The fourth batch post 2.14
[git/git.git] / diff.c
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
24 #ifdef NO_FAST_WORKING_DIRECTORY
25 #define FAST_WORKING_DIRECTORY 0
26 #else
27 #define FAST_WORKING_DIRECTORY 1
28 #endif
29
30 static int diff_detect_rename_default;
31 static int diff_indent_heuristic = 1;
32 static int diff_rename_limit_default = 400;
33 static int diff_suppress_blank_empty;
34 static int diff_use_color_default = -1;
35 static int diff_context_default = 3;
36 static int diff_interhunk_context_default;
37 static const char *diff_word_regex_cfg;
38 static const char *external_diff_cmd_cfg;
39 static const char *diff_order_file_cfg;
40 int diff_auto_refresh_index = 1;
41 static int diff_mnemonic_prefix;
42 static int diff_no_prefix;
43 static int diff_stat_graph_width;
44 static int diff_dirstat_permille_default = 30;
45 static struct diff_options default_diff_options;
46 static long diff_algorithm;
47 static unsigned ws_error_highlight_default = WSEH_NEW;
48
49 static 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 };
60
61 static NORETURN void die_want_option(const char *option_name)
62 {
63 die(_("option '%s' requires a value"), option_name);
64 }
65
66 static 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 }
86
87 static 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
95 if (*params_copy)
96 string_list_split_in_place(&params, 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
134 }
135 string_list_clear(&params, 0);
136 free(params_copy);
137 return ret;
138 }
139
140 static 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 }
152
153 static 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 }
161
162 long 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 }
176
177 static 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 }
186
187 static int parse_ws_error_highlight(const char *arg)
188 {
189 const char *orig_arg = arg;
190 unsigned val = 0;
191
192 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
214 /*
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 */
220 void init_diff_ui_defaults(void)
221 {
222 diff_detect_rename_default = 1;
223 }
224
225 int 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 }
231
232 int 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
277 if (!strcmp(var, "diff.ignoresubmodules"))
278 handle_ignore_submodules_arg(&default_diff_options, value);
279
280 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
287 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
294 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
302 return git_diff_basic_config(var, value, cb);
303 }
304
305 int git_diff_basic_config(const char *var, const char *value, void *cb)
306 {
307 const char *name;
308
309 if (!strcmp(var, "diff.renamelimit")) {
310 diff_rename_limit_default = git_config_int(var, value);
311 return 0;
312 }
313
314 if (userdiff_config(var, value) < 0)
315 return -1;
316
317 if (skip_prefix(var, "diff.color.", &name) ||
318 skip_prefix(var, "color.diff.", &name)) {
319 int slot = parse_diff_color_slot(name);
320 if (slot < 0)
321 return 0;
322 if (!value)
323 return config_error_nonbool(var);
324 return color_parse(value, diff_colors[slot]);
325 }
326
327 /* like GNU diff's --suppress-blank-empty option */
328 if (!strcmp(var, "diff.suppressblankempty") ||
329 /* for backwards compatibility */
330 !strcmp(var, "diff.suppress-blank-empty")) {
331 diff_suppress_blank_empty = git_config_bool(var, value);
332 return 0;
333 }
334
335 if (!strcmp(var, "diff.dirstat")) {
336 struct strbuf errmsg = STRBUF_INIT;
337 default_diff_options.dirstat_permille = diff_dirstat_permille_default;
338 if (parse_dirstat_params(&default_diff_options, value, &errmsg))
339 warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
340 errmsg.buf);
341 strbuf_release(&errmsg);
342 diff_dirstat_permille_default = default_diff_options.dirstat_permille;
343 return 0;
344 }
345
346 if (starts_with(var, "submodule."))
347 return parse_submodule_config_option(var, value);
348
349 if (git_diff_heuristic_config(var, value, cb) < 0)
350 return -1;
351
352 return git_default_config(var, value, cb);
353 }
354
355 static char *quote_two(const char *one, const char *two)
356 {
357 int need_one = quote_c_style(one, NULL, NULL, 1);
358 int need_two = quote_c_style(two, NULL, NULL, 1);
359 struct strbuf res = STRBUF_INIT;
360
361 if (need_one + need_two) {
362 strbuf_addch(&res, '"');
363 quote_c_style(one, &res, NULL, 1);
364 quote_c_style(two, &res, NULL, 1);
365 strbuf_addch(&res, '"');
366 } else {
367 strbuf_addstr(&res, one);
368 strbuf_addstr(&res, two);
369 }
370 return strbuf_detach(&res, NULL);
371 }
372
373 static const char *external_diff(void)
374 {
375 static const char *external_diff_cmd = NULL;
376 static int done_preparing = 0;
377
378 if (done_preparing)
379 return external_diff_cmd;
380 external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
381 if (!external_diff_cmd)
382 external_diff_cmd = external_diff_cmd_cfg;
383 done_preparing = 1;
384 return external_diff_cmd;
385 }
386
387 /*
388 * Keep track of files used for diffing. Sometimes such an entry
389 * refers to a temporary file, sometimes to an existing file, and
390 * sometimes to "/dev/null".
391 */
392 static struct diff_tempfile {
393 /*
394 * filename external diff should read from, or NULL if this
395 * entry is currently not in use:
396 */
397 const char *name;
398
399 char hex[GIT_MAX_HEXSZ + 1];
400 char mode[10];
401
402 /*
403 * If this diff_tempfile instance refers to a temporary file,
404 * this tempfile object is used to manage its lifetime.
405 */
406 struct tempfile tempfile;
407 } diff_temp[2];
408
409 struct emit_callback {
410 int color_diff;
411 unsigned ws_rule;
412 int blank_at_eof_in_preimage;
413 int blank_at_eof_in_postimage;
414 int lno_in_preimage;
415 int lno_in_postimage;
416 const char **label_path;
417 struct diff_words_data *diff_words;
418 struct diff_options *opt;
419 struct strbuf *header;
420 };
421
422 static int count_lines(const char *data, int size)
423 {
424 int count, ch, completely_empty = 1, nl_just_seen = 0;
425 count = 0;
426 while (0 < size--) {
427 ch = *data++;
428 if (ch == '\n') {
429 count++;
430 nl_just_seen = 1;
431 completely_empty = 0;
432 }
433 else {
434 nl_just_seen = 0;
435 completely_empty = 0;
436 }
437 }
438 if (completely_empty)
439 return 0;
440 if (!nl_just_seen)
441 count++; /* no trailing newline */
442 return count;
443 }
444
445 static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
446 {
447 if (!DIFF_FILE_VALID(one)) {
448 mf->ptr = (char *)""; /* does not matter */
449 mf->size = 0;
450 return 0;
451 }
452 else if (diff_populate_filespec(one, 0))
453 return -1;
454
455 mf->ptr = one->data;
456 mf->size = one->size;
457 return 0;
458 }
459
460 /* like fill_mmfile, but only for size, so we can avoid retrieving blob */
461 static unsigned long diff_filespec_size(struct diff_filespec *one)
462 {
463 if (!DIFF_FILE_VALID(one))
464 return 0;
465 diff_populate_filespec(one, CHECK_SIZE_ONLY);
466 return one->size;
467 }
468
469 static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
470 {
471 char *ptr = mf->ptr;
472 long size = mf->size;
473 int cnt = 0;
474
475 if (!size)
476 return cnt;
477 ptr += size - 1; /* pointing at the very end */
478 if (*ptr != '\n')
479 ; /* incomplete line */
480 else
481 ptr--; /* skip the last LF */
482 while (mf->ptr < ptr) {
483 char *prev_eol;
484 for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
485 if (*prev_eol == '\n')
486 break;
487 if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
488 break;
489 cnt++;
490 ptr = prev_eol - 1;
491 }
492 return cnt;
493 }
494
495 static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
496 struct emit_callback *ecbdata)
497 {
498 int l1, l2, at;
499 unsigned ws_rule = ecbdata->ws_rule;
500 l1 = count_trailing_blank(mf1, ws_rule);
501 l2 = count_trailing_blank(mf2, ws_rule);
502 if (l2 <= l1) {
503 ecbdata->blank_at_eof_in_preimage = 0;
504 ecbdata->blank_at_eof_in_postimage = 0;
505 return;
506 }
507 at = count_lines(mf1->ptr, mf1->size);
508 ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
509
510 at = count_lines(mf2->ptr, mf2->size);
511 ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
512 }
513
514 static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
515 int first, const char *line, int len)
516 {
517 int has_trailing_newline, has_trailing_carriage_return;
518 int nofirst;
519 FILE *file = o->file;
520
521 fputs(diff_line_prefix(o), file);
522
523 if (len == 0) {
524 has_trailing_newline = (first == '\n');
525 has_trailing_carriage_return = (!has_trailing_newline &&
526 (first == '\r'));
527 nofirst = has_trailing_newline || has_trailing_carriage_return;
528 } else {
529 has_trailing_newline = (len > 0 && line[len-1] == '\n');
530 if (has_trailing_newline)
531 len--;
532 has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
533 if (has_trailing_carriage_return)
534 len--;
535 nofirst = 0;
536 }
537
538 if (len || !nofirst) {
539 fputs(set, file);
540 if (!nofirst)
541 fputc(first, file);
542 fwrite(line, len, 1, file);
543 fputs(reset, file);
544 }
545 if (has_trailing_carriage_return)
546 fputc('\r', file);
547 if (has_trailing_newline)
548 fputc('\n', file);
549 }
550
551 static void emit_line(struct diff_options *o, const char *set, const char *reset,
552 const char *line, int len)
553 {
554 emit_line_0(o, set, reset, line[0], line+1, len-1);
555 }
556
557 static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
558 {
559 if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
560 ecbdata->blank_at_eof_in_preimage &&
561 ecbdata->blank_at_eof_in_postimage &&
562 ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
563 ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
564 return 0;
565 return ws_blank_line(line, len, ecbdata->ws_rule);
566 }
567
568 static void emit_line_checked(const char *reset,
569 struct emit_callback *ecbdata,
570 const char *line, int len,
571 enum color_diff color,
572 unsigned ws_error_highlight,
573 char sign)
574 {
575 const char *set = diff_get_color(ecbdata->color_diff, color);
576 const char *ws = NULL;
577
578 if (ecbdata->opt->ws_error_highlight & ws_error_highlight) {
579 ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
580 if (!*ws)
581 ws = NULL;
582 }
583
584 if (!ws)
585 emit_line_0(ecbdata->opt, set, reset, sign, line, len);
586 else if (sign == '+' && new_blank_line_at_eof(ecbdata, line, len))
587 /* Blank line at EOF - paint '+' as well */
588 emit_line_0(ecbdata->opt, ws, reset, sign, line, len);
589 else {
590 /* Emit just the prefix, then the rest. */
591 emit_line_0(ecbdata->opt, set, reset, sign, "", 0);
592 ws_check_emit(line, len, ecbdata->ws_rule,
593 ecbdata->opt->file, set, reset, ws);
594 }
595 }
596
597 static void emit_add_line(const char *reset,
598 struct emit_callback *ecbdata,
599 const char *line, int len)
600 {
601 emit_line_checked(reset, ecbdata, line, len,
602 DIFF_FILE_NEW, WSEH_NEW, '+');
603 }
604
605 static void emit_del_line(const char *reset,
606 struct emit_callback *ecbdata,
607 const char *line, int len)
608 {
609 emit_line_checked(reset, ecbdata, line, len,
610 DIFF_FILE_OLD, WSEH_OLD, '-');
611 }
612
613 static void emit_context_line(const char *reset,
614 struct emit_callback *ecbdata,
615 const char *line, int len)
616 {
617 emit_line_checked(reset, ecbdata, line, len,
618 DIFF_CONTEXT, WSEH_CONTEXT, ' ');
619 }
620
621 static void emit_hunk_header(struct emit_callback *ecbdata,
622 const char *line, int len)
623 {
624 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
625 const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
626 const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
627 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
628 static const char atat[2] = { '@', '@' };
629 const char *cp, *ep;
630 struct strbuf msgbuf = STRBUF_INIT;
631 int org_len = len;
632 int i = 1;
633
634 /*
635 * As a hunk header must begin with "@@ -<old>, +<new> @@",
636 * it always is at least 10 bytes long.
637 */
638 if (len < 10 ||
639 memcmp(line, atat, 2) ||
640 !(ep = memmem(line + 2, len - 2, atat, 2))) {
641 emit_line(ecbdata->opt, context, reset, line, len);
642 return;
643 }
644 ep += 2; /* skip over @@ */
645
646 /* The hunk header in fraginfo color */
647 strbuf_addstr(&msgbuf, frag);
648 strbuf_add(&msgbuf, line, ep - line);
649 strbuf_addstr(&msgbuf, reset);
650
651 /*
652 * trailing "\r\n"
653 */
654 for ( ; i < 3; i++)
655 if (line[len - i] == '\r' || line[len - i] == '\n')
656 len--;
657
658 /* blank before the func header */
659 for (cp = ep; ep - line < len; ep++)
660 if (*ep != ' ' && *ep != '\t')
661 break;
662 if (ep != cp) {
663 strbuf_addstr(&msgbuf, context);
664 strbuf_add(&msgbuf, cp, ep - cp);
665 strbuf_addstr(&msgbuf, reset);
666 }
667
668 if (ep < line + len) {
669 strbuf_addstr(&msgbuf, func);
670 strbuf_add(&msgbuf, ep, line + len - ep);
671 strbuf_addstr(&msgbuf, reset);
672 }
673
674 strbuf_add(&msgbuf, line + len, org_len - len);
675 emit_line(ecbdata->opt, "", "", msgbuf.buf, msgbuf.len);
676 strbuf_release(&msgbuf);
677 }
678
679 static struct diff_tempfile *claim_diff_tempfile(void) {
680 int i;
681 for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
682 if (!diff_temp[i].name)
683 return diff_temp + i;
684 die("BUG: diff is failing to clean up its tempfiles");
685 }
686
687 static void remove_tempfile(void)
688 {
689 int i;
690 for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
691 if (is_tempfile_active(&diff_temp[i].tempfile))
692 delete_tempfile(&diff_temp[i].tempfile);
693 diff_temp[i].name = NULL;
694 }
695 }
696
697 static void print_line_count(FILE *file, int count)
698 {
699 switch (count) {
700 case 0:
701 fprintf(file, "0,0");
702 break;
703 case 1:
704 fprintf(file, "1");
705 break;
706 default:
707 fprintf(file, "1,%d", count);
708 break;
709 }
710 }
711
712 static void emit_rewrite_lines(struct emit_callback *ecb,
713 int prefix, const char *data, int size)
714 {
715 const char *endp = NULL;
716 static const char *nneof = " No newline at end of file\n";
717 const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
718
719 while (0 < size) {
720 int len;
721
722 endp = memchr(data, '\n', size);
723 len = endp ? (endp - data + 1) : size;
724 if (prefix != '+') {
725 ecb->lno_in_preimage++;
726 emit_del_line(reset, ecb, data, len);
727 } else {
728 ecb->lno_in_postimage++;
729 emit_add_line(reset, ecb, data, len);
730 }
731 size -= len;
732 data += len;
733 }
734 if (!endp) {
735 const char *context = diff_get_color(ecb->color_diff,
736 DIFF_CONTEXT);
737 putc('\n', ecb->opt->file);
738 emit_line_0(ecb->opt, context, reset, '\\',
739 nneof, strlen(nneof));
740 }
741 }
742
743 static void emit_rewrite_diff(const char *name_a,
744 const char *name_b,
745 struct diff_filespec *one,
746 struct diff_filespec *two,
747 struct userdiff_driver *textconv_one,
748 struct userdiff_driver *textconv_two,
749 struct diff_options *o)
750 {
751 int lc_a, lc_b;
752 const char *name_a_tab, *name_b_tab;
753 const char *metainfo = diff_get_color(o->use_color, DIFF_METAINFO);
754 const char *fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
755 const char *reset = diff_get_color(o->use_color, DIFF_RESET);
756 static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
757 const char *a_prefix, *b_prefix;
758 char *data_one, *data_two;
759 size_t size_one, size_two;
760 struct emit_callback ecbdata;
761 const char *line_prefix = diff_line_prefix(o);
762
763 if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
764 a_prefix = o->b_prefix;
765 b_prefix = o->a_prefix;
766 } else {
767 a_prefix = o->a_prefix;
768 b_prefix = o->b_prefix;
769 }
770
771 name_a += (*name_a == '/');
772 name_b += (*name_b == '/');
773 name_a_tab = strchr(name_a, ' ') ? "\t" : "";
774 name_b_tab = strchr(name_b, ' ') ? "\t" : "";
775
776 strbuf_reset(&a_name);
777 strbuf_reset(&b_name);
778 quote_two_c_style(&a_name, a_prefix, name_a, 0);
779 quote_two_c_style(&b_name, b_prefix, name_b, 0);
780
781 size_one = fill_textconv(textconv_one, one, &data_one);
782 size_two = fill_textconv(textconv_two, two, &data_two);
783
784 memset(&ecbdata, 0, sizeof(ecbdata));
785 ecbdata.color_diff = want_color(o->use_color);
786 ecbdata.ws_rule = whitespace_rule(name_b);
787 ecbdata.opt = o;
788 if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
789 mmfile_t mf1, mf2;
790 mf1.ptr = (char *)data_one;
791 mf2.ptr = (char *)data_two;
792 mf1.size = size_one;
793 mf2.size = size_two;
794 check_blank_at_eof(&mf1, &mf2, &ecbdata);
795 }
796 ecbdata.lno_in_preimage = 1;
797 ecbdata.lno_in_postimage = 1;
798
799 lc_a = count_lines(data_one, size_one);
800 lc_b = count_lines(data_two, size_two);
801 fprintf(o->file,
802 "%s%s--- %s%s%s\n%s%s+++ %s%s%s\n%s%s@@ -",
803 line_prefix, metainfo, a_name.buf, name_a_tab, reset,
804 line_prefix, metainfo, b_name.buf, name_b_tab, reset,
805 line_prefix, fraginfo);
806 if (!o->irreversible_delete)
807 print_line_count(o->file, lc_a);
808 else
809 fprintf(o->file, "?,?");
810 fprintf(o->file, " +");
811 print_line_count(o->file, lc_b);
812 fprintf(o->file, " @@%s\n", reset);
813 if (lc_a && !o->irreversible_delete)
814 emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
815 if (lc_b)
816 emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
817 if (textconv_one)
818 free((char *)data_one);
819 if (textconv_two)
820 free((char *)data_two);
821 }
822
823 struct diff_words_buffer {
824 mmfile_t text;
825 long alloc;
826 struct diff_words_orig {
827 const char *begin, *end;
828 } *orig;
829 int orig_nr, orig_alloc;
830 };
831
832 static void diff_words_append(char *line, unsigned long len,
833 struct diff_words_buffer *buffer)
834 {
835 ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
836 line++;
837 len--;
838 memcpy(buffer->text.ptr + buffer->text.size, line, len);
839 buffer->text.size += len;
840 buffer->text.ptr[buffer->text.size] = '\0';
841 }
842
843 struct diff_words_style_elem {
844 const char *prefix;
845 const char *suffix;
846 const char *color; /* NULL; filled in by the setup code if
847 * color is enabled */
848 };
849
850 struct diff_words_style {
851 enum diff_words_type type;
852 struct diff_words_style_elem new, old, ctx;
853 const char *newline;
854 };
855
856 static struct diff_words_style diff_words_styles[] = {
857 { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
858 { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
859 { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
860 };
861
862 struct diff_words_data {
863 struct diff_words_buffer minus, plus;
864 const char *current_plus;
865 int last_minus;
866 struct diff_options *opt;
867 regex_t *word_regex;
868 enum diff_words_type type;
869 struct diff_words_style *style;
870 };
871
872 static int fn_out_diff_words_write_helper(FILE *fp,
873 struct diff_words_style_elem *st_el,
874 const char *newline,
875 size_t count, const char *buf,
876 const char *line_prefix)
877 {
878 int print = 0;
879
880 while (count) {
881 char *p = memchr(buf, '\n', count);
882 if (print)
883 fputs(line_prefix, fp);
884 if (p != buf) {
885 if (st_el->color && fputs(st_el->color, fp) < 0)
886 return -1;
887 if (fputs(st_el->prefix, fp) < 0 ||
888 fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
889 fputs(st_el->suffix, fp) < 0)
890 return -1;
891 if (st_el->color && *st_el->color
892 && fputs(GIT_COLOR_RESET, fp) < 0)
893 return -1;
894 }
895 if (!p)
896 return 0;
897 if (fputs(newline, fp) < 0)
898 return -1;
899 count -= p + 1 - buf;
900 buf = p + 1;
901 print = 1;
902 }
903 return 0;
904 }
905
906 /*
907 * '--color-words' algorithm can be described as:
908 *
909 * 1. collect the minus/plus lines of a diff hunk, divided into
910 * minus-lines and plus-lines;
911 *
912 * 2. break both minus-lines and plus-lines into words and
913 * place them into two mmfile_t with one word for each line;
914 *
915 * 3. use xdiff to run diff on the two mmfile_t to get the words level diff;
916 *
917 * And for the common parts of the both file, we output the plus side text.
918 * diff_words->current_plus is used to trace the current position of the plus file
919 * which printed. diff_words->last_minus is used to trace the last minus word
920 * printed.
921 *
922 * For '--graph' to work with '--color-words', we need to output the graph prefix
923 * on each line of color words output. Generally, there are two conditions on
924 * which we should output the prefix.
925 *
926 * 1. diff_words->last_minus == 0 &&
927 * diff_words->current_plus == diff_words->plus.text.ptr
928 *
929 * that is: the plus text must start as a new line, and if there is no minus
930 * word printed, a graph prefix must be printed.
931 *
932 * 2. diff_words->current_plus > diff_words->plus.text.ptr &&
933 * *(diff_words->current_plus - 1) == '\n'
934 *
935 * that is: a graph prefix must be printed following a '\n'
936 */
937 static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
938 {
939 if ((diff_words->last_minus == 0 &&
940 diff_words->current_plus == diff_words->plus.text.ptr) ||
941 (diff_words->current_plus > diff_words->plus.text.ptr &&
942 *(diff_words->current_plus - 1) == '\n')) {
943 return 1;
944 } else {
945 return 0;
946 }
947 }
948
949 static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
950 {
951 struct diff_words_data *diff_words = priv;
952 struct diff_words_style *style = diff_words->style;
953 int minus_first, minus_len, plus_first, plus_len;
954 const char *minus_begin, *minus_end, *plus_begin, *plus_end;
955 struct diff_options *opt = diff_words->opt;
956 const char *line_prefix;
957
958 if (line[0] != '@' || parse_hunk_header(line, len,
959 &minus_first, &minus_len, &plus_first, &plus_len))
960 return;
961
962 assert(opt);
963 line_prefix = diff_line_prefix(opt);
964
965 /* POSIX requires that first be decremented by one if len == 0... */
966 if (minus_len) {
967 minus_begin = diff_words->minus.orig[minus_first].begin;
968 minus_end =
969 diff_words->minus.orig[minus_first + minus_len - 1].end;
970 } else
971 minus_begin = minus_end =
972 diff_words->minus.orig[minus_first].end;
973
974 if (plus_len) {
975 plus_begin = diff_words->plus.orig[plus_first].begin;
976 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
977 } else
978 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
979
980 if (color_words_output_graph_prefix(diff_words)) {
981 fputs(line_prefix, diff_words->opt->file);
982 }
983 if (diff_words->current_plus != plus_begin) {
984 fn_out_diff_words_write_helper(diff_words->opt->file,
985 &style->ctx, style->newline,
986 plus_begin - diff_words->current_plus,
987 diff_words->current_plus, line_prefix);
988 if (*(plus_begin - 1) == '\n')
989 fputs(line_prefix, diff_words->opt->file);
990 }
991 if (minus_begin != minus_end) {
992 fn_out_diff_words_write_helper(diff_words->opt->file,
993 &style->old, style->newline,
994 minus_end - minus_begin, minus_begin,
995 line_prefix);
996 }
997 if (plus_begin != plus_end) {
998 fn_out_diff_words_write_helper(diff_words->opt->file,
999 &style->new, style->newline,
1000 plus_end - plus_begin, plus_begin,
1001 line_prefix);
1002 }
1003
1004 diff_words->current_plus = plus_end;
1005 diff_words->last_minus = minus_first;
1006 }
1007
1008 /* This function starts looking at *begin, and returns 0 iff a word was found. */
1009 static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
1010 int *begin, int *end)
1011 {
1012 if (word_regex && *begin < buffer->size) {
1013 regmatch_t match[1];
1014 if (!regexec_buf(word_regex, buffer->ptr + *begin,
1015 buffer->size - *begin, 1, match, 0)) {
1016 char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
1017 '\n', match[0].rm_eo - match[0].rm_so);
1018 *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
1019 *begin += match[0].rm_so;
1020 return *begin >= *end;
1021 }
1022 return -1;
1023 }
1024
1025 /* find the next word */
1026 while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
1027 (*begin)++;
1028 if (*begin >= buffer->size)
1029 return -1;
1030
1031 /* find the end of the word */
1032 *end = *begin + 1;
1033 while (*end < buffer->size && !isspace(buffer->ptr[*end]))
1034 (*end)++;
1035
1036 return 0;
1037 }
1038
1039 /*
1040 * This function splits the words in buffer->text, stores the list with
1041 * newline separator into out, and saves the offsets of the original words
1042 * in buffer->orig.
1043 */
1044 static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
1045 regex_t *word_regex)
1046 {
1047 int i, j;
1048 long alloc = 0;
1049
1050 out->size = 0;
1051 out->ptr = NULL;
1052
1053 /* fake an empty "0th" word */
1054 ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
1055 buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
1056 buffer->orig_nr = 1;
1057
1058 for (i = 0; i < buffer->text.size; i++) {
1059 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
1060 return;
1061
1062 /* store original boundaries */
1063 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
1064 buffer->orig_alloc);
1065 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
1066 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
1067 buffer->orig_nr++;
1068
1069 /* store one word */
1070 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
1071 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
1072 out->ptr[out->size + j - i] = '\n';
1073 out->size += j - i + 1;
1074
1075 i = j - 1;
1076 }
1077 }
1078
1079 /* this executes the word diff on the accumulated buffers */
1080 static void diff_words_show(struct diff_words_data *diff_words)
1081 {
1082 xpparam_t xpp;
1083 xdemitconf_t xecfg;
1084 mmfile_t minus, plus;
1085 struct diff_words_style *style = diff_words->style;
1086
1087 struct diff_options *opt = diff_words->opt;
1088 const char *line_prefix;
1089
1090 assert(opt);
1091 line_prefix = diff_line_prefix(opt);
1092
1093 /* special case: only removal */
1094 if (!diff_words->plus.text.size) {
1095 fputs(line_prefix, diff_words->opt->file);
1096 fn_out_diff_words_write_helper(diff_words->opt->file,
1097 &style->old, style->newline,
1098 diff_words->minus.text.size,
1099 diff_words->minus.text.ptr, line_prefix);
1100 diff_words->minus.text.size = 0;
1101 return;
1102 }
1103
1104 diff_words->current_plus = diff_words->plus.text.ptr;
1105 diff_words->last_minus = 0;
1106
1107 memset(&xpp, 0, sizeof(xpp));
1108 memset(&xecfg, 0, sizeof(xecfg));
1109 diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
1110 diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
1111 xpp.flags = 0;
1112 /* as only the hunk header will be parsed, we need a 0-context */
1113 xecfg.ctxlen = 0;
1114 if (xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
1115 &xpp, &xecfg))
1116 die("unable to generate word diff");
1117 free(minus.ptr);
1118 free(plus.ptr);
1119 if (diff_words->current_plus != diff_words->plus.text.ptr +
1120 diff_words->plus.text.size) {
1121 if (color_words_output_graph_prefix(diff_words))
1122 fputs(line_prefix, diff_words->opt->file);
1123 fn_out_diff_words_write_helper(diff_words->opt->file,
1124 &style->ctx, style->newline,
1125 diff_words->plus.text.ptr + diff_words->plus.text.size
1126 - diff_words->current_plus, diff_words->current_plus,
1127 line_prefix);
1128 }
1129 diff_words->minus.text.size = diff_words->plus.text.size = 0;
1130 }
1131
1132 /* In "color-words" mode, show word-diff of words accumulated in the buffer */
1133 static void diff_words_flush(struct emit_callback *ecbdata)
1134 {
1135 if (ecbdata->diff_words->minus.text.size ||
1136 ecbdata->diff_words->plus.text.size)
1137 diff_words_show(ecbdata->diff_words);
1138 }
1139
1140 static void diff_filespec_load_driver(struct diff_filespec *one)
1141 {
1142 /* Use already-loaded driver */
1143 if (one->driver)
1144 return;
1145
1146 if (S_ISREG(one->mode))
1147 one->driver = userdiff_find_by_path(one->path);
1148
1149 /* Fallback to default settings */
1150 if (!one->driver)
1151 one->driver = userdiff_find_by_name("default");
1152 }
1153
1154 static const char *userdiff_word_regex(struct diff_filespec *one)
1155 {
1156 diff_filespec_load_driver(one);
1157 return one->driver->word_regex;
1158 }
1159
1160 static void init_diff_words_data(struct emit_callback *ecbdata,
1161 struct diff_options *orig_opts,
1162 struct diff_filespec *one,
1163 struct diff_filespec *two)
1164 {
1165 int i;
1166 struct diff_options *o = xmalloc(sizeof(struct diff_options));
1167 memcpy(o, orig_opts, sizeof(struct diff_options));
1168
1169 ecbdata->diff_words =
1170 xcalloc(1, sizeof(struct diff_words_data));
1171 ecbdata->diff_words->type = o->word_diff;
1172 ecbdata->diff_words->opt = o;
1173 if (!o->word_regex)
1174 o->word_regex = userdiff_word_regex(one);
1175 if (!o->word_regex)
1176 o->word_regex = userdiff_word_regex(two);
1177 if (!o->word_regex)
1178 o->word_regex = diff_word_regex_cfg;
1179 if (o->word_regex) {
1180 ecbdata->diff_words->word_regex = (regex_t *)
1181 xmalloc(sizeof(regex_t));
1182 if (regcomp(ecbdata->diff_words->word_regex,
1183 o->word_regex,
1184 REG_EXTENDED | REG_NEWLINE))
1185 die ("Invalid regular expression: %s",
1186 o->word_regex);
1187 }
1188 for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1189 if (o->word_diff == diff_words_styles[i].type) {
1190 ecbdata->diff_words->style =
1191 &diff_words_styles[i];
1192 break;
1193 }
1194 }
1195 if (want_color(o->use_color)) {
1196 struct diff_words_style *st = ecbdata->diff_words->style;
1197 st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1198 st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1199 st->ctx.color = diff_get_color_opt(o, DIFF_CONTEXT);
1200 }
1201 }
1202
1203 static void free_diff_words_data(struct emit_callback *ecbdata)
1204 {
1205 if (ecbdata->diff_words) {
1206 diff_words_flush(ecbdata);
1207 free (ecbdata->diff_words->opt);
1208 free (ecbdata->diff_words->minus.text.ptr);
1209 free (ecbdata->diff_words->minus.orig);
1210 free (ecbdata->diff_words->plus.text.ptr);
1211 free (ecbdata->diff_words->plus.orig);
1212 if (ecbdata->diff_words->word_regex) {
1213 regfree(ecbdata->diff_words->word_regex);
1214 free(ecbdata->diff_words->word_regex);
1215 }
1216 FREE_AND_NULL(ecbdata->diff_words);
1217 }
1218 }
1219
1220 const char *diff_get_color(int diff_use_color, enum color_diff ix)
1221 {
1222 if (want_color(diff_use_color))
1223 return diff_colors[ix];
1224 return "";
1225 }
1226
1227 const char *diff_line_prefix(struct diff_options *opt)
1228 {
1229 struct strbuf *msgbuf;
1230 if (!opt->output_prefix)
1231 return "";
1232
1233 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
1234 return msgbuf->buf;
1235 }
1236
1237 static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1238 {
1239 const char *cp;
1240 unsigned long allot;
1241 size_t l = len;
1242
1243 cp = line;
1244 allot = l;
1245 while (0 < l) {
1246 (void) utf8_width(&cp, &l);
1247 if (!cp)
1248 break; /* truncated in the middle? */
1249 }
1250 return allot - l;
1251 }
1252
1253 static void find_lno(const char *line, struct emit_callback *ecbdata)
1254 {
1255 const char *p;
1256 ecbdata->lno_in_preimage = 0;
1257 ecbdata->lno_in_postimage = 0;
1258 p = strchr(line, '-');
1259 if (!p)
1260 return; /* cannot happen */
1261 ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1262 p = strchr(p, '+');
1263 if (!p)
1264 return; /* cannot happen */
1265 ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1266 }
1267
1268 static void fn_out_consume(void *priv, char *line, unsigned long len)
1269 {
1270 struct emit_callback *ecbdata = priv;
1271 const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
1272 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
1273 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1274 struct diff_options *o = ecbdata->opt;
1275 const char *line_prefix = diff_line_prefix(o);
1276
1277 o->found_changes = 1;
1278
1279 if (ecbdata->header) {
1280 fprintf(o->file, "%s", ecbdata->header->buf);
1281 strbuf_reset(ecbdata->header);
1282 ecbdata->header = NULL;
1283 }
1284
1285 if (ecbdata->label_path[0]) {
1286 const char *name_a_tab, *name_b_tab;
1287
1288 name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
1289 name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
1290
1291 fprintf(o->file, "%s%s--- %s%s%s\n",
1292 line_prefix, meta, ecbdata->label_path[0], reset, name_a_tab);
1293 fprintf(o->file, "%s%s+++ %s%s%s\n",
1294 line_prefix, meta, ecbdata->label_path[1], reset, name_b_tab);
1295 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1296 }
1297
1298 if (diff_suppress_blank_empty
1299 && len == 2 && line[0] == ' ' && line[1] == '\n') {
1300 line[0] = '\n';
1301 len = 1;
1302 }
1303
1304 if (line[0] == '@') {
1305 if (ecbdata->diff_words)
1306 diff_words_flush(ecbdata);
1307 len = sane_truncate_line(ecbdata, line, len);
1308 find_lno(line, ecbdata);
1309 emit_hunk_header(ecbdata, line, len);
1310 if (line[len-1] != '\n')
1311 putc('\n', o->file);
1312 return;
1313 }
1314
1315 if (ecbdata->diff_words) {
1316 if (line[0] == '-') {
1317 diff_words_append(line, len,
1318 &ecbdata->diff_words->minus);
1319 return;
1320 } else if (line[0] == '+') {
1321 diff_words_append(line, len,
1322 &ecbdata->diff_words->plus);
1323 return;
1324 } else if (starts_with(line, "\\ ")) {
1325 /*
1326 * Eat the "no newline at eof" marker as if we
1327 * saw a "+" or "-" line with nothing on it,
1328 * and return without diff_words_flush() to
1329 * defer processing. If this is the end of
1330 * preimage, more "+" lines may come after it.
1331 */
1332 return;
1333 }
1334 diff_words_flush(ecbdata);
1335 if (ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN) {
1336 emit_line(o, context, reset, line, len);
1337 fputs("~\n", o->file);
1338 } else {
1339 /*
1340 * Skip the prefix character, if any. With
1341 * diff_suppress_blank_empty, there may be
1342 * none.
1343 */
1344 if (line[0] != '\n') {
1345 line++;
1346 len--;
1347 }
1348 emit_line(o, context, reset, line, len);
1349 }
1350 return;
1351 }
1352
1353 switch (line[0]) {
1354 case '+':
1355 ecbdata->lno_in_postimage++;
1356 emit_add_line(reset, ecbdata, line + 1, len - 1);
1357 break;
1358 case '-':
1359 ecbdata->lno_in_preimage++;
1360 emit_del_line(reset, ecbdata, line + 1, len - 1);
1361 break;
1362 case ' ':
1363 ecbdata->lno_in_postimage++;
1364 ecbdata->lno_in_preimage++;
1365 emit_context_line(reset, ecbdata, line + 1, len - 1);
1366 break;
1367 default:
1368 /* incomplete line at the end */
1369 ecbdata->lno_in_preimage++;
1370 emit_line(o, diff_get_color(ecbdata->color_diff, DIFF_CONTEXT),
1371 reset, line, len);
1372 break;
1373 }
1374 }
1375
1376 static char *pprint_rename(const char *a, const char *b)
1377 {
1378 const char *old = a;
1379 const char *new = b;
1380 struct strbuf name = STRBUF_INIT;
1381 int pfx_length, sfx_length;
1382 int pfx_adjust_for_slash;
1383 int len_a = strlen(a);
1384 int len_b = strlen(b);
1385 int a_midlen, b_midlen;
1386 int qlen_a = quote_c_style(a, NULL, NULL, 0);
1387 int qlen_b = quote_c_style(b, NULL, NULL, 0);
1388
1389 if (qlen_a || qlen_b) {
1390 quote_c_style(a, &name, NULL, 0);
1391 strbuf_addstr(&name, " => ");
1392 quote_c_style(b, &name, NULL, 0);
1393 return strbuf_detach(&name, NULL);
1394 }
1395
1396 /* Find common prefix */
1397 pfx_length = 0;
1398 while (*old && *new && *old == *new) {
1399 if (*old == '/')
1400 pfx_length = old - a + 1;
1401 old++;
1402 new++;
1403 }
1404
1405 /* Find common suffix */
1406 old = a + len_a;
1407 new = b + len_b;
1408 sfx_length = 0;
1409 /*
1410 * If there is a common prefix, it must end in a slash. In
1411 * that case we let this loop run 1 into the prefix to see the
1412 * same slash.
1413 *
1414 * If there is no common prefix, we cannot do this as it would
1415 * underrun the input strings.
1416 */
1417 pfx_adjust_for_slash = (pfx_length ? 1 : 0);
1418 while (a + pfx_length - pfx_adjust_for_slash <= old &&
1419 b + pfx_length - pfx_adjust_for_slash <= new &&
1420 *old == *new) {
1421 if (*old == '/')
1422 sfx_length = len_a - (old - a);
1423 old--;
1424 new--;
1425 }
1426
1427 /*
1428 * pfx{mid-a => mid-b}sfx
1429 * {pfx-a => pfx-b}sfx
1430 * pfx{sfx-a => sfx-b}
1431 * name-a => name-b
1432 */
1433 a_midlen = len_a - pfx_length - sfx_length;
1434 b_midlen = len_b - pfx_length - sfx_length;
1435 if (a_midlen < 0)
1436 a_midlen = 0;
1437 if (b_midlen < 0)
1438 b_midlen = 0;
1439
1440 strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1441 if (pfx_length + sfx_length) {
1442 strbuf_add(&name, a, pfx_length);
1443 strbuf_addch(&name, '{');
1444 }
1445 strbuf_add(&name, a + pfx_length, a_midlen);
1446 strbuf_addstr(&name, " => ");
1447 strbuf_add(&name, b + pfx_length, b_midlen);
1448 if (pfx_length + sfx_length) {
1449 strbuf_addch(&name, '}');
1450 strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1451 }
1452 return strbuf_detach(&name, NULL);
1453 }
1454
1455 struct diffstat_t {
1456 int nr;
1457 int alloc;
1458 struct diffstat_file {
1459 char *from_name;
1460 char *name;
1461 char *print_name;
1462 unsigned is_unmerged:1;
1463 unsigned is_binary:1;
1464 unsigned is_renamed:1;
1465 unsigned is_interesting:1;
1466 uintmax_t added, deleted;
1467 } **files;
1468 };
1469
1470 static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1471 const char *name_a,
1472 const char *name_b)
1473 {
1474 struct diffstat_file *x;
1475 x = xcalloc(1, sizeof(*x));
1476 ALLOC_GROW(diffstat->files, diffstat->nr + 1, diffstat->alloc);
1477 diffstat->files[diffstat->nr++] = x;
1478 if (name_b) {
1479 x->from_name = xstrdup(name_a);
1480 x->name = xstrdup(name_b);
1481 x->is_renamed = 1;
1482 }
1483 else {
1484 x->from_name = NULL;
1485 x->name = xstrdup(name_a);
1486 }
1487 return x;
1488 }
1489
1490 static void diffstat_consume(void *priv, char *line, unsigned long len)
1491 {
1492 struct diffstat_t *diffstat = priv;
1493 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1494
1495 if (line[0] == '+')
1496 x->added++;
1497 else if (line[0] == '-')
1498 x->deleted++;
1499 }
1500
1501 const char mime_boundary_leader[] = "------------";
1502
1503 static int scale_linear(int it, int width, int max_change)
1504 {
1505 if (!it)
1506 return 0;
1507 /*
1508 * make sure that at least one '-' or '+' is printed if
1509 * there is any change to this path. The easiest way is to
1510 * scale linearly as if the alloted width is one column shorter
1511 * than it is, and then add 1 to the result.
1512 */
1513 return 1 + (it * (width - 1) / max_change);
1514 }
1515
1516 static void show_name(FILE *file,
1517 const char *prefix, const char *name, int len)
1518 {
1519 fprintf(file, " %s%-*s |", prefix, len, name);
1520 }
1521
1522 static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1523 {
1524 if (cnt <= 0)
1525 return;
1526 fprintf(file, "%s", set);
1527 while (cnt--)
1528 putc(ch, file);
1529 fprintf(file, "%s", reset);
1530 }
1531
1532 static void fill_print_name(struct diffstat_file *file)
1533 {
1534 char *pname;
1535
1536 if (file->print_name)
1537 return;
1538
1539 if (!file->is_renamed) {
1540 struct strbuf buf = STRBUF_INIT;
1541 if (quote_c_style(file->name, &buf, NULL, 0)) {
1542 pname = strbuf_detach(&buf, NULL);
1543 } else {
1544 pname = file->name;
1545 strbuf_release(&buf);
1546 }
1547 } else {
1548 pname = pprint_rename(file->from_name, file->name);
1549 }
1550 file->print_name = pname;
1551 }
1552
1553 int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1554 {
1555 struct strbuf sb = STRBUF_INIT;
1556 int ret;
1557
1558 if (!files) {
1559 assert(insertions == 0 && deletions == 0);
1560 return fprintf(fp, "%s\n", " 0 files changed");
1561 }
1562
1563 strbuf_addf(&sb,
1564 (files == 1) ? " %d file changed" : " %d files changed",
1565 files);
1566
1567 /*
1568 * For binary diff, the caller may want to print "x files
1569 * changed" with insertions == 0 && deletions == 0.
1570 *
1571 * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1572 * is probably less confusing (i.e skip over "2 files changed
1573 * but nothing about added/removed lines? Is this a bug in Git?").
1574 */
1575 if (insertions || deletions == 0) {
1576 strbuf_addf(&sb,
1577 (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
1578 insertions);
1579 }
1580
1581 if (deletions || insertions == 0) {
1582 strbuf_addf(&sb,
1583 (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
1584 deletions);
1585 }
1586 strbuf_addch(&sb, '\n');
1587 ret = fputs(sb.buf, fp);
1588 strbuf_release(&sb);
1589 return ret;
1590 }
1591
1592 static void show_stats(struct diffstat_t *data, struct diff_options *options)
1593 {
1594 int i, len, add, del, adds = 0, dels = 0;
1595 uintmax_t max_change = 0, max_len = 0;
1596 int total_files = data->nr, count;
1597 int width, name_width, graph_width, number_width = 0, bin_width = 0;
1598 const char *reset, *add_c, *del_c;
1599 const char *line_prefix = "";
1600 int extra_shown = 0;
1601
1602 if (data->nr == 0)
1603 return;
1604
1605 line_prefix = diff_line_prefix(options);
1606 count = options->stat_count ? options->stat_count : data->nr;
1607
1608 reset = diff_get_color_opt(options, DIFF_RESET);
1609 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1610 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1611
1612 /*
1613 * Find the longest filename and max number of changes
1614 */
1615 for (i = 0; (i < count) && (i < data->nr); i++) {
1616 struct diffstat_file *file = data->files[i];
1617 uintmax_t change = file->added + file->deleted;
1618
1619 if (!file->is_interesting && (change == 0)) {
1620 count++; /* not shown == room for one more */
1621 continue;
1622 }
1623 fill_print_name(file);
1624 len = strlen(file->print_name);
1625 if (max_len < len)
1626 max_len = len;
1627
1628 if (file->is_unmerged) {
1629 /* "Unmerged" is 8 characters */
1630 bin_width = bin_width < 8 ? 8 : bin_width;
1631 continue;
1632 }
1633 if (file->is_binary) {
1634 /* "Bin XXX -> YYY bytes" */
1635 int w = 14 + decimal_width(file->added)
1636 + decimal_width(file->deleted);
1637 bin_width = bin_width < w ? w : bin_width;
1638 /* Display change counts aligned with "Bin" */
1639 number_width = 3;
1640 continue;
1641 }
1642
1643 if (max_change < change)
1644 max_change = change;
1645 }
1646 count = i; /* where we can stop scanning in data->files[] */
1647
1648 /*
1649 * We have width = stat_width or term_columns() columns total.
1650 * We want a maximum of min(max_len, stat_name_width) for the name part.
1651 * We want a maximum of min(max_change, stat_graph_width) for the +- part.
1652 * We also need 1 for " " and 4 + decimal_width(max_change)
1653 * for " | NNNN " and one the empty column at the end, altogether
1654 * 6 + decimal_width(max_change).
1655 *
1656 * If there's not enough space, we will use the smaller of
1657 * stat_name_width (if set) and 5/8*width for the filename,
1658 * and the rest for constant elements + graph part, but no more
1659 * than stat_graph_width for the graph part.
1660 * (5/8 gives 50 for filename and 30 for the constant parts + graph
1661 * for the standard terminal size).
1662 *
1663 * In other words: stat_width limits the maximum width, and
1664 * stat_name_width fixes the maximum width of the filename,
1665 * and is also used to divide available columns if there
1666 * aren't enough.
1667 *
1668 * Binary files are displayed with "Bin XXX -> YYY bytes"
1669 * instead of the change count and graph. This part is treated
1670 * similarly to the graph part, except that it is not
1671 * "scaled". If total width is too small to accommodate the
1672 * guaranteed minimum width of the filename part and the
1673 * separators and this message, this message will "overflow"
1674 * making the line longer than the maximum width.
1675 */
1676
1677 if (options->stat_width == -1)
1678 width = term_columns() - strlen(line_prefix);
1679 else
1680 width = options->stat_width ? options->stat_width : 80;
1681 number_width = decimal_width(max_change) > number_width ?
1682 decimal_width(max_change) : number_width;
1683
1684 if (options->stat_graph_width == -1)
1685 options->stat_graph_width = diff_stat_graph_width;
1686
1687 /*
1688 * Guarantee 3/8*16==6 for the graph part
1689 * and 5/8*16==10 for the filename part
1690 */
1691 if (width < 16 + 6 + number_width)
1692 width = 16 + 6 + number_width;
1693
1694 /*
1695 * First assign sizes that are wanted, ignoring available width.
1696 * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
1697 * starting from "XXX" should fit in graph_width.
1698 */
1699 graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
1700 if (options->stat_graph_width &&
1701 options->stat_graph_width < graph_width)
1702 graph_width = options->stat_graph_width;
1703
1704 name_width = (options->stat_name_width > 0 &&
1705 options->stat_name_width < max_len) ?
1706 options->stat_name_width : max_len;
1707
1708 /*
1709 * Adjust adjustable widths not to exceed maximum width
1710 */
1711 if (name_width + number_width + 6 + graph_width > width) {
1712 if (graph_width > width * 3/8 - number_width - 6) {
1713 graph_width = width * 3/8 - number_width - 6;
1714 if (graph_width < 6)
1715 graph_width = 6;
1716 }
1717
1718 if (options->stat_graph_width &&
1719 graph_width > options->stat_graph_width)
1720 graph_width = options->stat_graph_width;
1721 if (name_width > width - number_width - 6 - graph_width)
1722 name_width = width - number_width - 6 - graph_width;
1723 else
1724 graph_width = width - number_width - 6 - name_width;
1725 }
1726
1727 /*
1728 * From here name_width is the width of the name area,
1729 * and graph_width is the width of the graph area.
1730 * max_change is used to scale graph properly.
1731 */
1732 for (i = 0; i < count; i++) {
1733 const char *prefix = "";
1734 struct diffstat_file *file = data->files[i];
1735 char *name = file->print_name;
1736 uintmax_t added = file->added;
1737 uintmax_t deleted = file->deleted;
1738 int name_len;
1739
1740 if (!file->is_interesting && (added + deleted == 0))
1741 continue;
1742
1743 /*
1744 * "scale" the filename
1745 */
1746 len = name_width;
1747 name_len = strlen(name);
1748 if (name_width < name_len) {
1749 char *slash;
1750 prefix = "...";
1751 len -= 3;
1752 name += name_len - len;
1753 slash = strchr(name, '/');
1754 if (slash)
1755 name = slash;
1756 }
1757
1758 if (file->is_binary) {
1759 fprintf(options->file, "%s", line_prefix);
1760 show_name(options->file, prefix, name, len);
1761 fprintf(options->file, " %*s", number_width, "Bin");
1762 if (!added && !deleted) {
1763 putc('\n', options->file);
1764 continue;
1765 }
1766 fprintf(options->file, " %s%"PRIuMAX"%s",
1767 del_c, deleted, reset);
1768 fprintf(options->file, " -> ");
1769 fprintf(options->file, "%s%"PRIuMAX"%s",
1770 add_c, added, reset);
1771 fprintf(options->file, " bytes");
1772 fprintf(options->file, "\n");
1773 continue;
1774 }
1775 else if (file->is_unmerged) {
1776 fprintf(options->file, "%s", line_prefix);
1777 show_name(options->file, prefix, name, len);
1778 fprintf(options->file, " Unmerged\n");
1779 continue;
1780 }
1781
1782 /*
1783 * scale the add/delete
1784 */
1785 add = added;
1786 del = deleted;
1787
1788 if (graph_width <= max_change) {
1789 int total = scale_linear(add + del, graph_width, max_change);
1790 if (total < 2 && add && del)
1791 /* width >= 2 due to the sanity check */
1792 total = 2;
1793 if (add < del) {
1794 add = scale_linear(add, graph_width, max_change);
1795 del = total - add;
1796 } else {
1797 del = scale_linear(del, graph_width, max_change);
1798 add = total - del;
1799 }
1800 }
1801 fprintf(options->file, "%s", line_prefix);
1802 show_name(options->file, prefix, name, len);
1803 fprintf(options->file, " %*"PRIuMAX"%s",
1804 number_width, added + deleted,
1805 added + deleted ? " " : "");
1806 show_graph(options->file, '+', add, add_c, reset);
1807 show_graph(options->file, '-', del, del_c, reset);
1808 fprintf(options->file, "\n");
1809 }
1810
1811 for (i = 0; i < data->nr; i++) {
1812 struct diffstat_file *file = data->files[i];
1813 uintmax_t added = file->added;
1814 uintmax_t deleted = file->deleted;
1815
1816 if (file->is_unmerged ||
1817 (!file->is_interesting && (added + deleted == 0))) {
1818 total_files--;
1819 continue;
1820 }
1821
1822 if (!file->is_binary) {
1823 adds += added;
1824 dels += deleted;
1825 }
1826 if (i < count)
1827 continue;
1828 if (!extra_shown)
1829 fprintf(options->file, "%s ...\n", line_prefix);
1830 extra_shown = 1;
1831 }
1832 fprintf(options->file, "%s", line_prefix);
1833 print_stat_summary(options->file, total_files, adds, dels);
1834 }
1835
1836 static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1837 {
1838 int i, adds = 0, dels = 0, total_files = data->nr;
1839
1840 if (data->nr == 0)
1841 return;
1842
1843 for (i = 0; i < data->nr; i++) {
1844 int added = data->files[i]->added;
1845 int deleted= data->files[i]->deleted;
1846
1847 if (data->files[i]->is_unmerged ||
1848 (!data->files[i]->is_interesting && (added + deleted == 0))) {
1849 total_files--;
1850 } else if (!data->files[i]->is_binary) { /* don't count bytes */
1851 adds += added;
1852 dels += deleted;
1853 }
1854 }
1855 fprintf(options->file, "%s", diff_line_prefix(options));
1856 print_stat_summary(options->file, total_files, adds, dels);
1857 }
1858
1859 static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1860 {
1861 int i;
1862
1863 if (data->nr == 0)
1864 return;
1865
1866 for (i = 0; i < data->nr; i++) {
1867 struct diffstat_file *file = data->files[i];
1868
1869 fprintf(options->file, "%s", diff_line_prefix(options));
1870
1871 if (file->is_binary)
1872 fprintf(options->file, "-\t-\t");
1873 else
1874 fprintf(options->file,
1875 "%"PRIuMAX"\t%"PRIuMAX"\t",
1876 file->added, file->deleted);
1877 if (options->line_termination) {
1878 fill_print_name(file);
1879 if (!file->is_renamed)
1880 write_name_quoted(file->name, options->file,
1881 options->line_termination);
1882 else {
1883 fputs(file->print_name, options->file);
1884 putc(options->line_termination, options->file);
1885 }
1886 } else {
1887 if (file->is_renamed) {
1888 putc('\0', options->file);
1889 write_name_quoted(file->from_name, options->file, '\0');
1890 }
1891 write_name_quoted(file->name, options->file, '\0');
1892 }
1893 }
1894 }
1895
1896 struct dirstat_file {
1897 const char *name;
1898 unsigned long changed;
1899 };
1900
1901 struct dirstat_dir {
1902 struct dirstat_file *files;
1903 int alloc, nr, permille, cumulative;
1904 };
1905
1906 static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1907 unsigned long changed, const char *base, int baselen)
1908 {
1909 unsigned long this_dir = 0;
1910 unsigned int sources = 0;
1911 const char *line_prefix = diff_line_prefix(opt);
1912
1913 while (dir->nr) {
1914 struct dirstat_file *f = dir->files;
1915 int namelen = strlen(f->name);
1916 unsigned long this;
1917 char *slash;
1918
1919 if (namelen < baselen)
1920 break;
1921 if (memcmp(f->name, base, baselen))
1922 break;
1923 slash = strchr(f->name + baselen, '/');
1924 if (slash) {
1925 int newbaselen = slash + 1 - f->name;
1926 this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
1927 sources++;
1928 } else {
1929 this = f->changed;
1930 dir->files++;
1931 dir->nr--;
1932 sources += 2;
1933 }
1934 this_dir += this;
1935 }
1936
1937 /*
1938 * We don't report dirstat's for
1939 * - the top level
1940 * - or cases where everything came from a single directory
1941 * under this directory (sources == 1).
1942 */
1943 if (baselen && sources != 1) {
1944 if (this_dir) {
1945 int permille = this_dir * 1000 / changed;
1946 if (permille >= dir->permille) {
1947 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
1948 permille / 10, permille % 10, baselen, base);
1949 if (!dir->cumulative)
1950 return 0;
1951 }
1952 }
1953 }
1954 return this_dir;
1955 }
1956
1957 static int dirstat_compare(const void *_a, const void *_b)
1958 {
1959 const struct dirstat_file *a = _a;
1960 const struct dirstat_file *b = _b;
1961 return strcmp(a->name, b->name);
1962 }
1963
1964 static void show_dirstat(struct diff_options *options)
1965 {
1966 int i;
1967 unsigned long changed;
1968 struct dirstat_dir dir;
1969 struct diff_queue_struct *q = &diff_queued_diff;
1970
1971 dir.files = NULL;
1972 dir.alloc = 0;
1973 dir.nr = 0;
1974 dir.permille = options->dirstat_permille;
1975 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1976
1977 changed = 0;
1978 for (i = 0; i < q->nr; i++) {
1979 struct diff_filepair *p = q->queue[i];
1980 const char *name;
1981 unsigned long copied, added, damage;
1982 int content_changed;
1983
1984 name = p->two->path ? p->two->path : p->one->path;
1985
1986 if (p->one->oid_valid && p->two->oid_valid)
1987 content_changed = oidcmp(&p->one->oid, &p->two->oid);
1988 else
1989 content_changed = 1;
1990
1991 if (!content_changed) {
1992 /*
1993 * The SHA1 has not changed, so pre-/post-content is
1994 * identical. We can therefore skip looking at the
1995 * file contents altogether.
1996 */
1997 damage = 0;
1998 goto found_damage;
1999 }
2000
2001 if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
2002 /*
2003 * In --dirstat-by-file mode, we don't really need to
2004 * look at the actual file contents at all.
2005 * The fact that the SHA1 changed is enough for us to
2006 * add this file to the list of results
2007 * (with each file contributing equal damage).
2008 */
2009 damage = 1;
2010 goto found_damage;
2011 }
2012
2013 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
2014 diff_populate_filespec(p->one, 0);
2015 diff_populate_filespec(p->two, 0);
2016 diffcore_count_changes(p->one, p->two, NULL, NULL,
2017 &copied, &added);
2018 diff_free_filespec_data(p->one);
2019 diff_free_filespec_data(p->two);
2020 } else if (DIFF_FILE_VALID(p->one)) {
2021 diff_populate_filespec(p->one, CHECK_SIZE_ONLY);
2022 copied = added = 0;
2023 diff_free_filespec_data(p->one);
2024 } else if (DIFF_FILE_VALID(p->two)) {
2025 diff_populate_filespec(p->two, CHECK_SIZE_ONLY);
2026 copied = 0;
2027 added = p->two->size;
2028 diff_free_filespec_data(p->two);
2029 } else
2030 continue;
2031
2032 /*
2033 * Original minus copied is the removed material,
2034 * added is the new material. They are both damages
2035 * made to the preimage.
2036 * If the resulting damage is zero, we know that
2037 * diffcore_count_changes() considers the two entries to
2038 * be identical, but since content_changed is true, we
2039 * know that there must have been _some_ kind of change,
2040 * so we force all entries to have damage > 0.
2041 */
2042 damage = (p->one->size - copied) + added;
2043 if (!damage)
2044 damage = 1;
2045
2046 found_damage:
2047 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2048 dir.files[dir.nr].name = name;
2049 dir.files[dir.nr].changed = damage;
2050 changed += damage;
2051 dir.nr++;
2052 }
2053
2054 /* This can happen even with many files, if everything was renames */
2055 if (!changed)
2056 return;
2057
2058 /* Show all directories with more than x% of the changes */
2059 QSORT(dir.files, dir.nr, dirstat_compare);
2060 gather_dirstat(options, &dir, changed, "", 0);
2061 }
2062
2063 static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
2064 {
2065 int i;
2066 unsigned long changed;
2067 struct dirstat_dir dir;
2068
2069 if (data->nr == 0)
2070 return;
2071
2072 dir.files = NULL;
2073 dir.alloc = 0;
2074 dir.nr = 0;
2075 dir.permille = options->dirstat_permille;
2076 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
2077
2078 changed = 0;
2079 for (i = 0; i < data->nr; i++) {
2080 struct diffstat_file *file = data->files[i];
2081 unsigned long damage = file->added + file->deleted;
2082 if (file->is_binary)
2083 /*
2084 * binary files counts bytes, not lines. Must find some
2085 * way to normalize binary bytes vs. textual lines.
2086 * The following heuristic assumes that there are 64
2087 * bytes per "line".
2088 * This is stupid and ugly, but very cheap...
2089 */
2090 damage = DIV_ROUND_UP(damage, 64);
2091 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2092 dir.files[dir.nr].name = file->name;
2093 dir.files[dir.nr].changed = damage;
2094 changed += damage;
2095 dir.nr++;
2096 }
2097
2098 /* This can happen even with many files, if everything was renames */
2099 if (!changed)
2100 return;
2101
2102 /* Show all directories with more than x% of the changes */
2103 QSORT(dir.files, dir.nr, dirstat_compare);
2104 gather_dirstat(options, &dir, changed, "", 0);
2105 }
2106
2107 static void free_diffstat_info(struct diffstat_t *diffstat)
2108 {
2109 int i;
2110 for (i = 0; i < diffstat->nr; i++) {
2111 struct diffstat_file *f = diffstat->files[i];
2112 if (f->name != f->print_name)
2113 free(f->print_name);
2114 free(f->name);
2115 free(f->from_name);
2116 free(f);
2117 }
2118 free(diffstat->files);
2119 }
2120
2121 struct checkdiff_t {
2122 const char *filename;
2123 int lineno;
2124 int conflict_marker_size;
2125 struct diff_options *o;
2126 unsigned ws_rule;
2127 unsigned status;
2128 };
2129
2130 static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2131 {
2132 char firstchar;
2133 int cnt;
2134
2135 if (len < marker_size + 1)
2136 return 0;
2137 firstchar = line[0];
2138 switch (firstchar) {
2139 case '=': case '>': case '<': case '|':
2140 break;
2141 default:
2142 return 0;
2143 }
2144 for (cnt = 1; cnt < marker_size; cnt++)
2145 if (line[cnt] != firstchar)
2146 return 0;
2147 /* line[1] thru line[marker_size-1] are same as firstchar */
2148 if (len < marker_size + 1 || !isspace(line[marker_size]))
2149 return 0;
2150 return 1;
2151 }
2152
2153 static void checkdiff_consume(void *priv, char *line, unsigned long len)
2154 {
2155 struct checkdiff_t *data = priv;
2156 int marker_size = data->conflict_marker_size;
2157 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2158 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2159 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2160 char *err;
2161 const char *line_prefix;
2162
2163 assert(data->o);
2164 line_prefix = diff_line_prefix(data->o);
2165
2166 if (line[0] == '+') {
2167 unsigned bad;
2168 data->lineno++;
2169 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2170 data->status |= 1;
2171 fprintf(data->o->file,
2172 "%s%s:%d: leftover conflict marker\n",
2173 line_prefix, data->filename, data->lineno);
2174 }
2175 bad = ws_check(line + 1, len - 1, data->ws_rule);
2176 if (!bad)
2177 return;
2178 data->status |= bad;
2179 err = whitespace_error_string(bad);
2180 fprintf(data->o->file, "%s%s:%d: %s.\n",
2181 line_prefix, data->filename, data->lineno, err);
2182 free(err);
2183 emit_line(data->o, set, reset, line, 1);
2184 ws_check_emit(line + 1, len - 1, data->ws_rule,
2185 data->o->file, set, reset, ws);
2186 } else if (line[0] == ' ') {
2187 data->lineno++;
2188 } else if (line[0] == '@') {
2189 char *plus = strchr(line, '+');
2190 if (plus)
2191 data->lineno = strtol(plus, NULL, 10) - 1;
2192 else
2193 die("invalid diff");
2194 }
2195 }
2196
2197 static unsigned char *deflate_it(char *data,
2198 unsigned long size,
2199 unsigned long *result_size)
2200 {
2201 int bound;
2202 unsigned char *deflated;
2203 git_zstream stream;
2204
2205 git_deflate_init(&stream, zlib_compression_level);
2206 bound = git_deflate_bound(&stream, size);
2207 deflated = xmalloc(bound);
2208 stream.next_out = deflated;
2209 stream.avail_out = bound;
2210
2211 stream.next_in = (unsigned char *)data;
2212 stream.avail_in = size;
2213 while (git_deflate(&stream, Z_FINISH) == Z_OK)
2214 ; /* nothing */
2215 git_deflate_end(&stream);
2216 *result_size = stream.total_out;
2217 return deflated;
2218 }
2219
2220 static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two,
2221 const char *prefix)
2222 {
2223 void *cp;
2224 void *delta;
2225 void *deflated;
2226 void *data;
2227 unsigned long orig_size;
2228 unsigned long delta_size;
2229 unsigned long deflate_size;
2230 unsigned long data_size;
2231
2232 /* We could do deflated delta, or we could do just deflated two,
2233 * whichever is smaller.
2234 */
2235 delta = NULL;
2236 deflated = deflate_it(two->ptr, two->size, &deflate_size);
2237 if (one->size && two->size) {
2238 delta = diff_delta(one->ptr, one->size,
2239 two->ptr, two->size,
2240 &delta_size, deflate_size);
2241 if (delta) {
2242 void *to_free = delta;
2243 orig_size = delta_size;
2244 delta = deflate_it(delta, delta_size, &delta_size);
2245 free(to_free);
2246 }
2247 }
2248
2249 if (delta && delta_size < deflate_size) {
2250 fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2251 free(deflated);
2252 data = delta;
2253 data_size = delta_size;
2254 }
2255 else {
2256 fprintf(file, "%sliteral %lu\n", prefix, two->size);
2257 free(delta);
2258 data = deflated;
2259 data_size = deflate_size;
2260 }
2261
2262 /* emit data encoded in base85 */
2263 cp = data;
2264 while (data_size) {
2265 int bytes = (52 < data_size) ? 52 : data_size;
2266 char line[70];
2267 data_size -= bytes;
2268 if (bytes <= 26)
2269 line[0] = bytes + 'A' - 1;
2270 else
2271 line[0] = bytes - 26 + 'a' - 1;
2272 encode_85(line + 1, cp, bytes);
2273 cp = (char *) cp + bytes;
2274 fprintf(file, "%s", prefix);
2275 fputs(line, file);
2276 fputc('\n', file);
2277 }
2278 fprintf(file, "%s\n", prefix);
2279 free(data);
2280 }
2281
2282 static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two,
2283 const char *prefix)
2284 {
2285 fprintf(file, "%sGIT binary patch\n", prefix);
2286 emit_binary_diff_body(file, one, two, prefix);
2287 emit_binary_diff_body(file, two, one, prefix);
2288 }
2289
2290 int diff_filespec_is_binary(struct diff_filespec *one)
2291 {
2292 if (one->is_binary == -1) {
2293 diff_filespec_load_driver(one);
2294 if (one->driver->binary != -1)
2295 one->is_binary = one->driver->binary;
2296 else {
2297 if (!one->data && DIFF_FILE_VALID(one))
2298 diff_populate_filespec(one, CHECK_BINARY);
2299 if (one->is_binary == -1 && one->data)
2300 one->is_binary = buffer_is_binary(one->data,
2301 one->size);
2302 if (one->is_binary == -1)
2303 one->is_binary = 0;
2304 }
2305 }
2306 return one->is_binary;
2307 }
2308
2309 static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2310 {
2311 diff_filespec_load_driver(one);
2312 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2313 }
2314
2315 void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2316 {
2317 if (!options->a_prefix)
2318 options->a_prefix = a;
2319 if (!options->b_prefix)
2320 options->b_prefix = b;
2321 }
2322
2323 struct userdiff_driver *get_textconv(struct diff_filespec *one)
2324 {
2325 if (!DIFF_FILE_VALID(one))
2326 return NULL;
2327
2328 diff_filespec_load_driver(one);
2329 return userdiff_get_textconv(one->driver);
2330 }
2331
2332 static void builtin_diff(const char *name_a,
2333 const char *name_b,
2334 struct diff_filespec *one,
2335 struct diff_filespec *two,
2336 const char *xfrm_msg,
2337 int must_show_header,
2338 struct diff_options *o,
2339 int complete_rewrite)
2340 {
2341 mmfile_t mf1, mf2;
2342 const char *lbl[2];
2343 char *a_one, *b_two;
2344 const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
2345 const char *reset = diff_get_color_opt(o, DIFF_RESET);
2346 const char *a_prefix, *b_prefix;
2347 struct userdiff_driver *textconv_one = NULL;
2348 struct userdiff_driver *textconv_two = NULL;
2349 struct strbuf header = STRBUF_INIT;
2350 const char *line_prefix = diff_line_prefix(o);
2351
2352 diff_set_mnemonic_prefix(o, "a/", "b/");
2353 if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2354 a_prefix = o->b_prefix;
2355 b_prefix = o->a_prefix;
2356 } else {
2357 a_prefix = o->a_prefix;
2358 b_prefix = o->b_prefix;
2359 }
2360
2361 if (o->submodule_format == DIFF_SUBMODULE_LOG &&
2362 (!one->mode || S_ISGITLINK(one->mode)) &&
2363 (!two->mode || S_ISGITLINK(two->mode))) {
2364 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2365 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2366 show_submodule_summary(o->file, one->path ? one->path : two->path,
2367 line_prefix,
2368 &one->oid, &two->oid,
2369 two->dirty_submodule,
2370 meta, del, add, reset);
2371 return;
2372 } else if (o->submodule_format == DIFF_SUBMODULE_INLINE_DIFF &&
2373 (!one->mode || S_ISGITLINK(one->mode)) &&
2374 (!two->mode || S_ISGITLINK(two->mode))) {
2375 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2376 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2377 show_submodule_inline_diff(o->file, one->path ? one->path : two->path,
2378 line_prefix,
2379 &one->oid, &two->oid,
2380 two->dirty_submodule,
2381 meta, del, add, reset, o);
2382 return;
2383 }
2384
2385 if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2386 textconv_one = get_textconv(one);
2387 textconv_two = get_textconv(two);
2388 }
2389
2390 /* Never use a non-valid filename anywhere if at all possible */
2391 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2392 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2393
2394 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2395 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2396 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2397 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2398 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
2399 if (lbl[0][0] == '/') {
2400 /* /dev/null */
2401 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
2402 if (xfrm_msg)
2403 strbuf_addstr(&header, xfrm_msg);
2404 must_show_header = 1;
2405 }
2406 else if (lbl[1][0] == '/') {
2407 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
2408 if (xfrm_msg)
2409 strbuf_addstr(&header, xfrm_msg);
2410 must_show_header = 1;
2411 }
2412 else {
2413 if (one->mode != two->mode) {
2414 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
2415 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
2416 must_show_header = 1;
2417 }
2418 if (xfrm_msg)
2419 strbuf_addstr(&header, xfrm_msg);
2420
2421 /*
2422 * we do not run diff between different kind
2423 * of objects.
2424 */
2425 if ((one->mode ^ two->mode) & S_IFMT)
2426 goto free_ab_and_return;
2427 if (complete_rewrite &&
2428 (textconv_one || !diff_filespec_is_binary(one)) &&
2429 (textconv_two || !diff_filespec_is_binary(two))) {
2430 fprintf(o->file, "%s", header.buf);
2431 strbuf_reset(&header);
2432 emit_rewrite_diff(name_a, name_b, one, two,
2433 textconv_one, textconv_two, o);
2434 o->found_changes = 1;
2435 goto free_ab_and_return;
2436 }
2437 }
2438
2439 if (o->irreversible_delete && lbl[1][0] == '/') {
2440 fprintf(o->file, "%s", header.buf);
2441 strbuf_reset(&header);
2442 goto free_ab_and_return;
2443 } else if (!DIFF_OPT_TST(o, TEXT) &&
2444 ( (!textconv_one && diff_filespec_is_binary(one)) ||
2445 (!textconv_two && diff_filespec_is_binary(two)) )) {
2446 if (!one->data && !two->data &&
2447 S_ISREG(one->mode) && S_ISREG(two->mode) &&
2448 !DIFF_OPT_TST(o, BINARY)) {
2449 if (!oidcmp(&one->oid, &two->oid)) {
2450 if (must_show_header)
2451 fprintf(o->file, "%s", header.buf);
2452 goto free_ab_and_return;
2453 }
2454 fprintf(o->file, "%s", header.buf);
2455 fprintf(o->file, "%sBinary files %s and %s differ\n",
2456 line_prefix, lbl[0], lbl[1]);
2457 goto free_ab_and_return;
2458 }
2459 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2460 die("unable to read files to diff");
2461 /* Quite common confusing case */
2462 if (mf1.size == mf2.size &&
2463 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2464 if (must_show_header)
2465 fprintf(o->file, "%s", header.buf);
2466 goto free_ab_and_return;
2467 }
2468 fprintf(o->file, "%s", header.buf);
2469 strbuf_reset(&header);
2470 if (DIFF_OPT_TST(o, BINARY))
2471 emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2472 else
2473 fprintf(o->file, "%sBinary files %s and %s differ\n",
2474 line_prefix, lbl[0], lbl[1]);
2475 o->found_changes = 1;
2476 } else {
2477 /* Crazy xdl interfaces.. */
2478 const char *diffopts = getenv("GIT_DIFF_OPTS");
2479 const char *v;
2480 xpparam_t xpp;
2481 xdemitconf_t xecfg;
2482 struct emit_callback ecbdata;
2483 const struct userdiff_funcname *pe;
2484
2485 if (must_show_header) {
2486 fprintf(o->file, "%s", header.buf);
2487 strbuf_reset(&header);
2488 }
2489
2490 mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2491 mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2492
2493 pe = diff_funcname_pattern(one);
2494 if (!pe)
2495 pe = diff_funcname_pattern(two);
2496
2497 memset(&xpp, 0, sizeof(xpp));
2498 memset(&xecfg, 0, sizeof(xecfg));
2499 memset(&ecbdata, 0, sizeof(ecbdata));
2500 ecbdata.label_path = lbl;
2501 ecbdata.color_diff = want_color(o->use_color);
2502 ecbdata.ws_rule = whitespace_rule(name_b);
2503 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2504 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2505 ecbdata.opt = o;
2506 ecbdata.header = header.len ? &header : NULL;
2507 xpp.flags = o->xdl_opts;
2508 xecfg.ctxlen = o->context;
2509 xecfg.interhunkctxlen = o->interhunkcontext;
2510 xecfg.flags = XDL_EMIT_FUNCNAMES;
2511 if (DIFF_OPT_TST(o, FUNCCONTEXT))
2512 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2513 if (pe)
2514 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2515 if (!diffopts)
2516 ;
2517 else if (skip_prefix(diffopts, "--unified=", &v))
2518 xecfg.ctxlen = strtoul(v, NULL, 10);
2519 else if (skip_prefix(diffopts, "-u", &v))
2520 xecfg.ctxlen = strtoul(v, NULL, 10);
2521 if (o->word_diff)
2522 init_diff_words_data(&ecbdata, o, one, two);
2523 if (xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2524 &xpp, &xecfg))
2525 die("unable to generate diff for %s", one->path);
2526 if (o->word_diff)
2527 free_diff_words_data(&ecbdata);
2528 if (textconv_one)
2529 free(mf1.ptr);
2530 if (textconv_two)
2531 free(mf2.ptr);
2532 xdiff_clear_find_func(&xecfg);
2533 }
2534
2535 free_ab_and_return:
2536 strbuf_release(&header);
2537 diff_free_filespec_data(one);
2538 diff_free_filespec_data(two);
2539 free(a_one);
2540 free(b_two);
2541 return;
2542 }
2543
2544 static void builtin_diffstat(const char *name_a, const char *name_b,
2545 struct diff_filespec *one,
2546 struct diff_filespec *two,
2547 struct diffstat_t *diffstat,
2548 struct diff_options *o,
2549 struct diff_filepair *p)
2550 {
2551 mmfile_t mf1, mf2;
2552 struct diffstat_file *data;
2553 int same_contents;
2554 int complete_rewrite = 0;
2555
2556 if (!DIFF_PAIR_UNMERGED(p)) {
2557 if (p->status == DIFF_STATUS_MODIFIED && p->score)
2558 complete_rewrite = 1;
2559 }
2560
2561 data = diffstat_add(diffstat, name_a, name_b);
2562 data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
2563
2564 if (!one || !two) {
2565 data->is_unmerged = 1;
2566 return;
2567 }
2568
2569 same_contents = !oidcmp(&one->oid, &two->oid);
2570
2571 if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2572 data->is_binary = 1;
2573 if (same_contents) {
2574 data->added = 0;
2575 data->deleted = 0;
2576 } else {
2577 data->added = diff_filespec_size(two);
2578 data->deleted = diff_filespec_size(one);
2579 }
2580 }
2581
2582 else if (complete_rewrite) {
2583 diff_populate_filespec(one, 0);
2584 diff_populate_filespec(two, 0);
2585 data->deleted = count_lines(one->data, one->size);
2586 data->added = count_lines(two->data, two->size);
2587 }
2588
2589 else if (!same_contents) {
2590 /* Crazy xdl interfaces.. */
2591 xpparam_t xpp;
2592 xdemitconf_t xecfg;
2593
2594 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2595 die("unable to read files to diff");
2596
2597 memset(&xpp, 0, sizeof(xpp));
2598 memset(&xecfg, 0, sizeof(xecfg));
2599 xpp.flags = o->xdl_opts;
2600 xecfg.ctxlen = o->context;
2601 xecfg.interhunkctxlen = o->interhunkcontext;
2602 if (xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2603 &xpp, &xecfg))
2604 die("unable to generate diffstat for %s", one->path);
2605 }
2606
2607 diff_free_filespec_data(one);
2608 diff_free_filespec_data(two);
2609 }
2610
2611 static void builtin_checkdiff(const char *name_a, const char *name_b,
2612 const char *attr_path,
2613 struct diff_filespec *one,
2614 struct diff_filespec *two,
2615 struct diff_options *o)
2616 {
2617 mmfile_t mf1, mf2;
2618 struct checkdiff_t data;
2619
2620 if (!two)
2621 return;
2622
2623 memset(&data, 0, sizeof(data));
2624 data.filename = name_b ? name_b : name_a;
2625 data.lineno = 0;
2626 data.o = o;
2627 data.ws_rule = whitespace_rule(attr_path);
2628 data.conflict_marker_size = ll_merge_marker_size(attr_path);
2629
2630 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2631 die("unable to read files to diff");
2632
2633 /*
2634 * All the other codepaths check both sides, but not checking
2635 * the "old" side here is deliberate. We are checking the newly
2636 * introduced changes, and as long as the "new" side is text, we
2637 * can and should check what it introduces.
2638 */
2639 if (diff_filespec_is_binary(two))
2640 goto free_and_return;
2641 else {
2642 /* Crazy xdl interfaces.. */
2643 xpparam_t xpp;
2644 xdemitconf_t xecfg;
2645
2646 memset(&xpp, 0, sizeof(xpp));
2647 memset(&xecfg, 0, sizeof(xecfg));
2648 xecfg.ctxlen = 1; /* at least one context line */
2649 xpp.flags = 0;
2650 if (xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2651 &xpp, &xecfg))
2652 die("unable to generate checkdiff for %s", one->path);
2653
2654 if (data.ws_rule & WS_BLANK_AT_EOF) {
2655 struct emit_callback ecbdata;
2656 int blank_at_eof;
2657
2658 ecbdata.ws_rule = data.ws_rule;
2659 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2660 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2661
2662 if (blank_at_eof) {
2663 static char *err;
2664 if (!err)
2665 err = whitespace_error_string(WS_BLANK_AT_EOF);
2666 fprintf(o->file, "%s:%d: %s.\n",
2667 data.filename, blank_at_eof, err);
2668 data.status = 1; /* report errors */
2669 }
2670 }
2671 }
2672 free_and_return:
2673 diff_free_filespec_data(one);
2674 diff_free_filespec_data(two);
2675 if (data.status)
2676 DIFF_OPT_SET(o, CHECK_FAILED);
2677 }
2678
2679 struct diff_filespec *alloc_filespec(const char *path)
2680 {
2681 struct diff_filespec *spec;
2682
2683 FLEXPTR_ALLOC_STR(spec, path, path);
2684 spec->count = 1;
2685 spec->is_binary = -1;
2686 return spec;
2687 }
2688
2689 void free_filespec(struct diff_filespec *spec)
2690 {
2691 if (!--spec->count) {
2692 diff_free_filespec_data(spec);
2693 free(spec);
2694 }
2695 }
2696
2697 void fill_filespec(struct diff_filespec *spec, const struct object_id *oid,
2698 int oid_valid, unsigned short mode)
2699 {
2700 if (mode) {
2701 spec->mode = canon_mode(mode);
2702 oidcpy(&spec->oid, oid);
2703 spec->oid_valid = oid_valid;
2704 }
2705 }
2706
2707 /*
2708 * Given a name and sha1 pair, if the index tells us the file in
2709 * the work tree has that object contents, return true, so that
2710 * prepare_temp_file() does not have to inflate and extract.
2711 */
2712 static int reuse_worktree_file(const char *name, const struct object_id *oid, int want_file)
2713 {
2714 const struct cache_entry *ce;
2715 struct stat st;
2716 int pos, len;
2717
2718 /*
2719 * We do not read the cache ourselves here, because the
2720 * benchmark with my previous version that always reads cache
2721 * shows that it makes things worse for diff-tree comparing
2722 * two linux-2.6 kernel trees in an already checked out work
2723 * tree. This is because most diff-tree comparisons deal with
2724 * only a small number of files, while reading the cache is
2725 * expensive for a large project, and its cost outweighs the
2726 * savings we get by not inflating the object to a temporary
2727 * file. Practically, this code only helps when we are used
2728 * by diff-cache --cached, which does read the cache before
2729 * calling us.
2730 */
2731 if (!active_cache)
2732 return 0;
2733
2734 /* We want to avoid the working directory if our caller
2735 * doesn't need the data in a normal file, this system
2736 * is rather slow with its stat/open/mmap/close syscalls,
2737 * and the object is contained in a pack file. The pack
2738 * is probably already open and will be faster to obtain
2739 * the data through than the working directory. Loose
2740 * objects however would tend to be slower as they need
2741 * to be individually opened and inflated.
2742 */
2743 if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(oid->hash))
2744 return 0;
2745
2746 /*
2747 * Similarly, if we'd have to convert the file contents anyway, that
2748 * makes the optimization not worthwhile.
2749 */
2750 if (!want_file && would_convert_to_git(&the_index, name))
2751 return 0;
2752
2753 len = strlen(name);
2754 pos = cache_name_pos(name, len);
2755 if (pos < 0)
2756 return 0;
2757 ce = active_cache[pos];
2758
2759 /*
2760 * This is not the sha1 we are looking for, or
2761 * unreusable because it is not a regular file.
2762 */
2763 if (oidcmp(oid, &ce->oid) || !S_ISREG(ce->ce_mode))
2764 return 0;
2765
2766 /*
2767 * If ce is marked as "assume unchanged", there is no
2768 * guarantee that work tree matches what we are looking for.
2769 */
2770 if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2771 return 0;
2772
2773 /*
2774 * If ce matches the file in the work tree, we can reuse it.
2775 */
2776 if (ce_uptodate(ce) ||
2777 (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2778 return 1;
2779
2780 return 0;
2781 }
2782
2783 static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2784 {
2785 struct strbuf buf = STRBUF_INIT;
2786 char *dirty = "";
2787
2788 /* Are we looking at the work tree? */
2789 if (s->dirty_submodule)
2790 dirty = "-dirty";
2791
2792 strbuf_addf(&buf, "Subproject commit %s%s\n",
2793 oid_to_hex(&s->oid), dirty);
2794 s->size = buf.len;
2795 if (size_only) {
2796 s->data = NULL;
2797 strbuf_release(&buf);
2798 } else {
2799 s->data = strbuf_detach(&buf, NULL);
2800 s->should_free = 1;
2801 }
2802 return 0;
2803 }
2804
2805 /*
2806 * While doing rename detection and pickaxe operation, we may need to
2807 * grab the data for the blob (or file) for our own in-core comparison.
2808 * diff_filespec has data and size fields for this purpose.
2809 */
2810 int diff_populate_filespec(struct diff_filespec *s, unsigned int flags)
2811 {
2812 int size_only = flags & CHECK_SIZE_ONLY;
2813 int err = 0;
2814 /*
2815 * demote FAIL to WARN to allow inspecting the situation
2816 * instead of refusing.
2817 */
2818 enum safe_crlf crlf_warn = (safe_crlf == SAFE_CRLF_FAIL
2819 ? SAFE_CRLF_WARN
2820 : safe_crlf);
2821
2822 if (!DIFF_FILE_VALID(s))
2823 die("internal error: asking to populate invalid file.");
2824 if (S_ISDIR(s->mode))
2825 return -1;
2826
2827 if (s->data)
2828 return 0;
2829
2830 if (size_only && 0 < s->size)
2831 return 0;
2832
2833 if (S_ISGITLINK(s->mode))
2834 return diff_populate_gitlink(s, size_only);
2835
2836 if (!s->oid_valid ||
2837 reuse_worktree_file(s->path, &s->oid, 0)) {
2838 struct strbuf buf = STRBUF_INIT;
2839 struct stat st;
2840 int fd;
2841
2842 if (lstat(s->path, &st) < 0) {
2843 if (errno == ENOENT) {
2844 err_empty:
2845 err = -1;
2846 empty:
2847 s->data = (char *)"";
2848 s->size = 0;
2849 return err;
2850 }
2851 }
2852 s->size = xsize_t(st.st_size);
2853 if (!s->size)
2854 goto empty;
2855 if (S_ISLNK(st.st_mode)) {
2856 struct strbuf sb = STRBUF_INIT;
2857
2858 if (strbuf_readlink(&sb, s->path, s->size))
2859 goto err_empty;
2860 s->size = sb.len;
2861 s->data = strbuf_detach(&sb, NULL);
2862 s->should_free = 1;
2863 return 0;
2864 }
2865
2866 /*
2867 * Even if the caller would be happy with getting
2868 * only the size, we cannot return early at this
2869 * point if the path requires us to run the content
2870 * conversion.
2871 */
2872 if (size_only && !would_convert_to_git(&the_index, s->path))
2873 return 0;
2874
2875 /*
2876 * Note: this check uses xsize_t(st.st_size) that may
2877 * not be the true size of the blob after it goes
2878 * through convert_to_git(). This may not strictly be
2879 * correct, but the whole point of big_file_threshold
2880 * and is_binary check being that we want to avoid
2881 * opening the file and inspecting the contents, this
2882 * is probably fine.
2883 */
2884 if ((flags & CHECK_BINARY) &&
2885 s->size > big_file_threshold && s->is_binary == -1) {
2886 s->is_binary = 1;
2887 return 0;
2888 }
2889 fd = open(s->path, O_RDONLY);
2890 if (fd < 0)
2891 goto err_empty;
2892 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2893 close(fd);
2894 s->should_munmap = 1;
2895
2896 /*
2897 * Convert from working tree format to canonical git format
2898 */
2899 if (convert_to_git(&the_index, s->path, s->data, s->size, &buf, crlf_warn)) {
2900 size_t size = 0;
2901 munmap(s->data, s->size);
2902 s->should_munmap = 0;
2903 s->data = strbuf_detach(&buf, &size);
2904 s->size = size;
2905 s->should_free = 1;
2906 }
2907 }
2908 else {
2909 enum object_type type;
2910 if (size_only || (flags & CHECK_BINARY)) {
2911 type = sha1_object_info(s->oid.hash, &s->size);
2912 if (type < 0)
2913 die("unable to read %s",
2914 oid_to_hex(&s->oid));
2915 if (size_only)
2916 return 0;
2917 if (s->size > big_file_threshold && s->is_binary == -1) {
2918 s->is_binary = 1;
2919 return 0;
2920 }
2921 }
2922 s->data = read_sha1_file(s->oid.hash, &type, &s->size);
2923 if (!s->data)
2924 die("unable to read %s", oid_to_hex(&s->oid));
2925 s->should_free = 1;
2926 }
2927 return 0;
2928 }
2929
2930 void diff_free_filespec_blob(struct diff_filespec *s)
2931 {
2932 if (s->should_free)
2933 free(s->data);
2934 else if (s->should_munmap)
2935 munmap(s->data, s->size);
2936
2937 if (s->should_free || s->should_munmap) {
2938 s->should_free = s->should_munmap = 0;
2939 s->data = NULL;
2940 }
2941 }
2942
2943 void diff_free_filespec_data(struct diff_filespec *s)
2944 {
2945 diff_free_filespec_blob(s);
2946 FREE_AND_NULL(s->cnt_data);
2947 }
2948
2949 static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2950 void *blob,
2951 unsigned long size,
2952 const struct object_id *oid,
2953 int mode)
2954 {
2955 int fd;
2956 struct strbuf buf = STRBUF_INIT;
2957 struct strbuf template = STRBUF_INIT;
2958 char *path_dup = xstrdup(path);
2959 const char *base = basename(path_dup);
2960
2961 /* Generate "XXXXXX_basename.ext" */
2962 strbuf_addstr(&template, "XXXXXX_");
2963 strbuf_addstr(&template, base);
2964
2965 fd = mks_tempfile_ts(&temp->tempfile, template.buf, strlen(base) + 1);
2966 if (fd < 0)
2967 die_errno("unable to create temp-file");
2968 if (convert_to_working_tree(path,
2969 (const char *)blob, (size_t)size, &buf)) {
2970 blob = buf.buf;
2971 size = buf.len;
2972 }
2973 if (write_in_full(fd, blob, size) != size)
2974 die_errno("unable to write temp-file");
2975 close_tempfile(&temp->tempfile);
2976 temp->name = get_tempfile_path(&temp->tempfile);
2977 oid_to_hex_r(temp->hex, oid);
2978 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", mode);
2979 strbuf_release(&buf);
2980 strbuf_release(&template);
2981 free(path_dup);
2982 }
2983
2984 static struct diff_tempfile *prepare_temp_file(const char *name,
2985 struct diff_filespec *one)
2986 {
2987 struct diff_tempfile *temp = claim_diff_tempfile();
2988
2989 if (!DIFF_FILE_VALID(one)) {
2990 not_a_valid_file:
2991 /* A '-' entry produces this for file-2, and
2992 * a '+' entry produces this for file-1.
2993 */
2994 temp->name = "/dev/null";
2995 xsnprintf(temp->hex, sizeof(temp->hex), ".");
2996 xsnprintf(temp->mode, sizeof(temp->mode), ".");
2997 return temp;
2998 }
2999
3000 if (!S_ISGITLINK(one->mode) &&
3001 (!one->oid_valid ||
3002 reuse_worktree_file(name, &one->oid, 1))) {
3003 struct stat st;
3004 if (lstat(name, &st) < 0) {
3005 if (errno == ENOENT)
3006 goto not_a_valid_file;
3007 die_errno("stat(%s)", name);
3008 }
3009 if (S_ISLNK(st.st_mode)) {
3010 struct strbuf sb = STRBUF_INIT;
3011 if (strbuf_readlink(&sb, name, st.st_size) < 0)
3012 die_errno("readlink(%s)", name);
3013 prep_temp_blob(name, temp, sb.buf, sb.len,
3014 (one->oid_valid ?
3015 &one->oid : &null_oid),
3016 (one->oid_valid ?
3017 one->mode : S_IFLNK));
3018 strbuf_release(&sb);
3019 }
3020 else {
3021 /* we can borrow from the file in the work tree */
3022 temp->name = name;
3023 if (!one->oid_valid)
3024 oid_to_hex_r(temp->hex, &null_oid);
3025 else
3026 oid_to_hex_r(temp->hex, &one->oid);
3027 /* Even though we may sometimes borrow the
3028 * contents from the work tree, we always want
3029 * one->mode. mode is trustworthy even when
3030 * !(one->oid_valid), as long as
3031 * DIFF_FILE_VALID(one).
3032 */
3033 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", one->mode);
3034 }
3035 return temp;
3036 }
3037 else {
3038 if (diff_populate_filespec(one, 0))
3039 die("cannot read data blob for %s", one->path);
3040 prep_temp_blob(name, temp, one->data, one->size,
3041 &one->oid, one->mode);
3042 }
3043 return temp;
3044 }
3045
3046 static void add_external_diff_name(struct argv_array *argv,
3047 const char *name,
3048 struct diff_filespec *df)
3049 {
3050 struct diff_tempfile *temp = prepare_temp_file(name, df);
3051 argv_array_push(argv, temp->name);
3052 argv_array_push(argv, temp->hex);
3053 argv_array_push(argv, temp->mode);
3054 }
3055
3056 /* An external diff command takes:
3057 *
3058 * diff-cmd name infile1 infile1-sha1 infile1-mode \
3059 * infile2 infile2-sha1 infile2-mode [ rename-to ]
3060 *
3061 */
3062 static void run_external_diff(const char *pgm,
3063 const char *name,
3064 const char *other,
3065 struct diff_filespec *one,
3066 struct diff_filespec *two,
3067 const char *xfrm_msg,
3068 int complete_rewrite,
3069 struct diff_options *o)
3070 {
3071 struct argv_array argv = ARGV_ARRAY_INIT;
3072 struct argv_array env = ARGV_ARRAY_INIT;
3073 struct diff_queue_struct *q = &diff_queued_diff;
3074
3075 argv_array_push(&argv, pgm);
3076 argv_array_push(&argv, name);
3077
3078 if (one && two) {
3079 add_external_diff_name(&argv, name, one);
3080 if (!other)
3081 add_external_diff_name(&argv, name, two);
3082 else {
3083 add_external_diff_name(&argv, other, two);
3084 argv_array_push(&argv, other);
3085 argv_array_push(&argv, xfrm_msg);
3086 }
3087 }
3088
3089 argv_array_pushf(&env, "GIT_DIFF_PATH_COUNTER=%d", ++o->diff_path_counter);
3090 argv_array_pushf(&env, "GIT_DIFF_PATH_TOTAL=%d", q->nr);
3091
3092 if (run_command_v_opt_cd_env(argv.argv, RUN_USING_SHELL, NULL, env.argv))
3093 die(_("external diff died, stopping at %s"), name);
3094
3095 remove_tempfile();
3096 argv_array_clear(&argv);
3097 argv_array_clear(&env);
3098 }
3099
3100 static int similarity_index(struct diff_filepair *p)
3101 {
3102 return p->score * 100 / MAX_SCORE;
3103 }
3104
3105 static const char *diff_abbrev_oid(const struct object_id *oid, int abbrev)
3106 {
3107 if (startup_info->have_repository)
3108 return find_unique_abbrev(oid->hash, abbrev);
3109 else {
3110 char *hex = oid_to_hex(oid);
3111 if (abbrev < 0)
3112 abbrev = FALLBACK_DEFAULT_ABBREV;
3113 if (abbrev > GIT_SHA1_HEXSZ)
3114 die("BUG: oid abbreviation out of range: %d", abbrev);
3115 if (abbrev)
3116 hex[abbrev] = '\0';
3117 return hex;
3118 }
3119 }
3120
3121 static void fill_metainfo(struct strbuf *msg,
3122 const char *name,
3123 const char *other,
3124 struct diff_filespec *one,
3125 struct diff_filespec *two,
3126 struct diff_options *o,
3127 struct diff_filepair *p,
3128 int *must_show_header,
3129 int use_color)
3130 {
3131 const char *set = diff_get_color(use_color, DIFF_METAINFO);
3132 const char *reset = diff_get_color(use_color, DIFF_RESET);
3133 const char *line_prefix = diff_line_prefix(o);
3134
3135 *must_show_header = 1;
3136 strbuf_init(msg, PATH_MAX * 2 + 300);
3137 switch (p->status) {
3138 case DIFF_STATUS_COPIED:
3139 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3140 line_prefix, set, similarity_index(p));
3141 strbuf_addf(msg, "%s\n%s%scopy from ",
3142 reset, line_prefix, set);
3143 quote_c_style(name, msg, NULL, 0);
3144 strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
3145 quote_c_style(other, msg, NULL, 0);
3146 strbuf_addf(msg, "%s\n", reset);
3147 break;
3148 case DIFF_STATUS_RENAMED:
3149 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3150 line_prefix, set, similarity_index(p));
3151 strbuf_addf(msg, "%s\n%s%srename from ",
3152 reset, line_prefix, set);
3153 quote_c_style(name, msg, NULL, 0);
3154 strbuf_addf(msg, "%s\n%s%srename to ",
3155 reset, line_prefix, set);
3156 quote_c_style(other, msg, NULL, 0);
3157 strbuf_addf(msg, "%s\n", reset);
3158 break;
3159 case DIFF_STATUS_MODIFIED:
3160 if (p->score) {
3161 strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
3162 line_prefix,
3163 set, similarity_index(p), reset);
3164 break;
3165 }
3166 /* fallthru */
3167 default:
3168 *must_show_header = 0;
3169 }
3170 if (one && two && oidcmp(&one->oid, &two->oid)) {
3171 int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
3172
3173 if (DIFF_OPT_TST(o, BINARY)) {
3174 mmfile_t mf;
3175 if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
3176 (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
3177 abbrev = 40;
3178 }
3179 strbuf_addf(msg, "%s%sindex %s..%s", line_prefix, set,
3180 diff_abbrev_oid(&one->oid, abbrev),
3181 diff_abbrev_oid(&two->oid, abbrev));
3182 if (one->mode == two->mode)
3183 strbuf_addf(msg, " %06o", one->mode);
3184 strbuf_addf(msg, "%s\n", reset);
3185 }
3186 }
3187
3188 static void run_diff_cmd(const char *pgm,
3189 const char *name,
3190 const char *other,
3191 const char *attr_path,
3192 struct diff_filespec *one,
3193 struct diff_filespec *two,
3194 struct strbuf *msg,
3195 struct diff_options *o,
3196 struct diff_filepair *p)
3197 {
3198 const char *xfrm_msg = NULL;
3199 int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
3200 int must_show_header = 0;
3201
3202
3203 if (DIFF_OPT_TST(o, ALLOW_EXTERNAL)) {
3204 struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
3205 if (drv && drv->external)
3206 pgm = drv->external;
3207 }
3208
3209 if (msg) {
3210 /*
3211 * don't use colors when the header is intended for an
3212 * external diff driver
3213 */
3214 fill_metainfo(msg, name, other, one, two, o, p,
3215 &must_show_header,
3216 want_color(o->use_color) && !pgm);
3217 xfrm_msg = msg->len ? msg->buf : NULL;
3218 }
3219
3220 if (pgm) {
3221 run_external_diff(pgm, name, other, one, two, xfrm_msg,
3222 complete_rewrite, o);
3223 return;
3224 }
3225 if (one && two)
3226 builtin_diff(name, other ? other : name,
3227 one, two, xfrm_msg, must_show_header,
3228 o, complete_rewrite);
3229 else
3230 fprintf(o->file, "* Unmerged path %s\n", name);
3231 }
3232
3233 static void diff_fill_oid_info(struct diff_filespec *one)
3234 {
3235 if (DIFF_FILE_VALID(one)) {
3236 if (!one->oid_valid) {
3237 struct stat st;
3238 if (one->is_stdin) {
3239 oidclr(&one->oid);
3240 return;
3241 }
3242 if (lstat(one->path, &st) < 0)
3243 die_errno("stat '%s'", one->path);
3244 if (index_path(one->oid.hash, one->path, &st, 0))
3245 die("cannot hash %s", one->path);
3246 }
3247 }
3248 else
3249 oidclr(&one->oid);
3250 }
3251
3252 static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
3253 {
3254 /* Strip the prefix but do not molest /dev/null and absolute paths */
3255 if (*namep && **namep != '/') {
3256 *namep += prefix_length;
3257 if (**namep == '/')
3258 ++*namep;
3259 }
3260 if (*otherp && **otherp != '/') {
3261 *otherp += prefix_length;
3262 if (**otherp == '/')
3263 ++*otherp;
3264 }
3265 }
3266
3267 static void run_diff(struct diff_filepair *p, struct diff_options *o)
3268 {
3269 const char *pgm = external_diff();
3270 struct strbuf msg;
3271 struct diff_filespec *one = p->one;
3272 struct diff_filespec *two = p->two;
3273 const char *name;
3274 const char *other;
3275 const char *attr_path;
3276
3277 name = p->one->path;
3278 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3279 attr_path = name;
3280 if (o->prefix_length)
3281 strip_prefix(o->prefix_length, &name, &other);
3282
3283 if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
3284 pgm = NULL;
3285
3286 if (DIFF_PAIR_UNMERGED(p)) {
3287 run_diff_cmd(pgm, name, NULL, attr_path,
3288 NULL, NULL, NULL, o, p);
3289 return;
3290 }
3291
3292 diff_fill_oid_info(one);
3293 diff_fill_oid_info(two);
3294
3295 if (!pgm &&
3296 DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3297 (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3298 /*
3299 * a filepair that changes between file and symlink
3300 * needs to be split into deletion and creation.
3301 */
3302 struct diff_filespec *null = alloc_filespec(two->path);
3303 run_diff_cmd(NULL, name, other, attr_path,
3304 one, null, &msg, o, p);
3305 free(null);
3306 strbuf_release(&msg);
3307
3308 null = alloc_filespec(one->path);
3309 run_diff_cmd(NULL, name, other, attr_path,
3310 null, two, &msg, o, p);
3311 free(null);
3312 }
3313 else
3314 run_diff_cmd(pgm, name, other, attr_path,
3315 one, two, &msg, o, p);
3316
3317 strbuf_release(&msg);
3318 }
3319
3320 static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3321 struct diffstat_t *diffstat)
3322 {
3323 const char *name;
3324 const char *other;
3325
3326 if (DIFF_PAIR_UNMERGED(p)) {
3327 /* unmerged */
3328 builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, p);
3329 return;
3330 }
3331
3332 name = p->one->path;
3333 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3334
3335 if (o->prefix_length)
3336 strip_prefix(o->prefix_length, &name, &other);
3337
3338 diff_fill_oid_info(p->one);
3339 diff_fill_oid_info(p->two);
3340
3341 builtin_diffstat(name, other, p->one, p->two, diffstat, o, p);
3342 }
3343
3344 static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3345 {
3346 const char *name;
3347 const char *other;
3348 const char *attr_path;
3349
3350 if (DIFF_PAIR_UNMERGED(p)) {
3351 /* unmerged */
3352 return;
3353 }
3354
3355 name = p->one->path;
3356 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3357 attr_path = other ? other : name;
3358
3359 if (o->prefix_length)
3360 strip_prefix(o->prefix_length, &name, &other);
3361
3362 diff_fill_oid_info(p->one);
3363 diff_fill_oid_info(p->two);
3364
3365 builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3366 }
3367
3368 void diff_setup(struct diff_options *options)
3369 {
3370 memcpy(options, &default_diff_options, sizeof(*options));
3371
3372 options->file = stdout;
3373
3374 options->abbrev = DEFAULT_ABBREV;
3375 options->line_termination = '\n';
3376 options->break_opt = -1;
3377 options->rename_limit = -1;
3378 options->dirstat_permille = diff_dirstat_permille_default;
3379 options->context = diff_context_default;
3380 options->interhunkcontext = diff_interhunk_context_default;
3381 options->ws_error_highlight = ws_error_highlight_default;
3382 DIFF_OPT_SET(options, RENAME_EMPTY);
3383
3384 /* pathchange left =NULL by default */
3385 options->change = diff_change;
3386 options->add_remove = diff_addremove;
3387 options->use_color = diff_use_color_default;
3388 options->detect_rename = diff_detect_rename_default;
3389 options->xdl_opts |= diff_algorithm;
3390 if (diff_indent_heuristic)
3391 DIFF_XDL_SET(options, INDENT_HEURISTIC);
3392
3393 options->orderfile = diff_order_file_cfg;
3394
3395 if (diff_no_prefix) {
3396 options->a_prefix = options->b_prefix = "";
3397 } else if (!diff_mnemonic_prefix) {
3398 options->a_prefix = "a/";
3399 options->b_prefix = "b/";
3400 }
3401 }
3402
3403 void diff_setup_done(struct diff_options *options)
3404 {
3405 int count = 0;
3406
3407 if (options->set_default)
3408 options->set_default(options);
3409
3410 if (options->output_format & DIFF_FORMAT_NAME)
3411 count++;
3412 if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3413 count++;
3414 if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3415 count++;
3416 if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3417 count++;
3418 if (count > 1)
3419 die(_("--name-only, --name-status, --check and -s are mutually exclusive"));
3420
3421 /*
3422 * Most of the time we can say "there are changes"
3423 * only by checking if there are changed paths, but
3424 * --ignore-whitespace* options force us to look
3425 * inside contents.
3426 */
3427
3428 if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3429 DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3430 DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3431 DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3432 else
3433 DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3434
3435 if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3436 options->detect_rename = DIFF_DETECT_COPY;
3437
3438 if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3439 options->prefix = NULL;
3440 if (options->prefix)
3441 options->prefix_length = strlen(options->prefix);
3442 else
3443 options->prefix_length = 0;
3444
3445 if (options->output_format & (DIFF_FORMAT_NAME |
3446 DIFF_FORMAT_NAME_STATUS |
3447 DIFF_FORMAT_CHECKDIFF |
3448 DIFF_FORMAT_NO_OUTPUT))
3449 options->output_format &= ~(DIFF_FORMAT_RAW |
3450 DIFF_FORMAT_NUMSTAT |
3451 DIFF_FORMAT_DIFFSTAT |
3452 DIFF_FORMAT_SHORTSTAT |
3453 DIFF_FORMAT_DIRSTAT |
3454 DIFF_FORMAT_SUMMARY |
3455 DIFF_FORMAT_PATCH);
3456
3457 /*
3458 * These cases always need recursive; we do not drop caller-supplied
3459 * recursive bits for other formats here.
3460 */
3461 if (options->output_format & (DIFF_FORMAT_PATCH |
3462 DIFF_FORMAT_NUMSTAT |
3463 DIFF_FORMAT_DIFFSTAT |
3464 DIFF_FORMAT_SHORTSTAT |
3465 DIFF_FORMAT_DIRSTAT |
3466 DIFF_FORMAT_SUMMARY |
3467 DIFF_FORMAT_CHECKDIFF))
3468 DIFF_OPT_SET(options, RECURSIVE);
3469 /*
3470 * Also pickaxe would not work very well if you do not say recursive
3471 */
3472 if (options->pickaxe)
3473 DIFF_OPT_SET(options, RECURSIVE);
3474 /*
3475 * When patches are generated, submodules diffed against the work tree
3476 * must be checked for dirtiness too so it can be shown in the output
3477 */
3478 if (options->output_format & DIFF_FORMAT_PATCH)
3479 DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3480
3481 if (options->detect_rename && options->rename_limit < 0)
3482 options->rename_limit = diff_rename_limit_default;
3483 if (options->setup & DIFF_SETUP_USE_CACHE) {
3484 if (!active_cache)
3485 /* read-cache does not die even when it fails
3486 * so it is safe for us to do this here. Also
3487 * it does not smudge active_cache or active_nr
3488 * when it fails, so we do not have to worry about
3489 * cleaning it up ourselves either.
3490 */
3491 read_cache();
3492 }
3493 if (40 < options->abbrev)
3494 options->abbrev = 40; /* full */
3495
3496 /*
3497 * It does not make sense to show the first hit we happened
3498 * to have found. It does not make sense not to return with
3499 * exit code in such a case either.
3500 */
3501 if (DIFF_OPT_TST(options, QUICK)) {
3502 options->output_format = DIFF_FORMAT_NO_OUTPUT;
3503 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3504 }
3505
3506 options->diff_path_counter = 0;
3507
3508 if (DIFF_OPT_TST(options, FOLLOW_RENAMES) && options->pathspec.nr != 1)
3509 die(_("--follow requires exactly one pathspec"));
3510 }
3511
3512 static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3513 {
3514 char c, *eq;
3515 int len;
3516
3517 if (*arg != '-')
3518 return 0;
3519 c = *++arg;
3520 if (!c)
3521 return 0;
3522 if (c == arg_short) {
3523 c = *++arg;
3524 if (!c)
3525 return 1;
3526 if (val && isdigit(c)) {
3527 char *end;
3528 int n = strtoul(arg, &end, 10);
3529 if (*end)
3530 return 0;
3531 *val = n;
3532 return 1;
3533 }
3534 return 0;
3535 }
3536 if (c != '-')
3537 return 0;
3538 arg++;
3539 eq = strchrnul(arg, '=');
3540 len = eq - arg;
3541 if (!len || strncmp(arg, arg_long, len))
3542 return 0;
3543 if (*eq) {
3544 int n;
3545 char *end;
3546 if (!isdigit(*++eq))
3547 return 0;
3548 n = strtoul(eq, &end, 10);
3549 if (*end)
3550 return 0;
3551 *val = n;
3552 }
3553 return 1;
3554 }
3555
3556 static int diff_scoreopt_parse(const char *opt);
3557
3558 static inline int short_opt(char opt, const char **argv,
3559 const char **optarg)
3560 {
3561 const char *arg = argv[0];
3562 if (arg[0] != '-' || arg[1] != opt)
3563 return 0;
3564 if (arg[2] != '\0') {
3565 *optarg = arg + 2;
3566 return 1;
3567 }
3568 if (!argv[1])
3569 die("Option '%c' requires a value", opt);
3570 *optarg = argv[1];
3571 return 2;
3572 }
3573
3574 int parse_long_opt(const char *opt, const char **argv,
3575 const char **optarg)
3576 {
3577 const char *arg = argv[0];
3578 if (!skip_prefix(arg, "--", &arg))
3579 return 0;
3580 if (!skip_prefix(arg, opt, &arg))
3581 return 0;
3582 if (*arg == '=') { /* stuck form: --option=value */
3583 *optarg = arg + 1;
3584 return 1;
3585 }
3586 if (*arg != '\0')
3587 return 0;
3588 /* separate form: --option value */
3589 if (!argv[1])
3590 die("Option '--%s' requires a value", opt);
3591 *optarg = argv[1];
3592 return 2;
3593 }
3594
3595 static int stat_opt(struct diff_options *options, const char **av)
3596 {
3597 const char *arg = av[0];
3598 char *end;
3599 int width = options->stat_width;
3600 int name_width = options->stat_name_width;
3601 int graph_width = options->stat_graph_width;
3602 int count = options->stat_count;
3603 int argcount = 1;
3604
3605 if (!skip_prefix(arg, "--stat", &arg))
3606 die("BUG: stat option does not begin with --stat: %s", arg);
3607 end = (char *)arg;
3608
3609 switch (*arg) {
3610 case '-':
3611 if (skip_prefix(arg, "-width", &arg)) {
3612 if (*arg == '=')
3613 width = strtoul(arg + 1, &end, 10);
3614 else if (!*arg && !av[1])
3615 die_want_option("--stat-width");
3616 else if (!*arg) {
3617 width = strtoul(av[1], &end, 10);
3618 argcount = 2;
3619 }
3620 } else if (skip_prefix(arg, "-name-width", &arg)) {
3621 if (*arg == '=')
3622 name_width = strtoul(arg + 1, &end, 10);
3623 else if (!*arg && !av[1])
3624 die_want_option("--stat-name-width");
3625 else if (!*arg) {
3626 name_width = strtoul(av[1], &end, 10);
3627 argcount = 2;
3628 }
3629 } else if (skip_prefix(arg, "-graph-width", &arg)) {
3630 if (*arg == '=')
3631 graph_width = strtoul(arg + 1, &end, 10);
3632 else if (!*arg && !av[1])
3633 die_want_option("--stat-graph-width");
3634 else if (!*arg) {
3635 graph_width = strtoul(av[1], &end, 10);
3636 argcount = 2;
3637 }
3638 } else if (skip_prefix(arg, "-count", &arg)) {
3639 if (*arg == '=')
3640 count = strtoul(arg + 1, &end, 10);
3641 else if (!*arg && !av[1])
3642 die_want_option("--stat-count");
3643 else if (!*arg) {
3644 count = strtoul(av[1], &end, 10);
3645 argcount = 2;
3646 }
3647 }
3648 break;
3649 case '=':
3650 width = strtoul(arg+1, &end, 10);
3651 if (*end == ',')
3652 name_width = strtoul(end+1, &end, 10);
3653 if (*end == ',')
3654 count = strtoul(end+1, &end, 10);
3655 }
3656
3657 /* Important! This checks all the error cases! */
3658 if (*end)
3659 return 0;
3660 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3661 options->stat_name_width = name_width;
3662 options->stat_graph_width = graph_width;
3663 options->stat_width = width;
3664 options->stat_count = count;
3665 return argcount;
3666 }
3667
3668 static int parse_dirstat_opt(struct diff_options *options, const char *params)
3669 {
3670 struct strbuf errmsg = STRBUF_INIT;
3671 if (parse_dirstat_params(options, params, &errmsg))
3672 die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3673 errmsg.buf);
3674 strbuf_release(&errmsg);
3675 /*
3676 * The caller knows a dirstat-related option is given from the command
3677 * line; allow it to say "return this_function();"
3678 */
3679 options->output_format |= DIFF_FORMAT_DIRSTAT;
3680 return 1;
3681 }
3682
3683 static int parse_submodule_opt(struct diff_options *options, const char *value)
3684 {
3685 if (parse_submodule_params(options, value))
3686 die(_("Failed to parse --submodule option parameter: '%s'"),
3687 value);
3688 return 1;
3689 }
3690
3691 static const char diff_status_letters[] = {
3692 DIFF_STATUS_ADDED,
3693 DIFF_STATUS_COPIED,
3694 DIFF_STATUS_DELETED,
3695 DIFF_STATUS_MODIFIED,
3696 DIFF_STATUS_RENAMED,