (status)
1903 /* otherwise we already gave an error message */
1904 return status;
1905
1906 reverse = parse_binary_hunk(&buffer, &size, &status, &used_1);
1907 if (reverse)
1908 used += used_1;
1909 else if (status) {
81bf96bb
JH
1910 /*
1911 * Not having reverse hunk is not an error, but having
3cd4f5e8
JH
1912 * a corrupt reverse hunk is.
1913 */
1914 free((void*) forward->patch);
1915 free(forward);
1916 return status;
1917 }
1918 forward->next = reverse;
1919 patch->fragments = forward;
1920 patch->is_binary = 1;
1921 return used;
051308f6
JH
1922}
1923
d487b0ba
JH
1924static void prefix_one(char **name)
1925{
1926 char *old_name = *name;
1927 if (!old_name)
1928 return;
1929 *name = xstrdup(prefix_filename(prefix, prefix_length, *name));
1930 free(old_name);
1931}
1932
1933static void prefix_patch(struct patch *p)
1934{
1935 if (!prefix || p->is_toplevel_relative)
1936 return;
1937 prefix_one(&p->new_name);
1938 prefix_one(&p->old_name);
1939}
1940
3ee2ad14
JH
1941/*
1942 * include/exclude
1943 */
1944
1945static struct string_list limit_by_name;
1946static int has_include;
1947static void add_name_limit(const char *name, int exclude)
1948{
1949 struct string_list_item *it;
1950
1951 it = string_list_append(&limit_by_name, name);
1952 it->util = exclude ? NULL : (void *) 1;
1953}
1954
1955static int use_patch(struct patch *p)
1956{
1957 const char *pathname = p->new_name ? p->new_name : p->old_name;
1958 int i;
1959
1960 /* Paths outside are not touched regardless of "--include" */
1961 if (0 < prefix_length) {
1962 int pathlen = strlen(pathname);
1963 if (pathlen <= prefix_length ||
1964 memcmp(prefix, pathname, prefix_length))
1965 return 0;
1966 }
1967
1968 /* See if it matches any of exclude/include rule */
1969 for (i = 0; i < limit_by_name.nr; i++) {
1970 struct string_list_item *it = &limit_by_name.items[i];
ead51a75 1971 if (!wildmatch(it->string, pathname, 0, NULL))
3ee2ad14
JH
1972 return (it->util != NULL);
1973 }
1974
1975 /*
1976 * If we had any include, a path that does not match any rule is
1977 * not used. Otherwise, we saw bunch of exclude rules (or none)
1978 * and such a path is used.
1979 */
1980 return !has_include;
1981}
1982
1983
92737a22 1984/*
41ccfdd9 1985 * Read the patch text in "buffer" that extends for "size" bytes; stop
92737a22
JH
1986 * reading after seeing a single patch (i.e. changes to a single file).
1987 * Create fragments (i.e. patch hunks) and hang them to the given patch.
1988 * Return the number of bytes consumed, so that the caller can call us
1989 * again for the next patch.
1990 */
19c58fb8 1991static int parse_chunk(char *buffer, unsigned long size, struct patch *patch)
c1bb9350
LT
1992{
1993 int hdrsize, patchsize;
19c58fb8 1994 int offset = find_header(buffer, size, &hdrsize, patch);
c1bb9350
LT
1995
1996 if (offset < 0)
1997 return offset;
c1bb9350 1998
d487b0ba
JH
1999 prefix_patch(patch);
2000
477a08af
JH
2001 if (!use_patch(patch))
2002 patch->ws_rule = 0;
2003 else
2004 patch->ws_rule = whitespace_rule(patch->new_name
2005 ? patch->new_name
2006 : patch->old_name);
cf1b7869 2007
81bf96bb
JH
2008 patchsize = parse_single_patch(buffer + offset + hdrsize,
2009 size - offset - hdrsize, patch);
c1bb9350 2010
92927ed0 2011 if (!patchsize) {
051308f6 2012 static const char git_binary[] = "GIT binary patch\n";
3200d1ae
JH
2013 int hd = hdrsize + offset;
2014 unsigned long llen = linelen(buffer + hd, size - hd);
2015
051308f6
JH
2016 if (llen == sizeof(git_binary) - 1 &&
2017 !memcmp(git_binary, buffer + hd, llen)) {
2018 int used;
2019 linenr++;
2020 used = parse_binary(buffer + hd + llen,
2021 size - hd - llen, patch);
2022 if (used)
2023 patchsize = used + llen;
2024 else
2025 patchsize = 0;
2026 }
2027 else if (!memcmp(" differ\n", buffer + hd + llen - 8, 8)) {
e36f3a8a
EP
2028 static const char *binhdr[] = {
2029 "Binary files ",
2030 "Files ",
2031 NULL,
2032 };
2033 int i;
3200d1ae
JH
2034 for (i = 0; binhdr[i]; i++) {
2035 int len = strlen(binhdr[i]);
2036 if (len < size - hd &&
2037 !memcmp(binhdr[i], buffer + hd, len)) {
051308f6 2038 linenr++;
3200d1ae 2039 patch->is_binary = 1;
051308f6 2040 patchsize = llen;
3200d1ae
JH
2041 break;
2042 }
2043 }
051308f6 2044 }
ff36de08 2045
2b6eef94
JH
2046 /* Empty patch cannot be applied if it is a text patch
2047 * without metadata change. A binary patch appears
2048 * empty to us here.
92927ed0
JH
2049 */
2050 if ((apply || check) &&
2b6eef94 2051 (!patch->is_binary && !metadata_changes(patch)))
3638eb43 2052 die(_("patch with only garbage at line %d"), linenr);
ff36de08 2053 }
1fea629f 2054
c1bb9350
LT
2055 return offset + hdrsize + patchsize;
2056}
2057
e5a94313
JS
2058#define swap(a,b) myswap((a),(b),sizeof(a))
2059
2060#define myswap(a, b, size) do { \
2061 unsigned char mytmp[size]; \
2062 memcpy(mytmp, &a, size); \
2063 memcpy(&a, &b, size); \
2064 memcpy(&b, mytmp, size); \
2065} while (0)
2066
2067static void reverse_patches(struct patch *p)
2068{
2069 for (; p; p = p->next) {
2070 struct fragment *frag = p->fragments;
2071
2072 swap(p->new_name, p->old_name);
2073 swap(p->new_mode, p->old_mode);
2074 swap(p->is_new, p->is_delete);
2075 swap(p->lines_added, p->lines_deleted);
2076 swap(p->old_sha1_prefix, p->new_sha1_prefix);
2077
2078 for (; frag; frag = frag->next) {
2079 swap(frag->newpos, frag->oldpos);
2080 swap(frag->newlines, frag->oldlines);
2081 }
e5a94313
JS
2082 }
2083}
2084
81bf96bb
JH
2085static const char pluses[] =
2086"++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++";
2087static const char minuses[]=
2088"----------------------------------------------------------------------";
3f40315a
LT
2089
2090static void show_stats(struct patch *patch)
2091{
f285a2d7 2092 struct strbuf qname = STRBUF_INIT;
663af342
PH
2093 char *cp = patch->new_name ? patch->new_name : patch->old_name;
2094 int max, add, del;
3f40315a 2095
663af342 2096 quote_c_style(cp, &qname, NULL, 0);
22943f1a 2097
3f40315a
LT
2098 /*
2099 * "scale" the filename
2100 */
3f40315a
LT
2101 max = max_len;
2102 if (max > 50)
2103 max = 50;
663af342
PH
2104
2105 if (qname.len > max) {
2106 cp = strchr(qname.buf + qname.len + 3 - max, '/');
2107 if (!cp)
2108 cp = qname.buf + qname.len + 3 - max;
2109 strbuf_splice(&qname, 0, cp - qname.buf, "...", 3);
2110 }
2111
2112 if (patch->is_binary) {
2113 printf(" %-*s | Bin\n", max, qname.buf);
2114 strbuf_release(&qname);
2115 return;
62917097 2116 }
663af342
PH
2117
2118 printf(" %-*s |", max, qname.buf);
2119 strbuf_release(&qname);
3f40315a
LT
2120
2121 /*
2122 * scale the add/delete
2123 */
663af342 2124 max = max + max_change > 70 ? 70 - max : max_change;
95bedc9e
LT
2125 add = patch->lines_added;
2126 del = patch->lines_deleted;
95bedc9e 2127
69f956e1 2128 if (max_change > 0) {
663af342 2129 int total = ((add + del) * max + max_change / 2) / max_change;
69f956e1
SV
2130 add = (add * max + max_change / 2) / max_change;
2131 del = total - add;
2132 }
663af342
PH
2133 printf("%5d %.*s%.*s\n", patch->lines_added + patch->lines_deleted,
2134 add, pluses, del, minuses);
3f40315a
LT
2135}
2136
c7f9cb14 2137static int read_old_data(struct stat *st, const char *path, struct strbuf *buf)
3cca928d 2138{
3cca928d
LT
2139 switch (st->st_mode & S_IFMT) {
2140 case S_IFLNK:
b11b7e13 2141 if (strbuf_readlink(buf, path, st->st_size) < 0)
3638eb43 2142 return error(_("unable to read symlink %s"), path);
c7f9cb14 2143 return 0;
3cca928d 2144 case S_IFREG:
387e7e19 2145 if (strbuf_read_file(buf, path, st->st_size) != st->st_size)
3638eb43 2146 return error(_("unable to open or read %s"), path);
21e5ad50 2147 convert_to_git(path, buf->buf, buf->len, buf, 0);
c7f9cb14 2148 return 0;
3cca928d
LT
2149 default:
2150 return -1;
2151 }
2152}
2153
86c91f91
GB
2154/*
2155 * Update the preimage, and the common lines in postimage,
2156 * from buffer buf of length len. If postlen is 0 the postimage
2157 * is updated in place, otherwise it's updated on a new buffer
2158 * of length postlen
2159 */
2160
c1beba5b
JH
2161static void update_pre_post_images(struct image *preimage,
2162 struct image *postimage,
2163 char *buf,
86c91f91 2164 size_t len, size_t postlen)
3cca928d 2165{
5de7166d 2166 int i, ctx, reduced;
c1beba5b
JH
2167 char *new, *old, *fixed;
2168 struct image fixed_preimage;
3cca928d 2169
c1beba5b
JH
2170 /*
2171 * Update the preimage with whitespace fixes. Note that we
2172 * are not losing preimage->buf -- apply_one_fragment() will
2173 * free "oldlines".
2174 */
2175 prepare_image(&fixed_preimage, buf, len, 1);
5de7166d
JH
2176 assert(postlen
2177 ? fixed_preimage.nr == preimage->nr
2178 : fixed_preimage.nr <= preimage->nr);
2179 for (i = 0; i < fixed_preimage.nr; i++)
c1beba5b
JH
2180 fixed_preimage.line[i].flag = preimage->line[i].flag;
2181 free(preimage->line_allocated);
2182 *preimage = fixed_preimage;
3cca928d 2183
c1beba5b 2184 /*
86c91f91 2185 * Adjust the common context lines in postimage. This can be
250b3c6c
JH
2186 * done in-place when we are shrinking it with whitespace
2187 * fixing, but needs a new buffer when ignoring whitespace or
2188 * expanding leading tabs to spaces.
2189 *
86c91f91
GB
2190 * We trust the caller to tell us if the update can be done
2191 * in place (postlen==0) or not.
c1beba5b 2192 */
86c91f91
GB
2193 old = postimage->buf;
2194 if (postlen)
2195 new = postimage->buf = xmalloc(postlen);
2196 else
2197 new = old;
c1beba5b 2198 fixed = preimage->buf;
5de7166d
JH
2199
2200 for (i = reduced = ctx = 0; i < postimage->nr; i++) {
c1beba5b
JH
2201 size_t len = postimage->line[i].len;
2202 if (!(postimage->line[i].flag & LINE_COMMON)) {
2203 /* an added line -- no counterparts in preimage */
2204 memmove(new, old, len);
2205 old += len;
2206 new += len;
2207 continue;
3cca928d 2208 }
c1beba5b
JH
2209
2210 /* a common context -- skip it in the original postimage */
2211 old += len;
2212
2213 /* and find the corresponding one in the fixed preimage */
2214 while (ctx < preimage->nr &&
2215 !(preimage->line[ctx].flag & LINE_COMMON)) {
2216 fixed += preimage->line[ctx].len;
2217 ctx++;
2218 }
5de7166d
JH
2219
2220 /*
2221 * preimage is expected to run out, if the caller
2222 * fixed addition of trailing blank lines.
2223 */
2224 if (preimage->nr <= ctx) {
2225 reduced++;
2226 continue;
2227 }
c1beba5b
JH
2228
2229 /* and copy it in, while fixing the line length */
2230 len = preimage->line[ctx].len;
2231 memcpy(new, fixed, len);
2232 new += len;
2233 fixed += len;
2234 postimage->line[i].len = len;
2235 ctx++;
2236 }
2237
2238 /* Fix the length of the whole thing */
2239 postimage->len = new - postimage->buf;
5de7166d 2240 postimage->nr -= reduced;
c1beba5b
JH
2241}
2242
b94f2eda
JH
2243static int match_fragment(struct image *img,
2244 struct image *preimage,
2245 struct image *postimage,
c89fb6b1 2246 unsigned long try,
b94f2eda 2247 int try_lno,
c607aaa2 2248 unsigned ws_rule,
dc41976a 2249 int match_beginning, int match_end)
c89fb6b1 2250{
b94f2eda 2251 int i;
c1beba5b 2252 char *fixed_buf, *buf, *orig, *target;
d511bd33 2253 struct strbuf fixed;
250b3c6c 2254 size_t fixed_len, postlen;
51667147 2255 int preimage_limit;
b94f2eda 2256
51667147
BG
2257 if (preimage->nr + try_lno <= img->nr) {
2258 /*
2259 * The hunk falls within the boundaries of img.
2260 */
2261 preimage_limit = preimage->nr;
2262 if (match_end && (preimage->nr + try_lno != img->nr))
2263 return 0;
2264 } else if (ws_error_action == correct_ws_error &&
0c3ef984 2265 (ws_rule & WS_BLANK_AT_EOF)) {
51667147 2266 /*
0c3ef984
BG
2267 * This hunk extends beyond the end of img, and we are
2268 * removing blank lines at the end of the file. This
2269 * many lines from the beginning of the preimage must
2270 * match with img, and the remainder of the preimage
2271 * must be blank.
51667147
BG
2272 */
2273 preimage_limit = img->nr - try_lno;
2274 } else {
2275 /*
2276 * The hunk extends beyond the end of the img and
2277 * we are not removing blanks at the end, so we
2278 * should reject the hunk at this position.
2279 */
b94f2eda 2280 return 0;
51667147 2281 }
b94f2eda
JH
2282
2283 if (match_beginning && try_lno)
c89fb6b1 2284 return 0;
dc41976a 2285
b94f2eda 2286 /* Quick hash check */
51667147 2287 for (i = 0; i < preimage_limit; i++)
9d158601
JH
2288 if ((img->line[try_lno + i].flag & LINE_PATCHED) ||
2289 (preimage->line[i].hash != img->line[try_lno + i].hash))
b94f2eda
JH
2290 return 0;
2291
51667147
BG
2292 if (preimage_limit == preimage->nr) {
2293 /*
2294 * Do we have an exact match? If we were told to match
2295 * at the end, size must be exactly at try+fragsize,
2296 * otherwise try+fragsize must be still within the preimage,
2297 * and either case, the old piece should match the preimage
2298 * exactly.
2299 */
2300 if ((match_end
2301 ? (try + preimage->len == img->len)
2302 : (try + preimage->len <= img->len)) &&
2303 !memcmp(img->buf + try, preimage->buf, preimage->len))
2304 return 1;
2305 } else {
2306 /*
2307 * The preimage extends beyond the end of img, so
2308 * there cannot be an exact match.
2309 *
2310 * There must be one non-blank context line that match
2311 * a line before the end of img.
2312 */
2313 char *buf_end;
2314
2315 buf = preimage->buf;
2316 buf_end = buf;
2317 for (i = 0; i < preimage_limit; i++)
2318 buf_end += preimage->line[i].len;
2319
2320 for ( ; buf < buf_end; buf++)
2321 if (!isspace(*buf))
2322 break;
2323 if (buf == buf_end)
2324 return 0;
2325 }
dc41976a 2326
86c91f91
GB
2327 /*
2328 * No exact match. If we are ignoring whitespace, run a line-by-line
2329 * fuzzy matching. We collect all the line length information because
2330 * we need it to adjust whitespace if we match.
2331 */
2332 if (ws_ignore_action == ignore_ws_change) {
2333 size_t imgoff = 0;
2334 size_t preoff = 0;
2335 size_t postlen = postimage->len;
51667147
BG
2336 size_t extra_chars;
2337 char *preimage_eof;
2338 char *preimage_end;
2339 for (i = 0; i < preimage_limit; i++) {
86c91f91 2340 size_t prelen = preimage->line[i].len;
0b1fac32 2341 size_t imglen = img->line[try_lno+i].len;
86c91f91 2342
0b1fac32
JH
2343 if (!fuzzy_matchlines(img->buf + try + imgoff, imglen,
2344 preimage->buf + preoff, prelen))
86c91f91
GB
2345 return 0;
2346 if (preimage->line[i].flag & LINE_COMMON)
0b1fac32
JH
2347 postlen += imglen - prelen;
2348 imgoff += imglen;
86c91f91
GB
2349 preoff += prelen;
2350 }
2351
2352 /*
9b25949a
BG
2353 * Ok, the preimage matches with whitespace fuzz.
2354 *
2355 * imgoff now holds the true length of the target that
51667147
BG
2356 * matches the preimage before the end of the file.
2357 *
2358 * Count the number of characters in the preimage that fall
2359 * beyond the end of the file and make sure that all of them
2360 * are whitespace characters. (This can only happen if
2361 * we are removing blank lines at the end of the file.)
86c91f91 2362 */
51667147
BG
2363 buf = preimage_eof = preimage->buf + preoff;
2364 for ( ; i < preimage->nr; i++)
2365 preoff += preimage->line[i].len;
2366 preimage_end = preimage->buf + preoff;
2367 for ( ; buf < preimage_end; buf++)
2368 if (!isspace(*buf))
2369 return 0;
2370
2371 /*
2372 * Update the preimage and the common postimage context
2373 * lines to use the same whitespace as the target.
2374 * If whitespace is missing in the target (i.e.
2375 * if the preimage extends beyond the end of the file),
2376 * use the whitespace from the preimage.
2377 */
2378 extra_chars = preimage_end - preimage_eof;
d511bd33
CW
2379 strbuf_init(&fixed, imgoff + extra_chars);
2380 strbuf_add(&fixed, img->buf + try, imgoff);
2381 strbuf_add(&fixed, preimage_eof, extra_chars);
2382 fixed_buf = strbuf_detach(&fixed, &fixed_len);
86c91f91 2383 update_pre_post_images(preimage, postimage,
d511bd33 2384 fixed_buf, fixed_len, postlen);
86c91f91
GB
2385 return 1;
2386 }
2387
c1beba5b
JH
2388 if (ws_error_action != correct_ws_error)
2389 return 0;
2390
dc41976a 2391 /*
c1beba5b 2392 * The hunk does not apply byte-by-byte, but the hash says
86c91f91
GB
2393 * it might with whitespace fuzz. We haven't been asked to
2394 * ignore whitespace, we were asked to correct whitespace
2395 * errors, so let's try matching after whitespace correction.
51667147
BG
2396 *
2397 * The preimage may extend beyond the end of the file,
2398 * but in this loop we will only handle the part of the
2399 * preimage that falls within the file.
dc41976a 2400 */
d511bd33 2401 strbuf_init(&fixed, preimage->len + 1);
c1beba5b
JH
2402 orig = preimage->buf;
2403 target = img->buf + try;
250b3c6c 2404 postlen = 0;
51667147 2405 for (i = 0; i < preimage_limit; i++) {
c1beba5b
JH
2406 size_t oldlen = preimage->line[i].len;
2407 size_t tgtlen = img->line[try_lno + i].len;
d511bd33
CW
2408 size_t fixstart = fixed.len;
2409 struct strbuf tgtfix;
c1beba5b
JH
2410 int match;
2411
2412 /* Try fixing the line in the preimage */
d511bd33 2413 ws_fix_copy(&fixed, orig, oldlen, ws_rule, NULL);
c1beba5b
JH
2414
2415 /* Try fixing the line in the target */
d511bd33
CW
2416 strbuf_init(&tgtfix, tgtlen);
2417 ws_fix_copy(&tgtfix, target, tgtlen, ws_rule, NULL);
c1beba5b
JH
2418
2419 /*
2420 * If they match, either the preimage was based on
2421 * a version before our tree fixed whitespace breakage,
2422 * or we are lacking a whitespace-fix patch the tree
2423 * the preimage was based on already had (i.e. target
2424 * has whitespace breakage, the preimage doesn't).
2425 * In either case, we are fixing the whitespace breakages
2426 * so we might as well take the fix together with their
2427 * real change.
2428 */
d511bd33
CW
2429 match = (tgtfix.len == fixed.len - fixstart &&
2430 !memcmp(tgtfix.buf, fixed.buf + fixstart,
2431 fixed.len - fixstart));
250b3c6c 2432 postlen += tgtfix.len;
c1beba5b 2433
d511bd33 2434 strbuf_release(&tgtfix);
c1beba5b
JH
2435 if (!match)
2436 goto unmatch_exit;
2437
2438 orig += oldlen;
c1beba5b 2439 target += tgtlen;
3cca928d
LT
2440 }
2441
51667147
BG
2442
2443 /*
2444 * Now handle the lines in the preimage that falls beyond the
2445 * end of the file (if any). They will only match if they are
2446 * empty or only contain whitespace (if WS_BLANK_AT_EOL is
2447 * false).
2448 */
2449 for ( ; i < preimage->nr; i++) {
d511bd33 2450 size_t fixstart = fixed.len; /* start of the fixed preimage */
51667147
BG
2451 size_t oldlen = preimage->line[i].len;
2452 int j;
2453
2454 /* Try fixing the line in the preimage */
d511bd33 2455 ws_fix_copy(&fixed, orig, oldlen, ws_rule, NULL);
51667147 2456
d511bd33
CW
2457 for (j = fixstart; j < fixed.len; j++)
2458 if (!isspace(fixed.buf[j]))
51667147
BG
2459 goto unmatch_exit;
2460
2461 orig += oldlen;
51667147
BG
2462 }
2463
c1beba5b
JH
2464 /*
2465 * Yes, the preimage is based on an older version that still
2466 * has whitespace breakages unfixed, and fixing them makes the
2467 * hunk match. Update the context lines in the postimage.
2468 */
d511bd33 2469 fixed_buf = strbuf_detach(&fixed, &fixed_len);
250b3c6c
JH
2470 if (postlen < postimage->len)
2471 postlen = 0;
c1beba5b 2472 update_pre_post_images(preimage, postimage,
250b3c6c 2473 fixed_buf, fixed_len, postlen);
c1beba5b
JH
2474 return 1;
2475
2476 unmatch_exit:
d511bd33 2477 strbuf_release(&fixed);
dc41976a 2478 return 0;
c89fb6b1
JH
2479}
2480
b94f2eda
JH
2481static int find_pos(struct image *img,
2482 struct image *preimage,
2483 struct image *postimage,
2484 int line,
c607aaa2 2485 unsigned ws_rule,
b94f2eda 2486 int match_beginning, int match_end)
3cca928d 2487{
b94f2eda
JH
2488 int i;
2489 unsigned long backwards, forwards, try;
2490 int backwards_lno, forwards_lno, try_lno;
3cca928d 2491
ecf4c2ec 2492 /*
24ff4d56 2493 * If match_beginning or match_end is specified, there is no
ecf4c2ec
JH
2494 * point starting from a wrong line that will never match and
2495 * wander around and wait for a match at the specified end.
2496 */
2497 if (match_beginning)
2498 line = 0;
2499 else if (match_end)
2500 line = img->nr - preimage->nr;
2501
24ff4d56
BG
2502 /*
2503 * Because the comparison is unsigned, the following test
2504 * will also take care of a negative line number that can
2505 * result when match_end and preimage is larger than the target.
2506 */
2507 if ((size_t) line > img->nr)
52f3c81a
JH
2508 line = img->nr;
2509
b94f2eda
JH
2510 try = 0;
2511 for (i = 0; i < line; i++)
2512 try += img->line[i].len;
3cca928d 2513
6e7c92a9
LT
2514 /*
2515 * There's probably some smart way to do this, but I'll leave
2516 * that to the smart and beautiful people. I'm simple and stupid.
2517 */
b94f2eda
JH
2518 backwards = try;
2519 backwards_lno = line;
2520 forwards = try;
2521 forwards_lno = line;
2522 try_lno = line;
fcb77bc5 2523
6e7c92a9 2524 for (i = 0; ; i++) {
b94f2eda 2525 if (match_fragment(img, preimage, postimage,
c607aaa2 2526 try, try_lno, ws_rule,
b94f2eda
JH
2527 match_beginning, match_end))
2528 return try_lno;
fcb77bc5
JH
2529
2530 again:
b94f2eda 2531 if (backwards_lno == 0 && forwards_lno == img->nr)
fcb77bc5 2532 break;
6e7c92a9 2533
6e7c92a9 2534 if (i & 1) {
b94f2eda 2535 if (backwards_lno == 0) {
fcb77bc5
JH
2536 i++;
2537 goto again;
6e7c92a9 2538 }
b94f2eda
JH
2539 backwards_lno--;
2540 backwards -= img->line[backwards_lno].len;
6e7c92a9 2541 try = backwards;
b94f2eda 2542 try_lno = backwards_lno;
6e7c92a9 2543 } else {
b94f2eda 2544 if (forwards_lno == img->nr) {
fcb77bc5
JH
2545 i++;
2546 goto again;
6e7c92a9 2547 }
b94f2eda
JH
2548 forwards += img->line[forwards_lno].len;
2549 forwards_lno++;
6e7c92a9 2550 try = forwards;
b94f2eda 2551 try_lno = forwards_lno;
6e7c92a9
LT
2552 }
2553
6e7c92a9 2554 }
3cca928d
LT
2555 return -1;
2556}
2557
b94f2eda 2558static void remove_first_line(struct image *img)
47495887 2559{
b94f2eda
JH
2560 img->buf += img->line[0].len;
2561 img->len -= img->line[0].len;
2562 img->line++;
2563 img->nr--;
47495887
EB
2564}
2565
b94f2eda 2566static void remove_last_line(struct image *img)
47495887 2567{
b94f2eda 2568 img->len -= img->line[--img->nr].len;
47495887
EB
2569}
2570
92737a22
JH
2571/*
2572 * The change from "preimage" and "postimage" has been found to
2573 * apply at applied_pos (counts in line numbers) in "img".
2574 * Update "img" to remove "preimage" and replace it with "postimage".
2575 */
b94f2eda
JH
2576static void update_image(struct image *img,
2577 int applied_pos,
2578 struct image *preimage,
2579 struct image *postimage)
b5767dd6 2580{
81bf96bb 2581 /*
b94f2eda
JH
2582 * remove the copy of preimage at offset in img
2583 * and replace it with postimage
81bf96bb 2584 */
b94f2eda
JH
2585 int i, nr;
2586 size_t remove_count, insert_count, applied_at = 0;
2587 char *result;
51667147
BG
2588 int preimage_limit;
2589
2590 /*
2591 * If we are removing blank lines at the end of img,
2592 * the preimage may extend beyond the end.
2593 * If that is the case, we must be careful only to
2594 * remove the part of the preimage that falls within
2595 * the boundaries of img. Initialize preimage_limit
2596 * to the number of lines in the preimage that falls
2597 * within the boundaries.
2598 */
2599 preimage_limit = preimage->nr;
2600 if (preimage_limit > img->nr - applied_pos)
2601 preimage_limit = img->nr - applied_pos;
d5a41641 2602
b94f2eda
JH
2603 for (i = 0; i < applied_pos; i++)
2604 applied_at += img->line[i].len;
2605
2606 remove_count = 0;
51667147 2607 for (i = 0; i < preimage_limit; i++)
b94f2eda
JH
2608 remove_count += img->line[applied_pos + i].len;
2609 insert_count = postimage->len;
2610
2611 /* Adjust the contents */
2612 result = xmalloc(img->len + insert_count - remove_count + 1);
2613 memcpy(result, img->buf, applied_at);
2614 memcpy(result + applied_at, postimage->buf, postimage->len);
2615 memcpy(result + applied_at + postimage->len,
2616 img->buf + (applied_at + remove_count),
2617 img->len - (applied_at + remove_count));
2618 free(img->buf);
2619 img->buf = result;
2620 img->len += insert_count - remove_count;
2621 result[img->len] = '\0';
2622
2623 /* Adjust the line table */
51667147
BG
2624 nr = img->nr + postimage->nr - preimage_limit;
2625 if (preimage_limit < postimage->nr) {
81bf96bb 2626 /*
b94f2eda
JH
2627 * NOTE: this knows that we never call remove_first_line()
2628 * on anything other than pre/post image.
d0c25035 2629 */
2756ca43 2630 REALLOC_ARRAY(img->line, nr);
b94f2eda 2631 img->line_allocated = img->line;
d0c25035 2632 }
51667147 2633 if (preimage_limit != postimage->nr)
b94f2eda 2634 memmove(img->line + applied_pos + postimage->nr,
51667147
BG
2635 img->line + applied_pos + preimage_limit,
2636 (img->nr - (applied_pos + preimage_limit)) *
b94f2eda
JH
2637 sizeof(*img->line));
2638 memcpy(img->line + applied_pos,
2639 postimage->line,
2640 postimage->nr * sizeof(*img->line));
933e44d3
JH
2641 if (!allow_overlap)
2642 for (i = 0; i < postimage->nr; i++)
2643 img->line[applied_pos + i].flag |= LINE_PATCHED;
b94f2eda 2644 img->nr = nr;
b5767dd6
JH
2645}
2646
92737a22
JH
2647/*
2648 * Use the patch-hunk text in "frag" to prepare two images (preimage and
2649 * postimage) for the hunk. Find lines that match "preimage" in "img" and
2650 * replace the part of "img" with "postimage" text.
2651 */
b94f2eda 2652static int apply_one_fragment(struct image *img, struct fragment *frag,
334f8cb2
JH
2653 int inaccurate_eof, unsigned ws_rule,
2654 int nth_fragment)
3cca928d 2655{
65aadb92 2656 int match_beginning, match_end;
3cca928d 2657 const char *patch = frag->patch;
b94f2eda 2658 int size = frag->size;
d511bd33
CW
2659 char *old, *oldlines;
2660 struct strbuf newlines;
077e1af5 2661 int new_blank_lines_at_end = 0;
85572639
JH
2662 int found_new_blank_lines_at_end = 0;
2663 int hunk_linenr = frag->linenr;
47495887 2664 unsigned long leading, trailing;
b94f2eda
JH
2665 int pos, applied_pos;
2666 struct image preimage;
2667 struct image postimage;
3cca928d 2668
c330fdd4
JH
2669 memset(&preimage, 0, sizeof(preimage));
2670 memset(&postimage, 0, sizeof(postimage));
61e08cca 2671 oldlines = xmalloc(size);
d511bd33 2672 strbuf_init(&newlines, size);
c330fdd4 2673
61e08cca 2674 old = oldlines;
3cca928d 2675 while (size > 0) {
e5a94313 2676 char first;
3cca928d 2677 int len = linelen(patch, size);
d511bd33 2678 int plen;
077e1af5 2679 int added_blank_line = 0;
efa57443 2680 int is_blank_context = 0;
d511bd33 2681 size_t start;
3cca928d
LT
2682
2683 if (!len)
2684 break;
2685
2686 /*
2687 * "plen" is how much of the line we should use for
2688 * the actual patch data. Normally we just remove the
2689 * first character on the line, but if the line is
2690 * followed by "\ No newline", then we also remove the
2691 * last one (which is the newline, of course).
2692 */
61e08cca 2693 plen = len - 1;
8b64647d 2694 if (len < size && patch[len] == '\\')
3cca928d 2695 plen--;
e5a94313 2696 first = *patch;
f686d030 2697 if (apply_in_reverse) {
e5a94313
JS
2698 if (first == '-')
2699 first = '+';
2700 else if (first == '+')
2701 first = '-';
2702 }
efe7f358 2703
e5a94313 2704 switch (first) {
b507b465
LT
2705 case '\n':
2706 /* Newer GNU diff, empty context line */
2707 if (plen < 0)
2708 /* ... followed by '\No newline'; nothing */
2709 break;
61e08cca 2710 *old++ = '\n';
d511bd33 2711 strbuf_addch(&newlines, '\n');
c330fdd4
JH
2712 add_line_info(&preimage, "\n", 1, LINE_COMMON);
2713 add_line_info(&postimage, "\n", 1, LINE_COMMON);
efa57443 2714 is_blank_context = 1;
b507b465 2715 break;
3cca928d 2716 case ' ':
94ea026b
JH
2717 if (plen && (ws_rule & WS_BLANK_AT_EOF) &&
2718 ws_blank_line(patch + 1, plen, ws_rule))
efa57443 2719 is_blank_context = 1;
3cca928d 2720 case '-':
61e08cca
JH
2721 memcpy(old, patch + 1, plen);
2722 add_line_info(&preimage, old, plen,
c330fdd4 2723 (first == ' ' ? LINE_COMMON : 0));
61e08cca 2724 old += plen;
e5a94313 2725 if (first == '-')
3cca928d
LT
2726 break;
2727 /* Fall-through for ' ' */
2728 case '+':
8441a9a8
JH
2729 /* --no-add does not add new lines */
2730 if (first == '+' && no_add)
2731 break;
2732
d511bd33 2733 start = newlines.len;
8441a9a8
JH
2734 if (first != '+' ||
2735 !whitespace_error ||
2736 ws_error_action != correct_ws_error) {
d511bd33 2737 strbuf_add(&newlines, patch + 1, plen);
8441a9a8
JH
2738 }
2739 else {
d511bd33 2740 ws_fix_copy(&newlines, patch + 1, plen, ws_rule, &applied_after_fixing_ws);
077e1af5 2741 }
d511bd33 2742 add_line_info(&postimage, newlines.buf + start, newlines.len - start,
8441a9a8 2743 (first == '+' ? 0 : LINE_COMMON));
8441a9a8 2744 if (first == '+' &&
94ea026b
JH
2745 (ws_rule & WS_BLANK_AT_EOF) &&
2746 ws_blank_line(patch + 1, plen, ws_rule))
8441a9a8 2747 added_blank_line = 1;
3cca928d
LT
2748 break;
2749 case '@': case '\\':
2750 /* Ignore it, we already handled it */
2751 break;
2752 default:
aeabfa07 2753 if (apply_verbosely)
3638eb43 2754 error(_("invalid start of line: '%c'"), first);
3cca928d
LT
2755 return -1;
2756 }
85572639
JH
2757 if (added_blank_line) {
2758 if (!new_blank_lines_at_end)
2759 found_new_blank_lines_at_end = hunk_linenr;
077e1af5 2760 new_blank_lines_at_end++;
85572639 2761 }
efa57443
JH
2762 else if (is_blank_context)
2763 ;
077e1af5
JH
2764 else
2765 new_blank_lines_at_end = 0;
3cca928d
LT
2766 patch += len;
2767 size -= len;
85572639 2768 hunk_linenr++;
3cca928d 2769 }
81bf96bb 2770 if (inaccurate_eof &&
61e08cca 2771 old > oldlines && old[-1] == '\n' &&
d511bd33 2772 newlines.len > 0 && newlines.buf[newlines.len - 1] == '\n') {
61e08cca 2773 old--;
d511bd33 2774 strbuf_setlen(&newlines, newlines.len - 1);
5b5d4d9e 2775 }
47495887 2776
47495887
EB
2777 leading = frag->leading;
2778 trailing = frag->trailing;
1bf1a859
LT
2779
2780 /*
ee5a317e
JH
2781 * A hunk to change lines at the beginning would begin with
2782 * @@ -1,L +N,M @@
ed0f47a8
JH
2783 * but we need to be careful. -U0 that inserts before the second
2784 * line also has this pattern.
4be60962 2785 *
ee5a317e
JH
2786 * And a hunk to add to an empty file would begin with
2787 * @@ -0,0 +N,M @@
2788 *
2789 * In other words, a hunk that is (frag->oldpos <= 1) with or
2790 * without leading context must match at the beginning.
1bf1a859 2791 */
ed0f47a8
JH
2792 match_beginning = (!frag->oldpos ||
2793 (frag->oldpos == 1 && !unidiff_zero));
ee5a317e
JH
2794
2795 /*
2796 * A hunk without trailing lines must match at the end.
2797 * However, we simply cannot tell if a hunk must match end
2798 * from the lack of trailing lines if the patch was generated
2799 * with unidiff without any context.
2800 */
2801 match_end = !unidiff_zero && !trailing;
1bf1a859 2802
b94f2eda 2803 pos = frag->newpos ? (frag->newpos - 1) : 0;
61e08cca
JH
2804 preimage.buf = oldlines;
2805 preimage.len = old - oldlines;
d511bd33
CW
2806 postimage.buf = newlines.buf;
2807 postimage.len = newlines.len;
c330fdd4
JH
2808 preimage.line = preimage.line_allocated;
2809 postimage.line = postimage.line_allocated;
2810
47495887 2811 for (;;) {
efe7f358 2812
c607aaa2
JH
2813 applied_pos = find_pos(img, &preimage, &postimage, pos,
2814 ws_rule, match_beginning, match_end);
b94f2eda
JH
2815
2816 if (applied_pos >= 0)
47495887 2817 break;
47495887
EB
2818
2819 /* Am I at my context limits? */
2820 if ((leading <= p_context) && (trailing <= p_context))
2821 break;
65aadb92
JH
2822 if (match_beginning || match_end) {
2823 match_beginning = match_end = 0;
1bf1a859
LT
2824 continue;
2825 }
b94f2eda 2826
81bf96bb
JH
2827 /*
2828 * Reduce the number of context lines; reduce both
2829 * leading and trailing if they are equal otherwise
2830 * just reduce the larger context.
47495887
EB
2831 */
2832 if (leading >= trailing) {
b94f2eda
JH
2833 remove_first_line(&preimage);
2834 remove_first_line(&postimage);
47495887
EB
2835 pos--;
2836 leading--;
2837 }
2838 if (trailing > leading) {
b94f2eda
JH
2839 remove_last_line(&preimage);
2840 remove_last_line(&postimage);
47495887 2841 trailing--;
6e7c92a9 2842 }
3cca928d
LT
2843 }
2844
b94f2eda 2845 if (applied_pos >= 0) {
77b15bbd 2846 if (new_blank_lines_at_end &&
51667147 2847 preimage.nr + applied_pos >= img->nr &&
77b15bbd
JH
2848 (ws_rule & WS_BLANK_AT_EOF) &&
2849 ws_error_action != nowarn_ws_error) {
85572639
JH
2850 record_ws_error(WS_BLANK_AT_EOF, "+", 1,
2851 found_new_blank_lines_at_end);
77b15bbd
JH
2852 if (ws_error_action == correct_ws_error) {
2853 while (new_blank_lines_at_end--)
2854 remove_last_line(&postimage);
2855 }
b94f2eda 2856 /*
77b15bbd
JH
2857 * We would want to prevent write_out_results()
2858 * from taking place in apply_patch() that follows
2859 * the callchain led us here, which is:
2860 * apply_patch->check_patch_list->check_patch->
2861 * apply_data->apply_fragments->apply_one_fragment
b94f2eda 2862 */
77b15bbd
JH
2863 if (ws_error_action == die_on_ws_error)
2864 apply = 0;
b94f2eda 2865 }
aeabfa07 2866
334f8cb2
JH
2867 if (apply_verbosely && applied_pos != pos) {
2868 int offset = applied_pos - pos;
2869 if (apply_in_reverse)
2870 offset = 0 - offset;
3638eb43
NTND
2871 fprintf_ln(stderr,
2872 Q_("Hunk #%d succeeded at %d (offset %d line).",
2873 "Hunk #%d succeeded at %d (offset %d lines).",
2874 offset),
2875 nth_fragment, applied_pos + 1, offset);
334f8cb2
JH
2876 }
2877
b94f2eda
JH
2878 /*
2879 * Warn if it was necessary to reduce the number
2880 * of context lines.
2881 */
2882 if ((leading != frag->leading) ||
2883 (trailing != frag->trailing))
3638eb43
NTND
2884 fprintf_ln(stderr, _("Context reduced to (%ld/%ld)"
2885 " to apply fragment at %d"),
2886 leading, trailing, applied_pos+1);
b94f2eda
JH
2887 update_image(img, applied_pos, &preimage, &postimage);
2888 } else {
2889 if (apply_verbosely)
3638eb43 2890 error(_("while searching for:\n%.*s"),
61e08cca 2891 (int)(old - oldlines), oldlines);
b94f2eda 2892 }
aeabfa07 2893
61e08cca 2894 free(oldlines);
d511bd33 2895 strbuf_release(&newlines);
b94f2eda
JH
2896 free(preimage.line_allocated);
2897 free(postimage.line_allocated);
2898
2899 return (applied_pos < 0);
3cca928d
LT
2900}
2901
b94f2eda 2902static int apply_binary_fragment(struct image *img, struct patch *patch)
0660626c 2903{
0660626c 2904 struct fragment *fragment = patch->fragments;
c7f9cb14
PH
2905 unsigned long len;
2906 void *dst;
0660626c 2907
24305cd7 2908 if (!fragment)
3638eb43 2909 return error(_("missing binary patch data for '%s'"),
24305cd7
JK
2910 patch->new_name ?
2911 patch->new_name :
2912 patch->old_name);
2913
3cd4f5e8
JH
2914 /* Binary patch is irreversible without the optional second hunk */
2915 if (apply_in_reverse) {
2916 if (!fragment->next)
2917 return error("cannot reverse-apply a binary patch "
2918 "without the reverse hunk to '%s'",
2919 patch->new_name
2920 ? patch->new_name : patch->old_name);
03eb8f8a 2921 fragment = fragment->next;
3cd4f5e8 2922 }
3cd4f5e8 2923 switch (fragment->binary_patch_method) {
0660626c 2924 case BINARY_DELTA_DEFLATED:
b94f2eda 2925 dst = patch_delta(img->buf, img->len, fragment->patch,
c7f9cb14
PH
2926 fragment->size, &len);
2927 if (!dst)
2928 return -1;
b94f2eda
JH
2929 clear_image(img);
2930 img->buf = dst;
2931 img->len = len;
c7f9cb14 2932 return 0;
0660626c 2933 case BINARY_LITERAL_DEFLATED:
b94f2eda
JH
2934 clear_image(img);
2935 img->len = fragment->size;
5c0b13f8 2936 img->buf = xmemdupz(fragment->patch, img->len);
c7f9cb14 2937 return 0;
0660626c 2938 }
c7f9cb14 2939 return -1;
0660626c
JH
2940}
2941
92737a22
JH
2942/*
2943 * Replace "img" with the result of applying the binary patch.
2944 * The binary patch data itself in patch->fragment is still kept
2945 * but the preimage prepared by the caller in "img" is freed here
2946 * or in the helper function apply_binary_fragment() this calls.
2947 */
b94f2eda 2948static int apply_binary(struct image *img, struct patch *patch)
3cca928d 2949{
011f4274 2950 const char *name = patch->old_name ? patch->old_name : patch->new_name;
051308f6 2951 unsigned char sha1[20];
011f4274 2952
81bf96bb
JH
2953 /*
2954 * For safety, we require patch index line to contain
051308f6
JH
2955 * full 40-byte textual SHA1 for old and new, at least for now.
2956 */
2957 if (strlen(patch->old_sha1_prefix) != 40 ||
2958 strlen(patch->new_sha1_prefix) != 40 ||
2959 get_sha1_hex(patch->old_sha1_prefix, sha1) ||
2960 get_sha1_hex(patch->new_sha1_prefix, sha1))
2961 return error("cannot apply binary patch to '%s' "
2962 "without full index line", name);
011f4274 2963
051308f6 2964 if (patch->old_name) {
81bf96bb
JH
2965 /*
2966 * See if the old one matches what the patch
051308f6 2967 * applies to.
011f4274 2968 */
b94f2eda 2969 hash_sha1_file(img->buf, img->len, blob_type, sha1);
051308f6
JH
2970 if (strcmp(sha1_to_hex(sha1), patch->old_sha1_prefix))
2971 return error("the patch applies to '%s' (%s), "
2972 "which does not match the "
2973 "current contents.",
2974 name, sha1_to_hex(sha1));
2975 }
2976 else {
2977 /* Otherwise, the old one must be empty. */
b94f2eda 2978 if (img->len)
051308f6
JH
2979 return error("the patch applies to an empty "
2980 "'%s' but it is not empty", name);
2981 }
011f4274 2982
0660626c 2983 get_sha1_hex(patch->new_sha1_prefix, sha1);
0bef57ee 2984 if (is_null_sha1(sha1)) {
b94f2eda 2985 clear_image(img);
051308f6 2986 return 0; /* deletion patch */
0660626c 2987 }
011f4274 2988
051308f6 2989 if (has_sha1_file(sha1)) {
0660626c 2990 /* We already have the postimage */
21666f1a 2991 enum object_type type;
051308f6 2992 unsigned long size;
c7f9cb14 2993 char *result;
051308f6 2994
c7f9cb14
PH
2995 result = read_sha1_file(sha1, &type, &size);
2996 if (!result)
051308f6
JH
2997 return error("the necessary postimage %s for "
2998 "'%s' cannot be read",
2999 patch->new_sha1_prefix, name);
b94f2eda
JH
3000 clear_image(img);
3001 img->buf = result;
3002 img->len = size;
c7f9cb14 3003 } else {
81bf96bb
JH
3004 /*
3005 * We have verified buf matches the preimage;
0660626c
JH
3006 * apply the patch data to it, which is stored
3007 * in the patch->fragments->{patch,size}.
011f4274 3008 */
b94f2eda 3009 if (apply_binary_fragment(img, patch))
3638eb43 3010 return error(_("binary patch does not apply to '%s'"),
051308f6 3011 name);
011f4274 3012
051308f6 3013 /* verify that the result matches */
b94f2eda 3014 hash_sha1_file(img->buf, img->len, blob_type, sha1);
051308f6 3015 if (strcmp(sha1_to_hex(sha1), patch->new_sha1_prefix))
3638eb43 3016 return error(_("binary patch to '%s' creates incorrect result (expecting %s, got %s)"),
c7f9cb14 3017 name, patch->new_sha1_prefix, sha1_to_hex(sha1));
011f4274 3018 }
3cca928d 3019
051308f6
JH
3020 return 0;
3021}
3022
b94f2eda 3023static int apply_fragments(struct image *img, struct patch *patch)
051308f6
JH
3024{
3025 struct fragment *frag = patch->fragments;
3026 const char *name = patch->old_name ? patch->old_name : patch->new_name;
cf1b7869
JH
3027 unsigned ws_rule = patch->ws_rule;
3028 unsigned inaccurate_eof = patch->inaccurate_eof;
334f8cb2 3029 int nth = 0;
051308f6
JH
3030
3031 if (patch->is_binary)
b94f2eda 3032 return apply_binary(img, patch);
051308f6 3033
3cca928d 3034 while (frag) {
334f8cb2
JH
3035 nth++;
3036 if (apply_one_fragment(img, frag, inaccurate_eof, ws_rule, nth)) {
3638eb43 3037 error(_("patch failed: %s:%ld"), name, frag->oldpos);
57dc397c
JH
3038 if (!apply_with_reject)
3039 return -1;
3040 frag->rejected = 1;
3041 }
3cca928d
LT
3042 frag = frag->next;
3043 }
30996652 3044 return 0;
3cca928d
LT
3045}
3046
e42a96e7 3047static int read_blob_object(struct strbuf *buf, const unsigned char *sha1, unsigned mode)
e06c5a6c 3048{
e42a96e7 3049 if (S_ISGITLINK(mode)) {
c7f9cb14 3050 strbuf_grow(buf, 100);
e42a96e7 3051 strbuf_addf(buf, "Subproject commit %s\n", sha1_to_hex(sha1));
e06c5a6c
SV
3052 } else {
3053 enum object_type type;
c7f9cb14
PH
3054 unsigned long sz;
3055 char *result;
3056
e42a96e7 3057 result = read_sha1_file(sha1, &type, &sz);
c7f9cb14 3058 if (!result)
e06c5a6c 3059 return -1;
c7f9cb14
PH
3060 /* XXX read_sha1_file NUL-terminates */
3061 strbuf_attach(buf, result, sz, sz + 1);
e06c5a6c
SV
3062 }
3063 return 0;
3064}
3065
9c5e6c80 3066static int read_file_or_gitlink(const struct cache_entry *ce, struct strbuf *buf)
e42a96e7
JH
3067{
3068 if (!ce)
3069 return 0;
3070 return read_blob_object(buf, ce->sha1, ce->ce_mode);
3071}
3072
7a07841c
DZ
3073static struct patch *in_fn_table(const char *name)
3074{
c455c87c 3075 struct string_list_item *item;
7a07841c
DZ
3076
3077 if (name == NULL)
3078 return NULL;
3079
e8c8b713 3080 item = string_list_lookup(&fn_table, name);
7a07841c
DZ
3081 if (item != NULL)
3082 return (struct patch *)item->util;
3083
3084 return NULL;
3085}
3086
7fac0eef
MK
3087/*
3088 * item->util in the filename table records the status of the path.
3089 * Usually it points at a patch (whose result records the contents
3090 * of it after applying it), but it could be PATH_WAS_DELETED for a
f3b8f91a
JH
3091 * path that a previously applied patch has already removed, or
3092 * PATH_TO_BE_DELETED for a path that a later patch would remove.
3093 *
3094 * The latter is needed to deal with a case where two paths A and B
3095 * are swapped by first renaming A to B and then renaming B to A;
41ccfdd9 3096 * moving A to B should not be prevented due to presence of B as we
f3b8f91a 3097 * will remove it in a later patch.
7fac0eef 3098 */
f3b8f91a 3099#define PATH_TO_BE_DELETED ((struct patch *) -2)
7fac0eef
MK
3100#define PATH_WAS_DELETED ((struct patch *) -1)
3101
3102static int to_be_deleted(struct patch *patch)
3103{
3104 return patch == PATH_TO_BE_DELETED;
3105}
3106
3107static int was_deleted(struct patch *patch)
3108{
3109 return patch == PATH_WAS_DELETED;
3110}
3111
7a07841c
DZ
3112static void add_to_fn_table(struct patch *patch)
3113{
c455c87c 3114 struct string_list_item *item;
7a07841c
DZ
3115
3116 /*
3117 * Always add new_name unless patch is a deletion
3118 * This should cover the cases for normal diffs,
3119 * file creations and copies
3120 */
3121 if (patch->new_name != NULL) {
78a395d3 3122 item = string_list_insert(&fn_table, patch->new_name);
7a07841c
DZ
3123 item->util = patch;
3124 }
3125
3126 /*
3127 * store a failure on rename/deletion cases because
3128 * later chunks shouldn't patch old names
3129 */
3130 if ((patch->new_name == NULL) || (patch->is_rename)) {
78a395d3 3131 item = string_list_insert(&fn_table, patch->old_name);
7fac0eef
MK
3132 item->util = PATH_WAS_DELETED;
3133 }
3134}
3135
3136static void prepare_fn_table(struct patch *patch)
3137{
3138 /*
3139 * store information about incoming file deletion
3140 */
3141 while (patch) {
3142 if ((patch->new_name == NULL) || (patch->is_rename)) {
3143 struct string_list_item *item;
78a395d3 3144 item = string_list_insert(&fn_table, patch->old_name);
7fac0eef
MK
3145 item->util = PATH_TO_BE_DELETED;
3146 }
3147 patch = patch->next;
7a07841c
DZ
3148 }
3149}
3150
d4a2024a
NTND
3151static int checkout_target(struct index_state *istate,
3152 struct cache_entry *ce, struct stat *st)
f4c66eed
JH
3153{
3154 struct checkout costate;
3155
3156 memset(&costate, 0, sizeof(costate));
3157 costate.base_dir = "";
3158 costate.refresh_cache = 1;
d4a2024a 3159 costate.istate = istate;
f4c66eed
JH
3160 if (checkout_entry(ce, &costate, NULL) || lstat(ce->name, st))
3161 return error(_("cannot checkout %s"), ce->name);
3162 return 0;
3163}
3164
ccf998b2
JH
3165static struct patch *previous_patch(struct patch *patch, int *gone)
3166{
3167 struct patch *previous;
3168
3169 *gone = 0;
3170 if (patch->is_copy || patch->is_rename)
3171 return NULL; /* "git" patches do not depend on the order */
3172
3173 previous = in_fn_table(patch->old_name);
3174 if (!previous)
3175 return NULL;
3176
3177 if (to_be_deleted(previous))
3178 return NULL; /* the deletion hasn't happened yet */
3179
3180 if (was_deleted(previous))
3181 *gone = 1;
3182
3183 return previous;
3184}
3185
9c5e6c80 3186static int verify_index_match(const struct cache_entry *ce, struct stat *st)
e09837e2
JH
3187{
3188 if (S_ISGITLINK(ce->ce_mode)) {
3189 if (!S_ISDIR(st->st_mode))
3190 return -1;
3191 return 0;
3192 }
3193 return ce_match_stat(ce, st, CE_MATCH_IGNORE_VALID|CE_MATCH_IGNORE_SKIP_WORKTREE);
3194}
3195
5a812661
JH
3196#define SUBMODULE_PATCH_WITHOUT_INDEX 1
3197
3198static int load_patch_target(struct strbuf *buf,
9c5e6c80 3199 const struct cache_entry *ce,
5a812661
JH
3200 struct stat *st,
3201 const char *name,
3202 unsigned expected_mode)
3203{
3204 if (cached) {
3205 if (read_file_or_gitlink(ce, buf))
3206 return error(_("read of %s failed"), name);
3207 } else if (name) {
3208 if (S_ISGITLINK(expected_mode)) {
3209 if (ce)
3210 return read_file_or_gitlink(ce, buf);
3211 else
3212 return SUBMODULE_PATCH_WITHOUT_INDEX;
3213 } else {
3214 if (read_old_data(st, name, buf))
3215 return error(_("read of %s failed"), name);
3216 }
3217 }
3218 return 0;
3219}
3220
ccf998b2
JH
3221/*
3222 * We are about to apply "patch"; populate the "image" with the
3223 * current version we have, from the working tree or from the index,
3224 * depending on the situation e.g. --cached/--index. If we are
3225 * applying a non-git patch that incrementally updates the tree,
3226 * we read from the result of a previous diff.
3227 */
37b9c903 3228static int load_preimage(struct image *image,
9c5e6c80
NTND
3229 struct patch *patch, struct stat *st,
3230 const struct cache_entry *ce)
3cca928d 3231{
f285a2d7 3232 struct strbuf buf = STRBUF_INIT;
b94f2eda
JH
3233 size_t len;
3234 char *img;
ccf998b2
JH
3235 struct patch *previous;
3236 int status;
3cca928d 3237
ccf998b2
JH
3238 previous = previous_patch(patch, &status);
3239 if (status)
3240 return error(_("path %s has been renamed/deleted"),
3241 patch->old_name);
3242 if (previous) {
92737a22 3243 /* We have a patched copy in memory; use that. */
ccf998b2 3244 strbuf_add(&buf, previous->result, previous->resultsize);
5a812661
JH
3245 } else {
3246 status = load_patch_target(&buf, ce, st,
3247 patch->old_name, patch->old_mode);
3248 if (status < 0)
3249 return status;
3250 else if (status == SUBMODULE_PATCH_WITHOUT_INDEX) {
3251 /*
3252 * There is no way to apply subproject
3253 * patch without looking at the index.
3254 * NEEDSWORK: shouldn't this be flagged
3255 * as an error???
3256 */
3257 free_fragment_list(patch->fragments);
3258 patch->fragments = NULL;
3259 } else if (status) {
3638eb43 3260 return error(_("read of %s failed"), patch->old_name);
04e4888e
JH
3261 }
3262 }
6e7c92a9 3263
b94f2eda 3264 img = strbuf_detach(&buf, &len);
37b9c903
JH
3265 prepare_image(image, img, len, !patch->is_binary);
3266 return 0;
3267}
b94f2eda 3268
519d1a5b
JH
3269static int three_way_merge(struct image *image,
3270 char *path,
3271 const unsigned char *base,
3272 const unsigned char *ours,
3273 const unsigned char *theirs)
3274{
28ff0512
JH
3275 mmfile_t base_file, our_file, their_file;
3276 mmbuffer_t result = { NULL };
3277 int status;
5aa7d94c 3278
28ff0512
JH
3279 read_mmblob(&base_file, base);
3280 read_mmblob(&our_file, ours);
3281 read_mmblob(&their_file, theirs);
3282 status = ll_merge(&result, path,
3283 &base_file, "base",
3284 &our_file, "ours",
3285 &their_file, "theirs", NULL);
3286 free(base_file.ptr);
3287 free(our_file.ptr);
3288 free(their_file.ptr);
3289 if (status < 0 || !result.ptr) {
3290 free(result.ptr);
3291 return -1;
3292 }
3293 clear_image(image);
3294 image->buf = result.ptr;
3295 image->len = result.size;
5aa7d94c 3296
28ff0512 3297 return status;
519d1a5b
JH
3298}
3299
099f3c42
JH
3300/*
3301 * When directly falling back to add/add three-way merge, we read from
3302 * the current contents of the new_name. In no cases other than that
3303 * this function will be called.
3304 */
3305static int load_current(struct image *image, struct patch *patch)
3306{
3307 struct strbuf buf = STRBUF_INIT;
3308 int status, pos;
3309 size_t len;
3310 char *img;
3311 struct stat st;
3312 struct cache_entry *ce;
3313 char *name = patch->new_name;
3314 unsigned mode = patch->new_mode;
3315
3316 if (!patch->is_new)
3317 die("BUG: patch to %s is not a creation", patch->old_name);
3318
3319 pos = cache_name_pos(name, strlen(name));
3320 if (pos < 0)
3321 return error(_("%s: does not exist in index"), name);
3322 ce = active_cache[pos];
3323 if (lstat(name, &st)) {
3324 if (errno != ENOENT)
3325 return error(_("%s: %s"), name, strerror(errno));
d4a2024a 3326 if (checkout_target(&the_index, ce, &st))
099f3c42
JH
3327 return -1;
3328 }
3329 if (verify_index_match(ce, &st))
3330 return error(_("%s: does not match index"), name);
3331
3332 status = load_patch_target(&buf, ce, &st, name, mode);
3333 if (status < 0)
3334 return status;
3335 else if (status)
3336 return -1;
3337 img = strbuf_detach(&buf, &len);
3338 prepare_image(image, img, len, !patch->is_binary);
3cca928d
LT
3339 return 0;
3340}
3341
cfb6f9ac 3342static int try_threeway(struct image *image, struct patch *patch,
9c5e6c80 3343 struct stat *st, const struct cache_entry *ce)
64cab591 3344{
519d1a5b
JH
3345 unsigned char pre_sha1[20], post_sha1[20], our_sha1[20];
3346 struct strbuf buf = STRBUF_INIT;
3347 size_t len;
28ff0512 3348 int status;
519d1a5b
JH
3349 char *img;
3350 struct image tmp_image;
3351
3352 /* No point falling back to 3-way merge in these cases */
099f3c42 3353 if (patch->is_delete ||
519d1a5b
JH
3354 S_ISGITLINK(patch->old_mode) || S_ISGITLINK(patch->new_mode))
3355 return -1;
64cab591 3356
519d1a5b 3357 /* Preimage the patch was prepared for */
099f3c42
JH
3358 if (patch->is_new)
3359 write_sha1_file("", 0, blob_type, pre_sha1);
3360 else if (get_sha1(patch->old_sha1_prefix, pre_sha1) ||
3361 read_blob_object(&buf, pre_sha1, patch->old_mode))
519d1a5b 3362 return error("repository lacks the necessary blob to fall back on 3-way merge.");
28ff0512
JH
3363
3364 fprintf(stderr, "Falling back to three-way merge...\n");
3365
519d1a5b
JH
3366 img = strbuf_detach(&buf, &len);
3367 prepare_image(&tmp_image, img, len, 1);
3368 /* Apply the patch to get the post image */
3369 if (apply_fragments(&tmp_image, patch) < 0) {
3370 clear_image(&tmp_image);
3371 return -1;
3372 }
3373 /* post_sha1[] is theirs */
3374 write_sha1_file(tmp_image.buf, tmp_image.len, blob_type, post_sha1);
3375 clear_image(&tmp_image);
3376
3377 /* our_sha1[] is ours */
099f3c42
JH
3378 if (patch->is_new) {
3379 if (load_current(&tmp_image, patch))
3380 return error("cannot read the current contents of '%s'",
3381 patch->new_name);
3382 } else {
3383 if (load_preimage(&tmp_image, patch, st, ce))
3384 return error("cannot read the current contents of '%s'",
3385 patch->old_name);
3386 }
519d1a5b
JH
3387 write_sha1_file(tmp_image.buf, tmp_image.len, blob_type, our_sha1);
3388 clear_image(&tmp_image);
3389
3390 /* in-core three-way merge between post and our using pre as base */
28ff0512
JH
3391 status = three_way_merge(image, patch->new_name,
3392 pre_sha1, our_sha1, post_sha1);
3393 if (status < 0) {
3394 fprintf(stderr, "Failed to fall back on three-way merge...\n");
3395 return status;
3396 }
3397
3398 if (status) {
3399 patch->conflicted_threeway = 1;
4f4a6cb9
JH
3400 if (patch->is_new)
3401 hashclr(patch->threeway_stage[0]);
3402 else
3403 hashcpy(patch->threeway_stage[0], pre_sha1);
28ff0512
JH
3404 hashcpy(patch->threeway_stage[1], our_sha1);
3405 hashcpy(patch->threeway_stage[2], post_sha1);
3406 fprintf(stderr, "Applied patch to '%s' with conflicts.\n", patch->new_name);
3407 } else {
3408 fprintf(stderr, "Applied patch to '%s' cleanly.\n", patch->new_name);
64cab591 3409 }
64cab591
JH
3410 return 0;
3411}
3412
9c5e6c80 3413static int apply_data(struct patch *patch, struct stat *st, const struct cache_entry *ce)
e06c5a6c 3414{
37b9c903
JH
3415 struct image image;
3416
3417 if (load_preimage(&image, patch, st, ce) < 0)
3418 return -1;
b94f2eda 3419
099f3c42
JH
3420 if (patch->direct_to_threeway ||
3421 apply_fragments(&image, patch) < 0) {
cfb6f9ac
JH
3422 /* Note: with --reject, apply_fragments() returns 0 */
3423 if (!threeway || try_threeway(&image, patch, st, ce) < 0)
e06c5a6c 3424 return -1;
e06c5a6c 3425 }
b94f2eda
JH
3426 patch->result = image.buf;
3427 patch->resultsize = image.len;
7a07841c 3428 add_to_fn_table(patch);
b94f2eda 3429 free(image.line_allocated);
5aa7d94c 3430
4be60962 3431 if (0 < patch->is_delete && patch->resultsize)
3638eb43 3432 return error(_("removal patch leaves file contents"));
5aa7d94c 3433
3cca928d 3434 return 0;
e06c5a6c
SV
3435}
3436
ccf998b2
JH
3437/*
3438 * If "patch" that we are looking at modifies or deletes what we have,
3439 * we would want it not to lose any local modification we have, either
3440 * in the working tree or in the index.
3441 *
3442 * This also decides if a non-git patch is a creation patch or a
3443 * modification to an existing empty file. We do not check the state
3444 * of the current tree for a creation patch in this function; the caller
3445 * check_patch() separately makes sure (and errors out otherwise) that
3446 * the path the patch creates does not exist in the current tree.
3447 */
5c47f4c6 3448static int check_preimage(struct patch *patch, struct cache_entry **ce, struct stat *st)
fab2c257
LT
3449{
3450 const char *old_name = patch->old_name;
ccf998b2
JH
3451 struct patch *previous = NULL;
3452 int stat_ret = 0, status;
5c47f4c6 3453 unsigned st_mode = 0;
e06c5a6c 3454
5c47f4c6
JH
3455 if (!old_name)
3456 return 0;
04e4888e 3457
5c47f4c6 3458 assert(patch->is_new <= 0);
ccf998b2 3459 previous = previous_patch(patch, &status);
a9a3e82e 3460
ccf998b2
JH
3461 if (status)
3462 return error(_("path %s has been renamed/deleted"), old_name);
3463 if (previous) {
3464 st_mode = previous->new_mode;
7a07841c 3465 } else if (!cached) {
5c47f4c6
JH
3466 stat_ret = lstat(old_name, st);
3467 if (stat_ret && errno != ENOENT)
3638eb43 3468 return error(_("%s: %s"), old_name, strerror(errno));
5c47f4c6 3469 }
a9a3e82e 3470
ccf998b2 3471 if (check_index && !previous) {
5c47f4c6
JH
3472 int pos = cache_name_pos(old_name, strlen(old_name));
3473 if (pos < 0) {
3474 if (patch->is_new < 0)
3475 goto is_new;
3638eb43 3476 return error(_("%s: does not exist in index"), old_name);
5c47f4c6
JH
3477 }
3478 *ce = active_cache[pos];
3479 if (stat_ret < 0) {
d4a2024a 3480 if (checkout_target(&the_index, *ce, st))
5c47f4c6
JH
3481 return -1;
3482 }
3483 if (!cached && verify_index_match(*ce, st))
3638eb43 3484 return error(_("%s: does not match index"), old_name);
5c47f4c6
JH
3485 if (cached)
3486 st_mode = (*ce)->ce_mode;
3487 } else if (stat_ret < 0) {
3cca928d 3488 if (patch->is_new < 0)
5c47f4c6 3489 goto is_new;
3638eb43 3490 return error(_("%s: %s"), old_name, strerror(errno));
fab2c257 3491 }
a577284a 3492
ccf998b2 3493 if (!cached && !previous)
5c47f4c6
JH
3494 st_mode = ce_mode_from_stat(*ce, st->st_mode);
3495
3496 if (patch->is_new < 0)