?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
test-sha1: add a binary output mode
[git/gitweb.git] / diff.c
blob71e4df96d48e3409f2bb84ca0c145e4f6278e985
1 /*
2 * Copyright (C) 2005 Junio C Hamano
3 */
4 #include "cache.h"
5 #include "quote.h"
6 #include "diff.h"
7 #include "diffcore.h"
8 #include "delta.h"
9 #include "xdiff-interface.h"
10 #include "color.h"
11 #include "attr.h"
12 #include "run-command.h"
13 #include "utf8.h"
14 #include "userdiff.h"
15 #include "sigchain.h"
16 #include "submodule.h"
17 #include "ll-merge.h"
18 #include "string-list.h"
19
20 #ifdef NO_FAST_WORKING_DIRECTORY
21 #define FAST_WORKING_DIRECTORY 0
22 #else
23 #define FAST_WORKING_DIRECTORY 1
24 #endif
25
26 static int diff_detect_rename_default;
27 static int diff_rename_limit_default = 400;
28 static int diff_suppress_blank_empty;
29 static int diff_use_color_default = -1;
30 static int diff_context_default = 3;
31 static const char *diff_word_regex_cfg;
32 static const char *external_diff_cmd_cfg;
33 int diff_auto_refresh_index = 1;
34 static int diff_mnemonic_prefix;
35 static int diff_no_prefix;
36 static int diff_stat_graph_width;
37 static int diff_dirstat_permille_default = 30;
38 static struct diff_options default_diff_options;
39 static long diff_algorithm;
40
41 static char diff_colors[][COLOR_MAXLEN] = {
42 GIT_COLOR_RESET,
43 GIT_COLOR_NORMAL, /* PLAIN */
44 GIT_COLOR_BOLD, /* METAINFO */
45 GIT_COLOR_CYAN, /* FRAGINFO */
46 GIT_COLOR_RED, /* OLD */
47 GIT_COLOR_GREEN, /* NEW */
48 GIT_COLOR_YELLOW, /* COMMIT */
49 GIT_COLOR_BG_RED, /* WHITESPACE */
50 GIT_COLOR_NORMAL, /* FUNCINFO */
51 };
52
53 static int parse_diff_color_slot(const char *var, int ofs)
54 {
55 if (!strcasecmp(var+ofs, "plain"))
56 return DIFF_PLAIN;
57 if (!strcasecmp(var+ofs, "meta"))
58 return DIFF_METAINFO;
59 if (!strcasecmp(var+ofs, "frag"))
60 return DIFF_FRAGINFO;
61 if (!strcasecmp(var+ofs, "old"))
62 return DIFF_FILE_OLD;
63 if (!strcasecmp(var+ofs, "new"))
64 return DIFF_FILE_NEW;
65 if (!strcasecmp(var+ofs, "commit"))
66 return DIFF_COMMIT;
67 if (!strcasecmp(var+ofs, "whitespace"))
68 return DIFF_WHITESPACE;
69 if (!strcasecmp(var+ofs, "func"))
70 return DIFF_FUNCINFO;
71 return -1;
72 }
73
74 static int parse_dirstat_params(struct diff_options *options, const char *params_string,
75 struct strbuf *errmsg)
76 {
77 char *params_copy = xstrdup(params_string);
78 struct string_list params = STRING_LIST_INIT_NODUP;
79 int ret = 0;
80 int i;
81
82 if (*params_copy)
83 string_list_split_in_place(&params, params_copy, ',', -1);
84 for (i = 0; i < params.nr; i++) {
85 const char *p = params.items[i].string;
86 if (!strcmp(p, "changes")) {
87 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
88 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
89 } else if (!strcmp(p, "lines")) {
90 DIFF_OPT_SET(options, DIRSTAT_BY_LINE);
91 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
92 } else if (!strcmp(p, "files")) {
93 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
94 DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
95 } else if (!strcmp(p, "noncumulative")) {
96 DIFF_OPT_CLR(options, DIRSTAT_CUMULATIVE);
97 } else if (!strcmp(p, "cumulative")) {
98 DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
99 } else if (isdigit(*p)) {
100 char *end;
101 int permille = strtoul(p, &end, 10) * 10;
102 if (*end == '.' && isdigit(*++end)) {
103 /* only use first digit */
104 permille += *end - '0';
105 /* .. and ignore any further digits */
106 while (isdigit(*++end))
107 ; /* nothing */
108 }
109 if (!*end)
110 options->dirstat_permille = permille;
111 else {
112 strbuf_addf(errmsg, _(" Failed to parse dirstat cut-off percentage '%s'\n"),
113 p);
114 ret++;
115 }
116 } else {
117 strbuf_addf(errmsg, _(" Unknown dirstat parameter '%s'\n"), p);
118 ret++;
119 }
120
121 }
122 string_list_clear(&params, 0);
123 free(params_copy);
124 return ret;
125 }
126
127 static int parse_submodule_params(struct diff_options *options, const char *value)
128 {
129 if (!strcmp(value, "log"))
130 DIFF_OPT_SET(options, SUBMODULE_LOG);
131 else if (!strcmp(value, "short"))
132 DIFF_OPT_CLR(options, SUBMODULE_LOG);
133 else
134 return -1;
135 return 0;
136 }
137
138 static int git_config_rename(const char *var, const char *value)
139 {
140 if (!value)
141 return DIFF_DETECT_RENAME;
142 if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
143 return DIFF_DETECT_COPY;
144 return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
145 }
146
147 long parse_algorithm_value(const char *value)
148 {
149 if (!value)
150 return -1;
151 else if (!strcasecmp(value, "myers") || !strcasecmp(value, "default"))
152 return 0;
153 else if (!strcasecmp(value, "minimal"))
154 return XDF_NEED_MINIMAL;
155 else if (!strcasecmp(value, "patience"))
156 return XDF_PATIENCE_DIFF;
157 else if (!strcasecmp(value, "histogram"))
158 return XDF_HISTOGRAM_DIFF;
159 return -1;
160 }
161
162 /*
163 * These are to give UI layer defaults.
164 * The core-level commands such as git-diff-files should
165 * never be affected by the setting of diff.renames
166 * the user happens to have in the configuration file.
167 */
168 int git_diff_ui_config(const char *var, const char *value, void *cb)
169 {
170 if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
171 diff_use_color_default = git_config_colorbool(var, value);
172 return 0;
173 }
174 if (!strcmp(var, "diff.context")) {
175 diff_context_default = git_config_int(var, value);
176 if (diff_context_default < 0)
177 return -1;
178 return 0;
179 }
180 if (!strcmp(var, "diff.renames")) {
181 diff_detect_rename_default = git_config_rename(var, value);
182 return 0;
183 }
184 if (!strcmp(var, "diff.autorefreshindex")) {
185 diff_auto_refresh_index = git_config_bool(var, value);
186 return 0;
187 }
188 if (!strcmp(var, "diff.mnemonicprefix")) {
189 diff_mnemonic_prefix = git_config_bool(var, value);
190 return 0;
191 }
192 if (!strcmp(var, "diff.noprefix")) {
193 diff_no_prefix = git_config_bool(var, value);
194 return 0;
195 }
196 if (!strcmp(var, "diff.statgraphwidth")) {
197 diff_stat_graph_width = git_config_int(var, value);
198 return 0;
199 }
200 if (!strcmp(var, "diff.external"))
201 return git_config_string(&external_diff_cmd_cfg, var, value);
202 if (!strcmp(var, "diff.wordregex"))
203 return git_config_string(&diff_word_regex_cfg, var, value);
204
205 if (!strcmp(var, "diff.ignoresubmodules"))
206 handle_ignore_submodules_arg(&default_diff_options, value);
207
208 if (!strcmp(var, "diff.submodule")) {
209 if (parse_submodule_params(&default_diff_options, value))
210 warning(_("Unknown value for 'diff.submodule' config variable: '%s'"),
211 value);
212 return 0;
213 }
214
215 if (!strcmp(var, "diff.algorithm")) {
216 diff_algorithm = parse_algorithm_value(value);
217 if (diff_algorithm < 0)
218 return -1;
219 return 0;
220 }
221
222 if (git_color_config(var, value, cb) < 0)
223 return -1;
224
225 return git_diff_basic_config(var, value, cb);
226 }
227
228 int git_diff_basic_config(const char *var, const char *value, void *cb)
229 {
230 if (!strcmp(var, "diff.renamelimit")) {
231 diff_rename_limit_default = git_config_int(var, value);
232 return 0;
233 }
234
235 if (userdiff_config(var, value) < 0)
236 return -1;
237
238 if (!prefixcmp(var, "diff.color.") || !prefixcmp(var, "color.diff.")) {
239 int slot = parse_diff_color_slot(var, 11);
240 if (slot < 0)
241 return 0;
242 if (!value)
243 return config_error_nonbool(var);
244 color_parse(value, var, diff_colors[slot]);
245 return 0;
246 }
247
248 /* like GNU diff's --suppress-blank-empty option */
249 if (!strcmp(var, "diff.suppressblankempty") ||
250 /* for backwards compatibility */
251 !strcmp(var, "diff.suppress-blank-empty")) {
252 diff_suppress_blank_empty = git_config_bool(var, value);
253 return 0;
254 }
255
256 if (!strcmp(var, "diff.dirstat")) {
257 struct strbuf errmsg = STRBUF_INIT;
258 default_diff_options.dirstat_permille = diff_dirstat_permille_default;
259 if (parse_dirstat_params(&default_diff_options, value, &errmsg))
260 warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
261 errmsg.buf);
262 strbuf_release(&errmsg);
263 diff_dirstat_permille_default = default_diff_options.dirstat_permille;
264 return 0;
265 }
266
267 if (!prefixcmp(var, "submodule."))
268 return parse_submodule_config_option(var, value);
269
270 return git_default_config(var, value, cb);
271 }
272
273 static char *quote_two(const char *one, const char *two)
274 {
275 int need_one = quote_c_style(one, NULL, NULL, 1);
276 int need_two = quote_c_style(two, NULL, NULL, 1);
277 struct strbuf res = STRBUF_INIT;
278
279 if (need_one + need_two) {
280 strbuf_addch(&res, '"');
281 quote_c_style(one, &res, NULL, 1);
282 quote_c_style(two, &res, NULL, 1);
283 strbuf_addch(&res, '"');
284 } else {
285 strbuf_addstr(&res, one);
286 strbuf_addstr(&res, two);
287 }
288 return strbuf_detach(&res, NULL);
289 }
290
291 static const char *external_diff(void)
292 {
293 static const char *external_diff_cmd = NULL;
294 static int done_preparing = 0;
295
296 if (done_preparing)
297 return external_diff_cmd;
298 external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
299 if (!external_diff_cmd)
300 external_diff_cmd = external_diff_cmd_cfg;
301 done_preparing = 1;
302 return external_diff_cmd;
303 }
304
305 static struct diff_tempfile {
306 const char *name; /* filename external diff should read from */
307 char hex[41];
308 char mode[10];
309 char tmp_path[PATH_MAX];
310 } diff_temp[2];
311
312 typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
313
314 struct emit_callback {
315 int color_diff;
316 unsigned ws_rule;
317 int blank_at_eof_in_preimage;
318 int blank_at_eof_in_postimage;
319 int lno_in_preimage;
320 int lno_in_postimage;
321 sane_truncate_fn truncate;
322 const char **label_path;
323 struct diff_words_data *diff_words;
324 struct diff_options *opt;
325 int *found_changesp;
326 struct strbuf *header;
327 };
328
329 static int count_lines(const char *data, int size)
330 {
331 int count, ch, completely_empty = 1, nl_just_seen = 0;
332 count = 0;
333 while (0 < size--) {
334 ch = *data++;
335 if (ch == '\n') {
336 count++;
337 nl_just_seen = 1;
338 completely_empty = 0;
339 }
340 else {
341 nl_just_seen = 0;
342 completely_empty = 0;
343 }
344 }
345 if (completely_empty)
346 return 0;
347 if (!nl_just_seen)
348 count++; /* no trailing newline */
349 return count;
350 }
351
352 static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
353 {
354 if (!DIFF_FILE_VALID(one)) {
355 mf->ptr = (char *)""; /* does not matter */
356 mf->size = 0;
357 return 0;
358 }
359 else if (diff_populate_filespec(one, 0))
360 return -1;
361
362 mf->ptr = one->data;
363 mf->size = one->size;
364 return 0;
365 }
366
367 /* like fill_mmfile, but only for size, so we can avoid retrieving blob */
368 static unsigned long diff_filespec_size(struct diff_filespec *one)
369 {
370 if (!DIFF_FILE_VALID(one))
371 return 0;
372 diff_populate_filespec(one, 1);
373 return one->size;
374 }
375
376 static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
377 {
378 char *ptr = mf->ptr;
379 long size = mf->size;
380 int cnt = 0;
381
382 if (!size)
383 return cnt;
384 ptr += size - 1; /* pointing at the very end */
385 if (*ptr != '\n')
386 ; /* incomplete line */
387 else
388 ptr--; /* skip the last LF */
389 while (mf->ptr < ptr) {
390 char *prev_eol;
391 for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
392 if (*prev_eol == '\n')
393 break;
394 if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
395 break;
396 cnt++;
397 ptr = prev_eol - 1;
398 }
399 return cnt;
400 }
401
402 static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
403 struct emit_callback *ecbdata)
404 {
405 int l1, l2, at;
406 unsigned ws_rule = ecbdata->ws_rule;
407 l1 = count_trailing_blank(mf1, ws_rule);
408 l2 = count_trailing_blank(mf2, ws_rule);
409 if (l2 <= l1) {
410 ecbdata->blank_at_eof_in_preimage = 0;
411 ecbdata->blank_at_eof_in_postimage = 0;
412 return;
413 }
414 at = count_lines(mf1->ptr, mf1->size);
415 ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
416
417 at = count_lines(mf2->ptr, mf2->size);
418 ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
419 }
420
421 static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
422 int first, const char *line, int len)
423 {
424 int has_trailing_newline, has_trailing_carriage_return;
425 int nofirst;
426 FILE *file = o->file;
427
428 fputs(diff_line_prefix(o), file);
429
430 if (len == 0) {
431 has_trailing_newline = (first == '\n');
432 has_trailing_carriage_return = (!has_trailing_newline &&
433 (first == '\r'));
434 nofirst = has_trailing_newline || has_trailing_carriage_return;
435 } else {
436 has_trailing_newline = (len > 0 && line[len-1] == '\n');
437 if (has_trailing_newline)
438 len--;
439 has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
440 if (has_trailing_carriage_return)
441 len--;
442 nofirst = 0;
443 }
444
445 if (len || !nofirst) {
446 fputs(set, file);
447 if (!nofirst)
448 fputc(first, file);
449 fwrite(line, len, 1, file);
450 fputs(reset, file);
451 }
452 if (has_trailing_carriage_return)
453 fputc('\r', file);
454 if (has_trailing_newline)
455 fputc('\n', file);
456 }
457
458 static void emit_line(struct diff_options *o, const char *set, const char *reset,
459 const char *line, int len)
460 {
461 emit_line_0(o, set, reset, line[0], line+1, len-1);
462 }
463
464 static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
465 {
466 if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
467 ecbdata->blank_at_eof_in_preimage &&
468 ecbdata->blank_at_eof_in_postimage &&
469 ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
470 ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
471 return 0;
472 return ws_blank_line(line, len, ecbdata->ws_rule);
473 }
474
475 static void emit_add_line(const char *reset,
476 struct emit_callback *ecbdata,
477 const char *line, int len)
478 {
479 const char *ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
480 const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_NEW);
481
482 if (!*ws)
483 emit_line_0(ecbdata->opt, set, reset, '+', line, len);
484 else if (new_blank_line_at_eof(ecbdata, line, len))
485 /* Blank line at EOF - paint '+' as well */
486 emit_line_0(ecbdata->opt, ws, reset, '+', line, len);
487 else {
488 /* Emit just the prefix, then the rest. */
489 emit_line_0(ecbdata->opt, set, reset, '+', "", 0);
490 ws_check_emit(line, len, ecbdata->ws_rule,
491 ecbdata->opt->file, set, reset, ws);
492 }
493 }
494
495 static void emit_hunk_header(struct emit_callback *ecbdata,
496 const char *line, int len)
497 {
498 const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
499 const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
500 const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
501 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
502 static const char atat[2] = { '@', '@' };
503 const char *cp, *ep;
504 struct strbuf msgbuf = STRBUF_INIT;
505 int org_len = len;
506 int i = 1;
507
508 /*
509 * As a hunk header must begin with "@@ -<old>, +<new> @@",
510 * it always is at least 10 bytes long.
511 */
512 if (len < 10 ||
513 memcmp(line, atat, 2) ||
514 !(ep = memmem(line + 2, len - 2, atat, 2))) {
515 emit_line(ecbdata->opt, plain, reset, line, len);
516 return;
517 }
518 ep += 2; /* skip over @@ */
519
520 /* The hunk header in fraginfo color */
521 strbuf_add(&msgbuf, frag, strlen(frag));
522 strbuf_add(&msgbuf, line, ep - line);
523 strbuf_add(&msgbuf, reset, strlen(reset));
524
525 /*
526 * trailing "\r\n"
527 */
528 for ( ; i < 3; i++)
529 if (line[len - i] == '\r' || line[len - i] == '\n')
530 len--;
531
532 /* blank before the func header */
533 for (cp = ep; ep - line < len; ep++)
534 if (*ep != ' ' && *ep != '\t')
535 break;
536 if (ep != cp) {
537 strbuf_add(&msgbuf, plain, strlen(plain));
538 strbuf_add(&msgbuf, cp, ep - cp);
539 strbuf_add(&msgbuf, reset, strlen(reset));
540 }
541
542 if (ep < line + len) {
543 strbuf_add(&msgbuf, func, strlen(func));
544 strbuf_add(&msgbuf, ep, line + len - ep);
545 strbuf_add(&msgbuf, reset, strlen(reset));
546 }
547
548 strbuf_add(&msgbuf, line + len, org_len - len);
549 emit_line(ecbdata->opt, "", "", msgbuf.buf, msgbuf.len);
550 strbuf_release(&msgbuf);
551 }
552
553 static struct diff_tempfile *claim_diff_tempfile(void) {
554 int i;
555 for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
556 if (!diff_temp[i].name)
557 return diff_temp + i;
558 die("BUG: diff is failing to clean up its tempfiles");
559 }
560
561 static int remove_tempfile_installed;
562
563 static void remove_tempfile(void)
564 {
565 int i;
566 for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
567 if (diff_temp[i].name == diff_temp[i].tmp_path)
568 unlink_or_warn(diff_temp[i].name);
569 diff_temp[i].name = NULL;
570 }
571 }
572
573 static void remove_tempfile_on_signal(int signo)
574 {
575 remove_tempfile();
576 sigchain_pop(signo);
577 raise(signo);
578 }
579
580 static void print_line_count(FILE *file, int count)
581 {
582 switch (count) {
583 case 0:
584 fprintf(file, "0,0");
585 break;
586 case 1:
587 fprintf(file, "1");
588 break;
589 default:
590 fprintf(file, "1,%d", count);
591 break;
592 }
593 }
594
595 static void emit_rewrite_lines(struct emit_callback *ecb,
596 int prefix, const char *data, int size)
597 {
598 const char *endp = NULL;
599 static const char *nneof = " No newline at end of file\n";
600 const char *old = diff_get_color(ecb->color_diff, DIFF_FILE_OLD);
601 const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
602
603 while (0 < size) {
604 int len;
605
606 endp = memchr(data, '\n', size);
607 len = endp ? (endp - data + 1) : size;
608 if (prefix != '+') {
609 ecb->lno_in_preimage++;
610 emit_line_0(ecb->opt, old, reset, '-',
611 data, len);
612 } else {
613 ecb->lno_in_postimage++;
614 emit_add_line(reset, ecb, data, len);
615 }
616 size -= len;
617 data += len;
618 }
619 if (!endp) {
620 const char *plain = diff_get_color(ecb->color_diff,
621 DIFF_PLAIN);
622 putc('\n', ecb->opt->file);
623 emit_line_0(ecb->opt, plain, reset, '\\',
624 nneof, strlen(nneof));
625 }
626 }
627
628 static void emit_rewrite_diff(const char *name_a,
629 const char *name_b,
630 struct diff_filespec *one,
631 struct diff_filespec *two,
632 struct userdiff_driver *textconv_one,
633 struct userdiff_driver *textconv_two,
634 struct diff_options *o)
635 {
636 int lc_a, lc_b;
637 const char *name_a_tab, *name_b_tab;
638 const char *metainfo = diff_get_color(o->use_color, DIFF_METAINFO);
639 const char *fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
640 const char *reset = diff_get_color(o->use_color, DIFF_RESET);
641 static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
642 const char *a_prefix, *b_prefix;
643 char *data_one, *data_two;
644 size_t size_one, size_two;
645 struct emit_callback ecbdata;
646 const char *line_prefix = diff_line_prefix(o);
647
648 if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
649 a_prefix = o->b_prefix;
650 b_prefix = o->a_prefix;
651 } else {
652 a_prefix = o->a_prefix;
653 b_prefix = o->b_prefix;
654 }
655
656 name_a += (*name_a == '/');
657 name_b += (*name_b == '/');
658 name_a_tab = strchr(name_a, ' ') ? "\t" : "";
659 name_b_tab = strchr(name_b, ' ') ? "\t" : "";
660
661 strbuf_reset(&a_name);
662 strbuf_reset(&b_name);
663 quote_two_c_style(&a_name, a_prefix, name_a, 0);
664 quote_two_c_style(&b_name, b_prefix, name_b, 0);
665
666 size_one = fill_textconv(textconv_one, one, &data_one);
667 size_two = fill_textconv(textconv_two, two, &data_two);
668
669 memset(&ecbdata, 0, sizeof(ecbdata));
670 ecbdata.color_diff = want_color(o->use_color);
671 ecbdata.found_changesp = &o->found_changes;
672 ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
673 ecbdata.opt = o;
674 if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
675 mmfile_t mf1, mf2;
676 mf1.ptr = (char *)data_one;
677 mf2.ptr = (char *)data_two;
678 mf1.size = size_one;
679 mf2.size = size_two;
680 check_blank_at_eof(&mf1, &mf2, &ecbdata);
681 }
682 ecbdata.lno_in_preimage = 1;
683 ecbdata.lno_in_postimage = 1;
684
685 lc_a = count_lines(data_one, size_one);
686 lc_b = count_lines(data_two, size_two);
687 fprintf(o->file,
688 "%s%s--- %s%s%s\n%s%s+++ %s%s%s\n%s%s@@ -",
689 line_prefix, metainfo, a_name.buf, name_a_tab, reset,
690 line_prefix, metainfo, b_name.buf, name_b_tab, reset,
691 line_prefix, fraginfo);
692 if (!o->irreversible_delete)
693 print_line_count(o->file, lc_a);
694 else
695 fprintf(o->file, "?,?");
696 fprintf(o->file, " +");
697 print_line_count(o->file, lc_b);
698 fprintf(o->file, " @@%s\n", reset);
699 if (lc_a && !o->irreversible_delete)
700 emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
701 if (lc_b)
702 emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
703 if (textconv_one)
704 free((char *)data_one);
705 if (textconv_two)
706 free((char *)data_two);
707 }
708
709 struct diff_words_buffer {
710 mmfile_t text;
711 long alloc;
712 struct diff_words_orig {
713 const char *begin, *end;
714 } *orig;
715 int orig_nr, orig_alloc;
716 };
717
718 static void diff_words_append(char *line, unsigned long len,
719 struct diff_words_buffer *buffer)
720 {
721 ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
722 line++;
723 len--;
724 memcpy(buffer->text.ptr + buffer->text.size, line, len);
725 buffer->text.size += len;
726 buffer->text.ptr[buffer->text.size] = '\0';
727 }
728
729 struct diff_words_style_elem {
730 const char *prefix;
731 const char *suffix;
732 const char *color; /* NULL; filled in by the setup code if
733 * color is enabled */
734 };
735
736 struct diff_words_style {
737 enum diff_words_type type;
738 struct diff_words_style_elem new, old, ctx;
739 const char *newline;
740 };
741
742 static struct diff_words_style diff_words_styles[] = {
743 { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
744 { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
745 { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
746 };
747
748 struct diff_words_data {
749 struct diff_words_buffer minus, plus;
750 const char *current_plus;
751 int last_minus;
752 struct diff_options *opt;
753 regex_t *word_regex;
754 enum diff_words_type type;
755 struct diff_words_style *style;
756 };
757
758 static int fn_out_diff_words_write_helper(FILE *fp,
759 struct diff_words_style_elem *st_el,
760 const char *newline,
761 size_t count, const char *buf,
762 const char *line_prefix)
763 {
764 int print = 0;
765
766 while (count) {
767 char *p = memchr(buf, '\n', count);
768 if (print)
769 fputs(line_prefix, fp);
770 if (p != buf) {
771 if (st_el->color && fputs(st_el->color, fp) < 0)
772 return -1;
773 if (fputs(st_el->prefix, fp) < 0 ||
774 fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
775 fputs(st_el->suffix, fp) < 0)
776 return -1;
777 if (st_el->color && *st_el->color
778 && fputs(GIT_COLOR_RESET, fp) < 0)
779 return -1;
780 }
781 if (!p)
782 return 0;
783 if (fputs(newline, fp) < 0)
784 return -1;
785 count -= p + 1 - buf;
786 buf = p + 1;
787 print = 1;
788 }
789 return 0;
790 }
791
792 /*
793 * '--color-words' algorithm can be described as:
794 *
795 * 1. collect a the minus/plus lines of a diff hunk, divided into
796 * minus-lines and plus-lines;
797 *
798 * 2. break both minus-lines and plus-lines into words and
799 * place them into two mmfile_t with one word for each line;
800 *
801 * 3. use xdiff to run diff on the two mmfile_t to get the words level diff;
802 *
803 * And for the common parts of the both file, we output the plus side text.
804 * diff_words->current_plus is used to trace the current position of the plus file
805 * which printed. diff_words->last_minus is used to trace the last minus word
806 * printed.
807 *
808 * For '--graph' to work with '--color-words', we need to output the graph prefix
809 * on each line of color words output. Generally, there are two conditions on
810 * which we should output the prefix.
811 *
812 * 1. diff_words->last_minus == 0 &&
813 * diff_words->current_plus == diff_words->plus.text.ptr
814 *
815 * that is: the plus text must start as a new line, and if there is no minus
816 * word printed, a graph prefix must be printed.
817 *
818 * 2. diff_words->current_plus > diff_words->plus.text.ptr &&
819 * *(diff_words->current_plus - 1) == '\n'
820 *
821 * that is: a graph prefix must be printed following a '\n'
822 */
823 static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
824 {
825 if ((diff_words->last_minus == 0 &&
826 diff_words->current_plus == diff_words->plus.text.ptr) ||
827 (diff_words->current_plus > diff_words->plus.text.ptr &&
828 *(diff_words->current_plus - 1) == '\n')) {
829 return 1;
830 } else {
831 return 0;
832 }
833 }
834
835 static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
836 {
837 struct diff_words_data *diff_words = priv;
838 struct diff_words_style *style = diff_words->style;
839 int minus_first, minus_len, plus_first, plus_len;
840 const char *minus_begin, *minus_end, *plus_begin, *plus_end;
841 struct diff_options *opt = diff_words->opt;
842 const char *line_prefix;
843
844 if (line[0] != '@' || parse_hunk_header(line, len,
845 &minus_first, &minus_len, &plus_first, &plus_len))
846 return;
847
848 assert(opt);
849 line_prefix = diff_line_prefix(opt);
850
851 /* POSIX requires that first be decremented by one if len == 0... */
852 if (minus_len) {
853 minus_begin = diff_words->minus.orig[minus_first].begin;
854 minus_end =
855 diff_words->minus.orig[minus_first + minus_len - 1].end;
856 } else
857 minus_begin = minus_end =
858 diff_words->minus.orig[minus_first].end;
859
860 if (plus_len) {
861 plus_begin = diff_words->plus.orig[plus_first].begin;
862 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
863 } else
864 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
865
866 if (color_words_output_graph_prefix(diff_words)) {
867 fputs(line_prefix, diff_words->opt->file);
868 }
869 if (diff_words->current_plus != plus_begin) {
870 fn_out_diff_words_write_helper(diff_words->opt->file,
871 &style->ctx, style->newline,
872 plus_begin - diff_words->current_plus,
873 diff_words->current_plus, line_prefix);
874 if (*(plus_begin - 1) == '\n')
875 fputs(line_prefix, diff_words->opt->file);
876 }
877 if (minus_begin != minus_end) {
878 fn_out_diff_words_write_helper(diff_words->opt->file,
879 &style->old, style->newline,
880 minus_end - minus_begin, minus_begin,
881 line_prefix);
882 }
883 if (plus_begin != plus_end) {
884 fn_out_diff_words_write_helper(diff_words->opt->file,
885 &style->new, style->newline,
886 plus_end - plus_begin, plus_begin,
887 line_prefix);
888 }
889
890 diff_words->current_plus = plus_end;
891 diff_words->last_minus = minus_first;
892 }
893
894 /* This function starts looking at *begin, and returns 0 iff a word was found. */
895 static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
896 int *begin, int *end)
897 {
898 if (word_regex && *begin < buffer->size) {
899 regmatch_t match[1];
900 if (!regexec(word_regex, buffer->ptr + *begin, 1, match, 0)) {
901 char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
902 '\n', match[0].rm_eo - match[0].rm_so);
903 *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
904 *begin += match[0].rm_so;
905 return *begin >= *end;
906 }
907 return -1;
908 }
909
910 /* find the next word */
911 while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
912 (*begin)++;
913 if (*begin >= buffer->size)
914 return -1;
915
916 /* find the end of the word */
917 *end = *begin + 1;
918 while (*end < buffer->size && !isspace(buffer->ptr[*end]))
919 (*end)++;
920
921 return 0;
922 }
923
924 /*
925 * This function splits the words in buffer->text, stores the list with
926 * newline separator into out, and saves the offsets of the original words
927 * in buffer->orig.
928 */
929 static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
930 regex_t *word_regex)
931 {
932 int i, j;
933 long alloc = 0;
934
935 out->size = 0;
936 out->ptr = NULL;
937
938 /* fake an empty "0th" word */
939 ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
940 buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
941 buffer->orig_nr = 1;
942
943 for (i = 0; i < buffer->text.size; i++) {
944 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
945 return;
946
947 /* store original boundaries */
948 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
949 buffer->orig_alloc);
950 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
951 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
952 buffer->orig_nr++;
953
954 /* store one word */
955 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
956 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
957 out->ptr[out->size + j - i] = '\n';
958 out->size += j - i + 1;
959
960 i = j - 1;
961 }
962 }
963
964 /* this executes the word diff on the accumulated buffers */
965 static void diff_words_show(struct diff_words_data *diff_words)
966 {
967 xpparam_t xpp;
968 xdemitconf_t xecfg;
969 mmfile_t minus, plus;
970 struct diff_words_style *style = diff_words->style;
971
972 struct diff_options *opt = diff_words->opt;
973 const char *line_prefix;
974
975 assert(opt);
976 line_prefix = diff_line_prefix(opt);
977
978 /* special case: only removal */
979 if (!diff_words->plus.text.size) {
980 fputs(line_prefix, diff_words->opt->file);
981 fn_out_diff_words_write_helper(diff_words->opt->file,
982 &style->old, style->newline,
983 diff_words->minus.text.size,
984 diff_words->minus.text.ptr, line_prefix);
985 diff_words->minus.text.size = 0;
986 return;
987 }
988
989 diff_words->current_plus = diff_words->plus.text.ptr;
990 diff_words->last_minus = 0;
991
992 memset(&xpp, 0, sizeof(xpp));
993 memset(&xecfg, 0, sizeof(xecfg));
994 diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
995 diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
996 xpp.flags = 0;
997 /* as only the hunk header will be parsed, we need a 0-context */
998 xecfg.ctxlen = 0;
999 xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
1000 &xpp, &xecfg);
1001 free(minus.ptr);
1002 free(plus.ptr);
1003 if (diff_words->current_plus != diff_words->plus.text.ptr +
1004 diff_words->plus.text.size) {
1005 if (color_words_output_graph_prefix(diff_words))
1006 fputs(line_prefix, diff_words->opt->file);
1007 fn_out_diff_words_write_helper(diff_words->opt->file,
1008 &style->ctx, style->newline,
1009 diff_words->plus.text.ptr + diff_words->plus.text.size
1010 - diff_words->current_plus, diff_words->current_plus,
1011 line_prefix);
1012 }
1013 diff_words->minus.text.size = diff_words->plus.text.size = 0;
1014 }
1015
1016 /* In "color-words" mode, show word-diff of words accumulated in the buffer */
1017 static void diff_words_flush(struct emit_callback *ecbdata)
1018 {
1019 if (ecbdata->diff_words->minus.text.size ||
1020 ecbdata->diff_words->plus.text.size)
1021 diff_words_show(ecbdata->diff_words);
1022 }
1023
1024 static void diff_filespec_load_driver(struct diff_filespec *one)
1025 {
1026 /* Use already-loaded driver */
1027 if (one->driver)
1028 return;
1029
1030 if (S_ISREG(one->mode))
1031 one->driver = userdiff_find_by_path(one->path);
1032
1033 /* Fallback to default settings */
1034 if (!one->driver)
1035 one->driver = userdiff_find_by_name("default");
1036 }
1037
1038 static const char *userdiff_word_regex(struct diff_filespec *one)
1039 {
1040 diff_filespec_load_driver(one);
1041 return one->driver->word_regex;
1042 }
1043
1044 static void init_diff_words_data(struct emit_callback *ecbdata,
1045 struct diff_options *orig_opts,
1046 struct diff_filespec *one,
1047 struct diff_filespec *two)
1048 {
1049 int i;
1050 struct diff_options *o = xmalloc(sizeof(struct diff_options));
1051 memcpy(o, orig_opts, sizeof(struct diff_options));
1052
1053 ecbdata->diff_words =
1054 xcalloc(1, sizeof(struct diff_words_data));
1055 ecbdata->diff_words->type = o->word_diff;
1056 ecbdata->diff_words->opt = o;
1057 if (!o->word_regex)
1058 o->word_regex = userdiff_word_regex(one);
1059 if (!o->word_regex)
1060 o->word_regex = userdiff_word_regex(two);
1061 if (!o->word_regex)
1062 o->word_regex = diff_word_regex_cfg;
1063 if (o->word_regex) {
1064 ecbdata->diff_words->word_regex = (regex_t *)
1065 xmalloc(sizeof(regex_t));
1066 if (regcomp(ecbdata->diff_words->word_regex,
1067 o->word_regex,
1068 REG_EXTENDED | REG_NEWLINE))
1069 die ("Invalid regular expression: %s",
1070 o->word_regex);
1071 }
1072 for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1073 if (o->word_diff == diff_words_styles[i].type) {
1074 ecbdata->diff_words->style =
1075 &diff_words_styles[i];
1076 break;
1077 }
1078 }
1079 if (want_color(o->use_color)) {
1080 struct diff_words_style *st = ecbdata->diff_words->style;
1081 st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1082 st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1083 st->ctx.color = diff_get_color_opt(o, DIFF_PLAIN);
1084 }
1085 }
1086
1087 static void free_diff_words_data(struct emit_callback *ecbdata)
1088 {
1089 if (ecbdata->diff_words) {
1090 diff_words_flush(ecbdata);
1091 free (ecbdata->diff_words->opt);
1092 free (ecbdata->diff_words->minus.text.ptr);
1093 free (ecbdata->diff_words->minus.orig);
1094 free (ecbdata->diff_words->plus.text.ptr);
1095 free (ecbdata->diff_words->plus.orig);
1096 if (ecbdata->diff_words->word_regex) {
1097 regfree(ecbdata->diff_words->word_regex);
1098 free(ecbdata->diff_words->word_regex);
1099 }
1100 free(ecbdata->diff_words);
1101 ecbdata->diff_words = NULL;
1102 }
1103 }
1104
1105 const char *diff_get_color(int diff_use_color, enum color_diff ix)
1106 {
1107 if (want_color(diff_use_color))
1108 return diff_colors[ix];
1109 return "";
1110 }
1111
1112 const char *diff_line_prefix(struct diff_options *opt)
1113 {
1114 struct strbuf *msgbuf;
1115 if (!opt->output_prefix)
1116 return "";
1117
1118 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
1119 return msgbuf->buf;
1120 }
1121
1122 static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1123 {
1124 const char *cp;
1125 unsigned long allot;
1126 size_t l = len;
1127
1128 if (ecb->truncate)
1129 return ecb->truncate(line, len);
1130 cp = line;
1131 allot = l;
1132 while (0 < l) {
1133 (void) utf8_width(&cp, &l);
1134 if (!cp)
1135 break; /* truncated in the middle? */
1136 }
1137 return allot - l;
1138 }
1139
1140 static void find_lno(const char *line, struct emit_callback *ecbdata)
1141 {
1142 const char *p;
1143 ecbdata->lno_in_preimage = 0;
1144 ecbdata->lno_in_postimage = 0;
1145 p = strchr(line, '-');
1146 if (!p)
1147 return; /* cannot happen */
1148 ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1149 p = strchr(p, '+');
1150 if (!p)
1151 return; /* cannot happen */
1152 ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1153 }
1154
1155 static void fn_out_consume(void *priv, char *line, unsigned long len)
1156 {
1157 struct emit_callback *ecbdata = priv;
1158 const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
1159 const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
1160 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1161 struct diff_options *o = ecbdata->opt;
1162 const char *line_prefix = diff_line_prefix(o);
1163
1164 if (ecbdata->header) {
1165 fprintf(ecbdata->opt->file, "%s", ecbdata->header->buf);
1166 strbuf_reset(ecbdata->header);
1167 ecbdata->header = NULL;
1168 }
1169 *(ecbdata->found_changesp) = 1;
1170
1171 if (ecbdata->label_path[0]) {
1172 const char *name_a_tab, *name_b_tab;
1173
1174 name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
1175 name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
1176
1177 fprintf(ecbdata->opt->file, "%s%s--- %s%s%s\n",
1178 line_prefix, meta, ecbdata->label_path[0], reset, name_a_tab);
1179 fprintf(ecbdata->opt->file, "%s%s+++ %s%s%s\n",
1180 line_prefix, meta, ecbdata->label_path[1], reset, name_b_tab);
1181 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1182 }
1183
1184 if (diff_suppress_blank_empty
1185 && len == 2 && line[0] == ' ' && line[1] == '\n') {
1186 line[0] = '\n';
1187 len = 1;
1188 }
1189
1190 if (line[0] == '@') {
1191 if (ecbdata->diff_words)
1192 diff_words_flush(ecbdata);
1193 len = sane_truncate_line(ecbdata, line, len);
1194 find_lno(line, ecbdata);
1195 emit_hunk_header(ecbdata, line, len);
1196 if (line[len-1] != '\n')
1197 putc('\n', ecbdata->opt->file);
1198 return;
1199 }
1200
1201 if (len < 1) {
1202 emit_line(ecbdata->opt, reset, reset, line, len);
1203 if (ecbdata->diff_words
1204 && ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN)
1205 fputs("~\n", ecbdata->opt->file);
1206 return;
1207 }
1208
1209 if (ecbdata->diff_words) {
1210 if (line[0] == '-') {
1211 diff_words_append(line, len,
1212 &ecbdata->diff_words->minus);
1213 return;
1214 } else if (line[0] == '+') {
1215 diff_words_append(line, len,
1216 &ecbdata->diff_words->plus);
1217 return;
1218 } else if (!prefixcmp(line, "\\ ")) {
1219 /*
1220 * Eat the "no newline at eof" marker as if we
1221 * saw a "+" or "-" line with nothing on it,
1222 * and return without diff_words_flush() to
1223 * defer processing. If this is the end of
1224 * preimage, more "+" lines may come after it.
1225 */
1226 return;
1227 }
1228 diff_words_flush(ecbdata);
1229 if (ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN) {
1230 emit_line(ecbdata->opt, plain, reset, line, len);
1231 fputs("~\n", ecbdata->opt->file);
1232 } else {
1233 /*
1234 * Skip the prefix character, if any. With
1235 * diff_suppress_blank_empty, there may be
1236 * none.
1237 */
1238 if (line[0] != '\n') {
1239 line++;
1240 len--;
1241 }
1242 emit_line(ecbdata->opt, plain, reset, line, len);
1243 }
1244 return;
1245 }
1246
1247 if (line[0] != '+') {
1248 const char *color =
1249 diff_get_color(ecbdata->color_diff,
1250 line[0] == '-' ? DIFF_FILE_OLD : DIFF_PLAIN);
1251 ecbdata->lno_in_preimage++;
1252 if (line[0] == ' ')
1253 ecbdata->lno_in_postimage++;
1254 emit_line(ecbdata->opt, color, reset, line, len);
1255 } else {
1256 ecbdata->lno_in_postimage++;
1257 emit_add_line(reset, ecbdata, line + 1, len - 1);
1258 }
1259 }
1260
1261 static char *pprint_rename(const char *a, const char *b)
1262 {
1263 const char *old = a;
1264 const char *new = b;
1265 struct strbuf name = STRBUF_INIT;
1266 int pfx_length, sfx_length;
1267 int pfx_adjust_for_slash;
1268 int len_a = strlen(a);
1269 int len_b = strlen(b);
1270 int a_midlen, b_midlen;
1271 int qlen_a = quote_c_style(a, NULL, NULL, 0);
1272 int qlen_b = quote_c_style(b, NULL, NULL, 0);
1273
1274 if (qlen_a || qlen_b) {
1275 quote_c_style(a, &name, NULL, 0);
1276 strbuf_addstr(&name, " => ");
1277 quote_c_style(b, &name, NULL, 0);
1278 return strbuf_detach(&name, NULL);
1279 }
1280
1281 /* Find common prefix */
1282 pfx_length = 0;
1283 while (*old && *new && *old == *new) {
1284 if (*old == '/')
1285 pfx_length = old - a + 1;
1286 old++;
1287 new++;
1288 }
1289
1290 /* Find common suffix */
1291 old = a + len_a;
1292 new = b + len_b;
1293 sfx_length = 0;
1294 /*
1295 * If there is a common prefix, it must end in a slash. In
1296 * that case we let this loop run 1 into the prefix to see the
1297 * same slash.
1298 *
1299 * If there is no common prefix, we cannot do this as it would
1300 * underrun the input strings.
1301 */
1302 pfx_adjust_for_slash = (pfx_length ? 1 : 0);
1303 while (a + pfx_length - pfx_adjust_for_slash <= old &&
1304 b + pfx_length - pfx_adjust_for_slash <= new &&
1305 *old == *new) {
1306 if (*old == '/')
1307 sfx_length = len_a - (old - a);
1308 old--;
1309 new--;
1310 }
1311
1312 /*
1313 * pfx{mid-a => mid-b}sfx
1314 * {pfx-a => pfx-b}sfx
1315 * pfx{sfx-a => sfx-b}
1316 * name-a => name-b
1317 */
1318 a_midlen = len_a - pfx_length - sfx_length;
1319 b_midlen = len_b - pfx_length - sfx_length;
1320 if (a_midlen < 0)
1321 a_midlen = 0;
1322 if (b_midlen < 0)
1323 b_midlen = 0;
1324
1325 strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1326 if (pfx_length + sfx_length) {
1327 strbuf_add(&name, a, pfx_length);
1328 strbuf_addch(&name, '{');
1329 }
1330 strbuf_add(&name, a + pfx_length, a_midlen);
1331 strbuf_addstr(&name, " => ");
1332 strbuf_add(&name, b + pfx_length, b_midlen);
1333 if (pfx_length + sfx_length) {
1334 strbuf_addch(&name, '}');
1335 strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1336 }
1337 return strbuf_detach(&name, NULL);
1338 }
1339
1340 struct diffstat_t {
1341 int nr;
1342 int alloc;
1343 struct diffstat_file {
1344 char *from_name;
1345 char *name;
1346 char *print_name;
1347 unsigned is_unmerged:1;
1348 unsigned is_binary:1;
1349 unsigned is_renamed:1;
1350 unsigned is_interesting:1;
1351 uintmax_t added, deleted;
1352 } **files;
1353 };
1354
1355 static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1356 const char *name_a,
1357 const char *name_b)
1358 {
1359 struct diffstat_file *x;
1360 x = xcalloc(sizeof (*x), 1);
1361 if (diffstat->nr == diffstat->alloc) {
1362 diffstat->alloc = alloc_nr(diffstat->alloc);
1363 diffstat->files = xrealloc(diffstat->files,
1364 diffstat->alloc * sizeof(x));
1365 }
1366 diffstat->files[diffstat->nr++] = x;
1367 if (name_b) {
1368 x->from_name = xstrdup(name_a);
1369 x->name = xstrdup(name_b);
1370 x->is_renamed = 1;
1371 }
1372 else {
1373 x->from_name = NULL;
1374 x->name = xstrdup(name_a);
1375 }
1376 return x;
1377 }
1378
1379 static void diffstat_consume(void *priv, char *line, unsigned long len)
1380 {
1381 struct diffstat_t *diffstat = priv;
1382 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1383
1384 if (line[0] == '+')
1385 x->added++;
1386 else if (line[0] == '-')
1387 x->deleted++;
1388 }
1389
1390 const char mime_boundary_leader[] = "------------";
1391
1392 static int scale_linear(int it, int width, int max_change)
1393 {
1394 if (!it)
1395 return 0;
1396 /*
1397 * make sure that at least one '-' or '+' is printed if
1398 * there is any change to this path. The easiest way is to
1399 * scale linearly as if the alloted width is one column shorter
1400 * than it is, and then add 1 to the result.
1401 */
1402 return 1 + (it * (width - 1) / max_change);
1403 }
1404
1405 static void show_name(FILE *file,
1406 const char *prefix, const char *name, int len)
1407 {
1408 fprintf(file, " %s%-*s |", prefix, len, name);
1409 }
1410
1411 static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1412 {
1413 if (cnt <= 0)
1414 return;
1415 fprintf(file, "%s", set);
1416 while (cnt--)
1417 putc(ch, file);
1418 fprintf(file, "%s", reset);
1419 }
1420
1421 static void fill_print_name(struct diffstat_file *file)
1422 {
1423 char *pname;
1424
1425 if (file->print_name)
1426 return;
1427
1428 if (!file->is_renamed) {
1429 struct strbuf buf = STRBUF_INIT;
1430 if (quote_c_style(file->name, &buf, NULL, 0)) {
1431 pname = strbuf_detach(&buf, NULL);
1432 } else {
1433 pname = file->name;
1434 strbuf_release(&buf);
1435 }
1436 } else {
1437 pname = pprint_rename(file->from_name, file->name);
1438 }
1439 file->print_name = pname;
1440 }
1441
1442 int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1443 {
1444 struct strbuf sb = STRBUF_INIT;
1445 int ret;
1446
1447 if (!files) {
1448 assert(insertions == 0 && deletions == 0);
1449 return fprintf(fp, "%s\n", " 0 files changed");
1450 }
1451
1452 strbuf_addf(&sb,
1453 (files == 1) ? " %d file changed" : " %d files changed",
1454 files);
1455
1456 /*
1457 * For binary diff, the caller may want to print "x files
1458 * changed" with insertions == 0 && deletions == 0.
1459 *
1460 * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1461 * is probably less confusing (i.e skip over "2 files changed
1462 * but nothing about added/removed lines? Is this a bug in Git?").
1463 */
1464 if (insertions || deletions == 0) {
1465 /*
1466 * TRANSLATORS: "+" in (+) is a line addition marker;
1467 * do not translate it.
1468 */
1469 strbuf_addf(&sb,
1470 (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
1471 insertions);
1472 }
1473
1474 if (deletions || insertions == 0) {
1475 /*
1476 * TRANSLATORS: "-" in (-) is a line removal marker;
1477 * do not translate it.
1478 */
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 = add + del;
1687
1688 total = scale_linear(add + del, graph_width, max_change);
1689 if (total < 2 && add && del)
1690 /* width >= 2 due to the sanity check */
1691 total = 2;
1692 if (add < del) {
1693 add = scale_linear(add, graph_width, max_change);
1694 del = total - add;
1695 } else {
1696 del = scale_linear(del, graph_width, max_change);
1697 add = total - del;
1698 }
1699 }
1700 fprintf(options->file, "%s", line_prefix);
1701 show_name(options->file, prefix, name, len);
1702 fprintf(options->file, " %*"PRIuMAX"%s",
1703 number_width, added + deleted,
1704 added + deleted ? " " : "");
1705 show_graph(options->file, '+', add, add_c, reset);
1706 show_graph(options->file, '-', del, del_c, reset);
1707 fprintf(options->file, "\n");
1708 }
1709
1710 for (i = 0; i < data->nr; i++) {
1711 struct diffstat_file *file = data->files[i];
1712 uintmax_t added = file->added;
1713 uintmax_t deleted = file->deleted;
1714
1715 if (file->is_unmerged ||
1716 (!file->is_interesting && (added + deleted == 0))) {
1717 total_files--;
1718 continue;
1719 }
1720
1721 if (!file->is_binary) {
1722 adds += added;
1723 dels += deleted;
1724 }
1725 if (i < count)
1726 continue;
1727 if (!extra_shown)
1728 fprintf(options->file, "%s ...\n", line_prefix);
1729 extra_shown = 1;
1730 }
1731 fprintf(options->file, "%s", line_prefix);
1732 print_stat_summary(options->file, total_files, adds, dels);
1733 }
1734
1735 static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1736 {
1737 int i, adds = 0, dels = 0, total_files = data->nr;
1738
1739 if (data->nr == 0)
1740 return;
1741
1742 for (i = 0; i < data->nr; i++) {
1743 int added = data->files[i]->added;
1744 int deleted= data->files[i]->deleted;
1745
1746 if (data->files[i]->is_unmerged ||
1747 (!data->files[i]->is_interesting && (added + deleted == 0))) {
1748 total_files--;
1749 } else if (!data->files[i]->is_binary) { /* don't count bytes */
1750 adds += added;
1751 dels += deleted;
1752 }
1753 }
1754 fprintf(options->file, "%s", diff_line_prefix(options));
1755 print_stat_summary(options->file, total_files, adds, dels);
1756 }
1757
1758 static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1759 {
1760 int i;
1761
1762 if (data->nr == 0)
1763 return;
1764
1765 for (i = 0; i < data->nr; i++) {
1766 struct diffstat_file *file = data->files[i];
1767
1768 fprintf(options->file, "%s", diff_line_prefix(options));
1769
1770 if (file->is_binary)
1771 fprintf(options->file, "-\t-\t");
1772 else
1773 fprintf(options->file,
1774 "%"PRIuMAX"\t%"PRIuMAX"\t",
1775 file->added, file->deleted);
1776 if (options->line_termination) {
1777 fill_print_name(file);
1778 if (!file->is_renamed)
1779 write_name_quoted(file->name, options->file,
1780 options->line_termination);
1781 else {
1782 fputs(file->print_name, options->file);
1783 putc(options->line_termination, options->file);
1784 }
1785 } else {
1786 if (file->is_renamed) {
1787 putc('\0', options->file);
1788 write_name_quoted(file->from_name, options->file, '\0');
1789 }
1790 write_name_quoted(file->name, options->file, '\0');
1791 }
1792 }
1793 }
1794
1795 struct dirstat_file {
1796 const char *name;
1797 unsigned long changed;
1798 };
1799
1800 struct dirstat_dir {
1801 struct dirstat_file *files;
1802 int alloc, nr, permille, cumulative;
1803 };
1804
1805 static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1806 unsigned long changed, const char *base, int baselen)
1807 {
1808 unsigned long this_dir = 0;
1809 unsigned int sources = 0;
1810 const char *line_prefix = diff_line_prefix(opt);
1811
1812 while (dir->nr) {
1813 struct dirstat_file *f = dir->files;
1814 int namelen = strlen(f->name);
1815 unsigned long this;
1816 char *slash;
1817
1818 if (namelen < baselen)
1819 break;
1820 if (memcmp(f->name, base, baselen))
1821 break;
1822 slash = strchr(f->name + baselen, '/');
1823 if (slash) {
1824 int newbaselen = slash + 1 - f->name;
1825 this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
1826 sources++;
1827 } else {
1828 this = f->changed;
1829 dir->files++;
1830 dir->nr--;
1831 sources += 2;
1832 }
1833 this_dir += this;
1834 }
1835
1836 /*
1837 * We don't report dirstat's for
1838 * - the top level
1839 * - or cases where everything came from a single directory
1840 * under this directory (sources == 1).
1841 */
1842 if (baselen && sources != 1) {
1843 if (this_dir) {
1844 int permille = this_dir * 1000 / changed;
1845 if (permille >= dir->permille) {
1846 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
1847 permille / 10, permille % 10, baselen, base);
1848 if (!dir->cumulative)
1849 return 0;
1850 }
1851 }
1852 }
1853 return this_dir;
1854 }
1855
1856 static int dirstat_compare(const void *_a, const void *_b)
1857 {
1858 const struct dirstat_file *a = _a;
1859 const struct dirstat_file *b = _b;
1860 return strcmp(a->name, b->name);
1861 }
1862
1863 static void show_dirstat(struct diff_options *options)
1864 {
1865 int i;
1866 unsigned long changed;
1867 struct dirstat_dir dir;
1868 struct diff_queue_struct *q = &diff_queued_diff;
1869
1870 dir.files = NULL;
1871 dir.alloc = 0;
1872 dir.nr = 0;
1873 dir.permille = options->dirstat_permille;
1874 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1875
1876 changed = 0;
1877 for (i = 0; i < q->nr; i++) {
1878 struct diff_filepair *p = q->queue[i];
1879 const char *name;
1880 unsigned long copied, added, damage;
1881 int content_changed;
1882
1883 name = p->two->path ? p->two->path : p->one->path;
1884
1885 if (p->one->sha1_valid && p->two->sha1_valid)
1886 content_changed = hashcmp(p->one->sha1, p->two->sha1);
1887 else
1888 content_changed = 1;
1889
1890 if (!content_changed) {
1891 /*
1892 * The SHA1 has not changed, so pre-/post-content is
1893 * identical. We can therefore skip looking at the
1894 * file contents altogether.
1895 */
1896 damage = 0;
1897 goto found_damage;
1898 }
1899
1900 if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
1901 /*
1902 * In --dirstat-by-file mode, we don't really need to
1903 * look at the actual file contents at all.
1904 * The fact that the SHA1 changed is enough for us to
1905 * add this file to the list of results
1906 * (with each file contributing equal damage).
1907 */
1908 damage = 1;
1909 goto found_damage;
1910 }
1911
1912 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
1913 diff_populate_filespec(p->one, 0);
1914 diff_populate_filespec(p->two, 0);
1915 diffcore_count_changes(p->one, p->two, NULL, NULL, 0,
1916 &copied, &added);
1917 diff_free_filespec_data(p->one);
1918 diff_free_filespec_data(p->two);
1919 } else if (DIFF_FILE_VALID(p->one)) {
1920 diff_populate_filespec(p->one, 1);
1921 copied = added = 0;
1922 diff_free_filespec_data(p->one);
1923 } else if (DIFF_FILE_VALID(p->two)) {
1924 diff_populate_filespec(p->two, 1);
1925 copied = 0;
1926 added = p->two->size;
1927 diff_free_filespec_data(p->two);
1928 } else
1929 continue;
1930
1931 /*
1932 * Original minus copied is the removed material,
1933 * added is the new material. They are both damages
1934 * made to the preimage.
1935 * If the resulting damage is zero, we know that
1936 * diffcore_count_changes() considers the two entries to
1937 * be identical, but since content_changed is true, we
1938 * know that there must have been _some_ kind of change,
1939 * so we force all entries to have damage > 0.
1940 */
1941 damage = (p->one->size - copied) + added;
1942 if (!damage)
1943 damage = 1;
1944
1945 found_damage:
1946 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1947 dir.files[dir.nr].name = name;
1948 dir.files[dir.nr].changed = damage;
1949 changed += damage;
1950 dir.nr++;
1951 }
1952
1953 /* This can happen even with many files, if everything was renames */
1954 if (!changed)
1955 return;
1956
1957 /* Show all directories with more than x% of the changes */
1958 qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1959 gather_dirstat(options, &dir, changed, "", 0);
1960 }
1961
1962 static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
1963 {
1964 int i;
1965 unsigned long changed;
1966 struct dirstat_dir dir;
1967
1968 if (data->nr == 0)
1969 return;
1970
1971 dir.files = NULL;
1972 dir.alloc = 0;
1973 dir.nr = 0;
1974 dir.permille = options->dirstat_permille;
1975 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1976
1977 changed = 0;
1978 for (i = 0; i < data->nr; i++) {
1979 struct diffstat_file *file = data->files[i];
1980 unsigned long damage = file->added + file->deleted;
1981 if (file->is_binary)
1982 /*
1983 * binary files counts bytes, not lines. Must find some
1984 * way to normalize binary bytes vs. textual lines.
1985 * The following heuristic assumes that there are 64
1986 * bytes per "line".
1987 * This is stupid and ugly, but very cheap...
1988 */
1989 damage = (damage + 63) / 64;
1990 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1991 dir.files[dir.nr].name = file->name;
1992 dir.files[dir.nr].changed = damage;
1993 changed += damage;
1994 dir.nr++;
1995 }
1996
1997 /* This can happen even with many files, if everything was renames */
1998 if (!changed)
1999 return;
2000
2001 /* Show all directories with more than x% of the changes */
2002 qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
2003 gather_dirstat(options, &dir, changed, "", 0);
2004 }
2005
2006 static void free_diffstat_info(struct diffstat_t *diffstat)
2007 {
2008 int i;
2009 for (i = 0; i < diffstat->nr; i++) {
2010 struct diffstat_file *f = diffstat->files[i];
2011 if (f->name != f->print_name)
2012 free(f->print_name);
2013 free(f->name);
2014 free(f->from_name);
2015 free(f);
2016 }
2017 free(diffstat->files);
2018 }
2019
2020 struct checkdiff_t {
2021 const char *filename;
2022 int lineno;
2023 int conflict_marker_size;
2024 struct diff_options *o;
2025 unsigned ws_rule;
2026 unsigned status;
2027 };
2028
2029 static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2030 {
2031 char firstchar;
2032 int cnt;
2033
2034 if (len < marker_size + 1)
2035 return 0;
2036 firstchar = line[0];
2037 switch (firstchar) {
2038 case '=': case '>': case '<': case '|':
2039 break;
2040 default:
2041 return 0;
2042 }
2043 for (cnt = 1; cnt < marker_size; cnt++)
2044 if (line[cnt] != firstchar)
2045 return 0;
2046 /* line[1] thru line[marker_size-1] are same as firstchar */
2047 if (len < marker_size + 1 || !isspace(line[marker_size]))
2048 return 0;
2049 return 1;
2050 }
2051
2052 static void checkdiff_consume(void *priv, char *line, unsigned long len)
2053 {
2054 struct checkdiff_t *data = priv;
2055 int marker_size = data->conflict_marker_size;
2056 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2057 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2058 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2059 char *err;
2060 const char *line_prefix;
2061
2062 assert(data->o);
2063 line_prefix = diff_line_prefix(data->o);
2064
2065 if (line[0] == '+') {
2066 unsigned bad;
2067 data->lineno++;
2068 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2069 data->status |= 1;
2070 fprintf(data->o->file,
2071 "%s%s:%d: leftover conflict marker\n",
2072 line_prefix, data->filename, data->lineno);
2073 }
2074 bad = ws_check(line + 1, len - 1, data->ws_rule);
2075 if (!bad)
2076 return;
2077 data->status |= bad;
2078 err = whitespace_error_string(bad);
2079 fprintf(data->o->file, "%s%s:%d: %s.\n",
2080 line_prefix, data->filename, data->lineno, err);
2081 free(err);
2082 emit_line(data->o, set, reset, line, 1);
2083 ws_check_emit(line + 1, len - 1, data->ws_rule,
2084 data->o->file, set, reset, ws);
2085 } else if (line[0] == ' ') {
2086 data->lineno++;
2087 } else if (line[0] == '@') {
2088 char *plus = strchr(line, '+');
2089 if (plus)
2090 data->lineno = strtol(plus, NULL, 10) - 1;
2091 else
2092 die("invalid diff");
2093 }
2094 }
2095
2096 static unsigned char *deflate_it(char *data,
2097 unsigned long size,
2098 unsigned long *result_size)
2099 {
2100 int bound;
2101 unsigned char *deflated;
2102 git_zstream stream;
2103
2104 memset(&stream, 0, sizeof(stream));
2105 git_deflate_init(&stream, zlib_compression_level);
2106 bound = git_deflate_bound(&stream, size);
2107 deflated = xmalloc(bound);
2108 stream.next_out = deflated;
2109 stream.avail_out = bound;
2110
2111 stream.next_in = (unsigned char *)data;
2112 stream.avail_in = size;
2113 while (git_deflate(&stream, Z_FINISH) == Z_OK)
2114 ; /* nothing */
2115 git_deflate_end(&stream);
2116 *result_size = stream.total_out;
2117 return deflated;
2118 }
2119
2120 static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two,
2121 const char *prefix)
2122 {
2123 void *cp;
2124 void *delta;
2125 void *deflated;
2126 void *data;
2127 unsigned long orig_size;
2128 unsigned long delta_size;
2129 unsigned long deflate_size;
2130 unsigned long data_size;
2131
2132 /* We could do deflated delta, or we could do just deflated two,
2133 * whichever is smaller.
2134 */
2135 delta = NULL;
2136 deflated = deflate_it(two->ptr, two->size, &deflate_size);
2137 if (one->size && two->size) {
2138 delta = diff_delta(one->ptr, one->size,
2139 two->ptr, two->size,
2140 &delta_size, deflate_size);
2141 if (delta) {
2142 void *to_free = delta;
2143 orig_size = delta_size;
2144 delta = deflate_it(delta, delta_size, &delta_size);
2145 free(to_free);
2146 }
2147 }
2148
2149 if (delta && delta_size < deflate_size) {
2150 fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2151 free(deflated);
2152 data = delta;
2153 data_size = delta_size;
2154 }
2155 else {
2156 fprintf(file, "%sliteral %lu\n", prefix, two->size);
2157 free(delta);
2158 data = deflated;
2159 data_size = deflate_size;
2160 }
2161
2162 /* emit data encoded in base85 */
2163 cp = data;
2164 while (data_size) {
2165 int bytes = (52 < data_size) ? 52 : data_size;
2166 char line[70];
2167 data_size -= bytes;
2168 if (bytes <= 26)
2169 line[0] = bytes + 'A' - 1;
2170 else
2171 line[0] = bytes - 26 + 'a' - 1;
2172 encode_85(line + 1, cp, bytes);
2173 cp = (char *) cp + bytes;
2174 fprintf(file, "%s", prefix);
2175 fputs(line, file);
2176 fputc('\n', file);
2177 }
2178 fprintf(file, "%s\n", prefix);
2179 free(data);
2180 }
2181
2182 static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two,
2183 const char *prefix)
2184 {
2185 fprintf(file, "%sGIT binary patch\n", prefix);
2186 emit_binary_diff_body(file, one, two, prefix);
2187 emit_binary_diff_body(file, two, one, prefix);
2188 }
2189
2190 int diff_filespec_is_binary(struct diff_filespec *one)
2191 {
2192 if (one->is_binary == -1) {
2193 diff_filespec_load_driver(one);
2194 if (one->driver->binary != -1)
2195 one->is_binary = one->driver->binary;
2196 else {
2197 if (!one->data && DIFF_FILE_VALID(one))
2198 diff_populate_filespec(one, 0);
2199 if (one->data)
2200 one->is_binary = buffer_is_binary(one->data,
2201 one->size);
2202 if (one->is_binary == -1)
2203 one->is_binary = 0;
2204 }
2205 }
2206 return one->is_binary;
2207 }
2208
2209 static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2210 {
2211 diff_filespec_load_driver(one);
2212 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2213 }
2214
2215 void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2216 {
2217 if (!options->a_prefix)
2218 options->a_prefix = a;
2219 if (!options->b_prefix)
2220 options->b_prefix = b;
2221 }
2222
2223 struct userdiff_driver *get_textconv(struct diff_filespec *one)
2224 {
2225 if (!DIFF_FILE_VALID(one))
2226 return NULL;
2227
2228 diff_filespec_load_driver(one);
2229 return userdiff_get_textconv(one->driver);
2230 }
2231
2232 static void builtin_diff(const char *name_a,
2233 const char *name_b,
2234 struct diff_filespec *one,
2235 struct diff_filespec *two,
2236 const char *xfrm_msg,
2237 int must_show_header,
2238 struct diff_options *o,
2239 int complete_rewrite)
2240 {
2241 mmfile_t mf1, mf2;
2242 const char *lbl[2];
2243 char *a_one, *b_two;
2244 const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
2245 const char *reset = diff_get_color_opt(o, DIFF_RESET);
2246 const char *a_prefix, *b_prefix;
2247 struct userdiff_driver *textconv_one = NULL;
2248 struct userdiff_driver *textconv_two = NULL;
2249 struct strbuf header = STRBUF_INIT;
2250 const char *line_prefix = diff_line_prefix(o);
2251
2252 if (DIFF_OPT_TST(o, SUBMODULE_LOG) &&
2253 (!one->mode || S_ISGITLINK(one->mode)) &&
2254 (!two->mode || S_ISGITLINK(two->mode))) {
2255 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2256 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2257 show_submodule_summary(o->file, one ? one->path : two->path,
2258 line_prefix,
2259 one->sha1, two->sha1, two->dirty_submodule,
2260 meta, del, add, reset);
2261 return;
2262 }
2263
2264 if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2265 textconv_one = get_textconv(one);
2266 textconv_two = get_textconv(two);
2267 }
2268
2269 diff_set_mnemonic_prefix(o, "a/", "b/");
2270 if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2271 a_prefix = o->b_prefix;
2272 b_prefix = o->a_prefix;
2273 } else {
2274 a_prefix = o->a_prefix;
2275 b_prefix = o->b_prefix;
2276 }
2277
2278 /* Never use a non-valid filename anywhere if at all possible */
2279 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2280 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2281
2282 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2283 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2284 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2285 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2286 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
2287 if (lbl[0][0] == '/') {
2288 /* /dev/null */
2289 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
2290 if (xfrm_msg)
2291 strbuf_addstr(&header, xfrm_msg);
2292 must_show_header = 1;
2293 }
2294 else if (lbl[1][0] == '/') {
2295 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
2296 if (xfrm_msg)
2297 strbuf_addstr(&header, xfrm_msg);
2298 must_show_header = 1;
2299 }
2300 else {
2301 if (one->mode != two->mode) {
2302 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
2303 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
2304 must_show_header = 1;
2305 }
2306 if (xfrm_msg)
2307 strbuf_addstr(&header, xfrm_msg);
2308
2309 /*
2310 * we do not run diff between different kind
2311 * of objects.
2312 */
2313 if ((one->mode ^ two->mode) & S_IFMT)
2314 goto free_ab_and_return;
2315 if (complete_rewrite &&
2316 (textconv_one || !diff_filespec_is_binary(one)) &&
2317 (textconv_two || !diff_filespec_is_binary(two))) {
2318 fprintf(o->file, "%s", header.buf);
2319 strbuf_reset(&header);
2320 emit_rewrite_diff(name_a, name_b, one, two,
2321 textconv_one, textconv_two, o);
2322 o->found_changes = 1;
2323 goto free_ab_and_return;
2324 }
2325 }
2326
2327 if (o->irreversible_delete && lbl[1][0] == '/') {
2328 fprintf(o->file, "%s", header.buf);
2329 strbuf_reset(&header);
2330 goto free_ab_and_return;
2331 } else if (!DIFF_OPT_TST(o, TEXT) &&
2332 ( (!textconv_one && diff_filespec_is_binary(one)) ||
2333 (!textconv_two && diff_filespec_is_binary(two)) )) {
2334 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2335 die("unable to read files to diff");
2336 /* Quite common confusing case */
2337 if (mf1.size == mf2.size &&
2338 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2339 if (must_show_header)
2340 fprintf(o->file, "%s", header.buf);
2341 goto free_ab_and_return;
2342 }
2343 fprintf(o->file, "%s", header.buf);
2344 strbuf_reset(&header);
2345 if (DIFF_OPT_TST(o, BINARY))
2346 emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2347 else
2348 fprintf(o->file, "%sBinary files %s and %s differ\n",
2349 line_prefix, lbl[0], lbl[1]);
2350 o->found_changes = 1;
2351 } else {
2352 /* Crazy xdl interfaces.. */
2353 const char *diffopts = getenv("GIT_DIFF_OPTS");
2354 xpparam_t xpp;
2355 xdemitconf_t xecfg;
2356 struct emit_callback ecbdata;
2357 const struct userdiff_funcname *pe;
2358
2359 if (must_show_header) {
2360 fprintf(o->file, "%s", header.buf);
2361 strbuf_reset(&header);
2362 }
2363
2364 mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2365 mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2366
2367 pe = diff_funcname_pattern(one);
2368 if (!pe)
2369 pe = diff_funcname_pattern(two);
2370
2371 memset(&xpp, 0, sizeof(xpp));
2372 memset(&xecfg, 0, sizeof(xecfg));
2373 memset(&ecbdata, 0, sizeof(ecbdata));
2374 ecbdata.label_path = lbl;
2375 ecbdata.color_diff = want_color(o->use_color);
2376 ecbdata.found_changesp = &o->found_changes;
2377 ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
2378 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2379 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2380 ecbdata.opt = o;
2381 ecbdata.header = header.len ? &header : NULL;
2382 xpp.flags = o->xdl_opts;
2383 xecfg.ctxlen = o->context;
2384 xecfg.interhunkctxlen = o->interhunkcontext;
2385 xecfg.flags = XDL_EMIT_FUNCNAMES;
2386 if (DIFF_OPT_TST(o, FUNCCONTEXT))
2387 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2388 if (pe)
2389 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2390 if (!diffopts)
2391 ;
2392 else if (!prefixcmp(diffopts, "--unified="))
2393 xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
2394 else if (!prefixcmp(diffopts, "-u"))
2395 xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
2396 if (o->word_diff)
2397 init_diff_words_data(&ecbdata, o, one, two);
2398 xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2399 &xpp, &xecfg);
2400 if (o->word_diff)
2401 free_diff_words_data(&ecbdata);
2402 if (textconv_one)
2403 free(mf1.ptr);
2404 if (textconv_two)
2405 free(mf2.ptr);
2406 xdiff_clear_find_func(&xecfg);
2407 }
2408
2409 free_ab_and_return:
2410 strbuf_release(&header);
2411 diff_free_filespec_data(one);
2412 diff_free_filespec_data(two);
2413 free(a_one);
2414 free(b_two);
2415 return;
2416 }
2417
2418 static void builtin_diffstat(const char *name_a, const char *name_b,
2419 struct diff_filespec *one,
2420 struct diff_filespec *two,
2421 struct diffstat_t *diffstat,
2422 struct diff_options *o,
2423 struct diff_filepair *p)
2424 {
2425 mmfile_t mf1, mf2;
2426 struct diffstat_file *data;
2427 int same_contents;
2428 int complete_rewrite = 0;
2429
2430 if (!DIFF_PAIR_UNMERGED(p)) {
2431 if (p->status == DIFF_STATUS_MODIFIED && p->score)
2432 complete_rewrite = 1;
2433 }
2434
2435 data = diffstat_add(diffstat, name_a, name_b);
2436 data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
2437
2438 if (!one || !two) {
2439 data->is_unmerged = 1;
2440 return;
2441 }
2442
2443 same_contents = !hashcmp(one->sha1, two->sha1);
2444
2445 if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2446 data->is_binary = 1;
2447 if (same_contents) {
2448 data->added = 0;
2449 data->deleted = 0;
2450 } else {
2451 data->added = diff_filespec_size(two);
2452 data->deleted = diff_filespec_size(one);
2453 }
2454 }
2455
2456 else if (complete_rewrite) {
2457 diff_populate_filespec(one, 0);
2458 diff_populate_filespec(two, 0);
2459 data->deleted = count_lines(one->data, one->size);
2460 data->added = count_lines(two->data, two->size);
2461 }
2462
2463 else if (!same_contents) {
2464 /* Crazy xdl interfaces.. */
2465 xpparam_t xpp;
2466 xdemitconf_t xecfg;
2467
2468 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2469 die("unable to read files to diff");
2470
2471 memset(&xpp, 0, sizeof(xpp));
2472 memset(&xecfg, 0, sizeof(xecfg));
2473 xpp.flags = o->xdl_opts;
2474 xecfg.ctxlen = o->context;
2475 xecfg.interhunkctxlen = o->interhunkcontext;
2476 xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2477 &xpp, &xecfg);
2478 }
2479
2480 diff_free_filespec_data(one);
2481 diff_free_filespec_data(two);
2482 }
2483
2484 static void builtin_checkdiff(const char *name_a, const char *name_b,
2485 const char *attr_path,
2486 struct diff_filespec *one,
2487 struct diff_filespec *two,
2488 struct diff_options *o)
2489 {
2490 mmfile_t mf1, mf2;
2491 struct checkdiff_t data;
2492
2493 if (!two)
2494 return;
2495
2496 memset(&data, 0, sizeof(data));
2497 data.filename = name_b ? name_b : name_a;
2498 data.lineno = 0;
2499 data.o = o;
2500 data.ws_rule = whitespace_rule(attr_path);
2501 data.conflict_marker_size = ll_merge_marker_size(attr_path);
2502
2503 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2504 die("unable to read files to diff");
2505
2506 /*
2507 * All the other codepaths check both sides, but not checking
2508 * the "old" side here is deliberate. We are checking the newly
2509 * introduced changes, and as long as the "new" side is text, we
2510 * can and should check what it introduces.
2511 */
2512 if (diff_filespec_is_binary(two))
2513 goto free_and_return;
2514 else {
2515 /* Crazy xdl interfaces.. */
2516 xpparam_t xpp;
2517 xdemitconf_t xecfg;
2518
2519 memset(&xpp, 0, sizeof(xpp));
2520 memset(&xecfg, 0, sizeof(xecfg));
2521 xecfg.ctxlen = 1; /* at least one context line */
2522 xpp.flags = 0;
2523 xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2524 &xpp, &xecfg);
2525
2526 if (data.ws_rule & WS_BLANK_AT_EOF) {
2527 struct emit_callback ecbdata;
2528 int blank_at_eof;
2529
2530 ecbdata.ws_rule = data.ws_rule;
2531 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2532 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2533
2534 if (blank_at_eof) {
2535 static char *err;
2536 if (!err)
2537 err = whitespace_error_string(WS_BLANK_AT_EOF);
2538 fprintf(o->file, "%s:%d: %s.\n",
2539 data.filename, blank_at_eof, err);
2540 data.status = 1; /* report errors */
2541 }
2542 }
2543 }
2544 free_and_return:
2545 diff_free_filespec_data(one);
2546 diff_free_filespec_data(two);
2547 if (data.status)
2548 DIFF_OPT_SET(o, CHECK_FAILED);
2549 }
2550
2551 struct diff_filespec *alloc_filespec(const char *path)
2552 {
2553 int namelen = strlen(path);
2554 struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
2555
2556 memset(spec, 0, sizeof(*spec));
2557 spec->path = (char *)(spec + 1);
2558 memcpy(spec->path, path, namelen+1);
2559 spec->count = 1;
2560 spec->is_binary = -1;
2561 return spec;
2562 }
2563
2564 void free_filespec(struct diff_filespec *spec)
2565 {
2566 if (!--spec->count) {
2567 diff_free_filespec_data(spec);
2568 free(spec);
2569 }
2570 }
2571
2572 void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
2573 int sha1_valid, unsigned short mode)
2574 {
2575 if (mode) {
2576 spec->mode = canon_mode(mode);
2577 hashcpy(spec->sha1, sha1);
2578 spec->sha1_valid = sha1_valid;
2579 }
2580 }
2581
2582 /*
2583 * Given a name and sha1 pair, if the index tells us the file in
2584 * the work tree has that object contents, return true, so that
2585 * prepare_temp_file() does not have to inflate and extract.
2586 */
2587 static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
2588 {
2589 struct cache_entry *ce;
2590 struct stat st;
2591 int pos, len;
2592
2593 /*
2594 * We do not read the cache ourselves here, because the
2595 * benchmark with my previous version that always reads cache
2596 * shows that it makes things worse for diff-tree comparing
2597 * two linux-2.6 kernel trees in an already checked out work
2598 * tree. This is because most diff-tree comparisons deal with
2599 * only a small number of files, while reading the cache is
2600 * expensive for a large project, and its cost outweighs the
2601 * savings we get by not inflating the object to a temporary
2602 * file. Practically, this code only helps when we are used
2603 * by diff-cache --cached, which does read the cache before
2604 * calling us.
2605 */
2606 if (!active_cache)
2607 return 0;
2608
2609 /* We want to avoid the working directory if our caller
2610 * doesn't need the data in a normal file, this system
2611 * is rather slow with its stat/open/mmap/close syscalls,
2612 * and the object is contained in a pack file. The pack
2613 * is probably already open and will be faster to obtain
2614 * the data through than the working directory. Loose
2615 * objects however would tend to be slower as they need
2616 * to be individually opened and inflated.
2617 */
2618 if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1))
2619 return 0;
2620
2621 len = strlen(name);
2622 pos = cache_name_pos(name, len);
2623 if (pos < 0)
2624 return 0;
2625 ce = active_cache[pos];
2626
2627 /*
2628 * This is not the sha1 we are looking for, or
2629 * unreusable because it is not a regular file.
2630 */
2631 if (hashcmp(sha1, ce->sha1) || !S_ISREG(ce->ce_mode))
2632 return 0;
2633
2634 /*
2635 * If ce is marked as "assume unchanged", there is no
2636 * guarantee that work tree matches what we are looking for.
2637 */
2638 if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2639 return 0;
2640
2641 /*
2642 * If ce matches the file in the work tree, we can reuse it.
2643 */
2644 if (ce_uptodate(ce) ||
2645 (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2646 return 1;
2647
2648 return 0;
2649 }
2650
2651 static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2652 {
2653 int len;
2654 char *data = xmalloc(100), *dirty = "";
2655
2656 /* Are we looking at the work tree? */
2657 if (s->dirty_submodule)
2658 dirty = "-dirty";
2659
2660 len = snprintf(data, 100,
2661 "Subproject commit %s%s\n", sha1_to_hex(s->sha1), dirty);
2662 s->data = data;
2663 s->size = len;
2664 s->should_free = 1;
2665 if (size_only) {
2666 s->data = NULL;
2667 free(data);
2668 }
2669 return 0;
2670 }
2671
2672 /*
2673 * While doing rename detection and pickaxe operation, we may need to
2674 * grab the data for the blob (or file) for our own in-core comparison.
2675 * diff_filespec has data and size fields for this purpose.
2676 */
2677 int diff_populate_filespec(struct diff_filespec *s, int size_only)
2678 {
2679 int err = 0;
2680 /*
2681 * demote FAIL to WARN to allow inspecting the situation
2682 * instead of refusing.
2683 */
2684 enum safe_crlf crlf_warn = (safe_crlf == SAFE_CRLF_FAIL
2685 ? SAFE_CRLF_WARN
2686 : safe_crlf);
2687
2688 if (!DIFF_FILE_VALID(s))
2689 die("internal error: asking to populate invalid file.");
2690 if (S_ISDIR(s->mode))
2691 return -1;
2692
2693 if (s->data)
2694 return 0;
2695
2696 if (size_only && 0 < s->size)
2697 return 0;
2698
2699 if (S_ISGITLINK(s->mode))
2700 return diff_populate_gitlink(s, size_only);
2701
2702 if (!s->sha1_valid ||
2703 reuse_worktree_file(s->path, s->sha1, 0)) {
2704 struct strbuf buf = STRBUF_INIT;
2705 struct stat st;
2706 int fd;
2707
2708 if (lstat(s->path, &st) < 0) {
2709 if (errno == ENOENT) {
2710 err_empty:
2711 err = -1;
2712 empty:
2713 s->data = (char *)"";
2714 s->size = 0;
2715 return err;
2716 }
2717 }
2718 s->size = xsize_t(st.st_size);
2719 if (!s->size)
2720 goto empty;
2721 if (S_ISLNK(st.st_mode)) {
2722 struct strbuf sb = STRBUF_INIT;
2723
2724 if (strbuf_readlink(&sb, s->path, s->size))
2725 goto err_empty;
2726 s->size = sb.len;
2727 s->data = strbuf_detach(&sb, NULL);
2728 s->should_free = 1;
2729 return 0;
2730 }
2731 if (size_only)
2732 return 0;
2733 fd = open(s->path, O_RDONLY);
2734 if (fd < 0)
2735 goto err_empty;
2736 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2737 close(fd);
2738 s->should_munmap = 1;
2739
2740 /*
2741 * Convert from working tree format to canonical git format
2742 */
2743 if (convert_to_git(s->path, s->data, s->size, &buf, crlf_warn)) {
2744 size_t size = 0;
2745 munmap(s->data, s->size);
2746 s->should_munmap = 0;
2747 s->data = strbuf_detach(&buf, &size);
2748 s->size = size;
2749 s->should_free = 1;
2750 }
2751 }
2752 else {
2753 enum object_type type;
2754 if (size_only) {
2755 type = sha1_object_info(s->sha1, &s->size);
2756 if (type < 0)
2757 die("unable to read %s", sha1_to_hex(s->sha1));
2758 } else {
2759 s->data = read_sha1_file(s->sha1, &type, &s->size);
2760 if (!s->data)
2761 die("unable to read %s", sha1_to_hex(s->sha1));
2762 s->should_free = 1;
2763 }
2764 }
2765 return 0;
2766 }
2767
2768 void diff_free_filespec_blob(struct diff_filespec *s)
2769 {
2770 if (s->should_free)
2771 free(s->data);
2772 else if (s->should_munmap)
2773 munmap(s->data, s->size);
2774
2775 if (s->should_free || s->should_munmap) {
2776 s->should_free = s->should_munmap = 0;
2777 s->data = NULL;
2778 }
2779 }
2780
2781 void diff_free_filespec_data(struct diff_filespec *s)
2782 {
2783 diff_free_filespec_blob(s);
2784 free(s->cnt_data);
2785 s->cnt_data = NULL;
2786 }
2787
2788 static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2789 void *blob,
2790 unsigned long size,
2791 const unsigned char *sha1,
2792 int mode)
2793 {
2794 int fd;
2795 struct strbuf buf = STRBUF_INIT;
2796 struct strbuf template = STRBUF_INIT;
2797 char *path_dup = xstrdup(path);
2798 const char *base = basename(path_dup);
2799
2800 /* Generate "XXXXXX_basename.ext" */
2801 strbuf_addstr(&template, "XXXXXX_");
2802 strbuf_addstr(&template, base);
2803
2804 fd = git_mkstemps(temp->tmp_path, PATH_MAX, template.buf,
2805 strlen(base) + 1);
2806 if (fd < 0)
2807 die_errno("unable to create temp-file");
2808 if (convert_to_working_tree(path,
2809 (const char *)blob, (size_t)size, &buf)) {
2810 blob = buf.buf;
2811 size = buf.len;
2812 }
2813 if (write_in_full(fd, blob, size) != size)
2814 die_errno("unable to write temp-file");
2815 close(fd);
2816 temp->name = temp->tmp_path;
2817 strcpy(temp->hex, sha1_to_hex(sha1));
2818 temp->hex[40] = 0;
2819 sprintf(temp->mode, "%06o", mode);
2820 strbuf_release(&buf);
2821 strbuf_release(&template);
2822 free(path_dup);
2823 }
2824
2825 static struct diff_tempfile *prepare_temp_file(const char *name,
2826 struct diff_filespec *one)
2827 {
2828 struct diff_tempfile *temp = claim_diff_tempfile();
2829
2830 if (!DIFF_FILE_VALID(one)) {
2831 not_a_valid_file:
2832 /* A '-' entry produces this for file-2, and
2833 * a '+' entry produces this for file-1.
2834 */
2835 temp->name = "/dev/null";
2836 strcpy(temp->hex, ".");
2837 strcpy(temp->mode, ".");
2838 return temp;
2839 }
2840
2841 if (!remove_tempfile_installed) {
2842 atexit(remove_tempfile);
2843 sigchain_push_common(remove_tempfile_on_signal);
2844 remove_tempfile_installed = 1;
2845 }
2846
2847 if (!one->sha1_valid ||
2848 reuse_worktree_file(name, one->sha1, 1)) {
2849 struct stat st;
2850 if (lstat(name, &st) < 0) {
2851 if (errno == ENOENT)
2852 goto not_a_valid_file;
2853 die_errno("stat(%s)", name);
2854 }
2855 if (S_ISLNK(st.st_mode)) {
2856 struct strbuf sb = STRBUF_INIT;
2857 if (strbuf_readlink(&sb, name, st.st_size) < 0)
2858 die_errno("readlink(%s)", name);
2859 prep_temp_blob(name, temp, sb.buf, sb.len,
2860 (one->sha1_valid ?
2861 one->sha1 : null_sha1),
2862 (one->sha1_valid ?
2863 one->mode : S_IFLNK));
2864 strbuf_release(&sb);
2865 }
2866 else {
2867 /* we can borrow from the file in the work tree */
2868 temp->name = name;
2869 if (!one->sha1_valid)
2870 strcpy(temp->hex, sha1_to_hex(null_sha1));
2871 else
2872 strcpy(temp->hex, sha1_to_hex(one->sha1));
2873 /* Even though we may sometimes borrow the
2874 * contents from the work tree, we always want
2875 * one->mode. mode is trustworthy even when
2876 * !(one->sha1_valid), as long as
2877 * DIFF_FILE_VALID(one).
2878 */
2879 sprintf(temp->mode, "%06o", one->mode);
2880 }
2881 return temp;
2882 }
2883 else {
2884 if (diff_populate_filespec(one, 0))
2885 die("cannot read data blob for %s", one->path);
2886 prep_temp_blob(name, temp, one->data, one->size,
2887 one->sha1, one->mode);
2888 }
2889 return temp;
2890 }
2891
2892 /* An external diff command takes:
2893 *
2894 * diff-cmd name infile1 infile1-sha1 infile1-mode \
2895 * infile2 infile2-sha1 infile2-mode [ rename-to ]
2896 *
2897 */
2898 static void run_external_diff(const char *pgm,
2899 const char *name,
2900 const char *other,
2901 struct diff_filespec *one,
2902 struct diff_filespec *two,
2903 const char *xfrm_msg,
2904 int complete_rewrite)
2905 {
2906 const char *spawn_arg[10];
2907 int retval;
2908 const char **arg = &spawn_arg[0];
2909
2910 if (one && two) {
2911 struct diff_tempfile *temp_one, *temp_two;
2912 const char *othername = (other ? other : name);
2913 temp_one = prepare_temp_file(name, one);
2914 temp_two = prepare_temp_file(othername, two);
2915 *arg++ = pgm;
2916 *arg++ = name;
2917 *arg++ = temp_one->name;
2918 *arg++ = temp_one->hex;
2919 *arg++ = temp_one->mode;
2920 *arg++ = temp_two->name;
2921 *arg++ = temp_two->hex;
2922 *arg++ = temp_two->mode;
2923 if (other) {
2924 *arg++ = other;
2925 *arg++ = xfrm_msg;
2926 }
2927 } else {
2928 *arg++ = pgm;
2929 *arg++ = name;
2930 }
2931 *arg = NULL;
2932 fflush(NULL);
2933 retval = run_command_v_opt(spawn_arg, RUN_USING_SHELL);
2934 remove_tempfile();
2935 if (retval) {
2936 fprintf(stderr, "external diff died, stopping at %s.\n", name);
2937 exit(1);
2938 }
2939 }
2940
2941 static int similarity_index(struct diff_filepair *p)
2942 {
2943 return p->score * 100 / MAX_SCORE;
2944 }
2945
2946 static void fill_metainfo(struct strbuf *msg,
2947 const char *name,
2948 const char *other,
2949 struct diff_filespec *one,
2950 struct diff_filespec *two,
2951 struct diff_options *o,
2952 struct diff_filepair *p,
2953 int *must_show_header,
2954 int use_color)
2955 {
2956 const char *set = diff_get_color(use_color, DIFF_METAINFO);
2957 const char *reset = diff_get_color(use_color, DIFF_RESET);
2958 const char *line_prefix = diff_line_prefix(o);
2959
2960 *must_show_header = 1;
2961 strbuf_init(msg, PATH_MAX * 2 + 300);
2962 switch (p->status) {
2963 case DIFF_STATUS_COPIED:
2964 strbuf_addf(msg, "%s%ssimilarity index %d%%",
2965 line_prefix, set, similarity_index(p));
2966 strbuf_addf(msg, "%s\n%s%scopy from ",
2967 reset, line_prefix, set);
2968 quote_c_style(name, msg, NULL, 0);
2969 strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
2970 quote_c_style(other, msg, NULL, 0);
2971 strbuf_addf(msg, "%s\n", reset);
2972 break;
2973 case DIFF_STATUS_RENAMED:
2974 strbuf_addf(msg, "%s%ssimilarity index %d%%",
2975 line_prefix, set, similarity_index(p));
2976 strbuf_addf(msg, "%s\n%s%srename from ",
2977 reset, line_prefix, set);
2978 quote_c_style(name, msg, NULL, 0);
2979 strbuf_addf(msg, "%s\n%s%srename to ",
2980 reset, line_prefix, set);
2981 quote_c_style(other, msg, NULL, 0);
2982 strbuf_addf(msg, "%s\n", reset);
2983 break;
2984 case DIFF_STATUS_MODIFIED:
2985 if (p->score) {
2986 strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
2987 line_prefix,
2988 set, similarity_index(p), reset);
2989 break;
2990 }
2991 /* fallthru */
2992 default:
2993 *must_show_header = 0;
2994 }
2995 if (one && two && hashcmp(one->sha1, two->sha1)) {
2996 int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
2997
2998 if (DIFF_OPT_TST(o, BINARY)) {
2999 mmfile_t mf;
3000 if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
3001 (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
3002 abbrev = 40;
3003 }
3004 strbuf_addf(msg, "%s%sindex %s..", line_prefix, set,
3005 find_unique_abbrev(one->sha1, abbrev));
3006 strbuf_addstr(msg, find_unique_abbrev(two->sha1, abbrev));
3007 if (one->mode == two->mode)
3008 strbuf_addf(msg, " %06o", one->mode);
3009 strbuf_addf(msg, "%s\n", reset);
3010 }
3011 }
3012
3013 static void run_diff_cmd(const char *pgm,
3014 const char *name,
3015 const char *other,
3016 const char *attr_path,
3017 struct diff_filespec *one,
3018 struct diff_filespec *two,
3019 struct strbuf *msg,
3020 struct diff_options *o,
3021 struct diff_filepair *p)
3022 {
3023 const char *xfrm_msg = NULL;
3024 int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
3025 int must_show_header = 0;
3026
3027
3028 if (DIFF_OPT_TST(o, ALLOW_EXTERNAL)) {
3029 struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
3030 if (drv && drv->external)
3031 pgm = drv->external;
3032 }
3033
3034 if (msg) {
3035 /*
3036 * don't use colors when the header is intended for an
3037 * external diff driver
3038 */
3039 fill_metainfo(msg, name, other, one, two, o, p,
3040 &must_show_header,
3041 want_color(o->use_color) && !pgm);
3042 xfrm_msg = msg->len ? msg->buf : NULL;
3043 }
3044
3045 if (pgm) {
3046 run_external_diff(pgm, name, other, one, two, xfrm_msg,
3047 complete_rewrite);
3048 return;
3049 }
3050 if (one && two)
3051 builtin_diff(name, other ? other : name,
3052 one, two, xfrm_msg, must_show_header,
3053 o, complete_rewrite);
3054 else
3055 fprintf(o->file, "* Unmerged path %s\n", name);
3056 }
3057
3058 static void diff_fill_sha1_info(struct diff_filespec *one)
3059 {
3060 if (DIFF_FILE_VALID(one)) {
3061 if (!one->sha1_valid) {
3062 struct stat st;
3063 if (one->is_stdin) {
3064 hashcpy(one->sha1, null_sha1);
3065 return;
3066 }
3067 if (lstat(one->path, &st) < 0)
3068 die_errno("stat '%s'", one->path);
3069 if (index_path(one->sha1, one->path, &st, 0))
3070 die("cannot hash %s", one->path);
3071 }
3072 }
3073 else
3074 hashclr(one->sha1);
3075 }
3076
3077 static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
3078 {
3079 /* Strip the prefix but do not molest /dev/null and absolute paths */
3080 if (*namep && **namep != '/') {
3081 *namep += prefix_length;
3082 if (**namep == '/')
3083 ++*namep;
3084 }
3085 if (*otherp && **otherp != '/') {
3086 *otherp += prefix_length;
3087 if (**otherp == '/')
3088 ++*otherp;
3089 }
3090 }
3091
3092 static void run_diff(struct diff_filepair *p, struct diff_options *o)
3093 {
3094 const char *pgm = external_diff();
3095 struct strbuf msg;
3096 struct diff_filespec *one = p->one;
3097 struct diff_filespec *two = p->two;
3098 const char *name;
3099 const char *other;
3100 const char *attr_path;
3101
3102 name = p->one->path;
3103 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3104 attr_path = name;
3105 if (o->prefix_length)
3106 strip_prefix(o->prefix_length, &name, &other);
3107
3108 if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
3109 pgm = NULL;
3110
3111 if (DIFF_PAIR_UNMERGED(p)) {
3112 run_diff_cmd(pgm, name, NULL, attr_path,
3113 NULL, NULL, NULL, o, p);
3114 return;
3115 }
3116
3117 diff_fill_sha1_info(one);
3118 diff_fill_sha1_info(two);
3119
3120 if (!pgm &&
3121 DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3122 (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3123 /*
3124 * a filepair that changes between file and symlink
3125 * needs to be split into deletion and creation.
3126 */
3127 struct diff_filespec *null = alloc_filespec(two->path);
3128 run_diff_cmd(NULL, name, other, attr_path,
3129 one, null, &msg, o, p);
3130 free(null);
3131 strbuf_release(&msg);
3132
3133 null = alloc_filespec(one->path);
3134 run_diff_cmd(NULL, name, other, attr_path,
3135 null, two, &msg, o, p);
3136 free(null);
3137 }
3138 else
3139 run_diff_cmd(pgm, name, other, attr_path,
3140 one, two, &msg, o, p);
3141
3142 strbuf_release(&msg);
3143 }
3144
3145 static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3146 struct diffstat_t *diffstat)
3147 {
3148 const char *name;
3149 const char *other;
3150
3151 if (DIFF_PAIR_UNMERGED(p)) {
3152 /* unmerged */
3153 builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, p);
3154 return;
3155 }
3156
3157 name = p->one->path;
3158 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3159
3160 if (o->prefix_length)
3161 strip_prefix(o->prefix_length, &name, &other);
3162
3163 diff_fill_sha1_info(p->one);
3164 diff_fill_sha1_info(p->two);
3165
3166 builtin_diffstat(name, other, p->one, p->two, diffstat, o, p);
3167 }
3168
3169 static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3170 {
3171 const char *name;
3172 const char *other;
3173 const char *attr_path;
3174
3175 if (DIFF_PAIR_UNMERGED(p)) {
3176 /* unmerged */
3177 return;
3178 }
3179
3180 name = p->one->path;
3181 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3182 attr_path = other ? other : name;
3183
3184 if (o->prefix_length)
3185 strip_prefix(o->prefix_length, &name, &other);
3186
3187 diff_fill_sha1_info(p->one);
3188 diff_fill_sha1_info(p->two);
3189
3190 builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3191 }
3192
3193 void diff_setup(struct diff_options *options)
3194 {
3195 memcpy(options, &default_diff_options, sizeof(*options));
3196
3197 options->file = stdout;
3198
3199 options->line_termination = '\n';
3200 options->break_opt = -1;
3201 options->rename_limit = -1;
3202 options->dirstat_permille = diff_dirstat_permille_default;
3203 options->context = diff_context_default;
3204 DIFF_OPT_SET(options, RENAME_EMPTY);
3205
3206 options->change = diff_change;
3207 options->add_remove = diff_addremove;
3208 options->use_color = diff_use_color_default;
3209 options->detect_rename = diff_detect_rename_default;
3210 options->xdl_opts |= diff_algorithm;
3211
3212 if (diff_no_prefix) {
3213 options->a_prefix = options->b_prefix = "";
3214 } else if (!diff_mnemonic_prefix) {
3215 options->a_prefix = "a/";
3216 options->b_prefix = "b/";
3217 }
3218 }
3219
3220 void diff_setup_done(struct diff_options *options)
3221 {
3222 int count = 0;
3223
3224 if (options->output_format & DIFF_FORMAT_NAME)
3225 count++;
3226 if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3227 count++;
3228 if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3229 count++;
3230 if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3231 count++;
3232 if (count > 1)
3233 die("--name-only, --name-status, --check and -s are mutually exclusive");
3234
3235 /*
3236 * Most of the time we can say "there are changes"
3237 * only by checking if there are changed paths, but
3238 * --ignore-whitespace* options force us to look
3239 * inside contents.
3240 */
3241
3242 if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3243 DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3244 DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3245 DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3246 else
3247 DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3248
3249 if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3250 options->detect_rename = DIFF_DETECT_COPY;
3251
3252 if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3253 options->prefix = NULL;
3254 if (options->prefix)
3255 options->prefix_length = strlen(options->prefix);
3256 else
3257 options->prefix_length = 0;
3258
3259 if (options->output_format & (DIFF_FORMAT_NAME |
3260 DIFF_FORMAT_NAME_STATUS |
3261 DIFF_FORMAT_CHECKDIFF |
3262 DIFF_FORMAT_NO_OUTPUT))
3263 options->output_format &= ~(DIFF_FORMAT_RAW |
3264 DIFF_FORMAT_NUMSTAT |
3265 DIFF_FORMAT_DIFFSTAT |
3266 DIFF_FORMAT_SHORTSTAT |
3267 DIFF_FORMAT_DIRSTAT |
3268 DIFF_FORMAT_SUMMARY |
3269 DIFF_FORMAT_PATCH);
3270
3271 /*
3272 * These cases always need recursive; we do not drop caller-supplied
3273 * recursive bits for other formats here.
3274 */
3275 if (options->output_format & (DIFF_FORMAT_PATCH |
3276 DIFF_FORMAT_NUMSTAT |
3277 DIFF_FORMAT_DIFFSTAT |
3278 DIFF_FORMAT_SHORTSTAT |
3279 DIFF_FORMAT_DIRSTAT |
3280 DIFF_FORMAT_SUMMARY |
3281 DIFF_FORMAT_CHECKDIFF))
3282 DIFF_OPT_SET(options, RECURSIVE);
3283 /*
3284 * Also pickaxe would not work very well if you do not say recursive
3285 */
3286 if (options->pickaxe)
3287 DIFF_OPT_SET(options, RECURSIVE);
3288 /*
3289 * When patches are generated, submodules diffed against the work tree
3290 * must be checked for dirtiness too so it can be shown in the output
3291 */
3292 if (options->output_format & DIFF_FORMAT_PATCH)
3293 DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3294
3295 if (options->detect_rename && options->rename_limit < 0)
3296 options->rename_limit = diff_rename_limit_default;
3297 if (options->setup & DIFF_SETUP_USE_CACHE) {
3298 if (!active_cache)
3299 /* read-cache does not die even when it fails
3300 * so it is safe for us to do this here. Also
3301 * it does not smudge active_cache or active_nr
3302 * when it fails, so we do not have to worry about
3303 * cleaning it up ourselves either.
3304 */
3305 read_cache();
3306 }
3307 if (options->abbrev <= 0 || 40 < options->abbrev)
3308 options->abbrev = 40; /* full */
3309
3310 /*
3311 * It does not make sense to show the first hit we happened
3312 * to have found. It does not make sense not to return with
3313 * exit code in such a case either.
3314 */
3315 if (DIFF_OPT_TST(options, QUICK)) {
3316 options->output_format = DIFF_FORMAT_NO_OUTPUT;
3317 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3318 }
3319 }
3320
3321 static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3322 {
3323 char c, *eq;
3324 int len;
3325
3326 if (*arg != '-')
3327 return 0;
3328 c = *++arg;
3329 if (!c)
3330 return 0;
3331 if (c == arg_short) {
3332 c = *++arg;
3333 if (!c)
3334 return 1;
3335 if (val && isdigit(c)) {
3336 char *end;
3337 int n = strtoul(arg, &end, 10);
3338 if (*end)
3339 return 0;
3340 *val = n;
3341 return 1;
3342 }
3343 return 0;
3344 }
3345 if (c != '-')
3346 return 0;
3347 arg++;
3348 eq = strchr(arg, '=');
3349 if (eq)
3350 len = eq - arg;
3351 else
3352 len = strlen(arg);
3353 if (!len || strncmp(arg, arg_long, len))
3354 return 0;
3355 if (eq) {
3356 int n;
3357 char *end;
3358 if (!isdigit(*++eq))
3359 return 0;
3360 n = strtoul(eq, &end, 10);
3361 if (*end)
3362 return 0;
3363 *val = n;
3364 }
3365 return 1;
3366 }
3367
3368 static int diff_scoreopt_parse(const char *opt);
3369
3370 static inline int short_opt(char opt, const char **argv,
3371 const char **optarg)
3372 {
3373 const char *arg = argv[0];
3374 if (arg[0] != '-' || arg[1] != opt)
3375 return 0;
3376 if (arg[2] != '\0') {
3377 *optarg = arg + 2;
3378 return 1;
3379 }
3380 if (!argv[1])
3381 die("Option '%c' requires a value", opt);
3382 *optarg = argv[1];
3383 return 2;
3384 }
3385
3386 int parse_long_opt(const char *opt, const char **argv,
3387 const char **optarg)
3388 {
3389 const char *arg = argv[0];
3390 if (arg[0] != '-' || arg[1] != '-')
3391 return 0;
3392 arg += strlen("--");
3393 if (prefixcmp(arg, opt))
3394 return 0;
3395 arg += strlen(opt);
3396 if (*arg == '=') { /* sticked form: --option=value */
3397 *optarg = arg + 1;
3398 return 1;
3399 }
3400 if (*arg != '\0')
3401 return 0;
3402 /* separate form: --option value */
3403 if (!argv[1])
3404 die("Option '--%s' requires a value", opt);
3405 *optarg = argv[1];
3406 return 2;
3407 }
3408
3409 static int stat_opt(struct diff_options *options, const char **av)
3410 {
3411 const char *arg = av[0];
3412 char *end;
3413 int width = options->stat_width;
3414 int name_width = options->stat_name_width;
3415 int graph_width = options->stat_graph_width;
3416 int count = options->stat_count;
3417 int argcount = 1;
3418
3419 arg += strlen("--stat");
3420 end = (char *)arg;
3421
3422 switch (*arg) {
3423 case '-':
3424 if (!prefixcmp(arg, "-width")) {
3425 arg += strlen("-width");
3426 if (*arg == '=')
3427 width = strtoul(arg + 1, &end, 10);
3428 else if (!*arg && !av[1])
3429 die("Option '--stat-width' requires a value");
3430 else if (!*arg) {
3431 width = strtoul(av[1], &end, 10);
3432 argcount = 2;
3433 }
3434 } else if (!prefixcmp(arg, "-name-width")) {
3435 arg += strlen("-name-width");
3436 if (*arg == '=')
3437 name_width = strtoul(arg + 1, &end, 10);
3438 else if (!*arg && !av[1])
3439 die("Option '--stat-name-width' requires a value");
3440 else if (!*arg) {
3441 name_width = strtoul(av[1], &end, 10);
3442 argcount = 2;
3443 }
3444 } else if (!prefixcmp(arg, "-graph-width")) {
3445 arg += strlen("-graph-width");
3446 if (*arg == '=')
3447 graph_width = strtoul(arg + 1, &end, 10);
3448 else if (!*arg && !av[1])
3449 die("Option '--stat-graph-width' requires a value");
3450 else if (!*arg) {
3451 graph_width = strtoul(av[1], &end, 10);
3452 argcount = 2;
3453 }
3454 } else if (!prefixcmp(arg, "-count")) {
3455 arg += strlen("-count");
3456 if (*arg == '=')
3457 count = strtoul(arg + 1, &end, 10);
3458 else if (!*arg && !av[1])
3459 die("Option '--stat-count' requires a value");
3460 else if (!*arg) {
3461 count = strtoul(av[1], &end, 10);
3462 argcount = 2;
3463 }
3464 }
3465 break;
3466 case '=':
3467 width = strtoul(arg+1, &end, 10);
3468 if (*end == ',')
3469 name_width = strtoul(end+1, &end, 10);
3470 if (*end == ',')
3471 count = strtoul(end+1, &end, 10);
3472 }
3473
3474 /* Important! This checks all the error cases! */
3475 if (*end)
3476 return 0;
3477 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3478 options->stat_name_width = name_width;
3479 options->stat_graph_width = graph_width;
3480 options->stat_width = width;
3481 options->stat_count = count;
3482 return argcount;
3483 }
3484
3485 static int parse_dirstat_opt(struct diff_options *options, const char *params)
3486 {
3487 struct strbuf errmsg = STRBUF_INIT;
3488 if (parse_dirstat_params(options, params, &errmsg))
3489 die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3490 errmsg.buf);
3491 strbuf_release(&errmsg);
3492 /*
3493 * The caller knows a dirstat-related option is given from the command
3494 * line; allow it to say "return this_function();"
3495 */
3496 options->output_format |= DIFF_FORMAT_DIRSTAT;
3497 return 1;
3498 }
3499
3500 static int parse_submodule_opt(struct diff_options *options, const char *value)
3501 {
3502 if (parse_submodule_params(options, value))
3503 die(_("Failed to parse --submodule option parameter: '%s'"),
3504 value);
3505 return 1;
3506 }
3507
3508 int diff_opt_parse(struct diff_options *options, const char **av, int ac)
3509 {
3510 const char *arg = av[0];
3511 const char *optarg;
3512 int argcount;
3513
3514 /* Output format options */
3515 if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch"))
3516 options->output_format |= DIFF_FORMAT_PATCH;
3517 else if (opt_arg(arg, 'U', "unified", &options->context))
3518 options->output_format |= DIFF_FORMAT_PATCH;
3519 else if (!strcmp(arg, "--raw"))
3520 options->output_format |= DIFF_FORMAT_RAW;
3521 else if (!strcmp(arg, "--patch-with-raw"))
3522 options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_RAW;
3523 else if (!strcmp(arg, "--numstat"))
3524 options->output_format |= DIFF_FORMAT_NUMSTAT;
3525 else if (!strcmp(arg, "--shortstat"))
3526 options->output_format |= DIFF_FORMAT_SHORTSTAT;
3527 else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3528 return parse_dirstat_opt(options, "");
3529 else if (!prefixcmp(arg, "-X"))
3530 return parse_dirstat_opt(options, arg + 2);
3531 else if (!prefixcmp(arg, "--dirstat="))
3532 return parse_dirstat_opt(options, arg + 10);
3533 else if (!strcmp(arg, "--cumulative"))
3534 return parse_dirstat_opt(options, "cumulative");
3535 else if (!strcmp(arg, "--dirstat-by-file"))
3536 return parse_dirstat_opt(options, "files");
3537 else if (!prefixcmp(arg, "--dirstat-by-file=")) {
3538 parse_dirstat_opt(options, "files");
3539 return parse_dirstat_opt(options, arg + 18);
3540 }
3541 else if (!strcmp(arg, "--check"))
3542 options->output_format |= DIFF_FORMAT_CHECKDIFF;
3543 else if (!strcmp(arg, "--summary"))
3544 options->output_format |= DIFF_FORMAT_SUMMARY;
3545 else if (!strcmp(arg, "--patch-with-stat"))
3546 options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_DIFFSTAT;
3547 else if (!strcmp(arg, "--name-only"))
3548 options->output_format |= DIFF_FORMAT_NAME;
3549 else if (!strcmp(arg, "--name-status"))
3550 options->output_format |= DIFF_FORMAT_NAME_STATUS;
3551 else if (!strcmp(arg, "-s"))
3552 options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3553 else if (!prefixcmp(arg, "--stat"))
3554 /* --stat, --stat-width, --stat-name-width, or --stat-count */
3555 return stat_opt(options, av);
3556
3557 /* renames options */
3558 else if (!prefixcmp(arg, "-B") || !prefixcmp(arg, "--break-rewrites=") ||
3559 !strcmp(arg, "--break-rewrites")) {
3560 if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3561 return error("invalid argument to -B: %s", arg+2);
3562 }
3563 else if (!prefixcmp(arg, "-M") || !prefixcmp(arg, "--find-renames=") ||
3564 !strcmp(arg, "--find-renames")) {
3565 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3566 return error("invalid argument to -M: %s", arg+2);
3567 options->detect_rename = DIFF_DETECT_RENAME;
3568 }
3569 else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3570 options->irreversible_delete = 1;
3571 }
3572 else if (!prefixcmp(arg, "-C") || !prefixcmp(arg, "--find-copies=") ||
3573 !strcmp(arg, "--find-copies")) {
3574 if (options->detect_rename == DIFF_DETECT_COPY)
3575 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3576 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3577 return error("invalid argument to -C: %s", arg+2);
3578 options->detect_rename = DIFF_DETECT_COPY;
3579 }
3580 else if (!strcmp(arg, "--no-renames"))
3581 options->detect_rename = 0;
3582 else if (!strcmp(arg, "--rename-empty"))
3583 DIFF_OPT_SET(options, RENAME_EMPTY);
3584 else if (!strcmp(arg, "--no-rename-empty"))
3585 DIFF_OPT_CLR(options, RENAME_EMPTY);
3586 else if (!strcmp(arg, "--relative"))
3587 DIFF_OPT_SET(options, RELATIVE_NAME);
3588 else if (!prefixcmp(arg, "--relative=")) {
3589 DIFF_OPT_SET(options, RELATIVE_NAME);
3590 options->prefix = arg + 11;
3591 }
3592
3593 /* xdiff options */
3594 else if (!strcmp(arg, "--minimal"))
3595 DIFF_XDL_SET(options, NEED_MINIMAL);
3596 else if (!strcmp(arg, "--no-minimal"))
3597 DIFF_XDL_CLR(options, NEED_MINIMAL);
3598 else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3599 DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3600 else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3601 DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3602 else if (!strcmp(arg, "--ignore-space-at-eol"))
3603 DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3604 else if (!strcmp(arg, "--patience"))
3605 options->xdl_opts = DIFF_WITH_ALG(options, PATIENCE_DIFF);
3606 else if (!strcmp(arg, "--histogram"))
3607 options->xdl_opts = DIFF_WITH_ALG(options, HISTOGRAM_DIFF);
3608 else if ((argcount = parse_long_opt("diff-algorithm", av, &optarg))) {
3609 long value = parse_algorithm_value(optarg);
3610 if (value < 0)
3611 return error("option diff-algorithm accepts \"myers\", "
3612 "\"minimal\", \"patience\" and \"histogram\"");
3613 /* clear out previous settings */
3614 DIFF_XDL_CLR(options, NEED_MINIMAL);
3615 options->xdl_opts &= ~XDF_DIFF_ALGORITHM_MASK;
3616 options->xdl_opts |= value;
3617 return argcount;
3618 }
3619
3620 /* flags options */
3621 else if (!strcmp(arg, "--binary")) {
3622 options->output_format |= DIFF_FORMAT_PATCH;
3623 DIFF_OPT_SET(options, BINARY);
3624 }
3625 else if (!strcmp(arg, "--full-index"))
3626 DIFF_OPT_SET(options, FULL_INDEX);
3627 else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
3628 DIFF_OPT_SET(options, TEXT);
3629 else if (!strcmp(arg, "-R"))
3630 DIFF_OPT_SET(options, REVERSE_DIFF);
3631 else if (!strcmp(arg, "--find-copies-harder"))
3632 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3633 else if (!strcmp(arg, "--follow"))
3634 DIFF_OPT_SET(options, FOLLOW_RENAMES);
3635 else if (!strcmp(arg, "--no-follow"))
3636 DIFF_OPT_CLR(options, FOLLOW_RENAMES);
3637 else if (!strcmp(arg, "--color"))
3638 options->use_color = 1;
3639 else if (!prefixcmp(arg, "--color=")) {
3640 int value = git_config_colorbool(NULL, arg+8);
3641 if (value < 0)
3642 return error("option `color' expects \"always\", \"auto\", or \"never\"");
3643 options->use_color = value;
3644 }
3645 else if (!strcmp(arg, "--no-color"))
3646 options->use_color = 0;
3647 else if (!strcmp(arg, "--color-words")) {
3648 options->use_color = 1;
3649 options->word_diff = DIFF_WORDS_COLOR;
3650 }
3651 else if (!prefixcmp(arg, "--color-words=")) {
3652 options->use_color = 1;
3653 options->word_diff = DIFF_WORDS_COLOR;
3654 options->word_regex = arg + 14;
3655 }
3656 else if (!strcmp(arg, "--word-diff")) {
3657 if (options->word_diff == DIFF_WORDS_NONE)
3658 options->word_diff = DIFF_WORDS_PLAIN;
3659 }
3660 else if (!prefixcmp(arg, "--word-diff=")) {
3661 const char *type = arg + 12;
3662 if (!strcmp(type, "plain"))
3663 options->word_diff = DIFF_WORDS_PLAIN;
3664 else if (!strcmp(type, "color")) {
3665 options->use_color = 1;
3666 options->word_diff = DIFF_WORDS_COLOR;
3667 }
3668 else if (!strcmp(type, "porcelain"))
3669 options->word_diff = DIFF_WORDS_PORCELAIN;
3670 else if (!strcmp(type, "none"))
3671 options->word_diff = DIFF_WORDS_NONE;
3672 else
3673 die("bad --word-diff argument: %s", type);
3674 }
3675 else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
3676 if (options->word_diff == DIFF_WORDS_NONE)
3677 options->word_diff = DIFF_WORDS_PLAIN;
3678 options->word_regex = optarg;
3679 return argcount;
3680 }
3681 else if (!strcmp(arg, "--exit-code"))
3682 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3683 else if (!strcmp(arg, "--quiet"))
3684 DIFF_OPT_SET(options, QUICK);
3685 else if (!strcmp(arg, "--ext-diff"))
3686 DIFF_OPT_SET(options, ALLOW_EXTERNAL);
3687 else if (!strcmp(arg, "--no-ext-diff"))
3688 DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
3689 else if (!strcmp(arg, "--textconv"))
3690 DIFF_OPT_SET(options, ALLOW_TEXTCONV);
3691 else if (!strcmp(arg, "--no-textconv"))
3692 DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
3693 else if (!strcmp(arg, "--ignore-submodules")) {
3694 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3695 handle_ignore_submodules_arg(options, "all");
3696 } else if (!prefixcmp(arg, "--ignore-submodules=")) {
3697 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3698 handle_ignore_submodules_arg(options, arg + 20);
3699 } else if (!strcmp(arg, "--submodule"))
3700 DIFF_OPT_SET(options, SUBMODULE_LOG);
3701 else if (!prefixcmp(arg, "--submodule="))
3702 return parse_submodule_opt(options, arg + 12);
3703
3704 /* misc options */
3705 else if (!strcmp(arg, "-z"))
3706 options->line_termination = 0;
3707 else if ((argcount = short_opt('l', av, &optarg))) {
3708 options->rename_limit = strtoul(optarg, NULL, 10);
3709 return argcount;
3710 }
3711 else if ((argcount = short_opt('S', av, &optarg))) {
3712 options->pickaxe = optarg;
3713 options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
3714 return argcount;
3715 } else if ((argcount = short_opt('G', av, &optarg))) {
3716 options->pickaxe = optarg;
3717 options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
3718 return argcount;
3719 }
3720 else if (!strcmp(arg, "--pickaxe-all"))
3721 options->pickaxe_opts |= DIFF_PICKAXE_ALL;
3722 else if (!strcmp(arg, "--pickaxe-regex"))
3723 options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
3724 else if ((argcount = short_opt('O', av, &optarg))) {
3725 options->orderfile = optarg;
3726 return argcount;
3727 }
3728 else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
3729 options->filter = optarg;
3730 return argcount;
3731 }
3732 else if (!strcmp(arg, "--abbrev"))
3733 options->abbrev = DEFAULT_ABBREV;
3734 else if (!prefixcmp(arg, "--abbrev=")) {
3735 options->abbrev = strtoul(arg + 9, NULL, 10);
3736 if (options->abbrev < MINIMUM_ABBREV)
3737 options->abbrev = MINIMUM_ABBREV;
3738 else if (40 < options->abbrev)
3739 options->abbrev = 40;
3740 }
3741 else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
3742 options->a_prefix = optarg;
3743 return argcount;
3744 }
3745 else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
3746 options->b_prefix = optarg;
3747 return argcount;
3748 }
3749 else if (!strcmp(arg, "--no-prefix"))
3750 options->a_prefix = options->b_prefix = "";
3751 else if (opt_arg(arg, '\0', "inter-hunk-context",
3752 &options->interhunkcontext))
3753 ;
3754 else if (!strcmp(arg, "-W"))
3755 DIFF_OPT_SET(options, FUNCCONTEXT);
3756 else if (!strcmp(arg, "--function-context"))
3757 DIFF_OPT_SET(options, FUNCCONTEXT);
3758 else if (!strcmp(arg, "--no-function-context"))
3759 DIFF_OPT_CLR(options, FUNCCONTEXT);
3760 else if ((argcount = parse_long_opt("output", av, &optarg))) {
3761 options->file = fopen(optarg, "w");
3762 if (!options->file)
3763 die_errno("Could not open '%s'", optarg);
3764 options->close_file = 1;
3765 return argcount;
3766 } else
3767 return 0;
3768 return 1;
3769 }
3770
3771 int parse_rename_score(const char **cp_p)
3772 {
3773 unsigned long num, scale;
3774 int ch, dot;
3775 const char *cp = *cp_p;
3776
3777 num = 0;
3778 scale = 1;
3779 dot = 0;
3780 for (;;) {
3781 ch = *cp;
3782 if ( !dot && ch == '.' ) {
3783 scale = 1;
3784 dot = 1;
3785 } else if ( ch == '%' ) {
3786 scale = dot ? scale*100 : 100;
3787 cp++; /* % is always at the end */
3788 break;
3789 } else if ( ch >= '0' && ch <= '9' ) {
3790 if ( scale < 100000 ) {
3791 scale *= 10;
3792 num = (num*10) + (ch-'0');
3793 }
3794 } else {
3795 break;
3796 }
3797 cp++;
3798 }
3799 *cp_p = cp;
3800
3801 /* user says num divided by scale and we say internally that
3802 * is MAX_SCORE * num / scale.
3803 */
3804 return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
3805 }
3806
3807 static int diff_scoreopt_parse(const char *opt)
3808 {
3809 int opt1, opt2, cmd;
3810
3811 if (*opt++ != '-')
3812 return -1;
3813 cmd = *opt++;
3814 if (cmd == '-') {
3815 /* convert the long-form arguments into short-form versions */
3816 if (!prefixcmp(opt, "break-rewrites")) {
3817 opt += strlen("break-rewrites");
3818 if (*opt == 0 || *opt++ == '=')
3819 cmd = 'B';
3820 } else if (!prefixcmp(opt, "find-copies")) {
3821 opt += strlen("find-copies");
3822 if (*opt == 0 || *opt++ == '=')
3823 cmd = 'C';
3824 } else if (!prefixcmp(opt, "find-renames")) {
3825 opt += strlen("find-renames");
3826 if (*opt == 0 || *opt++ == '=')
3827 cmd = 'M';
3828 }
3829 }
3830 if (cmd != 'M' && cmd != 'C' && cmd != 'B')
3831 return -1; /* that is not a -M, -C nor -B option */
3832
3833 opt1 = parse_rename_score(&opt);
3834 if (cmd != 'B')
3835 opt2 = 0;
3836 else {
3837 if (*opt == 0)
3838 opt2 = 0;
3839 else if (*opt != '/')
3840 return -1; /* we expect -B80/99 or -B80 */
3841 else {
3842 opt++;
3843 opt2 = parse_rename_score(&opt);
3844 }
3845 }
3846 if (*opt != 0)
3847 return -1;
3848 return opt1 | (opt2 << 16);
3849 }
3850
3851 struct diff_queue_struct diff_queued_diff;
3852
3853 void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
3854 {
3855 if (queue->alloc <= queue->nr) {
3856 queue->alloc = alloc_nr(queue->alloc);
3857 queue->queue = xrealloc(queue->queue,
3858 sizeof(dp) * queue->alloc);
3859 }
3860 queue->queue[queue->nr++] = dp;
3861 }
3862
3863 struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
3864 struct diff_filespec *one,
3865 struct diff_filespec *two)
3866 {
3867 struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
3868 dp->one = one;
3869 dp->two = two;
3870 if (queue)
3871 diff_q(queue, dp);
3872 return dp;
3873 }
3874
3875 void diff_free_filepair(struct diff_filepair *p)
3876 {
3877 free_filespec(p->one);
3878 free_filespec(p->two);
3879 free(p);
3880 }
3881
3882 /* This is different from find_unique_abbrev() in that
3883 * it stuffs the result with dots for alignment.
3884 */
3885 const char *diff_unique_abbrev(const unsigned char *sha1, int len)
3886 {
3887 int abblen;
3888 const char *abbrev;
3889 if (len == 40)
3890 return sha1_to_hex(sha1);
3891
3892 abbrev = find_unique_abbrev(sha1, len);
3893 abblen = strlen(abbrev);
3894 if (abblen < 37) {
3895 static char hex[41];
3896 if (len < abblen && abblen <= len + 2)
3897 sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
3898 else
3899 sprintf(hex, "%s...", abbrev);
3900 return hex;
3901 }
3902 return sha1_to_hex(sha1);
3903 }
3904
3905 static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
3906 {
3907 int line_termination = opt->line_termination;
3908 int inter_name_termination = line_termination ? '\t' : '\0';
3909
3910 fprintf(opt->file, "%s", diff_line_prefix(opt));
3911 if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
3912 fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
3913 diff_unique_abbrev(p->one->sha1, opt->abbrev));
3914 fprintf(opt->file, "%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
3915 }
3916 if (p->score) {
3917 fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
3918 inter_name_termination);
3919 } else {
3920 fprintf(opt->file, "%c%c", p->status, inter_name_termination);
3921 }
3922
3923 if (p->status == DIFF_STATUS_COPIED ||
3924 p->status == DIFF_STATUS_RENAMED) {
3925 const char *name_a, *name_b;
3926 name_a = p->one->path;
3927 name_b = p->two->path;
3928 strip_prefix(opt->prefix_length, &name_a, &name_b);
3929 write_name_quoted(name_a, opt->file, inter_name_termination);
3930 write_name_quoted(name_b, opt->file, line_termination);
3931 } else {
3932 const char *name_a, *name_b;
3933 name_a = p->one->mode ? p->one->path : p->two->path;
3934 name_b = NULL;
3935 strip_prefix(opt->prefix_length, &name_a, &name_b);
3936 write_name_quoted(name_a, opt->file, line_termination);
3937 }
3938 }
3939
3940 int diff_unmodified_pair(struct diff_filepair *p)
3941 {
3942 /* This function is written stricter than necessary to support
3943 * the currently implemented transformers, but the idea is to
3944 * let transformers to produce diff_filepairs any way they want,
3945 * and filter and clean them up here before producing the output.
3946 */
3947 struct diff_filespec *one = p->one, *two = p->two;
3948
3949 if (DIFF_PAIR_UNMERGED(p))
3950 return 0; /* unmerged is interesting */
3951
3952 /* deletion, addition, mode or type change
3953 * and rename are all interesting.
3954 */
3955 if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
3956 DIFF_PAIR_MODE_CHANGED(p) ||
3957 strcmp(one->path, two->path))
3958 return 0;
3959
3960 /* both are valid and point at the same path. that is, we are
3961 * dealing with a change.
3962 */
3963 if (one->sha1_valid && two->sha1_valid &&
3964 !hashcmp(one->sha1, two->sha1) &&
3965 !one->dirty_submodule && !two->dirty_submodule)
3966 return 1; /* no change */
3967 if (!one->sha1_valid && !two->sha1_valid)
3968 return 1; /* both look at the same file on the filesystem. */
3969 return 0;
3970 }
3971
3972 static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
3973 {
3974 if (diff_unmodified_pair(p))
3975 return;
3976
3977 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3978 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3979 return; /* no tree diffs in patch format */
3980
3981 run_diff(p, o);
3982 }
3983
3984 static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
3985 struct diffstat_t *diffstat)
3986 {
3987 if (diff_unmodified_pair(p))
3988 return;
3989
3990 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3991 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3992 return; /* no useful stat for tree diffs */
3993
3994 run_diffstat(p, o, diffstat);
3995 }
3996
3997 static void diff_flush_checkdiff(struct diff_filepair *p,
3998 struct diff_options *o)
3999 {
4000 if (diff_unmodified_pair(p))
4001 return;
4002
4003 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4004 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4005 return; /* nothing to check in tree diffs */
4006
4007 run_checkdiff(p, o);
4008 }
4009
4010 int diff_queue_is_empty(void)
4011 {
4012 struct diff_queue_struct *q = &diff_queued_diff;
4013 int i;
4014 for (i = 0; i < q->nr; i++)
4015 if (!diff_unmodified_pair(q->queue[i]))
4016 return 0;
4017 return 1;
4018 }
4019
4020 #if DIFF_DEBUG
4021 void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
4022 {
4023 fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
4024 x, one ? one : "",
4025 s->path,
4026 DIFF_FILE_VALID(s) ? "valid" : "invalid",
4027 s->mode,
4028 s->sha1_valid ? sha1_to_hex(s->sha1) : "");
4029 fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
4030 x, one ? one : "",
4031 s->size, s->xfrm_flags);
4032 }
4033
4034 void diff_debug_filepair(const struct diff_filepair *p, int i)
4035 {
4036 diff_debug_filespec(p->one, i, "one");
4037 diff_debug_filespec(p->two, i, "two");
4038 fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
4039 p->score, p->status ? p->status : '?',
4040 p->one->rename_used, p->broken_pair);
4041 }
4042
4043 void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
4044 {
4045 int i;
4046 if (msg)
4047 fprintf(stderr, "%s\n", msg);
4048 fprintf(stderr, "q->nr = %d\n", q->nr);
4049 for (i = 0; i < q->nr; i++) {
4050 struct diff_filepair *p = q->queue[i];
4051 diff_debug_filepair(p, i);
4052 }
4053 }
4054 #endif
4055
4056 static void diff_resolve_rename_copy(void)
4057 {
4058 int i;
4059 struct diff_filepair *p;
4060 struct diff_queue_struct *q = &diff_queued_diff;
4061
4062 diff_debug_queue("resolve-rename-copy", q);
4063
4064 for (i = 0; i < q->nr; i++) {
4065 p = q->queue[i];
4066 p->status = 0; /* undecided */
4067 if (DIFF_PAIR_UNMERGED(p))
4068 p->status = DIFF_STATUS_UNMERGED;
4069 else if (!DIFF_FILE_VALID(p->one))
4070 p->status = DIFF_STATUS_ADDED;
4071 else if (!DIFF_FILE_VALID(p->two))
4072 p->status = DIFF_STATUS_DELETED;
4073 else if (DIFF_PAIR_TYPE_CHANGED(p))
4074 p->status = DIFF_STATUS_TYPE_CHANGED;
4075
4076 /* from this point on, we are dealing with a pair
4077 * whose both sides are valid and of the same type, i.e.
4078 * either in-place edit or rename/copy edit.
4079 */
4080 else if (DIFF_PAIR_RENAME(p)) {
4081 /*
4082 * A rename might have re-connected a broken
4083 * pair up, causing the pathnames to be the
4084 * same again. If so, that's not a rename at
4085 * all, just a modification..
4086 *
4087 * Otherwise, see if this source was used for
4088 * multiple renames, in which case we decrement
4089 * the count, and call it a copy.
4090 */
4091 if (!strcmp(p->one->path, p->two->path))
4092 p->status = DIFF_STATUS_MODIFIED;
4093 else if (--p->one->rename_used > 0)
4094 p->status = DIFF_STATUS_COPIED;
4095 else
4096 p->status = DIFF_STATUS_RENAMED;
4097 }
4098 else if (hashcmp(p->one->sha1, p->two->sha1) ||
4099 p->one->mode != p->two->mode ||
4100 p->one->dirty_submodule ||
4101 p->two->dirty_submodule ||
4102 is_null_sha1(p->one->sha1))
4103 p->status = DIFF_STATUS_MODIFIED;
4104 else {
4105 /* This is a "no-change" entry and should not
4106 * happen anymore, but prepare for broken callers.
4107 */
4108 error("feeding unmodified %s to diffcore",
4109 p->one->path);
4110 p->status = DIFF_STATUS_UNKNOWN;
4111 }
4112 }
4113 diff_debug_queue("resolve-rename-copy done", q);
4114 }
4115
4116 static int check_pair_status(struct diff_filepair *p)
4117 {
4118 switch (p->status) {
4119 case DIFF_STATUS_UNKNOWN:
4120 return 0;
4121 case 0:
4122 die("internal error in diff-resolve-rename-copy");
4123 default:
4124 return 1;
4125 }
4126 }
4127
4128 static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
4129 {
4130 int fmt = opt->output_format;
4131
4132 if (fmt & DIFF_FORMAT_CHECKDIFF)
4133 diff_flush_checkdiff(p, opt);
4134 else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
4135 diff_flush_raw(p, opt);
4136 else if (fmt & DIFF_FORMAT_NAME) {
4137 const char *name_a, *name_b;
4138 name_a = p->two->path;
4139 name_b = NULL;
4140 strip_prefix(opt->prefix_length, &name_a, &name_b);
4141 write_name_quoted(name_a, opt->file, opt->line_termination);
4142 }
4143 }
4144
4145 static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
4146 {
4147 if (fs->mode)
4148 fprintf(file, " %s mode %06o ", newdelete, fs->mode);
4149 else
4150 fprintf(file, " %s ", newdelete);
4151 write_name_quoted(fs->path, file, '\n');
4152 }
4153
4154
4155 static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
4156 const char *line_prefix)
4157 {
4158 if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
4159 fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
4160 p->two->mode, show_name ? ' ' : '\n');
4161 if (show_name) {
4162 write_name_quoted(p->two->path, file, '\n');
4163 }
4164 }
4165 }
4166
4167 static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
4168 const char *line_prefix)
4169 {
4170 char *names = pprint_rename(p->one->path, p->two->path);
4171
4172 fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
4173 free(names);
4174 show_mode_change(file, p, 0, line_prefix);
4175 }
4176
4177 static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
4178 {
4179 FILE *file = opt->file;
4180 const char *line_prefix = diff_line_prefix(opt);
4181
4182 switch(p->status) {
4183 case DIFF_STATUS_DELETED:
4184 fputs(line_prefix, file);
4185 show_file_mode_name(file, "delete", p->one);
4186 break;
4187 case DIFF_STATUS_ADDED:
4188 fputs(line_prefix, file);
4189 show_file_mode_name(file, "create", p->two);
4190 break;
4191 case DIFF_STATUS_COPIED:
4192 fputs(line_prefix, file);
4193 show_rename_copy(file, "copy", p, line_prefix);
4194 break;
4195 case DIFF_STATUS_RENAMED:
4196 fputs(line_prefix, file);
4197 show_rename_copy(file, "rename", p, line_prefix);
4198 break;
4199 default:
4200 if (p->score) {
4201 fprintf(file, "%s rewrite ", line_prefix);
4202 write_name_quoted(p->two->path, file, ' ');
4203 fprintf(file, "(%d%%)\n", similarity_index(p));
4204 }
4205 show_mode_change(file, p, !p->score, line_prefix);
4206 break;
4207 }
4208 }
4209
4210 struct patch_id_t {
4211 git_SHA_CTX *ctx;
4212 int patchlen;
4213 };
4214
4215 static int remove_space(char *line, int len)
4216 {
4217 int i;
4218 char *dst = line;
4219 unsigned char c;
4220
4221 for (i = 0; i < len; i++)
4222 if (!isspace((c = line[i])))
4223 *dst++ = c;
4224
4225 return dst - line;
4226 }
4227
4228 static void patch_id_consume(void *priv, char *line, unsigned long len)
4229 {
4230 struct patch_id_t *data = priv;
4231 int new_len;
4232
4233 /* Ignore line numbers when computing the SHA1 of the patch */
4234 if (!prefixcmp(line, "@@ -"))
4235 return;
4236
4237 new_len = remove_space(line, len);
4238
4239 git_SHA1_Update(data->ctx, line, new_len);
4240 data->patchlen += new_len;
4241 }
4242
4243 /* returns 0 upon success, and writes result into sha1 */
4244 static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
4245 {
4246 struct diff_queue_struct *q = &diff_queued_diff;
4247 int i;
4248 git_SHA_CTX ctx;
4249 struct patch_id_t data;
4250 char buffer[PATH_MAX * 4 + 20];
4251
4252 git_SHA1_Init(&ctx);
4253 memset(&data, 0, sizeof(struct patch_id_t));
4254 data.ctx = &ctx;
4255
4256 for (i = 0; i < q->nr; i++) {
4257 xpparam_t xpp;
4258 xdemitconf_t xecfg;
4259 mmfile_t mf1, mf2;
4260 struct diff_filepair *p = q->queue[i];
4261 int len1, len2;
4262
4263 memset(&xpp, 0, sizeof(xpp));
4264 memset(&xecfg, 0, sizeof(xecfg));
4265 if (p->status == 0)
4266 return error("internal diff status error");
4267 if (p->status == DIFF_STATUS_UNKNOWN)
4268 continue;
4269 if (diff_unmodified_pair(p))
4270 continue;
4271 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4272 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4273 continue;
4274 if (DIFF_PAIR_UNMERGED(p))
4275 continue;
4276
4277 diff_fill_sha1_info(p->one);
4278 diff_fill_sha1_info(p->two);
4279 if (fill_mmfile(&mf1, p->one) < 0 ||
4280 fill_mmfile(&mf2, p->two) < 0)
4281 return error("unable to read files to diff");
4282
4283 len1 = remove_space(p->one->path, strlen(p->one->path));
4284 len2 = remove_space(p->two->path, strlen(p->two->path));
4285 if (p->one->mode == 0)
4286 len1 = snprintf(buffer, sizeof(buffer),
4287 "diff--gita/%.*sb/%.*s"
4288 "newfilemode%06o"
4289 "---/dev/null"
4290 "+++b/%.*s",
4291 len1, p->one->path,
4292 len2, p->two->path,
4293 p->two->mode,
4294 len2, p->two->path);
4295 else if (p->two->mode == 0)
4296 len1 = snprintf(buffer, sizeof(buffer),
4297 "diff--gita/%.*sb/%.*s"
4298 "deletedfilemode%06o"
4299 "---a/%.*s"
4300 "+++/dev/null",
4301 len1, p->one->path,
4302 len2, p->two->path,
4303 p->one->mode,
4304 len1, p->one->path);
4305 else
4306 len1 = snprintf(buffer, sizeof(buffer),
4307 "diff--gita/%.*sb/%.*s"
4308 "---a/%.*s"
4309 "+++b/%.*s",
4310 len1, p->one->path,
4311 len2, p->two->path,
4312 len1, p->one->path,
4313 len2, p->two->path);
4314 git_SHA1_Update(&ctx, buffer, len1);
4315
4316 if (diff_filespec_is_binary(p->one) ||
4317 diff_filespec_is_binary(p->two)) {
4318 git_SHA1_Update(&ctx, sha1_to_hex(p->one->sha1), 40);
4319 git_SHA1_Update(&ctx, sha1_to_hex(p->two->sha1), 40);
4320 continue;
4321 }
4322
4323 xpp.flags = 0;
4324 xecfg.ctxlen = 3;
4325 xecfg.flags = 0;
4326 xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4327 &xpp, &xecfg);
4328 }
4329
4330 git_SHA1_Final(sha1, &ctx);
4331 return 0;
4332 }
4333
4334 int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
4335 {
4336 struct diff_queue_struct *q = &diff_queued_diff;
4337 int i;
4338 int result = diff_get_patch_id(options, sha1);
4339
4340 for (i = 0; i < q->nr; i++)
4341 diff_free_filepair(q->queue[i]);
4342
4343 free(q->queue);
4344 DIFF_QUEUE_CLEAR(q);
4345
4346 return result;
4347 }
4348
4349 static int is_summary_empty(const struct diff_queue_struct *q)
4350 {
4351 int i;
4352
4353 for (i = 0; i < q->nr; i++) {
4354 const struct diff_filepair *p = q->queue[i];
4355
4356 switch (p->status) {
4357 case DIFF_STATUS_DELETED:
4358 case DIFF_STATUS_ADDED:
4359 case DIFF_STATUS_COPIED:
4360 case DIFF_STATUS_RENAMED:
4361 return 0;
4362 default:
4363 if (p->score)
4364 return 0;
4365 if (p->one->mode && p->two->mode &&
4366 p->one->mode != p->two->mode)
4367 return 0;
4368 break;
4369 }
4370 }
4371 return 1;
4372 }
4373
4374 static const char rename_limit_warning[] =
4375 "inexact rename detection was skipped due to too many files.";
4376
4377 static const char degrade_cc_to_c_warning[] =
4378 "only found copies from modified paths due to too many files.";
4379
4380 static const char rename_limit_advice[] =
4381 "you may want to set your %s variable to at least "
4382 "%d and retry the command.";
4383
4384 void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4385 {
4386 if (degraded_cc)
4387 warning(degrade_cc_to_c_warning);
4388 else if (needed)
4389 warning(rename_limit_warning);
4390 else
4391 return;
4392 if (0 < needed && needed < 32767)
4393 warning(rename_limit_advice, varname, needed);
4394 }
4395
4396 void diff_flush(struct diff_options *options)
4397 {
4398 struct diff_queue_struct *q = &diff_queued_diff;
4399 int i, output_format = options->output_format;
4400 int separator = 0;
4401 int dirstat_by_line = 0;
4402
4403 /*
4404 * Order: raw, stat, summary, patch
4405 * or: name/name-status/checkdiff (other bits clear)
4406 */
4407 if (!q->nr)
4408 goto free_queue;
4409
4410 if (output_format & (DIFF_FORMAT_RAW |
4411 DIFF_FORMAT_NAME |
4412 DIFF_FORMAT_NAME_STATUS |
4413 DIFF_FORMAT_CHECKDIFF)) {
4414 for (i = 0; i < q->nr; i++) {
4415 struct diff_filepair *p = q->queue[i];
4416 if (check_pair_status(p))
4417 flush_one_pair(p, options);
4418 }
4419 separator++;
4420 }
4421
4422 if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4423 dirstat_by_line = 1;
4424
4425 if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4426 dirstat_by_line) {
4427 struct diffstat_t diffstat;
4428
4429 memset(&diffstat, 0, sizeof(struct diffstat_t));
4430 for (i = 0; i < q->nr; i++) {
4431 struct diff_filepair *p = q->queue[i];
4432 if (check_pair_status(p))
4433 diff_flush_stat(p, options, &diffstat);
4434 }
4435 if (output_format & DIFF_FORMAT_NUMSTAT)
4436 show_numstat(&diffstat, options);
4437 if (output_format & DIFF_FORMAT_DIFFSTAT)
4438 show_stats(&diffstat, options);
4439 if (output_format & DIFF_FORMAT_SHORTSTAT)
4440 show_shortstats(&diffstat, options);
4441 if (output_format & DIFF_FORMAT_DIRSTAT)
4442 show_dirstat_by_line(&diffstat, options);
4443 free_diffstat_info(&diffstat);
4444 separator++;
4445 }
4446 if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4447 show_dirstat(options);
4448
4449 if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4450 for (i = 0; i < q->nr; i++) {
4451 diff_summary(options, q->queue[i]);
4452 }
4453 separator++;
4454 }
4455
4456 if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4457 DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4458 DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4459 /*
4460 * run diff_flush_patch for the exit status. setting
4461 * options->file to /dev/null should be safe, becaue we
4462 * aren't supposed to produce any output anyway.
4463 */
4464 if (options->close_file)
4465 fclose(options->file);
4466 options->file = fopen("/dev/null", "w");
4467 if (!options->file)
4468 die_errno("Could not open /dev/null");
4469 options->close_file = 1;
4470 for (i = 0; i < q->nr; i++) {
4471 struct diff_filepair *p = q->queue[i];
4472 if (check_pair_status(p))
4473 diff_flush_patch(p, options);
4474 if (options->found_changes)
4475 break;
4476 }
4477 }
4478
4479 if (output_format & DIFF_FORMAT_PATCH) {
4480 if (separator) {
4481 fprintf(options->file, "%s%c",
4482 diff_line_prefix(options),
4483 options->line_termination);
4484 if (options->stat_sep) {
4485 /* attach patch instead of inline */
4486 fputs(options->stat_sep, options->file);
4487 }
4488 }
4489
4490 for (i = 0; i < q->nr; i++) {
4491 struct diff_filepair *p = q->queue[i];
4492 if (check_pair_status(p))
4493 diff_flush_patch(p, options);
4494 }
4495 }
4496
4497 if (output_format & DIFF_FORMAT_CALLBACK)
4498 options->format_callback(q, options, options->format_callback_data);
4499
4500 for (i = 0; i < q->nr; i++)
4501 diff_free_filepair(q->queue[i]);
4502 free_queue:
4503 free(q->queue);
4504 DIFF_QUEUE_CLEAR(q);
4505 if (options->close_file)
4506 fclose(options->file);
4507
4508 /*
4509 * Report the content-level differences with HAS_CHANGES;
4510 * diff_addremove/diff_change does not set the bit when
4511 * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4512 */
4513 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4514 if (options->found_changes)
4515 DIFF_OPT_SET(options, HAS_CHANGES);
4516 else
4517 DIFF_OPT_CLR(options, HAS_CHANGES);
4518 }
4519 }
4520
4521 static void diffcore_apply_filter(const char *filter)
4522 {
4523 int i;
4524 struct diff_queue_struct *q = &diff_queued_diff;
4525 struct diff_queue_struct outq;
4526 DIFF_QUEUE_CLEAR(&outq);
4527
4528 if (!filter)
4529 return;
4530
4531 if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
4532 int found;
4533 for (i = found = 0; !found && i < q->nr; i++) {
4534 struct diff_filepair *p = q->queue[i];
4535 if (((p->status == DIFF_STATUS_MODIFIED) &&
4536 ((p->score &&
4537 strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
4538 (!p->score &&
4539 strchr(filter, DIFF_STATUS_MODIFIED)))) ||
4540 ((p->status != DIFF_STATUS_MODIFIED) &&
4541 strchr(filter, p->status)))
4542 found++;
4543 }
4544 if (found)
4545 return;
4546
4547 /* otherwise we will clear the whole queue
4548 * by copying the empty outq at the end of this
4549 * function, but first clear the current entries
4550 * in the queue.
4551 */
4552 for (i = 0; i < q->nr; i++)
4553 diff_free_filepair(q->queue[i]);
4554 }
4555 else {
4556 /* Only the matching ones */
4557 for (i = 0; i < q->nr; i++) {
4558 struct diff_filepair *p = q->queue[i];
4559
4560 if (((p->status == DIFF_STATUS_MODIFIED) &&
4561 ((p->score &&
4562 strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
4563 (!p->score &&
4564 strchr(filter, DIFF_STATUS_MODIFIED)))) ||
4565 ((p->status != DIFF_STATUS_MODIFIED) &&
4566 strchr(filter, p->status)))
4567 diff_q(&outq, p);
4568 else
4569 diff_free_filepair(p);
4570 }
4571 }
4572 free(q->queue);
4573 *q = outq;
4574 }
4575
4576 /* Check whether two filespecs with the same mode and size are identical */
4577 static int diff_filespec_is_identical(struct diff_filespec *one,
4578 struct diff_filespec *two)
4579 {
4580 if (S_ISGITLINK(one->mode))
4581 return 0;
4582 if (diff_populate_filespec(one, 0))
4583 return 0;
4584 if (diff_populate_filespec(two, 0))
4585 return 0;
4586 return !memcmp(one->data, two->data, one->size);
4587 }
4588
4589 static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
4590 {
4591 int i;
4592 struct diff_queue_struct *q = &diff_queued_diff;
4593 struct diff_queue_struct outq;
4594 DIFF_QUEUE_CLEAR(&outq);
4595
4596 for (i = 0; i < q->nr; i++) {
4597 struct diff_filepair *p = q->queue[i];
4598
4599 /*
4600 * 1. Entries that come from stat info dirtiness
4601 * always have both sides (iow, not create/delete),
4602 * one side of the object name is unknown, with
4603 * the same mode and size. Keep the ones that
4604 * do not match these criteria. They have real
4605 * differences.
4606 *
4607 * 2. At this point, the file is known to be modified,
4608 * with the same mode and size, and the object
4609 * name of one side is unknown. Need to inspect
4610 * the identical contents.
4611 */
4612 if (!DIFF_FILE_VALID(p->one) || /* (1) */
4613 !DIFF_FILE_VALID(p->two) ||
4614 (p->one->sha1_valid && p->two->sha1_valid) ||
4615 (p->one->mode != p->two->mode) ||
4616 diff_populate_filespec(p->one, 1) ||
4617 diff_populate_filespec(p->two, 1) ||
4618 (p->one->size != p->two->size) ||
4619 !diff_filespec_is_identical(p->one, p->two)) /* (2) */
4620 diff_q(&outq, p);
4621 else {
4622 /*
4623 * The caller can subtract 1 from skip_stat_unmatch
4624 * to determine how many paths were dirty only
4625 * due to stat info mismatch.
4626 */
4627 if (!DIFF_OPT_TST(diffopt, NO_INDEX))
4628 diffopt->skip_stat_unmatch++;
4629 diff_free_filepair(p);
4630 }
4631 }
4632 free(q->queue);
4633 *q = outq;
4634 }
4635
4636 static int diffnamecmp(const void *a_, const void *b_)
4637 {
4638 const struct diff_filepair *a = *((const struct diff_filepair **)a_);
4639 const struct diff_filepair *b = *((const struct diff_filepair **)b_);
4640 const char *name_a, *name_b;
4641
4642 name_a = a->one ? a->one->path : a->two->path;
4643 name_b = b->one ? b->one->path : b->two->path;
4644 return strcmp(name_a, name_b);
4645 }
4646
4647 void diffcore_fix_diff_index(struct diff_options *options)
4648 {
4649 struct diff_queue_struct *q = &diff_queued_diff;
4650 qsort(q->queue, q->nr, sizeof(q->queue[0]), diffnamecmp);
4651 }
4652
4653 void diffcore_std(struct diff_options *options)
4654 {
4655 if (options->skip_stat_unmatch)
4656 diffcore_skip_stat_unmatch(options);
4657 if (!options->found_follow) {
4658 /* See try_to_follow_renames() in tree-diff.c */
4659 if (options->break_opt != -1)
4660 diffcore_break(options->break_opt);
4661 if (options->detect_rename)
4662 diffcore_rename(options);
4663 if (options->break_opt != -1)
4664 diffcore_merge_broken();
4665 }
4666 if (options->pickaxe)
4667 diffcore_pickaxe(options);
4668 if (options->orderfile)
4669 diffcore_order(options->orderfile);
4670 if (!options->found_follow)
4671 /* See try_to_follow_renames() in tree-diff.c */
4672 diff_resolve_rename_copy();
4673 diffcore_apply_filter(options->filter);
4674
4675 if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4676 DIFF_OPT_SET(options, HAS_CHANGES);
4677 else
4678 DIFF_OPT_CLR(options, HAS_CHANGES);
4679
4680 options->found_follow = 0;
4681 }
4682
4683 int diff_result_code(struct diff_options *opt, int status)
4684 {
4685 int result = 0;
4686
4687 diff_warn_rename_limit("diff.renameLimit",
4688 opt->needed_rename_limit,
4689 opt->degraded_cc_to_c);
4690 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4691 !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
4692 return status;
4693 if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4694 DIFF_OPT_TST(opt, HAS_CHANGES))
4695 result |= 01;
4696 if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
4697 DIFF_OPT_TST(opt, CHECK_FAILED))
4698 result |= 02;
4699 return result;
4700 }
4701
4702 int diff_can_quit_early(struct diff_options *opt)
4703 {
4704 return (DIFF_OPT_TST(opt, QUICK) &&
4705 !opt->filter &&
4706 DIFF_OPT_TST(opt, HAS_CHANGES));
4707 }
4708
4709 /*
4710 * Shall changes to this submodule be ignored?
4711 *
4712 * Submodule changes can be configured to be ignored separately for each path,
4713 * but that configuration can be overridden from the command line.
4714 */
4715 static int is_submodule_ignored(const char *path, struct diff_options *options)
4716 {
4717 int ignored = 0;
4718 unsigned orig_flags = options->flags;
4719 if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
4720 set_diffopt_flags_from_submodule_config(options, path);
4721 if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
4722 ignored = 1;
4723 options->flags = orig_flags;
4724 return ignored;
4725 }
4726
4727 void diff_addremove(struct diff_options *options,
4728 int addremove, unsigned mode,
4729 const unsigned char *sha1,
4730 int sha1_valid,
4731 const char *concatpath, unsigned dirty_submodule)
4732 {
4733 struct diff_filespec *one, *two;
4734
4735 if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
4736 return;
4737
4738 /* This may look odd, but it is a preparation for
4739 * feeding "there are unchanged files which should
4740 * not produce diffs, but when you are doing copy
4741 * detection you would need them, so here they are"
4742 * entries to the diff-core. They will be prefixed
4743 * with something like '=' or '*' (I haven't decided
4744 * which but should not make any difference).
4745 * Feeding the same new and old to diff_change()
4746 * also has the same effect.
4747 * Before the final output happens, they are pruned after
4748 * merged into rename/copy pairs as appropriate.
4749 */
4750 if (DIFF_OPT_TST(options, REVERSE_DIFF))
4751 addremove = (addremove == '+' ? '-' :
4752 addremove == '-' ? '+' : addremove);
4753
4754 if (options->prefix &&
4755 strncmp(concatpath, options->prefix, options->prefix_length))
4756 return;
4757
4758 one = alloc_filespec(concatpath);
4759 two = alloc_filespec(concatpath);
4760
4761 if (addremove != '+')
4762 fill_filespec(one, sha1, sha1_valid, mode);
4763 if (addremove != '-') {
4764 fill_filespec(two, sha1, sha1_valid, mode);
4765 two->dirty_submodule = dirty_submodule;
4766 }
4767
4768 diff_queue(&diff_queued_diff, one, two);
4769 if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4770 DIFF_OPT_SET(options, HAS_CHANGES);
4771 }
4772
4773 void diff_change(struct diff_options *options,
4774 unsigned old_mode, unsigned new_mode,
4775 const unsigned char *old_sha1,
4776 const unsigned char *new_sha1,
4777 int old_sha1_valid, int new_sha1_valid,
4778 const char *concatpath,
4779 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
4780 {
4781 struct diff_filespec *one, *two;
4782
4783 if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
4784 is_submodule_ignored(concatpath, options))
4785 return;
4786
4787 if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
4788 unsigned tmp;
4789 const unsigned char *tmp_c;
4790 tmp = old_mode; old_mode = new_mode; new_mode = tmp;
4791 tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
4792 tmp = old_sha1_valid; old_sha1_valid = new_sha1_valid;
4793 new_sha1_valid = tmp;
4794 tmp = old_dirty_submodule; old_dirty_submodule = new_dirty_submodule;
4795 new_dirty_submodule = tmp;
4796 }
4797
4798 if (options->prefix &&
4799 strncmp(concatpath, options->prefix, options->prefix_length))
4800 return;
4801
4802 one = alloc_filespec(concatpath);
4803 two = alloc_filespec(concatpath);
4804 fill_filespec(one, old_sha1, old_sha1_valid, old_mode);
4805 fill_filespec(two, new_sha1, new_sha1_valid, new_mode);
4806 one->dirty_submodule = old_dirty_submodule;
4807 two->dirty_submodule = new_dirty_submodule;
4808
4809 diff_queue(&diff_queued_diff, one, two);
4810 if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4811 DIFF_OPT_SET(options, HAS_CHANGES);
4812 }
4813
4814 struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
4815 {
4816 struct diff_filepair *pair;
4817 struct diff_filespec *one, *two;
4818
4819 if (options->prefix &&
4820 strncmp(path, options->prefix, options->prefix_length))
4821 return NULL;
4822
4823 one = alloc_filespec(path);
4824 two = alloc_filespec(path);
4825 pair = diff_queue(&diff_queued_diff, one, two);
4826 pair->is_unmerged = 1;
4827 return pair;
4828 }
4829
4830 static char *run_textconv(const char *pgm, struct diff_filespec *spec,
4831 size_t *outsize)
4832 {
4833 struct diff_tempfile *temp;
4834 const char *argv[3];
4835 const char **arg = argv;
4836 struct child_process child;
4837 struct strbuf buf = STRBUF_INIT;
4838 int err = 0;
4839
4840 temp = prepare_temp_file(spec->path, spec);
4841 *arg++ = pgm;
4842 *arg++ = temp->name;
4843 *arg = NULL;
4844
4845 memset(&child, 0, sizeof(child));
4846 child.use_shell = 1;
4847 child.argv = argv;
4848 child.out = -1;
4849 if (start_command(&child)) {
4850 remove_tempfile();
4851 return NULL;
4852 }
4853
4854 if (strbuf_read(&buf, child.out, 0) < 0)
4855 err = error("error reading from textconv command '%s'", pgm);
4856 close(child.out);
4857
4858 if (finish_command(&child) || err) {
4859 strbuf_release(&buf);
4860 remove_tempfile();
4861 return NULL;
4862 }
4863 remove_tempfile();
4864
4865 return strbuf_detach(&buf, outsize);
4866 }
4867
4868 size_t fill_textconv(struct userdiff_driver *driver,
4869 struct diff_filespec *df,
4870 char **outbuf)
4871 {
4872 size_t size;
4873
4874 if (!driver || !driver->textconv) {
4875 if (!DIFF_FILE_VALID(df)) {
4876 *outbuf = "";
4877 return 0;
4878 }
4879 if (diff_populate_filespec(df, 0))
4880 die("unable to read files to diff");
4881 *outbuf = df->data;
4882 return df->size;
4883 }
4884
4885 if (driver->textconv_cache && df->sha1_valid) {
4886 *outbuf = notes_cache_get(driver->textconv_cache, df->sha1,
4887 &size);
4888 if (*outbuf)
4889 return size;
4890 }
4891
4892 *outbuf = run_textconv(driver->textconv, df, &size);
4893 if (!*outbuf)
4894 die("unable to read files to diff");
4895
4896 if (driver->textconv_cache && df->sha1_valid) {
4897 /* ignore errors, as we might be in a readonly repository */
4898 notes_cache_put(driver->textconv_cache, df->sha1, *outbuf,
4899 size);
4900 /*
4901 * we could save up changes and flush them all at the end,
4902 * but we would need an extra call after all diffing is done.
4903 * Since generating a cache entry is the slow path anyway,
4904 * this extra overhead probably isn't a big deal.
4905 */
4906 notes_cache_write(driver->textconv_cache);
4907 }
4908
4909 return size;
4910 }
4911
4912 void setup_diff_pager(struct diff_options *opt)
4913 {
4914 /*
4915 * If the user asked for our exit code, then either they want --quiet
4916 * or --exit-code. We should definitely not bother with a pager in the
4917 * former case, as we will generate no output. Since we still properly
4918 * report our exit code even when a pager is run, we _could_ run a
4919 * pager with --exit-code. But since we have not done so historically,
4920 * and because it is easy to find people oneline advising "git diff
4921 * --exit-code" in hooks and other scripts, we do not do so.
4922 */
4923 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4924 check_pager_config("diff") != 0)
4925 setup_pager();
4926 }
Gitweb development nest