]>
Commit | Line | Data |
---|---|---|
13b5af22 CC |
1 | /* |
2 | * apply.c | |
3 | * | |
4 | * Copyright (C) Linus Torvalds, 2005 | |
5 | * | |
6 | * This applies patches on top of some (arbitrary) version of the SCM. | |
7 | * | |
8 | */ | |
9 | ||
bb493a5c | 10 | #include "cache.h" |
13b5af22 CC |
11 | #include "blob.h" |
12 | #include "delta.h" | |
13 | #include "diff.h" | |
14 | #include "dir.h" | |
15 | #include "xdiff-interface.h" | |
16 | #include "ll-merge.h" | |
bb493a5c | 17 | #include "lockfile.h" |
13b5af22 CC |
18 | #include "parse-options.h" |
19 | #include "quote.h" | |
20 | #include "rerere.h" | |
bb493a5c CC |
21 | #include "apply.h" |
22 | ||
23 | static void git_apply_config(void) | |
24 | { | |
25 | git_config_get_string_const("apply.whitespace", &apply_default_whitespace); | |
26 | git_config_get_string_const("apply.ignorewhitespace", &apply_default_ignorewhitespace); | |
27 | git_config(git_default_config, NULL); | |
28 | } | |
29 | ||
9123d5dd | 30 | static int parse_whitespace_option(struct apply_state *state, const char *option) |
bb493a5c CC |
31 | { |
32 | if (!option) { | |
33 | state->ws_error_action = warn_on_ws_error; | |
34 | return 0; | |
35 | } | |
36 | if (!strcmp(option, "warn")) { | |
37 | state->ws_error_action = warn_on_ws_error; | |
38 | return 0; | |
39 | } | |
40 | if (!strcmp(option, "nowarn")) { | |
41 | state->ws_error_action = nowarn_ws_error; | |
42 | return 0; | |
43 | } | |
44 | if (!strcmp(option, "error")) { | |
45 | state->ws_error_action = die_on_ws_error; | |
46 | return 0; | |
47 | } | |
48 | if (!strcmp(option, "error-all")) { | |
49 | state->ws_error_action = die_on_ws_error; | |
50 | state->squelch_whitespace_errors = 0; | |
51 | return 0; | |
52 | } | |
53 | if (!strcmp(option, "strip") || !strcmp(option, "fix")) { | |
54 | state->ws_error_action = correct_ws_error; | |
55 | return 0; | |
56 | } | |
57 | return error(_("unrecognized whitespace option '%s'"), option); | |
58 | } | |
59 | ||
9123d5dd CC |
60 | static int parse_ignorewhitespace_option(struct apply_state *state, |
61 | const char *option) | |
bb493a5c CC |
62 | { |
63 | if (!option || !strcmp(option, "no") || | |
64 | !strcmp(option, "false") || !strcmp(option, "never") || | |
65 | !strcmp(option, "none")) { | |
66 | state->ws_ignore_action = ignore_ws_none; | |
67 | return 0; | |
68 | } | |
69 | if (!strcmp(option, "change")) { | |
70 | state->ws_ignore_action = ignore_ws_change; | |
71 | return 0; | |
72 | } | |
73 | return error(_("unrecognized whitespace ignore option '%s'"), option); | |
74 | } | |
75 | ||
2f5a6d12 CC |
76 | int init_apply_state(struct apply_state *state, |
77 | const char *prefix, | |
78 | struct lock_file *lock_file) | |
bb493a5c CC |
79 | { |
80 | memset(state, 0, sizeof(*state)); | |
81 | state->prefix = prefix; | |
82 | state->prefix_length = state->prefix ? strlen(state->prefix) : 0; | |
83 | state->lock_file = lock_file; | |
84 | state->newfd = -1; | |
85 | state->apply = 1; | |
86 | state->line_termination = '\n'; | |
87 | state->p_value = 1; | |
88 | state->p_context = UINT_MAX; | |
89 | state->squelch_whitespace_errors = 5; | |
90 | state->ws_error_action = warn_on_ws_error; | |
91 | state->ws_ignore_action = ignore_ws_none; | |
92 | state->linenr = 1; | |
93 | string_list_init(&state->fn_table, 0); | |
94 | string_list_init(&state->limit_by_name, 0); | |
95 | string_list_init(&state->symlink_changes, 0); | |
96 | strbuf_init(&state->root, 0); | |
97 | ||
98 | git_apply_config(); | |
99 | if (apply_default_whitespace && parse_whitespace_option(state, apply_default_whitespace)) | |
2f5a6d12 | 100 | return -1; |
bb493a5c | 101 | if (apply_default_ignorewhitespace && parse_ignorewhitespace_option(state, apply_default_ignorewhitespace)) |
2f5a6d12 CC |
102 | return -1; |
103 | return 0; | |
bb493a5c CC |
104 | } |
105 | ||
106 | void clear_apply_state(struct apply_state *state) | |
107 | { | |
108 | string_list_clear(&state->limit_by_name, 0); | |
109 | string_list_clear(&state->symlink_changes, 0); | |
110 | strbuf_release(&state->root); | |
111 | ||
112 | /* &state->fn_table is cleared at the end of apply_patch() */ | |
113 | } | |
b6446d54 CC |
114 | |
115 | int check_apply_state(struct apply_state *state, int force_apply) | |
116 | { | |
117 | int is_not_gitdir = !startup_info->have_repository; | |
118 | ||
119 | if (state->apply_with_reject && state->threeway) | |
120 | return error("--reject and --3way cannot be used together."); | |
121 | if (state->cached && state->threeway) | |
122 | return error("--cached and --3way cannot be used together."); | |
123 | if (state->threeway) { | |
124 | if (is_not_gitdir) | |
125 | return error(_("--3way outside a repository")); | |
126 | state->check_index = 1; | |
127 | } | |
128 | if (state->apply_with_reject) | |
129 | state->apply = state->apply_verbosely = 1; | |
130 | if (!force_apply && (state->diffstat || state->numstat || state->summary || state->check || state->fake_ancestor)) | |
131 | state->apply = 0; | |
132 | if (state->check_index && is_not_gitdir) | |
133 | return error(_("--index outside a repository")); | |
134 | if (state->cached) { | |
135 | if (is_not_gitdir) | |
136 | return error(_("--cached outside a repository")); | |
137 | state->check_index = 1; | |
138 | } | |
139 | if (state->check_index) | |
140 | state->unsafe_paths = 0; | |
141 | if (!state->lock_file) | |
142 | return error("BUG: state->lock_file should not be NULL"); | |
143 | ||
144 | return 0; | |
145 | } | |
13b5af22 CC |
146 | |
147 | static void set_default_whitespace_mode(struct apply_state *state) | |
148 | { | |
149 | if (!state->whitespace_option && !apply_default_whitespace) | |
150 | state->ws_error_action = (state->apply ? warn_on_ws_error : nowarn_ws_error); | |
151 | } | |
152 | ||
153 | /* | |
154 | * This represents one "hunk" from a patch, starting with | |
155 | * "@@ -oldpos,oldlines +newpos,newlines @@" marker. The | |
156 | * patch text is pointed at by patch, and its byte length | |
157 | * is stored in size. leading and trailing are the number | |
158 | * of context lines. | |
159 | */ | |
160 | struct fragment { | |
161 | unsigned long leading, trailing; | |
162 | unsigned long oldpos, oldlines; | |
163 | unsigned long newpos, newlines; | |
164 | /* | |
165 | * 'patch' is usually borrowed from buf in apply_patch(), | |
166 | * but some codepaths store an allocated buffer. | |
167 | */ | |
168 | const char *patch; | |
169 | unsigned free_patch:1, | |
170 | rejected:1; | |
171 | int size; | |
172 | int linenr; | |
173 | struct fragment *next; | |
174 | }; | |
175 | ||
176 | /* | |
177 | * When dealing with a binary patch, we reuse "leading" field | |
178 | * to store the type of the binary hunk, either deflated "delta" | |
179 | * or deflated "literal". | |
180 | */ | |
181 | #define binary_patch_method leading | |
182 | #define BINARY_DELTA_DEFLATED 1 | |
183 | #define BINARY_LITERAL_DEFLATED 2 | |
184 | ||
185 | /* | |
186 | * This represents a "patch" to a file, both metainfo changes | |
187 | * such as creation/deletion, filemode and content changes represented | |
188 | * as a series of fragments. | |
189 | */ | |
190 | struct patch { | |
191 | char *new_name, *old_name, *def_name; | |
192 | unsigned int old_mode, new_mode; | |
193 | int is_new, is_delete; /* -1 = unknown, 0 = false, 1 = true */ | |
194 | int rejected; | |
195 | unsigned ws_rule; | |
196 | int lines_added, lines_deleted; | |
197 | int score; | |
198 | unsigned int is_toplevel_relative:1; | |
199 | unsigned int inaccurate_eof:1; | |
200 | unsigned int is_binary:1; | |
201 | unsigned int is_copy:1; | |
202 | unsigned int is_rename:1; | |
203 | unsigned int recount:1; | |
204 | unsigned int conflicted_threeway:1; | |
205 | unsigned int direct_to_threeway:1; | |
206 | struct fragment *fragments; | |
207 | char *result; | |
208 | size_t resultsize; | |
209 | char old_sha1_prefix[41]; | |
210 | char new_sha1_prefix[41]; | |
211 | struct patch *next; | |
212 | ||
213 | /* three-way fallback result */ | |
214 | struct object_id threeway_stage[3]; | |
215 | }; | |
216 | ||
217 | static void free_fragment_list(struct fragment *list) | |
218 | { | |
219 | while (list) { | |
220 | struct fragment *next = list->next; | |
221 | if (list->free_patch) | |
222 | free((char *)list->patch); | |
223 | free(list); | |
224 | list = next; | |
225 | } | |
226 | } | |
227 | ||
228 | static void free_patch(struct patch *patch) | |
229 | { | |
230 | free_fragment_list(patch->fragments); | |
231 | free(patch->def_name); | |
232 | free(patch->old_name); | |
233 | free(patch->new_name); | |
234 | free(patch->result); | |
235 | free(patch); | |
236 | } | |
237 | ||
238 | static void free_patch_list(struct patch *list) | |
239 | { | |
240 | while (list) { | |
241 | struct patch *next = list->next; | |
242 | free_patch(list); | |
243 | list = next; | |
244 | } | |
245 | } | |
246 | ||
247 | /* | |
248 | * A line in a file, len-bytes long (includes the terminating LF, | |
249 | * except for an incomplete line at the end if the file ends with | |
250 | * one), and its contents hashes to 'hash'. | |
251 | */ | |
252 | struct line { | |
253 | size_t len; | |
254 | unsigned hash : 24; | |
255 | unsigned flag : 8; | |
256 | #define LINE_COMMON 1 | |
257 | #define LINE_PATCHED 2 | |
258 | }; | |
259 | ||
260 | /* | |
261 | * This represents a "file", which is an array of "lines". | |
262 | */ | |
263 | struct image { | |
264 | char *buf; | |
265 | size_t len; | |
266 | size_t nr; | |
267 | size_t alloc; | |
268 | struct line *line_allocated; | |
269 | struct line *line; | |
270 | }; | |
271 | ||
272 | static uint32_t hash_line(const char *cp, size_t len) | |
273 | { | |
274 | size_t i; | |
275 | uint32_t h; | |
276 | for (i = 0, h = 0; i < len; i++) { | |
277 | if (!isspace(cp[i])) { | |
278 | h = h * 3 + (cp[i] & 0xff); | |
279 | } | |
280 | } | |
281 | return h; | |
282 | } | |
283 | ||
284 | /* | |
285 | * Compare lines s1 of length n1 and s2 of length n2, ignoring | |
286 | * whitespace difference. Returns 1 if they match, 0 otherwise | |
287 | */ | |
288 | static int fuzzy_matchlines(const char *s1, size_t n1, | |
289 | const char *s2, size_t n2) | |
290 | { | |
291 | const char *last1 = s1 + n1 - 1; | |
292 | const char *last2 = s2 + n2 - 1; | |
293 | int result = 0; | |
294 | ||
295 | /* ignore line endings */ | |
296 | while ((*last1 == '\r') || (*last1 == '\n')) | |
297 | last1--; | |
298 | while ((*last2 == '\r') || (*last2 == '\n')) | |
299 | last2--; | |
300 | ||
301 | /* skip leading whitespaces, if both begin with whitespace */ | |
302 | if (s1 <= last1 && s2 <= last2 && isspace(*s1) && isspace(*s2)) { | |
303 | while (isspace(*s1) && (s1 <= last1)) | |
304 | s1++; | |
305 | while (isspace(*s2) && (s2 <= last2)) | |
306 | s2++; | |
307 | } | |
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 | ||
339 | static 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 | ||
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 | */ | |
353 | static void prepare_image(struct image *image, char *buf, size_t len, | |
354 | int prepare_linetable) | |
355 | { | |
356 | const char *cp, *ep; | |
357 | ||
358 | memset(image, 0, sizeof(*image)); | |
359 | image->buf = buf; | |
360 | image->len = len; | |
361 | ||
362 | if (!prepare_linetable) | |
363 | return; | |
364 | ||
365 | ep = image->buf + image->len; | |
366 | cp = image->buf; | |
367 | while (cp < ep) { | |
368 | const char *next; | |
369 | for (next = cp; next < ep && *next != '\n'; next++) | |
370 | ; | |
371 | if (next < ep) | |
372 | next++; | |
373 | add_line_info(image, cp, next - cp, 0); | |
374 | cp = next; | |
375 | } | |
376 | image->line = image->line_allocated; | |
377 | } | |
378 | ||
379 | static void clear_image(struct image *image) | |
380 | { | |
381 | free(image->buf); | |
382 | free(image->line_allocated); | |
383 | memset(image, 0, sizeof(*image)); | |
384 | } | |
385 | ||
386 | /* fmt must contain _one_ %s and no other substitution */ | |
387 | static void say_patch_name(FILE *output, const char *fmt, struct patch *patch) | |
388 | { | |
389 | struct strbuf sb = STRBUF_INIT; | |
390 | ||
391 | if (patch->old_name && patch->new_name && | |
392 | strcmp(patch->old_name, patch->new_name)) { | |
393 | quote_c_style(patch->old_name, &sb, NULL, 0); | |
394 | strbuf_addstr(&sb, " => "); | |
395 | quote_c_style(patch->new_name, &sb, NULL, 0); | |
396 | } else { | |
397 | const char *n = patch->new_name; | |
398 | if (!n) | |
399 | n = patch->old_name; | |
400 | quote_c_style(n, &sb, NULL, 0); | |
401 | } | |
402 | fprintf(output, fmt, sb.buf); | |
403 | fputc('\n', output); | |
404 | strbuf_release(&sb); | |
405 | } | |
406 | ||
407 | #define SLOP (16) | |
408 | ||
409 | static int read_patch_file(struct strbuf *sb, int fd) | |
410 | { | |
411 | if (strbuf_read(sb, fd, 0) < 0) | |
412 | return error_errno("git apply: failed to read"); | |
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 | */ | |
419 | strbuf_grow(sb, SLOP); | |
420 | memset(sb->buf + sb->len, 0, SLOP); | |
421 | return 0; | |
422 | } | |
423 | ||
424 | static unsigned long linelen(const char *buffer, unsigned long size) | |
425 | { | |
426 | unsigned long len = 0; | |
427 | while (size--) { | |
428 | len++; | |
429 | if (*buffer++ == '\n') | |
430 | break; | |
431 | } | |
432 | return len; | |
433 | } | |
434 | ||
435 | static int is_dev_null(const char *str) | |
436 | { | |
437 | return skip_prefix(str, "/dev/null", &str) && isspace(*str); | |
438 | } | |
439 | ||
440 | #define TERM_SPACE 1 | |
441 | #define TERM_TAB 2 | |
442 | ||
443 | static int name_terminate(int c, int terminate) | |
444 | { | |
445 | if (c == ' ' && !(terminate & TERM_SPACE)) | |
446 | return 0; | |
447 | if (c == '\t' && !(terminate & TERM_TAB)) | |
448 | return 0; | |
449 | ||
450 | return 1; | |
451 | } | |
452 | ||
453 | /* remove double slashes to make --index work with such filenames */ | |
454 | static char *squash_slash(char *name) | |
455 | { | |
456 | int i = 0, j = 0; | |
457 | ||
458 | if (!name) | |
459 | return NULL; | |
460 | ||
461 | while (name[i]) { | |
462 | if ((name[j++] = name[i++]) == '/') | |
463 | while (name[i] == '/') | |
464 | i++; | |
465 | } | |
466 | name[j] = '\0'; | |
467 | return name; | |
468 | } | |
469 | ||
470 | static char *find_name_gnu(struct apply_state *state, | |
471 | const char *line, | |
472 | const char *def, | |
473 | int p_value) | |
474 | { | |
475 | struct strbuf name = STRBUF_INIT; | |
476 | char *cp; | |
477 | ||
478 | /* | |
479 | * Proposed "new-style" GNU patch/diff format; see | |
480 | * http://marc.info/?l=git&m=112927316408690&w=2 | |
481 | */ | |
482 | if (unquote_c_style(&name, line, NULL)) { | |
483 | strbuf_release(&name); | |
484 | return NULL; | |
485 | } | |
486 | ||
487 | for (cp = name.buf; p_value; p_value--) { | |
488 | cp = strchr(cp, '/'); | |
489 | if (!cp) { | |
490 | strbuf_release(&name); | |
491 | return NULL; | |
492 | } | |
493 | cp++; | |
494 | } | |
495 | ||
496 | strbuf_remove(&name, 0, cp - name.buf); | |
497 | if (state->root.len) | |
498 | strbuf_insert(&name, 0, state->root.buf, state->root.len); | |
499 | return squash_slash(strbuf_detach(&name, NULL)); | |
500 | } | |
501 | ||
502 | static size_t sane_tz_len(const char *line, size_t len) | |
503 | { | |
504 | const char *tz, *p; | |
505 | ||
506 | if (len < strlen(" +0500") || line[len-strlen(" +0500")] != ' ') | |
507 | return 0; | |
508 | tz = line + len - strlen(" +0500"); | |
509 | ||
510 | if (tz[1] != '+' && tz[1] != '-') | |
511 | return 0; | |
512 | ||
513 | for (p = tz + 2; p != line + len; p++) | |
514 | if (!isdigit(*p)) | |
515 | return 0; | |
516 | ||
517 | return line + len - tz; | |
518 | } | |
519 | ||
520 | static size_t tz_with_colon_len(const char *line, size_t len) | |
521 | { | |
522 | const char *tz, *p; | |
523 | ||
524 | if (len < strlen(" +08:00") || line[len - strlen(":00")] != ':') | |
525 | return 0; | |
526 | tz = line + len - strlen(" +08:00"); | |
527 | ||
528 | if (tz[0] != ' ' || (tz[1] != '+' && tz[1] != '-')) | |
529 | return 0; | |
530 | p = tz + 2; | |
531 | if (!isdigit(*p++) || !isdigit(*p++) || *p++ != ':' || | |
532 | !isdigit(*p++) || !isdigit(*p++)) | |
533 | return 0; | |
534 | ||
535 | return line + len - tz; | |
536 | } | |
537 | ||
538 | static size_t date_len(const char *line, size_t len) | |
539 | { | |
540 | const char *date, *p; | |
541 | ||
542 | if (len < strlen("72-02-05") || line[len-strlen("-05")] != '-') | |
543 | return 0; | |
544 | p = date = line + len - strlen("72-02-05"); | |
545 | ||
546 | if (!isdigit(*p++) || !isdigit(*p++) || *p++ != '-' || | |
547 | !isdigit(*p++) || !isdigit(*p++) || *p++ != '-' || | |
548 | !isdigit(*p++) || !isdigit(*p++)) /* Not a date. */ | |
549 | return 0; | |
550 | ||
551 | if (date - line >= strlen("19") && | |
552 | isdigit(date[-1]) && isdigit(date[-2])) /* 4-digit year */ | |
553 | date -= strlen("19"); | |
554 | ||
555 | return line + len - date; | |
556 | } | |
557 | ||
558 | static size_t short_time_len(const char *line, size_t len) | |
559 | { | |
560 | const char *time, *p; | |
561 | ||
562 | if (len < strlen(" 07:01:32") || line[len-strlen(":32")] != ':') | |
563 | return 0; | |
564 | p = time = line + len - strlen(" 07:01:32"); | |
565 | ||
566 | /* Permit 1-digit hours? */ | |
567 | if (*p++ != ' ' || | |
568 | !isdigit(*p++) || !isdigit(*p++) || *p++ != ':' || | |
569 | !isdigit(*p++) || !isdigit(*p++) || *p++ != ':' || | |
570 | !isdigit(*p++) || !isdigit(*p++)) /* Not a time. */ | |
571 | return 0; | |
572 | ||
573 | return line + len - time; | |
574 | } | |
575 | ||
576 | static size_t fractional_time_len(const char *line, size_t len) | |
577 | { | |
578 | const char *p; | |
579 | size_t n; | |
580 | ||
581 | /* Expected format: 19:41:17.620000023 */ | |
582 | if (!len || !isdigit(line[len - 1])) | |
583 | return 0; | |
584 | p = line + len - 1; | |
585 | ||
586 | /* Fractional seconds. */ | |
587 | while (p > line && isdigit(*p)) | |
588 | p--; | |
589 | if (*p != '.') | |
590 | return 0; | |
591 | ||
592 | /* Hours, minutes, and whole seconds. */ | |
593 | n = short_time_len(line, p - line); | |
594 | if (!n) | |
595 | return 0; | |
596 | ||
597 | return line + len - p + n; | |
598 | } | |
599 | ||
600 | static size_t trailing_spaces_len(const char *line, size_t len) | |
601 | { | |
602 | const char *p; | |
603 | ||
604 | /* Expected format: ' ' x (1 or more) */ | |
605 | if (!len || line[len - 1] != ' ') | |
606 | return 0; | |
607 | ||
608 | p = line + len; | |
609 | while (p != line) { | |
610 | p--; | |
611 | if (*p != ' ') | |
612 | return line + len - (p + 1); | |
613 | } | |
614 | ||
615 | /* All spaces! */ | |
616 | return len; | |
617 | } | |
618 | ||
619 | static size_t diff_timestamp_len(const char *line, size_t len) | |
620 | { | |
621 | const char *end = line + len; | |
622 | size_t n; | |
623 | ||
624 | /* | |
625 | * Posix: 2010-07-05 19:41:17 | |
626 | * GNU: 2010-07-05 19:41:17.620000023 -0500 | |
627 | */ | |
628 | ||
629 | if (!isdigit(end[-1])) | |
630 | return 0; | |
631 | ||
632 | n = sane_tz_len(line, end - line); | |
633 | if (!n) | |
634 | n = tz_with_colon_len(line, end - line); | |
635 | end -= n; | |
636 | ||
637 | n = short_time_len(line, end - line); | |
638 | if (!n) | |
639 | n = fractional_time_len(line, end - line); | |
640 | end -= n; | |
641 | ||
642 | n = date_len(line, end - line); | |
643 | if (!n) /* No date. Too bad. */ | |
644 | return 0; | |
645 | end -= n; | |
646 | ||
647 | if (end == line) /* No space before date. */ | |
648 | return 0; | |
649 | if (end[-1] == '\t') { /* Success! */ | |
650 | end--; | |
651 | return line + len - end; | |
652 | } | |
653 | if (end[-1] != ' ') /* No space before date. */ | |
654 | return 0; | |
655 | ||
656 | /* Whitespace damage. */ | |
657 | end -= trailing_spaces_len(line, end - line); | |
658 | return line + len - end; | |
659 | } | |
660 | ||
661 | static char *find_name_common(struct apply_state *state, | |
662 | const char *line, | |
663 | const char *def, | |
664 | int p_value, | |
665 | const char *end, | |
666 | int terminate) | |
667 | { | |
668 | int len; | |
669 | const char *start = NULL; | |
670 | ||
671 | if (p_value == 0) | |
672 | start = line; | |
673 | while (line != end) { | |
674 | char c = *line; | |
675 | ||
676 | if (!end && isspace(c)) { | |
677 | if (c == '\n') | |
678 | break; | |
679 | if (name_terminate(c, terminate)) | |
680 | break; | |
681 | } | |
682 | line++; | |
683 | if (c == '/' && !--p_value) | |
684 | start = line; | |
685 | } | |
686 | if (!start) | |
687 | return squash_slash(xstrdup_or_null(def)); | |
688 | len = line - start; | |
689 | if (!len) | |
690 | return squash_slash(xstrdup_or_null(def)); | |
691 | ||
692 | /* | |
693 | * Generally we prefer the shorter name, especially | |
694 | * if the other one is just a variation of that with | |
695 | * something else tacked on to the end (ie "file.orig" | |
696 | * or "file~"). | |
697 | */ | |
698 | if (def) { | |
699 | int deflen = strlen(def); | |
700 | if (deflen < len && !strncmp(start, def, deflen)) | |
701 | return squash_slash(xstrdup(def)); | |
702 | } | |
703 | ||
704 | if (state->root.len) { | |
705 | char *ret = xstrfmt("%s%.*s", state->root.buf, len, start); | |
706 | return squash_slash(ret); | |
707 | } | |
708 | ||
709 | return squash_slash(xmemdupz(start, len)); | |
710 | } | |
711 | ||
712 | static char *find_name(struct apply_state *state, | |
713 | const char *line, | |
714 | char *def, | |
715 | int p_value, | |
716 | int terminate) | |
717 | { | |
718 | if (*line == '"') { | |
719 | char *name = find_name_gnu(state, line, def, p_value); | |
720 | if (name) | |
721 | return name; | |
722 | } | |
723 | ||
724 | return find_name_common(state, line, def, p_value, NULL, terminate); | |
725 | } | |
726 | ||
727 | static char *find_name_traditional(struct apply_state *state, | |
728 | const char *line, | |
729 | char *def, | |
730 | int p_value) | |
731 | { | |
732 | size_t len; | |
733 | size_t date_len; | |
734 | ||
735 | if (*line == '"') { | |
736 | char *name = find_name_gnu(state, line, def, p_value); | |
737 | if (name) | |
738 | return name; | |
739 | } | |
740 | ||
741 | len = strchrnul(line, '\n') - line; | |
742 | date_len = diff_timestamp_len(line, len); | |
743 | if (!date_len) | |
744 | return find_name_common(state, line, def, p_value, NULL, TERM_TAB); | |
745 | len -= date_len; | |
746 | ||
747 | return find_name_common(state, line, def, p_value, line + len, 0); | |
748 | } | |
749 | ||
750 | static int count_slashes(const char *cp) | |
751 | { | |
752 | int cnt = 0; | |
753 | char ch; | |
754 | ||
755 | while ((ch = *cp++)) | |
756 | if (ch == '/') | |
757 | cnt++; | |
758 | return cnt; | |
759 | } | |
760 | ||
761 | /* | |
762 | * Given the string after "--- " or "+++ ", guess the appropriate | |
763 | * p_value for the given patch. | |
764 | */ | |
765 | static int guess_p_value(struct apply_state *state, const char *nameline) | |
766 | { | |
767 | char *name, *cp; | |
768 | int val = -1; | |
769 | ||
770 | if (is_dev_null(nameline)) | |
771 | return -1; | |
772 | name = find_name_traditional(state, nameline, NULL, 0); | |
773 | if (!name) | |
774 | return -1; | |
775 | cp = strchr(name, '/'); | |
776 | if (!cp) | |
777 | val = 0; | |
778 | else if (state->prefix) { | |
779 | /* | |
780 | * Does it begin with "a/$our-prefix" and such? Then this is | |
781 | * very likely to apply to our directory. | |
782 | */ | |
783 | if (!strncmp(name, state->prefix, state->prefix_length)) | |
784 | val = count_slashes(state->prefix); | |
785 | else { | |
786 | cp++; | |
787 | if (!strncmp(cp, state->prefix, state->prefix_length)) | |
788 | val = count_slashes(state->prefix) + 1; | |
789 | } | |
790 | } | |
791 | free(name); | |
792 | return val; | |
793 | } | |
794 | ||
795 | /* | |
796 | * Does the ---/+++ line have the POSIX timestamp after the last HT? | |
797 | * GNU diff puts epoch there to signal a creation/deletion event. Is | |
798 | * this such a timestamp? | |
799 | */ | |
800 | static int has_epoch_timestamp(const char *nameline) | |
801 | { | |
802 | /* | |
803 | * We are only interested in epoch timestamp; any non-zero | |
804 | * fraction cannot be one, hence "(\.0+)?" in the regexp below. | |
805 | * For the same reason, the date must be either 1969-12-31 or | |
806 | * 1970-01-01, and the seconds part must be "00". | |
807 | */ | |
808 | const char stamp_regexp[] = | |
809 | "^(1969-12-31|1970-01-01)" | |
810 | " " | |
811 | "[0-2][0-9]:[0-5][0-9]:00(\\.0+)?" | |
812 | " " | |
813 | "([-+][0-2][0-9]:?[0-5][0-9])\n"; | |
814 | const char *timestamp = NULL, *cp, *colon; | |
815 | static regex_t *stamp; | |
816 | regmatch_t m[10]; | |
817 | int zoneoffset; | |
818 | int hourminute; | |
819 | int status; | |
820 | ||
821 | for (cp = nameline; *cp != '\n'; cp++) { | |
822 | if (*cp == '\t') | |
823 | timestamp = cp + 1; | |
824 | } | |
825 | if (!timestamp) | |
826 | return 0; | |
827 | if (!stamp) { | |
828 | stamp = xmalloc(sizeof(*stamp)); | |
829 | if (regcomp(stamp, stamp_regexp, REG_EXTENDED)) { | |
830 | warning(_("Cannot prepare timestamp regexp %s"), | |
831 | stamp_regexp); | |
832 | return 0; | |
833 | } | |
834 | } | |
835 | ||
836 | status = regexec(stamp, timestamp, ARRAY_SIZE(m), m, 0); | |
837 | if (status) { | |
838 | if (status != REG_NOMATCH) | |
839 | warning(_("regexec returned %d for input: %s"), | |
840 | status, timestamp); | |
841 | return 0; | |
842 | } | |
843 | ||
844 | zoneoffset = strtol(timestamp + m[3].rm_so + 1, (char **) &colon, 10); | |
845 | if (*colon == ':') | |
846 | zoneoffset = zoneoffset * 60 + strtol(colon + 1, NULL, 10); | |
847 | else | |
848 | zoneoffset = (zoneoffset / 100) * 60 + (zoneoffset % 100); | |
849 | if (timestamp[m[3].rm_so] == '-') | |
850 | zoneoffset = -zoneoffset; | |
851 | ||
852 | /* | |
853 | * YYYY-MM-DD hh:mm:ss must be from either 1969-12-31 | |
854 | * (west of GMT) or 1970-01-01 (east of GMT) | |
855 | */ | |
856 | if ((zoneoffset < 0 && memcmp(timestamp, "1969-12-31", 10)) || | |
857 | (0 <= zoneoffset && memcmp(timestamp, "1970-01-01", 10))) | |
858 | return 0; | |
859 | ||
860 | hourminute = (strtol(timestamp + 11, NULL, 10) * 60 + | |
861 | strtol(timestamp + 14, NULL, 10) - | |
862 | zoneoffset); | |
863 | ||
864 | return ((zoneoffset < 0 && hourminute == 1440) || | |
865 | (0 <= zoneoffset && !hourminute)); | |
866 | } | |
867 | ||
868 | /* | |
869 | * Get the name etc info from the ---/+++ lines of a traditional patch header | |
870 | * | |
871 | * FIXME! The end-of-filename heuristics are kind of screwy. For existing | |
872 | * files, we can happily check the index for a match, but for creating a | |
873 | * new file we should try to match whatever "patch" does. I have no idea. | |
874 | */ | |
875 | static int parse_traditional_patch(struct apply_state *state, | |
876 | const char *first, | |
877 | const char *second, | |
878 | struct patch *patch) | |
879 | { | |
880 | char *name; | |
881 | ||
882 | first += 4; /* skip "--- " */ | |
883 | second += 4; /* skip "+++ " */ | |
884 | if (!state->p_value_known) { | |
885 | int p, q; | |
886 | p = guess_p_value(state, first); | |
887 | q = guess_p_value(state, second); | |
888 | if (p < 0) p = q; | |
889 | if (0 <= p && p == q) { | |
890 | state->p_value = p; | |
891 | state->p_value_known = 1; | |
892 | } | |
893 | } | |
894 | if (is_dev_null(first)) { | |
895 | patch->is_new = 1; | |
896 | patch->is_delete = 0; | |
897 | name = find_name_traditional(state, second, NULL, state->p_value); | |
898 | patch->new_name = name; | |
899 | } else if (is_dev_null(second)) { | |
900 | patch->is_new = 0; | |
901 | patch->is_delete = 1; | |
902 | name = find_name_traditional(state, first, NULL, state->p_value); | |
903 | patch->old_name = name; | |
904 | } else { | |
905 | char *first_name; | |
906 | first_name = find_name_traditional(state, first, NULL, state->p_value); | |
907 | name = find_name_traditional(state, second, first_name, state->p_value); | |
908 | free(first_name); | |
909 | if (has_epoch_timestamp(first)) { | |
910 | patch->is_new = 1; | |
911 | patch->is_delete = 0; | |
912 | patch->new_name = name; | |
913 | } else if (has_epoch_timestamp(second)) { | |
914 | patch->is_new = 0; | |
915 | patch->is_delete = 1; | |
916 | patch->old_name = name; | |
917 | } else { | |
918 | patch->old_name = name; | |
919 | patch->new_name = xstrdup_or_null(name); | |
920 | } | |
921 | } | |
922 | if (!name) | |
923 | return error(_("unable to find filename in patch at line %d"), state->linenr); | |
924 | ||
925 | return 0; | |
926 | } | |
927 | ||
928 | static int gitdiff_hdrend(struct apply_state *state, | |
929 | const char *line, | |
930 | struct patch *patch) | |
931 | { | |
932 | return 1; | |
933 | } | |
934 | ||
935 | /* | |
936 | * We're anal about diff header consistency, to make | |
937 | * sure that we don't end up having strange ambiguous | |
938 | * patches floating around. | |
939 | * | |
940 | * As a result, gitdiff_{old|new}name() will check | |
941 | * their names against any previous information, just | |
942 | * to make sure.. | |
943 | */ | |
944 | #define DIFF_OLD_NAME 0 | |
945 | #define DIFF_NEW_NAME 1 | |
946 | ||
947 | static int gitdiff_verify_name(struct apply_state *state, | |
948 | const char *line, | |
949 | int isnull, | |
950 | char **name, | |
951 | int side) | |
952 | { | |
953 | if (!*name && !isnull) { | |
954 | *name = find_name(state, line, NULL, state->p_value, TERM_TAB); | |
955 | return 0; | |
956 | } | |
957 | ||
958 | if (*name) { | |
959 | int len = strlen(*name); | |
960 | char *another; | |
961 | if (isnull) | |
962 | return error(_("git apply: bad git-diff - expected /dev/null, got %s on line %d"), | |
963 | *name, state->linenr); | |
964 | another = find_name(state, line, NULL, state->p_value, TERM_TAB); | |
965 | if (!another || memcmp(another, *name, len + 1)) { | |
966 | free(another); | |
967 | return error((side == DIFF_NEW_NAME) ? | |
968 | _("git apply: bad git-diff - inconsistent new filename on line %d") : | |
969 | _("git apply: bad git-diff - inconsistent old filename on line %d"), state->linenr); | |
970 | } | |
971 | free(another); | |
972 | } else { | |
973 | /* expect "/dev/null" */ | |
974 | if (memcmp("/dev/null", line, 9) || line[9] != '\n') | |
975 | return error(_("git apply: bad git-diff - expected /dev/null on line %d"), state->linenr); | |
976 | } | |
977 | ||
978 | return 0; | |
979 | } | |
980 | ||
981 | static int gitdiff_oldname(struct apply_state *state, | |
982 | const char *line, | |
983 | struct patch *patch) | |
984 | { | |
985 | return gitdiff_verify_name(state, line, | |
986 | patch->is_new, &patch->old_name, | |
987 | DIFF_OLD_NAME); | |
988 | } | |
989 | ||
990 | static int gitdiff_newname(struct apply_state *state, | |
991 | const char *line, | |
992 | struct patch *patch) | |
993 | { | |
994 | return gitdiff_verify_name(state, line, | |
995 | patch->is_delete, &patch->new_name, | |
996 | DIFF_NEW_NAME); | |
997 | } | |
998 | ||
999 | static int gitdiff_oldmode(struct apply_state *state, | |
1000 | const char *line, | |
1001 | struct patch *patch) | |
1002 | { | |
1003 | patch->old_mode = strtoul(line, NULL, 8); | |
1004 | return 0; | |
1005 | } | |
1006 | ||
1007 | static int gitdiff_newmode(struct apply_state *state, | |
1008 | const char *line, | |
1009 | struct patch *patch) | |
1010 | { | |
1011 | patch->new_mode = strtoul(line, NULL, 8); | |
1012 | return 0; | |
1013 | } | |
1014 | ||
1015 | static int gitdiff_delete(struct apply_state *state, | |
1016 | const char *line, | |
1017 | struct patch *patch) | |
1018 | { | |
1019 | patch->is_delete = 1; | |
1020 | free(patch->old_name); | |
1021 | patch->old_name = xstrdup_or_null(patch->def_name); | |
1022 | return gitdiff_oldmode(state, line, patch); | |
1023 | } | |
1024 | ||
1025 | static int gitdiff_newfile(struct apply_state *state, | |
1026 | const char *line, | |
1027 | struct patch *patch) | |
1028 | { | |
1029 | patch->is_new = 1; | |
1030 | free(patch->new_name); | |
1031 | patch->new_name = xstrdup_or_null(patch->def_name); | |
1032 | return gitdiff_newmode(state, line, patch); | |
1033 | } | |
1034 | ||
1035 | static int gitdiff_copysrc(struct apply_state *state, | |
1036 | const char *line, | |
1037 | struct patch *patch) | |
1038 | { | |
1039 | patch->is_copy = 1; | |
1040 | free(patch->old_name); | |
1041 | patch->old_name = find_name(state, line, NULL, state->p_value ? state->p_value - 1 : 0, 0); | |
1042 | return 0; | |
1043 | } | |
1044 | ||
1045 | static int gitdiff_copydst(struct apply_state *state, | |
1046 | const char *line, | |
1047 | struct patch *patch) | |
1048 | { | |
1049 | patch->is_copy = 1; | |
1050 | free(patch->new_name); | |
1051 | patch->new_name = find_name(state, line, NULL, state->p_value ? state->p_value - 1 : 0, 0); | |
1052 | return 0; | |
1053 | } | |
1054 | ||
1055 | static int gitdiff_renamesrc(struct apply_state *state, | |
1056 | const char *line, | |
1057 | struct patch *patch) | |
1058 | { | |
1059 | patch->is_rename = 1; | |
1060 | free(patch->old_name); | |
1061 | patch->old_name = find_name(state, line, NULL, state->p_value ? state->p_value - 1 : 0, 0); | |
1062 | return 0; | |
1063 | } | |
1064 | ||
1065 | static int gitdiff_renamedst(struct apply_state *state, | |
1066 | const char *line, | |
1067 | struct patch *patch) | |
1068 | { | |
1069 | patch->is_rename = 1; | |
1070 | free(patch->new_name); | |
1071 | patch->new_name = find_name(state, line, NULL, state->p_value ? state->p_value - 1 : 0, 0); | |
1072 | return 0; | |
1073 | } | |
1074 | ||
1075 | static int gitdiff_similarity(struct apply_state *state, | |
1076 | const char *line, | |
1077 | struct patch *patch) | |
1078 | { | |
1079 | unsigned long val = strtoul(line, NULL, 10); | |
1080 | if (val <= 100) | |
1081 | patch->score = val; | |
1082 | return 0; | |
1083 | } | |
1084 | ||
1085 | static int gitdiff_dissimilarity(struct apply_state *state, | |
1086 | const char *line, | |
1087 | struct patch *patch) | |
1088 | { | |
1089 | unsigned long val = strtoul(line, NULL, 10); | |
1090 | if (val <= 100) | |
1091 | patch->score = val; | |
1092 | return 0; | |
1093 | } | |
1094 | ||
1095 | static int gitdiff_index(struct apply_state *state, | |
1096 | const char *line, | |
1097 | struct patch *patch) | |
1098 | { | |
1099 | /* | |
1100 | * index line is N hexadecimal, "..", N hexadecimal, | |
1101 | * and optional space with octal mode. | |
1102 | */ | |
1103 | const char *ptr, *eol; | |
1104 | int len; | |
1105 | ||
1106 | ptr = strchr(line, '.'); | |
1107 | if (!ptr || ptr[1] != '.' || 40 < ptr - line) | |
1108 | return 0; | |
1109 | len = ptr - line; | |
1110 | memcpy(patch->old_sha1_prefix, line, len); | |
1111 | patch->old_sha1_prefix[len] = 0; | |
1112 | ||
1113 | line = ptr + 2; | |
1114 | ptr = strchr(line, ' '); | |
1115 | eol = strchrnul(line, '\n'); | |
1116 | ||
1117 | if (!ptr || eol < ptr) | |
1118 | ptr = eol; | |
1119 | len = ptr - line; | |
1120 | ||
1121 | if (40 < len) | |
1122 | return 0; | |
1123 | memcpy(patch->new_sha1_prefix, line, len); | |
1124 | patch->new_sha1_prefix[len] = 0; | |
1125 | if (*ptr == ' ') | |
1126 | patch->old_mode = strtoul(ptr+1, NULL, 8); | |
1127 | return 0; | |
1128 | } | |
1129 | ||
1130 | /* | |
1131 | * This is normal for a diff that doesn't change anything: we'll fall through | |
1132 | * into the next diff. Tell the parser to break out. | |
1133 | */ | |
1134 | static int gitdiff_unrecognized(struct apply_state *state, | |
1135 | const char *line, | |
1136 | struct patch *patch) | |
1137 | { | |
1138 | return 1; | |
1139 | } | |
1140 | ||
1141 | /* | |
1142 | * Skip p_value leading components from "line"; as we do not accept | |
1143 | * absolute paths, return NULL in that case. | |
1144 | */ | |
1145 | static const char *skip_tree_prefix(struct apply_state *state, | |
1146 | const char *line, | |
1147 | int llen) | |
1148 | { | |
1149 | int nslash; | |
1150 | int i; | |
1151 | ||
1152 | if (!state->p_value) | |
1153 | return (llen && line[0] == '/') ? NULL : line; | |
1154 | ||
1155 | nslash = state->p_value; | |
1156 | for (i = 0; i < llen; i++) { | |
1157 | int ch = line[i]; | |
1158 | if (ch == '/' && --nslash <= 0) | |
1159 | return (i == 0) ? NULL : &line[i + 1]; | |
1160 | } | |
1161 | return NULL; | |
1162 | } | |
1163 | ||
1164 | /* | |
1165 | * This is to extract the same name that appears on "diff --git" | |
1166 | * line. We do not find and return anything if it is a rename | |
1167 | * patch, and it is OK because we will find the name elsewhere. | |
1168 | * We need to reliably find name only when it is mode-change only, | |
1169 | * creation or deletion of an empty file. In any of these cases, | |
1170 | * both sides are the same name under a/ and b/ respectively. | |
1171 | */ | |
1172 | static char *git_header_name(struct apply_state *state, | |
1173 | const char *line, | |
1174 | int llen) | |
1175 | { | |
1176 | const char *name; | |
1177 | const char *second = NULL; | |
1178 | size_t len, line_len; | |
1179 | ||
1180 | line += strlen("diff --git "); | |
1181 | llen -= strlen("diff --git "); | |
1182 | ||
1183 | if (*line == '"') { | |
1184 | const char *cp; | |
1185 | struct strbuf first = STRBUF_INIT; | |
1186 | struct strbuf sp = STRBUF_INIT; | |
1187 | ||
1188 | if (unquote_c_style(&first, line, &second)) | |
1189 | goto free_and_fail1; | |
1190 | ||
1191 | /* strip the a/b prefix including trailing slash */ | |
1192 | cp = skip_tree_prefix(state, first.buf, first.len); | |
1193 | if (!cp) | |
1194 | goto free_and_fail1; | |
1195 | strbuf_remove(&first, 0, cp - first.buf); | |
1196 | ||
1197 | /* | |
1198 | * second points at one past closing dq of name. | |
1199 | * find the second name. | |
1200 | */ | |
1201 | while ((second < line + llen) && isspace(*second)) | |
1202 | second++; | |
1203 | ||
1204 | if (line + llen <= second) | |
1205 | goto free_and_fail1; | |
1206 | if (*second == '"') { | |
1207 | if (unquote_c_style(&sp, second, NULL)) | |
1208 | goto free_and_fail1; | |
1209 | cp = skip_tree_prefix(state, sp.buf, sp.len); | |
1210 | if (!cp) | |
1211 | goto free_and_fail1; | |
1212 | /* They must match, otherwise ignore */ | |
1213 | if (strcmp(cp, first.buf)) | |
1214 | goto free_and_fail1; | |
1215 | strbuf_release(&sp); | |
1216 | return strbuf_detach(&first, NULL); | |
1217 | } | |
1218 | ||
1219 | /* unquoted second */ | |
1220 | cp = skip_tree_prefix(state, second, line + llen - second); | |
1221 | if (!cp) | |
1222 | goto free_and_fail1; | |
1223 | if (line + llen - cp != first.len || | |
1224 | memcmp(first.buf, cp, first.len)) | |
1225 | goto free_and_fail1; | |
1226 | return strbuf_detach(&first, NULL); | |
1227 | ||
1228 | free_and_fail1: | |
1229 | strbuf_release(&first); | |
1230 | strbuf_release(&sp); | |
1231 | return NULL; | |
1232 | } | |
1233 | ||
1234 | /* unquoted first name */ | |
1235 | name = skip_tree_prefix(state, line, llen); | |
1236 | if (!name) | |
1237 | return NULL; | |
1238 | ||
1239 | /* | |
1240 | * since the first name is unquoted, a dq if exists must be | |
1241 | * the beginning of the second name. | |
1242 | */ | |
1243 | for (second = name; second < line + llen; second++) { | |
1244 | if (*second == '"') { | |
1245 | struct strbuf sp = STRBUF_INIT; | |
1246 | const char *np; | |
1247 | ||
1248 | if (unquote_c_style(&sp, second, NULL)) | |
1249 | goto free_and_fail2; | |
1250 | ||
1251 | np = skip_tree_prefix(state, sp.buf, sp.len); | |
1252 | if (!np) | |
1253 | goto free_and_fail2; | |
1254 | ||
1255 | len = sp.buf + sp.len - np; | |
1256 | if (len < second - name && | |
1257 | !strncmp(np, name, len) && | |
1258 | isspace(name[len])) { | |
1259 | /* Good */ | |
1260 | strbuf_remove(&sp, 0, np - sp.buf); | |
1261 | return strbuf_detach(&sp, NULL); | |
1262 | } | |
1263 | ||
1264 | free_and_fail2: | |
1265 | strbuf_release(&sp); | |
1266 | return NULL; | |
1267 | } | |
1268 | } | |
1269 | ||
1270 | /* | |
1271 | * Accept a name only if it shows up twice, exactly the same | |
1272 | * form. | |
1273 | */ | |
1274 | second = strchr(name, '\n'); | |
1275 | if (!second) | |
1276 | return NULL; | |
1277 | line_len = second - name; | |
1278 | for (len = 0 ; ; len++) { | |
1279 | switch (name[len]) { | |
1280 | default: | |
1281 | continue; | |
1282 | case '\n': | |
1283 | return NULL; | |
1284 | case '\t': case ' ': | |
1285 | /* | |
1286 | * Is this the separator between the preimage | |
1287 | * and the postimage pathname? Again, we are | |
1288 | * only interested in the case where there is | |
1289 | * no rename, as this is only to set def_name | |
1290 | * and a rename patch has the names elsewhere | |
1291 | * in an unambiguous form. | |
1292 | */ | |
1293 | if (!name[len + 1]) | |
1294 | return NULL; /* no postimage name */ | |
1295 | second = skip_tree_prefix(state, name + len + 1, | |
1296 | line_len - (len + 1)); | |
1297 | if (!second) | |
1298 | return NULL; | |
1299 | /* | |
1300 | * Does len bytes starting at "name" and "second" | |
1301 | * (that are separated by one HT or SP we just | |
1302 | * found) exactly match? | |
1303 | */ | |
1304 | if (second[len] == '\n' && !strncmp(name, second, len)) | |
1305 | return xmemdupz(name, len); | |
1306 | } | |
1307 | } | |
1308 | } | |
1309 | ||
1310 | /* Verify that we recognize the lines following a git header */ | |
1311 | static int parse_git_header(struct apply_state *state, | |
1312 | const char *line, | |
1313 | int len, | |
1314 | unsigned int size, | |
1315 | struct patch *patch) | |
1316 | { | |
1317 | unsigned long offset; | |
1318 | ||
1319 | /* A git diff has explicit new/delete information, so we don't guess */ | |
1320 | patch->is_new = 0; | |
1321 | patch->is_delete = 0; | |
1322 | ||
1323 | /* | |
1324 | * Some things may not have the old name in the | |
1325 | * rest of the headers anywhere (pure mode changes, | |
1326 | * or removing or adding empty files), so we get | |
1327 | * the default name from the header. | |
1328 | */ | |
1329 | patch->def_name = git_header_name(state, line, len); | |
1330 | if (patch->def_name && state->root.len) { | |
1331 | char *s = xstrfmt("%s%s", state->root.buf, patch->def_name); | |
1332 | free(patch->def_name); | |
1333 | patch->def_name = s; | |
1334 | } | |
1335 | ||
1336 | line += len; | |
1337 | size -= len; | |
1338 | state->linenr++; | |
1339 | for (offset = len ; size > 0 ; offset += len, size -= len, line += len, state->linenr++) { | |
1340 | static const struct opentry { | |
1341 | const char *str; | |
1342 | int (*fn)(struct apply_state *, const char *, struct patch *); | |
1343 | } optable[] = { | |
1344 | { "@@ -", gitdiff_hdrend }, | |
1345 | { "--- ", gitdiff_oldname }, | |
1346 | { "+++ ", gitdiff_newname }, | |
1347 | { "old mode ", gitdiff_oldmode }, | |
1348 | { "new mode ", gitdiff_newmode }, | |
1349 | { "deleted file mode ", gitdiff_delete }, | |
1350 | { "new file mode ", gitdiff_newfile }, | |
1351 | { "copy from ", gitdiff_copysrc }, | |
1352 | { "copy to ", gitdiff_copydst }, | |
1353 | { "rename old ", gitdiff_renamesrc }, | |
1354 | { "rename new ", gitdiff_renamedst }, | |
1355 | { "rename from ", gitdiff_renamesrc }, | |
1356 | { "rename to ", gitdiff_renamedst }, | |
1357 | { "similarity index ", gitdiff_similarity }, | |
1358 | { "dissimilarity index ", gitdiff_dissimilarity }, | |
1359 | { "index ", gitdiff_index }, | |
1360 | { "", gitdiff_unrecognized }, | |
1361 | }; | |
1362 | int i; | |
1363 | ||
1364 | len = linelen(line, size); | |
1365 | if (!len || line[len-1] != '\n') | |
1366 | break; | |
1367 | for (i = 0; i < ARRAY_SIZE(optable); i++) { | |
1368 | const struct opentry *p = optable + i; | |
1369 | int oplen = strlen(p->str); | |
1370 | int res; | |
1371 | if (len < oplen || memcmp(p->str, line, oplen)) | |
1372 | continue; | |
1373 | res = p->fn(state, line + oplen, patch); | |
1374 | if (res < 0) | |
1375 | return -1; | |
1376 | if (res > 0) | |
1377 | return offset; | |
1378 | break; | |
1379 | } | |
1380 | } | |
1381 | ||
1382 | return offset; | |
1383 | } | |
1384 | ||
1385 | static int parse_num(const char *line, unsigned long *p) | |
1386 | { | |
1387 | char *ptr; | |
1388 | ||
1389 | if (!isdigit(*line)) | |
1390 | return 0; | |
1391 | *p = strtoul(line, &ptr, 10); | |
1392 | return ptr - line; | |
1393 | } | |
1394 | ||
1395 | static int parse_range(const char *line, int len, int offset, const char *expect, | |
1396 | unsigned long *p1, unsigned long *p2) | |
1397 | { | |
1398 | int digits, ex; | |
1399 | ||
1400 | if (offset < 0 || offset >= len) | |
1401 | return -1; | |
1402 | line += offset; | |
1403 | len -= offset; | |
1404 | ||
1405 | digits = parse_num(line, p1); | |
1406 | if (!digits) | |
1407 | return -1; | |
1408 | ||
1409 | offset += digits; | |
1410 | line += digits; | |
1411 | len -= digits; | |
1412 | ||
1413 | *p2 = 1; | |
1414 | if (*line == ',') { | |
1415 | digits = parse_num(line+1, p2); | |
1416 | if (!digits) | |
1417 | return -1; | |
1418 | ||
1419 | offset += digits+1; | |
1420 | line += digits+1; | |
1421 | len -= digits+1; | |
1422 | } | |
1423 | ||
1424 | ex = strlen(expect); | |
1425 | if (ex > len) | |
1426 | return -1; | |
1427 | if (memcmp(line, expect, ex)) | |
1428 | return -1; | |
1429 | ||
1430 | return offset + ex; | |
1431 | } | |
1432 | ||
1433 | static void recount_diff(const char *line, int size, struct fragment *fragment) | |
1434 | { | |
1435 | int oldlines = 0, newlines = 0, ret = 0; | |
1436 | ||
1437 | if (size < 1) { | |
1438 | warning("recount: ignore empty hunk"); | |
1439 | return; | |
1440 | } | |
1441 | ||
1442 | for (;;) { | |
1443 | int len = linelen(line, size); | |
1444 | size -= len; | |
1445 | line += len; | |
1446 | ||
1447 | if (size < 1) | |
1448 | break; | |
1449 | ||
1450 | switch (*line) { | |
1451 | case ' ': case '\n': | |
1452 | newlines++; | |
1453 | /* fall through */ | |
1454 | case '-': | |
1455 | oldlines++; | |
1456 | continue; | |
1457 | case '+': | |
1458 | newlines++; | |
1459 | continue; | |
1460 | case '\\': | |
1461 | continue; | |
1462 | case '@': | |
1463 | ret = size < 3 || !starts_with(line, "@@ "); | |
1464 | break; | |
1465 | case 'd': | |
1466 | ret = size < 5 || !starts_with(line, "diff "); | |
1467 | break; | |
1468 | default: | |
1469 | ret = -1; | |
1470 | break; | |
1471 | } | |
1472 | if (ret) { | |
1473 | warning(_("recount: unexpected line: %.*s"), | |
1474 | (int)linelen(line, size), line); | |
1475 | return; | |
1476 | } | |
1477 | break; | |
1478 | } | |
1479 | fragment->oldlines = oldlines; | |
1480 | fragment->newlines = newlines; | |
1481 | } | |
1482 | ||
1483 | /* | |
1484 | * Parse a unified diff fragment header of the | |
1485 | * form "@@ -a,b +c,d @@" | |
1486 | */ | |
1487 | static int parse_fragment_header(const char *line, int len, struct fragment *fragment) | |
1488 | { | |
1489 | int offset; | |
1490 | ||
1491 | if (!len || line[len-1] != '\n') | |
1492 | return -1; | |
1493 | ||
1494 | /* Figure out the number of lines in a fragment */ | |
1495 | offset = parse_range(line, len, 4, " +", &fragment->oldpos, &fragment->oldlines); | |
1496 | offset = parse_range(line, len, offset, " @@", &fragment->newpos, &fragment->newlines); | |
1497 | ||
1498 | return offset; | |
1499 | } | |
1500 | ||
1501 | /* | |
1502 | * Find file diff header | |
1503 | * | |
1504 | * Returns: | |
1505 | * -1 if no header was found | |
1506 | * -128 in case of error | |
1507 | * the size of the header in bytes (called "offset") otherwise | |
1508 | */ | |
1509 | static int find_header(struct apply_state *state, | |
1510 | const char *line, | |
1511 | unsigned long size, | |
1512 | int *hdrsize, | |
1513 | struct patch *patch) | |
1514 | { | |
1515 | unsigned long offset, len; | |
1516 | ||
1517 | patch->is_toplevel_relative = 0; | |
1518 | patch->is_rename = patch->is_copy = 0; | |
1519 | patch->is_new = patch->is_delete = -1; | |
1520 | patch->old_mode = patch->new_mode = 0; | |
1521 | patch->old_name = patch->new_name = NULL; | |
1522 | for (offset = 0; size > 0; offset += len, size -= len, line += len, state->linenr++) { | |
1523 | unsigned long nextlen; | |
1524 | ||
1525 | len = linelen(line, size); | |
1526 | if (!len) | |
1527 | break; | |
1528 | ||
1529 | /* Testing this early allows us to take a few shortcuts.. */ | |
1530 | if (len < 6) | |
1531 | continue; | |
1532 | ||
1533 | /* | |
1534 | * Make sure we don't find any unconnected patch fragments. | |
1535 | * That's a sign that we didn't find a header, and that a | |
1536 | * patch has become corrupted/broken up. | |
1537 | */ | |
1538 | if (!memcmp("@@ -", line, 4)) { | |
1539 | struct fragment dummy; | |
1540 | if (parse_fragment_header(line, len, &dummy) < 0) | |
1541 | continue; | |
1542 | error(_("patch fragment without header at line %d: %.*s"), | |
1543 | state->linenr, (int)len-1, line); | |
1544 | return -128; | |
1545 | } | |
1546 | ||
1547 | if (size < len + 6) | |
1548 | break; | |
1549 | ||
1550 | /* | |
1551 | * Git patch? It might not have a real patch, just a rename | |
1552 | * or mode change, so we handle that specially | |
1553 | */ | |
1554 | if (!memcmp("diff --git ", line, 11)) { | |
1555 | int git_hdr_len = parse_git_header(state, line, len, size, patch); | |
1556 | if (git_hdr_len < 0) | |
1557 | return -128; | |
1558 | if (git_hdr_len <= len) | |
1559 | continue; | |
1560 | if (!patch->old_name && !patch->new_name) { | |
1561 | if (!patch->def_name) { | |
1562 | error(Q_("git diff header lacks filename information when removing " | |
1563 | "%d leading pathname component (line %d)", | |
1564 | "git diff header lacks filename information when removing " | |
1565 | "%d leading pathname components (line %d)", | |
1566 | state->p_value), | |
1567 | state->p_value, state->linenr); | |
1568 | return -128; | |
1569 | } | |
1570 | patch->old_name = xstrdup(patch->def_name); | |
1571 | patch->new_name = xstrdup(patch->def_name); | |
1572 | } | |
1573 | if (!patch->is_delete && !patch->new_name) { | |
1574 | error("git diff header lacks filename information " | |
1575 | "(line %d)", state->linenr); | |
1576 | return -128; | |
1577 | } | |
1578 | patch->is_toplevel_relative = 1; | |
1579 | *hdrsize = git_hdr_len; | |
1580 | return offset; | |
1581 | } | |
1582 | ||
1583 | /* --- followed by +++ ? */ | |
1584 | if (memcmp("--- ", line, 4) || memcmp("+++ ", line + len, 4)) | |
1585 | continue; | |
1586 | ||
1587 | /* | |
1588 | * We only accept unified patches, so we want it to | |
1589 | * at least have "@@ -a,b +c,d @@\n", which is 14 chars | |
1590 | * minimum ("@@ -0,0 +1 @@\n" is the shortest). | |
1591 | */ | |
1592 | nextlen = linelen(line + len, size - len); | |
1593 | if (size < nextlen + 14 || memcmp("@@ -", line + len + nextlen, 4)) | |
1594 | continue; | |
1595 | ||
1596 | /* Ok, we'll consider it a patch */ | |
1597 | if (parse_traditional_patch(state, line, line+len, patch)) | |
1598 | return -128; | |
1599 | *hdrsize = len + nextlen; | |
1600 | state->linenr += 2; | |
1601 | return offset; | |
1602 | } | |
1603 | return -1; | |
1604 | } | |
1605 | ||
1606 | static void record_ws_error(struct apply_state *state, | |
1607 | unsigned result, | |
1608 | const char *line, | |
1609 | int len, | |
1610 | int linenr) | |
1611 | { | |
1612 | char *err; | |
1613 | ||
1614 | if (!result) | |
1615 | return; | |
1616 | ||
1617 | state->whitespace_error++; | |
1618 | if (state->squelch_whitespace_errors && | |
1619 | state->squelch_whitespace_errors < state->whitespace_error) | |
1620 | return; | |
1621 | ||
1622 | err = whitespace_error_string(result); | |
1623 | fprintf(stderr, "%s:%d: %s.\n%.*s\n", | |
1624 | state->patch_input_file, linenr, err, len, line); | |
1625 | free(err); | |
1626 | } | |
1627 | ||
1628 | static void check_whitespace(struct apply_state *state, | |
1629 | const char *line, | |
1630 | int len, | |
1631 | unsigned ws_rule) | |
1632 | { | |
1633 | unsigned result = ws_check(line + 1, len - 1, ws_rule); | |
1634 | ||
1635 | record_ws_error(state, result, line + 1, len - 2, state->linenr); | |
1636 | } | |
1637 | ||
1638 | /* | |
1639 | * Parse a unified diff. Note that this really needs to parse each | |
1640 | * fragment separately, since the only way to know the difference | |
1641 | * between a "---" that is part of a patch, and a "---" that starts | |
1642 | * the next patch is to look at the line counts.. | |
1643 | */ | |
1644 | static int parse_fragment(struct apply_state *state, | |
1645 | const char *line, | |
1646 | unsigned long size, | |
1647 | struct patch *patch, | |
1648 | struct fragment *fragment) | |
1649 | { | |
1650 | int added, deleted; | |
1651 | int len = linelen(line, size), offset; | |
1652 | unsigned long oldlines, newlines; | |
1653 | unsigned long leading, trailing; | |
1654 | ||
1655 | offset = parse_fragment_header(line, len, fragment); | |
1656 | if (offset < 0) | |
1657 | return -1; | |
1658 | if (offset > 0 && patch->recount) | |
1659 | recount_diff(line + offset, size - offset, fragment); | |
1660 | oldlines = fragment->oldlines; | |
1661 | newlines = fragment->newlines; | |
1662 | leading = 0; | |
1663 | trailing = 0; | |
1664 | ||
1665 | /* Parse the thing.. */ | |
1666 | line += len; | |
1667 | size -= len; | |
1668 | state->linenr++; | |
1669 | added = deleted = 0; | |
1670 | for (offset = len; | |
1671 | 0 < size; | |
1672 | offset += len, size -= len, line += len, state->linenr++) { | |
1673 | if (!oldlines && !newlines) | |
1674 | break; | |
1675 | len = linelen(line, size); | |
1676 | if (!len || line[len-1] != '\n') | |
1677 | return -1; | |
1678 | switch (*line) { | |
1679 | default: | |
1680 | return -1; | |
1681 | case '\n': /* newer GNU diff, an empty context line */ | |
1682 | case ' ': | |
1683 | oldlines--; | |
1684 | newlines--; | |
1685 | if (!deleted && !added) | |
1686 | leading++; | |
1687 | trailing++; | |
1688 | if (!state->apply_in_reverse && | |
1689 | state->ws_error_action == correct_ws_error) | |
1690 | check_whitespace(state, line, len, patch->ws_rule); | |
1691 | break; | |
1692 | case '-': | |
1693 | if (state->apply_in_reverse && | |
1694 | state->ws_error_action != nowarn_ws_error) | |
1695 | check_whitespace(state, line, len, patch->ws_rule); | |
1696 | deleted++; | |
1697 | oldlines--; | |
1698 | trailing = 0; | |
1699 | break; | |
1700 | case '+': | |
1701 | if (!state->apply_in_reverse && | |
1702 | state->ws_error_action != nowarn_ws_error) | |
1703 | check_whitespace(state, line, len, patch->ws_rule); | |
1704 | added++; | |
1705 | newlines--; | |
1706 | trailing = 0; | |
1707 | break; | |
1708 | ||
1709 | /* | |
1710 | * We allow "\ No newline at end of file". Depending | |
1711 | * on locale settings when the patch was produced we | |
1712 | * don't know what this line looks like. The only | |
1713 | * thing we do know is that it begins with "\ ". | |
1714 | * Checking for 12 is just for sanity check -- any | |
1715 | * l10n of "\ No newline..." is at least that long. | |
1716 | */ | |
1717 | case '\\': | |
1718 | if (len < 12 || memcmp(line, "\\ ", 2)) | |
1719 | return -1; | |
1720 | break; | |
1721 | } | |
1722 | } | |
1723 | if (oldlines || newlines) | |
1724 | return -1; | |
1725 | if (!deleted && !added) | |
1726 | return -1; | |
1727 | ||
1728 | fragment->leading = leading; | |
1729 | fragment->trailing = trailing; | |
1730 | ||
1731 | /* | |
1732 | * If a fragment ends with an incomplete line, we failed to include | |
1733 | * it in the above loop because we hit oldlines == newlines == 0 | |
1734 | * before seeing it. | |
1735 | */ | |
1736 | if (12 < size && !memcmp(line, "\\ ", 2)) | |
1737 | offset += linelen(line, size); | |
1738 | ||
1739 | patch->lines_added += added; | |
1740 | patch->lines_deleted += deleted; | |
1741 | ||
1742 | if (0 < patch->is_new && oldlines) | |
1743 | return error(_("new file depends on old contents")); | |
1744 | if (0 < patch->is_delete && newlines) | |
1745 | return error(_("deleted file still has contents")); | |
1746 | return offset; | |
1747 | } | |
1748 | ||
1749 | /* | |
1750 | * We have seen "diff --git a/... b/..." header (or a traditional patch | |
1751 | * header). Read hunks that belong to this patch into fragments and hang | |
1752 | * them to the given patch structure. | |
1753 | * | |
1754 | * The (fragment->patch, fragment->size) pair points into the memory given | |
1755 | * by the caller, not a copy, when we return. | |
1756 | * | |
1757 | * Returns: | |
1758 | * -1 in case of error, | |
1759 | * the number of bytes in the patch otherwise. | |
1760 | */ | |
1761 | static int parse_single_patch(struct apply_state *state, | |
1762 | const char *line, | |
1763 | unsigned long size, | |
1764 | struct patch *patch) | |
1765 | { | |
1766 | unsigned long offset = 0; | |
1767 | unsigned long oldlines = 0, newlines = 0, context = 0; | |
1768 | struct fragment **fragp = &patch->fragments; | |
1769 | ||
1770 | while (size > 4 && !memcmp(line, "@@ -", 4)) { | |
1771 | struct fragment *fragment; | |
1772 | int len; | |
1773 | ||
1774 | fragment = xcalloc(1, sizeof(*fragment)); | |
1775 | fragment->linenr = state->linenr; | |
1776 | len = parse_fragment(state, line, size, patch, fragment); | |
1777 | if (len <= 0) { | |
1778 | free(fragment); | |
1779 | return error(_("corrupt patch at line %d"), state->linenr); | |
1780 | } | |
1781 | fragment->patch = line; | |
1782 | fragment->size = len; | |
1783 | oldlines += fragment->oldlines; | |
1784 | newlines += fragment->newlines; | |
1785 | context += fragment->leading + fragment->trailing; | |
1786 | ||
1787 | *fragp = fragment; | |
1788 | fragp = &fragment->next; | |
1789 | ||
1790 | offset += len; | |
1791 | line += len; | |
1792 | size -= len; | |
1793 | } | |
1794 | ||
1795 | /* | |
1796 | * If something was removed (i.e. we have old-lines) it cannot | |
1797 | * be creation, and if something was added it cannot be | |
1798 | * deletion. However, the reverse is not true; --unified=0 | |
1799 | * patches that only add are not necessarily creation even | |
1800 | * though they do not have any old lines, and ones that only | |
1801 | * delete are not necessarily deletion. | |
1802 | * | |
1803 | * Unfortunately, a real creation/deletion patch do _not_ have | |
1804 | * any context line by definition, so we cannot safely tell it | |
1805 | * apart with --unified=0 insanity. At least if the patch has | |
1806 | * more than one hunk it is not creation or deletion. | |
1807 | */ | |
1808 | if (patch->is_new < 0 && | |
1809 | (oldlines || (patch->fragments && patch->fragments->next))) | |
1810 | patch->is_new = 0; | |
1811 | if (patch->is_delete < 0 && | |
1812 | (newlines || (patch->fragments && patch->fragments->next))) | |
1813 | patch->is_delete = 0; | |
1814 | ||
1815 | if (0 < patch->is_new && oldlines) | |
1816 | return error(_("new file %s depends on old contents"), patch->new_name); | |
1817 | if (0 < patch->is_delete && newlines) | |
1818 | return error(_("deleted file %s still has contents"), patch->old_name); | |
1819 | if (!patch->is_delete && !newlines && context) | |
1820 | fprintf_ln(stderr, | |
1821 | _("** warning: " | |
1822 | "file %s becomes empty but is not deleted"), | |
1823 | patch->new_name); | |
1824 | ||
1825 | return offset; | |
1826 | } | |
1827 | ||
1828 | static inline int metadata_changes(struct patch *patch) | |
1829 | { | |
1830 | return patch->is_rename > 0 || | |
1831 | patch->is_copy > 0 || | |
1832 | patch->is_new > 0 || | |
1833 | patch->is_delete || | |
1834 | (patch->old_mode && patch->new_mode && | |
1835 | patch->old_mode != patch->new_mode); | |
1836 | } | |
1837 | ||
1838 | static char *inflate_it(const void *data, unsigned long size, | |
1839 | unsigned long inflated_size) | |
1840 | { | |
1841 | git_zstream stream; | |
1842 | void *out; | |
1843 | int st; | |
1844 | ||
1845 | memset(&stream, 0, sizeof(stream)); | |
1846 | ||
1847 | stream.next_in = (unsigned char *)data; | |
1848 | stream.avail_in = size; | |
1849 | stream.next_out = out = xmalloc(inflated_size); | |
1850 | stream.avail_out = inflated_size; | |
1851 | git_inflate_init(&stream); | |
1852 | st = git_inflate(&stream, Z_FINISH); | |
1853 | git_inflate_end(&stream); | |
1854 | if ((st != Z_STREAM_END) || stream.total_out != inflated_size) { | |
1855 | free(out); | |
1856 | return NULL; | |
1857 | } | |
1858 | return out; | |
1859 | } | |
1860 | ||
1861 | /* | |
1862 | * Read a binary hunk and return a new fragment; fragment->patch | |
1863 | * points at an allocated memory that the caller must free, so | |
1864 | * it is marked as "->free_patch = 1". | |
1865 | */ | |
1866 | static struct fragment *parse_binary_hunk(struct apply_state *state, | |
1867 | char **buf_p, | |
1868 | unsigned long *sz_p, | |
1869 | int *status_p, | |
1870 | int *used_p) | |
1871 | { | |
1872 | /* | |
1873 | * Expect a line that begins with binary patch method ("literal" | |
1874 | * or "delta"), followed by the length of data before deflating. | |
1875 | * a sequence of 'length-byte' followed by base-85 encoded data | |
1876 | * should follow, terminated by a newline. | |
1877 | * | |
1878 | * Each 5-byte sequence of base-85 encodes up to 4 bytes, | |
1879 | * and we would limit the patch line to 66 characters, | |
1880 | * so one line can fit up to 13 groups that would decode | |
1881 | * to 52 bytes max. The length byte 'A'-'Z' corresponds | |
1882 | * to 1-26 bytes, and 'a'-'z' corresponds to 27-52 bytes. | |
1883 | */ | |
1884 | int llen, used; | |
1885 | unsigned long size = *sz_p; | |
1886 | char *buffer = *buf_p; | |
1887 | int patch_method; | |
1888 | unsigned long origlen; | |
1889 | char *data = NULL; | |
1890 | int hunk_size = 0; | |
1891 | struct fragment *frag; | |
1892 | ||
1893 | llen = linelen(buffer, size); | |
1894 | used = llen; | |
1895 | ||
1896 | *status_p = 0; | |
1897 | ||
1898 | if (starts_with(buffer, "delta ")) { | |
1899 | patch_method = BINARY_DELTA_DEFLATED; | |
1900 | origlen = strtoul(buffer + 6, NULL, 10); | |
1901 | } | |
1902 | else if (starts_with(buffer, "literal ")) { | |
1903 | patch_method = BINARY_LITERAL_DEFLATED; | |
1904 | origlen = strtoul(buffer + 8, NULL, 10); | |
1905 | } | |
1906 | else | |
1907 | return NULL; | |
1908 | ||
1909 | state->linenr++; | |
1910 | buffer += llen; | |
1911 | while (1) { | |
1912 | int byte_length, max_byte_length, newsize; | |
1913 | llen = linelen(buffer, size); | |
1914 | used += llen; | |
1915 | state->linenr++; | |
1916 | if (llen == 1) { | |
1917 | /* consume the blank line */ | |
1918 | buffer++; | |
1919 | size--; | |
1920 | break; | |
1921 | } | |
1922 | /* | |
1923 | * Minimum line is "A00000\n" which is 7-byte long, | |
1924 | * and the line length must be multiple of 5 plus 2. | |
1925 | */ | |
1926 | if ((llen < 7) || (llen-2) % 5) | |
1927 | goto corrupt; | |
1928 | max_byte_length = (llen - 2) / 5 * 4; | |
1929 | byte_length = *buffer; | |
1930 | if ('A' <= byte_length && byte_length <= 'Z') | |
1931 | byte_length = byte_length - 'A' + 1; | |
1932 | else if ('a' <= byte_length && byte_length <= 'z') | |
1933 | byte_length = byte_length - 'a' + 27; | |
1934 | else | |
1935 | goto corrupt; | |
1936 | /* if the input length was not multiple of 4, we would | |
1937 | * have filler at the end but the filler should never | |
1938 | * exceed 3 bytes | |
1939 | */ | |
1940 | if (max_byte_length < byte_length || | |
1941 | byte_length <= max_byte_length - 4) | |
1942 | goto corrupt; | |
1943 | newsize = hunk_size + byte_length; | |
1944 | data = xrealloc(data, newsize); | |
1945 | if (decode_85(data + hunk_size, buffer + 1, byte_length)) | |
1946 | goto corrupt; | |
1947 | hunk_size = newsize; | |
1948 | buffer += llen; | |
1949 | size -= llen; | |
1950 | } | |
1951 | ||
1952 | frag = xcalloc(1, sizeof(*frag)); | |
1953 | frag->patch = inflate_it(data, hunk_size, origlen); | |
1954 | frag->free_patch = 1; | |
1955 | if (!frag->patch) | |
1956 | goto corrupt; | |
1957 | free(data); | |
1958 | frag->size = origlen; | |
1959 | *buf_p = buffer; | |
1960 | *sz_p = size; | |
1961 | *used_p = used; | |
1962 | frag->binary_patch_method = patch_method; | |
1963 | return frag; | |
1964 | ||
1965 | corrupt: | |
1966 | free(data); | |
1967 | *status_p = -1; | |
1968 | error(_("corrupt binary patch at line %d: %.*s"), | |
1969 | state->linenr-1, llen-1, buffer); | |
1970 | return NULL; | |
1971 | } | |
1972 | ||
1973 | /* | |
1974 | * Returns: | |
1975 | * -1 in case of error, | |
1976 | * the length of the parsed binary patch otherwise | |
1977 | */ | |
1978 | static int parse_binary(struct apply_state *state, | |
1979 | char *buffer, | |
1980 | unsigned long size, | |
1981 | struct patch *patch) | |
1982 | { | |
1983 | /* | |
1984 | * We have read "GIT binary patch\n"; what follows is a line | |
1985 | * that says the patch method (currently, either "literal" or | |
1986 | * "delta") and the length of data before deflating; a | |
1987 | * sequence of 'length-byte' followed by base-85 encoded data | |
1988 | * follows. | |
1989 | * | |
1990 | * When a binary patch is reversible, there is another binary | |
1991 | * hunk in the same format, starting with patch method (either | |
1992 | * "literal" or "delta") with the length of data, and a sequence | |
1993 | * of length-byte + base-85 encoded data, terminated with another | |
1994 | * empty line. This data, when applied to the postimage, produces | |
1995 | * the preimage. | |
1996 | */ | |
1997 | struct fragment *forward; | |
1998 | struct fragment *reverse; | |
1999 | int status; | |
2000 | int used, used_1; | |
2001 | ||
2002 | forward = parse_binary_hunk(state, &buffer, &size, &status, &used); | |
2003 | if (!forward && !status) | |
2004 | /* there has to be one hunk (forward hunk) */ | |
2005 | return error(_("unrecognized binary patch at line %d"), state->linenr-1); | |
2006 | if (status) | |
2007 | /* otherwise we already gave an error message */ | |
2008 | return status; | |
2009 | ||
2010 | reverse = parse_binary_hunk(state, &buffer, &size, &status, &used_1); | |
2011 | if (reverse) | |
2012 | used += used_1; | |
2013 | else if (status) { | |
2014 | /* | |
2015 | * Not having reverse hunk is not an error, but having | |
2016 | * a corrupt reverse hunk is. | |
2017 | */ | |
2018 | free((void*) forward->patch); | |
2019 | free(forward); | |
2020 | return status; | |
2021 | } | |
2022 | forward->next = reverse; | |
2023 | patch->fragments = forward; | |
2024 | patch->is_binary = 1; | |
2025 | return used; | |
2026 | } | |
2027 | ||
2028 | static void prefix_one(struct apply_state *state, char **name) | |
2029 | { | |
2030 | char *old_name = *name; | |
2031 | if (!old_name) | |
2032 | return; | |
2033 | *name = xstrdup(prefix_filename(state->prefix, state->prefix_length, *name)); | |
2034 | free(old_name); | |
2035 | } | |
2036 | ||
2037 | static void prefix_patch(struct apply_state *state, struct patch *p) | |
2038 | { | |
2039 | if (!state->prefix || p->is_toplevel_relative) | |
2040 | return; | |
2041 | prefix_one(state, &p->new_name); | |
2042 | prefix_one(state, &p->old_name); | |
2043 | } | |
2044 | ||
2045 | /* | |
2046 | * include/exclude | |
2047 | */ | |
2048 | ||
2049 | static void add_name_limit(struct apply_state *state, | |
2050 | const char *name, | |
2051 | int exclude) | |
2052 | { | |
2053 | struct string_list_item *it; | |
2054 | ||
2055 | it = string_list_append(&state->limit_by_name, name); | |
2056 | it->util = exclude ? NULL : (void *) 1; | |
2057 | } | |
2058 | ||
2059 | static int use_patch(struct apply_state *state, struct patch *p) | |
2060 | { | |
2061 | const char *pathname = p->new_name ? p->new_name : p->old_name; | |
2062 | int i; | |
2063 | ||
2064 | /* Paths outside are not touched regardless of "--include" */ | |
2065 | if (0 < state->prefix_length) { | |
2066 | int pathlen = strlen(pathname); | |
2067 | if (pathlen <= state->prefix_length || | |
2068 | memcmp(state->prefix, pathname, state->prefix_length)) | |
2069 | return 0; | |
2070 | } | |
2071 | ||
2072 | /* See if it matches any of exclude/include rule */ | |
2073 | for (i = 0; i < state->limit_by_name.nr; i++) { | |
2074 | struct string_list_item *it = &state->limit_by_name.items[i]; | |
2075 | if (!wildmatch(it->string, pathname, 0, NULL)) | |
2076 | return (it->util != NULL); | |
2077 | } | |
2078 | ||
2079 | /* | |
2080 | * If we had any include, a path that does not match any rule is | |
2081 | * not used. Otherwise, we saw bunch of exclude rules (or none) | |
2082 | * and such a path is used. | |
2083 | */ | |
2084 | return !state->has_include; | |
2085 | } | |
2086 | ||
2087 | /* | |
2088 | * Read the patch text in "buffer" that extends for "size" bytes; stop | |
2089 | * reading after seeing a single patch (i.e. changes to a single file). | |
2090 | * Create fragments (i.e. patch hunks) and hang them to the given patch. | |
2091 | * | |
2092 | * Returns: | |
2093 | * -1 if no header was found or parse_binary() failed, | |
2094 | * -128 on another error, | |
2095 | * the number of bytes consumed otherwise, | |
2096 | * so that the caller can call us again for the next patch. | |
2097 | */ | |
2098 | static int parse_chunk(struct apply_state *state, char *buffer, unsigned long size, struct patch *patch) | |
2099 | { | |
2100 | int hdrsize, patchsize; | |
2101 | int offset = find_header(state, buffer, size, &hdrsize, patch); | |
2102 | ||
2103 | if (offset < 0) | |
2104 | return offset; | |
2105 | ||
2106 | prefix_patch(state, patch); | |
2107 | ||
2108 | if (!use_patch(state, patch)) | |
2109 | patch->ws_rule = 0; | |
2110 | else | |
2111 | patch->ws_rule = whitespace_rule(patch->new_name | |
2112 | ? patch->new_name | |
2113 | : patch->old_name); | |
2114 | ||
2115 | patchsize = parse_single_patch(state, | |
2116 | buffer + offset + hdrsize, | |
2117 | size - offset - hdrsize, | |
2118 | patch); | |
2119 | ||
2120 | if (patchsize < 0) | |
2121 | return -128; | |
2122 | ||
2123 | if (!patchsize) { | |
2124 | static const char git_binary[] = "GIT binary patch\n"; | |
2125 | int hd = hdrsize + offset; | |
2126 | unsigned long llen = linelen(buffer + hd, size - hd); | |
2127 | ||
2128 | if (llen == sizeof(git_binary) - 1 && | |
2129 | !memcmp(git_binary, buffer + hd, llen)) { | |
2130 | int used; | |
2131 | state->linenr++; | |
2132 | used = parse_binary(state, buffer + hd + llen, | |
2133 | size - hd - llen, patch); | |
2134 | if (used < 0) | |
2135 | return -1; | |
2136 | if (used) | |
2137 | patchsize = used + llen; | |
2138 | else | |
2139 | patchsize = 0; | |
2140 | } | |
2141 | else if (!memcmp(" differ\n", buffer + hd + llen - 8, 8)) { | |
2142 | static const char *binhdr[] = { | |
2143 | "Binary files ", | |
2144 | "Files ", | |
2145 | NULL, | |
2146 | }; | |
2147 | int i; | |
2148 | for (i = 0; binhdr[i]; i++) { | |
2149 | int len = strlen(binhdr[i]); | |
2150 | if (len < size - hd && | |
2151 | !memcmp(binhdr[i], buffer + hd, len)) { | |
2152 | state->linenr++; | |
2153 | patch->is_binary = 1; | |
2154 | patchsize = llen; | |
2155 | break; | |
2156 | } | |
2157 | } | |
2158 | } | |
2159 | ||
2160 | /* Empty patch cannot be applied if it is a text patch | |
2161 | * without metadata change. A binary patch appears | |
2162 | * empty to us here. | |
2163 | */ | |
2164 | if ((state->apply || state->check) && | |
2165 | (!patch->is_binary && !metadata_changes(patch))) { | |
2166 | error(_("patch with only garbage at line %d"), state->linenr); | |
2167 | return -128; | |
2168 | } | |
2169 | } | |
2170 | ||
2171 | return offset + hdrsize + patchsize; | |
2172 | } | |
2173 | ||
2174 | #define swap(a,b) myswap((a),(b),sizeof(a)) | |
2175 | ||
2176 | #define myswap(a, b, size) do { \ | |
2177 | unsigned char mytmp[size]; \ | |
2178 | memcpy(mytmp, &a, size); \ | |
2179 | memcpy(&a, &b, size); \ | |
2180 | memcpy(&b, mytmp, size); \ | |
2181 | } while (0) | |
2182 | ||
2183 | static void reverse_patches(struct patch *p) | |
2184 | { | |
2185 | for (; p; p = p->next) { | |
2186 | struct fragment *frag = p->fragments; | |
2187 | ||
2188 | swap(p->new_name, p->old_name); | |
2189 | swap(p->new_mode, p->old_mode); | |
2190 | swap(p->is_new, p->is_delete); | |
2191 | swap(p->lines_added, p->lines_deleted); | |
2192 | swap(p->old_sha1_prefix, p->new_sha1_prefix); | |
2193 | ||
2194 | for (; frag; frag = frag->next) { | |
2195 | swap(frag->newpos, frag->oldpos); | |
2196 | swap(frag->newlines, frag->oldlines); | |
2197 | } | |
2198 | } | |
2199 | } | |
2200 | ||
2201 | static const char pluses[] = | |
2202 | "++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++"; | |
2203 | static const char minuses[]= | |
2204 | "----------------------------------------------------------------------"; | |
2205 | ||
2206 | static void show_stats(struct apply_state *state, struct patch *patch) | |
2207 | { | |
2208 | struct strbuf qname = STRBUF_INIT; | |
2209 | char *cp = patch->new_name ? patch->new_name : patch->old_name; | |
2210 | int max, add, del; | |
2211 | ||
2212 | quote_c_style(cp, &qname, NULL, 0); | |
2213 | ||
2214 | /* | |
2215 | * "scale" the filename | |
2216 | */ | |
2217 | max = state->max_len; | |
2218 | if (max > 50) | |
2219 | max = 50; | |
2220 | ||
2221 | if (qname.len > max) { | |
2222 | cp = strchr(qname.buf + qname.len + 3 - max, '/'); | |
2223 | if (!cp) | |
2224 | cp = qname.buf + qname.len + 3 - max; | |
2225 | strbuf_splice(&qname, 0, cp - qname.buf, "...", 3); | |
2226 | } | |
2227 | ||
2228 | if (patch->is_binary) { | |
2229 | printf(" %-*s | Bin\n", max, qname.buf); | |
2230 | strbuf_release(&qname); | |
2231 | return; | |
2232 | } | |
2233 | ||
2234 | printf(" %-*s |", max, qname.buf); | |
2235 | strbuf_release(&qname); | |
2236 | ||
2237 | /* | |
2238 | * scale the add/delete | |
2239 | */ | |
2240 | max = max + state->max_change > 70 ? 70 - max : state->max_change; | |
2241 | add = patch->lines_added; | |
2242 | del = patch->lines_deleted; | |
2243 | ||
2244 | if (state->max_change > 0) { | |
2245 | int total = ((add + del) * max + state->max_change / 2) / state->max_change; | |
2246 | add = (add * max + state->max_change / 2) / state->max_change; | |
2247 | del = total - add; | |
2248 | } | |
2249 | printf("%5d %.*s%.*s\n", patch->lines_added + patch->lines_deleted, | |
2250 | add, pluses, del, minuses); | |
2251 | } | |
2252 | ||
2253 | static int read_old_data(struct stat *st, const char *path, struct strbuf *buf) | |
2254 | { | |
2255 | switch (st->st_mode & S_IFMT) { | |
2256 | case S_IFLNK: | |
2257 | if (strbuf_readlink(buf, path, st->st_size) < 0) | |
2258 | return error(_("unable to read symlink %s"), path); | |
2259 | return 0; | |
2260 | case S_IFREG: | |
2261 | if (strbuf_read_file(buf, path, st->st_size) != st->st_size) | |
2262 | return error(_("unable to open or read %s"), path); | |
2263 | convert_to_git(path, buf->buf, buf->len, buf, 0); | |
2264 | return 0; | |
2265 | default: | |
2266 | return -1; | |
2267 | } | |
2268 | } | |
2269 | ||
2270 | /* | |
2271 | * Update the preimage, and the common lines in postimage, | |
2272 | * from buffer buf of length len. If postlen is 0 the postimage | |
2273 | * is updated in place, otherwise it's updated on a new buffer | |
2274 | * of length postlen | |
2275 | */ | |
2276 | ||
2277 | static void update_pre_post_images(struct image *preimage, | |
2278 | struct image *postimage, | |
2279 | char *buf, | |
2280 | size_t len, size_t postlen) | |
2281 | { | |
2282 | int i, ctx, reduced; | |
2283 | char *new, *old, *fixed; | |
2284 | struct image fixed_preimage; | |
2285 | ||
2286 | /* | |
2287 | * Update the preimage with whitespace fixes. Note that we | |
2288 | * are not losing preimage->buf -- apply_one_fragment() will | |
2289 | * free "oldlines". | |
2290 | */ | |
2291 | prepare_image(&fixed_preimage, buf, len, 1); | |
2292 | assert(postlen | |
2293 | ? fixed_preimage.nr == preimage->nr | |
2294 | : fixed_preimage.nr <= preimage->nr); | |
2295 | for (i = 0; i < fixed_preimage.nr; i++) | |
2296 | fixed_preimage.line[i].flag = preimage->line[i].flag; | |
2297 | free(preimage->line_allocated); | |
2298 | *preimage = fixed_preimage; | |
2299 | ||
2300 | /* | |
2301 | * Adjust the common context lines in postimage. This can be | |
2302 | * done in-place when we are shrinking it with whitespace | |
2303 | * fixing, but needs a new buffer when ignoring whitespace or | |
2304 | * expanding leading tabs to spaces. | |
2305 | * | |
2306 | * We trust the caller to tell us if the update can be done | |
2307 | * in place (postlen==0) or not. | |
2308 | */ | |
2309 | old = postimage->buf; | |
2310 | if (postlen) | |
2311 | new = postimage->buf = xmalloc(postlen); | |
2312 | else | |
2313 | new = old; | |
2314 | fixed = preimage->buf; | |
2315 | ||
2316 | for (i = reduced = ctx = 0; i < postimage->nr; i++) { | |
2317 | size_t l_len = postimage->line[i].len; | |
2318 | if (!(postimage->line[i].flag & LINE_COMMON)) { | |
2319 | /* an added line -- no counterparts in preimage */ | |
2320 | memmove(new, old, l_len); | |
2321 | old += l_len; | |
2322 | new += l_len; | |
2323 | continue; | |
2324 | } | |
2325 | ||
2326 | /* a common context -- skip it in the original postimage */ | |
2327 | old += l_len; | |
2328 | ||
2329 | /* and find the corresponding one in the fixed preimage */ | |
2330 | while (ctx < preimage->nr && | |
2331 | !(preimage->line[ctx].flag & LINE_COMMON)) { | |
2332 | fixed += preimage->line[ctx].len; | |
2333 | ctx++; | |
2334 | } | |
2335 | ||
2336 | /* | |
2337 | * preimage is expected to run out, if the caller | |
2338 | * fixed addition of trailing blank lines. | |
2339 | */ | |
2340 | if (preimage->nr <= ctx) { | |
2341 | reduced++; | |
2342 | continue; | |
2343 | } | |
2344 | ||
2345 | /* and copy it in, while fixing the line length */ | |
2346 | l_len = preimage->line[ctx].len; | |
2347 | memcpy(new, fixed, l_len); | |
2348 | new += l_len; | |
2349 | fixed += l_len; | |
2350 | postimage->line[i].len = l_len; | |
2351 | ctx++; | |
2352 | } | |
2353 | ||
2354 | if (postlen | |
2355 | ? postlen < new - postimage->buf | |
2356 | : postimage->len < new - postimage->buf) | |
2357 | die("BUG: caller miscounted postlen: asked %d, orig = %d, used = %d", | |
2358 | (int)postlen, (int) postimage->len, (int)(new - postimage->buf)); | |
2359 | ||
2360 | /* Fix the length of the whole thing */ | |
2361 | postimage->len = new - postimage->buf; | |
2362 | postimage->nr -= reduced; | |
2363 | } | |
2364 | ||
2365 | static int line_by_line_fuzzy_match(struct image *img, | |
2366 | struct image *preimage, | |
2367 | struct image *postimage, | |
2368 | unsigned long try, | |
2369 | int try_lno, | |
2370 | int preimage_limit) | |
2371 | { | |
2372 | int i; | |
2373 | size_t imgoff = 0; | |
2374 | size_t preoff = 0; | |
2375 | size_t postlen = postimage->len; | |
2376 | size_t extra_chars; | |
2377 | char *buf; | |
2378 | char *preimage_eof; | |
2379 | char *preimage_end; | |
2380 | struct strbuf fixed; | |
2381 | char *fixed_buf; | |
2382 | size_t fixed_len; | |
2383 | ||
2384 | for (i = 0; i < preimage_limit; i++) { | |
2385 | size_t prelen = preimage->line[i].len; | |
2386 | size_t imglen = img->line[try_lno+i].len; | |
2387 | ||
2388 | if (!fuzzy_matchlines(img->buf + try + imgoff, imglen, | |
2389 | preimage->buf + preoff, prelen)) | |
2390 | return 0; | |
2391 | if (preimage->line[i].flag & LINE_COMMON) | |
2392 | postlen += imglen - prelen; | |
2393 | imgoff += imglen; | |
2394 | preoff += prelen; | |
2395 | } | |
2396 | ||
2397 | /* | |
2398 | * Ok, the preimage matches with whitespace fuzz. | |
2399 | * | |
2400 | * imgoff now holds the true length of the target that | |
2401 | * matches the preimage before the end of the file. | |
2402 | * | |
2403 | * Count the number of characters in the preimage that fall | |
2404 | * beyond the end of the file and make sure that all of them | |
2405 | * are whitespace characters. (This can only happen if | |
2406 | * we are removing blank lines at the end of the file.) | |
2407 | */ | |
2408 | buf = preimage_eof = preimage->buf + preoff; | |
2409 | for ( ; i < preimage->nr; i++) | |
2410 | preoff += preimage->line[i].len; | |
2411 | preimage_end = preimage->buf + preoff; | |
2412 | for ( ; buf < preimage_end; buf++) | |
2413 | if (!isspace(*buf)) | |
2414 | return 0; | |
2415 | ||
2416 | /* | |
2417 | * Update the preimage and the common postimage context | |
2418 | * lines to use the same whitespace as the target. | |
2419 | * If whitespace is missing in the target (i.e. | |
2420 | * if the preimage extends beyond the end of the file), | |
2421 | * use the whitespace from the preimage. | |
2422 | */ | |
2423 | extra_chars = preimage_end - preimage_eof; | |
2424 | strbuf_init(&fixed, imgoff + extra_chars); | |
2425 | strbuf_add(&fixed, img->buf + try, imgoff); | |
2426 | strbuf_add(&fixed, preimage_eof, extra_chars); | |
2427 | fixed_buf = strbuf_detach(&fixed, &fixed_len); | |
2428 | update_pre_post_images(preimage, postimage, | |
2429 | fixed_buf, fixed_len, postlen); | |
2430 | return 1; | |
2431 | } | |
2432 | ||
2433 | static int match_fragment(struct apply_state *state, | |
2434 | struct image *img, | |
2435 | struct image *preimage, | |
2436 | struct image *postimage, | |
2437 | unsigned long try, | |
2438 | int try_lno, | |
2439 | unsigned ws_rule, | |
2440 | int match_beginning, int match_end) | |
2441 | { | |
2442 | int i; | |
2443 | char *fixed_buf, *buf, *orig, *target; | |
2444 | struct strbuf fixed; | |
2445 | size_t fixed_len, postlen; | |
2446 | int preimage_limit; | |
2447 | ||
2448 | if (preimage->nr + try_lno <= img->nr) { | |
2449 | /* | |
2450 | * The hunk falls within the boundaries of img. | |
2451 | */ | |
2452 | preimage_limit = preimage->nr; | |
2453 | if (match_end && (preimage->nr + try_lno != img->nr)) | |
2454 | return 0; | |
2455 | } else if (state->ws_error_action == correct_ws_error && | |
2456 | (ws_rule & WS_BLANK_AT_EOF)) { | |
2457 | /* | |
2458 | * This hunk extends beyond the end of img, and we are | |
2459 | * removing blank lines at the end of the file. This | |
2460 | * many lines from the beginning of the preimage must | |
2461 | * match with img, and the remainder of the preimage | |
2462 | * must be blank. | |
2463 | */ | |
2464 | preimage_limit = img->nr - try_lno; | |
2465 | } else { | |
2466 | /* | |
2467 | * The hunk extends beyond the end of the img and | |
2468 | * we are not removing blanks at the end, so we | |
2469 | * should reject the hunk at this position. | |
2470 | */ | |
2471 | return 0; | |
2472 | } | |
2473 | ||
2474 | if (match_beginning && try_lno) | |
2475 | return 0; | |
2476 | ||
2477 | /* Quick hash check */ | |
2478 | for (i = 0; i < preimage_limit; i++) | |
2479 | if ((img->line[try_lno + i].flag & LINE_PATCHED) || | |
2480 | (preimage->line[i].hash != img->line[try_lno + i].hash)) | |
2481 | return 0; | |
2482 | ||
2483 | if (preimage_limit == preimage->nr) { | |
2484 | /* | |
2485 | * Do we have an exact match? If we were told to match | |
2486 | * at the end, size must be exactly at try+fragsize, | |
2487 | * otherwise try+fragsize must be still within the preimage, | |
2488 | * and either case, the old piece should match the preimage | |
2489 | * exactly. | |
2490 | */ | |
2491 | if ((match_end | |
2492 | ? (try + preimage->len == img->len) | |
2493 | : (try + preimage->len <= img->len)) && | |
2494 | !memcmp(img->buf + try, preimage->buf, preimage->len)) | |
2495 | return 1; | |
2496 | } else { | |
2497 | /* | |
2498 | * The preimage extends beyond the end of img, so | |
2499 | * there cannot be an exact match. | |
2500 | * | |
2501 | * There must be one non-blank context line that match | |
2502 | * a line before the end of img. | |
2503 | */ | |
2504 | char *buf_end; | |
2505 | ||
2506 | buf = preimage->buf; | |
2507 | buf_end = buf; | |
2508 | for (i = 0; i < preimage_limit; i++) | |
2509 | buf_end += preimage->line[i].len; | |
2510 | ||
2511 | for ( ; buf < buf_end; buf++) | |
2512 | if (!isspace(*buf)) | |
2513 | break; | |
2514 | if (buf == buf_end) | |
2515 | return 0; | |
2516 | } | |
2517 | ||
2518 | /* | |
2519 | * No exact match. If we are ignoring whitespace, run a line-by-line | |
2520 | * fuzzy matching. We collect all the line length information because | |
2521 | * we need it to adjust whitespace if we match. | |
2522 | */ | |
2523 | if (state->ws_ignore_action == ignore_ws_change) | |
2524 | return line_by_line_fuzzy_match(img, preimage, postimage, | |
2525 | try, try_lno, preimage_limit); | |
2526 | ||
2527 | if (state->ws_error_action != correct_ws_error) | |
2528 | return 0; | |
2529 | ||
2530 | /* | |
2531 | * The hunk does not apply byte-by-byte, but the hash says | |
2532 | * it might with whitespace fuzz. We weren't asked to | |
2533 | * ignore whitespace, we were asked to correct whitespace | |
2534 | * errors, so let's try matching after whitespace correction. | |
2535 | * | |
2536 | * While checking the preimage against the target, whitespace | |
2537 | * errors in both fixed, we count how large the corresponding | |
2538 | * postimage needs to be. The postimage prepared by | |
2539 | * apply_one_fragment() has whitespace errors fixed on added | |
2540 | * lines already, but the common lines were propagated as-is, | |
2541 | * which may become longer when their whitespace errors are | |
2542 | * fixed. | |
2543 | */ | |
2544 | ||
2545 | /* First count added lines in postimage */ | |
2546 | postlen = 0; | |
2547 | for (i = 0; i < postimage->nr; i++) { | |
2548 | if (!(postimage->line[i].flag & LINE_COMMON)) | |
2549 | postlen += postimage->line[i].len; | |
2550 | } | |
2551 | ||
2552 | /* | |
2553 | * The preimage may extend beyond the end of the file, | |
2554 | * but in this loop we will only handle the part of the | |
2555 | * preimage that falls within the file. | |
2556 | */ | |
2557 | strbuf_init(&fixed, preimage->len + 1); | |
2558 | orig = preimage->buf; | |
2559 | target = img->buf + try; | |
2560 | for (i = 0; i < preimage_limit; i++) { | |
2561 | size_t oldlen = preimage->line[i].len; | |
2562 | size_t tgtlen = img->line[try_lno + i].len; | |
2563 | size_t fixstart = fixed.len; | |
2564 | struct strbuf tgtfix; | |
2565 | int match; | |
2566 | ||
2567 | /* Try fixing the line in the preimage */ | |
2568 | ws_fix_copy(&fixed, orig, oldlen, ws_rule, NULL); | |
2569 | ||
2570 | /* Try fixing the line in the target */ | |
2571 | strbuf_init(&tgtfix, tgtlen); | |
2572 | ws_fix_copy(&tgtfix, target, tgtlen, ws_rule, NULL); | |
2573 | ||
2574 | /* | |
2575 | * If they match, either the preimage was based on | |
2576 | * a version before our tree fixed whitespace breakage, | |
2577 | * or we are lacking a whitespace-fix patch the tree | |
2578 | * the preimage was based on already had (i.e. target | |
2579 | * has whitespace breakage, the preimage doesn't). | |
2580 | * In either case, we are fixing the whitespace breakages | |
2581 | * so we might as well take the fix together with their | |
2582 | * real change. | |
2583 | */ | |
2584 | match = (tgtfix.len == fixed.len - fixstart && | |
2585 | !memcmp(tgtfix.buf, fixed.buf + fixstart, | |
2586 | fixed.len - fixstart)); | |
2587 | ||
2588 | /* Add the length if this is common with the postimage */ | |
2589 | if (preimage->line[i].flag & LINE_COMMON) | |
2590 | postlen += tgtfix.len; | |
2591 | ||
2592 | strbuf_release(&tgtfix); | |
2593 | if (!match) | |
2594 | goto unmatch_exit; | |
2595 | ||
2596 | orig += oldlen; | |
2597 | target += tgtlen; | |
2598 | } | |
2599 | ||
2600 | ||
2601 | /* | |
2602 | * Now handle the lines in the preimage that falls beyond the | |
2603 | * end of the file (if any). They will only match if they are | |
2604 | * empty or only contain whitespace (if WS_BLANK_AT_EOL is | |
2605 | * false). | |
2606 | */ | |
2607 | for ( ; i < preimage->nr; i++) { | |
2608 | size_t fixstart = fixed.len; /* start of the fixed preimage */ | |
2609 | size_t oldlen = preimage->line[i].len; | |
2610 | int j; | |
2611 | ||
2612 | /* Try fixing the line in the preimage */ | |
2613 | ws_fix_copy(&fixed, orig, oldlen, ws_rule, NULL); | |
2614 | ||
2615 | for (j = fixstart; j < fixed.len; j++) | |
2616 | if (!isspace(fixed.buf[j])) | |
2617 | goto unmatch_exit; | |
2618 | ||
2619 | orig += oldlen; | |
2620 | } | |
2621 | ||
2622 | /* | |
2623 | * Yes, the preimage is based on an older version that still | |
2624 | * has whitespace breakages unfixed, and fixing them makes the | |
2625 | * hunk match. Update the context lines in the postimage. | |
2626 | */ | |
2627 | fixed_buf = strbuf_detach(&fixed, &fixed_len); | |
2628 | if (postlen < postimage->len) | |
2629 | postlen = 0; | |
2630 | update_pre_post_images(preimage, postimage, | |
2631 | fixed_buf, fixed_len, postlen); | |
2632 | return 1; | |
2633 | ||
2634 | unmatch_exit: | |
2635 | strbuf_release(&fixed); | |
2636 | return 0; | |
2637 | } | |
2638 | ||
2639 | static int find_pos(struct apply_state *state, | |
2640 | struct image *img, | |
2641 | struct image *preimage, | |
2642 | struct image *postimage, | |
2643 | int line, | |
2644 | unsigned ws_rule, | |
2645 | int match_beginning, int match_end) | |
2646 | { | |
2647 | int i; | |
2648 | unsigned long backwards, forwards, try; | |
2649 | int backwards_lno, forwards_lno, try_lno; | |
2650 | ||
2651 | /* | |
2652 | * If match_beginning or match_end is specified, there is no | |
2653 | * point starting from a wrong line that will never match and | |
2654 | * wander around and wait for a match at the specified end. | |
2655 | */ | |
2656 | if (match_beginning) | |
2657 | line = 0; | |
2658 | else if (match_end) | |
2659 | line = img->nr - preimage->nr; | |
2660 | ||
2661 | /* | |
2662 | * Because the comparison is unsigned, the following test | |
2663 | * will also take care of a negative line number that can | |
2664 | * result when match_end and preimage is larger than the target. | |
2665 | */ | |
2666 | if ((size_t) line > img->nr) | |
2667 | line = img->nr; | |
2668 | ||
2669 | try = 0; | |
2670 | for (i = 0; i < line; i++) | |
2671 | try += img->line[i].len; | |
2672 | ||
2673 | /* | |
2674 | * There's probably some smart way to do this, but I'll leave | |
2675 | * that to the smart and beautiful people. I'm simple and stupid. | |
2676 | */ | |
2677 | backwards = try; | |
2678 | backwards_lno = line; | |
2679 | forwards = try; | |
2680 | forwards_lno = line; | |
2681 | try_lno = line; | |
2682 | ||
2683 | for (i = 0; ; i++) { | |
2684 | if (match_fragment(state, img, preimage, postimage, | |
2685 | try, try_lno, ws_rule, | |
2686 | match_beginning, match_end)) | |
2687 | return try_lno; | |
2688 | ||
2689 | again: | |
2690 | if (backwards_lno == 0 && forwards_lno == img->nr) | |
2691 | break; | |
2692 | ||
2693 | if (i & 1) { | |
2694 | if (backwards_lno == 0) { | |
2695 | i++; | |
2696 | goto again; | |
2697 | } | |
2698 | backwards_lno--; | |
2699 | backwards -= img->line[backwards_lno].len; | |
2700 | try = backwards; | |
2701 | try_lno = backwards_lno; | |
2702 | } else { | |
2703 | if (forwards_lno == img->nr) { | |
2704 | i++; | |
2705 | goto again; | |
2706 | } | |
2707 | forwards += img->line[forwards_lno].len; | |
2708 | forwards_lno++; | |
2709 | try = forwards; | |
2710 | try_lno = forwards_lno; | |
2711 | } | |
2712 | ||
2713 | } | |
2714 | return -1; | |
2715 | } | |
2716 | ||
2717 | static void remove_first_line(struct image *img) | |
2718 | { | |
2719 | img->buf += img->line[0].len; | |
2720 | img->len -= img->line[0].len; | |
2721 | img->line++; | |
2722 | img->nr--; | |
2723 | } | |
2724 | ||
2725 | static void remove_last_line(struct image *img) | |
2726 | { | |
2727 | img->len -= img->line[--img->nr].len; | |
2728 | } | |
2729 | ||
2730 | /* | |
2731 | * The change from "preimage" and "postimage" has been found to | |
2732 | * apply at applied_pos (counts in line numbers) in "img". | |
2733 | * Update "img" to remove "preimage" and replace it with "postimage". | |
2734 | */ | |
2735 | static void update_image(struct apply_state *state, | |
2736 | struct image *img, | |
2737 | int applied_pos, | |
2738 | struct image *preimage, | |
2739 | struct image *postimage) | |
2740 | { | |
2741 | /* | |
2742 | * remove the copy of preimage at offset in img | |
2743 | * and replace it with postimage | |
2744 | */ | |
2745 | int i, nr; | |
2746 | size_t remove_count, insert_count, applied_at = 0; | |
2747 | char *result; | |
2748 | int preimage_limit; | |
2749 | ||
2750 | /* | |
2751 | * If we are removing blank lines at the end of img, | |
2752 | * the preimage may extend beyond the end. | |
2753 | * If that is the case, we must be careful only to | |
2754 | * remove the part of the preimage that falls within | |
2755 | * the boundaries of img. Initialize preimage_limit | |
2756 | * to the number of lines in the preimage that falls | |
2757 | * within the boundaries. | |
2758 | */ | |
2759 | preimage_limit = preimage->nr; | |
2760 | if (preimage_limit > img->nr - applied_pos) | |
2761 | preimage_limit = img->nr - applied_pos; | |
2762 | ||
2763 | for (i = 0; i < applied_pos; i++) | |
2764 | applied_at += img->line[i].len; | |
2765 | ||
2766 | remove_count = 0; | |
2767 | for (i = 0; i < preimage_limit; i++) | |
2768 | remove_count += img->line[applied_pos + i].len; | |
2769 | insert_count = postimage->len; | |
2770 | ||
2771 | /* Adjust the contents */ | |
2772 | result = xmalloc(st_add3(st_sub(img->len, remove_count), insert_count, 1)); | |
2773 | memcpy(result, img->buf, applied_at); | |
2774 | memcpy(result + applied_at, postimage->buf, postimage->len); | |
2775 | memcpy(result + applied_at + postimage->len, | |
2776 | img->buf + (applied_at + remove_count), | |
2777 | img->len - (applied_at + remove_count)); | |
2778 | free(img->buf); | |
2779 | img->buf = result; | |
2780 | img->len += insert_count - remove_count; | |
2781 | result[img->len] = '\0'; | |
2782 | ||
2783 | /* Adjust the line table */ | |
2784 | nr = img->nr + postimage->nr - preimage_limit; | |
2785 | if (preimage_limit < postimage->nr) { | |
2786 | /* | |
2787 | * NOTE: this knows that we never call remove_first_line() | |
2788 | * on anything other than pre/post image. | |
2789 | */ | |
2790 | REALLOC_ARRAY(img->line, nr); | |
2791 | img->line_allocated = img->line; | |
2792 | } | |
2793 | if (preimage_limit != postimage->nr) | |
2794 | memmove(img->line + applied_pos + postimage->nr, | |
2795 | img->line + applied_pos + preimage_limit, | |
2796 | (img->nr - (applied_pos + preimage_limit)) * | |
2797 | sizeof(*img->line)); | |
2798 | memcpy(img->line + applied_pos, | |
2799 | postimage->line, | |
2800 | postimage->nr * sizeof(*img->line)); | |
2801 | if (!state->allow_overlap) | |
2802 | for (i = 0; i < postimage->nr; i++) | |
2803 | img->line[applied_pos + i].flag |= LINE_PATCHED; | |
2804 | img->nr = nr; | |
2805 | } | |
2806 | ||
2807 | /* | |
2808 | * Use the patch-hunk text in "frag" to prepare two images (preimage and | |
2809 | * postimage) for the hunk. Find lines that match "preimage" in "img" and | |
2810 | * replace the part of "img" with "postimage" text. | |
2811 | */ | |
2812 | static int apply_one_fragment(struct apply_state *state, | |
2813 | struct image *img, struct fragment *frag, | |
2814 | int inaccurate_eof, unsigned ws_rule, | |
2815 | int nth_fragment) | |
2816 | { | |
2817 | int match_beginning, match_end; | |
2818 | const char *patch = frag->patch; | |
2819 | int size = frag->size; | |
2820 | char *old, *oldlines; | |
2821 | struct strbuf newlines; | |
2822 | int new_blank_lines_at_end = 0; | |
2823 | int found_new_blank_lines_at_end = 0; | |
2824 | int hunk_linenr = frag->linenr; | |
2825 | unsigned long leading, trailing; | |
2826 | int pos, applied_pos; | |
2827 | struct image preimage; | |
2828 | struct image postimage; | |
2829 | ||
2830 | memset(&preimage, 0, sizeof(preimage)); | |
2831 | memset(&postimage, 0, sizeof(postimage)); | |
2832 | oldlines = xmalloc(size); | |
2833 | strbuf_init(&newlines, size); | |
2834 | ||
2835 | old = oldlines; | |
2836 | while (size > 0) { | |
2837 | char first; | |
2838 | int len = linelen(patch, size); | |
2839 | int plen; | |
2840 | int added_blank_line = 0; | |
2841 | int is_blank_context = 0; | |
2842 | size_t start; | |
2843 | ||
2844 | if (!len) | |
2845 | break; | |
2846 | ||
2847 | /* | |
2848 | * "plen" is how much of the line we should use for | |
2849 | * the actual patch data. Normally we just remove the | |
2850 | * first character on the line, but if the line is | |
2851 | * followed by "\ No newline", then we also remove the | |
2852 | * last one (which is the newline, of course). | |
2853 | */ | |
2854 | plen = len - 1; | |
2855 | if (len < size && patch[len] == '\\') | |
2856 | plen--; | |
2857 | first = *patch; | |
2858 | if (state->apply_in_reverse) { | |
2859 | if (first == '-') | |
2860 | first = '+'; | |
2861 | else if (first == '+') | |
2862 | first = '-'; | |
2863 | } | |
2864 | ||
2865 | switch (first) { | |
2866 | case '\n': | |
2867 | /* Newer GNU diff, empty context line */ | |
2868 | if (plen < 0) | |
2869 | /* ... followed by '\No newline'; nothing */ | |
2870 | break; | |
2871 | *old++ = '\n'; | |
2872 | strbuf_addch(&newlines, '\n'); | |
2873 | add_line_info(&preimage, "\n", 1, LINE_COMMON); | |
2874 | add_line_info(&postimage, "\n", 1, LINE_COMMON); | |
2875 | is_blank_context = 1; | |
2876 | break; | |
2877 | case ' ': | |
2878 | if (plen && (ws_rule & WS_BLANK_AT_EOF) && | |
2879 | ws_blank_line(patch + 1, plen, ws_rule)) | |
2880 | is_blank_context = 1; | |
2881 | case '-': | |
2882 | memcpy(old, patch + 1, plen); | |
2883 | add_line_info(&preimage, old, plen, | |
2884 | (first == ' ' ? LINE_COMMON : 0)); | |
2885 | old += plen; | |
2886 | if (first == '-') | |
2887 | break; | |
2888 | /* Fall-through for ' ' */ | |
2889 | case '+': | |
2890 | /* --no-add does not add new lines */ | |
2891 | if (first == '+' && state->no_add) | |
2892 | break; | |
2893 | ||
2894 | start = newlines.len; | |
2895 | if (first != '+' || | |
2896 | !state->whitespace_error || | |
2897 | state->ws_error_action != correct_ws_error) { | |
2898 | strbuf_add(&newlines, patch + 1, plen); | |
2899 | } | |
2900 | else { | |
2901 | ws_fix_copy(&newlines, patch + 1, plen, ws_rule, &state->applied_after_fixing_ws); | |
2902 | } | |
2903 | add_line_info(&postimage, newlines.buf + start, newlines.len - start, | |
2904 | (first == '+' ? 0 : LINE_COMMON)); | |
2905 | if (first == '+' && | |
2906 | (ws_rule & WS_BLANK_AT_EOF) && | |
2907 | ws_blank_line(patch + 1, plen, ws_rule)) | |
2908 | added_blank_line = 1; | |
2909 | break; | |
2910 | case '@': case '\\': | |
2911 | /* Ignore it, we already handled it */ | |
2912 | break; | |
2913 | default: | |
2914 | if (state->apply_verbosely) | |
2915 | error(_("invalid start of line: '%c'"), first); | |
2916 | applied_pos = -1; | |
2917 | goto out; | |
2918 | } | |
2919 | if (added_blank_line) { | |
2920 | if (!new_blank_lines_at_end) | |
2921 | found_new_blank_lines_at_end = hunk_linenr; | |
2922 | new_blank_lines_at_end++; | |
2923 | } | |
2924 | else if (is_blank_context) | |
2925 | ; | |
2926 | else | |
2927 | new_blank_lines_at_end = 0; | |
2928 | patch += len; | |
2929 | size -= len; | |
2930 | hunk_linenr++; | |
2931 | } | |
2932 | if (inaccurate_eof && | |
2933 | old > oldlines && old[-1] == '\n' && | |
2934 | newlines.len > 0 && newlines.buf[newlines.len - 1] == '\n') { | |
2935 | old--; | |
2936 | strbuf_setlen(&newlines, newlines.len - 1); | |
2937 | } | |
2938 | ||
2939 | leading = frag->leading; | |
2940 | trailing = frag->trailing; | |
2941 | ||
2942 | /* | |
2943 | * A hunk to change lines at the beginning would begin with | |
2944 | * @@ -1,L +N,M @@ | |
2945 | * but we need to be careful. -U0 that inserts before the second | |
2946 | * line also has this pattern. | |
2947 | * | |
2948 | * And a hunk to add to an empty file would begin with | |
2949 | * @@ -0,0 +N,M @@ | |
2950 | * | |
2951 | * In other words, a hunk that is (frag->oldpos <= 1) with or | |
2952 | * without leading context must match at the beginning. | |
2953 | */ | |
2954 | match_beginning = (!frag->oldpos || | |
2955 | (frag->oldpos == 1 && !state->unidiff_zero)); | |
2956 | ||
2957 | /* | |
2958 | * A hunk without trailing lines must match at the end. | |
2959 | * However, we simply cannot tell if a hunk must match end | |
2960 | * from the lack of trailing lines if the patch was generated | |
2961 | * with unidiff without any context. | |
2962 | */ | |
2963 | match_end = !state->unidiff_zero && !trailing; | |
2964 | ||
2965 | pos = frag->newpos ? (frag->newpos - 1) : 0; | |
2966 | preimage.buf = oldlines; | |
2967 | preimage.len = old - oldlines; | |
2968 | postimage.buf = newlines.buf; | |
2969 | postimage.len = newlines.len; | |
2970 | preimage.line = preimage.line_allocated; | |
2971 | postimage.line = postimage.line_allocated; | |
2972 | ||
2973 | for (;;) { | |
2974 | ||
2975 | applied_pos = find_pos(state, img, &preimage, &postimage, pos, | |
2976 | ws_rule, match_beginning, match_end); | |
2977 | ||
2978 | if (applied_pos >= 0) | |
2979 | break; | |
2980 | ||
2981 | /* Am I at my context limits? */ | |
2982 | if ((leading <= state->p_context) && (trailing <= state->p_context)) | |
2983 | break; | |
2984 | if (match_beginning || match_end) { | |
2985 | match_beginning = match_end = 0; | |
2986 | continue; | |
2987 | } | |
2988 | ||
2989 | /* | |
2990 | * Reduce the number of context lines; reduce both | |
2991 | * leading and trailing if they are equal otherwise | |
2992 | * just reduce the larger context. | |
2993 | */ | |
2994 | if (leading >= trailing) { | |
2995 | remove_first_line(&preimage); | |
2996 | remove_first_line(&postimage); | |
2997 | pos--; | |
2998 | leading--; | |
2999 | } | |
3000 | if (trailing > leading) { | |
3001 | remove_last_line(&preimage); | |
3002 | remove_last_line(&postimage); | |
3003 | trailing--; | |
3004 | } | |
3005 | } | |
3006 | ||
3007 | if (applied_pos >= 0) { | |
3008 | if (new_blank_lines_at_end && | |
3009 | preimage.nr + applied_pos >= img->nr && | |
3010 | (ws_rule & WS_BLANK_AT_EOF) && | |
3011 | state->ws_error_action != nowarn_ws_error) { | |
3012 | record_ws_error(state, WS_BLANK_AT_EOF, "+", 1, | |
3013 | found_new_blank_lines_at_end); | |
3014 | if (state->ws_error_action == correct_ws_error) { | |
3015 | while (new_blank_lines_at_end--) | |
3016 | remove_last_line(&postimage); | |
3017 | } | |
3018 | /* | |
3019 | * We would want to prevent write_out_results() | |
3020 | * from taking place in apply_patch() that follows | |
3021 | * the callchain led us here, which is: | |
3022 | * apply_patch->check_patch_list->check_patch-> | |
3023 | * apply_data->apply_fragments->apply_one_fragment | |
3024 | */ | |
3025 | if (state->ws_error_action == die_on_ws_error) | |
3026 | state->apply = 0; | |
3027 | } | |
3028 | ||
3029 | if (state->apply_verbosely && applied_pos != pos) { | |
3030 | int offset = applied_pos - pos; | |
3031 | if (state->apply_in_reverse) | |
3032 | offset = 0 - offset; | |
3033 | fprintf_ln(stderr, | |
3034 | Q_("Hunk #%d succeeded at %d (offset %d line).", | |
3035 | "Hunk #%d succeeded at %d (offset %d lines).", | |
3036 | offset), | |
3037 | nth_fragment, applied_pos + 1, offset); | |
3038 | } | |
3039 | ||
3040 | /* | |
3041 | * Warn if it was necessary to reduce the number | |
3042 | * of context lines. | |
3043 | */ | |
3044 | if ((leading != frag->leading) || | |
3045 | (trailing != frag->trailing)) | |
3046 | fprintf_ln(stderr, _("Context reduced to (%ld/%ld)" | |
3047 | " to apply fragment at %d"), | |
3048 | leading, trailing, applied_pos+1); | |
3049 | update_image(state, img, applied_pos, &preimage, &postimage); | |
3050 | } else { | |
3051 | if (state->apply_verbosely) | |
3052 | error(_("while searching for:\n%.*s"), | |
3053 | (int)(old - oldlines), oldlines); | |
3054 | } | |
3055 | ||
3056 | out: | |
3057 | free(oldlines); | |
3058 | strbuf_release(&newlines); | |
3059 | free(preimage.line_allocated); | |
3060 | free(postimage.line_allocated); | |
3061 | ||
3062 | return (applied_pos < 0); | |
3063 | } | |
3064 | ||
3065 | static int apply_binary_fragment(struct apply_state *state, | |
3066 | struct image *img, | |
3067 | struct patch *patch) | |
3068 | { | |
3069 | struct fragment *fragment = patch->fragments; | |
3070 | unsigned long len; | |
3071 | void *dst; | |
3072 | ||
3073 | if (!fragment) | |
3074 | return error(_("missing binary patch data for '%s'"), | |
3075 | patch->new_name ? | |
3076 | patch->new_name : | |
3077 | patch->old_name); | |
3078 | ||
3079 | /* Binary patch is irreversible without the optional second hunk */ | |
3080 | if (state->apply_in_reverse) { | |
3081 | if (!fragment->next) | |
3082 | return error("cannot reverse-apply a binary patch " | |
3083 | "without the reverse hunk to '%s'", | |
3084 | patch->new_name | |
3085 | ? patch->new_name : patch->old_name); | |
3086 | fragment = fragment->next; | |
3087 | } | |
3088 | switch (fragment->binary_patch_method) { | |
3089 | case BINARY_DELTA_DEFLATED: | |
3090 | dst = patch_delta(img->buf, img->len, fragment->patch, | |
3091 | fragment->size, &len); | |
3092 | if (!dst) | |
3093 | return -1; | |
3094 | clear_image(img); | |
3095 | img->buf = dst; | |
3096 | img->len = len; | |
3097 | return 0; | |
3098 | case BINARY_LITERAL_DEFLATED: | |
3099 | clear_image(img); | |
3100 | img->len = fragment->size; | |
3101 | img->buf = xmemdupz(fragment->patch, img->len); | |
3102 | return 0; | |
3103 | } | |
3104 | return -1; | |
3105 | } | |
3106 | ||
3107 | /* | |
3108 | * Replace "img" with the result of applying the binary patch. | |
3109 | * The binary patch data itself in patch->fragment is still kept | |
3110 | * but the preimage prepared by the caller in "img" is freed here | |
3111 | * or in the helper function apply_binary_fragment() this calls. | |
3112 | */ | |
3113 | static int apply_binary(struct apply_state *state, | |
3114 | struct image *img, | |
3115 | struct patch *patch) | |
3116 | { | |
3117 | const char *name = patch->old_name ? patch->old_name : patch->new_name; | |
3118 | unsigned char sha1[20]; | |
3119 | ||
3120 | /* | |
3121 | * For safety, we require patch index line to contain | |
3122 | * full 40-byte textual SHA1 for old and new, at least for now. | |
3123 | */ | |
3124 | if (strlen(patch->old_sha1_prefix) != 40 || | |
3125 | strlen(patch->new_sha1_prefix) != 40 || | |
3126 | get_sha1_hex(patch->old_sha1_prefix, sha1) || | |
3127 | get_sha1_hex(patch->new_sha1_prefix, sha1)) | |
3128 | return error("cannot apply binary patch to '%s' " | |
3129 | "without full index line", name); | |
3130 | ||
3131 | if (patch->old_name) { | |
3132 | /* | |
3133 | * See if the old one matches what the patch | |
3134 | * applies to. | |
3135 | */ | |
3136 | hash_sha1_file(img->buf, img->len, blob_type, sha1); | |
3137 | if (strcmp(sha1_to_hex(sha1), patch->old_sha1_prefix)) | |
3138 | return error("the patch applies to '%s' (%s), " | |
3139 | "which does not match the " | |
3140 | "current contents.", | |
3141 | name, sha1_to_hex(sha1)); | |
3142 | } | |
3143 | else { | |
3144 | /* Otherwise, the old one must be empty. */ | |
3145 | if (img->len) | |
3146 | return error("the patch applies to an empty " | |
3147 | "'%s' but it is not empty", name); | |
3148 | } | |
3149 | ||
3150 | get_sha1_hex(patch->new_sha1_prefix, sha1); | |
3151 | if (is_null_sha1(sha1)) { | |
3152 | clear_image(img); | |
3153 | return 0; /* deletion patch */ | |
3154 | } | |
3155 | ||
3156 | if (has_sha1_file(sha1)) { | |
3157 | /* We already have the postimage */ | |
3158 | enum object_type type; | |
3159 | unsigned long size; | |
3160 | char *result; | |
3161 | ||
3162 | result = read_sha1_file(sha1, &type, &size); | |
3163 | if (!result) | |
3164 | return error("the necessary postimage %s for " | |
3165 | "'%s' cannot be read", | |
3166 | patch->new_sha1_prefix, name); | |
3167 | clear_image(img); | |
3168 | img->buf = result; | |
3169 | img->len = size; | |
3170 | } else { | |
3171 | /* | |
3172 | * We have verified buf matches the preimage; | |
3173 | * apply the patch data to it, which is stored | |
3174 | * in the patch->fragments->{patch,size}. | |
3175 | */ | |
3176 | if (apply_binary_fragment(state, img, patch)) | |
3177 | return error(_("binary patch does not apply to '%s'"), | |
3178 | name); | |
3179 | ||
3180 | /* verify that the result matches */ | |
3181 | hash_sha1_file(img->buf, img->len, blob_type, sha1); | |
3182 | if (strcmp(sha1_to_hex(sha1), patch->new_sha1_prefix)) | |
3183 | return error(_("binary patch to '%s' creates incorrect result (expecting %s, got %s)"), | |
3184 | name, patch->new_sha1_prefix, sha1_to_hex(sha1)); | |
3185 | } | |
3186 | ||
3187 | return 0; | |
3188 | } | |
3189 | ||
3190 | static int apply_fragments(struct apply_state *state, struct image *img, struct patch *patch) | |
3191 | { | |
3192 | struct fragment *frag = patch->fragments; | |
3193 | const char *name = patch->old_name ? patch->old_name : patch->new_name; | |
3194 | unsigned ws_rule = patch->ws_rule; | |
3195 | unsigned inaccurate_eof = patch->inaccurate_eof; | |
3196 | int nth = 0; | |
3197 | ||
3198 | if (patch->is_binary) | |
3199 | return apply_binary(state, img, patch); | |
3200 | ||
3201 | while (frag) { | |
3202 | nth++; | |
3203 | if (apply_one_fragment(state, img, frag, inaccurate_eof, ws_rule, nth)) { | |
3204 | error(_("patch failed: %s:%ld"), name, frag->oldpos); | |
3205 | if (!state->apply_with_reject) | |
3206 | return -1; | |
3207 | frag->rejected = 1; | |
3208 | } | |
3209 | frag = frag->next; | |
3210 | } | |
3211 | return 0; | |
3212 | } | |
3213 | ||
3214 | static int read_blob_object(struct strbuf *buf, const unsigned char *sha1, unsigned mode) | |
3215 | { | |
3216 | if (S_ISGITLINK(mode)) { | |
3217 | strbuf_grow(buf, 100); | |
3218 | strbuf_addf(buf, "Subproject commit %s\n", sha1_to_hex(sha1)); | |
3219 | } else { | |
3220 | enum object_type type; | |
3221 | unsigned long sz; | |
3222 | char *result; | |
3223 | ||
3224 | result = read_sha1_file(sha1, &type, &sz); | |
3225 | if (!result) | |
3226 | return -1; | |
3227 | /* XXX read_sha1_file NUL-terminates */ | |
3228 | strbuf_attach(buf, result, sz, sz + 1); | |
3229 | } | |
3230 | return 0; | |
3231 | } | |
3232 | ||
3233 | static int read_file_or_gitlink(const struct cache_entry *ce, struct strbuf *buf) | |
3234 | { | |
3235 | if (!ce) | |
3236 | return 0; | |
3237 | return read_blob_object(buf, ce->sha1, ce->ce_mode); | |
3238 | } | |
3239 | ||
3240 | static struct patch *in_fn_table(struct apply_state *state, const char *name) | |
3241 | { | |
3242 | struct string_list_item *item; | |
3243 | ||
3244 | if (name == NULL) | |
3245 | return NULL; | |
3246 | ||
3247 | item = string_list_lookup(&state->fn_table, name); | |
3248 | if (item != NULL) | |
3249 | return (struct patch *)item->util; | |
3250 | ||
3251 | return NULL; | |
3252 | } | |
3253 | ||
3254 | /* | |
3255 | * item->util in the filename table records the status of the path. | |
3256 | * Usually it points at a patch (whose result records the contents | |
3257 | * of it after applying it), but it could be PATH_WAS_DELETED for a | |
3258 | * path that a previously applied patch has already removed, or | |
3259 | * PATH_TO_BE_DELETED for a path that a later patch would remove. | |
3260 | * | |
3261 | * The latter is needed to deal with a case where two paths A and B | |
3262 | * are swapped by first renaming A to B and then renaming B to A; | |
3263 | * moving A to B should not be prevented due to presence of B as we | |
3264 | * will remove it in a later patch. | |
3265 | */ | |
3266 | #define PATH_TO_BE_DELETED ((struct patch *) -2) | |
3267 | #define PATH_WAS_DELETED ((struct patch *) -1) | |
3268 | ||
3269 | static int to_be_deleted(struct patch *patch) | |
3270 | { | |
3271 | return patch == PATH_TO_BE_DELETED; | |
3272 | } | |
3273 | ||
3274 | static int was_deleted(struct patch *patch) | |
3275 | { | |
3276 | return patch == PATH_WAS_DELETED; | |
3277 | } | |
3278 | ||
3279 | static void add_to_fn_table(struct apply_state *state, struct patch *patch) | |
3280 | { | |
3281 | struct string_list_item *item; | |
3282 | ||
3283 | /* | |
3284 | * Always add new_name unless patch is a deletion | |
3285 | * This should cover the cases for normal diffs, | |
3286 | * file creations and copies | |
3287 | */ | |
3288 | if (patch->new_name != NULL) { | |
3289 | item = string_list_insert(&state->fn_table, patch->new_name); | |
3290 | item->util = patch; | |
3291 | } | |
3292 | ||
3293 | /* | |
3294 | * store a failure on rename/deletion cases because | |
3295 | * later chunks shouldn't patch old names | |
3296 | */ | |
3297 | if ((patch->new_name == NULL) || (patch->is_rename)) { | |
3298 | item = string_list_insert(&state->fn_table, patch->old_name); | |
3299 | item->util = PATH_WAS_DELETED; | |
3300 | } | |
3301 | } | |
3302 | ||
3303 | static void prepare_fn_table(struct apply_state *state, struct patch *patch) | |
3304 | { | |
3305 | /* | |
3306 | * store information about incoming file deletion | |
3307 | */ | |
3308 | while (patch) { | |
3309 | if ((patch->new_name == NULL) || (patch->is_rename)) { | |
3310 | struct string_list_item *item; | |
3311 | item = string_list_insert(&state->fn_table, patch->old_name); | |
3312 | item->util = PATH_TO_BE_DELETED; | |
3313 | } | |
3314 | patch = patch->next; | |
3315 | } | |
3316 | } | |
3317 | ||
3318 | static int checkout_target(struct index_state *istate, | |
3319 | struct cache_entry *ce, struct stat *st) | |
3320 | { | |
3321 | struct checkout costate; | |
3322 | ||
3323 | memset(&costate, 0, sizeof(costate)); | |
3324 | costate.base_dir = ""; | |
3325 | costate.refresh_cache = 1; | |
3326 | costate.istate = istate; | |
3327 | if (checkout_entry(ce, &costate, NULL) || lstat(ce->name, st)) | |
3328 | return error(_("cannot checkout %s"), ce->name); | |
3329 | return 0; | |
3330 | } | |
3331 | ||
3332 | static struct patch *previous_patch(struct apply_state *state, | |
3333 | struct patch *patch, | |
3334 | int *gone) | |
3335 | { | |
3336 | struct patch *previous; | |
3337 | ||
3338 | *gone = 0; | |
3339 | if (patch->is_copy || patch->is_rename) | |
3340 | return NULL; /* "git" patches do not depend on the order */ | |
3341 | ||
3342 | previous = in_fn_table(state, patch->old_name); | |
3343 | if (!previous) | |
3344 | return NULL; | |
3345 | ||
3346 | if (to_be_deleted(previous)) | |
3347 | return NULL; /* the deletion hasn't happened yet */ | |
3348 | ||
3349 | if (was_deleted(previous)) | |
3350 | *gone = 1; | |
3351 | ||
3352 | return previous; | |
3353 | } | |
3354 | ||
3355 | static int verify_index_match(const struct cache_entry *ce, struct stat *st) | |
3356 | { | |
3357 | if (S_ISGITLINK(ce->ce_mode)) { | |
3358 | if (!S_ISDIR(st->st_mode)) | |
3359 | return -1; | |
3360 | return 0; | |
3361 | } | |
3362 | return ce_match_stat(ce, st, CE_MATCH_IGNORE_VALID|CE_MATCH_IGNORE_SKIP_WORKTREE); | |
3363 | } | |
3364 | ||
3365 | #define SUBMODULE_PATCH_WITHOUT_INDEX 1 | |
3366 | ||
3367 | static int load_patch_target(struct apply_state *state, | |
3368 | struct strbuf *buf, | |
3369 | const struct cache_entry *ce, | |
3370 | struct stat *st, | |
3371 | const char *name, | |
3372 | unsigned expected_mode) | |
3373 | { | |
3374 | if (state->cached || state->check_index) { | |
3375 | if (read_file_or_gitlink(ce, buf)) | |
3376 | return error(_("failed to read %s"), name); | |
3377 | } else if (name) { | |
3378 | if (S_ISGITLINK(expected_mode)) { | |
3379 | if (ce) | |
3380 | return read_file_or_gitlink(ce, buf); | |
3381 | else | |
3382 | return SUBMODULE_PATCH_WITHOUT_INDEX; | |
3383 | } else if (has_symlink_leading_path(name, strlen(name))) { | |
3384 | return error(_("reading from '%s' beyond a symbolic link"), name); | |
3385 | } else { | |
3386 | if (read_old_data(st, name, buf)) | |
3387 | return error(_("failed to read %s"), name); | |
3388 | } | |
3389 | } | |
3390 | return 0; | |
3391 | } | |
3392 | ||
3393 | /* | |
3394 | * We are about to apply "patch"; populate the "image" with the | |
3395 | * current version we have, from the working tree or from the index, | |
3396 | * depending on the situation e.g. --cached/--index. If we are | |
3397 | * applying a non-git patch that incrementally updates the tree, | |
3398 | * we read from the result of a previous diff. | |
3399 | */ | |
3400 | static int load_preimage(struct apply_state *state, | |
3401 | struct image *image, | |
3402 | struct patch *patch, struct stat *st, | |
3403 | const struct cache_entry *ce) | |
3404 | { | |
3405 | struct strbuf buf = STRBUF_INIT; | |
3406 | size_t len; | |
3407 | char *img; | |
3408 | struct patch *previous; | |
3409 | int status; | |
3410 | ||
3411 | previous = previous_patch(state, patch, &status); | |
3412 | if (status) | |
3413 | return error(_("path %s has been renamed/deleted"), | |
3414 | patch->old_name); | |
3415 | if (previous) { | |
3416 | /* We have a patched copy in memory; use that. */ | |
3417 | strbuf_add(&buf, previous->result, previous->resultsize); | |
3418 | } else { | |
3419 | status = load_patch_target(state, &buf, ce, st, | |
3420 | patch->old_name, patch->old_mode); | |
3421 | if (status < 0) | |
3422 | return status; | |
3423 | else if (status == SUBMODULE_PATCH_WITHOUT_INDEX) { | |
3424 | /* | |
3425 | * There is no way to apply subproject | |
3426 | * patch without looking at the index. | |
3427 | * NEEDSWORK: shouldn't this be flagged | |
3428 | * as an error??? | |
3429 | */ | |
3430 | free_fragment_list(patch->fragments); | |
3431 | patch->fragments = NULL; | |
3432 | } else if (status) { | |
3433 | return error(_("failed to read %s"), patch->old_name); | |
3434 | } | |
3435 | } | |
3436 | ||
3437 | img = strbuf_detach(&buf, &len); | |
3438 | prepare_image(image, img, len, !patch->is_binary); | |
3439 | return 0; | |
3440 | } | |
3441 | ||
3442 | static int three_way_merge(struct image *image, | |
3443 | char *path, | |
3444 | const unsigned char *base, | |
3445 | const unsigned char *ours, | |
3446 | const unsigned char *theirs) | |
3447 | { | |
3448 | mmfile_t base_file, our_file, their_file; | |
3449 | mmbuffer_t result = { NULL }; | |
3450 | int status; | |
3451 | ||
3452 | read_mmblob(&base_file, base); | |
3453 | read_mmblob(&our_file, ours); | |
3454 | read_mmblob(&their_file, theirs); | |
3455 | status = ll_merge(&result, path, | |
3456 | &base_file, "base", | |
3457 | &our_file, "ours", | |
3458 | &their_file, "theirs", NULL); | |
3459 | free(base_file.ptr); | |
3460 | free(our_file.ptr); | |
3461 | free(their_file.ptr); | |
3462 | if (status < 0 || !result.ptr) { | |
3463 | free(result.ptr); | |
3464 | return -1; | |
3465 | } | |
3466 | clear_image(image); | |
3467 | image->buf = result.ptr; | |
3468 | image->len = result.size; | |
3469 | ||
3470 | return status; | |
3471 | } | |
3472 | ||
3473 | /* | |
3474 | * When directly falling back to add/add three-way merge, we read from | |
3475 | * the current contents of the new_name. In no cases other than that | |
3476 | * this function will be called. | |
3477 | */ | |
3478 | static int load_current(struct apply_state *state, | |
3479 | struct image *image, | |
3480 | struct patch *patch) | |
3481 | { | |
3482 | struct strbuf buf = STRBUF_INIT; | |
3483 | int status, pos; | |
3484 | size_t len; | |
3485 | char *img; | |
3486 | struct stat st; | |
3487 | struct cache_entry *ce; | |
3488 | char *name = patch->new_name; | |
3489 | unsigned mode = patch->new_mode; | |
3490 | ||
3491 | if (!patch->is_new) | |
3492 | die("BUG: patch to %s is not a creation", patch->old_name); | |
3493 | ||
3494 | pos = cache_name_pos(name, strlen(name)); | |
3495 | if (pos < 0) | |
3496 | return error(_("%s: does not exist in index"), name); | |
3497 | ce = active_cache[pos]; | |
3498 | if (lstat(name, &st)) { | |
3499 | if (errno != ENOENT) | |
90875eca | 3500 | return error_errno("%s", name); |
13b5af22 CC |
3501 | if (checkout_target(&the_index, ce, &st)) |
3502 | return -1; | |
3503 | } | |
3504 | if (verify_index_match(ce, &st)) | |
3505 | return error(_("%s: does not match index"), name); | |
3506 | ||
3507 | status = load_patch_target(state, &buf, ce, &st, name, mode); | |
3508 | if (status < 0) | |
3509 | return status; | |
3510 | else if (status) | |
3511 | return -1; | |
3512 | img = strbuf_detach(&buf, &len); | |
3513 | prepare_image(image, img, len, !patch->is_binary); | |
3514 | return 0; | |
3515 | } | |
3516 | ||
3517 | static int try_threeway(struct apply_state *state, | |
3518 | struct image *image, | |
3519 | struct patch *patch, | |
3520 | struct stat *st, | |
3521 | const struct cache_entry *ce) | |
3522 | { | |
3523 | unsigned char pre_sha1[20], post_sha1[20], our_sha1[20]; | |
3524 | struct strbuf buf = STRBUF_INIT; | |
3525 | size_t len; | |
3526 | int status; | |
3527 | char *img; | |
3528 | struct image tmp_image; | |
3529 | ||
3530 | /* No point falling back to 3-way merge in these cases */ | |
3531 | if (patch->is_delete || | |
3532 | S_ISGITLINK(patch->old_mode) || S_ISGITLINK(patch->new_mode)) | |
3533 | return -1; | |
3534 | ||
3535 | /* Preimage the patch was prepared for */ | |
3536 | if (patch->is_new) | |
3537 | write_sha1_file("", 0, blob_type, pre_sha1); | |
3538 | else if (get_sha1(patch->old_sha1_prefix, pre_sha1) || | |
3539 | read_blob_object(&buf, pre_sha1, patch->old_mode)) | |
3540 | return error("repository lacks the necessary blob to fall back on 3-way merge."); | |
3541 | ||
3542 | fprintf(stderr, "Falling back to three-way merge...\n"); | |
3543 | ||
3544 | img = strbuf_detach(&buf, &len); | |
3545 | prepare_image(&tmp_image, img, len, 1); | |
3546 | /* Apply the patch to get the post image */ | |
3547 | if (apply_fragments(state, &tmp_image, patch) < 0) { | |
3548 | clear_image(&tmp_image); | |
3549 | return -1; | |
3550 | } | |
3551 | /* post_sha1[] is theirs */ | |
3552 | write_sha1_file(tmp_image.buf, tmp_image.len, blob_type, post_sha1); | |
3553 | clear_image(&tmp_image); | |
3554 | ||
3555 | /* our_sha1[] is ours */ | |
3556 | if (patch->is_new) { | |
3557 | if (load_current(state, &tmp_image, patch)) | |
3558 | return error("cannot read the current contents of '%s'", | |
3559 | patch->new_name); | |
3560 | } else { | |
3561 | if (load_preimage(state, &tmp_image, patch, st, ce)) | |
3562 | return error("cannot read the current contents of '%s'", | |
3563 | patch->old_name); | |
3564 | } | |
3565 | write_sha1_file(tmp_image.buf, tmp_image.len, blob_type, our_sha1); | |
3566 | clear_image(&tmp_image); | |
3567 | ||
3568 | /* in-core three-way merge between post and our using pre as base */ | |
3569 | status = three_way_merge(image, patch->new_name, | |
3570 | pre_sha1, our_sha1, post_sha1); | |
3571 | if (status < 0) { | |
3572 | fprintf(stderr, "Failed to fall back on three-way merge...\n"); | |
3573 | return status; | |
3574 | } | |
3575 | ||
3576 | if (status) { | |
3577 | patch->conflicted_threeway = 1; | |
3578 | if (patch->is_new) | |
3579 | oidclr(&patch->threeway_stage[0]); | |
3580 | else | |
3581 | hashcpy(patch->threeway_stage[0].hash, pre_sha1); | |
3582 | hashcpy(patch->threeway_stage[1].hash, our_sha1); | |
3583 | hashcpy(patch->threeway_stage[2].hash, post_sha1); | |
3584 | fprintf(stderr, "Applied patch to '%s' with conflicts.\n", patch->new_name); | |
3585 | } else { | |
3586 | fprintf(stderr, "Applied patch to '%s' cleanly.\n", patch->new_name); | |
3587 | } | |
3588 | return 0; | |
3589 | } | |
3590 | ||
3591 | static int apply_data(struct apply_state *state, struct patch *patch, | |
3592 | struct stat *st, const struct cache_entry *ce) | |
3593 | { | |
3594 | struct image image; | |
3595 | ||
3596 | if (load_preimage(state, &image, patch, st, ce) < 0) | |
3597 | return -1; | |
3598 | ||
3599 | if (patch->direct_to_threeway || | |
3600 | apply_fragments(state, &image, patch) < 0) { | |
3601 | /* Note: with --reject, apply_fragments() returns 0 */ | |
3602 | if (!state->threeway || try_threeway(state, &image, patch, st, ce) < 0) | |
3603 | return -1; | |
3604 | } | |
3605 | patch->result = image.buf; | |
3606 | patch->resultsize = image.len; | |
3607 | add_to_fn_table(state, patch); | |
3608 | free(image.line_allocated); | |
3609 | ||
3610 | if (0 < patch->is_delete && patch->resultsize) | |
3611 | return error(_("removal patch leaves file contents")); | |
3612 | ||
3613 | return 0; | |
3614 | } | |
3615 | ||
3616 | /* | |
3617 | * If "patch" that we are looking at modifies or deletes what we have, | |
3618 | * we would want it not to lose any local modification we have, either | |
3619 | * in the working tree or in the index. | |
3620 | * | |
3621 | * This also decides if a non-git patch is a creation patch or a | |
3622 | * modification to an existing empty file. We do not check the state | |
3623 | * of the current tree for a creation patch in this function; the caller | |
3624 | * check_patch() separately makes sure (and errors out otherwise) that | |
3625 | * the path the patch creates does not exist in the current tree. | |
3626 | */ | |
3627 | static int check_preimage(struct apply_state *state, | |
3628 | struct patch *patch, | |
3629 | struct cache_entry **ce, | |
3630 | struct stat *st) | |
3631 | { | |
3632 | const char *old_name = patch->old_name; | |
3633 | struct patch *previous = NULL; | |
3634 | int stat_ret = 0, status; | |
3635 | unsigned st_mode = 0; | |
3636 | ||
3637 | if (!old_name) | |
3638 | return 0; | |
3639 | ||
3640 | assert(patch->is_new <= 0); | |
3641 | previous = previous_patch(state, patch, &status); | |
3642 | ||
3643 | if (status) | |
3644 | return error(_("path %s has been renamed/deleted"), old_name); | |
3645 | if (previous) { | |
3646 | st_mode = previous->new_mode; | |
3647 | } else if (!state->cached) { | |
3648 | stat_ret = lstat(old_name, st); | |
3649 | if (stat_ret && errno != ENOENT) | |
90875eca | 3650 | return error_errno("%s", old_name); |
13b5af22 CC |
3651 | } |
3652 | ||
3653 | if (state->check_index && !previous) { | |
3654 | int pos = cache_name_pos(old_name, strlen(old_name)); | |
3655 | if (pos < 0) { | |
3656 | if (patch->is_new < 0) | |
3657 | goto is_new; | |
3658 | return error(_("%s: does not exist in index"), old_name); | |
3659 | } | |
3660 | *ce = active_cache[pos]; | |
3661 | if (stat_ret < 0) { | |
3662 | if (checkout_target(&the_index, *ce, st)) | |
3663 | return -1; | |
3664 | } | |
3665 | if (!state->cached && verify_index_match(*ce, st)) | |
3666 | return error(_("%s: does not match index"), old_name); | |
3667 | if (state->cached) | |
3668 | st_mode = (*ce)->ce_mode; | |
3669 | } else if (stat_ret < 0) { | |
3670 | if (patch->is_new < 0) | |
3671 | goto is_new; | |
90875eca | 3672 | return error_errno("%s", old_name); |
13b5af22 CC |
3673 | } |
3674 | ||
3675 | if (!state->cached && !previous) | |
3676 | st_mode = ce_mode_from_stat(*ce, st->st_mode); | |
3677 | ||
3678 | if (patch->is_new < 0) | |
3679 | patch->is_new = 0; | |
3680 | if (!patch->old_mode) | |
3681 | patch->old_mode = st_mode; | |
3682 | if ((st_mode ^ patch->old_mode) & S_IFMT) | |
3683 | return error(_("%s: wrong type"), old_name); | |
3684 | if (st_mode != patch->old_mode) | |
3685 | warning(_("%s has type %o, expected %o"), | |
3686 | old_name, st_mode, patch->old_mode); | |
3687 | if (!patch->new_mode && !patch->is_delete) | |
3688 | patch->new_mode = st_mode; | |
3689 | return 0; | |
3690 | ||
3691 | is_new: | |
3692 | patch->is_new = 1; | |
3693 | patch->is_delete = 0; | |
3694 | free(patch->old_name); | |
3695 | patch->old_name = NULL; | |
3696 | return 0; | |
3697 | } | |
3698 | ||
3699 | ||
3700 | #define EXISTS_IN_INDEX 1 | |
3701 | #define EXISTS_IN_WORKTREE 2 | |
3702 | ||
3703 | static int check_to_create(struct apply_state *state, | |
3704 | const char *new_name, | |
3705 | int ok_if_exists) | |
3706 | { | |
3707 | struct stat nst; | |
3708 | ||
3709 | if (state->check_index && | |
3710 | cache_name_pos(new_name, strlen(new_name)) >= 0 && | |
3711 | !ok_if_exists) | |
3712 | return EXISTS_IN_INDEX; | |
3713 | if (state->cached) | |
3714 | return 0; | |
3715 | ||
3716 | if (!lstat(new_name, &nst)) { | |
3717 | if (S_ISDIR(nst.st_mode) || ok_if_exists) | |
3718 | return 0; | |
3719 | /* | |
3720 | * A leading component of new_name might be a symlink | |
3721 | * that is going to be removed with this patch, but | |
3722 | * still pointing at somewhere that has the path. | |
3723 | * In such a case, path "new_name" does not exist as | |
3724 | * far as git is concerned. | |
3725 | */ | |
3726 | if (has_symlink_leading_path(new_name, strlen(new_name))) | |
3727 | return 0; | |
3728 | ||
3729 | return EXISTS_IN_WORKTREE; | |
3730 | } else if ((errno != ENOENT) && (errno != ENOTDIR)) { | |
90875eca | 3731 | return error_errno("%s", new_name); |
13b5af22 CC |
3732 | } |
3733 | return 0; | |
3734 | } | |
3735 | ||
3736 | static uintptr_t register_symlink_changes(struct apply_state *state, | |
3737 | const char *path, | |
3738 | uintptr_t what) | |
3739 | { | |
3740 | struct string_list_item *ent; | |
3741 | ||
3742 | ent = string_list_lookup(&state->symlink_changes, path); | |
3743 | if (!ent) { | |
3744 | ent = string_list_insert(&state->symlink_changes, path); | |
3745 | ent->util = (void *)0; | |
3746 | } | |
3747 | ent->util = (void *)(what | ((uintptr_t)ent->util)); | |
3748 | return (uintptr_t)ent->util; | |
3749 | } | |
3750 | ||
3751 | static uintptr_t check_symlink_changes(struct apply_state *state, const char *path) | |
3752 | { | |
3753 | struct string_list_item *ent; | |
3754 | ||
3755 | ent = string_list_lookup(&state->symlink_changes, path); | |
3756 | if (!ent) | |
3757 | return 0; | |
3758 | return (uintptr_t)ent->util; | |
3759 | } | |
3760 | ||
3761 | static void prepare_symlink_changes(struct apply_state *state, struct patch *patch) | |
3762 | { | |
3763 | for ( ; patch; patch = patch->next) { | |
3764 | if ((patch->old_name && S_ISLNK(patch->old_mode)) && | |
3765 | (patch->is_rename || patch->is_delete)) | |
3766 | /* the symlink at patch->old_name is removed */ | |
3767 | register_symlink_changes(state, patch->old_name, APPLY_SYMLINK_GOES_AWAY); | |
3768 | ||
3769 | if (patch->new_name && S_ISLNK(patch->new_mode)) | |
3770 | /* the symlink at patch->new_name is created or remains */ | |
3771 | register_symlink_changes(state, patch->new_name, APPLY_SYMLINK_IN_RESULT); | |
3772 | } | |
3773 | } | |
3774 | ||
3775 | static int path_is_beyond_symlink_1(struct apply_state *state, struct strbuf *name) | |
3776 | { | |
3777 | do { | |
3778 | unsigned int change; | |
3779 | ||
3780 | while (--name->len && name->buf[name->len] != '/') | |
3781 | ; /* scan backwards */ | |
3782 | if (!name->len) | |
3783 | break; | |
3784 | name->buf[name->len] = '\0'; | |
3785 | change = check_symlink_changes(state, name->buf); | |
3786 | if (change & APPLY_SYMLINK_IN_RESULT) | |
3787 | return 1; | |
3788 | if (change & APPLY_SYMLINK_GOES_AWAY) | |
3789 | /* | |
3790 | * This cannot be "return 0", because we may | |
3791 | * see a new one created at a higher level. | |
3792 | */ | |
3793 | continue; | |
3794 | ||
3795 | /* otherwise, check the preimage */ | |
3796 | if (state->check_index) { | |
3797 | struct cache_entry *ce; | |
3798 | ||
3799 | ce = cache_file_exists(name->buf, name->len, ignore_case); | |
3800 | if (ce && S_ISLNK(ce->ce_mode)) | |
3801 | return 1; | |
3802 | } else { | |
3803 | struct stat st; | |
3804 | if (!lstat(name->buf, &st) && S_ISLNK(st.st_mode)) | |
3805 | return 1; | |
3806 | } | |
3807 | } while (1); | |
3808 | return 0; | |
3809 | } | |
3810 | ||
3811 | static int path_is_beyond_symlink(struct apply_state *state, const char *name_) | |
3812 | { | |
3813 | int ret; | |
3814 | struct strbuf name = STRBUF_INIT; | |
3815 | ||
3816 | assert(*name_ != '\0'); | |
3817 | strbuf_addstr(&name, name_); | |
3818 | ret = path_is_beyond_symlink_1(state, &name); | |
3819 | strbuf_release(&name); | |
3820 | ||
3821 | return ret; | |
3822 | } | |
3823 | ||
3824 | static int check_unsafe_path(struct patch *patch) | |
3825 | { | |
3826 | const char *old_name = NULL; | |
3827 | const char *new_name = NULL; | |
3828 | if (patch->is_delete) | |
3829 | old_name = patch->old_name; | |
3830 | else if (!patch->is_new && !patch->is_copy) | |
3831 | old_name = patch->old_name; | |
3832 | if (!patch->is_delete) | |
3833 | new_name = patch->new_name; | |
3834 | ||
3835 | if (old_name && !verify_path(old_name)) | |
3836 | return error(_("invalid path '%s'"), old_name); | |
3837 | if (new_name && !verify_path(new_name)) | |
3838 | return error(_("invalid path '%s'"), new_name); | |
3839 | return 0; | |
3840 | } | |
3841 | ||
3842 | /* | |
3843 | * Check and apply the patch in-core; leave the result in patch->result | |
3844 | * for the caller to write it out to the final destination. | |
3845 | */ | |
3846 | static int check_patch(struct apply_state *state, struct patch *patch) | |
3847 | { | |
3848 | struct stat st; | |
3849 | const char *old_name = patch->old_name; | |
3850 | const char *new_name = patch->new_name; | |
3851 | const char *name = old_name ? old_name : new_name; | |
3852 | struct cache_entry *ce = NULL; | |
3853 | struct patch *tpatch; | |
3854 | int ok_if_exists; | |
3855 | int status; | |
3856 | ||
3857 | patch->rejected = 1; /* we will drop this after we succeed */ | |
3858 | ||
3859 | status = check_preimage(state, patch, &ce, &st); | |
3860 | if (status) | |
3861 | return status; | |
3862 | old_name = patch->old_name; | |
3863 | ||
3864 | /* | |
3865 | * A type-change diff is always split into a patch to delete | |
3866 | * old, immediately followed by a patch to create new (see | |
3867 | * diff.c::run_diff()); in such a case it is Ok that the entry | |
3868 | * to be deleted by the previous patch is still in the working | |
3869 | * tree and in the index. | |
3870 | * | |
3871 | * A patch to swap-rename between A and B would first rename A | |
3872 | * to B and then rename B to A. While applying the first one, | |
3873 | * the presence of B should not stop A from getting renamed to | |
3874 | * B; ask to_be_deleted() about the later rename. Removal of | |
3875 | * B and rename from A to B is handled the same way by asking | |
3876 | * was_deleted(). | |
3877 | */ | |
3878 | if ((tpatch = in_fn_table(state, new_name)) && | |
3879 | (was_deleted(tpatch) || to_be_deleted(tpatch))) | |
3880 | ok_if_exists = 1; | |
3881 | else | |
3882 | ok_if_exists = 0; | |
3883 | ||
3884 | if (new_name && | |
3885 | ((0 < patch->is_new) || patch->is_rename || patch->is_copy)) { | |
3886 | int err = check_to_create(state, new_name, ok_if_exists); | |
3887 | ||
3888 | if (err && state->threeway) { | |
3889 | patch->direct_to_threeway = 1; | |
3890 | } else switch (err) { | |
3891 | case 0: | |
3892 | break; /* happy */ | |
3893 | case EXISTS_IN_INDEX: | |
3894 | return error(_("%s: already exists in index"), new_name); | |
3895 | break; | |
3896 | case EXISTS_IN_WORKTREE: | |
3897 | return error(_("%s: already exists in working directory"), | |
3898 | new_name); | |
3899 | default: | |
3900 | return err; | |
3901 | } | |
3902 | ||
3903 | if (!patch->new_mode) { | |
3904 | if (0 < patch->is_new) | |
3905 | patch->new_mode = S_IFREG | 0644; | |
3906 | else | |
3907 | patch->new_mode = patch->old_mode; | |
3908 | } | |
3909 | } | |
3910 | ||
3911 | if (new_name && old_name) { | |
3912 | int same = !strcmp(old_name, new_name); | |
3913 | if (!patch->new_mode) | |
3914 | patch->new_mode = patch->old_mode; | |
3915 | if ((patch->old_mode ^ patch->new_mode) & S_IFMT) { | |
3916 | if (same) | |
3917 | return error(_("new mode (%o) of %s does not " | |
3918 | "match old mode (%o)"), | |
3919 | patch->new_mode, new_name, | |
3920 | patch->old_mode); | |
3921 | else | |
3922 | return error(_("new mode (%o) of %s does not " | |
3923 | "match old mode (%o) of %s"), | |
3924 | patch->new_mode, new_name, | |
3925 | patch->old_mode, old_name); | |
3926 | } | |
3927 | } | |
3928 | ||
3929 | if (!state->unsafe_paths && check_unsafe_path(patch)) | |
3930 | return -128; | |
3931 | ||
3932 | /* | |
3933 | * An attempt to read from or delete a path that is beyond a | |
3934 | * symbolic link will be prevented by load_patch_target() that | |
3935 | * is called at the beginning of apply_data() so we do not | |
3936 | * have to worry about a patch marked with "is_delete" bit | |
3937 | * here. We however need to make sure that the patch result | |
3938 | * is not deposited to a path that is beyond a symbolic link | |
3939 | * here. | |
3940 | */ | |
3941 | if (!patch->is_delete && path_is_beyond_symlink(state, patch->new_name)) | |
3942 | return error(_("affected file '%s' is beyond a symbolic link"), | |
3943 | patch->new_name); | |
3944 | ||
3945 | if (apply_data(state, patch, &st, ce) < 0) | |
3946 | return error(_("%s: patch does not apply"), name); | |
3947 | patch->rejected = 0; | |
3948 | return 0; | |
3949 | } | |
3950 | ||
3951 | static int check_patch_list(struct apply_state *state, struct patch *patch) | |
3952 | { | |
3953 | int err = 0; | |
3954 | ||
3955 | prepare_symlink_changes(state, patch); | |
3956 | prepare_fn_table(state, patch); | |
3957 | while (patch) { | |
3958 | int res; | |
3959 | if (state->apply_verbosely) | |
3960 | say_patch_name(stderr, | |
3961 | _("Checking patch %s..."), patch); | |
3962 | res = check_patch(state, patch); | |
3963 | if (res == -128) | |
3964 | return -128; | |
3965 | err |= res; | |
3966 | patch = patch->next; | |
3967 | } | |
3968 | return err; | |
3969 | } | |
3970 | ||
3971 | /* This function tries to read the sha1 from the current index */ | |
3972 | static int get_current_sha1(const char *path, unsigned char *sha1) | |
3973 | { | |
3974 | int pos; | |
3975 | ||
3976 | if (read_cache() < 0) | |
3977 | return -1; | |
3978 | pos = cache_name_pos(path, strlen(path)); | |
3979 | if (pos < 0) | |
3980 | return -1; | |
3981 | hashcpy(sha1, active_cache[pos]->sha1); | |
3982 | return 0; | |
3983 | } | |
3984 | ||
3985 | static int preimage_sha1_in_gitlink_patch(struct patch *p, unsigned char sha1[20]) | |
3986 | { | |
3987 | /* | |
3988 | * A usable gitlink patch has only one fragment (hunk) that looks like: | |
3989 | * @@ -1 +1 @@ | |
3990 | * -Subproject commit <old sha1> | |
3991 | * +Subproject commit <new sha1> | |
3992 | * or | |
3993 | * @@ -1 +0,0 @@ | |
3994 | * -Subproject commit <old sha1> | |
3995 | * for a removal patch. | |
3996 | */ | |
3997 | struct fragment *hunk = p->fragments; | |
3998 | static const char heading[] = "-Subproject commit "; | |
3999 | char *preimage; | |
4000 | ||
4001 | if (/* does the patch have only one hunk? */ | |
4002 | hunk && !hunk->next && | |
4003 | /* is its preimage one line? */ | |
4004 | hunk->oldpos == 1 && hunk->oldlines == 1 && | |
4005 | /* does preimage begin with the heading? */ | |
4006 | (preimage = memchr(hunk->patch, '\n', hunk->size)) != NULL && | |
4007 | starts_with(++preimage, heading) && | |
4008 | /* does it record full SHA-1? */ | |
4009 | !get_sha1_hex(preimage + sizeof(heading) - 1, sha1) && | |
4010 | preimage[sizeof(heading) + 40 - 1] == '\n' && | |
4011 | /* does the abbreviated name on the index line agree with it? */ | |
4012 | starts_with(preimage + sizeof(heading) - 1, p->old_sha1_prefix)) | |
4013 | return 0; /* it all looks fine */ | |
4014 | ||
4015 | /* we may have full object name on the index line */ | |
4016 | return get_sha1_hex(p->old_sha1_prefix, sha1); | |
4017 | } | |
4018 | ||
4019 | /* Build an index that contains the just the files needed for a 3way merge */ | |
4020 | static int build_fake_ancestor(struct patch *list, const char *filename) | |
4021 | { | |
4022 | struct patch *patch; | |
4023 | struct index_state result = { NULL }; | |
4024 | static struct lock_file lock; | |
4025 | int res; | |
4026 | ||
4027 | /* Once we start supporting the reverse patch, it may be | |
4028 | * worth showing the new sha1 prefix, but until then... | |
4029 | */ | |
4030 | for (patch = list; patch; patch = patch->next) { | |
4031 | unsigned char sha1[20]; | |
4032 | struct cache_entry *ce; | |
4033 | const char *name; | |
4034 | ||
4035 | name = patch->old_name ? patch->old_name : patch->new_name; | |
4036 | if (0 < patch->is_new) | |
4037 | continue; | |
4038 | ||
4039 | if (S_ISGITLINK(patch->old_mode)) { | |
4040 | if (!preimage_sha1_in_gitlink_patch(patch, sha1)) | |
4041 | ; /* ok, the textual part looks sane */ | |
4042 | else | |
4043 | return error("sha1 information is lacking or " | |
4044 | "useless for submodule %s", name); | |
4045 | } else if (!get_sha1_blob(patch->old_sha1_prefix, sha1)) { | |
4046 | ; /* ok */ | |
4047 | } else if (!patch->lines_added && !patch->lines_deleted) { | |
4048 | /* mode-only change: update the current */ | |
4049 | if (get_current_sha1(patch->old_name, sha1)) | |
4050 | return error("mode change for %s, which is not " | |
4051 | "in current HEAD", name); | |
4052 | } else | |
4053 | return error("sha1 information is lacking or useless " | |
4054 | "(%s).", name); | |
4055 | ||
4056 | ce = make_cache_entry(patch->old_mode, sha1, name, 0, 0); | |
4057 | if (!ce) | |
4058 | return error(_("make_cache_entry failed for path '%s'"), | |
4059 | name); | |
4060 | if (add_index_entry(&result, ce, ADD_CACHE_OK_TO_ADD)) { | |
4061 | free(ce); | |
4062 | return error("Could not add %s to temporary index", | |
4063 | name); | |
4064 | } | |
4065 | } | |
4066 | ||
4067 | hold_lock_file_for_update(&lock, filename, LOCK_DIE_ON_ERROR); | |
4068 | res = write_locked_index(&result, &lock, COMMIT_LOCK); | |
4069 | discard_index(&result); | |
4070 | ||
4071 | if (res) | |
4072 | return error("Could not write temporary index to %s", filename); | |
4073 | ||
4074 | return 0; | |
4075 | } | |
4076 | ||
4077 | static void stat_patch_list(struct apply_state *state, struct patch *patch) | |
4078 | { | |
4079 | int files, adds, dels; | |
4080 | ||
4081 | for (files = adds = dels = 0 ; patch ; patch = patch->next) { | |
4082 | files++; | |
4083 | adds += patch->lines_added; | |
4084 | dels += patch->lines_deleted; | |
4085 | show_stats(state, patch); | |
4086 | } | |
4087 | ||
4088 | print_stat_summary(stdout, files, adds, dels); | |
4089 | } | |
4090 | ||
4091 | static void numstat_patch_list(struct apply_state *state, | |
4092 | struct patch *patch) | |
4093 | { | |
4094 | for ( ; patch; patch = patch->next) { | |
4095 | const char *name; | |
4096 | name = patch->new_name ? patch->new_name : patch->old_name; | |
4097 | if (patch->is_binary) | |
4098 | printf("-\t-\t"); | |
4099 | else | |
4100 | printf("%d\t%d\t", patch->lines_added, patch->lines_deleted); | |
4101 | write_name_quoted(name, stdout, state->line_termination); | |
4102 | } | |
4103 | } | |
4104 | ||
4105 | static void show_file_mode_name(const char *newdelete, unsigned int mode, const char *name) | |
4106 | { | |
4107 | if (mode) | |
4108 | printf(" %s mode %06o %s\n", newdelete, mode, name); | |
4109 | else | |
4110 | printf(" %s %s\n", newdelete, name); | |
4111 | } | |
4112 | ||
4113 | static void show_mode_change(struct patch *p, int show_name) | |
4114 | { | |
4115 | if (p->old_mode && p->new_mode && p->old_mode != p->new_mode) { | |
4116 | if (show_name) | |
4117 | printf(" mode change %06o => %06o %s\n", | |
4118 | p->old_mode, p->new_mode, p->new_name); | |
4119 | else | |
4120 | printf(" mode change %06o => %06o\n", | |
4121 | p->old_mode, p->new_mode); | |
4122 | } | |
4123 | } | |
4124 | ||
4125 | static void show_rename_copy(struct patch *p) | |
4126 | { | |
4127 | const char *renamecopy = p->is_rename ? "rename" : "copy"; | |
4128 | const char *old, *new; | |
4129 | ||
4130 | /* Find common prefix */ | |
4131 | old = p->old_name; | |
4132 | new = p->new_name; | |
4133 | while (1) { | |
4134 | const char *slash_old, *slash_new; | |
4135 | slash_old = strchr(old, '/'); | |
4136 | slash_new = strchr(new, '/'); | |
4137 | if (!slash_old || | |
4138 | !slash_new || | |
4139 | slash_old - old != slash_new - new || | |
4140 | memcmp(old, new, slash_new - new)) | |
4141 | break; | |
4142 | old = slash_old + 1; | |
4143 | new = slash_new + 1; | |
4144 | } | |
4145 | /* p->old_name thru old is the common prefix, and old and new | |
4146 | * through the end of names are renames | |
4147 | */ | |
4148 | if (old != p->old_name) | |
4149 | printf(" %s %.*s{%s => %s} (%d%%)\n", renamecopy, | |
4150 | (int)(old - p->old_name), p->old_name, | |
4151 | old, new, p->score); | |
4152 | else | |
4153 | printf(" %s %s => %s (%d%%)\n", renamecopy, | |
4154 | p->old_name, p->new_name, p->score); | |
4155 | show_mode_change(p, 0); | |
4156 | } | |
4157 | ||
4158 | static void summary_patch_list(struct patch *patch) | |
4159 | { | |
4160 | struct patch *p; | |
4161 | ||
4162 | for (p = patch; p; p = p->next) { | |
4163 | if (p->is_new) | |
4164 | show_file_mode_name("create", p->new_mode, p->new_name); | |
4165 | else if (p->is_delete) | |
4166 | show_file_mode_name("delete", p->old_mode, p->old_name); | |
4167 | else { | |
4168 | if (p->is_rename || p->is_copy) | |
4169 | show_rename_copy(p); | |
4170 | else { | |
4171 | if (p->score) { | |
4172 | printf(" rewrite %s (%d%%)\n", | |
4173 | p->new_name, p->score); | |
4174 | show_mode_change(p, 0); | |
4175 | } | |
4176 | else | |
4177 | show_mode_change(p, 1); | |
4178 | } | |
4179 | } | |
4180 | } | |
4181 | } | |
4182 | ||
4183 | static void patch_stats(struct apply_state *state, struct patch *patch) | |
4184 | { | |
4185 | int lines = patch->lines_added + patch->lines_deleted; | |
4186 | ||
4187 | if (lines > state->max_change) | |
4188 | state->max_change = lines; | |
4189 | if (patch->old_name) { | |
4190 | int len = quote_c_style(patch->old_name, NULL, NULL, 0); | |
4191 | if (!len) | |
4192 | len = strlen(patch->old_name); | |
4193 | if (len > state->max_len) | |
4194 | state->max_len = len; | |
4195 | } | |
4196 | if (patch->new_name) { | |
4197 | int len = quote_c_style(patch->new_name, NULL, NULL, 0); | |
4198 | if (!len) | |
4199 | len = strlen(patch->new_name); | |
4200 | if (len > state->max_len) | |
4201 | state->max_len = len; | |
4202 | } | |
4203 | } | |
4204 | ||
4205 | static int remove_file(struct apply_state *state, struct patch *patch, int rmdir_empty) | |
4206 | { | |
4207 | if (state->update_index) { | |
4208 | if (remove_file_from_cache(patch->old_name) < 0) | |
4209 | return error(_("unable to remove %s from index"), patch->old_name); | |
4210 | } | |
4211 | if (!state->cached) { | |
4212 | if (!remove_or_warn(patch->old_mode, patch->old_name) && rmdir_empty) { | |
4213 | remove_path(patch->old_name); | |
4214 | } | |
4215 | } | |
4216 | return 0; | |
4217 | } | |
4218 | ||
4219 | static int add_index_file(struct apply_state *state, | |
4220 | const char *path, | |
4221 | unsigned mode, | |
4222 | void *buf, | |
4223 | unsigned long size) | |
4224 | { | |
4225 | struct stat st; | |
4226 | struct cache_entry *ce; | |
4227 | int namelen = strlen(path); | |
4228 | unsigned ce_size = cache_entry_size(namelen); | |
4229 | ||
4230 | if (!state->update_index) | |
4231 | return 0; | |
4232 | ||
4233 | ce = xcalloc(1, ce_size); | |
4234 | memcpy(ce->name, path, namelen); | |
4235 | ce->ce_mode = create_ce_mode(mode); | |
4236 | ce->ce_flags = create_ce_flags(0); | |
4237 | ce->ce_namelen = namelen; | |
4238 | if (S_ISGITLINK(mode)) { | |
4239 | const char *s; | |
4240 | ||
4241 | if (!skip_prefix(buf, "Subproject commit ", &s) || | |
4242 | get_sha1_hex(s, ce->sha1)) { | |
4243 | free(ce); | |
4244 | return error(_("corrupt patch for submodule %s"), path); | |
4245 | } | |
4246 | } else { | |
4247 | if (!state->cached) { | |
4248 | if (lstat(path, &st) < 0) { | |
4249 | free(ce); | |
90875eca CC |
4250 | return error_errno(_("unable to stat newly " |
4251 | "created file '%s'"), | |
4252 | path); | |
13b5af22 CC |
4253 | } |
4254 | fill_stat_cache_info(ce, &st); | |
4255 | } | |
4256 | if (write_sha1_file(buf, size, blob_type, ce->sha1) < 0) { | |
4257 | free(ce); | |
4258 | return error(_("unable to create backing store " | |
4259 | "for newly created file %s"), path); | |
4260 | } | |
4261 | } | |
4262 | if (add_cache_entry(ce, ADD_CACHE_OK_TO_ADD) < 0) { | |
4263 | free(ce); | |
4264 | return error(_("unable to add cache entry for %s"), path); | |
4265 | } | |
4266 | ||
4267 | return 0; | |
4268 | } | |
4269 | ||
4270 | /* | |
4271 | * Returns: | |
4272 | * -1 if an unrecoverable error happened | |
4273 | * 0 if everything went well | |
4274 | * 1 if a recoverable error happened | |
4275 | */ | |
4276 | static int try_create_file(const char *path, unsigned int mode, const char *buf, unsigned long size) | |
4277 | { | |
4278 | int fd, res; | |
4279 | struct strbuf nbuf = STRBUF_INIT; | |
4280 | ||
4281 | if (S_ISGITLINK(mode)) { | |
4282 | struct stat st; | |
4283 | if (!lstat(path, &st) && S_ISDIR(st.st_mode)) | |
4284 | return 0; | |
4285 | return !!mkdir(path, 0777); | |
4286 | } | |
4287 | ||
4288 | if (has_symlinks && S_ISLNK(mode)) | |
4289 | /* Although buf:size is counted string, it also is NUL | |
4290 | * terminated. | |
4291 | */ | |
4292 | return !!symlink(buf, path); | |
4293 | ||
4294 | fd = open(path, O_CREAT | O_EXCL | O_WRONLY, (mode & 0100) ? 0777 : 0666); | |
4295 | if (fd < 0) | |
4296 | return 1; | |
4297 | ||
4298 | if (convert_to_working_tree(path, buf, size, &nbuf)) { | |
4299 | size = nbuf.len; | |
4300 | buf = nbuf.buf; | |
4301 | } | |
4302 | ||
4303 | res = write_in_full(fd, buf, size) < 0; | |
4304 | if (res) | |
4305 | error_errno(_("failed to write to '%s'"), path); | |
4306 | strbuf_release(&nbuf); | |
4307 | ||
4308 | if (close(fd) < 0 && !res) | |
4309 | return error_errno(_("closing file '%s'"), path); | |
4310 | ||
4311 | return res ? -1 : 0; | |
4312 | } | |
4313 | ||
4314 | /* | |
4315 | * We optimistically assume that the directories exist, | |
4316 | * which is true 99% of the time anyway. If they don't, | |
4317 | * we create them and try again. | |
4318 | * | |
4319 | * Returns: | |
4320 | * -1 on error | |
4321 | * 0 otherwise | |
4322 | */ | |
4323 | static int create_one_file(struct apply_state *state, | |
4324 | char *path, | |
4325 | unsigned mode, | |
4326 | const char *buf, | |
4327 | unsigned long size) | |
4328 | { | |
4329 | int res; | |
4330 | ||
4331 | if (state->cached) | |
4332 | return 0; | |
4333 | ||
4334 | res = try_create_file(path, mode, buf, size); | |
4335 | if (res < 0) | |
4336 | return -1; | |
4337 | if (!res) | |
4338 | return 0; | |
4339 | ||
4340 | if (errno == ENOENT) { | |
4341 | if (safe_create_leading_directories(path)) | |
4342 | return 0; | |
4343 | res = try_create_file(path, mode, buf, size); | |
4344 | if (res < 0) | |
4345 | return -1; | |
4346 | if (!res) | |
4347 | return 0; | |
4348 | } | |
4349 | ||
4350 | if (errno == EEXIST || errno == EACCES) { | |
4351 | /* We may be trying to create a file where a directory | |
4352 | * used to be. | |
4353 | */ | |
4354 | struct stat st; | |
4355 | if (!lstat(path, &st) && (!S_ISDIR(st.st_mode) || !rmdir(path))) | |
4356 | errno = EEXIST; | |
4357 | } | |
4358 | ||
4359 | if (errno == EEXIST) { | |
4360 | unsigned int nr = getpid(); | |
4361 | ||
4362 | for (;;) { | |
4363 | char newpath[PATH_MAX]; | |
4364 | mksnpath(newpath, sizeof(newpath), "%s~%u", path, nr); | |
4365 | res = try_create_file(newpath, mode, buf, size); | |
4366 | if (res < 0) | |
4367 | return -1; | |
4368 | if (!res) { | |
4369 | if (!rename(newpath, path)) | |
4370 | return 0; | |
4371 | unlink_or_warn(newpath); | |
4372 | break; | |
4373 | } | |
4374 | if (errno != EEXIST) | |
4375 | break; | |
4376 | ++nr; | |
4377 | } | |
4378 | } | |
4379 | return error_errno(_("unable to write file '%s' mode %o"), | |
4380 | path, mode); | |
4381 | } | |
4382 | ||
4383 | static int add_conflicted_stages_file(struct apply_state *state, | |
4384 | struct patch *patch) | |
4385 | { | |
4386 | int stage, namelen; | |
4387 | unsigned ce_size, mode; | |
4388 | struct cache_entry *ce; | |
4389 | ||
4390 | if (!state->update_index) | |
4391 | return 0; | |
4392 | namelen = strlen(patch->new_name); | |
4393 | ce_size = cache_entry_size(namelen); | |
4394 | mode = patch->new_mode ? patch->new_mode : (S_IFREG | 0644); | |
4395 | ||
4396 | remove_file_from_cache(patch->new_name); | |
4397 | for (stage = 1; stage < 4; stage++) { | |
4398 | if (is_null_oid(&patch->threeway_stage[stage - 1])) | |
4399 | continue; | |
4400 | ce = xcalloc(1, ce_size); | |
4401 | memcpy(ce->name, patch->new_name, namelen); | |
4402 | ce->ce_mode = create_ce_mode(mode); | |
4403 | ce->ce_flags = create_ce_flags(stage); | |
4404 | ce->ce_namelen = namelen; | |
4405 | hashcpy(ce->sha1, patch->threeway_stage[stage - 1].hash); | |
4406 | if (add_cache_entry(ce, ADD_CACHE_OK_TO_ADD) < 0) { | |
4407 | free(ce); | |
4408 | return error(_("unable to add cache entry for %s"), | |
4409 | patch->new_name); | |
4410 | } | |
4411 | } | |
4412 | ||
4413 | return 0; | |
4414 | } | |
4415 | ||
4416 | static int create_file(struct apply_state *state, struct patch *patch) | |
4417 | { | |
4418 | char *path = patch->new_name; | |
4419 | unsigned mode = patch->new_mode; | |
4420 | unsigned long size = patch->resultsize; | |
4421 | char *buf = patch->result; | |
4422 | ||
4423 | if (!mode) | |
4424 | mode = S_IFREG | 0644; | |
4425 | if (create_one_file(state, path, mode, buf, size)) | |
4426 | return -1; | |
4427 | ||
4428 | if (patch->conflicted_threeway) | |
4429 | return add_conflicted_stages_file(state, patch); | |
4430 | else | |
4431 | return add_index_file(state, path, mode, buf, size); | |
4432 | } | |
4433 | ||
4434 | /* phase zero is to remove, phase one is to create */ | |
4435 | static int write_out_one_result(struct apply_state *state, | |
4436 | struct patch *patch, | |
4437 | int phase) | |
4438 | { | |
4439 | if (patch->is_delete > 0) { | |
4440 | if (phase == 0) | |
4441 | return remove_file(state, patch, 1); | |
4442 | return 0; | |
4443 | } | |
4444 | if (patch->is_new > 0 || patch->is_copy) { | |
4445 | if (phase == 1) | |
4446 | return create_file(state, patch); | |
4447 | return 0; | |
4448 | } | |
4449 | /* | |
4450 | * Rename or modification boils down to the same | |
4451 | * thing: remove the old, write the new | |
4452 | */ | |
4453 | if (phase == 0) | |
4454 | return remove_file(state, patch, patch->is_rename); | |
4455 | if (phase == 1) | |
4456 | return create_file(state, patch); | |
4457 | return 0; | |
4458 | } | |
4459 | ||
4460 | static int write_out_one_reject(struct apply_state *state, struct patch *patch) | |
4461 | { | |
4462 | FILE *rej; | |
4463 | char namebuf[PATH_MAX]; | |
4464 | struct fragment *frag; | |
4465 | int cnt = 0; | |
4466 | struct strbuf sb = STRBUF_INIT; | |
4467 | ||
4468 | for (cnt = 0, frag = patch->fragments; frag; frag = frag->next) { | |
4469 | if (!frag->rejected) | |
4470 | continue; | |
4471 | cnt++; | |
4472 | } | |
4473 | ||
4474 | if (!cnt) { | |
4475 | if (state->apply_verbosely) | |
4476 | say_patch_name(stderr, | |
4477 | _("Applied patch %s cleanly."), patch); | |
4478 | return 0; | |
4479 | } | |
4480 | ||
4481 | /* This should not happen, because a removal patch that leaves | |
4482 | * contents are marked "rejected" at the patch level. | |
4483 | */ | |
4484 | if (!patch->new_name) | |
4485 | die(_("internal error")); | |
4486 | ||
4487 | /* Say this even without --verbose */ | |
4488 | strbuf_addf(&sb, Q_("Applying patch %%s with %d reject...", | |
4489 | "Applying patch %%s with %d rejects...", | |
4490 | cnt), | |
4491 | cnt); | |
4492 | say_patch_name(stderr, sb.buf, patch); | |
4493 | strbuf_release(&sb); | |
4494 | ||
4495 | cnt = strlen(patch->new_name); | |
4496 | if (ARRAY_SIZE(namebuf) <= cnt + 5) { | |
4497 | cnt = ARRAY_SIZE(namebuf) - 5; | |
4498 | warning(_("truncating .rej filename to %.*s.rej"), | |
4499 | cnt - 1, patch->new_name); | |
4500 | } | |
4501 | memcpy(namebuf, patch->new_name, cnt); | |
4502 | memcpy(namebuf + cnt, ".rej", 5); | |
4503 | ||
4504 | rej = fopen(namebuf, "w"); | |
4505 | if (!rej) | |
90875eca | 4506 | return error_errno(_("cannot open %s"), namebuf); |
13b5af22 CC |
4507 | |
4508 | /* Normal git tools never deal with .rej, so do not pretend | |
4509 | * this is a git patch by saying --git or giving extended | |
4510 | * headers. While at it, maybe please "kompare" that wants | |
4511 | * the trailing TAB and some garbage at the end of line ;-). | |
4512 | */ | |
4513 | fprintf(rej, "diff a/%s b/%s\t(rejected hunks)\n", | |
4514 | patch->new_name, patch->new_name); | |
4515 | for (cnt = 1, frag = patch->fragments; | |
4516 | frag; | |
4517 | cnt++, frag = frag->next) { | |
4518 | if (!frag->rejected) { | |
4519 | fprintf_ln(stderr, _("Hunk #%d applied cleanly."), cnt); | |
4520 | continue; | |
4521 | } | |
4522 | fprintf_ln(stderr, _("Rejected hunk #%d."), cnt); | |
4523 | fprintf(rej, "%.*s", frag->size, frag->patch); | |
4524 | if (frag->patch[frag->size-1] != '\n') | |
4525 | fputc('\n', rej); | |
4526 | } | |
4527 | fclose(rej); | |
4528 | return -1; | |
4529 | } | |
4530 | ||
4531 | /* | |
4532 | * Returns: | |
4533 | * -1 if an error happened | |
4534 | * 0 if the patch applied cleanly | |
4535 | * 1 if the patch did not apply cleanly | |
4536 | */ | |
4537 | static int write_out_results(struct apply_state *state, struct patch *list) | |
4538 | { | |
4539 | int phase; | |
4540 | int errs = 0; | |
4541 | struct patch *l; | |
4542 | struct string_list cpath = STRING_LIST_INIT_DUP; | |
4543 | ||
4544 | for (phase = 0; phase < 2; phase++) { | |
4545 | l = list; | |
4546 | while (l) { | |
4547 | if (l->rejected) | |
4548 | errs = 1; | |
4549 | else { | |
4550 | if (write_out_one_result(state, l, phase)) { | |
4551 | string_list_clear(&cpath, 0); | |
4552 | return -1; | |
4553 | } | |
4554 | if (phase == 1) { | |
4555 | if (write_out_one_reject(state, l)) | |
4556 | errs = 1; | |
4557 | if (l->conflicted_threeway) { | |
4558 | string_list_append(&cpath, l->new_name); | |
4559 | errs = 1; | |
4560 | } | |
4561 | } | |
4562 | } | |
4563 | l = l->next; | |
4564 | } | |
4565 | } | |
4566 | ||
4567 | if (cpath.nr) { | |
4568 | struct string_list_item *item; | |
4569 | ||
4570 | string_list_sort(&cpath); | |
4571 | for_each_string_list_item(item, &cpath) | |
4572 | fprintf(stderr, "U %s\n", item->string); | |
4573 | string_list_clear(&cpath, 0); | |
4574 | ||
4575 | rerere(0); | |
4576 | } | |
4577 | ||
4578 | return errs; | |
4579 | } | |
4580 | ||
4581 | /* | |
4582 | * Try to apply a patch. | |
4583 | * | |
4584 | * Returns: | |
4585 | * -128 if a bad error happened (like patch unreadable) | |
4586 | * -1 if patch did not apply and user cannot deal with it | |
4587 | * 0 if the patch applied | |
4588 | * 1 if the patch did not apply but user might fix it | |
4589 | */ | |
4590 | static int apply_patch(struct apply_state *state, | |
4591 | int fd, | |
4592 | const char *filename, | |
4593 | int options) | |
4594 | { | |
4595 | size_t offset; | |
4596 | struct strbuf buf = STRBUF_INIT; /* owns the patch text */ | |
4597 | struct patch *list = NULL, **listp = &list; | |
4598 | int skipped_patch = 0; | |
4599 | int res = 0; | |
4600 | ||
4601 | state->patch_input_file = filename; | |
4602 | if (read_patch_file(&buf, fd) < 0) | |
4603 | return -128; | |
4604 | offset = 0; | |
4605 | while (offset < buf.len) { | |
4606 | struct patch *patch; | |
4607 | int nr; | |
4608 | ||
4609 | patch = xcalloc(1, sizeof(*patch)); | |
4610 | patch->inaccurate_eof = !!(options & APPLY_OPT_INACCURATE_EOF); | |
4611 | patch->recount = !!(options & APPLY_OPT_RECOUNT); | |
4612 | nr = parse_chunk(state, buf.buf + offset, buf.len - offset, patch); | |
4613 | if (nr < 0) { | |
4614 | free_patch(patch); | |
4615 | if (nr == -128) { | |
4616 | res = -128; | |
4617 | goto end; | |
4618 | } | |
4619 | break; | |
4620 | } | |
4621 | if (state->apply_in_reverse) | |
4622 | reverse_patches(patch); | |
4623 | if (use_patch(state, patch)) { | |
4624 | patch_stats(state, patch); | |
4625 | *listp = patch; | |
4626 | listp = &patch->next; | |
4627 | } | |
4628 | else { | |
4629 | if (state->apply_verbosely) | |
4630 | say_patch_name(stderr, _("Skipped patch '%s'."), patch); | |
4631 | free_patch(patch); | |
4632 | skipped_patch++; | |
4633 | } | |
4634 | offset += nr; | |
4635 | } | |
4636 | ||
4637 | if (!list && !skipped_patch) { | |
4638 | error(_("unrecognized input")); | |
4639 | res = -128; | |
4640 | goto end; | |
4641 | } | |
4642 | ||
4643 | if (state->whitespace_error && (state->ws_error_action == die_on_ws_error)) | |
4644 | state->apply = 0; | |
4645 | ||
4646 | state->update_index = state->check_index && state->apply; | |
4647 | if (state->update_index && state->newfd < 0) | |
4648 | state->newfd = hold_locked_index(state->lock_file, 1); | |
4649 | ||
4650 | if (state->check_index && read_cache() < 0) { | |
4651 | error(_("unable to read index file")); | |
4652 | res = -128; | |
4653 | goto end; | |
4654 | } | |
4655 | ||
4656 | if (state->check || state->apply) { | |
4657 | int r = check_patch_list(state, list); | |
4658 | if (r == -128) { | |
4659 | res = -128; | |
4660 | goto end; | |
4661 | } | |
4662 | if (r < 0 && !state->apply_with_reject) { | |
4663 | res = -1; | |
4664 | goto end; | |
4665 | } | |
4666 | } | |
4667 | ||
4668 | if (state->apply) { | |
4669 | int write_res = write_out_results(state, list); | |
4670 | if (write_res < 0) { | |
4671 | res = -128; | |
4672 | goto end; | |
4673 | } | |
4674 | if (write_res > 0) { | |
4675 | /* with --3way, we still need to write the index out */ | |
4676 | res = state->apply_with_reject ? -1 : 1; | |
4677 | goto end; | |
4678 | } | |
4679 | } | |
4680 | ||
4681 | if (state->fake_ancestor && | |
4682 | build_fake_ancestor(list, state->fake_ancestor)) { | |
4683 | res = -128; | |
4684 | goto end; | |
4685 | } | |
4686 | ||
4687 | if (state->diffstat) | |
4688 | stat_patch_list(state, list); | |
4689 | ||
4690 | if (state->numstat) | |
4691 | numstat_patch_list(state, list); | |
4692 | ||
4693 | if (state->summary) | |
4694 | summary_patch_list(list); | |
4695 | ||
4696 | end: | |
4697 | free_patch_list(list); | |
4698 | strbuf_release(&buf); | |
4699 | string_list_clear(&state->fn_table, 0); | |
4700 | return res; | |
4701 | } | |
4702 | ||
4703 | int apply_option_parse_exclude(const struct option *opt, | |
4704 | const char *arg, int unset) | |
4705 | { | |
4706 | struct apply_state *state = opt->value; | |
4707 | add_name_limit(state, arg, 1); | |
4708 | return 0; | |
4709 | } | |
4710 | ||
4711 | int apply_option_parse_include(const struct option *opt, | |
4712 | const char *arg, int unset) | |
4713 | { | |
4714 | struct apply_state *state = opt->value; | |
4715 | add_name_limit(state, arg, 0); | |
4716 | state->has_include = 1; | |
4717 | return 0; | |
4718 | } | |
4719 | ||
4720 | int apply_option_parse_p(const struct option *opt, | |
4721 | const char *arg, | |
4722 | int unset) | |
4723 | { | |
4724 | struct apply_state *state = opt->value; | |
4725 | state->p_value = atoi(arg); | |
4726 | state->p_value_known = 1; | |
4727 | return 0; | |
4728 | } | |
4729 | ||
4730 | int apply_option_parse_space_change(const struct option *opt, | |
4731 | const char *arg, int unset) | |
4732 | { | |
4733 | struct apply_state *state = opt->value; | |
4734 | if (unset) | |
4735 | state->ws_ignore_action = ignore_ws_none; | |
4736 | else | |
4737 | state->ws_ignore_action = ignore_ws_change; | |
4738 | return 0; | |
4739 | } | |
4740 | ||
4741 | int apply_option_parse_whitespace(const struct option *opt, | |
4742 | const char *arg, int unset) | |
4743 | { | |
4744 | struct apply_state *state = opt->value; | |
4745 | state->whitespace_option = arg; | |
4746 | if (parse_whitespace_option(state, arg)) | |
4747 | exit(1); | |
4748 | return 0; | |
4749 | } | |
4750 | ||
4751 | int apply_option_parse_directory(const struct option *opt, | |
4752 | const char *arg, int unset) | |
4753 | { | |
4754 | struct apply_state *state = opt->value; | |
4755 | strbuf_reset(&state->root); | |
4756 | strbuf_addstr(&state->root, arg); | |
4757 | strbuf_complete(&state->root, '/'); | |
4758 | return 0; | |
4759 | } | |
4760 | ||
4761 | int apply_all_patches(struct apply_state *state, | |
4762 | int argc, | |
4763 | const char **argv, | |
4764 | int options) | |
4765 | { | |
4766 | int i; | |
4767 | int res; | |
4768 | int errs = 0; | |
4769 | int read_stdin = 1; | |
4770 | ||
4771 | for (i = 0; i < argc; i++) { | |
4772 | const char *arg = argv[i]; | |
4773 | int fd; | |
4774 | ||
4775 | if (!strcmp(arg, "-")) { | |
4776 | res = apply_patch(state, 0, "<stdin>", options); | |
4777 | if (res < 0) | |
4778 | goto end; | |
4779 | errs |= res; | |
4780 | read_stdin = 0; | |
4781 | continue; | |
4782 | } else if (0 < state->prefix_length) | |
4783 | arg = prefix_filename(state->prefix, | |
4784 | state->prefix_length, | |
4785 | arg); | |
4786 | ||
4787 | fd = open(arg, O_RDONLY); | |
4788 | if (fd < 0) { | |
4789 | error(_("can't open patch '%s': %s"), arg, strerror(errno)); | |
4790 | res = -128; | |
4791 | goto end; | |
4792 | } | |
4793 | read_stdin = 0; | |
4794 | set_default_whitespace_mode(state); | |
4795 | res = apply_patch(state, fd, arg, options); | |
4796 | close(fd); | |
4797 | if (res < 0) | |
4798 | goto end; | |
4799 | errs |= res; | |
4800 | } | |
4801 | set_default_whitespace_mode(state); | |
4802 | if (read_stdin) { | |
4803 | res = apply_patch(state, 0, "<stdin>", options); | |
4804 | if (res < 0) | |
4805 | goto end; | |
4806 | errs |= res; | |
4807 | } | |
4808 | ||
4809 | if (state->whitespace_error) { | |
4810 | if (state->squelch_whitespace_errors && | |
4811 | state->squelch_whitespace_errors < state->whitespace_error) { | |
4812 | int squelched = | |
4813 | state->whitespace_error - state->squelch_whitespace_errors; | |
4814 | warning(Q_("squelched %d whitespace error", | |
4815 | "squelched %d whitespace errors", | |
4816 | squelched), | |
4817 | squelched); | |
4818 | } | |
4819 | if (state->ws_error_action == die_on_ws_error) { | |
4820 | error(Q_("%d line adds whitespace errors.", | |
4821 | "%d lines add whitespace errors.", | |
4822 | state->whitespace_error), | |
4823 | state->whitespace_error); | |
4824 | res = -128; | |
4825 | goto end; | |
4826 | } | |
4827 | if (state->applied_after_fixing_ws && state->apply) | |
4828 | warning("%d line%s applied after" | |
4829 | " fixing whitespace errors.", | |
4830 | state->applied_after_fixing_ws, | |
4831 | state->applied_after_fixing_ws == 1 ? "" : "s"); | |
4832 | else if (state->whitespace_error) | |
4833 | warning(Q_("%d line adds whitespace errors.", | |
4834 | "%d lines add whitespace errors.", | |
4835 | state->whitespace_error), | |
4836 | state->whitespace_error); | |
4837 | } | |
4838 | ||
4839 | if (state->update_index) { | |
4840 | res = write_locked_index(&the_index, state->lock_file, COMMIT_LOCK); | |
4841 | if (res) { | |
4842 | error(_("Unable to write new index file")); | |
4843 | res = -128; | |
4844 | goto end; | |
4845 | } | |
4846 | state->newfd = -1; | |
4847 | } | |
4848 | ||
4849 | return !!errs; | |
4850 | ||
4851 | end: | |
4852 | if (state->newfd >= 0) { | |
4853 | rollback_lock_file(state->lock_file); | |
4854 | state->newfd = -1; | |
4855 | } | |
4856 | ||
4857 | return (res == -1 ? 1 : 128); | |
4858 | } |