Merge branch 'mh/tempfile'
[git/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 "sigchain.h"
17 #include "submodule.h"
18 #include "ll-merge.h"
19 #include "string-list.h"
20 #include "argv-array.h"
21
22 #ifdef NO_FAST_WORKING_DIRECTORY
23 #define FAST_WORKING_DIRECTORY 0
24 #else
25 #define FAST_WORKING_DIRECTORY 1
26 #endif
27
28 static int diff_detect_rename_default;
29 static int diff_rename_limit_default = 400;
30 static int diff_suppress_blank_empty;
31 static int diff_use_color_default = -1;
32 static int diff_context_default = 3;
33 static const char *diff_word_regex_cfg;
34 static const char *external_diff_cmd_cfg;
35 static const char *diff_order_file_cfg;
36 int diff_auto_refresh_index = 1;
37 static int diff_mnemonic_prefix;
38 static int diff_no_prefix;
39 static int diff_stat_graph_width;
40 static int diff_dirstat_permille_default = 30;
41 static struct diff_options default_diff_options;
42 static long diff_algorithm;
43
44 static char diff_colors[][COLOR_MAXLEN] = {
45 GIT_COLOR_RESET,
46 GIT_COLOR_NORMAL, /* CONTEXT */
47 GIT_COLOR_BOLD, /* METAINFO */
48 GIT_COLOR_CYAN, /* FRAGINFO */
49 GIT_COLOR_RED, /* OLD */
50 GIT_COLOR_GREEN, /* NEW */
51 GIT_COLOR_YELLOW, /* COMMIT */
52 GIT_COLOR_BG_RED, /* WHITESPACE */
53 GIT_COLOR_NORMAL, /* FUNCINFO */
54 };
55
56 static int parse_diff_color_slot(const char *var)
57 {
58 if (!strcasecmp(var, "context") || !strcasecmp(var, "plain"))
59 return DIFF_CONTEXT;
60 if (!strcasecmp(var, "meta"))
61 return DIFF_METAINFO;
62 if (!strcasecmp(var, "frag"))
63 return DIFF_FRAGINFO;
64 if (!strcasecmp(var, "old"))
65 return DIFF_FILE_OLD;
66 if (!strcasecmp(var, "new"))
67 return DIFF_FILE_NEW;
68 if (!strcasecmp(var, "commit"))
69 return DIFF_COMMIT;
70 if (!strcasecmp(var, "whitespace"))
71 return DIFF_WHITESPACE;
72 if (!strcasecmp(var, "func"))
73 return DIFF_FUNCINFO;
74 return -1;
75 }
76
77 static int parse_dirstat_params(struct diff_options *options, const char *params_string,
78 struct strbuf *errmsg)
79 {
80 char *params_copy = xstrdup(params_string);
81 struct string_list params = STRING_LIST_INIT_NODUP;
82 int ret = 0;
83 int i;
84
85 if (*params_copy)
86 string_list_split_in_place(&params, params_copy, ',', -1);
87 for (i = 0; i < params.nr; i++) {
88 const char *p = params.items[i].string;
89 if (!strcmp(p, "changes")) {
90 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
91 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
92 } else if (!strcmp(p, "lines")) {
93 DIFF_OPT_SET(options, DIRSTAT_BY_LINE);
94 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
95 } else if (!strcmp(p, "files")) {
96 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
97 DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
98 } else if (!strcmp(p, "noncumulative")) {
99 DIFF_OPT_CLR(options, DIRSTAT_CUMULATIVE);
100 } else if (!strcmp(p, "cumulative")) {
101 DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
102 } else if (isdigit(*p)) {
103 char *end;
104 int permille = strtoul(p, &end, 10) * 10;
105 if (*end == '.' && isdigit(*++end)) {
106 /* only use first digit */
107 permille += *end - '0';
108 /* .. and ignore any further digits */
109 while (isdigit(*++end))
110 ; /* nothing */
111 }
112 if (!*end)
113 options->dirstat_permille = permille;
114 else {
115 strbuf_addf(errmsg, _(" Failed to parse dirstat cut-off percentage '%s'\n"),
116 p);
117 ret++;
118 }
119 } else {
120 strbuf_addf(errmsg, _(" Unknown dirstat parameter '%s'\n"), p);
121 ret++;
122 }
123
124 }
125 string_list_clear(&params, 0);
126 free(params_copy);
127 return ret;
128 }
129
130 static int parse_submodule_params(struct diff_options *options, const char *value)
131 {
132 if (!strcmp(value, "log"))
133 DIFF_OPT_SET(options, SUBMODULE_LOG);
134 else if (!strcmp(value, "short"))
135 DIFF_OPT_CLR(options, SUBMODULE_LOG);
136 else
137 return -1;
138 return 0;
139 }
140
141 static int git_config_rename(const char *var, const char *value)
142 {
143 if (!value)
144 return DIFF_DETECT_RENAME;
145 if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
146 return DIFF_DETECT_COPY;
147 return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
148 }
149
150 long parse_algorithm_value(const char *value)
151 {
152 if (!value)
153 return -1;
154 else if (!strcasecmp(value, "myers") || !strcasecmp(value, "default"))
155 return 0;
156 else if (!strcasecmp(value, "minimal"))
157 return XDF_NEED_MINIMAL;
158 else if (!strcasecmp(value, "patience"))
159 return XDF_PATIENCE_DIFF;
160 else if (!strcasecmp(value, "histogram"))
161 return XDF_HISTOGRAM_DIFF;
162 return -1;
163 }
164
165 /*
166 * These are to give UI layer defaults.
167 * The core-level commands such as git-diff-files should
168 * never be affected by the setting of diff.renames
169 * the user happens to have in the configuration file.
170 */
171 int git_diff_ui_config(const char *var, const char *value, void *cb)
172 {
173 if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
174 diff_use_color_default = git_config_colorbool(var, value);
175 return 0;
176 }
177 if (!strcmp(var, "diff.context")) {
178 diff_context_default = git_config_int(var, value);
179 if (diff_context_default < 0)
180 return -1;
181 return 0;
182 }
183 if (!strcmp(var, "diff.renames")) {
184 diff_detect_rename_default = git_config_rename(var, value);
185 return 0;
186 }
187 if (!strcmp(var, "diff.autorefreshindex")) {
188 diff_auto_refresh_index = git_config_bool(var, value);
189 return 0;
190 }
191 if (!strcmp(var, "diff.mnemonicprefix")) {
192 diff_mnemonic_prefix = git_config_bool(var, value);
193 return 0;
194 }
195 if (!strcmp(var, "diff.noprefix")) {
196 diff_no_prefix = git_config_bool(var, value);
197 return 0;
198 }
199 if (!strcmp(var, "diff.statgraphwidth")) {
200 diff_stat_graph_width = git_config_int(var, value);
201 return 0;
202 }
203 if (!strcmp(var, "diff.external"))
204 return git_config_string(&external_diff_cmd_cfg, var, value);
205 if (!strcmp(var, "diff.wordregex"))
206 return git_config_string(&diff_word_regex_cfg, var, value);
207 if (!strcmp(var, "diff.orderfile"))
208 return git_config_pathname(&diff_order_file_cfg, var, value);
209
210 if (!strcmp(var, "diff.ignoresubmodules"))
211 handle_ignore_submodules_arg(&default_diff_options, value);
212
213 if (!strcmp(var, "diff.submodule")) {
214 if (parse_submodule_params(&default_diff_options, value))
215 warning(_("Unknown value for 'diff.submodule' config variable: '%s'"),
216 value);
217 return 0;
218 }
219
220 if (!strcmp(var, "diff.algorithm")) {
221 diff_algorithm = parse_algorithm_value(value);
222 if (diff_algorithm < 0)
223 return -1;
224 return 0;
225 }
226
227 if (git_color_config(var, value, cb) < 0)
228 return -1;
229
230 return git_diff_basic_config(var, value, cb);
231 }
232
233 int git_diff_basic_config(const char *var, const char *value, void *cb)
234 {
235 const char *name;
236
237 if (!strcmp(var, "diff.renamelimit")) {
238 diff_rename_limit_default = git_config_int(var, value);
239 return 0;
240 }
241
242 if (userdiff_config(var, value) < 0)
243 return -1;
244
245 if (skip_prefix(var, "diff.color.", &name) ||
246 skip_prefix(var, "color.diff.", &name)) {
247 int slot = parse_diff_color_slot(name);
248 if (slot < 0)
249 return 0;
250 if (!value)
251 return config_error_nonbool(var);
252 return color_parse(value, diff_colors[slot]);
253 }
254
255 /* like GNU diff's --suppress-blank-empty option */
256 if (!strcmp(var, "diff.suppressblankempty") ||
257 /* for backwards compatibility */
258 !strcmp(var, "diff.suppress-blank-empty")) {
259 diff_suppress_blank_empty = git_config_bool(var, value);
260 return 0;
261 }
262
263 if (!strcmp(var, "diff.dirstat")) {
264 struct strbuf errmsg = STRBUF_INIT;
265 default_diff_options.dirstat_permille = diff_dirstat_permille_default;
266 if (parse_dirstat_params(&default_diff_options, value, &errmsg))
267 warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
268 errmsg.buf);
269 strbuf_release(&errmsg);
270 diff_dirstat_permille_default = default_diff_options.dirstat_permille;
271 return 0;
272 }
273
274 if (starts_with(var, "submodule."))
275 return parse_submodule_config_option(var, value);
276
277 return git_default_config(var, value, cb);
278 }
279
280 static char *quote_two(const char *one, const char *two)
281 {
282 int need_one = quote_c_style(one, NULL, NULL, 1);
283 int need_two = quote_c_style(two, NULL, NULL, 1);
284 struct strbuf res = STRBUF_INIT;
285
286 if (need_one + need_two) {
287 strbuf_addch(&res, '"');
288 quote_c_style(one, &res, NULL, 1);
289 quote_c_style(two, &res, NULL, 1);
290 strbuf_addch(&res, '"');
291 } else {
292 strbuf_addstr(&res, one);
293 strbuf_addstr(&res, two);
294 }
295 return strbuf_detach(&res, NULL);
296 }
297
298 static const char *external_diff(void)
299 {
300 static const char *external_diff_cmd = NULL;
301 static int done_preparing = 0;
302
303 if (done_preparing)
304 return external_diff_cmd;
305 external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
306 if (!external_diff_cmd)
307 external_diff_cmd = external_diff_cmd_cfg;
308 done_preparing = 1;
309 return external_diff_cmd;
310 }
311
312 /*
313 * Keep track of files used for diffing. Sometimes such an entry
314 * refers to a temporary file, sometimes to an existing file, and
315 * sometimes to "/dev/null".
316 */
317 static struct diff_tempfile {
318 /*
319 * filename external diff should read from, or NULL if this
320 * entry is currently not in use:
321 */
322 const char *name;
323
324 char hex[41];
325 char mode[10];
326
327 /*
328 * If this diff_tempfile instance refers to a temporary file,
329 * this tempfile object is used to manage its lifetime.
330 */
331 struct tempfile tempfile;
332 } diff_temp[2];
333
334 typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
335
336 struct emit_callback {
337 int color_diff;
338 unsigned ws_rule;
339 int blank_at_eof_in_preimage;
340 int blank_at_eof_in_postimage;
341 int lno_in_preimage;
342 int lno_in_postimage;
343 sane_truncate_fn truncate;
344 const char **label_path;
345 struct diff_words_data *diff_words;
346 struct diff_options *opt;
347 int *found_changesp;
348 struct strbuf *header;
349 };
350
351 static int count_lines(const char *data, int size)
352 {
353 int count, ch, completely_empty = 1, nl_just_seen = 0;
354 count = 0;
355 while (0 < size--) {
356 ch = *data++;
357 if (ch == '\n') {
358 count++;
359 nl_just_seen = 1;
360 completely_empty = 0;
361 }
362 else {
363 nl_just_seen = 0;
364 completely_empty = 0;
365 }
366 }
367 if (completely_empty)
368 return 0;
369 if (!nl_just_seen)
370 count++; /* no trailing newline */
371 return count;
372 }
373
374 static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
375 {
376 if (!DIFF_FILE_VALID(one)) {
377 mf->ptr = (char *)""; /* does not matter */
378 mf->size = 0;
379 return 0;
380 }
381 else if (diff_populate_filespec(one, 0))
382 return -1;
383
384 mf->ptr = one->data;
385 mf->size = one->size;
386 return 0;
387 }
388
389 /* like fill_mmfile, but only for size, so we can avoid retrieving blob */
390 static unsigned long diff_filespec_size(struct diff_filespec *one)
391 {
392 if (!DIFF_FILE_VALID(one))
393 return 0;
394 diff_populate_filespec(one, CHECK_SIZE_ONLY);
395 return one->size;
396 }
397
398 static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
399 {
400 char *ptr = mf->ptr;
401 long size = mf->size;
402 int cnt = 0;
403
404 if (!size)
405 return cnt;
406 ptr += size - 1; /* pointing at the very end */
407 if (*ptr != '\n')
408 ; /* incomplete line */
409 else
410 ptr--; /* skip the last LF */
411 while (mf->ptr < ptr) {
412 char *prev_eol;
413 for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
414 if (*prev_eol == '\n')
415 break;
416 if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
417 break;
418 cnt++;
419 ptr = prev_eol - 1;
420 }
421 return cnt;
422 }
423
424 static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
425 struct emit_callback *ecbdata)
426 {
427 int l1, l2, at;
428 unsigned ws_rule = ecbdata->ws_rule;
429 l1 = count_trailing_blank(mf1, ws_rule);
430 l2 = count_trailing_blank(mf2, ws_rule);
431 if (l2 <= l1) {
432 ecbdata->blank_at_eof_in_preimage = 0;
433 ecbdata->blank_at_eof_in_postimage = 0;
434 return;
435 }
436 at = count_lines(mf1->ptr, mf1->size);
437 ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
438
439 at = count_lines(mf2->ptr, mf2->size);
440 ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
441 }
442
443 static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
444 int first, const char *line, int len)
445 {
446 int has_trailing_newline, has_trailing_carriage_return;
447 int nofirst;
448 FILE *file = o->file;
449
450 fputs(diff_line_prefix(o), file);
451
452 if (len == 0) {
453 has_trailing_newline = (first == '\n');
454 has_trailing_carriage_return = (!has_trailing_newline &&
455 (first == '\r'));
456 nofirst = has_trailing_newline || has_trailing_carriage_return;
457 } else {
458 has_trailing_newline = (len > 0 && line[len-1] == '\n');
459 if (has_trailing_newline)
460 len--;
461 has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
462 if (has_trailing_carriage_return)
463 len--;
464 nofirst = 0;
465 }
466
467 if (len || !nofirst) {
468 fputs(set, file);
469 if (!nofirst)
470 fputc(first, file);
471 fwrite(line, len, 1, file);
472 fputs(reset, file);
473 }
474 if (has_trailing_carriage_return)
475 fputc('\r', file);
476 if (has_trailing_newline)
477 fputc('\n', file);
478 }
479
480 static void emit_line(struct diff_options *o, const char *set, const char *reset,
481 const char *line, int len)
482 {
483 emit_line_0(o, set, reset, line[0], line+1, len-1);
484 }
485
486 static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
487 {
488 if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
489 ecbdata->blank_at_eof_in_preimage &&
490 ecbdata->blank_at_eof_in_postimage &&
491 ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
492 ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
493 return 0;
494 return ws_blank_line(line, len, ecbdata->ws_rule);
495 }
496
497 static void emit_line_checked(const char *reset,
498 struct emit_callback *ecbdata,
499 const char *line, int len,
500 enum color_diff color,
501 unsigned ws_error_highlight,
502 char sign)
503 {
504 const char *set = diff_get_color(ecbdata->color_diff, color);
505 const char *ws = NULL;
506
507 if (ecbdata->opt->ws_error_highlight & ws_error_highlight) {
508 ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
509 if (!*ws)
510 ws = NULL;
511 }
512
513 if (!ws)
514 emit_line_0(ecbdata->opt, set, reset, sign, line, len);
515 else if (sign == '+' && new_blank_line_at_eof(ecbdata, line, len))
516 /* Blank line at EOF - paint '+' as well */
517 emit_line_0(ecbdata->opt, ws, reset, sign, line, len);
518 else {
519 /* Emit just the prefix, then the rest. */
520 emit_line_0(ecbdata->opt, set, reset, sign, "", 0);
521 ws_check_emit(line, len, ecbdata->ws_rule,
522 ecbdata->opt->file, set, reset, ws);
523 }
524 }
525
526 static void emit_add_line(const char *reset,
527 struct emit_callback *ecbdata,
528 const char *line, int len)
529 {
530 emit_line_checked(reset, ecbdata, line, len,
531 DIFF_FILE_NEW, WSEH_NEW, '+');
532 }
533
534 static void emit_del_line(const char *reset,
535 struct emit_callback *ecbdata,
536 const char *line, int len)
537 {
538 emit_line_checked(reset, ecbdata, line, len,
539 DIFF_FILE_OLD, WSEH_OLD, '-');
540 }
541
542 static void emit_context_line(const char *reset,
543 struct emit_callback *ecbdata,
544 const char *line, int len)
545 {
546 emit_line_checked(reset, ecbdata, line, len,
547 DIFF_CONTEXT, WSEH_CONTEXT, ' ');
548 }
549
550 static void emit_hunk_header(struct emit_callback *ecbdata,
551 const char *line, int len)
552 {
553 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
554 const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
555 const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
556 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
557 static const char atat[2] = { '@', '@' };
558 const char *cp, *ep;
559 struct strbuf msgbuf = STRBUF_INIT;
560 int org_len = len;
561 int i = 1;
562
563 /*
564 * As a hunk header must begin with "@@ -<old>, +<new> @@",
565 * it always is at least 10 bytes long.
566 */
567 if (len < 10 ||
568 memcmp(line, atat, 2) ||
569 !(ep = memmem(line + 2, len - 2, atat, 2))) {
570 emit_line(ecbdata->opt, context, reset, line, len);
571 return;
572 }
573 ep += 2; /* skip over @@ */
574
575 /* The hunk header in fraginfo color */
576 strbuf_addstr(&msgbuf, frag);
577 strbuf_add(&msgbuf, line, ep - line);
578 strbuf_addstr(&msgbuf, reset);
579
580 /*
581 * trailing "\r\n"
582 */
583 for ( ; i < 3; i++)
584 if (line[len - i] == '\r' || line[len - i] == '\n')
585 len--;
586
587 /* blank before the func header */
588 for (cp = ep; ep - line < len; ep++)
589 if (*ep != ' ' && *ep != '\t')
590 break;
591 if (ep != cp) {
592 strbuf_addstr(&msgbuf, context);
593 strbuf_add(&msgbuf, cp, ep - cp);
594 strbuf_addstr(&msgbuf, reset);
595 }
596
597 if (ep < line + len) {
598 strbuf_addstr(&msgbuf, func);
599 strbuf_add(&msgbuf, ep, line + len - ep);
600 strbuf_addstr(&msgbuf, reset);
601 }
602
603 strbuf_add(&msgbuf, line + len, org_len - len);
604 emit_line(ecbdata->opt, "", "", msgbuf.buf, msgbuf.len);
605 strbuf_release(&msgbuf);
606 }
607
608 static struct diff_tempfile *claim_diff_tempfile(void) {
609 int i;
610 for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
611 if (!diff_temp[i].name)
612 return diff_temp + i;
613 die("BUG: diff is failing to clean up its tempfiles");
614 }
615
616 static void remove_tempfile(void)
617 {
618 int i;
619 for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
620 if (is_tempfile_active(&diff_temp[i].tempfile))
621 delete_tempfile(&diff_temp[i].tempfile);
622 diff_temp[i].name = NULL;
623 }
624 }
625
626 static void print_line_count(FILE *file, int count)
627 {
628 switch (count) {
629 case 0:
630 fprintf(file, "0,0");
631 break;
632 case 1:
633 fprintf(file, "1");
634 break;
635 default:
636 fprintf(file, "1,%d", count);
637 break;
638 }
639 }
640
641 static void emit_rewrite_lines(struct emit_callback *ecb,
642 int prefix, const char *data, int size)
643 {
644 const char *endp = NULL;
645 static const char *nneof = " No newline at end of file\n";
646 const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
647
648 while (0 < size) {
649 int len;
650
651 endp = memchr(data, '\n', size);
652 len = endp ? (endp - data + 1) : size;
653 if (prefix != '+') {
654 ecb->lno_in_preimage++;
655 emit_del_line(reset, ecb, data, len);
656 } else {
657 ecb->lno_in_postimage++;
658 emit_add_line(reset, ecb, data, len);
659 }
660 size -= len;
661 data += len;
662 }
663 if (!endp) {
664 const char *context = diff_get_color(ecb->color_diff,
665 DIFF_CONTEXT);
666 putc('\n', ecb->opt->file);
667 emit_line_0(ecb->opt, context, reset, '\\',
668 nneof, strlen(nneof));
669 }
670 }
671
672 static void emit_rewrite_diff(const char *name_a,
673 const char *name_b,
674 struct diff_filespec *one,
675 struct diff_filespec *two,
676 struct userdiff_driver *textconv_one,
677 struct userdiff_driver *textconv_two,
678 struct diff_options *o)
679 {
680 int lc_a, lc_b;
681 const char *name_a_tab, *name_b_tab;
682 const char *metainfo = diff_get_color(o->use_color, DIFF_METAINFO);
683 const char *fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
684 const char *reset = diff_get_color(o->use_color, DIFF_RESET);
685 static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
686 const char *a_prefix, *b_prefix;
687 char *data_one, *data_two;
688 size_t size_one, size_two;
689 struct emit_callback ecbdata;
690 const char *line_prefix = diff_line_prefix(o);
691
692 if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
693 a_prefix = o->b_prefix;
694 b_prefix = o->a_prefix;
695 } else {
696 a_prefix = o->a_prefix;
697 b_prefix = o->b_prefix;
698 }
699
700 name_a += (*name_a == '/');
701 name_b += (*name_b == '/');
702 name_a_tab = strchr(name_a, ' ') ? "\t" : "";
703 name_b_tab = strchr(name_b, ' ') ? "\t" : "";
704
705 strbuf_reset(&a_name);
706 strbuf_reset(&b_name);
707 quote_two_c_style(&a_name, a_prefix, name_a, 0);
708 quote_two_c_style(&b_name, b_prefix, name_b, 0);
709
710 size_one = fill_textconv(textconv_one, one, &data_one);
711 size_two = fill_textconv(textconv_two, two, &data_two);
712
713 memset(&ecbdata, 0, sizeof(ecbdata));
714 ecbdata.color_diff = want_color(o->use_color);
715 ecbdata.found_changesp = &o->found_changes;
716 ecbdata.ws_rule = whitespace_rule(name_b);
717 ecbdata.opt = o;
718 if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
719 mmfile_t mf1, mf2;
720 mf1.ptr = (char *)data_one;
721 mf2.ptr = (char *)data_two;
722 mf1.size = size_one;
723 mf2.size = size_two;
724 check_blank_at_eof(&mf1, &mf2, &ecbdata);
725 }
726 ecbdata.lno_in_preimage = 1;
727 ecbdata.lno_in_postimage = 1;
728
729 lc_a = count_lines(data_one, size_one);
730 lc_b = count_lines(data_two, size_two);
731 fprintf(o->file,
732 "%s%s--- %s%s%s\n%s%s+++ %s%s%s\n%s%s@@ -",
733 line_prefix, metainfo, a_name.buf, name_a_tab, reset,
734 line_prefix, metainfo, b_name.buf, name_b_tab, reset,
735 line_prefix, fraginfo);
736 if (!o->irreversible_delete)
737 print_line_count(o->file, lc_a);
738 else
739 fprintf(o->file, "?,?");
740 fprintf(o->file, " +");
741 print_line_count(o->file, lc_b);
742 fprintf(o->file, " @@%s\n", reset);
743 if (lc_a && !o->irreversible_delete)
744 emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
745 if (lc_b)
746 emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
747 if (textconv_one)
748 free((char *)data_one);
749 if (textconv_two)
750 free((char *)data_two);
751 }
752
753 struct diff_words_buffer {
754 mmfile_t text;
755 long alloc;
756 struct diff_words_orig {
757 const char *begin, *end;
758 } *orig;
759 int orig_nr, orig_alloc;
760 };
761
762 static void diff_words_append(char *line, unsigned long len,
763 struct diff_words_buffer *buffer)
764 {
765 ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
766 line++;
767 len--;
768 memcpy(buffer->text.ptr + buffer->text.size, line, len);
769 buffer->text.size += len;
770 buffer->text.ptr[buffer->text.size] = '\0';
771 }
772
773 struct diff_words_style_elem {
774 const char *prefix;
775 const char *suffix;
776 const char *color; /* NULL; filled in by the setup code if
777 * color is enabled */
778 };
779
780 struct diff_words_style {
781 enum diff_words_type type;
782 struct diff_words_style_elem new, old, ctx;
783 const char *newline;
784 };
785
786 static struct diff_words_style diff_words_styles[] = {
787 { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
788 { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
789 { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
790 };
791
792 struct diff_words_data {
793 struct diff_words_buffer minus, plus;
794 const char *current_plus;
795 int last_minus;
796 struct diff_options *opt;
797 regex_t *word_regex;
798 enum diff_words_type type;
799 struct diff_words_style *style;
800 };
801
802 static int fn_out_diff_words_write_helper(FILE *fp,
803 struct diff_words_style_elem *st_el,
804 const char *newline,
805 size_t count, const char *buf,
806 const char *line_prefix)
807 {
808 int print = 0;
809
810 while (count) {
811 char *p = memchr(buf, '\n', count);
812 if (print)
813 fputs(line_prefix, fp);
814 if (p != buf) {
815 if (st_el->color && fputs(st_el->color, fp) < 0)
816 return -1;
817 if (fputs(st_el->prefix, fp) < 0 ||
818 fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
819 fputs(st_el->suffix, fp) < 0)
820 return -1;
821 if (st_el->color && *st_el->color
822 && fputs(GIT_COLOR_RESET, fp) < 0)
823 return -1;
824 }
825 if (!p)
826 return 0;
827 if (fputs(newline, fp) < 0)
828 return -1;
829 count -= p + 1 - buf;
830 buf = p + 1;
831 print = 1;
832 }
833 return 0;
834 }
835
836 /*
837 * '--color-words' algorithm can be described as:
838 *
839 * 1. collect a the minus/plus lines of a diff hunk, divided into
840 * minus-lines and plus-lines;
841 *
842 * 2. break both minus-lines and plus-lines into words and
843 * place them into two mmfile_t with one word for each line;
844 *
845 * 3. use xdiff to run diff on the two mmfile_t to get the words level diff;
846 *
847 * And for the common parts of the both file, we output the plus side text.
848 * diff_words->current_plus is used to trace the current position of the plus file
849 * which printed. diff_words->last_minus is used to trace the last minus word
850 * printed.
851 *
852 * For '--graph' to work with '--color-words', we need to output the graph prefix
853 * on each line of color words output. Generally, there are two conditions on
854 * which we should output the prefix.
855 *
856 * 1. diff_words->last_minus == 0 &&
857 * diff_words->current_plus == diff_words->plus.text.ptr
858 *
859 * that is: the plus text must start as a new line, and if there is no minus
860 * word printed, a graph prefix must be printed.
861 *
862 * 2. diff_words->current_plus > diff_words->plus.text.ptr &&
863 * *(diff_words->current_plus - 1) == '\n'
864 *
865 * that is: a graph prefix must be printed following a '\n'
866 */
867 static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
868 {
869 if ((diff_words->last_minus == 0 &&
870 diff_words->current_plus == diff_words->plus.text.ptr) ||
871 (diff_words->current_plus > diff_words->plus.text.ptr &&
872 *(diff_words->current_plus - 1) == '\n')) {
873 return 1;
874 } else {
875 return 0;
876 }
877 }
878
879 static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
880 {
881 struct diff_words_data *diff_words = priv;
882 struct diff_words_style *style = diff_words->style;
883 int minus_first, minus_len, plus_first, plus_len;
884 const char *minus_begin, *minus_end, *plus_begin, *plus_end;
885 struct diff_options *opt = diff_words->opt;
886 const char *line_prefix;
887
888 if (line[0] != '@' || parse_hunk_header(line, len,
889 &minus_first, &minus_len, &plus_first, &plus_len))
890 return;
891
892 assert(opt);
893 line_prefix = diff_line_prefix(opt);
894
895 /* POSIX requires that first be decremented by one if len == 0... */
896 if (minus_len) {
897 minus_begin = diff_words->minus.orig[minus_first].begin;
898 minus_end =
899 diff_words->minus.orig[minus_first + minus_len - 1].end;
900 } else
901 minus_begin = minus_end =
902 diff_words->minus.orig[minus_first].end;
903
904 if (plus_len) {
905 plus_begin = diff_words->plus.orig[plus_first].begin;
906 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
907 } else
908 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
909
910 if (color_words_output_graph_prefix(diff_words)) {
911 fputs(line_prefix, diff_words->opt->file);
912 }
913 if (diff_words->current_plus != plus_begin) {
914 fn_out_diff_words_write_helper(diff_words->opt->file,
915 &style->ctx, style->newline,
916 plus_begin - diff_words->current_plus,
917 diff_words->current_plus, line_prefix);
918 if (*(plus_begin - 1) == '\n')
919 fputs(line_prefix, diff_words->opt->file);
920 }
921 if (minus_begin != minus_end) {
922 fn_out_diff_words_write_helper(diff_words->opt->file,
923 &style->old, style->newline,
924 minus_end - minus_begin, minus_begin,
925 line_prefix);
926 }
927 if (plus_begin != plus_end) {
928 fn_out_diff_words_write_helper(diff_words->opt->file,
929 &style->new, style->newline,
930 plus_end - plus_begin, plus_begin,
931 line_prefix);
932 }
933
934 diff_words->current_plus = plus_end;
935 diff_words->last_minus = minus_first;
936 }
937
938 /* This function starts looking at *begin, and returns 0 iff a word was found. */
939 static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
940 int *begin, int *end)
941 {
942 if (word_regex && *begin < buffer->size) {
943 regmatch_t match[1];
944 if (!regexec(word_regex, buffer->ptr + *begin, 1, match, 0)) {
945 char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
946 '\n', match[0].rm_eo - match[0].rm_so);
947 *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
948 *begin += match[0].rm_so;
949 return *begin >= *end;
950 }
951 return -1;
952 }
953
954 /* find the next word */
955 while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
956 (*begin)++;
957 if (*begin >= buffer->size)
958 return -1;
959
960 /* find the end of the word */
961 *end = *begin + 1;
962 while (*end < buffer->size && !isspace(buffer->ptr[*end]))
963 (*end)++;
964
965 return 0;
966 }
967
968 /*
969 * This function splits the words in buffer->text, stores the list with
970 * newline separator into out, and saves the offsets of the original words
971 * in buffer->orig.
972 */
973 static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
974 regex_t *word_regex)
975 {
976 int i, j;
977 long alloc = 0;
978
979 out->size = 0;
980 out->ptr = NULL;
981
982 /* fake an empty "0th" word */
983 ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
984 buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
985 buffer->orig_nr = 1;
986
987 for (i = 0; i < buffer->text.size; i++) {
988 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
989 return;
990
991 /* store original boundaries */
992 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
993 buffer->orig_alloc);
994 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
995 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
996 buffer->orig_nr++;
997
998 /* store one word */
999 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
1000 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
1001 out->ptr[out->size + j - i] = '\n';
1002 out->size += j - i + 1;
1003
1004 i = j - 1;
1005 }
1006 }
1007
1008 /* this executes the word diff on the accumulated buffers */
1009 static void diff_words_show(struct diff_words_data *diff_words)
1010 {
1011 xpparam_t xpp;
1012 xdemitconf_t xecfg;
1013 mmfile_t minus, plus;
1014 struct diff_words_style *style = diff_words->style;
1015
1016 struct diff_options *opt = diff_words->opt;
1017 const char *line_prefix;
1018
1019 assert(opt);
1020 line_prefix = diff_line_prefix(opt);
1021
1022 /* special case: only removal */
1023 if (!diff_words->plus.text.size) {
1024 fputs(line_prefix, diff_words->opt->file);
1025 fn_out_diff_words_write_helper(diff_words->opt->file,
1026 &style->old, style->newline,
1027 diff_words->minus.text.size,
1028 diff_words->minus.text.ptr, line_prefix);
1029 diff_words->minus.text.size = 0;
1030 return;
1031 }
1032
1033 diff_words->current_plus = diff_words->plus.text.ptr;
1034 diff_words->last_minus = 0;
1035
1036 memset(&xpp, 0, sizeof(xpp));
1037 memset(&xecfg, 0, sizeof(xecfg));
1038 diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
1039 diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
1040 xpp.flags = 0;
1041 /* as only the hunk header will be parsed, we need a 0-context */
1042 xecfg.ctxlen = 0;
1043 xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
1044 &xpp, &xecfg);
1045 free(minus.ptr);
1046 free(plus.ptr);
1047 if (diff_words->current_plus != diff_words->plus.text.ptr +
1048 diff_words->plus.text.size) {
1049 if (color_words_output_graph_prefix(diff_words))
1050 fputs(line_prefix, diff_words->opt->file);
1051 fn_out_diff_words_write_helper(diff_words->opt->file,
1052 &style->ctx, style->newline,
1053 diff_words->plus.text.ptr + diff_words->plus.text.size
1054 - diff_words->current_plus, diff_words->current_plus,
1055 line_prefix);
1056 }
1057 diff_words->minus.text.size = diff_words->plus.text.size = 0;
1058 }
1059
1060 /* In "color-words" mode, show word-diff of words accumulated in the buffer */
1061 static void diff_words_flush(struct emit_callback *ecbdata)
1062 {
1063 if (ecbdata->diff_words->minus.text.size ||
1064 ecbdata->diff_words->plus.text.size)
1065 diff_words_show(ecbdata->diff_words);
1066 }
1067
1068 static void diff_filespec_load_driver(struct diff_filespec *one)
1069 {
1070 /* Use already-loaded driver */
1071 if (one->driver)
1072 return;
1073
1074 if (S_ISREG(one->mode))
1075 one->driver = userdiff_find_by_path(one->path);
1076
1077 /* Fallback to default settings */
1078 if (!one->driver)
1079 one->driver = userdiff_find_by_name("default");
1080 }
1081
1082 static const char *userdiff_word_regex(struct diff_filespec *one)
1083 {
1084 diff_filespec_load_driver(one);
1085 return one->driver->word_regex;
1086 }
1087
1088 static void init_diff_words_data(struct emit_callback *ecbdata,
1089 struct diff_options *orig_opts,
1090 struct diff_filespec *one,
1091 struct diff_filespec *two)
1092 {
1093 int i;
1094 struct diff_options *o = xmalloc(sizeof(struct diff_options));
1095 memcpy(o, orig_opts, sizeof(struct diff_options));
1096
1097 ecbdata->diff_words =
1098 xcalloc(1, sizeof(struct diff_words_data));
1099 ecbdata->diff_words->type = o->word_diff;
1100 ecbdata->diff_words->opt = o;
1101 if (!o->word_regex)
1102 o->word_regex = userdiff_word_regex(one);
1103 if (!o->word_regex)
1104 o->word_regex = userdiff_word_regex(two);
1105 if (!o->word_regex)
1106 o->word_regex = diff_word_regex_cfg;
1107 if (o->word_regex) {
1108 ecbdata->diff_words->word_regex = (regex_t *)
1109 xmalloc(sizeof(regex_t));
1110 if (regcomp(ecbdata->diff_words->word_regex,
1111 o->word_regex,
1112 REG_EXTENDED | REG_NEWLINE))
1113 die ("Invalid regular expression: %s",
1114 o->word_regex);
1115 }
1116 for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1117 if (o->word_diff == diff_words_styles[i].type) {
1118 ecbdata->diff_words->style =
1119 &diff_words_styles[i];
1120 break;
1121 }
1122 }
1123 if (want_color(o->use_color)) {
1124 struct diff_words_style *st = ecbdata->diff_words->style;
1125 st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1126 st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1127 st->ctx.color = diff_get_color_opt(o, DIFF_CONTEXT);
1128 }
1129 }
1130
1131 static void free_diff_words_data(struct emit_callback *ecbdata)
1132 {
1133 if (ecbdata->diff_words) {
1134 diff_words_flush(ecbdata);
1135 free (ecbdata->diff_words->opt);
1136 free (ecbdata->diff_words->minus.text.ptr);
1137 free (ecbdata->diff_words->minus.orig);
1138 free (ecbdata->diff_words->plus.text.ptr);
1139 free (ecbdata->diff_words->plus.orig);
1140 if (ecbdata->diff_words->word_regex) {
1141 regfree(ecbdata->diff_words->word_regex);
1142 free(ecbdata->diff_words->word_regex);
1143 }
1144 free(ecbdata->diff_words);
1145 ecbdata->diff_words = NULL;
1146 }
1147 }
1148
1149 const char *diff_get_color(int diff_use_color, enum color_diff ix)
1150 {
1151 if (want_color(diff_use_color))
1152 return diff_colors[ix];
1153 return "";
1154 }
1155
1156 const char *diff_line_prefix(struct diff_options *opt)
1157 {
1158 struct strbuf *msgbuf;
1159 if (!opt->output_prefix)
1160 return "";
1161
1162 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
1163 return msgbuf->buf;
1164 }
1165
1166 static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1167 {
1168 const char *cp;
1169 unsigned long allot;
1170 size_t l = len;
1171
1172 if (ecb->truncate)
1173 return ecb->truncate(line, len);
1174 cp = line;
1175 allot = l;
1176 while (0 < l) {
1177 (void) utf8_width(&cp, &l);
1178 if (!cp)
1179 break; /* truncated in the middle? */
1180 }
1181 return allot - l;
1182 }
1183
1184 static void find_lno(const char *line, struct emit_callback *ecbdata)
1185 {
1186 const char *p;
1187 ecbdata->lno_in_preimage = 0;
1188 ecbdata->lno_in_postimage = 0;
1189 p = strchr(line, '-');
1190 if (!p)
1191 return; /* cannot happen */
1192 ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1193 p = strchr(p, '+');
1194 if (!p)
1195 return; /* cannot happen */
1196 ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1197 }
1198
1199 static void fn_out_consume(void *priv, char *line, unsigned long len)
1200 {
1201 struct emit_callback *ecbdata = priv;
1202 const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
1203 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
1204 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1205 struct diff_options *o = ecbdata->opt;
1206 const char *line_prefix = diff_line_prefix(o);
1207
1208 if (ecbdata->header) {
1209 fprintf(ecbdata->opt->file, "%s", ecbdata->header->buf);
1210 strbuf_reset(ecbdata->header);
1211 ecbdata->header = NULL;
1212 }
1213 *(ecbdata->found_changesp) = 1;
1214
1215 if (ecbdata->label_path[0]) {
1216 const char *name_a_tab, *name_b_tab;
1217
1218 name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
1219 name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
1220
1221 fprintf(ecbdata->opt->file, "%s%s--- %s%s%s\n",
1222 line_prefix, meta, ecbdata->label_path[0], reset, name_a_tab);
1223 fprintf(ecbdata->opt->file, "%s%s+++ %s%s%s\n",
1224 line_prefix, meta, ecbdata->label_path[1], reset, name_b_tab);
1225 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1226 }
1227
1228 if (diff_suppress_blank_empty
1229 && len == 2 && line[0] == ' ' && line[1] == '\n') {
1230 line[0] = '\n';
1231 len = 1;
1232 }
1233
1234 if (line[0] == '@') {
1235 if (ecbdata->diff_words)
1236 diff_words_flush(ecbdata);
1237 len = sane_truncate_line(ecbdata, line, len);
1238 find_lno(line, ecbdata);
1239 emit_hunk_header(ecbdata, line, len);
1240 if (line[len-1] != '\n')
1241 putc('\n', ecbdata->opt->file);
1242 return;
1243 }
1244
1245 if (len < 1) {
1246 emit_line(ecbdata->opt, reset, reset, line, len);
1247 if (ecbdata->diff_words
1248 && ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN)
1249 fputs("~\n", ecbdata->opt->file);
1250 return;
1251 }
1252
1253 if (ecbdata->diff_words) {
1254 if (line[0] == '-') {
1255 diff_words_append(line, len,
1256 &ecbdata->diff_words->minus);
1257 return;
1258 } else if (line[0] == '+') {
1259 diff_words_append(line, len,
1260 &ecbdata->diff_words->plus);
1261 return;
1262 } else if (starts_with(line, "\\ ")) {
1263 /*
1264 * Eat the "no newline at eof" marker as if we
1265 * saw a "+" or "-" line with nothing on it,
1266 * and return without diff_words_flush() to
1267 * defer processing. If this is the end of
1268 * preimage, more "+" lines may come after it.
1269 */
1270 return;
1271 }
1272 diff_words_flush(ecbdata);
1273 if (ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN) {
1274 emit_line(ecbdata->opt, context, reset, line, len);
1275 fputs("~\n", ecbdata->opt->file);
1276 } else {
1277 /*
1278 * Skip the prefix character, if any. With
1279 * diff_suppress_blank_empty, there may be
1280 * none.
1281 */
1282 if (line[0] != '\n') {
1283 line++;
1284 len--;
1285 }
1286 emit_line(ecbdata->opt, context, reset, line, len);
1287 }
1288 return;
1289 }
1290
1291 switch (line[0]) {
1292 case '+':
1293 ecbdata->lno_in_postimage++;
1294 emit_add_line(reset, ecbdata, line + 1, len - 1);
1295 break;
1296 case '-':
1297 ecbdata->lno_in_preimage++;
1298 emit_del_line(reset, ecbdata, line + 1, len - 1);
1299 break;
1300 case ' ':
1301 ecbdata->lno_in_postimage++;
1302 ecbdata->lno_in_preimage++;
1303 emit_context_line(reset, ecbdata, line + 1, len - 1);
1304 break;
1305 default:
1306 /* incomplete line at the end */
1307 ecbdata->lno_in_preimage++;
1308 emit_line(ecbdata->opt,
1309 diff_get_color(ecbdata->color_diff, DIFF_CONTEXT),
1310 reset, line, len);
1311 break;
1312 }
1313 }
1314
1315 static char *pprint_rename(const char *a, const char *b)
1316 {
1317 const char *old = a;
1318 const char *new = b;
1319 struct strbuf name = STRBUF_INIT;
1320 int pfx_length, sfx_length;
1321 int pfx_adjust_for_slash;
1322 int len_a = strlen(a);
1323 int len_b = strlen(b);
1324 int a_midlen, b_midlen;
1325 int qlen_a = quote_c_style(a, NULL, NULL, 0);
1326 int qlen_b = quote_c_style(b, NULL, NULL, 0);
1327
1328 if (qlen_a || qlen_b) {
1329 quote_c_style(a, &name, NULL, 0);
1330 strbuf_addstr(&name, " => ");
1331 quote_c_style(b, &name, NULL, 0);
1332 return strbuf_detach(&name, NULL);
1333 }
1334
1335 /* Find common prefix */
1336 pfx_length = 0;
1337 while (*old && *new && *old == *new) {
1338 if (*old == '/')
1339 pfx_length = old - a + 1;
1340 old++;
1341 new++;
1342 }
1343
1344 /* Find common suffix */
1345 old = a + len_a;
1346 new = b + len_b;
1347 sfx_length = 0;
1348 /*
1349 * If there is a common prefix, it must end in a slash. In
1350 * that case we let this loop run 1 into the prefix to see the
1351 * same slash.
1352 *
1353 * If there is no common prefix, we cannot do this as it would
1354 * underrun the input strings.
1355 */
1356 pfx_adjust_for_slash = (pfx_length ? 1 : 0);
1357 while (a + pfx_length - pfx_adjust_for_slash <= old &&
1358 b + pfx_length - pfx_adjust_for_slash <= new &&
1359 *old == *new) {
1360 if (*old == '/')
1361 sfx_length = len_a - (old - a);
1362 old--;
1363 new--;
1364 }
1365
1366 /*
1367 * pfx{mid-a => mid-b}sfx
1368 * {pfx-a => pfx-b}sfx
1369 * pfx{sfx-a => sfx-b}
1370 * name-a => name-b
1371 */
1372 a_midlen = len_a - pfx_length - sfx_length;
1373 b_midlen = len_b - pfx_length - sfx_length;
1374 if (a_midlen < 0)
1375 a_midlen = 0;
1376 if (b_midlen < 0)
1377 b_midlen = 0;
1378
1379 strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1380 if (pfx_length + sfx_length) {
1381 strbuf_add(&name, a, pfx_length);
1382 strbuf_addch(&name, '{');
1383 }
1384 strbuf_add(&name, a + pfx_length, a_midlen);
1385 strbuf_addstr(&name, " => ");
1386 strbuf_add(&name, b + pfx_length, b_midlen);
1387 if (pfx_length + sfx_length) {
1388 strbuf_addch(&name, '}');
1389 strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1390 }
1391 return strbuf_detach(&name, NULL);
1392 }
1393
1394 struct diffstat_t {
1395 int nr;
1396 int alloc;
1397 struct diffstat_file {
1398 char *from_name;
1399 char *name;
1400 char *print_name;
1401 unsigned is_unmerged:1;
1402 unsigned is_binary:1;
1403 unsigned is_renamed:1;
1404 unsigned is_interesting:1;
1405 uintmax_t added, deleted;
1406 } **files;
1407 };
1408
1409 static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1410 const char *name_a,
1411 const char *name_b)
1412 {
1413 struct diffstat_file *x;
1414 x = xcalloc(1, sizeof(*x));
1415 ALLOC_GROW(diffstat->files, diffstat->nr + 1, diffstat->alloc);
1416 diffstat->files[diffstat->nr++] = x;
1417 if (name_b) {
1418 x->from_name = xstrdup(name_a);
1419 x->name = xstrdup(name_b);
1420 x->is_renamed = 1;
1421 }
1422 else {
1423 x->from_name = NULL;
1424 x->name = xstrdup(name_a);
1425 }
1426 return x;
1427 }
1428
1429 static void diffstat_consume(void *priv, char *line, unsigned long len)
1430 {
1431 struct diffstat_t *diffstat = priv;
1432 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1433
1434 if (line[0] == '+')
1435 x->added++;
1436 else if (line[0] == '-')
1437 x->deleted++;
1438 }
1439
1440 const char mime_boundary_leader[] = "------------";
1441
1442 static int scale_linear(int it, int width, int max_change)
1443 {
1444 if (!it)
1445 return 0;
1446 /*
1447 * make sure that at least one '-' or '+' is printed if
1448 * there is any change to this path. The easiest way is to
1449 * scale linearly as if the alloted width is one column shorter
1450 * than it is, and then add 1 to the result.
1451 */
1452 return 1 + (it * (width - 1) / max_change);
1453 }
1454
1455 static void show_name(FILE *file,
1456 const char *prefix, const char *name, int len)
1457 {
1458 fprintf(file, " %s%-*s |", prefix, len, name);
1459 }
1460
1461 static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1462 {
1463 if (cnt <= 0)
1464 return;
1465 fprintf(file, "%s", set);
1466 while (cnt--)
1467 putc(ch, file);
1468 fprintf(file, "%s", reset);
1469 }
1470
1471 static void fill_print_name(struct diffstat_file *file)
1472 {
1473 char *pname;
1474
1475 if (file->print_name)
1476 return;
1477
1478 if (!file->is_renamed) {
1479 struct strbuf buf = STRBUF_INIT;
1480 if (quote_c_style(file->name, &buf, NULL, 0)) {
1481 pname = strbuf_detach(&buf, NULL);
1482 } else {
1483 pname = file->name;
1484 strbuf_release(&buf);
1485 }
1486 } else {
1487 pname = pprint_rename(file->from_name, file->name);
1488 }
1489 file->print_name = pname;
1490 }
1491
1492 int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1493 {
1494 struct strbuf sb = STRBUF_INIT;
1495 int ret;
1496
1497 if (!files) {
1498 assert(insertions == 0 && deletions == 0);
1499 return fprintf(fp, "%s\n", " 0 files changed");
1500 }
1501
1502 strbuf_addf(&sb,
1503 (files == 1) ? " %d file changed" : " %d files changed",
1504 files);
1505
1506 /*
1507 * For binary diff, the caller may want to print "x files
1508 * changed" with insertions == 0 && deletions == 0.
1509 *
1510 * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1511 * is probably less confusing (i.e skip over "2 files changed
1512 * but nothing about added/removed lines? Is this a bug in Git?").
1513 */
1514 if (insertions || deletions == 0) {
1515 strbuf_addf(&sb,
1516 (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
1517 insertions);
1518 }
1519
1520 if (deletions || insertions == 0) {
1521 strbuf_addf(&sb,
1522 (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
1523 deletions);
1524 }
1525 strbuf_addch(&sb, '\n');
1526 ret = fputs(sb.buf, fp);
1527 strbuf_release(&sb);
1528 return ret;
1529 }
1530
1531 static void show_stats(struct diffstat_t *data, struct diff_options *options)
1532 {
1533 int i, len, add, del, adds = 0, dels = 0;
1534 uintmax_t max_change = 0, max_len = 0;
1535 int total_files = data->nr, count;
1536 int width, name_width, graph_width, number_width = 0, bin_width = 0;
1537 const char *reset, *add_c, *del_c;
1538 const char *line_prefix = "";
1539 int extra_shown = 0;
1540
1541 if (data->nr == 0)
1542 return;
1543
1544 line_prefix = diff_line_prefix(options);
1545 count = options->stat_count ? options->stat_count : data->nr;
1546
1547 reset = diff_get_color_opt(options, DIFF_RESET);
1548 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1549 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1550
1551 /*
1552 * Find the longest filename and max number of changes
1553 */
1554 for (i = 0; (i < count) && (i < data->nr); i++) {
1555 struct diffstat_file *file = data->files[i];
1556 uintmax_t change = file->added + file->deleted;
1557
1558 if (!file->is_interesting && (change == 0)) {
1559 count++; /* not shown == room for one more */
1560 continue;
1561 }
1562 fill_print_name(file);
1563 len = strlen(file->print_name);
1564 if (max_len < len)
1565 max_len = len;
1566
1567 if (file->is_unmerged) {
1568 /* "Unmerged" is 8 characters */
1569 bin_width = bin_width < 8 ? 8 : bin_width;
1570 continue;
1571 }
1572 if (file->is_binary) {
1573 /* "Bin XXX -> YYY bytes" */
1574 int w = 14 + decimal_width(file->added)
1575 + decimal_width(file->deleted);
1576 bin_width = bin_width < w ? w : bin_width;
1577 /* Display change counts aligned with "Bin" */
1578 number_width = 3;
1579 continue;
1580 }
1581
1582 if (max_change < change)
1583 max_change = change;
1584 }
1585 count = i; /* where we can stop scanning in data->files[] */
1586
1587 /*
1588 * We have width = stat_width or term_columns() columns total.
1589 * We want a maximum of min(max_len, stat_name_width) for the name part.
1590 * We want a maximum of min(max_change, stat_graph_width) for the +- part.
1591 * We also need 1 for " " and 4 + decimal_width(max_change)
1592 * for " | NNNN " and one the empty column at the end, altogether
1593 * 6 + decimal_width(max_change).
1594 *
1595 * If there's not enough space, we will use the smaller of
1596 * stat_name_width (if set) and 5/8*width for the filename,
1597 * and the rest for constant elements + graph part, but no more
1598 * than stat_graph_width for the graph part.
1599 * (5/8 gives 50 for filename and 30 for the constant parts + graph
1600 * for the standard terminal size).
1601 *
1602 * In other words: stat_width limits the maximum width, and
1603 * stat_name_width fixes the maximum width of the filename,
1604 * and is also used to divide available columns if there
1605 * aren't enough.
1606 *
1607 * Binary files are displayed with "Bin XXX -> YYY bytes"
1608 * instead of the change count and graph. This part is treated
1609 * similarly to the graph part, except that it is not
1610 * "scaled". If total width is too small to accommodate the
1611 * guaranteed minimum width of the filename part and the
1612 * separators and this message, this message will "overflow"
1613 * making the line longer than the maximum width.
1614 */
1615
1616 if (options->stat_width == -1)
1617 width = term_columns() - options->output_prefix_length;
1618 else
1619 width = options->stat_width ? options->stat_width : 80;
1620 number_width = decimal_width(max_change) > number_width ?
1621 decimal_width(max_change) : number_width;
1622
1623 if (options->stat_graph_width == -1)
1624 options->stat_graph_width = diff_stat_graph_width;
1625
1626 /*
1627 * Guarantee 3/8*16==6 for the graph part
1628 * and 5/8*16==10 for the filename part
1629 */
1630 if (width < 16 + 6 + number_width)
1631 width = 16 + 6 + number_width;
1632
1633 /*
1634 * First assign sizes that are wanted, ignoring available width.
1635 * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
1636 * starting from "XXX" should fit in graph_width.
1637 */
1638 graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
1639 if (options->stat_graph_width &&
1640 options->stat_graph_width < graph_width)
1641 graph_width = options->stat_graph_width;
1642
1643 name_width = (options->stat_name_width > 0 &&
1644 options->stat_name_width < max_len) ?
1645 options->stat_name_width : max_len;
1646
1647 /*
1648 * Adjust adjustable widths not to exceed maximum width
1649 */
1650 if (name_width + number_width + 6 + graph_width > width) {
1651 if (graph_width > width * 3/8 - number_width - 6) {
1652 graph_width = width * 3/8 - number_width - 6;
1653 if (graph_width < 6)
1654 graph_width = 6;
1655 }
1656
1657 if (options->stat_graph_width &&
1658 graph_width > options->stat_graph_width)
1659 graph_width = options->stat_graph_width;
1660 if (name_width > width - number_width - 6 - graph_width)
1661 name_width = width - number_width - 6 - graph_width;
1662 else
1663 graph_width = width - number_width - 6 - name_width;
1664 }
1665
1666 /*
1667 * From here name_width is the width of the name area,
1668 * and graph_width is the width of the graph area.
1669 * max_change is used to scale graph properly.
1670 */
1671 for (i = 0; i < count; i++) {
1672 const char *prefix = "";
1673 struct diffstat_file *file = data->files[i];
1674 char *name = file->print_name;
1675 uintmax_t added = file->added;
1676 uintmax_t deleted = file->deleted;
1677 int name_len;
1678
1679 if (!file->is_interesting && (added + deleted == 0))
1680 continue;
1681
1682 /*
1683 * "scale" the filename
1684 */
1685 len = name_width;
1686 name_len = strlen(name);
1687 if (name_width < name_len) {
1688 char *slash;
1689 prefix = "...";
1690 len -= 3;
1691 name += name_len - len;
1692 slash = strchr(name, '/');
1693 if (slash)
1694 name = slash;
1695 }
1696
1697 if (file->is_binary) {
1698 fprintf(options->file, "%s", line_prefix);
1699 show_name(options->file, prefix, name, len);
1700 fprintf(options->file, " %*s", number_width, "Bin");
1701 if (!added && !deleted) {
1702 putc('\n', options->file);
1703 continue;
1704 }
1705 fprintf(options->file, " %s%"PRIuMAX"%s",
1706 del_c, deleted, reset);
1707 fprintf(options->file, " -> ");
1708 fprintf(options->file, "%s%"PRIuMAX"%s",
1709 add_c, added, reset);
1710 fprintf(options->file, " bytes");
1711 fprintf(options->file, "\n");
1712 continue;
1713 }
1714 else if (file->is_unmerged) {
1715 fprintf(options->file, "%s", line_prefix);
1716 show_name(options->file, prefix, name, len);
1717 fprintf(options->file, " Unmerged\n");
1718 continue;
1719 }
1720
1721 /*
1722 * scale the add/delete
1723 */
1724 add = added;
1725 del = deleted;
1726
1727 if (graph_width <= max_change) {
1728 int total = scale_linear(add + del, graph_width, max_change);
1729 if (total < 2 && add && del)
1730 /* width >= 2 due to the sanity check */
1731 total = 2;
1732 if (add < del) {
1733 add = scale_linear(add, graph_width, max_change);
1734 del = total - add;
1735 } else {
1736 del = scale_linear(del, graph_width, max_change);
1737 add = total - del;
1738 }
1739 }
1740 fprintf(options->file, "%s", line_prefix);
1741 show_name(options->file, prefix, name, len);
1742 fprintf(options->file, " %*"PRIuMAX"%s",
1743 number_width, added + deleted,
1744 added + deleted ? " " : "");
1745 show_graph(options->file, '+', add, add_c, reset);
1746 show_graph(options->file, '-', del, del_c, reset);
1747 fprintf(options->file, "\n");
1748 }
1749
1750 for (i = 0; i < data->nr; i++) {
1751 struct diffstat_file *file = data->files[i];
1752 uintmax_t added = file->added;
1753 uintmax_t deleted = file->deleted;
1754
1755 if (file->is_unmerged ||
1756 (!file->is_interesting && (added + deleted == 0))) {
1757 total_files--;
1758 continue;
1759 }
1760
1761 if (!file->is_binary) {
1762 adds += added;
1763 dels += deleted;
1764 }
1765 if (i < count)
1766 continue;
1767 if (!extra_shown)
1768 fprintf(options->file, "%s ...\n", line_prefix);
1769 extra_shown = 1;
1770 }
1771 fprintf(options->file, "%s", line_prefix);
1772 print_stat_summary(options->file, total_files, adds, dels);
1773 }
1774
1775 static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1776 {
1777 int i, adds = 0, dels = 0, total_files = data->nr;
1778
1779 if (data->nr == 0)
1780 return;
1781
1782 for (i = 0; i < data->nr; i++) {
1783 int added = data->files[i]->added;
1784 int deleted= data->files[i]->deleted;
1785
1786 if (data->files[i]->is_unmerged ||
1787 (!data->files[i]->is_interesting && (added + deleted == 0))) {
1788 total_files--;
1789 } else if (!data->files[i]->is_binary) { /* don't count bytes */
1790 adds += added;
1791 dels += deleted;
1792 }
1793 }
1794 fprintf(options->file, "%s", diff_line_prefix(options));
1795 print_stat_summary(options->file, total_files, adds, dels);
1796 }
1797
1798 static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1799 {
1800 int i;
1801
1802 if (data->nr == 0)
1803 return;
1804
1805 for (i = 0; i < data->nr; i++) {
1806 struct diffstat_file *file = data->files[i];
1807
1808 fprintf(options->file, "%s", diff_line_prefix(options));
1809
1810 if (file->is_binary)
1811 fprintf(options->file, "-\t-\t");
1812 else
1813 fprintf(options->file,
1814 "%"PRIuMAX"\t%"PRIuMAX"\t",
1815 file->added, file->deleted);
1816 if (options->line_termination) {
1817 fill_print_name(file);
1818 if (!file->is_renamed)
1819 write_name_quoted(file->name, options->file,
1820 options->line_termination);
1821 else {
1822 fputs(file->print_name, options->file);
1823 putc(options->line_termination, options->file);
1824 }
1825 } else {
1826 if (file->is_renamed) {
1827 putc('\0', options->file);
1828 write_name_quoted(file->from_name, options->file, '\0');
1829 }
1830 write_name_quoted(file->name, options->file, '\0');
1831 }
1832 }
1833 }
1834
1835 struct dirstat_file {
1836 const char *name;
1837 unsigned long changed;
1838 };
1839
1840 struct dirstat_dir {
1841 struct dirstat_file *files;
1842 int alloc, nr, permille, cumulative;
1843 };
1844
1845 static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1846 unsigned long changed, const char *base, int baselen)
1847 {
1848 unsigned long this_dir = 0;
1849 unsigned int sources = 0;
1850 const char *line_prefix = diff_line_prefix(opt);
1851
1852 while (dir->nr) {
1853 struct dirstat_file *f = dir->files;
1854 int namelen = strlen(f->name);
1855 unsigned long this;
1856 char *slash;
1857
1858 if (namelen < baselen)
1859 break;
1860 if (memcmp(f->name, base, baselen))
1861 break;
1862 slash = strchr(f->name + baselen, '/');
1863 if (slash) {
1864 int newbaselen = slash + 1 - f->name;
1865 this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
1866 sources++;
1867 } else {
1868 this = f->changed;
1869 dir->files++;
1870 dir->nr--;
1871 sources += 2;
1872 }
1873 this_dir += this;
1874 }
1875
1876 /*
1877 * We don't report dirstat's for
1878 * - the top level
1879 * - or cases where everything came from a single directory
1880 * under this directory (sources == 1).
1881 */
1882 if (baselen && sources != 1) {
1883 if (this_dir) {
1884 int permille = this_dir * 1000 / changed;
1885 if (permille >= dir->permille) {
1886 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
1887 permille / 10, permille % 10, baselen, base);
1888 if (!dir->cumulative)
1889 return 0;
1890 }
1891 }
1892 }
1893 return this_dir;
1894 }
1895
1896 static int dirstat_compare(const void *_a, const void *_b)
1897 {
1898 const struct dirstat_file *a = _a;
1899 const struct dirstat_file *b = _b;
1900 return strcmp(a->name, b->name);
1901 }
1902
1903 static void show_dirstat(struct diff_options *options)
1904 {
1905 int i;
1906 unsigned long changed;
1907 struct dirstat_dir dir;
1908 struct diff_queue_struct *q = &diff_queued_diff;
1909
1910 dir.files = NULL;
1911 dir.alloc = 0;
1912 dir.nr = 0;
1913 dir.permille = options->dirstat_permille;
1914 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1915
1916 changed = 0;
1917 for (i = 0; i < q->nr; i++) {
1918 struct diff_filepair *p = q->queue[i];
1919 const char *name;
1920 unsigned long copied, added, damage;
1921 int content_changed;
1922
1923 name = p->two->path ? p->two->path : p->one->path;
1924
1925 if (p->one->sha1_valid && p->two->sha1_valid)
1926 content_changed = hashcmp(p->one->sha1, p->two->sha1);
1927 else
1928 content_changed = 1;
1929
1930 if (!content_changed) {
1931 /*
1932 * The SHA1 has not changed, so pre-/post-content is
1933 * identical. We can therefore skip looking at the
1934 * file contents altogether.
1935 */
1936 damage = 0;
1937 goto found_damage;
1938 }
1939
1940 if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
1941 /*
1942 * In --dirstat-by-file mode, we don't really need to
1943 * look at the actual file contents at all.
1944 * The fact that the SHA1 changed is enough for us to
1945 * add this file to the list of results
1946 * (with each file contributing equal damage).
1947 */
1948 damage = 1;
1949 goto found_damage;
1950 }
1951
1952 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
1953 diff_populate_filespec(p->one, 0);
1954 diff_populate_filespec(p->two, 0);
1955 diffcore_count_changes(p->one, p->two, NULL, NULL, 0,
1956 &copied, &added);
1957 diff_free_filespec_data(p->one);
1958 diff_free_filespec_data(p->two);
1959 } else if (DIFF_FILE_VALID(p->one)) {
1960 diff_populate_filespec(p->one, CHECK_SIZE_ONLY);
1961 copied = added = 0;
1962 diff_free_filespec_data(p->one);
1963 } else if (DIFF_FILE_VALID(p->two)) {
1964 diff_populate_filespec(p->two, CHECK_SIZE_ONLY);
1965 copied = 0;
1966 added = p->two->size;
1967 diff_free_filespec_data(p->two);
1968 } else
1969 continue;
1970
1971 /*
1972 * Original minus copied is the removed material,
1973 * added is the new material. They are both damages
1974 * made to the preimage.
1975 * If the resulting damage is zero, we know that
1976 * diffcore_count_changes() considers the two entries to
1977 * be identical, but since content_changed is true, we
1978 * know that there must have been _some_ kind of change,
1979 * so we force all entries to have damage > 0.
1980 */
1981 damage = (p->one->size - copied) + added;
1982 if (!damage)
1983 damage = 1;
1984
1985 found_damage:
1986 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1987 dir.files[dir.nr].name = name;
1988 dir.files[dir.nr].changed = damage;
1989 changed += damage;
1990 dir.nr++;
1991 }
1992
1993 /* This can happen even with many files, if everything was renames */
1994 if (!changed)
1995 return;
1996
1997 /* Show all directories with more than x% of the changes */
1998 qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1999 gather_dirstat(options, &dir, changed, "", 0);
2000 }
2001
2002 static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
2003 {
2004 int i;
2005 unsigned long changed;
2006 struct dirstat_dir dir;
2007
2008 if (data->nr == 0)
2009 return;
2010
2011 dir.files = NULL;
2012 dir.alloc = 0;
2013 dir.nr = 0;
2014 dir.permille = options->dirstat_permille;
2015 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
2016
2017 changed = 0;
2018 for (i = 0; i < data->nr; i++) {
2019 struct diffstat_file *file = data->files[i];
2020 unsigned long damage = file->added + file->deleted;
2021 if (file->is_binary)
2022 /*
2023 * binary files counts bytes, not lines. Must find some
2024 * way to normalize binary bytes vs. textual lines.
2025 * The following heuristic assumes that there are 64
2026 * bytes per "line".
2027 * This is stupid and ugly, but very cheap...
2028 */
2029 damage = (damage + 63) / 64;
2030 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2031 dir.files[dir.nr].name = file->name;
2032 dir.files[dir.nr].changed = damage;
2033 changed += damage;
2034 dir.nr++;
2035 }
2036
2037 /* This can happen even with many files, if everything was renames */
2038 if (!changed)
2039 return;
2040
2041 /* Show all directories with more than x% of the changes */
2042 qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
2043 gather_dirstat(options, &dir, changed, "", 0);
2044 }
2045
2046 static void free_diffstat_info(struct diffstat_t *diffstat)
2047 {
2048 int i;
2049 for (i = 0; i < diffstat->nr; i++) {
2050 struct diffstat_file *f = diffstat->files[i];
2051 if (f->name != f->print_name)
2052 free(f->print_name);
2053 free(f->name);
2054 free(f->from_name);
2055 free(f);
2056 }
2057 free(diffstat->files);
2058 }
2059
2060 struct checkdiff_t {
2061 const char *filename;
2062 int lineno;
2063 int conflict_marker_size;
2064 struct diff_options *o;
2065 unsigned ws_rule;
2066 unsigned status;
2067 };
2068
2069 static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2070 {
2071 char firstchar;
2072 int cnt;
2073
2074 if (len < marker_size + 1)
2075 return 0;
2076 firstchar = line[0];
2077 switch (firstchar) {
2078 case '=': case '>': case '<': case '|':
2079 break;
2080 default:
2081 return 0;
2082 }
2083 for (cnt = 1; cnt < marker_size; cnt++)
2084 if (line[cnt] != firstchar)
2085 return 0;
2086 /* line[1] thru line[marker_size-1] are same as firstchar */
2087 if (len < marker_size + 1 || !isspace(line[marker_size]))
2088 return 0;
2089 return 1;
2090 }
2091
2092 static void checkdiff_consume(void *priv, char *line, unsigned long len)
2093 {
2094 struct checkdiff_t *data = priv;
2095 int marker_size = data->conflict_marker_size;
2096 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2097 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2098 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2099 char *err;
2100 const char *line_prefix;
2101
2102 assert(data->o);
2103 line_prefix = diff_line_prefix(data->o);
2104
2105 if (line[0] == '+') {
2106 unsigned bad;
2107 data->lineno++;
2108 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2109 data->status |= 1;
2110 fprintf(data->o->file,
2111 "%s%s:%d: leftover conflict marker\n",
2112 line_prefix, data->filename, data->lineno);
2113 }
2114 bad = ws_check(line + 1, len - 1, data->ws_rule);
2115 if (!bad)
2116 return;
2117 data->status |= bad;
2118 err = whitespace_error_string(bad);
2119 fprintf(data->o->file, "%s%s:%d: %s.\n",
2120 line_prefix, data->filename, data->lineno, err);
2121 free(err);
2122 emit_line(data->o, set, reset, line, 1);
2123 ws_check_emit(line + 1, len - 1, data->ws_rule,
2124 data->o->file, set, reset, ws);
2125 } else if (line[0] == ' ') {
2126 data->lineno++;
2127 } else if (line[0] == '@') {
2128 char *plus = strchr(line, '+');
2129 if (plus)
2130 data->lineno = strtol(plus, NULL, 10) - 1;
2131 else
2132 die("invalid diff");
2133 }
2134 }
2135
2136 static unsigned char *deflate_it(char *data,
2137 unsigned long size,
2138 unsigned long *result_size)
2139 {
2140 int bound;
2141 unsigned char *deflated;
2142 git_zstream stream;
2143
2144 git_deflate_init(&stream, zlib_compression_level);
2145 bound = git_deflate_bound(&stream, size);
2146 deflated = xmalloc(bound);
2147 stream.next_out = deflated;
2148 stream.avail_out = bound;
2149
2150 stream.next_in = (unsigned char *)data;
2151 stream.avail_in = size;
2152 while (git_deflate(&stream, Z_FINISH) == Z_OK)
2153 ; /* nothing */
2154 git_deflate_end(&stream);
2155 *result_size = stream.total_out;
2156 return deflated;
2157 }
2158
2159 static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two,
2160 const char *prefix)
2161 {
2162 void *cp;
2163 void *delta;
2164 void *deflated;
2165 void *data;
2166 unsigned long orig_size;
2167 unsigned long delta_size;
2168 unsigned long deflate_size;
2169 unsigned long data_size;
2170
2171 /* We could do deflated delta, or we could do just deflated two,
2172 * whichever is smaller.
2173 */
2174 delta = NULL;
2175 deflated = deflate_it(two->ptr, two->size, &deflate_size);
2176 if (one->size && two->size) {
2177 delta = diff_delta(one->ptr, one->size,
2178 two->ptr, two->size,
2179 &delta_size, deflate_size);
2180 if (delta) {
2181 void *to_free = delta;
2182 orig_size = delta_size;
2183 delta = deflate_it(delta, delta_size, &delta_size);
2184 free(to_free);
2185 }
2186 }
2187
2188 if (delta && delta_size < deflate_size) {
2189 fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2190 free(deflated);
2191 data = delta;
2192 data_size = delta_size;
2193 }
2194 else {
2195 fprintf(file, "%sliteral %lu\n", prefix, two->size);
2196 free(delta);
2197 data = deflated;
2198 data_size = deflate_size;
2199 }
2200
2201 /* emit data encoded in base85 */
2202 cp = data;
2203 while (data_size) {
2204 int bytes = (52 < data_size) ? 52 : data_size;
2205 char line[70];
2206 data_size -= bytes;
2207 if (bytes <= 26)
2208 line[0] = bytes + 'A' - 1;
2209 else
2210 line[0] = bytes - 26 + 'a' - 1;
2211 encode_85(line + 1, cp, bytes);
2212 cp = (char *) cp + bytes;
2213 fprintf(file, "%s", prefix);
2214 fputs(line, file);
2215 fputc('\n', file);
2216 }
2217 fprintf(file, "%s\n", prefix);
2218 free(data);
2219 }
2220
2221 static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two,
2222 const char *prefix)
2223 {
2224 fprintf(file, "%sGIT binary patch\n", prefix);
2225 emit_binary_diff_body(file, one, two, prefix);
2226 emit_binary_diff_body(file, two, one, prefix);
2227 }
2228
2229 int diff_filespec_is_binary(struct diff_filespec *one)
2230 {
2231 if (one->is_binary == -1) {
2232 diff_filespec_load_driver(one);
2233 if (one->driver->binary != -1)
2234 one->is_binary = one->driver->binary;
2235 else {
2236 if (!one->data && DIFF_FILE_VALID(one))
2237 diff_populate_filespec(one, CHECK_BINARY);
2238 if (one->is_binary == -1 && one->data)
2239 one->is_binary = buffer_is_binary(one->data,
2240 one->size);
2241 if (one->is_binary == -1)
2242 one->is_binary = 0;
2243 }
2244 }
2245 return one->is_binary;
2246 }
2247
2248 static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2249 {
2250 diff_filespec_load_driver(one);
2251 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2252 }
2253
2254 void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2255 {
2256 if (!options->a_prefix)
2257 options->a_prefix = a;
2258 if (!options->b_prefix)
2259 options->b_prefix = b;
2260 }
2261
2262 struct userdiff_driver *get_textconv(struct diff_filespec *one)
2263 {
2264 if (!DIFF_FILE_VALID(one))
2265 return NULL;
2266
2267 diff_filespec_load_driver(one);
2268 return userdiff_get_textconv(one->driver);
2269 }
2270
2271 static void builtin_diff(const char *name_a,
2272 const char *name_b,
2273 struct diff_filespec *one,
2274 struct diff_filespec *two,
2275 const char *xfrm_msg,
2276 int must_show_header,
2277 struct diff_options *o,
2278 int complete_rewrite)
2279 {
2280 mmfile_t mf1, mf2;
2281 const char *lbl[2];
2282 char *a_one, *b_two;
2283 const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
2284 const char *reset = diff_get_color_opt(o, DIFF_RESET);
2285 const char *a_prefix, *b_prefix;
2286 struct userdiff_driver *textconv_one = NULL;
2287 struct userdiff_driver *textconv_two = NULL;
2288 struct strbuf header = STRBUF_INIT;
2289 const char *line_prefix = diff_line_prefix(o);
2290
2291 if (DIFF_OPT_TST(o, SUBMODULE_LOG) &&
2292 (!one->mode || S_ISGITLINK(one->mode)) &&
2293 (!two->mode || S_ISGITLINK(two->mode))) {
2294 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2295 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2296 show_submodule_summary(o->file, one->path ? one->path : two->path,
2297 line_prefix,
2298 one->sha1, two->sha1, two->dirty_submodule,
2299 meta, del, add, reset);
2300 return;
2301 }
2302
2303 if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2304 textconv_one = get_textconv(one);
2305 textconv_two = get_textconv(two);
2306 }
2307
2308 diff_set_mnemonic_prefix(o, "a/", "b/");
2309 if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2310 a_prefix = o->b_prefix;
2311 b_prefix = o->a_prefix;
2312 } else {
2313 a_prefix = o->a_prefix;
2314 b_prefix = o->b_prefix;
2315 }
2316
2317 /* Never use a non-valid filename anywhere if at all possible */
2318 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2319 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2320
2321 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2322 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2323 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2324 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2325 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
2326 if (lbl[0][0] == '/') {
2327 /* /dev/null */
2328 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
2329 if (xfrm_msg)
2330 strbuf_addstr(&header, xfrm_msg);
2331 must_show_header = 1;
2332 }
2333 else if (lbl[1][0] == '/') {
2334 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
2335 if (xfrm_msg)
2336 strbuf_addstr(&header, xfrm_msg);
2337 must_show_header = 1;
2338 }
2339 else {
2340 if (one->mode != two->mode) {
2341 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
2342 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
2343 must_show_header = 1;
2344 }
2345 if (xfrm_msg)
2346 strbuf_addstr(&header, xfrm_msg);
2347
2348 /*
2349 * we do not run diff between different kind
2350 * of objects.
2351 */
2352 if ((one->mode ^ two->mode) & S_IFMT)
2353 goto free_ab_and_return;
2354 if (complete_rewrite &&
2355 (textconv_one || !diff_filespec_is_binary(one)) &&
2356 (textconv_two || !diff_filespec_is_binary(two))) {
2357 fprintf(o->file, "%s", header.buf);
2358 strbuf_reset(&header);
2359 emit_rewrite_diff(name_a, name_b, one, two,
2360 textconv_one, textconv_two, o);
2361 o->found_changes = 1;
2362 goto free_ab_and_return;
2363 }
2364 }
2365
2366 if (o->irreversible_delete && lbl[1][0] == '/') {
2367 fprintf(o->file, "%s", header.buf);
2368 strbuf_reset(&header);
2369 goto free_ab_and_return;
2370 } else if (!DIFF_OPT_TST(o, TEXT) &&
2371 ( (!textconv_one && diff_filespec_is_binary(one)) ||
2372 (!textconv_two && diff_filespec_is_binary(two)) )) {
2373 if (!one->data && !two->data &&
2374 S_ISREG(one->mode) && S_ISREG(two->mode) &&
2375 !DIFF_OPT_TST(o, BINARY)) {
2376 if (!hashcmp(one->sha1, two->sha1)) {
2377 if (must_show_header)
2378 fprintf(o->file, "%s", header.buf);
2379 goto free_ab_and_return;
2380 }
2381 fprintf(o->file, "%s", header.buf);
2382 fprintf(o->file, "%sBinary files %s and %s differ\n",
2383 line_prefix, lbl[0], lbl[1]);
2384 goto free_ab_and_return;
2385 }
2386 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2387 die("unable to read files to diff");
2388 /* Quite common confusing case */
2389 if (mf1.size == mf2.size &&
2390 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2391 if (must_show_header)
2392 fprintf(o->file, "%s", header.buf);
2393 goto free_ab_and_return;
2394 }
2395 fprintf(o->file, "%s", header.buf);
2396 strbuf_reset(&header);
2397 if (DIFF_OPT_TST(o, BINARY))
2398 emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2399 else
2400 fprintf(o->file, "%sBinary files %s and %s differ\n",
2401 line_prefix, lbl[0], lbl[1]);
2402 o->found_changes = 1;
2403 } else {
2404 /* Crazy xdl interfaces.. */
2405 const char *diffopts = getenv("GIT_DIFF_OPTS");
2406 const char *v;
2407 xpparam_t xpp;
2408 xdemitconf_t xecfg;
2409 struct emit_callback ecbdata;
2410 const struct userdiff_funcname *pe;
2411
2412 if (must_show_header) {
2413 fprintf(o->file, "%s", header.buf);
2414 strbuf_reset(&header);
2415 }
2416
2417 mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2418 mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2419
2420 pe = diff_funcname_pattern(one);
2421 if (!pe)
2422 pe = diff_funcname_pattern(two);
2423
2424 memset(&xpp, 0, sizeof(xpp));
2425 memset(&xecfg, 0, sizeof(xecfg));
2426 memset(&ecbdata, 0, sizeof(ecbdata));
2427 ecbdata.label_path = lbl;
2428 ecbdata.color_diff = want_color(o->use_color);
2429 ecbdata.found_changesp = &o->found_changes;
2430 ecbdata.ws_rule = whitespace_rule(name_b);
2431 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2432 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2433 ecbdata.opt = o;
2434 ecbdata.header = header.len ? &header : NULL;
2435 xpp.flags = o->xdl_opts;
2436 xecfg.ctxlen = o->context;
2437 xecfg.interhunkctxlen = o->interhunkcontext;
2438 xecfg.flags = XDL_EMIT_FUNCNAMES;
2439 if (DIFF_OPT_TST(o, FUNCCONTEXT))
2440 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2441 if (pe)
2442 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2443 if (!diffopts)
2444 ;
2445 else if (skip_prefix(diffopts, "--unified=", &v))
2446 xecfg.ctxlen = strtoul(v, NULL, 10);
2447 else if (skip_prefix(diffopts, "-u", &v))
2448 xecfg.ctxlen = strtoul(v, NULL, 10);
2449 if (o->word_diff)
2450 init_diff_words_data(&ecbdata, o, one, two);
2451 xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2452 &xpp, &xecfg);
2453 if (o->word_diff)
2454 free_diff_words_data(&ecbdata);
2455 if (textconv_one)
2456 free(mf1.ptr);
2457 if (textconv_two)
2458 free(mf2.ptr);
2459 xdiff_clear_find_func(&xecfg);
2460 }
2461
2462 free_ab_and_return:
2463 strbuf_release(&header);
2464 diff_free_filespec_data(one);
2465 diff_free_filespec_data(two);
2466 free(a_one);
2467 free(b_two);
2468 return;
2469 }
2470
2471 static void builtin_diffstat(const char *name_a, const char *name_b,
2472 struct diff_filespec *one,
2473 struct diff_filespec *two,
2474 struct diffstat_t *diffstat,
2475 struct diff_options *o,
2476 struct diff_filepair *p)
2477 {
2478 mmfile_t mf1, mf2;
2479 struct diffstat_file *data;
2480 int same_contents;
2481 int complete_rewrite = 0;
2482
2483 if (!DIFF_PAIR_UNMERGED(p)) {
2484 if (p->status == DIFF_STATUS_MODIFIED && p->score)
2485 complete_rewrite = 1;
2486 }
2487
2488 data = diffstat_add(diffstat, name_a, name_b);
2489 data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
2490
2491 if (!one || !two) {
2492 data->is_unmerged = 1;
2493 return;
2494 }
2495
2496 same_contents = !hashcmp(one->sha1, two->sha1);
2497
2498 if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2499 data->is_binary = 1;
2500 if (same_contents) {
2501 data->added = 0;
2502 data->deleted = 0;
2503 } else {
2504 data->added = diff_filespec_size(two);
2505 data->deleted = diff_filespec_size(one);
2506 }
2507 }
2508
2509 else if (complete_rewrite) {
2510 diff_populate_filespec(one, 0);
2511 diff_populate_filespec(two, 0);
2512 data->deleted = count_lines(one->data, one->size);
2513 data->added = count_lines(two->data, two->size);
2514 }
2515
2516 else if (!same_contents) {
2517 /* Crazy xdl interfaces.. */
2518 xpparam_t xpp;
2519 xdemitconf_t xecfg;
2520
2521 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2522 die("unable to read files to diff");
2523
2524 memset(&xpp, 0, sizeof(xpp));
2525 memset(&xecfg, 0, sizeof(xecfg));
2526 xpp.flags = o->xdl_opts;
2527 xecfg.ctxlen = o->context;
2528 xecfg.interhunkctxlen = o->interhunkcontext;
2529 xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2530 &xpp, &xecfg);
2531 }
2532
2533 diff_free_filespec_data(one);
2534 diff_free_filespec_data(two);
2535 }
2536
2537 static void builtin_checkdiff(const char *name_a, const char *name_b,
2538 const char *attr_path,
2539 struct diff_filespec *one,
2540 struct diff_filespec *two,
2541 struct diff_options *o)
2542 {
2543 mmfile_t mf1, mf2;
2544 struct checkdiff_t data;
2545
2546 if (!two)
2547 return;
2548
2549 memset(&data, 0, sizeof(data));
2550 data.filename = name_b ? name_b : name_a;
2551 data.lineno = 0;
2552 data.o = o;
2553 data.ws_rule = whitespace_rule(attr_path);
2554 data.conflict_marker_size = ll_merge_marker_size(attr_path);
2555
2556 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2557 die("unable to read files to diff");
2558
2559 /*
2560 * All the other codepaths check both sides, but not checking
2561 * the "old" side here is deliberate. We are checking the newly
2562 * introduced changes, and as long as the "new" side is text, we
2563 * can and should check what it introduces.
2564 */
2565 if (diff_filespec_is_binary(two))
2566 goto free_and_return;
2567 else {
2568 /* Crazy xdl interfaces.. */
2569 xpparam_t xpp;
2570 xdemitconf_t xecfg;
2571
2572 memset(&xpp, 0, sizeof(xpp));
2573 memset(&xecfg, 0, sizeof(xecfg));
2574 xecfg.ctxlen = 1; /* at least one context line */
2575 xpp.flags = 0;
2576 xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2577 &xpp, &xecfg);
2578
2579 if (data.ws_rule & WS_BLANK_AT_EOF) {
2580 struct emit_callback ecbdata;
2581 int blank_at_eof;
2582
2583 ecbdata.ws_rule = data.ws_rule;
2584 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2585 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2586
2587 if (blank_at_eof) {
2588 static char *err;
2589 if (!err)
2590 err = whitespace_error_string(WS_BLANK_AT_EOF);
2591 fprintf(o->file, "%s:%d: %s.\n",
2592 data.filename, blank_at_eof, err);
2593 data.status = 1; /* report errors */
2594 }
2595 }
2596 }
2597 free_and_return:
2598 diff_free_filespec_data(one);
2599 diff_free_filespec_data(two);
2600 if (data.status)
2601 DIFF_OPT_SET(o, CHECK_FAILED);
2602 }
2603
2604 struct diff_filespec *alloc_filespec(const char *path)
2605 {
2606 int namelen = strlen(path);
2607 struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
2608
2609 memset(spec, 0, sizeof(*spec));
2610 spec->path = (char *)(spec + 1);
2611 memcpy(spec->path, path, namelen+1);
2612 spec->count = 1;
2613 spec->is_binary = -1;
2614 return spec;
2615 }
2616
2617 void free_filespec(struct diff_filespec *spec)
2618 {
2619 if (!--spec->count) {
2620 diff_free_filespec_data(spec);
2621 free(spec);
2622 }
2623 }
2624
2625 void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
2626 int sha1_valid, unsigned short mode)
2627 {
2628 if (mode) {
2629 spec->mode = canon_mode(mode);
2630 hashcpy(spec->sha1, sha1);
2631 spec->sha1_valid = sha1_valid;
2632 }
2633 }
2634
2635 /*
2636 * Given a name and sha1 pair, if the index tells us the file in
2637 * the work tree has that object contents, return true, so that
2638 * prepare_temp_file() does not have to inflate and extract.
2639 */
2640 static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
2641 {
2642 const struct cache_entry *ce;
2643 struct stat st;
2644 int pos, len;
2645
2646 /*
2647 * We do not read the cache ourselves here, because the
2648 * benchmark with my previous version that always reads cache
2649 * shows that it makes things worse for diff-tree comparing
2650 * two linux-2.6 kernel trees in an already checked out work
2651 * tree. This is because most diff-tree comparisons deal with
2652 * only a small number of files, while reading the cache is
2653 * expensive for a large project, and its cost outweighs the
2654 * savings we get by not inflating the object to a temporary
2655 * file. Practically, this code only helps when we are used
2656 * by diff-cache --cached, which does read the cache before
2657 * calling us.
2658 */
2659 if (!active_cache)
2660 return 0;
2661
2662 /* We want to avoid the working directory if our caller
2663 * doesn't need the data in a normal file, this system
2664 * is rather slow with its stat/open/mmap/close syscalls,
2665 * and the object is contained in a pack file. The pack
2666 * is probably already open and will be faster to obtain
2667 * the data through than the working directory. Loose
2668 * objects however would tend to be slower as they need
2669 * to be individually opened and inflated.
2670 */
2671 if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1))
2672 return 0;
2673
2674 len = strlen(name);
2675 pos = cache_name_pos(name, len);
2676 if (pos < 0)
2677 return 0;
2678 ce = active_cache[pos];
2679
2680 /*
2681 * This is not the sha1 we are looking for, or
2682 * unreusable because it is not a regular file.
2683 */
2684 if (hashcmp(sha1, ce->sha1) || !S_ISREG(ce->ce_mode))
2685 return 0;
2686
2687 /*
2688 * If ce is marked as "assume unchanged", there is no
2689 * guarantee that work tree matches what we are looking for.
2690 */
2691 if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2692 return 0;
2693
2694 /*
2695 * If ce matches the file in the work tree, we can reuse it.
2696 */
2697 if (ce_uptodate(ce) ||
2698 (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2699 return 1;
2700
2701 return 0;
2702 }
2703
2704 static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2705 {
2706 int len;
2707 char *data = xmalloc(100), *dirty = "";
2708
2709 /* Are we looking at the work tree? */
2710 if (s->dirty_submodule)
2711 dirty = "-dirty";
2712
2713 len = snprintf(data, 100,
2714 "Subproject commit %s%s\n", sha1_to_hex(s->sha1), dirty);
2715 s->data = data;
2716 s->size = len;
2717 s->should_free = 1;
2718 if (size_only) {
2719 s->data = NULL;
2720 free(data);
2721 }
2722 return 0;
2723 }
2724
2725 /*
2726 * While doing rename detection and pickaxe operation, we may need to
2727 * grab the data for the blob (or file) for our own in-core comparison.
2728 * diff_filespec has data and size fields for this purpose.
2729 */
2730 int diff_populate_filespec(struct diff_filespec *s, unsigned int flags)
2731 {
2732 int size_only = flags & CHECK_SIZE_ONLY;
2733 int err = 0;
2734 /*
2735 * demote FAIL to WARN to allow inspecting the situation
2736 * instead of refusing.
2737 */
2738 enum safe_crlf crlf_warn = (safe_crlf == SAFE_CRLF_FAIL
2739 ? SAFE_CRLF_WARN
2740 : safe_crlf);
2741
2742 if (!DIFF_FILE_VALID(s))
2743 die("internal error: asking to populate invalid file.");
2744 if (S_ISDIR(s->mode))
2745 return -1;
2746
2747 if (s->data)
2748 return 0;
2749
2750 if (size_only && 0 < s->size)
2751 return 0;
2752
2753 if (S_ISGITLINK(s->mode))
2754 return diff_populate_gitlink(s, size_only);
2755
2756 if (!s->sha1_valid ||
2757 reuse_worktree_file(s->path, s->sha1, 0)) {
2758 struct strbuf buf = STRBUF_INIT;
2759 struct stat st;
2760 int fd;
2761
2762 if (lstat(s->path, &st) < 0) {
2763 if (errno == ENOENT) {
2764 err_empty:
2765 err = -1;
2766 empty:
2767 s->data = (char *)"";
2768 s->size = 0;
2769 return err;
2770 }
2771 }
2772 s->size = xsize_t(st.st_size);
2773 if (!s->size)
2774 goto empty;
2775 if (S_ISLNK(st.st_mode)) {
2776 struct strbuf sb = STRBUF_INIT;
2777
2778 if (strbuf_readlink(&sb, s->path, s->size))
2779 goto err_empty;
2780 s->size = sb.len;
2781 s->data = strbuf_detach(&sb, NULL);
2782 s->should_free = 1;
2783 return 0;
2784 }
2785 if (size_only)
2786 return 0;
2787 if ((flags & CHECK_BINARY) &&
2788 s->size > big_file_threshold && s->is_binary == -1) {
2789 s->is_binary = 1;
2790 return 0;
2791 }
2792 fd = open(s->path, O_RDONLY);
2793 if (fd < 0)
2794 goto err_empty;
2795 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2796 close(fd);
2797 s->should_munmap = 1;
2798
2799 /*
2800 * Convert from working tree format to canonical git format
2801 */
2802 if (convert_to_git(s->path, s->data, s->size, &buf, crlf_warn)) {
2803 size_t size = 0;
2804 munmap(s->data, s->size);
2805 s->should_munmap = 0;
2806 s->data = strbuf_detach(&buf, &size);
2807 s->size = size;
2808 s->should_free = 1;
2809 }
2810 }
2811 else {
2812 enum object_type type;
2813 if (size_only || (flags & CHECK_BINARY)) {
2814 type = sha1_object_info(s->sha1, &s->size);
2815 if (type < 0)
2816 die("unable to read %s", sha1_to_hex(s->sha1));
2817 if (size_only)
2818 return 0;
2819 if (s->size > big_file_threshold && s->is_binary == -1) {
2820 s->is_binary = 1;
2821 return 0;
2822 }
2823 }
2824 s->data = read_sha1_file(s->sha1, &type, &s->size);
2825 if (!s->data)
2826 die("unable to read %s", sha1_to_hex(s->sha1));
2827 s->should_free = 1;
2828 }
2829 return 0;
2830 }
2831
2832 void diff_free_filespec_blob(struct diff_filespec *s)
2833 {
2834 if (s->should_free)
2835 free(s->data);
2836 else if (s->should_munmap)
2837 munmap(s->data, s->size);
2838
2839 if (s->should_free || s->should_munmap) {
2840 s->should_free = s->should_munmap = 0;
2841 s->data = NULL;
2842 }
2843 }
2844
2845 void diff_free_filespec_data(struct diff_filespec *s)
2846 {
2847 diff_free_filespec_blob(s);
2848 free(s->cnt_data);
2849 s->cnt_data = NULL;
2850 }
2851
2852 static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2853 void *blob,
2854 unsigned long size,
2855 const unsigned char *sha1,
2856 int mode)
2857 {
2858 int fd;
2859 struct strbuf buf = STRBUF_INIT;
2860 struct strbuf template = STRBUF_INIT;
2861 char *path_dup = xstrdup(path);
2862 const char *base = basename(path_dup);
2863
2864 /* Generate "XXXXXX_basename.ext" */
2865 strbuf_addstr(&template, "XXXXXX_");
2866 strbuf_addstr(&template, base);
2867
2868 fd = mks_tempfile_ts(&temp->tempfile, template.buf, strlen(base) + 1);
2869 if (fd < 0)
2870 die_errno("unable to create temp-file");
2871 if (convert_to_working_tree(path,
2872 (const char *)blob, (size_t)size, &buf)) {
2873 blob = buf.buf;
2874 size = buf.len;
2875 }
2876 if (write_in_full(fd, blob, size) != size)
2877 die_errno("unable to write temp-file");
2878 close_tempfile(&temp->tempfile);
2879 temp->name = get_tempfile_path(&temp->tempfile);
2880 strcpy(temp->hex, sha1_to_hex(sha1));
2881 temp->hex[40] = 0;
2882 sprintf(temp->mode, "%06o", mode);
2883 strbuf_release(&buf);
2884 strbuf_release(&template);
2885 free(path_dup);
2886 }
2887
2888 static struct diff_tempfile *prepare_temp_file(const char *name,
2889 struct diff_filespec *one)
2890 {
2891 struct diff_tempfile *temp = claim_diff_tempfile();
2892
2893 if (!DIFF_FILE_VALID(one)) {
2894 not_a_valid_file:
2895 /* A '-' entry produces this for file-2, and
2896 * a '+' entry produces this for file-1.
2897 */
2898 temp->name = "/dev/null";
2899 strcpy(temp->hex, ".");
2900 strcpy(temp->mode, ".");
2901 return temp;
2902 }
2903
2904 if (!S_ISGITLINK(one->mode) &&
2905 (!one->sha1_valid ||
2906 reuse_worktree_file(name, one->sha1, 1))) {
2907 struct stat st;
2908 if (lstat(name, &st) < 0) {
2909 if (errno == ENOENT)
2910 goto not_a_valid_file;
2911 die_errno("stat(%s)", name);
2912 }
2913 if (S_ISLNK(st.st_mode)) {
2914 struct strbuf sb = STRBUF_INIT;
2915 if (strbuf_readlink(&sb, name, st.st_size) < 0)
2916 die_errno("readlink(%s)", name);
2917 prep_temp_blob(name, temp, sb.buf, sb.len,
2918 (one->sha1_valid ?
2919 one->sha1 : null_sha1),
2920 (one->sha1_valid ?
2921 one->mode : S_IFLNK));
2922 strbuf_release(&sb);
2923 }
2924 else {
2925 /* we can borrow from the file in the work tree */
2926 temp->name = name;
2927 if (!one->sha1_valid)
2928 strcpy(temp->hex, sha1_to_hex(null_sha1));
2929 else
2930 strcpy(temp->hex, sha1_to_hex(one->sha1));
2931 /* Even though we may sometimes borrow the
2932 * contents from the work tree, we always want
2933 * one->mode. mode is trustworthy even when
2934 * !(one->sha1_valid), as long as
2935 * DIFF_FILE_VALID(one).
2936 */
2937 sprintf(temp->mode, "%06o", one->mode);
2938 }
2939 return temp;
2940 }
2941 else {
2942 if (diff_populate_filespec(one, 0))
2943 die("cannot read data blob for %s", one->path);
2944 prep_temp_blob(name, temp, one->data, one->size,
2945 one->sha1, one->mode);
2946 }
2947 return temp;
2948 }
2949
2950 static void add_external_diff_name(struct argv_array *argv,
2951 const char *name,
2952 struct diff_filespec *df)
2953 {
2954 struct diff_tempfile *temp = prepare_temp_file(name, df);
2955 argv_array_push(argv, temp->name);
2956 argv_array_push(argv, temp->hex);
2957 argv_array_push(argv, temp->mode);
2958 }
2959
2960 /* An external diff command takes:
2961 *
2962 * diff-cmd name infile1 infile1-sha1 infile1-mode \
2963 * infile2 infile2-sha1 infile2-mode [ rename-to ]
2964 *
2965 */
2966 static void run_external_diff(const char *pgm,
2967 const char *name,
2968 const char *other,
2969 struct diff_filespec *one,
2970 struct diff_filespec *two,
2971 const char *xfrm_msg,
2972 int complete_rewrite,
2973 struct diff_options *o)
2974 {
2975 struct argv_array argv = ARGV_ARRAY_INIT;
2976 struct argv_array env = ARGV_ARRAY_INIT;
2977 struct diff_queue_struct *q = &diff_queued_diff;
2978
2979 argv_array_push(&argv, pgm);
2980 argv_array_push(&argv, name);
2981
2982 if (one && two) {
2983 add_external_diff_name(&argv, name, one);
2984 if (!other)
2985 add_external_diff_name(&argv, name, two);
2986 else {
2987 add_external_diff_name(&argv, other, two);
2988 argv_array_push(&argv, other);
2989 argv_array_push(&argv, xfrm_msg);
2990 }
2991 }
2992
2993 argv_array_pushf(&env, "GIT_DIFF_PATH_COUNTER=%d", ++o->diff_path_counter);
2994 argv_array_pushf(&env, "GIT_DIFF_PATH_TOTAL=%d", q->nr);
2995
2996 if (run_command_v_opt_cd_env(argv.argv, RUN_USING_SHELL, NULL, env.argv))
2997 die(_("external diff died, stopping at %s"), name);
2998
2999 remove_tempfile();
3000 argv_array_clear(&argv);
3001 argv_array_clear(&env);
3002 }
3003
3004 static int similarity_index(struct diff_filepair *p)
3005 {
3006 return p->score * 100 / MAX_SCORE;
3007 }
3008
3009 static void fill_metainfo(struct strbuf *msg,
3010 const char *name,
3011 const char *other,
3012 struct diff_filespec *one,
3013 struct diff_filespec *two,
3014 struct diff_options *o,
3015 struct diff_filepair *p,
3016 int *must_show_header,
3017 int use_color)
3018 {
3019 const char *set = diff_get_color(use_color, DIFF_METAINFO);
3020 const char *reset = diff_get_color(use_color, DIFF_RESET);
3021 const char *line_prefix = diff_line_prefix(o);
3022
3023 *must_show_header = 1;
3024 strbuf_init(msg, PATH_MAX * 2 + 300);
3025 switch (p->status) {
3026 case DIFF_STATUS_COPIED:
3027 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3028 line_prefix, set, similarity_index(p));
3029 strbuf_addf(msg, "%s\n%s%scopy from ",
3030 reset, line_prefix, set);
3031 quote_c_style(name, msg, NULL, 0);
3032 strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
3033 quote_c_style(other, msg, NULL, 0);
3034 strbuf_addf(msg, "%s\n", reset);
3035 break;
3036 case DIFF_STATUS_RENAMED:
3037 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3038 line_prefix, set, similarity_index(p));
3039 strbuf_addf(msg, "%s\n%s%srename from ",
3040 reset, line_prefix, set);
3041 quote_c_style(name, msg, NULL, 0);
3042 strbuf_addf(msg, "%s\n%s%srename to ",
3043 reset, line_prefix, set);
3044 quote_c_style(other, msg, NULL, 0);
3045 strbuf_addf(msg, "%s\n", reset);
3046 break;
3047 case DIFF_STATUS_MODIFIED:
3048 if (p->score) {
3049 strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
3050 line_prefix,
3051 set, similarity_index(p), reset);
3052 break;
3053 }
3054 /* fallthru */
3055 default:
3056 *must_show_header = 0;
3057 }
3058 if (one && two && hashcmp(one->sha1, two->sha1)) {
3059 int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
3060
3061 if (DIFF_OPT_TST(o, BINARY)) {
3062 mmfile_t mf;
3063 if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
3064 (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
3065 abbrev = 40;
3066 }
3067 strbuf_addf(msg, "%s%sindex %s..", line_prefix, set,
3068 find_unique_abbrev(one->sha1, abbrev));
3069 strbuf_addstr(msg, find_unique_abbrev(two->sha1, abbrev));
3070 if (one->mode == two->mode)
3071 strbuf_addf(msg, " %06o", one->mode);
3072 strbuf_addf(msg, "%s\n", reset);
3073 }
3074 }
3075
3076 static void run_diff_cmd(const char *pgm,
3077 const char *name,
3078 const char *other,
3079 const char *attr_path,
3080 struct diff_filespec *one,
3081 struct diff_filespec *two,
3082 struct strbuf *msg,
3083 struct diff_options *o,
3084 struct diff_filepair *p)
3085 {
3086 const char *xfrm_msg = NULL;
3087 int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
3088 int must_show_header = 0;
3089
3090
3091 if (DIFF_OPT_TST(o, ALLOW_EXTERNAL)) {
3092 struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
3093 if (drv && drv->external)
3094 pgm = drv->external;
3095 }
3096
3097 if (msg) {
3098 /*
3099 * don't use colors when the header is intended for an
3100 * external diff driver
3101 */
3102 fill_metainfo(msg, name, other, one, two, o, p,
3103 &must_show_header,
3104 want_color(o->use_color) && !pgm);
3105 xfrm_msg = msg->len ? msg->buf : NULL;
3106 }
3107
3108 if (pgm) {
3109 run_external_diff(pgm, name, other, one, two, xfrm_msg,
3110 complete_rewrite, o);
3111 return;
3112 }
3113 if (one && two)
3114 builtin_diff(name, other ? other : name,
3115 one, two, xfrm_msg, must_show_header,
3116 o, complete_rewrite);
3117 else
3118 fprintf(o->file, "* Unmerged path %s\n", name);
3119 }
3120
3121 static void diff_fill_sha1_info(struct diff_filespec *one)
3122 {
3123 if (DIFF_FILE_VALID(one)) {
3124 if (!one->sha1_valid) {
3125 struct stat st;
3126 if (one->is_stdin) {
3127 hashcpy(one->sha1, null_sha1);
3128 return;
3129 }
3130 if (lstat(one->path, &st) < 0)
3131 die_errno("stat '%s'", one->path);
3132 if (index_path(one->sha1, one->path, &st, 0))
3133 die("cannot hash %s", one->path);
3134 }
3135 }
3136 else
3137 hashclr(one->sha1);
3138 }
3139
3140 static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
3141 {
3142 /* Strip the prefix but do not molest /dev/null and absolute paths */
3143 if (*namep && **namep != '/') {
3144 *namep += prefix_length;
3145 if (**namep == '/')
3146 ++*namep;
3147 }
3148 if (*otherp && **otherp != '/') {
3149 *otherp += prefix_length;
3150 if (**otherp == '/')
3151 ++*otherp;
3152 }
3153 }
3154
3155 static void run_diff(struct diff_filepair *p, struct diff_options *o)
3156 {
3157 const char *pgm = external_diff();
3158 struct strbuf msg;
3159 struct diff_filespec *one = p->one;
3160 struct diff_filespec *two = p->two;
3161 const char *name;
3162 const char *other;
3163 const char *attr_path;
3164
3165 name = p->one->path;
3166 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3167 attr_path = name;
3168 if (o->prefix_length)
3169 strip_prefix(o->prefix_length, &name, &other);
3170
3171 if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
3172 pgm = NULL;
3173
3174 if (DIFF_PAIR_UNMERGED(p)) {
3175 run_diff_cmd(pgm, name, NULL, attr_path,
3176 NULL, NULL, NULL, o, p);
3177 return;
3178 }
3179
3180 diff_fill_sha1_info(one);
3181 diff_fill_sha1_info(two);
3182
3183 if (!pgm &&
3184 DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3185 (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3186 /*
3187 * a filepair that changes between file and symlink
3188 * needs to be split into deletion and creation.
3189 */
3190 struct diff_filespec *null = alloc_filespec(two->path);
3191 run_diff_cmd(NULL, name, other, attr_path,
3192 one, null, &msg, o, p);
3193 free(null);
3194 strbuf_release(&msg);
3195
3196 null = alloc_filespec(one->path);
3197 run_diff_cmd(NULL, name, other, attr_path,
3198 null, two, &msg, o, p);
3199 free(null);
3200 }
3201 else
3202 run_diff_cmd(pgm, name, other, attr_path,
3203 one, two, &msg, o, p);
3204
3205 strbuf_release(&msg);
3206 }
3207
3208 static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3209 struct diffstat_t *diffstat)
3210 {
3211 const char *name;
3212 const char *other;
3213
3214 if (DIFF_PAIR_UNMERGED(p)) {
3215 /* unmerged */
3216 builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, p);
3217 return;
3218 }
3219
3220 name = p->one->path;
3221 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3222
3223 if (o->prefix_length)
3224 strip_prefix(o->prefix_length, &name, &other);
3225
3226 diff_fill_sha1_info(p->one);
3227 diff_fill_sha1_info(p->two);
3228
3229 builtin_diffstat(name, other, p->one, p->two, diffstat, o, p);
3230 }
3231
3232 static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3233 {
3234 const char *name;
3235 const char *other;
3236 const char *attr_path;
3237
3238 if (DIFF_PAIR_UNMERGED(p)) {
3239 /* unmerged */
3240 return;
3241 }
3242
3243 name = p->one->path;
3244 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3245 attr_path = other ? other : name;
3246
3247 if (o->prefix_length)
3248 strip_prefix(o->prefix_length, &name, &other);
3249
3250 diff_fill_sha1_info(p->one);
3251 diff_fill_sha1_info(p->two);
3252
3253 builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3254 }
3255
3256 void diff_setup(struct diff_options *options)
3257 {
3258 memcpy(options, &default_diff_options, sizeof(*options));
3259
3260 options->file = stdout;
3261
3262 options->line_termination = '\n';
3263 options->break_opt = -1;
3264 options->rename_limit = -1;
3265 options->dirstat_permille = diff_dirstat_permille_default;
3266 options->context = diff_context_default;
3267 options->ws_error_highlight = WSEH_NEW;
3268 DIFF_OPT_SET(options, RENAME_EMPTY);
3269
3270 /* pathchange left =NULL by default */
3271 options->change = diff_change;
3272 options->add_remove = diff_addremove;
3273 options->use_color = diff_use_color_default;
3274 options->detect_rename = diff_detect_rename_default;
3275 options->xdl_opts |= diff_algorithm;
3276
3277 options->orderfile = diff_order_file_cfg;
3278
3279 if (diff_no_prefix) {
3280 options->a_prefix = options->b_prefix = "";
3281 } else if (!diff_mnemonic_prefix) {
3282 options->a_prefix = "a/";
3283 options->b_prefix = "b/";
3284 }
3285 }
3286
3287 void diff_setup_done(struct diff_options *options)
3288 {
3289 int count = 0;
3290
3291 if (options->set_default)
3292 options->set_default(options);
3293
3294 if (options->output_format & DIFF_FORMAT_NAME)
3295 count++;
3296 if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3297 count++;
3298 if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3299 count++;
3300 if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3301 count++;
3302 if (count > 1)
3303 die("--name-only, --name-status, --check and -s are mutually exclusive");
3304
3305 /*
3306 * Most of the time we can say "there are changes"
3307 * only by checking if there are changed paths, but
3308 * --ignore-whitespace* options force us to look
3309 * inside contents.
3310 */
3311
3312 if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3313 DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3314 DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3315 DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3316 else
3317 DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3318
3319 if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3320 options->detect_rename = DIFF_DETECT_COPY;
3321
3322 if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3323 options->prefix = NULL;
3324 if (options->prefix)
3325 options->prefix_length = strlen(options->prefix);
3326 else
3327 options->prefix_length = 0;
3328
3329 if (options->output_format & (DIFF_FORMAT_NAME |
3330 DIFF_FORMAT_NAME_STATUS |
3331 DIFF_FORMAT_CHECKDIFF |
3332 DIFF_FORMAT_NO_OUTPUT))
3333 options->output_format &= ~(DIFF_FORMAT_RAW |
3334 DIFF_FORMAT_NUMSTAT |
3335 DIFF_FORMAT_DIFFSTAT |
3336 DIFF_FORMAT_SHORTSTAT |
3337 DIFF_FORMAT_DIRSTAT |
3338 DIFF_FORMAT_SUMMARY |
3339 DIFF_FORMAT_PATCH);
3340
3341 /*
3342 * These cases always need recursive; we do not drop caller-supplied
3343 * recursive bits for other formats here.
3344 */
3345 if (options->output_format & (DIFF_FORMAT_PATCH |
3346 DIFF_FORMAT_NUMSTAT |
3347 DIFF_FORMAT_DIFFSTAT |
3348 DIFF_FORMAT_SHORTSTAT |
3349 DIFF_FORMAT_DIRSTAT |
3350 DIFF_FORMAT_SUMMARY |
3351 DIFF_FORMAT_CHECKDIFF))
3352 DIFF_OPT_SET(options, RECURSIVE);
3353 /*
3354 * Also pickaxe would not work very well if you do not say recursive
3355 */
3356 if (options->pickaxe)
3357 DIFF_OPT_SET(options, RECURSIVE);
3358 /*
3359 * When patches are generated, submodules diffed against the work tree
3360 * must be checked for dirtiness too so it can be shown in the output
3361 */
3362 if (options->output_format & DIFF_FORMAT_PATCH)
3363 DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3364
3365 if (options->detect_rename && options->rename_limit < 0)
3366 options->rename_limit = diff_rename_limit_default;
3367 if (options->setup & DIFF_SETUP_USE_CACHE) {
3368 if (!active_cache)
3369 /* read-cache does not die even when it fails
3370 * so it is safe for us to do this here. Also
3371 * it does not smudge active_cache or active_nr
3372 * when it fails, so we do not have to worry about
3373 * cleaning it up ourselves either.
3374 */
3375 read_cache();
3376 }
3377 if (options->abbrev <= 0 || 40 < options->abbrev)
3378 options->abbrev = 40; /* full */
3379
3380 /*
3381 * It does not make sense to show the first hit we happened
3382 * to have found. It does not make sense not to return with
3383 * exit code in such a case either.
3384 */
3385 if (DIFF_OPT_TST(options, QUICK)) {
3386 options->output_format = DIFF_FORMAT_NO_OUTPUT;
3387 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3388 }
3389
3390 options->diff_path_counter = 0;
3391
3392 if (DIFF_OPT_TST(options, FOLLOW_RENAMES) && options->pathspec.nr != 1)
3393 die(_("--follow requires exactly one pathspec"));
3394 }
3395
3396 static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3397 {
3398 char c, *eq;
3399 int len;
3400
3401 if (*arg != '-')
3402 return 0;
3403 c = *++arg;
3404 if (!c)
3405 return 0;
3406 if (c == arg_short) {
3407 c = *++arg;
3408 if (!c)
3409 return 1;
3410 if (val && isdigit(c)) {
3411 char *end;
3412 int n = strtoul(arg, &end, 10);
3413 if (*end)
3414 return 0;
3415 *val = n;
3416 return 1;
3417 }
3418 return 0;
3419 }
3420 if (c != '-')
3421 return 0;
3422 arg++;
3423 eq = strchrnul(arg, '=');
3424 len = eq - arg;
3425 if (!len || strncmp(arg, arg_long, len))
3426 return 0;
3427 if (*eq) {
3428 int n;
3429 char *end;
3430 if (!isdigit(*++eq))
3431 return 0;
3432 n = strtoul(eq, &end, 10);
3433 if (*end)
3434 return 0;
3435 *val = n;
3436 }
3437 return 1;
3438 }
3439
3440 static int diff_scoreopt_parse(const char *opt);
3441
3442 static inline int short_opt(char opt, const char **argv,
3443 const char **optarg)
3444 {
3445 const char *arg = argv[0];
3446 if (arg[0] != '-' || arg[1] != opt)
3447 return 0;
3448 if (arg[2] != '\0') {
3449 *optarg = arg + 2;
3450 return 1;
3451 }
3452 if (!argv[1])
3453 die("Option '%c' requires a value", opt);
3454 *optarg = argv[1];
3455 return 2;
3456 }
3457
3458 int parse_long_opt(const char *opt, const char **argv,
3459 const char **optarg)
3460 {
3461 const char *arg = argv[0];
3462 if (!skip_prefix(arg, "--", &arg))
3463 return 0;
3464 if (!skip_prefix(arg, opt, &arg))
3465 return 0;
3466 if (*arg == '=') { /* stuck form: --option=value */
3467 *optarg = arg + 1;
3468 return 1;
3469 }
3470 if (*arg != '\0')
3471 return 0;
3472 /* separate form: --option value */
3473 if (!argv[1])
3474 die("Option '--%s' requires a value", opt);
3475 *optarg = argv[1];
3476 return 2;
3477 }
3478
3479 static int stat_opt(struct diff_options *options, const char **av)
3480 {
3481 const char *arg = av[0];
3482 char *end;
3483 int width = options->stat_width;
3484 int name_width = options->stat_name_width;
3485 int graph_width = options->stat_graph_width;
3486 int count = options->stat_count;
3487 int argcount = 1;
3488
3489 if (!skip_prefix(arg, "--stat", &arg))
3490 die("BUG: stat option does not begin with --stat: %s", arg);
3491 end = (char *)arg;
3492
3493 switch (*arg) {
3494 case '-':
3495 if (skip_prefix(arg, "-width", &arg)) {
3496 if (*arg == '=')
3497 width = strtoul(arg + 1, &end, 10);
3498 else if (!*arg && !av[1])
3499 die("Option '--stat-width' requires a value");
3500 else if (!*arg) {
3501 width = strtoul(av[1], &end, 10);
3502 argcount = 2;
3503 }
3504 } else if (skip_prefix(arg, "-name-width", &arg)) {
3505 if (*arg == '=')
3506 name_width = strtoul(arg + 1, &end, 10);
3507 else if (!*arg && !av[1])
3508 die("Option '--stat-name-width' requires a value");
3509 else if (!*arg) {
3510 name_width = strtoul(av[1], &end, 10);
3511 argcount = 2;
3512 }
3513 } else if (skip_prefix(arg, "-graph-width", &arg)) {
3514 if (*arg == '=')
3515 graph_width = strtoul(arg + 1, &end, 10);
3516 else if (!*arg && !av[1])
3517 die("Option '--stat-graph-width' requires a value");
3518 else if (!*arg) {
3519 graph_width = strtoul(av[1], &end, 10);
3520 argcount = 2;
3521 }
3522 } else if (skip_prefix(arg, "-count", &arg)) {
3523 if (*arg == '=')
3524 count = strtoul(arg + 1, &end, 10);
3525 else if (!*arg && !av[1])
3526 die("Option '--stat-count' requires a value");
3527 else if (!*arg) {
3528 count = strtoul(av[1], &end, 10);
3529 argcount = 2;
3530 }
3531 }
3532 break;
3533 case '=':
3534 width = strtoul(arg+1, &end, 10);
3535 if (*end == ',')
3536 name_width = strtoul(end+1, &end, 10);
3537 if (*end == ',')
3538 count = strtoul(end+1, &end, 10);
3539 }
3540
3541 /* Important! This checks all the error cases! */
3542 if (*end)
3543 return 0;
3544 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3545 options->stat_name_width = name_width;
3546 options->stat_graph_width = graph_width;
3547 options->stat_width = width;
3548 options->stat_count = count;
3549 return argcount;
3550 }
3551
3552 static int parse_dirstat_opt(struct diff_options *options, const char *params)
3553 {
3554 struct strbuf errmsg = STRBUF_INIT;
3555 if (parse_dirstat_params(options, params, &errmsg))
3556 die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3557 errmsg.buf);
3558 strbuf_release(&errmsg);
3559 /*
3560 * The caller knows a dirstat-related option is given from the command
3561 * line; allow it to say "return this_function();"
3562 */
3563 options->output_format |= DIFF_FORMAT_DIRSTAT;
3564 return 1;
3565 }
3566
3567 static int parse_submodule_opt(struct diff_options *options, const char *value)
3568 {
3569 if (parse_submodule_params(options, value))
3570 die(_("Failed to parse --submodule option parameter: '%s'"),
3571 value);
3572 return 1;
3573 }
3574
3575 static const char diff_status_letters[] = {
3576 DIFF_STATUS_ADDED,
3577 DIFF_STATUS_COPIED,
3578 DIFF_STATUS_DELETED,
3579 DIFF_STATUS_MODIFIED,
3580 DIFF_STATUS_RENAMED,
3581 DIFF_STATUS_TYPE_CHANGED,
3582 DIFF_STATUS_UNKNOWN,
3583 DIFF_STATUS_UNMERGED,
3584 DIFF_STATUS_FILTER_AON,
3585 DIFF_STATUS_FILTER_BROKEN,
3586 '\0',
3587 };
3588
3589 static unsigned int filter_bit['Z' + 1];
3590
3591 static void prepare_filter_bits(void)
3592 {
3593 int i;
3594
3595 if (!filter_bit[DIFF_STATUS_ADDED]) {
3596 for (i = 0; diff_status_letters[i]; i++)
3597 filter_bit[(int) diff_status_letters[i]] = (1 << i);
3598 }
3599 }
3600
3601 static unsigned filter_bit_tst(char status, const struct diff_options *opt)
3602 {
3603 return opt->filter & filter_bit[(int) status];
3604 }
3605
3606 static int parse_diff_filter_opt(const char *optarg, struct diff_options *opt)
3607 {
3608 int i, optch;
3609
3610 prepare_filter_bits();
3611
3612 /*
3613 * If there is a negation e.g. 'd' in the input, and we haven't
3614 * initialized the filter field with another --diff-filter, start
3615 * from full set of bits, except for AON.
3616 */
3617 if (!opt->filter) {
3618 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3619 if (optch < 'a' || 'z' < optch)
3620 continue;
3621 opt->filter = (1 << (ARRAY_SIZE(diff_status_letters) - 1)) - 1;
3622 opt->filter &= ~filter_bit[DIFF_STATUS_FILTER_AON];
3623 break;
3624 }
3625 }
3626
3627 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3628 unsigned int bit;
3629 int negate;
3630
3631 if ('a' <= optch && optch <= 'z') {
3632 negate = 1;
3633 optch = toupper(optch);
3634 } else {
3635 negate = 0;
3636 }
3637
3638 bit = (0 <= optch && optch <= 'Z') ? filter_bit[optch] : 0;
3639 if (!bit)
3640 return optarg[i];
3641 if (negate)
3642 opt->filter &= ~bit;
3643 else
3644 opt->filter |= bit;
3645 }
3646 return 0;
3647 }
3648
3649 static void enable_patch_output(int *fmt) {
3650 *fmt &= ~DIFF_FORMAT_NO_OUTPUT;
3651 *fmt |= DIFF_FORMAT_PATCH;
3652 }
3653
3654 static int parse_one_token(const char **arg, const char *token)
3655 {
3656 const char *rest;
3657 if (skip_prefix(*arg, token, &rest) && (!*rest || *rest == ',')) {
3658 *arg = rest;
3659 return 1;
3660 }
3661 return 0;
3662 }
3663
3664 static int parse_ws_error_highlight(struct diff_options *opt, const char *arg)
3665 {
3666 const char *orig_arg = arg;
3667 unsigned val = 0;
3668 while (*arg) {
3669 if (parse_one_token(&arg, "none"))
3670 val = 0;
3671 else if (parse_one_token(&arg, "default"))
3672 val = WSEH_NEW;
3673 else if (parse_one_token(&arg, "all"))
3674 val = WSEH_NEW | WSEH_OLD | WSEH_CONTEXT;
3675 else if (parse_one_token(&arg, "new"))
3676 val |= WSEH_NEW;
3677 else if (parse_one_token(&arg, "old"))
3678 val |= WSEH_OLD;
3679 else if (parse_one_token(&arg, "context"))
3680 val |= WSEH_CONTEXT;
3681 else {
3682 error("unknown value after ws-error-highlight=%.*s",
3683 (int)(arg - orig_arg), orig_arg);
3684 return 0;
3685 }
3686 if (*arg)
3687 arg++;
3688 }
3689 opt->ws_error_highlight = val;
3690 return 1;
3691 }
3692
3693 int diff_opt_parse(struct diff_options *options, const char **av, int ac)
3694 {
3695 const char *arg = av[0];
3696 const char *optarg