?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
diff: undocument the compaction heuristic knobs for experimentation
[git.git] / diff.c
blob05ca3ce08e21082f66ea9a8c6d0b51d1e56c6815
1 /*
2 * Copyright (C) 2005 Junio C Hamano
3 */
4 #include "cache.h"
5 #include "quote.h"
6 #include "diff.h"
7 #include "diffcore.h"
8 #include "delta.h"
9 #include "xdiff-interface.h"
10 #include "color.h"
11 #include "attr.h"
12 #include "run-command.h"
13 #include "utf8.h"
14 #include "userdiff.h"
15 #include "sigchain.h"
16 #include "submodule.h"
17 #include "ll-merge.h"
18 #include "string-list.h"
19 #include "argv-array.h"
20
21 #ifdef NO_FAST_WORKING_DIRECTORY
22 #define FAST_WORKING_DIRECTORY 0
23 #else
24 #define FAST_WORKING_DIRECTORY 1
25 #endif
26
27 static int diff_detect_rename_default;
28 static int diff_compaction_heuristic = 1;
29 static int diff_rename_limit_default = 400;
30 static int diff_suppress_blank_empty;
31 static int diff_use_color_default = -1;
32 static int diff_context_default = 3;
33 static const char *diff_word_regex_cfg;
34 static const char *external_diff_cmd_cfg;
35 static const char *diff_order_file_cfg;
36 int diff_auto_refresh_index = 1;
37 static int diff_mnemonic_prefix;
38 static int diff_no_prefix;
39 static int diff_stat_graph_width;
40 static int diff_dirstat_permille_default = 30;
41 static struct diff_options default_diff_options;
42 static long diff_algorithm;
43
44 static char diff_colors[][COLOR_MAXLEN] = {
45 GIT_COLOR_RESET,
46 GIT_COLOR_NORMAL, /* CONTEXT */
47 GIT_COLOR_BOLD, /* METAINFO */
48 GIT_COLOR_CYAN, /* FRAGINFO */
49 GIT_COLOR_RED, /* OLD */
50 GIT_COLOR_GREEN, /* NEW */
51 GIT_COLOR_YELLOW, /* COMMIT */
52 GIT_COLOR_BG_RED, /* WHITESPACE */
53 GIT_COLOR_NORMAL, /* FUNCINFO */
54 };
55
56 static int parse_diff_color_slot(const char *var)
57 {
58 if (!strcasecmp(var, "context") || !strcasecmp(var, "plain"))
59 return DIFF_CONTEXT;
60 if (!strcasecmp(var, "meta"))
61 return DIFF_METAINFO;
62 if (!strcasecmp(var, "frag"))
63 return DIFF_FRAGINFO;
64 if (!strcasecmp(var, "old"))
65 return DIFF_FILE_OLD;
66 if (!strcasecmp(var, "new"))
67 return DIFF_FILE_NEW;
68 if (!strcasecmp(var, "commit"))
69 return DIFF_COMMIT;
70 if (!strcasecmp(var, "whitespace"))
71 return DIFF_WHITESPACE;
72 if (!strcasecmp(var, "func"))
73 return DIFF_FUNCINFO;
74 return -1;
75 }
76
77 static int parse_dirstat_params(struct diff_options *options, const char *params_string,
78 struct strbuf *errmsg)
79 {
80 char *params_copy = xstrdup(params_string);
81 struct string_list params = STRING_LIST_INIT_NODUP;
82 int ret = 0;
83 int i;
84
85 if (*params_copy)
86 string_list_split_in_place(&params, params_copy, ',', -1);
87 for (i = 0; i < params.nr; i++) {
88 const char *p = params.items[i].string;
89 if (!strcmp(p, "changes")) {
90 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
91 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
92 } else if (!strcmp(p, "lines")) {
93 DIFF_OPT_SET(options, DIRSTAT_BY_LINE);
94 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
95 } else if (!strcmp(p, "files")) {
96 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
97 DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
98 } else if (!strcmp(p, "noncumulative")) {
99 DIFF_OPT_CLR(options, DIRSTAT_CUMULATIVE);
100 } else if (!strcmp(p, "cumulative")) {
101 DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
102 } else if (isdigit(*p)) {
103 char *end;
104 int permille = strtoul(p, &end, 10) * 10;
105 if (*end == '.' && isdigit(*++end)) {
106 /* only use first digit */
107 permille += *end - '0';
108 /* .. and ignore any further digits */
109 while (isdigit(*++end))
110 ; /* nothing */
111 }
112 if (!*end)
113 options->dirstat_permille = permille;
114 else {
115 strbuf_addf(errmsg, _(" Failed to parse dirstat cut-off percentage '%s'\n"),
116 p);
117 ret++;
118 }
119 } else {
120 strbuf_addf(errmsg, _(" Unknown dirstat parameter '%s'\n"), p);
121 ret++;
122 }
123
124 }
125 string_list_clear(&params, 0);
126 free(params_copy);
127 return ret;
128 }
129
130 static int parse_submodule_params(struct diff_options *options, const char *value)
131 {
132 if (!strcmp(value, "log"))
133 DIFF_OPT_SET(options, SUBMODULE_LOG);
134 else if (!strcmp(value, "short"))
135 DIFF_OPT_CLR(options, SUBMODULE_LOG);
136 else
137 return -1;
138 return 0;
139 }
140
141 static int git_config_rename(const char *var, const char *value)
142 {
143 if (!value)
144 return DIFF_DETECT_RENAME;
145 if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
146 return DIFF_DETECT_COPY;
147 return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
148 }
149
150 long parse_algorithm_value(const char *value)
151 {
152 if (!value)
153 return -1;
154 else if (!strcasecmp(value, "myers") || !strcasecmp(value, "default"))
155 return 0;
156 else if (!strcasecmp(value, "minimal"))
157 return XDF_NEED_MINIMAL;
158 else if (!strcasecmp(value, "patience"))
159 return XDF_PATIENCE_DIFF;
160 else if (!strcasecmp(value, "histogram"))
161 return XDF_HISTOGRAM_DIFF;
162 return -1;
163 }
164
165 /*
166 * These are to give UI layer defaults.
167 * The core-level commands such as git-diff-files should
168 * never be affected by the setting of diff.renames
169 * the user happens to have in the configuration file.
170 */
171 int git_diff_ui_config(const char *var, const char *value, void *cb)
172 {
173 if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
174 diff_use_color_default = git_config_colorbool(var, value);
175 return 0;
176 }
177 if (!strcmp(var, "diff.context")) {
178 diff_context_default = git_config_int(var, value);
179 if (diff_context_default < 0)
180 return -1;
181 return 0;
182 }
183 if (!strcmp(var, "diff.renames")) {
184 diff_detect_rename_default = git_config_rename(var, value);
185 return 0;
186 }
187 if (!strcmp(var, "diff.compactionheuristic")) {
188 diff_compaction_heuristic = git_config_bool(var, value);
189 return 0;
190 }
191 if (!strcmp(var, "diff.autorefreshindex")) {
192 diff_auto_refresh_index = git_config_bool(var, value);
193 return 0;
194 }
195 if (!strcmp(var, "diff.mnemonicprefix")) {
196 diff_mnemonic_prefix = git_config_bool(var, value);
197 return 0;
198 }
199 if (!strcmp(var, "diff.noprefix")) {
200 diff_no_prefix = git_config_bool(var, value);
201 return 0;
202 }
203 if (!strcmp(var, "diff.statgraphwidth")) {
204 diff_stat_graph_width = git_config_int(var, value);
205 return 0;
206 }
207 if (!strcmp(var, "diff.external"))
208 return git_config_string(&external_diff_cmd_cfg, var, value);
209 if (!strcmp(var, "diff.wordregex"))
210 return git_config_string(&diff_word_regex_cfg, var, value);
211 if (!strcmp(var, "diff.orderfile"))
212 return git_config_pathname(&diff_order_file_cfg, var, value);
213
214 if (!strcmp(var, "diff.ignoresubmodules"))
215 handle_ignore_submodules_arg(&default_diff_options, value);
216
217 if (!strcmp(var, "diff.submodule")) {
218 if (parse_submodule_params(&default_diff_options, value))
219 warning(_("Unknown value for 'diff.submodule' config variable: '%s'"),
220 value);
221 return 0;
222 }
223
224 if (!strcmp(var, "diff.algorithm")) {
225 diff_algorithm = parse_algorithm_value(value);
226 if (diff_algorithm < 0)
227 return -1;
228 return 0;
229 }
230
231 if (git_color_config(var, value, cb) < 0)
232 return -1;
233
234 return git_diff_basic_config(var, value, cb);
235 }
236
237 int git_diff_basic_config(const char *var, const char *value, void *cb)
238 {
239 const char *name;
240
241 if (!strcmp(var, "diff.renamelimit")) {
242 diff_rename_limit_default = git_config_int(var, value);
243 return 0;
244 }
245
246 if (userdiff_config(var, value) < 0)
247 return -1;
248
249 if (skip_prefix(var, "diff.color.", &name) ||
250 skip_prefix(var, "color.diff.", &name)) {
251 int slot = parse_diff_color_slot(name);
252 if (slot < 0)
253 return 0;
254 if (!value)
255 return config_error_nonbool(var);
256 return color_parse(value, diff_colors[slot]);
257 }
258
259 /* like GNU diff's --suppress-blank-empty option */
260 if (!strcmp(var, "diff.suppressblankempty") ||
261 /* for backwards compatibility */
262 !strcmp(var, "diff.suppress-blank-empty")) {
263 diff_suppress_blank_empty = git_config_bool(var, value);
264 return 0;
265 }
266
267 if (!strcmp(var, "diff.dirstat")) {
268 struct strbuf errmsg = STRBUF_INIT;
269 default_diff_options.dirstat_permille = diff_dirstat_permille_default;
270 if (parse_dirstat_params(&default_diff_options, value, &errmsg))
271 warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
272 errmsg.buf);
273 strbuf_release(&errmsg);
274 diff_dirstat_permille_default = default_diff_options.dirstat_permille;
275 return 0;
276 }
277
278 if (starts_with(var, "submodule."))
279 return parse_submodule_config_option(var, value);
280
281 return git_default_config(var, value, cb);
282 }
283
284 static char *quote_two(const char *one, const char *two)
285 {
286 int need_one = quote_c_style(one, NULL, NULL, 1);
287 int need_two = quote_c_style(two, NULL, NULL, 1);
288 struct strbuf res = STRBUF_INIT;
289
290 if (need_one + need_two) {
291 strbuf_addch(&res, '"');
292 quote_c_style(one, &res, NULL, 1);
293 quote_c_style(two, &res, NULL, 1);
294 strbuf_addch(&res, '"');
295 } else {
296 strbuf_addstr(&res, one);
297 strbuf_addstr(&res, two);
298 }
299 return strbuf_detach(&res, NULL);
300 }
301
302 static const char *external_diff(void)
303 {
304 static const char *external_diff_cmd = NULL;
305 static int done_preparing = 0;
306
307 if (done_preparing)
308 return external_diff_cmd;
309 external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
310 if (!external_diff_cmd)
311 external_diff_cmd = external_diff_cmd_cfg;
312 done_preparing = 1;
313 return external_diff_cmd;
314 }
315
316 static struct diff_tempfile {
317 const char *name; /* filename external diff should read from */
318 char hex[41];
319 char mode[10];
320 char tmp_path[PATH_MAX];
321 } diff_temp[2];
322
323 typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
324
325 struct emit_callback {
326 int color_diff;
327 unsigned ws_rule;
328 int blank_at_eof_in_preimage;
329 int blank_at_eof_in_postimage;
330 int lno_in_preimage;
331 int lno_in_postimage;
332 sane_truncate_fn truncate;
333 const char **label_path;
334 struct diff_words_data *diff_words;
335 struct diff_options *opt;
336 int *found_changesp;
337 struct strbuf *header;
338 };
339
340 static int count_lines(const char *data, int size)
341 {
342 int count, ch, completely_empty = 1, nl_just_seen = 0;
343 count = 0;
344 while (0 < size--) {
345 ch = *data++;
346 if (ch == '\n') {
347 count++;
348 nl_just_seen = 1;
349 completely_empty = 0;
350 }
351 else {
352 nl_just_seen = 0;
353 completely_empty = 0;
354 }
355 }
356 if (completely_empty)
357 return 0;
358 if (!nl_just_seen)
359 count++; /* no trailing newline */
360 return count;
361 }
362
363 static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
364 {
365 if (!DIFF_FILE_VALID(one)) {
366 mf->ptr = (char *)""; /* does not matter */
367 mf->size = 0;
368 return 0;
369 }
370 else if (diff_populate_filespec(one, 0))
371 return -1;
372
373 mf->ptr = one->data;
374 mf->size = one->size;
375 return 0;
376 }
377
378 /* like fill_mmfile, but only for size, so we can avoid retrieving blob */
379 static unsigned long diff_filespec_size(struct diff_filespec *one)
380 {
381 if (!DIFF_FILE_VALID(one))
382 return 0;
383 diff_populate_filespec(one, CHECK_SIZE_ONLY);
384 return one->size;
385 }
386
387 static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
388 {
389 char *ptr = mf->ptr;
390 long size = mf->size;
391 int cnt = 0;
392
393 if (!size)
394 return cnt;
395 ptr += size - 1; /* pointing at the very end */
396 if (*ptr != '\n')
397 ; /* incomplete line */
398 else
399 ptr--; /* skip the last LF */
400 while (mf->ptr < ptr) {
401 char *prev_eol;
402 for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
403 if (*prev_eol == '\n')
404 break;
405 if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
406 break;
407 cnt++;
408 ptr = prev_eol - 1;
409 }
410 return cnt;
411 }
412
413 static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
414 struct emit_callback *ecbdata)
415 {
416 int l1, l2, at;
417 unsigned ws_rule = ecbdata->ws_rule;
418 l1 = count_trailing_blank(mf1, ws_rule);
419 l2 = count_trailing_blank(mf2, ws_rule);
420 if (l2 <= l1) {
421 ecbdata->blank_at_eof_in_preimage = 0;
422 ecbdata->blank_at_eof_in_postimage = 0;
423 return;
424 }
425 at = count_lines(mf1->ptr, mf1->size);
426 ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
427
428 at = count_lines(mf2->ptr, mf2->size);
429 ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
430 }
431
432 static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
433 int first, const char *line, int len)
434 {
435 int has_trailing_newline, has_trailing_carriage_return;
436 int nofirst;
437 FILE *file = o->file;
438
439 fputs(diff_line_prefix(o), file);
440
441 if (len == 0) {
442 has_trailing_newline = (first == '\n');
443 has_trailing_carriage_return = (!has_trailing_newline &&
444 (first == '\r'));
445 nofirst = has_trailing_newline || has_trailing_carriage_return;
446 } else {
447 has_trailing_newline = (len > 0 && line[len-1] == '\n');
448 if (has_trailing_newline)
449 len--;
450 has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
451 if (has_trailing_carriage_return)
452 len--;
453 nofirst = 0;
454 }
455
456 if (len || !nofirst) {
457 fputs(set, file);
458 if (!nofirst)
459 fputc(first, file);
460 fwrite(line, len, 1, file);
461 fputs(reset, file);
462 }
463 if (has_trailing_carriage_return)
464 fputc('\r', file);
465 if (has_trailing_newline)
466 fputc('\n', file);
467 }
468
469 static void emit_line(struct diff_options *o, const char *set, const char *reset,
470 const char *line, int len)
471 {
472 emit_line_0(o, set, reset, line[0], line+1, len-1);
473 }
474
475 static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
476 {
477 if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
478 ecbdata->blank_at_eof_in_preimage &&
479 ecbdata->blank_at_eof_in_postimage &&
480 ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
481 ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
482 return 0;
483 return ws_blank_line(line, len, ecbdata->ws_rule);
484 }
485
486 static void emit_add_line(const char *reset,
487 struct emit_callback *ecbdata,
488 const char *line, int len)
489 {
490 const char *ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
491 const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_NEW);
492
493 if (!*ws)
494 emit_line_0(ecbdata->opt, set, reset, '+', line, len);
495 else if (new_blank_line_at_eof(ecbdata, line, len))
496 /* Blank line at EOF - paint '+' as well */
497 emit_line_0(ecbdata->opt, ws, reset, '+', line, len);
498 else {
499 /* Emit just the prefix, then the rest. */
500 emit_line_0(ecbdata->opt, set, reset, '+', "", 0);
501 ws_check_emit(line, len, ecbdata->ws_rule,
502 ecbdata->opt->file, set, reset, ws);
503 }
504 }
505
506 static void emit_hunk_header(struct emit_callback *ecbdata,
507 const char *line, int len)
508 {
509 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
510 const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
511 const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
512 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
513 static const char atat[2] = { '@', '@' };
514 const char *cp, *ep;
515 struct strbuf msgbuf = STRBUF_INIT;
516 int org_len = len;
517 int i = 1;
518
519 /*
520 * As a hunk header must begin with "@@ -<old>, +<new> @@",
521 * it always is at least 10 bytes long.
522 */
523 if (len < 10 ||
524 memcmp(line, atat, 2) ||
525 !(ep = memmem(line + 2, len - 2, atat, 2))) {
526 emit_line(ecbdata->opt, context, reset, line, len);
527 return;
528 }
529 ep += 2; /* skip over @@ */
530
531 /* The hunk header in fraginfo color */
532 strbuf_addstr(&msgbuf, frag);
533 strbuf_add(&msgbuf, line, ep - line);
534 strbuf_addstr(&msgbuf, reset);
535
536 /*
537 * trailing "\r\n"
538 */
539 for ( ; i < 3; i++)
540 if (line[len - i] == '\r' || line[len - i] == '\n')
541 len--;
542
543 /* blank before the func header */
544 for (cp = ep; ep - line < len; ep++)
545 if (*ep != ' ' && *ep != '\t')
546 break;
547 if (ep != cp) {
548 strbuf_addstr(&msgbuf, context);
549 strbuf_add(&msgbuf, cp, ep - cp);
550 strbuf_addstr(&msgbuf, reset);
551 }
552
553 if (ep < line + len) {
554 strbuf_addstr(&msgbuf, func);
555 strbuf_add(&msgbuf, ep, line + len - ep);
556 strbuf_addstr(&msgbuf, reset);
557 }
558
559 strbuf_add(&msgbuf, line + len, org_len - len);
560 emit_line(ecbdata->opt, "", "", msgbuf.buf, msgbuf.len);
561 strbuf_release(&msgbuf);
562 }
563
564 static struct diff_tempfile *claim_diff_tempfile(void) {
565 int i;
566 for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
567 if (!diff_temp[i].name)
568 return diff_temp + i;
569 die("BUG: diff is failing to clean up its tempfiles");
570 }
571
572 static int remove_tempfile_installed;
573
574 static void remove_tempfile(void)
575 {
576 int i;
577 for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
578 if (diff_temp[i].name == diff_temp[i].tmp_path)
579 unlink_or_warn(diff_temp[i].name);
580 diff_temp[i].name = NULL;
581 }
582 }
583
584 static void remove_tempfile_on_signal(int signo)
585 {
586 remove_tempfile();
587 sigchain_pop(signo);
588 raise(signo);
589 }
590
591 static void print_line_count(FILE *file, int count)
592 {
593 switch (count) {
594 case 0:
595 fprintf(file, "0,0");
596 break;
597 case 1:
598 fprintf(file, "1");
599 break;
600 default:
601 fprintf(file, "1,%d", count);
602 break;
603 }
604 }
605
606 static void emit_rewrite_lines(struct emit_callback *ecb,
607 int prefix, const char *data, int size)
608 {
609 const char *endp = NULL;
610 static const char *nneof = " No newline at end of file\n";
611 const char *old = diff_get_color(ecb->color_diff, DIFF_FILE_OLD);
612 const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
613
614 while (0 < size) {
615 int len;
616
617 endp = memchr(data, '\n', size);
618 len = endp ? (endp - data + 1) : size;
619 if (prefix != '+') {
620 ecb->lno_in_preimage++;
621 emit_line_0(ecb->opt, old, reset, '-',
622 data, len);
623 } else {
624 ecb->lno_in_postimage++;
625 emit_add_line(reset, ecb, data, len);
626 }
627 size -= len;
628 data += len;
629 }
630 if (!endp) {
631 const char *context = diff_get_color(ecb->color_diff,
632 DIFF_CONTEXT);
633 putc('\n', ecb->opt->file);
634 emit_line_0(ecb->opt, context, reset, '\\',
635 nneof, strlen(nneof));
636 }
637 }
638
639 static void emit_rewrite_diff(const char *name_a,
640 const char *name_b,
641 struct diff_filespec *one,
642 struct diff_filespec *two,
643 struct userdiff_driver *textconv_one,
644 struct userdiff_driver *textconv_two,
645 struct diff_options *o)
646 {
647 int lc_a, lc_b;
648 const char *name_a_tab, *name_b_tab;
649 const char *metainfo = diff_get_color(o->use_color, DIFF_METAINFO);
650 const char *fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
651 const char *reset = diff_get_color(o->use_color, DIFF_RESET);
652 static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
653 const char *a_prefix, *b_prefix;
654 char *data_one, *data_two;
655 size_t size_one, size_two;
656 struct emit_callback ecbdata;
657 const char *line_prefix = diff_line_prefix(o);
658
659 if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
660 a_prefix = o->b_prefix;
661 b_prefix = o->a_prefix;
662 } else {
663 a_prefix = o->a_prefix;
664 b_prefix = o->b_prefix;
665 }
666
667 name_a += (*name_a == '/');
668 name_b += (*name_b == '/');
669 name_a_tab = strchr(name_a, ' ') ? "\t" : "";
670 name_b_tab = strchr(name_b, ' ') ? "\t" : "";
671
672 strbuf_reset(&a_name);
673 strbuf_reset(&b_name);
674 quote_two_c_style(&a_name, a_prefix, name_a, 0);
675 quote_two_c_style(&b_name, b_prefix, name_b, 0);
676
677 size_one = fill_textconv(textconv_one, one, &data_one);
678 size_two = fill_textconv(textconv_two, two, &data_two);
679
680 memset(&ecbdata, 0, sizeof(ecbdata));
681 ecbdata.color_diff = want_color(o->use_color);
682 ecbdata.found_changesp = &o->found_changes;
683 ecbdata.ws_rule = whitespace_rule(name_b);
684 ecbdata.opt = o;
685 if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
686 mmfile_t mf1, mf2;
687 mf1.ptr = (char *)data_one;
688 mf2.ptr = (char *)data_two;
689 mf1.size = size_one;
690 mf2.size = size_two;
691 check_blank_at_eof(&mf1, &mf2, &ecbdata);
692 }
693 ecbdata.lno_in_preimage = 1;
694 ecbdata.lno_in_postimage = 1;
695
696 lc_a = count_lines(data_one, size_one);
697 lc_b = count_lines(data_two, size_two);
698 fprintf(o->file,
699 "%s%s--- %s%s%s\n%s%s+++ %s%s%s\n%s%s@@ -",
700 line_prefix, metainfo, a_name.buf, name_a_tab, reset,
701 line_prefix, metainfo, b_name.buf, name_b_tab, reset,
702 line_prefix, fraginfo);
703 if (!o->irreversible_delete)
704 print_line_count(o->file, lc_a);
705 else
706 fprintf(o->file, "?,?");
707 fprintf(o->file, " +");
708 print_line_count(o->file, lc_b);
709 fprintf(o->file, " @@%s\n", reset);
710 if (lc_a && !o->irreversible_delete)
711 emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
712 if (lc_b)
713 emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
714 if (textconv_one)
715 free((char *)data_one);
716 if (textconv_two)
717 free((char *)data_two);
718 }
719
720 struct diff_words_buffer {
721 mmfile_t text;
722 long alloc;
723 struct diff_words_orig {
724 const char *begin, *end;
725 } *orig;
726 int orig_nr, orig_alloc;
727 };
728
729 static void diff_words_append(char *line, unsigned long len,
730 struct diff_words_buffer *buffer)
731 {
732 ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
733 line++;
734 len--;
735 memcpy(buffer->text.ptr + buffer->text.size, line, len);
736 buffer->text.size += len;
737 buffer->text.ptr[buffer->text.size] = '\0';
738 }
739
740 struct diff_words_style_elem {
741 const char *prefix;
742 const char *suffix;
743 const char *color; /* NULL; filled in by the setup code if
744 * color is enabled */
745 };
746
747 struct diff_words_style {
748 enum diff_words_type type;
749 struct diff_words_style_elem new, old, ctx;
750 const char *newline;
751 };
752
753 static struct diff_words_style diff_words_styles[] = {
754 { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
755 { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
756 { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
757 };
758
759 struct diff_words_data {
760 struct diff_words_buffer minus, plus;
761 const char *current_plus;
762 int last_minus;
763 struct diff_options *opt;
764 regex_t *word_regex;
765 enum diff_words_type type;
766 struct diff_words_style *style;
767 };
768
769 static int fn_out_diff_words_write_helper(FILE *fp,
770 struct diff_words_style_elem *st_el,
771 const char *newline,
772 size_t count, const char *buf,
773 const char *line_prefix)
774 {
775 int print = 0;
776
777 while (count) {
778 char *p = memchr(buf, '\n', count);
779 if (print)
780 fputs(line_prefix, fp);
781 if (p != buf) {
782 if (st_el->color && fputs(st_el->color, fp) < 0)
783 return -1;
784 if (fputs(st_el->prefix, fp) < 0 ||
785 fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
786 fputs(st_el->suffix, fp) < 0)
787 return -1;
788 if (st_el->color && *st_el->color
789 && fputs(GIT_COLOR_RESET, fp) < 0)
790 return -1;
791 }
792 if (!p)
793 return 0;
794 if (fputs(newline, fp) < 0)
795 return -1;
796 count -= p + 1 - buf;
797 buf = p + 1;
798 print = 1;
799 }
800 return 0;
801 }
802
803 /*
804 * '--color-words' algorithm can be described as:
805 *
806 * 1. collect a the minus/plus lines of a diff hunk, divided into
807 * minus-lines and plus-lines;
808 *
809 * 2. break both minus-lines and plus-lines into words and
810 * place them into two mmfile_t with one word for each line;
811 *
812 * 3. use xdiff to run diff on the two mmfile_t to get the words level diff;
813 *
814 * And for the common parts of the both file, we output the plus side text.
815 * diff_words->current_plus is used to trace the current position of the plus file
816 * which printed. diff_words->last_minus is used to trace the last minus word
817 * printed.
818 *
819 * For '--graph' to work with '--color-words', we need to output the graph prefix
820 * on each line of color words output. Generally, there are two conditions on
821 * which we should output the prefix.
822 *
823 * 1. diff_words->last_minus == 0 &&
824 * diff_words->current_plus == diff_words->plus.text.ptr
825 *
826 * that is: the plus text must start as a new line, and if there is no minus
827 * word printed, a graph prefix must be printed.
828 *
829 * 2. diff_words->current_plus > diff_words->plus.text.ptr &&
830 * *(diff_words->current_plus - 1) == '\n'
831 *
832 * that is: a graph prefix must be printed following a '\n'
833 */
834 static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
835 {
836 if ((diff_words->last_minus == 0 &&
837 diff_words->current_plus == diff_words->plus.text.ptr) ||
838 (diff_words->current_plus > diff_words->plus.text.ptr &&
839 *(diff_words->current_plus - 1) == '\n')) {
840 return 1;
841 } else {
842 return 0;
843 }
844 }
845
846 static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
847 {
848 struct diff_words_data *diff_words = priv;
849 struct diff_words_style *style = diff_words->style;
850 int minus_first, minus_len, plus_first, plus_len;
851 const char *minus_begin, *minus_end, *plus_begin, *plus_end;
852 struct diff_options *opt = diff_words->opt;
853 const char *line_prefix;
854
855 if (line[0] != '@' || parse_hunk_header(line, len,
856 &minus_first, &minus_len, &plus_first, &plus_len))
857 return;
858
859 assert(opt);
860 line_prefix = diff_line_prefix(opt);
861
862 /* POSIX requires that first be decremented by one if len == 0... */
863 if (minus_len) {
864 minus_begin = diff_words->minus.orig[minus_first].begin;
865 minus_end =
866 diff_words->minus.orig[minus_first + minus_len - 1].end;
867 } else
868 minus_begin = minus_end =
869 diff_words->minus.orig[minus_first].end;
870
871 if (plus_len) {
872 plus_begin = diff_words->plus.orig[plus_first].begin;
873 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
874 } else
875 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
876
877 if (color_words_output_graph_prefix(diff_words)) {
878 fputs(line_prefix, diff_words->opt->file);
879 }
880 if (diff_words->current_plus != plus_begin) {
881 fn_out_diff_words_write_helper(diff_words->opt->file,
882 &style->ctx, style->newline,
883 plus_begin - diff_words->current_plus,
884 diff_words->current_plus, line_prefix);
885 if (*(plus_begin - 1) == '\n')
886 fputs(line_prefix, diff_words->opt->file);
887 }
888 if (minus_begin != minus_end) {
889 fn_out_diff_words_write_helper(diff_words->opt->file,
890 &style->old, style->newline,
891 minus_end - minus_begin, minus_begin,
892 line_prefix);
893 }
894 if (plus_begin != plus_end) {
895 fn_out_diff_words_write_helper(diff_words->opt->file,
896 &style->new, style->newline,
897 plus_end - plus_begin, plus_begin,
898 line_prefix);
899 }
900
901 diff_words->current_plus = plus_end;
902 diff_words->last_minus = minus_first;
903 }
904
905 /* This function starts looking at *begin, and returns 0 iff a word was found. */
906 static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
907 int *begin, int *end)
908 {
909 if (word_regex && *begin < buffer->size) {
910 regmatch_t match[1];
911 if (!regexec(word_regex, buffer->ptr + *begin, 1, match, 0)) {
912 char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
913 '\n', match[0].rm_eo - match[0].rm_so);
914 *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
915 *begin += match[0].rm_so;
916 return *begin >= *end;
917 }
918 return -1;
919 }
920
921 /* find the next word */
922 while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
923 (*begin)++;
924 if (*begin >= buffer->size)
925 return -1;
926
927 /* find the end of the word */
928 *end = *begin + 1;
929 while (*end < buffer->size && !isspace(buffer->ptr[*end]))
930 (*end)++;
931
932 return 0;
933 }
934
935 /*
936 * This function splits the words in buffer->text, stores the list with
937 * newline separator into out, and saves the offsets of the original words
938 * in buffer->orig.
939 */
940 static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
941 regex_t *word_regex)
942 {
943 int i, j;
944 long alloc = 0;
945
946 out->size = 0;
947 out->ptr = NULL;
948
949 /* fake an empty "0th" word */
950 ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
951 buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
952 buffer->orig_nr = 1;
953
954 for (i = 0; i < buffer->text.size; i++) {
955 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
956 return;
957
958 /* store original boundaries */
959 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
960 buffer->orig_alloc);
961 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
962 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
963 buffer->orig_nr++;
964
965 /* store one word */
966 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
967 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
968 out->ptr[out->size + j - i] = '\n';
969 out->size += j - i + 1;
970
971 i = j - 1;
972 }
973 }
974
975 /* this executes the word diff on the accumulated buffers */
976 static void diff_words_show(struct diff_words_data *diff_words)
977 {
978 xpparam_t xpp;
979 xdemitconf_t xecfg;
980 mmfile_t minus, plus;
981 struct diff_words_style *style = diff_words->style;
982
983 struct diff_options *opt = diff_words->opt;
984 const char *line_prefix;
985
986 assert(opt);
987 line_prefix = diff_line_prefix(opt);
988
989 /* special case: only removal */
990 if (!diff_words->plus.text.size) {
991 fputs(line_prefix, diff_words->opt->file);
992 fn_out_diff_words_write_helper(diff_words->opt->file,
993 &style->old, style->newline,
994 diff_words->minus.text.size,
995 diff_words->minus.text.ptr, line_prefix);
996 diff_words->minus.text.size = 0;
997 return;
998 }
999
1000 diff_words->current_plus = diff_words->plus.text.ptr;
1001 diff_words->last_minus = 0;
1002
1003 memset(&xpp, 0, sizeof(xpp));
1004 memset(&xecfg, 0, sizeof(xecfg));
1005 diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
1006 diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
1007 xpp.flags = 0;
1008 /* as only the hunk header will be parsed, we need a 0-context */
1009 xecfg.ctxlen = 0;
1010 if (xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
1011 &xpp, &xecfg))
1012 die("unable to generate word diff");
1013 free(minus.ptr);
1014 free(plus.ptr);
1015 if (diff_words->current_plus != diff_words->plus.text.ptr +
1016 diff_words->plus.text.size) {
1017 if (color_words_output_graph_prefix(diff_words))
1018 fputs(line_prefix, diff_words->opt->file);
1019 fn_out_diff_words_write_helper(diff_words->opt->file,
1020 &style->ctx, style->newline,
1021 diff_words->plus.text.ptr + diff_words->plus.text.size
1022 - diff_words->current_plus, diff_words->current_plus,
1023 line_prefix);
1024 }
1025 diff_words->minus.text.size = diff_words->plus.text.size = 0;
1026 }
1027
1028 /* In "color-words" mode, show word-diff of words accumulated in the buffer */
1029 static void diff_words_flush(struct emit_callback *ecbdata)
1030 {
1031 if (ecbdata->diff_words->minus.text.size ||
1032 ecbdata->diff_words->plus.text.size)
1033 diff_words_show(ecbdata->diff_words);
1034 }
1035
1036 static void diff_filespec_load_driver(struct diff_filespec *one)
1037 {
1038 /* Use already-loaded driver */
1039 if (one->driver)
1040 return;
1041
1042 if (S_ISREG(one->mode))
1043 one->driver = userdiff_find_by_path(one->path);
1044
1045 /* Fallback to default settings */
1046 if (!one->driver)
1047 one->driver = userdiff_find_by_name("default");
1048 }
1049
1050 static const char *userdiff_word_regex(struct diff_filespec *one)
1051 {
1052 diff_filespec_load_driver(one);
1053 return one->driver->word_regex;
1054 }
1055
1056 static void init_diff_words_data(struct emit_callback *ecbdata,
1057 struct diff_options *orig_opts,
1058 struct diff_filespec *one,
1059 struct diff_filespec *two)
1060 {
1061 int i;
1062 struct diff_options *o = xmalloc(sizeof(struct diff_options));
1063 memcpy(o, orig_opts, sizeof(struct diff_options));
1064
1065 ecbdata->diff_words =
1066 xcalloc(1, sizeof(struct diff_words_data));
1067 ecbdata->diff_words->type = o->word_diff;
1068 ecbdata->diff_words->opt = o;
1069 if (!o->word_regex)
1070 o->word_regex = userdiff_word_regex(one);
1071 if (!o->word_regex)
1072 o->word_regex = userdiff_word_regex(two);
1073 if (!o->word_regex)
1074 o->word_regex = diff_word_regex_cfg;
1075 if (o->word_regex) {
1076 ecbdata->diff_words->word_regex = (regex_t *)
1077 xmalloc(sizeof(regex_t));
1078 if (regcomp(ecbdata->diff_words->word_regex,
1079 o->word_regex,
1080 REG_EXTENDED | REG_NEWLINE))
1081 die ("Invalid regular expression: %s",
1082 o->word_regex);
1083 }
1084 for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1085 if (o->word_diff == diff_words_styles[i].type) {
1086 ecbdata->diff_words->style =
1087 &diff_words_styles[i];
1088 break;
1089 }
1090 }
1091 if (want_color(o->use_color)) {
1092 struct diff_words_style *st = ecbdata->diff_words->style;
1093 st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1094 st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1095 st->ctx.color = diff_get_color_opt(o, DIFF_CONTEXT);
1096 }
1097 }
1098
1099 static void free_diff_words_data(struct emit_callback *ecbdata)
1100 {
1101 if (ecbdata->diff_words) {
1102 diff_words_flush(ecbdata);
1103 free (ecbdata->diff_words->opt);
1104 free (ecbdata->diff_words->minus.text.ptr);
1105 free (ecbdata->diff_words->minus.orig);
1106 free (ecbdata->diff_words->plus.text.ptr);
1107 free (ecbdata->diff_words->plus.orig);
1108 if (ecbdata->diff_words->word_regex) {
1109 regfree(ecbdata->diff_words->word_regex);
1110 free(ecbdata->diff_words->word_regex);
1111 }
1112 free(ecbdata->diff_words);
1113 ecbdata->diff_words = NULL;
1114 }
1115 }
1116
1117 const char *diff_get_color(int diff_use_color, enum color_diff ix)
1118 {
1119 if (want_color(diff_use_color))
1120 return diff_colors[ix];
1121 return "";
1122 }
1123
1124 const char *diff_line_prefix(struct diff_options *opt)
1125 {
1126 struct strbuf *msgbuf;
1127 if (!opt->output_prefix)
1128 return "";
1129
1130 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
1131 return msgbuf->buf;
1132 }
1133
1134 static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1135 {
1136 const char *cp;
1137 unsigned long allot;
1138 size_t l = len;
1139
1140 if (ecb->truncate)
1141 return ecb->truncate(line, len);
1142 cp = line;
1143 allot = l;
1144 while (0 < l) {
1145 (void) utf8_width(&cp, &l);
1146 if (!cp)
1147 break; /* truncated in the middle? */
1148 }
1149 return allot - l;
1150 }
1151
1152 static void find_lno(const char *line, struct emit_callback *ecbdata)
1153 {
1154 const char *p;
1155 ecbdata->lno_in_preimage = 0;
1156 ecbdata->lno_in_postimage = 0;
1157 p = strchr(line, '-');
1158 if (!p)
1159 return; /* cannot happen */
1160 ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1161 p = strchr(p, '+');
1162 if (!p)
1163 return; /* cannot happen */
1164 ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1165 }
1166
1167 static void fn_out_consume(void *priv, char *line, unsigned long len)
1168 {
1169 struct emit_callback *ecbdata = priv;
1170 const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
1171 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
1172 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1173 struct diff_options *o = ecbdata->opt;
1174 const char *line_prefix = diff_line_prefix(o);
1175
1176 if (ecbdata->header) {
1177 fprintf(ecbdata->opt->file, "%s", ecbdata->header->buf);
1178 strbuf_reset(ecbdata->header);
1179 ecbdata->header = NULL;
1180 }
1181 *(ecbdata->found_changesp) = 1;
1182
1183 if (ecbdata->label_path[0]) {
1184 const char *name_a_tab, *name_b_tab;
1185
1186 name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
1187 name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
1188
1189 fprintf(ecbdata->opt->file, "%s%s--- %s%s%s\n",
1190 line_prefix, meta, ecbdata->label_path[0], reset, name_a_tab);
1191 fprintf(ecbdata->opt->file, "%s%s+++ %s%s%s\n",
1192 line_prefix, meta, ecbdata->label_path[1], reset, name_b_tab);
1193 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1194 }
1195
1196 if (diff_suppress_blank_empty
1197 && len == 2 && line[0] == ' ' && line[1] == '\n') {
1198 line[0] = '\n';
1199 len = 1;
1200 }
1201
1202 if (line[0] == '@') {
1203 if (ecbdata->diff_words)
1204 diff_words_flush(ecbdata);
1205 len = sane_truncate_line(ecbdata, line, len);
1206 find_lno(line, ecbdata);
1207 emit_hunk_header(ecbdata, line, len);
1208 if (line[len-1] != '\n')
1209 putc('\n', ecbdata->opt->file);
1210 return;
1211 }
1212
1213 if (len < 1) {
1214 emit_line(ecbdata->opt, reset, reset, line, len);
1215 if (ecbdata->diff_words
1216 && ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN)
1217 fputs("~\n", ecbdata->opt->file);
1218 return;
1219 }
1220
1221 if (ecbdata->diff_words) {
1222 if (line[0] == '-') {
1223 diff_words_append(line, len,
1224 &ecbdata->diff_words->minus);
1225 return;
1226 } else if (line[0] == '+') {
1227 diff_words_append(line, len,
1228 &ecbdata->diff_words->plus);
1229 return;
1230 } else if (starts_with(line, "\\ ")) {
1231 /*
1232 * Eat the "no newline at eof" marker as if we
1233 * saw a "+" or "-" line with nothing on it,
1234 * and return without diff_words_flush() to
1235 * defer processing. If this is the end of
1236 * preimage, more "+" lines may come after it.
1237 */
1238 return;
1239 }
1240 diff_words_flush(ecbdata);
1241 if (ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN) {
1242 emit_line(ecbdata->opt, context, reset, line, len);
1243 fputs("~\n", ecbdata->opt->file);
1244 } else {
1245 /*
1246 * Skip the prefix character, if any. With
1247 * diff_suppress_blank_empty, there may be
1248 * none.
1249 */
1250 if (line[0] != '\n') {
1251 line++;
1252 len--;
1253 }
1254 emit_line(ecbdata->opt, context, reset, line, len);
1255 }
1256 return;
1257 }
1258
1259 if (line[0] != '+') {
1260 const char *color =
1261 diff_get_color(ecbdata->color_diff,
1262 line[0] == '-' ? DIFF_FILE_OLD : DIFF_CONTEXT);
1263 ecbdata->lno_in_preimage++;
1264 if (line[0] == ' ')
1265 ecbdata->lno_in_postimage++;
1266 emit_line(ecbdata->opt, color, reset, line, len);
1267 } else {
1268 ecbdata->lno_in_postimage++;
1269 emit_add_line(reset, ecbdata, line + 1, len - 1);
1270 }
1271 }
1272
1273 static char *pprint_rename(const char *a, const char *b)
1274 {
1275 const char *old = a;
1276 const char *new = b;
1277 struct strbuf name = STRBUF_INIT;
1278 int pfx_length, sfx_length;
1279 int pfx_adjust_for_slash;
1280 int len_a = strlen(a);
1281 int len_b = strlen(b);
1282 int a_midlen, b_midlen;
1283 int qlen_a = quote_c_style(a, NULL, NULL, 0);
1284 int qlen_b = quote_c_style(b, NULL, NULL, 0);
1285
1286 if (qlen_a || qlen_b) {
1287 quote_c_style(a, &name, NULL, 0);
1288 strbuf_addstr(&name, " => ");
1289 quote_c_style(b, &name, NULL, 0);
1290 return strbuf_detach(&name, NULL);
1291 }
1292
1293 /* Find common prefix */
1294 pfx_length = 0;
1295 while (*old && *new && *old == *new) {
1296 if (*old == '/')
1297 pfx_length = old - a + 1;
1298 old++;
1299 new++;
1300 }
1301
1302 /* Find common suffix */
1303 old = a + len_a;
1304 new = b + len_b;
1305 sfx_length = 0;
1306 /*
1307 * If there is a common prefix, it must end in a slash. In
1308 * that case we let this loop run 1 into the prefix to see the
1309 * same slash.
1310 *
1311 * If there is no common prefix, we cannot do this as it would
1312 * underrun the input strings.
1313 */
1314 pfx_adjust_for_slash = (pfx_length ? 1 : 0);
1315 while (a + pfx_length - pfx_adjust_for_slash <= old &&
1316 b + pfx_length - pfx_adjust_for_slash <= new &&
1317 *old == *new) {
1318 if (*old == '/')
1319 sfx_length = len_a - (old - a);
1320 old--;
1321 new--;
1322 }
1323
1324 /*
1325 * pfx{mid-a => mid-b}sfx
1326 * {pfx-a => pfx-b}sfx
1327 * pfx{sfx-a => sfx-b}
1328 * name-a => name-b
1329 */
1330 a_midlen = len_a - pfx_length - sfx_length;
1331 b_midlen = len_b - pfx_length - sfx_length;
1332 if (a_midlen < 0)
1333 a_midlen = 0;
1334 if (b_midlen < 0)
1335 b_midlen = 0;
1336
1337 strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1338 if (pfx_length + sfx_length) {
1339 strbuf_add(&name, a, pfx_length);
1340 strbuf_addch(&name, '{');
1341 }
1342 strbuf_add(&name, a + pfx_length, a_midlen);
1343 strbuf_addstr(&name, " => ");
1344 strbuf_add(&name, b + pfx_length, b_midlen);
1345 if (pfx_length + sfx_length) {
1346 strbuf_addch(&name, '}');
1347 strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1348 }
1349 return strbuf_detach(&name, NULL);
1350 }
1351
1352 struct diffstat_t {
1353 int nr;
1354 int alloc;
1355 struct diffstat_file {
1356 char *from_name;
1357 char *name;
1358 char *print_name;
1359 unsigned is_unmerged:1;
1360 unsigned is_binary:1;
1361 unsigned is_renamed:1;
1362 unsigned is_interesting:1;
1363 uintmax_t added, deleted;
1364 } **files;
1365 };
1366
1367 static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1368 const char *name_a,
1369 const char *name_b)
1370 {
1371 struct diffstat_file *x;
1372 x = xcalloc(1, sizeof(*x));
1373 ALLOC_GROW(diffstat->files, diffstat->nr + 1, diffstat->alloc);
1374 diffstat->files[diffstat->nr++] = x;
1375 if (name_b) {
1376 x->from_name = xstrdup(name_a);
1377 x->name = xstrdup(name_b);
1378 x->is_renamed = 1;
1379 }
1380 else {
1381 x->from_name = NULL;
1382 x->name = xstrdup(name_a);
1383 }
1384 return x;
1385 }
1386
1387 static void diffstat_consume(void *priv, char *line, unsigned long len)
1388 {
1389 struct diffstat_t *diffstat = priv;
1390 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1391
1392 if (line[0] == '+')
1393 x->added++;
1394 else if (line[0] == '-')
1395 x->deleted++;
1396 }
1397
1398 const char mime_boundary_leader[] = "------------";
1399
1400 static int scale_linear(int it, int width, int max_change)
1401 {
1402 if (!it)
1403 return 0;
1404 /*
1405 * make sure that at least one '-' or '+' is printed if
1406 * there is any change to this path. The easiest way is to
1407 * scale linearly as if the alloted width is one column shorter
1408 * than it is, and then add 1 to the result.
1409 */
1410 return 1 + (it * (width - 1) / max_change);
1411 }
1412
1413 static void show_name(FILE *file,
1414 const char *prefix, const char *name, int len)
1415 {
1416 fprintf(file, " %s%-*s |", prefix, len, name);
1417 }
1418
1419 static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1420 {
1421 if (cnt <= 0)
1422 return;
1423 fprintf(file, "%s", set);
1424 while (cnt--)
1425 putc(ch, file);
1426 fprintf(file, "%s", reset);
1427 }
1428
1429 static void fill_print_name(struct diffstat_file *file)
1430 {
1431 char *pname;
1432
1433 if (file->print_name)
1434 return;
1435
1436 if (!file->is_renamed) {
1437 struct strbuf buf = STRBUF_INIT;
1438 if (quote_c_style(file->name, &buf, NULL, 0)) {
1439 pname = strbuf_detach(&buf, NULL);
1440 } else {
1441 pname = file->name;
1442 strbuf_release(&buf);
1443 }
1444 } else {
1445 pname = pprint_rename(file->from_name, file->name);
1446 }
1447 file->print_name = pname;
1448 }
1449
1450 int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1451 {
1452 struct strbuf sb = STRBUF_INIT;
1453 int ret;
1454
1455 if (!files) {
1456 assert(insertions == 0 && deletions == 0);
1457 return fprintf(fp, "%s\n", " 0 files changed");
1458 }
1459
1460 strbuf_addf(&sb,
1461 (files == 1) ? " %d file changed" : " %d files changed",
1462 files);
1463
1464 /*
1465 * For binary diff, the caller may want to print "x files
1466 * changed" with insertions == 0 && deletions == 0.
1467 *
1468 * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1469 * is probably less confusing (i.e skip over "2 files changed
1470 * but nothing about added/removed lines? Is this a bug in Git?").
1471 */
1472 if (insertions || deletions == 0) {
1473 strbuf_addf(&sb,
1474 (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
1475 insertions);
1476 }
1477
1478 if (deletions || insertions == 0) {
1479 strbuf_addf(&sb,
1480 (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
1481 deletions);
1482 }
1483 strbuf_addch(&sb, '\n');
1484 ret = fputs(sb.buf, fp);
1485 strbuf_release(&sb);
1486 return ret;
1487 }
1488
1489 static void show_stats(struct diffstat_t *data, struct diff_options *options)
1490 {
1491 int i, len, add, del, adds = 0, dels = 0;
1492 uintmax_t max_change = 0, max_len = 0;
1493 int total_files = data->nr, count;
1494 int width, name_width, graph_width, number_width = 0, bin_width = 0;
1495 const char *reset, *add_c, *del_c;
1496 const char *line_prefix = "";
1497 int extra_shown = 0;
1498
1499 if (data->nr == 0)
1500 return;
1501
1502 line_prefix = diff_line_prefix(options);
1503 count = options->stat_count ? options->stat_count : data->nr;
1504
1505 reset = diff_get_color_opt(options, DIFF_RESET);
1506 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1507 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1508
1509 /*
1510 * Find the longest filename and max number of changes
1511 */
1512 for (i = 0; (i < count) && (i < data->nr); i++) {
1513 struct diffstat_file *file = data->files[i];
1514 uintmax_t change = file->added + file->deleted;
1515
1516 if (!file->is_interesting && (change == 0)) {
1517 count++; /* not shown == room for one more */
1518 continue;
1519 }
1520 fill_print_name(file);
1521 len = strlen(file->print_name);
1522 if (max_len < len)
1523 max_len = len;
1524
1525 if (file->is_unmerged) {
1526 /* "Unmerged" is 8 characters */
1527 bin_width = bin_width < 8 ? 8 : bin_width;
1528 continue;
1529 }
1530 if (file->is_binary) {
1531 /* "Bin XXX -> YYY bytes" */
1532 int w = 14 + decimal_width(file->added)
1533 + decimal_width(file->deleted);
1534 bin_width = bin_width < w ? w : bin_width;
1535 /* Display change counts aligned with "Bin" */
1536 number_width = 3;
1537 continue;
1538 }
1539
1540 if (max_change < change)
1541 max_change = change;
1542 }
1543 count = i; /* where we can stop scanning in data->files[] */
1544
1545 /*
1546 * We have width = stat_width or term_columns() columns total.
1547 * We want a maximum of min(max_len, stat_name_width) for the name part.
1548 * We want a maximum of min(max_change, stat_graph_width) for the +- part.
1549 * We also need 1 for " " and 4 + decimal_width(max_change)
1550 * for " | NNNN " and one the empty column at the end, altogether
1551 * 6 + decimal_width(max_change).
1552 *
1553 * If there's not enough space, we will use the smaller of
1554 * stat_name_width (if set) and 5/8*width for the filename,
1555 * and the rest for constant elements + graph part, but no more
1556 * than stat_graph_width for the graph part.
1557 * (5/8 gives 50 for filename and 30 for the constant parts + graph
1558 * for the standard terminal size).
1559 *
1560 * In other words: stat_width limits the maximum width, and
1561 * stat_name_width fixes the maximum width of the filename,
1562 * and is also used to divide available columns if there
1563 * aren't enough.
1564 *
1565 * Binary files are displayed with "Bin XXX -> YYY bytes"
1566 * instead of the change count and graph. This part is treated
1567 * similarly to the graph part, except that it is not
1568 * "scaled". If total width is too small to accommodate the
1569 * guaranteed minimum width of the filename part and the
1570 * separators and this message, this message will "overflow"
1571 * making the line longer than the maximum width.
1572 */
1573
1574 if (options->stat_width == -1)
1575 width = term_columns() - options->output_prefix_length;
1576 else
1577 width = options->stat_width ? options->stat_width : 80;
1578 number_width = decimal_width(max_change) > number_width ?
1579 decimal_width(max_change) : number_width;
1580
1581 if (options->stat_graph_width == -1)
1582 options->stat_graph_width = diff_stat_graph_width;
1583
1584 /*
1585 * Guarantee 3/8*16==6 for the graph part
1586 * and 5/8*16==10 for the filename part
1587 */
1588 if (width < 16 + 6 + number_width)
1589 width = 16 + 6 + number_width;
1590
1591 /*
1592 * First assign sizes that are wanted, ignoring available width.
1593 * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
1594 * starting from "XXX" should fit in graph_width.
1595 */
1596 graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
1597 if (options->stat_graph_width &&
1598 options->stat_graph_width < graph_width)
1599 graph_width = options->stat_graph_width;
1600
1601 name_width = (options->stat_name_width > 0 &&
1602 options->stat_name_width < max_len) ?
1603 options->stat_name_width : max_len;
1604
1605 /*
1606 * Adjust adjustable widths not to exceed maximum width
1607 */
1608 if (name_width + number_width + 6 + graph_width > width) {
1609 if (graph_width > width * 3/8 - number_width - 6) {
1610 graph_width = width * 3/8 - number_width - 6;
1611 if (graph_width < 6)
1612 graph_width = 6;
1613 }
1614
1615 if (options->stat_graph_width &&
1616 graph_width > options->stat_graph_width)
1617 graph_width = options->stat_graph_width;
1618 if (name_width > width - number_width - 6 - graph_width)
1619 name_width = width - number_width - 6 - graph_width;
1620 else
1621 graph_width = width - number_width - 6 - name_width;
1622 }
1623
1624 /*
1625 * From here name_width is the width of the name area,
1626 * and graph_width is the width of the graph area.
1627 * max_change is used to scale graph properly.
1628 */
1629 for (i = 0; i < count; i++) {
1630 const char *prefix = "";
1631 struct diffstat_file *file = data->files[i];
1632 char *name = file->print_name;
1633 uintmax_t added = file->added;
1634 uintmax_t deleted = file->deleted;
1635 int name_len;
1636
1637 if (!file->is_interesting && (added + deleted == 0))
1638 continue;
1639
1640 /*
1641 * "scale" the filename
1642 */
1643 len = name_width;
1644 name_len = strlen(name);
1645 if (name_width < name_len) {
1646 char *slash;
1647 prefix = "...";
1648 len -= 3;
1649 name += name_len - len;
1650 slash = strchr(name, '/');
1651 if (slash)
1652 name = slash;
1653 }
1654
1655 if (file->is_binary) {
1656 fprintf(options->file, "%s", line_prefix);
1657 show_name(options->file, prefix, name, len);
1658 fprintf(options->file, " %*s", number_width, "Bin");
1659 if (!added && !deleted) {
1660 putc('\n', options->file);
1661 continue;
1662 }
1663 fprintf(options->file, " %s%"PRIuMAX"%s",
1664 del_c, deleted, reset);
1665 fprintf(options->file, " -> ");
1666 fprintf(options->file, "%s%"PRIuMAX"%s",
1667 add_c, added, reset);
1668 fprintf(options->file, " bytes");
1669 fprintf(options->file, "\n");
1670 continue;
1671 }
1672 else if (file->is_unmerged) {
1673 fprintf(options->file, "%s", line_prefix);
1674 show_name(options->file, prefix, name, len);
1675 fprintf(options->file, " Unmerged\n");
1676 continue;
1677 }
1678
1679 /*
1680 * scale the add/delete
1681 */
1682 add = added;
1683 del = deleted;
1684
1685 if (graph_width <= max_change) {
1686 int total = scale_linear(add + del, graph_width, max_change);
1687 if (total < 2 && add && del)
1688 /* width >= 2 due to the sanity check */
1689 total = 2;
1690 if (add < del) {
1691 add = scale_linear(add, graph_width, max_change);
1692 del = total - add;
1693 } else {
1694 del = scale_linear(del, graph_width, max_change);
1695 add = total - del;
1696 }
1697 }
1698 fprintf(options->file, "%s", line_prefix);
1699 show_name(options->file, prefix, name, len);
1700 fprintf(options->file, " %*"PRIuMAX"%s",
1701 number_width, added + deleted,
1702 added + deleted ? " " : "");
1703 show_graph(options->file, '+', add, add_c, reset);
1704 show_graph(options->file, '-', del, del_c, reset);
1705 fprintf(options->file, "\n");
1706 }
1707
1708 for (i = 0; i < data->nr; i++) {
1709 struct diffstat_file *file = data->files[i];
1710 uintmax_t added = file->added;
1711 uintmax_t deleted = file->deleted;
1712
1713 if (file->is_unmerged ||
1714 (!file->is_interesting && (added + deleted == 0))) {
1715 total_files--;
1716 continue;
1717 }
1718
1719 if (!file->is_binary) {
1720 adds += added;
1721 dels += deleted;
1722 }
1723 if (i < count)
1724 continue;
1725 if (!extra_shown)
1726 fprintf(options->file, "%s ...\n", line_prefix);
1727 extra_shown = 1;
1728 }
1729 fprintf(options->file, "%s", line_prefix);
1730 print_stat_summary(options->file, total_files, adds, dels);
1731 }
1732
1733 static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1734 {
1735 int i, adds = 0, dels = 0, total_files = data->nr;
1736
1737 if (data->nr == 0)
1738 return;
1739
1740 for (i = 0; i < data->nr; i++) {
1741 int added = data->files[i]->added;
1742 int deleted= data->files[i]->deleted;
1743
1744 if (data->files[i]->is_unmerged ||
1745 (!data->files[i]->is_interesting && (added + deleted == 0))) {
1746 total_files--;
1747 } else if (!data->files[i]->is_binary) { /* don't count bytes */
1748 adds += added;
1749 dels += deleted;
1750 }
1751 }
1752 fprintf(options->file, "%s", diff_line_prefix(options));
1753 print_stat_summary(options->file, total_files, adds, dels);
1754 }
1755
1756 static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1757 {
1758 int i;
1759
1760 if (data->nr == 0)
1761 return;
1762
1763 for (i = 0; i < data->nr; i++) {
1764 struct diffstat_file *file = data->files[i];
1765
1766 fprintf(options->file, "%s", diff_line_prefix(options));
1767
1768 if (file->is_binary)
1769 fprintf(options->file, "-\t-\t");
1770 else
1771 fprintf(options->file,
1772 "%"PRIuMAX"\t%"PRIuMAX"\t",
1773 file->added, file->deleted);
1774 if (options->line_termination) {
1775 fill_print_name(file);
1776 if (!file->is_renamed)
1777 write_name_quoted(file->name, options->file,
1778 options->line_termination);
1779 else {
1780 fputs(file->print_name, options->file);
1781 putc(options->line_termination, options->file);
1782 }
1783 } else {
1784 if (file->is_renamed) {
1785 putc('\0', options->file);
1786 write_name_quoted(file->from_name, options->file, '\0');
1787 }
1788 write_name_quoted(file->name, options->file, '\0');
1789 }
1790 }
1791 }
1792
1793 struct dirstat_file {
1794 const char *name;
1795 unsigned long changed;
1796 };
1797
1798 struct dirstat_dir {
1799 struct dirstat_file *files;
1800 int alloc, nr, permille, cumulative;
1801 };
1802
1803 static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1804 unsigned long changed, const char *base, int baselen)
1805 {
1806 unsigned long this_dir = 0;
1807 unsigned int sources = 0;
1808 const char *line_prefix = diff_line_prefix(opt);
1809
1810 while (dir->nr) {
1811 struct dirstat_file *f = dir->files;
1812 int namelen = strlen(f->name);
1813 unsigned long this;
1814 char *slash;
1815
1816 if (namelen < baselen)
1817 break;
1818 if (memcmp(f->name, base, baselen))
1819 break;
1820 slash = strchr(f->name + baselen, '/');
1821 if (slash) {
1822 int newbaselen = slash + 1 - f->name;
1823 this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
1824 sources++;
1825 } else {
1826 this = f->changed;
1827 dir->files++;
1828 dir->nr--;
1829 sources += 2;
1830 }
1831 this_dir += this;
1832 }
1833
1834 /*
1835 * We don't report dirstat's for
1836 * - the top level
1837 * - or cases where everything came from a single directory
1838 * under this directory (sources == 1).
1839 */
1840 if (baselen && sources != 1) {
1841 if (this_dir) {
1842 int permille = this_dir * 1000 / changed;
1843 if (permille >= dir->permille) {
1844 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
1845 permille / 10, permille % 10, baselen, base);
1846 if (!dir->cumulative)
1847 return 0;
1848 }
1849 }
1850 }
1851 return this_dir;
1852 }
1853
1854 static int dirstat_compare(const void *_a, const void *_b)
1855 {
1856 const struct dirstat_file *a = _a;
1857 const struct dirstat_file *b = _b;
1858 return strcmp(a->name, b->name);
1859 }
1860
1861 static void show_dirstat(struct diff_options *options)
1862 {
1863 int i;
1864 unsigned long changed;
1865 struct dirstat_dir dir;
1866 struct diff_queue_struct *q = &diff_queued_diff;
1867
1868 dir.files = NULL;
1869 dir.alloc = 0;
1870 dir.nr = 0;
1871 dir.permille = options->dirstat_permille;
1872 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1873
1874 changed = 0;
1875 for (i = 0; i < q->nr; i++) {
1876 struct diff_filepair *p = q->queue[i];
1877 const char *name;
1878 unsigned long copied, added, damage;
1879 int content_changed;
1880
1881 name = p->two->path ? p->two->path : p->one->path;
1882
1883 if (p->one->sha1_valid && p->two->sha1_valid)
1884 content_changed = hashcmp(p->one->sha1, p->two->sha1);
1885 else
1886 content_changed = 1;
1887
1888 if (!content_changed) {
1889 /*
1890 * The SHA1 has not changed, so pre-/post-content is
1891 * identical. We can therefore skip looking at the
1892 * file contents altogether.
1893 */
1894 damage = 0;
1895 goto found_damage;
1896 }
1897
1898 if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
1899 /*
1900 * In --dirstat-by-file mode, we don't really need to
1901 * look at the actual file contents at all.
1902 * The fact that the SHA1 changed is enough for us to
1903 * add this file to the list of results
1904 * (with each file contributing equal damage).
1905 */
1906 damage = 1;
1907 goto found_damage;
1908 }
1909
1910 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
1911 diff_populate_filespec(p->one, 0);
1912 diff_populate_filespec(p->two, 0);
1913 diffcore_count_changes(p->one, p->two, NULL, NULL, 0,
1914 &copied, &added);
1915 diff_free_filespec_data(p->one);
1916 diff_free_filespec_data(p->two);
1917 } else if (DIFF_FILE_VALID(p->one)) {
1918 diff_populate_filespec(p->one, CHECK_SIZE_ONLY);
1919 copied = added = 0;
1920 diff_free_filespec_data(p->one);
1921 } else if (DIFF_FILE_VALID(p->two)) {
1922 diff_populate_filespec(p->two, CHECK_SIZE_ONLY);
1923 copied = 0;
1924 added = p->two->size;
1925 diff_free_filespec_data(p->two);
1926 } else
1927 continue;
1928
1929 /*
1930 * Original minus copied is the removed material,
1931 * added is the new material. They are both damages
1932 * made to the preimage.
1933 * If the resulting damage is zero, we know that
1934 * diffcore_count_changes() considers the two entries to
1935 * be identical, but since content_changed is true, we
1936 * know that there must have been _some_ kind of change,
1937 * so we force all entries to have damage > 0.
1938 */
1939 damage = (p->one->size - copied) + added;
1940 if (!damage)
1941 damage = 1;
1942
1943 found_damage:
1944 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1945 dir.files[dir.nr].name = name;
1946 dir.files[dir.nr].changed = damage;
1947 changed += damage;
1948 dir.nr++;
1949 }
1950
1951 /* This can happen even with many files, if everything was renames */
1952 if (!changed)
1953 return;
1954
1955 /* Show all directories with more than x% of the changes */
1956 qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1957 gather_dirstat(options, &dir, changed, "", 0);
1958 }
1959
1960 static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
1961 {
1962 int i;
1963 unsigned long changed;
1964 struct dirstat_dir dir;
1965
1966 if (data->nr == 0)
1967 return;
1968
1969 dir.files = NULL;
1970 dir.alloc = 0;
1971 dir.nr = 0;
1972 dir.permille = options->dirstat_permille;
1973 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1974
1975 changed = 0;
1976 for (i = 0; i < data->nr; i++) {
1977 struct diffstat_file *file = data->files[i];
1978 unsigned long damage = file->added + file->deleted;
1979 if (file->is_binary)
1980 /*
1981 * binary files counts bytes, not lines. Must find some
1982 * way to normalize binary bytes vs. textual lines.
1983 * The following heuristic assumes that there are 64
1984 * bytes per "line".
1985 * This is stupid and ugly, but very cheap...
1986 */
1987 damage = (damage + 63) / 64;
1988 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1989 dir.files[dir.nr].name = file->name;
1990 dir.files[dir.nr].changed = damage;
1991 changed += damage;
1992 dir.nr++;
1993 }
1994
1995 /* This can happen even with many files, if everything was renames */
1996 if (!changed)
1997 return;
1998
1999 /* Show all directories with more than x% of the changes */
2000 qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
2001 gather_dirstat(options, &dir, changed, "", 0);
2002 }
2003
2004 static void free_diffstat_info(struct diffstat_t *diffstat)
2005 {
2006 int i;
2007 for (i = 0; i < diffstat->nr; i++) {
2008 struct diffstat_file *f = diffstat->files[i];
2009 if (f->name != f->print_name)
2010 free(f->print_name);
2011 free(f->name);
2012 free(f->from_name);
2013 free(f);
2014 }
2015 free(diffstat->files);
2016 }
2017
2018 struct checkdiff_t {
2019 const char *filename;
2020 int lineno;
2021 int conflict_marker_size;
2022 struct diff_options *o;
2023 unsigned ws_rule;
2024 unsigned status;
2025 };
2026
2027 static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2028 {
2029 char firstchar;
2030 int cnt;
2031
2032 if (len < marker_size + 1)
2033 return 0;
2034 firstchar = line[0];
2035 switch (firstchar) {
2036 case '=': case '>': case '<': case '|':
2037 break;
2038 default:
2039 return 0;
2040 }
2041 for (cnt = 1; cnt < marker_size; cnt++)
2042 if (line[cnt] != firstchar)
2043 return 0;
2044 /* line[1] thru line[marker_size-1] are same as firstchar */
2045 if (len < marker_size + 1 || !isspace(line[marker_size]))
2046 return 0;
2047 return 1;
2048 }
2049
2050 static void checkdiff_consume(void *priv, char *line, unsigned long len)
2051 {
2052 struct checkdiff_t *data = priv;
2053 int marker_size = data->conflict_marker_size;
2054 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2055 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2056 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2057 char *err;
2058 const char *line_prefix;
2059
2060 assert(data->o);
2061 line_prefix = diff_line_prefix(data->o);
2062
2063 if (line[0] == '+') {
2064 unsigned bad;
2065 data->lineno++;
2066 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2067 data->status |= 1;
2068 fprintf(data->o->file,
2069 "%s%s:%d: leftover conflict marker\n",
2070 line_prefix, data->filename, data->lineno);
2071 }
2072 bad = ws_check(line + 1, len - 1, data->ws_rule);
2073 if (!bad)
2074 return;
2075 data->status |= bad;
2076 err = whitespace_error_string(bad);
2077 fprintf(data->o->file, "%s%s:%d: %s.\n",
2078 line_prefix, data->filename, data->lineno, err);
2079 free(err);
2080 emit_line(data->o, set, reset, line, 1);
2081 ws_check_emit(line + 1, len - 1, data->ws_rule,
2082 data->o->file, set, reset, ws);
2083 } else if (line[0] == ' ') {
2084 data->lineno++;
2085 } else if (line[0] == '@') {
2086 char *plus = strchr(line, '+');
2087 if (plus)
2088 data->lineno = strtol(plus, NULL, 10) - 1;
2089 else
2090 die("invalid diff");
2091 }
2092 }
2093
2094 static unsigned char *deflate_it(char *data,
2095 unsigned long size,
2096 unsigned long *result_size)
2097 {
2098 int bound;
2099 unsigned char *deflated;
2100 git_zstream stream;
2101
2102 git_deflate_init(&stream, zlib_compression_level);
2103 bound = git_deflate_bound(&stream, size);
2104 deflated = xmalloc(bound);
2105 stream.next_out = deflated;
2106 stream.avail_out = bound;
2107
2108 stream.next_in = (unsigned char *)data;
2109 stream.avail_in = size;
2110 while (git_deflate(&stream, Z_FINISH) == Z_OK)
2111 ; /* nothing */
2112 git_deflate_end(&stream);
2113 *result_size = stream.total_out;
2114 return deflated;
2115 }
2116
2117 static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two,
2118 const char *prefix)
2119 {
2120 void *cp;
2121 void *delta;
2122 void *deflated;
2123 void *data;
2124 unsigned long orig_size;
2125 unsigned long delta_size;
2126 unsigned long deflate_size;
2127 unsigned long data_size;
2128
2129 /* We could do deflated delta, or we could do just deflated two,
2130 * whichever is smaller.
2131 */
2132 delta = NULL;
2133 deflated = deflate_it(two->ptr, two->size, &deflate_size);
2134 if (one->size && two->size) {
2135 delta = diff_delta(one->ptr, one->size,
2136 two->ptr, two->size,
2137 &delta_size, deflate_size);
2138 if (delta) {
2139 void *to_free = delta;
2140 orig_size = delta_size;
2141 delta = deflate_it(delta, delta_size, &delta_size);
2142 free(to_free);
2143 }
2144 }
2145
2146 if (delta && delta_size < deflate_size) {
2147 fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2148 free(deflated);
2149 data = delta;
2150 data_size = delta_size;
2151 }
2152 else {
2153 fprintf(file, "%sliteral %lu\n", prefix, two->size);
2154 free(delta);
2155 data = deflated;
2156 data_size = deflate_size;
2157 }
2158
2159 /* emit data encoded in base85 */
2160 cp = data;
2161 while (data_size) {
2162 int bytes = (52 < data_size) ? 52 : data_size;
2163 char line[70];
2164 data_size -= bytes;
2165 if (bytes <= 26)
2166 line[0] = bytes + 'A' - 1;
2167 else
2168 line[0] = bytes - 26 + 'a' - 1;
2169 encode_85(line + 1, cp, bytes);
2170 cp = (char *) cp + bytes;
2171 fprintf(file, "%s", prefix);
2172 fputs(line, file);
2173 fputc('\n', file);
2174 }
2175 fprintf(file, "%s\n", prefix);
2176 free(data);
2177 }
2178
2179 static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two,
2180 const char *prefix)
2181 {
2182 fprintf(file, "%sGIT binary patch\n", prefix);
2183 emit_binary_diff_body(file, one, two, prefix);
2184 emit_binary_diff_body(file, two, one, prefix);
2185 }
2186
2187 int diff_filespec_is_binary(struct diff_filespec *one)
2188 {
2189 if (one->is_binary == -1) {
2190 diff_filespec_load_driver(one);
2191 if (one->driver->binary != -1)
2192 one->is_binary = one->driver->binary;
2193 else {
2194 if (!one->data && DIFF_FILE_VALID(one))
2195 diff_populate_filespec(one, CHECK_BINARY);
2196 if (one->is_binary == -1 && one->data)
2197 one->is_binary = buffer_is_binary(one->data,
2198 one->size);
2199 if (one->is_binary == -1)
2200 one->is_binary = 0;
2201 }
2202 }
2203 return one->is_binary;
2204 }
2205
2206 static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2207 {
2208 diff_filespec_load_driver(one);
2209 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2210 }
2211
2212 void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2213 {
2214 if (!options->a_prefix)
2215 options->a_prefix = a;
2216 if (!options->b_prefix)
2217 options->b_prefix = b;
2218 }
2219
2220 struct userdiff_driver *get_textconv(struct diff_filespec *one)
2221 {
2222 if (!DIFF_FILE_VALID(one))
2223 return NULL;
2224
2225 diff_filespec_load_driver(one);
2226 return userdiff_get_textconv(one->driver);
2227 }
2228
2229 static void builtin_diff(const char *name_a,
2230 const char *name_b,
2231 struct diff_filespec *one,
2232 struct diff_filespec *two,
2233 const char *xfrm_msg,
2234 int must_show_header,
2235 struct diff_options *o,
2236 int complete_rewrite)
2237 {
2238 mmfile_t mf1, mf2;
2239 const char *lbl[2];
2240 char *a_one, *b_two;
2241 const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
2242 const char *reset = diff_get_color_opt(o, DIFF_RESET);
2243 const char *a_prefix, *b_prefix;
2244 struct userdiff_driver *textconv_one = NULL;
2245 struct userdiff_driver *textconv_two = NULL;
2246 struct strbuf header = STRBUF_INIT;
2247 const char *line_prefix = diff_line_prefix(o);
2248
2249 if (DIFF_OPT_TST(o, SUBMODULE_LOG) &&
2250 (!one->mode || S_ISGITLINK(one->mode)) &&
2251 (!two->mode || S_ISGITLINK(two->mode))) {
2252 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2253 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2254 show_submodule_summary(o->file, one->path ? one->path : two->path,
2255 line_prefix,
2256 one->sha1, two->sha1, two->dirty_submodule,
2257 meta, del, add, reset);
2258 return;
2259 }
2260
2261 if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2262 textconv_one = get_textconv(one);
2263 textconv_two = get_textconv(two);
2264 }
2265
2266 diff_set_mnemonic_prefix(o, "a/", "b/");
2267 if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2268 a_prefix = o->b_prefix;
2269 b_prefix = o->a_prefix;
2270 } else {
2271 a_prefix = o->a_prefix;
2272 b_prefix = o->b_prefix;
2273 }
2274
2275 /* Never use a non-valid filename anywhere if at all possible */
2276 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2277 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2278
2279 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2280 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2281 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2282 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2283 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
2284 if (lbl[0][0] == '/') {
2285 /* /dev/null */
2286 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
2287 if (xfrm_msg)
2288 strbuf_addstr(&header, xfrm_msg);
2289 must_show_header = 1;
2290 }
2291 else if (lbl[1][0] == '/') {
2292 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
2293 if (xfrm_msg)
2294 strbuf_addstr(&header, xfrm_msg);
2295 must_show_header = 1;
2296 }
2297 else {
2298 if (one->mode != two->mode) {
2299 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
2300 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
2301 must_show_header = 1;
2302 }
2303 if (xfrm_msg)
2304 strbuf_addstr(&header, xfrm_msg);
2305
2306 /*
2307 * we do not run diff between different kind
2308 * of objects.
2309 */
2310 if ((one->mode ^ two->mode) & S_IFMT)
2311 goto free_ab_and_return;
2312 if (complete_rewrite &&
2313 (textconv_one || !diff_filespec_is_binary(one)) &&
2314 (textconv_two || !diff_filespec_is_binary(two))) {
2315 fprintf(o->file, "%s", header.buf);
2316 strbuf_reset(&header);
2317 emit_rewrite_diff(name_a, name_b, one, two,
2318 textconv_one, textconv_two, o);
2319 o->found_changes = 1;
2320 goto free_ab_and_return;
2321 }
2322 }
2323
2324 if (o->irreversible_delete && lbl[1][0] == '/') {
2325 fprintf(o->file, "%s", header.buf);
2326 strbuf_reset(&header);
2327 goto free_ab_and_return;
2328 } else if (!DIFF_OPT_TST(o, TEXT) &&
2329 ( (!textconv_one && diff_filespec_is_binary(one)) ||
2330 (!textconv_two && diff_filespec_is_binary(two)) )) {
2331 if (!one->data && !two->data &&
2332 S_ISREG(one->mode) && S_ISREG(two->mode) &&
2333 !DIFF_OPT_TST(o, BINARY)) {
2334 if (!hashcmp(one->sha1, two->sha1)) {
2335 if (must_show_header)
2336 fprintf(o->file, "%s", header.buf);
2337 goto free_ab_and_return;
2338 }
2339 fprintf(o->file, "%s", header.buf);
2340 fprintf(o->file, "%sBinary files %s and %s differ\n",
2341 line_prefix, lbl[0], lbl[1]);
2342 goto free_ab_and_return;
2343 }
2344 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2345 die("unable to read files to diff");
2346 /* Quite common confusing case */
2347 if (mf1.size == mf2.size &&
2348 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2349 if (must_show_header)
2350 fprintf(o->file, "%s", header.buf);
2351 goto free_ab_and_return;
2352 }
2353 fprintf(o->file, "%s", header.buf);
2354 strbuf_reset(&header);
2355 if (DIFF_OPT_TST(o, BINARY))
2356 emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2357 else
2358 fprintf(o->file, "%sBinary files %s and %s differ\n",
2359 line_prefix, lbl[0], lbl[1]);
2360 o->found_changes = 1;
2361 } else {
2362 /* Crazy xdl interfaces.. */
2363 const char *diffopts = getenv("GIT_DIFF_OPTS");
2364 const char *v;
2365 xpparam_t xpp;
2366 xdemitconf_t xecfg;
2367 struct emit_callback ecbdata;
2368 const struct userdiff_funcname *pe;
2369
2370 if (must_show_header) {
2371 fprintf(o->file, "%s", header.buf);
2372 strbuf_reset(&header);
2373 }
2374
2375 mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2376 mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2377
2378 pe = diff_funcname_pattern(one);
2379 if (!pe)
2380 pe = diff_funcname_pattern(two);
2381
2382 memset(&xpp, 0, sizeof(xpp));
2383 memset(&xecfg, 0, sizeof(xecfg));
2384 memset(&ecbdata, 0, sizeof(ecbdata));
2385 ecbdata.label_path = lbl;
2386 ecbdata.color_diff = want_color(o->use_color);
2387 ecbdata.found_changesp = &o->found_changes;
2388 ecbdata.ws_rule = whitespace_rule(name_b);
2389 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2390 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2391 ecbdata.opt = o;
2392 ecbdata.header = header.len ? &header : NULL;
2393 xpp.flags = o->xdl_opts;
2394 xecfg.ctxlen = o->context;
2395 xecfg.interhunkctxlen = o->interhunkcontext;
2396 xecfg.flags = XDL_EMIT_FUNCNAMES;
2397 if (DIFF_OPT_TST(o, FUNCCONTEXT))
2398 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2399 if (pe)
2400 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2401 if (!diffopts)
2402 ;
2403 else if (skip_prefix(diffopts, "--unified=", &v))
2404 xecfg.ctxlen = strtoul(v, NULL, 10);
2405 else if (skip_prefix(diffopts, "-u", &v))
2406 xecfg.ctxlen = strtoul(v, NULL, 10);
2407 if (o->word_diff)
2408 init_diff_words_data(&ecbdata, o, one, two);
2409 if (xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2410 &xpp, &xecfg))
2411 die("unable to generate diff for %s", one->path);
2412 if (o->word_diff)
2413 free_diff_words_data(&ecbdata);
2414 if (textconv_one)
2415 free(mf1.ptr);
2416 if (textconv_two)
2417 free(mf2.ptr);
2418 xdiff_clear_find_func(&xecfg);
2419 }
2420
2421 free_ab_and_return:
2422 strbuf_release(&header);
2423 diff_free_filespec_data(one);
2424 diff_free_filespec_data(two);
2425 free(a_one);
2426 free(b_two);
2427 return;
2428 }
2429
2430 static void builtin_diffstat(const char *name_a, const char *name_b,
2431 struct diff_filespec *one,
2432 struct diff_filespec *two,
2433 struct diffstat_t *diffstat,
2434 struct diff_options *o,
2435 struct diff_filepair *p)
2436 {
2437 mmfile_t mf1, mf2;
2438 struct diffstat_file *data;
2439 int same_contents;
2440 int complete_rewrite = 0;
2441
2442 if (!DIFF_PAIR_UNMERGED(p)) {
2443 if (p->status == DIFF_STATUS_MODIFIED && p->score)
2444 complete_rewrite = 1;
2445 }
2446
2447 data = diffstat_add(diffstat, name_a, name_b);
2448 data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
2449
2450 if (!one || !two) {
2451 data->is_unmerged = 1;
2452 return;
2453 }
2454
2455 same_contents = !hashcmp(one->sha1, two->sha1);
2456
2457 if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2458 data->is_binary = 1;
2459 if (same_contents) {
2460 data->added = 0;
2461 data->deleted = 0;
2462 } else {
2463 data->added = diff_filespec_size(two);
2464 data->deleted = diff_filespec_size(one);
2465 }
2466 }
2467
2468 else if (complete_rewrite) {
2469 diff_populate_filespec(one, 0);
2470 diff_populate_filespec(two, 0);
2471 data->deleted = count_lines(one->data, one->size);
2472 data->added = count_lines(two->data, two->size);
2473 }
2474
2475 else if (!same_contents) {
2476 /* Crazy xdl interfaces.. */
2477 xpparam_t xpp;
2478 xdemitconf_t xecfg;
2479
2480 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2481 die("unable to read files to diff");
2482
2483 memset(&xpp, 0, sizeof(xpp));
2484 memset(&xecfg, 0, sizeof(xecfg));
2485 xpp.flags = o->xdl_opts;
2486 xecfg.ctxlen = o->context;
2487 xecfg.interhunkctxlen = o->interhunkcontext;
2488 if (xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2489 &xpp, &xecfg))
2490 die("unable to generate diffstat for %s", one->path);
2491 }
2492
2493 diff_free_filespec_data(one);
2494 diff_free_filespec_data(two);
2495 }
2496
2497 static void builtin_checkdiff(const char *name_a, const char *name_b,
2498 const char *attr_path,
2499 struct diff_filespec *one,
2500 struct diff_filespec *two,
2501 struct diff_options *o)
2502 {
2503 mmfile_t mf1, mf2;
2504 struct checkdiff_t data;
2505
2506 if (!two)
2507 return;
2508
2509 memset(&data, 0, sizeof(data));
2510 data.filename = name_b ? name_b : name_a;
2511 data.lineno = 0;
2512 data.o = o;
2513 data.ws_rule = whitespace_rule(attr_path);
2514 data.conflict_marker_size = ll_merge_marker_size(attr_path);
2515
2516 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2517 die("unable to read files to diff");
2518
2519 /*
2520 * All the other codepaths check both sides, but not checking
2521 * the "old" side here is deliberate. We are checking the newly
2522 * introduced changes, and as long as the "new" side is text, we
2523 * can and should check what it introduces.
2524 */
2525 if (diff_filespec_is_binary(two))
2526 goto free_and_return;
2527 else {
2528 /* Crazy xdl interfaces.. */
2529 xpparam_t xpp;
2530 xdemitconf_t xecfg;
2531
2532 memset(&xpp, 0, sizeof(xpp));
2533 memset(&xecfg, 0, sizeof(xecfg));
2534 xecfg.ctxlen = 1; /* at least one context line */
2535 xpp.flags = 0;
2536 if (xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2537 &xpp, &xecfg))
2538 die("unable to generate checkdiff for %s", one->path);
2539
2540 if (data.ws_rule & WS_BLANK_AT_EOF) {
2541 struct emit_callback ecbdata;
2542 int blank_at_eof;
2543
2544 ecbdata.ws_rule = data.ws_rule;
2545 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2546 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2547
2548 if (blank_at_eof) {
2549 static char *err;
2550 if (!err)
2551 err = whitespace_error_string(WS_BLANK_AT_EOF);
2552 fprintf(o->file, "%s:%d: %s.\n",
2553 data.filename, blank_at_eof, err);
2554 data.status = 1; /* report errors */
2555 }
2556 }
2557 }
2558 free_and_return:
2559 diff_free_filespec_data(one);
2560 diff_free_filespec_data(two);
2561 if (data.status)
2562 DIFF_OPT_SET(o, CHECK_FAILED);
2563 }
2564
2565 struct diff_filespec *alloc_filespec(const char *path)
2566 {
2567 int namelen = strlen(path);
2568 struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
2569
2570 memset(spec, 0, sizeof(*spec));
2571 spec->path = (char *)(spec + 1);
2572 memcpy(spec->path, path, namelen+1);
2573 spec->count = 1;
2574 spec->is_binary = -1;
2575 return spec;
2576 }
2577
2578 void free_filespec(struct diff_filespec *spec)
2579 {
2580 if (!--spec->count) {
2581 diff_free_filespec_data(spec);
2582 free(spec);
2583 }
2584 }
2585
2586 void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
2587 int sha1_valid, unsigned short mode)
2588 {
2589 if (mode) {
2590 spec->mode = canon_mode(mode);
2591 hashcpy(spec->sha1, sha1);
2592 spec->sha1_valid = sha1_valid;
2593 }
2594 }
2595
2596 /*
2597 * Given a name and sha1 pair, if the index tells us the file in
2598 * the work tree has that object contents, return true, so that
2599 * prepare_temp_file() does not have to inflate and extract.
2600 */
2601 static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
2602 {
2603 const struct cache_entry *ce;
2604 struct stat st;
2605 int pos, len;
2606
2607 /*
2608 * We do not read the cache ourselves here, because the
2609 * benchmark with my previous version that always reads cache
2610 * shows that it makes things worse for diff-tree comparing
2611 * two linux-2.6 kernel trees in an already checked out work
2612 * tree. This is because most diff-tree comparisons deal with
2613 * only a small number of files, while reading the cache is
2614 * expensive for a large project, and its cost outweighs the
2615 * savings we get by not inflating the object to a temporary
2616 * file. Practically, this code only helps when we are used
2617 * by diff-cache --cached, which does read the cache before
2618 * calling us.
2619 */
2620 if (!active_cache)
2621 return 0;
2622
2623 /* We want to avoid the working directory if our caller
2624 * doesn't need the data in a normal file, this system
2625 * is rather slow with its stat/open/mmap/close syscalls,
2626 * and the object is contained in a pack file. The pack
2627 * is probably already open and will be faster to obtain
2628 * the data through than the working directory. Loose
2629 * objects however would tend to be slower as they need
2630 * to be individually opened and inflated.
2631 */
2632 if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1))
2633 return 0;
2634
2635 len = strlen(name);
2636 pos = cache_name_pos(name, len);
2637 if (pos < 0)
2638 return 0;
2639 ce = active_cache[pos];
2640
2641 /*
2642 * This is not the sha1 we are looking for, or
2643 * unreusable because it is not a regular file.
2644 */
2645 if (hashcmp(sha1, ce->sha1) || !S_ISREG(ce->ce_mode))
2646 return 0;
2647
2648 /*
2649 * If ce is marked as "assume unchanged", there is no
2650 * guarantee that work tree matches what we are looking for.
2651 */
2652 if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2653 return 0;
2654
2655 /*
2656 * If ce matches the file in the work tree, we can reuse it.
2657 */
2658 if (ce_uptodate(ce) ||
2659 (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2660 return 1;
2661
2662 return 0;
2663 }
2664
2665 static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2666 {
2667 int len;
2668 char *data = xmalloc(100), *dirty = "";
2669
2670 /* Are we looking at the work tree? */
2671 if (s->dirty_submodule)
2672 dirty = "-dirty";
2673
2674 len = snprintf(data, 100,
2675 "Subproject commit %s%s\n", sha1_to_hex(s->sha1), dirty);
2676 s->data = data;
2677 s->size = len;
2678 s->should_free = 1;
2679 if (size_only) {
2680 s->data = NULL;
2681 free(data);
2682 }
2683 return 0;
2684 }
2685
2686 /*
2687 * While doing rename detection and pickaxe operation, we may need to
2688 * grab the data for the blob (or file) for our own in-core comparison.
2689 * diff_filespec has data and size fields for this purpose.
2690 */
2691 int diff_populate_filespec(struct diff_filespec *s, unsigned int flags)
2692 {
2693 int size_only = flags & CHECK_SIZE_ONLY;
2694 int err = 0;
2695 /*
2696 * demote FAIL to WARN to allow inspecting the situation
2697 * instead of refusing.
2698 */
2699 enum safe_crlf crlf_warn = (safe_crlf == SAFE_CRLF_FAIL
2700 ? SAFE_CRLF_WARN
2701 : safe_crlf);
2702
2703 if (!DIFF_FILE_VALID(s))
2704 die("internal error: asking to populate invalid file.");
2705 if (S_ISDIR(s->mode))
2706 return -1;
2707
2708 if (s->data)
2709 return 0;
2710
2711 if (size_only && 0 < s->size)
2712 return 0;
2713
2714 if (S_ISGITLINK(s->mode))
2715 return diff_populate_gitlink(s, size_only);
2716
2717 if (!s->sha1_valid ||
2718 reuse_worktree_file(s->path, s->sha1, 0)) {
2719 struct strbuf buf = STRBUF_INIT;
2720 struct stat st;
2721 int fd;
2722
2723 if (lstat(s->path, &st) < 0) {
2724 if (errno == ENOENT) {
2725 err_empty:
2726 err = -1;
2727 empty:
2728 s->data = (char *)"";
2729 s->size = 0;
2730 return err;
2731 }
2732 }
2733 s->size = xsize_t(st.st_size);
2734 if (!s->size)
2735 goto empty;
2736 if (S_ISLNK(st.st_mode)) {
2737 struct strbuf sb = STRBUF_INIT;
2738
2739 if (strbuf_readlink(&sb, s->path, s->size))
2740 goto err_empty;
2741 s->size = sb.len;
2742 s->data = strbuf_detach(&sb, NULL);
2743 s->should_free = 1;
2744 return 0;
2745 }
2746 if (size_only)
2747 return 0;
2748 if ((flags & CHECK_BINARY) &&
2749 s->size > big_file_threshold && s->is_binary == -1) {
2750 s->is_binary = 1;
2751 return 0;
2752 }
2753 fd = open(s->path, O_RDONLY);
2754 if (fd < 0)
2755 goto err_empty;
2756 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2757 close(fd);
2758 s->should_munmap = 1;
2759
2760 /*
2761 * Convert from working tree format to canonical git format
2762 */
2763 if (convert_to_git(s->path, s->data, s->size, &buf, crlf_warn)) {
2764 size_t size = 0;
2765 munmap(s->data, s->size);
2766 s->should_munmap = 0;
2767 s->data = strbuf_detach(&buf, &size);
2768 s->size = size;
2769 s->should_free = 1;
2770 }
2771 }
2772 else {
2773 enum object_type type;
2774 if (size_only || (flags & CHECK_BINARY)) {
2775 type = sha1_object_info(s->sha1, &s->size);
2776 if (type < 0)
2777 die("unable to read %s", sha1_to_hex(s->sha1));
2778 if (size_only)
2779 return 0;
2780 if (s->size > big_file_threshold && s->is_binary == -1) {
2781 s->is_binary = 1;
2782 return 0;
2783 }
2784 }
2785 s->data = read_sha1_file(s->sha1, &type, &s->size);
2786 if (!s->data)
2787 die("unable to read %s", sha1_to_hex(s->sha1));
2788 s->should_free = 1;
2789 }
2790 return 0;
2791 }
2792
2793 void diff_free_filespec_blob(struct diff_filespec *s)
2794 {
2795 if (s->should_free)
2796 free(s->data);
2797 else if (s->should_munmap)
2798 munmap(s->data, s->size);
2799
2800 if (s->should_free || s->should_munmap) {
2801 s->should_free = s->should_munmap = 0;
2802 s->data = NULL;
2803 }
2804 }
2805
2806 void diff_free_filespec_data(struct diff_filespec *s)
2807 {
2808 diff_free_filespec_blob(s);
2809 free(s->cnt_data);
2810 s->cnt_data = NULL;
2811 }
2812
2813 static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2814 void *blob,
2815 unsigned long size,
2816 const unsigned char *sha1,
2817 int mode)
2818 {
2819 int fd;
2820 struct strbuf buf = STRBUF_INIT;
2821 struct strbuf template = STRBUF_INIT;
2822 char *path_dup = xstrdup(path);
2823 const char *base = basename(path_dup);
2824
2825 /* Generate "XXXXXX_basename.ext" */
2826 strbuf_addstr(&template, "XXXXXX_");
2827 strbuf_addstr(&template, base);
2828
2829 fd = git_mkstemps(temp->tmp_path, PATH_MAX, template.buf,
2830 strlen(base) + 1);
2831 if (fd < 0)
2832 die_errno("unable to create temp-file");
2833 if (convert_to_working_tree(path,
2834 (const char *)blob, (size_t)size, &buf)) {
2835 blob = buf.buf;
2836 size = buf.len;
2837 }
2838 if (write_in_full(fd, blob, size) != size)
2839 die_errno("unable to write temp-file");
2840 close(fd);
2841 temp->name = temp->tmp_path;
2842 strcpy(temp->hex, sha1_to_hex(sha1));
2843 temp->hex[40] = 0;
2844 sprintf(temp->mode, "%06o", mode);
2845 strbuf_release(&buf);
2846 strbuf_release(&template);
2847 free(path_dup);
2848 }
2849
2850 static struct diff_tempfile *prepare_temp_file(const char *name,
2851 struct diff_filespec *one)
2852 {
2853 struct diff_tempfile *temp = claim_diff_tempfile();
2854
2855 if (!DIFF_FILE_VALID(one)) {
2856 not_a_valid_file:
2857 /* A '-' entry produces this for file-2, and
2858 * a '+' entry produces this for file-1.
2859 */
2860 temp->name = "/dev/null";
2861 strcpy(temp->hex, ".");
2862 strcpy(temp->mode, ".");
2863 return temp;
2864 }
2865
2866 if (!remove_tempfile_installed) {
2867 atexit(remove_tempfile);
2868 sigchain_push_common(remove_tempfile_on_signal);
2869 remove_tempfile_installed = 1;
2870 }
2871
2872 if (!S_ISGITLINK(one->mode) &&
2873 (!one->sha1_valid ||
2874 reuse_worktree_file(name, one->sha1, 1))) {
2875 struct stat st;
2876 if (lstat(name, &st) < 0) {
2877 if (errno == ENOENT)
2878 goto not_a_valid_file;
2879 die_errno("stat(%s)", name);
2880 }
2881 if (S_ISLNK(st.st_mode)) {
2882 struct strbuf sb = STRBUF_INIT;
2883 if (strbuf_readlink(&sb, name, st.st_size) < 0)
2884 die_errno("readlink(%s)", name);
2885 prep_temp_blob(name, temp, sb.buf, sb.len,
2886 (one->sha1_valid ?
2887 one->sha1 : null_sha1),
2888 (one->sha1_valid ?
2889 one->mode : S_IFLNK));
2890 strbuf_release(&sb);
2891 }
2892 else {
2893 /* we can borrow from the file in the work tree */
2894 temp->name = name;
2895 if (!one->sha1_valid)
2896 strcpy(temp->hex, sha1_to_hex(null_sha1));
2897 else
2898 strcpy(temp->hex, sha1_to_hex(one->sha1));
2899 /* Even though we may sometimes borrow the
2900 * contents from the work tree, we always want
2901 * one->mode. mode is trustworthy even when
2902 * !(one->sha1_valid), as long as
2903 * DIFF_FILE_VALID(one).
2904 */
2905 sprintf(temp->mode, "%06o", one->mode);
2906 }
2907 return temp;
2908 }
2909 else {
2910 if (diff_populate_filespec(one, 0))
2911 die("cannot read data blob for %s", one->path);
2912 prep_temp_blob(name, temp, one->data, one->size,
2913 one->sha1, one->mode);
2914 }
2915 return temp;
2916 }
2917
2918 static void add_external_diff_name(struct argv_array *argv,
2919 const char *name,
2920 struct diff_filespec *df)
2921 {
2922 struct diff_tempfile *temp = prepare_temp_file(name, df);
2923 argv_array_push(argv, temp->name);
2924 argv_array_push(argv, temp->hex);
2925 argv_array_push(argv, temp->mode);
2926 }
2927
2928 /* An external diff command takes:
2929 *
2930 * diff-cmd name infile1 infile1-sha1 infile1-mode \
2931 * infile2 infile2-sha1 infile2-mode [ rename-to ]
2932 *
2933 */
2934 static void run_external_diff(const char *pgm,
2935 const char *name,
2936 const char *other,
2937 struct diff_filespec *one,
2938 struct diff_filespec *two,
2939 const char *xfrm_msg,
2940 int complete_rewrite,
2941 struct diff_options *o)
2942 {
2943 struct argv_array argv = ARGV_ARRAY_INIT;
2944 struct argv_array env = ARGV_ARRAY_INIT;
2945 struct diff_queue_struct *q = &diff_queued_diff;
2946
2947 argv_array_push(&argv, pgm);
2948 argv_array_push(&argv, name);
2949
2950 if (one && two) {
2951 add_external_diff_name(&argv, name, one);
2952 if (!other)
2953 add_external_diff_name(&argv, name, two);
2954 else {
2955 add_external_diff_name(&argv, other, two);
2956 argv_array_push(&argv, other);
2957 argv_array_push(&argv, xfrm_msg);
2958 }
2959 }
2960
2961 argv_array_pushf(&env, "GIT_DIFF_PATH_COUNTER=%d", ++o->diff_path_counter);
2962 argv_array_pushf(&env, "GIT_DIFF_PATH_TOTAL=%d", q->nr);
2963
2964 if (run_command_v_opt_cd_env(argv.argv, RUN_USING_SHELL, NULL, env.argv))
2965 die(_("external diff died, stopping at %s"), name);
2966
2967 remove_tempfile();
2968 argv_array_clear(&argv);
2969 argv_array_clear(&env);
2970 }
2971
2972 static int similarity_index(struct diff_filepair *p)
2973 {
2974 return p->score * 100 / MAX_SCORE;
2975 }
2976
2977 static void fill_metainfo(struct strbuf *msg,
2978 const char *name,
2979 const char *other,
2980 struct diff_filespec *one,
2981 struct diff_filespec *two,
2982 struct diff_options *o,
2983 struct diff_filepair *p,
2984 int *must_show_header,
2985 int use_color)
2986 {
2987 const char *set = diff_get_color(use_color, DIFF_METAINFO);
2988 const char *reset = diff_get_color(use_color, DIFF_RESET);
2989 const char *line_prefix = diff_line_prefix(o);
2990
2991 *must_show_header = 1;
2992 strbuf_init(msg, PATH_MAX * 2 + 300);
2993 switch (p->status) {
2994 case DIFF_STATUS_COPIED:
2995 strbuf_addf(msg, "%s%ssimilarity index %d%%",
2996 line_prefix, set, similarity_index(p));
2997 strbuf_addf(msg, "%s\n%s%scopy from ",
2998 reset, line_prefix, set);
2999 quote_c_style(name, msg, NULL, 0);
3000 strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
3001 quote_c_style(other, msg, NULL, 0);
3002 strbuf_addf(msg, "%s\n", reset);
3003 break;
3004 case DIFF_STATUS_RENAMED:
3005 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3006 line_prefix, set, similarity_index(p));
3007 strbuf_addf(msg, "%s\n%s%srename from ",
3008 reset, line_prefix, set);
3009 quote_c_style(name, msg, NULL, 0);
3010 strbuf_addf(msg, "%s\n%s%srename to ",
3011 reset, line_prefix, set);
3012 quote_c_style(other, msg, NULL, 0);
3013 strbuf_addf(msg, "%s\n", reset);
3014 break;
3015 case DIFF_STATUS_MODIFIED:
3016 if (p->score) {
3017 strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
3018 line_prefix,
3019 set, similarity_index(p), reset);
3020 break;
3021 }
3022 /* fallthru */
3023 default:
3024 *must_show_header = 0;
3025 }
3026 if (one && two && hashcmp(one->sha1, two->sha1)) {
3027 int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
3028
3029 if (DIFF_OPT_TST(o, BINARY)) {
3030 mmfile_t mf;
3031 if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
3032 (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
3033 abbrev = 40;
3034 }
3035 strbuf_addf(msg, "%s%sindex %s..", line_prefix, set,
3036 find_unique_abbrev(one->sha1, abbrev));
3037 strbuf_addstr(msg, find_unique_abbrev(two->sha1, abbrev));
3038 if (one->mode == two->mode)
3039 strbuf_addf(msg, " %06o", one->mode);
3040 strbuf_addf(msg, "%s\n", reset);
3041 }
3042 }
3043
3044 static void run_diff_cmd(const char *pgm,
3045 const char *name,
3046 const char *other,
3047 const char *attr_path,
3048 struct diff_filespec *one,
3049 struct diff_filespec *two,
3050 struct strbuf *msg,
3051 struct diff_options *o,
3052 struct diff_filepair *p)
3053 {
3054 const char *xfrm_msg = NULL;
3055 int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
3056 int must_show_header = 0;
3057
3058
3059 if (DIFF_OPT_TST(o, ALLOW_EXTERNAL)) {
3060 struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
3061 if (drv && drv->external)
3062 pgm = drv->external;
3063 }
3064
3065 if (msg) {
3066 /*
3067 * don't use colors when the header is intended for an
3068 * external diff driver
3069 */
3070 fill_metainfo(msg, name, other, one, two, o, p,
3071 &must_show_header,
3072 want_color(o->use_color) && !pgm);
3073 xfrm_msg = msg->len ? msg->buf : NULL;
3074 }
3075
3076 if (pgm) {
3077 run_external_diff(pgm, name, other, one, two, xfrm_msg,
3078 complete_rewrite, o);
3079 return;
3080 }
3081 if (one && two)
3082 builtin_diff(name, other ? other : name,
3083 one, two, xfrm_msg, must_show_header,
3084 o, complete_rewrite);
3085 else
3086 fprintf(o->file, "* Unmerged path %s\n", name);
3087 }
3088
3089 static void diff_fill_sha1_info(struct diff_filespec *one)
3090 {
3091 if (DIFF_FILE_VALID(one)) {
3092 if (!one->sha1_valid) {
3093 struct stat st;
3094 if (one->is_stdin) {
3095 hashcpy(one->sha1, null_sha1);
3096 return;
3097 }
3098 if (lstat(one->path, &st) < 0)
3099 die_errno("stat '%s'", one->path);
3100 if (index_path(one->sha1, one->path, &st, 0))
3101 die("cannot hash %s", one->path);
3102 }
3103 }
3104 else
3105 hashclr(one->sha1);
3106 }
3107
3108 static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
3109 {
3110 /* Strip the prefix but do not molest /dev/null and absolute paths */
3111 if (*namep && **namep != '/') {
3112 *namep += prefix_length;
3113 if (**namep == '/')
3114 ++*namep;
3115 }
3116 if (*otherp && **otherp != '/') {
3117 *otherp += prefix_length;
3118 if (**otherp == '/')
3119 ++*otherp;
3120 }
3121 }
3122
3123 static void run_diff(struct diff_filepair *p, struct diff_options *o)
3124 {
3125 const char *pgm = external_diff();
3126 struct strbuf msg;
3127 struct diff_filespec *one = p->one;
3128 struct diff_filespec *two = p->two;
3129 const char *name;
3130 const char *other;
3131 const char *attr_path;
3132
3133 name = p->one->path;
3134 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3135 attr_path = name;
3136 if (o->prefix_length)
3137 strip_prefix(o->prefix_length, &name, &other);
3138
3139 if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
3140 pgm = NULL;
3141
3142 if (DIFF_PAIR_UNMERGED(p)) {
3143 run_diff_cmd(pgm, name, NULL, attr_path,
3144 NULL, NULL, NULL, o, p);
3145 return;
3146 }
3147
3148 diff_fill_sha1_info(one);
3149 diff_fill_sha1_info(two);
3150
3151 if (!pgm &&
3152 DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3153 (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3154 /*
3155 * a filepair that changes between file and symlink
3156 * needs to be split into deletion and creation.
3157 */
3158 struct diff_filespec *null = alloc_filespec(two->path);
3159 run_diff_cmd(NULL, name, other, attr_path,
3160 one, null, &msg, o, p);
3161 free(null);
3162 strbuf_release(&msg);
3163
3164 null = alloc_filespec(one->path);
3165 run_diff_cmd(NULL, name, other, attr_path,
3166 null, two, &msg, o, p);
3167 free(null);
3168 }
3169 else
3170 run_diff_cmd(pgm, name, other, attr_path,
3171 one, two, &msg, o, p);
3172
3173 strbuf_release(&msg);
3174 }
3175
3176 static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3177 struct diffstat_t *diffstat)
3178 {
3179 const char *name;
3180 const char *other;
3181
3182 if (DIFF_PAIR_UNMERGED(p)) {
3183 /* unmerged */
3184 builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, p);
3185 return;
3186 }
3187
3188 name = p->one->path;
3189 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3190
3191 if (o->prefix_length)
3192 strip_prefix(o->prefix_length, &name, &other);
3193
3194 diff_fill_sha1_info(p->one);
3195 diff_fill_sha1_info(p->two);
3196
3197 builtin_diffstat(name, other, p->one, p->two, diffstat, o, p);
3198 }
3199
3200 static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3201 {
3202 const char *name;
3203 const char *other;
3204 const char *attr_path;
3205
3206 if (DIFF_PAIR_UNMERGED(p)) {
3207 /* unmerged */
3208 return;
3209 }
3210
3211 name = p->one->path;
3212 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3213 attr_path = other ? other : name;
3214
3215 if (o->prefix_length)
3216 strip_prefix(o->prefix_length, &name, &other);
3217
3218 diff_fill_sha1_info(p->one);
3219 diff_fill_sha1_info(p->two);
3220
3221 builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3222 }
3223
3224 void diff_setup(struct diff_options *options)
3225 {
3226 memcpy(options, &default_diff_options, sizeof(*options));
3227
3228 options->file = stdout;
3229
3230 options->line_termination = '\n';
3231 options->break_opt = -1;
3232 options->rename_limit = -1;
3233 options->dirstat_permille = diff_dirstat_permille_default;
3234 options->context = diff_context_default;
3235 DIFF_OPT_SET(options, RENAME_EMPTY);
3236
3237 /* pathchange left =NULL by default */
3238 options->change = diff_change;
3239 options->add_remove = diff_addremove;
3240 options->use_color = diff_use_color_default;
3241 options->detect_rename = diff_detect_rename_default;
3242 options->xdl_opts |= diff_algorithm;
3243 if (diff_compaction_heuristic)
3244 DIFF_XDL_SET(options, COMPACTION_HEURISTIC);
3245
3246 options->orderfile = diff_order_file_cfg;
3247
3248 if (diff_no_prefix) {
3249 options->a_prefix = options->b_prefix = "";
3250 } else if (!diff_mnemonic_prefix) {
3251 options->a_prefix = "a/";
3252 options->b_prefix = "b/";
3253 }
3254 }
3255
3256 void diff_setup_done(struct diff_options *options)
3257 {
3258 int count = 0;
3259
3260 if (options->set_default)
3261 options->set_default(options);
3262
3263 if (options->output_format & DIFF_FORMAT_NAME)
3264 count++;
3265 if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3266 count++;
3267 if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3268 count++;
3269 if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3270 count++;
3271 if (count > 1)
3272 die("--name-only, --name-status, --check and -s are mutually exclusive");
3273
3274 /*
3275 * Most of the time we can say "there are changes"
3276 * only by checking if there are changed paths, but
3277 * --ignore-whitespace* options force us to look
3278 * inside contents.
3279 */
3280
3281 if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3282 DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3283 DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3284 DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3285 else
3286 DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3287
3288 if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3289 options->detect_rename = DIFF_DETECT_COPY;
3290
3291 if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3292 options->prefix = NULL;
3293 if (options->prefix)
3294 options->prefix_length = strlen(options->prefix);
3295 else
3296 options->prefix_length = 0;
3297
3298 if (options->output_format & (DIFF_FORMAT_NAME |
3299 DIFF_FORMAT_NAME_STATUS |
3300 DIFF_FORMAT_CHECKDIFF |
3301 DIFF_FORMAT_NO_OUTPUT))
3302 options->output_format &= ~(DIFF_FORMAT_RAW |
3303 DIFF_FORMAT_NUMSTAT |
3304 DIFF_FORMAT_DIFFSTAT |
3305 DIFF_FORMAT_SHORTSTAT |
3306 DIFF_FORMAT_DIRSTAT |
3307 DIFF_FORMAT_SUMMARY |
3308 DIFF_FORMAT_PATCH);
3309
3310 /*
3311 * These cases always need recursive; we do not drop caller-supplied
3312 * recursive bits for other formats here.
3313 */
3314 if (options->output_format & (DIFF_FORMAT_PATCH |
3315 DIFF_FORMAT_NUMSTAT |
3316 DIFF_FORMAT_DIFFSTAT |
3317 DIFF_FORMAT_SHORTSTAT |
3318 DIFF_FORMAT_DIRSTAT |
3319 DIFF_FORMAT_SUMMARY |
3320 DIFF_FORMAT_CHECKDIFF))
3321 DIFF_OPT_SET(options, RECURSIVE);
3322 /*
3323 * Also pickaxe would not work very well if you do not say recursive
3324 */
3325 if (options->pickaxe)
3326 DIFF_OPT_SET(options, RECURSIVE);
3327 /*
3328 * When patches are generated, submodules diffed against the work tree
3329 * must be checked for dirtiness too so it can be shown in the output
3330 */
3331 if (options->output_format & DIFF_FORMAT_PATCH)
3332 DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3333
3334 if (options->detect_rename && options->rename_limit < 0)
3335 options->rename_limit = diff_rename_limit_default;
3336 if (options->setup & DIFF_SETUP_USE_CACHE) {
3337 if (!active_cache)
3338 /* read-cache does not die even when it fails
3339 * so it is safe for us to do this here. Also
3340 * it does not smudge active_cache or active_nr
3341 * when it fails, so we do not have to worry about
3342 * cleaning it up ourselves either.
3343 */
3344 read_cache();
3345 }
3346 if (options->abbrev <= 0 || 40 < options->abbrev)
3347 options->abbrev = 40; /* full */
3348
3349 /*
3350 * It does not make sense to show the first hit we happened
3351 * to have found. It does not make sense not to return with
3352 * exit code in such a case either.
3353 */
3354 if (DIFF_OPT_TST(options, QUICK)) {
3355 options->output_format = DIFF_FORMAT_NO_OUTPUT;
3356 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3357 }
3358
3359 options->diff_path_counter = 0;
3360
3361 if (DIFF_OPT_TST(options, FOLLOW_RENAMES) && options->pathspec.nr != 1)
3362 die(_("--follow requires exactly one pathspec"));
3363 }
3364
3365 static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3366 {
3367 char c, *eq;
3368 int len;
3369
3370 if (*arg != '-')
3371 return 0;
3372 c = *++arg;
3373 if (!c)
3374 return 0;
3375 if (c == arg_short) {
3376 c = *++arg;
3377 if (!c)
3378 return 1;
3379 if (val && isdigit(c)) {
3380 char *end;
3381 int n = strtoul(arg, &end, 10);
3382 if (*end)
3383 return 0;
3384 *val = n;
3385 return 1;
3386 }
3387 return 0;
3388 }
3389 if (c != '-')
3390 return 0;
3391 arg++;
3392 eq = strchrnul(arg, '=');
3393 len = eq - arg;
3394 if (!len || strncmp(arg, arg_long, len))
3395 return 0;
3396 if (*eq) {
3397 int n;
3398 char *end;
3399 if (!isdigit(*++eq))
3400 return 0;
3401 n = strtoul(eq, &end, 10);
3402 if (*end)
3403 return 0;
3404 *val = n;
3405 }
3406 return 1;
3407 }
3408
3409 static int diff_scoreopt_parse(const char *opt);
3410
3411 static inline int short_opt(char opt, const char **argv,
3412 const char **optarg)
3413 {
3414 const char *arg = argv[0];
3415 if (arg[0] != '-' || arg[1] != opt)
3416 return 0;
3417 if (arg[2] != '\0') {
3418 *optarg = arg + 2;
3419 return 1;
3420 }
3421 if (!argv[1])
3422 die("Option '%c' requires a value", opt);
3423 *optarg = argv[1];
3424 return 2;
3425 }
3426
3427 int parse_long_opt(const char *opt, const char **argv,
3428 const char **optarg)
3429 {
3430 const char *arg = argv[0];
3431 if (!skip_prefix(arg, "--", &arg))
3432 return 0;
3433 if (!skip_prefix(arg, opt, &arg))
3434 return 0;
3435 if (*arg == '=') { /* stuck form: --option=value */
3436 *optarg = arg + 1;
3437 return 1;
3438 }
3439 if (*arg != '\0')
3440 return 0;
3441 /* separate form: --option value */
3442 if (!argv[1])
3443 die("Option '--%s' requires a value", opt);
3444 *optarg = argv[1];
3445 return 2;
3446 }
3447
3448 static int stat_opt(struct diff_options *options, const char **av)
3449 {
3450 const char *arg = av[0];
3451 char *end;
3452 int width = options->stat_width;
3453 int name_width = options->stat_name_width;
3454 int graph_width = options->stat_graph_width;
3455 int count = options->stat_count;
3456 int argcount = 1;
3457
3458 if (!skip_prefix(arg, "--stat", &arg))
3459 die("BUG: stat option does not begin with --stat: %s", arg);
3460 end = (char *)arg;
3461
3462 switch (*arg) {
3463 case '-':
3464 if (skip_prefix(arg, "-width", &arg)) {
3465 if (*arg == '=')
3466 width = strtoul(arg + 1, &end, 10);
3467 else if (!*arg && !av[1])
3468 die("Option '--stat-width' requires a value");
3469 else if (!*arg) {
3470 width = strtoul(av[1], &end, 10);
3471 argcount = 2;
3472 }
3473 } else if (skip_prefix(arg, "-name-width", &arg)) {
3474 if (*arg == '=')
3475 name_width = strtoul(arg + 1, &end, 10);
3476 else if (!*arg && !av[1])
3477 die("Option '--stat-name-width' requires a value");
3478 else if (!*arg) {
3479 name_width = strtoul(av[1], &end, 10);
3480 argcount = 2;
3481 }
3482 } else if (skip_prefix(arg, "-graph-width", &arg)) {
3483 if (*arg == '=')
3484 graph_width = strtoul(arg + 1, &end, 10);
3485 else if (!*arg && !av[1])
3486 die("Option '--stat-graph-width' requires a value");
3487 else if (!*arg) {
3488 graph_width = strtoul(av[1], &end, 10);
3489 argcount = 2;
3490 }
3491 } else if (skip_prefix(arg, "-count", &arg)) {
3492 if (*arg == '=')
3493 count = strtoul(arg + 1, &end, 10);
3494 else if (!*arg && !av[1])
3495 die("Option '--stat-count' requires a value");
3496 else if (!*arg) {
3497 count = strtoul(av[1], &end, 10);
3498 argcount = 2;
3499 }
3500 }
3501 break;
3502 case '=':
3503 width = strtoul(arg+1, &end, 10);
3504 if (*end == ',')
3505 name_width = strtoul(end+1, &end, 10);
3506 if (*end == ',')
3507 count = strtoul(end+1, &end, 10);
3508 }
3509
3510 /* Important! This checks all the error cases! */
3511 if (*end)
3512 return 0;
3513 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3514 options->stat_name_width = name_width;
3515 options->stat_graph_width = graph_width;
3516 options->stat_width = width;
3517 options->stat_count = count;
3518 return argcount;
3519 }
3520
3521 static int parse_dirstat_opt(struct diff_options *options, const char *params)
3522 {
3523 struct strbuf errmsg = STRBUF_INIT;
3524 if (parse_dirstat_params(options, params, &errmsg))
3525 die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3526 errmsg.buf);
3527 strbuf_release(&errmsg);
3528 /*
3529 * The caller knows a dirstat-related option is given from the command
3530 * line; allow it to say "return this_function();"
3531 */
3532 options->output_format |= DIFF_FORMAT_DIRSTAT;
3533 return 1;
3534 }
3535
3536 static int parse_submodule_opt(struct diff_options *options, const char *value)
3537 {
3538 if (parse_submodule_params(options, value))
3539 die(_("Failed to parse --submodule option parameter: '%s'"),
3540 value);
3541 return 1;
3542 }
3543
3544 static const char diff_status_letters[] = {
3545 DIFF_STATUS_ADDED,
3546 DIFF_STATUS_COPIED,
3547 DIFF_STATUS_DELETED,
3548 DIFF_STATUS_MODIFIED,
3549 DIFF_STATUS_RENAMED,
3550 DIFF_STATUS_TYPE_CHANGED,
3551 DIFF_STATUS_UNKNOWN,
3552 DIFF_STATUS_UNMERGED,
3553 DIFF_STATUS_FILTER_AON,
3554 DIFF_STATUS_FILTER_BROKEN,
3555 '\0',
3556 };
3557
3558 static unsigned int filter_bit['Z' + 1];
3559
3560 static void prepare_filter_bits(void)
3561 {
3562 int i;
3563
3564 if (!filter_bit[DIFF_STATUS_ADDED]) {
3565 for (i = 0; diff_status_letters[i]; i++)
3566 filter_bit[(int) diff_status_letters[i]] = (1 << i);
3567 }
3568 }
3569
3570 static unsigned filter_bit_tst(char status, const struct diff_options *opt)
3571 {
3572 return opt->filter & filter_bit[(int) status];
3573 }
3574
3575 static int parse_diff_filter_opt(const char *optarg, struct diff_options *opt)
3576 {
3577 int i, optch;
3578
3579 prepare_filter_bits();
3580
3581 /*
3582 * If there is a negation e.g. 'd' in the input, and we haven't
3583 * initialized the filter field with another --diff-filter, start
3584 * from full set of bits, except for AON.
3585 */
3586 if (!opt->filter) {
3587 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3588 if (optch < 'a' || 'z' < optch)
3589 continue;
3590 opt->filter = (1 << (ARRAY_SIZE(diff_status_letters) - 1)) - 1;
3591 opt->filter &= ~filter_bit[DIFF_STATUS_FILTER_AON];
3592 break;
3593 }
3594 }
3595
3596 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3597 unsigned int bit;
3598 int negate;
3599
3600 if ('a' <= optch && optch <= 'z') {
3601 negate = 1;
3602 optch = toupper(optch);
3603 } else {
3604 negate = 0;
3605 }
3606
3607 bit = (0 <= optch && optch <= 'Z') ? filter_bit[optch] : 0;
3608 if (!bit)
3609 return optarg[i];
3610 if (negate)
3611 opt->filter &= ~bit;
3612 else
3613 opt->filter |= bit;
3614 }
3615 return 0;
3616 }
3617
3618 static void enable_patch_output(int *fmt) {
3619 *fmt &= ~DIFF_FORMAT_NO_OUTPUT;
3620 *fmt |= DIFF_FORMAT_PATCH;
3621 }
3622
3623 int diff_opt_parse(struct diff_options *options, const char **av, int ac)
3624 {
3625 const char *arg = av[0];
3626 const char *optarg;
3627 int argcount;
3628
3629 /* Output format options */
3630 if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch")
3631 || opt_arg(arg, 'U', "unified", &options->context))
3632 enable_patch_output(&options->output_format);
3633 else if (!strcmp(arg, "--raw"))
3634 options->output_format |= DIFF_FORMAT_RAW;
3635 else if (!strcmp(arg, "--patch-with-raw")) {
3636 enable_patch_output(&options->output_format);
3637 options->output_format |= DIFF_FORMAT_RAW;
3638 } else if (!strcmp(arg, "--numstat"))
3639 options->output_format |= DIFF_FORMAT_NUMSTAT;
3640 else if (!strcmp(arg, "--shortstat"))
3641 options->output_format |= DIFF_FORMAT_SHORTSTAT;
3642 else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3643 return parse_dirstat_opt(options, "");
3644 else if (skip_prefix(arg, "-X", &arg))
3645 return parse_dirstat_opt(options, arg);
3646 else if (skip_prefix(arg, "--dirstat=", &arg))
3647 return parse_dirstat_opt(options, arg);
3648 else if (!strcmp(arg, "--cumulative"))
3649 return parse_dirstat_opt(options, "cumulative");
3650 else if (!strcmp(arg, "--dirstat-by-file"))
3651 return parse_dirstat_opt(options, "files");
3652 else if (skip_prefix(arg, "--dirstat-by-file=", &arg)) {
3653 parse_dirstat_opt(options, "files");
3654 return parse_dirstat_opt(options, arg);
3655 }
3656 else if (!strcmp(arg, "--check"))
3657 options->output_format |= DIFF_FORMAT_CHECKDIFF;
3658 else if (!strcmp(arg, "--summary"))
3659 options->output_format |= DIFF_FORMAT_SUMMARY;
3660 else if (!strcmp(arg, "--patch-with-stat")) {
3661 enable_patch_output(&options->output_format);
3662 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3663 } else if (!strcmp(arg, "--name-only"))
3664 options->output_format |= DIFF_FORMAT_NAME;
3665 else if (!strcmp(arg, "--name-status"))
3666 options->output_format |= DIFF_FORMAT_NAME_STATUS;
3667 else if (!strcmp(arg, "-s") || !strcmp(arg, "--no-patch"))
3668 options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3669 else if (starts_with(arg, "--stat"))
3670 /* --stat, --stat-width, --stat-name-width, or --stat-count */
3671 return stat_opt(options, av);
3672
3673 /* renames options */
3674 else if (starts_with(arg, "-B") || starts_with(arg, "--break-rewrites=") ||
3675 !strcmp(arg, "--break-rewrites")) {
3676 if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3677 return error("invalid argument to -B: %s", arg+2);
3678 }
3679 else if (starts_with(arg, "-M") || starts_with(arg, "--find-renames=") ||
3680 !strcmp(arg, "--find-renames")) {
3681 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3682 return error("invalid argument to -M: %s", arg+2);
3683 options->detect_rename = DIFF_DETECT_RENAME;
3684 }
3685 else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3686 options->irreversible_delete = 1;
3687 }
3688 else if (starts_with(arg, "-C") || starts_with(arg, "--find-copies=") ||
3689 !strcmp(arg, "--find-copies")) {
3690 if (options->detect_rename == DIFF_DETECT_COPY)
3691 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3692 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3693 return error("invalid argument to -C: %s", arg+2);
3694 options->detect_rename = DIFF_DETECT_COPY;
3695 }
3696 else if (!strcmp(arg, "--no-renames"))
3697 options->detect_rename = 0;
3698 else if (!strcmp(arg, "--rename-empty"))
3699 DIFF_OPT_SET(options, RENAME_EMPTY);
3700 else if (!strcmp(arg, "--no-rename-empty"))
3701 DIFF_OPT_CLR(options, RENAME_EMPTY);
3702 else if (!strcmp(arg, "--relative"))
3703 DIFF_OPT_SET(options, RELATIVE_NAME);
3704 else if (skip_prefix(arg, "--relative=", &arg)) {
3705 DIFF_OPT_SET(options, RELATIVE_NAME);
3706 options->prefix = arg;
3707 }
3708
3709 /* xdiff options */
3710 else if (!strcmp(arg, "--minimal"))
3711 DIFF_XDL_SET(options, NEED_MINIMAL);
3712 else if (!strcmp(arg, "--no-minimal"))
3713 DIFF_XDL_CLR(options, NEED_MINIMAL);
3714 else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3715 DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3716 else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3717 DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3718 else if (!strcmp(arg, "--ignore-space-at-eol"))
3719 DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3720 else if (!strcmp(arg, "--ignore-blank-lines"))
3721 DIFF_XDL_SET(options, IGNORE_BLANK_LINES);
3722 else if (!strcmp(arg, "--compaction-heuristic"))
3723 DIFF_XDL_SET(options, COMPACTION_HEURISTIC);
3724 else if (!strcmp(arg, "--no-compaction-heuristic"))
3725 DIFF_XDL_CLR(options, COMPACTION_HEURISTIC);
3726 else if (!strcmp(arg, "--patience"))
3727 options->xdl_opts = DIFF_WITH_ALG(options, PATIENCE_DIFF);
3728 else if (!strcmp(arg, "--histogram"))
3729 options->xdl_opts = DIFF_WITH_ALG(options, HISTOGRAM_DIFF);
3730 else if ((argcount = parse_long_opt("diff-algorithm", av, &optarg))) {
3731 long value = parse_algorithm_value(optarg);
3732 if (value < 0)
3733 return error("option diff-algorithm accepts \"myers\", "
3734 "\"minimal\", \"patience\" and \"histogram\"");
3735 /* clear out previous settings */
3736 DIFF_XDL_CLR(options, NEED_MINIMAL);
3737 options->xdl_opts &= ~XDF_DIFF_ALGORITHM_MASK;
3738 options->xdl_opts |= value;
3739 return argcount;
3740 }
3741
3742 /* flags options */
3743 else if (!strcmp(arg, "--binary")) {
3744 enable_patch_output(&options->output_format);
3745 DIFF_OPT_SET(options, BINARY);
3746 }
3747 else if (!strcmp(arg, "--full-index"))
3748 DIFF_OPT_SET(options, FULL_INDEX);
3749 else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
3750 DIFF_OPT_SET(options, TEXT);
3751 else if (!strcmp(arg, "-R"))
3752 DIFF_OPT_SET(options, REVERSE_DIFF);
3753 else if (!strcmp(arg, "--find-copies-harder"))
3754 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3755 else if (!strcmp(arg, "--follow"))
3756 DIFF_OPT_SET(options, FOLLOW_RENAMES);
3757 else if (!strcmp(arg, "--no-follow"))
3758 DIFF_OPT_CLR(options, FOLLOW_RENAMES);
3759 else if (!strcmp(arg, "--color"))
3760 options->use_color = 1;
3761 else if (skip_prefix(arg, "--color=", &arg)) {
3762 int value = git_config_colorbool(NULL, arg);
3763 if (value < 0)
3764 return error("option `color' expects \"always\", \"auto\", or \"never\"");
3765 options->use_color = value;
3766 }
3767 else if (!strcmp(arg, "--no-color"))
3768 options->use_color = 0;
3769 else if (!strcmp(arg, "--color-words")) {
3770 options->use_color = 1;
3771 options->word_diff = DIFF_WORDS_COLOR;
3772 }
3773 else if (skip_prefix(arg, "--color-words=", &arg)) {
3774 options->use_color = 1;
3775 options->word_diff = DIFF_WORDS_COLOR;
3776 options->word_regex = arg;
3777 }
3778 else if (!strcmp(arg, "--word-diff")) {
3779 if (options->word_diff == DIFF_WORDS_NONE)
3780 options->word_diff = DIFF_WORDS_PLAIN;
3781 }
3782 else if (skip_prefix(arg, "--word-diff=", &arg)) {
3783 if (!strcmp(arg, "plain"))
3784 options->word_diff = DIFF_WORDS_PLAIN;
3785 else if (!strcmp(arg, "color")) {
3786 options->use_color = 1;
3787 options->word_diff = DIFF_WORDS_COLOR;
3788 }
3789 else if (!strcmp(arg, "porcelain"))
3790 options->word_diff = DIFF_WORDS_PORCELAIN;
3791 else if (!strcmp(arg, "none"))
3792 options->word_diff = DIFF_WORDS_NONE;
3793 else
3794 die("bad --word-diff argument: %s", arg);
3795 }
3796 else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
3797 if (options->word_diff == DIFF_WORDS_NONE)
3798 options->word_diff = DIFF_WORDS_PLAIN;
3799 options->word_regex = optarg;
3800 return argcount;
3801 }
3802 else if (!strcmp(arg, "--exit-code"))
3803 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3804 else if (!strcmp(arg, "--quiet"))
3805 DIFF_OPT_SET(options, QUICK);
3806 else if (!strcmp(arg, "--ext-diff"))
3807 DIFF_OPT_SET(options, ALLOW_EXTERNAL);
3808 else if (!strcmp(arg, "--no-ext-diff"))
3809 DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
3810 else if (!strcmp(arg, "--textconv"))
3811 DIFF_OPT_SET(options, ALLOW_TEXTCONV);
3812 else if (!strcmp(arg, "--no-textconv"))
3813 DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
3814 else if (!strcmp(arg, "--ignore-submodules")) {
3815 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3816 handle_ignore_submodules_arg(options, "all");
3817 } else if (skip_prefix(arg, "--ignore-submodules=", &arg)) {
3818 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3819 handle_ignore_submodules_arg(options, arg);
3820 } else if (!strcmp(arg, "--submodule"))
3821 DIFF_OPT_SET(options, SUBMODULE_LOG);
3822 else if (skip_prefix(arg, "--submodule=", &arg))
3823 return parse_submodule_opt(options, arg);
3824
3825 /* misc options */
3826 else if (!strcmp(arg, "-z"))
3827 options->line_termination = 0;
3828 else if ((argcount = short_opt('l', av, &optarg))) {
3829 options->rename_limit = strtoul(optarg, NULL, 10);
3830 return argcount;
3831 }
3832 else if ((argcount = short_opt('S', av, &optarg))) {
3833 options->pickaxe = optarg;
3834 options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
3835 return argcount;
3836 } else if ((argcount = short_opt('G', av, &optarg))) {
3837 options->pickaxe = optarg;
3838 options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
3839 return argcount;
3840 }
3841 else if (!strcmp(arg, "--pickaxe-all"))
3842 options->pickaxe_opts |= DIFF_PICKAXE_ALL;
3843 else if (!strcmp(arg, "--pickaxe-regex"))
3844 options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
3845 else if ((argcount = short_opt('O', av, &optarg))) {
3846 options->orderfile = optarg;
3847 return argcount;
3848 }
3849 else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
3850 int offending = parse_diff_filter_opt(optarg, options);
3851 if (offending)
3852 die("unknown change class '%c' in --diff-filter=%s",
3853 offending, optarg);
3854 return argcount;
3855 }
3856 else if (!strcmp(arg, "--abbrev"))
3857 options->abbrev = DEFAULT_ABBREV;
3858 else if (skip_prefix(arg, "--abbrev=", &arg)) {
3859 options->abbrev = strtoul(arg, NULL, 10);
3860 if (options->abbrev < MINIMUM_ABBREV)
3861 options->abbrev = MINIMUM_ABBREV;
3862 else if (40 < options->abbrev)
3863 options->abbrev = 40;
3864 }
3865 else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
3866 options->a_prefix = optarg;
3867 return argcount;
3868 }
3869 else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
3870 options->b_prefix = optarg;
3871 return argcount;
3872 }
3873 else if (!strcmp(arg, "--no-prefix"))
3874 options->a_prefix = options->b_prefix = "";
3875 else if (opt_arg(arg, '\0', "inter-hunk-context",
3876 &options->interhunkcontext))
3877 ;
3878 else if (!strcmp(arg, "-W"))
3879 DIFF_OPT_SET(options, FUNCCONTEXT);
3880 else if (!strcmp(arg, "--function-context"))
3881 DIFF_OPT_SET(options, FUNCCONTEXT);
3882 else if (!strcmp(arg, "--no-function-context"))
3883 DIFF_OPT_CLR(options, FUNCCONTEXT);
3884 else if ((argcount = parse_long_opt("output", av, &optarg))) {
3885 options->file = fopen(optarg, "w");
3886 if (!options->file)
3887 die_errno("Could not open '%s'", optarg);
3888 options->close_file = 1;
3889 return argcount;
3890 } else
3891 return 0;
3892 return 1;
3893 }
3894
3895 int parse_rename_score(const char **cp_p)
3896 {
3897 unsigned long num, scale;
3898 int ch, dot;
3899 const char *cp = *cp_p;
3900
3901 num = 0;
3902 scale = 1;
3903 dot = 0;
3904 for (;;) {
3905 ch = *cp;
3906 if ( !dot && ch == '.' ) {
3907 scale = 1;
3908 dot = 1;
3909 } else if ( ch == '%' ) {
3910 scale = dot ? scale*100 : 100;
3911 cp++; /* % is always at the end */
3912 break;
3913 } else if ( ch >= '0' && ch <= '9' ) {
3914 if ( scale < 100000 ) {
3915 scale *= 10;
3916 num = (num*10) + (ch-'0');
3917 }
3918 } else {
3919 break;
3920 }
3921 cp++;
3922 }
3923 *cp_p = cp;
3924
3925 /* user says num divided by scale and we say internally that
3926 * is MAX_SCORE * num / scale.
3927 */
3928 return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
3929 }
3930
3931 static int diff_scoreopt_parse(const char *opt)
3932 {
3933 int opt1, opt2, cmd;
3934
3935 if (*opt++ != '-')
3936 return -1;
3937 cmd = *opt++;
3938 if (cmd == '-') {
3939 /* convert the long-form arguments into short-form versions */
3940 if (skip_prefix(opt, "break-rewrites", &opt)) {
3941 if (*opt == 0 || *opt++ == '=')
3942 cmd = 'B';
3943 } else if (skip_prefix(opt, "find-copies", &opt)) {
3944 if (*opt == 0 || *opt++ == '=')
3945 cmd = 'C';
3946 } else if (skip_prefix(opt, "find-renames", &opt)) {
3947 if (*opt == 0 || *opt++ == '=')
3948 cmd = 'M';
3949 }
3950 }
3951 if (cmd != 'M' && cmd != 'C' && cmd != 'B')
3952 return -1; /* that is not a -M, -C, or -B option */
3953
3954 opt1 = parse_rename_score(&opt);
3955 if (cmd != 'B')
3956 opt2 = 0;
3957 else {
3958 if (*opt == 0)
3959 opt2 = 0;
3960 else if (*opt != '/')
3961 return -1; /* we expect -B80/99 or -B80 */
3962 else {
3963 opt++;
3964 opt2 = parse_rename_score(&opt);
3965 }
3966 }
3967 if (*opt != 0)
3968 return -1;
3969 return opt1 | (opt2 << 16);
3970 }
3971
3972 struct diff_queue_struct diff_queued_diff;
3973
3974 void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
3975 {
3976 ALLOC_GROW(queue->queue, queue->nr + 1, queue->alloc);
3977 queue->queue[queue->nr++] = dp;
3978 }
3979
3980 struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
3981 struct diff_filespec *one,
3982 struct diff_filespec *two)
3983 {
3984 struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
3985 dp->one = one;
3986 dp->two = two;
3987 if (queue)
3988 diff_q(queue, dp);
3989 return dp;
3990 }
3991
3992 void diff_free_filepair(struct diff_filepair *p)
3993 {
3994 free_filespec(p->one);
3995 free_filespec(p->two);
3996 free(p);
3997 }
3998
3999 /* This is different from find_unique_abbrev() in that
4000 * it stuffs the result with dots for alignment.
4001 */
4002 const char *diff_unique_abbrev(const unsigned char *sha1, int len)
4003 {
4004 int abblen;
4005 const char *abbrev;
4006 if (len == 40)
4007 return sha1_to_hex(sha1);
4008
4009 abbrev = find_unique_abbrev(sha1, len);
4010 abblen = strlen(abbrev);
4011 if (abblen < 37) {
4012 static char hex[41];
4013 if (len < abblen && abblen <= len + 2)
4014 sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
4015 else
4016 sprintf(hex, "%s...", abbrev);
4017 return hex;
4018 }
4019 return sha1_to_hex(sha1);
4020 }
4021
4022 static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
4023 {
4024 int line_termination = opt->line_termination;
4025 int inter_name_termination = line_termination ? '\t' : '\0';
4026
4027 fprintf(opt->file, "%s", diff_line_prefix(opt));
4028 if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
4029 fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
4030 diff_unique_abbrev(p->one->sha1, opt->abbrev));
4031 fprintf(opt->file, "%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
4032 }
4033 if (p->score) {
4034 fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
4035 inter_name_termination);
4036 } else {
4037 fprintf(opt->file, "%c%c", p->status, inter_name_termination);
4038 }
4039
4040 if (p->status == DIFF_STATUS_COPIED ||
4041 p->status == DIFF_STATUS_RENAMED) {
4042 const char *name_a, *name_b;
4043 name_a = p->one->path;
4044 name_b = p->two->path;
4045 strip_prefix(opt->prefix_length, &name_a, &name_b);
4046 write_name_quoted(name_a, opt->file, inter_name_termination);
4047 write_name_quoted(name_b, opt->file, line_termination);
4048 } else {
4049 const char *name_a, *name_b;
4050 name_a = p->one->mode ? p->one->path : p->two->path;
4051 name_b = NULL;
4052 strip_prefix(opt->prefix_length, &name_a, &name_b);
4053 write_name_quoted(name_a, opt->file, line_termination);
4054 }
4055 }
4056
4057 int diff_unmodified_pair(struct diff_filepair *p)
4058 {
4059 /* This function is written stricter than necessary to support
4060 * the currently implemented transformers, but the idea is to
4061 * let transformers to produce diff_filepairs any way they want,
4062 * and filter and clean them up here before producing the output.
4063 */
4064 struct diff_filespec *one = p->one, *two = p->two;
4065
4066 if (DIFF_PAIR_UNMERGED(p))
4067 return 0; /* unmerged is interesting */
4068
4069 /* deletion, addition, mode or type change
4070 * and rename are all interesting.
4071 */
4072 if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
4073 DIFF_PAIR_MODE_CHANGED(p) ||
4074 strcmp(one->path, two->path))
4075 return 0;
4076
4077 /* both are valid and point at the same path. that is, we are
4078 * dealing with a change.
4079 */
4080 if (one->sha1_valid && two->sha1_valid &&
4081 !hashcmp(one->sha1, two->sha1) &&
4082 !one->dirty_submodule && !two->dirty_submodule)
4083 return 1; /* no change */
4084 if (!one->sha1_valid && !two->sha1_valid)
4085 return 1; /* both look at the same file on the filesystem. */
4086 return 0;
4087 }
4088
4089 static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
4090 {
4091 if (diff_unmodified_pair(p))
4092 return;
4093
4094 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4095 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4096 return; /* no tree diffs in patch format */
4097
4098 run_diff(p, o);
4099 }
4100
4101 static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
4102 struct diffstat_t *diffstat)
4103 {
4104 if (diff_unmodified_pair(p))
4105 return;
4106
4107 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4108 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4109 return; /* no useful stat for tree diffs */
4110
4111 run_diffstat(p, o, diffstat);
4112 }
4113
4114 static void diff_flush_checkdiff(struct diff_filepair *p,
4115 struct diff_options *o)
4116 {
4117 if (diff_unmodified_pair(p))
4118 return;
4119
4120 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4121 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4122 return; /* nothing to check in tree diffs */
4123
4124 run_checkdiff(p, o);
4125 }
4126
4127 int diff_queue_is_empty(void)
4128 {
4129 struct diff_queue_struct *q = &diff_queued_diff;
4130 int i;
4131 for (i = 0; i < q->nr; i++)
4132 if (!diff_unmodified_pair(q->queue[i]))
4133 return 0;
4134 return 1;
4135 }
4136
4137 #if DIFF_DEBUG
4138 void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
4139 {
4140 fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
4141 x, one ? one : "",
4142 s->path,
4143 DIFF_FILE_VALID(s) ? "valid" : "invalid",
4144 s->mode,
4145 s->sha1_valid ? sha1_to_hex(s->sha1) : "");
4146 fprintf(stderr, "queue[%d] %s size %lu\n",
4147 x, one ? one : "",
4148 s->size);
4149 }
4150
4151 void diff_debug_filepair(const struct diff_filepair *p, int i)
4152 {
4153 diff_debug_filespec(p->one, i, "one");
4154 diff_debug_filespec(p->two, i, "two");
4155 fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
4156 p->score, p->status ? p->status : '?',
4157 p->one->rename_used, p->broken_pair);
4158 }
4159
4160 void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
4161 {
4162 int i;
4163 if (msg)
4164 fprintf(stderr, "%s\n", msg);
4165 fprintf(stderr, "q->nr = %d\n", q->nr);
4166 for (i = 0; i < q->nr; i++) {
4167 struct diff_filepair *p = q->queue[i];
4168 diff_debug_filepair(p, i);
4169 }
4170 }
4171 #endif
4172
4173 static void diff_resolve_rename_copy(void)
4174 {
4175 int i;
4176 struct diff_filepair *p;
4177 struct diff_queue_struct *q = &diff_queued_diff;
4178
4179 diff_debug_queue("resolve-rename-copy", q);
4180
4181 for (i = 0; i < q->nr; i++) {
4182 p = q->queue[i];
4183 p->status = 0; /* undecided */
4184 if (DIFF_PAIR_UNMERGED(p))
4185 p->status = DIFF_STATUS_UNMERGED;
4186 else if (!DIFF_FILE_VALID(p->one))
4187 p->status = DIFF_STATUS_ADDED;
4188 else if (!DIFF_FILE_VALID(p->two))
4189 p->status = DIFF_STATUS_DELETED;
4190 else if (DIFF_PAIR_TYPE_CHANGED(p))
4191 p->status = DIFF_STATUS_TYPE_CHANGED;
4192
4193 /* from this point on, we are dealing with a pair
4194 * whose both sides are valid and of the same type, i.e.
4195 * either in-place edit or rename/copy edit.
4196 */
4197 else if (DIFF_PAIR_RENAME(p)) {
4198 /*
4199 * A rename might have re-connected a broken
4200 * pair up, causing the pathnames to be the
4201 * same again. If so, that's not a rename at
4202 * all, just a modification..
4203 *
4204 * Otherwise, see if this source was used for
4205 * multiple renames, in which case we decrement
4206 * the count, and call it a copy.
4207 */
4208 if (!strcmp(p->one->path, p->two->path))
4209 p->status = DIFF_STATUS_MODIFIED;
4210 else if (--p->one->rename_used > 0)
4211 p->status = DIFF_STATUS_COPIED;
4212 else
4213 p->status = DIFF_STATUS_RENAMED;
4214 }
4215 else if (hashcmp(p->one->sha1, p->two->sha1) ||
4216 p->one->mode != p->two->mode ||
4217 p->one->dirty_submodule ||
4218 p->two->dirty_submodule ||
4219 is_null_sha1(p->one->sha1))
4220 p->status = DIFF_STATUS_MODIFIED;
4221 else {
4222 /* This is a "no-change" entry and should not
4223 * happen anymore, but prepare for broken callers.
4224 */
4225 error("feeding unmodified %s to diffcore",
4226 p->one->path);
4227 p->status = DIFF_STATUS_UNKNOWN;
4228 }
4229 }
4230 diff_debug_queue("resolve-rename-copy done", q);
4231 }
4232
4233 static int check_pair_status(struct diff_filepair *p)
4234 {
4235 switch (p->status) {
4236 case DIFF_STATUS_UNKNOWN:
4237 return 0;
4238 case 0:
4239 die("internal error in diff-resolve-rename-copy");
4240 default:
4241 return 1;
4242 }
4243 }
4244
4245 static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
4246 {
4247 int fmt = opt->output_format;
4248
4249 if (fmt & DIFF_FORMAT_CHECKDIFF)
4250 diff_flush_checkdiff(p, opt);
4251 else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
4252 diff_flush_raw(p, opt);
4253 else if (fmt & DIFF_FORMAT_NAME) {
4254 const char *name_a, *name_b;
4255 name_a = p->two->path;
4256 name_b = NULL;
4257 strip_prefix(opt->prefix_length, &name_a, &name_b);
4258 write_name_quoted(name_a, opt->file, opt->line_termination);
4259 }
4260 }
4261
4262 static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
4263 {
4264 if (fs->mode)
4265 fprintf(file, " %s mode %06o ", newdelete, fs->mode);
4266 else
4267 fprintf(file, " %s ", newdelete);
4268 write_name_quoted(fs->path, file, '\n');
4269 }
4270
4271
4272 static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
4273 const char *line_prefix)
4274 {
4275 if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
4276 fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
4277 p->two->mode, show_name ? ' ' : '\n');
4278 if (show_name) {
4279 write_name_quoted(p->two->path, file, '\n');
4280 }
4281 }
4282 }
4283
4284 static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
4285 const char *line_prefix)
4286 {
4287 char *names = pprint_rename(p->one->path, p->two->path);
4288
4289 fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
4290 free(names);
4291 show_mode_change(file, p, 0, line_prefix);
4292 }
4293
4294 static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
4295 {
4296 FILE *file = opt->file;
4297 const char *line_prefix = diff_line_prefix(opt);
4298
4299 switch(p->status) {
4300 case DIFF_STATUS_DELETED:
4301 fputs(line_prefix, file);
4302 show_file_mode_name(file, "delete", p->one);
4303 break;
4304 case DIFF_STATUS_ADDED:
4305 fputs(line_prefix, file);
4306 show_file_mode_name(file, "create", p->two);
4307 break;
4308 case DIFF_STATUS_COPIED:
4309 fputs(line_prefix, file);
4310 show_rename_copy(file, "copy", p, line_prefix);
4311 break;
4312 case DIFF_STATUS_RENAMED:
4313 fputs(line_prefix, file);
4314 show_rename_copy(file, "rename", p, line_prefix);
4315 break;
4316 default:
4317 if (p->score) {
4318 fprintf(file, "%s rewrite ", line_prefix);
4319 write_name_quoted(p->two->path, file, ' ');
4320 fprintf(file, "(%d%%)\n", similarity_index(p));
4321 }
4322 show_mode_change(file, p, !p->score, line_prefix);
4323 break;
4324 }
4325 }
4326
4327 struct patch_id_t {
4328 git_SHA_CTX *ctx;
4329 int patchlen;
4330 };
4331
4332 static int remove_space(char *line, int len)
4333 {
4334 int i;
4335 char *dst = line;
4336 unsigned char c;
4337
4338 for (i = 0; i < len; i++)
4339 if (!isspace((c = line[i])))
4340 *dst++ = c;
4341
4342 return dst - line;
4343 }
4344
4345 static void patch_id_consume(void *priv, char *line, unsigned long len)
4346 {
4347 struct patch_id_t *data = priv;
4348 int new_len;
4349
4350 /* Ignore line numbers when computing the SHA1 of the patch */
4351 if (starts_with(line, "@@ -"))
4352 return;
4353
4354 new_len = remove_space(line, len);
4355
4356 git_SHA1_Update(data->ctx, line, new_len);
4357 data->patchlen += new_len;
4358 }
4359
4360 /* returns 0 upon success, and writes result into sha1 */
4361 static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
4362 {
4363 struct diff_queue_struct *q = &diff_queued_diff;
4364 int i;
4365 git_SHA_CTX ctx;
4366 struct patch_id_t data;
4367 char buffer[PATH_MAX * 4 + 20];
4368
4369 git_SHA1_Init(&ctx);
4370 memset(&data, 0, sizeof(struct patch_id_t));
4371 data.ctx = &ctx;
4372
4373 for (i = 0; i < q->nr; i++) {
4374 xpparam_t xpp;
4375 xdemitconf_t xecfg;
4376 mmfile_t mf1, mf2;
4377 struct diff_filepair *p = q->queue[i];
4378 int len1, len2;
4379
4380 memset(&xpp, 0, sizeof(xpp));
4381 memset(&xecfg, 0, sizeof(xecfg));
4382 if (p->status == 0)
4383 return error("internal diff status error");
4384 if (p->status == DIFF_STATUS_UNKNOWN)
4385 continue;
4386 if (diff_unmodified_pair(p))
4387 continue;
4388 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4389 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4390 continue;
4391 if (DIFF_PAIR_UNMERGED(p))
4392 continue;
4393
4394 diff_fill_sha1_info(p->one);
4395 diff_fill_sha1_info(p->two);
4396 if (fill_mmfile(&mf1, p->one) < 0 ||
4397 fill_mmfile(&mf2, p->two) < 0)
4398 return error("unable to read files to diff");
4399
4400 len1 = remove_space(p->one->path, strlen(p->one->path));
4401 len2 = remove_space(p->two->path, strlen(p->two->path));
4402 if (p->one->mode == 0)
4403 len1 = snprintf(buffer, sizeof(buffer),
4404 "diff--gita/%.*sb/%.*s"
4405 "newfilemode%06o"
4406 "---/dev/null"
4407 "+++b/%.*s",
4408 len1, p->one->path,
4409 len2, p->two->path,
4410 p->two->mode,
4411 len2, p->two->path);
4412 else if (p->two->mode == 0)
4413 len1 = snprintf(buffer, sizeof(buffer),
4414 "diff--gita/%.*sb/%.*s"
4415 "deletedfilemode%06o"
4416 "---a/%.*s"
4417 "+++/dev/null",
4418 len1, p->one->path,
4419 len2, p->two->path,
4420 p->one->mode,
4421 len1, p->one->path);
4422 else
4423 len1 = snprintf(buffer, sizeof(buffer),
4424 "diff--gita/%.*sb/%.*s"
4425 "---a/%.*s"
4426 "+++b/%.*s",
4427 len1, p->one->path,
4428 len2, p->two->path,
4429 len1, p->one->path,
4430 len2, p->two->path);
4431 git_SHA1_Update(&ctx, buffer, len1);
4432
4433 if (diff_filespec_is_binary(p->one) ||
4434 diff_filespec_is_binary(p->two)) {
4435 git_SHA1_Update(&ctx, sha1_to_hex(p->one->sha1), 40);
4436 git_SHA1_Update(&ctx, sha1_to_hex(p->two->sha1), 40);
4437 continue;
4438 }
4439
4440 xpp.flags = 0;
4441 xecfg.ctxlen = 3;
4442 xecfg.flags = 0;
4443 if (xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4444 &xpp, &xecfg))
4445 return error("unable to generate patch-id diff for %s",
4446 p->one->path);
4447 }
4448
4449 git_SHA1_Final(sha1, &ctx);
4450 return 0;
4451 }
4452
4453 int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
4454 {
4455 struct diff_queue_struct *q = &diff_queued_diff;
4456 int i;
4457 int result = diff_get_patch_id(options, sha1);
4458
4459 for (i = 0; i < q->nr; i++)
4460 diff_free_filepair(q->queue[i]);
4461
4462 free(q->queue);
4463 DIFF_QUEUE_CLEAR(q);
4464
4465 return result;
4466 }
4467
4468 static int is_summary_empty(const struct diff_queue_struct *q)
4469 {
4470 int i;
4471
4472 for (i = 0; i < q->nr; i++) {
4473 const struct diff_filepair *p = q->queue[i];
4474
4475 switch (p->status) {
4476 case DIFF_STATUS_DELETED:
4477 case DIFF_STATUS_ADDED:
4478 case DIFF_STATUS_COPIED:
4479 case DIFF_STATUS_RENAMED:
4480 return 0;
4481 default:
4482 if (p->score)
4483 return 0;
4484 if (p->one->mode && p->two->mode &&
4485 p->one->mode != p->two->mode)
4486 return 0;
4487 break;
4488 }
4489 }
4490 return 1;
4491 }
4492
4493 static const char rename_limit_warning[] =
4494 "inexact rename detection was skipped due to too many files.";
4495
4496 static const char degrade_cc_to_c_warning[] =
4497 "only found copies from modified paths due to too many files.";
4498
4499 static const char rename_limit_advice[] =
4500 "you may want to set your %s variable to at least "
4501 "%d and retry the command.";
4502
4503 void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4504 {
4505 if (degraded_cc)
4506 warning(degrade_cc_to_c_warning);
4507 else if (needed)
4508 warning(rename_limit_warning);
4509 else
4510 return;
4511 if (0 < needed && needed < 32767)
4512 warning(rename_limit_advice, varname, needed);
4513 }
4514
4515 void diff_flush(struct diff_options *options)
4516 {
4517 struct diff_queue_struct *q = &diff_queued_diff;
4518 int i, output_format = options->output_format;
4519 int separator = 0;
4520 int dirstat_by_line = 0;
4521
4522 /*
4523 * Order: raw, stat, summary, patch
4524 * or: name/name-status/checkdiff (other bits clear)
4525 */
4526 if (!q->nr)
4527 goto free_queue;
4528
4529 if (output_format & (DIFF_FORMAT_RAW |
4530 DIFF_FORMAT_NAME |
4531 DIFF_FORMAT_NAME_STATUS |
4532 DIFF_FORMAT_CHECKDIFF)) {
4533 for (i = 0; i < q->nr; i++) {
4534 struct diff_filepair *p = q->queue[i];
4535 if (check_pair_status(p))
4536 flush_one_pair(p, options);
4537 }
4538 separator++;
4539 }
4540
4541 if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4542 dirstat_by_line = 1;
4543
4544 if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4545 dirstat_by_line) {
4546 struct diffstat_t diffstat;
4547
4548 memset(&diffstat, 0, sizeof(struct diffstat_t));
4549 for (i = 0; i < q->nr; i++) {
4550 struct diff_filepair *p = q->queue[i];
4551 if (check_pair_status(p))
4552 diff_flush_stat(p, options, &diffstat);
4553 }
4554 if (output_format & DIFF_FORMAT_NUMSTAT)
4555 show_numstat(&diffstat, options);
4556 if (output_format & DIFF_FORMAT_DIFFSTAT)
4557 show_stats(&diffstat, options);
4558 if (output_format & DIFF_FORMAT_SHORTSTAT)
4559 show_shortstats(&diffstat, options);
4560 if (output_format & DIFF_FORMAT_DIRSTAT && dirstat_by_line)
4561 show_dirstat_by_line(&diffstat, options);
4562 free_diffstat_info(&diffstat);
4563 separator++;
4564 }
4565 if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4566 show_dirstat(options);
4567
4568 if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4569 for (i = 0; i < q->nr; i++) {
4570 diff_summary(options, q->queue[i]);
4571 }
4572 separator++;
4573 }
4574
4575 if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4576 DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4577 DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4578 /*
4579 * run diff_flush_patch for the exit status. setting
4580 * options->file to /dev/null should be safe, because we
4581 * aren't supposed to produce any output anyway.
4582 */
4583 if (options->close_file)
4584 fclose(options->file);
4585 options->file = fopen("/dev/null", "w");
4586 if (!options->file)
4587 die_errno("Could not open /dev/null");
4588 options->close_file = 1;
4589 for (i = 0; i < q->nr; i++) {
4590 struct diff_filepair *p = q->queue[i];
4591 if (check_pair_status(p))
4592 diff_flush_patch(p, options);
4593 if (options->found_changes)
4594 break;
4595 }
4596 }
4597
4598 if (output_format & DIFF_FORMAT_PATCH) {
4599 if (separator) {
4600 fprintf(options->file, "%s%c",
4601 diff_line_prefix(options),
4602 options->line_termination);
4603 if (options->stat_sep) {
4604 /* attach patch instead of inline */
4605 fputs(options->stat_sep, options->file);
4606 }
4607 }
4608
4609 for (i = 0; i < q->nr; i++) {
4610 struct diff_filepair *p = q->queue[i];
4611 if (check_pair_status(p))
4612 diff_flush_patch(p, options);
4613 }
4614 }
4615
4616 if (output_format & DIFF_FORMAT_CALLBACK)
4617 options->format_callback(q, options, options->format_callback_data);
4618
4619 for (i = 0; i < q->nr; i++)
4620 diff_free_filepair(q->queue[i]);
4621 free_queue:
4622 free(q->queue);
4623 DIFF_QUEUE_CLEAR(q);
4624 if (options->close_file)
4625 fclose(options->file);
4626
4627 /*
4628 * Report the content-level differences with HAS_CHANGES;
4629 * diff_addremove/diff_change does not set the bit when
4630 * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4631 */
4632 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4633 if (options->found_changes)
4634 DIFF_OPT_SET(options, HAS_CHANGES);
4635 else
4636 DIFF_OPT_CLR(options, HAS_CHANGES);
4637 }
4638 }
4639
4640 static int match_filter(const struct diff_options *options, const struct diff_filepair *p)
4641 {
4642 return (((p->status == DIFF_STATUS_MODIFIED) &&
4643 ((p->score &&
4644 filter_bit_tst(DIFF_STATUS_FILTER_BROKEN, options)) ||
4645 (!p->score &&
4646 filter_bit_tst(DIFF_STATUS_MODIFIED, options)))) ||
4647 ((p->status != DIFF_STATUS_MODIFIED) &&
4648 filter_bit_tst(p->status, options)));
4649 }
4650
4651 static void diffcore_apply_filter(struct diff_options *options)
4652 {
4653 int i;
4654 struct diff_queue_struct *q = &diff_queued_diff;
4655 struct diff_queue_struct outq;
4656
4657 DIFF_QUEUE_CLEAR(&outq);
4658
4659 if (!options->filter)
4660 return;
4661
4662 if (filter_bit_tst(DIFF_STATUS_FILTER_AON, options)) {
4663 int found;
4664 for (i = found = 0; !found && i < q->nr; i++) {
4665 if (match_filter(options, q->queue[i]))
4666 found++;
4667 }
4668 if (found)
4669 return;
4670
4671 /* otherwise we will clear the whole queue
4672 * by copying the empty outq at the end of this
4673 * function, but first clear the current entries
4674 * in the queue.
4675 */
4676 for (i = 0; i < q->nr; i++)
4677 diff_free_filepair(q->queue[i]);
4678 }
4679 else {
4680 /* Only the matching ones */
4681 for (i = 0; i < q->nr; i++) {
4682 struct diff_filepair *p = q->queue[i];
4683 if (match_filter(options, p))
4684 diff_q(&outq, p);
4685 else
4686 diff_free_filepair(p);
4687 }
4688 }
4689 free(q->queue);
4690 *q = outq;
4691 }
4692
4693 /* Check whether two filespecs with the same mode and size are identical */
4694 static int diff_filespec_is_identical(struct diff_filespec *one,
4695 struct diff_filespec *two)
4696 {
4697 if (S_ISGITLINK(one->mode))
4698 return 0;
4699 if (diff_populate_filespec(one, 0))
4700 return 0;
4701 if (diff_populate_filespec(two, 0))
4702 return 0;
4703 return !memcmp(one->data, two->data, one->size);
4704 }
4705
4706 static int diff_filespec_check_stat_unmatch(struct diff_filepair *p)
4707 {
4708 if (p->done_skip_stat_unmatch)
4709 return p->skip_stat_unmatch_result;
4710
4711 p->done_skip_stat_unmatch = 1;
4712 p->skip_stat_unmatch_result = 0;
4713 /*
4714 * 1. Entries that come from stat info dirtiness
4715 * always have both sides (iow, not create/delete),
4716 * one side of the object name is unknown, with
4717 * the same mode and size. Keep the ones that
4718 * do not match these criteria. They have real
4719 * differences.
4720 *
4721 * 2. At this point, the file is known to be modified,
4722 * with the same mode and size, and the object
4723 * name of one side is unknown. Need to inspect
4724 * the identical contents.
4725 */
4726 if (!DIFF_FILE_VALID(p->one) || /* (1) */
4727 !DIFF_FILE_VALID(p->two) ||
4728 (p->one->sha1_valid && p->two->sha1_valid) ||
4729 (p->one->mode != p->two->mode) ||
4730 diff_populate_filespec(p->one, CHECK_SIZE_ONLY) ||
4731 diff_populate_filespec(p->two, CHECK_SIZE_ONLY) ||
4732 (p->one->size != p->two->size) ||
4733 !diff_filespec_is_identical(p->one, p->two)) /* (2) */
4734 p->skip_stat_unmatch_result = 1;
4735 return p->skip_stat_unmatch_result;
4736 }
4737
4738 static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
4739 {
4740 int i;
4741 struct diff_queue_struct *q = &diff_queued_diff;
4742 struct diff_queue_struct outq;
4743 DIFF_QUEUE_CLEAR(&outq);
4744
4745 for (i = 0; i < q->nr; i++) {
4746 struct diff_filepair *p = q->queue[i];
4747
4748 if (diff_filespec_check_stat_unmatch(p))
4749 diff_q(&outq, p);
4750 else {
4751 /*
4752 * The caller can subtract 1 from skip_stat_unmatch
4753 * to determine how many paths were dirty only
4754 * due to stat info mismatch.
4755 */
4756 if (!DIFF_OPT_TST(diffopt, NO_INDEX))
4757 diffopt->skip_stat_unmatch++;
4758 diff_free_filepair(p);
4759 }
4760 }
4761 free(q->queue);
4762 *q = outq;
4763 }
4764
4765 static int diffnamecmp(const void *a_, const void *b_)
4766 {
4767 const struct diff_filepair *a = *((const struct diff_filepair **)a_);
4768 const struct diff_filepair *b = *((const struct diff_filepair **)b_);
4769 const char *name_a, *name_b;
4770
4771 name_a = a->one ? a->one->path : a->two->path;
4772 name_b = b->one ? b->one->path : b->two->path;
4773 return strcmp(name_a, name_b);
4774 }
4775
4776 void diffcore_fix_diff_index(struct diff_options *options)
4777 {
4778 struct diff_queue_struct *q = &diff_queued_diff;
4779 qsort(q->queue, q->nr, sizeof(q->queue[0]), diffnamecmp);
4780 }
4781
4782 void diffcore_std(struct diff_options *options)
4783 {
4784 /* NOTE please keep the following in sync with diff_tree_combined() */
4785 if (options->skip_stat_unmatch)
4786 diffcore_skip_stat_unmatch(options);
4787 if (!options->found_follow) {
4788 /* See try_to_follow_renames() in tree-diff.c */
4789 if (options->break_opt != -1)
4790 diffcore_break(options->break_opt);
4791 if (options->detect_rename)
4792 diffcore_rename(options);
4793 if (options->break_opt != -1)
4794 diffcore_merge_broken();
4795 }
4796 if (options->pickaxe)
4797 diffcore_pickaxe(options);
4798 if (options->orderfile)
4799 diffcore_order(options->orderfile);
4800 if (!options->found_follow)
4801 /* See try_to_follow_renames() in tree-diff.c */
4802 diff_resolve_rename_copy();
4803 diffcore_apply_filter(options);
4804
4805 if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4806 DIFF_OPT_SET(options, HAS_CHANGES);
4807 else
4808 DIFF_OPT_CLR(options, HAS_CHANGES);
4809
4810 options->found_follow = 0;
4811 }
4812
4813 int diff_result_code(struct diff_options *opt, int status)
4814 {
4815 int result = 0;
4816
4817 diff_warn_rename_limit("diff.renameLimit",
4818 opt->needed_rename_limit,
4819 opt->degraded_cc_to_c);
4820 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4821 !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
4822 return status;
4823 if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4824 DIFF_OPT_TST(opt, HAS_CHANGES))
4825 result |= 01;
4826 if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
4827 DIFF_OPT_TST(opt, CHECK_FAILED))
4828 result |= 02;
4829 return result;
4830 }
4831
4832 int diff_can_quit_early(struct diff_options *opt)
4833 {
4834 return (DIFF_OPT_TST(opt, QUICK) &&
4835 !opt->filter &&
4836 DIFF_OPT_TST(opt, HAS_CHANGES));
4837 }
4838
4839 /*
4840 * Shall changes to this submodule be ignored?
4841 *
4842 * Submodule changes can be configured to be ignored separately for each path,
4843 * but that configuration can be overridden from the command line.
4844 */
4845 static int is_submodule_ignored(const char *path, struct diff_options *options)
4846 {
4847 int ignored = 0;
4848 unsigned orig_flags = options->flags;
4849 if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
4850 set_diffopt_flags_from_submodule_config(options, path);
4851 if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
4852 ignored = 1;
4853 options->flags = orig_flags;
4854 return ignored;
4855 }
4856
4857 void diff_addremove(struct diff_options *options,
4858 int addremove, unsigned mode,
4859 const unsigned char *sha1,
4860 int sha1_valid,
4861 const char *concatpath, unsigned dirty_submodule)
4862 {
4863 struct diff_filespec *one, *two;
4864
4865 if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
4866 return;
4867
4868 /* This may look odd, but it is a preparation for
4869 * feeding "there are unchanged files which should
4870 * not produce diffs, but when you are doing copy
4871 * detection you would need them, so here they are"
4872 * entries to the diff-core. They will be prefixed
4873 * with something like '=' or '*' (I haven't decided
4874 * which but should not make any difference).
4875 * Feeding the same new and old to diff_change()
4876 * also has the same effect.
4877 * Before the final output happens, they are pruned after
4878 * merged into rename/copy pairs as appropriate.
4879 */
4880 if (DIFF_OPT_TST(options, REVERSE_DIFF))
4881 addremove = (addremove == '+' ? '-' :
4882 addremove == '-' ? '+' : addremove);
4883
4884 if (options->prefix &&
4885 strncmp(concatpath, options->prefix, options->prefix_length))
4886 return;
4887
4888 one = alloc_filespec(concatpath);
4889 two = alloc_filespec(concatpath);
4890
4891 if (addremove != '+')
4892 fill_filespec(one, sha1, sha1_valid, mode);
4893 if (addremove != '-') {
4894 fill_filespec(two, sha1, sha1_valid, mode);
4895 two->dirty_submodule = dirty_submodule;
4896 }
4897
4898 diff_queue(&diff_queued_diff, one, two);
4899 if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4900 DIFF_OPT_SET(options, HAS_CHANGES);
4901 }
4902
4903 void diff_change(struct diff_options *options,
4904 unsigned old_mode, unsigned new_mode,
4905 const unsigned char *old_sha1,
4906 const unsigned char *new_sha1,
4907 int old_sha1_valid, int new_sha1_valid,
4908 const char *concatpath,
4909 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
4910 {
4911 struct diff_filespec *one, *two;
4912 struct diff_filepair *p;
4913
4914 if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
4915 is_submodule_ignored(concatpath, options))
4916 return;
4917
4918 if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
4919 unsigned tmp;
4920 const unsigned char *tmp_c;
4921 tmp = old_mode; old_mode = new_mode; new_mode = tmp;
4922 tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
4923 tmp = old_sha1_valid; old_sha1_valid = new_sha1_valid;
4924 new_sha1_valid = tmp;
4925 tmp = old_dirty_submodule; old_dirty_submodule = new_dirty_submodule;
4926 new_dirty_submodule = tmp;
4927 }
4928
4929 if (options->prefix &&
4930 strncmp(concatpath, options->prefix, options->prefix_length))
4931 return;
4932
4933 one = alloc_filespec(concatpath);
4934 two = alloc_filespec(concatpath);
4935 fill_filespec(one, old_sha1, old_sha1_valid, old_mode);
4936 fill_filespec(two, new_sha1, new_sha1_valid, new_mode);
4937 one->dirty_submodule = old_dirty_submodule;
4938 two->dirty_submodule = new_dirty_submodule;
4939 p = diff_queue(&diff_queued_diff, one, two);
4940
4941 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4942 return;
4943
4944 if (DIFF_OPT_TST(options, QUICK) && options->skip_stat_unmatch &&
4945 !diff_filespec_check_stat_unmatch(p))
4946 return;
4947
4948 DIFF_OPT_SET(options, HAS_CHANGES);
4949 }
4950
4951 struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
4952 {
4953 struct diff_filepair *pair;
4954 struct diff_filespec *one, *two;
4955
4956 if (options->prefix &&
4957 strncmp(path, options->prefix, options->prefix_length))
4958 return NULL;
4959
4960 one = alloc_filespec(path);
4961 two = alloc_filespec(path);
4962 pair = diff_queue(&diff_queued_diff, one, two);
4963 pair->is_unmerged = 1;
4964 return pair;
4965 }
4966
4967 static char *run_textconv(const char *pgm, struct diff_filespec *spec,
4968 size_t *outsize)
4969 {
4970 struct diff_tempfile *temp;
4971 const char *argv[3];
4972 const char **arg = argv;
4973 struct child_process child = CHILD_PROCESS_INIT;
4974 struct strbuf buf = STRBUF_INIT;
4975 int err = 0;
4976
4977 temp = prepare_temp_file(spec->path, spec);
4978 *arg++ = pgm;
4979 *arg++ = temp->name;
4980 *arg = NULL;
4981
4982 child.use_shell = 1;
4983 child.argv = argv;
4984 child.out = -1;
4985 if (start_command(&child)) {
4986 remove_tempfile();
4987 return NULL;
4988 }
4989
4990 if (strbuf_read(&buf, child.out, 0) < 0)
4991 err = error("error reading from textconv command '%s'", pgm);
4992 close(child.out);
4993
4994 if (finish_command(&child) || err) {
4995 strbuf_release(&buf);
4996 remove_tempfile();
4997 return NULL;
4998 }
4999 remove_tempfile();
5000
5001 return strbuf_detach(&buf, outsize);
5002 }
5003
5004 size_t fill_textconv(struct userdiff_driver *driver,
5005 struct diff_filespec *df,
5006 char **outbuf)
5007 {
5008 size_t size;
5009
5010 if (!driver || !driver->textconv) {
5011 if (!DIFF_FILE_VALID(df)) {
5012 *outbuf = "";
5013 return 0;
5014 }
5015 if (diff_populate_filespec(df, 0))
5016 die("unable to read files to diff");
5017 *outbuf = df->data;
5018 return df->size;
5019 }
5020
5021 if (driver->textconv_cache && df->sha1_valid) {
5022 *outbuf = notes_cache_get(driver->textconv_cache, df->sha1,
5023 &size);
5024 if (*outbuf)
5025 return size;
5026 }
5027
5028 *outbuf = run_textconv(driver->textconv, df, &size);
5029 if (!*outbuf)
5030 die("unable to read files to diff");
5031
5032 if (driver->textconv_cache && df->sha1_valid) {
5033 /* ignore errors, as we might be in a readonly repository */
5034 notes_cache_put(driver->textconv_cache, df->sha1, *outbuf,
5035 size);
5036 /*
5037 * we could save up changes and flush them all at the end,
5038 * but we would need an extra call after all diffing is done.
5039 * Since generating a cache entry is the slow path anyway,
5040 * this extra overhead probably isn't a big deal.
5041 */
5042 notes_cache_write(driver->textconv_cache);
5043 }
5044
5045 return size;
5046 }
5047
5048 void setup_diff_pager(struct diff_options *opt)
5049 {
5050 /*
5051 * If the user asked for our exit code, then either they want --quiet
5052 * or --exit-code. We should definitely not bother with a pager in the
5053 * former case, as we will generate no output. Since we still properly
5054 * report our exit code even when a pager is run, we _could_ run a
5055 * pager with --exit-code. But since we have not done so historically,
5056 * and because it is easy to find people oneline advising "git diff
5057 * --exit-code" in hooks and other scripts, we do not do so.
5058 */
5059 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5060 check_pager_config("diff") != 0)
5061 setup_pager();
5062 }
The Best VCS Around