1/*
2 * Copyright (C) 2005 Junio C Hamano
3 */
4#include "cache.h"
5#include "config.h"
6#include "tempfile.h"
7#include "quote.h"
8#include "diff.h"
9#include "diffcore.h"
10#include "delta.h"
11#include "xdiff-interface.h"
12#include "color.h"
13#include "attr.h"
14#include "run-command.h"
15#include "utf8.h"
16#include "userdiff.h"
17#include "submodule-config.h"
18#include "submodule.h"
19#include "ll-merge.h"
20#include "string-list.h"
21#include "argv-array.h"
22#include "graph.h"
23#include "packfile.h"
24#ifdef NO_FAST_WORKING_DIRECTORY
26#define FAST_WORKING_DIRECTORY 0
27#else
28#define FAST_WORKING_DIRECTORY 1
29#endif
30static int diff_detect_rename_default;
32static int diff_indent_heuristic = 1;
33static int diff_rename_limit_default = 400;
34static int diff_suppress_blank_empty;
35static int diff_use_color_default = -1;
36static int diff_context_default = 3;
37static int diff_interhunk_context_default;
38static const char *diff_word_regex_cfg;
39static const char *external_diff_cmd_cfg;
40static const char *diff_order_file_cfg;
41int diff_auto_refresh_index = 1;
42static int diff_mnemonic_prefix;
43static int diff_no_prefix;
44static int diff_stat_graph_width;
45static int diff_dirstat_permille_default = 30;
46static struct diff_options default_diff_options;
47static long diff_algorithm;
48static unsigned ws_error_highlight_default = WSEH_NEW;
49static char diff_colors[][COLOR_MAXLEN] = {
51 GIT_COLOR_RESET,
52 GIT_COLOR_NORMAL, /* CONTEXT */
53 GIT_COLOR_BOLD, /* METAINFO */
54 GIT_COLOR_CYAN, /* FRAGINFO */
55 GIT_COLOR_RED, /* OLD */
56 GIT_COLOR_GREEN, /* NEW */
57 GIT_COLOR_YELLOW, /* COMMIT */
58 GIT_COLOR_BG_RED, /* WHITESPACE */
59 GIT_COLOR_NORMAL, /* FUNCINFO */
60};
61static NORETURN void die_want_option(const char *option_name)
63{
64 die(_("option '%s' requires a value"), option_name);
65}
66static int parse_diff_color_slot(const char *var)
68{
69 if (!strcasecmp(var, "context") || !strcasecmp(var, "plain"))
70 return DIFF_CONTEXT;
71 if (!strcasecmp(var, "meta"))
72 return DIFF_METAINFO;
73 if (!strcasecmp(var, "frag"))
74 return DIFF_FRAGINFO;
75 if (!strcasecmp(var, "old"))
76 return DIFF_FILE_OLD;
77 if (!strcasecmp(var, "new"))
78 return DIFF_FILE_NEW;
79 if (!strcasecmp(var, "commit"))
80 return DIFF_COMMIT;
81 if (!strcasecmp(var, "whitespace"))
82 return DIFF_WHITESPACE;
83 if (!strcasecmp(var, "func"))
84 return DIFF_FUNCINFO;
85 return -1;
86}
87static int parse_dirstat_params(struct diff_options *options, const char *params_string,
89 struct strbuf *errmsg)
90{
91 char *params_copy = xstrdup(params_string);
92 struct string_list params = STRING_LIST_INIT_NODUP;
93 int ret = 0;
94 int i;
95 if (*params_copy)
97 string_list_split_in_place(¶ms, params_copy, ',', -1);
98 for (i = 0; i < params.nr; i++) {
99 const char *p = params.items[i].string;
100 if (!strcmp(p, "changes")) {
101 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
102 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
103 } else if (!strcmp(p, "lines")) {
104 DIFF_OPT_SET(options, DIRSTAT_BY_LINE);
105 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
106 } else if (!strcmp(p, "files")) {
107 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
108 DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
109 } else if (!strcmp(p, "noncumulative")) {
110 DIFF_OPT_CLR(options, DIRSTAT_CUMULATIVE);
111 } else if (!strcmp(p, "cumulative")) {
112 DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
113 } else if (isdigit(*p)) {
114 char *end;
115 int permille = strtoul(p, &end, 10) * 10;
116 if (*end == '.' && isdigit(*++end)) {
117 /* only use first digit */
118 permille += *end - '0';
119 /* .. and ignore any further digits */
120 while (isdigit(*++end))
121 ; /* nothing */
122 }
123 if (!*end)
124 options->dirstat_permille = permille;
125 else {
126 strbuf_addf(errmsg, _(" Failed to parse dirstat cut-off percentage '%s'\n"),
127 p);
128 ret++;
129 }
130 } else {
131 strbuf_addf(errmsg, _(" Unknown dirstat parameter '%s'\n"), p);
132 ret++;
133 }
134 }
136 string_list_clear(¶ms, 0);
137 free(params_copy);
138 return ret;
139}
140static int parse_submodule_params(struct diff_options *options, const char *value)
142{
143 if (!strcmp(value, "log"))
144 options->submodule_format = DIFF_SUBMODULE_LOG;
145 else if (!strcmp(value, "short"))
146 options->submodule_format = DIFF_SUBMODULE_SHORT;
147 else if (!strcmp(value, "diff"))
148 options->submodule_format = DIFF_SUBMODULE_INLINE_DIFF;
149 else
150 return -1;
151 return 0;
152}
153static int git_config_rename(const char *var, const char *value)
155{
156 if (!value)
157 return DIFF_DETECT_RENAME;
158 if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
159 return DIFF_DETECT_COPY;
160 return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
161}
162long parse_algorithm_value(const char *value)
164{
165 if (!value)
166 return -1;
167 else if (!strcasecmp(value, "myers") || !strcasecmp(value, "default"))
168 return 0;
169 else if (!strcasecmp(value, "minimal"))
170 return XDF_NEED_MINIMAL;
171 else if (!strcasecmp(value, "patience"))
172 return XDF_PATIENCE_DIFF;
173 else if (!strcasecmp(value, "histogram"))
174 return XDF_HISTOGRAM_DIFF;
175 return -1;
176}
177static int parse_one_token(const char **arg, const char *token)
179{
180 const char *rest;
181 if (skip_prefix(*arg, token, &rest) && (!*rest || *rest == ',')) {
182 *arg = rest;
183 return 1;
184 }
185 return 0;
186}
187static int parse_ws_error_highlight(const char *arg)
189{
190 const char *orig_arg = arg;
191 unsigned val = 0;
192 while (*arg) {
194 if (parse_one_token(&arg, "none"))
195 val = 0;
196 else if (parse_one_token(&arg, "default"))
197 val = WSEH_NEW;
198 else if (parse_one_token(&arg, "all"))
199 val = WSEH_NEW | WSEH_OLD | WSEH_CONTEXT;
200 else if (parse_one_token(&arg, "new"))
201 val |= WSEH_NEW;
202 else if (parse_one_token(&arg, "old"))
203 val |= WSEH_OLD;
204 else if (parse_one_token(&arg, "context"))
205 val |= WSEH_CONTEXT;
206 else {
207 return -1 - (int)(arg - orig_arg);
208 }
209 if (*arg)
210 arg++;
211 }
212 return val;
213}
214/*
216 * These are to give UI layer defaults.
217 * The core-level commands such as git-diff-files should
218 * never be affected by the setting of diff.renames
219 * the user happens to have in the configuration file.
220 */
221void init_diff_ui_defaults(void)
222{
223 diff_detect_rename_default = 1;
224}
225int git_diff_heuristic_config(const char *var, const char *value, void *cb)
227{
228 if (!strcmp(var, "diff.indentheuristic"))
229 diff_indent_heuristic = git_config_bool(var, value);
230 return 0;
231}
232int git_diff_ui_config(const char *var, const char *value, void *cb)
234{
235 if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
236 diff_use_color_default = git_config_colorbool(var, value);
237 return 0;
238 }
239 if (!strcmp(var, "diff.context")) {
240 diff_context_default = git_config_int(var, value);
241 if (diff_context_default < 0)
242 return -1;
243 return 0;
244 }
245 if (!strcmp(var, "diff.interhunkcontext")) {
246 diff_interhunk_context_default = git_config_int(var, value);
247 if (diff_interhunk_context_default < 0)
248 return -1;
249 return 0;
250 }
251 if (!strcmp(var, "diff.renames")) {
252 diff_detect_rename_default = git_config_rename(var, value);
253 return 0;
254 }
255 if (!strcmp(var, "diff.autorefreshindex")) {
256 diff_auto_refresh_index = git_config_bool(var, value);
257 return 0;
258 }
259 if (!strcmp(var, "diff.mnemonicprefix")) {
260 diff_mnemonic_prefix = git_config_bool(var, value);
261 return 0;
262 }
263 if (!strcmp(var, "diff.noprefix")) {
264 diff_no_prefix = git_config_bool(var, value);
265 return 0;
266 }
267 if (!strcmp(var, "diff.statgraphwidth")) {
268 diff_stat_graph_width = git_config_int(var, value);
269 return 0;
270 }
271 if (!strcmp(var, "diff.external"))
272 return git_config_string(&external_diff_cmd_cfg, var, value);
273 if (!strcmp(var, "diff.wordregex"))
274 return git_config_string(&diff_word_regex_cfg, var, value);
275 if (!strcmp(var, "diff.orderfile"))
276 return git_config_pathname(&diff_order_file_cfg, var, value);
277 if (!strcmp(var, "diff.ignoresubmodules"))
279 handle_ignore_submodules_arg(&default_diff_options, value);
280 if (!strcmp(var, "diff.submodule")) {
282 if (parse_submodule_params(&default_diff_options, value))
283 warning(_("Unknown value for 'diff.submodule' config variable: '%s'"),
284 value);
285 return 0;
286 }
287 if (!strcmp(var, "diff.algorithm")) {
289 diff_algorithm = parse_algorithm_value(value);
290 if (diff_algorithm < 0)
291 return -1;
292 return 0;
293 }
294 if (!strcmp(var, "diff.wserrorhighlight")) {
296 int val = parse_ws_error_highlight(value);
297 if (val < 0)
298 return -1;
299 ws_error_highlight_default = val;
300 return 0;
301 }
302 return git_diff_basic_config(var, value, cb);
304}
305int git_diff_basic_config(const char *var, const char *value, void *cb)
307{
308 const char *name;
309 if (!strcmp(var, "diff.renamelimit")) {
311 diff_rename_limit_default = git_config_int(var, value);
312 return 0;
313 }
314 if (userdiff_config(var, value) < 0)
316 return -1;
317 if (skip_prefix(var, "diff.color.", &name) ||
319 skip_prefix(var, "color.diff.", &name)) {
320 int slot = parse_diff_color_slot(name);
321 if (slot < 0)
322 return 0;
323 if (!value)
324 return config_error_nonbool(var);
325 return color_parse(value, diff_colors[slot]);
326 }
327 /* like GNU diff's --suppress-blank-empty option */
329 if (!strcmp(var, "diff.suppressblankempty") ||
330 /* for backwards compatibility */
331 !strcmp(var, "diff.suppress-blank-empty")) {
332 diff_suppress_blank_empty = git_config_bool(var, value);
333 return 0;
334 }
335 if (!strcmp(var, "diff.dirstat")) {
337 struct strbuf errmsg = STRBUF_INIT;
338 default_diff_options.dirstat_permille = diff_dirstat_permille_default;
339 if (parse_dirstat_params(&default_diff_options, value, &errmsg))
340 warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
341 errmsg.buf);
342 strbuf_release(&errmsg);
343 diff_dirstat_permille_default = default_diff_options.dirstat_permille;
344 return 0;
345 }
346 if (starts_with(var, "submodule."))
348 return parse_submodule_config_option(var, value);
349 if (git_diff_heuristic_config(var, value, cb) < 0)
351 return -1;
352 return git_default_config(var, value, cb);
354}
355static char *quote_two(const char *one, const char *two)
357{
358 int need_one = quote_c_style(one, NULL, NULL, 1);
359 int need_two = quote_c_style(two, NULL, NULL, 1);
360 struct strbuf res = STRBUF_INIT;
361 if (need_one + need_two) {
363 strbuf_addch(&res, '"');
364 quote_c_style(one, &res, NULL, 1);
365 quote_c_style(two, &res, NULL, 1);
366 strbuf_addch(&res, '"');
367 } else {
368 strbuf_addstr(&res, one);
369 strbuf_addstr(&res, two);
370 }
371 return strbuf_detach(&res, NULL);
372}
373static const char *external_diff(void)
375{
376 static const char *external_diff_cmd = NULL;
377 static int done_preparing = 0;
378 if (done_preparing)
380 return external_diff_cmd;
381 external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
382 if (!external_diff_cmd)
383 external_diff_cmd = external_diff_cmd_cfg;
384 done_preparing = 1;
385 return external_diff_cmd;
386}
387/*
389 * Keep track of files used for diffing. Sometimes such an entry
390 * refers to a temporary file, sometimes to an existing file, and
391 * sometimes to "/dev/null".
392 */
393static struct diff_tempfile {
394 /*
395 * filename external diff should read from, or NULL if this
396 * entry is currently not in use:
397 */
398 const char *name;
399 char hex[GIT_MAX_HEXSZ + 1];
401 char mode[10];
402 /*
404 * If this diff_tempfile instance refers to a temporary file,
405 * this tempfile object is used to manage its lifetime.
406 */
407 struct tempfile tempfile;
408} diff_temp[2];
409struct emit_callback {
411 int color_diff;
412 unsigned ws_rule;
413 int blank_at_eof_in_preimage;
414 int blank_at_eof_in_postimage;
415 int lno_in_preimage;
416 int lno_in_postimage;
417 const char **label_path;
418 struct diff_words_data *diff_words;
419 struct diff_options *opt;
420 struct strbuf *header;
421};
422static int count_lines(const char *data, int size)
424{
425 int count, ch, completely_empty = 1, nl_just_seen = 0;
426 count = 0;
427 while (0 < size--) {
428 ch = *data++;
429 if (ch == '\n') {
430 count++;
431 nl_just_seen = 1;
432 completely_empty = 0;
433 }
434 else {
435 nl_just_seen = 0;
436 completely_empty = 0;
437 }
438 }
439 if (completely_empty)
440 return 0;
441 if (!nl_just_seen)
442 count++; /* no trailing newline */
443 return count;
444}
445static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
447{
448 if (!DIFF_FILE_VALID(one)) {
449 mf->ptr = (char *)""; /* does not matter */
450 mf->size = 0;
451 return 0;
452 }
453 else if (diff_populate_filespec(one, 0))
454 return -1;
455 mf->ptr = one->data;
457 mf->size = one->size;
458 return 0;
459}
460/* like fill_mmfile, but only for size, so we can avoid retrieving blob */
462static unsigned long diff_filespec_size(struct diff_filespec *one)
463{
464 if (!DIFF_FILE_VALID(one))
465 return 0;
466 diff_populate_filespec(one, CHECK_SIZE_ONLY);
467 return one->size;
468}
469static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
471{
472 char *ptr = mf->ptr;
473 long size = mf->size;
474 int cnt = 0;
475 if (!size)
477 return cnt;
478 ptr += size - 1; /* pointing at the very end */
479 if (*ptr != '\n')
480 ; /* incomplete line */
481 else
482 ptr--; /* skip the last LF */
483 while (mf->ptr < ptr) {
484 char *prev_eol;
485 for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
486 if (*prev_eol == '\n')
487 break;
488 if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
489 break;
490 cnt++;
491 ptr = prev_eol - 1;
492 }
493 return cnt;
494}
495static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
497 struct emit_callback *ecbdata)
498{
499 int l1, l2, at;
500 unsigned ws_rule = ecbdata->ws_rule;
501 l1 = count_trailing_blank(mf1, ws_rule);
502 l2 = count_trailing_blank(mf2, ws_rule);
503 if (l2 <= l1) {
504 ecbdata->blank_at_eof_in_preimage = 0;
505 ecbdata->blank_at_eof_in_postimage = 0;
506 return;
507 }
508 at = count_lines(mf1->ptr, mf1->size);
509 ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
510 at = count_lines(mf2->ptr, mf2->size);
512 ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
513}
514static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
516 int first, const char *line, int len)
517{
518 int has_trailing_newline, has_trailing_carriage_return;
519 int nofirst;
520 FILE *file = o->file;
521 fputs(diff_line_prefix(o), file);
523 if (len == 0) {
525 has_trailing_newline = (first == '\n');
526 has_trailing_carriage_return = (!has_trailing_newline &&
527 (first == '\r'));
528 nofirst = has_trailing_newline || has_trailing_carriage_return;
529 } else {
530 has_trailing_newline = (len > 0 && line[len-1] == '\n');
531 if (has_trailing_newline)
532 len--;
533 has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
534 if (has_trailing_carriage_return)
535 len--;
536 nofirst = 0;
537 }
538 if (len || !nofirst) {
540 fputs(set, file);
541 if (!nofirst)
542 fputc(first, file);
543 fwrite(line, len, 1, file);
544 fputs(reset, file);
545 }
546 if (has_trailing_carriage_return)
547 fputc('\r', file);
548 if (has_trailing_newline)
549 fputc('\n', file);
550}
551static void emit_line(struct diff_options *o, const char *set, const char *reset,
553 const char *line, int len)
554{
555 emit_line_0(o, set, reset, line[0], line+1, len-1);
556}
557static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
559{
560 if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
561 ecbdata->blank_at_eof_in_preimage &&
562 ecbdata->blank_at_eof_in_postimage &&
563 ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
564 ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
565 return 0;
566 return ws_blank_line(line, len, ecbdata->ws_rule);
567}
568static void emit_line_checked(const char *reset,
570 struct emit_callback *ecbdata,
571 const char *line, int len,
572 enum color_diff color,
573 unsigned ws_error_highlight,
574 char sign)
575{
576 const char *set = diff_get_color(ecbdata->color_diff, color);
577 const char *ws = NULL;
578 if (ecbdata->opt->ws_error_highlight & ws_error_highlight) {
580 ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
581 if (!*ws)
582 ws = NULL;
583 }
584 if (!ws)
586 emit_line_0(ecbdata->opt, set, reset, sign, line, len);
587 else if (sign == '+' && new_blank_line_at_eof(ecbdata, line, len))
588 /* Blank line at EOF - paint '+' as well */
589 emit_line_0(ecbdata->opt, ws, reset, sign, line, len);
590 else {
591 /* Emit just the prefix, then the rest. */
592 emit_line_0(ecbdata->opt, set, reset, sign, "", 0);
593 ws_check_emit(line, len, ecbdata->ws_rule,
594 ecbdata->opt->file, set, reset, ws);
595 }
596}
597static void emit_add_line(const char *reset,
599 struct emit_callback *ecbdata,
600 const char *line, int len)
601{
602 emit_line_checked(reset, ecbdata, line, len,
603 DIFF_FILE_NEW, WSEH_NEW, '+');
604}
605static void emit_del_line(const char *reset,
607 struct emit_callback *ecbdata,
608 const char *line, int len)
609{
610 emit_line_checked(reset, ecbdata, line, len,
611 DIFF_FILE_OLD, WSEH_OLD, '-');
612}
613static void emit_context_line(const char *reset,
615 struct emit_callback *ecbdata,
616 const char *line, int len)
617{
618 emit_line_checked(reset, ecbdata, line, len,
619 DIFF_CONTEXT, WSEH_CONTEXT, ' ');
620}
621static void emit_hunk_header(struct emit_callback *ecbdata,
623 const char *line, int len)
624{
625 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
626 const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
627 const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
628 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
629 static const char atat[2] = { '@', '@' };
630 const char *cp, *ep;
631 struct strbuf msgbuf = STRBUF_INIT;
632 int org_len = len;
633 int i = 1;
634 /*
636 * As a hunk header must begin with "@@ -<old>, +<new> @@",
637 * it always is at least 10 bytes long.
638 */
639 if (len < 10 ||
640 memcmp(line, atat, 2) ||
641 !(ep = memmem(line + 2, len - 2, atat, 2))) {
642 emit_line(ecbdata->opt, context, reset, line, len);
643 return;
644 }
645 ep += 2; /* skip over @@ */
646 /* The hunk header in fraginfo color */
648 strbuf_addstr(&msgbuf, frag);
649 strbuf_add(&msgbuf, line, ep - line);
650 strbuf_addstr(&msgbuf, reset);
651 /*
653 * trailing "\r\n"
654 */
655 for ( ; i < 3; i++)
656 if (line[len - i] == '\r' || line[len - i] == '\n')
657 len--;
658 /* blank before the func header */
660 for (cp = ep; ep - line < len; ep++)
661 if (*ep != ' ' && *ep != '\t')
662 break;
663 if (ep != cp) {
664 strbuf_addstr(&msgbuf, context);
665 strbuf_add(&msgbuf, cp, ep - cp);
666 strbuf_addstr(&msgbuf, reset);
667 }
668 if (ep < line + len) {
670 strbuf_addstr(&msgbuf, func);
671 strbuf_add(&msgbuf, ep, line + len - ep);
672 strbuf_addstr(&msgbuf, reset);
673 }
674 strbuf_add(&msgbuf, line + len, org_len - len);
676 emit_line(ecbdata->opt, "", "", msgbuf.buf, msgbuf.len);
677 strbuf_release(&msgbuf);
678}
679static struct diff_tempfile *claim_diff_tempfile(void) {
681 int i;
682 for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
683 if (!diff_temp[i].name)
684 return diff_temp + i;
685 die("BUG: diff is failing to clean up its tempfiles");
686}
687static void remove_tempfile(void)
689{
690 int i;
691 for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
692 if (is_tempfile_active(&diff_temp[i].tempfile))
693 delete_tempfile(&diff_temp[i].tempfile);
694 diff_temp[i].name = NULL;
695 }
696}
697static void print_line_count(FILE *file, int count)
699{
700 switch (count) {
701 case 0:
702 fprintf(file, "0,0");
703 break;
704 case 1:
705 fprintf(file, "1");
706 break;
707 default:
708 fprintf(file, "1,%d", count);
709 break;
710 }
711}
712static void emit_rewrite_lines(struct emit_callback *ecb,
714 int prefix, const char *data, int size)
715{
716 const char *endp = NULL;
717 static const char *nneof = " No newline at end of file\n";
718 const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
719 while (0 < size) {
721 int len;
722 endp = memchr(data, '\n', size);
724 len = endp ? (endp - data + 1) : size;
725 if (prefix != '+') {
726 ecb->lno_in_preimage++;
727 emit_del_line(reset, ecb, data, len);
728 } else {
729 ecb->lno_in_postimage++;
730 emit_add_line(reset, ecb, data, len);
731 }
732 size -= len;
733 data += len;
734 }
735 if (!endp) {
736 const char *context = diff_get_color(ecb->color_diff,
737 DIFF_CONTEXT);
738 putc('\n', ecb->opt->file);
739 emit_line_0(ecb->opt, context, reset, '\\',
740 nneof, strlen(nneof));
741 }
742}
743static void emit_rewrite_diff(const char *name_a,
745 const char *name_b,
746 struct diff_filespec *one,
747 struct diff_filespec *two,
748 struct userdiff_driver *textconv_one,
749 struct userdiff_driver *textconv_two,
750 struct diff_options *o)
751{
752 int lc_a, lc_b;
753 const char *name_a_tab, *name_b_tab;
754 const char *metainfo = diff_get_color(o->use_color, DIFF_METAINFO);
755 const char *fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
756 const char *reset = diff_get_color(o->use_color, DIFF_RESET);
757 static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
758 const char *a_prefix, *b_prefix;
759 char *data_one, *data_two;
760 size_t size_one, size_two;
761 struct emit_callback ecbdata;
762 const char *line_prefix = diff_line_prefix(o);
763 if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
765 a_prefix = o->b_prefix;
766 b_prefix = o->a_prefix;
767 } else {
768 a_prefix = o->a_prefix;
769 b_prefix = o->b_prefix;
770 }
771 name_a += (*name_a == '/');
773 name_b += (*name_b == '/');
774 name_a_tab = strchr(name_a, ' ') ? "\t" : "";
775 name_b_tab = strchr(name_b, ' ') ? "\t" : "";
776 strbuf_reset(&a_name);
778 strbuf_reset(&b_name);
779 quote_two_c_style(&a_name, a_prefix, name_a, 0);
780 quote_two_c_style(&b_name, b_prefix, name_b, 0);
781 size_one = fill_textconv(textconv_one, one, &data_one);
783 size_two = fill_textconv(textconv_two, two, &data_two);
784 memset(&ecbdata, 0, sizeof(ecbdata));
786 ecbdata.color_diff = want_color(o->use_color);
787 ecbdata.ws_rule = whitespace_rule(name_b);
788 ecbdata.opt = o;
789 if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
790 mmfile_t mf1, mf2;
791 mf1.ptr = (char *)data_one;
792 mf2.ptr = (char *)data_two;
793 mf1.size = size_one;
794 mf2.size = size_two;
795 check_blank_at_eof(&mf1, &mf2, &ecbdata);
796 }
797 ecbdata.lno_in_preimage = 1;
798 ecbdata.lno_in_postimage = 1;
799 lc_a = count_lines(data_one, size_one);
801 lc_b = count_lines(data_two, size_two);
802 fprintf(o->file,
803 "%s%s--- %s%s%s\n%s%s+++ %s%s%s\n%s%s@@ -",
804 line_prefix, metainfo, a_name.buf, name_a_tab, reset,
805 line_prefix, metainfo, b_name.buf, name_b_tab, reset,
806 line_prefix, fraginfo);
807 if (!o->irreversible_delete)
808 print_line_count(o->file, lc_a);
809 else
810 fprintf(o->file, "?,?");
811 fprintf(o->file, " +");
812 print_line_count(o->file, lc_b);
813 fprintf(o->file, " @@%s\n", reset);
814 if (lc_a && !o->irreversible_delete)
815 emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
816 if (lc_b)
817 emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
818 if (textconv_one)
819 free((char *)data_one);
820 if (textconv_two)
821 free((char *)data_two);
822}
823struct diff_words_buffer {
825 mmfile_t text;
826 long alloc;
827 struct diff_words_orig {
828 const char *begin, *end;
829 } *orig;
830 int orig_nr, orig_alloc;
831};
832static void diff_words_append(char *line, unsigned long len,
834 struct diff_words_buffer *buffer)
835{
836 ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
837 line++;
838 len--;
839 memcpy(buffer->text.ptr + buffer->text.size, line, len);
840 buffer->text.size += len;
841 buffer->text.ptr[buffer->text.size] = '\0';
842}
843struct diff_words_style_elem {
845 const char *prefix;
846 const char *suffix;
847 const char *color; /* NULL; filled in by the setup code if
848 * color is enabled */
849};
850struct diff_words_style {
852 enum diff_words_type type;
853 struct diff_words_style_elem new, old, ctx;
854 const char *newline;
855};
856static struct diff_words_style diff_words_styles[] = {
858 { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
859 { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
860 { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
861};
862struct diff_words_data {
864 struct diff_words_buffer minus, plus;
865 const char *current_plus;
866 int last_minus;
867 struct diff_options *opt;
868 regex_t *word_regex;
869 enum diff_words_type type;
870 struct diff_words_style *style;
871};
872static int fn_out_diff_words_write_helper(FILE *fp,
874 struct diff_words_style_elem *st_el,
875 const char *newline,
876 size_t count, const char *buf,
877 const char *line_prefix)
878{
879 int print = 0;
880 while (count) {
882 char *p = memchr(buf, '\n', count);
883 if (print)
884 fputs(line_prefix, fp);
885 if (p != buf) {
886 if (st_el->color && fputs(st_el->color, fp) < 0)
887 return -1;
888 if (fputs(st_el->prefix, fp) < 0 ||
889 fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
890 fputs(st_el->suffix, fp) < 0)
891 return -1;
892 if (st_el->color && *st_el->color
893 && fputs(GIT_COLOR_RESET, fp) < 0)
894 return -1;
895 }
896 if (!p)
897 return 0;
898 if (fputs(newline, fp) < 0)
899 return -1;
900 count -= p + 1 - buf;
901 buf = p + 1;
902 print = 1;
903 }
904 return 0;
905}
906/*
908 * '--color-words' algorithm can be described as:
909 *
910 * 1. collect the minus/plus lines of a diff hunk, divided into
911 * minus-lines and plus-lines;
912 *
913 * 2. break both minus-lines and plus-lines into words and
914 * place them into two mmfile_t with one word for each line;
915 *
916 * 3. use xdiff to run diff on the two mmfile_t to get the words level diff;
917 *
918 * And for the common parts of the both file, we output the plus side text.
919 * diff_words->current_plus is used to trace the current position of the plus file
920 * which printed. diff_words->last_minus is used to trace the last minus word
921 * printed.
922 *
923 * For '--graph' to work with '--color-words', we need to output the graph prefix
924 * on each line of color words output. Generally, there are two conditions on
925 * which we should output the prefix.
926 *
927 * 1. diff_words->last_minus == 0 &&
928 * diff_words->current_plus == diff_words->plus.text.ptr
929 *
930 * that is: the plus text must start as a new line, and if there is no minus
931 * word printed, a graph prefix must be printed.
932 *
933 * 2. diff_words->current_plus > diff_words->plus.text.ptr &&
934 * *(diff_words->current_plus - 1) == '\n'
935 *
936 * that is: a graph prefix must be printed following a '\n'
937 */
938static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
939{
940 if ((diff_words->last_minus == 0 &&
941 diff_words->current_plus == diff_words->plus.text.ptr) ||
942 (diff_words->current_plus > diff_words->plus.text.ptr &&
943 *(diff_words->current_plus - 1) == '\n')) {
944 return 1;
945 } else {
946 return 0;
947 }
948}
949static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
951{
952 struct diff_words_data *diff_words = priv;
953 struct diff_words_style *style = diff_words->style;
954 int minus_first, minus_len, plus_first, plus_len;
955 const char *minus_begin, *minus_end, *plus_begin, *plus_end;
956 struct diff_options *opt = diff_words->opt;
957 const char *line_prefix;
958 if (line[0] != '@' || parse_hunk_header(line, len,
960 &minus_first, &minus_len, &plus_first, &plus_len))
961 return;
962 assert(opt);
964 line_prefix = diff_line_prefix(opt);
965 /* POSIX requires that first be decremented by one if len == 0... */
967 if (minus_len) {
968 minus_begin = diff_words->minus.orig[minus_first].begin;
969 minus_end =
970 diff_words->minus.orig[minus_first + minus_len - 1].end;
971 } else
972 minus_begin = minus_end =
973 diff_words->minus.orig[minus_first].end;
974 if (plus_len) {
976 plus_begin = diff_words->plus.orig[plus_first].begin;
977 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
978 } else
979 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
980 if (color_words_output_graph_prefix(diff_words)) {
982 fputs(line_prefix, diff_words->opt->file);
983 }
984 if (diff_words->current_plus != plus_begin) {
985 fn_out_diff_words_write_helper(diff_words->opt->file,
986 &style->ctx, style->newline,
987 plus_begin - diff_words->current_plus,
988 diff_words->current_plus, line_prefix);
989 if (*(plus_begin - 1) == '\n')
990 fputs(line_prefix, diff_words->opt->file);
991 }
992 if (minus_begin != minus_end) {
993 fn_out_diff_words_write_helper(diff_words->opt->file,
994 &style->old, style->newline,
995 minus_end - minus_begin, minus_begin,
996 line_prefix);
997 }
998 if (plus_begin != plus_end) {
999 fn_out_diff_words_write_helper(diff_words->opt->file,
1000 &style->new, style->newline,
1001 plus_end - plus_begin, plus_begin,
1002 line_prefix);
1003 }
1004 diff_words->current_plus = plus_end;
1006 diff_words->last_minus = minus_first;
1007}
1008/* This function starts looking at *begin, and returns 0 iff a word was found. */
1010static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
1011 int *begin, int *end)
1012{
1013 if (word_regex && *begin < buffer->size) {
1014 regmatch_t match[1];
1015 if (!regexec_buf(word_regex, buffer->ptr + *begin,
1016 buffer->size - *begin, 1, match, 0)) {
1017 char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
1018 '\n', match[0].rm_eo - match[0].rm_so);
1019 *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
1020 *begin += match[0].rm_so;
1021 return *begin >= *end;
1022 }
1023 return -1;
1024 }
1025 /* find the next word */
1027 while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
1028 (*begin)++;
1029 if (*begin >= buffer->size)
1030 return -1;
1031 /* find the end of the word */
1033 *end = *begin + 1;
1034 while (*end < buffer->size && !isspace(buffer->ptr[*end]))
1035 (*end)++;
1036 return 0;
1038}
1039/*
1041 * This function splits the words in buffer->text, stores the list with
1042 * newline separator into out, and saves the offsets of the original words
1043 * in buffer->orig.
1044 */
1045static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
1046 regex_t *word_regex)
1047{
1048 int i, j;
1049 long alloc = 0;
1050 out->size = 0;
1052 out->ptr = NULL;
1053 /* fake an empty "0th" word */
1055 ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
1056 buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
1057 buffer->orig_nr = 1;
1058 for (i = 0; i < buffer->text.size; i++) {
1060 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
1061 return;
1062 /* store original boundaries */
1064 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
1065 buffer->orig_alloc);
1066 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
1067 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
1068 buffer->orig_nr++;
1069 /* store one word */
1071 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
1072 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
1073 out->ptr[out->size + j - i] = '\n';
1074 out->size += j - i + 1;
1075 i = j - 1;
1077 }
1078}
1079/* this executes the word diff on the accumulated buffers */
1081static void diff_words_show(struct diff_words_data *diff_words)
1082{
1083 xpparam_t xpp;
1084 xdemitconf_t xecfg;
1085 mmfile_t minus, plus;
1086 struct diff_words_style *style = diff_words->style;
1087 struct diff_options *opt = diff_words->opt;
1089 const char *line_prefix;
1090 assert(opt);
1092 line_prefix = diff_line_prefix(opt);
1093 /* special case: only removal */
1095 if (!diff_words->plus.text.size) {
1096 fputs(line_prefix, diff_words->opt->file);
1097 fn_out_diff_words_write_helper(diff_words->opt->file,
1098 &style->old, style->newline,
1099 diff_words->minus.text.size,
1100 diff_words->minus.text.ptr, line_prefix);
1101 diff_words->minus.text.size = 0;
1102 return;
1103 }
1104 diff_words->current_plus = diff_words->plus.text.ptr;
1106 diff_words->last_minus = 0;
1107 memset(&xpp, 0, sizeof(xpp));
1109 memset(&xecfg, 0, sizeof(xecfg));
1110 diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
1111 diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
1112 xpp.flags = 0;
1113 /* as only the hunk header will be parsed, we need a 0-context */
1114 xecfg.ctxlen = 0;
1115 if (xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
1116 &xpp, &xecfg))
1117 die("unable to generate word diff");
1118 free(minus.ptr);
1119 free(plus.ptr);
1120 if (diff_words->current_plus != diff_words->plus.text.ptr +
1121 diff_words->plus.text.size) {
1122 if (color_words_output_graph_prefix(diff_words))
1123 fputs(line_prefix, diff_words->opt->file);
1124 fn_out_diff_words_write_helper(diff_words->opt->file,
1125 &style->ctx, style->newline,
1126 diff_words->plus.text.ptr + diff_words->plus.text.size
1127 - diff_words->current_plus, diff_words->current_plus,
1128 line_prefix);
1129 }
1130 diff_words->minus.text.size = diff_words->plus.text.size = 0;
1131}
1132/* In "color-words" mode, show word-diff of words accumulated in the buffer */
1134static void diff_words_flush(struct emit_callback *ecbdata)
1135{
1136 if (ecbdata->diff_words->minus.text.size ||
1137 ecbdata->diff_words->plus.text.size)
1138 diff_words_show(ecbdata->diff_words);
1139}
1140static void diff_filespec_load_driver(struct diff_filespec *one)
1142{
1143 /* Use already-loaded driver */
1144 if (one->driver)
1145 return;
1146 if (S_ISREG(one->mode))
1148 one->driver = userdiff_find_by_path(one->path);
1149 /* Fallback to default settings */
1151 if (!one->driver)
1152 one->driver = userdiff_find_by_name("default");
1153}
1154static const char *userdiff_word_regex(struct diff_filespec *one)
1156{
1157 diff_filespec_load_driver(one);
1158 return one->driver->word_regex;
1159}
1160static void init_diff_words_data(struct emit_callback *ecbdata,
1162 struct diff_options *orig_opts,
1163 struct diff_filespec *one,
1164 struct diff_filespec *two)
1165{
1166 int i;
1167 struct diff_options *o = xmalloc(sizeof(struct diff_options));
1168 memcpy(o, orig_opts, sizeof(struct diff_options));
1169 ecbdata->diff_words =
1171 xcalloc(1, sizeof(struct diff_words_data));
1172 ecbdata->diff_words->type = o->word_diff;
1173 ecbdata->diff_words->opt = o;
1174 if (!o->word_regex)
1175 o->word_regex = userdiff_word_regex(one);
1176 if (!o->word_regex)
1177 o->word_regex = userdiff_word_regex(two);
1178 if (!o->word_regex)
1179 o->word_regex = diff_word_regex_cfg;
1180 if (o->word_regex) {
1181 ecbdata->diff_words->word_regex = (regex_t *)
1182 xmalloc(sizeof(regex_t));
1183 if (regcomp(ecbdata->diff_words->word_regex,
1184 o->word_regex,
1185 REG_EXTENDED | REG_NEWLINE))
1186 die ("Invalid regular expression: %s",
1187 o->word_regex);
1188 }
1189 for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1190 if (o->word_diff == diff_words_styles[i].type) {
1191 ecbdata->diff_words->style =
1192 &diff_words_styles[i];
1193 break;
1194 }
1195 }
1196 if (want_color(o->use_color)) {
1197 struct diff_words_style *st = ecbdata->diff_words->style;
1198 st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1199 st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1200 st->ctx.color = diff_get_color_opt(o, DIFF_CONTEXT);
1201 }
1202}
1203static void free_diff_words_data(struct emit_callback *ecbdata)
1205{
1206 if (ecbdata->diff_words) {
1207 diff_words_flush(ecbdata);
1208 free (ecbdata->diff_words->opt);
1209 free (ecbdata->diff_words->minus.text.ptr);
1210 free (ecbdata->diff_words->minus.orig);
1211 free (ecbdata->diff_words->plus.text.ptr);
1212 free (ecbdata->diff_words->plus.orig);
1213 if (ecbdata->diff_words->word_regex) {
1214 regfree(ecbdata->diff_words->word_regex);
1215 free(ecbdata->diff_words->word_regex);
1216 }
1217 FREE_AND_NULL(ecbdata->diff_words);
1218 }
1219}
1220const char *diff_get_color(int diff_use_color, enum color_diff ix)
1222{
1223 if (want_color(diff_use_color))
1224 return diff_colors[ix];
1225 return "";
1226}
1227const char *diff_line_prefix(struct diff_options *opt)
1229{
1230 struct strbuf *msgbuf;
1231 if (!opt->output_prefix)
1232 return "";
1233 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
1235 return msgbuf->buf;
1236}
1237static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1239{
1240 const char *cp;
1241 unsigned long allot;
1242 size_t l = len;
1243 cp = line;
1245 allot = l;
1246 while (0 < l) {
1247 (void) utf8_width(&cp, &l);
1248 if (!cp)
1249 break; /* truncated in the middle? */
1250 }
1251 return allot - l;
1252}
1253static void find_lno(const char *line, struct emit_callback *ecbdata)
1255{
1256 const char *p;
1257 ecbdata->lno_in_preimage = 0;
1258 ecbdata->lno_in_postimage = 0;
1259 p = strchr(line, '-');
1260 if (!p)
1261 return; /* cannot happen */
1262 ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1263 p = strchr(p, '+');
1264 if (!p)
1265 return; /* cannot happen */
1266 ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1267}
1268static void fn_out_consume(void *priv, char *line, unsigned long len)
1270{
1271 struct emit_callback *ecbdata = priv;
1272 const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
1273 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
1274 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1275 struct diff_options *o = ecbdata->opt;
1276 const char *line_prefix = diff_line_prefix(o);
1277 o->found_changes = 1;
1279 if (ecbdata->header) {
1281 fprintf(o->file, "%s", ecbdata->header->buf);
1282 strbuf_reset(ecbdata->header);
1283 ecbdata->header = NULL;
1284 }
1285 if (ecbdata->label_path[0]) {
1287 const char *name_a_tab, *name_b_tab;
1288 name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
1290 name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
1291 fprintf(o->file, "%s%s--- %s%s%s\n",
1293 line_prefix, meta, ecbdata->label_path[0], reset, name_a_tab);
1294 fprintf(o->file, "%s%s+++ %s%s%s\n",
1295 line_prefix, meta, ecbdata->label_path[1], reset, name_b_tab);
1296 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1297 }
1298 if (diff_suppress_blank_empty
1300 && len == 2 && line[0] == ' ' && line[1] == '\n') {
1301 line[0] = '\n';
1302 len = 1;
1303 }
1304 if (line[0] == '@') {
1306 if (ecbdata->diff_words)
1307 diff_words_flush(ecbdata);
1308 len = sane_truncate_line(ecbdata, line, len);
1309 find_lno(line, ecbdata);
1310 emit_hunk_header(ecbdata, line, len);
1311 if (line[len-1] != '\n')
1312 putc('\n', o->file);
1313 return;
1314 }
1315 if (ecbdata->diff_words) {
1317 if (line[0] == '-') {
1318 diff_words_append(line, len,
1319 &ecbdata->diff_words->minus);
1320 return;
1321 } else if (line[0] == '+') {
1322 diff_words_append(line, len,
1323 &ecbdata->diff_words->plus);
1324 return;
1325 } else if (starts_with(line, "\\ ")) {
1326 /*
1327 * Eat the "no newline at eof" marker as if we
1328 * saw a "+" or "-" line with nothing on it,
1329 * and return without diff_words_flush() to
1330 * defer processing. If this is the end of
1331 * preimage, more "+" lines may come after it.
1332 */
1333 return;
1334 }
1335 diff_words_flush(ecbdata);
1336 if (ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN) {
1337 emit_line(o, context, reset, line, len);
1338 fputs("~\n", o->file);
1339 } else {
1340 /*
1341 * Skip the prefix character, if any. With
1342 * diff_suppress_blank_empty, there may be
1343 * none.
1344 */
1345 if (line[0] != '\n') {
1346 line++;
1347 len--;
1348 }
1349 emit_line(o, context, reset, line, len);
1350 }
1351 return;
1352 }
1353 switch (line[0]) {
1355 case '+':
1356 ecbdata->lno_in_postimage++;
1357 emit_add_line(reset, ecbdata, line + 1, len - 1);
1358 break;
1359 case '-':
1360 ecbdata->lno_in_preimage++;
1361 emit_del_line(reset, ecbdata, line + 1, len - 1);
1362 break;
1363 case ' ':
1364 ecbdata->lno_in_postimage++;
1365 ecbdata->lno_in_preimage++;
1366 emit_context_line(reset, ecbdata, line + 1, len - 1);
1367 break;
1368 default:
1369 /* incomplete line at the end */
1370 ecbdata->lno_in_preimage++;
1371 emit_line(o, diff_get_color(ecbdata->color_diff, DIFF_CONTEXT),
1372 reset, line, len);
1373 break;
1374 }
1375}
1376static char *pprint_rename(const char *a, const char *b)
1378{
1379 const char *old = a;
1380 const char *new = b;
1381 struct strbuf name = STRBUF_INIT;
1382 int pfx_length, sfx_length;
1383 int pfx_adjust_for_slash;
1384 int len_a = strlen(a);
1385 int len_b = strlen(b);
1386 int a_midlen, b_midlen;
1387 int qlen_a = quote_c_style(a, NULL, NULL, 0);
1388 int qlen_b = quote_c_style(b, NULL, NULL, 0);
1389 if (qlen_a || qlen_b) {
1391 quote_c_style(a, &name, NULL, 0);
1392 strbuf_addstr(&name, " => ");
1393 quote_c_style(b, &name, NULL, 0);
1394 return strbuf_detach(&name, NULL);
1395 }
1396 /* Find common prefix */
1398 pfx_length = 0;
1399 while (*old && *new && *old == *new) {
1400 if (*old == '/')
1401 pfx_length = old - a + 1;
1402 old++;
1403 new++;
1404 }
1405 /* Find common suffix */
1407 old = a + len_a;
1408 new = b + len_b;
1409 sfx_length = 0;
1410 /*
1411 * If there is a common prefix, it must end in a slash. In
1412 * that case we let this loop run 1 into the prefix to see the
1413 * same slash.
1414 *
1415 * If there is no common prefix, we cannot do this as it would
1416 * underrun the input strings.
1417 */
1418 pfx_adjust_for_slash = (pfx_length ? 1 : 0);
1419 while (a + pfx_length - pfx_adjust_for_slash <= old &&
1420 b + pfx_length - pfx_adjust_for_slash <= new &&
1421 *old == *new) {
1422 if (*old == '/')
1423 sfx_length = len_a - (old - a);
1424 old--;
1425 new--;
1426 }
1427 /*
1429 * pfx{mid-a => mid-b}sfx
1430 * {pfx-a => pfx-b}sfx
1431 * pfx{sfx-a => sfx-b}
1432 * name-a => name-b
1433 */
1434 a_midlen = len_a - pfx_length - sfx_length;
1435 b_midlen = len_b - pfx_length - sfx_length;
1436 if (a_midlen < 0)
1437 a_midlen = 0;
1438 if (b_midlen < 0)
1439 b_midlen = 0;
1440 strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1442 if (pfx_length + sfx_length) {
1443 strbuf_add(&name, a, pfx_length);
1444 strbuf_addch(&name, '{');
1445 }
1446 strbuf_add(&name, a + pfx_length, a_midlen);
1447 strbuf_addstr(&name, " => ");
1448 strbuf_add(&name, b + pfx_length, b_midlen);
1449 if (pfx_length + sfx_length) {
1450 strbuf_addch(&name, '}');
1451 strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1452 }
1453 return strbuf_detach(&name, NULL);
1454}
1455struct diffstat_t {
1457 int nr;
1458 int alloc;
1459 struct diffstat_file {
1460 char *from_name;
1461 char *name;
1462 char *print_name;
1463 unsigned is_unmerged:1;
1464 unsigned is_binary:1;
1465 unsigned is_renamed:1;
1466 unsigned is_interesting:1;
1467 uintmax_t added, deleted;
1468 } **files;
1469};
1470static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1472 const char *name_a,
1473 const char *name_b)
1474{
1475 struct diffstat_file *x;
1476 x = xcalloc(1, sizeof(*x));
1477 ALLOC_GROW(diffstat->files, diffstat->nr + 1, diffstat->alloc);
1478 diffstat->files[diffstat->nr++] = x;
1479 if (name_b) {
1480 x->from_name = xstrdup(name_a);
1481 x->name = xstrdup(name_b);
1482 x->is_renamed = 1;
1483 }
1484 else {
1485 x->from_name = NULL;
1486 x->name = xstrdup(name_a);
1487 }
1488 return x;
1489}
1490static void diffstat_consume(void *priv, char *line, unsigned long len)
1492{
1493 struct diffstat_t *diffstat = priv;
1494 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1495 if (line[0] == '+')
1497 x->added++;
1498 else if (line[0] == '-')
1499 x->deleted++;
1500}
1501const char mime_boundary_leader[] = "------------";
1503static int scale_linear(int it, int width, int max_change)
1505{
1506 if (!it)
1507 return 0;
1508 /*
1509 * make sure that at least one '-' or '+' is printed if
1510 * there is any change to this path. The easiest way is to
1511 * scale linearly as if the alloted width is one column shorter
1512 * than it is, and then add 1 to the result.
1513 */
1514 return 1 + (it * (width - 1) / max_change);
1515}
1516static void show_name(FILE *file,
1518 const char *prefix, const char *name, int len)
1519{
1520 fprintf(file, " %s%-*s |", prefix, len, name);
1521}
1522static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1524{
1525 if (cnt <= 0)
1526 return;
1527 fprintf(file, "%s", set);
1528 while (cnt--)
1529 putc(ch, file);
1530 fprintf(file, "%s", reset);
1531}
1532static void fill_print_name(struct diffstat_file *file)
1534{
1535 char *pname;
1536 if (file->print_name)
1538 return;
1539 if (!file->is_renamed) {
1541 struct strbuf buf = STRBUF_INIT;
1542 if (quote_c_style(file->name, &buf, NULL, 0)) {
1543 pname = strbuf_detach(&buf, NULL);
1544 } else {
1545 pname = file->name;
1546 strbuf_release(&buf);
1547 }
1548 } else {
1549 pname = pprint_rename(file->from_name, file->name);
1550 }
1551 file->print_name = pname;
1552}
1553int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1555{
1556 struct strbuf sb = STRBUF_INIT;
1557 int ret;
1558 if (!files) {
1560 assert(insertions == 0 && deletions == 0);
1561 return fprintf(fp, "%s\n", " 0 files changed");
1562 }
1563 strbuf_addf(&sb,
1565 (files == 1) ? " %d file changed" : " %d files changed",
1566 files);
1567 /*
1569 * For binary diff, the caller may want to print "x files
1570 * changed" with insertions == 0 && deletions == 0.
1571 *
1572 * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1573 * is probably less confusing (i.e skip over "2 files changed
1574 * but nothing about added/removed lines? Is this a bug in Git?").
1575 */
1576 if (insertions || deletions == 0) {
1577 strbuf_addf(&sb,
1578 (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
1579 insertions);
1580 }
1581 if (deletions || insertions == 0) {
1583 strbuf_addf(&sb,
1584 (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
1585 deletions);
1586 }
1587 strbuf_addch(&sb, '\n');
1588 ret = fputs(sb.buf, fp);
1589 strbuf_release(&sb);
1590 return ret;
1591}
1592static void show_stats(struct diffstat_t *data, struct diff_options *options)
1594{
1595 int i, len, add, del, adds = 0, dels = 0;
1596 uintmax_t max_change = 0, max_len = 0;
1597 int total_files = data->nr, count;
1598 int width, name_width, graph_width, number_width = 0, bin_width = 0;
1599 const char *reset, *add_c, *del_c;
1600 const char *line_prefix = "";
1601 int extra_shown = 0;
1602 if (data->nr == 0)
1604 return;
1605 line_prefix = diff_line_prefix(options);
1607 count = options->stat_count ? options->stat_count : data->nr;
1608 reset = diff_get_color_opt(options, DIFF_RESET);
1610 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1611 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1612 /*
1614 * Find the longest filename and max number of changes
1615 */
1616 for (i = 0; (i < count) && (i < data->nr); i++) {
1617 struct diffstat_file *file = data->files[i];
1618 uintmax_t change = file->added + file->deleted;
1619 if (!file->is_interesting && (change == 0)) {
1621 count++; /* not shown == room for one more */
1622 continue;
1623 }
1624 fill_print_name(file);
1625 len = strlen(file->print_name);
1626 if (max_len < len)
1627 max_len = len;
1628 if (file->is_unmerged) {
1630 /* "Unmerged" is 8 characters */
1631 bin_width = bin_width < 8 ? 8 : bin_width;
1632 continue;
1633 }
1634 if (file->is_binary) {
1635 /* "Bin XXX -> YYY bytes" */
1636 int w = 14 + decimal_width(file->added)
1637 + decimal_width(file->deleted);
1638 bin_width = bin_width < w ? w : bin_width;
1639 /* Display change counts aligned with "Bin" */
1640 number_width = 3;
1641 continue;
1642 }
1643 if (max_change < change)
1645 max_change = change;
1646 }
1647 count = i; /* where we can stop scanning in data->files[] */
1648 /*
1650 * We have width = stat_width or term_columns() columns total.
1651 * We want a maximum of min(max_len, stat_name_width) for the name part.
1652 * We want a maximum of min(max_change, stat_graph_width) for the +- part.
1653 * We also need 1 for " " and 4 + decimal_width(max_change)
1654 * for " | NNNN " and one the empty column at the end, altogether
1655 * 6 + decimal_width(max_change).
1656 *
1657 * If there's not enough space, we will use the smaller of
1658 * stat_name_width (if set) and 5/8*width for the filename,
1659 * and the rest for constant elements + graph part, but no more
1660 * than stat_graph_width for the graph part.
1661 * (5/8 gives 50 for filename and 30 for the constant parts + graph
1662 * for the standard terminal size).
1663 *
1664 * In other words: stat_width limits the maximum width, and
1665 * stat_name_width fixes the maximum width of the filename,
1666 * and is also used to divide available columns if there
1667 * aren't enough.
1668 *
1669 * Binary files are displayed with "Bin XXX -> YYY bytes"
1670 * instead of the change count and graph. This part is treated
1671 * similarly to the graph part, except that it is not
1672 * "scaled". If total width is too small to accommodate the
1673 * guaranteed minimum width of the filename part and the
1674 * separators and this message, this message will "overflow"
1675 * making the line longer than the maximum width.
1676 */
1677 if (options->stat_width == -1)
1679 width = term_columns() - strlen(line_prefix);
1680 else
1681 width = options->stat_width ? options->stat_width : 80;
1682 number_width = decimal_width(max_change) > number_width ?
1683 decimal_width(max_change) : number_width;
1684 if (options->stat_graph_width == -1)
1686 options->stat_graph_width = diff_stat_graph_width;
1687 /*
1689 * Guarantee 3/8*16==6 for the graph part
1690 * and 5/8*16==10 for the filename part
1691 */
1692 if (width < 16 + 6 + number_width)
1693 width = 16 + 6 + number_width;
1694 /*
1696 * First assign sizes that are wanted, ignoring available width.
1697 * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
1698 * starting from "XXX" should fit in graph_width.
1699 */
1700 graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
1701 if (options->stat_graph_width &&
1702 options->stat_graph_width < graph_width)
1703 graph_width = options->stat_graph_width;
1704 name_width = (options->stat_name_width > 0 &&
1706 options->stat_name_width < max_len) ?
1707 options->stat_name_width : max_len;
1708 /*
1710 * Adjust adjustable widths not to exceed maximum width
1711 */
1712 if (name_width + number_width + 6 + graph_width > width) {
1713 if (graph_width > width * 3/8 - number_width - 6) {
1714 graph_width = width * 3/8 - number_width - 6;
1715 if (graph_width < 6)
1716 graph_width = 6;
1717 }
1718 if (options->stat_graph_width &&
1720 graph_width > options->stat_graph_width)
1721 graph_width = options->stat_graph_width;
1722 if (name_width > width - number_width - 6 - graph_width)
1723 name_width = width - number_width - 6 - graph_width;
1724 else
1725 graph_width = width - number_width - 6 - name_width;
1726 }
1727 /*
1729 * From here name_width is the width of the name area,
1730 * and graph_width is the width of the graph area.
1731 * max_change is used to scale graph properly.
1732 */
1733 for (i = 0; i < count; i++) {
1734 const char *prefix = "";
1735 struct diffstat_file *file = data->files[i];
1736 char *name = file->print_name;
1737 uintmax_t added = file->added;
1738 uintmax_t deleted = file->deleted;
1739 int name_len;
1740 if (!file->is_interesting && (added + deleted == 0))
1742 continue;
1743 /*
1745 * "scale" the filename
1746 */
1747 len = name_width;
1748 name_len = strlen(name);
1749 if (name_width < name_len) {
1750 char *slash;
1751 prefix = "...";
1752 len -= 3;
1753 name += name_len - len;
1754 slash = strchr(name, '/');
1755 if (slash)
1756 name = slash;
1757 }
1758 if (file->is_binary) {
1760 fprintf(options->file, "%s", line_prefix);
1761 show_name(options->file, prefix, name, len);
1762 fprintf(options->file, " %*s", number_width, "Bin");
1763 if (!added && !deleted) {
1764 putc('\n', options->file);
1765 continue;
1766 }
1767 fprintf(options->file, " %s%"PRIuMAX"%s",
1768 del_c, deleted, reset);
1769 fprintf(options->file, " -> ");
1770 fprintf(options->file, "%s%"PRIuMAX"%s",
1771 add_c, added, reset);
1772 fprintf(options->file, " bytes");
1773 fprintf(options->file, "\n");
1774 continue;
1775 }
1776 else if (file->is_unmerged) {
1777 fprintf(options->file, "%s", line_prefix);
1778 show_name(options->file, prefix, name, len);
1779 fprintf(options->file, " Unmerged\n");
1780 continue;
1781 }
1782 /*
1784 * scale the add/delete
1785 */
1786 add = added;
1787 del = deleted;
1788 if (graph_width <= max_change) {
1790 int total = scale_linear(add + del, graph_width, max_change);
1791 if (total < 2 && add && del)
1792 /* width >= 2 due to the sanity check */
1793 total = 2;
1794 if (add < del) {
1795 add = scale_linear(add, graph_width, max_change);
1796 del = total - add;
1797 } else {
1798 del = scale_linear(del, graph_width, max_change);
1799 add = total - del;
1800 }
1801 }
1802 fprintf(options->file, "%s", line_prefix);
1803 show_name(options->file, prefix, name, len);
1804 fprintf(options->file, " %*"PRIuMAX"%s",
1805 number_width, added + deleted,
1806 added + deleted ? " " : "");
1807 show_graph(options->file, '+', add, add_c, reset);
1808 show_graph(options->file, '-', del, del_c, reset);
1809 fprintf(options->file, "\n");
1810 }
1811 for (i = 0; i < data->nr; i++) {
1813 struct diffstat_file *file = data->files[i];
1814 uintmax_t added = file->added;
1815 uintmax_t deleted = file->deleted;
1816 if (file->is_unmerged ||
1818 (!file->is_interesting && (added + deleted == 0))) {
1819 total_files--;
1820 continue;
1821 }
1822 if (!file->is_binary) {
1824 adds += added;
1825 dels += deleted;
1826 }
1827 if (i < count)
1828 continue;
1829 if (!extra_shown)
1830 fprintf(options->file, "%s ...\n", line_prefix);
1831 extra_shown = 1;
1832 }
1833 fprintf(options->file, "%s", line_prefix);
1834 print_stat_summary(options->file, total_files, adds, dels);
1835}
1836static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1838{
1839 int i, adds = 0, dels = 0, total_files = data->nr;
1840 if (data->nr == 0)
1842 return;
1843 for (i = 0; i < data->nr; i++) {
1845 int added = data->files[i]->added;
1846 int deleted= data->files[i]->deleted;
1847 if (data->files[i]->is_unmerged ||
1849 (!data->files[i]->is_interesting && (added + deleted == 0))) {
1850 total_files--;
1851 } else if (!data->files[i]->is_binary) { /* don't count bytes */
1852 adds += added;
1853 dels += deleted;
1854 }
1855 }
1856 fprintf(options->file, "%s", diff_line_prefix(options));
1857 print_stat_summary(options->file, total_files, adds, dels);
1858}
1859static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1861{
1862 int i;
1863 if (data->nr == 0)
1865 return;
1866 for (i = 0; i < data->nr; i++) {
1868 struct diffstat_file *file = data->files[i];
1869 fprintf(options->file, "%s", diff_line_prefix(options));
1871 if (file->is_binary)
1873 fprintf(options->file, "-\t-\t");
1874 else
1875 fprintf(options->file,
1876 "%"PRIuMAX"\t%"PRIuMAX"\t",
1877 file->added, file->deleted);
1878 if (options->line_termination) {
1879 fill_print_name(file);
1880 if (!file->is_renamed)
1881 write_name_quoted(file->name, options->file,
1882 options->line_termination);
1883 else {
1884 fputs(file->print_name, options->file);
1885 putc(options->line_termination, options->file);
1886 }
1887 } else {
1888 if (file->is_renamed) {
1889 putc('\0', options->file);
1890 write_name_quoted(file->from_name, options->file, '\0');
1891 }
1892 write_name_quoted(file->name, options->file, '\0');
1893 }
1894 }
1895}
1896struct dirstat_file {
1898 const char *name;
1899 unsigned long changed;
1900};
1901struct dirstat_dir {
1903 struct dirstat_file *files;
1904 int alloc, nr, permille, cumulative;
1905};
1906static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1908 unsigned long changed, const char *base, int baselen)
1909{
1910 unsigned long this_dir = 0;
1911 unsigned int sources = 0;
1912 const char *line_prefix = diff_line_prefix(opt);
1913 while (dir->nr) {
1915 struct dirstat_file *f = dir->files;
1916 int namelen = strlen(f->name);
1917 unsigned long this;
1918 char *slash;
1919 if (namelen < baselen)
1921 break;
1922 if (memcmp(f->name, base, baselen))
1923 break;
1924 slash = strchr(f->name + baselen, '/');
1925 if (slash) {
1926 int newbaselen = slash + 1 - f->name;
1927 this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
1928 sources++;
1929 } else {
1930 this = f->changed;
1931 dir->files++;
1932 dir->nr--;
1933 sources += 2;
1934 }
1935 this_dir += this;
1936 }
1937 /*
1939 * We don't report dirstat's for
1940 * - the top level
1941 * - or cases where everything came from a single directory
1942 * under this directory (sources == 1).
1943 */
1944 if (baselen && sources != 1) {
1945 if (this_dir) {
1946 int permille = this_dir * 1000 / changed;
1947 if (permille >= dir->permille) {
1948 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
1949 permille / 10, permille % 10, baselen, base);
1950 if (!dir->cumulative)
1951 return 0;
1952 }
1953 }
1954 }
1955 return this_dir;
1956}
1957static int dirstat_compare(const void *_a, const void *_b)
1959{
1960 const struct dirstat_file *a = _a;
1961 const struct dirstat_file *b = _b;
1962 return strcmp(a->name, b->name);
1963}
1964static void show_dirstat(struct diff_options *options)
1966{
1967 int i;
1968 unsigned long changed;
1969 struct dirstat_dir dir;
1970 struct diff_queue_struct *q = &diff_queued_diff;
1971 dir.files = NULL;
1973 dir.alloc = 0;
1974 dir.nr = 0;
1975 dir.permille = options->dirstat_permille;
1976 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1977 changed = 0;
1979 for (i = 0; i < q->nr; i++) {
1980 struct diff_filepair *p = q->queue[i];
1981 const char *name;
1982 unsigned long copied, added, damage;
1983 int content_changed;
1984 name = p->two->path ? p->two->path : p->one->path;
1986 if (p->one->oid_valid && p->two->oid_valid)
1988 content_changed = oidcmp(&p->one->oid, &p->two->oid);
1989 else
1990 content_changed = 1;
1991 if (!content_changed) {
1993 /*
1994 * The SHA1 has not changed, so pre-/post-content is
1995 * identical. We can therefore skip looking at the
1996 * file contents altogether.
1997 */
1998 damage = 0;
1999 goto found_damage;
2000 }
2001 if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
2003 /*
2004 * In --dirstat-by-file mode, we don't really need to
2005 * look at the actual file contents at all.
2006 * The fact that the SHA1 changed is enough for us to
2007 * add this file to the list of results
2008 * (with each file contributing equal damage).
2009 */
2010 damage = 1;
2011 goto found_damage;
2012 }
2013 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
2015 diff_populate_filespec(p->one, 0);
2016 diff_populate_filespec(p->two, 0);
2017 diffcore_count_changes(p->one, p->two, NULL, NULL,
2018 &copied, &added);
2019 diff_free_filespec_data(p->one);
2020 diff_free_filespec_data(p->two);
2021 } else if (DIFF_FILE_VALID(p->one)) {
2022 diff_populate_filespec(p->one, CHECK_SIZE_ONLY);
2023 copied = added = 0;
2024 diff_free_filespec_data(p->one);
2025 } else if (DIFF_FILE_VALID(p->two)) {
2026 diff_populate_filespec(p->two, CHECK_SIZE_ONLY);
2027 copied = 0;
2028 added = p->two->size;
2029 diff_free_filespec_data(p->two);
2030 } else
2031 continue;
2032 /*
2034 * Original minus copied is the removed material,
2035 * added is the new material. They are both damages
2036 * made to the preimage.
2037 * If the resulting damage is zero, we know that
2038 * diffcore_count_changes() considers the two entries to
2039 * be identical, but since content_changed is true, we
2040 * know that there must have been _some_ kind of change,
2041 * so we force all entries to have damage > 0.
2042 */
2043 damage = (p->one->size - copied) + added;
2044 if (!damage)
2045 damage = 1;
2046found_damage:
2048 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2049 dir.files[dir.nr].name = name;
2050 dir.files[dir.nr].changed = damage;
2051 changed += damage;
2052 dir.nr++;
2053 }
2054 /* This can happen even with many files, if everything was renames */
2056 if (!changed)
2057 return;
2058 /* Show all directories with more than x% of the changes */
2060 QSORT(dir.files, dir.nr, dirstat_compare);
2061 gather_dirstat(options, &dir, changed, "", 0);
2062}
2063static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
2065{
2066 int i;
2067 unsigned long changed;
2068 struct dirstat_dir dir;
2069 if (data->nr == 0)
2071 return;
2072 dir.files = NULL;
2074 dir.alloc = 0;
2075 dir.nr = 0;
2076 dir.permille = options->dirstat_permille;
2077 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
2078 changed = 0;
2080 for (i = 0; i < data->nr; i++) {
2081 struct diffstat_file *file = data->files[i];
2082 unsigned long damage = file->added + file->deleted;
2083 if (file->is_binary)
2084 /*
2085 * binary files counts bytes, not lines. Must find some
2086 * way to normalize binary bytes vs. textual lines.
2087 * The following heuristic assumes that there are 64
2088 * bytes per "line".
2089 * This is stupid and ugly, but very cheap...
2090 */
2091 damage = DIV_ROUND_UP(damage, 64);
2092 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2093 dir.files[dir.nr].name = file->name;
2094 dir.files[dir.nr].changed = damage;
2095 changed += damage;
2096 dir.nr++;
2097 }
2098 /* This can happen even with many files, if everything was renames */
2100 if (!changed)
2101 return;
2102 /* Show all directories with more than x% of the changes */
2104 QSORT(dir.files, dir.nr, dirstat_compare);
2105 gather_dirstat(options, &dir, changed, "", 0);
2106}
2107static void free_diffstat_info(struct diffstat_t *diffstat)
2109{
2110 int i;
2111 for (i = 0; i < diffstat->nr; i++) {
2112 struct diffstat_file *f = diffstat->files[i];
2113 if (f->name != f->print_name)
2114 free(f->print_name);
2115 free(f->name);
2116 free(f->from_name);
2117 free(f);
2118 }
2119 free(diffstat->files);
2120}
2121struct checkdiff_t {
2123 const char *filename;
2124 int lineno;
2125 int conflict_marker_size;
2126 struct diff_options *o;
2127 unsigned ws_rule;
2128 unsigned status;
2129};
2130static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2132{
2133 char firstchar;
2134 int cnt;
2135 if (len < marker_size + 1)
2137 return 0;
2138 firstchar = line[0];
2139 switch (firstchar) {
2140 case '=': case '>': case '<': case '|':
2141 break;
2142 default:
2143 return 0;
2144 }
2145 for (cnt = 1; cnt < marker_size; cnt++)
2146 if (line[cnt] != firstchar)
2147 return 0;
2148 /* line[1] thru line[marker_size-1] are same as firstchar */
2149 if (len < marker_size + 1 || !isspace(line[marker_size]))
2150 return 0;
2151 return 1;
2152}
2153static void checkdiff_consume(void *priv, char *line, unsigned long len)
2155{
2156 struct checkdiff_t *data = priv;
2157 int marker_size = data->conflict_marker_size;
2158 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2159 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2160 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2161 char *err;
2162 const char *line_prefix;
2163 assert(data->o);
2165 line_prefix = diff_line_prefix(data->o);
2166 if (line[0] == '+') {
2168 unsigned bad;
2169 data->lineno++;
2170 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2171 data->status |= 1;
2172 fprintf(data->o->file,
2173 "%s%s:%d: leftover conflict marker\n",
2174 line_prefix, data->filename, data->lineno);
2175 }
2176 bad = ws_check(line + 1, len - 1, data->ws_rule);
2177 if (!bad)
2178 return;
2179 data->status |= bad;
2180 err = whitespace_error_string(bad);
2181 fprintf(data->o->file, "%s%s:%d: %s.\n",
2182 line_prefix, data->filename, data->lineno, err);
2183 free(err);
2184 emit_line(data->o, set, reset, line, 1);
2185 ws_check_emit(line + 1, len - 1, data->ws_rule,
2186 data->o->file, set, reset, ws);
2187 } else if (line[0] == ' ') {
2188 data->lineno++;
2189 } else if (line[0] == '@') {
2190 char *plus = strchr(line, '+');
2191 if (plus)
2192 data->lineno = strtol(plus, NULL, 10) - 1;
2193 else
2194 die("invalid diff");
2195 }
2196}
2197static unsigned char *deflate_it(char *data,
2199 unsigned long size,
2200 unsigned long *result_size)
2201{
2202 int bound;
2203 unsigned char *deflated;
2204 git_zstream stream;
2205 git_deflate_init(&stream, zlib_compression_level);
2207 bound = git_deflate_bound(&stream, size);
2208 deflated = xmalloc(bound);
2209 stream.next_out = deflated;
2210 stream.avail_out = bound;
2211 stream.next_in = (unsigned char *)data;
2213 stream.avail_in = size;
2214 while (git_deflate(&stream, Z_FINISH) == Z_OK)
2215 ; /* nothing */
2216 git_deflate_end(&stream);
2217 *result_size = stream.total_out;
2218 return deflated;
2219}
2220static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two,
2222 const char *prefix)
2223{
2224 void *cp;
2225 void *delta;
2226 void *deflated;
2227 void *data;
2228 unsigned long orig_size;
2229 unsigned long delta_size;
2230 unsigned long deflate_size;
2231 unsigned long data_size;
2232 /* We could do deflated delta, or we could do just deflated two,
2234 * whichever is smaller.
2235 */
2236 delta = NULL;
2237 deflated = deflate_it(two->ptr, two->size, &deflate_size);
2238 if (one->size && two->size) {
2239 delta = diff_delta(one->ptr, one->size,
2240 two->ptr, two->size,
2241 &delta_size, deflate_size);
2242 if (delta) {
2243 void *to_free = delta;
2244 orig_size = delta_size;
2245 delta = deflate_it(delta, delta_size, &delta_size);
2246 free(to_free);
2247 }
2248 }
2249 if (delta && delta_size < deflate_size) {
2251 fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2252 free(deflated);
2253 data = delta;
2254 data_size = delta_size;
2255 }
2256 else {
2257 fprintf(file, "%sliteral %lu\n", prefix, two->size);
2258 free(delta);
2259 data = deflated;
2260 data_size = deflate_size;
2261 }
2262 /* emit data encoded in base85 */
2264 cp = data;
2265 while (data_size) {
2266 int bytes = (52 < data_size) ? 52 : data_size;
2267 char line[70];
2268 data_size -= bytes;
2269 if (bytes <= 26)
2270 line[0] = bytes + 'A' - 1;
2271 else
2272 line[0] = bytes - 26 + 'a' - 1;
2273 encode_85(line + 1, cp, bytes);
2274 cp = (char *) cp + bytes;
2275 fprintf(file, "%s", prefix);
2276 fputs(line, file);
2277 fputc('\n', file);
2278 }
2279 fprintf(file, "%s\n", prefix);
2280 free(data);
2281}
2282static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two,
2284 const char *prefix)
2285{
2286 fprintf(file, "%sGIT binary patch\n", prefix);
2287 emit_binary_diff_body(file, one, two, prefix);
2288 emit_binary_diff_body(file, two, one, prefix);
2289}
2290int diff_filespec_is_binary(struct diff_filespec *one)
2292{
2293 if (one->is_binary == -1) {
2294 diff_filespec_load_driver(one);
2295 if (one->driver->binary != -1)
2296 one->is_binary = one->driver->binary;
2297 else {
2298 if (!one->data && DIFF_FILE_VALID(one))
2299 diff_populate_filespec(one, CHECK_BINARY);
2300 if (one->is_binary == -1 && one->data)
2301 one->is_binary = buffer_is_binary(one->data,
2302 one->size);
2303 if (one->is_binary == -1)
2304 one->is_binary = 0;
2305 }
2306 }
2307 return one->is_binary;
2308}
2309static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2311{
2312 diff_filespec_load_driver(one);
2313 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2314}
2315void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2317{
2318 if (!options->a_prefix)
2319 options->a_prefix = a;
2320 if (!options->b_prefix)
2321 options->b_prefix = b;
2322}
2323struct userdiff_driver *get_textconv(struct diff_filespec *one)
2325{
2326 if (!DIFF_FILE_VALID(one))
2327 return NULL;
2328 diff_filespec_load_driver(one);
2330 return userdiff_get_textconv(one->driver);
2331}
2332static void builtin_diff(const char *name_a,
2334 const char *name_b,
2335 struct diff_filespec *one,
2336 struct diff_filespec *two,
2337 const char *xfrm_msg,
2338 int must_show_header,
2339 struct diff_options *o,
2340 int complete_rewrite)
2341{
2342 mmfile_t mf1, mf2;
2343 const char *lbl[2];
2344 char *a_one, *b_two;
2345 const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
2346 const char *reset = diff_get_color_opt(o, DIFF_RESET);
2347 const char *a_prefix, *b_prefix;
2348 struct userdiff_driver *textconv_one = NULL;
2349 struct userdiff_driver *textconv_two = NULL;
2350 struct strbuf header = STRBUF_INIT;
2351 const char *line_prefix = diff_line_prefix(o);
2352 diff_set_mnemonic_prefix(o, "a/", "b/");
2354 if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2355 a_prefix = o->b_prefix;
2356 b_prefix = o->a_prefix;
2357 } else {
2358 a_prefix = o->a_prefix;
2359 b_prefix = o->b_prefix;
2360 }
2361 if (o->submodule_format == DIFF_SUBMODULE_LOG &&
2363 (!one->mode || S_ISGITLINK(one->mode)) &&
2364 (!two->mode || S_ISGITLINK(two->mode))) {
2365 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2366 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2367 show_submodule_summary(o->file, one->path ? one->path : two->path,
2368 line_prefix,
2369 &one->oid, &two->oid,
2370 two->dirty_submodule,
2371 meta, del, add, reset);
2372 return;
2373 } else if (o->submodule_format == DIFF_SUBMODULE_INLINE_DIFF &&
2374 (!one->mode || S_ISGITLINK(one->mode)) &&
2375 (!two->mode || S_ISGITLINK(two->mode))) {
2376 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2377 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2378 show_submodule_inline_diff(o->file, one->path ? one->path : two->path,
2379 line_prefix,
2380 &one->oid, &two->oid,
2381 two->dirty_submodule,
2382 meta, del, add, reset, o);
2383 return;
2384 }
2385 if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2387 textconv_one = get_textconv(one);
2388 textconv_two = get_textconv(two);
2389 }
2390 /* Never use a non-valid filename anywhere if at all possible */
2392 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2393 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2394 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2396 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2397 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2398 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2399 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
2400 if (lbl[0][0] == '/') {
2401 /* /dev/null */
2402 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
2403 if (xfrm_msg)
2404 strbuf_addstr(&header, xfrm_msg);
2405 must_show_header = 1;
2406 }
2407 else if (lbl[1][0] == '/') {
2408 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
2409 if (xfrm_msg)
2410 strbuf_addstr(&header, xfrm_msg);
2411 must_show_header = 1;
2412 }
2413 else {
2414 if (one->mode != two->mode) {
2415 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
2416 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
2417 must_show_header = 1;
2418 }
2419 if (xfrm_msg)
2420 strbuf_addstr(&header, xfrm_msg);
2421 /*
2423 * we do not run diff between different kind
2424 * of objects.
2425 */
2426 if ((one->mode ^ two->mode) & S_IFMT)
2427 goto free_ab_and_return;
2428 if (complete_rewrite &&
2429 (textconv_one || !diff_filespec_is_binary(one)) &&
2430 (textconv_two || !diff_filespec_is_binary(two))) {
2431 fprintf(o->file, "%s", header.buf);
2432 strbuf_reset(&header);
2433 emit_rewrite_diff(name_a, name_b, one, two,
2434 textconv_one, textconv_two, o);
2435 o->found_changes = 1;
2436 goto free_ab_and_return;
2437 }
2438 }
2439 if (o->irreversible_delete && lbl[1][0] == '/') {
2441 fprintf(o->file, "%s", header.buf);
2442 strbuf_reset(&header);
2443 goto free_ab_and_return;
2444 } else if (!DIFF_OPT_TST(o, TEXT) &&
2445 ( (!textconv_one && diff_filespec_is_binary(one)) ||
2446 (!textconv_two && diff_filespec_is_binary(two)) )) {
2447 if (!one->data && !two->data &&
2448 S_ISREG(one->mode) && S_ISREG(two->mode) &&
2449 !DIFF_OPT_TST(o, BINARY)) {
2450 if (!oidcmp(&one->oid, &two->oid)) {
2451 if (must_show_header)
2452 fprintf(o->file, "%s", header.buf);
2453 goto free_ab_and_return;
2454 }
2455 fprintf(o->file, "%s", header.buf);
2456 fprintf(o->file, "%sBinary files %s and %s differ\n",
2457 line_prefix, lbl[0], lbl[1]);
2458 goto free_ab_and_return;
2459 }
2460 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2461 die("unable to read files to diff");
2462 /* Quite common confusing case */
2463 if (mf1.size == mf2.size &&
2464 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2465 if (must_show_header)
2466 fprintf(o->file, "%s", header.buf);
2467 goto free_ab_and_return;
2468 }
2469 fprintf(o->file, "%s", header.buf);
2470 strbuf_reset(&header);
2471 if (DIFF_OPT_TST(o, BINARY))
2472 emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2473 else
2474 fprintf(o->file, "%sBinary files %s and %s differ\n",
2475 line_prefix, lbl[0], lbl[1]);
2476 o->found_changes = 1;
2477 } else {
2478 /* Crazy xdl interfaces.. */
2479 const char *diffopts = getenv("GIT_DIFF_OPTS");
2480 const char *v;
2481 xpparam_t xpp;
2482 xdemitconf_t xecfg;
2483 struct emit_callback ecbdata;
2484 const struct userdiff_funcname *pe;
2485 if (must_show_header) {
2487 fprintf(o->file, "%s", header.buf);
2488 strbuf_reset(&header);
2489 }
2490 mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2492 mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2493 pe = diff_funcname_pattern(one);
2495 if (!pe)
2496 pe = diff_funcname_pattern(two);
2497 memset(&xpp, 0, sizeof(xpp));
2499 memset(&xecfg, 0, sizeof(xecfg));
2500 memset(&ecbdata, 0, sizeof(ecbdata));
2501 ecbdata.label_path = lbl;
2502 ecbdata.color_diff = want_color(o->use_color);
2503 ecbdata.ws_rule = whitespace_rule(name_b);
2504 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2505 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2506 ecbdata.opt = o;
2507 ecbdata.header = header.len ? &header : NULL;
2508 xpp.flags = o->xdl_opts;
2509 xecfg.ctxlen = o->context;
2510 xecfg.interhunkctxlen = o->interhunkcontext;
2511 xecfg.flags = XDL_EMIT_FUNCNAMES;
2512 if (DIFF_OPT_TST(o, FUNCCONTEXT))
2513 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2514 if (pe)
2515 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2516 if (!diffopts)
2517 ;
2518 else if (skip_prefix(diffopts, "--unified=", &v))
2519 xecfg.ctxlen = strtoul(v, NULL, 10);
2520 else if (skip_prefix(diffopts, "-u", &v))
2521 xecfg.ctxlen = strtoul(v, NULL, 10);
2522 if (o->word_diff)
2523 init_diff_words_data(&ecbdata, o, one, two);
2524 if (xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2525 &xpp, &xecfg))
2526 die("unable to generate diff for %s", one->path);
2527 if (o->word_diff)
2528 free_diff_words_data(&ecbdata);
2529 if (textconv_one)
2530 free(mf1.ptr);
2531 if (textconv_two)
2532 free(mf2.ptr);
2533 xdiff_clear_find_func(&xecfg);
2534 }
2535 free_ab_and_return:
2537 strbuf_release(&header);
2538 diff_free_filespec_data(one);
2539 diff_free_filespec_data(two);
2540 free(a_one);
2541 free(b_two);
2542 return;
2543}
2544static void builtin_diffstat(const char *name_a, const char *name_b,
2546 struct diff_filespec *one,
2547 struct diff_filespec *two,
2548 struct diffstat_t *diffstat,
2549 struct diff_options *o,
2550 struct diff_filepair *p)
2551{
2552 mmfile_t mf1, mf2;
2553 struct diffstat_file *data;
2554 int same_contents;
2555 int complete_rewrite = 0;
2556 if (!DIFF_PAIR_UNMERGED(p)) {
2558 if (p->status == DIFF_STATUS_MODIFIED && p->score)
2559 complete_rewrite = 1;
2560 }
2561 data = diffstat_add(diffstat, name_a, name_b);
2563 data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
2564 if (!one || !two) {
2566 data->is_unmerged = 1;
2567 return;
2568 }
2569 same_contents = !oidcmp(&one->oid, &two->oid);
2571 if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2573 data->is_binary = 1;
2574 if (same_contents) {
2575 data->added = 0;
2576 data->deleted = 0;
2577 } else {
2578 data->added = diff_filespec_size(two);
2579 data->deleted = diff_filespec_size(one);
2580 }
2581 }
2582 else if (complete_rewrite) {
2584 diff_populate_filespec(one, 0);
2585 diff_populate_filespec(two, 0);
2586 data->deleted = count_lines(one->data, one->size);
2587 data->added = count_lines(two->data, two->size);
2588 }
2589 else if (!same_contents) {
2591 /* Crazy xdl interfaces.. */
2592 xpparam_t xpp;
2593 xdemitconf_t xecfg;
2594 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2596 die("unable to read files to diff");
2597 memset(&xpp, 0, sizeof(xpp));
2599 memset(&xecfg, 0, sizeof(xecfg));
2600 xpp.flags = o->xdl_opts;
2601 xecfg.ctxlen = o->context;
2602 xecfg.interhunkctxlen = o->interhunkcontext;
2603 if (xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2604 &xpp, &xecfg))
2605 die("unable to generate diffstat for %s", one->path);
2606 }
2607 diff_free_filespec_data(one);
2609 diff_free_filespec_data(two);
2610}
2611static void builtin_checkdiff(const char *name_a, const char *name_b,
2613 const char *attr_path,
2614 struct diff_filespec *one,
2615 struct diff_filespec *two,
2616 struct diff_options *o)
2617{
2618 mmfile_t mf1, mf2;
2619 struct checkdiff_t data;
2620 if (!two)
2622 return;
2623 memset(&data, 0, sizeof(data));
2625 data.filename = name_b ? name_b : name_a;
2626 data.lineno = 0;
2627 data.o = o;
2628 data.ws_rule = whitespace_rule(attr_path);
2629 data.conflict_marker_size = ll_merge_marker_size(attr_path);
2630 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2632 die("unable to read files to diff");
2633 /*
2635 * All the other codepaths check both sides, but not checking
2636 * the "old" side here is deliberate. We are checking the newly
2637 * introduced changes, and as long as the "new" side is text, we
2638 * can and should check what it introduces.
2639 */
2640 if (diff_filespec_is_binary(two))
2641 goto free_and_return;
2642 else {
2643 /* Crazy xdl interfaces.. */
2644 xpparam_t xpp;
2645 xdemitconf_t xecfg;
2646 memset(&xpp, 0, sizeof(xpp));
2648 memset(&xecfg, 0, sizeof(xecfg));
2649 xecfg.ctxlen = 1; /* at least one context line */
2650 xpp.flags = 0;
2651 if (xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2652 &xpp, &xecfg))
2653 die("unable to generate checkdiff for %s", one->path);
2654 if (data.ws_rule & WS_BLANK_AT_EOF) {
2656 struct emit_callback ecbdata;
2657 int blank_at_eof;
2658 ecbdata.ws_rule = data.ws_rule;
2660 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2661 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2662 if (blank_at_eof) {
2664 static char *err;
2665 if (!err)
2666 err = whitespace_error_string(WS_BLANK_AT_EOF);
2667 fprintf(o->file, "%s:%d: %s.\n",
2668 data.filename, blank_at_eof, err);
2669 data.status = 1; /* report errors */
2670 }
2671 }
2672 }
2673 free_and_return:
2674 diff_free_filespec_data(one);
2675 diff_free_filespec_data(two);
2676 if (data.status)
2677 DIFF_OPT_SET(o, CHECK_FAILED);
2678}
2679struct diff_filespec *alloc_filespec(const char *path)
2681{
2682 struct diff_filespec *spec;
2683 FLEXPTR_ALLOC_STR(spec, path, path);
2685 spec->count = 1;
2686 spec->is_binary = -1;
2687 return spec;
2688}
2689void free_filespec(struct diff_filespec *spec)
2691{
2692 if (!--spec->count) {
2693 diff_free_filespec_data(spec);
2694 free(spec);
2695 }
2696}
2697void fill_filespec(struct diff_filespec *spec, const struct object_id *oid,
2699 int oid_valid, unsigned short mode)
2700{
2701 if (mode) {
2702 spec->mode = canon_mode(mode);
2703 oidcpy(&spec->oid, oid);
2704 spec->oid_valid = oid_valid;
2705 }
2706}
2707/*
2709 * Given a name and sha1 pair, if the index tells us the file in
2710 * the work tree has that object contents, return true, so that
2711 * prepare_temp_file() does not have to inflate and extract.
2712 */
2713static int reuse_worktree_file(const char *name, const struct object_id *oid, int want_file)
2714{
2715 const struct cache_entry *ce;
2716 struct stat st;
2717 int pos, len;
2718 /*
2720 * We do not read the cache ourselves here, because the
2721 * benchmark with my previous version that always reads cache
2722 * shows that it makes things worse for diff-tree comparing
2723 * two linux-2.6 kernel trees in an already checked out work
2724 * tree. This is because most diff-tree comparisons deal with
2725 * only a small number of files, while reading the cache is
2726 * expensive for a large project, and its cost outweighs the
2727 * savings we get by not inflating the object to a temporary
2728 * file. Practically, this code only helps when we are used
2729 * by diff-cache --cached, which does read the cache before
2730 * calling us.
2731 */
2732 if (!active_cache)
2733 return 0;
2734 /* We want to avoid the working directory if our caller
2736 * doesn't need the data in a normal file, this system
2737 * is rather slow with its stat/open/mmap/close syscalls,
2738 * and the object is contained in a pack file. The pack
2739 * is probably already open and will be faster to obtain
2740 * the data through than the working directory. Loose
2741 * objects however would tend to be slower as they need
2742 * to be individually opened and inflated.
2743 */
2744 if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(oid->hash))
2745 return 0;
2746 /*
2748 * Similarly, if we'd have to convert the file contents anyway, that
2749 * makes the optimization not worthwhile.
2750 */
2751 if (!want_file && would_convert_to_git(&the_index, name))
2752 return 0;
2753 len = strlen(name);
2755 pos = cache_name_pos(name, len);
2756 if (pos < 0)
2757 return 0;
2758 ce = active_cache[pos];
2759 /*
2761 * This is not the sha1 we are looking for, or
2762 * unreusable because it is not a regular file.
2763 */
2764 if (oidcmp(oid, &ce->oid) || !S_ISREG(ce->ce_mode))
2765 return 0;
2766 /*
2768 * If ce is marked as "assume unchanged", there is no
2769 * guarantee that work tree matches what we are looking for.
2770 */
2771 if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2772 return 0;
2773 /*
2775 * If ce matches the file in the work tree, we can reuse it.
2776 */
2777 if (ce_uptodate(ce) ||
2778 (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2779 return 1;
2780 return 0;
2782}
2783static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2785{
2786 struct strbuf buf = STRBUF_INIT;
2787 char *dirty = "";
2788 /* Are we looking at the work tree? */
2790 if (s->dirty_submodule)
2791 dirty = "-dirty";
2792 strbuf_addf(&buf, "Subproject commit %s%s\n",
2794 oid_to_hex(&s->oid), dirty);
2795 s->size = buf.len;
2796 if (size_only) {
2797 s->data = NULL;
2798 strbuf_release(&buf);
2799 } else {
2800 s->data = strbuf_detach(&buf, NULL);
2801 s->should_free = 1;
2802 }
2803 return 0;
2804}
2805/*
2807 * While doing rename detection and pickaxe operation, we may need to
2808 * grab the data for the blob (or file) for our own in-core comparison.
2809 * diff_filespec has data and size fields for this purpose.
2810 */
2811int diff_populate_filespec(struct diff_filespec *s, unsigned int flags)
2812{
2813 int size_only = flags & CHECK_SIZE_ONLY;
2814 int err = 0;
2815 /*
2816 * demote FAIL to WARN to allow inspecting the situation
2817 * instead of refusing.
2818 */
2819 enum safe_crlf crlf_warn = (safe_crlf == SAFE_CRLF_FAIL
2820 ? SAFE_CRLF_WARN
2821 : safe_crlf);
2822 if (!DIFF_FILE_VALID(s))
2824 die("internal error: asking to populate invalid file.");
2825 if (S_ISDIR(s->mode))
2826 return -1;
2827 if (s->data)
2829 return 0;
2830 if (size_only && 0 < s->size)
2832 return 0;
2833 if (S_ISGITLINK(s->mode))
2835 return diff_populate_gitlink(s, size_only);
2836 if (!s->oid_valid ||
2838 reuse_worktree_file(s->path, &s->oid, 0)) {
2839 struct strbuf buf = STRBUF_INIT;
2840 struct stat st;
2841 int fd;
2842 if (lstat(s->path, &st) < 0) {
2844 if (errno == ENOENT) {
2845 err_empty:
2846 err = -1;
2847 empty:
2848 s->data = (char *)"";
2849 s->size = 0;
2850 return err;
2851 }
2852 }
2853 s->size = xsize_t(st.st_size);
2854 if (!s->size)
2855 goto empty;
2856 if (S_ISLNK(st.st_mode)) {
2857 struct strbuf sb = STRBUF_INIT;
2858 if (strbuf_readlink(&sb, s->path, s->size))
2860 goto err_empty;
2861 s->size = sb.len;
2862 s->data = strbuf_detach(&sb, NULL);
2863 s->should_free = 1;
2864 return 0;
2865 }
2866 /*
2868 * Even if the caller would be happy with getting
2869 * only the size, we cannot return early at this
2870 * point if the path requires us to run the content
2871 * conversion.
2872 */
2873 if (size_only && !would_convert_to_git(&the_index, s->path))
2874 return 0;
2875 /*
2877 * Note: this check uses xsize_t(st.st_size) that may
2878 * not be the true size of the blob after it goes
2879 * through convert_to_git(). This may not strictly be
2880 * correct, but the whole point of big_file_threshold
2881 * and is_binary check being that we want to avoid
2882 * opening the file and inspecting the contents, this
2883 * is probably fine.
2884 */
2885 if ((flags & CHECK_BINARY) &&
2886 s->size > big_file_threshold && s->is_binary == -1) {
2887 s->is_binary = 1;
2888 return 0;
2889 }
2890 fd = open(s->path, O_RDONLY);
2891 if (fd < 0)
2892 goto err_empty;
2893 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2894 close(fd);
2895 s->should_munmap = 1;
2896 /*
2898 * Convert from working tree format to canonical git format
2899 */
2900 if (convert_to_git(&the_index, s->path, s->data, s->size, &buf, crlf_warn)) {
2901 size_t size = 0;
2902 munmap(s->data, s->size);
2903 s->should_munmap = 0;
2904 s->data = strbuf_detach(&buf, &size);
2905 s->size = size;
2906 s->should_free = 1;
2907 }
2908 }
2909 else {
2910 enum object_type type;
2911 if (size_only || (flags & CHECK_BINARY)) {
2912 type = sha1_object_info(s->oid.hash, &s->size);
2913 if (type < 0)
2914 die("unable to read %s",
2915 oid_to_hex(&s->oid));
2916 if (size_only)
2917 return 0;
2918 if (s->size > big_file_threshold && s->is_binary == -1) {
2919 s->is_binary = 1;
2920 return 0;
2921 }
2922 }
2923 s->data = read_sha1_file(s->oid.hash, &type, &s->size);
2924 if (!s->data)
2925 die("unable to read %s", oid_to_hex(&s->oid));
2926 s->should_free = 1;
2927 }
2928 return 0;
2929}
2930void diff_free_filespec_blob(struct diff_filespec *s)
2932{
2933 if (s->should_free)
2934 free(s->data);
2935 else if (s->should_munmap)
2936 munmap(s->data, s->size);
2937 if (s->should_free || s->should_munmap) {
2939 s->should_free = s->should_munmap = 0;
2940 s->data = NULL;
2941 }
2942}
2943void diff_free_filespec_data(struct diff_filespec *s)
2945{
2946 diff_free_filespec_blob(s);
2947 FREE_AND_NULL(s->cnt_data);
2948}
2949static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2951 void *blob,
2952 unsigned long size,
2953 const struct object_id *oid,
2954 int mode)
2955{
2956 int fd;
2957 struct strbuf buf = STRBUF_INIT;
2958 struct strbuf template = STRBUF_INIT;
2959 char *path_dup = xstrdup(path);
2960 const char *base = basename(path_dup);
2961 /* Generate "XXXXXX_basename.ext" */
2963 strbuf_addstr(&template, "XXXXXX_");
2964 strbuf_addstr(&template, base);
2965 fd = mks_tempfile_ts(&temp->tempfile, template.buf, strlen(base) + 1);
2967 if (fd < 0)
2968 die_errno("unable to create temp-file");
2969 if (convert_to_working_tree(path,
2970 (const char *)blob, (size_t)size, &buf)) {
2971 blob = buf.buf;
2972 size = buf.len;
2973 }
2974 if (write_in_full(fd, blob, size) != size)
2975 die_errno("unable to write temp-file");
2976 close_tempfile(&temp->tempfile);
2977 temp->name = get_tempfile_path(&temp->tempfile);
2978 oid_to_hex_r(temp->hex, oid);
2979 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", mode);
2980 strbuf_release(&buf);
2981 strbuf_release(&template);
2982 free(path_dup);
2983}
2984static struct diff_tempfile *prepare_temp_file(const char *name,
2986 struct diff_filespec *one)
2987{
2988 struct diff_tempfile *temp = claim_diff_tempfile();
2989 if (!DIFF_FILE_VALID(one)) {
2991 not_a_valid_file:
2992 /* A '-' entry produces this for file-2, and
2993 * a '+' entry produces this for file-1.
2994 */
2995 temp->name = "/dev/null";
2996 xsnprintf(temp->hex, sizeof(temp->hex), ".");
2997 xsnprintf(temp->mode, sizeof(temp->mode), ".");
2998 return temp;
2999 }
3000 if (!S_ISGITLINK(one->mode) &&
3002 (!one->oid_valid ||
3003 reuse_worktree_file(name, &one->oid, 1))) {
3004 struct stat st;
3005 if (lstat(name, &st) < 0) {
3006 if (errno == ENOENT)
3007 goto not_a_valid_file;
3008 die_errno("stat(%s)", name);
3009 }
3010 if (S_ISLNK(st.st_mode)) {
3011 struct strbuf sb = STRBUF_INIT;
3012 if (strbuf_readlink(&sb, name, st.st_size) < 0)
3013 die_errno("readlink(%s)", name);
3014 prep_temp_blob(name, temp, sb.buf, sb.len,
3015 (one->oid_valid ?
3016 &one->oid : &null_oid),
3017 (one->oid_valid ?
3018 one->mode : S_IFLNK));
3019 strbuf_release(&sb);
3020 }
3021 else {
3022 /* we can borrow from the file in the work tree */
3023 temp->name = name;
3024 if (!one->oid_valid)
3025 oid_to_hex_r(temp->hex, &null_oid);
3026 else
3027 oid_to_hex_r(temp->hex, &one->oid);
3028 /* Even though we may sometimes borrow the
3029 * contents from the work tree, we always want
3030 * one->mode. mode is trustworthy even when
3031 * !(one->oid_valid), as long as
3032 * DIFF_FILE_VALID(one).
3033 */
3034 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", one->mode);
3035 }
3036 return temp;
3037 }
3038 else {
3039 if (diff_populate_filespec(one, 0))
3040 die("cannot read data blob for %s", one->path);
3041 prep_temp_blob(name, temp, one->data, one->size,
3042 &one->oid, one->mode);
3043 }
3044 return temp;
3045}
3046static void add_external_diff_name(struct argv_array *argv,
3048 const char *name,
3049 struct diff_filespec *df)
3050{
3051 struct diff_tempfile *temp = prepare_temp_file(name, df);
3052 argv_array_push(argv, temp->name);
3053 argv_array_push(argv, temp->hex);
3054 argv_array_push(argv, temp->mode);
3055}
3056/* An external diff command takes:
3058 *
3059 * diff-cmd name infile1 infile1-sha1 infile1-mode \
3060 * infile2 infile2-sha1 infile2-mode [ rename-to ]
3061 *
3062 */
3063static void run_external_diff(const char *pgm,
3064 const char *name,
3065 const char *other,
3066 struct diff_filespec *one,
3067 struct diff_filespec *two,
3068 const char *xfrm_msg,
3069 int complete_rewrite,
3070 struct diff_options *o)
3071{
3072 struct argv_array argv = ARGV_ARRAY_INIT;
3073 struct argv_array env = ARGV_ARRAY_INIT;
3074 struct diff_queue_struct *q = &diff_queued_diff;
3075 argv_array_push(&argv, pgm);
3077 argv_array_push(&argv, name);
3078 if (one && two) {
3080 add_external_diff_name(&argv, name, one);
3081 if (!other)
3082 add_external_diff_name(&argv, name, two);
3083 else {
3084 add_external_diff_name(&argv, other, two);
3085 argv_array_push(&argv, other);
3086 argv_array_push(&argv, xfrm_msg);
3087 }
3088 }
3089 argv_array_pushf(&env, "GIT_DIFF_PATH_COUNTER=%d", ++o->diff_path_counter);
3091 argv_array_pushf(&env, "GIT_DIFF_PATH_TOTAL=%d", q->nr);
3092 if (run_command_v_opt_cd_env(argv.argv, RUN_USING_SHELL, NULL, env.argv))
3094 die(_("external diff died, stopping at %s"), name);
3095 remove_tempfile();
3097 argv_array_clear(&argv);
3098 argv_array_clear(&env);
3099}
3100static int similarity_index(struct diff_filepair *p)
3102{
3103 return p->score * 100 / MAX_SCORE;
3104}
3105static const char *diff_abbrev_oid(const struct object_id *oid, int abbrev)
3107{
3108 if (startup_info->have_repository)
3109 return find_unique_abbrev(oid->hash, abbrev);
3110 else {
3111 char *hex = oid_to_hex(oid);
3112 if (abbrev < 0)
3113 abbrev = FALLBACK_DEFAULT_ABBREV;
3114 if (abbrev > GIT_SHA1_HEXSZ)
3115 die("BUG: oid abbreviation out of range: %d", abbrev);
3116 if (abbrev)
3117 hex[abbrev] = '\0';
3118 return hex;
3119 }
3120}
3121static void fill_metainfo(struct strbuf *msg,
3123 const char *name,
3124 const char *other,
3125 struct diff_filespec *one,
3126 struct diff_filespec *two,
3127 struct diff_options *o,
3128 struct diff_filepair *p,
3129 int *must_show_header,
3130 int use_color)
3131{
3132 const char *set = diff_get_color(use_color, DIFF_METAINFO);
3133 const char *reset = diff_get_color(use_color, DIFF_RESET);
3134 const char *line_prefix = diff_line_prefix(o);
3135 *must_show_header = 1;
3137 strbuf_init(msg, PATH_MAX * 2 + 300);
3138 switch (p->status) {
3139 case DIFF_STATUS_COPIED:
3140 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3141 line_prefix, set, similarity_index(p));
3142 strbuf_addf(msg, "%s\n%s%scopy from ",
3143 reset, line_prefix, set);
3144 quote_c_style(name, msg, NULL, 0);
3145 strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
3146 quote_c_style(other, msg, NULL, 0);
3147 strbuf_addf(msg, "%s\n", reset);
3148 break;
3149 case DIFF_STATUS_RENAMED:
3150 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3151 line_prefix, set, similarity_index(p));
3152 strbuf_addf(msg, "%s\n%s%srename from ",
3153 reset, line_prefix, set);
3154 quote_c_style(name, msg, NULL, 0);
3155 strbuf_addf(msg, "%s\n%s%srename to ",
3156 reset, line_prefix, set);
3157 quote_c_style(other, msg, NULL, 0);
3158 strbuf_addf(msg, "%s\n", reset);
3159 break;
3160 case DIFF_STATUS_MODIFIED:
3161 if (p->score) {
3162 strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
3163 line_prefix,
3164 set, similarity_index(p), reset);
3165 break;
3166 }
3167 /* fallthru */
3168 default:
3169 *must_show_header = 0;
3170 }
3171 if (one && two && oidcmp(&one->oid, &two->oid)) {
3172 int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
3173 if (DIFF_OPT_TST(o, BINARY)) {
3175 mmfile_t mf;
3176 if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
3177 (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
3178 abbrev = 40;
3179 }
3180 strbuf_addf(msg, "%s%sindex %s..%s", line_prefix, set,
3181 diff_abbrev_oid(&one->oid, abbrev),
3182 diff_abbrev_oid(&two->oid, abbrev));
3183 if (one->mode == two->mode)
3184 strbuf_addf(msg, " %06o", one->mode);
3185 strbuf_addf(msg, "%s\n", reset);
3186 }
3187}
3188static void run_diff_cmd(const char *pgm,
3190 const char *name,
3191 const char *other,
3192 const char *attr_path,
3193 struct diff_filespec *one,
3194 struct diff_filespec *two,
3195 struct strbuf *msg,
3196 struct diff_options *o,
3197 struct diff_filepair *p)
3198{
3199 const char *xfrm_msg = NULL;
3200 int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
3201 int must_show_header = 0;
3202 if (DIFF_OPT_TST(o, ALLOW_EXTERNAL)) {
3205 struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
3206 if (drv && drv->external)
3207 pgm = drv->external;
3208 }
3209 if (msg) {
3211 /*
3212 * don't use colors when the header is intended for an
3213 * external diff driver
3214 */
3215 fill_metainfo(msg, name, other, one, two, o, p,
3216 &must_show_header,
3217 want_color(o->use_color) && !pgm);
3218 xfrm_msg = msg->len ? msg->buf : NULL;
3219 }
3220 if (pgm) {
3222 run_external_diff(pgm, name, other, one, two, xfrm_msg,
3223 complete_rewrite, o);
3224 return;
3225 }
3226 if (one && two)
3227 builtin_diff(name, other ? other : name,
3228 one, two, xfrm_msg, must_show_header,
3229 o, complete_rewrite);
3230 else
3231 fprintf(o->file, "* Unmerged path %s\n", name);
3232}
3233static void diff_fill_oid_info(struct diff_filespec *one)
3235{
3236 if (DIFF_FILE_VALID(one)) {
3237 if (!one->oid_valid) {
3238 struct stat st;
3239 if (one->is_stdin) {
3240 oidclr(&one->oid);
3241 return;
3242 }
3243 if (lstat(one->path, &st) < 0)
3244 die_errno("stat '%s'", one->path);
3245 if (index_path(one->oid.hash, one->path, &st, 0))
3246 die("cannot hash %s", one->path);
3247 }
3248 }
3249 else
3250 oidclr(&one->oid);
3251}
3252static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
3254{
3255 /* Strip the prefix but do not molest /dev/null and absolute paths */
3256 if (*namep && **namep != '/') {
3257 *namep += prefix_length;
3258 if (**namep == '/')
3259 ++*namep;
3260 }
3261 if (*otherp && **otherp != '/') {
3262 *otherp += prefix_length;
3263 if (**otherp == '/')
3264 ++*otherp;
3265 }
3266}
3267static void run_diff(struct diff_filepair *p, struct diff_options *o)
3269{
3270 const char *pgm = external_diff();
3271 struct strbuf msg;
3272 struct diff_filespec *one = p->one;
3273 struct diff_filespec *two = p->two;
3274 const char *name;
3275 const char *other;
3276 const char *attr_path;
3277 name = p->one->path;
3279 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3280 attr_path = name;
3281 if (o->prefix_length)
3282 strip_prefix(o->prefix_length, &name, &other);
3283 if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
3285 pgm = NULL;
3286 if (DIFF_PAIR_UNMERGED(p)) {
3288 run_diff_cmd(pgm, name, NULL, attr_path,
3289 NULL, NULL, NULL, o, p);
3290 return;
3291 }
3292 diff_fill_oid_info(one);
3294 diff_fill_oid_info(two);
3295 if (!pgm &&
3297 DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3298 (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3299 /*
3300 * a filepair that changes between file and symlink
3301 * needs to be split into deletion and creation.
3302 */
3303 struct diff_filespec *null = alloc_filespec(two->path);
3304 run_diff_cmd(NULL, name, other, attr_path,
3305 one, null, &msg, o, p);
3306 free(null);
3307 strbuf_release(&msg);
3308 null = alloc_filespec(one->path);
3310 run_diff_cmd(NULL, name, other, attr_path,
3311 null, two, &msg, o, p);
3312 free(null);
3313 }
3314 else
3315 run_diff_cmd(pgm, name, other, attr_path,
3316 one, two, &msg, o, p);
3317 strbuf_release(&msg);
3319}
3320static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3322 struct diffstat_t *diffstat)
3323{
3324 const char *name;
3325 const char *other;
3326 if (DIFF_PAIR_UNMERGED(p)) {
3328 /* unmerged */
3329 builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, p);
3330 return;
3331 }
3332 name = p->one->path;
3334 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3335 if (o->prefix_length)
3337 strip_prefix(o->prefix_length, &name, &other);
3338 diff_fill_oid_info(p->one);
3340 diff_fill_oid_info(p->two);
3341 builtin_diffstat(name, other, p->one, p->two, diffstat, o, p);
3343}
3344static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3346{
3347 const char *name;
3348 const char *other;
3349 const char *attr_path;
3350 if (DIFF_PAIR_UNMERGED(p)) {
3352 /* unmerged */
3353 return;
3354 }
3355 name = p->one->path;
3357 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3358 attr_path = other ? other : name;
3359 if (o->prefix_length)
3361 strip_prefix(o->prefix_length, &name, &other);
3362 diff_fill_oid_info(p->one);
3364 diff_fill_oid_info(p->two);
3365 builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3367}
3368void diff_setup(struct diff_options *options)
3370{
3371 memcpy(options, &default_diff_options, sizeof(*options));
3372 options->file = stdout;
3374 options->abbrev = DEFAULT_ABBREV;
3376 options->line_termination = '\n';
3377 options->break_opt = -1;
3378 options->rename_limit = -1;
3379 options->dirstat_permille = diff_dirstat_permille_default;
3380 options->context = diff_context_default;
3381 options->interhunkcontext = diff_interhunk_context_default;
3382 options->ws_error_highlight = ws_error_highlight_default;
3383 DIFF_OPT_SET(options, RENAME_EMPTY);
3384 /* pathchange left =NULL by default */
3386 options->change = diff_change;
3387 options->add_remove = diff_addremove;
3388 options->use_color = diff_use_color_default;
3389 options->detect_rename = diff_detect_rename_default;
3390 options->xdl_opts |= diff_algorithm;
3391 if (diff_indent_heuristic)
3392 DIFF_XDL_SET(options, INDENT_HEURISTIC);
3393 options->orderfile = diff_order_file_cfg;
3395 if (diff_no_prefix) {
3397 options->a_prefix = options->b_prefix = "";
3398 } else if (!diff_mnemonic_prefix) {
3399 options->a_prefix = "a/";
3400 options->b_prefix = "b/";
3401 }
3402}
3403void diff_setup_done(struct diff_options *options)
3405{
3406 int count = 0;
3407 if (options->set_default)
3409 options->set_default(options);
3410 if (options->output_format & DIFF_FORMAT_NAME)
3412 count++;
3413 if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3414 count++;
3415 if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3416 count++;
3417 if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3418 count++;
3419 if (count > 1)
3420 die(_("--name-only, --name-status, --check and -s are mutually exclusive"));
3421 /*
3423 * Most of the time we can say "there are changes"
3424 * only by checking if there are changed paths, but
3425 * --ignore-whitespace* options force us to look
3426 * inside contents.
3427 */
3428 if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3430 DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3431 DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3432 DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3433 else
3434 DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3435 if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3437 options->detect_rename = DIFF_DETECT_COPY;
3438 if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3440 options->prefix = NULL;
3441 if (options->prefix)
3442 options->prefix_length = strlen(options->prefix);
3443 else
3444 options->prefix_length = 0;
3445 if (options->output_format & (DIFF_FORMAT_NAME |
3447 DIFF_FORMAT_NAME_STATUS |
3448 DIFF_FORMAT_CHECKDIFF |
3449 DIFF_FORMAT_NO_OUTPUT))
3450 options->output_format &= ~(DIFF_FORMAT_RAW |
3451 DIFF_FORMAT_NUMSTAT |
3452 DIFF_FORMAT_DIFFSTAT |
3453 DIFF_FORMAT_SHORTSTAT |
3454 DIFF_FORMAT_DIRSTAT |
3455 DIFF_FORMAT_SUMMARY |
3456 DIFF_FORMAT_PATCH);
3457 /*
3459 * These cases always need recursive; we do not drop caller-supplied
3460 * recursive bits for other formats here.
3461 */
3462 if (options->output_format & (DIFF_FORMAT_PATCH |
3463 DIFF_FORMAT_NUMSTAT |
3464 DIFF_FORMAT_DIFFSTAT |
3465 DIFF_FORMAT_SHORTSTAT |
3466 DIFF_FORMAT_DIRSTAT |
3467 DIFF_FORMAT_SUMMARY |
3468 DIFF_FORMAT_CHECKDIFF))
3469 DIFF_OPT_SET(options, RECURSIVE);
3470 /*
3471 * Also pickaxe would not work very well if you do not say recursive
3472 */
3473 if (options->pickaxe)
3474 DIFF_OPT_SET(options, RECURSIVE);
3475 /*
3476 * When patches are generated, submodules diffed against the work tree
3477 * must be checked for dirtiness too so it can be shown in the output
3478 */
3479 if (options->output_format & DIFF_FORMAT_PATCH)
3480 DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3481 if (options->detect_rename && options->rename_limit < 0)
3483 options->rename_limit = diff_rename_limit_default;
3484 if (options->setup & DIFF_SETUP_USE_CACHE) {
3485 if (!active_cache)
3486 /* read-cache does not die even when it fails
3487 * so it is safe for us to do this here. Also
3488 * it does not smudge active_cache or active_nr
3489 * when it fails, so we do not have to worry about
3490 * cleaning it up ourselves either.
3491 */
3492 read_cache();
3493 }
3494 if (40 < options->abbrev)
3495 options->abbrev = 40; /* full */
3496 /*
3498 * It does not make sense to show the first hit we happened
3499 * to have found. It does not make sense not to return with
3500 * exit code in such a case either.
3501 */
3502 if (DIFF_OPT_TST(options, QUICK)) {
3503 options->output_format = DIFF_FORMAT_NO_OUTPUT;
3504 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3505 }
3506 options->diff_path_counter = 0;
3508 if (DIFF_OPT_TST(options, FOLLOW_RENAMES) && options->pathspec.nr != 1)
3510 die(_("--follow requires exactly one pathspec"));
3511}
3512static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3514{
3515 char c, *eq;
3516 int len;
3517 if (*arg != '-')
3519 return 0;
3520 c = *++arg;
3521 if (!c)
3522 return 0;
3523 if (c == arg_short) {
3524 c = *++arg;
3525 if (!c)
3526 return 1;
3527 if (val && isdigit(c)) {
3528 char *end;
3529 int n = strtoul(arg, &end, 10);
3530 if (*end)
3531 return 0;
3532 *val = n;
3533 return 1;
3534 }
3535 return 0;
3536 }
3537 if (c != '-')
3538 return 0;
3539 arg++;
3540 eq = strchrnul(arg, '=');
3541 len = eq - arg;
3542 if (!len || strncmp(arg, arg_long, len))
3543 return 0;
3544 if (*eq) {
3545 int n;
3546 char *end;
3547 if (!isdigit(*++eq))
3548 return 0;
3549 n = strtoul(eq, &end, 10);
3550 if (*end)
3551 return 0;
3552 *val = n;
3553 }
3554 return 1;
3555}
3556static int diff_scoreopt_parse(const char *opt);
3558static inline int short_opt(char opt, const char **argv,
3560 const char **optarg)
3561{
3562 const char *arg = argv[0];
3563 if (arg[0] != '-' || arg[1] != opt)
3564 return 0;
3565 if (arg[2] != '\0') {
3566 *optarg = arg + 2;
3567 return 1;
3568 }
3569 if (!argv[1])
3570 die("Option '%c' requires a value", opt);
3571 *optarg = argv[1];
3572 return 2;
3573}
3574int parse_long_opt(const char *opt, const char **argv,
3576 const char **optarg)
3577{
3578 const char *arg = argv[0];
3579 if (!skip_prefix(arg, "--", &arg))
3580 return 0;
3581 if (!skip_prefix(arg, opt, &arg))
3582 return 0;
3583 if (*arg == '=') { /* stuck form: --option=value */
3584 *optarg = arg + 1;
3585 return 1;
3586 }
3587 if (*arg != '\0')
3588 return 0;
3589 /* separate form: --option value */
3590 if (!argv[1])
3591 die("Option '--%s' requires a value", opt);
3592 *optarg = argv[1];
3593 return 2;
3594}
3595static int stat_opt(struct diff_options *options, const char **av)
3597{
3598 const char *arg = av[0];
3599 char *end;
3600 int width = options->stat_width;
3601 int name_width = options->stat_name_width;
3602 int graph_width = options->stat_graph_width;
3603 int count = options->stat_count;
3604 int argcount = 1;
3605 if (!skip_prefix(arg, "--stat", &arg))
3607 die("BUG: stat option does not begin with --stat: %s", arg);
3608 end = (char *)arg;
3609 switch (*arg) {
3611 case '-':
3612 if (skip_prefix(arg, "-width", &arg)) {
3613 if (*arg == '=')
3614 width = strtoul(arg + 1, &end, 10);
3615 else if (!*arg && !av[1])
3616 die_want_option("--stat-width");
3617 else if (!*arg) {
3618 width = strtoul(av[1], &end, 10);
3619 argcount = 2;
3620 }
3621 } else if (skip_prefix(arg, "-name-width", &arg)) {
3622 if (*arg == '=')
3623 name_width = strtoul(arg + 1, &end, 10);
3624 else if (!*arg && !av[1])
3625 die_want_option("--stat-name-width");
3626 else if (!*arg) {
3627 name_width = strtoul(av[1], &end, 10);
3628 argcount = 2;
3629 }
3630 } else if (skip_prefix(arg, "-graph-width", &arg)) {
3631 if (*arg == '=')
3632 graph_width = strtoul(arg + 1, &end, 10);
3633 else if (!*arg && !av[1])
3634 die_want_option("--stat-graph-width");
3635 else if (!*arg) {
3636 graph_width = strtoul(av[1], &end, 10);
3637 argcount = 2;
3638 }
3639 } else if (skip_prefix(arg, "-count", &arg)) {
3640 if (*arg == '=')
3641 count = strtoul(arg + 1, &end, 10);
3642 else if (!*arg && !av[1])
3643 die_want_option("--stat-count");
3644 else if (!*arg) {
3645 count = strtoul(av[1], &end, 10);
3646 argcount = 2;
3647 }
3648 }
3649 break;
3650 case '=':
3651 width = strtoul(arg+1, &end, 10);
3652 if (*end == ',')
3653 name_width = strtoul(end+1, &end, 10);
3654 if (*end == ',')
3655 count = strtoul(end+1, &end, 10);
3656 }
3657 /* Important! This checks all the error cases! */
3659 if (*end)
3660 return 0;
3661 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3662 options->stat_name_width = name_width;
3663 options->stat_graph_width = graph_width;
3664 options->stat_width = width;
3665 options->stat_count = count;
3666 return argcount;
3667}
3668static int parse_dirstat_opt(struct diff_options *options, const char *params)
3670{
3671 struct strbuf errmsg = STRBUF_INIT;
3672 if (parse_dirstat_params(options, params, &errmsg))
3673 die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3674 errmsg.buf);
3675 strbuf_release(&errmsg);
3676 /*
3677 * The caller knows a dirstat-related option is given from the command
3678 * line; allow it to say "return this_function();"
3679 */
3680 options->output_format |= DIFF_FORMAT_DIRSTAT;
3681 return 1;
3682}
3683static int parse_submodule_opt(struct diff_options *options, const char *value)
3685{
3686 if (parse_submodule_params(options, value))
3687 die(_("Failed to parse --submodule option parameter: '%s'"),
3688 value);
3689 return 1;
3690}
3691static const char diff_status_letters[] = {
3693 DIFF_STATUS_ADDED,
3694 DIFF_STATUS_COPIED,
3695 DIFF_STATUS_DELETED,
3696 DIFF_STATUS_MODIFIED,
3697 DIFF_STATUS_RENAMED,
3698 DIFF_STATUS_TYPE_CHANGED,
3699 DIFF_STATUS_UNKNOWN,
3700 DIFF_STATUS_UNMERGED,
3701 DIFF_STATUS_FILTER_AON,
3702 DIFF_STATUS_FILTER_BROKEN,
3703 '\0',
3704};
3705static unsigned int filter_bit['Z' + 1];
3707static void prepare_filter_bits(void)
3709{
3710 int i;
3711 if (!filter_bit[DIFF_STATUS_ADDED]) {
3713 for (i = 0; diff_status_letters[i]; i++)
3714 filter_bit[(int) diff_status_letters[i]] = (1 << i);
3715 }
3716}
3717static unsigned filter_bit_tst(char status, const struct diff_options *opt)
3719{
3720 return opt->filter & filter_bit[(int) status];
3721}
3722static int parse_diff_filter_opt(const char *optarg, struct diff_options *opt)
3724{
3725 int i, optch;
3726 prepare_filter_bits();
3728 /*
3730 * If there is a negation e.g. 'd' in the input, and we haven't
3731 * initialized the filter field with another --diff-filter, start
3732 * from full set of bits, except for AON.
3733 */
3734 if (!opt->filter) {
3735 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3736 if (optch < 'a' || 'z' < optch)
3737 continue;
3738 opt->filter = (1 << (ARRAY_SIZE(diff_status_letters) - 1)) - 1;
3739 opt->filter &= ~filter_bit[DIFF_STATUS_FILTER_AON];
3740 break;
3741 }
3742 }
3743 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3745 unsigned int bit;
3746 int negate;
3747 if ('a' <= optch && optch <= 'z') {
3749 negate = 1;
3750 optch = toupper(optch);
3751 } else {
3752 negate = 0;
3753 }
3754 bit = (0 <= optch && optch <= 'Z') ? filter_bit[optch] : 0;
3756 if (!bit)
3757 return optarg[i];
3758 if (negate)
3759 opt->filter &= ~bit;
3760 else
3761 opt->filter |= bit;
3762 }
3763 return 0;
3764}
3765static void enable_patch_output(int *fmt) {
3767 *fmt &= ~DIFF_FORMAT_NO_OUTPUT;
3768 *fmt |= DIFF_FORMAT_PATCH;
3769}
3770static int parse_ws_error_highlight_opt(struct diff_options *opt, const char *arg)
3772{
3773 int val = parse_ws_error_highlight(arg);
3774 if (val < 0) {
3776 error("unknown value after ws-error-highlight=%.*s",
3777 -1 - val, arg);
3778 return 0;
3779 }
3780 opt->ws_error_highlight = val;
3781 return 1;
3782}
3783int diff_opt_parse(struct diff_options *options,
3785 const char **av, int ac, const char *prefix)
3786{
3787 const char *arg = av[0];
3788 const char *optarg;
3789 int argcount;
3790 if (!prefix)
3792 prefix = "";
3793 /* Output format options */
3795 if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch")
3796 || opt_arg(arg, 'U', "unified", &options->context))
3797 enable_patch_output(&options->output_format);
3798 else if (!strcmp(arg, "--raw"))
3799 options->output_format |= DIFF_FORMAT_RAW;
3800 else if (!strcmp(arg, "--patch-with-raw")) {
3801 enable_patch_output(&options->output_format);
3802 options->output_format |= DIFF_FORMAT_RAW;
3803 } else if (!strcmp(arg, "--numstat"))
3804 options->output_format |= DIFF_FORMAT_NUMSTAT;
3805 else if (!strcmp(arg, "--shortstat"))
3806 options->output_format |= DIFF_FORMAT_SHORTSTAT;
3807 else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3808 return parse_dirstat_opt(options, "");
3809 else if (skip_prefix(arg, "-X", &arg))
3810 return parse_dirstat_opt(options, arg);
3811 else if (skip_prefix(arg, "--dirstat=", &arg))
3812 return parse_dirstat_opt(options, arg);
3813 else if (!strcmp(arg, "--cumulative"))
3814 return parse_dirstat_opt(options, "cumulative");
3815 else if (!strcmp(arg, "--dirstat-by-file"))
3816 return parse_dirstat_opt(options, "files");
3817 else if (skip_prefix(arg, "--dirstat-by-file=", &arg)) {
3818 parse_dirstat_opt(options, "files");
3819 return parse_dirstat_opt(options, arg);
3820 }
3821 else if (!strcmp(arg, "--check"))
3822 options->output_format |= DIFF_FORMAT_CHECKDIFF;
3823 else if (!strcmp(arg, "--summary"))
3824 options->output_format |= DIFF_FORMAT_SUMMARY;
3825 else if (!strcmp(arg, "--patch-with-stat")) {
3826 enable_patch_output(&options->output_format);
3827 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3828 } else if (!strcmp(arg, "--name-only"))
3829 options->output_format |= DIFF_FORMAT_NAME;
3830 else if (!strcmp(arg, "--name-status"))
3831 options->output_format |= DIFF_FORMAT_NAME_STATUS;
3832 else if (!strcmp(arg, "-s") || !strcmp(arg, "--no-patch"))
3833 options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3834 else if (starts_with(arg, "--stat"))
3835 /* --stat, --stat-width, --stat-name-width, or --stat-count */
3836 return stat_opt(options, av);
3837 /* renames options */
3839 else if (starts_with(arg, "-B") || starts_with(arg, "--break-rewrites=") ||
3840 !strcmp(arg, "--break-rewrites")) {
3841 if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3842 return error("invalid argument to -B: %s", arg+2);
3843 }
3844 else if (starts_with(arg, "-M") || starts_with(arg, "--find-renames=") ||
3845 !strcmp(arg, "--find-renames")) {
3846 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3847 return error("invalid argument to -M: %s", arg+2);
3848 options->detect_rename = DIFF_DETECT_RENAME;
3849 }
3850 else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3851 options->irreversible_delete = 1;
3852 }
3853 else if (starts_with(arg, "-C") || starts_with(arg, "--find-copies=") ||
3854 !strcmp(arg, "--find-copies")) {
3855 if (options->detect_rename == DIFF_DETECT_COPY)
3856 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3857 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3858 return error("invalid argument to -C: %s", arg+2);
3859 options->detect_rename = DIFF_DETECT_COPY;
3860 }
3861 else if (!strcmp(arg, "--no-renames"))
3862 options->detect_rename = 0;
3863 else if (!strcmp(arg, "--rename-empty"))
3864 DIFF_OPT_SET(options, RENAME_EMPTY);
3865 else if (!strcmp(arg, "--no-rename-empty"))
3866 DIFF_OPT_CLR(options, RENAME_EMPTY);
3867 else if (!strcmp(arg, "--relative"))
3868 DIFF_OPT_SET(options, RELATIVE_NAME);
3869 else if (skip_prefix(arg, "--relative=", &arg)) {
3870 DIFF_OPT_SET(options, RELATIVE_NAME);
3871 options->prefix = arg;
3872 }
3873 /* xdiff options */
3875 else if (!strcmp(arg, "--minimal"))
3876 DIFF_XDL_SET(options, NEED_MINIMAL);
3877 else if (!strcmp(arg, "--no-minimal"))
3878 DIFF_XDL_CLR(options, NEED_MINIMAL);
3879 else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3880 DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3881 else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3882 DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3883 else if (!strcmp(arg, "--ignore-space-at-eol"))
3884 DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3885 else if (!strcmp(arg, "--ignore-blank-lines"))
3886 DIFF_XDL_SET(options, IGNORE_BLANK_LINES);
3887 else if (!strcmp(arg, "--indent-heuristic"))
3888 DIFF_XDL_SET(options, INDENT_HEURISTIC);
3889 else if (!strcmp(arg, "--no-indent-heuristic"))
3890 DIFF_XDL_CLR(options, INDENT_HEURISTIC);
3891 else if (!strcmp(arg, "--patience"))
3892 options->xdl_opts = DIFF_WITH_ALG(options, PATIENCE_DIFF);
3893 else if (!strcmp(arg, "--histogram"))
3894 options->xdl_opts = DIFF_WITH_ALG(options, HISTOGRAM_DIFF);
3895 else if ((argcount = parse_long_opt("diff-algorithm", av, &optarg))) {
3896 long value = parse_algorithm_value(optarg);
3897 if (value < 0)
3898 return error("option diff-algorithm accepts \"myers\", "
3899 "\"minimal\", \"patience\" and \"histogram\"");
3900 /* clear out previous settings */
3901 DIFF_XDL_CLR(options, NEED_MINIMAL);
3902 options->xdl_opts &= ~XDF_DIFF_ALGORITHM_MASK;
3903 options->xdl_opts |= value;
3904 return argcount;
3905 }
3906 /* flags options */
3908 else if (!strcmp(arg, "--binary")) {
3909 enable_patch_output(&options->output_format);
3910 DIFF_OPT_SET(options, BINARY);
3911 }
3912 else if (!strcmp(arg, "--full-index"))
3913 DIFF_OPT_SET(options, FULL_INDEX);
3914 else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
3915 DIFF_OPT_SET(options, TEXT);
3916 else if (!strcmp(arg, "-R"))
3917 DIFF_OPT_SET(options, REVERSE_DIFF);
3918 else if (!strcmp(arg, "--find-copies-harder"))
3919 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3920 else if (!strcmp(arg, "--follow"))
3921 DIFF_OPT_SET(options, FOLLOW_RENAMES);
3922 else if (!strcmp(arg, "--no-follow")) {
3923 DIFF_OPT_CLR(options, FOLLOW_RENAMES);
3924 DIFF_OPT_CLR(options, DEFAULT_FOLLOW_RENAMES);
3925 } else if (!strcmp(arg, "--color"))
3926 options->use_color = 1;
3927 else if (skip_prefix(arg, "--color=", &arg)) {
3928 int value = git_config_colorbool(NULL, arg);
3929 if (value < 0)
3930 return error("option `color' expects \"always\", \"auto\", or \"never\"");
3931 options->use_color = value;
3932 }
3933 else if (!strcmp(arg, "--no-color"))
3934 options->use_color = 0;
3935 else if (!strcmp(arg, "--color-words")) {
3936 options->use_color = 1;
3937 options->word_diff = DIFF_WORDS_COLOR;
3938 }
3939 else if (skip_prefix(arg, "--color-words=", &arg)) {
3940 options->use_color = 1;
3941 options->word_diff = DIFF_WORDS_COLOR;
3942 options->word_regex = arg;
3943 }
3944 else if (!strcmp(arg, "--word-diff")) {
3945 if (options->word_diff == DIFF_WORDS_NONE)
3946 options->word_diff = DIFF_WORDS_PLAIN;
3947 }
3948 else if (skip_prefix(arg, "--word-diff=", &arg)) {
3949 if (!strcmp(arg, "plain"))
3950 options->word_diff = DIFF_WORDS_PLAIN;
3951 else if (!strcmp(arg, "color")) {
3952 options->use_color = 1;
3953 options->word_diff = DIFF_WORDS_COLOR;
3954 }
3955 else if (!strcmp(arg, "porcelain"))
3956 options->word_diff = DIFF_WORDS_PORCELAIN;
3957 else if (!strcmp(arg, "none"))
3958 options->word_diff = DIFF_WORDS_NONE;
3959 else
3960 die("bad --word-diff argument: %s", arg);
3961 }
3962 else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
3963 if (options->word_diff == DIFF_WORDS_NONE)
3964 options->word_diff = DIFF_WORDS_PLAIN;
3965 options->word_regex = optarg;
3966 return argcount;
3967 }
3968 else if (!strcmp(arg, "--exit-code"))
3969 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3970 else if (!strcmp(arg, "--quiet"))
3971 DIFF_OPT_SET(options, QUICK);
3972 else if (!strcmp(arg, "--ext-diff"))
3973 DIFF_OPT_SET(options, ALLOW_EXTERNAL);
3974 else if (!strcmp(arg, "--no-ext-diff"))
3975 DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
3976 else if (!strcmp(arg, "--textconv"))
3977 DIFF_OPT_SET(options, ALLOW_TEXTCONV);
3978 else if (!strcmp(arg, "--no-textconv"))
3979 DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
3980 else if (!strcmp(arg, "--ignore-submodules")) {
3981 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3982 handle_ignore_submodules_arg(options, "all");
3983 } else if (skip_prefix(arg, "--ignore-submodules=", &arg)) {
3984 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3985 handle_ignore_submodules_arg(options, arg);
3986 } else if (!strcmp(arg, "--submodule"))
3987 options->submodule_format = DIFF_SUBMODULE_LOG;
3988 else if (skip_prefix(arg, "--submodule=", &arg))
3989 return parse_submodule_opt(options, arg);
3990 else if (skip_prefix(arg, "--ws-error-highlight=", &arg))
3991 return parse_ws_error_highlight_opt(options, arg);
3992 else if (!strcmp(arg, "--ita-invisible-in-index"))
3993 options->ita_invisible_in_index = 1;
3994 else if (!strcmp(arg, "--ita-visible-in-index"))
3995 options->ita_invisible_in_index = 0;
3996 /* misc options */
3998 else if (!strcmp(arg, "-z"))
3999 options->line_termination = 0;
4000 else if ((argcount = short_opt('l', av, &optarg))) {
4001 options->rename_limit = strtoul(optarg, NULL, 10);
4002 return argcount;
4003 }
4004 else if ((argcount = short_opt('S', av, &optarg))) {
4005 options->pickaxe = optarg;
4006 options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
4007 return argcount;
4008 } else if ((argcount = short_opt('G', av, &optarg))) {
4009 options->pickaxe = optarg;
4010 options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
4011 return argcount;
4012 }
4013 else if (!strcmp(arg, "--pickaxe-all"))
4014 options->pickaxe_opts |= DIFF_PICKAXE_ALL;
4015 else if (!strcmp(arg, "--pickaxe-regex"))
4016 options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
4017 else if ((argcount = short_opt('O', av, &optarg))) {
4018 options->orderfile = prefix_filename(prefix, optarg);
4019 return argcount;
4020 }
4021 else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
4022 int offending = parse_diff_filter_opt(optarg, options);
4023 if (offending)
4024 die("unknown change class '%c' in --diff-filter=%s",
4025 offending, optarg);
4026 return argcount;
4027 }
4028 else if (!strcmp(arg, "--no-abbrev"))
4029 options->abbrev = 0;
4030 else if (!strcmp(arg, "--abbrev"))
4031 options->abbrev = DEFAULT_ABBREV;
4032 else if (skip_prefix(arg, "--abbrev=", &arg)) {
4033 options->abbrev = strtoul(arg, NULL, 10);
4034 if (options->abbrev < MINIMUM_ABBREV)
4035 options->abbrev = MINIMUM_ABBREV;
4036 else if (40 < options->abbrev)
4037 options->abbrev = 40;
4038 }
4039 else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
4040 options->a_prefix = optarg;
4041 return argcount;
4042 }
4043 else if ((argcount = parse_long_opt("line-prefix", av, &optarg))) {
4044 options->line_prefix = optarg;
4045 options->line_prefix_length = strlen(options->line_prefix);
4046 graph_setup_line_prefix(options);
4047 return argcount;
4048 }
4049 else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
4050 options->b_prefix = optarg;
4051 return argcount;
4052 }
4053 else if (!strcmp(arg, "--no-prefix"))
4054 options->a_prefix = options->b_prefix = "";
4055 else if (opt_arg(arg, '\0', "inter-hunk-context",
4056 &options->interhunkcontext))
4057 ;
4058 else if (!strcmp(arg, "-W"))
4059 DIFF_OPT_SET(options, FUNCCONTEXT);
4060 else if (!strcmp(arg, "--function-context"))
4061 DIFF_OPT_SET(options, FUNCCONTEXT);
4062 else if (!strcmp(arg, "--no-function-context"))
4063 DIFF_OPT_CLR(options, FUNCCONTEXT);
4064 else if ((argcount = parse_long_opt("output", av, &optarg))) {
4065 char *path = prefix_filename(prefix, optarg);
4066 options->file = xfopen(path, "w");
4067 options->close_file = 1;
4068 if (options->use_color != GIT_COLOR_ALWAYS)
4069 options->use_color = GIT_COLOR_NEVER;
4070 free(path);
4071 return argcount;
4072 } else
4073 return 0;
4074 return 1;
4075}
4076int parse_rename_score(const char **cp_p)
4078{
4079 unsigned long num, scale;
4080 int ch, dot;
4081 const char *cp = *cp_p;
4082 num = 0;
4084 scale = 1;
4085 dot = 0;
4086 for (;;) {
4087 ch = *cp;
4088 if ( !dot && ch == '.' ) {
4089 scale = 1;
4090 dot = 1;
4091 } else if ( ch == '%' ) {
4092 scale = dot ? scale*100 : 100;
4093 cp++; /* % is always at the end */
4094 break;
4095 } else if ( ch >= '0' && ch <= '9' ) {
4096 if ( scale < 100000 ) {
4097 scale *= 10;
4098 num = (num*10) + (ch-'0');
4099 }
4100 } else {
4101 break;
4102 }
4103 cp++;
4104 }
4105 *cp_p = cp;
4106 /* user says num divided by scale and we say internally that
4108 * is MAX_SCORE * num / scale.
4109 */
4110 return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
4111}
4112static int diff_scoreopt_parse(const char *opt)
4114{
4115 int opt1, opt2, cmd;
4116 if (*opt++ != '-')
4118 return -1;
4119 cmd = *opt++;
4120 if (cmd == '-') {
4121 /* convert the long-form arguments into short-form versions */
4122 if (skip_prefix(opt, "break-rewrites", &opt)) {
4123 if (*opt == 0 || *opt++ == '=')
4124 cmd = 'B';
4125 } else if (skip_prefix(opt, "find-copies", &opt)) {
4126 if (*opt == 0 || *opt++ == '=')
4127 cmd = 'C';
4128 } else if (skip_prefix(opt, "find-renames", &opt)) {
4129 if (*opt == 0 || *opt++ == '=')
4130 cmd = 'M';
4131 }
4132 }
4133 if (cmd != 'M' && cmd != 'C' && cmd != 'B')
4134 return -1; /* that is not a -M, -C, or -B option */
4135 opt1 = parse_rename_score(&opt);
4137 if (cmd != 'B')
4138 opt2 = 0;
4139 else {
4140 if (*opt == 0)
4141 opt2 = 0;
4142 else if (*opt != '/')
4143 return -1; /* we expect -B80/99 or -B80 */
4144 else {
4145 opt++;
4146 opt2 = parse_rename_score(&opt);
4147 }
4148 }
4149 if (*opt != 0)
4150 return -1;
4151 return opt1 | (opt2 << 16);
4152}
4153struct diff_queue_struct diff_queued_diff;
4155void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
4157{
4158 ALLOC_GROW(queue->queue, queue->nr + 1, queue->alloc);
4159 queue->queue[queue->nr++] = dp;
4160}
4161struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
4163 struct diff_filespec *one,
4164 struct diff_filespec *two)
4165{
4166 struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
4167 dp->one = one;
4168 dp->two = two;
4169 if (queue)
4170 diff_q(queue, dp);
4171 return dp;
4172}
4173void diff_free_filepair(struct diff_filepair *p)
4175{
4176 free_filespec(p->one);
4177 free_filespec(p->two);
4178 free(p);
4179}
4180const char *diff_aligned_abbrev(const struct object_id *oid, int len)
4182{
4183 int abblen;
4184 const char *abbrev;
4185 if (len == GIT_SHA1_HEXSZ)
4187 return oid_to_hex(oid);
4188 abbrev = diff_abbrev_oid(oid, len);
4190 abblen = strlen(abbrev);
4191 /*
4193 * In well-behaved cases, where the abbbreviated result is the
4194 * same as the requested length, append three dots after the
4195 * abbreviation (hence the whole logic is limited to the case
4196 * where abblen < 37); when the actual abbreviated result is a
4197 * bit longer than the requested length, we reduce the number
4198 * of dots so that they match the well-behaved ones. However,
4199 * if the actual abbreviation is longer than the requested
4200 * length by more than three, we give up on aligning, and add
4201 * three dots anyway, to indicate that the output is not the
4202 * full object name. Yes, this may be suboptimal, but this
4203 * appears only in "diff --raw --abbrev" output and it is not
4204 * worth the effort to change it now. Note that this would
4205 * likely to work fine when the automatic sizing of default
4206 * abbreviation length is used--we would be fed -1 in "len" in
4207 * that case, and will end up always appending three-dots, but
4208 * the automatic sizing is supposed to give abblen that ensures
4209 * uniqueness across all objects (statistically speaking).
4210 */
4211 if (abblen < GIT_SHA1_HEXSZ - 3) {
4212 static char hex[GIT_MAX_HEXSZ + 1];
4213 if (len < abblen && abblen <= len + 2)
4214 xsnprintf(hex, sizeof(hex), "%s%.*s", abbrev, len+3-abblen, "..");
4215 else
4216 xsnprintf(hex, sizeof(hex), "%s...", abbrev);
4217 return hex;
4218 }
4219 return oid_to_hex(oid);
4221}
4222static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
4224{
4225 int line_termination = opt->line_termination;
4226 int inter_name_termination = line_termination ? '\t' : '\0';
4227 fprintf(opt->file, "%s", diff_line_prefix(opt));
4229 if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
4230 fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
4231 diff_aligned_abbrev(&p->one->oid, opt->abbrev));
4232 fprintf(opt->file, "%s ",
4233 diff_aligned_abbrev(&p->two->oid, opt->abbrev));
4234 }
4235 if (p->score) {
4236 fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
4237 inter_name_termination);
4238 } else {
4239 fprintf(opt->file, "%c%c", p->status, inter_name_termination);
4240 }
4241 if (p->status == DIFF_STATUS_COPIED ||
4243 p->status == DIFF_STATUS_RENAMED) {
4244 const char *name_a, *name_b;
4245 name_a = p->one->path;
4246 name_b = p->two->path;
4247 strip_prefix(opt->prefix_length, &name_a, &name_b);
4248 write_name_quoted(name_a, opt->file, inter_name_termination);
4249 write_name_quoted(name_b, opt->file, line_termination);
4250 } else {
4251 const char *name_a, *name_b;
4252 name_a = p->one->mode ? p->one->path : p->two->path;
4253 name_b = NULL;
4254 strip_prefix(opt->prefix_length, &name_a, &name_b);
4255 write_name_quoted(name_a, opt->file, line_termination);
4256 }
4257}
4258int diff_unmodified_pair(struct diff_filepair *p)
4260{
4261 /* This function is written stricter than necessary to support
4262 * the currently implemented transformers, but the idea is to
4263 * let transformers to produce diff_filepairs any way they want,
4264 * and filter and clean them up here before producing the output.
4265 */
4266 struct diff_filespec *one = p->one, *two = p->two;
4267 if (DIFF_PAIR_UNMERGED(p))
4269 return 0; /* unmerged is interesting */
4270 /* deletion, addition, mode or type change
4272 * and rename are all interesting.
4273 */
4274 if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
4275 DIFF_PAIR_MODE_CHANGED(p) ||
4276 strcmp(one->path, two->path))
4277 return 0;
4278 /* both are valid and point at the same path. that is, we are
4280 * dealing with a change.
4281 */
4282 if (one->oid_valid && two->oid_valid &&
4283 !oidcmp(&one->oid, &two->oid) &&
4284 !one->dirty_submodule && !two->dirty_submodule)
4285 return 1; /* no change */
4286 if (!one->oid_valid && !two->oid_valid)
4287 return 1; /* both look at the same file on the filesystem. */
4288 return 0;
4289}
4290static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
4292{
4293 if (diff_unmodified_pair(p))
4294 return;
4295 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4297 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4298 return; /* no tree diffs in patch format */
4299 run_diff(p, o);
4301}
4302static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
4304 struct diffstat_t *diffstat)
4305{
4306 if (diff_unmodified_pair(p))
4307 return;
4308 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4310 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4311 return; /* no useful stat for tree diffs */
4312 run_diffstat(p, o, diffstat);
4314}
4315static void diff_flush_checkdiff(struct diff_filepair *p,
4317 struct diff_options *o)
4318{
4319 if (diff_unmodified_pair(p))
4320 return;
4321 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4323 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4324 return; /* nothing to check in tree diffs */
4325 run_checkdiff(p, o);
4327}
4328int diff_queue_is_empty(void)
4330{
4331 struct diff_queue_struct *q = &diff_queued_diff;
4332 int i;
4333 for (i = 0; i < q->nr; i++)
4334 if (!diff_unmodified_pair(q->queue[i]))
4335 return 0;
4336 return 1;
4337}
4338#if DIFF_DEBUG
4340void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
4341{
4342 fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
4343 x, one ? one : "",
4344 s->path,
4345 DIFF_FILE_VALID(s) ? "valid" : "invalid",
4346 s->mode,
4347 s->oid_valid ? oid_to_hex(&s->oid) : "");
4348 fprintf(stderr, "queue[%d] %s size %lu\n",
4349 x, one ? one : "",
4350 s->size);
4351}
4352void diff_debug_filepair(const struct diff_filepair *p, int i)
4354{
4355 diff_debug_filespec(p->one, i, "one");
4356 diff_debug_filespec(p->two, i, "two");
4357 fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
4358 p->score, p->status ? p->status : '?',
4359 p->one->rename_used, p->broken_pair);
4360}
4361void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
4363{
4364 int i;
4365 if (msg)
4366 fprintf(stderr, "%s\n", msg);
4367 fprintf(stderr, "q->nr = %d\n", q->nr);
4368 for (i = 0; i < q->nr; i++) {
4369 struct diff_filepair *p = q->queue[i];
4370 diff_debug_filepair(p, i);
4371 }
4372}
4373#endif
4374static void diff_resolve_rename_copy(void)
4376{
4377 int i;
4378 struct diff_filepair *p;
4379 struct diff_queue_struct *q = &diff_queued_diff;
4380 diff_debug_queue("resolve-rename-copy", q);
4382 for (i = 0; i < q->nr; i++) {
4384 p = q->queue[i];
4385 p->status = 0; /* undecided */
4386 if (DIFF_PAIR_UNMERGED(p))
4387 p->status = DIFF_STATUS_UNMERGED;
4388 else if (!DIFF_FILE_VALID(p->one))
4389 p->status = DIFF_STATUS_ADDED;
4390 else if (!DIFF_FILE_VALID(p->two))
4391 p->status = DIFF_STATUS_DELETED;
4392 else if (DIFF_PAIR_TYPE_CHANGED(p))
4393 p->status = DIFF_STATUS_TYPE_CHANGED;
4394 /* from this point on, we are dealing with a pair
4396 * whose both sides are valid and of the same type, i.e.
4397 * either in-place edit or rename/copy edit.
4398 */
4399 else if (DIFF_PAIR_RENAME(p)) {
4400 /*
4401 * A rename might have re-connected a broken
4402 * pair up, causing the pathnames to be the
4403 * same again. If so, that's not a rename at
4404 * all, just a modification..
4405 *
4406 * Otherwise, see if this source was used for
4407 * multiple renames, in which case we decrement
4408 * the count, and call it a copy.
4409 */
4410 if (!strcmp(p->one->path, p->two->path))
4411 p->status = DIFF_STATUS_MODIFIED;
4412 else if (--p->one->rename_used > 0)
4413 p->status = DIFF_STATUS_COPIED;
4414 else
4415 p->status = DIFF_STATUS_RENAMED;
4416 }
4417 else if (oidcmp(&p->one->oid, &p->two->oid) ||
4418 p->one->mode != p->two->mode ||
4419 p->one->dirty_submodule ||
4420 p->two->dirty_submodule ||
4421 is_null_oid(&p->one->oid))
4422 p->status = DIFF_STATUS_MODIFIED;
4423 else {
4424 /* This is a "no-change" entry and should not
4425 * happen anymore, but prepare for broken callers.
4426 */
4427 error("feeding unmodified %s to diffcore",
4428 p->one->path);
4429 p->status = DIFF_STATUS_UNKNOWN;
4430 }
4431 }
4432 diff_debug_queue("resolve-rename-copy done", q);
4433}
4434static int check_pair_status(struct diff_filepair *p)
4436{
4437 switch (p->status) {
4438 case DIFF_STATUS_UNKNOWN:
4439 return 0;
4440 case 0:
4441 die("internal error in diff-resolve-rename-copy");
4442 default:
4443 return 1;
4444 }
4445}
4446static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
4448{
4449 int fmt = opt->output_format;
4450 if (fmt & DIFF_FORMAT_CHECKDIFF)
4452 diff_flush_checkdiff(p, opt);
4453 else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
4454 diff_flush_raw(p, opt);
4455 else if (fmt & DIFF_FORMAT_NAME) {
4456 const char *name_a, *name_b;
4457 name_a = p->two->path;
4458 name_b = NULL;
4459 strip_prefix(opt->prefix_length, &name_a, &name_b);
4460 fprintf(opt->file, "%s", diff_line_prefix(opt));
4461 write_name_quoted(name_a, opt->file, opt->line_termination);
4462 }
4463}
4464static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
4466{
4467 if (fs->mode)
4468 fprintf(file, " %s mode %06o ", newdelete, fs->mode);
4469 else
4470 fprintf(file, " %s ", newdelete);
4471 write_name_quoted(fs->path, file, '\n');
4472}
4473static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
4476 const char *line_prefix)
4477{
4478 if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
4479 fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
4480 p->two->mode, show_name ? ' ' : '\n');
4481 if (show_name) {
4482 write_name_quoted(p->two->path, file, '\n');
4483 }
4484 }
4485}
4486static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
4488 const char *line_prefix)
4489{
4490 char *names = pprint_rename(p->one->path, p->two->path);
4491 fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
4493 free(names);
4494 show_mode_change(file, p, 0, line_prefix);
4495}
4496static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
4498{
4499 FILE *file = opt->file;
4500 const char *line_prefix = diff_line_prefix(opt);
4501 switch(p->status) {
4503 case DIFF_STATUS_DELETED:
4504 fputs(line_prefix, file);
4505 show_file_mode_name(file, "delete", p->one);
4506 break;
4507 case DIFF_STATUS_ADDED:
4508 fputs(line_prefix, file);
4509 show_file_mode_name(file, "create", p->two);
4510 break;
4511 case DIFF_STATUS_COPIED:
4512 fputs(line_prefix, file);
4513 show_rename_copy(file, "copy", p, line_prefix);
4514 break;
4515 case DIFF_STATUS_RENAMED:
4516 fputs(line_prefix, file);
4517 show_rename_copy(file, "rename", p, line_prefix);
4518 break;
4519 default:
4520 if (p->score) {
4521 fprintf(file, "%s rewrite ", line_prefix);
4522 write_name_quoted(p->two->path, file, ' ');
4523 fprintf(file, "(%d%%)\n", similarity_index(p));
4524 }
4525 show_mode_change(file, p, !p->score, line_prefix);
4526 break;
4527 }
4528}
4529struct patch_id_t {
4531 git_SHA_CTX *ctx;
4532 int patchlen;
4533};
4534static int remove_space(char *line, int len)
4536{
4537 int i;
4538 char *dst = line;
4539 unsigned char c;
4540 for (i = 0; i < len; i++)
4542 if (!isspace((c = line[i])))
4543 *dst++ = c;
4544 return dst - line;
4546}
4547static void patch_id_consume(void *priv, char *line, unsigned long len)
4549{
4550 struct patch_id_t *data = priv;
4551 int new_len;
4552 /* Ignore line numbers when computing the SHA1 of the patch */
4554 if (starts_with(line, "@@ -"))
4555 return;
4556 new_len = remove_space(line, len);
4558 git_SHA1_Update(data->ctx, line, new_len);
4560 data->patchlen += new_len;
4561}
4562static void patch_id_add_string(git_SHA_CTX *ctx, const char *str)
4564{
4565 git_SHA1_Update(ctx, str, strlen(str));
4566}
4567static void patch_id_add_mode(git_SHA_CTX *ctx, unsigned mode)
4569{
4570 /* large enough for 2^32 in octal */
4571 char buf[12];
4572 int len = xsnprintf(buf, sizeof(buf), "%06o", mode);
4573 git_SHA1_Update(ctx, buf, len);
4574}
4575/* returns 0 upon success, and writes result into sha1 */
4577static int diff_get_patch_id(struct diff_options *options, struct object_id *oid, int diff_header_only)
4578{
4579 struct diff_queue_struct *q = &diff_queued_diff;
4580 int i;
4581 git_SHA_CTX ctx;
4582 struct patch_id_t data;
4583 git_SHA1_Init(&ctx);
4585 memset(&data, 0, sizeof(struct patch_id_t));
4586 data.ctx = &ctx;
4587 for (i = 0; i < q->nr; i++) {
4589 xpparam_t xpp;
4590 xdemitconf_t xecfg;
4591 mmfile_t mf1, mf2;
4592 struct diff_filepair *p = q->queue[i];
4593 int len1, len2;
4594 memset(&xpp, 0, sizeof(xpp));
4596 memset(&xecfg, 0, sizeof(xecfg));
4597 if (p->status == 0)
4598 return error("internal diff status error");
4599 if (p->status == DIFF_STATUS_UNKNOWN)
4600 continue;
4601 if (diff_unmodified_pair(p))
4602 continue;
4603 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4604 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4605 continue;
4606 if (DIFF_PAIR_UNMERGED(p))
4607 continue;
4608 diff_fill_oid_info(p->one);
4610 diff_fill_oid_info(p->two);
4611 len1 = remove_space(p->one->path, strlen(p->one->path));
4613 len2 = remove_space(p->two->path, strlen(p->two->path));
4614 patch_id_add_string(&ctx, "diff--git");
4615 patch_id_add_string(&ctx, "a/");
4616 git_SHA1_Update(&ctx, p->one->path, len1);
4617 patch_id_add_string(&ctx, "b/");
4618 git_SHA1_Update(&ctx, p->two->path, len2);
4619 if (p->one->mode == 0) {
4621 patch_id_add_string(&ctx, "newfilemode");
4622 patch_id_add_mode(&ctx, p->two->mode);
4623 patch_id_add_string(&ctx, "---/dev/null");
4624 patch_id_add_string(&ctx, "+++b/");
4625 git_SHA1_Update(&ctx, p->two->path, len2);
4626 } else if (p->two->mode == 0) {
4627 patch_id_add_string(&ctx, "deletedfilemode");
4628 patch_id_add_mode(&ctx, p->one->mode);
4629 patch_id_add_string(&ctx, "---a/");
4630 git_SHA1_Update(&ctx, p->one->path, len1);
4631 patch_id_add_string(&ctx, "+++/dev/null");
4632 } else {
4633 patch_id_add_string(&ctx, "---a/");
4634 git_SHA1_Update(&ctx, p->one->path, len1);
4635 patch_id_add_string(&ctx, "+++b/");
4636 git_SHA1_Update(&ctx, p->two->path, len2);
4637 }
4638 if (diff_header_only)
4640 continue;
4641 if (fill_mmfile(&mf1, p->one) < 0 ||
4643 fill_mmfile(&mf2, p->two) < 0)
4644 return error("unable to read files to diff");
4645 if (diff_filespec_is_binary(p->one) ||
4647 diff_filespec_is_binary(p->two)) {
4648 git_SHA1_Update(&ctx, oid_to_hex(&p->one->oid),
4649 GIT_SHA1_HEXSZ);
4650 git_SHA1_Update(&ctx, oid_to_hex(&p->two->oid),
4651 GIT_SHA1_HEXSZ);
4652 continue;
4653 }
4654 xpp.flags = 0;
4656 xecfg.ctxlen = 3;
4657 xecfg.flags = 0;
4658 if (xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4659 &xpp, &xecfg))
4660 return error("unable to generate patch-id diff for %s",
4661 p->one->path);
4662 }
4663 git_SHA1_Final(oid->hash, &ctx);
4665 return 0;
4666}
4667int diff_flush_patch_id(struct diff_options *options, struct object_id *oid, int diff_header_only)
4669{
4670 struct diff_queue_struct *q = &diff_queued_diff;
4671 int i;
4672 int result = diff_get_patch_id(options, oid, diff_header_only);
4673 for (i = 0; i < q->nr; i++)
4675 diff_free_filepair(q->queue[i]);
4676 free(q->queue);
4678 DIFF_QUEUE_CLEAR(q);
4679 return result;
4681}
4682static int is_summary_empty(const struct diff_queue_struct *q)
4684{
4685 int i;
4686 for (i = 0; i < q->nr; i++) {
4688 const struct diff_filepair *p = q->queue[i];
4689 switch (p->status) {
4691 case DIFF_STATUS_DELETED:
4692 case DIFF_STATUS_ADDED:
4693 case DIFF_STATUS_COPIED:
4694 case DIFF_STATUS_RENAMED:
4695 return 0;
4696 default:
4697 if (p->score)
4698 return 0;
4699 if (p->one->mode && p->two->mode &&
4700 p->one->mode != p->two->mode)
4701 return 0;
4702 break;
4703 }
4704 }
4705 return 1;
4706}
4707static const char rename_limit_warning[] =
4709N_("inexact rename detection was skipped due to too many files.");
4710static const char degrade_cc_to_c_warning[] =
4712N_("only found copies from modified paths due to too many files.");
4713static const char rename_limit_advice[] =
4715N_("you may want to set your %s variable to at least "
4716 "%d and retry the command.");
4717void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4719{
4720 if (degraded_cc)
4721 warning(_(degrade_cc_to_c_warning));
4722 else if (needed)
4723 warning(_(rename_limit_warning));
4724 else
4725 return;
4726 if (0 < needed && needed < 32767)
4727 warning(_(rename_limit_advice), varname, needed);
4728}
4729void diff_flush(struct diff_options *options)
4731{
4732 struct diff_queue_struct *q = &diff_queued_diff;
4733 int i, output_format = options->output_format;
4734 int separator = 0;
4735 int dirstat_by_line = 0;
4736 /*
4738 * Order: raw, stat, summary, patch
4739 * or: name/name-status/checkdiff (other bits clear)
4740 */
4741 if (!q->nr)
4742 goto free_queue;
4743 if (output_format & (DIFF_FORMAT_RAW |
4745 DIFF_FORMAT_NAME |
4746 DIFF_FORMAT_NAME_STATUS |
4747 DIFF_FORMAT_CHECKDIFF)) {
4748 for (i = 0; i < q->nr; i++) {
4749 struct diff_filepair *p = q->queue[i];
4750 if (check_pair_status(p))
4751 flush_one_pair(p, options);
4752 }
4753 separator++;
4754 }
4755 if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4757 dirstat_by_line = 1;
4758 if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4760 dirstat_by_line) {
4761 struct diffstat_t diffstat;
4762 memset(&diffstat, 0, sizeof(struct diffstat_t));
4764 for (i = 0; i < q->nr; i++) {
4765 struct diff_filepair *p = q->queue[i];
4766 if (check_pair_status(p))
4767 diff_flush_stat(p, options, &diffstat);
4768 }
4769 if (output_format & DIFF_FORMAT_NUMSTAT)
4770 show_numstat(&diffstat, options);
4771 if (output_format & DIFF_FORMAT_DIFFSTAT)
4772 show_stats(&diffstat, options);
4773 if (output_format & DIFF_FORMAT_SHORTSTAT)
4774 show_shortstats(&diffstat, options);
4775 if (output_format & DIFF_FORMAT_DIRSTAT && dirstat_by_line)
4776 show_dirstat_by_line(&diffstat, options);
4777 free_diffstat_info(&diffstat);
4778 separator++;
4779 }
4780 if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4781 show_dirstat(options);
4782 if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4784 for (i = 0; i < q->nr; i++) {
4785 diff_summary(options, q->queue[i]);
4786 }
4787 separator++;
4788 }
4789 if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4791 DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4792 DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4793 /*
4794 * run diff_flush_patch for the exit status. setting
4795 * options->file to /dev/null should be safe, because we
4796 * aren't supposed to produce any output anyway.
4797 */
4798 if (options->close_file)
4799 fclose(options->file);
4800 options->file = xfopen("/dev/null", "w");
4801 options->close_file = 1;
4802 for (i = 0; i < q->nr; i++) {
4803 struct diff_filepair *p = q->queue[i];
4804 if (check_pair_status(p))
4805 diff_flush_patch(p, options);
4806 if (options->found_changes)
4807 break;
4808 }
4809 }
4810 if (output_format & DIFF_FORMAT_PATCH) {
4812 if (separator) {
4813 fprintf(options->file, "%s%c",
4814 diff_line_prefix(options),
4815 options->line_termination);
4816 if (options->stat_sep) {
4817 /* attach patch instead of inline */
4818 fputs(options->stat_sep, options->file);
4819 }
4820 }
4821 for (i = 0; i < q->nr; i++) {
4823 struct diff_filepair *p = q->queue[i];
4824 if (check_pair_status(p))
4825 diff_flush_patch(p, options);
4826 }
4827 }
4828 if (output_format & DIFF_FORMAT_CALLBACK)
4830 options->format_callback(q, options, options->format_callback_data);
4831 for (i = 0; i < q->nr; i++)
4833 diff_free_filepair(q->queue[i]);
4834free_queue:
4835 free(q->queue);
4836 DIFF_QUEUE_CLEAR(q);
4837 if (options->close_file)
4838 fclose(options->file);
4839 /*
4841 * Report the content-level differences with HAS_CHANGES;
4842 * diff_addremove/diff_change does not set the bit when
4843 * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4844 */
4845 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4846 if (options->found_changes)
4847 DIFF_OPT_SET(options, HAS_CHANGES);
4848 else
4849 DIFF_OPT_CLR(options, HAS_CHANGES);
4850 }
4851}
4852static int match_filter(const struct diff_options *options, const struct diff_filepair *p)
4854{
4855 return (((p->status == DIFF_STATUS_MODIFIED) &&
4856 ((p->score &&
4857 filter_bit_tst(DIFF_STATUS_FILTER_BROKEN, options)) ||
4858 (!p->score &&
4859 filter_bit_tst(DIFF_STATUS_MODIFIED, options)))) ||
4860 ((p->status != DIFF_STATUS_MODIFIED) &&
4861 filter_bit_tst(p->status, options)));
4862}
4863static void diffcore_apply_filter(struct diff_options *options)
4865{
4866 int i;
4867 struct diff_queue_struct *q = &diff_queued_diff;
4868 struct diff_queue_struct outq;
4869 DIFF_QUEUE_CLEAR(&outq);
4871 if (!options->filter)
4873 return;
4874 if (filter_bit_tst(DIFF_STATUS_FILTER_AON, options)) {
4876 int found;
4877 for (i = found = 0; !found && i < q->nr; i++) {
4878 if (match_filter(options, q->queue[i]))
4879 found++;
4880 }
4881 if (found)
4882 return;
4883 /* otherwise we will clear the whole queue
4885 * by copying the empty outq at the end of this
4886 * function, but first clear the current entries
4887 * in the queue.
4888 */
4889 for (i = 0; i < q->nr; i++)
4890 diff_free_filepair(q->queue[i]);
4891 }
4892 else {
4893 /* Only the matching ones */
4894 for (i = 0; i < q->nr; i++) {
4895 struct diff_filepair *p = q->queue[i];
4896 if (match_filter(options, p))
4897 diff_q(&outq, p);
4898 else
4899 diff_free_filepair(p);
4900 }
4901 }
4902 free(q->queue);
4903 *q = outq;
4904}
4905/* Check whether two filespecs with the same mode and size are identical */
4907static int diff_filespec_is_identical(struct diff_filespec *one,
4908 struct diff_filespec *two)
4909{
4910 if (S_ISGITLINK(one->mode))
4911 return 0;
4912 if (diff_populate_filespec(one, 0))
4913 return 0;
4914 if (diff_populate_filespec(two, 0))
4915 return 0;
4916 return !memcmp(one->data, two->data, one->size);
4917}
4918static int diff_filespec_check_stat_unmatch(struct diff_filepair *p)
4920{
4921 if (p->done_skip_stat_unmatch)
4922 return p->skip_stat_unmatch_result;
4923 p->done_skip_stat_unmatch = 1;
4925 p->skip_stat_unmatch_result = 0;
4926 /*
4927 * 1. Entries that come from stat info dirtiness
4928 * always have both sides (iow, not create/delete),
4929 * one side of the object name is unknown, with
4930 * the same mode and size. Keep the ones that
4931 * do not match these criteria. They have real
4932 * differences.
4933 *
4934 * 2. At this point, the file is known to be modified,
4935 * with the same mode and size, and the object
4936 * name of one side is unknown. Need to inspect
4937 * the identical contents.
4938 */
4939 if (!DIFF_FILE_VALID(p->one) || /* (1) */
4940 !DIFF_FILE_VALID(p->two) ||
4941 (p->one->oid_valid && p->two->oid_valid) ||
4942 (p->one->mode != p->two->mode) ||
4943 diff_populate_filespec(p->one, CHECK_SIZE_ONLY) ||
4944 diff_populate_filespec(p->two, CHECK_SIZE_ONLY) ||
4945 (p->one->size != p->two->size) ||
4946 !diff_filespec_is_identical(p->one, p->two)) /* (2) */
4947 p->skip_stat_unmatch_result = 1;
4948 return p->skip_stat_unmatch_result;
4949}
4950static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
4952{
4953 int i;
4954 struct diff_queue_struct *q = &diff_queued_diff;
4955 struct diff_queue_struct outq;
4956 DIFF_QUEUE_CLEAR(&outq);
4957 for (i = 0; i < q->nr; i++) {
4959 struct diff_filepair *p = q->queue[i];
4960 if (diff_filespec_check_stat_unmatch(p))
4962 diff_q(&outq, p);
4963 else {
4964 /*
4965 * The caller can subtract 1 from skip_stat_unmatch
4966 * to determine how many paths were dirty only
4967 * due to stat info mismatch.
4968 */
4969 if (!DIFF_OPT_TST(diffopt, NO_INDEX))
4970 diffopt->skip_stat_unmatch++;
4971 diff_free_filepair(p);
4972 }
4973 }
4974 free(q->queue);
4975 *q = outq;
4976}
4977static int diffnamecmp(const void *a_, const void *b_)
4979{
4980 const struct diff_filepair *a = *((const struct diff_filepair **)a_);
4981 const struct diff_filepair *b = *((const struct diff_filepair **)b_);
4982 const char *name_a, *name_b;
4983 name_a = a->one ? a->one->path : a->two->path;
4985 name_b = b->one ? b->one->path : b->two->path;
4986 return strcmp(name_a, name_b);
4987}
4988void diffcore_fix_diff_index(struct diff_options *options)
4990{
4991 struct diff_queue_struct *q = &diff_queued_diff;
4992 QSORT(q->queue, q->nr, diffnamecmp);
4993}
4994void diffcore_std(struct diff_options *options)
4996{
4997 /* NOTE please keep the following in sync with diff_tree_combined() */
4998 if (options->skip_stat_unmatch)
4999 diffcore_skip_stat_unmatch(options);
5000 if (!options->found_follow) {
5001 /* See try_to_follow_renames() in tree-diff.c */
5002 if (options->break_opt != -1)
5003 diffcore_break(options->break_opt);
5004 if (options->detect_rename)
5005 diffcore_rename(options);
5006 if (options->break_opt != -1)
5007 diffcore_merge_broken();
5008 }
5009 if (options->pickaxe)
5010 diffcore_pickaxe(options);
5011 if (options->orderfile)
5012 diffcore_order(options->orderfile);
5013 if (!options->found_follow)
5014 /* See try_to_follow_renames() in tree-diff.c */
5015 diff_resolve_rename_copy();
5016 diffcore_apply_filter(options);
5017 if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5019 DIFF_OPT_SET(options, HAS_CHANGES);
5020 else
5021 DIFF_OPT_CLR(options, HAS_CHANGES);
5022 options->found_follow = 0;
5024}
5025int diff_result_code(struct diff_options *opt, int status)
5027{
5028 int result = 0;
5029 diff_warn_rename_limit("diff.renameLimit",
5031 opt->needed_rename_limit,
5032 opt->degraded_cc_to_c);
5033 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5034 !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
5035 return status;
5036 if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5037 DIFF_OPT_TST(opt, HAS_CHANGES))
5038 result |= 01;
5039 if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
5040 DIFF_OPT_TST(opt, CHECK_FAILED))
5041 result |= 02;
5042 return result;
5043}
5044int diff_can_quit_early(struct diff_options *opt)
5046{
5047 return (DIFF_OPT_TST(opt, QUICK) &&
5048 !opt->filter &&
5049 DIFF_OPT_TST(opt, HAS_CHANGES));
5050}
5051/*
5053 * Shall changes to this submodule be ignored?
5054 *
5055 * Submodule changes can be configured to be ignored separately for each path,
5056 * but that configuration can be overridden from the command line.
5057 */
5058static int is_submodule_ignored(const char *path, struct diff_options *options)
5059{
5060 int ignored = 0;
5061 unsigned orig_flags = options->flags;
5062 if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
5063 set_diffopt_flags_from_submodule_config(options, path);
5064 if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
5065 ignored = 1;
5066 options->flags = orig_flags;
5067 return ignored;
5068}
5069void diff_addremove(struct diff_options *options,
5071 int addremove, unsigned mode,
5072 const struct object_id *oid,
5073 int oid_valid,
5074 const char *concatpath, unsigned dirty_submodule)
5075{
5076 struct diff_filespec *one, *two;
5077 if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
5079 return;
5080 /* This may look odd, but it is a preparation for
5082 * feeding "there are unchanged files which should
5083 * not produce diffs, but when you are doing copy
5084 * detection you would need them, so here they are"
5085 * entries to the diff-core. They will be prefixed
5086 * with something like '=' or '*' (I haven't decided
5087 * which but should not make any difference).
5088 * Feeding the same new and old to diff_change()
5089 * also has the same effect.
5090 * Before the final output happens, they are pruned after
5091 * merged into rename/copy pairs as appropriate.
5092 */
5093 if (DIFF_OPT_TST(options, REVERSE_DIFF))
5094 addremove = (addremove == '+' ? '-' :
5095 addremove == '-' ? '+' : addremove);
5096 if (options->prefix &&
5098 strncmp(concatpath, options->prefix, options->prefix_length))
5099 return;
5100 one = alloc_filespec(concatpath);
5102 two = alloc_filespec(concatpath);
5103 if (addremove != '+')
5105 fill_filespec(one, oid, oid_valid, mode);
5106 if (addremove != '-') {
5107 fill_filespec(two, oid, oid_valid, mode);
5108 two->dirty_submodule = dirty_submodule;
5109 }
5110 diff_queue(&diff_queued_diff, one, two);
5112 if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5113 DIFF_OPT_SET(options, HAS_CHANGES);
5114}
5115void diff_change(struct diff_options *options,
5117 unsigned old_mode, unsigned new_mode,
5118 const struct object_id *old_oid,
5119 const struct object_id *new_oid,
5120 int old_oid_valid, int new_oid_valid,
5121 const char *concatpath,
5122 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
5123{
5124 struct diff_filespec *one, *two;
5125 struct diff_filepair *p;
5126 if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
5128 is_submodule_ignored(concatpath, options))
5129 return;
5130 if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
5132 SWAP(old_mode, new_mode);
5133 SWAP(old_oid, new_oid);
5134 SWAP(old_oid_valid, new_oid_valid);
5135 SWAP(old_dirty_submodule, new_dirty_submodule);
5136 }
5137 if (options->prefix &&
5139 strncmp(concatpath, options->prefix, options->prefix_length))
5140 return;
5141 one = alloc_filespec(concatpath);
5143 two = alloc_filespec(concatpath);
5144 fill_filespec(one, old_oid, old_oid_valid, old_mode);
5145 fill_filespec(two, new_oid, new_oid_valid, new_mode);
5146 one->dirty_submodule = old_dirty_submodule;
5147 two->dirty_submodule = new_dirty_submodule;
5148 p = diff_queue(&diff_queued_diff, one, two);
5149 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5151 return;
5152 if (DIFF_OPT_TST(options, QUICK) && options->skip_stat_unmatch &&
5154 !diff_filespec_check_stat_unmatch(p))
5155 return;
5156 DIFF_OPT_SET(options, HAS_CHANGES);
5158}
5159struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
5161{
5162 struct diff_filepair *pair;
5163 struct diff_filespec *one, *two;
5164 if (options->prefix &&
5166 strncmp(path, options->prefix, options->prefix_length))
5167 return NULL;
5168 one = alloc_filespec(path);
5170 two = alloc_filespec(path);
5171 pair = diff_queue(&diff_queued_diff, one, two);
5172 pair->is_unmerged = 1;
5173 return pair;
5174}
5175static char *run_textconv(const char *pgm, struct diff_filespec *spec,
5177 size_t *outsize)
5178{
5179 struct diff_tempfile *temp;
5180 const char *argv[3];
5181 const char **arg = argv;
5182 struct child_process child = CHILD_PROCESS_INIT;
5183 struct strbuf buf = STRBUF_INIT;
5184 int err = 0;
5185 temp = prepare_temp_file(spec->path, spec);
5187 *arg++ = pgm;
5188 *arg++ = temp->name;
5189 *arg = NULL;
5190 child.use_shell = 1;
5192 child.argv = argv;
5193 child.out = -1;
5194 if (start_command(&child)) {
5195 remove_tempfile();
5196 return NULL;
5197 }
5198 if (strbuf_read(&buf, child.out, 0) < 0)
5200 err = error("error reading from textconv command '%s'", pgm);
5201 close(child.out);
5202 if (finish_command(&child) || err) {
5204 strbuf_release(&buf);
5205 remove_tempfile();
5206 return NULL;
5207 }
5208 remove_tempfile();
5209 return strbuf_detach(&buf, outsize);
5211}
5212size_t fill_textconv(struct userdiff_driver *driver,
5214 struct diff_filespec *df,
5215 char **outbuf)
5216{
5217 size_t size;
5218 if (!driver) {
5220 if (!DIFF_FILE_VALID(df)) {
5221 *outbuf = "";
5222 return 0;
5223 }
5224 if (diff_populate_filespec(df, 0))
5225 die("unable to read files to diff");
5226 *outbuf = df->data;
5227 return df->size;
5228 }
5229 if (!driver->textconv)
5231 die("BUG: fill_textconv called with non-textconv driver");
5232 if (driver->textconv_cache && df->oid_valid) {
5234 *outbuf = notes_cache_get(driver->textconv_cache,
5235 &df->oid,
5236 &size);
5237 if (*outbuf)
5238 return size;
5239 }
5240 *outbuf = run_textconv(driver->textconv, df, &size);
5242 if (!*outbuf)
5243 die("unable to read files to diff");
5244 if (driver->textconv_cache && df->oid_valid) {
5246 /* ignore errors, as we might be in a readonly repository */
5247 notes_cache_put(driver->textconv_cache, &df->oid, *outbuf,
5248 size);
5249 /*
5250 * we could save up changes and flush them all at the end,
5251 * but we would need an extra call after all diffing is done.
5252 * Since generating a cache entry is the slow path anyway,
5253 * this extra overhead probably isn't a big deal.
5254 */
5255 notes_cache_write(driver->textconv_cache);
5256 }
5257 return size;
5259}
5260int textconv_object(const char *path,
5262 unsigned mode,
5263 const struct object_id *oid,
5264 int oid_valid,
5265 char **buf,
5266 unsigned long *buf_size)
5267{
5268 struct diff_filespec *df;
5269 struct userdiff_driver *textconv;
5270 df = alloc_filespec(path);
5272 fill_filespec(df, oid, oid_valid, mode);
5273 textconv = get_textconv(df);
5274 if (!textconv) {
5275 free_filespec(df);
5276 return 0;
5277 }
5278 *buf_size = fill_textconv(textconv, df, buf);
5280 free_filespec(df);
5281 return 1;
5282}
5283void setup_diff_pager(struct diff_options *opt)
5285{
5286 /*
5287 * If the user asked for our exit code, then either they want --quiet
5288 * or --exit-code. We should definitely not bother with a pager in the
5289 * former case, as we will generate no output. Since we still properly
5290 * report our exit code even when a pager is run, we _could_ run a
5291 * pager with --exit-code. But since we have not done so historically,
5292 * and because it is easy to find people oneline advising "git diff
5293 * --exit-code" in hooks and other scripts, we do not do so.
5294 */
5295 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5296 check_pager_config("diff") != 0)
5297 setup_pager();
5298}