?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
compate/clipped-write.c: large write(2) fails on Mac OS X/XNU
[git/jrn.git] / diff.c
blob156fec447061f83855edf27d578f055bb606cacf
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 len_a = strlen(a);
1268 int len_b = strlen(b);
1269 int a_midlen, b_midlen;
1270 int qlen_a = quote_c_style(a, NULL, NULL, 0);
1271 int qlen_b = quote_c_style(b, NULL, NULL, 0);
1272
1273 if (qlen_a || qlen_b) {
1274 quote_c_style(a, &name, NULL, 0);
1275 strbuf_addstr(&name, " => ");
1276 quote_c_style(b, &name, NULL, 0);
1277 return strbuf_detach(&name, NULL);
1278 }
1279
1280 /* Find common prefix */
1281 pfx_length = 0;
1282 while (*old && *new && *old == *new) {
1283 if (*old == '/')
1284 pfx_length = old - a + 1;
1285 old++;
1286 new++;
1287 }
1288
1289 /* Find common suffix */
1290 old = a + len_a;
1291 new = b + len_b;
1292 sfx_length = 0;
1293 while (a <= old && b <= new && *old == *new) {
1294 if (*old == '/')
1295 sfx_length = len_a - (old - a);
1296 old--;
1297 new--;
1298 }
1299
1300 /*
1301 * pfx{mid-a => mid-b}sfx
1302 * {pfx-a => pfx-b}sfx
1303 * pfx{sfx-a => sfx-b}
1304 * name-a => name-b
1305 */
1306 a_midlen = len_a - pfx_length - sfx_length;
1307 b_midlen = len_b - pfx_length - sfx_length;
1308 if (a_midlen < 0)
1309 a_midlen = 0;
1310 if (b_midlen < 0)
1311 b_midlen = 0;
1312
1313 strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1314 if (pfx_length + sfx_length) {
1315 strbuf_add(&name, a, pfx_length);
1316 strbuf_addch(&name, '{');
1317 }
1318 strbuf_add(&name, a + pfx_length, a_midlen);
1319 strbuf_addstr(&name, " => ");
1320 strbuf_add(&name, b + pfx_length, b_midlen);
1321 if (pfx_length + sfx_length) {
1322 strbuf_addch(&name, '}');
1323 strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1324 }
1325 return strbuf_detach(&name, NULL);
1326 }
1327
1328 struct diffstat_t {
1329 int nr;
1330 int alloc;
1331 struct diffstat_file {
1332 char *from_name;
1333 char *name;
1334 char *print_name;
1335 unsigned is_unmerged:1;
1336 unsigned is_binary:1;
1337 unsigned is_renamed:1;
1338 unsigned is_interesting:1;
1339 uintmax_t added, deleted;
1340 } **files;
1341 };
1342
1343 static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1344 const char *name_a,
1345 const char *name_b)
1346 {
1347 struct diffstat_file *x;
1348 x = xcalloc(sizeof (*x), 1);
1349 if (diffstat->nr == diffstat->alloc) {
1350 diffstat->alloc = alloc_nr(diffstat->alloc);
1351 diffstat->files = xrealloc(diffstat->files,
1352 diffstat->alloc * sizeof(x));
1353 }
1354 diffstat->files[diffstat->nr++] = x;
1355 if (name_b) {
1356 x->from_name = xstrdup(name_a);
1357 x->name = xstrdup(name_b);
1358 x->is_renamed = 1;
1359 }
1360 else {
1361 x->from_name = NULL;
1362 x->name = xstrdup(name_a);
1363 }
1364 return x;
1365 }
1366
1367 static void diffstat_consume(void *priv, char *line, unsigned long len)
1368 {
1369 struct diffstat_t *diffstat = priv;
1370 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1371
1372 if (line[0] == '+')
1373 x->added++;
1374 else if (line[0] == '-')
1375 x->deleted++;
1376 }
1377
1378 const char mime_boundary_leader[] = "------------";
1379
1380 static int scale_linear(int it, int width, int max_change)
1381 {
1382 if (!it)
1383 return 0;
1384 /*
1385 * make sure that at least one '-' or '+' is printed if
1386 * there is any change to this path. The easiest way is to
1387 * scale linearly as if the alloted width is one column shorter
1388 * than it is, and then add 1 to the result.
1389 */
1390 return 1 + (it * (width - 1) / max_change);
1391 }
1392
1393 static void show_name(FILE *file,
1394 const char *prefix, const char *name, int len)
1395 {
1396 fprintf(file, " %s%-*s |", prefix, len, name);
1397 }
1398
1399 static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1400 {
1401 if (cnt <= 0)
1402 return;
1403 fprintf(file, "%s", set);
1404 while (cnt--)
1405 putc(ch, file);
1406 fprintf(file, "%s", reset);
1407 }
1408
1409 static void fill_print_name(struct diffstat_file *file)
1410 {
1411 char *pname;
1412
1413 if (file->print_name)
1414 return;
1415
1416 if (!file->is_renamed) {
1417 struct strbuf buf = STRBUF_INIT;
1418 if (quote_c_style(file->name, &buf, NULL, 0)) {
1419 pname = strbuf_detach(&buf, NULL);
1420 } else {
1421 pname = file->name;
1422 strbuf_release(&buf);
1423 }
1424 } else {
1425 pname = pprint_rename(file->from_name, file->name);
1426 }
1427 file->print_name = pname;
1428 }
1429
1430 int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1431 {
1432 struct strbuf sb = STRBUF_INIT;
1433 int ret;
1434
1435 if (!files) {
1436 assert(insertions == 0 && deletions == 0);
1437 return fprintf(fp, "%s\n", " 0 files changed");
1438 }
1439
1440 strbuf_addf(&sb,
1441 (files == 1) ? " %d file changed" : " %d files changed",
1442 files);
1443
1444 /*
1445 * For binary diff, the caller may want to print "x files
1446 * changed" with insertions == 0 && deletions == 0.
1447 *
1448 * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1449 * is probably less confusing (i.e skip over "2 files changed
1450 * but nothing about added/removed lines? Is this a bug in Git?").
1451 */
1452 if (insertions || deletions == 0) {
1453 /*
1454 * TRANSLATORS: "+" in (+) is a line addition marker;
1455 * do not translate it.
1456 */
1457 strbuf_addf(&sb,
1458 (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
1459 insertions);
1460 }
1461
1462 if (deletions || insertions == 0) {
1463 /*
1464 * TRANSLATORS: "-" in (-) is a line removal marker;
1465 * do not translate it.
1466 */
1467 strbuf_addf(&sb,
1468 (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
1469 deletions);
1470 }
1471 strbuf_addch(&sb, '\n');
1472 ret = fputs(sb.buf, fp);
1473 strbuf_release(&sb);
1474 return ret;
1475 }
1476
1477 static void show_stats(struct diffstat_t *data, struct diff_options *options)
1478 {
1479 int i, len, add, del, adds = 0, dels = 0;
1480 uintmax_t max_change = 0, max_len = 0;
1481 int total_files = data->nr, count;
1482 int width, name_width, graph_width, number_width = 0, bin_width = 0;
1483 const char *reset, *add_c, *del_c;
1484 const char *line_prefix = "";
1485 int extra_shown = 0;
1486
1487 if (data->nr == 0)
1488 return;
1489
1490 line_prefix = diff_line_prefix(options);
1491 count = options->stat_count ? options->stat_count : data->nr;
1492
1493 reset = diff_get_color_opt(options, DIFF_RESET);
1494 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1495 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1496
1497 /*
1498 * Find the longest filename and max number of changes
1499 */
1500 for (i = 0; (i < count) && (i < data->nr); i++) {
1501 struct diffstat_file *file = data->files[i];
1502 uintmax_t change = file->added + file->deleted;
1503
1504 if (!file->is_interesting && (change == 0)) {
1505 count++; /* not shown == room for one more */
1506 continue;
1507 }
1508 fill_print_name(file);
1509 len = strlen(file->print_name);
1510 if (max_len < len)
1511 max_len = len;
1512
1513 if (file->is_unmerged) {
1514 /* "Unmerged" is 8 characters */
1515 bin_width = bin_width < 8 ? 8 : bin_width;
1516 continue;
1517 }
1518 if (file->is_binary) {
1519 /* "Bin XXX -> YYY bytes" */
1520 int w = 14 + decimal_width(file->added)
1521 + decimal_width(file->deleted);
1522 bin_width = bin_width < w ? w : bin_width;
1523 /* Display change counts aligned with "Bin" */
1524 number_width = 3;
1525 continue;
1526 }
1527
1528 if (max_change < change)
1529 max_change = change;
1530 }
1531 count = i; /* where we can stop scanning in data->files[] */
1532
1533 /*
1534 * We have width = stat_width or term_columns() columns total.
1535 * We want a maximum of min(max_len, stat_name_width) for the name part.
1536 * We want a maximum of min(max_change, stat_graph_width) for the +- part.
1537 * We also need 1 for " " and 4 + decimal_width(max_change)
1538 * for " | NNNN " and one the empty column at the end, altogether
1539 * 6 + decimal_width(max_change).
1540 *
1541 * If there's not enough space, we will use the smaller of
1542 * stat_name_width (if set) and 5/8*width for the filename,
1543 * and the rest for constant elements + graph part, but no more
1544 * than stat_graph_width for the graph part.
1545 * (5/8 gives 50 for filename and 30 for the constant parts + graph
1546 * for the standard terminal size).
1547 *
1548 * In other words: stat_width limits the maximum width, and
1549 * stat_name_width fixes the maximum width of the filename,
1550 * and is also used to divide available columns if there
1551 * aren't enough.
1552 *
1553 * Binary files are displayed with "Bin XXX -> YYY bytes"
1554 * instead of the change count and graph. This part is treated
1555 * similarly to the graph part, except that it is not
1556 * "scaled". If total width is too small to accomodate the
1557 * guaranteed minimum width of the filename part and the
1558 * separators and this message, this message will "overflow"
1559 * making the line longer than the maximum width.
1560 */
1561
1562 if (options->stat_width == -1)
1563 width = term_columns() - options->output_prefix_length;
1564 else
1565 width = options->stat_width ? options->stat_width : 80;
1566 number_width = decimal_width(max_change) > number_width ?
1567 decimal_width(max_change) : number_width;
1568
1569 if (options->stat_graph_width == -1)
1570 options->stat_graph_width = diff_stat_graph_width;
1571
1572 /*
1573 * Guarantee 3/8*16==6 for the graph part
1574 * and 5/8*16==10 for the filename part
1575 */
1576 if (width < 16 + 6 + number_width)
1577 width = 16 + 6 + number_width;
1578
1579 /*
1580 * First assign sizes that are wanted, ignoring available width.
1581 * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
1582 * starting from "XXX" should fit in graph_width.
1583 */
1584 graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
1585 if (options->stat_graph_width &&
1586 options->stat_graph_width < graph_width)
1587 graph_width = options->stat_graph_width;
1588
1589 name_width = (options->stat_name_width > 0 &&
1590 options->stat_name_width < max_len) ?
1591 options->stat_name_width : max_len;
1592
1593 /*
1594 * Adjust adjustable widths not to exceed maximum width
1595 */
1596 if (name_width + number_width + 6 + graph_width > width) {
1597 if (graph_width > width * 3/8 - number_width - 6) {
1598 graph_width = width * 3/8 - number_width - 6;
1599 if (graph_width < 6)
1600 graph_width = 6;
1601 }
1602
1603 if (options->stat_graph_width &&
1604 graph_width > options->stat_graph_width)
1605 graph_width = options->stat_graph_width;
1606 if (name_width > width - number_width - 6 - graph_width)
1607 name_width = width - number_width - 6 - graph_width;
1608 else
1609 graph_width = width - number_width - 6 - name_width;
1610 }
1611
1612 /*
1613 * From here name_width is the width of the name area,
1614 * and graph_width is the width of the graph area.
1615 * max_change is used to scale graph properly.
1616 */
1617 for (i = 0; i < count; i++) {
1618 const char *prefix = "";
1619 struct diffstat_file *file = data->files[i];
1620 char *name = file->print_name;
1621 uintmax_t added = file->added;
1622 uintmax_t deleted = file->deleted;
1623 int name_len;
1624
1625 if (!file->is_interesting && (added + deleted == 0))
1626 continue;
1627
1628 /*
1629 * "scale" the filename
1630 */
1631 len = name_width;
1632 name_len = strlen(name);
1633 if (name_width < name_len) {
1634 char *slash;
1635 prefix = "...";
1636 len -= 3;
1637 name += name_len - len;
1638 slash = strchr(name, '/');
1639 if (slash)
1640 name = slash;
1641 }
1642
1643 if (file->is_binary) {
1644 fprintf(options->file, "%s", line_prefix);
1645 show_name(options->file, prefix, name, len);
1646 fprintf(options->file, " %*s", number_width, "Bin");
1647 if (!added && !deleted) {
1648 putc('\n', options->file);
1649 continue;
1650 }
1651 fprintf(options->file, " %s%"PRIuMAX"%s",
1652 del_c, deleted, reset);
1653 fprintf(options->file, " -> ");
1654 fprintf(options->file, "%s%"PRIuMAX"%s",
1655 add_c, added, reset);
1656 fprintf(options->file, " bytes");
1657 fprintf(options->file, "\n");
1658 continue;
1659 }
1660 else if (file->is_unmerged) {
1661 fprintf(options->file, "%s", line_prefix);
1662 show_name(options->file, prefix, name, len);
1663 fprintf(options->file, " Unmerged\n");
1664 continue;
1665 }
1666
1667 /*
1668 * scale the add/delete
1669 */
1670 add = added;
1671 del = deleted;
1672
1673 if (graph_width <= max_change) {
1674 int total = add + del;
1675
1676 total = scale_linear(add + del, graph_width, max_change);
1677 if (total < 2 && add && del)
1678 /* width >= 2 due to the sanity check */
1679 total = 2;
1680 if (add < del) {
1681 add = scale_linear(add, graph_width, max_change);
1682 del = total - add;
1683 } else {
1684 del = scale_linear(del, graph_width, max_change);
1685 add = total - del;
1686 }
1687 }
1688 fprintf(options->file, "%s", line_prefix);
1689 show_name(options->file, prefix, name, len);
1690 fprintf(options->file, " %*"PRIuMAX"%s",
1691 number_width, added + deleted,
1692 added + deleted ? " " : "");
1693 show_graph(options->file, '+', add, add_c, reset);
1694 show_graph(options->file, '-', del, del_c, reset);
1695 fprintf(options->file, "\n");
1696 }
1697
1698 for (i = 0; i < data->nr; i++) {
1699 struct diffstat_file *file = data->files[i];
1700 uintmax_t added = file->added;
1701 uintmax_t deleted = file->deleted;
1702
1703 if (file->is_unmerged ||
1704 (!file->is_interesting && (added + deleted == 0))) {
1705 total_files--;
1706 continue;
1707 }
1708
1709 if (!file->is_binary) {
1710 adds += added;
1711 dels += deleted;
1712 }
1713 if (i < count)
1714 continue;
1715 if (!extra_shown)
1716 fprintf(options->file, "%s ...\n", line_prefix);
1717 extra_shown = 1;
1718 }
1719 fprintf(options->file, "%s", line_prefix);
1720 print_stat_summary(options->file, total_files, adds, dels);
1721 }
1722
1723 static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1724 {
1725 int i, adds = 0, dels = 0, total_files = data->nr;
1726
1727 if (data->nr == 0)
1728 return;
1729
1730 for (i = 0; i < data->nr; i++) {
1731 int added = data->files[i]->added;
1732 int deleted= data->files[i]->deleted;
1733
1734 if (data->files[i]->is_unmerged ||
1735 (!data->files[i]->is_interesting && (added + deleted == 0))) {
1736 total_files--;
1737 } else if (!data->files[i]->is_binary) { /* don't count bytes */
1738 adds += added;
1739 dels += deleted;
1740 }
1741 }
1742 fprintf(options->file, "%s", diff_line_prefix(options));
1743 print_stat_summary(options->file, total_files, adds, dels);
1744 }
1745
1746 static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1747 {
1748 int i;
1749
1750 if (data->nr == 0)
1751 return;
1752
1753 for (i = 0; i < data->nr; i++) {
1754 struct diffstat_file *file = data->files[i];
1755
1756 fprintf(options->file, "%s", diff_line_prefix(options));
1757
1758 if (file->is_binary)
1759 fprintf(options->file, "-\t-\t");
1760 else
1761 fprintf(options->file,
1762 "%"PRIuMAX"\t%"PRIuMAX"\t",
1763 file->added, file->deleted);
1764 if (options->line_termination) {
1765 fill_print_name(file);
1766 if (!file->is_renamed)
1767 write_name_quoted(file->name, options->file,
1768 options->line_termination);
1769 else {
1770 fputs(file->print_name, options->file);
1771 putc(options->line_termination, options->file);
1772 }
1773 } else {
1774 if (file->is_renamed) {
1775 putc('\0', options->file);
1776 write_name_quoted(file->from_name, options->file, '\0');
1777 }
1778 write_name_quoted(file->name, options->file, '\0');
1779 }
1780 }
1781 }
1782
1783 struct dirstat_file {
1784 const char *name;
1785 unsigned long changed;
1786 };
1787
1788 struct dirstat_dir {
1789 struct dirstat_file *files;
1790 int alloc, nr, permille, cumulative;
1791 };
1792
1793 static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1794 unsigned long changed, const char *base, int baselen)
1795 {
1796 unsigned long this_dir = 0;
1797 unsigned int sources = 0;
1798 const char *line_prefix = diff_line_prefix(opt);
1799
1800 while (dir->nr) {
1801 struct dirstat_file *f = dir->files;
1802 int namelen = strlen(f->name);
1803 unsigned long this;
1804 char *slash;
1805
1806 if (namelen < baselen)
1807 break;
1808 if (memcmp(f->name, base, baselen))
1809 break;
1810 slash = strchr(f->name + baselen, '/');
1811 if (slash) {
1812 int newbaselen = slash + 1 - f->name;
1813 this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
1814 sources++;
1815 } else {
1816 this = f->changed;
1817 dir->files++;
1818 dir->nr--;
1819 sources += 2;
1820 }
1821 this_dir += this;
1822 }
1823
1824 /*
1825 * We don't report dirstat's for
1826 * - the top level
1827 * - or cases where everything came from a single directory
1828 * under this directory (sources == 1).
1829 */
1830 if (baselen && sources != 1) {
1831 if (this_dir) {
1832 int permille = this_dir * 1000 / changed;
1833 if (permille >= dir->permille) {
1834 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
1835 permille / 10, permille % 10, baselen, base);
1836 if (!dir->cumulative)
1837 return 0;
1838 }
1839 }
1840 }
1841 return this_dir;
1842 }
1843
1844 static int dirstat_compare(const void *_a, const void *_b)
1845 {
1846 const struct dirstat_file *a = _a;
1847 const struct dirstat_file *b = _b;
1848 return strcmp(a->name, b->name);
1849 }
1850
1851 static void show_dirstat(struct diff_options *options)
1852 {
1853 int i;
1854 unsigned long changed;
1855 struct dirstat_dir dir;
1856 struct diff_queue_struct *q = &diff_queued_diff;
1857
1858 dir.files = NULL;
1859 dir.alloc = 0;
1860 dir.nr = 0;
1861 dir.permille = options->dirstat_permille;
1862 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1863
1864 changed = 0;
1865 for (i = 0; i < q->nr; i++) {
1866 struct diff_filepair *p = q->queue[i];
1867 const char *name;
1868 unsigned long copied, added, damage;
1869 int content_changed;
1870
1871 name = p->two->path ? p->two->path : p->one->path;
1872
1873 if (p->one->sha1_valid && p->two->sha1_valid)
1874 content_changed = hashcmp(p->one->sha1, p->two->sha1);
1875 else
1876 content_changed = 1;
1877
1878 if (!content_changed) {
1879 /*
1880 * The SHA1 has not changed, so pre-/post-content is
1881 * identical. We can therefore skip looking at the
1882 * file contents altogether.
1883 */
1884 damage = 0;
1885 goto found_damage;
1886 }
1887
1888 if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
1889 /*
1890 * In --dirstat-by-file mode, we don't really need to
1891 * look at the actual file contents at all.
1892 * The fact that the SHA1 changed is enough for us to
1893 * add this file to the list of results
1894 * (with each file contributing equal damage).
1895 */
1896 damage = 1;
1897 goto found_damage;
1898 }
1899
1900 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
1901 diff_populate_filespec(p->one, 0);
1902 diff_populate_filespec(p->two, 0);
1903 diffcore_count_changes(p->one, p->two, NULL, NULL, 0,
1904 &copied, &added);
1905 diff_free_filespec_data(p->one);
1906 diff_free_filespec_data(p->two);
1907 } else if (DIFF_FILE_VALID(p->one)) {
1908 diff_populate_filespec(p->one, 1);
1909 copied = added = 0;
1910 diff_free_filespec_data(p->one);
1911 } else if (DIFF_FILE_VALID(p->two)) {
1912 diff_populate_filespec(p->two, 1);
1913 copied = 0;
1914 added = p->two->size;
1915 diff_free_filespec_data(p->two);
1916 } else
1917 continue;
1918
1919 /*
1920 * Original minus copied is the removed material,
1921 * added is the new material. They are both damages
1922 * made to the preimage.
1923 * If the resulting damage is zero, we know that
1924 * diffcore_count_changes() considers the two entries to
1925 * be identical, but since content_changed is true, we
1926 * know that there must have been _some_ kind of change,
1927 * so we force all entries to have damage > 0.
1928 */
1929 damage = (p->one->size - copied) + added;
1930 if (!damage)
1931 damage = 1;
1932
1933 found_damage:
1934 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1935 dir.files[dir.nr].name = name;
1936 dir.files[dir.nr].changed = damage;
1937 changed += damage;
1938 dir.nr++;
1939 }
1940
1941 /* This can happen even with many files, if everything was renames */
1942 if (!changed)
1943 return;
1944
1945 /* Show all directories with more than x% of the changes */
1946 qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1947 gather_dirstat(options, &dir, changed, "", 0);
1948 }
1949
1950 static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
1951 {
1952 int i;
1953 unsigned long changed;
1954 struct dirstat_dir dir;
1955
1956 if (data->nr == 0)
1957 return;
1958
1959 dir.files = NULL;
1960 dir.alloc = 0;
1961 dir.nr = 0;
1962 dir.permille = options->dirstat_permille;
1963 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1964
1965 changed = 0;
1966 for (i = 0; i < data->nr; i++) {
1967 struct diffstat_file *file = data->files[i];
1968 unsigned long damage = file->added + file->deleted;
1969 if (file->is_binary)
1970 /*
1971 * binary files counts bytes, not lines. Must find some
1972 * way to normalize binary bytes vs. textual lines.
1973 * The following heuristic assumes that there are 64
1974 * bytes per "line".
1975 * This is stupid and ugly, but very cheap...
1976 */
1977 damage = (damage + 63) / 64;
1978 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1979 dir.files[dir.nr].name = file->name;
1980 dir.files[dir.nr].changed = damage;
1981 changed += damage;
1982 dir.nr++;
1983 }
1984
1985 /* This can happen even with many files, if everything was renames */
1986 if (!changed)
1987 return;
1988
1989 /* Show all directories with more than x% of the changes */
1990 qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1991 gather_dirstat(options, &dir, changed, "", 0);
1992 }
1993
1994 static void free_diffstat_info(struct diffstat_t *diffstat)
1995 {
1996 int i;
1997 for (i = 0; i < diffstat->nr; i++) {
1998 struct diffstat_file *f = diffstat->files[i];
1999 if (f->name != f->print_name)
2000 free(f->print_name);
2001 free(f->name);
2002 free(f->from_name);
2003 free(f);
2004 }
2005 free(diffstat->files);
2006 }
2007
2008 struct checkdiff_t {
2009 const char *filename;
2010 int lineno;
2011 int conflict_marker_size;
2012 struct diff_options *o;
2013 unsigned ws_rule;
2014 unsigned status;
2015 };
2016
2017 static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2018 {
2019 char firstchar;
2020 int cnt;
2021
2022 if (len < marker_size + 1)
2023 return 0;
2024 firstchar = line[0];
2025 switch (firstchar) {
2026 case '=': case '>': case '<': case '|':
2027 break;
2028 default:
2029 return 0;
2030 }
2031 for (cnt = 1; cnt < marker_size; cnt++)
2032 if (line[cnt] != firstchar)
2033 return 0;
2034 /* line[1] thru line[marker_size-1] are same as firstchar */
2035 if (len < marker_size + 1 || !isspace(line[marker_size]))
2036 return 0;
2037 return 1;
2038 }
2039
2040 static void checkdiff_consume(void *priv, char *line, unsigned long len)
2041 {
2042 struct checkdiff_t *data = priv;
2043 int marker_size = data->conflict_marker_size;
2044 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2045 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2046 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2047 char *err;
2048 const char *line_prefix;
2049
2050 assert(data->o);
2051 line_prefix = diff_line_prefix(data->o);
2052
2053 if (line[0] == '+') {
2054 unsigned bad;
2055 data->lineno++;
2056 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2057 data->status |= 1;
2058 fprintf(data->o->file,
2059 "%s%s:%d: leftover conflict marker\n",
2060 line_prefix, data->filename, data->lineno);
2061 }
2062 bad = ws_check(line + 1, len - 1, data->ws_rule);
2063 if (!bad)
2064 return;
2065 data->status |= bad;
2066 err = whitespace_error_string(bad);
2067 fprintf(data->o->file, "%s%s:%d: %s.\n",
2068 line_prefix, data->filename, data->lineno, err);
2069 free(err);
2070 emit_line(data->o, set, reset, line, 1);
2071 ws_check_emit(line + 1, len - 1, data->ws_rule,
2072 data->o->file, set, reset, ws);
2073 } else if (line[0] == ' ') {
2074 data->lineno++;
2075 } else if (line[0] == '@') {
2076 char *plus = strchr(line, '+');
2077 if (plus)
2078 data->lineno = strtol(plus, NULL, 10) - 1;
2079 else
2080 die("invalid diff");
2081 }
2082 }
2083
2084 static unsigned char *deflate_it(char *data,
2085 unsigned long size,
2086 unsigned long *result_size)
2087 {
2088 int bound;
2089 unsigned char *deflated;
2090 git_zstream stream;
2091
2092 memset(&stream, 0, sizeof(stream));
2093 git_deflate_init(&stream, zlib_compression_level);
2094 bound = git_deflate_bound(&stream, size);
2095 deflated = xmalloc(bound);
2096 stream.next_out = deflated;
2097 stream.avail_out = bound;
2098
2099 stream.next_in = (unsigned char *)data;
2100 stream.avail_in = size;
2101 while (git_deflate(&stream, Z_FINISH) == Z_OK)
2102 ; /* nothing */
2103 git_deflate_end(&stream);
2104 *result_size = stream.total_out;
2105 return deflated;
2106 }
2107
2108 static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two,
2109 const char *prefix)
2110 {
2111 void *cp;
2112 void *delta;
2113 void *deflated;
2114 void *data;
2115 unsigned long orig_size;
2116 unsigned long delta_size;
2117 unsigned long deflate_size;
2118 unsigned long data_size;
2119
2120 /* We could do deflated delta, or we could do just deflated two,
2121 * whichever is smaller.
2122 */
2123 delta = NULL;
2124 deflated = deflate_it(two->ptr, two->size, &deflate_size);
2125 if (one->size && two->size) {
2126 delta = diff_delta(one->ptr, one->size,
2127 two->ptr, two->size,
2128 &delta_size, deflate_size);
2129 if (delta) {
2130 void *to_free = delta;
2131 orig_size = delta_size;
2132 delta = deflate_it(delta, delta_size, &delta_size);
2133 free(to_free);
2134 }
2135 }
2136
2137 if (delta && delta_size < deflate_size) {
2138 fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2139 free(deflated);
2140 data = delta;
2141 data_size = delta_size;
2142 }
2143 else {
2144 fprintf(file, "%sliteral %lu\n", prefix, two->size);
2145 free(delta);
2146 data = deflated;
2147 data_size = deflate_size;
2148 }
2149
2150 /* emit data encoded in base85 */
2151 cp = data;
2152 while (data_size) {
2153 int bytes = (52 < data_size) ? 52 : data_size;
2154 char line[70];
2155 data_size -= bytes;
2156 if (bytes <= 26)
2157 line[0] = bytes + 'A' - 1;
2158 else
2159 line[0] = bytes - 26 + 'a' - 1;
2160 encode_85(line + 1, cp, bytes);
2161 cp = (char *) cp + bytes;
2162 fprintf(file, "%s", prefix);
2163 fputs(line, file);
2164 fputc('\n', file);
2165 }
2166 fprintf(file, "%s\n", prefix);
2167 free(data);
2168 }
2169
2170 static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two,
2171 const char *prefix)
2172 {
2173 fprintf(file, "%sGIT binary patch\n", prefix);
2174 emit_binary_diff_body(file, one, two, prefix);
2175 emit_binary_diff_body(file, two, one, prefix);
2176 }
2177
2178 int diff_filespec_is_binary(struct diff_filespec *one)
2179 {
2180 if (one->is_binary == -1) {
2181 diff_filespec_load_driver(one);
2182 if (one->driver->binary != -1)
2183 one->is_binary = one->driver->binary;
2184 else {
2185 if (!one->data && DIFF_FILE_VALID(one))
2186 diff_populate_filespec(one, 0);
2187 if (one->data)
2188 one->is_binary = buffer_is_binary(one->data,
2189 one->size);
2190 if (one->is_binary == -1)
2191 one->is_binary = 0;
2192 }
2193 }
2194 return one->is_binary;
2195 }
2196
2197 static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2198 {
2199 diff_filespec_load_driver(one);
2200 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2201 }
2202
2203 void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2204 {
2205 if (!options->a_prefix)
2206 options->a_prefix = a;
2207 if (!options->b_prefix)
2208 options->b_prefix = b;
2209 }
2210
2211 struct userdiff_driver *get_textconv(struct diff_filespec *one)
2212 {
2213 if (!DIFF_FILE_VALID(one))
2214 return NULL;
2215
2216 diff_filespec_load_driver(one);
2217 return userdiff_get_textconv(one->driver);
2218 }
2219
2220 static void builtin_diff(const char *name_a,
2221 const char *name_b,
2222 struct diff_filespec *one,
2223 struct diff_filespec *two,
2224 const char *xfrm_msg,
2225 int must_show_header,
2226 struct diff_options *o,
2227 int complete_rewrite)
2228 {
2229 mmfile_t mf1, mf2;
2230 const char *lbl[2];
2231 char *a_one, *b_two;
2232 const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
2233 const char *reset = diff_get_color_opt(o, DIFF_RESET);
2234 const char *a_prefix, *b_prefix;
2235 struct userdiff_driver *textconv_one = NULL;
2236 struct userdiff_driver *textconv_two = NULL;
2237 struct strbuf header = STRBUF_INIT;
2238 const char *line_prefix = diff_line_prefix(o);
2239
2240 if (DIFF_OPT_TST(o, SUBMODULE_LOG) &&
2241 (!one->mode || S_ISGITLINK(one->mode)) &&
2242 (!two->mode || S_ISGITLINK(two->mode))) {
2243 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2244 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2245 show_submodule_summary(o->file, one ? one->path : two->path,
2246 one->sha1, two->sha1, two->dirty_submodule,
2247 meta, del, add, reset);
2248 return;
2249 }
2250
2251 if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2252 textconv_one = get_textconv(one);
2253 textconv_two = get_textconv(two);
2254 }
2255
2256 diff_set_mnemonic_prefix(o, "a/", "b/");
2257 if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2258 a_prefix = o->b_prefix;
2259 b_prefix = o->a_prefix;
2260 } else {
2261 a_prefix = o->a_prefix;
2262 b_prefix = o->b_prefix;
2263 }
2264
2265 /* Never use a non-valid filename anywhere if at all possible */
2266 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2267 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2268
2269 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2270 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2271 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2272 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2273 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
2274 if (lbl[0][0] == '/') {
2275 /* /dev/null */
2276 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
2277 if (xfrm_msg)
2278 strbuf_addstr(&header, xfrm_msg);
2279 must_show_header = 1;
2280 }
2281 else if (lbl[1][0] == '/') {
2282 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
2283 if (xfrm_msg)
2284 strbuf_addstr(&header, xfrm_msg);
2285 must_show_header = 1;
2286 }
2287 else {
2288 if (one->mode != two->mode) {
2289 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
2290 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
2291 must_show_header = 1;
2292 }
2293 if (xfrm_msg)
2294 strbuf_addstr(&header, xfrm_msg);
2295
2296 /*
2297 * we do not run diff between different kind
2298 * of objects.
2299 */
2300 if ((one->mode ^ two->mode) & S_IFMT)
2301 goto free_ab_and_return;
2302 if (complete_rewrite &&
2303 (textconv_one || !diff_filespec_is_binary(one)) &&
2304 (textconv_two || !diff_filespec_is_binary(two))) {
2305 fprintf(o->file, "%s", header.buf);
2306 strbuf_reset(&header);
2307 emit_rewrite_diff(name_a, name_b, one, two,
2308 textconv_one, textconv_two, o);
2309 o->found_changes = 1;
2310 goto free_ab_and_return;
2311 }
2312 }
2313
2314 if (o->irreversible_delete && lbl[1][0] == '/') {
2315 fprintf(o->file, "%s", header.buf);
2316 strbuf_reset(&header);
2317 goto free_ab_and_return;
2318 } else if (!DIFF_OPT_TST(o, TEXT) &&
2319 ( (!textconv_one && diff_filespec_is_binary(one)) ||
2320 (!textconv_two && diff_filespec_is_binary(two)) )) {
2321 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2322 die("unable to read files to diff");
2323 /* Quite common confusing case */
2324 if (mf1.size == mf2.size &&
2325 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2326 if (must_show_header)
2327 fprintf(o->file, "%s", header.buf);
2328 goto free_ab_and_return;
2329 }
2330 fprintf(o->file, "%s", header.buf);
2331 strbuf_reset(&header);
2332 if (DIFF_OPT_TST(o, BINARY))
2333 emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2334 else
2335 fprintf(o->file, "%sBinary files %s and %s differ\n",
2336 line_prefix, lbl[0], lbl[1]);
2337 o->found_changes = 1;
2338 } else {
2339 /* Crazy xdl interfaces.. */
2340 const char *diffopts = getenv("GIT_DIFF_OPTS");
2341 xpparam_t xpp;
2342 xdemitconf_t xecfg;
2343 struct emit_callback ecbdata;
2344 const struct userdiff_funcname *pe;
2345
2346 if (must_show_header) {
2347 fprintf(o->file, "%s", header.buf);
2348 strbuf_reset(&header);
2349 }
2350
2351 mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2352 mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2353
2354 pe = diff_funcname_pattern(one);
2355 if (!pe)
2356 pe = diff_funcname_pattern(two);
2357
2358 memset(&xpp, 0, sizeof(xpp));
2359 memset(&xecfg, 0, sizeof(xecfg));
2360 memset(&ecbdata, 0, sizeof(ecbdata));
2361 ecbdata.label_path = lbl;
2362 ecbdata.color_diff = want_color(o->use_color);
2363 ecbdata.found_changesp = &o->found_changes;
2364 ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
2365 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2366 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2367 ecbdata.opt = o;
2368 ecbdata.header = header.len ? &header : NULL;
2369 xpp.flags = o->xdl_opts;
2370 xecfg.ctxlen = o->context;
2371 xecfg.interhunkctxlen = o->interhunkcontext;
2372 xecfg.flags = XDL_EMIT_FUNCNAMES;
2373 if (DIFF_OPT_TST(o, FUNCCONTEXT))
2374 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2375 if (pe)
2376 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2377 if (!diffopts)
2378 ;
2379 else if (!prefixcmp(diffopts, "--unified="))
2380 xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
2381 else if (!prefixcmp(diffopts, "-u"))
2382 xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
2383 if (o->word_diff)
2384 init_diff_words_data(&ecbdata, o, one, two);
2385 xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2386 &xpp, &xecfg);
2387 if (o->word_diff)
2388 free_diff_words_data(&ecbdata);
2389 if (textconv_one)
2390 free(mf1.ptr);
2391 if (textconv_two)
2392 free(mf2.ptr);
2393 xdiff_clear_find_func(&xecfg);
2394 }
2395
2396 free_ab_and_return:
2397 strbuf_release(&header);
2398 diff_free_filespec_data(one);
2399 diff_free_filespec_data(two);
2400 free(a_one);
2401 free(b_two);
2402 return;
2403 }
2404
2405 static void builtin_diffstat(const char *name_a, const char *name_b,
2406 struct diff_filespec *one,
2407 struct diff_filespec *two,
2408 struct diffstat_t *diffstat,
2409 struct diff_options *o,
2410 struct diff_filepair *p)
2411 {
2412 mmfile_t mf1, mf2;
2413 struct diffstat_file *data;
2414 int same_contents;
2415 int complete_rewrite = 0;
2416
2417 if (!DIFF_PAIR_UNMERGED(p)) {
2418 if (p->status == DIFF_STATUS_MODIFIED && p->score)
2419 complete_rewrite = 1;
2420 }
2421
2422 data = diffstat_add(diffstat, name_a, name_b);
2423 data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
2424
2425 if (!one || !two) {
2426 data->is_unmerged = 1;
2427 return;
2428 }
2429
2430 same_contents = !hashcmp(one->sha1, two->sha1);
2431
2432 if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2433 data->is_binary = 1;
2434 if (same_contents) {
2435 data->added = 0;
2436 data->deleted = 0;
2437 } else {
2438 data->added = diff_filespec_size(two);
2439 data->deleted = diff_filespec_size(one);
2440 }
2441 }
2442
2443 else if (complete_rewrite) {
2444 diff_populate_filespec(one, 0);
2445 diff_populate_filespec(two, 0);
2446 data->deleted = count_lines(one->data, one->size);
2447 data->added = count_lines(two->data, two->size);
2448 }
2449
2450 else if (!same_contents) {
2451 /* Crazy xdl interfaces.. */
2452 xpparam_t xpp;
2453 xdemitconf_t xecfg;
2454
2455 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2456 die("unable to read files to diff");
2457
2458 memset(&xpp, 0, sizeof(xpp));
2459 memset(&xecfg, 0, sizeof(xecfg));
2460 xpp.flags = o->xdl_opts;
2461 xecfg.ctxlen = o->context;
2462 xecfg.interhunkctxlen = o->interhunkcontext;
2463 xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2464 &xpp, &xecfg);
2465 }
2466
2467 diff_free_filespec_data(one);
2468 diff_free_filespec_data(two);
2469 }
2470
2471 static void builtin_checkdiff(const char *name_a, const char *name_b,
2472 const char *attr_path,
2473 struct diff_filespec *one,
2474 struct diff_filespec *two,
2475 struct diff_options *o)
2476 {
2477 mmfile_t mf1, mf2;
2478 struct checkdiff_t data;
2479
2480 if (!two)
2481 return;
2482
2483 memset(&data, 0, sizeof(data));
2484 data.filename = name_b ? name_b : name_a;
2485 data.lineno = 0;
2486 data.o = o;
2487 data.ws_rule = whitespace_rule(attr_path);
2488 data.conflict_marker_size = ll_merge_marker_size(attr_path);
2489
2490 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2491 die("unable to read files to diff");
2492
2493 /*
2494 * All the other codepaths check both sides, but not checking
2495 * the "old" side here is deliberate. We are checking the newly
2496 * introduced changes, and as long as the "new" side is text, we
2497 * can and should check what it introduces.
2498 */
2499 if (diff_filespec_is_binary(two))
2500 goto free_and_return;
2501 else {
2502 /* Crazy xdl interfaces.. */
2503 xpparam_t xpp;
2504 xdemitconf_t xecfg;
2505
2506 memset(&xpp, 0, sizeof(xpp));
2507 memset(&xecfg, 0, sizeof(xecfg));
2508 xecfg.ctxlen = 1; /* at least one context line */
2509 xpp.flags = 0;
2510 xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2511 &xpp, &xecfg);
2512
2513 if (data.ws_rule & WS_BLANK_AT_EOF) {
2514 struct emit_callback ecbdata;
2515 int blank_at_eof;
2516
2517 ecbdata.ws_rule = data.ws_rule;
2518 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2519 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2520
2521 if (blank_at_eof) {
2522 static char *err;
2523 if (!err)
2524 err = whitespace_error_string(WS_BLANK_AT_EOF);
2525 fprintf(o->file, "%s:%d: %s.\n",
2526 data.filename, blank_at_eof, err);
2527 data.status = 1; /* report errors */
2528 }
2529 }
2530 }
2531 free_and_return:
2532 diff_free_filespec_data(one);
2533 diff_free_filespec_data(two);
2534 if (data.status)
2535 DIFF_OPT_SET(o, CHECK_FAILED);
2536 }
2537
2538 struct diff_filespec *alloc_filespec(const char *path)
2539 {
2540 int namelen = strlen(path);
2541 struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
2542
2543 memset(spec, 0, sizeof(*spec));
2544 spec->path = (char *)(spec + 1);
2545 memcpy(spec->path, path, namelen+1);
2546 spec->count = 1;
2547 spec->is_binary = -1;
2548 return spec;
2549 }
2550
2551 void free_filespec(struct diff_filespec *spec)
2552 {
2553 if (!--spec->count) {
2554 diff_free_filespec_data(spec);
2555 free(spec);
2556 }
2557 }
2558
2559 void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
2560 int sha1_valid, unsigned short mode)
2561 {
2562 if (mode) {
2563 spec->mode = canon_mode(mode);
2564 hashcpy(spec->sha1, sha1);
2565 spec->sha1_valid = sha1_valid;
2566 }
2567 }
2568
2569 /*
2570 * Given a name and sha1 pair, if the index tells us the file in
2571 * the work tree has that object contents, return true, so that
2572 * prepare_temp_file() does not have to inflate and extract.
2573 */
2574 static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
2575 {
2576 struct cache_entry *ce;
2577 struct stat st;
2578 int pos, len;
2579
2580 /*
2581 * We do not read the cache ourselves here, because the
2582 * benchmark with my previous version that always reads cache
2583 * shows that it makes things worse for diff-tree comparing
2584 * two linux-2.6 kernel trees in an already checked out work
2585 * tree. This is because most diff-tree comparisons deal with
2586 * only a small number of files, while reading the cache is
2587 * expensive for a large project, and its cost outweighs the
2588 * savings we get by not inflating the object to a temporary
2589 * file. Practically, this code only helps when we are used
2590 * by diff-cache --cached, which does read the cache before
2591 * calling us.
2592 */
2593 if (!active_cache)
2594 return 0;
2595
2596 /* We want to avoid the working directory if our caller
2597 * doesn't need the data in a normal file, this system
2598 * is rather slow with its stat/open/mmap/close syscalls,
2599 * and the object is contained in a pack file. The pack
2600 * is probably already open and will be faster to obtain
2601 * the data through than the working directory. Loose
2602 * objects however would tend to be slower as they need
2603 * to be individually opened and inflated.
2604 */
2605 if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1))
2606 return 0;
2607
2608 len = strlen(name);
2609 pos = cache_name_pos(name, len);
2610 if (pos < 0)
2611 return 0;
2612 ce = active_cache[pos];
2613
2614 /*
2615 * This is not the sha1 we are looking for, or
2616 * unreusable because it is not a regular file.
2617 */
2618 if (hashcmp(sha1, ce->sha1) || !S_ISREG(ce->ce_mode))
2619 return 0;
2620
2621 /*
2622 * If ce is marked as "assume unchanged", there is no
2623 * guarantee that work tree matches what we are looking for.
2624 */
2625 if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2626 return 0;
2627
2628 /*
2629 * If ce matches the file in the work tree, we can reuse it.
2630 */
2631 if (ce_uptodate(ce) ||
2632 (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2633 return 1;
2634
2635 return 0;
2636 }
2637
2638 static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2639 {
2640 int len;
2641 char *data = xmalloc(100), *dirty = "";
2642
2643 /* Are we looking at the work tree? */
2644 if (s->dirty_submodule)
2645 dirty = "-dirty";
2646
2647 len = snprintf(data, 100,
2648 "Subproject commit %s%s\n", sha1_to_hex(s->sha1), dirty);
2649 s->data = data;
2650 s->size = len;
2651 s->should_free = 1;
2652 if (size_only) {
2653 s->data = NULL;
2654 free(data);
2655 }
2656 return 0;
2657 }
2658
2659 /*
2660 * While doing rename detection and pickaxe operation, we may need to
2661 * grab the data for the blob (or file) for our own in-core comparison.
2662 * diff_filespec has data and size fields for this purpose.
2663 */
2664 int diff_populate_filespec(struct diff_filespec *s, int size_only)
2665 {
2666 int err = 0;
2667 if (!DIFF_FILE_VALID(s))
2668 die("internal error: asking to populate invalid file.");
2669 if (S_ISDIR(s->mode))
2670 return -1;
2671
2672 if (s->data)
2673 return 0;
2674
2675 if (size_only && 0 < s->size)
2676 return 0;
2677
2678 if (S_ISGITLINK(s->mode))
2679 return diff_populate_gitlink(s, size_only);
2680
2681 if (!s->sha1_valid ||
2682 reuse_worktree_file(s->path, s->sha1, 0)) {
2683 struct strbuf buf = STRBUF_INIT;
2684 struct stat st;
2685 int fd;
2686
2687 if (lstat(s->path, &st) < 0) {
2688 if (errno == ENOENT) {
2689 err_empty:
2690 err = -1;
2691 empty:
2692 s->data = (char *)"";
2693 s->size = 0;
2694 return err;
2695 }
2696 }
2697 s->size = xsize_t(st.st_size);
2698 if (!s->size)
2699 goto empty;
2700 if (S_ISLNK(st.st_mode)) {
2701 struct strbuf sb = STRBUF_INIT;
2702
2703 if (strbuf_readlink(&sb, s->path, s->size))
2704 goto err_empty;
2705 s->size = sb.len;
2706 s->data = strbuf_detach(&sb, NULL);
2707 s->should_free = 1;
2708 return 0;
2709 }
2710 if (size_only)
2711 return 0;
2712 fd = open(s->path, O_RDONLY);
2713 if (fd < 0)
2714 goto err_empty;
2715 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2716 close(fd);
2717 s->should_munmap = 1;
2718
2719 /*
2720 * Convert from working tree format to canonical git format
2721 */
2722 if (convert_to_git(s->path, s->data, s->size, &buf, safe_crlf)) {
2723 size_t size = 0;
2724 munmap(s->data, s->size);
2725 s->should_munmap = 0;
2726 s->data = strbuf_detach(&buf, &size);
2727 s->size = size;
2728 s->should_free = 1;
2729 }
2730 }
2731 else {
2732 enum object_type type;
2733 if (size_only) {
2734 type = sha1_object_info(s->sha1, &s->size);
2735 if (type < 0)
2736 die("unable to read %s", sha1_to_hex(s->sha1));
2737 } else {
2738 s->data = read_sha1_file(s->sha1, &type, &s->size);
2739 if (!s->data)
2740 die("unable to read %s", sha1_to_hex(s->sha1));
2741 s->should_free = 1;
2742 }
2743 }
2744 return 0;
2745 }
2746
2747 void diff_free_filespec_blob(struct diff_filespec *s)
2748 {
2749 if (s->should_free)
2750 free(s->data);
2751 else if (s->should_munmap)
2752 munmap(s->data, s->size);
2753
2754 if (s->should_free || s->should_munmap) {
2755 s->should_free = s->should_munmap = 0;
2756 s->data = NULL;
2757 }
2758 }
2759
2760 void diff_free_filespec_data(struct diff_filespec *s)
2761 {
2762 diff_free_filespec_blob(s);
2763 free(s->cnt_data);
2764 s->cnt_data = NULL;
2765 }
2766
2767 static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2768 void *blob,
2769 unsigned long size,
2770 const unsigned char *sha1,
2771 int mode)
2772 {
2773 int fd;
2774 struct strbuf buf = STRBUF_INIT;
2775 struct strbuf template = STRBUF_INIT;
2776 char *path_dup = xstrdup(path);
2777 const char *base = basename(path_dup);
2778
2779 /* Generate "XXXXXX_basename.ext" */
2780 strbuf_addstr(&template, "XXXXXX_");
2781 strbuf_addstr(&template, base);
2782
2783 fd = git_mkstemps(temp->tmp_path, PATH_MAX, template.buf,
2784 strlen(base) + 1);
2785 if (fd < 0)
2786 die_errno("unable to create temp-file");
2787 if (convert_to_working_tree(path,
2788 (const char *)blob, (size_t)size, &buf)) {
2789 blob = buf.buf;
2790 size = buf.len;
2791 }
2792 if (write_in_full(fd, blob, size) != size)
2793 die_errno("unable to write temp-file");
2794 close(fd);
2795 temp->name = temp->tmp_path;
2796 strcpy(temp->hex, sha1_to_hex(sha1));
2797 temp->hex[40] = 0;
2798 sprintf(temp->mode, "%06o", mode);
2799 strbuf_release(&buf);
2800 strbuf_release(&template);
2801 free(path_dup);
2802 }
2803
2804 static struct diff_tempfile *prepare_temp_file(const char *name,
2805 struct diff_filespec *one)
2806 {
2807 struct diff_tempfile *temp = claim_diff_tempfile();
2808
2809 if (!DIFF_FILE_VALID(one)) {
2810 not_a_valid_file:
2811 /* A '-' entry produces this for file-2, and
2812 * a '+' entry produces this for file-1.
2813 */
2814 temp->name = "/dev/null";
2815 strcpy(temp->hex, ".");
2816 strcpy(temp->mode, ".");
2817 return temp;
2818 }
2819
2820 if (!remove_tempfile_installed) {
2821 atexit(remove_tempfile);
2822 sigchain_push_common(remove_tempfile_on_signal);
2823 remove_tempfile_installed = 1;
2824 }
2825
2826 if (!one->sha1_valid ||
2827 reuse_worktree_file(name, one->sha1, 1)) {
2828 struct stat st;
2829 if (lstat(name, &st) < 0) {
2830 if (errno == ENOENT)
2831 goto not_a_valid_file;
2832 die_errno("stat(%s)", name);
2833 }
2834 if (S_ISLNK(st.st_mode)) {
2835 struct strbuf sb = STRBUF_INIT;
2836 if (strbuf_readlink(&sb, name, st.st_size) < 0)
2837 die_errno("readlink(%s)", name);
2838 prep_temp_blob(name, temp, sb.buf, sb.len,
2839 (one->sha1_valid ?
2840 one->sha1 : null_sha1),
2841 (one->sha1_valid ?
2842 one->mode : S_IFLNK));
2843 strbuf_release(&sb);
2844 }
2845 else {
2846 /* we can borrow from the file in the work tree */
2847 temp->name = name;
2848 if (!one->sha1_valid)
2849 strcpy(temp->hex, sha1_to_hex(null_sha1));
2850 else
2851 strcpy(temp->hex, sha1_to_hex(one->sha1));
2852 /* Even though we may sometimes borrow the
2853 * contents from the work tree, we always want
2854 * one->mode. mode is trustworthy even when
2855 * !(one->sha1_valid), as long as
2856 * DIFF_FILE_VALID(one).
2857 */
2858 sprintf(temp->mode, "%06o", one->mode);
2859 }
2860 return temp;
2861 }
2862 else {
2863 if (diff_populate_filespec(one, 0))
2864 die("cannot read data blob for %s", one->path);
2865 prep_temp_blob(name, temp, one->data, one->size,
2866 one->sha1, one->mode);
2867 }
2868 return temp;
2869 }
2870
2871 /* An external diff command takes:
2872 *
2873 * diff-cmd name infile1 infile1-sha1 infile1-mode \
2874 * infile2 infile2-sha1 infile2-mode [ rename-to ]
2875 *
2876 */
2877 static void run_external_diff(const char *pgm,
2878 const char *name,
2879 const char *other,
2880 struct diff_filespec *one,
2881 struct diff_filespec *two,
2882 const char *xfrm_msg,
2883 int complete_rewrite)
2884 {
2885 const char *spawn_arg[10];
2886 int retval;
2887 const char **arg = &spawn_arg[0];
2888
2889 if (one && two) {
2890 struct diff_tempfile *temp_one, *temp_two;
2891 const char *othername = (other ? other : name);
2892 temp_one = prepare_temp_file(name, one);
2893 temp_two = prepare_temp_file(othername, two);
2894 *arg++ = pgm;
2895 *arg++ = name;
2896 *arg++ = temp_one->name;
2897 *arg++ = temp_one->hex;
2898 *arg++ = temp_one->mode;
2899 *arg++ = temp_two->name;
2900 *arg++ = temp_two->hex;
2901 *arg++ = temp_two->mode;
2902 if (other) {
2903 *arg++ = other;
2904 *arg++ = xfrm_msg;
2905 }
2906 } else {
2907 *arg++ = pgm;
2908 *arg++ = name;
2909 }
2910 *arg = NULL;
2911 fflush(NULL);
2912 retval = run_command_v_opt(spawn_arg, RUN_USING_SHELL);
2913 remove_tempfile();
2914 if (retval) {
2915 fprintf(stderr, "external diff died, stopping at %s.\n", name);
2916 exit(1);
2917 }
2918 }
2919
2920 static int similarity_index(struct diff_filepair *p)
2921 {
2922 return p->score * 100 / MAX_SCORE;
2923 }
2924
2925 static void fill_metainfo(struct strbuf *msg,
2926 const char *name,
2927 const char *other,
2928 struct diff_filespec *one,
2929 struct diff_filespec *two,
2930 struct diff_options *o,
2931 struct diff_filepair *p,
2932 int *must_show_header,
2933 int use_color)
2934 {
2935 const char *set = diff_get_color(use_color, DIFF_METAINFO);
2936 const char *reset = diff_get_color(use_color, DIFF_RESET);
2937 const char *line_prefix = diff_line_prefix(o);
2938
2939 *must_show_header = 1;
2940 strbuf_init(msg, PATH_MAX * 2 + 300);
2941 switch (p->status) {
2942 case DIFF_STATUS_COPIED:
2943 strbuf_addf(msg, "%s%ssimilarity index %d%%",
2944 line_prefix, set, similarity_index(p));
2945 strbuf_addf(msg, "%s\n%s%scopy from ",
2946 reset, line_prefix, set);
2947 quote_c_style(name, msg, NULL, 0);
2948 strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
2949 quote_c_style(other, msg, NULL, 0);
2950 strbuf_addf(msg, "%s\n", reset);
2951 break;
2952 case DIFF_STATUS_RENAMED:
2953 strbuf_addf(msg, "%s%ssimilarity index %d%%",
2954 line_prefix, set, similarity_index(p));
2955 strbuf_addf(msg, "%s\n%s%srename from ",
2956 reset, line_prefix, set);
2957 quote_c_style(name, msg, NULL, 0);
2958 strbuf_addf(msg, "%s\n%s%srename to ",
2959 reset, line_prefix, set);
2960 quote_c_style(other, msg, NULL, 0);
2961 strbuf_addf(msg, "%s\n", reset);
2962 break;
2963 case DIFF_STATUS_MODIFIED:
2964 if (p->score) {
2965 strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
2966 line_prefix,
2967 set, similarity_index(p), reset);
2968 break;
2969 }
2970 /* fallthru */
2971 default:
2972 *must_show_header = 0;
2973 }
2974 if (one && two && hashcmp(one->sha1, two->sha1)) {
2975 int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
2976
2977 if (DIFF_OPT_TST(o, BINARY)) {
2978 mmfile_t mf;
2979 if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
2980 (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
2981 abbrev = 40;
2982 }
2983 strbuf_addf(msg, "%s%sindex %s..", line_prefix, set,
2984 find_unique_abbrev(one->sha1, abbrev));
2985 strbuf_addstr(msg, find_unique_abbrev(two->sha1, abbrev));
2986 if (one->mode == two->mode)
2987 strbuf_addf(msg, " %06o", one->mode);
2988 strbuf_addf(msg, "%s\n", reset);
2989 }
2990 }
2991
2992 static void run_diff_cmd(const char *pgm,
2993 const char *name,
2994 const char *other,
2995 const char *attr_path,
2996 struct diff_filespec *one,
2997 struct diff_filespec *two,
2998 struct strbuf *msg,
2999 struct diff_options *o,
3000 struct diff_filepair *p)
3001 {
3002 const char *xfrm_msg = NULL;
3003 int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
3004 int must_show_header = 0;
3005
3006
3007 if (DIFF_OPT_TST(o, ALLOW_EXTERNAL)) {
3008 struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
3009 if (drv && drv->external)
3010 pgm = drv->external;
3011 }
3012
3013 if (msg) {
3014 /*
3015 * don't use colors when the header is intended for an
3016 * external diff driver
3017 */
3018 fill_metainfo(msg, name, other, one, two, o, p,
3019 &must_show_header,
3020 want_color(o->use_color) && !pgm);
3021 xfrm_msg = msg->len ? msg->buf : NULL;
3022 }
3023
3024 if (pgm) {
3025 run_external_diff(pgm, name, other, one, two, xfrm_msg,
3026 complete_rewrite);
3027 return;
3028 }
3029 if (one && two)
3030 builtin_diff(name, other ? other : name,
3031 one, two, xfrm_msg, must_show_header,
3032 o, complete_rewrite);
3033 else
3034 fprintf(o->file, "* Unmerged path %s\n", name);
3035 }
3036
3037 static void diff_fill_sha1_info(struct diff_filespec *one)
3038 {
3039 if (DIFF_FILE_VALID(one)) {
3040 if (!one->sha1_valid) {
3041 struct stat st;
3042 if (one->is_stdin) {
3043 hashcpy(one->sha1, null_sha1);
3044 return;
3045 }
3046 if (lstat(one->path, &st) < 0)
3047 die_errno("stat '%s'", one->path);
3048 if (index_path(one->sha1, one->path, &st, 0))
3049 die("cannot hash %s", one->path);
3050 }
3051 }
3052 else
3053 hashclr(one->sha1);
3054 }
3055
3056 static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
3057 {
3058 /* Strip the prefix but do not molest /dev/null and absolute paths */
3059 if (*namep && **namep != '/') {
3060 *namep += prefix_length;
3061 if (**namep == '/')
3062 ++*namep;
3063 }
3064 if (*otherp && **otherp != '/') {
3065 *otherp += prefix_length;
3066 if (**otherp == '/')
3067 ++*otherp;
3068 }
3069 }
3070
3071 static void run_diff(struct diff_filepair *p, struct diff_options *o)
3072 {
3073 const char *pgm = external_diff();
3074 struct strbuf msg;
3075 struct diff_filespec *one = p->one;
3076 struct diff_filespec *two = p->two;
3077 const char *name;
3078 const char *other;
3079 const char *attr_path;
3080
3081 name = p->one->path;
3082 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3083 attr_path = name;
3084 if (o->prefix_length)
3085 strip_prefix(o->prefix_length, &name, &other);
3086
3087 if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
3088 pgm = NULL;
3089
3090 if (DIFF_PAIR_UNMERGED(p)) {
3091 run_diff_cmd(pgm, name, NULL, attr_path,
3092 NULL, NULL, NULL, o, p);
3093 return;
3094 }
3095
3096 diff_fill_sha1_info(one);
3097 diff_fill_sha1_info(two);
3098
3099 if (!pgm &&
3100 DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3101 (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3102 /*
3103 * a filepair that changes between file and symlink
3104 * needs to be split into deletion and creation.
3105 */
3106 struct diff_filespec *null = alloc_filespec(two->path);
3107 run_diff_cmd(NULL, name, other, attr_path,
3108 one, null, &msg, o, p);
3109 free(null);
3110 strbuf_release(&msg);
3111
3112 null = alloc_filespec(one->path);
3113 run_diff_cmd(NULL, name, other, attr_path,
3114 null, two, &msg, o, p);
3115 free(null);
3116 }
3117 else
3118 run_diff_cmd(pgm, name, other, attr_path,
3119 one, two, &msg, o, p);
3120
3121 strbuf_release(&msg);
3122 }
3123
3124 static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3125 struct diffstat_t *diffstat)
3126 {
3127 const char *name;
3128 const char *other;
3129
3130 if (DIFF_PAIR_UNMERGED(p)) {
3131 /* unmerged */
3132 builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, p);
3133 return;
3134 }
3135
3136 name = p->one->path;
3137 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3138
3139 if (o->prefix_length)
3140 strip_prefix(o->prefix_length, &name, &other);
3141
3142 diff_fill_sha1_info(p->one);
3143 diff_fill_sha1_info(p->two);
3144
3145 builtin_diffstat(name, other, p->one, p->two, diffstat, o, p);
3146 }
3147
3148 static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3149 {
3150 const char *name;
3151 const char *other;
3152 const char *attr_path;
3153
3154 if (DIFF_PAIR_UNMERGED(p)) {
3155 /* unmerged */
3156 return;
3157 }
3158
3159 name = p->one->path;
3160 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3161 attr_path = other ? other : name;
3162
3163 if (o->prefix_length)
3164 strip_prefix(o->prefix_length, &name, &other);
3165
3166 diff_fill_sha1_info(p->one);
3167 diff_fill_sha1_info(p->two);
3168
3169 builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3170 }
3171
3172 void diff_setup(struct diff_options *options)
3173 {
3174 memcpy(options, &default_diff_options, sizeof(*options));
3175
3176 options->file = stdout;
3177
3178 options->line_termination = '\n';
3179 options->break_opt = -1;
3180 options->rename_limit = -1;
3181 options->dirstat_permille = diff_dirstat_permille_default;
3182 options->context = diff_context_default;
3183 DIFF_OPT_SET(options, RENAME_EMPTY);
3184
3185 options->change = diff_change;
3186 options->add_remove = diff_addremove;
3187 options->use_color = diff_use_color_default;
3188 options->detect_rename = diff_detect_rename_default;
3189 options->xdl_opts |= diff_algorithm;
3190
3191 if (diff_no_prefix) {
3192 options->a_prefix = options->b_prefix = "";
3193 } else if (!diff_mnemonic_prefix) {
3194 options->a_prefix = "a/";
3195 options->b_prefix = "b/";
3196 }
3197 }
3198
3199 void diff_setup_done(struct diff_options *options)
3200 {
3201 int count = 0;
3202
3203 if (options->output_format & DIFF_FORMAT_NAME)
3204 count++;
3205 if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3206 count++;
3207 if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3208 count++;
3209 if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3210 count++;
3211 if (count > 1)
3212 die("--name-only, --name-status, --check and -s are mutually exclusive");
3213
3214 /*
3215 * Most of the time we can say "there are changes"
3216 * only by checking if there are changed paths, but
3217 * --ignore-whitespace* options force us to look
3218 * inside contents.
3219 */
3220
3221 if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3222 DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3223 DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3224 DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3225 else
3226 DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3227
3228 if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3229 options->detect_rename = DIFF_DETECT_COPY;
3230
3231 if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3232 options->prefix = NULL;
3233 if (options->prefix)
3234 options->prefix_length = strlen(options->prefix);
3235 else
3236 options->prefix_length = 0;
3237
3238 if (options->output_format & (DIFF_FORMAT_NAME |
3239 DIFF_FORMAT_NAME_STATUS |
3240 DIFF_FORMAT_CHECKDIFF |
3241 DIFF_FORMAT_NO_OUTPUT))
3242 options->output_format &= ~(DIFF_FORMAT_RAW |
3243 DIFF_FORMAT_NUMSTAT |
3244 DIFF_FORMAT_DIFFSTAT |
3245 DIFF_FORMAT_SHORTSTAT |
3246 DIFF_FORMAT_DIRSTAT |
3247 DIFF_FORMAT_SUMMARY |
3248 DIFF_FORMAT_PATCH);
3249
3250 /*
3251 * These cases always need recursive; we do not drop caller-supplied
3252 * recursive bits for other formats here.
3253 */
3254 if (options->output_format & (DIFF_FORMAT_PATCH |
3255 DIFF_FORMAT_NUMSTAT |
3256 DIFF_FORMAT_DIFFSTAT |
3257 DIFF_FORMAT_SHORTSTAT |
3258 DIFF_FORMAT_DIRSTAT |
3259 DIFF_FORMAT_SUMMARY |
3260 DIFF_FORMAT_CHECKDIFF))
3261 DIFF_OPT_SET(options, RECURSIVE);
3262 /*
3263 * Also pickaxe would not work very well if you do not say recursive
3264 */
3265 if (options->pickaxe)
3266 DIFF_OPT_SET(options, RECURSIVE);
3267 /*
3268 * When patches are generated, submodules diffed against the work tree
3269 * must be checked for dirtiness too so it can be shown in the output
3270 */
3271 if (options->output_format & DIFF_FORMAT_PATCH)
3272 DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3273
3274 if (options->detect_rename && options->rename_limit < 0)
3275 options->rename_limit = diff_rename_limit_default;
3276 if (options->setup & DIFF_SETUP_USE_CACHE) {
3277 if (!active_cache)
3278 /* read-cache does not die even when it fails
3279 * so it is safe for us to do this here. Also
3280 * it does not smudge active_cache or active_nr
3281 * when it fails, so we do not have to worry about
3282 * cleaning it up ourselves either.
3283 */
3284 read_cache();
3285 }
3286 if (options->abbrev <= 0 || 40 < options->abbrev)
3287 options->abbrev = 40; /* full */
3288
3289 /*
3290 * It does not make sense to show the first hit we happened
3291 * to have found. It does not make sense not to return with
3292 * exit code in such a case either.
3293 */
3294 if (DIFF_OPT_TST(options, QUICK)) {
3295 options->output_format = DIFF_FORMAT_NO_OUTPUT;
3296 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3297 }
3298 }
3299
3300 static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3301 {
3302 char c, *eq;
3303 int len;
3304
3305 if (*arg != '-')
3306 return 0;
3307 c = *++arg;
3308 if (!c)
3309 return 0;
3310 if (c == arg_short) {
3311 c = *++arg;
3312 if (!c)
3313 return 1;
3314 if (val && isdigit(c)) {
3315 char *end;
3316 int n = strtoul(arg, &end, 10);
3317 if (*end)
3318 return 0;
3319 *val = n;
3320 return 1;
3321 }
3322 return 0;
3323 }
3324 if (c != '-')
3325 return 0;
3326 arg++;
3327 eq = strchr(arg, '=');
3328 if (eq)
3329 len = eq - arg;
3330 else
3331 len = strlen(arg);
3332 if (!len || strncmp(arg, arg_long, len))
3333 return 0;
3334 if (eq) {
3335 int n;
3336 char *end;
3337 if (!isdigit(*++eq))
3338 return 0;
3339 n = strtoul(eq, &end, 10);
3340 if (*end)
3341 return 0;
3342 *val = n;
3343 }
3344 return 1;
3345 }
3346
3347 static int diff_scoreopt_parse(const char *opt);
3348
3349 static inline int short_opt(char opt, const char **argv,
3350 const char **optarg)
3351 {
3352 const char *arg = argv[0];
3353 if (arg[0] != '-' || arg[1] != opt)
3354 return 0;
3355 if (arg[2] != '\0') {
3356 *optarg = arg + 2;
3357 return 1;
3358 }
3359 if (!argv[1])
3360 die("Option '%c' requires a value", opt);
3361 *optarg = argv[1];
3362 return 2;
3363 }
3364
3365 int parse_long_opt(const char *opt, const char **argv,
3366 const char **optarg)
3367 {
3368 const char *arg = argv[0];
3369 if (arg[0] != '-' || arg[1] != '-')
3370 return 0;
3371 arg += strlen("--");
3372 if (prefixcmp(arg, opt))
3373 return 0;
3374 arg += strlen(opt);
3375 if (*arg == '=') { /* sticked form: --option=value */
3376 *optarg = arg + 1;
3377 return 1;
3378 }
3379 if (*arg != '\0')
3380 return 0;
3381 /* separate form: --option value */
3382 if (!argv[1])
3383 die("Option '--%s' requires a value", opt);
3384 *optarg = argv[1];
3385 return 2;
3386 }
3387
3388 static int stat_opt(struct diff_options *options, const char **av)
3389 {
3390 const char *arg = av[0];
3391 char *end;
3392 int width = options->stat_width;
3393 int name_width = options->stat_name_width;
3394 int graph_width = options->stat_graph_width;
3395 int count = options->stat_count;
3396 int argcount = 1;
3397
3398 arg += strlen("--stat");
3399 end = (char *)arg;
3400
3401 switch (*arg) {
3402 case '-':
3403 if (!prefixcmp(arg, "-width")) {
3404 arg += strlen("-width");
3405 if (*arg == '=')
3406 width = strtoul(arg + 1, &end, 10);
3407 else if (!*arg && !av[1])
3408 die("Option '--stat-width' requires a value");
3409 else if (!*arg) {
3410 width = strtoul(av[1], &end, 10);
3411 argcount = 2;
3412 }
3413 } else if (!prefixcmp(arg, "-name-width")) {
3414 arg += strlen("-name-width");
3415 if (*arg == '=')
3416 name_width = strtoul(arg + 1, &end, 10);
3417 else if (!*arg && !av[1])
3418 die("Option '--stat-name-width' requires a value");
3419 else if (!*arg) {
3420 name_width = strtoul(av[1], &end, 10);
3421 argcount = 2;
3422 }
3423 } else if (!prefixcmp(arg, "-graph-width")) {
3424 arg += strlen("-graph-width");
3425 if (*arg == '=')
3426 graph_width = strtoul(arg + 1, &end, 10);
3427 else if (!*arg && !av[1])
3428 die("Option '--stat-graph-width' requires a value");
3429 else if (!*arg) {
3430 graph_width = strtoul(av[1], &end, 10);
3431 argcount = 2;
3432 }
3433 } else if (!prefixcmp(arg, "-count")) {
3434 arg += strlen("-count");
3435 if (*arg == '=')
3436 count = strtoul(arg + 1, &end, 10);
3437 else if (!*arg && !av[1])
3438 die("Option '--stat-count' requires a value");
3439 else if (!*arg) {
3440 count = strtoul(av[1], &end, 10);
3441 argcount = 2;
3442 }
3443 }
3444 break;
3445 case '=':
3446 width = strtoul(arg+1, &end, 10);
3447 if (*end == ',')
3448 name_width = strtoul(end+1, &end, 10);
3449 if (*end == ',')
3450 count = strtoul(end+1, &end, 10);
3451 }
3452
3453 /* Important! This checks all the error cases! */
3454 if (*end)
3455 return 0;
3456 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3457 options->stat_name_width = name_width;
3458 options->stat_graph_width = graph_width;
3459 options->stat_width = width;
3460 options->stat_count = count;
3461 return argcount;
3462 }
3463
3464 static int parse_dirstat_opt(struct diff_options *options, const char *params)
3465 {
3466 struct strbuf errmsg = STRBUF_INIT;
3467 if (parse_dirstat_params(options, params, &errmsg))
3468 die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3469 errmsg.buf);
3470 strbuf_release(&errmsg);
3471 /*
3472 * The caller knows a dirstat-related option is given from the command
3473 * line; allow it to say "return this_function();"
3474 */
3475 options->output_format |= DIFF_FORMAT_DIRSTAT;
3476 return 1;
3477 }
3478
3479 static int parse_submodule_opt(struct diff_options *options, const char *value)
3480 {
3481 if (parse_submodule_params(options, value))
3482 die(_("Failed to parse --submodule option parameter: '%s'"),
3483 value);
3484 return 1;
3485 }
3486
3487 int diff_opt_parse(struct diff_options *options, const char **av, int ac)
3488 {
3489 const char *arg = av[0];
3490 const char *optarg;
3491 int argcount;
3492
3493 /* Output format options */
3494 if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch"))
3495 options->output_format |= DIFF_FORMAT_PATCH;
3496 else if (opt_arg(arg, 'U', "unified", &options->context))
3497 options->output_format |= DIFF_FORMAT_PATCH;
3498 else if (!strcmp(arg, "--raw"))
3499 options->output_format |= DIFF_FORMAT_RAW;
3500 else if (!strcmp(arg, "--patch-with-raw"))
3501 options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_RAW;
3502 else if (!strcmp(arg, "--numstat"))
3503 options->output_format |= DIFF_FORMAT_NUMSTAT;
3504 else if (!strcmp(arg, "--shortstat"))
3505 options->output_format |= DIFF_FORMAT_SHORTSTAT;
3506 else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3507 return parse_dirstat_opt(options, "");
3508 else if (!prefixcmp(arg, "-X"))
3509 return parse_dirstat_opt(options, arg + 2);
3510 else if (!prefixcmp(arg, "--dirstat="))
3511 return parse_dirstat_opt(options, arg + 10);
3512 else if (!strcmp(arg, "--cumulative"))
3513 return parse_dirstat_opt(options, "cumulative");
3514 else if (!strcmp(arg, "--dirstat-by-file"))
3515 return parse_dirstat_opt(options, "files");
3516 else if (!prefixcmp(arg, "--dirstat-by-file=")) {
3517 parse_dirstat_opt(options, "files");
3518 return parse_dirstat_opt(options, arg + 18);
3519 }
3520 else if (!strcmp(arg, "--check"))
3521 options->output_format |= DIFF_FORMAT_CHECKDIFF;
3522 else if (!strcmp(arg, "--summary"))
3523 options->output_format |= DIFF_FORMAT_SUMMARY;
3524 else if (!strcmp(arg, "--patch-with-stat"))
3525 options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_DIFFSTAT;
3526 else if (!strcmp(arg, "--name-only"))
3527 options->output_format |= DIFF_FORMAT_NAME;
3528 else if (!strcmp(arg, "--name-status"))
3529 options->output_format |= DIFF_FORMAT_NAME_STATUS;
3530 else if (!strcmp(arg, "-s"))
3531 options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3532 else if (!prefixcmp(arg, "--stat"))
3533 /* --stat, --stat-width, --stat-name-width, or --stat-count */
3534 return stat_opt(options, av);
3535
3536 /* renames options */
3537 else if (!prefixcmp(arg, "-B") || !prefixcmp(arg, "--break-rewrites=") ||
3538 !strcmp(arg, "--break-rewrites")) {
3539 if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3540 return error("invalid argument to -B: %s", arg+2);
3541 }
3542 else if (!prefixcmp(arg, "-M") || !prefixcmp(arg, "--find-renames=") ||
3543 !strcmp(arg, "--find-renames")) {
3544 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3545 return error("invalid argument to -M: %s", arg+2);
3546 options->detect_rename = DIFF_DETECT_RENAME;
3547 }
3548 else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3549 options->irreversible_delete = 1;
3550 }
3551 else if (!prefixcmp(arg, "-C") || !prefixcmp(arg, "--find-copies=") ||
3552 !strcmp(arg, "--find-copies")) {
3553 if (options->detect_rename == DIFF_DETECT_COPY)
3554 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3555 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3556 return error("invalid argument to -C: %s", arg+2);
3557 options->detect_rename = DIFF_DETECT_COPY;
3558 }
3559 else if (!strcmp(arg, "--no-renames"))
3560 options->detect_rename = 0;
3561 else if (!strcmp(arg, "--rename-empty"))
3562 DIFF_OPT_SET(options, RENAME_EMPTY);
3563 else if (!strcmp(arg, "--no-rename-empty"))
3564 DIFF_OPT_CLR(options, RENAME_EMPTY);
3565 else if (!strcmp(arg, "--relative"))
3566 DIFF_OPT_SET(options, RELATIVE_NAME);
3567 else if (!prefixcmp(arg, "--relative=")) {
3568 DIFF_OPT_SET(options, RELATIVE_NAME);
3569 options->prefix = arg + 11;
3570 }
3571
3572 /* xdiff options */
3573 else if (!strcmp(arg, "--minimal"))
3574 DIFF_XDL_SET(options, NEED_MINIMAL);
3575 else if (!strcmp(arg, "--no-minimal"))
3576 DIFF_XDL_CLR(options, NEED_MINIMAL);
3577 else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3578 DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3579 else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3580 DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3581 else if (!strcmp(arg, "--ignore-space-at-eol"))
3582 DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3583 else if (!strcmp(arg, "--patience"))
3584 options->xdl_opts = DIFF_WITH_ALG(options, PATIENCE_DIFF);
3585 else if (!strcmp(arg, "--histogram"))
3586 options->xdl_opts = DIFF_WITH_ALG(options, HISTOGRAM_DIFF);
3587 else if (!prefixcmp(arg, "--diff-algorithm=")) {
3588 long value = parse_algorithm_value(arg+17);
3589 if (value < 0)
3590 return error("option diff-algorithm accepts \"myers\", "
3591 "\"minimal\", \"patience\" and \"histogram\"");
3592 /* clear out previous settings */
3593 DIFF_XDL_CLR(options, NEED_MINIMAL);
3594 options->xdl_opts &= ~XDF_DIFF_ALGORITHM_MASK;
3595 options->xdl_opts |= value;
3596 }
3597
3598 /* flags options */
3599 else if (!strcmp(arg, "--binary")) {
3600 options->output_format |= DIFF_FORMAT_PATCH;
3601 DIFF_OPT_SET(options, BINARY);
3602 }
3603 else if (!strcmp(arg, "--full-index"))
3604 DIFF_OPT_SET(options, FULL_INDEX);
3605 else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
3606 DIFF_OPT_SET(options, TEXT);
3607 else if (!strcmp(arg, "-R"))
3608 DIFF_OPT_SET(options, REVERSE_DIFF);
3609 else if (!strcmp(arg, "--find-copies-harder"))
3610 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3611 else if (!strcmp(arg, "--follow"))
3612 DIFF_OPT_SET(options, FOLLOW_RENAMES);
3613 else if (!strcmp(arg, "--no-follow"))
3614 DIFF_OPT_CLR(options, FOLLOW_RENAMES);
3615 else if (!strcmp(arg, "--color"))
3616 options->use_color = 1;
3617 else if (!prefixcmp(arg, "--color=")) {
3618 int value = git_config_colorbool(NULL, arg+8);
3619 if (value < 0)
3620 return error("option `color' expects \"always\", \"auto\", or \"never\"");
3621 options->use_color = value;
3622 }
3623 else if (!strcmp(arg, "--no-color"))
3624 options->use_color = 0;
3625 else if (!strcmp(arg, "--color-words")) {
3626 options->use_color = 1;
3627 options->word_diff = DIFF_WORDS_COLOR;
3628 }
3629 else if (!prefixcmp(arg, "--color-words=")) {
3630 options->use_color = 1;
3631 options->word_diff = DIFF_WORDS_COLOR;
3632 options->word_regex = arg + 14;
3633 }
3634 else if (!strcmp(arg, "--word-diff")) {
3635 if (options->word_diff == DIFF_WORDS_NONE)
3636 options->word_diff = DIFF_WORDS_PLAIN;
3637 }
3638 else if (!prefixcmp(arg, "--word-diff=")) {
3639 const char *type = arg + 12;
3640 if (!strcmp(type, "plain"))
3641 options->word_diff = DIFF_WORDS_PLAIN;
3642 else if (!strcmp(type, "color")) {
3643 options->use_color = 1;
3644 options->word_diff = DIFF_WORDS_COLOR;
3645 }
3646 else if (!strcmp(type, "porcelain"))
3647 options->word_diff = DIFF_WORDS_PORCELAIN;
3648 else if (!strcmp(type, "none"))
3649 options->word_diff = DIFF_WORDS_NONE;
3650 else
3651 die("bad --word-diff argument: %s", type);
3652 }
3653 else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
3654 if (options->word_diff == DIFF_WORDS_NONE)
3655 options->word_diff = DIFF_WORDS_PLAIN;
3656 options->word_regex = optarg;
3657 return argcount;
3658 }
3659 else if (!strcmp(arg, "--exit-code"))
3660 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3661 else if (!strcmp(arg, "--quiet"))
3662 DIFF_OPT_SET(options, QUICK);
3663 else if (!strcmp(arg, "--ext-diff"))
3664 DIFF_OPT_SET(options, ALLOW_EXTERNAL);
3665 else if (!strcmp(arg, "--no-ext-diff"))
3666 DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
3667 else if (!strcmp(arg, "--textconv"))
3668 DIFF_OPT_SET(options, ALLOW_TEXTCONV);
3669 else if (!strcmp(arg, "--no-textconv"))
3670 DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
3671 else if (!strcmp(arg, "--ignore-submodules")) {
3672 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3673 handle_ignore_submodules_arg(options, "all");
3674 } else if (!prefixcmp(arg, "--ignore-submodules=")) {
3675 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3676 handle_ignore_submodules_arg(options, arg + 20);
3677 } else if (!strcmp(arg, "--submodule"))
3678 DIFF_OPT_SET(options, SUBMODULE_LOG);
3679 else if (!prefixcmp(arg, "--submodule="))
3680 return parse_submodule_opt(options, arg + 12);
3681
3682 /* misc options */
3683 else if (!strcmp(arg, "-z"))
3684 options->line_termination = 0;
3685 else if ((argcount = short_opt('l', av, &optarg))) {
3686 options->rename_limit = strtoul(optarg, NULL, 10);
3687 return argcount;
3688 }
3689 else if ((argcount = short_opt('S', av, &optarg))) {
3690 options->pickaxe = optarg;
3691 options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
3692 return argcount;
3693 } else if ((argcount = short_opt('G', av, &optarg))) {
3694 options->pickaxe = optarg;
3695 options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
3696 return argcount;
3697 }
3698 else if (!strcmp(arg, "--pickaxe-all"))
3699 options->pickaxe_opts |= DIFF_PICKAXE_ALL;
3700 else if (!strcmp(arg, "--pickaxe-regex"))
3701 options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
3702 else if ((argcount = short_opt('O', av, &optarg))) {
3703 options->orderfile = optarg;
3704 return argcount;
3705 }
3706 else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
3707 options->filter = optarg;
3708 return argcount;
3709 }
3710 else if (!strcmp(arg, "--abbrev"))
3711 options->abbrev = DEFAULT_ABBREV;
3712 else if (!prefixcmp(arg, "--abbrev=")) {
3713 options->abbrev = strtoul(arg + 9, NULL, 10);
3714 if (options->abbrev < MINIMUM_ABBREV)
3715 options->abbrev = MINIMUM_ABBREV;
3716 else if (40 < options->abbrev)
3717 options->abbrev = 40;
3718 }
3719 else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
3720 options->a_prefix = optarg;
3721 return argcount;
3722 }
3723 else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
3724 options->b_prefix = optarg;
3725 return argcount;
3726 }
3727 else if (!strcmp(arg, "--no-prefix"))
3728 options->a_prefix = options->b_prefix = "";
3729 else if (opt_arg(arg, '\0', "inter-hunk-context",
3730 &options->interhunkcontext))
3731 ;
3732 else if (!strcmp(arg, "-W"))
3733 DIFF_OPT_SET(options, FUNCCONTEXT);
3734 else if (!strcmp(arg, "--function-context"))
3735 DIFF_OPT_SET(options, FUNCCONTEXT);
3736 else if (!strcmp(arg, "--no-function-context"))
3737 DIFF_OPT_CLR(options, FUNCCONTEXT);
3738 else if ((argcount = parse_long_opt("output", av, &optarg))) {
3739 options->file = fopen(optarg, "w");
3740 if (!options->file)
3741 die_errno("Could not open '%s'", optarg);
3742 options->close_file = 1;
3743 return argcount;
3744 } else
3745 return 0;
3746 return 1;
3747 }
3748
3749 int parse_rename_score(const char **cp_p)
3750 {
3751 unsigned long num, scale;
3752 int ch, dot;
3753 const char *cp = *cp_p;
3754
3755 num = 0;
3756 scale = 1;
3757 dot = 0;
3758 for (;;) {
3759 ch = *cp;
3760 if ( !dot && ch == '.' ) {
3761 scale = 1;
3762 dot = 1;
3763 } else if ( ch == '%' ) {
3764 scale = dot ? scale*100 : 100;
3765 cp++; /* % is always at the end */
3766 break;
3767 } else if ( ch >= '0' && ch <= '9' ) {
3768 if ( scale < 100000 ) {
3769 scale *= 10;
3770 num = (num*10) + (ch-'0');
3771 }
3772 } else {
3773 break;
3774 }
3775 cp++;
3776 }
3777 *cp_p = cp;
3778
3779 /* user says num divided by scale and we say internally that
3780 * is MAX_SCORE * num / scale.
3781 */
3782 return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
3783 }
3784
3785 static int diff_scoreopt_parse(const char *opt)
3786 {
3787 int opt1, opt2, cmd;
3788
3789 if (*opt++ != '-')
3790 return -1;
3791 cmd = *opt++;
3792 if (cmd == '-') {
3793 /* convert the long-form arguments into short-form versions */
3794 if (!prefixcmp(opt, "break-rewrites")) {
3795 opt += strlen("break-rewrites");
3796 if (*opt == 0 || *opt++ == '=')
3797 cmd = 'B';
3798 } else if (!prefixcmp(opt, "find-copies")) {
3799 opt += strlen("find-copies");
3800 if (*opt == 0 || *opt++ == '=')
3801 cmd = 'C';
3802 } else if (!prefixcmp(opt, "find-renames")) {
3803 opt += strlen("find-renames");
3804 if (*opt == 0 || *opt++ == '=')
3805 cmd = 'M';
3806 }
3807 }
3808 if (cmd != 'M' && cmd != 'C' && cmd != 'B')
3809 return -1; /* that is not a -M, -C nor -B option */
3810
3811 opt1 = parse_rename_score(&opt);
3812 if (cmd != 'B')
3813 opt2 = 0;
3814 else {
3815 if (*opt == 0)
3816 opt2 = 0;
3817 else if (*opt != '/')
3818 return -1; /* we expect -B80/99 or -B80 */
3819 else {
3820 opt++;
3821 opt2 = parse_rename_score(&opt);
3822 }
3823 }
3824 if (*opt != 0)
3825 return -1;
3826 return opt1 | (opt2 << 16);
3827 }
3828
3829 struct diff_queue_struct diff_queued_diff;
3830
3831 void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
3832 {
3833 if (queue->alloc <= queue->nr) {
3834 queue->alloc = alloc_nr(queue->alloc);
3835 queue->queue = xrealloc(queue->queue,
3836 sizeof(dp) * queue->alloc);
3837 }
3838 queue->queue[queue->nr++] = dp;
3839 }
3840
3841 struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
3842 struct diff_filespec *one,
3843 struct diff_filespec *two)
3844 {
3845 struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
3846 dp->one = one;
3847 dp->two = two;
3848 if (queue)
3849 diff_q(queue, dp);
3850 return dp;
3851 }
3852
3853 void diff_free_filepair(struct diff_filepair *p)
3854 {
3855 free_filespec(p->one);
3856 free_filespec(p->two);
3857 free(p);
3858 }
3859
3860 /* This is different from find_unique_abbrev() in that
3861 * it stuffs the result with dots for alignment.
3862 */
3863 const char *diff_unique_abbrev(const unsigned char *sha1, int len)
3864 {
3865 int abblen;
3866 const char *abbrev;
3867 if (len == 40)
3868 return sha1_to_hex(sha1);
3869
3870 abbrev = find_unique_abbrev(sha1, len);
3871 abblen = strlen(abbrev);
3872 if (abblen < 37) {
3873 static char hex[41];
3874 if (len < abblen && abblen <= len + 2)
3875 sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
3876 else
3877 sprintf(hex, "%s...", abbrev);
3878 return hex;
3879 }
3880 return sha1_to_hex(sha1);
3881 }
3882
3883 static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
3884 {
3885 int line_termination = opt->line_termination;
3886 int inter_name_termination = line_termination ? '\t' : '\0';
3887
3888 fprintf(opt->file, "%s", diff_line_prefix(opt));
3889 if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
3890 fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
3891 diff_unique_abbrev(p->one->sha1, opt->abbrev));
3892 fprintf(opt->file, "%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
3893 }
3894 if (p->score) {
3895 fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
3896 inter_name_termination);
3897 } else {
3898 fprintf(opt->file, "%c%c", p->status, inter_name_termination);
3899 }
3900
3901 if (p->status == DIFF_STATUS_COPIED ||
3902 p->status == DIFF_STATUS_RENAMED) {
3903 const char *name_a, *name_b;
3904 name_a = p->one->path;
3905 name_b = p->two->path;
3906 strip_prefix(opt->prefix_length, &name_a, &name_b);
3907 write_name_quoted(name_a, opt->file, inter_name_termination);
3908 write_name_quoted(name_b, opt->file, line_termination);
3909 } else {
3910 const char *name_a, *name_b;
3911 name_a = p->one->mode ? p->one->path : p->two->path;
3912 name_b = NULL;
3913 strip_prefix(opt->prefix_length, &name_a, &name_b);
3914 write_name_quoted(name_a, opt->file, line_termination);
3915 }
3916 }
3917
3918 int diff_unmodified_pair(struct diff_filepair *p)
3919 {
3920 /* This function is written stricter than necessary to support
3921 * the currently implemented transformers, but the idea is to
3922 * let transformers to produce diff_filepairs any way they want,
3923 * and filter and clean them up here before producing the output.
3924 */
3925 struct diff_filespec *one = p->one, *two = p->two;
3926
3927 if (DIFF_PAIR_UNMERGED(p))
3928 return 0; /* unmerged is interesting */
3929
3930 /* deletion, addition, mode or type change
3931 * and rename are all interesting.
3932 */
3933 if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
3934 DIFF_PAIR_MODE_CHANGED(p) ||
3935 strcmp(one->path, two->path))
3936 return 0;
3937
3938 /* both are valid and point at the same path. that is, we are
3939 * dealing with a change.
3940 */
3941 if (one->sha1_valid && two->sha1_valid &&
3942 !hashcmp(one->sha1, two->sha1) &&
3943 !one->dirty_submodule && !two->dirty_submodule)
3944 return 1; /* no change */
3945 if (!one->sha1_valid && !two->sha1_valid)
3946 return 1; /* both look at the same file on the filesystem. */
3947 return 0;
3948 }
3949
3950 static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
3951 {
3952 if (diff_unmodified_pair(p))
3953 return;
3954
3955 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3956 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3957 return; /* no tree diffs in patch format */
3958
3959 run_diff(p, o);
3960 }
3961
3962 static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
3963 struct diffstat_t *diffstat)
3964 {
3965 if (diff_unmodified_pair(p))
3966 return;
3967
3968 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3969 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3970 return; /* no useful stat for tree diffs */
3971
3972 run_diffstat(p, o, diffstat);
3973 }
3974
3975 static void diff_flush_checkdiff(struct diff_filepair *p,
3976 struct diff_options *o)
3977 {
3978 if (diff_unmodified_pair(p))
3979 return;
3980
3981 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3982 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3983 return; /* nothing to check in tree diffs */
3984
3985 run_checkdiff(p, o);
3986 }
3987
3988 int diff_queue_is_empty(void)
3989 {
3990 struct diff_queue_struct *q = &diff_queued_diff;
3991 int i;
3992 for (i = 0; i < q->nr; i++)
3993 if (!diff_unmodified_pair(q->queue[i]))
3994 return 0;
3995 return 1;
3996 }
3997
3998 #if DIFF_DEBUG
3999 void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
4000 {
4001 fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
4002 x, one ? one : "",
4003 s->path,
4004 DIFF_FILE_VALID(s) ? "valid" : "invalid",
4005 s->mode,
4006 s->sha1_valid ? sha1_to_hex(s->sha1) : "");
4007 fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
4008 x, one ? one : "",
4009 s->size, s->xfrm_flags);
4010 }
4011
4012 void diff_debug_filepair(const struct diff_filepair *p, int i)
4013 {
4014 diff_debug_filespec(p->one, i, "one");
4015 diff_debug_filespec(p->two, i, "two");
4016 fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
4017 p->score, p->status ? p->status : '?',
4018 p->one->rename_used, p->broken_pair);
4019 }
4020
4021 void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
4022 {
4023 int i;
4024 if (msg)
4025 fprintf(stderr, "%s\n", msg);
4026 fprintf(stderr, "q->nr = %d\n", q->nr);
4027 for (i = 0; i < q->nr; i++) {
4028 struct diff_filepair *p = q->queue[i];
4029 diff_debug_filepair(p, i);
4030 }
4031 }
4032 #endif
4033
4034 static void diff_resolve_rename_copy(void)
4035 {
4036 int i;
4037 struct diff_filepair *p;
4038 struct diff_queue_struct *q = &diff_queued_diff;
4039
4040 diff_debug_queue("resolve-rename-copy", q);
4041
4042 for (i = 0; i < q->nr; i++) {
4043 p = q->queue[i];
4044 p->status = 0; /* undecided */
4045 if (DIFF_PAIR_UNMERGED(p))
4046 p->status = DIFF_STATUS_UNMERGED;
4047 else if (!DIFF_FILE_VALID(p->one))
4048 p->status = DIFF_STATUS_ADDED;
4049 else if (!DIFF_FILE_VALID(p->two))
4050 p->status = DIFF_STATUS_DELETED;
4051 else if (DIFF_PAIR_TYPE_CHANGED(p))
4052 p->status = DIFF_STATUS_TYPE_CHANGED;
4053
4054 /* from this point on, we are dealing with a pair
4055 * whose both sides are valid and of the same type, i.e.
4056 * either in-place edit or rename/copy edit.
4057 */
4058 else if (DIFF_PAIR_RENAME(p)) {
4059 /*
4060 * A rename might have re-connected a broken
4061 * pair up, causing the pathnames to be the
4062 * same again. If so, that's not a rename at
4063 * all, just a modification..
4064 *
4065 * Otherwise, see if this source was used for
4066 * multiple renames, in which case we decrement
4067 * the count, and call it a copy.
4068 */
4069 if (!strcmp(p->one->path, p->two->path))
4070 p->status = DIFF_STATUS_MODIFIED;
4071 else if (--p->one->rename_used > 0)
4072 p->status = DIFF_STATUS_COPIED;
4073 else
4074 p->status = DIFF_STATUS_RENAMED;
4075 }
4076 else if (hashcmp(p->one->sha1, p->two->sha1) ||
4077 p->one->mode != p->two->mode ||
4078 p->one->dirty_submodule ||
4079 p->two->dirty_submodule ||
4080 is_null_sha1(p->one->sha1))
4081 p->status = DIFF_STATUS_MODIFIED;
4082 else {
4083 /* This is a "no-change" entry and should not
4084 * happen anymore, but prepare for broken callers.
4085 */
4086 error("feeding unmodified %s to diffcore",
4087 p->one->path);
4088 p->status = DIFF_STATUS_UNKNOWN;
4089 }
4090 }
4091 diff_debug_queue("resolve-rename-copy done", q);
4092 }
4093
4094 static int check_pair_status(struct diff_filepair *p)
4095 {
4096 switch (p->status) {
4097 case DIFF_STATUS_UNKNOWN:
4098 return 0;
4099 case 0:
4100 die("internal error in diff-resolve-rename-copy");
4101 default:
4102 return 1;
4103 }
4104 }
4105
4106 static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
4107 {
4108 int fmt = opt->output_format;
4109
4110 if (fmt & DIFF_FORMAT_CHECKDIFF)
4111 diff_flush_checkdiff(p, opt);
4112 else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
4113 diff_flush_raw(p, opt);
4114 else if (fmt & DIFF_FORMAT_NAME) {
4115 const char *name_a, *name_b;
4116 name_a = p->two->path;
4117 name_b = NULL;
4118 strip_prefix(opt->prefix_length, &name_a, &name_b);
4119 write_name_quoted(name_a, opt->file, opt->line_termination);
4120 }
4121 }
4122
4123 static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
4124 {
4125 if (fs->mode)
4126 fprintf(file, " %s mode %06o ", newdelete, fs->mode);
4127 else
4128 fprintf(file, " %s ", newdelete);
4129 write_name_quoted(fs->path, file, '\n');
4130 }
4131
4132
4133 static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
4134 const char *line_prefix)
4135 {
4136 if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
4137 fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
4138 p->two->mode, show_name ? ' ' : '\n');
4139 if (show_name) {
4140 write_name_quoted(p->two->path, file, '\n');
4141 }
4142 }
4143 }
4144
4145 static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
4146 const char *line_prefix)
4147 {
4148 char *names = pprint_rename(p->one->path, p->two->path);
4149
4150 fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
4151 free(names);
4152 show_mode_change(file, p, 0, line_prefix);
4153 }
4154
4155 static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
4156 {
4157 FILE *file = opt->file;
4158 const char *line_prefix = diff_line_prefix(opt);
4159
4160 switch(p->status) {
4161 case DIFF_STATUS_DELETED:
4162 fputs(line_prefix, file);
4163 show_file_mode_name(file, "delete", p->one);
4164 break;
4165 case DIFF_STATUS_ADDED:
4166 fputs(line_prefix, file);
4167 show_file_mode_name(file, "create", p->two);
4168 break;
4169 case DIFF_STATUS_COPIED:
4170 fputs(line_prefix, file);
4171 show_rename_copy(file, "copy", p, line_prefix);
4172 break;
4173 case DIFF_STATUS_RENAMED:
4174 fputs(line_prefix, file);
4175 show_rename_copy(file, "rename", p, line_prefix);
4176 break;
4177 default:
4178 if (p->score) {
4179 fprintf(file, "%s rewrite ", line_prefix);
4180 write_name_quoted(p->two->path, file, ' ');
4181 fprintf(file, "(%d%%)\n", similarity_index(p));
4182 }
4183 show_mode_change(file, p, !p->score, line_prefix);
4184 break;
4185 }
4186 }
4187
4188 struct patch_id_t {
4189 git_SHA_CTX *ctx;
4190 int patchlen;
4191 };
4192
4193 static int remove_space(char *line, int len)
4194 {
4195 int i;
4196 char *dst = line;
4197 unsigned char c;
4198
4199 for (i = 0; i < len; i++)
4200 if (!isspace((c = line[i])))
4201 *dst++ = c;
4202
4203 return dst - line;
4204 }
4205
4206 static void patch_id_consume(void *priv, char *line, unsigned long len)
4207 {
4208 struct patch_id_t *data = priv;
4209 int new_len;
4210
4211 /* Ignore line numbers when computing the SHA1 of the patch */
4212 if (!prefixcmp(line, "@@ -"))
4213 return;
4214
4215 new_len = remove_space(line, len);
4216
4217 git_SHA1_Update(data->ctx, line, new_len);
4218 data->patchlen += new_len;
4219 }
4220
4221 /* returns 0 upon success, and writes result into sha1 */
4222 static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
4223 {
4224 struct diff_queue_struct *q = &diff_queued_diff;
4225 int i;
4226 git_SHA_CTX ctx;
4227 struct patch_id_t data;
4228 char buffer[PATH_MAX * 4 + 20];
4229
4230 git_SHA1_Init(&ctx);
4231 memset(&data, 0, sizeof(struct patch_id_t));
4232 data.ctx = &ctx;
4233
4234 for (i = 0; i < q->nr; i++) {
4235 xpparam_t xpp;
4236 xdemitconf_t xecfg;
4237 mmfile_t mf1, mf2;
4238 struct diff_filepair *p = q->queue[i];
4239 int len1, len2;
4240
4241 memset(&xpp, 0, sizeof(xpp));
4242 memset(&xecfg, 0, sizeof(xecfg));
4243 if (p->status == 0)
4244 return error("internal diff status error");
4245 if (p->status == DIFF_STATUS_UNKNOWN)
4246 continue;
4247 if (diff_unmodified_pair(p))
4248 continue;
4249 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4250 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4251 continue;
4252 if (DIFF_PAIR_UNMERGED(p))
4253 continue;
4254
4255 diff_fill_sha1_info(p->one);
4256 diff_fill_sha1_info(p->two);
4257 if (fill_mmfile(&mf1, p->one) < 0 ||
4258 fill_mmfile(&mf2, p->two) < 0)
4259 return error("unable to read files to diff");
4260
4261 len1 = remove_space(p->one->path, strlen(p->one->path));
4262 len2 = remove_space(p->two->path, strlen(p->two->path));
4263 if (p->one->mode == 0)
4264 len1 = snprintf(buffer, sizeof(buffer),
4265 "diff--gita/%.*sb/%.*s"
4266 "newfilemode%06o"
4267 "---/dev/null"
4268 "+++b/%.*s",
4269 len1, p->one->path,
4270 len2, p->two->path,
4271 p->two->mode,
4272 len2, p->two->path);
4273 else if (p->two->mode == 0)
4274 len1 = snprintf(buffer, sizeof(buffer),
4275 "diff--gita/%.*sb/%.*s"
4276 "deletedfilemode%06o"
4277 "---a/%.*s"
4278 "+++/dev/null",
4279 len1, p->one->path,
4280 len2, p->two->path,
4281 p->one->mode,
4282 len1, p->one->path);
4283 else
4284 len1 = snprintf(buffer, sizeof(buffer),
4285 "diff--gita/%.*sb/%.*s"
4286 "---a/%.*s"
4287 "+++b/%.*s",
4288 len1, p->one->path,
4289 len2, p->two->path,
4290 len1, p->one->path,
4291 len2, p->two->path);
4292 git_SHA1_Update(&ctx, buffer, len1);
4293
4294 if (diff_filespec_is_binary(p->one) ||
4295 diff_filespec_is_binary(p->two)) {
4296 git_SHA1_Update(&ctx, sha1_to_hex(p->one->sha1), 40);
4297 git_SHA1_Update(&ctx, sha1_to_hex(p->two->sha1), 40);
4298 continue;
4299 }
4300
4301 xpp.flags = 0;
4302 xecfg.ctxlen = 3;
4303 xecfg.flags = 0;
4304 xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4305 &xpp, &xecfg);
4306 }
4307
4308 git_SHA1_Final(sha1, &ctx);
4309 return 0;
4310 }
4311
4312 int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
4313 {
4314 struct diff_queue_struct *q = &diff_queued_diff;
4315 int i;
4316 int result = diff_get_patch_id(options, sha1);
4317
4318 for (i = 0; i < q->nr; i++)
4319 diff_free_filepair(q->queue[i]);
4320
4321 free(q->queue);
4322 DIFF_QUEUE_CLEAR(q);
4323
4324 return result;
4325 }
4326
4327 static int is_summary_empty(const struct diff_queue_struct *q)
4328 {
4329 int i;
4330
4331 for (i = 0; i < q->nr; i++) {
4332 const struct diff_filepair *p = q->queue[i];
4333
4334 switch (p->status) {
4335 case DIFF_STATUS_DELETED:
4336 case DIFF_STATUS_ADDED:
4337 case DIFF_STATUS_COPIED:
4338 case DIFF_STATUS_RENAMED:
4339 return 0;
4340 default:
4341 if (p->score)
4342 return 0;
4343 if (p->one->mode && p->two->mode &&
4344 p->one->mode != p->two->mode)
4345 return 0;
4346 break;
4347 }
4348 }
4349 return 1;
4350 }
4351
4352 static const char rename_limit_warning[] =
4353 "inexact rename detection was skipped due to too many files.";
4354
4355 static const char degrade_cc_to_c_warning[] =
4356 "only found copies from modified paths due to too many files.";
4357
4358 static const char rename_limit_advice[] =
4359 "you may want to set your %s variable to at least "
4360 "%d and retry the command.";
4361
4362 void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4363 {
4364 if (degraded_cc)
4365 warning(degrade_cc_to_c_warning);
4366 else if (needed)
4367 warning(rename_limit_warning);
4368 else
4369 return;
4370 if (0 < needed && needed < 32767)
4371 warning(rename_limit_advice, varname, needed);
4372 }
4373
4374 void diff_flush(struct diff_options *options)
4375 {
4376 struct diff_queue_struct *q = &diff_queued_diff;
4377 int i, output_format = options->output_format;
4378 int separator = 0;
4379 int dirstat_by_line = 0;
4380
4381 /*
4382 * Order: raw, stat, summary, patch
4383 * or: name/name-status/checkdiff (other bits clear)
4384 */
4385 if (!q->nr)
4386 goto free_queue;
4387
4388 if (output_format & (DIFF_FORMAT_RAW |
4389 DIFF_FORMAT_NAME |
4390 DIFF_FORMAT_NAME_STATUS |
4391 DIFF_FORMAT_CHECKDIFF)) {
4392 for (i = 0; i < q->nr; i++) {
4393 struct diff_filepair *p = q->queue[i];
4394 if (check_pair_status(p))
4395 flush_one_pair(p, options);
4396 }
4397 separator++;
4398 }
4399
4400 if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4401 dirstat_by_line = 1;
4402
4403 if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4404 dirstat_by_line) {
4405 struct diffstat_t diffstat;
4406
4407 memset(&diffstat, 0, sizeof(struct diffstat_t));
4408 for (i = 0; i < q->nr; i++) {
4409 struct diff_filepair *p = q->queue[i];
4410 if (check_pair_status(p))
4411 diff_flush_stat(p, options, &diffstat);
4412 }
4413 if (output_format & DIFF_FORMAT_NUMSTAT)
4414 show_numstat(&diffstat, options);
4415 if (output_format & DIFF_FORMAT_DIFFSTAT)
4416 show_stats(&diffstat, options);
4417 if (output_format & DIFF_FORMAT_SHORTSTAT)
4418 show_shortstats(&diffstat, options);
4419 if (output_format & DIFF_FORMAT_DIRSTAT)
4420 show_dirstat_by_line(&diffstat, options);
4421 free_diffstat_info(&diffstat);
4422 separator++;
4423 }
4424 if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4425 show_dirstat(options);
4426
4427 if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4428 for (i = 0; i < q->nr; i++) {
4429 diff_summary(options, q->queue[i]);
4430 }
4431 separator++;
4432 }
4433
4434 if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4435 DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4436 DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4437 /*
4438 * run diff_flush_patch for the exit status. setting
4439 * options->file to /dev/null should be safe, becaue we
4440 * aren't supposed to produce any output anyway.
4441 */
4442 if (options->close_file)
4443 fclose(options->file);
4444 options->file = fopen("/dev/null", "w");
4445 if (!options->file)
4446 die_errno("Could not open /dev/null");
4447 options->close_file = 1;
4448 for (i = 0; i < q->nr; i++) {
4449 struct diff_filepair *p = q->queue[i];
4450 if (check_pair_status(p))
4451 diff_flush_patch(p, options);
4452 if (options->found_changes)
4453 break;
4454 }
4455 }
4456
4457 if (output_format & DIFF_FORMAT_PATCH) {
4458 if (separator) {
4459 fprintf(options->file, "%s%c",
4460 diff_line_prefix(options),
4461 options->line_termination);
4462 if (options->stat_sep) {
4463 /* attach patch instead of inline */
4464 fputs(options->stat_sep, options->file);
4465 }
4466 }
4467
4468 for (i = 0; i < q->nr; i++) {
4469 struct diff_filepair *p = q->queue[i];
4470 if (check_pair_status(p))
4471 diff_flush_patch(p, options);
4472 }
4473 }
4474
4475 if (output_format & DIFF_FORMAT_CALLBACK)
4476 options->format_callback(q, options, options->format_callback_data);
4477
4478 for (i = 0; i < q->nr; i++)
4479 diff_free_filepair(q->queue[i]);
4480 free_queue:
4481 free(q->queue);
4482 DIFF_QUEUE_CLEAR(q);
4483 if (options->close_file)
4484 fclose(options->file);
4485
4486 /*
4487 * Report the content-level differences with HAS_CHANGES;
4488 * diff_addremove/diff_change does not set the bit when
4489 * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4490 */
4491 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4492 if (options->found_changes)
4493 DIFF_OPT_SET(options, HAS_CHANGES);
4494 else
4495 DIFF_OPT_CLR(options, HAS_CHANGES);
4496 }
4497 }
4498
4499 static void diffcore_apply_filter(const char *filter)
4500 {
4501 int i;
4502 struct diff_queue_struct *q = &diff_queued_diff;
4503 struct diff_queue_struct outq;
4504 DIFF_QUEUE_CLEAR(&outq);
4505
4506 if (!filter)
4507 return;
4508
4509 if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
4510 int found;
4511 for (i = found = 0; !found && i < q->nr; i++) {
4512 struct diff_filepair *p = q->queue[i];
4513 if (((p->status == DIFF_STATUS_MODIFIED) &&
4514 ((p->score &&
4515 strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
4516 (!p->score &&
4517 strchr(filter, DIFF_STATUS_MODIFIED)))) ||
4518 ((p->status != DIFF_STATUS_MODIFIED) &&
4519 strchr(filter, p->status)))
4520 found++;
4521 }
4522 if (found)
4523 return;
4524
4525 /* otherwise we will clear the whole queue
4526 * by copying the empty outq at the end of this
4527 * function, but first clear the current entries
4528 * in the queue.
4529 */
4530 for (i = 0; i < q->nr; i++)
4531 diff_free_filepair(q->queue[i]);
4532 }
4533 else {
4534 /* Only the matching ones */
4535 for (i = 0; i < q->nr; i++) {
4536 struct diff_filepair *p = q->queue[i];
4537
4538 if (((p->status == DIFF_STATUS_MODIFIED) &&
4539 ((p->score &&
4540 strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
4541 (!p->score &&
4542 strchr(filter, DIFF_STATUS_MODIFIED)))) ||
4543 ((p->status != DIFF_STATUS_MODIFIED) &&
4544 strchr(filter, p->status)))
4545 diff_q(&outq, p);
4546 else
4547 diff_free_filepair(p);
4548 }
4549 }
4550 free(q->queue);
4551 *q = outq;
4552 }
4553
4554 /* Check whether two filespecs with the same mode and size are identical */
4555 static int diff_filespec_is_identical(struct diff_filespec *one,
4556 struct diff_filespec *two)
4557 {
4558 if (S_ISGITLINK(one->mode))
4559 return 0;
4560 if (diff_populate_filespec(one, 0))
4561 return 0;
4562 if (diff_populate_filespec(two, 0))
4563 return 0;
4564 return !memcmp(one->data, two->data, one->size);
4565 }
4566
4567 static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
4568 {
4569 int i;
4570 struct diff_queue_struct *q = &diff_queued_diff;
4571 struct diff_queue_struct outq;
4572 DIFF_QUEUE_CLEAR(&outq);
4573
4574 for (i = 0; i < q->nr; i++) {
4575 struct diff_filepair *p = q->queue[i];
4576
4577 /*
4578 * 1. Entries that come from stat info dirtiness
4579 * always have both sides (iow, not create/delete),
4580 * one side of the object name is unknown, with
4581 * the same mode and size. Keep the ones that
4582 * do not match these criteria. They have real
4583 * differences.
4584 *
4585 * 2. At this point, the file is known to be modified,
4586 * with the same mode and size, and the object
4587 * name of one side is unknown. Need to inspect
4588 * the identical contents.
4589 */
4590 if (!DIFF_FILE_VALID(p->one) || /* (1) */
4591 !DIFF_FILE_VALID(p->two) ||
4592 (p->one->sha1_valid && p->two->sha1_valid) ||
4593 (p->one->mode != p->two->mode) ||
4594 diff_populate_filespec(p->one, 1) ||
4595 diff_populate_filespec(p->two, 1) ||
4596 (p->one->size != p->two->size) ||
4597 !diff_filespec_is_identical(p->one, p->two)) /* (2) */
4598 diff_q(&outq, p);
4599 else {
4600 /*
4601 * The caller can subtract 1 from skip_stat_unmatch
4602 * to determine how many paths were dirty only
4603 * due to stat info mismatch.
4604 */
4605 if (!DIFF_OPT_TST(diffopt, NO_INDEX))
4606 diffopt->skip_stat_unmatch++;
4607 diff_free_filepair(p);
4608 }
4609 }
4610 free(q->queue);
4611 *q = outq;
4612 }
4613
4614 static int diffnamecmp(const void *a_, const void *b_)
4615 {
4616 const struct diff_filepair *a = *((const struct diff_filepair **)a_);
4617 const struct diff_filepair *b = *((const struct diff_filepair **)b_);
4618 const char *name_a, *name_b;
4619
4620 name_a = a->one ? a->one->path : a->two->path;
4621 name_b = b->one ? b->one->path : b->two->path;
4622 return strcmp(name_a, name_b);
4623 }
4624
4625 void diffcore_fix_diff_index(struct diff_options *options)
4626 {
4627 struct diff_queue_struct *q = &diff_queued_diff;
4628 qsort(q->queue, q->nr, sizeof(q->queue[0]), diffnamecmp);
4629 }
4630
4631 void diffcore_std(struct diff_options *options)
4632 {
4633 if (options->skip_stat_unmatch)
4634 diffcore_skip_stat_unmatch(options);
4635 if (!options->found_follow) {
4636 /* See try_to_follow_renames() in tree-diff.c */
4637 if (options->break_opt != -1)
4638 diffcore_break(options->break_opt);
4639 if (options->detect_rename)
4640 diffcore_rename(options);
4641 if (options->break_opt != -1)
4642 diffcore_merge_broken();
4643 }
4644 if (options->pickaxe)
4645 diffcore_pickaxe(options);
4646 if (options->orderfile)
4647 diffcore_order(options->orderfile);
4648 if (!options->found_follow)
4649 /* See try_to_follow_renames() in tree-diff.c */
4650 diff_resolve_rename_copy();
4651 diffcore_apply_filter(options->filter);
4652
4653 if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4654 DIFF_OPT_SET(options, HAS_CHANGES);
4655 else
4656 DIFF_OPT_CLR(options, HAS_CHANGES);
4657
4658 options->found_follow = 0;
4659 }
4660
4661 int diff_result_code(struct diff_options *opt, int status)
4662 {
4663 int result = 0;
4664
4665 diff_warn_rename_limit("diff.renamelimit",
4666 opt->needed_rename_limit,
4667 opt->degraded_cc_to_c);
4668 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4669 !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
4670 return status;
4671 if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4672 DIFF_OPT_TST(opt, HAS_CHANGES))
4673 result |= 01;
4674 if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
4675 DIFF_OPT_TST(opt, CHECK_FAILED))
4676 result |= 02;
4677 return result;
4678 }
4679
4680 int diff_can_quit_early(struct diff_options *opt)
4681 {
4682 return (DIFF_OPT_TST(opt, QUICK) &&
4683 !opt->filter &&
4684 DIFF_OPT_TST(opt, HAS_CHANGES));
4685 }
4686
4687 /*
4688 * Shall changes to this submodule be ignored?
4689 *
4690 * Submodule changes can be configured to be ignored separately for each path,
4691 * but that configuration can be overridden from the command line.
4692 */
4693 static int is_submodule_ignored(const char *path, struct diff_options *options)
4694 {
4695 int ignored = 0;
4696 unsigned orig_flags = options->flags;
4697 if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
4698 set_diffopt_flags_from_submodule_config(options, path);
4699 if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
4700 ignored = 1;
4701 options->flags = orig_flags;
4702 return ignored;
4703 }
4704
4705 void diff_addremove(struct diff_options *options,
4706 int addremove, unsigned mode,
4707 const unsigned char *sha1,
4708 int sha1_valid,
4709 const char *concatpath, unsigned dirty_submodule)
4710 {
4711 struct diff_filespec *one, *two;
4712
4713 if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
4714 return;
4715
4716 /* This may look odd, but it is a preparation for
4717 * feeding "there are unchanged files which should
4718 * not produce diffs, but when you are doing copy
4719 * detection you would need them, so here they are"
4720 * entries to the diff-core. They will be prefixed
4721 * with something like '=' or '*' (I haven't decided
4722 * which but should not make any difference).
4723 * Feeding the same new and old to diff_change()
4724 * also has the same effect.
4725 * Before the final output happens, they are pruned after
4726 * merged into rename/copy pairs as appropriate.
4727 */
4728 if (DIFF_OPT_TST(options, REVERSE_DIFF))
4729 addremove = (addremove == '+' ? '-' :
4730 addremove == '-' ? '+' : addremove);
4731
4732 if (options->prefix &&
4733 strncmp(concatpath, options->prefix, options->prefix_length))
4734 return;
4735
4736 one = alloc_filespec(concatpath);
4737 two = alloc_filespec(concatpath);
4738
4739 if (addremove != '+')
4740 fill_filespec(one, sha1, sha1_valid, mode);
4741 if (addremove != '-') {
4742 fill_filespec(two, sha1, sha1_valid, mode);
4743 two->dirty_submodule = dirty_submodule;
4744 }
4745
4746 diff_queue(&diff_queued_diff, one, two);
4747 if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4748 DIFF_OPT_SET(options, HAS_CHANGES);
4749 }
4750
4751 void diff_change(struct diff_options *options,
4752 unsigned old_mode, unsigned new_mode,
4753 const unsigned char *old_sha1,
4754 const unsigned char *new_sha1,
4755 int old_sha1_valid, int new_sha1_valid,
4756 const char *concatpath,
4757 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
4758 {
4759 struct diff_filespec *one, *two;
4760
4761 if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
4762 is_submodule_ignored(concatpath, options))
4763 return;
4764
4765 if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
4766 unsigned tmp;
4767 const unsigned char *tmp_c;
4768 tmp = old_mode; old_mode = new_mode; new_mode = tmp;
4769 tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
4770 tmp = old_sha1_valid; old_sha1_valid = new_sha1_valid;
4771 new_sha1_valid = tmp;
4772 tmp = old_dirty_submodule; old_dirty_submodule = new_dirty_submodule;
4773 new_dirty_submodule = tmp;
4774 }
4775
4776 if (options->prefix &&
4777 strncmp(concatpath, options->prefix, options->prefix_length))
4778 return;
4779
4780 one = alloc_filespec(concatpath);
4781 two = alloc_filespec(concatpath);
4782 fill_filespec(one, old_sha1, old_sha1_valid, old_mode);
4783 fill_filespec(two, new_sha1, new_sha1_valid, new_mode);
4784 one->dirty_submodule = old_dirty_submodule;
4785 two->dirty_submodule = new_dirty_submodule;
4786
4787 diff_queue(&diff_queued_diff, one, two);
4788 if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4789 DIFF_OPT_SET(options, HAS_CHANGES);
4790 }
4791
4792 struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
4793 {
4794 struct diff_filepair *pair;
4795 struct diff_filespec *one, *two;
4796
4797 if (options->prefix &&
4798 strncmp(path, options->prefix, options->prefix_length))
4799 return NULL;
4800
4801 one = alloc_filespec(path);
4802 two = alloc_filespec(path);
4803 pair = diff_queue(&diff_queued_diff, one, two);
4804 pair->is_unmerged = 1;
4805 return pair;
4806 }
4807
4808 static char *run_textconv(const char *pgm, struct diff_filespec *spec,
4809 size_t *outsize)
4810 {
4811 struct diff_tempfile *temp;
4812 const char *argv[3];
4813 const char **arg = argv;
4814 struct child_process child;
4815 struct strbuf buf = STRBUF_INIT;
4816 int err = 0;
4817
4818 temp = prepare_temp_file(spec->path, spec);
4819 *arg++ = pgm;
4820 *arg++ = temp->name;
4821 *arg = NULL;
4822
4823 memset(&child, 0, sizeof(child));
4824 child.use_shell = 1;
4825 child.argv = argv;
4826 child.out = -1;
4827 if (start_command(&child)) {
4828 remove_tempfile();
4829 return NULL;
4830 }
4831
4832 if (strbuf_read(&buf, child.out, 0) < 0)
4833 err = error("error reading from textconv command '%s'", pgm);
4834 close(child.out);
4835
4836 if (finish_command(&child) || err) {
4837 strbuf_release(&buf);
4838 remove_tempfile();
4839 return NULL;
4840 }
4841 remove_tempfile();
4842
4843 return strbuf_detach(&buf, outsize);
4844 }
4845
4846 size_t fill_textconv(struct userdiff_driver *driver,
4847 struct diff_filespec *df,
4848 char **outbuf)
4849 {
4850 size_t size;
4851
4852 if (!driver || !driver->textconv) {
4853 if (!DIFF_FILE_VALID(df)) {
4854 *outbuf = "";
4855 return 0;
4856 }
4857 if (diff_populate_filespec(df, 0))
4858 die("unable to read files to diff");
4859 *outbuf = df->data;
4860 return df->size;
4861 }
4862
4863 if (driver->textconv_cache && df->sha1_valid) {
4864 *outbuf = notes_cache_get(driver->textconv_cache, df->sha1,
4865 &size);
4866 if (*outbuf)
4867 return size;
4868 }
4869
4870 *outbuf = run_textconv(driver->textconv, df, &size);
4871 if (!*outbuf)
4872 die("unable to read files to diff");
4873
4874 if (driver->textconv_cache && df->sha1_valid) {
4875 /* ignore errors, as we might be in a readonly repository */
4876 notes_cache_put(driver->textconv_cache, df->sha1, *outbuf,
4877 size);
4878 /*
4879 * we could save up changes and flush them all at the end,
4880 * but we would need an extra call after all diffing is done.
4881 * Since generating a cache entry is the slow path anyway,
4882 * this extra overhead probably isn't a big deal.
4883 */
4884 notes_cache_write(driver->textconv_cache);
4885 }
4886
4887 return size;
4888 }
4889
4890 void setup_diff_pager(struct diff_options *opt)
4891 {
4892 /*
4893 * If the user asked for our exit code, then either they want --quiet
4894 * or --exit-code. We should definitely not bother with a pager in the
4895 * former case, as we will generate no output. Since we still properly
4896 * report our exit code even when a pager is run, we _could_ run a
4897 * pager with --exit-code. But since we have not done so historically,
4898 * and because it is easy to find people oneline advising "git diff
4899 * --exit-code" in hooks and other scripts, we do not do so.
4900 */
4901 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4902 check_pager_config("diff") != 0)
4903 setup_pager();
4904 }
Jonathan Nieder’s miscellaneous git branches