?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
diff.c: add emit_del_line() and emit_context_line()
[git/raj.git] / diff.c
blobc575c452f16fb781d86f5fe748a216c192c5ad41
1 /*
2 * Copyright (C) 2005 Junio C Hamano
3 */
4 #include "cache.h"
5 #include "quote.h"
6 #include "diff.h"
7 #include "diffcore.h"
8 #include "delta.h"
9 #include "xdiff-interface.h"
10 #include "color.h"
11 #include "attr.h"
12 #include "run-command.h"
13 #include "utf8.h"
14 #include "userdiff.h"
15 #include "sigchain.h"
16 #include "submodule.h"
17 #include "ll-merge.h"
18 #include "string-list.h"
19 #include "argv-array.h"
20
21 #ifdef NO_FAST_WORKING_DIRECTORY
22 #define FAST_WORKING_DIRECTORY 0
23 #else
24 #define FAST_WORKING_DIRECTORY 1
25 #endif
26
27 static int diff_detect_rename_default;
28 static int diff_rename_limit_default = 400;
29 static int diff_suppress_blank_empty;
30 static int diff_use_color_default = -1;
31 static int diff_context_default = 3;
32 static const char *diff_word_regex_cfg;
33 static const char *external_diff_cmd_cfg;
34 static const char *diff_order_file_cfg;
35 int diff_auto_refresh_index = 1;
36 static int diff_mnemonic_prefix;
37 static int diff_no_prefix;
38 static int diff_stat_graph_width;
39 static int diff_dirstat_permille_default = 30;
40 static struct diff_options default_diff_options;
41 static long diff_algorithm;
42
43 static char diff_colors[][COLOR_MAXLEN] = {
44 GIT_COLOR_RESET,
45 GIT_COLOR_NORMAL, /* PLAIN */
46 GIT_COLOR_BOLD, /* METAINFO */
47 GIT_COLOR_CYAN, /* FRAGINFO */
48 GIT_COLOR_RED, /* OLD */
49 GIT_COLOR_GREEN, /* NEW */
50 GIT_COLOR_YELLOW, /* COMMIT */
51 GIT_COLOR_BG_RED, /* WHITESPACE */
52 GIT_COLOR_NORMAL, /* FUNCINFO */
53 };
54
55 static int parse_diff_color_slot(const char *var)
56 {
57 if (!strcasecmp(var, "plain"))
58 return DIFF_PLAIN;
59 if (!strcasecmp(var, "meta"))
60 return DIFF_METAINFO;
61 if (!strcasecmp(var, "frag"))
62 return DIFF_FRAGINFO;
63 if (!strcasecmp(var, "old"))
64 return DIFF_FILE_OLD;
65 if (!strcasecmp(var, "new"))
66 return DIFF_FILE_NEW;
67 if (!strcasecmp(var, "commit"))
68 return DIFF_COMMIT;
69 if (!strcasecmp(var, "whitespace"))
70 return DIFF_WHITESPACE;
71 if (!strcasecmp(var, "func"))
72 return DIFF_FUNCINFO;
73 return -1;
74 }
75
76 static int parse_dirstat_params(struct diff_options *options, const char *params_string,
77 struct strbuf *errmsg)
78 {
79 char *params_copy = xstrdup(params_string);
80 struct string_list params = STRING_LIST_INIT_NODUP;
81 int ret = 0;
82 int i;
83
84 if (*params_copy)
85 string_list_split_in_place(&params, params_copy, ',', -1);
86 for (i = 0; i < params.nr; i++) {
87 const char *p = params.items[i].string;
88 if (!strcmp(p, "changes")) {
89 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
90 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
91 } else if (!strcmp(p, "lines")) {
92 DIFF_OPT_SET(options, DIRSTAT_BY_LINE);
93 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
94 } else if (!strcmp(p, "files")) {
95 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
96 DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
97 } else if (!strcmp(p, "noncumulative")) {
98 DIFF_OPT_CLR(options, DIRSTAT_CUMULATIVE);
99 } else if (!strcmp(p, "cumulative")) {
100 DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
101 } else if (isdigit(*p)) {
102 char *end;
103 int permille = strtoul(p, &end, 10) * 10;
104 if (*end == '.' && isdigit(*++end)) {
105 /* only use first digit */
106 permille += *end - '0';
107 /* .. and ignore any further digits */
108 while (isdigit(*++end))
109 ; /* nothing */
110 }
111 if (!*end)
112 options->dirstat_permille = permille;
113 else {
114 strbuf_addf(errmsg, _(" Failed to parse dirstat cut-off percentage '%s'\n"),
115 p);
116 ret++;
117 }
118 } else {
119 strbuf_addf(errmsg, _(" Unknown dirstat parameter '%s'\n"), p);
120 ret++;
121 }
122
123 }
124 string_list_clear(&params, 0);
125 free(params_copy);
126 return ret;
127 }
128
129 static int parse_submodule_params(struct diff_options *options, const char *value)
130 {
131 if (!strcmp(value, "log"))
132 DIFF_OPT_SET(options, SUBMODULE_LOG);
133 else if (!strcmp(value, "short"))
134 DIFF_OPT_CLR(options, SUBMODULE_LOG);
135 else
136 return -1;
137 return 0;
138 }
139
140 static int git_config_rename(const char *var, const char *value)
141 {
142 if (!value)
143 return DIFF_DETECT_RENAME;
144 if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
145 return DIFF_DETECT_COPY;
146 return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
147 }
148
149 long parse_algorithm_value(const char *value)
150 {
151 if (!value)
152 return -1;
153 else if (!strcasecmp(value, "myers") || !strcasecmp(value, "default"))
154 return 0;
155 else if (!strcasecmp(value, "minimal"))
156 return XDF_NEED_MINIMAL;
157 else if (!strcasecmp(value, "patience"))
158 return XDF_PATIENCE_DIFF;
159 else if (!strcasecmp(value, "histogram"))
160 return XDF_HISTOGRAM_DIFF;
161 return -1;
162 }
163
164 /*
165 * These are to give UI layer defaults.
166 * The core-level commands such as git-diff-files should
167 * never be affected by the setting of diff.renames
168 * the user happens to have in the configuration file.
169 */
170 int git_diff_ui_config(const char *var, const char *value, void *cb)
171 {
172 if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
173 diff_use_color_default = git_config_colorbool(var, value);
174 return 0;
175 }
176 if (!strcmp(var, "diff.context")) {
177 diff_context_default = git_config_int(var, value);
178 if (diff_context_default < 0)
179 return -1;
180 return 0;
181 }
182 if (!strcmp(var, "diff.renames")) {
183 diff_detect_rename_default = git_config_rename(var, value);
184 return 0;
185 }
186 if (!strcmp(var, "diff.autorefreshindex")) {
187 diff_auto_refresh_index = git_config_bool(var, value);
188 return 0;
189 }
190 if (!strcmp(var, "diff.mnemonicprefix")) {
191 diff_mnemonic_prefix = git_config_bool(var, value);
192 return 0;
193 }
194 if (!strcmp(var, "diff.noprefix")) {
195 diff_no_prefix = git_config_bool(var, value);
196 return 0;
197 }
198 if (!strcmp(var, "diff.statgraphwidth")) {
199 diff_stat_graph_width = git_config_int(var, value);
200 return 0;
201 }
202 if (!strcmp(var, "diff.external"))
203 return git_config_string(&external_diff_cmd_cfg, var, value);
204 if (!strcmp(var, "diff.wordregex"))
205 return git_config_string(&diff_word_regex_cfg, var, value);
206 if (!strcmp(var, "diff.orderfile"))
207 return git_config_pathname(&diff_order_file_cfg, var, value);
208
209 if (!strcmp(var, "diff.ignoresubmodules"))
210 handle_ignore_submodules_arg(&default_diff_options, value);
211
212 if (!strcmp(var, "diff.submodule")) {
213 if (parse_submodule_params(&default_diff_options, value))
214 warning(_("Unknown value for 'diff.submodule' config variable: '%s'"),
215 value);
216 return 0;
217 }
218
219 if (!strcmp(var, "diff.algorithm")) {
220 diff_algorithm = parse_algorithm_value(value);
221 if (diff_algorithm < 0)
222 return -1;
223 return 0;
224 }
225
226 if (git_color_config(var, value, cb) < 0)
227 return -1;
228
229 return git_diff_basic_config(var, value, cb);
230 }
231
232 int git_diff_basic_config(const char *var, const char *value, void *cb)
233 {
234 const char *name;
235
236 if (!strcmp(var, "diff.renamelimit")) {
237 diff_rename_limit_default = git_config_int(var, value);
238 return 0;
239 }
240
241 if (userdiff_config(var, value) < 0)
242 return -1;
243
244 if (skip_prefix(var, "diff.color.", &name) ||
245 skip_prefix(var, "color.diff.", &name)) {
246 int slot = parse_diff_color_slot(name);
247 if (slot < 0)
248 return 0;
249 if (!value)
250 return config_error_nonbool(var);
251 return color_parse(value, diff_colors[slot]);
252 }
253
254 /* like GNU diff's --suppress-blank-empty option */
255 if (!strcmp(var, "diff.suppressblankempty") ||
256 /* for backwards compatibility */
257 !strcmp(var, "diff.suppress-blank-empty")) {
258 diff_suppress_blank_empty = git_config_bool(var, value);
259 return 0;
260 }
261
262 if (!strcmp(var, "diff.dirstat")) {
263 struct strbuf errmsg = STRBUF_INIT;
264 default_diff_options.dirstat_permille = diff_dirstat_permille_default;
265 if (parse_dirstat_params(&default_diff_options, value, &errmsg))
266 warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
267 errmsg.buf);
268 strbuf_release(&errmsg);
269 diff_dirstat_permille_default = default_diff_options.dirstat_permille;
270 return 0;
271 }
272
273 if (starts_with(var, "submodule."))
274 return parse_submodule_config_option(var, value);
275
276 return git_default_config(var, value, cb);
277 }
278
279 static char *quote_two(const char *one, const char *two)
280 {
281 int need_one = quote_c_style(one, NULL, NULL, 1);
282 int need_two = quote_c_style(two, NULL, NULL, 1);
283 struct strbuf res = STRBUF_INIT;
284
285 if (need_one + need_two) {
286 strbuf_addch(&res, '"');
287 quote_c_style(one, &res, NULL, 1);
288 quote_c_style(two, &res, NULL, 1);
289 strbuf_addch(&res, '"');
290 } else {
291 strbuf_addstr(&res, one);
292 strbuf_addstr(&res, two);
293 }
294 return strbuf_detach(&res, NULL);
295 }
296
297 static const char *external_diff(void)
298 {
299 static const char *external_diff_cmd = NULL;
300 static int done_preparing = 0;
301
302 if (done_preparing)
303 return external_diff_cmd;
304 external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
305 if (!external_diff_cmd)
306 external_diff_cmd = external_diff_cmd_cfg;
307 done_preparing = 1;
308 return external_diff_cmd;
309 }
310
311 static struct diff_tempfile {
312 const char *name; /* filename external diff should read from */
313 char hex[41];
314 char mode[10];
315 char tmp_path[PATH_MAX];
316 } diff_temp[2];
317
318 typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
319
320 struct emit_callback {
321 int color_diff;
322 unsigned ws_rule;
323 int blank_at_eof_in_preimage;
324 int blank_at_eof_in_postimage;
325 int lno_in_preimage;
326 int lno_in_postimage;
327 sane_truncate_fn truncate;
328 const char **label_path;
329 struct diff_words_data *diff_words;
330 struct diff_options *opt;
331 int *found_changesp;
332 struct strbuf *header;
333 };
334
335 static int count_lines(const char *data, int size)
336 {
337 int count, ch, completely_empty = 1, nl_just_seen = 0;
338 count = 0;
339 while (0 < size--) {
340 ch = *data++;
341 if (ch == '\n') {
342 count++;
343 nl_just_seen = 1;
344 completely_empty = 0;
345 }
346 else {
347 nl_just_seen = 0;
348 completely_empty = 0;
349 }
350 }
351 if (completely_empty)
352 return 0;
353 if (!nl_just_seen)
354 count++; /* no trailing newline */
355 return count;
356 }
357
358 static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
359 {
360 if (!DIFF_FILE_VALID(one)) {
361 mf->ptr = (char *)""; /* does not matter */
362 mf->size = 0;
363 return 0;
364 }
365 else if (diff_populate_filespec(one, 0))
366 return -1;
367
368 mf->ptr = one->data;
369 mf->size = one->size;
370 return 0;
371 }
372
373 /* like fill_mmfile, but only for size, so we can avoid retrieving blob */
374 static unsigned long diff_filespec_size(struct diff_filespec *one)
375 {
376 if (!DIFF_FILE_VALID(one))
377 return 0;
378 diff_populate_filespec(one, CHECK_SIZE_ONLY);
379 return one->size;
380 }
381
382 static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
383 {
384 char *ptr = mf->ptr;
385 long size = mf->size;
386 int cnt = 0;
387
388 if (!size)
389 return cnt;
390 ptr += size - 1; /* pointing at the very end */
391 if (*ptr != '\n')
392 ; /* incomplete line */
393 else
394 ptr--; /* skip the last LF */
395 while (mf->ptr < ptr) {
396 char *prev_eol;
397 for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
398 if (*prev_eol == '\n')
399 break;
400 if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
401 break;
402 cnt++;
403 ptr = prev_eol - 1;
404 }
405 return cnt;
406 }
407
408 static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
409 struct emit_callback *ecbdata)
410 {
411 int l1, l2, at;
412 unsigned ws_rule = ecbdata->ws_rule;
413 l1 = count_trailing_blank(mf1, ws_rule);
414 l2 = count_trailing_blank(mf2, ws_rule);
415 if (l2 <= l1) {
416 ecbdata->blank_at_eof_in_preimage = 0;
417 ecbdata->blank_at_eof_in_postimage = 0;
418 return;
419 }
420 at = count_lines(mf1->ptr, mf1->size);
421 ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
422
423 at = count_lines(mf2->ptr, mf2->size);
424 ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
425 }
426
427 static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
428 int first, const char *line, int len)
429 {
430 int has_trailing_newline, has_trailing_carriage_return;
431 int nofirst;
432 FILE *file = o->file;
433
434 fputs(diff_line_prefix(o), file);
435
436 if (len == 0) {
437 has_trailing_newline = (first == '\n');
438 has_trailing_carriage_return = (!has_trailing_newline &&
439 (first == '\r'));
440 nofirst = has_trailing_newline || has_trailing_carriage_return;
441 } else {
442 has_trailing_newline = (len > 0 && line[len-1] == '\n');
443 if (has_trailing_newline)
444 len--;
445 has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
446 if (has_trailing_carriage_return)
447 len--;
448 nofirst = 0;
449 }
450
451 if (len || !nofirst) {
452 fputs(set, file);
453 if (!nofirst)
454 fputc(first, file);
455 fwrite(line, len, 1, file);
456 fputs(reset, file);
457 }
458 if (has_trailing_carriage_return)
459 fputc('\r', file);
460 if (has_trailing_newline)
461 fputc('\n', file);
462 }
463
464 static void emit_line(struct diff_options *o, const char *set, const char *reset,
465 const char *line, int len)
466 {
467 emit_line_0(o, set, reset, line[0], line+1, len-1);
468 }
469
470 static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
471 {
472 if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
473 ecbdata->blank_at_eof_in_preimage &&
474 ecbdata->blank_at_eof_in_postimage &&
475 ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
476 ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
477 return 0;
478 return ws_blank_line(line, len, ecbdata->ws_rule);
479 }
480
481 static void emit_add_line(const char *reset,
482 struct emit_callback *ecbdata,
483 const char *line, int len)
484 {
485 const char *ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
486 const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_NEW);
487
488 if (!*ws)
489 emit_line_0(ecbdata->opt, set, reset, '+', line, len);
490 else if (new_blank_line_at_eof(ecbdata, line, len))
491 /* Blank line at EOF - paint '+' as well */
492 emit_line_0(ecbdata->opt, ws, reset, '+', line, len);
493 else {
494 /* Emit just the prefix, then the rest. */
495 emit_line_0(ecbdata->opt, set, reset, '+', "", 0);
496 ws_check_emit(line, len, ecbdata->ws_rule,
497 ecbdata->opt->file, set, reset, ws);
498 }
499 }
500
501 static void emit_del_line(const char *reset,
502 struct emit_callback *ecbdata,
503 const char *line, int len)
504 {
505 const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_OLD);
506
507 emit_line_0(ecbdata->opt, set, reset, '-', line, len);
508 }
509
510 static void emit_context_line(const char *reset,
511 struct emit_callback *ecbdata,
512 const char *line, int len)
513 {
514 const char *set = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
515
516 emit_line_0(ecbdata->opt, set, reset, ' ', line, len);
517 }
518
519 static void emit_hunk_header(struct emit_callback *ecbdata,
520 const char *line, int len)
521 {
522 const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
523 const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
524 const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
525 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
526 static const char atat[2] = { '@', '@' };
527 const char *cp, *ep;
528 struct strbuf msgbuf = STRBUF_INIT;
529 int org_len = len;
530 int i = 1;
531
532 /*
533 * As a hunk header must begin with "@@ -<old>, +<new> @@",
534 * it always is at least 10 bytes long.
535 */
536 if (len < 10 ||
537 memcmp(line, atat, 2) ||
538 !(ep = memmem(line + 2, len - 2, atat, 2))) {
539 emit_line(ecbdata->opt, plain, reset, line, len);
540 return;
541 }
542 ep += 2; /* skip over @@ */
543
544 /* The hunk header in fraginfo color */
545 strbuf_addstr(&msgbuf, frag);
546 strbuf_add(&msgbuf, line, ep - line);
547 strbuf_addstr(&msgbuf, reset);
548
549 /*
550 * trailing "\r\n"
551 */
552 for ( ; i < 3; i++)
553 if (line[len - i] == '\r' || line[len - i] == '\n')
554 len--;
555
556 /* blank before the func header */
557 for (cp = ep; ep - line < len; ep++)
558 if (*ep != ' ' && *ep != '\t')
559 break;
560 if (ep != cp) {
561 strbuf_addstr(&msgbuf, plain);
562 strbuf_add(&msgbuf, cp, ep - cp);
563 strbuf_addstr(&msgbuf, reset);
564 }
565
566 if (ep < line + len) {
567 strbuf_addstr(&msgbuf, func);
568 strbuf_add(&msgbuf, ep, line + len - ep);
569 strbuf_addstr(&msgbuf, reset);
570 }
571
572 strbuf_add(&msgbuf, line + len, org_len - len);
573 emit_line(ecbdata->opt, "", "", msgbuf.buf, msgbuf.len);
574 strbuf_release(&msgbuf);
575 }
576
577 static struct diff_tempfile *claim_diff_tempfile(void) {
578 int i;
579 for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
580 if (!diff_temp[i].name)
581 return diff_temp + i;
582 die("BUG: diff is failing to clean up its tempfiles");
583 }
584
585 static int remove_tempfile_installed;
586
587 static void remove_tempfile(void)
588 {
589 int i;
590 for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
591 if (diff_temp[i].name == diff_temp[i].tmp_path)
592 unlink_or_warn(diff_temp[i].name);
593 diff_temp[i].name = NULL;
594 }
595 }
596
597 static void remove_tempfile_on_signal(int signo)
598 {
599 remove_tempfile();
600 sigchain_pop(signo);
601 raise(signo);
602 }
603
604 static void print_line_count(FILE *file, int count)
605 {
606 switch (count) {
607 case 0:
608 fprintf(file, "0,0");
609 break;
610 case 1:
611 fprintf(file, "1");
612 break;
613 default:
614 fprintf(file, "1,%d", count);
615 break;
616 }
617 }
618
619 static void emit_rewrite_lines(struct emit_callback *ecb,
620 int prefix, const char *data, int size)
621 {
622 const char *endp = NULL;
623 static const char *nneof = " No newline at end of file\n";
624 const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
625
626 while (0 < size) {
627 int len;
628
629 endp = memchr(data, '\n', size);
630 len = endp ? (endp - data + 1) : size;
631 if (prefix != '+') {
632 ecb->lno_in_preimage++;
633 emit_del_line(reset, ecb, data, len);
634 } else {
635 ecb->lno_in_postimage++;
636 emit_add_line(reset, ecb, data, len);
637 }
638 size -= len;
639 data += len;
640 }
641 if (!endp) {
642 const char *plain = diff_get_color(ecb->color_diff,
643 DIFF_PLAIN);
644 putc('\n', ecb->opt->file);
645 emit_line_0(ecb->opt, plain, reset, '\\',
646 nneof, strlen(nneof));
647 }
648 }
649
650 static void emit_rewrite_diff(const char *name_a,
651 const char *name_b,
652 struct diff_filespec *one,
653 struct diff_filespec *two,
654 struct userdiff_driver *textconv_one,
655 struct userdiff_driver *textconv_two,
656 struct diff_options *o)
657 {
658 int lc_a, lc_b;
659 const char *name_a_tab, *name_b_tab;
660 const char *metainfo = diff_get_color(o->use_color, DIFF_METAINFO);
661 const char *fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
662 const char *reset = diff_get_color(o->use_color, DIFF_RESET);
663 static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
664 const char *a_prefix, *b_prefix;
665 char *data_one, *data_two;
666 size_t size_one, size_two;
667 struct emit_callback ecbdata;
668 const char *line_prefix = diff_line_prefix(o);
669
670 if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
671 a_prefix = o->b_prefix;
672 b_prefix = o->a_prefix;
673 } else {
674 a_prefix = o->a_prefix;
675 b_prefix = o->b_prefix;
676 }
677
678 name_a += (*name_a == '/');
679 name_b += (*name_b == '/');
680 name_a_tab = strchr(name_a, ' ') ? "\t" : "";
681 name_b_tab = strchr(name_b, ' ') ? "\t" : "";
682
683 strbuf_reset(&a_name);
684 strbuf_reset(&b_name);
685 quote_two_c_style(&a_name, a_prefix, name_a, 0);
686 quote_two_c_style(&b_name, b_prefix, name_b, 0);
687
688 size_one = fill_textconv(textconv_one, one, &data_one);
689 size_two = fill_textconv(textconv_two, two, &data_two);
690
691 memset(&ecbdata, 0, sizeof(ecbdata));
692 ecbdata.color_diff = want_color(o->use_color);
693 ecbdata.found_changesp = &o->found_changes;
694 ecbdata.ws_rule = whitespace_rule(name_b);
695 ecbdata.opt = o;
696 if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
697 mmfile_t mf1, mf2;
698 mf1.ptr = (char *)data_one;
699 mf2.ptr = (char *)data_two;
700 mf1.size = size_one;
701 mf2.size = size_two;
702 check_blank_at_eof(&mf1, &mf2, &ecbdata);
703 }
704 ecbdata.lno_in_preimage = 1;
705 ecbdata.lno_in_postimage = 1;
706
707 lc_a = count_lines(data_one, size_one);
708 lc_b = count_lines(data_two, size_two);
709 fprintf(o->file,
710 "%s%s--- %s%s%s\n%s%s+++ %s%s%s\n%s%s@@ -",
711 line_prefix, metainfo, a_name.buf, name_a_tab, reset,
712 line_prefix, metainfo, b_name.buf, name_b_tab, reset,
713 line_prefix, fraginfo);
714 if (!o->irreversible_delete)
715 print_line_count(o->file, lc_a);
716 else
717 fprintf(o->file, "?,?");
718 fprintf(o->file, " +");
719 print_line_count(o->file, lc_b);
720 fprintf(o->file, " @@%s\n", reset);
721 if (lc_a && !o->irreversible_delete)
722 emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
723 if (lc_b)
724 emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
725 if (textconv_one)
726 free((char *)data_one);
727 if (textconv_two)
728 free((char *)data_two);
729 }
730
731 struct diff_words_buffer {
732 mmfile_t text;
733 long alloc;
734 struct diff_words_orig {
735 const char *begin, *end;
736 } *orig;
737 int orig_nr, orig_alloc;
738 };
739
740 static void diff_words_append(char *line, unsigned long len,
741 struct diff_words_buffer *buffer)
742 {
743 ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
744 line++;
745 len--;
746 memcpy(buffer->text.ptr + buffer->text.size, line, len);
747 buffer->text.size += len;
748 buffer->text.ptr[buffer->text.size] = '\0';
749 }
750
751 struct diff_words_style_elem {
752 const char *prefix;
753 const char *suffix;
754 const char *color; /* NULL; filled in by the setup code if
755 * color is enabled */
756 };
757
758 struct diff_words_style {
759 enum diff_words_type type;
760 struct diff_words_style_elem new, old, ctx;
761 const char *newline;
762 };
763
764 static struct diff_words_style diff_words_styles[] = {
765 { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
766 { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
767 { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
768 };
769
770 struct diff_words_data {
771 struct diff_words_buffer minus, plus;
772 const char *current_plus;
773 int last_minus;
774 struct diff_options *opt;
775 regex_t *word_regex;
776 enum diff_words_type type;
777 struct diff_words_style *style;
778 };
779
780 static int fn_out_diff_words_write_helper(FILE *fp,
781 struct diff_words_style_elem *st_el,
782 const char *newline,
783 size_t count, const char *buf,
784 const char *line_prefix)
785 {
786 int print = 0;
787
788 while (count) {
789 char *p = memchr(buf, '\n', count);
790 if (print)
791 fputs(line_prefix, fp);
792 if (p != buf) {
793 if (st_el->color && fputs(st_el->color, fp) < 0)
794 return -1;
795 if (fputs(st_el->prefix, fp) < 0 ||
796 fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
797 fputs(st_el->suffix, fp) < 0)
798 return -1;
799 if (st_el->color && *st_el->color
800 && fputs(GIT_COLOR_RESET, fp) < 0)
801 return -1;
802 }
803 if (!p)
804 return 0;
805 if (fputs(newline, fp) < 0)
806 return -1;
807 count -= p + 1 - buf;
808 buf = p + 1;
809 print = 1;
810 }
811 return 0;
812 }
813
814 /*
815 * '--color-words' algorithm can be described as:
816 *
817 * 1. collect a the minus/plus lines of a diff hunk, divided into
818 * minus-lines and plus-lines;
819 *
820 * 2. break both minus-lines and plus-lines into words and
821 * place them into two mmfile_t with one word for each line;
822 *
823 * 3. use xdiff to run diff on the two mmfile_t to get the words level diff;
824 *
825 * And for the common parts of the both file, we output the plus side text.
826 * diff_words->current_plus is used to trace the current position of the plus file
827 * which printed. diff_words->last_minus is used to trace the last minus word
828 * printed.
829 *
830 * For '--graph' to work with '--color-words', we need to output the graph prefix
831 * on each line of color words output. Generally, there are two conditions on
832 * which we should output the prefix.
833 *
834 * 1. diff_words->last_minus == 0 &&
835 * diff_words->current_plus == diff_words->plus.text.ptr
836 *
837 * that is: the plus text must start as a new line, and if there is no minus
838 * word printed, a graph prefix must be printed.
839 *
840 * 2. diff_words->current_plus > diff_words->plus.text.ptr &&
841 * *(diff_words->current_plus - 1) == '\n'
842 *
843 * that is: a graph prefix must be printed following a '\n'
844 */
845 static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
846 {
847 if ((diff_words->last_minus == 0 &&
848 diff_words->current_plus == diff_words->plus.text.ptr) ||
849 (diff_words->current_plus > diff_words->plus.text.ptr &&
850 *(diff_words->current_plus - 1) == '\n')) {
851 return 1;
852 } else {
853 return 0;
854 }
855 }
856
857 static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
858 {
859 struct diff_words_data *diff_words = priv;
860 struct diff_words_style *style = diff_words->style;
861 int minus_first, minus_len, plus_first, plus_len;
862 const char *minus_begin, *minus_end, *plus_begin, *plus_end;
863 struct diff_options *opt = diff_words->opt;
864 const char *line_prefix;
865
866 if (line[0] != '@' || parse_hunk_header(line, len,
867 &minus_first, &minus_len, &plus_first, &plus_len))
868 return;
869
870 assert(opt);
871 line_prefix = diff_line_prefix(opt);
872
873 /* POSIX requires that first be decremented by one if len == 0... */
874 if (minus_len) {
875 minus_begin = diff_words->minus.orig[minus_first].begin;
876 minus_end =
877 diff_words->minus.orig[minus_first + minus_len - 1].end;
878 } else
879 minus_begin = minus_end =
880 diff_words->minus.orig[minus_first].end;
881
882 if (plus_len) {
883 plus_begin = diff_words->plus.orig[plus_first].begin;
884 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
885 } else
886 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
887
888 if (color_words_output_graph_prefix(diff_words)) {
889 fputs(line_prefix, diff_words->opt->file);
890 }
891 if (diff_words->current_plus != plus_begin) {
892 fn_out_diff_words_write_helper(diff_words->opt->file,
893 &style->ctx, style->newline,
894 plus_begin - diff_words->current_plus,
895 diff_words->current_plus, line_prefix);
896 if (*(plus_begin - 1) == '\n')
897 fputs(line_prefix, diff_words->opt->file);
898 }
899 if (minus_begin != minus_end) {
900 fn_out_diff_words_write_helper(diff_words->opt->file,
901 &style->old, style->newline,
902 minus_end - minus_begin, minus_begin,
903 line_prefix);
904 }
905 if (plus_begin != plus_end) {
906 fn_out_diff_words_write_helper(diff_words->opt->file,
907 &style->new, style->newline,
908 plus_end - plus_begin, plus_begin,
909 line_prefix);
910 }
911
912 diff_words->current_plus = plus_end;
913 diff_words->last_minus = minus_first;
914 }
915
916 /* This function starts looking at *begin, and returns 0 iff a word was found. */
917 static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
918 int *begin, int *end)
919 {
920 if (word_regex && *begin < buffer->size) {
921 regmatch_t match[1];
922 if (!regexec(word_regex, buffer->ptr + *begin, 1, match, 0)) {
923 char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
924 '\n', match[0].rm_eo - match[0].rm_so);
925 *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
926 *begin += match[0].rm_so;
927 return *begin >= *end;
928 }
929 return -1;
930 }
931
932 /* find the next word */
933 while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
934 (*begin)++;
935 if (*begin >= buffer->size)
936 return -1;
937
938 /* find the end of the word */
939 *end = *begin + 1;
940 while (*end < buffer->size && !isspace(buffer->ptr[*end]))
941 (*end)++;
942
943 return 0;
944 }
945
946 /*
947 * This function splits the words in buffer->text, stores the list with
948 * newline separator into out, and saves the offsets of the original words
949 * in buffer->orig.
950 */
951 static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
952 regex_t *word_regex)
953 {
954 int i, j;
955 long alloc = 0;
956
957 out->size = 0;
958 out->ptr = NULL;
959
960 /* fake an empty "0th" word */
961 ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
962 buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
963 buffer->orig_nr = 1;
964
965 for (i = 0; i < buffer->text.size; i++) {
966 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
967 return;
968
969 /* store original boundaries */
970 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
971 buffer->orig_alloc);
972 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
973 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
974 buffer->orig_nr++;
975
976 /* store one word */
977 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
978 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
979 out->ptr[out->size + j - i] = '\n';
980 out->size += j - i + 1;
981
982 i = j - 1;
983 }
984 }
985
986 /* this executes the word diff on the accumulated buffers */
987 static void diff_words_show(struct diff_words_data *diff_words)
988 {
989 xpparam_t xpp;
990 xdemitconf_t xecfg;
991 mmfile_t minus, plus;
992 struct diff_words_style *style = diff_words->style;
993
994 struct diff_options *opt = diff_words->opt;
995 const char *line_prefix;
996
997 assert(opt);
998 line_prefix = diff_line_prefix(opt);
999
1000 /* special case: only removal */
1001 if (!diff_words->plus.text.size) {
1002 fputs(line_prefix, diff_words->opt->file);
1003 fn_out_diff_words_write_helper(diff_words->opt->file,
1004 &style->old, style->newline,
1005 diff_words->minus.text.size,
1006 diff_words->minus.text.ptr, line_prefix);
1007 diff_words->minus.text.size = 0;
1008 return;
1009 }
1010
1011 diff_words->current_plus = diff_words->plus.text.ptr;
1012 diff_words->last_minus = 0;
1013
1014 memset(&xpp, 0, sizeof(xpp));
1015 memset(&xecfg, 0, sizeof(xecfg));
1016 diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
1017 diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
1018 xpp.flags = 0;
1019 /* as only the hunk header will be parsed, we need a 0-context */
1020 xecfg.ctxlen = 0;
1021 xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
1022 &xpp, &xecfg);
1023 free(minus.ptr);
1024 free(plus.ptr);
1025 if (diff_words->current_plus != diff_words->plus.text.ptr +
1026 diff_words->plus.text.size) {
1027 if (color_words_output_graph_prefix(diff_words))
1028 fputs(line_prefix, diff_words->opt->file);
1029 fn_out_diff_words_write_helper(diff_words->opt->file,
1030 &style->ctx, style->newline,
1031 diff_words->plus.text.ptr + diff_words->plus.text.size
1032 - diff_words->current_plus, diff_words->current_plus,
1033 line_prefix);
1034 }
1035 diff_words->minus.text.size = diff_words->plus.text.size = 0;
1036 }
1037
1038 /* In "color-words" mode, show word-diff of words accumulated in the buffer */
1039 static void diff_words_flush(struct emit_callback *ecbdata)
1040 {
1041 if (ecbdata->diff_words->minus.text.size ||
1042 ecbdata->diff_words->plus.text.size)
1043 diff_words_show(ecbdata->diff_words);
1044 }
1045
1046 static void diff_filespec_load_driver(struct diff_filespec *one)
1047 {
1048 /* Use already-loaded driver */
1049 if (one->driver)
1050 return;
1051
1052 if (S_ISREG(one->mode))
1053 one->driver = userdiff_find_by_path(one->path);
1054
1055 /* Fallback to default settings */
1056 if (!one->driver)
1057 one->driver = userdiff_find_by_name("default");
1058 }
1059
1060 static const char *userdiff_word_regex(struct diff_filespec *one)
1061 {
1062 diff_filespec_load_driver(one);
1063 return one->driver->word_regex;
1064 }
1065
1066 static void init_diff_words_data(struct emit_callback *ecbdata,
1067 struct diff_options *orig_opts,
1068 struct diff_filespec *one,
1069 struct diff_filespec *two)
1070 {
1071 int i;
1072 struct diff_options *o = xmalloc(sizeof(struct diff_options));
1073 memcpy(o, orig_opts, sizeof(struct diff_options));
1074
1075 ecbdata->diff_words =
1076 xcalloc(1, sizeof(struct diff_words_data));
1077 ecbdata->diff_words->type = o->word_diff;
1078 ecbdata->diff_words->opt = o;
1079 if (!o->word_regex)
1080 o->word_regex = userdiff_word_regex(one);
1081 if (!o->word_regex)
1082 o->word_regex = userdiff_word_regex(two);
1083 if (!o->word_regex)
1084 o->word_regex = diff_word_regex_cfg;
1085 if (o->word_regex) {
1086 ecbdata->diff_words->word_regex = (regex_t *)
1087 xmalloc(sizeof(regex_t));
1088 if (regcomp(ecbdata->diff_words->word_regex,
1089 o->word_regex,
1090 REG_EXTENDED | REG_NEWLINE))
1091 die ("Invalid regular expression: %s",
1092 o->word_regex);
1093 }
1094 for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1095 if (o->word_diff == diff_words_styles[i].type) {
1096 ecbdata->diff_words->style =
1097 &diff_words_styles[i];
1098 break;
1099 }
1100 }
1101 if (want_color(o->use_color)) {
1102 struct diff_words_style *st = ecbdata->diff_words->style;
1103 st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1104 st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1105 st->ctx.color = diff_get_color_opt(o, DIFF_PLAIN);
1106 }
1107 }
1108
1109 static void free_diff_words_data(struct emit_callback *ecbdata)
1110 {
1111 if (ecbdata->diff_words) {
1112 diff_words_flush(ecbdata);
1113 free (ecbdata->diff_words->opt);
1114 free (ecbdata->diff_words->minus.text.ptr);
1115 free (ecbdata->diff_words->minus.orig);
1116 free (ecbdata->diff_words->plus.text.ptr);
1117 free (ecbdata->diff_words->plus.orig);
1118 if (ecbdata->diff_words->word_regex) {
1119 regfree(ecbdata->diff_words->word_regex);
1120 free(ecbdata->diff_words->word_regex);
1121 }
1122 free(ecbdata->diff_words);
1123 ecbdata->diff_words = NULL;
1124 }
1125 }
1126
1127 const char *diff_get_color(int diff_use_color, enum color_diff ix)
1128 {
1129 if (want_color(diff_use_color))
1130 return diff_colors[ix];
1131 return "";
1132 }
1133
1134 const char *diff_line_prefix(struct diff_options *opt)
1135 {
1136 struct strbuf *msgbuf;
1137 if (!opt->output_prefix)
1138 return "";
1139
1140 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
1141 return msgbuf->buf;
1142 }
1143
1144 static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1145 {
1146 const char *cp;
1147 unsigned long allot;
1148 size_t l = len;
1149
1150 if (ecb->truncate)
1151 return ecb->truncate(line, len);
1152 cp = line;
1153 allot = l;
1154 while (0 < l) {
1155 (void) utf8_width(&cp, &l);
1156 if (!cp)
1157 break; /* truncated in the middle? */
1158 }
1159 return allot - l;
1160 }
1161
1162 static void find_lno(const char *line, struct emit_callback *ecbdata)
1163 {
1164 const char *p;
1165 ecbdata->lno_in_preimage = 0;
1166 ecbdata->lno_in_postimage = 0;
1167 p = strchr(line, '-');
1168 if (!p)
1169 return; /* cannot happen */
1170 ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1171 p = strchr(p, '+');
1172 if (!p)
1173 return; /* cannot happen */
1174 ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1175 }
1176
1177 static void fn_out_consume(void *priv, char *line, unsigned long len)
1178 {
1179 struct emit_callback *ecbdata = priv;
1180 const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
1181 const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
1182 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1183 struct diff_options *o = ecbdata->opt;
1184 const char *line_prefix = diff_line_prefix(o);
1185
1186 if (ecbdata->header) {
1187 fprintf(ecbdata->opt->file, "%s", ecbdata->header->buf);
1188 strbuf_reset(ecbdata->header);
1189 ecbdata->header = NULL;
1190 }
1191 *(ecbdata->found_changesp) = 1;
1192
1193 if (ecbdata->label_path[0]) {
1194 const char *name_a_tab, *name_b_tab;
1195
1196 name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
1197 name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
1198
1199 fprintf(ecbdata->opt->file, "%s%s--- %s%s%s\n",
1200 line_prefix, meta, ecbdata->label_path[0], reset, name_a_tab);
1201 fprintf(ecbdata->opt->file, "%s%s+++ %s%s%s\n",
1202 line_prefix, meta, ecbdata->label_path[1], reset, name_b_tab);
1203 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1204 }
1205
1206 if (diff_suppress_blank_empty
1207 && len == 2 && line[0] == ' ' && line[1] == '\n') {
1208 line[0] = '\n';
1209 len = 1;
1210 }
1211
1212 if (line[0] == '@') {
1213 if (ecbdata->diff_words)
1214 diff_words_flush(ecbdata);
1215 len = sane_truncate_line(ecbdata, line, len);
1216 find_lno(line, ecbdata);
1217 emit_hunk_header(ecbdata, line, len);
1218 if (line[len-1] != '\n')
1219 putc('\n', ecbdata->opt->file);
1220 return;
1221 }
1222
1223 if (len < 1) {
1224 emit_line(ecbdata->opt, reset, reset, line, len);
1225 if (ecbdata->diff_words
1226 && ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN)
1227 fputs("~\n", ecbdata->opt->file);
1228 return;
1229 }
1230
1231 if (ecbdata->diff_words) {
1232 if (line[0] == '-') {
1233 diff_words_append(line, len,
1234 &ecbdata->diff_words->minus);
1235 return;
1236 } else if (line[0] == '+') {
1237 diff_words_append(line, len,
1238 &ecbdata->diff_words->plus);
1239 return;
1240 } else if (starts_with(line, "\\ ")) {
1241 /*
1242 * Eat the "no newline at eof" marker as if we
1243 * saw a "+" or "-" line with nothing on it,
1244 * and return without diff_words_flush() to
1245 * defer processing. If this is the end of
1246 * preimage, more "+" lines may come after it.
1247 */
1248 return;
1249 }
1250 diff_words_flush(ecbdata);
1251 if (ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN) {
1252 emit_line(ecbdata->opt, plain, reset, line, len);
1253 fputs("~\n", ecbdata->opt->file);
1254 } else {
1255 /*
1256 * Skip the prefix character, if any. With
1257 * diff_suppress_blank_empty, there may be
1258 * none.
1259 */
1260 if (line[0] != '\n') {
1261 line++;
1262 len--;
1263 }
1264 emit_line(ecbdata->opt, plain, reset, line, len);
1265 }
1266 return;
1267 }
1268
1269 switch (line[0]) {
1270 case '+':
1271 ecbdata->lno_in_postimage++;
1272 emit_add_line(reset, ecbdata, line + 1, len - 1);
1273 break;
1274 case '-':
1275 ecbdata->lno_in_preimage++;
1276 emit_del_line(reset, ecbdata, line + 1, len - 1);
1277 break;
1278 case ' ':
1279 ecbdata->lno_in_postimage++;
1280 ecbdata->lno_in_preimage++;
1281 emit_context_line(reset, ecbdata, line + 1, len - 1);
1282 break;
1283 default:
1284 /* incomplete line at the end */
1285 ecbdata->lno_in_preimage++;
1286 emit_line(ecbdata->opt,
1287 diff_get_color(ecbdata->color_diff, DIFF_PLAIN),
1288 reset, line, len);
1289 break;
1290 }
1291 }
1292
1293 static char *pprint_rename(const char *a, const char *b)
1294 {
1295 const char *old = a;
1296 const char *new = b;
1297 struct strbuf name = STRBUF_INIT;
1298 int pfx_length, sfx_length;
1299 int pfx_adjust_for_slash;
1300 int len_a = strlen(a);
1301 int len_b = strlen(b);
1302 int a_midlen, b_midlen;
1303 int qlen_a = quote_c_style(a, NULL, NULL, 0);
1304 int qlen_b = quote_c_style(b, NULL, NULL, 0);
1305
1306 if (qlen_a || qlen_b) {
1307 quote_c_style(a, &name, NULL, 0);
1308 strbuf_addstr(&name, " => ");
1309 quote_c_style(b, &name, NULL, 0);
1310 return strbuf_detach(&name, NULL);
1311 }
1312
1313 /* Find common prefix */
1314 pfx_length = 0;
1315 while (*old && *new && *old == *new) {
1316 if (*old == '/')
1317 pfx_length = old - a + 1;
1318 old++;
1319 new++;
1320 }
1321
1322 /* Find common suffix */
1323 old = a + len_a;
1324 new = b + len_b;
1325 sfx_length = 0;
1326 /*
1327 * If there is a common prefix, it must end in a slash. In
1328 * that case we let this loop run 1 into the prefix to see the
1329 * same slash.
1330 *
1331 * If there is no common prefix, we cannot do this as it would
1332 * underrun the input strings.
1333 */
1334 pfx_adjust_for_slash = (pfx_length ? 1 : 0);
1335 while (a + pfx_length - pfx_adjust_for_slash <= old &&
1336 b + pfx_length - pfx_adjust_for_slash <= new &&
1337 *old == *new) {
1338 if (*old == '/')
1339 sfx_length = len_a - (old - a);
1340 old--;
1341 new--;
1342 }
1343
1344 /*
1345 * pfx{mid-a => mid-b}sfx
1346 * {pfx-a => pfx-b}sfx
1347 * pfx{sfx-a => sfx-b}
1348 * name-a => name-b
1349 */
1350 a_midlen = len_a - pfx_length - sfx_length;
1351 b_midlen = len_b - pfx_length - sfx_length;
1352 if (a_midlen < 0)
1353 a_midlen = 0;
1354 if (b_midlen < 0)
1355 b_midlen = 0;
1356
1357 strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1358 if (pfx_length + sfx_length) {
1359 strbuf_add(&name, a, pfx_length);
1360 strbuf_addch(&name, '{');
1361 }
1362 strbuf_add(&name, a + pfx_length, a_midlen);
1363 strbuf_addstr(&name, " => ");
1364 strbuf_add(&name, b + pfx_length, b_midlen);
1365 if (pfx_length + sfx_length) {
1366 strbuf_addch(&name, '}');
1367 strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1368 }
1369 return strbuf_detach(&name, NULL);
1370 }
1371
1372 struct diffstat_t {
1373 int nr;
1374 int alloc;
1375 struct diffstat_file {
1376 char *from_name;
1377 char *name;
1378 char *print_name;
1379 unsigned is_unmerged:1;
1380 unsigned is_binary:1;
1381 unsigned is_renamed:1;
1382 unsigned is_interesting:1;
1383 uintmax_t added, deleted;
1384 } **files;
1385 };
1386
1387 static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1388 const char *name_a,
1389 const char *name_b)
1390 {
1391 struct diffstat_file *x;
1392 x = xcalloc(1, sizeof(*x));
1393 ALLOC_GROW(diffstat->files, diffstat->nr + 1, diffstat->alloc);
1394 diffstat->files[diffstat->nr++] = x;
1395 if (name_b) {
1396 x->from_name = xstrdup(name_a);
1397 x->name = xstrdup(name_b);
1398 x->is_renamed = 1;
1399 }
1400 else {
1401 x->from_name = NULL;
1402 x->name = xstrdup(name_a);
1403 }
1404 return x;
1405 }
1406
1407 static void diffstat_consume(void *priv, char *line, unsigned long len)
1408 {
1409 struct diffstat_t *diffstat = priv;
1410 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1411
1412 if (line[0] == '+')
1413 x->added++;
1414 else if (line[0] == '-')
1415 x->deleted++;
1416 }
1417
1418 const char mime_boundary_leader[] = "------------";
1419
1420 static int scale_linear(int it, int width, int max_change)
1421 {
1422 if (!it)
1423 return 0;
1424 /*
1425 * make sure that at least one '-' or '+' is printed if
1426 * there is any change to this path. The easiest way is to
1427 * scale linearly as if the alloted width is one column shorter
1428 * than it is, and then add 1 to the result.
1429 */
1430 return 1 + (it * (width - 1) / max_change);
1431 }
1432
1433 static void show_name(FILE *file,
1434 const char *prefix, const char *name, int len)
1435 {
1436 fprintf(file, " %s%-*s |", prefix, len, name);
1437 }
1438
1439 static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1440 {
1441 if (cnt <= 0)
1442 return;
1443 fprintf(file, "%s", set);
1444 while (cnt--)
1445 putc(ch, file);
1446 fprintf(file, "%s", reset);
1447 }
1448
1449 static void fill_print_name(struct diffstat_file *file)
1450 {
1451 char *pname;
1452
1453 if (file->print_name)
1454 return;
1455
1456 if (!file->is_renamed) {
1457 struct strbuf buf = STRBUF_INIT;
1458 if (quote_c_style(file->name, &buf, NULL, 0)) {
1459 pname = strbuf_detach(&buf, NULL);
1460 } else {
1461 pname = file->name;
1462 strbuf_release(&buf);
1463 }
1464 } else {
1465 pname = pprint_rename(file->from_name, file->name);
1466 }
1467 file->print_name = pname;
1468 }
1469
1470 int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1471 {
1472 struct strbuf sb = STRBUF_INIT;
1473 int ret;
1474
1475 if (!files) {
1476 assert(insertions == 0 && deletions == 0);
1477 return fprintf(fp, "%s\n", " 0 files changed");
1478 }
1479
1480 strbuf_addf(&sb,
1481 (files == 1) ? " %d file changed" : " %d files changed",
1482 files);
1483
1484 /*
1485 * For binary diff, the caller may want to print "x files
1486 * changed" with insertions == 0 && deletions == 0.
1487 *
1488 * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1489 * is probably less confusing (i.e skip over "2 files changed
1490 * but nothing about added/removed lines? Is this a bug in Git?").
1491 */
1492 if (insertions || deletions == 0) {
1493 strbuf_addf(&sb,
1494 (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
1495 insertions);
1496 }
1497
1498 if (deletions || insertions == 0) {
1499 strbuf_addf(&sb,
1500 (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
1501 deletions);
1502 }
1503 strbuf_addch(&sb, '\n');
1504 ret = fputs(sb.buf, fp);
1505 strbuf_release(&sb);
1506 return ret;
1507 }
1508
1509 static void show_stats(struct diffstat_t *data, struct diff_options *options)
1510 {
1511 int i, len, add, del, adds = 0, dels = 0;
1512 uintmax_t max_change = 0, max_len = 0;
1513 int total_files = data->nr, count;
1514 int width, name_width, graph_width, number_width = 0, bin_width = 0;
1515 const char *reset, *add_c, *del_c;
1516 const char *line_prefix = "";
1517 int extra_shown = 0;
1518
1519 if (data->nr == 0)
1520 return;
1521
1522 line_prefix = diff_line_prefix(options);
1523 count = options->stat_count ? options->stat_count : data->nr;
1524
1525 reset = diff_get_color_opt(options, DIFF_RESET);
1526 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1527 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1528
1529 /*
1530 * Find the longest filename and max number of changes
1531 */
1532 for (i = 0; (i < count) && (i < data->nr); i++) {
1533 struct diffstat_file *file = data->files[i];
1534 uintmax_t change = file->added + file->deleted;
1535
1536 if (!file->is_interesting && (change == 0)) {
1537 count++; /* not shown == room for one more */
1538 continue;
1539 }
1540 fill_print_name(file);
1541 len = strlen(file->print_name);
1542 if (max_len < len)
1543 max_len = len;
1544
1545 if (file->is_unmerged) {
1546 /* "Unmerged" is 8 characters */
1547 bin_width = bin_width < 8 ? 8 : bin_width;
1548 continue;
1549 }
1550 if (file->is_binary) {
1551 /* "Bin XXX -> YYY bytes" */
1552 int w = 14 + decimal_width(file->added)
1553 + decimal_width(file->deleted);
1554 bin_width = bin_width < w ? w : bin_width;
1555 /* Display change counts aligned with "Bin" */
1556 number_width = 3;
1557 continue;
1558 }
1559
1560 if (max_change < change)
1561 max_change = change;
1562 }
1563 count = i; /* where we can stop scanning in data->files[] */
1564
1565 /*
1566 * We have width = stat_width or term_columns() columns total.
1567 * We want a maximum of min(max_len, stat_name_width) for the name part.
1568 * We want a maximum of min(max_change, stat_graph_width) for the +- part.
1569 * We also need 1 for " " and 4 + decimal_width(max_change)
1570 * for " | NNNN " and one the empty column at the end, altogether
1571 * 6 + decimal_width(max_change).
1572 *
1573 * If there's not enough space, we will use the smaller of
1574 * stat_name_width (if set) and 5/8*width for the filename,
1575 * and the rest for constant elements + graph part, but no more
1576 * than stat_graph_width for the graph part.
1577 * (5/8 gives 50 for filename and 30 for the constant parts + graph
1578 * for the standard terminal size).
1579 *
1580 * In other words: stat_width limits the maximum width, and
1581 * stat_name_width fixes the maximum width of the filename,
1582 * and is also used to divide available columns if there
1583 * aren't enough.
1584 *
1585 * Binary files are displayed with "Bin XXX -> YYY bytes"
1586 * instead of the change count and graph. This part is treated
1587 * similarly to the graph part, except that it is not
1588 * "scaled". If total width is too small to accommodate the
1589 * guaranteed minimum width of the filename part and the
1590 * separators and this message, this message will "overflow"
1591 * making the line longer than the maximum width.
1592 */
1593
1594 if (options->stat_width == -1)
1595 width = term_columns() - options->output_prefix_length;
1596 else
1597 width = options->stat_width ? options->stat_width : 80;
1598 number_width = decimal_width(max_change) > number_width ?
1599 decimal_width(max_change) : number_width;
1600
1601 if (options->stat_graph_width == -1)
1602 options->stat_graph_width = diff_stat_graph_width;
1603
1604 /*
1605 * Guarantee 3/8*16==6 for the graph part
1606 * and 5/8*16==10 for the filename part
1607 */
1608 if (width < 16 + 6 + number_width)
1609 width = 16 + 6 + number_width;
1610
1611 /*
1612 * First assign sizes that are wanted, ignoring available width.
1613 * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
1614 * starting from "XXX" should fit in graph_width.
1615 */
1616 graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
1617 if (options->stat_graph_width &&
1618 options->stat_graph_width < graph_width)
1619 graph_width = options->stat_graph_width;
1620
1621 name_width = (options->stat_name_width > 0 &&
1622 options->stat_name_width < max_len) ?
1623 options->stat_name_width : max_len;
1624
1625 /*
1626 * Adjust adjustable widths not to exceed maximum width
1627 */
1628 if (name_width + number_width + 6 + graph_width > width) {
1629 if (graph_width > width * 3/8 - number_width - 6) {
1630 graph_width = width * 3/8 - number_width - 6;
1631 if (graph_width < 6)
1632 graph_width = 6;
1633 }
1634
1635 if (options->stat_graph_width &&
1636 graph_width > options->stat_graph_width)
1637 graph_width = options->stat_graph_width;
1638 if (name_width > width - number_width - 6 - graph_width)
1639 name_width = width - number_width - 6 - graph_width;
1640 else
1641 graph_width = width - number_width - 6 - name_width;
1642 }
1643
1644 /*
1645 * From here name_width is the width of the name area,
1646 * and graph_width is the width of the graph area.
1647 * max_change is used to scale graph properly.
1648 */
1649 for (i = 0; i < count; i++) {
1650 const char *prefix = "";
1651 struct diffstat_file *file = data->files[i];
1652 char *name = file->print_name;
1653 uintmax_t added = file->added;
1654 uintmax_t deleted = file->deleted;
1655 int name_len;
1656
1657 if (!file->is_interesting && (added + deleted == 0))
1658 continue;
1659
1660 /*
1661 * "scale" the filename
1662 */
1663 len = name_width;
1664 name_len = strlen(name);
1665 if (name_width < name_len) {
1666 char *slash;
1667 prefix = "...";
1668 len -= 3;
1669 name += name_len - len;
1670 slash = strchr(name, '/');
1671 if (slash)
1672 name = slash;
1673 }
1674
1675 if (file->is_binary) {
1676 fprintf(options->file, "%s", line_prefix);
1677 show_name(options->file, prefix, name, len);
1678 fprintf(options->file, " %*s", number_width, "Bin");
1679 if (!added && !deleted) {
1680 putc('\n', options->file);
1681 continue;
1682 }
1683 fprintf(options->file, " %s%"PRIuMAX"%s",
1684 del_c, deleted, reset);
1685 fprintf(options->file, " -> ");
1686 fprintf(options->file, "%s%"PRIuMAX"%s",
1687 add_c, added, reset);
1688 fprintf(options->file, " bytes");
1689 fprintf(options->file, "\n");
1690 continue;
1691 }
1692 else if (file->is_unmerged) {
1693 fprintf(options->file, "%s", line_prefix);
1694 show_name(options->file, prefix, name, len);
1695 fprintf(options->file, " Unmerged\n");
1696 continue;
1697 }
1698
1699 /*
1700 * scale the add/delete
1701 */
1702 add = added;
1703 del = deleted;
1704
1705 if (graph_width <= max_change) {
1706 int total = scale_linear(add + del, graph_width, max_change);
1707 if (total < 2 && add && del)
1708 /* width >= 2 due to the sanity check */
1709 total = 2;
1710 if (add < del) {
1711 add = scale_linear(add, graph_width, max_change);
1712 del = total - add;
1713 } else {
1714 del = scale_linear(del, graph_width, max_change);
1715 add = total - del;
1716 }
1717 }
1718 fprintf(options->file, "%s", line_prefix);
1719 show_name(options->file, prefix, name, len);
1720 fprintf(options->file, " %*"PRIuMAX"%s",
1721 number_width, added + deleted,
1722 added + deleted ? " " : "");
1723 show_graph(options->file, '+', add, add_c, reset);
1724 show_graph(options->file, '-', del, del_c, reset);
1725 fprintf(options->file, "\n");
1726 }
1727
1728 for (i = 0; i < data->nr; i++) {
1729 struct diffstat_file *file = data->files[i];
1730 uintmax_t added = file->added;
1731 uintmax_t deleted = file->deleted;
1732
1733 if (file->is_unmerged ||
1734 (!file->is_interesting && (added + deleted == 0))) {
1735 total_files--;
1736 continue;
1737 }
1738
1739 if (!file->is_binary) {
1740 adds += added;
1741 dels += deleted;
1742 }
1743 if (i < count)
1744 continue;
1745 if (!extra_shown)
1746 fprintf(options->file, "%s ...\n", line_prefix);
1747 extra_shown = 1;
1748 }
1749 fprintf(options->file, "%s", line_prefix);
1750 print_stat_summary(options->file, total_files, adds, dels);
1751 }
1752
1753 static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1754 {
1755 int i, adds = 0, dels = 0, total_files = data->nr;
1756
1757 if (data->nr == 0)
1758 return;
1759
1760 for (i = 0; i < data->nr; i++) {
1761 int added = data->files[i]->added;
1762 int deleted= data->files[i]->deleted;
1763
1764 if (data->files[i]->is_unmerged ||
1765 (!data->files[i]->is_interesting && (added + deleted == 0))) {
1766 total_files--;
1767 } else if (!data->files[i]->is_binary) { /* don't count bytes */
1768 adds += added;
1769 dels += deleted;
1770 }
1771 }
1772 fprintf(options->file, "%s", diff_line_prefix(options));
1773 print_stat_summary(options->file, total_files, adds, dels);
1774 }
1775
1776 static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1777 {
1778 int i;
1779
1780 if (data->nr == 0)
1781 return;
1782
1783 for (i = 0; i < data->nr; i++) {
1784 struct diffstat_file *file = data->files[i];
1785
1786 fprintf(options->file, "%s", diff_line_prefix(options));
1787
1788 if (file->is_binary)
1789 fprintf(options->file, "-\t-\t");
1790 else
1791 fprintf(options->file,
1792 "%"PRIuMAX"\t%"PRIuMAX"\t",
1793 file->added, file->deleted);
1794 if (options->line_termination) {
1795 fill_print_name(file);
1796 if (!file->is_renamed)
1797 write_name_quoted(file->name, options->file,
1798 options->line_termination);
1799 else {
1800 fputs(file->print_name, options->file);
1801 putc(options->line_termination, options->file);
1802 }
1803 } else {
1804 if (file->is_renamed) {
1805 putc('\0', options->file);
1806 write_name_quoted(file->from_name, options->file, '\0');
1807 }
1808 write_name_quoted(file->name, options->file, '\0');
1809 }
1810 }
1811 }
1812
1813 struct dirstat_file {
1814 const char *name;
1815 unsigned long changed;
1816 };
1817
1818 struct dirstat_dir {
1819 struct dirstat_file *files;
1820 int alloc, nr, permille, cumulative;
1821 };
1822
1823 static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1824 unsigned long changed, const char *base, int baselen)
1825 {
1826 unsigned long this_dir = 0;
1827 unsigned int sources = 0;
1828 const char *line_prefix = diff_line_prefix(opt);
1829
1830 while (dir->nr) {
1831 struct dirstat_file *f = dir->files;
1832 int namelen = strlen(f->name);
1833 unsigned long this;
1834 char *slash;
1835
1836 if (namelen < baselen)
1837 break;
1838 if (memcmp(f->name, base, baselen))
1839 break;
1840 slash = strchr(f->name + baselen, '/');
1841 if (slash) {
1842 int newbaselen = slash + 1 - f->name;
1843 this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
1844 sources++;
1845 } else {
1846 this = f->changed;
1847 dir->files++;
1848 dir->nr--;
1849 sources += 2;
1850 }
1851 this_dir += this;
1852 }
1853
1854 /*
1855 * We don't report dirstat's for
1856 * - the top level
1857 * - or cases where everything came from a single directory
1858 * under this directory (sources == 1).
1859 */
1860 if (baselen && sources != 1) {
1861 if (this_dir) {
1862 int permille = this_dir * 1000 / changed;
1863 if (permille >= dir->permille) {
1864 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
1865 permille / 10, permille % 10, baselen, base);
1866 if (!dir->cumulative)
1867 return 0;
1868 }
1869 }
1870 }
1871 return this_dir;
1872 }
1873
1874 static int dirstat_compare(const void *_a, const void *_b)
1875 {
1876 const struct dirstat_file *a = _a;
1877 const struct dirstat_file *b = _b;
1878 return strcmp(a->name, b->name);
1879 }
1880
1881 static void show_dirstat(struct diff_options *options)
1882 {
1883 int i;
1884 unsigned long changed;
1885 struct dirstat_dir dir;
1886 struct diff_queue_struct *q = &diff_queued_diff;
1887
1888 dir.files = NULL;
1889 dir.alloc = 0;
1890 dir.nr = 0;
1891 dir.permille = options->dirstat_permille;
1892 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1893
1894 changed = 0;
1895 for (i = 0; i < q->nr; i++) {
1896 struct diff_filepair *p = q->queue[i];
1897 const char *name;
1898 unsigned long copied, added, damage;
1899 int content_changed;
1900
1901 name = p->two->path ? p->two->path : p->one->path;
1902
1903 if (p->one->sha1_valid && p->two->sha1_valid)
1904 content_changed = hashcmp(p->one->sha1, p->two->sha1);
1905 else
1906 content_changed = 1;
1907
1908 if (!content_changed) {
1909 /*
1910 * The SHA1 has not changed, so pre-/post-content is
1911 * identical. We can therefore skip looking at the
1912 * file contents altogether.
1913 */
1914 damage = 0;
1915 goto found_damage;
1916 }
1917
1918 if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
1919 /*
1920 * In --dirstat-by-file mode, we don't really need to
1921 * look at the actual file contents at all.
1922 * The fact that the SHA1 changed is enough for us to
1923 * add this file to the list of results
1924 * (with each file contributing equal damage).
1925 */
1926 damage = 1;
1927 goto found_damage;
1928 }
1929
1930 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
1931 diff_populate_filespec(p->one, 0);
1932 diff_populate_filespec(p->two, 0);
1933 diffcore_count_changes(p->one, p->two, NULL, NULL, 0,
1934 &copied, &added);
1935 diff_free_filespec_data(p->one);
1936 diff_free_filespec_data(p->two);
1937 } else if (DIFF_FILE_VALID(p->one)) {
1938 diff_populate_filespec(p->one, CHECK_SIZE_ONLY);
1939 copied = added = 0;
1940 diff_free_filespec_data(p->one);
1941 } else if (DIFF_FILE_VALID(p->two)) {
1942 diff_populate_filespec(p->two, CHECK_SIZE_ONLY);
1943 copied = 0;
1944 added = p->two->size;
1945 diff_free_filespec_data(p->two);
1946 } else
1947 continue;
1948
1949 /*
1950 * Original minus copied is the removed material,
1951 * added is the new material. They are both damages
1952 * made to the preimage.
1953 * If the resulting damage is zero, we know that
1954 * diffcore_count_changes() considers the two entries to
1955 * be identical, but since content_changed is true, we
1956 * know that there must have been _some_ kind of change,
1957 * so we force all entries to have damage > 0.
1958 */
1959 damage = (p->one->size - copied) + added;
1960 if (!damage)
1961 damage = 1;
1962
1963 found_damage:
1964 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1965 dir.files[dir.nr].name = name;
1966 dir.files[dir.nr].changed = damage;
1967 changed += damage;
1968 dir.nr++;
1969 }
1970
1971 /* This can happen even with many files, if everything was renames */
1972 if (!changed)
1973 return;
1974
1975 /* Show all directories with more than x% of the changes */
1976 qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1977 gather_dirstat(options, &dir, changed, "", 0);
1978 }
1979
1980 static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
1981 {
1982 int i;
1983 unsigned long changed;
1984 struct dirstat_dir dir;
1985
1986 if (data->nr == 0)
1987 return;
1988
1989 dir.files = NULL;
1990 dir.alloc = 0;
1991 dir.nr = 0;
1992 dir.permille = options->dirstat_permille;
1993 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1994
1995 changed = 0;
1996 for (i = 0; i < data->nr; i++) {
1997 struct diffstat_file *file = data->files[i];
1998 unsigned long damage = file->added + file->deleted;
1999 if (file->is_binary)
2000 /*
2001 * binary files counts bytes, not lines. Must find some
2002 * way to normalize binary bytes vs. textual lines.
2003 * The following heuristic assumes that there are 64
2004 * bytes per "line".
2005 * This is stupid and ugly, but very cheap...
2006 */
2007 damage = (damage + 63) / 64;
2008 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2009 dir.files[dir.nr].name = file->name;
2010 dir.files[dir.nr].changed = damage;
2011 changed += damage;
2012 dir.nr++;
2013 }
2014
2015 /* This can happen even with many files, if everything was renames */
2016 if (!changed)
2017 return;
2018
2019 /* Show all directories with more than x% of the changes */
2020 qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
2021 gather_dirstat(options, &dir, changed, "", 0);
2022 }
2023
2024 static void free_diffstat_info(struct diffstat_t *diffstat)
2025 {
2026 int i;
2027 for (i = 0; i < diffstat->nr; i++) {
2028 struct diffstat_file *f = diffstat->files[i];
2029 if (f->name != f->print_name)
2030 free(f->print_name);
2031 free(f->name);
2032 free(f->from_name);
2033 free(f);
2034 }
2035 free(diffstat->files);
2036 }
2037
2038 struct checkdiff_t {
2039 const char *filename;
2040 int lineno;
2041 int conflict_marker_size;
2042 struct diff_options *o;
2043 unsigned ws_rule;
2044 unsigned status;
2045 };
2046
2047 static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2048 {
2049 char firstchar;
2050 int cnt;
2051
2052 if (len < marker_size + 1)
2053 return 0;
2054 firstchar = line[0];
2055 switch (firstchar) {
2056 case '=': case '>': case '<': case '|':
2057 break;
2058 default:
2059 return 0;
2060 }
2061 for (cnt = 1; cnt < marker_size; cnt++)
2062 if (line[cnt] != firstchar)
2063 return 0;
2064 /* line[1] thru line[marker_size-1] are same as firstchar */
2065 if (len < marker_size + 1 || !isspace(line[marker_size]))
2066 return 0;
2067 return 1;
2068 }
2069
2070 static void checkdiff_consume(void *priv, char *line, unsigned long len)
2071 {
2072 struct checkdiff_t *data = priv;
2073 int marker_size = data->conflict_marker_size;
2074 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2075 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2076 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2077 char *err;
2078 const char *line_prefix;
2079
2080 assert(data->o);
2081 line_prefix = diff_line_prefix(data->o);
2082
2083 if (line[0] == '+') {
2084 unsigned bad;
2085 data->lineno++;
2086 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2087 data->status |= 1;
2088 fprintf(data->o->file,
2089 "%s%s:%d: leftover conflict marker\n",
2090 line_prefix, data->filename, data->lineno);
2091 }
2092 bad = ws_check(line + 1, len - 1, data->ws_rule);
2093 if (!bad)
2094 return;
2095 data->status |= bad;
2096 err = whitespace_error_string(bad);
2097 fprintf(data->o->file, "%s%s:%d: %s.\n",
2098 line_prefix, data->filename, data->lineno, err);
2099 free(err);
2100 emit_line(data->o, set, reset, line, 1);
2101 ws_check_emit(line + 1, len - 1, data->ws_rule,
2102 data->o->file, set, reset, ws);
2103 } else if (line[0] == ' ') {
2104 data->lineno++;
2105 } else if (line[0] == '@') {
2106 char *plus = strchr(line, '+');
2107 if (plus)
2108 data->lineno = strtol(plus, NULL, 10) - 1;
2109 else
2110 die("invalid diff");
2111 }
2112 }
2113
2114 static unsigned char *deflate_it(char *data,
2115 unsigned long size,
2116 unsigned long *result_size)
2117 {
2118 int bound;
2119 unsigned char *deflated;
2120 git_zstream stream;
2121
2122 memset(&stream, 0, sizeof(stream));
2123 git_deflate_init(&stream, zlib_compression_level);
2124 bound = git_deflate_bound(&stream, size);
2125 deflated = xmalloc(bound);
2126 stream.next_out = deflated;
2127 stream.avail_out = bound;
2128
2129 stream.next_in = (unsigned char *)data;
2130 stream.avail_in = size;
2131 while (git_deflate(&stream, Z_FINISH) == Z_OK)
2132 ; /* nothing */
2133 git_deflate_end(&stream);
2134 *result_size = stream.total_out;
2135 return deflated;
2136 }
2137
2138 static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two,
2139 const char *prefix)
2140 {
2141 void *cp;
2142 void *delta;
2143 void *deflated;
2144 void *data;
2145 unsigned long orig_size;
2146 unsigned long delta_size;
2147 unsigned long deflate_size;
2148 unsigned long data_size;
2149
2150 /* We could do deflated delta, or we could do just deflated two,
2151 * whichever is smaller.
2152 */
2153 delta = NULL;
2154 deflated = deflate_it(two->ptr, two->size, &deflate_size);
2155 if (one->size && two->size) {
2156 delta = diff_delta(one->ptr, one->size,
2157 two->ptr, two->size,
2158 &delta_size, deflate_size);
2159 if (delta) {
2160 void *to_free = delta;
2161 orig_size = delta_size;
2162 delta = deflate_it(delta, delta_size, &delta_size);
2163 free(to_free);
2164 }
2165 }
2166
2167 if (delta && delta_size < deflate_size) {
2168 fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2169 free(deflated);
2170 data = delta;
2171 data_size = delta_size;
2172 }
2173 else {
2174 fprintf(file, "%sliteral %lu\n", prefix, two->size);
2175 free(delta);
2176 data = deflated;
2177 data_size = deflate_size;
2178 }
2179
2180 /* emit data encoded in base85 */
2181 cp = data;
2182 while (data_size) {
2183 int bytes = (52 < data_size) ? 52 : data_size;
2184 char line[70];
2185 data_size -= bytes;
2186 if (bytes <= 26)
2187 line[0] = bytes + 'A' - 1;
2188 else
2189 line[0] = bytes - 26 + 'a' - 1;
2190 encode_85(line + 1, cp, bytes);
2191 cp = (char *) cp + bytes;
2192 fprintf(file, "%s", prefix);
2193 fputs(line, file);
2194 fputc('\n', file);
2195 }
2196 fprintf(file, "%s\n", prefix);
2197 free(data);
2198 }
2199
2200 static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two,
2201 const char *prefix)
2202 {
2203 fprintf(file, "%sGIT binary patch\n", prefix);
2204 emit_binary_diff_body(file, one, two, prefix);
2205 emit_binary_diff_body(file, two, one, prefix);
2206 }
2207
2208 int diff_filespec_is_binary(struct diff_filespec *one)
2209 {
2210 if (one->is_binary == -1) {
2211 diff_filespec_load_driver(one);
2212 if (one->driver->binary != -1)
2213 one->is_binary = one->driver->binary;
2214 else {
2215 if (!one->data && DIFF_FILE_VALID(one))
2216 diff_populate_filespec(one, CHECK_BINARY);
2217 if (one->is_binary == -1 && one->data)
2218 one->is_binary = buffer_is_binary(one->data,
2219 one->size);
2220 if (one->is_binary == -1)
2221 one->is_binary = 0;
2222 }
2223 }
2224 return one->is_binary;
2225 }
2226
2227 static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2228 {
2229 diff_filespec_load_driver(one);
2230 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2231 }
2232
2233 void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2234 {
2235 if (!options->a_prefix)
2236 options->a_prefix = a;
2237 if (!options->b_prefix)
2238 options->b_prefix = b;
2239 }
2240
2241 struct userdiff_driver *get_textconv(struct diff_filespec *one)
2242 {
2243 if (!DIFF_FILE_VALID(one))
2244 return NULL;
2245
2246 diff_filespec_load_driver(one);
2247 return userdiff_get_textconv(one->driver);
2248 }
2249
2250 static void builtin_diff(const char *name_a,
2251 const char *name_b,
2252 struct diff_filespec *one,
2253 struct diff_filespec *two,
2254 const char *xfrm_msg,
2255 int must_show_header,
2256 struct diff_options *o,
2257 int complete_rewrite)
2258 {
2259 mmfile_t mf1, mf2;
2260 const char *lbl[2];
2261 char *a_one, *b_two;
2262 const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
2263 const char *reset = diff_get_color_opt(o, DIFF_RESET);
2264 const char *a_prefix, *b_prefix;
2265 struct userdiff_driver *textconv_one = NULL;
2266 struct userdiff_driver *textconv_two = NULL;
2267 struct strbuf header = STRBUF_INIT;
2268 const char *line_prefix = diff_line_prefix(o);
2269
2270 if (DIFF_OPT_TST(o, SUBMODULE_LOG) &&
2271 (!one->mode || S_ISGITLINK(one->mode)) &&
2272 (!two->mode || S_ISGITLINK(two->mode))) {
2273 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2274 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2275 show_submodule_summary(o->file, one->path ? one->path : two->path,
2276 line_prefix,
2277 one->sha1, two->sha1, two->dirty_submodule,
2278 meta, del, add, reset);
2279 return;
2280 }
2281
2282 if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2283 textconv_one = get_textconv(one);
2284 textconv_two = get_textconv(two);
2285 }
2286
2287 diff_set_mnemonic_prefix(o, "a/", "b/");
2288 if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2289 a_prefix = o->b_prefix;
2290 b_prefix = o->a_prefix;
2291 } else {
2292 a_prefix = o->a_prefix;
2293 b_prefix = o->b_prefix;
2294 }
2295
2296 /* Never use a non-valid filename anywhere if at all possible */
2297 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2298 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2299
2300 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2301 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2302 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2303 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2304 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
2305 if (lbl[0][0] == '/') {
2306 /* /dev/null */
2307 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
2308 if (xfrm_msg)
2309 strbuf_addstr(&header, xfrm_msg);
2310 must_show_header = 1;
2311 }
2312 else if (lbl[1][0] == '/') {
2313 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
2314 if (xfrm_msg)
2315 strbuf_addstr(&header, xfrm_msg);
2316 must_show_header = 1;
2317 }
2318 else {
2319 if (one->mode != two->mode) {
2320 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
2321 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
2322 must_show_header = 1;
2323 }
2324 if (xfrm_msg)
2325 strbuf_addstr(&header, xfrm_msg);
2326
2327 /*
2328 * we do not run diff between different kind
2329 * of objects.
2330 */
2331 if ((one->mode ^ two->mode) & S_IFMT)
2332 goto free_ab_and_return;
2333 if (complete_rewrite &&
2334 (textconv_one || !diff_filespec_is_binary(one)) &&
2335 (textconv_two || !diff_filespec_is_binary(two))) {
2336 fprintf(o->file, "%s", header.buf);
2337 strbuf_reset(&header);
2338 emit_rewrite_diff(name_a, name_b, one, two,
2339 textconv_one, textconv_two, o);
2340 o->found_changes = 1;
2341 goto free_ab_and_return;
2342 }
2343 }
2344
2345 if (o->irreversible_delete && lbl[1][0] == '/') {
2346 fprintf(o->file, "%s", header.buf);
2347 strbuf_reset(&header);
2348 goto free_ab_and_return;
2349 } else if (!DIFF_OPT_TST(o, TEXT) &&
2350 ( (!textconv_one && diff_filespec_is_binary(one)) ||
2351 (!textconv_two && diff_filespec_is_binary(two)) )) {
2352 if (!one->data && !two->data &&
2353 S_ISREG(one->mode) && S_ISREG(two->mode) &&
2354 !DIFF_OPT_TST(o, BINARY)) {
2355 if (!hashcmp(one->sha1, two->sha1)) {
2356 if (must_show_header)
2357 fprintf(o->file, "%s", header.buf);
2358 goto free_ab_and_return;
2359 }
2360 fprintf(o->file, "%s", header.buf);
2361 fprintf(o->file, "%sBinary files %s and %s differ\n",
2362 line_prefix, lbl[0], lbl[1]);
2363 goto free_ab_and_return;
2364 }
2365 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2366 die("unable to read files to diff");
2367 /* Quite common confusing case */
2368 if (mf1.size == mf2.size &&
2369 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2370 if (must_show_header)
2371 fprintf(o->file, "%s", header.buf);
2372 goto free_ab_and_return;
2373 }
2374 fprintf(o->file, "%s", header.buf);
2375 strbuf_reset(&header);
2376 if (DIFF_OPT_TST(o, BINARY))
2377 emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2378 else
2379 fprintf(o->file, "%sBinary files %s and %s differ\n",
2380 line_prefix, lbl[0], lbl[1]);
2381 o->found_changes = 1;
2382 } else {
2383 /* Crazy xdl interfaces.. */
2384 const char *diffopts = getenv("GIT_DIFF_OPTS");
2385 const char *v;
2386 xpparam_t xpp;
2387 xdemitconf_t xecfg;
2388 struct emit_callback ecbdata;
2389 const struct userdiff_funcname *pe;
2390
2391 if (must_show_header) {
2392 fprintf(o->file, "%s", header.buf);
2393 strbuf_reset(&header);
2394 }
2395
2396 mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2397 mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2398
2399 pe = diff_funcname_pattern(one);
2400 if (!pe)
2401 pe = diff_funcname_pattern(two);
2402
2403 memset(&xpp, 0, sizeof(xpp));
2404 memset(&xecfg, 0, sizeof(xecfg));
2405 memset(&ecbdata, 0, sizeof(ecbdata));
2406 ecbdata.label_path = lbl;
2407 ecbdata.color_diff = want_color(o->use_color);
2408 ecbdata.found_changesp = &o->found_changes;
2409 ecbdata.ws_rule = whitespace_rule(name_b);
2410 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2411 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2412 ecbdata.opt = o;
2413 ecbdata.header = header.len ? &header : NULL;
2414 xpp.flags = o->xdl_opts;
2415 xecfg.ctxlen = o->context;
2416 xecfg.interhunkctxlen = o->interhunkcontext;
2417 xecfg.flags = XDL_EMIT_FUNCNAMES;
2418 if (DIFF_OPT_TST(o, FUNCCONTEXT))
2419 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2420 if (pe)
2421 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2422 if (!diffopts)
2423 ;
2424 else if (skip_prefix(diffopts, "--unified=", &v))
2425 xecfg.ctxlen = strtoul(v, NULL, 10);
2426 else if (skip_prefix(diffopts, "-u", &v))
2427 xecfg.ctxlen = strtoul(v, NULL, 10);
2428 if (o->word_diff)
2429 init_diff_words_data(&ecbdata, o, one, two);
2430 xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2431 &xpp, &xecfg);
2432 if (o->word_diff)
2433 free_diff_words_data(&ecbdata);
2434 if (textconv_one)
2435 free(mf1.ptr);
2436 if (textconv_two)
2437 free(mf2.ptr);
2438 xdiff_clear_find_func(&xecfg);
2439 }
2440
2441 free_ab_and_return:
2442 strbuf_release(&header);
2443 diff_free_filespec_data(one);
2444 diff_free_filespec_data(two);
2445 free(a_one);
2446 free(b_two);
2447 return;
2448 }
2449
2450 static void builtin_diffstat(const char *name_a, const char *name_b,
2451 struct diff_filespec *one,
2452 struct diff_filespec *two,
2453 struct diffstat_t *diffstat,
2454 struct diff_options *o,
2455 struct diff_filepair *p)
2456 {
2457 mmfile_t mf1, mf2;
2458 struct diffstat_file *data;
2459 int same_contents;
2460 int complete_rewrite = 0;
2461
2462 if (!DIFF_PAIR_UNMERGED(p)) {
2463 if (p->status == DIFF_STATUS_MODIFIED && p->score)
2464 complete_rewrite = 1;
2465 }
2466
2467 data = diffstat_add(diffstat, name_a, name_b);
2468 data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
2469
2470 if (!one || !two) {
2471 data->is_unmerged = 1;
2472 return;
2473 }
2474
2475 same_contents = !hashcmp(one->sha1, two->sha1);
2476
2477 if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2478 data->is_binary = 1;
2479 if (same_contents) {
2480 data->added = 0;
2481 data->deleted = 0;
2482 } else {
2483 data->added = diff_filespec_size(two);
2484 data->deleted = diff_filespec_size(one);
2485 }
2486 }
2487
2488 else if (complete_rewrite) {
2489 diff_populate_filespec(one, 0);
2490 diff_populate_filespec(two, 0);
2491 data->deleted = count_lines(one->data, one->size);
2492 data->added = count_lines(two->data, two->size);
2493 }
2494
2495 else if (!same_contents) {
2496 /* Crazy xdl interfaces.. */
2497 xpparam_t xpp;
2498 xdemitconf_t xecfg;
2499
2500 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2501 die("unable to read files to diff");
2502
2503 memset(&xpp, 0, sizeof(xpp));
2504 memset(&xecfg, 0, sizeof(xecfg));
2505 xpp.flags = o->xdl_opts;
2506 xecfg.ctxlen = o->context;
2507 xecfg.interhunkctxlen = o->interhunkcontext;
2508 xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2509 &xpp, &xecfg);
2510 }
2511
2512 diff_free_filespec_data(one);
2513 diff_free_filespec_data(two);
2514 }
2515
2516 static void builtin_checkdiff(const char *name_a, const char *name_b,
2517 const char *attr_path,
2518 struct diff_filespec *one,
2519 struct diff_filespec *two,
2520 struct diff_options *o)
2521 {
2522 mmfile_t mf1, mf2;
2523 struct checkdiff_t data;
2524
2525 if (!two)
2526 return;
2527
2528 memset(&data, 0, sizeof(data));
2529 data.filename = name_b ? name_b : name_a;
2530 data.lineno = 0;
2531 data.o = o;
2532 data.ws_rule = whitespace_rule(attr_path);
2533 data.conflict_marker_size = ll_merge_marker_size(attr_path);
2534
2535 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2536 die("unable to read files to diff");
2537
2538 /*
2539 * All the other codepaths check both sides, but not checking
2540 * the "old" side here is deliberate. We are checking the newly
2541 * introduced changes, and as long as the "new" side is text, we
2542 * can and should check what it introduces.
2543 */
2544 if (diff_filespec_is_binary(two))
2545 goto free_and_return;
2546 else {
2547 /* Crazy xdl interfaces.. */
2548 xpparam_t xpp;
2549 xdemitconf_t xecfg;
2550
2551 memset(&xpp, 0, sizeof(xpp));
2552 memset(&xecfg, 0, sizeof(xecfg));
2553 xecfg.ctxlen = 1; /* at least one context line */
2554 xpp.flags = 0;
2555 xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2556 &xpp, &xecfg);
2557
2558 if (data.ws_rule & WS_BLANK_AT_EOF) {
2559 struct emit_callback ecbdata;
2560 int blank_at_eof;
2561
2562 ecbdata.ws_rule = data.ws_rule;
2563 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2564 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2565
2566 if (blank_at_eof) {
2567 static char *err;
2568 if (!err)
2569 err = whitespace_error_string(WS_BLANK_AT_EOF);
2570 fprintf(o->file, "%s:%d: %s.\n",
2571 data.filename, blank_at_eof, err);
2572 data.status = 1; /* report errors */
2573 }
2574 }
2575 }
2576 free_and_return:
2577 diff_free_filespec_data(one);
2578 diff_free_filespec_data(two);
2579 if (data.status)
2580 DIFF_OPT_SET(o, CHECK_FAILED);
2581 }
2582
2583 struct diff_filespec *alloc_filespec(const char *path)
2584 {
2585 int namelen = strlen(path);
2586 struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
2587
2588 memset(spec, 0, sizeof(*spec));
2589 spec->path = (char *)(spec + 1);
2590 memcpy(spec->path, path, namelen+1);
2591 spec->count = 1;
2592 spec->is_binary = -1;
2593 return spec;
2594 }
2595
2596 void free_filespec(struct diff_filespec *spec)
2597 {
2598 if (!--spec->count) {
2599 diff_free_filespec_data(spec);
2600 free(spec);
2601 }
2602 }
2603
2604 void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
2605 int sha1_valid, unsigned short mode)
2606 {
2607 if (mode) {
2608 spec->mode = canon_mode(mode);
2609 hashcpy(spec->sha1, sha1);
2610 spec->sha1_valid = sha1_valid;
2611 }
2612 }
2613
2614 /*
2615 * Given a name and sha1 pair, if the index tells us the file in
2616 * the work tree has that object contents, return true, so that
2617 * prepare_temp_file() does not have to inflate and extract.
2618 */
2619 static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
2620 {
2621 const struct cache_entry *ce;
2622 struct stat st;
2623 int pos, len;
2624
2625 /*
2626 * We do not read the cache ourselves here, because the
2627 * benchmark with my previous version that always reads cache
2628 * shows that it makes things worse for diff-tree comparing
2629 * two linux-2.6 kernel trees in an already checked out work
2630 * tree. This is because most diff-tree comparisons deal with
2631 * only a small number of files, while reading the cache is
2632 * expensive for a large project, and its cost outweighs the
2633 * savings we get by not inflating the object to a temporary
2634 * file. Practically, this code only helps when we are used
2635 * by diff-cache --cached, which does read the cache before
2636 * calling us.
2637 */
2638 if (!active_cache)
2639 return 0;
2640
2641 /* We want to avoid the working directory if our caller
2642 * doesn't need the data in a normal file, this system
2643 * is rather slow with its stat/open/mmap/close syscalls,
2644 * and the object is contained in a pack file. The pack
2645 * is probably already open and will be faster to obtain
2646 * the data through than the working directory. Loose
2647 * objects however would tend to be slower as they need
2648 * to be individually opened and inflated.
2649 */
2650 if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1))
2651 return 0;
2652
2653 len = strlen(name);
2654 pos = cache_name_pos(name, len);
2655 if (pos < 0)
2656 return 0;
2657 ce = active_cache[pos];
2658
2659 /*
2660 * This is not the sha1 we are looking for, or
2661 * unreusable because it is not a regular file.
2662 */
2663 if (hashcmp(sha1, ce->sha1) || !S_ISREG(ce->ce_mode))
2664 return 0;
2665
2666 /*
2667 * If ce is marked as "assume unchanged", there is no
2668 * guarantee that work tree matches what we are looking for.
2669 */
2670 if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2671 return 0;
2672
2673 /*
2674 * If ce matches the file in the work tree, we can reuse it.
2675 */
2676 if (ce_uptodate(ce) ||
2677 (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2678 return 1;
2679
2680 return 0;
2681 }
2682
2683 static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2684 {
2685 int len;
2686 char *data = xmalloc(100), *dirty = "";
2687
2688 /* Are we looking at the work tree? */
2689 if (s->dirty_submodule)
2690 dirty = "-dirty";
2691
2692 len = snprintf(data, 100,
2693 "Subproject commit %s%s\n", sha1_to_hex(s->sha1), dirty);
2694 s->data = data;
2695 s->size = len;
2696 s->should_free = 1;
2697 if (size_only) {
2698 s->data = NULL;
2699 free(data);
2700 }
2701 return 0;
2702 }
2703
2704 /*
2705 * While doing rename detection and pickaxe operation, we may need to
2706 * grab the data for the blob (or file) for our own in-core comparison.
2707 * diff_filespec has data and size fields for this purpose.
2708 */
2709 int diff_populate_filespec(struct diff_filespec *s, unsigned int flags)
2710 {
2711 int size_only = flags & CHECK_SIZE_ONLY;
2712 int err = 0;
2713 /*
2714 * demote FAIL to WARN to allow inspecting the situation
2715 * instead of refusing.
2716 */
2717 enum safe_crlf crlf_warn = (safe_crlf == SAFE_CRLF_FAIL
2718 ? SAFE_CRLF_WARN
2719 : safe_crlf);
2720
2721 if (!DIFF_FILE_VALID(s))
2722 die("internal error: asking to populate invalid file.");
2723 if (S_ISDIR(s->mode))
2724 return -1;
2725
2726 if (s->data)
2727 return 0;
2728
2729 if (size_only && 0 < s->size)
2730 return 0;
2731
2732 if (S_ISGITLINK(s->mode))
2733 return diff_populate_gitlink(s, size_only);
2734
2735 if (!s->sha1_valid ||
2736 reuse_worktree_file(s->path, s->sha1, 0)) {
2737 struct strbuf buf = STRBUF_INIT;
2738 struct stat st;
2739 int fd;
2740
2741 if (lstat(s->path, &st) < 0) {
2742 if (errno == ENOENT) {
2743 err_empty:
2744 err = -1;
2745 empty:
2746 s->data = (char *)"";
2747 s->size = 0;
2748 return err;
2749 }
2750 }
2751 s->size = xsize_t(st.st_size);
2752 if (!s->size)
2753 goto empty;
2754 if (S_ISLNK(st.st_mode)) {
2755 struct strbuf sb = STRBUF_INIT;
2756
2757 if (strbuf_readlink(&sb, s->path, s->size))
2758 goto err_empty;
2759 s->size = sb.len;
2760 s->data = strbuf_detach(&sb, NULL);
2761 s->should_free = 1;
2762 return 0;
2763 }
2764 if (size_only)
2765 return 0;
2766 if ((flags & CHECK_BINARY) &&
2767 s->size > big_file_threshold && s->is_binary == -1) {
2768 s->is_binary = 1;
2769 return 0;
2770 }
2771 fd = open(s->path, O_RDONLY);
2772 if (fd < 0)
2773 goto err_empty;
2774 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2775 close(fd);
2776 s->should_munmap = 1;
2777
2778 /*
2779 * Convert from working tree format to canonical git format
2780 */
2781 if (convert_to_git(s->path, s->data, s->size, &buf, crlf_warn)) {
2782 size_t size = 0;
2783 munmap(s->data, s->size);
2784 s->should_munmap = 0;
2785 s->data = strbuf_detach(&buf, &size);
2786 s->size = size;
2787 s->should_free = 1;
2788 }
2789 }
2790 else {
2791 enum object_type type;
2792 if (size_only || (flags & CHECK_BINARY)) {
2793 type = sha1_object_info(s->sha1, &s->size);
2794 if (type < 0)
2795 die("unable to read %s", sha1_to_hex(s->sha1));
2796 if (size_only)
2797 return 0;
2798 if (s->size > big_file_threshold && s->is_binary == -1) {
2799 s->is_binary = 1;
2800 return 0;
2801 }
2802 }
2803 s->data = read_sha1_file(s->sha1, &type, &s->size);
2804 if (!s->data)
2805 die("unable to read %s", sha1_to_hex(s->sha1));
2806 s->should_free = 1;
2807 }
2808 return 0;
2809 }
2810
2811 void diff_free_filespec_blob(struct diff_filespec *s)
2812 {
2813 if (s->should_free)
2814 free(s->data);
2815 else if (s->should_munmap)
2816 munmap(s->data, s->size);
2817
2818 if (s->should_free || s->should_munmap) {
2819 s->should_free = s->should_munmap = 0;
2820 s->data = NULL;
2821 }
2822 }
2823
2824 void diff_free_filespec_data(struct diff_filespec *s)
2825 {
2826 diff_free_filespec_blob(s);
2827 free(s->cnt_data);
2828 s->cnt_data = NULL;
2829 }
2830
2831 static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2832 void *blob,
2833 unsigned long size,
2834 const unsigned char *sha1,
2835 int mode)
2836 {
2837 int fd;
2838 struct strbuf buf = STRBUF_INIT;
2839 struct strbuf template = STRBUF_INIT;
2840 char *path_dup = xstrdup(path);
2841 const char *base = basename(path_dup);
2842
2843 /* Generate "XXXXXX_basename.ext" */
2844 strbuf_addstr(&template, "XXXXXX_");
2845 strbuf_addstr(&template, base);
2846
2847 fd = git_mkstemps(temp->tmp_path, PATH_MAX, template.buf,
2848 strlen(base) + 1);
2849 if (fd < 0)
2850 die_errno("unable to create temp-file");
2851 if (convert_to_working_tree(path,
2852 (const char *)blob, (size_t)size, &buf)) {
2853 blob = buf.buf;
2854 size = buf.len;
2855 }
2856 if (write_in_full(fd, blob, size) != size)
2857 die_errno("unable to write temp-file");
2858 close(fd);
2859 temp->name = temp->tmp_path;
2860 strcpy(temp->hex, sha1_to_hex(sha1));
2861 temp->hex[40] = 0;
2862 sprintf(temp->mode, "%06o", mode);
2863 strbuf_release(&buf);
2864 strbuf_release(&template);
2865 free(path_dup);
2866 }
2867
2868 static struct diff_tempfile *prepare_temp_file(const char *name,
2869 struct diff_filespec *one)
2870 {
2871 struct diff_tempfile *temp = claim_diff_tempfile();
2872
2873 if (!DIFF_FILE_VALID(one)) {
2874 not_a_valid_file:
2875 /* A '-' entry produces this for file-2, and
2876 * a '+' entry produces this for file-1.
2877 */
2878 temp->name = "/dev/null";
2879 strcpy(temp->hex, ".");
2880 strcpy(temp->mode, ".");
2881 return temp;
2882 }
2883
2884 if (!remove_tempfile_installed) {
2885 atexit(remove_tempfile);
2886 sigchain_push_common(remove_tempfile_on_signal);
2887 remove_tempfile_installed = 1;
2888 }
2889
2890 if (!S_ISGITLINK(one->mode) &&
2891 (!one->sha1_valid ||
2892 reuse_worktree_file(name, one->sha1, 1))) {
2893 struct stat st;
2894 if (lstat(name, &st) < 0) {
2895 if (errno == ENOENT)
2896 goto not_a_valid_file;
2897 die_errno("stat(%s)", name);
2898 }
2899 if (S_ISLNK(st.st_mode)) {
2900 struct strbuf sb = STRBUF_INIT;
2901 if (strbuf_readlink(&sb, name, st.st_size) < 0)
2902 die_errno("readlink(%s)", name);
2903 prep_temp_blob(name, temp, sb.buf, sb.len,
2904 (one->sha1_valid ?
2905 one->sha1 : null_sha1),
2906 (one->sha1_valid ?
2907 one->mode : S_IFLNK));
2908 strbuf_release(&sb);
2909 }
2910 else {
2911 /* we can borrow from the file in the work tree */
2912 temp->name = name;
2913 if (!one->sha1_valid)
2914 strcpy(temp->hex, sha1_to_hex(null_sha1));
2915 else
2916 strcpy(temp->hex, sha1_to_hex(one->sha1));
2917 /* Even though we may sometimes borrow the
2918 * contents from the work tree, we always want
2919 * one->mode. mode is trustworthy even when
2920 * !(one->sha1_valid), as long as
2921 * DIFF_FILE_VALID(one).
2922 */
2923 sprintf(temp->mode, "%06o", one->mode);
2924 }
2925 return temp;
2926 }
2927 else {
2928 if (diff_populate_filespec(one, 0))
2929 die("cannot read data blob for %s", one->path);
2930 prep_temp_blob(name, temp, one->data, one->size,
2931 one->sha1, one->mode);
2932 }
2933 return temp;
2934 }
2935
2936 static void add_external_diff_name(struct argv_array *argv,
2937 const char *name,
2938 struct diff_filespec *df)
2939 {
2940 struct diff_tempfile *temp = prepare_temp_file(name, df);
2941 argv_array_push(argv, temp->name);
2942 argv_array_push(argv, temp->hex);
2943 argv_array_push(argv, temp->mode);
2944 }
2945
2946 /* An external diff command takes:
2947 *
2948 * diff-cmd name infile1 infile1-sha1 infile1-mode \
2949 * infile2 infile2-sha1 infile2-mode [ rename-to ]
2950 *
2951 */
2952 static void run_external_diff(const char *pgm,
2953 const char *name,
2954 const char *other,
2955 struct diff_filespec *one,
2956 struct diff_filespec *two,
2957 const char *xfrm_msg,
2958 int complete_rewrite,
2959 struct diff_options *o)
2960 {
2961 struct argv_array argv = ARGV_ARRAY_INIT;
2962 struct argv_array env = ARGV_ARRAY_INIT;
2963 struct diff_queue_struct *q = &diff_queued_diff;
2964
2965 argv_array_push(&argv, pgm);
2966 argv_array_push(&argv, name);
2967
2968 if (one && two) {
2969 add_external_diff_name(&argv, name, one);
2970 if (!other)
2971 add_external_diff_name(&argv, name, two);
2972 else {
2973 add_external_diff_name(&argv, other, two);
2974 argv_array_push(&argv, other);
2975 argv_array_push(&argv, xfrm_msg);
2976 }
2977 }
2978
2979 argv_array_pushf(&env, "GIT_DIFF_PATH_COUNTER=%d", ++o->diff_path_counter);
2980 argv_array_pushf(&env, "GIT_DIFF_PATH_TOTAL=%d", q->nr);
2981
2982 if (run_command_v_opt_cd_env(argv.argv, RUN_USING_SHELL, NULL, env.argv))
2983 die(_("external diff died, stopping at %s"), name);
2984
2985 remove_tempfile();
2986 argv_array_clear(&argv);
2987 argv_array_clear(&env);
2988 }
2989
2990 static int similarity_index(struct diff_filepair *p)
2991 {
2992 return p->score * 100 / MAX_SCORE;
2993 }
2994
2995 static void fill_metainfo(struct strbuf *msg,
2996 const char *name,
2997 const char *other,
2998 struct diff_filespec *one,
2999 struct diff_filespec *two,
3000 struct diff_options *o,
3001 struct diff_filepair *p,
3002 int *must_show_header,
3003 int use_color)
3004 {
3005 const char *set = diff_get_color(use_color, DIFF_METAINFO);
3006 const char *reset = diff_get_color(use_color, DIFF_RESET);
3007 const char *line_prefix = diff_line_prefix(o);
3008
3009 *must_show_header = 1;
3010 strbuf_init(msg, PATH_MAX * 2 + 300);
3011 switch (p->status) {
3012 case DIFF_STATUS_COPIED:
3013 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3014 line_prefix, set, similarity_index(p));
3015 strbuf_addf(msg, "%s\n%s%scopy from ",
3016 reset, line_prefix, set);
3017 quote_c_style(name, msg, NULL, 0);
3018 strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
3019 quote_c_style(other, msg, NULL, 0);
3020 strbuf_addf(msg, "%s\n", reset);
3021 break;
3022 case DIFF_STATUS_RENAMED:
3023 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3024 line_prefix, set, similarity_index(p));
3025 strbuf_addf(msg, "%s\n%s%srename from ",
3026 reset, line_prefix, set);
3027 quote_c_style(name, msg, NULL, 0);
3028 strbuf_addf(msg, "%s\n%s%srename to ",
3029 reset, line_prefix, set);
3030 quote_c_style(other, msg, NULL, 0);
3031 strbuf_addf(msg, "%s\n", reset);
3032 break;
3033 case DIFF_STATUS_MODIFIED:
3034 if (p->score) {
3035 strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
3036 line_prefix,
3037 set, similarity_index(p), reset);
3038 break;
3039 }
3040 /* fallthru */
3041 default:
3042 *must_show_header = 0;
3043 }
3044 if (one && two && hashcmp(one->sha1, two->sha1)) {
3045 int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
3046
3047 if (DIFF_OPT_TST(o, BINARY)) {
3048 mmfile_t mf;
3049 if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
3050 (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
3051 abbrev = 40;
3052 }
3053 strbuf_addf(msg, "%s%sindex %s..", line_prefix, set,
3054 find_unique_abbrev(one->sha1, abbrev));
3055 strbuf_addstr(msg, find_unique_abbrev(two->sha1, abbrev));
3056 if (one->mode == two->mode)
3057 strbuf_addf(msg, " %06o", one->mode);
3058 strbuf_addf(msg, "%s\n", reset);
3059 }
3060 }
3061
3062 static void run_diff_cmd(const char *pgm,
3063 const char *name,
3064 const char *other,
3065 const char *attr_path,
3066 struct diff_filespec *one,
3067 struct diff_filespec *two,
3068 struct strbuf *msg,
3069 struct diff_options *o,
3070 struct diff_filepair *p)
3071 {
3072 const char *xfrm_msg = NULL;
3073 int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
3074 int must_show_header = 0;
3075
3076
3077 if (DIFF_OPT_TST(o, ALLOW_EXTERNAL)) {
3078 struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
3079 if (drv && drv->external)
3080 pgm = drv->external;
3081 }
3082
3083 if (msg) {
3084 /*
3085 * don't use colors when the header is intended for an
3086 * external diff driver
3087 */
3088 fill_metainfo(msg, name, other, one, two, o, p,
3089 &must_show_header,
3090 want_color(o->use_color) && !pgm);
3091 xfrm_msg = msg->len ? msg->buf : NULL;
3092 }
3093
3094 if (pgm) {
3095 run_external_diff(pgm, name, other, one, two, xfrm_msg,
3096 complete_rewrite, o);
3097 return;
3098 }
3099 if (one && two)
3100 builtin_diff(name, other ? other : name,
3101 one, two, xfrm_msg, must_show_header,
3102 o, complete_rewrite);
3103 else
3104 fprintf(o->file, "* Unmerged path %s\n", name);
3105 }
3106
3107 static void diff_fill_sha1_info(struct diff_filespec *one)
3108 {
3109 if (DIFF_FILE_VALID(one)) {
3110 if (!one->sha1_valid) {
3111 struct stat st;
3112 if (one->is_stdin) {
3113 hashcpy(one->sha1, null_sha1);
3114 return;
3115 }
3116 if (lstat(one->path, &st) < 0)
3117 die_errno("stat '%s'", one->path);
3118 if (index_path(one->sha1, one->path, &st, 0))
3119 die("cannot hash %s", one->path);
3120 }
3121 }
3122 else
3123 hashclr(one->sha1);
3124 }
3125
3126 static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
3127 {
3128 /* Strip the prefix but do not molest /dev/null and absolute paths */
3129 if (*namep && **namep != '/') {
3130 *namep += prefix_length;
3131 if (**namep == '/')
3132 ++*namep;
3133 }
3134 if (*otherp && **otherp != '/') {
3135 *otherp += prefix_length;
3136 if (**otherp == '/')
3137 ++*otherp;
3138 }
3139 }
3140
3141 static void run_diff(struct diff_filepair *p, struct diff_options *o)
3142 {
3143 const char *pgm = external_diff();
3144 struct strbuf msg;
3145 struct diff_filespec *one = p->one;
3146 struct diff_filespec *two = p->two;
3147 const char *name;
3148 const char *other;
3149 const char *attr_path;
3150
3151 name = p->one->path;
3152 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3153 attr_path = name;
3154 if (o->prefix_length)
3155 strip_prefix(o->prefix_length, &name, &other);
3156
3157 if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
3158 pgm = NULL;
3159
3160 if (DIFF_PAIR_UNMERGED(p)) {
3161 run_diff_cmd(pgm, name, NULL, attr_path,
3162 NULL, NULL, NULL, o, p);
3163 return;
3164 }
3165
3166 diff_fill_sha1_info(one);
3167 diff_fill_sha1_info(two);
3168
3169 if (!pgm &&
3170 DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3171 (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3172 /*
3173 * a filepair that changes between file and symlink
3174 * needs to be split into deletion and creation.
3175 */
3176 struct diff_filespec *null = alloc_filespec(two->path);
3177 run_diff_cmd(NULL, name, other, attr_path,
3178 one, null, &msg, o, p);
3179 free(null);
3180 strbuf_release(&msg);
3181
3182 null = alloc_filespec(one->path);
3183 run_diff_cmd(NULL, name, other, attr_path,
3184 null, two, &msg, o, p);
3185 free(null);
3186 }
3187 else
3188 run_diff_cmd(pgm, name, other, attr_path,
3189 one, two, &msg, o, p);
3190
3191 strbuf_release(&msg);
3192 }
3193
3194 static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3195 struct diffstat_t *diffstat)
3196 {
3197 const char *name;
3198 const char *other;
3199
3200 if (DIFF_PAIR_UNMERGED(p)) {
3201 /* unmerged */
3202 builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, p);
3203 return;
3204 }
3205
3206 name = p->one->path;
3207 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3208
3209 if (o->prefix_length)
3210 strip_prefix(o->prefix_length, &name, &other);
3211
3212 diff_fill_sha1_info(p->one);
3213 diff_fill_sha1_info(p->two);
3214
3215 builtin_diffstat(name, other, p->one, p->two, diffstat, o, p);
3216 }
3217
3218 static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3219 {
3220 const char *name;
3221 const char *other;
3222 const char *attr_path;
3223
3224 if (DIFF_PAIR_UNMERGED(p)) {
3225 /* unmerged */
3226 return;
3227 }
3228
3229 name = p->one->path;
3230 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3231 attr_path = other ? other : name;
3232
3233 if (o->prefix_length)
3234 strip_prefix(o->prefix_length, &name, &other);
3235
3236 diff_fill_sha1_info(p->one);
3237 diff_fill_sha1_info(p->two);
3238
3239 builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3240 }
3241
3242 void diff_setup(struct diff_options *options)
3243 {
3244 memcpy(options, &default_diff_options, sizeof(*options));
3245
3246 options->file = stdout;
3247
3248 options->line_termination = '\n';
3249 options->break_opt = -1;
3250 options->rename_limit = -1;
3251 options->dirstat_permille = diff_dirstat_permille_default;
3252 options->context = diff_context_default;
3253 DIFF_OPT_SET(options, RENAME_EMPTY);
3254
3255 /* pathchange left =NULL by default */
3256 options->change = diff_change;
3257 options->add_remove = diff_addremove;
3258 options->use_color = diff_use_color_default;
3259 options->detect_rename = diff_detect_rename_default;
3260 options->xdl_opts |= diff_algorithm;
3261
3262 options->orderfile = diff_order_file_cfg;
3263
3264 if (diff_no_prefix) {
3265 options->a_prefix = options->b_prefix = "";
3266 } else if (!diff_mnemonic_prefix) {
3267 options->a_prefix = "a/";
3268 options->b_prefix = "b/";
3269 }
3270 }
3271
3272 void diff_setup_done(struct diff_options *options)
3273 {
3274 int count = 0;
3275
3276 if (options->set_default)
3277 options->set_default(options);
3278
3279 if (options->output_format & DIFF_FORMAT_NAME)
3280 count++;
3281 if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3282 count++;
3283 if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3284 count++;
3285 if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3286 count++;
3287 if (count > 1)
3288 die("--name-only, --name-status, --check and -s are mutually exclusive");
3289
3290 /*
3291 * Most of the time we can say "there are changes"
3292 * only by checking if there are changed paths, but
3293 * --ignore-whitespace* options force us to look
3294 * inside contents.
3295 */
3296
3297 if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3298 DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3299 DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3300 DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3301 else
3302 DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3303
3304 if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3305 options->detect_rename = DIFF_DETECT_COPY;
3306
3307 if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3308 options->prefix = NULL;
3309 if (options->prefix)
3310 options->prefix_length = strlen(options->prefix);
3311 else
3312 options->prefix_length = 0;
3313
3314 if (options->output_format & (DIFF_FORMAT_NAME |
3315 DIFF_FORMAT_NAME_STATUS |
3316 DIFF_FORMAT_CHECKDIFF |
3317 DIFF_FORMAT_NO_OUTPUT))
3318 options->output_format &= ~(DIFF_FORMAT_RAW |
3319 DIFF_FORMAT_NUMSTAT |
3320 DIFF_FORMAT_DIFFSTAT |
3321 DIFF_FORMAT_SHORTSTAT |
3322 DIFF_FORMAT_DIRSTAT |
3323 DIFF_FORMAT_SUMMARY |
3324 DIFF_FORMAT_PATCH);
3325
3326 /*
3327 * These cases always need recursive; we do not drop caller-supplied
3328 * recursive bits for other formats here.
3329 */
3330 if (options->output_format & (DIFF_FORMAT_PATCH |
3331 DIFF_FORMAT_NUMSTAT |
3332 DIFF_FORMAT_DIFFSTAT |
3333 DIFF_FORMAT_SHORTSTAT |
3334 DIFF_FORMAT_DIRSTAT |
3335 DIFF_FORMAT_SUMMARY |
3336 DIFF_FORMAT_CHECKDIFF))
3337 DIFF_OPT_SET(options, RECURSIVE);
3338 /*
3339 * Also pickaxe would not work very well if you do not say recursive
3340 */
3341 if (options->pickaxe)
3342 DIFF_OPT_SET(options, RECURSIVE);
3343 /*
3344 * When patches are generated, submodules diffed against the work tree
3345 * must be checked for dirtiness too so it can be shown in the output
3346 */
3347 if (options->output_format & DIFF_FORMAT_PATCH)
3348 DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3349
3350 if (options->detect_rename && options->rename_limit < 0)
3351 options->rename_limit = diff_rename_limit_default;
3352 if (options->setup & DIFF_SETUP_USE_CACHE) {
3353 if (!active_cache)
3354 /* read-cache does not die even when it fails
3355 * so it is safe for us to do this here. Also
3356 * it does not smudge active_cache or active_nr
3357 * when it fails, so we do not have to worry about
3358 * cleaning it up ourselves either.
3359 */
3360 read_cache();
3361 }
3362 if (options->abbrev <= 0 || 40 < options->abbrev)
3363 options->abbrev = 40; /* full */
3364
3365 /*
3366 * It does not make sense to show the first hit we happened
3367 * to have found. It does not make sense not to return with
3368 * exit code in such a case either.
3369 */
3370 if (DIFF_OPT_TST(options, QUICK)) {
3371 options->output_format = DIFF_FORMAT_NO_OUTPUT;
3372 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3373 }
3374
3375 options->diff_path_counter = 0;
3376
3377 if (DIFF_OPT_TST(options, FOLLOW_RENAMES) && options->pathspec.nr != 1)
3378 die(_("--follow requires exactly one pathspec"));
3379 }
3380
3381 static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3382 {
3383 char c, *eq;
3384 int len;
3385
3386 if (*arg != '-')
3387 return 0;
3388 c = *++arg;
3389 if (!c)
3390 return 0;
3391 if (c == arg_short) {
3392 c = *++arg;
3393 if (!c)
3394 return 1;
3395 if (val && isdigit(c)) {
3396 char *end;
3397 int n = strtoul(arg, &end, 10);
3398 if (*end)
3399 return 0;
3400 *val = n;
3401 return 1;
3402 }
3403 return 0;
3404 }
3405 if (c != '-')
3406 return 0;
3407 arg++;
3408 eq = strchrnul(arg, '=');
3409 len = eq - arg;
3410 if (!len || strncmp(arg, arg_long, len))
3411 return 0;
3412 if (*eq) {
3413 int n;
3414 char *end;
3415 if (!isdigit(*++eq))
3416 return 0;
3417 n = strtoul(eq, &end, 10);
3418 if (*end)
3419 return 0;
3420 *val = n;
3421 }
3422 return 1;
3423 }
3424
3425 static int diff_scoreopt_parse(const char *opt);
3426
3427 static inline int short_opt(char opt, const char **argv,
3428 const char **optarg)
3429 {
3430 const char *arg = argv[0];
3431 if (arg[0] != '-' || arg[1] != opt)
3432 return 0;
3433 if (arg[2] != '\0') {
3434 *optarg = arg + 2;
3435 return 1;
3436 }
3437 if (!argv[1])
3438 die("Option '%c' requires a value", opt);
3439 *optarg = argv[1];
3440 return 2;
3441 }
3442
3443 int parse_long_opt(const char *opt, const char **argv,
3444 const char **optarg)
3445 {
3446 const char *arg = argv[0];
3447 if (!skip_prefix(arg, "--", &arg))
3448 return 0;
3449 if (!skip_prefix(arg, opt, &arg))
3450 return 0;
3451 if (*arg == '=') { /* stuck form: --option=value */
3452 *optarg = arg + 1;
3453 return 1;
3454 }
3455 if (*arg != '\0')
3456 return 0;
3457 /* separate form: --option value */
3458 if (!argv[1])
3459 die("Option '--%s' requires a value", opt);
3460 *optarg = argv[1];
3461 return 2;
3462 }
3463
3464 static int stat_opt(struct diff_options *options, const char **av)
3465 {
3466 const char *arg = av[0];
3467 char *end;
3468 int width = options->stat_width;
3469 int name_width = options->stat_name_width;
3470 int graph_width = options->stat_graph_width;
3471 int count = options->stat_count;
3472 int argcount = 1;
3473
3474 if (!skip_prefix(arg, "--stat", &arg))
3475 die("BUG: stat option does not begin with --stat: %s", arg);
3476 end = (char *)arg;
3477
3478 switch (*arg) {
3479 case '-':
3480 if (skip_prefix(arg, "-width", &arg)) {
3481 if (*arg == '=')
3482 width = strtoul(arg + 1, &end, 10);
3483 else if (!*arg && !av[1])
3484 die("Option '--stat-width' requires a value");
3485 else if (!*arg) {
3486 width = strtoul(av[1], &end, 10);
3487 argcount = 2;
3488 }
3489 } else if (skip_prefix(arg, "-name-width", &arg)) {
3490 if (*arg == '=')
3491 name_width = strtoul(arg + 1, &end, 10);
3492 else if (!*arg && !av[1])
3493 die("Option '--stat-name-width' requires a value");
3494 else if (!*arg) {
3495 name_width = strtoul(av[1], &end, 10);
3496 argcount = 2;
3497 }
3498 } else if (skip_prefix(arg, "-graph-width", &arg)) {
3499 if (*arg == '=')
3500 graph_width = strtoul(arg + 1, &end, 10);
3501 else if (!*arg && !av[1])
3502 die("Option '--stat-graph-width' requires a value");
3503 else if (!*arg) {
3504 graph_width = strtoul(av[1], &end, 10);
3505 argcount = 2;
3506 }
3507 } else if (skip_prefix(arg, "-count", &arg)) {
3508 if (*arg == '=')
3509 count = strtoul(arg + 1, &end, 10);
3510 else if (!*arg && !av[1])
3511 die("Option '--stat-count' requires a value");
3512 else if (!*arg) {
3513 count = strtoul(av[1], &end, 10);
3514 argcount = 2;
3515 }
3516 }
3517 break;
3518 case '=':
3519 width = strtoul(arg+1, &end, 10);
3520 if (*end == ',')
3521 name_width = strtoul(end+1, &end, 10);
3522 if (*end == ',')
3523 count = strtoul(end+1, &end, 10);
3524 }
3525
3526 /* Important! This checks all the error cases! */
3527 if (*end)
3528 return 0;
3529 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3530 options->stat_name_width = name_width;
3531 options->stat_graph_width = graph_width;
3532 options->stat_width = width;
3533 options->stat_count = count;
3534 return argcount;
3535 }
3536
3537 static int parse_dirstat_opt(struct diff_options *options, const char *params)
3538 {
3539 struct strbuf errmsg = STRBUF_INIT;
3540 if (parse_dirstat_params(options, params, &errmsg))
3541 die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3542 errmsg.buf);
3543 strbuf_release(&errmsg);
3544 /*
3545 * The caller knows a dirstat-related option is given from the command
3546 * line; allow it to say "return this_function();"
3547 */
3548 options->output_format |= DIFF_FORMAT_DIRSTAT;
3549 return 1;
3550 }
3551
3552 static int parse_submodule_opt(struct diff_options *options, const char *value)
3553 {
3554 if (parse_submodule_params(options, value))
3555 die(_("Failed to parse --submodule option parameter: '%s'"),
3556 value);
3557 return 1;
3558 }
3559
3560 static const char diff_status_letters[] = {
3561 DIFF_STATUS_ADDED,
3562 DIFF_STATUS_COPIED,
3563 DIFF_STATUS_DELETED,
3564 DIFF_STATUS_MODIFIED,
3565 DIFF_STATUS_RENAMED,
3566 DIFF_STATUS_TYPE_CHANGED,
3567 DIFF_STATUS_UNKNOWN,
3568 DIFF_STATUS_UNMERGED,
3569 DIFF_STATUS_FILTER_AON,
3570 DIFF_STATUS_FILTER_BROKEN,
3571 '\0',
3572 };
3573
3574 static unsigned int filter_bit['Z' + 1];
3575
3576 static void prepare_filter_bits(void)
3577 {
3578 int i;
3579
3580 if (!filter_bit[DIFF_STATUS_ADDED]) {
3581 for (i = 0; diff_status_letters[i]; i++)
3582 filter_bit[(int) diff_status_letters[i]] = (1 << i);
3583 }
3584 }
3585
3586 static unsigned filter_bit_tst(char status, const struct diff_options *opt)
3587 {
3588 return opt->filter & filter_bit[(int) status];
3589 }
3590
3591 static int parse_diff_filter_opt(const char *optarg, struct diff_options *opt)
3592 {
3593 int i, optch;
3594
3595 prepare_filter_bits();
3596
3597 /*
3598 * If there is a negation e.g. 'd' in the input, and we haven't
3599 * initialized the filter field with another --diff-filter, start
3600 * from full set of bits, except for AON.
3601 */
3602 if (!opt->filter) {
3603 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3604 if (optch < 'a' || 'z' < optch)
3605 continue;
3606 opt->filter = (1 << (ARRAY_SIZE(diff_status_letters) - 1)) - 1;
3607 opt->filter &= ~filter_bit[DIFF_STATUS_FILTER_AON];
3608 break;
3609 }
3610 }
3611
3612 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3613 unsigned int bit;
3614 int negate;
3615
3616 if ('a' <= optch && optch <= 'z') {
3617 negate = 1;
3618 optch = toupper(optch);
3619 } else {
3620 negate = 0;
3621 }
3622
3623 bit = (0 <= optch && optch <= 'Z') ? filter_bit[optch] : 0;
3624 if (!bit)
3625 return optarg[i];
3626 if (negate)
3627 opt->filter &= ~bit;
3628 else
3629 opt->filter |= bit;
3630 }
3631 return 0;
3632 }
3633
3634 static void enable_patch_output(int *fmt) {
3635 *fmt &= ~DIFF_FORMAT_NO_OUTPUT;
3636 *fmt |= DIFF_FORMAT_PATCH;
3637 }
3638
3639 int diff_opt_parse(struct diff_options *options, const char **av, int ac)
3640 {
3641 const char *arg = av[0];
3642 const char *optarg;
3643 int argcount;
3644
3645 /* Output format options */
3646 if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch")
3647 || opt_arg(arg, 'U', "unified", &options->context))
3648 enable_patch_output(&options->output_format);
3649 else if (!strcmp(arg, "--raw"))
3650 options->output_format |= DIFF_FORMAT_RAW;
3651 else if (!strcmp(arg, "--patch-with-raw")) {
3652 enable_patch_output(&options->output_format);
3653 options->output_format |= DIFF_FORMAT_RAW;
3654 } else if (!strcmp(arg, "--numstat"))
3655 options->output_format |= DIFF_FORMAT_NUMSTAT;
3656 else if (!strcmp(arg, "--shortstat"))
3657 options->output_format |= DIFF_FORMAT_SHORTSTAT;
3658 else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3659 return parse_dirstat_opt(options, "");
3660 else if (skip_prefix(arg, "-X", &arg))
3661 return parse_dirstat_opt(options, arg);
3662 else if (skip_prefix(arg, "--dirstat=", &arg))
3663 return parse_dirstat_opt(options, arg);
3664 else if (!strcmp(arg, "--cumulative"))
3665 return parse_dirstat_opt(options, "cumulative");
3666 else if (!strcmp(arg, "--dirstat-by-file"))
3667 return parse_dirstat_opt(options, "files");
3668 else if (skip_prefix(arg, "--dirstat-by-file=", &arg)) {
3669 parse_dirstat_opt(options, "files");
3670 return parse_dirstat_opt(options, arg);
3671 }
3672 else if (!strcmp(arg, "--check"))
3673 options->output_format |= DIFF_FORMAT_CHECKDIFF;
3674 else if (!strcmp(arg, "--summary"))
3675 options->output_format |= DIFF_FORMAT_SUMMARY;
3676 else if (!strcmp(arg, "--patch-with-stat")) {
3677 enable_patch_output(&options->output_format);
3678 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3679 } else if (!strcmp(arg, "--name-only"))
3680 options->output_format |= DIFF_FORMAT_NAME;
3681 else if (!strcmp(arg, "--name-status"))
3682 options->output_format |= DIFF_FORMAT_NAME_STATUS;
3683 else if (!strcmp(arg, "-s") || !strcmp(arg, "--no-patch"))
3684 options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3685 else if (starts_with(arg, "--stat"))
3686 /* --stat, --stat-width, --stat-name-width, or --stat-count */
3687 return stat_opt(options, av);
3688
3689 /* renames options */
3690 else if (starts_with(arg, "-B") || starts_with(arg, "--break-rewrites=") ||
3691 !strcmp(arg, "--break-rewrites")) {
3692 if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3693 return error("invalid argument to -B: %s", arg+2);
3694 }
3695 else if (starts_with(arg, "-M") || starts_with(arg, "--find-renames=") ||
3696 !strcmp(arg, "--find-renames")) {
3697 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3698 return error("invalid argument to -M: %s", arg+2);
3699 options->detect_rename = DIFF_DETECT_RENAME;
3700 }
3701 else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3702 options->irreversible_delete = 1;
3703 }
3704 else if (starts_with(arg, "-C") || starts_with(arg, "--find-copies=") ||
3705 !strcmp(arg, "--find-copies")) {
3706 if (options->detect_rename == DIFF_DETECT_COPY)
3707 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3708 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3709 return error("invalid argument to -C: %s", arg+2);
3710 options->detect_rename = DIFF_DETECT_COPY;
3711 }
3712 else if (!strcmp(arg, "--no-renames"))
3713 options->detect_rename = 0;
3714 else if (!strcmp(arg, "--rename-empty"))
3715 DIFF_OPT_SET(options, RENAME_EMPTY);
3716 else if (!strcmp(arg, "--no-rename-empty"))
3717 DIFF_OPT_CLR(options, RENAME_EMPTY);
3718 else if (!strcmp(arg, "--relative"))
3719 DIFF_OPT_SET(options, RELATIVE_NAME);
3720 else if (skip_prefix(arg, "--relative=", &arg)) {
3721 DIFF_OPT_SET(options, RELATIVE_NAME);
3722 options->prefix = arg;
3723 }
3724
3725 /* xdiff options */
3726 else if (!strcmp(arg, "--minimal"))
3727 DIFF_XDL_SET(options, NEED_MINIMAL);
3728 else if (!strcmp(arg, "--no-minimal"))
3729 DIFF_XDL_CLR(options, NEED_MINIMAL);
3730 else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3731 DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3732 else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3733 DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3734 else if (!strcmp(arg, "--ignore-space-at-eol"))
3735 DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3736 else if (!strcmp(arg, "--ignore-blank-lines"))
3737 DIFF_XDL_SET(options, IGNORE_BLANK_LINES);
3738 else if (!strcmp(arg, "--patience"))
3739 options->xdl_opts = DIFF_WITH_ALG(options, PATIENCE_DIFF);
3740 else if (!strcmp(arg, "--histogram"))
3741 options->xdl_opts = DIFF_WITH_ALG(options, HISTOGRAM_DIFF);
3742 else if ((argcount = parse_long_opt("diff-algorithm", av, &optarg))) {
3743 long value = parse_algorithm_value(optarg);
3744 if (value < 0)
3745 return error("option diff-algorithm accepts \"myers\", "
3746 "\"minimal\", \"patience\" and \"histogram\"");
3747 /* clear out previous settings */
3748 DIFF_XDL_CLR(options, NEED_MINIMAL);
3749 options->xdl_opts &= ~XDF_DIFF_ALGORITHM_MASK;
3750 options->xdl_opts |= value;
3751 return argcount;
3752 }
3753
3754 /* flags options */
3755 else if (!strcmp(arg, "--binary")) {
3756 enable_patch_output(&options->output_format);
3757 DIFF_OPT_SET(options, BINARY);
3758 }
3759 else if (!strcmp(arg, "--full-index"))
3760 DIFF_OPT_SET(options, FULL_INDEX);
3761 else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
3762 DIFF_OPT_SET(options, TEXT);
3763 else if (!strcmp(arg, "-R"))
3764 DIFF_OPT_SET(options, REVERSE_DIFF);
3765 else if (!strcmp(arg, "--find-copies-harder"))
3766 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3767 else if (!strcmp(arg, "--follow"))
3768 DIFF_OPT_SET(options, FOLLOW_RENAMES);
3769 else if (!strcmp(arg, "--no-follow"))
3770 DIFF_OPT_CLR(options, FOLLOW_RENAMES);
3771 else if (!strcmp(arg, "--color"))
3772 options->use_color = 1;
3773 else if (skip_prefix(arg, "--color=", &arg)) {
3774 int value = git_config_colorbool(NULL, arg);
3775 if (value < 0)
3776 return error("option `color' expects \"always\", \"auto\", or \"never\"");
3777 options->use_color = value;
3778 }
3779 else if (!strcmp(arg, "--no-color"))
3780 options->use_color = 0;
3781 else if (!strcmp(arg, "--color-words")) {
3782 options->use_color = 1;
3783 options->word_diff = DIFF_WORDS_COLOR;
3784 }
3785 else if (skip_prefix(arg, "--color-words=", &arg)) {
3786 options->use_color = 1;
3787 options->word_diff = DIFF_WORDS_COLOR;
3788 options->word_regex = arg;
3789 }
3790 else if (!strcmp(arg, "--word-diff")) {
3791 if (options->word_diff == DIFF_WORDS_NONE)
3792 options->word_diff = DIFF_WORDS_PLAIN;
3793 }
3794 else if (skip_prefix(arg, "--word-diff=", &arg)) {
3795 if (!strcmp(arg, "plain"))
3796 options->word_diff = DIFF_WORDS_PLAIN;
3797 else if (!strcmp(arg, "color")) {
3798 options->use_color = 1;
3799 options->word_diff = DIFF_WORDS_COLOR;
3800 }
3801 else if (!strcmp(arg, "porcelain"))
3802 options->word_diff = DIFF_WORDS_PORCELAIN;
3803 else if (!strcmp(arg, "none"))
3804 options->word_diff = DIFF_WORDS_NONE;
3805 else
3806 die("bad --word-diff argument: %s", arg);
3807 }
3808 else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
3809 if (options->word_diff == DIFF_WORDS_NONE)
3810 options->word_diff = DIFF_WORDS_PLAIN;
3811 options->word_regex = optarg;
3812 return argcount;
3813 }
3814 else if (!strcmp(arg, "--exit-code"))
3815 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3816 else if (!strcmp(arg, "--quiet"))
3817 DIFF_OPT_SET(options, QUICK);
3818 else if (!strcmp(arg, "--ext-diff"))
3819 DIFF_OPT_SET(options, ALLOW_EXTERNAL);
3820 else if (!strcmp(arg, "--no-ext-diff"))
3821 DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
3822 else if (!strcmp(arg, "--textconv"))
3823 DIFF_OPT_SET(options, ALLOW_TEXTCONV);
3824 else if (!strcmp(arg, "--no-textconv"))
3825 DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
3826 else if (!strcmp(arg, "--ignore-submodules")) {
3827 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3828 handle_ignore_submodules_arg(options, "all");
3829 } else if (skip_prefix(arg, "--ignore-submodules=", &arg)) {
3830 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3831 handle_ignore_submodules_arg(options, arg);
3832 } else if (!strcmp(arg, "--submodule"))
3833 DIFF_OPT_SET(options, SUBMODULE_LOG);
3834 else if (skip_prefix(arg, "--submodule=", &arg))
3835 return parse_submodule_opt(options, arg);
3836
3837 /* misc options */
3838 else if (!strcmp(arg, "-z"))
3839 options->line_termination = 0;
3840 else if ((argcount = short_opt('l', av, &optarg))) {
3841 options->rename_limit = strtoul(optarg, NULL, 10);
3842 return argcount;
3843 }
3844 else if ((argcount = short_opt('S', av, &optarg))) {
3845 options->pickaxe = optarg;
3846 options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
3847 return argcount;
3848 } else if ((argcount = short_opt('G', av, &optarg))) {
3849 options->pickaxe = optarg;
3850 options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
3851 return argcount;
3852 }
3853 else if (!strcmp(arg, "--pickaxe-all"))
3854 options->pickaxe_opts |= DIFF_PICKAXE_ALL;
3855 else if (!strcmp(arg, "--pickaxe-regex"))
3856 options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
3857 else if ((argcount = short_opt('O', av, &optarg))) {
3858 options->orderfile = optarg;
3859 return argcount;
3860 }
3861 else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
3862 int offending = parse_diff_filter_opt(optarg, options);
3863 if (offending)
3864 die("unknown change class '%c' in --diff-filter=%s",
3865 offending, optarg);
3866 return argcount;
3867 }
3868 else if (!strcmp(arg, "--abbrev"))
3869 options->abbrev = DEFAULT_ABBREV;
3870 else if (skip_prefix(arg, "--abbrev=", &arg)) {
3871 options->abbrev = strtoul(arg, NULL, 10);
3872 if (options->abbrev < MINIMUM_ABBREV)
3873 options->abbrev = MINIMUM_ABBREV;
3874 else if (40 < options->abbrev)
3875 options->abbrev = 40;
3876 }
3877 else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
3878 options->a_prefix = optarg;
3879 return argcount;
3880 }
3881 else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
3882 options->b_prefix = optarg;
3883 return argcount;
3884 }
3885 else if (!strcmp(arg, "--no-prefix"))
3886 options->a_prefix = options->b_prefix = "";
3887 else if (opt_arg(arg, '\0', "inter-hunk-context",
3888 &options->interhunkcontext))
3889 ;
3890 else if (!strcmp(arg, "-W"))
3891 DIFF_OPT_SET(options, FUNCCONTEXT);
3892 else if (!strcmp(arg, "--function-context"))
3893 DIFF_OPT_SET(options, FUNCCONTEXT);
3894 else if (!strcmp(arg, "--no-function-context"))
3895 DIFF_OPT_CLR(options, FUNCCONTEXT);
3896 else if ((argcount = parse_long_opt("output", av, &optarg))) {
3897 options->file = fopen(optarg, "w");
3898 if (!options->file)
3899 die_errno("Could not open '%s'", optarg);
3900 options->close_file = 1;
3901 return argcount;
3902 } else
3903 return 0;
3904 return 1;
3905 }
3906
3907 int parse_rename_score(const char **cp_p)
3908 {
3909 unsigned long num, scale;
3910 int ch, dot;
3911 const char *cp = *cp_p;
3912
3913 num = 0;
3914 scale = 1;
3915 dot = 0;
3916 for (;;) {
3917 ch = *cp;
3918 if ( !dot && ch == '.' ) {
3919 scale = 1;
3920 dot = 1;
3921 } else if ( ch == '%' ) {
3922 scale = dot ? scale*100 : 100;
3923 cp++; /* % is always at the end */
3924 break;
3925 } else if ( ch >= '0' && ch <= '9' ) {
3926 if ( scale < 100000 ) {
3927 scale *= 10;
3928 num = (num*10) + (ch-'0');
3929 }
3930 } else {
3931 break;
3932 }
3933 cp++;
3934 }
3935 *cp_p = cp;
3936
3937 /* user says num divided by scale and we say internally that
3938 * is MAX_SCORE * num / scale.
3939 */
3940 return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
3941 }
3942
3943 static int diff_scoreopt_parse(const char *opt)
3944 {
3945 int opt1, opt2, cmd;
3946
3947 if (*opt++ != '-')
3948 return -1;
3949 cmd = *opt++;
3950 if (cmd == '-') {
3951 /* convert the long-form arguments into short-form versions */
3952 if (skip_prefix(opt, "break-rewrites", &opt)) {
3953 if (*opt == 0 || *opt++ == '=')
3954 cmd = 'B';
3955 } else if (skip_prefix(opt, "find-copies", &opt)) {
3956 if (*opt == 0 || *opt++ == '=')
3957 cmd = 'C';
3958 } else if (skip_prefix(opt, "find-renames", &opt)) {
3959 if (*opt == 0 || *opt++ == '=')
3960 cmd = 'M';
3961 }
3962 }
3963 if (cmd != 'M' && cmd != 'C' && cmd != 'B')
3964 return -1; /* that is not a -M, -C, or -B option */
3965
3966 opt1 = parse_rename_score(&opt);
3967 if (cmd != 'B')
3968 opt2 = 0;
3969 else {
3970 if (*opt == 0)
3971 opt2 = 0;
3972 else if (*opt != '/')
3973 return -1; /* we expect -B80/99 or -B80 */
3974 else {
3975 opt++;
3976 opt2 = parse_rename_score(&opt);
3977 }
3978 }
3979 if (*opt != 0)
3980 return -1;
3981 return opt1 | (opt2 << 16);
3982 }
3983
3984 struct diff_queue_struct diff_queued_diff;
3985
3986 void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
3987 {
3988 ALLOC_GROW(queue->queue, queue->nr + 1, queue->alloc);
3989 queue->queue[queue->nr++] = dp;
3990 }
3991
3992 struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
3993 struct diff_filespec *one,
3994 struct diff_filespec *two)
3995 {
3996 struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
3997 dp->one = one;
3998 dp->two = two;
3999 if (queue)
4000 diff_q(queue, dp);
4001 return dp;
4002 }
4003
4004 void diff_free_filepair(struct diff_filepair *p)
4005 {
4006 free_filespec(p->one);
4007 free_filespec(p->two);
4008 free(p);
4009 }
4010
4011 /* This is different from find_unique_abbrev() in that
4012 * it stuffs the result with dots for alignment.
4013 */
4014 const char *diff_unique_abbrev(const unsigned char *sha1, int len)
4015 {
4016 int abblen;
4017 const char *abbrev;
4018 if (len == 40)
4019 return sha1_to_hex(sha1);
4020
4021 abbrev = find_unique_abbrev(sha1, len);
4022 abblen = strlen(abbrev);
4023 if (abblen < 37) {
4024 static char hex[41];
4025 if (len < abblen && abblen <= len + 2)
4026 sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
4027 else
4028 sprintf(hex, "%s...", abbrev);
4029 return hex;
4030 }
4031 return sha1_to_hex(sha1);
4032 }
4033
4034 static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
4035 {
4036 int line_termination = opt->line_termination;
4037 int inter_name_termination = line_termination ? '\t' : '\0';
4038
4039 fprintf(opt->file, "%s", diff_line_prefix(opt));
4040 if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
4041 fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
4042 diff_unique_abbrev(p->one->sha1, opt->abbrev));
4043 fprintf(opt->file, "%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
4044 }
4045 if (p->score) {
4046 fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
4047 inter_name_termination);
4048 } else {
4049 fprintf(opt->file, "%c%c", p->status, inter_name_termination);
4050 }
4051
4052 if (p->status == DIFF_STATUS_COPIED ||
4053 p->status == DIFF_STATUS_RENAMED) {
4054 const char *name_a, *name_b;
4055 name_a = p->one->path;
4056 name_b = p->two->path;
4057 strip_prefix(opt->prefix_length, &name_a, &name_b);
4058 write_name_quoted(name_a, opt->file, inter_name_termination);
4059 write_name_quoted(name_b, opt->file, line_termination);
4060 } else {
4061 const char *name_a, *name_b;
4062 name_a = p->one->mode ? p->one->path : p->two->path;
4063 name_b = NULL;
4064 strip_prefix(opt->prefix_length, &name_a, &name_b);
4065 write_name_quoted(name_a, opt->file, line_termination);
4066 }
4067 }
4068
4069 int diff_unmodified_pair(struct diff_filepair *p)
4070 {
4071 /* This function is written stricter than necessary to support
4072 * the currently implemented transformers, but the idea is to
4073 * let transformers to produce diff_filepairs any way they want,
4074 * and filter and clean them up here before producing the output.
4075 */
4076 struct diff_filespec *one = p->one, *two = p->two;
4077
4078 if (DIFF_PAIR_UNMERGED(p))
4079 return 0; /* unmerged is interesting */
4080
4081 /* deletion, addition, mode or type change
4082 * and rename are all interesting.
4083 */
4084 if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
4085 DIFF_PAIR_MODE_CHANGED(p) ||
4086 strcmp(one->path, two->path))
4087 return 0;
4088
4089 /* both are valid and point at the same path. that is, we are
4090 * dealing with a change.
4091 */
4092 if (one->sha1_valid && two->sha1_valid &&
4093 !hashcmp(one->sha1, two->sha1) &&
4094 !one->dirty_submodule && !two->dirty_submodule)
4095 return 1; /* no change */
4096 if (!one->sha1_valid && !two->sha1_valid)
4097 return 1; /* both look at the same file on the filesystem. */
4098 return 0;
4099 }
4100
4101 static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
4102 {
4103 if (diff_unmodified_pair(p))
4104 return;
4105
4106 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4107 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4108 return; /* no tree diffs in patch format */
4109
4110 run_diff(p, o);
4111 }
4112
4113 static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
4114 struct diffstat_t *diffstat)
4115 {
4116 if (diff_unmodified_pair(p))
4117 return;
4118
4119 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4120 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4121 return; /* no useful stat for tree diffs */
4122
4123 run_diffstat(p, o, diffstat);
4124 }
4125
4126 static void diff_flush_checkdiff(struct diff_filepair *p,
4127 struct diff_options *o)
4128 {
4129 if (diff_unmodified_pair(p))
4130 return;
4131
4132 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4133 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4134 return; /* nothing to check in tree diffs */
4135
4136 run_checkdiff(p, o);
4137 }
4138
4139 int diff_queue_is_empty(void)
4140 {
4141 struct diff_queue_struct *q = &diff_queued_diff;
4142 int i;
4143 for (i = 0; i < q->nr; i++)
4144 if (!diff_unmodified_pair(q->queue[i]))
4145 return 0;
4146 return 1;
4147 }
4148
4149 #if DIFF_DEBUG
4150 void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
4151 {
4152 fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
4153 x, one ? one : "",
4154 s->path,
4155 DIFF_FILE_VALID(s) ? "valid" : "invalid",
4156 s->mode,
4157 s->sha1_valid ? sha1_to_hex(s->sha1) : "");
4158 fprintf(stderr, "queue[%d] %s size %lu\n",
4159 x, one ? one : "",
4160 s->size);
4161 }
4162
4163 void diff_debug_filepair(const struct diff_filepair *p, int i)
4164 {
4165 diff_debug_filespec(p->one, i, "one");
4166 diff_debug_filespec(p->two, i, "two");
4167 fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
4168 p->score, p->status ? p->status : '?',
4169 p->one->rename_used, p->broken_pair);
4170 }
4171
4172 void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
4173 {
4174 int i;
4175 if (msg)
4176 fprintf(stderr, "%s\n", msg);
4177 fprintf(stderr, "q->nr = %d\n", q->nr);
4178 for (i = 0; i < q->nr; i++) {
4179 struct diff_filepair *p = q->queue[i];
4180 diff_debug_filepair(p, i);
4181 }
4182 }
4183 #endif
4184
4185 static void diff_resolve_rename_copy(void)
4186 {
4187 int i;
4188 struct diff_filepair *p;
4189 struct diff_queue_struct *q = &diff_queued_diff;
4190
4191 diff_debug_queue("resolve-rename-copy", q);
4192
4193 for (i = 0; i < q->nr; i++) {
4194 p = q->queue[i];
4195 p->status = 0; /* undecided */
4196 if (DIFF_PAIR_UNMERGED(p))
4197 p->status = DIFF_STATUS_UNMERGED;
4198 else if (!DIFF_FILE_VALID(p->one))
4199 p->status = DIFF_STATUS_ADDED;
4200 else if (!DIFF_FILE_VALID(p->two))
4201 p->status = DIFF_STATUS_DELETED;
4202 else if (DIFF_PAIR_TYPE_CHANGED(p))
4203 p->status = DIFF_STATUS_TYPE_CHANGED;
4204
4205 /* from this point on, we are dealing with a pair
4206 * whose both sides are valid and of the same type, i.e.
4207 * either in-place edit or rename/copy edit.
4208 */
4209 else if (DIFF_PAIR_RENAME(p)) {
4210 /*
4211 * A rename might have re-connected a broken
4212 * pair up, causing the pathnames to be the
4213 * same again. If so, that's not a rename at
4214 * all, just a modification..
4215 *
4216 * Otherwise, see if this source was used for
4217 * multiple renames, in which case we decrement
4218 * the count, and call it a copy.
4219 */
4220 if (!strcmp(p->one->path, p->two->path))
4221 p->status = DIFF_STATUS_MODIFIED;
4222 else if (--p->one->rename_used > 0)
4223 p->status = DIFF_STATUS_COPIED;
4224 else
4225 p->status = DIFF_STATUS_RENAMED;
4226 }
4227 else if (hashcmp(p->one->sha1, p->two->sha1) ||
4228 p->one->mode != p->two->mode ||
4229 p->one->dirty_submodule ||
4230 p->two->dirty_submodule ||
4231 is_null_sha1(p->one->sha1))
4232 p->status = DIFF_STATUS_MODIFIED;
4233 else {
4234 /* This is a "no-change" entry and should not
4235 * happen anymore, but prepare for broken callers.
4236 */
4237 error("feeding unmodified %s to diffcore",
4238 p->one->path);
4239 p->status = DIFF_STATUS_UNKNOWN;
4240 }
4241 }
4242 diff_debug_queue("resolve-rename-copy done", q);
4243 }
4244
4245 static int check_pair_status(struct diff_filepair *p)
4246 {
4247 switch (p->status) {
4248 case DIFF_STATUS_UNKNOWN:
4249 return 0;
4250 case 0:
4251 die("internal error in diff-resolve-rename-copy");
4252 default:
4253 return 1;
4254 }
4255 }
4256
4257 static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
4258 {
4259 int fmt = opt->output_format;
4260
4261 if (fmt & DIFF_FORMAT_CHECKDIFF)
4262 diff_flush_checkdiff(p, opt);
4263 else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
4264 diff_flush_raw(p, opt);
4265 else if (fmt & DIFF_FORMAT_NAME) {
4266 const char *name_a, *name_b;
4267 name_a = p->two->path;
4268 name_b = NULL;
4269 strip_prefix(opt->prefix_length, &name_a, &name_b);
4270 write_name_quoted(name_a, opt->file, opt->line_termination);
4271 }
4272 }
4273
4274 static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
4275 {
4276 if (fs->mode)
4277 fprintf(file, " %s mode %06o ", newdelete, fs->mode);
4278 else
4279 fprintf(file, " %s ", newdelete);
4280 write_name_quoted(fs->path, file, '\n');
4281 }
4282
4283
4284 static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
4285 const char *line_prefix)
4286 {
4287 if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
4288 fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
4289 p->two->mode, show_name ? ' ' : '\n');
4290 if (show_name) {
4291 write_name_quoted(p->two->path, file, '\n');
4292 }
4293 }
4294 }
4295
4296 static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
4297 const char *line_prefix)
4298 {
4299 char *names = pprint_rename(p->one->path, p->two->path);
4300
4301 fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
4302 free(names);
4303 show_mode_change(file, p, 0, line_prefix);
4304 }
4305
4306 static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
4307 {
4308 FILE *file = opt->file;
4309 const char *line_prefix = diff_line_prefix(opt);
4310
4311 switch(p->status) {
4312 case DIFF_STATUS_DELETED:
4313 fputs(line_prefix, file);
4314 show_file_mode_name(file, "delete", p->one);
4315 break;
4316 case DIFF_STATUS_ADDED:
4317 fputs(line_prefix, file);
4318 show_file_mode_name(file, "create", p->two);
4319 break;
4320 case DIFF_STATUS_COPIED:
4321 fputs(line_prefix, file);
4322 show_rename_copy(file, "copy", p, line_prefix);
4323 break;
4324 case DIFF_STATUS_RENAMED:
4325 fputs(line_prefix, file);
4326 show_rename_copy(file, "rename", p, line_prefix);
4327 break;
4328 default:
4329 if (p->score) {
4330 fprintf(file, "%s rewrite ", line_prefix);
4331 write_name_quoted(p->two->path, file, ' ');
4332 fprintf(file, "(%d%%)\n", similarity_index(p));
4333 }
4334 show_mode_change(file, p, !p->score, line_prefix);
4335 break;
4336 }
4337 }
4338
4339 struct patch_id_t {
4340 git_SHA_CTX *ctx;
4341 int patchlen;
4342 };
4343
4344 static int remove_space(char *line, int len)
4345 {
4346 int i;
4347 char *dst = line;
4348 unsigned char c;
4349
4350 for (i = 0; i < len; i++)
4351 if (!isspace((c = line[i])))
4352 *dst++ = c;
4353
4354 return dst - line;
4355 }
4356
4357 static void patch_id_consume(void *priv, char *line, unsigned long len)
4358 {
4359 struct patch_id_t *data = priv;
4360 int new_len;
4361
4362 /* Ignore line numbers when computing the SHA1 of the patch */
4363 if (starts_with(line, "@@ -"))
4364 return;
4365
4366 new_len = remove_space(line, len);
4367
4368 git_SHA1_Update(data->ctx, line, new_len);
4369 data->patchlen += new_len;
4370 }
4371
4372 /* returns 0 upon success, and writes result into sha1 */
4373 static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
4374 {
4375 struct diff_queue_struct *q = &diff_queued_diff;
4376 int i;
4377 git_SHA_CTX ctx;
4378 struct patch_id_t data;
4379 char buffer[PATH_MAX * 4 + 20];
4380
4381 git_SHA1_Init(&ctx);
4382 memset(&data, 0, sizeof(struct patch_id_t));
4383 data.ctx = &ctx;
4384
4385 for (i = 0; i < q->nr; i++) {
4386 xpparam_t xpp;
4387 xdemitconf_t xecfg;
4388 mmfile_t mf1, mf2;
4389 struct diff_filepair *p = q->queue[i];
4390 int len1, len2;
4391
4392 memset(&xpp, 0, sizeof(xpp));
4393 memset(&xecfg, 0, sizeof(xecfg));
4394 if (p->status == 0)
4395 return error("internal diff status error");
4396 if (p->status == DIFF_STATUS_UNKNOWN)
4397 continue;
4398 if (diff_unmodified_pair(p))
4399 continue;
4400 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4401 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4402 continue;
4403 if (DIFF_PAIR_UNMERGED(p))
4404 continue;
4405
4406 diff_fill_sha1_info(p->one);
4407 diff_fill_sha1_info(p->two);
4408 if (fill_mmfile(&mf1, p->one) < 0 ||
4409 fill_mmfile(&mf2, p->two) < 0)
4410 return error("unable to read files to diff");
4411
4412 len1 = remove_space(p->one->path, strlen(p->one->path));
4413 len2 = remove_space(p->two->path, strlen(p->two->path));
4414 if (p->one->mode == 0)
4415 len1 = snprintf(buffer, sizeof(buffer),
4416 "diff--gita/%.*sb/%.*s"
4417 "newfilemode%06o"
4418 "---/dev/null"
4419 "+++b/%.*s",
4420 len1, p->one->path,
4421 len2, p->two->path,
4422 p->two->mode,
4423 len2, p->two->path);
4424 else if (p->two->mode == 0)
4425 len1 = snprintf(buffer, sizeof(buffer),
4426 "diff--gita/%.*sb/%.*s"
4427 "deletedfilemode%06o"
4428 "---a/%.*s"
4429 "+++/dev/null",
4430 len1, p->one->path,
4431 len2, p->two->path,
4432 p->one->mode,
4433 len1, p->one->path);
4434 else
4435 len1 = snprintf(buffer, sizeof(buffer),
4436 "diff--gita/%.*sb/%.*s"
4437 "---a/%.*s"
4438 "+++b/%.*s",
4439 len1, p->one->path,
4440 len2, p->two->path,
4441 len1, p->one->path,
4442 len2, p->two->path);
4443 git_SHA1_Update(&ctx, buffer, len1);
4444
4445 if (diff_filespec_is_binary(p->one) ||
4446 diff_filespec_is_binary(p->two)) {
4447 git_SHA1_Update(&ctx, sha1_to_hex(p->one->sha1), 40);
4448 git_SHA1_Update(&ctx, sha1_to_hex(p->two->sha1), 40);
4449 continue;
4450 }
4451
4452 xpp.flags = 0;
4453 xecfg.ctxlen = 3;
4454 xecfg.flags = 0;
4455 xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4456 &xpp, &xecfg);
4457 }
4458
4459 git_SHA1_Final(sha1, &ctx);
4460 return 0;
4461 }
4462
4463 int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
4464 {
4465 struct diff_queue_struct *q = &diff_queued_diff;
4466 int i;
4467 int result = diff_get_patch_id(options, sha1);
4468
4469 for (i = 0; i < q->nr; i++)
4470 diff_free_filepair(q->queue[i]);
4471
4472 free(q->queue);
4473 DIFF_QUEUE_CLEAR(q);
4474
4475 return result;
4476 }
4477
4478 static int is_summary_empty(const struct diff_queue_struct *q)
4479 {
4480 int i;
4481
4482 for (i = 0; i < q->nr; i++) {
4483 const struct diff_filepair *p = q->queue[i];
4484
4485 switch (p->status) {
4486 case DIFF_STATUS_DELETED:
4487 case DIFF_STATUS_ADDED:
4488 case DIFF_STATUS_COPIED:
4489 case DIFF_STATUS_RENAMED:
4490 return 0;
4491 default:
4492 if (p->score)
4493 return 0;
4494 if (p->one->mode && p->two->mode &&
4495 p->one->mode != p->two->mode)
4496 return 0;
4497 break;
4498 }
4499 }
4500 return 1;
4501 }
4502
4503 static const char rename_limit_warning[] =
4504 "inexact rename detection was skipped due to too many files.";
4505
4506 static const char degrade_cc_to_c_warning[] =
4507 "only found copies from modified paths due to too many files.";
4508
4509 static const char rename_limit_advice[] =
4510 "you may want to set your %s variable to at least "
4511 "%d and retry the command.";
4512
4513 void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4514 {
4515 if (degraded_cc)
4516 warning(degrade_cc_to_c_warning);
4517 else if (needed)
4518 warning(rename_limit_warning);
4519 else
4520 return;
4521 if (0 < needed && needed < 32767)
4522 warning(rename_limit_advice, varname, needed);
4523 }
4524
4525 void diff_flush(struct diff_options *options)
4526 {
4527 struct diff_queue_struct *q = &diff_queued_diff;
4528 int i, output_format = options->output_format;
4529 int separator = 0;
4530 int dirstat_by_line = 0;
4531
4532 /*
4533 * Order: raw, stat, summary, patch
4534 * or: name/name-status/checkdiff (other bits clear)
4535 */
4536 if (!q->nr)
4537 goto free_queue;
4538
4539 if (output_format & (DIFF_FORMAT_RAW |
4540 DIFF_FORMAT_NAME |
4541 DIFF_FORMAT_NAME_STATUS |
4542 DIFF_FORMAT_CHECKDIFF)) {
4543 for (i = 0; i < q->nr; i++) {
4544 struct diff_filepair *p = q->queue[i];
4545 if (check_pair_status(p))
4546 flush_one_pair(p, options);
4547 }
4548 separator++;
4549 }
4550
4551 if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4552 dirstat_by_line = 1;
4553
4554 if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4555 dirstat_by_line) {
4556 struct diffstat_t diffstat;
4557
4558 memset(&diffstat, 0, sizeof(struct diffstat_t));
4559 for (i = 0; i < q->nr; i++) {
4560 struct diff_filepair *p = q->queue[i];
4561 if (check_pair_status(p))
4562 diff_flush_stat(p, options, &diffstat);
4563 }
4564 if (output_format & DIFF_FORMAT_NUMSTAT)
4565 show_numstat(&diffstat, options);
4566 if (output_format & DIFF_FORMAT_DIFFSTAT)
4567 show_stats(&diffstat, options);
4568 if (output_format & DIFF_FORMAT_SHORTSTAT)
4569 show_shortstats(&diffstat, options);
4570 if (output_format & DIFF_FORMAT_DIRSTAT)
4571 show_dirstat_by_line(&diffstat, options);
4572 free_diffstat_info(&diffstat);
4573 separator++;
4574 }
4575 if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4576 show_dirstat(options);
4577
4578 if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4579 for (i = 0; i < q->nr; i++) {
4580 diff_summary(options, q->queue[i]);
4581 }
4582 separator++;
4583 }
4584
4585 if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4586 DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4587 DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4588 /*
4589 * run diff_flush_patch for the exit status. setting
4590 * options->file to /dev/null should be safe, because we
4591 * aren't supposed to produce any output anyway.
4592 */
4593 if (options->close_file)
4594 fclose(options->file);
4595 options->file = fopen("/dev/null", "w");
4596 if (!options->file)
4597 die_errno("Could not open /dev/null");
4598 options->close_file = 1;
4599 for (i = 0; i < q->nr; i++) {
4600 struct diff_filepair *p = q->queue[i];
4601 if (check_pair_status(p))
4602 diff_flush_patch(p, options);
4603 if (options->found_changes)
4604 break;
4605 }
4606 }
4607
4608 if (output_format & DIFF_FORMAT_PATCH) {
4609 if (separator) {
4610 fprintf(options->file, "%s%c",
4611 diff_line_prefix(options),
4612 options->line_termination);
4613 if (options->stat_sep) {
4614 /* attach patch instead of inline */
4615 fputs(options->stat_sep, options->file);
4616 }
4617 }
4618
4619 for (i = 0; i < q->nr; i++) {
4620 struct diff_filepair *p = q->queue[i];
4621 if (check_pair_status(p))
4622 diff_flush_patch(p, options);
4623 }
4624 }
4625
4626 if (output_format & DIFF_FORMAT_CALLBACK)
4627 options->format_callback(q, options, options->format_callback_data);
4628
4629 for (i = 0; i < q->nr; i++)
4630 diff_free_filepair(q->queue[i]);
4631 free_queue:
4632 free(q->queue);
4633 DIFF_QUEUE_CLEAR(q);
4634 if (options->close_file)
4635 fclose(options->file);
4636
4637 /*
4638 * Report the content-level differences with HAS_CHANGES;
4639 * diff_addremove/diff_change does not set the bit when
4640 * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4641 */
4642 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4643 if (options->found_changes)
4644 DIFF_OPT_SET(options, HAS_CHANGES);
4645 else
4646 DIFF_OPT_CLR(options, HAS_CHANGES);
4647 }
4648 }
4649
4650 static int match_filter(const struct diff_options *options, const struct diff_filepair *p)
4651 {
4652 return (((p->status == DIFF_STATUS_MODIFIED) &&
4653 ((p->score &&
4654 filter_bit_tst(DIFF_STATUS_FILTER_BROKEN, options)) ||
4655 (!p->score &&
4656 filter_bit_tst(DIFF_STATUS_MODIFIED, options)))) ||
4657 ((p->status != DIFF_STATUS_MODIFIED) &&
4658 filter_bit_tst(p->status, options)));
4659 }
4660
4661 static void diffcore_apply_filter(struct diff_options *options)
4662 {
4663 int i;
4664 struct diff_queue_struct *q = &diff_queued_diff;
4665 struct diff_queue_struct outq;
4666
4667 DIFF_QUEUE_CLEAR(&outq);
4668
4669 if (!options->filter)
4670 return;
4671
4672 if (filter_bit_tst(DIFF_STATUS_FILTER_AON, options)) {
4673 int found;
4674 for (i = found = 0; !found && i < q->nr; i++) {
4675 if (match_filter(options, q->queue[i]))
4676 found++;
4677 }
4678 if (found)
4679 return;
4680
4681 /* otherwise we will clear the whole queue
4682 * by copying the empty outq at the end of this
4683 * function, but first clear the current entries
4684 * in the queue.
4685 */
4686 for (i = 0; i < q->nr; i++)
4687 diff_free_filepair(q->queue[i]);
4688 }
4689 else {
4690 /* Only the matching ones */
4691 for (i = 0; i < q->nr; i++) {
4692 struct diff_filepair *p = q->queue[i];
4693 if (match_filter(options, p))
4694 diff_q(&outq, p);
4695 else
4696 diff_free_filepair(p);
4697 }
4698 }
4699 free(q->queue);
4700 *q = outq;
4701 }
4702
4703 /* Check whether two filespecs with the same mode and size are identical */
4704 static int diff_filespec_is_identical(struct diff_filespec *one,
4705 struct diff_filespec *two)
4706 {
4707 if (S_ISGITLINK(one->mode))
4708 return 0;
4709 if (diff_populate_filespec(one, 0))
4710 return 0;
4711 if (diff_populate_filespec(two, 0))
4712 return 0;
4713 return !memcmp(one->data, two->data, one->size);
4714 }
4715
4716 static int diff_filespec_check_stat_unmatch(struct diff_filepair *p)
4717 {
4718 if (p->done_skip_stat_unmatch)
4719 return p->skip_stat_unmatch_result;
4720
4721 p->done_skip_stat_unmatch = 1;
4722 p->skip_stat_unmatch_result = 0;
4723 /*
4724 * 1. Entries that come from stat info dirtiness
4725 * always have both sides (iow, not create/delete),
4726 * one side of the object name is unknown, with
4727 * the same mode and size. Keep the ones that
4728 * do not match these criteria. They have real
4729 * differences.
4730 *
4731 * 2. At this point, the file is known to be modified,
4732 * with the same mode and size, and the object
4733 * name of one side is unknown. Need to inspect
4734 * the identical contents.
4735 */
4736 if (!DIFF_FILE_VALID(p->one) || /* (1) */
4737 !DIFF_FILE_VALID(p->two) ||
4738 (p->one->sha1_valid && p->two->sha1_valid) ||
4739 (p->one->mode != p->two->mode) ||
4740 diff_populate_filespec(p->one, CHECK_SIZE_ONLY) ||
4741 diff_populate_filespec(p->two, CHECK_SIZE_ONLY) ||
4742 (p->one->size != p->two->size) ||
4743 !diff_filespec_is_identical(p->one, p->two)) /* (2) */
4744 p->skip_stat_unmatch_result = 1;
4745 return p->skip_stat_unmatch_result;
4746 }
4747
4748 static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
4749 {
4750 int i;
4751 struct diff_queue_struct *q = &diff_queued_diff;
4752 struct diff_queue_struct outq;
4753 DIFF_QUEUE_CLEAR(&outq);
4754
4755 for (i = 0; i < q->nr; i++) {
4756 struct diff_filepair *p = q->queue[i];
4757
4758 if (diff_filespec_check_stat_unmatch(p))
4759 diff_q(&outq, p);
4760 else {
4761 /*
4762 * The caller can subtract 1 from skip_stat_unmatch
4763 * to determine how many paths were dirty only
4764 * due to stat info mismatch.
4765 */
4766 if (!DIFF_OPT_TST(diffopt, NO_INDEX))
4767 diffopt->skip_stat_unmatch++;
4768 diff_free_filepair(p);
4769 }
4770 }
4771 free(q->queue);
4772 *q = outq;
4773 }
4774
4775 static int diffnamecmp(const void *a_, const void *b_)
4776 {
4777 const struct diff_filepair *a = *((const struct diff_filepair **)a_);
4778 const struct diff_filepair *b = *((const struct diff_filepair **)b_);
4779 const char *name_a, *name_b;
4780
4781 name_a = a->one ? a->one->path : a->two->path;
4782 name_b = b->one ? b->one->path : b->two->path;
4783 return strcmp(name_a, name_b);
4784 }
4785
4786 void diffcore_fix_diff_index(struct diff_options *options)
4787 {
4788 struct diff_queue_struct *q = &diff_queued_diff;
4789 qsort(q->queue, q->nr, sizeof(q->queue[0]), diffnamecmp);
4790 }
4791
4792 void diffcore_std(struct diff_options *options)
4793 {
4794 /* NOTE please keep the following in sync with diff_tree_combined() */
4795 if (options->skip_stat_unmatch)
4796 diffcore_skip_stat_unmatch(options);
4797 if (!options->found_follow) {
4798 /* See try_to_follow_renames() in tree-diff.c */
4799 if (options->break_opt != -1)
4800 diffcore_break(options->break_opt);
4801 if (options->detect_rename)
4802 diffcore_rename(options);
4803 if (options->break_opt != -1)
4804 diffcore_merge_broken();
4805 }
4806 if (options->pickaxe)
4807 diffcore_pickaxe(options);
4808 if (options->orderfile)
4809 diffcore_order(options->orderfile);
4810 if (!options->found_follow)
4811 /* See try_to_follow_renames() in tree-diff.c */
4812 diff_resolve_rename_copy();
4813 diffcore_apply_filter(options);
4814
4815 if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4816 DIFF_OPT_SET(options, HAS_CHANGES);
4817 else
4818 DIFF_OPT_CLR(options, HAS_CHANGES);
4819
4820 options->found_follow = 0;
4821 }
4822
4823 int diff_result_code(struct diff_options *opt, int status)
4824 {
4825 int result = 0;
4826
4827 diff_warn_rename_limit("diff.renameLimit",
4828 opt->needed_rename_limit,
4829 opt->degraded_cc_to_c);
4830 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4831 !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
4832 return status;
4833 if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4834 DIFF_OPT_TST(opt, HAS_CHANGES))
4835 result |= 01;
4836 if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
4837 DIFF_OPT_TST(opt, CHECK_FAILED))
4838 result |= 02;
4839 return result;
4840 }
4841
4842 int diff_can_quit_early(struct diff_options *opt)
4843 {
4844 return (DIFF_OPT_TST(opt, QUICK) &&
4845 !opt->filter &&
4846 DIFF_OPT_TST(opt, HAS_CHANGES));
4847 }
4848
4849 /*
4850 * Shall changes to this submodule be ignored?
4851 *
4852 * Submodule changes can be configured to be ignored separately for each path,
4853 * but that configuration can be overridden from the command line.
4854 */
4855 static int is_submodule_ignored(const char *path, struct diff_options *options)
4856 {
4857 int ignored = 0;
4858 unsigned orig_flags = options->flags;
4859 if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
4860 set_diffopt_flags_from_submodule_config(options, path);
4861 if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
4862 ignored = 1;
4863 options->flags = orig_flags;
4864 return ignored;
4865 }
4866
4867 void diff_addremove(struct diff_options *options,
4868 int addremove, unsigned mode,
4869 const unsigned char *sha1,
4870 int sha1_valid,
4871 const char *concatpath, unsigned dirty_submodule)
4872 {
4873 struct diff_filespec *one, *two;
4874
4875 if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
4876 return;
4877
4878 /* This may look odd, but it is a preparation for
4879 * feeding "there are unchanged files which should
4880 * not produce diffs, but when you are doing copy
4881 * detection you would need them, so here they are"
4882 * entries to the diff-core. They will be prefixed
4883 * with something like '=' or '*' (I haven't decided
4884 * which but should not make any difference).
4885 * Feeding the same new and old to diff_change()
4886 * also has the same effect.
4887 * Before the final output happens, they are pruned after
4888 * merged into rename/copy pairs as appropriate.
4889 */
4890 if (DIFF_OPT_TST(options, REVERSE_DIFF))
4891 addremove = (addremove == '+' ? '-' :
4892 addremove == '-' ? '+' : addremove);
4893
4894 if (options->prefix &&
4895 strncmp(concatpath, options->prefix, options->prefix_length))
4896 return;
4897
4898 one = alloc_filespec(concatpath);
4899 two = alloc_filespec(concatpath);
4900
4901 if (addremove != '+')
4902 fill_filespec(one, sha1, sha1_valid, mode);
4903 if (addremove != '-') {
4904 fill_filespec(two, sha1, sha1_valid, mode);
4905 two->dirty_submodule = dirty_submodule;
4906 }
4907
4908 diff_queue(&diff_queued_diff, one, two);
4909 if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4910 DIFF_OPT_SET(options, HAS_CHANGES);
4911 }
4912
4913 void diff_change(struct diff_options *options,
4914 unsigned old_mode, unsigned new_mode,
4915 const unsigned char *old_sha1,
4916 const unsigned char *new_sha1,
4917 int old_sha1_valid, int new_sha1_valid,
4918 const char *concatpath,
4919 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
4920 {
4921 struct diff_filespec *one, *two;
4922 struct diff_filepair *p;
4923
4924 if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
4925 is_submodule_ignored(concatpath, options))
4926 return;
4927
4928 if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
4929 unsigned tmp;
4930 const unsigned char *tmp_c;
4931 tmp = old_mode; old_mode = new_mode; new_mode = tmp;
4932 tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
4933 tmp = old_sha1_valid; old_sha1_valid = new_sha1_valid;
4934 new_sha1_valid = tmp;
4935 tmp = old_dirty_submodule; old_dirty_submodule = new_dirty_submodule;
4936 new_dirty_submodule = tmp;
4937 }
4938
4939 if (options->prefix &&
4940 strncmp(concatpath, options->prefix, options->prefix_length))
4941 return;
4942
4943 one = alloc_filespec(concatpath);
4944 two = alloc_filespec(concatpath);
4945 fill_filespec(one, old_sha1, old_sha1_valid, old_mode);
4946 fill_filespec(two, new_sha1, new_sha1_valid, new_mode);
4947 one->dirty_submodule = old_dirty_submodule;
4948 two->dirty_submodule = new_dirty_submodule;
4949 p = diff_queue(&diff_queued_diff, one, two);
4950
4951 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4952 return;
4953
4954 if (DIFF_OPT_TST(options, QUICK) && options->skip_stat_unmatch &&
4955 !diff_filespec_check_stat_unmatch(p))
4956 return;
4957
4958 DIFF_OPT_SET(options, HAS_CHANGES);
4959 }
4960
4961 struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
4962 {
4963 struct diff_filepair *pair;
4964 struct diff_filespec *one, *two;
4965
4966 if (options->prefix &&
4967 strncmp(path, options->prefix, options->prefix_length))
4968 return NULL;
4969
4970 one = alloc_filespec(path);
4971 two = alloc_filespec(path);
4972 pair = diff_queue(&diff_queued_diff, one, two);
4973 pair->is_unmerged = 1;
4974 return pair;
4975 }
4976
4977 static char *run_textconv(const char *pgm, struct diff_filespec *spec,
4978 size_t *outsize)
4979 {
4980 struct diff_tempfile *temp;
4981 const char *argv[3];
4982 const char **arg = argv;
4983 struct child_process child = CHILD_PROCESS_INIT;
4984 struct strbuf buf = STRBUF_INIT;
4985 int err = 0;
4986
4987 temp = prepare_temp_file(spec->path, spec);
4988 *arg++ = pgm;
4989 *arg++ = temp->name;
4990 *arg = NULL;
4991
4992 child.use_shell = 1;
4993 child.argv = argv;
4994 child.out = -1;
4995 if (start_command(&child)) {
4996 remove_tempfile();
4997 return NULL;
4998 }
4999
5000 if (strbuf_read(&buf, child.out, 0) < 0)
5001 err = error("error reading from textconv command '%s'", pgm);
5002 close(child.out);
5003
5004 if (finish_command(&child) || err) {
5005 strbuf_release(&buf);
5006 remove_tempfile();
5007 return NULL;
5008 }
5009 remove_tempfile();
5010
5011 return strbuf_detach(&buf, outsize);
5012 }
5013
5014 size_t fill_textconv(struct userdiff_driver *driver,
5015 struct diff_filespec *df,
5016 char **outbuf)
5017 {
5018 size_t size;
5019
5020 if (!driver || !driver->textconv) {
5021 if (!DIFF_FILE_VALID(df)) {
5022 *outbuf = "";
5023 return 0;
5024 }
5025 if (diff_populate_filespec(df, 0))
5026 die("unable to read files to diff");
5027 *outbuf = df->data;
5028 return df->size;
5029 }
5030
5031 if (driver->textconv_cache && df->sha1_valid) {
5032 *outbuf = notes_cache_get(driver->textconv_cache, df->sha1,
5033 &size);
5034 if (*outbuf)
5035 return size;
5036 }
5037
5038 *outbuf = run_textconv(driver->textconv, df, &size);
5039 if (!*outbuf)
5040 die("unable to read files to diff");
5041
5042 if (driver->textconv_cache && df->sha1_valid) {
5043 /* ignore errors, as we might be in a readonly repository */
5044 notes_cache_put(driver->textconv_cache, df->sha1, *outbuf,
5045 size);
5046 /*
5047 * we could save up changes and flush them all at the end,
5048 * but we would need an extra call after all diffing is done.
5049 * Since generating a cache entry is the slow path anyway,
5050 * this extra overhead probably isn't a big deal.
5051 */
5052 notes_cache_write(driver->textconv_cache);
5053 }
5054
5055 return size;
5056 }
5057
5058 void setup_diff_pager(struct diff_options *opt)
5059 {
5060 /*
5061 * If the user asked for our exit code, then either they want --quiet
5062 * or --exit-code. We should definitely not bother with a pager in the
5063 * former case, as we will generate no output. Since we still properly
5064 * report our exit code even when a pager is run, we _could_ run a
5065 * pager with --exit-code. But since we have not done so historically,
5066 * and because it is easy to find people oneline advising "git diff
5067 * --exit-code" in hooks and other scripts, we do not do so.
5068 */
5069 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5070 check_pager_config("diff") != 0)
5071 setup_pager();
5072 }
ramsay's fork of git