75b996c4cf17d712d7430c24869208263d05f064
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#ifdef NO_FAST_WORKING_DIRECTORY
25#define FAST_WORKING_DIRECTORY 0
26#else
27#define FAST_WORKING_DIRECTORY 1
28#endif
29static int diff_detect_rename_default;
31static int diff_indent_heuristic = 1;
32static int diff_rename_limit_default = 400;
33static int diff_suppress_blank_empty;
34static int diff_use_color_default = -1;
35static int diff_context_default = 3;
36static int diff_interhunk_context_default;
37static const char *diff_word_regex_cfg;
38static const char *external_diff_cmd_cfg;
39static const char *diff_order_file_cfg;
40int diff_auto_refresh_index = 1;
41static int diff_mnemonic_prefix;
42static int diff_no_prefix;
43static int diff_stat_graph_width;
44static int diff_dirstat_permille_default = 30;
45static struct diff_options default_diff_options;
46static long diff_algorithm;
47static unsigned ws_error_highlight_default = WSEH_NEW;
48static char diff_colors[][COLOR_MAXLEN] = {
50 GIT_COLOR_RESET,
51 GIT_COLOR_NORMAL, /* CONTEXT */
52 GIT_COLOR_BOLD, /* METAINFO */
53 GIT_COLOR_CYAN, /* FRAGINFO */
54 GIT_COLOR_RED, /* OLD */
55 GIT_COLOR_GREEN, /* NEW */
56 GIT_COLOR_YELLOW, /* COMMIT */
57 GIT_COLOR_BG_RED, /* WHITESPACE */
58 GIT_COLOR_NORMAL, /* FUNCINFO */
59};
60static NORETURN void die_want_option(const char *option_name)
62{
63 die(_("option '%s' requires a value"), option_name);
64}
65static int parse_diff_color_slot(const char *var)
67{
68 if (!strcasecmp(var, "context") || !strcasecmp(var, "plain"))
69 return DIFF_CONTEXT;
70 if (!strcasecmp(var, "meta"))
71 return DIFF_METAINFO;
72 if (!strcasecmp(var, "frag"))
73 return DIFF_FRAGINFO;
74 if (!strcasecmp(var, "old"))
75 return DIFF_FILE_OLD;
76 if (!strcasecmp(var, "new"))
77 return DIFF_FILE_NEW;
78 if (!strcasecmp(var, "commit"))
79 return DIFF_COMMIT;
80 if (!strcasecmp(var, "whitespace"))
81 return DIFF_WHITESPACE;
82 if (!strcasecmp(var, "func"))
83 return DIFF_FUNCINFO;
84 return -1;
85}
86static int parse_dirstat_params(struct diff_options *options, const char *params_string,
88 struct strbuf *errmsg)
89{
90 char *params_copy = xstrdup(params_string);
91 struct string_list params = STRING_LIST_INIT_NODUP;
92 int ret = 0;
93 int i;
94 if (*params_copy)
96 string_list_split_in_place(¶ms, params_copy, ',', -1);
97 for (i = 0; i < params.nr; i++) {
98 const char *p = params.items[i].string;
99 if (!strcmp(p, "changes")) {
100 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
101 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
102 } else if (!strcmp(p, "lines")) {
103 DIFF_OPT_SET(options, DIRSTAT_BY_LINE);
104 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
105 } else if (!strcmp(p, "files")) {
106 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
107 DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
108 } else if (!strcmp(p, "noncumulative")) {
109 DIFF_OPT_CLR(options, DIRSTAT_CUMULATIVE);
110 } else if (!strcmp(p, "cumulative")) {
111 DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
112 } else if (isdigit(*p)) {
113 char *end;
114 int permille = strtoul(p, &end, 10) * 10;
115 if (*end == '.' && isdigit(*++end)) {
116 /* only use first digit */
117 permille += *end - '0';
118 /* .. and ignore any further digits */
119 while (isdigit(*++end))
120 ; /* nothing */
121 }
122 if (!*end)
123 options->dirstat_permille = permille;
124 else {
125 strbuf_addf(errmsg, _(" Failed to parse dirstat cut-off percentage '%s'\n"),
126 p);
127 ret++;
128 }
129 } else {
130 strbuf_addf(errmsg, _(" Unknown dirstat parameter '%s'\n"), p);
131 ret++;
132 }
133 }
135 string_list_clear(¶ms, 0);
136 free(params_copy);
137 return ret;
138}
139static int parse_submodule_params(struct diff_options *options, const char *value)
141{
142 if (!strcmp(value, "log"))
143 options->submodule_format = DIFF_SUBMODULE_LOG;
144 else if (!strcmp(value, "short"))
145 options->submodule_format = DIFF_SUBMODULE_SHORT;
146 else if (!strcmp(value, "diff"))
147 options->submodule_format = DIFF_SUBMODULE_INLINE_DIFF;
148 else
149 return -1;
150 return 0;
151}
152static int git_config_rename(const char *var, const char *value)
154{
155 if (!value)
156 return DIFF_DETECT_RENAME;
157 if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
158 return DIFF_DETECT_COPY;
159 return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
160}
161long parse_algorithm_value(const char *value)
163{
164 if (!value)
165 return -1;
166 else if (!strcasecmp(value, "myers") || !strcasecmp(value, "default"))
167 return 0;
168 else if (!strcasecmp(value, "minimal"))
169 return XDF_NEED_MINIMAL;
170 else if (!strcasecmp(value, "patience"))
171 return XDF_PATIENCE_DIFF;
172 else if (!strcasecmp(value, "histogram"))
173 return XDF_HISTOGRAM_DIFF;
174 return -1;
175}
176static int parse_one_token(const char **arg, const char *token)
178{
179 const char *rest;
180 if (skip_prefix(*arg, token, &rest) && (!*rest || *rest == ',')) {
181 *arg = rest;
182 return 1;
183 }
184 return 0;
185}
186static int parse_ws_error_highlight(const char *arg)
188{
189 const char *orig_arg = arg;
190 unsigned val = 0;
191 while (*arg) {
193 if (parse_one_token(&arg, "none"))
194 val = 0;
195 else if (parse_one_token(&arg, "default"))
196 val = WSEH_NEW;
197 else if (parse_one_token(&arg, "all"))
198 val = WSEH_NEW | WSEH_OLD | WSEH_CONTEXT;
199 else if (parse_one_token(&arg, "new"))
200 val |= WSEH_NEW;
201 else if (parse_one_token(&arg, "old"))
202 val |= WSEH_OLD;
203 else if (parse_one_token(&arg, "context"))
204 val |= WSEH_CONTEXT;
205 else {
206 return -1 - (int)(arg - orig_arg);
207 }
208 if (*arg)
209 arg++;
210 }
211 return val;
212}
213/*
215 * These are to give UI layer defaults.
216 * The core-level commands such as git-diff-files should
217 * never be affected by the setting of diff.renames
218 * the user happens to have in the configuration file.
219 */
220void init_diff_ui_defaults(void)
221{
222 diff_detect_rename_default = 1;
223}
224int git_diff_heuristic_config(const char *var, const char *value, void *cb)
226{
227 if (!strcmp(var, "diff.indentheuristic"))
228 diff_indent_heuristic = git_config_bool(var, value);
229 return 0;
230}
231int git_diff_ui_config(const char *var, const char *value, void *cb)
233{
234 if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
235 diff_use_color_default = git_config_colorbool(var, value);
236 return 0;
237 }
238 if (!strcmp(var, "diff.context")) {
239 diff_context_default = git_config_int(var, value);
240 if (diff_context_default < 0)
241 return -1;
242 return 0;
243 }
244 if (!strcmp(var, "diff.interhunkcontext")) {
245 diff_interhunk_context_default = git_config_int(var, value);
246 if (diff_interhunk_context_default < 0)
247 return -1;
248 return 0;
249 }
250 if (!strcmp(var, "diff.renames")) {
251 diff_detect_rename_default = git_config_rename(var, value);
252 return 0;
253 }
254 if (!strcmp(var, "diff.autorefreshindex")) {
255 diff_auto_refresh_index = git_config_bool(var, value);
256 return 0;
257 }
258 if (!strcmp(var, "diff.mnemonicprefix")) {
259 diff_mnemonic_prefix = git_config_bool(var, value);
260 return 0;
261 }
262 if (!strcmp(var, "diff.noprefix")) {
263 diff_no_prefix = git_config_bool(var, value);
264 return 0;
265 }
266 if (!strcmp(var, "diff.statgraphwidth")) {
267 diff_stat_graph_width = git_config_int(var, value);
268 return 0;
269 }
270 if (!strcmp(var, "diff.external"))
271 return git_config_string(&external_diff_cmd_cfg, var, value);
272 if (!strcmp(var, "diff.wordregex"))
273 return git_config_string(&diff_word_regex_cfg, var, value);
274 if (!strcmp(var, "diff.orderfile"))
275 return git_config_pathname(&diff_order_file_cfg, var, value);
276 if (!strcmp(var, "diff.ignoresubmodules"))
278 handle_ignore_submodules_arg(&default_diff_options, value);
279 if (!strcmp(var, "diff.submodule")) {
281 if (parse_submodule_params(&default_diff_options, value))
282 warning(_("Unknown value for 'diff.submodule' config variable: '%s'"),
283 value);
284 return 0;
285 }
286 if (!strcmp(var, "diff.algorithm")) {
288 diff_algorithm = parse_algorithm_value(value);
289 if (diff_algorithm < 0)
290 return -1;
291 return 0;
292 }
293 if (!strcmp(var, "diff.wserrorhighlight")) {
295 int val = parse_ws_error_highlight(value);
296 if (val < 0)
297 return -1;
298 ws_error_highlight_default = val;
299 return 0;
300 }
301 if (git_color_config(var, value, cb) < 0)
303 return -1;
304 return git_diff_basic_config(var, value, cb);
306}
307int git_diff_basic_config(const char *var, const char *value, void *cb)
309{
310 const char *name;
311 if (!strcmp(var, "diff.renamelimit")) {
313 diff_rename_limit_default = git_config_int(var, value);
314 return 0;
315 }
316 if (userdiff_config(var, value) < 0)
318 return -1;
319 if (skip_prefix(var, "diff.color.", &name) ||
321 skip_prefix(var, "color.diff.", &name)) {
322 int slot = parse_diff_color_slot(name);
323 if (slot < 0)
324 return 0;
325 if (!value)
326 return config_error_nonbool(var);
327 return color_parse(value, diff_colors[slot]);
328 }
329 /* like GNU diff's --suppress-blank-empty option */
331 if (!strcmp(var, "diff.suppressblankempty") ||
332 /* for backwards compatibility */
333 !strcmp(var, "diff.suppress-blank-empty")) {
334 diff_suppress_blank_empty = git_config_bool(var, value);
335 return 0;
336 }
337 if (!strcmp(var, "diff.dirstat")) {
339 struct strbuf errmsg = STRBUF_INIT;
340 default_diff_options.dirstat_permille = diff_dirstat_permille_default;
341 if (parse_dirstat_params(&default_diff_options, value, &errmsg))
342 warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
343 errmsg.buf);
344 strbuf_release(&errmsg);
345 diff_dirstat_permille_default = default_diff_options.dirstat_permille;
346 return 0;
347 }
348 if (starts_with(var, "submodule."))
350 return parse_submodule_config_option(var, value);
351 if (git_diff_heuristic_config(var, value, cb) < 0)
353 return -1;
354 return git_default_config(var, value, cb);
356}
357static char *quote_two(const char *one, const char *two)
359{
360 int need_one = quote_c_style(one, NULL, NULL, 1);
361 int need_two = quote_c_style(two, NULL, NULL, 1);
362 struct strbuf res = STRBUF_INIT;
363 if (need_one + need_two) {
365 strbuf_addch(&res, '"');
366 quote_c_style(one, &res, NULL, 1);
367 quote_c_style(two, &res, NULL, 1);
368 strbuf_addch(&res, '"');
369 } else {
370 strbuf_addstr(&res, one);
371 strbuf_addstr(&res, two);
372 }
373 return strbuf_detach(&res, NULL);
374}
375static const char *external_diff(void)
377{
378 static const char *external_diff_cmd = NULL;
379 static int done_preparing = 0;
380 if (done_preparing)
382 return external_diff_cmd;
383 external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
384 if (!external_diff_cmd)
385 external_diff_cmd = external_diff_cmd_cfg;
386 done_preparing = 1;
387 return external_diff_cmd;
388}
389/*
391 * Keep track of files used for diffing. Sometimes such an entry
392 * refers to a temporary file, sometimes to an existing file, and
393 * sometimes to "/dev/null".
394 */
395static struct diff_tempfile {
396 /*
397 * filename external diff should read from, or NULL if this
398 * entry is currently not in use:
399 */
400 const char *name;
401 char hex[GIT_MAX_HEXSZ + 1];
403 char mode[10];
404 /*
406 * If this diff_tempfile instance refers to a temporary file,
407 * this tempfile object is used to manage its lifetime.
408 */
409 struct tempfile tempfile;
410} diff_temp[2];
411typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
413struct emit_callback {
415 int color_diff;
416 unsigned ws_rule;
417 int blank_at_eof_in_preimage;
418 int blank_at_eof_in_postimage;
419 int lno_in_preimage;
420 int lno_in_postimage;
421 sane_truncate_fn truncate;
422 const char **label_path;
423 struct diff_words_data *diff_words;
424 struct diff_options *opt;
425 struct strbuf *header;
426};
427static int count_lines(const char *data, int size)
429{
430 int count, ch, completely_empty = 1, nl_just_seen = 0;
431 count = 0;
432 while (0 < size--) {
433 ch = *data++;
434 if (ch == '\n') {
435 count++;
436 nl_just_seen = 1;
437 completely_empty = 0;
438 }
439 else {
440 nl_just_seen = 0;
441 completely_empty = 0;
442 }
443 }
444 if (completely_empty)
445 return 0;
446 if (!nl_just_seen)
447 count++; /* no trailing newline */
448 return count;
449}
450static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
452{
453 if (!DIFF_FILE_VALID(one)) {
454 mf->ptr = (char *)""; /* does not matter */
455 mf->size = 0;
456 return 0;
457 }
458 else if (diff_populate_filespec(one, 0))
459 return -1;
460 mf->ptr = one->data;
462 mf->size = one->size;
463 return 0;
464}
465/* like fill_mmfile, but only for size, so we can avoid retrieving blob */
467static unsigned long diff_filespec_size(struct diff_filespec *one)
468{
469 if (!DIFF_FILE_VALID(one))
470 return 0;
471 diff_populate_filespec(one, CHECK_SIZE_ONLY);
472 return one->size;
473}
474static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
476{
477 char *ptr = mf->ptr;
478 long size = mf->size;
479 int cnt = 0;
480 if (!size)
482 return cnt;
483 ptr += size - 1; /* pointing at the very end */
484 if (*ptr != '\n')
485 ; /* incomplete line */
486 else
487 ptr--; /* skip the last LF */
488 while (mf->ptr < ptr) {
489 char *prev_eol;
490 for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
491 if (*prev_eol == '\n')
492 break;
493 if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
494 break;
495 cnt++;
496 ptr = prev_eol - 1;
497 }
498 return cnt;
499}
500static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
502 struct emit_callback *ecbdata)
503{
504 int l1, l2, at;
505 unsigned ws_rule = ecbdata->ws_rule;
506 l1 = count_trailing_blank(mf1, ws_rule);
507 l2 = count_trailing_blank(mf2, ws_rule);
508 if (l2 <= l1) {
509 ecbdata->blank_at_eof_in_preimage = 0;
510 ecbdata->blank_at_eof_in_postimage = 0;
511 return;
512 }
513 at = count_lines(mf1->ptr, mf1->size);
514 ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
515 at = count_lines(mf2->ptr, mf2->size);
517 ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
518}
519static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
521 int first, const char *line, int len)
522{
523 int has_trailing_newline, has_trailing_carriage_return;
524 int nofirst;
525 FILE *file = o->file;
526 fputs(diff_line_prefix(o), file);
528 if (len == 0) {
530 has_trailing_newline = (first == '\n');
531 has_trailing_carriage_return = (!has_trailing_newline &&
532 (first == '\r'));
533 nofirst = has_trailing_newline || has_trailing_carriage_return;
534 } else {
535 has_trailing_newline = (len > 0 && line[len-1] == '\n');
536 if (has_trailing_newline)
537 len--;
538 has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
539 if (has_trailing_carriage_return)
540 len--;
541 nofirst = 0;
542 }
543 if (len || !nofirst) {
545 fputs(set, file);
546 if (!nofirst)
547 fputc(first, file);
548 fwrite(line, len, 1, file);
549 fputs(reset, file);
550 }
551 if (has_trailing_carriage_return)
552 fputc('\r', file);
553 if (has_trailing_newline)
554 fputc('\n', file);
555}
556static void emit_line(struct diff_options *o, const char *set, const char *reset,
558 const char *line, int len)
559{
560 emit_line_0(o, set, reset, line[0], line+1, len-1);
561}
562enum diff_symbol {
564 DIFF_SYMBOL_CONTEXT_MARKER,
565 DIFF_SYMBOL_SEPARATOR
566};
567static void emit_diff_symbol(struct diff_options *o, enum diff_symbol s,
569 const char *line, int len)
570{
571 const char *context, *reset;
572 switch (s) {
573 case DIFF_SYMBOL_CONTEXT_MARKER:
574 context = diff_get_color_opt(o, DIFF_CONTEXT);
575 reset = diff_get_color_opt(o, DIFF_RESET);
576 emit_line(o, context, reset, line, len);
577 break;
578 case DIFF_SYMBOL_SEPARATOR:
579 fprintf(o->file, "%s%c",
580 diff_line_prefix(o),
581 o->line_termination);
582 break;
583 default:
584 die("BUG: unknown diff symbol");
585 }
586}
587static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
589{
590 if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
591 ecbdata->blank_at_eof_in_preimage &&
592 ecbdata->blank_at_eof_in_postimage &&
593 ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
594 ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
595 return 0;
596 return ws_blank_line(line, len, ecbdata->ws_rule);
597}
598static void emit_line_checked(const char *reset,
600 struct emit_callback *ecbdata,
601 const char *line, int len,
602 enum color_diff color,
603 unsigned ws_error_highlight,
604 char sign)
605{
606 const char *set = diff_get_color(ecbdata->color_diff, color);
607 const char *ws = NULL;
608 if (ecbdata->opt->ws_error_highlight & ws_error_highlight) {
610 ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
611 if (!*ws)
612 ws = NULL;
613 }
614 if (!ws)
616 emit_line_0(ecbdata->opt, set, reset, sign, line, len);
617 else if (sign == '+' && new_blank_line_at_eof(ecbdata, line, len))
618 /* Blank line at EOF - paint '+' as well */
619 emit_line_0(ecbdata->opt, ws, reset, sign, line, len);
620 else {
621 /* Emit just the prefix, then the rest. */
622 emit_line_0(ecbdata->opt, set, reset, sign, "", 0);
623 ws_check_emit(line, len, ecbdata->ws_rule,
624 ecbdata->opt->file, set, reset, ws);
625 }
626}
627static void emit_add_line(const char *reset,
629 struct emit_callback *ecbdata,
630 const char *line, int len)
631{
632 emit_line_checked(reset, ecbdata, line, len,
633 DIFF_FILE_NEW, WSEH_NEW, '+');
634}
635static void emit_del_line(const char *reset,
637 struct emit_callback *ecbdata,
638 const char *line, int len)
639{
640 emit_line_checked(reset, ecbdata, line, len,
641 DIFF_FILE_OLD, WSEH_OLD, '-');
642}
643static void emit_context_line(const char *reset,
645 struct emit_callback *ecbdata,
646 const char *line, int len)
647{
648 emit_line_checked(reset, ecbdata, line, len,
649 DIFF_CONTEXT, WSEH_CONTEXT, ' ');
650}
651static void emit_hunk_header(struct emit_callback *ecbdata,
653 const char *line, int len)
654{
655 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
656 const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
657 const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
658 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
659 static const char atat[2] = { '@', '@' };
660 const char *cp, *ep;
661 struct strbuf msgbuf = STRBUF_INIT;
662 int org_len = len;
663 int i = 1;
664 /*
666 * As a hunk header must begin with "@@ -<old>, +<new> @@",
667 * it always is at least 10 bytes long.
668 */
669 if (len < 10 ||
670 memcmp(line, atat, 2) ||
671 !(ep = memmem(line + 2, len - 2, atat, 2))) {
672 emit_diff_symbol(ecbdata->opt,
673 DIFF_SYMBOL_CONTEXT_MARKER, line, len);
674 return;
675 }
676 ep += 2; /* skip over @@ */
677 /* The hunk header in fraginfo color */
679 strbuf_addstr(&msgbuf, frag);
680 strbuf_add(&msgbuf, line, ep - line);
681 strbuf_addstr(&msgbuf, reset);
682 /*
684 * trailing "\r\n"
685 */
686 for ( ; i < 3; i++)
687 if (line[len - i] == '\r' || line[len - i] == '\n')
688 len--;
689 /* blank before the func header */
691 for (cp = ep; ep - line < len; ep++)
692 if (*ep != ' ' && *ep != '\t')
693 break;
694 if (ep != cp) {
695 strbuf_addstr(&msgbuf, context);
696 strbuf_add(&msgbuf, cp, ep - cp);
697 strbuf_addstr(&msgbuf, reset);
698 }
699 if (ep < line + len) {
701 strbuf_addstr(&msgbuf, func);
702 strbuf_add(&msgbuf, ep, line + len - ep);
703 strbuf_addstr(&msgbuf, reset);
704 }
705 strbuf_add(&msgbuf, line + len, org_len - len);
707 strbuf_complete_line(&msgbuf);
708 emit_line(ecbdata->opt, "", "", msgbuf.buf, msgbuf.len);
710 strbuf_release(&msgbuf);
711}
712static struct diff_tempfile *claim_diff_tempfile(void) {
714 int i;
715 for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
716 if (!diff_temp[i].name)
717 return diff_temp + i;
718 die("BUG: diff is failing to clean up its tempfiles");
719}
720static void remove_tempfile(void)
722{
723 int i;
724 for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
725 if (is_tempfile_active(&diff_temp[i].tempfile))
726 delete_tempfile(&diff_temp[i].tempfile);
727 diff_temp[i].name = NULL;
728 }
729}
730static void print_line_count(FILE *file, int count)
732{
733 switch (count) {
734 case 0:
735 fprintf(file, "0,0");
736 break;
737 case 1:
738 fprintf(file, "1");
739 break;
740 default:
741 fprintf(file, "1,%d", count);
742 break;
743 }
744}
745static void emit_rewrite_lines(struct emit_callback *ecb,
747 int prefix, const char *data, int size)
748{
749 const char *endp = NULL;
750 static const char *nneof = " No newline at end of file\n";
751 const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
752 while (0 < size) {
754 int len;
755 endp = memchr(data, '\n', size);
757 len = endp ? (endp - data + 1) : size;
758 if (prefix != '+') {
759 ecb->lno_in_preimage++;
760 emit_del_line(reset, ecb, data, len);
761 } else {
762 ecb->lno_in_postimage++;
763 emit_add_line(reset, ecb, data, len);
764 }
765 size -= len;
766 data += len;
767 }
768 if (!endp) {
769 const char *context = diff_get_color(ecb->color_diff,
770 DIFF_CONTEXT);
771 putc('\n', ecb->opt->file);
772 emit_line_0(ecb->opt, context, reset, '\\',
773 nneof, strlen(nneof));
774 }
775}
776static void emit_rewrite_diff(const char *name_a,
778 const char *name_b,
779 struct diff_filespec *one,
780 struct diff_filespec *two,
781 struct userdiff_driver *textconv_one,
782 struct userdiff_driver *textconv_two,
783 struct diff_options *o)
784{
785 int lc_a, lc_b;
786 const char *name_a_tab, *name_b_tab;
787 const char *metainfo = diff_get_color(o->use_color, DIFF_METAINFO);
788 const char *fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
789 const char *reset = diff_get_color(o->use_color, DIFF_RESET);
790 static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
791 const char *a_prefix, *b_prefix;
792 char *data_one, *data_two;
793 size_t size_one, size_two;
794 struct emit_callback ecbdata;
795 const char *line_prefix = diff_line_prefix(o);
796 if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
798 a_prefix = o->b_prefix;
799 b_prefix = o->a_prefix;
800 } else {
801 a_prefix = o->a_prefix;
802 b_prefix = o->b_prefix;
803 }
804 name_a += (*name_a == '/');
806 name_b += (*name_b == '/');
807 name_a_tab = strchr(name_a, ' ') ? "\t" : "";
808 name_b_tab = strchr(name_b, ' ') ? "\t" : "";
809 strbuf_reset(&a_name);
811 strbuf_reset(&b_name);
812 quote_two_c_style(&a_name, a_prefix, name_a, 0);
813 quote_two_c_style(&b_name, b_prefix, name_b, 0);
814 size_one = fill_textconv(textconv_one, one, &data_one);
816 size_two = fill_textconv(textconv_two, two, &data_two);
817 memset(&ecbdata, 0, sizeof(ecbdata));
819 ecbdata.color_diff = want_color(o->use_color);
820 ecbdata.ws_rule = whitespace_rule(name_b);
821 ecbdata.opt = o;
822 if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
823 mmfile_t mf1, mf2;
824 mf1.ptr = (char *)data_one;
825 mf2.ptr = (char *)data_two;
826 mf1.size = size_one;
827 mf2.size = size_two;
828 check_blank_at_eof(&mf1, &mf2, &ecbdata);
829 }
830 ecbdata.lno_in_preimage = 1;
831 ecbdata.lno_in_postimage = 1;
832 lc_a = count_lines(data_one, size_one);
834 lc_b = count_lines(data_two, size_two);
835 fprintf(o->file,
836 "%s%s--- %s%s%s\n%s%s+++ %s%s%s\n%s%s@@ -",
837 line_prefix, metainfo, a_name.buf, name_a_tab, reset,
838 line_prefix, metainfo, b_name.buf, name_b_tab, reset,
839 line_prefix, fraginfo);
840 if (!o->irreversible_delete)
841 print_line_count(o->file, lc_a);
842 else
843 fprintf(o->file, "?,?");
844 fprintf(o->file, " +");
845 print_line_count(o->file, lc_b);
846 fprintf(o->file, " @@%s\n", reset);
847 if (lc_a && !o->irreversible_delete)
848 emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
849 if (lc_b)
850 emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
851 if (textconv_one)
852 free((char *)data_one);
853 if (textconv_two)
854 free((char *)data_two);
855}
856struct diff_words_buffer {
858 mmfile_t text;
859 long alloc;
860 struct diff_words_orig {
861 const char *begin, *end;
862 } *orig;
863 int orig_nr, orig_alloc;
864};
865static void diff_words_append(char *line, unsigned long len,
867 struct diff_words_buffer *buffer)
868{
869 ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
870 line++;
871 len--;
872 memcpy(buffer->text.ptr + buffer->text.size, line, len);
873 buffer->text.size += len;
874 buffer->text.ptr[buffer->text.size] = '\0';
875}
876struct diff_words_style_elem {
878 const char *prefix;
879 const char *suffix;
880 const char *color; /* NULL; filled in by the setup code if
881 * color is enabled */
882};
883struct diff_words_style {
885 enum diff_words_type type;
886 struct diff_words_style_elem new, old, ctx;
887 const char *newline;
888};
889static struct diff_words_style diff_words_styles[] = {
891 { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
892 { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
893 { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
894};
895struct diff_words_data {
897 struct diff_words_buffer minus, plus;
898 const char *current_plus;
899 int last_minus;
900 struct diff_options *opt;
901 regex_t *word_regex;
902 enum diff_words_type type;
903 struct diff_words_style *style;
904};
905static int fn_out_diff_words_write_helper(FILE *fp,
907 struct diff_words_style_elem *st_el,
908 const char *newline,
909 size_t count, const char *buf,
910 const char *line_prefix)
911{
912 int print = 0;
913 while (count) {
915 char *p = memchr(buf, '\n', count);
916 if (print)
917 fputs(line_prefix, fp);
918 if (p != buf) {
919 if (st_el->color && fputs(st_el->color, fp) < 0)
920 return -1;
921 if (fputs(st_el->prefix, fp) < 0 ||
922 fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
923 fputs(st_el->suffix, fp) < 0)
924 return -1;
925 if (st_el->color && *st_el->color
926 && fputs(GIT_COLOR_RESET, fp) < 0)
927 return -1;
928 }
929 if (!p)
930 return 0;
931 if (fputs(newline, fp) < 0)
932 return -1;
933 count -= p + 1 - buf;
934 buf = p + 1;
935 print = 1;
936 }
937 return 0;
938}
939/*
941 * '--color-words' algorithm can be described as:
942 *
943 * 1. collect the minus/plus lines of a diff hunk, divided into
944 * minus-lines and plus-lines;
945 *
946 * 2. break both minus-lines and plus-lines into words and
947 * place them into two mmfile_t with one word for each line;
948 *
949 * 3. use xdiff to run diff on the two mmfile_t to get the words level diff;
950 *
951 * And for the common parts of the both file, we output the plus side text.
952 * diff_words->current_plus is used to trace the current position of the plus file
953 * which printed. diff_words->last_minus is used to trace the last minus word
954 * printed.
955 *
956 * For '--graph' to work with '--color-words', we need to output the graph prefix
957 * on each line of color words output. Generally, there are two conditions on
958 * which we should output the prefix.
959 *
960 * 1. diff_words->last_minus == 0 &&
961 * diff_words->current_plus == diff_words->plus.text.ptr
962 *
963 * that is: the plus text must start as a new line, and if there is no minus
964 * word printed, a graph prefix must be printed.
965 *
966 * 2. diff_words->current_plus > diff_words->plus.text.ptr &&
967 * *(diff_words->current_plus - 1) == '\n'
968 *
969 * that is: a graph prefix must be printed following a '\n'
970 */
971static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
972{
973 if ((diff_words->last_minus == 0 &&
974 diff_words->current_plus == diff_words->plus.text.ptr) ||
975 (diff_words->current_plus > diff_words->plus.text.ptr &&
976 *(diff_words->current_plus - 1) == '\n')) {
977 return 1;
978 } else {
979 return 0;
980 }
981}
982static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
984{
985 struct diff_words_data *diff_words = priv;
986 struct diff_words_style *style = diff_words->style;
987 int minus_first, minus_len, plus_first, plus_len;
988 const char *minus_begin, *minus_end, *plus_begin, *plus_end;
989 struct diff_options *opt = diff_words->opt;
990 const char *line_prefix;
991 if (line[0] != '@' || parse_hunk_header(line, len,
993 &minus_first, &minus_len, &plus_first, &plus_len))
994 return;
995 assert(opt);
997 line_prefix = diff_line_prefix(opt);
998 /* POSIX requires that first be decremented by one if len == 0... */
1000 if (minus_len) {
1001 minus_begin = diff_words->minus.orig[minus_first].begin;
1002 minus_end =
1003 diff_words->minus.orig[minus_first + minus_len - 1].end;
1004 } else
1005 minus_begin = minus_end =
1006 diff_words->minus.orig[minus_first].end;
1007 if (plus_len) {
1009 plus_begin = diff_words->plus.orig[plus_first].begin;
1010 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
1011 } else
1012 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
1013 if (color_words_output_graph_prefix(diff_words)) {
1015 fputs(line_prefix, diff_words->opt->file);
1016 }
1017 if (diff_words->current_plus != plus_begin) {
1018 fn_out_diff_words_write_helper(diff_words->opt->file,
1019 &style->ctx, style->newline,
1020 plus_begin - diff_words->current_plus,
1021 diff_words->current_plus, line_prefix);
1022 if (*(plus_begin - 1) == '\n')
1023 fputs(line_prefix, diff_words->opt->file);
1024 }
1025 if (minus_begin != minus_end) {
1026 fn_out_diff_words_write_helper(diff_words->opt->file,
1027 &style->old, style->newline,
1028 minus_end - minus_begin, minus_begin,
1029 line_prefix);
1030 }
1031 if (plus_begin != plus_end) {
1032 fn_out_diff_words_write_helper(diff_words->opt->file,
1033 &style->new, style->newline,
1034 plus_end - plus_begin, plus_begin,
1035 line_prefix);
1036 }
1037 diff_words->current_plus = plus_end;
1039 diff_words->last_minus = minus_first;
1040}
1041/* This function starts looking at *begin, and returns 0 iff a word was found. */
1043static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
1044 int *begin, int *end)
1045{
1046 if (word_regex && *begin < buffer->size) {
1047 regmatch_t match[1];
1048 if (!regexec_buf(word_regex, buffer->ptr + *begin,
1049 buffer->size - *begin, 1, match, 0)) {
1050 char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
1051 '\n', match[0].rm_eo - match[0].rm_so);
1052 *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
1053 *begin += match[0].rm_so;
1054 return *begin >= *end;
1055 }
1056 return -1;
1057 }
1058 /* find the next word */
1060 while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
1061 (*begin)++;
1062 if (*begin >= buffer->size)
1063 return -1;
1064 /* find the end of the word */
1066 *end = *begin + 1;
1067 while (*end < buffer->size && !isspace(buffer->ptr[*end]))
1068 (*end)++;
1069 return 0;
1071}
1072/*
1074 * This function splits the words in buffer->text, stores the list with
1075 * newline separator into out, and saves the offsets of the original words
1076 * in buffer->orig.
1077 */
1078static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
1079 regex_t *word_regex)
1080{
1081 int i, j;
1082 long alloc = 0;
1083 out->size = 0;
1085 out->ptr = NULL;
1086 /* fake an empty "0th" word */
1088 ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
1089 buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
1090 buffer->orig_nr = 1;
1091 for (i = 0; i < buffer->text.size; i++) {
1093 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
1094 return;
1095 /* store original boundaries */
1097 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
1098 buffer->orig_alloc);
1099 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
1100 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
1101 buffer->orig_nr++;
1102 /* store one word */
1104 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
1105 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
1106 out->ptr[out->size + j - i] = '\n';
1107 out->size += j - i + 1;
1108 i = j - 1;
1110 }
1111}
1112/* this executes the word diff on the accumulated buffers */
1114static void diff_words_show(struct diff_words_data *diff_words)
1115{
1116 xpparam_t xpp;
1117 xdemitconf_t xecfg;
1118 mmfile_t minus, plus;
1119 struct diff_words_style *style = diff_words->style;
1120 struct diff_options *opt = diff_words->opt;
1122 const char *line_prefix;
1123 assert(opt);
1125 line_prefix = diff_line_prefix(opt);
1126 /* special case: only removal */
1128 if (!diff_words->plus.text.size) {
1129 fputs(line_prefix, diff_words->opt->file);
1130 fn_out_diff_words_write_helper(diff_words->opt->file,
1131 &style->old, style->newline,
1132 diff_words->minus.text.size,
1133 diff_words->minus.text.ptr, line_prefix);
1134 diff_words->minus.text.size = 0;
1135 return;
1136 }
1137 diff_words->current_plus = diff_words->plus.text.ptr;
1139 diff_words->last_minus = 0;
1140 memset(&xpp, 0, sizeof(xpp));
1142 memset(&xecfg, 0, sizeof(xecfg));
1143 diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
1144 diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
1145 xpp.flags = 0;
1146 /* as only the hunk header will be parsed, we need a 0-context */
1147 xecfg.ctxlen = 0;
1148 if (xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
1149 &xpp, &xecfg))
1150 die("unable to generate word diff");
1151 free(minus.ptr);
1152 free(plus.ptr);
1153 if (diff_words->current_plus != diff_words->plus.text.ptr +
1154 diff_words->plus.text.size) {
1155 if (color_words_output_graph_prefix(diff_words))
1156 fputs(line_prefix, diff_words->opt->file);
1157 fn_out_diff_words_write_helper(diff_words->opt->file,
1158 &style->ctx, style->newline,
1159 diff_words->plus.text.ptr + diff_words->plus.text.size
1160 - diff_words->current_plus, diff_words->current_plus,
1161 line_prefix);
1162 }
1163 diff_words->minus.text.size = diff_words->plus.text.size = 0;
1164}
1165/* In "color-words" mode, show word-diff of words accumulated in the buffer */
1167static void diff_words_flush(struct emit_callback *ecbdata)
1168{
1169 if (ecbdata->diff_words->minus.text.size ||
1170 ecbdata->diff_words->plus.text.size)
1171 diff_words_show(ecbdata->diff_words);
1172}
1173static void diff_filespec_load_driver(struct diff_filespec *one)
1175{
1176 /* Use already-loaded driver */
1177 if (one->driver)
1178 return;
1179 if (S_ISREG(one->mode))
1181 one->driver = userdiff_find_by_path(one->path);
1182 /* Fallback to default settings */
1184 if (!one->driver)
1185 one->driver = userdiff_find_by_name("default");
1186}
1187static const char *userdiff_word_regex(struct diff_filespec *one)
1189{
1190 diff_filespec_load_driver(one);
1191 return one->driver->word_regex;
1192}
1193static void init_diff_words_data(struct emit_callback *ecbdata,
1195 struct diff_options *orig_opts,
1196 struct diff_filespec *one,
1197 struct diff_filespec *two)
1198{
1199 int i;
1200 struct diff_options *o = xmalloc(sizeof(struct diff_options));
1201 memcpy(o, orig_opts, sizeof(struct diff_options));
1202 ecbdata->diff_words =
1204 xcalloc(1, sizeof(struct diff_words_data));
1205 ecbdata->diff_words->type = o->word_diff;
1206 ecbdata->diff_words->opt = o;
1207 if (!o->word_regex)
1208 o->word_regex = userdiff_word_regex(one);
1209 if (!o->word_regex)
1210 o->word_regex = userdiff_word_regex(two);
1211 if (!o->word_regex)
1212 o->word_regex = diff_word_regex_cfg;
1213 if (o->word_regex) {
1214 ecbdata->diff_words->word_regex = (regex_t *)
1215 xmalloc(sizeof(regex_t));
1216 if (regcomp(ecbdata->diff_words->word_regex,
1217 o->word_regex,
1218 REG_EXTENDED | REG_NEWLINE))
1219 die ("Invalid regular expression: %s",
1220 o->word_regex);
1221 }
1222 for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1223 if (o->word_diff == diff_words_styles[i].type) {
1224 ecbdata->diff_words->style =
1225 &diff_words_styles[i];
1226 break;
1227 }
1228 }
1229 if (want_color(o->use_color)) {
1230 struct diff_words_style *st = ecbdata->diff_words->style;
1231 st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1232 st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1233 st->ctx.color = diff_get_color_opt(o, DIFF_CONTEXT);
1234 }
1235}
1236static void free_diff_words_data(struct emit_callback *ecbdata)
1238{
1239 if (ecbdata->diff_words) {
1240 diff_words_flush(ecbdata);
1241 free (ecbdata->diff_words->opt);
1242 free (ecbdata->diff_words->minus.text.ptr);
1243 free (ecbdata->diff_words->minus.orig);
1244 free (ecbdata->diff_words->plus.text.ptr);
1245 free (ecbdata->diff_words->plus.orig);
1246 if (ecbdata->diff_words->word_regex) {
1247 regfree(ecbdata->diff_words->word_regex);
1248 free(ecbdata->diff_words->word_regex);
1249 }
1250 FREE_AND_NULL(ecbdata->diff_words);
1251 }
1252}
1253const char *diff_get_color(int diff_use_color, enum color_diff ix)
1255{
1256 if (want_color(diff_use_color))
1257 return diff_colors[ix];
1258 return "";
1259}
1260const char *diff_line_prefix(struct diff_options *opt)
1262{
1263 struct strbuf *msgbuf;
1264 if (!opt->output_prefix)
1265 return "";
1266 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
1268 return msgbuf->buf;
1269}
1270static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1272{
1273 const char *cp;
1274 unsigned long allot;
1275 size_t l = len;
1276 if (ecb->truncate)
1278 return ecb->truncate(line, len);
1279 cp = line;
1280 allot = l;
1281 while (0 < l) {
1282 (void) utf8_width(&cp, &l);
1283 if (!cp)
1284 break; /* truncated in the middle? */
1285 }
1286 return allot - l;
1287}
1288static void find_lno(const char *line, struct emit_callback *ecbdata)
1290{
1291 const char *p;
1292 ecbdata->lno_in_preimage = 0;
1293 ecbdata->lno_in_postimage = 0;
1294 p = strchr(line, '-');
1295 if (!p)
1296 return; /* cannot happen */
1297 ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1298 p = strchr(p, '+');
1299 if (!p)
1300 return; /* cannot happen */
1301 ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1302}
1303static void fn_out_consume(void *priv, char *line, unsigned long len)
1305{
1306 struct emit_callback *ecbdata = priv;
1307 const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
1308 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
1309 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1310 struct diff_options *o = ecbdata->opt;
1311 const char *line_prefix = diff_line_prefix(o);
1312 o->found_changes = 1;
1314 if (ecbdata->header) {
1316 fprintf(o->file, "%s", ecbdata->header->buf);
1317 strbuf_reset(ecbdata->header);
1318 ecbdata->header = NULL;
1319 }
1320 if (ecbdata->label_path[0]) {
1322 const char *name_a_tab, *name_b_tab;
1323 name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
1325 name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
1326 fprintf(o->file, "%s%s--- %s%s%s\n",
1328 line_prefix, meta, ecbdata->label_path[0], reset, name_a_tab);
1329 fprintf(o->file, "%s%s+++ %s%s%s\n",
1330 line_prefix, meta, ecbdata->label_path[1], reset, name_b_tab);
1331 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1332 }
1333 if (diff_suppress_blank_empty
1335 && len == 2 && line[0] == ' ' && line[1] == '\n') {
1336 line[0] = '\n';
1337 len = 1;
1338 }
1339 if (line[0] == '@') {
1341 if (ecbdata->diff_words)
1342 diff_words_flush(ecbdata);
1343 len = sane_truncate_line(ecbdata, line, len);
1344 find_lno(line, ecbdata);
1345 emit_hunk_header(ecbdata, line, len);
1346 return;
1347 }
1348 if (ecbdata->diff_words) {
1350 if (line[0] == '-') {
1351 diff_words_append(line, len,
1352 &ecbdata->diff_words->minus);
1353 return;
1354 } else if (line[0] == '+') {
1355 diff_words_append(line, len,
1356 &ecbdata->diff_words->plus);
1357 return;
1358 } else if (starts_with(line, "\\ ")) {
1359 /*
1360 * Eat the "no newline at eof" marker as if we
1361 * saw a "+" or "-" line with nothing on it,
1362 * and return without diff_words_flush() to
1363 * defer processing. If this is the end of
1364 * preimage, more "+" lines may come after it.
1365 */
1366 return;
1367 }
1368 diff_words_flush(ecbdata);
1369 if (ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN) {
1370 emit_line(o, context, reset, line, len);
1371 fputs("~\n", o->file);
1372 } else {
1373 /*
1374 * Skip the prefix character, if any. With
1375 * diff_suppress_blank_empty, there may be
1376 * none.
1377 */
1378 if (line[0] != '\n') {
1379 line++;
1380 len--;
1381 }
1382 emit_line(o, context, reset, line, len);
1383 }
1384 return;
1385 }
1386 switch (line[0]) {
1388 case '+':
1389 ecbdata->lno_in_postimage++;
1390 emit_add_line(reset, ecbdata, line + 1, len - 1);
1391 break;
1392 case '-':
1393 ecbdata->lno_in_preimage++;
1394 emit_del_line(reset, ecbdata, line + 1, len - 1);
1395 break;
1396 case ' ':
1397 ecbdata->lno_in_postimage++;
1398 ecbdata->lno_in_preimage++;
1399 emit_context_line(reset, ecbdata, line + 1, len - 1);
1400 break;
1401 default:
1402 /* incomplete line at the end */
1403 ecbdata->lno_in_preimage++;
1404 emit_line(o, diff_get_color(ecbdata->color_diff, DIFF_CONTEXT),
1405 reset, line, len);
1406 break;
1407 }
1408}
1409static char *pprint_rename(const char *a, const char *b)
1411{
1412 const char *old = a;
1413 const char *new = b;
1414 struct strbuf name = STRBUF_INIT;
1415 int pfx_length, sfx_length;
1416 int pfx_adjust_for_slash;
1417 int len_a = strlen(a);
1418 int len_b = strlen(b);
1419 int a_midlen, b_midlen;
1420 int qlen_a = quote_c_style(a, NULL, NULL, 0);
1421 int qlen_b = quote_c_style(b, NULL, NULL, 0);
1422 if (qlen_a || qlen_b) {
1424 quote_c_style(a, &name, NULL, 0);
1425 strbuf_addstr(&name, " => ");
1426 quote_c_style(b, &name, NULL, 0);
1427 return strbuf_detach(&name, NULL);
1428 }
1429 /* Find common prefix */
1431 pfx_length = 0;
1432 while (*old && *new && *old == *new) {
1433 if (*old == '/')
1434 pfx_length = old - a + 1;
1435 old++;
1436 new++;
1437 }
1438 /* Find common suffix */
1440 old = a + len_a;
1441 new = b + len_b;
1442 sfx_length = 0;
1443 /*
1444 * If there is a common prefix, it must end in a slash. In
1445 * that case we let this loop run 1 into the prefix to see the
1446 * same slash.
1447 *
1448 * If there is no common prefix, we cannot do this as it would
1449 * underrun the input strings.
1450 */
1451 pfx_adjust_for_slash = (pfx_length ? 1 : 0);
1452 while (a + pfx_length - pfx_adjust_for_slash <= old &&
1453 b + pfx_length - pfx_adjust_for_slash <= new &&
1454 *old == *new) {
1455 if (*old == '/')
1456 sfx_length = len_a - (old - a);
1457 old--;
1458 new--;
1459 }
1460 /*
1462 * pfx{mid-a => mid-b}sfx
1463 * {pfx-a => pfx-b}sfx
1464 * pfx{sfx-a => sfx-b}
1465 * name-a => name-b
1466 */
1467 a_midlen = len_a - pfx_length - sfx_length;
1468 b_midlen = len_b - pfx_length - sfx_length;
1469 if (a_midlen < 0)
1470 a_midlen = 0;
1471 if (b_midlen < 0)
1472 b_midlen = 0;
1473 strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1475 if (pfx_length + sfx_length) {
1476 strbuf_add(&name, a, pfx_length);
1477 strbuf_addch(&name, '{');
1478 }
1479 strbuf_add(&name, a + pfx_length, a_midlen);
1480 strbuf_addstr(&name, " => ");
1481 strbuf_add(&name, b + pfx_length, b_midlen);
1482 if (pfx_length + sfx_length) {
1483 strbuf_addch(&name, '}');
1484 strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1485 }
1486 return strbuf_detach(&name, NULL);
1487}
1488struct diffstat_t {
1490 int nr;
1491 int alloc;
1492 struct diffstat_file {
1493 char *from_name;
1494 char *name;
1495 char *print_name;
1496 unsigned is_unmerged:1;
1497 unsigned is_binary:1;
1498 unsigned is_renamed:1;
1499 unsigned is_interesting:1;
1500 uintmax_t added, deleted;
1501 } **files;
1502};
1503static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1505 const char *name_a,
1506 const char *name_b)
1507{
1508 struct diffstat_file *x;
1509 x = xcalloc(1, sizeof(*x));
1510 ALLOC_GROW(diffstat->files, diffstat->nr + 1, diffstat->alloc);
1511 diffstat->files[diffstat->nr++] = x;
1512 if (name_b) {
1513 x->from_name = xstrdup(name_a);
1514 x->name = xstrdup(name_b);
1515 x->is_renamed = 1;
1516 }
1517 else {
1518 x->from_name = NULL;
1519 x->name = xstrdup(name_a);
1520 }
1521 return x;
1522}
1523static void diffstat_consume(void *priv, char *line, unsigned long len)
1525{
1526 struct diffstat_t *diffstat = priv;
1527 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1528 if (line[0] == '+')
1530 x->added++;
1531 else if (line[0] == '-')
1532 x->deleted++;
1533}
1534const char mime_boundary_leader[] = "------------";
1536static int scale_linear(int it, int width, int max_change)
1538{
1539 if (!it)
1540 return 0;
1541 /*
1542 * make sure that at least one '-' or '+' is printed if
1543 * there is any change to this path. The easiest way is to
1544 * scale linearly as if the alloted width is one column shorter
1545 * than it is, and then add 1 to the result.
1546 */
1547 return 1 + (it * (width - 1) / max_change);
1548}
1549static void show_name(FILE *file,
1551 const char *prefix, const char *name, int len)
1552{
1553 fprintf(file, " %s%-*s |", prefix, len, name);
1554}
1555static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1557{
1558 if (cnt <= 0)
1559 return;
1560 fprintf(file, "%s", set);
1561 while (cnt--)
1562 putc(ch, file);
1563 fprintf(file, "%s", reset);
1564}
1565static void fill_print_name(struct diffstat_file *file)
1567{
1568 char *pname;
1569 if (file->print_name)
1571 return;
1572 if (!file->is_renamed) {
1574 struct strbuf buf = STRBUF_INIT;
1575 if (quote_c_style(file->name, &buf, NULL, 0)) {
1576 pname = strbuf_detach(&buf, NULL);
1577 } else {
1578 pname = file->name;
1579 strbuf_release(&buf);
1580 }
1581 } else {
1582 pname = pprint_rename(file->from_name, file->name);
1583 }
1584 file->print_name = pname;
1585}
1586int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1588{
1589 struct strbuf sb = STRBUF_INIT;
1590 int ret;
1591 if (!files) {
1593 assert(insertions == 0 && deletions == 0);
1594 return fprintf(fp, "%s\n", " 0 files changed");
1595 }
1596 strbuf_addf(&sb,
1598 (files == 1) ? " %d file changed" : " %d files changed",
1599 files);
1600 /*
1602 * For binary diff, the caller may want to print "x files
1603 * changed" with insertions == 0 && deletions == 0.
1604 *
1605 * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1606 * is probably less confusing (i.e skip over "2 files changed
1607 * but nothing about added/removed lines? Is this a bug in Git?").
1608 */
1609 if (insertions || deletions == 0) {
1610 strbuf_addf(&sb,
1611 (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
1612 insertions);
1613 }
1614 if (deletions || insertions == 0) {
1616 strbuf_addf(&sb,
1617 (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
1618 deletions);
1619 }
1620 strbuf_addch(&sb, '\n');
1621 ret = fputs(sb.buf, fp);
1622 strbuf_release(&sb);
1623 return ret;
1624}
1625static void show_stats(struct diffstat_t *data, struct diff_options *options)
1627{
1628 int i, len, add, del, adds = 0, dels = 0;
1629 uintmax_t max_change = 0, max_len = 0;
1630 int total_files = data->nr, count;
1631 int width, name_width, graph_width, number_width = 0, bin_width = 0;
1632 const char *reset, *add_c, *del_c;
1633 const char *line_prefix = "";
1634 int extra_shown = 0;
1635 if (data->nr == 0)
1637 return;
1638 line_prefix = diff_line_prefix(options);
1640 count = options->stat_count ? options->stat_count : data->nr;
1641 reset = diff_get_color_opt(options, DIFF_RESET);
1643 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1644 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1645 /*
1647 * Find the longest filename and max number of changes
1648 */
1649 for (i = 0; (i < count) && (i < data->nr); i++) {
1650 struct diffstat_file *file = data->files[i];
1651 uintmax_t change = file->added + file->deleted;
1652 if (!file->is_interesting && (change == 0)) {
1654 count++; /* not shown == room for one more */
1655 continue;
1656 }
1657 fill_print_name(file);
1658 len = strlen(file->print_name);
1659 if (max_len < len)
1660 max_len = len;
1661 if (file->is_unmerged) {
1663 /* "Unmerged" is 8 characters */
1664 bin_width = bin_width < 8 ? 8 : bin_width;
1665 continue;
1666 }
1667 if (file->is_binary) {
1668 /* "Bin XXX -> YYY bytes" */
1669 int w = 14 + decimal_width(file->added)
1670 + decimal_width(file->deleted);
1671 bin_width = bin_width < w ? w : bin_width;
1672 /* Display change counts aligned with "Bin" */
1673 number_width = 3;
1674 continue;
1675 }
1676 if (max_change < change)
1678 max_change = change;
1679 }
1680 count = i; /* where we can stop scanning in data->files[] */
1681 /*
1683 * We have width = stat_width or term_columns() columns total.
1684 * We want a maximum of min(max_len, stat_name_width) for the name part.
1685 * We want a maximum of min(max_change, stat_graph_width) for the +- part.
1686 * We also need 1 for " " and 4 + decimal_width(max_change)
1687 * for " | NNNN " and one the empty column at the end, altogether
1688 * 6 + decimal_width(max_change).
1689 *
1690 * If there's not enough space, we will use the smaller of
1691 * stat_name_width (if set) and 5/8*width for the filename,
1692 * and the rest for constant elements + graph part, but no more
1693 * than stat_graph_width for the graph part.
1694 * (5/8 gives 50 for filename and 30 for the constant parts + graph
1695 * for the standard terminal size).
1696 *
1697 * In other words: stat_width limits the maximum width, and
1698 * stat_name_width fixes the maximum width of the filename,
1699 * and is also used to divide available columns if there
1700 * aren't enough.
1701 *
1702 * Binary files are displayed with "Bin XXX -> YYY bytes"
1703 * instead of the change count and graph. This part is treated
1704 * similarly to the graph part, except that it is not
1705 * "scaled". If total width is too small to accommodate the
1706 * guaranteed minimum width of the filename part and the
1707 * separators and this message, this message will "overflow"
1708 * making the line longer than the maximum width.
1709 */
1710 if (options->stat_width == -1)
1712 width = term_columns() - strlen(line_prefix);
1713 else
1714 width = options->stat_width ? options->stat_width : 80;
1715 number_width = decimal_width(max_change) > number_width ?
1716 decimal_width(max_change) : number_width;
1717 if (options->stat_graph_width == -1)
1719 options->stat_graph_width = diff_stat_graph_width;
1720 /*
1722 * Guarantee 3/8*16==6 for the graph part
1723 * and 5/8*16==10 for the filename part
1724 */
1725 if (width < 16 + 6 + number_width)
1726 width = 16 + 6 + number_width;
1727 /*
1729 * First assign sizes that are wanted, ignoring available width.
1730 * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
1731 * starting from "XXX" should fit in graph_width.
1732 */
1733 graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
1734 if (options->stat_graph_width &&
1735 options->stat_graph_width < graph_width)
1736 graph_width = options->stat_graph_width;
1737 name_width = (options->stat_name_width > 0 &&
1739 options->stat_name_width < max_len) ?
1740 options->stat_name_width : max_len;
1741 /*
1743 * Adjust adjustable widths not to exceed maximum width
1744 */
1745 if (name_width + number_width + 6 + graph_width > width) {
1746 if (graph_width > width * 3/8 - number_width - 6) {
1747 graph_width = width * 3/8 - number_width - 6;
1748 if (graph_width < 6)
1749 graph_width = 6;
1750 }
1751 if (options->stat_graph_width &&
1753 graph_width > options->stat_graph_width)
1754 graph_width = options->stat_graph_width;
1755 if (name_width > width - number_width - 6 - graph_width)
1756 name_width = width - number_width - 6 - graph_width;
1757 else
1758 graph_width = width - number_width - 6 - name_width;
1759 }
1760 /*
1762 * From here name_width is the width of the name area,
1763 * and graph_width is the width of the graph area.
1764 * max_change is used to scale graph properly.
1765 */
1766 for (i = 0; i < count; i++) {
1767 const char *prefix = "";
1768 struct diffstat_file *file = data->files[i];
1769 char *name = file->print_name;
1770 uintmax_t added = file->added;
1771 uintmax_t deleted = file->deleted;
1772 int name_len;
1773 if (!file->is_interesting && (added + deleted == 0))
1775 continue;
1776 /*
1778 * "scale" the filename
1779 */
1780 len = name_width;
1781 name_len = strlen(name);
1782 if (name_width < name_len) {
1783 char *slash;
1784 prefix = "...";
1785 len -= 3;
1786 name += name_len - len;
1787 slash = strchr(name, '/');
1788 if (slash)
1789 name = slash;
1790 }
1791 if (file->is_binary) {
1793 fprintf(options->file, "%s", line_prefix);
1794 show_name(options->file, prefix, name, len);
1795 fprintf(options->file, " %*s", number_width, "Bin");
1796 if (!added && !deleted) {
1797 putc('\n', options->file);
1798 continue;
1799 }
1800 fprintf(options->file, " %s%"PRIuMAX"%s",
1801 del_c, deleted, reset);
1802 fprintf(options->file, " -> ");
1803 fprintf(options->file, "%s%"PRIuMAX"%s",
1804 add_c, added, reset);
1805 fprintf(options->file, " bytes");
1806 fprintf(options->file, "\n");
1807 continue;
1808 }
1809 else if (file->is_unmerged) {
1810 fprintf(options->file, "%s", line_prefix);
1811 show_name(options->file, prefix, name, len);
1812 fprintf(options->file, " Unmerged\n");
1813 continue;
1814 }
1815 /*
1817 * scale the add/delete
1818 */
1819 add = added;
1820 del = deleted;
1821 if (graph_width <= max_change) {
1823 int total = scale_linear(add + del, graph_width, max_change);
1824 if (total < 2 && add && del)
1825 /* width >= 2 due to the sanity check */
1826 total = 2;
1827 if (add < del) {
1828 add = scale_linear(add, graph_width, max_change);
1829 del = total - add;
1830 } else {
1831 del = scale_linear(del, graph_width, max_change);
1832 add = total - del;
1833 }
1834 }
1835 fprintf(options->file, "%s", line_prefix);
1836 show_name(options->file, prefix, name, len);
1837 fprintf(options->file, " %*"PRIuMAX"%s",
1838 number_width, added + deleted,
1839 added + deleted ? " " : "");
1840 show_graph(options->file, '+', add, add_c, reset);
1841 show_graph(options->file, '-', del, del_c, reset);
1842 fprintf(options->file, "\n");
1843 }
1844 for (i = 0; i < data->nr; i++) {
1846 struct diffstat_file *file = data->files[i];
1847 uintmax_t added = file->added;
1848 uintmax_t deleted = file->deleted;
1849 if (file->is_unmerged ||
1851 (!file->is_interesting && (added + deleted == 0))) {
1852 total_files--;
1853 continue;
1854 }
1855 if (!file->is_binary) {
1857 adds += added;
1858 dels += deleted;
1859 }
1860 if (i < count)
1861 continue;
1862 if (!extra_shown)
1863 fprintf(options->file, "%s ...\n", line_prefix);
1864 extra_shown = 1;
1865 }
1866 fprintf(options->file, "%s", line_prefix);
1867 print_stat_summary(options->file, total_files, adds, dels);
1868}
1869static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1871{
1872 int i, adds = 0, dels = 0, total_files = data->nr;
1873 if (data->nr == 0)
1875 return;
1876 for (i = 0; i < data->nr; i++) {
1878 int added = data->files[i]->added;
1879 int deleted= data->files[i]->deleted;
1880 if (data->files[i]->is_unmerged ||
1882 (!data->files[i]->is_interesting && (added + deleted == 0))) {
1883 total_files--;
1884 } else if (!data->files[i]->is_binary) { /* don't count bytes */
1885 adds += added;
1886 dels += deleted;
1887 }
1888 }
1889 fprintf(options->file, "%s", diff_line_prefix(options));
1890 print_stat_summary(options->file, total_files, adds, dels);
1891}
1892static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1894{
1895 int i;
1896 if (data->nr == 0)
1898 return;
1899 for (i = 0; i < data->nr; i++) {
1901 struct diffstat_file *file = data->files[i];
1902 fprintf(options->file, "%s", diff_line_prefix(options));
1904 if (file->is_binary)
1906 fprintf(options->file, "-\t-\t");
1907 else
1908 fprintf(options->file,
1909 "%"PRIuMAX"\t%"PRIuMAX"\t",
1910 file->added, file->deleted);
1911 if (options->line_termination) {
1912 fill_print_name(file);
1913 if (!file->is_renamed)
1914 write_name_quoted(file->name, options->file,
1915 options->line_termination);
1916 else {
1917 fputs(file->print_name, options->file);
1918 putc(options->line_termination, options->file);
1919 }
1920 } else {
1921 if (file->is_renamed) {
1922 putc('\0', options->file);
1923 write_name_quoted(file->from_name, options->file, '\0');
1924 }
1925 write_name_quoted(file->name, options->file, '\0');
1926 }
1927 }
1928}
1929struct dirstat_file {
1931 const char *name;
1932 unsigned long changed;
1933};
1934struct dirstat_dir {
1936 struct dirstat_file *files;
1937 int alloc, nr, permille, cumulative;
1938};
1939static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1941 unsigned long changed, const char *base, int baselen)
1942{
1943 unsigned long this_dir = 0;
1944 unsigned int sources = 0;
1945 const char *line_prefix = diff_line_prefix(opt);
1946 while (dir->nr) {
1948 struct dirstat_file *f = dir->files;
1949 int namelen = strlen(f->name);
1950 unsigned long this;
1951 char *slash;
1952 if (namelen < baselen)
1954 break;
1955 if (memcmp(f->name, base, baselen))
1956 break;
1957 slash = strchr(f->name + baselen, '/');
1958 if (slash) {
1959 int newbaselen = slash + 1 - f->name;
1960 this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
1961 sources++;
1962 } else {
1963 this = f->changed;
1964 dir->files++;
1965 dir->nr--;
1966 sources += 2;
1967 }
1968 this_dir += this;
1969 }
1970 /*
1972 * We don't report dirstat's for
1973 * - the top level
1974 * - or cases where everything came from a single directory
1975 * under this directory (sources == 1).
1976 */
1977 if (baselen && sources != 1) {
1978 if (this_dir) {
1979 int permille = this_dir * 1000 / changed;
1980 if (permille >= dir->permille) {
1981 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
1982 permille / 10, permille % 10, baselen, base);
1983 if (!dir->cumulative)
1984 return 0;
1985 }
1986 }
1987 }
1988 return this_dir;
1989}
1990static int dirstat_compare(const void *_a, const void *_b)
1992{
1993 const struct dirstat_file *a = _a;
1994 const struct dirstat_file *b = _b;
1995 return strcmp(a->name, b->name);
1996}
1997static void show_dirstat(struct diff_options *options)
1999{
2000 int i;
2001 unsigned long changed;
2002 struct dirstat_dir dir;
2003 struct diff_queue_struct *q = &diff_queued_diff;
2004 dir.files = NULL;
2006 dir.alloc = 0;
2007 dir.nr = 0;
2008 dir.permille = options->dirstat_permille;
2009 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
2010 changed = 0;
2012 for (i = 0; i < q->nr; i++) {
2013 struct diff_filepair *p = q->queue[i];
2014 const char *name;
2015 unsigned long copied, added, damage;
2016 int content_changed;
2017 name = p->two->path ? p->two->path : p->one->path;
2019 if (p->one->oid_valid && p->two->oid_valid)
2021 content_changed = oidcmp(&p->one->oid, &p->two->oid);
2022 else
2023 content_changed = 1;
2024 if (!content_changed) {
2026 /*
2027 * The SHA1 has not changed, so pre-/post-content is
2028 * identical. We can therefore skip looking at the
2029 * file contents altogether.
2030 */
2031 damage = 0;
2032 goto found_damage;
2033 }
2034 if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
2036 /*
2037 * In --dirstat-by-file mode, we don't really need to
2038 * look at the actual file contents at all.
2039 * The fact that the SHA1 changed is enough for us to
2040 * add this file to the list of results
2041 * (with each file contributing equal damage).
2042 */
2043 damage = 1;
2044 goto found_damage;
2045 }
2046 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
2048 diff_populate_filespec(p->one, 0);
2049 diff_populate_filespec(p->two, 0);
2050 diffcore_count_changes(p->one, p->two, NULL, NULL,
2051 &copied, &added);
2052 diff_free_filespec_data(p->one);
2053 diff_free_filespec_data(p->two);
2054 } else if (DIFF_FILE_VALID(p->one)) {
2055 diff_populate_filespec(p->one, CHECK_SIZE_ONLY);
2056 copied = added = 0;
2057 diff_free_filespec_data(p->one);
2058 } else if (DIFF_FILE_VALID(p->two)) {
2059 diff_populate_filespec(p->two, CHECK_SIZE_ONLY);
2060 copied = 0;
2061 added = p->two->size;
2062 diff_free_filespec_data(p->two);
2063 } else
2064 continue;
2065 /*
2067 * Original minus copied is the removed material,
2068 * added is the new material. They are both damages
2069 * made to the preimage.
2070 * If the resulting damage is zero, we know that
2071 * diffcore_count_changes() considers the two entries to
2072 * be identical, but since content_changed is true, we
2073 * know that there must have been _some_ kind of change,
2074 * so we force all entries to have damage > 0.
2075 */
2076 damage = (p->one->size - copied) + added;
2077 if (!damage)
2078 damage = 1;
2079found_damage:
2081 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2082 dir.files[dir.nr].name = name;
2083 dir.files[dir.nr].changed = damage;
2084 changed += damage;
2085 dir.nr++;
2086 }
2087 /* This can happen even with many files, if everything was renames */
2089 if (!changed)
2090 return;
2091 /* Show all directories with more than x% of the changes */
2093 QSORT(dir.files, dir.nr, dirstat_compare);
2094 gather_dirstat(options, &dir, changed, "", 0);
2095}
2096static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
2098{
2099 int i;
2100 unsigned long changed;
2101 struct dirstat_dir dir;
2102 if (data->nr == 0)
2104 return;
2105 dir.files = NULL;
2107 dir.alloc = 0;
2108 dir.nr = 0;
2109 dir.permille = options->dirstat_permille;
2110 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
2111 changed = 0;
2113 for (i = 0; i < data->nr; i++) {
2114 struct diffstat_file *file = data->files[i];
2115 unsigned long damage = file->added + file->deleted;
2116 if (file->is_binary)
2117 /*
2118 * binary files counts bytes, not lines. Must find some
2119 * way to normalize binary bytes vs. textual lines.
2120 * The following heuristic assumes that there are 64
2121 * bytes per "line".
2122 * This is stupid and ugly, but very cheap...
2123 */
2124 damage = (damage + 63) / 64;
2125 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2126 dir.files[dir.nr].name = file->name;
2127 dir.files[dir.nr].changed = damage;
2128 changed += damage;
2129 dir.nr++;
2130 }
2131 /* This can happen even with many files, if everything was renames */
2133 if (!changed)
2134 return;
2135 /* Show all directories with more than x% of the changes */
2137 QSORT(dir.files, dir.nr, dirstat_compare);
2138 gather_dirstat(options, &dir, changed, "", 0);
2139}
2140static void free_diffstat_info(struct diffstat_t *diffstat)
2142{
2143 int i;
2144 for (i = 0; i < diffstat->nr; i++) {
2145 struct diffstat_file *f = diffstat->files[i];
2146 if (f->name != f->print_name)
2147 free(f->print_name);
2148 free(f->name);
2149 free(f->from_name);
2150 free(f);
2151 }
2152 free(diffstat->files);
2153}
2154struct checkdiff_t {
2156 const char *filename;
2157 int lineno;
2158 int conflict_marker_size;
2159 struct diff_options *o;
2160 unsigned ws_rule;
2161 unsigned status;
2162};
2163static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2165{
2166 char firstchar;
2167 int cnt;
2168 if (len < marker_size + 1)
2170 return 0;
2171 firstchar = line[0];
2172 switch (firstchar) {
2173 case '=': case '>': case '<': case '|':
2174 break;
2175 default:
2176 return 0;
2177 }
2178 for (cnt = 1; cnt < marker_size; cnt++)
2179 if (line[cnt] != firstchar)
2180 return 0;
2181 /* line[1] thru line[marker_size-1] are same as firstchar */
2182 if (len < marker_size + 1 || !isspace(line[marker_size]))
2183 return 0;
2184 return 1;
2185}
2186static void checkdiff_consume(void *priv, char *line, unsigned long len)
2188{
2189 struct checkdiff_t *data = priv;
2190 int marker_size = data->conflict_marker_size;
2191 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2192 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2193 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2194 char *err;
2195 const char *line_prefix;
2196 assert(data->o);
2198 line_prefix = diff_line_prefix(data->o);
2199 if (line[0] == '+') {
2201 unsigned bad;
2202 data->lineno++;
2203 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2204 data->status |= 1;
2205 fprintf(data->o->file,
2206 "%s%s:%d: leftover conflict marker\n",
2207 line_prefix, data->filename, data->lineno);
2208 }
2209 bad = ws_check(line + 1, len - 1, data->ws_rule);
2210 if (!bad)
2211 return;
2212 data->status |= bad;
2213 err = whitespace_error_string(bad);
2214 fprintf(data->o->file, "%s%s:%d: %s.\n",
2215 line_prefix, data->filename, data->lineno, err);
2216 free(err);
2217 emit_line(data->o, set, reset, line, 1);
2218 ws_check_emit(line + 1, len - 1, data->ws_rule,
2219 data->o->file, set, reset, ws);
2220 } else if (line[0] == ' ') {
2221 data->lineno++;
2222 } else if (line[0] == '@') {
2223 char *plus = strchr(line, '+');
2224 if (plus)
2225 data->lineno = strtol(plus, NULL, 10) - 1;
2226 else
2227 die("invalid diff");
2228 }
2229}
2230static unsigned char *deflate_it(char *data,
2232 unsigned long size,
2233 unsigned long *result_size)
2234{
2235 int bound;
2236 unsigned char *deflated;
2237 git_zstream stream;
2238 git_deflate_init(&stream, zlib_compression_level);
2240 bound = git_deflate_bound(&stream, size);
2241 deflated = xmalloc(bound);
2242 stream.next_out = deflated;
2243 stream.avail_out = bound;
2244 stream.next_in = (unsigned char *)data;
2246 stream.avail_in = size;
2247 while (git_deflate(&stream, Z_FINISH) == Z_OK)
2248 ; /* nothing */
2249 git_deflate_end(&stream);
2250 *result_size = stream.total_out;
2251 return deflated;
2252}
2253static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two,
2255 const char *prefix)
2256{
2257 void *cp;
2258 void *delta;
2259 void *deflated;
2260 void *data;
2261 unsigned long orig_size;
2262 unsigned long delta_size;
2263 unsigned long deflate_size;
2264 unsigned long data_size;
2265 /* We could do deflated delta, or we could do just deflated two,
2267 * whichever is smaller.
2268 */
2269 delta = NULL;
2270 deflated = deflate_it(two->ptr, two->size, &deflate_size);
2271 if (one->size && two->size) {
2272 delta = diff_delta(one->ptr, one->size,
2273 two->ptr, two->size,
2274 &delta_size, deflate_size);
2275 if (delta) {
2276 void *to_free = delta;
2277 orig_size = delta_size;
2278 delta = deflate_it(delta, delta_size, &delta_size);
2279 free(to_free);
2280 }
2281 }
2282 if (delta && delta_size < deflate_size) {
2284 fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2285 free(deflated);
2286 data = delta;
2287 data_size = delta_size;
2288 }
2289 else {
2290 fprintf(file, "%sliteral %lu\n", prefix, two->size);
2291 free(delta);
2292 data = deflated;
2293 data_size = deflate_size;
2294 }
2295 /* emit data encoded in base85 */
2297 cp = data;
2298 while (data_size) {
2299 int bytes = (52 < data_size) ? 52 : data_size;
2300 char line[70];
2301 data_size -= bytes;
2302 if (bytes <= 26)
2303 line[0] = bytes + 'A' - 1;
2304 else
2305 line[0] = bytes - 26 + 'a' - 1;
2306 encode_85(line + 1, cp, bytes);
2307 cp = (char *) cp + bytes;
2308 fprintf(file, "%s", prefix);
2309 fputs(line, file);
2310 fputc('\n', file);
2311 }
2312 fprintf(file, "%s\n", prefix);
2313 free(data);
2314}
2315static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two,
2317 const char *prefix)
2318{
2319 fprintf(file, "%sGIT binary patch\n", prefix);
2320 emit_binary_diff_body(file, one, two, prefix);
2321 emit_binary_diff_body(file, two, one, prefix);
2322}
2323int diff_filespec_is_binary(struct diff_filespec *one)
2325{
2326 if (one->is_binary == -1) {
2327 diff_filespec_load_driver(one);
2328 if (one->driver->binary != -1)
2329 one->is_binary = one->driver->binary;
2330 else {
2331 if (!one->data && DIFF_FILE_VALID(one))
2332 diff_populate_filespec(one, CHECK_BINARY);
2333 if (one->is_binary == -1 && one->data)
2334 one->is_binary = buffer_is_binary(one->data,
2335 one->size);
2336 if (one->is_binary == -1)
2337 one->is_binary = 0;
2338 }
2339 }
2340 return one->is_binary;
2341}
2342static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2344{
2345 diff_filespec_load_driver(one);
2346 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2347}
2348void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2350{
2351 if (!options->a_prefix)
2352 options->a_prefix = a;
2353 if (!options->b_prefix)
2354 options->b_prefix = b;
2355}
2356struct userdiff_driver *get_textconv(struct diff_filespec *one)
2358{
2359 if (!DIFF_FILE_VALID(one))
2360 return NULL;
2361 diff_filespec_load_driver(one);
2363 return userdiff_get_textconv(one->driver);
2364}
2365static void builtin_diff(const char *name_a,
2367 const char *name_b,
2368 struct diff_filespec *one,
2369 struct diff_filespec *two,
2370 const char *xfrm_msg,
2371 int must_show_header,
2372 struct diff_options *o,
2373 int complete_rewrite)
2374{
2375 mmfile_t mf1, mf2;
2376 const char *lbl[2];
2377 char *a_one, *b_two;
2378 const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
2379 const char *reset = diff_get_color_opt(o, DIFF_RESET);
2380 const char *a_prefix, *b_prefix;
2381 struct userdiff_driver *textconv_one = NULL;
2382 struct userdiff_driver *textconv_two = NULL;
2383 struct strbuf header = STRBUF_INIT;
2384 const char *line_prefix = diff_line_prefix(o);
2385 diff_set_mnemonic_prefix(o, "a/", "b/");
2387 if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2388 a_prefix = o->b_prefix;
2389 b_prefix = o->a_prefix;
2390 } else {
2391 a_prefix = o->a_prefix;
2392 b_prefix = o->b_prefix;
2393 }
2394 if (o->submodule_format == DIFF_SUBMODULE_LOG &&
2396 (!one->mode || S_ISGITLINK(one->mode)) &&
2397 (!two->mode || S_ISGITLINK(two->mode))) {
2398 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2399 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2400 show_submodule_summary(o->file, one->path ? one->path : two->path,
2401 line_prefix,
2402 &one->oid, &two->oid,
2403 two->dirty_submodule,
2404 meta, del, add, reset);
2405 return;
2406 } else if (o->submodule_format == DIFF_SUBMODULE_INLINE_DIFF &&
2407 (!one->mode || S_ISGITLINK(one->mode)) &&
2408 (!two->mode || S_ISGITLINK(two->mode))) {
2409 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2410 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2411 show_submodule_inline_diff(o->file, one->path ? one->path : two->path,
2412 line_prefix,
2413 &one->oid, &two->oid,
2414 two->dirty_submodule,
2415 meta, del, add, reset, o);
2416 return;
2417 }
2418 if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2420 textconv_one = get_textconv(one);
2421 textconv_two = get_textconv(two);
2422 }
2423 /* Never use a non-valid filename anywhere if at all possible */
2425 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2426 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2427 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2429 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2430 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2431 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2432 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
2433 if (lbl[0][0] == '/') {
2434 /* /dev/null */
2435 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
2436 if (xfrm_msg)
2437 strbuf_addstr(&header, xfrm_msg);
2438 must_show_header = 1;
2439 }
2440 else if (lbl[1][0] == '/') {
2441 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
2442 if (xfrm_msg)
2443 strbuf_addstr(&header, xfrm_msg);
2444 must_show_header = 1;
2445 }
2446 else {
2447 if (one->mode != two->mode) {
2448 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
2449 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
2450 must_show_header = 1;
2451 }
2452 if (xfrm_msg)
2453 strbuf_addstr(&header, xfrm_msg);
2454 /*
2456 * we do not run diff between different kind
2457 * of objects.
2458 */
2459 if ((one->mode ^ two->mode) & S_IFMT)
2460 goto free_ab_and_return;
2461 if (complete_rewrite &&
2462 (textconv_one || !diff_filespec_is_binary(one)) &&
2463 (textconv_two || !diff_filespec_is_binary(two))) {
2464 fprintf(o->file, "%s", header.buf);
2465 strbuf_reset(&header);
2466 emit_rewrite_diff(name_a, name_b, one, two,
2467 textconv_one, textconv_two, o);
2468 o->found_changes = 1;
2469 goto free_ab_and_return;
2470 }
2471 }
2472 if (o->irreversible_delete && lbl[1][0] == '/') {
2474 fprintf(o->file, "%s", header.buf);
2475 strbuf_reset(&header);
2476 goto free_ab_and_return;
2477 } else if (!DIFF_OPT_TST(o, TEXT) &&
2478 ( (!textconv_one && diff_filespec_is_binary(one)) ||
2479 (!textconv_two && diff_filespec_is_binary(two)) )) {
2480 if (!one->data && !two->data &&
2481 S_ISREG(one->mode) && S_ISREG(two->mode) &&
2482 !DIFF_OPT_TST(o, BINARY)) {
2483 if (!oidcmp(&one->oid, &two->oid)) {
2484 if (must_show_header)
2485 fprintf(o->file, "%s", header.buf);
2486 goto free_ab_and_return;
2487 }
2488 fprintf(o->file, "%s", header.buf);
2489 fprintf(o->file, "%sBinary files %s and %s differ\n",
2490 line_prefix, lbl[0], lbl[1]);
2491 goto free_ab_and_return;
2492 }
2493 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2494 die("unable to read files to diff");
2495 /* Quite common confusing case */
2496 if (mf1.size == mf2.size &&
2497 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2498 if (must_show_header)
2499 fprintf(o->file, "%s", header.buf);
2500 goto free_ab_and_return;
2501 }
2502 fprintf(o->file, "%s", header.buf);
2503 strbuf_reset(&header);
2504 if (DIFF_OPT_TST(o, BINARY))
2505 emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2506 else
2507 fprintf(o->file, "%sBinary files %s and %s differ\n",
2508 line_prefix, lbl[0], lbl[1]);
2509 o->found_changes = 1;
2510 } else {
2511 /* Crazy xdl interfaces.. */
2512 const char *diffopts = getenv("GIT_DIFF_OPTS");
2513 const char *v;
2514 xpparam_t xpp;
2515 xdemitconf_t xecfg;
2516 struct emit_callback ecbdata;
2517 const struct userdiff_funcname *pe;
2518 if (must_show_header) {
2520 fprintf(o->file, "%s", header.buf);
2521 strbuf_reset(&header);
2522 }
2523 mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2525 mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2526 pe = diff_funcname_pattern(one);
2528 if (!pe)
2529 pe = diff_funcname_pattern(two);
2530 memset(&xpp, 0, sizeof(xpp));
2532 memset(&xecfg, 0, sizeof(xecfg));
2533 memset(&ecbdata, 0, sizeof(ecbdata));
2534 ecbdata.label_path = lbl;
2535 ecbdata.color_diff = want_color(o->use_color);
2536 ecbdata.ws_rule = whitespace_rule(name_b);
2537 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2538 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2539 ecbdata.opt = o;
2540 ecbdata.header = header.len ? &header : NULL;
2541 xpp.flags = o->xdl_opts;
2542 xecfg.ctxlen = o->context;
2543 xecfg.interhunkctxlen = o->interhunkcontext;
2544 xecfg.flags = XDL_EMIT_FUNCNAMES;
2545 if (DIFF_OPT_TST(o, FUNCCONTEXT))
2546 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2547 if (pe)
2548 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2549 if (!diffopts)
2550 ;
2551 else if (skip_prefix(diffopts, "--unified=", &v))
2552 xecfg.ctxlen = strtoul(v, NULL, 10);
2553 else if (skip_prefix(diffopts, "-u", &v))
2554 xecfg.ctxlen = strtoul(v, NULL, 10);
2555 if (o->word_diff)
2556 init_diff_words_data(&ecbdata, o, one, two);
2557 if (xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2558 &xpp, &xecfg))
2559 die("unable to generate diff for %s", one->path);
2560 if (o->word_diff)
2561 free_diff_words_data(&ecbdata);
2562 if (textconv_one)
2563 free(mf1.ptr);
2564 if (textconv_two)
2565 free(mf2.ptr);
2566 xdiff_clear_find_func(&xecfg);
2567 }
2568 free_ab_and_return:
2570 strbuf_release(&header);
2571 diff_free_filespec_data(one);
2572 diff_free_filespec_data(two);
2573 free(a_one);
2574 free(b_two);
2575 return;
2576}
2577static void builtin_diffstat(const char *name_a, const char *name_b,
2579 struct diff_filespec *one,
2580 struct diff_filespec *two,
2581 struct diffstat_t *diffstat,
2582 struct diff_options *o,
2583 struct diff_filepair *p)
2584{
2585 mmfile_t mf1, mf2;
2586 struct diffstat_file *data;
2587 int same_contents;
2588 int complete_rewrite = 0;
2589 if (!DIFF_PAIR_UNMERGED(p)) {
2591 if (p->status == DIFF_STATUS_MODIFIED && p->score)
2592 complete_rewrite = 1;
2593 }
2594 data = diffstat_add(diffstat, name_a, name_b);
2596 data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
2597 if (!one || !two) {
2599 data->is_unmerged = 1;
2600 return;
2601 }
2602 same_contents = !oidcmp(&one->oid, &two->oid);
2604 if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2606 data->is_binary = 1;
2607 if (same_contents) {
2608 data->added = 0;
2609 data->deleted = 0;
2610 } else {
2611 data->added = diff_filespec_size(two);
2612 data->deleted = diff_filespec_size(one);
2613 }
2614 }
2615 else if (complete_rewrite) {
2617 diff_populate_filespec(one, 0);
2618 diff_populate_filespec(two, 0);
2619 data->deleted = count_lines(one->data, one->size);
2620 data->added = count_lines(two->data, two->size);
2621 }
2622 else if (!same_contents) {
2624 /* Crazy xdl interfaces.. */
2625 xpparam_t xpp;
2626 xdemitconf_t xecfg;
2627 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2629 die("unable to read files to diff");
2630 memset(&xpp, 0, sizeof(xpp));
2632 memset(&xecfg, 0, sizeof(xecfg));
2633 xpp.flags = o->xdl_opts;
2634 xecfg.ctxlen = o->context;
2635 xecfg.interhunkctxlen = o->interhunkcontext;
2636 if (xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2637 &xpp, &xecfg))
2638 die("unable to generate diffstat for %s", one->path);
2639 }
2640 diff_free_filespec_data(one);
2642 diff_free_filespec_data(two);
2643}
2644static void builtin_checkdiff(const char *name_a, const char *name_b,
2646 const char *attr_path,
2647 struct diff_filespec *one,
2648 struct diff_filespec *two,
2649 struct diff_options *o)
2650{
2651 mmfile_t mf1, mf2;
2652 struct checkdiff_t data;
2653 if (!two)
2655 return;
2656 memset(&data, 0, sizeof(data));
2658 data.filename = name_b ? name_b : name_a;
2659 data.lineno = 0;
2660 data.o = o;
2661 data.ws_rule = whitespace_rule(attr_path);
2662 data.conflict_marker_size = ll_merge_marker_size(attr_path);
2663 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2665 die("unable to read files to diff");
2666 /*
2668 * All the other codepaths check both sides, but not checking
2669 * the "old" side here is deliberate. We are checking the newly
2670 * introduced changes, and as long as the "new" side is text, we
2671 * can and should check what it introduces.
2672 */
2673 if (diff_filespec_is_binary(two))
2674 goto free_and_return;
2675 else {
2676 /* Crazy xdl interfaces.. */
2677 xpparam_t xpp;
2678 xdemitconf_t xecfg;
2679 memset(&xpp, 0, sizeof(xpp));
2681 memset(&xecfg, 0, sizeof(xecfg));
2682 xecfg.ctxlen = 1; /* at least one context line */
2683 xpp.flags = 0;
2684 if (xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2685 &xpp, &xecfg))
2686 die("unable to generate checkdiff for %s", one->path);
2687 if (data.ws_rule & WS_BLANK_AT_EOF) {
2689 struct emit_callback ecbdata;
2690 int blank_at_eof;
2691 ecbdata.ws_rule = data.ws_rule;
2693 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2694 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2695 if (blank_at_eof) {
2697 static char *err;
2698 if (!err)
2699 err = whitespace_error_string(WS_BLANK_AT_EOF);
2700 fprintf(o->file, "%s:%d: %s.\n",
2701 data.filename, blank_at_eof, err);
2702 data.status = 1; /* report errors */
2703 }
2704 }
2705 }
2706 free_and_return:
2707 diff_free_filespec_data(one);
2708 diff_free_filespec_data(two);
2709 if (data.status)
2710 DIFF_OPT_SET(o, CHECK_FAILED);
2711}
2712struct diff_filespec *alloc_filespec(const char *path)
2714{
2715 struct diff_filespec *spec;
2716 FLEXPTR_ALLOC_STR(spec, path, path);
2718 spec->count = 1;
2719 spec->is_binary = -1;
2720 return spec;
2721}
2722void free_filespec(struct diff_filespec *spec)
2724{
2725 if (!--spec->count) {
2726 diff_free_filespec_data(spec);
2727 free(spec);
2728 }
2729}
2730void fill_filespec(struct diff_filespec *spec, const struct object_id *oid,
2732 int oid_valid, unsigned short mode)
2733{
2734 if (mode) {
2735 spec->mode = canon_mode(mode);
2736 oidcpy(&spec->oid, oid);
2737 spec->oid_valid = oid_valid;
2738 }
2739}
2740/*
2742 * Given a name and sha1 pair, if the index tells us the file in
2743 * the work tree has that object contents, return true, so that
2744 * prepare_temp_file() does not have to inflate and extract.
2745 */
2746static int reuse_worktree_file(const char *name, const struct object_id *oid, int want_file)
2747{
2748 const struct cache_entry *ce;
2749 struct stat st;
2750 int pos, len;
2751 /*
2753 * We do not read the cache ourselves here, because the
2754 * benchmark with my previous version that always reads cache
2755 * shows that it makes things worse for diff-tree comparing
2756 * two linux-2.6 kernel trees in an already checked out work
2757 * tree. This is because most diff-tree comparisons deal with
2758 * only a small number of files, while reading the cache is
2759 * expensive for a large project, and its cost outweighs the
2760 * savings we get by not inflating the object to a temporary
2761 * file. Practically, this code only helps when we are used
2762 * by diff-cache --cached, which does read the cache before
2763 * calling us.
2764 */
2765 if (!active_cache)
2766 return 0;
2767 /* We want to avoid the working directory if our caller
2769 * doesn't need the data in a normal file, this system
2770 * is rather slow with its stat/open/mmap/close syscalls,
2771 * and the object is contained in a pack file. The pack
2772 * is probably already open and will be faster to obtain
2773 * the data through than the working directory. Loose
2774 * objects however would tend to be slower as they need
2775 * to be individually opened and inflated.
2776 */
2777 if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(oid->hash))
2778 return 0;
2779 /*
2781 * Similarly, if we'd have to convert the file contents anyway, that
2782 * makes the optimization not worthwhile.
2783 */
2784 if (!want_file && would_convert_to_git(&the_index, name))
2785 return 0;
2786 len = strlen(name);
2788 pos = cache_name_pos(name, len);
2789 if (pos < 0)
2790 return 0;
2791 ce = active_cache[pos];
2792 /*
2794 * This is not the sha1 we are looking for, or
2795 * unreusable because it is not a regular file.
2796 */
2797 if (oidcmp(oid, &ce->oid) || !S_ISREG(ce->ce_mode))
2798 return 0;
2799 /*
2801 * If ce is marked as "assume unchanged", there is no
2802 * guarantee that work tree matches what we are looking for.
2803 */
2804 if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2805 return 0;
2806 /*
2808 * If ce matches the file in the work tree, we can reuse it.
2809 */
2810 if (ce_uptodate(ce) ||
2811 (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2812 return 1;
2813 return 0;
2815}
2816static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2818{
2819 struct strbuf buf = STRBUF_INIT;
2820 char *dirty = "";
2821 /* Are we looking at the work tree? */
2823 if (s->dirty_submodule)
2824 dirty = "-dirty";
2825 strbuf_addf(&buf, "Subproject commit %s%s\n",
2827 oid_to_hex(&s->oid), dirty);
2828 s->size = buf.len;
2829 if (size_only) {
2830 s->data = NULL;
2831 strbuf_release(&buf);
2832 } else {
2833 s->data = strbuf_detach(&buf, NULL);
2834 s->should_free = 1;
2835 }
2836 return 0;
2837}
2838/*
2840 * While doing rename detection and pickaxe operation, we may need to
2841 * grab the data for the blob (or file) for our own in-core comparison.
2842 * diff_filespec has data and size fields for this purpose.
2843 */
2844int diff_populate_filespec(struct diff_filespec *s, unsigned int flags)
2845{
2846 int size_only = flags & CHECK_SIZE_ONLY;
2847 int err = 0;
2848 /*
2849 * demote FAIL to WARN to allow inspecting the situation
2850 * instead of refusing.
2851 */
2852 enum safe_crlf crlf_warn = (safe_crlf == SAFE_CRLF_FAIL
2853 ? SAFE_CRLF_WARN
2854 : safe_crlf);
2855 if (!DIFF_FILE_VALID(s))
2857 die("internal error: asking to populate invalid file.");
2858 if (S_ISDIR(s->mode))
2859 return -1;
2860 if (s->data)
2862 return 0;
2863 if (size_only && 0 < s->size)
2865 return 0;
2866 if (S_ISGITLINK(s->mode))
2868 return diff_populate_gitlink(s, size_only);
2869 if (!s->oid_valid ||
2871 reuse_worktree_file(s->path, &s->oid, 0)) {
2872 struct strbuf buf = STRBUF_INIT;
2873 struct stat st;
2874 int fd;
2875 if (lstat(s->path, &st) < 0) {
2877 if (errno == ENOENT) {
2878 err_empty:
2879 err = -1;
2880 empty:
2881 s->data = (char *)"";
2882 s->size = 0;
2883 return err;
2884 }
2885 }
2886 s->size = xsize_t(st.st_size);
2887 if (!s->size)
2888 goto empty;
2889 if (S_ISLNK(st.st_mode)) {
2890 struct strbuf sb = STRBUF_INIT;
2891 if (strbuf_readlink(&sb, s->path, s->size))
2893 goto err_empty;
2894 s->size = sb.len;
2895 s->data = strbuf_detach(&sb, NULL);
2896 s->should_free = 1;
2897 return 0;
2898 }
2899 /*
2901 * Even if the caller would be happy with getting
2902 * only the size, we cannot return early at this
2903 * point if the path requires us to run the content
2904 * conversion.
2905 */
2906 if (size_only && !would_convert_to_git(&the_index, s->path))
2907 return 0;
2908 /*
2910 * Note: this check uses xsize_t(st.st_size) that may
2911 * not be the true size of the blob after it goes
2912 * through convert_to_git(). This may not strictly be
2913 * correct, but the whole point of big_file_threshold
2914 * and is_binary check being that we want to avoid
2915 * opening the file and inspecting the contents, this
2916 * is probably fine.
2917 */
2918 if ((flags & CHECK_BINARY) &&
2919 s->size > big_file_threshold && s->is_binary == -1) {
2920 s->is_binary = 1;
2921 return 0;
2922 }
2923 fd = open(s->path, O_RDONLY);
2924 if (fd < 0)
2925 goto err_empty;
2926 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2927 close(fd);
2928 s->should_munmap = 1;
2929 /*
2931 * Convert from working tree format to canonical git format
2932 */
2933 if (convert_to_git(&the_index, s->path, s->data, s->size, &buf, crlf_warn)) {
2934 size_t size = 0;
2935 munmap(s->data, s->size);
2936 s->should_munmap = 0;
2937 s->data = strbuf_detach(&buf, &size);
2938 s->size = size;
2939 s->should_free = 1;
2940 }
2941 }
2942 else {
2943 enum object_type type;
2944 if (size_only || (flags & CHECK_BINARY)) {
2945 type = sha1_object_info(s->oid.hash, &s->size);
2946 if (type < 0)
2947 die("unable to read %s",
2948 oid_to_hex(&s->oid));
2949 if (size_only)
2950 return 0;
2951 if (s->size > big_file_threshold && s->is_binary == -1) {
2952 s->is_binary = 1;
2953 return 0;
2954 }
2955 }
2956 s->data = read_sha1_file(s->oid.hash, &type, &s->size);
2957 if (!s->data)
2958 die("unable to read %s", oid_to_hex(&s->oid));
2959 s->should_free = 1;
2960 }
2961 return 0;
2962}
2963void diff_free_filespec_blob(struct diff_filespec *s)
2965{
2966 if (s->should_free)
2967 free(s->data);
2968 else if (s->should_munmap)
2969 munmap(s->data, s->size);
2970 if (s->should_free || s->should_munmap) {
2972 s->should_free = s->should_munmap = 0;
2973 s->data = NULL;
2974 }
2975}
2976void diff_free_filespec_data(struct diff_filespec *s)
2978{
2979 diff_free_filespec_blob(s);
2980 FREE_AND_NULL(s->cnt_data);
2981}
2982static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2984 void *blob,
2985 unsigned long size,
2986 const struct object_id *oid,
2987 int mode)
2988{
2989 int fd;
2990 struct strbuf buf = STRBUF_INIT;
2991 struct strbuf template = STRBUF_INIT;
2992 char *path_dup = xstrdup(path);
2993 const char *base = basename(path_dup);
2994 /* Generate "XXXXXX_basename.ext" */
2996 strbuf_addstr(&template, "XXXXXX_");
2997 strbuf_addstr(&template, base);
2998 fd = mks_tempfile_ts(&temp->tempfile, template.buf, strlen(base) + 1);
3000 if (fd < 0)
3001 die_errno("unable to create temp-file");
3002 if (convert_to_working_tree(path,
3003 (const char *)blob, (size_t)size, &buf)) {
3004 blob = buf.buf;
3005 size = buf.len;
3006 }
3007 if (write_in_full(fd, blob, size) != size)
3008 die_errno("unable to write temp-file");
3009 close_tempfile(&temp->tempfile);
3010 temp->name = get_tempfile_path(&temp->tempfile);
3011 oid_to_hex_r(temp->hex, oid);
3012 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", mode);
3013 strbuf_release(&buf);
3014 strbuf_release(&template);
3015 free(path_dup);
3016}
3017static struct diff_tempfile *prepare_temp_file(const char *name,
3019 struct diff_filespec *one)
3020{
3021 struct diff_tempfile *temp = claim_diff_tempfile();
3022 if (!DIFF_FILE_VALID(one)) {
3024 not_a_valid_file:
3025 /* A '-' entry produces this for file-2, and
3026 * a '+' entry produces this for file-1.
3027 */
3028 temp->name = "/dev/null";
3029 xsnprintf(temp->hex, sizeof(temp->hex), ".");
3030 xsnprintf(temp->mode, sizeof(temp->mode), ".");
3031 return temp;
3032 }
3033 if (!S_ISGITLINK(one->mode) &&
3035 (!one->oid_valid ||
3036 reuse_worktree_file(name, &one->oid, 1))) {
3037 struct stat st;
3038 if (lstat(name, &st) < 0) {
3039 if (errno == ENOENT)
3040 goto not_a_valid_file;
3041 die_errno("stat(%s)", name);
3042 }
3043 if (S_ISLNK(st.st_mode)) {
3044 struct strbuf sb = STRBUF_INIT;
3045 if (strbuf_readlink(&sb, name, st.st_size) < 0)
3046 die_errno("readlink(%s)", name);
3047 prep_temp_blob(name, temp, sb.buf, sb.len,
3048 (one->oid_valid ?
3049 &one->oid : &null_oid),
3050 (one->oid_valid ?
3051 one->mode : S_IFLNK));
3052 strbuf_release(&sb);
3053 }
3054 else {
3055 /* we can borrow from the file in the work tree */
3056 temp->name = name;
3057 if (!one->oid_valid)
3058 oid_to_hex_r(temp->hex, &null_oid);
3059 else
3060 oid_to_hex_r(temp->hex, &one->oid);
3061 /* Even though we may sometimes borrow the
3062 * contents from the work tree, we always want
3063 * one->mode. mode is trustworthy even when
3064 * !(one->oid_valid), as long as
3065 * DIFF_FILE_VALID(one).
3066 */
3067 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", one->mode);
3068 }
3069 return temp;
3070 }
3071 else {
3072 if (diff_populate_filespec(one, 0))
3073 die("cannot read data blob for %s", one->path);
3074 prep_temp_blob(name, temp, one->data, one->size,
3075 &one->oid, one->mode);
3076 }
3077 return temp;
3078}
3079static void add_external_diff_name(struct argv_array *argv,
3081 const char *name,
3082 struct diff_filespec *df)
3083{
3084 struct diff_tempfile *temp = prepare_temp_file(name, df);
3085 argv_array_push(argv, temp->name);
3086 argv_array_push(argv, temp->hex);
3087 argv_array_push(argv, temp->mode);
3088}
3089/* An external diff command takes:
3091 *
3092 * diff-cmd name infile1 infile1-sha1 infile1-mode \
3093 * infile2 infile2-sha1 infile2-mode [ rename-to ]
3094 *
3095 */
3096static void run_external_diff(const char *pgm,
3097 const char *name,
3098 const char *other,
3099 struct diff_filespec *one,
3100 struct diff_filespec *two,
3101 const char *xfrm_msg,
3102 int complete_rewrite,
3103 struct diff_options *o)
3104{
3105 struct argv_array argv = ARGV_ARRAY_INIT;
3106 struct argv_array env = ARGV_ARRAY_INIT;
3107 struct diff_queue_struct *q = &diff_queued_diff;
3108 argv_array_push(&argv, pgm);
3110 argv_array_push(&argv, name);
3111 if (one && two) {
3113 add_external_diff_name(&argv, name, one);
3114 if (!other)
3115 add_external_diff_name(&argv, name, two);
3116 else {
3117 add_external_diff_name(&argv, other, two);
3118 argv_array_push(&argv, other);
3119 argv_array_push(&argv, xfrm_msg);
3120 }
3121 }
3122 argv_array_pushf(&env, "GIT_DIFF_PATH_COUNTER=%d", ++o->diff_path_counter);
3124 argv_array_pushf(&env, "GIT_DIFF_PATH_TOTAL=%d", q->nr);
3125 if (run_command_v_opt_cd_env(argv.argv, RUN_USING_SHELL, NULL, env.argv))
3127 die(_("external diff died, stopping at %s"), name);
3128 remove_tempfile();
3130 argv_array_clear(&argv);
3131 argv_array_clear(&env);
3132}
3133static int similarity_index(struct diff_filepair *p)
3135{
3136 return p->score * 100 / MAX_SCORE;
3137}
3138static const char *diff_abbrev_oid(const struct object_id *oid, int abbrev)
3140{
3141 if (startup_info->have_repository)
3142 return find_unique_abbrev(oid->hash, abbrev);
3143 else {
3144 char *hex = oid_to_hex(oid);
3145 if (abbrev < 0)
3146 abbrev = FALLBACK_DEFAULT_ABBREV;
3147 if (abbrev > GIT_SHA1_HEXSZ)
3148 die("BUG: oid abbreviation out of range: %d", abbrev);
3149 if (abbrev)
3150 hex[abbrev] = '\0';
3151 return hex;
3152 }
3153}
3154static void fill_metainfo(struct strbuf *msg,
3156 const char *name,
3157 const char *other,
3158 struct diff_filespec *one,
3159 struct diff_filespec *two,
3160 struct diff_options *o,
3161 struct diff_filepair *p,
3162 int *must_show_header,
3163 int use_color)
3164{
3165 const char *set = diff_get_color(use_color, DIFF_METAINFO);
3166 const char *reset = diff_get_color(use_color, DIFF_RESET);
3167 const char *line_prefix = diff_line_prefix(o);
3168 *must_show_header = 1;
3170 strbuf_init(msg, PATH_MAX * 2 + 300);
3171 switch (p->status) {
3172 case DIFF_STATUS_COPIED:
3173 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3174 line_prefix, set, similarity_index(p));
3175 strbuf_addf(msg, "%s\n%s%scopy from ",
3176 reset, line_prefix, set);
3177 quote_c_style(name, msg, NULL, 0);
3178 strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
3179 quote_c_style(other, msg, NULL, 0);
3180 strbuf_addf(msg, "%s\n", reset);
3181 break;
3182 case DIFF_STATUS_RENAMED:
3183 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3184 line_prefix, set, similarity_index(p));
3185 strbuf_addf(msg, "%s\n%s%srename from ",
3186 reset, line_prefix, set);
3187 quote_c_style(name, msg, NULL, 0);
3188 strbuf_addf(msg, "%s\n%s%srename to ",
3189 reset, line_prefix, set);
3190 quote_c_style(other, msg, NULL, 0);
3191 strbuf_addf(msg, "%s\n", reset);
3192 break;
3193 case DIFF_STATUS_MODIFIED:
3194 if (p->score) {
3195 strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
3196 line_prefix,
3197 set, similarity_index(p), reset);
3198 break;
3199 }
3200 /* fallthru */
3201 default:
3202 *must_show_header = 0;
3203 }
3204 if (one && two && oidcmp(&one->oid, &two->oid)) {
3205 int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
3206 if (DIFF_OPT_TST(o, BINARY)) {
3208 mmfile_t mf;
3209 if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
3210 (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
3211 abbrev = 40;
3212 }
3213 strbuf_addf(msg, "%s%sindex %s..%s", line_prefix, set,
3214 diff_abbrev_oid(&one->oid, abbrev),
3215 diff_abbrev_oid(&two->oid, abbrev));
3216 if (one->mode == two->mode)
3217 strbuf_addf(msg, " %06o", one->mode);
3218 strbuf_addf(msg, "%s\n", reset);
3219 }
3220}
3221static void run_diff_cmd(const char *pgm,
3223 const char *name,
3224 const char *other,
3225 const char *attr_path,
3226 struct diff_filespec *one,
3227 struct diff_filespec *two,
3228 struct strbuf *msg,
3229 struct diff_options *o,
3230 struct diff_filepair *p)
3231{
3232 const char *xfrm_msg = NULL;
3233 int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
3234 int must_show_header = 0;
3235 if (DIFF_OPT_TST(o, ALLOW_EXTERNAL)) {
3238 struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
3239 if (drv && drv->external)
3240 pgm = drv->external;
3241 }
3242 if (msg) {
3244 /*
3245 * don't use colors when the header is intended for an
3246 * external diff driver
3247 */
3248 fill_metainfo(msg, name, other, one, two, o, p,
3249 &must_show_header,
3250 want_color(o->use_color) && !pgm);
3251 xfrm_msg = msg->len ? msg->buf : NULL;
3252 }
3253 if (pgm) {
3255 run_external_diff(pgm, name, other, one, two, xfrm_msg,
3256 complete_rewrite, o);
3257 return;
3258 }
3259 if (one && two)
3260 builtin_diff(name, other ? other : name,
3261 one, two, xfrm_msg, must_show_header,
3262 o, complete_rewrite);
3263 else
3264 fprintf(o->file, "* Unmerged path %s\n", name);
3265}
3266static void diff_fill_oid_info(struct diff_filespec *one)
3268{
3269 if (DIFF_FILE_VALID(one)) {
3270 if (!one->oid_valid) {
3271 struct stat st;
3272 if (one->is_stdin) {
3273 oidclr(&one->oid);
3274 return;
3275 }
3276 if (lstat(one->path, &st) < 0)
3277 die_errno("stat '%s'", one->path);
3278 if (index_path(one->oid.hash, one->path, &st, 0))
3279 die("cannot hash %s", one->path);
3280 }
3281 }
3282 else
3283 oidclr(&one->oid);
3284}
3285static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
3287{
3288 /* Strip the prefix but do not molest /dev/null and absolute paths */
3289 if (*namep && **namep != '/') {
3290 *namep += prefix_length;
3291 if (**namep == '/')
3292 ++*namep;
3293 }
3294 if (*otherp && **otherp != '/') {
3295 *otherp += prefix_length;
3296 if (**otherp == '/')
3297 ++*otherp;
3298 }
3299}
3300static void run_diff(struct diff_filepair *p, struct diff_options *o)
3302{
3303 const char *pgm = external_diff();
3304 struct strbuf msg;
3305 struct diff_filespec *one = p->one;
3306 struct diff_filespec *two = p->two;
3307 const char *name;
3308 const char *other;
3309 const char *attr_path;
3310 name = one->path;
3312 other = (strcmp(name, two->path) ? two->path : NULL);
3313 attr_path = name;
3314 if (o->prefix_length)
3315 strip_prefix(o->prefix_length, &name, &other);
3316 if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
3318 pgm = NULL;
3319 if (DIFF_PAIR_UNMERGED(p)) {
3321 run_diff_cmd(pgm, name, NULL, attr_path,
3322 NULL, NULL, NULL, o, p);
3323 return;
3324 }
3325 diff_fill_oid_info(one);
3327 diff_fill_oid_info(two);
3328 if (!pgm &&
3330 DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3331 (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3332 /*
3333 * a filepair that changes between file and symlink
3334 * needs to be split into deletion and creation.
3335 */
3336 struct diff_filespec *null = alloc_filespec(two->path);
3337 run_diff_cmd(NULL, name, other, attr_path,
3338 one, null, &msg, o, p);
3339 free(null);
3340 strbuf_release(&msg);
3341 null = alloc_filespec(one->path);
3343 run_diff_cmd(NULL, name, other, attr_path,
3344 null, two, &msg, o, p);
3345 free(null);
3346 }
3347 else
3348 run_diff_cmd(pgm, name, other, attr_path,
3349 one, two, &msg, o, p);
3350 strbuf_release(&msg);
3352}
3353static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3355 struct diffstat_t *diffstat)
3356{
3357 const char *name;
3358 const char *other;
3359 if (DIFF_PAIR_UNMERGED(p)) {
3361 /* unmerged */
3362 builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, p);
3363 return;
3364 }
3365 name = p->one->path;
3367 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3368 if (o->prefix_length)
3370 strip_prefix(o->prefix_length, &name, &other);
3371 diff_fill_oid_info(p->one);
3373 diff_fill_oid_info(p->two);
3374 builtin_diffstat(name, other, p->one, p->two, diffstat, o, p);
3376}
3377static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3379{
3380 const char *name;
3381 const char *other;
3382 const char *attr_path;
3383 if (DIFF_PAIR_UNMERGED(p)) {
3385 /* unmerged */
3386 return;
3387 }
3388 name = p->one->path;
3390 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3391 attr_path = other ? other : name;
3392 if (o->prefix_length)
3394 strip_prefix(o->prefix_length, &name, &other);
3395 diff_fill_oid_info(p->one);
3397 diff_fill_oid_info(p->two);
3398 builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3400}
3401void diff_setup(struct diff_options *options)
3403{
3404 memcpy(options, &default_diff_options, sizeof(*options));
3405 options->file = stdout;
3407 options->abbrev = DEFAULT_ABBREV;
3409 options->line_termination = '\n';
3410 options->break_opt = -1;
3411 options->rename_limit = -1;
3412 options->dirstat_permille = diff_dirstat_permille_default;
3413 options->context = diff_context_default;
3414 options->interhunkcontext = diff_interhunk_context_default;
3415 options->ws_error_highlight = ws_error_highlight_default;
3416 DIFF_OPT_SET(options, RENAME_EMPTY);
3417 /* pathchange left =NULL by default */
3419 options->change = diff_change;
3420 options->add_remove = diff_addremove;
3421 options->use_color = diff_use_color_default;
3422 options->detect_rename = diff_detect_rename_default;
3423 options->xdl_opts |= diff_algorithm;
3424 if (diff_indent_heuristic)
3425 DIFF_XDL_SET(options, INDENT_HEURISTIC);
3426 options->orderfile = diff_order_file_cfg;
3428 if (diff_no_prefix) {
3430 options->a_prefix = options->b_prefix = "";
3431 } else if (!diff_mnemonic_prefix) {
3432 options->a_prefix = "a/";
3433 options->b_prefix = "b/";
3434 }
3435}
3436void diff_setup_done(struct diff_options *options)
3438{
3439 int count = 0;
3440 if (options->set_default)
3442 options->set_default(options);
3443 if (options->output_format & DIFF_FORMAT_NAME)
3445 count++;
3446 if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3447 count++;
3448 if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3449 count++;
3450 if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3451 count++;
3452 if (count > 1)
3453 die(_("--name-only, --name-status, --check and -s are mutually exclusive"));
3454 /*
3456 * Most of the time we can say "there are changes"
3457 * only by checking if there are changed paths, but
3458 * --ignore-whitespace* options force us to look
3459 * inside contents.
3460 */
3461 if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3463 DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3464 DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3465 DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3466 else
3467 DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3468 if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3470 options->detect_rename = DIFF_DETECT_COPY;
3471 if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3473 options->prefix = NULL;
3474 if (options->prefix)
3475 options->prefix_length = strlen(options->prefix);
3476 else
3477 options->prefix_length = 0;
3478 if (options->output_format & (DIFF_FORMAT_NAME |
3480 DIFF_FORMAT_NAME_STATUS |
3481 DIFF_FORMAT_CHECKDIFF |
3482 DIFF_FORMAT_NO_OUTPUT))
3483 options->output_format &= ~(DIFF_FORMAT_RAW |
3484 DIFF_FORMAT_NUMSTAT |
3485 DIFF_FORMAT_DIFFSTAT |
3486 DIFF_FORMAT_SHORTSTAT |
3487 DIFF_FORMAT_DIRSTAT |
3488 DIFF_FORMAT_SUMMARY |
3489 DIFF_FORMAT_PATCH);
3490 /*
3492 * These cases always need recursive; we do not drop caller-supplied
3493 * recursive bits for other formats here.
3494 */
3495 if (options->output_format & (DIFF_FORMAT_PATCH |
3496 DIFF_FORMAT_NUMSTAT |
3497 DIFF_FORMAT_DIFFSTAT |
3498 DIFF_FORMAT_SHORTSTAT |
3499 DIFF_FORMAT_DIRSTAT |
3500 DIFF_FORMAT_SUMMARY |
3501 DIFF_FORMAT_CHECKDIFF))
3502 DIFF_OPT_SET(options, RECURSIVE);
3503 /*
3504 * Also pickaxe would not work very well if you do not say recursive
3505 */
3506 if (options->pickaxe)
3507 DIFF_OPT_SET(options, RECURSIVE);
3508 /*
3509 * When patches are generated, submodules diffed against the work tree
3510 * must be checked for dirtiness too so it can be shown in the output
3511 */
3512 if (options->output_format & DIFF_FORMAT_PATCH)
3513 DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3514 if (options->detect_rename && options->rename_limit < 0)
3516 options->rename_limit = diff_rename_limit_default;
3517 if (options->setup & DIFF_SETUP_USE_CACHE) {
3518 if (!active_cache)
3519 /* read-cache does not die even when it fails
3520 * so it is safe for us to do this here. Also
3521 * it does not smudge active_cache or active_nr
3522 * when it fails, so we do not have to worry about
3523 * cleaning it up ourselves either.
3524 */
3525 read_cache();
3526 }
3527 if (40 < options->abbrev)
3528 options->abbrev = 40; /* full */
3529 /*
3531 * It does not make sense to show the first hit we happened
3532 * to have found. It does not make sense not to return with
3533 * exit code in such a case either.
3534 */
3535 if (DIFF_OPT_TST(options, QUICK)) {
3536 options->output_format = DIFF_FORMAT_NO_OUTPUT;
3537 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3538 }
3539 options->diff_path_counter = 0;
3541 if (DIFF_OPT_TST(options, FOLLOW_RENAMES) && options->pathspec.nr != 1)
3543 die(_("--follow requires exactly one pathspec"));
3544}
3545static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3547{
3548 char c, *eq;
3549 int len;
3550 if (*arg != '-')
3552 return 0;
3553 c = *++arg;
3554 if (!c)
3555 return 0;
3556 if (c == arg_short) {
3557 c = *++arg;
3558 if (!c)
3559 return 1;
3560 if (val && isdigit(c)) {
3561 char *end;
3562 int n = strtoul(arg, &end, 10);
3563 if (*end)
3564 return 0;
3565 *val = n;
3566 return 1;
3567 }
3568 return 0;
3569 }
3570 if (c != '-')
3571 return 0;
3572 arg++;
3573 eq = strchrnul(arg, '=');
3574 len = eq - arg;
3575 if (!len || strncmp(arg, arg_long, len))
3576 return 0;
3577 if (*eq) {
3578 int n;
3579 char *end;
3580 if (!isdigit(*++eq))
3581 return 0;
3582 n = strtoul(eq, &end, 10);
3583 if (*end)
3584 return 0;
3585 *val = n;
3586 }
3587 return 1;
3588}
3589static int diff_scoreopt_parse(const char *opt);
3591static inline int short_opt(char opt, const char **argv,
3593 const char **optarg)
3594{
3595 const char *arg = argv[0];
3596 if (arg[0] != '-' || arg[1] != opt)
3597 return 0;
3598 if (arg[2] != '\0') {
3599 *optarg = arg + 2;
3600 return 1;
3601 }
3602 if (!argv[1])
3603 die("Option '%c' requires a value", opt);
3604 *optarg = argv[1];
3605 return 2;
3606}
3607int parse_long_opt(const char *opt, const char **argv,
3609 const char **optarg)
3610{
3611 const char *arg = argv[0];
3612 if (!skip_prefix(arg, "--", &arg))
3613 return 0;
3614 if (!skip_prefix(arg, opt, &arg))
3615 return 0;
3616 if (*arg == '=') { /* stuck form: --option=value */
3617 *optarg = arg + 1;
3618 return 1;
3619 }
3620 if (*arg != '\0')
3621 return 0;
3622 /* separate form: --option value */
3623 if (!argv[1])
3624 die("Option '--%s' requires a value", opt);
3625 *optarg = argv[1];
3626 return 2;
3627}
3628static int stat_opt(struct diff_options *options, const char **av)
3630{
3631 const char *arg = av[0];
3632 char *end;
3633 int width = options->stat_width;
3634 int name_width = options->stat_name_width;
3635 int graph_width = options->stat_graph_width;
3636 int count = options->stat_count;
3637 int argcount = 1;
3638 if (!skip_prefix(arg, "--stat", &arg))
3640 die("BUG: stat option does not begin with --stat: %s", arg);
3641 end = (char *)arg;
3642 switch (*arg) {
3644 case '-':
3645 if (skip_prefix(arg, "-width", &arg)) {
3646 if (*arg == '=')
3647 width = strtoul(arg + 1, &end, 10);
3648 else if (!*arg && !av[1])
3649 die_want_option("--stat-width");
3650 else if (!*arg) {
3651 width = strtoul(av[1], &end, 10);
3652 argcount = 2;
3653 }
3654 } else if (skip_prefix(arg, "-name-width", &arg)) {
3655 if (*arg == '=')
3656 name_width = strtoul(arg + 1, &end, 10);
3657 else if (!*arg && !av[1])
3658 die_want_option("--stat-name-width");
3659 else if (!*arg) {
3660 name_width = strtoul(av[1], &end, 10);
3661 argcount = 2;
3662 }
3663 } else if (skip_prefix(arg, "-graph-width", &arg)) {
3664 if (*arg == '=')
3665 graph_width = strtoul(arg + 1, &end, 10);
3666 else if (!*arg && !av[1])
3667 die_want_option("--stat-graph-width");
3668 else if (!*arg) {
3669 graph_width = strtoul(av[1], &end, 10);
3670 argcount = 2;
3671 }
3672 } else if (skip_prefix(arg, "-count", &arg)) {
3673 if (*arg == '=')
3674 count = strtoul(arg + 1, &end, 10);
3675 else if (!*arg && !av[1])
3676 die_want_option("--stat-count");
3677 else if (!*arg) {
3678 count = strtoul(av[1], &end, 10);
3679 argcount = 2;
3680 }
3681 }
3682 break;
3683 case '=':
3684 width = strtoul(arg+1, &end, 10);
3685 if (*end == ',')
3686 name_width = strtoul(end+1, &end, 10);
3687 if (*end == ',')
3688 count = strtoul(end+1, &end, 10);
3689 }
3690 /* Important! This checks all the error cases! */
3692 if (*end)
3693 return 0;
3694 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3695 options->stat_name_width = name_width;
3696 options->stat_graph_width = graph_width;
3697 options->stat_width = width;
3698 options->stat_count = count;
3699 return argcount;
3700}
3701static int parse_dirstat_opt(struct diff_options *options, const char *params)
3703{
3704 struct strbuf errmsg = STRBUF_INIT;
3705 if (parse_dirstat_params(options, params, &errmsg))
3706 die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3707 errmsg.buf);
3708 strbuf_release(&errmsg);
3709 /*
3710 * The caller knows a dirstat-related option is given from the command
3711 * line; allow it to say "return this_function();"
3712 */
3713 options->output_format |= DIFF_FORMAT_DIRSTAT;
3714 return 1;
3715}
3716static int parse_submodule_opt(struct diff_options *options, const char *value)
3718{
3719 if (parse_submodule_params(options, value))
3720 die(_("Failed to parse --submodule option parameter: '%s'"),
3721 value);
3722 return 1;
3723}
3724static const char diff_status_letters[] = {
3726 DIFF_STATUS_ADDED,
3727 DIFF_STATUS_COPIED,
3728 DIFF_STATUS_DELETED,
3729 DIFF_STATUS_MODIFIED,
3730 DIFF_STATUS_RENAMED,
3731 DIFF_STATUS_TYPE_CHANGED,
3732 DIFF_STATUS_UNKNOWN,
3733 DIFF_STATUS_UNMERGED,
3734 DIFF_STATUS_FILTER_AON,
3735 DIFF_STATUS_FILTER_BROKEN,
3736 '\0',
3737};
3738static unsigned int filter_bit['Z' + 1];
3740static void prepare_filter_bits(void)
3742{
3743 int i;
3744 if (!filter_bit[DIFF_STATUS_ADDED]) {
3746 for (i = 0; diff_status_letters[i]; i++)
3747 filter_bit[(int) diff_status_letters[i]] = (1 << i);
3748 }
3749}
3750static unsigned filter_bit_tst(char status, const struct diff_options *opt)
3752{
3753 return opt->filter & filter_bit[(int) status];
3754}
3755static int parse_diff_filter_opt(const char *optarg, struct diff_options *opt)
3757{
3758 int i, optch;
3759 prepare_filter_bits();
3761 /*
3763 * If there is a negation e.g. 'd' in the input, and we haven't
3764 * initialized the filter field with another --diff-filter, start
3765 * from full set of bits, except for AON.
3766 */
3767 if (!opt->filter) {
3768 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3769 if (optch < 'a' || 'z' < optch)
3770 continue;
3771 opt->filter = (1 << (ARRAY_SIZE(diff_status_letters) - 1)) - 1;
3772 opt->filter &= ~filter_bit[DIFF_STATUS_FILTER_AON];
3773 break;
3774 }
3775 }
3776 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3778 unsigned int bit;
3779 int negate;
3780 if ('a' <= optch && optch <= 'z') {
3782 negate = 1;
3783 optch = toupper(optch);
3784 } else {
3785 negate = 0;
3786 }
3787 bit = (0 <= optch && optch <= 'Z') ? filter_bit[optch] : 0;
3789 if (!bit)
3790 return optarg[i];
3791 if (negate)
3792 opt->filter &= ~bit;
3793 else
3794 opt->filter |= bit;
3795 }
3796 return 0;
3797}
3798static void enable_patch_output(int *fmt) {
3800 *fmt &= ~DIFF_FORMAT_NO_OUTPUT;
3801 *fmt |= DIFF_FORMAT_PATCH;
3802}
3803static int parse_ws_error_highlight_opt(struct diff_options *opt, const char *arg)
3805{
3806 int val = parse_ws_error_highlight(arg);
3807 if (val < 0) {
3809 error("unknown value after ws-error-highlight=%.*s",
3810 -1 - val, arg);
3811 return 0;
3812 }
3813 opt->ws_error_highlight = val;
3814 return 1;
3815}
3816int diff_opt_parse(struct diff_options *options,
3818 const char **av, int ac, const char *prefix)
3819{
3820 const char *arg = av[0];
3821 const char *optarg;
3822 int argcount;
3823 if (!prefix)
3825 prefix = "";
3826 /* Output format options */
3828 if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch")
3829 || opt_arg(arg, 'U', "unified", &options->context))
3830 enable_patch_output(&options->output_format);
3831 else if (!strcmp(arg, "--raw"))
3832 options->output_format |= DIFF_FORMAT_RAW;
3833 else if (!strcmp(arg, "--patch-with-raw")) {
3834 enable_patch_output(&options->output_format);
3835 options->output_format |= DIFF_FORMAT_RAW;
3836 } else if (!strcmp(arg, "--numstat"))
3837 options->output_format |= DIFF_FORMAT_NUMSTAT;
3838 else if (!strcmp(arg, "--shortstat"))
3839 options->output_format |= DIFF_FORMAT_SHORTSTAT;
3840 else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3841 return parse_dirstat_opt(options, "");
3842 else if (skip_prefix(arg, "-X", &arg))
3843 return parse_dirstat_opt(options, arg);
3844 else if (skip_prefix(arg, "--dirstat=", &arg))
3845 return parse_dirstat_opt(options, arg);
3846 else if (!strcmp(arg, "--cumulative"))
3847 return parse_dirstat_opt(options, "cumulative");
3848 else if (!strcmp(arg, "--dirstat-by-file"))
3849 return parse_dirstat_opt(options, "files");
3850 else if (skip_prefix(arg, "--dirstat-by-file=", &arg)) {
3851 parse_dirstat_opt(options, "files");
3852 return parse_dirstat_opt(options, arg);
3853 }
3854 else if (!strcmp(arg, "--check"))
3855 options->output_format |= DIFF_FORMAT_CHECKDIFF;
3856 else if (!strcmp(arg, "--summary"))
3857 options->output_format |= DIFF_FORMAT_SUMMARY;
3858 else if (!strcmp(arg, "--patch-with-stat")) {
3859 enable_patch_output(&options->output_format);
3860 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3861 } else if (!strcmp(arg, "--name-only"))
3862 options->output_format |= DIFF_FORMAT_NAME;
3863 else if (!strcmp(arg, "--name-status"))
3864 options->output_format |= DIFF_FORMAT_NAME_STATUS;
3865 else if (!strcmp(arg, "-s") || !strcmp(arg, "--no-patch"))
3866 options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3867 else if (starts_with(arg, "--stat"))
3868 /* --stat, --stat-width, --stat-name-width, or --stat-count */
3869 return stat_opt(options, av);
3870 /* renames options */
3872 else if (starts_with(arg, "-B") || starts_with(arg, "--break-rewrites=") ||
3873 !strcmp(arg, "--break-rewrites")) {
3874 if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3875 return error("invalid argument to -B: %s", arg+2);
3876 }
3877 else if (starts_with(arg, "-M") || starts_with(arg, "--find-renames=") ||
3878 !strcmp(arg, "--find-renames")) {
3879 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3880 return error("invalid argument to -M: %s", arg+2);
3881 options->detect_rename = DIFF_DETECT_RENAME;
3882 }
3883 else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3884 options->irreversible_delete = 1;
3885 }
3886 else if (starts_with(arg, "-C") || starts_with(arg, "--find-copies=") ||
3887 !strcmp(arg, "--find-copies")) {
3888 if (options->detect_rename == DIFF_DETECT_COPY)
3889 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3890 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3891 return error("invalid argument to -C: %s", arg+2);
3892 options->detect_rename = DIFF_DETECT_COPY;
3893 }
3894 else if (!strcmp(arg, "--no-renames"))
3895 options->detect_rename = 0;
3896 else if (!strcmp(arg, "--rename-empty"))
3897 DIFF_OPT_SET(options, RENAME_EMPTY);
3898 else if (!strcmp(arg, "--no-rename-empty"))
3899 DIFF_OPT_CLR(options, RENAME_EMPTY);
3900 else if (!strcmp(arg, "--relative"))
3901 DIFF_OPT_SET(options, RELATIVE_NAME);
3902 else if (skip_prefix(arg, "--relative=", &arg)) {
3903 DIFF_OPT_SET(options, RELATIVE_NAME);
3904 options->prefix = arg;
3905 }
3906 /* xdiff options */
3908 else if (!strcmp(arg, "--minimal"))
3909 DIFF_XDL_SET(options, NEED_MINIMAL);
3910 else if (!strcmp(arg, "--no-minimal"))
3911 DIFF_XDL_CLR(options, NEED_MINIMAL);
3912 else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3913 DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3914 else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3915 DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3916 else if (!strcmp(arg, "--ignore-space-at-eol"))
3917 DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3918 else if (!strcmp(arg, "--ignore-blank-lines"))
3919 DIFF_XDL_SET(options, IGNORE_BLANK_LINES);
3920 else if (!strcmp(arg, "--indent-heuristic"))
3921 DIFF_XDL_SET(options, INDENT_HEURISTIC);
3922 else if (!strcmp(arg, "--no-indent-heuristic"))
3923 DIFF_XDL_CLR(options, INDENT_HEURISTIC);
3924 else if (!strcmp(arg, "--patience"))
3925 options->xdl_opts = DIFF_WITH_ALG(options, PATIENCE_DIFF);
3926 else if (!strcmp(arg, "--histogram"))
3927 options->xdl_opts = DIFF_WITH_ALG(options, HISTOGRAM_DIFF);
3928 else if ((argcount = parse_long_opt("diff-algorithm", av, &optarg))) {
3929 long value = parse_algorithm_value(optarg);
3930 if (value < 0)
3931 return error("option diff-algorithm accepts \"myers\", "
3932 "\"minimal\", \"patience\" and \"histogram\"");
3933 /* clear out previous settings */
3934 DIFF_XDL_CLR(options, NEED_MINIMAL);
3935 options->xdl_opts &= ~XDF_DIFF_ALGORITHM_MASK;
3936 options->xdl_opts |= value;
3937 return argcount;
3938 }
3939 /* flags options */
3941 else if (!strcmp(arg, "--binary")) {
3942 enable_patch_output(&options->output_format);
3943 DIFF_OPT_SET(options, BINARY);
3944 }
3945 else if (!strcmp(arg, "--full-index"))
3946 DIFF_OPT_SET(options, FULL_INDEX);
3947 else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
3948 DIFF_OPT_SET(options, TEXT);
3949 else if (!strcmp(arg, "-R"))
3950 DIFF_OPT_SET(options, REVERSE_DIFF);
3951 else if (!strcmp(arg, "--find-copies-harder"))
3952 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3953 else if (!strcmp(arg, "--follow"))
3954 DIFF_OPT_SET(options, FOLLOW_RENAMES);
3955 else if (!strcmp(arg, "--no-follow")) {
3956 DIFF_OPT_CLR(options, FOLLOW_RENAMES);
3957 DIFF_OPT_CLR(options, DEFAULT_FOLLOW_RENAMES);
3958 } else if (!strcmp(arg, "--color"))
3959 options->use_color = 1;
3960 else if (skip_prefix(arg, "--color=", &arg)) {
3961 int value = git_config_colorbool(NULL, arg);
3962 if (value < 0)
3963 return error("option `color' expects \"always\", \"auto\", or \"never\"");
3964 options->use_color = value;
3965 }
3966 else if (!strcmp(arg, "--no-color"))
3967 options->use_color = 0;
3968 else if (!strcmp(arg, "--color-words")) {
3969 options->use_color = 1;
3970 options->word_diff = DIFF_WORDS_COLOR;
3971 }
3972 else if (skip_prefix(arg, "--color-words=", &arg)) {
3973 options->use_color = 1;
3974 options->word_diff = DIFF_WORDS_COLOR;
3975 options->word_regex = arg;
3976 }
3977 else if (!strcmp(arg, "--word-diff")) {
3978 if (options->word_diff == DIFF_WORDS_NONE)
3979 options->word_diff = DIFF_WORDS_PLAIN;
3980 }
3981 else if (skip_prefix(arg, "--word-diff=", &arg)) {
3982 if (!strcmp(arg, "plain"))
3983 options->word_diff = DIFF_WORDS_PLAIN;
3984 else if (!strcmp(arg, "color")) {
3985 options->use_color = 1;
3986 options->word_diff = DIFF_WORDS_COLOR;
3987 }
3988 else if (!strcmp(arg, "porcelain"))
3989 options->word_diff = DIFF_WORDS_PORCELAIN;
3990 else if (!strcmp(arg, "none"))
3991 options->word_diff = DIFF_WORDS_NONE;
3992 else
3993 die("bad --word-diff argument: %s", arg);
3994 }
3995 else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
3996 if (options->word_diff == DIFF_WORDS_NONE)
3997 options->word_diff = DIFF_WORDS_PLAIN;
3998 options->word_regex = optarg;
3999 return argcount;
4000 }
4001 else if (!strcmp(arg, "--exit-code"))
4002 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
4003 else if (!strcmp(arg, "--quiet"))
4004 DIFF_OPT_SET(options, QUICK);
4005 else if (!strcmp(arg, "--ext-diff"))
4006 DIFF_OPT_SET(options, ALLOW_EXTERNAL);
4007 else if (!strcmp(arg, "--no-ext-diff"))
4008 DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
4009 else if (!strcmp(arg, "--textconv"))
4010 DIFF_OPT_SET(options, ALLOW_TEXTCONV);
4011 else if (!strcmp(arg, "--no-textconv"))
4012 DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
4013 else if (!strcmp(arg, "--ignore-submodules")) {
4014 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
4015 handle_ignore_submodules_arg(options, "all");
4016 } else if (skip_prefix(arg, "--ignore-submodules=", &arg)) {
4017 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
4018 handle_ignore_submodules_arg(options, arg);
4019 } else if (!strcmp(arg, "--submodule"))
4020 options->submodule_format = DIFF_SUBMODULE_LOG;
4021 else if (skip_prefix(arg, "--submodule=", &arg))
4022 return parse_submodule_opt(options, arg);
4023 else if (skip_prefix(arg, "--ws-error-highlight=", &arg))
4024 return parse_ws_error_highlight_opt(options, arg);
4025 else if (!strcmp(arg, "--ita-invisible-in-index"))
4026 options->ita_invisible_in_index = 1;
4027 else if (!strcmp(arg, "--ita-visible-in-index"))
4028 options->ita_invisible_in_index = 0;
4029 /* misc options */
4031 else if (!strcmp(arg, "-z"))
4032 options->line_termination = 0;
4033 else if ((argcount = short_opt('l', av, &optarg))) {
4034 options->rename_limit = strtoul(optarg, NULL, 10);
4035 return argcount;
4036 }
4037 else if ((argcount = short_opt('S', av, &optarg))) {
4038 options->pickaxe = optarg;
4039 options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
4040 return argcount;
4041 } else if ((argcount = short_opt('G', av, &optarg))) {
4042 options->pickaxe = optarg;
4043 options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
4044 return argcount;
4045 }
4046 else if (!strcmp(arg, "--pickaxe-all"))
4047 options->pickaxe_opts |= DIFF_PICKAXE_ALL;
4048 else if (!strcmp(arg, "--pickaxe-regex"))
4049 options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
4050 else if ((argcount = short_opt('O', av, &optarg))) {
4051 options->orderfile = prefix_filename(prefix, optarg);
4052 return argcount;
4053 }
4054 else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
4055 int offending = parse_diff_filter_opt(optarg, options);
4056 if (offending)
4057 die("unknown change class '%c' in --diff-filter=%s",
4058 offending, optarg);
4059 return argcount;
4060 }
4061 else if (!strcmp(arg, "--no-abbrev"))
4062 options->abbrev = 0;
4063 else if (!strcmp(arg, "--abbrev"))
4064 options->abbrev = DEFAULT_ABBREV;
4065 else if (skip_prefix(arg, "--abbrev=", &arg)) {
4066 options->abbrev = strtoul(arg, NULL, 10);
4067 if (options->abbrev < MINIMUM_ABBREV)
4068 options->abbrev = MINIMUM_ABBREV;
4069 else if (40 < options->abbrev)
4070 options->abbrev = 40;
4071 }
4072 else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
4073 options->a_prefix = optarg;
4074 return argcount;
4075 }
4076 else if ((argcount = parse_long_opt("line-prefix", av, &optarg))) {
4077 options->line_prefix = optarg;
4078 options->line_prefix_length = strlen(options->line_prefix);
4079 graph_setup_line_prefix(options);
4080 return argcount;
4081 }
4082 else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
4083 options->b_prefix = optarg;
4084 return argcount;
4085 }
4086 else if (!strcmp(arg, "--no-prefix"))
4087 options->a_prefix = options->b_prefix = "";
4088 else if (opt_arg(arg, '\0', "inter-hunk-context",
4089 &options->interhunkcontext))
4090 ;
4091 else if (!strcmp(arg, "-W"))
4092 DIFF_OPT_SET(options, FUNCCONTEXT);
4093 else if (!strcmp(arg, "--function-context"))
4094 DIFF_OPT_SET(options, FUNCCONTEXT);
4095 else if (!strcmp(arg, "--no-function-context"))
4096 DIFF_OPT_CLR(options, FUNCCONTEXT);
4097 else if ((argcount = parse_long_opt("output", av, &optarg))) {
4098 char *path = prefix_filename(prefix, optarg);
4099 options->file = xfopen(path, "w");
4100 options->close_file = 1;
4101 if (options->use_color != GIT_COLOR_ALWAYS)
4102 options->use_color = GIT_COLOR_NEVER;
4103 free(path);
4104 return argcount;
4105 } else
4106 return 0;
4107 return 1;
4108}
4109int parse_rename_score(const char **cp_p)
4111{
4112 unsigned long num, scale;
4113 int ch, dot;
4114 const char *cp = *cp_p;
4115 num = 0;
4117 scale = 1;
4118 dot = 0;
4119 for (;;) {
4120 ch = *cp;
4121 if ( !dot && ch == '.' ) {
4122 scale = 1;
4123 dot = 1;
4124 } else if ( ch == '%' ) {
4125 scale = dot ? scale*100 : 100;
4126 cp++; /* % is always at the end */
4127 break;
4128 } else if ( ch >= '0' && ch <= '9' ) {
4129 if ( scale < 100000 ) {
4130 scale *= 10;
4131 num = (num*10) + (ch-'0');
4132 }
4133 } else {
4134 break;
4135 }
4136 cp++;
4137 }
4138 *cp_p = cp;
4139 /* user says num divided by scale and we say internally that
4141 * is MAX_SCORE * num / scale.
4142 */
4143 return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
4144}
4145static int diff_scoreopt_parse(const char *opt)
4147{
4148 int opt1, opt2, cmd;
4149 if (*opt++ != '-')
4151 return -1;
4152 cmd = *opt++;
4153 if (cmd == '-') {
4154 /* convert the long-form arguments into short-form versions */
4155 if (skip_prefix(opt, "break-rewrites", &opt)) {
4156 if (*opt == 0 || *opt++ == '=')
4157 cmd = 'B';
4158 } else if (skip_prefix(opt, "find-copies", &opt)) {
4159 if (*opt == 0 || *opt++ == '=')
4160 cmd = 'C';
4161 } else if (skip_prefix(opt, "find-renames", &opt)) {
4162 if (*opt == 0 || *opt++ == '=')
4163 cmd = 'M';
4164 }
4165 }
4166 if (cmd != 'M' && cmd != 'C' && cmd != 'B')
4167 return -1; /* that is not a -M, -C, or -B option */
4168 opt1 = parse_rename_score(&opt);
4170 if (cmd != 'B')
4171 opt2 = 0;
4172 else {
4173 if (*opt == 0)
4174 opt2 = 0;
4175 else if (*opt != '/')
4176 return -1; /* we expect -B80/99 or -B80 */
4177 else {
4178 opt++;
4179 opt2 = parse_rename_score(&opt);
4180 }
4181 }
4182 if (*opt != 0)
4183 return -1;
4184 return opt1 | (opt2 << 16);
4185}
4186struct diff_queue_struct diff_queued_diff;
4188void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
4190{
4191 ALLOC_GROW(queue->queue, queue->nr + 1, queue->alloc);
4192 queue->queue[queue->nr++] = dp;
4193}
4194struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
4196 struct diff_filespec *one,
4197 struct diff_filespec *two)
4198{
4199 struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
4200 dp->one = one;
4201 dp->two = two;
4202 if (queue)
4203 diff_q(queue, dp);
4204 return dp;
4205}
4206void diff_free_filepair(struct diff_filepair *p)
4208{
4209 free_filespec(p->one);
4210 free_filespec(p->two);
4211 free(p);
4212}
4213const char *diff_aligned_abbrev(const struct object_id *oid, int len)
4215{
4216 int abblen;
4217 const char *abbrev;
4218 if (len == GIT_SHA1_HEXSZ)
4220 return oid_to_hex(oid);
4221 abbrev = diff_abbrev_oid(oid, len);
4223 abblen = strlen(abbrev);
4224 /*
4226 * In well-behaved cases, where the abbbreviated result is the
4227 * same as the requested length, append three dots after the
4228 * abbreviation (hence the whole logic is limited to the case
4229 * where abblen < 37); when the actual abbreviated result is a
4230 * bit longer than the requested length, we reduce the number
4231 * of dots so that they match the well-behaved ones. However,
4232 * if the actual abbreviation is longer than the requested
4233 * length by more than three, we give up on aligning, and add
4234 * three dots anyway, to indicate that the output is not the
4235 * full object name. Yes, this may be suboptimal, but this
4236 * appears only in "diff --raw --abbrev" output and it is not
4237 * worth the effort to change it now. Note that this would
4238 * likely to work fine when the automatic sizing of default
4239 * abbreviation length is used--we would be fed -1 in "len" in
4240 * that case, and will end up always appending three-dots, but
4241 * the automatic sizing is supposed to give abblen that ensures
4242 * uniqueness across all objects (statistically speaking).
4243 */
4244 if (abblen < GIT_SHA1_HEXSZ - 3) {
4245 static char hex[GIT_MAX_HEXSZ + 1];
4246 if (len < abblen && abblen <= len + 2)
4247 xsnprintf(hex, sizeof(hex), "%s%.*s", abbrev, len+3-abblen, "..");
4248 else
4249 xsnprintf(hex, sizeof(hex), "%s...", abbrev);
4250 return hex;
4251 }
4252 return oid_to_hex(oid);
4254}
4255static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
4257{
4258 int line_termination = opt->line_termination;
4259 int inter_name_termination = line_termination ? '\t' : '\0';
4260 fprintf(opt->file, "%s", diff_line_prefix(opt));
4262 if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
4263 fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
4264 diff_aligned_abbrev(&p->one->oid, opt->abbrev));
4265 fprintf(opt->file, "%s ",
4266 diff_aligned_abbrev(&p->two->oid, opt->abbrev));
4267 }
4268 if (p->score) {
4269 fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
4270 inter_name_termination);
4271 } else {
4272 fprintf(opt->file, "%c%c", p->status, inter_name_termination);
4273 }
4274 if (p->status == DIFF_STATUS_COPIED ||
4276 p->status == DIFF_STATUS_RENAMED) {
4277 const char *name_a, *name_b;
4278 name_a = p->one->path;
4279 name_b = p->two->path;
4280 strip_prefix(opt->prefix_length, &name_a, &name_b);
4281 write_name_quoted(name_a, opt->file, inter_name_termination);
4282 write_name_quoted(name_b, opt->file, line_termination);
4283 } else {
4284 const char *name_a, *name_b;
4285 name_a = p->one->mode ? p->one->path : p->two->path;
4286 name_b = NULL;
4287 strip_prefix(opt->prefix_length, &name_a, &name_b);
4288 write_name_quoted(name_a, opt->file, line_termination);
4289 }
4290}
4291int diff_unmodified_pair(struct diff_filepair *p)
4293{
4294 /* This function is written stricter than necessary to support
4295 * the currently implemented transformers, but the idea is to
4296 * let transformers to produce diff_filepairs any way they want,
4297 * and filter and clean them up here before producing the output.
4298 */
4299 struct diff_filespec *one = p->one, *two = p->two;
4300 if (DIFF_PAIR_UNMERGED(p))
4302 return 0; /* unmerged is interesting */
4303 /* deletion, addition, mode or type change
4305 * and rename are all interesting.
4306 */
4307 if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
4308 DIFF_PAIR_MODE_CHANGED(p) ||
4309 strcmp(one->path, two->path))
4310 return 0;
4311 /* both are valid and point at the same path. that is, we are
4313 * dealing with a change.
4314 */
4315 if (one->oid_valid && two->oid_valid &&
4316 !oidcmp(&one->oid, &two->oid) &&
4317 !one->dirty_submodule && !two->dirty_submodule)
4318 return 1; /* no change */
4319 if (!one->oid_valid && !two->oid_valid)
4320 return 1; /* both look at the same file on the filesystem. */
4321 return 0;
4322}
4323static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
4325{
4326 if (diff_unmodified_pair(p))
4327 return;
4328 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4330 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4331 return; /* no tree diffs in patch format */
4332 run_diff(p, o);
4334}
4335static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
4337 struct diffstat_t *diffstat)
4338{
4339 if (diff_unmodified_pair(p))
4340 return;
4341 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4343 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4344 return; /* no useful stat for tree diffs */
4345 run_diffstat(p, o, diffstat);
4347}
4348static void diff_flush_checkdiff(struct diff_filepair *p,
4350 struct diff_options *o)
4351{
4352 if (diff_unmodified_pair(p))
4353 return;
4354 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4356 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4357 return; /* nothing to check in tree diffs */
4358 run_checkdiff(p, o);
4360}
4361int diff_queue_is_empty(void)
4363{
4364 struct diff_queue_struct *q = &diff_queued_diff;
4365 int i;
4366 for (i = 0; i < q->nr; i++)
4367 if (!diff_unmodified_pair(q->queue[i]))
4368 return 0;
4369 return 1;
4370}
4371#if DIFF_DEBUG
4373void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
4374{
4375 fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
4376 x, one ? one : "",
4377 s->path,
4378 DIFF_FILE_VALID(s) ? "valid" : "invalid",
4379 s->mode,
4380 s->oid_valid ? oid_to_hex(&s->oid) : "");
4381 fprintf(stderr, "queue[%d] %s size %lu\n",
4382 x, one ? one : "",
4383 s->size);
4384}
4385void diff_debug_filepair(const struct diff_filepair *p, int i)
4387{
4388 diff_debug_filespec(p->one, i, "one");
4389 diff_debug_filespec(p->two, i, "two");
4390 fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
4391 p->score, p->status ? p->status : '?',
4392 p->one->rename_used, p->broken_pair);
4393}
4394void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
4396{
4397 int i;
4398 if (msg)
4399 fprintf(stderr, "%s\n", msg);
4400 fprintf(stderr, "q->nr = %d\n", q->nr);
4401 for (i = 0; i < q->nr; i++) {
4402 struct diff_filepair *p = q->queue[i];
4403 diff_debug_filepair(p, i);
4404 }
4405}
4406#endif
4407static void diff_resolve_rename_copy(void)
4409{
4410 int i;
4411 struct diff_filepair *p;
4412 struct diff_queue_struct *q = &diff_queued_diff;
4413 diff_debug_queue("resolve-rename-copy", q);
4415 for (i = 0; i < q->nr; i++) {
4417 p = q->queue[i];
4418 p->status = 0; /* undecided */
4419 if (DIFF_PAIR_UNMERGED(p))
4420 p->status = DIFF_STATUS_UNMERGED;
4421 else if (!DIFF_FILE_VALID(p->one))
4422 p->status = DIFF_STATUS_ADDED;
4423 else if (!DIFF_FILE_VALID(p->two))
4424 p->status = DIFF_STATUS_DELETED;
4425 else if (DIFF_PAIR_TYPE_CHANGED(p))
4426 p->status = DIFF_STATUS_TYPE_CHANGED;
4427 /* from this point on, we are dealing with a pair
4429 * whose both sides are valid and of the same type, i.e.
4430 * either in-place edit or rename/copy edit.
4431 */
4432 else if (DIFF_PAIR_RENAME(p)) {
4433 /*
4434 * A rename might have re-connected a broken
4435 * pair up, causing the pathnames to be the
4436 * same again. If so, that's not a rename at
4437 * all, just a modification..
4438 *
4439 * Otherwise, see if this source was used for
4440 * multiple renames, in which case we decrement
4441 * the count, and call it a copy.
4442 */
4443 if (!strcmp(p->one->path, p->two->path))
4444 p->status = DIFF_STATUS_MODIFIED;
4445 else if (--p->one->rename_used > 0)
4446 p->status = DIFF_STATUS_COPIED;
4447 else
4448 p->status = DIFF_STATUS_RENAMED;
4449 }
4450 else if (oidcmp(&p->one->oid, &p->two->oid) ||
4451 p->one->mode != p->two->mode ||
4452 p->one->dirty_submodule ||
4453 p->two->dirty_submodule ||
4454 is_null_oid(&p->one->oid))
4455 p->status = DIFF_STATUS_MODIFIED;
4456 else {
4457 /* This is a "no-change" entry and should not
4458 * happen anymore, but prepare for broken callers.
4459 */
4460 error("feeding unmodified %s to diffcore",
4461 p->one->path);
4462 p->status = DIFF_STATUS_UNKNOWN;
4463 }
4464 }
4465 diff_debug_queue("resolve-rename-copy done", q);
4466}
4467static int check_pair_status(struct diff_filepair *p)
4469{
4470 switch (p->status) {
4471 case DIFF_STATUS_UNKNOWN:
4472 return 0;
4473 case 0:
4474 die("internal error in diff-resolve-rename-copy");
4475 default:
4476 return 1;
4477 }
4478}
4479static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
4481{
4482 int fmt = opt->output_format;
4483 if (fmt & DIFF_FORMAT_CHECKDIFF)
4485 diff_flush_checkdiff(p, opt);
4486 else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
4487 diff_flush_raw(p, opt);
4488 else if (fmt & DIFF_FORMAT_NAME) {
4489 const char *name_a, *name_b;
4490 name_a = p->two->path;
4491 name_b = NULL;
4492 strip_prefix(opt->prefix_length, &name_a, &name_b);
4493 fprintf(opt->file, "%s", diff_line_prefix(opt));
4494 write_name_quoted(name_a, opt->file, opt->line_termination);
4495 }
4496}
4497static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
4499{
4500 if (fs->mode)
4501 fprintf(file, " %s mode %06o ", newdelete, fs->mode);
4502 else
4503 fprintf(file, " %s ", newdelete);
4504 write_name_quoted(fs->path, file, '\n');
4505}
4506static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
4509 const char *line_prefix)
4510{
4511 if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
4512 fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
4513 p->two->mode, show_name ? ' ' : '\n');
4514 if (show_name) {
4515 write_name_quoted(p->two->path, file, '\n');
4516 }
4517 }
4518}
4519static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
4521 const char *line_prefix)
4522{
4523 char *names = pprint_rename(p->one->path, p->two->path);
4524 fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
4526 free(names);
4527 show_mode_change(file, p, 0, line_prefix);
4528}
4529static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
4531{
4532 FILE *file = opt->file;
4533 const char *line_prefix = diff_line_prefix(opt);
4534 switch(p->status) {
4536 case DIFF_STATUS_DELETED:
4537 fputs(line_prefix, file);
4538 show_file_mode_name(file, "delete", p->one);
4539 break;
4540 case DIFF_STATUS_ADDED:
4541 fputs(line_prefix, file);
4542 show_file_mode_name(file, "create", p->two);
4543 break;
4544 case DIFF_STATUS_COPIED:
4545 fputs(line_prefix, file);
4546 show_rename_copy(file, "copy", p, line_prefix);
4547 break;
4548 case DIFF_STATUS_RENAMED:
4549 fputs(line_prefix, file);
4550 show_rename_copy(file, "rename", p, line_prefix);
4551 break;
4552 default:
4553 if (p->score) {
4554 fprintf(file, "%s rewrite ", line_prefix);
4555 write_name_quoted(p->two->path, file, ' ');
4556 fprintf(file, "(%d%%)\n", similarity_index(p));
4557 }
4558 show_mode_change(file, p, !p->score, line_prefix);
4559 break;
4560 }
4561}
4562struct patch_id_t {
4564 git_SHA_CTX *ctx;
4565 int patchlen;
4566};
4567static int remove_space(char *line, int len)
4569{
4570 int i;
4571 char *dst = line;
4572 unsigned char c;
4573 for (i = 0; i < len; i++)
4575 if (!isspace((c = line[i])))
4576 *dst++ = c;
4577 return dst - line;
4579}
4580static void patch_id_consume(void *priv, char *line, unsigned long len)
4582{
4583 struct patch_id_t *data = priv;
4584 int new_len;
4585 /* Ignore line numbers when computing the SHA1 of the patch */
4587 if (starts_with(line, "@@ -"))
4588 return;
4589 new_len = remove_space(line, len);
4591 git_SHA1_Update(data->ctx, line, new_len);
4593 data->patchlen += new_len;
4594}
4595static void patch_id_add_string(git_SHA_CTX *ctx, const char *str)
4597{
4598 git_SHA1_Update(ctx, str, strlen(str));
4599}
4600static void patch_id_add_mode(git_SHA_CTX *ctx, unsigned mode)
4602{
4603 /* large enough for 2^32 in octal */
4604 char buf[12];
4605 int len = xsnprintf(buf, sizeof(buf), "%06o", mode);
4606 git_SHA1_Update(ctx, buf, len);
4607}
4608/* returns 0 upon success, and writes result into sha1 */
4610static int diff_get_patch_id(struct diff_options *options, struct object_id *oid, int diff_header_only)
4611{
4612 struct diff_queue_struct *q = &diff_queued_diff;
4613 int i;
4614 git_SHA_CTX ctx;
4615 struct patch_id_t data;
4616 git_SHA1_Init(&ctx);
4618 memset(&data, 0, sizeof(struct patch_id_t));
4619 data.ctx = &ctx;
4620 for (i = 0; i < q->nr; i++) {
4622 xpparam_t xpp;
4623 xdemitconf_t xecfg;
4624 mmfile_t mf1, mf2;
4625 struct diff_filepair *p = q->queue[i];
4626 int len1, len2;
4627 memset(&xpp, 0, sizeof(xpp));
4629 memset(&xecfg, 0, sizeof(xecfg));
4630 if (p->status == 0)
4631 return error("internal diff status error");
4632 if (p->status == DIFF_STATUS_UNKNOWN)
4633 continue;
4634 if (diff_unmodified_pair(p))
4635 continue;
4636 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4637 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4638 continue;
4639 if (DIFF_PAIR_UNMERGED(p))
4640 continue;
4641 diff_fill_oid_info(p->one);
4643 diff_fill_oid_info(p->two);
4644 len1 = remove_space(p->one->path, strlen(p->one->path));
4646 len2 = remove_space(p->two->path, strlen(p->two->path));
4647 patch_id_add_string(&ctx, "diff--git");
4648 patch_id_add_string(&ctx, "a/");
4649 git_SHA1_Update(&ctx, p->one->path, len1);
4650 patch_id_add_string(&ctx, "b/");
4651 git_SHA1_Update(&ctx, p->two->path, len2);
4652 if (p->one->mode == 0) {
4654 patch_id_add_string(&ctx, "newfilemode");
4655 patch_id_add_mode(&ctx, p->two->mode);
4656 patch_id_add_string(&ctx, "---/dev/null");
4657 patch_id_add_string(&ctx, "+++b/");
4658 git_SHA1_Update(&ctx, p->two->path, len2);
4659 } else if (p->two->mode == 0) {
4660 patch_id_add_string(&ctx, "deletedfilemode");
4661 patch_id_add_mode(&ctx, p->one->mode);
4662 patch_id_add_string(&ctx, "---a/");
4663 git_SHA1_Update(&ctx, p->one->path, len1);
4664 patch_id_add_string(&ctx, "+++/dev/null");
4665 } else {
4666 patch_id_add_string(&ctx, "---a/");
4667 git_SHA1_Update(&ctx, p->one->path, len1);
4668 patch_id_add_string(&ctx, "+++b/");
4669 git_SHA1_Update(&ctx, p->two->path, len2);
4670 }
4671 if (diff_header_only)
4673 continue;
4674 if (fill_mmfile(&mf1, p->one) < 0 ||
4676 fill_mmfile(&mf2, p->two) < 0)
4677 return error("unable to read files to diff");
4678 if (diff_filespec_is_binary(p->one) ||
4680 diff_filespec_is_binary(p->two)) {
4681 git_SHA1_Update(&ctx, oid_to_hex(&p->one->oid),
4682 GIT_SHA1_HEXSZ);
4683 git_SHA1_Update(&ctx, oid_to_hex(&p->two->oid),
4684 GIT_SHA1_HEXSZ);
4685 continue;
4686 }
4687 xpp.flags = 0;
4689 xecfg.ctxlen = 3;
4690 xecfg.flags = 0;
4691 if (xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4692 &xpp, &xecfg))
4693 return error("unable to generate patch-id diff for %s",
4694 p->one->path);
4695 }
4696 git_SHA1_Final(oid->hash, &ctx);
4698 return 0;
4699}
4700int diff_flush_patch_id(struct diff_options *options, struct object_id *oid, int diff_header_only)
4702{
4703 struct diff_queue_struct *q = &diff_queued_diff;
4704 int i;
4705 int result = diff_get_patch_id(options, oid, diff_header_only);
4706 for (i = 0; i < q->nr; i++)
4708 diff_free_filepair(q->queue[i]);
4709 free(q->queue);
4711 DIFF_QUEUE_CLEAR(q);
4712 return result;
4714}
4715static int is_summary_empty(const struct diff_queue_struct *q)
4717{
4718 int i;
4719 for (i = 0; i < q->nr; i++) {
4721 const struct diff_filepair *p = q->queue[i];
4722 switch (p->status) {
4724 case DIFF_STATUS_DELETED:
4725 case DIFF_STATUS_ADDED:
4726 case DIFF_STATUS_COPIED:
4727 case DIFF_STATUS_RENAMED:
4728 return 0;
4729 default:
4730 if (p->score)
4731 return 0;
4732 if (p->one->mode && p->two->mode &&
4733 p->one->mode != p->two->mode)
4734 return 0;
4735 break;
4736 }
4737 }
4738 return 1;
4739}
4740static const char rename_limit_warning[] =
4742N_("inexact rename detection was skipped due to too many files.");
4743static const char degrade_cc_to_c_warning[] =
4745N_("only found copies from modified paths due to too many files.");
4746static const char rename_limit_advice[] =
4748N_("you may want to set your %s variable to at least "
4749 "%d and retry the command.");
4750void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4752{
4753 if (degraded_cc)
4754 warning(_(degrade_cc_to_c_warning));
4755 else if (needed)
4756 warning(_(rename_limit_warning));
4757 else
4758 return;
4759 if (0 < needed && needed < 32767)
4760 warning(_(rename_limit_advice), varname, needed);
4761}
4762static void diff_flush_patch_all_file_pairs(struct diff_options *o)
4764{
4765 int i;
4766 struct diff_queue_struct *q = &diff_queued_diff;
4767 for (i = 0; i < q->nr; i++) {
4768 struct diff_filepair *p = q->queue[i];
4769 if (check_pair_status(p))
4770 diff_flush_patch(p, o);
4771 }
4772}
4773void diff_flush(struct diff_options *options)
4775{
4776 struct diff_queue_struct *q = &diff_queued_diff;
4777 int i, output_format = options->output_format;
4778 int separator = 0;
4779 int dirstat_by_line = 0;
4780 /*
4782 * Order: raw, stat, summary, patch
4783 * or: name/name-status/checkdiff (other bits clear)
4784 */
4785 if (!q->nr)
4786 goto free_queue;
4787 if (output_format & (DIFF_FORMAT_RAW |
4789 DIFF_FORMAT_NAME |
4790 DIFF_FORMAT_NAME_STATUS |
4791 DIFF_FORMAT_CHECKDIFF)) {
4792 for (i = 0; i < q->nr; i++) {
4793 struct diff_filepair *p = q->queue[i];
4794 if (check_pair_status(p))
4795 flush_one_pair(p, options);
4796 }
4797 separator++;
4798 }
4799 if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4801 dirstat_by_line = 1;
4802 if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4804 dirstat_by_line) {
4805 struct diffstat_t diffstat;
4806 memset(&diffstat, 0, sizeof(struct diffstat_t));
4808 for (i = 0; i < q->nr; i++) {
4809 struct diff_filepair *p = q->queue[i];
4810 if (check_pair_status(p))
4811 diff_flush_stat(p, options, &diffstat);
4812 }
4813 if (output_format & DIFF_FORMAT_NUMSTAT)
4814 show_numstat(&diffstat, options);
4815 if (output_format & DIFF_FORMAT_DIFFSTAT)
4816 show_stats(&diffstat, options);
4817 if (output_format & DIFF_FORMAT_SHORTSTAT)
4818 show_shortstats(&diffstat, options);
4819 if (output_format & DIFF_FORMAT_DIRSTAT && dirstat_by_line)
4820 show_dirstat_by_line(&diffstat, options);
4821 free_diffstat_info(&diffstat);
4822 separator++;
4823 }
4824 if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4825 show_dirstat(options);
4826 if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4828 for (i = 0; i < q->nr; i++) {
4829 diff_summary(options, q->queue[i]);
4830 }
4831 separator++;
4832 }
4833 if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4835 DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4836 DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4837 /*
4838 * run diff_flush_patch for the exit status. setting
4839 * options->file to /dev/null should be safe, because we
4840 * aren't supposed to produce any output anyway.
4841 */
4842 if (options->close_file)
4843 fclose(options->file);
4844 options->file = xfopen("/dev/null", "w");
4845 options->close_file = 1;
4846 for (i = 0; i < q->nr; i++) {
4847 struct diff_filepair *p = q->queue[i];
4848 if (check_pair_status(p))
4849 diff_flush_patch(p, options);
4850 if (options->found_changes)
4851 break;
4852 }
4853 }
4854 if (output_format & DIFF_FORMAT_PATCH) {
4856 if (separator) {
4857 emit_diff_symbol(options, DIFF_SYMBOL_SEPARATOR, NULL, 0);
4858 if (options->stat_sep) {
4859 /* attach patch instead of inline */
4860 fputs(options->stat_sep, options->file);
4861 }
4862 }
4863 diff_flush_patch_all_file_pairs(options);
4865 }
4866 if (output_format & DIFF_FORMAT_CALLBACK)
4868 options->format_callback(q, options, options->format_callback_data);
4869 for (i = 0; i < q->nr; i++)
4871 diff_free_filepair(q->queue[i]);
4872free_queue:
4873 free(q->queue);
4874 DIFF_QUEUE_CLEAR(q);
4875 if (options->close_file)
4876 fclose(options->file);
4877 /*
4879 * Report the content-level differences with HAS_CHANGES;
4880 * diff_addremove/diff_change does not set the bit when
4881 * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4882 */
4883 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4884 if (options->found_changes)
4885 DIFF_OPT_SET(options, HAS_CHANGES);
4886 else
4887 DIFF_OPT_CLR(options, HAS_CHANGES);
4888 }
4889}
4890static int match_filter(const struct diff_options *options, const struct diff_filepair *p)
4892{
4893 return (((p->status == DIFF_STATUS_MODIFIED) &&
4894 ((p->score &&
4895 filter_bit_tst(DIFF_STATUS_FILTER_BROKEN, options)) ||
4896 (!p->score &&
4897 filter_bit_tst(DIFF_STATUS_MODIFIED, options)))) ||
4898 ((p->status != DIFF_STATUS_MODIFIED) &&
4899 filter_bit_tst(p->status, options)));
4900}
4901static void diffcore_apply_filter(struct diff_options *options)
4903{
4904 int i;
4905 struct diff_queue_struct *q = &diff_queued_diff;
4906 struct diff_queue_struct outq;
4907 DIFF_QUEUE_CLEAR(&outq);
4909 if (!options->filter)
4911 return;
4912 if (filter_bit_tst(DIFF_STATUS_FILTER_AON, options)) {
4914 int found;
4915 for (i = found = 0; !found && i < q->nr; i++) {
4916 if (match_filter(options, q->queue[i]))
4917 found++;
4918 }
4919 if (found)
4920 return;
4921 /* otherwise we will clear the whole queue
4923 * by copying the empty outq at the end of this
4924 * function, but first clear the current entries
4925 * in the queue.
4926 */
4927 for (i = 0; i < q->nr; i++)
4928 diff_free_filepair(q->queue[i]);
4929 }
4930 else {
4931 /* Only the matching ones */
4932 for (i = 0; i < q->nr; i++) {
4933 struct diff_filepair *p = q->queue[i];
4934 if (match_filter(options, p))
4935 diff_q(&outq, p);
4936 else
4937 diff_free_filepair(p);
4938 }
4939 }
4940 free(q->queue);
4941 *q = outq;
4942}
4943/* Check whether two filespecs with the same mode and size are identical */
4945static int diff_filespec_is_identical(struct diff_filespec *one,
4946 struct diff_filespec *two)
4947{
4948 if (S_ISGITLINK(one->mode))
4949 return 0;
4950 if (diff_populate_filespec(one, 0))
4951 return 0;
4952 if (diff_populate_filespec(two, 0))
4953 return 0;
4954 return !memcmp(one->data, two->data, one->size);
4955}
4956static int diff_filespec_check_stat_unmatch(struct diff_filepair *p)
4958{
4959 if (p->done_skip_stat_unmatch)
4960 return p->skip_stat_unmatch_result;
4961 p->done_skip_stat_unmatch = 1;
4963 p->skip_stat_unmatch_result = 0;
4964 /*
4965 * 1. Entries that come from stat info dirtiness
4966 * always have both sides (iow, not create/delete),
4967 * one side of the object name is unknown, with
4968 * the same mode and size. Keep the ones that
4969 * do not match these criteria. They have real
4970 * differences.
4971 *
4972 * 2. At this point, the file is known to be modified,
4973 * with the same mode and size, and the object
4974 * name of one side is unknown. Need to inspect
4975 * the identical contents.
4976 */
4977 if (!DIFF_FILE_VALID(p->one) || /* (1) */
4978 !DIFF_FILE_VALID(p->two) ||
4979 (p->one->oid_valid && p->two->oid_valid) ||
4980 (p->one->mode != p->two->mode) ||
4981 diff_populate_filespec(p->one, CHECK_SIZE_ONLY) ||
4982 diff_populate_filespec(p->two, CHECK_SIZE_ONLY) ||
4983 (p->one->size != p->two->size) ||
4984 !diff_filespec_is_identical(p->one, p->two)) /* (2) */
4985 p->skip_stat_unmatch_result = 1;
4986 return p->skip_stat_unmatch_result;
4987}
4988static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
4990{
4991 int i;
4992 struct diff_queue_struct *q = &diff_queued_diff;
4993 struct diff_queue_struct outq;
4994 DIFF_QUEUE_CLEAR(&outq);
4995 for (i = 0; i < q->nr; i++) {
4997 struct diff_filepair *p = q->queue[i];
4998 if (diff_filespec_check_stat_unmatch(p))
5000 diff_q(&outq, p);
5001 else {
5002 /*
5003 * The caller can subtract 1 from skip_stat_unmatch
5004 * to determine how many paths were dirty only
5005 * due to stat info mismatch.
5006 */
5007 if (!DIFF_OPT_TST(diffopt, NO_INDEX))
5008 diffopt->skip_stat_unmatch++;
5009 diff_free_filepair(p);
5010 }
5011 }
5012 free(q->queue);
5013 *q = outq;
5014}
5015static int diffnamecmp(const void *a_, const void *b_)
5017{
5018 const struct diff_filepair *a = *((const struct diff_filepair **)a_);
5019 const struct diff_filepair *b = *((const struct diff_filepair **)b_);
5020 const char *name_a, *name_b;
5021 name_a = a->one ? a->one->path : a->two->path;
5023 name_b = b->one ? b->one->path : b->two->path;
5024 return strcmp(name_a, name_b);
5025}
5026void diffcore_fix_diff_index(struct diff_options *options)
5028{
5029 struct diff_queue_struct *q = &diff_queued_diff;
5030 QSORT(q->queue, q->nr, diffnamecmp);
5031}
5032void diffcore_std(struct diff_options *options)
5034{
5035 /* NOTE please keep the following in sync with diff_tree_combined() */
5036 if (options->skip_stat_unmatch)
5037 diffcore_skip_stat_unmatch(options);
5038 if (!options->found_follow) {
5039 /* See try_to_follow_renames() in tree-diff.c */
5040 if (options->break_opt != -1)
5041 diffcore_break(options->break_opt);
5042 if (options->detect_rename)
5043 diffcore_rename(options);
5044 if (options->break_opt != -1)
5045 diffcore_merge_broken();
5046 }
5047 if (options->pickaxe)
5048 diffcore_pickaxe(options);
5049 if (options->orderfile)
5050 diffcore_order(options->orderfile);
5051 if (!options->found_follow)
5052 /* See try_to_follow_renames() in tree-diff.c */
5053 diff_resolve_rename_copy();
5054 diffcore_apply_filter(options);
5055 if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5057 DIFF_OPT_SET(options, HAS_CHANGES);
5058 else
5059 DIFF_OPT_CLR(options, HAS_CHANGES);
5060 options->found_follow = 0;
5062}
5063int diff_result_code(struct diff_options *opt, int status)
5065{
5066 int result = 0;
5067 diff_warn_rename_limit("diff.renameLimit",
5069 opt->needed_rename_limit,
5070 opt->degraded_cc_to_c);
5071 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5072 !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
5073 return status;
5074 if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5075 DIFF_OPT_TST(opt, HAS_CHANGES))
5076 result |= 01;
5077 if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
5078 DIFF_OPT_TST(opt, CHECK_FAILED))
5079 result |= 02;
5080 return result;
5081}
5082int diff_can_quit_early(struct diff_options *opt)
5084{
5085 return (DIFF_OPT_TST(opt, QUICK) &&
5086 !opt->filter &&
5087 DIFF_OPT_TST(opt, HAS_CHANGES));
5088}
5089/*
5091 * Shall changes to this submodule be ignored?
5092 *
5093 * Submodule changes can be configured to be ignored separately for each path,
5094 * but that configuration can be overridden from the command line.
5095 */
5096static int is_submodule_ignored(const char *path, struct diff_options *options)
5097{
5098 int ignored = 0;
5099 unsigned orig_flags = options->flags;
5100 if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
5101 set_diffopt_flags_from_submodule_config(options, path);
5102 if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
5103 ignored = 1;
5104 options->flags = orig_flags;
5105 return ignored;
5106}
5107void diff_addremove(struct diff_options *options,
5109 int addremove, unsigned mode,
5110 const struct object_id *oid,
5111 int oid_valid,
5112 const char *concatpath, unsigned dirty_submodule)
5113{
5114 struct diff_filespec *one, *two;
5115 if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
5117 return;
5118 /* This may look odd, but it is a preparation for
5120 * feeding "there are unchanged files which should
5121 * not produce diffs, but when you are doing copy
5122 * detection you would need them, so here they are"
5123 * entries to the diff-core. They will be prefixed
5124 * with something like '=' or '*' (I haven't decided
5125 * which but should not make any difference).
5126 * Feeding the same new and old to diff_change()
5127 * also has the same effect.
5128 * Before the final output happens, they are pruned after
5129 * merged into rename/copy pairs as appropriate.
5130 */
5131 if (DIFF_OPT_TST(options, REVERSE_DIFF))
5132 addremove = (addremove == '+' ? '-' :
5133 addremove == '-' ? '+' : addremove);
5134 if (options->prefix &&
5136 strncmp(concatpath, options->prefix, options->prefix_length))
5137 return;
5138 one = alloc_filespec(concatpath);
5140 two = alloc_filespec(concatpath);
5141 if (addremove != '+')
5143 fill_filespec(one, oid, oid_valid, mode);
5144 if (addremove != '-') {
5145 fill_filespec(two, oid, oid_valid, mode);
5146 two->dirty_submodule = dirty_submodule;
5147 }
5148 diff_queue(&diff_queued_diff, one, two);
5150 if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5151 DIFF_OPT_SET(options, HAS_CHANGES);
5152}
5153void diff_change(struct diff_options *options,
5155 unsigned old_mode, unsigned new_mode,
5156 const struct object_id *old_oid,
5157 const struct object_id *new_oid,
5158 int old_oid_valid, int new_oid_valid,
5159 const char *concatpath,
5160 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
5161{
5162 struct diff_filespec *one, *two;
5163 struct diff_filepair *p;
5164 if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
5166 is_submodule_ignored(concatpath, options))
5167 return;
5168 if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
5170 SWAP(old_mode, new_mode);
5171 SWAP(old_oid, new_oid);
5172 SWAP(old_oid_valid, new_oid_valid);
5173 SWAP(old_dirty_submodule, new_dirty_submodule);
5174 }
5175 if (options->prefix &&
5177 strncmp(concatpath, options->prefix, options->prefix_length))
5178 return;
5179 one = alloc_filespec(concatpath);
5181 two = alloc_filespec(concatpath);
5182 fill_filespec(one, old_oid, old_oid_valid, old_mode);
5183 fill_filespec(two, new_oid, new_oid_valid, new_mode);
5184 one->dirty_submodule = old_dirty_submodule;
5185 two->dirty_submodule = new_dirty_submodule;
5186 p = diff_queue(&diff_queued_diff, one, two);
5187 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5189 return;
5190 if (DIFF_OPT_TST(options, QUICK) && options->skip_stat_unmatch &&
5192 !diff_filespec_check_stat_unmatch(p))
5193 return;
5194 DIFF_OPT_SET(options, HAS_CHANGES);
5196}
5197struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
5199{
5200 struct diff_filepair *pair;
5201 struct diff_filespec *one, *two;
5202 if (options->prefix &&
5204 strncmp(path, options->prefix, options->prefix_length))
5205 return NULL;
5206 one = alloc_filespec(path);
5208 two = alloc_filespec(path);
5209 pair = diff_queue(&diff_queued_diff, one, two);
5210 pair->is_unmerged = 1;
5211 return pair;
5212}
5213static char *run_textconv(const char *pgm, struct diff_filespec *spec,
5215 size_t *outsize)
5216{
5217 struct diff_tempfile *temp;
5218 const char *argv[3];
5219 const char **arg = argv;
5220 struct child_process child = CHILD_PROCESS_INIT;
5221 struct strbuf buf = STRBUF_INIT;
5222 int err = 0;
5223 temp = prepare_temp_file(spec->path, spec);
5225 *arg++ = pgm;
5226 *arg++ = temp->name;
5227 *arg = NULL;
5228 child.use_shell = 1;
5230 child.argv = argv;
5231 child.out = -1;
5232 if (start_command(&child)) {
5233 remove_tempfile();
5234 return NULL;
5235 }
5236 if (strbuf_read(&buf, child.out, 0) < 0)
5238 err = error("error reading from textconv command '%s'", pgm);
5239 close(child.out);
5240 if (finish_command(&child) || err) {
5242 strbuf_release(&buf);
5243 remove_tempfile();
5244 return NULL;
5245 }
5246 remove_tempfile();
5247 return strbuf_detach(&buf, outsize);
5249}
5250size_t fill_textconv(struct userdiff_driver *driver,
5252 struct diff_filespec *df,
5253 char **outbuf)
5254{
5255 size_t size;
5256 if (!driver) {
5258 if (!DIFF_FILE_VALID(df)) {
5259 *outbuf = "";
5260 return 0;
5261 }
5262 if (diff_populate_filespec(df, 0))
5263 die("unable to read files to diff");
5264 *outbuf = df->data;
5265 return df->size;
5266 }
5267 if (!driver->textconv)
5269 die("BUG: fill_textconv called with non-textconv driver");
5270 if (driver->textconv_cache && df->oid_valid) {
5272 *outbuf = notes_cache_get(driver->textconv_cache,
5273 &df->oid,
5274 &size);
5275 if (*outbuf)
5276 return size;
5277 }
5278 *outbuf = run_textconv(driver->textconv, df, &size);
5280 if (!*outbuf)
5281 die("unable to read files to diff");
5282 if (driver->textconv_cache && df->oid_valid) {
5284 /* ignore errors, as we might be in a readonly repository */
5285 notes_cache_put(driver->textconv_cache, &df->oid, *outbuf,
5286 size);
5287 /*
5288 * we could save up changes and flush them all at the end,
5289 * but we would need an extra call after all diffing is done.
5290 * Since generating a cache entry is the slow path anyway,
5291 * this extra overhead probably isn't a big deal.
5292 */
5293 notes_cache_write(driver->textconv_cache);
5294 }
5295 return size;
5297}
5298int textconv_object(const char *path,
5300 unsigned mode,
5301 const struct object_id *oid,
5302 int oid_valid,
5303 char **buf,
5304 unsigned long *buf_size)
5305{
5306 struct diff_filespec *df;
5307 struct userdiff_driver *textconv;
5308 df = alloc_filespec(path);
5310 fill_filespec(df, oid, oid_valid, mode);
5311 textconv = get_textconv(df);
5312 if (!textconv) {
5313 free_filespec(df);
5314 return 0;
5315 }
5316 *buf_size = fill_textconv(textconv, df, buf);
5318 free_filespec(df);
5319 return 1;
5320}
5321void setup_diff_pager(struct diff_options *opt)
5323{
5324 /*
5325 * If the user asked for our exit code, then either they want --quiet
5326 * or --exit-code. We should definitely not bother with a pager in the
5327 * former case, as we will generate no output. Since we still properly
5328 * report our exit code even when a pager is run, we _could_ run a
5329 * pager with --exit-code. But since we have not done so historically,
5330 * and because it is easy to find people oneline advising "git diff
5331 * --exit-code" in hooks and other scripts, we do not do so.
5332 */
5333 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5334 check_pager_config("diff") != 0)
5335 setup_pager();
5336}