Merge branches zj/decimal-width, zj/term-columns and jc/diff-stat-scaler
[alt-git.git] / diff.c
blobe0590a305655b6da3c5903875b3f2080f47a57c7
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_dirstat_permille_default = 30;
35 static struct diff_options default_diff_options;
37 static char diff_colors[][COLOR_MAXLEN] = {
38 GIT_COLOR_RESET,
39 GIT_COLOR_NORMAL, /* PLAIN */
40 GIT_COLOR_BOLD, /* METAINFO */
41 GIT_COLOR_CYAN, /* FRAGINFO */
42 GIT_COLOR_RED, /* OLD */
43 GIT_COLOR_GREEN, /* NEW */
44 GIT_COLOR_YELLOW, /* COMMIT */
45 GIT_COLOR_BG_RED, /* WHITESPACE */
46 GIT_COLOR_NORMAL, /* FUNCINFO */
49 static int parse_diff_color_slot(const char *var, int ofs)
51 if (!strcasecmp(var+ofs, "plain"))
52 return DIFF_PLAIN;
53 if (!strcasecmp(var+ofs, "meta"))
54 return DIFF_METAINFO;
55 if (!strcasecmp(var+ofs, "frag"))
56 return DIFF_FRAGINFO;
57 if (!strcasecmp(var+ofs, "old"))
58 return DIFF_FILE_OLD;
59 if (!strcasecmp(var+ofs, "new"))
60 return DIFF_FILE_NEW;
61 if (!strcasecmp(var+ofs, "commit"))
62 return DIFF_COMMIT;
63 if (!strcasecmp(var+ofs, "whitespace"))
64 return DIFF_WHITESPACE;
65 if (!strcasecmp(var+ofs, "func"))
66 return DIFF_FUNCINFO;
67 return -1;
70 static int parse_dirstat_params(struct diff_options *options, const char *params,
71 struct strbuf *errmsg)
73 const char *p = params;
74 int p_len, ret = 0;
76 while (*p) {
77 p_len = strchrnul(p, ',') - p;
78 if (!memcmp(p, "changes", p_len)) {
79 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
80 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
81 } else if (!memcmp(p, "lines", p_len)) {
82 DIFF_OPT_SET(options, DIRSTAT_BY_LINE);
83 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
84 } else if (!memcmp(p, "files", p_len)) {
85 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
86 DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
87 } else if (!memcmp(p, "noncumulative", p_len)) {
88 DIFF_OPT_CLR(options, DIRSTAT_CUMULATIVE);
89 } else if (!memcmp(p, "cumulative", p_len)) {
90 DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
91 } else if (isdigit(*p)) {
92 char *end;
93 int permille = strtoul(p, &end, 10) * 10;
94 if (*end == '.' && isdigit(*++end)) {
95 /* only use first digit */
96 permille += *end - '0';
97 /* .. and ignore any further digits */
98 while (isdigit(*++end))
99 ; /* nothing */
101 if (end - p == p_len)
102 options->dirstat_permille = permille;
103 else {
104 strbuf_addf(errmsg, _(" Failed to parse dirstat cut-off percentage '%.*s'\n"),
105 p_len, p);
106 ret++;
108 } else {
109 strbuf_addf(errmsg, _(" Unknown dirstat parameter '%.*s'\n"),
110 p_len, p);
111 ret++;
114 p += p_len;
116 if (*p)
117 p++; /* more parameters, swallow separator */
119 return ret;
122 static int git_config_rename(const char *var, const char *value)
124 if (!value)
125 return DIFF_DETECT_RENAME;
126 if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
127 return DIFF_DETECT_COPY;
128 return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
132 * These are to give UI layer defaults.
133 * The core-level commands such as git-diff-files should
134 * never be affected by the setting of diff.renames
135 * the user happens to have in the configuration file.
137 int git_diff_ui_config(const char *var, const char *value, void *cb)
139 if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
140 diff_use_color_default = git_config_colorbool(var, value);
141 return 0;
143 if (!strcmp(var, "diff.renames")) {
144 diff_detect_rename_default = git_config_rename(var, value);
145 return 0;
147 if (!strcmp(var, "diff.autorefreshindex")) {
148 diff_auto_refresh_index = git_config_bool(var, value);
149 return 0;
151 if (!strcmp(var, "diff.mnemonicprefix")) {
152 diff_mnemonic_prefix = git_config_bool(var, value);
153 return 0;
155 if (!strcmp(var, "diff.noprefix")) {
156 diff_no_prefix = git_config_bool(var, value);
157 return 0;
159 if (!strcmp(var, "diff.external"))
160 return git_config_string(&external_diff_cmd_cfg, var, value);
161 if (!strcmp(var, "diff.wordregex"))
162 return git_config_string(&diff_word_regex_cfg, var, value);
164 if (!strcmp(var, "diff.ignoresubmodules"))
165 handle_ignore_submodules_arg(&default_diff_options, value);
167 if (git_color_config(var, value, cb) < 0)
168 return -1;
170 return git_diff_basic_config(var, value, cb);
173 int git_diff_basic_config(const char *var, const char *value, void *cb)
175 if (!strcmp(var, "diff.renamelimit")) {
176 diff_rename_limit_default = git_config_int(var, value);
177 return 0;
180 switch (userdiff_config(var, value)) {
181 case 0: break;
182 case -1: return -1;
183 default: return 0;
186 if (!prefixcmp(var, "diff.color.") || !prefixcmp(var, "color.diff.")) {
187 int slot = parse_diff_color_slot(var, 11);
188 if (slot < 0)
189 return 0;
190 if (!value)
191 return config_error_nonbool(var);
192 color_parse(value, var, diff_colors[slot]);
193 return 0;
196 /* like GNU diff's --suppress-blank-empty option */
197 if (!strcmp(var, "diff.suppressblankempty") ||
198 /* for backwards compatibility */
199 !strcmp(var, "diff.suppress-blank-empty")) {
200 diff_suppress_blank_empty = git_config_bool(var, value);
201 return 0;
204 if (!strcmp(var, "diff.dirstat")) {
205 struct strbuf errmsg = STRBUF_INIT;
206 default_diff_options.dirstat_permille = diff_dirstat_permille_default;
207 if (parse_dirstat_params(&default_diff_options, value, &errmsg))
208 warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
209 errmsg.buf);
210 strbuf_release(&errmsg);
211 diff_dirstat_permille_default = default_diff_options.dirstat_permille;
212 return 0;
215 if (!prefixcmp(var, "submodule."))
216 return parse_submodule_config_option(var, value);
218 return git_default_config(var, value, cb);
221 static char *quote_two(const char *one, const char *two)
223 int need_one = quote_c_style(one, NULL, NULL, 1);
224 int need_two = quote_c_style(two, NULL, NULL, 1);
225 struct strbuf res = STRBUF_INIT;
227 if (need_one + need_two) {
228 strbuf_addch(&res, '"');
229 quote_c_style(one, &res, NULL, 1);
230 quote_c_style(two, &res, NULL, 1);
231 strbuf_addch(&res, '"');
232 } else {
233 strbuf_addstr(&res, one);
234 strbuf_addstr(&res, two);
236 return strbuf_detach(&res, NULL);
239 static const char *external_diff(void)
241 static const char *external_diff_cmd = NULL;
242 static int done_preparing = 0;
244 if (done_preparing)
245 return external_diff_cmd;
246 external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
247 if (!external_diff_cmd)
248 external_diff_cmd = external_diff_cmd_cfg;
249 done_preparing = 1;
250 return external_diff_cmd;
253 static struct diff_tempfile {
254 const char *name; /* filename external diff should read from */
255 char hex[41];
256 char mode[10];
257 char tmp_path[PATH_MAX];
258 } diff_temp[2];
260 typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
262 struct emit_callback {
263 int color_diff;
264 unsigned ws_rule;
265 int blank_at_eof_in_preimage;
266 int blank_at_eof_in_postimage;
267 int lno_in_preimage;
268 int lno_in_postimage;
269 sane_truncate_fn truncate;
270 const char **label_path;
271 struct diff_words_data *diff_words;
272 struct diff_options *opt;
273 int *found_changesp;
274 struct strbuf *header;
277 static int count_lines(const char *data, int size)
279 int count, ch, completely_empty = 1, nl_just_seen = 0;
280 count = 0;
281 while (0 < size--) {
282 ch = *data++;
283 if (ch == '\n') {
284 count++;
285 nl_just_seen = 1;
286 completely_empty = 0;
288 else {
289 nl_just_seen = 0;
290 completely_empty = 0;
293 if (completely_empty)
294 return 0;
295 if (!nl_just_seen)
296 count++; /* no trailing newline */
297 return count;
300 static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
302 if (!DIFF_FILE_VALID(one)) {
303 mf->ptr = (char *)""; /* does not matter */
304 mf->size = 0;
305 return 0;
307 else if (diff_populate_filespec(one, 0))
308 return -1;
310 mf->ptr = one->data;
311 mf->size = one->size;
312 return 0;
315 /* like fill_mmfile, but only for size, so we can avoid retrieving blob */
316 static unsigned long diff_filespec_size(struct diff_filespec *one)
318 if (!DIFF_FILE_VALID(one))
319 return 0;
320 diff_populate_filespec(one, 1);
321 return one->size;
324 static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
326 char *ptr = mf->ptr;
327 long size = mf->size;
328 int cnt = 0;
330 if (!size)
331 return cnt;
332 ptr += size - 1; /* pointing at the very end */
333 if (*ptr != '\n')
334 ; /* incomplete line */
335 else
336 ptr--; /* skip the last LF */
337 while (mf->ptr < ptr) {
338 char *prev_eol;
339 for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
340 if (*prev_eol == '\n')
341 break;
342 if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
343 break;
344 cnt++;
345 ptr = prev_eol - 1;
347 return cnt;
350 static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
351 struct emit_callback *ecbdata)
353 int l1, l2, at;
354 unsigned ws_rule = ecbdata->ws_rule;
355 l1 = count_trailing_blank(mf1, ws_rule);
356 l2 = count_trailing_blank(mf2, ws_rule);
357 if (l2 <= l1) {
358 ecbdata->blank_at_eof_in_preimage = 0;
359 ecbdata->blank_at_eof_in_postimage = 0;
360 return;
362 at = count_lines(mf1->ptr, mf1->size);
363 ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
365 at = count_lines(mf2->ptr, mf2->size);
366 ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
369 static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
370 int first, const char *line, int len)
372 int has_trailing_newline, has_trailing_carriage_return;
373 int nofirst;
374 FILE *file = o->file;
376 if (o->output_prefix) {
377 struct strbuf *msg = NULL;
378 msg = o->output_prefix(o, o->output_prefix_data);
379 assert(msg);
380 fwrite(msg->buf, msg->len, 1, file);
383 if (len == 0) {
384 has_trailing_newline = (first == '\n');
385 has_trailing_carriage_return = (!has_trailing_newline &&
386 (first == '\r'));
387 nofirst = has_trailing_newline || has_trailing_carriage_return;
388 } else {
389 has_trailing_newline = (len > 0 && line[len-1] == '\n');
390 if (has_trailing_newline)
391 len--;
392 has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
393 if (has_trailing_carriage_return)
394 len--;
395 nofirst = 0;
398 if (len || !nofirst) {
399 fputs(set, file);
400 if (!nofirst)
401 fputc(first, file);
402 fwrite(line, len, 1, file);
403 fputs(reset, file);
405 if (has_trailing_carriage_return)
406 fputc('\r', file);
407 if (has_trailing_newline)
408 fputc('\n', file);
411 static void emit_line(struct diff_options *o, const char *set, const char *reset,
412 const char *line, int len)
414 emit_line_0(o, set, reset, line[0], line+1, len-1);
417 static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
419 if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
420 ecbdata->blank_at_eof_in_preimage &&
421 ecbdata->blank_at_eof_in_postimage &&
422 ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
423 ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
424 return 0;
425 return ws_blank_line(line, len, ecbdata->ws_rule);
428 static void emit_add_line(const char *reset,
429 struct emit_callback *ecbdata,
430 const char *line, int len)
432 const char *ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
433 const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_NEW);
435 if (!*ws)
436 emit_line_0(ecbdata->opt, set, reset, '+', line, len);
437 else if (new_blank_line_at_eof(ecbdata, line, len))
438 /* Blank line at EOF - paint '+' as well */
439 emit_line_0(ecbdata->opt, ws, reset, '+', line, len);
440 else {
441 /* Emit just the prefix, then the rest. */
442 emit_line_0(ecbdata->opt, set, reset, '+', "", 0);
443 ws_check_emit(line, len, ecbdata->ws_rule,
444 ecbdata->opt->file, set, reset, ws);
448 static void emit_hunk_header(struct emit_callback *ecbdata,
449 const char *line, int len)
451 const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
452 const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
453 const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
454 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
455 static const char atat[2] = { '@', '@' };
456 const char *cp, *ep;
457 struct strbuf msgbuf = STRBUF_INIT;
458 int org_len = len;
459 int i = 1;
462 * As a hunk header must begin with "@@ -<old>, +<new> @@",
463 * it always is at least 10 bytes long.
465 if (len < 10 ||
466 memcmp(line, atat, 2) ||
467 !(ep = memmem(line + 2, len - 2, atat, 2))) {
468 emit_line(ecbdata->opt, plain, reset, line, len);
469 return;
471 ep += 2; /* skip over @@ */
473 /* The hunk header in fraginfo color */
474 strbuf_add(&msgbuf, frag, strlen(frag));
475 strbuf_add(&msgbuf, line, ep - line);
476 strbuf_add(&msgbuf, reset, strlen(reset));
479 * trailing "\r\n"
481 for ( ; i < 3; i++)
482 if (line[len - i] == '\r' || line[len - i] == '\n')
483 len--;
485 /* blank before the func header */
486 for (cp = ep; ep - line < len; ep++)
487 if (*ep != ' ' && *ep != '\t')
488 break;
489 if (ep != cp) {
490 strbuf_add(&msgbuf, plain, strlen(plain));
491 strbuf_add(&msgbuf, cp, ep - cp);
492 strbuf_add(&msgbuf, reset, strlen(reset));
495 if (ep < line + len) {
496 strbuf_add(&msgbuf, func, strlen(func));
497 strbuf_add(&msgbuf, ep, line + len - ep);
498 strbuf_add(&msgbuf, reset, strlen(reset));
501 strbuf_add(&msgbuf, line + len, org_len - len);
502 emit_line(ecbdata->opt, "", "", msgbuf.buf, msgbuf.len);
503 strbuf_release(&msgbuf);
506 static struct diff_tempfile *claim_diff_tempfile(void) {
507 int i;
508 for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
509 if (!diff_temp[i].name)
510 return diff_temp + i;
511 die("BUG: diff is failing to clean up its tempfiles");
514 static int remove_tempfile_installed;
516 static void remove_tempfile(void)
518 int i;
519 for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
520 if (diff_temp[i].name == diff_temp[i].tmp_path)
521 unlink_or_warn(diff_temp[i].name);
522 diff_temp[i].name = NULL;
526 static void remove_tempfile_on_signal(int signo)
528 remove_tempfile();
529 sigchain_pop(signo);
530 raise(signo);
533 static void print_line_count(FILE *file, int count)
535 switch (count) {
536 case 0:
537 fprintf(file, "0,0");
538 break;
539 case 1:
540 fprintf(file, "1");
541 break;
542 default:
543 fprintf(file, "1,%d", count);
544 break;
548 static void emit_rewrite_lines(struct emit_callback *ecb,
549 int prefix, const char *data, int size)
551 const char *endp = NULL;
552 static const char *nneof = " No newline at end of file\n";
553 const char *old = diff_get_color(ecb->color_diff, DIFF_FILE_OLD);
554 const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
556 while (0 < size) {
557 int len;
559 endp = memchr(data, '\n', size);
560 len = endp ? (endp - data + 1) : size;
561 if (prefix != '+') {
562 ecb->lno_in_preimage++;
563 emit_line_0(ecb->opt, old, reset, '-',
564 data, len);
565 } else {
566 ecb->lno_in_postimage++;
567 emit_add_line(reset, ecb, data, len);
569 size -= len;
570 data += len;
572 if (!endp) {
573 const char *plain = diff_get_color(ecb->color_diff,
574 DIFF_PLAIN);
575 emit_line_0(ecb->opt, plain, reset, '\\',
576 nneof, strlen(nneof));
580 static void emit_rewrite_diff(const char *name_a,
581 const char *name_b,
582 struct diff_filespec *one,
583 struct diff_filespec *two,
584 struct userdiff_driver *textconv_one,
585 struct userdiff_driver *textconv_two,
586 struct diff_options *o)
588 int lc_a, lc_b;
589 const char *name_a_tab, *name_b_tab;
590 const char *metainfo = diff_get_color(o->use_color, DIFF_METAINFO);
591 const char *fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
592 const char *reset = diff_get_color(o->use_color, DIFF_RESET);
593 static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
594 const char *a_prefix, *b_prefix;
595 char *data_one, *data_two;
596 size_t size_one, size_two;
597 struct emit_callback ecbdata;
598 char *line_prefix = "";
599 struct strbuf *msgbuf;
601 if (o && o->output_prefix) {
602 msgbuf = o->output_prefix(o, o->output_prefix_data);
603 line_prefix = msgbuf->buf;
606 if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
607 a_prefix = o->b_prefix;
608 b_prefix = o->a_prefix;
609 } else {
610 a_prefix = o->a_prefix;
611 b_prefix = o->b_prefix;
614 name_a += (*name_a == '/');
615 name_b += (*name_b == '/');
616 name_a_tab = strchr(name_a, ' ') ? "\t" : "";
617 name_b_tab = strchr(name_b, ' ') ? "\t" : "";
619 strbuf_reset(&a_name);
620 strbuf_reset(&b_name);
621 quote_two_c_style(&a_name, a_prefix, name_a, 0);
622 quote_two_c_style(&b_name, b_prefix, name_b, 0);
624 size_one = fill_textconv(textconv_one, one, &data_one);
625 size_two = fill_textconv(textconv_two, two, &data_two);
627 memset(&ecbdata, 0, sizeof(ecbdata));
628 ecbdata.color_diff = want_color(o->use_color);
629 ecbdata.found_changesp = &o->found_changes;
630 ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
631 ecbdata.opt = o;
632 if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
633 mmfile_t mf1, mf2;
634 mf1.ptr = (char *)data_one;
635 mf2.ptr = (char *)data_two;
636 mf1.size = size_one;
637 mf2.size = size_two;
638 check_blank_at_eof(&mf1, &mf2, &ecbdata);
640 ecbdata.lno_in_preimage = 1;
641 ecbdata.lno_in_postimage = 1;
643 lc_a = count_lines(data_one, size_one);
644 lc_b = count_lines(data_two, size_two);
645 fprintf(o->file,
646 "%s%s--- %s%s%s\n%s%s+++ %s%s%s\n%s%s@@ -",
647 line_prefix, metainfo, a_name.buf, name_a_tab, reset,
648 line_prefix, metainfo, b_name.buf, name_b_tab, reset,
649 line_prefix, fraginfo);
650 if (!o->irreversible_delete)
651 print_line_count(o->file, lc_a);
652 else
653 fprintf(o->file, "?,?");
654 fprintf(o->file, " +");
655 print_line_count(o->file, lc_b);
656 fprintf(o->file, " @@%s\n", reset);
657 if (lc_a && !o->irreversible_delete)
658 emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
659 if (lc_b)
660 emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
661 if (textconv_one)
662 free((char *)data_one);
663 if (textconv_two)
664 free((char *)data_two);
667 struct diff_words_buffer {
668 mmfile_t text;
669 long alloc;
670 struct diff_words_orig {
671 const char *begin, *end;
672 } *orig;
673 int orig_nr, orig_alloc;
676 static void diff_words_append(char *line, unsigned long len,
677 struct diff_words_buffer *buffer)
679 ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
680 line++;
681 len--;
682 memcpy(buffer->text.ptr + buffer->text.size, line, len);
683 buffer->text.size += len;
684 buffer->text.ptr[buffer->text.size] = '\0';
687 struct diff_words_style_elem {
688 const char *prefix;
689 const char *suffix;
690 const char *color; /* NULL; filled in by the setup code if
691 * color is enabled */
694 struct diff_words_style {
695 enum diff_words_type type;
696 struct diff_words_style_elem new, old, ctx;
697 const char *newline;
700 static struct diff_words_style diff_words_styles[] = {
701 { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
702 { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
703 { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
706 struct diff_words_data {
707 struct diff_words_buffer minus, plus;
708 const char *current_plus;
709 int last_minus;
710 struct diff_options *opt;
711 regex_t *word_regex;
712 enum diff_words_type type;
713 struct diff_words_style *style;
716 static int fn_out_diff_words_write_helper(FILE *fp,
717 struct diff_words_style_elem *st_el,
718 const char *newline,
719 size_t count, const char *buf,
720 const char *line_prefix)
722 int print = 0;
724 while (count) {
725 char *p = memchr(buf, '\n', count);
726 if (print)
727 fputs(line_prefix, fp);
728 if (p != buf) {
729 if (st_el->color && fputs(st_el->color, fp) < 0)
730 return -1;
731 if (fputs(st_el->prefix, fp) < 0 ||
732 fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
733 fputs(st_el->suffix, fp) < 0)
734 return -1;
735 if (st_el->color && *st_el->color
736 && fputs(GIT_COLOR_RESET, fp) < 0)
737 return -1;
739 if (!p)
740 return 0;
741 if (fputs(newline, fp) < 0)
742 return -1;
743 count -= p + 1 - buf;
744 buf = p + 1;
745 print = 1;
747 return 0;
751 * '--color-words' algorithm can be described as:
753 * 1. collect a the minus/plus lines of a diff hunk, divided into
754 * minus-lines and plus-lines;
756 * 2. break both minus-lines and plus-lines into words and
757 * place them into two mmfile_t with one word for each line;
759 * 3. use xdiff to run diff on the two mmfile_t to get the words level diff;
761 * And for the common parts of the both file, we output the plus side text.
762 * diff_words->current_plus is used to trace the current position of the plus file
763 * which printed. diff_words->last_minus is used to trace the last minus word
764 * printed.
766 * For '--graph' to work with '--color-words', we need to output the graph prefix
767 * on each line of color words output. Generally, there are two conditions on
768 * which we should output the prefix.
770 * 1. diff_words->last_minus == 0 &&
771 * diff_words->current_plus == diff_words->plus.text.ptr
773 * that is: the plus text must start as a new line, and if there is no minus
774 * word printed, a graph prefix must be printed.
776 * 2. diff_words->current_plus > diff_words->plus.text.ptr &&
777 * *(diff_words->current_plus - 1) == '\n'
779 * that is: a graph prefix must be printed following a '\n'
781 static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
783 if ((diff_words->last_minus == 0 &&
784 diff_words->current_plus == diff_words->plus.text.ptr) ||
785 (diff_words->current_plus > diff_words->plus.text.ptr &&
786 *(diff_words->current_plus - 1) == '\n')) {
787 return 1;
788 } else {
789 return 0;
793 static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
795 struct diff_words_data *diff_words = priv;
796 struct diff_words_style *style = diff_words->style;
797 int minus_first, minus_len, plus_first, plus_len;
798 const char *minus_begin, *minus_end, *plus_begin, *plus_end;
799 struct diff_options *opt = diff_words->opt;
800 struct strbuf *msgbuf;
801 char *line_prefix = "";
803 if (line[0] != '@' || parse_hunk_header(line, len,
804 &minus_first, &minus_len, &plus_first, &plus_len))
805 return;
807 assert(opt);
808 if (opt->output_prefix) {
809 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
810 line_prefix = msgbuf->buf;
813 /* POSIX requires that first be decremented by one if len == 0... */
814 if (minus_len) {
815 minus_begin = diff_words->minus.orig[minus_first].begin;
816 minus_end =
817 diff_words->minus.orig[minus_first + minus_len - 1].end;
818 } else
819 minus_begin = minus_end =
820 diff_words->minus.orig[minus_first].end;
822 if (plus_len) {
823 plus_begin = diff_words->plus.orig[plus_first].begin;
824 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
825 } else
826 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
828 if (color_words_output_graph_prefix(diff_words)) {
829 fputs(line_prefix, diff_words->opt->file);
831 if (diff_words->current_plus != plus_begin) {
832 fn_out_diff_words_write_helper(diff_words->opt->file,
833 &style->ctx, style->newline,
834 plus_begin - diff_words->current_plus,
835 diff_words->current_plus, line_prefix);
836 if (*(plus_begin - 1) == '\n')
837 fputs(line_prefix, diff_words->opt->file);
839 if (minus_begin != minus_end) {
840 fn_out_diff_words_write_helper(diff_words->opt->file,
841 &style->old, style->newline,
842 minus_end - minus_begin, minus_begin,
843 line_prefix);
845 if (plus_begin != plus_end) {
846 fn_out_diff_words_write_helper(diff_words->opt->file,
847 &style->new, style->newline,
848 plus_end - plus_begin, plus_begin,
849 line_prefix);
852 diff_words->current_plus = plus_end;
853 diff_words->last_minus = minus_first;
856 /* This function starts looking at *begin, and returns 0 iff a word was found. */
857 static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
858 int *begin, int *end)
860 if (word_regex && *begin < buffer->size) {
861 regmatch_t match[1];
862 if (!regexec(word_regex, buffer->ptr + *begin, 1, match, 0)) {
863 char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
864 '\n', match[0].rm_eo - match[0].rm_so);
865 *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
866 *begin += match[0].rm_so;
867 return *begin >= *end;
869 return -1;
872 /* find the next word */
873 while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
874 (*begin)++;
875 if (*begin >= buffer->size)
876 return -1;
878 /* find the end of the word */
879 *end = *begin + 1;
880 while (*end < buffer->size && !isspace(buffer->ptr[*end]))
881 (*end)++;
883 return 0;
887 * This function splits the words in buffer->text, stores the list with
888 * newline separator into out, and saves the offsets of the original words
889 * in buffer->orig.
891 static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
892 regex_t *word_regex)
894 int i, j;
895 long alloc = 0;
897 out->size = 0;
898 out->ptr = NULL;
900 /* fake an empty "0th" word */
901 ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
902 buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
903 buffer->orig_nr = 1;
905 for (i = 0; i < buffer->text.size; i++) {
906 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
907 return;
909 /* store original boundaries */
910 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
911 buffer->orig_alloc);
912 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
913 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
914 buffer->orig_nr++;
916 /* store one word */
917 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
918 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
919 out->ptr[out->size + j - i] = '\n';
920 out->size += j - i + 1;
922 i = j - 1;
926 /* this executes the word diff on the accumulated buffers */
927 static void diff_words_show(struct diff_words_data *diff_words)
929 xpparam_t xpp;
930 xdemitconf_t xecfg;
931 mmfile_t minus, plus;
932 struct diff_words_style *style = diff_words->style;
934 struct diff_options *opt = diff_words->opt;
935 struct strbuf *msgbuf;
936 char *line_prefix = "";
938 assert(opt);
939 if (opt->output_prefix) {
940 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
941 line_prefix = msgbuf->buf;
944 /* special case: only removal */
945 if (!diff_words->plus.text.size) {
946 fputs(line_prefix, diff_words->opt->file);
947 fn_out_diff_words_write_helper(diff_words->opt->file,
948 &style->old, style->newline,
949 diff_words->minus.text.size,
950 diff_words->minus.text.ptr, line_prefix);
951 diff_words->minus.text.size = 0;
952 return;
955 diff_words->current_plus = diff_words->plus.text.ptr;
956 diff_words->last_minus = 0;
958 memset(&xpp, 0, sizeof(xpp));
959 memset(&xecfg, 0, sizeof(xecfg));
960 diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
961 diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
962 xpp.flags = 0;
963 /* as only the hunk header will be parsed, we need a 0-context */
964 xecfg.ctxlen = 0;
965 xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
966 &xpp, &xecfg);
967 free(minus.ptr);
968 free(plus.ptr);
969 if (diff_words->current_plus != diff_words->plus.text.ptr +
970 diff_words->plus.text.size) {
971 if (color_words_output_graph_prefix(diff_words))
972 fputs(line_prefix, diff_words->opt->file);
973 fn_out_diff_words_write_helper(diff_words->opt->file,
974 &style->ctx, style->newline,
975 diff_words->plus.text.ptr + diff_words->plus.text.size
976 - diff_words->current_plus, diff_words->current_plus,
977 line_prefix);
979 diff_words->minus.text.size = diff_words->plus.text.size = 0;
982 /* In "color-words" mode, show word-diff of words accumulated in the buffer */
983 static void diff_words_flush(struct emit_callback *ecbdata)
985 if (ecbdata->diff_words->minus.text.size ||
986 ecbdata->diff_words->plus.text.size)
987 diff_words_show(ecbdata->diff_words);
990 static void free_diff_words_data(struct emit_callback *ecbdata)
992 if (ecbdata->diff_words) {
993 diff_words_flush(ecbdata);
994 free (ecbdata->diff_words->minus.text.ptr);
995 free (ecbdata->diff_words->minus.orig);
996 free (ecbdata->diff_words->plus.text.ptr);
997 free (ecbdata->diff_words->plus.orig);
998 if (ecbdata->diff_words->word_regex) {
999 regfree(ecbdata->diff_words->word_regex);
1000 free(ecbdata->diff_words->word_regex);
1002 free(ecbdata->diff_words);
1003 ecbdata->diff_words = NULL;
1007 const char *diff_get_color(int diff_use_color, enum color_diff ix)
1009 if (want_color(diff_use_color))
1010 return diff_colors[ix];
1011 return "";
1014 static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1016 const char *cp;
1017 unsigned long allot;
1018 size_t l = len;
1020 if (ecb->truncate)
1021 return ecb->truncate(line, len);
1022 cp = line;
1023 allot = l;
1024 while (0 < l) {
1025 (void) utf8_width(&cp, &l);
1026 if (!cp)
1027 break; /* truncated in the middle? */
1029 return allot - l;
1032 static void find_lno(const char *line, struct emit_callback *ecbdata)
1034 const char *p;
1035 ecbdata->lno_in_preimage = 0;
1036 ecbdata->lno_in_postimage = 0;
1037 p = strchr(line, '-');
1038 if (!p)
1039 return; /* cannot happen */
1040 ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1041 p = strchr(p, '+');
1042 if (!p)
1043 return; /* cannot happen */
1044 ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1047 static void fn_out_consume(void *priv, char *line, unsigned long len)
1049 struct emit_callback *ecbdata = priv;
1050 const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
1051 const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
1052 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1053 struct diff_options *o = ecbdata->opt;
1054 char *line_prefix = "";
1055 struct strbuf *msgbuf;
1057 if (o && o->output_prefix) {
1058 msgbuf = o->output_prefix(o, o->output_prefix_data);
1059 line_prefix = msgbuf->buf;
1062 if (ecbdata->header) {
1063 fprintf(ecbdata->opt->file, "%s", ecbdata->header->buf);
1064 strbuf_reset(ecbdata->header);
1065 ecbdata->header = NULL;
1067 *(ecbdata->found_changesp) = 1;
1069 if (ecbdata->label_path[0]) {
1070 const char *name_a_tab, *name_b_tab;
1072 name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
1073 name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
1075 fprintf(ecbdata->opt->file, "%s%s--- %s%s%s\n",
1076 line_prefix, meta, ecbdata->label_path[0], reset, name_a_tab);
1077 fprintf(ecbdata->opt->file, "%s%s+++ %s%s%s\n",
1078 line_prefix, meta, ecbdata->label_path[1], reset, name_b_tab);
1079 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1082 if (diff_suppress_blank_empty
1083 && len == 2 && line[0] == ' ' && line[1] == '\n') {
1084 line[0] = '\n';
1085 len = 1;
1088 if (line[0] == '@') {
1089 if (ecbdata->diff_words)
1090 diff_words_flush(ecbdata);
1091 len = sane_truncate_line(ecbdata, line, len);
1092 find_lno(line, ecbdata);
1093 emit_hunk_header(ecbdata, line, len);
1094 if (line[len-1] != '\n')
1095 putc('\n', ecbdata->opt->file);
1096 return;
1099 if (len < 1) {
1100 emit_line(ecbdata->opt, reset, reset, line, len);
1101 if (ecbdata->diff_words
1102 && ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN)
1103 fputs("~\n", ecbdata->opt->file);
1104 return;
1107 if (ecbdata->diff_words) {
1108 if (line[0] == '-') {
1109 diff_words_append(line, len,
1110 &ecbdata->diff_words->minus);
1111 return;
1112 } else if (line[0] == '+') {
1113 diff_words_append(line, len,
1114 &ecbdata->diff_words->plus);
1115 return;
1116 } else if (!prefixcmp(line, "\\ ")) {
1118 * Eat the "no newline at eof" marker as if we
1119 * saw a "+" or "-" line with nothing on it,
1120 * and return without diff_words_flush() to
1121 * defer processing. If this is the end of
1122 * preimage, more "+" lines may come after it.
1124 return;
1126 diff_words_flush(ecbdata);
1127 if (ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN) {
1128 emit_line(ecbdata->opt, plain, reset, line, len);
1129 fputs("~\n", ecbdata->opt->file);
1130 } else {
1132 * Skip the prefix character, if any. With
1133 * diff_suppress_blank_empty, there may be
1134 * none.
1136 if (line[0] != '\n') {
1137 line++;
1138 len--;
1140 emit_line(ecbdata->opt, plain, reset, line, len);
1142 return;
1145 if (line[0] != '+') {
1146 const char *color =
1147 diff_get_color(ecbdata->color_diff,
1148 line[0] == '-' ? DIFF_FILE_OLD : DIFF_PLAIN);
1149 ecbdata->lno_in_preimage++;
1150 if (line[0] == ' ')
1151 ecbdata->lno_in_postimage++;
1152 emit_line(ecbdata->opt, color, reset, line, len);
1153 } else {
1154 ecbdata->lno_in_postimage++;
1155 emit_add_line(reset, ecbdata, line + 1, len - 1);
1159 static char *pprint_rename(const char *a, const char *b)
1161 const char *old = a;
1162 const char *new = b;
1163 struct strbuf name = STRBUF_INIT;
1164 int pfx_length, sfx_length;
1165 int len_a = strlen(a);
1166 int len_b = strlen(b);
1167 int a_midlen, b_midlen;
1168 int qlen_a = quote_c_style(a, NULL, NULL, 0);
1169 int qlen_b = quote_c_style(b, NULL, NULL, 0);
1171 if (qlen_a || qlen_b) {
1172 quote_c_style(a, &name, NULL, 0);
1173 strbuf_addstr(&name, " => ");
1174 quote_c_style(b, &name, NULL, 0);
1175 return strbuf_detach(&name, NULL);
1178 /* Find common prefix */
1179 pfx_length = 0;
1180 while (*old && *new && *old == *new) {
1181 if (*old == '/')
1182 pfx_length = old - a + 1;
1183 old++;
1184 new++;
1187 /* Find common suffix */
1188 old = a + len_a;
1189 new = b + len_b;
1190 sfx_length = 0;
1191 while (a <= old && b <= new && *old == *new) {
1192 if (*old == '/')
1193 sfx_length = len_a - (old - a);
1194 old--;
1195 new--;
1199 * pfx{mid-a => mid-b}sfx
1200 * {pfx-a => pfx-b}sfx
1201 * pfx{sfx-a => sfx-b}
1202 * name-a => name-b
1204 a_midlen = len_a - pfx_length - sfx_length;
1205 b_midlen = len_b - pfx_length - sfx_length;
1206 if (a_midlen < 0)
1207 a_midlen = 0;
1208 if (b_midlen < 0)
1209 b_midlen = 0;
1211 strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1212 if (pfx_length + sfx_length) {
1213 strbuf_add(&name, a, pfx_length);
1214 strbuf_addch(&name, '{');
1216 strbuf_add(&name, a + pfx_length, a_midlen);
1217 strbuf_addstr(&name, " => ");
1218 strbuf_add(&name, b + pfx_length, b_midlen);
1219 if (pfx_length + sfx_length) {
1220 strbuf_addch(&name, '}');
1221 strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1223 return strbuf_detach(&name, NULL);
1226 struct diffstat_t {
1227 int nr;
1228 int alloc;
1229 struct diffstat_file {
1230 char *from_name;
1231 char *name;
1232 char *print_name;
1233 unsigned is_unmerged:1;
1234 unsigned is_binary:1;
1235 unsigned is_renamed:1;
1236 uintmax_t added, deleted;
1237 } **files;
1240 static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1241 const char *name_a,
1242 const char *name_b)
1244 struct diffstat_file *x;
1245 x = xcalloc(sizeof (*x), 1);
1246 if (diffstat->nr == diffstat->alloc) {
1247 diffstat->alloc = alloc_nr(diffstat->alloc);
1248 diffstat->files = xrealloc(diffstat->files,
1249 diffstat->alloc * sizeof(x));
1251 diffstat->files[diffstat->nr++] = x;
1252 if (name_b) {
1253 x->from_name = xstrdup(name_a);
1254 x->name = xstrdup(name_b);
1255 x->is_renamed = 1;
1257 else {
1258 x->from_name = NULL;
1259 x->name = xstrdup(name_a);
1261 return x;
1264 static void diffstat_consume(void *priv, char *line, unsigned long len)
1266 struct diffstat_t *diffstat = priv;
1267 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1269 if (line[0] == '+')
1270 x->added++;
1271 else if (line[0] == '-')
1272 x->deleted++;
1275 const char mime_boundary_leader[] = "------------";
1277 static int scale_linear(int it, int width, int max_change)
1279 if (!it)
1280 return 0;
1282 * make sure that at least one '-' or '+' is printed if
1283 * there is any change to this path. The easiest way is to
1284 * scale linearly as if the alloted width is one column shorter
1285 * than it is, and then add 1 to the result.
1287 return 1 + (it * (width - 1) / max_change);
1290 static void show_name(FILE *file,
1291 const char *prefix, const char *name, int len)
1293 fprintf(file, " %s%-*s |", prefix, len, name);
1296 static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1298 if (cnt <= 0)
1299 return;
1300 fprintf(file, "%s", set);
1301 while (cnt--)
1302 putc(ch, file);
1303 fprintf(file, "%s", reset);
1306 static void fill_print_name(struct diffstat_file *file)
1308 char *pname;
1310 if (file->print_name)
1311 return;
1313 if (!file->is_renamed) {
1314 struct strbuf buf = STRBUF_INIT;
1315 if (quote_c_style(file->name, &buf, NULL, 0)) {
1316 pname = strbuf_detach(&buf, NULL);
1317 } else {
1318 pname = file->name;
1319 strbuf_release(&buf);
1321 } else {
1322 pname = pprint_rename(file->from_name, file->name);
1324 file->print_name = pname;
1327 static void show_stats(struct diffstat_t *data, struct diff_options *options)
1329 int i, len, add, del, adds = 0, dels = 0;
1330 uintmax_t max_change = 0, max_len = 0;
1331 int total_files = data->nr;
1332 int width, name_width, count;
1333 const char *reset, *add_c, *del_c;
1334 const char *line_prefix = "";
1335 int extra_shown = 0;
1336 struct strbuf *msg = NULL;
1338 if (data->nr == 0)
1339 return;
1341 if (options->output_prefix) {
1342 msg = options->output_prefix(options, options->output_prefix_data);
1343 line_prefix = msg->buf;
1346 width = options->stat_width ? options->stat_width : 80;
1347 name_width = options->stat_name_width ? options->stat_name_width : 50;
1348 count = options->stat_count ? options->stat_count : data->nr;
1350 /* Sanity: give at least 5 columns to the graph,
1351 * but leave at least 10 columns for the name.
1353 if (width < 25)
1354 width = 25;
1355 if (name_width < 10)
1356 name_width = 10;
1357 else if (width < name_width + 15)
1358 name_width = width - 15;
1360 /* Find the longest filename and max number of changes */
1361 reset = diff_get_color_opt(options, DIFF_RESET);
1362 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1363 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1365 for (i = 0; (i < count) && (i < data->nr); i++) {
1366 struct diffstat_file *file = data->files[i];
1367 uintmax_t change = file->added + file->deleted;
1368 if (!data->files[i]->is_renamed &&
1369 (change == 0)) {
1370 count++; /* not shown == room for one more */
1371 continue;
1373 fill_print_name(file);
1374 len = strlen(file->print_name);
1375 if (max_len < len)
1376 max_len = len;
1378 if (file->is_binary || file->is_unmerged)
1379 continue;
1380 if (max_change < change)
1381 max_change = change;
1383 count = i; /* min(count, data->nr) */
1385 /* Compute the width of the graph part;
1386 * 10 is for one blank at the beginning of the line plus
1387 * " | count " between the name and the graph.
1389 * From here on, name_width is the width of the name area,
1390 * and width is the width of the graph area.
1392 name_width = (name_width < max_len) ? name_width : max_len;
1393 if (width < (name_width + 10) + max_change)
1394 width = width - (name_width + 10);
1395 else
1396 width = max_change;
1398 for (i = 0; i < count; i++) {
1399 const char *prefix = "";
1400 char *name = data->files[i]->print_name;
1401 uintmax_t added = data->files[i]->added;
1402 uintmax_t deleted = data->files[i]->deleted;
1403 int name_len;
1405 if (!data->files[i]->is_renamed &&
1406 (added + deleted == 0)) {
1407 total_files--;
1408 continue;
1411 * "scale" the filename
1413 len = name_width;
1414 name_len = strlen(name);
1415 if (name_width < name_len) {
1416 char *slash;
1417 prefix = "...";
1418 len -= 3;
1419 name += name_len - len;
1420 slash = strchr(name, '/');
1421 if (slash)
1422 name = slash;
1425 if (data->files[i]->is_binary) {
1426 fprintf(options->file, "%s", line_prefix);
1427 show_name(options->file, prefix, name, len);
1428 fprintf(options->file, " Bin ");
1429 fprintf(options->file, "%s%"PRIuMAX"%s",
1430 del_c, deleted, reset);
1431 fprintf(options->file, " -> ");
1432 fprintf(options->file, "%s%"PRIuMAX"%s",
1433 add_c, added, reset);
1434 fprintf(options->file, " bytes");
1435 fprintf(options->file, "\n");
1436 continue;
1438 else if (data->files[i]->is_unmerged) {
1439 fprintf(options->file, "%s", line_prefix);
1440 show_name(options->file, prefix, name, len);
1441 fprintf(options->file, " Unmerged\n");
1442 continue;
1446 * scale the add/delete
1448 add = added;
1449 del = deleted;
1450 adds += add;
1451 dels += del;
1453 if (width <= max_change) {
1454 int total = add + del;
1456 total = scale_linear(add + del, width, max_change);
1457 if (total < 2 && add && del)
1458 /* width >= 2 due to the sanity check */
1459 total = 2;
1460 if (add < del) {
1461 add = scale_linear(add, width, max_change);
1462 del = total - add;
1463 } else {
1464 del = scale_linear(del, width, max_change);
1465 add = total - del;
1468 fprintf(options->file, "%s", line_prefix);
1469 show_name(options->file, prefix, name, len);
1470 fprintf(options->file, "%5"PRIuMAX"%s", added + deleted,
1471 added + deleted ? " " : "");
1472 show_graph(options->file, '+', add, add_c, reset);
1473 show_graph(options->file, '-', del, del_c, reset);
1474 fprintf(options->file, "\n");
1476 for (i = count; i < data->nr; i++) {
1477 uintmax_t added = data->files[i]->added;
1478 uintmax_t deleted = data->files[i]->deleted;
1479 if (!data->files[i]->is_renamed &&
1480 (added + deleted == 0)) {
1481 total_files--;
1482 continue;
1484 adds += added;
1485 dels += deleted;
1486 if (!extra_shown)
1487 fprintf(options->file, "%s ...\n", line_prefix);
1488 extra_shown = 1;
1490 fprintf(options->file, "%s", line_prefix);
1491 fprintf(options->file,
1492 " %d files changed, %d insertions(+), %d deletions(-)\n",
1493 total_files, adds, dels);
1496 static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1498 int i, adds = 0, dels = 0, total_files = data->nr;
1500 if (data->nr == 0)
1501 return;
1503 for (i = 0; i < data->nr; i++) {
1504 if (!data->files[i]->is_binary &&
1505 !data->files[i]->is_unmerged) {
1506 int added = data->files[i]->added;
1507 int deleted= data->files[i]->deleted;
1508 if (!data->files[i]->is_renamed &&
1509 (added + deleted == 0)) {
1510 total_files--;
1511 } else {
1512 adds += added;
1513 dels += deleted;
1517 if (options->output_prefix) {
1518 struct strbuf *msg = NULL;
1519 msg = options->output_prefix(options,
1520 options->output_prefix_data);
1521 fprintf(options->file, "%s", msg->buf);
1523 fprintf(options->file, " %d files changed, %d insertions(+), %d deletions(-)\n",
1524 total_files, adds, dels);
1527 static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1529 int i;
1531 if (data->nr == 0)
1532 return;
1534 for (i = 0; i < data->nr; i++) {
1535 struct diffstat_file *file = data->files[i];
1537 if (options->output_prefix) {
1538 struct strbuf *msg = NULL;
1539 msg = options->output_prefix(options,
1540 options->output_prefix_data);
1541 fprintf(options->file, "%s", msg->buf);
1544 if (file->is_binary)
1545 fprintf(options->file, "-\t-\t");
1546 else
1547 fprintf(options->file,
1548 "%"PRIuMAX"\t%"PRIuMAX"\t",
1549 file->added, file->deleted);
1550 if (options->line_termination) {
1551 fill_print_name(file);
1552 if (!file->is_renamed)
1553 write_name_quoted(file->name, options->file,
1554 options->line_termination);
1555 else {
1556 fputs(file->print_name, options->file);
1557 putc(options->line_termination, options->file);
1559 } else {
1560 if (file->is_renamed) {
1561 putc('\0', options->file);
1562 write_name_quoted(file->from_name, options->file, '\0');
1564 write_name_quoted(file->name, options->file, '\0');
1569 struct dirstat_file {
1570 const char *name;
1571 unsigned long changed;
1574 struct dirstat_dir {
1575 struct dirstat_file *files;
1576 int alloc, nr, permille, cumulative;
1579 static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1580 unsigned long changed, const char *base, int baselen)
1582 unsigned long this_dir = 0;
1583 unsigned int sources = 0;
1584 const char *line_prefix = "";
1585 struct strbuf *msg = NULL;
1587 if (opt->output_prefix) {
1588 msg = opt->output_prefix(opt, opt->output_prefix_data);
1589 line_prefix = msg->buf;
1592 while (dir->nr) {
1593 struct dirstat_file *f = dir->files;
1594 int namelen = strlen(f->name);
1595 unsigned long this;
1596 char *slash;
1598 if (namelen < baselen)
1599 break;
1600 if (memcmp(f->name, base, baselen))
1601 break;
1602 slash = strchr(f->name + baselen, '/');
1603 if (slash) {
1604 int newbaselen = slash + 1 - f->name;
1605 this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
1606 sources++;
1607 } else {
1608 this = f->changed;
1609 dir->files++;
1610 dir->nr--;
1611 sources += 2;
1613 this_dir += this;
1617 * We don't report dirstat's for
1618 * - the top level
1619 * - or cases where everything came from a single directory
1620 * under this directory (sources == 1).
1622 if (baselen && sources != 1) {
1623 if (this_dir) {
1624 int permille = this_dir * 1000 / changed;
1625 if (permille >= dir->permille) {
1626 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
1627 permille / 10, permille % 10, baselen, base);
1628 if (!dir->cumulative)
1629 return 0;
1633 return this_dir;
1636 static int dirstat_compare(const void *_a, const void *_b)
1638 const struct dirstat_file *a = _a;
1639 const struct dirstat_file *b = _b;
1640 return strcmp(a->name, b->name);
1643 static void show_dirstat(struct diff_options *options)
1645 int i;
1646 unsigned long changed;
1647 struct dirstat_dir dir;
1648 struct diff_queue_struct *q = &diff_queued_diff;
1650 dir.files = NULL;
1651 dir.alloc = 0;
1652 dir.nr = 0;
1653 dir.permille = options->dirstat_permille;
1654 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1656 changed = 0;
1657 for (i = 0; i < q->nr; i++) {
1658 struct diff_filepair *p = q->queue[i];
1659 const char *name;
1660 unsigned long copied, added, damage;
1661 int content_changed;
1663 name = p->two->path ? p->two->path : p->one->path;
1665 if (p->one->sha1_valid && p->two->sha1_valid)
1666 content_changed = hashcmp(p->one->sha1, p->two->sha1);
1667 else
1668 content_changed = 1;
1670 if (!content_changed) {
1672 * The SHA1 has not changed, so pre-/post-content is
1673 * identical. We can therefore skip looking at the
1674 * file contents altogether.
1676 damage = 0;
1677 goto found_damage;
1680 if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
1682 * In --dirstat-by-file mode, we don't really need to
1683 * look at the actual file contents at all.
1684 * The fact that the SHA1 changed is enough for us to
1685 * add this file to the list of results
1686 * (with each file contributing equal damage).
1688 damage = 1;
1689 goto found_damage;
1692 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
1693 diff_populate_filespec(p->one, 0);
1694 diff_populate_filespec(p->two, 0);
1695 diffcore_count_changes(p->one, p->two, NULL, NULL, 0,
1696 &copied, &added);
1697 diff_free_filespec_data(p->one);
1698 diff_free_filespec_data(p->two);
1699 } else if (DIFF_FILE_VALID(p->one)) {
1700 diff_populate_filespec(p->one, 1);
1701 copied = added = 0;
1702 diff_free_filespec_data(p->one);
1703 } else if (DIFF_FILE_VALID(p->two)) {
1704 diff_populate_filespec(p->two, 1);
1705 copied = 0;
1706 added = p->two->size;
1707 diff_free_filespec_data(p->two);
1708 } else
1709 continue;
1712 * Original minus copied is the removed material,
1713 * added is the new material. They are both damages
1714 * made to the preimage.
1715 * If the resulting damage is zero, we know that
1716 * diffcore_count_changes() considers the two entries to
1717 * be identical, but since content_changed is true, we
1718 * know that there must have been _some_ kind of change,
1719 * so we force all entries to have damage > 0.
1721 damage = (p->one->size - copied) + added;
1722 if (!damage)
1723 damage = 1;
1725 found_damage:
1726 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1727 dir.files[dir.nr].name = name;
1728 dir.files[dir.nr].changed = damage;
1729 changed += damage;
1730 dir.nr++;
1733 /* This can happen even with many files, if everything was renames */
1734 if (!changed)
1735 return;
1737 /* Show all directories with more than x% of the changes */
1738 qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1739 gather_dirstat(options, &dir, changed, "", 0);
1742 static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
1744 int i;
1745 unsigned long changed;
1746 struct dirstat_dir dir;
1748 if (data->nr == 0)
1749 return;
1751 dir.files = NULL;
1752 dir.alloc = 0;
1753 dir.nr = 0;
1754 dir.permille = options->dirstat_permille;
1755 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1757 changed = 0;
1758 for (i = 0; i < data->nr; i++) {
1759 struct diffstat_file *file = data->files[i];
1760 unsigned long damage = file->added + file->deleted;
1761 if (file->is_binary)
1763 * binary files counts bytes, not lines. Must find some
1764 * way to normalize binary bytes vs. textual lines.
1765 * The following heuristic assumes that there are 64
1766 * bytes per "line".
1767 * This is stupid and ugly, but very cheap...
1769 damage = (damage + 63) / 64;
1770 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1771 dir.files[dir.nr].name = file->name;
1772 dir.files[dir.nr].changed = damage;
1773 changed += damage;
1774 dir.nr++;
1777 /* This can happen even with many files, if everything was renames */
1778 if (!changed)
1779 return;
1781 /* Show all directories with more than x% of the changes */
1782 qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1783 gather_dirstat(options, &dir, changed, "", 0);
1786 static void free_diffstat_info(struct diffstat_t *diffstat)
1788 int i;
1789 for (i = 0; i < diffstat->nr; i++) {
1790 struct diffstat_file *f = diffstat->files[i];
1791 if (f->name != f->print_name)
1792 free(f->print_name);
1793 free(f->name);
1794 free(f->from_name);
1795 free(f);
1797 free(diffstat->files);
1800 struct checkdiff_t {
1801 const char *filename;
1802 int lineno;
1803 int conflict_marker_size;
1804 struct diff_options *o;
1805 unsigned ws_rule;
1806 unsigned status;
1809 static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
1811 char firstchar;
1812 int cnt;
1814 if (len < marker_size + 1)
1815 return 0;
1816 firstchar = line[0];
1817 switch (firstchar) {
1818 case '=': case '>': case '<': case '|':
1819 break;
1820 default:
1821 return 0;
1823 for (cnt = 1; cnt < marker_size; cnt++)
1824 if (line[cnt] != firstchar)
1825 return 0;
1826 /* line[1] thru line[marker_size-1] are same as firstchar */
1827 if (len < marker_size + 1 || !isspace(line[marker_size]))
1828 return 0;
1829 return 1;
1832 static void checkdiff_consume(void *priv, char *line, unsigned long len)
1834 struct checkdiff_t *data = priv;
1835 int marker_size = data->conflict_marker_size;
1836 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
1837 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
1838 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
1839 char *err;
1840 char *line_prefix = "";
1841 struct strbuf *msgbuf;
1843 assert(data->o);
1844 if (data->o->output_prefix) {
1845 msgbuf = data->o->output_prefix(data->o,
1846 data->o->output_prefix_data);
1847 line_prefix = msgbuf->buf;
1850 if (line[0] == '+') {
1851 unsigned bad;
1852 data->lineno++;
1853 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
1854 data->status |= 1;
1855 fprintf(data->o->file,
1856 "%s%s:%d: leftover conflict marker\n",
1857 line_prefix, data->filename, data->lineno);
1859 bad = ws_check(line + 1, len - 1, data->ws_rule);
1860 if (!bad)
1861 return;
1862 data->status |= bad;
1863 err = whitespace_error_string(bad);
1864 fprintf(data->o->file, "%s%s:%d: %s.\n",
1865 line_prefix, data->filename, data->lineno, err);
1866 free(err);
1867 emit_line(data->o, set, reset, line, 1);
1868 ws_check_emit(line + 1, len - 1, data->ws_rule,
1869 data->o->file, set, reset, ws);
1870 } else if (line[0] == ' ') {
1871 data->lineno++;
1872 } else if (line[0] == '@') {
1873 char *plus = strchr(line, '+');
1874 if (plus)
1875 data->lineno = strtol(plus, NULL, 10) - 1;
1876 else
1877 die("invalid diff");
1881 static unsigned char *deflate_it(char *data,
1882 unsigned long size,
1883 unsigned long *result_size)
1885 int bound;
1886 unsigned char *deflated;
1887 git_zstream stream;
1889 memset(&stream, 0, sizeof(stream));
1890 git_deflate_init(&stream, zlib_compression_level);
1891 bound = git_deflate_bound(&stream, size);
1892 deflated = xmalloc(bound);
1893 stream.next_out = deflated;
1894 stream.avail_out = bound;
1896 stream.next_in = (unsigned char *)data;
1897 stream.avail_in = size;
1898 while (git_deflate(&stream, Z_FINISH) == Z_OK)
1899 ; /* nothing */
1900 git_deflate_end(&stream);
1901 *result_size = stream.total_out;
1902 return deflated;
1905 static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two, char *prefix)
1907 void *cp;
1908 void *delta;
1909 void *deflated;
1910 void *data;
1911 unsigned long orig_size;
1912 unsigned long delta_size;
1913 unsigned long deflate_size;
1914 unsigned long data_size;
1916 /* We could do deflated delta, or we could do just deflated two,
1917 * whichever is smaller.
1919 delta = NULL;
1920 deflated = deflate_it(two->ptr, two->size, &deflate_size);
1921 if (one->size && two->size) {
1922 delta = diff_delta(one->ptr, one->size,
1923 two->ptr, two->size,
1924 &delta_size, deflate_size);
1925 if (delta) {
1926 void *to_free = delta;
1927 orig_size = delta_size;
1928 delta = deflate_it(delta, delta_size, &delta_size);
1929 free(to_free);
1933 if (delta && delta_size < deflate_size) {
1934 fprintf(file, "%sdelta %lu\n", prefix, orig_size);
1935 free(deflated);
1936 data = delta;
1937 data_size = delta_size;
1939 else {
1940 fprintf(file, "%sliteral %lu\n", prefix, two->size);
1941 free(delta);
1942 data = deflated;
1943 data_size = deflate_size;
1946 /* emit data encoded in base85 */
1947 cp = data;
1948 while (data_size) {
1949 int bytes = (52 < data_size) ? 52 : data_size;
1950 char line[70];
1951 data_size -= bytes;
1952 if (bytes <= 26)
1953 line[0] = bytes + 'A' - 1;
1954 else
1955 line[0] = bytes - 26 + 'a' - 1;
1956 encode_85(line + 1, cp, bytes);
1957 cp = (char *) cp + bytes;
1958 fprintf(file, "%s", prefix);
1959 fputs(line, file);
1960 fputc('\n', file);
1962 fprintf(file, "%s\n", prefix);
1963 free(data);
1966 static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two, char *prefix)
1968 fprintf(file, "%sGIT binary patch\n", prefix);
1969 emit_binary_diff_body(file, one, two, prefix);
1970 emit_binary_diff_body(file, two, one, prefix);
1973 static void diff_filespec_load_driver(struct diff_filespec *one)
1975 /* Use already-loaded driver */
1976 if (one->driver)
1977 return;
1979 if (S_ISREG(one->mode))
1980 one->driver = userdiff_find_by_path(one->path);
1982 /* Fallback to default settings */
1983 if (!one->driver)
1984 one->driver = userdiff_find_by_name("default");
1987 int diff_filespec_is_binary(struct diff_filespec *one)
1989 if (one->is_binary == -1) {
1990 diff_filespec_load_driver(one);
1991 if (one->driver->binary != -1)
1992 one->is_binary = one->driver->binary;
1993 else {
1994 if (!one->data && DIFF_FILE_VALID(one))
1995 diff_populate_filespec(one, 0);
1996 if (one->data)
1997 one->is_binary = buffer_is_binary(one->data,
1998 one->size);
1999 if (one->is_binary == -1)
2000 one->is_binary = 0;
2003 return one->is_binary;
2006 static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2008 diff_filespec_load_driver(one);
2009 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2012 static const char *userdiff_word_regex(struct diff_filespec *one)
2014 diff_filespec_load_driver(one);
2015 return one->driver->word_regex;
2018 void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2020 if (!options->a_prefix)
2021 options->a_prefix = a;
2022 if (!options->b_prefix)
2023 options->b_prefix = b;
2026 struct userdiff_driver *get_textconv(struct diff_filespec *one)
2028 if (!DIFF_FILE_VALID(one))
2029 return NULL;
2031 diff_filespec_load_driver(one);
2032 return userdiff_get_textconv(one->driver);
2035 static void builtin_diff(const char *name_a,
2036 const char *name_b,
2037 struct diff_filespec *one,
2038 struct diff_filespec *two,
2039 const char *xfrm_msg,
2040 int must_show_header,
2041 struct diff_options *o,
2042 int complete_rewrite)
2044 mmfile_t mf1, mf2;
2045 const char *lbl[2];
2046 char *a_one, *b_two;
2047 const char *set = diff_get_color_opt(o, DIFF_METAINFO);
2048 const char *reset = diff_get_color_opt(o, DIFF_RESET);
2049 const char *a_prefix, *b_prefix;
2050 struct userdiff_driver *textconv_one = NULL;
2051 struct userdiff_driver *textconv_two = NULL;
2052 struct strbuf header = STRBUF_INIT;
2053 struct strbuf *msgbuf;
2054 char *line_prefix = "";
2056 if (o->output_prefix) {
2057 msgbuf = o->output_prefix(o, o->output_prefix_data);
2058 line_prefix = msgbuf->buf;
2061 if (DIFF_OPT_TST(o, SUBMODULE_LOG) &&
2062 (!one->mode || S_ISGITLINK(one->mode)) &&
2063 (!two->mode || S_ISGITLINK(two->mode))) {
2064 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2065 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2066 show_submodule_summary(o->file, one ? one->path : two->path,
2067 one->sha1, two->sha1, two->dirty_submodule,
2068 del, add, reset);
2069 return;
2072 if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2073 textconv_one = get_textconv(one);
2074 textconv_two = get_textconv(two);
2077 diff_set_mnemonic_prefix(o, "a/", "b/");
2078 if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2079 a_prefix = o->b_prefix;
2080 b_prefix = o->a_prefix;
2081 } else {
2082 a_prefix = o->a_prefix;
2083 b_prefix = o->b_prefix;
2086 /* Never use a non-valid filename anywhere if at all possible */
2087 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2088 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2090 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2091 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2092 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2093 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2094 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, set, a_one, b_two, reset);
2095 if (lbl[0][0] == '/') {
2096 /* /dev/null */
2097 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, set, two->mode, reset);
2098 if (xfrm_msg)
2099 strbuf_addstr(&header, xfrm_msg);
2100 must_show_header = 1;
2102 else if (lbl[1][0] == '/') {
2103 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, set, one->mode, reset);
2104 if (xfrm_msg)
2105 strbuf_addstr(&header, xfrm_msg);
2106 must_show_header = 1;
2108 else {
2109 if (one->mode != two->mode) {
2110 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, set, one->mode, reset);
2111 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, set, two->mode, reset);
2112 must_show_header = 1;
2114 if (xfrm_msg)
2115 strbuf_addstr(&header, xfrm_msg);
2118 * we do not run diff between different kind
2119 * of objects.
2121 if ((one->mode ^ two->mode) & S_IFMT)
2122 goto free_ab_and_return;
2123 if (complete_rewrite &&
2124 (textconv_one || !diff_filespec_is_binary(one)) &&
2125 (textconv_two || !diff_filespec_is_binary(two))) {
2126 fprintf(o->file, "%s", header.buf);
2127 strbuf_reset(&header);
2128 emit_rewrite_diff(name_a, name_b, one, two,
2129 textconv_one, textconv_two, o);
2130 o->found_changes = 1;
2131 goto free_ab_and_return;
2135 if (o->irreversible_delete && lbl[1][0] == '/') {
2136 fprintf(o->file, "%s", header.buf);
2137 strbuf_reset(&header);
2138 goto free_ab_and_return;
2139 } else if (!DIFF_OPT_TST(o, TEXT) &&
2140 ( (!textconv_one && diff_filespec_is_binary(one)) ||
2141 (!textconv_two && diff_filespec_is_binary(two)) )) {
2142 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2143 die("unable to read files to diff");
2144 /* Quite common confusing case */
2145 if (mf1.size == mf2.size &&
2146 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2147 if (must_show_header)
2148 fprintf(o->file, "%s", header.buf);
2149 goto free_ab_and_return;
2151 fprintf(o->file, "%s", header.buf);
2152 strbuf_reset(&header);
2153 if (DIFF_OPT_TST(o, BINARY))
2154 emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2155 else
2156 fprintf(o->file, "%sBinary files %s and %s differ\n",
2157 line_prefix, lbl[0], lbl[1]);
2158 o->found_changes = 1;
2159 } else {
2160 /* Crazy xdl interfaces.. */
2161 const char *diffopts = getenv("GIT_DIFF_OPTS");
2162 xpparam_t xpp;
2163 xdemitconf_t xecfg;
2164 struct emit_callback ecbdata;
2165 const struct userdiff_funcname *pe;
2167 if (!DIFF_XDL_TST(o, WHITESPACE_FLAGS) || must_show_header) {
2168 fprintf(o->file, "%s", header.buf);
2169 strbuf_reset(&header);
2172 mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2173 mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2175 pe = diff_funcname_pattern(one);
2176 if (!pe)
2177 pe = diff_funcname_pattern(two);
2179 memset(&xpp, 0, sizeof(xpp));
2180 memset(&xecfg, 0, sizeof(xecfg));
2181 memset(&ecbdata, 0, sizeof(ecbdata));
2182 ecbdata.label_path = lbl;
2183 ecbdata.color_diff = want_color(o->use_color);
2184 ecbdata.found_changesp = &o->found_changes;
2185 ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
2186 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2187 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2188 ecbdata.opt = o;
2189 ecbdata.header = header.len ? &header : NULL;
2190 xpp.flags = o->xdl_opts;
2191 xecfg.ctxlen = o->context;
2192 xecfg.interhunkctxlen = o->interhunkcontext;
2193 xecfg.flags = XDL_EMIT_FUNCNAMES;
2194 if (DIFF_OPT_TST(o, FUNCCONTEXT))
2195 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2196 if (pe)
2197 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2198 if (!diffopts)
2200 else if (!prefixcmp(diffopts, "--unified="))
2201 xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
2202 else if (!prefixcmp(diffopts, "-u"))
2203 xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
2204 if (o->word_diff) {
2205 int i;
2207 ecbdata.diff_words =
2208 xcalloc(1, sizeof(struct diff_words_data));
2209 ecbdata.diff_words->type = o->word_diff;
2210 ecbdata.diff_words->opt = o;
2211 if (!o->word_regex)
2212 o->word_regex = userdiff_word_regex(one);
2213 if (!o->word_regex)
2214 o->word_regex = userdiff_word_regex(two);
2215 if (!o->word_regex)
2216 o->word_regex = diff_word_regex_cfg;
2217 if (o->word_regex) {
2218 ecbdata.diff_words->word_regex = (regex_t *)
2219 xmalloc(sizeof(regex_t));
2220 if (regcomp(ecbdata.diff_words->word_regex,
2221 o->word_regex,
2222 REG_EXTENDED | REG_NEWLINE))
2223 die ("Invalid regular expression: %s",
2224 o->word_regex);
2226 for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
2227 if (o->word_diff == diff_words_styles[i].type) {
2228 ecbdata.diff_words->style =
2229 &diff_words_styles[i];
2230 break;
2233 if (want_color(o->use_color)) {
2234 struct diff_words_style *st = ecbdata.diff_words->style;
2235 st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
2236 st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
2237 st->ctx.color = diff_get_color_opt(o, DIFF_PLAIN);
2240 xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2241 &xpp, &xecfg);
2242 if (o->word_diff)
2243 free_diff_words_data(&ecbdata);
2244 if (textconv_one)
2245 free(mf1.ptr);
2246 if (textconv_two)
2247 free(mf2.ptr);
2248 xdiff_clear_find_func(&xecfg);
2251 free_ab_and_return:
2252 strbuf_release(&header);
2253 diff_free_filespec_data(one);
2254 diff_free_filespec_data(two);
2255 free(a_one);
2256 free(b_two);
2257 return;
2260 static void builtin_diffstat(const char *name_a, const char *name_b,
2261 struct diff_filespec *one,
2262 struct diff_filespec *two,
2263 struct diffstat_t *diffstat,
2264 struct diff_options *o,
2265 int complete_rewrite)
2267 mmfile_t mf1, mf2;
2268 struct diffstat_file *data;
2270 data = diffstat_add(diffstat, name_a, name_b);
2272 if (!one || !two) {
2273 data->is_unmerged = 1;
2274 return;
2277 if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2278 data->is_binary = 1;
2279 data->added = diff_filespec_size(two);
2280 data->deleted = diff_filespec_size(one);
2283 else if (complete_rewrite) {
2284 diff_populate_filespec(one, 0);
2285 diff_populate_filespec(two, 0);
2286 data->deleted = count_lines(one->data, one->size);
2287 data->added = count_lines(two->data, two->size);
2290 else {
2291 /* Crazy xdl interfaces.. */
2292 xpparam_t xpp;
2293 xdemitconf_t xecfg;
2295 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2296 die("unable to read files to diff");
2298 memset(&xpp, 0, sizeof(xpp));
2299 memset(&xecfg, 0, sizeof(xecfg));
2300 xpp.flags = o->xdl_opts;
2301 xecfg.ctxlen = o->context;
2302 xecfg.interhunkctxlen = o->interhunkcontext;
2303 xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2304 &xpp, &xecfg);
2307 diff_free_filespec_data(one);
2308 diff_free_filespec_data(two);
2311 static void builtin_checkdiff(const char *name_a, const char *name_b,
2312 const char *attr_path,
2313 struct diff_filespec *one,
2314 struct diff_filespec *two,
2315 struct diff_options *o)
2317 mmfile_t mf1, mf2;
2318 struct checkdiff_t data;
2320 if (!two)
2321 return;
2323 memset(&data, 0, sizeof(data));
2324 data.filename = name_b ? name_b : name_a;
2325 data.lineno = 0;
2326 data.o = o;
2327 data.ws_rule = whitespace_rule(attr_path);
2328 data.conflict_marker_size = ll_merge_marker_size(attr_path);
2330 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2331 die("unable to read files to diff");
2334 * All the other codepaths check both sides, but not checking
2335 * the "old" side here is deliberate. We are checking the newly
2336 * introduced changes, and as long as the "new" side is text, we
2337 * can and should check what it introduces.
2339 if (diff_filespec_is_binary(two))
2340 goto free_and_return;
2341 else {
2342 /* Crazy xdl interfaces.. */
2343 xpparam_t xpp;
2344 xdemitconf_t xecfg;
2346 memset(&xpp, 0, sizeof(xpp));
2347 memset(&xecfg, 0, sizeof(xecfg));
2348 xecfg.ctxlen = 1; /* at least one context line */
2349 xpp.flags = 0;
2350 xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2351 &xpp, &xecfg);
2353 if (data.ws_rule & WS_BLANK_AT_EOF) {
2354 struct emit_callback ecbdata;
2355 int blank_at_eof;
2357 ecbdata.ws_rule = data.ws_rule;
2358 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2359 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2361 if (blank_at_eof) {
2362 static char *err;
2363 if (!err)
2364 err = whitespace_error_string(WS_BLANK_AT_EOF);
2365 fprintf(o->file, "%s:%d: %s.\n",
2366 data.filename, blank_at_eof, err);
2367 data.status = 1; /* report errors */
2371 free_and_return:
2372 diff_free_filespec_data(one);
2373 diff_free_filespec_data(two);
2374 if (data.status)
2375 DIFF_OPT_SET(o, CHECK_FAILED);
2378 struct diff_filespec *alloc_filespec(const char *path)
2380 int namelen = strlen(path);
2381 struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
2383 memset(spec, 0, sizeof(*spec));
2384 spec->path = (char *)(spec + 1);
2385 memcpy(spec->path, path, namelen+1);
2386 spec->count = 1;
2387 spec->is_binary = -1;
2388 return spec;
2391 void free_filespec(struct diff_filespec *spec)
2393 if (!--spec->count) {
2394 diff_free_filespec_data(spec);
2395 free(spec);
2399 void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
2400 unsigned short mode)
2402 if (mode) {
2403 spec->mode = canon_mode(mode);
2404 hashcpy(spec->sha1, sha1);
2405 spec->sha1_valid = !is_null_sha1(sha1);
2410 * Given a name and sha1 pair, if the index tells us the file in
2411 * the work tree has that object contents, return true, so that
2412 * prepare_temp_file() does not have to inflate and extract.
2414 static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
2416 struct cache_entry *ce;
2417 struct stat st;
2418 int pos, len;
2421 * We do not read the cache ourselves here, because the
2422 * benchmark with my previous version that always reads cache
2423 * shows that it makes things worse for diff-tree comparing
2424 * two linux-2.6 kernel trees in an already checked out work
2425 * tree. This is because most diff-tree comparisons deal with
2426 * only a small number of files, while reading the cache is
2427 * expensive for a large project, and its cost outweighs the
2428 * savings we get by not inflating the object to a temporary
2429 * file. Practically, this code only helps when we are used
2430 * by diff-cache --cached, which does read the cache before
2431 * calling us.
2433 if (!active_cache)
2434 return 0;
2436 /* We want to avoid the working directory if our caller
2437 * doesn't need the data in a normal file, this system
2438 * is rather slow with its stat/open/mmap/close syscalls,
2439 * and the object is contained in a pack file. The pack
2440 * is probably already open and will be faster to obtain
2441 * the data through than the working directory. Loose
2442 * objects however would tend to be slower as they need
2443 * to be individually opened and inflated.
2445 if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1))
2446 return 0;
2448 len = strlen(name);
2449 pos = cache_name_pos(name, len);
2450 if (pos < 0)
2451 return 0;
2452 ce = active_cache[pos];
2455 * This is not the sha1 we are looking for, or
2456 * unreusable because it is not a regular file.
2458 if (hashcmp(sha1, ce->sha1) || !S_ISREG(ce->ce_mode))
2459 return 0;
2462 * If ce is marked as "assume unchanged", there is no
2463 * guarantee that work tree matches what we are looking for.
2465 if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2466 return 0;
2469 * If ce matches the file in the work tree, we can reuse it.
2471 if (ce_uptodate(ce) ||
2472 (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2473 return 1;
2475 return 0;
2478 static int populate_from_stdin(struct diff_filespec *s)
2480 struct strbuf buf = STRBUF_INIT;
2481 size_t size = 0;
2483 if (strbuf_read(&buf, 0, 0) < 0)
2484 return error("error while reading from stdin %s",
2485 strerror(errno));
2487 s->should_munmap = 0;
2488 s->data = strbuf_detach(&buf, &size);
2489 s->size = size;
2490 s->should_free = 1;
2491 return 0;
2494 static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2496 int len;
2497 char *data = xmalloc(100), *dirty = "";
2499 /* Are we looking at the work tree? */
2500 if (s->dirty_submodule)
2501 dirty = "-dirty";
2503 len = snprintf(data, 100,
2504 "Subproject commit %s%s\n", sha1_to_hex(s->sha1), dirty);
2505 s->data = data;
2506 s->size = len;
2507 s->should_free = 1;
2508 if (size_only) {
2509 s->data = NULL;
2510 free(data);
2512 return 0;
2516 * While doing rename detection and pickaxe operation, we may need to
2517 * grab the data for the blob (or file) for our own in-core comparison.
2518 * diff_filespec has data and size fields for this purpose.
2520 int diff_populate_filespec(struct diff_filespec *s, int size_only)
2522 int err = 0;
2523 if (!DIFF_FILE_VALID(s))
2524 die("internal error: asking to populate invalid file.");
2525 if (S_ISDIR(s->mode))
2526 return -1;
2528 if (s->data)
2529 return 0;
2531 if (size_only && 0 < s->size)
2532 return 0;
2534 if (S_ISGITLINK(s->mode))
2535 return diff_populate_gitlink(s, size_only);
2537 if (!s->sha1_valid ||
2538 reuse_worktree_file(s->path, s->sha1, 0)) {
2539 struct strbuf buf = STRBUF_INIT;
2540 struct stat st;
2541 int fd;
2543 if (!strcmp(s->path, "-"))
2544 return populate_from_stdin(s);
2546 if (lstat(s->path, &st) < 0) {
2547 if (errno == ENOENT) {
2548 err_empty:
2549 err = -1;
2550 empty:
2551 s->data = (char *)"";
2552 s->size = 0;
2553 return err;
2556 s->size = xsize_t(st.st_size);
2557 if (!s->size)
2558 goto empty;
2559 if (S_ISLNK(st.st_mode)) {
2560 struct strbuf sb = STRBUF_INIT;
2562 if (strbuf_readlink(&sb, s->path, s->size))
2563 goto err_empty;
2564 s->size = sb.len;
2565 s->data = strbuf_detach(&sb, NULL);
2566 s->should_free = 1;
2567 return 0;
2569 if (size_only)
2570 return 0;
2571 fd = open(s->path, O_RDONLY);
2572 if (fd < 0)
2573 goto err_empty;
2574 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2575 close(fd);
2576 s->should_munmap = 1;
2579 * Convert from working tree format to canonical git format
2581 if (convert_to_git(s->path, s->data, s->size, &buf, safe_crlf)) {
2582 size_t size = 0;
2583 munmap(s->data, s->size);
2584 s->should_munmap = 0;
2585 s->data = strbuf_detach(&buf, &size);
2586 s->size = size;
2587 s->should_free = 1;
2590 else {
2591 enum object_type type;
2592 if (size_only) {
2593 type = sha1_object_info(s->sha1, &s->size);
2594 if (type < 0)
2595 die("unable to read %s", sha1_to_hex(s->sha1));
2596 } else {
2597 s->data = read_sha1_file(s->sha1, &type, &s->size);
2598 if (!s->data)
2599 die("unable to read %s", sha1_to_hex(s->sha1));
2600 s->should_free = 1;
2603 return 0;
2606 void diff_free_filespec_blob(struct diff_filespec *s)
2608 if (s->should_free)
2609 free(s->data);
2610 else if (s->should_munmap)
2611 munmap(s->data, s->size);
2613 if (s->should_free || s->should_munmap) {
2614 s->should_free = s->should_munmap = 0;
2615 s->data = NULL;
2619 void diff_free_filespec_data(struct diff_filespec *s)
2621 diff_free_filespec_blob(s);
2622 free(s->cnt_data);
2623 s->cnt_data = NULL;
2626 static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2627 void *blob,
2628 unsigned long size,
2629 const unsigned char *sha1,
2630 int mode)
2632 int fd;
2633 struct strbuf buf = STRBUF_INIT;
2634 struct strbuf template = STRBUF_INIT;
2635 char *path_dup = xstrdup(path);
2636 const char *base = basename(path_dup);
2638 /* Generate "XXXXXX_basename.ext" */
2639 strbuf_addstr(&template, "XXXXXX_");
2640 strbuf_addstr(&template, base);
2642 fd = git_mkstemps(temp->tmp_path, PATH_MAX, template.buf,
2643 strlen(base) + 1);
2644 if (fd < 0)
2645 die_errno("unable to create temp-file");
2646 if (convert_to_working_tree(path,
2647 (const char *)blob, (size_t)size, &buf)) {
2648 blob = buf.buf;
2649 size = buf.len;
2651 if (write_in_full(fd, blob, size) != size)
2652 die_errno("unable to write temp-file");
2653 close(fd);
2654 temp->name = temp->tmp_path;
2655 strcpy(temp->hex, sha1_to_hex(sha1));
2656 temp->hex[40] = 0;
2657 sprintf(temp->mode, "%06o", mode);
2658 strbuf_release(&buf);
2659 strbuf_release(&template);
2660 free(path_dup);
2663 static struct diff_tempfile *prepare_temp_file(const char *name,
2664 struct diff_filespec *one)
2666 struct diff_tempfile *temp = claim_diff_tempfile();
2668 if (!DIFF_FILE_VALID(one)) {
2669 not_a_valid_file:
2670 /* A '-' entry produces this for file-2, and
2671 * a '+' entry produces this for file-1.
2673 temp->name = "/dev/null";
2674 strcpy(temp->hex, ".");
2675 strcpy(temp->mode, ".");
2676 return temp;
2679 if (!remove_tempfile_installed) {
2680 atexit(remove_tempfile);
2681 sigchain_push_common(remove_tempfile_on_signal);
2682 remove_tempfile_installed = 1;
2685 if (!one->sha1_valid ||
2686 reuse_worktree_file(name, one->sha1, 1)) {
2687 struct stat st;
2688 if (lstat(name, &st) < 0) {
2689 if (errno == ENOENT)
2690 goto not_a_valid_file;
2691 die_errno("stat(%s)", name);
2693 if (S_ISLNK(st.st_mode)) {
2694 struct strbuf sb = STRBUF_INIT;
2695 if (strbuf_readlink(&sb, name, st.st_size) < 0)
2696 die_errno("readlink(%s)", name);
2697 prep_temp_blob(name, temp, sb.buf, sb.len,
2698 (one->sha1_valid ?
2699 one->sha1 : null_sha1),
2700 (one->sha1_valid ?
2701 one->mode : S_IFLNK));
2702 strbuf_release(&sb);
2704 else {
2705 /* we can borrow from the file in the work tree */
2706 temp->name = name;
2707 if (!one->sha1_valid)
2708 strcpy(temp->hex, sha1_to_hex(null_sha1));
2709 else
2710 strcpy(temp->hex, sha1_to_hex(one->sha1));
2711 /* Even though we may sometimes borrow the
2712 * contents from the work tree, we always want
2713 * one->mode. mode is trustworthy even when
2714 * !(one->sha1_valid), as long as
2715 * DIFF_FILE_VALID(one).
2717 sprintf(temp->mode, "%06o", one->mode);
2719 return temp;
2721 else {
2722 if (diff_populate_filespec(one, 0))
2723 die("cannot read data blob for %s", one->path);
2724 prep_temp_blob(name, temp, one->data, one->size,
2725 one->sha1, one->mode);
2727 return temp;
2730 /* An external diff command takes:
2732 * diff-cmd name infile1 infile1-sha1 infile1-mode \
2733 * infile2 infile2-sha1 infile2-mode [ rename-to ]
2736 static void run_external_diff(const char *pgm,
2737 const char *name,
2738 const char *other,
2739 struct diff_filespec *one,
2740 struct diff_filespec *two,
2741 const char *xfrm_msg,
2742 int complete_rewrite)
2744 const char *spawn_arg[10];
2745 int retval;
2746 const char **arg = &spawn_arg[0];
2748 if (one && two) {
2749 struct diff_tempfile *temp_one, *temp_two;
2750 const char *othername = (other ? other : name);
2751 temp_one = prepare_temp_file(name, one);
2752 temp_two = prepare_temp_file(othername, two);
2753 *arg++ = pgm;
2754 *arg++ = name;
2755 *arg++ = temp_one->name;
2756 *arg++ = temp_one->hex;
2757 *arg++ = temp_one->mode;
2758 *arg++ = temp_two->name;
2759 *arg++ = temp_two->hex;
2760 *arg++ = temp_two->mode;
2761 if (other) {
2762 *arg++ = other;
2763 *arg++ = xfrm_msg;
2765 } else {
2766 *arg++ = pgm;
2767 *arg++ = name;
2769 *arg = NULL;
2770 fflush(NULL);
2771 retval = run_command_v_opt(spawn_arg, RUN_USING_SHELL);
2772 remove_tempfile();
2773 if (retval) {
2774 fprintf(stderr, "external diff died, stopping at %s.\n", name);
2775 exit(1);
2779 static int similarity_index(struct diff_filepair *p)
2781 return p->score * 100 / MAX_SCORE;
2784 static void fill_metainfo(struct strbuf *msg,
2785 const char *name,
2786 const char *other,
2787 struct diff_filespec *one,
2788 struct diff_filespec *two,
2789 struct diff_options *o,
2790 struct diff_filepair *p,
2791 int *must_show_header,
2792 int use_color)
2794 const char *set = diff_get_color(use_color, DIFF_METAINFO);
2795 const char *reset = diff_get_color(use_color, DIFF_RESET);
2796 struct strbuf *msgbuf;
2797 char *line_prefix = "";
2799 *must_show_header = 1;
2800 if (o->output_prefix) {
2801 msgbuf = o->output_prefix(o, o->output_prefix_data);
2802 line_prefix = msgbuf->buf;
2804 strbuf_init(msg, PATH_MAX * 2 + 300);
2805 switch (p->status) {
2806 case DIFF_STATUS_COPIED:
2807 strbuf_addf(msg, "%s%ssimilarity index %d%%",
2808 line_prefix, set, similarity_index(p));
2809 strbuf_addf(msg, "%s\n%s%scopy from ",
2810 reset, line_prefix, set);
2811 quote_c_style(name, msg, NULL, 0);
2812 strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
2813 quote_c_style(other, msg, NULL, 0);
2814 strbuf_addf(msg, "%s\n", reset);
2815 break;
2816 case DIFF_STATUS_RENAMED:
2817 strbuf_addf(msg, "%s%ssimilarity index %d%%",
2818 line_prefix, set, similarity_index(p));
2819 strbuf_addf(msg, "%s\n%s%srename from ",
2820 reset, line_prefix, set);
2821 quote_c_style(name, msg, NULL, 0);
2822 strbuf_addf(msg, "%s\n%s%srename to ",
2823 reset, line_prefix, set);
2824 quote_c_style(other, msg, NULL, 0);
2825 strbuf_addf(msg, "%s\n", reset);
2826 break;
2827 case DIFF_STATUS_MODIFIED:
2828 if (p->score) {
2829 strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
2830 line_prefix,
2831 set, similarity_index(p), reset);
2832 break;
2834 /* fallthru */
2835 default:
2836 *must_show_header = 0;
2838 if (one && two && hashcmp(one->sha1, two->sha1)) {
2839 int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
2841 if (DIFF_OPT_TST(o, BINARY)) {
2842 mmfile_t mf;
2843 if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
2844 (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
2845 abbrev = 40;
2847 strbuf_addf(msg, "%s%sindex %s..", line_prefix, set,
2848 find_unique_abbrev(one->sha1, abbrev));
2849 strbuf_addstr(msg, find_unique_abbrev(two->sha1, abbrev));
2850 if (one->mode == two->mode)
2851 strbuf_addf(msg, " %06o", one->mode);
2852 strbuf_addf(msg, "%s\n", reset);
2856 static void run_diff_cmd(const char *pgm,
2857 const char *name,
2858 const char *other,
2859 const char *attr_path,
2860 struct diff_filespec *one,
2861 struct diff_filespec *two,
2862 struct strbuf *msg,
2863 struct diff_options *o,
2864 struct diff_filepair *p)
2866 const char *xfrm_msg = NULL;
2867 int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
2868 int must_show_header = 0;
2870 if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
2871 pgm = NULL;
2872 else {
2873 struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
2874 if (drv && drv->external)
2875 pgm = drv->external;
2878 if (msg) {
2880 * don't use colors when the header is intended for an
2881 * external diff driver
2883 fill_metainfo(msg, name, other, one, two, o, p,
2884 &must_show_header,
2885 want_color(o->use_color) && !pgm);
2886 xfrm_msg = msg->len ? msg->buf : NULL;
2889 if (pgm) {
2890 run_external_diff(pgm, name, other, one, two, xfrm_msg,
2891 complete_rewrite);
2892 return;
2894 if (one && two)
2895 builtin_diff(name, other ? other : name,
2896 one, two, xfrm_msg, must_show_header,
2897 o, complete_rewrite);
2898 else
2899 fprintf(o->file, "* Unmerged path %s\n", name);
2902 static void diff_fill_sha1_info(struct diff_filespec *one)
2904 if (DIFF_FILE_VALID(one)) {
2905 if (!one->sha1_valid) {
2906 struct stat st;
2907 if (!strcmp(one->path, "-")) {
2908 hashcpy(one->sha1, null_sha1);
2909 return;
2911 if (lstat(one->path, &st) < 0)
2912 die_errno("stat '%s'", one->path);
2913 if (index_path(one->sha1, one->path, &st, 0))
2914 die("cannot hash %s", one->path);
2917 else
2918 hashclr(one->sha1);
2921 static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
2923 /* Strip the prefix but do not molest /dev/null and absolute paths */
2924 if (*namep && **namep != '/') {
2925 *namep += prefix_length;
2926 if (**namep == '/')
2927 ++*namep;
2929 if (*otherp && **otherp != '/') {
2930 *otherp += prefix_length;
2931 if (**otherp == '/')
2932 ++*otherp;
2936 static void run_diff(struct diff_filepair *p, struct diff_options *o)
2938 const char *pgm = external_diff();
2939 struct strbuf msg;
2940 struct diff_filespec *one = p->one;
2941 struct diff_filespec *two = p->two;
2942 const char *name;
2943 const char *other;
2944 const char *attr_path;
2946 name = p->one->path;
2947 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2948 attr_path = name;
2949 if (o->prefix_length)
2950 strip_prefix(o->prefix_length, &name, &other);
2952 if (DIFF_PAIR_UNMERGED(p)) {
2953 run_diff_cmd(pgm, name, NULL, attr_path,
2954 NULL, NULL, NULL, o, p);
2955 return;
2958 diff_fill_sha1_info(one);
2959 diff_fill_sha1_info(two);
2961 if (!pgm &&
2962 DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
2963 (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
2965 * a filepair that changes between file and symlink
2966 * needs to be split into deletion and creation.
2968 struct diff_filespec *null = alloc_filespec(two->path);
2969 run_diff_cmd(NULL, name, other, attr_path,
2970 one, null, &msg, o, p);
2971 free(null);
2972 strbuf_release(&msg);
2974 null = alloc_filespec(one->path);
2975 run_diff_cmd(NULL, name, other, attr_path,
2976 null, two, &msg, o, p);
2977 free(null);
2979 else
2980 run_diff_cmd(pgm, name, other, attr_path,
2981 one, two, &msg, o, p);
2983 strbuf_release(&msg);
2986 static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
2987 struct diffstat_t *diffstat)
2989 const char *name;
2990 const char *other;
2991 int complete_rewrite = 0;
2993 if (DIFF_PAIR_UNMERGED(p)) {
2994 /* unmerged */
2995 builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, 0);
2996 return;
2999 name = p->one->path;
3000 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3002 if (o->prefix_length)
3003 strip_prefix(o->prefix_length, &name, &other);
3005 diff_fill_sha1_info(p->one);
3006 diff_fill_sha1_info(p->two);
3008 if (p->status == DIFF_STATUS_MODIFIED && p->score)
3009 complete_rewrite = 1;
3010 builtin_diffstat(name, other, p->one, p->two, diffstat, o, complete_rewrite);
3013 static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3015 const char *name;
3016 const char *other;
3017 const char *attr_path;
3019 if (DIFF_PAIR_UNMERGED(p)) {
3020 /* unmerged */
3021 return;
3024 name = p->one->path;
3025 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3026 attr_path = other ? other : name;
3028 if (o->prefix_length)
3029 strip_prefix(o->prefix_length, &name, &other);
3031 diff_fill_sha1_info(p->one);
3032 diff_fill_sha1_info(p->two);
3034 builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3037 void diff_setup(struct diff_options *options)
3039 memcpy(options, &default_diff_options, sizeof(*options));
3041 options->file = stdout;
3043 options->line_termination = '\n';
3044 options->break_opt = -1;
3045 options->rename_limit = -1;
3046 options->dirstat_permille = diff_dirstat_permille_default;
3047 options->context = 3;
3049 options->change = diff_change;
3050 options->add_remove = diff_addremove;
3051 options->use_color = diff_use_color_default;
3052 options->detect_rename = diff_detect_rename_default;
3054 if (diff_no_prefix) {
3055 options->a_prefix = options->b_prefix = "";
3056 } else if (!diff_mnemonic_prefix) {
3057 options->a_prefix = "a/";
3058 options->b_prefix = "b/";
3062 int diff_setup_done(struct diff_options *options)
3064 int count = 0;
3066 if (options->output_format & DIFF_FORMAT_NAME)
3067 count++;
3068 if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3069 count++;
3070 if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3071 count++;
3072 if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3073 count++;
3074 if (count > 1)
3075 die("--name-only, --name-status, --check and -s are mutually exclusive");
3078 * Most of the time we can say "there are changes"
3079 * only by checking if there are changed paths, but
3080 * --ignore-whitespace* options force us to look
3081 * inside contents.
3084 if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3085 DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3086 DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3087 DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3088 else
3089 DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3091 if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3092 options->detect_rename = DIFF_DETECT_COPY;
3094 if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3095 options->prefix = NULL;
3096 if (options->prefix)
3097 options->prefix_length = strlen(options->prefix);
3098 else
3099 options->prefix_length = 0;
3101 if (options->output_format & (DIFF_FORMAT_NAME |
3102 DIFF_FORMAT_NAME_STATUS |
3103 DIFF_FORMAT_CHECKDIFF |
3104 DIFF_FORMAT_NO_OUTPUT))
3105 options->output_format &= ~(DIFF_FORMAT_RAW |
3106 DIFF_FORMAT_NUMSTAT |
3107 DIFF_FORMAT_DIFFSTAT |
3108 DIFF_FORMAT_SHORTSTAT |
3109 DIFF_FORMAT_DIRSTAT |
3110 DIFF_FORMAT_SUMMARY |
3111 DIFF_FORMAT_PATCH);
3114 * These cases always need recursive; we do not drop caller-supplied
3115 * recursive bits for other formats here.
3117 if (options->output_format & (DIFF_FORMAT_PATCH |
3118 DIFF_FORMAT_NUMSTAT |
3119 DIFF_FORMAT_DIFFSTAT |
3120 DIFF_FORMAT_SHORTSTAT |
3121 DIFF_FORMAT_DIRSTAT |
3122 DIFF_FORMAT_SUMMARY |
3123 DIFF_FORMAT_CHECKDIFF))
3124 DIFF_OPT_SET(options, RECURSIVE);
3126 * Also pickaxe would not work very well if you do not say recursive
3128 if (options->pickaxe)
3129 DIFF_OPT_SET(options, RECURSIVE);
3131 * When patches are generated, submodules diffed against the work tree
3132 * must be checked for dirtiness too so it can be shown in the output
3134 if (options->output_format & DIFF_FORMAT_PATCH)
3135 DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3137 if (options->detect_rename && options->rename_limit < 0)
3138 options->rename_limit = diff_rename_limit_default;
3139 if (options->setup & DIFF_SETUP_USE_CACHE) {
3140 if (!active_cache)
3141 /* read-cache does not die even when it fails
3142 * so it is safe for us to do this here. Also
3143 * it does not smudge active_cache or active_nr
3144 * when it fails, so we do not have to worry about
3145 * cleaning it up ourselves either.
3147 read_cache();
3149 if (options->abbrev <= 0 || 40 < options->abbrev)
3150 options->abbrev = 40; /* full */
3153 * It does not make sense to show the first hit we happened
3154 * to have found. It does not make sense not to return with
3155 * exit code in such a case either.
3157 if (DIFF_OPT_TST(options, QUICK)) {
3158 options->output_format = DIFF_FORMAT_NO_OUTPUT;
3159 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3162 return 0;
3165 static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3167 char c, *eq;
3168 int len;
3170 if (*arg != '-')
3171 return 0;
3172 c = *++arg;
3173 if (!c)
3174 return 0;
3175 if (c == arg_short) {
3176 c = *++arg;
3177 if (!c)
3178 return 1;
3179 if (val && isdigit(c)) {
3180 char *end;
3181 int n = strtoul(arg, &end, 10);
3182 if (*end)
3183 return 0;
3184 *val = n;
3185 return 1;
3187 return 0;
3189 if (c != '-')
3190 return 0;
3191 arg++;
3192 eq = strchr(arg, '=');
3193 if (eq)
3194 len = eq - arg;
3195 else
3196 len = strlen(arg);
3197 if (!len || strncmp(arg, arg_long, len))
3198 return 0;
3199 if (eq) {
3200 int n;
3201 char *end;
3202 if (!isdigit(*++eq))
3203 return 0;
3204 n = strtoul(eq, &end, 10);
3205 if (*end)
3206 return 0;
3207 *val = n;
3209 return 1;
3212 static int diff_scoreopt_parse(const char *opt);
3214 static inline int short_opt(char opt, const char **argv,
3215 const char **optarg)
3217 const char *arg = argv[0];
3218 if (arg[0] != '-' || arg[1] != opt)
3219 return 0;
3220 if (arg[2] != '\0') {
3221 *optarg = arg + 2;
3222 return 1;
3224 if (!argv[1])
3225 die("Option '%c' requires a value", opt);
3226 *optarg = argv[1];
3227 return 2;
3230 int parse_long_opt(const char *opt, const char **argv,
3231 const char **optarg)
3233 const char *arg = argv[0];
3234 if (arg[0] != '-' || arg[1] != '-')
3235 return 0;
3236 arg += strlen("--");
3237 if (prefixcmp(arg, opt))
3238 return 0;
3239 arg += strlen(opt);
3240 if (*arg == '=') { /* sticked form: --option=value */
3241 *optarg = arg + 1;
3242 return 1;
3244 if (*arg != '\0')
3245 return 0;
3246 /* separate form: --option value */
3247 if (!argv[1])
3248 die("Option '--%s' requires a value", opt);
3249 *optarg = argv[1];
3250 return 2;
3253 static int stat_opt(struct diff_options *options, const char **av)
3255 const char *arg = av[0];
3256 char *end;
3257 int width = options->stat_width;
3258 int name_width = options->stat_name_width;
3259 int count = options->stat_count;
3260 int argcount = 1;
3262 arg += strlen("--stat");
3263 end = (char *)arg;
3265 switch (*arg) {
3266 case '-':
3267 if (!prefixcmp(arg, "-width")) {
3268 arg += strlen("-width");
3269 if (*arg == '=')
3270 width = strtoul(arg + 1, &end, 10);
3271 else if (!*arg && !av[1])
3272 die("Option '--stat-width' requires a value");
3273 else if (!*arg) {
3274 width = strtoul(av[1], &end, 10);
3275 argcount = 2;
3277 } else if (!prefixcmp(arg, "-name-width")) {
3278 arg += strlen("-name-width");
3279 if (*arg == '=')
3280 name_width = strtoul(arg + 1, &end, 10);
3281 else if (!*arg && !av[1])
3282 die("Option '--stat-name-width' requires a value");
3283 else if (!*arg) {
3284 name_width = strtoul(av[1], &end, 10);
3285 argcount = 2;
3287 } else if (!prefixcmp(arg, "-count")) {
3288 arg += strlen("-count");
3289 if (*arg == '=')
3290 count = strtoul(arg + 1, &end, 10);
3291 else if (!*arg && !av[1])
3292 die("Option '--stat-count' requires a value");
3293 else if (!*arg) {
3294 count = strtoul(av[1], &end, 10);
3295 argcount = 2;
3298 break;
3299 case '=':
3300 width = strtoul(arg+1, &end, 10);
3301 if (*end == ',')
3302 name_width = strtoul(end+1, &end, 10);
3303 if (*end == ',')
3304 count = strtoul(end+1, &end, 10);
3307 /* Important! This checks all the error cases! */
3308 if (*end)
3309 return 0;
3310 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3311 options->stat_name_width = name_width;
3312 options->stat_width = width;
3313 options->stat_count = count;
3314 return argcount;
3317 static int parse_dirstat_opt(struct diff_options *options, const char *params)
3319 struct strbuf errmsg = STRBUF_INIT;
3320 if (parse_dirstat_params(options, params, &errmsg))
3321 die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3322 errmsg.buf);
3323 strbuf_release(&errmsg);
3325 * The caller knows a dirstat-related option is given from the command
3326 * line; allow it to say "return this_function();"
3328 options->output_format |= DIFF_FORMAT_DIRSTAT;
3329 return 1;
3332 int diff_opt_parse(struct diff_options *options, const char **av, int ac)
3334 const char *arg = av[0];
3335 const char *optarg;
3336 int argcount;
3338 /* Output format options */
3339 if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch"))
3340 options->output_format |= DIFF_FORMAT_PATCH;
3341 else if (opt_arg(arg, 'U', "unified", &options->context))
3342 options->output_format |= DIFF_FORMAT_PATCH;
3343 else if (!strcmp(arg, "--raw"))
3344 options->output_format |= DIFF_FORMAT_RAW;
3345 else if (!strcmp(arg, "--patch-with-raw"))
3346 options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_RAW;
3347 else if (!strcmp(arg, "--numstat"))
3348 options->output_format |= DIFF_FORMAT_NUMSTAT;
3349 else if (!strcmp(arg, "--shortstat"))
3350 options->output_format |= DIFF_FORMAT_SHORTSTAT;
3351 else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3352 return parse_dirstat_opt(options, "");
3353 else if (!prefixcmp(arg, "-X"))
3354 return parse_dirstat_opt(options, arg + 2);
3355 else if (!prefixcmp(arg, "--dirstat="))
3356 return parse_dirstat_opt(options, arg + 10);
3357 else if (!strcmp(arg, "--cumulative"))
3358 return parse_dirstat_opt(options, "cumulative");
3359 else if (!strcmp(arg, "--dirstat-by-file"))
3360 return parse_dirstat_opt(options, "files");
3361 else if (!prefixcmp(arg, "--dirstat-by-file=")) {
3362 parse_dirstat_opt(options, "files");
3363 return parse_dirstat_opt(options, arg + 18);
3365 else if (!strcmp(arg, "--check"))
3366 options->output_format |= DIFF_FORMAT_CHECKDIFF;
3367 else if (!strcmp(arg, "--summary"))
3368 options->output_format |= DIFF_FORMAT_SUMMARY;
3369 else if (!strcmp(arg, "--patch-with-stat"))
3370 options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_DIFFSTAT;
3371 else if (!strcmp(arg, "--name-only"))
3372 options->output_format |= DIFF_FORMAT_NAME;
3373 else if (!strcmp(arg, "--name-status"))
3374 options->output_format |= DIFF_FORMAT_NAME_STATUS;
3375 else if (!strcmp(arg, "-s"))
3376 options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3377 else if (!prefixcmp(arg, "--stat"))
3378 /* --stat, --stat-width, --stat-name-width, or --stat-count */
3379 return stat_opt(options, av);
3381 /* renames options */
3382 else if (!prefixcmp(arg, "-B") || !prefixcmp(arg, "--break-rewrites=") ||
3383 !strcmp(arg, "--break-rewrites")) {
3384 if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3385 return error("invalid argument to -B: %s", arg+2);
3387 else if (!prefixcmp(arg, "-M") || !prefixcmp(arg, "--find-renames=") ||
3388 !strcmp(arg, "--find-renames")) {
3389 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3390 return error("invalid argument to -M: %s", arg+2);
3391 options->detect_rename = DIFF_DETECT_RENAME;
3393 else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3394 options->irreversible_delete = 1;
3396 else if (!prefixcmp(arg, "-C") || !prefixcmp(arg, "--find-copies=") ||
3397 !strcmp(arg, "--find-copies")) {
3398 if (options->detect_rename == DIFF_DETECT_COPY)
3399 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3400 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3401 return error("invalid argument to -C: %s", arg+2);
3402 options->detect_rename = DIFF_DETECT_COPY;
3404 else if (!strcmp(arg, "--no-renames"))
3405 options->detect_rename = 0;
3406 else if (!strcmp(arg, "--relative"))
3407 DIFF_OPT_SET(options, RELATIVE_NAME);
3408 else if (!prefixcmp(arg, "--relative=")) {
3409 DIFF_OPT_SET(options, RELATIVE_NAME);
3410 options->prefix = arg + 11;
3413 /* xdiff options */
3414 else if (!strcmp(arg, "--minimal"))
3415 DIFF_XDL_SET(options, NEED_MINIMAL);
3416 else if (!strcmp(arg, "--no-minimal"))
3417 DIFF_XDL_CLR(options, NEED_MINIMAL);
3418 else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3419 DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3420 else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3421 DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3422 else if (!strcmp(arg, "--ignore-space-at-eol"))
3423 DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3424 else if (!strcmp(arg, "--patience"))
3425 DIFF_XDL_SET(options, PATIENCE_DIFF);
3426 else if (!strcmp(arg, "--histogram"))
3427 DIFF_XDL_SET(options, HISTOGRAM_DIFF);
3429 /* flags options */
3430 else if (!strcmp(arg, "--binary")) {
3431 options->output_format |= DIFF_FORMAT_PATCH;
3432 DIFF_OPT_SET(options, BINARY);
3434 else if (!strcmp(arg, "--full-index"))
3435 DIFF_OPT_SET(options, FULL_INDEX);
3436 else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
3437 DIFF_OPT_SET(options, TEXT);
3438 else if (!strcmp(arg, "-R"))
3439 DIFF_OPT_SET(options, REVERSE_DIFF);
3440 else if (!strcmp(arg, "--find-copies-harder"))
3441 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3442 else if (!strcmp(arg, "--follow"))
3443 DIFF_OPT_SET(options, FOLLOW_RENAMES);
3444 else if (!strcmp(arg, "--color"))
3445 options->use_color = 1;
3446 else if (!prefixcmp(arg, "--color=")) {
3447 int value = git_config_colorbool(NULL, arg+8);
3448 if (value < 0)
3449 return error("option `color' expects \"always\", \"auto\", or \"never\"");
3450 options->use_color = value;
3452 else if (!strcmp(arg, "--no-color"))
3453 options->use_color = 0;
3454 else if (!strcmp(arg, "--color-words")) {
3455 options->use_color = 1;
3456 options->word_diff = DIFF_WORDS_COLOR;
3458 else if (!prefixcmp(arg, "--color-words=")) {
3459 options->use_color = 1;
3460 options->word_diff = DIFF_WORDS_COLOR;
3461 options->word_regex = arg + 14;
3463 else if (!strcmp(arg, "--word-diff")) {
3464 if (options->word_diff == DIFF_WORDS_NONE)
3465 options->word_diff = DIFF_WORDS_PLAIN;
3467 else if (!prefixcmp(arg, "--word-diff=")) {
3468 const char *type = arg + 12;
3469 if (!strcmp(type, "plain"))
3470 options->word_diff = DIFF_WORDS_PLAIN;
3471 else if (!strcmp(type, "color")) {
3472 options->use_color = 1;
3473 options->word_diff = DIFF_WORDS_COLOR;
3475 else if (!strcmp(type, "porcelain"))
3476 options->word_diff = DIFF_WORDS_PORCELAIN;
3477 else if (!strcmp(type, "none"))
3478 options->word_diff = DIFF_WORDS_NONE;
3479 else
3480 die("bad --word-diff argument: %s", type);
3482 else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
3483 if (options->word_diff == DIFF_WORDS_NONE)
3484 options->word_diff = DIFF_WORDS_PLAIN;
3485 options->word_regex = optarg;
3486 return argcount;
3488 else if (!strcmp(arg, "--exit-code"))
3489 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3490 else if (!strcmp(arg, "--quiet"))
3491 DIFF_OPT_SET(options, QUICK);
3492 else if (!strcmp(arg, "--ext-diff"))
3493 DIFF_OPT_SET(options, ALLOW_EXTERNAL);
3494 else if (!strcmp(arg, "--no-ext-diff"))
3495 DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
3496 else if (!strcmp(arg, "--textconv"))
3497 DIFF_OPT_SET(options, ALLOW_TEXTCONV);
3498 else if (!strcmp(arg, "--no-textconv"))
3499 DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
3500 else if (!strcmp(arg, "--ignore-submodules")) {
3501 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3502 handle_ignore_submodules_arg(options, "all");
3503 } else if (!prefixcmp(arg, "--ignore-submodules=")) {
3504 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3505 handle_ignore_submodules_arg(options, arg + 20);
3506 } else if (!strcmp(arg, "--submodule"))
3507 DIFF_OPT_SET(options, SUBMODULE_LOG);
3508 else if (!prefixcmp(arg, "--submodule=")) {
3509 if (!strcmp(arg + 12, "log"))
3510 DIFF_OPT_SET(options, SUBMODULE_LOG);
3513 /* misc options */
3514 else if (!strcmp(arg, "-z"))
3515 options->line_termination = 0;
3516 else if ((argcount = short_opt('l', av, &optarg))) {
3517 options->rename_limit = strtoul(optarg, NULL, 10);
3518 return argcount;
3520 else if ((argcount = short_opt('S', av, &optarg))) {
3521 options->pickaxe = optarg;
3522 options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
3523 return argcount;
3524 } else if ((argcount = short_opt('G', av, &optarg))) {
3525 options->pickaxe = optarg;
3526 options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
3527 return argcount;
3529 else if (!strcmp(arg, "--pickaxe-all"))
3530 options->pickaxe_opts |= DIFF_PICKAXE_ALL;
3531 else if (!strcmp(arg, "--pickaxe-regex"))
3532 options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
3533 else if ((argcount = short_opt('O', av, &optarg))) {
3534 options->orderfile = optarg;
3535 return argcount;
3537 else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
3538 options->filter = optarg;
3539 return argcount;
3541 else if (!strcmp(arg, "--abbrev"))
3542 options->abbrev = DEFAULT_ABBREV;
3543 else if (!prefixcmp(arg, "--abbrev=")) {
3544 options->abbrev = strtoul(arg + 9, NULL, 10);
3545 if (options->abbrev < MINIMUM_ABBREV)
3546 options->abbrev = MINIMUM_ABBREV;
3547 else if (40 < options->abbrev)
3548 options->abbrev = 40;
3550 else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
3551 options->a_prefix = optarg;
3552 return argcount;
3554 else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
3555 options->b_prefix = optarg;
3556 return argcount;
3558 else if (!strcmp(arg, "--no-prefix"))
3559 options->a_prefix = options->b_prefix = "";
3560 else if (opt_arg(arg, '\0', "inter-hunk-context",
3561 &options->interhunkcontext))
3563 else if (!strcmp(arg, "-W"))
3564 DIFF_OPT_SET(options, FUNCCONTEXT);
3565 else if (!strcmp(arg, "--function-context"))
3566 DIFF_OPT_SET(options, FUNCCONTEXT);
3567 else if (!strcmp(arg, "--no-function-context"))
3568 DIFF_OPT_CLR(options, FUNCCONTEXT);
3569 else if ((argcount = parse_long_opt("output", av, &optarg))) {
3570 options->file = fopen(optarg, "w");
3571 if (!options->file)
3572 die_errno("Could not open '%s'", optarg);
3573 options->close_file = 1;
3574 return argcount;
3575 } else
3576 return 0;
3577 return 1;
3580 int parse_rename_score(const char **cp_p)
3582 unsigned long num, scale;
3583 int ch, dot;
3584 const char *cp = *cp_p;
3586 num = 0;
3587 scale = 1;
3588 dot = 0;
3589 for (;;) {
3590 ch = *cp;
3591 if ( !dot && ch == '.' ) {
3592 scale = 1;
3593 dot = 1;
3594 } else if ( ch == '%' ) {
3595 scale = dot ? scale*100 : 100;
3596 cp++; /* % is always at the end */
3597 break;
3598 } else if ( ch >= '0' && ch <= '9' ) {
3599 if ( scale < 100000 ) {
3600 scale *= 10;
3601 num = (num*10) + (ch-'0');
3603 } else {
3604 break;
3606 cp++;
3608 *cp_p = cp;
3610 /* user says num divided by scale and we say internally that
3611 * is MAX_SCORE * num / scale.
3613 return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
3616 static int diff_scoreopt_parse(const char *opt)
3618 int opt1, opt2, cmd;
3620 if (*opt++ != '-')
3621 return -1;
3622 cmd = *opt++;
3623 if (cmd == '-') {
3624 /* convert the long-form arguments into short-form versions */
3625 if (!prefixcmp(opt, "break-rewrites")) {
3626 opt += strlen("break-rewrites");
3627 if (*opt == 0 || *opt++ == '=')
3628 cmd = 'B';
3629 } else if (!prefixcmp(opt, "find-copies")) {
3630 opt += strlen("find-copies");
3631 if (*opt == 0 || *opt++ == '=')
3632 cmd = 'C';
3633 } else if (!prefixcmp(opt, "find-renames")) {
3634 opt += strlen("find-renames");
3635 if (*opt == 0 || *opt++ == '=')
3636 cmd = 'M';
3639 if (cmd != 'M' && cmd != 'C' && cmd != 'B')
3640 return -1; /* that is not a -M, -C nor -B option */
3642 opt1 = parse_rename_score(&opt);
3643 if (cmd != 'B')
3644 opt2 = 0;
3645 else {
3646 if (*opt == 0)
3647 opt2 = 0;
3648 else if (*opt != '/')
3649 return -1; /* we expect -B80/99 or -B80 */
3650 else {
3651 opt++;
3652 opt2 = parse_rename_score(&opt);
3655 if (*opt != 0)
3656 return -1;
3657 return opt1 | (opt2 << 16);
3660 struct diff_queue_struct diff_queued_diff;
3662 void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
3664 if (queue->alloc <= queue->nr) {
3665 queue->alloc = alloc_nr(queue->alloc);
3666 queue->queue = xrealloc(queue->queue,
3667 sizeof(dp) * queue->alloc);
3669 queue->queue[queue->nr++] = dp;
3672 struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
3673 struct diff_filespec *one,
3674 struct diff_filespec *two)
3676 struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
3677 dp->one = one;
3678 dp->two = two;
3679 if (queue)
3680 diff_q(queue, dp);
3681 return dp;
3684 void diff_free_filepair(struct diff_filepair *p)
3686 free_filespec(p->one);
3687 free_filespec(p->two);
3688 free(p);
3691 /* This is different from find_unique_abbrev() in that
3692 * it stuffs the result with dots for alignment.
3694 const char *diff_unique_abbrev(const unsigned char *sha1, int len)
3696 int abblen;
3697 const char *abbrev;
3698 if (len == 40)
3699 return sha1_to_hex(sha1);
3701 abbrev = find_unique_abbrev(sha1, len);
3702 abblen = strlen(abbrev);
3703 if (abblen < 37) {
3704 static char hex[41];
3705 if (len < abblen && abblen <= len + 2)
3706 sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
3707 else
3708 sprintf(hex, "%s...", abbrev);
3709 return hex;
3711 return sha1_to_hex(sha1);
3714 static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
3716 int line_termination = opt->line_termination;
3717 int inter_name_termination = line_termination ? '\t' : '\0';
3718 if (opt->output_prefix) {
3719 struct strbuf *msg = NULL;
3720 msg = opt->output_prefix(opt, opt->output_prefix_data);
3721 fprintf(opt->file, "%s", msg->buf);
3724 if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
3725 fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
3726 diff_unique_abbrev(p->one->sha1, opt->abbrev));
3727 fprintf(opt->file, "%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
3729 if (p->score) {
3730 fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
3731 inter_name_termination);
3732 } else {
3733 fprintf(opt->file, "%c%c", p->status, inter_name_termination);
3736 if (p->status == DIFF_STATUS_COPIED ||
3737 p->status == DIFF_STATUS_RENAMED) {
3738 const char *name_a, *name_b;
3739 name_a = p->one->path;
3740 name_b = p->two->path;
3741 strip_prefix(opt->prefix_length, &name_a, &name_b);
3742 write_name_quoted(name_a, opt->file, inter_name_termination);
3743 write_name_quoted(name_b, opt->file, line_termination);
3744 } else {
3745 const char *name_a, *name_b;
3746 name_a = p->one->mode ? p->one->path : p->two->path;
3747 name_b = NULL;
3748 strip_prefix(opt->prefix_length, &name_a, &name_b);
3749 write_name_quoted(name_a, opt->file, line_termination);
3753 int diff_unmodified_pair(struct diff_filepair *p)
3755 /* This function is written stricter than necessary to support
3756 * the currently implemented transformers, but the idea is to
3757 * let transformers to produce diff_filepairs any way they want,
3758 * and filter and clean them up here before producing the output.
3760 struct diff_filespec *one = p->one, *two = p->two;
3762 if (DIFF_PAIR_UNMERGED(p))
3763 return 0; /* unmerged is interesting */
3765 /* deletion, addition, mode or type change
3766 * and rename are all interesting.
3768 if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
3769 DIFF_PAIR_MODE_CHANGED(p) ||
3770 strcmp(one->path, two->path))
3771 return 0;
3773 /* both are valid and point at the same path. that is, we are
3774 * dealing with a change.
3776 if (one->sha1_valid && two->sha1_valid &&
3777 !hashcmp(one->sha1, two->sha1) &&
3778 !one->dirty_submodule && !two->dirty_submodule)
3779 return 1; /* no change */
3780 if (!one->sha1_valid && !two->sha1_valid)
3781 return 1; /* both look at the same file on the filesystem. */
3782 return 0;
3785 static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
3787 if (diff_unmodified_pair(p))
3788 return;
3790 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3791 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3792 return; /* no tree diffs in patch format */
3794 run_diff(p, o);
3797 static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
3798 struct diffstat_t *diffstat)
3800 if (diff_unmodified_pair(p))
3801 return;
3803 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3804 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3805 return; /* no useful stat for tree diffs */
3807 run_diffstat(p, o, diffstat);
3810 static void diff_flush_checkdiff(struct diff_filepair *p,
3811 struct diff_options *o)
3813 if (diff_unmodified_pair(p))
3814 return;
3816 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3817 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3818 return; /* nothing to check in tree diffs */
3820 run_checkdiff(p, o);
3823 int diff_queue_is_empty(void)
3825 struct diff_queue_struct *q = &diff_queued_diff;
3826 int i;
3827 for (i = 0; i < q->nr; i++)
3828 if (!diff_unmodified_pair(q->queue[i]))
3829 return 0;
3830 return 1;
3833 #if DIFF_DEBUG
3834 void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
3836 fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
3837 x, one ? one : "",
3838 s->path,
3839 DIFF_FILE_VALID(s) ? "valid" : "invalid",
3840 s->mode,
3841 s->sha1_valid ? sha1_to_hex(s->sha1) : "");
3842 fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
3843 x, one ? one : "",
3844 s->size, s->xfrm_flags);
3847 void diff_debug_filepair(const struct diff_filepair *p, int i)
3849 diff_debug_filespec(p->one, i, "one");
3850 diff_debug_filespec(p->two, i, "two");
3851 fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
3852 p->score, p->status ? p->status : '?',
3853 p->one->rename_used, p->broken_pair);
3856 void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
3858 int i;
3859 if (msg)
3860 fprintf(stderr, "%s\n", msg);
3861 fprintf(stderr, "q->nr = %d\n", q->nr);
3862 for (i = 0; i < q->nr; i++) {
3863 struct diff_filepair *p = q->queue[i];
3864 diff_debug_filepair(p, i);
3867 #endif
3869 static void diff_resolve_rename_copy(void)
3871 int i;
3872 struct diff_filepair *p;
3873 struct diff_queue_struct *q = &diff_queued_diff;
3875 diff_debug_queue("resolve-rename-copy", q);
3877 for (i = 0; i < q->nr; i++) {
3878 p = q->queue[i];
3879 p->status = 0; /* undecided */
3880 if (DIFF_PAIR_UNMERGED(p))
3881 p->status = DIFF_STATUS_UNMERGED;
3882 else if (!DIFF_FILE_VALID(p->one))
3883 p->status = DIFF_STATUS_ADDED;
3884 else if (!DIFF_FILE_VALID(p->two))
3885 p->status = DIFF_STATUS_DELETED;
3886 else if (DIFF_PAIR_TYPE_CHANGED(p))
3887 p->status = DIFF_STATUS_TYPE_CHANGED;
3889 /* from this point on, we are dealing with a pair
3890 * whose both sides are valid and of the same type, i.e.
3891 * either in-place edit or rename/copy edit.
3893 else if (DIFF_PAIR_RENAME(p)) {
3895 * A rename might have re-connected a broken
3896 * pair up, causing the pathnames to be the
3897 * same again. If so, that's not a rename at
3898 * all, just a modification..
3900 * Otherwise, see if this source was used for
3901 * multiple renames, in which case we decrement
3902 * the count, and call it a copy.
3904 if (!strcmp(p->one->path, p->two->path))
3905 p->status = DIFF_STATUS_MODIFIED;
3906 else if (--p->one->rename_used > 0)
3907 p->status = DIFF_STATUS_COPIED;
3908 else
3909 p->status = DIFF_STATUS_RENAMED;
3911 else if (hashcmp(p->one->sha1, p->two->sha1) ||
3912 p->one->mode != p->two->mode ||
3913 p->one->dirty_submodule ||
3914 p->two->dirty_submodule ||
3915 is_null_sha1(p->one->sha1))
3916 p->status = DIFF_STATUS_MODIFIED;
3917 else {
3918 /* This is a "no-change" entry and should not
3919 * happen anymore, but prepare for broken callers.
3921 error("feeding unmodified %s to diffcore",
3922 p->one->path);
3923 p->status = DIFF_STATUS_UNKNOWN;
3926 diff_debug_queue("resolve-rename-copy done", q);
3929 static int check_pair_status(struct diff_filepair *p)
3931 switch (p->status) {
3932 case DIFF_STATUS_UNKNOWN:
3933 return 0;
3934 case 0:
3935 die("internal error in diff-resolve-rename-copy");
3936 default:
3937 return 1;
3941 static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
3943 int fmt = opt->output_format;
3945 if (fmt & DIFF_FORMAT_CHECKDIFF)
3946 diff_flush_checkdiff(p, opt);
3947 else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
3948 diff_flush_raw(p, opt);
3949 else if (fmt & DIFF_FORMAT_NAME) {
3950 const char *name_a, *name_b;
3951 name_a = p->two->path;
3952 name_b = NULL;
3953 strip_prefix(opt->prefix_length, &name_a, &name_b);
3954 write_name_quoted(name_a, opt->file, opt->line_termination);
3958 static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
3960 if (fs->mode)
3961 fprintf(file, " %s mode %06o ", newdelete, fs->mode);
3962 else
3963 fprintf(file, " %s ", newdelete);
3964 write_name_quoted(fs->path, file, '\n');
3968 static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
3969 const char *line_prefix)
3971 if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
3972 fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
3973 p->two->mode, show_name ? ' ' : '\n');
3974 if (show_name) {
3975 write_name_quoted(p->two->path, file, '\n');
3980 static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
3981 const char *line_prefix)
3983 char *names = pprint_rename(p->one->path, p->two->path);
3985 fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
3986 free(names);
3987 show_mode_change(file, p, 0, line_prefix);
3990 static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
3992 FILE *file = opt->file;
3993 char *line_prefix = "";
3995 if (opt->output_prefix) {
3996 struct strbuf *buf = opt->output_prefix(opt, opt->output_prefix_data);
3997 line_prefix = buf->buf;
4000 switch(p->status) {
4001 case DIFF_STATUS_DELETED:
4002 fputs(line_prefix, file);
4003 show_file_mode_name(file, "delete", p->one);
4004 break;
4005 case DIFF_STATUS_ADDED:
4006 fputs(line_prefix, file);
4007 show_file_mode_name(file, "create", p->two);
4008 break;
4009 case DIFF_STATUS_COPIED:
4010 fputs(line_prefix, file);
4011 show_rename_copy(file, "copy", p, line_prefix);
4012 break;
4013 case DIFF_STATUS_RENAMED:
4014 fputs(line_prefix, file);
4015 show_rename_copy(file, "rename", p, line_prefix);
4016 break;
4017 default:
4018 if (p->score) {
4019 fprintf(file, "%s rewrite ", line_prefix);
4020 write_name_quoted(p->two->path, file, ' ');
4021 fprintf(file, "(%d%%)\n", similarity_index(p));
4023 show_mode_change(file, p, !p->score, line_prefix);
4024 break;
4028 struct patch_id_t {
4029 git_SHA_CTX *ctx;
4030 int patchlen;
4033 static int remove_space(char *line, int len)
4035 int i;
4036 char *dst = line;
4037 unsigned char c;
4039 for (i = 0; i < len; i++)
4040 if (!isspace((c = line[i])))
4041 *dst++ = c;
4043 return dst - line;
4046 static void patch_id_consume(void *priv, char *line, unsigned long len)
4048 struct patch_id_t *data = priv;
4049 int new_len;
4051 /* Ignore line numbers when computing the SHA1 of the patch */
4052 if (!prefixcmp(line, "@@ -"))
4053 return;
4055 new_len = remove_space(line, len);
4057 git_SHA1_Update(data->ctx, line, new_len);
4058 data->patchlen += new_len;
4061 /* returns 0 upon success, and writes result into sha1 */
4062 static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
4064 struct diff_queue_struct *q = &diff_queued_diff;
4065 int i;
4066 git_SHA_CTX ctx;
4067 struct patch_id_t data;
4068 char buffer[PATH_MAX * 4 + 20];
4070 git_SHA1_Init(&ctx);
4071 memset(&data, 0, sizeof(struct patch_id_t));
4072 data.ctx = &ctx;
4074 for (i = 0; i < q->nr; i++) {
4075 xpparam_t xpp;
4076 xdemitconf_t xecfg;
4077 mmfile_t mf1, mf2;
4078 struct diff_filepair *p = q->queue[i];
4079 int len1, len2;
4081 memset(&xpp, 0, sizeof(xpp));
4082 memset(&xecfg, 0, sizeof(xecfg));
4083 if (p->status == 0)
4084 return error("internal diff status error");
4085 if (p->status == DIFF_STATUS_UNKNOWN)
4086 continue;
4087 if (diff_unmodified_pair(p))
4088 continue;
4089 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4090 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4091 continue;
4092 if (DIFF_PAIR_UNMERGED(p))
4093 continue;
4095 diff_fill_sha1_info(p->one);
4096 diff_fill_sha1_info(p->two);
4097 if (fill_mmfile(&mf1, p->one) < 0 ||
4098 fill_mmfile(&mf2, p->two) < 0)
4099 return error("unable to read files to diff");
4101 len1 = remove_space(p->one->path, strlen(p->one->path));
4102 len2 = remove_space(p->two->path, strlen(p->two->path));
4103 if (p->one->mode == 0)
4104 len1 = snprintf(buffer, sizeof(buffer),
4105 "diff--gita/%.*sb/%.*s"
4106 "newfilemode%06o"
4107 "---/dev/null"
4108 "+++b/%.*s",
4109 len1, p->one->path,
4110 len2, p->two->path,
4111 p->two->mode,
4112 len2, p->two->path);
4113 else if (p->two->mode == 0)
4114 len1 = snprintf(buffer, sizeof(buffer),
4115 "diff--gita/%.*sb/%.*s"
4116 "deletedfilemode%06o"
4117 "---a/%.*s"
4118 "+++/dev/null",
4119 len1, p->one->path,
4120 len2, p->two->path,
4121 p->one->mode,
4122 len1, p->one->path);
4123 else
4124 len1 = snprintf(buffer, sizeof(buffer),
4125 "diff--gita/%.*sb/%.*s"
4126 "---a/%.*s"
4127 "+++b/%.*s",
4128 len1, p->one->path,
4129 len2, p->two->path,
4130 len1, p->one->path,
4131 len2, p->two->path);
4132 git_SHA1_Update(&ctx, buffer, len1);
4134 if (diff_filespec_is_binary(p->one) ||
4135 diff_filespec_is_binary(p->two)) {
4136 git_SHA1_Update(&ctx, sha1_to_hex(p->one->sha1), 40);
4137 git_SHA1_Update(&ctx, sha1_to_hex(p->two->sha1), 40);
4138 continue;
4141 xpp.flags = 0;
4142 xecfg.ctxlen = 3;
4143 xecfg.flags = 0;
4144 xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4145 &xpp, &xecfg);
4148 git_SHA1_Final(sha1, &ctx);
4149 return 0;
4152 int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
4154 struct diff_queue_struct *q = &diff_queued_diff;
4155 int i;
4156 int result = diff_get_patch_id(options, sha1);
4158 for (i = 0; i < q->nr; i++)
4159 diff_free_filepair(q->queue[i]);
4161 free(q->queue);
4162 DIFF_QUEUE_CLEAR(q);
4164 return result;
4167 static int is_summary_empty(const struct diff_queue_struct *q)
4169 int i;
4171 for (i = 0; i < q->nr; i++) {
4172 const struct diff_filepair *p = q->queue[i];
4174 switch (p->status) {
4175 case DIFF_STATUS_DELETED:
4176 case DIFF_STATUS_ADDED:
4177 case DIFF_STATUS_COPIED:
4178 case DIFF_STATUS_RENAMED:
4179 return 0;
4180 default:
4181 if (p->score)
4182 return 0;
4183 if (p->one->mode && p->two->mode &&
4184 p->one->mode != p->two->mode)
4185 return 0;
4186 break;
4189 return 1;
4192 static const char rename_limit_warning[] =
4193 "inexact rename detection was skipped due to too many files.";
4195 static const char degrade_cc_to_c_warning[] =
4196 "only found copies from modified paths due to too many files.";
4198 static const char rename_limit_advice[] =
4199 "you may want to set your %s variable to at least "
4200 "%d and retry the command.";
4202 void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4204 if (degraded_cc)
4205 warning(degrade_cc_to_c_warning);
4206 else if (needed)
4207 warning(rename_limit_warning);
4208 else
4209 return;
4210 if (0 < needed && needed < 32767)
4211 warning(rename_limit_advice, varname, needed);
4214 void diff_flush(struct diff_options *options)
4216 struct diff_queue_struct *q = &diff_queued_diff;
4217 int i, output_format = options->output_format;
4218 int separator = 0;
4219 int dirstat_by_line = 0;
4222 * Order: raw, stat, summary, patch
4223 * or: name/name-status/checkdiff (other bits clear)
4225 if (!q->nr)
4226 goto free_queue;
4228 if (output_format & (DIFF_FORMAT_RAW |
4229 DIFF_FORMAT_NAME |
4230 DIFF_FORMAT_NAME_STATUS |
4231 DIFF_FORMAT_CHECKDIFF)) {
4232 for (i = 0; i < q->nr; i++) {
4233 struct diff_filepair *p = q->queue[i];
4234 if (check_pair_status(p))
4235 flush_one_pair(p, options);
4237 separator++;
4240 if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4241 dirstat_by_line = 1;
4243 if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4244 dirstat_by_line) {
4245 struct diffstat_t diffstat;
4247 memset(&diffstat, 0, sizeof(struct diffstat_t));
4248 for (i = 0; i < q->nr; i++) {
4249 struct diff_filepair *p = q->queue[i];
4250 if (check_pair_status(p))
4251 diff_flush_stat(p, options, &diffstat);
4253 if (output_format & DIFF_FORMAT_NUMSTAT)
4254 show_numstat(&diffstat, options);
4255 if (output_format & DIFF_FORMAT_DIFFSTAT)
4256 show_stats(&diffstat, options);
4257 if (output_format & DIFF_FORMAT_SHORTSTAT)
4258 show_shortstats(&diffstat, options);
4259 if (output_format & DIFF_FORMAT_DIRSTAT)
4260 show_dirstat_by_line(&diffstat, options);
4261 free_diffstat_info(&diffstat);
4262 separator++;
4264 if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4265 show_dirstat(options);
4267 if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4268 for (i = 0; i < q->nr; i++) {
4269 diff_summary(options, q->queue[i]);
4271 separator++;
4274 if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4275 DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4276 DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4278 * run diff_flush_patch for the exit status. setting
4279 * options->file to /dev/null should be safe, becaue we
4280 * aren't supposed to produce any output anyway.
4282 if (options->close_file)
4283 fclose(options->file);
4284 options->file = fopen("/dev/null", "w");
4285 if (!options->file)
4286 die_errno("Could not open /dev/null");
4287 options->close_file = 1;
4288 for (i = 0; i < q->nr; i++) {
4289 struct diff_filepair *p = q->queue[i];
4290 if (check_pair_status(p))
4291 diff_flush_patch(p, options);
4292 if (options->found_changes)
4293 break;
4297 if (output_format & DIFF_FORMAT_PATCH) {
4298 if (separator) {
4299 putc(options->line_termination, options->file);
4300 if (options->stat_sep) {
4301 /* attach patch instead of inline */
4302 fputs(options->stat_sep, options->file);
4306 for (i = 0; i < q->nr; i++) {
4307 struct diff_filepair *p = q->queue[i];
4308 if (check_pair_status(p))
4309 diff_flush_patch(p, options);
4313 if (output_format & DIFF_FORMAT_CALLBACK)
4314 options->format_callback(q, options, options->format_callback_data);
4316 for (i = 0; i < q->nr; i++)
4317 diff_free_filepair(q->queue[i]);
4318 free_queue:
4319 free(q->queue);
4320 DIFF_QUEUE_CLEAR(q);
4321 if (options->close_file)
4322 fclose(options->file);
4325 * Report the content-level differences with HAS_CHANGES;
4326 * diff_addremove/diff_change does not set the bit when
4327 * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4329 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4330 if (options->found_changes)
4331 DIFF_OPT_SET(options, HAS_CHANGES);
4332 else
4333 DIFF_OPT_CLR(options, HAS_CHANGES);
4337 static void diffcore_apply_filter(const char *filter)
4339 int i;
4340 struct diff_queue_struct *q = &diff_queued_diff;
4341 struct diff_queue_struct outq;
4342 DIFF_QUEUE_CLEAR(&outq);
4344 if (!filter)
4345 return;
4347 if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
4348 int found;
4349 for (i = found = 0; !found && i < q->nr; i++) {
4350 struct diff_filepair *p = q->queue[i];
4351 if (((p->status == DIFF_STATUS_MODIFIED) &&
4352 ((p->score &&
4353 strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
4354 (!p->score &&
4355 strchr(filter, DIFF_STATUS_MODIFIED)))) ||
4356 ((p->status != DIFF_STATUS_MODIFIED) &&
4357 strchr(filter, p->status)))
4358 found++;
4360 if (found)
4361 return;
4363 /* otherwise we will clear the whole queue
4364 * by copying the empty outq at the end of this
4365 * function, but first clear the current entries
4366 * in the queue.
4368 for (i = 0; i < q->nr; i++)
4369 diff_free_filepair(q->queue[i]);
4371 else {
4372 /* Only the matching ones */
4373 for (i = 0; i < q->nr; i++) {
4374 struct diff_filepair *p = q->queue[i];
4376 if (((p->status == DIFF_STATUS_MODIFIED) &&
4377 ((p->score &&
4378 strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
4379 (!p->score &&
4380 strchr(filter, DIFF_STATUS_MODIFIED)))) ||
4381 ((p->status != DIFF_STATUS_MODIFIED) &&
4382 strchr(filter, p->status)))
4383 diff_q(&outq, p);
4384 else
4385 diff_free_filepair(p);
4388 free(q->queue);
4389 *q = outq;
4392 /* Check whether two filespecs with the same mode and size are identical */
4393 static int diff_filespec_is_identical(struct diff_filespec *one,
4394 struct diff_filespec *two)
4396 if (S_ISGITLINK(one->mode))
4397 return 0;
4398 if (diff_populate_filespec(one, 0))
4399 return 0;
4400 if (diff_populate_filespec(two, 0))
4401 return 0;
4402 return !memcmp(one->data, two->data, one->size);
4405 static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
4407 int i;
4408 struct diff_queue_struct *q = &diff_queued_diff;
4409 struct diff_queue_struct outq;
4410 DIFF_QUEUE_CLEAR(&outq);
4412 for (i = 0; i < q->nr; i++) {
4413 struct diff_filepair *p = q->queue[i];
4416 * 1. Entries that come from stat info dirtiness
4417 * always have both sides (iow, not create/delete),
4418 * one side of the object name is unknown, with
4419 * the same mode and size. Keep the ones that
4420 * do not match these criteria. They have real
4421 * differences.
4423 * 2. At this point, the file is known to be modified,
4424 * with the same mode and size, and the object
4425 * name of one side is unknown. Need to inspect
4426 * the identical contents.
4428 if (!DIFF_FILE_VALID(p->one) || /* (1) */
4429 !DIFF_FILE_VALID(p->two) ||
4430 (p->one->sha1_valid && p->two->sha1_valid) ||
4431 (p->one->mode != p->two->mode) ||
4432 diff_populate_filespec(p->one, 1) ||
4433 diff_populate_filespec(p->two, 1) ||
4434 (p->one->size != p->two->size) ||
4435 !diff_filespec_is_identical(p->one, p->two)) /* (2) */
4436 diff_q(&outq, p);
4437 else {
4439 * The caller can subtract 1 from skip_stat_unmatch
4440 * to determine how many paths were dirty only
4441 * due to stat info mismatch.
4443 if (!DIFF_OPT_TST(diffopt, NO_INDEX))
4444 diffopt->skip_stat_unmatch++;
4445 diff_free_filepair(p);
4448 free(q->queue);
4449 *q = outq;
4452 static int diffnamecmp(const void *a_, const void *b_)
4454 const struct diff_filepair *a = *((const struct diff_filepair **)a_);
4455 const struct diff_filepair *b = *((const struct diff_filepair **)b_);
4456 const char *name_a, *name_b;
4458 name_a = a->one ? a->one->path : a->two->path;
4459 name_b = b->one ? b->one->path : b->two->path;
4460 return strcmp(name_a, name_b);
4463 void diffcore_fix_diff_index(struct diff_options *options)
4465 struct diff_queue_struct *q = &diff_queued_diff;
4466 qsort(q->queue, q->nr, sizeof(q->queue[0]), diffnamecmp);
4469 void diffcore_std(struct diff_options *options)
4471 if (options->skip_stat_unmatch)
4472 diffcore_skip_stat_unmatch(options);
4473 if (!options->found_follow) {
4474 /* See try_to_follow_renames() in tree-diff.c */
4475 if (options->break_opt != -1)
4476 diffcore_break(options->break_opt);
4477 if (options->detect_rename)
4478 diffcore_rename(options);
4479 if (options->break_opt != -1)
4480 diffcore_merge_broken();
4482 if (options->pickaxe)
4483 diffcore_pickaxe(options);
4484 if (options->orderfile)
4485 diffcore_order(options->orderfile);
4486 if (!options->found_follow)
4487 /* See try_to_follow_renames() in tree-diff.c */
4488 diff_resolve_rename_copy();
4489 diffcore_apply_filter(options->filter);
4491 if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4492 DIFF_OPT_SET(options, HAS_CHANGES);
4493 else
4494 DIFF_OPT_CLR(options, HAS_CHANGES);
4496 options->found_follow = 0;
4499 int diff_result_code(struct diff_options *opt, int status)
4501 int result = 0;
4503 diff_warn_rename_limit("diff.renamelimit",
4504 opt->needed_rename_limit,
4505 opt->degraded_cc_to_c);
4506 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4507 !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
4508 return status;
4509 if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4510 DIFF_OPT_TST(opt, HAS_CHANGES))
4511 result |= 01;
4512 if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
4513 DIFF_OPT_TST(opt, CHECK_FAILED))
4514 result |= 02;
4515 return result;
4518 int diff_can_quit_early(struct diff_options *opt)
4520 return (DIFF_OPT_TST(opt, QUICK) &&
4521 !opt->filter &&
4522 DIFF_OPT_TST(opt, HAS_CHANGES));
4526 * Shall changes to this submodule be ignored?
4528 * Submodule changes can be configured to be ignored separately for each path,
4529 * but that configuration can be overridden from the command line.
4531 static int is_submodule_ignored(const char *path, struct diff_options *options)
4533 int ignored = 0;
4534 unsigned orig_flags = options->flags;
4535 if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
4536 set_diffopt_flags_from_submodule_config(options, path);
4537 if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
4538 ignored = 1;
4539 options->flags = orig_flags;
4540 return ignored;
4543 void diff_addremove(struct diff_options *options,
4544 int addremove, unsigned mode,
4545 const unsigned char *sha1,
4546 const char *concatpath, unsigned dirty_submodule)
4548 struct diff_filespec *one, *two;
4550 if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
4551 return;
4553 /* This may look odd, but it is a preparation for
4554 * feeding "there are unchanged files which should
4555 * not produce diffs, but when you are doing copy
4556 * detection you would need them, so here they are"
4557 * entries to the diff-core. They will be prefixed
4558 * with something like '=' or '*' (I haven't decided
4559 * which but should not make any difference).
4560 * Feeding the same new and old to diff_change()
4561 * also has the same effect.
4562 * Before the final output happens, they are pruned after
4563 * merged into rename/copy pairs as appropriate.
4565 if (DIFF_OPT_TST(options, REVERSE_DIFF))
4566 addremove = (addremove == '+' ? '-' :
4567 addremove == '-' ? '+' : addremove);
4569 if (options->prefix &&
4570 strncmp(concatpath, options->prefix, options->prefix_length))
4571 return;
4573 one = alloc_filespec(concatpath);
4574 two = alloc_filespec(concatpath);
4576 if (addremove != '+')
4577 fill_filespec(one, sha1, mode);
4578 if (addremove != '-') {
4579 fill_filespec(two, sha1, mode);
4580 two->dirty_submodule = dirty_submodule;
4583 diff_queue(&diff_queued_diff, one, two);
4584 if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4585 DIFF_OPT_SET(options, HAS_CHANGES);
4588 void diff_change(struct diff_options *options,
4589 unsigned old_mode, unsigned new_mode,
4590 const unsigned char *old_sha1,
4591 const unsigned char *new_sha1,
4592 const char *concatpath,
4593 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
4595 struct diff_filespec *one, *two;
4597 if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
4598 is_submodule_ignored(concatpath, options))
4599 return;
4601 if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
4602 unsigned tmp;
4603 const unsigned char *tmp_c;
4604 tmp = old_mode; old_mode = new_mode; new_mode = tmp;
4605 tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
4606 tmp = old_dirty_submodule; old_dirty_submodule = new_dirty_submodule;
4607 new_dirty_submodule = tmp;
4610 if (options->prefix &&
4611 strncmp(concatpath, options->prefix, options->prefix_length))
4612 return;
4614 one = alloc_filespec(concatpath);
4615 two = alloc_filespec(concatpath);
4616 fill_filespec(one, old_sha1, old_mode);
4617 fill_filespec(two, new_sha1, new_mode);
4618 one->dirty_submodule = old_dirty_submodule;
4619 two->dirty_submodule = new_dirty_submodule;
4621 diff_queue(&diff_queued_diff, one, two);
4622 if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4623 DIFF_OPT_SET(options, HAS_CHANGES);
4626 struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
4628 struct diff_filepair *pair;
4629 struct diff_filespec *one, *two;
4631 if (options->prefix &&
4632 strncmp(path, options->prefix, options->prefix_length))
4633 return NULL;
4635 one = alloc_filespec(path);
4636 two = alloc_filespec(path);
4637 pair = diff_queue(&diff_queued_diff, one, two);
4638 pair->is_unmerged = 1;
4639 return pair;
4642 static char *run_textconv(const char *pgm, struct diff_filespec *spec,
4643 size_t *outsize)
4645 struct diff_tempfile *temp;
4646 const char *argv[3];
4647 const char **arg = argv;
4648 struct child_process child;
4649 struct strbuf buf = STRBUF_INIT;
4650 int err = 0;
4652 temp = prepare_temp_file(spec->path, spec);
4653 *arg++ = pgm;
4654 *arg++ = temp->name;
4655 *arg = NULL;
4657 memset(&child, 0, sizeof(child));
4658 child.use_shell = 1;
4659 child.argv = argv;
4660 child.out = -1;
4661 if (start_command(&child)) {
4662 remove_tempfile();
4663 return NULL;
4666 if (strbuf_read(&buf, child.out, 0) < 0)
4667 err = error("error reading from textconv command '%s'", pgm);
4668 close(child.out);
4670 if (finish_command(&child) || err) {
4671 strbuf_release(&buf);
4672 remove_tempfile();
4673 return NULL;
4675 remove_tempfile();
4677 return strbuf_detach(&buf, outsize);
4680 size_t fill_textconv(struct userdiff_driver *driver,
4681 struct diff_filespec *df,
4682 char **outbuf)
4684 size_t size;
4686 if (!driver || !driver->textconv) {
4687 if (!DIFF_FILE_VALID(df)) {
4688 *outbuf = "";
4689 return 0;
4691 if (diff_populate_filespec(df, 0))
4692 die("unable to read files to diff");
4693 *outbuf = df->data;
4694 return df->size;
4697 if (driver->textconv_cache && df->sha1_valid) {
4698 *outbuf = notes_cache_get(driver->textconv_cache, df->sha1,
4699 &size);
4700 if (*outbuf)
4701 return size;
4704 *outbuf = run_textconv(driver->textconv, df, &size);
4705 if (!*outbuf)
4706 die("unable to read files to diff");
4708 if (driver->textconv_cache && df->sha1_valid) {
4709 /* ignore errors, as we might be in a readonly repository */
4710 notes_cache_put(driver->textconv_cache, df->sha1, *outbuf,
4711 size);
4713 * we could save up changes and flush them all at the end,
4714 * but we would need an extra call after all diffing is done.
4715 * Since generating a cache entry is the slow path anyway,
4716 * this extra overhead probably isn't a big deal.
4718 notes_cache_write(driver->textconv_cache);
4721 return size;