shallow: remove unused #include "sigchain.h"
[git/git.git] / diff.c
1 /*
2 * Copyright (C) 2005 Junio C Hamano
3 */
4 #include "cache.h"
5 #include "tempfile.h"
6 #include "quote.h"
7 #include "diff.h"
8 #include "diffcore.h"
9 #include "delta.h"
10 #include "xdiff-interface.h"
11 #include "color.h"
12 #include "attr.h"
13 #include "run-command.h"
14 #include "utf8.h"
15 #include "userdiff.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 /*
312 * Keep track of files used for diffing. Sometimes such an entry
313 * refers to a temporary file, sometimes to an existing file, and
314 * sometimes to "/dev/null".
315 */
316 static struct diff_tempfile {
317 /*
318 * filename external diff should read from, or NULL if this
319 * entry is currently not in use:
320 */
321 const char *name;
322
323 char hex[41];
324 char mode[10];
325
326 /*
327 * If this diff_tempfile instance refers to a temporary file,
328 * this tempfile object is used to manage its lifetime.
329 */
330 struct tempfile tempfile;
331 } diff_temp[2];
332
333 typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
334
335 struct emit_callback {
336 int color_diff;
337 unsigned ws_rule;
338 int blank_at_eof_in_preimage;
339 int blank_at_eof_in_postimage;
340 int lno_in_preimage;
341 int lno_in_postimage;
342 sane_truncate_fn truncate;
343 const char **label_path;
344 struct diff_words_data *diff_words;
345 struct diff_options *opt;
346 int *found_changesp;
347 struct strbuf *header;
348 };
349
350 static int count_lines(const char *data, int size)
351 {
352 int count, ch, completely_empty = 1, nl_just_seen = 0;
353 count = 0;
354 while (0 < size--) {
355 ch = *data++;
356 if (ch == '\n') {
357 count++;
358 nl_just_seen = 1;
359 completely_empty = 0;
360 }
361 else {
362 nl_just_seen = 0;
363 completely_empty = 0;
364 }
365 }
366 if (completely_empty)
367 return 0;
368 if (!nl_just_seen)
369 count++; /* no trailing newline */
370 return count;
371 }
372
373 static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
374 {
375 if (!DIFF_FILE_VALID(one)) {
376 mf->ptr = (char *)""; /* does not matter */
377 mf->size = 0;
378 return 0;
379 }
380 else if (diff_populate_filespec(one, 0))
381 return -1;
382
383 mf->ptr = one->data;
384 mf->size = one->size;
385 return 0;
386 }
387
388 /* like fill_mmfile, but only for size, so we can avoid retrieving blob */
389 static unsigned long diff_filespec_size(struct diff_filespec *one)
390 {
391 if (!DIFF_FILE_VALID(one))
392 return 0;
393 diff_populate_filespec(one, CHECK_SIZE_ONLY);
394 return one->size;
395 }
396
397 static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
398 {
399 char *ptr = mf->ptr;
400 long size = mf->size;
401 int cnt = 0;
402
403 if (!size)
404 return cnt;
405 ptr += size - 1; /* pointing at the very end */
406 if (*ptr != '\n')
407 ; /* incomplete line */
408 else
409 ptr--; /* skip the last LF */
410 while (mf->ptr < ptr) {
411 char *prev_eol;
412 for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
413 if (*prev_eol == '\n')
414 break;
415 if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
416 break;
417 cnt++;
418 ptr = prev_eol - 1;
419 }
420 return cnt;
421 }
422
423 static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
424 struct emit_callback *ecbdata)
425 {
426 int l1, l2, at;
427 unsigned ws_rule = ecbdata->ws_rule;
428 l1 = count_trailing_blank(mf1, ws_rule);
429 l2 = count_trailing_blank(mf2, ws_rule);
430 if (l2 <= l1) {
431 ecbdata->blank_at_eof_in_preimage = 0;
432 ecbdata->blank_at_eof_in_postimage = 0;
433 return;
434 }
435 at = count_lines(mf1->ptr, mf1->size);
436 ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
437
438 at = count_lines(mf2->ptr, mf2->size);
439 ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
440 }
441
442 static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
443 int first, const char *line, int len)
444 {
445 int has_trailing_newline, has_trailing_carriage_return;
446 int nofirst;
447 FILE *file = o->file;
448
449 fputs(diff_line_prefix(o), file);
450
451 if (len == 0) {
452 has_trailing_newline = (first == '\n');
453 has_trailing_carriage_return = (!has_trailing_newline &&
454 (first == '\r'));
455 nofirst = has_trailing_newline || has_trailing_carriage_return;
456 } else {
457 has_trailing_newline = (len > 0 && line[len-1] == '\n');
458 if (has_trailing_newline)
459 len--;
460 has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
461 if (has_trailing_carriage_return)
462 len--;
463 nofirst = 0;
464 }
465
466 if (len || !nofirst) {
467 fputs(set, file);
468 if (!nofirst)
469 fputc(first, file);
470 fwrite(line, len, 1, file);
471 fputs(reset, file);
472 }
473 if (has_trailing_carriage_return)
474 fputc('\r', file);
475 if (has_trailing_newline)
476 fputc('\n', file);
477 }
478
479 static void emit_line(struct diff_options *o, const char *set, const char *reset,
480 const char *line, int len)
481 {
482 emit_line_0(o, set, reset, line[0], line+1, len-1);
483 }
484
485 static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
486 {
487 if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
488 ecbdata->blank_at_eof_in_preimage &&
489 ecbdata->blank_at_eof_in_postimage &&
490 ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
491 ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
492 return 0;
493 return ws_blank_line(line, len, ecbdata->ws_rule);
494 }
495
496 static void emit_add_line(const char *reset,
497 struct emit_callback *ecbdata,
498 const char *line, int len)
499 {
500 const char *ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
501 const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_NEW);
502
503 if (!*ws)
504 emit_line_0(ecbdata->opt, set, reset, '+', line, len);
505 else if (new_blank_line_at_eof(ecbdata, line, len))
506 /* Blank line at EOF - paint '+' as well */
507 emit_line_0(ecbdata->opt, ws, reset, '+', line, len);
508 else {
509 /* Emit just the prefix, then the rest. */
510 emit_line_0(ecbdata->opt, set, reset, '+', "", 0);
511 ws_check_emit(line, len, ecbdata->ws_rule,
512 ecbdata->opt->file, set, reset, ws);
513 }
514 }
515
516 static void emit_hunk_header(struct emit_callback *ecbdata,
517 const char *line, int len)
518 {
519 const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
520 const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
521 const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
522 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
523 static const char atat[2] = { '@', '@' };
524 const char *cp, *ep;
525 struct strbuf msgbuf = STRBUF_INIT;
526 int org_len = len;
527 int i = 1;
528
529 /*
530 * As a hunk header must begin with "@@ -<old>, +<new> @@",
531 * it always is at least 10 bytes long.
532 */
533 if (len < 10 ||
534 memcmp(line, atat, 2) ||
535 !(ep = memmem(line + 2, len - 2, atat, 2))) {
536 emit_line(ecbdata->opt, plain, reset, line, len);
537 return;
538 }
539 ep += 2; /* skip over @@ */
540
541 /* The hunk header in fraginfo color */
542 strbuf_addstr(&msgbuf, frag);
543 strbuf_add(&msgbuf, line, ep - line);
544 strbuf_addstr(&msgbuf, reset);
545
546 /*
547 * trailing "\r\n"
548 */
549 for ( ; i < 3; i++)
550 if (line[len - i] == '\r' || line[len - i] == '\n')
551 len--;
552
553 /* blank before the func header */
554 for (cp = ep; ep - line < len; ep++)
555 if (*ep != ' ' && *ep != '\t')
556 break;
557 if (ep != cp) {
558 strbuf_addstr(&msgbuf, plain);
559 strbuf_add(&msgbuf, cp, ep - cp);
560 strbuf_addstr(&msgbuf, reset);
561 }
562
563 if (ep < line + len) {
564 strbuf_addstr(&msgbuf, func);
565 strbuf_add(&msgbuf, ep, line + len - ep);
566 strbuf_addstr(&msgbuf, reset);
567 }
568
569 strbuf_add(&msgbuf, line + len, org_len - len);
570 emit_line(ecbdata->opt, "", "", msgbuf.buf, msgbuf.len);
571 strbuf_release(&msgbuf);
572 }
573
574 static struct diff_tempfile *claim_diff_tempfile(void) {
575 int i;
576 for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
577 if (!diff_temp[i].name)
578 return diff_temp + i;
579 die("BUG: diff is failing to clean up its tempfiles");
580 }
581
582 static void remove_tempfile(void)
583 {
584 int i;
585 for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
586 if (is_tempfile_active(&diff_temp[i].tempfile))
587 delete_tempfile(&diff_temp[i].tempfile);
588 diff_temp[i].name = NULL;
589 }
590 }
591
592 static void print_line_count(FILE *file, int count)
593 {
594 switch (count) {
595 case 0:
596 fprintf(file, "0,0");
597 break;
598 case 1:
599 fprintf(file, "1");
600 break;
601 default:
602 fprintf(file, "1,%d", count);
603 break;
604 }
605 }
606
607 static void emit_rewrite_lines(struct emit_callback *ecb,
608 int prefix, const char *data, int size)
609 {
610 const char *endp = NULL;
611 static const char *nneof = " No newline at end of file\n";
612 const char *old = diff_get_color(ecb->color_diff, DIFF_FILE_OLD);
613 const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
614
615 while (0 < size) {
616 int len;
617
618 endp = memchr(data, '\n', size);
619 len = endp ? (endp - data + 1) : size;
620 if (prefix != '+') {
621 ecb->lno_in_preimage++;
622 emit_line_0(ecb->opt, old, reset, '-',
623 data, len);
624 } else {
625 ecb->lno_in_postimage++;
626 emit_add_line(reset, ecb, data, len);
627 }
628 size -= len;
629 data += len;
630 }
631 if (!endp) {
632 const char *plain = diff_get_color(ecb->color_diff,
633 DIFF_PLAIN);
634 putc('\n', ecb->opt->file);
635 emit_line_0(ecb->opt, plain, reset, '\\',
636 nneof, strlen(nneof));
637 }
638 }
639
640 static void emit_rewrite_diff(const char *name_a,
641 const char *name_b,
642 struct diff_filespec *one,
643 struct diff_filespec *two,
644 struct userdiff_driver *textconv_one,
645 struct userdiff_driver *textconv_two,
646 struct diff_options *o)
647 {
648 int lc_a, lc_b;
649 const char *name_a_tab, *name_b_tab;
650 const char *metainfo = diff_get_color(o->use_color, DIFF_METAINFO);
651 const char *fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
652 const char *reset = diff_get_color(o->use_color, DIFF_RESET);
653 static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
654 const char *a_prefix, *b_prefix;
655 char *data_one, *data_two;
656 size_t size_one, size_two;
657 struct emit_callback ecbdata;
658 const char *line_prefix = diff_line_prefix(o);
659
660 if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
661 a_prefix = o->b_prefix;
662 b_prefix = o->a_prefix;
663 } else {
664 a_prefix = o->a_prefix;
665 b_prefix = o->b_prefix;
666 }
667
668 name_a += (*name_a == '/');
669 name_b += (*name_b == '/');
670 name_a_tab = strchr(name_a, ' ') ? "\t" : "";
671 name_b_tab = strchr(name_b, ' ') ? "\t" : "";
672
673 strbuf_reset(&a_name);
674 strbuf_reset(&b_name);
675 quote_two_c_style(&a_name, a_prefix, name_a, 0);
676 quote_two_c_style(&b_name, b_prefix, name_b, 0);
677
678 size_one = fill_textconv(textconv_one, one, &data_one);
679 size_two = fill_textconv(textconv_two, two, &data_two);
680
681 memset(&ecbdata, 0, sizeof(ecbdata));
682 ecbdata.color_diff = want_color(o->use_color);
683 ecbdata.found_changesp = &o->found_changes;
684 ecbdata.ws_rule = whitespace_rule(name_b);
685 ecbdata.opt = o;
686 if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
687 mmfile_t mf1, mf2;
688 mf1.ptr = (char *)data_one;
689 mf2.ptr = (char *)data_two;
690 mf1.size = size_one;
691 mf2.size = size_two;
692 check_blank_at_eof(&mf1, &mf2, &ecbdata);
693 }
694 ecbdata.lno_in_preimage = 1;
695 ecbdata.lno_in_postimage = 1;
696
697 lc_a = count_lines(data_one, size_one);
698 lc_b = count_lines(data_two, size_two);
699 fprintf(o->file,
700 "%s%s--- %s%s%s\n%s%s+++ %s%s%s\n%s%s@@ -",
701 line_prefix, metainfo, a_name.buf, name_a_tab, reset,
702 line_prefix, metainfo, b_name.buf, name_b_tab, reset,
703 line_prefix, fraginfo);
704 if (!o->irreversible_delete)
705 print_line_count(o->file, lc_a);
706 else
707 fprintf(o->file, "?,?");
708 fprintf(o->file, " +");
709 print_line_count(o->file, lc_b);
710 fprintf(o->file, " @@%s\n", reset);
711 if (lc_a && !o->irreversible_delete)
712 emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
713 if (lc_b)
714 emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
715 if (textconv_one)
716 free((char *)data_one);
717 if (textconv_two)
718 free((char *)data_two);
719 }
720
721 struct diff_words_buffer {
722 mmfile_t text;
723 long alloc;
724 struct diff_words_orig {
725 const char *begin, *end;
726 } *orig;
727 int orig_nr, orig_alloc;
728 };
729
730 static void diff_words_append(char *line, unsigned long len,
731 struct diff_words_buffer *buffer)
732 {
733 ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
734 line++;
735 len--;
736 memcpy(buffer->text.ptr + buffer->text.size, line, len);
737 buffer->text.size += len;
738 buffer->text.ptr[buffer->text.size] = '\0';
739 }
740
741 struct diff_words_style_elem {
742 const char *prefix;
743 const char *suffix;
744 const char *color; /* NULL; filled in by the setup code if
745 * color is enabled */
746 };
747
748 struct diff_words_style {
749 enum diff_words_type type;
750 struct diff_words_style_elem new, old, ctx;
751 const char *newline;
752 };
753
754 static struct diff_words_style diff_words_styles[] = {
755 { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
756 { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
757 { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
758 };
759
760 struct diff_words_data {
761 struct diff_words_buffer minus, plus;
762 const char *current_plus;
763 int last_minus;
764 struct diff_options *opt;
765 regex_t *word_regex;
766 enum diff_words_type type;
767 struct diff_words_style *style;
768 };
769
770 static int fn_out_diff_words_write_helper(FILE *fp,
771 struct diff_words_style_elem *st_el,
772 const char *newline,
773 size_t count, const char *buf,
774 const char *line_prefix)
775 {
776 int print = 0;
777
778 while (count) {
779 char *p = memchr(buf, '\n', count);
780 if (print)
781 fputs(line_prefix, fp);
782 if (p != buf) {
783 if (st_el->color && fputs(st_el->color, fp) < 0)
784 return -1;
785 if (fputs(st_el->prefix, fp) < 0 ||
786 fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
787 fputs(st_el->suffix, fp) < 0)
788 return -1;
789 if (st_el->color && *st_el->color
790 && fputs(GIT_COLOR_RESET, fp) < 0)
791 return -1;
792 }
793 if (!p)
794 return 0;
795 if (fputs(newline, fp) < 0)
796 return -1;
797 count -= p + 1 - buf;
798 buf = p + 1;
799 print = 1;
800 }
801 return 0;
802 }
803
804 /*
805 * '--color-words' algorithm can be described as:
806 *
807 * 1. collect a the minus/plus lines of a diff hunk, divided into
808 * minus-lines and plus-lines;
809 *
810 * 2. break both minus-lines and plus-lines into words and
811 * place them into two mmfile_t with one word for each line;
812 *
813 * 3. use xdiff to run diff on the two mmfile_t to get the words level diff;
814 *
815 * And for the common parts of the both file, we output the plus side text.
816 * diff_words->current_plus is used to trace the current position of the plus file
817 * which printed. diff_words->last_minus is used to trace the last minus word
818 * printed.
819 *
820 * For '--graph' to work with '--color-words', we need to output the graph prefix
821 * on each line of color words output. Generally, there are two conditions on
822 * which we should output the prefix.
823 *
824 * 1. diff_words->last_minus == 0 &&
825 * diff_words->current_plus == diff_words->plus.text.ptr
826 *
827 * that is: the plus text must start as a new line, and if there is no minus
828 * word printed, a graph prefix must be printed.
829 *
830 * 2. diff_words->current_plus > diff_words->plus.text.ptr &&
831 * *(diff_words->current_plus - 1) == '\n'
832 *
833 * that is: a graph prefix must be printed following a '\n'
834 */
835 static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
836 {
837 if ((diff_words->last_minus == 0 &&
838 diff_words->current_plus == diff_words->plus.text.ptr) ||
839 (diff_words->current_plus > diff_words->plus.text.ptr &&
840 *(diff_words->current_plus - 1) == '\n')) {
841 return 1;
842 } else {
843 return 0;
844 }
845 }
846
847 static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
848 {
849 struct diff_words_data *diff_words = priv;
850 struct diff_words_style *style = diff_words->style;
851 int minus_first, minus_len, plus_first, plus_len;
852 const char *minus_begin, *minus_end, *plus_begin, *plus_end;
853 struct diff_options *opt = diff_words->opt;
854 const char *line_prefix;
855
856 if (line[0] != '@' || parse_hunk_header(line, len,
857 &minus_first, &minus_len, &plus_first, &plus_len))
858 return;
859
860 assert(opt);
861 line_prefix = diff_line_prefix(opt);
862
863 /* POSIX requires that first be decremented by one if len == 0... */
864 if (minus_len) {
865 minus_begin = diff_words->minus.orig[minus_first].begin;
866 minus_end =
867 diff_words->minus.orig[minus_first + minus_len - 1].end;
868 } else
869 minus_begin = minus_end =
870 diff_words->minus.orig[minus_first].end;
871
872 if (plus_len) {
873 plus_begin = diff_words->plus.orig[plus_first].begin;
874 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
875 } else
876 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
877
878 if (color_words_output_graph_prefix(diff_words)) {
879 fputs(line_prefix, diff_words->opt->file);
880 }
881 if (diff_words->current_plus != plus_begin) {
882 fn_out_diff_words_write_helper(diff_words->opt->file,
883 &style->ctx, style->newline,
884 plus_begin - diff_words->current_plus,
885 diff_words->current_plus, line_prefix);
886 if (*(plus_begin - 1) == '\n')
887 fputs(line_prefix, diff_words->opt->file);
888 }
889 if (minus_begin != minus_end) {
890 fn_out_diff_words_write_helper(diff_words->opt->file,
891 &style->old, style->newline,
892 minus_end - minus_begin, minus_begin,
893 line_prefix);
894 }
895 if (plus_begin != plus_end) {
896 fn_out_diff_words_write_helper(diff_words->opt->file,
897 &style->new, style->newline,
898 plus_end - plus_begin, plus_begin,
899 line_prefix);
900 }
901
902 diff_words->current_plus = plus_end;
903 diff_words->last_minus = minus_first;
904 }
905
906 /* This function starts looking at *begin, and returns 0 iff a word was found. */
907 static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
908 int *begin, int *end)
909 {
910 if (word_regex && *begin < buffer->size) {
911 regmatch_t match[1];
912 if (!regexec(word_regex, buffer->ptr + *begin, 1, match, 0)) {
913 char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
914 '\n', match[0].rm_eo - match[0].rm_so);
915 *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
916 *begin += match[0].rm_so;
917 return *begin >= *end;
918 }
919 return -1;
920 }
921
922 /* find the next word */
923 while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
924 (*begin)++;
925 if (*begin >= buffer->size)
926 return -1;
927
928 /* find the end of the word */
929 *end = *begin + 1;
930 while (*end < buffer->size && !isspace(buffer->ptr[*end]))
931 (*end)++;
932
933 return 0;
934 }
935
936 /*
937 * This function splits the words in buffer->text, stores the list with
938 * newline separator into out, and saves the offsets of the original words
939 * in buffer->orig.
940 */
941 static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
942 regex_t *word_regex)
943 {
944 int i, j;
945 long alloc = 0;
946
947 out->size = 0;
948 out->ptr = NULL;
949
950 /* fake an empty "0th" word */
951 ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
952 buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
953 buffer->orig_nr = 1;
954
955 for (i = 0; i < buffer->text.size; i++) {
956 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
957 return;
958
959 /* store original boundaries */
960 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
961 buffer->orig_alloc);
962 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
963 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
964 buffer->orig_nr++;
965
966 /* store one word */
967 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
968 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
969 out->ptr[out->size + j - i] = '\n';
970 out->size += j - i + 1;
971
972 i = j - 1;
973 }
974 }
975
976 /* this executes the word diff on the accumulated buffers */
977 static void diff_words_show(struct diff_words_data *diff_words)
978 {
979 xpparam_t xpp;
980 xdemitconf_t xecfg;
981 mmfile_t minus, plus;
982 struct diff_words_style *style = diff_words->style;
983
984 struct diff_options *opt = diff_words->opt;
985 const char *line_prefix;
986
987 assert(opt);
988 line_prefix = diff_line_prefix(opt);
989
990 /* special case: only removal */
991 if (!diff_words->plus.text.size) {
992 fputs(line_prefix, diff_words->opt->file);
993 fn_out_diff_words_write_helper(diff_words->opt->file,
994 &style->old, style->newline,
995 diff_words->minus.text.size,
996 diff_words->minus.text.ptr, line_prefix);
997 diff_words->minus.text.size = 0;
998 return;
999 }
1000
1001 diff_words->current_plus = diff_words->plus.text.ptr;
1002 diff_words->last_minus = 0;
1003
1004 memset(&xpp, 0, sizeof(xpp));
1005 memset(&xecfg, 0, sizeof(xecfg));
1006 diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
1007 diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
1008 xpp.flags = 0;
1009 /* as only the hunk header will be parsed, we need a 0-context */
1010 xecfg.ctxlen = 0;
1011 xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
1012 &xpp, &xecfg);
1013 free(minus.ptr);
1014 free(plus.ptr);
1015 if (diff_words->current_plus != diff_words->plus.text.ptr +
1016 diff_words->plus.text.size) {
1017 if (color_words_output_graph_prefix(diff_words))
1018 fputs(line_prefix, diff_words->opt->file);
1019 fn_out_diff_words_write_helper(diff_words->opt->file,
1020 &style->ctx, style->newline,
1021 diff_words->plus.text.ptr + diff_words->plus.text.size
1022 - diff_words->current_plus, diff_words->current_plus,
1023 line_prefix);
1024 }
1025 diff_words->minus.text.size = diff_words->plus.text.size = 0;
1026 }
1027
1028 /* In "color-words" mode, show word-diff of words accumulated in the buffer */
1029 static void diff_words_flush(struct emit_callback *ecbdata)
1030 {
1031 if (ecbdata->diff_words->minus.text.size ||
1032 ecbdata->diff_words->plus.text.size)
1033 diff_words_show(ecbdata->diff_words);
1034 }
1035
1036 static void diff_filespec_load_driver(struct diff_filespec *one)
1037 {
1038 /* Use already-loaded driver */
1039 if (one->driver)
1040 return;
1041
1042 if (S_ISREG(one->mode))
1043 one->driver = userdiff_find_by_path(one->path);
1044
1045 /* Fallback to default settings */
1046 if (!one->driver)
1047 one->driver = userdiff_find_by_name("default");
1048 }
1049
1050 static const char *userdiff_word_regex(struct diff_filespec *one)
1051 {
1052 diff_filespec_load_driver(one);
1053 return one->driver->word_regex;
1054 }
1055
1056 static void init_diff_words_data(struct emit_callback *ecbdata,
1057 struct diff_options *orig_opts,
1058 struct diff_filespec *one,
1059 struct diff_filespec *two)
1060 {
1061 int i;
1062 struct diff_options *o = xmalloc(sizeof(struct diff_options));
1063 memcpy(o, orig_opts, sizeof(struct diff_options));
1064
1065 ecbdata->diff_words =
1066 xcalloc(1, sizeof(struct diff_words_data));
1067 ecbdata->diff_words->type = o->word_diff;
1068 ecbdata->diff_words->opt = o;
1069 if (!o->word_regex)
1070 o->word_regex = userdiff_word_regex(one);
1071 if (!o->word_regex)
1072 o->word_regex = userdiff_word_regex(two);
1073 if (!o->word_regex)
1074 o->word_regex = diff_word_regex_cfg;
1075 if (o->word_regex) {
1076 ecbdata->diff_words->word_regex = (regex_t *)
1077 xmalloc(sizeof(regex_t));
1078 if (regcomp(ecbdata->diff_words->word_regex,
1079 o->word_regex,
1080 REG_EXTENDED | REG_NEWLINE))
1081 die ("Invalid regular expression: %s",
1082 o->word_regex);
1083 }
1084 for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1085 if (o->word_diff == diff_words_styles[i].type) {
1086 ecbdata->diff_words->style =
1087 &diff_words_styles[i];
1088 break;
1089 }
1090 }
1091 if (want_color(o->use_color)) {
1092 struct diff_words_style *st = ecbdata->diff_words->style;
1093 st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1094 st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1095 st->ctx.color = diff_get_color_opt(o, DIFF_PLAIN);
1096 }
1097 }
1098
1099 static void free_diff_words_data(struct emit_callback *ecbdata)
1100 {
1101 if (ecbdata->diff_words) {
1102 diff_words_flush(ecbdata);
1103 free (ecbdata->diff_words->opt);
1104 free (ecbdata->diff_words->minus.text.ptr);
1105 free (ecbdata->diff_words->minus.orig);
1106 free (ecbdata->diff_words->plus.text.ptr);
1107 free (ecbdata->diff_words->plus.orig);
1108 if (ecbdata->diff_words->word_regex) {
1109 regfree(ecbdata->diff_words->word_regex);
1110 free(ecbdata->diff_words->word_regex);
1111 }
1112 free(ecbdata->diff_words);
1113 ecbdata->diff_words = NULL;
1114 }
1115 }
1116
1117 const char *diff_get_color(int diff_use_color, enum color_diff ix)
1118 {
1119 if (want_color(diff_use_color))
1120 return diff_colors[ix];
1121 return "";
1122 }
1123
1124 const char *diff_line_prefix(struct diff_options *opt)
1125 {
1126 struct strbuf *msgbuf;
1127 if (!opt->output_prefix)
1128 return "";
1129
1130 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
1131 return msgbuf->buf;
1132 }
1133
1134 static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1135 {
1136 const char *cp;
1137 unsigned long allot;
1138 size_t l = len;
1139
1140 if (ecb->truncate)
1141 return ecb->truncate(line, len);
1142 cp = line;
1143 allot = l;
1144 while (0 < l) {
1145 (void) utf8_width(&cp, &l);
1146 if (!cp)
1147 break; /* truncated in the middle? */
1148 }
1149 return allot - l;
1150 }
1151
1152 static void find_lno(const char *line, struct emit_callback *ecbdata)
1153 {
1154 const char *p;
1155 ecbdata->lno_in_preimage = 0;
1156 ecbdata->lno_in_postimage = 0;
1157 p = strchr(line, '-');
1158 if (!p)
1159 return; /* cannot happen */
1160 ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1161 p = strchr(p, '+');
1162 if (!p)
1163 return; /* cannot happen */
1164 ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1165 }
1166
1167 static void fn_out_consume(void *priv, char *line, unsigned long len)
1168 {
1169 struct emit_callback *ecbdata = priv;
1170 const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
1171 const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
1172 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1173 struct diff_options *o = ecbdata->opt;
1174 const char *line_prefix = diff_line_prefix(o);
1175
1176 if (ecbdata->header) {
1177 fprintf(ecbdata->opt->file, "%s", ecbdata->header->buf);
1178 strbuf_reset(ecbdata->header);
1179 ecbdata->header = NULL;
1180 }
1181 *(ecbdata->found_changesp) = 1;
1182
1183 if (ecbdata->label_path[0]) {
1184 const char *name_a_tab, *name_b_tab;
1185
1186 name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
1187 name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
1188
1189 fprintf(ecbdata->opt->file, "%s%s--- %s%s%s\n",
1190 line_prefix, meta, ecbdata->label_path[0], reset, name_a_tab);
1191 fprintf(ecbdata->opt->file, "%s%s+++ %s%s%s\n",
1192 line_prefix, meta, ecbdata->label_path[1], reset, name_b_tab);
1193 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1194 }
1195
1196 if (diff_suppress_blank_empty
1197 && len == 2 && line[0] == ' ' && line[1] == '\n') {
1198 line[0] = '\n';
1199 len = 1;
1200 }
1201
1202 if (line[0] == '@') {
1203 if (ecbdata->diff_words)
1204 diff_words_flush(ecbdata);
1205 len = sane_truncate_line(ecbdata, line, len);
1206 find_lno(line, ecbdata);
1207 emit_hunk_header(ecbdata, line, len);
1208 if (line[len-1] != '\n')
1209 putc('\n', ecbdata->opt->file);
1210 return;
1211 }
1212
1213 if (len < 1) {
1214 emit_line(ecbdata->opt, reset, reset, line, len);
1215 if (ecbdata->diff_words
1216 && ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN)
1217 fputs("~\n", ecbdata->opt->file);
1218 return;
1219 }
1220
1221 if (ecbdata->diff_words) {
1222 if (line[0] == '-') {
1223 diff_words_append(line, len,
1224 &ecbdata->diff_words->minus);
1225 return;
1226 } else if (line[0] == '+') {
1227 diff_words_append(line, len,
1228 &ecbdata->diff_words->plus);
1229 return;
1230 } else if (starts_with(line, "\\ ")) {
1231 /*
1232 * Eat the "no newline at eof" marker as if we
1233 * saw a "+" or "-" line with nothing on it,
1234 * and return without diff_words_flush() to
1235 * defer processing. If this is the end of
1236 * preimage, more "+" lines may come after it.
1237 */
1238 return;
1239 }
1240 diff_words_flush(ecbdata);
1241 if (ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN) {
1242 emit_line(ecbdata->opt, plain, reset, line, len);
1243 fputs("~\n", ecbdata->opt->file);
1244 } else {
1245 /*
1246 * Skip the prefix character, if any. With
1247 * diff_suppress_blank_empty, there may be
1248 * none.
1249 */
1250 if (line[0] != '\n') {
1251 line++;
1252 len--;
1253 }
1254 emit_line(ecbdata->opt, plain, reset, line, len);
1255 }
1256 return;
1257 }
1258
1259 if (line[0] != '+') {
1260 const char *color =
1261 diff_get_color(ecbdata->color_diff,
1262 line[0] == '-' ? DIFF_FILE_OLD : DIFF_PLAIN);
1263 ecbdata->lno_in_preimage++;
1264 if (line[0] == ' ')
1265 ecbdata->lno_in_postimage++;
1266 emit_line(ecbdata->opt, color, reset, line, len);
1267 } else {
1268 ecbdata->lno_in_postimage++;
1269 emit_add_line(reset, ecbdata, line + 1, len - 1);
1270 }
1271 }
1272
1273 static char *pprint_rename(const char *a, const char *b)
1274 {
1275 const char *old = a;
1276 const char *new = b;
1277 struct strbuf name = STRBUF_INIT;
1278 int pfx_length, sfx_length;
1279 int pfx_adjust_for_slash;
1280 int len_a = strlen(a);
1281 int len_b = strlen(b);
1282 int a_midlen, b_midlen;
1283 int qlen_a = quote_c_style(a, NULL, NULL, 0);
1284 int qlen_b = quote_c_style(b, NULL, NULL, 0);
1285
1286 if (qlen_a || qlen_b) {
1287 quote_c_style(a, &name, NULL, 0);
1288 strbuf_addstr(&name, " => ");
1289 quote_c_style(b, &name, NULL, 0);
1290 return strbuf_detach(&name, NULL);
1291 }
1292
1293 /* Find common prefix */
1294 pfx_length = 0;
1295 while (*old && *new && *old == *new) {
1296 if (*old == '/')
1297 pfx_length = old - a + 1;
1298 old++;
1299 new++;
1300 }
1301
1302 /* Find common suffix */
1303 old = a + len_a;
1304 new = b + len_b;
1305 sfx_length = 0;
1306 /*
1307 * If there is a common prefix, it must end in a slash. In
1308 * that case we let this loop run 1 into the prefix to see the
1309 * same slash.
1310 *
1311 * If there is no common prefix, we cannot do this as it would
1312 * underrun the input strings.
1313 */
1314 pfx_adjust_for_slash = (pfx_length ? 1 : 0);
1315 while (a + pfx_length - pfx_adjust_for_slash <= old &&
1316 b + pfx_length - pfx_adjust_for_slash <= new &&
1317 *old == *new) {
1318 if (*old == '/')
1319 sfx_length = len_a - (old - a);
1320 old--;
1321 new--;
1322 }
1323
1324 /*
1325 * pfx{mid-a => mid-b}sfx
1326 * {pfx-a => pfx-b}sfx
1327 * pfx{sfx-a => sfx-b}
1328 * name-a => name-b
1329 */
1330 a_midlen = len_a - pfx_length - sfx_length;
1331 b_midlen = len_b - pfx_length - sfx_length;
1332 if (a_midlen < 0)
1333 a_midlen = 0;
1334 if (b_midlen < 0)
1335 b_midlen = 0;
1336
1337 strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1338 if (pfx_length + sfx_length) {
1339 strbuf_add(&name, a, pfx_length);
1340 strbuf_addch(&name, '{');
1341 }
1342 strbuf_add(&name, a + pfx_length, a_midlen);
1343 strbuf_addstr(&name, " => ");
1344 strbuf_add(&name, b + pfx_length, b_midlen);
1345 if (pfx_length + sfx_length) {
1346 strbuf_addch(&name, '}');
1347 strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1348 }
1349 return strbuf_detach(&name, NULL);
1350 }
1351
1352 struct diffstat_t {
1353 int nr;
1354 int alloc;
1355 struct diffstat_file {
1356 char *from_name;
1357 char *name;
1358 char *print_name;
1359 unsigned is_unmerged:1;
1360 unsigned is_binary:1;
1361 unsigned is_renamed:1;
1362 unsigned is_interesting:1;
1363 uintmax_t added, deleted;
1364 } **files;
1365 };
1366
1367 static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1368 const char *name_a,
1369 const char *name_b)
1370 {
1371 struct diffstat_file *x;
1372 x = xcalloc(1, sizeof(*x));
1373 ALLOC_GROW(diffstat->files, diffstat->nr + 1, diffstat->alloc);
1374 diffstat->files[diffstat->nr++] = x;
1375 if (name_b) {
1376 x->from_name = xstrdup(name_a);
1377 x->name = xstrdup(name_b);
1378 x->is_renamed = 1;
1379 }
1380 else {
1381 x->from_name = NULL;
1382 x->name = xstrdup(name_a);
1383 }
1384 return x;
1385 }
1386
1387 static void diffstat_consume(void *priv, char *line, unsigned long len)
1388 {
1389 struct diffstat_t *diffstat = priv;
1390 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1391
1392 if (line[0] == '+')
1393 x->added++;
1394 else if (line[0] == '-')
1395 x->deleted++;
1396 }
1397
1398 const char mime_boundary_leader[] = "------------";
1399
1400 static int scale_linear(int it, int width, int max_change)
1401 {
1402 if (!it)
1403 return 0;
1404 /*
1405 * make sure that at least one '-' or '+' is printed if
1406 * there is any change to this path. The easiest way is to
1407 * scale linearly as if the alloted width is one column shorter
1408 * than it is, and then add 1 to the result.
1409 */
1410 return 1 + (it * (width - 1) / max_change);
1411 }
1412
1413 static void show_name(FILE *file,
1414 const char *prefix, const char *name, int len)
1415 {
1416 fprintf(file, " %s%-*s |", prefix, len, name);
1417 }
1418
1419 static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1420 {
1421 if (cnt <= 0)
1422 return;
1423 fprintf(file, "%s", set);
1424 while (cnt--)
1425 putc(ch, file);
1426 fprintf(file, "%s", reset);
1427 }
1428
1429 static void fill_print_name(struct diffstat_file *file)
1430 {
1431 char *pname;
1432
1433 if (file->print_name)
1434 return;
1435
1436 if (!file->is_renamed) {
1437 struct strbuf buf = STRBUF_INIT;
1438 if (quote_c_style(file->name, &buf, NULL, 0)) {
1439 pname = strbuf_detach(&buf, NULL);
1440 } else {
1441 pname = file->name;
1442 strbuf_release(&buf);
1443 }
1444 } else {
1445 pname = pprint_rename(file->from_name, file->name);
1446 }
1447 file->print_name = pname;
1448 }
1449
1450 int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1451 {
1452 struct strbuf sb = STRBUF_INIT;
1453 int ret;
1454
1455 if (!files) {
1456 assert(insertions == 0 && deletions == 0);
1457 return fprintf(fp, "%s\n", " 0 files changed");
1458 }
1459
1460 strbuf_addf(&sb,
1461 (files == 1) ? " %d file changed" : " %d files changed",
1462 files);
1463
1464 /*
1465 * For binary diff, the caller may want to print "x files
1466 * changed" with insertions == 0 && deletions == 0.
1467 *
1468 * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1469 * is probably less confusing (i.e skip over "2 files changed
1470 * but nothing about added/removed lines? Is this a bug in Git?").
1471 */
1472 if (insertions || deletions == 0) {
1473 strbuf_addf(&sb,
1474 (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
1475 insertions);
1476 }
1477
1478 if (deletions || insertions == 0) {
1479 strbuf_addf(&sb,
1480 (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
1481 deletions);
1482 }
1483 strbuf_addch(&sb, '\n');
1484 ret = fputs(sb.buf, fp);
1485 strbuf_release(&sb);
1486 return ret;
1487 }
1488
1489 static void show_stats(struct diffstat_t *data, struct diff_options *options)
1490 {
1491 int i, len, add, del, adds = 0, dels = 0;
1492 uintmax_t max_change = 0, max_len = 0;
1493 int total_files = data->nr, count;
1494 int width, name_width, graph_width, number_width = 0, bin_width = 0;
1495 const char *reset, *add_c, *del_c;
1496 const char *line_prefix = "";
1497 int extra_shown = 0;
1498
1499 if (data->nr == 0)
1500 return;
1501
1502 line_prefix = diff_line_prefix(options);
1503 count = options->stat_count ? options->stat_count : data->nr;
1504
1505 reset = diff_get_color_opt(options, DIFF_RESET);
1506 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1507 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1508
1509 /*
1510 * Find the longest filename and max number of changes
1511 */
1512 for (i = 0; (i < count) && (i < data->nr); i++) {
1513 struct diffstat_file *file = data->files[i];
1514 uintmax_t change = file->added + file->deleted;
1515
1516 if (!file->is_interesting && (change == 0)) {
1517 count++; /* not shown == room for one more */
1518 continue;
1519 }
1520 fill_print_name(file);
1521 len = strlen(file->print_name);
1522 if (max_len < len)
1523 max_len = len;
1524
1525 if (file->is_unmerged) {
1526 /* "Unmerged" is 8 characters */
1527 bin_width = bin_width < 8 ? 8 : bin_width;
1528 continue;
1529 }
1530 if (file->is_binary) {
1531 /* "Bin XXX -> YYY bytes" */
1532 int w = 14 + decimal_width(file->added)
1533 + decimal_width(file->deleted);
1534 bin_width = bin_width < w ? w : bin_width;
1535 /* Display change counts aligned with "Bin" */
1536 number_width = 3;
1537 continue;
1538 }
1539
1540 if (max_change < change)
1541 max_change = change;
1542 }
1543 count = i; /* where we can stop scanning in data->files[] */
1544
1545 /*
1546 * We have width = stat_width or term_columns() columns total.
1547 * We want a maximum of min(max_len, stat_name_width) for the name part.
1548 * We want a maximum of min(max_change, stat_graph_width) for the +- part.
1549 * We also need 1 for " " and 4 + decimal_width(max_change)
1550 * for " | NNNN " and one the empty column at the end, altogether
1551 * 6 + decimal_width(max_change).
1552 *
1553 * If there's not enough space, we will use the smaller of
1554 * stat_name_width (if set) and 5/8*width for the filename,
1555 * and the rest for constant elements + graph part, but no more
1556 * than stat_graph_width for the graph part.
1557 * (5/8 gives 50 for filename and 30 for the constant parts + graph
1558 * for the standard terminal size).
1559 *
1560 * In other words: stat_width limits the maximum width, and
1561 * stat_name_width fixes the maximum width of the filename,
1562 * and is also used to divide available columns if there
1563 * aren't enough.
1564 *
1565 * Binary files are displayed with "Bin XXX -> YYY bytes"
1566 * instead of the change count and graph. This part is treated
1567 * similarly to the graph part, except that it is not
1568 * "scaled". If total width is too small to accommodate the
1569 * guaranteed minimum width of the filename part and the
1570 * separators and this message, this message will "overflow"
1571 * making the line longer than the maximum width.
1572 */
1573
1574 if (options->stat_width == -1)
1575 width = term_columns() - options->output_prefix_length;
1576 else
1577 width = options->stat_width ? options->stat_width : 80;
1578 number_width = decimal_width(max_change) > number_width ?
1579 decimal_width(max_change) : number_width;
1580
1581 if (options->stat_graph_width == -1)
1582 options->stat_graph_width = diff_stat_graph_width;
1583
1584 /*
1585 * Guarantee 3/8*16==6 for the graph part
1586 * and 5/8*16==10 for the filename part
1587 */
1588 if (width < 16 + 6 + number_width)
1589 width = 16 + 6 + number_width;
1590
1591 /*
1592 * First assign sizes that are wanted, ignoring available width.
1593 * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
1594 * starting from "XXX" should fit in graph_width.
1595 */
1596 graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
1597 if (options->stat_graph_width &&
1598 options->stat_graph_width < graph_width)
1599 graph_width = options->stat_graph_width;
1600
1601 name_width = (options->stat_name_width > 0 &&
1602 options->stat_name_width < max_len) ?
1603 options->stat_name_width : max_len;
1604
1605 /*
1606 * Adjust adjustable widths not to exceed maximum width
1607 */
1608 if (name_width + number_width + 6 + graph_width > width) {
1609 if (graph_width > width * 3/8 - number_width - 6) {
1610 graph_width = width * 3/8 - number_width - 6;
1611 if (graph_width < 6)
1612 graph_width = 6;
1613 }
1614
1615 if (options->stat_graph_width &&
1616 graph_width > options->stat_graph_width)
1617 graph_width = options->stat_graph_width;
1618 if (name_width > width - number_width - 6 - graph_width)
1619 name_width = width - number_width - 6 - graph_width;
1620 else
1621 graph_width = width - number_width - 6 - name_width;
1622 }
1623
1624 /*
1625 * From here name_width is the width of the name area,
1626 * and graph_width is the width of the graph area.
1627 * max_change is used to scale graph properly.
1628 */
1629 for (i = 0; i < count; i++) {
1630 const char *prefix = "";
1631 struct diffstat_file *file = data->files[i];
1632 char *name = file->print_name;
1633 uintmax_t added = file->added;
1634 uintmax_t deleted = file->deleted;
1635 int name_len;
1636
1637 if (!file->is_interesting && (added + deleted == 0))
1638 continue;
1639
1640 /*
1641 * "scale" the filename
1642 */
1643 len = name_width;
1644 name_len = strlen(name);
1645 if (name_width < name_len) {
1646 char *slash;
1647 prefix = "...";
1648 len -= 3;
1649 name += name_len - len;
1650 slash = strchr(name, '/');
1651 if (slash)
1652 name = slash;
1653 }
1654
1655 if (file->is_binary) {
1656 fprintf(options->file, "%s", line_prefix);
1657 show_name(options->file, prefix, name, len);
1658 fprintf(options->file, " %*s", number_width, "Bin");
1659 if (!added && !deleted) {
1660 putc('\n', options->file);
1661 continue;
1662 }
1663 fprintf(options->file, " %s%"PRIuMAX"%s",
1664 del_c, deleted, reset);
1665 fprintf(options->file, " -> ");
1666 fprintf(options->file, "%s%"PRIuMAX"%s",
1667 add_c, added, reset);
1668 fprintf(options->file, " bytes");
1669 fprintf(options->file, "\n");
1670 continue;
1671 }
1672 else if (file->is_unmerged) {
1673 fprintf(options->file, "%s", line_prefix);
1674 show_name(options->file, prefix, name, len);
1675 fprintf(options->file, " Unmerged\n");
1676 continue;
1677 }
1678
1679 /*
1680 * scale the add/delete
1681 */
1682 add = added;
1683 del = deleted;
1684
1685 if (graph_width <= max_change) {
1686 int total = scale_linear(add + del, graph_width, max_change);
1687 if (total < 2 && add && del)
1688 /* width >= 2 due to the sanity check */
1689 total = 2;
1690 if (add < del) {
1691 add = scale_linear(add, graph_width, max_change);
1692 del = total - add;
1693 } else {
1694 del = scale_linear(del, graph_width, max_change);
1695 add = total - del;
1696 }
1697 }
1698 fprintf(options->file, "%s", line_prefix);
1699 show_name(options->file, prefix, name, len);
1700 fprintf(options->file, " %*"PRIuMAX"%s",
1701 number_width, added + deleted,
1702 added + deleted ? " " : "");
1703 show_graph(options->file, '+', add, add_c, reset);
1704 show_graph(options->file, '-', del, del_c, reset);
1705 fprintf(options->file, "\n");
1706 }
1707
1708 for (i = 0; i < data->nr; i++) {
1709 struct diffstat_file *file = data->files[i];
1710 uintmax_t added = file->added;
1711 uintmax_t deleted = file->deleted;
1712
1713 if (file->is_unmerged ||
1714 (!file->is_interesting && (added + deleted == 0))) {
1715 total_files--;
1716 continue;
1717 }
1718
1719 if (!file->is_binary) {
1720 adds += added;
1721 dels += deleted;
1722 }
1723 if (i < count)
1724 continue;
1725 if (!extra_shown)
1726 fprintf(options->file, "%s ...\n", line_prefix);
1727 extra_shown = 1;
1728 }
1729 fprintf(options->file, "%s", line_prefix);
1730 print_stat_summary(options->file, total_files, adds, dels);
1731 }
1732
1733 static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1734 {
1735 int i, adds = 0, dels = 0, total_files = data->nr;
1736
1737 if (data->nr == 0)
1738 return;
1739
1740 for (i = 0; i < data->nr; i++) {
1741 int added = data->files[i]->added;
1742 int deleted= data->files[i]->deleted;
1743
1744 if (data->files[i]->is_unmerged ||
1745 (!data->files[i]->is_interesting && (added + deleted == 0))) {
1746 total_files--;
1747 } else if (!data->files[i]->is_binary) { /* don't count bytes */
1748 adds += added;
1749 dels += deleted;
1750 }
1751 }
1752 fprintf(options->file, "%s", diff_line_prefix(options));
1753 print_stat_summary(options->file, total_files, adds, dels);
1754 }
1755
1756 static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1757 {
1758 int i;
1759
1760 if (data->nr == 0)
1761 return;
1762
1763 for (i = 0; i < data->nr; i++) {
1764 struct diffstat_file *file = data->files[i];
1765
1766 fprintf(options->file, "%s", diff_line_prefix(options));
1767
1768 if (file->is_binary)
1769 fprintf(options->file, "-\t-\t");
1770 else
1771 fprintf(options->file,
1772 "%"PRIuMAX"\t%"PRIuMAX"\t",
1773 file->added, file->deleted);
1774 if (options->line_termination) {
1775 fill_print_name(file);
1776 if (!file->is_renamed)
1777 write_name_quoted(file->name, options->file,
1778 options->line_termination);
1779 else {
1780 fputs(file->print_name, options->file);
1781 putc(options->line_termination, options->file);
1782 }
1783 } else {
1784 if (file->is_renamed) {
1785 putc('\0', options->file);
1786 write_name_quoted(file->from_name, options->file, '\0');
1787 }
1788 write_name_quoted(file->name, options->file, '\0');
1789 }
1790 }
1791 }
1792
1793 struct dirstat_file {
1794 const char *name;
1795 unsigned long changed;
1796 };
1797
1798 struct dirstat_dir {
1799 struct dirstat_file *files;
1800 int alloc, nr, permille, cumulative;
1801 };
1802
1803 static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1804 unsigned long changed, const char *base, int baselen)
1805 {
1806 unsigned long this_dir = 0;
1807 unsigned int sources = 0;
1808 const char *line_prefix = diff_line_prefix(opt);
1809
1810 while (dir->nr) {
1811 struct dirstat_file *f = dir->files;
1812 int namelen = strlen(f->name);
1813 unsigned long this;
1814 char *slash;
1815
1816 if (namelen < baselen)
1817 break;
1818 if (memcmp(f->name, base, baselen))
1819 break;
1820 slash = strchr(f->name + baselen, '/');
1821 if (slash) {
1822 int newbaselen = slash + 1 - f->name;
1823 this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
1824 sources++;
1825 } else {
1826 this = f->changed;
1827 dir->files++;
1828 dir->nr--;
1829 sources += 2;
1830 }
1831 this_dir += this;
1832 }
1833
1834 /*
1835 * We don't report dirstat's for
1836 * - the top level
1837 * - or cases where everything came from a single directory
1838 * under this directory (sources == 1).
1839 */
1840 if (baselen && sources != 1) {
1841 if (this_dir) {
1842 int permille = this_dir * 1000 / changed;
1843 if (permille >= dir->permille) {
1844 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
1845 permille / 10, permille % 10, baselen, base);
1846 if (!dir->cumulative)
1847 return 0;
1848 }
1849 }
1850 }
1851 return this_dir;
1852 }
1853
1854 static int dirstat_compare(const void *_a, const void *_b)
1855 {
1856 const struct dirstat_file *a = _a;
1857 const struct dirstat_file *b = _b;
1858 return strcmp(a->name, b->name);
1859 }
1860
1861 static void show_dirstat(struct diff_options *options)
1862 {
1863 int i;
1864 unsigned long changed;
1865 struct dirstat_dir dir;
1866 struct diff_queue_struct *q = &diff_queued_diff;
1867
1868 dir.files = NULL;
1869 dir.alloc = 0;
1870 dir.nr = 0;
1871 dir.permille = options->dirstat_permille;
1872 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1873
1874 changed = 0;
1875 for (i = 0; i < q->nr; i++) {
1876 struct diff_filepair *p = q->queue[i];
1877 const char *name;
1878 unsigned long copied, added, damage;
1879 int content_changed;
1880
1881 name = p->two->path ? p->two->path : p->one->path;
1882
1883 if (p->one->sha1_valid && p->two->sha1_valid)
1884 content_changed = hashcmp(p->one->sha1, p->two->sha1);
1885 else
1886 content_changed = 1;
1887
1888 if (!content_changed) {
1889 /*
1890 * The SHA1 has not changed, so pre-/post-content is
1891 * identical. We can therefore skip looking at the
1892 * file contents altogether.
1893 */
1894 damage = 0;
1895 goto found_damage;
1896 }
1897
1898 if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
1899 /*
1900 * In --dirstat-by-file mode, we don't really need to
1901 * look at the actual file contents at all.
1902 * The fact that the SHA1 changed is enough for us to
1903 * add this file to the list of results
1904 * (with each file contributing equal damage).
1905 */
1906 damage = 1;
1907 goto found_damage;
1908 }
1909
1910 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
1911 diff_populate_filespec(p->one, 0);
1912 diff_populate_filespec(p->two, 0);
1913 diffcore_count_changes(p->one, p->two, NULL, NULL, 0,
1914 &copied, &added);
1915 diff_free_filespec_data(p->one);
1916 diff_free_filespec_data(p->two);
1917 } else if (DIFF_FILE_VALID(p->one)) {
1918 diff_populate_filespec(p->one, CHECK_SIZE_ONLY);
1919 copied = added = 0;
1920 diff_free_filespec_data(p->one);
1921 } else if (DIFF_FILE_VALID(p->two)) {
1922 diff_populate_filespec(p->two, CHECK_SIZE_ONLY);
1923 copied = 0;
1924 added = p->two->size;
1925 diff_free_filespec_data(p->two);
1926 } else
1927 continue;
1928
1929 /*
1930 * Original minus copied is the removed material,
1931 * added is the new material. They are both damages
1932 * made to the preimage.
1933 * If the resulting damage is zero, we know that
1934 * diffcore_count_changes() considers the two entries to
1935 * be identical, but since content_changed is true, we
1936 * know that there must have been _some_ kind of change,
1937 * so we force all entries to have damage > 0.
1938 */
1939 damage = (p->one->size - copied) + added;
1940 if (!damage)
1941 damage = 1;
1942
1943 found_damage:
1944 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1945 dir.files[dir.nr].name = name;
1946 dir.files[dir.nr].changed = damage;
1947 changed += damage;
1948 dir.nr++;
1949 }
1950
1951 /* This can happen even with many files, if everything was renames */
1952 if (!changed)
1953 return;
1954
1955 /* Show all directories with more than x% of the changes */
1956 qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1957 gather_dirstat(options, &dir, changed, "", 0);
1958 }
1959
1960 static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
1961 {
1962 int i;
1963 unsigned long changed;
1964 struct dirstat_dir dir;
1965
1966 if (data->nr == 0)
1967 return;
1968
1969 dir.files = NULL;
1970 dir.alloc = 0;
1971 dir.nr = 0;
1972 dir.permille = options->dirstat_permille;
1973 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1974
1975 changed = 0;
1976 for (i = 0; i < data->nr; i++) {
1977 struct diffstat_file *file = data->files[i];
1978 unsigned long damage = file->added + file->deleted;
1979 if (file->is_binary)
1980 /*
1981 * binary files counts bytes, not lines. Must find some
1982 * way to normalize binary bytes vs. textual lines.
1983 * The following heuristic assumes that there are 64
1984 * bytes per "line".
1985 * This is stupid and ugly, but very cheap...
1986 */
1987 damage = (damage + 63) / 64;
1988 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1989 dir.files[dir.nr].name = file->name;
1990 dir.files[dir.nr].changed = damage;
1991 changed += damage;
1992 dir.nr++;
1993 }
1994
1995 /* This can happen even with many files, if everything was renames */
1996 if (!changed)
1997 return;
1998
1999 /* Show all directories with more than x% of the changes */
2000 qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
2001 gather_dirstat(options, &dir, changed, "", 0);
2002 }
2003
2004 static void free_diffstat_info(struct diffstat_t *diffstat)
2005 {
2006 int i;
2007 for (i = 0; i < diffstat->nr; i++) {
2008 struct diffstat_file *f = diffstat->files[i];
2009 if (f->name != f->print_name)
2010 free(f->print_name);
2011 free(f->name);
2012 free(f->from_name);
2013 free(f);
2014 }
2015 free(diffstat->files);
2016 }
2017
2018 struct checkdiff_t {
2019 const char *filename;
2020 int lineno;
2021 int conflict_marker_size;
2022 struct diff_options *o;
2023 unsigned ws_rule;
2024 unsigned status;
2025 };
2026
2027 static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2028 {
2029 char firstchar;
2030 int cnt;
2031
2032 if (len < marker_size + 1)
2033 return 0;
2034 firstchar = line[0];
2035 switch (firstchar) {
2036 case '=': case '>': case '<': case '|':
2037 break;
2038 default:
2039 return 0;
2040 }
2041 for (cnt = 1; cnt < marker_size; cnt++)
2042 if (line[cnt] != firstchar)
2043 return 0;
2044 /* line[1] thru line[marker_size-1] are same as firstchar */
2045 if (len < marker_size + 1 || !isspace(line[marker_size]))
2046 return 0;
2047 return 1;
2048 }
2049
2050 static void checkdiff_consume(void *priv, char *line, unsigned long len)
2051 {
2052 struct checkdiff_t *data = priv;
2053 int marker_size = data->conflict_marker_size;
2054 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2055 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2056 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2057 char *err;
2058 const char *line_prefix;
2059
2060 assert(data->o);
2061 line_prefix = diff_line_prefix(data->o);
2062
2063 if (line[0] == '+') {
2064 unsigned bad;
2065 data->lineno++;
2066 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2067 data->status |= 1;
2068 fprintf(data->o->file,
2069 "%s%s:%d: leftover conflict marker\n",
2070 line_prefix, data->filename, data->lineno);
2071 }
2072 bad = ws_check(line + 1, len - 1, data->ws_rule);
2073 if (!bad)
2074 return;
2075 data->status |= bad;
2076 err = whitespace_error_string(bad);
2077 fprintf(data->o->file, "%s%s:%d: %s.\n",
2078 line_prefix, data->filename, data->lineno, err);
2079 free(err);
2080 emit_line(data->o, set, reset, line, 1);
2081 ws_check_emit(line + 1, len - 1, data->ws_rule,
2082 data->o->file, set, reset, ws);
2083 } else if (line[0] == ' ') {
2084 data->lineno++;
2085 } else if (line[0] == '@') {
2086 char *plus = strchr(line, '+');
2087 if (plus)
2088 data->lineno = strtol(plus, NULL, 10) - 1;
2089 else
2090 die("invalid diff");
2091 }
2092 }
2093
2094 static unsigned char *deflate_it(char *data,
2095 unsigned long size,
2096 unsigned long *result_size)
2097 {
2098 int bound;
2099 unsigned char *deflated;
2100 git_zstream stream;
2101
2102 git_deflate_init(&stream, zlib_compression_level);
2103 bound = git_deflate_bound(&stream, size);
2104 deflated = xmalloc(bound);
2105 stream.next_out = deflated;
2106 stream.avail_out = bound;
2107
2108 stream.next_in = (unsigned char *)data;
2109 stream.avail_in = size;
2110 while (git_deflate(&stream, Z_FINISH) == Z_OK)
2111 ; /* nothing */
2112 git_deflate_end(&stream);
2113 *result_size = stream.total_out;
2114 return deflated;
2115 }
2116
2117 static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two,
2118 const char *prefix)
2119 {
2120 void *cp;
2121 void *delta;
2122 void *deflated;
2123 void *data;
2124 unsigned long orig_size;
2125 unsigned long delta_size;
2126 unsigned long deflate_size;
2127 unsigned long data_size;
2128
2129 /* We could do deflated delta, or we could do just deflated two,
2130 * whichever is smaller.
2131 */
2132 delta = NULL;
2133 deflated = deflate_it(two->ptr, two->size, &deflate_size);
2134 if (one->size && two->size) {
2135 delta = diff_delta(one->ptr, one->size,
2136 two->ptr, two->size,
2137 &delta_size, deflate_size);
2138 if (delta) {
2139 void *to_free = delta;
2140 orig_size = delta_size;
2141 delta = deflate_it(delta, delta_size, &delta_size);
2142 free(to_free);
2143 }
2144 }
2145
2146 if (delta && delta_size < deflate_size) {
2147 fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2148 free(deflated);
2149 data = delta;
2150 data_size = delta_size;
2151 }
2152 else {
2153 fprintf(file, "%sliteral %lu\n", prefix, two->size);
2154 free(delta);
2155 data = deflated;
2156 data_size = deflate_size;
2157 }
2158
2159 /* emit data encoded in base85 */
2160 cp = data;
2161 while (data_size) {
2162 int bytes = (52 < data_size) ? 52 : data_size;
2163 char line[70];
2164 data_size -= bytes;
2165 if (bytes <= 26)
2166 line[0] = bytes + 'A' - 1;
2167 else
2168 line[0] = bytes - 26 + 'a' - 1;
2169 encode_85(line + 1, cp, bytes);
2170 cp = (char *) cp + bytes;
2171 fprintf(file, "%s", prefix);
2172 fputs(line, file);
2173 fputc('\n', file);
2174 }
2175 fprintf(file, "%s\n", prefix);
2176 free(data);
2177 }
2178
2179 static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two,
2180 const char *prefix)
2181 {
2182 fprintf(file, "%sGIT binary patch\n", prefix);
2183 emit_binary_diff_body(file, one, two, prefix);
2184 emit_binary_diff_body(file, two, one, prefix);
2185 }
2186
2187 int diff_filespec_is_binary(struct diff_filespec *one)
2188 {
2189 if (one->is_binary == -1) {
2190 diff_filespec_load_driver(one);
2191 if (one->driver->binary != -1)
2192 one->is_binary = one->driver->binary;
2193 else {
2194 if (!one->data && DIFF_FILE_VALID(one))
2195 diff_populate_filespec(one, CHECK_BINARY);
2196 if (one->is_binary == -1 && one->data)
2197 one->is_binary = buffer_is_binary(one->data,
2198 one->size);
2199 if (one->is_binary == -1)
2200 one->is_binary = 0;
2201 }
2202 }
2203 return one->is_binary;
2204 }
2205
2206 static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2207 {
2208 diff_filespec_load_driver(one);
2209 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2210 }
2211
2212 void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2213 {
2214 if (!options->a_prefix)
2215 options->a_prefix = a;
2216 if (!options->b_prefix)
2217 options->b_prefix = b;
2218 }
2219
2220 struct userdiff_driver *get_textconv(struct diff_filespec *one)
2221 {
2222 if (!DIFF_FILE_VALID(one))
2223 return NULL;
2224
2225 diff_filespec_load_driver(one);
2226 return userdiff_get_textconv(one->driver);
2227 }
2228
2229 static void builtin_diff(const char *name_a,
2230 const char *name_b,
2231 struct diff_filespec *one,
2232 struct diff_filespec *two,
2233 const char *xfrm_msg,
2234 int must_show_header,
2235 struct diff_options *o,
2236 int complete_rewrite)
2237 {
2238 mmfile_t mf1, mf2;
2239 const char *lbl[2];
2240 char *a_one, *b_two;
2241 const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
2242 const char *reset = diff_get_color_opt(o, DIFF_RESET);
2243 const char *a_prefix, *b_prefix;
2244 struct userdiff_driver *textconv_one = NULL;
2245 struct userdiff_driver *textconv_two = NULL;
2246 struct strbuf header = STRBUF_INIT;
2247 const char *line_prefix = diff_line_prefix(o);
2248
2249 if (DIFF_OPT_TST(o, SUBMODULE_LOG) &&
2250 (!one->mode || S_ISGITLINK(one->mode)) &&
2251 (!two->mode || S_ISGITLINK(two->mode))) {
2252 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2253 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2254 show_submodule_summary(o->file, one->path ? one->path : two->path,
2255 line_prefix,
2256 one->sha1, two->sha1, two->dirty_submodule,
2257 meta, del, add, reset);
2258 return;
2259 }
2260
2261 if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2262 textconv_one = get_textconv(one);
2263 textconv_two = get_textconv(two);
2264 }
2265
2266 diff_set_mnemonic_prefix(o, "a/", "b/");
2267 if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2268 a_prefix = o->b_prefix;
2269 b_prefix = o->a_prefix;
2270 } else {
2271 a_prefix = o->a_prefix;
2272 b_prefix = o->b_prefix;
2273 }
2274
2275 /* Never use a non-valid filename anywhere if at all possible */
2276 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2277 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2278
2279 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2280 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2281 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2282 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2283 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
2284 if (lbl[0][0] == '/') {
2285 /* /dev/null */
2286 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
2287 if (xfrm_msg)
2288 strbuf_addstr(&header, xfrm_msg);
2289 must_show_header = 1;
2290 }
2291 else if (lbl[1][0] == '/') {
2292 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
2293 if (xfrm_msg)
2294 strbuf_addstr(&header, xfrm_msg);
2295 must_show_header = 1;
2296 }
2297 else {
2298 if (one->mode != two->mode) {
2299 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
2300 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
2301 must_show_header = 1;
2302 }
2303 if (xfrm_msg)
2304 strbuf_addstr(&header, xfrm_msg);
2305
2306 /*
2307 * we do not run diff between different kind
2308 * of objects.
2309 */
2310 if ((one->mode ^ two->mode) & S_IFMT)
2311 goto free_ab_and_return;
2312 if (complete_rewrite &&
2313 (textconv_one || !diff_filespec_is_binary(one)) &&
2314 (textconv_two || !diff_filespec_is_binary(two))) {
2315 fprintf(o->file, "%s", header.buf);
2316 strbuf_reset(&header);
2317 emit_rewrite_diff(name_a, name_b, one, two,
2318 textconv_one, textconv_two, o);
2319 o->found_changes = 1;
2320 goto free_ab_and_return;
2321 }
2322 }
2323
2324 if (o->irreversible_delete && lbl[1][0] == '/') {
2325 fprintf(o->file, "%s", header.buf);
2326 strbuf_reset(&header);
2327 goto free_ab_and_return;
2328 } else if (!DIFF_OPT_TST(o, TEXT) &&
2329 ( (!textconv_one && diff_filespec_is_binary(one)) ||
2330 (!textconv_two && diff_filespec_is_binary(two)) )) {
2331 if (!one->data && !two->data &&
2332 S_ISREG(one->mode) && S_ISREG(two->mode) &&
2333 !DIFF_OPT_TST(o, BINARY)) {
2334 if (!hashcmp(one->sha1, two->sha1)) {
2335 if (must_show_header)
2336 fprintf(o->file, "%s", header.buf);
2337 goto free_ab_and_return;
2338 }
2339 fprintf(o->file, "%s", header.buf);
2340 fprintf(o->file, "%sBinary files %s and %s differ\n",
2341 line_prefix, lbl[0], lbl[1]);
2342 goto free_ab_and_return;
2343 }
2344 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2345 die("unable to read files to diff");
2346 /* Quite common confusing case */
2347 if (mf1.size == mf2.size &&
2348 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2349 if (must_show_header)
2350 fprintf(o->file, "%s", header.buf);
2351 goto free_ab_and_return;
2352 }
2353 fprintf(o->file, "%s", header.buf);
2354 strbuf_reset(&header);
2355 if (DIFF_OPT_TST(o, BINARY))
2356 emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2357 else
2358 fprintf(o->file, "%sBinary files %s and %s differ\n",
2359 line_prefix, lbl[0], lbl[1]);
2360 o->found_changes = 1;
2361 } else {
2362 /* Crazy xdl interfaces.. */
2363 const char *diffopts = getenv("GIT_DIFF_OPTS");
2364 const char *v;
2365 xpparam_t xpp;
2366 xdemitconf_t xecfg;
2367 struct emit_callback ecbdata;
2368 const struct userdiff_funcname *pe;
2369
2370 if (must_show_header) {
2371 fprintf(o->file, "%s", header.buf);
2372 strbuf_reset(&header);
2373 }
2374
2375 mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2376 mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2377
2378 pe = diff_funcname_pattern(one);
2379 if (!pe)
2380 pe = diff_funcname_pattern(two);
2381
2382 memset(&xpp, 0, sizeof(xpp));
2383 memset(&xecfg, 0, sizeof(xecfg));
2384 memset(&ecbdata, 0, sizeof(ecbdata));
2385 ecbdata.label_path = lbl;
2386 ecbdata.color_diff = want_color(o->use_color);
2387 ecbdata.found_changesp = &o->found_changes;
2388 ecbdata.ws_rule = whitespace_rule(name_b);
2389 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2390 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2391 ecbdata.opt = o;
2392 ecbdata.header = header.len ? &header : NULL;
2393 xpp.flags = o->xdl_opts;
2394 xecfg.ctxlen = o->context;
2395 xecfg.interhunkctxlen = o->interhunkcontext;
2396 xecfg.flags = XDL_EMIT_FUNCNAMES;
2397 if (DIFF_OPT_TST(o, FUNCCONTEXT))
2398 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2399 if (pe)
2400 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2401 if (!diffopts)
2402 ;
2403 else if (skip_prefix(diffopts, "--unified=", &v))
2404 xecfg.ctxlen = strtoul(v, NULL, 10);
2405 else if (skip_prefix(diffopts, "-u", &v))
2406 xecfg.ctxlen = strtoul(v, NULL, 10);
2407 if (o->word_diff)
2408 init_diff_words_data(&ecbdata, o, one, two);
2409 xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2410 &xpp, &xecfg);
2411 if (o->word_diff)
2412 free_diff_words_data(&ecbdata);
2413 if (textconv_one)
2414 free(mf1.ptr);
2415 if (textconv_two)
2416 free(mf2.ptr);
2417 xdiff_clear_find_func(&xecfg);
2418 }
2419
2420 free_ab_and_return:
2421 strbuf_release(&header);
2422 diff_free_filespec_data(one);
2423 diff_free_filespec_data(two);
2424 free(a_one);
2425 free(b_two);
2426 return;
2427 }
2428
2429 static void builtin_diffstat(const char *name_a, const char *name_b,
2430 struct diff_filespec *one,
2431 struct diff_filespec *two,
2432 struct diffstat_t *diffstat,
2433 struct diff_options *o,
2434 struct diff_filepair *p)
2435 {
2436 mmfile_t mf1, mf2;
2437 struct diffstat_file *data;
2438 int same_contents;
2439 int complete_rewrite = 0;
2440
2441 if (!DIFF_PAIR_UNMERGED(p)) {
2442 if (p->status == DIFF_STATUS_MODIFIED && p->score)
2443 complete_rewrite = 1;
2444 }
2445
2446 data = diffstat_add(diffstat, name_a, name_b);
2447 data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
2448
2449 if (!one || !two) {
2450 data->is_unmerged = 1;
2451 return;
2452 }
2453
2454 same_contents = !hashcmp(one->sha1, two->sha1);
2455
2456 if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2457 data->is_binary = 1;
2458 if (same_contents) {
2459 data->added = 0;
2460 data->deleted = 0;
2461 } else {
2462 data->added = diff_filespec_size(two);
2463 data->deleted = diff_filespec_size(one);
2464 }
2465 }
2466
2467 else if (complete_rewrite) {
2468 diff_populate_filespec(one, 0);
2469 diff_populate_filespec(two, 0);
2470 data->deleted = count_lines(one->data, one->size);
2471 data->added = count_lines(two->data, two->size);
2472 }
2473
2474 else if (!same_contents) {
2475 /* Crazy xdl interfaces.. */
2476 xpparam_t xpp;
2477 xdemitconf_t xecfg;
2478
2479 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2480 die("unable to read files to diff");
2481
2482 memset(&xpp, 0, sizeof(xpp));
2483 memset(&xecfg, 0, sizeof(xecfg));
2484 xpp.flags = o->xdl_opts;
2485 xecfg.ctxlen = o->context;
2486 xecfg.interhunkctxlen = o->interhunkcontext;
2487 xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2488 &xpp, &xecfg);
2489 }
2490
2491 diff_free_filespec_data(one);
2492 diff_free_filespec_data(two);
2493 }
2494
2495 static void builtin_checkdiff(const char *name_a, const char *name_b,
2496 const char *attr_path,
2497 struct diff_filespec *one,
2498 struct diff_filespec *two,
2499 struct diff_options *o)
2500 {
2501 mmfile_t mf1, mf2;
2502 struct checkdiff_t data;
2503
2504 if (!two)
2505 return;
2506
2507 memset(&data, 0, sizeof(data));
2508 data.filename = name_b ? name_b : name_a;
2509 data.lineno = 0;
2510 data.o = o;
2511 data.ws_rule = whitespace_rule(attr_path);
2512 data.conflict_marker_size = ll_merge_marker_size(attr_path);
2513
2514 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2515 die("unable to read files to diff");
2516
2517 /*
2518 * All the other codepaths check both sides, but not checking
2519 * the "old" side here is deliberate. We are checking the newly
2520 * introduced changes, and as long as the "new" side is text, we
2521 * can and should check what it introduces.
2522 */
2523 if (diff_filespec_is_binary(two))
2524 goto free_and_return;
2525 else {
2526 /* Crazy xdl interfaces.. */
2527 xpparam_t xpp;
2528 xdemitconf_t xecfg;
2529
2530 memset(&xpp, 0, sizeof(xpp));
2531 memset(&xecfg, 0, sizeof(xecfg));
2532 xecfg.ctxlen = 1; /* at least one context line */
2533 xpp.flags = 0;
2534 xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2535 &xpp, &xecfg);
2536
2537 if (data.ws_rule & WS_BLANK_AT_EOF) {
2538 struct emit_callback ecbdata;
2539 int blank_at_eof;
2540
2541 ecbdata.ws_rule = data.ws_rule;
2542 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2543 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2544
2545 if (blank_at_eof) {
2546 static char *err;
2547 if (!err)
2548 err = whitespace_error_string(WS_BLANK_AT_EOF);
2549 fprintf(o->file, "%s:%d: %s.\n",
2550 data.filename, blank_at_eof, err);
2551 data.status = 1; /* report errors */
2552 }
2553 }
2554 }
2555 free_and_return:
2556 diff_free_filespec_data(one);
2557 diff_free_filespec_data(two);
2558 if (data.status)
2559 DIFF_OPT_SET(o, CHECK_FAILED);
2560 }
2561
2562 struct diff_filespec *alloc_filespec(const char *path)
2563 {
2564 int namelen = strlen(path);
2565 struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
2566
2567 memset(spec, 0, sizeof(*spec));
2568 spec->path = (char *)(spec + 1);
2569 memcpy(spec->path, path, namelen+1);
2570 spec->count = 1;
2571 spec->is_binary = -1;
2572 return spec;
2573 }
2574
2575 void free_filespec(struct diff_filespec *spec)
2576 {
2577 if (!--spec->count) {
2578 diff_free_filespec_data(spec);
2579 free(spec);
2580 }
2581 }
2582
2583 void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
2584 int sha1_valid, unsigned short mode)
2585 {
2586 if (mode) {
2587 spec->mode = canon_mode(mode);
2588 hashcpy(spec->sha1, sha1);
2589 spec->sha1_valid = sha1_valid;
2590 }
2591 }
2592
2593 /*
2594 * Given a name and sha1 pair, if the index tells us the file in
2595 * the work tree has that object contents, return true, so that
2596 * prepare_temp_file() does not have to inflate and extract.
2597 */
2598 static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
2599 {
2600 const struct cache_entry *ce;
2601 struct stat st;
2602 int pos, len;
2603
2604 /*
2605 * We do not read the cache ourselves here, because the
2606 * benchmark with my previous version that always reads cache
2607 * shows that it makes things worse for diff-tree comparing
2608 * two linux-2.6 kernel trees in an already checked out work
2609 * tree. This is because most diff-tree comparisons deal with
2610 * only a small number of files, while reading the cache is
2611 * expensive for a large project, and its cost outweighs the
2612 * savings we get by not inflating the object to a temporary
2613 * file. Practically, this code only helps when we are used
2614 * by diff-cache --cached, which does read the cache before
2615 * calling us.
2616 */
2617 if (!active_cache)
2618 return 0;
2619
2620 /* We want to avoid the working directory if our caller
2621 * doesn't need the data in a normal file, this system
2622 * is rather slow with its stat/open/mmap/close syscalls,
2623 * and the object is contained in a pack file. The pack
2624 * is probably already open and will be faster to obtain
2625 * the data through than the working directory. Loose
2626 * objects however would tend to be slower as they need
2627 * to be individually opened and inflated.
2628 */
2629 if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1))
2630 return 0;
2631
2632 len = strlen(name);
2633 pos = cache_name_pos(name, len);
2634 if (pos < 0)
2635 return 0;
2636 ce = active_cache[pos];
2637
2638 /*
2639 * This is not the sha1 we are looking for, or
2640 * unreusable because it is not a regular file.
2641 */
2642 if (hashcmp(sha1, ce->sha1) || !S_ISREG(ce->ce_mode))
2643 return 0;
2644
2645 /*
2646 * If ce is marked as "assume unchanged", there is no
2647 * guarantee that work tree matches what we are looking for.
2648 */
2649 if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2650 return 0;
2651
2652 /*
2653 * If ce matches the file in the work tree, we can reuse it.
2654 */
2655 if (ce_uptodate(ce) ||
2656 (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2657 return 1;
2658
2659 return 0;
2660 }
2661
2662 static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2663 {
2664 int len;
2665 char *data = xmalloc(100), *dirty = "";
2666
2667 /* Are we looking at the work tree? */
2668 if (s->dirty_submodule)
2669 dirty = "-dirty";
2670
2671 len = snprintf(data, 100,
2672 "Subproject commit %s%s\n", sha1_to_hex(s->sha1), dirty);
2673 s->data = data;
2674 s->size = len;
2675 s->should_free = 1;
2676 if (size_only) {
2677 s->data = NULL;
2678 free(data);
2679 }
2680 return 0;
2681 }
2682
2683 /*
2684 * While doing rename detection and pickaxe operation, we may need to
2685 * grab the data for the blob (or file) for our own in-core comparison.
2686 * diff_filespec has data and size fields for this purpose.
2687 */
2688 int diff_populate_filespec(struct diff_filespec *s, unsigned int flags)
2689 {
2690 int size_only = flags & CHECK_SIZE_ONLY;
2691 int err = 0;
2692 /*
2693 * demote FAIL to WARN to allow inspecting the situation
2694 * instead of refusing.
2695 */
2696 enum safe_crlf crlf_warn = (safe_crlf == SAFE_CRLF_FAIL
2697 ? SAFE_CRLF_WARN
2698 : safe_crlf);
2699
2700 if (!DIFF_FILE_VALID(s))
2701 die("internal error: asking to populate invalid file.");
2702 if (S_ISDIR(s->mode))
2703 return -1;
2704
2705 if (s->data)
2706 return 0;
2707
2708 if (size_only && 0 < s->size)
2709 return 0;
2710
2711 if (S_ISGITLINK(s->mode))
2712 return diff_populate_gitlink(s, size_only);
2713
2714 if (!s->sha1_valid ||
2715 reuse_worktree_file(s->path, s->sha1, 0)) {
2716 struct strbuf buf = STRBUF_INIT;
2717 struct stat st;
2718 int fd;
2719
2720 if (lstat(s->path, &st) < 0) {
2721 if (errno == ENOENT) {
2722 err_empty:
2723 err = -1;
2724 empty:
2725 s->data = (char *)"";
2726 s->size = 0;
2727 return err;
2728 }
2729 }
2730 s->size = xsize_t(st.st_size);
2731 if (!s->size)
2732 goto empty;
2733 if (S_ISLNK(st.st_mode)) {
2734 struct strbuf sb = STRBUF_INIT;
2735
2736 if (strbuf_readlink(&sb, s->path, s->size))
2737 goto err_empty;
2738 s->size = sb.len;
2739 s->data = strbuf_detach(&sb, NULL);
2740 s->should_free = 1;
2741 return 0;
2742 }
2743 if (size_only)
2744 return 0;
2745 if ((flags & CHECK_BINARY) &&
2746 s->size > big_file_threshold && s->is_binary == -1) {
2747 s->is_binary = 1;
2748 return 0;
2749 }
2750 fd = open(s->path, O_RDONLY);
2751 if (fd < 0)
2752 goto err_empty;
2753 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2754 close(fd);
2755 s->should_munmap = 1;
2756
2757 /*
2758 * Convert from working tree format to canonical git format
2759 */
2760 if (convert_to_git(s->path, s->data, s->size, &buf, crlf_warn)) {
2761 size_t size = 0;
2762 munmap(s->data, s->size);
2763 s->should_munmap = 0;
2764 s->data = strbuf_detach(&buf, &size);
2765 s->size = size;
2766 s->should_free = 1;
2767 }
2768 }
2769 else {
2770 enum object_type type;
2771 if (size_only || (flags & CHECK_BINARY)) {
2772 type = sha1_object_info(s->sha1, &s->size);
2773 if (type < 0)
2774 die("unable to read %s", sha1_to_hex(s->sha1));
2775 if (size_only)
2776 return 0;
2777 if (s->size > big_file_threshold && s->is_binary == -1) {
2778 s->is_binary = 1;
2779 return 0;
2780 }
2781 }
2782 s->data = read_sha1_file(s->sha1, &type, &s->size);
2783 if (!s->data)
2784 die("unable to read %s", sha1_to_hex(s->sha1));
2785 s->should_free = 1;
2786 }
2787 return 0;
2788 }
2789
2790 void diff_free_filespec_blob(struct diff_filespec *s)
2791 {
2792 if (s->should_free)
2793 free(s->data);
2794 else if (s->should_munmap)
2795 munmap(s->data, s->size);
2796
2797 if (s->should_free || s->should_munmap) {
2798 s->should_free = s->should_munmap = 0;
2799 s->data = NULL;
2800 }
2801 }
2802
2803 void diff_free_filespec_data(struct diff_filespec *s)
2804 {
2805 diff_free_filespec_blob(s);
2806 free(s->cnt_data);
2807 s->cnt_data = NULL;
2808 }
2809
2810 static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2811 void *blob,
2812 unsigned long size,
2813 const unsigned char *sha1,
2814 int mode)
2815 {
2816 int fd;
2817 struct strbuf buf = STRBUF_INIT;
2818 struct strbuf template = STRBUF_INIT;
2819 char *path_dup = xstrdup(path);
2820 const char *base = basename(path_dup);
2821
2822 /* Generate "XXXXXX_basename.ext" */
2823 strbuf_addstr(&template, "XXXXXX_");
2824 strbuf_addstr(&template, base);
2825
2826 fd = mks_tempfile_ts(&temp->tempfile, template.buf, strlen(base) + 1);
2827 if (fd < 0)
2828 die_errno("unable to create temp-file");
2829 if (convert_to_working_tree(path,
2830 (const char *)blob, (size_t)size, &buf)) {
2831 blob = buf.buf;
2832 size = buf.len;
2833 }
2834 if (write_in_full(fd, blob, size) != size)
2835 die_errno("unable to write temp-file");
2836 close_tempfile(&temp->tempfile);
2837 temp->name = get_tempfile_path(&temp->tempfile);
2838 strcpy(temp->hex, sha1_to_hex(sha1));
2839 temp->hex[40] = 0;
2840 sprintf(temp->mode, "%06o", mode);
2841 strbuf_release(&buf);
2842 strbuf_release(&template);
2843 free(path_dup);
2844 }
2845
2846 static struct diff_tempfile *prepare_temp_file(const char *name,
2847 struct diff_filespec *one)
2848 {
2849 struct diff_tempfile *temp = claim_diff_tempfile();
2850
2851 if (!DIFF_FILE_VALID(one)) {
2852 not_a_valid_file:
2853 /* A '-' entry produces this for file-2, and
2854 * a '+' entry produces this for file-1.
2855 */
2856 temp->name = "/dev/null";
2857 strcpy(temp->hex, ".");
2858 strcpy(temp->mode, ".");
2859 return temp;
2860 }
2861
2862 if (!S_ISGITLINK(one->mode) &&
2863 (!one->sha1_valid ||
2864 reuse_worktree_file(name, one->sha1, 1))) {
2865 struct stat st;
2866 if (lstat(name, &st) < 0) {
2867 if (errno == ENOENT)
2868 goto not_a_valid_file;
2869 die_errno("stat(%s)", name);
2870 }
2871 if (S_ISLNK(st.st_mode)) {
2872 struct strbuf sb = STRBUF_INIT;
2873 if (strbuf_readlink(&sb, name, st.st_size) < 0)
2874 die_errno("readlink(%s)", name);
2875 prep_temp_blob(name, temp, sb.buf, sb.len,
2876 (one->sha1_valid ?
2877 one->sha1 : null_sha1),
2878 (one->sha1_valid ?
2879 one->mode : S_IFLNK));
2880 strbuf_release(&sb);
2881 }
2882 else {
2883 /* we can borrow from the file in the work tree */
2884 temp->name = name;
2885 if (!one->sha1_valid)
2886 strcpy(temp->hex, sha1_to_hex(null_sha1));
2887 else
2888 strcpy(temp->hex, sha1_to_hex(one->sha1));
2889 /* Even though we may sometimes borrow the
2890 * contents from the work tree, we always want
2891 * one->mode. mode is trustworthy even when
2892 * !(one->sha1_valid), as long as
2893 * DIFF_FILE_VALID(one).
2894 */
2895 sprintf(temp->mode, "%06o", one->mode);
2896 }
2897 return temp;
2898 }
2899 else {
2900 if (diff_populate_filespec(one, 0))
2901 die("cannot read data blob for %s", one->path);
2902 prep_temp_blob(name, temp, one->data, one->size,
2903 one->sha1, one->mode);
2904 }
2905 return temp;
2906 }
2907
2908 static void add_external_diff_name(struct argv_array *argv,
2909 const char *name,
2910 struct diff_filespec *df)
2911 {
2912 struct diff_tempfile *temp = prepare_temp_file(name, df);
2913 argv_array_push(argv, temp->name);
2914 argv_array_push(argv, temp->hex);
2915 argv_array_push(argv, temp->mode);
2916 }
2917
2918 /* An external diff command takes:
2919 *
2920 * diff-cmd name infile1 infile1-sha1 infile1-mode \
2921 * infile2 infile2-sha1 infile2-mode [ rename-to ]
2922 *
2923 */
2924 static void run_external_diff(const char *pgm,
2925 const char *name,
2926 const char *other,
2927 struct diff_filespec *one,
2928 struct diff_filespec *two,
2929 const char *xfrm_msg,
2930 int complete_rewrite,
2931 struct diff_options *o)
2932 {
2933 struct argv_array argv = ARGV_ARRAY_INIT;
2934 struct argv_array env = ARGV_ARRAY_INIT;
2935 struct diff_queue_struct *q = &diff_queued_diff;
2936
2937 argv_array_push(&argv, pgm);
2938 argv_array_push(&argv, name);
2939
2940 if (one && two) {
2941 add_external_diff_name(&argv, name, one);
2942 if (!other)
2943 add_external_diff_name(&argv, name, two);
2944 else {
2945 add_external_diff_name(&argv, other, two);
2946 argv_array_push(&argv, other);
2947 argv_array_push(&argv, xfrm_msg);
2948 }
2949 }
2950
2951 argv_array_pushf(&env, "GIT_DIFF_PATH_COUNTER=%d", ++o->diff_path_counter);
2952 argv_array_pushf(&env, "GIT_DIFF_PATH_TOTAL=%d", q->nr);
2953
2954 if (run_command_v_opt_cd_env(argv.argv, RUN_USING_SHELL, NULL, env.argv))
2955 die(_("external diff died, stopping at %s"), name);
2956
2957 remove_tempfile();
2958 argv_array_clear(&argv);
2959 argv_array_clear(&env);
2960 }
2961
2962 static int similarity_index(struct diff_filepair *p)
2963 {
2964 return p->score * 100 / MAX_SCORE;
2965 }
2966
2967 static void fill_metainfo(struct strbuf *msg,
2968 const char *name,
2969 const char *other,
2970 struct diff_filespec *one,
2971 struct diff_filespec *two,
2972 struct diff_options *o,
2973 struct diff_filepair *p,
2974 int *must_show_header,
2975 int use_color)
2976 {
2977 const char *set = diff_get_color(use_color, DIFF_METAINFO);
2978 const char *reset = diff_get_color(use_color, DIFF_RESET);
2979 const char *line_prefix = diff_line_prefix(o);
2980
2981 *must_show_header = 1;
2982 strbuf_init(msg, PATH_MAX * 2 + 300);
2983 switch (p->status) {
2984 case DIFF_STATUS_COPIED:
2985 strbuf_addf(msg, "%s%ssimilarity index %d%%",
2986 line_prefix, set, similarity_index(p));
2987 strbuf_addf(msg, "%s\n%s%scopy from ",
2988 reset, line_prefix, set);
2989 quote_c_style(name, msg, NULL, 0);
2990 strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
2991 quote_c_style(other, msg, NULL, 0);
2992 strbuf_addf(msg, "%s\n", reset);
2993 break;
2994 case DIFF_STATUS_RENAMED:
2995 strbuf_addf(msg, "%s%ssimilarity index %d%%",
2996 line_prefix, set, similarity_index(p));
2997 strbuf_addf(msg, "%s\n%s%srename from ",
2998 reset, line_prefix, set);
2999 quote_c_style(name, msg, NULL, 0);
3000 strbuf_addf(msg, "%s\n%s%srename to ",
3001 reset, line_prefix, set);
3002 quote_c_style(other, msg, NULL, 0);
3003 strbuf_addf(msg, "%s\n", reset);
3004 break;
3005 case DIFF_STATUS_MODIFIED:
3006 if (p->score) {
3007 strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
3008 line_prefix,
3009 set, similarity_index(p), reset);
3010 break;
3011 }
3012 /* fallthru */
3013 default:
3014 *must_show_header = 0;
3015 }
3016 if (one && two && hashcmp(one->sha1, two->sha1)) {
3017 int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
3018
3019 if (DIFF_OPT_TST(o, BINARY)) {
3020 mmfile_t mf;
3021 if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
3022 (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
3023 abbrev = 40;
3024 }
3025 strbuf_addf(msg, "%s%sindex %s..", line_prefix, set,
3026 find_unique_abbrev(one->sha1, abbrev));
3027 strbuf_addstr(msg, find_unique_abbrev(two->sha1, abbrev));
3028 if (one->mode == two->mode)
3029 strbuf_addf(msg, " %06o", one->mode);
3030 strbuf_addf(msg, "%s\n", reset);
3031 }
3032 }
3033
3034 static void run_diff_cmd(const char *pgm,
3035 const char *name,
3036 const char *other,
3037 const char *attr_path,
3038 struct diff_filespec *one,
3039 struct diff_filespec *two,
3040 struct strbuf *msg,
3041 struct diff_options *o,
3042 struct diff_filepair *p)
3043 {
3044 const char *xfrm_msg = NULL;
3045 int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
3046 int must_show_header = 0;
3047
3048
3049 if (DIFF_OPT_TST(o, ALLOW_EXTERNAL)) {
3050 struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
3051 if (drv && drv->external)
3052 pgm = drv->external;
3053 }
3054
3055 if (msg) {
3056 /*
3057 * don't use colors when the header is intended for an
3058 * external diff driver
3059 */
3060 fill_metainfo(msg, name, other, one, two, o, p,
3061 &must_show_header,
3062 want_color(o->use_color) && !pgm);
3063 xfrm_msg = msg->len ? msg->buf : NULL;
3064 }
3065
3066 if (pgm) {
3067 run_external_diff(pgm, name, other, one, two, xfrm_msg,
3068 complete_rewrite, o);
3069 return;
3070 }
3071 if (one && two)
3072 builtin_diff(name, other ? other : name,
3073 one, two, xfrm_msg, must_show_header,
3074 o, complete_rewrite);
3075 else
3076 fprintf(o->file, "* Unmerged path %s\n", name);
3077 }
3078
3079 static void diff_fill_sha1_info(struct diff_filespec *one)
3080 {
3081 if (DIFF_FILE_VALID(one)) {
3082 if (!one->sha1_valid) {
3083 struct stat st;
3084 if (one->is_stdin) {
3085 hashcpy(one->sha1, null_sha1);
3086 return;
3087 }
3088 if (lstat(one->path, &st) < 0)
3089 die_errno("stat '%s'", one->path);
3090 if (index_path(one->sha1, one->path, &st, 0))
3091 die("cannot hash %s", one->path);
3092 }
3093 }
3094 else
3095 hashclr(one->sha1);
3096 }
3097
3098 static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
3099 {
3100 /* Strip the prefix but do not molest /dev/null and absolute paths */
3101 if (*namep && **namep != '/') {
3102 *namep += prefix_length;
3103 if (**namep == '/')
3104 ++*namep;
3105 }
3106 if (*otherp && **otherp != '/') {
3107 *otherp += prefix_length;
3108 if (**otherp == '/')
3109 ++*otherp;
3110 }
3111 }
3112
3113 static void run_diff(struct diff_filepair *p, struct diff_options *o)
3114 {
3115 const char *pgm = external_diff();
3116 struct strbuf msg;
3117 struct diff_filespec *one = p->one;
3118 struct diff_filespec *two = p->two;
3119 const char *name;
3120 const char *other;
3121 const char *attr_path;
3122
3123 name = p->one->path;
3124 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3125 attr_path = name;
3126 if (o->prefix_length)
3127 strip_prefix(o->prefix_length, &name, &other);
3128
3129 if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
3130 pgm = NULL;
3131
3132 if (DIFF_PAIR_UNMERGED(p)) {
3133 run_diff_cmd(pgm, name, NULL, attr_path,
3134 NULL, NULL, NULL, o, p);
3135 return;
3136 }
3137
3138 diff_fill_sha1_info(one);
3139 diff_fill_sha1_info(two);
3140
3141 if (!pgm &&
3142 DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3143 (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3144 /*
3145 * a filepair that changes between file and symlink
3146 * needs to be split into deletion and creation.
3147 */
3148 struct diff_filespec *null = alloc_filespec(two->path);
3149 run_diff_cmd(NULL, name, other, attr_path,
3150 one, null, &msg, o, p);
3151 free(null);
3152 strbuf_release(&msg);
3153
3154 null = alloc_filespec(one->path);
3155 run_diff_cmd(NULL, name, other, attr_path,
3156 null, two, &msg, o, p);
3157 free(null);
3158 }
3159 else
3160 run_diff_cmd(pgm, name, other, attr_path,
3161 one, two, &msg, o, p);
3162
3163 strbuf_release(&msg);
3164 }
3165
3166 static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3167 struct diffstat_t *diffstat)
3168 {
3169 const char *name;
3170 const char *other;
3171
3172 if (DIFF_PAIR_UNMERGED(p)) {
3173 /* unmerged */
3174 builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, p);
3175 return;
3176 }
3177
3178 name = p->one->path;
3179 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3180
3181 if (o->prefix_length)
3182 strip_prefix(o->prefix_length, &name, &other);
3183
3184 diff_fill_sha1_info(p->one);
3185 diff_fill_sha1_info(p->two);
3186
3187 builtin_diffstat(name, other, p->one, p->two, diffstat, o, p);
3188 }
3189
3190 static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3191 {
3192 const char *name;
3193 const char *other;
3194 const char *attr_path;
3195
3196 if (DIFF_PAIR_UNMERGED(p)) {
3197 /* unmerged */
3198 return;
3199 }
3200
3201 name = p->one->path;
3202 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3203 attr_path = other ? other : name;
3204
3205 if (o->prefix_length)
3206 strip_prefix(o->prefix_length, &name, &other);
3207
3208 diff_fill_sha1_info(p->one);
3209 diff_fill_sha1_info(p->two);
3210
3211 builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3212 }
3213
3214 void diff_setup(struct diff_options *options)
3215 {
3216 memcpy(options, &default_diff_options, sizeof(*options));
3217
3218 options->file = stdout;
3219
3220 options->line_termination = '\n';
3221 options->break_opt = -1;
3222 options->rename_limit = -1;
3223 options->dirstat_permille = diff_dirstat_permille_default;
3224 options->context = diff_context_default;
3225 DIFF_OPT_SET(options, RENAME_EMPTY);
3226
3227 /* pathchange left =NULL by default */
3228 options->change = diff_change;
3229 options->add_remove = diff_addremove;
3230 options->use_color = diff_use_color_default;
3231 options->detect_rename = diff_detect_rename_default;
3232 options->xdl_opts |= diff_algorithm;
3233
3234 options->orderfile = diff_order_file_cfg;
3235
3236 if (diff_no_prefix) {
3237 options->a_prefix = options->b_prefix = "";
3238 } else if (!diff_mnemonic_prefix) {
3239 options->a_prefix = "a/";
3240 options->b_prefix = "b/";
3241 }
3242 }
3243
3244 void diff_setup_done(struct diff_options *options)
3245 {
3246 int count = 0;
3247
3248 if (options->set_default)
3249 options->set_default(options);
3250
3251 if (options->output_format & DIFF_FORMAT_NAME)
3252 count++;
3253 if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3254 count++;
3255 if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3256 count++;
3257 if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3258 count++;
3259 if (count > 1)
3260 die("--name-only, --name-status, --check and -s are mutually exclusive");
3261
3262 /*
3263 * Most of the time we can say "there are changes"
3264 * only by checking if there are changed paths, but
3265 * --ignore-whitespace* options force us to look
3266 * inside contents.
3267 */
3268
3269 if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3270 DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3271 DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3272 DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3273 else
3274 DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3275
3276 if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3277 options->detect_rename = DIFF_DETECT_COPY;
3278
3279 if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3280 options->prefix = NULL;
3281 if (options->prefix)
3282 options->prefix_length = strlen(options->prefix);
3283 else
3284 options->prefix_length = 0;
3285
3286 if (options->output_format & (DIFF_FORMAT_NAME |
3287 DIFF_FORMAT_NAME_STATUS |
3288 DIFF_FORMAT_CHECKDIFF |
3289 DIFF_FORMAT_NO_OUTPUT))
3290 options->output_format &= ~(DIFF_FORMAT_RAW |
3291 DIFF_FORMAT_NUMSTAT |
3292 DIFF_FORMAT_DIFFSTAT |
3293 DIFF_FORMAT_SHORTSTAT |
3294 DIFF_FORMAT_DIRSTAT |
3295 DIFF_FORMAT_SUMMARY |
3296 DIFF_FORMAT_PATCH);
3297
3298 /*
3299 * These cases always need recursive; we do not drop caller-supplied
3300 * recursive bits for other formats here.
3301 */
3302 if (options->output_format & (DIFF_FORMAT_PATCH |
3303 DIFF_FORMAT_NUMSTAT |
3304 DIFF_FORMAT_DIFFSTAT |
3305 DIFF_FORMAT_SHORTSTAT |
3306 DIFF_FORMAT_DIRSTAT |
3307 DIFF_FORMAT_SUMMARY |
3308 DIFF_FORMAT_CHECKDIFF))
3309 DIFF_OPT_SET(options, RECURSIVE);
3310 /*
3311 * Also pickaxe would not work very well if you do not say recursive
3312 */
3313 if (options->pickaxe)
3314 DIFF_OPT_SET(options, RECURSIVE);
3315 /*
3316 * When patches are generated, submodules diffed against the work tree
3317 * must be checked for dirtiness too so it can be shown in the output
3318 */
3319 if (options->output_format & DIFF_FORMAT_PATCH)
3320 DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3321
3322 if (options->detect_rename && options->rename_limit < 0)
3323 options->rename_limit = diff_rename_limit_default;
3324 if (options->setup & DIFF_SETUP_USE_CACHE) {
3325 if (!active_cache)
3326 /* read-cache does not die even when it fails
3327 * so it is safe for us to do this here. Also
3328 * it does not smudge active_cache or active_nr
3329 * when it fails, so we do not have to worry about
3330 * cleaning it up ourselves either.
3331 */
3332 read_cache();
3333 }
3334 if (options->abbrev <= 0 || 40 < options->abbrev)
3335 options->abbrev = 40; /* full */
3336
3337 /*
3338 * It does not make sense to show the first hit we happened
3339 * to have found. It does not make sense not to return with
3340 * exit code in such a case either.
3341 */
3342 if (DIFF_OPT_TST(options, QUICK)) {
3343 options->output_format = DIFF_FORMAT_NO_OUTPUT;
3344 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3345 }
3346
3347 options->diff_path_counter = 0;
3348
3349 if (DIFF_OPT_TST(options, FOLLOW_RENAMES) && options->pathspec.nr != 1)
3350 die(_("--follow requires exactly one pathspec"));
3351 }
3352
3353 static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3354 {
3355 char c, *eq;
3356 int len;
3357
3358 if (*arg != '-')
3359 return 0;
3360 c = *++arg;
3361 if (!c)
3362 return 0;
3363 if (c == arg_short) {
3364 c = *++arg;
3365 if (!c)
3366 return 1;
3367 if (val && isdigit(c)) {
3368 char *end;
3369 int n = strtoul(arg, &end, 10);
3370 if (*end)
3371 return 0;
3372 *val = n;
3373 return 1;
3374 }
3375 return 0;
3376 }
3377 if (c != '-')
3378 return 0;
3379 arg++;
3380 eq = strchrnul(arg, '=');
3381 len = eq - arg;
3382 if (!len || strncmp(arg, arg_long, len))
3383 return 0;
3384 if (*eq) {
3385 int n;
3386 char *end;
3387 if (!isdigit(*++eq))
3388 return 0;
3389 n = strtoul(eq, &end, 10);
3390 if (*end)
3391 return 0;
3392 *val = n;
3393 }
3394 return 1;
3395 }
3396
3397 static int diff_scoreopt_parse(const char *opt);
3398
3399 static inline int short_opt(char opt, const char **argv,
3400 const char **optarg)
3401 {
3402 const char *arg = argv[0];
3403 if (arg[0] != '-' || arg[1] != opt)
3404 return 0;
3405 if (arg[2] != '\0') {
3406 *optarg = arg + 2;
3407 return 1;
3408 }
3409 if (!argv[1])
3410 die("Option '%c' requires a value", opt);
3411 *optarg = argv[1];
3412 return 2;
3413 }
3414
3415 int parse_long_opt(const char *opt, const char **argv,
3416 const char **optarg)
3417 {
3418 const char *arg = argv[0];
3419 if (!skip_prefix(arg, "--", &arg))
3420 return 0;
3421 if (!skip_prefix(arg, opt, &arg))
3422 return 0;
3423 if (*arg == '=') { /* stuck form: --option=value */
3424 *optarg = arg + 1;
3425 return 1;
3426 }
3427 if (*arg != '\0')
3428 return 0;
3429 /* separate form: --option value */
3430 if (!argv[1])
3431 die("Option '--%s' requires a value", opt);
3432 *optarg = argv[1];
3433 return 2;
3434 }
3435
3436 static int stat_opt(struct diff_options *options, const char **av)
3437 {
3438 const char *arg = av[0];
3439 char *end;
3440 int width = options->stat_width;
3441 int name_width = options->stat_name_width;
3442 int graph_width = options->stat_graph_width;
3443 int count = options->stat_count;
3444 int argcount = 1;
3445
3446 if (!skip_prefix(arg, "--stat", &arg))
3447 die("BUG: stat option does not begin with --stat: %s", arg);
3448 end = (char *)arg;
3449
3450 switch (*arg) {
3451 case '-':
3452 if (skip_prefix(arg, "-width", &arg)) {
3453 if (*arg == '=')
3454 width = strtoul(arg + 1, &end, 10);
3455 else if (!*arg && !av[1])
3456 die("Option '--stat-width' requires a value");
3457 else if (!*arg) {
3458 width = strtoul(av[1], &end, 10);
3459 argcount = 2;
3460 }
3461 } else if (skip_prefix(arg, "-name-width", &arg)) {
3462 if (*arg == '=')
3463 name_width = strtoul(arg + 1, &end, 10);
3464 else if (!*arg && !av[1])
3465 die("Option '--stat-name-width' requires a value");
3466 else if (!*arg) {
3467 name_width = strtoul(av[1], &end, 10);
3468 argcount = 2;
3469 }
3470 } else if (skip_prefix(arg, "-graph-width", &arg)) {
3471 if (*arg == '=')
3472 graph_width = strtoul(arg + 1, &end, 10);
3473 else if (!*arg && !av[1])
3474 die("Option '--stat-graph-width' requires a value");
3475 else if (!*arg) {
3476 graph_width = strtoul(av[1], &end, 10);
3477 argcount = 2;
3478 }
3479 } else if (skip_prefix(arg, "-count", &arg)) {
3480 if (*arg == '=')
3481 count = strtoul(arg + 1, &end, 10);
3482 else if (!*arg && !av[1])
3483 die("Option '--stat-count' requires a value");
3484 else if (!*arg) {
3485 count = strtoul(av[1], &end, 10);
3486 argcount = 2;
3487 }
3488 }
3489 break;
3490 case '=':
3491 width = strtoul(arg+1, &end, 10);
3492 if (*end == ',')
3493 name_width = strtoul(end+1, &end, 10);
3494 if (*end == ',')
3495 count = strtoul(end+1, &end, 10);
3496 }
3497
3498 /* Important! This checks all the error cases! */
3499 if (*end)
3500 return 0;
3501 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3502 options->stat_name_width = name_width;
3503 options->stat_graph_width = graph_width;
3504 options->stat_width = width;
3505 options->stat_count = count;
3506 return argcount;
3507 }
3508
3509 static int parse_dirstat_opt(struct diff_options *options, const char *params)
3510 {
3511 struct strbuf errmsg = STRBUF_INIT;
3512 if (parse_dirstat_params(options, params, &errmsg))
3513 die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3514 errmsg.buf);
3515 strbuf_release(&errmsg);
3516 /*
3517 * The caller knows a dirstat-related option is given from the command
3518 * line; allow it to say "return this_function();"
3519 */
3520 options->output_format |= DIFF_FORMAT_DIRSTAT;
3521 return 1;
3522 }
3523
3524 static int parse_submodule_opt(struct diff_options *options, const char *value)
3525 {
3526 if (parse_submodule_params(options, value))
3527 die(_("Failed to parse --submodule option parameter: '%s'"),
3528 value);
3529 return 1;
3530 }
3531
3532 static const char diff_status_letters[] = {
3533 DIFF_STATUS_ADDED,
3534 DIFF_STATUS_COPIED,
3535 DIFF_STATUS_DELETED,
3536 DIFF_STATUS_MODIFIED,
3537 DIFF_STATUS_RENAMED,
3538 DIFF_STATUS_TYPE_CHANGED,
3539 DIFF_STATUS_UNKNOWN,
3540 DIFF_STATUS_UNMERGED,
3541 DIFF_STATUS_FILTER_AON,
3542 DIFF_STATUS_FILTER_BROKEN,
3543 '\0',
3544 };
3545
3546 static unsigned int filter_bit['Z' + 1];
3547
3548 static void prepare_filter_bits(void)
3549 {
3550 int i;
3551
3552 if (!filter_bit[DIFF_STATUS_ADDED]) {
3553 for (i = 0; diff_status_letters[i]; i++)
3554 filter_bit[(int) diff_status_letters[i]] = (1 << i);
3555 }
3556 }
3557
3558 static unsigned filter_bit_tst(char status, const struct diff_options *opt)
3559 {
3560 return opt->filter & filter_bit[(int) status];
3561 }
3562
3563 static int parse_diff_filter_opt(const char *optarg, struct diff_options *opt)
3564 {
3565 int i, optch;
3566
3567 prepare_filter_bits();
3568
3569 /*
3570 * If there is a negation e.g. 'd' in the input, and we haven't
3571 * initialized the filter field with another --diff-filter, start
3572 * from full set of bits, except for AON.
3573 */
3574 if (!opt->filter) {
3575 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3576 if (optch < 'a' || 'z' < optch)
3577 continue;
3578 opt->filter = (1 << (ARRAY_SIZE(diff_status_letters) - 1)) - 1;
3579 opt->filter &= ~filter_bit[DIFF_STATUS_FILTER_AON];
3580 break;
3581 }
3582 }
3583
3584 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3585 unsigned int bit;
3586 int negate;
3587
3588 if ('a' <= optch && optch <= 'z') {
3589 negate = 1;
3590 optch = toupper(optch);
3591 } else {
3592 negate = 0;
3593 }
3594
3595 bit = (0 <= optch && optch <= 'Z') ? filter_bit[optch] : 0;
3596 if (!bit)
3597 return optarg[i];
3598 if (negate)
3599 opt->filter &= ~bit;
3600 else
3601 opt->filter |= bit;
3602 }
3603 return 0;
3604 }
3605
3606 static void enable_patch_output(int *fmt) {
3607 *fmt &= ~DIFF_FORMAT_NO_OUTPUT;
3608 *fmt |= DIFF_FORMAT_PATCH;
3609 }
3610
3611 int diff_opt_parse(struct diff_options *options, const char **av, int ac)
3612 {
3613 const char *arg = av[0];
3614 const char *optarg;
3615 int argcount;
3616
3617 /* Output format options */
3618 if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch")
3619 || opt_arg(arg, 'U', "unified", &options->context))
3620 enable_patch_output(&options->output_format);
3621 else if (!strcmp(arg, "--raw"))
3622 options->output_format |= DIFF_FORMAT_RAW;
3623 else if (!strcmp(arg, "--patch-with-raw")) {
3624 enable_patch_output(&options->output_format);
3625 options->output_format |= DIFF_FORMAT_RAW;
3626 } else if (!strcmp(arg, "--numstat"))
3627 options->output_format |= DIFF_FORMAT_NUMSTAT;
3628 else if (!strcmp(arg, "--shortstat"))
3629 options->output_format |= DIFF_FORMAT_SHORTSTAT;
3630 else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3631 return parse_dirstat_opt(options, "");
3632 else if (skip_prefix(arg, "-X", &arg))
3633 return parse_dirstat_opt(options, arg);
3634 else if (skip_prefix(arg, "--dirstat=", &arg))
3635 return parse_dirstat_opt(options, arg);
3636 else if (!strcmp(arg, "--cumulative"))
3637 return parse_dirstat_opt(options, "cumulative");
3638 else if (!strcmp(arg, "--dirstat-by-file"))
3639 return parse_dirstat_opt(options, "files");
3640 else if (skip_prefix(arg, "--dirstat-by-file=", &arg)) {
3641 parse_dirstat_opt(options, "files");
3642 return parse_dirstat_opt(options, arg);
3643 }
3644 else if (!strcmp(arg, "--check"))
3645 options->output_format |= DIFF_FORMAT_CHECKDIFF;
3646 else if (!strcmp(arg, "--summary"))
3647 options->output_format |= DIFF_FORMAT_SUMMARY;
3648 else if (!strcmp(arg, "--patch-with-stat")) {
3649 enable_patch_output(&options->output_format);
3650 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3651 } else if (!strcmp(arg, "--name-only"))
3652 options->output_format |= DIFF_FORMAT_NAME;
3653 else if (!strcmp(arg, "--name-status"))
3654 options->output_format |= DIFF_FORMAT_NAME_STATUS;
3655 else if (!strcmp(arg, "-s") || !strcmp(arg, "--no-patch"))
3656 options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3657 else if (starts_with(arg, "--stat"))
3658 /* --stat, --stat-width, --stat-name-width, or --stat-count */
3659 return stat_opt(options, av);
3660
3661 /* renames options */
3662 else if (starts_with(arg, "-B") || starts_with(arg, "--break-rewrites=") ||
3663 !strcmp(arg, "--break-rewrites")) {
3664 if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3665 return error("invalid argument to -B: %s", arg+2);
3666 }
3667 else if (starts_with(arg, "-M") || starts_with(arg, "--find-renames=") ||
3668 !strcmp(arg, "--find-renames")) {
3669 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3670 return error("invalid argument to -M: %s", arg+2);
3671 options->detect_rename = DIFF_DETECT_RENAME;
3672 }
3673 else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3674 options->irreversible_delete = 1;
3675 }