]> git.ipfire.org Git - thirdparty/git.git/blob - diff.c
diff: improve positioning of add/delete blocks in diffs
[thirdparty/git.git] / diff.c
1 /*
2 * Copyright (C) 2005 Junio C Hamano
3 */
4 #include "cache.h"
5 #include "tempfile.h"
6 #include "quote.h"
7 #include "diff.h"
8 #include "diffcore.h"
9 #include "delta.h"
10 #include "xdiff-interface.h"
11 #include "color.h"
12 #include "attr.h"
13 #include "run-command.h"
14 #include "utf8.h"
15 #include "userdiff.h"
16 #include "submodule-config.h"
17 #include "submodule.h"
18 #include "ll-merge.h"
19 #include "string-list.h"
20 #include "argv-array.h"
21
22 #ifdef NO_FAST_WORKING_DIRECTORY
23 #define FAST_WORKING_DIRECTORY 0
24 #else
25 #define FAST_WORKING_DIRECTORY 1
26 #endif
27
28 static int diff_detect_rename_default;
29 static int diff_indent_heuristic; /* experimental */
30 static int diff_compaction_heuristic; /* experimental */
31 static int diff_rename_limit_default = 400;
32 static int diff_suppress_blank_empty;
33 static int diff_use_color_default = -1;
34 static int diff_context_default = 3;
35 static const char *diff_word_regex_cfg;
36 static const char *external_diff_cmd_cfg;
37 static const char *diff_order_file_cfg;
38 int diff_auto_refresh_index = 1;
39 static int diff_mnemonic_prefix;
40 static int diff_no_prefix;
41 static int diff_stat_graph_width;
42 static int diff_dirstat_permille_default = 30;
43 static struct diff_options default_diff_options;
44 static long diff_algorithm;
45
46 static char diff_colors[][COLOR_MAXLEN] = {
47 GIT_COLOR_RESET,
48 GIT_COLOR_NORMAL, /* CONTEXT */
49 GIT_COLOR_BOLD, /* METAINFO */
50 GIT_COLOR_CYAN, /* FRAGINFO */
51 GIT_COLOR_RED, /* OLD */
52 GIT_COLOR_GREEN, /* NEW */
53 GIT_COLOR_YELLOW, /* COMMIT */
54 GIT_COLOR_BG_RED, /* WHITESPACE */
55 GIT_COLOR_NORMAL, /* FUNCINFO */
56 };
57
58 static int parse_diff_color_slot(const char *var)
59 {
60 if (!strcasecmp(var, "context") || !strcasecmp(var, "plain"))
61 return DIFF_CONTEXT;
62 if (!strcasecmp(var, "meta"))
63 return DIFF_METAINFO;
64 if (!strcasecmp(var, "frag"))
65 return DIFF_FRAGINFO;
66 if (!strcasecmp(var, "old"))
67 return DIFF_FILE_OLD;
68 if (!strcasecmp(var, "new"))
69 return DIFF_FILE_NEW;
70 if (!strcasecmp(var, "commit"))
71 return DIFF_COMMIT;
72 if (!strcasecmp(var, "whitespace"))
73 return DIFF_WHITESPACE;
74 if (!strcasecmp(var, "func"))
75 return DIFF_FUNCINFO;
76 return -1;
77 }
78
79 static int parse_dirstat_params(struct diff_options *options, const char *params_string,
80 struct strbuf *errmsg)
81 {
82 char *params_copy = xstrdup(params_string);
83 struct string_list params = STRING_LIST_INIT_NODUP;
84 int ret = 0;
85 int i;
86
87 if (*params_copy)
88 string_list_split_in_place(&params, params_copy, ',', -1);
89 for (i = 0; i < params.nr; i++) {
90 const char *p = params.items[i].string;
91 if (!strcmp(p, "changes")) {
92 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
93 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
94 } else if (!strcmp(p, "lines")) {
95 DIFF_OPT_SET(options, DIRSTAT_BY_LINE);
96 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
97 } else if (!strcmp(p, "files")) {
98 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
99 DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
100 } else if (!strcmp(p, "noncumulative")) {
101 DIFF_OPT_CLR(options, DIRSTAT_CUMULATIVE);
102 } else if (!strcmp(p, "cumulative")) {
103 DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
104 } else if (isdigit(*p)) {
105 char *end;
106 int permille = strtoul(p, &end, 10) * 10;
107 if (*end == '.' && isdigit(*++end)) {
108 /* only use first digit */
109 permille += *end - '0';
110 /* .. and ignore any further digits */
111 while (isdigit(*++end))
112 ; /* nothing */
113 }
114 if (!*end)
115 options->dirstat_permille = permille;
116 else {
117 strbuf_addf(errmsg, _(" Failed to parse dirstat cut-off percentage '%s'\n"),
118 p);
119 ret++;
120 }
121 } else {
122 strbuf_addf(errmsg, _(" Unknown dirstat parameter '%s'\n"), p);
123 ret++;
124 }
125
126 }
127 string_list_clear(&params, 0);
128 free(params_copy);
129 return ret;
130 }
131
132 static int parse_submodule_params(struct diff_options *options, const char *value)
133 {
134 if (!strcmp(value, "log"))
135 DIFF_OPT_SET(options, SUBMODULE_LOG);
136 else if (!strcmp(value, "short"))
137 DIFF_OPT_CLR(options, SUBMODULE_LOG);
138 else
139 return -1;
140 return 0;
141 }
142
143 static int git_config_rename(const char *var, const char *value)
144 {
145 if (!value)
146 return DIFF_DETECT_RENAME;
147 if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
148 return DIFF_DETECT_COPY;
149 return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
150 }
151
152 long parse_algorithm_value(const char *value)
153 {
154 if (!value)
155 return -1;
156 else if (!strcasecmp(value, "myers") || !strcasecmp(value, "default"))
157 return 0;
158 else if (!strcasecmp(value, "minimal"))
159 return XDF_NEED_MINIMAL;
160 else if (!strcasecmp(value, "patience"))
161 return XDF_PATIENCE_DIFF;
162 else if (!strcasecmp(value, "histogram"))
163 return XDF_HISTOGRAM_DIFF;
164 return -1;
165 }
166
167 /*
168 * These are to give UI layer defaults.
169 * The core-level commands such as git-diff-files should
170 * never be affected by the setting of diff.renames
171 * the user happens to have in the configuration file.
172 */
173 void init_diff_ui_defaults(void)
174 {
175 diff_detect_rename_default = 1;
176 }
177
178 int git_diff_ui_config(const char *var, const char *value, void *cb)
179 {
180 if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
181 diff_use_color_default = git_config_colorbool(var, value);
182 return 0;
183 }
184 if (!strcmp(var, "diff.context")) {
185 diff_context_default = git_config_int(var, value);
186 if (diff_context_default < 0)
187 return -1;
188 return 0;
189 }
190 if (!strcmp(var, "diff.renames")) {
191 diff_detect_rename_default = git_config_rename(var, value);
192 return 0;
193 }
194 if (!strcmp(var, "diff.indentheuristic")) {
195 diff_indent_heuristic = git_config_bool(var, value);
196 if (diff_indent_heuristic)
197 diff_compaction_heuristic = 0;
198 return 0;
199 }
200 if (!strcmp(var, "diff.compactionheuristic")) {
201 diff_compaction_heuristic = git_config_bool(var, value);
202 if (diff_compaction_heuristic)
203 diff_indent_heuristic = 0;
204 return 0;
205 }
206 if (!strcmp(var, "diff.autorefreshindex")) {
207 diff_auto_refresh_index = git_config_bool(var, value);
208 return 0;
209 }
210 if (!strcmp(var, "diff.mnemonicprefix")) {
211 diff_mnemonic_prefix = git_config_bool(var, value);
212 return 0;
213 }
214 if (!strcmp(var, "diff.noprefix")) {
215 diff_no_prefix = git_config_bool(var, value);
216 return 0;
217 }
218 if (!strcmp(var, "diff.statgraphwidth")) {
219 diff_stat_graph_width = git_config_int(var, value);
220 return 0;
221 }
222 if (!strcmp(var, "diff.external"))
223 return git_config_string(&external_diff_cmd_cfg, var, value);
224 if (!strcmp(var, "diff.wordregex"))
225 return git_config_string(&diff_word_regex_cfg, var, value);
226 if (!strcmp(var, "diff.orderfile"))
227 return git_config_pathname(&diff_order_file_cfg, var, value);
228
229 if (!strcmp(var, "diff.ignoresubmodules"))
230 handle_ignore_submodules_arg(&default_diff_options, value);
231
232 if (!strcmp(var, "diff.submodule")) {
233 if (parse_submodule_params(&default_diff_options, value))
234 warning(_("Unknown value for 'diff.submodule' config variable: '%s'"),
235 value);
236 return 0;
237 }
238
239 if (!strcmp(var, "diff.algorithm")) {
240 diff_algorithm = parse_algorithm_value(value);
241 if (diff_algorithm < 0)
242 return -1;
243 return 0;
244 }
245
246 if (git_color_config(var, value, cb) < 0)
247 return -1;
248
249 return git_diff_basic_config(var, value, cb);
250 }
251
252 int git_diff_basic_config(const char *var, const char *value, void *cb)
253 {
254 const char *name;
255
256 if (!strcmp(var, "diff.renamelimit")) {
257 diff_rename_limit_default = git_config_int(var, value);
258 return 0;
259 }
260
261 if (userdiff_config(var, value) < 0)
262 return -1;
263
264 if (skip_prefix(var, "diff.color.", &name) ||
265 skip_prefix(var, "color.diff.", &name)) {
266 int slot = parse_diff_color_slot(name);
267 if (slot < 0)
268 return 0;
269 if (!value)
270 return config_error_nonbool(var);
271 return color_parse(value, diff_colors[slot]);
272 }
273
274 /* like GNU diff's --suppress-blank-empty option */
275 if (!strcmp(var, "diff.suppressblankempty") ||
276 /* for backwards compatibility */
277 !strcmp(var, "diff.suppress-blank-empty")) {
278 diff_suppress_blank_empty = git_config_bool(var, value);
279 return 0;
280 }
281
282 if (!strcmp(var, "diff.dirstat")) {
283 struct strbuf errmsg = STRBUF_INIT;
284 default_diff_options.dirstat_permille = diff_dirstat_permille_default;
285 if (parse_dirstat_params(&default_diff_options, value, &errmsg))
286 warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
287 errmsg.buf);
288 strbuf_release(&errmsg);
289 diff_dirstat_permille_default = default_diff_options.dirstat_permille;
290 return 0;
291 }
292
293 if (starts_with(var, "submodule."))
294 return parse_submodule_config_option(var, value);
295
296 return git_default_config(var, value, cb);
297 }
298
299 static char *quote_two(const char *one, const char *two)
300 {
301 int need_one = quote_c_style(one, NULL, NULL, 1);
302 int need_two = quote_c_style(two, NULL, NULL, 1);
303 struct strbuf res = STRBUF_INIT;
304
305 if (need_one + need_two) {
306 strbuf_addch(&res, '"');
307 quote_c_style(one, &res, NULL, 1);
308 quote_c_style(two, &res, NULL, 1);
309 strbuf_addch(&res, '"');
310 } else {
311 strbuf_addstr(&res, one);
312 strbuf_addstr(&res, two);
313 }
314 return strbuf_detach(&res, NULL);
315 }
316
317 static const char *external_diff(void)
318 {
319 static const char *external_diff_cmd = NULL;
320 static int done_preparing = 0;
321
322 if (done_preparing)
323 return external_diff_cmd;
324 external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
325 if (!external_diff_cmd)
326 external_diff_cmd = external_diff_cmd_cfg;
327 done_preparing = 1;
328 return external_diff_cmd;
329 }
330
331 /*
332 * Keep track of files used for diffing. Sometimes such an entry
333 * refers to a temporary file, sometimes to an existing file, and
334 * sometimes to "/dev/null".
335 */
336 static struct diff_tempfile {
337 /*
338 * filename external diff should read from, or NULL if this
339 * entry is currently not in use:
340 */
341 const char *name;
342
343 char hex[GIT_SHA1_HEXSZ + 1];
344 char mode[10];
345
346 /*
347 * If this diff_tempfile instance refers to a temporary file,
348 * this tempfile object is used to manage its lifetime.
349 */
350 struct tempfile tempfile;
351 } diff_temp[2];
352
353 typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
354
355 struct emit_callback {
356 int color_diff;
357 unsigned ws_rule;
358 int blank_at_eof_in_preimage;
359 int blank_at_eof_in_postimage;
360 int lno_in_preimage;
361 int lno_in_postimage;
362 sane_truncate_fn truncate;
363 const char **label_path;
364 struct diff_words_data *diff_words;
365 struct diff_options *opt;
366 int *found_changesp;
367 struct strbuf *header;
368 };
369
370 static int count_lines(const char *data, int size)
371 {
372 int count, ch, completely_empty = 1, nl_just_seen = 0;
373 count = 0;
374 while (0 < size--) {
375 ch = *data++;
376 if (ch == '\n') {
377 count++;
378 nl_just_seen = 1;
379 completely_empty = 0;
380 }
381 else {
382 nl_just_seen = 0;
383 completely_empty = 0;
384 }
385 }
386 if (completely_empty)
387 return 0;
388 if (!nl_just_seen)
389 count++; /* no trailing newline */
390 return count;
391 }
392
393 static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
394 {
395 if (!DIFF_FILE_VALID(one)) {
396 mf->ptr = (char *)""; /* does not matter */
397 mf->size = 0;
398 return 0;
399 }
400 else if (diff_populate_filespec(one, 0))
401 return -1;
402
403 mf->ptr = one->data;
404 mf->size = one->size;
405 return 0;
406 }
407
408 /* like fill_mmfile, but only for size, so we can avoid retrieving blob */
409 static unsigned long diff_filespec_size(struct diff_filespec *one)
410 {
411 if (!DIFF_FILE_VALID(one))
412 return 0;
413 diff_populate_filespec(one, CHECK_SIZE_ONLY);
414 return one->size;
415 }
416
417 static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
418 {
419 char *ptr = mf->ptr;
420 long size = mf->size;
421 int cnt = 0;
422
423 if (!size)
424 return cnt;
425 ptr += size - 1; /* pointing at the very end */
426 if (*ptr != '\n')
427 ; /* incomplete line */
428 else
429 ptr--; /* skip the last LF */
430 while (mf->ptr < ptr) {
431 char *prev_eol;
432 for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
433 if (*prev_eol == '\n')
434 break;
435 if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
436 break;
437 cnt++;
438 ptr = prev_eol - 1;
439 }
440 return cnt;
441 }
442
443 static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
444 struct emit_callback *ecbdata)
445 {
446 int l1, l2, at;
447 unsigned ws_rule = ecbdata->ws_rule;
448 l1 = count_trailing_blank(mf1, ws_rule);
449 l2 = count_trailing_blank(mf2, ws_rule);
450 if (l2 <= l1) {
451 ecbdata->blank_at_eof_in_preimage = 0;
452 ecbdata->blank_at_eof_in_postimage = 0;
453 return;
454 }
455 at = count_lines(mf1->ptr, mf1->size);
456 ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
457
458 at = count_lines(mf2->ptr, mf2->size);
459 ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
460 }
461
462 static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
463 int first, const char *line, int len)
464 {
465 int has_trailing_newline, has_trailing_carriage_return;
466 int nofirst;
467 FILE *file = o->file;
468
469 fputs(diff_line_prefix(o), file);
470
471 if (len == 0) {
472 has_trailing_newline = (first == '\n');
473 has_trailing_carriage_return = (!has_trailing_newline &&
474 (first == '\r'));
475 nofirst = has_trailing_newline || has_trailing_carriage_return;
476 } else {
477 has_trailing_newline = (len > 0 && line[len-1] == '\n');
478 if (has_trailing_newline)
479 len--;
480 has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
481 if (has_trailing_carriage_return)
482 len--;
483 nofirst = 0;
484 }
485
486 if (len || !nofirst) {
487 fputs(set, file);
488 if (!nofirst)
489 fputc(first, file);
490 fwrite(line, len, 1, file);
491 fputs(reset, file);
492 }
493 if (has_trailing_carriage_return)
494 fputc('\r', file);
495 if (has_trailing_newline)
496 fputc('\n', file);
497 }
498
499 static void emit_line(struct diff_options *o, const char *set, const char *reset,
500 const char *line, int len)
501 {
502 emit_line_0(o, set, reset, line[0], line+1, len-1);
503 }
504
505 static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
506 {
507 if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
508 ecbdata->blank_at_eof_in_preimage &&
509 ecbdata->blank_at_eof_in_postimage &&
510 ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
511 ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
512 return 0;
513 return ws_blank_line(line, len, ecbdata->ws_rule);
514 }
515
516 static void emit_line_checked(const char *reset,
517 struct emit_callback *ecbdata,
518 const char *line, int len,
519 enum color_diff color,
520 unsigned ws_error_highlight,
521 char sign)
522 {
523 const char *set = diff_get_color(ecbdata->color_diff, color);
524 const char *ws = NULL;
525
526 if (ecbdata->opt->ws_error_highlight & ws_error_highlight) {
527 ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
528 if (!*ws)
529 ws = NULL;
530 }
531
532 if (!ws)
533 emit_line_0(ecbdata->opt, set, reset, sign, line, len);
534 else if (sign == '+' && new_blank_line_at_eof(ecbdata, line, len))
535 /* Blank line at EOF - paint '+' as well */
536 emit_line_0(ecbdata->opt, ws, reset, sign, line, len);
537 else {
538 /* Emit just the prefix, then the rest. */
539 emit_line_0(ecbdata->opt, set, reset, sign, "", 0);
540 ws_check_emit(line, len, ecbdata->ws_rule,
541 ecbdata->opt->file, set, reset, ws);
542 }
543 }
544
545 static void emit_add_line(const char *reset,
546 struct emit_callback *ecbdata,
547 const char *line, int len)
548 {
549 emit_line_checked(reset, ecbdata, line, len,
550 DIFF_FILE_NEW, WSEH_NEW, '+');
551 }
552
553 static void emit_del_line(const char *reset,
554 struct emit_callback *ecbdata,
555 const char *line, int len)
556 {
557 emit_line_checked(reset, ecbdata, line, len,
558 DIFF_FILE_OLD, WSEH_OLD, '-');
559 }
560
561 static void emit_context_line(const char *reset,
562 struct emit_callback *ecbdata,
563 const char *line, int len)
564 {
565 emit_line_checked(reset, ecbdata, line, len,
566 DIFF_CONTEXT, WSEH_CONTEXT, ' ');
567 }
568
569 static void emit_hunk_header(struct emit_callback *ecbdata,
570 const char *line, int len)
571 {
572 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
573 const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
574 const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
575 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
576 static const char atat[2] = { '@', '@' };
577 const char *cp, *ep;
578 struct strbuf msgbuf = STRBUF_INIT;
579 int org_len = len;
580 int i = 1;
581
582 /*
583 * As a hunk header must begin with "@@ -<old>, +<new> @@",
584 * it always is at least 10 bytes long.
585 */
586 if (len < 10 ||
587 memcmp(line, atat, 2) ||
588 !(ep = memmem(line + 2, len - 2, atat, 2))) {
589 emit_line(ecbdata->opt, context, reset, line, len);
590 return;
591 }
592 ep += 2; /* skip over @@ */
593
594 /* The hunk header in fraginfo color */
595 strbuf_addstr(&msgbuf, frag);
596 strbuf_add(&msgbuf, line, ep - line);
597 strbuf_addstr(&msgbuf, reset);
598
599 /*
600 * trailing "\r\n"
601 */
602 for ( ; i < 3; i++)
603 if (line[len - i] == '\r' || line[len - i] == '\n')
604 len--;
605
606 /* blank before the func header */
607 for (cp = ep; ep - line < len; ep++)
608 if (*ep != ' ' && *ep != '\t')
609 break;
610 if (ep != cp) {
611 strbuf_addstr(&msgbuf, context);
612 strbuf_add(&msgbuf, cp, ep - cp);
613 strbuf_addstr(&msgbuf, reset);
614 }
615
616 if (ep < line + len) {
617 strbuf_addstr(&msgbuf, func);
618 strbuf_add(&msgbuf, ep, line + len - ep);
619 strbuf_addstr(&msgbuf, reset);
620 }
621
622 strbuf_add(&msgbuf, line + len, org_len - len);
623 emit_line(ecbdata->opt, "", "", msgbuf.buf, msgbuf.len);
624 strbuf_release(&msgbuf);
625 }
626
627 static struct diff_tempfile *claim_diff_tempfile(void) {
628 int i;
629 for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
630 if (!diff_temp[i].name)
631 return diff_temp + i;
632 die("BUG: diff is failing to clean up its tempfiles");
633 }
634
635 static void remove_tempfile(void)
636 {
637 int i;
638 for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
639 if (is_tempfile_active(&diff_temp[i].tempfile))
640 delete_tempfile(&diff_temp[i].tempfile);
641 diff_temp[i].name = NULL;
642 }
643 }
644
645 static void print_line_count(FILE *file, int count)
646 {
647 switch (count) {
648 case 0:
649 fprintf(file, "0,0");
650 break;
651 case 1:
652 fprintf(file, "1");
653 break;
654 default:
655 fprintf(file, "1,%d", count);
656 break;
657 }
658 }
659
660 static void emit_rewrite_lines(struct emit_callback *ecb,
661 int prefix, const char *data, int size)
662 {
663 const char *endp = NULL;
664 static const char *nneof = " No newline at end of file\n";
665 const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
666
667 while (0 < size) {
668 int len;
669
670 endp = memchr(data, '\n', size);
671 len = endp ? (endp - data + 1) : size;
672 if (prefix != '+') {
673 ecb->lno_in_preimage++;
674 emit_del_line(reset, ecb, data, len);
675 } else {
676 ecb->lno_in_postimage++;
677 emit_add_line(reset, ecb, data, len);
678 }
679 size -= len;
680 data += len;
681 }
682 if (!endp) {
683 const char *context = diff_get_color(ecb->color_diff,
684 DIFF_CONTEXT);
685 putc('\n', ecb->opt->file);
686 emit_line_0(ecb->opt, context, reset, '\\',
687 nneof, strlen(nneof));
688 }
689 }
690
691 static void emit_rewrite_diff(const char *name_a,
692 const char *name_b,
693 struct diff_filespec *one,
694 struct diff_filespec *two,
695 struct userdiff_driver *textconv_one,
696 struct userdiff_driver *textconv_two,
697 struct diff_options *o)
698 {
699 int lc_a, lc_b;
700 const char *name_a_tab, *name_b_tab;
701 const char *metainfo = diff_get_color(o->use_color, DIFF_METAINFO);
702 const char *fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
703 const char *reset = diff_get_color(o->use_color, DIFF_RESET);
704 static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
705 const char *a_prefix, *b_prefix;
706 char *data_one, *data_two;
707 size_t size_one, size_two;
708 struct emit_callback ecbdata;
709 const char *line_prefix = diff_line_prefix(o);
710
711 if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
712 a_prefix = o->b_prefix;
713 b_prefix = o->a_prefix;
714 } else {
715 a_prefix = o->a_prefix;
716 b_prefix = o->b_prefix;
717 }
718
719 name_a += (*name_a == '/');
720 name_b += (*name_b == '/');
721 name_a_tab = strchr(name_a, ' ') ? "\t" : "";
722 name_b_tab = strchr(name_b, ' ') ? "\t" : "";
723
724 strbuf_reset(&a_name);
725 strbuf_reset(&b_name);
726 quote_two_c_style(&a_name, a_prefix, name_a, 0);
727 quote_two_c_style(&b_name, b_prefix, name_b, 0);
728
729 size_one = fill_textconv(textconv_one, one, &data_one);
730 size_two = fill_textconv(textconv_two, two, &data_two);
731
732 memset(&ecbdata, 0, sizeof(ecbdata));
733 ecbdata.color_diff = want_color(o->use_color);
734 ecbdata.found_changesp = &o->found_changes;
735 ecbdata.ws_rule = whitespace_rule(name_b);
736 ecbdata.opt = o;
737 if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
738 mmfile_t mf1, mf2;
739 mf1.ptr = (char *)data_one;
740 mf2.ptr = (char *)data_two;
741 mf1.size = size_one;
742 mf2.size = size_two;
743 check_blank_at_eof(&mf1, &mf2, &ecbdata);
744 }
745 ecbdata.lno_in_preimage = 1;
746 ecbdata.lno_in_postimage = 1;
747
748 lc_a = count_lines(data_one, size_one);
749 lc_b = count_lines(data_two, size_two);
750 fprintf(o->file,
751 "%s%s--- %s%s%s\n%s%s+++ %s%s%s\n%s%s@@ -",
752 line_prefix, metainfo, a_name.buf, name_a_tab, reset,
753 line_prefix, metainfo, b_name.buf, name_b_tab, reset,
754 line_prefix, fraginfo);
755 if (!o->irreversible_delete)
756 print_line_count(o->file, lc_a);
757 else
758 fprintf(o->file, "?,?");
759 fprintf(o->file, " +");
760 print_line_count(o->file, lc_b);
761 fprintf(o->file, " @@%s\n", reset);
762 if (lc_a && !o->irreversible_delete)
763 emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
764 if (lc_b)
765 emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
766 if (textconv_one)
767 free((char *)data_one);
768 if (textconv_two)
769 free((char *)data_two);
770 }
771
772 struct diff_words_buffer {
773 mmfile_t text;
774 long alloc;
775 struct diff_words_orig {
776 const char *begin, *end;
777 } *orig;
778 int orig_nr, orig_alloc;
779 };
780
781 static void diff_words_append(char *line, unsigned long len,
782 struct diff_words_buffer *buffer)
783 {
784 ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
785 line++;
786 len--;
787 memcpy(buffer->text.ptr + buffer->text.size, line, len);
788 buffer->text.size += len;
789 buffer->text.ptr[buffer->text.size] = '\0';
790 }
791
792 struct diff_words_style_elem {
793 const char *prefix;
794 const char *suffix;
795 const char *color; /* NULL; filled in by the setup code if
796 * color is enabled */
797 };
798
799 struct diff_words_style {
800 enum diff_words_type type;
801 struct diff_words_style_elem new, old, ctx;
802 const char *newline;
803 };
804
805 static struct diff_words_style diff_words_styles[] = {
806 { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
807 { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
808 { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
809 };
810
811 struct diff_words_data {
812 struct diff_words_buffer minus, plus;
813 const char *current_plus;
814 int last_minus;
815 struct diff_options *opt;
816 regex_t *word_regex;
817 enum diff_words_type type;
818 struct diff_words_style *style;
819 };
820
821 static int fn_out_diff_words_write_helper(FILE *fp,
822 struct diff_words_style_elem *st_el,
823 const char *newline,
824 size_t count, const char *buf,
825 const char *line_prefix)
826 {
827 int print = 0;
828
829 while (count) {
830 char *p = memchr(buf, '\n', count);
831 if (print)
832 fputs(line_prefix, fp);
833 if (p != buf) {
834 if (st_el->color && fputs(st_el->color, fp) < 0)
835 return -1;
836 if (fputs(st_el->prefix, fp) < 0 ||
837 fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
838 fputs(st_el->suffix, fp) < 0)
839 return -1;
840 if (st_el->color && *st_el->color
841 && fputs(GIT_COLOR_RESET, fp) < 0)
842 return -1;
843 }
844 if (!p)
845 return 0;
846 if (fputs(newline, fp) < 0)
847 return -1;
848 count -= p + 1 - buf;
849 buf = p + 1;
850 print = 1;
851 }
852 return 0;
853 }
854
855 /*
856 * '--color-words' algorithm can be described as:
857 *
858 * 1. collect a the minus/plus lines of a diff hunk, divided into
859 * minus-lines and plus-lines;
860 *
861 * 2. break both minus-lines and plus-lines into words and
862 * place them into two mmfile_t with one word for each line;
863 *
864 * 3. use xdiff to run diff on the two mmfile_t to get the words level diff;
865 *
866 * And for the common parts of the both file, we output the plus side text.
867 * diff_words->current_plus is used to trace the current position of the plus file
868 * which printed. diff_words->last_minus is used to trace the last minus word
869 * printed.
870 *
871 * For '--graph' to work with '--color-words', we need to output the graph prefix
872 * on each line of color words output. Generally, there are two conditions on
873 * which we should output the prefix.
874 *
875 * 1. diff_words->last_minus == 0 &&
876 * diff_words->current_plus == diff_words->plus.text.ptr
877 *
878 * that is: the plus text must start as a new line, and if there is no minus
879 * word printed, a graph prefix must be printed.
880 *
881 * 2. diff_words->current_plus > diff_words->plus.text.ptr &&
882 * *(diff_words->current_plus - 1) == '\n'
883 *
884 * that is: a graph prefix must be printed following a '\n'
885 */
886 static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
887 {
888 if ((diff_words->last_minus == 0 &&
889 diff_words->current_plus == diff_words->plus.text.ptr) ||
890 (diff_words->current_plus > diff_words->plus.text.ptr &&
891 *(diff_words->current_plus - 1) == '\n')) {
892 return 1;
893 } else {
894 return 0;
895 }
896 }
897
898 static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
899 {
900 struct diff_words_data *diff_words = priv;
901 struct diff_words_style *style = diff_words->style;
902 int minus_first, minus_len, plus_first, plus_len;
903 const char *minus_begin, *minus_end, *plus_begin, *plus_end;
904 struct diff_options *opt = diff_words->opt;
905 const char *line_prefix;
906
907 if (line[0] != '@' || parse_hunk_header(line, len,
908 &minus_first, &minus_len, &plus_first, &plus_len))
909 return;
910
911 assert(opt);
912 line_prefix = diff_line_prefix(opt);
913
914 /* POSIX requires that first be decremented by one if len == 0... */
915 if (minus_len) {
916 minus_begin = diff_words->minus.orig[minus_first].begin;
917 minus_end =
918 diff_words->minus.orig[minus_first + minus_len - 1].end;
919 } else
920 minus_begin = minus_end =
921 diff_words->minus.orig[minus_first].end;
922
923 if (plus_len) {
924 plus_begin = diff_words->plus.orig[plus_first].begin;
925 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
926 } else
927 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
928
929 if (color_words_output_graph_prefix(diff_words)) {
930 fputs(line_prefix, diff_words->opt->file);
931 }
932 if (diff_words->current_plus != plus_begin) {
933 fn_out_diff_words_write_helper(diff_words->opt->file,
934 &style->ctx, style->newline,
935 plus_begin - diff_words->current_plus,
936 diff_words->current_plus, line_prefix);
937 if (*(plus_begin - 1) == '\n')
938 fputs(line_prefix, diff_words->opt->file);
939 }
940 if (minus_begin != minus_end) {
941 fn_out_diff_words_write_helper(diff_words->opt->file,
942 &style->old, style->newline,
943 minus_end - minus_begin, minus_begin,
944 line_prefix);
945 }
946 if (plus_begin != plus_end) {
947 fn_out_diff_words_write_helper(diff_words->opt->file,
948 &style->new, style->newline,
949 plus_end - plus_begin, plus_begin,
950 line_prefix);
951 }
952
953 diff_words->current_plus = plus_end;
954 diff_words->last_minus = minus_first;
955 }
956
957 /* This function starts looking at *begin, and returns 0 iff a word was found. */
958 static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
959 int *begin, int *end)
960 {
961 if (word_regex && *begin < buffer->size) {
962 regmatch_t match[1];
963 if (!regexec(word_regex, buffer->ptr + *begin, 1, match, 0)) {
964 char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
965 '\n', match[0].rm_eo - match[0].rm_so);
966 *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
967 *begin += match[0].rm_so;
968 return *begin >= *end;
969 }
970 return -1;
971 }
972
973 /* find the next word */
974 while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
975 (*begin)++;
976 if (*begin >= buffer->size)
977 return -1;
978
979 /* find the end of the word */
980 *end = *begin + 1;
981 while (*end < buffer->size && !isspace(buffer->ptr[*end]))
982 (*end)++;
983
984 return 0;
985 }
986
987 /*
988 * This function splits the words in buffer->text, stores the list with
989 * newline separator into out, and saves the offsets of the original words
990 * in buffer->orig.
991 */
992 static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
993 regex_t *word_regex)
994 {
995 int i, j;
996 long alloc = 0;
997
998 out->size = 0;
999 out->ptr = NULL;
1000
1001 /* fake an empty "0th" word */
1002 ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
1003 buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
1004 buffer->orig_nr = 1;
1005
1006 for (i = 0; i < buffer->text.size; i++) {
1007 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
1008 return;
1009
1010 /* store original boundaries */
1011 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
1012 buffer->orig_alloc);
1013 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
1014 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
1015 buffer->orig_nr++;
1016
1017 /* store one word */
1018 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
1019 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
1020 out->ptr[out->size + j - i] = '\n';
1021 out->size += j - i + 1;
1022
1023 i = j - 1;
1024 }
1025 }
1026
1027 /* this executes the word diff on the accumulated buffers */
1028 static void diff_words_show(struct diff_words_data *diff_words)
1029 {
1030 xpparam_t xpp;
1031 xdemitconf_t xecfg;
1032 mmfile_t minus, plus;
1033 struct diff_words_style *style = diff_words->style;
1034
1035 struct diff_options *opt = diff_words->opt;
1036 const char *line_prefix;
1037
1038 assert(opt);
1039 line_prefix = diff_line_prefix(opt);
1040
1041 /* special case: only removal */
1042 if (!diff_words->plus.text.size) {
1043 fputs(line_prefix, diff_words->opt->file);
1044 fn_out_diff_words_write_helper(diff_words->opt->file,
1045 &style->old, style->newline,
1046 diff_words->minus.text.size,
1047 diff_words->minus.text.ptr, line_prefix);
1048 diff_words->minus.text.size = 0;
1049 return;
1050 }
1051
1052 diff_words->current_plus = diff_words->plus.text.ptr;
1053 diff_words->last_minus = 0;
1054
1055 memset(&xpp, 0, sizeof(xpp));
1056 memset(&xecfg, 0, sizeof(xecfg));
1057 diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
1058 diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
1059 xpp.flags = 0;
1060 /* as only the hunk header will be parsed, we need a 0-context */
1061 xecfg.ctxlen = 0;
1062 if (xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
1063 &xpp, &xecfg))
1064 die("unable to generate word diff");
1065 free(minus.ptr);
1066 free(plus.ptr);
1067 if (diff_words->current_plus != diff_words->plus.text.ptr +
1068 diff_words->plus.text.size) {
1069 if (color_words_output_graph_prefix(diff_words))
1070 fputs(line_prefix, diff_words->opt->file);
1071 fn_out_diff_words_write_helper(diff_words->opt->file,
1072 &style->ctx, style->newline,
1073 diff_words->plus.text.ptr + diff_words->plus.text.size
1074 - diff_words->current_plus, diff_words->current_plus,
1075 line_prefix);
1076 }
1077 diff_words->minus.text.size = diff_words->plus.text.size = 0;
1078 }
1079
1080 /* In "color-words" mode, show word-diff of words accumulated in the buffer */
1081 static void diff_words_flush(struct emit_callback *ecbdata)
1082 {
1083 if (ecbdata->diff_words->minus.text.size ||
1084 ecbdata->diff_words->plus.text.size)
1085 diff_words_show(ecbdata->diff_words);
1086 }
1087
1088 static void diff_filespec_load_driver(struct diff_filespec *one)
1089 {
1090 /* Use already-loaded driver */
1091 if (one->driver)
1092 return;
1093
1094 if (S_ISREG(one->mode))
1095 one->driver = userdiff_find_by_path(one->path);
1096
1097 /* Fallback to default settings */
1098 if (!one->driver)
1099 one->driver = userdiff_find_by_name("default");
1100 }
1101
1102 static const char *userdiff_word_regex(struct diff_filespec *one)
1103 {
1104 diff_filespec_load_driver(one);
1105 return one->driver->word_regex;
1106 }
1107
1108 static void init_diff_words_data(struct emit_callback *ecbdata,
1109 struct diff_options *orig_opts,
1110 struct diff_filespec *one,
1111 struct diff_filespec *two)
1112 {
1113 int i;
1114 struct diff_options *o = xmalloc(sizeof(struct diff_options));
1115 memcpy(o, orig_opts, sizeof(struct diff_options));
1116
1117 ecbdata->diff_words =
1118 xcalloc(1, sizeof(struct diff_words_data));
1119 ecbdata->diff_words->type = o->word_diff;
1120 ecbdata->diff_words->opt = o;
1121 if (!o->word_regex)
1122 o->word_regex = userdiff_word_regex(one);
1123 if (!o->word_regex)
1124 o->word_regex = userdiff_word_regex(two);
1125 if (!o->word_regex)
1126 o->word_regex = diff_word_regex_cfg;
1127 if (o->word_regex) {
1128 ecbdata->diff_words->word_regex = (regex_t *)
1129 xmalloc(sizeof(regex_t));
1130 if (regcomp(ecbdata->diff_words->word_regex,
1131 o->word_regex,
1132 REG_EXTENDED | REG_NEWLINE))
1133 die ("Invalid regular expression: %s",
1134 o->word_regex);
1135 }
1136 for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1137 if (o->word_diff == diff_words_styles[i].type) {
1138 ecbdata->diff_words->style =
1139 &diff_words_styles[i];
1140 break;
1141 }
1142 }
1143 if (want_color(o->use_color)) {
1144 struct diff_words_style *st = ecbdata->diff_words->style;
1145 st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1146 st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1147 st->ctx.color = diff_get_color_opt(o, DIFF_CONTEXT);
1148 }
1149 }
1150
1151 static void free_diff_words_data(struct emit_callback *ecbdata)
1152 {
1153 if (ecbdata->diff_words) {
1154 diff_words_flush(ecbdata);
1155 free (ecbdata->diff_words->opt);
1156 free (ecbdata->diff_words->minus.text.ptr);
1157 free (ecbdata->diff_words->minus.orig);
1158 free (ecbdata->diff_words->plus.text.ptr);
1159 free (ecbdata->diff_words->plus.orig);
1160 if (ecbdata->diff_words->word_regex) {
1161 regfree(ecbdata->diff_words->word_regex);
1162 free(ecbdata->diff_words->word_regex);
1163 }
1164 free(ecbdata->diff_words);
1165 ecbdata->diff_words = NULL;
1166 }
1167 }
1168
1169 const char *diff_get_color(int diff_use_color, enum color_diff ix)
1170 {
1171 if (want_color(diff_use_color))
1172 return diff_colors[ix];
1173 return "";
1174 }
1175
1176 const char *diff_line_prefix(struct diff_options *opt)
1177 {
1178 struct strbuf *msgbuf;
1179 if (!opt->output_prefix)
1180 return "";
1181
1182 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
1183 return msgbuf->buf;
1184 }
1185
1186 static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1187 {
1188 const char *cp;
1189 unsigned long allot;
1190 size_t l = len;
1191
1192 if (ecb->truncate)
1193 return ecb->truncate(line, len);
1194 cp = line;
1195 allot = l;
1196 while (0 < l) {
1197 (void) utf8_width(&cp, &l);
1198 if (!cp)
1199 break; /* truncated in the middle? */
1200 }
1201 return allot - l;
1202 }
1203
1204 static void find_lno(const char *line, struct emit_callback *ecbdata)
1205 {
1206 const char *p;
1207 ecbdata->lno_in_preimage = 0;
1208 ecbdata->lno_in_postimage = 0;
1209 p = strchr(line, '-');
1210 if (!p)
1211 return; /* cannot happen */
1212 ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1213 p = strchr(p, '+');
1214 if (!p)
1215 return; /* cannot happen */
1216 ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1217 }
1218
1219 static void fn_out_consume(void *priv, char *line, unsigned long len)
1220 {
1221 struct emit_callback *ecbdata = priv;
1222 const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
1223 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
1224 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1225 struct diff_options *o = ecbdata->opt;
1226 const char *line_prefix = diff_line_prefix(o);
1227
1228 if (ecbdata->header) {
1229 fprintf(ecbdata->opt->file, "%s", ecbdata->header->buf);
1230 strbuf_reset(ecbdata->header);
1231 ecbdata->header = NULL;
1232 }
1233 *(ecbdata->found_changesp) = 1;
1234
1235 if (ecbdata->label_path[0]) {
1236 const char *name_a_tab, *name_b_tab;
1237
1238 name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
1239 name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
1240
1241 fprintf(ecbdata->opt->file, "%s%s--- %s%s%s\n",
1242 line_prefix, meta, ecbdata->label_path[0], reset, name_a_tab);
1243 fprintf(ecbdata->opt->file, "%s%s+++ %s%s%s\n",
1244 line_prefix, meta, ecbdata->label_path[1], reset, name_b_tab);
1245 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1246 }
1247
1248 if (diff_suppress_blank_empty
1249 && len == 2 && line[0] == ' ' && line[1] == '\n') {
1250 line[0] = '\n';
1251 len = 1;
1252 }
1253
1254 if (line[0] == '@') {
1255 if (ecbdata->diff_words)
1256 diff_words_flush(ecbdata);
1257 len = sane_truncate_line(ecbdata, line, len);
1258 find_lno(line, ecbdata);
1259 emit_hunk_header(ecbdata, line, len);
1260 if (line[len-1] != '\n')
1261 putc('\n', ecbdata->opt->file);
1262 return;
1263 }
1264
1265 if (len < 1) {
1266 emit_line(ecbdata->opt, reset, reset, line, len);
1267 if (ecbdata->diff_words
1268 && ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN)
1269 fputs("~\n", ecbdata->opt->file);
1270 return;
1271 }
1272
1273 if (ecbdata->diff_words) {
1274 if (line[0] == '-') {
1275 diff_words_append(line, len,
1276 &ecbdata->diff_words->minus);
1277 return;
1278 } else if (line[0] == '+') {
1279 diff_words_append(line, len,
1280 &ecbdata->diff_words->plus);
1281 return;
1282 } else if (starts_with(line, "\\ ")) {
1283 /*
1284 * Eat the "no newline at eof" marker as if we
1285 * saw a "+" or "-" line with nothing on it,
1286 * and return without diff_words_flush() to
1287 * defer processing. If this is the end of
1288 * preimage, more "+" lines may come after it.
1289 */
1290 return;
1291 }
1292 diff_words_flush(ecbdata);
1293 if (ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN) {
1294 emit_line(ecbdata->opt, context, reset, line, len);
1295 fputs("~\n", ecbdata->opt->file);
1296 } else {
1297 /*
1298 * Skip the prefix character, if any. With
1299 * diff_suppress_blank_empty, there may be
1300 * none.
1301 */
1302 if (line[0] != '\n') {
1303 line++;
1304 len--;
1305 }
1306 emit_line(ecbdata->opt, context, reset, line, len);
1307 }
1308 return;
1309 }
1310
1311 switch (line[0]) {
1312 case '+':
1313 ecbdata->lno_in_postimage++;
1314 emit_add_line(reset, ecbdata, line + 1, len - 1);
1315 break;
1316 case '-':
1317 ecbdata->lno_in_preimage++;
1318 emit_del_line(reset, ecbdata, line + 1, len - 1);
1319 break;
1320 case ' ':
1321 ecbdata->lno_in_postimage++;
1322 ecbdata->lno_in_preimage++;
1323 emit_context_line(reset, ecbdata, line + 1, len - 1);
1324 break;
1325 default:
1326 /* incomplete line at the end */
1327 ecbdata->lno_in_preimage++;
1328 emit_line(ecbdata->opt,
1329 diff_get_color(ecbdata->color_diff, DIFF_CONTEXT),
1330 reset, line, len);
1331 break;
1332 }
1333 }
1334
1335 static char *pprint_rename(const char *a, const char *b)
1336 {
1337 const char *old = a;
1338 const char *new = b;
1339 struct strbuf name = STRBUF_INIT;
1340 int pfx_length, sfx_length;
1341 int pfx_adjust_for_slash;
1342 int len_a = strlen(a);
1343 int len_b = strlen(b);
1344 int a_midlen, b_midlen;
1345 int qlen_a = quote_c_style(a, NULL, NULL, 0);
1346 int qlen_b = quote_c_style(b, NULL, NULL, 0);
1347
1348 if (qlen_a || qlen_b) {
1349 quote_c_style(a, &name, NULL, 0);
1350 strbuf_addstr(&name, " => ");
1351 quote_c_style(b, &name, NULL, 0);
1352 return strbuf_detach(&name, NULL);
1353 }
1354
1355 /* Find common prefix */
1356 pfx_length = 0;
1357 while (*old && *new && *old == *new) {
1358 if (*old == '/')
1359 pfx_length = old - a + 1;
1360 old++;
1361 new++;
1362 }
1363
1364 /* Find common suffix */
1365 old = a + len_a;
1366 new = b + len_b;
1367 sfx_length = 0;
1368 /*
1369 * If there is a common prefix, it must end in a slash. In
1370 * that case we let this loop run 1 into the prefix to see the
1371 * same slash.
1372 *
1373 * If there is no common prefix, we cannot do this as it would
1374 * underrun the input strings.
1375 */
1376 pfx_adjust_for_slash = (pfx_length ? 1 : 0);
1377 while (a + pfx_length - pfx_adjust_for_slash <= old &&
1378 b + pfx_length - pfx_adjust_for_slash <= new &&
1379 *old == *new) {
1380 if (*old == '/')
1381 sfx_length = len_a - (old - a);
1382 old--;
1383 new--;
1384 }
1385
1386 /*
1387 * pfx{mid-a => mid-b}sfx
1388 * {pfx-a => pfx-b}sfx
1389 * pfx{sfx-a => sfx-b}
1390 * name-a => name-b
1391 */
1392 a_midlen = len_a - pfx_length - sfx_length;
1393 b_midlen = len_b - pfx_length - sfx_length;
1394 if (a_midlen < 0)
1395 a_midlen = 0;
1396 if (b_midlen < 0)
1397 b_midlen = 0;
1398
1399 strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1400 if (pfx_length + sfx_length) {
1401 strbuf_add(&name, a, pfx_length);
1402 strbuf_addch(&name, '{');
1403 }
1404 strbuf_add(&name, a + pfx_length, a_midlen);
1405 strbuf_addstr(&name, " => ");
1406 strbuf_add(&name, b + pfx_length, b_midlen);
1407 if (pfx_length + sfx_length) {
1408 strbuf_addch(&name, '}');
1409 strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1410 }
1411 return strbuf_detach(&name, NULL);
1412 }
1413
1414 struct diffstat_t {
1415 int nr;
1416 int alloc;
1417 struct diffstat_file {
1418 char *from_name;
1419 char *name;
1420 char *print_name;
1421 unsigned is_unmerged:1;
1422 unsigned is_binary:1;
1423 unsigned is_renamed:1;
1424 unsigned is_interesting:1;
1425 uintmax_t added, deleted;
1426 } **files;
1427 };
1428
1429 static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1430 const char *name_a,
1431 const char *name_b)
1432 {
1433 struct diffstat_file *x;
1434 x = xcalloc(1, sizeof(*x));
1435 ALLOC_GROW(diffstat->files, diffstat->nr + 1, diffstat->alloc);
1436 diffstat->files[diffstat->nr++] = x;
1437 if (name_b) {
1438 x->from_name = xstrdup(name_a);
1439 x->name = xstrdup(name_b);
1440 x->is_renamed = 1;
1441 }
1442 else {
1443 x->from_name = NULL;
1444 x->name = xstrdup(name_a);
1445 }
1446 return x;
1447 }
1448
1449 static void diffstat_consume(void *priv, char *line, unsigned long len)
1450 {
1451 struct diffstat_t *diffstat = priv;
1452 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1453
1454 if (line[0] == '+')
1455 x->added++;
1456 else if (line[0] == '-')
1457 x->deleted++;
1458 }
1459
1460 const char mime_boundary_leader[] = "------------";
1461
1462 static int scale_linear(int it, int width, int max_change)
1463 {
1464 if (!it)
1465 return 0;
1466 /*
1467 * make sure that at least one '-' or '+' is printed if
1468 * there is any change to this path. The easiest way is to
1469 * scale linearly as if the alloted width is one column shorter
1470 * than it is, and then add 1 to the result.
1471 */
1472 return 1 + (it * (width - 1) / max_change);
1473 }
1474
1475 static void show_name(FILE *file,
1476 const char *prefix, const char *name, int len)
1477 {
1478 fprintf(file, " %s%-*s |", prefix, len, name);
1479 }
1480
1481 static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1482 {
1483 if (cnt <= 0)
1484 return;
1485 fprintf(file, "%s", set);
1486 while (cnt--)
1487 putc(ch, file);
1488 fprintf(file, "%s", reset);
1489 }
1490
1491 static void fill_print_name(struct diffstat_file *file)
1492 {
1493 char *pname;
1494
1495 if (file->print_name)
1496 return;
1497
1498 if (!file->is_renamed) {
1499 struct strbuf buf = STRBUF_INIT;
1500 if (quote_c_style(file->name, &buf, NULL, 0)) {
1501 pname = strbuf_detach(&buf, NULL);
1502 } else {
1503 pname = file->name;
1504 strbuf_release(&buf);
1505 }
1506 } else {
1507 pname = pprint_rename(file->from_name, file->name);
1508 }
1509 file->print_name = pname;
1510 }
1511
1512 int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1513 {
1514 struct strbuf sb = STRBUF_INIT;
1515 int ret;
1516
1517 if (!files) {
1518 assert(insertions == 0 && deletions == 0);
1519 return fprintf(fp, "%s\n", " 0 files changed");
1520 }
1521
1522 strbuf_addf(&sb,
1523 (files == 1) ? " %d file changed" : " %d files changed",
1524 files);
1525
1526 /*
1527 * For binary diff, the caller may want to print "x files
1528 * changed" with insertions == 0 && deletions == 0.
1529 *
1530 * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1531 * is probably less confusing (i.e skip over "2 files changed
1532 * but nothing about added/removed lines? Is this a bug in Git?").
1533 */
1534 if (insertions || deletions == 0) {
1535 strbuf_addf(&sb,
1536 (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
1537 insertions);
1538 }
1539
1540 if (deletions || insertions == 0) {
1541 strbuf_addf(&sb,
1542 (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
1543 deletions);
1544 }
1545 strbuf_addch(&sb, '\n');
1546 ret = fputs(sb.buf, fp);
1547 strbuf_release(&sb);
1548 return ret;
1549 }
1550
1551 static void show_stats(struct diffstat_t *data, struct diff_options *options)
1552 {
1553 int i, len, add, del, adds = 0, dels = 0;
1554 uintmax_t max_change = 0, max_len = 0;
1555 int total_files = data->nr, count;
1556 int width, name_width, graph_width, number_width = 0, bin_width = 0;
1557 const char *reset, *add_c, *del_c;
1558 const char *line_prefix = "";
1559 int extra_shown = 0;
1560
1561 if (data->nr == 0)
1562 return;
1563
1564 line_prefix = diff_line_prefix(options);
1565 count = options->stat_count ? options->stat_count : data->nr;
1566
1567 reset = diff_get_color_opt(options, DIFF_RESET);
1568 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1569 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1570
1571 /*
1572 * Find the longest filename and max number of changes
1573 */
1574 for (i = 0; (i < count) && (i < data->nr); i++) {
1575 struct diffstat_file *file = data->files[i];
1576 uintmax_t change = file->added + file->deleted;
1577
1578 if (!file->is_interesting && (change == 0)) {
1579 count++; /* not shown == room for one more */
1580 continue;
1581 }
1582 fill_print_name(file);
1583 len = strlen(file->print_name);
1584 if (max_len < len)
1585 max_len = len;
1586
1587 if (file->is_unmerged) {
1588 /* "Unmerged" is 8 characters */
1589 bin_width = bin_width < 8 ? 8 : bin_width;
1590 continue;
1591 }
1592 if (file->is_binary) {
1593 /* "Bin XXX -> YYY bytes" */
1594 int w = 14 + decimal_width(file->added)
1595 + decimal_width(file->deleted);
1596 bin_width = bin_width < w ? w : bin_width;
1597 /* Display change counts aligned with "Bin" */
1598 number_width = 3;
1599 continue;
1600 }
1601
1602 if (max_change < change)
1603 max_change = change;
1604 }
1605 count = i; /* where we can stop scanning in data->files[] */
1606
1607 /*
1608 * We have width = stat_width or term_columns() columns total.
1609 * We want a maximum of min(max_len, stat_name_width) for the name part.
1610 * We want a maximum of min(max_change, stat_graph_width) for the +- part.
1611 * We also need 1 for " " and 4 + decimal_width(max_change)
1612 * for " | NNNN " and one the empty column at the end, altogether
1613 * 6 + decimal_width(max_change).
1614 *
1615 * If there's not enough space, we will use the smaller of
1616 * stat_name_width (if set) and 5/8*width for the filename,
1617 * and the rest for constant elements + graph part, but no more
1618 * than stat_graph_width for the graph part.
1619 * (5/8 gives 50 for filename and 30 for the constant parts + graph
1620 * for the standard terminal size).
1621 *
1622 * In other words: stat_width limits the maximum width, and
1623 * stat_name_width fixes the maximum width of the filename,
1624 * and is also used to divide available columns if there
1625 * aren't enough.
1626 *
1627 * Binary files are displayed with "Bin XXX -> YYY bytes"
1628 * instead of the change count and graph. This part is treated
1629 * similarly to the graph part, except that it is not
1630 * "scaled". If total width is too small to accommodate the
1631 * guaranteed minimum width of the filename part and the
1632 * separators and this message, this message will "overflow"
1633 * making the line longer than the maximum width.
1634 */
1635
1636 if (options->stat_width == -1)
1637 width = term_columns() - options->output_prefix_length;
1638 else
1639 width = options->stat_width ? options->stat_width : 80;
1640 number_width = decimal_width(max_change) > number_width ?
1641 decimal_width(max_change) : number_width;
1642
1643 if (options->stat_graph_width == -1)
1644 options->stat_graph_width = diff_stat_graph_width;
1645
1646 /*
1647 * Guarantee 3/8*16==6 for the graph part
1648 * and 5/8*16==10 for the filename part
1649 */
1650 if (width < 16 + 6 + number_width)
1651 width = 16 + 6 + number_width;
1652
1653 /*
1654 * First assign sizes that are wanted, ignoring available width.
1655 * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
1656 * starting from "XXX" should fit in graph_width.
1657 */
1658 graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
1659 if (options->stat_graph_width &&
1660 options->stat_graph_width < graph_width)
1661 graph_width = options->stat_graph_width;
1662
1663 name_width = (options->stat_name_width > 0 &&
1664 options->stat_name_width < max_len) ?
1665 options->stat_name_width : max_len;
1666
1667 /*
1668 * Adjust adjustable widths not to exceed maximum width
1669 */
1670 if (name_width + number_width + 6 + graph_width > width) {
1671 if (graph_width > width * 3/8 - number_width - 6) {
1672 graph_width = width * 3/8 - number_width - 6;
1673 if (graph_width < 6)
1674 graph_width = 6;
1675 }
1676
1677 if (options->stat_graph_width &&
1678 graph_width > options->stat_graph_width)
1679 graph_width = options->stat_graph_width;
1680 if (name_width > width - number_width - 6 - graph_width)
1681 name_width = width - number_width - 6 - graph_width;
1682 else
1683 graph_width = width - number_width - 6 - name_width;
1684 }
1685
1686 /*
1687 * From here name_width is the width of the name area,
1688 * and graph_width is the width of the graph area.
1689 * max_change is used to scale graph properly.
1690 */
1691 for (i = 0; i < count; i++) {
1692 const char *prefix = "";
1693 struct diffstat_file *file = data->files[i];
1694 char *name = file->print_name;
1695 uintmax_t added = file->added;
1696 uintmax_t deleted = file->deleted;
1697 int name_len;
1698
1699 if (!file->is_interesting && (added + deleted == 0))
1700 continue;
1701
1702 /*
1703 * "scale" the filename
1704 */
1705 len = name_width;
1706 name_len = strlen(name);
1707 if (name_width < name_len) {
1708 char *slash;
1709 prefix = "...";
1710 len -= 3;
1711 name += name_len - len;
1712 slash = strchr(name, '/');
1713 if (slash)
1714 name = slash;
1715 }
1716
1717 if (file->is_binary) {
1718 fprintf(options->file, "%s", line_prefix);
1719 show_name(options->file, prefix, name, len);
1720 fprintf(options->file, " %*s", number_width, "Bin");
1721 if (!added && !deleted) {
1722 putc('\n', options->file);
1723 continue;
1724 }
1725 fprintf(options->file, " %s%"PRIuMAX"%s",
1726 del_c, deleted, reset);
1727 fprintf(options->file, " -> ");
1728 fprintf(options->file, "%s%"PRIuMAX"%s",
1729 add_c, added, reset);
1730 fprintf(options->file, " bytes");
1731 fprintf(options->file, "\n");
1732 continue;
1733 }
1734 else if (file->is_unmerged) {
1735 fprintf(options->file, "%s", line_prefix);
1736 show_name(options->file, prefix, name, len);
1737 fprintf(options->file, " Unmerged\n");
1738 continue;
1739 }
1740
1741 /*
1742 * scale the add/delete
1743 */
1744 add = added;
1745 del = deleted;
1746
1747 if (graph_width <= max_change) {
1748 int total = scale_linear(add + del, graph_width, max_change);
1749 if (total < 2 && add && del)
1750 /* width >= 2 due to the sanity check */
1751 total = 2;
1752 if (add < del) {
1753 add = scale_linear(add, graph_width, max_change);
1754 del = total - add;
1755 } else {
1756 del = scale_linear(del, graph_width, max_change);
1757 add = total - del;
1758 }
1759 }
1760 fprintf(options->file, "%s", line_prefix);
1761 show_name(options->file, prefix, name, len);
1762 fprintf(options->file, " %*"PRIuMAX"%s",
1763 number_width, added + deleted,
1764 added + deleted ? " " : "");
1765 show_graph(options->file, '+', add, add_c, reset);
1766 show_graph(options->file, '-', del, del_c, reset);
1767 fprintf(options->file, "\n");
1768 }
1769
1770 for (i = 0; i < data->nr; i++) {
1771 struct diffstat_file *file = data->files[i];
1772 uintmax_t added = file->added;
1773 uintmax_t deleted = file->deleted;
1774
1775 if (file->is_unmerged ||
1776 (!file->is_interesting && (added + deleted == 0))) {
1777 total_files--;
1778 continue;
1779 }
1780
1781 if (!file->is_binary) {
1782 adds += added;
1783 dels += deleted;
1784 }
1785 if (i < count)
1786 continue;
1787 if (!extra_shown)
1788 fprintf(options->file, "%s ...\n", line_prefix);
1789 extra_shown = 1;
1790 }
1791 fprintf(options->file, "%s", line_prefix);
1792 print_stat_summary(options->file, total_files, adds, dels);
1793 }
1794
1795 static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1796 {
1797 int i, adds = 0, dels = 0, total_files = data->nr;
1798
1799 if (data->nr == 0)
1800 return;
1801
1802 for (i = 0; i < data->nr; i++) {
1803 int added = data->files[i]->added;
1804 int deleted= data->files[i]->deleted;
1805
1806 if (data->files[i]->is_unmerged ||
1807 (!data->files[i]->is_interesting && (added + deleted == 0))) {
1808 total_files--;
1809 } else if (!data->files[i]->is_binary) { /* don't count bytes */
1810 adds += added;
1811 dels += deleted;
1812 }
1813 }
1814 fprintf(options->file, "%s", diff_line_prefix(options));
1815 print_stat_summary(options->file, total_files, adds, dels);
1816 }
1817
1818 static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1819 {
1820 int i;
1821
1822 if (data->nr == 0)
1823 return;
1824
1825 for (i = 0; i < data->nr; i++) {
1826 struct diffstat_file *file = data->files[i];
1827
1828 fprintf(options->file, "%s", diff_line_prefix(options));
1829
1830 if (file->is_binary)
1831 fprintf(options->file, "-\t-\t");
1832 else
1833 fprintf(options->file,
1834 "%"PRIuMAX"\t%"PRIuMAX"\t",
1835 file->added, file->deleted);
1836 if (options->line_termination) {
1837 fill_print_name(file);
1838 if (!file->is_renamed)
1839 write_name_quoted(file->name, options->file,
1840 options->line_termination);
1841 else {
1842 fputs(file->print_name, options->file);
1843 putc(options->line_termination, options->file);
1844 }
1845 } else {
1846 if (file->is_renamed) {
1847 putc('\0', options->file);
1848 write_name_quoted(file->from_name, options->file, '\0');
1849 }
1850 write_name_quoted(file->name, options->file, '\0');
1851 }
1852 }
1853 }
1854
1855 struct dirstat_file {
1856 const char *name;
1857 unsigned long changed;
1858 };
1859
1860 struct dirstat_dir {
1861 struct dirstat_file *files;
1862 int alloc, nr, permille, cumulative;
1863 };
1864
1865 static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1866 unsigned long changed, const char *base, int baselen)
1867 {
1868 unsigned long this_dir = 0;
1869 unsigned int sources = 0;
1870 const char *line_prefix = diff_line_prefix(opt);
1871
1872 while (dir->nr) {
1873 struct dirstat_file *f = dir->files;
1874 int namelen = strlen(f->name);
1875 unsigned long this;
1876 char *slash;
1877
1878 if (namelen < baselen)
1879 break;
1880 if (memcmp(f->name, base, baselen))
1881 break;
1882 slash = strchr(f->name + baselen, '/');
1883 if (slash) {
1884 int newbaselen = slash + 1 - f->name;
1885 this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
1886 sources++;
1887 } else {
1888 this = f->changed;
1889 dir->files++;
1890 dir->nr--;
1891 sources += 2;
1892 }
1893 this_dir += this;
1894 }
1895
1896 /*
1897 * We don't report dirstat's for
1898 * - the top level
1899 * - or cases where everything came from a single directory
1900 * under this directory (sources == 1).
1901 */
1902 if (baselen && sources != 1) {
1903 if (this_dir) {
1904 int permille = this_dir * 1000 / changed;
1905 if (permille >= dir->permille) {
1906 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
1907 permille / 10, permille % 10, baselen, base);
1908 if (!dir->cumulative)
1909 return 0;
1910 }
1911 }
1912 }
1913 return this_dir;
1914 }
1915
1916 static int dirstat_compare(const void *_a, const void *_b)
1917 {
1918 const struct dirstat_file *a = _a;
1919 const struct dirstat_file *b = _b;
1920 return strcmp(a->name, b->name);
1921 }
1922
1923 static void show_dirstat(struct diff_options *options)
1924 {
1925 int i;
1926 unsigned long changed;
1927 struct dirstat_dir dir;
1928 struct diff_queue_struct *q = &diff_queued_diff;
1929
1930 dir.files = NULL;
1931 dir.alloc = 0;
1932 dir.nr = 0;
1933 dir.permille = options->dirstat_permille;
1934 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1935
1936 changed = 0;
1937 for (i = 0; i < q->nr; i++) {
1938 struct diff_filepair *p = q->queue[i];
1939 const char *name;
1940 unsigned long copied, added, damage;
1941 int content_changed;
1942
1943 name = p->two->path ? p->two->path : p->one->path;
1944
1945 if (p->one->sha1_valid && p->two->sha1_valid)
1946 content_changed = hashcmp(p->one->sha1, p->two->sha1);
1947 else
1948 content_changed = 1;
1949
1950 if (!content_changed) {
1951 /*
1952 * The SHA1 has not changed, so pre-/post-content is
1953 * identical. We can therefore skip looking at the
1954 * file contents altogether.
1955 */
1956 damage = 0;
1957 goto found_damage;
1958 }
1959
1960 if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
1961 /*
1962 * In --dirstat-by-file mode, we don't really need to
1963 * look at the actual file contents at all.
1964 * The fact that the SHA1 changed is enough for us to
1965 * add this file to the list of results
1966 * (with each file contributing equal damage).
1967 */
1968 damage = 1;
1969 goto found_damage;
1970 }
1971
1972 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
1973 diff_populate_filespec(p->one, 0);
1974 diff_populate_filespec(p->two, 0);
1975 diffcore_count_changes(p->one, p->two, NULL, NULL, 0,
1976 &copied, &added);
1977 diff_free_filespec_data(p->one);
1978 diff_free_filespec_data(p->two);
1979 } else if (DIFF_FILE_VALID(p->one)) {
1980 diff_populate_filespec(p->one, CHECK_SIZE_ONLY);
1981 copied = added = 0;
1982 diff_free_filespec_data(p->one);
1983 } else if (DIFF_FILE_VALID(p->two)) {
1984 diff_populate_filespec(p->two, CHECK_SIZE_ONLY);
1985 copied = 0;
1986 added = p->two->size;
1987 diff_free_filespec_data(p->two);
1988 } else
1989 continue;
1990
1991 /*
1992 * Original minus copied is the removed material,
1993 * added is the new material. They are both damages
1994 * made to the preimage.
1995 * If the resulting damage is zero, we know that
1996 * diffcore_count_changes() considers the two entries to
1997 * be identical, but since content_changed is true, we
1998 * know that there must have been _some_ kind of change,
1999 * so we force all entries to have damage > 0.
2000 */
2001 damage = (p->one->size - copied) + added;
2002 if (!damage)
2003 damage = 1;
2004
2005 found_damage:
2006 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2007 dir.files[dir.nr].name = name;
2008 dir.files[dir.nr].changed = damage;
2009 changed += damage;
2010 dir.nr++;
2011 }
2012
2013 /* This can happen even with many files, if everything was renames */
2014 if (!changed)
2015 return;
2016
2017 /* Show all directories with more than x% of the changes */
2018 qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
2019 gather_dirstat(options, &dir, changed, "", 0);
2020 }
2021
2022 static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
2023 {
2024 int i;
2025 unsigned long changed;
2026 struct dirstat_dir dir;
2027
2028 if (data->nr == 0)
2029 return;
2030
2031 dir.files = NULL;
2032 dir.alloc = 0;
2033 dir.nr = 0;
2034 dir.permille = options->dirstat_permille;
2035 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
2036
2037 changed = 0;
2038 for (i = 0; i < data->nr; i++) {
2039 struct diffstat_file *file = data->files[i];
2040 unsigned long damage = file->added + file->deleted;
2041 if (file->is_binary)
2042 /*
2043 * binary files counts bytes, not lines. Must find some
2044 * way to normalize binary bytes vs. textual lines.
2045 * The following heuristic assumes that there are 64
2046 * bytes per "line".
2047 * This is stupid and ugly, but very cheap...
2048 */
2049 damage = (damage + 63) / 64;
2050 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2051 dir.files[dir.nr].name = file->name;
2052 dir.files[dir.nr].changed = damage;
2053 changed += damage;
2054 dir.nr++;
2055 }
2056
2057 /* This can happen even with many files, if everything was renames */
2058 if (!changed)
2059 return;
2060
2061 /* Show all directories with more than x% of the changes */
2062 qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
2063 gather_dirstat(options, &dir, changed, "", 0);
2064 }
2065
2066 static void free_diffstat_info(struct diffstat_t *diffstat)
2067 {
2068 int i;
2069 for (i = 0; i < diffstat->nr; i++) {
2070 struct diffstat_file *f = diffstat->files[i];
2071 if (f->name != f->print_name)
2072 free(f->print_name);
2073 free(f->name);
2074 free(f->from_name);
2075 free(f);
2076 }
2077 free(diffstat->files);
2078 }
2079
2080 struct checkdiff_t {
2081 const char *filename;
2082 int lineno;
2083 int conflict_marker_size;
2084 struct diff_options *o;
2085 unsigned ws_rule;
2086 unsigned status;
2087 };
2088
2089 static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2090 {
2091 char firstchar;
2092 int cnt;
2093
2094 if (len < marker_size + 1)
2095 return 0;
2096 firstchar = line[0];
2097 switch (firstchar) {
2098 case '=': case '>': case '<': case '|':
2099 break;
2100 default:
2101 return 0;
2102 }
2103 for (cnt = 1; cnt < marker_size; cnt++)
2104 if (line[cnt] != firstchar)
2105 return 0;
2106 /* line[1] thru line[marker_size-1] are same as firstchar */
2107 if (len < marker_size + 1 || !isspace(line[marker_size]))
2108 return 0;
2109 return 1;
2110 }
2111
2112 static void checkdiff_consume(void *priv, char *line, unsigned long len)
2113 {
2114 struct checkdiff_t *data = priv;
2115 int marker_size = data->conflict_marker_size;
2116 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2117 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2118 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2119 char *err;
2120 const char *line_prefix;
2121
2122 assert(data->o);
2123 line_prefix = diff_line_prefix(data->o);
2124
2125 if (line[0] == '+') {
2126 unsigned bad;
2127 data->lineno++;
2128 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2129 data->status |= 1;
2130 fprintf(data->o->file,
2131 "%s%s:%d: leftover conflict marker\n",
2132 line_prefix, data->filename, data->lineno);
2133 }
2134 bad = ws_check(line + 1, len - 1, data->ws_rule);
2135 if (!bad)
2136 return;
2137 data->status |= bad;
2138 err = whitespace_error_string(bad);
2139 fprintf(data->o->file, "%s%s:%d: %s.\n",
2140 line_prefix, data->filename, data->lineno, err);
2141 free(err);
2142 emit_line(data->o, set, reset, line, 1);
2143 ws_check_emit(line + 1, len - 1, data->ws_rule,
2144 data->o->file, set, reset, ws);
2145 } else if (line[0] == ' ') {
2146 data->lineno++;
2147 } else if (line[0] == '@') {
2148 char *plus = strchr(line, '+');
2149 if (plus)
2150 data->lineno = strtol(plus, NULL, 10) - 1;
2151 else
2152 die("invalid diff");
2153 }
2154 }
2155
2156 static unsigned char *deflate_it(char *data,
2157 unsigned long size,
2158 unsigned long *result_size)
2159 {
2160 int bound;
2161 unsigned char *deflated;
2162 git_zstream stream;
2163
2164 git_deflate_init(&stream, zlib_compression_level);
2165 bound = git_deflate_bound(&stream, size);
2166 deflated = xmalloc(bound);
2167 stream.next_out = deflated;
2168 stream.avail_out = bound;
2169
2170 stream.next_in = (unsigned char *)data;
2171 stream.avail_in = size;
2172 while (git_deflate(&stream, Z_FINISH) == Z_OK)
2173 ; /* nothing */
2174 git_deflate_end(&stream);
2175 *result_size = stream.total_out;
2176 return deflated;
2177 }
2178
2179 static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two,
2180 const char *prefix)
2181 {
2182 void *cp;
2183 void *delta;
2184 void *deflated;
2185 void *data;
2186 unsigned long orig_size;
2187 unsigned long delta_size;
2188 unsigned long deflate_size;
2189 unsigned long data_size;
2190
2191 /* We could do deflated delta, or we could do just deflated two,
2192 * whichever is smaller.
2193 */
2194 delta = NULL;
2195 deflated = deflate_it(two->ptr, two->size, &deflate_size);
2196 if (one->size && two->size) {
2197 delta = diff_delta(one->ptr, one->size,
2198 two->ptr, two->size,
2199 &delta_size, deflate_size);
2200 if (delta) {
2201 void *to_free = delta;
2202 orig_size = delta_size;
2203 delta = deflate_it(delta, delta_size, &delta_size);
2204 free(to_free);
2205 }
2206 }
2207
2208 if (delta && delta_size < deflate_size) {
2209 fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2210 free(deflated);
2211 data = delta;
2212 data_size = delta_size;
2213 }
2214 else {
2215 fprintf(file, "%sliteral %lu\n", prefix, two->size);
2216 free(delta);
2217 data = deflated;
2218 data_size = deflate_size;
2219 }
2220
2221 /* emit data encoded in base85 */
2222 cp = data;
2223 while (data_size) {
2224 int bytes = (52 < data_size) ? 52 : data_size;
2225 char line[70];
2226 data_size -= bytes;
2227 if (bytes <= 26)
2228 line[0] = bytes + 'A' - 1;
2229 else
2230 line[0] = bytes - 26 + 'a' - 1;
2231 encode_85(line + 1, cp, bytes);
2232 cp = (char *) cp + bytes;
2233 fprintf(file, "%s", prefix);
2234 fputs(line, file);
2235 fputc('\n', file);
2236 }
2237 fprintf(file, "%s\n", prefix);
2238 free(data);
2239 }
2240
2241 static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two,
2242 const char *prefix)
2243 {
2244 fprintf(file, "%sGIT binary patch\n", prefix);
2245 emit_binary_diff_body(file, one, two, prefix);
2246 emit_binary_diff_body(file, two, one, prefix);
2247 }
2248
2249 int diff_filespec_is_binary(struct diff_filespec *one)
2250 {
2251 if (one->is_binary == -1) {
2252 diff_filespec_load_driver(one);
2253 if (one->driver->binary != -1)
2254 one->is_binary = one->driver->binary;
2255 else {
2256 if (!one->data && DIFF_FILE_VALID(one))
2257 diff_populate_filespec(one, CHECK_BINARY);
2258 if (one->is_binary == -1 && one->data)
2259 one->is_binary = buffer_is_binary(one->data,
2260 one->size);
2261 if (one->is_binary == -1)
2262 one->is_binary = 0;
2263 }
2264 }
2265 return one->is_binary;
2266 }
2267
2268 static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2269 {
2270 diff_filespec_load_driver(one);
2271 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2272 }
2273
2274 void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2275 {
2276 if (!options->a_prefix)
2277 options->a_prefix = a;
2278 if (!options->b_prefix)
2279 options->b_prefix = b;
2280 }
2281
2282 struct userdiff_driver *get_textconv(struct diff_filespec *one)
2283 {
2284 if (!DIFF_FILE_VALID(one))
2285 return NULL;
2286
2287 diff_filespec_load_driver(one);
2288 return userdiff_get_textconv(one->driver);
2289 }
2290
2291 static void builtin_diff(const char *name_a,
2292 const char *name_b,
2293 struct diff_filespec *one,
2294 struct diff_filespec *two,
2295 const char *xfrm_msg,
2296 int must_show_header,
2297 struct diff_options *o,
2298 int complete_rewrite)
2299 {
2300 mmfile_t mf1, mf2;
2301 const char *lbl[2];
2302 char *a_one, *b_two;
2303 const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
2304 const char *reset = diff_get_color_opt(o, DIFF_RESET);
2305 const char *a_prefix, *b_prefix;
2306 struct userdiff_driver *textconv_one = NULL;
2307 struct userdiff_driver *textconv_two = NULL;
2308 struct strbuf header = STRBUF_INIT;
2309 const char *line_prefix = diff_line_prefix(o);
2310
2311 if (DIFF_OPT_TST(o, SUBMODULE_LOG) &&
2312 (!one->mode || S_ISGITLINK(one->mode)) &&
2313 (!two->mode || S_ISGITLINK(two->mode))) {
2314 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2315 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2316 show_submodule_summary(o->file, one->path ? one->path : two->path,
2317 line_prefix,
2318 one->sha1, two->sha1, two->dirty_submodule,
2319 meta, del, add, reset);
2320 return;
2321 }
2322
2323 if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2324 textconv_one = get_textconv(one);
2325 textconv_two = get_textconv(two);
2326 }
2327
2328 diff_set_mnemonic_prefix(o, "a/", "b/");
2329 if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2330 a_prefix = o->b_prefix;
2331 b_prefix = o->a_prefix;
2332 } else {
2333 a_prefix = o->a_prefix;
2334 b_prefix = o->b_prefix;
2335 }
2336
2337 /* Never use a non-valid filename anywhere if at all possible */
2338 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2339 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2340
2341 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2342 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2343 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2344 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2345 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
2346 if (lbl[0][0] == '/') {
2347 /* /dev/null */
2348 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
2349 if (xfrm_msg)
2350 strbuf_addstr(&header, xfrm_msg);
2351 must_show_header = 1;
2352 }
2353 else if (lbl[1][0] == '/') {
2354 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
2355 if (xfrm_msg)
2356 strbuf_addstr(&header, xfrm_msg);
2357 must_show_header = 1;
2358 }
2359 else {
2360 if (one->mode != two->mode) {
2361 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
2362 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
2363 must_show_header = 1;
2364 }
2365 if (xfrm_msg)
2366 strbuf_addstr(&header, xfrm_msg);
2367
2368 /*
2369 * we do not run diff between different kind
2370 * of objects.
2371 */
2372 if ((one->mode ^ two->mode) & S_IFMT)
2373 goto free_ab_and_return;
2374 if (complete_rewrite &&
2375 (textconv_one || !diff_filespec_is_binary(one)) &&
2376 (textconv_two || !diff_filespec_is_binary(two))) {
2377 fprintf(o->file, "%s", header.buf);
2378 strbuf_reset(&header);
2379 emit_rewrite_diff(name_a, name_b, one, two,
2380 textconv_one, textconv_two, o);
2381 o->found_changes = 1;
2382 goto free_ab_and_return;
2383 }
2384 }
2385
2386 if (o->irreversible_delete && lbl[1][0] == '/') {
2387 fprintf(o->file, "%s", header.buf);
2388 strbuf_reset(&header);
2389 goto free_ab_and_return;
2390 } else if (!DIFF_OPT_TST(o, TEXT) &&
2391 ( (!textconv_one && diff_filespec_is_binary(one)) ||
2392 (!textconv_two && diff_filespec_is_binary(two)) )) {
2393 if (!one->data && !two->data &&
2394 S_ISREG(one->mode) && S_ISREG(two->mode) &&
2395 !DIFF_OPT_TST(o, BINARY)) {
2396 if (!hashcmp(one->sha1, two->sha1)) {
2397 if (must_show_header)
2398 fprintf(o->file, "%s", header.buf);
2399 goto free_ab_and_return;
2400 }
2401 fprintf(o->file, "%s", header.buf);
2402 fprintf(o->file, "%sBinary files %s and %s differ\n",
2403 line_prefix, lbl[0], lbl[1]);
2404 goto free_ab_and_return;
2405 }
2406 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2407 die("unable to read files to diff");
2408 /* Quite common confusing case */
2409 if (mf1.size == mf2.size &&
2410 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2411 if (must_show_header)
2412 fprintf(o->file, "%s", header.buf);
2413 goto free_ab_and_return;
2414 }
2415 fprintf(o->file, "%s", header.buf);
2416 strbuf_reset(&header);
2417 if (DIFF_OPT_TST(o, BINARY))
2418 emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2419 else
2420 fprintf(o->file, "%sBinary files %s and %s differ\n",
2421 line_prefix, lbl[0], lbl[1]);
2422 o->found_changes = 1;
2423 } else {
2424 /* Crazy xdl interfaces.. */
2425 const char *diffopts = getenv("GIT_DIFF_OPTS");
2426 const char *v;
2427 xpparam_t xpp;
2428 xdemitconf_t xecfg;
2429 struct emit_callback ecbdata;
2430 const struct userdiff_funcname *pe;
2431
2432 if (must_show_header) {
2433 fprintf(o->file, "%s", header.buf);
2434 strbuf_reset(&header);
2435 }
2436
2437 mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2438 mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2439
2440 pe = diff_funcname_pattern(one);
2441 if (!pe)
2442 pe = diff_funcname_pattern(two);
2443
2444 memset(&xpp, 0, sizeof(xpp));
2445 memset(&xecfg, 0, sizeof(xecfg));
2446 memset(&ecbdata, 0, sizeof(ecbdata));
2447 ecbdata.label_path = lbl;
2448 ecbdata.color_diff = want_color(o->use_color);
2449 ecbdata.found_changesp = &o->found_changes;
2450 ecbdata.ws_rule = whitespace_rule(name_b);
2451 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2452 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2453 ecbdata.opt = o;
2454 ecbdata.header = header.len ? &header : NULL;
2455 xpp.flags = o->xdl_opts;
2456 xecfg.ctxlen = o->context;
2457 xecfg.interhunkctxlen = o->interhunkcontext;
2458 xecfg.flags = XDL_EMIT_FUNCNAMES;
2459 if (DIFF_OPT_TST(o, FUNCCONTEXT))
2460 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2461 if (pe)
2462 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2463 if (!diffopts)
2464 ;
2465 else if (skip_prefix(diffopts, "--unified=", &v))
2466 xecfg.ctxlen = strtoul(v, NULL, 10);
2467 else if (skip_prefix(diffopts, "-u", &v))
2468 xecfg.ctxlen = strtoul(v, NULL, 10);
2469 if (o->word_diff)
2470 init_diff_words_data(&ecbdata, o, one, two);
2471 if (xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2472 &xpp, &xecfg))
2473 die("unable to generate diff for %s", one->path);
2474 if (o->word_diff)
2475 free_diff_words_data(&ecbdata);
2476 if (textconv_one)
2477 free(mf1.ptr);
2478 if (textconv_two)
2479 free(mf2.ptr);
2480 xdiff_clear_find_func(&xecfg);
2481 }
2482
2483 free_ab_and_return:
2484 strbuf_release(&header);
2485 diff_free_filespec_data(one);
2486 diff_free_filespec_data(two);
2487 free(a_one);
2488 free(b_two);
2489 return;
2490 }
2491
2492 static void builtin_diffstat(const char *name_a, const char *name_b,
2493 struct diff_filespec *one,
2494 struct diff_filespec *two,
2495 struct diffstat_t *diffstat,
2496 struct diff_options *o,
2497 struct diff_filepair *p)
2498 {
2499 mmfile_t mf1, mf2;
2500 struct diffstat_file *data;
2501 int same_contents;
2502 int complete_rewrite = 0;
2503
2504 if (!DIFF_PAIR_UNMERGED(p)) {
2505 if (p->status == DIFF_STATUS_MODIFIED && p->score)
2506 complete_rewrite = 1;
2507 }
2508
2509 data = diffstat_add(diffstat, name_a, name_b);
2510 data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
2511
2512 if (!one || !two) {
2513 data->is_unmerged = 1;
2514 return;
2515 }
2516
2517 same_contents = !hashcmp(one->sha1, two->sha1);
2518
2519 if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2520 data->is_binary = 1;
2521 if (same_contents) {
2522 data->added = 0;
2523 data->deleted = 0;
2524 } else {
2525 data->added = diff_filespec_size(two);
2526 data->deleted = diff_filespec_size(one);
2527 }
2528 }
2529
2530 else if (complete_rewrite) {
2531 diff_populate_filespec(one, 0);
2532 diff_populate_filespec(two, 0);
2533 data->deleted = count_lines(one->data, one->size);
2534 data->added = count_lines(two->data, two->size);
2535 }
2536
2537 else if (!same_contents) {
2538 /* Crazy xdl interfaces.. */
2539 xpparam_t xpp;
2540 xdemitconf_t xecfg;
2541
2542 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2543 die("unable to read files to diff");
2544
2545 memset(&xpp, 0, sizeof(xpp));
2546 memset(&xecfg, 0, sizeof(xecfg));
2547 xpp.flags = o->xdl_opts;
2548 xecfg.ctxlen = o->context;
2549 xecfg.interhunkctxlen = o->interhunkcontext;
2550 if (xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2551 &xpp, &xecfg))
2552 die("unable to generate diffstat for %s", one->path);
2553 }
2554
2555 diff_free_filespec_data(one);
2556 diff_free_filespec_data(two);
2557 }
2558
2559 static void builtin_checkdiff(const char *name_a, const char *name_b,
2560 const char *attr_path,
2561 struct diff_filespec *one,
2562 struct diff_filespec *two,
2563 struct diff_options *o)
2564 {
2565 mmfile_t mf1, mf2;
2566 struct checkdiff_t data;
2567
2568 if (!two)
2569 return;
2570
2571 memset(&data, 0, sizeof(data));
2572 data.filename = name_b ? name_b : name_a;
2573 data.lineno = 0;
2574 data.o = o;
2575 data.ws_rule = whitespace_rule(attr_path);
2576 data.conflict_marker_size = ll_merge_marker_size(attr_path);
2577
2578 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2579 die("unable to read files to diff");
2580
2581 /*
2582 * All the other codepaths check both sides, but not checking
2583 * the "old" side here is deliberate. We are checking the newly
2584 * introduced changes, and as long as the "new" side is text, we
2585 * can and should check what it introduces.
2586 */
2587 if (diff_filespec_is_binary(two))
2588 goto free_and_return;
2589 else {
2590 /* Crazy xdl interfaces.. */
2591 xpparam_t xpp;
2592 xdemitconf_t xecfg;
2593
2594 memset(&xpp, 0, sizeof(xpp));
2595 memset(&xecfg, 0, sizeof(xecfg));
2596 xecfg.ctxlen = 1; /* at least one context line */
2597 xpp.flags = 0;
2598 if (xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2599 &xpp, &xecfg))
2600 die("unable to generate checkdiff for %s", one->path);
2601
2602 if (data.ws_rule & WS_BLANK_AT_EOF) {
2603 struct emit_callback ecbdata;
2604 int blank_at_eof;
2605
2606 ecbdata.ws_rule = data.ws_rule;
2607 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2608 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2609
2610 if (blank_at_eof) {
2611 static char *err;
2612 if (!err)
2613 err = whitespace_error_string(WS_BLANK_AT_EOF);
2614 fprintf(o->file, "%s:%d: %s.\n",
2615 data.filename, blank_at_eof, err);
2616 data.status = 1; /* report errors */
2617 }
2618 }
2619 }
2620 free_and_return:
2621 diff_free_filespec_data(one);
2622 diff_free_filespec_data(two);
2623 if (data.status)
2624 DIFF_OPT_SET(o, CHECK_FAILED);
2625 }
2626
2627 struct diff_filespec *alloc_filespec(const char *path)
2628 {
2629 struct diff_filespec *spec;
2630
2631 FLEXPTR_ALLOC_STR(spec, path, path);
2632 spec->count = 1;
2633 spec->is_binary = -1;
2634 return spec;
2635 }
2636
2637 void free_filespec(struct diff_filespec *spec)
2638 {
2639 if (!--spec->count) {
2640 diff_free_filespec_data(spec);
2641 free(spec);
2642 }
2643 }
2644
2645 void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
2646 int sha1_valid, unsigned short mode)
2647 {
2648 if (mode) {
2649 spec->mode = canon_mode(mode);
2650 hashcpy(spec->sha1, sha1);
2651 spec->sha1_valid = sha1_valid;
2652 }
2653 }
2654
2655 /*
2656 * Given a name and sha1 pair, if the index tells us the file in
2657 * the work tree has that object contents, return true, so that
2658 * prepare_temp_file() does not have to inflate and extract.
2659 */
2660 static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
2661 {
2662 const struct cache_entry *ce;
2663 struct stat st;
2664 int pos, len;
2665
2666 /*
2667 * We do not read the cache ourselves here, because the
2668 * benchmark with my previous version that always reads cache
2669 * shows that it makes things worse for diff-tree comparing
2670 * two linux-2.6 kernel trees in an already checked out work
2671 * tree. This is because most diff-tree comparisons deal with
2672 * only a small number of files, while reading the cache is
2673 * expensive for a large project, and its cost outweighs the
2674 * savings we get by not inflating the object to a temporary
2675 * file. Practically, this code only helps when we are used
2676 * by diff-cache --cached, which does read the cache before
2677 * calling us.
2678 */
2679 if (!active_cache)
2680 return 0;
2681
2682 /* We want to avoid the working directory if our caller
2683 * doesn't need the data in a normal file, this system
2684 * is rather slow with its stat/open/mmap/close syscalls,
2685 * and the object is contained in a pack file. The pack
2686 * is probably already open and will be faster to obtain
2687 * the data through than the working directory. Loose
2688 * objects however would tend to be slower as they need
2689 * to be individually opened and inflated.
2690 */
2691 if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1))
2692 return 0;
2693
2694 len = strlen(name);
2695 pos = cache_name_pos(name, len);
2696 if (pos < 0)
2697 return 0;
2698 ce = active_cache[pos];
2699
2700 /*
2701 * This is not the sha1 we are looking for, or
2702 * unreusable because it is not a regular file.
2703 */
2704 if (hashcmp(sha1, ce->sha1) || !S_ISREG(ce->ce_mode))
2705 return 0;
2706
2707 /*
2708 * If ce is marked as "assume unchanged", there is no
2709 * guarantee that work tree matches what we are looking for.
2710 */
2711 if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2712 return 0;
2713
2714 /*
2715 * If ce matches the file in the work tree, we can reuse it.
2716 */
2717 if (ce_uptodate(ce) ||
2718 (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2719 return 1;
2720
2721 return 0;
2722 }
2723
2724 static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2725 {
2726 struct strbuf buf = STRBUF_INIT;
2727 char *dirty = "";
2728
2729 /* Are we looking at the work tree? */
2730 if (s->dirty_submodule)
2731 dirty = "-dirty";
2732
2733 strbuf_addf(&buf, "Subproject commit %s%s\n", sha1_to_hex(s->sha1), dirty);
2734 s->size = buf.len;
2735 if (size_only) {
2736 s->data = NULL;
2737 strbuf_release(&buf);
2738 } else {
2739 s->data = strbuf_detach(&buf, NULL);
2740 s->should_free = 1;
2741 }
2742 return 0;
2743 }
2744
2745 /*
2746 * While doing rename detection and pickaxe operation, we may need to
2747 * grab the data for the blob (or file) for our own in-core comparison.
2748 * diff_filespec has data and size fields for this purpose.
2749 */
2750 int diff_populate_filespec(struct diff_filespec *s, unsigned int flags)
2751 {
2752 int size_only = flags & CHECK_SIZE_ONLY;
2753 int err = 0;
2754 /*
2755 * demote FAIL to WARN to allow inspecting the situation
2756 * instead of refusing.
2757 */
2758 enum safe_crlf crlf_warn = (safe_crlf == SAFE_CRLF_FAIL
2759 ? SAFE_CRLF_WARN
2760 : safe_crlf);
2761
2762 if (!DIFF_FILE_VALID(s))
2763 die("internal error: asking to populate invalid file.");
2764 if (S_ISDIR(s->mode))
2765 return -1;
2766
2767 if (s->data)
2768 return 0;
2769
2770 if (size_only && 0 < s->size)
2771 return 0;
2772
2773 if (S_ISGITLINK(s->mode))
2774 return diff_populate_gitlink(s, size_only);
2775
2776 if (!s->sha1_valid ||
2777 reuse_worktree_file(s->path, s->sha1, 0)) {
2778 struct strbuf buf = STRBUF_INIT;
2779 struct stat st;
2780 int fd;
2781
2782 if (lstat(s->path, &st) < 0) {
2783 if (errno == ENOENT) {
2784 err_empty:
2785 err = -1;
2786 empty:
2787 s->data = (char *)"";
2788 s->size = 0;
2789 return err;
2790 }
2791 }
2792 s->size = xsize_t(st.st_size);
2793 if (!s->size)
2794 goto empty;
2795 if (S_ISLNK(st.st_mode)) {
2796 struct strbuf sb = STRBUF_INIT;
2797
2798 if (strbuf_readlink(&sb, s->path, s->size))
2799 goto err_empty;
2800 s->size = sb.len;
2801 s->data = strbuf_detach(&sb, NULL);
2802 s->should_free = 1;
2803 return 0;
2804 }
2805 if (size_only)
2806 return 0;
2807 if ((flags & CHECK_BINARY) &&
2808 s->size > big_file_threshold && s->is_binary == -1) {
2809 s->is_binary = 1;
2810 return 0;
2811 }
2812 fd = open(s->path, O_RDONLY);
2813 if (fd < 0)
2814 goto err_empty;
2815 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2816 close(fd);
2817 s->should_munmap = 1;
2818
2819 /*
2820 * Convert from working tree format to canonical git format
2821 */
2822 if (convert_to_git(s->path, s->data, s->size, &buf, crlf_warn)) {
2823 size_t size = 0;
2824 munmap(s->data, s->size);
2825 s->should_munmap = 0;
2826 s->data = strbuf_detach(&buf, &size);
2827 s->size = size;
2828 s->should_free = 1;
2829 }
2830 }
2831 else {
2832 enum object_type type;
2833 if (size_only || (flags & CHECK_BINARY)) {
2834 type = sha1_object_info(s->sha1, &s->size);
2835 if (type < 0)
2836 die("unable to read %s", sha1_to_hex(s->sha1));
2837 if (size_only)
2838 return 0;
2839 if (s->size > big_file_threshold && s->is_binary == -1) {
2840 s->is_binary = 1;
2841 return 0;
2842 }
2843 }
2844 s->data = read_sha1_file(s->sha1, &type, &s->size);
2845 if (!s->data)
2846 die("unable to read %s", sha1_to_hex(s->sha1));
2847 s->should_free = 1;
2848 }
2849 return 0;
2850 }
2851
2852 void diff_free_filespec_blob(struct diff_filespec *s)
2853 {
2854 if (s->should_free)
2855 free(s->data);
2856 else if (s->should_munmap)
2857 munmap(s->data, s->size);
2858
2859 if (s->should_free || s->should_munmap) {
2860 s->should_free = s->should_munmap = 0;
2861 s->data = NULL;
2862 }
2863 }
2864
2865 void diff_free_filespec_data(struct diff_filespec *s)
2866 {
2867 diff_free_filespec_blob(s);
2868 free(s->cnt_data);
2869 s->cnt_data = NULL;
2870 }
2871
2872 static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2873 void *blob,
2874 unsigned long size,
2875 const unsigned char *sha1,
2876 int mode)
2877 {
2878 int fd;
2879 struct strbuf buf = STRBUF_INIT;
2880 struct strbuf template = STRBUF_INIT;
2881 char *path_dup = xstrdup(path);
2882 const char *base = basename(path_dup);
2883
2884 /* Generate "XXXXXX_basename.ext" */
2885 strbuf_addstr(&template, "XXXXXX_");
2886 strbuf_addstr(&template, base);
2887
2888 fd = mks_tempfile_ts(&temp->tempfile, template.buf, strlen(base) + 1);
2889 if (fd < 0)
2890 die_errno("unable to create temp-file");
2891 if (convert_to_working_tree(path,
2892 (const char *)blob, (size_t)size, &buf)) {
2893 blob = buf.buf;
2894 size = buf.len;
2895 }
2896 if (write_in_full(fd, blob, size) != size)
2897 die_errno("unable to write temp-file");
2898 close_tempfile(&temp->tempfile);
2899 temp->name = get_tempfile_path(&temp->tempfile);
2900 sha1_to_hex_r(temp->hex, sha1);
2901 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", mode);
2902 strbuf_release(&buf);
2903 strbuf_release(&template);
2904 free(path_dup);
2905 }
2906
2907 static struct diff_tempfile *prepare_temp_file(const char *name,
2908 struct diff_filespec *one)
2909 {
2910 struct diff_tempfile *temp = claim_diff_tempfile();
2911
2912 if (!DIFF_FILE_VALID(one)) {
2913 not_a_valid_file:
2914 /* A '-' entry produces this for file-2, and
2915 * a '+' entry produces this for file-1.
2916 */
2917 temp->name = "/dev/null";
2918 xsnprintf(temp->hex, sizeof(temp->hex), ".");
2919 xsnprintf(temp->mode, sizeof(temp->mode), ".");
2920 return temp;
2921 }
2922
2923 if (!S_ISGITLINK(one->mode) &&
2924 (!one->sha1_valid ||
2925 reuse_worktree_file(name, one->sha1, 1))) {
2926 struct stat st;
2927 if (lstat(name, &st) < 0) {
2928 if (errno == ENOENT)
2929 goto not_a_valid_file;
2930 die_errno("stat(%s)", name);
2931 }
2932 if (S_ISLNK(st.st_mode)) {
2933 struct strbuf sb = STRBUF_INIT;
2934 if (strbuf_readlink(&sb, name, st.st_size) < 0)
2935 die_errno("readlink(%s)", name);
2936 prep_temp_blob(name, temp, sb.buf, sb.len,
2937 (one->sha1_valid ?
2938 one->sha1 : null_sha1),
2939 (one->sha1_valid ?
2940 one->mode : S_IFLNK));
2941 strbuf_release(&sb);
2942 }
2943 else {
2944 /* we can borrow from the file in the work tree */
2945 temp->name = name;
2946 if (!one->sha1_valid)
2947 sha1_to_hex_r(temp->hex, null_sha1);
2948 else
2949 sha1_to_hex_r(temp->hex, one->sha1);
2950 /* Even though we may sometimes borrow the
2951 * contents from the work tree, we always want
2952 * one->mode. mode is trustworthy even when
2953 * !(one->sha1_valid), as long as
2954 * DIFF_FILE_VALID(one).
2955 */
2956 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", one->mode);
2957 }
2958 return temp;
2959 }
2960 else {
2961 if (diff_populate_filespec(one, 0))
2962 die("cannot read data blob for %s", one->path);
2963 prep_temp_blob(name, temp, one->data, one->size,
2964 one->sha1, one->mode);
2965 }
2966 return temp;
2967 }
2968
2969 static void add_external_diff_name(struct argv_array *argv,
2970 const char *name,
2971 struct diff_filespec *df)
2972 {
2973 struct diff_tempfile *temp = prepare_temp_file(name, df);
2974 argv_array_push(argv, temp->name);
2975 argv_array_push(argv, temp->hex);
2976 argv_array_push(argv, temp->mode);
2977 }
2978
2979 /* An external diff command takes:
2980 *
2981 * diff-cmd name infile1 infile1-sha1 infile1-mode \
2982 * infile2 infile2-sha1 infile2-mode [ rename-to ]
2983 *
2984 */
2985 static void run_external_diff(const char *pgm,
2986 const char *name,
2987 const char *other,
2988 struct diff_filespec *one,
2989 struct diff_filespec *two,
2990 const char *xfrm_msg,
2991 int complete_rewrite,
2992 struct diff_options *o)
2993 {
2994 struct argv_array argv = ARGV_ARRAY_INIT;
2995 struct argv_array env = ARGV_ARRAY_INIT;
2996 struct diff_queue_struct *q = &diff_queued_diff;
2997
2998 argv_array_push(&argv, pgm);
2999 argv_array_push(&argv, name);
3000
3001 if (one && two) {
3002 add_external_diff_name(&argv, name, one);
3003 if (!other)
3004 add_external_diff_name(&argv, name, two);
3005 else {
3006 add_external_diff_name(&argv, other, two);
3007 argv_array_push(&argv, other);
3008 argv_array_push(&argv, xfrm_msg);
3009 }
3010 }
3011
3012 argv_array_pushf(&env, "GIT_DIFF_PATH_COUNTER=%d", ++o->diff_path_counter);
3013 argv_array_pushf(&env, "GIT_DIFF_PATH_TOTAL=%d", q->nr);
3014
3015 if (run_command_v_opt_cd_env(argv.argv, RUN_USING_SHELL, NULL, env.argv))
3016 die(_("external diff died, stopping at %s"), name);
3017
3018 remove_tempfile();
3019 argv_array_clear(&argv);
3020 argv_array_clear(&env);
3021 }
3022
3023 static int similarity_index(struct diff_filepair *p)
3024 {
3025 return p->score * 100 / MAX_SCORE;
3026 }
3027
3028 static void fill_metainfo(struct strbuf *msg,
3029 const char *name,
3030 const char *other,
3031 struct diff_filespec *one,
3032 struct diff_filespec *two,
3033 struct diff_options *o,
3034 struct diff_filepair *p,
3035 int *must_show_header,
3036 int use_color)
3037 {
3038 const char *set = diff_get_color(use_color, DIFF_METAINFO);
3039 const char *reset = diff_get_color(use_color, DIFF_RESET);
3040 const char *line_prefix = diff_line_prefix(o);
3041
3042 *must_show_header = 1;
3043 strbuf_init(msg, PATH_MAX * 2 + 300);
3044 switch (p->status) {
3045 case DIFF_STATUS_COPIED:
3046 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3047 line_prefix, set, similarity_index(p));
3048 strbuf_addf(msg, "%s\n%s%scopy from ",
3049 reset, line_prefix, set);
3050 quote_c_style(name, msg, NULL, 0);
3051 strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
3052 quote_c_style(other, msg, NULL, 0);
3053 strbuf_addf(msg, "%s\n", reset);
3054 break;
3055 case DIFF_STATUS_RENAMED:
3056 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3057 line_prefix, set, similarity_index(p));
3058 strbuf_addf(msg, "%s\n%s%srename from ",
3059 reset, line_prefix, set);
3060 quote_c_style(name, msg, NULL, 0);
3061 strbuf_addf(msg, "%s\n%s%srename to ",
3062 reset, line_prefix, set);
3063 quote_c_style(other, msg, NULL, 0);
3064 strbuf_addf(msg, "%s\n", reset);
3065 break;
3066 case DIFF_STATUS_MODIFIED:
3067 if (p->score) {
3068 strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
3069 line_prefix,
3070 set, similarity_index(p), reset);
3071 break;
3072 }
3073 /* fallthru */
3074 default:
3075 *must_show_header = 0;
3076 }
3077 if (one && two && hashcmp(one->sha1, two->sha1)) {
3078 int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
3079
3080 if (DIFF_OPT_TST(o, BINARY)) {
3081 mmfile_t mf;
3082 if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
3083 (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
3084 abbrev = 40;
3085 }
3086 strbuf_addf(msg, "%s%sindex %s..", line_prefix, set,
3087 find_unique_abbrev(one->sha1, abbrev));
3088 strbuf_addstr(msg, find_unique_abbrev(two->sha1, abbrev));
3089 if (one->mode == two->mode)
3090 strbuf_addf(msg, " %06o", one->mode);
3091 strbuf_addf(msg, "%s\n", reset);
3092 }
3093 }
3094
3095 static void run_diff_cmd(const char *pgm,
3096 const char *name,
3097 const char *other,
3098 const char *attr_path,
3099 struct diff_filespec *one,
3100 struct diff_filespec *two,
3101 struct strbuf *msg,
3102 struct diff_options *o,
3103 struct diff_filepair *p)
3104 {
3105 const char *xfrm_msg = NULL;
3106 int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
3107 int must_show_header = 0;
3108
3109
3110 if (DIFF_OPT_TST(o, ALLOW_EXTERNAL)) {
3111 struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
3112 if (drv && drv->external)
3113 pgm = drv->external;
3114 }
3115
3116 if (msg) {
3117 /*
3118 * don't use colors when the header is intended for an
3119 * external diff driver
3120 */
3121 fill_metainfo(msg, name, other, one, two, o, p,
3122 &must_show_header,
3123 want_color(o->use_color) && !pgm);
3124 xfrm_msg = msg->len ? msg->buf : NULL;
3125 }
3126
3127 if (pgm) {
3128 run_external_diff(pgm, name, other, one, two, xfrm_msg,
3129 complete_rewrite, o);
3130 return;
3131 }
3132 if (one && two)
3133 builtin_diff(name, other ? other : name,
3134 one, two, xfrm_msg, must_show_header,
3135 o, complete_rewrite);
3136 else
3137 fprintf(o->file, "* Unmerged path %s\n", name);
3138 }
3139
3140 static void diff_fill_sha1_info(struct diff_filespec *one)
3141 {
3142 if (DIFF_FILE_VALID(one)) {
3143 if (!one->sha1_valid) {
3144 struct stat st;
3145 if (one->is_stdin) {
3146 hashcpy(one->sha1, null_sha1);
3147 return;
3148 }
3149 if (lstat(one->path, &st) < 0)
3150 die_errno("stat '%s'", one->path);
3151 if (index_path(one->sha1, one->path, &st, 0))
3152 die("cannot hash %s", one->path);
3153 }
3154 }
3155 else
3156 hashclr(one->sha1);
3157 }
3158
3159 static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
3160 {
3161 /* Strip the prefix but do not molest /dev/null and absolute paths */
3162 if (*namep && **namep != '/') {
3163 *namep += prefix_length;
3164 if (**namep == '/')
3165 ++*namep;
3166 }
3167 if (*otherp && **otherp != '/') {
3168 *otherp += prefix_length;
3169 if (**otherp == '/')
3170 ++*otherp;
3171 }
3172 }
3173
3174 static void run_diff(struct diff_filepair *p, struct diff_options *o)
3175 {
3176 const char *pgm = external_diff();
3177 struct strbuf msg;
3178 struct diff_filespec *one = p->one;
3179 struct diff_filespec *two = p->two;
3180 const char *name;
3181 const char *other;
3182 const char *attr_path;
3183
3184 name = p->one->path;
3185 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3186 attr_path = name;
3187 if (o->prefix_length)
3188 strip_prefix(o->prefix_length, &name, &other);
3189
3190 if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
3191 pgm = NULL;
3192
3193 if (DIFF_PAIR_UNMERGED(p)) {
3194 run_diff_cmd(pgm, name, NULL, attr_path,
3195 NULL, NULL, NULL, o, p);
3196 return;
3197 }
3198
3199 diff_fill_sha1_info(one);
3200 diff_fill_sha1_info(two);
3201
3202 if (!pgm &&
3203 DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3204 (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3205 /*
3206 * a filepair that changes between file and symlink
3207 * needs to be split into deletion and creation.
3208 */
3209 struct diff_filespec *null = alloc_filespec(two->path);
3210 run_diff_cmd(NULL, name, other, attr_path,
3211 one, null, &msg, o, p);
3212 free(null);
3213 strbuf_release(&msg);
3214
3215 null = alloc_filespec(one->path);
3216 run_diff_cmd(NULL, name, other, attr_path,
3217 null, two, &msg, o, p);
3218 free(null);
3219 }
3220 else
3221 run_diff_cmd(pgm, name, other, attr_path,
3222 one, two, &msg, o, p);
3223
3224 strbuf_release(&msg);
3225 }
3226
3227 static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3228 struct diffstat_t *diffstat)
3229 {
3230 const char *name;
3231 const char *other;
3232
3233 if (DIFF_PAIR_UNMERGED(p)) {
3234 /* unmerged */
3235 builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, p);
3236 return;
3237 }
3238
3239 name = p->one->path;
3240 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3241
3242 if (o->prefix_length)
3243 strip_prefix(o->prefix_length, &name, &other);
3244
3245 diff_fill_sha1_info(p->one);
3246 diff_fill_sha1_info(p->two);
3247
3248 builtin_diffstat(name, other, p->one, p->two, diffstat, o, p);
3249 }
3250
3251 static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3252 {
3253 const char *name;
3254 const char *other;
3255 const char *attr_path;
3256
3257 if (DIFF_PAIR_UNMERGED(p)) {
3258 /* unmerged */
3259 return;
3260 }
3261
3262 name = p->one->path;
3263 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3264 attr_path = other ? other : name;
3265
3266 if (o->prefix_length)
3267 strip_prefix(o->prefix_length, &name, &other);
3268
3269 diff_fill_sha1_info(p->one);
3270 diff_fill_sha1_info(p->two);
3271
3272 builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3273 }
3274
3275 void diff_setup(struct diff_options *options)
3276 {
3277 memcpy(options, &default_diff_options, sizeof(*options));
3278
3279 options->file = stdout;
3280
3281 options->line_termination = '\n';
3282 options->break_opt = -1;
3283 options->rename_limit = -1;
3284 options->dirstat_permille = diff_dirstat_permille_default;
3285 options->context = diff_context_default;
3286 options->ws_error_highlight = WSEH_NEW;
3287 DIFF_OPT_SET(options, RENAME_EMPTY);
3288
3289 /* pathchange left =NULL by default */
3290 options->change = diff_change;
3291 options->add_remove = diff_addremove;
3292 options->use_color = diff_use_color_default;
3293 options->detect_rename = diff_detect_rename_default;
3294 options->xdl_opts |= diff_algorithm;
3295 if (diff_indent_heuristic)
3296 DIFF_XDL_SET(options, INDENT_HEURISTIC);
3297 else if (diff_compaction_heuristic)
3298 DIFF_XDL_SET(options, COMPACTION_HEURISTIC);
3299
3300 options->orderfile = diff_order_file_cfg;
3301
3302 if (diff_no_prefix) {
3303 options->a_prefix = options->b_prefix = "";
3304 } else if (!diff_mnemonic_prefix) {
3305 options->a_prefix = "a/";
3306 options->b_prefix = "b/";
3307 }
3308 }
3309
3310 void diff_setup_done(struct diff_options *options)
3311 {
3312 int count = 0;
3313
3314 if (options->set_default)
3315 options->set_default(options);
3316
3317 if (options->output_format & DIFF_FORMAT_NAME)
3318 count++;
3319 if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3320 count++;
3321 if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3322 count++;
3323 if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3324 count++;
3325 if (count > 1)
3326 die("--name-only, --name-status, --check and -s are mutually exclusive");
3327
3328 /*
3329 * Most of the time we can say "there are changes"
3330 * only by checking if there are changed paths, but
3331 * --ignore-whitespace* options force us to look
3332 * inside contents.
3333 */
3334
3335 if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3336 DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3337 DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3338 DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3339 else
3340 DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3341
3342 if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3343 options->detect_rename = DIFF_DETECT_COPY;
3344
3345 if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3346 options->prefix = NULL;
3347 if (options->prefix)
3348 options->prefix_length = strlen(options->prefix);
3349 else
3350 options->prefix_length = 0;
3351
3352 if (options->output_format & (DIFF_FORMAT_NAME |
3353 DIFF_FORMAT_NAME_STATUS |
3354 DIFF_FORMAT_CHECKDIFF |
3355 DIFF_FORMAT_NO_OUTPUT))
3356 options->output_format &= ~(DIFF_FORMAT_RAW |
3357 DIFF_FORMAT_NUMSTAT |
3358 DIFF_FORMAT_DIFFSTAT |
3359 DIFF_FORMAT_SHORTSTAT |
3360 DIFF_FORMAT_DIRSTAT |
3361 DIFF_FORMAT_SUMMARY |
3362 DIFF_FORMAT_PATCH);
3363
3364 /*
3365 * These cases always need recursive; we do not drop caller-supplied
3366 * recursive bits for other formats here.
3367 */
3368 if (options->output_format & (DIFF_FORMAT_PATCH |
3369 DIFF_FORMAT_NUMSTAT |
3370 DIFF_FORMAT_DIFFSTAT |
3371 DIFF_FORMAT_SHORTSTAT |
3372 DIFF_FORMAT_DIRSTAT |
3373 DIFF_FORMAT_SUMMARY |
3374 DIFF_FORMAT_CHECKDIFF))
3375 DIFF_OPT_SET(options, RECURSIVE);
3376 /*
3377 * Also pickaxe would not work very well if you do not say recursive
3378 */
3379 if (options->pickaxe)
3380 DIFF_OPT_SET(options, RECURSIVE);
3381 /*
3382 * When patches are generated, submodules diffed against the work tree
3383 * must be checked for dirtiness too so it can be shown in the output
3384 */
3385 if (options->output_format & DIFF_FORMAT_PATCH)
3386 DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3387
3388 if (options->detect_rename && options->rename_limit < 0)
3389 options->rename_limit = diff_rename_limit_default;
3390 if (options->setup & DIFF_SETUP_USE_CACHE) {
3391 if (!active_cache)
3392 /* read-cache does not die even when it fails
3393 * so it is safe for us to do this here. Also
3394 * it does not smudge active_cache or active_nr
3395 * when it fails, so we do not have to worry about
3396 * cleaning it up ourselves either.
3397 */
3398 read_cache();
3399 }
3400 if (options->abbrev <= 0 || 40 < options->abbrev)
3401 options->abbrev = 40; /* full */
3402
3403 /*
3404 * It does not make sense to show the first hit we happened
3405 * to have found. It does not make sense not to return with
3406 * exit code in such a case either.
3407 */
3408 if (DIFF_OPT_TST(options, QUICK)) {
3409 options->output_format = DIFF_FORMAT_NO_OUTPUT;
3410 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3411 }
3412
3413 options->diff_path_counter = 0;
3414
3415 if (DIFF_OPT_TST(options, FOLLOW_RENAMES) && options->pathspec.nr != 1)
3416 die(_("--follow requires exactly one pathspec"));
3417 }
3418
3419 static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3420 {
3421 char c, *eq;
3422 int len;
3423
3424 if (*arg != '-')
3425 return 0;
3426 c = *++arg;
3427 if (!c)
3428 return 0;
3429 if (c == arg_short) {
3430 c = *++arg;
3431 if (!c)
3432 return 1;
3433 if (val && isdigit(c)) {
3434 char *end;
3435 int n = strtoul(arg, &end, 10);
3436 if (*end)
3437 return 0;
3438 *val = n;
3439 return 1;
3440 }
3441 return 0;
3442 }
3443 if (c != '-')
3444 return 0;
3445 arg++;
3446 eq = strchrnul(arg, '=');
3447 len = eq - arg;
3448 if (!len || strncmp(arg, arg_long, len))
3449 return 0;
3450 if (*eq) {
3451 int n;
3452 char *end;
3453 if (!isdigit(*++eq))
3454 return 0;
3455 n = strtoul(eq, &end, 10);
3456 if (*end)
3457 return 0;
3458 *val = n;
3459 }
3460 return 1;
3461 }
3462
3463 static int diff_scoreopt_parse(const char *opt);
3464
3465 static inline int short_opt(char opt, const char **argv,
3466 const char **optarg)
3467 {
3468 const char *arg = argv[0];
3469 if (arg[0] != '-' || arg[1] != opt)
3470 return 0;
3471 if (arg[2] != '\0') {
3472 *optarg = arg + 2;
3473 return 1;
3474 }
3475 if (!argv[1])
3476 die("Option '%c' requires a value", opt);
3477 *optarg = argv[1];
3478 return 2;
3479 }
3480
3481 int parse_long_opt(const char *opt, const char **argv,
3482 const char **optarg)
3483 {
3484 const char *arg = argv[0];
3485 if (!skip_prefix(arg, "--", &arg))
3486 return 0;
3487 if (!skip_prefix(arg, opt, &arg))
3488 return 0;
3489 if (*arg == '=') { /* stuck form: --option=value */
3490 *optarg = arg + 1;
3491 return 1;
3492 }
3493 if (*arg != '\0')
3494 return 0;
3495 /* separate form: --option value */
3496 if (!argv[1])
3497 die("Option '--%s' requires a value", opt);
3498 *optarg = argv[1];
3499 return 2;
3500 }
3501
3502 static int stat_opt(struct diff_options *options, const char **av)
3503 {
3504 const char *arg = av[0];
3505 char *end;
3506 int width = options->stat_width;
3507 int name_width = options->stat_name_width;
3508 int graph_width = options->stat_graph_width;
3509 int count = options->stat_count;
3510 int argcount = 1;
3511
3512 if (!skip_prefix(arg, "--stat", &arg))
3513 die("BUG: stat option does not begin with --stat: %s", arg);
3514 end = (char *)arg;
3515
3516 switch (*arg) {
3517 case '-':
3518 if (skip_prefix(arg, "-width", &arg)) {
3519 if (*arg == '=')
3520 width = strtoul(arg + 1, &end, 10);
3521 else if (!*arg && !av[1])
3522 die("Option '--stat-width' requires a value");
3523 else if (!*arg) {
3524 width = strtoul(av[1], &end, 10);
3525 argcount = 2;
3526 }
3527 } else if (skip_prefix(arg, "-name-width", &arg)) {
3528 if (*arg == '=')
3529 name_width = strtoul(arg + 1, &end, 10);
3530 else if (!*arg && !av[1])
3531 die("Option '--stat-name-width' requires a value");
3532 else if (!*arg) {
3533 name_width = strtoul(av[1], &end, 10);
3534 argcount = 2;
3535 }
3536 } else if (skip_prefix(arg, "-graph-width", &arg)) {
3537 if (*arg == '=')
3538 graph_width = strtoul(arg + 1, &end, 10);
3539 else if (!*arg && !av[1])
3540 die("Option '--stat-graph-width' requires a value");
3541 else if (!*arg) {
3542 graph_width = strtoul(av[1], &end, 10);
3543 argcount = 2;
3544 }
3545 } else if (skip_prefix(arg, "-count", &arg)) {
3546 if (*arg == '=')
3547 count = strtoul(arg + 1, &end, 10);
3548 else if (!*arg && !av[1])
3549 die("Option '--stat-count' requires a value");
3550 else if (!*arg) {
3551 count = strtoul(av[1], &end, 10);
3552 argcount = 2;
3553 }
3554 }
3555 break;
3556 case '=':
3557 width = strtoul(arg+1, &end, 10);
3558 if (*end == ',')
3559 name_width = strtoul(end+1, &end, 10);
3560 if (*end == ',')
3561 count = strtoul(end+1, &end, 10);
3562 }
3563
3564 /* Important! This checks all the error cases! */
3565 if (*end)
3566 return 0;
3567 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3568 options->stat_name_width = name_width;
3569 options->stat_graph_width = graph_width;
3570 options->stat_width = width;
3571 options->stat_count = count;
3572 return argcount;
3573 }
3574
3575 static int parse_dirstat_opt(struct diff_options *options, const char *params)
3576 {
3577 struct strbuf errmsg = STRBUF_INIT;
3578 if (parse_dirstat_params(options, params, &errmsg))
3579 die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3580 errmsg.buf);
3581 strbuf_release(&errmsg);
3582 /*
3583 * The caller knows a dirstat-related option is given from the command
3584 * line; allow it to say "return this_function();"
3585 */
3586 options->output_format |= DIFF_FORMAT_DIRSTAT;
3587 return 1;
3588 }
3589
3590 static int parse_submodule_opt(struct diff_options *options, const char *value)
3591 {
3592 if (parse_submodule_params(options, value))
3593 die(_("Failed to parse --submodule option parameter: '%s'"),
3594 value);
3595 return 1;
3596 }
3597
3598 static const char diff_status_letters[] = {
3599 DIFF_STATUS_ADDED,
3600 DIFF_STATUS_COPIED,
3601 DIFF_STATUS_DELETED,
3602 DIFF_STATUS_MODIFIED,
3603 DIFF_STATUS_RENAMED,
3604 DIFF_STATUS_TYPE_CHANGED,
3605 DIFF_STATUS_UNKNOWN,
3606 DIFF_STATUS_UNMERGED,
3607 DIFF_STATUS_FILTER_AON,
3608 DIFF_STATUS_FILTER_BROKEN,
3609 '\0',
3610 };
3611
3612 static unsigned int filter_bit['Z' + 1];
3613
3614 static void prepare_filter_bits(void)
3615 {
3616 int i;
3617
3618 if (!filter_bit[DIFF_STATUS_ADDED]) {
3619 for (i = 0; diff_status_letters[i]; i++)
3620 filter_bit[(int) diff_status_letters[i]] = (1 << i);
3621 }
3622 }
3623
3624 static unsigned filter_bit_tst(char status, const struct diff_options *opt)
3625 {
3626 return opt->filter & filter_bit[(int) status];
3627 }
3628
3629 static int parse_diff_filter_opt(const char *optarg, struct diff_options *opt)
3630 {
3631 int i, optch;
3632
3633 prepare_filter_bits();
3634
3635 /*
3636 * If there is a negation e.g. 'd' in the input, and we haven't
3637 * initialized the filter field with another --diff-filter, start
3638 * from full set of bits, except for AON.
3639 */
3640 if (!opt->filter) {
3641 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3642 if (optch < 'a' || 'z' < optch)
3643 continue;
3644 opt->filter = (1 << (ARRAY_SIZE(diff_status_letters) - 1)) - 1;
3645 opt->filter &= ~filter_bit[DIFF_STATUS_FILTER_AON];
3646 break;
3647 }
3648 }
3649
3650 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3651 unsigned int bit;
3652 int negate;
3653
3654 if ('a' <= optch && optch <= 'z') {
3655 negate = 1;
3656 optch = toupper(optch);
3657 } else {
3658 negate = 0;
3659 }
3660
3661 bit = (0 <= optch && optch <= 'Z') ? filter_bit[optch] : 0;
3662 if (!bit)
3663 return optarg[i];
3664 if (negate)
3665 opt->filter &= ~bit;
3666 else
3667 opt->filter |= bit;
3668 }
3669 return 0;
3670 }
3671
3672 static void enable_patch_output(int *fmt) {
3673 *fmt &= ~DIFF_FORMAT_NO_OUTPUT;
3674 *fmt |= DIFF_FORMAT_PATCH;
3675 }
3676
3677 static int parse_one_token(const char **arg, const char *token)
3678 {
3679 const char *rest;
3680 if (skip_prefix(*arg, token, &rest) && (!*rest || *rest == ',')) {
3681 *arg = rest;
3682 return 1;
3683 }
3684 return 0;
3685 }
3686
3687 static int parse_ws_error_highlight(struct diff_options *opt, const char *arg)
3688 {
3689 const char *orig_arg = arg;
3690 unsigned val = 0;
3691 while (*arg) {
3692 if (parse_one_token(&arg, "none"))
3693 val = 0;
3694 else if (parse_one_token(&arg, "default"))
3695 val = WSEH_NEW;
3696 else if (parse_one_token(&arg, "all"))
3697 val = WSEH_NEW | WSEH_OLD | WSEH_CONTEXT;
3698 else if (parse_one_token(&arg, "new"))
3699 val |= WSEH_NEW;
3700 else if (parse_one_token(&arg, "old"))
3701 val |= WSEH_OLD;
3702 else if (parse_one_token(&arg, "context"))
3703 val |= WSEH_CONTEXT;
3704 else {
3705 error("unknown value after ws-error-highlight=%.*s",
3706 (int)(arg - orig_arg), orig_arg);
3707 return 0;
3708 }
3709 if (*arg)
3710 arg++;
3711 }
3712 opt->ws_error_highlight = val;
3713 return 1;
3714 }
3715
3716 int diff_opt_parse(struct diff_options *options,
3717 const char **av, int ac, const char *prefix)
3718 {
3719 const char *arg = av[0];
3720 const char *optarg;
3721 int argcount;
3722
3723 if (!prefix)
3724 prefix = "";
3725
3726 /* Output format options */
3727 if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch")
3728 || opt_arg(arg, 'U', "unified", &options->context))
3729 enable_patch_output(&options->output_format);
3730 else if (!strcmp(arg, "--raw"))
3731 options->output_format |= DIFF_FORMAT_RAW;
3732 else if (!strcmp(arg, "--patch-with-raw")) {
3733 enable_patch_output(&options->output_format);
3734 options->output_format |= DIFF_FORMAT_RAW;
3735 } else if (!strcmp(arg, "--numstat"))
3736 options->output_format |= DIFF_FORMAT_NUMSTAT;
3737 else if (!strcmp(arg, "--shortstat"))
3738 options->output_format |= DIFF_FORMAT_SHORTSTAT;
3739 else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3740 return parse_dirstat_opt(options, "");
3741 else if (skip_prefix(arg, "-X", &arg))
3742 return parse_dirstat_opt(options, arg);
3743 else if (skip_prefix(arg, "--dirstat=", &arg))
3744 return parse_dirstat_opt(options, arg);
3745 else if (!strcmp(arg, "--cumulative"))
3746 return parse_dirstat_opt(options, "cumulative");
3747 else if (!strcmp(arg, "--dirstat-by-file"))
3748 return parse_dirstat_opt(options, "files");
3749 else if (skip_prefix(arg, "--dirstat-by-file=", &arg)) {
3750 parse_dirstat_opt(options, "files");
3751 return parse_dirstat_opt(options, arg);
3752 }
3753 else if (!strcmp(arg, "--check"))
3754 options->output_format |= DIFF_FORMAT_CHECKDIFF;
3755 else if (!strcmp(arg, "--summary"))
3756 options->output_format |= DIFF_FORMAT_SUMMARY;
3757 else if (!strcmp(arg, "--patch-with-stat")) {
3758 enable_patch_output(&options->output_format);
3759 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3760 } else if (!strcmp(arg, "--name-only"))
3761 options->output_format |= DIFF_FORMAT_NAME;
3762 else if (!strcmp(arg, "--name-status"))
3763 options->output_format |= DIFF_FORMAT_NAME_STATUS;
3764 else if (!strcmp(arg, "-s") || !strcmp(arg, "--no-patch"))
3765 options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3766 else if (starts_with(arg, "--stat"))
3767 /* --stat, --stat-width, --stat-name-width, or --stat-count */
3768 return stat_opt(options, av);
3769
3770 /* renames options */
3771 else if (starts_with(arg, "-B") || starts_with(arg, "--break-rewrites=") ||
3772 !strcmp(arg, "--break-rewrites")) {
3773 if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3774 return error("invalid argument to -B: %s", arg+2);
3775 }
3776 else if (starts_with(arg, "-M") || starts_with(arg, "--find-renames=") ||
3777 !strcmp(arg, "--find-renames")) {
3778 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3779 return error("invalid argument to -M: %s", arg+2);
3780 options->detect_rename = DIFF_DETECT_RENAME;
3781 }
3782 else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3783 options->irreversible_delete = 1;
3784 }
3785 else if (starts_with(arg, "-C") || starts_with(arg, "--find-copies=") ||
3786 !strcmp(arg, "--find-copies")) {
3787 if (options->detect_rename == DIFF_DETECT_COPY)
3788 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3789 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3790 return error("invalid argument to -C: %s", arg+2);
3791 options->detect_rename = DIFF_DETECT_COPY;
3792 }
3793 else if (!strcmp(arg, "--no-renames"))
3794 options->detect_rename = 0;
3795 else if (!strcmp(arg, "--rename-empty"))
3796 DIFF_OPT_SET(options, RENAME_EMPTY);
3797 else if (!strcmp(arg, "--no-rename-empty"))
3798 DIFF_OPT_CLR(options, RENAME_EMPTY);
3799 else if (!strcmp(arg, "--relative"))
3800 DIFF_OPT_SET(options, RELATIVE_NAME);
3801 else if (skip_prefix(arg, "--relative=", &arg)) {
3802 DIFF_OPT_SET(options, RELATIVE_NAME);
3803 options->prefix = arg;
3804 }
3805
3806 /* xdiff options */
3807 else if (!strcmp(arg, "--minimal"))
3808 DIFF_XDL_SET(options, NEED_MINIMAL);
3809 else if (!strcmp(arg, "--no-minimal"))
3810 DIFF_XDL_CLR(options, NEED_MINIMAL);
3811 else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3812 DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3813 else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3814 DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3815 else if (!strcmp(arg, "--ignore-space-at-eol"))
3816 DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3817 else if (!strcmp(arg, "--ignore-blank-lines"))
3818 DIFF_XDL_SET(options, IGNORE_BLANK_LINES);
3819 else if (!strcmp(arg, "--indent-heuristic")) {
3820 DIFF_XDL_SET(options, INDENT_HEURISTIC);
3821 DIFF_XDL_CLR(options, COMPACTION_HEURISTIC);
3822 } else if (!strcmp(arg, "--no-indent-heuristic"))
3823 DIFF_XDL_CLR(options, INDENT_HEURISTIC);
3824 else if (!strcmp(arg, "--compaction-heuristic")) {
3825 DIFF_XDL_SET(options, COMPACTION_HEURISTIC);
3826 DIFF_XDL_CLR(options, INDENT_HEURISTIC);
3827 } else if (!strcmp(arg, "--no-compaction-heuristic"))
3828 DIFF_XDL_CLR(options, COMPACTION_HEURISTIC);
3829 else if (!strcmp(arg, "--patience"))
3830 options->xdl_opts = DIFF_WITH_ALG(options, PATIENCE_DIFF);
3831 else if (!strcmp(arg, "--histogram"))
3832 options->xdl_opts = DIFF_WITH_ALG(options, HISTOGRAM_DIFF);
3833 else if ((argcount = parse_long_opt("diff-algorithm", av, &optarg))) {
3834 long value = parse_algorithm_value(optarg);
3835 if (value < 0)
3836 return error("option diff-algorithm accepts \"myers\", "
3837 "\"minimal\", \"patience\" and \"histogram\"");
3838 /* clear out previous settings */
3839 DIFF_XDL_CLR(options, NEED_MINIMAL);
3840 options->xdl_opts &= ~XDF_DIFF_ALGORITHM_MASK;
3841 options->xdl_opts |= value;
3842 return argcount;
3843 }
3844
3845 /* flags options */
3846 else if (!strcmp(arg, "--binary")) {
3847 enable_patch_output(&options->output_format);
3848 DIFF_OPT_SET(options, BINARY);
3849 }
3850 else if (!strcmp(arg, "--full-index"))
3851 DIFF_OPT_SET(options, FULL_INDEX);
3852 else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
3853 DIFF_OPT_SET(options, TEXT);
3854 else if (!strcmp(arg, "-R"))
3855 DIFF_OPT_SET(options, REVERSE_DIFF);
3856 else if (!strcmp(arg, "--find-copies-harder"))
3857 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3858 else if (!strcmp(arg, "--follow"))
3859 DIFF_OPT_SET(options, FOLLOW_RENAMES);
3860 else if (!strcmp(arg, "--no-follow")) {
3861 DIFF_OPT_CLR(options, FOLLOW_RENAMES);
3862 DIFF_OPT_CLR(options, DEFAULT_FOLLOW_RENAMES);
3863 } else if (!strcmp(arg, "--color"))
3864 options->use_color = 1;
3865 else if (skip_prefix(arg, "--color=", &arg)) {
3866 int value = git_config_colorbool(NULL, arg);
3867 if (value < 0)
3868 return error("option `color' expects \"always\", \"auto\", or \"never\"");
3869 options->use_color = value;
3870 }
3871 else if (!strcmp(arg, "--no-color"))
3872 options->use_color = 0;
3873 else if (!strcmp(arg, "--color-words")) {
3874 options->use_color = 1;
3875 options->word_diff = DIFF_WORDS_COLOR;
3876 }
3877 else if (skip_prefix(arg, "--color-words=", &arg)) {
3878 options->use_color = 1;
3879 options->word_diff = DIFF_WORDS_COLOR;
3880 options->word_regex = arg;
3881 }
3882 else if (!strcmp(arg, "--word-diff")) {
3883 if (options->word_diff == DIFF_WORDS_NONE)
3884 options->word_diff = DIFF_WORDS_PLAIN;
3885 }
3886 else if (skip_prefix(arg, "--word-diff=", &arg)) {
3887 if (!strcmp(arg, "plain"))
3888 options->word_diff = DIFF_WORDS_PLAIN;
3889 else if (!strcmp(arg, "color")) {
3890 options->use_color = 1;
3891 options->word_diff = DIFF_WORDS_COLOR;
3892 }
3893 else if (!strcmp(arg, "porcelain"))
3894 options->word_diff = DIFF_WORDS_PORCELAIN;
3895 else if (!strcmp(arg, "none"))
3896 options->word_diff = DIFF_WORDS_NONE;
3897 else
3898 die("bad --word-diff argument: %s", arg);
3899 }
3900 else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
3901 if (options->word_diff == DIFF_WORDS_NONE)
3902 options->word_diff = DIFF_WORDS_PLAIN;
3903 options->word_regex = optarg;
3904 return argcount;
3905 }
3906 else if (!strcmp(arg, "--exit-code"))
3907 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3908 else if (!strcmp(arg, "--quiet"))
3909 DIFF_OPT_SET(options, QUICK);
3910 else if (!strcmp(arg, "--ext-diff"))
3911 DIFF_OPT_SET(options, ALLOW_EXTERNAL);
3912 else if (!strcmp(arg, "--no-ext-diff"))
3913 DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
3914 else if (!strcmp(arg, "--textconv"))
3915 DIFF_OPT_SET(options, ALLOW_TEXTCONV);
3916 else if (!strcmp(arg, "--no-textconv"))
3917 DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
3918 else if (!strcmp(arg, "--ignore-submodules")) {
3919 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3920 handle_ignore_submodules_arg(options, "all");
3921 } else if (skip_prefix(arg, "--ignore-submodules=", &arg)) {
3922 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3923 handle_ignore_submodules_arg(options, arg);
3924 } else if (!strcmp(arg, "--submodule"))
3925 DIFF_OPT_SET(options, SUBMODULE_LOG);
3926 else if (skip_prefix(arg, "--submodule=", &arg))
3927 return parse_submodule_opt(options, arg);
3928 else if (skip_prefix(arg, "--ws-error-highlight=", &arg))
3929 return parse_ws_error_highlight(options, arg);
3930
3931 /* misc options */
3932 else if (!strcmp(arg, "-z"))
3933 options->line_termination = 0;
3934 else if ((argcount = short_opt('l', av, &optarg))) {
3935 options->rename_limit = strtoul(optarg, NULL, 10);
3936 return argcount;
3937 }
3938 else if ((argcount = short_opt('S', av, &optarg))) {
3939 options->pickaxe = optarg;
3940 options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
3941 return argcount;
3942 } else if ((argcount = short_opt('G', av, &optarg))) {
3943 options->pickaxe = optarg;
3944 options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
3945 return argcount;
3946 }
3947 else if (!strcmp(arg, "--pickaxe-all"))
3948 options->pickaxe_opts |= DIFF_PICKAXE_ALL;
3949 else if (!strcmp(arg, "--pickaxe-regex"))
3950 options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
3951 else if ((argcount = short_opt('O', av, &optarg))) {
3952 const char *path = prefix_filename(prefix, strlen(prefix), optarg);
3953 options->orderfile = xstrdup(path);
3954 return argcount;
3955 }
3956 else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
3957 int offending = parse_diff_filter_opt(optarg, options);
3958 if (offending)
3959 die("unknown change class '%c' in --diff-filter=%s",
3960 offending, optarg);
3961 return argcount;
3962 }
3963 else if (!strcmp(arg, "--abbrev"))
3964 options->abbrev = DEFAULT_ABBREV;
3965 else if (skip_prefix(arg, "--abbrev=", &arg)) {
3966 options->abbrev = strtoul(arg, NULL, 10);
3967 if (options->abbrev < MINIMUM_ABBREV)
3968 options->abbrev = MINIMUM_ABBREV;
3969 else if (40 < options->abbrev)
3970 options->abbrev = 40;
3971 }
3972 else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
3973 options->a_prefix = optarg;
3974 return argcount;
3975 }
3976 else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
3977 options->b_prefix = optarg;
3978 return argcount;
3979 }
3980 else if (!strcmp(arg, "--no-prefix"))
3981 options->a_prefix = options->b_prefix = "";
3982 else if (opt_arg(arg, '\0', "inter-hunk-context",
3983 &options->interhunkcontext))
3984 ;
3985 else if (!strcmp(arg, "-W"))
3986 DIFF_OPT_SET(options, FUNCCONTEXT);
3987 else if (!strcmp(arg, "--function-context"))
3988 DIFF_OPT_SET(options, FUNCCONTEXT);
3989 else if (!strcmp(arg, "--no-function-context"))
3990 DIFF_OPT_CLR(options, FUNCCONTEXT);
3991 else if ((argcount = parse_long_opt("output", av, &optarg))) {
3992 const char *path = prefix_filename(prefix, strlen(prefix), optarg);
3993 options->file = fopen(path, "w");
3994 if (!options->file)
3995 die_errno("Could not open '%s'", path);
3996 options->close_file = 1;
3997 return argcount;
3998 } else
3999 return 0;
4000 return 1;
4001 }
4002
4003 int parse_rename_score(const char **cp_p)
4004 {
4005 unsigned long num, scale;
4006 int ch, dot;
4007 const char *cp = *cp_p;
4008
4009 num = 0;
4010 scale = 1;
4011 dot = 0;
4012 for (;;) {
4013 ch = *cp;
4014 if ( !dot && ch == '.' ) {
4015 scale = 1;
4016 dot = 1;
4017 } else if ( ch == '%' ) {
4018 scale = dot ? scale*100 : 100;
4019 cp++; /* % is always at the end */
4020 break;
4021 } else if ( ch >= '0' && ch <= '9' ) {
4022 if ( scale < 100000 ) {
4023 scale *= 10;
4024 num = (num*10) + (ch-'0');
4025 }
4026 } else {
4027 break;
4028 }
4029 cp++;
4030 }
4031 *cp_p = cp;
4032
4033 /* user says num divided by scale and we say internally that
4034 * is MAX_SCORE * num / scale.
4035 */
4036 return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
4037 }
4038
4039 static int diff_scoreopt_parse(const char *opt)
4040 {
4041 int opt1, opt2, cmd;
4042
4043 if (*opt++ != '-')
4044 return -1;
4045 cmd = *opt++;
4046 if (cmd == '-') {
4047 /* convert the long-form arguments into short-form versions */
4048 if (skip_prefix(opt, "break-rewrites", &opt)) {
4049 if (*opt == 0 || *opt++ == '=')
4050 cmd = 'B';
4051 } else if (skip_prefix(opt, "find-copies", &opt)) {
4052 if (*opt == 0 || *opt++ == '=')
4053 cmd = 'C';
4054 } else if (skip_prefix(opt, "find-renames", &opt)) {
4055 if (*opt == 0 || *opt++ == '=')
4056 cmd = 'M';
4057 }
4058 }
4059 if (cmd != 'M' && cmd != 'C' && cmd != 'B')
4060 return -1; /* that is not a -M, -C, or -B option */
4061
4062 opt1 = parse_rename_score(&opt);
4063 if (cmd != 'B')
4064 opt2 = 0;
4065 else {
4066 if (*opt == 0)
4067 opt2 = 0;
4068 else if (*opt != '/')
4069 return -1; /* we expect -B80/99 or -B80 */
4070 else {
4071 opt++;
4072 opt2 = parse_rename_score(&opt);
4073 }
4074 }
4075 if (*opt != 0)
4076 return -1;
4077 return opt1 | (opt2 << 16);
4078 }
4079
4080 struct diff_queue_struct diff_queued_diff;
4081
4082 void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
4083 {
4084 ALLOC_GROW(queue->queue, queue->nr + 1, queue->alloc);
4085 queue->queue[queue->nr++] = dp;
4086 }
4087
4088 struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
4089 struct diff_filespec *one,
4090 struct diff_filespec *two)
4091 {
4092 struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
4093 dp->one = one;
4094 dp->two = two;
4095 if (queue)
4096 diff_q(queue, dp);
4097 return dp;
4098 }
4099
4100 void diff_free_filepair(struct diff_filepair *p)
4101 {
4102 free_filespec(p->one);
4103 free_filespec(p->two);
4104 free(p);
4105 }
4106
4107 /* This is different from find_unique_abbrev() in that
4108 * it stuffs the result with dots for alignment.
4109 */
4110 const char *diff_unique_abbrev(const unsigned char *sha1, int len)
4111 {
4112 int abblen;
4113 const char *abbrev;
4114 if (len == 40)
4115 return sha1_to_hex(sha1);
4116
4117 abbrev = find_unique_abbrev(sha1, len);
4118 abblen = strlen(abbrev);
4119 if (abblen < 37) {
4120 static char hex[41];
4121 if (len < abblen && abblen <= len + 2)
4122 xsnprintf(hex, sizeof(hex), "%s%.*s", abbrev, len+3-abblen, "..");
4123 else
4124 xsnprintf(hex, sizeof(hex), "%s...", abbrev);
4125 return hex;
4126 }
4127 return sha1_to_hex(sha1);
4128 }
4129
4130 static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
4131 {
4132 int line_termination = opt->line_termination;
4133 int inter_name_termination = line_termination ? '\t' : '\0';
4134
4135 fprintf(opt->file, "%s", diff_line_prefix(opt));
4136 if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
4137 fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
4138 diff_unique_abbrev(p->one->sha1, opt->abbrev));
4139 fprintf(opt->file, "%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
4140 }
4141 if (p->score) {
4142 fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
4143 inter_name_termination);
4144 } else {
4145 fprintf(opt->file, "%c%c", p->status, inter_name_termination);
4146 }
4147
4148 if (p->status == DIFF_STATUS_COPIED ||
4149 p->status == DIFF_STATUS_RENAMED) {
4150 const char *name_a, *name_b;
4151 name_a = p->one->path;
4152 name_b = p->two->path;
4153 strip_prefix(opt->prefix_length, &name_a, &name_b);
4154 write_name_quoted(name_a, opt->file, inter_name_termination);
4155 write_name_quoted(name_b, opt->file, line_termination);
4156 } else {
4157 const char *name_a, *name_b;
4158 name_a = p->one->mode ? p->one->path : p->two->path;
4159 name_b = NULL;
4160 strip_prefix(opt->prefix_length, &name_a, &name_b);
4161 write_name_quoted(name_a, opt->file, line_termination);
4162 }
4163 }
4164
4165 int diff_unmodified_pair(struct diff_filepair *p)
4166 {
4167 /* This function is written stricter than necessary to support
4168 * the currently implemented transformers, but the idea is to
4169 * let transformers to produce diff_filepairs any way they want,
4170 * and filter and clean them up here before producing the output.
4171 */
4172 struct diff_filespec *one = p->one, *two = p->two;
4173
4174 if (DIFF_PAIR_UNMERGED(p))
4175 return 0; /* unmerged is interesting */
4176
4177 /* deletion, addition, mode or type change
4178 * and rename are all interesting.
4179 */
4180 if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
4181 DIFF_PAIR_MODE_CHANGED(p) ||
4182 strcmp(one->path, two->path))
4183 return 0;
4184
4185 /* both are valid and point at the same path. that is, we are
4186 * dealing with a change.
4187 */
4188 if (one->sha1_valid && two->sha1_valid &&
4189 !hashcmp(one->sha1, two->sha1) &&
4190 !one->dirty_submodule && !two->dirty_submodule)
4191 return 1; /* no change */
4192 if (!one->sha1_valid && !two->sha1_valid)
4193 return 1; /* both look at the same file on the filesystem. */
4194 return 0;
4195 }
4196
4197 static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
4198 {
4199 if (diff_unmodified_pair(p))
4200 return;
4201
4202 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4203 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4204 return; /* no tree diffs in patch format */
4205
4206 run_diff(p, o);
4207 }
4208
4209 static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
4210 struct diffstat_t *diffstat)
4211 {
4212 if (diff_unmodified_pair(p))
4213 return;
4214
4215 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4216 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4217 return; /* no useful stat for tree diffs */
4218
4219 run_diffstat(p, o, diffstat);
4220 }
4221
4222 static void diff_flush_checkdiff(struct diff_filepair *p,
4223 struct diff_options *o)
4224 {
4225 if (diff_unmodified_pair(p))
4226 return;
4227
4228 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4229 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4230 return; /* nothing to check in tree diffs */
4231
4232 run_checkdiff(p, o);
4233 }
4234
4235 int diff_queue_is_empty(void)
4236 {
4237 struct diff_queue_struct *q = &diff_queued_diff;
4238 int i;
4239 for (i = 0; i < q->nr; i++)
4240 if (!diff_unmodified_pair(q->queue[i]))
4241 return 0;
4242 return 1;
4243 }
4244
4245 #if DIFF_DEBUG
4246 void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
4247 {
4248 fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
4249 x, one ? one : "",
4250 s->path,
4251 DIFF_FILE_VALID(s) ? "valid" : "invalid",
4252 s->mode,
4253 s->sha1_valid ? sha1_to_hex(s->sha1) : "");
4254 fprintf(stderr, "queue[%d] %s size %lu\n",
4255 x, one ? one : "",
4256 s->size);
4257 }
4258
4259 void diff_debug_filepair(const struct diff_filepair *p, int i)
4260 {
4261 diff_debug_filespec(p->one, i, "one");
4262 diff_debug_filespec(p->two, i, "two");
4263 fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
4264 p->score, p->status ? p->status : '?',
4265 p->one->rename_used, p->broken_pair);
4266 }
4267
4268 void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
4269 {
4270 int i;
4271 if (msg)
4272 fprintf(stderr, "%s\n", msg);
4273 fprintf(stderr, "q->nr = %d\n", q->nr);
4274 for (i = 0; i < q->nr; i++) {
4275 struct diff_filepair *p = q->queue[i];
4276 diff_debug_filepair(p, i);
4277 }
4278 }
4279 #endif
4280
4281 static void diff_resolve_rename_copy(void)
4282 {
4283 int i;
4284 struct diff_filepair *p;
4285 struct diff_queue_struct *q = &diff_queued_diff;
4286
4287 diff_debug_queue("resolve-rename-copy", q);
4288
4289 for (i = 0; i < q->nr; i++) {
4290 p = q->queue[i];
4291 p->status = 0; /* undecided */
4292 if (DIFF_PAIR_UNMERGED(p))
4293 p->status = DIFF_STATUS_UNMERGED;
4294 else if (!DIFF_FILE_VALID(p->one))
4295 p->status = DIFF_STATUS_ADDED;
4296 else if (!DIFF_FILE_VALID(p->two))
4297 p->status = DIFF_STATUS_DELETED;
4298 else if (DIFF_PAIR_TYPE_CHANGED(p))
4299 p->status = DIFF_STATUS_TYPE_CHANGED;
4300
4301 /* from this point on, we are dealing with a pair
4302 * whose both sides are valid and of the same type, i.e.
4303 * either in-place edit or rename/copy edit.
4304 */
4305 else if (DIFF_PAIR_RENAME(p)) {
4306 /*
4307 * A rename might have re-connected a broken
4308 * pair up, causing the pathnames to be the
4309 * same again. If so, that's not a rename at
4310 * all, just a modification..
4311 *
4312 * Otherwise, see if this source was used for
4313 * multiple renames, in which case we decrement
4314 * the count, and call it a copy.
4315 */
4316 if (!strcmp(p->one->path, p->two->path))
4317 p->status = DIFF_STATUS_MODIFIED;
4318 else if (--p->one->rename_used > 0)
4319 p->status = DIFF_STATUS_COPIED;
4320 else
4321 p->status = DIFF_STATUS_RENAMED;
4322 }
4323 else if (hashcmp(p->one->sha1, p->two->sha1) ||
4324 p->one->mode != p->two->mode ||
4325 p->one->dirty_submodule ||
4326 p->two->dirty_submodule ||
4327 is_null_sha1(p->one->sha1))
4328 p->status = DIFF_STATUS_MODIFIED;
4329 else {
4330 /* This is a "no-change" entry and should not
4331 * happen anymore, but prepare for broken callers.
4332 */
4333 error("feeding unmodified %s to diffcore",
4334 p->one->path);
4335 p->status = DIFF_STATUS_UNKNOWN;
4336 }
4337 }
4338 diff_debug_queue("resolve-rename-copy done", q);
4339 }
4340
4341 static int check_pair_status(struct diff_filepair *p)
4342 {
4343 switch (p->status) {
4344 case DIFF_STATUS_UNKNOWN:
4345 return 0;
4346 case 0:
4347 die("internal error in diff-resolve-rename-copy");
4348 default:
4349 return 1;
4350 }
4351 }
4352
4353 static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
4354 {
4355 int fmt = opt->output_format;
4356
4357 if (fmt & DIFF_FORMAT_CHECKDIFF)
4358 diff_flush_checkdiff(p, opt);
4359 else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
4360 diff_flush_raw(p, opt);
4361 else if (fmt & DIFF_FORMAT_NAME) {
4362 const char *name_a, *name_b;
4363 name_a = p->two->path;
4364 name_b = NULL;
4365 strip_prefix(opt->prefix_length, &name_a, &name_b);
4366 write_name_quoted(name_a, opt->file, opt->line_termination);
4367 }
4368 }
4369
4370 static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
4371 {
4372 if (fs->mode)
4373 fprintf(file, " %s mode %06o ", newdelete, fs->mode);
4374 else
4375 fprintf(file, " %s ", newdelete);
4376 write_name_quoted(fs->path, file, '\n');
4377 }
4378
4379
4380 static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
4381 const char *line_prefix)
4382 {
4383 if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
4384 fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
4385 p->two->mode, show_name ? ' ' : '\n');
4386 if (show_name) {
4387 write_name_quoted(p->two->path, file, '\n');
4388 }
4389 }
4390 }
4391
4392 static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
4393 const char *line_prefix)
4394 {
4395 char *names = pprint_rename(p->one->path, p->two->path);
4396
4397 fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
4398 free(names);
4399 show_mode_change(file, p, 0, line_prefix);
4400 }
4401
4402 static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
4403 {
4404 FILE *file = opt->file;
4405 const char *line_prefix = diff_line_prefix(opt);
4406
4407 switch(p->status) {
4408 case DIFF_STATUS_DELETED:
4409 fputs(line_prefix, file);
4410 show_file_mode_name(file, "delete", p->one);
4411 break;
4412 case DIFF_STATUS_ADDED:
4413 fputs(line_prefix, file);
4414 show_file_mode_name(file, "create", p->two);
4415 break;
4416 case DIFF_STATUS_COPIED:
4417 fputs(line_prefix, file);
4418 show_rename_copy(file, "copy", p, line_prefix);
4419 break;
4420 case DIFF_STATUS_RENAMED:
4421 fputs(line_prefix, file);
4422 show_rename_copy(file, "rename", p, line_prefix);
4423 break;
4424 default:
4425 if (p->score) {
4426 fprintf(file, "%s rewrite ", line_prefix);
4427 write_name_quoted(p->two->path, file, ' ');
4428 fprintf(file, "(%d%%)\n", similarity_index(p));
4429 }
4430 show_mode_change(file, p, !p->score, line_prefix);
4431 break;
4432 }
4433 }
4434
4435 struct patch_id_t {
4436 git_SHA_CTX *ctx;
4437 int patchlen;
4438 };
4439
4440 static int remove_space(char *line, int len)
4441 {
4442 int i;
4443 char *dst = line;
4444 unsigned char c;
4445
4446 for (i = 0; i < len; i++)
4447 if (!isspace((c = line[i])))
4448 *dst++ = c;
4449
4450 return dst - line;
4451 }
4452
4453 static void patch_id_consume(void *priv, char *line, unsigned long len)
4454 {
4455 struct patch_id_t *data = priv;
4456 int new_len;
4457
4458 /* Ignore line numbers when computing the SHA1 of the patch */
4459 if (starts_with(line, "@@ -"))
4460 return;
4461
4462 new_len = remove_space(line, len);
4463
4464 git_SHA1_Update(data->ctx, line, new_len);
4465 data->patchlen += new_len;
4466 }
4467
4468 /* returns 0 upon success, and writes result into sha1 */
4469 static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
4470 {
4471 struct diff_queue_struct *q = &diff_queued_diff;
4472 int i;
4473 git_SHA_CTX ctx;
4474 struct patch_id_t data;
4475 char buffer[PATH_MAX * 4 + 20];
4476
4477 git_SHA1_Init(&ctx);
4478 memset(&data, 0, sizeof(struct patch_id_t));
4479 data.ctx = &ctx;
4480
4481 for (i = 0; i < q->nr; i++) {
4482 xpparam_t xpp;
4483 xdemitconf_t xecfg;
4484 mmfile_t mf1, mf2;
4485 struct diff_filepair *p = q->queue[i];
4486 int len1, len2;
4487
4488 memset(&xpp, 0, sizeof(xpp));
4489 memset(&xecfg, 0, sizeof(xecfg));
4490 if (p->status == 0)
4491 return error("internal diff status error");
4492 if (p->status == DIFF_STATUS_UNKNOWN)
4493 continue;
4494 if (diff_unmodified_pair(p))
4495 continue;
4496 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4497 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4498 continue;
4499 if (DIFF_PAIR_UNMERGED(p))
4500 continue;
4501
4502 diff_fill_sha1_info(p->one);
4503 diff_fill_sha1_info(p->two);
4504 if (fill_mmfile(&mf1, p->one) < 0 ||
4505 fill_mmfile(&mf2, p->two) < 0)
4506 return error("unable to read files to diff");
4507
4508 len1 = remove_space(p->one->path, strlen(p->one->path));
4509 len2 = remove_space(p->two->path, strlen(p->two->path));
4510 if (p->one->mode == 0)
4511 len1 = snprintf(buffer, sizeof(buffer),
4512 "diff--gita/%.*sb/%.*s"
4513 "newfilemode%06o"
4514 "---/dev/null"
4515 "+++b/%.*s",
4516 len1, p->one->path,
4517 len2, p->two->path,
4518 p->two->mode,
4519 len2, p->two->path);
4520 else if (p->two->mode == 0)
4521 len1 = snprintf(buffer, sizeof(buffer),
4522 "diff--gita/%.*sb/%.*s"
4523 "deletedfilemode%06o"
4524 "---a/%.*s"
4525 "+++/dev/null",
4526 len1, p->one->path,
4527 len2, p->two->path,
4528 p->one->mode,
4529 len1, p->one->path);
4530 else
4531 len1 = snprintf(buffer, sizeof(buffer),
4532 "diff--gita/%.*sb/%.*s"
4533 "---a/%.*s"
4534 "+++b/%.*s",
4535 len1, p->one->path,
4536 len2, p->two->path,
4537 len1, p->one->path,
4538 len2, p->two->path);
4539 git_SHA1_Update(&ctx, buffer, len1);
4540
4541 if (diff_filespec_is_binary(p->one) ||
4542 diff_filespec_is_binary(p->two)) {
4543 git_SHA1_Update(&ctx, sha1_to_hex(p->one->sha1), 40);
4544 git_SHA1_Update(&ctx, sha1_to_hex(p->two->sha1), 40);
4545 continue;
4546 }
4547
4548 xpp.flags = 0;
4549 xecfg.ctxlen = 3;
4550 xecfg.flags = 0;
4551 if (xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4552 &xpp, &xecfg))
4553 return error("unable to generate patch-id diff for %s",
4554 p->one->path);
4555 }
4556
4557 git_SHA1_Final(sha1, &ctx);
4558 return 0;
4559 }
4560
4561 int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
4562 {
4563 struct diff_queue_struct *q = &diff_queued_diff;
4564 int i;
4565 int result = diff_get_patch_id(options, sha1);
4566
4567 for (i = 0; i < q->nr; i++)
4568 diff_free_filepair(q->queue[i]);
4569
4570 free(q->queue);
4571 DIFF_QUEUE_CLEAR(q);
4572
4573 return result;
4574 }
4575
4576 static int is_summary_empty(const struct diff_queue_struct *q)
4577 {
4578 int i;
4579
4580 for (i = 0; i < q->nr; i++) {
4581 const struct diff_filepair *p = q->queue[i];
4582
4583 switch (p->status) {
4584 case DIFF_STATUS_DELETED:
4585 case DIFF_STATUS_ADDED:
4586 case DIFF_STATUS_COPIED:
4587 case DIFF_STATUS_RENAMED:
4588 return 0;
4589 default:
4590 if (p->score)
4591 return 0;
4592 if (p->one->mode && p->two->mode &&
4593 p->one->mode != p->two->mode)
4594 return 0;
4595 break;
4596 }
4597 }
4598 return 1;
4599 }
4600
4601 static const char rename_limit_warning[] =
4602 "inexact rename detection was skipped due to too many files.";
4603
4604 static const char degrade_cc_to_c_warning[] =
4605 "only found copies from modified paths due to too many files.";
4606
4607 static const char rename_limit_advice[] =
4608 "you may want to set your %s variable to at least "
4609 "%d and retry the command.";
4610
4611 void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4612 {
4613 if (degraded_cc)
4614 warning(degrade_cc_to_c_warning);
4615 else if (needed)
4616 warning(rename_limit_warning);
4617 else
4618 return;
4619 if (0 < needed && needed < 32767)
4620 warning(rename_limit_advice, varname, needed);
4621 }
4622
4623 void diff_flush(struct diff_options *options)
4624 {
4625 struct diff_queue_struct *q = &diff_queued_diff;
4626 int i, output_format = options->output_format;
4627 int separator = 0;
4628 int dirstat_by_line = 0;
4629
4630 /*
4631 * Order: raw, stat, summary, patch
4632 * or: name/name-status/checkdiff (other bits clear)
4633 */
4634 if (!q->nr)
4635 goto free_queue;
4636
4637 if (output_format & (DIFF_FORMAT_RAW |
4638 DIFF_FORMAT_NAME |
4639 DIFF_FORMAT_NAME_STATUS |
4640 DIFF_FORMAT_CHECKDIFF)) {
4641 for (i = 0; i < q->nr; i++) {
4642 struct diff_filepair *p = q->queue[i];
4643 if (check_pair_status(p))
4644 flush_one_pair(p, options);
4645 }
4646 separator++;
4647 }
4648
4649 if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4650 dirstat_by_line = 1;
4651
4652 if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4653 dirstat_by_line) {
4654 struct diffstat_t diffstat;
4655
4656 memset(&diffstat, 0, sizeof(struct diffstat_t));
4657 for (i = 0; i < q->nr; i++) {
4658 struct diff_filepair *p = q->queue[i];
4659 if (check_pair_status(p))
4660 diff_flush_stat(p, options, &diffstat);
4661 }
4662 if (output_format & DIFF_FORMAT_NUMSTAT)
4663 show_numstat(&diffstat, options);
4664 if (output_format & DIFF_FORMAT_DIFFSTAT)
4665 show_stats(&diffstat, options);
4666 if (output_format & DIFF_FORMAT_SHORTSTAT)
4667 show_shortstats(&diffstat, options);
4668 if (output_format & DIFF_FORMAT_DIRSTAT && dirstat_by_line)
4669 show_dirstat_by_line(&diffstat, options);
4670 free_diffstat_info(&diffstat);
4671 separator++;
4672 }
4673 if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4674 show_dirstat(options);
4675
4676 if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4677 for (i = 0; i < q->nr; i++) {
4678 diff_summary(options, q->queue[i]);
4679 }
4680 separator++;
4681 }
4682
4683 if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4684 DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4685 DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4686 /*
4687 * run diff_flush_patch for the exit status. setting
4688 * options->file to /dev/null should be safe, because we
4689 * aren't supposed to produce any output anyway.
4690 */
4691 if (options->close_file)
4692 fclose(options->file);
4693 options->file = fopen("/dev/null", "w");
4694 if (!options->file)
4695 die_errno("Could not open /dev/null");
4696 options->close_file = 1;
4697 for (i = 0; i < q->nr; i++) {
4698 struct diff_filepair *p = q->queue[i];
4699 if (check_pair_status(p))
4700 diff_flush_patch(p, options);
4701 if (options->found_changes)
4702 break;
4703 }
4704 }
4705
4706 if (output_format & DIFF_FORMAT_PATCH) {
4707 if (separator) {
4708 fprintf(options->file, "%s%c",
4709 diff_line_prefix(options),
4710 options->line_termination);
4711 if (options->stat_sep) {
4712 /* attach patch instead of inline */
4713 fputs(options->stat_sep, options->file);
4714 }
4715 }
4716
4717 for (i = 0; i < q->nr; i++) {
4718 struct diff_filepair *p = q->queue[i];
4719 if (check_pair_status(p))
4720 diff_flush_patch(p, options);
4721 }
4722 }
4723
4724 if (output_format & DIFF_FORMAT_CALLBACK)
4725 options->format_callback(q, options, options->format_callback_data);
4726
4727 for (i = 0; i < q->nr; i++)
4728 diff_free_filepair(q->queue[i]);
4729 free_queue:
4730 free(q->queue);
4731 DIFF_QUEUE_CLEAR(q);
4732 if (options->close_file)
4733 fclose(options->file);
4734
4735 /*
4736 * Report the content-level differences with HAS_CHANGES;
4737 * diff_addremove/diff_change does not set the bit when
4738 * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4739 */
4740 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4741 if (options->found_changes)
4742 DIFF_OPT_SET(options, HAS_CHANGES);
4743 else
4744 DIFF_OPT_CLR(options, HAS_CHANGES);
4745 }
4746 }
4747
4748 static int match_filter(const struct diff_options *options, const struct diff_filepair *p)
4749 {
4750 return (((p->status == DIFF_STATUS_MODIFIED) &&
4751 ((p->score &&
4752 filter_bit_tst(DIFF_STATUS_FILTER_BROKEN, options)) ||
4753 (!p->score &&
4754 filter_bit_tst(DIFF_STATUS_MODIFIED, options)))) ||
4755 ((p->status != DIFF_STATUS_MODIFIED) &&
4756 filter_bit_tst(p->status, options)));
4757 }
4758
4759 static void diffcore_apply_filter(struct diff_options *options)
4760 {
4761 int i;
4762 struct diff_queue_struct *q = &diff_queued_diff;
4763 struct diff_queue_struct outq;
4764
4765 DIFF_QUEUE_CLEAR(&outq);
4766
4767 if (!options->filter)
4768 return;
4769
4770 if (filter_bit_tst(DIFF_STATUS_FILTER_AON, options)) {
4771 int found;
4772 for (i = found = 0; !found && i < q->nr; i++) {
4773 if (match_filter(options, q->queue[i]))
4774 found++;
4775 }
4776 if (found)
4777 return;
4778
4779 /* otherwise we will clear the whole queue
4780 * by copying the empty outq at the end of this
4781 * function, but first clear the current entries
4782 * in the queue.
4783 */
4784 for (i = 0; i < q->nr; i++)
4785 diff_free_filepair(q->queue[i]);
4786 }
4787 else {
4788 /* Only the matching ones */
4789 for (i = 0; i < q->nr; i++) {
4790 struct diff_filepair *p = q->queue[i];
4791 if (match_filter(options, p))
4792 diff_q(&outq, p);
4793 else
4794 diff_free_filepair(p);
4795 }
4796 }
4797 free(q->queue);
4798 *q = outq;
4799 }
4800
4801 /* Check whether two filespecs with the same mode and size are identical */
4802 static int diff_filespec_is_identical(struct diff_filespec *one,
4803 struct diff_filespec *two)
4804 {
4805 if (S_ISGITLINK(one->mode))
4806 return 0;
4807 if (diff_populate_filespec(one, 0))
4808 return 0;
4809 if (diff_populate_filespec(two, 0))
4810 return 0;
4811 return !memcmp(one->data, two->data, one->size);
4812 }
4813
4814 static int diff_filespec_check_stat_unmatch(struct diff_filepair *p)
4815 {
4816 if (p->done_skip_stat_unmatch)
4817 return p->skip_stat_unmatch_result;
4818
4819 p->done_skip_stat_unmatch = 1;
4820 p->skip_stat_unmatch_result = 0;
4821 /*
4822 * 1. Entries that come from stat info dirtiness
4823 * always have both sides (iow, not create/delete),
4824 * one side of the object name is unknown, with
4825 * the same mode and size. Keep the ones that
4826 * do not match these criteria. They have real
4827 * differences.
4828 *
4829 * 2. At this point, the file is known to be modified,
4830 * with the same mode and size, and the object
4831 * name of one side is unknown. Need to inspect
4832 * the identical contents.
4833 */
4834 if (!DIFF_FILE_VALID(p->one) || /* (1) */
4835 !DIFF_FILE_VALID(p->two) ||
4836 (p->one->sha1_valid && p->two->sha1_valid) ||
4837 (p->one->mode != p->two->mode) ||
4838 diff_populate_filespec(p->one, CHECK_SIZE_ONLY) ||
4839 diff_populate_filespec(p->two, CHECK_SIZE_ONLY) ||
4840 (p->one->size != p->two->size) ||
4841 !diff_filespec_is_identical(p->one, p->two)) /* (2) */
4842 p->skip_stat_unmatch_result = 1;
4843 return p->skip_stat_unmatch_result;
4844 }
4845
4846 static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
4847 {
4848 int i;
4849 struct diff_queue_struct *q = &diff_queued_diff;
4850 struct diff_queue_struct outq;
4851 DIFF_QUEUE_CLEAR(&outq);
4852
4853 for (i = 0; i < q->nr; i++) {
4854 struct diff_filepair *p = q->queue[i];
4855
4856 if (diff_filespec_check_stat_unmatch(p))
4857 diff_q(&outq, p);
4858 else {
4859 /*
4860 * The caller can subtract 1 from skip_stat_unmatch
4861 * to determine how many paths were dirty only
4862 * due to stat info mismatch.
4863 */
4864 if (!DIFF_OPT_TST(diffopt, NO_INDEX))
4865 diffopt->skip_stat_unmatch++;
4866 diff_free_filepair(p);
4867 }
4868 }
4869 free(q->queue);
4870 *q = outq;
4871 }
4872
4873 static int diffnamecmp(const void *a_, const void *b_)
4874 {
4875 const struct diff_filepair *a = *((const struct diff_filepair **)a_);
4876 const struct diff_filepair *b = *((const struct diff_filepair **)b_);
4877 const char *name_a, *name_b;
4878
4879 name_a = a->one ? a->one->path : a->two->path;
4880 name_b = b->one ? b->one->path : b->two->path;
4881 return strcmp(name_a, name_b);
4882 }
4883
4884 void diffcore_fix_diff_index(struct diff_options *options)
4885 {
4886 struct diff_queue_struct *q = &diff_queued_diff;
4887 qsort(q->queue, q->nr, sizeof(q->queue[0]), diffnamecmp);
4888 }
4889
4890 void diffcore_std(struct diff_options *options)
4891 {
4892 /* NOTE please keep the following in sync with diff_tree_combined() */
4893 if (options->skip_stat_unmatch)
4894 diffcore_skip_stat_unmatch(options);
4895 if (!options->found_follow) {
4896 /* See try_to_follow_renames() in tree-diff.c */
4897 if (options->break_opt != -1)
4898 diffcore_break(options->break_opt);
4899 if (options->detect_rename)
4900 diffcore_rename(options);
4901 if (options->break_opt != -1)
4902 diffcore_merge_broken();
4903 }
4904 if (options->pickaxe)
4905 diffcore_pickaxe(options);
4906 if (options->orderfile)
4907 diffcore_order(options->orderfile);
4908 if (!options->found_follow)
4909 /* See try_to_follow_renames() in tree-diff.c */
4910 diff_resolve_rename_copy();
4911 diffcore_apply_filter(options);
4912
4913 if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4914 DIFF_OPT_SET(options, HAS_CHANGES);
4915 else
4916 DIFF_OPT_CLR(options, HAS_CHANGES);
4917
4918 options->found_follow = 0;
4919 }
4920
4921 int diff_result_code(struct diff_options *opt, int status)
4922 {
4923 int result = 0;
4924
4925 diff_warn_rename_limit("diff.renameLimit",
4926 opt->needed_rename_limit,
4927 opt->degraded_cc_to_c);
4928 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4929 !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
4930 return status;
4931 if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4932 DIFF_OPT_TST(opt, HAS_CHANGES))
4933 result |= 01;
4934 if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
4935 DIFF_OPT_TST(opt, CHECK_FAILED))
4936 result |= 02;
4937 return result;
4938 }
4939
4940 int diff_can_quit_early(struct diff_options *opt)
4941 {
4942 return (DIFF_OPT_TST(opt, QUICK) &&
4943 !opt->filter &&
4944 DIFF_OPT_TST(opt, HAS_CHANGES));
4945 }
4946
4947 /*
4948 * Shall changes to this submodule be ignored?
4949 *
4950 * Submodule changes can be configured to be ignored separately for each path,
4951 * but that configuration can be overridden from the command line.
4952 */
4953 static int is_submodule_ignored(const char *path, struct diff_options *options)
4954 {
4955 int ignored = 0;
4956 unsigned orig_flags = options->flags;
4957 if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
4958 set_diffopt_flags_from_submodule_config(options, path);
4959 if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
4960 ignored = 1;
4961 options->flags = orig_flags;
4962 return ignored;
4963 }
4964
4965 void diff_addremove(struct diff_options *options,
4966 int addremove, unsigned mode,
4967 const unsigned char *sha1,
4968 int sha1_valid,
4969 const char *concatpath, unsigned dirty_submodule)
4970 {
4971 struct diff_filespec *one, *two;
4972
4973 if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
4974 return;
4975
4976 /* This may look odd, but it is a preparation for
4977 * feeding "there are unchanged files which should
4978 * not produce diffs, but when you are doing copy
4979 * detection you would need them, so here they are"
4980 * entries to the diff-core. They will be prefixed
4981 * with something like '=' or '*' (I haven't decided
4982 * which but should not make any difference).
4983 * Feeding the same new and old to diff_change()
4984 * also has the same effect.
4985 * Before the final output happens, they are pruned after
4986 * merged into rename/copy pairs as appropriate.
4987 */
4988 if (DIFF_OPT_TST(options, REVERSE_DIFF))
4989 addremove = (addremove == '+' ? '-' :
4990 addremove == '-' ? '+' : addremove);
4991
4992 if (options->prefix &&
4993 strncmp(concatpath, options->prefix, options->prefix_length))
4994 return;
4995
4996 one = alloc_filespec(concatpath);
4997 two = alloc_filespec(concatpath);
4998
4999 if (addremove != '+')
5000 fill_filespec(one, sha1, sha1_valid, mode);
5001 if (addremove != '-') {
5002 fill_filespec(two, sha1, sha1_valid, mode);
5003 two->dirty_submodule = dirty_submodule;
5004 }
5005
5006 diff_queue(&diff_queued_diff, one, two);
5007 if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5008 DIFF_OPT_SET(options, HAS_CHANGES);
5009 }
5010
5011 void diff_change(struct diff_options *options,
5012 unsigned old_mode, unsigned new_mode,
5013 const unsigned char *old_sha1,
5014 const unsigned char *new_sha1,
5015 int old_sha1_valid, int new_sha1_valid,
5016 const char *concatpath,
5017 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
5018 {
5019 struct diff_filespec *one, *two;
5020 struct diff_filepair *p;
5021
5022 if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
5023 is_submodule_ignored(concatpath, options))
5024 return;
5025
5026 if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
5027 unsigned tmp;
5028 const unsigned char *tmp_c;
5029 tmp = old_mode; old_mode = new_mode; new_mode = tmp;
5030 tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
5031 tmp = old_sha1_valid; old_sha1_valid = new_sha1_valid;
5032 new_sha1_valid = tmp;
5033 tmp = old_dirty_submodule; old_dirty_submodule = new_dirty_submodule;
5034 new_dirty_submodule = tmp;
5035 }
5036
5037 if (options->prefix &&
5038 strncmp(concatpath, options->prefix, options->prefix_length))
5039 return;
5040
5041 one = alloc_filespec(concatpath);
5042 two = alloc_filespec(concatpath);
5043 fill_filespec(one, old_sha1, old_sha1_valid, old_mode);
5044 fill_filespec(two, new_sha1, new_sha1_valid, new_mode);
5045 one->dirty_submodule = old_dirty_submodule;
5046 two->dirty_submodule = new_dirty_submodule;
5047 p = diff_queue(&diff_queued_diff, one, two);
5048
5049 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5050 return;
5051
5052 if (DIFF_OPT_TST(options, QUICK) && options->skip_stat_unmatch &&
5053 !diff_filespec_check_stat_unmatch(p))
5054 return;
5055
5056 DIFF_OPT_SET(options, HAS_CHANGES);
5057 }
5058
5059 struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
5060 {
5061 struct diff_filepair *pair;
5062 struct diff_filespec *one, *two;
5063
5064 if (options->prefix &&
5065 strncmp(path, options->prefix, options->prefix_length))
5066 return NULL;
5067
5068 one = alloc_filespec(path);
5069 two = alloc_filespec(path);
5070 pair = diff_queue(&diff_queued_diff, one, two);
5071 pair->is_unmerged = 1;
5072 return pair;
5073 }
5074
5075 static char *run_textconv(const char *pgm, struct diff_filespec *spec,
5076 size_t *outsize)
5077 {
5078 struct diff_tempfile *temp;
5079 const char *argv[3];
5080 const char **arg = argv;
5081 struct child_process child = CHILD_PROCESS_INIT;
5082 struct strbuf buf = STRBUF_INIT;
5083 int err = 0;
5084
5085 temp = prepare_temp_file(spec->path, spec);
5086 *arg++ = pgm;
5087 *arg++ = temp->name;
5088 *arg = NULL;
5089
5090 child.use_shell = 1;
5091 child.argv = argv;
5092 child.out = -1;
5093 if (start_command(&child)) {
5094 remove_tempfile();
5095 return NULL;
5096 }
5097
5098 if (strbuf_read(&buf, child.out, 0) < 0)
5099 err = error("error reading from textconv command '%s'", pgm);
5100 close(child.out);
5101
5102 if (finish_command(&child) || err) {
5103 strbuf_release(&buf);
5104 remove_tempfile();
5105 return NULL;
5106 }
5107 remove_tempfile();
5108
5109 return strbuf_detach(&buf, outsize);
5110 }
5111
5112 size_t fill_textconv(struct userdiff_driver *driver,
5113 struct diff_filespec *df,
5114 char **outbuf)
5115 {
5116 size_t size;
5117
5118 if (!driver) {
5119 if (!DIFF_FILE_VALID(df)) {
5120 *outbuf = "";
5121 return 0;
5122 }
5123 if (diff_populate_filespec(df, 0))
5124 die("unable to read files to diff");
5125 *outbuf = df->data;
5126 return df->size;
5127 }
5128
5129 if (!driver->textconv)
5130 die("BUG: fill_textconv called with non-textconv driver");
5131
5132 if (driver->textconv_cache && df->sha1_valid) {
5133 *outbuf = notes_cache_get(driver->textconv_cache, df->sha1,
5134 &size);
5135 if (*outbuf)
5136 return size;
5137 }
5138
5139 *outbuf = run_textconv(driver->textconv, df, &size);
5140 if (!*outbuf)
5141 die("unable to read files to diff");
5142
5143 if (driver->textconv_cache && df->sha1_valid) {
5144 /* ignore errors, as we might be in a readonly repository */
5145 notes_cache_put(driver->textconv_cache, df->sha1, *outbuf,
5146 size);
5147 /*
5148 * we could save up changes and flush them all at the end,
5149 * but we would need an extra call after all diffing is done.
5150 * Since generating a cache entry is the slow path anyway,
5151 * this extra overhead probably isn't a big deal.
5152 */
5153 notes_cache_write(driver->textconv_cache);
5154 }
5155
5156 return size;
5157 }
5158
5159 void setup_diff_pager(struct diff_options *opt)
5160 {
5161 /*
5162 * If the user asked for our exit code, then either they want --quiet
5163 * or --exit-code. We should definitely not bother with a pager in the
5164 * former case, as we will generate no output. Since we still properly
5165 * report our exit code even when a pager is run, we _could_ run a
5166 * pager with --exit-code. But since we have not done so historically,
5167 * and because it is easy to find people oneline advising "git diff
5168 * --exit-code" in hooks and other scripts, we do not do so.
5169 */
5170 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5171 check_pager_config("diff") != 0)
5172 setup_pager();
5173 }