Merge branch 'cn/branch-set-upstream-to'
[git.git] / diff.c
blobe6846ca7507aa8a423c5e1f89390fa37723248ff
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"
14 #include "userdiff.h"
15 #include "sigchain.h"
16 #include "submodule.h"
17 #include "ll-merge.h"
19 #ifdef NO_FAST_WORKING_DIRECTORY
20 #define FAST_WORKING_DIRECTORY 0
21 #else
22 #define FAST_WORKING_DIRECTORY 1
23 #endif
25 static int diff_detect_rename_default;
26 static int diff_rename_limit_default = 400;
27 static int diff_suppress_blank_empty;
28 int diff_use_color_default = -1;
29 static const char *diff_word_regex_cfg;
30 static const char *external_diff_cmd_cfg;
31 int diff_auto_refresh_index = 1;
32 static int diff_mnemonic_prefix;
33 static int diff_no_prefix;
34 static int diff_stat_graph_width;
35 static int diff_dirstat_permille_default = 30;
36 static struct diff_options default_diff_options;
38 static char diff_colors[][COLOR_MAXLEN] = {
39 GIT_COLOR_RESET,
40 GIT_COLOR_NORMAL, /* PLAIN */
41 GIT_COLOR_BOLD, /* METAINFO */
42 GIT_COLOR_CYAN, /* FRAGINFO */
43 GIT_COLOR_RED, /* OLD */
44 GIT_COLOR_GREEN, /* NEW */
45 GIT_COLOR_YELLOW, /* COMMIT */
46 GIT_COLOR_BG_RED, /* WHITESPACE */
47 GIT_COLOR_NORMAL, /* FUNCINFO */
50 static int parse_diff_color_slot(const char *var, int ofs)
52 if (!strcasecmp(var+ofs, "plain"))
53 return DIFF_PLAIN;
54 if (!strcasecmp(var+ofs, "meta"))
55 return DIFF_METAINFO;
56 if (!strcasecmp(var+ofs, "frag"))
57 return DIFF_FRAGINFO;
58 if (!strcasecmp(var+ofs, "old"))
59 return DIFF_FILE_OLD;
60 if (!strcasecmp(var+ofs, "new"))
61 return DIFF_FILE_NEW;
62 if (!strcasecmp(var+ofs, "commit"))
63 return DIFF_COMMIT;
64 if (!strcasecmp(var+ofs, "whitespace"))
65 return DIFF_WHITESPACE;
66 if (!strcasecmp(var+ofs, "func"))
67 return DIFF_FUNCINFO;
68 return -1;
71 static int parse_dirstat_params(struct diff_options *options, const char *params,
72 struct strbuf *errmsg)
74 const char *p = params;
75 int p_len, ret = 0;
77 while (*p) {
78 p_len = strchrnul(p, ',') - p;
79 if (!memcmp(p, "changes", p_len)) {
80 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
81 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
82 } else if (!memcmp(p, "lines", p_len)) {
83 DIFF_OPT_SET(options, DIRSTAT_BY_LINE);
84 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
85 } else if (!memcmp(p, "files", p_len)) {
86 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
87 DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
88 } else if (!memcmp(p, "noncumulative", p_len)) {
89 DIFF_OPT_CLR(options, DIRSTAT_CUMULATIVE);
90 } else if (!memcmp(p, "cumulative", p_len)) {
91 DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
92 } else if (isdigit(*p)) {
93 char *end;
94 int permille = strtoul(p, &end, 10) * 10;
95 if (*end == '.' && isdigit(*++end)) {
96 /* only use first digit */
97 permille += *end - '0';
98 /* .. and ignore any further digits */
99 while (isdigit(*++end))
100 ; /* nothing */
102 if (end - p == p_len)
103 options->dirstat_permille = permille;
104 else {
105 strbuf_addf(errmsg, _(" Failed to parse dirstat cut-off percentage '%.*s'\n"),
106 p_len, p);
107 ret++;
109 } else {
110 strbuf_addf(errmsg, _(" Unknown dirstat parameter '%.*s'\n"),
111 p_len, p);
112 ret++;
115 p += p_len;
117 if (*p)
118 p++; /* more parameters, swallow separator */
120 return ret;
123 static int git_config_rename(const char *var, const char *value)
125 if (!value)
126 return DIFF_DETECT_RENAME;
127 if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
128 return DIFF_DETECT_COPY;
129 return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
133 * These are to give UI layer defaults.
134 * The core-level commands such as git-diff-files should
135 * never be affected by the setting of diff.renames
136 * the user happens to have in the configuration file.
138 int git_diff_ui_config(const char *var, const char *value, void *cb)
140 if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
141 diff_use_color_default = git_config_colorbool(var, value);
142 return 0;
144 if (!strcmp(var, "diff.renames")) {
145 diff_detect_rename_default = git_config_rename(var, value);
146 return 0;
148 if (!strcmp(var, "diff.autorefreshindex")) {
149 diff_auto_refresh_index = git_config_bool(var, value);
150 return 0;
152 if (!strcmp(var, "diff.mnemonicprefix")) {
153 diff_mnemonic_prefix = git_config_bool(var, value);
154 return 0;
156 if (!strcmp(var, "diff.noprefix")) {
157 diff_no_prefix = git_config_bool(var, value);
158 return 0;
160 if (!strcmp(var, "diff.statgraphwidth")) {
161 diff_stat_graph_width = git_config_int(var, value);
162 return 0;
164 if (!strcmp(var, "diff.external"))
165 return git_config_string(&external_diff_cmd_cfg, var, value);
166 if (!strcmp(var, "diff.wordregex"))
167 return git_config_string(&diff_word_regex_cfg, var, value);
169 if (!strcmp(var, "diff.ignoresubmodules"))
170 handle_ignore_submodules_arg(&default_diff_options, value);
172 if (git_color_config(var, value, cb) < 0)
173 return -1;
175 return git_diff_basic_config(var, value, cb);
178 int git_diff_basic_config(const char *var, const char *value, void *cb)
180 if (!strcmp(var, "diff.renamelimit")) {
181 diff_rename_limit_default = git_config_int(var, value);
182 return 0;
185 if (userdiff_config(var, value) < 0)
186 return -1;
188 if (!prefixcmp(var, "diff.color.") || !prefixcmp(var, "color.diff.")) {
189 int slot = parse_diff_color_slot(var, 11);
190 if (slot < 0)
191 return 0;
192 if (!value)
193 return config_error_nonbool(var);
194 color_parse(value, var, diff_colors[slot]);
195 return 0;
198 /* like GNU diff's --suppress-blank-empty option */
199 if (!strcmp(var, "diff.suppressblankempty") ||
200 /* for backwards compatibility */
201 !strcmp(var, "diff.suppress-blank-empty")) {
202 diff_suppress_blank_empty = git_config_bool(var, value);
203 return 0;
206 if (!strcmp(var, "diff.dirstat")) {
207 struct strbuf errmsg = STRBUF_INIT;
208 default_diff_options.dirstat_permille = diff_dirstat_permille_default;
209 if (parse_dirstat_params(&default_diff_options, value, &errmsg))
210 warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
211 errmsg.buf);
212 strbuf_release(&errmsg);
213 diff_dirstat_permille_default = default_diff_options.dirstat_permille;
214 return 0;
217 if (!prefixcmp(var, "submodule."))
218 return parse_submodule_config_option(var, value);
220 return git_default_config(var, value, cb);
223 static char *quote_two(const char *one, const char *two)
225 int need_one = quote_c_style(one, NULL, NULL, 1);
226 int need_two = quote_c_style(two, NULL, NULL, 1);
227 struct strbuf res = STRBUF_INIT;
229 if (need_one + need_two) {
230 strbuf_addch(&res, '"');
231 quote_c_style(one, &res, NULL, 1);
232 quote_c_style(two, &res, NULL, 1);
233 strbuf_addch(&res, '"');
234 } else {
235 strbuf_addstr(&res, one);
236 strbuf_addstr(&res, two);
238 return strbuf_detach(&res, NULL);
241 static const char *external_diff(void)
243 static const char *external_diff_cmd = NULL;
244 static int done_preparing = 0;
246 if (done_preparing)
247 return external_diff_cmd;
248 external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
249 if (!external_diff_cmd)
250 external_diff_cmd = external_diff_cmd_cfg;
251 done_preparing = 1;
252 return external_diff_cmd;
255 static struct diff_tempfile {
256 const char *name; /* filename external diff should read from */
257 char hex[41];
258 char mode[10];
259 char tmp_path[PATH_MAX];
260 } diff_temp[2];
262 typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
264 struct emit_callback {
265 int color_diff;
266 unsigned ws_rule;
267 int blank_at_eof_in_preimage;
268 int blank_at_eof_in_postimage;
269 int lno_in_preimage;
270 int lno_in_postimage;
271 sane_truncate_fn truncate;
272 const char **label_path;
273 struct diff_words_data *diff_words;
274 struct diff_options *opt;
275 int *found_changesp;
276 struct strbuf *header;
279 static int count_lines(const char *data, int size)
281 int count, ch, completely_empty = 1, nl_just_seen = 0;
282 count = 0;
283 while (0 < size--) {
284 ch = *data++;
285 if (ch == '\n') {
286 count++;
287 nl_just_seen = 1;
288 completely_empty = 0;
290 else {
291 nl_just_seen = 0;
292 completely_empty = 0;
295 if (completely_empty)
296 return 0;
297 if (!nl_just_seen)
298 count++; /* no trailing newline */
299 return count;
302 static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
304 if (!DIFF_FILE_VALID(one)) {
305 mf->ptr = (char *)""; /* does not matter */
306 mf->size = 0;
307 return 0;
309 else if (diff_populate_filespec(one, 0))
310 return -1;
312 mf->ptr = one->data;
313 mf->size = one->size;
314 return 0;
317 /* like fill_mmfile, but only for size, so we can avoid retrieving blob */
318 static unsigned long diff_filespec_size(struct diff_filespec *one)
320 if (!DIFF_FILE_VALID(one))
321 return 0;
322 diff_populate_filespec(one, 1);
323 return one->size;
326 static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
328 char *ptr = mf->ptr;
329 long size = mf->size;
330 int cnt = 0;
332 if (!size)
333 return cnt;
334 ptr += size - 1; /* pointing at the very end */
335 if (*ptr != '\n')
336 ; /* incomplete line */
337 else
338 ptr--; /* skip the last LF */
339 while (mf->ptr < ptr) {
340 char *prev_eol;
341 for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
342 if (*prev_eol == '\n')
343 break;
344 if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
345 break;
346 cnt++;
347 ptr = prev_eol - 1;
349 return cnt;
352 static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
353 struct emit_callback *ecbdata)
355 int l1, l2, at;
356 unsigned ws_rule = ecbdata->ws_rule;
357 l1 = count_trailing_blank(mf1, ws_rule);
358 l2 = count_trailing_blank(mf2, ws_rule);
359 if (l2 <= l1) {
360 ecbdata->blank_at_eof_in_preimage = 0;
361 ecbdata->blank_at_eof_in_postimage = 0;
362 return;
364 at = count_lines(mf1->ptr, mf1->size);
365 ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
367 at = count_lines(mf2->ptr, mf2->size);
368 ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
371 static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
372 int first, const char *line, int len)
374 int has_trailing_newline, has_trailing_carriage_return;
375 int nofirst;
376 FILE *file = o->file;
378 if (o->output_prefix) {
379 struct strbuf *msg = NULL;
380 msg = o->output_prefix(o, o->output_prefix_data);
381 assert(msg);
382 fwrite(msg->buf, msg->len, 1, file);
385 if (len == 0) {
386 has_trailing_newline = (first == '\n');
387 has_trailing_carriage_return = (!has_trailing_newline &&
388 (first == '\r'));
389 nofirst = has_trailing_newline || has_trailing_carriage_return;
390 } else {
391 has_trailing_newline = (len > 0 && line[len-1] == '\n');
392 if (has_trailing_newline)
393 len--;
394 has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
395 if (has_trailing_carriage_return)
396 len--;
397 nofirst = 0;
400 if (len || !nofirst) {
401 fputs(set, file);
402 if (!nofirst)
403 fputc(first, file);
404 fwrite(line, len, 1, file);
405 fputs(reset, file);
407 if (has_trailing_carriage_return)
408 fputc('\r', file);
409 if (has_trailing_newline)
410 fputc('\n', file);
413 static void emit_line(struct diff_options *o, const char *set, const char *reset,
414 const char *line, int len)
416 emit_line_0(o, set, reset, line[0], line+1, len-1);
419 static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
421 if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
422 ecbdata->blank_at_eof_in_preimage &&
423 ecbdata->blank_at_eof_in_postimage &&
424 ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
425 ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
426 return 0;
427 return ws_blank_line(line, len, ecbdata->ws_rule);
430 static void emit_add_line(const char *reset,
431 struct emit_callback *ecbdata,
432 const char *line, int len)
434 const char *ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
435 const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_NEW);
437 if (!*ws)
438 emit_line_0(ecbdata->opt, set, reset, '+', line, len);
439 else if (new_blank_line_at_eof(ecbdata, line, len))
440 /* Blank line at EOF - paint '+' as well */
441 emit_line_0(ecbdata->opt, ws, reset, '+', line, len);
442 else {
443 /* Emit just the prefix, then the rest. */
444 emit_line_0(ecbdata->opt, set, reset, '+', "", 0);
445 ws_check_emit(line, len, ecbdata->ws_rule,
446 ecbdata->opt->file, set, reset, ws);
450 static void emit_hunk_header(struct emit_callback *ecbdata,
451 const char *line, int len)
453 const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
454 const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
455 const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
456 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
457 static const char atat[2] = { '@', '@' };
458 const char *cp, *ep;
459 struct strbuf msgbuf = STRBUF_INIT;
460 int org_len = len;
461 int i = 1;
464 * As a hunk header must begin with "@@ -<old>, +<new> @@",
465 * it always is at least 10 bytes long.
467 if (len < 10 ||
468 memcmp(line, atat, 2) ||
469 !(ep = memmem(line + 2, len - 2, atat, 2))) {
470 emit_line(ecbdata->opt, plain, reset, line, len);
471 return;
473 ep += 2; /* skip over @@ */
475 /* The hunk header in fraginfo color */
476 strbuf_add(&msgbuf, frag, strlen(frag));
477 strbuf_add(&msgbuf, line, ep - line);
478 strbuf_add(&msgbuf, reset, strlen(reset));
481 * trailing "\r\n"
483 for ( ; i < 3; i++)
484 if (line[len - i] == '\r' || line[len - i] == '\n')
485 len--;
487 /* blank before the func header */
488 for (cp = ep; ep - line < len; ep++)
489 if (*ep != ' ' && *ep != '\t')
490 break;
491 if (ep != cp) {
492 strbuf_add(&msgbuf, plain, strlen(plain));
493 strbuf_add(&msgbuf, cp, ep - cp);
494 strbuf_add(&msgbuf, reset, strlen(reset));
497 if (ep < line + len) {
498 strbuf_add(&msgbuf, func, strlen(func));
499 strbuf_add(&msgbuf, ep, line + len - ep);
500 strbuf_add(&msgbuf, reset, strlen(reset));
503 strbuf_add(&msgbuf, line + len, org_len - len);
504 emit_line(ecbdata->opt, "", "", msgbuf.buf, msgbuf.len);
505 strbuf_release(&msgbuf);
508 static struct diff_tempfile *claim_diff_tempfile(void) {
509 int i;
510 for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
511 if (!diff_temp[i].name)
512 return diff_temp + i;
513 die("BUG: diff is failing to clean up its tempfiles");
516 static int remove_tempfile_installed;
518 static void remove_tempfile(void)
520 int i;
521 for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
522 if (diff_temp[i].name == diff_temp[i].tmp_path)
523 unlink_or_warn(diff_temp[i].name);
524 diff_temp[i].name = NULL;
528 static void remove_tempfile_on_signal(int signo)
530 remove_tempfile();
531 sigchain_pop(signo);
532 raise(signo);
535 static void print_line_count(FILE *file, int count)
537 switch (count) {
538 case 0:
539 fprintf(file, "0,0");
540 break;
541 case 1:
542 fprintf(file, "1");
543 break;
544 default:
545 fprintf(file, "1,%d", count);
546 break;
550 static void emit_rewrite_lines(struct emit_callback *ecb,
551 int prefix, const char *data, int size)
553 const char *endp = NULL;
554 static const char *nneof = " No newline at end of file\n";
555 const char *old = diff_get_color(ecb->color_diff, DIFF_FILE_OLD);
556 const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
558 while (0 < size) {
559 int len;
561 endp = memchr(data, '\n', size);
562 len = endp ? (endp - data + 1) : size;
563 if (prefix != '+') {
564 ecb->lno_in_preimage++;
565 emit_line_0(ecb->opt, old, reset, '-',
566 data, len);
567 } else {
568 ecb->lno_in_postimage++;
569 emit_add_line(reset, ecb, data, len);
571 size -= len;
572 data += len;
574 if (!endp) {
575 const char *plain = diff_get_color(ecb->color_diff,
576 DIFF_PLAIN);
577 putc('\n', ecb->opt->file);
578 emit_line_0(ecb->opt, plain, reset, '\\',
579 nneof, strlen(nneof));
583 static void emit_rewrite_diff(const char *name_a,
584 const char *name_b,
585 struct diff_filespec *one,
586 struct diff_filespec *two,
587 struct userdiff_driver *textconv_one,
588 struct userdiff_driver *textconv_two,
589 struct diff_options *o)
591 int lc_a, lc_b;
592 const char *name_a_tab, *name_b_tab;
593 const char *metainfo = diff_get_color(o->use_color, DIFF_METAINFO);
594 const char *fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
595 const char *reset = diff_get_color(o->use_color, DIFF_RESET);
596 static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
597 const char *a_prefix, *b_prefix;
598 char *data_one, *data_two;
599 size_t size_one, size_two;
600 struct emit_callback ecbdata;
601 char *line_prefix = "";
602 struct strbuf *msgbuf;
604 if (o && o->output_prefix) {
605 msgbuf = o->output_prefix(o, o->output_prefix_data);
606 line_prefix = msgbuf->buf;
609 if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
610 a_prefix = o->b_prefix;
611 b_prefix = o->a_prefix;
612 } else {
613 a_prefix = o->a_prefix;
614 b_prefix = o->b_prefix;
617 name_a += (*name_a == '/');
618 name_b += (*name_b == '/');
619 name_a_tab = strchr(name_a, ' ') ? "\t" : "";
620 name_b_tab = strchr(name_b, ' ') ? "\t" : "";
622 strbuf_reset(&a_name);
623 strbuf_reset(&b_name);
624 quote_two_c_style(&a_name, a_prefix, name_a, 0);
625 quote_two_c_style(&b_name, b_prefix, name_b, 0);
627 size_one = fill_textconv(textconv_one, one, &data_one);
628 size_two = fill_textconv(textconv_two, two, &data_two);
630 memset(&ecbdata, 0, sizeof(ecbdata));
631 ecbdata.color_diff = want_color(o->use_color);
632 ecbdata.found_changesp = &o->found_changes;
633 ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
634 ecbdata.opt = o;
635 if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
636 mmfile_t mf1, mf2;
637 mf1.ptr = (char *)data_one;
638 mf2.ptr = (char *)data_two;
639 mf1.size = size_one;
640 mf2.size = size_two;
641 check_blank_at_eof(&mf1, &mf2, &ecbdata);
643 ecbdata.lno_in_preimage = 1;
644 ecbdata.lno_in_postimage = 1;
646 lc_a = count_lines(data_one, size_one);
647 lc_b = count_lines(data_two, size_two);
648 fprintf(o->file,
649 "%s%s--- %s%s%s\n%s%s+++ %s%s%s\n%s%s@@ -",
650 line_prefix, metainfo, a_name.buf, name_a_tab, reset,
651 line_prefix, metainfo, b_name.buf, name_b_tab, reset,
652 line_prefix, fraginfo);
653 if (!o->irreversible_delete)
654 print_line_count(o->file, lc_a);
655 else
656 fprintf(o->file, "?,?");
657 fprintf(o->file, " +");
658 print_line_count(o->file, lc_b);
659 fprintf(o->file, " @@%s\n", reset);
660 if (lc_a && !o->irreversible_delete)
661 emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
662 if (lc_b)
663 emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
664 if (textconv_one)
665 free((char *)data_one);
666 if (textconv_two)
667 free((char *)data_two);
670 struct diff_words_buffer {
671 mmfile_t text;
672 long alloc;
673 struct diff_words_orig {
674 const char *begin, *end;
675 } *orig;
676 int orig_nr, orig_alloc;
679 static void diff_words_append(char *line, unsigned long len,
680 struct diff_words_buffer *buffer)
682 ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
683 line++;
684 len--;
685 memcpy(buffer->text.ptr + buffer->text.size, line, len);
686 buffer->text.size += len;
687 buffer->text.ptr[buffer->text.size] = '\0';
690 struct diff_words_style_elem {
691 const char *prefix;
692 const char *suffix;
693 const char *color; /* NULL; filled in by the setup code if
694 * color is enabled */
697 struct diff_words_style {
698 enum diff_words_type type;
699 struct diff_words_style_elem new, old, ctx;
700 const char *newline;
703 static struct diff_words_style diff_words_styles[] = {
704 { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
705 { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
706 { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
709 struct diff_words_data {
710 struct diff_words_buffer minus, plus;
711 const char *current_plus;
712 int last_minus;
713 struct diff_options *opt;
714 regex_t *word_regex;
715 enum diff_words_type type;
716 struct diff_words_style *style;
719 static int fn_out_diff_words_write_helper(FILE *fp,
720 struct diff_words_style_elem *st_el,
721 const char *newline,
722 size_t count, const char *buf,
723 const char *line_prefix)
725 int print = 0;
727 while (count) {
728 char *p = memchr(buf, '\n', count);
729 if (print)
730 fputs(line_prefix, fp);
731 if (p != buf) {
732 if (st_el->color && fputs(st_el->color, fp) < 0)
733 return -1;
734 if (fputs(st_el->prefix, fp) < 0 ||
735 fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
736 fputs(st_el->suffix, fp) < 0)
737 return -1;
738 if (st_el->color && *st_el->color
739 && fputs(GIT_COLOR_RESET, fp) < 0)
740 return -1;
742 if (!p)
743 return 0;
744 if (fputs(newline, fp) < 0)
745 return -1;
746 count -= p + 1 - buf;
747 buf = p + 1;
748 print = 1;
750 return 0;
754 * '--color-words' algorithm can be described as:
756 * 1. collect a the minus/plus lines of a diff hunk, divided into
757 * minus-lines and plus-lines;
759 * 2. break both minus-lines and plus-lines into words and
760 * place them into two mmfile_t with one word for each line;
762 * 3. use xdiff to run diff on the two mmfile_t to get the words level diff;
764 * And for the common parts of the both file, we output the plus side text.
765 * diff_words->current_plus is used to trace the current position of the plus file
766 * which printed. diff_words->last_minus is used to trace the last minus word
767 * printed.
769 * For '--graph' to work with '--color-words', we need to output the graph prefix
770 * on each line of color words output. Generally, there are two conditions on
771 * which we should output the prefix.
773 * 1. diff_words->last_minus == 0 &&
774 * diff_words->current_plus == diff_words->plus.text.ptr
776 * that is: the plus text must start as a new line, and if there is no minus
777 * word printed, a graph prefix must be printed.
779 * 2. diff_words->current_plus > diff_words->plus.text.ptr &&
780 * *(diff_words->current_plus - 1) == '\n'
782 * that is: a graph prefix must be printed following a '\n'
784 static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
786 if ((diff_words->last_minus == 0 &&
787 diff_words->current_plus == diff_words->plus.text.ptr) ||
788 (diff_words->current_plus > diff_words->plus.text.ptr &&
789 *(diff_words->current_plus - 1) == '\n')) {
790 return 1;
791 } else {
792 return 0;
796 static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
798 struct diff_words_data *diff_words = priv;
799 struct diff_words_style *style = diff_words->style;
800 int minus_first, minus_len, plus_first, plus_len;
801 const char *minus_begin, *minus_end, *plus_begin, *plus_end;
802 struct diff_options *opt = diff_words->opt;
803 struct strbuf *msgbuf;
804 char *line_prefix = "";
806 if (line[0] != '@' || parse_hunk_header(line, len,
807 &minus_first, &minus_len, &plus_first, &plus_len))
808 return;
810 assert(opt);
811 if (opt->output_prefix) {
812 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
813 line_prefix = msgbuf->buf;
816 /* POSIX requires that first be decremented by one if len == 0... */
817 if (minus_len) {
818 minus_begin = diff_words->minus.orig[minus_first].begin;
819 minus_end =
820 diff_words->minus.orig[minus_first + minus_len - 1].end;
821 } else
822 minus_begin = minus_end =
823 diff_words->minus.orig[minus_first].end;
825 if (plus_len) {
826 plus_begin = diff_words->plus.orig[plus_first].begin;
827 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
828 } else
829 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
831 if (color_words_output_graph_prefix(diff_words)) {
832 fputs(line_prefix, diff_words->opt->file);
834 if (diff_words->current_plus != plus_begin) {
835 fn_out_diff_words_write_helper(diff_words->opt->file,
836 &style->ctx, style->newline,
837 plus_begin - diff_words->current_plus,
838 diff_words->current_plus, line_prefix);
839 if (*(plus_begin - 1) == '\n')
840 fputs(line_prefix, diff_words->opt->file);
842 if (minus_begin != minus_end) {
843 fn_out_diff_words_write_helper(diff_words->opt->file,
844 &style->old, style->newline,
845 minus_end - minus_begin, minus_begin,
846 line_prefix);
848 if (plus_begin != plus_end) {
849 fn_out_diff_words_write_helper(diff_words->opt->file,
850 &style->new, style->newline,
851 plus_end - plus_begin, plus_begin,
852 line_prefix);
855 diff_words->current_plus = plus_end;
856 diff_words->last_minus = minus_first;
859 /* This function starts looking at *begin, and returns 0 iff a word was found. */
860 static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
861 int *begin, int *end)
863 if (word_regex && *begin < buffer->size) {
864 regmatch_t match[1];
865 if (!regexec(word_regex, buffer->ptr + *begin, 1, match, 0)) {
866 char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
867 '\n', match[0].rm_eo - match[0].rm_so);
868 *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
869 *begin += match[0].rm_so;
870 return *begin >= *end;
872 return -1;
875 /* find the next word */
876 while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
877 (*begin)++;
878 if (*begin >= buffer->size)
879 return -1;
881 /* find the end of the word */
882 *end = *begin + 1;
883 while (*end < buffer->size && !isspace(buffer->ptr[*end]))
884 (*end)++;
886 return 0;
890 * This function splits the words in buffer->text, stores the list with
891 * newline separator into out, and saves the offsets of the original words
892 * in buffer->orig.
894 static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
895 regex_t *word_regex)
897 int i, j;
898 long alloc = 0;
900 out->size = 0;
901 out->ptr = NULL;
903 /* fake an empty "0th" word */
904 ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
905 buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
906 buffer->orig_nr = 1;
908 for (i = 0; i < buffer->text.size; i++) {
909 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
910 return;
912 /* store original boundaries */
913 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
914 buffer->orig_alloc);
915 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
916 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
917 buffer->orig_nr++;
919 /* store one word */
920 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
921 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
922 out->ptr[out->size + j - i] = '\n';
923 out->size += j - i + 1;
925 i = j - 1;
929 /* this executes the word diff on the accumulated buffers */
930 static void diff_words_show(struct diff_words_data *diff_words)
932 xpparam_t xpp;
933 xdemitconf_t xecfg;
934 mmfile_t minus, plus;
935 struct diff_words_style *style = diff_words->style;
937 struct diff_options *opt = diff_words->opt;
938 struct strbuf *msgbuf;
939 char *line_prefix = "";
941 assert(opt);
942 if (opt->output_prefix) {
943 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
944 line_prefix = msgbuf->buf;
947 /* special case: only removal */
948 if (!diff_words->plus.text.size) {
949 fputs(line_prefix, diff_words->opt->file);
950 fn_out_diff_words_write_helper(diff_words->opt->file,
951 &style->old, style->newline,
952 diff_words->minus.text.size,
953 diff_words->minus.text.ptr, line_prefix);
954 diff_words->minus.text.size = 0;
955 return;
958 diff_words->current_plus = diff_words->plus.text.ptr;
959 diff_words->last_minus = 0;
961 memset(&xpp, 0, sizeof(xpp));
962 memset(&xecfg, 0, sizeof(xecfg));
963 diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
964 diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
965 xpp.flags = 0;
966 /* as only the hunk header will be parsed, we need a 0-context */
967 xecfg.ctxlen = 0;
968 xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
969 &xpp, &xecfg);
970 free(minus.ptr);
971 free(plus.ptr);
972 if (diff_words->current_plus != diff_words->plus.text.ptr +
973 diff_words->plus.text.size) {
974 if (color_words_output_graph_prefix(diff_words))
975 fputs(line_prefix, diff_words->opt->file);
976 fn_out_diff_words_write_helper(diff_words->opt->file,
977 &style->ctx, style->newline,
978 diff_words->plus.text.ptr + diff_words->plus.text.size
979 - diff_words->current_plus, diff_words->current_plus,
980 line_prefix);
982 diff_words->minus.text.size = diff_words->plus.text.size = 0;
985 /* In "color-words" mode, show word-diff of words accumulated in the buffer */
986 static void diff_words_flush(struct emit_callback *ecbdata)
988 if (ecbdata->diff_words->minus.text.size ||
989 ecbdata->diff_words->plus.text.size)
990 diff_words_show(ecbdata->diff_words);
993 static void diff_filespec_load_driver(struct diff_filespec *one)
995 /* Use already-loaded driver */
996 if (one->driver)
997 return;
999 if (S_ISREG(one->mode))
1000 one->driver = userdiff_find_by_path(one->path);
1002 /* Fallback to default settings */
1003 if (!one->driver)
1004 one->driver = userdiff_find_by_name("default");
1007 static const char *userdiff_word_regex(struct diff_filespec *one)
1009 diff_filespec_load_driver(one);
1010 return one->driver->word_regex;
1013 static void init_diff_words_data(struct emit_callback *ecbdata,
1014 struct diff_options *orig_opts,
1015 struct diff_filespec *one,
1016 struct diff_filespec *two)
1018 int i;
1019 struct diff_options *o = xmalloc(sizeof(struct diff_options));
1020 memcpy(o, orig_opts, sizeof(struct diff_options));
1022 ecbdata->diff_words =
1023 xcalloc(1, sizeof(struct diff_words_data));
1024 ecbdata->diff_words->type = o->word_diff;
1025 ecbdata->diff_words->opt = o;
1026 if (!o->word_regex)
1027 o->word_regex = userdiff_word_regex(one);
1028 if (!o->word_regex)
1029 o->word_regex = userdiff_word_regex(two);
1030 if (!o->word_regex)
1031 o->word_regex = diff_word_regex_cfg;
1032 if (o->word_regex) {
1033 ecbdata->diff_words->word_regex = (regex_t *)
1034 xmalloc(sizeof(regex_t));
1035 if (regcomp(ecbdata->diff_words->word_regex,
1036 o->word_regex,
1037 REG_EXTENDED | REG_NEWLINE))
1038 die ("Invalid regular expression: %s",
1039 o->word_regex);
1041 for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1042 if (o->word_diff == diff_words_styles[i].type) {
1043 ecbdata->diff_words->style =
1044 &diff_words_styles[i];
1045 break;
1048 if (want_color(o->use_color)) {
1049 struct diff_words_style *st = ecbdata->diff_words->style;
1050 st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1051 st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1052 st->ctx.color = diff_get_color_opt(o, DIFF_PLAIN);
1056 static void free_diff_words_data(struct emit_callback *ecbdata)
1058 if (ecbdata->diff_words) {
1059 diff_words_flush(ecbdata);
1060 free (ecbdata->diff_words->opt);
1061 free (ecbdata->diff_words->minus.text.ptr);
1062 free (ecbdata->diff_words->minus.orig);
1063 free (ecbdata->diff_words->plus.text.ptr);
1064 free (ecbdata->diff_words->plus.orig);
1065 if (ecbdata->diff_words->word_regex) {
1066 regfree(ecbdata->diff_words->word_regex);
1067 free(ecbdata->diff_words->word_regex);
1069 free(ecbdata->diff_words);
1070 ecbdata->diff_words = NULL;
1074 const char *diff_get_color(int diff_use_color, enum color_diff ix)
1076 if (want_color(diff_use_color))
1077 return diff_colors[ix];
1078 return "";
1081 static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1083 const char *cp;
1084 unsigned long allot;
1085 size_t l = len;
1087 if (ecb->truncate)
1088 return ecb->truncate(line, len);
1089 cp = line;
1090 allot = l;
1091 while (0 < l) {
1092 (void) utf8_width(&cp, &l);
1093 if (!cp)
1094 break; /* truncated in the middle? */
1096 return allot - l;
1099 static void find_lno(const char *line, struct emit_callback *ecbdata)
1101 const char *p;
1102 ecbdata->lno_in_preimage = 0;
1103 ecbdata->lno_in_postimage = 0;
1104 p = strchr(line, '-');
1105 if (!p)
1106 return; /* cannot happen */
1107 ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1108 p = strchr(p, '+');
1109 if (!p)
1110 return; /* cannot happen */
1111 ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1114 static void fn_out_consume(void *priv, char *line, unsigned long len)
1116 struct emit_callback *ecbdata = priv;
1117 const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
1118 const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
1119 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1120 struct diff_options *o = ecbdata->opt;
1121 char *line_prefix = "";
1122 struct strbuf *msgbuf;
1124 if (o && o->output_prefix) {
1125 msgbuf = o->output_prefix(o, o->output_prefix_data);
1126 line_prefix = msgbuf->buf;
1129 if (ecbdata->header) {
1130 fprintf(ecbdata->opt->file, "%s", ecbdata->header->buf);
1131 strbuf_reset(ecbdata->header);
1132 ecbdata->header = NULL;
1134 *(ecbdata->found_changesp) = 1;
1136 if (ecbdata->label_path[0]) {
1137 const char *name_a_tab, *name_b_tab;
1139 name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
1140 name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
1142 fprintf(ecbdata->opt->file, "%s%s--- %s%s%s\n",
1143 line_prefix, meta, ecbdata->label_path[0], reset, name_a_tab);
1144 fprintf(ecbdata->opt->file, "%s%s+++ %s%s%s\n",
1145 line_prefix, meta, ecbdata->label_path[1], reset, name_b_tab);
1146 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1149 if (diff_suppress_blank_empty
1150 && len == 2 && line[0] == ' ' && line[1] == '\n') {
1151 line[0] = '\n';
1152 len = 1;
1155 if (line[0] == '@') {
1156 if (ecbdata->diff_words)
1157 diff_words_flush(ecbdata);
1158 len = sane_truncate_line(ecbdata, line, len);
1159 find_lno(line, ecbdata);
1160 emit_hunk_header(ecbdata, line, len);
1161 if (line[len-1] != '\n')
1162 putc('\n', ecbdata->opt->file);
1163 return;
1166 if (len < 1) {
1167 emit_line(ecbdata->opt, reset, reset, line, len);
1168 if (ecbdata->diff_words
1169 && ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN)
1170 fputs("~\n", ecbdata->opt->file);
1171 return;
1174 if (ecbdata->diff_words) {
1175 if (line[0] == '-') {
1176 diff_words_append(line, len,
1177 &ecbdata->diff_words->minus);
1178 return;
1179 } else if (line[0] == '+') {
1180 diff_words_append(line, len,
1181 &ecbdata->diff_words->plus);
1182 return;
1183 } else if (!prefixcmp(line, "\\ ")) {
1185 * Eat the "no newline at eof" marker as if we
1186 * saw a "+" or "-" line with nothing on it,
1187 * and return without diff_words_flush() to
1188 * defer processing. If this is the end of
1189 * preimage, more "+" lines may come after it.
1191 return;
1193 diff_words_flush(ecbdata);
1194 if (ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN) {
1195 emit_line(ecbdata->opt, plain, reset, line, len);
1196 fputs("~\n", ecbdata->opt->file);
1197 } else {
1199 * Skip the prefix character, if any. With
1200 * diff_suppress_blank_empty, there may be
1201 * none.
1203 if (line[0] != '\n') {
1204 line++;
1205 len--;
1207 emit_line(ecbdata->opt, plain, reset, line, len);
1209 return;
1212 if (line[0] != '+') {
1213 const char *color =
1214 diff_get_color(ecbdata->color_diff,
1215 line[0] == '-' ? DIFF_FILE_OLD : DIFF_PLAIN);
1216 ecbdata->lno_in_preimage++;
1217 if (line[0] == ' ')
1218 ecbdata->lno_in_postimage++;
1219 emit_line(ecbdata->opt, color, reset, line, len);
1220 } else {
1221 ecbdata->lno_in_postimage++;
1222 emit_add_line(reset, ecbdata, line + 1, len - 1);
1226 static char *pprint_rename(const char *a, const char *b)
1228 const char *old = a;
1229 const char *new = b;
1230 struct strbuf name = STRBUF_INIT;
1231 int pfx_length, sfx_length;
1232 int len_a = strlen(a);
1233 int len_b = strlen(b);
1234 int a_midlen, b_midlen;
1235 int qlen_a = quote_c_style(a, NULL, NULL, 0);
1236 int qlen_b = quote_c_style(b, NULL, NULL, 0);
1238 if (qlen_a || qlen_b) {
1239 quote_c_style(a, &name, NULL, 0);
1240 strbuf_addstr(&name, " => ");
1241 quote_c_style(b, &name, NULL, 0);
1242 return strbuf_detach(&name, NULL);
1245 /* Find common prefix */
1246 pfx_length = 0;
1247 while (*old && *new && *old == *new) {
1248 if (*old == '/')
1249 pfx_length = old - a + 1;
1250 old++;
1251 new++;
1254 /* Find common suffix */
1255 old = a + len_a;
1256 new = b + len_b;
1257 sfx_length = 0;
1258 while (a <= old && b <= new && *old == *new) {
1259 if (*old == '/')
1260 sfx_length = len_a - (old - a);
1261 old--;
1262 new--;
1266 * pfx{mid-a => mid-b}sfx
1267 * {pfx-a => pfx-b}sfx
1268 * pfx{sfx-a => sfx-b}
1269 * name-a => name-b
1271 a_midlen = len_a - pfx_length - sfx_length;
1272 b_midlen = len_b - pfx_length - sfx_length;
1273 if (a_midlen < 0)
1274 a_midlen = 0;
1275 if (b_midlen < 0)
1276 b_midlen = 0;
1278 strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1279 if (pfx_length + sfx_length) {
1280 strbuf_add(&name, a, pfx_length);
1281 strbuf_addch(&name, '{');
1283 strbuf_add(&name, a + pfx_length, a_midlen);
1284 strbuf_addstr(&name, " => ");
1285 strbuf_add(&name, b + pfx_length, b_midlen);
1286 if (pfx_length + sfx_length) {
1287 strbuf_addch(&name, '}');
1288 strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1290 return strbuf_detach(&name, NULL);
1293 struct diffstat_t {
1294 int nr;
1295 int alloc;
1296 struct diffstat_file {
1297 char *from_name;
1298 char *name;
1299 char *print_name;
1300 unsigned is_unmerged:1;
1301 unsigned is_binary:1;
1302 unsigned is_renamed:1;
1303 uintmax_t added, deleted;
1304 } **files;
1307 static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1308 const char *name_a,
1309 const char *name_b)
1311 struct diffstat_file *x;
1312 x = xcalloc(sizeof (*x), 1);
1313 if (diffstat->nr == diffstat->alloc) {
1314 diffstat->alloc = alloc_nr(diffstat->alloc);
1315 diffstat->files = xrealloc(diffstat->files,
1316 diffstat->alloc * sizeof(x));
1318 diffstat->files[diffstat->nr++] = x;
1319 if (name_b) {
1320 x->from_name = xstrdup(name_a);
1321 x->name = xstrdup(name_b);
1322 x->is_renamed = 1;
1324 else {
1325 x->from_name = NULL;
1326 x->name = xstrdup(name_a);
1328 return x;
1331 static void diffstat_consume(void *priv, char *line, unsigned long len)
1333 struct diffstat_t *diffstat = priv;
1334 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1336 if (line[0] == '+')
1337 x->added++;
1338 else if (line[0] == '-')
1339 x->deleted++;
1342 const char mime_boundary_leader[] = "------------";
1344 static int scale_linear(int it, int width, int max_change)
1346 if (!it)
1347 return 0;
1349 * make sure that at least one '-' or '+' is printed if
1350 * there is any change to this path. The easiest way is to
1351 * scale linearly as if the alloted width is one column shorter
1352 * than it is, and then add 1 to the result.
1354 return 1 + (it * (width - 1) / max_change);
1357 static void show_name(FILE *file,
1358 const char *prefix, const char *name, int len)
1360 fprintf(file, " %s%-*s |", prefix, len, name);
1363 static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1365 if (cnt <= 0)
1366 return;
1367 fprintf(file, "%s", set);
1368 while (cnt--)
1369 putc(ch, file);
1370 fprintf(file, "%s", reset);
1373 static void fill_print_name(struct diffstat_file *file)
1375 char *pname;
1377 if (file->print_name)
1378 return;
1380 if (!file->is_renamed) {
1381 struct strbuf buf = STRBUF_INIT;
1382 if (quote_c_style(file->name, &buf, NULL, 0)) {
1383 pname = strbuf_detach(&buf, NULL);
1384 } else {
1385 pname = file->name;
1386 strbuf_release(&buf);
1388 } else {
1389 pname = pprint_rename(file->from_name, file->name);
1391 file->print_name = pname;
1394 int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1396 struct strbuf sb = STRBUF_INIT;
1397 int ret;
1399 if (!files) {
1400 assert(insertions == 0 && deletions == 0);
1401 return fprintf(fp, "%s\n", _(" 0 files changed"));
1404 strbuf_addf(&sb,
1405 Q_(" %d file changed", " %d files changed", files),
1406 files);
1409 * For binary diff, the caller may want to print "x files
1410 * changed" with insertions == 0 && deletions == 0.
1412 * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1413 * is probably less confusing (i.e skip over "2 files changed
1414 * but nothing about added/removed lines? Is this a bug in Git?").
1416 if (insertions || deletions == 0) {
1418 * TRANSLATORS: "+" in (+) is a line addition marker;
1419 * do not translate it.
1421 strbuf_addf(&sb,
1422 Q_(", %d insertion(+)", ", %d insertions(+)",
1423 insertions),
1424 insertions);
1427 if (deletions || insertions == 0) {
1429 * TRANSLATORS: "-" in (-) is a line removal marker;
1430 * do not translate it.
1432 strbuf_addf(&sb,
1433 Q_(", %d deletion(-)", ", %d deletions(-)",
1434 deletions),
1435 deletions);
1437 strbuf_addch(&sb, '\n');
1438 ret = fputs(sb.buf, fp);
1439 strbuf_release(&sb);
1440 return ret;
1443 static void show_stats(struct diffstat_t *data, struct diff_options *options)
1445 int i, len, add, del, adds = 0, dels = 0;
1446 uintmax_t max_change = 0, max_len = 0;
1447 int total_files = data->nr, count;
1448 int width, name_width, graph_width, number_width = 0, bin_width = 0;
1449 const char *reset, *add_c, *del_c;
1450 const char *line_prefix = "";
1451 int extra_shown = 0;
1452 struct strbuf *msg = NULL;
1454 if (data->nr == 0)
1455 return;
1457 if (options->output_prefix) {
1458 msg = options->output_prefix(options, options->output_prefix_data);
1459 line_prefix = msg->buf;
1462 count = options->stat_count ? options->stat_count : data->nr;
1464 reset = diff_get_color_opt(options, DIFF_RESET);
1465 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1466 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1469 * Find the longest filename and max number of changes
1471 for (i = 0; (i < count) && (i < data->nr); i++) {
1472 struct diffstat_file *file = data->files[i];
1473 uintmax_t change = file->added + file->deleted;
1474 if (!data->files[i]->is_renamed &&
1475 (change == 0)) {
1476 count++; /* not shown == room for one more */
1477 continue;
1479 fill_print_name(file);
1480 len = strlen(file->print_name);
1481 if (max_len < len)
1482 max_len = len;
1484 if (file->is_unmerged) {
1485 /* "Unmerged" is 8 characters */
1486 bin_width = bin_width < 8 ? 8 : bin_width;
1487 continue;
1489 if (file->is_binary) {
1490 /* "Bin XXX -> YYY bytes" */
1491 int w = 14 + decimal_width(file->added)
1492 + decimal_width(file->deleted);
1493 bin_width = bin_width < w ? w : bin_width;
1494 /* Display change counts aligned with "Bin" */
1495 number_width = 3;
1496 continue;
1499 if (max_change < change)
1500 max_change = change;
1502 count = i; /* min(count, data->nr) */
1505 * We have width = stat_width or term_columns() columns total.
1506 * We want a maximum of min(max_len, stat_name_width) for the name part.
1507 * We want a maximum of min(max_change, stat_graph_width) for the +- part.
1508 * We also need 1 for " " and 4 + decimal_width(max_change)
1509 * for " | NNNN " and one the empty column at the end, altogether
1510 * 6 + decimal_width(max_change).
1512 * If there's not enough space, we will use the smaller of
1513 * stat_name_width (if set) and 5/8*width for the filename,
1514 * and the rest for constant elements + graph part, but no more
1515 * than stat_graph_width for the graph part.
1516 * (5/8 gives 50 for filename and 30 for the constant parts + graph
1517 * for the standard terminal size).
1519 * In other words: stat_width limits the maximum width, and
1520 * stat_name_width fixes the maximum width of the filename,
1521 * and is also used to divide available columns if there
1522 * aren't enough.
1524 * Binary files are displayed with "Bin XXX -> YYY bytes"
1525 * instead of the change count and graph. This part is treated
1526 * similarly to the graph part, except that it is not
1527 * "scaled". If total width is too small to accomodate the
1528 * guaranteed minimum width of the filename part and the
1529 * separators and this message, this message will "overflow"
1530 * making the line longer than the maximum width.
1533 if (options->stat_width == -1)
1534 width = term_columns() - options->output_prefix_length;
1535 else
1536 width = options->stat_width ? options->stat_width : 80;
1537 number_width = decimal_width(max_change) > number_width ?
1538 decimal_width(max_change) : number_width;
1540 if (options->stat_graph_width == -1)
1541 options->stat_graph_width = diff_stat_graph_width;
1544 * Guarantee 3/8*16==6 for the graph part
1545 * and 5/8*16==10 for the filename part
1547 if (width < 16 + 6 + number_width)
1548 width = 16 + 6 + number_width;
1551 * First assign sizes that are wanted, ignoring available width.
1552 * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
1553 * starting from "XXX" should fit in graph_width.
1555 graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
1556 if (options->stat_graph_width &&
1557 options->stat_graph_width < graph_width)
1558 graph_width = options->stat_graph_width;
1560 name_width = (options->stat_name_width > 0 &&
1561 options->stat_name_width < max_len) ?
1562 options->stat_name_width : max_len;
1565 * Adjust adjustable widths not to exceed maximum width
1567 if (name_width + number_width + 6 + graph_width > width) {
1568 if (graph_width > width * 3/8 - number_width - 6) {
1569 graph_width = width * 3/8 - number_width - 6;
1570 if (graph_width < 6)
1571 graph_width = 6;
1574 if (options->stat_graph_width &&
1575 graph_width > options->stat_graph_width)
1576 graph_width = options->stat_graph_width;
1577 if (name_width > width - number_width - 6 - graph_width)
1578 name_width = width - number_width - 6 - graph_width;
1579 else
1580 graph_width = width - number_width - 6 - name_width;
1584 * From here name_width is the width of the name area,
1585 * and graph_width is the width of the graph area.
1586 * max_change is used to scale graph properly.
1588 for (i = 0; i < count; i++) {
1589 const char *prefix = "";
1590 char *name = data->files[i]->print_name;
1591 uintmax_t added = data->files[i]->added;
1592 uintmax_t deleted = data->files[i]->deleted;
1593 int name_len;
1595 if (!data->files[i]->is_renamed &&
1596 (added + deleted == 0)) {
1597 total_files--;
1598 continue;
1601 * "scale" the filename
1603 len = name_width;
1604 name_len = strlen(name);
1605 if (name_width < name_len) {
1606 char *slash;
1607 prefix = "...";
1608 len -= 3;
1609 name += name_len - len;
1610 slash = strchr(name, '/');
1611 if (slash)
1612 name = slash;
1615 if (data->files[i]->is_binary) {
1616 fprintf(options->file, "%s", line_prefix);
1617 show_name(options->file, prefix, name, len);
1618 fprintf(options->file, " %*s", number_width, "Bin");
1619 if (!added && !deleted) {
1620 putc('\n', options->file);
1621 continue;
1623 fprintf(options->file, " %s%"PRIuMAX"%s",
1624 del_c, deleted, reset);
1625 fprintf(options->file, " -> ");
1626 fprintf(options->file, "%s%"PRIuMAX"%s",
1627 add_c, added, reset);
1628 fprintf(options->file, " bytes");
1629 fprintf(options->file, "\n");
1630 continue;
1632 else if (data->files[i]->is_unmerged) {
1633 fprintf(options->file, "%s", line_prefix);
1634 show_name(options->file, prefix, name, len);
1635 fprintf(options->file, " Unmerged\n");
1636 continue;
1640 * scale the add/delete
1642 add = added;
1643 del = deleted;
1644 adds += add;
1645 dels += del;
1647 if (graph_width <= max_change) {
1648 int total = add + del;
1650 total = scale_linear(add + del, graph_width, max_change);
1651 if (total < 2 && add && del)
1652 /* width >= 2 due to the sanity check */
1653 total = 2;
1654 if (add < del) {
1655 add = scale_linear(add, graph_width, max_change);
1656 del = total - add;
1657 } else {
1658 del = scale_linear(del, graph_width, max_change);
1659 add = total - del;
1662 fprintf(options->file, "%s", line_prefix);
1663 show_name(options->file, prefix, name, len);
1664 fprintf(options->file, " %*"PRIuMAX"%s",
1665 number_width, added + deleted,
1666 added + deleted ? " " : "");
1667 show_graph(options->file, '+', add, add_c, reset);
1668 show_graph(options->file, '-', del, del_c, reset);
1669 fprintf(options->file, "\n");
1671 for (i = count; i < data->nr; i++) {
1672 uintmax_t added = data->files[i]->added;
1673 uintmax_t deleted = data->files[i]->deleted;
1674 if (!data->files[i]->is_renamed &&
1675 (added + deleted == 0)) {
1676 total_files--;
1677 continue;
1679 adds += added;
1680 dels += deleted;
1681 if (!extra_shown)
1682 fprintf(options->file, "%s ...\n", line_prefix);
1683 extra_shown = 1;
1685 fprintf(options->file, "%s", line_prefix);
1686 print_stat_summary(options->file, total_files, adds, dels);
1689 static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1691 int i, adds = 0, dels = 0, total_files = data->nr;
1693 if (data->nr == 0)
1694 return;
1696 for (i = 0; i < data->nr; i++) {
1697 int added = data->files[i]->added;
1698 int deleted= data->files[i]->deleted;
1700 if (data->files[i]->is_unmerged)
1701 continue;
1702 if (!data->files[i]->is_renamed && (added + deleted == 0)) {
1703 total_files--;
1704 } else if (!data->files[i]->is_binary) { /* don't count bytes */
1705 adds += added;
1706 dels += deleted;
1709 if (options->output_prefix) {
1710 struct strbuf *msg = NULL;
1711 msg = options->output_prefix(options,
1712 options->output_prefix_data);
1713 fprintf(options->file, "%s", msg->buf);
1715 print_stat_summary(options->file, total_files, adds, dels);
1718 static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1720 int i;
1722 if (data->nr == 0)
1723 return;
1725 for (i = 0; i < data->nr; i++) {
1726 struct diffstat_file *file = data->files[i];
1728 if (options->output_prefix) {
1729 struct strbuf *msg = NULL;
1730 msg = options->output_prefix(options,
1731 options->output_prefix_data);
1732 fprintf(options->file, "%s", msg->buf);
1735 if (file->is_binary)
1736 fprintf(options->file, "-\t-\t");
1737 else
1738 fprintf(options->file,
1739 "%"PRIuMAX"\t%"PRIuMAX"\t",
1740 file->added, file->deleted);
1741 if (options->line_termination) {
1742 fill_print_name(file);
1743 if (!file->is_renamed)
1744 write_name_quoted(file->name, options->file,
1745 options->line_termination);
1746 else {
1747 fputs(file->print_name, options->file);
1748 putc(options->line_termination, options->file);
1750 } else {
1751 if (file->is_renamed) {
1752 putc('\0', options->file);
1753 write_name_quoted(file->from_name, options->file, '\0');
1755 write_name_quoted(file->name, options->file, '\0');
1760 struct dirstat_file {
1761 const char *name;
1762 unsigned long changed;
1765 struct dirstat_dir {
1766 struct dirstat_file *files;
1767 int alloc, nr, permille, cumulative;
1770 static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1771 unsigned long changed, const char *base, int baselen)
1773 unsigned long this_dir = 0;
1774 unsigned int sources = 0;
1775 const char *line_prefix = "";
1776 struct strbuf *msg = NULL;
1778 if (opt->output_prefix) {
1779 msg = opt->output_prefix(opt, opt->output_prefix_data);
1780 line_prefix = msg->buf;
1783 while (dir->nr) {
1784 struct dirstat_file *f = dir->files;
1785 int namelen = strlen(f->name);
1786 unsigned long this;
1787 char *slash;
1789 if (namelen < baselen)
1790 break;
1791 if (memcmp(f->name, base, baselen))
1792 break;
1793 slash = strchr(f->name + baselen, '/');
1794 if (slash) {
1795 int newbaselen = slash + 1 - f->name;
1796 this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
1797 sources++;
1798 } else {
1799 this = f->changed;
1800 dir->files++;
1801 dir->nr--;
1802 sources += 2;
1804 this_dir += this;
1808 * We don't report dirstat's for
1809 * - the top level
1810 * - or cases where everything came from a single directory
1811 * under this directory (sources == 1).
1813 if (baselen && sources != 1) {
1814 if (this_dir) {
1815 int permille = this_dir * 1000 / changed;
1816 if (permille >= dir->permille) {
1817 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
1818 permille / 10, permille % 10, baselen, base);
1819 if (!dir->cumulative)
1820 return 0;
1824 return this_dir;
1827 static int dirstat_compare(const void *_a, const void *_b)
1829 const struct dirstat_file *a = _a;
1830 const struct dirstat_file *b = _b;
1831 return strcmp(a->name, b->name);
1834 static void show_dirstat(struct diff_options *options)
1836 int i;
1837 unsigned long changed;
1838 struct dirstat_dir dir;
1839 struct diff_queue_struct *q = &diff_queued_diff;
1841 dir.files = NULL;
1842 dir.alloc = 0;
1843 dir.nr = 0;
1844 dir.permille = options->dirstat_permille;
1845 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1847 changed = 0;
1848 for (i = 0; i < q->nr; i++) {
1849 struct diff_filepair *p = q->queue[i];
1850 const char *name;
1851 unsigned long copied, added, damage;
1852 int content_changed;
1854 name = p->two->path ? p->two->path : p->one->path;
1856 if (p->one->sha1_valid && p->two->sha1_valid)
1857 content_changed = hashcmp(p->one->sha1, p->two->sha1);
1858 else
1859 content_changed = 1;
1861 if (!content_changed) {
1863 * The SHA1 has not changed, so pre-/post-content is
1864 * identical. We can therefore skip looking at the
1865 * file contents altogether.
1867 damage = 0;
1868 goto found_damage;
1871 if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
1873 * In --dirstat-by-file mode, we don't really need to
1874 * look at the actual file contents at all.
1875 * The fact that the SHA1 changed is enough for us to
1876 * add this file to the list of results
1877 * (with each file contributing equal damage).
1879 damage = 1;
1880 goto found_damage;
1883 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
1884 diff_populate_filespec(p->one, 0);
1885 diff_populate_filespec(p->two, 0);
1886 diffcore_count_changes(p->one, p->two, NULL, NULL, 0,
1887 &copied, &added);
1888 diff_free_filespec_data(p->one);
1889 diff_free_filespec_data(p->two);
1890 } else if (DIFF_FILE_VALID(p->one)) {
1891 diff_populate_filespec(p->one, 1);
1892 copied = added = 0;
1893 diff_free_filespec_data(p->one);
1894 } else if (DIFF_FILE_VALID(p->two)) {
1895 diff_populate_filespec(p->two, 1);
1896 copied = 0;
1897 added = p->two->size;
1898 diff_free_filespec_data(p->two);
1899 } else
1900 continue;
1903 * Original minus copied is the removed material,
1904 * added is the new material. They are both damages
1905 * made to the preimage.
1906 * If the resulting damage is zero, we know that
1907 * diffcore_count_changes() considers the two entries to
1908 * be identical, but since content_changed is true, we
1909 * know that there must have been _some_ kind of change,
1910 * so we force all entries to have damage > 0.
1912 damage = (p->one->size - copied) + added;
1913 if (!damage)
1914 damage = 1;
1916 found_damage:
1917 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1918 dir.files[dir.nr].name = name;
1919 dir.files[dir.nr].changed = damage;
1920 changed += damage;
1921 dir.nr++;
1924 /* This can happen even with many files, if everything was renames */
1925 if (!changed)
1926 return;
1928 /* Show all directories with more than x% of the changes */
1929 qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1930 gather_dirstat(options, &dir, changed, "", 0);
1933 static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
1935 int i;
1936 unsigned long changed;
1937 struct dirstat_dir dir;
1939 if (data->nr == 0)
1940 return;
1942 dir.files = NULL;
1943 dir.alloc = 0;
1944 dir.nr = 0;
1945 dir.permille = options->dirstat_permille;
1946 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1948 changed = 0;
1949 for (i = 0; i < data->nr; i++) {
1950 struct diffstat_file *file = data->files[i];
1951 unsigned long damage = file->added + file->deleted;
1952 if (file->is_binary)
1954 * binary files counts bytes, not lines. Must find some
1955 * way to normalize binary bytes vs. textual lines.
1956 * The following heuristic assumes that there are 64
1957 * bytes per "line".
1958 * This is stupid and ugly, but very cheap...
1960 damage = (damage + 63) / 64;
1961 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1962 dir.files[dir.nr].name = file->name;
1963 dir.files[dir.nr].changed = damage;
1964 changed += damage;
1965 dir.nr++;
1968 /* This can happen even with many files, if everything was renames */
1969 if (!changed)
1970 return;
1972 /* Show all directories with more than x% of the changes */
1973 qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1974 gather_dirstat(options, &dir, changed, "", 0);
1977 static void free_diffstat_info(struct diffstat_t *diffstat)
1979 int i;
1980 for (i = 0; i < diffstat->nr; i++) {
1981 struct diffstat_file *f = diffstat->files[i];
1982 if (f->name != f->print_name)
1983 free(f->print_name);
1984 free(f->name);
1985 free(f->from_name);
1986 free(f);
1988 free(diffstat->files);
1991 struct checkdiff_t {
1992 const char *filename;
1993 int lineno;
1994 int conflict_marker_size;
1995 struct diff_options *o;
1996 unsigned ws_rule;
1997 unsigned status;
2000 static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2002 char firstchar;
2003 int cnt;
2005 if (len < marker_size + 1)
2006 return 0;
2007 firstchar = line[0];
2008 switch (firstchar) {
2009 case '=': case '>': case '<': case '|':
2010 break;
2011 default:
2012 return 0;
2014 for (cnt = 1; cnt < marker_size; cnt++)
2015 if (line[cnt] != firstchar)
2016 return 0;
2017 /* line[1] thru line[marker_size-1] are same as firstchar */
2018 if (len < marker_size + 1 || !isspace(line[marker_size]))
2019 return 0;
2020 return 1;
2023 static void checkdiff_consume(void *priv, char *line, unsigned long len)
2025 struct checkdiff_t *data = priv;
2026 int marker_size = data->conflict_marker_size;
2027 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2028 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2029 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2030 char *err;
2031 char *line_prefix = "";
2032 struct strbuf *msgbuf;
2034 assert(data->o);
2035 if (data->o->output_prefix) {
2036 msgbuf = data->o->output_prefix(data->o,
2037 data->o->output_prefix_data);
2038 line_prefix = msgbuf->buf;
2041 if (line[0] == '+') {
2042 unsigned bad;
2043 data->lineno++;
2044 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2045 data->status |= 1;
2046 fprintf(data->o->file,
2047 "%s%s:%d: leftover conflict marker\n",
2048 line_prefix, data->filename, data->lineno);
2050 bad = ws_check(line + 1, len - 1, data->ws_rule);
2051 if (!bad)
2052 return;
2053 data->status |= bad;
2054 err = whitespace_error_string(bad);
2055 fprintf(data->o->file, "%s%s:%d: %s.\n",
2056 line_prefix, data->filename, data->lineno, err);
2057 free(err);
2058 emit_line(data->o, set, reset, line, 1);
2059 ws_check_emit(line + 1, len - 1, data->ws_rule,
2060 data->o->file, set, reset, ws);
2061 } else if (line[0] == ' ') {
2062 data->lineno++;
2063 } else if (line[0] == '@') {
2064 char *plus = strchr(line, '+');
2065 if (plus)
2066 data->lineno = strtol(plus, NULL, 10) - 1;
2067 else
2068 die("invalid diff");
2072 static unsigned char *deflate_it(char *data,
2073 unsigned long size,
2074 unsigned long *result_size)
2076 int bound;
2077 unsigned char *deflated;
2078 git_zstream stream;
2080 memset(&stream, 0, sizeof(stream));
2081 git_deflate_init(&stream, zlib_compression_level);
2082 bound = git_deflate_bound(&stream, size);
2083 deflated = xmalloc(bound);
2084 stream.next_out = deflated;
2085 stream.avail_out = bound;
2087 stream.next_in = (unsigned char *)data;
2088 stream.avail_in = size;
2089 while (git_deflate(&stream, Z_FINISH) == Z_OK)
2090 ; /* nothing */
2091 git_deflate_end(&stream);
2092 *result_size = stream.total_out;
2093 return deflated;
2096 static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two, char *prefix)
2098 void *cp;
2099 void *delta;
2100 void *deflated;
2101 void *data;
2102 unsigned long orig_size;
2103 unsigned long delta_size;
2104 unsigned long deflate_size;
2105 unsigned long data_size;
2107 /* We could do deflated delta, or we could do just deflated two,
2108 * whichever is smaller.
2110 delta = NULL;
2111 deflated = deflate_it(two->ptr, two->size, &deflate_size);
2112 if (one->size && two->size) {
2113 delta = diff_delta(one->ptr, one->size,
2114 two->ptr, two->size,
2115 &delta_size, deflate_size);
2116 if (delta) {
2117 void *to_free = delta;
2118 orig_size = delta_size;
2119 delta = deflate_it(delta, delta_size, &delta_size);
2120 free(to_free);
2124 if (delta && delta_size < deflate_size) {
2125 fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2126 free(deflated);
2127 data = delta;
2128 data_size = delta_size;
2130 else {
2131 fprintf(file, "%sliteral %lu\n", prefix, two->size);
2132 free(delta);
2133 data = deflated;
2134 data_size = deflate_size;
2137 /* emit data encoded in base85 */
2138 cp = data;
2139 while (data_size) {
2140 int bytes = (52 < data_size) ? 52 : data_size;
2141 char line[70];
2142 data_size -= bytes;
2143 if (bytes <= 26)
2144 line[0] = bytes + 'A' - 1;
2145 else
2146 line[0] = bytes - 26 + 'a' - 1;
2147 encode_85(line + 1, cp, bytes);
2148 cp = (char *) cp + bytes;
2149 fprintf(file, "%s", prefix);
2150 fputs(line, file);
2151 fputc('\n', file);
2153 fprintf(file, "%s\n", prefix);
2154 free(data);
2157 static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two, char *prefix)
2159 fprintf(file, "%sGIT binary patch\n", prefix);
2160 emit_binary_diff_body(file, one, two, prefix);
2161 emit_binary_diff_body(file, two, one, prefix);
2164 int diff_filespec_is_binary(struct diff_filespec *one)
2166 if (one->is_binary == -1) {
2167 diff_filespec_load_driver(one);
2168 if (one->driver->binary != -1)
2169 one->is_binary = one->driver->binary;
2170 else {
2171 if (!one->data && DIFF_FILE_VALID(one))
2172 diff_populate_filespec(one, 0);
2173 if (one->data)
2174 one->is_binary = buffer_is_binary(one->data,
2175 one->size);
2176 if (one->is_binary == -1)
2177 one->is_binary = 0;
2180 return one->is_binary;
2183 static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2185 diff_filespec_load_driver(one);
2186 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2189 void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2191 if (!options->a_prefix)
2192 options->a_prefix = a;
2193 if (!options->b_prefix)
2194 options->b_prefix = b;
2197 struct userdiff_driver *get_textconv(struct diff_filespec *one)
2199 if (!DIFF_FILE_VALID(one))
2200 return NULL;
2202 diff_filespec_load_driver(one);
2203 return userdiff_get_textconv(one->driver);
2206 static void builtin_diff(const char *name_a,
2207 const char *name_b,
2208 struct diff_filespec *one,
2209 struct diff_filespec *two,
2210 const char *xfrm_msg,
2211 int must_show_header,
2212 struct diff_options *o,
2213 int complete_rewrite)
2215 mmfile_t mf1, mf2;
2216 const char *lbl[2];
2217 char *a_one, *b_two;
2218 const char *set = diff_get_color_opt(o, DIFF_METAINFO);
2219 const char *reset = diff_get_color_opt(o, DIFF_RESET);
2220 const char *a_prefix, *b_prefix;
2221 struct userdiff_driver *textconv_one = NULL;
2222 struct userdiff_driver *textconv_two = NULL;
2223 struct strbuf header = STRBUF_INIT;
2224 struct strbuf *msgbuf;
2225 char *line_prefix = "";
2227 if (o->output_prefix) {
2228 msgbuf = o->output_prefix(o, o->output_prefix_data);
2229 line_prefix = msgbuf->buf;
2232 if (DIFF_OPT_TST(o, SUBMODULE_LOG) &&
2233 (!one->mode || S_ISGITLINK(one->mode)) &&
2234 (!two->mode || S_ISGITLINK(two->mode))) {
2235 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2236 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2237 show_submodule_summary(o->file, one ? one->path : two->path,
2238 one->sha1, two->sha1, two->dirty_submodule,
2239 del, add, reset);
2240 return;
2243 if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2244 textconv_one = get_textconv(one);
2245 textconv_two = get_textconv(two);
2248 diff_set_mnemonic_prefix(o, "a/", "b/");
2249 if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2250 a_prefix = o->b_prefix;
2251 b_prefix = o->a_prefix;
2252 } else {
2253 a_prefix = o->a_prefix;
2254 b_prefix = o->b_prefix;
2257 /* Never use a non-valid filename anywhere if at all possible */
2258 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2259 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2261 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2262 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2263 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2264 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2265 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, set, a_one, b_two, reset);
2266 if (lbl[0][0] == '/') {
2267 /* /dev/null */
2268 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, set, two->mode, reset);
2269 if (xfrm_msg)
2270 strbuf_addstr(&header, xfrm_msg);
2271 must_show_header = 1;
2273 else if (lbl[1][0] == '/') {
2274 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, set, one->mode, reset);
2275 if (xfrm_msg)
2276 strbuf_addstr(&header, xfrm_msg);
2277 must_show_header = 1;
2279 else {
2280 if (one->mode != two->mode) {
2281 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, set, one->mode, reset);
2282 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, set, two->mode, reset);
2283 must_show_header = 1;
2285 if (xfrm_msg)
2286 strbuf_addstr(&header, xfrm_msg);
2289 * we do not run diff between different kind
2290 * of objects.
2292 if ((one->mode ^ two->mode) & S_IFMT)
2293 goto free_ab_and_return;
2294 if (complete_rewrite &&
2295 (textconv_one || !diff_filespec_is_binary(one)) &&
2296 (textconv_two || !diff_filespec_is_binary(two))) {
2297 fprintf(o->file, "%s", header.buf);
2298 strbuf_reset(&header);
2299 emit_rewrite_diff(name_a, name_b, one, two,
2300 textconv_one, textconv_two, o);
2301 o->found_changes = 1;
2302 goto free_ab_and_return;
2306 if (o->irreversible_delete && lbl[1][0] == '/') {
2307 fprintf(o->file, "%s", header.buf);
2308 strbuf_reset(&header);
2309 goto free_ab_and_return;
2310 } else if (!DIFF_OPT_TST(o, TEXT) &&
2311 ( (!textconv_one && diff_filespec_is_binary(one)) ||
2312 (!textconv_two && diff_filespec_is_binary(two)) )) {
2313 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2314 die("unable to read files to diff");
2315 /* Quite common confusing case */
2316 if (mf1.size == mf2.size &&
2317 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2318 if (must_show_header)
2319 fprintf(o->file, "%s", header.buf);
2320 goto free_ab_and_return;
2322 fprintf(o->file, "%s", header.buf);
2323 strbuf_reset(&header);
2324 if (DIFF_OPT_TST(o, BINARY))
2325 emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2326 else
2327 fprintf(o->file, "%sBinary files %s and %s differ\n",
2328 line_prefix, lbl[0], lbl[1]);
2329 o->found_changes = 1;
2330 } else {
2331 /* Crazy xdl interfaces.. */
2332 const char *diffopts = getenv("GIT_DIFF_OPTS");
2333 xpparam_t xpp;
2334 xdemitconf_t xecfg;
2335 struct emit_callback ecbdata;
2336 const struct userdiff_funcname *pe;
2338 if (must_show_header) {
2339 fprintf(o->file, "%s", header.buf);
2340 strbuf_reset(&header);
2343 mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2344 mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2346 pe = diff_funcname_pattern(one);
2347 if (!pe)
2348 pe = diff_funcname_pattern(two);
2350 memset(&xpp, 0, sizeof(xpp));
2351 memset(&xecfg, 0, sizeof(xecfg));
2352 memset(&ecbdata, 0, sizeof(ecbdata));
2353 ecbdata.label_path = lbl;
2354 ecbdata.color_diff = want_color(o->use_color);
2355 ecbdata.found_changesp = &o->found_changes;
2356 ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
2357 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2358 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2359 ecbdata.opt = o;
2360 ecbdata.header = header.len ? &header : NULL;
2361 xpp.flags = o->xdl_opts;
2362 xecfg.ctxlen = o->context;
2363 xecfg.interhunkctxlen = o->interhunkcontext;
2364 xecfg.flags = XDL_EMIT_FUNCNAMES;
2365 if (DIFF_OPT_TST(o, FUNCCONTEXT))
2366 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2367 if (pe)
2368 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2369 if (!diffopts)
2371 else if (!prefixcmp(diffopts, "--unified="))
2372 xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
2373 else if (!prefixcmp(diffopts, "-u"))
2374 xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
2375 if (o->word_diff)
2376 init_diff_words_data(&ecbdata, o, one, two);
2377 xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2378 &xpp, &xecfg);
2379 if (o->word_diff)
2380 free_diff_words_data(&ecbdata);
2381 if (textconv_one)
2382 free(mf1.ptr);
2383 if (textconv_two)
2384 free(mf2.ptr);
2385 xdiff_clear_find_func(&xecfg);
2388 free_ab_and_return:
2389 strbuf_release(&header);
2390 diff_free_filespec_data(one);
2391 diff_free_filespec_data(two);
2392 free(a_one);
2393 free(b_two);
2394 return;
2397 static void builtin_diffstat(const char *name_a, const char *name_b,
2398 struct diff_filespec *one,
2399 struct diff_filespec *two,
2400 struct diffstat_t *diffstat,
2401 struct diff_options *o,
2402 int complete_rewrite)
2404 mmfile_t mf1, mf2;
2405 struct diffstat_file *data;
2406 int same_contents;
2408 data = diffstat_add(diffstat, name_a, name_b);
2410 if (!one || !two) {
2411 data->is_unmerged = 1;
2412 return;
2415 same_contents = !hashcmp(one->sha1, two->sha1);
2417 if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2418 data->is_binary = 1;
2419 if (same_contents) {
2420 data->added = 0;
2421 data->deleted = 0;
2422 } else {
2423 data->added = diff_filespec_size(two);
2424 data->deleted = diff_filespec_size(one);
2428 else if (complete_rewrite) {
2429 diff_populate_filespec(one, 0);
2430 diff_populate_filespec(two, 0);
2431 data->deleted = count_lines(one->data, one->size);
2432 data->added = count_lines(two->data, two->size);
2435 else if (!same_contents) {
2436 /* Crazy xdl interfaces.. */
2437 xpparam_t xpp;
2438 xdemitconf_t xecfg;
2440 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2441 die("unable to read files to diff");
2443 memset(&xpp, 0, sizeof(xpp));
2444 memset(&xecfg, 0, sizeof(xecfg));
2445 xpp.flags = o->xdl_opts;
2446 xecfg.ctxlen = o->context;
2447 xecfg.interhunkctxlen = o->interhunkcontext;
2448 xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2449 &xpp, &xecfg);
2452 diff_free_filespec_data(one);
2453 diff_free_filespec_data(two);
2456 static void builtin_checkdiff(const char *name_a, const char *name_b,
2457 const char *attr_path,
2458 struct diff_filespec *one,
2459 struct diff_filespec *two,
2460 struct diff_options *o)
2462 mmfile_t mf1, mf2;
2463 struct checkdiff_t data;
2465 if (!two)
2466 return;
2468 memset(&data, 0, sizeof(data));
2469 data.filename = name_b ? name_b : name_a;
2470 data.lineno = 0;
2471 data.o = o;
2472 data.ws_rule = whitespace_rule(attr_path);
2473 data.conflict_marker_size = ll_merge_marker_size(attr_path);
2475 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2476 die("unable to read files to diff");
2479 * All the other codepaths check both sides, but not checking
2480 * the "old" side here is deliberate. We are checking the newly
2481 * introduced changes, and as long as the "new" side is text, we
2482 * can and should check what it introduces.
2484 if (diff_filespec_is_binary(two))
2485 goto free_and_return;
2486 else {
2487 /* Crazy xdl interfaces.. */
2488 xpparam_t xpp;
2489 xdemitconf_t xecfg;
2491 memset(&xpp, 0, sizeof(xpp));
2492 memset(&xecfg, 0, sizeof(xecfg));
2493 xecfg.ctxlen = 1; /* at least one context line */
2494 xpp.flags = 0;
2495 xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2496 &xpp, &xecfg);
2498 if (data.ws_rule & WS_BLANK_AT_EOF) {
2499 struct emit_callback ecbdata;
2500 int blank_at_eof;
2502 ecbdata.ws_rule = data.ws_rule;
2503 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2504 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2506 if (blank_at_eof) {
2507 static char *err;
2508 if (!err)
2509 err = whitespace_error_string(WS_BLANK_AT_EOF);
2510 fprintf(o->file, "%s:%d: %s.\n",
2511 data.filename, blank_at_eof, err);
2512 data.status = 1; /* report errors */
2516 free_and_return:
2517 diff_free_filespec_data(one);
2518 diff_free_filespec_data(two);
2519 if (data.status)
2520 DIFF_OPT_SET(o, CHECK_FAILED);
2523 struct diff_filespec *alloc_filespec(const char *path)
2525 int namelen = strlen(path);
2526 struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
2528 memset(spec, 0, sizeof(*spec));
2529 spec->path = (char *)(spec + 1);
2530 memcpy(spec->path, path, namelen+1);
2531 spec->count = 1;
2532 spec->is_binary = -1;
2533 return spec;
2536 void free_filespec(struct diff_filespec *spec)
2538 if (!--spec->count) {
2539 diff_free_filespec_data(spec);
2540 free(spec);
2544 void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
2545 int sha1_valid, unsigned short mode)
2547 if (mode) {
2548 spec->mode = canon_mode(mode);
2549 hashcpy(spec->sha1, sha1);
2550 spec->sha1_valid = sha1_valid;
2555 * Given a name and sha1 pair, if the index tells us the file in
2556 * the work tree has that object contents, return true, so that
2557 * prepare_temp_file() does not have to inflate and extract.
2559 static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
2561 struct cache_entry *ce;
2562 struct stat st;
2563 int pos, len;
2566 * We do not read the cache ourselves here, because the
2567 * benchmark with my previous version that always reads cache
2568 * shows that it makes things worse for diff-tree comparing
2569 * two linux-2.6 kernel trees in an already checked out work
2570 * tree. This is because most diff-tree comparisons deal with
2571 * only a small number of files, while reading the cache is
2572 * expensive for a large project, and its cost outweighs the
2573 * savings we get by not inflating the object to a temporary
2574 * file. Practically, this code only helps when we are used
2575 * by diff-cache --cached, which does read the cache before
2576 * calling us.
2578 if (!active_cache)
2579 return 0;
2581 /* We want to avoid the working directory if our caller
2582 * doesn't need the data in a normal file, this system
2583 * is rather slow with its stat/open/mmap/close syscalls,
2584 * and the object is contained in a pack file. The pack
2585 * is probably already open and will be faster to obtain
2586 * the data through than the working directory. Loose
2587 * objects however would tend to be slower as they need
2588 * to be individually opened and inflated.
2590 if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1))
2591 return 0;
2593 len = strlen(name);
2594 pos = cache_name_pos(name, len);
2595 if (pos < 0)
2596 return 0;
2597 ce = active_cache[pos];
2600 * This is not the sha1 we are looking for, or
2601 * unreusable because it is not a regular file.
2603 if (hashcmp(sha1, ce->sha1) || !S_ISREG(ce->ce_mode))
2604 return 0;
2607 * If ce is marked as "assume unchanged", there is no
2608 * guarantee that work tree matches what we are looking for.
2610 if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2611 return 0;
2614 * If ce matches the file in the work tree, we can reuse it.
2616 if (ce_uptodate(ce) ||
2617 (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2618 return 1;
2620 return 0;
2623 static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2625 int len;
2626 char *data = xmalloc(100), *dirty = "";
2628 /* Are we looking at the work tree? */
2629 if (s->dirty_submodule)
2630 dirty = "-dirty";
2632 len = snprintf(data, 100,
2633 "Subproject commit %s%s\n", sha1_to_hex(s->sha1), dirty);
2634 s->data = data;
2635 s->size = len;
2636 s->should_free = 1;
2637 if (size_only) {
2638 s->data = NULL;
2639 free(data);
2641 return 0;
2645 * While doing rename detection and pickaxe operation, we may need to
2646 * grab the data for the blob (or file) for our own in-core comparison.
2647 * diff_filespec has data and size fields for this purpose.
2649 int diff_populate_filespec(struct diff_filespec *s, int size_only)
2651 int err = 0;
2652 if (!DIFF_FILE_VALID(s))
2653 die("internal error: asking to populate invalid file.");
2654 if (S_ISDIR(s->mode))
2655 return -1;
2657 if (s->data)
2658 return 0;
2660 if (size_only && 0 < s->size)
2661 return 0;
2663 if (S_ISGITLINK(s->mode))
2664 return diff_populate_gitlink(s, size_only);
2666 if (!s->sha1_valid ||
2667 reuse_worktree_file(s->path, s->sha1, 0)) {
2668 struct strbuf buf = STRBUF_INIT;
2669 struct stat st;
2670 int fd;
2672 if (lstat(s->path, &st) < 0) {
2673 if (errno == ENOENT) {
2674 err_empty:
2675 err = -1;
2676 empty:
2677 s->data = (char *)"";
2678 s->size = 0;
2679 return err;
2682 s->size = xsize_t(st.st_size);
2683 if (!s->size)
2684 goto empty;
2685 if (S_ISLNK(st.st_mode)) {
2686 struct strbuf sb = STRBUF_INIT;
2688 if (strbuf_readlink(&sb, s->path, s->size))
2689 goto err_empty;
2690 s->size = sb.len;
2691 s->data = strbuf_detach(&sb, NULL);
2692 s->should_free = 1;
2693 return 0;
2695 if (size_only)
2696 return 0;
2697 fd = open(s->path, O_RDONLY);
2698 if (fd < 0)
2699 goto err_empty;
2700 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2701 close(fd);
2702 s->should_munmap = 1;
2705 * Convert from working tree format to canonical git format
2707 if (convert_to_git(s->path, s->data, s->size, &buf, safe_crlf)) {
2708 size_t size = 0;
2709 munmap(s->data, s->size);
2710 s->should_munmap = 0;
2711 s->data = strbuf_detach(&buf, &size);
2712 s->size = size;
2713 s->should_free = 1;
2716 else {
2717 enum object_type type;
2718 if (size_only) {
2719 type = sha1_object_info(s->sha1, &s->size);
2720 if (type < 0)
2721 die("unable to read %s", sha1_to_hex(s->sha1));
2722 } else {
2723 s->data = read_sha1_file(s->sha1, &type, &s->size);
2724 if (!s->data)
2725 die("unable to read %s", sha1_to_hex(s->sha1));
2726 s->should_free = 1;
2729 return 0;
2732 void diff_free_filespec_blob(struct diff_filespec *s)
2734 if (s->should_free)
2735 free(s->data);
2736 else if (s->should_munmap)
2737 munmap(s->data, s->size);
2739 if (s->should_free || s->should_munmap) {
2740 s->should_free = s->should_munmap = 0;
2741 s->data = NULL;
2745 void diff_free_filespec_data(struct diff_filespec *s)
2747 diff_free_filespec_blob(s);
2748 free(s->cnt_data);
2749 s->cnt_data = NULL;
2752 static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2753 void *blob,
2754 unsigned long size,
2755 const unsigned char *sha1,
2756 int mode)
2758 int fd;
2759 struct strbuf buf = STRBUF_INIT;
2760 struct strbuf template = STRBUF_INIT;
2761 char *path_dup = xstrdup(path);
2762 const char *base = basename(path_dup);
2764 /* Generate "XXXXXX_basename.ext" */
2765 strbuf_addstr(&template, "XXXXXX_");
2766 strbuf_addstr(&template, base);
2768 fd = git_mkstemps(temp->tmp_path, PATH_MAX, template.buf,
2769 strlen(base) + 1);
2770 if (fd < 0)
2771 die_errno("unable to create temp-file");
2772 if (convert_to_working_tree(path,
2773 (const char *)blob, (size_t)size, &buf)) {
2774 blob = buf.buf;
2775 size = buf.len;
2777 if (write_in_full(fd, blob, size) != size)
2778 die_errno("unable to write temp-file");
2779 close(fd);
2780 temp->name = temp->tmp_path;
2781 strcpy(temp->hex, sha1_to_hex(sha1));
2782 temp->hex[40] = 0;
2783 sprintf(temp->mode, "%06o", mode);
2784 strbuf_release(&buf);
2785 strbuf_release(&template);
2786 free(path_dup);
2789 static struct diff_tempfile *prepare_temp_file(const char *name,
2790 struct diff_filespec *one)
2792 struct diff_tempfile *temp = claim_diff_tempfile();
2794 if (!DIFF_FILE_VALID(one)) {
2795 not_a_valid_file:
2796 /* A '-' entry produces this for file-2, and
2797 * a '+' entry produces this for file-1.
2799 temp->name = "/dev/null";
2800 strcpy(temp->hex, ".");
2801 strcpy(temp->mode, ".");
2802 return temp;
2805 if (!remove_tempfile_installed) {
2806 atexit(remove_tempfile);
2807 sigchain_push_common(remove_tempfile_on_signal);
2808 remove_tempfile_installed = 1;
2811 if (!one->sha1_valid ||
2812 reuse_worktree_file(name, one->sha1, 1)) {
2813 struct stat st;
2814 if (lstat(name, &st) < 0) {
2815 if (errno == ENOENT)
2816 goto not_a_valid_file;
2817 die_errno("stat(%s)", name);
2819 if (S_ISLNK(st.st_mode)) {
2820 struct strbuf sb = STRBUF_INIT;
2821 if (strbuf_readlink(&sb, name, st.st_size) < 0)
2822 die_errno("readlink(%s)", name);
2823 prep_temp_blob(name, temp, sb.buf, sb.len,
2824 (one->sha1_valid ?
2825 one->sha1 : null_sha1),
2826 (one->sha1_valid ?
2827 one->mode : S_IFLNK));
2828 strbuf_release(&sb);
2830 else {
2831 /* we can borrow from the file in the work tree */
2832 temp->name = name;
2833 if (!one->sha1_valid)
2834 strcpy(temp->hex, sha1_to_hex(null_sha1));
2835 else
2836 strcpy(temp->hex, sha1_to_hex(one->sha1));
2837 /* Even though we may sometimes borrow the
2838 * contents from the work tree, we always want
2839 * one->mode. mode is trustworthy even when
2840 * !(one->sha1_valid), as long as
2841 * DIFF_FILE_VALID(one).
2843 sprintf(temp->mode, "%06o", one->mode);
2845 return temp;
2847 else {
2848 if (diff_populate_filespec(one, 0))
2849 die("cannot read data blob for %s", one->path);
2850 prep_temp_blob(name, temp, one->data, one->size,
2851 one->sha1, one->mode);
2853 return temp;
2856 /* An external diff command takes:
2858 * diff-cmd name infile1 infile1-sha1 infile1-mode \
2859 * infile2 infile2-sha1 infile2-mode [ rename-to ]
2862 static void run_external_diff(const char *pgm,
2863 const char *name,
2864 const char *other,
2865 struct diff_filespec *one,
2866 struct diff_filespec *two,
2867 const char *xfrm_msg,
2868 int complete_rewrite)
2870 const char *spawn_arg[10];
2871 int retval;
2872 const char **arg = &spawn_arg[0];
2874 if (one && two) {
2875 struct diff_tempfile *temp_one, *temp_two;
2876 const char *othername = (other ? other : name);
2877 temp_one = prepare_temp_file(name, one);
2878 temp_two = prepare_temp_file(othername, two);
2879 *arg++ = pgm;
2880 *arg++ = name;
2881 *arg++ = temp_one->name;
2882 *arg++ = temp_one->hex;
2883 *arg++ = temp_one->mode;
2884 *arg++ = temp_two->name;
2885 *arg++ = temp_two->hex;
2886 *arg++ = temp_two->mode;
2887 if (other) {
2888 *arg++ = other;
2889 *arg++ = xfrm_msg;
2891 } else {
2892 *arg++ = pgm;
2893 *arg++ = name;
2895 *arg = NULL;
2896 fflush(NULL);
2897 retval = run_command_v_opt(spawn_arg, RUN_USING_SHELL);
2898 remove_tempfile();
2899 if (retval) {
2900 fprintf(stderr, "external diff died, stopping at %s.\n", name);
2901 exit(1);
2905 static int similarity_index(struct diff_filepair *p)
2907 return p->score * 100 / MAX_SCORE;
2910 static void fill_metainfo(struct strbuf *msg,
2911 const char *name,
2912 const char *other,
2913 struct diff_filespec *one,
2914 struct diff_filespec *two,
2915 struct diff_options *o,
2916 struct diff_filepair *p,
2917 int *must_show_header,
2918 int use_color)
2920 const char *set = diff_get_color(use_color, DIFF_METAINFO);
2921 const char *reset = diff_get_color(use_color, DIFF_RESET);
2922 struct strbuf *msgbuf;
2923 char *line_prefix = "";
2925 *must_show_header = 1;
2926 if (o->output_prefix) {
2927 msgbuf = o->output_prefix(o, o->output_prefix_data);
2928 line_prefix = msgbuf->buf;
2930 strbuf_init(msg, PATH_MAX * 2 + 300);
2931 switch (p->status) {
2932 case DIFF_STATUS_COPIED:
2933 strbuf_addf(msg, "%s%ssimilarity index %d%%",
2934 line_prefix, set, similarity_index(p));
2935 strbuf_addf(msg, "%s\n%s%scopy from ",
2936 reset, line_prefix, set);
2937 quote_c_style(name, msg, NULL, 0);
2938 strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
2939 quote_c_style(other, msg, NULL, 0);
2940 strbuf_addf(msg, "%s\n", reset);
2941 break;
2942 case DIFF_STATUS_RENAMED:
2943 strbuf_addf(msg, "%s%ssimilarity index %d%%",
2944 line_prefix, set, similarity_index(p));
2945 strbuf_addf(msg, "%s\n%s%srename from ",
2946 reset, line_prefix, set);
2947 quote_c_style(name, msg, NULL, 0);
2948 strbuf_addf(msg, "%s\n%s%srename to ",
2949 reset, line_prefix, set);
2950 quote_c_style(other, msg, NULL, 0);
2951 strbuf_addf(msg, "%s\n", reset);
2952 break;
2953 case DIFF_STATUS_MODIFIED:
2954 if (p->score) {
2955 strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
2956 line_prefix,
2957 set, similarity_index(p), reset);
2958 break;
2960 /* fallthru */
2961 default:
2962 *must_show_header = 0;
2964 if (one && two && hashcmp(one->sha1, two->sha1)) {
2965 int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
2967 if (DIFF_OPT_TST(o, BINARY)) {
2968 mmfile_t mf;
2969 if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
2970 (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
2971 abbrev = 40;
2973 strbuf_addf(msg, "%s%sindex %s..", line_prefix, set,
2974 find_unique_abbrev(one->sha1, abbrev));
2975 strbuf_addstr(msg, find_unique_abbrev(two->sha1, abbrev));
2976 if (one->mode == two->mode)
2977 strbuf_addf(msg, " %06o", one->mode);
2978 strbuf_addf(msg, "%s\n", reset);
2982 static void run_diff_cmd(const char *pgm,
2983 const char *name,
2984 const char *other,
2985 const char *attr_path,
2986 struct diff_filespec *one,
2987 struct diff_filespec *two,
2988 struct strbuf *msg,
2989 struct diff_options *o,
2990 struct diff_filepair *p)
2992 const char *xfrm_msg = NULL;
2993 int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
2994 int must_show_header = 0;
2997 if (DIFF_OPT_TST(o, ALLOW_EXTERNAL)) {
2998 struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
2999 if (drv && drv->external)
3000 pgm = drv->external;
3003 if (msg) {
3005 * don't use colors when the header is intended for an
3006 * external diff driver
3008 fill_metainfo(msg, name, other, one, two, o, p,
3009 &must_show_header,
3010 want_color(o->use_color) && !pgm);
3011 xfrm_msg = msg->len ? msg->buf : NULL;
3014 if (pgm) {
3015 run_external_diff(pgm, name, other, one, two, xfrm_msg,
3016 complete_rewrite);
3017 return;
3019 if (one && two)
3020 builtin_diff(name, other ? other : name,
3021 one, two, xfrm_msg, must_show_header,
3022 o, complete_rewrite);
3023 else
3024 fprintf(o->file, "* Unmerged path %s\n", name);
3027 static void diff_fill_sha1_info(struct diff_filespec *one)
3029 if (DIFF_FILE_VALID(one)) {
3030 if (!one->sha1_valid) {
3031 struct stat st;
3032 if (one->is_stdin) {
3033 hashcpy(one->sha1, null_sha1);
3034 return;
3036 if (lstat(one->path, &st) < 0)
3037 die_errno("stat '%s'", one->path);
3038 if (index_path(one->sha1, one->path, &st, 0))
3039 die("cannot hash %s", one->path);
3042 else
3043 hashclr(one->sha1);
3046 static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
3048 /* Strip the prefix but do not molest /dev/null and absolute paths */
3049 if (*namep && **namep != '/') {
3050 *namep += prefix_length;
3051 if (**namep == '/')
3052 ++*namep;
3054 if (*otherp && **otherp != '/') {
3055 *otherp += prefix_length;
3056 if (**otherp == '/')
3057 ++*otherp;
3061 static void run_diff(struct diff_filepair *p, struct diff_options *o)
3063 const char *pgm = external_diff();
3064 struct strbuf msg;
3065 struct diff_filespec *one = p->one;
3066 struct diff_filespec *two = p->two;
3067 const char *name;
3068 const char *other;
3069 const char *attr_path;
3071 name = p->one->path;
3072 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3073 attr_path = name;
3074 if (o->prefix_length)
3075 strip_prefix(o->prefix_length, &name, &other);
3077 if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
3078 pgm = NULL;
3080 if (DIFF_PAIR_UNMERGED(p)) {
3081 run_diff_cmd(pgm, name, NULL, attr_path,
3082 NULL, NULL, NULL, o, p);
3083 return;
3086 diff_fill_sha1_info(one);
3087 diff_fill_sha1_info(two);
3089 if (!pgm &&
3090 DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3091 (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3093 * a filepair that changes between file and symlink
3094 * needs to be split into deletion and creation.
3096 struct diff_filespec *null = alloc_filespec(two->path);
3097 run_diff_cmd(NULL, name, other, attr_path,
3098 one, null, &msg, o, p);
3099 free(null);
3100 strbuf_release(&msg);
3102 null = alloc_filespec(one->path);
3103 run_diff_cmd(NULL, name, other, attr_path,
3104 null, two, &msg, o, p);
3105 free(null);
3107 else
3108 run_diff_cmd(pgm, name, other, attr_path,
3109 one, two, &msg, o, p);
3111 strbuf_release(&msg);
3114 static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3115 struct diffstat_t *diffstat)
3117 const char *name;
3118 const char *other;
3119 int complete_rewrite = 0;
3121 if (DIFF_PAIR_UNMERGED(p)) {
3122 /* unmerged */
3123 builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, 0);
3124 return;
3127 name = p->one->path;
3128 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3130 if (o->prefix_length)
3131 strip_prefix(o->prefix_length, &name, &other);
3133 diff_fill_sha1_info(p->one);
3134 diff_fill_sha1_info(p->two);
3136 if (p->status == DIFF_STATUS_MODIFIED && p->score)
3137 complete_rewrite = 1;
3138 builtin_diffstat(name, other, p->one, p->two, diffstat, o, complete_rewrite);
3141 static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3143 const char *name;
3144 const char *other;
3145 const char *attr_path;
3147 if (DIFF_PAIR_UNMERGED(p)) {
3148 /* unmerged */
3149 return;
3152 name = p->one->path;
3153 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3154 attr_path = other ? other : name;
3156 if (o->prefix_length)
3157 strip_prefix(o->prefix_length, &name, &other);
3159 diff_fill_sha1_info(p->one);
3160 diff_fill_sha1_info(p->two);
3162 builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3165 void diff_setup(struct diff_options *options)
3167 memcpy(options, &default_diff_options, sizeof(*options));
3169 options->file = stdout;
3171 options->line_termination = '\n';
3172 options->break_opt = -1;
3173 options->rename_limit = -1;
3174 options->dirstat_permille = diff_dirstat_permille_default;
3175 options->context = 3;
3176 DIFF_OPT_SET(options, RENAME_EMPTY);
3178 options->change = diff_change;
3179 options->add_remove = diff_addremove;
3180 options->use_color = diff_use_color_default;
3181 options->detect_rename = diff_detect_rename_default;
3183 if (diff_no_prefix) {
3184 options->a_prefix = options->b_prefix = "";
3185 } else if (!diff_mnemonic_prefix) {
3186 options->a_prefix = "a/";
3187 options->b_prefix = "b/";
3191 void diff_setup_done(struct diff_options *options)
3193 int count = 0;
3195 if (options->output_format & DIFF_FORMAT_NAME)
3196 count++;
3197 if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3198 count++;
3199 if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3200 count++;
3201 if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3202 count++;
3203 if (count > 1)
3204 die("--name-only, --name-status, --check and -s are mutually exclusive");
3207 * Most of the time we can say "there are changes"
3208 * only by checking if there are changed paths, but
3209 * --ignore-whitespace* options force us to look
3210 * inside contents.
3213 if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3214 DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3215 DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3216 DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3217 else
3218 DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3220 if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3221 options->detect_rename = DIFF_DETECT_COPY;
3223 if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3224 options->prefix = NULL;
3225 if (options->prefix)
3226 options->prefix_length = strlen(options->prefix);
3227 else
3228 options->prefix_length = 0;
3230 if (options->output_format & (DIFF_FORMAT_NAME |
3231 DIFF_FORMAT_NAME_STATUS |
3232 DIFF_FORMAT_CHECKDIFF |
3233 DIFF_FORMAT_NO_OUTPUT))
3234 options->output_format &= ~(DIFF_FORMAT_RAW |
3235 DIFF_FORMAT_NUMSTAT |
3236 DIFF_FORMAT_DIFFSTAT |
3237 DIFF_FORMAT_SHORTSTAT |
3238 DIFF_FORMAT_DIRSTAT |
3239 DIFF_FORMAT_SUMMARY |
3240 DIFF_FORMAT_PATCH);
3243 * These cases always need recursive; we do not drop caller-supplied
3244 * recursive bits for other formats here.
3246 if (options->output_format & (DIFF_FORMAT_PATCH |
3247 DIFF_FORMAT_NUMSTAT |
3248 DIFF_FORMAT_DIFFSTAT |
3249 DIFF_FORMAT_SHORTSTAT |
3250 DIFF_FORMAT_DIRSTAT |
3251 DIFF_FORMAT_SUMMARY |
3252 DIFF_FORMAT_CHECKDIFF))
3253 DIFF_OPT_SET(options, RECURSIVE);
3255 * Also pickaxe would not work very well if you do not say recursive
3257 if (options->pickaxe)
3258 DIFF_OPT_SET(options, RECURSIVE);
3260 * When patches are generated, submodules diffed against the work tree
3261 * must be checked for dirtiness too so it can be shown in the output
3263 if (options->output_format & DIFF_FORMAT_PATCH)
3264 DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3266 if (options->detect_rename && options->rename_limit < 0)
3267 options->rename_limit = diff_rename_limit_default;
3268 if (options->setup & DIFF_SETUP_USE_CACHE) {
3269 if (!active_cache)
3270 /* read-cache does not die even when it fails
3271 * so it is safe for us to do this here. Also
3272 * it does not smudge active_cache or active_nr
3273 * when it fails, so we do not have to worry about
3274 * cleaning it up ourselves either.
3276 read_cache();
3278 if (options->abbrev <= 0 || 40 < options->abbrev)
3279 options->abbrev = 40; /* full */
3282 * It does not make sense to show the first hit we happened
3283 * to have found. It does not make sense not to return with
3284 * exit code in such a case either.
3286 if (DIFF_OPT_TST(options, QUICK)) {
3287 options->output_format = DIFF_FORMAT_NO_OUTPUT;
3288 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3292 static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3294 char c, *eq;
3295 int len;
3297 if (*arg != '-')
3298 return 0;
3299 c = *++arg;
3300 if (!c)
3301 return 0;
3302 if (c == arg_short) {
3303 c = *++arg;
3304 if (!c)
3305 return 1;
3306 if (val && isdigit(c)) {
3307 char *end;
3308 int n = strtoul(arg, &end, 10);
3309 if (*end)
3310 return 0;
3311 *val = n;
3312 return 1;
3314 return 0;
3316 if (c != '-')
3317 return 0;
3318 arg++;
3319 eq = strchr(arg, '=');
3320 if (eq)
3321 len = eq - arg;
3322 else
3323 len = strlen(arg);
3324 if (!len || strncmp(arg, arg_long, len))
3325 return 0;
3326 if (eq) {
3327 int n;
3328 char *end;
3329 if (!isdigit(*++eq))
3330 return 0;
3331 n = strtoul(eq, &end, 10);
3332 if (*end)
3333 return 0;
3334 *val = n;
3336 return 1;
3339 static int diff_scoreopt_parse(const char *opt);
3341 static inline int short_opt(char opt, const char **argv,
3342 const char **optarg)
3344 const char *arg = argv[0];
3345 if (arg[0] != '-' || arg[1] != opt)
3346 return 0;
3347 if (arg[2] != '\0') {
3348 *optarg = arg + 2;
3349 return 1;
3351 if (!argv[1])
3352 die("Option '%c' requires a value", opt);
3353 *optarg = argv[1];
3354 return 2;
3357 int parse_long_opt(const char *opt, const char **argv,
3358 const char **optarg)
3360 const char *arg = argv[0];
3361 if (arg[0] != '-' || arg[1] != '-')
3362 return 0;
3363 arg += strlen("--");
3364 if (prefixcmp(arg, opt))
3365 return 0;
3366 arg += strlen(opt);
3367 if (*arg == '=') { /* sticked form: --option=value */
3368 *optarg = arg + 1;
3369 return 1;
3371 if (*arg != '\0')
3372 return 0;
3373 /* separate form: --option value */
3374 if (!argv[1])
3375 die("Option '--%s' requires a value", opt);
3376 *optarg = argv[1];
3377 return 2;
3380 static int stat_opt(struct diff_options *options, const char **av)
3382 const char *arg = av[0];
3383 char *end;
3384 int width = options->stat_width;
3385 int name_width = options->stat_name_width;
3386 int graph_width = options->stat_graph_width;
3387 int count = options->stat_count;
3388 int argcount = 1;
3390 arg += strlen("--stat");
3391 end = (char *)arg;
3393 switch (*arg) {
3394 case '-':
3395 if (!prefixcmp(arg, "-width")) {
3396 arg += strlen("-width");
3397 if (*arg == '=')
3398 width = strtoul(arg + 1, &end, 10);
3399 else if (!*arg && !av[1])
3400 die("Option '--stat-width' requires a value");
3401 else if (!*arg) {
3402 width = strtoul(av[1], &end, 10);
3403 argcount = 2;
3405 } else if (!prefixcmp(arg, "-name-width")) {
3406 arg += strlen("-name-width");
3407 if (*arg == '=')
3408 name_width = strtoul(arg + 1, &end, 10);
3409 else if (!*arg && !av[1])
3410 die("Option '--stat-name-width' requires a value");
3411 else if (!*arg) {
3412 name_width = strtoul(av[1], &end, 10);
3413 argcount = 2;
3415 } else if (!prefixcmp(arg, "-graph-width")) {
3416 arg += strlen("-graph-width");
3417 if (*arg == '=')
3418 graph_width = strtoul(arg + 1, &end, 10);
3419 else if (!*arg && !av[1])
3420 die("Option '--stat-graph-width' requires a value");
3421 else if (!*arg) {
3422 graph_width = strtoul(av[1], &end, 10);
3423 argcount = 2;
3425 } else if (!prefixcmp(arg, "-count")) {
3426 arg += strlen("-count");
3427 if (*arg == '=')
3428 count = strtoul(arg + 1, &end, 10);
3429 else if (!*arg && !av[1])
3430 die("Option '--stat-count' requires a value");
3431 else if (!*arg) {
3432 count = strtoul(av[1], &end, 10);
3433 argcount = 2;
3436 break;
3437 case '=':
3438 width = strtoul(arg+1, &end, 10);
3439 if (*end == ',')
3440 name_width = strtoul(end+1, &end, 10);
3441 if (*end == ',')
3442 count = strtoul(end+1, &end, 10);
3445 /* Important! This checks all the error cases! */
3446 if (*end)
3447 return 0;
3448 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3449 options->stat_name_width = name_width;
3450 options->stat_graph_width = graph_width;
3451 options->stat_width = width;
3452 options->stat_count = count;
3453 return argcount;
3456 static int parse_dirstat_opt(struct diff_options *options, const char *params)
3458 struct strbuf errmsg = STRBUF_INIT;
3459 if (parse_dirstat_params(options, params, &errmsg))
3460 die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3461 errmsg.buf);
3462 strbuf_release(&errmsg);
3464 * The caller knows a dirstat-related option is given from the command
3465 * line; allow it to say "return this_function();"
3467 options->output_format |= DIFF_FORMAT_DIRSTAT;
3468 return 1;
3471 int diff_opt_parse(struct diff_options *options, const char **av, int ac)
3473 const char *arg = av[0];
3474 const char *optarg;
3475 int argcount;
3477 /* Output format options */
3478 if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch"))
3479 options->output_format |= DIFF_FORMAT_PATCH;
3480 else if (opt_arg(arg, 'U', "unified", &options->context))
3481 options->output_format |= DIFF_FORMAT_PATCH;
3482 else if (!strcmp(arg, "--raw"))
3483 options->output_format |= DIFF_FORMAT_RAW;
3484 else if (!strcmp(arg, "--patch-with-raw"))
3485 options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_RAW;
3486 else if (!strcmp(arg, "--numstat"))
3487 options->output_format |= DIFF_FORMAT_NUMSTAT;
3488 else if (!strcmp(arg, "--shortstat"))
3489 options->output_format |= DIFF_FORMAT_SHORTSTAT;
3490 else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3491 return parse_dirstat_opt(options, "");
3492 else if (!prefixcmp(arg, "-X"))
3493 return parse_dirstat_opt(options, arg + 2);
3494 else if (!prefixcmp(arg, "--dirstat="))
3495 return parse_dirstat_opt(options, arg + 10);
3496 else if (!strcmp(arg, "--cumulative"))
3497 return parse_dirstat_opt(options, "cumulative");
3498 else if (!strcmp(arg, "--dirstat-by-file"))
3499 return parse_dirstat_opt(options, "files");
3500 else if (!prefixcmp(arg, "--dirstat-by-file=")) {
3501 parse_dirstat_opt(options, "files");
3502 return parse_dirstat_opt(options, arg + 18);
3504 else if (!strcmp(arg, "--check"))
3505 options->output_format |= DIFF_FORMAT_CHECKDIFF;
3506 else if (!strcmp(arg, "--summary"))
3507 options->output_format |= DIFF_FORMAT_SUMMARY;
3508 else if (!strcmp(arg, "--patch-with-stat"))
3509 options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_DIFFSTAT;
3510 else if (!strcmp(arg, "--name-only"))
3511 options->output_format |= DIFF_FORMAT_NAME;
3512 else if (!strcmp(arg, "--name-status"))
3513 options->output_format |= DIFF_FORMAT_NAME_STATUS;
3514 else if (!strcmp(arg, "-s"))
3515 options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3516 else if (!prefixcmp(arg, "--stat"))
3517 /* --stat, --stat-width, --stat-name-width, or --stat-count */
3518 return stat_opt(options, av);
3520 /* renames options */
3521 else if (!prefixcmp(arg, "-B") || !prefixcmp(arg, "--break-rewrites=") ||
3522 !strcmp(arg, "--break-rewrites")) {
3523 if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3524 return error("invalid argument to -B: %s", arg+2);
3526 else if (!prefixcmp(arg, "-M") || !prefixcmp(arg, "--find-renames=") ||
3527 !strcmp(arg, "--find-renames")) {
3528 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3529 return error("invalid argument to -M: %s", arg+2);
3530 options->detect_rename = DIFF_DETECT_RENAME;
3532 else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3533 options->irreversible_delete = 1;
3535 else if (!prefixcmp(arg, "-C") || !prefixcmp(arg, "--find-copies=") ||
3536 !strcmp(arg, "--find-copies")) {
3537 if (options->detect_rename == DIFF_DETECT_COPY)
3538 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3539 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3540 return error("invalid argument to -C: %s", arg+2);
3541 options->detect_rename = DIFF_DETECT_COPY;
3543 else if (!strcmp(arg, "--no-renames"))
3544 options->detect_rename = 0;
3545 else if (!strcmp(arg, "--rename-empty"))
3546 DIFF_OPT_SET(options, RENAME_EMPTY);
3547 else if (!strcmp(arg, "--no-rename-empty"))
3548 DIFF_OPT_CLR(options, RENAME_EMPTY);
3549 else if (!strcmp(arg, "--relative"))
3550 DIFF_OPT_SET(options, RELATIVE_NAME);
3551 else if (!prefixcmp(arg, "--relative=")) {
3552 DIFF_OPT_SET(options, RELATIVE_NAME);
3553 options->prefix = arg + 11;
3556 /* xdiff options */
3557 else if (!strcmp(arg, "--minimal"))
3558 DIFF_XDL_SET(options, NEED_MINIMAL);
3559 else if (!strcmp(arg, "--no-minimal"))
3560 DIFF_XDL_CLR(options, NEED_MINIMAL);
3561 else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3562 DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3563 else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3564 DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3565 else if (!strcmp(arg, "--ignore-space-at-eol"))
3566 DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3567 else if (!strcmp(arg, "--patience"))
3568 options->xdl_opts = DIFF_WITH_ALG(options, PATIENCE_DIFF);
3569 else if (!strcmp(arg, "--histogram"))
3570 options->xdl_opts = DIFF_WITH_ALG(options, HISTOGRAM_DIFF);
3572 /* flags options */
3573 else if (!strcmp(arg, "--binary")) {
3574 options->output_format |= DIFF_FORMAT_PATCH;
3575 DIFF_OPT_SET(options, BINARY);
3577 else if (!strcmp(arg, "--full-index"))
3578 DIFF_OPT_SET(options, FULL_INDEX);
3579 else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
3580 DIFF_OPT_SET(options, TEXT);
3581 else if (!strcmp(arg, "-R"))
3582 DIFF_OPT_SET(options, REVERSE_DIFF);
3583 else if (!strcmp(arg, "--find-copies-harder"))
3584 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3585 else if (!strcmp(arg, "--follow"))
3586 DIFF_OPT_SET(options, FOLLOW_RENAMES);
3587 else if (!strcmp(arg, "--color"))
3588 options->use_color = 1;
3589 else if (!prefixcmp(arg, "--color=")) {
3590 int value = git_config_colorbool(NULL, arg+8);
3591 if (value < 0)
3592 return error("option `color' expects \"always\", \"auto\", or \"never\"");
3593 options->use_color = value;
3595 else if (!strcmp(arg, "--no-color"))
3596 options->use_color = 0;
3597 else if (!strcmp(arg, "--color-words")) {
3598 options->use_color = 1;
3599 options->word_diff = DIFF_WORDS_COLOR;
3601 else if (!prefixcmp(arg, "--color-words=")) {
3602 options->use_color = 1;
3603 options->word_diff = DIFF_WORDS_COLOR;
3604 options->word_regex = arg + 14;
3606 else if (!strcmp(arg, "--word-diff")) {
3607 if (options->word_diff == DIFF_WORDS_NONE)
3608 options->word_diff = DIFF_WORDS_PLAIN;
3610 else if (!prefixcmp(arg, "--word-diff=")) {
3611 const char *type = arg + 12;
3612 if (!strcmp(type, "plain"))
3613 options->word_diff = DIFF_WORDS_PLAIN;
3614 else if (!strcmp(type, "color")) {
3615 options->use_color = 1;
3616 options->word_diff = DIFF_WORDS_COLOR;
3618 else if (!strcmp(type, "porcelain"))
3619 options->word_diff = DIFF_WORDS_PORCELAIN;
3620 else if (!strcmp(type, "none"))
3621 options->word_diff = DIFF_WORDS_NONE;
3622 else
3623 die("bad --word-diff argument: %s", type);
3625 else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
3626 if (options->word_diff == DIFF_WORDS_NONE)
3627 options->word_diff = DIFF_WORDS_PLAIN;
3628 options->word_regex = optarg;
3629 return argcount;
3631 else if (!strcmp(arg, "--exit-code"))
3632 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3633 else if (!strcmp(arg, "--quiet"))
3634 DIFF_OPT_SET(options, QUICK);
3635 else if (!strcmp(arg, "--ext-diff"))
3636 DIFF_OPT_SET(options, ALLOW_EXTERNAL);
3637 else if (!strcmp(arg, "--no-ext-diff"))
3638 DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
3639 else if (!strcmp(arg, "--textconv"))
3640 DIFF_OPT_SET(options, ALLOW_TEXTCONV);
3641 else if (!strcmp(arg, "--no-textconv"))
3642 DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
3643 else if (!strcmp(arg, "--ignore-submodules")) {
3644 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3645 handle_ignore_submodules_arg(options, "all");
3646 } else if (!prefixcmp(arg, "--ignore-submodules=")) {
3647 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3648 handle_ignore_submodules_arg(options, arg + 20);
3649 } else if (!strcmp(arg, "--submodule"))
3650 DIFF_OPT_SET(options, SUBMODULE_LOG);
3651 else if (!prefixcmp(arg, "--submodule=")) {
3652 if (!strcmp(arg + 12, "log"))
3653 DIFF_OPT_SET(options, SUBMODULE_LOG);
3656 /* misc options */
3657 else if (!strcmp(arg, "-z"))
3658 options->line_termination = 0;
3659 else if ((argcount = short_opt('l', av, &optarg))) {
3660 options->rename_limit = strtoul(optarg, NULL, 10);
3661 return argcount;
3663 else if ((argcount = short_opt('S', av, &optarg))) {
3664 options->pickaxe = optarg;
3665 options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
3666 return argcount;
3667 } else if ((argcount = short_opt('G', av, &optarg))) {
3668 options->pickaxe = optarg;
3669 options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
3670 return argcount;
3672 else if (!strcmp(arg, "--pickaxe-all"))
3673 options->pickaxe_opts |= DIFF_PICKAXE_ALL;
3674 else if (!strcmp(arg, "--pickaxe-regex"))
3675 options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
3676 else if ((argcount = short_opt('O', av, &optarg))) {
3677 options->orderfile = optarg;
3678 return argcount;
3680 else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
3681 options->filter = optarg;
3682 return argcount;
3684 else if (!strcmp(arg, "--abbrev"))
3685 options->abbrev = DEFAULT_ABBREV;
3686 else if (!prefixcmp(arg, "--abbrev=")) {
3687 options->abbrev = strtoul(arg + 9, NULL, 10);
3688 if (options->abbrev < MINIMUM_ABBREV)
3689 options->abbrev = MINIMUM_ABBREV;
3690 else if (40 < options->abbrev)
3691 options->abbrev = 40;
3693 else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
3694 options->a_prefix = optarg;
3695 return argcount;
3697 else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
3698 options->b_prefix = optarg;
3699 return argcount;
3701 else if (!strcmp(arg, "--no-prefix"))
3702 options->a_prefix = options->b_prefix = "";
3703 else if (opt_arg(arg, '\0', "inter-hunk-context",
3704 &options->interhunkcontext))
3706 else if (!strcmp(arg, "-W"))
3707 DIFF_OPT_SET(options, FUNCCONTEXT);
3708 else if (!strcmp(arg, "--function-context"))
3709 DIFF_OPT_SET(options, FUNCCONTEXT);
3710 else if (!strcmp(arg, "--no-function-context"))
3711 DIFF_OPT_CLR(options, FUNCCONTEXT);
3712 else if ((argcount = parse_long_opt("output", av, &optarg))) {
3713 options->file = fopen(optarg, "w");
3714 if (!options->file)
3715 die_errno("Could not open '%s'", optarg);
3716 options->close_file = 1;
3717 return argcount;
3718 } else
3719 return 0;
3720 return 1;
3723 int parse_rename_score(const char **cp_p)
3725 unsigned long num, scale;
3726 int ch, dot;
3727 const char *cp = *cp_p;
3729 num = 0;
3730 scale = 1;
3731 dot = 0;
3732 for (;;) {
3733 ch = *cp;
3734 if ( !dot && ch == '.' ) {
3735 scale = 1;
3736 dot = 1;
3737 } else if ( ch == '%' ) {
3738 scale = dot ? scale*100 : 100;
3739 cp++; /* % is always at the end */
3740 break;
3741 } else if ( ch >= '0' && ch <= '9' ) {
3742 if ( scale < 100000 ) {
3743 scale *= 10;
3744 num = (num*10) + (ch-'0');
3746 } else {
3747 break;
3749 cp++;
3751 *cp_p = cp;
3753 /* user says num divided by scale and we say internally that
3754 * is MAX_SCORE * num / scale.
3756 return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
3759 static int diff_scoreopt_parse(const char *opt)
3761 int opt1, opt2, cmd;
3763 if (*opt++ != '-')
3764 return -1;
3765 cmd = *opt++;
3766 if (cmd == '-') {
3767 /* convert the long-form arguments into short-form versions */
3768 if (!prefixcmp(opt, "break-rewrites")) {
3769 opt += strlen("break-rewrites");
3770 if (*opt == 0 || *opt++ == '=')
3771 cmd = 'B';
3772 } else if (!prefixcmp(opt, "find-copies")) {
3773 opt += strlen("find-copies");
3774 if (*opt == 0 || *opt++ == '=')
3775 cmd = 'C';
3776 } else if (!prefixcmp(opt, "find-renames")) {
3777 opt += strlen("find-renames");
3778 if (*opt == 0 || *opt++ == '=')
3779 cmd = 'M';
3782 if (cmd != 'M' && cmd != 'C' && cmd != 'B')
3783 return -1; /* that is not a -M, -C nor -B option */
3785 opt1 = parse_rename_score(&opt);
3786 if (cmd != 'B')
3787 opt2 = 0;
3788 else {
3789 if (*opt == 0)
3790 opt2 = 0;
3791 else if (*opt != '/')
3792 return -1; /* we expect -B80/99 or -B80 */
3793 else {
3794 opt++;
3795 opt2 = parse_rename_score(&opt);
3798 if (*opt != 0)
3799 return -1;
3800 return opt1 | (opt2 << 16);
3803 struct diff_queue_struct diff_queued_diff;
3805 void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
3807 if (queue->alloc <= queue->nr) {
3808 queue->alloc = alloc_nr(queue->alloc);
3809 queue->queue = xrealloc(queue->queue,
3810 sizeof(dp) * queue->alloc);
3812 queue->queue[queue->nr++] = dp;
3815 struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
3816 struct diff_filespec *one,
3817 struct diff_filespec *two)
3819 struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
3820 dp->one = one;
3821 dp->two = two;
3822 if (queue)
3823 diff_q(queue, dp);
3824 return dp;
3827 void diff_free_filepair(struct diff_filepair *p)
3829 free_filespec(p->one);
3830 free_filespec(p->two);
3831 free(p);
3834 /* This is different from find_unique_abbrev() in that
3835 * it stuffs the result with dots for alignment.
3837 const char *diff_unique_abbrev(const unsigned char *sha1, int len)
3839 int abblen;
3840 const char *abbrev;
3841 if (len == 40)
3842 return sha1_to_hex(sha1);
3844 abbrev = find_unique_abbrev(sha1, len);
3845 abblen = strlen(abbrev);
3846 if (abblen < 37) {
3847 static char hex[41];
3848 if (len < abblen && abblen <= len + 2)
3849 sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
3850 else
3851 sprintf(hex, "%s...", abbrev);
3852 return hex;
3854 return sha1_to_hex(sha1);
3857 static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
3859 int line_termination = opt->line_termination;
3860 int inter_name_termination = line_termination ? '\t' : '\0';
3861 if (opt->output_prefix) {
3862 struct strbuf *msg = NULL;
3863 msg = opt->output_prefix(opt, opt->output_prefix_data);
3864 fprintf(opt->file, "%s", msg->buf);
3867 if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
3868 fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
3869 diff_unique_abbrev(p->one->sha1, opt->abbrev));
3870 fprintf(opt->file, "%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
3872 if (p->score) {
3873 fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
3874 inter_name_termination);
3875 } else {
3876 fprintf(opt->file, "%c%c", p->status, inter_name_termination);
3879 if (p->status == DIFF_STATUS_COPIED ||
3880 p->status == DIFF_STATUS_RENAMED) {
3881 const char *name_a, *name_b;
3882 name_a = p->one->path;
3883 name_b = p->two->path;
3884 strip_prefix(opt->prefix_length, &name_a, &name_b);
3885 write_name_quoted(name_a, opt->file, inter_name_termination);
3886 write_name_quoted(name_b, opt->file, line_termination);
3887 } else {
3888 const char *name_a, *name_b;
3889 name_a = p->one->mode ? p->one->path : p->two->path;
3890 name_b = NULL;
3891 strip_prefix(opt->prefix_length, &name_a, &name_b);
3892 write_name_quoted(name_a, opt->file, line_termination);
3896 int diff_unmodified_pair(struct diff_filepair *p)
3898 /* This function is written stricter than necessary to support
3899 * the currently implemented transformers, but the idea is to
3900 * let transformers to produce diff_filepairs any way they want,
3901 * and filter and clean them up here before producing the output.
3903 struct diff_filespec *one = p->one, *two = p->two;
3905 if (DIFF_PAIR_UNMERGED(p))
3906 return 0; /* unmerged is interesting */
3908 /* deletion, addition, mode or type change
3909 * and rename are all interesting.
3911 if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
3912 DIFF_PAIR_MODE_CHANGED(p) ||
3913 strcmp(one->path, two->path))
3914 return 0;
3916 /* both are valid and point at the same path. that is, we are
3917 * dealing with a change.
3919 if (one->sha1_valid && two->sha1_valid &&
3920 !hashcmp(one->sha1, two->sha1) &&
3921 !one->dirty_submodule && !two->dirty_submodule)
3922 return 1; /* no change */
3923 if (!one->sha1_valid && !two->sha1_valid)
3924 return 1; /* both look at the same file on the filesystem. */
3925 return 0;
3928 static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
3930 if (diff_unmodified_pair(p))
3931 return;
3933 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3934 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3935 return; /* no tree diffs in patch format */
3937 run_diff(p, o);
3940 static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
3941 struct diffstat_t *diffstat)
3943 if (diff_unmodified_pair(p))
3944 return;
3946 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3947 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3948 return; /* no useful stat for tree diffs */
3950 run_diffstat(p, o, diffstat);
3953 static void diff_flush_checkdiff(struct diff_filepair *p,
3954 struct diff_options *o)
3956 if (diff_unmodified_pair(p))
3957 return;
3959 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3960 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3961 return; /* nothing to check in tree diffs */
3963 run_checkdiff(p, o);
3966 int diff_queue_is_empty(void)
3968 struct diff_queue_struct *q = &diff_queued_diff;
3969 int i;
3970 for (i = 0; i < q->nr; i++)
3971 if (!diff_unmodified_pair(q->queue[i]))
3972 return 0;
3973 return 1;
3976 #if DIFF_DEBUG
3977 void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
3979 fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
3980 x, one ? one : "",
3981 s->path,
3982 DIFF_FILE_VALID(s) ? "valid" : "invalid",
3983 s->mode,
3984 s->sha1_valid ? sha1_to_hex(s->sha1) : "");
3985 fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
3986 x, one ? one : "",
3987 s->size, s->xfrm_flags);
3990 void diff_debug_filepair(const struct diff_filepair *p, int i)
3992 diff_debug_filespec(p->one, i, "one");
3993 diff_debug_filespec(p->two, i, "two");
3994 fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
3995 p->score, p->status ? p->status : '?',
3996 p->one->rename_used, p->broken_pair);
3999 void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
4001 int i;
4002 if (msg)
4003 fprintf(stderr, "%s\n", msg);
4004 fprintf(stderr, "q->nr = %d\n", q->nr);
4005 for (i = 0; i < q->nr; i++) {
4006 struct diff_filepair *p = q->queue[i];
4007 diff_debug_filepair(p, i);
4010 #endif
4012 static void diff_resolve_rename_copy(void)
4014 int i;
4015 struct diff_filepair *p;
4016 struct diff_queue_struct *q = &diff_queued_diff;
4018 diff_debug_queue("resolve-rename-copy", q);
4020 for (i = 0; i < q->nr; i++) {
4021 p = q->queue[i];
4022 p->status = 0; /* undecided */
4023 if (DIFF_PAIR_UNMERGED(p))
4024 p->status = DIFF_STATUS_UNMERGED;
4025 else if (!DIFF_FILE_VALID(p->one))
4026 p->status = DIFF_STATUS_ADDED;
4027 else if (!DIFF_FILE_VALID(p->two))
4028 p->status = DIFF_STATUS_DELETED;
4029 else if (DIFF_PAIR_TYPE_CHANGED(p))
4030 p->status = DIFF_STATUS_TYPE_CHANGED;
4032 /* from this point on, we are dealing with a pair
4033 * whose both sides are valid and of the same type, i.e.
4034 * either in-place edit or rename/copy edit.
4036 else if (DIFF_PAIR_RENAME(p)) {
4038 * A rename might have re-connected a broken
4039 * pair up, causing the pathnames to be the
4040 * same again. If so, that's not a rename at
4041 * all, just a modification..
4043 * Otherwise, see if this source was used for
4044 * multiple renames, in which case we decrement
4045 * the count, and call it a copy.
4047 if (!strcmp(p->one->path, p->two->path))
4048 p->status = DIFF_STATUS_MODIFIED;
4049 else if (--p->one->rename_used > 0)
4050 p->status = DIFF_STATUS_COPIED;
4051 else
4052 p->status = DIFF_STATUS_RENAMED;
4054 else if (hashcmp(p->one->sha1, p->two->sha1) ||
4055 p->one->mode != p->two->mode ||
4056 p->one->dirty_submodule ||
4057 p->two->dirty_submodule ||
4058 is_null_sha1(p->one->sha1))
4059 p->status = DIFF_STATUS_MODIFIED;
4060 else {
4061 /* This is a "no-change" entry and should not
4062 * happen anymore, but prepare for broken callers.
4064 error("feeding unmodified %s to diffcore",
4065 p->one->path);
4066 p->status = DIFF_STATUS_UNKNOWN;
4069 diff_debug_queue("resolve-rename-copy done", q);
4072 static int check_pair_status(struct diff_filepair *p)
4074 switch (p->status) {
4075 case DIFF_STATUS_UNKNOWN:
4076 return 0;
4077 case 0:
4078 die("internal error in diff-resolve-rename-copy");
4079 default:
4080 return 1;
4084 static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
4086 int fmt = opt->output_format;
4088 if (fmt & DIFF_FORMAT_CHECKDIFF)
4089 diff_flush_checkdiff(p, opt);
4090 else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
4091 diff_flush_raw(p, opt);
4092 else if (fmt & DIFF_FORMAT_NAME) {
4093 const char *name_a, *name_b;
4094 name_a = p->two->path;
4095 name_b = NULL;
4096 strip_prefix(opt->prefix_length, &name_a, &name_b);
4097 write_name_quoted(name_a, opt->file, opt->line_termination);
4101 static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
4103 if (fs->mode)
4104 fprintf(file, " %s mode %06o ", newdelete, fs->mode);
4105 else
4106 fprintf(file, " %s ", newdelete);
4107 write_name_quoted(fs->path, file, '\n');
4111 static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
4112 const char *line_prefix)
4114 if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
4115 fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
4116 p->two->mode, show_name ? ' ' : '\n');
4117 if (show_name) {
4118 write_name_quoted(p->two->path, file, '\n');
4123 static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
4124 const char *line_prefix)
4126 char *names = pprint_rename(p->one->path, p->two->path);
4128 fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
4129 free(names);
4130 show_mode_change(file, p, 0, line_prefix);
4133 static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
4135 FILE *file = opt->file;
4136 char *line_prefix = "";
4138 if (opt->output_prefix) {
4139 struct strbuf *buf = opt->output_prefix(opt, opt->output_prefix_data);
4140 line_prefix = buf->buf;
4143 switch(p->status) {
4144 case DIFF_STATUS_DELETED:
4145 fputs(line_prefix, file);
4146 show_file_mode_name(file, "delete", p->one);
4147 break;
4148 case DIFF_STATUS_ADDED:
4149 fputs(line_prefix, file);
4150 show_file_mode_name(file, "create", p->two);
4151 break;
4152 case DIFF_STATUS_COPIED:
4153 fputs(line_prefix, file);
4154 show_rename_copy(file, "copy", p, line_prefix);
4155 break;
4156 case DIFF_STATUS_RENAMED:
4157 fputs(line_prefix, file);
4158 show_rename_copy(file, "rename", p, line_prefix);
4159 break;
4160 default:
4161 if (p->score) {
4162 fprintf(file, "%s rewrite ", line_prefix);
4163 write_name_quoted(p->two->path, file, ' ');
4164 fprintf(file, "(%d%%)\n", similarity_index(p));
4166 show_mode_change(file, p, !p->score, line_prefix);
4167 break;
4171 struct patch_id_t {
4172 git_SHA_CTX *ctx;
4173 int patchlen;
4176 static int remove_space(char *line, int len)
4178 int i;
4179 char *dst = line;
4180 unsigned char c;
4182 for (i = 0; i < len; i++)
4183 if (!isspace((c = line[i])))
4184 *dst++ = c;
4186 return dst - line;
4189 static void patch_id_consume(void *priv, char *line, unsigned long len)
4191 struct patch_id_t *data = priv;
4192 int new_len;
4194 /* Ignore line numbers when computing the SHA1 of the patch */
4195 if (!prefixcmp(line, "@@ -"))
4196 return;
4198 new_len = remove_space(line, len);
4200 git_SHA1_Update(data->ctx, line, new_len);
4201 data->patchlen += new_len;
4204 /* returns 0 upon success, and writes result into sha1 */
4205 static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
4207 struct diff_queue_struct *q = &diff_queued_diff;
4208 int i;
4209 git_SHA_CTX ctx;
4210 struct patch_id_t data;
4211 char buffer[PATH_MAX * 4 + 20];
4213 git_SHA1_Init(&ctx);
4214 memset(&data, 0, sizeof(struct patch_id_t));
4215 data.ctx = &ctx;
4217 for (i = 0; i < q->nr; i++) {
4218 xpparam_t xpp;
4219 xdemitconf_t xecfg;
4220 mmfile_t mf1, mf2;
4221 struct diff_filepair *p = q->queue[i];
4222 int len1, len2;
4224 memset(&xpp, 0, sizeof(xpp));
4225 memset(&xecfg, 0, sizeof(xecfg));
4226 if (p->status == 0)
4227 return error("internal diff status error");
4228 if (p->status == DIFF_STATUS_UNKNOWN)
4229 continue;
4230 if (diff_unmodified_pair(p))
4231 continue;
4232 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4233 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4234 continue;
4235 if (DIFF_PAIR_UNMERGED(p))
4236 continue;
4238 diff_fill_sha1_info(p->one);
4239 diff_fill_sha1_info(p->two);
4240 if (fill_mmfile(&mf1, p->one) < 0 ||
4241 fill_mmfile(&mf2, p->two) < 0)
4242 return error("unable to read files to diff");
4244 len1 = remove_space(p->one->path, strlen(p->one->path));
4245 len2 = remove_space(p->two->path, strlen(p->two->path));
4246 if (p->one->mode == 0)
4247 len1 = snprintf(buffer, sizeof(buffer),
4248 "diff--gita/%.*sb/%.*s"
4249 "newfilemode%06o"
4250 "---/dev/null"
4251 "+++b/%.*s",
4252 len1, p->one->path,
4253 len2, p->two->path,
4254 p->two->mode,
4255 len2, p->two->path);
4256 else if (p->two->mode == 0)
4257 len1 = snprintf(buffer, sizeof(buffer),
4258 "diff--gita/%.*sb/%.*s"
4259 "deletedfilemode%06o"
4260 "---a/%.*s"
4261 "+++/dev/null",
4262 len1, p->one->path,
4263 len2, p->two->path,
4264 p->one->mode,
4265 len1, p->one->path);
4266 else
4267 len1 = snprintf(buffer, sizeof(buffer),
4268 "diff--gita/%.*sb/%.*s"
4269 "---a/%.*s"
4270 "+++b/%.*s",
4271 len1, p->one->path,
4272 len2, p->two->path,
4273 len1, p->one->path,
4274 len2, p->two->path);
4275 git_SHA1_Update(&ctx, buffer, len1);
4277 if (diff_filespec_is_binary(p->one) ||
4278 diff_filespec_is_binary(p->two)) {
4279 git_SHA1_Update(&ctx, sha1_to_hex(p->one->sha1), 40);
4280 git_SHA1_Update(&ctx, sha1_to_hex(p->two->sha1), 40);
4281 continue;
4284 xpp.flags = 0;
4285 xecfg.ctxlen = 3;
4286 xecfg.flags = 0;
4287 xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4288 &xpp, &xecfg);
4291 git_SHA1_Final(sha1, &ctx);
4292 return 0;
4295 int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
4297 struct diff_queue_struct *q = &diff_queued_diff;
4298 int i;
4299 int result = diff_get_patch_id(options, sha1);
4301 for (i = 0; i < q->nr; i++)
4302 diff_free_filepair(q->queue[i]);
4304 free(q->queue);
4305 DIFF_QUEUE_CLEAR(q);
4307 return result;
4310 static int is_summary_empty(const struct diff_queue_struct *q)
4312 int i;
4314 for (i = 0; i < q->nr; i++) {
4315 const struct diff_filepair *p = q->queue[i];
4317 switch (p->status) {
4318 case DIFF_STATUS_DELETED:
4319 case DIFF_STATUS_ADDED:
4320 case DIFF_STATUS_COPIED:
4321 case DIFF_STATUS_RENAMED:
4322 return 0;
4323 default:
4324 if (p->score)
4325 return 0;
4326 if (p->one->mode && p->two->mode &&
4327 p->one->mode != p->two->mode)
4328 return 0;
4329 break;
4332 return 1;
4335 static const char rename_limit_warning[] =
4336 "inexact rename detection was skipped due to too many files.";
4338 static const char degrade_cc_to_c_warning[] =
4339 "only found copies from modified paths due to too many files.";
4341 static const char rename_limit_advice[] =
4342 "you may want to set your %s variable to at least "
4343 "%d and retry the command.";
4345 void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4347 if (degraded_cc)
4348 warning(degrade_cc_to_c_warning);
4349 else if (needed)
4350 warning(rename_limit_warning);
4351 else
4352 return;
4353 if (0 < needed && needed < 32767)
4354 warning(rename_limit_advice, varname, needed);
4357 void diff_flush(struct diff_options *options)
4359 struct diff_queue_struct *q = &diff_queued_diff;
4360 int i, output_format = options->output_format;
4361 int separator = 0;
4362 int dirstat_by_line = 0;
4365 * Order: raw, stat, summary, patch
4366 * or: name/name-status/checkdiff (other bits clear)
4368 if (!q->nr)
4369 goto free_queue;
4371 if (output_format & (DIFF_FORMAT_RAW |
4372 DIFF_FORMAT_NAME |
4373 DIFF_FORMAT_NAME_STATUS |
4374 DIFF_FORMAT_CHECKDIFF)) {
4375 for (i = 0; i < q->nr; i++) {
4376 struct diff_filepair *p = q->queue[i];
4377 if (check_pair_status(p))
4378 flush_one_pair(p, options);
4380 separator++;
4383 if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4384 dirstat_by_line = 1;
4386 if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4387 dirstat_by_line) {
4388 struct diffstat_t diffstat;
4390 memset(&diffstat, 0, sizeof(struct diffstat_t));
4391 for (i = 0; i < q->nr; i++) {
4392 struct diff_filepair *p = q->queue[i];
4393 if (check_pair_status(p))
4394 diff_flush_stat(p, options, &diffstat);
4396 if (output_format & DIFF_FORMAT_NUMSTAT)
4397 show_numstat(&diffstat, options);
4398 if (output_format & DIFF_FORMAT_DIFFSTAT)
4399 show_stats(&diffstat, options);
4400 if (output_format & DIFF_FORMAT_SHORTSTAT)
4401 show_shortstats(&diffstat, options);
4402 if (output_format & DIFF_FORMAT_DIRSTAT)
4403 show_dirstat_by_line(&diffstat, options);
4404 free_diffstat_info(&diffstat);
4405 separator++;
4407 if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4408 show_dirstat(options);
4410 if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4411 for (i = 0; i < q->nr; i++) {
4412 diff_summary(options, q->queue[i]);
4414 separator++;
4417 if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4418 DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4419 DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4421 * run diff_flush_patch for the exit status. setting
4422 * options->file to /dev/null should be safe, becaue we
4423 * aren't supposed to produce any output anyway.
4425 if (options->close_file)
4426 fclose(options->file);
4427 options->file = fopen("/dev/null", "w");
4428 if (!options->file)
4429 die_errno("Could not open /dev/null");
4430 options->close_file = 1;
4431 for (i = 0; i < q->nr; i++) {
4432 struct diff_filepair *p = q->queue[i];
4433 if (check_pair_status(p))
4434 diff_flush_patch(p, options);
4435 if (options->found_changes)
4436 break;
4440 if (output_format & DIFF_FORMAT_PATCH) {
4441 if (separator) {
4442 if (options->output_prefix) {
4443 struct strbuf *msg = NULL;
4444 msg = options->output_prefix(options,
4445 options->output_prefix_data);
4446 fwrite(msg->buf, msg->len, 1, stdout);
4448 putc(options->line_termination, options->file);
4449 if (options->stat_sep) {
4450 /* attach patch instead of inline */
4451 fputs(options->stat_sep, options->file);
4455 for (i = 0; i < q->nr; i++) {
4456 struct diff_filepair *p = q->queue[i];
4457 if (check_pair_status(p))
4458 diff_flush_patch(p, options);
4462 if (output_format & DIFF_FORMAT_CALLBACK)
4463 options->format_callback(q, options, options->format_callback_data);
4465 for (i = 0; i < q->nr; i++)
4466 diff_free_filepair(q->queue[i]);
4467 free_queue:
4468 free(q->queue);
4469 DIFF_QUEUE_CLEAR(q);
4470 if (options->close_file)
4471 fclose(options->file);
4474 * Report the content-level differences with HAS_CHANGES;
4475 * diff_addremove/diff_change does not set the bit when
4476 * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4478 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4479 if (options->found_changes)
4480 DIFF_OPT_SET(options, HAS_CHANGES);
4481 else
4482 DIFF_OPT_CLR(options, HAS_CHANGES);
4486 static void diffcore_apply_filter(const char *filter)
4488 int i;
4489 struct diff_queue_struct *q = &diff_queued_diff;
4490 struct diff_queue_struct outq;
4491 DIFF_QUEUE_CLEAR(&outq);
4493 if (!filter)
4494 return;
4496 if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
4497 int found;
4498 for (i = found = 0; !found && i < q->nr; i++) {
4499 struct diff_filepair *p = q->queue[i];
4500 if (((p->status == DIFF_STATUS_MODIFIED) &&
4501 ((p->score &&
4502 strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
4503 (!p->score &&
4504 strchr(filter, DIFF_STATUS_MODIFIED)))) ||
4505 ((p->status != DIFF_STATUS_MODIFIED) &&
4506 strchr(filter, p->status)))
4507 found++;
4509 if (found)
4510 return;
4512 /* otherwise we will clear the whole queue
4513 * by copying the empty outq at the end of this
4514 * function, but first clear the current entries
4515 * in the queue.
4517 for (i = 0; i < q->nr; i++)
4518 diff_free_filepair(q->queue[i]);
4520 else {
4521 /* Only the matching ones */
4522 for (i = 0; i < q->nr; i++) {
4523 struct diff_filepair *p = q->queue[i];
4525 if (((p->status == DIFF_STATUS_MODIFIED) &&
4526 ((p->score &&
4527 strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
4528 (!p->score &&
4529 strchr(filter, DIFF_STATUS_MODIFIED)))) ||
4530 ((p->status != DIFF_STATUS_MODIFIED) &&
4531 strchr(filter, p->status)))
4532 diff_q(&outq, p);
4533 else
4534 diff_free_filepair(p);
4537 free(q->queue);
4538 *q = outq;
4541 /* Check whether two filespecs with the same mode and size are identical */
4542 static int diff_filespec_is_identical(struct diff_filespec *one,
4543 struct diff_filespec *two)
4545 if (S_ISGITLINK(one->mode))
4546 return 0;
4547 if (diff_populate_filespec(one, 0))
4548 return 0;
4549 if (diff_populate_filespec(two, 0))
4550 return 0;
4551 return !memcmp(one->data, two->data, one->size);
4554 static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
4556 int i;
4557 struct diff_queue_struct *q = &diff_queued_diff;
4558 struct diff_queue_struct outq;
4559 DIFF_QUEUE_CLEAR(&outq);
4561 for (i = 0; i < q->nr; i++) {
4562 struct diff_filepair *p = q->queue[i];
4565 * 1. Entries that come from stat info dirtiness
4566 * always have both sides (iow, not create/delete),
4567 * one side of the object name is unknown, with
4568 * the same mode and size. Keep the ones that
4569 * do not match these criteria. They have real
4570 * differences.
4572 * 2. At this point, the file is known to be modified,
4573 * with the same mode and size, and the object
4574 * name of one side is unknown. Need to inspect
4575 * the identical contents.
4577 if (!DIFF_FILE_VALID(p->one) || /* (1) */
4578 !DIFF_FILE_VALID(p->two) ||
4579 (p->one->sha1_valid && p->two->sha1_valid) ||
4580 (p->one->mode != p->two->mode) ||
4581 diff_populate_filespec(p->one, 1) ||
4582 diff_populate_filespec(p->two, 1) ||
4583 (p->one->size != p->two->size) ||
4584 !diff_filespec_is_identical(p->one, p->two)) /* (2) */
4585 diff_q(&outq, p);
4586 else {
4588 * The caller can subtract 1 from skip_stat_unmatch
4589 * to determine how many paths were dirty only
4590 * due to stat info mismatch.
4592 if (!DIFF_OPT_TST(diffopt, NO_INDEX))
4593 diffopt->skip_stat_unmatch++;
4594 diff_free_filepair(p);
4597 free(q->queue);
4598 *q = outq;
4601 static int diffnamecmp(const void *a_, const void *b_)
4603 const struct diff_filepair *a = *((const struct diff_filepair **)a_);
4604 const struct diff_filepair *b = *((const struct diff_filepair **)b_);
4605 const char *name_a, *name_b;
4607 name_a = a->one ? a->one->path : a->two->path;
4608 name_b = b->one ? b->one->path : b->two->path;
4609 return strcmp(name_a, name_b);
4612 void diffcore_fix_diff_index(struct diff_options *options)
4614 struct diff_queue_struct *q = &diff_queued_diff;
4615 qsort(q->queue, q->nr, sizeof(q->queue[0]), diffnamecmp);
4618 void diffcore_std(struct diff_options *options)
4620 if (options->skip_stat_unmatch)
4621 diffcore_skip_stat_unmatch(options);
4622 if (!options->found_follow) {
4623 /* See try_to_follow_renames() in tree-diff.c */
4624 if (options->break_opt != -1)
4625 diffcore_break(options->break_opt);
4626 if (options->detect_rename)
4627 diffcore_rename(options);
4628 if (options->break_opt != -1)
4629 diffcore_merge_broken();
4631 if (options->pickaxe)
4632 diffcore_pickaxe(options);
4633 if (options->orderfile)
4634 diffcore_order(options->orderfile);
4635 if (!options->found_follow)
4636 /* See try_to_follow_renames() in tree-diff.c */
4637 diff_resolve_rename_copy();
4638 diffcore_apply_filter(options->filter);
4640 if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4641 DIFF_OPT_SET(options, HAS_CHANGES);
4642 else
4643 DIFF_OPT_CLR(options, HAS_CHANGES);
4645 options->found_follow = 0;
4648 int diff_result_code(struct diff_options *opt, int status)
4650 int result = 0;
4652 diff_warn_rename_limit("diff.renamelimit",
4653 opt->needed_rename_limit,
4654 opt->degraded_cc_to_c);
4655 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4656 !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
4657 return status;
4658 if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4659 DIFF_OPT_TST(opt, HAS_CHANGES))
4660 result |= 01;
4661 if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
4662 DIFF_OPT_TST(opt, CHECK_FAILED))
4663 result |= 02;
4664 return result;
4667 int diff_can_quit_early(struct diff_options *opt)
4669 return (DIFF_OPT_TST(opt, QUICK) &&
4670 !opt->filter &&
4671 DIFF_OPT_TST(opt, HAS_CHANGES));
4675 * Shall changes to this submodule be ignored?
4677 * Submodule changes can be configured to be ignored separately for each path,
4678 * but that configuration can be overridden from the command line.
4680 static int is_submodule_ignored(const char *path, struct diff_options *options)
4682 int ignored = 0;
4683 unsigned orig_flags = options->flags;
4684 if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
4685 set_diffopt_flags_from_submodule_config(options, path);
4686 if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
4687 ignored = 1;
4688 options->flags = orig_flags;
4689 return ignored;
4692 void diff_addremove(struct diff_options *options,
4693 int addremove, unsigned mode,
4694 const unsigned char *sha1,
4695 int sha1_valid,
4696 const char *concatpath, unsigned dirty_submodule)
4698 struct diff_filespec *one, *two;
4700 if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
4701 return;
4703 /* This may look odd, but it is a preparation for
4704 * feeding "there are unchanged files which should
4705 * not produce diffs, but when you are doing copy
4706 * detection you would need them, so here they are"
4707 * entries to the diff-core. They will be prefixed
4708 * with something like '=' or '*' (I haven't decided
4709 * which but should not make any difference).
4710 * Feeding the same new and old to diff_change()
4711 * also has the same effect.
4712 * Before the final output happens, they are pruned after
4713 * merged into rename/copy pairs as appropriate.
4715 if (DIFF_OPT_TST(options, REVERSE_DIFF))
4716 addremove = (addremove == '+' ? '-' :
4717 addremove == '-' ? '+' : addremove);
4719 if (options->prefix &&
4720 strncmp(concatpath, options->prefix, options->prefix_length))
4721 return;
4723 one = alloc_filespec(concatpath);
4724 two = alloc_filespec(concatpath);
4726 if (addremove != '+')
4727 fill_filespec(one, sha1, sha1_valid, mode);
4728 if (addremove != '-') {
4729 fill_filespec(two, sha1, sha1_valid, mode);
4730 two->dirty_submodule = dirty_submodule;
4733 diff_queue(&diff_queued_diff, one, two);
4734 if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4735 DIFF_OPT_SET(options, HAS_CHANGES);
4738 void diff_change(struct diff_options *options,
4739 unsigned old_mode, unsigned new_mode,
4740 const unsigned char *old_sha1,
4741 const unsigned char *new_sha1,
4742 int old_sha1_valid, int new_sha1_valid,
4743 const char *concatpath,
4744 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
4746 struct diff_filespec *one, *two;
4748 if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
4749 is_submodule_ignored(concatpath, options))
4750 return;
4752 if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
4753 unsigned tmp;
4754 const unsigned char *tmp_c;
4755 tmp = old_mode; old_mode = new_mode; new_mode = tmp;
4756 tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
4757 tmp = old_sha1_valid; old_sha1_valid = new_sha1_valid;
4758 new_sha1_valid = tmp;
4759 tmp = old_dirty_submodule; old_dirty_submodule = new_dirty_submodule;
4760 new_dirty_submodule = tmp;
4763 if (options->prefix &&
4764 strncmp(concatpath, options->prefix, options->prefix_length))
4765 return;
4767 one = alloc_filespec(concatpath);
4768 two = alloc_filespec(concatpath);
4769 fill_filespec(one, old_sha1, old_sha1_valid, old_mode);
4770 fill_filespec(two, new_sha1, new_sha1_valid, new_mode);
4771 one->dirty_submodule = old_dirty_submodule;
4772 two->dirty_submodule = new_dirty_submodule;
4774 diff_queue(&diff_queued_diff, one, two);
4775 if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4776 DIFF_OPT_SET(options, HAS_CHANGES);
4779 struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
4781 struct diff_filepair *pair;
4782 struct diff_filespec *one, *two;
4784 if (options->prefix &&
4785 strncmp(path, options->prefix, options->prefix_length))
4786 return NULL;
4788 one = alloc_filespec(path);
4789 two = alloc_filespec(path);
4790 pair = diff_queue(&diff_queued_diff, one, two);
4791 pair->is_unmerged = 1;
4792 return pair;
4795 static char *run_textconv(const char *pgm, struct diff_filespec *spec,
4796 size_t *outsize)
4798 struct diff_tempfile *temp;
4799 const char *argv[3];
4800 const char **arg = argv;
4801 struct child_process child;
4802 struct strbuf buf = STRBUF_INIT;
4803 int err = 0;
4805 temp = prepare_temp_file(spec->path, spec);
4806 *arg++ = pgm;
4807 *arg++ = temp->name;
4808 *arg = NULL;
4810 memset(&child, 0, sizeof(child));
4811 child.use_shell = 1;
4812 child.argv = argv;
4813 child.out = -1;
4814 if (start_command(&child)) {
4815 remove_tempfile();
4816 return NULL;
4819 if (strbuf_read(&buf, child.out, 0) < 0)
4820 err = error("error reading from textconv command '%s'", pgm);
4821 close(child.out);
4823 if (finish_command(&child) || err) {
4824 strbuf_release(&buf);
4825 remove_tempfile();
4826 return NULL;
4828 remove_tempfile();
4830 return strbuf_detach(&buf, outsize);
4833 size_t fill_textconv(struct userdiff_driver *driver,
4834 struct diff_filespec *df,
4835 char **outbuf)
4837 size_t size;
4839 if (!driver || !driver->textconv) {
4840 if (!DIFF_FILE_VALID(df)) {
4841 *outbuf = "";
4842 return 0;
4844 if (diff_populate_filespec(df, 0))
4845 die("unable to read files to diff");
4846 *outbuf = df->data;
4847 return df->size;
4850 if (driver->textconv_cache && df->sha1_valid) {
4851 *outbuf = notes_cache_get(driver->textconv_cache, df->sha1,
4852 &size);
4853 if (*outbuf)
4854 return size;
4857 *outbuf = run_textconv(driver->textconv, df, &size);
4858 if (!*outbuf)
4859 die("unable to read files to diff");
4861 if (driver->textconv_cache && df->sha1_valid) {
4862 /* ignore errors, as we might be in a readonly repository */
4863 notes_cache_put(driver->textconv_cache, df->sha1, *outbuf,
4864 size);
4866 * we could save up changes and flush them all at the end,
4867 * but we would need an extra call after all diffing is done.
4868 * Since generating a cache entry is the slow path anyway,
4869 * this extra overhead probably isn't a big deal.
4871 notes_cache_write(driver->textconv_cache);
4874 return size;