Teach diff -B about colours
[git/dscho.git] / diff.c
blob019ecbc1fa80807fc7f59e12a931b0f77224e71f
1 /*
2 * Copyright (C) 2005 Junio C Hamano
3 */
4 #include "cache.h"
5 #include "quote.h"
6 #include "diff.h"
7 #include "diffcore.h"
8 #include "delta.h"
9 #include "xdiff-interface.h"
10 #include "color.h"
12 #ifdef NO_FAST_WORKING_DIRECTORY
13 #define FAST_WORKING_DIRECTORY 0
14 #else
15 #define FAST_WORKING_DIRECTORY 1
16 #endif
18 static int use_size_cache;
20 static int diff_detect_rename_default;
21 static int diff_rename_limit_default = -1;
22 static int diff_use_color_default;
24 static char diff_colors[][COLOR_MAXLEN] = {
25 "\033[m", /* reset */
26 "", /* PLAIN (normal) */
27 "\033[1m", /* METAINFO (bold) */
28 "\033[36m", /* FRAGINFO (cyan) */
29 "\033[31m", /* OLD (red) */
30 "\033[32m", /* NEW (green) */
31 "\033[33m", /* COMMIT (yellow) */
32 "\033[41m", /* WHITESPACE (red background) */
35 static int parse_diff_color_slot(const char *var, int ofs)
37 if (!strcasecmp(var+ofs, "plain"))
38 return DIFF_PLAIN;
39 if (!strcasecmp(var+ofs, "meta"))
40 return DIFF_METAINFO;
41 if (!strcasecmp(var+ofs, "frag"))
42 return DIFF_FRAGINFO;
43 if (!strcasecmp(var+ofs, "old"))
44 return DIFF_FILE_OLD;
45 if (!strcasecmp(var+ofs, "new"))
46 return DIFF_FILE_NEW;
47 if (!strcasecmp(var+ofs, "commit"))
48 return DIFF_COMMIT;
49 if (!strcasecmp(var+ofs, "whitespace"))
50 return DIFF_WHITESPACE;
51 die("bad config variable '%s'", var);
55 * These are to give UI layer defaults.
56 * The core-level commands such as git-diff-files should
57 * never be affected by the setting of diff.renames
58 * the user happens to have in the configuration file.
60 int git_diff_ui_config(const char *var, const char *value)
62 if (!strcmp(var, "diff.renamelimit")) {
63 diff_rename_limit_default = git_config_int(var, value);
64 return 0;
66 if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
67 diff_use_color_default = git_config_colorbool(var, value);
68 return 0;
70 if (!strcmp(var, "diff.renames")) {
71 if (!value)
72 diff_detect_rename_default = DIFF_DETECT_RENAME;
73 else if (!strcasecmp(value, "copies") ||
74 !strcasecmp(value, "copy"))
75 diff_detect_rename_default = DIFF_DETECT_COPY;
76 else if (git_config_bool(var,value))
77 diff_detect_rename_default = DIFF_DETECT_RENAME;
78 return 0;
80 if (!prefixcmp(var, "diff.color.") || !prefixcmp(var, "color.diff.")) {
81 int slot = parse_diff_color_slot(var, 11);
82 color_parse(value, var, diff_colors[slot]);
83 return 0;
85 return git_default_config(var, value);
88 static char *quote_one(const char *str)
90 int needlen;
91 char *xp;
93 if (!str)
94 return NULL;
95 needlen = quote_c_style(str, NULL, NULL, 0);
96 if (!needlen)
97 return xstrdup(str);
98 xp = xmalloc(needlen + 1);
99 quote_c_style(str, xp, NULL, 0);
100 return xp;
103 static char *quote_two(const char *one, const char *two)
105 int need_one = quote_c_style(one, NULL, NULL, 1);
106 int need_two = quote_c_style(two, NULL, NULL, 1);
107 char *xp;
109 if (need_one + need_two) {
110 if (!need_one) need_one = strlen(one);
111 if (!need_two) need_one = strlen(two);
113 xp = xmalloc(need_one + need_two + 3);
114 xp[0] = '"';
115 quote_c_style(one, xp + 1, NULL, 1);
116 quote_c_style(two, xp + need_one + 1, NULL, 1);
117 strcpy(xp + need_one + need_two + 1, "\"");
118 return xp;
120 need_one = strlen(one);
121 need_two = strlen(two);
122 xp = xmalloc(need_one + need_two + 1);
123 strcpy(xp, one);
124 strcpy(xp + need_one, two);
125 return xp;
128 static const char *external_diff(void)
130 static const char *external_diff_cmd = NULL;
131 static int done_preparing = 0;
133 if (done_preparing)
134 return external_diff_cmd;
135 external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
136 done_preparing = 1;
137 return external_diff_cmd;
140 #define TEMPFILE_PATH_LEN 50
142 static struct diff_tempfile {
143 const char *name; /* filename external diff should read from */
144 char hex[41];
145 char mode[10];
146 char tmp_path[TEMPFILE_PATH_LEN];
147 } diff_temp[2];
149 static int count_lines(const char *data, int size)
151 int count, ch, completely_empty = 1, nl_just_seen = 0;
152 count = 0;
153 while (0 < size--) {
154 ch = *data++;
155 if (ch == '\n') {
156 count++;
157 nl_just_seen = 1;
158 completely_empty = 0;
160 else {
161 nl_just_seen = 0;
162 completely_empty = 0;
165 if (completely_empty)
166 return 0;
167 if (!nl_just_seen)
168 count++; /* no trailing newline */
169 return count;
172 static void print_line_count(int count)
174 switch (count) {
175 case 0:
176 printf("0,0");
177 break;
178 case 1:
179 printf("1");
180 break;
181 default:
182 printf("1,%d", count);
183 break;
187 static void copy_file(int prefix, const char *data, int size,
188 const char *set, const char *reset)
190 int ch, nl_just_seen = 1;
191 while (0 < size--) {
192 ch = *data++;
193 if (nl_just_seen) {
194 fputs(set, stdout);
195 putchar(prefix);
197 if (ch == '\n') {
198 nl_just_seen = 1;
199 fputs(reset, stdout);
200 } else
201 nl_just_seen = 0;
202 putchar(ch);
204 if (!nl_just_seen)
205 printf("%s\n\\ No newline at end of file\n", reset);
208 static void emit_rewrite_diff(const char *name_a,
209 const char *name_b,
210 struct diff_filespec *one,
211 struct diff_filespec *two,
212 int color_diff)
214 int lc_a, lc_b;
215 const char *name_a_tab, *name_b_tab;
216 const char *metainfo = diff_get_color(color_diff, DIFF_METAINFO);
217 const char *fraginfo = diff_get_color(color_diff, DIFF_FRAGINFO);
218 const char *old = diff_get_color(color_diff, DIFF_FILE_OLD);
219 const char *new = diff_get_color(color_diff, DIFF_FILE_NEW);
220 const char *reset = diff_get_color(color_diff, DIFF_RESET);
222 name_a_tab = strchr(name_a, ' ') ? "\t" : "";
223 name_b_tab = strchr(name_b, ' ') ? "\t" : "";
225 diff_populate_filespec(one, 0);
226 diff_populate_filespec(two, 0);
227 lc_a = count_lines(one->data, one->size);
228 lc_b = count_lines(two->data, two->size);
229 printf("%s--- a/%s%s%s\n%s+++ b/%s%s%s\n%s@@ -",
230 metainfo, name_a, name_a_tab, reset,
231 metainfo, name_b, name_b_tab, reset, fraginfo);
232 print_line_count(lc_a);
233 printf(" +");
234 print_line_count(lc_b);
235 printf(" @@%s\n", reset);
236 if (lc_a)
237 copy_file('-', one->data, one->size, old, reset);
238 if (lc_b)
239 copy_file('+', two->data, two->size, new, reset);
242 static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
244 if (!DIFF_FILE_VALID(one)) {
245 mf->ptr = (char *)""; /* does not matter */
246 mf->size = 0;
247 return 0;
249 else if (diff_populate_filespec(one, 0))
250 return -1;
251 mf->ptr = one->data;
252 mf->size = one->size;
253 return 0;
256 struct diff_words_buffer {
257 mmfile_t text;
258 long alloc;
259 long current; /* output pointer */
260 int suppressed_newline;
263 static void diff_words_append(char *line, unsigned long len,
264 struct diff_words_buffer *buffer)
266 if (buffer->text.size + len > buffer->alloc) {
267 buffer->alloc = (buffer->text.size + len) * 3 / 2;
268 buffer->text.ptr = xrealloc(buffer->text.ptr, buffer->alloc);
270 line++;
271 len--;
272 memcpy(buffer->text.ptr + buffer->text.size, line, len);
273 buffer->text.size += len;
276 struct diff_words_data {
277 struct xdiff_emit_state xm;
278 struct diff_words_buffer minus, plus;
281 static void print_word(struct diff_words_buffer *buffer, int len, int color,
282 int suppress_newline)
284 const char *ptr;
285 int eol = 0;
287 if (len == 0)
288 return;
290 ptr = buffer->text.ptr + buffer->current;
291 buffer->current += len;
293 if (ptr[len - 1] == '\n') {
294 eol = 1;
295 len--;
298 fputs(diff_get_color(1, color), stdout);
299 fwrite(ptr, len, 1, stdout);
300 fputs(diff_get_color(1, DIFF_RESET), stdout);
302 if (eol) {
303 if (suppress_newline)
304 buffer->suppressed_newline = 1;
305 else
306 putchar('\n');
310 static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
312 struct diff_words_data *diff_words = priv;
314 if (diff_words->minus.suppressed_newline) {
315 if (line[0] != '+')
316 putchar('\n');
317 diff_words->minus.suppressed_newline = 0;
320 len--;
321 switch (line[0]) {
322 case '-':
323 print_word(&diff_words->minus, len, DIFF_FILE_OLD, 1);
324 break;
325 case '+':
326 print_word(&diff_words->plus, len, DIFF_FILE_NEW, 0);
327 break;
328 case ' ':
329 print_word(&diff_words->plus, len, DIFF_PLAIN, 0);
330 diff_words->minus.current += len;
331 break;
335 /* this executes the word diff on the accumulated buffers */
336 static void diff_words_show(struct diff_words_data *diff_words)
338 xpparam_t xpp;
339 xdemitconf_t xecfg;
340 xdemitcb_t ecb;
341 mmfile_t minus, plus;
342 int i;
344 minus.size = diff_words->minus.text.size;
345 minus.ptr = xmalloc(minus.size);
346 memcpy(minus.ptr, diff_words->minus.text.ptr, minus.size);
347 for (i = 0; i < minus.size; i++)
348 if (isspace(minus.ptr[i]))
349 minus.ptr[i] = '\n';
350 diff_words->minus.current = 0;
352 plus.size = diff_words->plus.text.size;
353 plus.ptr = xmalloc(plus.size);
354 memcpy(plus.ptr, diff_words->plus.text.ptr, plus.size);
355 for (i = 0; i < plus.size; i++)
356 if (isspace(plus.ptr[i]))
357 plus.ptr[i] = '\n';
358 diff_words->plus.current = 0;
360 xpp.flags = XDF_NEED_MINIMAL;
361 xecfg.ctxlen = diff_words->minus.alloc + diff_words->plus.alloc;
362 xecfg.flags = 0;
363 ecb.outf = xdiff_outf;
364 ecb.priv = diff_words;
365 diff_words->xm.consume = fn_out_diff_words_aux;
366 xdl_diff(&minus, &plus, &xpp, &xecfg, &ecb);
368 free(minus.ptr);
369 free(plus.ptr);
370 diff_words->minus.text.size = diff_words->plus.text.size = 0;
372 if (diff_words->minus.suppressed_newline) {
373 putchar('\n');
374 diff_words->minus.suppressed_newline = 0;
378 struct emit_callback {
379 struct xdiff_emit_state xm;
380 int nparents, color_diff;
381 const char **label_path;
382 struct diff_words_data *diff_words;
385 static void free_diff_words_data(struct emit_callback *ecbdata)
387 if (ecbdata->diff_words) {
388 /* flush buffers */
389 if (ecbdata->diff_words->minus.text.size ||
390 ecbdata->diff_words->plus.text.size)
391 diff_words_show(ecbdata->diff_words);
393 if (ecbdata->diff_words->minus.text.ptr)
394 free (ecbdata->diff_words->minus.text.ptr);
395 if (ecbdata->diff_words->plus.text.ptr)
396 free (ecbdata->diff_words->plus.text.ptr);
397 free(ecbdata->diff_words);
398 ecbdata->diff_words = NULL;
402 const char *diff_get_color(int diff_use_color, enum color_diff ix)
404 if (diff_use_color)
405 return diff_colors[ix];
406 return "";
409 static void emit_line(const char *set, const char *reset, const char *line, int len)
411 if (len > 0 && line[len-1] == '\n')
412 len--;
413 fputs(set, stdout);
414 fwrite(line, len, 1, stdout);
415 puts(reset);
418 static void emit_line_with_ws(int nparents,
419 const char *set, const char *reset, const char *ws,
420 const char *line, int len)
422 int col0 = nparents;
423 int last_tab_in_indent = -1;
424 int last_space_in_indent = -1;
425 int i;
426 int tail = len;
427 int need_highlight_leading_space = 0;
428 /* The line is a newly added line. Does it have funny leading
429 * whitespaces? In indent, SP should never precede a TAB.
431 for (i = col0; i < len; i++) {
432 if (line[i] == '\t') {
433 last_tab_in_indent = i;
434 if (0 <= last_space_in_indent)
435 need_highlight_leading_space = 1;
437 else if (line[i] == ' ')
438 last_space_in_indent = i;
439 else
440 break;
442 fputs(set, stdout);
443 fwrite(line, col0, 1, stdout);
444 fputs(reset, stdout);
445 if (((i == len) || line[i] == '\n') && i != col0) {
446 /* The whole line was indent */
447 emit_line(ws, reset, line + col0, len - col0);
448 return;
450 i = col0;
451 if (need_highlight_leading_space) {
452 while (i < last_tab_in_indent) {
453 if (line[i] == ' ') {
454 fputs(ws, stdout);
455 putchar(' ');
456 fputs(reset, stdout);
458 else
459 putchar(line[i]);
460 i++;
463 tail = len - 1;
464 if (line[tail] == '\n' && i < tail)
465 tail--;
466 while (i < tail) {
467 if (!isspace(line[tail]))
468 break;
469 tail--;
471 if ((i < tail && line[tail + 1] != '\n')) {
472 /* This has whitespace between tail+1..len */
473 fputs(set, stdout);
474 fwrite(line + i, tail - i + 1, 1, stdout);
475 fputs(reset, stdout);
476 emit_line(ws, reset, line + tail + 1, len - tail - 1);
478 else
479 emit_line(set, reset, line + i, len - i);
482 static void emit_add_line(const char *reset, struct emit_callback *ecbdata, const char *line, int len)
484 const char *ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
485 const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_NEW);
487 if (!*ws)
488 emit_line(set, reset, line, len);
489 else
490 emit_line_with_ws(ecbdata->nparents, set, reset, ws,
491 line, len);
494 static void fn_out_consume(void *priv, char *line, unsigned long len)
496 int i;
497 int color;
498 struct emit_callback *ecbdata = priv;
499 const char *set = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
500 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
502 if (ecbdata->label_path[0]) {
503 const char *name_a_tab, *name_b_tab;
505 name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
506 name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
508 printf("%s--- %s%s%s\n",
509 set, ecbdata->label_path[0], reset, name_a_tab);
510 printf("%s+++ %s%s%s\n",
511 set, ecbdata->label_path[1], reset, name_b_tab);
512 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
515 /* This is not really necessary for now because
516 * this codepath only deals with two-way diffs.
518 for (i = 0; i < len && line[i] == '@'; i++)
520 if (2 <= i && i < len && line[i] == ' ') {
521 ecbdata->nparents = i - 1;
522 emit_line(diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO),
523 reset, line, len);
524 return;
527 if (len < ecbdata->nparents) {
528 set = reset;
529 emit_line(reset, reset, line, len);
530 return;
533 color = DIFF_PLAIN;
534 if (ecbdata->diff_words && ecbdata->nparents != 1)
535 /* fall back to normal diff */
536 free_diff_words_data(ecbdata);
537 if (ecbdata->diff_words) {
538 if (line[0] == '-') {
539 diff_words_append(line, len,
540 &ecbdata->diff_words->minus);
541 return;
542 } else if (line[0] == '+') {
543 diff_words_append(line, len,
544 &ecbdata->diff_words->plus);
545 return;
547 if (ecbdata->diff_words->minus.text.size ||
548 ecbdata->diff_words->plus.text.size)
549 diff_words_show(ecbdata->diff_words);
550 line++;
551 len--;
552 emit_line(set, reset, line, len);
553 return;
555 for (i = 0; i < ecbdata->nparents && len; i++) {
556 if (line[i] == '-')
557 color = DIFF_FILE_OLD;
558 else if (line[i] == '+')
559 color = DIFF_FILE_NEW;
562 if (color != DIFF_FILE_NEW) {
563 emit_line(diff_get_color(ecbdata->color_diff, color),
564 reset, line, len);
565 return;
567 emit_add_line(reset, ecbdata, line, len);
570 static char *pprint_rename(const char *a, const char *b)
572 const char *old = a;
573 const char *new = b;
574 char *name = NULL;
575 int pfx_length, sfx_length;
576 int len_a = strlen(a);
577 int len_b = strlen(b);
578 int qlen_a = quote_c_style(a, NULL, NULL, 0);
579 int qlen_b = quote_c_style(b, NULL, NULL, 0);
581 if (qlen_a || qlen_b) {
582 if (qlen_a) len_a = qlen_a;
583 if (qlen_b) len_b = qlen_b;
584 name = xmalloc( len_a + len_b + 5 );
585 if (qlen_a)
586 quote_c_style(a, name, NULL, 0);
587 else
588 memcpy(name, a, len_a);
589 memcpy(name + len_a, " => ", 4);
590 if (qlen_b)
591 quote_c_style(b, name + len_a + 4, NULL, 0);
592 else
593 memcpy(name + len_a + 4, b, len_b + 1);
594 return name;
597 /* Find common prefix */
598 pfx_length = 0;
599 while (*old && *new && *old == *new) {
600 if (*old == '/')
601 pfx_length = old - a + 1;
602 old++;
603 new++;
606 /* Find common suffix */
607 old = a + len_a;
608 new = b + len_b;
609 sfx_length = 0;
610 while (a <= old && b <= new && *old == *new) {
611 if (*old == '/')
612 sfx_length = len_a - (old - a);
613 old--;
614 new--;
618 * pfx{mid-a => mid-b}sfx
619 * {pfx-a => pfx-b}sfx
620 * pfx{sfx-a => sfx-b}
621 * name-a => name-b
623 if (pfx_length + sfx_length) {
624 int a_midlen = len_a - pfx_length - sfx_length;
625 int b_midlen = len_b - pfx_length - sfx_length;
626 if (a_midlen < 0) a_midlen = 0;
627 if (b_midlen < 0) b_midlen = 0;
629 name = xmalloc(pfx_length + a_midlen + b_midlen + sfx_length + 7);
630 sprintf(name, "%.*s{%.*s => %.*s}%s",
631 pfx_length, a,
632 a_midlen, a + pfx_length,
633 b_midlen, b + pfx_length,
634 a + len_a - sfx_length);
636 else {
637 name = xmalloc(len_a + len_b + 5);
638 sprintf(name, "%s => %s", a, b);
640 return name;
643 struct diffstat_t {
644 struct xdiff_emit_state xm;
646 int nr;
647 int alloc;
648 struct diffstat_file {
649 char *name;
650 unsigned is_unmerged:1;
651 unsigned is_binary:1;
652 unsigned is_renamed:1;
653 unsigned int added, deleted;
654 } **files;
657 static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
658 const char *name_a,
659 const char *name_b)
661 struct diffstat_file *x;
662 x = xcalloc(sizeof (*x), 1);
663 if (diffstat->nr == diffstat->alloc) {
664 diffstat->alloc = alloc_nr(diffstat->alloc);
665 diffstat->files = xrealloc(diffstat->files,
666 diffstat->alloc * sizeof(x));
668 diffstat->files[diffstat->nr++] = x;
669 if (name_b) {
670 x->name = pprint_rename(name_a, name_b);
671 x->is_renamed = 1;
673 else
674 x->name = xstrdup(name_a);
675 return x;
678 static void diffstat_consume(void *priv, char *line, unsigned long len)
680 struct diffstat_t *diffstat = priv;
681 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
683 if (line[0] == '+')
684 x->added++;
685 else if (line[0] == '-')
686 x->deleted++;
689 const char mime_boundary_leader[] = "------------";
691 static int scale_linear(int it, int width, int max_change)
694 * make sure that at least one '-' is printed if there were deletions,
695 * and likewise for '+'.
697 if (max_change < 2)
698 return it;
699 return ((it - 1) * (width - 1) + max_change - 1) / (max_change - 1);
702 static void show_name(const char *prefix, const char *name, int len,
703 const char *reset, const char *set)
705 printf(" %s%s%-*s%s |", set, prefix, len, name, reset);
708 static void show_graph(char ch, int cnt, const char *set, const char *reset)
710 if (cnt <= 0)
711 return;
712 printf("%s", set);
713 while (cnt--)
714 putchar(ch);
715 printf("%s", reset);
718 static void show_stats(struct diffstat_t* data, struct diff_options *options)
720 int i, len, add, del, total, adds = 0, dels = 0;
721 int max_change = 0, max_len = 0;
722 int total_files = data->nr;
723 int width, name_width;
724 const char *reset, *set, *add_c, *del_c;
726 if (data->nr == 0)
727 return;
729 width = options->stat_width ? options->stat_width : 80;
730 name_width = options->stat_name_width ? options->stat_name_width : 50;
732 /* Sanity: give at least 5 columns to the graph,
733 * but leave at least 10 columns for the name.
735 if (width < name_width + 15) {
736 if (name_width <= 25)
737 width = name_width + 15;
738 else
739 name_width = width - 15;
742 /* Find the longest filename and max number of changes */
743 reset = diff_get_color(options->color_diff, DIFF_RESET);
744 set = diff_get_color(options->color_diff, DIFF_PLAIN);
745 add_c = diff_get_color(options->color_diff, DIFF_FILE_NEW);
746 del_c = diff_get_color(options->color_diff, DIFF_FILE_OLD);
748 for (i = 0; i < data->nr; i++) {
749 struct diffstat_file *file = data->files[i];
750 int change = file->added + file->deleted;
752 if (!file->is_renamed) { /* renames are already quoted by pprint_rename */
753 len = quote_c_style(file->name, NULL, NULL, 0);
754 if (len) {
755 char *qname = xmalloc(len + 1);
756 quote_c_style(file->name, qname, NULL, 0);
757 free(file->name);
758 file->name = qname;
762 len = strlen(file->name);
763 if (max_len < len)
764 max_len = len;
766 if (file->is_binary || file->is_unmerged)
767 continue;
768 if (max_change < change)
769 max_change = change;
772 /* Compute the width of the graph part;
773 * 10 is for one blank at the beginning of the line plus
774 * " | count " between the name and the graph.
776 * From here on, name_width is the width of the name area,
777 * and width is the width of the graph area.
779 name_width = (name_width < max_len) ? name_width : max_len;
780 if (width < (name_width + 10) + max_change)
781 width = width - (name_width + 10);
782 else
783 width = max_change;
785 for (i = 0; i < data->nr; i++) {
786 const char *prefix = "";
787 char *name = data->files[i]->name;
788 int added = data->files[i]->added;
789 int deleted = data->files[i]->deleted;
790 int name_len;
793 * "scale" the filename
795 len = name_width;
796 name_len = strlen(name);
797 if (name_width < name_len) {
798 char *slash;
799 prefix = "...";
800 len -= 3;
801 name += name_len - len;
802 slash = strchr(name, '/');
803 if (slash)
804 name = slash;
807 if (data->files[i]->is_binary) {
808 show_name(prefix, name, len, reset, set);
809 printf(" Bin\n");
810 goto free_diffstat_file;
812 else if (data->files[i]->is_unmerged) {
813 show_name(prefix, name, len, reset, set);
814 printf(" Unmerged\n");
815 goto free_diffstat_file;
817 else if (!data->files[i]->is_renamed &&
818 (added + deleted == 0)) {
819 total_files--;
820 goto free_diffstat_file;
824 * scale the add/delete
826 add = added;
827 del = deleted;
828 total = add + del;
829 adds += add;
830 dels += del;
832 if (width <= max_change) {
833 add = scale_linear(add, width, max_change);
834 del = scale_linear(del, width, max_change);
835 total = add + del;
837 show_name(prefix, name, len, reset, set);
838 printf("%5d ", added + deleted);
839 show_graph('+', add, add_c, reset);
840 show_graph('-', del, del_c, reset);
841 putchar('\n');
842 free_diffstat_file:
843 free(data->files[i]->name);
844 free(data->files[i]);
846 free(data->files);
847 printf("%s %d files changed, %d insertions(+), %d deletions(-)%s\n",
848 set, total_files, adds, dels, reset);
851 static void show_shortstats(struct diffstat_t* data)
853 int i, adds = 0, dels = 0, total_files = data->nr;
855 if (data->nr == 0)
856 return;
858 for (i = 0; i < data->nr; i++) {
859 if (!data->files[i]->is_binary &&
860 !data->files[i]->is_unmerged) {
861 int added = data->files[i]->added;
862 int deleted= data->files[i]->deleted;
863 if (!data->files[i]->is_renamed &&
864 (added + deleted == 0)) {
865 total_files--;
866 } else {
867 adds += added;
868 dels += deleted;
871 free(data->files[i]->name);
872 free(data->files[i]);
874 free(data->files);
876 printf(" %d files changed, %d insertions(+), %d deletions(-)\n",
877 total_files, adds, dels);
880 static void show_numstat(struct diffstat_t* data, struct diff_options *options)
882 int i;
884 for (i = 0; i < data->nr; i++) {
885 struct diffstat_file *file = data->files[i];
887 if (file->is_binary)
888 printf("-\t-\t");
889 else
890 printf("%d\t%d\t", file->added, file->deleted);
891 if (options->line_termination && !file->is_renamed &&
892 quote_c_style(file->name, NULL, NULL, 0))
893 quote_c_style(file->name, NULL, stdout, 0);
894 else
895 fputs(file->name, stdout);
896 putchar(options->line_termination);
900 struct checkdiff_t {
901 struct xdiff_emit_state xm;
902 const char *filename;
903 int lineno, color_diff;
906 static void checkdiff_consume(void *priv, char *line, unsigned long len)
908 struct checkdiff_t *data = priv;
909 const char *ws = diff_get_color(data->color_diff, DIFF_WHITESPACE);
910 const char *reset = diff_get_color(data->color_diff, DIFF_RESET);
911 const char *set = diff_get_color(data->color_diff, DIFF_FILE_NEW);
913 if (line[0] == '+') {
914 int i, spaces = 0, space_before_tab = 0, white_space_at_end = 0;
916 /* check space before tab */
917 for (i = 1; i < len && (line[i] == ' ' || line[i] == '\t'); i++)
918 if (line[i] == ' ')
919 spaces++;
920 if (line[i - 1] == '\t' && spaces)
921 space_before_tab = 1;
923 /* check white space at line end */
924 if (line[len - 1] == '\n')
925 len--;
926 if (isspace(line[len - 1]))
927 white_space_at_end = 1;
929 if (space_before_tab || white_space_at_end) {
930 printf("%s:%d: %s", data->filename, data->lineno, ws);
931 if (space_before_tab) {
932 printf("space before tab");
933 if (white_space_at_end)
934 putchar(',');
936 if (white_space_at_end)
937 printf("white space at end");
938 printf(":%s ", reset);
939 emit_line_with_ws(1, set, reset, ws, line, len);
942 data->lineno++;
943 } else if (line[0] == ' ')
944 data->lineno++;
945 else if (line[0] == '@') {
946 char *plus = strchr(line, '+');
947 if (plus)
948 data->lineno = strtol(plus, NULL, 10);
949 else
950 die("invalid diff");
954 static unsigned char *deflate_it(char *data,
955 unsigned long size,
956 unsigned long *result_size)
958 int bound;
959 unsigned char *deflated;
960 z_stream stream;
962 memset(&stream, 0, sizeof(stream));
963 deflateInit(&stream, zlib_compression_level);
964 bound = deflateBound(&stream, size);
965 deflated = xmalloc(bound);
966 stream.next_out = deflated;
967 stream.avail_out = bound;
969 stream.next_in = (unsigned char *)data;
970 stream.avail_in = size;
971 while (deflate(&stream, Z_FINISH) == Z_OK)
972 ; /* nothing */
973 deflateEnd(&stream);
974 *result_size = stream.total_out;
975 return deflated;
978 static void emit_binary_diff_body(mmfile_t *one, mmfile_t *two)
980 void *cp;
981 void *delta;
982 void *deflated;
983 void *data;
984 unsigned long orig_size;
985 unsigned long delta_size;
986 unsigned long deflate_size;
987 unsigned long data_size;
989 /* We could do deflated delta, or we could do just deflated two,
990 * whichever is smaller.
992 delta = NULL;
993 deflated = deflate_it(two->ptr, two->size, &deflate_size);
994 if (one->size && two->size) {
995 delta = diff_delta(one->ptr, one->size,
996 two->ptr, two->size,
997 &delta_size, deflate_size);
998 if (delta) {
999 void *to_free = delta;
1000 orig_size = delta_size;
1001 delta = deflate_it(delta, delta_size, &delta_size);
1002 free(to_free);
1006 if (delta && delta_size < deflate_size) {
1007 printf("delta %lu\n", orig_size);
1008 free(deflated);
1009 data = delta;
1010 data_size = delta_size;
1012 else {
1013 printf("literal %lu\n", two->size);
1014 free(delta);
1015 data = deflated;
1016 data_size = deflate_size;
1019 /* emit data encoded in base85 */
1020 cp = data;
1021 while (data_size) {
1022 int bytes = (52 < data_size) ? 52 : data_size;
1023 char line[70];
1024 data_size -= bytes;
1025 if (bytes <= 26)
1026 line[0] = bytes + 'A' - 1;
1027 else
1028 line[0] = bytes - 26 + 'a' - 1;
1029 encode_85(line + 1, cp, bytes);
1030 cp = (char *) cp + bytes;
1031 puts(line);
1033 printf("\n");
1034 free(data);
1037 static void emit_binary_diff(mmfile_t *one, mmfile_t *two)
1039 printf("GIT binary patch\n");
1040 emit_binary_diff_body(one, two);
1041 emit_binary_diff_body(two, one);
1044 #define FIRST_FEW_BYTES 8000
1045 static int mmfile_is_binary(mmfile_t *mf)
1047 long sz = mf->size;
1048 if (FIRST_FEW_BYTES < sz)
1049 sz = FIRST_FEW_BYTES;
1050 return !!memchr(mf->ptr, 0, sz);
1053 static void builtin_diff(const char *name_a,
1054 const char *name_b,
1055 struct diff_filespec *one,
1056 struct diff_filespec *two,
1057 const char *xfrm_msg,
1058 struct diff_options *o,
1059 int complete_rewrite)
1061 mmfile_t mf1, mf2;
1062 const char *lbl[2];
1063 char *a_one, *b_two;
1064 const char *set = diff_get_color(o->color_diff, DIFF_METAINFO);
1065 const char *reset = diff_get_color(o->color_diff, DIFF_RESET);
1067 a_one = quote_two("a/", name_a);
1068 b_two = quote_two("b/", name_b);
1069 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
1070 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
1071 printf("%sdiff --git %s %s%s\n", set, a_one, b_two, reset);
1072 if (lbl[0][0] == '/') {
1073 /* /dev/null */
1074 printf("%snew file mode %06o%s\n", set, two->mode, reset);
1075 if (xfrm_msg && xfrm_msg[0])
1076 printf("%s%s%s\n", set, xfrm_msg, reset);
1078 else if (lbl[1][0] == '/') {
1079 printf("%sdeleted file mode %06o%s\n", set, one->mode, reset);
1080 if (xfrm_msg && xfrm_msg[0])
1081 printf("%s%s%s\n", set, xfrm_msg, reset);
1083 else {
1084 if (one->mode != two->mode) {
1085 printf("%sold mode %06o%s\n", set, one->mode, reset);
1086 printf("%snew mode %06o%s\n", set, two->mode, reset);
1088 if (xfrm_msg && xfrm_msg[0])
1089 printf("%s%s%s\n", set, xfrm_msg, reset);
1091 * we do not run diff between different kind
1092 * of objects.
1094 if ((one->mode ^ two->mode) & S_IFMT)
1095 goto free_ab_and_return;
1096 if (complete_rewrite) {
1097 emit_rewrite_diff(name_a, name_b, one, two,
1098 o->color_diff);
1099 goto free_ab_and_return;
1103 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1104 die("unable to read files to diff");
1106 if (!o->text && (mmfile_is_binary(&mf1) || mmfile_is_binary(&mf2))) {
1107 /* Quite common confusing case */
1108 if (mf1.size == mf2.size &&
1109 !memcmp(mf1.ptr, mf2.ptr, mf1.size))
1110 goto free_ab_and_return;
1111 if (o->binary)
1112 emit_binary_diff(&mf1, &mf2);
1113 else
1114 printf("Binary files %s and %s differ\n",
1115 lbl[0], lbl[1]);
1117 else {
1118 /* Crazy xdl interfaces.. */
1119 const char *diffopts = getenv("GIT_DIFF_OPTS");
1120 xpparam_t xpp;
1121 xdemitconf_t xecfg;
1122 xdemitcb_t ecb;
1123 struct emit_callback ecbdata;
1125 memset(&ecbdata, 0, sizeof(ecbdata));
1126 ecbdata.label_path = lbl;
1127 ecbdata.color_diff = o->color_diff;
1128 xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
1129 xecfg.ctxlen = o->context;
1130 xecfg.flags = XDL_EMIT_FUNCNAMES;
1131 if (!diffopts)
1133 else if (!prefixcmp(diffopts, "--unified="))
1134 xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
1135 else if (!prefixcmp(diffopts, "-u"))
1136 xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
1137 ecb.outf = xdiff_outf;
1138 ecb.priv = &ecbdata;
1139 ecbdata.xm.consume = fn_out_consume;
1140 if (o->color_diff_words)
1141 ecbdata.diff_words =
1142 xcalloc(1, sizeof(struct diff_words_data));
1143 xdl_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
1144 if (o->color_diff_words)
1145 free_diff_words_data(&ecbdata);
1148 free_ab_and_return:
1149 free(a_one);
1150 free(b_two);
1151 return;
1154 static void builtin_diffstat(const char *name_a, const char *name_b,
1155 struct diff_filespec *one,
1156 struct diff_filespec *two,
1157 struct diffstat_t *diffstat,
1158 struct diff_options *o,
1159 int complete_rewrite)
1161 mmfile_t mf1, mf2;
1162 struct diffstat_file *data;
1164 data = diffstat_add(diffstat, name_a, name_b);
1166 if (!one || !two) {
1167 data->is_unmerged = 1;
1168 return;
1170 if (complete_rewrite) {
1171 diff_populate_filespec(one, 0);
1172 diff_populate_filespec(two, 0);
1173 data->deleted = count_lines(one->data, one->size);
1174 data->added = count_lines(two->data, two->size);
1175 return;
1177 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1178 die("unable to read files to diff");
1180 if (mmfile_is_binary(&mf1) || mmfile_is_binary(&mf2))
1181 data->is_binary = 1;
1182 else {
1183 /* Crazy xdl interfaces.. */
1184 xpparam_t xpp;
1185 xdemitconf_t xecfg;
1186 xdemitcb_t ecb;
1188 xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
1189 xecfg.ctxlen = 0;
1190 xecfg.flags = 0;
1191 ecb.outf = xdiff_outf;
1192 ecb.priv = diffstat;
1193 xdl_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
1197 static void builtin_checkdiff(const char *name_a, const char *name_b,
1198 struct diff_filespec *one,
1199 struct diff_filespec *two, struct diff_options *o)
1201 mmfile_t mf1, mf2;
1202 struct checkdiff_t data;
1204 if (!two)
1205 return;
1207 memset(&data, 0, sizeof(data));
1208 data.xm.consume = checkdiff_consume;
1209 data.filename = name_b ? name_b : name_a;
1210 data.lineno = 0;
1211 data.color_diff = o->color_diff;
1213 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1214 die("unable to read files to diff");
1216 if (mmfile_is_binary(&mf2))
1217 return;
1218 else {
1219 /* Crazy xdl interfaces.. */
1220 xpparam_t xpp;
1221 xdemitconf_t xecfg;
1222 xdemitcb_t ecb;
1224 xpp.flags = XDF_NEED_MINIMAL;
1225 xecfg.ctxlen = 0;
1226 xecfg.flags = 0;
1227 ecb.outf = xdiff_outf;
1228 ecb.priv = &data;
1229 xdl_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
1233 struct diff_filespec *alloc_filespec(const char *path)
1235 int namelen = strlen(path);
1236 struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
1238 memset(spec, 0, sizeof(*spec));
1239 spec->path = (char *)(spec + 1);
1240 memcpy(spec->path, path, namelen+1);
1241 return spec;
1244 void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
1245 unsigned short mode)
1247 if (mode) {
1248 spec->mode = canon_mode(mode);
1249 hashcpy(spec->sha1, sha1);
1250 spec->sha1_valid = !is_null_sha1(sha1);
1255 * Given a name and sha1 pair, if the dircache tells us the file in
1256 * the work tree has that object contents, return true, so that
1257 * prepare_temp_file() does not have to inflate and extract.
1259 static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
1261 struct cache_entry *ce;
1262 struct stat st;
1263 int pos, len;
1265 /* We do not read the cache ourselves here, because the
1266 * benchmark with my previous version that always reads cache
1267 * shows that it makes things worse for diff-tree comparing
1268 * two linux-2.6 kernel trees in an already checked out work
1269 * tree. This is because most diff-tree comparisons deal with
1270 * only a small number of files, while reading the cache is
1271 * expensive for a large project, and its cost outweighs the
1272 * savings we get by not inflating the object to a temporary
1273 * file. Practically, this code only helps when we are used
1274 * by diff-cache --cached, which does read the cache before
1275 * calling us.
1277 if (!active_cache)
1278 return 0;
1280 /* We want to avoid the working directory if our caller
1281 * doesn't need the data in a normal file, this system
1282 * is rather slow with its stat/open/mmap/close syscalls,
1283 * and the object is contained in a pack file. The pack
1284 * is probably already open and will be faster to obtain
1285 * the data through than the working directory. Loose
1286 * objects however would tend to be slower as they need
1287 * to be individually opened and inflated.
1289 if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1, NULL))
1290 return 0;
1292 len = strlen(name);
1293 pos = cache_name_pos(name, len);
1294 if (pos < 0)
1295 return 0;
1296 ce = active_cache[pos];
1297 if ((lstat(name, &st) < 0) ||
1298 !S_ISREG(st.st_mode) || /* careful! */
1299 ce_match_stat(ce, &st, 0) ||
1300 hashcmp(sha1, ce->sha1))
1301 return 0;
1302 /* we return 1 only when we can stat, it is a regular file,
1303 * stat information matches, and sha1 recorded in the cache
1304 * matches. I.e. we know the file in the work tree really is
1305 * the same as the <name, sha1> pair.
1307 return 1;
1310 static struct sha1_size_cache {
1311 unsigned char sha1[20];
1312 unsigned long size;
1313 } **sha1_size_cache;
1314 static int sha1_size_cache_nr, sha1_size_cache_alloc;
1316 static struct sha1_size_cache *locate_size_cache(unsigned char *sha1,
1317 int find_only,
1318 unsigned long size)
1320 int first, last;
1321 struct sha1_size_cache *e;
1323 first = 0;
1324 last = sha1_size_cache_nr;
1325 while (last > first) {
1326 int cmp, next = (last + first) >> 1;
1327 e = sha1_size_cache[next];
1328 cmp = hashcmp(e->sha1, sha1);
1329 if (!cmp)
1330 return e;
1331 if (cmp < 0) {
1332 last = next;
1333 continue;
1335 first = next+1;
1337 /* not found */
1338 if (find_only)
1339 return NULL;
1340 /* insert to make it at "first" */
1341 if (sha1_size_cache_alloc <= sha1_size_cache_nr) {
1342 sha1_size_cache_alloc = alloc_nr(sha1_size_cache_alloc);
1343 sha1_size_cache = xrealloc(sha1_size_cache,
1344 sha1_size_cache_alloc *
1345 sizeof(*sha1_size_cache));
1347 sha1_size_cache_nr++;
1348 if (first < sha1_size_cache_nr)
1349 memmove(sha1_size_cache + first + 1, sha1_size_cache + first,
1350 (sha1_size_cache_nr - first - 1) *
1351 sizeof(*sha1_size_cache));
1352 e = xmalloc(sizeof(struct sha1_size_cache));
1353 sha1_size_cache[first] = e;
1354 hashcpy(e->sha1, sha1);
1355 e->size = size;
1356 return e;
1360 * While doing rename detection and pickaxe operation, we may need to
1361 * grab the data for the blob (or file) for our own in-core comparison.
1362 * diff_filespec has data and size fields for this purpose.
1364 int diff_populate_filespec(struct diff_filespec *s, int size_only)
1366 int err = 0;
1367 if (!DIFF_FILE_VALID(s))
1368 die("internal error: asking to populate invalid file.");
1369 if (S_ISDIR(s->mode))
1370 return -1;
1372 if (!use_size_cache)
1373 size_only = 0;
1375 if (s->data)
1376 return err;
1377 if (!s->sha1_valid ||
1378 reuse_worktree_file(s->path, s->sha1, 0)) {
1379 struct stat st;
1380 int fd;
1381 if (lstat(s->path, &st) < 0) {
1382 if (errno == ENOENT) {
1383 err_empty:
1384 err = -1;
1385 empty:
1386 s->data = (char *)"";
1387 s->size = 0;
1388 return err;
1391 s->size = st.st_size;
1392 if (!s->size)
1393 goto empty;
1394 if (size_only)
1395 return 0;
1396 if (S_ISLNK(st.st_mode)) {
1397 int ret;
1398 s->data = xmalloc(s->size);
1399 s->should_free = 1;
1400 ret = readlink(s->path, s->data, s->size);
1401 if (ret < 0) {
1402 free(s->data);
1403 goto err_empty;
1405 return 0;
1407 fd = open(s->path, O_RDONLY);
1408 if (fd < 0)
1409 goto err_empty;
1410 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
1411 close(fd);
1412 s->should_munmap = 1;
1413 /* FIXME! CRLF -> LF conversion goes here, based on "s->path" */
1415 else {
1416 char type[20];
1417 struct sha1_size_cache *e;
1419 if (size_only) {
1420 e = locate_size_cache(s->sha1, 1, 0);
1421 if (e) {
1422 s->size = e->size;
1423 return 0;
1425 if (!sha1_object_info(s->sha1, type, &s->size))
1426 locate_size_cache(s->sha1, 0, s->size);
1428 else {
1429 s->data = read_sha1_file(s->sha1, type, &s->size);
1430 s->should_free = 1;
1433 return 0;
1436 void diff_free_filespec_data(struct diff_filespec *s)
1438 if (s->should_free)
1439 free(s->data);
1440 else if (s->should_munmap)
1441 munmap(s->data, s->size);
1442 s->should_free = s->should_munmap = 0;
1443 s->data = NULL;
1444 free(s->cnt_data);
1445 s->cnt_data = NULL;
1448 static void prep_temp_blob(struct diff_tempfile *temp,
1449 void *blob,
1450 unsigned long size,
1451 const unsigned char *sha1,
1452 int mode)
1454 int fd;
1456 fd = git_mkstemp(temp->tmp_path, TEMPFILE_PATH_LEN, ".diff_XXXXXX");
1457 if (fd < 0)
1458 die("unable to create temp-file");
1459 if (write_in_full(fd, blob, size) != size)
1460 die("unable to write temp-file");
1461 close(fd);
1462 temp->name = temp->tmp_path;
1463 strcpy(temp->hex, sha1_to_hex(sha1));
1464 temp->hex[40] = 0;
1465 sprintf(temp->mode, "%06o", mode);
1468 static void prepare_temp_file(const char *name,
1469 struct diff_tempfile *temp,
1470 struct diff_filespec *one)
1472 if (!DIFF_FILE_VALID(one)) {
1473 not_a_valid_file:
1474 /* A '-' entry produces this for file-2, and
1475 * a '+' entry produces this for file-1.
1477 temp->name = "/dev/null";
1478 strcpy(temp->hex, ".");
1479 strcpy(temp->mode, ".");
1480 return;
1483 if (!one->sha1_valid ||
1484 reuse_worktree_file(name, one->sha1, 1)) {
1485 struct stat st;
1486 if (lstat(name, &st) < 0) {
1487 if (errno == ENOENT)
1488 goto not_a_valid_file;
1489 die("stat(%s): %s", name, strerror(errno));
1491 if (S_ISLNK(st.st_mode)) {
1492 int ret;
1493 char buf[PATH_MAX + 1]; /* ought to be SYMLINK_MAX */
1494 if (sizeof(buf) <= st.st_size)
1495 die("symlink too long: %s", name);
1496 ret = readlink(name, buf, st.st_size);
1497 if (ret < 0)
1498 die("readlink(%s)", name);
1499 prep_temp_blob(temp, buf, st.st_size,
1500 (one->sha1_valid ?
1501 one->sha1 : null_sha1),
1502 (one->sha1_valid ?
1503 one->mode : S_IFLNK));
1505 else {
1506 /* we can borrow from the file in the work tree */
1507 temp->name = name;
1508 if (!one->sha1_valid)
1509 strcpy(temp->hex, sha1_to_hex(null_sha1));
1510 else
1511 strcpy(temp->hex, sha1_to_hex(one->sha1));
1512 /* Even though we may sometimes borrow the
1513 * contents from the work tree, we always want
1514 * one->mode. mode is trustworthy even when
1515 * !(one->sha1_valid), as long as
1516 * DIFF_FILE_VALID(one).
1518 sprintf(temp->mode, "%06o", one->mode);
1520 return;
1522 else {
1523 if (diff_populate_filespec(one, 0))
1524 die("cannot read data blob for %s", one->path);
1525 prep_temp_blob(temp, one->data, one->size,
1526 one->sha1, one->mode);
1530 static void remove_tempfile(void)
1532 int i;
1534 for (i = 0; i < 2; i++)
1535 if (diff_temp[i].name == diff_temp[i].tmp_path) {
1536 unlink(diff_temp[i].name);
1537 diff_temp[i].name = NULL;
1541 static void remove_tempfile_on_signal(int signo)
1543 remove_tempfile();
1544 signal(SIGINT, SIG_DFL);
1545 raise(signo);
1548 static int spawn_prog(const char *pgm, const char **arg)
1550 pid_t pid;
1551 int status;
1553 fflush(NULL);
1554 pid = fork();
1555 if (pid < 0)
1556 die("unable to fork");
1557 if (!pid) {
1558 execvp(pgm, (char *const*) arg);
1559 exit(255);
1562 while (waitpid(pid, &status, 0) < 0) {
1563 if (errno == EINTR)
1564 continue;
1565 return -1;
1568 /* Earlier we did not check the exit status because
1569 * diff exits non-zero if files are different, and
1570 * we are not interested in knowing that. It was a
1571 * mistake which made it harder to quit a diff-*
1572 * session that uses the git-apply-patch-script as
1573 * the GIT_EXTERNAL_DIFF. A custom GIT_EXTERNAL_DIFF
1574 * should also exit non-zero only when it wants to
1575 * abort the entire diff-* session.
1577 if (WIFEXITED(status) && !WEXITSTATUS(status))
1578 return 0;
1579 return -1;
1582 /* An external diff command takes:
1584 * diff-cmd name infile1 infile1-sha1 infile1-mode \
1585 * infile2 infile2-sha1 infile2-mode [ rename-to ]
1588 static void run_external_diff(const char *pgm,
1589 const char *name,
1590 const char *other,
1591 struct diff_filespec *one,
1592 struct diff_filespec *two,
1593 const char *xfrm_msg,
1594 int complete_rewrite)
1596 const char *spawn_arg[10];
1597 struct diff_tempfile *temp = diff_temp;
1598 int retval;
1599 static int atexit_asked = 0;
1600 const char *othername;
1601 const char **arg = &spawn_arg[0];
1603 othername = (other? other : name);
1604 if (one && two) {
1605 prepare_temp_file(name, &temp[0], one);
1606 prepare_temp_file(othername, &temp[1], two);
1607 if (! atexit_asked &&
1608 (temp[0].name == temp[0].tmp_path ||
1609 temp[1].name == temp[1].tmp_path)) {
1610 atexit_asked = 1;
1611 atexit(remove_tempfile);
1613 signal(SIGINT, remove_tempfile_on_signal);
1616 if (one && two) {
1617 *arg++ = pgm;
1618 *arg++ = name;
1619 *arg++ = temp[0].name;
1620 *arg++ = temp[0].hex;
1621 *arg++ = temp[0].mode;
1622 *arg++ = temp[1].name;
1623 *arg++ = temp[1].hex;
1624 *arg++ = temp[1].mode;
1625 if (other) {
1626 *arg++ = other;
1627 *arg++ = xfrm_msg;
1629 } else {
1630 *arg++ = pgm;
1631 *arg++ = name;
1633 *arg = NULL;
1634 retval = spawn_prog(pgm, spawn_arg);
1635 remove_tempfile();
1636 if (retval) {
1637 fprintf(stderr, "external diff died, stopping at %s.\n", name);
1638 exit(1);
1642 static void run_diff_cmd(const char *pgm,
1643 const char *name,
1644 const char *other,
1645 struct diff_filespec *one,
1646 struct diff_filespec *two,
1647 const char *xfrm_msg,
1648 struct diff_options *o,
1649 int complete_rewrite)
1651 if (pgm) {
1652 run_external_diff(pgm, name, other, one, two, xfrm_msg,
1653 complete_rewrite);
1654 return;
1656 if (one && two)
1657 builtin_diff(name, other ? other : name,
1658 one, two, xfrm_msg, o, complete_rewrite);
1659 else
1660 printf("* Unmerged path %s\n", name);
1663 static void diff_fill_sha1_info(struct diff_filespec *one)
1665 if (DIFF_FILE_VALID(one)) {
1666 if (!one->sha1_valid) {
1667 struct stat st;
1668 if (lstat(one->path, &st) < 0)
1669 die("stat %s", one->path);
1670 if (index_path(one->sha1, one->path, &st, 0))
1671 die("cannot hash %s\n", one->path);
1674 else
1675 hashclr(one->sha1);
1678 static void run_diff(struct diff_filepair *p, struct diff_options *o)
1680 const char *pgm = external_diff();
1681 char msg[PATH_MAX*2+300], *xfrm_msg;
1682 struct diff_filespec *one;
1683 struct diff_filespec *two;
1684 const char *name;
1685 const char *other;
1686 char *name_munged, *other_munged;
1687 int complete_rewrite = 0;
1688 int len;
1690 if (DIFF_PAIR_UNMERGED(p)) {
1691 /* unmerged */
1692 run_diff_cmd(pgm, p->one->path, NULL, NULL, NULL, NULL, o, 0);
1693 return;
1696 name = p->one->path;
1697 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
1698 name_munged = quote_one(name);
1699 other_munged = quote_one(other);
1700 one = p->one; two = p->two;
1702 diff_fill_sha1_info(one);
1703 diff_fill_sha1_info(two);
1705 len = 0;
1706 switch (p->status) {
1707 case DIFF_STATUS_COPIED:
1708 len += snprintf(msg + len, sizeof(msg) - len,
1709 "similarity index %d%%\n"
1710 "copy from %s\n"
1711 "copy to %s\n",
1712 (int)(0.5 + p->score * 100.0/MAX_SCORE),
1713 name_munged, other_munged);
1714 break;
1715 case DIFF_STATUS_RENAMED:
1716 len += snprintf(msg + len, sizeof(msg) - len,
1717 "similarity index %d%%\n"
1718 "rename from %s\n"
1719 "rename to %s\n",
1720 (int)(0.5 + p->score * 100.0/MAX_SCORE),
1721 name_munged, other_munged);
1722 break;
1723 case DIFF_STATUS_MODIFIED:
1724 if (p->score) {
1725 len += snprintf(msg + len, sizeof(msg) - len,
1726 "dissimilarity index %d%%\n",
1727 (int)(0.5 + p->score *
1728 100.0/MAX_SCORE));
1729 complete_rewrite = 1;
1730 break;
1732 /* fallthru */
1733 default:
1734 /* nothing */
1738 if (hashcmp(one->sha1, two->sha1)) {
1739 int abbrev = o->full_index ? 40 : DEFAULT_ABBREV;
1741 if (o->binary) {
1742 mmfile_t mf;
1743 if ((!fill_mmfile(&mf, one) && mmfile_is_binary(&mf)) ||
1744 (!fill_mmfile(&mf, two) && mmfile_is_binary(&mf)))
1745 abbrev = 40;
1747 len += snprintf(msg + len, sizeof(msg) - len,
1748 "index %.*s..%.*s",
1749 abbrev, sha1_to_hex(one->sha1),
1750 abbrev, sha1_to_hex(two->sha1));
1751 if (one->mode == two->mode)
1752 len += snprintf(msg + len, sizeof(msg) - len,
1753 " %06o", one->mode);
1754 len += snprintf(msg + len, sizeof(msg) - len, "\n");
1757 if (len)
1758 msg[--len] = 0;
1759 xfrm_msg = len ? msg : NULL;
1761 if (!pgm &&
1762 DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
1763 (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
1764 /* a filepair that changes between file and symlink
1765 * needs to be split into deletion and creation.
1767 struct diff_filespec *null = alloc_filespec(two->path);
1768 run_diff_cmd(NULL, name, other, one, null, xfrm_msg, o, 0);
1769 free(null);
1770 null = alloc_filespec(one->path);
1771 run_diff_cmd(NULL, name, other, null, two, xfrm_msg, o, 0);
1772 free(null);
1774 else
1775 run_diff_cmd(pgm, name, other, one, two, xfrm_msg, o,
1776 complete_rewrite);
1778 free(name_munged);
1779 free(other_munged);
1782 static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
1783 struct diffstat_t *diffstat)
1785 const char *name;
1786 const char *other;
1787 int complete_rewrite = 0;
1789 if (DIFF_PAIR_UNMERGED(p)) {
1790 /* unmerged */
1791 builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, 0);
1792 return;
1795 name = p->one->path;
1796 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
1798 diff_fill_sha1_info(p->one);
1799 diff_fill_sha1_info(p->two);
1801 if (p->status == DIFF_STATUS_MODIFIED && p->score)
1802 complete_rewrite = 1;
1803 builtin_diffstat(name, other, p->one, p->two, diffstat, o, complete_rewrite);
1806 static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
1808 const char *name;
1809 const char *other;
1811 if (DIFF_PAIR_UNMERGED(p)) {
1812 /* unmerged */
1813 return;
1816 name = p->one->path;
1817 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
1819 diff_fill_sha1_info(p->one);
1820 diff_fill_sha1_info(p->two);
1822 builtin_checkdiff(name, other, p->one, p->two, o);
1825 void diff_setup(struct diff_options *options)
1827 memset(options, 0, sizeof(*options));
1828 options->line_termination = '\n';
1829 options->break_opt = -1;
1830 options->rename_limit = -1;
1831 options->context = 3;
1832 options->msg_sep = "";
1834 options->change = diff_change;
1835 options->add_remove = diff_addremove;
1836 options->color_diff = diff_use_color_default;
1837 options->detect_rename = diff_detect_rename_default;
1840 int diff_setup_done(struct diff_options *options)
1842 int count = 0;
1844 if (options->output_format & DIFF_FORMAT_NAME)
1845 count++;
1846 if (options->output_format & DIFF_FORMAT_NAME_STATUS)
1847 count++;
1848 if (options->output_format & DIFF_FORMAT_CHECKDIFF)
1849 count++;
1850 if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
1851 count++;
1852 if (count > 1)
1853 die("--name-only, --name-status, --check and -s are mutually exclusive");
1855 if (options->find_copies_harder)
1856 options->detect_rename = DIFF_DETECT_COPY;
1858 if (options->output_format & (DIFF_FORMAT_NAME |
1859 DIFF_FORMAT_NAME_STATUS |
1860 DIFF_FORMAT_CHECKDIFF |
1861 DIFF_FORMAT_NO_OUTPUT))
1862 options->output_format &= ~(DIFF_FORMAT_RAW |
1863 DIFF_FORMAT_NUMSTAT |
1864 DIFF_FORMAT_DIFFSTAT |
1865 DIFF_FORMAT_SHORTSTAT |
1866 DIFF_FORMAT_SUMMARY |
1867 DIFF_FORMAT_PATCH);
1870 * These cases always need recursive; we do not drop caller-supplied
1871 * recursive bits for other formats here.
1873 if (options->output_format & (DIFF_FORMAT_PATCH |
1874 DIFF_FORMAT_NUMSTAT |
1875 DIFF_FORMAT_DIFFSTAT |
1876 DIFF_FORMAT_SHORTSTAT |
1877 DIFF_FORMAT_SUMMARY |
1878 DIFF_FORMAT_CHECKDIFF))
1879 options->recursive = 1;
1881 * Also pickaxe would not work very well if you do not say recursive
1883 if (options->pickaxe)
1884 options->recursive = 1;
1886 if (options->detect_rename && options->rename_limit < 0)
1887 options->rename_limit = diff_rename_limit_default;
1888 if (options->setup & DIFF_SETUP_USE_CACHE) {
1889 if (!active_cache)
1890 /* read-cache does not die even when it fails
1891 * so it is safe for us to do this here. Also
1892 * it does not smudge active_cache or active_nr
1893 * when it fails, so we do not have to worry about
1894 * cleaning it up ourselves either.
1896 read_cache();
1898 if (options->setup & DIFF_SETUP_USE_SIZE_CACHE)
1899 use_size_cache = 1;
1900 if (options->abbrev <= 0 || 40 < options->abbrev)
1901 options->abbrev = 40; /* full */
1903 return 0;
1906 static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
1908 char c, *eq;
1909 int len;
1911 if (*arg != '-')
1912 return 0;
1913 c = *++arg;
1914 if (!c)
1915 return 0;
1916 if (c == arg_short) {
1917 c = *++arg;
1918 if (!c)
1919 return 1;
1920 if (val && isdigit(c)) {
1921 char *end;
1922 int n = strtoul(arg, &end, 10);
1923 if (*end)
1924 return 0;
1925 *val = n;
1926 return 1;
1928 return 0;
1930 if (c != '-')
1931 return 0;
1932 arg++;
1933 eq = strchr(arg, '=');
1934 if (eq)
1935 len = eq - arg;
1936 else
1937 len = strlen(arg);
1938 if (!len || strncmp(arg, arg_long, len))
1939 return 0;
1940 if (eq) {
1941 int n;
1942 char *end;
1943 if (!isdigit(*++eq))
1944 return 0;
1945 n = strtoul(eq, &end, 10);
1946 if (*end)
1947 return 0;
1948 *val = n;
1950 return 1;
1953 int diff_opt_parse(struct diff_options *options, const char **av, int ac)
1955 const char *arg = av[0];
1956 if (!strcmp(arg, "-p") || !strcmp(arg, "-u"))
1957 options->output_format |= DIFF_FORMAT_PATCH;
1958 else if (opt_arg(arg, 'U', "unified", &options->context))
1959 options->output_format |= DIFF_FORMAT_PATCH;
1960 else if (!strcmp(arg, "--raw"))
1961 options->output_format |= DIFF_FORMAT_RAW;
1962 else if (!strcmp(arg, "--patch-with-raw")) {
1963 options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_RAW;
1965 else if (!strcmp(arg, "--numstat")) {
1966 options->output_format |= DIFF_FORMAT_NUMSTAT;
1968 else if (!strcmp(arg, "--shortstat")) {
1969 options->output_format |= DIFF_FORMAT_SHORTSTAT;
1971 else if (!prefixcmp(arg, "--stat")) {
1972 char *end;
1973 int width = options->stat_width;
1974 int name_width = options->stat_name_width;
1975 arg += 6;
1976 end = (char *)arg;
1978 switch (*arg) {
1979 case '-':
1980 if (!prefixcmp(arg, "-width="))
1981 width = strtoul(arg + 7, &end, 10);
1982 else if (!prefixcmp(arg, "-name-width="))
1983 name_width = strtoul(arg + 12, &end, 10);
1984 break;
1985 case '=':
1986 width = strtoul(arg+1, &end, 10);
1987 if (*end == ',')
1988 name_width = strtoul(end+1, &end, 10);
1991 /* Important! This checks all the error cases! */
1992 if (*end)
1993 return 0;
1994 options->output_format |= DIFF_FORMAT_DIFFSTAT;
1995 options->stat_name_width = name_width;
1996 options->stat_width = width;
1998 else if (!strcmp(arg, "--check"))
1999 options->output_format |= DIFF_FORMAT_CHECKDIFF;
2000 else if (!strcmp(arg, "--summary"))
2001 options->output_format |= DIFF_FORMAT_SUMMARY;
2002 else if (!strcmp(arg, "--patch-with-stat")) {
2003 options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_DIFFSTAT;
2005 else if (!strcmp(arg, "-z"))
2006 options->line_termination = 0;
2007 else if (!prefixcmp(arg, "-l"))
2008 options->rename_limit = strtoul(arg+2, NULL, 10);
2009 else if (!strcmp(arg, "--full-index"))
2010 options->full_index = 1;
2011 else if (!strcmp(arg, "--binary")) {
2012 options->output_format |= DIFF_FORMAT_PATCH;
2013 options->binary = 1;
2015 else if (!strcmp(arg, "-a") || !strcmp(arg, "--text")) {
2016 options->text = 1;
2018 else if (!strcmp(arg, "--name-only"))
2019 options->output_format |= DIFF_FORMAT_NAME;
2020 else if (!strcmp(arg, "--name-status"))
2021 options->output_format |= DIFF_FORMAT_NAME_STATUS;
2022 else if (!strcmp(arg, "-R"))
2023 options->reverse_diff = 1;
2024 else if (!prefixcmp(arg, "-S"))
2025 options->pickaxe = arg + 2;
2026 else if (!strcmp(arg, "-s")) {
2027 options->output_format |= DIFF_FORMAT_NO_OUTPUT;
2029 else if (!prefixcmp(arg, "-O"))
2030 options->orderfile = arg + 2;
2031 else if (!prefixcmp(arg, "--diff-filter="))
2032 options->filter = arg + 14;
2033 else if (!strcmp(arg, "--pickaxe-all"))
2034 options->pickaxe_opts = DIFF_PICKAXE_ALL;
2035 else if (!strcmp(arg, "--pickaxe-regex"))
2036 options->pickaxe_opts = DIFF_PICKAXE_REGEX;
2037 else if (!prefixcmp(arg, "-B")) {
2038 if ((options->break_opt =
2039 diff_scoreopt_parse(arg)) == -1)
2040 return -1;
2042 else if (!prefixcmp(arg, "-M")) {
2043 if ((options->rename_score =
2044 diff_scoreopt_parse(arg)) == -1)
2045 return -1;
2046 options->detect_rename = DIFF_DETECT_RENAME;
2048 else if (!prefixcmp(arg, "-C")) {
2049 if ((options->rename_score =
2050 diff_scoreopt_parse(arg)) == -1)
2051 return -1;
2052 options->detect_rename = DIFF_DETECT_COPY;
2054 else if (!strcmp(arg, "--find-copies-harder"))
2055 options->find_copies_harder = 1;
2056 else if (!strcmp(arg, "--abbrev"))
2057 options->abbrev = DEFAULT_ABBREV;
2058 else if (!prefixcmp(arg, "--abbrev=")) {
2059 options->abbrev = strtoul(arg + 9, NULL, 10);
2060 if (options->abbrev < MINIMUM_ABBREV)
2061 options->abbrev = MINIMUM_ABBREV;
2062 else if (40 < options->abbrev)
2063 options->abbrev = 40;
2065 else if (!strcmp(arg, "--color"))
2066 options->color_diff = 1;
2067 else if (!strcmp(arg, "--no-color"))
2068 options->color_diff = 0;
2069 else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
2070 options->xdl_opts |= XDF_IGNORE_WHITESPACE;
2071 else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
2072 options->xdl_opts |= XDF_IGNORE_WHITESPACE_CHANGE;
2073 else if (!strcmp(arg, "--ignore-space-at-eol"))
2074 options->xdl_opts |= XDF_IGNORE_WHITESPACE_AT_EOL;
2075 else if (!strcmp(arg, "--color-words"))
2076 options->color_diff = options->color_diff_words = 1;
2077 else if (!strcmp(arg, "--no-renames"))
2078 options->detect_rename = 0;
2079 else
2080 return 0;
2081 return 1;
2084 static int parse_num(const char **cp_p)
2086 unsigned long num, scale;
2087 int ch, dot;
2088 const char *cp = *cp_p;
2090 num = 0;
2091 scale = 1;
2092 dot = 0;
2093 for(;;) {
2094 ch = *cp;
2095 if ( !dot && ch == '.' ) {
2096 scale = 1;
2097 dot = 1;
2098 } else if ( ch == '%' ) {
2099 scale = dot ? scale*100 : 100;
2100 cp++; /* % is always at the end */
2101 break;
2102 } else if ( ch >= '0' && ch <= '9' ) {
2103 if ( scale < 100000 ) {
2104 scale *= 10;
2105 num = (num*10) + (ch-'0');
2107 } else {
2108 break;
2110 cp++;
2112 *cp_p = cp;
2114 /* user says num divided by scale and we say internally that
2115 * is MAX_SCORE * num / scale.
2117 return (num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale);
2120 int diff_scoreopt_parse(const char *opt)
2122 int opt1, opt2, cmd;
2124 if (*opt++ != '-')
2125 return -1;
2126 cmd = *opt++;
2127 if (cmd != 'M' && cmd != 'C' && cmd != 'B')
2128 return -1; /* that is not a -M, -C nor -B option */
2130 opt1 = parse_num(&opt);
2131 if (cmd != 'B')
2132 opt2 = 0;
2133 else {
2134 if (*opt == 0)
2135 opt2 = 0;
2136 else if (*opt != '/')
2137 return -1; /* we expect -B80/99 or -B80 */
2138 else {
2139 opt++;
2140 opt2 = parse_num(&opt);
2143 if (*opt != 0)
2144 return -1;
2145 return opt1 | (opt2 << 16);
2148 struct diff_queue_struct diff_queued_diff;
2150 void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
2152 if (queue->alloc <= queue->nr) {
2153 queue->alloc = alloc_nr(queue->alloc);
2154 queue->queue = xrealloc(queue->queue,
2155 sizeof(dp) * queue->alloc);
2157 queue->queue[queue->nr++] = dp;
2160 struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
2161 struct diff_filespec *one,
2162 struct diff_filespec *two)
2164 struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
2165 dp->one = one;
2166 dp->two = two;
2167 if (queue)
2168 diff_q(queue, dp);
2169 return dp;
2172 void diff_free_filepair(struct diff_filepair *p)
2174 diff_free_filespec_data(p->one);
2175 diff_free_filespec_data(p->two);
2176 free(p->one);
2177 free(p->two);
2178 free(p);
2181 /* This is different from find_unique_abbrev() in that
2182 * it stuffs the result with dots for alignment.
2184 const char *diff_unique_abbrev(const unsigned char *sha1, int len)
2186 int abblen;
2187 const char *abbrev;
2188 if (len == 40)
2189 return sha1_to_hex(sha1);
2191 abbrev = find_unique_abbrev(sha1, len);
2192 if (!abbrev)
2193 return sha1_to_hex(sha1);
2194 abblen = strlen(abbrev);
2195 if (abblen < 37) {
2196 static char hex[41];
2197 if (len < abblen && abblen <= len + 2)
2198 sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
2199 else
2200 sprintf(hex, "%s...", abbrev);
2201 return hex;
2203 return sha1_to_hex(sha1);
2206 static void diff_flush_raw(struct diff_filepair *p,
2207 struct diff_options *options)
2209 int two_paths;
2210 char status[10];
2211 int abbrev = options->abbrev;
2212 const char *path_one, *path_two;
2213 int inter_name_termination = '\t';
2214 int line_termination = options->line_termination;
2216 if (!line_termination)
2217 inter_name_termination = 0;
2219 path_one = p->one->path;
2220 path_two = p->two->path;
2221 if (line_termination) {
2222 path_one = quote_one(path_one);
2223 path_two = quote_one(path_two);
2226 if (p->score)
2227 sprintf(status, "%c%03d", p->status,
2228 (int)(0.5 + p->score * 100.0/MAX_SCORE));
2229 else {
2230 status[0] = p->status;
2231 status[1] = 0;
2233 switch (p->status) {
2234 case DIFF_STATUS_COPIED:
2235 case DIFF_STATUS_RENAMED:
2236 two_paths = 1;
2237 break;
2238 case DIFF_STATUS_ADDED:
2239 case DIFF_STATUS_DELETED:
2240 two_paths = 0;
2241 break;
2242 default:
2243 two_paths = 0;
2244 break;
2246 if (!(options->output_format & DIFF_FORMAT_NAME_STATUS)) {
2247 printf(":%06o %06o %s ",
2248 p->one->mode, p->two->mode,
2249 diff_unique_abbrev(p->one->sha1, abbrev));
2250 printf("%s ",
2251 diff_unique_abbrev(p->two->sha1, abbrev));
2253 printf("%s%c%s", status, inter_name_termination, path_one);
2254 if (two_paths)
2255 printf("%c%s", inter_name_termination, path_two);
2256 putchar(line_termination);
2257 if (path_one != p->one->path)
2258 free((void*)path_one);
2259 if (path_two != p->two->path)
2260 free((void*)path_two);
2263 static void diff_flush_name(struct diff_filepair *p, struct diff_options *opt)
2265 char *path = p->two->path;
2267 if (opt->line_termination)
2268 path = quote_one(p->two->path);
2269 printf("%s%c", path, opt->line_termination);
2270 if (p->two->path != path)
2271 free(path);
2274 int diff_unmodified_pair(struct diff_filepair *p)
2276 /* This function is written stricter than necessary to support
2277 * the currently implemented transformers, but the idea is to
2278 * let transformers to produce diff_filepairs any way they want,
2279 * and filter and clean them up here before producing the output.
2281 struct diff_filespec *one, *two;
2283 if (DIFF_PAIR_UNMERGED(p))
2284 return 0; /* unmerged is interesting */
2286 one = p->one;
2287 two = p->two;
2289 /* deletion, addition, mode or type change
2290 * and rename are all interesting.
2292 if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
2293 DIFF_PAIR_MODE_CHANGED(p) ||
2294 strcmp(one->path, two->path))
2295 return 0;
2297 /* both are valid and point at the same path. that is, we are
2298 * dealing with a change.
2300 if (one->sha1_valid && two->sha1_valid &&
2301 !hashcmp(one->sha1, two->sha1))
2302 return 1; /* no change */
2303 if (!one->sha1_valid && !two->sha1_valid)
2304 return 1; /* both look at the same file on the filesystem. */
2305 return 0;
2308 static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
2310 if (diff_unmodified_pair(p))
2311 return;
2313 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2314 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2315 return; /* no tree diffs in patch format */
2317 run_diff(p, o);
2320 static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
2321 struct diffstat_t *diffstat)
2323 if (diff_unmodified_pair(p))
2324 return;
2326 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2327 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2328 return; /* no tree diffs in patch format */
2330 run_diffstat(p, o, diffstat);
2333 static void diff_flush_checkdiff(struct diff_filepair *p,
2334 struct diff_options *o)
2336 if (diff_unmodified_pair(p))
2337 return;
2339 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2340 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2341 return; /* no tree diffs in patch format */
2343 run_checkdiff(p, o);
2346 int diff_queue_is_empty(void)
2348 struct diff_queue_struct *q = &diff_queued_diff;
2349 int i;
2350 for (i = 0; i < q->nr; i++)
2351 if (!diff_unmodified_pair(q->queue[i]))
2352 return 0;
2353 return 1;
2356 #if DIFF_DEBUG
2357 void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
2359 fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
2360 x, one ? one : "",
2361 s->path,
2362 DIFF_FILE_VALID(s) ? "valid" : "invalid",
2363 s->mode,
2364 s->sha1_valid ? sha1_to_hex(s->sha1) : "");
2365 fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
2366 x, one ? one : "",
2367 s->size, s->xfrm_flags);
2370 void diff_debug_filepair(const struct diff_filepair *p, int i)
2372 diff_debug_filespec(p->one, i, "one");
2373 diff_debug_filespec(p->two, i, "two");
2374 fprintf(stderr, "score %d, status %c stays %d broken %d\n",
2375 p->score, p->status ? p->status : '?',
2376 p->source_stays, p->broken_pair);
2379 void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
2381 int i;
2382 if (msg)
2383 fprintf(stderr, "%s\n", msg);
2384 fprintf(stderr, "q->nr = %d\n", q->nr);
2385 for (i = 0; i < q->nr; i++) {
2386 struct diff_filepair *p = q->queue[i];
2387 diff_debug_filepair(p, i);
2390 #endif
2392 static void diff_resolve_rename_copy(void)
2394 int i, j;
2395 struct diff_filepair *p, *pp;
2396 struct diff_queue_struct *q = &diff_queued_diff;
2398 diff_debug_queue("resolve-rename-copy", q);
2400 for (i = 0; i < q->nr; i++) {
2401 p = q->queue[i];
2402 p->status = 0; /* undecided */
2403 if (DIFF_PAIR_UNMERGED(p))
2404 p->status = DIFF_STATUS_UNMERGED;
2405 else if (!DIFF_FILE_VALID(p->one))
2406 p->status = DIFF_STATUS_ADDED;
2407 else if (!DIFF_FILE_VALID(p->two))
2408 p->status = DIFF_STATUS_DELETED;
2409 else if (DIFF_PAIR_TYPE_CHANGED(p))
2410 p->status = DIFF_STATUS_TYPE_CHANGED;
2412 /* from this point on, we are dealing with a pair
2413 * whose both sides are valid and of the same type, i.e.
2414 * either in-place edit or rename/copy edit.
2416 else if (DIFF_PAIR_RENAME(p)) {
2417 if (p->source_stays) {
2418 p->status = DIFF_STATUS_COPIED;
2419 continue;
2421 /* See if there is some other filepair that
2422 * copies from the same source as us. If so
2423 * we are a copy. Otherwise we are either a
2424 * copy if the path stays, or a rename if it
2425 * does not, but we already handled "stays" case.
2427 for (j = i + 1; j < q->nr; j++) {
2428 pp = q->queue[j];
2429 if (strcmp(pp->one->path, p->one->path))
2430 continue; /* not us */
2431 if (!DIFF_PAIR_RENAME(pp))
2432 continue; /* not a rename/copy */
2433 /* pp is a rename/copy from the same source */
2434 p->status = DIFF_STATUS_COPIED;
2435 break;
2437 if (!p->status)
2438 p->status = DIFF_STATUS_RENAMED;
2440 else if (hashcmp(p->one->sha1, p->two->sha1) ||
2441 p->one->mode != p->two->mode)
2442 p->status = DIFF_STATUS_MODIFIED;
2443 else {
2444 /* This is a "no-change" entry and should not
2445 * happen anymore, but prepare for broken callers.
2447 error("feeding unmodified %s to diffcore",
2448 p->one->path);
2449 p->status = DIFF_STATUS_UNKNOWN;
2452 diff_debug_queue("resolve-rename-copy done", q);
2455 static int check_pair_status(struct diff_filepair *p)
2457 switch (p->status) {
2458 case DIFF_STATUS_UNKNOWN:
2459 return 0;
2460 case 0:
2461 die("internal error in diff-resolve-rename-copy");
2462 default:
2463 return 1;
2467 static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
2469 int fmt = opt->output_format;
2471 if (fmt & DIFF_FORMAT_CHECKDIFF)
2472 diff_flush_checkdiff(p, opt);
2473 else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
2474 diff_flush_raw(p, opt);
2475 else if (fmt & DIFF_FORMAT_NAME)
2476 diff_flush_name(p, opt);
2479 static void show_file_mode_name(const char *newdelete, struct diff_filespec *fs)
2481 char *name = quote_one(fs->path);
2482 if (fs->mode)
2483 printf(" %s mode %06o %s\n", newdelete, fs->mode, name);
2484 else
2485 printf(" %s %s\n", newdelete, name);
2486 free(name);
2490 static void show_mode_change(struct diff_filepair *p, int show_name)
2492 if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
2493 if (show_name) {
2494 char *name = quote_one(p->two->path);
2495 printf(" mode change %06o => %06o %s\n",
2496 p->one->mode, p->two->mode, name);
2497 free(name);
2499 else
2500 printf(" mode change %06o => %06o\n",
2501 p->one->mode, p->two->mode);
2505 static void show_rename_copy(const char *renamecopy, struct diff_filepair *p)
2507 char *names = pprint_rename(p->one->path, p->two->path);
2509 printf(" %s %s (%d%%)\n", renamecopy, names,
2510 (int)(0.5 + p->score * 100.0/MAX_SCORE));
2511 free(names);
2512 show_mode_change(p, 0);
2515 static void diff_summary(struct diff_filepair *p)
2517 switch(p->status) {
2518 case DIFF_STATUS_DELETED:
2519 show_file_mode_name("delete", p->one);
2520 break;
2521 case DIFF_STATUS_ADDED:
2522 show_file_mode_name("create", p->two);
2523 break;
2524 case DIFF_STATUS_COPIED:
2525 show_rename_copy("copy", p);
2526 break;
2527 case DIFF_STATUS_RENAMED:
2528 show_rename_copy("rename", p);
2529 break;
2530 default:
2531 if (p->score) {
2532 char *name = quote_one(p->two->path);
2533 printf(" rewrite %s (%d%%)\n", name,
2534 (int)(0.5 + p->score * 100.0/MAX_SCORE));
2535 free(name);
2536 show_mode_change(p, 0);
2537 } else show_mode_change(p, 1);
2538 break;
2542 struct patch_id_t {
2543 struct xdiff_emit_state xm;
2544 SHA_CTX *ctx;
2545 int patchlen;
2548 static int remove_space(char *line, int len)
2550 int i;
2551 char *dst = line;
2552 unsigned char c;
2554 for (i = 0; i < len; i++)
2555 if (!isspace((c = line[i])))
2556 *dst++ = c;
2558 return dst - line;
2561 static void patch_id_consume(void *priv, char *line, unsigned long len)
2563 struct patch_id_t *data = priv;
2564 int new_len;
2566 /* Ignore line numbers when computing the SHA1 of the patch */
2567 if (!prefixcmp(line, "@@ -"))
2568 return;
2570 new_len = remove_space(line, len);
2572 SHA1_Update(data->ctx, line, new_len);
2573 data->patchlen += new_len;
2576 /* returns 0 upon success, and writes result into sha1 */
2577 static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
2579 struct diff_queue_struct *q = &diff_queued_diff;
2580 int i;
2581 SHA_CTX ctx;
2582 struct patch_id_t data;
2583 char buffer[PATH_MAX * 4 + 20];
2585 SHA1_Init(&ctx);
2586 memset(&data, 0, sizeof(struct patch_id_t));
2587 data.ctx = &ctx;
2588 data.xm.consume = patch_id_consume;
2590 for (i = 0; i < q->nr; i++) {
2591 xpparam_t xpp;
2592 xdemitconf_t xecfg;
2593 xdemitcb_t ecb;
2594 mmfile_t mf1, mf2;
2595 struct diff_filepair *p = q->queue[i];
2596 int len1, len2;
2598 if (p->status == 0)
2599 return error("internal diff status error");
2600 if (p->status == DIFF_STATUS_UNKNOWN)
2601 continue;
2602 if (diff_unmodified_pair(p))
2603 continue;
2604 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2605 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2606 continue;
2607 if (DIFF_PAIR_UNMERGED(p))
2608 continue;
2610 diff_fill_sha1_info(p->one);
2611 diff_fill_sha1_info(p->two);
2612 if (fill_mmfile(&mf1, p->one) < 0 ||
2613 fill_mmfile(&mf2, p->two) < 0)
2614 return error("unable to read files to diff");
2616 /* Maybe hash p->two? into the patch id? */
2617 if (mmfile_is_binary(&mf2))
2618 continue;
2620 len1 = remove_space(p->one->path, strlen(p->one->path));
2621 len2 = remove_space(p->two->path, strlen(p->two->path));
2622 if (p->one->mode == 0)
2623 len1 = snprintf(buffer, sizeof(buffer),
2624 "diff--gita/%.*sb/%.*s"
2625 "newfilemode%06o"
2626 "---/dev/null"
2627 "+++b/%.*s",
2628 len1, p->one->path,
2629 len2, p->two->path,
2630 p->two->mode,
2631 len2, p->two->path);
2632 else if (p->two->mode == 0)
2633 len1 = snprintf(buffer, sizeof(buffer),
2634 "diff--gita/%.*sb/%.*s"
2635 "deletedfilemode%06o"
2636 "---a/%.*s"
2637 "+++/dev/null",
2638 len1, p->one->path,
2639 len2, p->two->path,
2640 p->one->mode,
2641 len1, p->one->path);
2642 else
2643 len1 = snprintf(buffer, sizeof(buffer),
2644 "diff--gita/%.*sb/%.*s"
2645 "---a/%.*s"
2646 "+++b/%.*s",
2647 len1, p->one->path,
2648 len2, p->two->path,
2649 len1, p->one->path,
2650 len2, p->two->path);
2651 SHA1_Update(&ctx, buffer, len1);
2653 xpp.flags = XDF_NEED_MINIMAL;
2654 xecfg.ctxlen = 3;
2655 xecfg.flags = XDL_EMIT_FUNCNAMES;
2656 ecb.outf = xdiff_outf;
2657 ecb.priv = &data;
2658 xdl_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
2661 SHA1_Final(sha1, &ctx);
2662 return 0;
2665 int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
2667 struct diff_queue_struct *q = &diff_queued_diff;
2668 int i;
2669 int result = diff_get_patch_id(options, sha1);
2671 for (i = 0; i < q->nr; i++)
2672 diff_free_filepair(q->queue[i]);
2674 free(q->queue);
2675 q->queue = NULL;
2676 q->nr = q->alloc = 0;
2678 return result;
2681 static int is_summary_empty(const struct diff_queue_struct *q)
2683 int i;
2685 for (i = 0; i < q->nr; i++) {
2686 const struct diff_filepair *p = q->queue[i];
2688 switch (p->status) {
2689 case DIFF_STATUS_DELETED:
2690 case DIFF_STATUS_ADDED:
2691 case DIFF_STATUS_COPIED:
2692 case DIFF_STATUS_RENAMED:
2693 return 0;
2694 default:
2695 if (p->score)
2696 return 0;
2697 if (p->one->mode && p->two->mode &&
2698 p->one->mode != p->two->mode)
2699 return 0;
2700 break;
2703 return 1;
2706 void diff_flush(struct diff_options *options)
2708 struct diff_queue_struct *q = &diff_queued_diff;
2709 int i, output_format = options->output_format;
2710 int separator = 0;
2713 * Order: raw, stat, summary, patch
2714 * or: name/name-status/checkdiff (other bits clear)
2716 if (!q->nr)
2717 goto free_queue;
2719 if (output_format & (DIFF_FORMAT_RAW |
2720 DIFF_FORMAT_NAME |
2721 DIFF_FORMAT_NAME_STATUS |
2722 DIFF_FORMAT_CHECKDIFF)) {
2723 for (i = 0; i < q->nr; i++) {
2724 struct diff_filepair *p = q->queue[i];
2725 if (check_pair_status(p))
2726 flush_one_pair(p, options);
2728 separator++;
2731 if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT)) {
2732 struct diffstat_t diffstat;
2734 memset(&diffstat, 0, sizeof(struct diffstat_t));
2735 diffstat.xm.consume = diffstat_consume;
2736 for (i = 0; i < q->nr; i++) {
2737 struct diff_filepair *p = q->queue[i];
2738 if (check_pair_status(p))
2739 diff_flush_stat(p, options, &diffstat);
2741 if (output_format & DIFF_FORMAT_NUMSTAT)
2742 show_numstat(&diffstat, options);
2743 if (output_format & DIFF_FORMAT_DIFFSTAT)
2744 show_stats(&diffstat, options);
2745 else if (output_format & DIFF_FORMAT_SHORTSTAT)
2746 show_shortstats(&diffstat);
2747 separator++;
2750 if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
2751 for (i = 0; i < q->nr; i++)
2752 diff_summary(q->queue[i]);
2753 separator++;
2756 if (output_format & DIFF_FORMAT_PATCH) {
2757 if (separator) {
2758 if (options->stat_sep) {
2759 /* attach patch instead of inline */
2760 fputs(options->stat_sep, stdout);
2761 } else {
2762 putchar(options->line_termination);
2766 for (i = 0; i < q->nr; i++) {
2767 struct diff_filepair *p = q->queue[i];
2768 if (check_pair_status(p))
2769 diff_flush_patch(p, options);
2773 if (output_format & DIFF_FORMAT_CALLBACK)
2774 options->format_callback(q, options, options->format_callback_data);
2776 for (i = 0; i < q->nr; i++)
2777 diff_free_filepair(q->queue[i]);
2778 free_queue:
2779 free(q->queue);
2780 q->queue = NULL;
2781 q->nr = q->alloc = 0;
2784 static void diffcore_apply_filter(const char *filter)
2786 int i;
2787 struct diff_queue_struct *q = &diff_queued_diff;
2788 struct diff_queue_struct outq;
2789 outq.queue = NULL;
2790 outq.nr = outq.alloc = 0;
2792 if (!filter)
2793 return;
2795 if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
2796 int found;
2797 for (i = found = 0; !found && i < q->nr; i++) {
2798 struct diff_filepair *p = q->queue[i];
2799 if (((p->status == DIFF_STATUS_MODIFIED) &&
2800 ((p->score &&
2801 strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
2802 (!p->score &&
2803 strchr(filter, DIFF_STATUS_MODIFIED)))) ||
2804 ((p->status != DIFF_STATUS_MODIFIED) &&
2805 strchr(filter, p->status)))
2806 found++;
2808 if (found)
2809 return;
2811 /* otherwise we will clear the whole queue
2812 * by copying the empty outq at the end of this
2813 * function, but first clear the current entries
2814 * in the queue.
2816 for (i = 0; i < q->nr; i++)
2817 diff_free_filepair(q->queue[i]);
2819 else {
2820 /* Only the matching ones */
2821 for (i = 0; i < q->nr; i++) {
2822 struct diff_filepair *p = q->queue[i];
2824 if (((p->status == DIFF_STATUS_MODIFIED) &&
2825 ((p->score &&
2826 strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
2827 (!p->score &&
2828 strchr(filter, DIFF_STATUS_MODIFIED)))) ||
2829 ((p->status != DIFF_STATUS_MODIFIED) &&
2830 strchr(filter, p->status)))
2831 diff_q(&outq, p);
2832 else
2833 diff_free_filepair(p);
2836 free(q->queue);
2837 *q = outq;
2840 void diffcore_std(struct diff_options *options)
2842 if (options->break_opt != -1)
2843 diffcore_break(options->break_opt);
2844 if (options->detect_rename)
2845 diffcore_rename(options);
2846 if (options->break_opt != -1)
2847 diffcore_merge_broken();
2848 if (options->pickaxe)
2849 diffcore_pickaxe(options->pickaxe, options->pickaxe_opts);
2850 if (options->orderfile)
2851 diffcore_order(options->orderfile);
2852 diff_resolve_rename_copy();
2853 diffcore_apply_filter(options->filter);
2857 void diffcore_std_no_resolve(struct diff_options *options)
2859 if (options->pickaxe)
2860 diffcore_pickaxe(options->pickaxe, options->pickaxe_opts);
2861 if (options->orderfile)
2862 diffcore_order(options->orderfile);
2863 diffcore_apply_filter(options->filter);
2866 void diff_addremove(struct diff_options *options,
2867 int addremove, unsigned mode,
2868 const unsigned char *sha1,
2869 const char *base, const char *path)
2871 char concatpath[PATH_MAX];
2872 struct diff_filespec *one, *two;
2874 /* This may look odd, but it is a preparation for
2875 * feeding "there are unchanged files which should
2876 * not produce diffs, but when you are doing copy
2877 * detection you would need them, so here they are"
2878 * entries to the diff-core. They will be prefixed
2879 * with something like '=' or '*' (I haven't decided
2880 * which but should not make any difference).
2881 * Feeding the same new and old to diff_change()
2882 * also has the same effect.
2883 * Before the final output happens, they are pruned after
2884 * merged into rename/copy pairs as appropriate.
2886 if (options->reverse_diff)
2887 addremove = (addremove == '+' ? '-' :
2888 addremove == '-' ? '+' : addremove);
2890 if (!path) path = "";
2891 sprintf(concatpath, "%s%s", base, path);
2892 one = alloc_filespec(concatpath);
2893 two = alloc_filespec(concatpath);
2895 if (addremove != '+')
2896 fill_filespec(one, sha1, mode);
2897 if (addremove != '-')
2898 fill_filespec(two, sha1, mode);
2900 diff_queue(&diff_queued_diff, one, two);
2903 void diff_change(struct diff_options *options,
2904 unsigned old_mode, unsigned new_mode,
2905 const unsigned char *old_sha1,
2906 const unsigned char *new_sha1,
2907 const char *base, const char *path)
2909 char concatpath[PATH_MAX];
2910 struct diff_filespec *one, *two;
2912 if (options->reverse_diff) {
2913 unsigned tmp;
2914 const unsigned char *tmp_c;
2915 tmp = old_mode; old_mode = new_mode; new_mode = tmp;
2916 tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
2918 if (!path) path = "";
2919 sprintf(concatpath, "%s%s", base, path);
2920 one = alloc_filespec(concatpath);
2921 two = alloc_filespec(concatpath);
2922 fill_filespec(one, old_sha1, old_mode);
2923 fill_filespec(two, new_sha1, new_mode);
2925 diff_queue(&diff_queued_diff, one, two);
2928 void diff_unmerge(struct diff_options *options,
2929 const char *path,
2930 unsigned mode, const unsigned char *sha1)
2932 struct diff_filespec *one, *two;
2933 one = alloc_filespec(path);
2934 two = alloc_filespec(path);
2935 fill_filespec(one, sha1, mode);
2936 diff_queue(&diff_queued_diff, one, two)->is_unmerged = 1;