fast-export --export-marks: fix off by one error
[git/git-p4.git] / diff.c
blob803fbba451dfe8226097dcb791c339ee2ccd6735
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"
11 #include "attr.h"
12 #include "run-command.h"
13 #include "utf8.h"
15 #ifdef NO_FAST_WORKING_DIRECTORY
16 #define FAST_WORKING_DIRECTORY 0
17 #else
18 #define FAST_WORKING_DIRECTORY 1
19 #endif
21 static int diff_detect_rename_default;
22 static int diff_rename_limit_default = 200;
23 int diff_use_color_default = -1;
24 static const char *external_diff_cmd_cfg;
25 int diff_auto_refresh_index = 1;
27 static char diff_colors[][COLOR_MAXLEN] = {
28 "\033[m", /* reset */
29 "", /* PLAIN (normal) */
30 "\033[1m", /* METAINFO (bold) */
31 "\033[36m", /* FRAGINFO (cyan) */
32 "\033[31m", /* OLD (red) */
33 "\033[32m", /* NEW (green) */
34 "\033[33m", /* COMMIT (yellow) */
35 "\033[41m", /* WHITESPACE (red background) */
38 static int parse_diff_color_slot(const char *var, int ofs)
40 if (!strcasecmp(var+ofs, "plain"))
41 return DIFF_PLAIN;
42 if (!strcasecmp(var+ofs, "meta"))
43 return DIFF_METAINFO;
44 if (!strcasecmp(var+ofs, "frag"))
45 return DIFF_FRAGINFO;
46 if (!strcasecmp(var+ofs, "old"))
47 return DIFF_FILE_OLD;
48 if (!strcasecmp(var+ofs, "new"))
49 return DIFF_FILE_NEW;
50 if (!strcasecmp(var+ofs, "commit"))
51 return DIFF_COMMIT;
52 if (!strcasecmp(var+ofs, "whitespace"))
53 return DIFF_WHITESPACE;
54 die("bad config variable '%s'", var);
57 static struct ll_diff_driver {
58 const char *name;
59 struct ll_diff_driver *next;
60 const char *cmd;
61 } *user_diff, **user_diff_tail;
64 * Currently there is only "diff.<drivername>.command" variable;
65 * because there are "diff.color.<slot>" variables, we are parsing
66 * this in a bit convoluted way to allow low level diff driver
67 * called "color".
69 static int parse_lldiff_command(const char *var, const char *ep, const char *value)
71 const char *name;
72 int namelen;
73 struct ll_diff_driver *drv;
75 name = var + 5;
76 namelen = ep - name;
77 for (drv = user_diff; drv; drv = drv->next)
78 if (!strncmp(drv->name, name, namelen) && !drv->name[namelen])
79 break;
80 if (!drv) {
81 drv = xcalloc(1, sizeof(struct ll_diff_driver));
82 drv->name = xmemdupz(name, namelen);
83 if (!user_diff_tail)
84 user_diff_tail = &user_diff;
85 *user_diff_tail = drv;
86 user_diff_tail = &(drv->next);
89 return git_config_string(&(drv->cmd), var, value);
93 * 'diff.<what>.funcname' attribute can be specified in the configuration
94 * to define a customized regexp to find the beginning of a function to
95 * be used for hunk header lines of "diff -p" style output.
97 static struct funcname_pattern {
98 char *name;
99 char *pattern;
100 struct funcname_pattern *next;
101 } *funcname_pattern_list;
103 static int parse_funcname_pattern(const char *var, const char *ep, const char *value)
105 const char *name;
106 int namelen;
107 struct funcname_pattern *pp;
109 name = var + 5; /* "diff." */
110 namelen = ep - name;
112 for (pp = funcname_pattern_list; pp; pp = pp->next)
113 if (!strncmp(pp->name, name, namelen) && !pp->name[namelen])
114 break;
115 if (!pp) {
116 pp = xcalloc(1, sizeof(*pp));
117 pp->name = xmemdupz(name, namelen);
118 pp->next = funcname_pattern_list;
119 funcname_pattern_list = pp;
121 free(pp->pattern);
122 pp->pattern = xstrdup(value);
123 return 0;
127 * These are to give UI layer defaults.
128 * The core-level commands such as git-diff-files should
129 * never be affected by the setting of diff.renames
130 * the user happens to have in the configuration file.
132 int git_diff_ui_config(const char *var, const char *value, void *cb)
134 if (!strcmp(var, "diff.renamelimit")) {
135 diff_rename_limit_default = git_config_int(var, value);
136 return 0;
138 if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
139 diff_use_color_default = git_config_colorbool(var, value, -1);
140 return 0;
142 if (!strcmp(var, "diff.renames")) {
143 if (!value)
144 diff_detect_rename_default = DIFF_DETECT_RENAME;
145 else if (!strcasecmp(value, "copies") ||
146 !strcasecmp(value, "copy"))
147 diff_detect_rename_default = DIFF_DETECT_COPY;
148 else if (git_config_bool(var,value))
149 diff_detect_rename_default = DIFF_DETECT_RENAME;
150 return 0;
152 if (!strcmp(var, "diff.autorefreshindex")) {
153 diff_auto_refresh_index = git_config_bool(var, value);
154 return 0;
156 if (!strcmp(var, "diff.external")) {
157 if (!value)
158 return config_error_nonbool(var);
159 external_diff_cmd_cfg = xstrdup(value);
160 return 0;
162 if (!prefixcmp(var, "diff.")) {
163 const char *ep = strrchr(var, '.');
165 if (ep != var + 4 && !strcmp(ep, ".command"))
166 return parse_lldiff_command(var, ep, value);
169 return git_diff_basic_config(var, value, cb);
172 int git_diff_basic_config(const char *var, const char *value, void *cb)
174 if (!prefixcmp(var, "diff.color.") || !prefixcmp(var, "color.diff.")) {
175 int slot = parse_diff_color_slot(var, 11);
176 if (!value)
177 return config_error_nonbool(var);
178 color_parse(value, var, diff_colors[slot]);
179 return 0;
182 if (!prefixcmp(var, "diff.")) {
183 const char *ep = strrchr(var, '.');
184 if (ep != var + 4) {
185 if (!strcmp(ep, ".funcname")) {
186 if (!value)
187 return config_error_nonbool(var);
188 return parse_funcname_pattern(var, ep, value);
193 return git_color_default_config(var, value, cb);
196 static char *quote_two(const char *one, const char *two)
198 int need_one = quote_c_style(one, NULL, NULL, 1);
199 int need_two = quote_c_style(two, NULL, NULL, 1);
200 struct strbuf res;
202 strbuf_init(&res, 0);
203 if (need_one + need_two) {
204 strbuf_addch(&res, '"');
205 quote_c_style(one, &res, NULL, 1);
206 quote_c_style(two, &res, NULL, 1);
207 strbuf_addch(&res, '"');
208 } else {
209 strbuf_addstr(&res, one);
210 strbuf_addstr(&res, two);
212 return strbuf_detach(&res, NULL);
215 static const char *external_diff(void)
217 static const char *external_diff_cmd = NULL;
218 static int done_preparing = 0;
220 if (done_preparing)
221 return external_diff_cmd;
222 external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
223 if (!external_diff_cmd)
224 external_diff_cmd = external_diff_cmd_cfg;
225 done_preparing = 1;
226 return external_diff_cmd;
229 static struct diff_tempfile {
230 const char *name; /* filename external diff should read from */
231 char hex[41];
232 char mode[10];
233 char tmp_path[PATH_MAX];
234 } diff_temp[2];
236 static int count_lines(const char *data, int size)
238 int count, ch, completely_empty = 1, nl_just_seen = 0;
239 count = 0;
240 while (0 < size--) {
241 ch = *data++;
242 if (ch == '\n') {
243 count++;
244 nl_just_seen = 1;
245 completely_empty = 0;
247 else {
248 nl_just_seen = 0;
249 completely_empty = 0;
252 if (completely_empty)
253 return 0;
254 if (!nl_just_seen)
255 count++; /* no trailing newline */
256 return count;
259 static void print_line_count(FILE *file, int count)
261 switch (count) {
262 case 0:
263 fprintf(file, "0,0");
264 break;
265 case 1:
266 fprintf(file, "1");
267 break;
268 default:
269 fprintf(file, "1,%d", count);
270 break;
274 static void copy_file_with_prefix(FILE *file,
275 int prefix, const char *data, int size,
276 const char *set, const char *reset)
278 int ch, nl_just_seen = 1;
279 while (0 < size--) {
280 ch = *data++;
281 if (nl_just_seen) {
282 fputs(set, file);
283 putc(prefix, file);
285 if (ch == '\n') {
286 nl_just_seen = 1;
287 fputs(reset, file);
288 } else
289 nl_just_seen = 0;
290 putc(ch, file);
292 if (!nl_just_seen)
293 fprintf(file, "%s\n\\ No newline at end of file\n", reset);
296 static void emit_rewrite_diff(const char *name_a,
297 const char *name_b,
298 struct diff_filespec *one,
299 struct diff_filespec *two,
300 struct diff_options *o)
302 int lc_a, lc_b;
303 int color_diff = DIFF_OPT_TST(o, COLOR_DIFF);
304 const char *name_a_tab, *name_b_tab;
305 const char *metainfo = diff_get_color(color_diff, DIFF_METAINFO);
306 const char *fraginfo = diff_get_color(color_diff, DIFF_FRAGINFO);
307 const char *old = diff_get_color(color_diff, DIFF_FILE_OLD);
308 const char *new = diff_get_color(color_diff, DIFF_FILE_NEW);
309 const char *reset = diff_get_color(color_diff, DIFF_RESET);
310 static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
312 name_a += (*name_a == '/');
313 name_b += (*name_b == '/');
314 name_a_tab = strchr(name_a, ' ') ? "\t" : "";
315 name_b_tab = strchr(name_b, ' ') ? "\t" : "";
317 strbuf_reset(&a_name);
318 strbuf_reset(&b_name);
319 quote_two_c_style(&a_name, o->a_prefix, name_a, 0);
320 quote_two_c_style(&b_name, o->b_prefix, name_b, 0);
322 diff_populate_filespec(one, 0);
323 diff_populate_filespec(two, 0);
324 lc_a = count_lines(one->data, one->size);
325 lc_b = count_lines(two->data, two->size);
326 fprintf(o->file,
327 "%s--- %s%s%s\n%s+++ %s%s%s\n%s@@ -",
328 metainfo, a_name.buf, name_a_tab, reset,
329 metainfo, b_name.buf, name_b_tab, reset, fraginfo);
330 print_line_count(o->file, lc_a);
331 fprintf(o->file, " +");
332 print_line_count(o->file, lc_b);
333 fprintf(o->file, " @@%s\n", reset);
334 if (lc_a)
335 copy_file_with_prefix(o->file, '-', one->data, one->size, old, reset);
336 if (lc_b)
337 copy_file_with_prefix(o->file, '+', two->data, two->size, new, reset);
340 static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
342 if (!DIFF_FILE_VALID(one)) {
343 mf->ptr = (char *)""; /* does not matter */
344 mf->size = 0;
345 return 0;
347 else if (diff_populate_filespec(one, 0))
348 return -1;
349 mf->ptr = one->data;
350 mf->size = one->size;
351 return 0;
354 struct diff_words_buffer {
355 mmfile_t text;
356 long alloc;
357 long current; /* output pointer */
358 int suppressed_newline;
361 static void diff_words_append(char *line, unsigned long len,
362 struct diff_words_buffer *buffer)
364 if (buffer->text.size + len > buffer->alloc) {
365 buffer->alloc = (buffer->text.size + len) * 3 / 2;
366 buffer->text.ptr = xrealloc(buffer->text.ptr, buffer->alloc);
368 line++;
369 len--;
370 memcpy(buffer->text.ptr + buffer->text.size, line, len);
371 buffer->text.size += len;
374 struct diff_words_data {
375 struct xdiff_emit_state xm;
376 struct diff_words_buffer minus, plus;
377 FILE *file;
380 static void print_word(FILE *file, struct diff_words_buffer *buffer, int len, int color,
381 int suppress_newline)
383 const char *ptr;
384 int eol = 0;
386 if (len == 0)
387 return;
389 ptr = buffer->text.ptr + buffer->current;
390 buffer->current += len;
392 if (ptr[len - 1] == '\n') {
393 eol = 1;
394 len--;
397 fputs(diff_get_color(1, color), file);
398 fwrite(ptr, len, 1, file);
399 fputs(diff_get_color(1, DIFF_RESET), file);
401 if (eol) {
402 if (suppress_newline)
403 buffer->suppressed_newline = 1;
404 else
405 putc('\n', file);
409 static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
411 struct diff_words_data *diff_words = priv;
413 if (diff_words->minus.suppressed_newline) {
414 if (line[0] != '+')
415 putc('\n', diff_words->file);
416 diff_words->minus.suppressed_newline = 0;
419 len--;
420 switch (line[0]) {
421 case '-':
422 print_word(diff_words->file,
423 &diff_words->minus, len, DIFF_FILE_OLD, 1);
424 break;
425 case '+':
426 print_word(diff_words->file,
427 &diff_words->plus, len, DIFF_FILE_NEW, 0);
428 break;
429 case ' ':
430 print_word(diff_words->file,
431 &diff_words->plus, len, DIFF_PLAIN, 0);
432 diff_words->minus.current += len;
433 break;
437 /* this executes the word diff on the accumulated buffers */
438 static void diff_words_show(struct diff_words_data *diff_words)
440 xpparam_t xpp;
441 xdemitconf_t xecfg;
442 xdemitcb_t ecb;
443 mmfile_t minus, plus;
444 int i;
446 memset(&xecfg, 0, sizeof(xecfg));
447 minus.size = diff_words->minus.text.size;
448 minus.ptr = xmalloc(minus.size);
449 memcpy(minus.ptr, diff_words->minus.text.ptr, minus.size);
450 for (i = 0; i < minus.size; i++)
451 if (isspace(minus.ptr[i]))
452 minus.ptr[i] = '\n';
453 diff_words->minus.current = 0;
455 plus.size = diff_words->plus.text.size;
456 plus.ptr = xmalloc(plus.size);
457 memcpy(plus.ptr, diff_words->plus.text.ptr, plus.size);
458 for (i = 0; i < plus.size; i++)
459 if (isspace(plus.ptr[i]))
460 plus.ptr[i] = '\n';
461 diff_words->plus.current = 0;
463 xpp.flags = XDF_NEED_MINIMAL;
464 xecfg.ctxlen = diff_words->minus.alloc + diff_words->plus.alloc;
465 ecb.outf = xdiff_outf;
466 ecb.priv = diff_words;
467 diff_words->xm.consume = fn_out_diff_words_aux;
468 xdi_diff(&minus, &plus, &xpp, &xecfg, &ecb);
470 free(minus.ptr);
471 free(plus.ptr);
472 diff_words->minus.text.size = diff_words->plus.text.size = 0;
474 if (diff_words->minus.suppressed_newline) {
475 putc('\n', diff_words->file);
476 diff_words->minus.suppressed_newline = 0;
480 typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
482 struct emit_callback {
483 struct xdiff_emit_state xm;
484 int nparents, color_diff;
485 unsigned ws_rule;
486 sane_truncate_fn truncate;
487 const char **label_path;
488 struct diff_words_data *diff_words;
489 int *found_changesp;
490 FILE *file;
493 static void free_diff_words_data(struct emit_callback *ecbdata)
495 if (ecbdata->diff_words) {
496 /* flush buffers */
497 if (ecbdata->diff_words->minus.text.size ||
498 ecbdata->diff_words->plus.text.size)
499 diff_words_show(ecbdata->diff_words);
501 free (ecbdata->diff_words->minus.text.ptr);
502 free (ecbdata->diff_words->plus.text.ptr);
503 free(ecbdata->diff_words);
504 ecbdata->diff_words = NULL;
508 const char *diff_get_color(int diff_use_color, enum color_diff ix)
510 if (diff_use_color)
511 return diff_colors[ix];
512 return "";
515 static void emit_line(FILE *file, const char *set, const char *reset, const char *line, int len)
517 int has_trailing_newline = (len > 0 && line[len-1] == '\n');
518 if (has_trailing_newline)
519 len--;
521 fputs(set, file);
522 fwrite(line, len, 1, file);
523 fputs(reset, file);
524 if (has_trailing_newline)
525 fputc('\n', file);
528 static void emit_add_line(const char *reset, struct emit_callback *ecbdata, const char *line, int len)
530 const char *ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
531 const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_NEW);
533 if (!*ws)
534 emit_line(ecbdata->file, set, reset, line, len);
535 else {
536 /* Emit just the prefix, then the rest. */
537 emit_line(ecbdata->file, set, reset, line, ecbdata->nparents);
538 ws_check_emit(line + ecbdata->nparents,
539 len - ecbdata->nparents, ecbdata->ws_rule,
540 ecbdata->file, set, reset, ws);
544 static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
546 const char *cp;
547 unsigned long allot;
548 size_t l = len;
550 if (ecb->truncate)
551 return ecb->truncate(line, len);
552 cp = line;
553 allot = l;
554 while (0 < l) {
555 (void) utf8_width(&cp, &l);
556 if (!cp)
557 break; /* truncated in the middle? */
559 return allot - l;
562 static void fn_out_consume(void *priv, char *line, unsigned long len)
564 int i;
565 int color;
566 struct emit_callback *ecbdata = priv;
567 const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
568 const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
569 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
571 *(ecbdata->found_changesp) = 1;
573 if (ecbdata->label_path[0]) {
574 const char *name_a_tab, *name_b_tab;
576 name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
577 name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
579 fprintf(ecbdata->file, "%s--- %s%s%s\n",
580 meta, ecbdata->label_path[0], reset, name_a_tab);
581 fprintf(ecbdata->file, "%s+++ %s%s%s\n",
582 meta, ecbdata->label_path[1], reset, name_b_tab);
583 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
586 /* This is not really necessary for now because
587 * this codepath only deals with two-way diffs.
589 for (i = 0; i < len && line[i] == '@'; i++)
591 if (2 <= i && i < len && line[i] == ' ') {
592 ecbdata->nparents = i - 1;
593 len = sane_truncate_line(ecbdata, line, len);
594 emit_line(ecbdata->file,
595 diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO),
596 reset, line, len);
597 if (line[len-1] != '\n')
598 putc('\n', ecbdata->file);
599 return;
602 if (len < ecbdata->nparents) {
603 emit_line(ecbdata->file, reset, reset, line, len);
604 return;
607 color = DIFF_PLAIN;
608 if (ecbdata->diff_words && ecbdata->nparents != 1)
609 /* fall back to normal diff */
610 free_diff_words_data(ecbdata);
611 if (ecbdata->diff_words) {
612 if (line[0] == '-') {
613 diff_words_append(line, len,
614 &ecbdata->diff_words->minus);
615 return;
616 } else if (line[0] == '+') {
617 diff_words_append(line, len,
618 &ecbdata->diff_words->plus);
619 return;
621 if (ecbdata->diff_words->minus.text.size ||
622 ecbdata->diff_words->plus.text.size)
623 diff_words_show(ecbdata->diff_words);
624 line++;
625 len--;
626 emit_line(ecbdata->file, plain, reset, line, len);
627 return;
629 for (i = 0; i < ecbdata->nparents && len; i++) {
630 if (line[i] == '-')
631 color = DIFF_FILE_OLD;
632 else if (line[i] == '+')
633 color = DIFF_FILE_NEW;
636 if (color != DIFF_FILE_NEW) {
637 emit_line(ecbdata->file,
638 diff_get_color(ecbdata->color_diff, color),
639 reset, line, len);
640 return;
642 emit_add_line(reset, ecbdata, line, len);
645 static char *pprint_rename(const char *a, const char *b)
647 const char *old = a;
648 const char *new = b;
649 struct strbuf name;
650 int pfx_length, sfx_length;
651 int len_a = strlen(a);
652 int len_b = strlen(b);
653 int a_midlen, b_midlen;
654 int qlen_a = quote_c_style(a, NULL, NULL, 0);
655 int qlen_b = quote_c_style(b, NULL, NULL, 0);
657 strbuf_init(&name, 0);
658 if (qlen_a || qlen_b) {
659 quote_c_style(a, &name, NULL, 0);
660 strbuf_addstr(&name, " => ");
661 quote_c_style(b, &name, NULL, 0);
662 return strbuf_detach(&name, NULL);
665 /* Find common prefix */
666 pfx_length = 0;
667 while (*old && *new && *old == *new) {
668 if (*old == '/')
669 pfx_length = old - a + 1;
670 old++;
671 new++;
674 /* Find common suffix */
675 old = a + len_a;
676 new = b + len_b;
677 sfx_length = 0;
678 while (a <= old && b <= new && *old == *new) {
679 if (*old == '/')
680 sfx_length = len_a - (old - a);
681 old--;
682 new--;
686 * pfx{mid-a => mid-b}sfx
687 * {pfx-a => pfx-b}sfx
688 * pfx{sfx-a => sfx-b}
689 * name-a => name-b
691 a_midlen = len_a - pfx_length - sfx_length;
692 b_midlen = len_b - pfx_length - sfx_length;
693 if (a_midlen < 0)
694 a_midlen = 0;
695 if (b_midlen < 0)
696 b_midlen = 0;
698 strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
699 if (pfx_length + sfx_length) {
700 strbuf_add(&name, a, pfx_length);
701 strbuf_addch(&name, '{');
703 strbuf_add(&name, a + pfx_length, a_midlen);
704 strbuf_addstr(&name, " => ");
705 strbuf_add(&name, b + pfx_length, b_midlen);
706 if (pfx_length + sfx_length) {
707 strbuf_addch(&name, '}');
708 strbuf_add(&name, a + len_a - sfx_length, sfx_length);
710 return strbuf_detach(&name, NULL);
713 struct diffstat_t {
714 struct xdiff_emit_state xm;
716 int nr;
717 int alloc;
718 struct diffstat_file {
719 char *from_name;
720 char *name;
721 char *print_name;
722 unsigned is_unmerged:1;
723 unsigned is_binary:1;
724 unsigned is_renamed:1;
725 unsigned int added, deleted;
726 } **files;
729 static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
730 const char *name_a,
731 const char *name_b)
733 struct diffstat_file *x;
734 x = xcalloc(sizeof (*x), 1);
735 if (diffstat->nr == diffstat->alloc) {
736 diffstat->alloc = alloc_nr(diffstat->alloc);
737 diffstat->files = xrealloc(diffstat->files,
738 diffstat->alloc * sizeof(x));
740 diffstat->files[diffstat->nr++] = x;
741 if (name_b) {
742 x->from_name = xstrdup(name_a);
743 x->name = xstrdup(name_b);
744 x->is_renamed = 1;
746 else {
747 x->from_name = NULL;
748 x->name = xstrdup(name_a);
750 return x;
753 static void diffstat_consume(void *priv, char *line, unsigned long len)
755 struct diffstat_t *diffstat = priv;
756 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
758 if (line[0] == '+')
759 x->added++;
760 else if (line[0] == '-')
761 x->deleted++;
764 const char mime_boundary_leader[] = "------------";
766 static int scale_linear(int it, int width, int max_change)
769 * make sure that at least one '-' is printed if there were deletions,
770 * and likewise for '+'.
772 if (max_change < 2)
773 return it;
774 return ((it - 1) * (width - 1) + max_change - 1) / (max_change - 1);
777 static void show_name(FILE *file,
778 const char *prefix, const char *name, int len,
779 const char *reset, const char *set)
781 fprintf(file, " %s%s%-*s%s |", set, prefix, len, name, reset);
784 static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
786 if (cnt <= 0)
787 return;
788 fprintf(file, "%s", set);
789 while (cnt--)
790 putc(ch, file);
791 fprintf(file, "%s", reset);
794 static void fill_print_name(struct diffstat_file *file)
796 char *pname;
798 if (file->print_name)
799 return;
801 if (!file->is_renamed) {
802 struct strbuf buf;
803 strbuf_init(&buf, 0);
804 if (quote_c_style(file->name, &buf, NULL, 0)) {
805 pname = strbuf_detach(&buf, NULL);
806 } else {
807 pname = file->name;
808 strbuf_release(&buf);
810 } else {
811 pname = pprint_rename(file->from_name, file->name);
813 file->print_name = pname;
816 static void show_stats(struct diffstat_t* data, struct diff_options *options)
818 int i, len, add, del, total, adds = 0, dels = 0;
819 int max_change = 0, max_len = 0;
820 int total_files = data->nr;
821 int width, name_width;
822 const char *reset, *set, *add_c, *del_c;
824 if (data->nr == 0)
825 return;
827 width = options->stat_width ? options->stat_width : 80;
828 name_width = options->stat_name_width ? options->stat_name_width : 50;
830 /* Sanity: give at least 5 columns to the graph,
831 * but leave at least 10 columns for the name.
833 if (width < 25)
834 width = 25;
835 if (name_width < 10)
836 name_width = 10;
837 else if (width < name_width + 15)
838 name_width = width - 15;
840 /* Find the longest filename and max number of changes */
841 reset = diff_get_color_opt(options, DIFF_RESET);
842 set = diff_get_color_opt(options, DIFF_PLAIN);
843 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
844 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
846 for (i = 0; i < data->nr; i++) {
847 struct diffstat_file *file = data->files[i];
848 int change = file->added + file->deleted;
849 fill_print_name(file);
850 len = strlen(file->print_name);
851 if (max_len < len)
852 max_len = len;
854 if (file->is_binary || file->is_unmerged)
855 continue;
856 if (max_change < change)
857 max_change = change;
860 /* Compute the width of the graph part;
861 * 10 is for one blank at the beginning of the line plus
862 * " | count " between the name and the graph.
864 * From here on, name_width is the width of the name area,
865 * and width is the width of the graph area.
867 name_width = (name_width < max_len) ? name_width : max_len;
868 if (width < (name_width + 10) + max_change)
869 width = width - (name_width + 10);
870 else
871 width = max_change;
873 for (i = 0; i < data->nr; i++) {
874 const char *prefix = "";
875 char *name = data->files[i]->print_name;
876 int added = data->files[i]->added;
877 int deleted = data->files[i]->deleted;
878 int name_len;
881 * "scale" the filename
883 len = name_width;
884 name_len = strlen(name);
885 if (name_width < name_len) {
886 char *slash;
887 prefix = "...";
888 len -= 3;
889 name += name_len - len;
890 slash = strchr(name, '/');
891 if (slash)
892 name = slash;
895 if (data->files[i]->is_binary) {
896 show_name(options->file, prefix, name, len, reset, set);
897 fprintf(options->file, " Bin ");
898 fprintf(options->file, "%s%d%s", del_c, deleted, reset);
899 fprintf(options->file, " -> ");
900 fprintf(options->file, "%s%d%s", add_c, added, reset);
901 fprintf(options->file, " bytes");
902 fprintf(options->file, "\n");
903 continue;
905 else if (data->files[i]->is_unmerged) {
906 show_name(options->file, prefix, name, len, reset, set);
907 fprintf(options->file, " Unmerged\n");
908 continue;
910 else if (!data->files[i]->is_renamed &&
911 (added + deleted == 0)) {
912 total_files--;
913 continue;
917 * scale the add/delete
919 add = added;
920 del = deleted;
921 total = add + del;
922 adds += add;
923 dels += del;
925 if (width <= max_change) {
926 add = scale_linear(add, width, max_change);
927 del = scale_linear(del, width, max_change);
928 total = add + del;
930 show_name(options->file, prefix, name, len, reset, set);
931 fprintf(options->file, "%5d%s", added + deleted,
932 added + deleted ? " " : "");
933 show_graph(options->file, '+', add, add_c, reset);
934 show_graph(options->file, '-', del, del_c, reset);
935 fprintf(options->file, "\n");
937 fprintf(options->file,
938 "%s %d files changed, %d insertions(+), %d deletions(-)%s\n",
939 set, total_files, adds, dels, reset);
942 static void show_shortstats(struct diffstat_t* data, struct diff_options *options)
944 int i, adds = 0, dels = 0, total_files = data->nr;
946 if (data->nr == 0)
947 return;
949 for (i = 0; i < data->nr; i++) {
950 if (!data->files[i]->is_binary &&
951 !data->files[i]->is_unmerged) {
952 int added = data->files[i]->added;
953 int deleted= data->files[i]->deleted;
954 if (!data->files[i]->is_renamed &&
955 (added + deleted == 0)) {
956 total_files--;
957 } else {
958 adds += added;
959 dels += deleted;
963 fprintf(options->file, " %d files changed, %d insertions(+), %d deletions(-)\n",
964 total_files, adds, dels);
967 static void show_numstat(struct diffstat_t* data, struct diff_options *options)
969 int i;
971 if (data->nr == 0)
972 return;
974 for (i = 0; i < data->nr; i++) {
975 struct diffstat_file *file = data->files[i];
977 if (file->is_binary)
978 fprintf(options->file, "-\t-\t");
979 else
980 fprintf(options->file,
981 "%d\t%d\t", file->added, file->deleted);
982 if (options->line_termination) {
983 fill_print_name(file);
984 if (!file->is_renamed)
985 write_name_quoted(file->name, options->file,
986 options->line_termination);
987 else {
988 fputs(file->print_name, options->file);
989 putc(options->line_termination, options->file);
991 } else {
992 if (file->is_renamed) {
993 putc('\0', options->file);
994 write_name_quoted(file->from_name, options->file, '\0');
996 write_name_quoted(file->name, options->file, '\0');
1001 struct dirstat_file {
1002 const char *name;
1003 unsigned long changed;
1006 struct dirstat_dir {
1007 struct dirstat_file *files;
1008 int alloc, nr, percent, cumulative;
1011 static long gather_dirstat(FILE *file, struct dirstat_dir *dir, unsigned long changed, const char *base, int baselen)
1013 unsigned long this_dir = 0;
1014 unsigned int sources = 0;
1016 while (dir->nr) {
1017 struct dirstat_file *f = dir->files;
1018 int namelen = strlen(f->name);
1019 unsigned long this;
1020 char *slash;
1022 if (namelen < baselen)
1023 break;
1024 if (memcmp(f->name, base, baselen))
1025 break;
1026 slash = strchr(f->name + baselen, '/');
1027 if (slash) {
1028 int newbaselen = slash + 1 - f->name;
1029 this = gather_dirstat(file, dir, changed, f->name, newbaselen);
1030 sources++;
1031 } else {
1032 this = f->changed;
1033 dir->files++;
1034 dir->nr--;
1035 sources += 2;
1037 this_dir += this;
1041 * We don't report dirstat's for
1042 * - the top level
1043 * - or cases where everything came from a single directory
1044 * under this directory (sources == 1).
1046 if (baselen && sources != 1) {
1047 int permille = this_dir * 1000 / changed;
1048 if (permille) {
1049 int percent = permille / 10;
1050 if (percent >= dir->percent) {
1051 fprintf(file, "%4d.%01d%% %.*s\n", percent, permille % 10, baselen, base);
1052 if (!dir->cumulative)
1053 return 0;
1057 return this_dir;
1060 static void show_dirstat(struct diff_options *options)
1062 int i;
1063 unsigned long changed;
1064 struct dirstat_dir dir;
1065 struct diff_queue_struct *q = &diff_queued_diff;
1067 dir.files = NULL;
1068 dir.alloc = 0;
1069 dir.nr = 0;
1070 dir.percent = options->dirstat_percent;
1071 dir.cumulative = options->output_format & DIFF_FORMAT_CUMULATIVE;
1073 changed = 0;
1074 for (i = 0; i < q->nr; i++) {
1075 struct diff_filepair *p = q->queue[i];
1076 const char *name;
1077 unsigned long copied, added, damage;
1079 name = p->one->path ? p->one->path : p->two->path;
1081 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
1082 diff_populate_filespec(p->one, 0);
1083 diff_populate_filespec(p->two, 0);
1084 diffcore_count_changes(p->one, p->two, NULL, NULL, 0,
1085 &copied, &added);
1086 diff_free_filespec_data(p->one);
1087 diff_free_filespec_data(p->two);
1088 } else if (DIFF_FILE_VALID(p->one)) {
1089 diff_populate_filespec(p->one, 1);
1090 copied = added = 0;
1091 diff_free_filespec_data(p->one);
1092 } else if (DIFF_FILE_VALID(p->two)) {
1093 diff_populate_filespec(p->two, 1);
1094 copied = 0;
1095 added = p->two->size;
1096 diff_free_filespec_data(p->two);
1097 } else
1098 continue;
1101 * Original minus copied is the removed material,
1102 * added is the new material. They are both damages
1103 * made to the preimage.
1105 damage = (p->one->size - copied) + added;
1107 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1108 dir.files[dir.nr].name = name;
1109 dir.files[dir.nr].changed = damage;
1110 changed += damage;
1111 dir.nr++;
1114 /* This can happen even with many files, if everything was renames */
1115 if (!changed)
1116 return;
1118 /* Show all directories with more than x% of the changes */
1119 gather_dirstat(options->file, &dir, changed, "", 0);
1122 static void free_diffstat_info(struct diffstat_t *diffstat)
1124 int i;
1125 for (i = 0; i < diffstat->nr; i++) {
1126 struct diffstat_file *f = diffstat->files[i];
1127 if (f->name != f->print_name)
1128 free(f->print_name);
1129 free(f->name);
1130 free(f->from_name);
1131 free(f);
1133 free(diffstat->files);
1136 struct checkdiff_t {
1137 struct xdiff_emit_state xm;
1138 const char *filename;
1139 int lineno;
1140 struct diff_options *o;
1141 unsigned ws_rule;
1142 unsigned status;
1143 int trailing_blanks_start;
1146 static int is_conflict_marker(const char *line, unsigned long len)
1148 char firstchar;
1149 int cnt;
1151 if (len < 8)
1152 return 0;
1153 firstchar = line[0];
1154 switch (firstchar) {
1155 case '=': case '>': case '<':
1156 break;
1157 default:
1158 return 0;
1160 for (cnt = 1; cnt < 7; cnt++)
1161 if (line[cnt] != firstchar)
1162 return 0;
1163 /* line[0] thru line[6] are same as firstchar */
1164 if (firstchar == '=') {
1165 /* divider between ours and theirs? */
1166 if (len != 8 || line[7] != '\n')
1167 return 0;
1168 } else if (len < 8 || !isspace(line[7])) {
1169 /* not divider before ours nor after theirs */
1170 return 0;
1172 return 1;
1175 static void checkdiff_consume(void *priv, char *line, unsigned long len)
1177 struct checkdiff_t *data = priv;
1178 int color_diff = DIFF_OPT_TST(data->o, COLOR_DIFF);
1179 const char *ws = diff_get_color(color_diff, DIFF_WHITESPACE);
1180 const char *reset = diff_get_color(color_diff, DIFF_RESET);
1181 const char *set = diff_get_color(color_diff, DIFF_FILE_NEW);
1182 char *err;
1184 if (line[0] == '+') {
1185 unsigned bad;
1186 data->lineno++;
1187 if (!ws_blank_line(line + 1, len - 1, data->ws_rule))
1188 data->trailing_blanks_start = 0;
1189 else if (!data->trailing_blanks_start)
1190 data->trailing_blanks_start = data->lineno;
1191 if (is_conflict_marker(line + 1, len - 1)) {
1192 data->status |= 1;
1193 fprintf(data->o->file,
1194 "%s:%d: leftover conflict marker\n",
1195 data->filename, data->lineno);
1197 bad = ws_check(line + 1, len - 1, data->ws_rule);
1198 if (!bad)
1199 return;
1200 data->status |= bad;
1201 err = whitespace_error_string(bad);
1202 fprintf(data->o->file, "%s:%d: %s.\n",
1203 data->filename, data->lineno, err);
1204 free(err);
1205 emit_line(data->o->file, set, reset, line, 1);
1206 ws_check_emit(line + 1, len - 1, data->ws_rule,
1207 data->o->file, set, reset, ws);
1208 } else if (line[0] == ' ') {
1209 data->lineno++;
1210 data->trailing_blanks_start = 0;
1211 } else if (line[0] == '@') {
1212 char *plus = strchr(line, '+');
1213 if (plus)
1214 data->lineno = strtol(plus, NULL, 10) - 1;
1215 else
1216 die("invalid diff");
1217 data->trailing_blanks_start = 0;
1221 static unsigned char *deflate_it(char *data,
1222 unsigned long size,
1223 unsigned long *result_size)
1225 int bound;
1226 unsigned char *deflated;
1227 z_stream stream;
1229 memset(&stream, 0, sizeof(stream));
1230 deflateInit(&stream, zlib_compression_level);
1231 bound = deflateBound(&stream, size);
1232 deflated = xmalloc(bound);
1233 stream.next_out = deflated;
1234 stream.avail_out = bound;
1236 stream.next_in = (unsigned char *)data;
1237 stream.avail_in = size;
1238 while (deflate(&stream, Z_FINISH) == Z_OK)
1239 ; /* nothing */
1240 deflateEnd(&stream);
1241 *result_size = stream.total_out;
1242 return deflated;
1245 static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two)
1247 void *cp;
1248 void *delta;
1249 void *deflated;
1250 void *data;
1251 unsigned long orig_size;
1252 unsigned long delta_size;
1253 unsigned long deflate_size;
1254 unsigned long data_size;
1256 /* We could do deflated delta, or we could do just deflated two,
1257 * whichever is smaller.
1259 delta = NULL;
1260 deflated = deflate_it(two->ptr, two->size, &deflate_size);
1261 if (one->size && two->size) {
1262 delta = diff_delta(one->ptr, one->size,
1263 two->ptr, two->size,
1264 &delta_size, deflate_size);
1265 if (delta) {
1266 void *to_free = delta;
1267 orig_size = delta_size;
1268 delta = deflate_it(delta, delta_size, &delta_size);
1269 free(to_free);
1273 if (delta && delta_size < deflate_size) {
1274 fprintf(file, "delta %lu\n", orig_size);
1275 free(deflated);
1276 data = delta;
1277 data_size = delta_size;
1279 else {
1280 fprintf(file, "literal %lu\n", two->size);
1281 free(delta);
1282 data = deflated;
1283 data_size = deflate_size;
1286 /* emit data encoded in base85 */
1287 cp = data;
1288 while (data_size) {
1289 int bytes = (52 < data_size) ? 52 : data_size;
1290 char line[70];
1291 data_size -= bytes;
1292 if (bytes <= 26)
1293 line[0] = bytes + 'A' - 1;
1294 else
1295 line[0] = bytes - 26 + 'a' - 1;
1296 encode_85(line + 1, cp, bytes);
1297 cp = (char *) cp + bytes;
1298 fputs(line, file);
1299 fputc('\n', file);
1301 fprintf(file, "\n");
1302 free(data);
1305 static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two)
1307 fprintf(file, "GIT binary patch\n");
1308 emit_binary_diff_body(file, one, two);
1309 emit_binary_diff_body(file, two, one);
1312 static void setup_diff_attr_check(struct git_attr_check *check)
1314 static struct git_attr *attr_diff;
1316 if (!attr_diff) {
1317 attr_diff = git_attr("diff", 4);
1319 check[0].attr = attr_diff;
1322 static void diff_filespec_check_attr(struct diff_filespec *one)
1324 struct git_attr_check attr_diff_check;
1325 int check_from_data = 0;
1327 if (one->checked_attr)
1328 return;
1330 setup_diff_attr_check(&attr_diff_check);
1331 one->is_binary = 0;
1332 one->funcname_pattern_ident = NULL;
1334 if (!git_checkattr(one->path, 1, &attr_diff_check)) {
1335 const char *value;
1337 /* binaryness */
1338 value = attr_diff_check.value;
1339 if (ATTR_TRUE(value))
1341 else if (ATTR_FALSE(value))
1342 one->is_binary = 1;
1343 else
1344 check_from_data = 1;
1346 /* funcname pattern ident */
1347 if (ATTR_TRUE(value) || ATTR_FALSE(value) || ATTR_UNSET(value))
1349 else
1350 one->funcname_pattern_ident = value;
1353 if (check_from_data) {
1354 if (!one->data && DIFF_FILE_VALID(one))
1355 diff_populate_filespec(one, 0);
1357 if (one->data)
1358 one->is_binary = buffer_is_binary(one->data, one->size);
1362 int diff_filespec_is_binary(struct diff_filespec *one)
1364 diff_filespec_check_attr(one);
1365 return one->is_binary;
1368 static const char *funcname_pattern(const char *ident)
1370 struct funcname_pattern *pp;
1372 for (pp = funcname_pattern_list; pp; pp = pp->next)
1373 if (!strcmp(ident, pp->name))
1374 return pp->pattern;
1375 return NULL;
1378 static struct builtin_funcname_pattern {
1379 const char *name;
1380 const char *pattern;
1381 } builtin_funcname_pattern[] = {
1382 { "java", "!^[ ]*\\(catch\\|do\\|for\\|if\\|instanceof\\|"
1383 "new\\|return\\|switch\\|throw\\|while\\)\n"
1384 "^[ ]*\\(\\([ ]*"
1385 "[A-Za-z_][A-Za-z_0-9]*\\)\\{2,\\}"
1386 "[ ]*([^;]*\\)$" },
1387 { "tex", "^\\(\\\\\\(sub\\)*section{.*\\)$" },
1390 static const char *diff_funcname_pattern(struct diff_filespec *one)
1392 const char *ident, *pattern;
1393 int i;
1395 diff_filespec_check_attr(one);
1396 ident = one->funcname_pattern_ident;
1398 if (!ident)
1400 * If the config file has "funcname.default" defined, that
1401 * regexp is used; otherwise NULL is returned and xemit uses
1402 * the built-in default.
1404 return funcname_pattern("default");
1406 /* Look up custom "funcname.$ident" regexp from config. */
1407 pattern = funcname_pattern(ident);
1408 if (pattern)
1409 return pattern;
1412 * And define built-in fallback patterns here. Note that
1413 * these can be overridden by the user's config settings.
1415 for (i = 0; i < ARRAY_SIZE(builtin_funcname_pattern); i++)
1416 if (!strcmp(ident, builtin_funcname_pattern[i].name))
1417 return builtin_funcname_pattern[i].pattern;
1419 return NULL;
1422 static void builtin_diff(const char *name_a,
1423 const char *name_b,
1424 struct diff_filespec *one,
1425 struct diff_filespec *two,
1426 const char *xfrm_msg,
1427 struct diff_options *o,
1428 int complete_rewrite)
1430 mmfile_t mf1, mf2;
1431 const char *lbl[2];
1432 char *a_one, *b_two;
1433 const char *set = diff_get_color_opt(o, DIFF_METAINFO);
1434 const char *reset = diff_get_color_opt(o, DIFF_RESET);
1436 a_one = quote_two(o->a_prefix, name_a + (*name_a == '/'));
1437 b_two = quote_two(o->b_prefix, name_b + (*name_b == '/'));
1438 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
1439 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
1440 fprintf(o->file, "%sdiff --git %s %s%s\n", set, a_one, b_two, reset);
1441 if (lbl[0][0] == '/') {
1442 /* /dev/null */
1443 fprintf(o->file, "%snew file mode %06o%s\n", set, two->mode, reset);
1444 if (xfrm_msg && xfrm_msg[0])
1445 fprintf(o->file, "%s%s%s\n", set, xfrm_msg, reset);
1447 else if (lbl[1][0] == '/') {
1448 fprintf(o->file, "%sdeleted file mode %06o%s\n", set, one->mode, reset);
1449 if (xfrm_msg && xfrm_msg[0])
1450 fprintf(o->file, "%s%s%s\n", set, xfrm_msg, reset);
1452 else {
1453 if (one->mode != two->mode) {
1454 fprintf(o->file, "%sold mode %06o%s\n", set, one->mode, reset);
1455 fprintf(o->file, "%snew mode %06o%s\n", set, two->mode, reset);
1457 if (xfrm_msg && xfrm_msg[0])
1458 fprintf(o->file, "%s%s%s\n", set, xfrm_msg, reset);
1460 * we do not run diff between different kind
1461 * of objects.
1463 if ((one->mode ^ two->mode) & S_IFMT)
1464 goto free_ab_and_return;
1465 if (complete_rewrite) {
1466 emit_rewrite_diff(name_a, name_b, one, two, o);
1467 o->found_changes = 1;
1468 goto free_ab_and_return;
1472 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1473 die("unable to read files to diff");
1475 if (!DIFF_OPT_TST(o, TEXT) &&
1476 (diff_filespec_is_binary(one) || diff_filespec_is_binary(two))) {
1477 /* Quite common confusing case */
1478 if (mf1.size == mf2.size &&
1479 !memcmp(mf1.ptr, mf2.ptr, mf1.size))
1480 goto free_ab_and_return;
1481 if (DIFF_OPT_TST(o, BINARY))
1482 emit_binary_diff(o->file, &mf1, &mf2);
1483 else
1484 fprintf(o->file, "Binary files %s and %s differ\n",
1485 lbl[0], lbl[1]);
1486 o->found_changes = 1;
1488 else {
1489 /* Crazy xdl interfaces.. */
1490 const char *diffopts = getenv("GIT_DIFF_OPTS");
1491 xpparam_t xpp;
1492 xdemitconf_t xecfg;
1493 xdemitcb_t ecb;
1494 struct emit_callback ecbdata;
1495 const char *funcname_pattern;
1497 funcname_pattern = diff_funcname_pattern(one);
1498 if (!funcname_pattern)
1499 funcname_pattern = diff_funcname_pattern(two);
1501 memset(&xecfg, 0, sizeof(xecfg));
1502 memset(&ecbdata, 0, sizeof(ecbdata));
1503 ecbdata.label_path = lbl;
1504 ecbdata.color_diff = DIFF_OPT_TST(o, COLOR_DIFF);
1505 ecbdata.found_changesp = &o->found_changes;
1506 ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
1507 ecbdata.file = o->file;
1508 xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
1509 xecfg.ctxlen = o->context;
1510 xecfg.flags = XDL_EMIT_FUNCNAMES;
1511 if (funcname_pattern)
1512 xdiff_set_find_func(&xecfg, funcname_pattern);
1513 if (!diffopts)
1515 else if (!prefixcmp(diffopts, "--unified="))
1516 xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
1517 else if (!prefixcmp(diffopts, "-u"))
1518 xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
1519 ecb.outf = xdiff_outf;
1520 ecb.priv = &ecbdata;
1521 ecbdata.xm.consume = fn_out_consume;
1522 if (DIFF_OPT_TST(o, COLOR_DIFF_WORDS)) {
1523 ecbdata.diff_words =
1524 xcalloc(1, sizeof(struct diff_words_data));
1525 ecbdata.diff_words->file = o->file;
1527 xdi_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
1528 if (DIFF_OPT_TST(o, COLOR_DIFF_WORDS))
1529 free_diff_words_data(&ecbdata);
1532 free_ab_and_return:
1533 diff_free_filespec_data(one);
1534 diff_free_filespec_data(two);
1535 free(a_one);
1536 free(b_two);
1537 return;
1540 static void builtin_diffstat(const char *name_a, const char *name_b,
1541 struct diff_filespec *one,
1542 struct diff_filespec *two,
1543 struct diffstat_t *diffstat,
1544 struct diff_options *o,
1545 int complete_rewrite)
1547 mmfile_t mf1, mf2;
1548 struct diffstat_file *data;
1550 data = diffstat_add(diffstat, name_a, name_b);
1552 if (!one || !two) {
1553 data->is_unmerged = 1;
1554 return;
1556 if (complete_rewrite) {
1557 diff_populate_filespec(one, 0);
1558 diff_populate_filespec(two, 0);
1559 data->deleted = count_lines(one->data, one->size);
1560 data->added = count_lines(two->data, two->size);
1561 goto free_and_return;
1563 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1564 die("unable to read files to diff");
1566 if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
1567 data->is_binary = 1;
1568 data->added = mf2.size;
1569 data->deleted = mf1.size;
1570 } else {
1571 /* Crazy xdl interfaces.. */
1572 xpparam_t xpp;
1573 xdemitconf_t xecfg;
1574 xdemitcb_t ecb;
1576 memset(&xecfg, 0, sizeof(xecfg));
1577 xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
1578 ecb.outf = xdiff_outf;
1579 ecb.priv = diffstat;
1580 xdi_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
1583 free_and_return:
1584 diff_free_filespec_data(one);
1585 diff_free_filespec_data(two);
1588 static void builtin_checkdiff(const char *name_a, const char *name_b,
1589 const char *attr_path,
1590 struct diff_filespec *one,
1591 struct diff_filespec *two,
1592 struct diff_options *o)
1594 mmfile_t mf1, mf2;
1595 struct checkdiff_t data;
1597 if (!two)
1598 return;
1600 memset(&data, 0, sizeof(data));
1601 data.xm.consume = checkdiff_consume;
1602 data.filename = name_b ? name_b : name_a;
1603 data.lineno = 0;
1604 data.o = o;
1605 data.ws_rule = whitespace_rule(attr_path);
1607 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1608 die("unable to read files to diff");
1611 * All the other codepaths check both sides, but not checking
1612 * the "old" side here is deliberate. We are checking the newly
1613 * introduced changes, and as long as the "new" side is text, we
1614 * can and should check what it introduces.
1616 if (diff_filespec_is_binary(two))
1617 goto free_and_return;
1618 else {
1619 /* Crazy xdl interfaces.. */
1620 xpparam_t xpp;
1621 xdemitconf_t xecfg;
1622 xdemitcb_t ecb;
1624 memset(&xecfg, 0, sizeof(xecfg));
1625 xpp.flags = XDF_NEED_MINIMAL;
1626 ecb.outf = xdiff_outf;
1627 ecb.priv = &data;
1628 xdi_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
1630 if (data.trailing_blanks_start) {
1631 fprintf(o->file, "%s:%d: ends with blank lines.\n",
1632 data.filename, data.trailing_blanks_start);
1633 data.status = 1; /* report errors */
1636 free_and_return:
1637 diff_free_filespec_data(one);
1638 diff_free_filespec_data(two);
1639 if (data.status)
1640 DIFF_OPT_SET(o, CHECK_FAILED);
1643 struct diff_filespec *alloc_filespec(const char *path)
1645 int namelen = strlen(path);
1646 struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
1648 memset(spec, 0, sizeof(*spec));
1649 spec->path = (char *)(spec + 1);
1650 memcpy(spec->path, path, namelen+1);
1651 spec->count = 1;
1652 return spec;
1655 void free_filespec(struct diff_filespec *spec)
1657 if (!--spec->count) {
1658 diff_free_filespec_data(spec);
1659 free(spec);
1663 void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
1664 unsigned short mode)
1666 if (mode) {
1667 spec->mode = canon_mode(mode);
1668 hashcpy(spec->sha1, sha1);
1669 spec->sha1_valid = !is_null_sha1(sha1);
1674 * Given a name and sha1 pair, if the index tells us the file in
1675 * the work tree has that object contents, return true, so that
1676 * prepare_temp_file() does not have to inflate and extract.
1678 static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
1680 struct cache_entry *ce;
1681 struct stat st;
1682 int pos, len;
1684 /* We do not read the cache ourselves here, because the
1685 * benchmark with my previous version that always reads cache
1686 * shows that it makes things worse for diff-tree comparing
1687 * two linux-2.6 kernel trees in an already checked out work
1688 * tree. This is because most diff-tree comparisons deal with
1689 * only a small number of files, while reading the cache is
1690 * expensive for a large project, and its cost outweighs the
1691 * savings we get by not inflating the object to a temporary
1692 * file. Practically, this code only helps when we are used
1693 * by diff-cache --cached, which does read the cache before
1694 * calling us.
1696 if (!active_cache)
1697 return 0;
1699 /* We want to avoid the working directory if our caller
1700 * doesn't need the data in a normal file, this system
1701 * is rather slow with its stat/open/mmap/close syscalls,
1702 * and the object is contained in a pack file. The pack
1703 * is probably already open and will be faster to obtain
1704 * the data through than the working directory. Loose
1705 * objects however would tend to be slower as they need
1706 * to be individually opened and inflated.
1708 if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1, NULL))
1709 return 0;
1711 len = strlen(name);
1712 pos = cache_name_pos(name, len);
1713 if (pos < 0)
1714 return 0;
1715 ce = active_cache[pos];
1718 * This is not the sha1 we are looking for, or
1719 * unreusable because it is not a regular file.
1721 if (hashcmp(sha1, ce->sha1) || !S_ISREG(ce->ce_mode))
1722 return 0;
1725 * If ce matches the file in the work tree, we can reuse it.
1727 if (ce_uptodate(ce) ||
1728 (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
1729 return 1;
1731 return 0;
1734 static int populate_from_stdin(struct diff_filespec *s)
1736 struct strbuf buf;
1737 size_t size = 0;
1739 strbuf_init(&buf, 0);
1740 if (strbuf_read(&buf, 0, 0) < 0)
1741 return error("error while reading from stdin %s",
1742 strerror(errno));
1744 s->should_munmap = 0;
1745 s->data = strbuf_detach(&buf, &size);
1746 s->size = size;
1747 s->should_free = 1;
1748 return 0;
1751 static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
1753 int len;
1754 char *data = xmalloc(100);
1755 len = snprintf(data, 100,
1756 "Subproject commit %s\n", sha1_to_hex(s->sha1));
1757 s->data = data;
1758 s->size = len;
1759 s->should_free = 1;
1760 if (size_only) {
1761 s->data = NULL;
1762 free(data);
1764 return 0;
1768 * While doing rename detection and pickaxe operation, we may need to
1769 * grab the data for the blob (or file) for our own in-core comparison.
1770 * diff_filespec has data and size fields for this purpose.
1772 int diff_populate_filespec(struct diff_filespec *s, int size_only)
1774 int err = 0;
1775 if (!DIFF_FILE_VALID(s))
1776 die("internal error: asking to populate invalid file.");
1777 if (S_ISDIR(s->mode))
1778 return -1;
1780 if (s->data)
1781 return 0;
1783 if (size_only && 0 < s->size)
1784 return 0;
1786 if (S_ISGITLINK(s->mode))
1787 return diff_populate_gitlink(s, size_only);
1789 if (!s->sha1_valid ||
1790 reuse_worktree_file(s->path, s->sha1, 0)) {
1791 struct strbuf buf;
1792 struct stat st;
1793 int fd;
1795 if (!strcmp(s->path, "-"))
1796 return populate_from_stdin(s);
1798 if (lstat(s->path, &st) < 0) {
1799 if (errno == ENOENT) {
1800 err_empty:
1801 err = -1;
1802 empty:
1803 s->data = (char *)"";
1804 s->size = 0;
1805 return err;
1808 s->size = xsize_t(st.st_size);
1809 if (!s->size)
1810 goto empty;
1811 if (size_only)
1812 return 0;
1813 if (S_ISLNK(st.st_mode)) {
1814 int ret;
1815 s->data = xmalloc(s->size);
1816 s->should_free = 1;
1817 ret = readlink(s->path, s->data, s->size);
1818 if (ret < 0) {
1819 free(s->data);
1820 goto err_empty;
1822 return 0;
1824 fd = open(s->path, O_RDONLY);
1825 if (fd < 0)
1826 goto err_empty;
1827 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
1828 close(fd);
1829 s->should_munmap = 1;
1832 * Convert from working tree format to canonical git format
1834 strbuf_init(&buf, 0);
1835 if (convert_to_git(s->path, s->data, s->size, &buf, safe_crlf)) {
1836 size_t size = 0;
1837 munmap(s->data, s->size);
1838 s->should_munmap = 0;
1839 s->data = strbuf_detach(&buf, &size);
1840 s->size = size;
1841 s->should_free = 1;
1844 else {
1845 enum object_type type;
1846 if (size_only)
1847 type = sha1_object_info(s->sha1, &s->size);
1848 else {
1849 s->data = read_sha1_file(s->sha1, &type, &s->size);
1850 s->should_free = 1;
1853 return 0;
1856 void diff_free_filespec_blob(struct diff_filespec *s)
1858 if (s->should_free)
1859 free(s->data);
1860 else if (s->should_munmap)
1861 munmap(s->data, s->size);
1863 if (s->should_free || s->should_munmap) {
1864 s->should_free = s->should_munmap = 0;
1865 s->data = NULL;
1869 void diff_free_filespec_data(struct diff_filespec *s)
1871 diff_free_filespec_blob(s);
1872 free(s->cnt_data);
1873 s->cnt_data = NULL;
1876 static void prep_temp_blob(struct diff_tempfile *temp,
1877 void *blob,
1878 unsigned long size,
1879 const unsigned char *sha1,
1880 int mode)
1882 int fd;
1884 fd = git_mkstemp(temp->tmp_path, PATH_MAX, ".diff_XXXXXX");
1885 if (fd < 0)
1886 die("unable to create temp-file: %s", strerror(errno));
1887 if (write_in_full(fd, blob, size) != size)
1888 die("unable to write temp-file");
1889 close(fd);
1890 temp->name = temp->tmp_path;
1891 strcpy(temp->hex, sha1_to_hex(sha1));
1892 temp->hex[40] = 0;
1893 sprintf(temp->mode, "%06o", mode);
1896 static void prepare_temp_file(const char *name,
1897 struct diff_tempfile *temp,
1898 struct diff_filespec *one)
1900 if (!DIFF_FILE_VALID(one)) {
1901 not_a_valid_file:
1902 /* A '-' entry produces this for file-2, and
1903 * a '+' entry produces this for file-1.
1905 temp->name = "/dev/null";
1906 strcpy(temp->hex, ".");
1907 strcpy(temp->mode, ".");
1908 return;
1911 if (!one->sha1_valid ||
1912 reuse_worktree_file(name, one->sha1, 1)) {
1913 struct stat st;
1914 if (lstat(name, &st) < 0) {
1915 if (errno == ENOENT)
1916 goto not_a_valid_file;
1917 die("stat(%s): %s", name, strerror(errno));
1919 if (S_ISLNK(st.st_mode)) {
1920 int ret;
1921 char buf[PATH_MAX + 1]; /* ought to be SYMLINK_MAX */
1922 size_t sz = xsize_t(st.st_size);
1923 if (sizeof(buf) <= st.st_size)
1924 die("symlink too long: %s", name);
1925 ret = readlink(name, buf, sz);
1926 if (ret < 0)
1927 die("readlink(%s)", name);
1928 prep_temp_blob(temp, buf, sz,
1929 (one->sha1_valid ?
1930 one->sha1 : null_sha1),
1931 (one->sha1_valid ?
1932 one->mode : S_IFLNK));
1934 else {
1935 /* we can borrow from the file in the work tree */
1936 temp->name = name;
1937 if (!one->sha1_valid)
1938 strcpy(temp->hex, sha1_to_hex(null_sha1));
1939 else
1940 strcpy(temp->hex, sha1_to_hex(one->sha1));
1941 /* Even though we may sometimes borrow the
1942 * contents from the work tree, we always want
1943 * one->mode. mode is trustworthy even when
1944 * !(one->sha1_valid), as long as
1945 * DIFF_FILE_VALID(one).
1947 sprintf(temp->mode, "%06o", one->mode);
1949 return;
1951 else {
1952 if (diff_populate_filespec(one, 0))
1953 die("cannot read data blob for %s", one->path);
1954 prep_temp_blob(temp, one->data, one->size,
1955 one->sha1, one->mode);
1959 static void remove_tempfile(void)
1961 int i;
1963 for (i = 0; i < 2; i++)
1964 if (diff_temp[i].name == diff_temp[i].tmp_path) {
1965 unlink(diff_temp[i].name);
1966 diff_temp[i].name = NULL;
1970 static void remove_tempfile_on_signal(int signo)
1972 remove_tempfile();
1973 signal(SIGINT, SIG_DFL);
1974 raise(signo);
1977 /* An external diff command takes:
1979 * diff-cmd name infile1 infile1-sha1 infile1-mode \
1980 * infile2 infile2-sha1 infile2-mode [ rename-to ]
1983 static void run_external_diff(const char *pgm,
1984 const char *name,
1985 const char *other,
1986 struct diff_filespec *one,
1987 struct diff_filespec *two,
1988 const char *xfrm_msg,
1989 int complete_rewrite)
1991 const char *spawn_arg[10];
1992 struct diff_tempfile *temp = diff_temp;
1993 int retval;
1994 static int atexit_asked = 0;
1995 const char *othername;
1996 const char **arg = &spawn_arg[0];
1998 othername = (other? other : name);
1999 if (one && two) {
2000 prepare_temp_file(name, &temp[0], one);
2001 prepare_temp_file(othername, &temp[1], two);
2002 if (! atexit_asked &&
2003 (temp[0].name == temp[0].tmp_path ||
2004 temp[1].name == temp[1].tmp_path)) {
2005 atexit_asked = 1;
2006 atexit(remove_tempfile);
2008 signal(SIGINT, remove_tempfile_on_signal);
2011 if (one && two) {
2012 *arg++ = pgm;
2013 *arg++ = name;
2014 *arg++ = temp[0].name;
2015 *arg++ = temp[0].hex;
2016 *arg++ = temp[0].mode;
2017 *arg++ = temp[1].name;
2018 *arg++ = temp[1].hex;
2019 *arg++ = temp[1].mode;
2020 if (other) {
2021 *arg++ = other;
2022 *arg++ = xfrm_msg;
2024 } else {
2025 *arg++ = pgm;
2026 *arg++ = name;
2028 *arg = NULL;
2029 fflush(NULL);
2030 retval = run_command_v_opt(spawn_arg, 0);
2031 remove_tempfile();
2032 if (retval) {
2033 fprintf(stderr, "external diff died, stopping at %s.\n", name);
2034 exit(1);
2038 static const char *external_diff_attr(const char *name)
2040 struct git_attr_check attr_diff_check;
2042 if (!name)
2043 return NULL;
2045 setup_diff_attr_check(&attr_diff_check);
2046 if (!git_checkattr(name, 1, &attr_diff_check)) {
2047 const char *value = attr_diff_check.value;
2048 if (!ATTR_TRUE(value) &&
2049 !ATTR_FALSE(value) &&
2050 !ATTR_UNSET(value)) {
2051 struct ll_diff_driver *drv;
2053 for (drv = user_diff; drv; drv = drv->next)
2054 if (!strcmp(drv->name, value))
2055 return drv->cmd;
2058 return NULL;
2061 static void run_diff_cmd(const char *pgm,
2062 const char *name,
2063 const char *other,
2064 const char *attr_path,
2065 struct diff_filespec *one,
2066 struct diff_filespec *two,
2067 const char *xfrm_msg,
2068 struct diff_options *o,
2069 int complete_rewrite)
2071 if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
2072 pgm = NULL;
2073 else {
2074 const char *cmd = external_diff_attr(attr_path);
2075 if (cmd)
2076 pgm = cmd;
2079 if (pgm) {
2080 run_external_diff(pgm, name, other, one, two, xfrm_msg,
2081 complete_rewrite);
2082 return;
2084 if (one && two)
2085 builtin_diff(name, other ? other : name,
2086 one, two, xfrm_msg, o, complete_rewrite);
2087 else
2088 fprintf(o->file, "* Unmerged path %s\n", name);
2091 static void diff_fill_sha1_info(struct diff_filespec *one)
2093 if (DIFF_FILE_VALID(one)) {
2094 if (!one->sha1_valid) {
2095 struct stat st;
2096 if (!strcmp(one->path, "-")) {
2097 hashcpy(one->sha1, null_sha1);
2098 return;
2100 if (lstat(one->path, &st) < 0)
2101 die("stat %s", one->path);
2102 if (index_path(one->sha1, one->path, &st, 0))
2103 die("cannot hash %s\n", one->path);
2106 else
2107 hashclr(one->sha1);
2110 static int similarity_index(struct diff_filepair *p)
2112 return p->score * 100 / MAX_SCORE;
2115 static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
2117 /* Strip the prefix but do not molest /dev/null and absolute paths */
2118 if (*namep && **namep != '/')
2119 *namep += prefix_length;
2120 if (*otherp && **otherp != '/')
2121 *otherp += prefix_length;
2124 static void run_diff(struct diff_filepair *p, struct diff_options *o)
2126 const char *pgm = external_diff();
2127 struct strbuf msg;
2128 char *xfrm_msg;
2129 struct diff_filespec *one = p->one;
2130 struct diff_filespec *two = p->two;
2131 const char *name;
2132 const char *other;
2133 const char *attr_path;
2134 int complete_rewrite = 0;
2136 name = p->one->path;
2137 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2138 attr_path = name;
2139 if (o->prefix_length)
2140 strip_prefix(o->prefix_length, &name, &other);
2142 if (DIFF_PAIR_UNMERGED(p)) {
2143 run_diff_cmd(pgm, name, NULL, attr_path,
2144 NULL, NULL, NULL, o, 0);
2145 return;
2148 diff_fill_sha1_info(one);
2149 diff_fill_sha1_info(two);
2151 strbuf_init(&msg, PATH_MAX * 2 + 300);
2152 switch (p->status) {
2153 case DIFF_STATUS_COPIED:
2154 strbuf_addf(&msg, "similarity index %d%%", similarity_index(p));
2155 strbuf_addstr(&msg, "\ncopy from ");
2156 quote_c_style(name, &msg, NULL, 0);
2157 strbuf_addstr(&msg, "\ncopy to ");
2158 quote_c_style(other, &msg, NULL, 0);
2159 strbuf_addch(&msg, '\n');
2160 break;
2161 case DIFF_STATUS_RENAMED:
2162 strbuf_addf(&msg, "similarity index %d%%", similarity_index(p));
2163 strbuf_addstr(&msg, "\nrename from ");
2164 quote_c_style(name, &msg, NULL, 0);
2165 strbuf_addstr(&msg, "\nrename to ");
2166 quote_c_style(other, &msg, NULL, 0);
2167 strbuf_addch(&msg, '\n');
2168 break;
2169 case DIFF_STATUS_MODIFIED:
2170 if (p->score) {
2171 strbuf_addf(&msg, "dissimilarity index %d%%\n",
2172 similarity_index(p));
2173 complete_rewrite = 1;
2174 break;
2176 /* fallthru */
2177 default:
2178 /* nothing */
2182 if (hashcmp(one->sha1, two->sha1)) {
2183 int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
2185 if (DIFF_OPT_TST(o, BINARY)) {
2186 mmfile_t mf;
2187 if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
2188 (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
2189 abbrev = 40;
2191 strbuf_addf(&msg, "index %.*s..%.*s",
2192 abbrev, sha1_to_hex(one->sha1),
2193 abbrev, sha1_to_hex(two->sha1));
2194 if (one->mode == two->mode)
2195 strbuf_addf(&msg, " %06o", one->mode);
2196 strbuf_addch(&msg, '\n');
2199 if (msg.len)
2200 strbuf_setlen(&msg, msg.len - 1);
2201 xfrm_msg = msg.len ? msg.buf : NULL;
2203 if (!pgm &&
2204 DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
2205 (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
2206 /* a filepair that changes between file and symlink
2207 * needs to be split into deletion and creation.
2209 struct diff_filespec *null = alloc_filespec(two->path);
2210 run_diff_cmd(NULL, name, other, attr_path,
2211 one, null, xfrm_msg, o, 0);
2212 free(null);
2213 null = alloc_filespec(one->path);
2214 run_diff_cmd(NULL, name, other, attr_path,
2215 null, two, xfrm_msg, o, 0);
2216 free(null);
2218 else
2219 run_diff_cmd(pgm, name, other, attr_path,
2220 one, two, xfrm_msg, o, complete_rewrite);
2222 strbuf_release(&msg);
2225 static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
2226 struct diffstat_t *diffstat)
2228 const char *name;
2229 const char *other;
2230 int complete_rewrite = 0;
2232 if (DIFF_PAIR_UNMERGED(p)) {
2233 /* unmerged */
2234 builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, 0);
2235 return;
2238 name = p->one->path;
2239 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2241 if (o->prefix_length)
2242 strip_prefix(o->prefix_length, &name, &other);
2244 diff_fill_sha1_info(p->one);
2245 diff_fill_sha1_info(p->two);
2247 if (p->status == DIFF_STATUS_MODIFIED && p->score)
2248 complete_rewrite = 1;
2249 builtin_diffstat(name, other, p->one, p->two, diffstat, o, complete_rewrite);
2252 static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
2254 const char *name;
2255 const char *other;
2256 const char *attr_path;
2258 if (DIFF_PAIR_UNMERGED(p)) {
2259 /* unmerged */
2260 return;
2263 name = p->one->path;
2264 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2265 attr_path = other ? other : name;
2267 if (o->prefix_length)
2268 strip_prefix(o->prefix_length, &name, &other);
2270 diff_fill_sha1_info(p->one);
2271 diff_fill_sha1_info(p->two);
2273 builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
2276 void diff_setup(struct diff_options *options)
2278 memset(options, 0, sizeof(*options));
2280 options->file = stdout;
2282 options->line_termination = '\n';
2283 options->break_opt = -1;
2284 options->rename_limit = -1;
2285 options->dirstat_percent = 3;
2286 options->context = 3;
2288 options->change = diff_change;
2289 options->add_remove = diff_addremove;
2290 if (diff_use_color_default > 0)
2291 DIFF_OPT_SET(options, COLOR_DIFF);
2292 else
2293 DIFF_OPT_CLR(options, COLOR_DIFF);
2294 options->detect_rename = diff_detect_rename_default;
2296 options->a_prefix = "a/";
2297 options->b_prefix = "b/";
2300 int diff_setup_done(struct diff_options *options)
2302 int count = 0;
2304 if (options->output_format & DIFF_FORMAT_NAME)
2305 count++;
2306 if (options->output_format & DIFF_FORMAT_NAME_STATUS)
2307 count++;
2308 if (options->output_format & DIFF_FORMAT_CHECKDIFF)
2309 count++;
2310 if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
2311 count++;
2312 if (count > 1)
2313 die("--name-only, --name-status, --check and -s are mutually exclusive");
2315 if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
2316 options->detect_rename = DIFF_DETECT_COPY;
2318 if (!DIFF_OPT_TST(options, RELATIVE_NAME))
2319 options->prefix = NULL;
2320 if (options->prefix)
2321 options->prefix_length = strlen(options->prefix);
2322 else
2323 options->prefix_length = 0;
2325 if (options->output_format & (DIFF_FORMAT_NAME |
2326 DIFF_FORMAT_NAME_STATUS |
2327 DIFF_FORMAT_CHECKDIFF |
2328 DIFF_FORMAT_NO_OUTPUT))
2329 options->output_format &= ~(DIFF_FORMAT_RAW |
2330 DIFF_FORMAT_NUMSTAT |
2331 DIFF_FORMAT_DIFFSTAT |
2332 DIFF_FORMAT_SHORTSTAT |
2333 DIFF_FORMAT_DIRSTAT |
2334 DIFF_FORMAT_SUMMARY |
2335 DIFF_FORMAT_PATCH);
2338 * These cases always need recursive; we do not drop caller-supplied
2339 * recursive bits for other formats here.
2341 if (options->output_format & (DIFF_FORMAT_PATCH |
2342 DIFF_FORMAT_NUMSTAT |
2343 DIFF_FORMAT_DIFFSTAT |
2344 DIFF_FORMAT_SHORTSTAT |
2345 DIFF_FORMAT_DIRSTAT |
2346 DIFF_FORMAT_SUMMARY |
2347 DIFF_FORMAT_CHECKDIFF))
2348 DIFF_OPT_SET(options, RECURSIVE);
2350 * Also pickaxe would not work very well if you do not say recursive
2352 if (options->pickaxe)
2353 DIFF_OPT_SET(options, RECURSIVE);
2355 if (options->detect_rename && options->rename_limit < 0)
2356 options->rename_limit = diff_rename_limit_default;
2357 if (options->setup & DIFF_SETUP_USE_CACHE) {
2358 if (!active_cache)
2359 /* read-cache does not die even when it fails
2360 * so it is safe for us to do this here. Also
2361 * it does not smudge active_cache or active_nr
2362 * when it fails, so we do not have to worry about
2363 * cleaning it up ourselves either.
2365 read_cache();
2367 if (options->abbrev <= 0 || 40 < options->abbrev)
2368 options->abbrev = 40; /* full */
2371 * It does not make sense to show the first hit we happened
2372 * to have found. It does not make sense not to return with
2373 * exit code in such a case either.
2375 if (DIFF_OPT_TST(options, QUIET)) {
2376 options->output_format = DIFF_FORMAT_NO_OUTPUT;
2377 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
2381 * If we postprocess in diffcore, we cannot simply return
2382 * upon the first hit. We need to run diff as usual.
2384 if (options->pickaxe || options->filter)
2385 DIFF_OPT_CLR(options, QUIET);
2387 return 0;
2390 static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
2392 char c, *eq;
2393 int len;
2395 if (*arg != '-')
2396 return 0;
2397 c = *++arg;
2398 if (!c)
2399 return 0;
2400 if (c == arg_short) {
2401 c = *++arg;
2402 if (!c)
2403 return 1;
2404 if (val && isdigit(c)) {
2405 char *end;
2406 int n = strtoul(arg, &end, 10);
2407 if (*end)
2408 return 0;
2409 *val = n;
2410 return 1;
2412 return 0;
2414 if (c != '-')
2415 return 0;
2416 arg++;
2417 eq = strchr(arg, '=');
2418 if (eq)
2419 len = eq - arg;
2420 else
2421 len = strlen(arg);
2422 if (!len || strncmp(arg, arg_long, len))
2423 return 0;
2424 if (eq) {
2425 int n;
2426 char *end;
2427 if (!isdigit(*++eq))
2428 return 0;
2429 n = strtoul(eq, &end, 10);
2430 if (*end)
2431 return 0;
2432 *val = n;
2434 return 1;
2437 static int diff_scoreopt_parse(const char *opt);
2439 int diff_opt_parse(struct diff_options *options, const char **av, int ac)
2441 const char *arg = av[0];
2443 /* Output format options */
2444 if (!strcmp(arg, "-p") || !strcmp(arg, "-u"))
2445 options->output_format |= DIFF_FORMAT_PATCH;
2446 else if (opt_arg(arg, 'U', "unified", &options->context))
2447 options->output_format |= DIFF_FORMAT_PATCH;
2448 else if (!strcmp(arg, "--raw"))
2449 options->output_format |= DIFF_FORMAT_RAW;
2450 else if (!strcmp(arg, "--patch-with-raw"))
2451 options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_RAW;
2452 else if (!strcmp(arg, "--numstat"))
2453 options->output_format |= DIFF_FORMAT_NUMSTAT;
2454 else if (!strcmp(arg, "--shortstat"))
2455 options->output_format |= DIFF_FORMAT_SHORTSTAT;
2456 else if (opt_arg(arg, 'X', "dirstat", &options->dirstat_percent))
2457 options->output_format |= DIFF_FORMAT_DIRSTAT;
2458 else if (!strcmp(arg, "--cumulative"))
2459 options->output_format |= DIFF_FORMAT_CUMULATIVE;
2460 else if (!strcmp(arg, "--check"))
2461 options->output_format |= DIFF_FORMAT_CHECKDIFF;
2462 else if (!strcmp(arg, "--summary"))
2463 options->output_format |= DIFF_FORMAT_SUMMARY;
2464 else if (!strcmp(arg, "--patch-with-stat"))
2465 options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_DIFFSTAT;
2466 else if (!strcmp(arg, "--name-only"))
2467 options->output_format |= DIFF_FORMAT_NAME;
2468 else if (!strcmp(arg, "--name-status"))
2469 options->output_format |= DIFF_FORMAT_NAME_STATUS;
2470 else if (!strcmp(arg, "-s"))
2471 options->output_format |= DIFF_FORMAT_NO_OUTPUT;
2472 else if (!prefixcmp(arg, "--stat")) {
2473 char *end;
2474 int width = options->stat_width;
2475 int name_width = options->stat_name_width;
2476 arg += 6;
2477 end = (char *)arg;
2479 switch (*arg) {
2480 case '-':
2481 if (!prefixcmp(arg, "-width="))
2482 width = strtoul(arg + 7, &end, 10);
2483 else if (!prefixcmp(arg, "-name-width="))
2484 name_width = strtoul(arg + 12, &end, 10);
2485 break;
2486 case '=':
2487 width = strtoul(arg+1, &end, 10);
2488 if (*end == ',')
2489 name_width = strtoul(end+1, &end, 10);
2492 /* Important! This checks all the error cases! */
2493 if (*end)
2494 return 0;
2495 options->output_format |= DIFF_FORMAT_DIFFSTAT;
2496 options->stat_name_width = name_width;
2497 options->stat_width = width;
2500 /* renames options */
2501 else if (!prefixcmp(arg, "-B")) {
2502 if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
2503 return -1;
2505 else if (!prefixcmp(arg, "-M")) {
2506 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
2507 return -1;
2508 options->detect_rename = DIFF_DETECT_RENAME;
2510 else if (!prefixcmp(arg, "-C")) {
2511 if (options->detect_rename == DIFF_DETECT_COPY)
2512 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
2513 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
2514 return -1;
2515 options->detect_rename = DIFF_DETECT_COPY;
2517 else if (!strcmp(arg, "--no-renames"))
2518 options->detect_rename = 0;
2519 else if (!strcmp(arg, "--relative"))
2520 DIFF_OPT_SET(options, RELATIVE_NAME);
2521 else if (!prefixcmp(arg, "--relative=")) {
2522 DIFF_OPT_SET(options, RELATIVE_NAME);
2523 options->prefix = arg + 11;
2526 /* xdiff options */
2527 else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
2528 options->xdl_opts |= XDF_IGNORE_WHITESPACE;
2529 else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
2530 options->xdl_opts |= XDF_IGNORE_WHITESPACE_CHANGE;
2531 else if (!strcmp(arg, "--ignore-space-at-eol"))
2532 options->xdl_opts |= XDF_IGNORE_WHITESPACE_AT_EOL;
2534 /* flags options */
2535 else if (!strcmp(arg, "--binary")) {
2536 options->output_format |= DIFF_FORMAT_PATCH;
2537 DIFF_OPT_SET(options, BINARY);
2539 else if (!strcmp(arg, "--full-index"))
2540 DIFF_OPT_SET(options, FULL_INDEX);
2541 else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
2542 DIFF_OPT_SET(options, TEXT);
2543 else if (!strcmp(arg, "-R"))
2544 DIFF_OPT_SET(options, REVERSE_DIFF);
2545 else if (!strcmp(arg, "--find-copies-harder"))
2546 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
2547 else if (!strcmp(arg, "--follow"))
2548 DIFF_OPT_SET(options, FOLLOW_RENAMES);
2549 else if (!strcmp(arg, "--color"))
2550 DIFF_OPT_SET(options, COLOR_DIFF);
2551 else if (!strcmp(arg, "--no-color"))
2552 DIFF_OPT_CLR(options, COLOR_DIFF);
2553 else if (!strcmp(arg, "--color-words"))
2554 options->flags |= DIFF_OPT_COLOR_DIFF | DIFF_OPT_COLOR_DIFF_WORDS;
2555 else if (!strcmp(arg, "--exit-code"))
2556 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
2557 else if (!strcmp(arg, "--quiet"))
2558 DIFF_OPT_SET(options, QUIET);
2559 else if (!strcmp(arg, "--ext-diff"))
2560 DIFF_OPT_SET(options, ALLOW_EXTERNAL);
2561 else if (!strcmp(arg, "--no-ext-diff"))
2562 DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
2563 else if (!strcmp(arg, "--ignore-submodules"))
2564 DIFF_OPT_SET(options, IGNORE_SUBMODULES);
2566 /* misc options */
2567 else if (!strcmp(arg, "-z"))
2568 options->line_termination = 0;
2569 else if (!prefixcmp(arg, "-l"))
2570 options->rename_limit = strtoul(arg+2, NULL, 10);
2571 else if (!prefixcmp(arg, "-S"))
2572 options->pickaxe = arg + 2;
2573 else if (!strcmp(arg, "--pickaxe-all"))
2574 options->pickaxe_opts = DIFF_PICKAXE_ALL;
2575 else if (!strcmp(arg, "--pickaxe-regex"))
2576 options->pickaxe_opts = DIFF_PICKAXE_REGEX;
2577 else if (!prefixcmp(arg, "-O"))
2578 options->orderfile = arg + 2;
2579 else if (!prefixcmp(arg, "--diff-filter="))
2580 options->filter = arg + 14;
2581 else if (!strcmp(arg, "--abbrev"))
2582 options->abbrev = DEFAULT_ABBREV;
2583 else if (!prefixcmp(arg, "--abbrev=")) {
2584 options->abbrev = strtoul(arg + 9, NULL, 10);
2585 if (options->abbrev < MINIMUM_ABBREV)
2586 options->abbrev = MINIMUM_ABBREV;
2587 else if (40 < options->abbrev)
2588 options->abbrev = 40;
2590 else if (!prefixcmp(arg, "--src-prefix="))
2591 options->a_prefix = arg + 13;
2592 else if (!prefixcmp(arg, "--dst-prefix="))
2593 options->b_prefix = arg + 13;
2594 else if (!strcmp(arg, "--no-prefix"))
2595 options->a_prefix = options->b_prefix = "";
2596 else if (!prefixcmp(arg, "--output=")) {
2597 options->file = fopen(arg + strlen("--output="), "w");
2598 options->close_file = 1;
2599 } else
2600 return 0;
2601 return 1;
2604 static int parse_num(const char **cp_p)
2606 unsigned long num, scale;
2607 int ch, dot;
2608 const char *cp = *cp_p;
2610 num = 0;
2611 scale = 1;
2612 dot = 0;
2613 for(;;) {
2614 ch = *cp;
2615 if ( !dot && ch == '.' ) {
2616 scale = 1;
2617 dot = 1;
2618 } else if ( ch == '%' ) {
2619 scale = dot ? scale*100 : 100;
2620 cp++; /* % is always at the end */
2621 break;
2622 } else if ( ch >= '0' && ch <= '9' ) {
2623 if ( scale < 100000 ) {
2624 scale *= 10;
2625 num = (num*10) + (ch-'0');
2627 } else {
2628 break;
2630 cp++;
2632 *cp_p = cp;
2634 /* user says num divided by scale and we say internally that
2635 * is MAX_SCORE * num / scale.
2637 return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
2640 static int diff_scoreopt_parse(const char *opt)
2642 int opt1, opt2, cmd;
2644 if (*opt++ != '-')
2645 return -1;
2646 cmd = *opt++;
2647 if (cmd != 'M' && cmd != 'C' && cmd != 'B')
2648 return -1; /* that is not a -M, -C nor -B option */
2650 opt1 = parse_num(&opt);
2651 if (cmd != 'B')
2652 opt2 = 0;
2653 else {
2654 if (*opt == 0)
2655 opt2 = 0;
2656 else if (*opt != '/')
2657 return -1; /* we expect -B80/99 or -B80 */
2658 else {
2659 opt++;
2660 opt2 = parse_num(&opt);
2663 if (*opt != 0)
2664 return -1;
2665 return opt1 | (opt2 << 16);
2668 struct diff_queue_struct diff_queued_diff;
2670 void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
2672 if (queue->alloc <= queue->nr) {
2673 queue->alloc = alloc_nr(queue->alloc);
2674 queue->queue = xrealloc(queue->queue,
2675 sizeof(dp) * queue->alloc);
2677 queue->queue[queue->nr++] = dp;
2680 struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
2681 struct diff_filespec *one,
2682 struct diff_filespec *two)
2684 struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
2685 dp->one = one;
2686 dp->two = two;
2687 if (queue)
2688 diff_q(queue, dp);
2689 return dp;
2692 void diff_free_filepair(struct diff_filepair *p)
2694 free_filespec(p->one);
2695 free_filespec(p->two);
2696 free(p);
2699 /* This is different from find_unique_abbrev() in that
2700 * it stuffs the result with dots for alignment.
2702 const char *diff_unique_abbrev(const unsigned char *sha1, int len)
2704 int abblen;
2705 const char *abbrev;
2706 if (len == 40)
2707 return sha1_to_hex(sha1);
2709 abbrev = find_unique_abbrev(sha1, len);
2710 abblen = strlen(abbrev);
2711 if (abblen < 37) {
2712 static char hex[41];
2713 if (len < abblen && abblen <= len + 2)
2714 sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
2715 else
2716 sprintf(hex, "%s...", abbrev);
2717 return hex;
2719 return sha1_to_hex(sha1);
2722 static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
2724 int line_termination = opt->line_termination;
2725 int inter_name_termination = line_termination ? '\t' : '\0';
2727 if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
2728 fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
2729 diff_unique_abbrev(p->one->sha1, opt->abbrev));
2730 fprintf(opt->file, "%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
2732 if (p->score) {
2733 fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
2734 inter_name_termination);
2735 } else {
2736 fprintf(opt->file, "%c%c", p->status, inter_name_termination);
2739 if (p->status == DIFF_STATUS_COPIED ||
2740 p->status == DIFF_STATUS_RENAMED) {
2741 const char *name_a, *name_b;
2742 name_a = p->one->path;
2743 name_b = p->two->path;
2744 strip_prefix(opt->prefix_length, &name_a, &name_b);
2745 write_name_quoted(name_a, opt->file, inter_name_termination);
2746 write_name_quoted(name_b, opt->file, line_termination);
2747 } else {
2748 const char *name_a, *name_b;
2749 name_a = p->one->mode ? p->one->path : p->two->path;
2750 name_b = NULL;
2751 strip_prefix(opt->prefix_length, &name_a, &name_b);
2752 write_name_quoted(name_a, opt->file, line_termination);
2756 int diff_unmodified_pair(struct diff_filepair *p)
2758 /* This function is written stricter than necessary to support
2759 * the currently implemented transformers, but the idea is to
2760 * let transformers to produce diff_filepairs any way they want,
2761 * and filter and clean them up here before producing the output.
2763 struct diff_filespec *one = p->one, *two = p->two;
2765 if (DIFF_PAIR_UNMERGED(p))
2766 return 0; /* unmerged is interesting */
2768 /* deletion, addition, mode or type change
2769 * and rename are all interesting.
2771 if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
2772 DIFF_PAIR_MODE_CHANGED(p) ||
2773 strcmp(one->path, two->path))
2774 return 0;
2776 /* both are valid and point at the same path. that is, we are
2777 * dealing with a change.
2779 if (one->sha1_valid && two->sha1_valid &&
2780 !hashcmp(one->sha1, two->sha1))
2781 return 1; /* no change */
2782 if (!one->sha1_valid && !two->sha1_valid)
2783 return 1; /* both look at the same file on the filesystem. */
2784 return 0;
2787 static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
2789 if (diff_unmodified_pair(p))
2790 return;
2792 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2793 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2794 return; /* no tree diffs in patch format */
2796 run_diff(p, o);
2799 static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
2800 struct diffstat_t *diffstat)
2802 if (diff_unmodified_pair(p))
2803 return;
2805 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2806 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2807 return; /* no tree diffs in patch format */
2809 run_diffstat(p, o, diffstat);
2812 static void diff_flush_checkdiff(struct diff_filepair *p,
2813 struct diff_options *o)
2815 if (diff_unmodified_pair(p))
2816 return;
2818 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2819 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2820 return; /* no tree diffs in patch format */
2822 run_checkdiff(p, o);
2825 int diff_queue_is_empty(void)
2827 struct diff_queue_struct *q = &diff_queued_diff;
2828 int i;
2829 for (i = 0; i < q->nr; i++)
2830 if (!diff_unmodified_pair(q->queue[i]))
2831 return 0;
2832 return 1;
2835 #if DIFF_DEBUG
2836 void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
2838 fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
2839 x, one ? one : "",
2840 s->path,
2841 DIFF_FILE_VALID(s) ? "valid" : "invalid",
2842 s->mode,
2843 s->sha1_valid ? sha1_to_hex(s->sha1) : "");
2844 fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
2845 x, one ? one : "",
2846 s->size, s->xfrm_flags);
2849 void diff_debug_filepair(const struct diff_filepair *p, int i)
2851 diff_debug_filespec(p->one, i, "one");
2852 diff_debug_filespec(p->two, i, "two");
2853 fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
2854 p->score, p->status ? p->status : '?',
2855 p->one->rename_used, p->broken_pair);
2858 void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
2860 int i;
2861 if (msg)
2862 fprintf(stderr, "%s\n", msg);
2863 fprintf(stderr, "q->nr = %d\n", q->nr);
2864 for (i = 0; i < q->nr; i++) {
2865 struct diff_filepair *p = q->queue[i];
2866 diff_debug_filepair(p, i);
2869 #endif
2871 static void diff_resolve_rename_copy(void)
2873 int i;
2874 struct diff_filepair *p;
2875 struct diff_queue_struct *q = &diff_queued_diff;
2877 diff_debug_queue("resolve-rename-copy", q);
2879 for (i = 0; i < q->nr; i++) {
2880 p = q->queue[i];
2881 p->status = 0; /* undecided */
2882 if (DIFF_PAIR_UNMERGED(p))
2883 p->status = DIFF_STATUS_UNMERGED;
2884 else if (!DIFF_FILE_VALID(p->one))
2885 p->status = DIFF_STATUS_ADDED;
2886 else if (!DIFF_FILE_VALID(p->two))
2887 p->status = DIFF_STATUS_DELETED;
2888 else if (DIFF_PAIR_TYPE_CHANGED(p))
2889 p->status = DIFF_STATUS_TYPE_CHANGED;
2891 /* from this point on, we are dealing with a pair
2892 * whose both sides are valid and of the same type, i.e.
2893 * either in-place edit or rename/copy edit.
2895 else if (DIFF_PAIR_RENAME(p)) {
2897 * A rename might have re-connected a broken
2898 * pair up, causing the pathnames to be the
2899 * same again. If so, that's not a rename at
2900 * all, just a modification..
2902 * Otherwise, see if this source was used for
2903 * multiple renames, in which case we decrement
2904 * the count, and call it a copy.
2906 if (!strcmp(p->one->path, p->two->path))
2907 p->status = DIFF_STATUS_MODIFIED;
2908 else if (--p->one->rename_used > 0)
2909 p->status = DIFF_STATUS_COPIED;
2910 else
2911 p->status = DIFF_STATUS_RENAMED;
2913 else if (hashcmp(p->one->sha1, p->two->sha1) ||
2914 p->one->mode != p->two->mode ||
2915 is_null_sha1(p->one->sha1))
2916 p->status = DIFF_STATUS_MODIFIED;
2917 else {
2918 /* This is a "no-change" entry and should not
2919 * happen anymore, but prepare for broken callers.
2921 error("feeding unmodified %s to diffcore",
2922 p->one->path);
2923 p->status = DIFF_STATUS_UNKNOWN;
2926 diff_debug_queue("resolve-rename-copy done", q);
2929 static int check_pair_status(struct diff_filepair *p)
2931 switch (p->status) {
2932 case DIFF_STATUS_UNKNOWN:
2933 return 0;
2934 case 0:
2935 die("internal error in diff-resolve-rename-copy");
2936 default:
2937 return 1;
2941 static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
2943 int fmt = opt->output_format;
2945 if (fmt & DIFF_FORMAT_CHECKDIFF)
2946 diff_flush_checkdiff(p, opt);
2947 else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
2948 diff_flush_raw(p, opt);
2949 else if (fmt & DIFF_FORMAT_NAME) {
2950 const char *name_a, *name_b;
2951 name_a = p->two->path;
2952 name_b = NULL;
2953 strip_prefix(opt->prefix_length, &name_a, &name_b);
2954 write_name_quoted(name_a, opt->file, opt->line_termination);
2958 static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
2960 if (fs->mode)
2961 fprintf(file, " %s mode %06o ", newdelete, fs->mode);
2962 else
2963 fprintf(file, " %s ", newdelete);
2964 write_name_quoted(fs->path, file, '\n');
2968 static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name)
2970 if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
2971 fprintf(file, " mode change %06o => %06o%c", p->one->mode, p->two->mode,
2972 show_name ? ' ' : '\n');
2973 if (show_name) {
2974 write_name_quoted(p->two->path, file, '\n');
2979 static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p)
2981 char *names = pprint_rename(p->one->path, p->two->path);
2983 fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
2984 free(names);
2985 show_mode_change(file, p, 0);
2988 static void diff_summary(FILE *file, struct diff_filepair *p)
2990 switch(p->status) {
2991 case DIFF_STATUS_DELETED:
2992 show_file_mode_name(file, "delete", p->one);
2993 break;
2994 case DIFF_STATUS_ADDED:
2995 show_file_mode_name(file, "create", p->two);
2996 break;
2997 case DIFF_STATUS_COPIED:
2998 show_rename_copy(file, "copy", p);
2999 break;
3000 case DIFF_STATUS_RENAMED:
3001 show_rename_copy(file, "rename", p);
3002 break;
3003 default:
3004 if (p->score) {
3005 fputs(" rewrite ", file);
3006 write_name_quoted(p->two->path, file, ' ');
3007 fprintf(file, "(%d%%)\n", similarity_index(p));
3009 show_mode_change(file, p, !p->score);
3010 break;
3014 struct patch_id_t {
3015 struct xdiff_emit_state xm;
3016 SHA_CTX *ctx;
3017 int patchlen;
3020 static int remove_space(char *line, int len)
3022 int i;
3023 char *dst = line;
3024 unsigned char c;
3026 for (i = 0; i < len; i++)
3027 if (!isspace((c = line[i])))
3028 *dst++ = c;
3030 return dst - line;
3033 static void patch_id_consume(void *priv, char *line, unsigned long len)
3035 struct patch_id_t *data = priv;
3036 int new_len;
3038 /* Ignore line numbers when computing the SHA1 of the patch */
3039 if (!prefixcmp(line, "@@ -"))
3040 return;
3042 new_len = remove_space(line, len);
3044 SHA1_Update(data->ctx, line, new_len);
3045 data->patchlen += new_len;
3048 /* returns 0 upon success, and writes result into sha1 */
3049 static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
3051 struct diff_queue_struct *q = &diff_queued_diff;
3052 int i;
3053 SHA_CTX ctx;
3054 struct patch_id_t data;
3055 char buffer[PATH_MAX * 4 + 20];
3057 SHA1_Init(&ctx);
3058 memset(&data, 0, sizeof(struct patch_id_t));
3059 data.ctx = &ctx;
3060 data.xm.consume = patch_id_consume;
3062 for (i = 0; i < q->nr; i++) {
3063 xpparam_t xpp;
3064 xdemitconf_t xecfg;
3065 xdemitcb_t ecb;
3066 mmfile_t mf1, mf2;
3067 struct diff_filepair *p = q->queue[i];
3068 int len1, len2;
3070 memset(&xecfg, 0, sizeof(xecfg));
3071 if (p->status == 0)
3072 return error("internal diff status error");
3073 if (p->status == DIFF_STATUS_UNKNOWN)
3074 continue;
3075 if (diff_unmodified_pair(p))
3076 continue;
3077 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3078 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3079 continue;
3080 if (DIFF_PAIR_UNMERGED(p))
3081 continue;
3083 diff_fill_sha1_info(p->one);
3084 diff_fill_sha1_info(p->two);
3085 if (fill_mmfile(&mf1, p->one) < 0 ||
3086 fill_mmfile(&mf2, p->two) < 0)
3087 return error("unable to read files to diff");
3089 len1 = remove_space(p->one->path, strlen(p->one->path));
3090 len2 = remove_space(p->two->path, strlen(p->two->path));
3091 if (p->one->mode == 0)
3092 len1 = snprintf(buffer, sizeof(buffer),
3093 "diff--gita/%.*sb/%.*s"
3094 "newfilemode%06o"
3095 "---/dev/null"
3096 "+++b/%.*s",
3097 len1, p->one->path,
3098 len2, p->two->path,
3099 p->two->mode,
3100 len2, p->two->path);
3101 else if (p->two->mode == 0)
3102 len1 = snprintf(buffer, sizeof(buffer),
3103 "diff--gita/%.*sb/%.*s"
3104 "deletedfilemode%06o"
3105 "---a/%.*s"
3106 "+++/dev/null",
3107 len1, p->one->path,
3108 len2, p->two->path,
3109 p->one->mode,
3110 len1, p->one->path);
3111 else
3112 len1 = snprintf(buffer, sizeof(buffer),
3113 "diff--gita/%.*sb/%.*s"
3114 "---a/%.*s"
3115 "+++b/%.*s",
3116 len1, p->one->path,
3117 len2, p->two->path,
3118 len1, p->one->path,
3119 len2, p->two->path);
3120 SHA1_Update(&ctx, buffer, len1);
3122 xpp.flags = XDF_NEED_MINIMAL;
3123 xecfg.ctxlen = 3;
3124 xecfg.flags = XDL_EMIT_FUNCNAMES;
3125 ecb.outf = xdiff_outf;
3126 ecb.priv = &data;
3127 xdi_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
3130 SHA1_Final(sha1, &ctx);
3131 return 0;
3134 int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
3136 struct diff_queue_struct *q = &diff_queued_diff;
3137 int i;
3138 int result = diff_get_patch_id(options, sha1);
3140 for (i = 0; i < q->nr; i++)
3141 diff_free_filepair(q->queue[i]);
3143 free(q->queue);
3144 q->queue = NULL;
3145 q->nr = q->alloc = 0;
3147 return result;
3150 static int is_summary_empty(const struct diff_queue_struct *q)
3152 int i;
3154 for (i = 0; i < q->nr; i++) {
3155 const struct diff_filepair *p = q->queue[i];
3157 switch (p->status) {
3158 case DIFF_STATUS_DELETED:
3159 case DIFF_STATUS_ADDED:
3160 case DIFF_STATUS_COPIED:
3161 case DIFF_STATUS_RENAMED:
3162 return 0;
3163 default:
3164 if (p->score)
3165 return 0;
3166 if (p->one->mode && p->two->mode &&
3167 p->one->mode != p->two->mode)
3168 return 0;
3169 break;
3172 return 1;
3175 void diff_flush(struct diff_options *options)
3177 struct diff_queue_struct *q = &diff_queued_diff;
3178 int i, output_format = options->output_format;
3179 int separator = 0;
3182 * Order: raw, stat, summary, patch
3183 * or: name/name-status/checkdiff (other bits clear)
3185 if (!q->nr)
3186 goto free_queue;
3188 if (output_format & (DIFF_FORMAT_RAW |
3189 DIFF_FORMAT_NAME |
3190 DIFF_FORMAT_NAME_STATUS |
3191 DIFF_FORMAT_CHECKDIFF)) {
3192 for (i = 0; i < q->nr; i++) {
3193 struct diff_filepair *p = q->queue[i];
3194 if (check_pair_status(p))
3195 flush_one_pair(p, options);
3197 separator++;
3200 if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT)) {
3201 struct diffstat_t diffstat;
3203 memset(&diffstat, 0, sizeof(struct diffstat_t));
3204 diffstat.xm.consume = diffstat_consume;
3205 for (i = 0; i < q->nr; i++) {
3206 struct diff_filepair *p = q->queue[i];
3207 if (check_pair_status(p))
3208 diff_flush_stat(p, options, &diffstat);
3210 if (output_format & DIFF_FORMAT_NUMSTAT)
3211 show_numstat(&diffstat, options);
3212 if (output_format & DIFF_FORMAT_DIFFSTAT)
3213 show_stats(&diffstat, options);
3214 if (output_format & DIFF_FORMAT_SHORTSTAT)
3215 show_shortstats(&diffstat, options);
3216 free_diffstat_info(&diffstat);
3217 separator++;
3219 if (output_format & DIFF_FORMAT_DIRSTAT)
3220 show_dirstat(options);
3222 if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
3223 for (i = 0; i < q->nr; i++)
3224 diff_summary(options->file, q->queue[i]);
3225 separator++;
3228 if (output_format & DIFF_FORMAT_PATCH) {
3229 if (separator) {
3230 if (options->stat_sep) {
3231 /* attach patch instead of inline */
3232 fputs(options->stat_sep, options->file);
3233 } else {
3234 putc(options->line_termination, options->file);
3238 for (i = 0; i < q->nr; i++) {
3239 struct diff_filepair *p = q->queue[i];
3240 if (check_pair_status(p))
3241 diff_flush_patch(p, options);
3245 if (output_format & DIFF_FORMAT_CALLBACK)
3246 options->format_callback(q, options, options->format_callback_data);
3248 for (i = 0; i < q->nr; i++)
3249 diff_free_filepair(q->queue[i]);
3250 free_queue:
3251 free(q->queue);
3252 q->queue = NULL;
3253 q->nr = q->alloc = 0;
3254 if (options->close_file)
3255 fclose(options->file);
3258 static void diffcore_apply_filter(const char *filter)
3260 int i;
3261 struct diff_queue_struct *q = &diff_queued_diff;
3262 struct diff_queue_struct outq;
3263 outq.queue = NULL;
3264 outq.nr = outq.alloc = 0;
3266 if (!filter)
3267 return;
3269 if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
3270 int found;
3271 for (i = found = 0; !found && i < q->nr; i++) {
3272 struct diff_filepair *p = q->queue[i];
3273 if (((p->status == DIFF_STATUS_MODIFIED) &&
3274 ((p->score &&
3275 strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
3276 (!p->score &&
3277 strchr(filter, DIFF_STATUS_MODIFIED)))) ||
3278 ((p->status != DIFF_STATUS_MODIFIED) &&
3279 strchr(filter, p->status)))
3280 found++;
3282 if (found)
3283 return;
3285 /* otherwise we will clear the whole queue
3286 * by copying the empty outq at the end of this
3287 * function, but first clear the current entries
3288 * in the queue.
3290 for (i = 0; i < q->nr; i++)
3291 diff_free_filepair(q->queue[i]);
3293 else {
3294 /* Only the matching ones */
3295 for (i = 0; i < q->nr; i++) {
3296 struct diff_filepair *p = q->queue[i];
3298 if (((p->status == DIFF_STATUS_MODIFIED) &&
3299 ((p->score &&
3300 strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
3301 (!p->score &&
3302 strchr(filter, DIFF_STATUS_MODIFIED)))) ||
3303 ((p->status != DIFF_STATUS_MODIFIED) &&
3304 strchr(filter, p->status)))
3305 diff_q(&outq, p);
3306 else
3307 diff_free_filepair(p);
3310 free(q->queue);
3311 *q = outq;
3314 /* Check whether two filespecs with the same mode and size are identical */
3315 static int diff_filespec_is_identical(struct diff_filespec *one,
3316 struct diff_filespec *two)
3318 if (S_ISGITLINK(one->mode))
3319 return 0;
3320 if (diff_populate_filespec(one, 0))
3321 return 0;
3322 if (diff_populate_filespec(two, 0))
3323 return 0;
3324 return !memcmp(one->data, two->data, one->size);
3327 static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
3329 int i;
3330 struct diff_queue_struct *q = &diff_queued_diff;
3331 struct diff_queue_struct outq;
3332 outq.queue = NULL;
3333 outq.nr = outq.alloc = 0;
3335 for (i = 0; i < q->nr; i++) {
3336 struct diff_filepair *p = q->queue[i];
3339 * 1. Entries that come from stat info dirtyness
3340 * always have both sides (iow, not create/delete),
3341 * one side of the object name is unknown, with
3342 * the same mode and size. Keep the ones that
3343 * do not match these criteria. They have real
3344 * differences.
3346 * 2. At this point, the file is known to be modified,
3347 * with the same mode and size, and the object
3348 * name of one side is unknown. Need to inspect
3349 * the identical contents.
3351 if (!DIFF_FILE_VALID(p->one) || /* (1) */
3352 !DIFF_FILE_VALID(p->two) ||
3353 (p->one->sha1_valid && p->two->sha1_valid) ||
3354 (p->one->mode != p->two->mode) ||
3355 diff_populate_filespec(p->one, 1) ||
3356 diff_populate_filespec(p->two, 1) ||
3357 (p->one->size != p->two->size) ||
3358 !diff_filespec_is_identical(p->one, p->two)) /* (2) */
3359 diff_q(&outq, p);
3360 else {
3362 * The caller can subtract 1 from skip_stat_unmatch
3363 * to determine how many paths were dirty only
3364 * due to stat info mismatch.
3366 if (!DIFF_OPT_TST(diffopt, NO_INDEX))
3367 diffopt->skip_stat_unmatch++;
3368 diff_free_filepair(p);
3371 free(q->queue);
3372 *q = outq;
3375 void diffcore_std(struct diff_options *options)
3377 if (DIFF_OPT_TST(options, QUIET))
3378 return;
3380 if (options->skip_stat_unmatch && !DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3381 diffcore_skip_stat_unmatch(options);
3382 if (options->break_opt != -1)
3383 diffcore_break(options->break_opt);
3384 if (options->detect_rename)
3385 diffcore_rename(options);
3386 if (options->break_opt != -1)
3387 diffcore_merge_broken();
3388 if (options->pickaxe)
3389 diffcore_pickaxe(options->pickaxe, options->pickaxe_opts);
3390 if (options->orderfile)
3391 diffcore_order(options->orderfile);
3392 diff_resolve_rename_copy();
3393 diffcore_apply_filter(options->filter);
3395 if (diff_queued_diff.nr)
3396 DIFF_OPT_SET(options, HAS_CHANGES);
3397 else
3398 DIFF_OPT_CLR(options, HAS_CHANGES);
3401 int diff_result_code(struct diff_options *opt, int status)
3403 int result = 0;
3404 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
3405 !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
3406 return status;
3407 if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
3408 DIFF_OPT_TST(opt, HAS_CHANGES))
3409 result |= 01;
3410 if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
3411 DIFF_OPT_TST(opt, CHECK_FAILED))
3412 result |= 02;
3413 return result;
3416 void diff_addremove(struct diff_options *options,
3417 int addremove, unsigned mode,
3418 const unsigned char *sha1,
3419 const char *base, const char *path)
3421 char concatpath[PATH_MAX];
3422 struct diff_filespec *one, *two;
3424 if (DIFF_OPT_TST(options, IGNORE_SUBMODULES) && S_ISGITLINK(mode))
3425 return;
3427 /* This may look odd, but it is a preparation for
3428 * feeding "there are unchanged files which should
3429 * not produce diffs, but when you are doing copy
3430 * detection you would need them, so here they are"
3431 * entries to the diff-core. They will be prefixed
3432 * with something like '=' or '*' (I haven't decided
3433 * which but should not make any difference).
3434 * Feeding the same new and old to diff_change()
3435 * also has the same effect.
3436 * Before the final output happens, they are pruned after
3437 * merged into rename/copy pairs as appropriate.
3439 if (DIFF_OPT_TST(options, REVERSE_DIFF))
3440 addremove = (addremove == '+' ? '-' :
3441 addremove == '-' ? '+' : addremove);
3443 if (!path) path = "";
3444 sprintf(concatpath, "%s%s", base, path);
3446 if (options->prefix &&
3447 strncmp(concatpath, options->prefix, options->prefix_length))
3448 return;
3450 one = alloc_filespec(concatpath);
3451 two = alloc_filespec(concatpath);
3453 if (addremove != '+')
3454 fill_filespec(one, sha1, mode);
3455 if (addremove != '-')
3456 fill_filespec(two, sha1, mode);
3458 diff_queue(&diff_queued_diff, one, two);
3459 DIFF_OPT_SET(options, HAS_CHANGES);
3462 void diff_change(struct diff_options *options,
3463 unsigned old_mode, unsigned new_mode,
3464 const unsigned char *old_sha1,
3465 const unsigned char *new_sha1,
3466 const char *base, const char *path)
3468 char concatpath[PATH_MAX];
3469 struct diff_filespec *one, *two;
3471 if (DIFF_OPT_TST(options, IGNORE_SUBMODULES) && S_ISGITLINK(old_mode)
3472 && S_ISGITLINK(new_mode))
3473 return;
3475 if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
3476 unsigned tmp;
3477 const unsigned char *tmp_c;
3478 tmp = old_mode; old_mode = new_mode; new_mode = tmp;
3479 tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
3481 if (!path) path = "";
3482 sprintf(concatpath, "%s%s", base, path);
3484 if (options->prefix &&
3485 strncmp(concatpath, options->prefix, options->prefix_length))
3486 return;
3488 one = alloc_filespec(concatpath);
3489 two = alloc_filespec(concatpath);
3490 fill_filespec(one, old_sha1, old_mode);
3491 fill_filespec(two, new_sha1, new_mode);
3493 diff_queue(&diff_queued_diff, one, two);
3494 DIFF_OPT_SET(options, HAS_CHANGES);
3497 void diff_unmerge(struct diff_options *options,
3498 const char *path,
3499 unsigned mode, const unsigned char *sha1)
3501 struct diff_filespec *one, *two;
3503 if (options->prefix &&
3504 strncmp(path, options->prefix, options->prefix_length))
3505 return;
3507 one = alloc_filespec(path);
3508 two = alloc_filespec(path);
3509 fill_filespec(one, sha1, mode);
3510 diff_queue(&diff_queued_diff, one, two)->is_unmerged = 1;