?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
path: convert do_git_path to take a 'struct repository'
[git/debian.git] / diff.c
blob41295d4ea9ff5564fa73d45aaa3afaab20103d87
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 if (git_color_config(var, value, cb) < 0)
303 return -1;
304
305 return git_diff_basic_config(var, value, cb);
306 }
307
308 int git_diff_basic_config(const char *var, const char *value, void *cb)
309 {
310 const char *name;
311
312 if (!strcmp(var, "diff.renamelimit")) {
313 diff_rename_limit_default = git_config_int(var, value);
314 return 0;
315 }
316
317 if (userdiff_config(var, value) < 0)
318 return -1;
319
320 if (skip_prefix(var, "diff.color.", &name) ||
321 skip_prefix(var, "color.diff.", &name)) {
322 int slot = parse_diff_color_slot(name);
323 if (slot < 0)
324 return 0;
325 if (!value)
326 return config_error_nonbool(var);
327 return color_parse(value, diff_colors[slot]);
328 }
329
330 /* like GNU diff's --suppress-blank-empty option */
331 if (!strcmp(var, "diff.suppressblankempty") ||
332 /* for backwards compatibility */
333 !strcmp(var, "diff.suppress-blank-empty")) {
334 diff_suppress_blank_empty = git_config_bool(var, value);
335 return 0;
336 }
337
338 if (!strcmp(var, "diff.dirstat")) {
339 struct strbuf errmsg = STRBUF_INIT;
340 default_diff_options.dirstat_permille = diff_dirstat_permille_default;
341 if (parse_dirstat_params(&default_diff_options, value, &errmsg))
342 warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
343 errmsg.buf);
344 strbuf_release(&errmsg);
345 diff_dirstat_permille_default = default_diff_options.dirstat_permille;
346 return 0;
347 }
348
349 if (starts_with(var, "submodule."))
350 return parse_submodule_config_option(var, value);
351
352 if (git_diff_heuristic_config(var, value, cb) < 0)
353 return -1;
354
355 return git_default_config(var, value, cb);
356 }
357
358 static char *quote_two(const char *one, const char *two)
359 {
360 int need_one = quote_c_style(one, NULL, NULL, 1);
361 int need_two = quote_c_style(two, NULL, NULL, 1);
362 struct strbuf res = STRBUF_INIT;
363
364 if (need_one + need_two) {
365 strbuf_addch(&res, '"');
366 quote_c_style(one, &res, NULL, 1);
367 quote_c_style(two, &res, NULL, 1);
368 strbuf_addch(&res, '"');
369 } else {
370 strbuf_addstr(&res, one);
371 strbuf_addstr(&res, two);
372 }
373 return strbuf_detach(&res, NULL);
374 }
375
376 static const char *external_diff(void)
377 {
378 static const char *external_diff_cmd = NULL;
379 static int done_preparing = 0;
380
381 if (done_preparing)
382 return external_diff_cmd;
383 external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
384 if (!external_diff_cmd)
385 external_diff_cmd = external_diff_cmd_cfg;
386 done_preparing = 1;
387 return external_diff_cmd;
388 }
389
390 /*
391 * Keep track of files used for diffing. Sometimes such an entry
392 * refers to a temporary file, sometimes to an existing file, and
393 * sometimes to "/dev/null".
394 */
395 static struct diff_tempfile {
396 /*
397 * filename external diff should read from, or NULL if this
398 * entry is currently not in use:
399 */
400 const char *name;
401
402 char hex[GIT_MAX_HEXSZ + 1];
403 char mode[10];
404
405 /*
406 * If this diff_tempfile instance refers to a temporary file,
407 * this tempfile object is used to manage its lifetime.
408 */
409 struct tempfile tempfile;
410 } diff_temp[2];
411
412 typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
413
414 struct emit_callback {
415 int color_diff;
416 unsigned ws_rule;
417 int blank_at_eof_in_preimage;
418 int blank_at_eof_in_postimage;
419 int lno_in_preimage;
420 int lno_in_postimage;
421 sane_truncate_fn truncate;
422 const char **label_path;
423 struct diff_words_data *diff_words;
424 struct diff_options *opt;
425 struct strbuf *header;
426 };
427
428 static int count_lines(const char *data, int size)
429 {
430 int count, ch, completely_empty = 1, nl_just_seen = 0;
431 count = 0;
432 while (0 < size--) {
433 ch = *data++;
434 if (ch == '\n') {
435 count++;
436 nl_just_seen = 1;
437 completely_empty = 0;
438 }
439 else {
440 nl_just_seen = 0;
441 completely_empty = 0;
442 }
443 }
444 if (completely_empty)
445 return 0;
446 if (!nl_just_seen)
447 count++; /* no trailing newline */
448 return count;
449 }
450
451 static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
452 {
453 if (!DIFF_FILE_VALID(one)) {
454 mf->ptr = (char *)""; /* does not matter */
455 mf->size = 0;
456 return 0;
457 }
458 else if (diff_populate_filespec(one, 0))
459 return -1;
460
461 mf->ptr = one->data;
462 mf->size = one->size;
463 return 0;
464 }
465
466 /* like fill_mmfile, but only for size, so we can avoid retrieving blob */
467 static unsigned long diff_filespec_size(struct diff_filespec *one)
468 {
469 if (!DIFF_FILE_VALID(one))
470 return 0;
471 diff_populate_filespec(one, CHECK_SIZE_ONLY);
472 return one->size;
473 }
474
475 static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
476 {
477 char *ptr = mf->ptr;
478 long size = mf->size;
479 int cnt = 0;
480
481 if (!size)
482 return cnt;
483 ptr += size - 1; /* pointing at the very end */
484 if (*ptr != '\n')
485 ; /* incomplete line */
486 else
487 ptr--; /* skip the last LF */
488 while (mf->ptr < ptr) {
489 char *prev_eol;
490 for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
491 if (*prev_eol == '\n')
492 break;
493 if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
494 break;
495 cnt++;
496 ptr = prev_eol - 1;
497 }
498 return cnt;
499 }
500
501 static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
502 struct emit_callback *ecbdata)
503 {
504 int l1, l2, at;
505 unsigned ws_rule = ecbdata->ws_rule;
506 l1 = count_trailing_blank(mf1, ws_rule);
507 l2 = count_trailing_blank(mf2, ws_rule);
508 if (l2 <= l1) {
509 ecbdata->blank_at_eof_in_preimage = 0;
510 ecbdata->blank_at_eof_in_postimage = 0;
511 return;
512 }
513 at = count_lines(mf1->ptr, mf1->size);
514 ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
515
516 at = count_lines(mf2->ptr, mf2->size);
517 ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
518 }
519
520 static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
521 int first, const char *line, int len)
522 {
523 int has_trailing_newline, has_trailing_carriage_return;
524 int nofirst;
525 FILE *file = o->file;
526
527 fputs(diff_line_prefix(o), file);
528
529 if (len == 0) {
530 has_trailing_newline = (first == '\n');
531 has_trailing_carriage_return = (!has_trailing_newline &&
532 (first == '\r'));
533 nofirst = has_trailing_newline || has_trailing_carriage_return;
534 } else {
535 has_trailing_newline = (len > 0 && line[len-1] == '\n');
536 if (has_trailing_newline)
537 len--;
538 has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
539 if (has_trailing_carriage_return)
540 len--;
541 nofirst = 0;
542 }
543
544 if (len || !nofirst) {
545 fputs(set, file);
546 if (!nofirst)
547 fputc(first, file);
548 fwrite(line, len, 1, file);
549 fputs(reset, file);
550 }
551 if (has_trailing_carriage_return)
552 fputc('\r', file);
553 if (has_trailing_newline)
554 fputc('\n', file);
555 }
556
557 static void emit_line(struct diff_options *o, const char *set, const char *reset,
558 const char *line, int len)
559 {
560 emit_line_0(o, set, reset, line[0], line+1, len-1);
561 }
562
563 static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
564 {
565 if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
566 ecbdata->blank_at_eof_in_preimage &&
567 ecbdata->blank_at_eof_in_postimage &&
568 ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
569 ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
570 return 0;
571 return ws_blank_line(line, len, ecbdata->ws_rule);
572 }
573
574 static void emit_line_checked(const char *reset,
575 struct emit_callback *ecbdata,
576 const char *line, int len,
577 enum color_diff color,
578 unsigned ws_error_highlight,
579 char sign)
580 {
581 const char *set = diff_get_color(ecbdata->color_diff, color);
582 const char *ws = NULL;
583
584 if (ecbdata->opt->ws_error_highlight & ws_error_highlight) {
585 ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
586 if (!*ws)
587 ws = NULL;
588 }
589
590 if (!ws)
591 emit_line_0(ecbdata->opt, set, reset, sign, line, len);
592 else if (sign == '+' && new_blank_line_at_eof(ecbdata, line, len))
593 /* Blank line at EOF - paint '+' as well */
594 emit_line_0(ecbdata->opt, ws, reset, sign, line, len);
595 else {
596 /* Emit just the prefix, then the rest. */
597 emit_line_0(ecbdata->opt, set, reset, sign, "", 0);
598 ws_check_emit(line, len, ecbdata->ws_rule,
599 ecbdata->opt->file, set, reset, ws);
600 }
601 }
602
603 static void emit_add_line(const char *reset,
604 struct emit_callback *ecbdata,
605 const char *line, int len)
606 {
607 emit_line_checked(reset, ecbdata, line, len,
608 DIFF_FILE_NEW, WSEH_NEW, '+');
609 }
610
611 static void emit_del_line(const char *reset,
612 struct emit_callback *ecbdata,
613 const char *line, int len)
614 {
615 emit_line_checked(reset, ecbdata, line, len,
616 DIFF_FILE_OLD, WSEH_OLD, '-');
617 }
618
619 static void emit_context_line(const char *reset,
620 struct emit_callback *ecbdata,
621 const char *line, int len)
622 {
623 emit_line_checked(reset, ecbdata, line, len,
624 DIFF_CONTEXT, WSEH_CONTEXT, ' ');
625 }
626
627 static void emit_hunk_header(struct emit_callback *ecbdata,
628 const char *line, int len)
629 {
630 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
631 const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
632 const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
633 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
634 static const char atat[2] = { '@', '@' };
635 const char *cp, *ep;
636 struct strbuf msgbuf = STRBUF_INIT;
637 int org_len = len;
638 int i = 1;
639
640 /*
641 * As a hunk header must begin with "@@ -<old>, +<new> @@",
642 * it always is at least 10 bytes long.
643 */
644 if (len < 10 ||
645 memcmp(line, atat, 2) ||
646 !(ep = memmem(line + 2, len - 2, atat, 2))) {
647 emit_line(ecbdata->opt, context, reset, line, len);
648 return;
649 }
650 ep += 2; /* skip over @@ */
651
652 /* The hunk header in fraginfo color */
653 strbuf_addstr(&msgbuf, frag);
654 strbuf_add(&msgbuf, line, ep - line);
655 strbuf_addstr(&msgbuf, reset);
656
657 /*
658 * trailing "\r\n"
659 */
660 for ( ; i < 3; i++)
661 if (line[len - i] == '\r' || line[len - i] == '\n')
662 len--;
663
664 /* blank before the func header */
665 for (cp = ep; ep - line < len; ep++)
666 if (*ep != ' ' && *ep != '\t')
667 break;
668 if (ep != cp) {
669 strbuf_addstr(&msgbuf, context);
670 strbuf_add(&msgbuf, cp, ep - cp);
671 strbuf_addstr(&msgbuf, reset);
672 }
673
674 if (ep < line + len) {
675 strbuf_addstr(&msgbuf, func);
676 strbuf_add(&msgbuf, ep, line + len - ep);
677 strbuf_addstr(&msgbuf, reset);
678 }
679
680 strbuf_add(&msgbuf, line + len, org_len - len);
681 emit_line(ecbdata->opt, "", "", msgbuf.buf, msgbuf.len);
682 strbuf_release(&msgbuf);
683 }
684
685 static struct diff_tempfile *claim_diff_tempfile(void) {
686 int i;
687 for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
688 if (!diff_temp[i].name)
689 return diff_temp + i;
690 die("BUG: diff is failing to clean up its tempfiles");
691 }
692
693 static void remove_tempfile(void)
694 {
695 int i;
696 for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
697 if (is_tempfile_active(&diff_temp[i].tempfile))
698 delete_tempfile(&diff_temp[i].tempfile);
699 diff_temp[i].name = NULL;
700 }
701 }
702
703 static void print_line_count(FILE *file, int count)
704 {
705 switch (count) {
706 case 0:
707 fprintf(file, "0,0");
708 break;
709 case 1:
710 fprintf(file, "1");
711 break;
712 default:
713 fprintf(file, "1,%d", count);
714 break;
715 }
716 }
717
718 static void emit_rewrite_lines(struct emit_callback *ecb,
719 int prefix, const char *data, int size)
720 {
721 const char *endp = NULL;
722 static const char *nneof = " No newline at end of file\n";
723 const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
724
725 while (0 < size) {
726 int len;
727
728 endp = memchr(data, '\n', size);
729 len = endp ? (endp - data + 1) : size;
730 if (prefix != '+') {
731 ecb->lno_in_preimage++;
732 emit_del_line(reset, ecb, data, len);
733 } else {
734 ecb->lno_in_postimage++;
735 emit_add_line(reset, ecb, data, len);
736 }
737 size -= len;
738 data += len;
739 }
740 if (!endp) {
741 const char *context = diff_get_color(ecb->color_diff,
742 DIFF_CONTEXT);
743 putc('\n', ecb->opt->file);
744 emit_line_0(ecb->opt, context, reset, '\\',
745 nneof, strlen(nneof));
746 }
747 }
748
749 static void emit_rewrite_diff(const char *name_a,
750 const char *name_b,
751 struct diff_filespec *one,
752 struct diff_filespec *two,
753 struct userdiff_driver *textconv_one,
754 struct userdiff_driver *textconv_two,
755 struct diff_options *o)
756 {
757 int lc_a, lc_b;
758 const char *name_a_tab, *name_b_tab;
759 const char *metainfo = diff_get_color(o->use_color, DIFF_METAINFO);
760 const char *fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
761 const char *reset = diff_get_color(o->use_color, DIFF_RESET);
762 static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
763 const char *a_prefix, *b_prefix;
764 char *data_one, *data_two;
765 size_t size_one, size_two;
766 struct emit_callback ecbdata;
767 const char *line_prefix = diff_line_prefix(o);
768
769 if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
770 a_prefix = o->b_prefix;
771 b_prefix = o->a_prefix;
772 } else {
773 a_prefix = o->a_prefix;
774 b_prefix = o->b_prefix;
775 }
776
777 name_a += (*name_a == '/');
778 name_b += (*name_b == '/');
779 name_a_tab = strchr(name_a, ' ') ? "\t" : "";
780 name_b_tab = strchr(name_b, ' ') ? "\t" : "";
781
782 strbuf_reset(&a_name);
783 strbuf_reset(&b_name);
784 quote_two_c_style(&a_name, a_prefix, name_a, 0);
785 quote_two_c_style(&b_name, b_prefix, name_b, 0);
786
787 size_one = fill_textconv(textconv_one, one, &data_one);
788 size_two = fill_textconv(textconv_two, two, &data_two);
789
790 memset(&ecbdata, 0, sizeof(ecbdata));
791 ecbdata.color_diff = want_color(o->use_color);
792 ecbdata.ws_rule = whitespace_rule(name_b);
793 ecbdata.opt = o;
794 if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
795 mmfile_t mf1, mf2;
796 mf1.ptr = (char *)data_one;
797 mf2.ptr = (char *)data_two;
798 mf1.size = size_one;
799 mf2.size = size_two;
800 check_blank_at_eof(&mf1, &mf2, &ecbdata);
801 }
802 ecbdata.lno_in_preimage = 1;
803 ecbdata.lno_in_postimage = 1;
804
805 lc_a = count_lines(data_one, size_one);
806 lc_b = count_lines(data_two, size_two);
807 fprintf(o->file,
808 "%s%s--- %s%s%s\n%s%s+++ %s%s%s\n%s%s@@ -",
809 line_prefix, metainfo, a_name.buf, name_a_tab, reset,
810 line_prefix, metainfo, b_name.buf, name_b_tab, reset,
811 line_prefix, fraginfo);
812 if (!o->irreversible_delete)
813 print_line_count(o->file, lc_a);
814 else
815 fprintf(o->file, "?,?");
816 fprintf(o->file, " +");
817 print_line_count(o->file, lc_b);
818 fprintf(o->file, " @@%s\n", reset);
819 if (lc_a && !o->irreversible_delete)
820 emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
821 if (lc_b)
822 emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
823 if (textconv_one)
824 free((char *)data_one);
825 if (textconv_two)
826 free((char *)data_two);
827 }
828
829 struct diff_words_buffer {
830 mmfile_t text;
831 long alloc;
832 struct diff_words_orig {
833 const char *begin, *end;
834 } *orig;
835 int orig_nr, orig_alloc;
836 };
837
838 static void diff_words_append(char *line, unsigned long len,
839 struct diff_words_buffer *buffer)
840 {
841 ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
842 line++;
843 len--;
844 memcpy(buffer->text.ptr + buffer->text.size, line, len);
845 buffer->text.size += len;
846 buffer->text.ptr[buffer->text.size] = '\0';
847 }
848
849 struct diff_words_style_elem {
850 const char *prefix;
851 const char *suffix;
852 const char *color; /* NULL; filled in by the setup code if
853 * color is enabled */
854 };
855
856 struct diff_words_style {
857 enum diff_words_type type;
858 struct diff_words_style_elem new, old, ctx;
859 const char *newline;
860 };
861
862 static struct diff_words_style diff_words_styles[] = {
863 { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
864 { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
865 { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
866 };
867
868 struct diff_words_data {
869 struct diff_words_buffer minus, plus;
870 const char *current_plus;
871 int last_minus;
872 struct diff_options *opt;
873 regex_t *word_regex;
874 enum diff_words_type type;
875 struct diff_words_style *style;
876 };
877
878 static int fn_out_diff_words_write_helper(FILE *fp,
879 struct diff_words_style_elem *st_el,
880 const char *newline,
881 size_t count, const char *buf,
882 const char *line_prefix)
883 {
884 int print = 0;
885
886 while (count) {
887 char *p = memchr(buf, '\n', count);
888 if (print)
889 fputs(line_prefix, fp);
890 if (p != buf) {
891 if (st_el->color && fputs(st_el->color, fp) < 0)
892 return -1;
893 if (fputs(st_el->prefix, fp) < 0 ||
894 fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
895 fputs(st_el->suffix, fp) < 0)
896 return -1;
897 if (st_el->color && *st_el->color
898 && fputs(GIT_COLOR_RESET, fp) < 0)
899 return -1;
900 }
901 if (!p)
902 return 0;
903 if (fputs(newline, fp) < 0)
904 return -1;
905 count -= p + 1 - buf;
906 buf = p + 1;
907 print = 1;
908 }
909 return 0;
910 }
911
912 /*
913 * '--color-words' algorithm can be described as:
914 *
915 * 1. collect the minus/plus lines of a diff hunk, divided into
916 * minus-lines and plus-lines;
917 *
918 * 2. break both minus-lines and plus-lines into words and
919 * place them into two mmfile_t with one word for each line;
920 *
921 * 3. use xdiff to run diff on the two mmfile_t to get the words level diff;
922 *
923 * And for the common parts of the both file, we output the plus side text.
924 * diff_words->current_plus is used to trace the current position of the plus file
925 * which printed. diff_words->last_minus is used to trace the last minus word
926 * printed.
927 *
928 * For '--graph' to work with '--color-words', we need to output the graph prefix
929 * on each line of color words output. Generally, there are two conditions on
930 * which we should output the prefix.
931 *
932 * 1. diff_words->last_minus == 0 &&
933 * diff_words->current_plus == diff_words->plus.text.ptr
934 *
935 * that is: the plus text must start as a new line, and if there is no minus
936 * word printed, a graph prefix must be printed.
937 *
938 * 2. diff_words->current_plus > diff_words->plus.text.ptr &&
939 * *(diff_words->current_plus - 1) == '\n'
940 *
941 * that is: a graph prefix must be printed following a '\n'
942 */
943 static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
944 {
945 if ((diff_words->last_minus == 0 &&
946 diff_words->current_plus == diff_words->plus.text.ptr) ||
947 (diff_words->current_plus > diff_words->plus.text.ptr &&
948 *(diff_words->current_plus - 1) == '\n')) {
949 return 1;
950 } else {
951 return 0;
952 }
953 }
954
955 static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
956 {
957 struct diff_words_data *diff_words = priv;
958 struct diff_words_style *style = diff_words->style;
959 int minus_first, minus_len, plus_first, plus_len;
960 const char *minus_begin, *minus_end, *plus_begin, *plus_end;
961 struct diff_options *opt = diff_words->opt;
962 const char *line_prefix;
963
964 if (line[0] != '@' || parse_hunk_header(line, len,
965 &minus_first, &minus_len, &plus_first, &plus_len))
966 return;
967
968 assert(opt);
969 line_prefix = diff_line_prefix(opt);
970
971 /* POSIX requires that first be decremented by one if len == 0... */
972 if (minus_len) {
973 minus_begin = diff_words->minus.orig[minus_first].begin;
974 minus_end =
975 diff_words->minus.orig[minus_first + minus_len - 1].end;
976 } else
977 minus_begin = minus_end =
978 diff_words->minus.orig[minus_first].end;
979
980 if (plus_len) {
981 plus_begin = diff_words->plus.orig[plus_first].begin;
982 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
983 } else
984 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
985
986 if (color_words_output_graph_prefix(diff_words)) {
987 fputs(line_prefix, diff_words->opt->file);
988 }
989 if (diff_words->current_plus != plus_begin) {
990 fn_out_diff_words_write_helper(diff_words->opt->file,
991 &style->ctx, style->newline,
992 plus_begin - diff_words->current_plus,
993 diff_words->current_plus, line_prefix);
994 if (*(plus_begin - 1) == '\n')
995 fputs(line_prefix, diff_words->opt->file);
996 }
997 if (minus_begin != minus_end) {
998 fn_out_diff_words_write_helper(diff_words->opt->file,
999 &style->old, style->newline,
1000 minus_end - minus_begin, minus_begin,
1001 line_prefix);
1002 }
1003 if (plus_begin != plus_end) {
1004 fn_out_diff_words_write_helper(diff_words->opt->file,
1005 &style->new, style->newline,
1006 plus_end - plus_begin, plus_begin,
1007 line_prefix);
1008 }
1009
1010 diff_words->current_plus = plus_end;
1011 diff_words->last_minus = minus_first;
1012 }
1013
1014 /* This function starts looking at *begin, and returns 0 iff a word was found. */
1015 static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
1016 int *begin, int *end)
1017 {
1018 if (word_regex && *begin < buffer->size) {
1019 regmatch_t match[1];
1020 if (!regexec_buf(word_regex, buffer->ptr + *begin,
1021 buffer->size - *begin, 1, match, 0)) {
1022 char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
1023 '\n', match[0].rm_eo - match[0].rm_so);
1024 *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
1025 *begin += match[0].rm_so;
1026 return *begin >= *end;
1027 }
1028 return -1;
1029 }
1030
1031 /* find the next word */
1032 while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
1033 (*begin)++;
1034 if (*begin >= buffer->size)
1035 return -1;
1036
1037 /* find the end of the word */
1038 *end = *begin + 1;
1039 while (*end < buffer->size && !isspace(buffer->ptr[*end]))
1040 (*end)++;
1041
1042 return 0;
1043 }
1044
1045 /*
1046 * This function splits the words in buffer->text, stores the list with
1047 * newline separator into out, and saves the offsets of the original words
1048 * in buffer->orig.
1049 */
1050 static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
1051 regex_t *word_regex)
1052 {
1053 int i, j;
1054 long alloc = 0;
1055
1056 out->size = 0;
1057 out->ptr = NULL;
1058
1059 /* fake an empty "0th" word */
1060 ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
1061 buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
1062 buffer->orig_nr = 1;
1063
1064 for (i = 0; i < buffer->text.size; i++) {
1065 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
1066 return;
1067
1068 /* store original boundaries */
1069 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
1070 buffer->orig_alloc);
1071 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
1072 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
1073 buffer->orig_nr++;
1074
1075 /* store one word */
1076 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
1077 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
1078 out->ptr[out->size + j - i] = '\n';
1079 out->size += j - i + 1;
1080
1081 i = j - 1;
1082 }
1083 }
1084
1085 /* this executes the word diff on the accumulated buffers */
1086 static void diff_words_show(struct diff_words_data *diff_words)
1087 {
1088 xpparam_t xpp;
1089 xdemitconf_t xecfg;
1090 mmfile_t minus, plus;
1091 struct diff_words_style *style = diff_words->style;
1092
1093 struct diff_options *opt = diff_words->opt;
1094 const char *line_prefix;
1095
1096 assert(opt);
1097 line_prefix = diff_line_prefix(opt);
1098
1099 /* special case: only removal */
1100 if (!diff_words->plus.text.size) {
1101 fputs(line_prefix, diff_words->opt->file);
1102 fn_out_diff_words_write_helper(diff_words->opt->file,
1103 &style->old, style->newline,
1104 diff_words->minus.text.size,
1105 diff_words->minus.text.ptr, line_prefix);
1106 diff_words->minus.text.size = 0;
1107 return;
1108 }
1109
1110 diff_words->current_plus = diff_words->plus.text.ptr;
1111 diff_words->last_minus = 0;
1112
1113 memset(&xpp, 0, sizeof(xpp));
1114 memset(&xecfg, 0, sizeof(xecfg));
1115 diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
1116 diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
1117 xpp.flags = 0;
1118 /* as only the hunk header will be parsed, we need a 0-context */
1119 xecfg.ctxlen = 0;
1120 if (xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
1121 &xpp, &xecfg))
1122 die("unable to generate word diff");
1123 free(minus.ptr);
1124 free(plus.ptr);
1125 if (diff_words->current_plus != diff_words->plus.text.ptr +
1126 diff_words->plus.text.size) {
1127 if (color_words_output_graph_prefix(diff_words))
1128 fputs(line_prefix, diff_words->opt->file);
1129 fn_out_diff_words_write_helper(diff_words->opt->file,
1130 &style->ctx, style->newline,
1131 diff_words->plus.text.ptr + diff_words->plus.text.size
1132 - diff_words->current_plus, diff_words->current_plus,
1133 line_prefix);
1134 }
1135 diff_words->minus.text.size = diff_words->plus.text.size = 0;
1136 }
1137
1138 /* In "color-words" mode, show word-diff of words accumulated in the buffer */
1139 static void diff_words_flush(struct emit_callback *ecbdata)
1140 {
1141 if (ecbdata->diff_words->minus.text.size ||
1142 ecbdata->diff_words->plus.text.size)
1143 diff_words_show(ecbdata->diff_words);
1144 }
1145
1146 static void diff_filespec_load_driver(struct diff_filespec *one)
1147 {
1148 /* Use already-loaded driver */
1149 if (one->driver)
1150 return;
1151
1152 if (S_ISREG(one->mode))
1153 one->driver = userdiff_find_by_path(one->path);
1154
1155 /* Fallback to default settings */
1156 if (!one->driver)
1157 one->driver = userdiff_find_by_name("default");
1158 }
1159
1160 static const char *userdiff_word_regex(struct diff_filespec *one)
1161 {
1162 diff_filespec_load_driver(one);
1163 return one->driver->word_regex;
1164 }
1165
1166 static void init_diff_words_data(struct emit_callback *ecbdata,
1167 struct diff_options *orig_opts,
1168 struct diff_filespec *one,
1169 struct diff_filespec *two)
1170 {
1171 int i;
1172 struct diff_options *o = xmalloc(sizeof(struct diff_options));
1173 memcpy(o, orig_opts, sizeof(struct diff_options));
1174
1175 ecbdata->diff_words =
1176 xcalloc(1, sizeof(struct diff_words_data));
1177 ecbdata->diff_words->type = o->word_diff;
1178 ecbdata->diff_words->opt = o;
1179 if (!o->word_regex)
1180 o->word_regex = userdiff_word_regex(one);
1181 if (!o->word_regex)
1182 o->word_regex = userdiff_word_regex(two);
1183 if (!o->word_regex)
1184 o->word_regex = diff_word_regex_cfg;
1185 if (o->word_regex) {
1186 ecbdata->diff_words->word_regex = (regex_t *)
1187 xmalloc(sizeof(regex_t));
1188 if (regcomp(ecbdata->diff_words->word_regex,
1189 o->word_regex,
1190 REG_EXTENDED | REG_NEWLINE))
1191 die ("Invalid regular expression: %s",
1192 o->word_regex);
1193 }
1194 for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1195 if (o->word_diff == diff_words_styles[i].type) {
1196 ecbdata->diff_words->style =
1197 &diff_words_styles[i];
1198 break;
1199 }
1200 }
1201 if (want_color(o->use_color)) {
1202 struct diff_words_style *st = ecbdata->diff_words->style;
1203 st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1204 st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1205 st->ctx.color = diff_get_color_opt(o, DIFF_CONTEXT);
1206 }
1207 }
1208
1209 static void free_diff_words_data(struct emit_callback *ecbdata)
1210 {
1211 if (ecbdata->diff_words) {
1212 diff_words_flush(ecbdata);
1213 free (ecbdata->diff_words->opt);
1214 free (ecbdata->diff_words->minus.text.ptr);
1215 free (ecbdata->diff_words->minus.orig);
1216 free (ecbdata->diff_words->plus.text.ptr);
1217 free (ecbdata->diff_words->plus.orig);
1218 if (ecbdata->diff_words->word_regex) {
1219 regfree(ecbdata->diff_words->word_regex);
1220 free(ecbdata->diff_words->word_regex);
1221 }
1222 free(ecbdata->diff_words);
1223 ecbdata->diff_words = NULL;
1224 }
1225 }
1226
1227 const char *diff_get_color(int diff_use_color, enum color_diff ix)
1228 {
1229 if (want_color(diff_use_color))
1230 return diff_colors[ix];
1231 return "";
1232 }
1233
1234 const char *diff_line_prefix(struct diff_options *opt)
1235 {
1236 struct strbuf *msgbuf;
1237 if (!opt->output_prefix)
1238 return "";
1239
1240 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
1241 return msgbuf->buf;
1242 }
1243
1244 static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1245 {
1246 const char *cp;
1247 unsigned long allot;
1248 size_t l = len;
1249
1250 if (ecb->truncate)
1251 return ecb->truncate(line, len);
1252 cp = line;
1253 allot = l;
1254 while (0 < l) {
1255 (void) utf8_width(&cp, &l);
1256 if (!cp)
1257 break; /* truncated in the middle? */
1258 }
1259 return allot - l;
1260 }
1261
1262 static void find_lno(const char *line, struct emit_callback *ecbdata)
1263 {
1264 const char *p;
1265 ecbdata->lno_in_preimage = 0;
1266 ecbdata->lno_in_postimage = 0;
1267 p = strchr(line, '-');
1268 if (!p)
1269 return; /* cannot happen */
1270 ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1271 p = strchr(p, '+');
1272 if (!p)
1273 return; /* cannot happen */
1274 ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1275 }
1276
1277 static void fn_out_consume(void *priv, char *line, unsigned long len)
1278 {
1279 struct emit_callback *ecbdata = priv;
1280 const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
1281 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
1282 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1283 struct diff_options *o = ecbdata->opt;
1284 const char *line_prefix = diff_line_prefix(o);
1285
1286 o->found_changes = 1;
1287
1288 if (ecbdata->header) {
1289 fprintf(o->file, "%s", ecbdata->header->buf);
1290 strbuf_reset(ecbdata->header);
1291 ecbdata->header = NULL;
1292 }
1293
1294 if (ecbdata->label_path[0]) {
1295 const char *name_a_tab, *name_b_tab;
1296
1297 name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
1298 name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
1299
1300 fprintf(o->file, "%s%s--- %s%s%s\n",
1301 line_prefix, meta, ecbdata->label_path[0], reset, name_a_tab);
1302 fprintf(o->file, "%s%s+++ %s%s%s\n",
1303 line_prefix, meta, ecbdata->label_path[1], reset, name_b_tab);
1304 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1305 }
1306
1307 if (diff_suppress_blank_empty
1308 && len == 2 && line[0] == ' ' && line[1] == '\n') {
1309 line[0] = '\n';
1310 len = 1;
1311 }
1312
1313 if (line[0] == '@') {
1314 if (ecbdata->diff_words)
1315 diff_words_flush(ecbdata);
1316 len = sane_truncate_line(ecbdata, line, len);
1317 find_lno(line, ecbdata);
1318 emit_hunk_header(ecbdata, line, len);
1319 if (line[len-1] != '\n')
1320 putc('\n', o->file);
1321 return;
1322 }
1323
1324 if (ecbdata->diff_words) {
1325 if (line[0] == '-') {
1326 diff_words_append(line, len,
1327 &ecbdata->diff_words->minus);
1328 return;
1329 } else if (line[0] == '+') {
1330 diff_words_append(line, len,
1331 &ecbdata->diff_words->plus);
1332 return;
1333 } else if (starts_with(line, "\\ ")) {
1334 /*
1335 * Eat the "no newline at eof" marker as if we
1336 * saw a "+" or "-" line with nothing on it,
1337 * and return without diff_words_flush() to
1338 * defer processing. If this is the end of
1339 * preimage, more "+" lines may come after it.
1340 */
1341 return;
1342 }
1343 diff_words_flush(ecbdata);
1344 if (ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN) {
1345 emit_line(o, context, reset, line, len);
1346 fputs("~\n", o->file);
1347 } else {
1348 /*
1349 * Skip the prefix character, if any. With
1350 * diff_suppress_blank_empty, there may be
1351 * none.
1352 */
1353 if (line[0] != '\n') {
1354 line++;
1355 len--;
1356 }
1357 emit_line(o, context, reset, line, len);
1358 }
1359 return;
1360 }
1361
1362 switch (line[0]) {
1363 case '+':
1364 ecbdata->lno_in_postimage++;
1365 emit_add_line(reset, ecbdata, line + 1, len - 1);
1366 break;
1367 case '-':
1368 ecbdata->lno_in_preimage++;
1369 emit_del_line(reset, ecbdata, line + 1, len - 1);
1370 break;
1371 case ' ':
1372 ecbdata->lno_in_postimage++;
1373 ecbdata->lno_in_preimage++;
1374 emit_context_line(reset, ecbdata, line + 1, len - 1);
1375 break;
1376 default:
1377 /* incomplete line at the end */
1378 ecbdata->lno_in_preimage++;
1379 emit_line(o, diff_get_color(ecbdata->color_diff, DIFF_CONTEXT),
1380 reset, line, len);
1381 break;
1382 }
1383 }
1384
1385 static char *pprint_rename(const char *a, const char *b)
1386 {
1387 const char *old = a;
1388 const char *new = b;
1389 struct strbuf name = STRBUF_INIT;
1390 int pfx_length, sfx_length;
1391 int pfx_adjust_for_slash;
1392 int len_a = strlen(a);
1393 int len_b = strlen(b);
1394 int a_midlen, b_midlen;
1395 int qlen_a = quote_c_style(a, NULL, NULL, 0);
1396 int qlen_b = quote_c_style(b, NULL, NULL, 0);
1397
1398 if (qlen_a || qlen_b) {
1399 quote_c_style(a, &name, NULL, 0);
1400 strbuf_addstr(&name, " => ");
1401 quote_c_style(b, &name, NULL, 0);
1402 return strbuf_detach(&name, NULL);
1403 }
1404
1405 /* Find common prefix */
1406 pfx_length = 0;
1407 while (*old && *new && *old == *new) {
1408 if (*old == '/')
1409 pfx_length = old - a + 1;
1410 old++;
1411 new++;
1412 }
1413
1414 /* Find common suffix */
1415 old = a + len_a;
1416 new = b + len_b;
1417 sfx_length = 0;
1418 /*
1419 * If there is a common prefix, it must end in a slash. In
1420 * that case we let this loop run 1 into the prefix to see the
1421 * same slash.
1422 *
1423 * If there is no common prefix, we cannot do this as it would
1424 * underrun the input strings.
1425 */
1426 pfx_adjust_for_slash = (pfx_length ? 1 : 0);
1427 while (a + pfx_length - pfx_adjust_for_slash <= old &&
1428 b + pfx_length - pfx_adjust_for_slash <= new &&
1429 *old == *new) {
1430 if (*old == '/')
1431 sfx_length = len_a - (old - a);
1432 old--;
1433 new--;
1434 }
1435
1436 /*
1437 * pfx{mid-a => mid-b}sfx
1438 * {pfx-a => pfx-b}sfx
1439 * pfx{sfx-a => sfx-b}
1440 * name-a => name-b
1441 */
1442 a_midlen = len_a - pfx_length - sfx_length;
1443 b_midlen = len_b - pfx_length - sfx_length;
1444 if (a_midlen < 0)
1445 a_midlen = 0;
1446 if (b_midlen < 0)
1447 b_midlen = 0;
1448
1449 strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1450 if (pfx_length + sfx_length) {
1451 strbuf_add(&name, a, pfx_length);
1452 strbuf_addch(&name, '{');
1453 }
1454 strbuf_add(&name, a + pfx_length, a_midlen);
1455 strbuf_addstr(&name, " => ");
1456 strbuf_add(&name, b + pfx_length, b_midlen);
1457 if (pfx_length + sfx_length) {
1458 strbuf_addch(&name, '}');
1459 strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1460 }
1461 return strbuf_detach(&name, NULL);
1462 }
1463
1464 struct diffstat_t {
1465 int nr;
1466 int alloc;
1467 struct diffstat_file {
1468 char *from_name;
1469 char *name;
1470 char *print_name;
1471 unsigned is_unmerged:1;
1472 unsigned is_binary:1;
1473 unsigned is_renamed:1;
1474 unsigned is_interesting:1;
1475 uintmax_t added, deleted;
1476 } **files;
1477 };
1478
1479 static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1480 const char *name_a,
1481 const char *name_b)
1482 {
1483 struct diffstat_file *x;
1484 x = xcalloc(1, sizeof(*x));
1485 ALLOC_GROW(diffstat->files, diffstat->nr + 1, diffstat->alloc);
1486 diffstat->files[diffstat->nr++] = x;
1487 if (name_b) {
1488 x->from_name = xstrdup(name_a);
1489 x->name = xstrdup(name_b);
1490 x->is_renamed = 1;
1491 }
1492 else {
1493 x->from_name = NULL;
1494 x->name = xstrdup(name_a);
1495 }
1496 return x;
1497 }
1498
1499 static void diffstat_consume(void *priv, char *line, unsigned long len)
1500 {
1501 struct diffstat_t *diffstat = priv;
1502 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1503
1504 if (line[0] == '+')
1505 x->added++;
1506 else if (line[0] == '-')
1507 x->deleted++;
1508 }
1509
1510 const char mime_boundary_leader[] = "------------";
1511
1512 static int scale_linear(int it, int width, int max_change)
1513 {
1514 if (!it)
1515 return 0;
1516 /*
1517 * make sure that at least one '-' or '+' is printed if
1518 * there is any change to this path. The easiest way is to
1519 * scale linearly as if the alloted width is one column shorter
1520 * than it is, and then add 1 to the result.
1521 */
1522 return 1 + (it * (width - 1) / max_change);
1523 }
1524
1525 static void show_name(FILE *file,
1526 const char *prefix, const char *name, int len)
1527 {
1528 fprintf(file, " %s%-*s |", prefix, len, name);
1529 }
1530
1531 static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1532 {
1533 if (cnt <= 0)
1534 return;
1535 fprintf(file, "%s", set);
1536 while (cnt--)
1537 putc(ch, file);
1538 fprintf(file, "%s", reset);
1539 }
1540
1541 static void fill_print_name(struct diffstat_file *file)
1542 {
1543 char *pname;
1544
1545 if (file->print_name)
1546 return;
1547
1548 if (!file->is_renamed) {
1549 struct strbuf buf = STRBUF_INIT;
1550 if (quote_c_style(file->name, &buf, NULL, 0)) {
1551 pname = strbuf_detach(&buf, NULL);
1552 } else {
1553 pname = file->name;
1554 strbuf_release(&buf);
1555 }
1556 } else {
1557 pname = pprint_rename(file->from_name, file->name);
1558 }
1559 file->print_name = pname;
1560 }
1561
1562 int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1563 {
1564 struct strbuf sb = STRBUF_INIT;
1565 int ret;
1566
1567 if (!files) {
1568 assert(insertions == 0 && deletions == 0);
1569 return fprintf(fp, "%s\n", " 0 files changed");
1570 }
1571
1572 strbuf_addf(&sb,
1573 (files == 1) ? " %d file changed" : " %d files changed",
1574 files);
1575
1576 /*
1577 * For binary diff, the caller may want to print "x files
1578 * changed" with insertions == 0 && deletions == 0.
1579 *
1580 * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1581 * is probably less confusing (i.e skip over "2 files changed
1582 * but nothing about added/removed lines? Is this a bug in Git?").
1583 */
1584 if (insertions || deletions == 0) {
1585 strbuf_addf(&sb,
1586 (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
1587 insertions);
1588 }
1589
1590 if (deletions || insertions == 0) {
1591 strbuf_addf(&sb,
1592 (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
1593 deletions);
1594 }
1595 strbuf_addch(&sb, '\n');
1596 ret = fputs(sb.buf, fp);
1597 strbuf_release(&sb);
1598 return ret;
1599 }
1600
1601 static void show_stats(struct diffstat_t *data, struct diff_options *options)
1602 {
1603 int i, len, add, del, adds = 0, dels = 0;
1604 uintmax_t max_change = 0, max_len = 0;
1605 int total_files = data->nr, count;
1606 int width, name_width, graph_width, number_width = 0, bin_width = 0;
1607 const char *reset, *add_c, *del_c;
1608 const char *line_prefix = "";
1609 int extra_shown = 0;
1610
1611 if (data->nr == 0)
1612 return;
1613
1614 line_prefix = diff_line_prefix(options);
1615 count = options->stat_count ? options->stat_count : data->nr;
1616
1617 reset = diff_get_color_opt(options, DIFF_RESET);
1618 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1619 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1620
1621 /*
1622 * Find the longest filename and max number of changes
1623 */
1624 for (i = 0; (i < count) && (i < data->nr); i++) {
1625 struct diffstat_file *file = data->files[i];
1626 uintmax_t change = file->added + file->deleted;
1627
1628 if (!file->is_interesting && (change == 0)) {
1629 count++; /* not shown == room for one more */
1630 continue;
1631 }
1632 fill_print_name(file);
1633 len = strlen(file->print_name);
1634 if (max_len < len)
1635 max_len = len;
1636
1637 if (file->is_unmerged) {
1638 /* "Unmerged" is 8 characters */
1639 bin_width = bin_width < 8 ? 8 : bin_width;
1640 continue;
1641 }
1642 if (file->is_binary) {
1643 /* "Bin XXX -> YYY bytes" */
1644 int w = 14 + decimal_width(file->added)
1645 + decimal_width(file->deleted);
1646 bin_width = bin_width < w ? w : bin_width;
1647 /* Display change counts aligned with "Bin" */
1648 number_width = 3;
1649 continue;
1650 }
1651
1652 if (max_change < change)
1653 max_change = change;
1654 }
1655 count = i; /* where we can stop scanning in data->files[] */
1656
1657 /*
1658 * We have width = stat_width or term_columns() columns total.
1659 * We want a maximum of min(max_len, stat_name_width) for the name part.
1660 * We want a maximum of min(max_change, stat_graph_width) for the +- part.
1661 * We also need 1 for " " and 4 + decimal_width(max_change)
1662 * for " | NNNN " and one the empty column at the end, altogether
1663 * 6 + decimal_width(max_change).
1664 *
1665 * If there's not enough space, we will use the smaller of
1666 * stat_name_width (if set) and 5/8*width for the filename,
1667 * and the rest for constant elements + graph part, but no more
1668 * than stat_graph_width for the graph part.
1669 * (5/8 gives 50 for filename and 30 for the constant parts + graph
1670 * for the standard terminal size).
1671 *
1672 * In other words: stat_width limits the maximum width, and
1673 * stat_name_width fixes the maximum width of the filename,
1674 * and is also used to divide available columns if there
1675 * aren't enough.
1676 *
1677 * Binary files are displayed with "Bin XXX -> YYY bytes"
1678 * instead of the change count and graph. This part is treated
1679 * similarly to the graph part, except that it is not
1680 * "scaled". If total width is too small to accommodate the
1681 * guaranteed minimum width of the filename part and the
1682 * separators and this message, this message will "overflow"
1683 * making the line longer than the maximum width.
1684 */
1685
1686 if (options->stat_width == -1)
1687 width = term_columns() - strlen(line_prefix);
1688 else
1689 width = options->stat_width ? options->stat_width : 80;
1690 number_width = decimal_width(max_change) > number_width ?
1691 decimal_width(max_change) : number_width;
1692
1693 if (options->stat_graph_width == -1)
1694 options->stat_graph_width = diff_stat_graph_width;
1695
1696 /*
1697 * Guarantee 3/8*16==6 for the graph part
1698 * and 5/8*16==10 for the filename part
1699 */
1700 if (width < 16 + 6 + number_width)
1701 width = 16 + 6 + number_width;
1702
1703 /*
1704 * First assign sizes that are wanted, ignoring available width.
1705 * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
1706 * starting from "XXX" should fit in graph_width.
1707 */
1708 graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
1709 if (options->stat_graph_width &&
1710 options->stat_graph_width < graph_width)
1711 graph_width = options->stat_graph_width;
1712
1713 name_width = (options->stat_name_width > 0 &&
1714 options->stat_name_width < max_len) ?
1715 options->stat_name_width : max_len;
1716
1717 /*
1718 * Adjust adjustable widths not to exceed maximum width
1719 */
1720 if (name_width + number_width + 6 + graph_width > width) {
1721 if (graph_width > width * 3/8 - number_width - 6) {
1722 graph_width = width * 3/8 - number_width - 6;
1723 if (graph_width < 6)
1724 graph_width = 6;
1725 }
1726
1727 if (options->stat_graph_width &&
1728 graph_width > options->stat_graph_width)
1729 graph_width = options->stat_graph_width;
1730 if (name_width > width - number_width - 6 - graph_width)
1731 name_width = width - number_width - 6 - graph_width;
1732 else
1733 graph_width = width - number_width - 6 - name_width;
1734 }
1735
1736 /*
1737 * From here name_width is the width of the name area,
1738 * and graph_width is the width of the graph area.
1739 * max_change is used to scale graph properly.
1740 */
1741 for (i = 0; i < count; i++) {
1742 const char *prefix = "";
1743 struct diffstat_file *file = data->files[i];
1744 char *name = file->print_name;
1745 uintmax_t added = file->added;
1746 uintmax_t deleted = file->deleted;
1747 int name_len;
1748
1749 if (!file->is_interesting && (added + deleted == 0))
1750 continue;
1751
1752 /*
1753 * "scale" the filename
1754 */
1755 len = name_width;
1756 name_len = strlen(name);
1757 if (name_width < name_len) {
1758 char *slash;
1759 prefix = "...";
1760 len -= 3;
1761 name += name_len - len;
1762 slash = strchr(name, '/');
1763 if (slash)
1764 name = slash;
1765 }
1766
1767 if (file->is_binary) {
1768 fprintf(options->file, "%s", line_prefix);
1769 show_name(options->file, prefix, name, len);
1770 fprintf(options->file, " %*s", number_width, "Bin");
1771 if (!added && !deleted) {
1772 putc('\n', options->file);
1773 continue;
1774 }
1775 fprintf(options->file, " %s%"PRIuMAX"%s",
1776 del_c, deleted, reset);
1777 fprintf(options->file, " -> ");
1778 fprintf(options->file, "%s%"PRIuMAX"%s",
1779 add_c, added, reset);
1780 fprintf(options->file, " bytes");
1781 fprintf(options->file, "\n");
1782 continue;
1783 }
1784 else if (file->is_unmerged) {
1785 fprintf(options->file, "%s", line_prefix);
1786 show_name(options->file, prefix, name, len);
1787 fprintf(options->file, " Unmerged\n");
1788 continue;
1789 }
1790
1791 /*
1792 * scale the add/delete
1793 */
1794 add = added;
1795 del = deleted;
1796
1797 if (graph_width <= max_change) {
1798 int total = scale_linear(add + del, graph_width, max_change);
1799 if (total < 2 && add && del)
1800 /* width >= 2 due to the sanity check */
1801 total = 2;
1802 if (add < del) {
1803 add = scale_linear(add, graph_width, max_change);
1804 del = total - add;
1805 } else {
1806 del = scale_linear(del, graph_width, max_change);
1807 add = total - del;
1808 }
1809 }
1810 fprintf(options->file, "%s", line_prefix);
1811 show_name(options->file, prefix, name, len);
1812 fprintf(options->file, " %*"PRIuMAX"%s",
1813 number_width, added + deleted,
1814 added + deleted ? " " : "");
1815 show_graph(options->file, '+', add, add_c, reset);
1816 show_graph(options->file, '-', del, del_c, reset);
1817 fprintf(options->file, "\n");
1818 }
1819
1820 for (i = 0; i < data->nr; i++) {
1821 struct diffstat_file *file = data->files[i];
1822 uintmax_t added = file->added;
1823 uintmax_t deleted = file->deleted;
1824
1825 if (file->is_unmerged ||
1826 (!file->is_interesting && (added + deleted == 0))) {
1827 total_files--;
1828 continue;
1829 }
1830
1831 if (!file->is_binary) {
1832 adds += added;
1833 dels += deleted;
1834 }
1835 if (i < count)
1836 continue;
1837 if (!extra_shown)
1838 fprintf(options->file, "%s ...\n", line_prefix);
1839 extra_shown = 1;
1840 }
1841 fprintf(options->file, "%s", line_prefix);
1842 print_stat_summary(options->file, total_files, adds, dels);
1843 }
1844
1845 static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1846 {
1847 int i, adds = 0, dels = 0, total_files = data->nr;
1848
1849 if (data->nr == 0)
1850 return;
1851
1852 for (i = 0; i < data->nr; i++) {
1853 int added = data->files[i]->added;
1854 int deleted= data->files[i]->deleted;
1855
1856 if (data->files[i]->is_unmerged ||
1857 (!data->files[i]->is_interesting && (added + deleted == 0))) {
1858 total_files--;
1859 } else if (!data->files[i]->is_binary) { /* don't count bytes */
1860 adds += added;
1861 dels += deleted;
1862 }
1863 }
1864 fprintf(options->file, "%s", diff_line_prefix(options));
1865 print_stat_summary(options->file, total_files, adds, dels);
1866 }
1867
1868 static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1869 {
1870 int i;
1871
1872 if (data->nr == 0)
1873 return;
1874
1875 for (i = 0; i < data->nr; i++) {
1876 struct diffstat_file *file = data->files[i];
1877
1878 fprintf(options->file, "%s", diff_line_prefix(options));
1879
1880 if (file->is_binary)
1881 fprintf(options->file, "-\t-\t");
1882 else
1883 fprintf(options->file,
1884 "%"PRIuMAX"\t%"PRIuMAX"\t",
1885 file->added, file->deleted);
1886 if (options->line_termination) {
1887 fill_print_name(file);
1888 if (!file->is_renamed)
1889 write_name_quoted(file->name, options->file,
1890 options->line_termination);
1891 else {
1892 fputs(file->print_name, options->file);
1893 putc(options->line_termination, options->file);
1894 }
1895 } else {
1896 if (file->is_renamed) {
1897 putc('\0', options->file);
1898 write_name_quoted(file->from_name, options->file, '\0');
1899 }
1900 write_name_quoted(file->name, options->file, '\0');
1901 }
1902 }
1903 }
1904
1905 struct dirstat_file {
1906 const char *name;
1907 unsigned long changed;
1908 };
1909
1910 struct dirstat_dir {
1911 struct dirstat_file *files;
1912 int alloc, nr, permille, cumulative;
1913 };
1914
1915 static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1916 unsigned long changed, const char *base, int baselen)
1917 {
1918 unsigned long this_dir = 0;
1919 unsigned int sources = 0;
1920 const char *line_prefix = diff_line_prefix(opt);
1921
1922 while (dir->nr) {
1923 struct dirstat_file *f = dir->files;
1924 int namelen = strlen(f->name);
1925 unsigned long this;
1926 char *slash;
1927
1928 if (namelen < baselen)
1929 break;
1930 if (memcmp(f->name, base, baselen))
1931 break;
1932 slash = strchr(f->name + baselen, '/');
1933 if (slash) {
1934 int newbaselen = slash + 1 - f->name;
1935 this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
1936 sources++;
1937 } else {
1938 this = f->changed;
1939 dir->files++;
1940 dir->nr--;
1941 sources += 2;
1942 }
1943 this_dir += this;
1944 }
1945
1946 /*
1947 * We don't report dirstat's for
1948 * - the top level
1949 * - or cases where everything came from a single directory
1950 * under this directory (sources == 1).
1951 */
1952 if (baselen && sources != 1) {
1953 if (this_dir) {
1954 int permille = this_dir * 1000 / changed;
1955 if (permille >= dir->permille) {
1956 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
1957 permille / 10, permille % 10, baselen, base);
1958 if (!dir->cumulative)
1959 return 0;
1960 }
1961 }
1962 }
1963 return this_dir;
1964 }
1965
1966 static int dirstat_compare(const void *_a, const void *_b)
1967 {
1968 const struct dirstat_file *a = _a;
1969 const struct dirstat_file *b = _b;
1970 return strcmp(a->name, b->name);
1971 }
1972
1973 static void show_dirstat(struct diff_options *options)
1974 {
1975 int i;
1976 unsigned long changed;
1977 struct dirstat_dir dir;
1978 struct diff_queue_struct *q = &diff_queued_diff;
1979
1980 dir.files = NULL;
1981 dir.alloc = 0;
1982 dir.nr = 0;
1983 dir.permille = options->dirstat_permille;
1984 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1985
1986 changed = 0;
1987 for (i = 0; i < q->nr; i++) {
1988 struct diff_filepair *p = q->queue[i];
1989 const char *name;
1990 unsigned long copied, added, damage;
1991 int content_changed;
1992
1993 name = p->two->path ? p->two->path : p->one->path;
1994
1995 if (p->one->oid_valid && p->two->oid_valid)
1996 content_changed = oidcmp(&p->one->oid, &p->two->oid);
1997 else
1998 content_changed = 1;
1999
2000 if (!content_changed) {
2001 /*
2002 * The SHA1 has not changed, so pre-/post-content is
2003 * identical. We can therefore skip looking at the
2004 * file contents altogether.
2005 */
2006 damage = 0;
2007 goto found_damage;
2008 }
2009
2010 if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
2011 /*
2012 * In --dirstat-by-file mode, we don't really need to
2013 * look at the actual file contents at all.
2014 * The fact that the SHA1 changed is enough for us to
2015 * add this file to the list of results
2016 * (with each file contributing equal damage).
2017 */
2018 damage = 1;
2019 goto found_damage;
2020 }
2021
2022 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
2023 diff_populate_filespec(p->one, 0);
2024 diff_populate_filespec(p->two, 0);
2025 diffcore_count_changes(p->one, p->two, NULL, NULL,
2026 &copied, &added);
2027 diff_free_filespec_data(p->one);
2028 diff_free_filespec_data(p->two);
2029 } else if (DIFF_FILE_VALID(p->one)) {
2030 diff_populate_filespec(p->one, CHECK_SIZE_ONLY);
2031 copied = added = 0;
2032 diff_free_filespec_data(p->one);
2033 } else if (DIFF_FILE_VALID(p->two)) {
2034 diff_populate_filespec(p->two, CHECK_SIZE_ONLY);
2035 copied = 0;
2036 added = p->two->size;
2037 diff_free_filespec_data(p->two);
2038 } else
2039 continue;
2040
2041 /*
2042 * Original minus copied is the removed material,
2043 * added is the new material. They are both damages
2044 * made to the preimage.
2045 * If the resulting damage is zero, we know that
2046 * diffcore_count_changes() considers the two entries to
2047 * be identical, but since content_changed is true, we
2048 * know that there must have been _some_ kind of change,
2049 * so we force all entries to have damage > 0.
2050 */
2051 damage = (p->one->size - copied) + added;
2052 if (!damage)
2053 damage = 1;
2054
2055 found_damage:
2056 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2057 dir.files[dir.nr].name = name;
2058 dir.files[dir.nr].changed = damage;
2059 changed += damage;
2060 dir.nr++;
2061 }
2062
2063 /* This can happen even with many files, if everything was renames */
2064 if (!changed)
2065 return;
2066
2067 /* Show all directories with more than x% of the changes */
2068 QSORT(dir.files, dir.nr, dirstat_compare);
2069 gather_dirstat(options, &dir, changed, "", 0);
2070 }
2071
2072 static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
2073 {
2074 int i;
2075 unsigned long changed;
2076 struct dirstat_dir dir;
2077
2078 if (data->nr == 0)
2079 return;
2080
2081 dir.files = NULL;
2082 dir.alloc = 0;
2083 dir.nr = 0;
2084 dir.permille = options->dirstat_permille;
2085 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
2086
2087 changed = 0;
2088 for (i = 0; i < data->nr; i++) {
2089 struct diffstat_file *file = data->files[i];
2090 unsigned long damage = file->added + file->deleted;
2091 if (file->is_binary)
2092 /*
2093 * binary files counts bytes, not lines. Must find some
2094 * way to normalize binary bytes vs. textual lines.
2095 * The following heuristic assumes that there are 64
2096 * bytes per "line".
2097 * This is stupid and ugly, but very cheap...
2098 */
2099 damage = (damage + 63) / 64;
2100 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2101 dir.files[dir.nr].name = file->name;
2102 dir.files[dir.nr].changed = damage;
2103 changed += damage;
2104 dir.nr++;
2105 }
2106
2107 /* This can happen even with many files, if everything was renames */
2108 if (!changed)
2109 return;
2110
2111 /* Show all directories with more than x% of the changes */
2112 QSORT(dir.files, dir.nr, dirstat_compare);
2113 gather_dirstat(options, &dir, changed, "", 0);
2114 }
2115
2116 static void free_diffstat_info(struct diffstat_t *diffstat)
2117 {
2118 int i;
2119 for (i = 0; i < diffstat->nr; i++) {
2120 struct diffstat_file *f = diffstat->files[i];
2121 if (f->name != f->print_name)
2122 free(f->print_name);
2123 free(f->name);
2124 free(f->from_name);
2125 free(f);
2126 }
2127 free(diffstat->files);
2128 }
2129
2130 struct checkdiff_t {
2131 const char *filename;
2132 int lineno;
2133 int conflict_marker_size;
2134 struct diff_options *o;
2135 unsigned ws_rule;
2136 unsigned status;
2137 };
2138
2139 static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2140 {
2141 char firstchar;
2142 int cnt;
2143
2144 if (len < marker_size + 1)
2145 return 0;
2146 firstchar = line[0];
2147 switch (firstchar) {
2148 case '=': case '>': case '<': case '|':
2149 break;
2150 default:
2151 return 0;
2152 }
2153 for (cnt = 1; cnt < marker_size; cnt++)
2154 if (line[cnt] != firstchar)
2155 return 0;
2156 /* line[1] thru line[marker_size-1] are same as firstchar */
2157 if (len < marker_size + 1 || !isspace(line[marker_size]))
2158 return 0;
2159 return 1;
2160 }
2161
2162 static void checkdiff_consume(void *priv, char *line, unsigned long len)
2163 {
2164 struct checkdiff_t *data = priv;
2165 int marker_size = data->conflict_marker_size;
2166 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2167 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2168 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2169 char *err;
2170 const char *line_prefix;
2171
2172 assert(data->o);
2173 line_prefix = diff_line_prefix(data->o);
2174
2175 if (line[0] == '+') {
2176 unsigned bad;
2177 data->lineno++;
2178 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2179 data->status |= 1;
2180 fprintf(data->o->file,
2181 "%s%s:%d: leftover conflict marker\n",
2182 line_prefix, data->filename, data->lineno);
2183 }
2184 bad = ws_check(line + 1, len - 1, data->ws_rule);
2185 if (!bad)
2186 return;
2187 data->status |= bad;
2188 err = whitespace_error_string(bad);
2189 fprintf(data->o->file, "%s%s:%d: %s.\n",
2190 line_prefix, data->filename, data->lineno, err);
2191 free(err);
2192 emit_line(data->o, set, reset, line, 1);
2193 ws_check_emit(line + 1, len - 1, data->ws_rule,
2194 data->o->file, set, reset, ws);
2195 } else if (line[0] == ' ') {
2196 data->lineno++;
2197 } else if (line[0] == '@') {
2198 char *plus = strchr(line, '+');
2199 if (plus)
2200 data->lineno = strtol(plus, NULL, 10) - 1;
2201 else
2202 die("invalid diff");
2203 }
2204 }
2205
2206 static unsigned char *deflate_it(char *data,
2207 unsigned long size,
2208 unsigned long *result_size)
2209 {
2210 int bound;
2211 unsigned char *deflated;
2212 git_zstream stream;
2213
2214 git_deflate_init(&stream, zlib_compression_level);
2215 bound = git_deflate_bound(&stream, size);
2216 deflated = xmalloc(bound);
2217 stream.next_out = deflated;
2218 stream.avail_out = bound;
2219
2220 stream.next_in = (unsigned char *)data;
2221 stream.avail_in = size;
2222 while (git_deflate(&stream, Z_FINISH) == Z_OK)
2223 ; /* nothing */
2224 git_deflate_end(&stream);
2225 *result_size = stream.total_out;
2226 return deflated;
2227 }
2228
2229 static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two,
2230 const char *prefix)
2231 {
2232 void *cp;
2233 void *delta;
2234 void *deflated;
2235 void *data;
2236 unsigned long orig_size;
2237 unsigned long delta_size;
2238 unsigned long deflate_size;
2239 unsigned long data_size;
2240
2241 /* We could do deflated delta, or we could do just deflated two,
2242 * whichever is smaller.
2243 */
2244 delta = NULL;
2245 deflated = deflate_it(two->ptr, two->size, &deflate_size);
2246 if (one->size && two->size) {
2247 delta = diff_delta(one->ptr, one->size,
2248 two->ptr, two->size,
2249 &delta_size, deflate_size);
2250 if (delta) {
2251 void *to_free = delta;
2252 orig_size = delta_size;
2253 delta = deflate_it(delta, delta_size, &delta_size);
2254 free(to_free);
2255 }
2256 }
2257
2258 if (delta && delta_size < deflate_size) {
2259 fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2260 free(deflated);
2261 data = delta;
2262 data_size = delta_size;
2263 }
2264 else {
2265 fprintf(file, "%sliteral %lu\n", prefix, two->size);
2266 free(delta);
2267 data = deflated;
2268 data_size = deflate_size;
2269 }
2270
2271 /* emit data encoded in base85 */
2272 cp = data;
2273 while (data_size) {
2274 int bytes = (52 < data_size) ? 52 : data_size;
2275 char line[70];
2276 data_size -= bytes;
2277 if (bytes <= 26)
2278 line[0] = bytes + 'A' - 1;
2279 else
2280 line[0] = bytes - 26 + 'a' - 1;
2281 encode_85(line + 1, cp, bytes);
2282 cp = (char *) cp + bytes;
2283 fprintf(file, "%s", prefix);
2284 fputs(line, file);
2285 fputc('\n', file);
2286 }
2287 fprintf(file, "%s\n", prefix);
2288 free(data);
2289 }
2290
2291 static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two,
2292 const char *prefix)
2293 {
2294 fprintf(file, "%sGIT binary patch\n", prefix);
2295 emit_binary_diff_body(file, one, two, prefix);
2296 emit_binary_diff_body(file, two, one, prefix);
2297 }
2298
2299 int diff_filespec_is_binary(struct diff_filespec *one)
2300 {
2301 if (one->is_binary == -1) {
2302 diff_filespec_load_driver(one);
2303 if (one->driver->binary != -1)
2304 one->is_binary = one->driver->binary;
2305 else {
2306 if (!one->data && DIFF_FILE_VALID(one))
2307 diff_populate_filespec(one, CHECK_BINARY);
2308 if (one->is_binary == -1 && one->data)
2309 one->is_binary = buffer_is_binary(one->data,
2310 one->size);
2311 if (one->is_binary == -1)
2312 one->is_binary = 0;
2313 }
2314 }
2315 return one->is_binary;
2316 }
2317
2318 static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2319 {
2320 diff_filespec_load_driver(one);
2321 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2322 }
2323
2324 void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2325 {
2326 if (!options->a_prefix)
2327 options->a_prefix = a;
2328 if (!options->b_prefix)
2329 options->b_prefix = b;
2330 }
2331
2332 struct userdiff_driver *get_textconv(struct diff_filespec *one)
2333 {
2334 if (!DIFF_FILE_VALID(one))
2335 return NULL;
2336
2337 diff_filespec_load_driver(one);
2338 return userdiff_get_textconv(one->driver);
2339 }
2340
2341 static void builtin_diff(const char *name_a,
2342 const char *name_b,
2343 struct diff_filespec *one,
2344 struct diff_filespec *two,
2345 const char *xfrm_msg,
2346 int must_show_header,
2347 struct diff_options *o,
2348 int complete_rewrite)
2349 {
2350 mmfile_t mf1, mf2;
2351 const char *lbl[2];
2352 char *a_one, *b_two;
2353 const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
2354 const char *reset = diff_get_color_opt(o, DIFF_RESET);
2355 const char *a_prefix, *b_prefix;
2356 struct userdiff_driver *textconv_one = NULL;
2357 struct userdiff_driver *textconv_two = NULL;
2358 struct strbuf header = STRBUF_INIT;
2359 const char *line_prefix = diff_line_prefix(o);
2360
2361 diff_set_mnemonic_prefix(o, "a/", "b/");
2362 if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2363 a_prefix = o->b_prefix;
2364 b_prefix = o->a_prefix;
2365 } else {
2366 a_prefix = o->a_prefix;
2367 b_prefix = o->b_prefix;
2368 }
2369
2370 if (o->submodule_format == DIFF_SUBMODULE_LOG &&
2371 (!one->mode || S_ISGITLINK(one->mode)) &&
2372 (!two->mode || S_ISGITLINK(two->mode))) {
2373 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2374 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2375 show_submodule_summary(o->file, one->path ? one->path : two->path,
2376 line_prefix,
2377 &one->oid, &two->oid,
2378 two->dirty_submodule,
2379 meta, del, add, reset);
2380 return;
2381 } else if (o->submodule_format == DIFF_SUBMODULE_INLINE_DIFF &&
2382 (!one->mode || S_ISGITLINK(one->mode)) &&
2383 (!two->mode || S_ISGITLINK(two->mode))) {
2384 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2385 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2386 show_submodule_inline_diff(o->file, one->path ? one->path : two->path,
2387 line_prefix,
2388 &one->oid, &two->oid,
2389 two->dirty_submodule,
2390 meta, del, add, reset, o);
2391 return;
2392 }
2393
2394 if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2395 textconv_one = get_textconv(one);
2396 textconv_two = get_textconv(two);
2397 }
2398
2399 /* Never use a non-valid filename anywhere if at all possible */
2400 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2401 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2402
2403 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2404 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2405 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2406 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2407 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
2408 if (lbl[0][0] == '/') {
2409 /* /dev/null */
2410 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
2411 if (xfrm_msg)
2412 strbuf_addstr(&header, xfrm_msg);
2413 must_show_header = 1;
2414 }
2415 else if (lbl[1][0] == '/') {
2416 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
2417 if (xfrm_msg)
2418 strbuf_addstr(&header, xfrm_msg);
2419 must_show_header = 1;
2420 }
2421 else {
2422 if (one->mode != two->mode) {
2423 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
2424 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
2425 must_show_header = 1;
2426 }
2427 if (xfrm_msg)
2428 strbuf_addstr(&header, xfrm_msg);
2429
2430 /*
2431 * we do not run diff between different kind
2432 * of objects.
2433 */
2434 if ((one->mode ^ two->mode) & S_IFMT)
2435 goto free_ab_and_return;
2436 if (complete_rewrite &&
2437 (textconv_one || !diff_filespec_is_binary(one)) &&
2438 (textconv_two || !diff_filespec_is_binary(two))) {
2439 fprintf(o->file, "%s", header.buf);
2440 strbuf_reset(&header);
2441 emit_rewrite_diff(name_a, name_b, one, two,
2442 textconv_one, textconv_two, o);
2443 o->found_changes = 1;
2444 goto free_ab_and_return;
2445 }
2446 }
2447
2448 if (o->irreversible_delete && lbl[1][0] == '/') {
2449 fprintf(o->file, "%s", header.buf);
2450 strbuf_reset(&header);
2451 goto free_ab_and_return;
2452 } else if (!DIFF_OPT_TST(o, TEXT) &&
2453 ( (!textconv_one && diff_filespec_is_binary(one)) ||
2454 (!textconv_two && diff_filespec_is_binary(two)) )) {
2455 if (!one->data && !two->data &&
2456 S_ISREG(one->mode) && S_ISREG(two->mode) &&
2457 !DIFF_OPT_TST(o, BINARY)) {
2458 if (!oidcmp(&one->oid, &two->oid)) {
2459 if (must_show_header)
2460 fprintf(o->file, "%s", header.buf);
2461 goto free_ab_and_return;
2462 }
2463 fprintf(o->file, "%s", header.buf);
2464 fprintf(o->file, "%sBinary files %s and %s differ\n",
2465 line_prefix, lbl[0], lbl[1]);
2466 goto free_ab_and_return;
2467 }
2468 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2469 die("unable to read files to diff");
2470 /* Quite common confusing case */
2471 if (mf1.size == mf2.size &&
2472 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2473 if (must_show_header)
2474 fprintf(o->file, "%s", header.buf);
2475 goto free_ab_and_return;
2476 }
2477 fprintf(o->file, "%s", header.buf);
2478 strbuf_reset(&header);
2479 if (DIFF_OPT_TST(o, BINARY))
2480 emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2481 else
2482 fprintf(o->file, "%sBinary files %s and %s differ\n",
2483 line_prefix, lbl[0], lbl[1]);
2484 o->found_changes = 1;
2485 } else {
2486 /* Crazy xdl interfaces.. */
2487 const char *diffopts = getenv("GIT_DIFF_OPTS");
2488 const char *v;
2489 xpparam_t xpp;
2490 xdemitconf_t xecfg;
2491 struct emit_callback ecbdata;
2492 const struct userdiff_funcname *pe;
2493
2494 if (must_show_header) {
2495 fprintf(o->file, "%s", header.buf);
2496 strbuf_reset(&header);
2497 }
2498
2499 mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2500 mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2501
2502 pe = diff_funcname_pattern(one);
2503 if (!pe)
2504 pe = diff_funcname_pattern(two);
2505
2506 memset(&xpp, 0, sizeof(xpp));
2507 memset(&xecfg, 0, sizeof(xecfg));
2508 memset(&ecbdata, 0, sizeof(ecbdata));
2509 ecbdata.label_path = lbl;
2510 ecbdata.color_diff = want_color(o->use_color);
2511 ecbdata.ws_rule = whitespace_rule(name_b);
2512 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2513 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2514 ecbdata.opt = o;
2515 ecbdata.header = header.len ? &header : NULL;
2516 xpp.flags = o->xdl_opts;
2517 xecfg.ctxlen = o->context;
2518 xecfg.interhunkctxlen = o->interhunkcontext;
2519 xecfg.flags = XDL_EMIT_FUNCNAMES;
2520 if (DIFF_OPT_TST(o, FUNCCONTEXT))
2521 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2522 if (pe)
2523 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2524 if (!diffopts)
2525 ;
2526 else if (skip_prefix(diffopts, "--unified=", &v))
2527 xecfg.ctxlen = strtoul(v, NULL, 10);
2528 else if (skip_prefix(diffopts, "-u", &v))
2529 xecfg.ctxlen = strtoul(v, NULL, 10);
2530 if (o->word_diff)
2531 init_diff_words_data(&ecbdata, o, one, two);
2532 if (xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2533 &xpp, &xecfg))
2534 die("unable to generate diff for %s", one->path);
2535 if (o->word_diff)
2536 free_diff_words_data(&ecbdata);
2537 if (textconv_one)
2538 free(mf1.ptr);
2539 if (textconv_two)
2540 free(mf2.ptr);
2541 xdiff_clear_find_func(&xecfg);
2542 }
2543
2544 free_ab_and_return:
2545 strbuf_release(&header);
2546 diff_free_filespec_data(one);
2547 diff_free_filespec_data(two);
2548 free(a_one);
2549 free(b_two);
2550 return;
2551 }
2552
2553 static void builtin_diffstat(const char *name_a, const char *name_b,
2554 struct diff_filespec *one,
2555 struct diff_filespec *two,
2556 struct diffstat_t *diffstat,
2557 struct diff_options *o,
2558 struct diff_filepair *p)
2559 {
2560 mmfile_t mf1, mf2;
2561 struct diffstat_file *data;
2562 int same_contents;
2563 int complete_rewrite = 0;
2564
2565 if (!DIFF_PAIR_UNMERGED(p)) {
2566 if (p->status == DIFF_STATUS_MODIFIED && p->score)
2567 complete_rewrite = 1;
2568 }
2569
2570 data = diffstat_add(diffstat, name_a, name_b);
2571 data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
2572
2573 if (!one || !two) {
2574 data->is_unmerged = 1;
2575 return;
2576 }
2577
2578 same_contents = !oidcmp(&one->oid, &two->oid);
2579
2580 if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2581 data->is_binary = 1;
2582 if (same_contents) {
2583 data->added = 0;
2584 data->deleted = 0;
2585 } else {
2586 data->added = diff_filespec_size(two);
2587 data->deleted = diff_filespec_size(one);
2588 }
2589 }
2590
2591 else if (complete_rewrite) {
2592 diff_populate_filespec(one, 0);
2593 diff_populate_filespec(two, 0);
2594 data->deleted = count_lines(one->data, one->size);
2595 data->added = count_lines(two->data, two->size);
2596 }
2597
2598 else if (!same_contents) {
2599 /* Crazy xdl interfaces.. */
2600 xpparam_t xpp;
2601 xdemitconf_t xecfg;
2602
2603 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2604 die("unable to read files to diff");
2605
2606 memset(&xpp, 0, sizeof(xpp));
2607 memset(&xecfg, 0, sizeof(xecfg));
2608 xpp.flags = o->xdl_opts;
2609 xecfg.ctxlen = o->context;
2610 xecfg.interhunkctxlen = o->interhunkcontext;
2611 if (xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2612 &xpp, &xecfg))
2613 die("unable to generate diffstat for %s", one->path);
2614 }
2615
2616 diff_free_filespec_data(one);
2617 diff_free_filespec_data(two);
2618 }
2619
2620 static void builtin_checkdiff(const char *name_a, const char *name_b,
2621 const char *attr_path,
2622 struct diff_filespec *one,
2623 struct diff_filespec *two,
2624 struct diff_options *o)
2625 {
2626 mmfile_t mf1, mf2;
2627 struct checkdiff_t data;
2628
2629 if (!two)
2630 return;
2631
2632 memset(&data, 0, sizeof(data));
2633 data.filename = name_b ? name_b : name_a;
2634 data.lineno = 0;
2635 data.o = o;
2636 data.ws_rule = whitespace_rule(attr_path);
2637 data.conflict_marker_size = ll_merge_marker_size(attr_path);
2638
2639 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2640 die("unable to read files to diff");
2641
2642 /*
2643 * All the other codepaths check both sides, but not checking
2644 * the "old" side here is deliberate. We are checking the newly
2645 * introduced changes, and as long as the "new" side is text, we
2646 * can and should check what it introduces.
2647 */
2648 if (diff_filespec_is_binary(two))
2649 goto free_and_return;
2650 else {
2651 /* Crazy xdl interfaces.. */
2652 xpparam_t xpp;
2653 xdemitconf_t xecfg;
2654
2655 memset(&xpp, 0, sizeof(xpp));
2656 memset(&xecfg, 0, sizeof(xecfg));
2657 xecfg.ctxlen = 1; /* at least one context line */
2658 xpp.flags = 0;
2659 if (xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2660 &xpp, &xecfg))
2661 die("unable to generate checkdiff for %s", one->path);
2662
2663 if (data.ws_rule & WS_BLANK_AT_EOF) {
2664 struct emit_callback ecbdata;
2665 int blank_at_eof;
2666
2667 ecbdata.ws_rule = data.ws_rule;
2668 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2669 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2670
2671 if (blank_at_eof) {
2672 static char *err;
2673 if (!err)
2674 err = whitespace_error_string(WS_BLANK_AT_EOF);
2675 fprintf(o->file, "%s:%d: %s.\n",
2676 data.filename, blank_at_eof, err);
2677 data.status = 1; /* report errors */
2678 }
2679 }
2680 }
2681 free_and_return:
2682 diff_free_filespec_data(one);
2683 diff_free_filespec_data(two);
2684 if (data.status)
2685 DIFF_OPT_SET(o, CHECK_FAILED);
2686 }
2687
2688 struct diff_filespec *alloc_filespec(const char *path)
2689 {
2690 struct diff_filespec *spec;
2691
2692 FLEXPTR_ALLOC_STR(spec, path, path);
2693 spec->count = 1;
2694 spec->is_binary = -1;
2695 return spec;
2696 }
2697
2698 void free_filespec(struct diff_filespec *spec)
2699 {
2700 if (!--spec->count) {
2701 diff_free_filespec_data(spec);
2702 free(spec);
2703 }
2704 }
2705
2706 void fill_filespec(struct diff_filespec *spec, const struct object_id *oid,
2707 int oid_valid, unsigned short mode)
2708 {
2709 if (mode) {
2710 spec->mode = canon_mode(mode);
2711 oidcpy(&spec->oid, oid);
2712 spec->oid_valid = oid_valid;
2713 }
2714 }
2715
2716 /*
2717 * Given a name and sha1 pair, if the index tells us the file in
2718 * the work tree has that object contents, return true, so that
2719 * prepare_temp_file() does not have to inflate and extract.
2720 */
2721 static int reuse_worktree_file(const char *name, const struct object_id *oid, int want_file)
2722 {
2723 const struct cache_entry *ce;
2724 struct stat st;
2725 int pos, len;
2726
2727 /*
2728 * We do not read the cache ourselves here, because the
2729 * benchmark with my previous version that always reads cache
2730 * shows that it makes things worse for diff-tree comparing
2731 * two linux-2.6 kernel trees in an already checked out work
2732 * tree. This is because most diff-tree comparisons deal with
2733 * only a small number of files, while reading the cache is
2734 * expensive for a large project, and its cost outweighs the
2735 * savings we get by not inflating the object to a temporary
2736 * file. Practically, this code only helps when we are used
2737 * by diff-cache --cached, which does read the cache before
2738 * calling us.
2739 */
2740 if (!active_cache)
2741 return 0;
2742
2743 /* We want to avoid the working directory if our caller
2744 * doesn't need the data in a normal file, this system
2745 * is rather slow with its stat/open/mmap/close syscalls,
2746 * and the object is contained in a pack file. The pack
2747 * is probably already open and will be faster to obtain
2748 * the data through than the working directory. Loose
2749 * objects however would tend to be slower as they need
2750 * to be individually opened and inflated.
2751 */
2752 if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(oid->hash))
2753 return 0;
2754
2755 /*
2756 * Similarly, if we'd have to convert the file contents anyway, that
2757 * makes the optimization not worthwhile.
2758 */
2759 if (!want_file && would_convert_to_git(&the_index, name))
2760 return 0;
2761
2762 len = strlen(name);
2763 pos = cache_name_pos(name, len);
2764 if (pos < 0)
2765 return 0;
2766 ce = active_cache[pos];
2767
2768 /*
2769 * This is not the sha1 we are looking for, or
2770 * unreusable because it is not a regular file.
2771 */
2772 if (oidcmp(oid, &ce->oid) || !S_ISREG(ce->ce_mode))
2773 return 0;
2774
2775 /*
2776 * If ce is marked as "assume unchanged", there is no
2777 * guarantee that work tree matches what we are looking for.
2778 */
2779 if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2780 return 0;
2781
2782 /*
2783 * If ce matches the file in the work tree, we can reuse it.
2784 */
2785 if (ce_uptodate(ce) ||
2786 (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2787 return 1;
2788
2789 return 0;
2790 }
2791
2792 static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2793 {
2794 struct strbuf buf = STRBUF_INIT;
2795 char *dirty = "";
2796
2797 /* Are we looking at the work tree? */
2798 if (s->dirty_submodule)
2799 dirty = "-dirty";
2800
2801 strbuf_addf(&buf, "Subproject commit %s%s\n",
2802 oid_to_hex(&s->oid), dirty);
2803 s->size = buf.len;
2804 if (size_only) {
2805 s->data = NULL;
2806 strbuf_release(&buf);
2807 } else {
2808 s->data = strbuf_detach(&buf, NULL);
2809 s->should_free = 1;
2810 }
2811 return 0;
2812 }
2813
2814 /*
2815 * While doing rename detection and pickaxe operation, we may need to
2816 * grab the data for the blob (or file) for our own in-core comparison.
2817 * diff_filespec has data and size fields for this purpose.
2818 */
2819 int diff_populate_filespec(struct diff_filespec *s, unsigned int flags)
2820 {
2821 int size_only = flags & CHECK_SIZE_ONLY;
2822 int err = 0;
2823 /*
2824 * demote FAIL to WARN to allow inspecting the situation
2825 * instead of refusing.
2826 */
2827 enum safe_crlf crlf_warn = (safe_crlf == SAFE_CRLF_FAIL
2828 ? SAFE_CRLF_WARN
2829 : safe_crlf);
2830
2831 if (!DIFF_FILE_VALID(s))
2832 die("internal error: asking to populate invalid file.");
2833 if (S_ISDIR(s->mode))
2834 return -1;
2835
2836 if (s->data)
2837 return 0;
2838
2839 if (size_only && 0 < s->size)
2840 return 0;
2841
2842 if (S_ISGITLINK(s->mode))
2843 return diff_populate_gitlink(s, size_only);
2844
2845 if (!s->oid_valid ||
2846 reuse_worktree_file(s->path, &s->oid, 0)) {
2847 struct strbuf buf = STRBUF_INIT;
2848 struct stat st;
2849 int fd;
2850
2851 if (lstat(s->path, &st) < 0) {
2852 if (errno == ENOENT) {
2853 err_empty:
2854 err = -1;
2855 empty:
2856 s->data = (char *)"";
2857 s->size = 0;
2858 return err;
2859 }
2860 }
2861 s->size = xsize_t(st.st_size);
2862 if (!s->size)
2863 goto empty;
2864 if (S_ISLNK(st.st_mode)) {
2865 struct strbuf sb = STRBUF_INIT;
2866
2867 if (strbuf_readlink(&sb, s->path, s->size))
2868 goto err_empty;
2869 s->size = sb.len;
2870 s->data = strbuf_detach(&sb, NULL);
2871 s->should_free = 1;
2872 return 0;
2873 }
2874
2875 /*
2876 * Even if the caller would be happy with getting
2877 * only the size, we cannot return early at this
2878 * point if the path requires us to run the content
2879 * conversion.
2880 */
2881 if (size_only && !would_convert_to_git(&the_index, s->path))
2882 return 0;
2883
2884 /*
2885 * Note: this check uses xsize_t(st.st_size) that may
2886 * not be the true size of the blob after it goes
2887 * through convert_to_git(). This may not strictly be
2888 * correct, but the whole point of big_file_threshold
2889 * and is_binary check being that we want to avoid
2890 * opening the file and inspecting the contents, this
2891 * is probably fine.
2892 */
2893 if ((flags & CHECK_BINARY) &&
2894 s->size > big_file_threshold && s->is_binary == -1) {
2895 s->is_binary = 1;
2896 return 0;
2897 }
2898 fd = open(s->path, O_RDONLY);
2899 if (fd < 0)
2900 goto err_empty;
2901 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2902 close(fd);
2903 s->should_munmap = 1;
2904
2905 /*
2906 * Convert from working tree format to canonical git format
2907 */
2908 if (convert_to_git(&the_index, s->path, s->data, s->size, &buf, crlf_warn)) {
2909 size_t size = 0;
2910 munmap(s->data, s->size);
2911 s->should_munmap = 0;
2912 s->data = strbuf_detach(&buf, &size);
2913 s->size = size;
2914 s->should_free = 1;
2915 }
2916 }
2917 else {
2918 enum object_type type;
2919 if (size_only || (flags & CHECK_BINARY)) {
2920 type = sha1_object_info(s->oid.hash, &s->size);
2921 if (type < 0)
2922 die("unable to read %s",
2923 oid_to_hex(&s->oid));
2924 if (size_only)
2925 return 0;
2926 if (s->size > big_file_threshold && s->is_binary == -1) {
2927 s->is_binary = 1;
2928 return 0;
2929 }
2930 }
2931 s->data = read_sha1_file(s->oid.hash, &type, &s->size);
2932 if (!s->data)
2933 die("unable to read %s", oid_to_hex(&s->oid));
2934 s->should_free = 1;
2935 }
2936 return 0;
2937 }
2938
2939 void diff_free_filespec_blob(struct diff_filespec *s)
2940 {
2941 if (s->should_free)
2942 free(s->data);
2943 else if (s->should_munmap)
2944 munmap(s->data, s->size);
2945
2946 if (s->should_free || s->should_munmap) {
2947 s->should_free = s->should_munmap = 0;
2948 s->data = NULL;
2949 }
2950 }
2951
2952 void diff_free_filespec_data(struct diff_filespec *s)
2953 {
2954 diff_free_filespec_blob(s);
2955 free(s->cnt_data);
2956 s->cnt_data = NULL;
2957 }
2958
2959 static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2960 void *blob,
2961 unsigned long size,
2962 const struct object_id *oid,
2963 int mode)
2964 {
2965 int fd;
2966 struct strbuf buf = STRBUF_INIT;
2967 struct strbuf template = STRBUF_INIT;
2968 char *path_dup = xstrdup(path);
2969 const char *base = basename(path_dup);
2970
2971 /* Generate "XXXXXX_basename.ext" */
2972 strbuf_addstr(&template, "XXXXXX_");
2973 strbuf_addstr(&template, base);
2974
2975 fd = mks_tempfile_ts(&temp->tempfile, template.buf, strlen(base) + 1);
2976 if (fd < 0)
2977 die_errno("unable to create temp-file");
2978 if (convert_to_working_tree(path,
2979 (const char *)blob, (size_t)size, &buf)) {
2980 blob = buf.buf;
2981 size = buf.len;
2982 }
2983 if (write_in_full(fd, blob, size) != size)
2984 die_errno("unable to write temp-file");
2985 close_tempfile(&temp->tempfile);
2986 temp->name = get_tempfile_path(&temp->tempfile);
2987 oid_to_hex_r(temp->hex, oid);
2988 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", mode);
2989 strbuf_release(&buf);
2990 strbuf_release(&template);
2991 free(path_dup);
2992 }
2993
2994 static struct diff_tempfile *prepare_temp_file(const char *name,
2995 struct diff_filespec *one)
2996 {
2997 struct diff_tempfile *temp = claim_diff_tempfile();
2998
2999 if (!DIFF_FILE_VALID(one)) {
3000 not_a_valid_file:
3001 /* A '-' entry produces this for file-2, and
3002 * a '+' entry produces this for file-1.
3003 */
3004 temp->name = "/dev/null";
3005 xsnprintf(temp->hex, sizeof(temp->hex), ".");
3006 xsnprintf(temp->mode, sizeof(temp->mode), ".");
3007 return temp;
3008 }
3009
3010 if (!S_ISGITLINK(one->mode) &&
3011 (!one->oid_valid ||
3012 reuse_worktree_file(name, &one->oid, 1))) {
3013 struct stat st;
3014 if (lstat(name, &st) < 0) {
3015 if (errno == ENOENT)
3016 goto not_a_valid_file;
3017 die_errno("stat(%s)", name);
3018 }
3019 if (S_ISLNK(st.st_mode)) {
3020 struct strbuf sb = STRBUF_INIT;
3021 if (strbuf_readlink(&sb, name, st.st_size) < 0)
3022 die_errno("readlink(%s)", name);
3023 prep_temp_blob(name, temp, sb.buf, sb.len,
3024 (one->oid_valid ?
3025 &one->oid : &null_oid),
3026 (one->oid_valid ?
3027 one->mode : S_IFLNK));
3028 strbuf_release(&sb);
3029 }
3030 else {
3031 /* we can borrow from the file in the work tree */
3032 temp->name = name;
3033 if (!one->oid_valid)
3034 oid_to_hex_r(temp->hex, &null_oid);
3035 else
3036 oid_to_hex_r(temp->hex, &one->oid);
3037 /* Even though we may sometimes borrow the
3038 * contents from the work tree, we always want
3039 * one->mode. mode is trustworthy even when
3040 * !(one->oid_valid), as long as
3041 * DIFF_FILE_VALID(one).
3042 */
3043 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", one->mode);
3044 }
3045 return temp;
3046 }
3047 else {
3048 if (diff_populate_filespec(one, 0))
3049 die("cannot read data blob for %s", one->path);
3050 prep_temp_blob(name, temp, one->data, one->size,
3051 &one->oid, one->mode);
3052 }
3053 return temp;
3054 }
3055
3056 static void add_external_diff_name(struct argv_array *argv,
3057 const char *name,
3058 struct diff_filespec *df)
3059 {
3060 struct diff_tempfile *temp = prepare_temp_file(name, df);
3061 argv_array_push(argv, temp->name);
3062 argv_array_push(argv, temp->hex);
3063 argv_array_push(argv, temp->mode);
3064 }
3065
3066 /* An external diff command takes:
3067 *
3068 * diff-cmd name infile1 infile1-sha1 infile1-mode \
3069 * infile2 infile2-sha1 infile2-mode [ rename-to ]
3070 *
3071 */
3072 static void run_external_diff(const char *pgm,
3073 const char *name,
3074 const char *other,
3075 struct diff_filespec *one,
3076 struct diff_filespec *two,
3077 const char *xfrm_msg,
3078 int complete_rewrite,
3079 struct diff_options *o)
3080 {
3081 struct argv_array argv = ARGV_ARRAY_INIT;
3082 struct argv_array env = ARGV_ARRAY_INIT;
3083 struct diff_queue_struct *q = &diff_queued_diff;
3084
3085 argv_array_push(&argv, pgm);
3086 argv_array_push(&argv, name);
3087
3088 if (one && two) {
3089 add_external_diff_name(&argv, name, one);
3090 if (!other)
3091 add_external_diff_name(&argv, name, two);
3092 else {
3093 add_external_diff_name(&argv, other, two);
3094 argv_array_push(&argv, other);
3095 argv_array_push(&argv, xfrm_msg);
3096 }
3097 }
3098
3099 argv_array_pushf(&env, "GIT_DIFF_PATH_COUNTER=%d", ++o->diff_path_counter);
3100 argv_array_pushf(&env, "GIT_DIFF_PATH_TOTAL=%d", q->nr);
3101
3102 if (run_command_v_opt_cd_env(argv.argv, RUN_USING_SHELL, NULL, env.argv))
3103 die(_("external diff died, stopping at %s"), name);
3104
3105 remove_tempfile();
3106 argv_array_clear(&argv);
3107 argv_array_clear(&env);
3108 }
3109
3110 static int similarity_index(struct diff_filepair *p)
3111 {
3112 return p->score * 100 / MAX_SCORE;
3113 }
3114
3115 static const char *diff_abbrev_oid(const struct object_id *oid, int abbrev)
3116 {
3117 if (startup_info->have_repository)
3118 return find_unique_abbrev(oid->hash, abbrev);
3119 else {
3120 char *hex = oid_to_hex(oid);
3121 if (abbrev < 0)
3122 abbrev = FALLBACK_DEFAULT_ABBREV;
3123 if (abbrev > GIT_SHA1_HEXSZ)
3124 die("BUG: oid abbreviation out of range: %d", abbrev);
3125 if (abbrev)
3126 hex[abbrev] = '\0';
3127 return hex;
3128 }
3129 }
3130
3131 static void fill_metainfo(struct strbuf *msg,
3132 const char *name,
3133 const char *other,
3134 struct diff_filespec *one,
3135 struct diff_filespec *two,
3136 struct diff_options *o,
3137 struct diff_filepair *p,
3138 int *must_show_header,
3139 int use_color)
3140 {
3141 const char *set = diff_get_color(use_color, DIFF_METAINFO);
3142 const char *reset = diff_get_color(use_color, DIFF_RESET);
3143 const char *line_prefix = diff_line_prefix(o);
3144
3145 *must_show_header = 1;
3146 strbuf_init(msg, PATH_MAX * 2 + 300);
3147 switch (p->status) {
3148 case DIFF_STATUS_COPIED:
3149 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3150 line_prefix, set, similarity_index(p));
3151 strbuf_addf(msg, "%s\n%s%scopy from ",
3152 reset, line_prefix, set);
3153 quote_c_style(name, msg, NULL, 0);
3154 strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
3155 quote_c_style(other, msg, NULL, 0);
3156 strbuf_addf(msg, "%s\n", reset);
3157 break;
3158 case DIFF_STATUS_RENAMED:
3159 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3160 line_prefix, set, similarity_index(p));
3161 strbuf_addf(msg, "%s\n%s%srename from ",
3162 reset, line_prefix, set);
3163 quote_c_style(name, msg, NULL, 0);
3164 strbuf_addf(msg, "%s\n%s%srename to ",
3165 reset, line_prefix, set);
3166 quote_c_style(other, msg, NULL, 0);
3167 strbuf_addf(msg, "%s\n", reset);
3168 break;
3169 case DIFF_STATUS_MODIFIED:
3170 if (p->score) {
3171 strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
3172 line_prefix,
3173 set, similarity_index(p), reset);
3174 break;
3175 }
3176 /* fallthru */
3177 default:
3178 *must_show_header = 0;
3179 }
3180 if (one && two && oidcmp(&one->oid, &two->oid)) {
3181 int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
3182
3183 if (DIFF_OPT_TST(o, BINARY)) {
3184 mmfile_t mf;
3185 if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
3186 (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
3187 abbrev = 40;
3188 }
3189 strbuf_addf(msg, "%s%sindex %s..%s", line_prefix, set,
3190 diff_abbrev_oid(&one->oid, abbrev),
3191 diff_abbrev_oid(&two->oid, abbrev));
3192 if (one->mode == two->mode)
3193 strbuf_addf(msg, " %06o", one->mode);
3194 strbuf_addf(msg, "%s\n", reset);
3195 }
3196 }
3197
3198 static void run_diff_cmd(const char *pgm,
3199 const char *name,
3200 const char *other,
3201 const char *attr_path,
3202 struct diff_filespec *one,
3203 struct diff_filespec *two,
3204 struct strbuf *msg,
3205 struct diff_options *o,
3206 struct diff_filepair *p)
3207 {
3208 const char *xfrm_msg = NULL;
3209 int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
3210 int must_show_header = 0;
3211
3212
3213 if (DIFF_OPT_TST(o, ALLOW_EXTERNAL)) {
3214 struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
3215 if (drv && drv->external)
3216 pgm = drv->external;
3217 }
3218
3219 if (msg) {
3220 /*
3221 * don't use colors when the header is intended for an
3222 * external diff driver
3223 */
3224 fill_metainfo(msg, name, other, one, two, o, p,
3225 &must_show_header,
3226 want_color(o->use_color) && !pgm);
3227 xfrm_msg = msg->len ? msg->buf : NULL;
3228 }
3229
3230 if (pgm) {
3231 run_external_diff(pgm, name, other, one, two, xfrm_msg,
3232 complete_rewrite, o);
3233 return;
3234 }
3235 if (one && two)
3236 builtin_diff(name, other ? other : name,
3237 one, two, xfrm_msg, must_show_header,
3238 o, complete_rewrite);
3239 else
3240 fprintf(o->file, "* Unmerged path %s\n", name);
3241 }
3242
3243 static void diff_fill_oid_info(struct diff_filespec *one)
3244 {
3245 if (DIFF_FILE_VALID(one)) {
3246 if (!one->oid_valid) {
3247 struct stat st;
3248 if (one->is_stdin) {
3249 oidclr(&one->oid);
3250 return;
3251 }
3252 if (lstat(one->path, &st) < 0)
3253 die_errno("stat '%s'", one->path);
3254 if (index_path(one->oid.hash, one->path, &st, 0))
3255 die("cannot hash %s", one->path);
3256 }
3257 }
3258 else
3259 oidclr(&one->oid);
3260 }
3261
3262 static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
3263 {
3264 /* Strip the prefix but do not molest /dev/null and absolute paths */
3265 if (*namep && **namep != '/') {
3266 *namep += prefix_length;
3267 if (**namep == '/')
3268 ++*namep;
3269 }
3270 if (*otherp && **otherp != '/') {
3271 *otherp += prefix_length;
3272 if (**otherp == '/')
3273 ++*otherp;
3274 }
3275 }
3276
3277 static void run_diff(struct diff_filepair *p, struct diff_options *o)
3278 {
3279 const char *pgm = external_diff();
3280 struct strbuf msg;
3281 struct diff_filespec *one = p->one;
3282 struct diff_filespec *two = p->two;
3283 const char *name;
3284 const char *other;
3285 const char *attr_path;
3286
3287 name = p->one->path;
3288 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3289 attr_path = name;
3290 if (o->prefix_length)
3291 strip_prefix(o->prefix_length, &name, &other);
3292
3293 if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
3294 pgm = NULL;
3295
3296 if (DIFF_PAIR_UNMERGED(p)) {
3297 run_diff_cmd(pgm, name, NULL, attr_path,
3298 NULL, NULL, NULL, o, p);
3299 return;
3300 }
3301
3302 diff_fill_oid_info(one);
3303 diff_fill_oid_info(two);
3304
3305 if (!pgm &&
3306 DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3307 (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3308 /*
3309 * a filepair that changes between file and symlink
3310 * needs to be split into deletion and creation.
3311 */
3312 struct diff_filespec *null = alloc_filespec(two->path);
3313 run_diff_cmd(NULL, name, other, attr_path,
3314 one, null, &msg, o, p);
3315 free(null);
3316 strbuf_release(&msg);
3317
3318 null = alloc_filespec(one->path);
3319 run_diff_cmd(NULL, name, other, attr_path,
3320 null, two, &msg, o, p);
3321 free(null);
3322 }
3323 else
3324 run_diff_cmd(pgm, name, other, attr_path,
3325 one, two, &msg, o, p);
3326
3327 strbuf_release(&msg);
3328 }
3329
3330 static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3331 struct diffstat_t *diffstat)
3332 {
3333 const char *name;
3334 const char *other;
3335
3336 if (DIFF_PAIR_UNMERGED(p)) {
3337 /* unmerged */
3338 builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, p);
3339 return;
3340 }
3341
3342 name = p->one->path;
3343 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3344
3345 if (o->prefix_length)
3346 strip_prefix(o->prefix_length, &name, &other);
3347
3348 diff_fill_oid_info(p->one);
3349 diff_fill_oid_info(p->two);
3350
3351 builtin_diffstat(name, other, p->one, p->two, diffstat, o, p);
3352 }
3353
3354 static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3355 {
3356 const char *name;
3357 const char *other;
3358 const char *attr_path;
3359
3360 if (DIFF_PAIR_UNMERGED(p)) {
3361 /* unmerged */
3362 return;
3363 }
3364
3365 name = p->one->path;
3366 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3367 attr_path = other ? other : name;
3368
3369 if (o->prefix_length)
3370 strip_prefix(o->prefix_length, &name, &other);
3371
3372 diff_fill_oid_info(p->one);
3373 diff_fill_oid_info(p->two);
3374
3375 builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3376 }
3377
3378 void diff_setup(struct diff_options *options)
3379 {
3380 memcpy(options, &default_diff_options, sizeof(*options));
3381
3382 options->file = stdout;
3383
3384 options->abbrev = DEFAULT_ABBREV;
3385 options->line_termination = '\n';
3386 options->break_opt = -1;
3387 options->rename_limit = -1;
3388 options->dirstat_permille = diff_dirstat_permille_default;
3389 options->context = diff_context_default;
3390 options->interhunkcontext = diff_interhunk_context_default;
3391 options->ws_error_highlight = ws_error_highlight_default;
3392 DIFF_OPT_SET(options, RENAME_EMPTY);
3393
3394 /* pathchange left =NULL by default */
3395 options->change = diff_change;
3396 options->add_remove = diff_addremove;
3397 options->use_color = diff_use_color_default;
3398 options->detect_rename = diff_detect_rename_default;
3399 options->xdl_opts |= diff_algorithm;
3400 if (diff_indent_heuristic)
3401 DIFF_XDL_SET(options, INDENT_HEURISTIC);
3402
3403 options->orderfile = diff_order_file_cfg;
3404
3405 if (diff_no_prefix) {
3406 options->a_prefix = options->b_prefix = "";
3407 } else if (!diff_mnemonic_prefix) {
3408 options->a_prefix = "a/";
3409 options->b_prefix = "b/";
3410 }
3411 }
3412
3413 void diff_setup_done(struct diff_options *options)
3414 {
3415 int count = 0;
3416
3417 if (options->set_default)
3418 options->set_default(options);
3419
3420 if (options->output_format & DIFF_FORMAT_NAME)
3421 count++;
3422 if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3423 count++;
3424 if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3425 count++;
3426 if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3427 count++;
3428 if (count > 1)
3429 die(_("--name-only, --name-status, --check and -s are mutually exclusive"));
3430
3431 /*
3432 * Most of the time we can say "there are changes"
3433 * only by checking if there are changed paths, but
3434 * --ignore-whitespace* options force us to look
3435 * inside contents.
3436 */
3437
3438 if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3439 DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3440 DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3441 DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3442 else
3443 DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3444
3445 if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3446 options->detect_rename = DIFF_DETECT_COPY;
3447
3448 if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3449 options->prefix = NULL;
3450 if (options->prefix)
3451 options->prefix_length = strlen(options->prefix);
3452 else
3453 options->prefix_length = 0;
3454
3455 if (options->output_format & (DIFF_FORMAT_NAME |
3456 DIFF_FORMAT_NAME_STATUS |
3457 DIFF_FORMAT_CHECKDIFF |
3458 DIFF_FORMAT_NO_OUTPUT))
3459 options->output_format &= ~(DIFF_FORMAT_RAW |
3460 DIFF_FORMAT_NUMSTAT |
3461 DIFF_FORMAT_DIFFSTAT |
3462 DIFF_FORMAT_SHORTSTAT |
3463 DIFF_FORMAT_DIRSTAT |
3464 DIFF_FORMAT_SUMMARY |
3465 DIFF_FORMAT_PATCH);
3466
3467 /*
3468 * These cases always need recursive; we do not drop caller-supplied
3469 * recursive bits for other formats here.
3470 */
3471 if (options->output_format & (DIFF_FORMAT_PATCH |
3472 DIFF_FORMAT_NUMSTAT |
3473 DIFF_FORMAT_DIFFSTAT |
3474 DIFF_FORMAT_SHORTSTAT |
3475 DIFF_FORMAT_DIRSTAT |
3476 DIFF_FORMAT_SUMMARY |
3477 DIFF_FORMAT_CHECKDIFF))
3478 DIFF_OPT_SET(options, RECURSIVE);
3479 /*
3480 * Also pickaxe would not work very well if you do not say recursive
3481 */
3482 if (options->pickaxe)
3483 DIFF_OPT_SET(options, RECURSIVE);
3484 /*
3485 * When patches are generated, submodules diffed against the work tree
3486 * must be checked for dirtiness too so it can be shown in the output
3487 */
3488 if (options->output_format & DIFF_FORMAT_PATCH)
3489 DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3490
3491 if (options->detect_rename && options->rename_limit < 0)
3492 options->rename_limit = diff_rename_limit_default;
3493 if (options->setup & DIFF_SETUP_USE_CACHE) {
3494 if (!active_cache)
3495 /* read-cache does not die even when it fails
3496 * so it is safe for us to do this here. Also
3497 * it does not smudge active_cache or active_nr
3498 * when it fails, so we do not have to worry about
3499 * cleaning it up ourselves either.
3500 */
3501 read_cache();
3502 }
3503 if (40 < options->abbrev)
3504 options->abbrev = 40; /* full */
3505
3506 /*
3507 * It does not make sense to show the first hit we happened
3508 * to have found. It does not make sense not to return with
3509 * exit code in such a case either.
3510 */
3511 if (DIFF_OPT_TST(options, QUICK)) {
3512 options->output_format = DIFF_FORMAT_NO_OUTPUT;
3513 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3514 }
3515
3516 options->diff_path_counter = 0;
3517
3518 if (DIFF_OPT_TST(options, FOLLOW_RENAMES) && options->pathspec.nr != 1)
3519 die(_("--follow requires exactly one pathspec"));
3520 }
3521
3522 static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3523 {
3524 char c, *eq;
3525 int len;
3526
3527 if (*arg != '-')
3528 return 0;
3529 c = *++arg;
3530 if (!c)
3531 return 0;
3532 if (c == arg_short) {
3533 c = *++arg;
3534 if (!c)
3535 return 1;
3536 if (val && isdigit(c)) {
3537 char *end;
3538 int n = strtoul(arg, &end, 10);
3539 if (*end)
3540 return 0;
3541 *val = n;
3542 return 1;
3543 }
3544 return 0;
3545 }
3546 if (c != '-')
3547 return 0;
3548 arg++;
3549 eq = strchrnul(arg, '=');
3550 len = eq - arg;
3551 if (!len || strncmp(arg, arg_long, len))
3552 return 0;
3553 if (*eq) {
3554 int n;
3555 char *end;
3556 if (!isdigit(*++eq))
3557 return 0;
3558 n = strtoul(eq, &end, 10);
3559 if (*end)
3560 return 0;
3561 *val = n;
3562 }
3563 return 1;
3564 }
3565
3566 static int diff_scoreopt_parse(const char *opt);
3567
3568 static inline int short_opt(char opt, const char **argv,
3569 const char **optarg)
3570 {
3571 const char *arg = argv[0];
3572 if (arg[0] != '-' || arg[1] != opt)
3573 return 0;
3574 if (arg[2] != '\0') {
3575 *optarg = arg + 2;
3576 return 1;
3577 }
3578 if (!argv[1])
3579 die("Option '%c' requires a value", opt);
3580 *optarg = argv[1];
3581 return 2;
3582 }
3583
3584 int parse_long_opt(const char *opt, const char **argv,
3585 const char **optarg)
3586 {
3587 const char *arg = argv[0];
3588 if (!skip_prefix(arg, "--", &arg))
3589 return 0;
3590 if (!skip_prefix(arg, opt, &arg))
3591 return 0;
3592 if (*arg == '=') { /* stuck form: --option=value */
3593 *optarg = arg + 1;
3594 return 1;
3595 }
3596 if (*arg != '\0')
3597 return 0;
3598 /* separate form: --option value */
3599 if (!argv[1])
3600 die("Option '--%s' requires a value", opt);
3601 *optarg = argv[1];
3602 return 2;
3603 }
3604
3605 static int stat_opt(struct diff_options *options, const char **av)
3606 {
3607 const char *arg = av[0];
3608 char *end;
3609 int width = options->stat_width;
3610 int name_width = options->stat_name_width;
3611 int graph_width = options->stat_graph_width;
3612 int count = options->stat_count;
3613 int argcount = 1;
3614
3615 if (!skip_prefix(arg, "--stat", &arg))
3616 die("BUG: stat option does not begin with --stat: %s", arg);
3617 end = (char *)arg;
3618
3619 switch (*arg) {
3620 case '-':
3621 if (skip_prefix(arg, "-width", &arg)) {
3622 if (*arg == '=')
3623 width = strtoul(arg + 1, &end, 10);
3624 else if (!*arg && !av[1])
3625 die_want_option("--stat-width");
3626 else if (!*arg) {
3627 width = strtoul(av[1], &end, 10);
3628 argcount = 2;
3629 }
3630 } else if (skip_prefix(arg, "-name-width", &arg)) {
3631 if (*arg == '=')
3632 name_width = strtoul(arg + 1, &end, 10);
3633 else if (!*arg && !av[1])
3634 die_want_option("--stat-name-width");
3635 else if (!*arg) {
3636 name_width = strtoul(av[1], &end, 10);
3637 argcount = 2;
3638 }
3639 } else if (skip_prefix(arg, "-graph-width", &arg)) {
3640 if (*arg == '=')
3641 graph_width = strtoul(arg + 1, &end, 10);
3642 else if (!*arg && !av[1])
3643 die_want_option("--stat-graph-width");
3644 else if (!*arg) {
3645 graph_width = strtoul(av[1], &end, 10);
3646 argcount = 2;
3647 }
3648 } else if (skip_prefix(arg, "-count", &arg)) {
3649 if (*arg == '=')
3650 count = strtoul(arg + 1, &end, 10);
3651 else if (!*arg && !av[1])
3652 die_want_option("--stat-count");
3653 else if (!*arg) {
3654 count = strtoul(av[1], &end, 10);
3655 argcount = 2;
3656 }
3657 }
3658 break;
3659 case '=':
3660 width = strtoul(arg+1, &end, 10);
3661 if (*end == ',')
3662 name_width = strtoul(end+1, &end, 10);
3663 if (*end == ',')
3664 count = strtoul(end+1, &end, 10);
3665 }
3666
3667 /* Important! This checks all the error cases! */
3668 if (*end)
3669 return 0;
3670 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3671 options->stat_name_width = name_width;
3672 options->stat_graph_width = graph_width;
3673 options->stat_width = width;
3674 options->stat_count = count;
3675 return argcount;
3676 }
3677
3678 static int parse_dirstat_opt(struct diff_options *options, const char *params)
3679 {
3680 struct strbuf errmsg = STRBUF_INIT;
3681 if (parse_dirstat_params(options, params, &errmsg))
3682 die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3683 errmsg.buf);
3684 strbuf_release(&errmsg);
3685 /*
3686 * The caller knows a dirstat-related option is given from the command
3687 * line; allow it to say "return this_function();"
3688 */
3689 options->output_format |= DIFF_FORMAT_DIRSTAT;
3690 return 1;
3691 }
3692
3693 static int parse_submodule_opt(struct diff_options *options, const char *value)
3694 {
3695 if (parse_submodule_params(options, value))
3696 die(_("Failed to parse --submodule option parameter: '%s'"),
3697 value);
3698 return 1;
3699 }
3700
3701 static const char diff_status_letters[] = {
3702 DIFF_STATUS_ADDED,
3703 DIFF_STATUS_COPIED,
3704 DIFF_STATUS_DELETED,
3705 DIFF_STATUS_MODIFIED,
3706 DIFF_STATUS_RENAMED,
3707 DIFF_STATUS_TYPE_CHANGED,
3708 DIFF_STATUS_UNKNOWN,
3709 DIFF_STATUS_UNMERGED,
3710 DIFF_STATUS_FILTER_AON,
3711 DIFF_STATUS_FILTER_BROKEN,
3712 '\0',
3713 };
3714
3715 static unsigned int filter_bit['Z' + 1];
3716
3717 static void prepare_filter_bits(void)
3718 {
3719 int i;
3720
3721 if (!filter_bit[DIFF_STATUS_ADDED]) {
3722 for (i = 0; diff_status_letters[i]; i++)
3723 filter_bit[(int) diff_status_letters[i]] = (1 << i);
3724 }
3725 }
3726
3727 static unsigned filter_bit_tst(char status, const struct diff_options *opt)
3728 {
3729 return opt->filter & filter_bit[(int) status];
3730 }
3731
3732 static int parse_diff_filter_opt(const char *optarg, struct diff_options *opt)
3733 {
3734 int i, optch;
3735
3736 prepare_filter_bits();
3737
3738 /*
3739 * If there is a negation e.g. 'd' in the input, and we haven't
3740 * initialized the filter field with another --diff-filter, start
3741 * from full set of bits, except for AON.
3742 */
3743 if (!opt->filter) {
3744 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3745 if (optch < 'a' || 'z' < optch)
3746 continue;
3747 opt->filter = (1 << (ARRAY_SIZE(diff_status_letters) - 1)) - 1;
3748 opt->filter &= ~filter_bit[DIFF_STATUS_FILTER_AON];
3749 break;
3750 }
3751 }
3752
3753 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3754 unsigned int bit;
3755 int negate;
3756
3757 if ('a' <= optch && optch <= 'z') {
3758 negate = 1;
3759 optch = toupper(optch);
3760 } else {
3761 negate = 0;
3762 }
3763
3764 bit = (0 <= optch && optch <= 'Z') ? filter_bit[optch] : 0;
3765 if (!bit)
3766 return optarg[i];
3767 if (negate)
3768 opt->filter &= ~bit;
3769 else
3770 opt->filter |= bit;
3771 }
3772 return 0;
3773 }
3774
3775 static void enable_patch_output(int *fmt) {
3776 *fmt &= ~DIFF_FORMAT_NO_OUTPUT;
3777 *fmt |= DIFF_FORMAT_PATCH;
3778 }
3779
3780 static int parse_ws_error_highlight_opt(struct diff_options *opt, const char *arg)
3781 {
3782 int val = parse_ws_error_highlight(arg);
3783
3784 if (val < 0) {
3785 error("unknown value after ws-error-highlight=%.*s",
3786 -1 - val, arg);
3787 return 0;
3788 }
3789 opt->ws_error_highlight = val;
3790 return 1;
3791 }
3792
3793 int diff_opt_parse(struct diff_options *options,
3794 const char **av, int ac, const char *prefix)
3795 {
3796 const char *arg = av[0];
3797 const char *optarg;
3798 int argcount;
3799
3800 if (!prefix)
3801 prefix = "";
3802
3803 /* Output format options */
3804 if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch")
3805 || opt_arg(arg, 'U', "unified", &options->context))
3806 enable_patch_output(&options->output_format);
3807 else if (!strcmp(arg, "--raw"))
3808 options->output_format |= DIFF_FORMAT_RAW;
3809 else if (!strcmp(arg, "--patch-with-raw")) {
3810 enable_patch_output(&options->output_format);
3811 options->output_format |= DIFF_FORMAT_RAW;
3812 } else if (!strcmp(arg, "--numstat"))
3813 options->output_format |= DIFF_FORMAT_NUMSTAT;
3814 else if (!strcmp(arg, "--shortstat"))
3815 options->output_format |= DIFF_FORMAT_SHORTSTAT;
3816 else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3817 return parse_dirstat_opt(options, "");
3818 else if (skip_prefix(arg, "-X", &arg))
3819 return parse_dirstat_opt(options, arg);
3820 else if (skip_prefix(arg, "--dirstat=", &arg))
3821 return parse_dirstat_opt(options, arg);
3822 else if (!strcmp(arg, "--cumulative"))
3823 return parse_dirstat_opt(options, "cumulative");
3824 else if (!strcmp(arg, "--dirstat-by-file"))
3825 return parse_dirstat_opt(options, "files");
3826 else if (skip_prefix(arg, "--dirstat-by-file=", &arg)) {
3827 parse_dirstat_opt(options, "files");
3828 return parse_dirstat_opt(options, arg);
3829 }
3830 else if (!strcmp(arg, "--check"))
3831 options->output_format |= DIFF_FORMAT_CHECKDIFF;
3832 else if (!strcmp(arg, "--summary"))
3833 options->output_format |= DIFF_FORMAT_SUMMARY;
3834 else if (!strcmp(arg, "--patch-with-stat")) {
3835 enable_patch_output(&options->output_format);
3836 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3837 } else if (!strcmp(arg, "--name-only"))
3838 options->output_format |= DIFF_FORMAT_NAME;
3839 else if (!strcmp(arg, "--name-status"))
3840 options->output_format |= DIFF_FORMAT_NAME_STATUS;
3841 else if (!strcmp(arg, "-s") || !strcmp(arg, "--no-patch"))
3842 options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3843 else if (starts_with(arg, "--stat"))
3844 /* --stat, --stat-width, --stat-name-width, or --stat-count */
3845 return stat_opt(options, av);
3846
3847 /* renames options */
3848 else if (starts_with(arg, "-B") || starts_with(arg, "--break-rewrites=") ||
3849 !strcmp(arg, "--break-rewrites")) {
3850 if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3851 return error("invalid argument to -B: %s", arg+2);
3852 }
3853 else if (starts_with(arg, "-M") || starts_with(arg, "--find-renames=") ||
3854 !strcmp(arg, "--find-renames")) {
3855 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3856 return error("invalid argument to -M: %s", arg+2);
3857 options->detect_rename = DIFF_DETECT_RENAME;
3858 }
3859 else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3860 options->irreversible_delete = 1;
3861 }
3862 else if (starts_with(arg, "-C") || starts_with(arg, "--find-copies=") ||
3863 !strcmp(arg, "--find-copies")) {
3864 if (options->detect_rename == DIFF_DETECT_COPY)
3865 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3866 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3867 return error("invalid argument to -C: %s", arg+2);
3868 options->detect_rename = DIFF_DETECT_COPY;
3869 }
3870 else if (!strcmp(arg, "--no-renames"))
3871 options->detect_rename = 0;
3872 else if (!strcmp(arg, "--rename-empty"))
3873 DIFF_OPT_SET(options, RENAME_EMPTY);
3874 else if (!strcmp(arg, "--no-rename-empty"))
3875 DIFF_OPT_CLR(options, RENAME_EMPTY);
3876 else if (!strcmp(arg, "--relative"))
3877 DIFF_OPT_SET(options, RELATIVE_NAME);
3878 else if (skip_prefix(arg, "--relative=", &arg)) {
3879 DIFF_OPT_SET(options, RELATIVE_NAME);
3880 options->prefix = arg;
3881 }
3882
3883 /* xdiff options */
3884 else if (!strcmp(arg, "--minimal"))
3885 DIFF_XDL_SET(options, NEED_MINIMAL);
3886 else if (!strcmp(arg, "--no-minimal"))
3887 DIFF_XDL_CLR(options, NEED_MINIMAL);
3888 else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3889 DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3890 else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3891 DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3892 else if (!strcmp(arg, "--ignore-space-at-eol"))
3893 DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3894 else if (!strcmp(arg, "--ignore-blank-lines"))
3895 DIFF_XDL_SET(options, IGNORE_BLANK_LINES);
3896 else if (!strcmp(arg, "--indent-heuristic"))
3897 DIFF_XDL_SET(options, INDENT_HEURISTIC);
3898 else if (!strcmp(arg, "--no-indent-heuristic"))
3899 DIFF_XDL_CLR(options, INDENT_HEURISTIC);
3900 else if (!strcmp(arg, "--patience"))
3901 options->xdl_opts = DIFF_WITH_ALG(options, PATIENCE_DIFF);
3902 else if (!strcmp(arg, "--histogram"))
3903 options->xdl_opts = DIFF_WITH_ALG(options, HISTOGRAM_DIFF);
3904 else if ((argcount = parse_long_opt("diff-algorithm", av, &optarg))) {
3905 long value = parse_algorithm_value(optarg);
3906 if (value < 0)
3907 return error("option diff-algorithm accepts \"myers\", "
3908 "\"minimal\", \"patience\" and \"histogram\"");
3909 /* clear out previous settings */
3910 DIFF_XDL_CLR(options, NEED_MINIMAL);
3911 options->xdl_opts &= ~XDF_DIFF_ALGORITHM_MASK;
3912 options->xdl_opts |= value;
3913 return argcount;
3914 }
3915
3916 /* flags options */
3917 else if (!strcmp(arg, "--binary")) {
3918 enable_patch_output(&options->output_format);
3919 DIFF_OPT_SET(options, BINARY);
3920 }
3921 else if (!strcmp(arg, "--full-index"))
3922 DIFF_OPT_SET(options, FULL_INDEX);
3923 else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
3924 DIFF_OPT_SET(options, TEXT);
3925 else if (!strcmp(arg, "-R"))
3926 DIFF_OPT_SET(options, REVERSE_DIFF);
3927 else if (!strcmp(arg, "--find-copies-harder"))
3928 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3929 else if (!strcmp(arg, "--follow"))
3930 DIFF_OPT_SET(options, FOLLOW_RENAMES);
3931 else if (!strcmp(arg, "--no-follow")) {
3932 DIFF_OPT_CLR(options, FOLLOW_RENAMES);
3933 DIFF_OPT_CLR(options, DEFAULT_FOLLOW_RENAMES);
3934 } else if (!strcmp(arg, "--color"))
3935 options->use_color = 1;
3936 else if (skip_prefix(arg, "--color=", &arg)) {
3937 int value = git_config_colorbool(NULL, arg);
3938 if (value < 0)
3939 return error("option `color' expects \"always\", \"auto\", or \"never\"");
3940 options->use_color = value;
3941 }
3942 else if (!strcmp(arg, "--no-color"))
3943 options->use_color = 0;
3944 else if (!strcmp(arg, "--color-words")) {
3945 options->use_color = 1;
3946 options->word_diff = DIFF_WORDS_COLOR;
3947 }
3948 else if (skip_prefix(arg, "--color-words=", &arg)) {
3949 options->use_color = 1;
3950 options->word_diff = DIFF_WORDS_COLOR;
3951 options->word_regex = arg;
3952 }
3953 else if (!strcmp(arg, "--word-diff")) {
3954 if (options->word_diff == DIFF_WORDS_NONE)
3955 options->word_diff = DIFF_WORDS_PLAIN;
3956 }
3957 else if (skip_prefix(arg, "--word-diff=", &arg)) {
3958 if (!strcmp(arg, "plain"))
3959 options->word_diff = DIFF_WORDS_PLAIN;
3960 else if (!strcmp(arg, "color")) {
3961 options->use_color = 1;
3962 options->word_diff = DIFF_WORDS_COLOR;
3963 }
3964 else if (!strcmp(arg, "porcelain"))
3965 options->word_diff = DIFF_WORDS_PORCELAIN;
3966 else if (!strcmp(arg, "none"))
3967 options->word_diff = DIFF_WORDS_NONE;
3968 else
3969 die("bad --word-diff argument: %s", arg);
3970 }
3971 else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
3972 if (options->word_diff == DIFF_WORDS_NONE)
3973 options->word_diff = DIFF_WORDS_PLAIN;
3974 options->word_regex = optarg;
3975 return argcount;
3976 }
3977 else if (!strcmp(arg, "--exit-code"))
3978 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3979 else if (!strcmp(arg, "--quiet"))
3980 DIFF_OPT_SET(options, QUICK);
3981 else if (!strcmp(arg, "--ext-diff"))
3982 DIFF_OPT_SET(options, ALLOW_EXTERNAL);
3983 else if (!strcmp(arg, "--no-ext-diff"))
3984 DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
3985 else if (!strcmp(arg, "--textconv"))
3986 DIFF_OPT_SET(options, ALLOW_TEXTCONV);
3987 else if (!strcmp(arg, "--no-textconv"))
3988 DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
3989 else if (!strcmp(arg, "--ignore-submodules")) {
3990 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3991 handle_ignore_submodules_arg(options, "all");
3992 } else if (skip_prefix(arg, "--ignore-submodules=", &arg)) {
3993 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3994 handle_ignore_submodules_arg(options, arg);
3995 } else if (!strcmp(arg, "--submodule"))
3996 options->submodule_format = DIFF_SUBMODULE_LOG;
3997 else if (skip_prefix(arg, "--submodule=", &arg))
3998 return parse_submodule_opt(options, arg);
3999 else if (skip_prefix(arg, "--ws-error-highlight=", &arg))
4000 return parse_ws_error_highlight_opt(options, arg);
4001 else if (!strcmp(arg, "--ita-invisible-in-index"))
4002 options->ita_invisible_in_index = 1;
4003 else if (!strcmp(arg, "--ita-visible-in-index"))
4004 options->ita_invisible_in_index = 0;
4005
4006 /* misc options */
4007 else if (!strcmp(arg, "-z"))
4008 options->line_termination = 0;
4009 else if ((argcount = short_opt('l', av, &optarg))) {
4010 options->rename_limit = strtoul(optarg, NULL, 10);
4011 return argcount;
4012 }
4013 else if ((argcount = short_opt('S', av, &optarg))) {
4014 options->pickaxe = optarg;
4015 options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
4016 return argcount;
4017 } else if ((argcount = short_opt('G', av, &optarg))) {
4018 options->pickaxe = optarg;
4019 options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
4020 return argcount;
4021 }
4022 else if (!strcmp(arg, "--pickaxe-all"))
4023 options->pickaxe_opts |= DIFF_PICKAXE_ALL;
4024 else if (!strcmp(arg, "--pickaxe-regex"))
4025 options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
4026 else if ((argcount = short_opt('O', av, &optarg))) {
4027 options->orderfile = prefix_filename(prefix, optarg);
4028 return argcount;
4029 }
4030 else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
4031 int offending = parse_diff_filter_opt(optarg, options);
4032 if (offending)
4033 die("unknown change class '%c' in --diff-filter=%s",
4034 offending, optarg);
4035 return argcount;
4036 }
4037 else if (!strcmp(arg, "--no-abbrev"))
4038 options->abbrev = 0;
4039 else if (!strcmp(arg, "--abbrev"))
4040 options->abbrev = DEFAULT_ABBREV;
4041 else if (skip_prefix(arg, "--abbrev=", &arg)) {
4042 options->abbrev = strtoul(arg, NULL, 10);
4043 if (options->abbrev < MINIMUM_ABBREV)
4044 options->abbrev = MINIMUM_ABBREV;
4045 else if (40 < options->abbrev)
4046 options->abbrev = 40;
4047 }
4048 else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
4049 options->a_prefix = optarg;
4050 return argcount;
4051 }
4052 else if ((argcount = parse_long_opt("line-prefix", av, &optarg))) {
4053 options->line_prefix = optarg;
4054 options->line_prefix_length = strlen(options->line_prefix);
4055 graph_setup_line_prefix(options);
4056 return argcount;
4057 }
4058 else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
4059 options->b_prefix = optarg;
4060 return argcount;
4061 }
4062 else if (!strcmp(arg, "--no-prefix"))
4063 options->a_prefix = options->b_prefix = "";
4064 else if (opt_arg(arg, '\0', "inter-hunk-context",
4065 &options->interhunkcontext))
4066 ;
4067 else if (!strcmp(arg, "-W"))
4068 DIFF_OPT_SET(options, FUNCCONTEXT);
4069 else if (!strcmp(arg, "--function-context"))
4070 DIFF_OPT_SET(options, FUNCCONTEXT);
4071 else if (!strcmp(arg, "--no-function-context"))
4072 DIFF_OPT_CLR(options, FUNCCONTEXT);
4073 else if ((argcount = parse_long_opt("output", av, &optarg))) {
4074 char *path = prefix_filename(prefix, optarg);
4075 options->file = xfopen(path, "w");
4076 options->close_file = 1;
4077 if (options->use_color != GIT_COLOR_ALWAYS)
4078 options->use_color = GIT_COLOR_NEVER;
4079 free(path);
4080 return argcount;
4081 } else
4082 return 0;
4083 return 1;
4084 }
4085
4086 int parse_rename_score(const char **cp_p)
4087 {
4088 unsigned long num, scale;
4089 int ch, dot;
4090 const char *cp = *cp_p;
4091
4092 num = 0;
4093 scale = 1;
4094 dot = 0;
4095 for (;;) {
4096 ch = *cp;
4097 if ( !dot && ch == '.' ) {
4098 scale = 1;
4099 dot = 1;
4100 } else if ( ch == '%' ) {
4101 scale = dot ? scale*100 : 100;
4102 cp++; /* % is always at the end */
4103 break;
4104 } else if ( ch >= '0' && ch <= '9' ) {
4105 if ( scale < 100000 ) {
4106 scale *= 10;
4107 num = (num*10) + (ch-'0');
4108 }
4109 } else {
4110 break;
4111 }
4112 cp++;
4113 }
4114 *cp_p = cp;
4115
4116 /* user says num divided by scale and we say internally that
4117 * is MAX_SCORE * num / scale.
4118 */
4119 return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
4120 }
4121
4122 static int diff_scoreopt_parse(const char *opt)
4123 {
4124 int opt1, opt2, cmd;
4125
4126 if (*opt++ != '-')
4127 return -1;
4128 cmd = *opt++;
4129 if (cmd == '-') {
4130 /* convert the long-form arguments into short-form versions */
4131 if (skip_prefix(opt, "break-rewrites", &opt)) {
4132 if (*opt == 0 || *opt++ == '=')
4133 cmd = 'B';
4134 } else if (skip_prefix(opt, "find-copies", &opt)) {
4135 if (*opt == 0 || *opt++ == '=')
4136 cmd = 'C';
4137 } else if (skip_prefix(opt, "find-renames", &opt)) {
4138 if (*opt == 0 || *opt++ == '=')
4139 cmd = 'M';
4140 }
4141 }
4142 if (cmd != 'M' && cmd != 'C' && cmd != 'B')
4143 return -1; /* that is not a -M, -C, or -B option */
4144
4145 opt1 = parse_rename_score(&opt);
4146 if (cmd != 'B')
4147 opt2 = 0;
4148 else {
4149 if (*opt == 0)
4150 opt2 = 0;
4151 else if (*opt != '/')
4152 return -1; /* we expect -B80/99 or -B80 */
4153 else {
4154 opt++;
4155 opt2 = parse_rename_score(&opt);
4156 }
4157 }
4158 if (*opt != 0)
4159 return -1;
4160 return opt1 | (opt2 << 16);
4161 }
4162
4163 struct diff_queue_struct diff_queued_diff;
4164
4165 void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
4166 {
4167 ALLOC_GROW(queue->queue, queue->nr + 1, queue->alloc);
4168 queue->queue[queue->nr++] = dp;
4169 }
4170
4171 struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
4172 struct diff_filespec *one,
4173 struct diff_filespec *two)
4174 {
4175 struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
4176 dp->one = one;
4177 dp->two = two;
4178 if (queue)
4179 diff_q(queue, dp);
4180 return dp;
4181 }
4182
4183 void diff_free_filepair(struct diff_filepair *p)
4184 {
4185 free_filespec(p->one);
4186 free_filespec(p->two);
4187 free(p);
4188 }
4189
4190 const char *diff_aligned_abbrev(const struct object_id *oid, int len)
4191 {
4192 int abblen;
4193 const char *abbrev;
4194
4195 if (len == GIT_SHA1_HEXSZ)
4196 return oid_to_hex(oid);
4197
4198 abbrev = diff_abbrev_oid(oid, len);
4199 abblen = strlen(abbrev);
4200
4201 /*
4202 * In well-behaved cases, where the abbbreviated result is the
4203 * same as the requested length, append three dots after the
4204 * abbreviation (hence the whole logic is limited to the case
4205 * where abblen < 37); when the actual abbreviated result is a
4206 * bit longer than the requested length, we reduce the number
4207 * of dots so that they match the well-behaved ones. However,
4208 * if the actual abbreviation is longer than the requested
4209 * length by more than three, we give up on aligning, and add
4210 * three dots anyway, to indicate that the output is not the
4211 * full object name. Yes, this may be suboptimal, but this
4212 * appears only in "diff --raw --abbrev" output and it is not
4213 * worth the effort to change it now. Note that this would
4214 * likely to work fine when the automatic sizing of default
4215 * abbreviation length is used--we would be fed -1 in "len" in
4216 * that case, and will end up always appending three-dots, but
4217 * the automatic sizing is supposed to give abblen that ensures
4218 * uniqueness across all objects (statistically speaking).
4219 */
4220 if (abblen < GIT_SHA1_HEXSZ - 3) {
4221 static char hex[GIT_MAX_HEXSZ + 1];
4222 if (len < abblen && abblen <= len + 2)
4223 xsnprintf(hex, sizeof(hex), "%s%.*s", abbrev, len+3-abblen, "..");
4224 else
4225 xsnprintf(hex, sizeof(hex), "%s...", abbrev);
4226 return hex;
4227 }
4228
4229 return oid_to_hex(oid);
4230 }
4231
4232 static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
4233 {
4234 int line_termination = opt->line_termination;
4235 int inter_name_termination = line_termination ? '\t' : '\0';
4236
4237 fprintf(opt->file, "%s", diff_line_prefix(opt));
4238 if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
4239 fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
4240 diff_aligned_abbrev(&p->one->oid, opt->abbrev));
4241 fprintf(opt->file, "%s ",
4242 diff_aligned_abbrev(&p->two->oid, opt->abbrev));
4243 }
4244 if (p->score) {
4245 fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
4246 inter_name_termination);
4247 } else {
4248 fprintf(opt->file, "%c%c", p->status, inter_name_termination);
4249 }
4250
4251 if (p->status == DIFF_STATUS_COPIED ||
4252 p->status == DIFF_STATUS_RENAMED) {
4253 const char *name_a, *name_b;
4254 name_a = p->one->path;
4255 name_b = p->two->path;
4256 strip_prefix(opt->prefix_length, &name_a, &name_b);
4257 write_name_quoted(name_a, opt->file, inter_name_termination);
4258 write_name_quoted(name_b, opt->file, line_termination);
4259 } else {
4260 const char *name_a, *name_b;
4261 name_a = p->one->mode ? p->one->path : p->two->path;
4262 name_b = NULL;
4263 strip_prefix(opt->prefix_length, &name_a, &name_b);
4264 write_name_quoted(name_a, opt->file, line_termination);
4265 }
4266 }
4267
4268 int diff_unmodified_pair(struct diff_filepair *p)
4269 {
4270 /* This function is written stricter than necessary to support
4271 * the currently implemented transformers, but the idea is to
4272 * let transformers to produce diff_filepairs any way they want,
4273 * and filter and clean them up here before producing the output.
4274 */
4275 struct diff_filespec *one = p->one, *two = p->two;
4276
4277 if (DIFF_PAIR_UNMERGED(p))
4278 return 0; /* unmerged is interesting */
4279
4280 /* deletion, addition, mode or type change
4281 * and rename are all interesting.
4282 */
4283 if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
4284 DIFF_PAIR_MODE_CHANGED(p) ||
4285 strcmp(one->path, two->path))
4286 return 0;
4287
4288 /* both are valid and point at the same path. that is, we are
4289 * dealing with a change.
4290 */
4291 if (one->oid_valid && two->oid_valid &&
4292 !oidcmp(&one->oid, &two->oid) &&
4293 !one->dirty_submodule && !two->dirty_submodule)
4294 return 1; /* no change */
4295 if (!one->oid_valid && !two->oid_valid)
4296 return 1; /* both look at the same file on the filesystem. */
4297 return 0;
4298 }
4299
4300 static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
4301 {
4302 if (diff_unmodified_pair(p))
4303 return;
4304
4305 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4306 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4307 return; /* no tree diffs in patch format */
4308
4309 run_diff(p, o);
4310 }
4311
4312 static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
4313 struct diffstat_t *diffstat)
4314 {
4315 if (diff_unmodified_pair(p))
4316 return;
4317
4318 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4319 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4320 return; /* no useful stat for tree diffs */
4321
4322 run_diffstat(p, o, diffstat);
4323 }
4324
4325 static void diff_flush_checkdiff(struct diff_filepair *p,
4326 struct diff_options *o)
4327 {
4328 if (diff_unmodified_pair(p))
4329 return;
4330
4331 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4332 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4333 return; /* nothing to check in tree diffs */
4334
4335 run_checkdiff(p, o);
4336 }
4337
4338 int diff_queue_is_empty(void)
4339 {
4340 struct diff_queue_struct *q = &diff_queued_diff;
4341 int i;
4342 for (i = 0; i < q->nr; i++)
4343 if (!diff_unmodified_pair(q->queue[i]))
4344 return 0;
4345 return 1;
4346 }
4347
4348 #if DIFF_DEBUG
4349 void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
4350 {
4351 fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
4352 x, one ? one : "",
4353 s->path,
4354 DIFF_FILE_VALID(s) ? "valid" : "invalid",
4355 s->mode,
4356 s->oid_valid ? oid_to_hex(&s->oid) : "");
4357 fprintf(stderr, "queue[%d] %s size %lu\n",
4358 x, one ? one : "",
4359 s->size);
4360 }
4361
4362 void diff_debug_filepair(const struct diff_filepair *p, int i)
4363 {
4364 diff_debug_filespec(p->one, i, "one");
4365 diff_debug_filespec(p->two, i, "two");
4366 fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
4367 p->score, p->status ? p->status : '?',
4368 p->one->rename_used, p->broken_pair);
4369 }
4370
4371 void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
4372 {
4373 int i;
4374 if (msg)
4375 fprintf(stderr, "%s\n", msg);
4376 fprintf(stderr, "q->nr = %d\n", q->nr);
4377 for (i = 0; i < q->nr; i++) {
4378 struct diff_filepair *p = q->queue[i];
4379 diff_debug_filepair(p, i);
4380 }
4381 }
4382 #endif
4383
4384 static void diff_resolve_rename_copy(void)
4385 {
4386 int i;
4387 struct diff_filepair *p;
4388 struct diff_queue_struct *q = &diff_queued_diff;
4389
4390 diff_debug_queue("resolve-rename-copy", q);
4391
4392 for (i = 0; i < q->nr; i++) {
4393 p = q->queue[i];
4394 p->status = 0; /* undecided */
4395 if (DIFF_PAIR_UNMERGED(p))
4396 p->status = DIFF_STATUS_UNMERGED;
4397 else if (!DIFF_FILE_VALID(p->one))
4398 p->status = DIFF_STATUS_ADDED;
4399 else if (!DIFF_FILE_VALID(p->two))
4400 p->status = DIFF_STATUS_DELETED;
4401 else if (DIFF_PAIR_TYPE_CHANGED(p))
4402 p->status = DIFF_STATUS_TYPE_CHANGED;
4403
4404 /* from this point on, we are dealing with a pair
4405 * whose both sides are valid and of the same type, i.e.
4406 * either in-place edit or rename/copy edit.
4407 */
4408 else if (DIFF_PAIR_RENAME(p)) {
4409 /*
4410 * A rename might have re-connected a broken
4411 * pair up, causing the pathnames to be the
4412 * same again. If so, that's not a rename at
4413 * all, just a modification..
4414 *
4415 * Otherwise, see if this source was used for
4416 * multiple renames, in which case we decrement
4417 * the count, and call it a copy.
4418 */
4419 if (!strcmp(p->one->path, p->two->path))
4420 p->status = DIFF_STATUS_MODIFIED;
4421 else if (--p->one->rename_used > 0)
4422 p->status = DIFF_STATUS_COPIED;
4423 else
4424 p->status = DIFF_STATUS_RENAMED;
4425 }
4426 else if (oidcmp(&p->one->oid, &p->two->oid) ||
4427 p->one->mode != p->two->mode ||
4428 p->one->dirty_submodule ||
4429 p->two->dirty_submodule ||
4430 is_null_oid(&p->one->oid))
4431 p->status = DIFF_STATUS_MODIFIED;
4432 else {
4433 /* This is a "no-change" entry and should not
4434 * happen anymore, but prepare for broken callers.
4435 */
4436 error("feeding unmodified %s to diffcore",
4437 p->one->path);
4438 p->status = DIFF_STATUS_UNKNOWN;
4439 }
4440 }
4441 diff_debug_queue("resolve-rename-copy done", q);
4442 }
4443
4444 static int check_pair_status(struct diff_filepair *p)
4445 {
4446 switch (p->status) {
4447 case DIFF_STATUS_UNKNOWN:
4448 return 0;
4449 case 0:
4450 die("internal error in diff-resolve-rename-copy");
4451 default:
4452 return 1;
4453 }
4454 }
4455
4456 static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
4457 {
4458 int fmt = opt->output_format;
4459
4460 if (fmt & DIFF_FORMAT_CHECKDIFF)
4461 diff_flush_checkdiff(p, opt);
4462 else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
4463 diff_flush_raw(p, opt);
4464 else if (fmt & DIFF_FORMAT_NAME) {
4465 const char *name_a, *name_b;
4466 name_a = p->two->path;
4467 name_b = NULL;
4468 strip_prefix(opt->prefix_length, &name_a, &name_b);
4469 fprintf(opt->file, "%s", diff_line_prefix(opt));
4470 write_name_quoted(name_a, opt->file, opt->line_termination);
4471 }
4472 }
4473
4474 static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
4475 {
4476 if (fs->mode)
4477 fprintf(file, " %s mode %06o ", newdelete, fs->mode);
4478 else
4479 fprintf(file, " %s ", newdelete);
4480 write_name_quoted(fs->path, file, '\n');
4481 }
4482
4483
4484 static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
4485 const char *line_prefix)
4486 {
4487 if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
4488 fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
4489 p->two->mode, show_name ? ' ' : '\n');
4490 if (show_name) {
4491 write_name_quoted(p->two->path, file, '\n');
4492 }
4493 }
4494 }
4495
4496 static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
4497 const char *line_prefix)
4498 {
4499 char *names = pprint_rename(p->one->path, p->two->path);
4500
4501 fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
4502 free(names);
4503 show_mode_change(file, p, 0, line_prefix);
4504 }
4505
4506 static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
4507 {
4508 FILE *file = opt->file;
4509 const char *line_prefix = diff_line_prefix(opt);
4510
4511 switch(p->status) {
4512 case DIFF_STATUS_DELETED:
4513 fputs(line_prefix, file);
4514 show_file_mode_name(file, "delete", p->one);
4515 break;
4516 case DIFF_STATUS_ADDED:
4517 fputs(line_prefix, file);
4518 show_file_mode_name(file, "create", p->two);
4519 break;
4520 case DIFF_STATUS_COPIED:
4521 fputs(line_prefix, file);
4522 show_rename_copy(file, "copy", p, line_prefix);
4523 break;
4524 case DIFF_STATUS_RENAMED:
4525 fputs(line_prefix, file);
4526 show_rename_copy(file, "rename", p, line_prefix);
4527 break;
4528 default:
4529 if (p->score) {
4530 fprintf(file, "%s rewrite ", line_prefix);
4531 write_name_quoted(p->two->path, file, ' ');
4532 fprintf(file, "(%d%%)\n", similarity_index(p));
4533 }
4534 show_mode_change(file, p, !p->score, line_prefix);
4535 break;
4536 }
4537 }
4538
4539 struct patch_id_t {
4540 git_SHA_CTX *ctx;
4541 int patchlen;
4542 };
4543
4544 static int remove_space(char *line, int len)
4545 {
4546 int i;
4547 char *dst = line;
4548 unsigned char c;
4549
4550 for (i = 0; i < len; i++)
4551 if (!isspace((c = line[i])))
4552 *dst++ = c;
4553
4554 return dst - line;
4555 }
4556
4557 static void patch_id_consume(void *priv, char *line, unsigned long len)
4558 {
4559 struct patch_id_t *data = priv;
4560 int new_len;
4561
4562 /* Ignore line numbers when computing the SHA1 of the patch */
4563 if (starts_with(line, "@@ -"))
4564 return;
4565
4566 new_len = remove_space(line, len);
4567
4568 git_SHA1_Update(data->ctx, line, new_len);
4569 data->patchlen += new_len;
4570 }
4571
4572 static void patch_id_add_string(git_SHA_CTX *ctx, const char *str)
4573 {
4574 git_SHA1_Update(ctx, str, strlen(str));
4575 }
4576
4577 static void patch_id_add_mode(git_SHA_CTX *ctx, unsigned mode)
4578 {
4579 /* large enough for 2^32 in octal */
4580 char buf[12];
4581 int len = xsnprintf(buf, sizeof(buf), "%06o", mode);
4582 git_SHA1_Update(ctx, buf, len);
4583 }
4584
4585 /* returns 0 upon success, and writes result into sha1 */
4586 static int diff_get_patch_id(struct diff_options *options, struct object_id *oid, int diff_header_only)
4587 {
4588 struct diff_queue_struct *q = &diff_queued_diff;
4589 int i;
4590 git_SHA_CTX ctx;
4591 struct patch_id_t data;
4592
4593 git_SHA1_Init(&ctx);
4594 memset(&data, 0, sizeof(struct patch_id_t));
4595 data.ctx = &ctx;
4596
4597 for (i = 0; i < q->nr; i++) {
4598 xpparam_t xpp;
4599 xdemitconf_t xecfg;
4600 mmfile_t mf1, mf2;
4601 struct diff_filepair *p = q->queue[i];
4602 int len1, len2;
4603
4604 memset(&xpp, 0, sizeof(xpp));
4605 memset(&xecfg, 0, sizeof(xecfg));
4606 if (p->status == 0)
4607 return error("internal diff status error");
4608 if (p->status == DIFF_STATUS_UNKNOWN)
4609 continue;
4610 if (diff_unmodified_pair(p))
4611 continue;
4612 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4613 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4614 continue;
4615 if (DIFF_PAIR_UNMERGED(p))
4616 continue;
4617
4618 diff_fill_oid_info(p->one);
4619 diff_fill_oid_info(p->two);
4620
4621 len1 = remove_space(p->one->path, strlen(p->one->path));
4622 len2 = remove_space(p->two->path, strlen(p->two->path));
4623 patch_id_add_string(&ctx, "diff--git");
4624 patch_id_add_string(&ctx, "a/");
4625 git_SHA1_Update(&ctx, p->one->path, len1);
4626 patch_id_add_string(&ctx, "b/");
4627 git_SHA1_Update(&ctx, p->two->path, len2);
4628
4629 if (p->one->mode == 0) {
4630 patch_id_add_string(&ctx, "newfilemode");
4631 patch_id_add_mode(&ctx, p->two->mode);
4632 patch_id_add_string(&ctx, "---/dev/null");
4633 patch_id_add_string(&ctx, "+++b/");
4634 git_SHA1_Update(&ctx, p->two->path, len2);
4635 } else if (p->two->mode == 0) {
4636 patch_id_add_string(&ctx, "deletedfilemode");
4637 patch_id_add_mode(&ctx, p->one->mode);
4638 patch_id_add_string(&ctx, "---a/");
4639 git_SHA1_Update(&ctx, p->one->path, len1);
4640 patch_id_add_string(&ctx, "+++/dev/null");
4641 } else {
4642 patch_id_add_string(&ctx, "---a/");
4643 git_SHA1_Update(&ctx, p->one->path, len1);
4644 patch_id_add_string(&ctx, "+++b/");
4645 git_SHA1_Update(&ctx, p->two->path, len2);
4646 }
4647
4648 if (diff_header_only)
4649 continue;
4650
4651 if (fill_mmfile(&mf1, p->one) < 0 ||
4652 fill_mmfile(&mf2, p->two) < 0)
4653 return error("unable to read files to diff");
4654
4655 if (diff_filespec_is_binary(p->one) ||
4656 diff_filespec_is_binary(p->two)) {
4657 git_SHA1_Update(&ctx, oid_to_hex(&p->one->oid),
4658 GIT_SHA1_HEXSZ);
4659 git_SHA1_Update(&ctx, oid_to_hex(&p->two->oid),
4660 GIT_SHA1_HEXSZ);
4661 continue;
4662 }
4663
4664 xpp.flags = 0;
4665 xecfg.ctxlen = 3;
4666 xecfg.flags = 0;
4667 if (xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4668 &xpp, &xecfg))
4669 return error("unable to generate patch-id diff for %s",
4670 p->one->path);
4671 }
4672
4673 git_SHA1_Final(oid->hash, &ctx);
4674 return 0;
4675 }
4676
4677 int diff_flush_patch_id(struct diff_options *options, struct object_id *oid, int diff_header_only)
4678 {
4679 struct diff_queue_struct *q = &diff_queued_diff;
4680 int i;
4681 int result = diff_get_patch_id(options, oid, diff_header_only);
4682
4683 for (i = 0; i < q->nr; i++)
4684 diff_free_filepair(q->queue[i]);
4685
4686 free(q->queue);
4687 DIFF_QUEUE_CLEAR(q);
4688
4689 return result;
4690 }
4691
4692 static int is_summary_empty(const struct diff_queue_struct *q)
4693 {
4694 int i;
4695
4696 for (i = 0; i < q->nr; i++) {
4697 const struct diff_filepair *p = q->queue[i];
4698
4699 switch (p->status) {
4700 case DIFF_STATUS_DELETED:
4701 case DIFF_STATUS_ADDED:
4702 case DIFF_STATUS_COPIED:
4703 case DIFF_STATUS_RENAMED:
4704 return 0;
4705 default:
4706 if (p->score)
4707 return 0;
4708 if (p->one->mode && p->two->mode &&
4709 p->one->mode != p->two->mode)
4710 return 0;
4711 break;
4712 }
4713 }
4714 return 1;
4715 }
4716
4717 static const char rename_limit_warning[] =
4718 N_("inexact rename detection was skipped due to too many files.");
4719
4720 static const char degrade_cc_to_c_warning[] =
4721 N_("only found copies from modified paths due to too many files.");
4722
4723 static const char rename_limit_advice[] =
4724 N_("you may want to set your %s variable to at least "
4725 "%d and retry the command.");
4726
4727 void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4728 {
4729 if (degraded_cc)
4730 warning(_(degrade_cc_to_c_warning));
4731 else if (needed)
4732 warning(_(rename_limit_warning));
4733 else
4734 return;
4735 if (0 < needed && needed < 32767)
4736 warning(_(rename_limit_advice), varname, needed);
4737 }
4738
4739 void diff_flush(struct diff_options *options)
4740 {
4741 struct diff_queue_struct *q = &diff_queued_diff;
4742 int i, output_format = options->output_format;
4743 int separator = 0;
4744 int dirstat_by_line = 0;
4745
4746 /*
4747 * Order: raw, stat, summary, patch
4748 * or: name/name-status/checkdiff (other bits clear)
4749 */
4750 if (!q->nr)
4751 goto free_queue;
4752
4753 if (output_format & (DIFF_FORMAT_RAW |
4754 DIFF_FORMAT_NAME |
4755 DIFF_FORMAT_NAME_STATUS |
4756 DIFF_FORMAT_CHECKDIFF)) {
4757 for (i = 0; i < q->nr; i++) {
4758 struct diff_filepair *p = q->queue[i];
4759 if (check_pair_status(p))
4760 flush_one_pair(p, options);
4761 }
4762 separator++;
4763 }
4764
4765 if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4766 dirstat_by_line = 1;
4767
4768 if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4769 dirstat_by_line) {
4770 struct diffstat_t diffstat;
4771
4772 memset(&diffstat, 0, sizeof(struct diffstat_t));
4773 for (i = 0; i < q->nr; i++) {
4774 struct diff_filepair *p = q->queue[i];
4775 if (check_pair_status(p))
4776 diff_flush_stat(p, options, &diffstat);
4777 }
4778 if (output_format & DIFF_FORMAT_NUMSTAT)
4779 show_numstat(&diffstat, options);
4780 if (output_format & DIFF_FORMAT_DIFFSTAT)
4781 show_stats(&diffstat, options);
4782 if (output_format & DIFF_FORMAT_SHORTSTAT)
4783 show_shortstats(&diffstat, options);
4784 if (output_format & DIFF_FORMAT_DIRSTAT && dirstat_by_line)
4785 show_dirstat_by_line(&diffstat, options);
4786 free_diffstat_info(&diffstat);
4787 separator++;
4788 }
4789 if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4790 show_dirstat(options);
4791
4792 if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4793 for (i = 0; i < q->nr; i++) {
4794 diff_summary(options, q->queue[i]);
4795 }
4796 separator++;
4797 }
4798
4799 if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4800 DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4801 DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4802 /*
4803 * run diff_flush_patch for the exit status. setting
4804 * options->file to /dev/null should be safe, because we
4805 * aren't supposed to produce any output anyway.
4806 */
4807 if (options->close_file)
4808 fclose(options->file);
4809 options->file = xfopen("/dev/null", "w");
4810 options->close_file = 1;
4811 for (i = 0; i < q->nr; i++) {
4812 struct diff_filepair *p = q->queue[i];
4813 if (check_pair_status(p))
4814 diff_flush_patch(p, options);
4815 if (options->found_changes)
4816 break;
4817 }
4818 }
4819
4820 if (output_format & DIFF_FORMAT_PATCH) {
4821 if (separator) {
4822 fprintf(options->file, "%s%c",
4823 diff_line_prefix(options),
4824 options->line_termination);
4825 if (options->stat_sep) {
4826 /* attach patch instead of inline */
4827 fputs(options->stat_sep, options->file);
4828 }
4829 }
4830
4831 for (i = 0; i < q->nr; i++) {
4832 struct diff_filepair *p = q->queue[i];
4833 if (check_pair_status(p))
4834 diff_flush_patch(p, options);
4835 }
4836 }
4837
4838 if (output_format & DIFF_FORMAT_CALLBACK)
4839 options->format_callback(q, options, options->format_callback_data);
4840
4841 for (i = 0; i < q->nr; i++)
4842 diff_free_filepair(q->queue[i]);
4843 free_queue:
4844 free(q->queue);
4845 DIFF_QUEUE_CLEAR(q);
4846 if (options->close_file)
4847 fclose(options->file);
4848
4849 /*
4850 * Report the content-level differences with HAS_CHANGES;
4851 * diff_addremove/diff_change does not set the bit when
4852 * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4853 */
4854 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4855 if (options->found_changes)
4856 DIFF_OPT_SET(options, HAS_CHANGES);
4857 else
4858 DIFF_OPT_CLR(options, HAS_CHANGES);
4859 }
4860 }
4861
4862 static int match_filter(const struct diff_options *options, const struct diff_filepair *p)
4863 {
4864 return (((p->status == DIFF_STATUS_MODIFIED) &&
4865 ((p->score &&
4866 filter_bit_tst(DIFF_STATUS_FILTER_BROKEN, options)) ||
4867 (!p->score &&
4868 filter_bit_tst(DIFF_STATUS_MODIFIED, options)))) ||
4869 ((p->status != DIFF_STATUS_MODIFIED) &&
4870 filter_bit_tst(p->status, options)));
4871 }
4872
4873 static void diffcore_apply_filter(struct diff_options *options)
4874 {
4875 int i;
4876 struct diff_queue_struct *q = &diff_queued_diff;
4877 struct diff_queue_struct outq;
4878
4879 DIFF_QUEUE_CLEAR(&outq);
4880
4881 if (!options->filter)
4882 return;
4883
4884 if (filter_bit_tst(DIFF_STATUS_FILTER_AON, options)) {
4885 int found;
4886 for (i = found = 0; !found && i < q->nr; i++) {
4887 if (match_filter(options, q->queue[i]))
4888 found++;
4889 }
4890 if (found)
4891 return;
4892
4893 /* otherwise we will clear the whole queue
4894 * by copying the empty outq at the end of this
4895 * function, but first clear the current entries
4896 * in the queue.
4897 */
4898 for (i = 0; i < q->nr; i++)
4899 diff_free_filepair(q->queue[i]);
4900 }
4901 else {
4902 /* Only the matching ones */
4903 for (i = 0; i < q->nr; i++) {
4904 struct diff_filepair *p = q->queue[i];
4905 if (match_filter(options, p))
4906 diff_q(&outq, p);
4907 else
4908 diff_free_filepair(p);
4909 }
4910 }
4911 free(q->queue);
4912 *q = outq;
4913 }
4914
4915 /* Check whether two filespecs with the same mode and size are identical */
4916 static int diff_filespec_is_identical(struct diff_filespec *one,
4917 struct diff_filespec *two)
4918 {
4919 if (S_ISGITLINK(one->mode))
4920 return 0;
4921 if (diff_populate_filespec(one, 0))
4922 return 0;
4923 if (diff_populate_filespec(two, 0))
4924 return 0;
4925 return !memcmp(one->data, two->data, one->size);
4926 }
4927
4928 static int diff_filespec_check_stat_unmatch(struct diff_filepair *p)
4929 {
4930 if (p->done_skip_stat_unmatch)
4931 return p->skip_stat_unmatch_result;
4932
4933 p->done_skip_stat_unmatch = 1;
4934 p->skip_stat_unmatch_result = 0;
4935 /*
4936 * 1. Entries that come from stat info dirtiness
4937 * always have both sides (iow, not create/delete),
4938 * one side of the object name is unknown, with
4939 * the same mode and size. Keep the ones that
4940 * do not match these criteria. They have real
4941 * differences.
4942 *
4943 * 2. At this point, the file is known to be modified,
4944 * with the same mode and size, and the object
4945 * name of one side is unknown. Need to inspect
4946 * the identical contents.
4947 */
4948 if (!DIFF_FILE_VALID(p->one) || /* (1) */
4949 !DIFF_FILE_VALID(p->two) ||
4950 (p->one->oid_valid && p->two->oid_valid) ||
4951 (p->one->mode != p->two->mode) ||
4952 diff_populate_filespec(p->one, CHECK_SIZE_ONLY) ||
4953 diff_populate_filespec(p->two, CHECK_SIZE_ONLY) ||
4954 (p->one->size != p->two->size) ||
4955 !diff_filespec_is_identical(p->one, p->two)) /* (2) */
4956 p->skip_stat_unmatch_result = 1;
4957 return p->skip_stat_unmatch_result;
4958 }
4959
4960 static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
4961 {
4962 int i;
4963 struct diff_queue_struct *q = &diff_queued_diff;
4964 struct diff_queue_struct outq;
4965 DIFF_QUEUE_CLEAR(&outq);
4966
4967 for (i = 0; i < q->nr; i++) {
4968 struct diff_filepair *p = q->queue[i];
4969
4970 if (diff_filespec_check_stat_unmatch(p))
4971 diff_q(&outq, p);
4972 else {
4973 /*
4974 * The caller can subtract 1 from skip_stat_unmatch
4975 * to determine how many paths were dirty only
4976 * due to stat info mismatch.
4977 */
4978 if (!DIFF_OPT_TST(diffopt, NO_INDEX))
4979 diffopt->skip_stat_unmatch++;
4980 diff_free_filepair(p);
4981 }
4982 }
4983 free(q->queue);
4984 *q = outq;
4985 }
4986
4987 static int diffnamecmp(const void *a_, const void *b_)
4988 {
4989 const struct diff_filepair *a = *((const struct diff_filepair **)a_);
4990 const struct diff_filepair *b = *((const struct diff_filepair **)b_);
4991 const char *name_a, *name_b;
4992
4993 name_a = a->one ? a->one->path : a->two->path;
4994 name_b = b->one ? b->one->path : b->two->path;
4995 return strcmp(name_a, name_b);
4996 }
4997
4998 void diffcore_fix_diff_index(struct diff_options *options)
4999 {
5000 struct diff_queue_struct *q = &diff_queued_diff;
5001 QSORT(q->queue, q->nr, diffnamecmp);
5002 }
5003
5004 void diffcore_std(struct diff_options *options)
5005 {
5006 /* NOTE please keep the following in sync with diff_tree_combined() */
5007 if (options->skip_stat_unmatch)
5008 diffcore_skip_stat_unmatch(options);
5009 if (!options->found_follow) {
5010 /* See try_to_follow_renames() in tree-diff.c */
5011 if (options->break_opt != -1)
5012 diffcore_break(options->break_opt);
5013 if (options->detect_rename)
5014 diffcore_rename(options);
5015 if (options->break_opt != -1)
5016 diffcore_merge_broken();
5017 }
5018 if (options->pickaxe)
5019 diffcore_pickaxe(options);
5020 if (options->orderfile)
5021 diffcore_order(options->orderfile);
5022 if (!options->found_follow)
5023 /* See try_to_follow_renames() in tree-diff.c */
5024 diff_resolve_rename_copy();
5025 diffcore_apply_filter(options);
5026
5027 if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5028 DIFF_OPT_SET(options, HAS_CHANGES);
5029 else
5030 DIFF_OPT_CLR(options, HAS_CHANGES);
5031
5032 options->found_follow = 0;
5033 }
5034
5035 int diff_result_code(struct diff_options *opt, int status)
5036 {
5037 int result = 0;
5038
5039 diff_warn_rename_limit("diff.renameLimit",
5040 opt->needed_rename_limit,
5041 opt->degraded_cc_to_c);
5042 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5043 !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
5044 return status;
5045 if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5046 DIFF_OPT_TST(opt, HAS_CHANGES))
5047 result |= 01;
5048 if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
5049 DIFF_OPT_TST(opt, CHECK_FAILED))
5050 result |= 02;
5051 return result;
5052 }
5053
5054 int diff_can_quit_early(struct diff_options *opt)
5055 {
5056 return (DIFF_OPT_TST(opt, QUICK) &&
5057 !opt->filter &&
5058 DIFF_OPT_TST(opt, HAS_CHANGES));
5059 }
5060
5061 /*
5062 * Shall changes to this submodule be ignored?
5063 *
5064 * Submodule changes can be configured to be ignored separately for each path,
5065 * but that configuration can be overridden from the command line.
5066 */
5067 static int is_submodule_ignored(const char *path, struct diff_options *options)
5068 {
5069 int ignored = 0;
5070 unsigned orig_flags = options->flags;
5071 if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
5072 set_diffopt_flags_from_submodule_config(options, path);
5073 if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
5074 ignored = 1;
5075 options->flags = orig_flags;
5076 return ignored;
5077 }
5078
5079 void diff_addremove(struct diff_options *options,
5080 int addremove, unsigned mode,
5081 const struct object_id *oid,
5082 int oid_valid,
5083 const char *concatpath, unsigned dirty_submodule)
5084 {
5085 struct diff_filespec *one, *two;
5086
5087 if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
5088 return;
5089
5090 /* This may look odd, but it is a preparation for
5091 * feeding "there are unchanged files which should
5092 * not produce diffs, but when you are doing copy
5093 * detection you would need them, so here they are"
5094 * entries to the diff-core. They will be prefixed
5095 * with something like '=' or '*' (I haven't decided
5096 * which but should not make any difference).
5097 * Feeding the same new and old to diff_change()
5098 * also has the same effect.
5099 * Before the final output happens, they are pruned after
5100 * merged into rename/copy pairs as appropriate.
5101 */
5102 if (DIFF_OPT_TST(options, REVERSE_DIFF))
5103 addremove = (addremove == '+' ? '-' :
5104 addremove == '-' ? '+' : addremove);
5105
5106 if (options->prefix &&
5107 strncmp(concatpath, options->prefix, options->prefix_length))
5108 return;
5109
5110 one = alloc_filespec(concatpath);
5111 two = alloc_filespec(concatpath);
5112
5113 if (addremove != '+')
5114 fill_filespec(one, oid, oid_valid, mode);
5115 if (addremove != '-') {
5116 fill_filespec(two, oid, oid_valid, mode);
5117 two->dirty_submodule = dirty_submodule;
5118 }
5119
5120 diff_queue(&diff_queued_diff, one, two);
5121 if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5122 DIFF_OPT_SET(options, HAS_CHANGES);
5123 }
5124
5125 void diff_change(struct diff_options *options,
5126 unsigned old_mode, unsigned new_mode,
5127 const struct object_id *old_oid,
5128 const struct object_id *new_oid,
5129 int old_oid_valid, int new_oid_valid,
5130 const char *concatpath,
5131 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
5132 {
5133 struct diff_filespec *one, *two;
5134 struct diff_filepair *p;
5135
5136 if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
5137 is_submodule_ignored(concatpath, options))
5138 return;
5139
5140 if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
5141 SWAP(old_mode, new_mode);
5142 SWAP(old_oid, new_oid);
5143 SWAP(old_oid_valid, new_oid_valid);
5144 SWAP(old_dirty_submodule, new_dirty_submodule);
5145 }
5146
5147 if (options->prefix &&
5148 strncmp(concatpath, options->prefix, options->prefix_length))
5149 return;
5150
5151 one = alloc_filespec(concatpath);
5152 two = alloc_filespec(concatpath);
5153 fill_filespec(one, old_oid, old_oid_valid, old_mode);
5154 fill_filespec(two, new_oid, new_oid_valid, new_mode);
5155 one->dirty_submodule = old_dirty_submodule;
5156 two->dirty_submodule = new_dirty_submodule;
5157 p = diff_queue(&diff_queued_diff, one, two);
5158
5159 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5160 return;
5161
5162 if (DIFF_OPT_TST(options, QUICK) && options->skip_stat_unmatch &&
5163 !diff_filespec_check_stat_unmatch(p))
5164 return;
5165
5166 DIFF_OPT_SET(options, HAS_CHANGES);
5167 }
5168
5169 struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
5170 {
5171 struct diff_filepair *pair;
5172 struct diff_filespec *one, *two;
5173
5174 if (options->prefix &&
5175 strncmp(path, options->prefix, options->prefix_length))
5176 return NULL;
5177
5178 one = alloc_filespec(path);
5179 two = alloc_filespec(path);
5180 pair = diff_queue(&diff_queued_diff, one, two);
5181 pair->is_unmerged = 1;
5182 return pair;
5183 }
5184
5185 static char *run_textconv(const char *pgm, struct diff_filespec *spec,
5186 size_t *outsize)
5187 {
5188 struct diff_tempfile *temp;
5189 const char *argv[3];
5190 const char **arg = argv;
5191 struct child_process child = CHILD_PROCESS_INIT;
5192 struct strbuf buf = STRBUF_INIT;
5193 int err = 0;
5194
5195 temp = prepare_temp_file(spec->path, spec);
5196 *arg++ = pgm;
5197 *arg++ = temp->name;
5198 *arg = NULL;
5199
5200 child.use_shell = 1;
5201 child.argv = argv;
5202 child.out = -1;
5203 if (start_command(&child)) {
5204 remove_tempfile();
5205 return NULL;
5206 }
5207
5208 if (strbuf_read(&buf, child.out, 0) < 0)
5209 err = error("error reading from textconv command '%s'", pgm);
5210 close(child.out);
5211
5212 if (finish_command(&child) || err) {
5213 strbuf_release(&buf);
5214 remove_tempfile();
5215 return NULL;
5216 }
5217 remove_tempfile();
5218
5219 return strbuf_detach(&buf, outsize);
5220 }
5221
5222 size_t fill_textconv(struct userdiff_driver *driver,
5223 struct diff_filespec *df,
5224 char **outbuf)
5225 {
5226 size_t size;
5227
5228 if (!driver) {
5229 if (!DIFF_FILE_VALID(df)) {
5230 *outbuf = "";
5231 return 0;
5232 }
5233 if (diff_populate_filespec(df, 0))
5234 die("unable to read files to diff");
5235 *outbuf = df->data;
5236 return df->size;
5237 }
5238
5239 if (!driver->textconv)
5240 die("BUG: fill_textconv called with non-textconv driver");
5241
5242 if (driver->textconv_cache && df->oid_valid) {
5243 *outbuf = notes_cache_get(driver->textconv_cache,
5244 &df->oid,
5245 &size);
5246 if (*outbuf)
5247 return size;
5248 }
5249
5250 *outbuf = run_textconv(driver->textconv, df, &size);
5251 if (!*outbuf)
5252 die("unable to read files to diff");
5253
5254 if (driver->textconv_cache && df->oid_valid) {
5255 /* ignore errors, as we might be in a readonly repository */
5256 notes_cache_put(driver->textconv_cache, &df->oid, *outbuf,
5257 size);
5258 /*
5259 * we could save up changes and flush them all at the end,
5260 * but we would need an extra call after all diffing is done.
5261 * Since generating a cache entry is the slow path anyway,
5262 * this extra overhead probably isn't a big deal.
5263 */
5264 notes_cache_write(driver->textconv_cache);
5265 }
5266
5267 return size;
5268 }
5269
5270 int textconv_object(const char *path,
5271 unsigned mode,
5272 const struct object_id *oid,
5273 int oid_valid,
5274 char **buf,
5275 unsigned long *buf_size)
5276 {
5277 struct diff_filespec *df;
5278 struct userdiff_driver *textconv;
5279
5280 df = alloc_filespec(path);
5281 fill_filespec(df, oid, oid_valid, mode);
5282 textconv = get_textconv(df);
5283 if (!textconv) {
5284 free_filespec(df);
5285 return 0;
5286 }
5287
5288 *buf_size = fill_textconv(textconv, df, buf);
5289 free_filespec(df);
5290 return 1;
5291 }
5292
5293 void setup_diff_pager(struct diff_options *opt)
5294 {
5295 /*
5296 * If the user asked for our exit code, then either they want --quiet
5297 * or --exit-code. We should definitely not bother with a pager in the
5298 * former case, as we will generate no output. Since we still properly
5299 * report our exit code even when a pager is run, we _could_ run a
5300 * pager with --exit-code. But since we have not done so historically,
5301 * and because it is easy to find people oneline advising "git diff
5302 * --exit-code" in hooks and other scripts, we do not do so.
5303 */
5304 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5305 check_pager_config("diff") != 0)
5306 setup_pager();
5307 }
git for Debian