488096b757377b4f86109253e3b275ea54254087
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_NO_LF_EOF,
565 DIFF_SYMBOL_CONTEXT_FRAGINFO,
566 DIFF_SYMBOL_CONTEXT_MARKER,
567 DIFF_SYMBOL_SEPARATOR
568};
569static void emit_diff_symbol(struct diff_options *o, enum diff_symbol s,
571 const char *line, int len)
572{
573 static const char *nneof = " No newline at end of file\n";
574 const char *context, *reset;
575 switch (s) {
576 case DIFF_SYMBOL_NO_LF_EOF:
577 context = diff_get_color_opt(o, DIFF_CONTEXT);
578 reset = diff_get_color_opt(o, DIFF_RESET);
579 putc('\n', o->file);
580 emit_line_0(o, context, reset, '\\',
581 nneof, strlen(nneof));
582 break;
583 case DIFF_SYMBOL_CONTEXT_FRAGINFO:
584 emit_line(o, "", "", line, len);
585 break;
586 case DIFF_SYMBOL_CONTEXT_MARKER:
587 context = diff_get_color_opt(o, DIFF_CONTEXT);
588 reset = diff_get_color_opt(o, DIFF_RESET);
589 emit_line(o, context, reset, line, len);
590 break;
591 case DIFF_SYMBOL_SEPARATOR:
592 fprintf(o->file, "%s%c",
593 diff_line_prefix(o),
594 o->line_termination);
595 break;
596 default:
597 die("BUG: unknown diff symbol");
598 }
599}
600static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
602{
603 if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
604 ecbdata->blank_at_eof_in_preimage &&
605 ecbdata->blank_at_eof_in_postimage &&
606 ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
607 ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
608 return 0;
609 return ws_blank_line(line, len, ecbdata->ws_rule);
610}
611static void emit_line_checked(const char *reset,
613 struct emit_callback *ecbdata,
614 const char *line, int len,
615 enum color_diff color,
616 unsigned ws_error_highlight,
617 char sign)
618{
619 const char *set = diff_get_color(ecbdata->color_diff, color);
620 const char *ws = NULL;
621 if (ecbdata->opt->ws_error_highlight & ws_error_highlight) {
623 ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
624 if (!*ws)
625 ws = NULL;
626 }
627 if (!ws)
629 emit_line_0(ecbdata->opt, set, reset, sign, line, len);
630 else if (sign == '+' && new_blank_line_at_eof(ecbdata, line, len))
631 /* Blank line at EOF - paint '+' as well */
632 emit_line_0(ecbdata->opt, ws, reset, sign, line, len);
633 else {
634 /* Emit just the prefix, then the rest. */
635 emit_line_0(ecbdata->opt, set, reset, sign, "", 0);
636 ws_check_emit(line, len, ecbdata->ws_rule,
637 ecbdata->opt->file, set, reset, ws);
638 }
639}
640static void emit_add_line(const char *reset,
642 struct emit_callback *ecbdata,
643 const char *line, int len)
644{
645 emit_line_checked(reset, ecbdata, line, len,
646 DIFF_FILE_NEW, WSEH_NEW, '+');
647}
648static void emit_del_line(const char *reset,
650 struct emit_callback *ecbdata,
651 const char *line, int len)
652{
653 emit_line_checked(reset, ecbdata, line, len,
654 DIFF_FILE_OLD, WSEH_OLD, '-');
655}
656static void emit_context_line(const char *reset,
658 struct emit_callback *ecbdata,
659 const char *line, int len)
660{
661 emit_line_checked(reset, ecbdata, line, len,
662 DIFF_CONTEXT, WSEH_CONTEXT, ' ');
663}
664static void emit_hunk_header(struct emit_callback *ecbdata,
666 const char *line, int len)
667{
668 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
669 const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
670 const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
671 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
672 static const char atat[2] = { '@', '@' };
673 const char *cp, *ep;
674 struct strbuf msgbuf = STRBUF_INIT;
675 int org_len = len;
676 int i = 1;
677 /*
679 * As a hunk header must begin with "@@ -<old>, +<new> @@",
680 * it always is at least 10 bytes long.
681 */
682 if (len < 10 ||
683 memcmp(line, atat, 2) ||
684 !(ep = memmem(line + 2, len - 2, atat, 2))) {
685 emit_diff_symbol(ecbdata->opt,
686 DIFF_SYMBOL_CONTEXT_MARKER, line, len);
687 return;
688 }
689 ep += 2; /* skip over @@ */
690 /* The hunk header in fraginfo color */
692 strbuf_addstr(&msgbuf, frag);
693 strbuf_add(&msgbuf, line, ep - line);
694 strbuf_addstr(&msgbuf, reset);
695 /*
697 * trailing "\r\n"
698 */
699 for ( ; i < 3; i++)
700 if (line[len - i] == '\r' || line[len - i] == '\n')
701 len--;
702 /* blank before the func header */
704 for (cp = ep; ep - line < len; ep++)
705 if (*ep != ' ' && *ep != '\t')
706 break;
707 if (ep != cp) {
708 strbuf_addstr(&msgbuf, context);
709 strbuf_add(&msgbuf, cp, ep - cp);
710 strbuf_addstr(&msgbuf, reset);
711 }
712 if (ep < line + len) {
714 strbuf_addstr(&msgbuf, func);
715 strbuf_add(&msgbuf, ep, line + len - ep);
716 strbuf_addstr(&msgbuf, reset);
717 }
718 strbuf_add(&msgbuf, line + len, org_len - len);
720 strbuf_complete_line(&msgbuf);
721 emit_diff_symbol(ecbdata->opt,
722 DIFF_SYMBOL_CONTEXT_FRAGINFO, msgbuf.buf, msgbuf.len);
723 strbuf_release(&msgbuf);
724}
725static struct diff_tempfile *claim_diff_tempfile(void) {
727 int i;
728 for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
729 if (!diff_temp[i].name)
730 return diff_temp + i;
731 die("BUG: diff is failing to clean up its tempfiles");
732}
733static void remove_tempfile(void)
735{
736 int i;
737 for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
738 if (is_tempfile_active(&diff_temp[i].tempfile))
739 delete_tempfile(&diff_temp[i].tempfile);
740 diff_temp[i].name = NULL;
741 }
742}
743static void print_line_count(FILE *file, int count)
745{
746 switch (count) {
747 case 0:
748 fprintf(file, "0,0");
749 break;
750 case 1:
751 fprintf(file, "1");
752 break;
753 default:
754 fprintf(file, "1,%d", count);
755 break;
756 }
757}
758static void emit_rewrite_lines(struct emit_callback *ecb,
760 int prefix, const char *data, int size)
761{
762 const char *endp = NULL;
763 const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
764 while (0 < size) {
766 int len;
767 endp = memchr(data, '\n', size);
769 len = endp ? (endp - data + 1) : size;
770 if (prefix != '+') {
771 ecb->lno_in_preimage++;
772 emit_del_line(reset, ecb, data, len);
773 } else {
774 ecb->lno_in_postimage++;
775 emit_add_line(reset, ecb, data, len);
776 }
777 size -= len;
778 data += len;
779 }
780 if (!endp)
781 emit_diff_symbol(ecb->opt, DIFF_SYMBOL_NO_LF_EOF, NULL, 0);
782}
783static void emit_rewrite_diff(const char *name_a,
785 const char *name_b,
786 struct diff_filespec *one,
787 struct diff_filespec *two,
788 struct userdiff_driver *textconv_one,
789 struct userdiff_driver *textconv_two,
790 struct diff_options *o)
791{
792 int lc_a, lc_b;
793 const char *name_a_tab, *name_b_tab;
794 const char *metainfo = diff_get_color(o->use_color, DIFF_METAINFO);
795 const char *fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
796 const char *reset = diff_get_color(o->use_color, DIFF_RESET);
797 static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
798 const char *a_prefix, *b_prefix;
799 char *data_one, *data_two;
800 size_t size_one, size_two;
801 struct emit_callback ecbdata;
802 const char *line_prefix = diff_line_prefix(o);
803 if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
805 a_prefix = o->b_prefix;
806 b_prefix = o->a_prefix;
807 } else {
808 a_prefix = o->a_prefix;
809 b_prefix = o->b_prefix;
810 }
811 name_a += (*name_a == '/');
813 name_b += (*name_b == '/');
814 name_a_tab = strchr(name_a, ' ') ? "\t" : "";
815 name_b_tab = strchr(name_b, ' ') ? "\t" : "";
816 strbuf_reset(&a_name);
818 strbuf_reset(&b_name);
819 quote_two_c_style(&a_name, a_prefix, name_a, 0);
820 quote_two_c_style(&b_name, b_prefix, name_b, 0);
821 size_one = fill_textconv(textconv_one, one, &data_one);
823 size_two = fill_textconv(textconv_two, two, &data_two);
824 memset(&ecbdata, 0, sizeof(ecbdata));
826 ecbdata.color_diff = want_color(o->use_color);
827 ecbdata.ws_rule = whitespace_rule(name_b);
828 ecbdata.opt = o;
829 if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
830 mmfile_t mf1, mf2;
831 mf1.ptr = (char *)data_one;
832 mf2.ptr = (char *)data_two;
833 mf1.size = size_one;
834 mf2.size = size_two;
835 check_blank_at_eof(&mf1, &mf2, &ecbdata);
836 }
837 ecbdata.lno_in_preimage = 1;
838 ecbdata.lno_in_postimage = 1;
839 lc_a = count_lines(data_one, size_one);
841 lc_b = count_lines(data_two, size_two);
842 fprintf(o->file,
843 "%s%s--- %s%s%s\n%s%s+++ %s%s%s\n%s%s@@ -",
844 line_prefix, metainfo, a_name.buf, name_a_tab, reset,
845 line_prefix, metainfo, b_name.buf, name_b_tab, reset,
846 line_prefix, fraginfo);
847 if (!o->irreversible_delete)
848 print_line_count(o->file, lc_a);
849 else
850 fprintf(o->file, "?,?");
851 fprintf(o->file, " +");
852 print_line_count(o->file, lc_b);
853 fprintf(o->file, " @@%s\n", reset);
854 if (lc_a && !o->irreversible_delete)
855 emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
856 if (lc_b)
857 emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
858 if (textconv_one)
859 free((char *)data_one);
860 if (textconv_two)
861 free((char *)data_two);
862}
863struct diff_words_buffer {
865 mmfile_t text;
866 long alloc;
867 struct diff_words_orig {
868 const char *begin, *end;
869 } *orig;
870 int orig_nr, orig_alloc;
871};
872static void diff_words_append(char *line, unsigned long len,
874 struct diff_words_buffer *buffer)
875{
876 ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
877 line++;
878 len--;
879 memcpy(buffer->text.ptr + buffer->text.size, line, len);
880 buffer->text.size += len;
881 buffer->text.ptr[buffer->text.size] = '\0';
882}
883struct diff_words_style_elem {
885 const char *prefix;
886 const char *suffix;
887 const char *color; /* NULL; filled in by the setup code if
888 * color is enabled */
889};
890struct diff_words_style {
892 enum diff_words_type type;
893 struct diff_words_style_elem new, old, ctx;
894 const char *newline;
895};
896static struct diff_words_style diff_words_styles[] = {
898 { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
899 { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
900 { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
901};
902struct diff_words_data {
904 struct diff_words_buffer minus, plus;
905 const char *current_plus;
906 int last_minus;
907 struct diff_options *opt;
908 regex_t *word_regex;
909 enum diff_words_type type;
910 struct diff_words_style *style;
911};
912static int fn_out_diff_words_write_helper(FILE *fp,
914 struct diff_words_style_elem *st_el,
915 const char *newline,
916 size_t count, const char *buf,
917 const char *line_prefix)
918{
919 int print = 0;
920 while (count) {
922 char *p = memchr(buf, '\n', count);
923 if (print)
924 fputs(line_prefix, fp);
925 if (p != buf) {
926 if (st_el->color && fputs(st_el->color, fp) < 0)
927 return -1;
928 if (fputs(st_el->prefix, fp) < 0 ||
929 fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
930 fputs(st_el->suffix, fp) < 0)
931 return -1;
932 if (st_el->color && *st_el->color
933 && fputs(GIT_COLOR_RESET, fp) < 0)
934 return -1;
935 }
936 if (!p)
937 return 0;
938 if (fputs(newline, fp) < 0)
939 return -1;
940 count -= p + 1 - buf;
941 buf = p + 1;
942 print = 1;
943 }
944 return 0;
945}
946/*
948 * '--color-words' algorithm can be described as:
949 *
950 * 1. collect the minus/plus lines of a diff hunk, divided into
951 * minus-lines and plus-lines;
952 *
953 * 2. break both minus-lines and plus-lines into words and
954 * place them into two mmfile_t with one word for each line;
955 *
956 * 3. use xdiff to run diff on the two mmfile_t to get the words level diff;
957 *
958 * And for the common parts of the both file, we output the plus side text.
959 * diff_words->current_plus is used to trace the current position of the plus file
960 * which printed. diff_words->last_minus is used to trace the last minus word
961 * printed.
962 *
963 * For '--graph' to work with '--color-words', we need to output the graph prefix
964 * on each line of color words output. Generally, there are two conditions on
965 * which we should output the prefix.
966 *
967 * 1. diff_words->last_minus == 0 &&
968 * diff_words->current_plus == diff_words->plus.text.ptr
969 *
970 * that is: the plus text must start as a new line, and if there is no minus
971 * word printed, a graph prefix must be printed.
972 *
973 * 2. diff_words->current_plus > diff_words->plus.text.ptr &&
974 * *(diff_words->current_plus - 1) == '\n'
975 *
976 * that is: a graph prefix must be printed following a '\n'
977 */
978static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
979{
980 if ((diff_words->last_minus == 0 &&
981 diff_words->current_plus == diff_words->plus.text.ptr) ||
982 (diff_words->current_plus > diff_words->plus.text.ptr &&
983 *(diff_words->current_plus - 1) == '\n')) {
984 return 1;
985 } else {
986 return 0;
987 }
988}
989static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
991{
992 struct diff_words_data *diff_words = priv;
993 struct diff_words_style *style = diff_words->style;
994 int minus_first, minus_len, plus_first, plus_len;
995 const char *minus_begin, *minus_end, *plus_begin, *plus_end;
996 struct diff_options *opt = diff_words->opt;
997 const char *line_prefix;
998 if (line[0] != '@' || parse_hunk_header(line, len,
1000 &minus_first, &minus_len, &plus_first, &plus_len))
1001 return;
1002 assert(opt);
1004 line_prefix = diff_line_prefix(opt);
1005 /* POSIX requires that first be decremented by one if len == 0... */
1007 if (minus_len) {
1008 minus_begin = diff_words->minus.orig[minus_first].begin;
1009 minus_end =
1010 diff_words->minus.orig[minus_first + minus_len - 1].end;
1011 } else
1012 minus_begin = minus_end =
1013 diff_words->minus.orig[minus_first].end;
1014 if (plus_len) {
1016 plus_begin = diff_words->plus.orig[plus_first].begin;
1017 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
1018 } else
1019 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
1020 if (color_words_output_graph_prefix(diff_words)) {
1022 fputs(line_prefix, diff_words->opt->file);
1023 }
1024 if (diff_words->current_plus != plus_begin) {
1025 fn_out_diff_words_write_helper(diff_words->opt->file,
1026 &style->ctx, style->newline,
1027 plus_begin - diff_words->current_plus,
1028 diff_words->current_plus, line_prefix);
1029 if (*(plus_begin - 1) == '\n')
1030 fputs(line_prefix, diff_words->opt->file);
1031 }
1032 if (minus_begin != minus_end) {
1033 fn_out_diff_words_write_helper(diff_words->opt->file,
1034 &style->old, style->newline,
1035 minus_end - minus_begin, minus_begin,
1036 line_prefix);
1037 }
1038 if (plus_begin != plus_end) {
1039 fn_out_diff_words_write_helper(diff_words->opt->file,
1040 &style->new, style->newline,
1041 plus_end - plus_begin, plus_begin,
1042 line_prefix);
1043 }
1044 diff_words->current_plus = plus_end;
1046 diff_words->last_minus = minus_first;
1047}
1048/* This function starts looking at *begin, and returns 0 iff a word was found. */
1050static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
1051 int *begin, int *end)
1052{
1053 if (word_regex && *begin < buffer->size) {
1054 regmatch_t match[1];
1055 if (!regexec_buf(word_regex, buffer->ptr + *begin,
1056 buffer->size - *begin, 1, match, 0)) {
1057 char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
1058 '\n', match[0].rm_eo - match[0].rm_so);
1059 *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
1060 *begin += match[0].rm_so;
1061 return *begin >= *end;
1062 }
1063 return -1;
1064 }
1065 /* find the next word */
1067 while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
1068 (*begin)++;
1069 if (*begin >= buffer->size)
1070 return -1;
1071 /* find the end of the word */
1073 *end = *begin + 1;
1074 while (*end < buffer->size && !isspace(buffer->ptr[*end]))
1075 (*end)++;
1076 return 0;
1078}
1079/*
1081 * This function splits the words in buffer->text, stores the list with
1082 * newline separator into out, and saves the offsets of the original words
1083 * in buffer->orig.
1084 */
1085static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
1086 regex_t *word_regex)
1087{
1088 int i, j;
1089 long alloc = 0;
1090 out->size = 0;
1092 out->ptr = NULL;
1093 /* fake an empty "0th" word */
1095 ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
1096 buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
1097 buffer->orig_nr = 1;
1098 for (i = 0; i < buffer->text.size; i++) {
1100 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
1101 return;
1102 /* store original boundaries */
1104 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
1105 buffer->orig_alloc);
1106 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
1107 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
1108 buffer->orig_nr++;
1109 /* store one word */
1111 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
1112 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
1113 out->ptr[out->size + j - i] = '\n';
1114 out->size += j - i + 1;
1115 i = j - 1;
1117 }
1118}
1119/* this executes the word diff on the accumulated buffers */
1121static void diff_words_show(struct diff_words_data *diff_words)
1122{
1123 xpparam_t xpp;
1124 xdemitconf_t xecfg;
1125 mmfile_t minus, plus;
1126 struct diff_words_style *style = diff_words->style;
1127 struct diff_options *opt = diff_words->opt;
1129 const char *line_prefix;
1130 assert(opt);
1132 line_prefix = diff_line_prefix(opt);
1133 /* special case: only removal */
1135 if (!diff_words->plus.text.size) {
1136 fputs(line_prefix, diff_words->opt->file);
1137 fn_out_diff_words_write_helper(diff_words->opt->file,
1138 &style->old, style->newline,
1139 diff_words->minus.text.size,
1140 diff_words->minus.text.ptr, line_prefix);
1141 diff_words->minus.text.size = 0;
1142 return;
1143 }
1144 diff_words->current_plus = diff_words->plus.text.ptr;
1146 diff_words->last_minus = 0;
1147 memset(&xpp, 0, sizeof(xpp));
1149 memset(&xecfg, 0, sizeof(xecfg));
1150 diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
1151 diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
1152 xpp.flags = 0;
1153 /* as only the hunk header will be parsed, we need a 0-context */
1154 xecfg.ctxlen = 0;
1155 if (xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
1156 &xpp, &xecfg))
1157 die("unable to generate word diff");
1158 free(minus.ptr);
1159 free(plus.ptr);
1160 if (diff_words->current_plus != diff_words->plus.text.ptr +
1161 diff_words->plus.text.size) {
1162 if (color_words_output_graph_prefix(diff_words))
1163 fputs(line_prefix, diff_words->opt->file);
1164 fn_out_diff_words_write_helper(diff_words->opt->file,
1165 &style->ctx, style->newline,
1166 diff_words->plus.text.ptr + diff_words->plus.text.size
1167 - diff_words->current_plus, diff_words->current_plus,
1168 line_prefix);
1169 }
1170 diff_words->minus.text.size = diff_words->plus.text.size = 0;
1171}
1172/* In "color-words" mode, show word-diff of words accumulated in the buffer */
1174static void diff_words_flush(struct emit_callback *ecbdata)
1175{
1176 if (ecbdata->diff_words->minus.text.size ||
1177 ecbdata->diff_words->plus.text.size)
1178 diff_words_show(ecbdata->diff_words);
1179}
1180static void diff_filespec_load_driver(struct diff_filespec *one)
1182{
1183 /* Use already-loaded driver */
1184 if (one->driver)
1185 return;
1186 if (S_ISREG(one->mode))
1188 one->driver = userdiff_find_by_path(one->path);
1189 /* Fallback to default settings */
1191 if (!one->driver)
1192 one->driver = userdiff_find_by_name("default");
1193}
1194static const char *userdiff_word_regex(struct diff_filespec *one)
1196{
1197 diff_filespec_load_driver(one);
1198 return one->driver->word_regex;
1199}
1200static void init_diff_words_data(struct emit_callback *ecbdata,
1202 struct diff_options *orig_opts,
1203 struct diff_filespec *one,
1204 struct diff_filespec *two)
1205{
1206 int i;
1207 struct diff_options *o = xmalloc(sizeof(struct diff_options));
1208 memcpy(o, orig_opts, sizeof(struct diff_options));
1209 ecbdata->diff_words =
1211 xcalloc(1, sizeof(struct diff_words_data));
1212 ecbdata->diff_words->type = o->word_diff;
1213 ecbdata->diff_words->opt = o;
1214 if (!o->word_regex)
1215 o->word_regex = userdiff_word_regex(one);
1216 if (!o->word_regex)
1217 o->word_regex = userdiff_word_regex(two);
1218 if (!o->word_regex)
1219 o->word_regex = diff_word_regex_cfg;
1220 if (o->word_regex) {
1221 ecbdata->diff_words->word_regex = (regex_t *)
1222 xmalloc(sizeof(regex_t));
1223 if (regcomp(ecbdata->diff_words->word_regex,
1224 o->word_regex,
1225 REG_EXTENDED | REG_NEWLINE))
1226 die ("Invalid regular expression: %s",
1227 o->word_regex);
1228 }
1229 for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1230 if (o->word_diff == diff_words_styles[i].type) {
1231 ecbdata->diff_words->style =
1232 &diff_words_styles[i];
1233 break;
1234 }
1235 }
1236 if (want_color(o->use_color)) {
1237 struct diff_words_style *st = ecbdata->diff_words->style;
1238 st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1239 st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1240 st->ctx.color = diff_get_color_opt(o, DIFF_CONTEXT);
1241 }
1242}
1243static void free_diff_words_data(struct emit_callback *ecbdata)
1245{
1246 if (ecbdata->diff_words) {
1247 diff_words_flush(ecbdata);
1248 free (ecbdata->diff_words->opt);
1249 free (ecbdata->diff_words->minus.text.ptr);
1250 free (ecbdata->diff_words->minus.orig);
1251 free (ecbdata->diff_words->plus.text.ptr);
1252 free (ecbdata->diff_words->plus.orig);
1253 if (ecbdata->diff_words->word_regex) {
1254 regfree(ecbdata->diff_words->word_regex);
1255 free(ecbdata->diff_words->word_regex);
1256 }
1257 FREE_AND_NULL(ecbdata->diff_words);
1258 }
1259}
1260const char *diff_get_color(int diff_use_color, enum color_diff ix)
1262{
1263 if (want_color(diff_use_color))
1264 return diff_colors[ix];
1265 return "";
1266}
1267const char *diff_line_prefix(struct diff_options *opt)
1269{
1270 struct strbuf *msgbuf;
1271 if (!opt->output_prefix)
1272 return "";
1273 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
1275 return msgbuf->buf;
1276}
1277static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1279{
1280 const char *cp;
1281 unsigned long allot;
1282 size_t l = len;
1283 if (ecb->truncate)
1285 return ecb->truncate(line, len);
1286 cp = line;
1287 allot = l;
1288 while (0 < l) {
1289 (void) utf8_width(&cp, &l);
1290 if (!cp)
1291 break; /* truncated in the middle? */
1292 }
1293 return allot - l;
1294}
1295static void find_lno(const char *line, struct emit_callback *ecbdata)
1297{
1298 const char *p;
1299 ecbdata->lno_in_preimage = 0;
1300 ecbdata->lno_in_postimage = 0;
1301 p = strchr(line, '-');
1302 if (!p)
1303 return; /* cannot happen */
1304 ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1305 p = strchr(p, '+');
1306 if (!p)
1307 return; /* cannot happen */
1308 ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1309}
1310static void fn_out_consume(void *priv, char *line, unsigned long len)
1312{
1313 struct emit_callback *ecbdata = priv;
1314 const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
1315 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
1316 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1317 struct diff_options *o = ecbdata->opt;
1318 const char *line_prefix = diff_line_prefix(o);
1319 o->found_changes = 1;
1321 if (ecbdata->header) {
1323 fprintf(o->file, "%s", ecbdata->header->buf);
1324 strbuf_reset(ecbdata->header);
1325 ecbdata->header = NULL;
1326 }
1327 if (ecbdata->label_path[0]) {
1329 const char *name_a_tab, *name_b_tab;
1330 name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
1332 name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
1333 fprintf(o->file, "%s%s--- %s%s%s\n",
1335 line_prefix, meta, ecbdata->label_path[0], reset, name_a_tab);
1336 fprintf(o->file, "%s%s+++ %s%s%s\n",
1337 line_prefix, meta, ecbdata->label_path[1], reset, name_b_tab);
1338 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1339 }
1340 if (diff_suppress_blank_empty
1342 && len == 2 && line[0] == ' ' && line[1] == '\n') {
1343 line[0] = '\n';
1344 len = 1;
1345 }
1346 if (line[0] == '@') {
1348 if (ecbdata->diff_words)
1349 diff_words_flush(ecbdata);
1350 len = sane_truncate_line(ecbdata, line, len);
1351 find_lno(line, ecbdata);
1352 emit_hunk_header(ecbdata, line, len);
1353 return;
1354 }
1355 if (ecbdata->diff_words) {
1357 if (line[0] == '-') {
1358 diff_words_append(line, len,
1359 &ecbdata->diff_words->minus);
1360 return;
1361 } else if (line[0] == '+') {
1362 diff_words_append(line, len,
1363 &ecbdata->diff_words->plus);
1364 return;
1365 } else if (starts_with(line, "\\ ")) {
1366 /*
1367 * Eat the "no newline at eof" marker as if we
1368 * saw a "+" or "-" line with nothing on it,
1369 * and return without diff_words_flush() to
1370 * defer processing. If this is the end of
1371 * preimage, more "+" lines may come after it.
1372 */
1373 return;
1374 }
1375 diff_words_flush(ecbdata);
1376 if (ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN) {
1377 emit_line(o, context, reset, line, len);
1378 fputs("~\n", o->file);
1379 } else {
1380 /*
1381 * Skip the prefix character, if any. With
1382 * diff_suppress_blank_empty, there may be
1383 * none.
1384 */
1385 if (line[0] != '\n') {
1386 line++;
1387 len--;
1388 }
1389 emit_line(o, context, reset, line, len);
1390 }
1391 return;
1392 }
1393 switch (line[0]) {
1395 case '+':
1396 ecbdata->lno_in_postimage++;
1397 emit_add_line(reset, ecbdata, line + 1, len - 1);
1398 break;
1399 case '-':
1400 ecbdata->lno_in_preimage++;
1401 emit_del_line(reset, ecbdata, line + 1, len - 1);
1402 break;
1403 case ' ':
1404 ecbdata->lno_in_postimage++;
1405 ecbdata->lno_in_preimage++;
1406 emit_context_line(reset, ecbdata, line + 1, len - 1);
1407 break;
1408 default:
1409 /* incomplete line at the end */
1410 ecbdata->lno_in_preimage++;
1411 emit_line(o, diff_get_color(ecbdata->color_diff, DIFF_CONTEXT),
1412 reset, line, len);
1413 break;
1414 }
1415}
1416static char *pprint_rename(const char *a, const char *b)
1418{
1419 const char *old = a;
1420 const char *new = b;
1421 struct strbuf name = STRBUF_INIT;
1422 int pfx_length, sfx_length;
1423 int pfx_adjust_for_slash;
1424 int len_a = strlen(a);
1425 int len_b = strlen(b);
1426 int a_midlen, b_midlen;
1427 int qlen_a = quote_c_style(a, NULL, NULL, 0);
1428 int qlen_b = quote_c_style(b, NULL, NULL, 0);
1429 if (qlen_a || qlen_b) {
1431 quote_c_style(a, &name, NULL, 0);
1432 strbuf_addstr(&name, " => ");
1433 quote_c_style(b, &name, NULL, 0);
1434 return strbuf_detach(&name, NULL);
1435 }
1436 /* Find common prefix */
1438 pfx_length = 0;
1439 while (*old && *new && *old == *new) {
1440 if (*old == '/')
1441 pfx_length = old - a + 1;
1442 old++;
1443 new++;
1444 }
1445 /* Find common suffix */
1447 old = a + len_a;
1448 new = b + len_b;
1449 sfx_length = 0;
1450 /*
1451 * If there is a common prefix, it must end in a slash. In
1452 * that case we let this loop run 1 into the prefix to see the
1453 * same slash.
1454 *
1455 * If there is no common prefix, we cannot do this as it would
1456 * underrun the input strings.
1457 */
1458 pfx_adjust_for_slash = (pfx_length ? 1 : 0);
1459 while (a + pfx_length - pfx_adjust_for_slash <= old &&
1460 b + pfx_length - pfx_adjust_for_slash <= new &&
1461 *old == *new) {
1462 if (*old == '/')
1463 sfx_length = len_a - (old - a);
1464 old--;
1465 new--;
1466 }
1467 /*
1469 * pfx{mid-a => mid-b}sfx
1470 * {pfx-a => pfx-b}sfx
1471 * pfx{sfx-a => sfx-b}
1472 * name-a => name-b
1473 */
1474 a_midlen = len_a - pfx_length - sfx_length;
1475 b_midlen = len_b - pfx_length - sfx_length;
1476 if (a_midlen < 0)
1477 a_midlen = 0;
1478 if (b_midlen < 0)
1479 b_midlen = 0;
1480 strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1482 if (pfx_length + sfx_length) {
1483 strbuf_add(&name, a, pfx_length);
1484 strbuf_addch(&name, '{');
1485 }
1486 strbuf_add(&name, a + pfx_length, a_midlen);
1487 strbuf_addstr(&name, " => ");
1488 strbuf_add(&name, b + pfx_length, b_midlen);
1489 if (pfx_length + sfx_length) {
1490 strbuf_addch(&name, '}');
1491 strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1492 }
1493 return strbuf_detach(&name, NULL);
1494}
1495struct diffstat_t {
1497 int nr;
1498 int alloc;
1499 struct diffstat_file {
1500 char *from_name;
1501 char *name;
1502 char *print_name;
1503 unsigned is_unmerged:1;
1504 unsigned is_binary:1;
1505 unsigned is_renamed:1;
1506 unsigned is_interesting:1;
1507 uintmax_t added, deleted;
1508 } **files;
1509};
1510static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1512 const char *name_a,
1513 const char *name_b)
1514{
1515 struct diffstat_file *x;
1516 x = xcalloc(1, sizeof(*x));
1517 ALLOC_GROW(diffstat->files, diffstat->nr + 1, diffstat->alloc);
1518 diffstat->files[diffstat->nr++] = x;
1519 if (name_b) {
1520 x->from_name = xstrdup(name_a);
1521 x->name = xstrdup(name_b);
1522 x->is_renamed = 1;
1523 }
1524 else {
1525 x->from_name = NULL;
1526 x->name = xstrdup(name_a);
1527 }
1528 return x;
1529}
1530static void diffstat_consume(void *priv, char *line, unsigned long len)
1532{
1533 struct diffstat_t *diffstat = priv;
1534 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1535 if (line[0] == '+')
1537 x->added++;
1538 else if (line[0] == '-')
1539 x->deleted++;
1540}
1541const char mime_boundary_leader[] = "------------";
1543static int scale_linear(int it, int width, int max_change)
1545{
1546 if (!it)
1547 return 0;
1548 /*
1549 * make sure that at least one '-' or '+' is printed if
1550 * there is any change to this path. The easiest way is to
1551 * scale linearly as if the alloted width is one column shorter
1552 * than it is, and then add 1 to the result.
1553 */
1554 return 1 + (it * (width - 1) / max_change);
1555}
1556static void show_name(FILE *file,
1558 const char *prefix, const char *name, int len)
1559{
1560 fprintf(file, " %s%-*s |", prefix, len, name);
1561}
1562static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1564{
1565 if (cnt <= 0)
1566 return;
1567 fprintf(file, "%s", set);
1568 while (cnt--)
1569 putc(ch, file);
1570 fprintf(file, "%s", reset);
1571}
1572static void fill_print_name(struct diffstat_file *file)
1574{
1575 char *pname;
1576 if (file->print_name)
1578 return;
1579 if (!file->is_renamed) {
1581 struct strbuf buf = STRBUF_INIT;
1582 if (quote_c_style(file->name, &buf, NULL, 0)) {
1583 pname = strbuf_detach(&buf, NULL);
1584 } else {
1585 pname = file->name;
1586 strbuf_release(&buf);
1587 }
1588 } else {
1589 pname = pprint_rename(file->from_name, file->name);
1590 }
1591 file->print_name = pname;
1592}
1593int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1595{
1596 struct strbuf sb = STRBUF_INIT;
1597 int ret;
1598 if (!files) {
1600 assert(insertions == 0 && deletions == 0);
1601 return fprintf(fp, "%s\n", " 0 files changed");
1602 }
1603 strbuf_addf(&sb,
1605 (files == 1) ? " %d file changed" : " %d files changed",
1606 files);
1607 /*
1609 * For binary diff, the caller may want to print "x files
1610 * changed" with insertions == 0 && deletions == 0.
1611 *
1612 * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1613 * is probably less confusing (i.e skip over "2 files changed
1614 * but nothing about added/removed lines? Is this a bug in Git?").
1615 */
1616 if (insertions || deletions == 0) {
1617 strbuf_addf(&sb,
1618 (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
1619 insertions);
1620 }
1621 if (deletions || insertions == 0) {
1623 strbuf_addf(&sb,
1624 (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
1625 deletions);
1626 }
1627 strbuf_addch(&sb, '\n');
1628 ret = fputs(sb.buf, fp);
1629 strbuf_release(&sb);
1630 return ret;
1631}
1632static void show_stats(struct diffstat_t *data, struct diff_options *options)
1634{
1635 int i, len, add, del, adds = 0, dels = 0;
1636 uintmax_t max_change = 0, max_len = 0;
1637 int total_files = data->nr, count;
1638 int width, name_width, graph_width, number_width = 0, bin_width = 0;
1639 const char *reset, *add_c, *del_c;
1640 const char *line_prefix = "";
1641 int extra_shown = 0;
1642 if (data->nr == 0)
1644 return;
1645 line_prefix = diff_line_prefix(options);
1647 count = options->stat_count ? options->stat_count : data->nr;
1648 reset = diff_get_color_opt(options, DIFF_RESET);
1650 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1651 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1652 /*
1654 * Find the longest filename and max number of changes
1655 */
1656 for (i = 0; (i < count) && (i < data->nr); i++) {
1657 struct diffstat_file *file = data->files[i];
1658 uintmax_t change = file->added + file->deleted;
1659 if (!file->is_interesting && (change == 0)) {
1661 count++; /* not shown == room for one more */
1662 continue;
1663 }
1664 fill_print_name(file);
1665 len = strlen(file->print_name);
1666 if (max_len < len)
1667 max_len = len;
1668 if (file->is_unmerged) {
1670 /* "Unmerged" is 8 characters */
1671 bin_width = bin_width < 8 ? 8 : bin_width;
1672 continue;
1673 }
1674 if (file->is_binary) {
1675 /* "Bin XXX -> YYY bytes" */
1676 int w = 14 + decimal_width(file->added)
1677 + decimal_width(file->deleted);
1678 bin_width = bin_width < w ? w : bin_width;
1679 /* Display change counts aligned with "Bin" */
1680 number_width = 3;
1681 continue;
1682 }
1683 if (max_change < change)
1685 max_change = change;
1686 }
1687 count = i; /* where we can stop scanning in data->files[] */
1688 /*
1690 * We have width = stat_width or term_columns() columns total.
1691 * We want a maximum of min(max_len, stat_name_width) for the name part.
1692 * We want a maximum of min(max_change, stat_graph_width) for the +- part.
1693 * We also need 1 for " " and 4 + decimal_width(max_change)
1694 * for " | NNNN " and one the empty column at the end, altogether
1695 * 6 + decimal_width(max_change).
1696 *
1697 * If there's not enough space, we will use the smaller of
1698 * stat_name_width (if set) and 5/8*width for the filename,
1699 * and the rest for constant elements + graph part, but no more
1700 * than stat_graph_width for the graph part.
1701 * (5/8 gives 50 for filename and 30 for the constant parts + graph
1702 * for the standard terminal size).
1703 *
1704 * In other words: stat_width limits the maximum width, and
1705 * stat_name_width fixes the maximum width of the filename,
1706 * and is also used to divide available columns if there
1707 * aren't enough.
1708 *
1709 * Binary files are displayed with "Bin XXX -> YYY bytes"
1710 * instead of the change count and graph. This part is treated
1711 * similarly to the graph part, except that it is not
1712 * "scaled". If total width is too small to accommodate the
1713 * guaranteed minimum width of the filename part and the
1714 * separators and this message, this message will "overflow"
1715 * making the line longer than the maximum width.
1716 */
1717 if (options->stat_width == -1)
1719 width = term_columns() - strlen(line_prefix);
1720 else
1721 width = options->stat_width ? options->stat_width : 80;
1722 number_width = decimal_width(max_change) > number_width ?
1723 decimal_width(max_change) : number_width;
1724 if (options->stat_graph_width == -1)
1726 options->stat_graph_width = diff_stat_graph_width;
1727 /*
1729 * Guarantee 3/8*16==6 for the graph part
1730 * and 5/8*16==10 for the filename part
1731 */
1732 if (width < 16 + 6 + number_width)
1733 width = 16 + 6 + number_width;
1734 /*
1736 * First assign sizes that are wanted, ignoring available width.
1737 * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
1738 * starting from "XXX" should fit in graph_width.
1739 */
1740 graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
1741 if (options->stat_graph_width &&
1742 options->stat_graph_width < graph_width)
1743 graph_width = options->stat_graph_width;
1744 name_width = (options->stat_name_width > 0 &&
1746 options->stat_name_width < max_len) ?
1747 options->stat_name_width : max_len;
1748 /*
1750 * Adjust adjustable widths not to exceed maximum width
1751 */
1752 if (name_width + number_width + 6 + graph_width > width) {
1753 if (graph_width > width * 3/8 - number_width - 6) {
1754 graph_width = width * 3/8 - number_width - 6;
1755 if (graph_width < 6)
1756 graph_width = 6;
1757 }
1758 if (options->stat_graph_width &&
1760 graph_width > options->stat_graph_width)
1761 graph_width = options->stat_graph_width;
1762 if (name_width > width - number_width - 6 - graph_width)
1763 name_width = width - number_width - 6 - graph_width;
1764 else
1765 graph_width = width - number_width - 6 - name_width;
1766 }
1767 /*
1769 * From here name_width is the width of the name area,
1770 * and graph_width is the width of the graph area.
1771 * max_change is used to scale graph properly.
1772 */
1773 for (i = 0; i < count; i++) {
1774 const char *prefix = "";
1775 struct diffstat_file *file = data->files[i];
1776 char *name = file->print_name;
1777 uintmax_t added = file->added;
1778 uintmax_t deleted = file->deleted;
1779 int name_len;
1780 if (!file->is_interesting && (added + deleted == 0))
1782 continue;
1783 /*
1785 * "scale" the filename
1786 */
1787 len = name_width;
1788 name_len = strlen(name);
1789 if (name_width < name_len) {
1790 char *slash;
1791 prefix = "...";
1792 len -= 3;
1793 name += name_len - len;
1794 slash = strchr(name, '/');
1795 if (slash)
1796 name = slash;
1797 }
1798 if (file->is_binary) {
1800 fprintf(options->file, "%s", line_prefix);
1801 show_name(options->file, prefix, name, len);
1802 fprintf(options->file, " %*s", number_width, "Bin");
1803 if (!added && !deleted) {
1804 putc('\n', options->file);
1805 continue;
1806 }
1807 fprintf(options->file, " %s%"PRIuMAX"%s",
1808 del_c, deleted, reset);
1809 fprintf(options->file, " -> ");
1810 fprintf(options->file, "%s%"PRIuMAX"%s",
1811 add_c, added, reset);
1812 fprintf(options->file, " bytes");
1813 fprintf(options->file, "\n");
1814 continue;
1815 }
1816 else if (file->is_unmerged) {
1817 fprintf(options->file, "%s", line_prefix);
1818 show_name(options->file, prefix, name, len);
1819 fprintf(options->file, " Unmerged\n");
1820 continue;
1821 }
1822 /*
1824 * scale the add/delete
1825 */
1826 add = added;
1827 del = deleted;
1828 if (graph_width <= max_change) {
1830 int total = scale_linear(add + del, graph_width, max_change);
1831 if (total < 2 && add && del)
1832 /* width >= 2 due to the sanity check */
1833 total = 2;
1834 if (add < del) {
1835 add = scale_linear(add, graph_width, max_change);
1836 del = total - add;
1837 } else {
1838 del = scale_linear(del, graph_width, max_change);
1839 add = total - del;
1840 }
1841 }
1842 fprintf(options->file, "%s", line_prefix);
1843 show_name(options->file, prefix, name, len);
1844 fprintf(options->file, " %*"PRIuMAX"%s",
1845 number_width, added + deleted,
1846 added + deleted ? " " : "");
1847 show_graph(options->file, '+', add, add_c, reset);
1848 show_graph(options->file, '-', del, del_c, reset);
1849 fprintf(options->file, "\n");
1850 }
1851 for (i = 0; i < data->nr; i++) {
1853 struct diffstat_file *file = data->files[i];
1854 uintmax_t added = file->added;
1855 uintmax_t deleted = file->deleted;
1856 if (file->is_unmerged ||
1858 (!file->is_interesting && (added + deleted == 0))) {
1859 total_files--;
1860 continue;
1861 }
1862 if (!file->is_binary) {
1864 adds += added;
1865 dels += deleted;
1866 }
1867 if (i < count)
1868 continue;
1869 if (!extra_shown)
1870 fprintf(options->file, "%s ...\n", line_prefix);
1871 extra_shown = 1;
1872 }
1873 fprintf(options->file, "%s", line_prefix);
1874 print_stat_summary(options->file, total_files, adds, dels);
1875}
1876static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1878{
1879 int i, adds = 0, dels = 0, total_files = data->nr;
1880 if (data->nr == 0)
1882 return;
1883 for (i = 0; i < data->nr; i++) {
1885 int added = data->files[i]->added;
1886 int deleted= data->files[i]->deleted;
1887 if (data->files[i]->is_unmerged ||
1889 (!data->files[i]->is_interesting && (added + deleted == 0))) {
1890 total_files--;
1891 } else if (!data->files[i]->is_binary) { /* don't count bytes */
1892 adds += added;
1893 dels += deleted;
1894 }
1895 }
1896 fprintf(options->file, "%s", diff_line_prefix(options));
1897 print_stat_summary(options->file, total_files, adds, dels);
1898}
1899static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1901{
1902 int i;
1903 if (data->nr == 0)
1905 return;
1906 for (i = 0; i < data->nr; i++) {
1908 struct diffstat_file *file = data->files[i];
1909 fprintf(options->file, "%s", diff_line_prefix(options));
1911 if (file->is_binary)
1913 fprintf(options->file, "-\t-\t");
1914 else
1915 fprintf(options->file,
1916 "%"PRIuMAX"\t%"PRIuMAX"\t",
1917 file->added, file->deleted);
1918 if (options->line_termination) {
1919 fill_print_name(file);
1920 if (!file->is_renamed)
1921 write_name_quoted(file->name, options->file,
1922 options->line_termination);
1923 else {
1924 fputs(file->print_name, options->file);
1925 putc(options->line_termination, options->file);
1926 }
1927 } else {
1928 if (file->is_renamed) {
1929 putc('\0', options->file);
1930 write_name_quoted(file->from_name, options->file, '\0');
1931 }
1932 write_name_quoted(file->name, options->file, '\0');
1933 }
1934 }
1935}
1936struct dirstat_file {
1938 const char *name;
1939 unsigned long changed;
1940};
1941struct dirstat_dir {
1943 struct dirstat_file *files;
1944 int alloc, nr, permille, cumulative;
1945};
1946static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1948 unsigned long changed, const char *base, int baselen)
1949{
1950 unsigned long this_dir = 0;
1951 unsigned int sources = 0;
1952 const char *line_prefix = diff_line_prefix(opt);
1953 while (dir->nr) {
1955 struct dirstat_file *f = dir->files;
1956 int namelen = strlen(f->name);
1957 unsigned long this;
1958 char *slash;
1959 if (namelen < baselen)
1961 break;
1962 if (memcmp(f->name, base, baselen))
1963 break;
1964 slash = strchr(f->name + baselen, '/');
1965 if (slash) {
1966 int newbaselen = slash + 1 - f->name;
1967 this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
1968 sources++;
1969 } else {
1970 this = f->changed;
1971 dir->files++;
1972 dir->nr--;
1973 sources += 2;
1974 }
1975 this_dir += this;
1976 }
1977 /*
1979 * We don't report dirstat's for
1980 * - the top level
1981 * - or cases where everything came from a single directory
1982 * under this directory (sources == 1).
1983 */
1984 if (baselen && sources != 1) {
1985 if (this_dir) {
1986 int permille = this_dir * 1000 / changed;
1987 if (permille >= dir->permille) {
1988 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
1989 permille / 10, permille % 10, baselen, base);
1990 if (!dir->cumulative)
1991 return 0;
1992 }
1993 }
1994 }
1995 return this_dir;
1996}
1997static int dirstat_compare(const void *_a, const void *_b)
1999{
2000 const struct dirstat_file *a = _a;
2001 const struct dirstat_file *b = _b;
2002 return strcmp(a->name, b->name);
2003}
2004static void show_dirstat(struct diff_options *options)
2006{
2007 int i;
2008 unsigned long changed;
2009 struct dirstat_dir dir;
2010 struct diff_queue_struct *q = &diff_queued_diff;
2011 dir.files = NULL;
2013 dir.alloc = 0;
2014 dir.nr = 0;
2015 dir.permille = options->dirstat_permille;
2016 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
2017 changed = 0;
2019 for (i = 0; i < q->nr; i++) {
2020 struct diff_filepair *p = q->queue[i];
2021 const char *name;
2022 unsigned long copied, added, damage;
2023 int content_changed;
2024 name = p->two->path ? p->two->path : p->one->path;
2026 if (p->one->oid_valid && p->two->oid_valid)
2028 content_changed = oidcmp(&p->one->oid, &p->two->oid);
2029 else
2030 content_changed = 1;
2031 if (!content_changed) {
2033 /*
2034 * The SHA1 has not changed, so pre-/post-content is
2035 * identical. We can therefore skip looking at the
2036 * file contents altogether.
2037 */
2038 damage = 0;
2039 goto found_damage;
2040 }
2041 if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
2043 /*
2044 * In --dirstat-by-file mode, we don't really need to
2045 * look at the actual file contents at all.
2046 * The fact that the SHA1 changed is enough for us to
2047 * add this file to the list of results
2048 * (with each file contributing equal damage).
2049 */
2050 damage = 1;
2051 goto found_damage;
2052 }
2053 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
2055 diff_populate_filespec(p->one, 0);
2056 diff_populate_filespec(p->two, 0);
2057 diffcore_count_changes(p->one, p->two, NULL, NULL,
2058 &copied, &added);
2059 diff_free_filespec_data(p->one);
2060 diff_free_filespec_data(p->two);
2061 } else if (DIFF_FILE_VALID(p->one)) {
2062 diff_populate_filespec(p->one, CHECK_SIZE_ONLY);
2063 copied = added = 0;
2064 diff_free_filespec_data(p->one);
2065 } else if (DIFF_FILE_VALID(p->two)) {
2066 diff_populate_filespec(p->two, CHECK_SIZE_ONLY);
2067 copied = 0;
2068 added = p->two->size;
2069 diff_free_filespec_data(p->two);
2070 } else
2071 continue;
2072 /*
2074 * Original minus copied is the removed material,
2075 * added is the new material. They are both damages
2076 * made to the preimage.
2077 * If the resulting damage is zero, we know that
2078 * diffcore_count_changes() considers the two entries to
2079 * be identical, but since content_changed is true, we
2080 * know that there must have been _some_ kind of change,
2081 * so we force all entries to have damage > 0.
2082 */
2083 damage = (p->one->size - copied) + added;
2084 if (!damage)
2085 damage = 1;
2086found_damage:
2088 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2089 dir.files[dir.nr].name = name;
2090 dir.files[dir.nr].changed = damage;
2091 changed += damage;
2092 dir.nr++;
2093 }
2094 /* This can happen even with many files, if everything was renames */
2096 if (!changed)
2097 return;
2098 /* Show all directories with more than x% of the changes */
2100 QSORT(dir.files, dir.nr, dirstat_compare);
2101 gather_dirstat(options, &dir, changed, "", 0);
2102}
2103static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
2105{
2106 int i;
2107 unsigned long changed;
2108 struct dirstat_dir dir;
2109 if (data->nr == 0)
2111 return;
2112 dir.files = NULL;
2114 dir.alloc = 0;
2115 dir.nr = 0;
2116 dir.permille = options->dirstat_permille;
2117 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
2118 changed = 0;
2120 for (i = 0; i < data->nr; i++) {
2121 struct diffstat_file *file = data->files[i];
2122 unsigned long damage = file->added + file->deleted;
2123 if (file->is_binary)
2124 /*
2125 * binary files counts bytes, not lines. Must find some
2126 * way to normalize binary bytes vs. textual lines.
2127 * The following heuristic assumes that there are 64
2128 * bytes per "line".
2129 * This is stupid and ugly, but very cheap...
2130 */
2131 damage = (damage + 63) / 64;
2132 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2133 dir.files[dir.nr].name = file->name;
2134 dir.files[dir.nr].changed = damage;
2135 changed += damage;
2136 dir.nr++;
2137 }
2138 /* This can happen even with many files, if everything was renames */
2140 if (!changed)
2141 return;
2142 /* Show all directories with more than x% of the changes */
2144 QSORT(dir.files, dir.nr, dirstat_compare);
2145 gather_dirstat(options, &dir, changed, "", 0);
2146}
2147static void free_diffstat_info(struct diffstat_t *diffstat)
2149{
2150 int i;
2151 for (i = 0; i < diffstat->nr; i++) {
2152 struct diffstat_file *f = diffstat->files[i];
2153 if (f->name != f->print_name)
2154 free(f->print_name);
2155 free(f->name);
2156 free(f->from_name);
2157 free(f);
2158 }
2159 free(diffstat->files);
2160}
2161struct checkdiff_t {
2163 const char *filename;
2164 int lineno;
2165 int conflict_marker_size;
2166 struct diff_options *o;
2167 unsigned ws_rule;
2168 unsigned status;
2169};
2170static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2172{
2173 char firstchar;
2174 int cnt;
2175 if (len < marker_size + 1)
2177 return 0;
2178 firstchar = line[0];
2179 switch (firstchar) {
2180 case '=': case '>': case '<': case '|':
2181 break;
2182 default:
2183 return 0;
2184 }
2185 for (cnt = 1; cnt < marker_size; cnt++)
2186 if (line[cnt] != firstchar)
2187 return 0;
2188 /* line[1] thru line[marker_size-1] are same as firstchar */
2189 if (len < marker_size + 1 || !isspace(line[marker_size]))
2190 return 0;
2191 return 1;
2192}
2193static void checkdiff_consume(void *priv, char *line, unsigned long len)
2195{
2196 struct checkdiff_t *data = priv;
2197 int marker_size = data->conflict_marker_size;
2198 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2199 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2200 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2201 char *err;
2202 const char *line_prefix;
2203 assert(data->o);
2205 line_prefix = diff_line_prefix(data->o);
2206 if (line[0] == '+') {
2208 unsigned bad;
2209 data->lineno++;
2210 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2211 data->status |= 1;
2212 fprintf(data->o->file,
2213 "%s%s:%d: leftover conflict marker\n",
2214 line_prefix, data->filename, data->lineno);
2215 }
2216 bad = ws_check(line + 1, len - 1, data->ws_rule);
2217 if (!bad)
2218 return;
2219 data->status |= bad;
2220 err = whitespace_error_string(bad);
2221 fprintf(data->o->file, "%s%s:%d: %s.\n",
2222 line_prefix, data->filename, data->lineno, err);
2223 free(err);
2224 emit_line(data->o, set, reset, line, 1);
2225 ws_check_emit(line + 1, len - 1, data->ws_rule,
2226 data->o->file, set, reset, ws);
2227 } else if (line[0] == ' ') {
2228 data->lineno++;
2229 } else if (line[0] == '@') {
2230 char *plus = strchr(line, '+');
2231 if (plus)
2232 data->lineno = strtol(plus, NULL, 10) - 1;
2233 else
2234 die("invalid diff");
2235 }
2236}
2237static unsigned char *deflate_it(char *data,
2239 unsigned long size,
2240 unsigned long *result_size)
2241{
2242 int bound;
2243 unsigned char *deflated;
2244 git_zstream stream;
2245 git_deflate_init(&stream, zlib_compression_level);
2247 bound = git_deflate_bound(&stream, size);
2248 deflated = xmalloc(bound);
2249 stream.next_out = deflated;
2250 stream.avail_out = bound;
2251 stream.next_in = (unsigned char *)data;
2253 stream.avail_in = size;
2254 while (git_deflate(&stream, Z_FINISH) == Z_OK)
2255 ; /* nothing */
2256 git_deflate_end(&stream);
2257 *result_size = stream.total_out;
2258 return deflated;
2259}
2260static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two,
2262 const char *prefix)
2263{
2264 void *cp;
2265 void *delta;
2266 void *deflated;
2267 void *data;
2268 unsigned long orig_size;
2269 unsigned long delta_size;
2270 unsigned long deflate_size;
2271 unsigned long data_size;
2272 /* We could do deflated delta, or we could do just deflated two,
2274 * whichever is smaller.
2275 */
2276 delta = NULL;
2277 deflated = deflate_it(two->ptr, two->size, &deflate_size);
2278 if (one->size && two->size) {
2279 delta = diff_delta(one->ptr, one->size,
2280 two->ptr, two->size,
2281 &delta_size, deflate_size);
2282 if (delta) {
2283 void *to_free = delta;
2284 orig_size = delta_size;
2285 delta = deflate_it(delta, delta_size, &delta_size);
2286 free(to_free);
2287 }
2288 }
2289 if (delta && delta_size < deflate_size) {
2291 fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2292 free(deflated);
2293 data = delta;
2294 data_size = delta_size;
2295 }
2296 else {
2297 fprintf(file, "%sliteral %lu\n", prefix, two->size);
2298 free(delta);
2299 data = deflated;
2300 data_size = deflate_size;
2301 }
2302 /* emit data encoded in base85 */
2304 cp = data;
2305 while (data_size) {
2306 int bytes = (52 < data_size) ? 52 : data_size;
2307 char line[70];
2308 data_size -= bytes;
2309 if (bytes <= 26)
2310 line[0] = bytes + 'A' - 1;
2311 else
2312 line[0] = bytes - 26 + 'a' - 1;
2313 encode_85(line + 1, cp, bytes);
2314 cp = (char *) cp + bytes;
2315 fprintf(file, "%s", prefix);
2316 fputs(line, file);
2317 fputc('\n', file);
2318 }
2319 fprintf(file, "%s\n", prefix);
2320 free(data);
2321}
2322static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two,
2324 const char *prefix)
2325{
2326 fprintf(file, "%sGIT binary patch\n", prefix);
2327 emit_binary_diff_body(file, one, two, prefix);
2328 emit_binary_diff_body(file, two, one, prefix);
2329}
2330int diff_filespec_is_binary(struct diff_filespec *one)
2332{
2333 if (one->is_binary == -1) {
2334 diff_filespec_load_driver(one);
2335 if (one->driver->binary != -1)
2336 one->is_binary = one->driver->binary;
2337 else {
2338 if (!one->data && DIFF_FILE_VALID(one))
2339 diff_populate_filespec(one, CHECK_BINARY);
2340 if (one->is_binary == -1 && one->data)
2341 one->is_binary = buffer_is_binary(one->data,
2342 one->size);
2343 if (one->is_binary == -1)
2344 one->is_binary = 0;
2345 }
2346 }
2347 return one->is_binary;
2348}
2349static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2351{
2352 diff_filespec_load_driver(one);
2353 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2354}
2355void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2357{
2358 if (!options->a_prefix)
2359 options->a_prefix = a;
2360 if (!options->b_prefix)
2361 options->b_prefix = b;
2362}
2363struct userdiff_driver *get_textconv(struct diff_filespec *one)
2365{
2366 if (!DIFF_FILE_VALID(one))
2367 return NULL;
2368 diff_filespec_load_driver(one);
2370 return userdiff_get_textconv(one->driver);
2371}
2372static void builtin_diff(const char *name_a,
2374 const char *name_b,
2375 struct diff_filespec *one,
2376 struct diff_filespec *two,
2377 const char *xfrm_msg,
2378 int must_show_header,
2379 struct diff_options *o,
2380 int complete_rewrite)
2381{
2382 mmfile_t mf1, mf2;
2383 const char *lbl[2];
2384 char *a_one, *b_two;
2385 const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
2386 const char *reset = diff_get_color_opt(o, DIFF_RESET);
2387 const char *a_prefix, *b_prefix;
2388 struct userdiff_driver *textconv_one = NULL;
2389 struct userdiff_driver *textconv_two = NULL;
2390 struct strbuf header = STRBUF_INIT;
2391 const char *line_prefix = diff_line_prefix(o);
2392 diff_set_mnemonic_prefix(o, "a/", "b/");
2394 if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2395 a_prefix = o->b_prefix;
2396 b_prefix = o->a_prefix;
2397 } else {
2398 a_prefix = o->a_prefix;
2399 b_prefix = o->b_prefix;
2400 }
2401 if (o->submodule_format == DIFF_SUBMODULE_LOG &&
2403 (!one->mode || S_ISGITLINK(one->mode)) &&
2404 (!two->mode || S_ISGITLINK(two->mode))) {
2405 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2406 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2407 show_submodule_summary(o->file, one->path ? one->path : two->path,
2408 line_prefix,
2409 &one->oid, &two->oid,
2410 two->dirty_submodule,
2411 meta, del, add, reset);
2412 return;
2413 } else if (o->submodule_format == DIFF_SUBMODULE_INLINE_DIFF &&
2414 (!one->mode || S_ISGITLINK(one->mode)) &&
2415 (!two->mode || S_ISGITLINK(two->mode))) {
2416 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2417 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2418 show_submodule_inline_diff(o->file, one->path ? one->path : two->path,
2419 line_prefix,
2420 &one->oid, &two->oid,
2421 two->dirty_submodule,
2422 meta, del, add, reset, o);
2423 return;
2424 }
2425 if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2427 textconv_one = get_textconv(one);
2428 textconv_two = get_textconv(two);
2429 }
2430 /* Never use a non-valid filename anywhere if at all possible */
2432 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2433 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2434 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2436 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2437 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2438 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2439 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
2440 if (lbl[0][0] == '/') {
2441 /* /dev/null */
2442 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
2443 if (xfrm_msg)
2444 strbuf_addstr(&header, xfrm_msg);
2445 must_show_header = 1;
2446 }
2447 else if (lbl[1][0] == '/') {
2448 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
2449 if (xfrm_msg)
2450 strbuf_addstr(&header, xfrm_msg);
2451 must_show_header = 1;
2452 }
2453 else {
2454 if (one->mode != two->mode) {
2455 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
2456 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
2457 must_show_header = 1;
2458 }
2459 if (xfrm_msg)
2460 strbuf_addstr(&header, xfrm_msg);
2461 /*
2463 * we do not run diff between different kind
2464 * of objects.
2465 */
2466 if ((one->mode ^ two->mode) & S_IFMT)
2467 goto free_ab_and_return;
2468 if (complete_rewrite &&
2469 (textconv_one || !diff_filespec_is_binary(one)) &&
2470 (textconv_two || !diff_filespec_is_binary(two))) {
2471 fprintf(o->file, "%s", header.buf);
2472 strbuf_reset(&header);
2473 emit_rewrite_diff(name_a, name_b, one, two,
2474 textconv_one, textconv_two, o);
2475 o->found_changes = 1;
2476 goto free_ab_and_return;
2477 }
2478 }
2479 if (o->irreversible_delete && lbl[1][0] == '/') {
2481 fprintf(o->file, "%s", header.buf);
2482 strbuf_reset(&header);
2483 goto free_ab_and_return;
2484 } else if (!DIFF_OPT_TST(o, TEXT) &&
2485 ( (!textconv_one && diff_filespec_is_binary(one)) ||
2486 (!textconv_two && diff_filespec_is_binary(two)) )) {
2487 if (!one->data && !two->data &&
2488 S_ISREG(one->mode) && S_ISREG(two->mode) &&
2489 !DIFF_OPT_TST(o, BINARY)) {
2490 if (!oidcmp(&one->oid, &two->oid)) {
2491 if (must_show_header)
2492 fprintf(o->file, "%s", header.buf);
2493 goto free_ab_and_return;
2494 }
2495 fprintf(o->file, "%s", header.buf);
2496 fprintf(o->file, "%sBinary files %s and %s differ\n",
2497 line_prefix, lbl[0], lbl[1]);
2498 goto free_ab_and_return;
2499 }
2500 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2501 die("unable to read files to diff");
2502 /* Quite common confusing case */
2503 if (mf1.size == mf2.size &&
2504 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2505 if (must_show_header)
2506 fprintf(o->file, "%s", header.buf);
2507 goto free_ab_and_return;
2508 }
2509 fprintf(o->file, "%s", header.buf);
2510 strbuf_reset(&header);
2511 if (DIFF_OPT_TST(o, BINARY))
2512 emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2513 else
2514 fprintf(o->file, "%sBinary files %s and %s differ\n",
2515 line_prefix, lbl[0], lbl[1]);
2516 o->found_changes = 1;
2517 } else {
2518 /* Crazy xdl interfaces.. */
2519 const char *diffopts = getenv("GIT_DIFF_OPTS");
2520 const char *v;
2521 xpparam_t xpp;
2522 xdemitconf_t xecfg;
2523 struct emit_callback ecbdata;
2524 const struct userdiff_funcname *pe;
2525 if (must_show_header) {
2527 fprintf(o->file, "%s", header.buf);
2528 strbuf_reset(&header);
2529 }
2530 mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2532 mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2533 pe = diff_funcname_pattern(one);
2535 if (!pe)
2536 pe = diff_funcname_pattern(two);
2537 memset(&xpp, 0, sizeof(xpp));
2539 memset(&xecfg, 0, sizeof(xecfg));
2540 memset(&ecbdata, 0, sizeof(ecbdata));
2541 ecbdata.label_path = lbl;
2542 ecbdata.color_diff = want_color(o->use_color);
2543 ecbdata.ws_rule = whitespace_rule(name_b);
2544 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2545 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2546 ecbdata.opt = o;
2547 ecbdata.header = header.len ? &header : NULL;
2548 xpp.flags = o->xdl_opts;
2549 xecfg.ctxlen = o->context;
2550 xecfg.interhunkctxlen = o->interhunkcontext;
2551 xecfg.flags = XDL_EMIT_FUNCNAMES;
2552 if (DIFF_OPT_TST(o, FUNCCONTEXT))
2553 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2554 if (pe)
2555 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2556 if (!diffopts)
2557 ;
2558 else if (skip_prefix(diffopts, "--unified=", &v))
2559 xecfg.ctxlen = strtoul(v, NULL, 10);
2560 else if (skip_prefix(diffopts, "-u", &v))
2561 xecfg.ctxlen = strtoul(v, NULL, 10);
2562 if (o->word_diff)
2563 init_diff_words_data(&ecbdata, o, one, two);
2564 if (xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2565 &xpp, &xecfg))
2566 die("unable to generate diff for %s", one->path);
2567 if (o->word_diff)
2568 free_diff_words_data(&ecbdata);
2569 if (textconv_one)
2570 free(mf1.ptr);
2571 if (textconv_two)
2572 free(mf2.ptr);
2573 xdiff_clear_find_func(&xecfg);
2574 }
2575 free_ab_and_return:
2577 strbuf_release(&header);
2578 diff_free_filespec_data(one);
2579 diff_free_filespec_data(two);
2580 free(a_one);
2581 free(b_two);
2582 return;
2583}
2584static void builtin_diffstat(const char *name_a, const char *name_b,
2586 struct diff_filespec *one,
2587 struct diff_filespec *two,
2588 struct diffstat_t *diffstat,
2589 struct diff_options *o,
2590 struct diff_filepair *p)
2591{
2592 mmfile_t mf1, mf2;
2593 struct diffstat_file *data;
2594 int same_contents;
2595 int complete_rewrite = 0;
2596 if (!DIFF_PAIR_UNMERGED(p)) {
2598 if (p->status == DIFF_STATUS_MODIFIED && p->score)
2599 complete_rewrite = 1;
2600 }
2601 data = diffstat_add(diffstat, name_a, name_b);
2603 data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
2604 if (!one || !two) {
2606 data->is_unmerged = 1;
2607 return;
2608 }
2609 same_contents = !oidcmp(&one->oid, &two->oid);
2611 if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2613 data->is_binary = 1;
2614 if (same_contents) {
2615 data->added = 0;
2616 data->deleted = 0;
2617 } else {
2618 data->added = diff_filespec_size(two);
2619 data->deleted = diff_filespec_size(one);
2620 }
2621 }
2622 else if (complete_rewrite) {
2624 diff_populate_filespec(one, 0);
2625 diff_populate_filespec(two, 0);
2626 data->deleted = count_lines(one->data, one->size);
2627 data->added = count_lines(two->data, two->size);
2628 }
2629 else if (!same_contents) {
2631 /* Crazy xdl interfaces.. */
2632 xpparam_t xpp;
2633 xdemitconf_t xecfg;
2634 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2636 die("unable to read files to diff");
2637 memset(&xpp, 0, sizeof(xpp));
2639 memset(&xecfg, 0, sizeof(xecfg));
2640 xpp.flags = o->xdl_opts;
2641 xecfg.ctxlen = o->context;
2642 xecfg.interhunkctxlen = o->interhunkcontext;
2643 if (xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2644 &xpp, &xecfg))
2645 die("unable to generate diffstat for %s", one->path);
2646 }
2647 diff_free_filespec_data(one);
2649 diff_free_filespec_data(two);
2650}
2651static void builtin_checkdiff(const char *name_a, const char *name_b,
2653 const char *attr_path,
2654 struct diff_filespec *one,
2655 struct diff_filespec *two,
2656 struct diff_options *o)
2657{
2658 mmfile_t mf1, mf2;
2659 struct checkdiff_t data;
2660 if (!two)
2662 return;
2663 memset(&data, 0, sizeof(data));
2665 data.filename = name_b ? name_b : name_a;
2666 data.lineno = 0;
2667 data.o = o;
2668 data.ws_rule = whitespace_rule(attr_path);
2669 data.conflict_marker_size = ll_merge_marker_size(attr_path);
2670 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2672 die("unable to read files to diff");
2673 /*
2675 * All the other codepaths check both sides, but not checking
2676 * the "old" side here is deliberate. We are checking the newly
2677 * introduced changes, and as long as the "new" side is text, we
2678 * can and should check what it introduces.
2679 */
2680 if (diff_filespec_is_binary(two))
2681 goto free_and_return;
2682 else {
2683 /* Crazy xdl interfaces.. */
2684 xpparam_t xpp;
2685 xdemitconf_t xecfg;
2686 memset(&xpp, 0, sizeof(xpp));
2688 memset(&xecfg, 0, sizeof(xecfg));
2689 xecfg.ctxlen = 1; /* at least one context line */
2690 xpp.flags = 0;
2691 if (xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2692 &xpp, &xecfg))
2693 die("unable to generate checkdiff for %s", one->path);
2694 if (data.ws_rule & WS_BLANK_AT_EOF) {
2696 struct emit_callback ecbdata;
2697 int blank_at_eof;
2698 ecbdata.ws_rule = data.ws_rule;
2700 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2701 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2702 if (blank_at_eof) {
2704 static char *err;
2705 if (!err)
2706 err = whitespace_error_string(WS_BLANK_AT_EOF);
2707 fprintf(o->file, "%s:%d: %s.\n",
2708 data.filename, blank_at_eof, err);
2709 data.status = 1; /* report errors */
2710 }
2711 }
2712 }
2713 free_and_return:
2714 diff_free_filespec_data(one);
2715 diff_free_filespec_data(two);
2716 if (data.status)
2717 DIFF_OPT_SET(o, CHECK_FAILED);
2718}
2719struct diff_filespec *alloc_filespec(const char *path)
2721{
2722 struct diff_filespec *spec;
2723 FLEXPTR_ALLOC_STR(spec, path, path);
2725 spec->count = 1;
2726 spec->is_binary = -1;
2727 return spec;
2728}
2729void free_filespec(struct diff_filespec *spec)
2731{
2732 if (!--spec->count) {
2733 diff_free_filespec_data(spec);
2734 free(spec);
2735 }
2736}
2737void fill_filespec(struct diff_filespec *spec, const struct object_id *oid,
2739 int oid_valid, unsigned short mode)
2740{
2741 if (mode) {
2742 spec->mode = canon_mode(mode);
2743 oidcpy(&spec->oid, oid);
2744 spec->oid_valid = oid_valid;
2745 }
2746}
2747/*
2749 * Given a name and sha1 pair, if the index tells us the file in
2750 * the work tree has that object contents, return true, so that
2751 * prepare_temp_file() does not have to inflate and extract.
2752 */
2753static int reuse_worktree_file(const char *name, const struct object_id *oid, int want_file)
2754{
2755 const struct cache_entry *ce;
2756 struct stat st;
2757 int pos, len;
2758 /*
2760 * We do not read the cache ourselves here, because the
2761 * benchmark with my previous version that always reads cache
2762 * shows that it makes things worse for diff-tree comparing
2763 * two linux-2.6 kernel trees in an already checked out work
2764 * tree. This is because most diff-tree comparisons deal with
2765 * only a small number of files, while reading the cache is
2766 * expensive for a large project, and its cost outweighs the
2767 * savings we get by not inflating the object to a temporary
2768 * file. Practically, this code only helps when we are used
2769 * by diff-cache --cached, which does read the cache before
2770 * calling us.
2771 */
2772 if (!active_cache)
2773 return 0;
2774 /* We want to avoid the working directory if our caller
2776 * doesn't need the data in a normal file, this system
2777 * is rather slow with its stat/open/mmap/close syscalls,
2778 * and the object is contained in a pack file. The pack
2779 * is probably already open and will be faster to obtain
2780 * the data through than the working directory. Loose
2781 * objects however would tend to be slower as they need
2782 * to be individually opened and inflated.
2783 */
2784 if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(oid->hash))
2785 return 0;
2786 /*
2788 * Similarly, if we'd have to convert the file contents anyway, that
2789 * makes the optimization not worthwhile.
2790 */
2791 if (!want_file && would_convert_to_git(&the_index, name))
2792 return 0;
2793 len = strlen(name);
2795 pos = cache_name_pos(name, len);
2796 if (pos < 0)
2797 return 0;
2798 ce = active_cache[pos];
2799 /*
2801 * This is not the sha1 we are looking for, or
2802 * unreusable because it is not a regular file.
2803 */
2804 if (oidcmp(oid, &ce->oid) || !S_ISREG(ce->ce_mode))
2805 return 0;
2806 /*
2808 * If ce is marked as "assume unchanged", there is no
2809 * guarantee that work tree matches what we are looking for.
2810 */
2811 if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2812 return 0;
2813 /*
2815 * If ce matches the file in the work tree, we can reuse it.
2816 */
2817 if (ce_uptodate(ce) ||
2818 (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2819 return 1;
2820 return 0;
2822}
2823static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2825{
2826 struct strbuf buf = STRBUF_INIT;
2827 char *dirty = "";
2828 /* Are we looking at the work tree? */
2830 if (s->dirty_submodule)
2831 dirty = "-dirty";
2832 strbuf_addf(&buf, "Subproject commit %s%s\n",
2834 oid_to_hex(&s->oid), dirty);
2835 s->size = buf.len;
2836 if (size_only) {
2837 s->data = NULL;
2838 strbuf_release(&buf);
2839 } else {
2840 s->data = strbuf_detach(&buf, NULL);
2841 s->should_free = 1;
2842 }
2843 return 0;
2844}
2845/*
2847 * While doing rename detection and pickaxe operation, we may need to
2848 * grab the data for the blob (or file) for our own in-core comparison.
2849 * diff_filespec has data and size fields for this purpose.
2850 */
2851int diff_populate_filespec(struct diff_filespec *s, unsigned int flags)
2852{
2853 int size_only = flags & CHECK_SIZE_ONLY;
2854 int err = 0;
2855 /*
2856 * demote FAIL to WARN to allow inspecting the situation
2857 * instead of refusing.
2858 */
2859 enum safe_crlf crlf_warn = (safe_crlf == SAFE_CRLF_FAIL
2860 ? SAFE_CRLF_WARN
2861 : safe_crlf);
2862 if (!DIFF_FILE_VALID(s))
2864 die("internal error: asking to populate invalid file.");
2865 if (S_ISDIR(s->mode))
2866 return -1;
2867 if (s->data)
2869 return 0;
2870 if (size_only && 0 < s->size)
2872 return 0;
2873 if (S_ISGITLINK(s->mode))
2875 return diff_populate_gitlink(s, size_only);
2876 if (!s->oid_valid ||
2878 reuse_worktree_file(s->path, &s->oid, 0)) {
2879 struct strbuf buf = STRBUF_INIT;
2880 struct stat st;
2881 int fd;
2882 if (lstat(s->path, &st) < 0) {
2884 if (errno == ENOENT) {
2885 err_empty:
2886 err = -1;
2887 empty:
2888 s->data = (char *)"";
2889 s->size = 0;
2890 return err;
2891 }
2892 }
2893 s->size = xsize_t(st.st_size);
2894 if (!s->size)
2895 goto empty;
2896 if (S_ISLNK(st.st_mode)) {
2897 struct strbuf sb = STRBUF_INIT;
2898 if (strbuf_readlink(&sb, s->path, s->size))
2900 goto err_empty;
2901 s->size = sb.len;
2902 s->data = strbuf_detach(&sb, NULL);
2903 s->should_free = 1;
2904 return 0;
2905 }
2906 /*
2908 * Even if the caller would be happy with getting
2909 * only the size, we cannot return early at this
2910 * point if the path requires us to run the content
2911 * conversion.
2912 */
2913 if (size_only && !would_convert_to_git(&the_index, s->path))
2914 return 0;
2915 /*
2917 * Note: this check uses xsize_t(st.st_size) that may
2918 * not be the true size of the blob after it goes
2919 * through convert_to_git(). This may not strictly be
2920 * correct, but the whole point of big_file_threshold
2921 * and is_binary check being that we want to avoid
2922 * opening the file and inspecting the contents, this
2923 * is probably fine.
2924 */
2925 if ((flags & CHECK_BINARY) &&
2926 s->size > big_file_threshold && s->is_binary == -1) {
2927 s->is_binary = 1;
2928 return 0;
2929 }
2930 fd = open(s->path, O_RDONLY);
2931 if (fd < 0)
2932 goto err_empty;
2933 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2934 close(fd);
2935 s->should_munmap = 1;
2936 /*
2938 * Convert from working tree format to canonical git format
2939 */
2940 if (convert_to_git(&the_index, s->path, s->data, s->size, &buf, crlf_warn)) {
2941 size_t size = 0;
2942 munmap(s->data, s->size);
2943 s->should_munmap = 0;
2944 s->data = strbuf_detach(&buf, &size);
2945 s->size = size;
2946 s->should_free = 1;
2947 }
2948 }
2949 else {
2950 enum object_type type;
2951 if (size_only || (flags & CHECK_BINARY)) {
2952 type = sha1_object_info(s->oid.hash, &s->size);
2953 if (type < 0)
2954 die("unable to read %s",
2955 oid_to_hex(&s->oid));
2956 if (size_only)
2957 return 0;
2958 if (s->size > big_file_threshold && s->is_binary == -1) {
2959 s->is_binary = 1;
2960 return 0;
2961 }
2962 }
2963 s->data = read_sha1_file(s->oid.hash, &type, &s->size);
2964 if (!s->data)
2965 die("unable to read %s", oid_to_hex(&s->oid));
2966 s->should_free = 1;
2967 }
2968 return 0;
2969}
2970void diff_free_filespec_blob(struct diff_filespec *s)
2972{
2973 if (s->should_free)
2974 free(s->data);
2975 else if (s->should_munmap)
2976 munmap(s->data, s->size);
2977 if (s->should_free || s->should_munmap) {
2979 s->should_free = s->should_munmap = 0;
2980 s->data = NULL;
2981 }
2982}
2983void diff_free_filespec_data(struct diff_filespec *s)
2985{
2986 diff_free_filespec_blob(s);
2987 FREE_AND_NULL(s->cnt_data);
2988}
2989static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2991 void *blob,
2992 unsigned long size,
2993 const struct object_id *oid,
2994 int mode)
2995{
2996 int fd;
2997 struct strbuf buf = STRBUF_INIT;
2998 struct strbuf template = STRBUF_INIT;
2999 char *path_dup = xstrdup(path);
3000 const char *base = basename(path_dup);
3001 /* Generate "XXXXXX_basename.ext" */
3003 strbuf_addstr(&template, "XXXXXX_");
3004 strbuf_addstr(&template, base);
3005 fd = mks_tempfile_ts(&temp->tempfile, template.buf, strlen(base) + 1);
3007 if (fd < 0)
3008 die_errno("unable to create temp-file");
3009 if (convert_to_working_tree(path,
3010 (const char *)blob, (size_t)size, &buf)) {
3011 blob = buf.buf;
3012 size = buf.len;
3013 }
3014 if (write_in_full(fd, blob, size) != size)
3015 die_errno("unable to write temp-file");
3016 close_tempfile(&temp->tempfile);
3017 temp->name = get_tempfile_path(&temp->tempfile);
3018 oid_to_hex_r(temp->hex, oid);
3019 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", mode);
3020 strbuf_release(&buf);
3021 strbuf_release(&template);
3022 free(path_dup);
3023}
3024static struct diff_tempfile *prepare_temp_file(const char *name,
3026 struct diff_filespec *one)
3027{
3028 struct diff_tempfile *temp = claim_diff_tempfile();
3029 if (!DIFF_FILE_VALID(one)) {
3031 not_a_valid_file:
3032 /* A '-' entry produces this for file-2, and
3033 * a '+' entry produces this for file-1.
3034 */
3035 temp->name = "/dev/null";
3036 xsnprintf(temp->hex, sizeof(temp->hex), ".");
3037 xsnprintf(temp->mode, sizeof(temp->mode), ".");
3038 return temp;
3039 }
3040 if (!S_ISGITLINK(one->mode) &&
3042 (!one->oid_valid ||
3043 reuse_worktree_file(name, &one->oid, 1))) {
3044 struct stat st;
3045 if (lstat(name, &st) < 0) {
3046 if (errno == ENOENT)
3047 goto not_a_valid_file;
3048 die_errno("stat(%s)", name);
3049 }
3050 if (S_ISLNK(st.st_mode)) {
3051 struct strbuf sb = STRBUF_INIT;
3052 if (strbuf_readlink(&sb, name, st.st_size) < 0)
3053 die_errno("readlink(%s)", name);
3054 prep_temp_blob(name, temp, sb.buf, sb.len,
3055 (one->oid_valid ?
3056 &one->oid : &null_oid),
3057 (one->oid_valid ?
3058 one->mode : S_IFLNK));
3059 strbuf_release(&sb);
3060 }
3061 else {
3062 /* we can borrow from the file in the work tree */
3063 temp->name = name;
3064 if (!one->oid_valid)
3065 oid_to_hex_r(temp->hex, &null_oid);
3066 else
3067 oid_to_hex_r(temp->hex, &one->oid);
3068 /* Even though we may sometimes borrow the
3069 * contents from the work tree, we always want
3070 * one->mode. mode is trustworthy even when
3071 * !(one->oid_valid), as long as
3072 * DIFF_FILE_VALID(one).
3073 */
3074 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", one->mode);
3075 }
3076 return temp;
3077 }
3078 else {
3079 if (diff_populate_filespec(one, 0))
3080 die("cannot read data blob for %s", one->path);
3081 prep_temp_blob(name, temp, one->data, one->size,
3082 &one->oid, one->mode);
3083 }
3084 return temp;
3085}
3086static void add_external_diff_name(struct argv_array *argv,
3088 const char *name,
3089 struct diff_filespec *df)
3090{
3091 struct diff_tempfile *temp = prepare_temp_file(name, df);
3092 argv_array_push(argv, temp->name);
3093 argv_array_push(argv, temp->hex);
3094 argv_array_push(argv, temp->mode);
3095}
3096/* An external diff command takes:
3098 *
3099 * diff-cmd name infile1 infile1-sha1 infile1-mode \
3100 * infile2 infile2-sha1 infile2-mode [ rename-to ]
3101 *
3102 */
3103static void run_external_diff(const char *pgm,
3104 const char *name,
3105 const char *other,
3106 struct diff_filespec *one,
3107 struct diff_filespec *two,
3108 const char *xfrm_msg,
3109 int complete_rewrite,
3110 struct diff_options *o)
3111{
3112 struct argv_array argv = ARGV_ARRAY_INIT;
3113 struct argv_array env = ARGV_ARRAY_INIT;
3114 struct diff_queue_struct *q = &diff_queued_diff;
3115 argv_array_push(&argv, pgm);
3117 argv_array_push(&argv, name);
3118 if (one && two) {
3120 add_external_diff_name(&argv, name, one);
3121 if (!other)
3122 add_external_diff_name(&argv, name, two);
3123 else {
3124 add_external_diff_name(&argv, other, two);
3125 argv_array_push(&argv, other);
3126 argv_array_push(&argv, xfrm_msg);
3127 }
3128 }
3129 argv_array_pushf(&env, "GIT_DIFF_PATH_COUNTER=%d", ++o->diff_path_counter);
3131 argv_array_pushf(&env, "GIT_DIFF_PATH_TOTAL=%d", q->nr);
3132 if (run_command_v_opt_cd_env(argv.argv, RUN_USING_SHELL, NULL, env.argv))
3134 die(_("external diff died, stopping at %s"), name);
3135 remove_tempfile();
3137 argv_array_clear(&argv);
3138 argv_array_clear(&env);
3139}
3140static int similarity_index(struct diff_filepair *p)
3142{
3143 return p->score * 100 / MAX_SCORE;
3144}
3145static const char *diff_abbrev_oid(const struct object_id *oid, int abbrev)
3147{
3148 if (startup_info->have_repository)
3149 return find_unique_abbrev(oid->hash, abbrev);
3150 else {
3151 char *hex = oid_to_hex(oid);
3152 if (abbrev < 0)
3153 abbrev = FALLBACK_DEFAULT_ABBREV;
3154 if (abbrev > GIT_SHA1_HEXSZ)
3155 die("BUG: oid abbreviation out of range: %d", abbrev);
3156 if (abbrev)
3157 hex[abbrev] = '\0';
3158 return hex;
3159 }
3160}
3161static void fill_metainfo(struct strbuf *msg,
3163 const char *name,
3164 const char *other,
3165 struct diff_filespec *one,
3166 struct diff_filespec *two,
3167 struct diff_options *o,
3168 struct diff_filepair *p,
3169 int *must_show_header,
3170 int use_color)
3171{
3172 const char *set = diff_get_color(use_color, DIFF_METAINFO);
3173 const char *reset = diff_get_color(use_color, DIFF_RESET);
3174 const char *line_prefix = diff_line_prefix(o);
3175 *must_show_header = 1;
3177 strbuf_init(msg, PATH_MAX * 2 + 300);
3178 switch (p->status) {
3179 case DIFF_STATUS_COPIED:
3180 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3181 line_prefix, set, similarity_index(p));
3182 strbuf_addf(msg, "%s\n%s%scopy from ",
3183 reset, line_prefix, set);
3184 quote_c_style(name, msg, NULL, 0);
3185 strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
3186 quote_c_style(other, msg, NULL, 0);
3187 strbuf_addf(msg, "%s\n", reset);
3188 break;
3189 case DIFF_STATUS_RENAMED:
3190 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3191 line_prefix, set, similarity_index(p));
3192 strbuf_addf(msg, "%s\n%s%srename from ",
3193 reset, line_prefix, set);
3194 quote_c_style(name, msg, NULL, 0);
3195 strbuf_addf(msg, "%s\n%s%srename to ",
3196 reset, line_prefix, set);
3197 quote_c_style(other, msg, NULL, 0);
3198 strbuf_addf(msg, "%s\n", reset);
3199 break;
3200 case DIFF_STATUS_MODIFIED:
3201 if (p->score) {
3202 strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
3203 line_prefix,
3204 set, similarity_index(p), reset);
3205 break;
3206 }
3207 /* fallthru */
3208 default:
3209 *must_show_header = 0;
3210 }
3211 if (one && two && oidcmp(&one->oid, &two->oid)) {
3212 int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
3213 if (DIFF_OPT_TST(o, BINARY)) {
3215 mmfile_t mf;
3216 if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
3217 (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
3218 abbrev = 40;
3219 }
3220 strbuf_addf(msg, "%s%sindex %s..%s", line_prefix, set,
3221 diff_abbrev_oid(&one->oid, abbrev),
3222 diff_abbrev_oid(&two->oid, abbrev));
3223 if (one->mode == two->mode)
3224 strbuf_addf(msg, " %06o", one->mode);
3225 strbuf_addf(msg, "%s\n", reset);
3226 }
3227}
3228static void run_diff_cmd(const char *pgm,
3230 const char *name,
3231 const char *other,
3232 const char *attr_path,
3233 struct diff_filespec *one,
3234 struct diff_filespec *two,
3235 struct strbuf *msg,
3236 struct diff_options *o,
3237 struct diff_filepair *p)
3238{
3239 const char *xfrm_msg = NULL;
3240 int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
3241 int must_show_header = 0;
3242 if (DIFF_OPT_TST(o, ALLOW_EXTERNAL)) {
3245 struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
3246 if (drv && drv->external)
3247 pgm = drv->external;
3248 }
3249 if (msg) {
3251 /*
3252 * don't use colors when the header is intended for an
3253 * external diff driver
3254 */
3255 fill_metainfo(msg, name, other, one, two, o, p,
3256 &must_show_header,
3257 want_color(o->use_color) && !pgm);
3258 xfrm_msg = msg->len ? msg->buf : NULL;
3259 }
3260 if (pgm) {
3262 run_external_diff(pgm, name, other, one, two, xfrm_msg,
3263 complete_rewrite, o);
3264 return;
3265 }
3266 if (one && two)
3267 builtin_diff(name, other ? other : name,
3268 one, two, xfrm_msg, must_show_header,
3269 o, complete_rewrite);
3270 else
3271 fprintf(o->file, "* Unmerged path %s\n", name);
3272}
3273static void diff_fill_oid_info(struct diff_filespec *one)
3275{
3276 if (DIFF_FILE_VALID(one)) {
3277 if (!one->oid_valid) {
3278 struct stat st;
3279 if (one->is_stdin) {
3280 oidclr(&one->oid);
3281 return;
3282 }
3283 if (lstat(one->path, &st) < 0)
3284 die_errno("stat '%s'", one->path);
3285 if (index_path(one->oid.hash, one->path, &st, 0))
3286 die("cannot hash %s", one->path);
3287 }
3288 }
3289 else
3290 oidclr(&one->oid);
3291}
3292static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
3294{
3295 /* Strip the prefix but do not molest /dev/null and absolute paths */
3296 if (*namep && **namep != '/') {
3297 *namep += prefix_length;
3298 if (**namep == '/')
3299 ++*namep;
3300 }
3301 if (*otherp && **otherp != '/') {
3302 *otherp += prefix_length;
3303 if (**otherp == '/')
3304 ++*otherp;
3305 }
3306}
3307static void run_diff(struct diff_filepair *p, struct diff_options *o)
3309{
3310 const char *pgm = external_diff();
3311 struct strbuf msg;
3312 struct diff_filespec *one = p->one;
3313 struct diff_filespec *two = p->two;
3314 const char *name;
3315 const char *other;
3316 const char *attr_path;
3317 name = one->path;
3319 other = (strcmp(name, two->path) ? two->path : NULL);
3320 attr_path = name;
3321 if (o->prefix_length)
3322 strip_prefix(o->prefix_length, &name, &other);
3323 if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
3325 pgm = NULL;
3326 if (DIFF_PAIR_UNMERGED(p)) {
3328 run_diff_cmd(pgm, name, NULL, attr_path,
3329 NULL, NULL, NULL, o, p);
3330 return;
3331 }
3332 diff_fill_oid_info(one);
3334 diff_fill_oid_info(two);
3335 if (!pgm &&
3337 DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3338 (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3339 /*
3340 * a filepair that changes between file and symlink
3341 * needs to be split into deletion and creation.
3342 */
3343 struct diff_filespec *null = alloc_filespec(two->path);
3344 run_diff_cmd(NULL, name, other, attr_path,
3345 one, null, &msg, o, p);
3346 free(null);
3347 strbuf_release(&msg);
3348 null = alloc_filespec(one->path);
3350 run_diff_cmd(NULL, name, other, attr_path,
3351 null, two, &msg, o, p);
3352 free(null);
3353 }
3354 else
3355 run_diff_cmd(pgm, name, other, attr_path,
3356 one, two, &msg, o, p);
3357 strbuf_release(&msg);
3359}
3360static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3362 struct diffstat_t *diffstat)
3363{
3364 const char *name;
3365 const char *other;
3366 if (DIFF_PAIR_UNMERGED(p)) {
3368 /* unmerged */
3369 builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, p);
3370 return;
3371 }
3372 name = p->one->path;
3374 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3375 if (o->prefix_length)
3377 strip_prefix(o->prefix_length, &name, &other);
3378 diff_fill_oid_info(p->one);
3380 diff_fill_oid_info(p->two);
3381 builtin_diffstat(name, other, p->one, p->two, diffstat, o, p);
3383}
3384static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3386{
3387 const char *name;
3388 const char *other;
3389 const char *attr_path;
3390 if (DIFF_PAIR_UNMERGED(p)) {
3392 /* unmerged */
3393 return;
3394 }
3395 name = p->one->path;
3397 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3398 attr_path = other ? other : name;
3399 if (o->prefix_length)
3401 strip_prefix(o->prefix_length, &name, &other);
3402 diff_fill_oid_info(p->one);
3404 diff_fill_oid_info(p->two);
3405 builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3407}
3408void diff_setup(struct diff_options *options)
3410{
3411 memcpy(options, &default_diff_options, sizeof(*options));
3412 options->file = stdout;
3414 options->abbrev = DEFAULT_ABBREV;
3416 options->line_termination = '\n';
3417 options->break_opt = -1;
3418 options->rename_limit = -1;
3419 options->dirstat_permille = diff_dirstat_permille_default;
3420 options->context = diff_context_default;
3421 options->interhunkcontext = diff_interhunk_context_default;
3422 options->ws_error_highlight = ws_error_highlight_default;
3423 DIFF_OPT_SET(options, RENAME_EMPTY);
3424 /* pathchange left =NULL by default */
3426 options->change = diff_change;
3427 options->add_remove = diff_addremove;
3428 options->use_color = diff_use_color_default;
3429 options->detect_rename = diff_detect_rename_default;
3430 options->xdl_opts |= diff_algorithm;
3431 if (diff_indent_heuristic)
3432 DIFF_XDL_SET(options, INDENT_HEURISTIC);
3433 options->orderfile = diff_order_file_cfg;
3435 if (diff_no_prefix) {
3437 options->a_prefix = options->b_prefix = "";
3438 } else if (!diff_mnemonic_prefix) {
3439 options->a_prefix = "a/";
3440 options->b_prefix = "b/";
3441 }
3442}
3443void diff_setup_done(struct diff_options *options)
3445{
3446 int count = 0;
3447 if (options->set_default)
3449 options->set_default(options);
3450 if (options->output_format & DIFF_FORMAT_NAME)
3452 count++;
3453 if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3454 count++;
3455 if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3456 count++;
3457 if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3458 count++;
3459 if (count > 1)
3460 die(_("--name-only, --name-status, --check and -s are mutually exclusive"));
3461 /*
3463 * Most of the time we can say "there are changes"
3464 * only by checking if there are changed paths, but
3465 * --ignore-whitespace* options force us to look
3466 * inside contents.
3467 */
3468 if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3470 DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3471 DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3472 DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3473 else
3474 DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3475 if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3477 options->detect_rename = DIFF_DETECT_COPY;
3478 if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3480 options->prefix = NULL;
3481 if (options->prefix)
3482 options->prefix_length = strlen(options->prefix);
3483 else
3484 options->prefix_length = 0;
3485 if (options->output_format & (DIFF_FORMAT_NAME |
3487 DIFF_FORMAT_NAME_STATUS |
3488 DIFF_FORMAT_CHECKDIFF |
3489 DIFF_FORMAT_NO_OUTPUT))
3490 options->output_format &= ~(DIFF_FORMAT_RAW |
3491 DIFF_FORMAT_NUMSTAT |
3492 DIFF_FORMAT_DIFFSTAT |
3493 DIFF_FORMAT_SHORTSTAT |
3494 DIFF_FORMAT_DIRSTAT |
3495 DIFF_FORMAT_SUMMARY |
3496 DIFF_FORMAT_PATCH);
3497 /*
3499 * These cases always need recursive; we do not drop caller-supplied
3500 * recursive bits for other formats here.
3501 */
3502 if (options->output_format & (DIFF_FORMAT_PATCH |
3503 DIFF_FORMAT_NUMSTAT |
3504 DIFF_FORMAT_DIFFSTAT |
3505 DIFF_FORMAT_SHORTSTAT |
3506 DIFF_FORMAT_DIRSTAT |
3507 DIFF_FORMAT_SUMMARY |
3508 DIFF_FORMAT_CHECKDIFF))
3509 DIFF_OPT_SET(options, RECURSIVE);
3510 /*
3511 * Also pickaxe would not work very well if you do not say recursive
3512 */
3513 if (options->pickaxe)
3514 DIFF_OPT_SET(options, RECURSIVE);
3515 /*
3516 * When patches are generated, submodules diffed against the work tree
3517 * must be checked for dirtiness too so it can be shown in the output
3518 */
3519 if (options->output_format & DIFF_FORMAT_PATCH)
3520 DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3521 if (options->detect_rename && options->rename_limit < 0)
3523 options->rename_limit = diff_rename_limit_default;
3524 if (options->setup & DIFF_SETUP_USE_CACHE) {
3525 if (!active_cache)
3526 /* read-cache does not die even when it fails
3527 * so it is safe for us to do this here. Also
3528 * it does not smudge active_cache or active_nr
3529 * when it fails, so we do not have to worry about
3530 * cleaning it up ourselves either.
3531 */
3532 read_cache();
3533 }
3534 if (40 < options->abbrev)
3535 options->abbrev = 40; /* full */
3536 /*
3538 * It does not make sense to show the first hit we happened
3539 * to have found. It does not make sense not to return with
3540 * exit code in such a case either.
3541 */
3542 if (DIFF_OPT_TST(options, QUICK)) {
3543 options->output_format = DIFF_FORMAT_NO_OUTPUT;
3544 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3545 }
3546 options->diff_path_counter = 0;
3548 if (DIFF_OPT_TST(options, FOLLOW_RENAMES) && options->pathspec.nr != 1)
3550 die(_("--follow requires exactly one pathspec"));
3551}
3552static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3554{
3555 char c, *eq;
3556 int len;
3557 if (*arg != '-')
3559 return 0;
3560 c = *++arg;
3561 if (!c)
3562 return 0;
3563 if (c == arg_short) {
3564 c = *++arg;
3565 if (!c)
3566 return 1;
3567 if (val && isdigit(c)) {
3568 char *end;
3569 int n = strtoul(arg, &end, 10);
3570 if (*end)
3571 return 0;
3572 *val = n;
3573 return 1;
3574 }
3575 return 0;
3576 }
3577 if (c != '-')
3578 return 0;
3579 arg++;
3580 eq = strchrnul(arg, '=');
3581 len = eq - arg;
3582 if (!len || strncmp(arg, arg_long, len))
3583 return 0;
3584 if (*eq) {
3585 int n;
3586 char *end;
3587 if (!isdigit(*++eq))
3588 return 0;
3589 n = strtoul(eq, &end, 10);
3590 if (*end)
3591 return 0;
3592 *val = n;
3593 }
3594 return 1;
3595}
3596static int diff_scoreopt_parse(const char *opt);
3598static inline int short_opt(char opt, const char **argv,
3600 const char **optarg)
3601{
3602 const char *arg = argv[0];
3603 if (arg[0] != '-' || arg[1] != opt)
3604 return 0;
3605 if (arg[2] != '\0') {
3606 *optarg = arg + 2;
3607 return 1;
3608 }
3609 if (!argv[1])
3610 die("Option '%c' requires a value", opt);
3611 *optarg = argv[1];
3612 return 2;
3613}
3614int parse_long_opt(const char *opt, const char **argv,
3616 const char **optarg)
3617{
3618 const char *arg = argv[0];
3619 if (!skip_prefix(arg, "--", &arg))
3620 return 0;
3621 if (!skip_prefix(arg, opt, &arg))
3622 return 0;
3623 if (*arg == '=') { /* stuck form: --option=value */
3624 *optarg = arg + 1;
3625 return 1;
3626 }
3627 if (*arg != '\0')
3628 return 0;
3629 /* separate form: --option value */
3630 if (!argv[1])
3631 die("Option '--%s' requires a value", opt);
3632 *optarg = argv[1];
3633 return 2;
3634}
3635static int stat_opt(struct diff_options *options, const char **av)
3637{
3638 const char *arg = av[0];
3639 char *end;
3640 int width = options->stat_width;
3641 int name_width = options->stat_name_width;
3642 int graph_width = options->stat_graph_width;
3643 int count = options->stat_count;
3644 int argcount = 1;
3645 if (!skip_prefix(arg, "--stat", &arg))
3647 die("BUG: stat option does not begin with --stat: %s", arg);
3648 end = (char *)arg;
3649 switch (*arg) {
3651 case '-':
3652 if (skip_prefix(arg, "-width", &arg)) {
3653 if (*arg == '=')
3654 width = strtoul(arg + 1, &end, 10);
3655 else if (!*arg && !av[1])
3656 die_want_option("--stat-width");
3657 else if (!*arg) {
3658 width = strtoul(av[1], &end, 10);
3659 argcount = 2;
3660 }
3661 } else if (skip_prefix(arg, "-name-width", &arg)) {
3662 if (*arg == '=')
3663 name_width = strtoul(arg + 1, &end, 10);
3664 else if (!*arg && !av[1])
3665 die_want_option("--stat-name-width");
3666 else if (!*arg) {
3667 name_width = strtoul(av[1], &end, 10);
3668 argcount = 2;
3669 }
3670 } else if (skip_prefix(arg, "-graph-width", &arg)) {
3671 if (*arg == '=')
3672 graph_width = strtoul(arg + 1, &end, 10);
3673 else if (!*arg && !av[1])
3674 die_want_option("--stat-graph-width");
3675 else if (!*arg) {
3676 graph_width = strtoul(av[1], &end, 10);
3677 argcount = 2;
3678 }
3679 } else if (skip_prefix(arg, "-count", &arg)) {
3680 if (*arg == '=')
3681 count = strtoul(arg + 1, &end, 10);
3682 else if (!*arg && !av[1])
3683 die_want_option("--stat-count");
3684 else if (!*arg) {
3685 count = strtoul(av[1], &end, 10);
3686 argcount = 2;
3687 }
3688 }
3689 break;
3690 case '=':
3691 width = strtoul(arg+1, &end, 10);
3692 if (*end == ',')
3693 name_width = strtoul(end+1, &end, 10);
3694 if (*end == ',')
3695 count = strtoul(end+1, &end, 10);
3696 }
3697 /* Important! This checks all the error cases! */
3699 if (*end)
3700 return 0;
3701 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3702 options->stat_name_width = name_width;
3703 options->stat_graph_width = graph_width;
3704 options->stat_width = width;
3705 options->stat_count = count;
3706 return argcount;
3707}
3708static int parse_dirstat_opt(struct diff_options *options, const char *params)
3710{
3711 struct strbuf errmsg = STRBUF_INIT;
3712 if (parse_dirstat_params(options, params, &errmsg))
3713 die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3714 errmsg.buf);
3715 strbuf_release(&errmsg);
3716 /*
3717 * The caller knows a dirstat-related option is given from the command
3718 * line; allow it to say "return this_function();"
3719 */
3720 options->output_format |= DIFF_FORMAT_DIRSTAT;
3721 return 1;
3722}
3723static int parse_submodule_opt(struct diff_options *options, const char *value)
3725{
3726 if (parse_submodule_params(options, value))
3727 die(_("Failed to parse --submodule option parameter: '%s'"),
3728 value);
3729 return 1;
3730}
3731static const char diff_status_letters[] = {
3733 DIFF_STATUS_ADDED,
3734 DIFF_STATUS_COPIED,
3735 DIFF_STATUS_DELETED,
3736 DIFF_STATUS_MODIFIED,
3737 DIFF_STATUS_RENAMED,
3738 DIFF_STATUS_TYPE_CHANGED,
3739 DIFF_STATUS_UNKNOWN,
3740 DIFF_STATUS_UNMERGED,
3741 DIFF_STATUS_FILTER_AON,
3742 DIFF_STATUS_FILTER_BROKEN,
3743 '\0',
3744};
3745static unsigned int filter_bit['Z' + 1];
3747static void prepare_filter_bits(void)
3749{
3750 int i;
3751 if (!filter_bit[DIFF_STATUS_ADDED]) {
3753 for (i = 0; diff_status_letters[i]; i++)
3754 filter_bit[(int) diff_status_letters[i]] = (1 << i);
3755 }
3756}
3757static unsigned filter_bit_tst(char status, const struct diff_options *opt)
3759{
3760 return opt->filter & filter_bit[(int) status];
3761}
3762static int parse_diff_filter_opt(const char *optarg, struct diff_options *opt)
3764{
3765 int i, optch;
3766 prepare_filter_bits();
3768 /*
3770 * If there is a negation e.g. 'd' in the input, and we haven't
3771 * initialized the filter field with another --diff-filter, start
3772 * from full set of bits, except for AON.
3773 */
3774 if (!opt->filter) {
3775 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3776 if (optch < 'a' || 'z' < optch)
3777 continue;
3778 opt->filter = (1 << (ARRAY_SIZE(diff_status_letters) - 1)) - 1;
3779 opt->filter &= ~filter_bit[DIFF_STATUS_FILTER_AON];
3780 break;
3781 }
3782 }
3783 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3785 unsigned int bit;
3786 int negate;
3787 if ('a' <= optch && optch <= 'z') {
3789 negate = 1;
3790 optch = toupper(optch);
3791 } else {
3792 negate = 0;
3793 }
3794 bit = (0 <= optch && optch <= 'Z') ? filter_bit[optch] : 0;
3796 if (!bit)
3797 return optarg[i];
3798 if (negate)
3799 opt->filter &= ~bit;
3800 else
3801 opt->filter |= bit;
3802 }
3803 return 0;
3804}
3805static void enable_patch_output(int *fmt) {
3807 *fmt &= ~DIFF_FORMAT_NO_OUTPUT;
3808 *fmt |= DIFF_FORMAT_PATCH;
3809}
3810static int parse_ws_error_highlight_opt(struct diff_options *opt, const char *arg)
3812{
3813 int val = parse_ws_error_highlight(arg);
3814 if (val < 0) {
3816 error("unknown value after ws-error-highlight=%.*s",
3817 -1 - val, arg);
3818 return 0;
3819 }
3820 opt->ws_error_highlight = val;
3821 return 1;
3822}
3823int diff_opt_parse(struct diff_options *options,
3825 const char **av, int ac, const char *prefix)
3826{
3827 const char *arg = av[0];
3828 const char *optarg;
3829 int argcount;
3830 if (!prefix)
3832 prefix = "";
3833 /* Output format options */
3835 if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch")
3836 || opt_arg(arg, 'U', "unified", &options->context))
3837 enable_patch_output(&options->output_format);
3838 else if (!strcmp(arg, "--raw"))
3839 options->output_format |= DIFF_FORMAT_RAW;
3840 else if (!strcmp(arg, "--patch-with-raw")) {
3841 enable_patch_output(&options->output_format);
3842 options->output_format |= DIFF_FORMAT_RAW;
3843 } else if (!strcmp(arg, "--numstat"))
3844 options->output_format |= DIFF_FORMAT_NUMSTAT;
3845 else if (!strcmp(arg, "--shortstat"))
3846 options->output_format |= DIFF_FORMAT_SHORTSTAT;
3847 else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3848 return parse_dirstat_opt(options, "");
3849 else if (skip_prefix(arg, "-X", &arg))
3850 return parse_dirstat_opt(options, arg);
3851 else if (skip_prefix(arg, "--dirstat=", &arg))
3852 return parse_dirstat_opt(options, arg);
3853 else if (!strcmp(arg, "--cumulative"))
3854 return parse_dirstat_opt(options, "cumulative");
3855 else if (!strcmp(arg, "--dirstat-by-file"))
3856 return parse_dirstat_opt(options, "files");
3857 else if (skip_prefix(arg, "--dirstat-by-file=", &arg)) {
3858 parse_dirstat_opt(options, "files");
3859 return parse_dirstat_opt(options, arg);
3860 }
3861 else if (!strcmp(arg, "--check"))
3862 options->output_format |= DIFF_FORMAT_CHECKDIFF;
3863 else if (!strcmp(arg, "--summary"))
3864 options->output_format |= DIFF_FORMAT_SUMMARY;
3865 else if (!strcmp(arg, "--patch-with-stat")) {
3866 enable_patch_output(&options->output_format);
3867 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3868 } else if (!strcmp(arg, "--name-only"))
3869 options->output_format |= DIFF_FORMAT_NAME;
3870 else if (!strcmp(arg, "--name-status"))
3871 options->output_format |= DIFF_FORMAT_NAME_STATUS;
3872 else if (!strcmp(arg, "-s") || !strcmp(arg, "--no-patch"))
3873 options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3874 else if (starts_with(arg, "--stat"))
3875 /* --stat, --stat-width, --stat-name-width, or --stat-count */
3876 return stat_opt(options, av);
3877 /* renames options */
3879 else if (starts_with(arg, "-B") || starts_with(arg, "--break-rewrites=") ||
3880 !strcmp(arg, "--break-rewrites")) {
3881 if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3882 return error("invalid argument to -B: %s", arg+2);
3883 }
3884 else if (starts_with(arg, "-M") || starts_with(arg, "--find-renames=") ||
3885 !strcmp(arg, "--find-renames")) {
3886 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3887 return error("invalid argument to -M: %s", arg+2);
3888 options->detect_rename = DIFF_DETECT_RENAME;
3889 }
3890 else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3891 options->irreversible_delete = 1;
3892 }
3893 else if (starts_with(arg, "-C") || starts_with(arg, "--find-copies=") ||
3894 !strcmp(arg, "--find-copies")) {
3895 if (options->detect_rename == DIFF_DETECT_COPY)
3896 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3897 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3898 return error("invalid argument to -C: %s", arg+2);
3899 options->detect_rename = DIFF_DETECT_COPY;
3900 }
3901 else if (!strcmp(arg, "--no-renames"))
3902 options->detect_rename = 0;
3903 else if (!strcmp(arg, "--rename-empty"))
3904 DIFF_OPT_SET(options, RENAME_EMPTY);
3905 else if (!strcmp(arg, "--no-rename-empty"))
3906 DIFF_OPT_CLR(options, RENAME_EMPTY);
3907 else if (!strcmp(arg, "--relative"))
3908 DIFF_OPT_SET(options, RELATIVE_NAME);
3909 else if (skip_prefix(arg, "--relative=", &arg)) {
3910 DIFF_OPT_SET(options, RELATIVE_NAME);
3911 options->prefix = arg;
3912 }
3913 /* xdiff options */
3915 else if (!strcmp(arg, "--minimal"))
3916 DIFF_XDL_SET(options, NEED_MINIMAL);
3917 else if (!strcmp(arg, "--no-minimal"))
3918 DIFF_XDL_CLR(options, NEED_MINIMAL);
3919 else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3920 DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3921 else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3922 DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3923 else if (!strcmp(arg, "--ignore-space-at-eol"))
3924 DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3925 else if (!strcmp(arg, "--ignore-blank-lines"))
3926 DIFF_XDL_SET(options, IGNORE_BLANK_LINES);
3927 else if (!strcmp(arg, "--indent-heuristic"))
3928 DIFF_XDL_SET(options, INDENT_HEURISTIC);
3929 else if (!strcmp(arg, "--no-indent-heuristic"))
3930 DIFF_XDL_CLR(options, INDENT_HEURISTIC);
3931 else if (!strcmp(arg, "--patience"))
3932 options->xdl_opts = DIFF_WITH_ALG(options, PATIENCE_DIFF);
3933 else if (!strcmp(arg, "--histogram"))
3934 options->xdl_opts = DIFF_WITH_ALG(options, HISTOGRAM_DIFF);
3935 else if ((argcount = parse_long_opt("diff-algorithm", av, &optarg))) {
3936 long value = parse_algorithm_value(optarg);
3937 if (value < 0)
3938 return error("option diff-algorithm accepts \"myers\", "
3939 "\"minimal\", \"patience\" and \"histogram\"");
3940 /* clear out previous settings */
3941 DIFF_XDL_CLR(options, NEED_MINIMAL);
3942 options->xdl_opts &= ~XDF_DIFF_ALGORITHM_MASK;
3943 options->xdl_opts |= value;
3944 return argcount;
3945 }
3946 /* flags options */
3948 else if (!strcmp(arg, "--binary")) {
3949 enable_patch_output(&options->output_format);
3950 DIFF_OPT_SET(options, BINARY);
3951 }
3952 else if (!strcmp(arg, "--full-index"))
3953 DIFF_OPT_SET(options, FULL_INDEX);
3954 else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
3955 DIFF_OPT_SET(options, TEXT);
3956 else if (!strcmp(arg, "-R"))
3957 DIFF_OPT_SET(options, REVERSE_DIFF);
3958 else if (!strcmp(arg, "--find-copies-harder"))
3959 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3960 else if (!strcmp(arg, "--follow"))
3961 DIFF_OPT_SET(options, FOLLOW_RENAMES);
3962 else if (!strcmp(arg, "--no-follow")) {
3963 DIFF_OPT_CLR(options, FOLLOW_RENAMES);
3964 DIFF_OPT_CLR(options, DEFAULT_FOLLOW_RENAMES);
3965 } else if (!strcmp(arg, "--color"))
3966 options->use_color = 1;
3967 else if (skip_prefix(arg, "--color=", &arg)) {
3968 int value = git_config_colorbool(NULL, arg);
3969 if (value < 0)
3970 return error("option `color' expects \"always\", \"auto\", or \"never\"");
3971 options->use_color = value;
3972 }
3973 else if (!strcmp(arg, "--no-color"))
3974 options->use_color = 0;
3975 else if (!strcmp(arg, "--color-words")) {
3976 options->use_color = 1;
3977 options->word_diff = DIFF_WORDS_COLOR;
3978 }
3979 else if (skip_prefix(arg, "--color-words=", &arg)) {
3980 options->use_color = 1;
3981 options->word_diff = DIFF_WORDS_COLOR;
3982 options->word_regex = arg;
3983 }
3984 else if (!strcmp(arg, "--word-diff")) {
3985 if (options->word_diff == DIFF_WORDS_NONE)
3986 options->word_diff = DIFF_WORDS_PLAIN;
3987 }
3988 else if (skip_prefix(arg, "--word-diff=", &arg)) {
3989 if (!strcmp(arg, "plain"))
3990 options->word_diff = DIFF_WORDS_PLAIN;
3991 else if (!strcmp(arg, "color")) {
3992 options->use_color = 1;
3993 options->word_diff = DIFF_WORDS_COLOR;
3994 }
3995 else if (!strcmp(arg, "porcelain"))
3996 options->word_diff = DIFF_WORDS_PORCELAIN;
3997 else if (!strcmp(arg, "none"))
3998 options->word_diff = DIFF_WORDS_NONE;
3999 else
4000 die("bad --word-diff argument: %s", arg);
4001 }
4002 else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
4003 if (options->word_diff == DIFF_WORDS_NONE)
4004 options->word_diff = DIFF_WORDS_PLAIN;
4005 options->word_regex = optarg;
4006 return argcount;
4007 }
4008 else if (!strcmp(arg, "--exit-code"))
4009 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
4010 else if (!strcmp(arg, "--quiet"))
4011 DIFF_OPT_SET(options, QUICK);
4012 else if (!strcmp(arg, "--ext-diff"))
4013 DIFF_OPT_SET(options, ALLOW_EXTERNAL);
4014 else if (!strcmp(arg, "--no-ext-diff"))
4015 DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
4016 else if (!strcmp(arg, "--textconv"))
4017 DIFF_OPT_SET(options, ALLOW_TEXTCONV);
4018 else if (!strcmp(arg, "--no-textconv"))
4019 DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
4020 else if (!strcmp(arg, "--ignore-submodules")) {
4021 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
4022 handle_ignore_submodules_arg(options, "all");
4023 } else if (skip_prefix(arg, "--ignore-submodules=", &arg)) {
4024 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
4025 handle_ignore_submodules_arg(options, arg);
4026 } else if (!strcmp(arg, "--submodule"))
4027 options->submodule_format = DIFF_SUBMODULE_LOG;
4028 else if (skip_prefix(arg, "--submodule=", &arg))
4029 return parse_submodule_opt(options, arg);
4030 else if (skip_prefix(arg, "--ws-error-highlight=", &arg))
4031 return parse_ws_error_highlight_opt(options, arg);
4032 else if (!strcmp(arg, "--ita-invisible-in-index"))
4033 options->ita_invisible_in_index = 1;
4034 else if (!strcmp(arg, "--ita-visible-in-index"))
4035 options->ita_invisible_in_index = 0;
4036 /* misc options */
4038 else if (!strcmp(arg, "-z"))
4039 options->line_termination = 0;
4040 else if ((argcount = short_opt('l', av, &optarg))) {
4041 options->rename_limit = strtoul(optarg, NULL, 10);
4042 return argcount;
4043 }
4044 else if ((argcount = short_opt('S', av, &optarg))) {
4045 options->pickaxe = optarg;
4046 options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
4047 return argcount;
4048 } else if ((argcount = short_opt('G', av, &optarg))) {
4049 options->pickaxe = optarg;
4050 options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
4051 return argcount;
4052 }
4053 else if (!strcmp(arg, "--pickaxe-all"))
4054 options->pickaxe_opts |= DIFF_PICKAXE_ALL;
4055 else if (!strcmp(arg, "--pickaxe-regex"))
4056 options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
4057 else if ((argcount = short_opt('O', av, &optarg))) {
4058 options->orderfile = prefix_filename(prefix, optarg);
4059 return argcount;
4060 }
4061 else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
4062 int offending = parse_diff_filter_opt(optarg, options);
4063 if (offending)
4064 die("unknown change class '%c' in --diff-filter=%s",
4065 offending, optarg);
4066 return argcount;
4067 }
4068 else if (!strcmp(arg, "--no-abbrev"))
4069 options->abbrev = 0;
4070 else if (!strcmp(arg, "--abbrev"))
4071 options->abbrev = DEFAULT_ABBREV;
4072 else if (skip_prefix(arg, "--abbrev=", &arg)) {
4073 options->abbrev = strtoul(arg, NULL, 10);
4074 if (options->abbrev < MINIMUM_ABBREV)
4075 options->abbrev = MINIMUM_ABBREV;
4076 else if (40 < options->abbrev)
4077 options->abbrev = 40;
4078 }
4079 else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
4080 options->a_prefix = optarg;
4081 return argcount;
4082 }
4083 else if ((argcount = parse_long_opt("line-prefix", av, &optarg))) {
4084 options->line_prefix = optarg;
4085 options->line_prefix_length = strlen(options->line_prefix);
4086 graph_setup_line_prefix(options);
4087 return argcount;
4088 }
4089 else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
4090 options->b_prefix = optarg;
4091 return argcount;
4092 }
4093 else if (!strcmp(arg, "--no-prefix"))
4094 options->a_prefix = options->b_prefix = "";
4095 else if (opt_arg(arg, '\0', "inter-hunk-context",
4096 &options->interhunkcontext))
4097 ;
4098 else if (!strcmp(arg, "-W"))
4099 DIFF_OPT_SET(options, FUNCCONTEXT);
4100 else if (!strcmp(arg, "--function-context"))
4101 DIFF_OPT_SET(options, FUNCCONTEXT);
4102 else if (!strcmp(arg, "--no-function-context"))
4103 DIFF_OPT_CLR(options, FUNCCONTEXT);
4104 else if ((argcount = parse_long_opt("output", av, &optarg))) {
4105 char *path = prefix_filename(prefix, optarg);
4106 options->file = xfopen(path, "w");
4107 options->close_file = 1;
4108 if (options->use_color != GIT_COLOR_ALWAYS)
4109 options->use_color = GIT_COLOR_NEVER;
4110 free(path);
4111 return argcount;
4112 } else
4113 return 0;
4114 return 1;
4115}
4116int parse_rename_score(const char **cp_p)
4118{
4119 unsigned long num, scale;
4120 int ch, dot;
4121 const char *cp = *cp_p;
4122 num = 0;
4124 scale = 1;
4125 dot = 0;
4126 for (;;) {
4127 ch = *cp;
4128 if ( !dot && ch == '.' ) {
4129 scale = 1;
4130 dot = 1;
4131 } else if ( ch == '%' ) {
4132 scale = dot ? scale*100 : 100;
4133 cp++; /* % is always at the end */
4134 break;
4135 } else if ( ch >= '0' && ch <= '9' ) {
4136 if ( scale < 100000 ) {
4137 scale *= 10;
4138 num = (num*10) + (ch-'0');
4139 }
4140 } else {
4141 break;
4142 }
4143 cp++;
4144 }
4145 *cp_p = cp;
4146 /* user says num divided by scale and we say internally that
4148 * is MAX_SCORE * num / scale.
4149 */
4150 return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
4151}
4152static int diff_scoreopt_parse(const char *opt)
4154{
4155 int opt1, opt2, cmd;
4156 if (*opt++ != '-')
4158 return -1;
4159 cmd = *opt++;
4160 if (cmd == '-') {
4161 /* convert the long-form arguments into short-form versions */
4162 if (skip_prefix(opt, "break-rewrites", &opt)) {
4163 if (*opt == 0 || *opt++ == '=')
4164 cmd = 'B';
4165 } else if (skip_prefix(opt, "find-copies", &opt)) {
4166 if (*opt == 0 || *opt++ == '=')
4167 cmd = 'C';
4168 } else if (skip_prefix(opt, "find-renames", &opt)) {
4169 if (*opt == 0 || *opt++ == '=')
4170 cmd = 'M';
4171 }
4172 }
4173 if (cmd != 'M' && cmd != 'C' && cmd != 'B')
4174 return -1; /* that is not a -M, -C, or -B option */
4175 opt1 = parse_rename_score(&opt);
4177 if (cmd != 'B')
4178 opt2 = 0;
4179 else {
4180 if (*opt == 0)
4181 opt2 = 0;
4182 else if (*opt != '/')
4183 return -1; /* we expect -B80/99 or -B80 */
4184 else {
4185 opt++;
4186 opt2 = parse_rename_score(&opt);
4187 }
4188 }
4189 if (*opt != 0)
4190 return -1;
4191 return opt1 | (opt2 << 16);
4192}
4193struct diff_queue_struct diff_queued_diff;
4195void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
4197{
4198 ALLOC_GROW(queue->queue, queue->nr + 1, queue->alloc);
4199 queue->queue[queue->nr++] = dp;
4200}
4201struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
4203 struct diff_filespec *one,
4204 struct diff_filespec *two)
4205{
4206 struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
4207 dp->one = one;
4208 dp->two = two;
4209 if (queue)
4210 diff_q(queue, dp);
4211 return dp;
4212}
4213void diff_free_filepair(struct diff_filepair *p)
4215{
4216 free_filespec(p->one);
4217 free_filespec(p->two);
4218 free(p);
4219}
4220const char *diff_aligned_abbrev(const struct object_id *oid, int len)
4222{
4223 int abblen;
4224 const char *abbrev;
4225 if (len == GIT_SHA1_HEXSZ)
4227 return oid_to_hex(oid);
4228 abbrev = diff_abbrev_oid(oid, len);
4230 abblen = strlen(abbrev);
4231 /*
4233 * In well-behaved cases, where the abbbreviated result is the
4234 * same as the requested length, append three dots after the
4235 * abbreviation (hence the whole logic is limited to the case
4236 * where abblen < 37); when the actual abbreviated result is a
4237 * bit longer than the requested length, we reduce the number
4238 * of dots so that they match the well-behaved ones. However,
4239 * if the actual abbreviation is longer than the requested
4240 * length by more than three, we give up on aligning, and add
4241 * three dots anyway, to indicate that the output is not the
4242 * full object name. Yes, this may be suboptimal, but this
4243 * appears only in "diff --raw --abbrev" output and it is not
4244 * worth the effort to change it now. Note that this would
4245 * likely to work fine when the automatic sizing of default
4246 * abbreviation length is used--we would be fed -1 in "len" in
4247 * that case, and will end up always appending three-dots, but
4248 * the automatic sizing is supposed to give abblen that ensures
4249 * uniqueness across all objects (statistically speaking).
4250 */
4251 if (abblen < GIT_SHA1_HEXSZ - 3) {
4252 static char hex[GIT_MAX_HEXSZ + 1];
4253 if (len < abblen && abblen <= len + 2)
4254 xsnprintf(hex, sizeof(hex), "%s%.*s", abbrev, len+3-abblen, "..");
4255 else
4256 xsnprintf(hex, sizeof(hex), "%s...", abbrev);
4257 return hex;
4258 }
4259 return oid_to_hex(oid);
4261}
4262static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
4264{
4265 int line_termination = opt->line_termination;
4266 int inter_name_termination = line_termination ? '\t' : '\0';
4267 fprintf(opt->file, "%s", diff_line_prefix(opt));
4269 if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
4270 fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
4271 diff_aligned_abbrev(&p->one->oid, opt->abbrev));
4272 fprintf(opt->file, "%s ",
4273 diff_aligned_abbrev(&p->two->oid, opt->abbrev));
4274 }
4275 if (p->score) {
4276 fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
4277 inter_name_termination);
4278 } else {
4279 fprintf(opt->file, "%c%c", p->status, inter_name_termination);
4280 }
4281 if (p->status == DIFF_STATUS_COPIED ||
4283 p->status == DIFF_STATUS_RENAMED) {
4284 const char *name_a, *name_b;
4285 name_a = p->one->path;
4286 name_b = p->two->path;
4287 strip_prefix(opt->prefix_length, &name_a, &name_b);
4288 write_name_quoted(name_a, opt->file, inter_name_termination);
4289 write_name_quoted(name_b, opt->file, line_termination);
4290 } else {
4291 const char *name_a, *name_b;
4292 name_a = p->one->mode ? p->one->path : p->two->path;
4293 name_b = NULL;
4294 strip_prefix(opt->prefix_length, &name_a, &name_b);
4295 write_name_quoted(name_a, opt->file, line_termination);
4296 }
4297}
4298int diff_unmodified_pair(struct diff_filepair *p)
4300{
4301 /* This function is written stricter than necessary to support
4302 * the currently implemented transformers, but the idea is to
4303 * let transformers to produce diff_filepairs any way they want,
4304 * and filter and clean them up here before producing the output.
4305 */
4306 struct diff_filespec *one = p->one, *two = p->two;
4307 if (DIFF_PAIR_UNMERGED(p))
4309 return 0; /* unmerged is interesting */
4310 /* deletion, addition, mode or type change
4312 * and rename are all interesting.
4313 */
4314 if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
4315 DIFF_PAIR_MODE_CHANGED(p) ||
4316 strcmp(one->path, two->path))
4317 return 0;
4318 /* both are valid and point at the same path. that is, we are
4320 * dealing with a change.
4321 */
4322 if (one->oid_valid && two->oid_valid &&
4323 !oidcmp(&one->oid, &two->oid) &&
4324 !one->dirty_submodule && !two->dirty_submodule)
4325 return 1; /* no change */
4326 if (!one->oid_valid && !two->oid_valid)
4327 return 1; /* both look at the same file on the filesystem. */
4328 return 0;
4329}
4330static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
4332{
4333 if (diff_unmodified_pair(p))
4334 return;
4335 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4337 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4338 return; /* no tree diffs in patch format */
4339 run_diff(p, o);
4341}
4342static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
4344 struct diffstat_t *diffstat)
4345{
4346 if (diff_unmodified_pair(p))
4347 return;
4348 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4350 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4351 return; /* no useful stat for tree diffs */
4352 run_diffstat(p, o, diffstat);
4354}
4355static void diff_flush_checkdiff(struct diff_filepair *p,
4357 struct diff_options *o)
4358{
4359 if (diff_unmodified_pair(p))
4360 return;
4361 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4363 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4364 return; /* nothing to check in tree diffs */
4365 run_checkdiff(p, o);
4367}
4368int diff_queue_is_empty(void)
4370{
4371 struct diff_queue_struct *q = &diff_queued_diff;
4372 int i;
4373 for (i = 0; i < q->nr; i++)
4374 if (!diff_unmodified_pair(q->queue[i]))
4375 return 0;
4376 return 1;
4377}
4378#if DIFF_DEBUG
4380void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
4381{
4382 fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
4383 x, one ? one : "",
4384 s->path,
4385 DIFF_FILE_VALID(s) ? "valid" : "invalid",
4386 s->mode,
4387 s->oid_valid ? oid_to_hex(&s->oid) : "");
4388 fprintf(stderr, "queue[%d] %s size %lu\n",
4389 x, one ? one : "",
4390 s->size);
4391}
4392void diff_debug_filepair(const struct diff_filepair *p, int i)
4394{
4395 diff_debug_filespec(p->one, i, "one");
4396 diff_debug_filespec(p->two, i, "two");
4397 fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
4398 p->score, p->status ? p->status : '?',
4399 p->one->rename_used, p->broken_pair);
4400}
4401void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
4403{
4404 int i;
4405 if (msg)
4406 fprintf(stderr, "%s\n", msg);
4407 fprintf(stderr, "q->nr = %d\n", q->nr);
4408 for (i = 0; i < q->nr; i++) {
4409 struct diff_filepair *p = q->queue[i];
4410 diff_debug_filepair(p, i);
4411 }
4412}
4413#endif
4414static void diff_resolve_rename_copy(void)
4416{
4417 int i;
4418 struct diff_filepair *p;
4419 struct diff_queue_struct *q = &diff_queued_diff;
4420 diff_debug_queue("resolve-rename-copy", q);
4422 for (i = 0; i < q->nr; i++) {
4424 p = q->queue[i];
4425 p->status = 0; /* undecided */
4426 if (DIFF_PAIR_UNMERGED(p))
4427 p->status = DIFF_STATUS_UNMERGED;
4428 else if (!DIFF_FILE_VALID(p->one))
4429 p->status = DIFF_STATUS_ADDED;
4430 else if (!DIFF_FILE_VALID(p->two))
4431 p->status = DIFF_STATUS_DELETED;
4432 else if (DIFF_PAIR_TYPE_CHANGED(p))
4433 p->status = DIFF_STATUS_TYPE_CHANGED;
4434 /* from this point on, we are dealing with a pair
4436 * whose both sides are valid and of the same type, i.e.
4437 * either in-place edit or rename/copy edit.
4438 */
4439 else if (DIFF_PAIR_RENAME(p)) {
4440 /*
4441 * A rename might have re-connected a broken
4442 * pair up, causing the pathnames to be the
4443 * same again. If so, that's not a rename at
4444 * all, just a modification..
4445 *
4446 * Otherwise, see if this source was used for
4447 * multiple renames, in which case we decrement
4448 * the count, and call it a copy.
4449 */
4450 if (!strcmp(p->one->path, p->two->path))
4451 p->status = DIFF_STATUS_MODIFIED;
4452 else if (--p->one->rename_used > 0)
4453 p->status = DIFF_STATUS_COPIED;
4454 else
4455 p->status = DIFF_STATUS_RENAMED;
4456 }
4457 else if (oidcmp(&p->one->oid, &p->two->oid) ||
4458 p->one->mode != p->two->mode ||
4459 p->one->dirty_submodule ||
4460 p->two->dirty_submodule ||
4461 is_null_oid(&p->one->oid))
4462 p->status = DIFF_STATUS_MODIFIED;
4463 else {
4464 /* This is a "no-change" entry and should not
4465 * happen anymore, but prepare for broken callers.
4466 */
4467 error("feeding unmodified %s to diffcore",
4468 p->one->path);
4469 p->status = DIFF_STATUS_UNKNOWN;
4470 }
4471 }
4472 diff_debug_queue("resolve-rename-copy done", q);
4473}
4474static int check_pair_status(struct diff_filepair *p)
4476{
4477 switch (p->status) {
4478 case DIFF_STATUS_UNKNOWN:
4479 return 0;
4480 case 0:
4481 die("internal error in diff-resolve-rename-copy");
4482 default:
4483 return 1;
4484 }
4485}
4486static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
4488{
4489 int fmt = opt->output_format;
4490 if (fmt & DIFF_FORMAT_CHECKDIFF)
4492 diff_flush_checkdiff(p, opt);
4493 else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
4494 diff_flush_raw(p, opt);
4495 else if (fmt & DIFF_FORMAT_NAME) {
4496 const char *name_a, *name_b;
4497 name_a = p->two->path;
4498 name_b = NULL;
4499 strip_prefix(opt->prefix_length, &name_a, &name_b);
4500 fprintf(opt->file, "%s", diff_line_prefix(opt));
4501 write_name_quoted(name_a, opt->file, opt->line_termination);
4502 }
4503}
4504static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
4506{
4507 if (fs->mode)
4508 fprintf(file, " %s mode %06o ", newdelete, fs->mode);
4509 else
4510 fprintf(file, " %s ", newdelete);
4511 write_name_quoted(fs->path, file, '\n');
4512}
4513static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
4516 const char *line_prefix)
4517{
4518 if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
4519 fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
4520 p->two->mode, show_name ? ' ' : '\n');
4521 if (show_name) {
4522 write_name_quoted(p->two->path, file, '\n');
4523 }
4524 }
4525}
4526static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
4528 const char *line_prefix)
4529{
4530 char *names = pprint_rename(p->one->path, p->two->path);
4531 fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
4533 free(names);
4534 show_mode_change(file, p, 0, line_prefix);
4535}
4536static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
4538{
4539 FILE *file = opt->file;
4540 const char *line_prefix = diff_line_prefix(opt);
4541 switch(p->status) {
4543 case DIFF_STATUS_DELETED:
4544 fputs(line_prefix, file);
4545 show_file_mode_name(file, "delete", p->one);
4546 break;
4547 case DIFF_STATUS_ADDED:
4548 fputs(line_prefix, file);
4549 show_file_mode_name(file, "create", p->two);
4550 break;
4551 case DIFF_STATUS_COPIED:
4552 fputs(line_prefix, file);
4553 show_rename_copy(file, "copy", p, line_prefix);
4554 break;
4555 case DIFF_STATUS_RENAMED:
4556 fputs(line_prefix, file);
4557 show_rename_copy(file, "rename", p, line_prefix);
4558 break;
4559 default:
4560 if (p->score) {
4561 fprintf(file, "%s rewrite ", line_prefix);
4562 write_name_quoted(p->two->path, file, ' ');
4563 fprintf(file, "(%d%%)\n", similarity_index(p));
4564 }
4565 show_mode_change(file, p, !p->score, line_prefix);
4566 break;
4567 }
4568}
4569struct patch_id_t {
4571 git_SHA_CTX *ctx;
4572 int patchlen;
4573};
4574static int remove_space(char *line, int len)
4576{
4577 int i;
4578 char *dst = line;
4579 unsigned char c;
4580 for (i = 0; i < len; i++)
4582 if (!isspace((c = line[i])))
4583 *dst++ = c;
4584 return dst - line;
4586}
4587static void patch_id_consume(void *priv, char *line, unsigned long len)
4589{
4590 struct patch_id_t *data = priv;
4591 int new_len;
4592 /* Ignore line numbers when computing the SHA1 of the patch */
4594 if (starts_with(line, "@@ -"))
4595 return;
4596 new_len = remove_space(line, len);
4598 git_SHA1_Update(data->ctx, line, new_len);
4600 data->patchlen += new_len;
4601}
4602static void patch_id_add_string(git_SHA_CTX *ctx, const char *str)
4604{
4605 git_SHA1_Update(ctx, str, strlen(str));
4606}
4607static void patch_id_add_mode(git_SHA_CTX *ctx, unsigned mode)
4609{
4610 /* large enough for 2^32 in octal */
4611 char buf[12];
4612 int len = xsnprintf(buf, sizeof(buf), "%06o", mode);
4613 git_SHA1_Update(ctx, buf, len);
4614}
4615/* returns 0 upon success, and writes result into sha1 */
4617static int diff_get_patch_id(struct diff_options *options, struct object_id *oid, int diff_header_only)
4618{
4619 struct diff_queue_struct *q = &diff_queued_diff;
4620 int i;
4621 git_SHA_CTX ctx;
4622 struct patch_id_t data;
4623 git_SHA1_Init(&ctx);
4625 memset(&data, 0, sizeof(struct patch_id_t));
4626 data.ctx = &ctx;
4627 for (i = 0; i < q->nr; i++) {
4629 xpparam_t xpp;
4630 xdemitconf_t xecfg;
4631 mmfile_t mf1, mf2;
4632 struct diff_filepair *p = q->queue[i];
4633 int len1, len2;
4634 memset(&xpp, 0, sizeof(xpp));
4636 memset(&xecfg, 0, sizeof(xecfg));
4637 if (p->status == 0)
4638 return error("internal diff status error");
4639 if (p->status == DIFF_STATUS_UNKNOWN)
4640 continue;
4641 if (diff_unmodified_pair(p))
4642 continue;
4643 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4644 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4645 continue;
4646 if (DIFF_PAIR_UNMERGED(p))
4647 continue;
4648 diff_fill_oid_info(p->one);
4650 diff_fill_oid_info(p->two);
4651 len1 = remove_space(p->one->path, strlen(p->one->path));
4653 len2 = remove_space(p->two->path, strlen(p->two->path));
4654 patch_id_add_string(&ctx, "diff--git");
4655 patch_id_add_string(&ctx, "a/");
4656 git_SHA1_Update(&ctx, p->one->path, len1);
4657 patch_id_add_string(&ctx, "b/");
4658 git_SHA1_Update(&ctx, p->two->path, len2);
4659 if (p->one->mode == 0) {
4661 patch_id_add_string(&ctx, "newfilemode");
4662 patch_id_add_mode(&ctx, p->two->mode);
4663 patch_id_add_string(&ctx, "---/dev/null");
4664 patch_id_add_string(&ctx, "+++b/");
4665 git_SHA1_Update(&ctx, p->two->path, len2);
4666 } else if (p->two->mode == 0) {
4667 patch_id_add_string(&ctx, "deletedfilemode");
4668 patch_id_add_mode(&ctx, p->one->mode);
4669 patch_id_add_string(&ctx, "---a/");
4670 git_SHA1_Update(&ctx, p->one->path, len1);
4671 patch_id_add_string(&ctx, "+++/dev/null");
4672 } else {
4673 patch_id_add_string(&ctx, "---a/");
4674 git_SHA1_Update(&ctx, p->one->path, len1);
4675 patch_id_add_string(&ctx, "+++b/");
4676 git_SHA1_Update(&ctx, p->two->path, len2);
4677 }
4678 if (diff_header_only)
4680 continue;
4681 if (fill_mmfile(&mf1, p->one) < 0 ||
4683 fill_mmfile(&mf2, p->two) < 0)
4684 return error("unable to read files to diff");
4685 if (diff_filespec_is_binary(p->one) ||
4687 diff_filespec_is_binary(p->two)) {
4688 git_SHA1_Update(&ctx, oid_to_hex(&p->one->oid),
4689 GIT_SHA1_HEXSZ);
4690 git_SHA1_Update(&ctx, oid_to_hex(&p->two->oid),
4691 GIT_SHA1_HEXSZ);
4692 continue;
4693 }
4694 xpp.flags = 0;
4696 xecfg.ctxlen = 3;
4697 xecfg.flags = 0;
4698 if (xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4699 &xpp, &xecfg))
4700 return error("unable to generate patch-id diff for %s",
4701 p->one->path);
4702 }
4703 git_SHA1_Final(oid->hash, &ctx);
4705 return 0;
4706}
4707int diff_flush_patch_id(struct diff_options *options, struct object_id *oid, int diff_header_only)
4709{
4710 struct diff_queue_struct *q = &diff_queued_diff;
4711 int i;
4712 int result = diff_get_patch_id(options, oid, diff_header_only);
4713 for (i = 0; i < q->nr; i++)
4715 diff_free_filepair(q->queue[i]);
4716 free(q->queue);
4718 DIFF_QUEUE_CLEAR(q);
4719 return result;
4721}
4722static int is_summary_empty(const struct diff_queue_struct *q)
4724{
4725 int i;
4726 for (i = 0; i < q->nr; i++) {
4728 const struct diff_filepair *p = q->queue[i];
4729 switch (p->status) {
4731 case DIFF_STATUS_DELETED:
4732 case DIFF_STATUS_ADDED:
4733 case DIFF_STATUS_COPIED:
4734 case DIFF_STATUS_RENAMED:
4735 return 0;
4736 default:
4737 if (p->score)
4738 return 0;
4739 if (p->one->mode && p->two->mode &&
4740 p->one->mode != p->two->mode)
4741 return 0;
4742 break;
4743 }
4744 }
4745 return 1;
4746}
4747static const char rename_limit_warning[] =
4749N_("inexact rename detection was skipped due to too many files.");
4750static const char degrade_cc_to_c_warning[] =
4752N_("only found copies from modified paths due to too many files.");
4753static const char rename_limit_advice[] =
4755N_("you may want to set your %s variable to at least "
4756 "%d and retry the command.");
4757void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4759{
4760 if (degraded_cc)
4761 warning(_(degrade_cc_to_c_warning));
4762 else if (needed)
4763 warning(_(rename_limit_warning));
4764 else
4765 return;
4766 if (0 < needed && needed < 32767)
4767 warning(_(rename_limit_advice), varname, needed);
4768}
4769static void diff_flush_patch_all_file_pairs(struct diff_options *o)
4771{
4772 int i;
4773 struct diff_queue_struct *q = &diff_queued_diff;
4774 for (i = 0; i < q->nr; i++) {
4775 struct diff_filepair *p = q->queue[i];
4776 if (check_pair_status(p))
4777 diff_flush_patch(p, o);
4778 }
4779}
4780void diff_flush(struct diff_options *options)
4782{
4783 struct diff_queue_struct *q = &diff_queued_diff;
4784 int i, output_format = options->output_format;
4785 int separator = 0;
4786 int dirstat_by_line = 0;
4787 /*
4789 * Order: raw, stat, summary, patch
4790 * or: name/name-status/checkdiff (other bits clear)
4791 */
4792 if (!q->nr)
4793 goto free_queue;
4794 if (output_format & (DIFF_FORMAT_RAW |
4796 DIFF_FORMAT_NAME |
4797 DIFF_FORMAT_NAME_STATUS |
4798 DIFF_FORMAT_CHECKDIFF)) {
4799 for (i = 0; i < q->nr; i++) {
4800 struct diff_filepair *p = q->queue[i];
4801 if (check_pair_status(p))
4802 flush_one_pair(p, options);
4803 }
4804 separator++;
4805 }
4806 if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4808 dirstat_by_line = 1;
4809 if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4811 dirstat_by_line) {
4812 struct diffstat_t diffstat;
4813 memset(&diffstat, 0, sizeof(struct diffstat_t));
4815 for (i = 0; i < q->nr; i++) {
4816 struct diff_filepair *p = q->queue[i];
4817 if (check_pair_status(p))
4818 diff_flush_stat(p, options, &diffstat);
4819 }
4820 if (output_format & DIFF_FORMAT_NUMSTAT)
4821 show_numstat(&diffstat, options);
4822 if (output_format & DIFF_FORMAT_DIFFSTAT)
4823 show_stats(&diffstat, options);
4824 if (output_format & DIFF_FORMAT_SHORTSTAT)
4825 show_shortstats(&diffstat, options);
4826 if (output_format & DIFF_FORMAT_DIRSTAT && dirstat_by_line)
4827 show_dirstat_by_line(&diffstat, options);
4828 free_diffstat_info(&diffstat);
4829 separator++;
4830 }
4831 if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4832 show_dirstat(options);
4833 if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4835 for (i = 0; i < q->nr; i++) {
4836 diff_summary(options, q->queue[i]);
4837 }
4838 separator++;
4839 }
4840 if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4842 DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4843 DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4844 /*
4845 * run diff_flush_patch for the exit status. setting
4846 * options->file to /dev/null should be safe, because we
4847 * aren't supposed to produce any output anyway.
4848 */
4849 if (options->close_file)
4850 fclose(options->file);
4851 options->file = xfopen("/dev/null", "w");
4852 options->close_file = 1;
4853 for (i = 0; i < q->nr; i++) {
4854 struct diff_filepair *p = q->queue[i];
4855 if (check_pair_status(p))
4856 diff_flush_patch(p, options);
4857 if (options->found_changes)
4858 break;
4859 }
4860 }
4861 if (output_format & DIFF_FORMAT_PATCH) {
4863 if (separator) {
4864 emit_diff_symbol(options, DIFF_SYMBOL_SEPARATOR, NULL, 0);
4865 if (options->stat_sep) {
4866 /* attach patch instead of inline */
4867 fputs(options->stat_sep, options->file);
4868 }
4869 }
4870 diff_flush_patch_all_file_pairs(options);
4872 }
4873 if (output_format & DIFF_FORMAT_CALLBACK)
4875 options->format_callback(q, options, options->format_callback_data);
4876 for (i = 0; i < q->nr; i++)
4878 diff_free_filepair(q->queue[i]);
4879free_queue:
4880 free(q->queue);
4881 DIFF_QUEUE_CLEAR(q);
4882 if (options->close_file)
4883 fclose(options->file);
4884 /*
4886 * Report the content-level differences with HAS_CHANGES;
4887 * diff_addremove/diff_change does not set the bit when
4888 * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4889 */
4890 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4891 if (options->found_changes)
4892 DIFF_OPT_SET(options, HAS_CHANGES);
4893 else
4894 DIFF_OPT_CLR(options, HAS_CHANGES);
4895 }
4896}
4897static int match_filter(const struct diff_options *options, const struct diff_filepair *p)
4899{
4900 return (((p->status == DIFF_STATUS_MODIFIED) &&
4901 ((p->score &&
4902 filter_bit_tst(DIFF_STATUS_FILTER_BROKEN, options)) ||
4903 (!p->score &&
4904 filter_bit_tst(DIFF_STATUS_MODIFIED, options)))) ||
4905 ((p->status != DIFF_STATUS_MODIFIED) &&
4906 filter_bit_tst(p->status, options)));
4907}
4908static void diffcore_apply_filter(struct diff_options *options)
4910{
4911 int i;
4912 struct diff_queue_struct *q = &diff_queued_diff;
4913 struct diff_queue_struct outq;
4914 DIFF_QUEUE_CLEAR(&outq);
4916 if (!options->filter)
4918 return;
4919 if (filter_bit_tst(DIFF_STATUS_FILTER_AON, options)) {
4921 int found;
4922 for (i = found = 0; !found && i < q->nr; i++) {
4923 if (match_filter(options, q->queue[i]))
4924 found++;
4925 }
4926 if (found)
4927 return;
4928 /* otherwise we will clear the whole queue
4930 * by copying the empty outq at the end of this
4931 * function, but first clear the current entries
4932 * in the queue.
4933 */
4934 for (i = 0; i < q->nr; i++)
4935 diff_free_filepair(q->queue[i]);
4936 }
4937 else {
4938 /* Only the matching ones */
4939 for (i = 0; i < q->nr; i++) {
4940 struct diff_filepair *p = q->queue[i];
4941 if (match_filter(options, p))
4942 diff_q(&outq, p);
4943 else
4944 diff_free_filepair(p);
4945 }
4946 }
4947 free(q->queue);
4948 *q = outq;
4949}
4950/* Check whether two filespecs with the same mode and size are identical */
4952static int diff_filespec_is_identical(struct diff_filespec *one,
4953 struct diff_filespec *two)
4954{
4955 if (S_ISGITLINK(one->mode))
4956 return 0;
4957 if (diff_populate_filespec(one, 0))
4958 return 0;
4959 if (diff_populate_filespec(two, 0))
4960 return 0;
4961 return !memcmp(one->data, two->data, one->size);
4962}
4963static int diff_filespec_check_stat_unmatch(struct diff_filepair *p)
4965{
4966 if (p->done_skip_stat_unmatch)
4967 return p->skip_stat_unmatch_result;
4968 p->done_skip_stat_unmatch = 1;
4970 p->skip_stat_unmatch_result = 0;
4971 /*
4972 * 1. Entries that come from stat info dirtiness
4973 * always have both sides (iow, not create/delete),
4974 * one side of the object name is unknown, with
4975 * the same mode and size. Keep the ones that
4976 * do not match these criteria. They have real
4977 * differences.
4978 *
4979 * 2. At this point, the file is known to be modified,
4980 * with the same mode and size, and the object
4981 * name of one side is unknown. Need to inspect
4982 * the identical contents.
4983 */
4984 if (!DIFF_FILE_VALID(p->one) || /* (1) */
4985 !DIFF_FILE_VALID(p->two) ||
4986 (p->one->oid_valid && p->two->oid_valid) ||
4987 (p->one->mode != p->two->mode) ||
4988 diff_populate_filespec(p->one, CHECK_SIZE_ONLY) ||
4989 diff_populate_filespec(p->two, CHECK_SIZE_ONLY) ||
4990 (p->one->size != p->two->size) ||
4991 !diff_filespec_is_identical(p->one, p->two)) /* (2) */
4992 p->skip_stat_unmatch_result = 1;
4993 return p->skip_stat_unmatch_result;
4994}
4995static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
4997{
4998 int i;
4999 struct diff_queue_struct *q = &diff_queued_diff;
5000 struct diff_queue_struct outq;
5001 DIFF_QUEUE_CLEAR(&outq);
5002 for (i = 0; i < q->nr; i++) {
5004 struct diff_filepair *p = q->queue[i];
5005 if (diff_filespec_check_stat_unmatch(p))
5007 diff_q(&outq, p);
5008 else {
5009 /*
5010 * The caller can subtract 1 from skip_stat_unmatch
5011 * to determine how many paths were dirty only
5012 * due to stat info mismatch.
5013 */
5014 if (!DIFF_OPT_TST(diffopt, NO_INDEX))
5015 diffopt->skip_stat_unmatch++;
5016 diff_free_filepair(p);
5017 }
5018 }
5019 free(q->queue);
5020 *q = outq;
5021}
5022static int diffnamecmp(const void *a_, const void *b_)
5024{
5025 const struct diff_filepair *a = *((const struct diff_filepair **)a_);
5026 const struct diff_filepair *b = *((const struct diff_filepair **)b_);
5027 const char *name_a, *name_b;
5028 name_a = a->one ? a->one->path : a->two->path;
5030 name_b = b->one ? b->one->path : b->two->path;
5031 return strcmp(name_a, name_b);
5032}
5033void diffcore_fix_diff_index(struct diff_options *options)
5035{
5036 struct diff_queue_struct *q = &diff_queued_diff;
5037 QSORT(q->queue, q->nr, diffnamecmp);
5038}
5039void diffcore_std(struct diff_options *options)
5041{
5042 /* NOTE please keep the following in sync with diff_tree_combined() */
5043 if (options->skip_stat_unmatch)
5044 diffcore_skip_stat_unmatch(options);
5045 if (!options->found_follow) {
5046 /* See try_to_follow_renames() in tree-diff.c */
5047 if (options->break_opt != -1)
5048 diffcore_break(options->break_opt);
5049 if (options->detect_rename)
5050 diffcore_rename(options);
5051 if (options->break_opt != -1)
5052 diffcore_merge_broken();
5053 }
5054 if (options->pickaxe)
5055 diffcore_pickaxe(options);
5056 if (options->orderfile)
5057 diffcore_order(options->orderfile);
5058 if (!options->found_follow)
5059 /* See try_to_follow_renames() in tree-diff.c */
5060 diff_resolve_rename_copy();
5061 diffcore_apply_filter(options);
5062 if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5064 DIFF_OPT_SET(options, HAS_CHANGES);
5065 else
5066 DIFF_OPT_CLR(options, HAS_CHANGES);
5067 options->found_follow = 0;
5069}
5070int diff_result_code(struct diff_options *opt, int status)
5072{
5073 int result = 0;
5074 diff_warn_rename_limit("diff.renameLimit",
5076 opt->needed_rename_limit,
5077 opt->degraded_cc_to_c);
5078 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5079 !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
5080 return status;
5081 if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5082 DIFF_OPT_TST(opt, HAS_CHANGES))
5083 result |= 01;
5084 if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
5085 DIFF_OPT_TST(opt, CHECK_FAILED))
5086 result |= 02;
5087 return result;
5088}
5089int diff_can_quit_early(struct diff_options *opt)
5091{
5092 return (DIFF_OPT_TST(opt, QUICK) &&
5093 !opt->filter &&
5094 DIFF_OPT_TST(opt, HAS_CHANGES));
5095}
5096/*
5098 * Shall changes to this submodule be ignored?
5099 *
5100 * Submodule changes can be configured to be ignored separately for each path,
5101 * but that configuration can be overridden from the command line.
5102 */
5103static int is_submodule_ignored(const char *path, struct diff_options *options)
5104{
5105 int ignored = 0;
5106 unsigned orig_flags = options->flags;
5107 if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
5108 set_diffopt_flags_from_submodule_config(options, path);
5109 if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
5110 ignored = 1;
5111 options->flags = orig_flags;
5112 return ignored;
5113}
5114void diff_addremove(struct diff_options *options,
5116 int addremove, unsigned mode,
5117 const struct object_id *oid,
5118 int oid_valid,
5119 const char *concatpath, unsigned dirty_submodule)
5120{
5121 struct diff_filespec *one, *two;
5122 if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
5124 return;
5125 /* This may look odd, but it is a preparation for
5127 * feeding "there are unchanged files which should
5128 * not produce diffs, but when you are doing copy
5129 * detection you would need them, so here they are"
5130 * entries to the diff-core. They will be prefixed
5131 * with something like '=' or '*' (I haven't decided
5132 * which but should not make any difference).
5133 * Feeding the same new and old to diff_change()
5134 * also has the same effect.
5135 * Before the final output happens, they are pruned after
5136 * merged into rename/copy pairs as appropriate.
5137 */
5138 if (DIFF_OPT_TST(options, REVERSE_DIFF))
5139 addremove = (addremove == '+' ? '-' :
5140 addremove == '-' ? '+' : addremove);
5141 if (options->prefix &&
5143 strncmp(concatpath, options->prefix, options->prefix_length))
5144 return;
5145 one = alloc_filespec(concatpath);
5147 two = alloc_filespec(concatpath);
5148 if (addremove != '+')
5150 fill_filespec(one, oid, oid_valid, mode);
5151 if (addremove != '-') {
5152 fill_filespec(two, oid, oid_valid, mode);
5153 two->dirty_submodule = dirty_submodule;
5154 }
5155 diff_queue(&diff_queued_diff, one, two);
5157 if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5158 DIFF_OPT_SET(options, HAS_CHANGES);
5159}
5160void diff_change(struct diff_options *options,
5162 unsigned old_mode, unsigned new_mode,
5163 const struct object_id *old_oid,
5164 const struct object_id *new_oid,
5165 int old_oid_valid, int new_oid_valid,
5166 const char *concatpath,
5167 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
5168{
5169 struct diff_filespec *one, *two;
5170 struct diff_filepair *p;
5171 if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
5173 is_submodule_ignored(concatpath, options))
5174 return;
5175 if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
5177 SWAP(old_mode, new_mode);
5178 SWAP(old_oid, new_oid);
5179 SWAP(old_oid_valid, new_oid_valid);
5180 SWAP(old_dirty_submodule, new_dirty_submodule);
5181 }
5182 if (options->prefix &&
5184 strncmp(concatpath, options->prefix, options->prefix_length))
5185 return;
5186 one = alloc_filespec(concatpath);
5188 two = alloc_filespec(concatpath);
5189 fill_filespec(one, old_oid, old_oid_valid, old_mode);
5190 fill_filespec(two, new_oid, new_oid_valid, new_mode);
5191 one->dirty_submodule = old_dirty_submodule;
5192 two->dirty_submodule = new_dirty_submodule;
5193 p = diff_queue(&diff_queued_diff, one, two);
5194 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5196 return;
5197 if (DIFF_OPT_TST(options, QUICK) && options->skip_stat_unmatch &&
5199 !diff_filespec_check_stat_unmatch(p))
5200 return;
5201 DIFF_OPT_SET(options, HAS_CHANGES);
5203}
5204struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
5206{
5207 struct diff_filepair *pair;
5208 struct diff_filespec *one, *two;
5209 if (options->prefix &&
5211 strncmp(path, options->prefix, options->prefix_length))
5212 return NULL;
5213 one = alloc_filespec(path);
5215 two = alloc_filespec(path);
5216 pair = diff_queue(&diff_queued_diff, one, two);
5217 pair->is_unmerged = 1;
5218 return pair;
5219}
5220static char *run_textconv(const char *pgm, struct diff_filespec *spec,
5222 size_t *outsize)
5223{
5224 struct diff_tempfile *temp;
5225 const char *argv[3];
5226 const char **arg = argv;
5227 struct child_process child = CHILD_PROCESS_INIT;
5228 struct strbuf buf = STRBUF_INIT;
5229 int err = 0;
5230 temp = prepare_temp_file(spec->path, spec);
5232 *arg++ = pgm;
5233 *arg++ = temp->name;
5234 *arg = NULL;
5235 child.use_shell = 1;
5237 child.argv = argv;
5238 child.out = -1;
5239 if (start_command(&child)) {
5240 remove_tempfile();
5241 return NULL;
5242 }
5243 if (strbuf_read(&buf, child.out, 0) < 0)
5245 err = error("error reading from textconv command '%s'", pgm);
5246 close(child.out);
5247 if (finish_command(&child) || err) {
5249 strbuf_release(&buf);
5250 remove_tempfile();
5251 return NULL;
5252 }
5253 remove_tempfile();
5254 return strbuf_detach(&buf, outsize);
5256}
5257size_t fill_textconv(struct userdiff_driver *driver,
5259 struct diff_filespec *df,
5260 char **outbuf)
5261{
5262 size_t size;
5263 if (!driver) {
5265 if (!DIFF_FILE_VALID(df)) {
5266 *outbuf = "";
5267 return 0;
5268 }
5269 if (diff_populate_filespec(df, 0))
5270 die("unable to read files to diff");
5271 *outbuf = df->data;
5272 return df->size;
5273 }
5274 if (!driver->textconv)
5276 die("BUG: fill_textconv called with non-textconv driver");
5277 if (driver->textconv_cache && df->oid_valid) {
5279 *outbuf = notes_cache_get(driver->textconv_cache,
5280 &df->oid,
5281 &size);
5282 if (*outbuf)
5283 return size;
5284 }
5285 *outbuf = run_textconv(driver->textconv, df, &size);
5287 if (!*outbuf)
5288 die("unable to read files to diff");
5289 if (driver->textconv_cache && df->oid_valid) {
5291 /* ignore errors, as we might be in a readonly repository */
5292 notes_cache_put(driver->textconv_cache, &df->oid, *outbuf,
5293 size);
5294 /*
5295 * we could save up changes and flush them all at the end,
5296 * but we would need an extra call after all diffing is done.
5297 * Since generating a cache entry is the slow path anyway,
5298 * this extra overhead probably isn't a big deal.
5299 */
5300 notes_cache_write(driver->textconv_cache);
5301 }
5302 return size;
5304}
5305int textconv_object(const char *path,
5307 unsigned mode,
5308 const struct object_id *oid,
5309 int oid_valid,
5310 char **buf,
5311 unsigned long *buf_size)
5312{
5313 struct diff_filespec *df;
5314 struct userdiff_driver *textconv;
5315 df = alloc_filespec(path);
5317 fill_filespec(df, oid, oid_valid, mode);
5318 textconv = get_textconv(df);
5319 if (!textconv) {
5320 free_filespec(df);
5321 return 0;
5322 }
5323 *buf_size = fill_textconv(textconv, df, buf);
5325 free_filespec(df);
5326 return 1;
5327}
5328void setup_diff_pager(struct diff_options *opt)
5330{
5331 /*
5332 * If the user asked for our exit code, then either they want --quiet
5333 * or --exit-code. We should definitely not bother with a pager in the
5334 * former case, as we will generate no output. Since we still properly
5335 * report our exit code even when a pager is run, we _could_ run a
5336 * pager with --exit-code. But since we have not done so historically,
5337 * and because it is easy to find people oneline advising "git diff
5338 * --exit-code" in hooks and other scripts, we do not do so.
5339 */
5340 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5341 check_pager_config("diff") != 0)
5342 setup_pager();
5343}