Merge in xdiff cleanup pieces
[git/mingw/4msysgit/kblees.git] / diff.c
blobe496905bad9853f912e1bbcc081c544524f8cb77
1 /*
2 * Copyright (C) 2005 Junio C Hamano
3 */
4 #include <sys/types.h>
5 #include <sys/wait.h>
6 #include <signal.h>
7 #include "cache.h"
8 #include "quote.h"
9 #include "diff.h"
10 #include "diffcore.h"
11 #include "xdiff/xdiff.h"
13 static int use_size_cache;
15 int diff_rename_limit_default = -1;
17 int git_diff_config(const char *var, const char *value)
19 if (!strcmp(var, "diff.renamelimit")) {
20 diff_rename_limit_default = git_config_int(var, value);
21 return 0;
24 return git_default_config(var, value);
27 static char *quote_one(const char *str)
29 int needlen;
30 char *xp;
32 if (!str)
33 return NULL;
34 needlen = quote_c_style(str, NULL, NULL, 0);
35 if (!needlen)
36 return strdup(str);
37 xp = xmalloc(needlen + 1);
38 quote_c_style(str, xp, NULL, 0);
39 return xp;
42 static char *quote_two(const char *one, const char *two)
44 int need_one = quote_c_style(one, NULL, NULL, 1);
45 int need_two = quote_c_style(two, NULL, NULL, 1);
46 char *xp;
48 if (need_one + need_two) {
49 if (!need_one) need_one = strlen(one);
50 if (!need_two) need_one = strlen(two);
52 xp = xmalloc(need_one + need_two + 3);
53 xp[0] = '"';
54 quote_c_style(one, xp + 1, NULL, 1);
55 quote_c_style(two, xp + need_one + 1, NULL, 1);
56 strcpy(xp + need_one + need_two + 1, "\"");
57 return xp;
59 need_one = strlen(one);
60 need_two = strlen(two);
61 xp = xmalloc(need_one + need_two + 1);
62 strcpy(xp, one);
63 strcpy(xp + need_one, two);
64 return xp;
67 static const char *external_diff(void)
69 static const char *external_diff_cmd = NULL;
70 static int done_preparing = 0;
72 if (done_preparing)
73 return external_diff_cmd;
74 external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
75 done_preparing = 1;
76 return external_diff_cmd;
79 #define TEMPFILE_PATH_LEN 50
81 static struct diff_tempfile {
82 const char *name; /* filename external diff should read from */
83 char hex[41];
84 char mode[10];
85 char tmp_path[TEMPFILE_PATH_LEN];
86 } diff_temp[2];
88 static int count_lines(const char *data, int size)
90 int count, ch, completely_empty = 1, nl_just_seen = 0;
91 count = 0;
92 while (0 < size--) {
93 ch = *data++;
94 if (ch == '\n') {
95 count++;
96 nl_just_seen = 1;
97 completely_empty = 0;
99 else {
100 nl_just_seen = 0;
101 completely_empty = 0;
104 if (completely_empty)
105 return 0;
106 if (!nl_just_seen)
107 count++; /* no trailing newline */
108 return count;
111 static void print_line_count(int count)
113 switch (count) {
114 case 0:
115 printf("0,0");
116 break;
117 case 1:
118 printf("1");
119 break;
120 default:
121 printf("1,%d", count);
122 break;
126 static void copy_file(int prefix, const char *data, int size)
128 int ch, nl_just_seen = 1;
129 while (0 < size--) {
130 ch = *data++;
131 if (nl_just_seen)
132 putchar(prefix);
133 putchar(ch);
134 if (ch == '\n')
135 nl_just_seen = 1;
136 else
137 nl_just_seen = 0;
139 if (!nl_just_seen)
140 printf("\n\\ No newline at end of file\n");
143 static void emit_rewrite_diff(const char *name_a,
144 const char *name_b,
145 struct diff_filespec *one,
146 struct diff_filespec *two)
148 /* Use temp[i].name as input, name_a and name_b as labels */
149 int lc_a, lc_b;
150 lc_a = count_lines(one->data, one->size);
151 lc_b = count_lines(two->data, two->size);
152 printf("--- %s\n+++ %s\n@@ -", name_a, name_b);
153 print_line_count(lc_a);
154 printf(" +");
155 print_line_count(lc_b);
156 printf(" @@\n");
157 if (lc_a)
158 copy_file('-', one->data, one->size);
159 if (lc_b)
160 copy_file('+', two->data, two->size);
163 static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
165 if (!DIFF_FILE_VALID(one)) {
166 mf->ptr = ""; /* does not matter */
167 mf->size = 0;
168 return 0;
170 else if (diff_populate_filespec(one, 0))
171 return -1;
172 mf->ptr = one->data;
173 mf->size = one->size;
174 return 0;
177 struct emit_callback {
178 const char **label_path;
181 static int fn_out(void *priv, mmbuffer_t *mb, int nbuf)
183 int i;
184 struct emit_callback *ecbdata = priv;
186 if (ecbdata->label_path[0]) {
187 printf("--- %s\n", ecbdata->label_path[0]);
188 printf("+++ %s\n", ecbdata->label_path[1]);
189 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
191 for (i = 0; i < nbuf; i++)
192 if (!fwrite(mb[i].ptr, mb[i].size, 1, stdout))
193 return -1;
194 return 0;
197 #define FIRST_FEW_BYTES 8000
198 static int mmfile_is_binary(mmfile_t *mf)
200 long sz = mf->size;
201 if (FIRST_FEW_BYTES < sz)
202 sz = FIRST_FEW_BYTES;
203 if (memchr(mf->ptr, 0, sz))
204 return 1;
205 return 0;
208 static void builtin_diff(const char *name_a,
209 const char *name_b,
210 struct diff_filespec *one,
211 struct diff_filespec *two,
212 const char *xfrm_msg,
213 int complete_rewrite)
215 mmfile_t mf1, mf2;
216 const char *lbl[2];
217 char *a_one, *b_two;
219 a_one = quote_two("a/", name_a);
220 b_two = quote_two("b/", name_b);
221 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
222 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
223 printf("diff --git %s %s\n", a_one, b_two);
224 if (lbl[0][0] == '/') {
225 /* /dev/null */
226 printf("new file mode %06o\n", two->mode);
227 if (xfrm_msg && xfrm_msg[0])
228 puts(xfrm_msg);
230 else if (lbl[1][0] == '/') {
231 printf("deleted file mode %06o\n", one->mode);
232 if (xfrm_msg && xfrm_msg[0])
233 puts(xfrm_msg);
235 else {
236 if (one->mode != two->mode) {
237 printf("old mode %06o\n", one->mode);
238 printf("new mode %06o\n", two->mode);
240 if (xfrm_msg && xfrm_msg[0])
241 puts(xfrm_msg);
243 * we do not run diff between different kind
244 * of objects.
246 if ((one->mode ^ two->mode) & S_IFMT)
247 goto free_ab_and_return;
248 if (complete_rewrite) {
249 emit_rewrite_diff(name_a, name_b, one, two);
250 goto free_ab_and_return;
254 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
255 die("unable to read files to diff");
257 if (mmfile_is_binary(&mf1) || mmfile_is_binary(&mf2))
258 printf("Binary files %s and %s differ\n", lbl[0], lbl[1]);
259 else {
260 /* Crazy xdl interfaces.. */
261 const char *diffopts = getenv("GIT_DIFF_OPTS");
262 xpparam_t xpp;
263 xdemitconf_t xecfg;
264 xdemitcb_t ecb;
265 struct emit_callback ecbdata;
267 ecbdata.label_path = lbl;
268 xpp.flags = XDF_NEED_MINIMAL;
269 xecfg.ctxlen = 3;
270 xecfg.flags = XDL_EMIT_FUNCNAMES;
271 if (!diffopts)
273 else if (!strncmp(diffopts, "--unified=", 10))
274 xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
275 else if (!strncmp(diffopts, "-u", 2))
276 xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
277 ecb.outf = fn_out;
278 ecb.priv = &ecbdata;
279 xdl_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
282 free_ab_and_return:
283 free(a_one);
284 free(b_two);
285 return;
288 struct diff_filespec *alloc_filespec(const char *path)
290 int namelen = strlen(path);
291 struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
293 memset(spec, 0, sizeof(*spec));
294 spec->path = (char *)(spec + 1);
295 memcpy(spec->path, path, namelen+1);
296 return spec;
299 void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
300 unsigned short mode)
302 if (mode) {
303 spec->mode = canon_mode(mode);
304 memcpy(spec->sha1, sha1, 20);
305 spec->sha1_valid = !!memcmp(sha1, null_sha1, 20);
310 * Given a name and sha1 pair, if the dircache tells us the file in
311 * the work tree has that object contents, return true, so that
312 * prepare_temp_file() does not have to inflate and extract.
314 static int work_tree_matches(const char *name, const unsigned char *sha1)
316 struct cache_entry *ce;
317 struct stat st;
318 int pos, len;
320 /* We do not read the cache ourselves here, because the
321 * benchmark with my previous version that always reads cache
322 * shows that it makes things worse for diff-tree comparing
323 * two linux-2.6 kernel trees in an already checked out work
324 * tree. This is because most diff-tree comparisons deal with
325 * only a small number of files, while reading the cache is
326 * expensive for a large project, and its cost outweighs the
327 * savings we get by not inflating the object to a temporary
328 * file. Practically, this code only helps when we are used
329 * by diff-cache --cached, which does read the cache before
330 * calling us.
332 if (!active_cache)
333 return 0;
335 len = strlen(name);
336 pos = cache_name_pos(name, len);
337 if (pos < 0)
338 return 0;
339 ce = active_cache[pos];
340 if ((lstat(name, &st) < 0) ||
341 !S_ISREG(st.st_mode) || /* careful! */
342 ce_match_stat(ce, &st, 0) ||
343 memcmp(sha1, ce->sha1, 20))
344 return 0;
345 /* we return 1 only when we can stat, it is a regular file,
346 * stat information matches, and sha1 recorded in the cache
347 * matches. I.e. we know the file in the work tree really is
348 * the same as the <name, sha1> pair.
350 return 1;
353 static struct sha1_size_cache {
354 unsigned char sha1[20];
355 unsigned long size;
356 } **sha1_size_cache;
357 static int sha1_size_cache_nr, sha1_size_cache_alloc;
359 static struct sha1_size_cache *locate_size_cache(unsigned char *sha1,
360 int find_only,
361 unsigned long size)
363 int first, last;
364 struct sha1_size_cache *e;
366 first = 0;
367 last = sha1_size_cache_nr;
368 while (last > first) {
369 int cmp, next = (last + first) >> 1;
370 e = sha1_size_cache[next];
371 cmp = memcmp(e->sha1, sha1, 20);
372 if (!cmp)
373 return e;
374 if (cmp < 0) {
375 last = next;
376 continue;
378 first = next+1;
380 /* not found */
381 if (find_only)
382 return NULL;
383 /* insert to make it at "first" */
384 if (sha1_size_cache_alloc <= sha1_size_cache_nr) {
385 sha1_size_cache_alloc = alloc_nr(sha1_size_cache_alloc);
386 sha1_size_cache = xrealloc(sha1_size_cache,
387 sha1_size_cache_alloc *
388 sizeof(*sha1_size_cache));
390 sha1_size_cache_nr++;
391 if (first < sha1_size_cache_nr)
392 memmove(sha1_size_cache + first + 1, sha1_size_cache + first,
393 (sha1_size_cache_nr - first - 1) *
394 sizeof(*sha1_size_cache));
395 e = xmalloc(sizeof(struct sha1_size_cache));
396 sha1_size_cache[first] = e;
397 memcpy(e->sha1, sha1, 20);
398 e->size = size;
399 return e;
403 * While doing rename detection and pickaxe operation, we may need to
404 * grab the data for the blob (or file) for our own in-core comparison.
405 * diff_filespec has data and size fields for this purpose.
407 int diff_populate_filespec(struct diff_filespec *s, int size_only)
409 int err = 0;
410 if (!DIFF_FILE_VALID(s))
411 die("internal error: asking to populate invalid file.");
412 if (S_ISDIR(s->mode))
413 return -1;
415 if (!use_size_cache)
416 size_only = 0;
418 if (s->data)
419 return err;
420 if (!s->sha1_valid ||
421 work_tree_matches(s->path, s->sha1)) {
422 struct stat st;
423 int fd;
424 if (lstat(s->path, &st) < 0) {
425 if (errno == ENOENT) {
426 err_empty:
427 err = -1;
428 empty:
429 s->data = "";
430 s->size = 0;
431 return err;
434 s->size = st.st_size;
435 if (!s->size)
436 goto empty;
437 if (size_only)
438 return 0;
439 if (S_ISLNK(st.st_mode)) {
440 int ret;
441 s->data = xmalloc(s->size);
442 s->should_free = 1;
443 ret = readlink(s->path, s->data, s->size);
444 if (ret < 0) {
445 free(s->data);
446 goto err_empty;
448 return 0;
450 fd = open(s->path, O_RDONLY);
451 if (fd < 0)
452 goto err_empty;
453 s->data = mmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
454 close(fd);
455 if (s->data == MAP_FAILED)
456 goto err_empty;
457 s->should_munmap = 1;
459 else {
460 char type[20];
461 struct sha1_size_cache *e;
463 if (size_only) {
464 e = locate_size_cache(s->sha1, 1, 0);
465 if (e) {
466 s->size = e->size;
467 return 0;
469 if (!sha1_object_info(s->sha1, type, &s->size))
470 locate_size_cache(s->sha1, 0, s->size);
472 else {
473 s->data = read_sha1_file(s->sha1, type, &s->size);
474 s->should_free = 1;
477 return 0;
480 void diff_free_filespec_data(struct diff_filespec *s)
482 if (s->should_free)
483 free(s->data);
484 else if (s->should_munmap)
485 munmap(s->data, s->size);
486 s->should_free = s->should_munmap = 0;
487 s->data = NULL;
488 free(s->cnt_data);
489 s->cnt_data = NULL;
492 static void prep_temp_blob(struct diff_tempfile *temp,
493 void *blob,
494 unsigned long size,
495 const unsigned char *sha1,
496 int mode)
498 int fd;
500 fd = git_mkstemp(temp->tmp_path, TEMPFILE_PATH_LEN, ".diff_XXXXXX");
501 if (fd < 0)
502 die("unable to create temp-file");
503 if (write(fd, blob, size) != size)
504 die("unable to write temp-file");
505 close(fd);
506 temp->name = temp->tmp_path;
507 strcpy(temp->hex, sha1_to_hex(sha1));
508 temp->hex[40] = 0;
509 sprintf(temp->mode, "%06o", mode);
512 static void prepare_temp_file(const char *name,
513 struct diff_tempfile *temp,
514 struct diff_filespec *one)
516 if (!DIFF_FILE_VALID(one)) {
517 not_a_valid_file:
518 /* A '-' entry produces this for file-2, and
519 * a '+' entry produces this for file-1.
521 temp->name = "/dev/null";
522 strcpy(temp->hex, ".");
523 strcpy(temp->mode, ".");
524 return;
527 if (!one->sha1_valid ||
528 work_tree_matches(name, one->sha1)) {
529 struct stat st;
530 if (lstat(name, &st) < 0) {
531 if (errno == ENOENT)
532 goto not_a_valid_file;
533 die("stat(%s): %s", name, strerror(errno));
535 if (S_ISLNK(st.st_mode)) {
536 int ret;
537 char buf[PATH_MAX + 1]; /* ought to be SYMLINK_MAX */
538 if (sizeof(buf) <= st.st_size)
539 die("symlink too long: %s", name);
540 ret = readlink(name, buf, st.st_size);
541 if (ret < 0)
542 die("readlink(%s)", name);
543 prep_temp_blob(temp, buf, st.st_size,
544 (one->sha1_valid ?
545 one->sha1 : null_sha1),
546 (one->sha1_valid ?
547 one->mode : S_IFLNK));
549 else {
550 /* we can borrow from the file in the work tree */
551 temp->name = name;
552 if (!one->sha1_valid)
553 strcpy(temp->hex, sha1_to_hex(null_sha1));
554 else
555 strcpy(temp->hex, sha1_to_hex(one->sha1));
556 /* Even though we may sometimes borrow the
557 * contents from the work tree, we always want
558 * one->mode. mode is trustworthy even when
559 * !(one->sha1_valid), as long as
560 * DIFF_FILE_VALID(one).
562 sprintf(temp->mode, "%06o", one->mode);
564 return;
566 else {
567 if (diff_populate_filespec(one, 0))
568 die("cannot read data blob for %s", one->path);
569 prep_temp_blob(temp, one->data, one->size,
570 one->sha1, one->mode);
574 static void remove_tempfile(void)
576 int i;
578 for (i = 0; i < 2; i++)
579 if (diff_temp[i].name == diff_temp[i].tmp_path) {
580 unlink(diff_temp[i].name);
581 diff_temp[i].name = NULL;
585 static void remove_tempfile_on_signal(int signo)
587 remove_tempfile();
588 signal(SIGINT, SIG_DFL);
589 raise(signo);
592 static int spawn_prog(const char *pgm, const char **arg)
594 pid_t pid;
595 int status;
597 fflush(NULL);
598 pid = fork();
599 if (pid < 0)
600 die("unable to fork");
601 if (!pid) {
602 execvp(pgm, (char *const*) arg);
603 exit(255);
606 while (waitpid(pid, &status, 0) < 0) {
607 if (errno == EINTR)
608 continue;
609 return -1;
612 /* Earlier we did not check the exit status because
613 * diff exits non-zero if files are different, and
614 * we are not interested in knowing that. It was a
615 * mistake which made it harder to quit a diff-*
616 * session that uses the git-apply-patch-script as
617 * the GIT_EXTERNAL_DIFF. A custom GIT_EXTERNAL_DIFF
618 * should also exit non-zero only when it wants to
619 * abort the entire diff-* session.
621 if (WIFEXITED(status) && !WEXITSTATUS(status))
622 return 0;
623 return -1;
626 /* An external diff command takes:
628 * diff-cmd name infile1 infile1-sha1 infile1-mode \
629 * infile2 infile2-sha1 infile2-mode [ rename-to ]
632 static void run_external_diff(const char *pgm,
633 const char *name,
634 const char *other,
635 struct diff_filespec *one,
636 struct diff_filespec *two,
637 const char *xfrm_msg,
638 int complete_rewrite)
640 const char *spawn_arg[10];
641 struct diff_tempfile *temp = diff_temp;
642 int retval;
643 static int atexit_asked = 0;
644 const char *othername;
645 const char **arg = &spawn_arg[0];
647 othername = (other? other : name);
648 if (one && two) {
649 prepare_temp_file(name, &temp[0], one);
650 prepare_temp_file(othername, &temp[1], two);
651 if (! atexit_asked &&
652 (temp[0].name == temp[0].tmp_path ||
653 temp[1].name == temp[1].tmp_path)) {
654 atexit_asked = 1;
655 atexit(remove_tempfile);
657 signal(SIGINT, remove_tempfile_on_signal);
660 if (one && two) {
661 *arg++ = pgm;
662 *arg++ = name;
663 *arg++ = temp[0].name;
664 *arg++ = temp[0].hex;
665 *arg++ = temp[0].mode;
666 *arg++ = temp[1].name;
667 *arg++ = temp[1].hex;
668 *arg++ = temp[1].mode;
669 if (other) {
670 *arg++ = other;
671 *arg++ = xfrm_msg;
673 } else {
674 *arg++ = pgm;
675 *arg++ = name;
677 *arg = NULL;
678 retval = spawn_prog(pgm, spawn_arg);
679 remove_tempfile();
680 if (retval) {
681 fprintf(stderr, "external diff died, stopping at %s.\n", name);
682 exit(1);
686 static void run_diff_cmd(const char *pgm,
687 const char *name,
688 const char *other,
689 struct diff_filespec *one,
690 struct diff_filespec *two,
691 const char *xfrm_msg,
692 int complete_rewrite)
694 if (pgm) {
695 run_external_diff(pgm, name, other, one, two, xfrm_msg,
696 complete_rewrite);
697 return;
699 if (one && two)
700 builtin_diff(name, other ? other : name,
701 one, two, xfrm_msg, complete_rewrite);
702 else
703 printf("* Unmerged path %s\n", name);
706 static void diff_fill_sha1_info(struct diff_filespec *one)
708 if (DIFF_FILE_VALID(one)) {
709 if (!one->sha1_valid) {
710 struct stat st;
711 if (lstat(one->path, &st) < 0)
712 die("stat %s", one->path);
713 if (index_path(one->sha1, one->path, &st, 0))
714 die("cannot hash %s\n", one->path);
717 else
718 memset(one->sha1, 0, 20);
721 static void run_diff(struct diff_filepair *p, struct diff_options *o)
723 const char *pgm = external_diff();
724 char msg[PATH_MAX*2+300], *xfrm_msg;
725 struct diff_filespec *one;
726 struct diff_filespec *two;
727 const char *name;
728 const char *other;
729 char *name_munged, *other_munged;
730 int complete_rewrite = 0;
731 int len;
733 if (DIFF_PAIR_UNMERGED(p)) {
734 /* unmerged */
735 run_diff_cmd(pgm, p->one->path, NULL, NULL, NULL, NULL, 0);
736 return;
739 name = p->one->path;
740 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
741 name_munged = quote_one(name);
742 other_munged = quote_one(other);
743 one = p->one; two = p->two;
745 diff_fill_sha1_info(one);
746 diff_fill_sha1_info(two);
748 len = 0;
749 switch (p->status) {
750 case DIFF_STATUS_COPIED:
751 len += snprintf(msg + len, sizeof(msg) - len,
752 "similarity index %d%%\n"
753 "copy from %s\n"
754 "copy to %s\n",
755 (int)(0.5 + p->score * 100.0/MAX_SCORE),
756 name_munged, other_munged);
757 break;
758 case DIFF_STATUS_RENAMED:
759 len += snprintf(msg + len, sizeof(msg) - len,
760 "similarity index %d%%\n"
761 "rename from %s\n"
762 "rename to %s\n",
763 (int)(0.5 + p->score * 100.0/MAX_SCORE),
764 name_munged, other_munged);
765 break;
766 case DIFF_STATUS_MODIFIED:
767 if (p->score) {
768 len += snprintf(msg + len, sizeof(msg) - len,
769 "dissimilarity index %d%%\n",
770 (int)(0.5 + p->score *
771 100.0/MAX_SCORE));
772 complete_rewrite = 1;
773 break;
775 /* fallthru */
776 default:
777 /* nothing */
781 if (memcmp(one->sha1, two->sha1, 20)) {
782 char one_sha1[41];
783 int abbrev = o->full_index ? 40 : DEFAULT_ABBREV;
784 memcpy(one_sha1, sha1_to_hex(one->sha1), 41);
786 len += snprintf(msg + len, sizeof(msg) - len,
787 "index %.*s..%.*s",
788 abbrev, one_sha1, abbrev,
789 sha1_to_hex(two->sha1));
790 if (one->mode == two->mode)
791 len += snprintf(msg + len, sizeof(msg) - len,
792 " %06o", one->mode);
793 len += snprintf(msg + len, sizeof(msg) - len, "\n");
796 if (len)
797 msg[--len] = 0;
798 xfrm_msg = len ? msg : NULL;
800 if (!pgm &&
801 DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
802 (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
803 /* a filepair that changes between file and symlink
804 * needs to be split into deletion and creation.
806 struct diff_filespec *null = alloc_filespec(two->path);
807 run_diff_cmd(NULL, name, other, one, null, xfrm_msg, 0);
808 free(null);
809 null = alloc_filespec(one->path);
810 run_diff_cmd(NULL, name, other, null, two, xfrm_msg, 0);
811 free(null);
813 else
814 run_diff_cmd(pgm, name, other, one, two, xfrm_msg,
815 complete_rewrite);
817 free(name_munged);
818 free(other_munged);
821 void diff_setup(struct diff_options *options)
823 memset(options, 0, sizeof(*options));
824 options->output_format = DIFF_FORMAT_RAW;
825 options->line_termination = '\n';
826 options->break_opt = -1;
827 options->rename_limit = -1;
829 options->change = diff_change;
830 options->add_remove = diff_addremove;
833 int diff_setup_done(struct diff_options *options)
835 if ((options->find_copies_harder &&
836 options->detect_rename != DIFF_DETECT_COPY) ||
837 (0 <= options->rename_limit && !options->detect_rename))
838 return -1;
839 if (options->detect_rename && options->rename_limit < 0)
840 options->rename_limit = diff_rename_limit_default;
841 if (options->setup & DIFF_SETUP_USE_CACHE) {
842 if (!active_cache)
843 /* read-cache does not die even when it fails
844 * so it is safe for us to do this here. Also
845 * it does not smudge active_cache or active_nr
846 * when it fails, so we do not have to worry about
847 * cleaning it up ourselves either.
849 read_cache();
851 if (options->setup & DIFF_SETUP_USE_SIZE_CACHE)
852 use_size_cache = 1;
853 if (options->abbrev <= 0 || 40 < options->abbrev)
854 options->abbrev = 40; /* full */
856 return 0;
859 int diff_opt_parse(struct diff_options *options, const char **av, int ac)
861 const char *arg = av[0];
862 if (!strcmp(arg, "-p") || !strcmp(arg, "-u"))
863 options->output_format = DIFF_FORMAT_PATCH;
864 else if (!strcmp(arg, "-z"))
865 options->line_termination = 0;
866 else if (!strncmp(arg, "-l", 2))
867 options->rename_limit = strtoul(arg+2, NULL, 10);
868 else if (!strcmp(arg, "--full-index"))
869 options->full_index = 1;
870 else if (!strcmp(arg, "--name-only"))
871 options->output_format = DIFF_FORMAT_NAME;
872 else if (!strcmp(arg, "--name-status"))
873 options->output_format = DIFF_FORMAT_NAME_STATUS;
874 else if (!strcmp(arg, "-R"))
875 options->reverse_diff = 1;
876 else if (!strncmp(arg, "-S", 2))
877 options->pickaxe = arg + 2;
878 else if (!strcmp(arg, "-s"))
879 options->output_format = DIFF_FORMAT_NO_OUTPUT;
880 else if (!strncmp(arg, "-O", 2))
881 options->orderfile = arg + 2;
882 else if (!strncmp(arg, "--diff-filter=", 14))
883 options->filter = arg + 14;
884 else if (!strcmp(arg, "--pickaxe-all"))
885 options->pickaxe_opts = DIFF_PICKAXE_ALL;
886 else if (!strncmp(arg, "-B", 2)) {
887 if ((options->break_opt =
888 diff_scoreopt_parse(arg)) == -1)
889 return -1;
891 else if (!strncmp(arg, "-M", 2)) {
892 if ((options->rename_score =
893 diff_scoreopt_parse(arg)) == -1)
894 return -1;
895 options->detect_rename = DIFF_DETECT_RENAME;
897 else if (!strncmp(arg, "-C", 2)) {
898 if ((options->rename_score =
899 diff_scoreopt_parse(arg)) == -1)
900 return -1;
901 options->detect_rename = DIFF_DETECT_COPY;
903 else if (!strcmp(arg, "--find-copies-harder"))
904 options->find_copies_harder = 1;
905 else if (!strcmp(arg, "--abbrev"))
906 options->abbrev = DEFAULT_ABBREV;
907 else if (!strncmp(arg, "--abbrev=", 9)) {
908 options->abbrev = strtoul(arg + 9, NULL, 10);
909 if (options->abbrev < MINIMUM_ABBREV)
910 options->abbrev = MINIMUM_ABBREV;
911 else if (40 < options->abbrev)
912 options->abbrev = 40;
914 else
915 return 0;
916 return 1;
919 static int parse_num(const char **cp_p)
921 unsigned long num, scale;
922 int ch, dot;
923 const char *cp = *cp_p;
925 num = 0;
926 scale = 1;
927 dot = 0;
928 for(;;) {
929 ch = *cp;
930 if ( !dot && ch == '.' ) {
931 scale = 1;
932 dot = 1;
933 } else if ( ch == '%' ) {
934 scale = dot ? scale*100 : 100;
935 cp++; /* % is always at the end */
936 break;
937 } else if ( ch >= '0' && ch <= '9' ) {
938 if ( scale < 100000 ) {
939 scale *= 10;
940 num = (num*10) + (ch-'0');
942 } else {
943 break;
945 cp++;
947 *cp_p = cp;
949 /* user says num divided by scale and we say internally that
950 * is MAX_SCORE * num / scale.
952 return (num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale);
955 int diff_scoreopt_parse(const char *opt)
957 int opt1, opt2, cmd;
959 if (*opt++ != '-')
960 return -1;
961 cmd = *opt++;
962 if (cmd != 'M' && cmd != 'C' && cmd != 'B')
963 return -1; /* that is not a -M, -C nor -B option */
965 opt1 = parse_num(&opt);
966 if (cmd != 'B')
967 opt2 = 0;
968 else {
969 if (*opt == 0)
970 opt2 = 0;
971 else if (*opt != '/')
972 return -1; /* we expect -B80/99 or -B80 */
973 else {
974 opt++;
975 opt2 = parse_num(&opt);
978 if (*opt != 0)
979 return -1;
980 return opt1 | (opt2 << 16);
983 struct diff_queue_struct diff_queued_diff;
985 void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
987 if (queue->alloc <= queue->nr) {
988 queue->alloc = alloc_nr(queue->alloc);
989 queue->queue = xrealloc(queue->queue,
990 sizeof(dp) * queue->alloc);
992 queue->queue[queue->nr++] = dp;
995 struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
996 struct diff_filespec *one,
997 struct diff_filespec *two)
999 struct diff_filepair *dp = xmalloc(sizeof(*dp));
1000 dp->one = one;
1001 dp->two = two;
1002 dp->score = 0;
1003 dp->status = 0;
1004 dp->source_stays = 0;
1005 dp->broken_pair = 0;
1006 if (queue)
1007 diff_q(queue, dp);
1008 return dp;
1011 void diff_free_filepair(struct diff_filepair *p)
1013 diff_free_filespec_data(p->one);
1014 diff_free_filespec_data(p->two);
1015 free(p->one);
1016 free(p->two);
1017 free(p);
1020 /* This is different from find_unique_abbrev() in that
1021 * it stuffs the result with dots for alignment.
1023 const char *diff_unique_abbrev(const unsigned char *sha1, int len)
1025 int abblen;
1026 const char *abbrev;
1027 if (len == 40)
1028 return sha1_to_hex(sha1);
1030 abbrev = find_unique_abbrev(sha1, len);
1031 if (!abbrev)
1032 return sha1_to_hex(sha1);
1033 abblen = strlen(abbrev);
1034 if (abblen < 37) {
1035 static char hex[41];
1036 if (len < abblen && abblen <= len + 2)
1037 sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
1038 else
1039 sprintf(hex, "%s...", abbrev);
1040 return hex;
1042 return sha1_to_hex(sha1);
1045 static void diff_flush_raw(struct diff_filepair *p,
1046 int line_termination,
1047 int inter_name_termination,
1048 struct diff_options *options)
1050 int two_paths;
1051 char status[10];
1052 int abbrev = options->abbrev;
1053 const char *path_one, *path_two;
1054 int output_format = options->output_format;
1056 path_one = p->one->path;
1057 path_two = p->two->path;
1058 if (line_termination) {
1059 path_one = quote_one(path_one);
1060 path_two = quote_one(path_two);
1063 if (p->score)
1064 sprintf(status, "%c%03d", p->status,
1065 (int)(0.5 + p->score * 100.0/MAX_SCORE));
1066 else {
1067 status[0] = p->status;
1068 status[1] = 0;
1070 switch (p->status) {
1071 case DIFF_STATUS_COPIED:
1072 case DIFF_STATUS_RENAMED:
1073 two_paths = 1;
1074 break;
1075 case DIFF_STATUS_ADDED:
1076 case DIFF_STATUS_DELETED:
1077 two_paths = 0;
1078 break;
1079 default:
1080 two_paths = 0;
1081 break;
1083 if (output_format != DIFF_FORMAT_NAME_STATUS) {
1084 printf(":%06o %06o %s ",
1085 p->one->mode, p->two->mode,
1086 diff_unique_abbrev(p->one->sha1, abbrev));
1087 printf("%s ",
1088 diff_unique_abbrev(p->two->sha1, abbrev));
1090 printf("%s%c%s", status, inter_name_termination, path_one);
1091 if (two_paths)
1092 printf("%c%s", inter_name_termination, path_two);
1093 putchar(line_termination);
1094 if (path_one != p->one->path)
1095 free((void*)path_one);
1096 if (path_two != p->two->path)
1097 free((void*)path_two);
1100 static void diff_flush_name(struct diff_filepair *p,
1101 int inter_name_termination,
1102 int line_termination)
1104 char *path = p->two->path;
1106 if (line_termination)
1107 path = quote_one(p->two->path);
1108 else
1109 path = p->two->path;
1110 printf("%s%c", path, line_termination);
1111 if (p->two->path != path)
1112 free(path);
1115 int diff_unmodified_pair(struct diff_filepair *p)
1117 /* This function is written stricter than necessary to support
1118 * the currently implemented transformers, but the idea is to
1119 * let transformers to produce diff_filepairs any way they want,
1120 * and filter and clean them up here before producing the output.
1122 struct diff_filespec *one, *two;
1124 if (DIFF_PAIR_UNMERGED(p))
1125 return 0; /* unmerged is interesting */
1127 one = p->one;
1128 two = p->two;
1130 /* deletion, addition, mode or type change
1131 * and rename are all interesting.
1133 if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
1134 DIFF_PAIR_MODE_CHANGED(p) ||
1135 strcmp(one->path, two->path))
1136 return 0;
1138 /* both are valid and point at the same path. that is, we are
1139 * dealing with a change.
1141 if (one->sha1_valid && two->sha1_valid &&
1142 !memcmp(one->sha1, two->sha1, sizeof(one->sha1)))
1143 return 1; /* no change */
1144 if (!one->sha1_valid && !two->sha1_valid)
1145 return 1; /* both look at the same file on the filesystem. */
1146 return 0;
1149 static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
1151 if (diff_unmodified_pair(p))
1152 return;
1154 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
1155 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
1156 return; /* no tree diffs in patch format */
1158 run_diff(p, o);
1161 int diff_queue_is_empty(void)
1163 struct diff_queue_struct *q = &diff_queued_diff;
1164 int i;
1165 for (i = 0; i < q->nr; i++)
1166 if (!diff_unmodified_pair(q->queue[i]))
1167 return 0;
1168 return 1;
1171 #if DIFF_DEBUG
1172 void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
1174 fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
1175 x, one ? one : "",
1176 s->path,
1177 DIFF_FILE_VALID(s) ? "valid" : "invalid",
1178 s->mode,
1179 s->sha1_valid ? sha1_to_hex(s->sha1) : "");
1180 fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
1181 x, one ? one : "",
1182 s->size, s->xfrm_flags);
1185 void diff_debug_filepair(const struct diff_filepair *p, int i)
1187 diff_debug_filespec(p->one, i, "one");
1188 diff_debug_filespec(p->two, i, "two");
1189 fprintf(stderr, "score %d, status %c stays %d broken %d\n",
1190 p->score, p->status ? p->status : '?',
1191 p->source_stays, p->broken_pair);
1194 void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
1196 int i;
1197 if (msg)
1198 fprintf(stderr, "%s\n", msg);
1199 fprintf(stderr, "q->nr = %d\n", q->nr);
1200 for (i = 0; i < q->nr; i++) {
1201 struct diff_filepair *p = q->queue[i];
1202 diff_debug_filepair(p, i);
1205 #endif
1207 static void diff_resolve_rename_copy(void)
1209 int i, j;
1210 struct diff_filepair *p, *pp;
1211 struct diff_queue_struct *q = &diff_queued_diff;
1213 diff_debug_queue("resolve-rename-copy", q);
1215 for (i = 0; i < q->nr; i++) {
1216 p = q->queue[i];
1217 p->status = 0; /* undecided */
1218 if (DIFF_PAIR_UNMERGED(p))
1219 p->status = DIFF_STATUS_UNMERGED;
1220 else if (!DIFF_FILE_VALID(p->one))
1221 p->status = DIFF_STATUS_ADDED;
1222 else if (!DIFF_FILE_VALID(p->two))
1223 p->status = DIFF_STATUS_DELETED;
1224 else if (DIFF_PAIR_TYPE_CHANGED(p))
1225 p->status = DIFF_STATUS_TYPE_CHANGED;
1227 /* from this point on, we are dealing with a pair
1228 * whose both sides are valid and of the same type, i.e.
1229 * either in-place edit or rename/copy edit.
1231 else if (DIFF_PAIR_RENAME(p)) {
1232 if (p->source_stays) {
1233 p->status = DIFF_STATUS_COPIED;
1234 continue;
1236 /* See if there is some other filepair that
1237 * copies from the same source as us. If so
1238 * we are a copy. Otherwise we are either a
1239 * copy if the path stays, or a rename if it
1240 * does not, but we already handled "stays" case.
1242 for (j = i + 1; j < q->nr; j++) {
1243 pp = q->queue[j];
1244 if (strcmp(pp->one->path, p->one->path))
1245 continue; /* not us */
1246 if (!DIFF_PAIR_RENAME(pp))
1247 continue; /* not a rename/copy */
1248 /* pp is a rename/copy from the same source */
1249 p->status = DIFF_STATUS_COPIED;
1250 break;
1252 if (!p->status)
1253 p->status = DIFF_STATUS_RENAMED;
1255 else if (memcmp(p->one->sha1, p->two->sha1, 20) ||
1256 p->one->mode != p->two->mode)
1257 p->status = DIFF_STATUS_MODIFIED;
1258 else {
1259 /* This is a "no-change" entry and should not
1260 * happen anymore, but prepare for broken callers.
1262 error("feeding unmodified %s to diffcore",
1263 p->one->path);
1264 p->status = DIFF_STATUS_UNKNOWN;
1267 diff_debug_queue("resolve-rename-copy done", q);
1270 void diff_flush(struct diff_options *options)
1272 struct diff_queue_struct *q = &diff_queued_diff;
1273 int i;
1274 int inter_name_termination = '\t';
1275 int diff_output_format = options->output_format;
1276 int line_termination = options->line_termination;
1278 if (!line_termination)
1279 inter_name_termination = 0;
1281 for (i = 0; i < q->nr; i++) {
1282 struct diff_filepair *p = q->queue[i];
1283 if ((diff_output_format == DIFF_FORMAT_NO_OUTPUT) ||
1284 (p->status == DIFF_STATUS_UNKNOWN))
1285 continue;
1286 if (p->status == 0)
1287 die("internal error in diff-resolve-rename-copy");
1288 switch (diff_output_format) {
1289 case DIFF_FORMAT_PATCH:
1290 diff_flush_patch(p, options);
1291 break;
1292 case DIFF_FORMAT_RAW:
1293 case DIFF_FORMAT_NAME_STATUS:
1294 diff_flush_raw(p, line_termination,
1295 inter_name_termination,
1296 options);
1297 break;
1298 case DIFF_FORMAT_NAME:
1299 diff_flush_name(p,
1300 inter_name_termination,
1301 line_termination);
1302 break;
1304 diff_free_filepair(q->queue[i]);
1306 free(q->queue);
1307 q->queue = NULL;
1308 q->nr = q->alloc = 0;
1311 static void diffcore_apply_filter(const char *filter)
1313 int i;
1314 struct diff_queue_struct *q = &diff_queued_diff;
1315 struct diff_queue_struct outq;
1316 outq.queue = NULL;
1317 outq.nr = outq.alloc = 0;
1319 if (!filter)
1320 return;
1322 if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
1323 int found;
1324 for (i = found = 0; !found && i < q->nr; i++) {
1325 struct diff_filepair *p = q->queue[i];
1326 if (((p->status == DIFF_STATUS_MODIFIED) &&
1327 ((p->score &&
1328 strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
1329 (!p->score &&
1330 strchr(filter, DIFF_STATUS_MODIFIED)))) ||
1331 ((p->status != DIFF_STATUS_MODIFIED) &&
1332 strchr(filter, p->status)))
1333 found++;
1335 if (found)
1336 return;
1338 /* otherwise we will clear the whole queue
1339 * by copying the empty outq at the end of this
1340 * function, but first clear the current entries
1341 * in the queue.
1343 for (i = 0; i < q->nr; i++)
1344 diff_free_filepair(q->queue[i]);
1346 else {
1347 /* Only the matching ones */
1348 for (i = 0; i < q->nr; i++) {
1349 struct diff_filepair *p = q->queue[i];
1351 if (((p->status == DIFF_STATUS_MODIFIED) &&
1352 ((p->score &&
1353 strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
1354 (!p->score &&
1355 strchr(filter, DIFF_STATUS_MODIFIED)))) ||
1356 ((p->status != DIFF_STATUS_MODIFIED) &&
1357 strchr(filter, p->status)))
1358 diff_q(&outq, p);
1359 else
1360 diff_free_filepair(p);
1363 free(q->queue);
1364 *q = outq;
1367 void diffcore_std(struct diff_options *options)
1369 if (options->paths && options->paths[0])
1370 diffcore_pathspec(options->paths);
1371 if (options->break_opt != -1)
1372 diffcore_break(options->break_opt);
1373 if (options->detect_rename)
1374 diffcore_rename(options);
1375 if (options->break_opt != -1)
1376 diffcore_merge_broken();
1377 if (options->pickaxe)
1378 diffcore_pickaxe(options->pickaxe, options->pickaxe_opts);
1379 if (options->orderfile)
1380 diffcore_order(options->orderfile);
1381 diff_resolve_rename_copy();
1382 diffcore_apply_filter(options->filter);
1386 void diffcore_std_no_resolve(struct diff_options *options)
1388 if (options->pickaxe)
1389 diffcore_pickaxe(options->pickaxe, options->pickaxe_opts);
1390 if (options->orderfile)
1391 diffcore_order(options->orderfile);
1392 diffcore_apply_filter(options->filter);
1395 void diff_addremove(struct diff_options *options,
1396 int addremove, unsigned mode,
1397 const unsigned char *sha1,
1398 const char *base, const char *path)
1400 char concatpath[PATH_MAX];
1401 struct diff_filespec *one, *two;
1403 /* This may look odd, but it is a preparation for
1404 * feeding "there are unchanged files which should
1405 * not produce diffs, but when you are doing copy
1406 * detection you would need them, so here they are"
1407 * entries to the diff-core. They will be prefixed
1408 * with something like '=' or '*' (I haven't decided
1409 * which but should not make any difference).
1410 * Feeding the same new and old to diff_change()
1411 * also has the same effect.
1412 * Before the final output happens, they are pruned after
1413 * merged into rename/copy pairs as appropriate.
1415 if (options->reverse_diff)
1416 addremove = (addremove == '+' ? '-' :
1417 addremove == '-' ? '+' : addremove);
1419 if (!path) path = "";
1420 sprintf(concatpath, "%s%s", base, path);
1421 one = alloc_filespec(concatpath);
1422 two = alloc_filespec(concatpath);
1424 if (addremove != '+')
1425 fill_filespec(one, sha1, mode);
1426 if (addremove != '-')
1427 fill_filespec(two, sha1, mode);
1429 diff_queue(&diff_queued_diff, one, two);
1432 void diff_change(struct diff_options *options,
1433 unsigned old_mode, unsigned new_mode,
1434 const unsigned char *old_sha1,
1435 const unsigned char *new_sha1,
1436 const char *base, const char *path)
1438 char concatpath[PATH_MAX];
1439 struct diff_filespec *one, *two;
1441 if (options->reverse_diff) {
1442 unsigned tmp;
1443 const unsigned char *tmp_c;
1444 tmp = old_mode; old_mode = new_mode; new_mode = tmp;
1445 tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
1447 if (!path) path = "";
1448 sprintf(concatpath, "%s%s", base, path);
1449 one = alloc_filespec(concatpath);
1450 two = alloc_filespec(concatpath);
1451 fill_filespec(one, old_sha1, old_mode);
1452 fill_filespec(two, new_sha1, new_mode);
1454 diff_queue(&diff_queued_diff, one, two);
1457 void diff_unmerge(struct diff_options *options,
1458 const char *path)
1460 struct diff_filespec *one, *two;
1461 one = alloc_filespec(path);
1462 two = alloc_filespec(path);
1463 diff_queue(&diff_queued_diff, one, two);