]> git.ipfire.org Git - thirdparty/git.git/blob - diff.c
Merge branch 'va/i18n-more'
[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 #include "graph.h"
22
23 #ifdef NO_FAST_WORKING_DIRECTORY
24 #define FAST_WORKING_DIRECTORY 0
25 #else
26 #define FAST_WORKING_DIRECTORY 1
27 #endif
28
29 static int diff_detect_rename_default;
30 static int diff_indent_heuristic; /* experimental */
31 static int diff_compaction_heuristic; /* experimental */
32 static int diff_rename_limit_default = 400;
33 static int diff_suppress_blank_empty;
34 static int diff_use_color_default = -1;
35 static int diff_context_default = 3;
36 static const char *diff_word_regex_cfg;
37 static const char *external_diff_cmd_cfg;
38 static const char *diff_order_file_cfg;
39 int diff_auto_refresh_index = 1;
40 static int diff_mnemonic_prefix;
41 static int diff_no_prefix;
42 static int diff_stat_graph_width;
43 static int diff_dirstat_permille_default = 30;
44 static struct diff_options default_diff_options;
45 static long diff_algorithm;
46
47 static char diff_colors[][COLOR_MAXLEN] = {
48 GIT_COLOR_RESET,
49 GIT_COLOR_NORMAL, /* CONTEXT */
50 GIT_COLOR_BOLD, /* METAINFO */
51 GIT_COLOR_CYAN, /* FRAGINFO */
52 GIT_COLOR_RED, /* OLD */
53 GIT_COLOR_GREEN, /* NEW */
54 GIT_COLOR_YELLOW, /* COMMIT */
55 GIT_COLOR_BG_RED, /* WHITESPACE */
56 GIT_COLOR_NORMAL, /* FUNCINFO */
57 };
58
59 static NORETURN void die_want_option(const char *option_name)
60 {
61 die(_("option '%s' requires a value"), option_name);
62 }
63
64 static int parse_diff_color_slot(const char *var)
65 {
66 if (!strcasecmp(var, "context") || !strcasecmp(var, "plain"))
67 return DIFF_CONTEXT;
68 if (!strcasecmp(var, "meta"))
69 return DIFF_METAINFO;
70 if (!strcasecmp(var, "frag"))
71 return DIFF_FRAGINFO;
72 if (!strcasecmp(var, "old"))
73 return DIFF_FILE_OLD;
74 if (!strcasecmp(var, "new"))
75 return DIFF_FILE_NEW;
76 if (!strcasecmp(var, "commit"))
77 return DIFF_COMMIT;
78 if (!strcasecmp(var, "whitespace"))
79 return DIFF_WHITESPACE;
80 if (!strcasecmp(var, "func"))
81 return DIFF_FUNCINFO;
82 return -1;
83 }
84
85 static int parse_dirstat_params(struct diff_options *options, const char *params_string,
86 struct strbuf *errmsg)
87 {
88 char *params_copy = xstrdup(params_string);
89 struct string_list params = STRING_LIST_INIT_NODUP;
90 int ret = 0;
91 int i;
92
93 if (*params_copy)
94 string_list_split_in_place(&params, params_copy, ',', -1);
95 for (i = 0; i < params.nr; i++) {
96 const char *p = params.items[i].string;
97 if (!strcmp(p, "changes")) {
98 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
99 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
100 } else if (!strcmp(p, "lines")) {
101 DIFF_OPT_SET(options, DIRSTAT_BY_LINE);
102 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
103 } else if (!strcmp(p, "files")) {
104 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
105 DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
106 } else if (!strcmp(p, "noncumulative")) {
107 DIFF_OPT_CLR(options, DIRSTAT_CUMULATIVE);
108 } else if (!strcmp(p, "cumulative")) {
109 DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
110 } else if (isdigit(*p)) {
111 char *end;
112 int permille = strtoul(p, &end, 10) * 10;
113 if (*end == '.' && isdigit(*++end)) {
114 /* only use first digit */
115 permille += *end - '0';
116 /* .. and ignore any further digits */
117 while (isdigit(*++end))
118 ; /* nothing */
119 }
120 if (!*end)
121 options->dirstat_permille = permille;
122 else {
123 strbuf_addf(errmsg, _(" Failed to parse dirstat cut-off percentage '%s'\n"),
124 p);
125 ret++;
126 }
127 } else {
128 strbuf_addf(errmsg, _(" Unknown dirstat parameter '%s'\n"), p);
129 ret++;
130 }
131
132 }
133 string_list_clear(&params, 0);
134 free(params_copy);
135 return ret;
136 }
137
138 static int parse_submodule_params(struct diff_options *options, const char *value)
139 {
140 if (!strcmp(value, "log"))
141 options->submodule_format = DIFF_SUBMODULE_LOG;
142 else if (!strcmp(value, "short"))
143 options->submodule_format = DIFF_SUBMODULE_SHORT;
144 else if (!strcmp(value, "diff"))
145 options->submodule_format = DIFF_SUBMODULE_INLINE_DIFF;
146 else
147 return -1;
148 return 0;
149 }
150
151 static int git_config_rename(const char *var, const char *value)
152 {
153 if (!value)
154 return DIFF_DETECT_RENAME;
155 if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
156 return DIFF_DETECT_COPY;
157 return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
158 }
159
160 long parse_algorithm_value(const char *value)
161 {
162 if (!value)
163 return -1;
164 else if (!strcasecmp(value, "myers") || !strcasecmp(value, "default"))
165 return 0;
166 else if (!strcasecmp(value, "minimal"))
167 return XDF_NEED_MINIMAL;
168 else if (!strcasecmp(value, "patience"))
169 return XDF_PATIENCE_DIFF;
170 else if (!strcasecmp(value, "histogram"))
171 return XDF_HISTOGRAM_DIFF;
172 return -1;
173 }
174
175 /*
176 * These are to give UI layer defaults.
177 * The core-level commands such as git-diff-files should
178 * never be affected by the setting of diff.renames
179 * the user happens to have in the configuration file.
180 */
181 void init_diff_ui_defaults(void)
182 {
183 diff_detect_rename_default = 1;
184 }
185
186 int git_diff_heuristic_config(const char *var, const char *value, void *cb)
187 {
188 if (!strcmp(var, "diff.indentheuristic")) {
189 diff_indent_heuristic = git_config_bool(var, value);
190 if (diff_indent_heuristic)
191 diff_compaction_heuristic = 0;
192 }
193 if (!strcmp(var, "diff.compactionheuristic")) {
194 diff_compaction_heuristic = git_config_bool(var, value);
195 if (diff_compaction_heuristic)
196 diff_indent_heuristic = 0;
197 }
198 return 0;
199 }
200
201 int git_diff_ui_config(const char *var, const char *value, void *cb)
202 {
203 if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
204 diff_use_color_default = git_config_colorbool(var, value);
205 return 0;
206 }
207 if (!strcmp(var, "diff.context")) {
208 diff_context_default = git_config_int(var, value);
209 if (diff_context_default < 0)
210 return -1;
211 return 0;
212 }
213 if (!strcmp(var, "diff.renames")) {
214 diff_detect_rename_default = git_config_rename(var, value);
215 return 0;
216 }
217 if (!strcmp(var, "diff.autorefreshindex")) {
218 diff_auto_refresh_index = git_config_bool(var, value);
219 return 0;
220 }
221 if (!strcmp(var, "diff.mnemonicprefix")) {
222 diff_mnemonic_prefix = git_config_bool(var, value);
223 return 0;
224 }
225 if (!strcmp(var, "diff.noprefix")) {
226 diff_no_prefix = git_config_bool(var, value);
227 return 0;
228 }
229 if (!strcmp(var, "diff.statgraphwidth")) {
230 diff_stat_graph_width = git_config_int(var, value);
231 return 0;
232 }
233 if (!strcmp(var, "diff.external"))
234 return git_config_string(&external_diff_cmd_cfg, var, value);
235 if (!strcmp(var, "diff.wordregex"))
236 return git_config_string(&diff_word_regex_cfg, var, value);
237 if (!strcmp(var, "diff.orderfile"))
238 return git_config_pathname(&diff_order_file_cfg, var, value);
239
240 if (!strcmp(var, "diff.ignoresubmodules"))
241 handle_ignore_submodules_arg(&default_diff_options, value);
242
243 if (!strcmp(var, "diff.submodule")) {
244 if (parse_submodule_params(&default_diff_options, value))
245 warning(_("Unknown value for 'diff.submodule' config variable: '%s'"),
246 value);
247 return 0;
248 }
249
250 if (!strcmp(var, "diff.algorithm")) {
251 diff_algorithm = parse_algorithm_value(value);
252 if (diff_algorithm < 0)
253 return -1;
254 return 0;
255 }
256
257 if (git_diff_heuristic_config(var, value, cb) < 0)
258 return -1;
259 if (git_color_config(var, value, cb) < 0)
260 return -1;
261
262 return git_diff_basic_config(var, value, cb);
263 }
264
265 int git_diff_basic_config(const char *var, const char *value, void *cb)
266 {
267 const char *name;
268
269 if (!strcmp(var, "diff.renamelimit")) {
270 diff_rename_limit_default = git_config_int(var, value);
271 return 0;
272 }
273
274 if (userdiff_config(var, value) < 0)
275 return -1;
276
277 if (skip_prefix(var, "diff.color.", &name) ||
278 skip_prefix(var, "color.diff.", &name)) {
279 int slot = parse_diff_color_slot(name);
280 if (slot < 0)
281 return 0;
282 if (!value)
283 return config_error_nonbool(var);
284 return color_parse(value, diff_colors[slot]);
285 }
286
287 /* like GNU diff's --suppress-blank-empty option */
288 if (!strcmp(var, "diff.suppressblankempty") ||
289 /* for backwards compatibility */
290 !strcmp(var, "diff.suppress-blank-empty")) {
291 diff_suppress_blank_empty = git_config_bool(var, value);
292 return 0;
293 }
294
295 if (!strcmp(var, "diff.dirstat")) {
296 struct strbuf errmsg = STRBUF_INIT;
297 default_diff_options.dirstat_permille = diff_dirstat_permille_default;
298 if (parse_dirstat_params(&default_diff_options, value, &errmsg))
299 warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
300 errmsg.buf);
301 strbuf_release(&errmsg);
302 diff_dirstat_permille_default = default_diff_options.dirstat_permille;
303 return 0;
304 }
305
306 if (starts_with(var, "submodule."))
307 return parse_submodule_config_option(var, value);
308
309 return git_default_config(var, value, cb);
310 }
311
312 static char *quote_two(const char *one, const char *two)
313 {
314 int need_one = quote_c_style(one, NULL, NULL, 1);
315 int need_two = quote_c_style(two, NULL, NULL, 1);
316 struct strbuf res = STRBUF_INIT;
317
318 if (need_one + need_two) {
319 strbuf_addch(&res, '"');
320 quote_c_style(one, &res, NULL, 1);
321 quote_c_style(two, &res, NULL, 1);
322 strbuf_addch(&res, '"');
323 } else {
324 strbuf_addstr(&res, one);
325 strbuf_addstr(&res, two);
326 }
327 return strbuf_detach(&res, NULL);
328 }
329
330 static const char *external_diff(void)
331 {
332 static const char *external_diff_cmd = NULL;
333 static int done_preparing = 0;
334
335 if (done_preparing)
336 return external_diff_cmd;
337 external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
338 if (!external_diff_cmd)
339 external_diff_cmd = external_diff_cmd_cfg;
340 done_preparing = 1;
341 return external_diff_cmd;
342 }
343
344 /*
345 * Keep track of files used for diffing. Sometimes such an entry
346 * refers to a temporary file, sometimes to an existing file, and
347 * sometimes to "/dev/null".
348 */
349 static struct diff_tempfile {
350 /*
351 * filename external diff should read from, or NULL if this
352 * entry is currently not in use:
353 */
354 const char *name;
355
356 char hex[GIT_SHA1_HEXSZ + 1];
357 char mode[10];
358
359 /*
360 * If this diff_tempfile instance refers to a temporary file,
361 * this tempfile object is used to manage its lifetime.
362 */
363 struct tempfile tempfile;
364 } diff_temp[2];
365
366 typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
367
368 struct emit_callback {
369 int color_diff;
370 unsigned ws_rule;
371 int blank_at_eof_in_preimage;
372 int blank_at_eof_in_postimage;
373 int lno_in_preimage;
374 int lno_in_postimage;
375 sane_truncate_fn truncate;
376 const char **label_path;
377 struct diff_words_data *diff_words;
378 struct diff_options *opt;
379 struct strbuf *header;
380 };
381
382 static int count_lines(const char *data, int size)
383 {
384 int count, ch, completely_empty = 1, nl_just_seen = 0;
385 count = 0;
386 while (0 < size--) {
387 ch = *data++;
388 if (ch == '\n') {
389 count++;
390 nl_just_seen = 1;
391 completely_empty = 0;
392 }
393 else {
394 nl_just_seen = 0;
395 completely_empty = 0;
396 }
397 }
398 if (completely_empty)
399 return 0;
400 if (!nl_just_seen)
401 count++; /* no trailing newline */
402 return count;
403 }
404
405 static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
406 {
407 if (!DIFF_FILE_VALID(one)) {
408 mf->ptr = (char *)""; /* does not matter */
409 mf->size = 0;
410 return 0;
411 }
412 else if (diff_populate_filespec(one, 0))
413 return -1;
414
415 mf->ptr = one->data;
416 mf->size = one->size;
417 return 0;
418 }
419
420 /* like fill_mmfile, but only for size, so we can avoid retrieving blob */
421 static unsigned long diff_filespec_size(struct diff_filespec *one)
422 {
423 if (!DIFF_FILE_VALID(one))
424 return 0;
425 diff_populate_filespec(one, CHECK_SIZE_ONLY);
426 return one->size;
427 }
428
429 static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
430 {
431 char *ptr = mf->ptr;
432 long size = mf->size;
433 int cnt = 0;
434
435 if (!size)
436 return cnt;
437 ptr += size - 1; /* pointing at the very end */
438 if (*ptr != '\n')
439 ; /* incomplete line */
440 else
441 ptr--; /* skip the last LF */
442 while (mf->ptr < ptr) {
443 char *prev_eol;
444 for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
445 if (*prev_eol == '\n')
446 break;
447 if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
448 break;
449 cnt++;
450 ptr = prev_eol - 1;
451 }
452 return cnt;
453 }
454
455 static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
456 struct emit_callback *ecbdata)
457 {
458 int l1, l2, at;
459 unsigned ws_rule = ecbdata->ws_rule;
460 l1 = count_trailing_blank(mf1, ws_rule);
461 l2 = count_trailing_blank(mf2, ws_rule);
462 if (l2 <= l1) {
463 ecbdata->blank_at_eof_in_preimage = 0;
464 ecbdata->blank_at_eof_in_postimage = 0;
465 return;
466 }
467 at = count_lines(mf1->ptr, mf1->size);
468 ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
469
470 at = count_lines(mf2->ptr, mf2->size);
471 ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
472 }
473
474 static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
475 int first, const char *line, int len)
476 {
477 int has_trailing_newline, has_trailing_carriage_return;
478 int nofirst;
479 FILE *file = o->file;
480
481 fputs(diff_line_prefix(o), file);
482
483 if (len == 0) {
484 has_trailing_newline = (first == '\n');
485 has_trailing_carriage_return = (!has_trailing_newline &&
486 (first == '\r'));
487 nofirst = has_trailing_newline || has_trailing_carriage_return;
488 } else {
489 has_trailing_newline = (len > 0 && line[len-1] == '\n');
490 if (has_trailing_newline)
491 len--;
492 has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
493 if (has_trailing_carriage_return)
494 len--;
495 nofirst = 0;
496 }
497
498 if (len || !nofirst) {
499 fputs(set, file);
500 if (!nofirst)
501 fputc(first, file);
502 fwrite(line, len, 1, file);
503 fputs(reset, file);
504 }
505 if (has_trailing_carriage_return)
506 fputc('\r', file);
507 if (has_trailing_newline)
508 fputc('\n', file);
509 }
510
511 static void emit_line(struct diff_options *o, const char *set, const char *reset,
512 const char *line, int len)
513 {
514 emit_line_0(o, set, reset, line[0], line+1, len-1);
515 }
516
517 static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
518 {
519 if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
520 ecbdata->blank_at_eof_in_preimage &&
521 ecbdata->blank_at_eof_in_postimage &&
522 ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
523 ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
524 return 0;
525 return ws_blank_line(line, len, ecbdata->ws_rule);
526 }
527
528 static void emit_line_checked(const char *reset,
529 struct emit_callback *ecbdata,
530 const char *line, int len,
531 enum color_diff color,
532 unsigned ws_error_highlight,
533 char sign)
534 {
535 const char *set = diff_get_color(ecbdata->color_diff, color);
536 const char *ws = NULL;
537
538 if (ecbdata->opt->ws_error_highlight & ws_error_highlight) {
539 ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
540 if (!*ws)
541 ws = NULL;
542 }
543
544 if (!ws)
545 emit_line_0(ecbdata->opt, set, reset, sign, line, len);
546 else if (sign == '+' && new_blank_line_at_eof(ecbdata, line, len))
547 /* Blank line at EOF - paint '+' as well */
548 emit_line_0(ecbdata->opt, ws, reset, sign, line, len);
549 else {
550 /* Emit just the prefix, then the rest. */
551 emit_line_0(ecbdata->opt, set, reset, sign, "", 0);
552 ws_check_emit(line, len, ecbdata->ws_rule,
553 ecbdata->opt->file, set, reset, ws);
554 }
555 }
556
557 static void emit_add_line(const char *reset,
558 struct emit_callback *ecbdata,
559 const char *line, int len)
560 {
561 emit_line_checked(reset, ecbdata, line, len,
562 DIFF_FILE_NEW, WSEH_NEW, '+');
563 }
564
565 static void emit_del_line(const char *reset,
566 struct emit_callback *ecbdata,
567 const char *line, int len)
568 {
569 emit_line_checked(reset, ecbdata, line, len,
570 DIFF_FILE_OLD, WSEH_OLD, '-');
571 }
572
573 static void emit_context_line(const char *reset,
574 struct emit_callback *ecbdata,
575 const char *line, int len)
576 {
577 emit_line_checked(reset, ecbdata, line, len,
578 DIFF_CONTEXT, WSEH_CONTEXT, ' ');
579 }
580
581 static void emit_hunk_header(struct emit_callback *ecbdata,
582 const char *line, int len)
583 {
584 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
585 const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
586 const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
587 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
588 static const char atat[2] = { '@', '@' };
589 const char *cp, *ep;
590 struct strbuf msgbuf = STRBUF_INIT;
591 int org_len = len;
592 int i = 1;
593
594 /*
595 * As a hunk header must begin with "@@ -<old>, +<new> @@",
596 * it always is at least 10 bytes long.
597 */
598 if (len < 10 ||
599 memcmp(line, atat, 2) ||
600 !(ep = memmem(line + 2, len - 2, atat, 2))) {
601 emit_line(ecbdata->opt, context, reset, line, len);
602 return;
603 }
604 ep += 2; /* skip over @@ */
605
606 /* The hunk header in fraginfo color */
607 strbuf_addstr(&msgbuf, frag);
608 strbuf_add(&msgbuf, line, ep - line);
609 strbuf_addstr(&msgbuf, reset);
610
611 /*
612 * trailing "\r\n"
613 */
614 for ( ; i < 3; i++)
615 if (line[len - i] == '\r' || line[len - i] == '\n')
616 len--;
617
618 /* blank before the func header */
619 for (cp = ep; ep - line < len; ep++)
620 if (*ep != ' ' && *ep != '\t')
621 break;
622 if (ep != cp) {
623 strbuf_addstr(&msgbuf, context);
624 strbuf_add(&msgbuf, cp, ep - cp);
625 strbuf_addstr(&msgbuf, reset);
626 }
627
628 if (ep < line + len) {
629 strbuf_addstr(&msgbuf, func);
630 strbuf_add(&msgbuf, ep, line + len - ep);
631 strbuf_addstr(&msgbuf, reset);
632 }
633
634 strbuf_add(&msgbuf, line + len, org_len - len);
635 emit_line(ecbdata->opt, "", "", msgbuf.buf, msgbuf.len);
636 strbuf_release(&msgbuf);
637 }
638
639 static struct diff_tempfile *claim_diff_tempfile(void) {
640 int i;
641 for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
642 if (!diff_temp[i].name)
643 return diff_temp + i;
644 die("BUG: diff is failing to clean up its tempfiles");
645 }
646
647 static void remove_tempfile(void)
648 {
649 int i;
650 for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
651 if (is_tempfile_active(&diff_temp[i].tempfile))
652 delete_tempfile(&diff_temp[i].tempfile);
653 diff_temp[i].name = NULL;
654 }
655 }
656
657 static void print_line_count(FILE *file, int count)
658 {
659 switch (count) {
660 case 0:
661 fprintf(file, "0,0");
662 break;
663 case 1:
664 fprintf(file, "1");
665 break;
666 default:
667 fprintf(file, "1,%d", count);
668 break;
669 }
670 }
671
672 static void emit_rewrite_lines(struct emit_callback *ecb,
673 int prefix, const char *data, int size)
674 {
675 const char *endp = NULL;
676 static const char *nneof = " No newline at end of file\n";
677 const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
678
679 while (0 < size) {
680 int len;
681
682 endp = memchr(data, '\n', size);
683 len = endp ? (endp - data + 1) : size;
684 if (prefix != '+') {
685 ecb->lno_in_preimage++;
686 emit_del_line(reset, ecb, data, len);
687 } else {
688 ecb->lno_in_postimage++;
689 emit_add_line(reset, ecb, data, len);
690 }
691 size -= len;
692 data += len;
693 }
694 if (!endp) {
695 const char *context = diff_get_color(ecb->color_diff,
696 DIFF_CONTEXT);
697 putc('\n', ecb->opt->file);
698 emit_line_0(ecb->opt, context, reset, '\\',
699 nneof, strlen(nneof));
700 }
701 }
702
703 static void emit_rewrite_diff(const char *name_a,
704 const char *name_b,
705 struct diff_filespec *one,
706 struct diff_filespec *two,
707 struct userdiff_driver *textconv_one,
708 struct userdiff_driver *textconv_two,
709 struct diff_options *o)
710 {
711 int lc_a, lc_b;
712 const char *name_a_tab, *name_b_tab;
713 const char *metainfo = diff_get_color(o->use_color, DIFF_METAINFO);
714 const char *fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
715 const char *reset = diff_get_color(o->use_color, DIFF_RESET);
716 static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
717 const char *a_prefix, *b_prefix;
718 char *data_one, *data_two;
719 size_t size_one, size_two;
720 struct emit_callback ecbdata;
721 const char *line_prefix = diff_line_prefix(o);
722
723 if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
724 a_prefix = o->b_prefix;
725 b_prefix = o->a_prefix;
726 } else {
727 a_prefix = o->a_prefix;
728 b_prefix = o->b_prefix;
729 }
730
731 name_a += (*name_a == '/');
732 name_b += (*name_b == '/');
733 name_a_tab = strchr(name_a, ' ') ? "\t" : "";
734 name_b_tab = strchr(name_b, ' ') ? "\t" : "";
735
736 strbuf_reset(&a_name);
737 strbuf_reset(&b_name);
738 quote_two_c_style(&a_name, a_prefix, name_a, 0);
739 quote_two_c_style(&b_name, b_prefix, name_b, 0);
740
741 size_one = fill_textconv(textconv_one, one, &data_one);
742 size_two = fill_textconv(textconv_two, two, &data_two);
743
744 memset(&ecbdata, 0, sizeof(ecbdata));
745 ecbdata.color_diff = want_color(o->use_color);
746 ecbdata.ws_rule = whitespace_rule(name_b);
747 ecbdata.opt = o;
748 if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
749 mmfile_t mf1, mf2;
750 mf1.ptr = (char *)data_one;
751 mf2.ptr = (char *)data_two;
752 mf1.size = size_one;
753 mf2.size = size_two;
754 check_blank_at_eof(&mf1, &mf2, &ecbdata);
755 }
756 ecbdata.lno_in_preimage = 1;
757 ecbdata.lno_in_postimage = 1;
758
759 lc_a = count_lines(data_one, size_one);
760 lc_b = count_lines(data_two, size_two);
761 fprintf(o->file,
762 "%s%s--- %s%s%s\n%s%s+++ %s%s%s\n%s%s@@ -",
763 line_prefix, metainfo, a_name.buf, name_a_tab, reset,
764 line_prefix, metainfo, b_name.buf, name_b_tab, reset,
765 line_prefix, fraginfo);
766 if (!o->irreversible_delete)
767 print_line_count(o->file, lc_a);
768 else
769 fprintf(o->file, "?,?");
770 fprintf(o->file, " +");
771 print_line_count(o->file, lc_b);
772 fprintf(o->file, " @@%s\n", reset);
773 if (lc_a && !o->irreversible_delete)
774 emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
775 if (lc_b)
776 emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
777 if (textconv_one)
778 free((char *)data_one);
779 if (textconv_two)
780 free((char *)data_two);
781 }
782
783 struct diff_words_buffer {
784 mmfile_t text;
785 long alloc;
786 struct diff_words_orig {
787 const char *begin, *end;
788 } *orig;
789 int orig_nr, orig_alloc;
790 };
791
792 static void diff_words_append(char *line, unsigned long len,
793 struct diff_words_buffer *buffer)
794 {
795 ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
796 line++;
797 len--;
798 memcpy(buffer->text.ptr + buffer->text.size, line, len);
799 buffer->text.size += len;
800 buffer->text.ptr[buffer->text.size] = '\0';
801 }
802
803 struct diff_words_style_elem {
804 const char *prefix;
805 const char *suffix;
806 const char *color; /* NULL; filled in by the setup code if
807 * color is enabled */
808 };
809
810 struct diff_words_style {
811 enum diff_words_type type;
812 struct diff_words_style_elem new, old, ctx;
813 const char *newline;
814 };
815
816 static struct diff_words_style diff_words_styles[] = {
817 { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
818 { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
819 { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
820 };
821
822 struct diff_words_data {
823 struct diff_words_buffer minus, plus;
824 const char *current_plus;
825 int last_minus;
826 struct diff_options *opt;
827 regex_t *word_regex;
828 enum diff_words_type type;
829 struct diff_words_style *style;
830 };
831
832 static int fn_out_diff_words_write_helper(FILE *fp,
833 struct diff_words_style_elem *st_el,
834 const char *newline,
835 size_t count, const char *buf,
836 const char *line_prefix)
837 {
838 int print = 0;
839
840 while (count) {
841 char *p = memchr(buf, '\n', count);
842 if (print)
843 fputs(line_prefix, fp);
844 if (p != buf) {
845 if (st_el->color && fputs(st_el->color, fp) < 0)
846 return -1;
847 if (fputs(st_el->prefix, fp) < 0 ||
848 fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
849 fputs(st_el->suffix, fp) < 0)
850 return -1;
851 if (st_el->color && *st_el->color
852 && fputs(GIT_COLOR_RESET, fp) < 0)
853 return -1;
854 }
855 if (!p)
856 return 0;
857 if (fputs(newline, fp) < 0)
858 return -1;
859 count -= p + 1 - buf;
860 buf = p + 1;
861 print = 1;
862 }
863 return 0;
864 }
865
866 /*
867 * '--color-words' algorithm can be described as:
868 *
869 * 1. collect a the minus/plus lines of a diff hunk, divided into
870 * minus-lines and plus-lines;
871 *
872 * 2. break both minus-lines and plus-lines into words and
873 * place them into two mmfile_t with one word for each line;
874 *
875 * 3. use xdiff to run diff on the two mmfile_t to get the words level diff;
876 *
877 * And for the common parts of the both file, we output the plus side text.
878 * diff_words->current_plus is used to trace the current position of the plus file
879 * which printed. diff_words->last_minus is used to trace the last minus word
880 * printed.
881 *
882 * For '--graph' to work with '--color-words', we need to output the graph prefix
883 * on each line of color words output. Generally, there are two conditions on
884 * which we should output the prefix.
885 *
886 * 1. diff_words->last_minus == 0 &&
887 * diff_words->current_plus == diff_words->plus.text.ptr
888 *
889 * that is: the plus text must start as a new line, and if there is no minus
890 * word printed, a graph prefix must be printed.
891 *
892 * 2. diff_words->current_plus > diff_words->plus.text.ptr &&
893 * *(diff_words->current_plus - 1) == '\n'
894 *
895 * that is: a graph prefix must be printed following a '\n'
896 */
897 static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
898 {
899 if ((diff_words->last_minus == 0 &&
900 diff_words->current_plus == diff_words->plus.text.ptr) ||
901 (diff_words->current_plus > diff_words->plus.text.ptr &&
902 *(diff_words->current_plus - 1) == '\n')) {
903 return 1;
904 } else {
905 return 0;
906 }
907 }
908
909 static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
910 {
911 struct diff_words_data *diff_words = priv;
912 struct diff_words_style *style = diff_words->style;
913 int minus_first, minus_len, plus_first, plus_len;
914 const char *minus_begin, *minus_end, *plus_begin, *plus_end;
915 struct diff_options *opt = diff_words->opt;
916 const char *line_prefix;
917
918 if (line[0] != '@' || parse_hunk_header(line, len,
919 &minus_first, &minus_len, &plus_first, &plus_len))
920 return;
921
922 assert(opt);
923 line_prefix = diff_line_prefix(opt);
924
925 /* POSIX requires that first be decremented by one if len == 0... */
926 if (minus_len) {
927 minus_begin = diff_words->minus.orig[minus_first].begin;
928 minus_end =
929 diff_words->minus.orig[minus_first + minus_len - 1].end;
930 } else
931 minus_begin = minus_end =
932 diff_words->minus.orig[minus_first].end;
933
934 if (plus_len) {
935 plus_begin = diff_words->plus.orig[plus_first].begin;
936 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
937 } else
938 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
939
940 if (color_words_output_graph_prefix(diff_words)) {
941 fputs(line_prefix, diff_words->opt->file);
942 }
943 if (diff_words->current_plus != plus_begin) {
944 fn_out_diff_words_write_helper(diff_words->opt->file,
945 &style->ctx, style->newline,
946 plus_begin - diff_words->current_plus,
947 diff_words->current_plus, line_prefix);
948 if (*(plus_begin - 1) == '\n')
949 fputs(line_prefix, diff_words->opt->file);
950 }
951 if (minus_begin != minus_end) {
952 fn_out_diff_words_write_helper(diff_words->opt->file,
953 &style->old, style->newline,
954 minus_end - minus_begin, minus_begin,
955 line_prefix);
956 }
957 if (plus_begin != plus_end) {
958 fn_out_diff_words_write_helper(diff_words->opt->file,
959 &style->new, style->newline,
960 plus_end - plus_begin, plus_begin,
961 line_prefix);
962 }
963
964 diff_words->current_plus = plus_end;
965 diff_words->last_minus = minus_first;
966 }
967
968 /* This function starts looking at *begin, and returns 0 iff a word was found. */
969 static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
970 int *begin, int *end)
971 {
972 if (word_regex && *begin < buffer->size) {
973 regmatch_t match[1];
974 if (!regexec(word_regex, buffer->ptr + *begin, 1, match, 0)) {
975 char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
976 '\n', match[0].rm_eo - match[0].rm_so);
977 *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
978 *begin += match[0].rm_so;
979 return *begin >= *end;
980 }
981 return -1;
982 }
983
984 /* find the next word */
985 while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
986 (*begin)++;
987 if (*begin >= buffer->size)
988 return -1;
989
990 /* find the end of the word */
991 *end = *begin + 1;
992 while (*end < buffer->size && !isspace(buffer->ptr[*end]))
993 (*end)++;
994
995 return 0;
996 }
997
998 /*
999 * This function splits the words in buffer->text, stores the list with
1000 * newline separator into out, and saves the offsets of the original words
1001 * in buffer->orig.
1002 */
1003 static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
1004 regex_t *word_regex)
1005 {
1006 int i, j;
1007 long alloc = 0;
1008
1009 out->size = 0;
1010 out->ptr = NULL;
1011
1012 /* fake an empty "0th" word */
1013 ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
1014 buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
1015 buffer->orig_nr = 1;
1016
1017 for (i = 0; i < buffer->text.size; i++) {
1018 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
1019 return;
1020
1021 /* store original boundaries */
1022 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
1023 buffer->orig_alloc);
1024 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
1025 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
1026 buffer->orig_nr++;
1027
1028 /* store one word */
1029 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
1030 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
1031 out->ptr[out->size + j - i] = '\n';
1032 out->size += j - i + 1;
1033
1034 i = j - 1;
1035 }
1036 }
1037
1038 /* this executes the word diff on the accumulated buffers */
1039 static void diff_words_show(struct diff_words_data *diff_words)
1040 {
1041 xpparam_t xpp;
1042 xdemitconf_t xecfg;
1043 mmfile_t minus, plus;
1044 struct diff_words_style *style = diff_words->style;
1045
1046 struct diff_options *opt = diff_words->opt;
1047 const char *line_prefix;
1048
1049 assert(opt);
1050 line_prefix = diff_line_prefix(opt);
1051
1052 /* special case: only removal */
1053 if (!diff_words->plus.text.size) {
1054 fputs(line_prefix, diff_words->opt->file);
1055 fn_out_diff_words_write_helper(diff_words->opt->file,
1056 &style->old, style->newline,
1057 diff_words->minus.text.size,
1058 diff_words->minus.text.ptr, line_prefix);
1059 diff_words->minus.text.size = 0;
1060 return;
1061 }
1062
1063 diff_words->current_plus = diff_words->plus.text.ptr;
1064 diff_words->last_minus = 0;
1065
1066 memset(&xpp, 0, sizeof(xpp));
1067 memset(&xecfg, 0, sizeof(xecfg));
1068 diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
1069 diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
1070 xpp.flags = 0;
1071 /* as only the hunk header will be parsed, we need a 0-context */
1072 xecfg.ctxlen = 0;
1073 if (xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
1074 &xpp, &xecfg))
1075 die("unable to generate word diff");
1076 free(minus.ptr);
1077 free(plus.ptr);
1078 if (diff_words->current_plus != diff_words->plus.text.ptr +
1079 diff_words->plus.text.size) {
1080 if (color_words_output_graph_prefix(diff_words))
1081 fputs(line_prefix, diff_words->opt->file);
1082 fn_out_diff_words_write_helper(diff_words->opt->file,
1083 &style->ctx, style->newline,
1084 diff_words->plus.text.ptr + diff_words->plus.text.size
1085 - diff_words->current_plus, diff_words->current_plus,
1086 line_prefix);
1087 }
1088 diff_words->minus.text.size = diff_words->plus.text.size = 0;
1089 }
1090
1091 /* In "color-words" mode, show word-diff of words accumulated in the buffer */
1092 static void diff_words_flush(struct emit_callback *ecbdata)
1093 {
1094 if (ecbdata->diff_words->minus.text.size ||
1095 ecbdata->diff_words->plus.text.size)
1096 diff_words_show(ecbdata->diff_words);
1097 }
1098
1099 static void diff_filespec_load_driver(struct diff_filespec *one)
1100 {
1101 /* Use already-loaded driver */
1102 if (one->driver)
1103 return;
1104
1105 if (S_ISREG(one->mode))
1106 one->driver = userdiff_find_by_path(one->path);
1107
1108 /* Fallback to default settings */
1109 if (!one->driver)
1110 one->driver = userdiff_find_by_name("default");
1111 }
1112
1113 static const char *userdiff_word_regex(struct diff_filespec *one)
1114 {
1115 diff_filespec_load_driver(one);
1116 return one->driver->word_regex;
1117 }
1118
1119 static void init_diff_words_data(struct emit_callback *ecbdata,
1120 struct diff_options *orig_opts,
1121 struct diff_filespec *one,
1122 struct diff_filespec *two)
1123 {
1124 int i;
1125 struct diff_options *o = xmalloc(sizeof(struct diff_options));
1126 memcpy(o, orig_opts, sizeof(struct diff_options));
1127
1128 ecbdata->diff_words =
1129 xcalloc(1, sizeof(struct diff_words_data));
1130 ecbdata->diff_words->type = o->word_diff;
1131 ecbdata->diff_words->opt = o;
1132 if (!o->word_regex)
1133 o->word_regex = userdiff_word_regex(one);
1134 if (!o->word_regex)
1135 o->word_regex = userdiff_word_regex(two);
1136 if (!o->word_regex)
1137 o->word_regex = diff_word_regex_cfg;
1138 if (o->word_regex) {
1139 ecbdata->diff_words->word_regex = (regex_t *)
1140 xmalloc(sizeof(regex_t));
1141 if (regcomp(ecbdata->diff_words->word_regex,
1142 o->word_regex,
1143 REG_EXTENDED | REG_NEWLINE))
1144 die ("Invalid regular expression: %s",
1145 o->word_regex);
1146 }
1147 for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1148 if (o->word_diff == diff_words_styles[i].type) {
1149 ecbdata->diff_words->style =
1150 &diff_words_styles[i];
1151 break;
1152 }
1153 }
1154 if (want_color(o->use_color)) {
1155 struct diff_words_style *st = ecbdata->diff_words->style;
1156 st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1157 st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1158 st->ctx.color = diff_get_color_opt(o, DIFF_CONTEXT);
1159 }
1160 }
1161
1162 static void free_diff_words_data(struct emit_callback *ecbdata)
1163 {
1164 if (ecbdata->diff_words) {
1165 diff_words_flush(ecbdata);
1166 free (ecbdata->diff_words->opt);
1167 free (ecbdata->diff_words->minus.text.ptr);
1168 free (ecbdata->diff_words->minus.orig);
1169 free (ecbdata->diff_words->plus.text.ptr);
1170 free (ecbdata->diff_words->plus.orig);
1171 if (ecbdata->diff_words->word_regex) {
1172 regfree(ecbdata->diff_words->word_regex);
1173 free(ecbdata->diff_words->word_regex);
1174 }
1175 free(ecbdata->diff_words);
1176 ecbdata->diff_words = NULL;
1177 }
1178 }
1179
1180 const char *diff_get_color(int diff_use_color, enum color_diff ix)
1181 {
1182 if (want_color(diff_use_color))
1183 return diff_colors[ix];
1184 return "";
1185 }
1186
1187 const char *diff_line_prefix(struct diff_options *opt)
1188 {
1189 struct strbuf *msgbuf;
1190 if (!opt->output_prefix)
1191 return "";
1192
1193 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
1194 return msgbuf->buf;
1195 }
1196
1197 static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1198 {
1199 const char *cp;
1200 unsigned long allot;
1201 size_t l = len;
1202
1203 if (ecb->truncate)
1204 return ecb->truncate(line, len);
1205 cp = line;
1206 allot = l;
1207 while (0 < l) {
1208 (void) utf8_width(&cp, &l);
1209 if (!cp)
1210 break; /* truncated in the middle? */
1211 }
1212 return allot - l;
1213 }
1214
1215 static void find_lno(const char *line, struct emit_callback *ecbdata)
1216 {
1217 const char *p;
1218 ecbdata->lno_in_preimage = 0;
1219 ecbdata->lno_in_postimage = 0;
1220 p = strchr(line, '-');
1221 if (!p)
1222 return; /* cannot happen */
1223 ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1224 p = strchr(p, '+');
1225 if (!p)
1226 return; /* cannot happen */
1227 ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1228 }
1229
1230 static void fn_out_consume(void *priv, char *line, unsigned long len)
1231 {
1232 struct emit_callback *ecbdata = priv;
1233 const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
1234 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
1235 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1236 struct diff_options *o = ecbdata->opt;
1237 const char *line_prefix = diff_line_prefix(o);
1238
1239 o->found_changes = 1;
1240
1241 if (ecbdata->header) {
1242 fprintf(o->file, "%s", ecbdata->header->buf);
1243 strbuf_reset(ecbdata->header);
1244 ecbdata->header = NULL;
1245 }
1246
1247 if (ecbdata->label_path[0]) {
1248 const char *name_a_tab, *name_b_tab;
1249
1250 name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
1251 name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
1252
1253 fprintf(o->file, "%s%s--- %s%s%s\n",
1254 line_prefix, meta, ecbdata->label_path[0], reset, name_a_tab);
1255 fprintf(o->file, "%s%s+++ %s%s%s\n",
1256 line_prefix, meta, ecbdata->label_path[1], reset, name_b_tab);
1257 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1258 }
1259
1260 if (diff_suppress_blank_empty
1261 && len == 2 && line[0] == ' ' && line[1] == '\n') {
1262 line[0] = '\n';
1263 len = 1;
1264 }
1265
1266 if (line[0] == '@') {
1267 if (ecbdata->diff_words)
1268 diff_words_flush(ecbdata);
1269 len = sane_truncate_line(ecbdata, line, len);
1270 find_lno(line, ecbdata);
1271 emit_hunk_header(ecbdata, line, len);
1272 if (line[len-1] != '\n')
1273 putc('\n', o->file);
1274 return;
1275 }
1276
1277 if (ecbdata->diff_words) {
1278 if (line[0] == '-') {
1279 diff_words_append(line, len,
1280 &ecbdata->diff_words->minus);
1281 return;
1282 } else if (line[0] == '+') {
1283 diff_words_append(line, len,
1284 &ecbdata->diff_words->plus);
1285 return;
1286 } else if (starts_with(line, "\\ ")) {
1287 /*
1288 * Eat the "no newline at eof" marker as if we
1289 * saw a "+" or "-" line with nothing on it,
1290 * and return without diff_words_flush() to
1291 * defer processing. If this is the end of
1292 * preimage, more "+" lines may come after it.
1293 */
1294 return;
1295 }
1296 diff_words_flush(ecbdata);
1297 if (ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN) {
1298 emit_line(o, context, reset, line, len);
1299 fputs("~\n", o->file);
1300 } else {
1301 /*
1302 * Skip the prefix character, if any. With
1303 * diff_suppress_blank_empty, there may be
1304 * none.
1305 */
1306 if (line[0] != '\n') {
1307 line++;
1308 len--;
1309 }
1310 emit_line(o, context, reset, line, len);
1311 }
1312 return;
1313 }
1314
1315 switch (line[0]) {
1316 case '+':
1317 ecbdata->lno_in_postimage++;
1318 emit_add_line(reset, ecbdata, line + 1, len - 1);
1319 break;
1320 case '-':
1321 ecbdata->lno_in_preimage++;
1322 emit_del_line(reset, ecbdata, line + 1, len - 1);
1323 break;
1324 case ' ':
1325 ecbdata->lno_in_postimage++;
1326 ecbdata->lno_in_preimage++;
1327 emit_context_line(reset, ecbdata, line + 1, len - 1);
1328 break;
1329 default:
1330 /* incomplete line at the end */
1331 ecbdata->lno_in_preimage++;
1332 emit_line(o, diff_get_color(ecbdata->color_diff, DIFF_CONTEXT),
1333 reset, line, len);
1334 break;
1335 }
1336 }
1337
1338 static char *pprint_rename(const char *a, const char *b)
1339 {
1340 const char *old = a;
1341 const char *new = b;
1342 struct strbuf name = STRBUF_INIT;
1343 int pfx_length, sfx_length;
1344 int pfx_adjust_for_slash;
1345 int len_a = strlen(a);
1346 int len_b = strlen(b);
1347 int a_midlen, b_midlen;
1348 int qlen_a = quote_c_style(a, NULL, NULL, 0);
1349 int qlen_b = quote_c_style(b, NULL, NULL, 0);
1350
1351 if (qlen_a || qlen_b) {
1352 quote_c_style(a, &name, NULL, 0);
1353 strbuf_addstr(&name, " => ");
1354 quote_c_style(b, &name, NULL, 0);
1355 return strbuf_detach(&name, NULL);
1356 }
1357
1358 /* Find common prefix */
1359 pfx_length = 0;
1360 while (*old && *new && *old == *new) {
1361 if (*old == '/')
1362 pfx_length = old - a + 1;
1363 old++;
1364 new++;
1365 }
1366
1367 /* Find common suffix */
1368 old = a + len_a;
1369 new = b + len_b;
1370 sfx_length = 0;
1371 /*
1372 * If there is a common prefix, it must end in a slash. In
1373 * that case we let this loop run 1 into the prefix to see the
1374 * same slash.
1375 *
1376 * If there is no common prefix, we cannot do this as it would
1377 * underrun the input strings.
1378 */
1379 pfx_adjust_for_slash = (pfx_length ? 1 : 0);
1380 while (a + pfx_length - pfx_adjust_for_slash <= old &&
1381 b + pfx_length - pfx_adjust_for_slash <= new &&
1382 *old == *new) {
1383 if (*old == '/')
1384 sfx_length = len_a - (old - a);
1385 old--;
1386 new--;
1387 }
1388
1389 /*
1390 * pfx{mid-a => mid-b}sfx
1391 * {pfx-a => pfx-b}sfx
1392 * pfx{sfx-a => sfx-b}
1393 * name-a => name-b
1394 */
1395 a_midlen = len_a - pfx_length - sfx_length;
1396 b_midlen = len_b - pfx_length - sfx_length;
1397 if (a_midlen < 0)
1398 a_midlen = 0;
1399 if (b_midlen < 0)
1400 b_midlen = 0;
1401
1402 strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1403 if (pfx_length + sfx_length) {
1404 strbuf_add(&name, a, pfx_length);
1405 strbuf_addch(&name, '{');
1406 }
1407 strbuf_add(&name, a + pfx_length, a_midlen);
1408 strbuf_addstr(&name, " => ");
1409 strbuf_add(&name, b + pfx_length, b_midlen);
1410 if (pfx_length + sfx_length) {
1411 strbuf_addch(&name, '}');
1412 strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1413 }
1414 return strbuf_detach(&name, NULL);
1415 }
1416
1417 struct diffstat_t {
1418 int nr;
1419 int alloc;
1420 struct diffstat_file {
1421 char *from_name;
1422 char *name;
1423 char *print_name;
1424 unsigned is_unmerged:1;
1425 unsigned is_binary:1;
1426 unsigned is_renamed:1;
1427 unsigned is_interesting:1;
1428 uintmax_t added, deleted;
1429 } **files;
1430 };
1431
1432 static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1433 const char *name_a,
1434 const char *name_b)
1435 {
1436 struct diffstat_file *x;
1437 x = xcalloc(1, sizeof(*x));
1438 ALLOC_GROW(diffstat->files, diffstat->nr + 1, diffstat->alloc);
1439 diffstat->files[diffstat->nr++] = x;
1440 if (name_b) {
1441 x->from_name = xstrdup(name_a);
1442 x->name = xstrdup(name_b);
1443 x->is_renamed = 1;
1444 }
1445 else {
1446 x->from_name = NULL;
1447 x->name = xstrdup(name_a);
1448 }
1449 return x;
1450 }
1451
1452 static void diffstat_consume(void *priv, char *line, unsigned long len)
1453 {
1454 struct diffstat_t *diffstat = priv;
1455 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1456
1457 if (line[0] == '+')
1458 x->added++;
1459 else if (line[0] == '-')
1460 x->deleted++;
1461 }
1462
1463 const char mime_boundary_leader[] = "------------";
1464
1465 static int scale_linear(int it, int width, int max_change)
1466 {
1467 if (!it)
1468 return 0;
1469 /*
1470 * make sure that at least one '-' or '+' is printed if
1471 * there is any change to this path. The easiest way is to
1472 * scale linearly as if the alloted width is one column shorter
1473 * than it is, and then add 1 to the result.
1474 */
1475 return 1 + (it * (width - 1) / max_change);
1476 }
1477
1478 static void show_name(FILE *file,
1479 const char *prefix, const char *name, int len)
1480 {
1481 fprintf(file, " %s%-*s |", prefix, len, name);
1482 }
1483
1484 static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1485 {
1486 if (cnt <= 0)
1487 return;
1488 fprintf(file, "%s", set);
1489 while (cnt--)
1490 putc(ch, file);
1491 fprintf(file, "%s", reset);
1492 }
1493
1494 static void fill_print_name(struct diffstat_file *file)
1495 {
1496 char *pname;
1497
1498 if (file->print_name)
1499 return;
1500
1501 if (!file->is_renamed) {
1502 struct strbuf buf = STRBUF_INIT;
1503 if (quote_c_style(file->name, &buf, NULL, 0)) {
1504 pname = strbuf_detach(&buf, NULL);
1505 } else {
1506 pname = file->name;
1507 strbuf_release(&buf);
1508 }
1509 } else {
1510 pname = pprint_rename(file->from_name, file->name);
1511 }
1512 file->print_name = pname;
1513 }
1514
1515 int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1516 {
1517 struct strbuf sb = STRBUF_INIT;
1518 int ret;
1519
1520 if (!files) {
1521 assert(insertions == 0 && deletions == 0);
1522 return fprintf(fp, "%s\n", " 0 files changed");
1523 }
1524
1525 strbuf_addf(&sb,
1526 (files == 1) ? " %d file changed" : " %d files changed",
1527 files);
1528
1529 /*
1530 * For binary diff, the caller may want to print "x files
1531 * changed" with insertions == 0 && deletions == 0.
1532 *
1533 * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1534 * is probably less confusing (i.e skip over "2 files changed
1535 * but nothing about added/removed lines? Is this a bug in Git?").
1536 */
1537 if (insertions || deletions == 0) {
1538 strbuf_addf(&sb,
1539 (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
1540 insertions);
1541 }
1542
1543 if (deletions || insertions == 0) {
1544 strbuf_addf(&sb,
1545 (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
1546 deletions);
1547 }
1548 strbuf_addch(&sb, '\n');
1549 ret = fputs(sb.buf, fp);
1550 strbuf_release(&sb);
1551 return ret;
1552 }
1553
1554 static void show_stats(struct diffstat_t *data, struct diff_options *options)
1555 {
1556 int i, len, add, del, adds = 0, dels = 0;
1557 uintmax_t max_change = 0, max_len = 0;
1558 int total_files = data->nr, count;
1559 int width, name_width, graph_width, number_width = 0, bin_width = 0;
1560 const char *reset, *add_c, *del_c;
1561 const char *line_prefix = "";
1562 int extra_shown = 0;
1563
1564 if (data->nr == 0)
1565 return;
1566
1567 line_prefix = diff_line_prefix(options);
1568 count = options->stat_count ? options->stat_count : data->nr;
1569
1570 reset = diff_get_color_opt(options, DIFF_RESET);
1571 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1572 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1573
1574 /*
1575 * Find the longest filename and max number of changes
1576 */
1577 for (i = 0; (i < count) && (i < data->nr); i++) {
1578 struct diffstat_file *file = data->files[i];
1579 uintmax_t change = file->added + file->deleted;
1580
1581 if (!file->is_interesting && (change == 0)) {
1582 count++; /* not shown == room for one more */
1583 continue;
1584 }
1585 fill_print_name(file);
1586 len = strlen(file->print_name);
1587 if (max_len < len)
1588 max_len = len;
1589
1590 if (file->is_unmerged) {
1591 /* "Unmerged" is 8 characters */
1592 bin_width = bin_width < 8 ? 8 : bin_width;
1593 continue;
1594 }
1595 if (file->is_binary) {
1596 /* "Bin XXX -> YYY bytes" */
1597 int w = 14 + decimal_width(file->added)
1598 + decimal_width(file->deleted);
1599 bin_width = bin_width < w ? w : bin_width;
1600 /* Display change counts aligned with "Bin" */
1601 number_width = 3;
1602 continue;
1603 }
1604
1605 if (max_change < change)
1606 max_change = change;
1607 }
1608 count = i; /* where we can stop scanning in data->files[] */
1609
1610 /*
1611 * We have width = stat_width or term_columns() columns total.
1612 * We want a maximum of min(max_len, stat_name_width) for the name part.
1613 * We want a maximum of min(max_change, stat_graph_width) for the +- part.
1614 * We also need 1 for " " and 4 + decimal_width(max_change)
1615 * for " | NNNN " and one the empty column at the end, altogether
1616 * 6 + decimal_width(max_change).
1617 *
1618 * If there's not enough space, we will use the smaller of
1619 * stat_name_width (if set) and 5/8*width for the filename,
1620 * and the rest for constant elements + graph part, but no more
1621 * than stat_graph_width for the graph part.
1622 * (5/8 gives 50 for filename and 30 for the constant parts + graph
1623 * for the standard terminal size).
1624 *
1625 * In other words: stat_width limits the maximum width, and
1626 * stat_name_width fixes the maximum width of the filename,
1627 * and is also used to divide available columns if there
1628 * aren't enough.
1629 *
1630 * Binary files are displayed with "Bin XXX -> YYY bytes"
1631 * instead of the change count and graph. This part is treated
1632 * similarly to the graph part, except that it is not
1633 * "scaled". If total width is too small to accommodate the
1634 * guaranteed minimum width of the filename part and the
1635 * separators and this message, this message will "overflow"
1636 * making the line longer than the maximum width.
1637 */
1638
1639 if (options->stat_width == -1)
1640 width = term_columns() - strlen(line_prefix);
1641 else
1642 width = options->stat_width ? options->stat_width : 80;
1643 number_width = decimal_width(max_change) > number_width ?
1644 decimal_width(max_change) : number_width;
1645
1646 if (options->stat_graph_width == -1)
1647 options->stat_graph_width = diff_stat_graph_width;
1648
1649 /*
1650 * Guarantee 3/8*16==6 for the graph part
1651 * and 5/8*16==10 for the filename part
1652 */
1653 if (width < 16 + 6 + number_width)
1654 width = 16 + 6 + number_width;
1655
1656 /*
1657 * First assign sizes that are wanted, ignoring available width.
1658 * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
1659 * starting from "XXX" should fit in graph_width.
1660 */
1661 graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
1662 if (options->stat_graph_width &&
1663 options->stat_graph_width < graph_width)
1664 graph_width = options->stat_graph_width;
1665
1666 name_width = (options->stat_name_width > 0 &&
1667 options->stat_name_width < max_len) ?
1668 options->stat_name_width : max_len;
1669
1670 /*
1671 * Adjust adjustable widths not to exceed maximum width
1672 */
1673 if (name_width + number_width + 6 + graph_width > width) {
1674 if (graph_width > width * 3/8 - number_width - 6) {
1675 graph_width = width * 3/8 - number_width - 6;
1676 if (graph_width < 6)
1677 graph_width = 6;
1678 }
1679
1680 if (options->stat_graph_width &&
1681 graph_width > options->stat_graph_width)
1682 graph_width = options->stat_graph_width;
1683 if (name_width > width - number_width - 6 - graph_width)
1684 name_width = width - number_width - 6 - graph_width;
1685 else
1686 graph_width = width - number_width - 6 - name_width;
1687 }
1688
1689 /*
1690 * From here name_width is the width of the name area,
1691 * and graph_width is the width of the graph area.
1692 * max_change is used to scale graph properly.
1693 */
1694 for (i = 0; i < count; i++) {
1695 const char *prefix = "";
1696 struct diffstat_file *file = data->files[i];
1697 char *name = file->print_name;
1698 uintmax_t added = file->added;
1699 uintmax_t deleted = file->deleted;
1700 int name_len;
1701
1702 if (!file->is_interesting && (added + deleted == 0))
1703 continue;
1704
1705 /*
1706 * "scale" the filename
1707 */
1708 len = name_width;
1709 name_len = strlen(name);
1710 if (name_width < name_len) {
1711 char *slash;
1712 prefix = "...";
1713 len -= 3;
1714 name += name_len - len;
1715 slash = strchr(name, '/');
1716 if (slash)
1717 name = slash;
1718 }
1719
1720 if (file->is_binary) {
1721 fprintf(options->file, "%s", line_prefix);
1722 show_name(options->file, prefix, name, len);
1723 fprintf(options->file, " %*s", number_width, "Bin");
1724 if (!added && !deleted) {
1725 putc('\n', options->file);
1726 continue;
1727 }
1728 fprintf(options->file, " %s%"PRIuMAX"%s",
1729 del_c, deleted, reset);
1730 fprintf(options->file, " -> ");
1731 fprintf(options->file, "%s%"PRIuMAX"%s",
1732 add_c, added, reset);
1733 fprintf(options->file, " bytes");
1734 fprintf(options->file, "\n");
1735 continue;
1736 }
1737 else if (file->is_unmerged) {
1738 fprintf(options->file, "%s", line_prefix);
1739 show_name(options->file, prefix, name, len);
1740 fprintf(options->file, " Unmerged\n");
1741 continue;
1742 }
1743
1744 /*
1745 * scale the add/delete
1746 */
1747 add = added;
1748 del = deleted;
1749
1750 if (graph_width <= max_change) {
1751 int total = scale_linear(add + del, graph_width, max_change);
1752 if (total < 2 && add && del)
1753 /* width >= 2 due to the sanity check */
1754 total = 2;
1755 if (add < del) {
1756 add = scale_linear(add, graph_width, max_change);
1757 del = total - add;
1758 } else {
1759 del = scale_linear(del, graph_width, max_change);
1760 add = total - del;
1761 }
1762 }
1763 fprintf(options->file, "%s", line_prefix);
1764 show_name(options->file, prefix, name, len);
1765 fprintf(options->file, " %*"PRIuMAX"%s",
1766 number_width, added + deleted,
1767 added + deleted ? " " : "");
1768 show_graph(options->file, '+', add, add_c, reset);
1769 show_graph(options->file, '-', del, del_c, reset);
1770 fprintf(options->file, "\n");
1771 }
1772
1773 for (i = 0; i < data->nr; i++) {
1774 struct diffstat_file *file = data->files[i];
1775 uintmax_t added = file->added;
1776 uintmax_t deleted = file->deleted;
1777
1778 if (file->is_unmerged ||
1779 (!file->is_interesting && (added + deleted == 0))) {
1780 total_files--;
1781 continue;
1782 }
1783
1784 if (!file->is_binary) {
1785 adds += added;
1786 dels += deleted;
1787 }
1788 if (i < count)
1789 continue;
1790 if (!extra_shown)
1791 fprintf(options->file, "%s ...\n", line_prefix);
1792 extra_shown = 1;
1793 }
1794 fprintf(options->file, "%s", line_prefix);
1795 print_stat_summary(options->file, total_files, adds, dels);
1796 }
1797
1798 static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1799 {
1800 int i, adds = 0, dels = 0, total_files = data->nr;
1801
1802 if (data->nr == 0)
1803 return;
1804
1805 for (i = 0; i < data->nr; i++) {
1806 int added = data->files[i]->added;
1807 int deleted= data->files[i]->deleted;
1808
1809 if (data->files[i]->is_unmerged ||
1810 (!data->files[i]->is_interesting && (added + deleted == 0))) {
1811 total_files--;
1812 } else if (!data->files[i]->is_binary) { /* don't count bytes */
1813 adds += added;
1814 dels += deleted;
1815 }
1816 }
1817 fprintf(options->file, "%s", diff_line_prefix(options));
1818 print_stat_summary(options->file, total_files, adds, dels);
1819 }
1820
1821 static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1822 {
1823 int i;
1824
1825 if (data->nr == 0)
1826 return;
1827
1828 for (i = 0; i < data->nr; i++) {
1829 struct diffstat_file *file = data->files[i];
1830
1831 fprintf(options->file, "%s", diff_line_prefix(options));
1832
1833 if (file->is_binary)
1834 fprintf(options->file, "-\t-\t");
1835 else
1836 fprintf(options->file,
1837 "%"PRIuMAX"\t%"PRIuMAX"\t",
1838 file->added, file->deleted);
1839 if (options->line_termination) {
1840 fill_print_name(file);
1841 if (!file->is_renamed)
1842 write_name_quoted(file->name, options->file,
1843 options->line_termination);
1844 else {
1845 fputs(file->print_name, options->file);
1846 putc(options->line_termination, options->file);
1847 }
1848 } else {
1849 if (file->is_renamed) {
1850 putc('\0', options->file);
1851 write_name_quoted(file->from_name, options->file, '\0');
1852 }
1853 write_name_quoted(file->name, options->file, '\0');
1854 }
1855 }
1856 }
1857
1858 struct dirstat_file {
1859 const char *name;
1860 unsigned long changed;
1861 };
1862
1863 struct dirstat_dir {
1864 struct dirstat_file *files;
1865 int alloc, nr, permille, cumulative;
1866 };
1867
1868 static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1869 unsigned long changed, const char *base, int baselen)
1870 {
1871 unsigned long this_dir = 0;
1872 unsigned int sources = 0;
1873 const char *line_prefix = diff_line_prefix(opt);
1874
1875 while (dir->nr) {
1876 struct dirstat_file *f = dir->files;
1877 int namelen = strlen(f->name);
1878 unsigned long this;
1879 char *slash;
1880
1881 if (namelen < baselen)
1882 break;
1883 if (memcmp(f->name, base, baselen))
1884 break;
1885 slash = strchr(f->name + baselen, '/');
1886 if (slash) {
1887 int newbaselen = slash + 1 - f->name;
1888 this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
1889 sources++;
1890 } else {
1891 this = f->changed;
1892 dir->files++;
1893 dir->nr--;
1894 sources += 2;
1895 }
1896 this_dir += this;
1897 }
1898
1899 /*
1900 * We don't report dirstat's for
1901 * - the top level
1902 * - or cases where everything came from a single directory
1903 * under this directory (sources == 1).
1904 */
1905 if (baselen && sources != 1) {
1906 if (this_dir) {
1907 int permille = this_dir * 1000 / changed;
1908 if (permille >= dir->permille) {
1909 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
1910 permille / 10, permille % 10, baselen, base);
1911 if (!dir->cumulative)
1912 return 0;
1913 }
1914 }
1915 }
1916 return this_dir;
1917 }
1918
1919 static int dirstat_compare(const void *_a, const void *_b)
1920 {
1921 const struct dirstat_file *a = _a;
1922 const struct dirstat_file *b = _b;
1923 return strcmp(a->name, b->name);
1924 }
1925
1926 static void show_dirstat(struct diff_options *options)
1927 {
1928 int i;
1929 unsigned long changed;
1930 struct dirstat_dir dir;
1931 struct diff_queue_struct *q = &diff_queued_diff;
1932
1933 dir.files = NULL;
1934 dir.alloc = 0;
1935 dir.nr = 0;
1936 dir.permille = options->dirstat_permille;
1937 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1938
1939 changed = 0;
1940 for (i = 0; i < q->nr; i++) {
1941 struct diff_filepair *p = q->queue[i];
1942 const char *name;
1943 unsigned long copied, added, damage;
1944 int content_changed;
1945
1946 name = p->two->path ? p->two->path : p->one->path;
1947
1948 if (p->one->oid_valid && p->two->oid_valid)
1949 content_changed = oidcmp(&p->one->oid, &p->two->oid);
1950 else
1951 content_changed = 1;
1952
1953 if (!content_changed) {
1954 /*
1955 * The SHA1 has not changed, so pre-/post-content is
1956 * identical. We can therefore skip looking at the
1957 * file contents altogether.
1958 */
1959 damage = 0;
1960 goto found_damage;
1961 }
1962
1963 if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
1964 /*
1965 * In --dirstat-by-file mode, we don't really need to
1966 * look at the actual file contents at all.
1967 * The fact that the SHA1 changed is enough for us to
1968 * add this file to the list of results
1969 * (with each file contributing equal damage).
1970 */
1971 damage = 1;
1972 goto found_damage;
1973 }
1974
1975 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
1976 diff_populate_filespec(p->one, 0);
1977 diff_populate_filespec(p->two, 0);
1978 diffcore_count_changes(p->one, p->two, NULL, NULL, 0,
1979 &copied, &added);
1980 diff_free_filespec_data(p->one);
1981 diff_free_filespec_data(p->two);
1982 } else if (DIFF_FILE_VALID(p->one)) {
1983 diff_populate_filespec(p->one, CHECK_SIZE_ONLY);
1984 copied = added = 0;
1985 diff_free_filespec_data(p->one);
1986 } else if (DIFF_FILE_VALID(p->two)) {
1987 diff_populate_filespec(p->two, CHECK_SIZE_ONLY);
1988 copied = 0;
1989 added = p->two->size;
1990 diff_free_filespec_data(p->two);
1991 } else
1992 continue;
1993
1994 /*
1995 * Original minus copied is the removed material,
1996 * added is the new material. They are both damages
1997 * made to the preimage.
1998 * If the resulting damage is zero, we know that
1999 * diffcore_count_changes() considers the two entries to
2000 * be identical, but since content_changed is true, we
2001 * know that there must have been _some_ kind of change,
2002 * so we force all entries to have damage > 0.
2003 */
2004 damage = (p->one->size - copied) + added;
2005 if (!damage)
2006 damage = 1;
2007
2008 found_damage:
2009 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2010 dir.files[dir.nr].name = name;
2011 dir.files[dir.nr].changed = damage;
2012 changed += damage;
2013 dir.nr++;
2014 }
2015
2016 /* This can happen even with many files, if everything was renames */
2017 if (!changed)
2018 return;
2019
2020 /* Show all directories with more than x% of the changes */
2021 qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
2022 gather_dirstat(options, &dir, changed, "", 0);
2023 }
2024
2025 static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
2026 {
2027 int i;
2028 unsigned long changed;
2029 struct dirstat_dir dir;
2030
2031 if (data->nr == 0)
2032 return;
2033
2034 dir.files = NULL;
2035 dir.alloc = 0;
2036 dir.nr = 0;
2037 dir.permille = options->dirstat_permille;
2038 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
2039
2040 changed = 0;
2041 for (i = 0; i < data->nr; i++) {
2042 struct diffstat_file *file = data->files[i];
2043 unsigned long damage = file->added + file->deleted;
2044 if (file->is_binary)
2045 /*
2046 * binary files counts bytes, not lines. Must find some
2047 * way to normalize binary bytes vs. textual lines.
2048 * The following heuristic assumes that there are 64
2049 * bytes per "line".
2050 * This is stupid and ugly, but very cheap...
2051 */
2052 damage = (damage + 63) / 64;
2053 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2054 dir.files[dir.nr].name = file->name;
2055 dir.files[dir.nr].changed = damage;
2056 changed += damage;
2057 dir.nr++;
2058 }
2059
2060 /* This can happen even with many files, if everything was renames */
2061 if (!changed)
2062 return;
2063
2064 /* Show all directories with more than x% of the changes */
2065 qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
2066 gather_dirstat(options, &dir, changed, "", 0);
2067 }
2068
2069 static void free_diffstat_info(struct diffstat_t *diffstat)
2070 {
2071 int i;
2072 for (i = 0; i < diffstat->nr; i++) {
2073 struct diffstat_file *f = diffstat->files[i];
2074 if (f->name != f->print_name)
2075 free(f->print_name);
2076 free(f->name);
2077 free(f->from_name);
2078 free(f);
2079 }
2080 free(diffstat->files);
2081 }
2082
2083 struct checkdiff_t {
2084 const char *filename;
2085 int lineno;
2086 int conflict_marker_size;
2087 struct diff_options *o;
2088 unsigned ws_rule;
2089 unsigned status;
2090 };
2091
2092 static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2093 {
2094 char firstchar;
2095 int cnt;
2096
2097 if (len < marker_size + 1)
2098 return 0;
2099 firstchar = line[0];
2100 switch (firstchar) {
2101 case '=': case '>': case '<': case '|':
2102 break;
2103 default:
2104 return 0;
2105 }
2106 for (cnt = 1; cnt < marker_size; cnt++)
2107 if (line[cnt] != firstchar)
2108 return 0;
2109 /* line[1] thru line[marker_size-1] are same as firstchar */
2110 if (len < marker_size + 1 || !isspace(line[marker_size]))
2111 return 0;
2112 return 1;
2113 }
2114
2115 static void checkdiff_consume(void *priv, char *line, unsigned long len)
2116 {
2117 struct checkdiff_t *data = priv;
2118 int marker_size = data->conflict_marker_size;
2119 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2120 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2121 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2122 char *err;
2123 const char *line_prefix;
2124
2125 assert(data->o);
2126 line_prefix = diff_line_prefix(data->o);
2127
2128 if (line[0] == '+') {
2129 unsigned bad;
2130 data->lineno++;
2131 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2132 data->status |= 1;
2133 fprintf(data->o->file,
2134 "%s%s:%d: leftover conflict marker\n",
2135 line_prefix, data->filename, data->lineno);
2136 }
2137 bad = ws_check(line + 1, len - 1, data->ws_rule);
2138 if (!bad)
2139 return;
2140 data->status |= bad;
2141 err = whitespace_error_string(bad);
2142 fprintf(data->o->file, "%s%s:%d: %s.\n",
2143 line_prefix, data->filename, data->lineno, err);
2144 free(err);
2145 emit_line(data->o, set, reset, line, 1);
2146 ws_check_emit(line + 1, len - 1, data->ws_rule,
2147 data->o->file, set, reset, ws);
2148 } else if (line[0] == ' ') {
2149 data->lineno++;
2150 } else if (line[0] == '@') {
2151 char *plus = strchr(line, '+');
2152 if (plus)
2153 data->lineno = strtol(plus, NULL, 10) - 1;
2154 else
2155 die("invalid diff");
2156 }
2157 }
2158
2159 static unsigned char *deflate_it(char *data,
2160 unsigned long size,
2161 unsigned long *result_size)
2162 {
2163 int bound;
2164 unsigned char *deflated;
2165 git_zstream stream;
2166
2167 git_deflate_init(&stream, zlib_compression_level);
2168 bound = git_deflate_bound(&stream, size);
2169 deflated = xmalloc(bound);
2170 stream.next_out = deflated;
2171 stream.avail_out = bound;
2172
2173 stream.next_in = (unsigned char *)data;
2174 stream.avail_in = size;
2175 while (git_deflate(&stream, Z_FINISH) == Z_OK)
2176 ; /* nothing */
2177 git_deflate_end(&stream);
2178 *result_size = stream.total_out;
2179 return deflated;
2180 }
2181
2182 static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two,
2183 const char *prefix)
2184 {
2185 void *cp;
2186 void *delta;
2187 void *deflated;
2188 void *data;
2189 unsigned long orig_size;
2190 unsigned long delta_size;
2191 unsigned long deflate_size;
2192 unsigned long data_size;
2193
2194 /* We could do deflated delta, or we could do just deflated two,
2195 * whichever is smaller.
2196 */
2197 delta = NULL;
2198 deflated = deflate_it(two->ptr, two->size, &deflate_size);
2199 if (one->size && two->size) {
2200 delta = diff_delta(one->ptr, one->size,
2201 two->ptr, two->size,
2202 &delta_size, deflate_size);
2203 if (delta) {
2204 void *to_free = delta;
2205 orig_size = delta_size;
2206 delta = deflate_it(delta, delta_size, &delta_size);
2207 free(to_free);
2208 }
2209 }
2210
2211 if (delta && delta_size < deflate_size) {
2212 fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2213 free(deflated);
2214 data = delta;
2215 data_size = delta_size;
2216 }
2217 else {
2218 fprintf(file, "%sliteral %lu\n", prefix, two->size);
2219 free(delta);
2220 data = deflated;
2221 data_size = deflate_size;
2222 }
2223
2224 /* emit data encoded in base85 */
2225 cp = data;
2226 while (data_size) {
2227 int bytes = (52 < data_size) ? 52 : data_size;
2228 char line[70];
2229 data_size -= bytes;
2230 if (bytes <= 26)
2231 line[0] = bytes + 'A' - 1;
2232 else
2233 line[0] = bytes - 26 + 'a' - 1;
2234 encode_85(line + 1, cp, bytes);
2235 cp = (char *) cp + bytes;
2236 fprintf(file, "%s", prefix);
2237 fputs(line, file);
2238 fputc('\n', file);
2239 }
2240 fprintf(file, "%s\n", prefix);
2241 free(data);
2242 }
2243
2244 static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two,
2245 const char *prefix)
2246 {
2247 fprintf(file, "%sGIT binary patch\n", prefix);
2248 emit_binary_diff_body(file, one, two, prefix);
2249 emit_binary_diff_body(file, two, one, prefix);
2250 }
2251
2252 int diff_filespec_is_binary(struct diff_filespec *one)
2253 {
2254 if (one->is_binary == -1) {
2255 diff_filespec_load_driver(one);
2256 if (one->driver->binary != -1)
2257 one->is_binary = one->driver->binary;
2258 else {
2259 if (!one->data && DIFF_FILE_VALID(one))
2260 diff_populate_filespec(one, CHECK_BINARY);
2261 if (one->is_binary == -1 && one->data)
2262 one->is_binary = buffer_is_binary(one->data,
2263 one->size);
2264 if (one->is_binary == -1)
2265 one->is_binary = 0;
2266 }
2267 }
2268 return one->is_binary;
2269 }
2270
2271 static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2272 {
2273 diff_filespec_load_driver(one);
2274 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2275 }
2276
2277 void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2278 {
2279 if (!options->a_prefix)
2280 options->a_prefix = a;
2281 if (!options->b_prefix)
2282 options->b_prefix = b;
2283 }
2284
2285 struct userdiff_driver *get_textconv(struct diff_filespec *one)
2286 {
2287 if (!DIFF_FILE_VALID(one))
2288 return NULL;
2289
2290 diff_filespec_load_driver(one);
2291 return userdiff_get_textconv(one->driver);
2292 }
2293
2294 static void builtin_diff(const char *name_a,
2295 const char *name_b,
2296 struct diff_filespec *one,
2297 struct diff_filespec *two,
2298 const char *xfrm_msg,
2299 int must_show_header,
2300 struct diff_options *o,
2301 int complete_rewrite)
2302 {
2303 mmfile_t mf1, mf2;
2304 const char *lbl[2];
2305 char *a_one, *b_two;
2306 const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
2307 const char *reset = diff_get_color_opt(o, DIFF_RESET);
2308 const char *a_prefix, *b_prefix;
2309 struct userdiff_driver *textconv_one = NULL;
2310 struct userdiff_driver *textconv_two = NULL;
2311 struct strbuf header = STRBUF_INIT;
2312 const char *line_prefix = diff_line_prefix(o);
2313
2314 diff_set_mnemonic_prefix(o, "a/", "b/");
2315 if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2316 a_prefix = o->b_prefix;
2317 b_prefix = o->a_prefix;
2318 } else {
2319 a_prefix = o->a_prefix;
2320 b_prefix = o->b_prefix;
2321 }
2322
2323 if (o->submodule_format == DIFF_SUBMODULE_LOG &&
2324 (!one->mode || S_ISGITLINK(one->mode)) &&
2325 (!two->mode || S_ISGITLINK(two->mode))) {
2326 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2327 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2328 show_submodule_summary(o->file, one->path ? one->path : two->path,
2329 line_prefix,
2330 &one->oid, &two->oid,
2331 two->dirty_submodule,
2332 meta, del, add, reset);
2333 return;
2334 } else if (o->submodule_format == DIFF_SUBMODULE_INLINE_DIFF &&
2335 (!one->mode || S_ISGITLINK(one->mode)) &&
2336 (!two->mode || S_ISGITLINK(two->mode))) {
2337 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2338 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2339 show_submodule_inline_diff(o->file, one->path ? one->path : two->path,
2340 line_prefix,
2341 &one->oid, &two->oid,
2342 two->dirty_submodule,
2343 meta, del, add, reset, o);
2344 return;
2345 }
2346
2347 if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2348 textconv_one = get_textconv(one);
2349 textconv_two = get_textconv(two);
2350 }
2351
2352 /* Never use a non-valid filename anywhere if at all possible */
2353 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2354 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2355
2356 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2357 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2358 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2359 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2360 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
2361 if (lbl[0][0] == '/') {
2362 /* /dev/null */
2363 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
2364 if (xfrm_msg)
2365 strbuf_addstr(&header, xfrm_msg);
2366 must_show_header = 1;
2367 }
2368 else if (lbl[1][0] == '/') {
2369 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
2370 if (xfrm_msg)
2371 strbuf_addstr(&header, xfrm_msg);
2372 must_show_header = 1;
2373 }
2374 else {
2375 if (one->mode != two->mode) {
2376 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
2377 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
2378 must_show_header = 1;
2379 }
2380 if (xfrm_msg)
2381 strbuf_addstr(&header, xfrm_msg);
2382
2383 /*
2384 * we do not run diff between different kind
2385 * of objects.
2386 */
2387 if ((one->mode ^ two->mode) & S_IFMT)
2388 goto free_ab_and_return;
2389 if (complete_rewrite &&
2390 (textconv_one || !diff_filespec_is_binary(one)) &&
2391 (textconv_two || !diff_filespec_is_binary(two))) {
2392 fprintf(o->file, "%s", header.buf);
2393 strbuf_reset(&header);
2394 emit_rewrite_diff(name_a, name_b, one, two,
2395 textconv_one, textconv_two, o);
2396 o->found_changes = 1;
2397 goto free_ab_and_return;
2398 }
2399 }
2400
2401 if (o->irreversible_delete && lbl[1][0] == '/') {
2402 fprintf(o->file, "%s", header.buf);
2403 strbuf_reset(&header);
2404 goto free_ab_and_return;
2405 } else if (!DIFF_OPT_TST(o, TEXT) &&
2406 ( (!textconv_one && diff_filespec_is_binary(one)) ||
2407 (!textconv_two && diff_filespec_is_binary(two)) )) {
2408 if (!one->data && !two->data &&
2409 S_ISREG(one->mode) && S_ISREG(two->mode) &&
2410 !DIFF_OPT_TST(o, BINARY)) {
2411 if (!oidcmp(&one->oid, &two->oid)) {
2412 if (must_show_header)
2413 fprintf(o->file, "%s", header.buf);
2414 goto free_ab_and_return;
2415 }
2416 fprintf(o->file, "%s", header.buf);
2417 fprintf(o->file, "%sBinary files %s and %s differ\n",
2418 line_prefix, lbl[0], lbl[1]);
2419 goto free_ab_and_return;
2420 }
2421 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2422 die("unable to read files to diff");
2423 /* Quite common confusing case */
2424 if (mf1.size == mf2.size &&
2425 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2426 if (must_show_header)
2427 fprintf(o->file, "%s", header.buf);
2428 goto free_ab_and_return;
2429 }
2430 fprintf(o->file, "%s", header.buf);
2431 strbuf_reset(&header);
2432 if (DIFF_OPT_TST(o, BINARY))
2433 emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2434 else
2435 fprintf(o->file, "%sBinary files %s and %s differ\n",
2436 line_prefix, lbl[0], lbl[1]);
2437 o->found_changes = 1;
2438 } else {
2439 /* Crazy xdl interfaces.. */
2440 const char *diffopts = getenv("GIT_DIFF_OPTS");
2441 const char *v;
2442 xpparam_t xpp;
2443 xdemitconf_t xecfg;
2444 struct emit_callback ecbdata;
2445 const struct userdiff_funcname *pe;
2446
2447 if (must_show_header) {
2448 fprintf(o->file, "%s", header.buf);
2449 strbuf_reset(&header);
2450 }
2451
2452 mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2453 mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2454
2455 pe = diff_funcname_pattern(one);
2456 if (!pe)
2457 pe = diff_funcname_pattern(two);
2458
2459 memset(&xpp, 0, sizeof(xpp));
2460 memset(&xecfg, 0, sizeof(xecfg));
2461 memset(&ecbdata, 0, sizeof(ecbdata));
2462 ecbdata.label_path = lbl;
2463 ecbdata.color_diff = want_color(o->use_color);
2464 ecbdata.ws_rule = whitespace_rule(name_b);
2465 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2466 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2467 ecbdata.opt = o;
2468 ecbdata.header = header.len ? &header : NULL;
2469 xpp.flags = o->xdl_opts;
2470 xecfg.ctxlen = o->context;
2471 xecfg.interhunkctxlen = o->interhunkcontext;
2472 xecfg.flags = XDL_EMIT_FUNCNAMES;
2473 if (DIFF_OPT_TST(o, FUNCCONTEXT))
2474 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2475 if (pe)
2476 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2477 if (!diffopts)
2478 ;
2479 else if (skip_prefix(diffopts, "--unified=", &v))
2480 xecfg.ctxlen = strtoul(v, NULL, 10);
2481 else if (skip_prefix(diffopts, "-u", &v))
2482 xecfg.ctxlen = strtoul(v, NULL, 10);
2483 if (o->word_diff)
2484 init_diff_words_data(&ecbdata, o, one, two);
2485 if (xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2486 &xpp, &xecfg))
2487 die("unable to generate diff for %s", one->path);
2488 if (o->word_diff)
2489 free_diff_words_data(&ecbdata);
2490 if (textconv_one)
2491 free(mf1.ptr);
2492 if (textconv_two)
2493 free(mf2.ptr);
2494 xdiff_clear_find_func(&xecfg);
2495 }
2496
2497 free_ab_and_return:
2498 strbuf_release(&header);
2499 diff_free_filespec_data(one);
2500 diff_free_filespec_data(two);
2501 free(a_one);
2502 free(b_two);
2503 return;
2504 }
2505
2506 static void builtin_diffstat(const char *name_a, const char *name_b,
2507 struct diff_filespec *one,
2508 struct diff_filespec *two,
2509 struct diffstat_t *diffstat,
2510 struct diff_options *o,
2511 struct diff_filepair *p)
2512 {
2513 mmfile_t mf1, mf2;
2514 struct diffstat_file *data;
2515 int same_contents;
2516 int complete_rewrite = 0;
2517
2518 if (!DIFF_PAIR_UNMERGED(p)) {
2519 if (p->status == DIFF_STATUS_MODIFIED && p->score)
2520 complete_rewrite = 1;
2521 }
2522
2523 data = diffstat_add(diffstat, name_a, name_b);
2524 data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
2525
2526 if (!one || !two) {
2527 data->is_unmerged = 1;
2528 return;
2529 }
2530
2531 same_contents = !oidcmp(&one->oid, &two->oid);
2532
2533 if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2534 data->is_binary = 1;
2535 if (same_contents) {
2536 data->added = 0;
2537 data->deleted = 0;
2538 } else {
2539 data->added = diff_filespec_size(two);
2540 data->deleted = diff_filespec_size(one);
2541 }
2542 }
2543
2544 else if (complete_rewrite) {
2545 diff_populate_filespec(one, 0);
2546 diff_populate_filespec(two, 0);
2547 data->deleted = count_lines(one->data, one->size);
2548 data->added = count_lines(two->data, two->size);
2549 }
2550
2551 else if (!same_contents) {
2552 /* Crazy xdl interfaces.. */
2553 xpparam_t xpp;
2554 xdemitconf_t xecfg;
2555
2556 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2557 die("unable to read files to diff");
2558
2559 memset(&xpp, 0, sizeof(xpp));
2560 memset(&xecfg, 0, sizeof(xecfg));
2561 xpp.flags = o->xdl_opts;
2562 xecfg.ctxlen = o->context;
2563 xecfg.interhunkctxlen = o->interhunkcontext;
2564 if (xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2565 &xpp, &xecfg))
2566 die("unable to generate diffstat for %s", one->path);
2567 }
2568
2569 diff_free_filespec_data(one);
2570 diff_free_filespec_data(two);
2571 }
2572
2573 static void builtin_checkdiff(const char *name_a, const char *name_b,
2574 const char *attr_path,
2575 struct diff_filespec *one,
2576 struct diff_filespec *two,
2577 struct diff_options *o)
2578 {
2579 mmfile_t mf1, mf2;
2580 struct checkdiff_t data;
2581
2582 if (!two)
2583 return;
2584
2585 memset(&data, 0, sizeof(data));
2586 data.filename = name_b ? name_b : name_a;
2587 data.lineno = 0;
2588 data.o = o;
2589 data.ws_rule = whitespace_rule(attr_path);
2590 data.conflict_marker_size = ll_merge_marker_size(attr_path);
2591
2592 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2593 die("unable to read files to diff");
2594
2595 /*
2596 * All the other codepaths check both sides, but not checking
2597 * the "old" side here is deliberate. We are checking the newly
2598 * introduced changes, and as long as the "new" side is text, we
2599 * can and should check what it introduces.
2600 */
2601 if (diff_filespec_is_binary(two))
2602 goto free_and_return;
2603 else {
2604 /* Crazy xdl interfaces.. */
2605 xpparam_t xpp;
2606 xdemitconf_t xecfg;
2607
2608 memset(&xpp, 0, sizeof(xpp));
2609 memset(&xecfg, 0, sizeof(xecfg));
2610 xecfg.ctxlen = 1; /* at least one context line */
2611 xpp.flags = 0;
2612 if (xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2613 &xpp, &xecfg))
2614 die("unable to generate checkdiff for %s", one->path);
2615
2616 if (data.ws_rule & WS_BLANK_AT_EOF) {
2617 struct emit_callback ecbdata;
2618 int blank_at_eof;
2619
2620 ecbdata.ws_rule = data.ws_rule;
2621 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2622 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2623
2624 if (blank_at_eof) {
2625 static char *err;
2626 if (!err)
2627 err = whitespace_error_string(WS_BLANK_AT_EOF);
2628 fprintf(o->file, "%s:%d: %s.\n",
2629 data.filename, blank_at_eof, err);
2630 data.status = 1; /* report errors */
2631 }
2632 }
2633 }
2634 free_and_return:
2635 diff_free_filespec_data(one);
2636 diff_free_filespec_data(two);
2637 if (data.status)
2638 DIFF_OPT_SET(o, CHECK_FAILED);
2639 }
2640
2641 struct diff_filespec *alloc_filespec(const char *path)
2642 {
2643 struct diff_filespec *spec;
2644
2645 FLEXPTR_ALLOC_STR(spec, path, path);
2646 spec->count = 1;
2647 spec->is_binary = -1;
2648 return spec;
2649 }
2650
2651 void free_filespec(struct diff_filespec *spec)
2652 {
2653 if (!--spec->count) {
2654 diff_free_filespec_data(spec);
2655 free(spec);
2656 }
2657 }
2658
2659 void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
2660 int sha1_valid, unsigned short mode)
2661 {
2662 if (mode) {
2663 spec->mode = canon_mode(mode);
2664 hashcpy(spec->oid.hash, sha1);
2665 spec->oid_valid = sha1_valid;
2666 }
2667 }
2668
2669 /*
2670 * Given a name and sha1 pair, if the index tells us the file in
2671 * the work tree has that object contents, return true, so that
2672 * prepare_temp_file() does not have to inflate and extract.
2673 */
2674 static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
2675 {
2676 const struct cache_entry *ce;
2677 struct stat st;
2678 int pos, len;
2679
2680 /*
2681 * We do not read the cache ourselves here, because the
2682 * benchmark with my previous version that always reads cache
2683 * shows that it makes things worse for diff-tree comparing
2684 * two linux-2.6 kernel trees in an already checked out work
2685 * tree. This is because most diff-tree comparisons deal with
2686 * only a small number of files, while reading the cache is
2687 * expensive for a large project, and its cost outweighs the
2688 * savings we get by not inflating the object to a temporary
2689 * file. Practically, this code only helps when we are used
2690 * by diff-cache --cached, which does read the cache before
2691 * calling us.
2692 */
2693 if (!active_cache)
2694 return 0;
2695
2696 /* We want to avoid the working directory if our caller
2697 * doesn't need the data in a normal file, this system
2698 * is rather slow with its stat/open/mmap/close syscalls,
2699 * and the object is contained in a pack file. The pack
2700 * is probably already open and will be faster to obtain
2701 * the data through than the working directory. Loose
2702 * objects however would tend to be slower as they need
2703 * to be individually opened and inflated.
2704 */
2705 if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1))
2706 return 0;
2707
2708 /*
2709 * Similarly, if we'd have to convert the file contents anyway, that
2710 * makes the optimization not worthwhile.
2711 */
2712 if (!want_file && would_convert_to_git(name))
2713 return 0;
2714
2715 len = strlen(name);
2716 pos = cache_name_pos(name, len);
2717 if (pos < 0)
2718 return 0;
2719 ce = active_cache[pos];
2720
2721 /*
2722 * This is not the sha1 we are looking for, or
2723 * unreusable because it is not a regular file.
2724 */
2725 if (hashcmp(sha1, ce->oid.hash) || !S_ISREG(ce->ce_mode))
2726 return 0;
2727
2728 /*
2729 * If ce is marked as "assume unchanged", there is no
2730 * guarantee that work tree matches what we are looking for.
2731 */
2732 if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2733 return 0;
2734
2735 /*
2736 * If ce matches the file in the work tree, we can reuse it.
2737 */
2738 if (ce_uptodate(ce) ||
2739 (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2740 return 1;
2741
2742 return 0;
2743 }
2744
2745 static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2746 {
2747 struct strbuf buf = STRBUF_INIT;
2748 char *dirty = "";
2749
2750 /* Are we looking at the work tree? */
2751 if (s->dirty_submodule)
2752 dirty = "-dirty";
2753
2754 strbuf_addf(&buf, "Subproject commit %s%s\n",
2755 oid_to_hex(&s->oid), dirty);
2756 s->size = buf.len;
2757 if (size_only) {
2758 s->data = NULL;
2759 strbuf_release(&buf);
2760 } else {
2761 s->data = strbuf_detach(&buf, NULL);
2762 s->should_free = 1;
2763 }
2764 return 0;
2765 }
2766
2767 /*
2768 * While doing rename detection and pickaxe operation, we may need to
2769 * grab the data for the blob (or file) for our own in-core comparison.
2770 * diff_filespec has data and size fields for this purpose.
2771 */
2772 int diff_populate_filespec(struct diff_filespec *s, unsigned int flags)
2773 {
2774 int size_only = flags & CHECK_SIZE_ONLY;
2775 int err = 0;
2776 /*
2777 * demote FAIL to WARN to allow inspecting the situation
2778 * instead of refusing.
2779 */
2780 enum safe_crlf crlf_warn = (safe_crlf == SAFE_CRLF_FAIL
2781 ? SAFE_CRLF_WARN
2782 : safe_crlf);
2783
2784 if (!DIFF_FILE_VALID(s))
2785 die("internal error: asking to populate invalid file.");
2786 if (S_ISDIR(s->mode))
2787 return -1;
2788
2789 if (s->data)
2790 return 0;
2791
2792 if (size_only && 0 < s->size)
2793 return 0;
2794
2795 if (S_ISGITLINK(s->mode))
2796 return diff_populate_gitlink(s, size_only);
2797
2798 if (!s->oid_valid ||
2799 reuse_worktree_file(s->path, s->oid.hash, 0)) {
2800 struct strbuf buf = STRBUF_INIT;
2801 struct stat st;
2802 int fd;
2803
2804 if (lstat(s->path, &st) < 0) {
2805 if (errno == ENOENT) {
2806 err_empty:
2807 err = -1;
2808 empty:
2809 s->data = (char *)"";
2810 s->size = 0;
2811 return err;
2812 }
2813 }
2814 s->size = xsize_t(st.st_size);
2815 if (!s->size)
2816 goto empty;
2817 if (S_ISLNK(st.st_mode)) {
2818 struct strbuf sb = STRBUF_INIT;
2819
2820 if (strbuf_readlink(&sb, s->path, s->size))
2821 goto err_empty;
2822 s->size = sb.len;
2823 s->data = strbuf_detach(&sb, NULL);
2824 s->should_free = 1;
2825 return 0;
2826 }
2827 if (size_only)
2828 return 0;
2829 if ((flags & CHECK_BINARY) &&
2830 s->size > big_file_threshold && s->is_binary == -1) {
2831 s->is_binary = 1;
2832 return 0;
2833 }
2834 fd = open(s->path, O_RDONLY);
2835 if (fd < 0)
2836 goto err_empty;
2837 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2838 close(fd);
2839 s->should_munmap = 1;
2840
2841 /*
2842 * Convert from working tree format to canonical git format
2843 */
2844 if (convert_to_git(s->path, s->data, s->size, &buf, crlf_warn)) {
2845 size_t size = 0;
2846 munmap(s->data, s->size);
2847 s->should_munmap = 0;
2848 s->data = strbuf_detach(&buf, &size);
2849 s->size = size;
2850 s->should_free = 1;
2851 }
2852 }
2853 else {
2854 enum object_type type;
2855 if (size_only || (flags & CHECK_BINARY)) {
2856 type = sha1_object_info(s->oid.hash, &s->size);
2857 if (type < 0)
2858 die("unable to read %s",
2859 oid_to_hex(&s->oid));
2860 if (size_only)
2861 return 0;
2862 if (s->size > big_file_threshold && s->is_binary == -1) {
2863 s->is_binary = 1;
2864 return 0;
2865 }
2866 }
2867 s->data = read_sha1_file(s->oid.hash, &type, &s->size);
2868 if (!s->data)
2869 die("unable to read %s", oid_to_hex(&s->oid));
2870 s->should_free = 1;
2871 }
2872 return 0;
2873 }
2874
2875 void diff_free_filespec_blob(struct diff_filespec *s)
2876 {
2877 if (s->should_free)
2878 free(s->data);
2879 else if (s->should_munmap)
2880 munmap(s->data, s->size);
2881
2882 if (s->should_free || s->should_munmap) {
2883 s->should_free = s->should_munmap = 0;
2884 s->data = NULL;
2885 }
2886 }
2887
2888 void diff_free_filespec_data(struct diff_filespec *s)
2889 {
2890 diff_free_filespec_blob(s);
2891 free(s->cnt_data);
2892 s->cnt_data = NULL;
2893 }
2894
2895 static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2896 void *blob,
2897 unsigned long size,
2898 const struct object_id *oid,
2899 int mode)
2900 {
2901 int fd;
2902 struct strbuf buf = STRBUF_INIT;
2903 struct strbuf template = STRBUF_INIT;
2904 char *path_dup = xstrdup(path);
2905 const char *base = basename(path_dup);
2906
2907 /* Generate "XXXXXX_basename.ext" */
2908 strbuf_addstr(&template, "XXXXXX_");
2909 strbuf_addstr(&template, base);
2910
2911 fd = mks_tempfile_ts(&temp->tempfile, template.buf, strlen(base) + 1);
2912 if (fd < 0)
2913 die_errno("unable to create temp-file");
2914 if (convert_to_working_tree(path,
2915 (const char *)blob, (size_t)size, &buf)) {
2916 blob = buf.buf;
2917 size = buf.len;
2918 }
2919 if (write_in_full(fd, blob, size) != size)
2920 die_errno("unable to write temp-file");
2921 close_tempfile(&temp->tempfile);
2922 temp->name = get_tempfile_path(&temp->tempfile);
2923 oid_to_hex_r(temp->hex, oid);
2924 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", mode);
2925 strbuf_release(&buf);
2926 strbuf_release(&template);
2927 free(path_dup);
2928 }
2929
2930 static struct diff_tempfile *prepare_temp_file(const char *name,
2931 struct diff_filespec *one)
2932 {
2933 struct diff_tempfile *temp = claim_diff_tempfile();
2934
2935 if (!DIFF_FILE_VALID(one)) {
2936 not_a_valid_file:
2937 /* A '-' entry produces this for file-2, and
2938 * a '+' entry produces this for file-1.
2939 */
2940 temp->name = "/dev/null";
2941 xsnprintf(temp->hex, sizeof(temp->hex), ".");
2942 xsnprintf(temp->mode, sizeof(temp->mode), ".");
2943 return temp;
2944 }
2945
2946 if (!S_ISGITLINK(one->mode) &&
2947 (!one->oid_valid ||
2948 reuse_worktree_file(name, one->oid.hash, 1))) {
2949 struct stat st;
2950 if (lstat(name, &st) < 0) {
2951 if (errno == ENOENT)
2952 goto not_a_valid_file;
2953 die_errno("stat(%s)", name);
2954 }
2955 if (S_ISLNK(st.st_mode)) {
2956 struct strbuf sb = STRBUF_INIT;
2957 if (strbuf_readlink(&sb, name, st.st_size) < 0)
2958 die_errno("readlink(%s)", name);
2959 prep_temp_blob(name, temp, sb.buf, sb.len,
2960 (one->oid_valid ?
2961 &one->oid : &null_oid),
2962 (one->oid_valid ?
2963 one->mode : S_IFLNK));
2964 strbuf_release(&sb);
2965 }
2966 else {
2967 /* we can borrow from the file in the work tree */
2968 temp->name = name;
2969 if (!one->oid_valid)
2970 sha1_to_hex_r(temp->hex, null_sha1);
2971 else
2972 sha1_to_hex_r(temp->hex, one->oid.hash);
2973 /* Even though we may sometimes borrow the
2974 * contents from the work tree, we always want
2975 * one->mode. mode is trustworthy even when
2976 * !(one->sha1_valid), as long as
2977 * DIFF_FILE_VALID(one).
2978 */
2979 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", one->mode);
2980 }
2981 return temp;
2982 }
2983 else {
2984 if (diff_populate_filespec(one, 0))
2985 die("cannot read data blob for %s", one->path);
2986 prep_temp_blob(name, temp, one->data, one->size,
2987 &one->oid, one->mode);
2988 }
2989 return temp;
2990 }
2991
2992 static void add_external_diff_name(struct argv_array *argv,
2993 const char *name,
2994 struct diff_filespec *df)
2995 {
2996 struct diff_tempfile *temp = prepare_temp_file(name, df);
2997 argv_array_push(argv, temp->name);
2998 argv_array_push(argv, temp->hex);
2999 argv_array_push(argv, temp->mode);
3000 }
3001
3002 /* An external diff command takes:
3003 *
3004 * diff-cmd name infile1 infile1-sha1 infile1-mode \
3005 * infile2 infile2-sha1 infile2-mode [ rename-to ]
3006 *
3007 */
3008 static void run_external_diff(const char *pgm,
3009 const char *name,
3010 const char *other,
3011 struct diff_filespec *one,
3012 struct diff_filespec *two,
3013 const char *xfrm_msg,
3014 int complete_rewrite,
3015 struct diff_options *o)
3016 {
3017 struct argv_array argv = ARGV_ARRAY_INIT;
3018 struct argv_array env = ARGV_ARRAY_INIT;
3019 struct diff_queue_struct *q = &diff_queued_diff;
3020
3021 argv_array_push(&argv, pgm);
3022 argv_array_push(&argv, name);
3023
3024 if (one && two) {
3025 add_external_diff_name(&argv, name, one);
3026 if (!other)
3027 add_external_diff_name(&argv, name, two);
3028 else {
3029 add_external_diff_name(&argv, other, two);
3030 argv_array_push(&argv, other);
3031 argv_array_push(&argv, xfrm_msg);
3032 }
3033 }
3034
3035 argv_array_pushf(&env, "GIT_DIFF_PATH_COUNTER=%d", ++o->diff_path_counter);
3036 argv_array_pushf(&env, "GIT_DIFF_PATH_TOTAL=%d", q->nr);
3037
3038 if (run_command_v_opt_cd_env(argv.argv, RUN_USING_SHELL, NULL, env.argv))
3039 die(_("external diff died, stopping at %s"), name);
3040
3041 remove_tempfile();
3042 argv_array_clear(&argv);
3043 argv_array_clear(&env);
3044 }
3045
3046 static int similarity_index(struct diff_filepair *p)
3047 {
3048 return p->score * 100 / MAX_SCORE;
3049 }
3050
3051 static void fill_metainfo(struct strbuf *msg,
3052 const char *name,
3053 const char *other,
3054 struct diff_filespec *one,
3055 struct diff_filespec *two,
3056 struct diff_options *o,
3057 struct diff_filepair *p,
3058 int *must_show_header,
3059 int use_color)
3060 {
3061 const char *set = diff_get_color(use_color, DIFF_METAINFO);
3062 const char *reset = diff_get_color(use_color, DIFF_RESET);
3063 const char *line_prefix = diff_line_prefix(o);
3064
3065 *must_show_header = 1;
3066 strbuf_init(msg, PATH_MAX * 2 + 300);
3067 switch (p->status) {
3068 case DIFF_STATUS_COPIED:
3069 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3070 line_prefix, set, similarity_index(p));
3071 strbuf_addf(msg, "%s\n%s%scopy from ",
3072 reset, line_prefix, set);
3073 quote_c_style(name, msg, NULL, 0);
3074 strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
3075 quote_c_style(other, msg, NULL, 0);
3076 strbuf_addf(msg, "%s\n", reset);
3077 break;
3078 case DIFF_STATUS_RENAMED:
3079 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3080 line_prefix, set, similarity_index(p));
3081 strbuf_addf(msg, "%s\n%s%srename from ",
3082 reset, line_prefix, set);
3083 quote_c_style(name, msg, NULL, 0);
3084 strbuf_addf(msg, "%s\n%s%srename to ",
3085 reset, line_prefix, set);
3086 quote_c_style(other, msg, NULL, 0);
3087 strbuf_addf(msg, "%s\n", reset);
3088 break;
3089 case DIFF_STATUS_MODIFIED:
3090 if (p->score) {
3091 strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
3092 line_prefix,
3093 set, similarity_index(p), reset);
3094 break;
3095 }
3096 /* fallthru */
3097 default:
3098 *must_show_header = 0;
3099 }
3100 if (one && two && oidcmp(&one->oid, &two->oid)) {
3101 int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
3102
3103 if (DIFF_OPT_TST(o, BINARY)) {
3104 mmfile_t mf;
3105 if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
3106 (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
3107 abbrev = 40;
3108 }
3109 strbuf_addf(msg, "%s%sindex %s..", line_prefix, set,
3110 find_unique_abbrev(one->oid.hash, abbrev));
3111 strbuf_addstr(msg, find_unique_abbrev(two->oid.hash, abbrev));
3112 if (one->mode == two->mode)
3113 strbuf_addf(msg, " %06o", one->mode);
3114 strbuf_addf(msg, "%s\n", reset);
3115 }
3116 }
3117
3118 static void run_diff_cmd(const char *pgm,
3119 const char *name,
3120 const char *other,
3121 const char *attr_path,
3122 struct diff_filespec *one,
3123 struct diff_filespec *two,
3124 struct strbuf *msg,
3125 struct diff_options *o,
3126 struct diff_filepair *p)
3127 {
3128 const char *xfrm_msg = NULL;
3129 int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
3130 int must_show_header = 0;
3131
3132
3133 if (DIFF_OPT_TST(o, ALLOW_EXTERNAL)) {
3134 struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
3135 if (drv && drv->external)
3136 pgm = drv->external;
3137 }
3138
3139 if (msg) {
3140 /*
3141 * don't use colors when the header is intended for an
3142 * external diff driver
3143 */
3144 fill_metainfo(msg, name, other, one, two, o, p,
3145 &must_show_header,
3146 want_color(o->use_color) && !pgm);
3147 xfrm_msg = msg->len ? msg->buf : NULL;
3148 }
3149
3150 if (pgm) {
3151 run_external_diff(pgm, name, other, one, two, xfrm_msg,
3152 complete_rewrite, o);
3153 return;
3154 }
3155 if (one && two)
3156 builtin_diff(name, other ? other : name,
3157 one, two, xfrm_msg, must_show_header,
3158 o, complete_rewrite);
3159 else
3160 fprintf(o->file, "* Unmerged path %s\n", name);
3161 }
3162
3163 static void diff_fill_sha1_info(struct diff_filespec *one)
3164 {
3165 if (DIFF_FILE_VALID(one)) {
3166 if (!one->oid_valid) {
3167 struct stat st;
3168 if (one->is_stdin) {
3169 oidclr(&one->oid);
3170 return;
3171 }
3172 if (lstat(one->path, &st) < 0)
3173 die_errno("stat '%s'", one->path);
3174 if (index_path(one->oid.hash, one->path, &st, 0))
3175 die("cannot hash %s", one->path);
3176 }
3177 }
3178 else
3179 oidclr(&one->oid);
3180 }
3181
3182 static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
3183 {
3184 /* Strip the prefix but do not molest /dev/null and absolute paths */
3185 if (*namep && **namep != '/') {
3186 *namep += prefix_length;
3187 if (**namep == '/')
3188 ++*namep;
3189 }
3190 if (*otherp && **otherp != '/') {
3191 *otherp += prefix_length;
3192 if (**otherp == '/')
3193 ++*otherp;
3194 }
3195 }
3196
3197 static void run_diff(struct diff_filepair *p, struct diff_options *o)
3198 {
3199 const char *pgm = external_diff();
3200 struct strbuf msg;
3201 struct diff_filespec *one = p->one;
3202 struct diff_filespec *two = p->two;
3203 const char *name;
3204 const char *other;
3205 const char *attr_path;
3206
3207 name = p->one->path;
3208 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3209 attr_path = name;
3210 if (o->prefix_length)
3211 strip_prefix(o->prefix_length, &name, &other);
3212
3213 if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
3214 pgm = NULL;
3215
3216 if (DIFF_PAIR_UNMERGED(p)) {
3217 run_diff_cmd(pgm, name, NULL, attr_path,
3218 NULL, NULL, NULL, o, p);
3219 return;
3220 }
3221
3222 diff_fill_sha1_info(one);
3223 diff_fill_sha1_info(two);
3224
3225 if (!pgm &&
3226 DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3227 (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3228 /*
3229 * a filepair that changes between file and symlink
3230 * needs to be split into deletion and creation.
3231 */
3232 struct diff_filespec *null = alloc_filespec(two->path);
3233 run_diff_cmd(NULL, name, other, attr_path,
3234 one, null, &msg, o, p);
3235 free(null);
3236 strbuf_release(&msg);
3237
3238 null = alloc_filespec(one->path);
3239 run_diff_cmd(NULL, name, other, attr_path,
3240 null, two, &msg, o, p);
3241 free(null);
3242 }
3243 else
3244 run_diff_cmd(pgm, name, other, attr_path,
3245 one, two, &msg, o, p);
3246
3247 strbuf_release(&msg);
3248 }
3249
3250 static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3251 struct diffstat_t *diffstat)
3252 {
3253 const char *name;
3254 const char *other;
3255
3256 if (DIFF_PAIR_UNMERGED(p)) {
3257 /* unmerged */
3258 builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, p);
3259 return;
3260 }
3261
3262 name = p->one->path;
3263 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3264
3265 if (o->prefix_length)
3266 strip_prefix(o->prefix_length, &name, &other);
3267
3268 diff_fill_sha1_info(p->one);
3269 diff_fill_sha1_info(p->two);
3270
3271 builtin_diffstat(name, other, p->one, p->two, diffstat, o, p);
3272 }
3273
3274 static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3275 {
3276 const char *name;
3277 const char *other;
3278 const char *attr_path;
3279
3280 if (DIFF_PAIR_UNMERGED(p)) {
3281 /* unmerged */
3282 return;
3283 }
3284
3285 name = p->one->path;
3286 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3287 attr_path = other ? other : name;
3288
3289 if (o->prefix_length)
3290 strip_prefix(o->prefix_length, &name, &other);
3291
3292 diff_fill_sha1_info(p->one);
3293 diff_fill_sha1_info(p->two);
3294
3295 builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3296 }
3297
3298 void diff_setup(struct diff_options *options)
3299 {
3300 memcpy(options, &default_diff_options, sizeof(*options));
3301
3302 options->file = stdout;
3303
3304 options->line_termination = '\n';
3305 options->break_opt = -1;
3306 options->rename_limit = -1;
3307 options->dirstat_permille = diff_dirstat_permille_default;
3308 options->context = diff_context_default;
3309 options->ws_error_highlight = WSEH_NEW;
3310 DIFF_OPT_SET(options, RENAME_EMPTY);
3311
3312 /* pathchange left =NULL by default */
3313 options->change = diff_change;
3314 options->add_remove = diff_addremove;
3315 options->use_color = diff_use_color_default;
3316 options->detect_rename = diff_detect_rename_default;
3317 options->xdl_opts |= diff_algorithm;
3318 if (diff_indent_heuristic)
3319 DIFF_XDL_SET(options, INDENT_HEURISTIC);
3320 else if (diff_compaction_heuristic)
3321 DIFF_XDL_SET(options, COMPACTION_HEURISTIC);
3322
3323 options->orderfile = diff_order_file_cfg;
3324
3325 if (diff_no_prefix) {
3326 options->a_prefix = options->b_prefix = "";
3327 } else if (!diff_mnemonic_prefix) {
3328 options->a_prefix = "a/";
3329 options->b_prefix = "b/";
3330 }
3331 }
3332
3333 void diff_setup_done(struct diff_options *options)
3334 {
3335 int count = 0;
3336
3337 if (options->set_default)
3338 options->set_default(options);
3339
3340 if (options->output_format & DIFF_FORMAT_NAME)
3341 count++;
3342 if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3343 count++;
3344 if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3345 count++;
3346 if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3347 count++;
3348 if (count > 1)
3349 die(_("--name-only, --name-status, --check and -s are mutually exclusive"));
3350
3351 /*
3352 * Most of the time we can say "there are changes"
3353 * only by checking if there are changed paths, but
3354 * --ignore-whitespace* options force us to look
3355 * inside contents.
3356 */
3357
3358 if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3359 DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3360 DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3361 DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3362 else
3363 DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3364
3365 if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3366 options->detect_rename = DIFF_DETECT_COPY;
3367
3368 if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3369 options->prefix = NULL;
3370 if (options->prefix)
3371 options->prefix_length = strlen(options->prefix);
3372 else
3373 options->prefix_length = 0;
3374
3375 if (options->output_format & (DIFF_FORMAT_NAME |
3376 DIFF_FORMAT_NAME_STATUS |
3377 DIFF_FORMAT_CHECKDIFF |
3378 DIFF_FORMAT_NO_OUTPUT))
3379 options->output_format &= ~(DIFF_FORMAT_RAW |
3380 DIFF_FORMAT_NUMSTAT |
3381 DIFF_FORMAT_DIFFSTAT |
3382 DIFF_FORMAT_SHORTSTAT |
3383 DIFF_FORMAT_DIRSTAT |
3384 DIFF_FORMAT_SUMMARY |
3385 DIFF_FORMAT_PATCH);
3386
3387 /*
3388 * These cases always need recursive; we do not drop caller-supplied
3389 * recursive bits for other formats here.
3390 */
3391 if (options->output_format & (DIFF_FORMAT_PATCH |
3392 DIFF_FORMAT_NUMSTAT |
3393 DIFF_FORMAT_DIFFSTAT |
3394 DIFF_FORMAT_SHORTSTAT |
3395 DIFF_FORMAT_DIRSTAT |
3396 DIFF_FORMAT_SUMMARY |
3397 DIFF_FORMAT_CHECKDIFF))
3398 DIFF_OPT_SET(options, RECURSIVE);
3399 /*
3400 * Also pickaxe would not work very well if you do not say recursive
3401 */
3402 if (options->pickaxe)
3403 DIFF_OPT_SET(options, RECURSIVE);
3404 /*
3405 * When patches are generated, submodules diffed against the work tree
3406 * must be checked for dirtiness too so it can be shown in the output
3407 */
3408 if (options->output_format & DIFF_FORMAT_PATCH)
3409 DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3410
3411 if (options->detect_rename && options->rename_limit < 0)
3412 options->rename_limit = diff_rename_limit_default;
3413 if (options->setup & DIFF_SETUP_USE_CACHE) {
3414 if (!active_cache)
3415 /* read-cache does not die even when it fails
3416 * so it is safe for us to do this here. Also
3417 * it does not smudge active_cache or active_nr
3418 * when it fails, so we do not have to worry about
3419 * cleaning it up ourselves either.
3420 */
3421 read_cache();
3422 }
3423 if (options->abbrev <= 0 || 40 < options->abbrev)
3424 options->abbrev = 40; /* full */
3425
3426 /*
3427 * It does not make sense to show the first hit we happened
3428 * to have found. It does not make sense not to return with
3429 * exit code in such a case either.
3430 */
3431 if (DIFF_OPT_TST(options, QUICK)) {
3432 options->output_format = DIFF_FORMAT_NO_OUTPUT;
3433 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3434 }
3435
3436 options->diff_path_counter = 0;
3437
3438 if (DIFF_OPT_TST(options, FOLLOW_RENAMES) && options->pathspec.nr != 1)
3439 die(_("--follow requires exactly one pathspec"));
3440 }
3441
3442 static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3443 {
3444 char c, *eq;
3445 int len;
3446
3447 if (*arg != '-')
3448 return 0;
3449 c = *++arg;
3450 if (!c)
3451 return 0;
3452 if (c == arg_short) {
3453 c = *++arg;
3454 if (!c)
3455 return 1;
3456 if (val && isdigit(c)) {
3457 char *end;
3458 int n = strtoul(arg, &end, 10);
3459 if (*end)
3460 return 0;
3461 *val = n;
3462 return 1;
3463 }
3464 return 0;
3465 }
3466 if (c != '-')
3467 return 0;
3468 arg++;
3469 eq = strchrnul(arg, '=');
3470 len = eq - arg;
3471 if (!len || strncmp(arg, arg_long, len))
3472 return 0;
3473 if (*eq) {
3474 int n;
3475 char *end;
3476 if (!isdigit(*++eq))
3477 return 0;
3478 n = strtoul(eq, &end, 10);
3479 if (*end)
3480 return 0;
3481 *val = n;
3482 }
3483 return 1;
3484 }
3485
3486 static int diff_scoreopt_parse(const char *opt);
3487
3488 static inline int short_opt(char opt, const char **argv,
3489 const char **optarg)
3490 {
3491 const char *arg = argv[0];
3492 if (arg[0] != '-' || arg[1] != opt)
3493 return 0;
3494 if (arg[2] != '\0') {
3495 *optarg = arg + 2;
3496 return 1;
3497 }
3498 if (!argv[1])
3499 die("Option '%c' requires a value", opt);
3500 *optarg = argv[1];
3501 return 2;
3502 }
3503
3504 int parse_long_opt(const char *opt, const char **argv,
3505 const char **optarg)
3506 {
3507 const char *arg = argv[0];
3508 if (!skip_prefix(arg, "--", &arg))
3509 return 0;
3510 if (!skip_prefix(arg, opt, &arg))
3511 return 0;
3512 if (*arg == '=') { /* stuck form: --option=value */
3513 *optarg = arg + 1;
3514 return 1;
3515 }
3516 if (*arg != '\0')
3517 return 0;
3518 /* separate form: --option value */
3519 if (!argv[1])
3520 die("Option '--%s' requires a value", opt);
3521 *optarg = argv[1];
3522 return 2;
3523 }
3524
3525 static int stat_opt(struct diff_options *options, const char **av)
3526 {
3527 const char *arg = av[0];
3528 char *end;
3529 int width = options->stat_width;
3530 int name_width = options->stat_name_width;
3531 int graph_width = options->stat_graph_width;
3532 int count = options->stat_count;
3533 int argcount = 1;
3534
3535 if (!skip_prefix(arg, "--stat", &arg))
3536 die("BUG: stat option does not begin with --stat: %s", arg);
3537 end = (char *)arg;
3538
3539 switch (*arg) {
3540 case '-':
3541 if (skip_prefix(arg, "-width", &arg)) {
3542 if (*arg == '=')
3543 width = strtoul(arg + 1, &end, 10);
3544 else if (!*arg && !av[1])
3545 die_want_option("--stat-width");
3546 else if (!*arg) {
3547 width = strtoul(av[1], &end, 10);
3548 argcount = 2;
3549 }
3550 } else if (skip_prefix(arg, "-name-width", &arg)) {
3551 if (*arg == '=')
3552 name_width = strtoul(arg + 1, &end, 10);
3553 else if (!*arg && !av[1])
3554 die_want_option("--stat-name-width");
3555 else if (!*arg) {
3556 name_width = strtoul(av[1], &end, 10);
3557 argcount = 2;
3558 }
3559 } else if (skip_prefix(arg, "-graph-width", &arg)) {
3560 if (*arg == '=')
3561 graph_width = strtoul(arg + 1, &end, 10);
3562 else if (!*arg && !av[1])
3563 die_want_option("--stat-graph-width");
3564 else if (!*arg) {
3565 graph_width = strtoul(av[1], &end, 10);
3566 argcount = 2;
3567 }
3568 } else if (skip_prefix(arg, "-count", &arg)) {
3569 if (*arg == '=')
3570 count = strtoul(arg + 1, &end, 10);
3571 else if (!*arg && !av[1])
3572 die_want_option("--stat-count");
3573 else if (!*arg) {
3574 count = strtoul(av[1], &end, 10);
3575 argcount = 2;
3576 }
3577 }
3578 break;
3579 case '=':
3580 width = strtoul(arg+1, &end, 10);
3581 if (*end == ',')
3582 name_width = strtoul(end+1, &end, 10);
3583 if (*end == ',')
3584 count = strtoul(end+1, &end, 10);
3585 }
3586
3587 /* Important! This checks all the error cases! */
3588 if (*end)
3589 return 0;
3590 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3591 options->stat_name_width = name_width;
3592 options->stat_graph_width = graph_width;
3593 options->stat_width = width;
3594 options->stat_count = count;
3595 return argcount;
3596 }
3597
3598 static int parse_dirstat_opt(struct diff_options *options, const char *params)
3599 {
3600 struct strbuf errmsg = STRBUF_INIT;
3601 if (parse_dirstat_params(options, params, &errmsg))
3602 die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3603 errmsg.buf);
3604 strbuf_release(&errmsg);
3605 /*
3606 * The caller knows a dirstat-related option is given from the command
3607 * line; allow it to say "return this_function();"
3608 */
3609 options->output_format |= DIFF_FORMAT_DIRSTAT;
3610 return 1;
3611 }
3612
3613 static int parse_submodule_opt(struct diff_options *options, const char *value)
3614 {
3615 if (parse_submodule_params(options, value))
3616 die(_("Failed to parse --submodule option parameter: '%s'"),
3617 value);
3618 return 1;
3619 }
3620
3621 static const char diff_status_letters[] = {
3622 DIFF_STATUS_ADDED,
3623 DIFF_STATUS_COPIED,
3624 DIFF_STATUS_DELETED,
3625 DIFF_STATUS_MODIFIED,
3626 DIFF_STATUS_RENAMED,
3627 DIFF_STATUS_TYPE_CHANGED,
3628 DIFF_STATUS_UNKNOWN,
3629 DIFF_STATUS_UNMERGED,
3630 DIFF_STATUS_FILTER_AON,
3631 DIFF_STATUS_FILTER_BROKEN,
3632 '\0',
3633 };
3634
3635 static unsigned int filter_bit['Z' + 1];
3636
3637 static void prepare_filter_bits(void)
3638 {
3639 int i;
3640
3641 if (!filter_bit[DIFF_STATUS_ADDED]) {
3642 for (i = 0; diff_status_letters[i]; i++)
3643 filter_bit[(int) diff_status_letters[i]] = (1 << i);
3644 }
3645 }
3646
3647 static unsigned filter_bit_tst(char status, const struct diff_options *opt)
3648 {
3649 return opt->filter & filter_bit[(int) status];
3650 }
3651
3652 static int parse_diff_filter_opt(const char *optarg, struct diff_options *opt)
3653 {
3654 int i, optch;
3655
3656 prepare_filter_bits();
3657
3658 /*
3659 * If there is a negation e.g. 'd' in the input, and we haven't
3660 * initialized the filter field with another --diff-filter, start
3661 * from full set of bits, except for AON.
3662 */
3663 if (!opt->filter) {
3664 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3665 if (optch < 'a' || 'z' < optch)
3666 continue;
3667 opt->filter = (1 << (ARRAY_SIZE(diff_status_letters) - 1)) - 1;
3668 opt->filter &= ~filter_bit[DIFF_STATUS_FILTER_AON];
3669 break;
3670 }
3671 }
3672
3673 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3674 unsigned int bit;
3675 int negate;
3676
3677 if ('a' <= optch && optch <= 'z') {
3678 negate = 1;
3679 optch = toupper(optch);
3680 } else {
3681 negate = 0;
3682 }
3683
3684 bit = (0 <= optch && optch <= 'Z') ? filter_bit[optch] : 0;
3685 if (!bit)
3686 return optarg[i];
3687 if (negate)
3688 opt->filter &= ~bit;
3689 else
3690 opt->filter |= bit;
3691 }
3692 return 0;
3693 }
3694
3695 static void enable_patch_output(int *fmt) {
3696 *fmt &= ~DIFF_FORMAT_NO_OUTPUT;
3697 *fmt |= DIFF_FORMAT_PATCH;
3698 }
3699
3700 static int parse_one_token(const char **arg, const char *token)
3701 {
3702 const char *rest;
3703 if (skip_prefix(*arg, token, &rest) && (!*rest || *rest == ',')) {
3704 *arg = rest;
3705 return 1;
3706 }
3707 return 0;
3708 }
3709
3710 static int parse_ws_error_highlight(struct diff_options *opt, const char *arg)
3711 {
3712 const char *orig_arg = arg;
3713 unsigned val = 0;
3714 while (*arg) {
3715 if (parse_one_token(&arg, "none"))
3716 val = 0;
3717 else if (parse_one_token(&arg, "default"))
3718 val = WSEH_NEW;
3719 else if (parse_one_token(&arg, "all"))
3720 val = WSEH_NEW | WSEH_OLD | WSEH_CONTEXT;
3721 else if (parse_one_token(&arg, "new"))
3722 val |= WSEH_NEW;
3723 else if (parse_one_token(&arg, "old"))
3724 val |= WSEH_OLD;
3725 else if (parse_one_token(&arg, "context"))
3726 val |= WSEH_CONTEXT;
3727 else {
3728 error("unknown value after ws-error-highlight=%.*s",
3729 (int)(arg - orig_arg), orig_arg);
3730 return 0;
3731 }
3732 if (*arg)
3733 arg++;
3734 }
3735 opt->ws_error_highlight = val;
3736 return 1;
3737 }
3738
3739 int diff_opt_parse(struct diff_options *options,
3740 const char **av, int ac, const char *prefix)
3741 {
3742 const char *arg = av[0];
3743 const char *optarg;
3744 int argcount;
3745
3746 if (!prefix)
3747 prefix = "";
3748
3749 /* Output format options */
3750 if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch")
3751 || opt_arg(arg, 'U', "unified", &options->context))
3752 enable_patch_output(&options->output_format);
3753 else if (!strcmp(arg, "--raw"))
3754 options->output_format |= DIFF_FORMAT_RAW;
3755 else if (!strcmp(arg, "--patch-with-raw")) {
3756 enable_patch_output(&options->output_format);
3757 options->output_format |= DIFF_FORMAT_RAW;
3758 } else if (!strcmp(arg, "--numstat"))
3759 options->output_format |= DIFF_FORMAT_NUMSTAT;
3760 else if (!strcmp(arg, "--shortstat"))
3761 options->output_format |= DIFF_FORMAT_SHORTSTAT;
3762 else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3763 return parse_dirstat_opt(options, "");
3764 else if (skip_prefix(arg, "-X", &arg))
3765 return parse_dirstat_opt(options, arg);
3766 else if (skip_prefix(arg, "--dirstat=", &arg))
3767 return parse_dirstat_opt(options, arg);
3768 else if (!strcmp(arg, "--cumulative"))
3769 return parse_dirstat_opt(options, "cumulative");
3770 else if (!strcmp(arg, "--dirstat-by-file"))
3771 return parse_dirstat_opt(options, "files");
3772 else if (skip_prefix(arg, "--dirstat-by-file=", &arg)) {
3773 parse_dirstat_opt(options, "files");
3774 return parse_dirstat_opt(options, arg);
3775 }
3776 else if (!strcmp(arg, "--check"))
3777 options->output_format |= DIFF_FORMAT_CHECKDIFF;
3778 else if (!strcmp(arg, "--summary"))
3779 options->output_format |= DIFF_FORMAT_SUMMARY;
3780 else if (!strcmp(arg, "--patch-with-stat")) {
3781 enable_patch_output(&options->output_format);
3782 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3783 } else if (!strcmp(arg, "--name-only"))
3784 options->output_format |= DIFF_FORMAT_NAME;
3785 else if (!strcmp(arg, "--name-status"))
3786 options->output_format |= DIFF_FORMAT_NAME_STATUS;
3787 else if (!strcmp(arg, "-s") || !strcmp(arg, "--no-patch"))
3788 options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3789 else if (starts_with(arg, "--stat"))
3790 /* --stat, --stat-width, --stat-name-width, or --stat-count */
3791 return stat_opt(options, av);
3792
3793 /* renames options */
3794 else if (starts_with(arg, "-B") || starts_with(arg, "--break-rewrites=") ||
3795 !strcmp(arg, "--break-rewrites")) {
3796 if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3797 return error("invalid argument to -B: %s", arg+2);
3798 }
3799 else if (starts_with(arg, "-M") || starts_with(arg, "--find-renames=") ||
3800 !strcmp(arg, "--find-renames")) {
3801 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3802 return error("invalid argument to -M: %s", arg+2);
3803 options->detect_rename = DIFF_DETECT_RENAME;
3804 }
3805 else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3806 options->irreversible_delete = 1;
3807 }
3808 else if (starts_with(arg, "-C") || starts_with(arg, "--find-copies=") ||
3809 !strcmp(arg, "--find-copies")) {
3810 if (options->detect_rename == DIFF_DETECT_COPY)
3811 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3812 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3813 return error("invalid argument to -C: %s", arg+2);
3814 options->detect_rename = DIFF_DETECT_COPY;
3815 }
3816 else if (!strcmp(arg, "--no-renames"))
3817 options->detect_rename = 0;
3818 else if (!strcmp(arg, "--rename-empty"))
3819 DIFF_OPT_SET(options, RENAME_EMPTY);
3820 else if (!strcmp(arg, "--no-rename-empty"))
3821 DIFF_OPT_CLR(options, RENAME_EMPTY);
3822 else if (!strcmp(arg, "--relative"))
3823 DIFF_OPT_SET(options, RELATIVE_NAME);
3824 else if (skip_prefix(arg, "--relative=", &arg)) {
3825 DIFF_OPT_SET(options, RELATIVE_NAME);
3826 options->prefix = arg;
3827 }
3828
3829 /* xdiff options */
3830 else if (!strcmp(arg, "--minimal"))
3831 DIFF_XDL_SET(options, NEED_MINIMAL);
3832 else if (!strcmp(arg, "--no-minimal"))
3833 DIFF_XDL_CLR(options, NEED_MINIMAL);
3834 else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3835 DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3836 else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3837 DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3838 else if (!strcmp(arg, "--ignore-space-at-eol"))
3839 DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3840 else if (!strcmp(arg, "--ignore-blank-lines"))
3841 DIFF_XDL_SET(options, IGNORE_BLANK_LINES);
3842 else if (!strcmp(arg, "--indent-heuristic")) {
3843 DIFF_XDL_SET(options, INDENT_HEURISTIC);
3844 DIFF_XDL_CLR(options, COMPACTION_HEURISTIC);
3845 } else if (!strcmp(arg, "--no-indent-heuristic"))
3846 DIFF_XDL_CLR(options, INDENT_HEURISTIC);
3847 else if (!strcmp(arg, "--compaction-heuristic")) {
3848 DIFF_XDL_SET(options, COMPACTION_HEURISTIC);
3849 DIFF_XDL_CLR(options, INDENT_HEURISTIC);
3850 } else if (!strcmp(arg, "--no-compaction-heuristic"))
3851 DIFF_XDL_CLR(options, COMPACTION_HEURISTIC);
3852 else if (!strcmp(arg, "--patience"))
3853 options->xdl_opts = DIFF_WITH_ALG(options, PATIENCE_DIFF);
3854 else if (!strcmp(arg, "--histogram"))
3855 options->xdl_opts = DIFF_WITH_ALG(options, HISTOGRAM_DIFF);
3856 else if ((argcount = parse_long_opt("diff-algorithm", av, &optarg))) {
3857 long value = parse_algorithm_value(optarg);
3858 if (value < 0)
3859 return error("option diff-algorithm accepts \"myers\", "
3860 "\"minimal\", \"patience\" and \"histogram\"");
3861 /* clear out previous settings */
3862 DIFF_XDL_CLR(options, NEED_MINIMAL);
3863 options->xdl_opts &= ~XDF_DIFF_ALGORITHM_MASK;
3864 options->xdl_opts |= value;
3865 return argcount;
3866 }
3867
3868 /* flags options */
3869 else if (!strcmp(arg, "--binary")) {
3870 enable_patch_output(&options->output_format);
3871 DIFF_OPT_SET(options, BINARY);
3872 }
3873 else if (!strcmp(arg, "--full-index"))
3874 DIFF_OPT_SET(options, FULL_INDEX);
3875 else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
3876 DIFF_OPT_SET(options, TEXT);
3877 else if (!strcmp(arg, "-R"))
3878 DIFF_OPT_SET(options, REVERSE_DIFF);
3879 else if (!strcmp(arg, "--find-copies-harder"))
3880 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3881 else if (!strcmp(arg, "--follow"))
3882 DIFF_OPT_SET(options, FOLLOW_RENAMES);
3883 else if (!strcmp(arg, "--no-follow")) {
3884 DIFF_OPT_CLR(options, FOLLOW_RENAMES);
3885 DIFF_OPT_CLR(options, DEFAULT_FOLLOW_RENAMES);
3886 } else if (!strcmp(arg, "--color"))
3887 options->use_color = 1;
3888 else if (skip_prefix(arg, "--color=", &arg)) {
3889 int value = git_config_colorbool(NULL, arg);
3890 if (value < 0)
3891 return error("option `color' expects \"always\", \"auto\", or \"never\"");
3892 options->use_color = value;
3893 }
3894 else if (!strcmp(arg, "--no-color"))
3895 options->use_color = 0;
3896 else if (!strcmp(arg, "--color-words")) {
3897 options->use_color = 1;
3898 options->word_diff = DIFF_WORDS_COLOR;
3899 }
3900 else if (skip_prefix(arg, "--color-words=", &arg)) {
3901 options->use_color = 1;
3902 options->word_diff = DIFF_WORDS_COLOR;
3903 options->word_regex = arg;
3904 }
3905 else if (!strcmp(arg, "--word-diff")) {
3906 if (options->word_diff == DIFF_WORDS_NONE)
3907 options->word_diff = DIFF_WORDS_PLAIN;
3908 }
3909 else if (skip_prefix(arg, "--word-diff=", &arg)) {
3910 if (!strcmp(arg, "plain"))
3911 options->word_diff = DIFF_WORDS_PLAIN;
3912 else if (!strcmp(arg, "color")) {
3913 options->use_color = 1;
3914 options->word_diff = DIFF_WORDS_COLOR;
3915 }
3916 else if (!strcmp(arg, "porcelain"))
3917 options->word_diff = DIFF_WORDS_PORCELAIN;
3918 else if (!strcmp(arg, "none"))
3919 options->word_diff = DIFF_WORDS_NONE;
3920 else
3921 die("bad --word-diff argument: %s", arg);
3922 }
3923 else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
3924 if (options->word_diff == DIFF_WORDS_NONE)
3925 options->word_diff = DIFF_WORDS_PLAIN;
3926 options->word_regex = optarg;
3927 return argcount;
3928 }
3929 else if (!strcmp(arg, "--exit-code"))
3930 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3931 else if (!strcmp(arg, "--quiet"))
3932 DIFF_OPT_SET(options, QUICK);
3933 else if (!strcmp(arg, "--ext-diff"))
3934 DIFF_OPT_SET(options, ALLOW_EXTERNAL);
3935 else if (!strcmp(arg, "--no-ext-diff"))
3936 DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
3937 else if (!strcmp(arg, "--textconv"))
3938 DIFF_OPT_SET(options, ALLOW_TEXTCONV);
3939 else if (!strcmp(arg, "--no-textconv"))
3940 DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
3941 else if (!strcmp(arg, "--ignore-submodules")) {
3942 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3943 handle_ignore_submodules_arg(options, "all");
3944 } else if (skip_prefix(arg, "--ignore-submodules=", &arg)) {
3945 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3946 handle_ignore_submodules_arg(options, arg);
3947 } else if (!strcmp(arg, "--submodule"))
3948 options->submodule_format = DIFF_SUBMODULE_LOG;
3949 else if (skip_prefix(arg, "--submodule=", &arg))
3950 return parse_submodule_opt(options, arg);
3951 else if (skip_prefix(arg, "--ws-error-highlight=", &arg))
3952 return parse_ws_error_highlight(options, arg);
3953
3954 /* misc options */
3955 else if (!strcmp(arg, "-z"))
3956 options->line_termination = 0;
3957 else if ((argcount = short_opt('l', av, &optarg))) {
3958 options->rename_limit = strtoul(optarg, NULL, 10);
3959 return argcount;
3960 }
3961 else if ((argcount = short_opt('S', av, &optarg))) {
3962 options->pickaxe = optarg;
3963 options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
3964 return argcount;
3965 } else if ((argcount = short_opt('G', av, &optarg))) {
3966 options->pickaxe = optarg;
3967 options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
3968 return argcount;
3969 }
3970 else if (!strcmp(arg, "--pickaxe-all"))
3971 options->pickaxe_opts |= DIFF_PICKAXE_ALL;
3972 else if (!strcmp(arg, "--pickaxe-regex"))
3973 options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
3974 else if ((argcount = short_opt('O', av, &optarg))) {
3975 const char *path = prefix_filename(prefix, strlen(prefix), optarg);
3976 options->orderfile = xstrdup(path);
3977 return argcount;
3978 }
3979 else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
3980 int offending = parse_diff_filter_opt(optarg, options);
3981 if (offending)
3982 die("unknown change class '%c' in --diff-filter=%s",
3983 offending, optarg);
3984 return argcount;
3985 }
3986 else if (!strcmp(arg, "--abbrev"))
3987 options->abbrev = DEFAULT_ABBREV;
3988 else if (skip_prefix(arg, "--abbrev=", &arg)) {
3989 options->abbrev = strtoul(arg, NULL, 10);
3990 if (options->abbrev < MINIMUM_ABBREV)
3991 options->abbrev = MINIMUM_ABBREV;
3992 else if (40 < options->abbrev)
3993 options->abbrev = 40;
3994 }
3995 else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
3996 options->a_prefix = optarg;
3997 return argcount;
3998 }
3999 else if ((argcount = parse_long_opt("line-prefix", av, &optarg))) {
4000 options->line_prefix = optarg;
4001 options->line_prefix_length = strlen(options->line_prefix);
4002 graph_setup_line_prefix(options);
4003 return argcount;
4004 }
4005 else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
4006 options->b_prefix = optarg;
4007 return argcount;
4008 }
4009 else if (!strcmp(arg, "--no-prefix"))
4010 options->a_prefix = options->b_prefix = "";
4011 else if (opt_arg(arg, '\0', "inter-hunk-context",
4012 &options->interhunkcontext))
4013 ;
4014 else if (!strcmp(arg, "-W"))
4015 DIFF_OPT_SET(options, FUNCCONTEXT);
4016 else if (!strcmp(arg, "--function-context"))
4017 DIFF_OPT_SET(options, FUNCCONTEXT);
4018 else if (!strcmp(arg, "--no-function-context"))
4019 DIFF_OPT_CLR(options, FUNCCONTEXT);
4020 else if ((argcount = parse_long_opt("output", av, &optarg))) {
4021 const char *path = prefix_filename(prefix, strlen(prefix), optarg);
4022 options->file = fopen(path, "w");
4023 if (!options->file)
4024 die_errno("Could not open '%s'", path);
4025 options->close_file = 1;
4026 if (options->use_color != GIT_COLOR_ALWAYS)
4027 options->use_color = GIT_COLOR_NEVER;
4028 return argcount;
4029 } else
4030 return 0;
4031 return 1;
4032 }
4033
4034 int parse_rename_score(const char **cp_p)
4035 {
4036 unsigned long num, scale;
4037 int ch, dot;
4038 const char *cp = *cp_p;
4039
4040 num = 0;
4041 scale = 1;
4042 dot = 0;
4043 for (;;) {
4044 ch = *cp;
4045 if ( !dot && ch == '.' ) {
4046 scale = 1;
4047 dot = 1;
4048 } else if ( ch == '%' ) {
4049 scale = dot ? scale*100 : 100;
4050 cp++; /* % is always at the end */
4051 break;
4052 } else if ( ch >= '0' && ch <= '9' ) {
4053 if ( scale < 100000 ) {
4054 scale *= 10;
4055 num = (num*10) + (ch-'0');
4056 }
4057 } else {
4058 break;
4059 }
4060 cp++;
4061 }
4062 *cp_p = cp;
4063
4064 /* user says num divided by scale and we say internally that
4065 * is MAX_SCORE * num / scale.
4066 */
4067 return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
4068 }
4069
4070 static int diff_scoreopt_parse(const char *opt)
4071 {
4072 int opt1, opt2, cmd;
4073
4074 if (*opt++ != '-')
4075 return -1;
4076 cmd = *opt++;
4077 if (cmd == '-') {
4078 /* convert the long-form arguments into short-form versions */
4079 if (skip_prefix(opt, "break-rewrites", &opt)) {
4080 if (*opt == 0 || *opt++ == '=')
4081 cmd = 'B';
4082 } else if (skip_prefix(opt, "find-copies", &opt)) {
4083 if (*opt == 0 || *opt++ == '=')
4084 cmd = 'C';
4085 } else if (skip_prefix(opt, "find-renames", &opt)) {
4086 if (*opt == 0 || *opt++ == '=')
4087 cmd = 'M';
4088 }
4089 }
4090 if (cmd != 'M' && cmd != 'C' && cmd != 'B')
4091 return -1; /* that is not a -M, -C, or -B option */
4092
4093 opt1 = parse_rename_score(&opt);
4094 if (cmd != 'B')
4095 opt2 = 0;
4096 else {
4097 if (*opt == 0)
4098 opt2 = 0;
4099 else if (*opt != '/')
4100 return -1; /* we expect -B80/99 or -B80 */
4101 else {
4102 opt++;
4103 opt2 = parse_rename_score(&opt);
4104 }
4105 }
4106 if (*opt != 0)
4107 return -1;
4108 return opt1 | (opt2 << 16);
4109 }
4110
4111 struct diff_queue_struct diff_queued_diff;
4112
4113 void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
4114 {
4115 ALLOC_GROW(queue->queue, queue->nr + 1, queue->alloc);
4116 queue->queue[queue->nr++] = dp;
4117 }
4118
4119 struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
4120 struct diff_filespec *one,
4121 struct diff_filespec *two)
4122 {
4123 struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
4124 dp->one = one;
4125 dp->two = two;
4126 if (queue)
4127 diff_q(queue, dp);
4128 return dp;
4129 }
4130
4131 void diff_free_filepair(struct diff_filepair *p)
4132 {
4133 free_filespec(p->one);
4134 free_filespec(p->two);
4135 free(p);
4136 }
4137
4138 /* This is different from find_unique_abbrev() in that
4139 * it stuffs the result with dots for alignment.
4140 */
4141 const char *diff_unique_abbrev(const unsigned char *sha1, int len)
4142 {
4143 int abblen;
4144 const char *abbrev;
4145 if (len == 40)
4146 return sha1_to_hex(sha1);
4147
4148 abbrev = find_unique_abbrev(sha1, len);
4149 abblen = strlen(abbrev);
4150 if (abblen < 37) {
4151 static char hex[41];
4152 if (len < abblen && abblen <= len + 2)
4153 xsnprintf(hex, sizeof(hex), "%s%.*s", abbrev, len+3-abblen, "..");
4154 else
4155 xsnprintf(hex, sizeof(hex), "%s...", abbrev);
4156 return hex;
4157 }
4158 return sha1_to_hex(sha1);
4159 }
4160
4161 static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
4162 {
4163 int line_termination = opt->line_termination;
4164 int inter_name_termination = line_termination ? '\t' : '\0';
4165
4166 fprintf(opt->file, "%s", diff_line_prefix(opt));
4167 if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
4168 fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
4169 diff_unique_abbrev(p->one->oid.hash, opt->abbrev));
4170 fprintf(opt->file, "%s ",
4171 diff_unique_abbrev(p->two->oid.hash, opt->abbrev));
4172 }
4173 if (p->score) {
4174 fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
4175 inter_name_termination);
4176 } else {
4177 fprintf(opt->file, "%c%c", p->status, inter_name_termination);
4178 }
4179
4180 if (p->status == DIFF_STATUS_COPIED ||
4181 p->status == DIFF_STATUS_RENAMED) {
4182 const char *name_a, *name_b;
4183 name_a = p->one->path;
4184 name_b = p->two->path;
4185 strip_prefix(opt->prefix_length, &name_a, &name_b);
4186 write_name_quoted(name_a, opt->file, inter_name_termination);
4187 write_name_quoted(name_b, opt->file, line_termination);
4188 } else {
4189 const char *name_a, *name_b;
4190 name_a = p->one->mode ? p->one->path : p->two->path;
4191 name_b = NULL;
4192 strip_prefix(opt->prefix_length, &name_a, &name_b);
4193 write_name_quoted(name_a, opt->file, line_termination);
4194 }
4195 }
4196
4197 int diff_unmodified_pair(struct diff_filepair *p)
4198 {
4199 /* This function is written stricter than necessary to support
4200 * the currently implemented transformers, but the idea is to
4201 * let transformers to produce diff_filepairs any way they want,
4202 * and filter and clean them up here before producing the output.
4203 */
4204 struct diff_filespec *one = p->one, *two = p->two;
4205
4206 if (DIFF_PAIR_UNMERGED(p))
4207 return 0; /* unmerged is interesting */
4208
4209 /* deletion, addition, mode or type change
4210 * and rename are all interesting.
4211 */
4212 if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
4213 DIFF_PAIR_MODE_CHANGED(p) ||
4214 strcmp(one->path, two->path))
4215 return 0;
4216
4217 /* both are valid and point at the same path. that is, we are
4218 * dealing with a change.
4219 */
4220 if (one->oid_valid && two->oid_valid &&
4221 !oidcmp(&one->oid, &two->oid) &&
4222 !one->dirty_submodule && !two->dirty_submodule)
4223 return 1; /* no change */
4224 if (!one->oid_valid && !two->oid_valid)
4225 return 1; /* both look at the same file on the filesystem. */
4226 return 0;
4227 }
4228
4229 static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
4230 {
4231 if (diff_unmodified_pair(p))
4232 return;
4233
4234 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4235 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4236 return; /* no tree diffs in patch format */
4237
4238 run_diff(p, o);
4239 }
4240
4241 static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
4242 struct diffstat_t *diffstat)
4243 {
4244 if (diff_unmodified_pair(p))
4245 return;
4246
4247 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4248 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4249 return; /* no useful stat for tree diffs */
4250
4251 run_diffstat(p, o, diffstat);
4252 }
4253
4254 static void diff_flush_checkdiff(struct diff_filepair *p,
4255 struct diff_options *o)
4256 {
4257 if (diff_unmodified_pair(p))
4258 return;
4259
4260 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4261 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4262 return; /* nothing to check in tree diffs */
4263
4264 run_checkdiff(p, o);
4265 }
4266
4267 int diff_queue_is_empty(void)
4268 {
4269 struct diff_queue_struct *q = &diff_queued_diff;
4270 int i;
4271 for (i = 0; i < q->nr; i++)
4272 if (!diff_unmodified_pair(q->queue[i]))
4273 return 0;
4274 return 1;
4275 }
4276
4277 #if DIFF_DEBUG
4278 void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
4279 {
4280 fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
4281 x, one ? one : "",
4282 s->path,
4283 DIFF_FILE_VALID(s) ? "valid" : "invalid",
4284 s->mode,
4285 s->oid_valid ? oid_to_hex(&s->oid) : "");
4286 fprintf(stderr, "queue[%d] %s size %lu\n",
4287 x, one ? one : "",
4288 s->size);
4289 }
4290
4291 void diff_debug_filepair(const struct diff_filepair *p, int i)
4292 {
4293 diff_debug_filespec(p->one, i, "one");
4294 diff_debug_filespec(p->two, i, "two");
4295 fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
4296 p->score, p->status ? p->status : '?',
4297 p->one->rename_used, p->broken_pair);
4298 }
4299
4300 void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
4301 {
4302 int i;
4303 if (msg)
4304 fprintf(stderr, "%s\n", msg);
4305 fprintf(stderr, "q->nr = %d\n", q->nr);
4306 for (i = 0; i < q->nr; i++) {
4307 struct diff_filepair *p = q->queue[i];
4308 diff_debug_filepair(p, i);
4309 }
4310 }
4311 #endif
4312
4313 static void diff_resolve_rename_copy(void)
4314 {
4315 int i;
4316 struct diff_filepair *p;
4317 struct diff_queue_struct *q = &diff_queued_diff;
4318
4319 diff_debug_queue("resolve-rename-copy", q);
4320
4321 for (i = 0; i < q->nr; i++) {
4322 p = q->queue[i];
4323 p->status = 0; /* undecided */
4324 if (DIFF_PAIR_UNMERGED(p))
4325 p->status = DIFF_STATUS_UNMERGED;
4326 else if (!DIFF_FILE_VALID(p->one))
4327 p->status = DIFF_STATUS_ADDED;
4328 else if (!DIFF_FILE_VALID(p->two))
4329 p->status = DIFF_STATUS_DELETED;
4330 else if (DIFF_PAIR_TYPE_CHANGED(p))
4331 p->status = DIFF_STATUS_TYPE_CHANGED;
4332
4333 /* from this point on, we are dealing with a pair
4334 * whose both sides are valid and of the same type, i.e.
4335 * either in-place edit or rename/copy edit.
4336 */
4337 else if (DIFF_PAIR_RENAME(p)) {
4338 /*
4339 * A rename might have re-connected a broken
4340 * pair up, causing the pathnames to be the
4341 * same again. If so, that's not a rename at
4342 * all, just a modification..
4343 *
4344 * Otherwise, see if this source was used for
4345 * multiple renames, in which case we decrement
4346 * the count, and call it a copy.
4347 */
4348 if (!strcmp(p->one->path, p->two->path))
4349 p->status = DIFF_STATUS_MODIFIED;
4350 else if (--p->one->rename_used > 0)
4351 p->status = DIFF_STATUS_COPIED;
4352 else
4353 p->status = DIFF_STATUS_RENAMED;
4354 }
4355 else if (oidcmp(&p->one->oid, &p->two->oid) ||
4356 p->one->mode != p->two->mode ||
4357 p->one->dirty_submodule ||
4358 p->two->dirty_submodule ||
4359 is_null_oid(&p->one->oid))
4360 p->status = DIFF_STATUS_MODIFIED;
4361 else {
4362 /* This is a "no-change" entry and should not
4363 * happen anymore, but prepare for broken callers.
4364 */
4365 error("feeding unmodified %s to diffcore",
4366 p->one->path);
4367 p->status = DIFF_STATUS_UNKNOWN;
4368 }
4369 }
4370 diff_debug_queue("resolve-rename-copy done", q);
4371 }
4372
4373 static int check_pair_status(struct diff_filepair *p)
4374 {
4375 switch (p->status) {
4376 case DIFF_STATUS_UNKNOWN:
4377 return 0;
4378 case 0:
4379 die("internal error in diff-resolve-rename-copy");
4380 default:
4381 return 1;
4382 }
4383 }
4384
4385 static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
4386 {
4387 int fmt = opt->output_format;
4388
4389 if (fmt & DIFF_FORMAT_CHECKDIFF)
4390 diff_flush_checkdiff(p, opt);
4391 else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
4392 diff_flush_raw(p, opt);
4393 else if (fmt & DIFF_FORMAT_NAME) {
4394 const char *name_a, *name_b;
4395 name_a = p->two->path;
4396 name_b = NULL;
4397 strip_prefix(opt->prefix_length, &name_a, &name_b);
4398 write_name_quoted(name_a, opt->file, opt->line_termination);
4399 }
4400 }
4401
4402 static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
4403 {
4404 if (fs->mode)
4405 fprintf(file, " %s mode %06o ", newdelete, fs->mode);
4406 else
4407 fprintf(file, " %s ", newdelete);
4408 write_name_quoted(fs->path, file, '\n');
4409 }
4410
4411
4412 static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
4413 const char *line_prefix)
4414 {
4415 if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
4416 fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
4417 p->two->mode, show_name ? ' ' : '\n');
4418 if (show_name) {
4419 write_name_quoted(p->two->path, file, '\n');
4420 }
4421 }
4422 }
4423
4424 static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
4425 const char *line_prefix)
4426 {
4427 char *names = pprint_rename(p->one->path, p->two->path);
4428
4429 fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
4430 free(names);
4431 show_mode_change(file, p, 0, line_prefix);
4432 }
4433
4434 static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
4435 {
4436 FILE *file = opt->file;
4437 const char *line_prefix = diff_line_prefix(opt);
4438
4439 switch(p->status) {
4440 case DIFF_STATUS_DELETED:
4441 fputs(line_prefix, file);
4442 show_file_mode_name(file, "delete", p->one);
4443 break;
4444 case DIFF_STATUS_ADDED:
4445 fputs(line_prefix, file);
4446 show_file_mode_name(file, "create", p->two);
4447 break;
4448 case DIFF_STATUS_COPIED:
4449 fputs(line_prefix, file);
4450 show_rename_copy(file, "copy", p, line_prefix);
4451 break;
4452 case DIFF_STATUS_RENAMED:
4453 fputs(line_prefix, file);
4454 show_rename_copy(file, "rename", p, line_prefix);
4455 break;
4456 default:
4457 if (p->score) {
4458 fprintf(file, "%s rewrite ", line_prefix);
4459 write_name_quoted(p->two->path, file, ' ');
4460 fprintf(file, "(%d%%)\n", similarity_index(p));
4461 }
4462 show_mode_change(file, p, !p->score, line_prefix);
4463 break;
4464 }
4465 }
4466
4467 struct patch_id_t {
4468 git_SHA_CTX *ctx;
4469 int patchlen;
4470 };
4471
4472 static int remove_space(char *line, int len)
4473 {
4474 int i;
4475 char *dst = line;
4476 unsigned char c;
4477
4478 for (i = 0; i < len; i++)
4479 if (!isspace((c = line[i])))
4480 *dst++ = c;
4481
4482 return dst - line;
4483 }
4484
4485 static void patch_id_consume(void *priv, char *line, unsigned long len)
4486 {
4487 struct patch_id_t *data = priv;
4488 int new_len;
4489
4490 /* Ignore line numbers when computing the SHA1 of the patch */
4491 if (starts_with(line, "@@ -"))
4492 return;
4493
4494 new_len = remove_space(line, len);
4495
4496 git_SHA1_Update(data->ctx, line, new_len);
4497 data->patchlen += new_len;
4498 }
4499
4500 /* returns 0 upon success, and writes result into sha1 */
4501 static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1, int diff_header_only)
4502 {
4503 struct diff_queue_struct *q = &diff_queued_diff;
4504 int i;
4505 git_SHA_CTX ctx;
4506 struct patch_id_t data;
4507 char buffer[PATH_MAX * 4 + 20];
4508
4509 git_SHA1_Init(&ctx);
4510 memset(&data, 0, sizeof(struct patch_id_t));
4511 data.ctx = &ctx;
4512
4513 for (i = 0; i < q->nr; i++) {
4514 xpparam_t xpp;
4515 xdemitconf_t xecfg;
4516 mmfile_t mf1, mf2;
4517 struct diff_filepair *p = q->queue[i];
4518 int len1, len2;
4519
4520 memset(&xpp, 0, sizeof(xpp));
4521 memset(&xecfg, 0, sizeof(xecfg));
4522 if (p->status == 0)
4523 return error("internal diff status error");
4524 if (p->status == DIFF_STATUS_UNKNOWN)
4525 continue;
4526 if (diff_unmodified_pair(p))
4527 continue;
4528 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4529 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4530 continue;
4531 if (DIFF_PAIR_UNMERGED(p))
4532 continue;
4533
4534 diff_fill_sha1_info(p->one);
4535 diff_fill_sha1_info(p->two);
4536
4537 len1 = remove_space(p->one->path, strlen(p->one->path));
4538 len2 = remove_space(p->two->path, strlen(p->two->path));
4539 if (p->one->mode == 0)
4540 len1 = snprintf(buffer, sizeof(buffer),
4541 "diff--gita/%.*sb/%.*s"
4542 "newfilemode%06o"
4543 "---/dev/null"
4544 "+++b/%.*s",
4545 len1, p->one->path,
4546 len2, p->two->path,
4547 p->two->mode,
4548 len2, p->two->path);
4549 else if (p->two->mode == 0)
4550 len1 = snprintf(buffer, sizeof(buffer),
4551 "diff--gita/%.*sb/%.*s"
4552 "deletedfilemode%06o"
4553 "---a/%.*s"
4554 "+++/dev/null",
4555 len1, p->one->path,
4556 len2, p->two->path,
4557 p->one->mode,
4558 len1, p->one->path);
4559 else
4560 len1 = snprintf(buffer, sizeof(buffer),
4561 "diff--gita/%.*sb/%.*s"
4562 "---a/%.*s"
4563 "+++b/%.*s",
4564 len1, p->one->path,
4565 len2, p->two->path,
4566 len1, p->one->path,
4567 len2, p->two->path);
4568 git_SHA1_Update(&ctx, buffer, len1);
4569
4570 if (diff_header_only)
4571 continue;
4572
4573 if (fill_mmfile(&mf1, p->one) < 0 ||
4574 fill_mmfile(&mf2, p->two) < 0)
4575 return error("unable to read files to diff");
4576
4577 if (diff_filespec_is_binary(p->one) ||
4578 diff_filespec_is_binary(p->two)) {
4579 git_SHA1_Update(&ctx, oid_to_hex(&p->one->oid),
4580 40);
4581 git_SHA1_Update(&ctx, oid_to_hex(&p->two->oid),
4582 40);
4583 continue;
4584 }
4585
4586 xpp.flags = 0;
4587 xecfg.ctxlen = 3;
4588 xecfg.flags = 0;
4589 if (xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4590 &xpp, &xecfg))
4591 return error("unable to generate patch-id diff for %s",
4592 p->one->path);
4593 }
4594
4595 git_SHA1_Final(sha1, &ctx);
4596 return 0;
4597 }
4598
4599 int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1, int diff_header_only)
4600 {
4601 struct diff_queue_struct *q = &diff_queued_diff;
4602 int i;
4603 int result = diff_get_patch_id(options, sha1, diff_header_only);
4604
4605 for (i = 0; i < q->nr; i++)
4606 diff_free_filepair(q->queue[i]);
4607
4608 free(q->queue);
4609 DIFF_QUEUE_CLEAR(q);
4610
4611 return result;
4612 }
4613
4614 static int is_summary_empty(const struct diff_queue_struct *q)
4615 {
4616 int i;
4617
4618 for (i = 0; i < q->nr; i++) {
4619 const struct diff_filepair *p = q->queue[i];
4620
4621 switch (p->status) {
4622 case DIFF_STATUS_DELETED:
4623 case DIFF_STATUS_ADDED:
4624 case DIFF_STATUS_COPIED:
4625 case DIFF_STATUS_RENAMED:
4626 return 0;
4627 default:
4628 if (p->score)
4629 return 0;
4630 if (p->one->mode && p->two->mode &&
4631 p->one->mode != p->two->mode)
4632 return 0;
4633 break;
4634 }
4635 }
4636 return 1;
4637 }
4638
4639 static const char rename_limit_warning[] =
4640 "inexact rename detection was skipped due to too many files.";
4641
4642 static const char degrade_cc_to_c_warning[] =
4643 "only found copies from modified paths due to too many files.";
4644
4645 static const char rename_limit_advice[] =
4646 "you may want to set your %s variable to at least "
4647 "%d and retry the command.";
4648
4649 void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4650 {
4651 if (degraded_cc)
4652 warning(degrade_cc_to_c_warning);
4653 else if (needed)
4654 warning(rename_limit_warning);
4655 else
4656 return;
4657 if (0 < needed && needed < 32767)
4658 warning(rename_limit_advice, varname, needed);
4659 }
4660
4661 void diff_flush(struct diff_options *options)
4662 {
4663 struct diff_queue_struct *q = &diff_queued_diff;
4664 int i, output_format = options->output_format;
4665 int separator = 0;
4666 int dirstat_by_line = 0;
4667
4668 /*
4669 * Order: raw, stat, summary, patch
4670 * or: name/name-status/checkdiff (other bits clear)
4671 */
4672 if (!q->nr)
4673 goto free_queue;
4674
4675 if (output_format & (DIFF_FORMAT_RAW |
4676 DIFF_FORMAT_NAME |
4677 DIFF_FORMAT_NAME_STATUS |
4678 DIFF_FORMAT_CHECKDIFF)) {
4679 for (i = 0; i < q->nr; i++) {
4680 struct diff_filepair *p = q->queue[i];
4681 if (check_pair_status(p))
4682 flush_one_pair(p, options);
4683 }
4684 separator++;
4685 }
4686
4687 if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4688 dirstat_by_line = 1;
4689
4690 if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4691 dirstat_by_line) {
4692 struct diffstat_t diffstat;
4693
4694 memset(&diffstat, 0, sizeof(struct diffstat_t));
4695 for (i = 0; i < q->nr; i++) {
4696 struct diff_filepair *p = q->queue[i];
4697 if (check_pair_status(p))
4698 diff_flush_stat(p, options, &diffstat);
4699 }
4700 if (output_format & DIFF_FORMAT_NUMSTAT)
4701 show_numstat(&diffstat, options);
4702 if (output_format & DIFF_FORMAT_DIFFSTAT)
4703 show_stats(&diffstat, options);
4704 if (output_format & DIFF_FORMAT_SHORTSTAT)
4705 show_shortstats(&diffstat, options);
4706 if (output_format & DIFF_FORMAT_DIRSTAT && dirstat_by_line)
4707 show_dirstat_by_line(&diffstat, options);
4708 free_diffstat_info(&diffstat);
4709 separator++;
4710 }
4711 if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4712 show_dirstat(options);
4713
4714 if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4715 for (i = 0; i < q->nr; i++) {
4716 diff_summary(options, q->queue[i]);
4717 }
4718 separator++;
4719 }
4720
4721 if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4722 DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4723 DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4724 /*
4725 * run diff_flush_patch for the exit status. setting
4726 * options->file to /dev/null should be safe, because we
4727 * aren't supposed to produce any output anyway.
4728 */
4729 if (options->close_file)
4730 fclose(options->file);
4731 options->file = fopen("/dev/null", "w");
4732 if (!options->file)
4733 die_errno("Could not open /dev/null");
4734 options->close_file = 1;
4735 for (i = 0; i < q->nr; i++) {
4736 struct diff_filepair *p = q->queue[i];
4737 if (check_pair_status(p))
4738 diff_flush_patch(p, options);
4739 if (options->found_changes)
4740 break;
4741 }
4742 }
4743
4744 if (output_format & DIFF_FORMAT_PATCH) {
4745 if (separator) {
4746 fprintf(options->file, "%s%c",
4747 diff_line_prefix(options),
4748 options->line_termination);
4749 if (options->stat_sep) {
4750 /* attach patch instead of inline */
4751 fputs(options->stat_sep, options->file);
4752 }
4753 }
4754
4755 for (i = 0; i < q->nr; i++) {
4756 struct diff_filepair *p = q->queue[i];
4757 if (check_pair_status(p))
4758 diff_flush_patch(p, options);
4759 }
4760 }
4761
4762 if (output_format & DIFF_FORMAT_CALLBACK)
4763 options->format_callback(q, options, options->format_callback_data);
4764
4765 for (i = 0; i < q->nr; i++)
4766 diff_free_filepair(q->queue[i]);
4767 free_queue:
4768 free(q->queue);
4769 DIFF_QUEUE_CLEAR(q);
4770 if (options->close_file)
4771 fclose(options->file);
4772
4773 /*
4774 * Report the content-level differences with HAS_CHANGES;
4775 * diff_addremove/diff_change does not set the bit when
4776 * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4777 */
4778 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4779 if (options->found_changes)
4780 DIFF_OPT_SET(options, HAS_CHANGES);
4781 else
4782 DIFF_OPT_CLR(options, HAS_CHANGES);
4783 }
4784 }
4785
4786 static int match_filter(const struct diff_options *options, const struct diff_filepair *p)
4787 {
4788 return (((p->status == DIFF_STATUS_MODIFIED) &&
4789 ((p->score &&
4790 filter_bit_tst(DIFF_STATUS_FILTER_BROKEN, options)) ||
4791 (!p->score &&
4792 filter_bit_tst(DIFF_STATUS_MODIFIED, options)))) ||
4793 ((p->status != DIFF_STATUS_MODIFIED) &&
4794 filter_bit_tst(p->status, options)));
4795 }
4796
4797 static void diffcore_apply_filter(struct diff_options *options)
4798 {
4799 int i;
4800 struct diff_queue_struct *q = &diff_queued_diff;
4801 struct diff_queue_struct outq;
4802
4803 DIFF_QUEUE_CLEAR(&outq);
4804
4805 if (!options->filter)
4806 return;
4807
4808 if (filter_bit_tst(DIFF_STATUS_FILTER_AON, options)) {
4809 int found;
4810 for (i = found = 0; !found && i < q->nr; i++) {
4811 if (match_filter(options, q->queue[i]))
4812 found++;
4813 }
4814 if (found)
4815 return;
4816
4817 /* otherwise we will clear the whole queue
4818 * by copying the empty outq at the end of this
4819 * function, but first clear the current entries
4820 * in the queue.
4821 */
4822 for (i = 0; i < q->nr; i++)
4823 diff_free_filepair(q->queue[i]);
4824 }
4825 else {
4826 /* Only the matching ones */
4827 for (i = 0; i < q->nr; i++) {
4828 struct diff_filepair *p = q->queue[i];
4829 if (match_filter(options, p))
4830 diff_q(&outq, p);
4831 else
4832 diff_free_filepair(p);
4833 }
4834 }
4835 free(q->queue);
4836 *q = outq;
4837 }
4838
4839 /* Check whether two filespecs with the same mode and size are identical */
4840 static int diff_filespec_is_identical(struct diff_filespec *one,
4841 struct diff_filespec *two)
4842 {
4843 if (S_ISGITLINK(one->mode))
4844 return 0;
4845 if (diff_populate_filespec(one, 0))
4846 return 0;
4847 if (diff_populate_filespec(two, 0))
4848 return 0;
4849 return !memcmp(one->data, two->data, one->size);
4850 }
4851
4852 static int diff_filespec_check_stat_unmatch(struct diff_filepair *p)
4853 {
4854 if (p->done_skip_stat_unmatch)
4855 return p->skip_stat_unmatch_result;
4856
4857 p->done_skip_stat_unmatch = 1;
4858 p->skip_stat_unmatch_result = 0;
4859 /*
4860 * 1. Entries that come from stat info dirtiness
4861 * always have both sides (iow, not create/delete),
4862 * one side of the object name is unknown, with
4863 * the same mode and size. Keep the ones that
4864 * do not match these criteria. They have real
4865 * differences.
4866 *
4867 * 2. At this point, the file is known to be modified,
4868 * with the same mode and size, and the object
4869 * name of one side is unknown. Need to inspect
4870 * the identical contents.
4871 */
4872 if (!DIFF_FILE_VALID(p->one) || /* (1) */
4873 !DIFF_FILE_VALID(p->two) ||
4874 (p->one->oid_valid && p->two->oid_valid) ||
4875 (p->one->mode != p->two->mode) ||
4876 diff_populate_filespec(p->one, CHECK_SIZE_ONLY) ||
4877 diff_populate_filespec(p->two, CHECK_SIZE_ONLY) ||
4878 (p->one->size != p->two->size) ||
4879 !diff_filespec_is_identical(p->one, p->two)) /* (2) */
4880 p->skip_stat_unmatch_result = 1;
4881 return p->skip_stat_unmatch_result;
4882 }
4883
4884 static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
4885 {
4886 int i;
4887 struct diff_queue_struct *q = &diff_queued_diff;
4888 struct diff_queue_struct outq;
4889 DIFF_QUEUE_CLEAR(&outq);
4890
4891 for (i = 0; i < q->nr; i++) {
4892 struct diff_filepair *p = q->queue[i];
4893
4894 if (diff_filespec_check_stat_unmatch(p))
4895 diff_q(&outq, p);
4896 else {
4897 /*
4898 * The caller can subtract 1 from skip_stat_unmatch
4899 * to determine how many paths were dirty only
4900 * due to stat info mismatch.
4901 */
4902 if (!DIFF_OPT_TST(diffopt, NO_INDEX))
4903 diffopt->skip_stat_unmatch++;
4904 diff_free_filepair(p);
4905 }
4906 }
4907 free(q->queue);
4908 *q = outq;
4909 }
4910
4911 static int diffnamecmp(const void *a_, const void *b_)
4912 {
4913 const struct diff_filepair *a = *((const struct diff_filepair **)a_);
4914 const struct diff_filepair *b = *((const struct diff_filepair **)b_);
4915 const char *name_a, *name_b;
4916
4917 name_a = a->one ? a->one->path : a->two->path;
4918 name_b = b->one ? b->one->path : b->two->path;
4919 return strcmp(name_a, name_b);
4920 }
4921
4922 void diffcore_fix_diff_index(struct diff_options *options)
4923 {
4924 struct diff_queue_struct *q = &diff_queued_diff;
4925 qsort(q->queue, q->nr, sizeof(q->queue[0]), diffnamecmp);
4926 }
4927
4928 void diffcore_std(struct diff_options *options)
4929 {
4930 /* NOTE please keep the following in sync with diff_tree_combined() */
4931 if (options->skip_stat_unmatch)
4932 diffcore_skip_stat_unmatch(options);
4933 if (!options->found_follow) {
4934 /* See try_to_follow_renames() in tree-diff.c */
4935 if (options->break_opt != -1)
4936 diffcore_break(options->break_opt);
4937 if (options->detect_rename)
4938 diffcore_rename(options);
4939 if (options->break_opt != -1)
4940 diffcore_merge_broken();
4941 }
4942 if (options->pickaxe)
4943 diffcore_pickaxe(options);
4944 if (options->orderfile)
4945 diffcore_order(options->orderfile);
4946 if (!options->found_follow)
4947 /* See try_to_follow_renames() in tree-diff.c */
4948 diff_resolve_rename_copy();
4949 diffcore_apply_filter(options);
4950
4951 if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4952 DIFF_OPT_SET(options, HAS_CHANGES);
4953 else
4954 DIFF_OPT_CLR(options, HAS_CHANGES);
4955
4956 options->found_follow = 0;
4957 }
4958
4959 int diff_result_code(struct diff_options *opt, int status)
4960 {
4961 int result = 0;
4962
4963 diff_warn_rename_limit("diff.renameLimit",
4964 opt->needed_rename_limit,
4965 opt->degraded_cc_to_c);
4966 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4967 !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
4968 return status;
4969 if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4970 DIFF_OPT_TST(opt, HAS_CHANGES))
4971 result |= 01;
4972 if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
4973 DIFF_OPT_TST(opt, CHECK_FAILED))
4974 result |= 02;
4975 return result;
4976 }
4977
4978 int diff_can_quit_early(struct diff_options *opt)
4979 {
4980 return (DIFF_OPT_TST(opt, QUICK) &&
4981 !opt->filter &&
4982 DIFF_OPT_TST(opt, HAS_CHANGES));
4983 }
4984
4985 /*
4986 * Shall changes to this submodule be ignored?
4987 *
4988 * Submodule changes can be configured to be ignored separately for each path,
4989 * but that configuration can be overridden from the command line.
4990 */
4991 static int is_submodule_ignored(const char *path, struct diff_options *options)
4992 {
4993 int ignored = 0;
4994 unsigned orig_flags = options->flags;
4995 if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
4996 set_diffopt_flags_from_submodule_config(options, path);
4997 if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
4998 ignored = 1;
4999 options->flags = orig_flags;
5000 return ignored;
5001 }
5002
5003 void diff_addremove(struct diff_options *options,
5004 int addremove, unsigned mode,
5005 const unsigned char *sha1,
5006 int sha1_valid,
5007 const char *concatpath, unsigned dirty_submodule)
5008 {
5009 struct diff_filespec *one, *two;
5010
5011 if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
5012 return;
5013
5014 /* This may look odd, but it is a preparation for
5015 * feeding "there are unchanged files which should
5016 * not produce diffs, but when you are doing copy
5017 * detection you would need them, so here they are"
5018 * entries to the diff-core. They will be prefixed
5019 * with something like '=' or '*' (I haven't decided
5020 * which but should not make any difference).
5021 * Feeding the same new and old to diff_change()
5022 * also has the same effect.
5023 * Before the final output happens, they are pruned after
5024 * merged into rename/copy pairs as appropriate.
5025 */
5026 if (DIFF_OPT_TST(options, REVERSE_DIFF))
5027 addremove = (addremove == '+' ? '-' :
5028 addremove == '-' ? '+' : addremove);
5029
5030 if (options->prefix &&
5031 strncmp(concatpath, options->prefix, options->prefix_length))
5032 return;
5033
5034 one = alloc_filespec(concatpath);
5035 two = alloc_filespec(concatpath);
5036
5037 if (addremove != '+')
5038 fill_filespec(one, sha1, sha1_valid, mode);
5039 if (addremove != '-') {
5040 fill_filespec(two, sha1, sha1_valid, mode);
5041 two->dirty_submodule = dirty_submodule;
5042 }
5043
5044 diff_queue(&diff_queued_diff, one, two);
5045 if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5046 DIFF_OPT_SET(options, HAS_CHANGES);
5047 }
5048
5049 void diff_change(struct diff_options *options,
5050 unsigned old_mode, unsigned new_mode,
5051 const unsigned char *old_sha1,
5052 const unsigned char *new_sha1,
5053 int old_sha1_valid, int new_sha1_valid,
5054 const char *concatpath,
5055 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
5056 {
5057 struct diff_filespec *one, *two;
5058 struct diff_filepair *p;
5059
5060 if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
5061 is_submodule_ignored(concatpath, options))
5062 return;
5063
5064 if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
5065 unsigned tmp;
5066 const unsigned char *tmp_c;
5067 tmp = old_mode; old_mode = new_mode; new_mode = tmp;
5068 tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
5069 tmp = old_sha1_valid; old_sha1_valid = new_sha1_valid;
5070 new_sha1_valid = tmp;
5071 tmp = old_dirty_submodule; old_dirty_submodule = new_dirty_submodule;
5072 new_dirty_submodule = tmp;
5073 }
5074
5075 if (options->prefix &&
5076 strncmp(concatpath, options->prefix, options->prefix_length))
5077 return;
5078
5079 one = alloc_filespec(concatpath);
5080 two = alloc_filespec(concatpath);
5081 fill_filespec(one, old_sha1, old_sha1_valid, old_mode);
5082 fill_filespec(two, new_sha1, new_sha1_valid, new_mode);
5083 one->dirty_submodule = old_dirty_submodule;
5084 two->dirty_submodule = new_dirty_submodule;
5085 p = diff_queue(&diff_queued_diff, one, two);
5086
5087 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5088 return;
5089
5090 if (DIFF_OPT_TST(options, QUICK) && options->skip_stat_unmatch &&
5091 !diff_filespec_check_stat_unmatch(p))
5092 return;
5093
5094 DIFF_OPT_SET(options, HAS_CHANGES);
5095 }
5096
5097 struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
5098 {
5099 struct diff_filepair *pair;
5100 struct diff_filespec *one, *two;
5101
5102 if (options->prefix &&
5103 strncmp(path, options->prefix, options->prefix_length))
5104 return NULL;
5105
5106 one = alloc_filespec(path);
5107 two = alloc_filespec(path);
5108 pair = diff_queue(&diff_queued_diff, one, two);
5109 pair->is_unmerged = 1;
5110 return pair;
5111 }
5112
5113 static char *run_textconv(const char *pgm, struct diff_filespec *spec,
5114 size_t *outsize)
5115 {
5116 struct diff_tempfile *temp;
5117 const char *argv[3];
5118 const char **arg = argv;
5119 struct child_process child = CHILD_PROCESS_INIT;
5120 struct strbuf buf = STRBUF_INIT;
5121 int err = 0;
5122
5123 temp = prepare_temp_file(spec->path, spec);
5124 *arg++ = pgm;
5125 *arg++ = temp->name;
5126 *arg = NULL;
5127
5128 child.use_shell = 1;
5129 child.argv = argv;
5130 child.out = -1;
5131 if (start_command(&child)) {
5132 remove_tempfile();
5133 return NULL;
5134 }
5135
5136 if (strbuf_read(&buf, child.out, 0) < 0)
5137 err = error("error reading from textconv command '%s'", pgm);
5138 close(child.out);
5139
5140 if (finish_command(&child) || err) {
5141 strbuf_release(&buf);
5142 remove_tempfile();
5143 return NULL;
5144 }
5145 remove_tempfile();
5146
5147 return strbuf_detach(&buf, outsize);
5148 }
5149
5150 size_t fill_textconv(struct userdiff_driver *driver,
5151 struct diff_filespec *df,
5152 char **outbuf)
5153 {
5154 size_t size;
5155
5156 if (!driver) {
5157 if (!DIFF_FILE_VALID(df)) {
5158 *outbuf = "";
5159 return 0;
5160 }
5161 if (diff_populate_filespec(df, 0))
5162 die("unable to read files to diff");
5163 *outbuf = df->data;
5164 return df->size;
5165 }
5166
5167 if (!driver->textconv)
5168 die("BUG: fill_textconv called with non-textconv driver");
5169
5170 if (driver->textconv_cache && df->oid_valid) {
5171 *outbuf = notes_cache_get(driver->textconv_cache,
5172 df->oid.hash,
5173 &size);
5174 if (*outbuf)
5175 return size;
5176 }
5177
5178 *outbuf = run_textconv(driver->textconv, df, &size);
5179 if (!*outbuf)
5180 die("unable to read files to diff");
5181
5182 if (driver->textconv_cache && df->oid_valid) {
5183 /* ignore errors, as we might be in a readonly repository */
5184 notes_cache_put(driver->textconv_cache, df->oid.hash, *outbuf,
5185 size);
5186 /*
5187 * we could save up changes and flush them all at the end,
5188 * but we would need an extra call after all diffing is done.
5189 * Since generating a cache entry is the slow path anyway,
5190 * this extra overhead probably isn't a big deal.
5191 */
5192 notes_cache_write(driver->textconv_cache);
5193 }
5194
5195 return size;
5196 }
5197
5198 void setup_diff_pager(struct diff_options *opt)
5199 {
5200 /*
5201 * If the user asked for our exit code, then either they want --quiet
5202 * or --exit-code. We should definitely not bother with a pager in the
5203 * former case, as we will generate no output. Since we still properly
5204 * report our exit code even when a pager is run, we _could_ run a
5205 * pager with --exit-code. But since we have not done so historically,
5206 * and because it is easy to find people oneline advising "git diff
5207 * --exit-code" in hooks and other scripts, we do not do so.
5208 */
5209 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5210 check_pager_config("diff") != 0)
5211 setup_pager();
5212 }