enable whitespace checking of test scripts
[4msysgit-hv.git] / diff.c
blobf77f9e944776e8b0bc684b7618227cf437cf6053
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 fputs(set, file);
518 fwrite(line, len, 1, file);
519 fputs(reset, file);
522 static void emit_add_line(const char *reset, struct emit_callback *ecbdata, const char *line, int len)
524 const char *ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
525 const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_NEW);
527 if (!*ws)
528 emit_line(ecbdata->file, set, reset, line, len);
529 else {
530 /* Emit just the prefix, then the rest. */
531 emit_line(ecbdata->file, set, reset, line, ecbdata->nparents);
532 (void)check_and_emit_line(line + ecbdata->nparents,
533 len - ecbdata->nparents, ecbdata->ws_rule,
534 ecbdata->file, set, reset, ws);
538 static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
540 const char *cp;
541 unsigned long allot;
542 size_t l = len;
544 if (ecb->truncate)
545 return ecb->truncate(line, len);
546 cp = line;
547 allot = l;
548 while (0 < l) {
549 (void) utf8_width(&cp, &l);
550 if (!cp)
551 break; /* truncated in the middle? */
553 return allot - l;
556 static void fn_out_consume(void *priv, char *line, unsigned long len)
558 int i;
559 int color;
560 struct emit_callback *ecbdata = priv;
561 const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
562 const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
563 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
565 *(ecbdata->found_changesp) = 1;
567 if (ecbdata->label_path[0]) {
568 const char *name_a_tab, *name_b_tab;
570 name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
571 name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
573 fprintf(ecbdata->file, "%s--- %s%s%s\n",
574 meta, ecbdata->label_path[0], reset, name_a_tab);
575 fprintf(ecbdata->file, "%s+++ %s%s%s\n",
576 meta, ecbdata->label_path[1], reset, name_b_tab);
577 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
580 /* This is not really necessary for now because
581 * this codepath only deals with two-way diffs.
583 for (i = 0; i < len && line[i] == '@'; i++)
585 if (2 <= i && i < len && line[i] == ' ') {
586 ecbdata->nparents = i - 1;
587 len = sane_truncate_line(ecbdata, line, len);
588 emit_line(ecbdata->file,
589 diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO),
590 reset, line, len);
591 if (line[len-1] != '\n')
592 putc('\n', ecbdata->file);
593 return;
596 if (len < ecbdata->nparents) {
597 emit_line(ecbdata->file, reset, reset, line, len);
598 return;
601 color = DIFF_PLAIN;
602 if (ecbdata->diff_words && ecbdata->nparents != 1)
603 /* fall back to normal diff */
604 free_diff_words_data(ecbdata);
605 if (ecbdata->diff_words) {
606 if (line[0] == '-') {
607 diff_words_append(line, len,
608 &ecbdata->diff_words->minus);
609 return;
610 } else if (line[0] == '+') {
611 diff_words_append(line, len,
612 &ecbdata->diff_words->plus);
613 return;
615 if (ecbdata->diff_words->minus.text.size ||
616 ecbdata->diff_words->plus.text.size)
617 diff_words_show(ecbdata->diff_words);
618 line++;
619 len--;
620 emit_line(ecbdata->file, plain, reset, line, len);
621 return;
623 for (i = 0; i < ecbdata->nparents && len; i++) {
624 if (line[i] == '-')
625 color = DIFF_FILE_OLD;
626 else if (line[i] == '+')
627 color = DIFF_FILE_NEW;
630 if (color != DIFF_FILE_NEW) {
631 emit_line(ecbdata->file,
632 diff_get_color(ecbdata->color_diff, color),
633 reset, line, len);
634 return;
636 emit_add_line(reset, ecbdata, line, len);
639 static char *pprint_rename(const char *a, const char *b)
641 const char *old = a;
642 const char *new = b;
643 struct strbuf name;
644 int pfx_length, sfx_length;
645 int len_a = strlen(a);
646 int len_b = strlen(b);
647 int a_midlen, b_midlen;
648 int qlen_a = quote_c_style(a, NULL, NULL, 0);
649 int qlen_b = quote_c_style(b, NULL, NULL, 0);
651 strbuf_init(&name, 0);
652 if (qlen_a || qlen_b) {
653 quote_c_style(a, &name, NULL, 0);
654 strbuf_addstr(&name, " => ");
655 quote_c_style(b, &name, NULL, 0);
656 return strbuf_detach(&name, NULL);
659 /* Find common prefix */
660 pfx_length = 0;
661 while (*old && *new && *old == *new) {
662 if (*old == '/')
663 pfx_length = old - a + 1;
664 old++;
665 new++;
668 /* Find common suffix */
669 old = a + len_a;
670 new = b + len_b;
671 sfx_length = 0;
672 while (a <= old && b <= new && *old == *new) {
673 if (*old == '/')
674 sfx_length = len_a - (old - a);
675 old--;
676 new--;
680 * pfx{mid-a => mid-b}sfx
681 * {pfx-a => pfx-b}sfx
682 * pfx{sfx-a => sfx-b}
683 * name-a => name-b
685 a_midlen = len_a - pfx_length - sfx_length;
686 b_midlen = len_b - pfx_length - sfx_length;
687 if (a_midlen < 0)
688 a_midlen = 0;
689 if (b_midlen < 0)
690 b_midlen = 0;
692 strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
693 if (pfx_length + sfx_length) {
694 strbuf_add(&name, a, pfx_length);
695 strbuf_addch(&name, '{');
697 strbuf_add(&name, a + pfx_length, a_midlen);
698 strbuf_addstr(&name, " => ");
699 strbuf_add(&name, b + pfx_length, b_midlen);
700 if (pfx_length + sfx_length) {
701 strbuf_addch(&name, '}');
702 strbuf_add(&name, a + len_a - sfx_length, sfx_length);
704 return strbuf_detach(&name, NULL);
707 struct diffstat_t {
708 struct xdiff_emit_state xm;
710 int nr;
711 int alloc;
712 struct diffstat_file {
713 char *from_name;
714 char *name;
715 char *print_name;
716 unsigned is_unmerged:1;
717 unsigned is_binary:1;
718 unsigned is_renamed:1;
719 unsigned int added, deleted;
720 } **files;
723 static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
724 const char *name_a,
725 const char *name_b)
727 struct diffstat_file *x;
728 x = xcalloc(sizeof (*x), 1);
729 if (diffstat->nr == diffstat->alloc) {
730 diffstat->alloc = alloc_nr(diffstat->alloc);
731 diffstat->files = xrealloc(diffstat->files,
732 diffstat->alloc * sizeof(x));
734 diffstat->files[diffstat->nr++] = x;
735 if (name_b) {
736 x->from_name = xstrdup(name_a);
737 x->name = xstrdup(name_b);
738 x->is_renamed = 1;
740 else {
741 x->from_name = NULL;
742 x->name = xstrdup(name_a);
744 return x;
747 static void diffstat_consume(void *priv, char *line, unsigned long len)
749 struct diffstat_t *diffstat = priv;
750 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
752 if (line[0] == '+')
753 x->added++;
754 else if (line[0] == '-')
755 x->deleted++;
758 const char mime_boundary_leader[] = "------------";
760 static int scale_linear(int it, int width, int max_change)
763 * make sure that at least one '-' is printed if there were deletions,
764 * and likewise for '+'.
766 if (max_change < 2)
767 return it;
768 return ((it - 1) * (width - 1) + max_change - 1) / (max_change - 1);
771 static void show_name(FILE *file,
772 const char *prefix, const char *name, int len,
773 const char *reset, const char *set)
775 fprintf(file, " %s%s%-*s%s |", set, prefix, len, name, reset);
778 static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
780 if (cnt <= 0)
781 return;
782 fprintf(file, "%s", set);
783 while (cnt--)
784 putc(ch, file);
785 fprintf(file, "%s", reset);
788 static void fill_print_name(struct diffstat_file *file)
790 char *pname;
792 if (file->print_name)
793 return;
795 if (!file->is_renamed) {
796 struct strbuf buf;
797 strbuf_init(&buf, 0);
798 if (quote_c_style(file->name, &buf, NULL, 0)) {
799 pname = strbuf_detach(&buf, NULL);
800 } else {
801 pname = file->name;
802 strbuf_release(&buf);
804 } else {
805 pname = pprint_rename(file->from_name, file->name);
807 file->print_name = pname;
810 static void show_stats(struct diffstat_t* data, struct diff_options *options)
812 int i, len, add, del, total, adds = 0, dels = 0;
813 int max_change = 0, max_len = 0;
814 int total_files = data->nr;
815 int width, name_width;
816 const char *reset, *set, *add_c, *del_c;
818 if (data->nr == 0)
819 return;
821 width = options->stat_width ? options->stat_width : 80;
822 name_width = options->stat_name_width ? options->stat_name_width : 50;
824 /* Sanity: give at least 5 columns to the graph,
825 * but leave at least 10 columns for the name.
827 if (width < name_width + 15) {
828 if (name_width <= 25)
829 width = name_width + 15;
830 else
831 name_width = width - 15;
834 /* Find the longest filename and max number of changes */
835 reset = diff_get_color_opt(options, DIFF_RESET);
836 set = diff_get_color_opt(options, DIFF_PLAIN);
837 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
838 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
840 for (i = 0; i < data->nr; i++) {
841 struct diffstat_file *file = data->files[i];
842 int change = file->added + file->deleted;
843 fill_print_name(file);
844 len = strlen(file->print_name);
845 if (max_len < len)
846 max_len = len;
848 if (file->is_binary || file->is_unmerged)
849 continue;
850 if (max_change < change)
851 max_change = change;
854 /* Compute the width of the graph part;
855 * 10 is for one blank at the beginning of the line plus
856 * " | count " between the name and the graph.
858 * From here on, name_width is the width of the name area,
859 * and width is the width of the graph area.
861 name_width = (name_width < max_len) ? name_width : max_len;
862 if (width < (name_width + 10) + max_change)
863 width = width - (name_width + 10);
864 else
865 width = max_change;
867 for (i = 0; i < data->nr; i++) {
868 const char *prefix = "";
869 char *name = data->files[i]->print_name;
870 int added = data->files[i]->added;
871 int deleted = data->files[i]->deleted;
872 int name_len;
875 * "scale" the filename
877 len = name_width;
878 name_len = strlen(name);
879 if (name_width < name_len) {
880 char *slash;
881 prefix = "...";
882 len -= 3;
883 name += name_len - len;
884 slash = strchr(name, '/');
885 if (slash)
886 name = slash;
889 if (data->files[i]->is_binary) {
890 show_name(options->file, prefix, name, len, reset, set);
891 fprintf(options->file, " Bin ");
892 fprintf(options->file, "%s%d%s", del_c, deleted, reset);
893 fprintf(options->file, " -> ");
894 fprintf(options->file, "%s%d%s", add_c, added, reset);
895 fprintf(options->file, " bytes");
896 fprintf(options->file, "\n");
897 continue;
899 else if (data->files[i]->is_unmerged) {
900 show_name(options->file, prefix, name, len, reset, set);
901 fprintf(options->file, " Unmerged\n");
902 continue;
904 else if (!data->files[i]->is_renamed &&
905 (added + deleted == 0)) {
906 total_files--;
907 continue;
911 * scale the add/delete
913 add = added;
914 del = deleted;
915 total = add + del;
916 adds += add;
917 dels += del;
919 if (width <= max_change) {
920 add = scale_linear(add, width, max_change);
921 del = scale_linear(del, width, max_change);
922 total = add + del;
924 show_name(options->file, prefix, name, len, reset, set);
925 fprintf(options->file, "%5d%s", added + deleted,
926 added + deleted ? " " : "");
927 show_graph(options->file, '+', add, add_c, reset);
928 show_graph(options->file, '-', del, del_c, reset);
929 fprintf(options->file, "\n");
931 fprintf(options->file,
932 "%s %d files changed, %d insertions(+), %d deletions(-)%s\n",
933 set, total_files, adds, dels, reset);
936 static void show_shortstats(struct diffstat_t* data, struct diff_options *options)
938 int i, adds = 0, dels = 0, total_files = data->nr;
940 if (data->nr == 0)
941 return;
943 for (i = 0; i < data->nr; i++) {
944 if (!data->files[i]->is_binary &&
945 !data->files[i]->is_unmerged) {
946 int added = data->files[i]->added;
947 int deleted= data->files[i]->deleted;
948 if (!data->files[i]->is_renamed &&
949 (added + deleted == 0)) {
950 total_files--;
951 } else {
952 adds += added;
953 dels += deleted;
957 fprintf(options->file, " %d files changed, %d insertions(+), %d deletions(-)\n",
958 total_files, adds, dels);
961 static void show_numstat(struct diffstat_t* data, struct diff_options *options)
963 int i;
965 if (data->nr == 0)
966 return;
968 for (i = 0; i < data->nr; i++) {
969 struct diffstat_file *file = data->files[i];
971 if (file->is_binary)
972 fprintf(options->file, "-\t-\t");
973 else
974 fprintf(options->file,
975 "%d\t%d\t", file->added, file->deleted);
976 if (options->line_termination) {
977 fill_print_name(file);
978 if (!file->is_renamed)
979 write_name_quoted(file->name, options->file,
980 options->line_termination);
981 else {
982 fputs(file->print_name, options->file);
983 putc(options->line_termination, options->file);
985 } else {
986 if (file->is_renamed) {
987 putc('\0', options->file);
988 write_name_quoted(file->from_name, options->file, '\0');
990 write_name_quoted(file->name, options->file, '\0');
995 struct dirstat_file {
996 const char *name;
997 unsigned long changed;
1000 struct dirstat_dir {
1001 struct dirstat_file *files;
1002 int alloc, nr, percent, cumulative;
1005 static long gather_dirstat(FILE *file, struct dirstat_dir *dir, unsigned long changed, const char *base, int baselen)
1007 unsigned long this_dir = 0;
1008 unsigned int sources = 0;
1010 while (dir->nr) {
1011 struct dirstat_file *f = dir->files;
1012 int namelen = strlen(f->name);
1013 unsigned long this;
1014 char *slash;
1016 if (namelen < baselen)
1017 break;
1018 if (memcmp(f->name, base, baselen))
1019 break;
1020 slash = strchr(f->name + baselen, '/');
1021 if (slash) {
1022 int newbaselen = slash + 1 - f->name;
1023 this = gather_dirstat(file, dir, changed, f->name, newbaselen);
1024 sources++;
1025 } else {
1026 this = f->changed;
1027 dir->files++;
1028 dir->nr--;
1029 sources += 2;
1031 this_dir += this;
1035 * We don't report dirstat's for
1036 * - the top level
1037 * - or cases where everything came from a single directory
1038 * under this directory (sources == 1).
1040 if (baselen && sources != 1) {
1041 int permille = this_dir * 1000 / changed;
1042 if (permille) {
1043 int percent = permille / 10;
1044 if (percent >= dir->percent) {
1045 fprintf(file, "%4d.%01d%% %.*s\n", percent, permille % 10, baselen, base);
1046 if (!dir->cumulative)
1047 return 0;
1051 return this_dir;
1054 static void show_dirstat(struct diff_options *options)
1056 int i;
1057 unsigned long changed;
1058 struct dirstat_dir dir;
1059 struct diff_queue_struct *q = &diff_queued_diff;
1061 dir.files = NULL;
1062 dir.alloc = 0;
1063 dir.nr = 0;
1064 dir.percent = options->dirstat_percent;
1065 dir.cumulative = options->output_format & DIFF_FORMAT_CUMULATIVE;
1067 changed = 0;
1068 for (i = 0; i < q->nr; i++) {
1069 struct diff_filepair *p = q->queue[i];
1070 const char *name;
1071 unsigned long copied, added, damage;
1073 name = p->one->path ? p->one->path : p->two->path;
1075 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
1076 diff_populate_filespec(p->one, 0);
1077 diff_populate_filespec(p->two, 0);
1078 diffcore_count_changes(p->one, p->two, NULL, NULL, 0,
1079 &copied, &added);
1080 diff_free_filespec_data(p->one);
1081 diff_free_filespec_data(p->two);
1082 } else if (DIFF_FILE_VALID(p->one)) {
1083 diff_populate_filespec(p->one, 1);
1084 copied = added = 0;
1085 diff_free_filespec_data(p->one);
1086 } else if (DIFF_FILE_VALID(p->two)) {
1087 diff_populate_filespec(p->two, 1);
1088 copied = 0;
1089 added = p->two->size;
1090 diff_free_filespec_data(p->two);
1091 } else
1092 continue;
1095 * Original minus copied is the removed material,
1096 * added is the new material. They are both damages
1097 * made to the preimage.
1099 damage = (p->one->size - copied) + added;
1101 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1102 dir.files[dir.nr].name = name;
1103 dir.files[dir.nr].changed = damage;
1104 changed += damage;
1105 dir.nr++;
1108 /* This can happen even with many files, if everything was renames */
1109 if (!changed)
1110 return;
1112 /* Show all directories with more than x% of the changes */
1113 gather_dirstat(options->file, &dir, changed, "", 0);
1116 static void free_diffstat_info(struct diffstat_t *diffstat)
1118 int i;
1119 for (i = 0; i < diffstat->nr; i++) {
1120 struct diffstat_file *f = diffstat->files[i];
1121 if (f->name != f->print_name)
1122 free(f->print_name);
1123 free(f->name);
1124 free(f->from_name);
1125 free(f);
1127 free(diffstat->files);
1130 struct checkdiff_t {
1131 struct xdiff_emit_state xm;
1132 const char *filename;
1133 int lineno, color_diff;
1134 unsigned ws_rule;
1135 unsigned status;
1136 FILE *file;
1139 static void checkdiff_consume(void *priv, char *line, unsigned long len)
1141 struct checkdiff_t *data = priv;
1142 const char *ws = diff_get_color(data->color_diff, DIFF_WHITESPACE);
1143 const char *reset = diff_get_color(data->color_diff, DIFF_RESET);
1144 const char *set = diff_get_color(data->color_diff, DIFF_FILE_NEW);
1145 char *err;
1147 if (line[0] == '+') {
1148 data->lineno++;
1149 data->status = check_and_emit_line(line + 1, len - 1,
1150 data->ws_rule, NULL, NULL, NULL, NULL);
1151 if (!data->status)
1152 return;
1153 err = whitespace_error_string(data->status);
1154 fprintf(data->file, "%s:%d: %s.\n", data->filename, data->lineno, err);
1155 free(err);
1156 emit_line(data->file, set, reset, line, 1);
1157 (void)check_and_emit_line(line + 1, len - 1, data->ws_rule,
1158 data->file, set, reset, ws);
1159 } else if (line[0] == ' ')
1160 data->lineno++;
1161 else if (line[0] == '@') {
1162 char *plus = strchr(line, '+');
1163 if (plus)
1164 data->lineno = strtol(plus, NULL, 10) - 1;
1165 else
1166 die("invalid diff");
1170 static unsigned char *deflate_it(char *data,
1171 unsigned long size,
1172 unsigned long *result_size)
1174 int bound;
1175 unsigned char *deflated;
1176 z_stream stream;
1178 memset(&stream, 0, sizeof(stream));
1179 deflateInit(&stream, zlib_compression_level);
1180 bound = deflateBound(&stream, size);
1181 deflated = xmalloc(bound);
1182 stream.next_out = deflated;
1183 stream.avail_out = bound;
1185 stream.next_in = (unsigned char *)data;
1186 stream.avail_in = size;
1187 while (deflate(&stream, Z_FINISH) == Z_OK)
1188 ; /* nothing */
1189 deflateEnd(&stream);
1190 *result_size = stream.total_out;
1191 return deflated;
1194 static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two)
1196 void *cp;
1197 void *delta;
1198 void *deflated;
1199 void *data;
1200 unsigned long orig_size;
1201 unsigned long delta_size;
1202 unsigned long deflate_size;
1203 unsigned long data_size;
1205 /* We could do deflated delta, or we could do just deflated two,
1206 * whichever is smaller.
1208 delta = NULL;
1209 deflated = deflate_it(two->ptr, two->size, &deflate_size);
1210 if (one->size && two->size) {
1211 delta = diff_delta(one->ptr, one->size,
1212 two->ptr, two->size,
1213 &delta_size, deflate_size);
1214 if (delta) {
1215 void *to_free = delta;
1216 orig_size = delta_size;
1217 delta = deflate_it(delta, delta_size, &delta_size);
1218 free(to_free);
1222 if (delta && delta_size < deflate_size) {
1223 fprintf(file, "delta %lu\n", orig_size);
1224 free(deflated);
1225 data = delta;
1226 data_size = delta_size;
1228 else {
1229 fprintf(file, "literal %lu\n", two->size);
1230 free(delta);
1231 data = deflated;
1232 data_size = deflate_size;
1235 /* emit data encoded in base85 */
1236 cp = data;
1237 while (data_size) {
1238 int bytes = (52 < data_size) ? 52 : data_size;
1239 char line[70];
1240 data_size -= bytes;
1241 if (bytes <= 26)
1242 line[0] = bytes + 'A' - 1;
1243 else
1244 line[0] = bytes - 26 + 'a' - 1;
1245 encode_85(line + 1, cp, bytes);
1246 cp = (char *) cp + bytes;
1247 fputs(line, file);
1248 fputc('\n', file);
1250 fprintf(file, "\n");
1251 free(data);
1254 static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two)
1256 fprintf(file, "GIT binary patch\n");
1257 emit_binary_diff_body(file, one, two);
1258 emit_binary_diff_body(file, two, one);
1261 static void setup_diff_attr_check(struct git_attr_check *check)
1263 static struct git_attr *attr_diff;
1265 if (!attr_diff) {
1266 attr_diff = git_attr("diff", 4);
1268 check[0].attr = attr_diff;
1271 static void diff_filespec_check_attr(struct diff_filespec *one)
1273 struct git_attr_check attr_diff_check;
1274 int check_from_data = 0;
1276 if (one->checked_attr)
1277 return;
1279 setup_diff_attr_check(&attr_diff_check);
1280 one->is_binary = 0;
1281 one->funcname_pattern_ident = NULL;
1283 if (!git_checkattr(one->path, 1, &attr_diff_check)) {
1284 const char *value;
1286 /* binaryness */
1287 value = attr_diff_check.value;
1288 if (ATTR_TRUE(value))
1290 else if (ATTR_FALSE(value))
1291 one->is_binary = 1;
1292 else
1293 check_from_data = 1;
1295 /* funcname pattern ident */
1296 if (ATTR_TRUE(value) || ATTR_FALSE(value) || ATTR_UNSET(value))
1298 else
1299 one->funcname_pattern_ident = value;
1302 if (check_from_data) {
1303 if (!one->data && DIFF_FILE_VALID(one))
1304 diff_populate_filespec(one, 0);
1306 if (one->data)
1307 one->is_binary = buffer_is_binary(one->data, one->size);
1311 int diff_filespec_is_binary(struct diff_filespec *one)
1313 diff_filespec_check_attr(one);
1314 return one->is_binary;
1317 static const char *funcname_pattern(const char *ident)
1319 struct funcname_pattern *pp;
1321 for (pp = funcname_pattern_list; pp; pp = pp->next)
1322 if (!strcmp(ident, pp->name))
1323 return pp->pattern;
1324 return NULL;
1327 static struct builtin_funcname_pattern {
1328 const char *name;
1329 const char *pattern;
1330 } builtin_funcname_pattern[] = {
1331 { "java", "!^[ ]*\\(catch\\|do\\|for\\|if\\|instanceof\\|"
1332 "new\\|return\\|switch\\|throw\\|while\\)\n"
1333 "^[ ]*\\(\\([ ]*"
1334 "[A-Za-z_][A-Za-z_0-9]*\\)\\{2,\\}"
1335 "[ ]*([^;]*\\)$" },
1336 { "tex", "^\\(\\\\\\(sub\\)*section{.*\\)$" },
1339 static const char *diff_funcname_pattern(struct diff_filespec *one)
1341 const char *ident, *pattern;
1342 int i;
1344 diff_filespec_check_attr(one);
1345 ident = one->funcname_pattern_ident;
1347 if (!ident)
1349 * If the config file has "funcname.default" defined, that
1350 * regexp is used; otherwise NULL is returned and xemit uses
1351 * the built-in default.
1353 return funcname_pattern("default");
1355 /* Look up custom "funcname.$ident" regexp from config. */
1356 pattern = funcname_pattern(ident);
1357 if (pattern)
1358 return pattern;
1361 * And define built-in fallback patterns here. Note that
1362 * these can be overridden by the user's config settings.
1364 for (i = 0; i < ARRAY_SIZE(builtin_funcname_pattern); i++)
1365 if (!strcmp(ident, builtin_funcname_pattern[i].name))
1366 return builtin_funcname_pattern[i].pattern;
1368 return NULL;
1371 static void builtin_diff(const char *name_a,
1372 const char *name_b,
1373 struct diff_filespec *one,
1374 struct diff_filespec *two,
1375 const char *xfrm_msg,
1376 struct diff_options *o,
1377 int complete_rewrite)
1379 mmfile_t mf1, mf2;
1380 const char *lbl[2];
1381 char *a_one, *b_two;
1382 const char *set = diff_get_color_opt(o, DIFF_METAINFO);
1383 const char *reset = diff_get_color_opt(o, DIFF_RESET);
1385 a_one = quote_two(o->a_prefix, name_a + (*name_a == '/'));
1386 b_two = quote_two(o->b_prefix, name_b + (*name_b == '/'));
1387 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
1388 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
1389 fprintf(o->file, "%sdiff --git %s %s%s\n", set, a_one, b_two, reset);
1390 if (lbl[0][0] == '/') {
1391 /* /dev/null */
1392 fprintf(o->file, "%snew file mode %06o%s\n", set, two->mode, reset);
1393 if (xfrm_msg && xfrm_msg[0])
1394 fprintf(o->file, "%s%s%s\n", set, xfrm_msg, reset);
1396 else if (lbl[1][0] == '/') {
1397 fprintf(o->file, "%sdeleted file mode %06o%s\n", set, one->mode, reset);
1398 if (xfrm_msg && xfrm_msg[0])
1399 fprintf(o->file, "%s%s%s\n", set, xfrm_msg, reset);
1401 else {
1402 if (one->mode != two->mode) {
1403 fprintf(o->file, "%sold mode %06o%s\n", set, one->mode, reset);
1404 fprintf(o->file, "%snew mode %06o%s\n", set, two->mode, reset);
1406 if (xfrm_msg && xfrm_msg[0])
1407 fprintf(o->file, "%s%s%s\n", set, xfrm_msg, reset);
1409 * we do not run diff between different kind
1410 * of objects.
1412 if ((one->mode ^ two->mode) & S_IFMT)
1413 goto free_ab_and_return;
1414 if (complete_rewrite) {
1415 emit_rewrite_diff(name_a, name_b, one, two, o);
1416 o->found_changes = 1;
1417 goto free_ab_and_return;
1421 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1422 die("unable to read files to diff");
1424 if (!DIFF_OPT_TST(o, TEXT) &&
1425 (diff_filespec_is_binary(one) || diff_filespec_is_binary(two))) {
1426 /* Quite common confusing case */
1427 if (mf1.size == mf2.size &&
1428 !memcmp(mf1.ptr, mf2.ptr, mf1.size))
1429 goto free_ab_and_return;
1430 if (DIFF_OPT_TST(o, BINARY))
1431 emit_binary_diff(o->file, &mf1, &mf2);
1432 else
1433 fprintf(o->file, "Binary files %s and %s differ\n",
1434 lbl[0], lbl[1]);
1435 o->found_changes = 1;
1437 else {
1438 /* Crazy xdl interfaces.. */
1439 const char *diffopts = getenv("GIT_DIFF_OPTS");
1440 xpparam_t xpp;
1441 xdemitconf_t xecfg;
1442 xdemitcb_t ecb;
1443 struct emit_callback ecbdata;
1444 const char *funcname_pattern;
1446 funcname_pattern = diff_funcname_pattern(one);
1447 if (!funcname_pattern)
1448 funcname_pattern = diff_funcname_pattern(two);
1450 memset(&xecfg, 0, sizeof(xecfg));
1451 memset(&ecbdata, 0, sizeof(ecbdata));
1452 ecbdata.label_path = lbl;
1453 ecbdata.color_diff = DIFF_OPT_TST(o, COLOR_DIFF);
1454 ecbdata.found_changesp = &o->found_changes;
1455 ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
1456 ecbdata.file = o->file;
1457 xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
1458 xecfg.ctxlen = o->context;
1459 xecfg.flags = XDL_EMIT_FUNCNAMES;
1460 if (funcname_pattern)
1461 xdiff_set_find_func(&xecfg, funcname_pattern);
1462 if (!diffopts)
1464 else if (!prefixcmp(diffopts, "--unified="))
1465 xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
1466 else if (!prefixcmp(diffopts, "-u"))
1467 xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
1468 ecb.outf = xdiff_outf;
1469 ecb.priv = &ecbdata;
1470 ecbdata.xm.consume = fn_out_consume;
1471 if (DIFF_OPT_TST(o, COLOR_DIFF_WORDS)) {
1472 ecbdata.diff_words =
1473 xcalloc(1, sizeof(struct diff_words_data));
1474 ecbdata.diff_words->file = o->file;
1476 xdi_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
1477 if (DIFF_OPT_TST(o, COLOR_DIFF_WORDS))
1478 free_diff_words_data(&ecbdata);
1481 free_ab_and_return:
1482 diff_free_filespec_data(one);
1483 diff_free_filespec_data(two);
1484 free(a_one);
1485 free(b_two);
1486 return;
1489 static void builtin_diffstat(const char *name_a, const char *name_b,
1490 struct diff_filespec *one,
1491 struct diff_filespec *two,
1492 struct diffstat_t *diffstat,
1493 struct diff_options *o,
1494 int complete_rewrite)
1496 mmfile_t mf1, mf2;
1497 struct diffstat_file *data;
1499 data = diffstat_add(diffstat, name_a, name_b);
1501 if (!one || !two) {
1502 data->is_unmerged = 1;
1503 return;
1505 if (complete_rewrite) {
1506 diff_populate_filespec(one, 0);
1507 diff_populate_filespec(two, 0);
1508 data->deleted = count_lines(one->data, one->size);
1509 data->added = count_lines(two->data, two->size);
1510 goto free_and_return;
1512 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1513 die("unable to read files to diff");
1515 if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
1516 data->is_binary = 1;
1517 data->added = mf2.size;
1518 data->deleted = mf1.size;
1519 } else {
1520 /* Crazy xdl interfaces.. */
1521 xpparam_t xpp;
1522 xdemitconf_t xecfg;
1523 xdemitcb_t ecb;
1525 memset(&xecfg, 0, sizeof(xecfg));
1526 xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
1527 ecb.outf = xdiff_outf;
1528 ecb.priv = diffstat;
1529 xdi_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
1532 free_and_return:
1533 diff_free_filespec_data(one);
1534 diff_free_filespec_data(two);
1537 static void builtin_checkdiff(const char *name_a, const char *name_b,
1538 const char *attr_path,
1539 struct diff_filespec *one,
1540 struct diff_filespec *two, struct diff_options *o)
1542 mmfile_t mf1, mf2;
1543 struct checkdiff_t data;
1545 if (!two)
1546 return;
1548 memset(&data, 0, sizeof(data));
1549 data.xm.consume = checkdiff_consume;
1550 data.filename = name_b ? name_b : name_a;
1551 data.lineno = 0;
1552 data.color_diff = DIFF_OPT_TST(o, COLOR_DIFF);
1553 data.ws_rule = whitespace_rule(attr_path);
1554 data.file = o->file;
1556 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1557 die("unable to read files to diff");
1559 if (diff_filespec_is_binary(two))
1560 goto free_and_return;
1561 else {
1562 /* Crazy xdl interfaces.. */
1563 xpparam_t xpp;
1564 xdemitconf_t xecfg;
1565 xdemitcb_t ecb;
1567 memset(&xecfg, 0, sizeof(xecfg));
1568 xpp.flags = XDF_NEED_MINIMAL;
1569 ecb.outf = xdiff_outf;
1570 ecb.priv = &data;
1571 xdi_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
1573 free_and_return:
1574 diff_free_filespec_data(one);
1575 diff_free_filespec_data(two);
1576 if (data.status)
1577 DIFF_OPT_SET(o, CHECK_FAILED);
1580 struct diff_filespec *alloc_filespec(const char *path)
1582 int namelen = strlen(path);
1583 struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
1585 memset(spec, 0, sizeof(*spec));
1586 spec->path = (char *)(spec + 1);
1587 memcpy(spec->path, path, namelen+1);
1588 spec->count = 1;
1589 return spec;
1592 void free_filespec(struct diff_filespec *spec)
1594 if (!--spec->count) {
1595 diff_free_filespec_data(spec);
1596 free(spec);
1600 void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
1601 unsigned short mode)
1603 if (mode) {
1604 spec->mode = canon_mode(mode);
1605 hashcpy(spec->sha1, sha1);
1606 spec->sha1_valid = !is_null_sha1(sha1);
1611 * Given a name and sha1 pair, if the index tells us the file in
1612 * the work tree has that object contents, return true, so that
1613 * prepare_temp_file() does not have to inflate and extract.
1615 static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
1617 struct cache_entry *ce;
1618 struct stat st;
1619 int pos, len;
1621 /* We do not read the cache ourselves here, because the
1622 * benchmark with my previous version that always reads cache
1623 * shows that it makes things worse for diff-tree comparing
1624 * two linux-2.6 kernel trees in an already checked out work
1625 * tree. This is because most diff-tree comparisons deal with
1626 * only a small number of files, while reading the cache is
1627 * expensive for a large project, and its cost outweighs the
1628 * savings we get by not inflating the object to a temporary
1629 * file. Practically, this code only helps when we are used
1630 * by diff-cache --cached, which does read the cache before
1631 * calling us.
1633 if (!active_cache)
1634 return 0;
1636 /* We want to avoid the working directory if our caller
1637 * doesn't need the data in a normal file, this system
1638 * is rather slow with its stat/open/mmap/close syscalls,
1639 * and the object is contained in a pack file. The pack
1640 * is probably already open and will be faster to obtain
1641 * the data through than the working directory. Loose
1642 * objects however would tend to be slower as they need
1643 * to be individually opened and inflated.
1645 if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1, NULL))
1646 return 0;
1648 len = strlen(name);
1649 pos = cache_name_pos(name, len);
1650 if (pos < 0)
1651 return 0;
1652 ce = active_cache[pos];
1655 * This is not the sha1 we are looking for, or
1656 * unreusable because it is not a regular file.
1658 if (hashcmp(sha1, ce->sha1) || !S_ISREG(ce->ce_mode))
1659 return 0;
1662 * If ce matches the file in the work tree, we can reuse it.
1664 if (ce_uptodate(ce) ||
1665 (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
1666 return 1;
1668 return 0;
1671 static int populate_from_stdin(struct diff_filespec *s)
1673 struct strbuf buf;
1674 size_t size = 0;
1676 strbuf_init(&buf, 0);
1677 if (strbuf_read(&buf, 0, 0) < 0)
1678 return error("error while reading from stdin %s",
1679 strerror(errno));
1681 s->should_munmap = 0;
1682 s->data = strbuf_detach(&buf, &size);
1683 s->size = size;
1684 s->should_free = 1;
1685 return 0;
1688 static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
1690 int len;
1691 char *data = xmalloc(100);
1692 len = snprintf(data, 100,
1693 "Subproject commit %s\n", sha1_to_hex(s->sha1));
1694 s->data = data;
1695 s->size = len;
1696 s->should_free = 1;
1697 if (size_only) {
1698 s->data = NULL;
1699 free(data);
1701 return 0;
1705 * While doing rename detection and pickaxe operation, we may need to
1706 * grab the data for the blob (or file) for our own in-core comparison.
1707 * diff_filespec has data and size fields for this purpose.
1709 int diff_populate_filespec(struct diff_filespec *s, int size_only)
1711 int err = 0;
1712 if (!DIFF_FILE_VALID(s))
1713 die("internal error: asking to populate invalid file.");
1714 if (S_ISDIR(s->mode))
1715 return -1;
1717 if (s->data)
1718 return 0;
1720 if (size_only && 0 < s->size)
1721 return 0;
1723 if (S_ISGITLINK(s->mode))
1724 return diff_populate_gitlink(s, size_only);
1726 if (!s->sha1_valid ||
1727 reuse_worktree_file(s->path, s->sha1, 0)) {
1728 struct strbuf buf;
1729 struct stat st;
1730 int fd;
1732 if (!strcmp(s->path, "-"))
1733 return populate_from_stdin(s);
1735 if (lstat(s->path, &st) < 0) {
1736 if (errno == ENOENT) {
1737 err_empty:
1738 err = -1;
1739 empty:
1740 s->data = (char *)"";
1741 s->size = 0;
1742 return err;
1745 s->size = xsize_t(st.st_size);
1746 if (!s->size)
1747 goto empty;
1748 if (size_only)
1749 return 0;
1750 if (S_ISLNK(st.st_mode)) {
1751 int ret;
1752 s->data = xmalloc(s->size);
1753 s->should_free = 1;
1754 ret = readlink(s->path, s->data, s->size);
1755 if (ret < 0) {
1756 free(s->data);
1757 goto err_empty;
1759 return 0;
1761 fd = open(s->path, O_RDONLY);
1762 if (fd < 0)
1763 goto err_empty;
1764 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
1765 close(fd);
1766 s->should_munmap = 1;
1769 * Convert from working tree format to canonical git format
1771 strbuf_init(&buf, 0);
1772 if (convert_to_git(s->path, s->data, s->size, &buf, safe_crlf)) {
1773 size_t size = 0;
1774 munmap(s->data, s->size);
1775 s->should_munmap = 0;
1776 s->data = strbuf_detach(&buf, &size);
1777 s->size = size;
1778 s->should_free = 1;
1781 else {
1782 enum object_type type;
1783 if (size_only)
1784 type = sha1_object_info(s->sha1, &s->size);
1785 else {
1786 s->data = read_sha1_file(s->sha1, &type, &s->size);
1787 s->should_free = 1;
1790 return 0;
1793 void diff_free_filespec_blob(struct diff_filespec *s)
1795 if (s->should_free)
1796 free(s->data);
1797 else if (s->should_munmap)
1798 munmap(s->data, s->size);
1800 if (s->should_free || s->should_munmap) {
1801 s->should_free = s->should_munmap = 0;
1802 s->data = NULL;
1806 void diff_free_filespec_data(struct diff_filespec *s)
1808 diff_free_filespec_blob(s);
1809 free(s->cnt_data);
1810 s->cnt_data = NULL;
1813 static void prep_temp_blob(struct diff_tempfile *temp,
1814 void *blob,
1815 unsigned long size,
1816 const unsigned char *sha1,
1817 int mode)
1819 int fd;
1821 fd = git_mkstemp(temp->tmp_path, PATH_MAX, ".diff_XXXXXX");
1822 if (fd < 0)
1823 die("unable to create temp-file: %s", strerror(errno));
1824 if (write_in_full(fd, blob, size) != size)
1825 die("unable to write temp-file");
1826 close(fd);
1827 temp->name = temp->tmp_path;
1828 strcpy(temp->hex, sha1_to_hex(sha1));
1829 temp->hex[40] = 0;
1830 sprintf(temp->mode, "%06o", mode);
1833 static void prepare_temp_file(const char *name,
1834 struct diff_tempfile *temp,
1835 struct diff_filespec *one)
1837 if (!DIFF_FILE_VALID(one)) {
1838 not_a_valid_file:
1839 /* A '-' entry produces this for file-2, and
1840 * a '+' entry produces this for file-1.
1842 temp->name = "/dev/null";
1843 strcpy(temp->hex, ".");
1844 strcpy(temp->mode, ".");
1845 return;
1848 if (!one->sha1_valid ||
1849 reuse_worktree_file(name, one->sha1, 1)) {
1850 struct stat st;
1851 if (lstat(name, &st) < 0) {
1852 if (errno == ENOENT)
1853 goto not_a_valid_file;
1854 die("stat(%s): %s", name, strerror(errno));
1856 if (S_ISLNK(st.st_mode)) {
1857 int ret;
1858 char buf[PATH_MAX + 1]; /* ought to be SYMLINK_MAX */
1859 size_t sz = xsize_t(st.st_size);
1860 if (sizeof(buf) <= st.st_size)
1861 die("symlink too long: %s", name);
1862 ret = readlink(name, buf, sz);
1863 if (ret < 0)
1864 die("readlink(%s)", name);
1865 prep_temp_blob(temp, buf, sz,
1866 (one->sha1_valid ?
1867 one->sha1 : null_sha1),
1868 (one->sha1_valid ?
1869 one->mode : S_IFLNK));
1871 else {
1872 /* we can borrow from the file in the work tree */
1873 temp->name = name;
1874 if (!one->sha1_valid)
1875 strcpy(temp->hex, sha1_to_hex(null_sha1));
1876 else
1877 strcpy(temp->hex, sha1_to_hex(one->sha1));
1878 /* Even though we may sometimes borrow the
1879 * contents from the work tree, we always want
1880 * one->mode. mode is trustworthy even when
1881 * !(one->sha1_valid), as long as
1882 * DIFF_FILE_VALID(one).
1884 sprintf(temp->mode, "%06o", one->mode);
1886 return;
1888 else {
1889 if (diff_populate_filespec(one, 0))
1890 die("cannot read data blob for %s", one->path);
1891 prep_temp_blob(temp, one->data, one->size,
1892 one->sha1, one->mode);
1896 static void remove_tempfile(void)
1898 int i;
1900 for (i = 0; i < 2; i++)
1901 if (diff_temp[i].name == diff_temp[i].tmp_path) {
1902 unlink(diff_temp[i].name);
1903 diff_temp[i].name = NULL;
1907 static void remove_tempfile_on_signal(int signo)
1909 remove_tempfile();
1910 signal(SIGINT, SIG_DFL);
1911 raise(signo);
1914 /* An external diff command takes:
1916 * diff-cmd name infile1 infile1-sha1 infile1-mode \
1917 * infile2 infile2-sha1 infile2-mode [ rename-to ]
1920 static void run_external_diff(const char *pgm,
1921 const char *name,
1922 const char *other,
1923 struct diff_filespec *one,
1924 struct diff_filespec *two,
1925 const char *xfrm_msg,
1926 int complete_rewrite)
1928 const char *spawn_arg[10];
1929 struct diff_tempfile *temp = diff_temp;
1930 int retval;
1931 static int atexit_asked = 0;
1932 const char *othername;
1933 const char **arg = &spawn_arg[0];
1935 othername = (other? other : name);
1936 if (one && two) {
1937 prepare_temp_file(name, &temp[0], one);
1938 prepare_temp_file(othername, &temp[1], two);
1939 if (! atexit_asked &&
1940 (temp[0].name == temp[0].tmp_path ||
1941 temp[1].name == temp[1].tmp_path)) {
1942 atexit_asked = 1;
1943 atexit(remove_tempfile);
1945 signal(SIGINT, remove_tempfile_on_signal);
1948 if (one && two) {
1949 *arg++ = pgm;
1950 *arg++ = name;
1951 *arg++ = temp[0].name;
1952 *arg++ = temp[0].hex;
1953 *arg++ = temp[0].mode;
1954 *arg++ = temp[1].name;
1955 *arg++ = temp[1].hex;
1956 *arg++ = temp[1].mode;
1957 if (other) {
1958 *arg++ = other;
1959 *arg++ = xfrm_msg;
1961 } else {
1962 *arg++ = pgm;
1963 *arg++ = name;
1965 *arg = NULL;
1966 fflush(NULL);
1967 retval = run_command_v_opt(spawn_arg, 0);
1968 remove_tempfile();
1969 if (retval) {
1970 fprintf(stderr, "external diff died, stopping at %s.\n", name);
1971 exit(1);
1975 static const char *external_diff_attr(const char *name)
1977 struct git_attr_check attr_diff_check;
1979 if (!name)
1980 return NULL;
1982 setup_diff_attr_check(&attr_diff_check);
1983 if (!git_checkattr(name, 1, &attr_diff_check)) {
1984 const char *value = attr_diff_check.value;
1985 if (!ATTR_TRUE(value) &&
1986 !ATTR_FALSE(value) &&
1987 !ATTR_UNSET(value)) {
1988 struct ll_diff_driver *drv;
1990 for (drv = user_diff; drv; drv = drv->next)
1991 if (!strcmp(drv->name, value))
1992 return drv->cmd;
1995 return NULL;
1998 static void run_diff_cmd(const char *pgm,
1999 const char *name,
2000 const char *other,
2001 const char *attr_path,
2002 struct diff_filespec *one,
2003 struct diff_filespec *two,
2004 const char *xfrm_msg,
2005 struct diff_options *o,
2006 int complete_rewrite)
2008 if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
2009 pgm = NULL;
2010 else {
2011 const char *cmd = external_diff_attr(attr_path);
2012 if (cmd)
2013 pgm = cmd;
2016 if (pgm) {
2017 run_external_diff(pgm, name, other, one, two, xfrm_msg,
2018 complete_rewrite);
2019 return;
2021 if (one && two)
2022 builtin_diff(name, other ? other : name,
2023 one, two, xfrm_msg, o, complete_rewrite);
2024 else
2025 fprintf(o->file, "* Unmerged path %s\n", name);
2028 static void diff_fill_sha1_info(struct diff_filespec *one)
2030 if (DIFF_FILE_VALID(one)) {
2031 if (!one->sha1_valid) {
2032 struct stat st;
2033 if (!strcmp(one->path, "-")) {
2034 hashcpy(one->sha1, null_sha1);
2035 return;
2037 if (lstat(one->path, &st) < 0)
2038 die("stat %s", one->path);
2039 if (index_path(one->sha1, one->path, &st, 0))
2040 die("cannot hash %s\n", one->path);
2043 else
2044 hashclr(one->sha1);
2047 static int similarity_index(struct diff_filepair *p)
2049 return p->score * 100 / MAX_SCORE;
2052 static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
2054 /* Strip the prefix but do not molest /dev/null and absolute paths */
2055 if (*namep && **namep != '/')
2056 *namep += prefix_length;
2057 if (*otherp && **otherp != '/')
2058 *otherp += prefix_length;
2061 static void run_diff(struct diff_filepair *p, struct diff_options *o)
2063 const char *pgm = external_diff();
2064 struct strbuf msg;
2065 char *xfrm_msg;
2066 struct diff_filespec *one = p->one;
2067 struct diff_filespec *two = p->two;
2068 const char *name;
2069 const char *other;
2070 const char *attr_path;
2071 int complete_rewrite = 0;
2073 name = p->one->path;
2074 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2075 attr_path = name;
2076 if (o->prefix_length)
2077 strip_prefix(o->prefix_length, &name, &other);
2079 if (DIFF_PAIR_UNMERGED(p)) {
2080 run_diff_cmd(pgm, name, NULL, attr_path,
2081 NULL, NULL, NULL, o, 0);
2082 return;
2085 diff_fill_sha1_info(one);
2086 diff_fill_sha1_info(two);
2088 strbuf_init(&msg, PATH_MAX * 2 + 300);
2089 switch (p->status) {
2090 case DIFF_STATUS_COPIED:
2091 strbuf_addf(&msg, "similarity index %d%%", similarity_index(p));
2092 strbuf_addstr(&msg, "\ncopy from ");
2093 quote_c_style(name, &msg, NULL, 0);
2094 strbuf_addstr(&msg, "\ncopy to ");
2095 quote_c_style(other, &msg, NULL, 0);
2096 strbuf_addch(&msg, '\n');
2097 break;
2098 case DIFF_STATUS_RENAMED:
2099 strbuf_addf(&msg, "similarity index %d%%", similarity_index(p));
2100 strbuf_addstr(&msg, "\nrename from ");
2101 quote_c_style(name, &msg, NULL, 0);
2102 strbuf_addstr(&msg, "\nrename to ");
2103 quote_c_style(other, &msg, NULL, 0);
2104 strbuf_addch(&msg, '\n');
2105 break;
2106 case DIFF_STATUS_MODIFIED:
2107 if (p->score) {
2108 strbuf_addf(&msg, "dissimilarity index %d%%\n",
2109 similarity_index(p));
2110 complete_rewrite = 1;
2111 break;
2113 /* fallthru */
2114 default:
2115 /* nothing */
2119 if (hashcmp(one->sha1, two->sha1)) {
2120 int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
2122 if (DIFF_OPT_TST(o, BINARY)) {
2123 mmfile_t mf;
2124 if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
2125 (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
2126 abbrev = 40;
2128 strbuf_addf(&msg, "index %.*s..%.*s",
2129 abbrev, sha1_to_hex(one->sha1),
2130 abbrev, sha1_to_hex(two->sha1));
2131 if (one->mode == two->mode)
2132 strbuf_addf(&msg, " %06o", one->mode);
2133 strbuf_addch(&msg, '\n');
2136 if (msg.len)
2137 strbuf_setlen(&msg, msg.len - 1);
2138 xfrm_msg = msg.len ? msg.buf : NULL;
2140 if (!pgm &&
2141 DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
2142 (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
2143 /* a filepair that changes between file and symlink
2144 * needs to be split into deletion and creation.
2146 struct diff_filespec *null = alloc_filespec(two->path);
2147 run_diff_cmd(NULL, name, other, attr_path,
2148 one, null, xfrm_msg, o, 0);
2149 free(null);
2150 null = alloc_filespec(one->path);
2151 run_diff_cmd(NULL, name, other, attr_path,
2152 null, two, xfrm_msg, o, 0);
2153 free(null);
2155 else
2156 run_diff_cmd(pgm, name, other, attr_path,
2157 one, two, xfrm_msg, o, complete_rewrite);
2159 strbuf_release(&msg);
2162 static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
2163 struct diffstat_t *diffstat)
2165 const char *name;
2166 const char *other;
2167 int complete_rewrite = 0;
2169 if (DIFF_PAIR_UNMERGED(p)) {
2170 /* unmerged */
2171 builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, 0);
2172 return;
2175 name = p->one->path;
2176 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2178 if (o->prefix_length)
2179 strip_prefix(o->prefix_length, &name, &other);
2181 diff_fill_sha1_info(p->one);
2182 diff_fill_sha1_info(p->two);
2184 if (p->status == DIFF_STATUS_MODIFIED && p->score)
2185 complete_rewrite = 1;
2186 builtin_diffstat(name, other, p->one, p->two, diffstat, o, complete_rewrite);
2189 static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
2191 const char *name;
2192 const char *other;
2193 const char *attr_path;
2195 if (DIFF_PAIR_UNMERGED(p)) {
2196 /* unmerged */
2197 return;
2200 name = p->one->path;
2201 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2202 attr_path = other ? other : name;
2204 if (o->prefix_length)
2205 strip_prefix(o->prefix_length, &name, &other);
2207 diff_fill_sha1_info(p->one);
2208 diff_fill_sha1_info(p->two);
2210 builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
2213 void diff_setup(struct diff_options *options)
2215 memset(options, 0, sizeof(*options));
2217 options->file = stdout;
2219 options->line_termination = '\n';
2220 options->break_opt = -1;
2221 options->rename_limit = -1;
2222 options->dirstat_percent = 3;
2223 options->context = 3;
2225 options->change = diff_change;
2226 options->add_remove = diff_addremove;
2227 if (diff_use_color_default > 0)
2228 DIFF_OPT_SET(options, COLOR_DIFF);
2229 else
2230 DIFF_OPT_CLR(options, COLOR_DIFF);
2231 options->detect_rename = diff_detect_rename_default;
2233 options->a_prefix = "a/";
2234 options->b_prefix = "b/";
2237 int diff_setup_done(struct diff_options *options)
2239 int count = 0;
2241 if (options->output_format & DIFF_FORMAT_NAME)
2242 count++;
2243 if (options->output_format & DIFF_FORMAT_NAME_STATUS)
2244 count++;
2245 if (options->output_format & DIFF_FORMAT_CHECKDIFF)
2246 count++;
2247 if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
2248 count++;
2249 if (count > 1)
2250 die("--name-only, --name-status, --check and -s are mutually exclusive");
2252 if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
2253 options->detect_rename = DIFF_DETECT_COPY;
2255 if (!DIFF_OPT_TST(options, RELATIVE_NAME))
2256 options->prefix = NULL;
2257 if (options->prefix)
2258 options->prefix_length = strlen(options->prefix);
2259 else
2260 options->prefix_length = 0;
2262 if (options->output_format & (DIFF_FORMAT_NAME |
2263 DIFF_FORMAT_NAME_STATUS |
2264 DIFF_FORMAT_CHECKDIFF |
2265 DIFF_FORMAT_NO_OUTPUT))
2266 options->output_format &= ~(DIFF_FORMAT_RAW |
2267 DIFF_FORMAT_NUMSTAT |
2268 DIFF_FORMAT_DIFFSTAT |
2269 DIFF_FORMAT_SHORTSTAT |
2270 DIFF_FORMAT_DIRSTAT |
2271 DIFF_FORMAT_SUMMARY |
2272 DIFF_FORMAT_PATCH);
2275 * These cases always need recursive; we do not drop caller-supplied
2276 * recursive bits for other formats here.
2278 if (options->output_format & (DIFF_FORMAT_PATCH |
2279 DIFF_FORMAT_NUMSTAT |
2280 DIFF_FORMAT_DIFFSTAT |
2281 DIFF_FORMAT_SHORTSTAT |
2282 DIFF_FORMAT_DIRSTAT |
2283 DIFF_FORMAT_SUMMARY |
2284 DIFF_FORMAT_CHECKDIFF))
2285 DIFF_OPT_SET(options, RECURSIVE);
2287 * Also pickaxe would not work very well if you do not say recursive
2289 if (options->pickaxe)
2290 DIFF_OPT_SET(options, RECURSIVE);
2292 if (options->detect_rename && options->rename_limit < 0)
2293 options->rename_limit = diff_rename_limit_default;
2294 if (options->setup & DIFF_SETUP_USE_CACHE) {
2295 if (!active_cache)
2296 /* read-cache does not die even when it fails
2297 * so it is safe for us to do this here. Also
2298 * it does not smudge active_cache or active_nr
2299 * when it fails, so we do not have to worry about
2300 * cleaning it up ourselves either.
2302 read_cache();
2304 if (options->abbrev <= 0 || 40 < options->abbrev)
2305 options->abbrev = 40; /* full */
2308 * It does not make sense to show the first hit we happened
2309 * to have found. It does not make sense not to return with
2310 * exit code in such a case either.
2312 if (DIFF_OPT_TST(options, QUIET)) {
2313 options->output_format = DIFF_FORMAT_NO_OUTPUT;
2314 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
2318 * If we postprocess in diffcore, we cannot simply return
2319 * upon the first hit. We need to run diff as usual.
2321 if (options->pickaxe || options->filter)
2322 DIFF_OPT_CLR(options, QUIET);
2324 return 0;
2327 static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
2329 char c, *eq;
2330 int len;
2332 if (*arg != '-')
2333 return 0;
2334 c = *++arg;
2335 if (!c)
2336 return 0;
2337 if (c == arg_short) {
2338 c = *++arg;
2339 if (!c)
2340 return 1;
2341 if (val && isdigit(c)) {
2342 char *end;
2343 int n = strtoul(arg, &end, 10);
2344 if (*end)
2345 return 0;
2346 *val = n;
2347 return 1;
2349 return 0;
2351 if (c != '-')
2352 return 0;
2353 arg++;
2354 eq = strchr(arg, '=');
2355 if (eq)
2356 len = eq - arg;
2357 else
2358 len = strlen(arg);
2359 if (!len || strncmp(arg, arg_long, len))
2360 return 0;
2361 if (eq) {
2362 int n;
2363 char *end;
2364 if (!isdigit(*++eq))
2365 return 0;
2366 n = strtoul(eq, &end, 10);
2367 if (*end)
2368 return 0;
2369 *val = n;
2371 return 1;
2374 static int diff_scoreopt_parse(const char *opt);
2376 int diff_opt_parse(struct diff_options *options, const char **av, int ac)
2378 const char *arg = av[0];
2380 /* Output format options */
2381 if (!strcmp(arg, "-p") || !strcmp(arg, "-u"))
2382 options->output_format |= DIFF_FORMAT_PATCH;
2383 else if (opt_arg(arg, 'U', "unified", &options->context))
2384 options->output_format |= DIFF_FORMAT_PATCH;
2385 else if (!strcmp(arg, "--raw"))
2386 options->output_format |= DIFF_FORMAT_RAW;
2387 else if (!strcmp(arg, "--patch-with-raw"))
2388 options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_RAW;
2389 else if (!strcmp(arg, "--numstat"))
2390 options->output_format |= DIFF_FORMAT_NUMSTAT;
2391 else if (!strcmp(arg, "--shortstat"))
2392 options->output_format |= DIFF_FORMAT_SHORTSTAT;
2393 else if (opt_arg(arg, 'X', "dirstat", &options->dirstat_percent))
2394 options->output_format |= DIFF_FORMAT_DIRSTAT;
2395 else if (!strcmp(arg, "--cumulative"))
2396 options->output_format |= DIFF_FORMAT_CUMULATIVE;
2397 else if (!strcmp(arg, "--check"))
2398 options->output_format |= DIFF_FORMAT_CHECKDIFF;
2399 else if (!strcmp(arg, "--summary"))
2400 options->output_format |= DIFF_FORMAT_SUMMARY;
2401 else if (!strcmp(arg, "--patch-with-stat"))
2402 options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_DIFFSTAT;
2403 else if (!strcmp(arg, "--name-only"))
2404 options->output_format |= DIFF_FORMAT_NAME;
2405 else if (!strcmp(arg, "--name-status"))
2406 options->output_format |= DIFF_FORMAT_NAME_STATUS;
2407 else if (!strcmp(arg, "-s"))
2408 options->output_format |= DIFF_FORMAT_NO_OUTPUT;
2409 else if (!prefixcmp(arg, "--stat")) {
2410 char *end;
2411 int width = options->stat_width;
2412 int name_width = options->stat_name_width;
2413 arg += 6;
2414 end = (char *)arg;
2416 switch (*arg) {
2417 case '-':
2418 if (!prefixcmp(arg, "-width="))
2419 width = strtoul(arg + 7, &end, 10);
2420 else if (!prefixcmp(arg, "-name-width="))
2421 name_width = strtoul(arg + 12, &end, 10);
2422 break;
2423 case '=':
2424 width = strtoul(arg+1, &end, 10);
2425 if (*end == ',')
2426 name_width = strtoul(end+1, &end, 10);
2429 /* Important! This checks all the error cases! */
2430 if (*end)
2431 return 0;
2432 options->output_format |= DIFF_FORMAT_DIFFSTAT;
2433 options->stat_name_width = name_width;
2434 options->stat_width = width;
2437 /* renames options */
2438 else if (!prefixcmp(arg, "-B")) {
2439 if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
2440 return -1;
2442 else if (!prefixcmp(arg, "-M")) {
2443 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
2444 return -1;
2445 options->detect_rename = DIFF_DETECT_RENAME;
2447 else if (!prefixcmp(arg, "-C")) {
2448 if (options->detect_rename == DIFF_DETECT_COPY)
2449 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
2450 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
2451 return -1;
2452 options->detect_rename = DIFF_DETECT_COPY;
2454 else if (!strcmp(arg, "--no-renames"))
2455 options->detect_rename = 0;
2456 else if (!strcmp(arg, "--relative"))
2457 DIFF_OPT_SET(options, RELATIVE_NAME);
2458 else if (!prefixcmp(arg, "--relative=")) {
2459 DIFF_OPT_SET(options, RELATIVE_NAME);
2460 options->prefix = arg + 11;
2463 /* xdiff options */
2464 else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
2465 options->xdl_opts |= XDF_IGNORE_WHITESPACE;
2466 else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
2467 options->xdl_opts |= XDF_IGNORE_WHITESPACE_CHANGE;
2468 else if (!strcmp(arg, "--ignore-space-at-eol"))
2469 options->xdl_opts |= XDF_IGNORE_WHITESPACE_AT_EOL;
2471 /* flags options */
2472 else if (!strcmp(arg, "--binary")) {
2473 options->output_format |= DIFF_FORMAT_PATCH;
2474 DIFF_OPT_SET(options, BINARY);
2476 else if (!strcmp(arg, "--full-index"))
2477 DIFF_OPT_SET(options, FULL_INDEX);
2478 else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
2479 DIFF_OPT_SET(options, TEXT);
2480 else if (!strcmp(arg, "-R"))
2481 DIFF_OPT_SET(options, REVERSE_DIFF);
2482 else if (!strcmp(arg, "--find-copies-harder"))
2483 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
2484 else if (!strcmp(arg, "--follow"))
2485 DIFF_OPT_SET(options, FOLLOW_RENAMES);
2486 else if (!strcmp(arg, "--color"))
2487 DIFF_OPT_SET(options, COLOR_DIFF);
2488 else if (!strcmp(arg, "--no-color"))
2489 DIFF_OPT_CLR(options, COLOR_DIFF);
2490 else if (!strcmp(arg, "--color-words"))
2491 options->flags |= DIFF_OPT_COLOR_DIFF | DIFF_OPT_COLOR_DIFF_WORDS;
2492 else if (!strcmp(arg, "--exit-code"))
2493 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
2494 else if (!strcmp(arg, "--quiet"))
2495 DIFF_OPT_SET(options, QUIET);
2496 else if (!strcmp(arg, "--ext-diff"))
2497 DIFF_OPT_SET(options, ALLOW_EXTERNAL);
2498 else if (!strcmp(arg, "--no-ext-diff"))
2499 DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
2500 else if (!strcmp(arg, "--ignore-submodules"))
2501 DIFF_OPT_SET(options, IGNORE_SUBMODULES);
2503 /* misc options */
2504 else if (!strcmp(arg, "-z"))
2505 options->line_termination = 0;
2506 else if (!prefixcmp(arg, "-l"))
2507 options->rename_limit = strtoul(arg+2, NULL, 10);
2508 else if (!prefixcmp(arg, "-S"))
2509 options->pickaxe = arg + 2;
2510 else if (!strcmp(arg, "--pickaxe-all"))
2511 options->pickaxe_opts = DIFF_PICKAXE_ALL;
2512 else if (!strcmp(arg, "--pickaxe-regex"))
2513 options->pickaxe_opts = DIFF_PICKAXE_REGEX;
2514 else if (!prefixcmp(arg, "-O"))
2515 options->orderfile = arg + 2;
2516 else if (!prefixcmp(arg, "--diff-filter="))
2517 options->filter = arg + 14;
2518 else if (!strcmp(arg, "--abbrev"))
2519 options->abbrev = DEFAULT_ABBREV;
2520 else if (!prefixcmp(arg, "--abbrev=")) {
2521 options->abbrev = strtoul(arg + 9, NULL, 10);
2522 if (options->abbrev < MINIMUM_ABBREV)
2523 options->abbrev = MINIMUM_ABBREV;
2524 else if (40 < options->abbrev)
2525 options->abbrev = 40;
2527 else if (!prefixcmp(arg, "--src-prefix="))
2528 options->a_prefix = arg + 13;
2529 else if (!prefixcmp(arg, "--dst-prefix="))
2530 options->b_prefix = arg + 13;
2531 else if (!strcmp(arg, "--no-prefix"))
2532 options->a_prefix = options->b_prefix = "";
2533 else if (!prefixcmp(arg, "--output=")) {
2534 options->file = fopen(arg + strlen("--output="), "w");
2535 options->close_file = 1;
2536 } else
2537 return 0;
2538 return 1;
2541 static int parse_num(const char **cp_p)
2543 unsigned long num, scale;
2544 int ch, dot;
2545 const char *cp = *cp_p;
2547 num = 0;
2548 scale = 1;
2549 dot = 0;
2550 for(;;) {
2551 ch = *cp;
2552 if ( !dot && ch == '.' ) {
2553 scale = 1;
2554 dot = 1;
2555 } else if ( ch == '%' ) {
2556 scale = dot ? scale*100 : 100;
2557 cp++; /* % is always at the end */
2558 break;
2559 } else if ( ch >= '0' && ch <= '9' ) {
2560 if ( scale < 100000 ) {
2561 scale *= 10;
2562 num = (num*10) + (ch-'0');
2564 } else {
2565 break;
2567 cp++;
2569 *cp_p = cp;
2571 /* user says num divided by scale and we say internally that
2572 * is MAX_SCORE * num / scale.
2574 return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
2577 static int diff_scoreopt_parse(const char *opt)
2579 int opt1, opt2, cmd;
2581 if (*opt++ != '-')
2582 return -1;
2583 cmd = *opt++;
2584 if (cmd != 'M' && cmd != 'C' && cmd != 'B')
2585 return -1; /* that is not a -M, -C nor -B option */
2587 opt1 = parse_num(&opt);
2588 if (cmd != 'B')
2589 opt2 = 0;
2590 else {
2591 if (*opt == 0)
2592 opt2 = 0;
2593 else if (*opt != '/')
2594 return -1; /* we expect -B80/99 or -B80 */
2595 else {
2596 opt++;
2597 opt2 = parse_num(&opt);
2600 if (*opt != 0)
2601 return -1;
2602 return opt1 | (opt2 << 16);
2605 struct diff_queue_struct diff_queued_diff;
2607 void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
2609 if (queue->alloc <= queue->nr) {
2610 queue->alloc = alloc_nr(queue->alloc);
2611 queue->queue = xrealloc(queue->queue,
2612 sizeof(dp) * queue->alloc);
2614 queue->queue[queue->nr++] = dp;
2617 struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
2618 struct diff_filespec *one,
2619 struct diff_filespec *two)
2621 struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
2622 dp->one = one;
2623 dp->two = two;
2624 if (queue)
2625 diff_q(queue, dp);
2626 return dp;
2629 void diff_free_filepair(struct diff_filepair *p)
2631 free_filespec(p->one);
2632 free_filespec(p->two);
2633 free(p);
2636 /* This is different from find_unique_abbrev() in that
2637 * it stuffs the result with dots for alignment.
2639 const char *diff_unique_abbrev(const unsigned char *sha1, int len)
2641 int abblen;
2642 const char *abbrev;
2643 if (len == 40)
2644 return sha1_to_hex(sha1);
2646 abbrev = find_unique_abbrev(sha1, len);
2647 abblen = strlen(abbrev);
2648 if (abblen < 37) {
2649 static char hex[41];
2650 if (len < abblen && abblen <= len + 2)
2651 sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
2652 else
2653 sprintf(hex, "%s...", abbrev);
2654 return hex;
2656 return sha1_to_hex(sha1);
2659 static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
2661 int line_termination = opt->line_termination;
2662 int inter_name_termination = line_termination ? '\t' : '\0';
2664 if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
2665 fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
2666 diff_unique_abbrev(p->one->sha1, opt->abbrev));
2667 fprintf(opt->file, "%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
2669 if (p->score) {
2670 fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
2671 inter_name_termination);
2672 } else {
2673 fprintf(opt->file, "%c%c", p->status, inter_name_termination);
2676 if (p->status == DIFF_STATUS_COPIED ||
2677 p->status == DIFF_STATUS_RENAMED) {
2678 const char *name_a, *name_b;
2679 name_a = p->one->path;
2680 name_b = p->two->path;
2681 strip_prefix(opt->prefix_length, &name_a, &name_b);
2682 write_name_quoted(name_a, opt->file, inter_name_termination);
2683 write_name_quoted(name_b, opt->file, line_termination);
2684 } else {
2685 const char *name_a, *name_b;
2686 name_a = p->one->mode ? p->one->path : p->two->path;
2687 name_b = NULL;
2688 strip_prefix(opt->prefix_length, &name_a, &name_b);
2689 write_name_quoted(name_a, opt->file, line_termination);
2693 int diff_unmodified_pair(struct diff_filepair *p)
2695 /* This function is written stricter than necessary to support
2696 * the currently implemented transformers, but the idea is to
2697 * let transformers to produce diff_filepairs any way they want,
2698 * and filter and clean them up here before producing the output.
2700 struct diff_filespec *one = p->one, *two = p->two;
2702 if (DIFF_PAIR_UNMERGED(p))
2703 return 0; /* unmerged is interesting */
2705 /* deletion, addition, mode or type change
2706 * and rename are all interesting.
2708 if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
2709 DIFF_PAIR_MODE_CHANGED(p) ||
2710 strcmp(one->path, two->path))
2711 return 0;
2713 /* both are valid and point at the same path. that is, we are
2714 * dealing with a change.
2716 if (one->sha1_valid && two->sha1_valid &&
2717 !hashcmp(one->sha1, two->sha1))
2718 return 1; /* no change */
2719 if (!one->sha1_valid && !two->sha1_valid)
2720 return 1; /* both look at the same file on the filesystem. */
2721 return 0;
2724 static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
2726 if (diff_unmodified_pair(p))
2727 return;
2729 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2730 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2731 return; /* no tree diffs in patch format */
2733 run_diff(p, o);
2736 static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
2737 struct diffstat_t *diffstat)
2739 if (diff_unmodified_pair(p))
2740 return;
2742 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2743 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2744 return; /* no tree diffs in patch format */
2746 run_diffstat(p, o, diffstat);
2749 static void diff_flush_checkdiff(struct diff_filepair *p,
2750 struct diff_options *o)
2752 if (diff_unmodified_pair(p))
2753 return;
2755 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2756 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2757 return; /* no tree diffs in patch format */
2759 run_checkdiff(p, o);
2762 int diff_queue_is_empty(void)
2764 struct diff_queue_struct *q = &diff_queued_diff;
2765 int i;
2766 for (i = 0; i < q->nr; i++)
2767 if (!diff_unmodified_pair(q->queue[i]))
2768 return 0;
2769 return 1;
2772 #if DIFF_DEBUG
2773 void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
2775 fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
2776 x, one ? one : "",
2777 s->path,
2778 DIFF_FILE_VALID(s) ? "valid" : "invalid",
2779 s->mode,
2780 s->sha1_valid ? sha1_to_hex(s->sha1) : "");
2781 fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
2782 x, one ? one : "",
2783 s->size, s->xfrm_flags);
2786 void diff_debug_filepair(const struct diff_filepair *p, int i)
2788 diff_debug_filespec(p->one, i, "one");
2789 diff_debug_filespec(p->two, i, "two");
2790 fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
2791 p->score, p->status ? p->status : '?',
2792 p->one->rename_used, p->broken_pair);
2795 void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
2797 int i;
2798 if (msg)
2799 fprintf(stderr, "%s\n", msg);
2800 fprintf(stderr, "q->nr = %d\n", q->nr);
2801 for (i = 0; i < q->nr; i++) {
2802 struct diff_filepair *p = q->queue[i];
2803 diff_debug_filepair(p, i);
2806 #endif
2808 static void diff_resolve_rename_copy(void)
2810 int i;
2811 struct diff_filepair *p;
2812 struct diff_queue_struct *q = &diff_queued_diff;
2814 diff_debug_queue("resolve-rename-copy", q);
2816 for (i = 0; i < q->nr; i++) {
2817 p = q->queue[i];
2818 p->status = 0; /* undecided */
2819 if (DIFF_PAIR_UNMERGED(p))
2820 p->status = DIFF_STATUS_UNMERGED;
2821 else if (!DIFF_FILE_VALID(p->one))
2822 p->status = DIFF_STATUS_ADDED;
2823 else if (!DIFF_FILE_VALID(p->two))
2824 p->status = DIFF_STATUS_DELETED;
2825 else if (DIFF_PAIR_TYPE_CHANGED(p))
2826 p->status = DIFF_STATUS_TYPE_CHANGED;
2828 /* from this point on, we are dealing with a pair
2829 * whose both sides are valid and of the same type, i.e.
2830 * either in-place edit or rename/copy edit.
2832 else if (DIFF_PAIR_RENAME(p)) {
2834 * A rename might have re-connected a broken
2835 * pair up, causing the pathnames to be the
2836 * same again. If so, that's not a rename at
2837 * all, just a modification..
2839 * Otherwise, see if this source was used for
2840 * multiple renames, in which case we decrement
2841 * the count, and call it a copy.
2843 if (!strcmp(p->one->path, p->two->path))
2844 p->status = DIFF_STATUS_MODIFIED;
2845 else if (--p->one->rename_used > 0)
2846 p->status = DIFF_STATUS_COPIED;
2847 else
2848 p->status = DIFF_STATUS_RENAMED;
2850 else if (hashcmp(p->one->sha1, p->two->sha1) ||
2851 p->one->mode != p->two->mode ||
2852 is_null_sha1(p->one->sha1))
2853 p->status = DIFF_STATUS_MODIFIED;
2854 else {
2855 /* This is a "no-change" entry and should not
2856 * happen anymore, but prepare for broken callers.
2858 error("feeding unmodified %s to diffcore",
2859 p->one->path);
2860 p->status = DIFF_STATUS_UNKNOWN;
2863 diff_debug_queue("resolve-rename-copy done", q);
2866 static int check_pair_status(struct diff_filepair *p)
2868 switch (p->status) {
2869 case DIFF_STATUS_UNKNOWN:
2870 return 0;
2871 case 0:
2872 die("internal error in diff-resolve-rename-copy");
2873 default:
2874 return 1;
2878 static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
2880 int fmt = opt->output_format;
2882 if (fmt & DIFF_FORMAT_CHECKDIFF)
2883 diff_flush_checkdiff(p, opt);
2884 else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
2885 diff_flush_raw(p, opt);
2886 else if (fmt & DIFF_FORMAT_NAME) {
2887 const char *name_a, *name_b;
2888 name_a = p->two->path;
2889 name_b = NULL;
2890 strip_prefix(opt->prefix_length, &name_a, &name_b);
2891 write_name_quoted(name_a, opt->file, opt->line_termination);
2895 static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
2897 if (fs->mode)
2898 fprintf(file, " %s mode %06o ", newdelete, fs->mode);
2899 else
2900 fprintf(file, " %s ", newdelete);
2901 write_name_quoted(fs->path, file, '\n');
2905 static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name)
2907 if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
2908 fprintf(file, " mode change %06o => %06o%c", p->one->mode, p->two->mode,
2909 show_name ? ' ' : '\n');
2910 if (show_name) {
2911 write_name_quoted(p->two->path, file, '\n');
2916 static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p)
2918 char *names = pprint_rename(p->one->path, p->two->path);
2920 fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
2921 free(names);
2922 show_mode_change(file, p, 0);
2925 static void diff_summary(FILE *file, struct diff_filepair *p)
2927 switch(p->status) {
2928 case DIFF_STATUS_DELETED:
2929 show_file_mode_name(file, "delete", p->one);
2930 break;
2931 case DIFF_STATUS_ADDED:
2932 show_file_mode_name(file, "create", p->two);
2933 break;
2934 case DIFF_STATUS_COPIED:
2935 show_rename_copy(file, "copy", p);
2936 break;
2937 case DIFF_STATUS_RENAMED:
2938 show_rename_copy(file, "rename", p);
2939 break;
2940 default:
2941 if (p->score) {
2942 fputs(" rewrite ", file);
2943 write_name_quoted(p->two->path, file, ' ');
2944 fprintf(file, "(%d%%)\n", similarity_index(p));
2946 show_mode_change(file, p, !p->score);
2947 break;
2951 struct patch_id_t {
2952 struct xdiff_emit_state xm;
2953 SHA_CTX *ctx;
2954 int patchlen;
2957 static int remove_space(char *line, int len)
2959 int i;
2960 char *dst = line;
2961 unsigned char c;
2963 for (i = 0; i < len; i++)
2964 if (!isspace((c = line[i])))
2965 *dst++ = c;
2967 return dst - line;
2970 static void patch_id_consume(void *priv, char *line, unsigned long len)
2972 struct patch_id_t *data = priv;
2973 int new_len;
2975 /* Ignore line numbers when computing the SHA1 of the patch */
2976 if (!prefixcmp(line, "@@ -"))
2977 return;
2979 new_len = remove_space(line, len);
2981 SHA1_Update(data->ctx, line, new_len);
2982 data->patchlen += new_len;
2985 /* returns 0 upon success, and writes result into sha1 */
2986 static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
2988 struct diff_queue_struct *q = &diff_queued_diff;
2989 int i;
2990 SHA_CTX ctx;
2991 struct patch_id_t data;
2992 char buffer[PATH_MAX * 4 + 20];
2994 SHA1_Init(&ctx);
2995 memset(&data, 0, sizeof(struct patch_id_t));
2996 data.ctx = &ctx;
2997 data.xm.consume = patch_id_consume;
2999 for (i = 0; i < q->nr; i++) {
3000 xpparam_t xpp;
3001 xdemitconf_t xecfg;
3002 xdemitcb_t ecb;
3003 mmfile_t mf1, mf2;
3004 struct diff_filepair *p = q->queue[i];
3005 int len1, len2;
3007 memset(&xecfg, 0, sizeof(xecfg));
3008 if (p->status == 0)
3009 return error("internal diff status error");
3010 if (p->status == DIFF_STATUS_UNKNOWN)
3011 continue;
3012 if (diff_unmodified_pair(p))
3013 continue;
3014 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3015 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3016 continue;
3017 if (DIFF_PAIR_UNMERGED(p))
3018 continue;
3020 diff_fill_sha1_info(p->one);
3021 diff_fill_sha1_info(p->two);
3022 if (fill_mmfile(&mf1, p->one) < 0 ||
3023 fill_mmfile(&mf2, p->two) < 0)
3024 return error("unable to read files to diff");
3026 len1 = remove_space(p->one->path, strlen(p->one->path));
3027 len2 = remove_space(p->two->path, strlen(p->two->path));
3028 if (p->one->mode == 0)
3029 len1 = snprintf(buffer, sizeof(buffer),
3030 "diff--gita/%.*sb/%.*s"
3031 "newfilemode%06o"
3032 "---/dev/null"
3033 "+++b/%.*s",
3034 len1, p->one->path,
3035 len2, p->two->path,
3036 p->two->mode,
3037 len2, p->two->path);
3038 else if (p->two->mode == 0)
3039 len1 = snprintf(buffer, sizeof(buffer),
3040 "diff--gita/%.*sb/%.*s"
3041 "deletedfilemode%06o"
3042 "---a/%.*s"
3043 "+++/dev/null",
3044 len1, p->one->path,
3045 len2, p->two->path,
3046 p->one->mode,
3047 len1, p->one->path);
3048 else
3049 len1 = snprintf(buffer, sizeof(buffer),
3050 "diff--gita/%.*sb/%.*s"
3051 "---a/%.*s"
3052 "+++b/%.*s",
3053 len1, p->one->path,
3054 len2, p->two->path,
3055 len1, p->one->path,
3056 len2, p->two->path);
3057 SHA1_Update(&ctx, buffer, len1);
3059 xpp.flags = XDF_NEED_MINIMAL;
3060 xecfg.ctxlen = 3;
3061 xecfg.flags = XDL_EMIT_FUNCNAMES;
3062 ecb.outf = xdiff_outf;
3063 ecb.priv = &data;
3064 xdi_diff(&mf1, &mf2, &xpp, &xecfg, &ecb);
3067 SHA1_Final(sha1, &ctx);
3068 return 0;
3071 int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
3073 struct diff_queue_struct *q = &diff_queued_diff;
3074 int i;
3075 int result = diff_get_patch_id(options, sha1);
3077 for (i = 0; i < q->nr; i++)
3078 diff_free_filepair(q->queue[i]);
3080 free(q->queue);
3081 q->queue = NULL;
3082 q->nr = q->alloc = 0;
3084 return result;
3087 static int is_summary_empty(const struct diff_queue_struct *q)
3089 int i;
3091 for (i = 0; i < q->nr; i++) {
3092 const struct diff_filepair *p = q->queue[i];
3094 switch (p->status) {
3095 case DIFF_STATUS_DELETED:
3096 case DIFF_STATUS_ADDED:
3097 case DIFF_STATUS_COPIED:
3098 case DIFF_STATUS_RENAMED:
3099 return 0;
3100 default:
3101 if (p->score)
3102 return 0;
3103 if (p->one->mode && p->two->mode &&
3104 p->one->mode != p->two->mode)
3105 return 0;
3106 break;
3109 return 1;
3112 void diff_flush(struct diff_options *options)
3114 struct diff_queue_struct *q = &diff_queued_diff;
3115 int i, output_format = options->output_format;
3116 int separator = 0;
3119 * Order: raw, stat, summary, patch
3120 * or: name/name-status/checkdiff (other bits clear)
3122 if (!q->nr)
3123 goto free_queue;
3125 if (output_format & (DIFF_FORMAT_RAW |
3126 DIFF_FORMAT_NAME |
3127 DIFF_FORMAT_NAME_STATUS |
3128 DIFF_FORMAT_CHECKDIFF)) {
3129 for (i = 0; i < q->nr; i++) {
3130 struct diff_filepair *p = q->queue[i];
3131 if (check_pair_status(p))
3132 flush_one_pair(p, options);
3134 separator++;
3137 if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT)) {
3138 struct diffstat_t diffstat;
3140 memset(&diffstat, 0, sizeof(struct diffstat_t));
3141 diffstat.xm.consume = diffstat_consume;
3142 for (i = 0; i < q->nr; i++) {
3143 struct diff_filepair *p = q->queue[i];
3144 if (check_pair_status(p))
3145 diff_flush_stat(p, options, &diffstat);
3147 if (output_format & DIFF_FORMAT_NUMSTAT)
3148 show_numstat(&diffstat, options);
3149 if (output_format & DIFF_FORMAT_DIFFSTAT)
3150 show_stats(&diffstat, options);
3151 if (output_format & DIFF_FORMAT_SHORTSTAT)
3152 show_shortstats(&diffstat, options);
3153 free_diffstat_info(&diffstat);
3154 separator++;
3156 if (output_format & DIFF_FORMAT_DIRSTAT)
3157 show_dirstat(options);
3159 if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
3160 for (i = 0; i < q->nr; i++)
3161 diff_summary(options->file, q->queue[i]);
3162 separator++;
3165 if (output_format & DIFF_FORMAT_PATCH) {
3166 if (separator) {
3167 if (options->stat_sep) {
3168 /* attach patch instead of inline */
3169 fputs(options->stat_sep, options->file);
3170 } else {
3171 putc(options->line_termination, options->file);
3175 for (i = 0; i < q->nr; i++) {
3176 struct diff_filepair *p = q->queue[i];
3177 if (check_pair_status(p))
3178 diff_flush_patch(p, options);
3182 if (output_format & DIFF_FORMAT_CALLBACK)
3183 options->format_callback(q, options, options->format_callback_data);
3185 for (i = 0; i < q->nr; i++)
3186 diff_free_filepair(q->queue[i]);
3187 free_queue:
3188 free(q->queue);
3189 q->queue = NULL;
3190 q->nr = q->alloc = 0;
3191 if (options->close_file)
3192 fclose(options->file);
3195 static void diffcore_apply_filter(const char *filter)
3197 int i;
3198 struct diff_queue_struct *q = &diff_queued_diff;
3199 struct diff_queue_struct outq;
3200 outq.queue = NULL;
3201 outq.nr = outq.alloc = 0;
3203 if (!filter)
3204 return;
3206 if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
3207 int found;
3208 for (i = found = 0; !found && i < q->nr; i++) {
3209 struct diff_filepair *p = q->queue[i];
3210 if (((p->status == DIFF_STATUS_MODIFIED) &&
3211 ((p->score &&
3212 strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
3213 (!p->score &&
3214 strchr(filter, DIFF_STATUS_MODIFIED)))) ||
3215 ((p->status != DIFF_STATUS_MODIFIED) &&
3216 strchr(filter, p->status)))
3217 found++;
3219 if (found)
3220 return;
3222 /* otherwise we will clear the whole queue
3223 * by copying the empty outq at the end of this
3224 * function, but first clear the current entries
3225 * in the queue.
3227 for (i = 0; i < q->nr; i++)
3228 diff_free_filepair(q->queue[i]);
3230 else {
3231 /* Only the matching ones */
3232 for (i = 0; i < q->nr; i++) {
3233 struct diff_filepair *p = q->queue[i];
3235 if (((p->status == DIFF_STATUS_MODIFIED) &&
3236 ((p->score &&
3237 strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
3238 (!p->score &&
3239 strchr(filter, DIFF_STATUS_MODIFIED)))) ||
3240 ((p->status != DIFF_STATUS_MODIFIED) &&
3241 strchr(filter, p->status)))
3242 diff_q(&outq, p);
3243 else
3244 diff_free_filepair(p);
3247 free(q->queue);
3248 *q = outq;
3251 /* Check whether two filespecs with the same mode and size are identical */
3252 static int diff_filespec_is_identical(struct diff_filespec *one,
3253 struct diff_filespec *two)
3255 if (S_ISGITLINK(one->mode))
3256 return 0;
3257 if (diff_populate_filespec(one, 0))
3258 return 0;
3259 if (diff_populate_filespec(two, 0))
3260 return 0;
3261 return !memcmp(one->data, two->data, one->size);
3264 static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
3266 int i;
3267 struct diff_queue_struct *q = &diff_queued_diff;
3268 struct diff_queue_struct outq;
3269 outq.queue = NULL;
3270 outq.nr = outq.alloc = 0;
3272 for (i = 0; i < q->nr; i++) {
3273 struct diff_filepair *p = q->queue[i];
3276 * 1. Entries that come from stat info dirtyness
3277 * always have both sides (iow, not create/delete),
3278 * one side of the object name is unknown, with
3279 * the same mode and size. Keep the ones that
3280 * do not match these criteria. They have real
3281 * differences.
3283 * 2. At this point, the file is known to be modified,
3284 * with the same mode and size, and the object
3285 * name of one side is unknown. Need to inspect
3286 * the identical contents.
3288 if (!DIFF_FILE_VALID(p->one) || /* (1) */
3289 !DIFF_FILE_VALID(p->two) ||
3290 (p->one->sha1_valid && p->two->sha1_valid) ||
3291 (p->one->mode != p->two->mode) ||
3292 diff_populate_filespec(p->one, 1) ||
3293 diff_populate_filespec(p->two, 1) ||
3294 (p->one->size != p->two->size) ||
3295 !diff_filespec_is_identical(p->one, p->two)) /* (2) */
3296 diff_q(&outq, p);
3297 else {
3299 * The caller can subtract 1 from skip_stat_unmatch
3300 * to determine how many paths were dirty only
3301 * due to stat info mismatch.
3303 if (!DIFF_OPT_TST(diffopt, NO_INDEX))
3304 diffopt->skip_stat_unmatch++;
3305 diff_free_filepair(p);
3308 free(q->queue);
3309 *q = outq;
3312 void diffcore_std(struct diff_options *options)
3314 if (DIFF_OPT_TST(options, QUIET))
3315 return;
3317 if (options->skip_stat_unmatch && !DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3318 diffcore_skip_stat_unmatch(options);
3319 if (options->break_opt != -1)
3320 diffcore_break(options->break_opt);
3321 if (options->detect_rename)
3322 diffcore_rename(options);
3323 if (options->break_opt != -1)
3324 diffcore_merge_broken();
3325 if (options->pickaxe)
3326 diffcore_pickaxe(options->pickaxe, options->pickaxe_opts);
3327 if (options->orderfile)
3328 diffcore_order(options->orderfile);
3329 diff_resolve_rename_copy();
3330 diffcore_apply_filter(options->filter);
3332 if (diff_queued_diff.nr)
3333 DIFF_OPT_SET(options, HAS_CHANGES);
3334 else
3335 DIFF_OPT_CLR(options, HAS_CHANGES);
3338 int diff_result_code(struct diff_options *opt, int status)
3340 int result = 0;
3341 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
3342 !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
3343 return status;
3344 if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
3345 DIFF_OPT_TST(opt, HAS_CHANGES))
3346 result |= 01;
3347 if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
3348 DIFF_OPT_TST(opt, CHECK_FAILED))
3349 result |= 02;
3350 return result;
3353 void diff_addremove(struct diff_options *options,
3354 int addremove, unsigned mode,
3355 const unsigned char *sha1,
3356 const char *base, const char *path)
3358 char concatpath[PATH_MAX];
3359 struct diff_filespec *one, *two;
3361 if (DIFF_OPT_TST(options, IGNORE_SUBMODULES) && S_ISGITLINK(mode))
3362 return;
3364 /* This may look odd, but it is a preparation for
3365 * feeding "there are unchanged files which should
3366 * not produce diffs, but when you are doing copy
3367 * detection you would need them, so here they are"
3368 * entries to the diff-core. They will be prefixed
3369 * with something like '=' or '*' (I haven't decided
3370 * which but should not make any difference).
3371 * Feeding the same new and old to diff_change()
3372 * also has the same effect.
3373 * Before the final output happens, they are pruned after
3374 * merged into rename/copy pairs as appropriate.
3376 if (DIFF_OPT_TST(options, REVERSE_DIFF))
3377 addremove = (addremove == '+' ? '-' :
3378 addremove == '-' ? '+' : addremove);
3380 if (!path) path = "";
3381 sprintf(concatpath, "%s%s", base, path);
3383 if (options->prefix &&
3384 strncmp(concatpath, options->prefix, options->prefix_length))
3385 return;
3387 one = alloc_filespec(concatpath);
3388 two = alloc_filespec(concatpath);
3390 if (addremove != '+')
3391 fill_filespec(one, sha1, mode);
3392 if (addremove != '-')
3393 fill_filespec(two, sha1, mode);
3395 diff_queue(&diff_queued_diff, one, two);
3396 DIFF_OPT_SET(options, HAS_CHANGES);
3399 void diff_change(struct diff_options *options,
3400 unsigned old_mode, unsigned new_mode,
3401 const unsigned char *old_sha1,
3402 const unsigned char *new_sha1,
3403 const char *base, const char *path)
3405 char concatpath[PATH_MAX];
3406 struct diff_filespec *one, *two;
3408 if (DIFF_OPT_TST(options, IGNORE_SUBMODULES) && S_ISGITLINK(old_mode)
3409 && S_ISGITLINK(new_mode))
3410 return;
3412 if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
3413 unsigned tmp;
3414 const unsigned char *tmp_c;
3415 tmp = old_mode; old_mode = new_mode; new_mode = tmp;
3416 tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
3418 if (!path) path = "";
3419 sprintf(concatpath, "%s%s", base, path);
3421 if (options->prefix &&
3422 strncmp(concatpath, options->prefix, options->prefix_length))
3423 return;
3425 one = alloc_filespec(concatpath);
3426 two = alloc_filespec(concatpath);
3427 fill_filespec(one, old_sha1, old_mode);
3428 fill_filespec(two, new_sha1, new_mode);
3430 diff_queue(&diff_queued_diff, one, two);
3431 DIFF_OPT_SET(options, HAS_CHANGES);
3434 void diff_unmerge(struct diff_options *options,
3435 const char *path,
3436 unsigned mode, const unsigned char *sha1)
3438 struct diff_filespec *one, *two;
3440 if (options->prefix &&
3441 strncmp(path, options->prefix, options->prefix_length))
3442 return;
3444 one = alloc_filespec(path);
3445 two = alloc_filespec(path);
3446 fill_filespec(one, sha1, mode);
3447 diff_queue(&diff_queued_diff, one, two)->is_unmerged = 1;