apply: avoid possible bogus pointer
[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"
03ac6e64 10#include "cache-tree.h"
22943f1a 11#include "quote.h"
8e440259 12#include "blob.h"
051308f6 13#include "delta.h"
ac6245e3 14#include "builtin.h"
c455c87c 15#include "string-list.h"
175a4948 16#include "dir.h"
7f814632 17#include "diff.h"
f26c4940 18#include "parse-options.h"
28ff0512
JH
19#include "xdiff-interface.h"
20#include "ll-merge.h"
f2633ebd 21#include "rerere.h"
c1bb9350 22
a9486b02
PR
23/*
24 * --check turns on checking that the working tree matches the
25 * files that are being modified, but doesn't apply the patch
26 * --stat does just a diffstat, and doesn't actually apply
27 * --numstat does numeric diffstat, and doesn't actually apply
28 * --index-info shows the old and new index info for paths if available.
29 * --index updates the cache as well.
30 * --cached updates only the cache without ever touching the working tree.
31 */
edf2e370
JH
32static const char *prefix;
33static int prefix_length = -1;
dbd0f7d3 34static int newfd = -1;
edf2e370 35
4be60962 36static int unidiff_zero;
e36f8b60 37static int p_value = 1;
3e8a5db9 38static int p_value_known;
96f1e58f 39static int check_index;
7da3bf37 40static int update_index;
96f1e58f
DR
41static int cached;
42static int diffstat;
43static int numstat;
44static int summary;
45static int check;
a577284a 46static int apply = 1;
96f1e58f 47static int apply_in_reverse;
57dc397c 48static int apply_with_reject;
a2bf404e 49static int apply_verbosely;
933e44d3 50static int allow_overlap;
96f1e58f 51static int no_add;
cfb6f9ac 52static int threeway;
7a988699 53static const char *fake_ancestor;
22943f1a 54static int line_termination = '\n';
f26c4940
MV
55static unsigned int p_context = UINT_MAX;
56static const char * const apply_usage[] = {
de373933 57 N_("git apply [options] [<patch>...]"),
f26c4940
MV
58 NULL
59};
c1bb9350 60
81bf96bb
JH
61static enum ws_error_action {
62 nowarn_ws_error,
63 warn_on_ws_error,
64 die_on_ws_error,
4b05548f 65 correct_ws_error
81bf96bb 66} ws_error_action = warn_on_ws_error;
96f1e58f 67static int whitespace_error;
fc96b7c9 68static int squelch_whitespace_errors = 5;
c94bf41c 69static int applied_after_fixing_ws;
86c91f91
GB
70
71static enum ws_ignore {
72 ignore_ws_none,
4b05548f 73 ignore_ws_change
86c91f91
GB
74} ws_ignore_action = ignore_ws_none;
75
76
96f1e58f 77static const char *patch_input_file;
c4730f35
JS
78static const char *root;
79static int root_len;
f26c4940
MV
80static int read_stdin = 1;
81static int options;
19bfcd5a 82
2ae1c53b
JH
83static void parse_whitespace_option(const char *option)
84{
85 if (!option) {
81bf96bb 86 ws_error_action = warn_on_ws_error;
2ae1c53b
JH
87 return;
88 }
89 if (!strcmp(option, "warn")) {
81bf96bb 90 ws_error_action = warn_on_ws_error;
2ae1c53b
JH
91 return;
92 }
621603b7 93 if (!strcmp(option, "nowarn")) {
81bf96bb 94 ws_error_action = nowarn_ws_error;
621603b7
JH
95 return;
96 }
2ae1c53b 97 if (!strcmp(option, "error")) {
81bf96bb 98 ws_error_action = die_on_ws_error;
2ae1c53b
JH
99 return;
100 }
101 if (!strcmp(option, "error-all")) {
81bf96bb 102 ws_error_action = die_on_ws_error;
2ae1c53b
JH
103 squelch_whitespace_errors = 0;
104 return;
105 }
81bf96bb
JH
106 if (!strcmp(option, "strip") || !strcmp(option, "fix")) {
107 ws_error_action = correct_ws_error;
2ae1c53b
JH
108 return;
109 }
3638eb43 110 die(_("unrecognized whitespace option '%s'"), option);
2ae1c53b
JH
111}
112
86c91f91
GB
113static void parse_ignorewhitespace_option(const char *option)
114{
115 if (!option || !strcmp(option, "no") ||
116 !strcmp(option, "false") || !strcmp(option, "never") ||
117 !strcmp(option, "none")) {
118 ws_ignore_action = ignore_ws_none;
119 return;
120 }
121 if (!strcmp(option, "change")) {
122 ws_ignore_action = ignore_ws_change;
123 return;
124 }
3638eb43 125 die(_("unrecognized whitespace ignore option '%s'"), option);
86c91f91
GB
126}
127
f21d6726
JH
128static void set_default_whitespace_mode(const char *whitespace_option)
129{
81bf96bb
JH
130 if (!whitespace_option && !apply_default_whitespace)
131 ws_error_action = (apply ? warn_on_ws_error : nowarn_ws_error);
f21d6726
JH
132}
133
3f40315a
LT
134/*
135 * For "diff-stat" like behaviour, we keep track of the biggest change
136 * we've seen, and the longest filename. That allows us to do simple
137 * scaling.
138 */
139static int max_change, max_len;
140
a4acb0eb
LT
141/*
142 * Various "current state", notably line numbers and what
143 * file (and how) we're patching right now.. The "is_xxxx"
144 * things are flags, where -1 means "don't know yet".
145 */
46979f56 146static int linenr = 1;
19c58fb8 147
3cd4f5e8
JH
148/*
149 * This represents one "hunk" from a patch, starting with
150 * "@@ -oldpos,oldlines +newpos,newlines @@" marker. The
151 * patch text is pointed at by patch, and its byte length
152 * is stored in size. leading and trailing are the number
153 * of context lines.
154 */
19c58fb8 155struct fragment {
47495887 156 unsigned long leading, trailing;
19c58fb8
LT
157 unsigned long oldpos, oldlines;
158 unsigned long newpos, newlines;
92737a22
JH
159 /*
160 * 'patch' is usually borrowed from buf in apply_patch(),
161 * but some codepaths store an allocated buffer.
162 */
19c58fb8 163 const char *patch;
6fe53908
JH
164 unsigned free_patch:1,
165 rejected:1;
19c58fb8 166 int size;
77b15bbd 167 int linenr;
19c58fb8
LT
168 struct fragment *next;
169};
170
3cd4f5e8
JH
171/*
172 * When dealing with a binary patch, we reuse "leading" field
173 * to store the type of the binary hunk, either deflated "delta"
174 * or deflated "literal".
175 */
176#define binary_patch_method leading
177#define BINARY_DELTA_DEFLATED 1
178#define BINARY_LITERAL_DEFLATED 2
179
81bf96bb
JH
180/*
181 * This represents a "patch" to a file, both metainfo changes
182 * such as creation/deletion, filemode and content changes represented
183 * as a series of fragments.
184 */
19c58fb8 185struct patch {
5041aa70 186 char *new_name, *old_name, *def_name;
19c58fb8 187 unsigned int old_mode, new_mode;
3dad11bf 188 int is_new, is_delete; /* -1 = unknown, 0 = false, 1 = true */
57dc397c 189 int rejected;
cf1b7869 190 unsigned ws_rule;
3f40315a 191 int lines_added, lines_deleted;
96c912a4 192 int score;
9987d7c5 193 unsigned int is_toplevel_relative:1;
3dad11bf
RS
194 unsigned int inaccurate_eof:1;
195 unsigned int is_binary:1;
196 unsigned int is_copy:1;
197 unsigned int is_rename:1;
c14b9d1e 198 unsigned int recount:1;
28ff0512 199 unsigned int conflicted_threeway:1;
099f3c42 200 unsigned int direct_to_threeway:1;
19c58fb8 201 struct fragment *fragments;
5aa7d94c 202 char *result;
c32f749f 203 size_t resultsize;
2cf67f1e
JH
204 char old_sha1_prefix[41];
205 char new_sha1_prefix[41];
19c58fb8 206 struct patch *next;
28ff0512
JH
207
208 /* three-way fallback result */
209 unsigned char threeway_stage[3][20];
19c58fb8 210};
46979f56 211
9d16c2d5 212static void free_fragment_list(struct fragment *list)
6fe53908 213{
9d16c2d5
JH
214 while (list) {
215 struct fragment *next = list->next;
216 if (list->free_patch)
217 free((char *)list->patch);
218 free(list);
219 list = next;
a604ddef 220 }
9d16c2d5
JH
221}
222
223static void free_patch(struct patch *patch)
224{
225 free_fragment_list(patch->fragments);
2901bbe5
JH
226 free(patch->def_name);
227 free(patch->old_name);
228 free(patch->new_name);
8192a2fa 229 free(patch->result);
a604ddef
JH
230 free(patch);
231}
232
233static void free_patch_list(struct patch *list)
234{
235 while (list) {
236 struct patch *next = list->next;
237 free_patch(list);
238 list = next;
6fe53908
JH
239 }
240}
241
b94f2eda
JH
242/*
243 * A line in a file, len-bytes long (includes the terminating LF,
244 * except for an incomplete line at the end if the file ends with
245 * one), and its contents hashes to 'hash'.
246 */
247struct line {
248 size_t len;
249 unsigned hash : 24;
250 unsigned flag : 8;
c330fdd4 251#define LINE_COMMON 1
9d158601 252#define LINE_PATCHED 2
b94f2eda
JH
253};
254
255/*
256 * This represents a "file", which is an array of "lines".
257 */
258struct image {
259 char *buf;
260 size_t len;
261 size_t nr;
c330fdd4 262 size_t alloc;
b94f2eda
JH
263 struct line *line_allocated;
264 struct line *line;
265};
266
7a07841c
DZ
267/*
268 * Records filenames that have been touched, in order to handle
269 * the case where more than one patches touch the same file.
270 */
271
c455c87c 272static struct string_list fn_table;
7a07841c 273
b94f2eda
JH
274static uint32_t hash_line(const char *cp, size_t len)
275{
276 size_t i;
277 uint32_t h;
278 for (i = 0, h = 0; i < len; i++) {
279 if (!isspace(cp[i])) {
280 h = h * 3 + (cp[i] & 0xff);
281 }
282 }
283 return h;
284}
285
86c91f91
GB
286/*
287 * Compare lines s1 of length n1 and s2 of length n2, ignoring
288 * whitespace difference. Returns 1 if they match, 0 otherwise
289 */
290static int fuzzy_matchlines(const char *s1, size_t n1,
291 const char *s2, size_t n2)
292{
293 const char *last1 = s1 + n1 - 1;
294 const char *last2 = s2 + n2 - 1;
295 int result = 0;
296
86c91f91
GB
297 /* ignore line endings */
298 while ((*last1 == '\r') || (*last1 == '\n'))
299 last1--;
300 while ((*last2 == '\r') || (*last2 == '\n'))
301 last2--;
302
303 /* skip leading whitespace */
304 while (isspace(*s1) && (s1 <= last1))
305 s1++;
306 while (isspace(*s2) && (s2 <= last2))
307 s2++;
308 /* early return if both lines are empty */
309 if ((s1 > last1) && (s2 > last2))
310 return 1;
311 while (!result) {
312 result = *s1++ - *s2++;
313 /*
314 * Skip whitespace inside. We check for whitespace on
315 * both buffers because we don't want "a b" to match
316 * "ab"
317 */
318 if (isspace(*s1) && isspace(*s2)) {
319 while (isspace(*s1) && s1 <= last1)
320 s1++;
321 while (isspace(*s2) && s2 <= last2)
322 s2++;
323 }
324 /*
325 * If we reached the end on one side only,
326 * lines don't match
327 */
328 if (
329 ((s2 > last2) && (s1 <= last1)) ||
330 ((s1 > last1) && (s2 <= last2)))
331 return 0;
332 if ((s1 > last1) && (s2 > last2))
333 break;
334 }
335
336 return !result;
337}
338
c330fdd4
JH
339static void add_line_info(struct image *img, const char *bol, size_t len, unsigned flag)
340{
341 ALLOC_GROW(img->line_allocated, img->nr + 1, img->alloc);
342 img->line_allocated[img->nr].len = len;
343 img->line_allocated[img->nr].hash = hash_line(bol, len);
344 img->line_allocated[img->nr].flag = flag;
345 img->nr++;
346}
347
92737a22
JH
348/*
349 * "buf" has the file contents to be patched (read from various sources).
350 * attach it to "image" and add line-based index to it.
351 * "image" now owns the "buf".
352 */
b94f2eda
JH
353static void prepare_image(struct image *image, char *buf, size_t len,
354 int prepare_linetable)
355{
356 const char *cp, *ep;
b94f2eda 357
c330fdd4 358 memset(image, 0, sizeof(*image));
b94f2eda
JH
359 image->buf = buf;
360 image->len = len;
361
c330fdd4 362 if (!prepare_linetable)
b94f2eda 363 return;
b94f2eda
JH
364
365 ep = image->buf + image->len;
b94f2eda 366 cp = image->buf;
b94f2eda
JH
367 while (cp < ep) {
368 const char *next;
369 for (next = cp; next < ep && *next != '\n'; next++)
370 ;
371 if (next < ep)
372 next++;
c330fdd4 373 add_line_info(image, cp, next - cp, 0);
b94f2eda 374 cp = next;
b94f2eda 375 }
c330fdd4 376 image->line = image->line_allocated;
b94f2eda
JH
377}
378
379static void clear_image(struct image *image)
380{
381 free(image->buf);
798b9ce8
JH
382 free(image->line_allocated);
383 memset(image, 0, sizeof(*image));
b94f2eda
JH
384}
385
5613e811
NTND
386/* fmt must contain _one_ %s and no other substitution */
387static void say_patch_name(FILE *output, const char *fmt, struct patch *patch)
a2bf404e 388{
5613e811
NTND
389 struct strbuf sb = STRBUF_INIT;
390
a2bf404e
JH
391 if (patch->old_name && patch->new_name &&
392 strcmp(patch->old_name, patch->new_name)) {
5613e811
NTND
393 quote_c_style(patch->old_name, &sb, NULL, 0);
394 strbuf_addstr(&sb, " => ");
395 quote_c_style(patch->new_name, &sb, NULL, 0);
663af342 396 } else {
a2bf404e
JH
397 const char *n = patch->new_name;
398 if (!n)
399 n = patch->old_name;
5613e811 400 quote_c_style(n, &sb, NULL, 0);
a2bf404e 401 }
5613e811
NTND
402 fprintf(output, fmt, sb.buf);
403 fputc('\n', output);
404 strbuf_release(&sb);
a2bf404e
JH
405}
406
a4acb0eb 407#define SLOP (16)
c1bb9350 408
9a76adeb 409static void read_patch_file(struct strbuf *sb, int fd)
c1bb9350 410{
9a76adeb 411 if (strbuf_read(sb, fd, 0) < 0)
d824cbba 412 die_errno("git apply: failed to read");
a4acb0eb
LT
413
414 /*
415 * Make sure that we have some slop in the buffer
416 * so that we can do speculative "memcmp" etc, and
417 * see to it that it is NUL-filled.
418 */
9a76adeb
PH
419 strbuf_grow(sb, SLOP);
420 memset(sb->buf + sb->len, 0, SLOP);
c1bb9350
LT
421}
422
3cca928d 423static unsigned long linelen(const char *buffer, unsigned long size)
c1bb9350
LT
424{
425 unsigned long len = 0;
426 while (size--) {
427 len++;
428 if (*buffer++ == '\n')
429 break;
430 }
431 return len;
432}
433
a4acb0eb
LT
434static int is_dev_null(const char *str)
435{
436 return !memcmp("/dev/null", str, 9) && isspace(str[9]);
437}
438
381ca9a3
LT
439#define TERM_SPACE 1
440#define TERM_TAB 2
9a4a100e
LT
441
442static int name_terminate(const char *name, int namelen, int c, int terminate)
443{
444 if (c == ' ' && !(terminate & TERM_SPACE))
445 return 0;
446 if (c == '\t' && !(terminate & TERM_TAB))
447 return 0;
448
9a4a100e
LT
449 return 1;
450}
451
33eb4dd9
MM
452/* remove double slashes to make --index work with such filenames */
453static char *squash_slash(char *name)
454{
455 int i = 0, j = 0;
456
15862087
AG
457 if (!name)
458 return NULL;
459
33eb4dd9
MM
460 while (name[i]) {
461 if ((name[j++] = name[i++]) == '/')
462 while (name[i] == '/')
463 i++;
464 }
465 name[j] = '\0';
466 return name;
467}
468
2901bbe5 469static char *find_name_gnu(const char *line, const char *def, int p_value)
c1bb9350 470{
bb7306b5
JN
471 struct strbuf name = STRBUF_INIT;
472 char *cp;
15862087 473
bb7306b5
JN
474 /*
475 * Proposed "new-style" GNU patch/diff format; see
efe6de6e 476 * http://marc.info/?l=git&m=112927316408690&w=2
bb7306b5
JN
477 */
478 if (unquote_c_style(&name, line, NULL)) {
479 strbuf_release(&name);
480 return NULL;
481 }
7fb1011e 482
bb7306b5
JN
483 for (cp = name.buf; p_value; p_value--) {
484 cp = strchr(cp, '/');
485 if (!cp) {
486 strbuf_release(&name);
487 return NULL;
22943f1a 488 }
bb7306b5
JN
489 cp++;
490 }
491
bb7306b5 492 strbuf_remove(&name, 0, cp - name.buf);
bb7306b5
JN
493 if (root)
494 strbuf_insert(&name, 0, root, root_len);
495 return squash_slash(strbuf_detach(&name, NULL));
496}
497
2d502e1f 498static size_t sane_tz_len(const char *line, size_t len)
c1bb9350 499{
5a12c886 500 const char *tz, *p;
15862087 501
5a12c886
JN
502 if (len < strlen(" +0500") || line[len-strlen(" +0500")] != ' ')
503 return 0;
504 tz = line + len - strlen(" +0500");
505
506 if (tz[1] != '+' && tz[1] != '-')
507 return 0;
508
509 for (p = tz + 2; p != line + len; p++)
510 if (!isdigit(*p))
511 return 0;
512
513 return line + len - tz;
514}
515
2d502e1f
JN
516static size_t tz_with_colon_len(const char *line, size_t len)
517{
518 const char *tz, *p;
519
520 if (len < strlen(" +08:00") || line[len - strlen(":00")] != ':')
521 return 0;
522 tz = line + len - strlen(" +08:00");
523
524 if (tz[0] != ' ' || (tz[1] != '+' && tz[1] != '-'))
525 return 0;
526 p = tz + 2;
527 if (!isdigit(*p++) || !isdigit(*p++) || *p++ != ':' ||
528 !isdigit(*p++) || !isdigit(*p++))
529 return 0;
530
531 return line + len - tz;
532}
533
5a12c886
JN
534static size_t date_len(const char *line, size_t len)
535{
536 const char *date, *p;
537
538 if (len < strlen("72-02-05") || line[len-strlen("-05")] != '-')
539 return 0;
540 p = date = line + len - strlen("72-02-05");
541
542 if (!isdigit(*p++) || !isdigit(*p++) || *p++ != '-' ||
543 !isdigit(*p++) || !isdigit(*p++) || *p++ != '-' ||
544 !isdigit(*p++) || !isdigit(*p++)) /* Not a date. */
545 return 0;
546
547 if (date - line >= strlen("19") &&
548 isdigit(date[-1]) && isdigit(date[-2])) /* 4-digit year */
549 date -= strlen("19");
550
551 return line + len - date;
552}
553
554static size_t short_time_len(const char *line, size_t len)
555{
556 const char *time, *p;
557
558 if (len < strlen(" 07:01:32") || line[len-strlen(":32")] != ':')
559 return 0;
560 p = time = line + len - strlen(" 07:01:32");
561
562 /* Permit 1-digit hours? */
563 if (*p++ != ' ' ||
564 !isdigit(*p++) || !isdigit(*p++) || *p++ != ':' ||
565 !isdigit(*p++) || !isdigit(*p++) || *p++ != ':' ||
566 !isdigit(*p++) || !isdigit(*p++)) /* Not a time. */
567 return 0;
568
569 return line + len - time;
570}
571
572static size_t fractional_time_len(const char *line, size_t len)
573{
574 const char *p;
575 size_t n;
576
577 /* Expected format: 19:41:17.620000023 */
578 if (!len || !isdigit(line[len - 1]))
579 return 0;
580 p = line + len - 1;
581
582 /* Fractional seconds. */
583 while (p > line && isdigit(*p))
584 p--;
585 if (*p != '.')
586 return 0;
587
588 /* Hours, minutes, and whole seconds. */
589 n = short_time_len(line, p - line);
590 if (!n)
591 return 0;
592
593 return line + len - p + n;
594}
595
596static size_t trailing_spaces_len(const char *line, size_t len)
597{
598 const char *p;
599
600 /* Expected format: ' ' x (1 or more) */
601 if (!len || line[len - 1] != ' ')
602 return 0;
603
604 p = line + len;
605 while (p != line) {
606 p--;
607 if (*p != ' ')
608 return line + len - (p + 1);
22943f1a
JH
609 }
610
5a12c886
JN
611 /* All spaces! */
612 return len;
613}
614
615static size_t diff_timestamp_len(const char *line, size_t len)
616{
617 const char *end = line + len;
618 size_t n;
619
620 /*
621 * Posix: 2010-07-05 19:41:17
622 * GNU: 2010-07-05 19:41:17.620000023 -0500
623 */
624
625 if (!isdigit(end[-1]))
626 return 0;
627
2d502e1f
JN
628 n = sane_tz_len(line, end - line);
629 if (!n)
630 n = tz_with_colon_len(line, end - line);
5a12c886
JN
631 end -= n;
632
633 n = short_time_len(line, end - line);
634 if (!n)
635 n = fractional_time_len(line, end - line);
636 end -= n;
637
638 n = date_len(line, end - line);
639 if (!n) /* No date. Too bad. */
640 return 0;
641 end -= n;
642
643 if (end == line) /* No space before date. */
644 return 0;
645 if (end[-1] == '\t') { /* Success! */
646 end--;
647 return line + len - end;
648 }
649 if (end[-1] != ' ') /* No space before date. */
650 return 0;
651
652 /* Whitespace damage. */
653 end -= trailing_spaces_len(line, end - line);
654 return line + len - end;
655}
656
2901bbe5
JH
657static char *null_strdup(const char *s)
658{
659 return s ? xstrdup(s) : NULL;
660}
661
662static char *find_name_common(const char *line, const char *def,
663 int p_value, const char *end, int terminate)
5a12c886
JN
664{
665 int len;
666 const char *start = NULL;
667
bb7306b5
JN
668 if (p_value == 0)
669 start = line;
5a12c886 670 while (line != end) {
a4acb0eb 671 char c = *line;
9a4a100e 672
5a12c886 673 if (!end && isspace(c)) {
9a4a100e
LT
674 if (c == '\n')
675 break;
676 if (name_terminate(start, line-start, c, terminate))
677 break;
678 }
a4acb0eb
LT
679 line++;
680 if (c == '/' && !--p_value)
681 start = line;
682 }
683 if (!start)
2901bbe5 684 return squash_slash(null_strdup(def));
a4acb0eb
LT
685 len = line - start;
686 if (!len)
2901bbe5 687 return squash_slash(null_strdup(def));
a4acb0eb
LT
688
689 /*
690 * Generally we prefer the shorter name, especially
691 * if the other one is just a variation of that with
692 * something else tacked on to the end (ie "file.orig"
693 * or "file~").
694 */
695 if (def) {
696 int deflen = strlen(def);
697 if (deflen < len && !strncmp(start, def, deflen))
2901bbe5 698 return squash_slash(xstrdup(def));
c1bb9350 699 }
a4acb0eb 700
c4730f35
JS
701 if (root) {
702 char *ret = xmalloc(root_len + len + 1);
703 strcpy(ret, root);
704 memcpy(ret + root_len, start, len);
705 ret[root_len + len] = '\0';
33eb4dd9 706 return squash_slash(ret);
c4730f35
JS
707 }
708
33eb4dd9 709 return squash_slash(xmemdupz(start, len));
a4acb0eb
LT
710}
711
5a12c886
JN
712static char *find_name(const char *line, char *def, int p_value, int terminate)
713{
714 if (*line == '"') {
715 char *name = find_name_gnu(line, def, p_value);
716 if (name)
717 return name;
718 }
719
720 return find_name_common(line, def, p_value, NULL, terminate);
721}
722
723static char *find_name_traditional(const char *line, char *def, int p_value)
724{
1f976bd0 725 size_t len;
5a12c886
JN
726 size_t date_len;
727
728 if (*line == '"') {
729 char *name = find_name_gnu(line, def, p_value);
730 if (name)
731 return name;
732 }
733
734 len = strchrnul(line, '\n') - line;
735 date_len = diff_timestamp_len(line, len);
736 if (!date_len)
737 return find_name_common(line, def, p_value, NULL, TERM_TAB);
738 len -= date_len;
739
740 return find_name_common(line, def, p_value, line + len, 0);
741}
742
3e8a5db9
JH
743static int count_slashes(const char *cp)
744{
745 int cnt = 0;
746 char ch;
747
748 while ((ch = *cp++))
749 if (ch == '/')
750 cnt++;
751 return cnt;
752}
753
754/*
755 * Given the string after "--- " or "+++ ", guess the appropriate
756 * p_value for the given patch.
757 */
758static int guess_p_value(const char *nameline)
759{
760 char *name, *cp;
761 int val = -1;
762
763 if (is_dev_null(nameline))
764 return -1;
5a12c886 765 name = find_name_traditional(nameline, NULL, 0);
3e8a5db9
JH
766 if (!name)
767 return -1;
768 cp = strchr(name, '/');
769 if (!cp)
770 val = 0;
771 else if (prefix) {
772 /*
773 * Does it begin with "a/$our-prefix" and such? Then this is
774 * very likely to apply to our directory.
775 */
776 if (!strncmp(name, prefix, prefix_length))
777 val = count_slashes(prefix);
778 else {
779 cp++;
780 if (!strncmp(cp, prefix, prefix_length))
781 val = count_slashes(prefix) + 1;
782 }
783 }
784 free(name);
785 return val;
786}
787
c4593faf
JH
788/*
789 * Does the ---/+++ line has the POSIX timestamp after the last HT?
790 * GNU diff puts epoch there to signal a creation/deletion event. Is
791 * this such a timestamp?
792 */
793static int has_epoch_timestamp(const char *nameline)
794{
795 /*
796 * We are only interested in epoch timestamp; any non-zero
797 * fraction cannot be one, hence "(\.0+)?" in the regexp below.
798 * For the same reason, the date must be either 1969-12-31 or
799 * 1970-01-01, and the seconds part must be "00".
800 */
801 const char stamp_regexp[] =
802 "^(1969-12-31|1970-01-01)"
803 " "
804 "[0-2][0-9]:[0-5][0-9]:00(\\.0+)?"
805 " "
a1980c4e
AK
806 "([-+][0-2][0-9]:?[0-5][0-9])\n";
807 const char *timestamp = NULL, *cp, *colon;
c4593faf
JH
808 static regex_t *stamp;
809 regmatch_t m[10];
810 int zoneoffset;
811 int hourminute;
812 int status;
813
814 for (cp = nameline; *cp != '\n'; cp++) {
815 if (*cp == '\t')
816 timestamp = cp + 1;
817 }
818 if (!timestamp)
819 return 0;
820 if (!stamp) {
821 stamp = xmalloc(sizeof(*stamp));
822 if (regcomp(stamp, stamp_regexp, REG_EXTENDED)) {
3638eb43 823 warning(_("Cannot prepare timestamp regexp %s"),
c4593faf
JH
824 stamp_regexp);
825 return 0;
826 }
827 }
828
829 status = regexec(stamp, timestamp, ARRAY_SIZE(m), m, 0);
830 if (status) {
831 if (status != REG_NOMATCH)
3638eb43 832 warning(_("regexec returned %d for input: %s"),
c4593faf
JH
833 status, timestamp);
834 return 0;
835 }
836
a1980c4e
AK
837 zoneoffset = strtol(timestamp + m[3].rm_so + 1, (char **) &colon, 10);
838 if (*colon == ':')
839 zoneoffset = zoneoffset * 60 + strtol(colon + 1, NULL, 10);
840 else
841 zoneoffset = (zoneoffset / 100) * 60 + (zoneoffset % 100);
c4593faf
JH
842 if (timestamp[m[3].rm_so] == '-')
843 zoneoffset = -zoneoffset;
844
845 /*
846 * YYYY-MM-DD hh:mm:ss must be from either 1969-12-31
847 * (west of GMT) or 1970-01-01 (east of GMT)
848 */
849 if ((zoneoffset < 0 && memcmp(timestamp, "1969-12-31", 10)) ||
850 (0 <= zoneoffset && memcmp(timestamp, "1970-01-01", 10)))
851 return 0;
852
853 hourminute = (strtol(timestamp + 11, NULL, 10) * 60 +
854 strtol(timestamp + 14, NULL, 10) -
855 zoneoffset);
856
857 return ((zoneoffset < 0 && hourminute == 1440) ||
858 (0 <= zoneoffset && !hourminute));
859}
860
a4acb0eb 861/*
88f6dbaf 862 * Get the name etc info from the ---/+++ lines of a traditional patch header
a4acb0eb 863 *
9a4a100e
LT
864 * FIXME! The end-of-filename heuristics are kind of screwy. For existing
865 * files, we can happily check the index for a match, but for creating a
866 * new file we should try to match whatever "patch" does. I have no idea.
a4acb0eb 867 */
19c58fb8 868static void parse_traditional_patch(const char *first, const char *second, struct patch *patch)
a4acb0eb 869{
a4acb0eb
LT
870 char *name;
871
a9486b02
PR
872 first += 4; /* skip "--- " */
873 second += 4; /* skip "+++ " */
3e8a5db9
JH
874 if (!p_value_known) {
875 int p, q;
876 p = guess_p_value(first);
877 q = guess_p_value(second);
878 if (p < 0) p = q;
879 if (0 <= p && p == q) {
880 p_value = p;
881 p_value_known = 1;
882 }
883 }
a4acb0eb 884 if (is_dev_null(first)) {
19c58fb8
LT
885 patch->is_new = 1;
886 patch->is_delete = 0;
5a12c886 887 name = find_name_traditional(second, NULL, p_value);
19c58fb8 888 patch->new_name = name;
a4acb0eb 889 } else if (is_dev_null(second)) {
19c58fb8
LT
890 patch->is_new = 0;
891 patch->is_delete = 1;
5a12c886 892 name = find_name_traditional(first, NULL, p_value);
19c58fb8 893 patch->old_name = name;
a4acb0eb 894 } else {
2901bbe5
JH
895 char *first_name;
896 first_name = find_name_traditional(first, NULL, p_value);
897 name = find_name_traditional(second, first_name, p_value);
898 free(first_name);
c4593faf
JH
899 if (has_epoch_timestamp(first)) {
900 patch->is_new = 1;
901 patch->is_delete = 0;
902 patch->new_name = name;
903 } else if (has_epoch_timestamp(second)) {
904 patch->is_new = 0;
905 patch->is_delete = 1;
906 patch->old_name = name;
907 } else {
2901bbe5 908 patch->old_name = name;
212eb96a 909 patch->new_name = null_strdup(name);
c4593faf 910 }
a4acb0eb
LT
911 }
912 if (!name)
3638eb43 913 die(_("unable to find filename in patch at line %d"), linenr);
a4acb0eb
LT
914}
915
19c58fb8 916static int gitdiff_hdrend(const char *line, struct patch *patch)
a4acb0eb
LT
917{
918 return -1;
919}
920
1e3f6b6e
LT
921/*
922 * We're anal about diff header consistency, to make
923 * sure that we don't end up having strange ambiguous
924 * patches floating around.
925 *
926 * As a result, gitdiff_{old|new}name() will check
927 * their names against any previous information, just
928 * to make sure..
929 */
4c5197d1
NTND
930#define DIFF_OLD_NAME 0
931#define DIFF_NEW_NAME 1
932
933static char *gitdiff_verify_name(const char *line, int isnull, char *orig_name, int side)
1e3f6b6e 934{
1e3f6b6e 935 if (!orig_name && !isnull)
79ee194e 936 return find_name(line, NULL, p_value, TERM_TAB);
1e3f6b6e 937
1e3f6b6e 938 if (orig_name) {
22943f1a
JH
939 int len;
940 const char *name;
941 char *another;
1e3f6b6e
LT
942 name = orig_name;
943 len = strlen(name);
944 if (isnull)
3638eb43 945 die(_("git apply: bad git-diff - expected /dev/null, got %s on line %d"), name, linenr);
79ee194e 946 another = find_name(line, NULL, p_value, TERM_TAB);
da915939 947 if (!another || memcmp(another, name, len + 1))
4c5197d1
NTND
948 die((side == DIFF_NEW_NAME) ?
949 _("git apply: bad git-diff - inconsistent new filename on line %d") :
950 _("git apply: bad git-diff - inconsistent old filename on line %d"), linenr);
22943f1a 951 free(another);
1e3f6b6e
LT
952 return orig_name;
953 }
22943f1a
JH
954 else {
955 /* expect "/dev/null" */
956 if (memcmp("/dev/null", line, 9) || line[9] != '\n')
3638eb43 957 die(_("git apply: bad git-diff - expected /dev/null on line %d"), linenr);
22943f1a
JH
958 return NULL;
959 }
1e3f6b6e
LT
960}
961
19c58fb8 962static int gitdiff_oldname(const char *line, struct patch *patch)
a4acb0eb 963{
2901bbe5 964 char *orig = patch->old_name;
4c5197d1
NTND
965 patch->old_name = gitdiff_verify_name(line, patch->is_new, patch->old_name,
966 DIFF_OLD_NAME);
2901bbe5
JH
967 if (orig != patch->old_name)
968 free(orig);
a4acb0eb
LT
969 return 0;
970}
971
19c58fb8 972static int gitdiff_newname(const char *line, struct patch *patch)
a4acb0eb 973{
2901bbe5 974 char *orig = patch->new_name;
4c5197d1
NTND
975 patch->new_name = gitdiff_verify_name(line, patch->is_delete, patch->new_name,
976 DIFF_NEW_NAME);
2901bbe5
JH
977 if (orig != patch->new_name)
978 free(orig);
a4acb0eb
LT
979 return 0;
980}
981
19c58fb8 982static int gitdiff_oldmode(const char *line, struct patch *patch)
a4acb0eb 983{
19c58fb8 984 patch->old_mode = strtoul(line, NULL, 8);
a4acb0eb
LT
985 return 0;
986}
987
19c58fb8 988static int gitdiff_newmode(const char *line, struct patch *patch)
a4acb0eb 989{
19c58fb8 990 patch->new_mode = strtoul(line, NULL, 8);
a4acb0eb
LT
991 return 0;
992}
993
19c58fb8 994static int gitdiff_delete(const char *line, struct patch *patch)
a4acb0eb 995{
19c58fb8 996 patch->is_delete = 1;
2901bbe5
JH
997 free(patch->old_name);
998 patch->old_name = null_strdup(patch->def_name);
19c58fb8 999 return gitdiff_oldmode(line, patch);
a4acb0eb
LT
1000}
1001
19c58fb8 1002static int gitdiff_newfile(const char *line, struct patch *patch)
a4acb0eb 1003{
19c58fb8 1004 patch->is_new = 1;
2901bbe5
JH
1005 free(patch->new_name);
1006 patch->new_name = null_strdup(patch->def_name);
19c58fb8 1007 return gitdiff_newmode(line, patch);
a4acb0eb
LT
1008}
1009
19c58fb8 1010static int gitdiff_copysrc(const char *line, struct patch *patch)
a4acb0eb 1011{
19c58fb8 1012 patch->is_copy = 1;
2901bbe5 1013 free(patch->old_name);
cefd43b7 1014 patch->old_name = find_name(line, NULL, p_value ? p_value - 1 : 0, 0);
a4acb0eb
LT
1015 return 0;
1016}
1017
19c58fb8 1018static int gitdiff_copydst(const char *line, struct patch *patch)
a4acb0eb 1019{
19c58fb8 1020 patch->is_copy = 1;
2901bbe5 1021 free(patch->new_name);
cefd43b7 1022 patch->new_name = find_name(line, NULL, p_value ? p_value - 1 : 0, 0);
a4acb0eb
LT
1023 return 0;
1024}
1025
19c58fb8 1026static int gitdiff_renamesrc(const char *line, struct patch *patch)
a4acb0eb 1027{
19c58fb8 1028 patch->is_rename = 1;
2901bbe5 1029 free(patch->old_name);
cefd43b7 1030 patch->old_name = find_name(line, NULL, p_value ? p_value - 1 : 0, 0);
a4acb0eb
LT
1031 return 0;
1032}
1033
19c58fb8 1034static int gitdiff_renamedst(const char *line, struct patch *patch)
a4acb0eb 1035{
19c58fb8 1036 patch->is_rename = 1;
2901bbe5 1037 free(patch->new_name);
cefd43b7 1038 patch->new_name = find_name(line, NULL, p_value ? p_value - 1 : 0, 0);
a4acb0eb
LT
1039 return 0;
1040}
1041
19c58fb8 1042static int gitdiff_similarity(const char *line, struct patch *patch)
a4acb0eb 1043{
afcb6ac8
JK
1044 unsigned long val = strtoul(line, NULL, 10);
1045 if (val <= 100)
1046 patch->score = val;
a4acb0eb 1047 return 0;
c1bb9350
LT
1048}
1049
70aadac0
JH
1050static int gitdiff_dissimilarity(const char *line, struct patch *patch)
1051{
afcb6ac8
JK
1052 unsigned long val = strtoul(line, NULL, 10);
1053 if (val <= 100)
1054 patch->score = val;
70aadac0
JH
1055 return 0;
1056}
1057
2cf67f1e
JH
1058static int gitdiff_index(const char *line, struct patch *patch)
1059{
81bf96bb
JH
1060 /*
1061 * index line is N hexadecimal, "..", N hexadecimal,
2cf67f1e
JH
1062 * and optional space with octal mode.
1063 */
1064 const char *ptr, *eol;
1065 int len;
1066
1067 ptr = strchr(line, '.');
9add69b1 1068 if (!ptr || ptr[1] != '.' || 40 < ptr - line)
2cf67f1e
JH
1069 return 0;
1070 len = ptr - line;
1071 memcpy(patch->old_sha1_prefix, line, len);
1072 patch->old_sha1_prefix[len] = 0;
1073
1074 line = ptr + 2;
1075 ptr = strchr(line, ' ');
31bb6d37 1076 eol = strchrnul(line, '\n');
2cf67f1e
JH
1077
1078 if (!ptr || eol < ptr)
1079 ptr = eol;
1080 len = ptr - line;
1081
9add69b1 1082 if (40 < len)
2cf67f1e
JH
1083 return 0;
1084 memcpy(patch->new_sha1_prefix, line, len);
1085 patch->new_sha1_prefix[len] = 0;
1086 if (*ptr == ' ')
1f7903a3 1087 patch->old_mode = strtoul(ptr+1, NULL, 8);
2cf67f1e
JH
1088 return 0;
1089}
1090
9a4a100e
LT
1091/*
1092 * This is normal for a diff that doesn't change anything: we'll fall through
1093 * into the next diff. Tell the parser to break out.
1094 */
19c58fb8 1095static int gitdiff_unrecognized(const char *line, struct patch *patch)
9a4a100e
LT
1096{
1097 return -1;
1098}
1099
6a2abdc1
JH
1100/*
1101 * Skip p_value leading components from "line"; as we do not accept
1102 * absolute paths, return NULL in that case.
1103 */
1104static const char *skip_tree_prefix(const char *line, int llen)
22943f1a 1105{
6a2abdc1 1106 int nslash;
22943f1a
JH
1107 int i;
1108
6a2abdc1
JH
1109 if (!p_value)
1110 return (llen && line[0] == '/') ? NULL : line;
1111
1112 nslash = p_value;
22943f1a
JH
1113 for (i = 0; i < llen; i++) {
1114 int ch = line[i];
ec7fc0b1 1115 if (ch == '/' && --nslash <= 0)
6a2abdc1 1116 return (i == 0) ? NULL : &line[i + 1];
22943f1a
JH
1117 }
1118 return NULL;
1119}
1120
81bf96bb
JH
1121/*
1122 * This is to extract the same name that appears on "diff --git"
22943f1a
JH
1123 * line. We do not find and return anything if it is a rename
1124 * patch, and it is OK because we will find the name elsewhere.
1125 * We need to reliably find name only when it is mode-change only,
1126 * creation or deletion of an empty file. In any of these cases,
1127 * both sides are the same name under a/ and b/ respectively.
1128 */
26693ba8 1129static char *git_header_name(const char *line, int llen)
5041aa70 1130{
22943f1a
JH
1131 const char *name;
1132 const char *second = NULL;
cefd43b7 1133 size_t len, line_len;
5041aa70 1134
22943f1a
JH
1135 line += strlen("diff --git ");
1136 llen -= strlen("diff --git ");
1137
1138 if (*line == '"') {
1139 const char *cp;
f285a2d7
BC
1140 struct strbuf first = STRBUF_INIT;
1141 struct strbuf sp = STRBUF_INIT;
7fb1011e
PH
1142
1143 if (unquote_c_style(&first, line, &second))
1144 goto free_and_fail1;
22943f1a 1145
6a2abdc1
JH
1146 /* strip the a/b prefix including trailing slash */
1147 cp = skip_tree_prefix(first.buf, first.len);
1148 if (!cp)
7fb1011e 1149 goto free_and_fail1;
6a2abdc1 1150 strbuf_remove(&first, 0, cp - first.buf);
22943f1a 1151
81bf96bb
JH
1152 /*
1153 * second points at one past closing dq of name.
22943f1a
JH
1154 * find the second name.
1155 */
1156 while ((second < line + llen) && isspace(*second))
1157 second++;
1158
1159 if (line + llen <= second)
7fb1011e 1160 goto free_and_fail1;
22943f1a 1161 if (*second == '"') {
7fb1011e
PH
1162 if (unquote_c_style(&sp, second, NULL))
1163 goto free_and_fail1;
6a2abdc1
JH
1164 cp = skip_tree_prefix(sp.buf, sp.len);
1165 if (!cp)
7fb1011e 1166 goto free_and_fail1;
22943f1a 1167 /* They must match, otherwise ignore */
6a2abdc1 1168 if (strcmp(cp, first.buf))
7fb1011e
PH
1169 goto free_and_fail1;
1170 strbuf_release(&sp);
b315c5c0 1171 return strbuf_detach(&first, NULL);
22943f1a
JH
1172 }
1173
1174 /* unquoted second */
6a2abdc1
JH
1175 cp = skip_tree_prefix(second, line + llen - second);
1176 if (!cp)
7fb1011e 1177 goto free_and_fail1;
6a2abdc1 1178 if (line + llen - cp != first.len ||
7fb1011e
PH
1179 memcmp(first.buf, cp, first.len))
1180 goto free_and_fail1;
b315c5c0 1181 return strbuf_detach(&first, NULL);
7fb1011e
PH
1182
1183 free_and_fail1:
1184 strbuf_release(&first);
1185 strbuf_release(&sp);
1186 return NULL;
5041aa70
LT
1187 }
1188
22943f1a 1189 /* unquoted first name */
6a2abdc1
JH
1190 name = skip_tree_prefix(line, llen);
1191 if (!name)
5041aa70 1192 return NULL;
22943f1a 1193
81bf96bb
JH
1194 /*
1195 * since the first name is unquoted, a dq if exists must be
22943f1a
JH
1196 * the beginning of the second name.
1197 */
1198 for (second = name; second < line + llen; second++) {
1199 if (*second == '"') {
f285a2d7 1200 struct strbuf sp = STRBUF_INIT;
22943f1a 1201 const char *np;
7fb1011e 1202
7fb1011e
PH
1203 if (unquote_c_style(&sp, second, NULL))
1204 goto free_and_fail2;
1205
6a2abdc1
JH
1206 np = skip_tree_prefix(sp.buf, sp.len);
1207 if (!np)
7fb1011e 1208 goto free_and_fail2;
7fb1011e
PH
1209
1210 len = sp.buf + sp.len - np;
1211 if (len < second - name &&
22943f1a
JH
1212 !strncmp(np, name, len) &&
1213 isspace(name[len])) {
1214 /* Good */
7fb1011e 1215 strbuf_remove(&sp, 0, np - sp.buf);
b315c5c0 1216 return strbuf_detach(&sp, NULL);
22943f1a 1217 }
7fb1011e
PH
1218
1219 free_and_fail2:
1220 strbuf_release(&sp);
1221 return NULL;
22943f1a
JH
1222 }
1223 }
1224
5041aa70
LT
1225 /*
1226 * Accept a name only if it shows up twice, exactly the same
1227 * form.
1228 */
cefd43b7
FC
1229 second = strchr(name, '\n');
1230 if (!second)
1231 return NULL;
1232 line_len = second - name;
5041aa70 1233 for (len = 0 ; ; len++) {
dd305c84 1234 switch (name[len]) {
5041aa70
LT
1235 default:
1236 continue;
1237 case '\n':
e70a165d 1238 return NULL;
5041aa70 1239 case '\t': case ' ':
6a2abdc1
JH
1240 /*
1241 * Is this the separator between the preimage
1242 * and the postimage pathname? Again, we are
1243 * only interested in the case where there is
1244 * no rename, as this is only to set def_name
1245 * and a rename patch has the names elsewhere
1246 * in an unambiguous form.
1247 */
1248 if (!name[len + 1])
1249 return NULL; /* no postimage name */
1250 second = skip_tree_prefix(name + len + 1,
1251 line_len - (len + 1));
cefd43b7
FC
1252 if (!second)
1253 return NULL;
6a2abdc1
JH
1254 /*
1255 * Does len bytes starting at "name" and "second"
1256 * (that are separated by one HT or SP we just
1257 * found) exactly match?
1258 */
1259 if (second[len] == '\n' && !strncmp(name, second, len))
182af834 1260 return xmemdupz(name, len);
5041aa70
LT
1261 }
1262 }
5041aa70
LT
1263}
1264
c1bb9350 1265/* Verify that we recognize the lines following a git header */
26693ba8 1266static int parse_git_header(const char *line, int len, unsigned int size, struct patch *patch)
c1bb9350 1267{
a4acb0eb
LT
1268 unsigned long offset;
1269
1270 /* A git diff has explicit new/delete information, so we don't guess */
19c58fb8
LT
1271 patch->is_new = 0;
1272 patch->is_delete = 0;
a4acb0eb 1273
5041aa70
LT
1274 /*
1275 * Some things may not have the old name in the
1276 * rest of the headers anywhere (pure mode changes,
1277 * or removing or adding empty files), so we get
1278 * the default name from the header.
1279 */
22943f1a 1280 patch->def_name = git_header_name(line, len);
969c8775
JK
1281 if (patch->def_name && root) {
1282 char *s = xmalloc(root_len + strlen(patch->def_name) + 1);
1283 strcpy(s, root);
1284 strcpy(s + root_len, patch->def_name);
1285 free(patch->def_name);
1286 patch->def_name = s;
1287 }
5041aa70 1288
a4acb0eb
LT
1289 line += len;
1290 size -= len;
1291 linenr++;
1292 for (offset = len ; size > 0 ; offset += len, size -= len, line += len, linenr++) {
1293 static const struct opentry {
1294 const char *str;
19c58fb8 1295 int (*fn)(const char *, struct patch *);
a4acb0eb
LT
1296 } optable[] = {
1297 { "@@ -", gitdiff_hdrend },
1298 { "--- ", gitdiff_oldname },
1299 { "+++ ", gitdiff_newname },
1300 { "old mode ", gitdiff_oldmode },
1301 { "new mode ", gitdiff_newmode },
1302 { "deleted file mode ", gitdiff_delete },
1303 { "new file mode ", gitdiff_newfile },
1304 { "copy from ", gitdiff_copysrc },
1305 { "copy to ", gitdiff_copydst },
33f4d087
LT
1306 { "rename old ", gitdiff_renamesrc },
1307 { "rename new ", gitdiff_renamedst },
dc938417
LT
1308 { "rename from ", gitdiff_renamesrc },
1309 { "rename to ", gitdiff_renamedst },
a4acb0eb 1310 { "similarity index ", gitdiff_similarity },
70aadac0 1311 { "dissimilarity index ", gitdiff_dissimilarity },
2cf67f1e 1312 { "index ", gitdiff_index },
9a4a100e 1313 { "", gitdiff_unrecognized },
a4acb0eb
LT
1314 };
1315 int i;
c1bb9350 1316
c1bb9350 1317 len = linelen(line, size);
a4acb0eb 1318 if (!len || line[len-1] != '\n')
c1bb9350 1319 break;
b4f2a6ac 1320 for (i = 0; i < ARRAY_SIZE(optable); i++) {
a4acb0eb
LT
1321 const struct opentry *p = optable + i;
1322 int oplen = strlen(p->str);
1323 if (len < oplen || memcmp(p->str, line, oplen))
1324 continue;
19c58fb8 1325 if (p->fn(line + oplen, patch) < 0)
a4acb0eb 1326 return offset;
9a4a100e 1327 break;
a4acb0eb 1328 }
c1bb9350
LT
1329 }
1330
a4acb0eb 1331 return offset;
c1bb9350
LT
1332}
1333
fab2c257 1334static int parse_num(const char *line, unsigned long *p)
46979f56
LT
1335{
1336 char *ptr;
fab2c257
LT
1337
1338 if (!isdigit(*line))
1339 return 0;
1340 *p = strtoul(line, &ptr, 10);
1341 return ptr - line;
1342}
1343
1344static int parse_range(const char *line, int len, int offset, const char *expect,
81bf96bb 1345 unsigned long *p1, unsigned long *p2)
fab2c257 1346{
46979f56
LT
1347 int digits, ex;
1348
1349 if (offset < 0 || offset >= len)
1350 return -1;
1351 line += offset;
1352 len -= offset;
1353
fab2c257
LT
1354 digits = parse_num(line, p1);
1355 if (!digits)
46979f56 1356 return -1;
46979f56
LT
1357
1358 offset += digits;
1359 line += digits;
1360 len -= digits;
1361
c1504628 1362 *p2 = 1;
fab2c257
LT
1363 if (*line == ',') {
1364 digits = parse_num(line+1, p2);
1365 if (!digits)
1366 return -1;
1367
1368 offset += digits+1;
1369 line += digits+1;
1370 len -= digits+1;
1371 }
1372
46979f56
LT
1373 ex = strlen(expect);
1374 if (ex > len)
1375 return -1;
1376 if (memcmp(line, expect, ex))
1377 return -1;
1378
1379 return offset + ex;
1380}
1381
26693ba8 1382static void recount_diff(const char *line, int size, struct fragment *fragment)
c14b9d1e
JS
1383{
1384 int oldlines = 0, newlines = 0, ret = 0;
1385
1386 if (size < 1) {
1387 warning("recount: ignore empty hunk");
1388 return;
1389 }
1390
1391 for (;;) {
1392 int len = linelen(line, size);
1393 size -= len;
1394 line += len;
1395
1396 if (size < 1)
1397 break;
1398
1399 switch (*line) {
1400 case ' ': case '\n':
1401 newlines++;
1402 /* fall through */
1403 case '-':
1404 oldlines++;
1405 continue;
1406 case '+':
1407 newlines++;
1408 continue;
1409 case '\\':
6cf91492 1410 continue;
c14b9d1e 1411 case '@':
59556548 1412 ret = size < 3 || !starts_with(line, "@@ ");
c14b9d1e
JS
1413 break;
1414 case 'd':
59556548 1415 ret = size < 5 || !starts_with(line, "diff ");
c14b9d1e
JS
1416 break;
1417 default:
1418 ret = -1;
1419 break;
1420 }
1421 if (ret) {
3638eb43 1422 warning(_("recount: unexpected line: %.*s"),
c14b9d1e
JS
1423 (int)linelen(line, size), line);
1424 return;
1425 }
1426 break;
1427 }
1428 fragment->oldlines = oldlines;
1429 fragment->newlines = newlines;
1430}
1431
46979f56
LT
1432/*
1433 * Parse a unified diff fragment header of the
1434 * form "@@ -a,b +c,d @@"
1435 */
26693ba8 1436static int parse_fragment_header(const char *line, int len, struct fragment *fragment)
46979f56
LT
1437{
1438 int offset;
1439
1440 if (!len || line[len-1] != '\n')
1441 return -1;
1442
1443 /* Figure out the number of lines in a fragment */
fab2c257
LT
1444 offset = parse_range(line, len, 4, " +", &fragment->oldpos, &fragment->oldlines);
1445 offset = parse_range(line, len, offset, " @@", &fragment->newpos, &fragment->newlines);
46979f56
LT
1446
1447 return offset;
1448}
1449
26693ba8 1450static int find_header(const char *line, unsigned long size, int *hdrsize, struct patch *patch)
c1bb9350
LT
1451{
1452 unsigned long offset, len;
1453
9987d7c5 1454 patch->is_toplevel_relative = 0;
19c58fb8
LT
1455 patch->is_rename = patch->is_copy = 0;
1456 patch->is_new = patch->is_delete = -1;
1457 patch->old_mode = patch->new_mode = 0;
1458 patch->old_name = patch->new_name = NULL;
46979f56 1459 for (offset = 0; size > 0; offset += len, size -= len, line += len, linenr++) {
c1bb9350
LT
1460 unsigned long nextlen;
1461
1462 len = linelen(line, size);
1463 if (!len)
1464 break;
1465
1466 /* Testing this early allows us to take a few shortcuts.. */
1467 if (len < 6)
1468 continue;
46979f56
LT
1469
1470 /*
82e5a82f 1471 * Make sure we don't find any unconnected patch fragments.
46979f56
LT
1472 * That's a sign that we didn't find a header, and that a
1473 * patch has become corrupted/broken up.
1474 */
1475 if (!memcmp("@@ -", line, 4)) {
19c58fb8
LT
1476 struct fragment dummy;
1477 if (parse_fragment_header(line, len, &dummy) < 0)
46979f56 1478 continue;
3638eb43 1479 die(_("patch fragment without header at line %d: %.*s"),
65341411 1480 linenr, (int)len-1, line);
46979f56
LT
1481 }
1482
c1bb9350
LT
1483 if (size < len + 6)
1484 break;
1485
1486 /*
1487 * Git patch? It might not have a real patch, just a rename
1488 * or mode change, so we handle that specially
1489 */
1490 if (!memcmp("diff --git ", line, 11)) {
19c58fb8 1491 int git_hdr_len = parse_git_header(line, len, size, patch);
206de27e 1492 if (git_hdr_len <= len)
c1bb9350 1493 continue;
b7e8039a
LT
1494 if (!patch->old_name && !patch->new_name) {
1495 if (!patch->def_name)
3638eb43
NTND
1496 die(Q_("git diff header lacks filename information when removing "
1497 "%d leading pathname component (line %d)",
1498 "git diff header lacks filename information when removing "
1499 "%d leading pathname components (line %d)",
1500 p_value),
1501 p_value, linenr);
2901bbe5
JH
1502 patch->old_name = xstrdup(patch->def_name);
1503 patch->new_name = xstrdup(patch->def_name);
b7e8039a 1504 }
2c93286a
JM
1505 if (!patch->is_delete && !patch->new_name)
1506 die("git diff header lacks filename information "
1507 "(line %d)", linenr);
9987d7c5 1508 patch->is_toplevel_relative = 1;
a4acb0eb 1509 *hdrsize = git_hdr_len;
c1bb9350
LT
1510 return offset;
1511 }
1512
81bf96bb 1513 /* --- followed by +++ ? */
c1bb9350
LT
1514 if (memcmp("--- ", line, 4) || memcmp("+++ ", line + len, 4))
1515 continue;
1516
1517 /*
1518 * We only accept unified patches, so we want it to
1519 * at least have "@@ -a,b +c,d @@\n", which is 14 chars
81bf96bb 1520 * minimum ("@@ -0,0 +1 @@\n" is the shortest).
c1bb9350
LT
1521 */
1522 nextlen = linelen(line + len, size - len);
1523 if (size < nextlen + 14 || memcmp("@@ -", line + len + nextlen, 4))
1524 continue;
1525
1526 /* Ok, we'll consider it a patch */
19c58fb8 1527 parse_traditional_patch(line, line+len, patch);
c1bb9350 1528 *hdrsize = len + nextlen;
46979f56 1529 linenr += 2;
c1bb9350
LT
1530 return offset;
1531 }
1532 return -1;
1533}
1534
92a1747e 1535static void record_ws_error(unsigned result, const char *line, int len, int linenr)
d0c25035 1536{
c1795bb0 1537 char *err;
92a1747e 1538
c1795bb0
WC
1539 if (!result)
1540 return;
d0c25035 1541
d0c25035
JH
1542 whitespace_error++;
1543 if (squelch_whitespace_errors &&
1544 squelch_whitespace_errors < whitespace_error)
92a1747e
JH
1545 return;
1546
1547 err = whitespace_error_string(result);
1548 fprintf(stderr, "%s:%d: %s.\n%.*s\n",
1549 patch_input_file, linenr, err, len, line);
1550 free(err);
1551}
1552
1553static void check_whitespace(const char *line, int len, unsigned ws_rule)
1554{
1555 unsigned result = ws_check(line + 1, len - 1, ws_rule);
1556
1557 record_ws_error(result, line + 1, len - 2, linenr);
d0c25035
JH
1558}
1559
c1bb9350 1560/*
4be60962
JH
1561 * Parse a unified diff. Note that this really needs to parse each
1562 * fragment separately, since the only way to know the difference
1563 * between a "---" that is part of a patch, and a "---" that starts
1564 * the next patch is to look at the line counts..
c1bb9350 1565 */
26693ba8 1566static int parse_fragment(const char *line, unsigned long size,
81bf96bb 1567 struct patch *patch, struct fragment *fragment)
c1bb9350 1568{
3f40315a 1569 int added, deleted;
c1bb9350 1570 int len = linelen(line, size), offset;
30996652 1571 unsigned long oldlines, newlines;
47495887 1572 unsigned long leading, trailing;
c1bb9350 1573
19c58fb8 1574 offset = parse_fragment_header(line, len, fragment);
c1bb9350
LT
1575 if (offset < 0)
1576 return -1;
c14b9d1e
JS
1577 if (offset > 0 && patch->recount)
1578 recount_diff(line + offset, size - offset, fragment);
19c58fb8
LT
1579 oldlines = fragment->oldlines;
1580 newlines = fragment->newlines;
47495887
EB
1581 leading = 0;
1582 trailing = 0;
c1bb9350
LT
1583
1584 /* Parse the thing.. */
1585 line += len;
1586 size -= len;
46979f56 1587 linenr++;
3f40315a 1588 added = deleted = 0;
4be60962
JH
1589 for (offset = len;
1590 0 < size;
1591 offset += len, size -= len, line += len, linenr++) {
c1bb9350
LT
1592 if (!oldlines && !newlines)
1593 break;
1594 len = linelen(line, size);
1595 if (!len || line[len-1] != '\n')
1596 return -1;
1597 switch (*line) {
1598 default:
1599 return -1;
b507b465 1600 case '\n': /* newer GNU diff, an empty context line */
c1bb9350
LT
1601 case ' ':
1602 oldlines--;
1603 newlines--;
47495887
EB
1604 if (!deleted && !added)
1605 leading++;
1606 trailing++;
c1bb9350
LT
1607 break;
1608 case '-':
5fda48d6 1609 if (apply_in_reverse &&
81bf96bb 1610 ws_error_action != nowarn_ws_error)
cf1b7869 1611 check_whitespace(line, len, patch->ws_rule);
3f40315a 1612 deleted++;
c1bb9350 1613 oldlines--;
47495887 1614 trailing = 0;
c1bb9350
LT
1615 break;
1616 case '+':
5fda48d6 1617 if (!apply_in_reverse &&
81bf96bb 1618 ws_error_action != nowarn_ws_error)
cf1b7869 1619 check_whitespace(line, len, patch->ws_rule);
3f40315a 1620 added++;
c1bb9350 1621 newlines--;
47495887 1622 trailing = 0;
c1bb9350 1623 break;
433ef8a2 1624
81bf96bb
JH
1625 /*
1626 * We allow "\ No newline at end of file". Depending
433ef8a2
FK
1627 * on locale settings when the patch was produced we
1628 * don't know what this line looks like. The only
56d33b11
JH
1629 * thing we do know is that it begins with "\ ".
1630 * Checking for 12 is just for sanity check -- any
1631 * l10n of "\ No newline..." is at least that long.
1632 */
fab2c257 1633 case '\\':
433ef8a2 1634 if (len < 12 || memcmp(line, "\\ ", 2))
3cca928d 1635 return -1;
fab2c257 1636 break;
c1bb9350
LT
1637 }
1638 }
c1504628
LT
1639 if (oldlines || newlines)
1640 return -1;
47495887
EB
1641 fragment->leading = leading;
1642 fragment->trailing = trailing;
1643
81bf96bb
JH
1644 /*
1645 * If a fragment ends with an incomplete line, we failed to include
8b64647d
JH
1646 * it in the above loop because we hit oldlines == newlines == 0
1647 * before seeing it.
1648 */
433ef8a2 1649 if (12 < size && !memcmp(line, "\\ ", 2))
8b64647d
JH
1650 offset += linelen(line, size);
1651
3f40315a
LT
1652 patch->lines_added += added;
1653 patch->lines_deleted += deleted;
4be60962
JH
1654
1655 if (0 < patch->is_new && oldlines)
3638eb43 1656 return error(_("new file depends on old contents"));
4be60962 1657 if (0 < patch->is_delete && newlines)
3638eb43 1658 return error(_("deleted file still has contents"));
c1bb9350
LT
1659 return offset;
1660}
1661
92737a22
JH
1662/*
1663 * We have seen "diff --git a/... b/..." header (or a traditional patch
1664 * header). Read hunks that belong to this patch into fragments and hang
1665 * them to the given patch structure.
1666 *
1667 * The (fragment->patch, fragment->size) pair points into the memory given
1668 * by the caller, not a copy, when we return.
1669 */
26693ba8 1670static int parse_single_patch(const char *line, unsigned long size, struct patch *patch)
c1bb9350
LT
1671{
1672 unsigned long offset = 0;
4be60962 1673 unsigned long oldlines = 0, newlines = 0, context = 0;
19c58fb8 1674 struct fragment **fragp = &patch->fragments;
c1bb9350
LT
1675
1676 while (size > 4 && !memcmp(line, "@@ -", 4)) {
19c58fb8
LT
1677 struct fragment *fragment;
1678 int len;
1679
90321c10 1680 fragment = xcalloc(1, sizeof(*fragment));
77b15bbd 1681 fragment->linenr = linenr;
19c58fb8 1682 len = parse_fragment(line, size, patch, fragment);
c1bb9350 1683 if (len <= 0)
3638eb43 1684 die(_("corrupt patch at line %d"), linenr);
19c58fb8
LT
1685 fragment->patch = line;
1686 fragment->size = len;
4be60962
JH
1687 oldlines += fragment->oldlines;
1688 newlines += fragment->newlines;
1689 context += fragment->leading + fragment->trailing;
19c58fb8
LT
1690
1691 *fragp = fragment;
1692 fragp = &fragment->next;
c1bb9350
LT
1693
1694 offset += len;
1695 line += len;
1696 size -= len;
1697 }
4be60962
JH
1698
1699 /*
1700 * If something was removed (i.e. we have old-lines) it cannot
1701 * be creation, and if something was added it cannot be
1702 * deletion. However, the reverse is not true; --unified=0
1703 * patches that only add are not necessarily creation even
1704 * though they do not have any old lines, and ones that only
1705 * delete are not necessarily deletion.
1706 *
1707 * Unfortunately, a real creation/deletion patch do _not_ have
1708 * any context line by definition, so we cannot safely tell it
1709 * apart with --unified=0 insanity. At least if the patch has
1710 * more than one hunk it is not creation or deletion.
1711 */
1712 if (patch->is_new < 0 &&
1713 (oldlines || (patch->fragments && patch->fragments->next)))
1714 patch->is_new = 0;
1715 if (patch->is_delete < 0 &&
1716 (newlines || (patch->fragments && patch->fragments->next)))
1717 patch->is_delete = 0;
4be60962
JH
1718
1719 if (0 < patch->is_new && oldlines)
3638eb43 1720 die(_("new file %s depends on old contents"), patch->new_name);
4be60962 1721 if (0 < patch->is_delete && newlines)
3638eb43 1722 die(_("deleted file %s still has contents"), patch->old_name);
4be60962 1723 if (!patch->is_delete && !newlines && context)
3638eb43
NTND
1724 fprintf_ln(stderr,
1725 _("** warning: "
1726 "file %s becomes empty but is not deleted"),
1727 patch->new_name);
4be60962 1728
c1bb9350
LT
1729 return offset;
1730}
1731
1fea629f
LT
1732static inline int metadata_changes(struct patch *patch)
1733{
1734 return patch->is_rename > 0 ||
1735 patch->is_copy > 0 ||
1736 patch->is_new > 0 ||
1737 patch->is_delete ||
1738 (patch->old_mode && patch->new_mode &&
1739 patch->old_mode != patch->new_mode);
1740}
1741
3cd4f5e8
JH
1742static char *inflate_it(const void *data, unsigned long size,
1743 unsigned long inflated_size)
051308f6 1744{
ef49a7a0 1745 git_zstream stream;
3cd4f5e8
JH
1746 void *out;
1747 int st;
1748
1749 memset(&stream, 0, sizeof(stream));
1750
1751 stream.next_in = (unsigned char *)data;
1752 stream.avail_in = size;
1753 stream.next_out = out = xmalloc(inflated_size);
1754 stream.avail_out = inflated_size;
39c68542
LT
1755 git_inflate_init(&stream);
1756 st = git_inflate(&stream, Z_FINISH);
1757 git_inflate_end(&stream);
3cd4f5e8
JH
1758 if ((st != Z_STREAM_END) || stream.total_out != inflated_size) {
1759 free(out);
1760 return NULL;
1761 }
1762 return out;
1763}
1764
92737a22
JH
1765/*
1766 * Read a binary hunk and return a new fragment; fragment->patch
1767 * points at an allocated memory that the caller must free, so
1768 * it is marked as "->free_patch = 1".
1769 */
3cd4f5e8
JH
1770static struct fragment *parse_binary_hunk(char **buf_p,
1771 unsigned long *sz_p,
1772 int *status_p,
1773 int *used_p)
1774{
81bf96bb
JH
1775 /*
1776 * Expect a line that begins with binary patch method ("literal"
3cd4f5e8
JH
1777 * or "delta"), followed by the length of data before deflating.
1778 * a sequence of 'length-byte' followed by base-85 encoded data
1779 * should follow, terminated by a newline.
051308f6
JH
1780 *
1781 * Each 5-byte sequence of base-85 encodes up to 4 bytes,
1782 * and we would limit the patch line to 66 characters,
1783 * so one line can fit up to 13 groups that would decode
1784 * to 52 bytes max. The length byte 'A'-'Z' corresponds
1785 * to 1-26 bytes, and 'a'-'z' corresponds to 27-52 bytes.
051308f6
JH
1786 */
1787 int llen, used;
3cd4f5e8
JH
1788 unsigned long size = *sz_p;
1789 char *buffer = *buf_p;
1790 int patch_method;
1791 unsigned long origlen;
0660626c 1792 char *data = NULL;
3cd4f5e8
JH
1793 int hunk_size = 0;
1794 struct fragment *frag;
051308f6 1795
0660626c
JH
1796 llen = linelen(buffer, size);
1797 used = llen;
3cd4f5e8
JH
1798
1799 *status_p = 0;
0660626c 1800
59556548 1801 if (starts_with(buffer, "delta ")) {
3cd4f5e8
JH
1802 patch_method = BINARY_DELTA_DEFLATED;
1803 origlen = strtoul(buffer + 6, NULL, 10);
0660626c 1804 }
59556548 1805 else if (starts_with(buffer, "literal ")) {
3cd4f5e8
JH
1806 patch_method = BINARY_LITERAL_DEFLATED;
1807 origlen = strtoul(buffer + 8, NULL, 10);
0660626c
JH
1808 }
1809 else
3cd4f5e8
JH
1810 return NULL;
1811
1812 linenr++;
0660626c 1813 buffer += llen;
051308f6
JH
1814 while (1) {
1815 int byte_length, max_byte_length, newsize;
1816 llen = linelen(buffer, size);
1817 used += llen;
1818 linenr++;
03eb8f8a
JH
1819 if (llen == 1) {
1820 /* consume the blank line */
1821 buffer++;
1822 size--;
051308f6 1823 break;
03eb8f8a 1824 }
81bf96bb
JH
1825 /*
1826 * Minimum line is "A00000\n" which is 7-byte long,
051308f6
JH
1827 * and the line length must be multiple of 5 plus 2.
1828 */
1829 if ((llen < 7) || (llen-2) % 5)
1830 goto corrupt;
1831 max_byte_length = (llen - 2) / 5 * 4;
1832 byte_length = *buffer;
1833 if ('A' <= byte_length && byte_length <= 'Z')
1834 byte_length = byte_length - 'A' + 1;
1835 else if ('a' <= byte_length && byte_length <= 'z')
1836 byte_length = byte_length - 'a' + 27;
1837 else
1838 goto corrupt;
1839 /* if the input length was not multiple of 4, we would
1840 * have filler at the end but the filler should never
1841 * exceed 3 bytes
1842 */
1843 if (max_byte_length < byte_length ||
1844 byte_length <= max_byte_length - 4)
1845 goto corrupt;
3cd4f5e8 1846 newsize = hunk_size + byte_length;
0660626c 1847 data = xrealloc(data, newsize);
3cd4f5e8 1848 if (decode_85(data + hunk_size, buffer + 1, byte_length))
051308f6 1849 goto corrupt;
3cd4f5e8 1850 hunk_size = newsize;
051308f6
JH
1851 buffer += llen;
1852 size -= llen;
1853 }
3cd4f5e8
JH
1854
1855 frag = xcalloc(1, sizeof(*frag));
1856 frag->patch = inflate_it(data, hunk_size, origlen);
6fe53908 1857 frag->free_patch = 1;
3cd4f5e8
JH
1858 if (!frag->patch)
1859 goto corrupt;
1860 free(data);
1861 frag->size = origlen;
1862 *buf_p = buffer;
1863 *sz_p = size;
1864 *used_p = used;
1865 frag->binary_patch_method = patch_method;
1866 return frag;
1867
051308f6 1868 corrupt:
4cac42b1 1869 free(data);
3cd4f5e8 1870 *status_p = -1;
3638eb43 1871 error(_("corrupt binary patch at line %d: %.*s"),
3cd4f5e8
JH
1872 linenr-1, llen-1, buffer);
1873 return NULL;
1874}
1875
1876static int parse_binary(char *buffer, unsigned long size, struct patch *patch)
1877{
81bf96bb
JH
1878 /*
1879 * We have read "GIT binary patch\n"; what follows is a line
3cd4f5e8
JH
1880 * that says the patch method (currently, either "literal" or
1881 * "delta") and the length of data before deflating; a
1882 * sequence of 'length-byte' followed by base-85 encoded data
1883 * follows.
1884 *
1885 * When a binary patch is reversible, there is another binary
1886 * hunk in the same format, starting with patch method (either
1887 * "literal" or "delta") with the length of data, and a sequence
1888 * of length-byte + base-85 encoded data, terminated with another
1889 * empty line. This data, when applied to the postimage, produces
1890 * the preimage.
1891 */
1892 struct fragment *forward;
1893 struct fragment *reverse;
1894 int status;
1895 int used, used_1;
1896
1897 forward = parse_binary_hunk(&buffer, &size, &status, &used);
1898 if (!forward && !status)
1899 /* there has to be one hunk (forward hunk) */
3638eb43 1900 return error(_("unrecognized binary patch at line %d"), linenr-1);
3cd4f5e8
JH
1901 if (status)
1902 /* otherwise we already gave an error message */
1903 return status;
1904
1905 reverse = parse_binary_hunk(&buffer, &size, &status, &used_1);
1906 if (reverse)
1907 used += used_1;
1908 else if (status) {
81bf96bb
JH
1909 /*
1910 * Not having reverse hunk is not an error, but having
3cd4f5e8
JH
1911 * a corrupt reverse hunk is.
1912 */
1913 free((void*) forward->patch);
1914 free(forward);
1915 return status;
1916 }
1917 forward->next = reverse;
1918 patch->fragments = forward;
1919 patch->is_binary = 1;
1920 return used;
051308f6
JH
1921}
1922
92737a22 1923/*
41ccfdd9 1924 * Read the patch text in "buffer" that extends for "size" bytes; stop
92737a22
JH
1925 * reading after seeing a single patch (i.e. changes to a single file).
1926 * Create fragments (i.e. patch hunks) and hang them to the given patch.
1927 * Return the number of bytes consumed, so that the caller can call us
1928 * again for the next patch.
1929 */
19c58fb8 1930static int parse_chunk(char *buffer, unsigned long size, struct patch *patch)
c1bb9350
LT
1931{
1932 int hdrsize, patchsize;
19c58fb8 1933 int offset = find_header(buffer, size, &hdrsize, patch);
c1bb9350
LT
1934
1935 if (offset < 0)
1936 return offset;
c1bb9350 1937
cf1b7869
JH
1938 patch->ws_rule = whitespace_rule(patch->new_name
1939 ? patch->new_name
1940 : patch->old_name);
1941
81bf96bb
JH
1942 patchsize = parse_single_patch(buffer + offset + hdrsize,
1943 size - offset - hdrsize, patch);
c1bb9350 1944
92927ed0 1945 if (!patchsize) {
051308f6 1946 static const char git_binary[] = "GIT binary patch\n";
3200d1ae
JH
1947 int hd = hdrsize + offset;
1948 unsigned long llen = linelen(buffer + hd, size - hd);
1949
051308f6
JH
1950 if (llen == sizeof(git_binary) - 1 &&
1951 !memcmp(git_binary, buffer + hd, llen)) {
1952 int used;
1953 linenr++;
1954 used = parse_binary(buffer + hd + llen,
1955 size - hd - llen, patch);
1956 if (used)
1957 patchsize = used + llen;
1958 else
1959 patchsize = 0;
1960 }
1961 else if (!memcmp(" differ\n", buffer + hd + llen - 8, 8)) {
e36f3a8a
EP
1962 static const char *binhdr[] = {
1963 "Binary files ",
1964 "Files ",
1965 NULL,
1966 };
1967 int i;
3200d1ae
JH
1968 for (i = 0; binhdr[i]; i++) {
1969 int len = strlen(binhdr[i]);
1970 if (len < size - hd &&
1971 !memcmp(binhdr[i], buffer + hd, len)) {
051308f6 1972 linenr++;
3200d1ae 1973 patch->is_binary = 1;
051308f6 1974 patchsize = llen;
3200d1ae
JH
1975 break;
1976 }
1977 }
051308f6 1978 }
ff36de08 1979
2b6eef94
JH
1980 /* Empty patch cannot be applied if it is a text patch
1981 * without metadata change. A binary patch appears
1982 * empty to us here.
92927ed0
JH
1983 */
1984 if ((apply || check) &&
2b6eef94 1985 (!patch->is_binary && !metadata_changes(patch)))
3638eb43 1986 die(_("patch with only garbage at line %d"), linenr);
ff36de08 1987 }
1fea629f 1988
c1bb9350
LT
1989 return offset + hdrsize + patchsize;
1990}
1991
e5a94313
JS
1992#define swap(a,b) myswap((a),(b),sizeof(a))
1993
1994#define myswap(a, b, size) do { \
1995 unsigned char mytmp[size]; \
1996 memcpy(mytmp, &a, size); \
1997 memcpy(&a, &b, size); \
1998 memcpy(&b, mytmp, size); \
1999} while (0)
2000
2001static void reverse_patches(struct patch *p)
2002{
2003 for (; p; p = p->next) {
2004 struct fragment *frag = p->fragments;
2005
2006 swap(p->new_name, p->old_name);
2007 swap(p->new_mode, p->old_mode);
2008 swap(p->is_new, p->is_delete);
2009 swap(p->lines_added, p->lines_deleted);
2010 swap(p->old_sha1_prefix, p->new_sha1_prefix);
2011
2012 for (; frag; frag = frag->next) {
2013 swap(frag->newpos, frag->oldpos);
2014 swap(frag->newlines, frag->oldlines);
2015 }
e5a94313
JS
2016 }
2017}
2018
81bf96bb
JH
2019static const char pluses[] =
2020"++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++";
2021static const char minuses[]=
2022"----------------------------------------------------------------------";
3f40315a
LT
2023
2024static void show_stats(struct patch *patch)
2025{
f285a2d7 2026 struct strbuf qname = STRBUF_INIT;
663af342
PH
2027 char *cp = patch->new_name ? patch->new_name : patch->old_name;
2028 int max, add, del;
3f40315a 2029
663af342 2030 quote_c_style(cp, &qname, NULL, 0);
22943f1a 2031
3f40315a
LT
2032 /*
2033 * "scale" the filename
2034 */
3f40315a
LT
2035 max = max_len;
2036 if (max > 50)
2037 max = 50;
663af342
PH
2038
2039 if (qname.len > max) {
2040 cp = strchr(qname.buf + qname.len + 3 - max, '/');
2041 if (!cp)
2042 cp = qname.buf + qname.len + 3 - max;
2043 strbuf_splice(&qname, 0, cp - qname.buf, "...", 3);
2044 }
2045
2046 if (patch->is_binary) {
2047 printf(" %-*s | Bin\n", max, qname.buf);
2048 strbuf_release(&qname);
2049 return;
62917097 2050 }
663af342
PH
2051
2052 printf(" %-*s |", max, qname.buf);
2053 strbuf_release(&qname);
3f40315a
LT
2054
2055 /*
2056 * scale the add/delete
2057 */
663af342 2058 max = max + max_change > 70 ? 70 - max : max_change;
95bedc9e
LT
2059 add = patch->lines_added;
2060 del = patch->lines_deleted;
95bedc9e 2061
69f956e1 2062 if (max_change > 0) {
663af342 2063 int total = ((add + del) * max + max_change / 2) / max_change;
69f956e1
SV
2064 add = (add * max + max_change / 2) / max_change;
2065 del = total - add;
2066 }
663af342
PH
2067 printf("%5d %.*s%.*s\n", patch->lines_added + patch->lines_deleted,
2068 add, pluses, del, minuses);
3f40315a
LT
2069}
2070
c7f9cb14 2071static int read_old_data(struct stat *st, const char *path, struct strbuf *buf)
3cca928d 2072{
3cca928d
LT
2073 switch (st->st_mode & S_IFMT) {
2074 case S_IFLNK:
b11b7e13 2075 if (strbuf_readlink(buf, path, st->st_size) < 0)
3638eb43 2076 return error(_("unable to read symlink %s"), path);
c7f9cb14 2077 return 0;
3cca928d 2078 case S_IFREG:
387e7e19 2079 if (strbuf_read_file(buf, path, st->st_size) != st->st_size)
3638eb43 2080 return error(_("unable to open or read %s"), path);
21e5ad50 2081 convert_to_git(path, buf->buf, buf->len, buf, 0);
c7f9cb14 2082 return 0;
3cca928d
LT
2083 default:
2084 return -1;
2085 }
2086}
2087
86c91f91
GB
2088/*
2089 * Update the preimage, and the common lines in postimage,
2090 * from buffer buf of length len. If postlen is 0 the postimage
2091 * is updated in place, otherwise it's updated on a new buffer
2092 * of length postlen
2093 */
2094
c1beba5b
JH
2095static void update_pre_post_images(struct image *preimage,
2096 struct image *postimage,
2097 char *buf,
86c91f91 2098 size_t len, size_t postlen)
3cca928d 2099{
5de7166d 2100 int i, ctx, reduced;
c1beba5b
JH
2101 char *new, *old, *fixed;
2102 struct image fixed_preimage;
3cca928d 2103
c1beba5b
JH
2104 /*
2105 * Update the preimage with whitespace fixes. Note that we
2106 * are not losing preimage->buf -- apply_one_fragment() will
2107 * free "oldlines".
2108 */
2109 prepare_image(&fixed_preimage, buf, len, 1);
5de7166d
JH
2110 assert(postlen
2111 ? fixed_preimage.nr == preimage->nr
2112 : fixed_preimage.nr <= preimage->nr);
2113 for (i = 0; i < fixed_preimage.nr; i++)
c1beba5b
JH
2114 fixed_preimage.line[i].flag = preimage->line[i].flag;
2115 free(preimage->line_allocated);
2116 *preimage = fixed_preimage;
3cca928d 2117
c1beba5b 2118 /*
86c91f91 2119 * Adjust the common context lines in postimage. This can be
250b3c6c
JH
2120 * done in-place when we are shrinking it with whitespace
2121 * fixing, but needs a new buffer when ignoring whitespace or
2122 * expanding leading tabs to spaces.
2123 *
86c91f91
GB
2124 * We trust the caller to tell us if the update can be done
2125 * in place (postlen==0) or not.
c1beba5b 2126 */
86c91f91
GB
2127 old = postimage->buf;
2128 if (postlen)
2129 new = postimage->buf = xmalloc(postlen);
2130 else
2131 new = old;
c1beba5b 2132 fixed = preimage->buf;
5de7166d
JH
2133
2134 for (i = reduced = ctx = 0; i < postimage->nr; i++) {
c1beba5b
JH
2135 size_t len = postimage->line[i].len;
2136 if (!(postimage->line[i].flag & LINE_COMMON)) {
2137 /* an added line -- no counterparts in preimage */
2138 memmove(new, old, len);
2139 old += len;
2140 new += len;
2141 continue;
3cca928d 2142 }
c1beba5b
JH
2143
2144 /* a common context -- skip it in the original postimage */
2145 old += len;
2146
2147 /* and find the corresponding one in the fixed preimage */
2148 while (ctx < preimage->nr &&
2149 !(preimage->line[ctx].flag & LINE_COMMON)) {
2150 fixed += preimage->line[ctx].len;
2151 ctx++;
2152 }
5de7166d
JH
2153
2154 /*
2155 * preimage is expected to run out, if the caller
2156 * fixed addition of trailing blank lines.
2157 */
2158 if (preimage->nr <= ctx) {
2159 reduced++;
2160 continue;
2161 }
c1beba5b
JH
2162
2163 /* and copy it in, while fixing the line length */
2164 len = preimage->line[ctx].len;
2165 memcpy(new, fixed, len);
2166 new += len;
2167 fixed += len;
2168 postimage->line[i].len = len;
2169 ctx++;
2170 }
2171
2172 /* Fix the length of the whole thing */
2173 postimage->len = new - postimage->buf;
5de7166d 2174 postimage->nr -= reduced;
c1beba5b
JH
2175}
2176
b94f2eda
JH
2177static int match_fragment(struct image *img,
2178 struct image *preimage,
2179 struct image *postimage,
c89fb6b1 2180 unsigned long try,
b94f2eda 2181 int try_lno,
c607aaa2 2182 unsigned ws_rule,
dc41976a 2183 int match_beginning, int match_end)
c89fb6b1 2184{
b94f2eda 2185 int i;
c1beba5b 2186 char *fixed_buf, *buf, *orig, *target;
d511bd33 2187 struct strbuf fixed;
250b3c6c 2188 size_t fixed_len, postlen;
51667147 2189 int preimage_limit;
b94f2eda 2190
51667147
BG
2191 if (preimage->nr + try_lno <= img->nr) {
2192 /*
2193 * The hunk falls within the boundaries of img.
2194 */
2195 preimage_limit = preimage->nr;
2196 if (match_end && (preimage->nr + try_lno != img->nr))
2197 return 0;
2198 } else if (ws_error_action == correct_ws_error &&
0c3ef984 2199 (ws_rule & WS_BLANK_AT_EOF)) {
51667147 2200 /*
0c3ef984
BG
2201 * This hunk extends beyond the end of img, and we are
2202 * removing blank lines at the end of the file. This
2203 * many lines from the beginning of the preimage must
2204 * match with img, and the remainder of the preimage
2205 * must be blank.
51667147
BG
2206 */
2207 preimage_limit = img->nr - try_lno;
2208 } else {
2209 /*
2210 * The hunk extends beyond the end of the img and
2211 * we are not removing blanks at the end, so we
2212 * should reject the hunk at this position.
2213 */
b94f2eda 2214 return 0;
51667147 2215 }
b94f2eda
JH
2216
2217 if (match_beginning && try_lno)
c89fb6b1 2218 return 0;
dc41976a 2219
b94f2eda 2220 /* Quick hash check */
51667147 2221 for (i = 0; i < preimage_limit; i++)
9d158601
JH
2222 if ((img->line[try_lno + i].flag & LINE_PATCHED) ||
2223 (preimage->line[i].hash != img->line[try_lno + i].hash))
b94f2eda
JH
2224 return 0;
2225
51667147
BG
2226 if (preimage_limit == preimage->nr) {
2227 /*
2228 * Do we have an exact match? If we were told to match
2229 * at the end, size must be exactly at try+fragsize,
2230 * otherwise try+fragsize must be still within the preimage,
2231 * and either case, the old piece should match the preimage
2232 * exactly.
2233 */
2234 if ((match_end
2235 ? (try + preimage->len == img->len)
2236 : (try + preimage->len <= img->len)) &&
2237 !memcmp(img->buf + try, preimage->buf, preimage->len))
2238 return 1;
2239 } else {
2240 /*
2241 * The preimage extends beyond the end of img, so
2242 * there cannot be an exact match.
2243 *
2244 * There must be one non-blank context line that match
2245 * a line before the end of img.
2246 */
2247 char *buf_end;
2248
2249 buf = preimage->buf;
2250 buf_end = buf;
2251 for (i = 0; i < preimage_limit; i++)
2252 buf_end += preimage->line[i].len;
2253
2254 for ( ; buf < buf_end; buf++)
2255 if (!isspace(*buf))
2256 break;
2257 if (buf == buf_end)
2258 return 0;
2259 }
dc41976a 2260
86c91f91
GB
2261 /*
2262 * No exact match. If we are ignoring whitespace, run a line-by-line
2263 * fuzzy matching. We collect all the line length information because
2264 * we need it to adjust whitespace if we match.
2265 */
2266 if (ws_ignore_action == ignore_ws_change) {
2267 size_t imgoff = 0;
2268 size_t preoff = 0;
2269 size_t postlen = postimage->len;
51667147
BG
2270 size_t extra_chars;
2271 char *preimage_eof;
2272 char *preimage_end;
2273 for (i = 0; i < preimage_limit; i++) {
86c91f91 2274 size_t prelen = preimage->line[i].len;
0b1fac32 2275 size_t imglen = img->line[try_lno+i].len;
86c91f91 2276
0b1fac32
JH
2277 if (!fuzzy_matchlines(img->buf + try + imgoff, imglen,
2278 preimage->buf + preoff, prelen))
86c91f91
GB
2279 return 0;
2280 if (preimage->line[i].flag & LINE_COMMON)
0b1fac32
JH
2281 postlen += imglen - prelen;
2282 imgoff += imglen;
86c91f91
GB
2283 preoff += prelen;
2284 }
2285
2286 /*
9b25949a
BG
2287 * Ok, the preimage matches with whitespace fuzz.
2288 *
2289 * imgoff now holds the true length of the target that
51667147
BG
2290 * matches the preimage before the end of the file.
2291 *
2292 * Count the number of characters in the preimage that fall
2293 * beyond the end of the file and make sure that all of them
2294 * are whitespace characters. (This can only happen if
2295 * we are removing blank lines at the end of the file.)
86c91f91 2296 */
51667147
BG
2297 buf = preimage_eof = preimage->buf + preoff;
2298 for ( ; i < preimage->nr; i++)
2299 preoff += preimage->line[i].len;
2300 preimage_end = preimage->buf + preoff;
2301 for ( ; buf < preimage_end; buf++)
2302 if (!isspace(*buf))
2303 return 0;
2304
2305 /*
2306 * Update the preimage and the common postimage context
2307 * lines to use the same whitespace as the target.
2308 * If whitespace is missing in the target (i.e.
2309 * if the preimage extends beyond the end of the file),
2310 * use the whitespace from the preimage.
2311 */
2312 extra_chars = preimage_end - preimage_eof;
d511bd33
CW
2313 strbuf_init(&fixed, imgoff + extra_chars);
2314 strbuf_add(&fixed, img->buf + try, imgoff);
2315 strbuf_add(&fixed, preimage_eof, extra_chars);
2316 fixed_buf = strbuf_detach(&fixed, &fixed_len);
86c91f91 2317 update_pre_post_images(preimage, postimage,
d511bd33 2318 fixed_buf, fixed_len, postlen);
86c91f91
GB
2319 return 1;
2320 }
2321
c1beba5b
JH
2322 if (ws_error_action != correct_ws_error)
2323 return 0;
2324
dc41976a 2325 /*
c1beba5b 2326 * The hunk does not apply byte-by-byte, but the hash says
86c91f91
GB
2327 * it might with whitespace fuzz. We haven't been asked to
2328 * ignore whitespace, we were asked to correct whitespace
2329 * errors, so let's try matching after whitespace correction.
51667147
BG
2330 *
2331 * The preimage may extend beyond the end of the file,
2332 * but in this loop we will only handle the part of the
2333 * preimage that falls within the file.
dc41976a 2334 */
d511bd33 2335 strbuf_init(&fixed, preimage->len + 1);
c1beba5b
JH
2336 orig = preimage->buf;
2337 target = img->buf + try;
250b3c6c 2338 postlen = 0;
51667147 2339 for (i = 0; i < preimage_limit; i++) {
c1beba5b
JH
2340 size_t oldlen = preimage->line[i].len;
2341 size_t tgtlen = img->line[try_lno + i].len;
d511bd33
CW
2342 size_t fixstart = fixed.len;
2343 struct strbuf tgtfix;
c1beba5b
JH
2344 int match;
2345
2346 /* Try fixing the line in the preimage */
d511bd33 2347 ws_fix_copy(&fixed, orig, oldlen, ws_rule, NULL);
c1beba5b
JH
2348
2349 /* Try fixing the line in the target */
d511bd33
CW
2350 strbuf_init(&tgtfix, tgtlen);
2351 ws_fix_copy(&tgtfix, target, tgtlen, ws_rule, NULL);
c1beba5b
JH
2352
2353 /*
2354 * If they match, either the preimage was based on
2355 * a version before our tree fixed whitespace breakage,
2356 * or we are lacking a whitespace-fix patch the tree
2357 * the preimage was based on already had (i.e. target
2358 * has whitespace breakage, the preimage doesn't).
2359 * In either case, we are fixing the whitespace breakages
2360 * so we might as well take the fix together with their
2361 * real change.
2362 */
d511bd33
CW
2363 match = (tgtfix.len == fixed.len - fixstart &&
2364 !memcmp(tgtfix.buf, fixed.buf + fixstart,
2365 fixed.len - fixstart));
250b3c6c 2366 postlen += tgtfix.len;
c1beba5b 2367
d511bd33 2368 strbuf_release(&tgtfix);
c1beba5b
JH
2369 if (!match)
2370 goto unmatch_exit;
2371
2372 orig += oldlen;
c1beba5b 2373 target += tgtlen;
3cca928d
LT
2374 }
2375
51667147
BG
2376
2377 /*
2378 * Now handle the lines in the preimage that falls beyond the
2379 * end of the file (if any). They will only match if they are
2380 * empty or only contain whitespace (if WS_BLANK_AT_EOL is
2381 * false).
2382 */
2383 for ( ; i < preimage->nr; i++) {
d511bd33 2384 size_t fixstart = fixed.len; /* start of the fixed preimage */
51667147
BG
2385 size_t oldlen = preimage->line[i].len;
2386 int j;
2387
2388 /* Try fixing the line in the preimage */
d511bd33 2389 ws_fix_copy(&fixed, orig, oldlen, ws_rule, NULL);
51667147 2390
d511bd33
CW
2391 for (j = fixstart; j < fixed.len; j++)
2392 if (!isspace(fixed.buf[j]))
51667147
BG
2393 goto unmatch_exit;
2394
2395 orig += oldlen;
51667147
BG
2396 }
2397
c1beba5b
JH
2398 /*
2399 * Yes, the preimage is based on an older version that still
2400 * has whitespace breakages unfixed, and fixing them makes the
2401 * hunk match. Update the context lines in the postimage.
2402 */
d511bd33 2403 fixed_buf = strbuf_detach(&fixed, &fixed_len);
250b3c6c
JH
2404 if (postlen < postimage->len)
2405 postlen = 0;
c1beba5b 2406 update_pre_post_images(preimage, postimage,
250b3c6c 2407 fixed_buf, fixed_len, postlen);
c1beba5b
JH
2408 return 1;
2409
2410 unmatch_exit:
d511bd33 2411 strbuf_release(&fixed);
dc41976a 2412 return 0;
c89fb6b1
JH
2413}
2414
b94f2eda
JH
2415static int find_pos(struct image *img,
2416 struct image *preimage,
2417 struct image *postimage,
2418 int line,
c607aaa2 2419 unsigned ws_rule,
b94f2eda 2420 int match_beginning, int match_end)
3cca928d 2421{
b94f2eda
JH
2422 int i;
2423 unsigned long backwards, forwards, try;
2424 int backwards_lno, forwards_lno, try_lno;
3cca928d 2425
ecf4c2ec 2426 /*
24ff4d56 2427 * If match_beginning or match_end is specified, there is no
ecf4c2ec
JH
2428 * point starting from a wrong line that will never match and
2429 * wander around and wait for a match at the specified end.
2430 */
2431 if (match_beginning)
2432 line = 0;
2433 else if (match_end)
2434 line = img->nr - preimage->nr;
2435
24ff4d56
BG
2436 /*
2437 * Because the comparison is unsigned, the following test
2438 * will also take care of a negative line number that can
2439 * result when match_end and preimage is larger than the target.
2440 */
2441 if ((size_t) line > img->nr)
52f3c81a
JH
2442 line = img->nr;
2443
b94f2eda
JH
2444 try = 0;
2445 for (i = 0; i < line; i++)
2446 try += img->line[i].len;
3cca928d 2447
6e7c92a9
LT
2448 /*
2449 * There's probably some smart way to do this, but I'll leave
2450 * that to the smart and beautiful people. I'm simple and stupid.
2451 */
b94f2eda
JH
2452 backwards = try;
2453 backwards_lno = line;
2454 forwards = try;
2455 forwards_lno = line;
2456 try_lno = line;
fcb77bc5 2457
6e7c92a9 2458 for (i = 0; ; i++) {
b94f2eda 2459 if (match_fragment(img, preimage, postimage,
c607aaa2 2460 try, try_lno, ws_rule,
b94f2eda
JH
2461 match_beginning, match_end))
2462 return try_lno;
fcb77bc5
JH
2463
2464 again:
b94f2eda 2465 if (backwards_lno == 0 && forwards_lno == img->nr)
fcb77bc5 2466 break;
6e7c92a9 2467
6e7c92a9 2468 if (i & 1) {
b94f2eda 2469 if (backwards_lno == 0) {
fcb77bc5
JH
2470 i++;
2471 goto again;
6e7c92a9 2472 }
b94f2eda
JH
2473 backwards_lno--;
2474 backwards -= img->line[backwards_lno].len;
6e7c92a9 2475 try = backwards;
b94f2eda 2476 try_lno = backwards_lno;
6e7c92a9 2477 } else {
b94f2eda 2478 if (forwards_lno == img->nr) {
fcb77bc5
JH
2479 i++;
2480 goto again;
6e7c92a9 2481 }
b94f2eda
JH
2482 forwards += img->line[forwards_lno].len;
2483 forwards_lno++;
6e7c92a9 2484 try = forwards;
b94f2eda 2485 try_lno = forwards_lno;
6e7c92a9
LT
2486 }
2487
6e7c92a9 2488 }
3cca928d
LT
2489 return -1;
2490}
2491
b94f2eda 2492static void remove_first_line(struct image *img)
47495887 2493{
b94f2eda
JH
2494 img->buf += img->line[0].len;
2495 img->len -= img->line[0].len;
2496 img->line++;
2497 img->nr--;
47495887
EB
2498}
2499
b94f2eda 2500static void remove_last_line(struct image *img)
47495887 2501{
b94f2eda 2502 img->len -= img->line[--img->nr].len;
47495887
EB
2503}
2504
92737a22
JH
2505/*
2506 * The change from "preimage" and "postimage" has been found to
2507 * apply at applied_pos (counts in line numbers) in "img".
2508 * Update "img" to remove "preimage" and replace it with "postimage".
2509 */
b94f2eda
JH
2510static void update_image(struct image *img,
2511 int applied_pos,
2512 struct image *preimage,
2513 struct image *postimage)
b5767dd6 2514{
81bf96bb 2515 /*
b94f2eda
JH
2516 * remove the copy of preimage at offset in img
2517 * and replace it with postimage
81bf96bb 2518 */
b94f2eda
JH
2519 int i, nr;
2520 size_t remove_count, insert_count, applied_at = 0;
2521 char *result;
51667147
BG
2522 int preimage_limit;
2523
2524 /*
2525 * If we are removing blank lines at the end of img,
2526 * the preimage may extend beyond the end.
2527 * If that is the case, we must be careful only to
2528 * remove the part of the preimage that falls within
2529 * the boundaries of img. Initialize preimage_limit
2530 * to the number of lines in the preimage that falls
2531 * within the boundaries.
2532 */
2533 preimage_limit = preimage->nr;
2534 if (preimage_limit > img->nr - applied_pos)
2535 preimage_limit = img->nr - applied_pos;
d5a41641 2536
b94f2eda
JH
2537 for (i = 0; i < applied_pos; i++)
2538 applied_at += img->line[i].len;
2539
2540 remove_count = 0;
51667147 2541 for (i = 0; i < preimage_limit; i++)
b94f2eda
JH
2542 remove_count += img->line[applied_pos + i].len;
2543 insert_count = postimage->len;
2544
2545 /* Adjust the contents */
2546 result = xmalloc(img->len + insert_count - remove_count + 1);
2547 memcpy(result, img->buf, applied_at);
2548 memcpy(result + applied_at, postimage->buf, postimage->len);
2549 memcpy(result + applied_at + postimage->len,
2550 img->buf + (applied_at + remove_count),
2551 img->len - (applied_at + remove_count));
2552 free(img->buf);
2553 img->buf = result;
2554 img->len += insert_count - remove_count;
2555 result[img->len] = '\0';
2556
2557 /* Adjust the line table */
51667147
BG
2558 nr = img->nr + postimage->nr - preimage_limit;
2559 if (preimage_limit < postimage->nr) {
81bf96bb 2560 /*
b94f2eda
JH
2561 * NOTE: this knows that we never call remove_first_line()
2562 * on anything other than pre/post image.
d0c25035 2563 */
b94f2eda
JH
2564 img->line = xrealloc(img->line, nr * sizeof(*img->line));
2565 img->line_allocated = img->line;
d0c25035 2566 }
51667147 2567 if (preimage_limit != postimage->nr)
b94f2eda 2568 memmove(img->line + applied_pos + postimage->nr,
51667147
BG
2569 img->line + applied_pos + preimage_limit,
2570 (img->nr - (applied_pos + preimage_limit)) *
b94f2eda
JH
2571 sizeof(*img->line));
2572 memcpy(img->line + applied_pos,
2573 postimage->line,
2574 postimage->nr * sizeof(*img->line));
933e44d3
JH
2575 if (!allow_overlap)
2576 for (i = 0; i < postimage->nr; i++)
2577 img->line[applied_pos + i].flag |= LINE_PATCHED;
b94f2eda 2578 img->nr = nr;
b5767dd6
JH
2579}
2580
92737a22
JH
2581/*
2582 * Use the patch-hunk text in "frag" to prepare two images (preimage and
2583 * postimage) for the hunk. Find lines that match "preimage" in "img" and
2584 * replace the part of "img" with "postimage" text.
2585 */
b94f2eda 2586static int apply_one_fragment(struct image *img, struct fragment *frag,
334f8cb2
JH
2587 int inaccurate_eof, unsigned ws_rule,
2588 int nth_fragment)
3cca928d 2589{
65aadb92 2590 int match_beginning, match_end;
3cca928d 2591 const char *patch = frag->patch;
b94f2eda 2592 int size = frag->size;
d511bd33
CW
2593 char *old, *oldlines;
2594 struct strbuf newlines;
077e1af5 2595 int new_blank_lines_at_end = 0;
85572639
JH
2596 int found_new_blank_lines_at_end = 0;
2597 int hunk_linenr = frag->linenr;
47495887 2598 unsigned long leading, trailing;
b94f2eda
JH
2599 int pos, applied_pos;
2600 struct image preimage;
2601 struct image postimage;
3cca928d 2602
c330fdd4
JH
2603 memset(&preimage, 0, sizeof(preimage));
2604 memset(&postimage, 0, sizeof(postimage));
61e08cca 2605 oldlines = xmalloc(size);
d511bd33 2606 strbuf_init(&newlines, size);
c330fdd4 2607
61e08cca 2608 old = oldlines;
3cca928d 2609 while (size > 0) {
e5a94313 2610 char first;
3cca928d 2611 int len = linelen(patch, size);
d511bd33 2612 int plen;
077e1af5 2613 int added_blank_line = 0;
efa57443 2614 int is_blank_context = 0;
d511bd33 2615 size_t start;
3cca928d
LT
2616
2617 if (!len)
2618 break;
2619
2620 /*
2621 * "plen" is how much of the line we should use for
2622 * the actual patch data. Normally we just remove the
2623 * first character on the line, but if the line is
2624 * followed by "\ No newline", then we also remove the
2625 * last one (which is the newline, of course).
2626 */
61e08cca 2627 plen = len - 1;
8b64647d 2628 if (len < size && patch[len] == '\\')
3cca928d 2629 plen--;
e5a94313 2630 first = *patch;
f686d030 2631 if (apply_in_reverse) {
e5a94313
JS
2632 if (first == '-')
2633 first = '+';
2634 else if (first == '+')
2635 first = '-';
2636 }
efe7f358 2637
e5a94313 2638 switch (first) {
b507b465
LT
2639 case '\n':
2640 /* Newer GNU diff, empty context line */
2641 if (plen < 0)
2642 /* ... followed by '\No newline'; nothing */
2643 break;
61e08cca 2644 *old++ = '\n';
d511bd33 2645 strbuf_addch(&newlines, '\n');
c330fdd4
JH
2646 add_line_info(&preimage, "\n", 1, LINE_COMMON);
2647 add_line_info(&postimage, "\n", 1, LINE_COMMON);
efa57443 2648 is_blank_context = 1;
b507b465 2649 break;
3cca928d 2650 case ' ':
94ea026b
JH
2651 if (plen && (ws_rule & WS_BLANK_AT_EOF) &&
2652 ws_blank_line(patch + 1, plen, ws_rule))
efa57443 2653 is_blank_context = 1;
3cca928d 2654 case '-':
61e08cca
JH
2655 memcpy(old, patch + 1, plen);
2656 add_line_info(&preimage, old, plen,
c330fdd4 2657 (first == ' ' ? LINE_COMMON : 0));
61e08cca 2658 old += plen;
e5a94313 2659 if (first == '-')
3cca928d
LT
2660 break;
2661 /* Fall-through for ' ' */
2662 case '+':
8441a9a8
JH
2663 /* --no-add does not add new lines */
2664 if (first == '+' && no_add)
2665 break;
2666
d511bd33 2667 start = newlines.len;
8441a9a8
JH
2668 if (first != '+' ||
2669 !whitespace_error ||
2670 ws_error_action != correct_ws_error) {
d511bd33 2671 strbuf_add(&newlines, patch + 1, plen);
8441a9a8
JH
2672 }
2673 else {
d511bd33 2674 ws_fix_copy(&newlines, patch + 1, plen, ws_rule, &applied_after_fixing_ws);
077e1af5 2675 }
d511bd33 2676 add_line_info(&postimage, newlines.buf + start, newlines.len - start,
8441a9a8 2677 (first == '+' ? 0 : LINE_COMMON));
8441a9a8 2678 if (first == '+' &&
94ea026b
JH
2679 (ws_rule & WS_BLANK_AT_EOF) &&
2680 ws_blank_line(patch + 1, plen, ws_rule))
8441a9a8 2681 added_blank_line = 1;
3cca928d
LT
2682 break;
2683 case '@': case '\\':
2684 /* Ignore it, we already handled it */
2685 break;
2686 default:
aeabfa07 2687 if (apply_verbosely)
3638eb43 2688 error(_("invalid start of line: '%c'"), first);
3cca928d
LT
2689 return -1;
2690 }
85572639
JH
2691 if (added_blank_line) {
2692 if (!new_blank_lines_at_end)
2693 found_new_blank_lines_at_end = hunk_linenr;
077e1af5 2694 new_blank_lines_at_end++;
85572639 2695 }
efa57443
JH
2696 else if (is_blank_context)
2697 ;
077e1af5
JH
2698 else
2699 new_blank_lines_at_end = 0;
3cca928d
LT
2700 patch += len;
2701 size -= len;
85572639 2702 hunk_linenr++;
3cca928d 2703 }
81bf96bb 2704 if (inaccurate_eof &&
61e08cca 2705 old > oldlines && old[-1] == '\n' &&
d511bd33 2706 newlines.len > 0 && newlines.buf[newlines.len - 1] == '\n') {
61e08cca 2707 old--;
d511bd33 2708 strbuf_setlen(&newlines, newlines.len - 1);
5b5d4d9e 2709 }
47495887 2710
47495887
EB
2711 leading = frag->leading;
2712 trailing = frag->trailing;
1bf1a859
LT
2713
2714 /*
ee5a317e
JH
2715 * A hunk to change lines at the beginning would begin with
2716 * @@ -1,L +N,M @@
ed0f47a8
JH
2717 * but we need to be careful. -U0 that inserts before the second
2718 * line also has this pattern.
4be60962 2719 *
ee5a317e
JH
2720 * And a hunk to add to an empty file would begin with
2721 * @@ -0,0 +N,M @@
2722 *
2723 * In other words, a hunk that is (frag->oldpos <= 1) with or
2724 * without leading context must match at the beginning.
1bf1a859 2725 */
ed0f47a8
JH
2726 match_beginning = (!frag->oldpos ||
2727 (frag->oldpos == 1 && !unidiff_zero));
ee5a317e
JH
2728
2729 /*
2730 * A hunk without trailing lines must match at the end.
2731 * However, we simply cannot tell if a hunk must match end
2732 * from the lack of trailing lines if the patch was generated
2733 * with unidiff without any context.
2734 */
2735 match_end = !unidiff_zero && !trailing;
1bf1a859 2736
b94f2eda 2737 pos = frag->newpos ? (frag->newpos - 1) : 0;
61e08cca
JH
2738 preimage.buf = oldlines;
2739 preimage.len = old - oldlines;
d511bd33
CW
2740 postimage.buf = newlines.buf;
2741 postimage.len = newlines.len;
c330fdd4
JH
2742 preimage.line = preimage.line_allocated;
2743 postimage.line = postimage.line_allocated;
2744
47495887 2745 for (;;) {
efe7f358 2746
c607aaa2
JH
2747 applied_pos = find_pos(img, &preimage, &postimage, pos,
2748 ws_rule, match_beginning, match_end);
b94f2eda
JH
2749
2750 if (applied_pos >= 0)
47495887 2751 break;
47495887
EB
2752
2753 /* Am I at my context limits? */
2754 if ((leading <= p_context) && (trailing <= p_context))
2755 break;
65aadb92
JH
2756 if (match_beginning || match_end) {
2757 match_beginning = match_end = 0;
1bf1a859
LT
2758 continue;
2759 }
b94f2eda 2760
81bf96bb
JH
2761 /*
2762 * Reduce the number of context lines; reduce both
2763 * leading and trailing if they are equal otherwise
2764 * just reduce the larger context.
47495887
EB
2765 */
2766 if (leading >= trailing) {
b94f2eda
JH
2767 remove_first_line(&preimage);
2768 remove_first_line(&postimage);
47495887
EB
2769 pos--;
2770 leading--;
2771 }
2772 if (trailing > leading) {
b94f2eda
JH
2773 remove_last_line(&preimage);
2774 remove_last_line(&postimage);
47495887 2775 trailing--;
6e7c92a9 2776 }
3cca928d
LT
2777 }
2778
b94f2eda 2779 if (applied_pos >= 0) {
77b15bbd 2780 if (new_blank_lines_at_end &&
51667147 2781 preimage.nr + applied_pos >= img->nr &&
77b15bbd
JH
2782 (ws_rule & WS_BLANK_AT_EOF) &&
2783 ws_error_action != nowarn_ws_error) {
85572639
JH
2784 record_ws_error(WS_BLANK_AT_EOF, "+", 1,
2785 found_new_blank_lines_at_end);
77b15bbd
JH
2786 if (ws_error_action == correct_ws_error) {
2787 while (new_blank_lines_at_end--)
2788 remove_last_line(&postimage);
2789 }
b94f2eda 2790 /*
77b15bbd
JH
2791 * We would want to prevent write_out_results()
2792 * from taking place in apply_patch() that follows
2793 * the callchain led us here, which is:
2794 * apply_patch->check_patch_list->check_patch->
2795 * apply_data->apply_fragments->apply_one_fragment
b94f2eda 2796 */
77b15bbd
JH
2797 if (ws_error_action == die_on_ws_error)
2798 apply = 0;
b94f2eda 2799 }
aeabfa07 2800
334f8cb2
JH
2801 if (apply_verbosely && applied_pos != pos) {
2802 int offset = applied_pos - pos;
2803 if (apply_in_reverse)
2804 offset = 0 - offset;
3638eb43
NTND
2805 fprintf_ln(stderr,
2806 Q_("Hunk #%d succeeded at %d (offset %d line).",
2807 "Hunk #%d succeeded at %d (offset %d lines).",
2808 offset),
2809 nth_fragment, applied_pos + 1, offset);
334f8cb2
JH
2810 }
2811
b94f2eda
JH
2812 /*
2813 * Warn if it was necessary to reduce the number
2814 * of context lines.
2815 */
2816 if ((leading != frag->leading) ||
2817 (trailing != frag->trailing))
3638eb43
NTND
2818 fprintf_ln(stderr, _("Context reduced to (%ld/%ld)"
2819 " to apply fragment at %d"),
2820 leading, trailing, applied_pos+1);
b94f2eda
JH
2821 update_image(img, applied_pos, &preimage, &postimage);
2822 } else {
2823 if (apply_verbosely)
3638eb43 2824 error(_("while searching for:\n%.*s"),
61e08cca 2825 (int)(old - oldlines), oldlines);
b94f2eda 2826 }
aeabfa07 2827
61e08cca 2828 free(oldlines);
d511bd33 2829 strbuf_release(&newlines);
b94f2eda
JH
2830 free(preimage.line_allocated);
2831 free(postimage.line_allocated);
2832
2833 return (applied_pos < 0);
3cca928d
LT
2834}
2835
b94f2eda 2836static int apply_binary_fragment(struct image *img, struct patch *patch)
0660626c 2837{
0660626c 2838 struct fragment *fragment = patch->fragments;
c7f9cb14
PH
2839 unsigned long len;
2840 void *dst;
0660626c 2841
24305cd7 2842 if (!fragment)
3638eb43 2843 return error(_("missing binary patch data for '%s'"),
24305cd7
JK
2844 patch->new_name ?
2845 patch->new_name :
2846 patch->old_name);
2847
3cd4f5e8
JH
2848 /* Binary patch is irreversible without the optional second hunk */
2849 if (apply_in_reverse) {
2850 if (!fragment->next)
2851 return error("cannot reverse-apply a binary patch "
2852 "without the reverse hunk to '%s'",
2853 patch->new_name
2854 ? patch->new_name : patch->old_name);
03eb8f8a 2855 fragment = fragment->next;
3cd4f5e8 2856 }
3cd4f5e8 2857 switch (fragment->binary_patch_method) {
0660626c 2858 case BINARY_DELTA_DEFLATED:
b94f2eda 2859 dst = patch_delta(img->buf, img->len, fragment->patch,
c7f9cb14
PH
2860 fragment->size, &len);
2861 if (!dst)
2862 return -1;
b94f2eda
JH
2863 clear_image(img);
2864 img->buf = dst;
2865 img->len = len;
c7f9cb14 2866 return 0;
0660626c 2867 case BINARY_LITERAL_DEFLATED:
b94f2eda
JH
2868 clear_image(img);
2869 img->len = fragment->size;
2870 img->buf = xmalloc(img->len+1);
2871 memcpy(img->buf, fragment->patch, img->len);
2872 img->buf[img->len] = '\0';
c7f9cb14 2873 return 0;
0660626c 2874 }
c7f9cb14 2875 return -1;
0660626c
JH
2876}
2877
92737a22
JH
2878/*
2879 * Replace "img" with the result of applying the binary patch.
2880 * The binary patch data itself in patch->fragment is still kept
2881 * but the preimage prepared by the caller in "img" is freed here
2882 * or in the helper function apply_binary_fragment() this calls.
2883 */
b94f2eda 2884static int apply_binary(struct image *img, struct patch *patch)
3cca928d 2885{
011f4274 2886 const char *name = patch->old_name ? patch->old_name : patch->new_name;
051308f6 2887 unsigned char sha1[20];
011f4274 2888
81bf96bb
JH
2889 /*
2890 * For safety, we require patch index line to contain
051308f6
JH
2891 * full 40-byte textual SHA1 for old and new, at least for now.
2892 */
2893 if (strlen(patch->old_sha1_prefix) != 40 ||
2894 strlen(patch->new_sha1_prefix) != 40 ||
2895 get_sha1_hex(patch->old_sha1_prefix, sha1) ||
2896 get_sha1_hex(patch->new_sha1_prefix, sha1))
2897 return error("cannot apply binary patch to '%s' "
2898 "without full index line", name);
011f4274 2899
051308f6 2900 if (patch->old_name) {
81bf96bb
JH
2901 /*
2902 * See if the old one matches what the patch
051308f6 2903 * applies to.
011f4274 2904 */
b94f2eda 2905 hash_sha1_file(img->buf, img->len, blob_type, sha1);
051308f6
JH
2906 if (strcmp(sha1_to_hex(sha1), patch->old_sha1_prefix))
2907 return error("the patch applies to '%s' (%s), "
2908 "which does not match the "
2909 "current contents.",
2910 name, sha1_to_hex(sha1));
2911 }
2912 else {
2913 /* Otherwise, the old one must be empty. */
b94f2eda 2914 if (img->len)
051308f6
JH
2915 return error("the patch applies to an empty "
2916 "'%s' but it is not empty", name);
2917 }
011f4274 2918
0660626c 2919 get_sha1_hex(patch->new_sha1_prefix, sha1);
0bef57ee 2920 if (is_null_sha1(sha1)) {
b94f2eda 2921 clear_image(img);
051308f6 2922 return 0; /* deletion patch */
0660626c 2923 }
011f4274 2924
051308f6 2925 if (has_sha1_file(sha1)) {
0660626c 2926 /* We already have the postimage */
21666f1a 2927 enum object_type type;
051308f6 2928 unsigned long size;
c7f9cb14 2929 char *result;
051308f6 2930
c7f9cb14
PH
2931 result = read_sha1_file(sha1, &type, &size);
2932 if (!result)
051308f6
JH
2933 return error("the necessary postimage %s for "
2934 "'%s' cannot be read",
2935 patch->new_sha1_prefix, name);
b94f2eda
JH
2936 clear_image(img);
2937 img->buf = result;
2938 img->len = size;
c7f9cb14 2939 } else {
81bf96bb
JH
2940 /*
2941 * We have verified buf matches the preimage;
0660626c
JH
2942 * apply the patch data to it, which is stored
2943 * in the patch->fragments->{patch,size}.
011f4274 2944 */
b94f2eda 2945 if (apply_binary_fragment(img, patch))
3638eb43 2946 return error(_("binary patch does not apply to '%s'"),
051308f6 2947 name);
011f4274 2948
051308f6 2949 /* verify that the result matches */
b94f2eda 2950 hash_sha1_file(img->buf, img->len, blob_type, sha1);
051308f6 2951 if (strcmp(sha1_to_hex(sha1), patch->new_sha1_prefix))
3638eb43 2952 return error(_("binary patch to '%s' creates incorrect result (expecting %s, got %s)"),
c7f9cb14 2953 name, patch->new_sha1_prefix, sha1_to_hex(sha1));
011f4274 2954 }
3cca928d 2955
051308f6
JH
2956 return 0;
2957}
2958
b94f2eda 2959static int apply_fragments(struct image *img, struct patch *patch)
051308f6
JH
2960{
2961 struct fragment *frag = patch->fragments;
2962 const char *name = patch->old_name ? patch->old_name : patch->new_name;
cf1b7869
JH
2963 unsigned ws_rule = patch->ws_rule;
2964 unsigned inaccurate_eof = patch->inaccurate_eof;
334f8cb2 2965 int nth = 0;
051308f6
JH
2966
2967 if (patch->is_binary)
b94f2eda 2968 return apply_binary(img, patch);
051308f6 2969
3cca928d 2970 while (frag) {
334f8cb2
JH
2971 nth++;
2972 if (apply_one_fragment(img, frag, inaccurate_eof, ws_rule, nth)) {
3638eb43 2973 error(_("patch failed: %s:%ld"), name, frag->oldpos);
57dc397c
JH
2974 if (!apply_with_reject)
2975 return -1;
2976 frag->rejected = 1;
2977 }
3cca928d
LT
2978 frag = frag->next;
2979 }
30996652 2980 return 0;
3cca928d
LT
2981}
2982
e42a96e7 2983static int read_blob_object(struct strbuf *buf, const unsigned char *sha1, unsigned mode)
e06c5a6c 2984{
e42a96e7 2985 if (S_ISGITLINK(mode)) {
c7f9cb14 2986 strbuf_grow(buf, 100);
e42a96e7 2987 strbuf_addf(buf, "Subproject commit %s\n", sha1_to_hex(sha1));
e06c5a6c
SV
2988 } else {
2989 enum object_type type;
c7f9cb14
PH
2990 unsigned long sz;
2991 char *result;
2992
e42a96e7 2993 result = read_sha1_file(sha1, &type, &sz);
c7f9cb14 2994 if (!result)
e06c5a6c 2995 return -1;
c7f9cb14
PH
2996 /* XXX read_sha1_file NUL-terminates */
2997 strbuf_attach(buf, result, sz, sz + 1);
e06c5a6c
SV
2998 }
2999 return 0;
3000}
3001
9c5e6c80 3002static int read_file_or_gitlink(const struct cache_entry *ce, struct strbuf *buf)
e42a96e7
JH
3003{
3004 if (!ce)
3005 return 0;
3006 return read_blob_object(buf, ce->sha1, ce->ce_mode);
3007}
3008
7a07841c
DZ
3009static struct patch *in_fn_table(const char *name)
3010{
c455c87c 3011 struct string_list_item *item;
7a07841c
DZ
3012
3013 if (name == NULL)
3014 return NULL;
3015
e8c8b713 3016 item = string_list_lookup(&fn_table, name);
7a07841c
DZ
3017 if (item != NULL)
3018 return (struct patch *)item->util;
3019
3020 return NULL;
3021}
3022
7fac0eef
MK
3023/*
3024 * item->util in the filename table records the status of the path.
3025 * Usually it points at a patch (whose result records the contents
3026 * of it after applying it), but it could be PATH_WAS_DELETED for a
f3b8f91a
JH
3027 * path that a previously applied patch has already removed, or
3028 * PATH_TO_BE_DELETED for a path that a later patch would remove.
3029 *
3030 * The latter is needed to deal with a case where two paths A and B
3031 * are swapped by first renaming A to B and then renaming B to A;
41ccfdd9 3032 * moving A to B should not be prevented due to presence of B as we
f3b8f91a 3033 * will remove it in a later patch.
7fac0eef 3034 */
f3b8f91a 3035#define PATH_TO_BE_DELETED ((struct patch *) -2)
7fac0eef
MK
3036#define PATH_WAS_DELETED ((struct patch *) -1)
3037
3038static int to_be_deleted(struct patch *patch)
3039{
3040 return patch == PATH_TO_BE_DELETED;
3041}
3042
3043static int was_deleted(struct patch *patch)
3044{
3045 return patch == PATH_WAS_DELETED;
3046}
3047
7a07841c
DZ
3048static void add_to_fn_table(struct patch *patch)
3049{
c455c87c 3050 struct string_list_item *item;
7a07841c
DZ
3051
3052 /*
3053 * Always add new_name unless patch is a deletion
3054 * This should cover the cases for normal diffs,
3055 * file creations and copies
3056 */
3057 if (patch->new_name != NULL) {
78a395d3 3058 item = string_list_insert(&fn_table, patch->new_name);
7a07841c
DZ
3059 item->util = patch;
3060 }
3061
3062 /*
3063 * store a failure on rename/deletion cases because
3064 * later chunks shouldn't patch old names
3065 */
3066 if ((patch->new_name == NULL) || (patch->is_rename)) {
78a395d3 3067 item = string_list_insert(&fn_table, patch->old_name);
7fac0eef
MK
3068 item->util = PATH_WAS_DELETED;
3069 }
3070}
3071
3072static void prepare_fn_table(struct patch *patch)
3073{
3074 /*
3075 * store information about incoming file deletion
3076 */
3077 while (patch) {
3078 if ((patch->new_name == NULL) || (patch->is_rename)) {
3079 struct string_list_item *item;
78a395d3 3080 item = string_list_insert(&fn_table, patch->old_name);
7fac0eef
MK
3081 item->util = PATH_TO_BE_DELETED;
3082 }
3083 patch = patch->next;
7a07841c
DZ
3084 }
3085}
3086
f4c66eed
JH
3087static int checkout_target(struct cache_entry *ce, struct stat *st)
3088{
3089 struct checkout costate;
3090
3091 memset(&costate, 0, sizeof(costate));
3092 costate.base_dir = "";
3093 costate.refresh_cache = 1;
3094 if (checkout_entry(ce, &costate, NULL) || lstat(ce->name, st))
3095 return error(_("cannot checkout %s"), ce->name);
3096 return 0;
3097}
3098
ccf998b2
JH
3099static struct patch *previous_patch(struct patch *patch, int *gone)
3100{
3101 struct patch *previous;
3102
3103 *gone = 0;
3104 if (patch->is_copy || patch->is_rename)
3105 return NULL; /* "git" patches do not depend on the order */
3106
3107 previous = in_fn_table(patch->old_name);
3108 if (!previous)
3109 return NULL;
3110
3111 if (to_be_deleted(previous))
3112 return NULL; /* the deletion hasn't happened yet */
3113
3114 if (was_deleted(previous))
3115 *gone = 1;
3116
3117 return previous;
3118}
3119
9c5e6c80 3120static int verify_index_match(const struct cache_entry *ce, struct stat *st)
e09837e2
JH
3121{
3122 if (S_ISGITLINK(ce->ce_mode)) {
3123 if (!S_ISDIR(st->st_mode))
3124 return -1;
3125 return 0;
3126 }
3127 return ce_match_stat(ce, st, CE_MATCH_IGNORE_VALID|CE_MATCH_IGNORE_SKIP_WORKTREE);
3128}
3129
5a812661
JH
3130#define SUBMODULE_PATCH_WITHOUT_INDEX 1
3131
3132static int load_patch_target(struct strbuf *buf,
9c5e6c80 3133 const struct cache_entry *ce,
5a812661
JH
3134 struct stat *st,
3135 const char *name,
3136 unsigned expected_mode)
3137{
3138 if (cached) {
3139 if (read_file_or_gitlink(ce, buf))
3140 return error(_("read of %s failed"), name);
3141 } else if (name) {
3142 if (S_ISGITLINK(expected_mode)) {
3143 if (ce)
3144 return read_file_or_gitlink(ce, buf);
3145 else
3146 return SUBMODULE_PATCH_WITHOUT_INDEX;
3147 } else {
3148 if (read_old_data(st, name, buf))
3149 return error(_("read of %s failed"), name);
3150 }
3151 }
3152 return 0;
3153}
3154
ccf998b2
JH
3155/*
3156 * We are about to apply "patch"; populate the "image" with the
3157 * current version we have, from the working tree or from the index,
3158 * depending on the situation e.g. --cached/--index. If we are
3159 * applying a non-git patch that incrementally updates the tree,
3160 * we read from the result of a previous diff.
3161 */
37b9c903 3162static int load_preimage(struct image *image,
9c5e6c80
NTND
3163 struct patch *patch, struct stat *st,
3164 const struct cache_entry *ce)
3cca928d 3165{
f285a2d7 3166 struct strbuf buf = STRBUF_INIT;
b94f2eda
JH
3167 size_t len;
3168 char *img;
ccf998b2
JH
3169 struct patch *previous;
3170 int status;
3cca928d 3171
ccf998b2
JH
3172 previous = previous_patch(patch, &status);
3173 if (status)
3174 return error(_("path %s has been renamed/deleted"),
3175 patch->old_name);
3176 if (previous) {
92737a22 3177 /* We have a patched copy in memory; use that. */
ccf998b2 3178 strbuf_add(&buf, previous->result, previous->resultsize);
5a812661
JH
3179 } else {
3180 status = load_patch_target(&buf, ce, st,
3181 patch->old_name, patch->old_mode);
3182 if (status < 0)
3183 return status;
3184 else if (status == SUBMODULE_PATCH_WITHOUT_INDEX) {
3185 /*
3186 * There is no way to apply subproject
3187 * patch without looking at the index.
3188 * NEEDSWORK: shouldn't this be flagged
3189 * as an error???
3190 */
3191 free_fragment_list(patch->fragments);
3192 patch->fragments = NULL;
3193 } else if (status) {
3638eb43 3194 return error(_("read of %s failed"), patch->old_name);
04e4888e
JH
3195 }
3196 }
6e7c92a9 3197
b94f2eda 3198 img = strbuf_detach(&buf, &len);
37b9c903
JH
3199 prepare_image(image, img, len, !patch->is_binary);
3200 return 0;
3201}
b94f2eda 3202
519d1a5b
JH
3203static int three_way_merge(struct image *image,
3204 char *path,
3205 const unsigned char *base,
3206 const unsigned char *ours,
3207 const unsigned char *theirs)
3208{
28ff0512
JH
3209 mmfile_t base_file, our_file, their_file;
3210 mmbuffer_t result = { NULL };
3211 int status;
5aa7d94c 3212
28ff0512
JH
3213 read_mmblob(&base_file, base);
3214 read_mmblob(&our_file, ours);
3215 read_mmblob(&their_file, theirs);
3216 status = ll_merge(&result, path,
3217 &base_file, "base",
3218 &our_file, "ours",
3219 &their_file, "theirs", NULL);
3220 free(base_file.ptr);
3221 free(our_file.ptr);
3222 free(their_file.ptr);
3223 if (status < 0 || !result.ptr) {
3224 free(result.ptr);
3225 return -1;
3226 }
3227 clear_image(image);
3228 image->buf = result.ptr;
3229 image->len = result.size;
5aa7d94c 3230
28ff0512 3231 return status;
519d1a5b
JH
3232}
3233
099f3c42
JH
3234/*
3235 * When directly falling back to add/add three-way merge, we read from
3236 * the current contents of the new_name. In no cases other than that
3237 * this function will be called.
3238 */
3239static int load_current(struct image *image, struct patch *patch)
3240{
3241 struct strbuf buf = STRBUF_INIT;
3242 int status, pos;
3243 size_t len;
3244 char *img;
3245 struct stat st;
3246 struct cache_entry *ce;
3247 char *name = patch->new_name;
3248 unsigned mode = patch->new_mode;
3249
3250 if (!patch->is_new)
3251 die("BUG: patch to %s is not a creation", patch->old_name);
3252
3253 pos = cache_name_pos(name, strlen(name));
3254 if (pos < 0)
3255 return error(_("%s: does not exist in index"), name);
3256 ce = active_cache[pos];
3257 if (lstat(name, &st)) {
3258 if (errno != ENOENT)
3259 return error(_("%s: %s"), name, strerror(errno));
3260 if (checkout_target(ce, &st))
3261 return -1;
3262 }
3263 if (verify_index_match(ce, &st))
3264 return error(_("%s: does not match index"), name);
3265
3266 status = load_patch_target(&buf, ce, &st, name, mode);
3267 if (status < 0)
3268 return status;
3269 else if (status)
3270 return -1;
3271 img = strbuf_detach(&buf, &len);
3272 prepare_image(image, img, len, !patch->is_binary);
3cca928d
LT
3273 return 0;
3274}
3275
cfb6f9ac 3276static int try_threeway(struct image *image, struct patch *patch,
9c5e6c80 3277 struct stat *st, const struct cache_entry *ce)
64cab591 3278{
519d1a5b
JH
3279 unsigned char pre_sha1[20], post_sha1[20], our_sha1[20];
3280 struct strbuf buf = STRBUF_INIT;
3281 size_t len;
28ff0512 3282 int status;
519d1a5b
JH
3283 char *img;
3284 struct image tmp_image;
3285
3286 /* No point falling back to 3-way merge in these cases */
099f3c42 3287 if (patch->is_delete ||
519d1a5b
JH
3288 S_ISGITLINK(patch->old_mode) || S_ISGITLINK(patch->new_mode))
3289 return -1;
64cab591 3290
519d1a5b 3291 /* Preimage the patch was prepared for */
099f3c42
JH
3292 if (patch->is_new)
3293 write_sha1_file("", 0, blob_type, pre_sha1);
3294 else if (get_sha1(patch->old_sha1_prefix, pre_sha1) ||
3295 read_blob_object(&buf, pre_sha1, patch->old_mode))
519d1a5b 3296 return error("repository lacks the necessary blob to fall back on 3-way merge.");
28ff0512
JH
3297
3298 fprintf(stderr, "Falling back to three-way merge...\n");
3299
519d1a5b
JH
3300 img = strbuf_detach(&buf, &len);
3301 prepare_image(&tmp_image, img, len, 1);
3302 /* Apply the patch to get the post image */
3303 if (apply_fragments(&tmp_image, patch) < 0) {
3304 clear_image(&tmp_image);
3305 return -1;
3306 }
3307 /* post_sha1[] is theirs */
3308 write_sha1_file(tmp_image.buf, tmp_image.len, blob_type, post_sha1);
3309 clear_image(&tmp_image);
3310
3311 /* our_sha1[] is ours */
099f3c42
JH
3312 if (patch->is_new) {
3313 if (load_current(&tmp_image, patch))
3314 return error("cannot read the current contents of '%s'",
3315 patch->new_name);
3316 } else {
3317 if (load_preimage(&tmp_image, patch, st, ce))
3318 return error("cannot read the current contents of '%s'",
3319 patch->old_name);
3320 }
519d1a5b
JH
3321 write_sha1_file(tmp_image.buf, tmp_image.len, blob_type, our_sha1);
3322 clear_image(&tmp_image);
3323
3324 /* in-core three-way merge between post and our using pre as base */
28ff0512
JH
3325 status = three_way_merge(image, patch->new_name,
3326 pre_sha1, our_sha1, post_sha1);
3327 if (status < 0) {
3328 fprintf(stderr, "Failed to fall back on three-way merge...\n");
3329 return status;
3330 }
3331
3332 if (status) {
3333 patch->conflicted_threeway = 1;
4f4a6cb9
JH
3334 if (patch->is_new)
3335 hashclr(patch->threeway_stage[0]);
3336 else
3337 hashcpy(patch->threeway_stage[0], pre_sha1);
28ff0512
JH
3338 hashcpy(patch->threeway_stage[1], our_sha1);
3339 hashcpy(patch->threeway_stage[2], post_sha1);
3340 fprintf(stderr, "Applied patch to '%s' with conflicts.\n", patch->new_name);
3341 } else {
3342 fprintf(stderr, "Applied patch to '%s' cleanly.\n", patch->new_name);
64cab591 3343 }
64cab591
JH
3344 return 0;
3345}
3346
9c5e6c80 3347static int apply_data(struct patch *patch, struct stat *st, const struct cache_entry *ce)
e06c5a6c 3348{
37b9c903
JH
3349 struct image image;
3350
3351 if (load_preimage(&image, patch, st, ce) < 0)
3352 return -1;
b94f2eda 3353
099f3c42
JH
3354 if (patch->direct_to_threeway ||
3355 apply_fragments(&image, patch) < 0) {
cfb6f9ac
JH
3356 /* Note: with --reject, apply_fragments() returns 0 */
3357 if (!threeway || try_threeway(&image, patch, st, ce) < 0)
e06c5a6c 3358 return -1;
e06c5a6c 3359 }
b94f2eda
JH
3360 patch->result = image.buf;
3361 patch->resultsize = image.len;
7a07841c 3362 add_to_fn_table(patch);
b94f2eda 3363 free(image.line_allocated);
5aa7d94c 3364
4be60962 3365 if (0 < patch->is_delete && patch->resultsize)
3638eb43 3366 return error(_("removal patch leaves file contents"));
5aa7d94c 3367
3cca928d 3368 return 0;
e06c5a6c
SV
3369}
3370
ccf998b2
JH
3371/*
3372 * If "patch" that we are looking at modifies or deletes what we have,
3373 * we would want it not to lose any local modification we have, either
3374 * in the working tree or in the index.
3375 *
3376 * This also decides if a non-git patch is a creation patch or a
3377 * modification to an existing empty file. We do not check the state
3378 * of the current tree for a creation patch in this function; the caller
3379 * check_patch() separately makes sure (and errors out otherwise) that
3380 * the path the patch creates does not exist in the current tree.
3381 */
5c47f4c6 3382static int check_preimage(struct patch *patch, struct cache_entry **ce, struct stat *st)
fab2c257
LT
3383{
3384 const char *old_name = patch->old_name;
ccf998b2
JH
3385 struct patch *previous = NULL;
3386 int stat_ret = 0, status;
5c47f4c6 3387 unsigned st_mode = 0;
e06c5a6c 3388
5c47f4c6
JH
3389 if (!old_name)
3390 return 0;
04e4888e 3391
5c47f4c6 3392 assert(patch->is_new <= 0);
ccf998b2 3393 previous = previous_patch(patch, &status);
a9a3e82e 3394
ccf998b2
JH
3395 if (status)
3396 return error(_("path %s has been renamed/deleted"), old_name);
3397 if (previous) {
3398 st_mode = previous->new_mode;
7a07841c 3399 } else if (!cached) {
5c47f4c6
JH
3400 stat_ret = lstat(old_name, st);
3401 if (stat_ret && errno != ENOENT)
3638eb43 3402 return error(_("%s: %s"), old_name, strerror(errno));
5c47f4c6 3403 }
a9a3e82e 3404
ccf998b2 3405 if (check_index && !previous) {
5c47f4c6
JH
3406 int pos = cache_name_pos(old_name, strlen(old_name));
3407 if (pos < 0) {
3408 if (patch->is_new < 0)
3409 goto is_new;
3638eb43 3410 return error(_("%s: does not exist in index"), old_name);
5c47f4c6
JH
3411 }
3412 *ce = active_cache[pos];
3413 if (stat_ret < 0) {
f4c66eed 3414 if (checkout_target(*ce, st))
5c47f4c6
JH
3415 return -1;
3416 }
3417 if (!cached && verify_index_match(*ce, st))
3638eb43 3418 return error(_("%s: does not match index"), old_name);
5c47f4c6
JH
3419 if (cached)
3420 st_mode = (*ce)->ce_mode;
3421 } else if (stat_ret < 0) {
3cca928d 3422 if (patch->is_new < 0)
5c47f4c6 3423 goto is_new;
3638eb43 3424 return error(_("%s: %s"), old_name, strerror(errno));
fab2c257 3425 }
a577284a 3426
ccf998b2 3427 if (!cached && !previous)
5c47f4c6
JH
3428 st_mode = ce_mode_from_stat(*ce, st->st_mode);
3429
3430 if (patch->is_new < 0)
3431 patch->is_new = 0;
3432 if (!patch->old_mode)
3433 patch->old_mode = st_mode;
3434 if ((st_mode ^ patch->old_mode) & S_IFMT)
3638eb43 3435 return error(_("%s: wrong type"), old_name);
5c47f4c6 3436 if (st_mode != patch->old_mode)
3638eb43 3437 warning(_("%s has type %o, expected %o"),
5c47f4c6 3438 old_name, st_mode, patch->old_mode);
a15080e5 3439 if (!patch->new_mode && !patch->is_delete)
1f7903a3 3440 patch->new_mode = st_mode;
5c47f4c6
JH
3441 return 0;
3442
3443 is_new:
3444 patch->is_new = 1;
3445 patch->is_delete = 0;
2901bbe5 3446 free(patch->old_name);
5c47f4c6
JH
3447 patch->old_name = NULL;
3448 return 0;
3449}
3450
ec15be02
JH
3451
3452#define EXISTS_IN_INDEX 1
3453#define EXISTS_IN_WORKTREE 2
3454
3455static int check_to_create(const char *new_name, int ok_if_exists)
813ebf82
JH
3456{
3457 struct stat nst;
ec15be02
JH
3458
3459 if (check_index &&
3460 cache_name_pos(new_name, strlen(new_name)) >= 0 &&
3461 !ok_if_exists)
3462 return EXISTS_IN_INDEX;
3463 if (cached)
3464 return 0;
3465
813ebf82
JH
3466 if (!lstat(new_name, &nst)) {
3467 if (S_ISDIR(nst.st_mode) || ok_if_exists)
3468 return 0;
3469 /*
3470 * A leading component of new_name might be a symlink
3471 * that is going to be removed with this patch, but
3472 * still pointing at somewhere that has the path.
3473 * In such a case, path "new_name" does not exist as
3474 * far as git is concerned.
3475 */
3476 if (has_symlink_leading_path(new_name, strlen(new_name)))
3477 return 0;
3478
ec15be02
JH
3479 return EXISTS_IN_WORKTREE;
3480 } else if ((errno != ENOENT) && (errno != ENOTDIR)) {
813ebf82 3481 return error("%s: %s", new_name, strerror(errno));
ec15be02 3482 }
813ebf82
JH
3483 return 0;
3484}
3485
92737a22
JH
3486/*
3487 * Check and apply the patch in-core; leave the result in patch->result
3488 * for the caller to write it out to the final destination.
3489 */
7a07841c 3490static int check_patch(struct patch *patch)
5c47f4c6
JH
3491{
3492 struct stat st;
3493 const char *old_name = patch->old_name;
3494 const char *new_name = patch->new_name;