Teach diff about -b and -w flags
[git.git] / diff.c
blob585be7e50da4a2222b6fe873729f6283e9a77b0b
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 "delta.h"
12 #include "xdiff-interface.h"
14 static int use_size_cache;
16 int diff_rename_limit_default = -1;
18 int git_diff_config(const char *var, const char *value)
20 if (!strcmp(var, "diff.renamelimit")) {
21 diff_rename_limit_default = git_config_int(var, value);
22 return 0;
25 return git_default_config(var, value);
28 enum color_diff {
29 DIFF_RESET = 0,
30 DIFF_PLAIN = 1,
31 DIFF_METAINFO = 2,
32 DIFF_FRAGINFO = 3,
33 DIFF_FILE_OLD = 4,
34 DIFF_FILE_NEW = 5,
37 #define COLOR_NORMAL ""
38 #define COLOR_BOLD "\033[1m"
39 #define COLOR_DIM "\033[2m"
40 #define COLOR_UL "\033[4m"
41 #define COLOR_BLINK "\033[5m"
42 #define COLOR_REVERSE "\033[7m"
43 #define COLOR_RESET "\033[m"
45 #define COLOR_BLACK "\033[30m"
46 #define COLOR_RED "\033[31m"
47 #define COLOR_GREEN "\033[32m"
48 #define COLOR_YELLOW "\033[33m"
49 #define COLOR_BLUE "\033[34m"
50 #define COLOR_MAGENTA "\033[35m"
51 #define COLOR_CYAN "\033[36m"
52 #define COLOR_WHITE "\033[37m"
54 #define COLOR_CYANBG "\033[46m"
55 #define COLOR_GRAYBG "\033[47m" // Good for xterm
57 static const char *diff_colors[] = {
58 [DIFF_RESET] = COLOR_RESET,
59 [DIFF_PLAIN] = COLOR_NORMAL,
60 [DIFF_METAINFO] = COLOR_BOLD,
61 [DIFF_FRAGINFO] = COLOR_CYAN,
62 [DIFF_FILE_OLD] = COLOR_RED,
63 [DIFF_FILE_NEW] = COLOR_GREEN,
66 static char *quote_one(const char *str)
68 int needlen;
69 char *xp;
71 if (!str)
72 return NULL;
73 needlen = quote_c_style(str, NULL, NULL, 0);
74 if (!needlen)
75 return strdup(str);
76 xp = xmalloc(needlen + 1);
77 quote_c_style(str, xp, NULL, 0);
78 return xp;
81 static char *quote_two(const char *one, const char *two)
83 int need_one = quote_c_style(one, NULL, NULL, 1);
84 int need_two = quote_c_style(two, NULL, NULL, 1);
85 char *xp;
87 if (need_one + need_two) {
88 if (!need_one) need_one = strlen(one);
89 if (!need_two) need_one = strlen(two);
91 xp = xmalloc(need_one + need_two + 3);
92 xp[0] = '"';
93 quote_c_style(one, xp + 1, NULL, 1);
94 quote_c_style(two, xp + need_one + 1, NULL, 1);
95 strcpy(xp + need_one + need_two + 1, "\"");
96 return xp;
98 need_one = strlen(one);
99 need_two = strlen(two);
100 xp = xmalloc(need_one + need_two + 1);
101 strcpy(xp, one);
102 strcpy(xp + need_one, two);
103 return xp;
106 static const char *external_diff(void)
108 static const char *external_diff_cmd = NULL;
109 static int done_preparing = 0;
111 if (done_preparing)
112 return external_diff_cmd;
113 external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
114 done_preparing = 1;
115 return external_diff_cmd;
118 #define TEMPFILE_PATH_LEN 50
120 static struct diff_tempfile {
121 const char *name; /* filename external diff should read from */
122 char hex[41];
123 char mode[10];
124 char tmp_path[TEMPFILE_PATH_LEN];
125 } diff_temp[2];
127 static int count_lines(const char *data, int size)
129 int count, ch, completely_empty = 1, nl_just_seen = 0;
130 count = 0;
131 while (0 < size--) {
132 ch = *data++;
133 if (ch == '\n') {
134 count++;
135 nl_just_seen = 1;
136 completely_empty = 0;
138 else {
139 nl_just_seen = 0;
140 completely_empty = 0;
143 if (completely_empty)
144 return 0;
145 if (!nl_just_seen)
146 count++; /* no trailing newline */
147 return count;
150 static void print_line_count(int count)
152 switch (count) {
153 case 0:
154 printf("0,0");
155 break;
156 case 1:
157 printf("1");
158 break;
159 default:
160 printf("1,%d", count);
161 break;
165 static void copy_file(int prefix, const char *data, int size)
167 int ch, nl_just_seen = 1;
168 while (0 < size--) {
169 ch = *data++;
170 if (nl_just_seen)
171 putchar(prefix);
172 putchar(ch);
173 if (ch == '\n')
174 nl_just_seen = 1;
175 else
176 nl_just_seen = 0;
178 if (!nl_just_seen)
179 printf("\n\\ No newline at end of file\n");
182 static void emit_rewrite_diff(const char *name_a,
183 const char *name_b,
184 struct diff_filespec *one,
185 struct diff_filespec *two)
187 int lc_a, lc_b;
188 diff_populate_filespec(one, 0);
189 diff_populate_filespec(two, 0);
190 lc_a = count_lines(one->data, one->size);
191 lc_b = count_lines(two->data, two->size);
192 printf("--- %s\n+++ %s\n@@ -", name_a, name_b);
193 print_line_count(lc_a);
194 printf(" +");
195 print_line_count(lc_b);
196 printf(" @@\n");
197 if (lc_a)
198 copy_file('-', one->data, one->size);
199 if (lc_b)
200 copy_file('+', two->data, two->size);
203 static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
205 if (!DIFF_FILE_VALID(one)) {
206 mf->ptr = ""; /* does not matter */
207 mf->size = 0;
208 return 0;
210 else if (diff_populate_filespec(one, 0))
211 return -1;
212 mf->ptr = one->data;
213 mf->size = one->size;
214 return 0;
217 struct emit_callback {
218 struct xdiff_emit_state xm;
219 int nparents, color_diff;
220 const char **label_path;
223 static inline const char *get_color(int diff_use_color, enum color_diff ix)
225 if (diff_use_color)
226 return diff_colors[ix];
227 return "";
230 static void fn_out_consume(void *priv, char *line, unsigned long len)
232 int i;
233 struct emit_callback *ecbdata = priv;
234 const char *set = get_color(ecbdata->color_diff, DIFF_METAINFO);
235 const char *reset = get_color(ecbdata->color_diff, DIFF_RESET);
237 if (ecbdata->label_path[0]) {
238 printf("%s--- %s%s\n", set, ecbdata->label_path[0], reset);
239 printf("%s+++ %s%s\n", set, ecbdata->label_path[1], reset);
240 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
243 /* This is not really necessary for now because
244 * this codepath only deals with two-way diffs.
246 for (i = 0; i < len && line[i] == '@'; i++)
248 if (2 <= i && i < len && line[i] == ' ') {
249 ecbdata->nparents = i - 1;
250 set = get_color(ecbdata->color_diff, DIFF_FRAGINFO);
252 else if (len < ecbdata->nparents)
253 set = reset;
254 else {
255 int nparents = ecbdata->nparents;
256 int color = DIFF_PLAIN;
257 for (i = 0; i < nparents && len; i++) {
258 if (line[i] == '-')
259 color = DIFF_FILE_OLD;
260 else if (line[i] == '+')
261 color = DIFF_FILE_NEW;
263 set = get_color(ecbdata->color_diff, color);
265 if (len > 0 && line[len-1] == '\n')
266 len--;
267 printf("%s%.*s%s\n", set, (int) len, line, reset);
270 static char *pprint_rename(const char *a, const char *b)
272 const char *old = a;
273 const char *new = b;
274 char *name = NULL;
275 int pfx_length, sfx_length;
276 int len_a = strlen(a);
277 int len_b = strlen(b);
279 /* Find common prefix */
280 pfx_length = 0;
281 while (*old && *new && *old == *new) {
282 if (*old == '/')
283 pfx_length = old - a + 1;
284 old++;
285 new++;
288 /* Find common suffix */
289 old = a + len_a;
290 new = b + len_b;
291 sfx_length = 0;
292 while (a <= old && b <= new && *old == *new) {
293 if (*old == '/')
294 sfx_length = len_a - (old - a);
295 old--;
296 new--;
300 * pfx{mid-a => mid-b}sfx
301 * {pfx-a => pfx-b}sfx
302 * pfx{sfx-a => sfx-b}
303 * name-a => name-b
305 if (pfx_length + sfx_length) {
306 int a_midlen = len_a - pfx_length - sfx_length;
307 int b_midlen = len_b - pfx_length - sfx_length;
308 if (a_midlen < 0) a_midlen = 0;
309 if (b_midlen < 0) b_midlen = 0;
311 name = xmalloc(pfx_length + a_midlen + b_midlen + sfx_length + 7);
312 sprintf(name, "%.*s{%.*s => %.*s}%s",
313 pfx_length, a,
314 a_midlen, a + pfx_length,
315 b_midlen, b + pfx_length,
316 a + len_a - sfx_length);
318 else {
319 name = xmalloc(len_a + len_b + 5);
320 sprintf(name, "%s => %s", a, b);
322 return name;
325 struct diffstat_t {
326 struct xdiff_emit_state xm;
328 int nr;
329 int alloc;
330 struct diffstat_file {
331 char *name;
332 unsigned is_unmerged:1;
333 unsigned is_binary:1;
334 unsigned is_renamed:1;
335 unsigned int added, deleted;
336 } **files;
339 static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
340 const char *name_a,
341 const char *name_b)
343 struct diffstat_file *x;
344 x = xcalloc(sizeof (*x), 1);
345 if (diffstat->nr == diffstat->alloc) {
346 diffstat->alloc = alloc_nr(diffstat->alloc);
347 diffstat->files = xrealloc(diffstat->files,
348 diffstat->alloc * sizeof(x));
350 diffstat->files[diffstat->nr++] = x;
351 if (name_b) {
352 x->name = pprint_rename(name_a, name_b);
353 x->is_renamed = 1;
355 else
356 x->name = strdup(name_a);
357 return x;
360 static void diffstat_consume(void *priv, char *line, unsigned long len)
362 struct diffstat_t *diffstat = priv;
363 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
365 if (line[0] == '+')
366 x->added++;
367 else if (line[0] == '-')
368 x->deleted++;
371 static const char pluses[] = "++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++";
372 static const char minuses[]= "----------------------------------------------------------------------";
373 const char mime_boundary_leader[] = "------------";
375 static void show_stats(struct diffstat_t* data)
377 int i, len, add, del, total, adds = 0, dels = 0;
378 int max, max_change = 0, max_len = 0;
379 int total_files = data->nr;
381 if (data->nr == 0)
382 return;
384 for (i = 0; i < data->nr; i++) {
385 struct diffstat_file *file = data->files[i];
387 len = strlen(file->name);
388 if (max_len < len)
389 max_len = len;
391 if (file->is_binary || file->is_unmerged)
392 continue;
393 if (max_change < file->added + file->deleted)
394 max_change = file->added + file->deleted;
397 for (i = 0; i < data->nr; i++) {
398 char *prefix = "";
399 char *name = data->files[i]->name;
400 int added = data->files[i]->added;
401 int deleted = data->files[i]->deleted;
403 if (0 < (len = quote_c_style(name, NULL, NULL, 0))) {
404 char *qname = xmalloc(len + 1);
405 quote_c_style(name, qname, NULL, 0);
406 free(name);
407 data->files[i]->name = name = qname;
411 * "scale" the filename
413 len = strlen(name);
414 max = max_len;
415 if (max > 50)
416 max = 50;
417 if (len > max) {
418 char *slash;
419 prefix = "...";
420 max -= 3;
421 name += len - max;
422 slash = strchr(name, '/');
423 if (slash)
424 name = slash;
426 len = max;
429 * scale the add/delete
431 max = max_change;
432 if (max + len > 70)
433 max = 70 - len;
435 if (data->files[i]->is_binary) {
436 printf(" %s%-*s | Bin\n", prefix, len, name);
437 goto free_diffstat_file;
439 else if (data->files[i]->is_unmerged) {
440 printf(" %s%-*s | Unmerged\n", prefix, len, name);
441 goto free_diffstat_file;
443 else if (!data->files[i]->is_renamed &&
444 (added + deleted == 0)) {
445 total_files--;
446 goto free_diffstat_file;
449 add = added;
450 del = deleted;
451 total = add + del;
452 adds += add;
453 dels += del;
455 if (max_change > 0) {
456 total = (total * max + max_change / 2) / max_change;
457 add = (add * max + max_change / 2) / max_change;
458 del = total - add;
460 printf(" %s%-*s |%5d %.*s%.*s\n", prefix,
461 len, name, added + deleted,
462 add, pluses, del, minuses);
463 free_diffstat_file:
464 free(data->files[i]->name);
465 free(data->files[i]);
467 free(data->files);
468 printf(" %d files changed, %d insertions(+), %d deletions(-)\n",
469 total_files, adds, dels);
472 struct checkdiff_t {
473 struct xdiff_emit_state xm;
474 const char *filename;
475 int lineno;
478 static void checkdiff_consume(void *priv, char *line, unsigned long len)
480 struct checkdiff_t *data = priv;
482 if (line[0] == '+') {
483 int i, spaces = 0;
485 data->lineno++;
487 /* check space before tab */
488 for (i = 1; i < len && (line[i] == ' ' || line[i] == '\t'); i++)
489 if (line[i] == ' ')
490 spaces++;
491 if (line[i - 1] == '\t' && spaces)
492 printf("%s:%d: space before tab:%.*s\n",
493 data->filename, data->lineno, (int)len, line);
495 /* check white space at line end */
496 if (line[len - 1] == '\n')
497 len--;
498 if (isspace(line[len - 1]))
499 printf("%s:%d: white space at end: %.*s\n",
500 data->filename, data->lineno, (int)len, line);
501 } else if (line[0] == ' ')
502 data->lineno++;
503 else if (line[0] == '@') {
504 char *plus = strchr(line, '+');
505 if (plus)
506 data->lineno = strtol(plus, NULL, 10);
507 else
508 die("invalid diff");
512 static unsigned char *deflate_it(char *data,
513 unsigned long size,
514 unsigned long *result_size)
516 int bound;
517 unsigned char *deflated;
518 z_stream stream;
520 memset(&stream, 0, sizeof(stream));
521 deflateInit(&stream, Z_BEST_COMPRESSION);
522 bound = deflateBound(&stream, size);
523 deflated = xmalloc(bound);
524 stream.next_out = deflated;
525 stream.avail_out = bound;
527 stream.next_in = (unsigned char *)data;
528 stream.avail_in = size;
529 while (deflate(&stream, Z_FINISH) == Z_OK)
530 ; /* nothing */
531 deflateEnd(&stream);
532 *result_size = stream.total_out;
533 return deflated;
536 static void emit_binary_diff(mmfile_t *one, mmfile_t *two)
538 void *cp;
539 void *delta;
540 void *deflated;
541 void *data;
542 unsigned long orig_size;
543 unsigned long delta_size;
544 unsigned long deflate_size;
545 unsigned long data_size;
547 printf("GIT binary patch\n");
548 /* We could do deflated delta, or we could do just deflated two,
549 * whichever is smaller.
551 delta = NULL;
552 deflated = deflate_it(two->ptr, two->size, &deflate_size);
553 if (one->size && two->size) {
554 delta = diff_delta(one->ptr, one->size,
555 two->ptr, two->size,
556 &delta_size, deflate_size);
557 if (delta) {
558 void *to_free = delta;
559 orig_size = delta_size;
560 delta = deflate_it(delta, delta_size, &delta_size);
561 free(to_free);
565 if (delta && delta_size < deflate_size) {
566 printf("delta %lu\n", orig_size);
567 free(deflated);
568 data = delta;
569 data_size = delta_size;
571 else {
572 printf("literal %lu\n", two->size);
573 free(delta);
574 data = deflated;
575 data_size = deflate_size;
578 /* emit data encoded in base85 */
579 cp = data;
580 while (data_size) {
581 int bytes = (52 < data_size) ? 52 : data_size;
582 char line[70];
583 data_size -= bytes;
584 if (bytes <= 26)
585 line[0] = bytes + 'A' - 1;
586 else
587 line[0] = bytes - 26 + 'a' - 1;
588 encode_85(line + 1, cp, bytes);
589 cp = (char *) cp + bytes;
590 puts(line);
592 printf("\n");
593 free(data);
596 #define FIRST_FEW_BYTES 8000
597 static int mmfile_is_binary(mmfile_t *mf)
599 long sz = mf->size;
600 if (FIRST_FEW_BYTES < sz)
601 sz = FIRST_FEW_BYTES;
602 if (memchr(mf->ptr, 0, sz))
603 return 1;
604 return 0;
607 static void builtin_diff(const char *name_a,
608 const char *name_b,
609 struct diff_filespec *one,
610 struct diff_filespec *two,
611 const char *xfrm_msg,
612 struct diff_options *o,
613 int complete_rewrite)
615 mmfile_t mf1, mf2;
616 const char *lbl[2];
617 char *a_one, *b_two;
618 const char *set = get_color(o->color_diff, DIFF_METAINFO);
619 const char *reset = get_color(o->color_diff, DIFF_PLAIN);
621 a_one = quote_two("a/", name_a);
622 b_two = quote_two("b/", name_b);
623 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
624 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
625 printf("%sdiff --git %s %s%s\n", set, a_one, b_two, reset);
626 if (lbl[0][0] == '/') {
627 /* /dev/null */
628 printf("%snew file mode %06o%s\n", set, two->mode, reset);
629 if (xfrm_msg && xfrm_msg[0])
630 printf("%s%s%s\n", set, xfrm_msg, reset);
632 else if (lbl[1][0] == '/') {
633 printf("%sdeleted file mode %06o%s\n", set, one->mode, reset);
634 if (xfrm_msg && xfrm_msg[0])
635 printf("%s%s%s\n", set, xfrm_msg, reset);
637 else {
638 if (one->mode != two->mode) {
639 printf("%sold mode %06o%s\n", set, one->mode, reset);
640 printf("%snew mode %06o%s\n", set, two->mode, reset);
642 if (xfrm_msg && xfrm_msg[0])
643 printf("%s%s%s\n", set, xfrm_msg, reset);
645 * we do not run diff between different kind
646 * of objects.
648 if ((one->mode ^ two->mode) & S_IFMT)
649 goto free_ab_and_return;
650 if (complete_rewrite) {
651 emit_rewrite_diff(name_a, name_b, one, two);
652 goto free_ab_and_return;
656 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
657 die("unable to read files to diff");
659 if (mmfile_is_binary(&mf1) || mmfile_is_binary(&mf2)) {
660 /* Quite common confusing case */
661 if (mf1.size == mf2.size &&
662 !memcmp(mf1.ptr, mf2.ptr, mf1.size))
663 goto free_ab_and_return;
664 if (o->binary)
665 emit_binary_diff(&mf1, &mf2);
666 else
667 printf("Binary files %s and %s differ\n",
668 lbl[0], lbl[1]);
670 else {
671 /* Crazy xdl interfaces.. */
672 const char *diffopts = getenv("GIT_DIFF_OPTS");
673 xpparam_t xpp;
674 xdemitconf_t xecfg;
675 xdemitcb_t ecb;
676 struct emit_callback ecbdata;
678 memset(&ecbdata, 0, sizeof(ecbdata));
679 ecbdata.label_path = lbl;
680 ecbdata.color_diff = o->color_diff;
681 xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
682 xecfg.ctxlen = o->context;
683 xecfg.flags = XDL_EMIT_FUNCNAMES;
684 if (!diffopts)
686 else if (!strncmp(diffopts, "--unified=", 10))
687 xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
688 else if (!strncmp(diffopts, "-u", 2))
689 xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
690 ecb.outf = xdiff_outf;
691 ecb.priv = &ecbdata;
692 ecbdata.xm.consume = fn_out_consume;
693 xdl_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
696 free_ab_and_return:
697 free(a_one);
698 free(b_two);
699 return;
702 static void builtin_diffstat(const char *name_a, const char *name_b,
703 struct diff_filespec *one,
704 struct diff_filespec *two,
705 struct diffstat_t *diffstat,
706 struct diff_options *o,
707 int complete_rewrite)
709 mmfile_t mf1, mf2;
710 struct diffstat_file *data;
712 data = diffstat_add(diffstat, name_a, name_b);
714 if (!one || !two) {
715 data->is_unmerged = 1;
716 return;
718 if (complete_rewrite) {
719 diff_populate_filespec(one, 0);
720 diff_populate_filespec(two, 0);
721 data->deleted = count_lines(one->data, one->size);
722 data->added = count_lines(two->data, two->size);
723 return;
725 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
726 die("unable to read files to diff");
728 if (mmfile_is_binary(&mf1) || mmfile_is_binary(&mf2))
729 data->is_binary = 1;
730 else {
731 /* Crazy xdl interfaces.. */
732 xpparam_t xpp;
733 xdemitconf_t xecfg;
734 xdemitcb_t ecb;
736 xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
737 xecfg.ctxlen = 0;
738 xecfg.flags = 0;
739 ecb.outf = xdiff_outf;
740 ecb.priv = diffstat;
741 xdl_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
745 static void builtin_checkdiff(const char *name_a, const char *name_b,
746 struct diff_filespec *one,
747 struct diff_filespec *two)
749 mmfile_t mf1, mf2;
750 struct checkdiff_t data;
752 if (!two)
753 return;
755 memset(&data, 0, sizeof(data));
756 data.xm.consume = checkdiff_consume;
757 data.filename = name_b ? name_b : name_a;
758 data.lineno = 0;
760 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
761 die("unable to read files to diff");
763 if (mmfile_is_binary(&mf2))
764 return;
765 else {
766 /* Crazy xdl interfaces.. */
767 xpparam_t xpp;
768 xdemitconf_t xecfg;
769 xdemitcb_t ecb;
771 xpp.flags = XDF_NEED_MINIMAL;
772 xecfg.ctxlen = 0;
773 xecfg.flags = 0;
774 ecb.outf = xdiff_outf;
775 ecb.priv = &data;
776 xdl_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
780 struct diff_filespec *alloc_filespec(const char *path)
782 int namelen = strlen(path);
783 struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
785 memset(spec, 0, sizeof(*spec));
786 spec->path = (char *)(spec + 1);
787 memcpy(spec->path, path, namelen+1);
788 return spec;
791 void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
792 unsigned short mode)
794 if (mode) {
795 spec->mode = canon_mode(mode);
796 memcpy(spec->sha1, sha1, 20);
797 spec->sha1_valid = !!memcmp(sha1, null_sha1, 20);
802 * Given a name and sha1 pair, if the dircache tells us the file in
803 * the work tree has that object contents, return true, so that
804 * prepare_temp_file() does not have to inflate and extract.
806 static int work_tree_matches(const char *name, const unsigned char *sha1)
808 struct cache_entry *ce;
809 struct stat st;
810 int pos, len;
812 /* We do not read the cache ourselves here, because the
813 * benchmark with my previous version that always reads cache
814 * shows that it makes things worse for diff-tree comparing
815 * two linux-2.6 kernel trees in an already checked out work
816 * tree. This is because most diff-tree comparisons deal with
817 * only a small number of files, while reading the cache is
818 * expensive for a large project, and its cost outweighs the
819 * savings we get by not inflating the object to a temporary
820 * file. Practically, this code only helps when we are used
821 * by diff-cache --cached, which does read the cache before
822 * calling us.
824 if (!active_cache)
825 return 0;
827 len = strlen(name);
828 pos = cache_name_pos(name, len);
829 if (pos < 0)
830 return 0;
831 ce = active_cache[pos];
832 if ((lstat(name, &st) < 0) ||
833 !S_ISREG(st.st_mode) || /* careful! */
834 ce_match_stat(ce, &st, 0) ||
835 memcmp(sha1, ce->sha1, 20))
836 return 0;
837 /* we return 1 only when we can stat, it is a regular file,
838 * stat information matches, and sha1 recorded in the cache
839 * matches. I.e. we know the file in the work tree really is
840 * the same as the <name, sha1> pair.
842 return 1;
845 static struct sha1_size_cache {
846 unsigned char sha1[20];
847 unsigned long size;
848 } **sha1_size_cache;
849 static int sha1_size_cache_nr, sha1_size_cache_alloc;
851 static struct sha1_size_cache *locate_size_cache(unsigned char *sha1,
852 int find_only,
853 unsigned long size)
855 int first, last;
856 struct sha1_size_cache *e;
858 first = 0;
859 last = sha1_size_cache_nr;
860 while (last > first) {
861 int cmp, next = (last + first) >> 1;
862 e = sha1_size_cache[next];
863 cmp = memcmp(e->sha1, sha1, 20);
864 if (!cmp)
865 return e;
866 if (cmp < 0) {
867 last = next;
868 continue;
870 first = next+1;
872 /* not found */
873 if (find_only)
874 return NULL;
875 /* insert to make it at "first" */
876 if (sha1_size_cache_alloc <= sha1_size_cache_nr) {
877 sha1_size_cache_alloc = alloc_nr(sha1_size_cache_alloc);
878 sha1_size_cache = xrealloc(sha1_size_cache,
879 sha1_size_cache_alloc *
880 sizeof(*sha1_size_cache));
882 sha1_size_cache_nr++;
883 if (first < sha1_size_cache_nr)
884 memmove(sha1_size_cache + first + 1, sha1_size_cache + first,
885 (sha1_size_cache_nr - first - 1) *
886 sizeof(*sha1_size_cache));
887 e = xmalloc(sizeof(struct sha1_size_cache));
888 sha1_size_cache[first] = e;
889 memcpy(e->sha1, sha1, 20);
890 e->size = size;
891 return e;
895 * While doing rename detection and pickaxe operation, we may need to
896 * grab the data for the blob (or file) for our own in-core comparison.
897 * diff_filespec has data and size fields for this purpose.
899 int diff_populate_filespec(struct diff_filespec *s, int size_only)
901 int err = 0;
902 if (!DIFF_FILE_VALID(s))
903 die("internal error: asking to populate invalid file.");
904 if (S_ISDIR(s->mode))
905 return -1;
907 if (!use_size_cache)
908 size_only = 0;
910 if (s->data)
911 return err;
912 if (!s->sha1_valid ||
913 work_tree_matches(s->path, s->sha1)) {
914 struct stat st;
915 int fd;
916 if (lstat(s->path, &st) < 0) {
917 if (errno == ENOENT) {
918 err_empty:
919 err = -1;
920 empty:
921 s->data = "";
922 s->size = 0;
923 return err;
926 s->size = st.st_size;
927 if (!s->size)
928 goto empty;
929 if (size_only)
930 return 0;
931 if (S_ISLNK(st.st_mode)) {
932 int ret;
933 s->data = xmalloc(s->size);
934 s->should_free = 1;
935 ret = readlink(s->path, s->data, s->size);
936 if (ret < 0) {
937 free(s->data);
938 goto err_empty;
940 return 0;
942 fd = open(s->path, O_RDONLY);
943 if (fd < 0)
944 goto err_empty;
945 s->data = mmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
946 close(fd);
947 if (s->data == MAP_FAILED)
948 goto err_empty;
949 s->should_munmap = 1;
951 else {
952 char type[20];
953 struct sha1_size_cache *e;
955 if (size_only) {
956 e = locate_size_cache(s->sha1, 1, 0);
957 if (e) {
958 s->size = e->size;
959 return 0;
961 if (!sha1_object_info(s->sha1, type, &s->size))
962 locate_size_cache(s->sha1, 0, s->size);
964 else {
965 s->data = read_sha1_file(s->sha1, type, &s->size);
966 s->should_free = 1;
969 return 0;
972 void diff_free_filespec_data(struct diff_filespec *s)
974 if (s->should_free)
975 free(s->data);
976 else if (s->should_munmap)
977 munmap(s->data, s->size);
978 s->should_free = s->should_munmap = 0;
979 s->data = NULL;
980 free(s->cnt_data);
981 s->cnt_data = NULL;
984 static void prep_temp_blob(struct diff_tempfile *temp,
985 void *blob,
986 unsigned long size,
987 const unsigned char *sha1,
988 int mode)
990 int fd;
992 fd = git_mkstemp(temp->tmp_path, TEMPFILE_PATH_LEN, ".diff_XXXXXX");
993 if (fd < 0)
994 die("unable to create temp-file");
995 if (write(fd, blob, size) != size)
996 die("unable to write temp-file");
997 close(fd);
998 temp->name = temp->tmp_path;
999 strcpy(temp->hex, sha1_to_hex(sha1));
1000 temp->hex[40] = 0;
1001 sprintf(temp->mode, "%06o", mode);
1004 static void prepare_temp_file(const char *name,
1005 struct diff_tempfile *temp,
1006 struct diff_filespec *one)
1008 if (!DIFF_FILE_VALID(one)) {
1009 not_a_valid_file:
1010 /* A '-' entry produces this for file-2, and
1011 * a '+' entry produces this for file-1.
1013 temp->name = "/dev/null";
1014 strcpy(temp->hex, ".");
1015 strcpy(temp->mode, ".");
1016 return;
1019 if (!one->sha1_valid ||
1020 work_tree_matches(name, one->sha1)) {
1021 struct stat st;
1022 if (lstat(name, &st) < 0) {
1023 if (errno == ENOENT)
1024 goto not_a_valid_file;
1025 die("stat(%s): %s", name, strerror(errno));
1027 if (S_ISLNK(st.st_mode)) {
1028 int ret;
1029 char buf[PATH_MAX + 1]; /* ought to be SYMLINK_MAX */
1030 if (sizeof(buf) <= st.st_size)
1031 die("symlink too long: %s", name);
1032 ret = readlink(name, buf, st.st_size);
1033 if (ret < 0)
1034 die("readlink(%s)", name);
1035 prep_temp_blob(temp, buf, st.st_size,
1036 (one->sha1_valid ?
1037 one->sha1 : null_sha1),
1038 (one->sha1_valid ?
1039 one->mode : S_IFLNK));
1041 else {
1042 /* we can borrow from the file in the work tree */
1043 temp->name = name;
1044 if (!one->sha1_valid)
1045 strcpy(temp->hex, sha1_to_hex(null_sha1));
1046 else
1047 strcpy(temp->hex, sha1_to_hex(one->sha1));
1048 /* Even though we may sometimes borrow the
1049 * contents from the work tree, we always want
1050 * one->mode. mode is trustworthy even when
1051 * !(one->sha1_valid), as long as
1052 * DIFF_FILE_VALID(one).
1054 sprintf(temp->mode, "%06o", one->mode);
1056 return;
1058 else {
1059 if (diff_populate_filespec(one, 0))
1060 die("cannot read data blob for %s", one->path);
1061 prep_temp_blob(temp, one->data, one->size,
1062 one->sha1, one->mode);
1066 static void remove_tempfile(void)
1068 int i;
1070 for (i = 0; i < 2; i++)
1071 if (diff_temp[i].name == diff_temp[i].tmp_path) {
1072 unlink(diff_temp[i].name);
1073 diff_temp[i].name = NULL;
1077 static void remove_tempfile_on_signal(int signo)
1079 remove_tempfile();
1080 signal(SIGINT, SIG_DFL);
1081 raise(signo);
1084 static int spawn_prog(const char *pgm, const char **arg)
1086 pid_t pid;
1087 int status;
1089 fflush(NULL);
1090 pid = fork();
1091 if (pid < 0)
1092 die("unable to fork");
1093 if (!pid) {
1094 execvp(pgm, (char *const*) arg);
1095 exit(255);
1098 while (waitpid(pid, &status, 0) < 0) {
1099 if (errno == EINTR)
1100 continue;
1101 return -1;
1104 /* Earlier we did not check the exit status because
1105 * diff exits non-zero if files are different, and
1106 * we are not interested in knowing that. It was a
1107 * mistake which made it harder to quit a diff-*
1108 * session that uses the git-apply-patch-script as
1109 * the GIT_EXTERNAL_DIFF. A custom GIT_EXTERNAL_DIFF
1110 * should also exit non-zero only when it wants to
1111 * abort the entire diff-* session.
1113 if (WIFEXITED(status) && !WEXITSTATUS(status))
1114 return 0;
1115 return -1;
1118 /* An external diff command takes:
1120 * diff-cmd name infile1 infile1-sha1 infile1-mode \
1121 * infile2 infile2-sha1 infile2-mode [ rename-to ]
1124 static void run_external_diff(const char *pgm,
1125 const char *name,
1126 const char *other,
1127 struct diff_filespec *one,
1128 struct diff_filespec *two,
1129 const char *xfrm_msg,
1130 int complete_rewrite)
1132 const char *spawn_arg[10];
1133 struct diff_tempfile *temp = diff_temp;
1134 int retval;
1135 static int atexit_asked = 0;
1136 const char *othername;
1137 const char **arg = &spawn_arg[0];
1139 othername = (other? other : name);
1140 if (one && two) {
1141 prepare_temp_file(name, &temp[0], one);
1142 prepare_temp_file(othername, &temp[1], two);
1143 if (! atexit_asked &&
1144 (temp[0].name == temp[0].tmp_path ||
1145 temp[1].name == temp[1].tmp_path)) {
1146 atexit_asked = 1;
1147 atexit(remove_tempfile);
1149 signal(SIGINT, remove_tempfile_on_signal);
1152 if (one && two) {
1153 *arg++ = pgm;
1154 *arg++ = name;
1155 *arg++ = temp[0].name;
1156 *arg++ = temp[0].hex;
1157 *arg++ = temp[0].mode;
1158 *arg++ = temp[1].name;
1159 *arg++ = temp[1].hex;
1160 *arg++ = temp[1].mode;
1161 if (other) {
1162 *arg++ = other;
1163 *arg++ = xfrm_msg;
1165 } else {
1166 *arg++ = pgm;
1167 *arg++ = name;
1169 *arg = NULL;
1170 retval = spawn_prog(pgm, spawn_arg);
1171 remove_tempfile();
1172 if (retval) {
1173 fprintf(stderr, "external diff died, stopping at %s.\n", name);
1174 exit(1);
1178 static void run_diff_cmd(const char *pgm,
1179 const char *name,
1180 const char *other,
1181 struct diff_filespec *one,
1182 struct diff_filespec *two,
1183 const char *xfrm_msg,
1184 struct diff_options *o,
1185 int complete_rewrite)
1187 if (pgm) {
1188 run_external_diff(pgm, name, other, one, two, xfrm_msg,
1189 complete_rewrite);
1190 return;
1192 if (one && two)
1193 builtin_diff(name, other ? other : name,
1194 one, two, xfrm_msg, o, complete_rewrite);
1195 else
1196 printf("* Unmerged path %s\n", name);
1199 static void diff_fill_sha1_info(struct diff_filespec *one)
1201 if (DIFF_FILE_VALID(one)) {
1202 if (!one->sha1_valid) {
1203 struct stat st;
1204 if (lstat(one->path, &st) < 0)
1205 die("stat %s", one->path);
1206 if (index_path(one->sha1, one->path, &st, 0))
1207 die("cannot hash %s\n", one->path);
1210 else
1211 memset(one->sha1, 0, 20);
1214 static void run_diff(struct diff_filepair *p, struct diff_options *o)
1216 const char *pgm = external_diff();
1217 char msg[PATH_MAX*2+300], *xfrm_msg;
1218 struct diff_filespec *one;
1219 struct diff_filespec *two;
1220 const char *name;
1221 const char *other;
1222 char *name_munged, *other_munged;
1223 int complete_rewrite = 0;
1224 int len;
1226 if (DIFF_PAIR_UNMERGED(p)) {
1227 /* unmerged */
1228 run_diff_cmd(pgm, p->one->path, NULL, NULL, NULL, NULL, o, 0);
1229 return;
1232 name = p->one->path;
1233 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
1234 name_munged = quote_one(name);
1235 other_munged = quote_one(other);
1236 one = p->one; two = p->two;
1238 diff_fill_sha1_info(one);
1239 diff_fill_sha1_info(two);
1241 len = 0;
1242 switch (p->status) {
1243 case DIFF_STATUS_COPIED:
1244 len += snprintf(msg + len, sizeof(msg) - len,
1245 "similarity index %d%%\n"
1246 "copy from %s\n"
1247 "copy to %s\n",
1248 (int)(0.5 + p->score * 100.0/MAX_SCORE),
1249 name_munged, other_munged);
1250 break;
1251 case DIFF_STATUS_RENAMED:
1252 len += snprintf(msg + len, sizeof(msg) - len,
1253 "similarity index %d%%\n"
1254 "rename from %s\n"
1255 "rename to %s\n",
1256 (int)(0.5 + p->score * 100.0/MAX_SCORE),
1257 name_munged, other_munged);
1258 break;
1259 case DIFF_STATUS_MODIFIED:
1260 if (p->score) {
1261 len += snprintf(msg + len, sizeof(msg) - len,
1262 "dissimilarity index %d%%\n",
1263 (int)(0.5 + p->score *
1264 100.0/MAX_SCORE));
1265 complete_rewrite = 1;
1266 break;
1268 /* fallthru */
1269 default:
1270 /* nothing */
1274 if (memcmp(one->sha1, two->sha1, 20)) {
1275 int abbrev = o->full_index ? 40 : DEFAULT_ABBREV;
1277 len += snprintf(msg + len, sizeof(msg) - len,
1278 "index %.*s..%.*s",
1279 abbrev, sha1_to_hex(one->sha1),
1280 abbrev, sha1_to_hex(two->sha1));
1281 if (one->mode == two->mode)
1282 len += snprintf(msg + len, sizeof(msg) - len,
1283 " %06o", one->mode);
1284 len += snprintf(msg + len, sizeof(msg) - len, "\n");
1287 if (len)
1288 msg[--len] = 0;
1289 xfrm_msg = len ? msg : NULL;
1291 if (!pgm &&
1292 DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
1293 (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
1294 /* a filepair that changes between file and symlink
1295 * needs to be split into deletion and creation.
1297 struct diff_filespec *null = alloc_filespec(two->path);
1298 run_diff_cmd(NULL, name, other, one, null, xfrm_msg, o, 0);
1299 free(null);
1300 null = alloc_filespec(one->path);
1301 run_diff_cmd(NULL, name, other, null, two, xfrm_msg, o, 0);
1302 free(null);
1304 else
1305 run_diff_cmd(pgm, name, other, one, two, xfrm_msg, o,
1306 complete_rewrite);
1308 free(name_munged);
1309 free(other_munged);
1312 static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
1313 struct diffstat_t *diffstat)
1315 const char *name;
1316 const char *other;
1317 int complete_rewrite = 0;
1319 if (DIFF_PAIR_UNMERGED(p)) {
1320 /* unmerged */
1321 builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, 0);
1322 return;
1325 name = p->one->path;
1326 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
1328 diff_fill_sha1_info(p->one);
1329 diff_fill_sha1_info(p->two);
1331 if (p->status == DIFF_STATUS_MODIFIED && p->score)
1332 complete_rewrite = 1;
1333 builtin_diffstat(name, other, p->one, p->two, diffstat, o, complete_rewrite);
1336 static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
1338 const char *name;
1339 const char *other;
1341 if (DIFF_PAIR_UNMERGED(p)) {
1342 /* unmerged */
1343 return;
1346 name = p->one->path;
1347 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
1349 diff_fill_sha1_info(p->one);
1350 diff_fill_sha1_info(p->two);
1352 builtin_checkdiff(name, other, p->one, p->two);
1355 void diff_setup(struct diff_options *options)
1357 memset(options, 0, sizeof(*options));
1358 options->output_format = DIFF_FORMAT_RAW;
1359 options->line_termination = '\n';
1360 options->break_opt = -1;
1361 options->rename_limit = -1;
1362 options->context = 3;
1364 options->change = diff_change;
1365 options->add_remove = diff_addremove;
1368 int diff_setup_done(struct diff_options *options)
1370 if ((options->find_copies_harder &&
1371 options->detect_rename != DIFF_DETECT_COPY) ||
1372 (0 <= options->rename_limit && !options->detect_rename))
1373 return -1;
1376 * These cases always need recursive; we do not drop caller-supplied
1377 * recursive bits for other formats here.
1379 if ((options->output_format == DIFF_FORMAT_PATCH) ||
1380 (options->output_format == DIFF_FORMAT_DIFFSTAT) ||
1381 (options->output_format == DIFF_FORMAT_CHECKDIFF))
1382 options->recursive = 1;
1385 * These combinations do not make sense.
1387 if (options->output_format == DIFF_FORMAT_RAW)
1388 options->with_raw = 0;
1389 if (options->output_format == DIFF_FORMAT_DIFFSTAT)
1390 options->with_stat = 0;
1392 if (options->detect_rename && options->rename_limit < 0)
1393 options->rename_limit = diff_rename_limit_default;
1394 if (options->setup & DIFF_SETUP_USE_CACHE) {
1395 if (!active_cache)
1396 /* read-cache does not die even when it fails
1397 * so it is safe for us to do this here. Also
1398 * it does not smudge active_cache or active_nr
1399 * when it fails, so we do not have to worry about
1400 * cleaning it up ourselves either.
1402 read_cache();
1404 if (options->setup & DIFF_SETUP_USE_SIZE_CACHE)
1405 use_size_cache = 1;
1406 if (options->abbrev <= 0 || 40 < options->abbrev)
1407 options->abbrev = 40; /* full */
1409 return 0;
1412 int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
1414 char c, *eq;
1415 int len;
1417 if (*arg != '-')
1418 return 0;
1419 c = *++arg;
1420 if (!c)
1421 return 0;
1422 if (c == arg_short) {
1423 c = *++arg;
1424 if (!c)
1425 return 1;
1426 if (val && isdigit(c)) {
1427 char *end;
1428 int n = strtoul(arg, &end, 10);
1429 if (*end)
1430 return 0;
1431 *val = n;
1432 return 1;
1434 return 0;
1436 if (c != '-')
1437 return 0;
1438 arg++;
1439 eq = strchr(arg, '=');
1440 if (eq)
1441 len = eq - arg;
1442 else
1443 len = strlen(arg);
1444 if (!len || strncmp(arg, arg_long, len))
1445 return 0;
1446 if (eq) {
1447 int n;
1448 char *end;
1449 if (!isdigit(*++eq))
1450 return 0;
1451 n = strtoul(eq, &end, 10);
1452 if (*end)
1453 return 0;
1454 *val = n;
1456 return 1;
1459 int diff_opt_parse(struct diff_options *options, const char **av, int ac)
1461 const char *arg = av[0];
1462 if (!strcmp(arg, "-p") || !strcmp(arg, "-u"))
1463 options->output_format = DIFF_FORMAT_PATCH;
1464 else if (opt_arg(arg, 'U', "unified", &options->context))
1465 options->output_format = DIFF_FORMAT_PATCH;
1466 else if (!strcmp(arg, "--patch-with-raw")) {
1467 options->output_format = DIFF_FORMAT_PATCH;
1468 options->with_raw = 1;
1470 else if (!strcmp(arg, "--stat"))
1471 options->output_format = DIFF_FORMAT_DIFFSTAT;
1472 else if (!strcmp(arg, "--check"))
1473 options->output_format = DIFF_FORMAT_CHECKDIFF;
1474 else if (!strcmp(arg, "--summary"))
1475 options->summary = 1;
1476 else if (!strcmp(arg, "--patch-with-stat")) {
1477 options->output_format = DIFF_FORMAT_PATCH;
1478 options->with_stat = 1;
1480 else if (!strcmp(arg, "-z"))
1481 options->line_termination = 0;
1482 else if (!strncmp(arg, "-l", 2))
1483 options->rename_limit = strtoul(arg+2, NULL, 10);
1484 else if (!strcmp(arg, "--full-index"))
1485 options->full_index = 1;
1486 else if (!strcmp(arg, "--binary")) {
1487 options->output_format = DIFF_FORMAT_PATCH;
1488 options->full_index = options->binary = 1;
1490 else if (!strcmp(arg, "--name-only"))
1491 options->output_format = DIFF_FORMAT_NAME;
1492 else if (!strcmp(arg, "--name-status"))
1493 options->output_format = DIFF_FORMAT_NAME_STATUS;
1494 else if (!strcmp(arg, "-R"))
1495 options->reverse_diff = 1;
1496 else if (!strncmp(arg, "-S", 2))
1497 options->pickaxe = arg + 2;
1498 else if (!strcmp(arg, "-s"))
1499 options->output_format = DIFF_FORMAT_NO_OUTPUT;
1500 else if (!strncmp(arg, "-O", 2))
1501 options->orderfile = arg + 2;
1502 else if (!strncmp(arg, "--diff-filter=", 14))
1503 options->filter = arg + 14;
1504 else if (!strcmp(arg, "--pickaxe-all"))
1505 options->pickaxe_opts = DIFF_PICKAXE_ALL;
1506 else if (!strcmp(arg, "--pickaxe-regex"))
1507 options->pickaxe_opts = DIFF_PICKAXE_REGEX;
1508 else if (!strncmp(arg, "-B", 2)) {
1509 if ((options->break_opt =
1510 diff_scoreopt_parse(arg)) == -1)
1511 return -1;
1513 else if (!strncmp(arg, "-M", 2)) {
1514 if ((options->rename_score =
1515 diff_scoreopt_parse(arg)) == -1)
1516 return -1;
1517 options->detect_rename = DIFF_DETECT_RENAME;
1519 else if (!strncmp(arg, "-C", 2)) {
1520 if ((options->rename_score =
1521 diff_scoreopt_parse(arg)) == -1)
1522 return -1;
1523 options->detect_rename = DIFF_DETECT_COPY;
1525 else if (!strcmp(arg, "--find-copies-harder"))
1526 options->find_copies_harder = 1;
1527 else if (!strcmp(arg, "--abbrev"))
1528 options->abbrev = DEFAULT_ABBREV;
1529 else if (!strncmp(arg, "--abbrev=", 9)) {
1530 options->abbrev = strtoul(arg + 9, NULL, 10);
1531 if (options->abbrev < MINIMUM_ABBREV)
1532 options->abbrev = MINIMUM_ABBREV;
1533 else if (40 < options->abbrev)
1534 options->abbrev = 40;
1536 else if (!strcmp(arg, "--color"))
1537 options->color_diff = 1;
1538 else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
1539 options->xdl_opts |= XDF_IGNORE_WHITESPACE;
1540 else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
1541 options->xdl_opts |= XDF_IGNORE_WHITESPACE_CHANGE;
1542 else
1543 return 0;
1544 return 1;
1547 static int parse_num(const char **cp_p)
1549 unsigned long num, scale;
1550 int ch, dot;
1551 const char *cp = *cp_p;
1553 num = 0;
1554 scale = 1;
1555 dot = 0;
1556 for(;;) {
1557 ch = *cp;
1558 if ( !dot && ch == '.' ) {
1559 scale = 1;
1560 dot = 1;
1561 } else if ( ch == '%' ) {
1562 scale = dot ? scale*100 : 100;
1563 cp++; /* % is always at the end */
1564 break;
1565 } else if ( ch >= '0' && ch <= '9' ) {
1566 if ( scale < 100000 ) {
1567 scale *= 10;
1568 num = (num*10) + (ch-'0');
1570 } else {
1571 break;
1573 cp++;
1575 *cp_p = cp;
1577 /* user says num divided by scale and we say internally that
1578 * is MAX_SCORE * num / scale.
1580 return (num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale);
1583 int diff_scoreopt_parse(const char *opt)
1585 int opt1, opt2, cmd;
1587 if (*opt++ != '-')
1588 return -1;
1589 cmd = *opt++;
1590 if (cmd != 'M' && cmd != 'C' && cmd != 'B')
1591 return -1; /* that is not a -M, -C nor -B option */
1593 opt1 = parse_num(&opt);
1594 if (cmd != 'B')
1595 opt2 = 0;
1596 else {
1597 if (*opt == 0)
1598 opt2 = 0;
1599 else if (*opt != '/')
1600 return -1; /* we expect -B80/99 or -B80 */
1601 else {
1602 opt++;
1603 opt2 = parse_num(&opt);
1606 if (*opt != 0)
1607 return -1;
1608 return opt1 | (opt2 << 16);
1611 struct diff_queue_struct diff_queued_diff;
1613 void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
1615 if (queue->alloc <= queue->nr) {
1616 queue->alloc = alloc_nr(queue->alloc);
1617 queue->queue = xrealloc(queue->queue,
1618 sizeof(dp) * queue->alloc);
1620 queue->queue[queue->nr++] = dp;
1623 struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
1624 struct diff_filespec *one,
1625 struct diff_filespec *two)
1627 struct diff_filepair *dp = xmalloc(sizeof(*dp));
1628 dp->one = one;
1629 dp->two = two;
1630 dp->score = 0;
1631 dp->status = 0;
1632 dp->source_stays = 0;
1633 dp->broken_pair = 0;
1634 if (queue)
1635 diff_q(queue, dp);
1636 return dp;
1639 void diff_free_filepair(struct diff_filepair *p)
1641 diff_free_filespec_data(p->one);
1642 diff_free_filespec_data(p->two);
1643 free(p->one);
1644 free(p->two);
1645 free(p);
1648 /* This is different from find_unique_abbrev() in that
1649 * it stuffs the result with dots for alignment.
1651 const char *diff_unique_abbrev(const unsigned char *sha1, int len)
1653 int abblen;
1654 const char *abbrev;
1655 if (len == 40)
1656 return sha1_to_hex(sha1);
1658 abbrev = find_unique_abbrev(sha1, len);
1659 if (!abbrev)
1660 return sha1_to_hex(sha1);
1661 abblen = strlen(abbrev);
1662 if (abblen < 37) {
1663 static char hex[41];
1664 if (len < abblen && abblen <= len + 2)
1665 sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
1666 else
1667 sprintf(hex, "%s...", abbrev);
1668 return hex;
1670 return sha1_to_hex(sha1);
1673 static void diff_flush_raw(struct diff_filepair *p,
1674 int line_termination,
1675 int inter_name_termination,
1676 struct diff_options *options,
1677 int output_format)
1679 int two_paths;
1680 char status[10];
1681 int abbrev = options->abbrev;
1682 const char *path_one, *path_two;
1684 path_one = p->one->path;
1685 path_two = p->two->path;
1686 if (line_termination) {
1687 path_one = quote_one(path_one);
1688 path_two = quote_one(path_two);
1691 if (p->score)
1692 sprintf(status, "%c%03d", p->status,
1693 (int)(0.5 + p->score * 100.0/MAX_SCORE));
1694 else {
1695 status[0] = p->status;
1696 status[1] = 0;
1698 switch (p->status) {
1699 case DIFF_STATUS_COPIED:
1700 case DIFF_STATUS_RENAMED:
1701 two_paths = 1;
1702 break;
1703 case DIFF_STATUS_ADDED:
1704 case DIFF_STATUS_DELETED:
1705 two_paths = 0;
1706 break;
1707 default:
1708 two_paths = 0;
1709 break;
1711 if (output_format != DIFF_FORMAT_NAME_STATUS) {
1712 printf(":%06o %06o %s ",
1713 p->one->mode, p->two->mode,
1714 diff_unique_abbrev(p->one->sha1, abbrev));
1715 printf("%s ",
1716 diff_unique_abbrev(p->two->sha1, abbrev));
1718 printf("%s%c%s", status, inter_name_termination, path_one);
1719 if (two_paths)
1720 printf("%c%s", inter_name_termination, path_two);
1721 putchar(line_termination);
1722 if (path_one != p->one->path)
1723 free((void*)path_one);
1724 if (path_two != p->two->path)
1725 free((void*)path_two);
1728 static void diff_flush_name(struct diff_filepair *p,
1729 int inter_name_termination,
1730 int line_termination)
1732 char *path = p->two->path;
1734 if (line_termination)
1735 path = quote_one(p->two->path);
1736 else
1737 path = p->two->path;
1738 printf("%s%c", path, line_termination);
1739 if (p->two->path != path)
1740 free(path);
1743 int diff_unmodified_pair(struct diff_filepair *p)
1745 /* This function is written stricter than necessary to support
1746 * the currently implemented transformers, but the idea is to
1747 * let transformers to produce diff_filepairs any way they want,
1748 * and filter and clean them up here before producing the output.
1750 struct diff_filespec *one, *two;
1752 if (DIFF_PAIR_UNMERGED(p))
1753 return 0; /* unmerged is interesting */
1755 one = p->one;
1756 two = p->two;
1758 /* deletion, addition, mode or type change
1759 * and rename are all interesting.
1761 if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
1762 DIFF_PAIR_MODE_CHANGED(p) ||
1763 strcmp(one->path, two->path))
1764 return 0;
1766 /* both are valid and point at the same path. that is, we are
1767 * dealing with a change.
1769 if (one->sha1_valid && two->sha1_valid &&
1770 !memcmp(one->sha1, two->sha1, sizeof(one->sha1)))
1771 return 1; /* no change */
1772 if (!one->sha1_valid && !two->sha1_valid)
1773 return 1; /* both look at the same file on the filesystem. */
1774 return 0;
1777 static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
1779 if (diff_unmodified_pair(p))
1780 return;
1782 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
1783 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
1784 return; /* no tree diffs in patch format */
1786 run_diff(p, o);
1789 static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
1790 struct diffstat_t *diffstat)
1792 if (diff_unmodified_pair(p))
1793 return;
1795 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
1796 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
1797 return; /* no tree diffs in patch format */
1799 run_diffstat(p, o, diffstat);
1802 static void diff_flush_checkdiff(struct diff_filepair *p,
1803 struct diff_options *o)
1805 if (diff_unmodified_pair(p))
1806 return;
1808 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
1809 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
1810 return; /* no tree diffs in patch format */
1812 run_checkdiff(p, o);
1815 int diff_queue_is_empty(void)
1817 struct diff_queue_struct *q = &diff_queued_diff;
1818 int i;
1819 for (i = 0; i < q->nr; i++)
1820 if (!diff_unmodified_pair(q->queue[i]))
1821 return 0;
1822 return 1;
1825 #if DIFF_DEBUG
1826 void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
1828 fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
1829 x, one ? one : "",
1830 s->path,
1831 DIFF_FILE_VALID(s) ? "valid" : "invalid",
1832 s->mode,
1833 s->sha1_valid ? sha1_to_hex(s->sha1) : "");
1834 fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
1835 x, one ? one : "",
1836 s->size, s->xfrm_flags);
1839 void diff_debug_filepair(const struct diff_filepair *p, int i)
1841 diff_debug_filespec(p->one, i, "one");
1842 diff_debug_filespec(p->two, i, "two");
1843 fprintf(stderr, "score %d, status %c stays %d broken %d\n",
1844 p->score, p->status ? p->status : '?',
1845 p->source_stays, p->broken_pair);
1848 void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
1850 int i;
1851 if (msg)
1852 fprintf(stderr, "%s\n", msg);
1853 fprintf(stderr, "q->nr = %d\n", q->nr);
1854 for (i = 0; i < q->nr; i++) {
1855 struct diff_filepair *p = q->queue[i];
1856 diff_debug_filepair(p, i);
1859 #endif
1861 static void diff_resolve_rename_copy(void)
1863 int i, j;
1864 struct diff_filepair *p, *pp;
1865 struct diff_queue_struct *q = &diff_queued_diff;
1867 diff_debug_queue("resolve-rename-copy", q);
1869 for (i = 0; i < q->nr; i++) {
1870 p = q->queue[i];
1871 p->status = 0; /* undecided */
1872 if (DIFF_PAIR_UNMERGED(p))
1873 p->status = DIFF_STATUS_UNMERGED;
1874 else if (!DIFF_FILE_VALID(p->one))
1875 p->status = DIFF_STATUS_ADDED;
1876 else if (!DIFF_FILE_VALID(p->two))
1877 p->status = DIFF_STATUS_DELETED;
1878 else if (DIFF_PAIR_TYPE_CHANGED(p))
1879 p->status = DIFF_STATUS_TYPE_CHANGED;
1881 /* from this point on, we are dealing with a pair
1882 * whose both sides are valid and of the same type, i.e.
1883 * either in-place edit or rename/copy edit.
1885 else if (DIFF_PAIR_RENAME(p)) {
1886 if (p->source_stays) {
1887 p->status = DIFF_STATUS_COPIED;
1888 continue;
1890 /* See if there is some other filepair that
1891 * copies from the same source as us. If so
1892 * we are a copy. Otherwise we are either a
1893 * copy if the path stays, or a rename if it
1894 * does not, but we already handled "stays" case.
1896 for (j = i + 1; j < q->nr; j++) {
1897 pp = q->queue[j];
1898 if (strcmp(pp->one->path, p->one->path))
1899 continue; /* not us */
1900 if (!DIFF_PAIR_RENAME(pp))
1901 continue; /* not a rename/copy */
1902 /* pp is a rename/copy from the same source */
1903 p->status = DIFF_STATUS_COPIED;
1904 break;
1906 if (!p->status)
1907 p->status = DIFF_STATUS_RENAMED;
1909 else if (memcmp(p->one->sha1, p->two->sha1, 20) ||
1910 p->one->mode != p->two->mode)
1911 p->status = DIFF_STATUS_MODIFIED;
1912 else {
1913 /* This is a "no-change" entry and should not
1914 * happen anymore, but prepare for broken callers.
1916 error("feeding unmodified %s to diffcore",
1917 p->one->path);
1918 p->status = DIFF_STATUS_UNKNOWN;
1921 diff_debug_queue("resolve-rename-copy done", q);
1924 static void flush_one_pair(struct diff_filepair *p,
1925 int diff_output_format,
1926 struct diff_options *options,
1927 struct diffstat_t *diffstat)
1929 int inter_name_termination = '\t';
1930 int line_termination = options->line_termination;
1931 if (!line_termination)
1932 inter_name_termination = 0;
1934 switch (p->status) {
1935 case DIFF_STATUS_UNKNOWN:
1936 break;
1937 case 0:
1938 die("internal error in diff-resolve-rename-copy");
1939 break;
1940 default:
1941 switch (diff_output_format) {
1942 case DIFF_FORMAT_DIFFSTAT:
1943 diff_flush_stat(p, options, diffstat);
1944 break;
1945 case DIFF_FORMAT_CHECKDIFF:
1946 diff_flush_checkdiff(p, options);
1947 break;
1948 case DIFF_FORMAT_PATCH:
1949 diff_flush_patch(p, options);
1950 break;
1951 case DIFF_FORMAT_RAW:
1952 case DIFF_FORMAT_NAME_STATUS:
1953 diff_flush_raw(p, line_termination,
1954 inter_name_termination,
1955 options, diff_output_format);
1956 break;
1957 case DIFF_FORMAT_NAME:
1958 diff_flush_name(p,
1959 inter_name_termination,
1960 line_termination);
1961 break;
1962 case DIFF_FORMAT_NO_OUTPUT:
1963 break;
1968 static void show_file_mode_name(const char *newdelete, struct diff_filespec *fs)
1970 if (fs->mode)
1971 printf(" %s mode %06o %s\n", newdelete, fs->mode, fs->path);
1972 else
1973 printf(" %s %s\n", newdelete, fs->path);
1977 static void show_mode_change(struct diff_filepair *p, int show_name)
1979 if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
1980 if (show_name)
1981 printf(" mode change %06o => %06o %s\n",
1982 p->one->mode, p->two->mode, p->two->path);
1983 else
1984 printf(" mode change %06o => %06o\n",
1985 p->one->mode, p->two->mode);
1989 static void show_rename_copy(const char *renamecopy, struct diff_filepair *p)
1991 const char *old, *new;
1993 /* Find common prefix */
1994 old = p->one->path;
1995 new = p->two->path;
1996 while (1) {
1997 const char *slash_old, *slash_new;
1998 slash_old = strchr(old, '/');
1999 slash_new = strchr(new, '/');
2000 if (!slash_old ||
2001 !slash_new ||
2002 slash_old - old != slash_new - new ||
2003 memcmp(old, new, slash_new - new))
2004 break;
2005 old = slash_old + 1;
2006 new = slash_new + 1;
2008 /* p->one->path thru old is the common prefix, and old and new
2009 * through the end of names are renames
2011 if (old != p->one->path)
2012 printf(" %s %.*s{%s => %s} (%d%%)\n", renamecopy,
2013 (int)(old - p->one->path), p->one->path,
2014 old, new, (int)(0.5 + p->score * 100.0/MAX_SCORE));
2015 else
2016 printf(" %s %s => %s (%d%%)\n", renamecopy,
2017 p->one->path, p->two->path,
2018 (int)(0.5 + p->score * 100.0/MAX_SCORE));
2019 show_mode_change(p, 0);
2022 static void diff_summary(struct diff_filepair *p)
2024 switch(p->status) {
2025 case DIFF_STATUS_DELETED:
2026 show_file_mode_name("delete", p->one);
2027 break;
2028 case DIFF_STATUS_ADDED:
2029 show_file_mode_name("create", p->two);
2030 break;
2031 case DIFF_STATUS_COPIED:
2032 show_rename_copy("copy", p);
2033 break;
2034 case DIFF_STATUS_RENAMED:
2035 show_rename_copy("rename", p);
2036 break;
2037 default:
2038 if (p->score) {
2039 printf(" rewrite %s (%d%%)\n", p->two->path,
2040 (int)(0.5 + p->score * 100.0/MAX_SCORE));
2041 show_mode_change(p, 0);
2042 } else show_mode_change(p, 1);
2043 break;
2047 void diff_flush(struct diff_options *options)
2049 struct diff_queue_struct *q = &diff_queued_diff;
2050 int i;
2051 int diff_output_format = options->output_format;
2052 struct diffstat_t *diffstat = NULL;
2054 if (diff_output_format == DIFF_FORMAT_DIFFSTAT || options->with_stat) {
2055 diffstat = xcalloc(sizeof (struct diffstat_t), 1);
2056 diffstat->xm.consume = diffstat_consume;
2059 if (options->with_raw) {
2060 for (i = 0; i < q->nr; i++) {
2061 struct diff_filepair *p = q->queue[i];
2062 flush_one_pair(p, DIFF_FORMAT_RAW, options, NULL);
2064 putchar(options->line_termination);
2066 if (options->with_stat) {
2067 for (i = 0; i < q->nr; i++) {
2068 struct diff_filepair *p = q->queue[i];
2069 flush_one_pair(p, DIFF_FORMAT_DIFFSTAT, options,
2070 diffstat);
2072 show_stats(diffstat);
2073 free(diffstat);
2074 diffstat = NULL;
2075 if (options->summary)
2076 for (i = 0; i < q->nr; i++)
2077 diff_summary(q->queue[i]);
2078 if (options->stat_sep)
2079 fputs(options->stat_sep, stdout);
2080 else
2081 putchar(options->line_termination);
2083 for (i = 0; i < q->nr; i++) {
2084 struct diff_filepair *p = q->queue[i];
2085 flush_one_pair(p, diff_output_format, options, diffstat);
2088 if (diffstat) {
2089 show_stats(diffstat);
2090 free(diffstat);
2093 for (i = 0; i < q->nr; i++) {
2094 if (diffstat && options->summary)
2095 diff_summary(q->queue[i]);
2096 diff_free_filepair(q->queue[i]);
2099 free(q->queue);
2100 q->queue = NULL;
2101 q->nr = q->alloc = 0;
2104 static void diffcore_apply_filter(const char *filter)
2106 int i;
2107 struct diff_queue_struct *q = &diff_queued_diff;
2108 struct diff_queue_struct outq;
2109 outq.queue = NULL;
2110 outq.nr = outq.alloc = 0;
2112 if (!filter)
2113 return;
2115 if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
2116 int found;
2117 for (i = found = 0; !found && i < q->nr; i++) {
2118 struct diff_filepair *p = q->queue[i];
2119 if (((p->status == DIFF_STATUS_MODIFIED) &&
2120 ((p->score &&
2121 strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
2122 (!p->score &&
2123 strchr(filter, DIFF_STATUS_MODIFIED)))) ||
2124 ((p->status != DIFF_STATUS_MODIFIED) &&
2125 strchr(filter, p->status)))
2126 found++;
2128 if (found)
2129 return;
2131 /* otherwise we will clear the whole queue
2132 * by copying the empty outq at the end of this
2133 * function, but first clear the current entries
2134 * in the queue.
2136 for (i = 0; i < q->nr; i++)
2137 diff_free_filepair(q->queue[i]);
2139 else {
2140 /* Only the matching ones */
2141 for (i = 0; i < q->nr; i++) {
2142 struct diff_filepair *p = q->queue[i];
2144 if (((p->status == DIFF_STATUS_MODIFIED) &&
2145 ((p->score &&
2146 strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
2147 (!p->score &&
2148 strchr(filter, DIFF_STATUS_MODIFIED)))) ||
2149 ((p->status != DIFF_STATUS_MODIFIED) &&
2150 strchr(filter, p->status)))
2151 diff_q(&outq, p);
2152 else
2153 diff_free_filepair(p);
2156 free(q->queue);
2157 *q = outq;
2160 void diffcore_std(struct diff_options *options)
2162 if (options->break_opt != -1)
2163 diffcore_break(options->break_opt);
2164 if (options->detect_rename)
2165 diffcore_rename(options);
2166 if (options->break_opt != -1)
2167 diffcore_merge_broken();
2168 if (options->pickaxe)
2169 diffcore_pickaxe(options->pickaxe, options->pickaxe_opts);
2170 if (options->orderfile)
2171 diffcore_order(options->orderfile);
2172 diff_resolve_rename_copy();
2173 diffcore_apply_filter(options->filter);
2177 void diffcore_std_no_resolve(struct diff_options *options)
2179 if (options->pickaxe)
2180 diffcore_pickaxe(options->pickaxe, options->pickaxe_opts);
2181 if (options->orderfile)
2182 diffcore_order(options->orderfile);
2183 diffcore_apply_filter(options->filter);
2186 void diff_addremove(struct diff_options *options,
2187 int addremove, unsigned mode,
2188 const unsigned char *sha1,
2189 const char *base, const char *path)
2191 char concatpath[PATH_MAX];
2192 struct diff_filespec *one, *two;
2194 /* This may look odd, but it is a preparation for
2195 * feeding "there are unchanged files which should
2196 * not produce diffs, but when you are doing copy
2197 * detection you would need them, so here they are"
2198 * entries to the diff-core. They will be prefixed
2199 * with something like '=' or '*' (I haven't decided
2200 * which but should not make any difference).
2201 * Feeding the same new and old to diff_change()
2202 * also has the same effect.
2203 * Before the final output happens, they are pruned after
2204 * merged into rename/copy pairs as appropriate.
2206 if (options->reverse_diff)
2207 addremove = (addremove == '+' ? '-' :
2208 addremove == '-' ? '+' : addremove);
2210 if (!path) path = "";
2211 sprintf(concatpath, "%s%s", base, path);
2212 one = alloc_filespec(concatpath);
2213 two = alloc_filespec(concatpath);
2215 if (addremove != '+')
2216 fill_filespec(one, sha1, mode);
2217 if (addremove != '-')
2218 fill_filespec(two, sha1, mode);
2220 diff_queue(&diff_queued_diff, one, two);
2223 void diff_change(struct diff_options *options,
2224 unsigned old_mode, unsigned new_mode,
2225 const unsigned char *old_sha1,
2226 const unsigned char *new_sha1,
2227 const char *base, const char *path)
2229 char concatpath[PATH_MAX];
2230 struct diff_filespec *one, *two;
2232 if (options->reverse_diff) {
2233 unsigned tmp;
2234 const unsigned char *tmp_c;
2235 tmp = old_mode; old_mode = new_mode; new_mode = tmp;
2236 tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
2238 if (!path) path = "";
2239 sprintf(concatpath, "%s%s", base, path);
2240 one = alloc_filespec(concatpath);
2241 two = alloc_filespec(concatpath);
2242 fill_filespec(one, old_sha1, old_mode);
2243 fill_filespec(two, new_sha1, new_mode);
2245 diff_queue(&diff_queued_diff, one, two);
2248 void diff_unmerge(struct diff_options *options,
2249 const char *path)
2251 struct diff_filespec *one, *two;
2252 one = alloc_filespec(path);
2253 two = alloc_filespec(path);
2254 diff_queue(&diff_queued_diff, one, two);