]> git.ipfire.org Git - thirdparty/git.git/blob - diff.c
projects / thirdparty / git.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
diff.c: emit_diff_symbol learns DIFF_SYMBOL_CONTEXT_INCOMPLETE
[thirdparty/git.git] / diff.c
1 /*
2 * Copyright (C) 2005 Junio C Hamano
3 */
4 #include "cache.h"
5 #include "config.h"
6 #include "tempfile.h"
7 #include "quote.h"
8 #include "diff.h"
9 #include "diffcore.h"
10 #include "delta.h"
11 #include "xdiff-interface.h"
12 #include "color.h"
13 #include "attr.h"
14 #include "run-command.h"
15 #include "utf8.h"
16 #include "userdiff.h"
17 #include "submodule-config.h"
18 #include "submodule.h"
19 #include "ll-merge.h"
20 #include "string-list.h"
21 #include "argv-array.h"
22 #include "graph.h"
23
24 #ifdef NO_FAST_WORKING_DIRECTORY
25 #define FAST_WORKING_DIRECTORY 0
26 #else
27 #define FAST_WORKING_DIRECTORY 1
28 #endif
29
30 static int diff_detect_rename_default;
31 static int diff_indent_heuristic = 1;
32 static int diff_rename_limit_default = 400;
33 static int diff_suppress_blank_empty;
34 static int diff_use_color_default = -1;
35 static int diff_context_default = 3;
36 static int diff_interhunk_context_default;
37 static const char *diff_word_regex_cfg;
38 static const char *external_diff_cmd_cfg;
39 static const char *diff_order_file_cfg;
40 int diff_auto_refresh_index = 1;
41 static int diff_mnemonic_prefix;
42 static int diff_no_prefix;
43 static int diff_stat_graph_width;
44 static int diff_dirstat_permille_default = 30;
45 static struct diff_options default_diff_options;
46 static long diff_algorithm;
47 static unsigned ws_error_highlight_default = WSEH_NEW;
48
49 static char diff_colors[][COLOR_MAXLEN] = {
50 GIT_COLOR_RESET,
51 GIT_COLOR_NORMAL, /* CONTEXT */
52 GIT_COLOR_BOLD, /* METAINFO */
53 GIT_COLOR_CYAN, /* FRAGINFO */
54 GIT_COLOR_RED, /* OLD */
55 GIT_COLOR_GREEN, /* NEW */
56 GIT_COLOR_YELLOW, /* COMMIT */
57 GIT_COLOR_BG_RED, /* WHITESPACE */
58 GIT_COLOR_NORMAL, /* FUNCINFO */
59 };
60
61 static NORETURN void die_want_option(const char *option_name)
62 {
63 die(_("option '%s' requires a value"), option_name);
64 }
65
66 static int parse_diff_color_slot(const char *var)
67 {
68 if (!strcasecmp(var, "context") || !strcasecmp(var, "plain"))
69 return DIFF_CONTEXT;
70 if (!strcasecmp(var, "meta"))
71 return DIFF_METAINFO;
72 if (!strcasecmp(var, "frag"))
73 return DIFF_FRAGINFO;
74 if (!strcasecmp(var, "old"))
75 return DIFF_FILE_OLD;
76 if (!strcasecmp(var, "new"))
77 return DIFF_FILE_NEW;
78 if (!strcasecmp(var, "commit"))
79 return DIFF_COMMIT;
80 if (!strcasecmp(var, "whitespace"))
81 return DIFF_WHITESPACE;
82 if (!strcasecmp(var, "func"))
83 return DIFF_FUNCINFO;
84 return -1;
85 }
86
87 static int parse_dirstat_params(struct diff_options *options, const char *params_string,
88 struct strbuf *errmsg)
89 {
90 char *params_copy = xstrdup(params_string);
91 struct string_list params = STRING_LIST_INIT_NODUP;
92 int ret = 0;
93 int i;
94
95 if (*params_copy)
96 string_list_split_in_place(&params, params_copy, ',', -1);
97 for (i = 0; i < params.nr; i++) {
98 const char *p = params.items[i].string;
99 if (!strcmp(p, "changes")) {
100 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
101 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
102 } else if (!strcmp(p, "lines")) {
103 DIFF_OPT_SET(options, DIRSTAT_BY_LINE);
104 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
105 } else if (!strcmp(p, "files")) {
106 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
107 DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
108 } else if (!strcmp(p, "noncumulative")) {
109 DIFF_OPT_CLR(options, DIRSTAT_CUMULATIVE);
110 } else if (!strcmp(p, "cumulative")) {
111 DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
112 } else if (isdigit(*p)) {
113 char *end;
114 int permille = strtoul(p, &end, 10) * 10;
115 if (*end == '.' && isdigit(*++end)) {
116 /* only use first digit */
117 permille += *end - '0';
118 /* .. and ignore any further digits */
119 while (isdigit(*++end))
120 ; /* nothing */
121 }
122 if (!*end)
123 options->dirstat_permille = permille;
124 else {
125 strbuf_addf(errmsg, _(" Failed to parse dirstat cut-off percentage '%s'\n"),
126 p);
127 ret++;
128 }
129 } else {
130 strbuf_addf(errmsg, _(" Unknown dirstat parameter '%s'\n"), p);
131 ret++;
132 }
133
134 }
135 string_list_clear(&params, 0);
136 free(params_copy);
137 return ret;
138 }
139
140 static int parse_submodule_params(struct diff_options *options, const char *value)
141 {
142 if (!strcmp(value, "log"))
143 options->submodule_format = DIFF_SUBMODULE_LOG;
144 else if (!strcmp(value, "short"))
145 options->submodule_format = DIFF_SUBMODULE_SHORT;
146 else if (!strcmp(value, "diff"))
147 options->submodule_format = DIFF_SUBMODULE_INLINE_DIFF;
148 else
149 return -1;
150 return 0;
151 }
152
153 static int git_config_rename(const char *var, const char *value)
154 {
155 if (!value)
156 return DIFF_DETECT_RENAME;
157 if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
158 return DIFF_DETECT_COPY;
159 return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
160 }
161
162 long parse_algorithm_value(const char *value)
163 {
164 if (!value)
165 return -1;
166 else if (!strcasecmp(value, "myers") || !strcasecmp(value, "default"))
167 return 0;
168 else if (!strcasecmp(value, "minimal"))
169 return XDF_NEED_MINIMAL;
170 else if (!strcasecmp(value, "patience"))
171 return XDF_PATIENCE_DIFF;
172 else if (!strcasecmp(value, "histogram"))
173 return XDF_HISTOGRAM_DIFF;
174 return -1;
175 }
176
177 static int parse_one_token(const char **arg, const char *token)
178 {
179 const char *rest;
180 if (skip_prefix(*arg, token, &rest) && (!*rest || *rest == ',')) {
181 *arg = rest;
182 return 1;
183 }
184 return 0;
185 }
186
187 static int parse_ws_error_highlight(const char *arg)
188 {
189 const char *orig_arg = arg;
190 unsigned val = 0;
191
192 while (*arg) {
193 if (parse_one_token(&arg, "none"))
194 val = 0;
195 else if (parse_one_token(&arg, "default"))
196 val = WSEH_NEW;
197 else if (parse_one_token(&arg, "all"))
198 val = WSEH_NEW | WSEH_OLD | WSEH_CONTEXT;
199 else if (parse_one_token(&arg, "new"))
200 val |= WSEH_NEW;
201 else if (parse_one_token(&arg, "old"))
202 val |= WSEH_OLD;
203 else if (parse_one_token(&arg, "context"))
204 val |= WSEH_CONTEXT;
205 else {
206 return -1 - (int)(arg - orig_arg);
207 }
208 if (*arg)
209 arg++;
210 }
211 return val;
212 }
213
214 /*
215 * These are to give UI layer defaults.
216 * The core-level commands such as git-diff-files should
217 * never be affected by the setting of diff.renames
218 * the user happens to have in the configuration file.
219 */
220 void init_diff_ui_defaults(void)
221 {
222 diff_detect_rename_default = 1;
223 }
224
225 int git_diff_heuristic_config(const char *var, const char *value, void *cb)
226 {
227 if (!strcmp(var, "diff.indentheuristic"))
228 diff_indent_heuristic = git_config_bool(var, value);
229 return 0;
230 }
231
232 int git_diff_ui_config(const char *var, const char *value, void *cb)
233 {
234 if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
235 diff_use_color_default = git_config_colorbool(var, value);
236 return 0;
237 }
238 if (!strcmp(var, "diff.context")) {
239 diff_context_default = git_config_int(var, value);
240 if (diff_context_default < 0)
241 return -1;
242 return 0;
243 }
244 if (!strcmp(var, "diff.interhunkcontext")) {
245 diff_interhunk_context_default = git_config_int(var, value);
246 if (diff_interhunk_context_default < 0)
247 return -1;
248 return 0;
249 }
250 if (!strcmp(var, "diff.renames")) {
251 diff_detect_rename_default = git_config_rename(var, value);
252 return 0;
253 }
254 if (!strcmp(var, "diff.autorefreshindex")) {
255 diff_auto_refresh_index = git_config_bool(var, value);
256 return 0;
257 }
258 if (!strcmp(var, "diff.mnemonicprefix")) {
259 diff_mnemonic_prefix = git_config_bool(var, value);
260 return 0;
261 }
262 if (!strcmp(var, "diff.noprefix")) {
263 diff_no_prefix = git_config_bool(var, value);
264 return 0;
265 }
266 if (!strcmp(var, "diff.statgraphwidth")) {
267 diff_stat_graph_width = git_config_int(var, value);
268 return 0;
269 }
270 if (!strcmp(var, "diff.external"))
271 return git_config_string(&external_diff_cmd_cfg, var, value);
272 if (!strcmp(var, "diff.wordregex"))
273 return git_config_string(&diff_word_regex_cfg, var, value);
274 if (!strcmp(var, "diff.orderfile"))
275 return git_config_pathname(&diff_order_file_cfg, var, value);
276
277 if (!strcmp(var, "diff.ignoresubmodules"))
278 handle_ignore_submodules_arg(&default_diff_options, value);
279
280 if (!strcmp(var, "diff.submodule")) {
281 if (parse_submodule_params(&default_diff_options, value))
282 warning(_("Unknown value for 'diff.submodule' config variable: '%s'"),
283 value);
284 return 0;
285 }
286
287 if (!strcmp(var, "diff.algorithm")) {
288 diff_algorithm = parse_algorithm_value(value);
289 if (diff_algorithm < 0)
290 return -1;
291 return 0;
292 }
293
294 if (!strcmp(var, "diff.wserrorhighlight")) {
295 int val = parse_ws_error_highlight(value);
296 if (val < 0)
297 return -1;
298 ws_error_highlight_default = val;
299 return 0;
300 }
301
302 if (git_color_config(var, value, cb) < 0)
303 return -1;
304
305 return git_diff_basic_config(var, value, cb);
306 }
307
308 int git_diff_basic_config(const char *var, const char *value, void *cb)
309 {
310 const char *name;
311
312 if (!strcmp(var, "diff.renamelimit")) {
313 diff_rename_limit_default = git_config_int(var, value);
314 return 0;
315 }
316
317 if (userdiff_config(var, value) < 0)
318 return -1;
319
320 if (skip_prefix(var, "diff.color.", &name) ||
321 skip_prefix(var, "color.diff.", &name)) {
322 int slot = parse_diff_color_slot(name);
323 if (slot < 0)
324 return 0;
325 if (!value)
326 return config_error_nonbool(var);
327 return color_parse(value, diff_colors[slot]);
328 }
329
330 /* like GNU diff's --suppress-blank-empty option */
331 if (!strcmp(var, "diff.suppressblankempty") ||
332 /* for backwards compatibility */
333 !strcmp(var, "diff.suppress-blank-empty")) {
334 diff_suppress_blank_empty = git_config_bool(var, value);
335 return 0;
336 }
337
338 if (!strcmp(var, "diff.dirstat")) {
339 struct strbuf errmsg = STRBUF_INIT;
340 default_diff_options.dirstat_permille = diff_dirstat_permille_default;
341 if (parse_dirstat_params(&default_diff_options, value, &errmsg))
342 warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
343 errmsg.buf);
344 strbuf_release(&errmsg);
345 diff_dirstat_permille_default = default_diff_options.dirstat_permille;
346 return 0;
347 }
348
349 if (starts_with(var, "submodule."))
350 return parse_submodule_config_option(var, value);
351
352 if (git_diff_heuristic_config(var, value, cb) < 0)
353 return -1;
354
355 return git_default_config(var, value, cb);
356 }
357
358 static char *quote_two(const char *one, const char *two)
359 {
360 int need_one = quote_c_style(one, NULL, NULL, 1);
361 int need_two = quote_c_style(two, NULL, NULL, 1);
362 struct strbuf res = STRBUF_INIT;
363
364 if (need_one + need_two) {
365 strbuf_addch(&res, '"');
366 quote_c_style(one, &res, NULL, 1);
367 quote_c_style(two, &res, NULL, 1);
368 strbuf_addch(&res, '"');
369 } else {
370 strbuf_addstr(&res, one);
371 strbuf_addstr(&res, two);
372 }
373 return strbuf_detach(&res, NULL);
374 }
375
376 static const char *external_diff(void)
377 {
378 static const char *external_diff_cmd = NULL;
379 static int done_preparing = 0;
380
381 if (done_preparing)
382 return external_diff_cmd;
383 external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
384 if (!external_diff_cmd)
385 external_diff_cmd = external_diff_cmd_cfg;
386 done_preparing = 1;
387 return external_diff_cmd;
388 }
389
390 /*
391 * Keep track of files used for diffing. Sometimes such an entry
392 * refers to a temporary file, sometimes to an existing file, and
393 * sometimes to "/dev/null".
394 */
395 static struct diff_tempfile {
396 /*
397 * filename external diff should read from, or NULL if this
398 * entry is currently not in use:
399 */
400 const char *name;
401
402 char hex[GIT_MAX_HEXSZ + 1];
403 char mode[10];
404
405 /*
406 * If this diff_tempfile instance refers to a temporary file,
407 * this tempfile object is used to manage its lifetime.
408 */
409 struct tempfile tempfile;
410 } diff_temp[2];
411
412 typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
413
414 struct emit_callback {
415 int color_diff;
416 unsigned ws_rule;
417 int blank_at_eof_in_preimage;
418 int blank_at_eof_in_postimage;
419 int lno_in_preimage;
420 int lno_in_postimage;
421 sane_truncate_fn truncate;
422 const char **label_path;
423 struct diff_words_data *diff_words;
424 struct diff_options *opt;
425 struct strbuf *header;
426 };
427
428 static int count_lines(const char *data, int size)
429 {
430 int count, ch, completely_empty = 1, nl_just_seen = 0;
431 count = 0;
432 while (0 < size--) {
433 ch = *data++;
434 if (ch == '\n') {
435 count++;
436 nl_just_seen = 1;
437 completely_empty = 0;
438 }
439 else {
440 nl_just_seen = 0;
441 completely_empty = 0;
442 }
443 }
444 if (completely_empty)
445 return 0;
446 if (!nl_just_seen)
447 count++; /* no trailing newline */
448 return count;
449 }
450
451 static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
452 {
453 if (!DIFF_FILE_VALID(one)) {
454 mf->ptr = (char *)""; /* does not matter */
455 mf->size = 0;
456 return 0;
457 }
458 else if (diff_populate_filespec(one, 0))
459 return -1;
460
461 mf->ptr = one->data;
462 mf->size = one->size;
463 return 0;
464 }
465
466 /* like fill_mmfile, but only for size, so we can avoid retrieving blob */
467 static unsigned long diff_filespec_size(struct diff_filespec *one)
468 {
469 if (!DIFF_FILE_VALID(one))
470 return 0;
471 diff_populate_filespec(one, CHECK_SIZE_ONLY);
472 return one->size;
473 }
474
475 static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
476 {
477 char *ptr = mf->ptr;
478 long size = mf->size;
479 int cnt = 0;
480
481 if (!size)
482 return cnt;
483 ptr += size - 1; /* pointing at the very end */
484 if (*ptr != '\n')
485 ; /* incomplete line */
486 else
487 ptr--; /* skip the last LF */
488 while (mf->ptr < ptr) {
489 char *prev_eol;
490 for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
491 if (*prev_eol == '\n')
492 break;
493 if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
494 break;
495 cnt++;
496 ptr = prev_eol - 1;
497 }
498 return cnt;
499 }
500
501 static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
502 struct emit_callback *ecbdata)
503 {
504 int l1, l2, at;
505 unsigned ws_rule = ecbdata->ws_rule;
506 l1 = count_trailing_blank(mf1, ws_rule);
507 l2 = count_trailing_blank(mf2, ws_rule);
508 if (l2 <= l1) {
509 ecbdata->blank_at_eof_in_preimage = 0;
510 ecbdata->blank_at_eof_in_postimage = 0;
511 return;
512 }
513 at = count_lines(mf1->ptr, mf1->size);
514 ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
515
516 at = count_lines(mf2->ptr, mf2->size);
517 ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
518 }
519
520 static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
521 int first, const char *line, int len)
522 {
523 int has_trailing_newline, has_trailing_carriage_return;
524 int nofirst;
525 FILE *file = o->file;
526
527 fputs(diff_line_prefix(o), file);
528
529 if (len == 0) {
530 has_trailing_newline = (first == '\n');
531 has_trailing_carriage_return = (!has_trailing_newline &&
532 (first == '\r'));
533 nofirst = has_trailing_newline || has_trailing_carriage_return;
534 } else {
535 has_trailing_newline = (len > 0 && line[len-1] == '\n');
536 if (has_trailing_newline)
537 len--;
538 has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
539 if (has_trailing_carriage_return)
540 len--;
541 nofirst = 0;
542 }
543
544 if (len || !nofirst) {
545 fputs(set, file);
546 if (!nofirst)
547 fputc(first, file);
548 fwrite(line, len, 1, file);
549 fputs(reset, file);
550 }
551 if (has_trailing_carriage_return)
552 fputc('\r', file);
553 if (has_trailing_newline)
554 fputc('\n', file);
555 }
556
557 static void emit_line(struct diff_options *o, const char *set, const char *reset,
558 const char *line, int len)
559 {
560 emit_line_0(o, set, reset, line[0], line+1, len-1);
561 }
562
563 enum diff_symbol {
564 DIFF_SYMBOL_WORDS_PORCELAIN,
565 DIFF_SYMBOL_WORDS,
566 DIFF_SYMBOL_CONTEXT,
567 DIFF_SYMBOL_CONTEXT_INCOMPLETE,
568 DIFF_SYMBOL_PLUS,
569 DIFF_SYMBOL_MINUS,
570 DIFF_SYMBOL_NO_LF_EOF,
571 DIFF_SYMBOL_CONTEXT_FRAGINFO,
572 DIFF_SYMBOL_CONTEXT_MARKER,
573 DIFF_SYMBOL_SEPARATOR
574 };
575 /*
576 * Flags for content lines:
577 * 0..12 are whitespace rules
578 * 13-15 are WSEH_NEW | WSEH_OLD | WSEH_CONTEXT
579 * 16 is marking if the line is blank at EOF
580 */
581 #define DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF (1<<16)
582 #define DIFF_SYMBOL_CONTENT_WS_MASK (WSEH_NEW | WSEH_OLD | WSEH_CONTEXT | WS_RULE_MASK)
583
584 static void emit_line_ws_markup(struct diff_options *o,
585 const char *set, const char *reset,
586 const char *line, int len, char sign,
587 unsigned ws_rule, int blank_at_eof)
588 {
589 const char *ws = NULL;
590
591 if (o->ws_error_highlight & ws_rule) {
592 ws = diff_get_color_opt(o, DIFF_WHITESPACE);
593 if (!*ws)
594 ws = NULL;
595 }
596
597 if (!ws)
598 emit_line_0(o, set, reset, sign, line, len);
599 else if (blank_at_eof)
600 /* Blank line at EOF - paint '+' as well */
601 emit_line_0(o, ws, reset, sign, line, len);
602 else {
603 /* Emit just the prefix, then the rest. */
604 emit_line_0(o, set, reset, sign, "", 0);
605 ws_check_emit(line, len, ws_rule,
606 o->file, set, reset, ws);
607 }
608 }
609
610 static void emit_diff_symbol(struct diff_options *o, enum diff_symbol s,
611 const char *line, int len, unsigned flags)
612 {
613 static const char *nneof = " No newline at end of file\n";
614 const char *context, *reset, *set;
615 switch (s) {
616 case DIFF_SYMBOL_NO_LF_EOF:
617 context = diff_get_color_opt(o, DIFF_CONTEXT);
618 reset = diff_get_color_opt(o, DIFF_RESET);
619 putc('\n', o->file);
620 emit_line_0(o, context, reset, '\\',
621 nneof, strlen(nneof));
622 break;
623 case DIFF_SYMBOL_CONTEXT_FRAGINFO:
624 emit_line(o, "", "", line, len);
625 break;
626 case DIFF_SYMBOL_CONTEXT_INCOMPLETE:
627 case DIFF_SYMBOL_CONTEXT_MARKER:
628 context = diff_get_color_opt(o, DIFF_CONTEXT);
629 reset = diff_get_color_opt(o, DIFF_RESET);
630 emit_line(o, context, reset, line, len);
631 break;
632 case DIFF_SYMBOL_SEPARATOR:
633 fprintf(o->file, "%s%c",
634 diff_line_prefix(o),
635 o->line_termination);
636 break;
637 case DIFF_SYMBOL_CONTEXT:
638 set = diff_get_color_opt(o, DIFF_CONTEXT);
639 reset = diff_get_color_opt(o, DIFF_RESET);
640 emit_line_ws_markup(o, set, reset, line, len, ' ',
641 flags & (DIFF_SYMBOL_CONTENT_WS_MASK), 0);
642 break;
643 case DIFF_SYMBOL_PLUS:
644 set = diff_get_color_opt(o, DIFF_FILE_NEW);
645 reset = diff_get_color_opt(o, DIFF_RESET);
646 emit_line_ws_markup(o, set, reset, line, len, '+',
647 flags & DIFF_SYMBOL_CONTENT_WS_MASK,
648 flags & DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF);
649 break;
650 case DIFF_SYMBOL_MINUS:
651 set = diff_get_color_opt(o, DIFF_FILE_OLD);
652 reset = diff_get_color_opt(o, DIFF_RESET);
653 emit_line_ws_markup(o, set, reset, line, len, '-',
654 flags & DIFF_SYMBOL_CONTENT_WS_MASK, 0);
655 break;
656 case DIFF_SYMBOL_WORDS_PORCELAIN:
657 context = diff_get_color_opt(o, DIFF_CONTEXT);
658 reset = diff_get_color_opt(o, DIFF_RESET);
659 emit_line(o, context, reset, line, len);
660 fputs("~\n", o->file);
661 break;
662 case DIFF_SYMBOL_WORDS:
663 context = diff_get_color_opt(o, DIFF_CONTEXT);
664 reset = diff_get_color_opt(o, DIFF_RESET);
665 /*
666 * Skip the prefix character, if any. With
667 * diff_suppress_blank_empty, there may be
668 * none.
669 */
670 if (line[0] != '\n') {
671 line++;
672 len--;
673 }
674 emit_line(o, context, reset, line, len);
675 break;
676 default:
677 die("BUG: unknown diff symbol");
678 }
679 }
680
681 static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
682 {
683 if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
684 ecbdata->blank_at_eof_in_preimage &&
685 ecbdata->blank_at_eof_in_postimage &&
686 ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
687 ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
688 return 0;
689 return ws_blank_line(line, len, ecbdata->ws_rule);
690 }
691
692 static void emit_add_line(const char *reset,
693 struct emit_callback *ecbdata,
694 const char *line, int len)
695 {
696 unsigned flags = WSEH_NEW | ecbdata->ws_rule;
697 if (new_blank_line_at_eof(ecbdata, line, len))
698 flags |= DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF;
699
700 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_PLUS, line, len, flags);
701 }
702
703 static void emit_del_line(const char *reset,
704 struct emit_callback *ecbdata,
705 const char *line, int len)
706 {
707 unsigned flags = WSEH_OLD | ecbdata->ws_rule;
708 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_MINUS, line, len, flags);
709 }
710
711 static void emit_context_line(const char *reset,
712 struct emit_callback *ecbdata,
713 const char *line, int len)
714 {
715 unsigned flags = WSEH_CONTEXT | ecbdata->ws_rule;
716 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_CONTEXT, line, len, flags);
717 }
718
719 static void emit_hunk_header(struct emit_callback *ecbdata,
720 const char *line, int len)
721 {
722 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
723 const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
724 const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
725 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
726 static const char atat[2] = { '@', '@' };
727 const char *cp, *ep;
728 struct strbuf msgbuf = STRBUF_INIT;
729 int org_len = len;
730 int i = 1;
731
732 /*
733 * As a hunk header must begin with "@@ -<old>, +<new> @@",
734 * it always is at least 10 bytes long.
735 */
736 if (len < 10 ||
737 memcmp(line, atat, 2) ||
738 !(ep = memmem(line + 2, len - 2, atat, 2))) {
739 emit_diff_symbol(ecbdata->opt,
740 DIFF_SYMBOL_CONTEXT_MARKER, line, len, 0);
741 return;
742 }
743 ep += 2; /* skip over @@ */
744
745 /* The hunk header in fraginfo color */
746 strbuf_addstr(&msgbuf, frag);
747 strbuf_add(&msgbuf, line, ep - line);
748 strbuf_addstr(&msgbuf, reset);
749
750 /*
751 * trailing "\r\n"
752 */
753 for ( ; i < 3; i++)
754 if (line[len - i] == '\r' || line[len - i] == '\n')
755 len--;
756
757 /* blank before the func header */
758 for (cp = ep; ep - line < len; ep++)
759 if (*ep != ' ' && *ep != '\t')
760 break;
761 if (ep != cp) {
762 strbuf_addstr(&msgbuf, context);
763 strbuf_add(&msgbuf, cp, ep - cp);
764 strbuf_addstr(&msgbuf, reset);
765 }
766
767 if (ep < line + len) {
768 strbuf_addstr(&msgbuf, func);
769 strbuf_add(&msgbuf, ep, line + len - ep);
770 strbuf_addstr(&msgbuf, reset);
771 }
772
773 strbuf_add(&msgbuf, line + len, org_len - len);
774 strbuf_complete_line(&msgbuf);
775 emit_diff_symbol(ecbdata->opt,
776 DIFF_SYMBOL_CONTEXT_FRAGINFO, msgbuf.buf, msgbuf.len, 0);
777 strbuf_release(&msgbuf);
778 }
779
780 static struct diff_tempfile *claim_diff_tempfile(void) {
781 int i;
782 for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
783 if (!diff_temp[i].name)
784 return diff_temp + i;
785 die("BUG: diff is failing to clean up its tempfiles");
786 }
787
788 static void remove_tempfile(void)
789 {
790 int i;
791 for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
792 if (is_tempfile_active(&diff_temp[i].tempfile))
793 delete_tempfile(&diff_temp[i].tempfile);
794 diff_temp[i].name = NULL;
795 }
796 }
797
798 static void print_line_count(FILE *file, int count)
799 {
800 switch (count) {
801 case 0:
802 fprintf(file, "0,0");
803 break;
804 case 1:
805 fprintf(file, "1");
806 break;
807 default:
808 fprintf(file, "1,%d", count);
809 break;
810 }
811 }
812
813 static void emit_rewrite_lines(struct emit_callback *ecb,
814 int prefix, const char *data, int size)
815 {
816 const char *endp = NULL;
817 const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
818
819 while (0 < size) {
820 int len;
821
822 endp = memchr(data, '\n', size);
823 len = endp ? (endp - data + 1) : size;
824 if (prefix != '+') {
825 ecb->lno_in_preimage++;
826 emit_del_line(reset, ecb, data, len);
827 } else {
828 ecb->lno_in_postimage++;
829 emit_add_line(reset, ecb, data, len);
830 }
831 size -= len;
832 data += len;
833 }
834 if (!endp)
835 emit_diff_symbol(ecb->opt, DIFF_SYMBOL_NO_LF_EOF, NULL, 0, 0);
836 }
837
838 static void emit_rewrite_diff(const char *name_a,
839 const char *name_b,
840 struct diff_filespec *one,
841 struct diff_filespec *two,
842 struct userdiff_driver *textconv_one,
843 struct userdiff_driver *textconv_two,
844 struct diff_options *o)
845 {
846 int lc_a, lc_b;
847 const char *name_a_tab, *name_b_tab;
848 const char *metainfo = diff_get_color(o->use_color, DIFF_METAINFO);
849 const char *fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
850 const char *reset = diff_get_color(o->use_color, DIFF_RESET);
851 static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
852 const char *a_prefix, *b_prefix;
853 char *data_one, *data_two;
854 size_t size_one, size_two;
855 struct emit_callback ecbdata;
856 const char *line_prefix = diff_line_prefix(o);
857
858 if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
859 a_prefix = o->b_prefix;
860 b_prefix = o->a_prefix;
861 } else {
862 a_prefix = o->a_prefix;
863 b_prefix = o->b_prefix;
864 }
865
866 name_a += (*name_a == '/');
867 name_b += (*name_b == '/');
868 name_a_tab = strchr(name_a, ' ') ? "\t" : "";
869 name_b_tab = strchr(name_b, ' ') ? "\t" : "";
870
871 strbuf_reset(&a_name);
872 strbuf_reset(&b_name);
873 quote_two_c_style(&a_name, a_prefix, name_a, 0);
874 quote_two_c_style(&b_name, b_prefix, name_b, 0);
875
876 size_one = fill_textconv(textconv_one, one, &data_one);
877 size_two = fill_textconv(textconv_two, two, &data_two);
878
879 memset(&ecbdata, 0, sizeof(ecbdata));
880 ecbdata.color_diff = want_color(o->use_color);
881 ecbdata.ws_rule = whitespace_rule(name_b);
882 ecbdata.opt = o;
883 if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
884 mmfile_t mf1, mf2;
885 mf1.ptr = (char *)data_one;
886 mf2.ptr = (char *)data_two;
887 mf1.size = size_one;
888 mf2.size = size_two;
889 check_blank_at_eof(&mf1, &mf2, &ecbdata);
890 }
891 ecbdata.lno_in_preimage = 1;
892 ecbdata.lno_in_postimage = 1;
893
894 lc_a = count_lines(data_one, size_one);
895 lc_b = count_lines(data_two, size_two);
896 fprintf(o->file,
897 "%s%s--- %s%s%s\n%s%s+++ %s%s%s\n%s%s@@ -",
898 line_prefix, metainfo, a_name.buf, name_a_tab, reset,
899 line_prefix, metainfo, b_name.buf, name_b_tab, reset,
900 line_prefix, fraginfo);
901 if (!o->irreversible_delete)
902 print_line_count(o->file, lc_a);
903 else
904 fprintf(o->file, "?,?");
905 fprintf(o->file, " +");
906 print_line_count(o->file, lc_b);
907 fprintf(o->file, " @@%s\n", reset);
908 if (lc_a && !o->irreversible_delete)
909 emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
910 if (lc_b)
911 emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
912 if (textconv_one)
913 free((char *)data_one);
914 if (textconv_two)
915 free((char *)data_two);
916 }
917
918 struct diff_words_buffer {
919 mmfile_t text;
920 long alloc;
921 struct diff_words_orig {
922 const char *begin, *end;
923 } *orig;
924 int orig_nr, orig_alloc;
925 };
926
927 static void diff_words_append(char *line, unsigned long len,
928 struct diff_words_buffer *buffer)
929 {
930 ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
931 line++;
932 len--;
933 memcpy(buffer->text.ptr + buffer->text.size, line, len);
934 buffer->text.size += len;
935 buffer->text.ptr[buffer->text.size] = '\0';
936 }
937
938 struct diff_words_style_elem {
939 const char *prefix;
940 const char *suffix;
941 const char *color; /* NULL; filled in by the setup code if
942 * color is enabled */
943 };
944
945 struct diff_words_style {
946 enum diff_words_type type;
947 struct diff_words_style_elem new, old, ctx;
948 const char *newline;
949 };
950
951 static struct diff_words_style diff_words_styles[] = {
952 { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
953 { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
954 { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
955 };
956
957 struct diff_words_data {
958 struct diff_words_buffer minus, plus;
959 const char *current_plus;
960 int last_minus;
961 struct diff_options *opt;
962 regex_t *word_regex;
963 enum diff_words_type type;
964 struct diff_words_style *style;
965 };
966
967 static int fn_out_diff_words_write_helper(FILE *fp,
968 struct diff_words_style_elem *st_el,
969 const char *newline,
970 size_t count, const char *buf,
971 const char *line_prefix)
972 {
973 int print = 0;
974
975 while (count) {
976 char *p = memchr(buf, '\n', count);
977 if (print)
978 fputs(line_prefix, fp);
979 if (p != buf) {
980 if (st_el->color && fputs(st_el->color, fp) < 0)
981 return -1;
982 if (fputs(st_el->prefix, fp) < 0 ||
983 fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
984 fputs(st_el->suffix, fp) < 0)
985 return -1;
986 if (st_el->color && *st_el->color
987 && fputs(GIT_COLOR_RESET, fp) < 0)
988 return -1;
989 }
990 if (!p)
991 return 0;
992 if (fputs(newline, fp) < 0)
993 return -1;
994 count -= p + 1 - buf;
995 buf = p + 1;
996 print = 1;
997 }
998 return 0;
999 }
1000
1001 /*
1002 * '--color-words' algorithm can be described as:
1003 *
1004 * 1. collect the minus/plus lines of a diff hunk, divided into
1005 * minus-lines and plus-lines;
1006 *
1007 * 2. break both minus-lines and plus-lines into words and
1008 * place them into two mmfile_t with one word for each line;
1009 *
1010 * 3. use xdiff to run diff on the two mmfile_t to get the words level diff;
1011 *
1012 * And for the common parts of the both file, we output the plus side text.
1013 * diff_words->current_plus is used to trace the current position of the plus file
1014 * which printed. diff_words->last_minus is used to trace the last minus word
1015 * printed.
1016 *
1017 * For '--graph' to work with '--color-words', we need to output the graph prefix
1018 * on each line of color words output. Generally, there are two conditions on
1019 * which we should output the prefix.
1020 *
1021 * 1. diff_words->last_minus == 0 &&
1022 * diff_words->current_plus == diff_words->plus.text.ptr
1023 *
1024 * that is: the plus text must start as a new line, and if there is no minus
1025 * word printed, a graph prefix must be printed.
1026 *
1027 * 2. diff_words->current_plus > diff_words->plus.text.ptr &&
1028 * *(diff_words->current_plus - 1) == '\n'
1029 *
1030 * that is: a graph prefix must be printed following a '\n'
1031 */
1032 static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
1033 {
1034 if ((diff_words->last_minus == 0 &&
1035 diff_words->current_plus == diff_words->plus.text.ptr) ||
1036 (diff_words->current_plus > diff_words->plus.text.ptr &&
1037 *(diff_words->current_plus - 1) == '\n')) {
1038 return 1;
1039 } else {
1040 return 0;
1041 }
1042 }
1043
1044 static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
1045 {
1046 struct diff_words_data *diff_words = priv;
1047 struct diff_words_style *style = diff_words->style;
1048 int minus_first, minus_len, plus_first, plus_len;
1049 const char *minus_begin, *minus_end, *plus_begin, *plus_end;
1050 struct diff_options *opt = diff_words->opt;
1051 const char *line_prefix;
1052
1053 if (line[0] != '@' || parse_hunk_header(line, len,
1054 &minus_first, &minus_len, &plus_first, &plus_len))
1055 return;
1056
1057 assert(opt);
1058 line_prefix = diff_line_prefix(opt);
1059
1060 /* POSIX requires that first be decremented by one if len == 0... */
1061 if (minus_len) {
1062 minus_begin = diff_words->minus.orig[minus_first].begin;
1063 minus_end =
1064 diff_words->minus.orig[minus_first + minus_len - 1].end;
1065 } else
1066 minus_begin = minus_end =
1067 diff_words->minus.orig[minus_first].end;
1068
1069 if (plus_len) {
1070 plus_begin = diff_words->plus.orig[plus_first].begin;
1071 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
1072 } else
1073 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
1074
1075 if (color_words_output_graph_prefix(diff_words)) {
1076 fputs(line_prefix, diff_words->opt->file);
1077 }
1078 if (diff_words->current_plus != plus_begin) {
1079 fn_out_diff_words_write_helper(diff_words->opt->file,
1080 &style->ctx, style->newline,
1081 plus_begin - diff_words->current_plus,
1082 diff_words->current_plus, line_prefix);
1083 if (*(plus_begin - 1) == '\n')
1084 fputs(line_prefix, diff_words->opt->file);
1085 }
1086 if (minus_begin != minus_end) {
1087 fn_out_diff_words_write_helper(diff_words->opt->file,
1088 &style->old, style->newline,
1089 minus_end - minus_begin, minus_begin,
1090 line_prefix);
1091 }
1092 if (plus_begin != plus_end) {
1093 fn_out_diff_words_write_helper(diff_words->opt->file,
1094 &style->new, style->newline,
1095 plus_end - plus_begin, plus_begin,
1096 line_prefix);
1097 }
1098
1099 diff_words->current_plus = plus_end;
1100 diff_words->last_minus = minus_first;
1101 }
1102
1103 /* This function starts looking at *begin, and returns 0 iff a word was found. */
1104 static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
1105 int *begin, int *end)
1106 {
1107 if (word_regex && *begin < buffer->size) {
1108 regmatch_t match[1];
1109 if (!regexec_buf(word_regex, buffer->ptr + *begin,
1110 buffer->size - *begin, 1, match, 0)) {
1111 char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
1112 '\n', match[0].rm_eo - match[0].rm_so);
1113 *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
1114 *begin += match[0].rm_so;
1115 return *begin >= *end;
1116 }
1117 return -1;
1118 }
1119
1120 /* find the next word */
1121 while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
1122 (*begin)++;
1123 if (*begin >= buffer->size)
1124 return -1;
1125
1126 /* find the end of the word */
1127 *end = *begin + 1;
1128 while (*end < buffer->size && !isspace(buffer->ptr[*end]))
1129 (*end)++;
1130
1131 return 0;
1132 }
1133
1134 /*
1135 * This function splits the words in buffer->text, stores the list with
1136 * newline separator into out, and saves the offsets of the original words
1137 * in buffer->orig.
1138 */
1139 static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
1140 regex_t *word_regex)
1141 {
1142 int i, j;
1143 long alloc = 0;
1144
1145 out->size = 0;
1146 out->ptr = NULL;
1147
1148 /* fake an empty "0th" word */
1149 ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
1150 buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
1151 buffer->orig_nr = 1;
1152
1153 for (i = 0; i < buffer->text.size; i++) {
1154 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
1155 return;
1156
1157 /* store original boundaries */
1158 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
1159 buffer->orig_alloc);
1160 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
1161 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
1162 buffer->orig_nr++;
1163
1164 /* store one word */
1165 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
1166 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
1167 out->ptr[out->size + j - i] = '\n';
1168 out->size += j - i + 1;
1169
1170 i = j - 1;
1171 }
1172 }
1173
1174 /* this executes the word diff on the accumulated buffers */
1175 static void diff_words_show(struct diff_words_data *diff_words)
1176 {
1177 xpparam_t xpp;
1178 xdemitconf_t xecfg;
1179 mmfile_t minus, plus;
1180 struct diff_words_style *style = diff_words->style;
1181
1182 struct diff_options *opt = diff_words->opt;
1183 const char *line_prefix;
1184
1185 assert(opt);
1186 line_prefix = diff_line_prefix(opt);
1187
1188 /* special case: only removal */
1189 if (!diff_words->plus.text.size) {
1190 fputs(line_prefix, diff_words->opt->file);
1191 fn_out_diff_words_write_helper(diff_words->opt->file,
1192 &style->old, style->newline,
1193 diff_words->minus.text.size,
1194 diff_words->minus.text.ptr, line_prefix);
1195 diff_words->minus.text.size = 0;
1196 return;
1197 }
1198
1199 diff_words->current_plus = diff_words->plus.text.ptr;
1200 diff_words->last_minus = 0;
1201
1202 memset(&xpp, 0, sizeof(xpp));
1203 memset(&xecfg, 0, sizeof(xecfg));
1204 diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
1205 diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
1206 xpp.flags = 0;
1207 /* as only the hunk header will be parsed, we need a 0-context */
1208 xecfg.ctxlen = 0;
1209 if (xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
1210 &xpp, &xecfg))
1211 die("unable to generate word diff");
1212 free(minus.ptr);
1213 free(plus.ptr);
1214 if (diff_words->current_plus != diff_words->plus.text.ptr +
1215 diff_words->plus.text.size) {
1216 if (color_words_output_graph_prefix(diff_words))
1217 fputs(line_prefix, diff_words->opt->file);
1218 fn_out_diff_words_write_helper(diff_words->opt->file,
1219 &style->ctx, style->newline,
1220 diff_words->plus.text.ptr + diff_words->plus.text.size
1221 - diff_words->current_plus, diff_words->current_plus,
1222 line_prefix);
1223 }
1224 diff_words->minus.text.size = diff_words->plus.text.size = 0;
1225 }
1226
1227 /* In "color-words" mode, show word-diff of words accumulated in the buffer */
1228 static void diff_words_flush(struct emit_callback *ecbdata)
1229 {
1230 if (ecbdata->diff_words->minus.text.size ||
1231 ecbdata->diff_words->plus.text.size)
1232 diff_words_show(ecbdata->diff_words);
1233 }
1234
1235 static void diff_filespec_load_driver(struct diff_filespec *one)
1236 {
1237 /* Use already-loaded driver */
1238 if (one->driver)
1239 return;
1240
1241 if (S_ISREG(one->mode))
1242 one->driver = userdiff_find_by_path(one->path);
1243
1244 /* Fallback to default settings */
1245 if (!one->driver)
1246 one->driver = userdiff_find_by_name("default");
1247 }
1248
1249 static const char *userdiff_word_regex(struct diff_filespec *one)
1250 {
1251 diff_filespec_load_driver(one);
1252 return one->driver->word_regex;
1253 }
1254
1255 static void init_diff_words_data(struct emit_callback *ecbdata,
1256 struct diff_options *orig_opts,
1257 struct diff_filespec *one,
1258 struct diff_filespec *two)
1259 {
1260 int i;
1261 struct diff_options *o = xmalloc(sizeof(struct diff_options));
1262 memcpy(o, orig_opts, sizeof(struct diff_options));
1263
1264 ecbdata->diff_words =
1265 xcalloc(1, sizeof(struct diff_words_data));
1266 ecbdata->diff_words->type = o->word_diff;
1267 ecbdata->diff_words->opt = o;
1268 if (!o->word_regex)
1269 o->word_regex = userdiff_word_regex(one);
1270 if (!o->word_regex)
1271 o->word_regex = userdiff_word_regex(two);
1272 if (!o->word_regex)
1273 o->word_regex = diff_word_regex_cfg;
1274 if (o->word_regex) {
1275 ecbdata->diff_words->word_regex = (regex_t *)
1276 xmalloc(sizeof(regex_t));
1277 if (regcomp(ecbdata->diff_words->word_regex,
1278 o->word_regex,
1279 REG_EXTENDED | REG_NEWLINE))
1280 die ("Invalid regular expression: %s",
1281 o->word_regex);
1282 }
1283 for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1284 if (o->word_diff == diff_words_styles[i].type) {
1285 ecbdata->diff_words->style =
1286 &diff_words_styles[i];
1287 break;
1288 }
1289 }
1290 if (want_color(o->use_color)) {
1291 struct diff_words_style *st = ecbdata->diff_words->style;
1292 st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1293 st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1294 st->ctx.color = diff_get_color_opt(o, DIFF_CONTEXT);
1295 }
1296 }
1297
1298 static void free_diff_words_data(struct emit_callback *ecbdata)
1299 {
1300 if (ecbdata->diff_words) {
1301 diff_words_flush(ecbdata);
1302 free (ecbdata->diff_words->opt);
1303 free (ecbdata->diff_words->minus.text.ptr);
1304 free (ecbdata->diff_words->minus.orig);
1305 free (ecbdata->diff_words->plus.text.ptr);
1306 free (ecbdata->diff_words->plus.orig);
1307 if (ecbdata->diff_words->word_regex) {
1308 regfree(ecbdata->diff_words->word_regex);
1309 free(ecbdata->diff_words->word_regex);
1310 }
1311 FREE_AND_NULL(ecbdata->diff_words);
1312 }
1313 }
1314
1315 const char *diff_get_color(int diff_use_color, enum color_diff ix)
1316 {
1317 if (want_color(diff_use_color))
1318 return diff_colors[ix];
1319 return "";
1320 }
1321
1322 const char *diff_line_prefix(struct diff_options *opt)
1323 {
1324 struct strbuf *msgbuf;
1325 if (!opt->output_prefix)
1326 return "";
1327
1328 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
1329 return msgbuf->buf;
1330 }
1331
1332 static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1333 {
1334 const char *cp;
1335 unsigned long allot;
1336 size_t l = len;
1337
1338 if (ecb->truncate)
1339 return ecb->truncate(line, len);
1340 cp = line;
1341 allot = l;
1342 while (0 < l) {
1343 (void) utf8_width(&cp, &l);
1344 if (!cp)
1345 break; /* truncated in the middle? */
1346 }
1347 return allot - l;
1348 }
1349
1350 static void find_lno(const char *line, struct emit_callback *ecbdata)
1351 {
1352 const char *p;
1353 ecbdata->lno_in_preimage = 0;
1354 ecbdata->lno_in_postimage = 0;
1355 p = strchr(line, '-');
1356 if (!p)
1357 return; /* cannot happen */
1358 ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1359 p = strchr(p, '+');
1360 if (!p)
1361 return; /* cannot happen */
1362 ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1363 }
1364
1365 static void fn_out_consume(void *priv, char *line, unsigned long len)
1366 {
1367 struct emit_callback *ecbdata = priv;
1368 const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
1369 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1370 struct diff_options *o = ecbdata->opt;
1371 const char *line_prefix = diff_line_prefix(o);
1372
1373 o->found_changes = 1;
1374
1375 if (ecbdata->header) {
1376 fprintf(o->file, "%s", ecbdata->header->buf);
1377 strbuf_reset(ecbdata->header);
1378 ecbdata->header = NULL;
1379 }
1380
1381 if (ecbdata->label_path[0]) {
1382 const char *name_a_tab, *name_b_tab;
1383
1384 name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
1385 name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
1386
1387 fprintf(o->file, "%s%s--- %s%s%s\n",
1388 line_prefix, meta, ecbdata->label_path[0], reset, name_a_tab);
1389 fprintf(o->file, "%s%s+++ %s%s%s\n",
1390 line_prefix, meta, ecbdata->label_path[1], reset, name_b_tab);
1391 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1392 }
1393
1394 if (diff_suppress_blank_empty
1395 && len == 2 && line[0] == ' ' && line[1] == '\n') {
1396 line[0] = '\n';
1397 len = 1;
1398 }
1399
1400 if (line[0] == '@') {
1401 if (ecbdata->diff_words)
1402 diff_words_flush(ecbdata);
1403 len = sane_truncate_line(ecbdata, line, len);
1404 find_lno(line, ecbdata);
1405 emit_hunk_header(ecbdata, line, len);
1406 return;
1407 }
1408
1409 if (ecbdata->diff_words) {
1410 enum diff_symbol s =
1411 ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN ?
1412 DIFF_SYMBOL_WORDS_PORCELAIN : DIFF_SYMBOL_WORDS;
1413 if (line[0] == '-') {
1414 diff_words_append(line, len,
1415 &ecbdata->diff_words->minus);
1416 return;
1417 } else if (line[0] == '+') {
1418 diff_words_append(line, len,
1419 &ecbdata->diff_words->plus);
1420 return;
1421 } else if (starts_with(line, "\\ ")) {
1422 /*
1423 * Eat the "no newline at eof" marker as if we
1424 * saw a "+" or "-" line with nothing on it,
1425 * and return without diff_words_flush() to
1426 * defer processing. If this is the end of
1427 * preimage, more "+" lines may come after it.
1428 */
1429 return;
1430 }
1431 diff_words_flush(ecbdata);
1432 emit_diff_symbol(o, s, line, len, 0);
1433 return;
1434 }
1435
1436 switch (line[0]) {
1437 case '+':
1438 ecbdata->lno_in_postimage++;
1439 emit_add_line(reset, ecbdata, line + 1, len - 1);
1440 break;
1441 case '-':
1442 ecbdata->lno_in_preimage++;
1443 emit_del_line(reset, ecbdata, line + 1, len - 1);
1444 break;
1445 case ' ':
1446 ecbdata->lno_in_postimage++;
1447 ecbdata->lno_in_preimage++;
1448 emit_context_line(reset, ecbdata, line + 1, len - 1);
1449 break;
1450 default:
1451 /* incomplete line at the end */
1452 ecbdata->lno_in_preimage++;
1453 emit_diff_symbol(o, DIFF_SYMBOL_CONTEXT_INCOMPLETE,
1454 line, len, 0);
1455 break;
1456 }
1457 }
1458
1459 static char *pprint_rename(const char *a, const char *b)
1460 {
1461 const char *old = a;
1462 const char *new = b;
1463 struct strbuf name = STRBUF_INIT;
1464 int pfx_length, sfx_length;
1465 int pfx_adjust_for_slash;
1466 int len_a = strlen(a);
1467 int len_b = strlen(b);
1468 int a_midlen, b_midlen;
1469 int qlen_a = quote_c_style(a, NULL, NULL, 0);
1470 int qlen_b = quote_c_style(b, NULL, NULL, 0);
1471
1472 if (qlen_a || qlen_b) {
1473 quote_c_style(a, &name, NULL, 0);
1474 strbuf_addstr(&name, " => ");
1475 quote_c_style(b, &name, NULL, 0);
1476 return strbuf_detach(&name, NULL);
1477 }
1478
1479 /* Find common prefix */
1480 pfx_length = 0;
1481 while (*old && *new && *old == *new) {
1482 if (*old == '/')
1483 pfx_length = old - a + 1;
1484 old++;
1485 new++;
1486 }
1487
1488 /* Find common suffix */
1489 old = a + len_a;
1490 new = b + len_b;
1491 sfx_length = 0;
1492 /*
1493 * If there is a common prefix, it must end in a slash. In
1494 * that case we let this loop run 1 into the prefix to see the
1495 * same slash.
1496 *
1497 * If there is no common prefix, we cannot do this as it would
1498 * underrun the input strings.
1499 */
1500 pfx_adjust_for_slash = (pfx_length ? 1 : 0);
1501 while (a + pfx_length - pfx_adjust_for_slash <= old &&
1502 b + pfx_length - pfx_adjust_for_slash <= new &&
1503 *old == *new) {
1504 if (*old == '/')
1505 sfx_length = len_a - (old - a);
1506 old--;
1507 new--;
1508 }
1509
1510 /*
1511 * pfx{mid-a => mid-b}sfx
1512 * {pfx-a => pfx-b}sfx
1513 * pfx{sfx-a => sfx-b}
1514 * name-a => name-b
1515 */
1516 a_midlen = len_a - pfx_length - sfx_length;
1517 b_midlen = len_b - pfx_length - sfx_length;
1518 if (a_midlen < 0)
1519 a_midlen = 0;
1520 if (b_midlen < 0)
1521 b_midlen = 0;
1522
1523 strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1524 if (pfx_length + sfx_length) {
1525 strbuf_add(&name, a, pfx_length);
1526 strbuf_addch(&name, '{');
1527 }
1528 strbuf_add(&name, a + pfx_length, a_midlen);
1529 strbuf_addstr(&name, " => ");
1530 strbuf_add(&name, b + pfx_length, b_midlen);
1531 if (pfx_length + sfx_length) {
1532 strbuf_addch(&name, '}');
1533 strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1534 }
1535 return strbuf_detach(&name, NULL);
1536 }
1537
1538 struct diffstat_t {
1539 int nr;
1540 int alloc;
1541 struct diffstat_file {
1542 char *from_name;
1543 char *name;
1544 char *print_name;
1545 unsigned is_unmerged:1;
1546 unsigned is_binary:1;
1547 unsigned is_renamed:1;
1548 unsigned is_interesting:1;
1549 uintmax_t added, deleted;
1550 } **files;
1551 };
1552
1553 static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1554 const char *name_a,
1555 const char *name_b)
1556 {
1557 struct diffstat_file *x;
1558 x = xcalloc(1, sizeof(*x));
1559 ALLOC_GROW(diffstat->files, diffstat->nr + 1, diffstat->alloc);
1560 diffstat->files[diffstat->nr++] = x;
1561 if (name_b) {
1562 x->from_name = xstrdup(name_a);
1563 x->name = xstrdup(name_b);
1564 x->is_renamed = 1;
1565 }
1566 else {
1567 x->from_name = NULL;
1568 x->name = xstrdup(name_a);
1569 }
1570 return x;
1571 }
1572
1573 static void diffstat_consume(void *priv, char *line, unsigned long len)
1574 {
1575 struct diffstat_t *diffstat = priv;
1576 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1577
1578 if (line[0] == '+')
1579 x->added++;
1580 else if (line[0] == '-')
1581 x->deleted++;
1582 }
1583
1584 const char mime_boundary_leader[] = "------------";
1585
1586 static int scale_linear(int it, int width, int max_change)
1587 {
1588 if (!it)
1589 return 0;
1590 /*
1591 * make sure that at least one '-' or '+' is printed if
1592 * there is any change to this path. The easiest way is to
1593 * scale linearly as if the alloted width is one column shorter
1594 * than it is, and then add 1 to the result.
1595 */
1596 return 1 + (it * (width - 1) / max_change);
1597 }
1598
1599 static void show_name(FILE *file,
1600 const char *prefix, const char *name, int len)
1601 {
1602 fprintf(file, " %s%-*s |", prefix, len, name);
1603 }
1604
1605 static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1606 {
1607 if (cnt <= 0)
1608 return;
1609 fprintf(file, "%s", set);
1610 while (cnt--)
1611 putc(ch, file);
1612 fprintf(file, "%s", reset);
1613 }
1614
1615 static void fill_print_name(struct diffstat_file *file)
1616 {
1617 char *pname;
1618
1619 if (file->print_name)
1620 return;
1621
1622 if (!file->is_renamed) {
1623 struct strbuf buf = STRBUF_INIT;
1624 if (quote_c_style(file->name, &buf, NULL, 0)) {
1625 pname = strbuf_detach(&buf, NULL);
1626 } else {
1627 pname = file->name;
1628 strbuf_release(&buf);
1629 }
1630 } else {
1631 pname = pprint_rename(file->from_name, file->name);
1632 }
1633 file->print_name = pname;
1634 }
1635
1636 int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1637 {
1638 struct strbuf sb = STRBUF_INIT;
1639 int ret;
1640
1641 if (!files) {
1642 assert(insertions == 0 && deletions == 0);
1643 return fprintf(fp, "%s\n", " 0 files changed");
1644 }
1645
1646 strbuf_addf(&sb,
1647 (files == 1) ? " %d file changed" : " %d files changed",
1648 files);
1649
1650 /*
1651 * For binary diff, the caller may want to print "x files
1652 * changed" with insertions == 0 && deletions == 0.
1653 *
1654 * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1655 * is probably less confusing (i.e skip over "2 files changed
1656 * but nothing about added/removed lines? Is this a bug in Git?").
1657 */
1658 if (insertions || deletions == 0) {
1659 strbuf_addf(&sb,
1660 (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
1661 insertions);
1662 }
1663
1664 if (deletions || insertions == 0) {
1665 strbuf_addf(&sb,
1666 (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
1667 deletions);
1668 }
1669 strbuf_addch(&sb, '\n');
1670 ret = fputs(sb.buf, fp);
1671 strbuf_release(&sb);
1672 return ret;
1673 }
1674
1675 static void show_stats(struct diffstat_t *data, struct diff_options *options)
1676 {
1677 int i, len, add, del, adds = 0, dels = 0;
1678 uintmax_t max_change = 0, max_len = 0;
1679 int total_files = data->nr, count;
1680 int width, name_width, graph_width, number_width = 0, bin_width = 0;
1681 const char *reset, *add_c, *del_c;
1682 const char *line_prefix = "";
1683 int extra_shown = 0;
1684
1685 if (data->nr == 0)
1686 return;
1687
1688 line_prefix = diff_line_prefix(options);
1689 count = options->stat_count ? options->stat_count : data->nr;
1690
1691 reset = diff_get_color_opt(options, DIFF_RESET);
1692 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1693 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1694
1695 /*
1696 * Find the longest filename and max number of changes
1697 */
1698 for (i = 0; (i < count) && (i < data->nr); i++) {
1699 struct diffstat_file *file = data->files[i];
1700 uintmax_t change = file->added + file->deleted;
1701
1702 if (!file->is_interesting && (change == 0)) {
1703 count++; /* not shown == room for one more */
1704 continue;
1705 }
1706 fill_print_name(file);
1707 len = strlen(file->print_name);
1708 if (max_len < len)
1709 max_len = len;
1710
1711 if (file->is_unmerged) {
1712 /* "Unmerged" is 8 characters */
1713 bin_width = bin_width < 8 ? 8 : bin_width;
1714 continue;
1715 }
1716 if (file->is_binary) {
1717 /* "Bin XXX -> YYY bytes" */
1718 int w = 14 + decimal_width(file->added)
1719 + decimal_width(file->deleted);
1720 bin_width = bin_width < w ? w : bin_width;
1721 /* Display change counts aligned with "Bin" */
1722 number_width = 3;
1723 continue;
1724 }
1725
1726 if (max_change < change)
1727 max_change = change;
1728 }
1729 count = i; /* where we can stop scanning in data->files[] */
1730
1731 /*
1732 * We have width = stat_width or term_columns() columns total.
1733 * We want a maximum of min(max_len, stat_name_width) for the name part.
1734 * We want a maximum of min(max_change, stat_graph_width) for the +- part.
1735 * We also need 1 for " " and 4 + decimal_width(max_change)
1736 * for " | NNNN " and one the empty column at the end, altogether
1737 * 6 + decimal_width(max_change).
1738 *
1739 * If there's not enough space, we will use the smaller of
1740 * stat_name_width (if set) and 5/8*width for the filename,
1741 * and the rest for constant elements + graph part, but no more
1742 * than stat_graph_width for the graph part.
1743 * (5/8 gives 50 for filename and 30 for the constant parts + graph
1744 * for the standard terminal size).
1745 *
1746 * In other words: stat_width limits the maximum width, and
1747 * stat_name_width fixes the maximum width of the filename,
1748 * and is also used to divide available columns if there
1749 * aren't enough.
1750 *
1751 * Binary files are displayed with "Bin XXX -> YYY bytes"
1752 * instead of the change count and graph. This part is treated
1753 * similarly to the graph part, except that it is not
1754 * "scaled". If total width is too small to accommodate the
1755 * guaranteed minimum width of the filename part and the
1756 * separators and this message, this message will "overflow"
1757 * making the line longer than the maximum width.
1758 */
1759
1760 if (options->stat_width == -1)
1761 width = term_columns() - strlen(line_prefix);
1762 else
1763 width = options->stat_width ? options->stat_width : 80;
1764 number_width = decimal_width(max_change) > number_width ?
1765 decimal_width(max_change) : number_width;
1766
1767 if (options->stat_graph_width == -1)
1768 options->stat_graph_width = diff_stat_graph_width;
1769
1770 /*
1771 * Guarantee 3/8*16==6 for the graph part
1772 * and 5/8*16==10 for the filename part
1773 */
1774 if (width < 16 + 6 + number_width)
1775 width = 16 + 6 + number_width;
1776
1777 /*
1778 * First assign sizes that are wanted, ignoring available width.
1779 * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
1780 * starting from "XXX" should fit in graph_width.
1781 */
1782 graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
1783 if (options->stat_graph_width &&
1784 options->stat_graph_width < graph_width)
1785 graph_width = options->stat_graph_width;
1786
1787 name_width = (options->stat_name_width > 0 &&
1788 options->stat_name_width < max_len) ?
1789 options->stat_name_width : max_len;
1790
1791 /*
1792 * Adjust adjustable widths not to exceed maximum width
1793 */
1794 if (name_width + number_width + 6 + graph_width > width) {
1795 if (graph_width > width * 3/8 - number_width - 6) {
1796 graph_width = width * 3/8 - number_width - 6;
1797 if (graph_width < 6)
1798 graph_width = 6;
1799 }
1800
1801 if (options->stat_graph_width &&
1802 graph_width > options->stat_graph_width)
1803 graph_width = options->stat_graph_width;
1804 if (name_width > width - number_width - 6 - graph_width)
1805 name_width = width - number_width - 6 - graph_width;
1806 else
1807 graph_width = width - number_width - 6 - name_width;
1808 }
1809
1810 /*
1811 * From here name_width is the width of the name area,
1812 * and graph_width is the width of the graph area.
1813 * max_change is used to scale graph properly.
1814 */
1815 for (i = 0; i < count; i++) {
1816 const char *prefix = "";
1817 struct diffstat_file *file = data->files[i];
1818 char *name = file->print_name;
1819 uintmax_t added = file->added;
1820 uintmax_t deleted = file->deleted;
1821 int name_len;
1822
1823 if (!file->is_interesting && (added + deleted == 0))
1824 continue;
1825
1826 /*
1827 * "scale" the filename
1828 */
1829 len = name_width;
1830 name_len = strlen(name);
1831 if (name_width < name_len) {
1832 char *slash;
1833 prefix = "...";
1834 len -= 3;
1835 name += name_len - len;
1836 slash = strchr(name, '/');
1837 if (slash)
1838 name = slash;
1839 }
1840
1841 if (file->is_binary) {
1842 fprintf(options->file, "%s", line_prefix);
1843 show_name(options->file, prefix, name, len);
1844 fprintf(options->file, " %*s", number_width, "Bin");
1845 if (!added && !deleted) {
1846 putc('\n', options->file);
1847 continue;
1848 }
1849 fprintf(options->file, " %s%"PRIuMAX"%s",
1850 del_c, deleted, reset);
1851 fprintf(options->file, " -> ");
1852 fprintf(options->file, "%s%"PRIuMAX"%s",
1853 add_c, added, reset);
1854 fprintf(options->file, " bytes");
1855 fprintf(options->file, "\n");
1856 continue;
1857 }
1858 else if (file->is_unmerged) {
1859 fprintf(options->file, "%s", line_prefix);
1860 show_name(options->file, prefix, name, len);
1861 fprintf(options->file, " Unmerged\n");
1862 continue;
1863 }
1864
1865 /*
1866 * scale the add/delete
1867 */
1868 add = added;
1869 del = deleted;
1870
1871 if (graph_width <= max_change) {
1872 int total = scale_linear(add + del, graph_width, max_change);
1873 if (total < 2 && add && del)
1874 /* width >= 2 due to the sanity check */
1875 total = 2;
1876 if (add < del) {
1877 add = scale_linear(add, graph_width, max_change);
1878 del = total - add;
1879 } else {
1880 del = scale_linear(del, graph_width, max_change);
1881 add = total - del;
1882 }
1883 }
1884 fprintf(options->file, "%s", line_prefix);
1885 show_name(options->file, prefix, name, len);
1886 fprintf(options->file, " %*"PRIuMAX"%s",
1887 number_width, added + deleted,
1888 added + deleted ? " " : "");
1889 show_graph(options->file, '+', add, add_c, reset);
1890 show_graph(options->file, '-', del, del_c, reset);
1891 fprintf(options->file, "\n");
1892 }
1893
1894 for (i = 0; i < data->nr; i++) {
1895 struct diffstat_file *file = data->files[i];
1896 uintmax_t added = file->added;
1897 uintmax_t deleted = file->deleted;
1898
1899 if (file->is_unmerged ||
1900 (!file->is_interesting && (added + deleted == 0))) {
1901 total_files--;
1902 continue;
1903 }
1904
1905 if (!file->is_binary) {
1906 adds += added;
1907 dels += deleted;
1908 }
1909 if (i < count)
1910 continue;
1911 if (!extra_shown)
1912 fprintf(options->file, "%s ...\n", line_prefix);
1913 extra_shown = 1;
1914 }
1915 fprintf(options->file, "%s", line_prefix);
1916 print_stat_summary(options->file, total_files, adds, dels);
1917 }
1918
1919 static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1920 {
1921 int i, adds = 0, dels = 0, total_files = data->nr;
1922
1923 if (data->nr == 0)
1924 return;
1925
1926 for (i = 0; i < data->nr; i++) {
1927 int added = data->files[i]->added;
1928 int deleted= data->files[i]->deleted;
1929
1930 if (data->files[i]->is_unmerged ||
1931 (!data->files[i]->is_interesting && (added + deleted == 0))) {
1932 total_files--;
1933 } else if (!data->files[i]->is_binary) { /* don't count bytes */
1934 adds += added;
1935 dels += deleted;
1936 }
1937 }
1938 fprintf(options->file, "%s", diff_line_prefix(options));
1939 print_stat_summary(options->file, total_files, adds, dels);
1940 }
1941
1942 static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1943 {
1944 int i;
1945
1946 if (data->nr == 0)
1947 return;
1948
1949 for (i = 0; i < data->nr; i++) {
1950 struct diffstat_file *file = data->files[i];
1951
1952 fprintf(options->file, "%s", diff_line_prefix(options));
1953
1954 if (file->is_binary)
1955 fprintf(options->file, "-\t-\t");
1956 else
1957 fprintf(options->file,
1958 "%"PRIuMAX"\t%"PRIuMAX"\t",
1959 file->added, file->deleted);
1960 if (options->line_termination) {
1961 fill_print_name(file);
1962 if (!file->is_renamed)
1963 write_name_quoted(file->name, options->file,
1964 options->line_termination);
1965 else {
1966 fputs(file->print_name, options->file);
1967 putc(options->line_termination, options->file);
1968 }
1969 } else {
1970 if (file->is_renamed) {
1971 putc('\0', options->file);
1972 write_name_quoted(file->from_name, options->file, '\0');
1973 }
1974 write_name_quoted(file->name, options->file, '\0');
1975 }
1976 }
1977 }
1978
1979 struct dirstat_file {
1980 const char *name;
1981 unsigned long changed;
1982 };
1983
1984 struct dirstat_dir {
1985 struct dirstat_file *files;
1986 int alloc, nr, permille, cumulative;
1987 };
1988
1989 static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1990 unsigned long changed, const char *base, int baselen)
1991 {
1992 unsigned long this_dir = 0;
1993 unsigned int sources = 0;
1994 const char *line_prefix = diff_line_prefix(opt);
1995
1996 while (dir->nr) {
1997 struct dirstat_file *f = dir->files;
1998 int namelen = strlen(f->name);
1999 unsigned long this;
2000 char *slash;
2001
2002 if (namelen < baselen)
2003 break;
2004 if (memcmp(f->name, base, baselen))
2005 break;
2006 slash = strchr(f->name + baselen, '/');
2007 if (slash) {
2008 int newbaselen = slash + 1 - f->name;
2009 this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
2010 sources++;
2011 } else {
2012 this = f->changed;
2013 dir->files++;
2014 dir->nr--;
2015 sources += 2;
2016 }
2017 this_dir += this;
2018 }
2019
2020 /*
2021 * We don't report dirstat's for
2022 * - the top level
2023 * - or cases where everything came from a single directory
2024 * under this directory (sources == 1).
2025 */
2026 if (baselen && sources != 1) {
2027 if (this_dir) {
2028 int permille = this_dir * 1000 / changed;
2029 if (permille >= dir->permille) {
2030 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
2031 permille / 10, permille % 10, baselen, base);
2032 if (!dir->cumulative)
2033 return 0;
2034 }
2035 }
2036 }
2037 return this_dir;
2038 }
2039
2040 static int dirstat_compare(const void *_a, const void *_b)
2041 {
2042 const struct dirstat_file *a = _a;
2043 const struct dirstat_file *b = _b;
2044 return strcmp(a->name, b->name);
2045 }
2046
2047 static void show_dirstat(struct diff_options *options)
2048 {
2049 int i;
2050 unsigned long changed;
2051 struct dirstat_dir dir;
2052 struct diff_queue_struct *q = &diff_queued_diff;
2053
2054 dir.files = NULL;
2055 dir.alloc = 0;
2056 dir.nr = 0;
2057 dir.permille = options->dirstat_permille;
2058 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
2059
2060 changed = 0;
2061 for (i = 0; i < q->nr; i++) {
2062 struct diff_filepair *p = q->queue[i];
2063 const char *name;
2064 unsigned long copied, added, damage;
2065 int content_changed;
2066
2067 name = p->two->path ? p->two->path : p->one->path;
2068
2069 if (p->one->oid_valid && p->two->oid_valid)
2070 content_changed = oidcmp(&p->one->oid, &p->two->oid);
2071 else
2072 content_changed = 1;
2073
2074 if (!content_changed) {
2075 /*
2076 * The SHA1 has not changed, so pre-/post-content is
2077 * identical. We can therefore skip looking at the
2078 * file contents altogether.
2079 */
2080 damage = 0;
2081 goto found_damage;
2082 }
2083
2084 if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
2085 /*
2086 * In --dirstat-by-file mode, we don't really need to
2087 * look at the actual file contents at all.
2088 * The fact that the SHA1 changed is enough for us to
2089 * add this file to the list of results
2090 * (with each file contributing equal damage).
2091 */
2092 damage = 1;
2093 goto found_damage;
2094 }
2095
2096 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
2097 diff_populate_filespec(p->one, 0);
2098 diff_populate_filespec(p->two, 0);
2099 diffcore_count_changes(p->one, p->two, NULL, NULL,
2100 &copied, &added);
2101 diff_free_filespec_data(p->one);
2102 diff_free_filespec_data(p->two);
2103 } else if (DIFF_FILE_VALID(p->one)) {
2104 diff_populate_filespec(p->one, CHECK_SIZE_ONLY);
2105 copied = added = 0;
2106 diff_free_filespec_data(p->one);
2107 } else if (DIFF_FILE_VALID(p->two)) {
2108 diff_populate_filespec(p->two, CHECK_SIZE_ONLY);
2109 copied = 0;
2110 added = p->two->size;
2111 diff_free_filespec_data(p->two);
2112 } else
2113 continue;
2114
2115 /*
2116 * Original minus copied is the removed material,
2117 * added is the new material. They are both damages
2118 * made to the preimage.
2119 * If the resulting damage is zero, we know that
2120 * diffcore_count_changes() considers the two entries to
2121 * be identical, but since content_changed is true, we
2122 * know that there must have been _some_ kind of change,
2123 * so we force all entries to have damage > 0.
2124 */
2125 damage = (p->one->size - copied) + added;
2126 if (!damage)
2127 damage = 1;
2128
2129 found_damage:
2130 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2131 dir.files[dir.nr].name = name;
2132 dir.files[dir.nr].changed = damage;
2133 changed += damage;
2134 dir.nr++;
2135 }
2136
2137 /* This can happen even with many files, if everything was renames */
2138 if (!changed)
2139 return;
2140
2141 /* Show all directories with more than x% of the changes */
2142 QSORT(dir.files, dir.nr, dirstat_compare);
2143 gather_dirstat(options, &dir, changed, "", 0);
2144 }
2145
2146 static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
2147 {
2148 int i;
2149 unsigned long changed;
2150 struct dirstat_dir dir;
2151
2152 if (data->nr == 0)
2153 return;
2154
2155 dir.files = NULL;
2156 dir.alloc = 0;
2157 dir.nr = 0;
2158 dir.permille = options->dirstat_permille;
2159 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
2160
2161 changed = 0;
2162 for (i = 0; i < data->nr; i++) {
2163 struct diffstat_file *file = data->files[i];
2164 unsigned long damage = file->added + file->deleted;
2165 if (file->is_binary)
2166 /*
2167 * binary files counts bytes, not lines. Must find some
2168 * way to normalize binary bytes vs. textual lines.
2169 * The following heuristic assumes that there are 64
2170 * bytes per "line".
2171 * This is stupid and ugly, but very cheap...
2172 */
2173 damage = (damage + 63) / 64;
2174 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2175 dir.files[dir.nr].name = file->name;
2176 dir.files[dir.nr].changed = damage;
2177 changed += damage;
2178 dir.nr++;
2179 }
2180
2181 /* This can happen even with many files, if everything was renames */
2182 if (!changed)
2183 return;
2184
2185 /* Show all directories with more than x% of the changes */
2186 QSORT(dir.files, dir.nr, dirstat_compare);
2187 gather_dirstat(options, &dir, changed, "", 0);
2188 }
2189
2190 static void free_diffstat_info(struct diffstat_t *diffstat)
2191 {
2192 int i;
2193 for (i = 0; i < diffstat->nr; i++) {
2194 struct diffstat_file *f = diffstat->files[i];
2195 if (f->name != f->print_name)
2196 free(f->print_name);
2197 free(f->name);
2198 free(f->from_name);
2199 free(f);
2200 }
2201 free(diffstat->files);
2202 }
2203
2204 struct checkdiff_t {
2205 const char *filename;
2206 int lineno;
2207 int conflict_marker_size;
2208 struct diff_options *o;
2209 unsigned ws_rule;
2210 unsigned status;
2211 };
2212
2213 static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2214 {
2215 char firstchar;
2216 int cnt;
2217
2218 if (len < marker_size + 1)
2219 return 0;
2220 firstchar = line[0];
2221 switch (firstchar) {
2222 case '=': case '>': case '<': case '|':
2223 break;
2224 default:
2225 return 0;
2226 }
2227 for (cnt = 1; cnt < marker_size; cnt++)
2228 if (line[cnt] != firstchar)
2229 return 0;
2230 /* line[1] thru line[marker_size-1] are same as firstchar */
2231 if (len < marker_size + 1 || !isspace(line[marker_size]))
2232 return 0;
2233 return 1;
2234 }
2235
2236 static void checkdiff_consume(void *priv, char *line, unsigned long len)
2237 {
2238 struct checkdiff_t *data = priv;
2239 int marker_size = data->conflict_marker_size;
2240 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2241 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2242 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2243 char *err;
2244 const char *line_prefix;
2245
2246 assert(data->o);
2247 line_prefix = diff_line_prefix(data->o);
2248
2249 if (line[0] == '+') {
2250 unsigned bad;
2251 data->lineno++;
2252 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2253 data->status |= 1;
2254 fprintf(data->o->file,
2255 "%s%s:%d: leftover conflict marker\n",
2256 line_prefix, data->filename, data->lineno);
2257 }
2258 bad = ws_check(line + 1, len - 1, data->ws_rule);
2259 if (!bad)
2260 return;
2261 data->status |= bad;
2262 err = whitespace_error_string(bad);
2263 fprintf(data->o->file, "%s%s:%d: %s.\n",
2264 line_prefix, data->filename, data->lineno, err);
2265 free(err);
2266 emit_line(data->o, set, reset, line, 1);
2267 ws_check_emit(line + 1, len - 1, data->ws_rule,
2268 data->o->file, set, reset, ws);
2269 } else if (line[0] == ' ') {
2270 data->lineno++;
2271 } else if (line[0] == '@') {
2272 char *plus = strchr(line, '+');
2273 if (plus)
2274 data->lineno = strtol(plus, NULL, 10) - 1;
2275 else
2276 die("invalid diff");
2277 }
2278 }
2279
2280 static unsigned char *deflate_it(char *data,
2281 unsigned long size,
2282 unsigned long *result_size)
2283 {
2284 int bound;
2285 unsigned char *deflated;
2286 git_zstream stream;
2287
2288 git_deflate_init(&stream, zlib_compression_level);
2289 bound = git_deflate_bound(&stream, size);
2290 deflated = xmalloc(bound);
2291 stream.next_out = deflated;
2292 stream.avail_out = bound;
2293
2294 stream.next_in = (unsigned char *)data;
2295 stream.avail_in = size;
2296 while (git_deflate(&stream, Z_FINISH) == Z_OK)
2297 ; /* nothing */
2298 git_deflate_end(&stream);
2299 *result_size = stream.total_out;
2300 return deflated;
2301 }
2302
2303 static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two,
2304 const char *prefix)
2305 {
2306 void *cp;
2307 void *delta;
2308 void *deflated;
2309 void *data;
2310 unsigned long orig_size;
2311 unsigned long delta_size;
2312 unsigned long deflate_size;
2313 unsigned long data_size;
2314
2315 /* We could do deflated delta, or we could do just deflated two,
2316 * whichever is smaller.
2317 */
2318 delta = NULL;
2319 deflated = deflate_it(two->ptr, two->size, &deflate_size);
2320 if (one->size && two->size) {
2321 delta = diff_delta(one->ptr, one->size,
2322 two->ptr, two->size,
2323 &delta_size, deflate_size);
2324 if (delta) {
2325 void *to_free = delta;
2326 orig_size = delta_size;
2327 delta = deflate_it(delta, delta_size, &delta_size);
2328 free(to_free);
2329 }
2330 }
2331
2332 if (delta && delta_size < deflate_size) {
2333 fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2334 free(deflated);
2335 data = delta;
2336 data_size = delta_size;
2337 }
2338 else {
2339 fprintf(file, "%sliteral %lu\n", prefix, two->size);
2340 free(delta);
2341 data = deflated;
2342 data_size = deflate_size;
2343 }
2344
2345 /* emit data encoded in base85 */
2346 cp = data;
2347 while (data_size) {
2348 int bytes = (52 < data_size) ? 52 : data_size;
2349 char line[70];
2350 data_size -= bytes;
2351 if (bytes <= 26)
2352 line[0] = bytes + 'A' - 1;
2353 else
2354 line[0] = bytes - 26 + 'a' - 1;
2355 encode_85(line + 1, cp, bytes);
2356 cp = (char *) cp + bytes;
2357 fprintf(file, "%s", prefix);
2358 fputs(line, file);
2359 fputc('\n', file);
2360 }
2361 fprintf(file, "%s\n", prefix);
2362 free(data);
2363 }
2364
2365 static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two,
2366 const char *prefix)
2367 {
2368 fprintf(file, "%sGIT binary patch\n", prefix);
2369 emit_binary_diff_body(file, one, two, prefix);
2370 emit_binary_diff_body(file, two, one, prefix);
2371 }
2372
2373 int diff_filespec_is_binary(struct diff_filespec *one)
2374 {
2375 if (one->is_binary == -1) {
2376 diff_filespec_load_driver(one);
2377 if (one->driver->binary != -1)
2378 one->is_binary = one->driver->binary;
2379 else {
2380 if (!one->data && DIFF_FILE_VALID(one))
2381 diff_populate_filespec(one, CHECK_BINARY);
2382 if (one->is_binary == -1 && one->data)
2383 one->is_binary = buffer_is_binary(one->data,
2384 one->size);
2385 if (one->is_binary == -1)
2386 one->is_binary = 0;
2387 }
2388 }
2389 return one->is_binary;
2390 }
2391
2392 static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2393 {
2394 diff_filespec_load_driver(one);
2395 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2396 }
2397
2398 void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2399 {
2400 if (!options->a_prefix)
2401 options->a_prefix = a;
2402 if (!options->b_prefix)
2403 options->b_prefix = b;
2404 }
2405
2406 struct userdiff_driver *get_textconv(struct diff_filespec *one)
2407 {
2408 if (!DIFF_FILE_VALID(one))
2409 return NULL;
2410
2411 diff_filespec_load_driver(one);
2412 return userdiff_get_textconv(one->driver);
2413 }
2414
2415 static void builtin_diff(const char *name_a,
2416 const char *name_b,
2417 struct diff_filespec *one,
2418 struct diff_filespec *two,
2419 const char *xfrm_msg,
2420 int must_show_header,
2421 struct diff_options *o,
2422 int complete_rewrite)
2423 {
2424 mmfile_t mf1, mf2;
2425 const char *lbl[2];
2426 char *a_one, *b_two;
2427 const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
2428 const char *reset = diff_get_color_opt(o, DIFF_RESET);
2429 const char *a_prefix, *b_prefix;
2430 struct userdiff_driver *textconv_one = NULL;
2431 struct userdiff_driver *textconv_two = NULL;
2432 struct strbuf header = STRBUF_INIT;
2433 const char *line_prefix = diff_line_prefix(o);
2434
2435 diff_set_mnemonic_prefix(o, "a/", "b/");
2436 if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2437 a_prefix = o->b_prefix;
2438 b_prefix = o->a_prefix;
2439 } else {
2440 a_prefix = o->a_prefix;
2441 b_prefix = o->b_prefix;
2442 }
2443
2444 if (o->submodule_format == DIFF_SUBMODULE_LOG &&
2445 (!one->mode || S_ISGITLINK(one->mode)) &&
2446 (!two->mode || S_ISGITLINK(two->mode))) {
2447 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2448 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2449 show_submodule_summary(o->file, one->path ? one->path : two->path,
2450 line_prefix,
2451 &one->oid, &two->oid,
2452 two->dirty_submodule,
2453 meta, del, add, reset);
2454 return;
2455 } else if (o->submodule_format == DIFF_SUBMODULE_INLINE_DIFF &&
2456 (!one->mode || S_ISGITLINK(one->mode)) &&
2457 (!two->mode || S_ISGITLINK(two->mode))) {
2458 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2459 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2460 show_submodule_inline_diff(o->file, one->path ? one->path : two->path,
2461 line_prefix,
2462 &one->oid, &two->oid,
2463 two->dirty_submodule,
2464 meta, del, add, reset, o);
2465 return;
2466 }
2467
2468 if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2469 textconv_one = get_textconv(one);
2470 textconv_two = get_textconv(two);
2471 }
2472
2473 /* Never use a non-valid filename anywhere if at all possible */
2474 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2475 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2476
2477 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2478 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2479 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2480 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2481 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
2482 if (lbl[0][0] == '/') {
2483 /* /dev/null */
2484 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
2485 if (xfrm_msg)
2486 strbuf_addstr(&header, xfrm_msg);
2487 must_show_header = 1;
2488 }
2489 else if (lbl[1][0] == '/') {
2490 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
2491 if (xfrm_msg)
2492 strbuf_addstr(&header, xfrm_msg);
2493 must_show_header = 1;
2494 }
2495 else {
2496 if (one->mode != two->mode) {
2497 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
2498 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
2499 must_show_header = 1;
2500 }
2501 if (xfrm_msg)
2502 strbuf_addstr(&header, xfrm_msg);
2503
2504 /*
2505 * we do not run diff between different kind
2506 * of objects.
2507 */
2508 if ((one->mode ^ two->mode) & S_IFMT)
2509 goto free_ab_and_return;
2510 if (complete_rewrite &&
2511 (textconv_one || !diff_filespec_is_binary(one)) &&
2512 (textconv_two || !diff_filespec_is_binary(two))) {
2513 fprintf(o->file, "%s", header.buf);
2514 strbuf_reset(&header);
2515 emit_rewrite_diff(name_a, name_b, one, two,
2516 textconv_one, textconv_two, o);
2517 o->found_changes = 1;
2518 goto free_ab_and_return;
2519 }
2520 }
2521
2522 if (o->irreversible_delete && lbl[1][0] == '/') {
2523 fprintf(o->file, "%s", header.buf);
2524 strbuf_reset(&header);
2525 goto free_ab_and_return;
2526 } else if (!DIFF_OPT_TST(o, TEXT) &&
2527 ( (!textconv_one && diff_filespec_is_binary(one)) ||
2528 (!textconv_two && diff_filespec_is_binary(two)) )) {
2529 if (!one->data && !two->data &&
2530 S_ISREG(one->mode) && S_ISREG(two->mode) &&
2531 !DIFF_OPT_TST(o, BINARY)) {
2532 if (!oidcmp(&one->oid, &two->oid)) {
2533 if (must_show_header)
2534 fprintf(o->file, "%s", header.buf);
2535 goto free_ab_and_return;
2536 }
2537 fprintf(o->file, "%s", header.buf);
2538 fprintf(o->file, "%sBinary files %s and %s differ\n",
2539 line_prefix, lbl[0], lbl[1]);
2540 goto free_ab_and_return;
2541 }
2542 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2543 die("unable to read files to diff");
2544 /* Quite common confusing case */
2545 if (mf1.size == mf2.size &&
2546 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2547 if (must_show_header)
2548 fprintf(o->file, "%s", header.buf);
2549 goto free_ab_and_return;
2550 }
2551 fprintf(o->file, "%s", header.buf);
2552 strbuf_reset(&header);
2553 if (DIFF_OPT_TST(o, BINARY))
2554 emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2555 else
2556 fprintf(o->file, "%sBinary files %s and %s differ\n",
2557 line_prefix, lbl[0], lbl[1]);
2558 o->found_changes = 1;
2559 } else {
2560 /* Crazy xdl interfaces.. */
2561 const char *diffopts = getenv("GIT_DIFF_OPTS");
2562 const char *v;
2563 xpparam_t xpp;
2564 xdemitconf_t xecfg;
2565 struct emit_callback ecbdata;
2566 const struct userdiff_funcname *pe;
2567
2568 if (must_show_header) {
2569 fprintf(o->file, "%s", header.buf);
2570 strbuf_reset(&header);
2571 }
2572
2573 mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2574 mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2575
2576 pe = diff_funcname_pattern(one);
2577 if (!pe)
2578 pe = diff_funcname_pattern(two);
2579
2580 memset(&xpp, 0, sizeof(xpp));
2581 memset(&xecfg, 0, sizeof(xecfg));
2582 memset(&ecbdata, 0, sizeof(ecbdata));
2583 ecbdata.label_path = lbl;
2584 ecbdata.color_diff = want_color(o->use_color);
2585 ecbdata.ws_rule = whitespace_rule(name_b);
2586 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2587 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2588 ecbdata.opt = o;
2589 ecbdata.header = header.len ? &header : NULL;
2590 xpp.flags = o->xdl_opts;
2591 xecfg.ctxlen = o->context;
2592 xecfg.interhunkctxlen = o->interhunkcontext;
2593 xecfg.flags = XDL_EMIT_FUNCNAMES;
2594 if (DIFF_OPT_TST(o, FUNCCONTEXT))
2595 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2596 if (pe)
2597 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2598 if (!diffopts)
2599 ;
2600 else if (skip_prefix(diffopts, "--unified=", &v))
2601 xecfg.ctxlen = strtoul(v, NULL, 10);
2602 else if (skip_prefix(diffopts, "-u", &v))
2603 xecfg.ctxlen = strtoul(v, NULL, 10);
2604 if (o->word_diff)
2605 init_diff_words_data(&ecbdata, o, one, two);
2606 if (xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2607 &xpp, &xecfg))
2608 die("unable to generate diff for %s", one->path);
2609 if (o->word_diff)
2610 free_diff_words_data(&ecbdata);
2611 if (textconv_one)
2612 free(mf1.ptr);
2613 if (textconv_two)
2614 free(mf2.ptr);
2615 xdiff_clear_find_func(&xecfg);
2616 }
2617
2618 free_ab_and_return:
2619 strbuf_release(&header);
2620 diff_free_filespec_data(one);
2621 diff_free_filespec_data(two);
2622 free(a_one);
2623 free(b_two);
2624 return;
2625 }
2626
2627 static void builtin_diffstat(const char *name_a, const char *name_b,
2628 struct diff_filespec *one,
2629 struct diff_filespec *two,
2630 struct diffstat_t *diffstat,
2631 struct diff_options *o,
2632 struct diff_filepair *p)
2633 {
2634 mmfile_t mf1, mf2;
2635 struct diffstat_file *data;
2636 int same_contents;
2637 int complete_rewrite = 0;
2638
2639 if (!DIFF_PAIR_UNMERGED(p)) {
2640 if (p->status == DIFF_STATUS_MODIFIED && p->score)
2641 complete_rewrite = 1;
2642 }
2643
2644 data = diffstat_add(diffstat, name_a, name_b);
2645 data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
2646
2647 if (!one || !two) {
2648 data->is_unmerged = 1;
2649 return;
2650 }
2651
2652 same_contents = !oidcmp(&one->oid, &two->oid);
2653
2654 if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2655 data->is_binary = 1;
2656 if (same_contents) {
2657 data->added = 0;
2658 data->deleted = 0;
2659 } else {
2660 data->added = diff_filespec_size(two);
2661 data->deleted = diff_filespec_size(one);
2662 }
2663 }
2664
2665 else if (complete_rewrite) {
2666 diff_populate_filespec(one, 0);
2667 diff_populate_filespec(two, 0);
2668 data->deleted = count_lines(one->data, one->size);
2669 data->added = count_lines(two->data, two->size);
2670 }
2671
2672 else if (!same_contents) {
2673 /* Crazy xdl interfaces.. */
2674 xpparam_t xpp;
2675 xdemitconf_t xecfg;
2676
2677 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2678 die("unable to read files to diff");
2679
2680 memset(&xpp, 0, sizeof(xpp));
2681 memset(&xecfg, 0, sizeof(xecfg));
2682 xpp.flags = o->xdl_opts;
2683 xecfg.ctxlen = o->context;
2684 xecfg.interhunkctxlen = o->interhunkcontext;
2685 if (xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2686 &xpp, &xecfg))
2687 die("unable to generate diffstat for %s", one->path);
2688 }
2689
2690 diff_free_filespec_data(one);
2691 diff_free_filespec_data(two);
2692 }
2693
2694 static void builtin_checkdiff(const char *name_a, const char *name_b,
2695 const char *attr_path,
2696 struct diff_filespec *one,
2697 struct diff_filespec *two,
2698 struct diff_options *o)
2699 {
2700 mmfile_t mf1, mf2;
2701 struct checkdiff_t data;
2702
2703 if (!two)
2704 return;
2705
2706 memset(&data, 0, sizeof(data));
2707 data.filename = name_b ? name_b : name_a;
2708 data.lineno = 0;
2709 data.o = o;
2710 data.ws_rule = whitespace_rule(attr_path);
2711 data.conflict_marker_size = ll_merge_marker_size(attr_path);
2712
2713 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2714 die("unable to read files to diff");
2715
2716 /*
2717 * All the other codepaths check both sides, but not checking
2718 * the "old" side here is deliberate. We are checking the newly
2719 * introduced changes, and as long as the "new" side is text, we
2720 * can and should check what it introduces.
2721 */
2722 if (diff_filespec_is_binary(two))
2723 goto free_and_return;
2724 else {
2725 /* Crazy xdl interfaces.. */
2726 xpparam_t xpp;
2727 xdemitconf_t xecfg;
2728
2729 memset(&xpp, 0, sizeof(xpp));
2730 memset(&xecfg, 0, sizeof(xecfg));
2731 xecfg.ctxlen = 1; /* at least one context line */
2732 xpp.flags = 0;
2733 if (xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2734 &xpp, &xecfg))
2735 die("unable to generate checkdiff for %s", one->path);
2736
2737 if (data.ws_rule & WS_BLANK_AT_EOF) {
2738 struct emit_callback ecbdata;
2739 int blank_at_eof;
2740
2741 ecbdata.ws_rule = data.ws_rule;
2742 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2743 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2744
2745 if (blank_at_eof) {
2746 static char *err;
2747 if (!err)
2748 err = whitespace_error_string(WS_BLANK_AT_EOF);
2749 fprintf(o->file, "%s:%d: %s.\n",
2750 data.filename, blank_at_eof, err);
2751 data.status = 1; /* report errors */
2752 }
2753 }
2754 }
2755 free_and_return:
2756 diff_free_filespec_data(one);
2757 diff_free_filespec_data(two);
2758 if (data.status)
2759 DIFF_OPT_SET(o, CHECK_FAILED);
2760 }
2761
2762 struct diff_filespec *alloc_filespec(const char *path)
2763 {
2764 struct diff_filespec *spec;
2765
2766 FLEXPTR_ALLOC_STR(spec, path, path);
2767 spec->count = 1;
2768 spec->is_binary = -1;
2769 return spec;
2770 }
2771
2772 void free_filespec(struct diff_filespec *spec)
2773 {
2774 if (!--spec->count) {
2775 diff_free_filespec_data(spec);
2776 free(spec);
2777 }
2778 }
2779
2780 void fill_filespec(struct diff_filespec *spec, const struct object_id *oid,
2781 int oid_valid, unsigned short mode)
2782 {
2783 if (mode) {
2784 spec->mode = canon_mode(mode);
2785 oidcpy(&spec->oid, oid);
2786 spec->oid_valid = oid_valid;
2787 }
2788 }
2789
2790 /*
2791 * Given a name and sha1 pair, if the index tells us the file in
2792 * the work tree has that object contents, return true, so that
2793 * prepare_temp_file() does not have to inflate and extract.
2794 */
2795 static int reuse_worktree_file(const char *name, const struct object_id *oid, int want_file)
2796 {
2797 const struct cache_entry *ce;
2798 struct stat st;
2799 int pos, len;
2800
2801 /*
2802 * We do not read the cache ourselves here, because the
2803 * benchmark with my previous version that always reads cache
2804 * shows that it makes things worse for diff-tree comparing
2805 * two linux-2.6 kernel trees in an already checked out work
2806 * tree. This is because most diff-tree comparisons deal with
2807 * only a small number of files, while reading the cache is
2808 * expensive for a large project, and its cost outweighs the
2809 * savings we get by not inflating the object to a temporary
2810 * file. Practically, this code only helps when we are used
2811 * by diff-cache --cached, which does read the cache before
2812 * calling us.
2813 */
2814 if (!active_cache)
2815 return 0;
2816
2817 /* We want to avoid the working directory if our caller
2818 * doesn't need the data in a normal file, this system
2819 * is rather slow with its stat/open/mmap/close syscalls,
2820 * and the object is contained in a pack file. The pack
2821 * is probably already open and will be faster to obtain
2822 * the data through than the working directory. Loose
2823 * objects however would tend to be slower as they need
2824 * to be individually opened and inflated.
2825 */
2826 if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(oid->hash))
2827 return 0;
2828
2829 /*
2830 * Similarly, if we'd have to convert the file contents anyway, that
2831 * makes the optimization not worthwhile.
2832 */
2833 if (!want_file && would_convert_to_git(&the_index, name))
2834 return 0;
2835
2836 len = strlen(name);
2837 pos = cache_name_pos(name, len);
2838 if (pos < 0)
2839 return 0;
2840 ce = active_cache[pos];
2841
2842 /*
2843 * This is not the sha1 we are looking for, or
2844 * unreusable because it is not a regular file.
2845 */
2846 if (oidcmp(oid, &ce->oid) || !S_ISREG(ce->ce_mode))
2847 return 0;
2848
2849 /*
2850 * If ce is marked as "assume unchanged", there is no
2851 * guarantee that work tree matches what we are looking for.
2852 */
2853 if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2854 return 0;
2855
2856 /*
2857 * If ce matches the file in the work tree, we can reuse it.
2858 */
2859 if (ce_uptodate(ce) ||
2860 (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2861 return 1;
2862
2863 return 0;
2864 }
2865
2866 static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2867 {
2868 struct strbuf buf = STRBUF_INIT;
2869 char *dirty = "";
2870
2871 /* Are we looking at the work tree? */
2872 if (s->dirty_submodule)
2873 dirty = "-dirty";
2874
2875 strbuf_addf(&buf, "Subproject commit %s%s\n",
2876 oid_to_hex(&s->oid), dirty);
2877 s->size = buf.len;
2878 if (size_only) {
2879 s->data = NULL;
2880 strbuf_release(&buf);
2881 } else {
2882 s->data = strbuf_detach(&buf, NULL);
2883 s->should_free = 1;
2884 }
2885 return 0;
2886 }
2887
2888 /*
2889 * While doing rename detection and pickaxe operation, we may need to
2890 * grab the data for the blob (or file) for our own in-core comparison.
2891 * diff_filespec has data and size fields for this purpose.
2892 */
2893 int diff_populate_filespec(struct diff_filespec *s, unsigned int flags)
2894 {
2895 int size_only = flags & CHECK_SIZE_ONLY;
2896 int err = 0;
2897 /*
2898 * demote FAIL to WARN to allow inspecting the situation
2899 * instead of refusing.
2900 */
2901 enum safe_crlf crlf_warn = (safe_crlf == SAFE_CRLF_FAIL
2902 ? SAFE_CRLF_WARN
2903 : safe_crlf);
2904
2905 if (!DIFF_FILE_VALID(s))
2906 die("internal error: asking to populate invalid file.");
2907 if (S_ISDIR(s->mode))
2908 return -1;
2909
2910 if (s->data)
2911 return 0;
2912
2913 if (size_only && 0 < s->size)
2914 return 0;
2915
2916 if (S_ISGITLINK(s->mode))
2917 return diff_populate_gitlink(s, size_only);
2918
2919 if (!s->oid_valid ||
2920 reuse_worktree_file(s->path, &s->oid, 0)) {
2921 struct strbuf buf = STRBUF_INIT;
2922 struct stat st;
2923 int fd;
2924
2925 if (lstat(s->path, &st) < 0) {
2926 if (errno == ENOENT) {
2927 err_empty:
2928 err = -1;
2929 empty:
2930 s->data = (char *)"";
2931 s->size = 0;
2932 return err;
2933 }
2934 }
2935 s->size = xsize_t(st.st_size);
2936 if (!s->size)
2937 goto empty;
2938 if (S_ISLNK(st.st_mode)) {
2939 struct strbuf sb = STRBUF_INIT;
2940
2941 if (strbuf_readlink(&sb, s->path, s->size))
2942 goto err_empty;
2943 s->size = sb.len;
2944 s->data = strbuf_detach(&sb, NULL);
2945 s->should_free = 1;
2946 return 0;
2947 }
2948
2949 /*
2950 * Even if the caller would be happy with getting
2951 * only the size, we cannot return early at this
2952 * point if the path requires us to run the content
2953 * conversion.
2954 */
2955 if (size_only && !would_convert_to_git(&the_index, s->path))
2956 return 0;
2957
2958 /*
2959 * Note: this check uses xsize_t(st.st_size) that may
2960 * not be the true size of the blob after it goes
2961 * through convert_to_git(). This may not strictly be
2962 * correct, but the whole point of big_file_threshold
2963 * and is_binary check being that we want to avoid
2964 * opening the file and inspecting the contents, this
2965 * is probably fine.
2966 */
2967 if ((flags & CHECK_BINARY) &&
2968 s->size > big_file_threshold && s->is_binary == -1) {
2969 s->is_binary = 1;
2970 return 0;
2971 }
2972 fd = open(s->path, O_RDONLY);
2973 if (fd < 0)
2974 goto err_empty;
2975 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2976 close(fd);
2977 s->should_munmap = 1;
2978
2979 /*
2980 * Convert from working tree format to canonical git format
2981 */
2982 if (convert_to_git(&the_index, s->path, s->data, s->size, &buf, crlf_warn)) {
2983 size_t size = 0;
2984 munmap(s->data, s->size);
2985 s->should_munmap = 0;
2986 s->data = strbuf_detach(&buf, &size);
2987 s->size = size;
2988 s->should_free = 1;
2989 }
2990 }
2991 else {
2992 enum object_type type;
2993 if (size_only || (flags & CHECK_BINARY)) {
2994 type = sha1_object_info(s->oid.hash, &s->size);
2995 if (type < 0)
2996 die("unable to read %s",
2997 oid_to_hex(&s->oid));
2998 if (size_only)
2999 return 0;
3000 if (s->size > big_file_threshold && s->is_binary == -1) {
3001 s->is_binary = 1;
3002 return 0;
3003 }
3004 }
3005 s->data = read_sha1_file(s->oid.hash, &type, &s->size);
3006 if (!s->data)
3007 die("unable to read %s", oid_to_hex(&s->oid));
3008 s->should_free = 1;
3009 }
3010 return 0;
3011 }
3012
3013 void diff_free_filespec_blob(struct diff_filespec *s)
3014 {
3015 if (s->should_free)
3016 free(s->data);
3017 else if (s->should_munmap)
3018 munmap(s->data, s->size);
3019
3020 if (s->should_free || s->should_munmap) {
3021 s->should_free = s->should_munmap = 0;
3022 s->data = NULL;
3023 }
3024 }
3025
3026 void diff_free_filespec_data(struct diff_filespec *s)
3027 {
3028 diff_free_filespec_blob(s);
3029 FREE_AND_NULL(s->cnt_data);
3030 }
3031
3032 static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
3033 void *blob,
3034 unsigned long size,
3035 const struct object_id *oid,
3036 int mode)
3037 {
3038 int fd;
3039 struct strbuf buf = STRBUF_INIT;
3040 struct strbuf template = STRBUF_INIT;
3041 char *path_dup = xstrdup(path);
3042 const char *base = basename(path_dup);
3043
3044 /* Generate "XXXXXX_basename.ext" */
3045 strbuf_addstr(&template, "XXXXXX_");
3046 strbuf_addstr(&template, base);
3047
3048 fd = mks_tempfile_ts(&temp->tempfile, template.buf, strlen(base) + 1);
3049 if (fd < 0)
3050 die_errno("unable to create temp-file");
3051 if (convert_to_working_tree(path,
3052 (const char *)blob, (size_t)size, &buf)) {
3053 blob = buf.buf;
3054 size = buf.len;
3055 }
3056 if (write_in_full(fd, blob, size) != size)
3057 die_errno("unable to write temp-file");
3058 close_tempfile(&temp->tempfile);
3059 temp->name = get_tempfile_path(&temp->tempfile);
3060 oid_to_hex_r(temp->hex, oid);
3061 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", mode);
3062 strbuf_release(&buf);
3063 strbuf_release(&template);
3064 free(path_dup);
3065 }
3066
3067 static struct diff_tempfile *prepare_temp_file(const char *name,
3068 struct diff_filespec *one)
3069 {
3070 struct diff_tempfile *temp = claim_diff_tempfile();
3071
3072 if (!DIFF_FILE_VALID(one)) {
3073 not_a_valid_file:
3074 /* A '-' entry produces this for file-2, and
3075 * a '+' entry produces this for file-1.
3076 */
3077 temp->name = "/dev/null";
3078 xsnprintf(temp->hex, sizeof(temp->hex), ".");
3079 xsnprintf(temp->mode, sizeof(temp->mode), ".");
3080 return temp;
3081 }
3082
3083 if (!S_ISGITLINK(one->mode) &&
3084 (!one->oid_valid ||
3085 reuse_worktree_file(name, &one->oid, 1))) {
3086 struct stat st;
3087 if (lstat(name, &st) < 0) {
3088 if (errno == ENOENT)
3089 goto not_a_valid_file;
3090 die_errno("stat(%s)", name);
3091 }
3092 if (S_ISLNK(st.st_mode)) {
3093 struct strbuf sb = STRBUF_INIT;
3094 if (strbuf_readlink(&sb, name, st.st_size) < 0)
3095 die_errno("readlink(%s)", name);
3096 prep_temp_blob(name, temp, sb.buf, sb.len,
3097 (one->oid_valid ?
3098 &one->oid : &null_oid),
3099 (one->oid_valid ?
3100 one->mode : S_IFLNK));
3101 strbuf_release(&sb);
3102 }
3103 else {
3104 /* we can borrow from the file in the work tree */
3105 temp->name = name;
3106 if (!one->oid_valid)
3107 oid_to_hex_r(temp->hex, &null_oid);
3108 else
3109 oid_to_hex_r(temp->hex, &one->oid);
3110 /* Even though we may sometimes borrow the
3111 * contents from the work tree, we always want
3112 * one->mode. mode is trustworthy even when
3113 * !(one->oid_valid), as long as
3114 * DIFF_FILE_VALID(one).
3115 */
3116 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", one->mode);
3117 }
3118 return temp;
3119 }
3120 else {
3121 if (diff_populate_filespec(one, 0))
3122 die("cannot read data blob for %s", one->path);
3123 prep_temp_blob(name, temp, one->data, one->size,
3124 &one->oid, one->mode);
3125 }
3126 return temp;
3127 }
3128
3129 static void add_external_diff_name(struct argv_array *argv,
3130 const char *name,
3131 struct diff_filespec *df)
3132 {
3133 struct diff_tempfile *temp = prepare_temp_file(name, df);
3134 argv_array_push(argv, temp->name);
3135 argv_array_push(argv, temp->hex);
3136 argv_array_push(argv, temp->mode);
3137 }
3138
3139 /* An external diff command takes:
3140 *
3141 * diff-cmd name infile1 infile1-sha1 infile1-mode \
3142 * infile2 infile2-sha1 infile2-mode [ rename-to ]
3143 *
3144 */
3145 static void run_external_diff(const char *pgm,
3146 const char *name,
3147 const char *other,
3148 struct diff_filespec *one,
3149 struct diff_filespec *two,
3150 const char *xfrm_msg,
3151 int complete_rewrite,
3152 struct diff_options *o)
3153 {
3154 struct argv_array argv = ARGV_ARRAY_INIT;
3155 struct argv_array env = ARGV_ARRAY_INIT;
3156 struct diff_queue_struct *q = &diff_queued_diff;
3157
3158 argv_array_push(&argv, pgm);
3159 argv_array_push(&argv, name);
3160
3161 if (one && two) {
3162 add_external_diff_name(&argv, name, one);
3163 if (!other)
3164 add_external_diff_name(&argv, name, two);
3165 else {
3166 add_external_diff_name(&argv, other, two);
3167 argv_array_push(&argv, other);
3168 argv_array_push(&argv, xfrm_msg);
3169 }
3170 }
3171
3172 argv_array_pushf(&env, "GIT_DIFF_PATH_COUNTER=%d", ++o->diff_path_counter);
3173 argv_array_pushf(&env, "GIT_DIFF_PATH_TOTAL=%d", q->nr);
3174
3175 if (run_command_v_opt_cd_env(argv.argv, RUN_USING_SHELL, NULL, env.argv))
3176 die(_("external diff died, stopping at %s"), name);
3177
3178 remove_tempfile();
3179 argv_array_clear(&argv);
3180 argv_array_clear(&env);
3181 }
3182
3183 static int similarity_index(struct diff_filepair *p)
3184 {
3185 return p->score * 100 / MAX_SCORE;
3186 }
3187
3188 static const char *diff_abbrev_oid(const struct object_id *oid, int abbrev)
3189 {
3190 if (startup_info->have_repository)
3191 return find_unique_abbrev(oid->hash, abbrev);
3192 else {
3193 char *hex = oid_to_hex(oid);
3194 if (abbrev < 0)
3195 abbrev = FALLBACK_DEFAULT_ABBREV;
3196 if (abbrev > GIT_SHA1_HEXSZ)
3197 die("BUG: oid abbreviation out of range: %d", abbrev);
3198 if (abbrev)
3199 hex[abbrev] = '\0';
3200 return hex;
3201 }
3202 }
3203
3204 static void fill_metainfo(struct strbuf *msg,
3205 const char *name,
3206 const char *other,
3207 struct diff_filespec *one,
3208 struct diff_filespec *two,
3209 struct diff_options *o,
3210 struct diff_filepair *p,
3211 int *must_show_header,
3212 int use_color)
3213 {
3214 const char *set = diff_get_color(use_color, DIFF_METAINFO);
3215 const char *reset = diff_get_color(use_color, DIFF_RESET);
3216 const char *line_prefix = diff_line_prefix(o);
3217
3218 *must_show_header = 1;
3219 strbuf_init(msg, PATH_MAX * 2 + 300);
3220 switch (p->status) {
3221 case DIFF_STATUS_COPIED:
3222 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3223 line_prefix, set, similarity_index(p));
3224 strbuf_addf(msg, "%s\n%s%scopy from ",
3225 reset, line_prefix, set);
3226 quote_c_style(name, msg, NULL, 0);
3227 strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
3228 quote_c_style(other, msg, NULL, 0);
3229 strbuf_addf(msg, "%s\n", reset);
3230 break;
3231 case DIFF_STATUS_RENAMED:
3232 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3233 line_prefix, set, similarity_index(p));
3234 strbuf_addf(msg, "%s\n%s%srename from ",
3235 reset, line_prefix, set);
3236 quote_c_style(name, msg, NULL, 0);
3237 strbuf_addf(msg, "%s\n%s%srename to ",
3238 reset, line_prefix, set);
3239 quote_c_style(other, msg, NULL, 0);
3240 strbuf_addf(msg, "%s\n", reset);
3241 break;
3242 case DIFF_STATUS_MODIFIED:
3243 if (p->score) {
3244 strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
3245 line_prefix,
3246 set, similarity_index(p), reset);
3247 break;
3248 }
3249 /* fallthru */
3250 default:
3251 *must_show_header = 0;
3252 }
3253 if (one && two && oidcmp(&one->oid, &two->oid)) {
3254 int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
3255
3256 if (DIFF_OPT_TST(o, BINARY)) {
3257 mmfile_t mf;
3258 if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
3259 (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
3260 abbrev = 40;
3261 }
3262 strbuf_addf(msg, "%s%sindex %s..%s", line_prefix, set,
3263 diff_abbrev_oid(&one->oid, abbrev),
3264 diff_abbrev_oid(&two->oid, abbrev));
3265 if (one->mode == two->mode)
3266 strbuf_addf(msg, " %06o", one->mode);
3267 strbuf_addf(msg, "%s\n", reset);
3268 }
3269 }
3270
3271 static void run_diff_cmd(const char *pgm,
3272 const char *name,
3273 const char *other,
3274 const char *attr_path,
3275 struct diff_filespec *one,
3276 struct diff_filespec *two,
3277 struct strbuf *msg,
3278 struct diff_options *o,
3279 struct diff_filepair *p)
3280 {
3281 const char *xfrm_msg = NULL;
3282 int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
3283 int must_show_header = 0;
3284
3285
3286 if (DIFF_OPT_TST(o, ALLOW_EXTERNAL)) {
3287 struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
3288 if (drv && drv->external)
3289 pgm = drv->external;
3290 }
3291
3292 if (msg) {
3293 /*
3294 * don't use colors when the header is intended for an
3295 * external diff driver
3296 */
3297 fill_metainfo(msg, name, other, one, two, o, p,
3298 &must_show_header,
3299 want_color(o->use_color) && !pgm);
3300 xfrm_msg = msg->len ? msg->buf : NULL;
3301 }
3302
3303 if (pgm) {
3304 run_external_diff(pgm, name, other, one, two, xfrm_msg,
3305 complete_rewrite, o);
3306 return;
3307 }
3308 if (one && two)
3309 builtin_diff(name, other ? other : name,
3310 one, two, xfrm_msg, must_show_header,
3311 o, complete_rewrite);
3312 else
3313 fprintf(o->file, "* Unmerged path %s\n", name);
3314 }
3315
3316 static void diff_fill_oid_info(struct diff_filespec *one)
3317 {
3318 if (DIFF_FILE_VALID(one)) {
3319 if (!one->oid_valid) {
3320 struct stat st;
3321 if (one->is_stdin) {
3322 oidclr(&one->oid);
3323 return;
3324 }
3325 if (lstat(one->path, &st) < 0)
3326 die_errno("stat '%s'", one->path);
3327 if (index_path(one->oid.hash, one->path, &st, 0))
3328 die("cannot hash %s", one->path);
3329 }
3330 }
3331 else
3332 oidclr(&one->oid);
3333 }
3334
3335 static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
3336 {
3337 /* Strip the prefix but do not molest /dev/null and absolute paths */
3338 if (*namep && **namep != '/') {
3339 *namep += prefix_length;
3340 if (**namep == '/')
3341 ++*namep;
3342 }
3343 if (*otherp && **otherp != '/') {
3344 *otherp += prefix_length;
3345 if (**otherp == '/')
3346 ++*otherp;
3347 }
3348 }
3349
3350 static void run_diff(struct diff_filepair *p, struct diff_options *o)
3351 {
3352 const char *pgm = external_diff();
3353 struct strbuf msg;
3354 struct diff_filespec *one = p->one;
3355 struct diff_filespec *two = p->two;
3356 const char *name;
3357 const char *other;
3358 const char *attr_path;
3359
3360 name = one->path;
3361 other = (strcmp(name, two->path) ? two->path : NULL);
3362 attr_path = name;
3363 if (o->prefix_length)
3364 strip_prefix(o->prefix_length, &name, &other);
3365
3366 if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
3367 pgm = NULL;
3368
3369 if (DIFF_PAIR_UNMERGED(p)) {
3370 run_diff_cmd(pgm, name, NULL, attr_path,
3371 NULL, NULL, NULL, o, p);
3372 return;
3373 }
3374
3375 diff_fill_oid_info(one);
3376 diff_fill_oid_info(two);
3377
3378 if (!pgm &&
3379 DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3380 (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3381 /*
3382 * a filepair that changes between file and symlink
3383 * needs to be split into deletion and creation.
3384 */
3385 struct diff_filespec *null = alloc_filespec(two->path);
3386 run_diff_cmd(NULL, name, other, attr_path,
3387 one, null, &msg, o, p);
3388 free(null);
3389 strbuf_release(&msg);
3390
3391 null = alloc_filespec(one->path);
3392 run_diff_cmd(NULL, name, other, attr_path,
3393 null, two, &msg, o, p);
3394 free(null);
3395 }
3396 else
3397 run_diff_cmd(pgm, name, other, attr_path,
3398 one, two, &msg, o, p);
3399
3400 strbuf_release(&msg);
3401 }
3402
3403 static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3404 struct diffstat_t *diffstat)
3405 {
3406 const char *name;
3407 const char *other;
3408
3409 if (DIFF_PAIR_UNMERGED(p)) {
3410 /* unmerged */
3411 builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, p);
3412 return;
3413 }
3414
3415 name = p->one->path;
3416 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3417
3418 if (o->prefix_length)
3419 strip_prefix(o->prefix_length, &name, &other);
3420
3421 diff_fill_oid_info(p->one);
3422 diff_fill_oid_info(p->two);
3423
3424 builtin_diffstat(name, other, p->one, p->two, diffstat, o, p);
3425 }
3426
3427 static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3428 {
3429 const char *name;
3430 const char *other;
3431 const char *attr_path;
3432
3433 if (DIFF_PAIR_UNMERGED(p)) {
3434 /* unmerged */
3435 return;
3436 }
3437
3438 name = p->one->path;
3439 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3440 attr_path = other ? other : name;
3441
3442 if (o->prefix_length)
3443 strip_prefix(o->prefix_length, &name, &other);
3444
3445 diff_fill_oid_info(p->one);
3446 diff_fill_oid_info(p->two);
3447
3448 builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3449 }
3450
3451 void diff_setup(struct diff_options *options)
3452 {
3453 memcpy(options, &default_diff_options, sizeof(*options));
3454
3455 options->file = stdout;
3456
3457 options->abbrev = DEFAULT_ABBREV;
3458 options->line_termination = '\n';
3459 options->break_opt = -1;
3460 options->rename_limit = -1;
3461 options->dirstat_permille = diff_dirstat_permille_default;
3462 options->context = diff_context_default;
3463 options->interhunkcontext = diff_interhunk_context_default;
3464 options->ws_error_highlight = ws_error_highlight_default;
3465 DIFF_OPT_SET(options, RENAME_EMPTY);
3466
3467 /* pathchange left =NULL by default */
3468 options->change = diff_change;
3469 options->add_remove = diff_addremove;
3470 options->use_color = diff_use_color_default;
3471 options->detect_rename = diff_detect_rename_default;
3472 options->xdl_opts |= diff_algorithm;
3473 if (diff_indent_heuristic)
3474 DIFF_XDL_SET(options, INDENT_HEURISTIC);
3475
3476 options->orderfile = diff_order_file_cfg;
3477
3478 if (diff_no_prefix) {
3479 options->a_prefix = options->b_prefix = "";
3480 } else if (!diff_mnemonic_prefix) {
3481 options->a_prefix = "a/";
3482 options->b_prefix = "b/";
3483 }
3484 }
3485
3486 void diff_setup_done(struct diff_options *options)
3487 {
3488 int count = 0;
3489
3490 if (options->set_default)
3491 options->set_default(options);
3492
3493 if (options->output_format & DIFF_FORMAT_NAME)
3494 count++;
3495 if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3496 count++;
3497 if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3498 count++;
3499 if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3500 count++;
3501 if (count > 1)
3502 die(_("--name-only, --name-status, --check and -s are mutually exclusive"));
3503
3504 /*
3505 * Most of the time we can say "there are changes"
3506 * only by checking if there are changed paths, but
3507 * --ignore-whitespace* options force us to look
3508 * inside contents.
3509 */
3510
3511 if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3512 DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3513 DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3514 DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3515 else
3516 DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3517
3518 if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3519 options->detect_rename = DIFF_DETECT_COPY;
3520
3521 if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3522 options->prefix = NULL;
3523 if (options->prefix)
3524 options->prefix_length = strlen(options->prefix);
3525 else
3526 options->prefix_length = 0;
3527
3528 if (options->output_format & (DIFF_FORMAT_NAME |
3529 DIFF_FORMAT_NAME_STATUS |
3530 DIFF_FORMAT_CHECKDIFF |
3531 DIFF_FORMAT_NO_OUTPUT))
3532 options->output_format &= ~(DIFF_FORMAT_RAW |
3533 DIFF_FORMAT_NUMSTAT |
3534 DIFF_FORMAT_DIFFSTAT |
3535 DIFF_FORMAT_SHORTSTAT |
3536 DIFF_FORMAT_DIRSTAT |
3537 DIFF_FORMAT_SUMMARY |
3538 DIFF_FORMAT_PATCH);
3539
3540 /*
3541 * These cases always need recursive; we do not drop caller-supplied
3542 * recursive bits for other formats here.
3543 */
3544 if (options->output_format & (DIFF_FORMAT_PATCH |
3545 DIFF_FORMAT_NUMSTAT |
3546 DIFF_FORMAT_DIFFSTAT |
3547 DIFF_FORMAT_SHORTSTAT |
3548 DIFF_FORMAT_DIRSTAT |
3549 DIFF_FORMAT_SUMMARY |
3550 DIFF_FORMAT_CHECKDIFF))
3551 DIFF_OPT_SET(options, RECURSIVE);
3552 /*
3553 * Also pickaxe would not work very well if you do not say recursive
3554 */
3555 if (options->pickaxe)
3556 DIFF_OPT_SET(options, RECURSIVE);
3557 /*
3558 * When patches are generated, submodules diffed against the work tree
3559 * must be checked for dirtiness too so it can be shown in the output
3560 */
3561 if (options->output_format & DIFF_FORMAT_PATCH)
3562 DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3563
3564 if (options->detect_rename && options->rename_limit < 0)
3565 options->rename_limit = diff_rename_limit_default;
3566 if (options->setup & DIFF_SETUP_USE_CACHE) {
3567 if (!active_cache)
3568 /* read-cache does not die even when it fails
3569 * so it is safe for us to do this here. Also
3570 * it does not smudge active_cache or active_nr
3571 * when it fails, so we do not have to worry about
3572 * cleaning it up ourselves either.
3573 */
3574 read_cache();
3575 }
3576 if (40 < options->abbrev)
3577 options->abbrev = 40; /* full */
3578
3579 /*
3580 * It does not make sense to show the first hit we happened
3581 * to have found. It does not make sense not to return with
3582 * exit code in such a case either.
3583 */
3584 if (DIFF_OPT_TST(options, QUICK)) {
3585 options->output_format = DIFF_FORMAT_NO_OUTPUT;
3586 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3587 }
3588
3589 options->diff_path_counter = 0;
3590
3591 if (DIFF_OPT_TST(options, FOLLOW_RENAMES) && options->pathspec.nr != 1)
3592 die(_("--follow requires exactly one pathspec"));
3593 }
3594
3595 static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3596 {
3597 char c, *eq;
3598 int len;
3599
3600 if (*arg != '-')
3601 return 0;
3602 c = *++arg;
3603 if (!c)
3604 return 0;
3605 if (c == arg_short) {
3606 c = *++arg;
3607 if (!c)
3608 return 1;
3609 if (val && isdigit(c)) {
3610 char *end;
3611 int n = strtoul(arg, &end, 10);
3612 if (*end)
3613 return 0;
3614 *val = n;
3615 return 1;
3616 }
3617 return 0;
3618 }
3619 if (c != '-')
3620 return 0;
3621 arg++;
3622 eq = strchrnul(arg, '=');
3623 len = eq - arg;
3624 if (!len || strncmp(arg, arg_long, len))
3625 return 0;
3626 if (*eq) {
3627 int n;
3628 char *end;
3629 if (!isdigit(*++eq))
3630 return 0;
3631 n = strtoul(eq, &end, 10);
3632 if (*end)
3633 return 0;
3634 *val = n;
3635 }
3636 return 1;
3637 }
3638
3639 static int diff_scoreopt_parse(const char *opt);
3640
3641 static inline int short_opt(char opt, const char **argv,
3642 const char **optarg)
3643 {
3644 const char *arg = argv[0];
3645 if (arg[0] != '-' || arg[1] != opt)
3646 return 0;
3647 if (arg[2] != '\0') {
3648 *optarg = arg + 2;
3649 return 1;
3650 }
3651 if (!argv[1])
3652 die("Option '%c' requires a value", opt);
3653 *optarg = argv[1];
3654 return 2;
3655 }
3656
3657 int parse_long_opt(const char *opt, const char **argv,
3658 const char **optarg)
3659 {
3660 const char *arg = argv[0];
3661 if (!skip_prefix(arg, "--", &arg))
3662 return 0;
3663 if (!skip_prefix(arg, opt, &arg))
3664 return 0;
3665 if (*arg == '=') { /* stuck form: --option=value */
3666 *optarg = arg + 1;
3667 return 1;
3668 }
3669 if (*arg != '\0')
3670 return 0;
3671 /* separate form: --option value */
3672 if (!argv[1])
3673 die("Option '--%s' requires a value", opt);
3674 *optarg = argv[1];
3675 return 2;
3676 }
3677
3678 static int stat_opt(struct diff_options *options, const char **av)
3679 {
3680 const char *arg = av[0];
3681 char *end;
3682 int width = options->stat_width;
3683 int name_width = options->stat_name_width;
3684 int graph_width = options->stat_graph_width;
3685 int count = options->stat_count;
3686 int argcount = 1;
3687
3688 if (!skip_prefix(arg, "--stat", &arg))
3689 die("BUG: stat option does not begin with --stat: %s", arg);
3690 end = (char *)arg;
3691
3692 switch (*arg) {
3693 case '-':
3694 if (skip_prefix(arg, "-width", &arg)) {
3695 if (*arg == '=')
3696 width = strtoul(arg + 1, &end, 10);
3697 else if (!*arg && !av[1])
3698 die_want_option("--stat-width");
3699 else if (!*arg) {
3700 width = strtoul(av[1], &end, 10);
3701 argcount = 2;
3702 }
3703 } else if (skip_prefix(arg, "-name-width", &arg)) {
3704 if (*arg == '=')
3705 name_width = strtoul(arg + 1, &end, 10);
3706 else if (!*arg && !av[1])
3707 die_want_option("--stat-name-width");
3708 else if (!*arg) {
3709 name_width = strtoul(av[1], &end, 10);
3710 argcount = 2;
3711 }
3712 } else if (skip_prefix(arg, "-graph-width", &arg)) {
3713 if (*arg == '=')
3714 graph_width = strtoul(arg + 1, &end, 10);
3715 else if (!*arg && !av[1])
3716 die_want_option("--stat-graph-width");
3717 else if (!*arg) {
3718 graph_width = strtoul(av[1], &end, 10);
3719 argcount = 2;
3720 }
3721 } else if (skip_prefix(arg, "-count", &arg)) {
3722 if (*arg == '=')
3723 count = strtoul(arg + 1, &end, 10);
3724 else if (!*arg && !av[1])
3725 die_want_option("--stat-count");
3726 else if (!*arg) {
3727 count = strtoul(av[1], &end, 10);
3728 argcount = 2;
3729 }
3730 }
3731 break;
3732 case '=':
3733 width = strtoul(arg+1, &end, 10);
3734 if (*end == ',')
3735 name_width = strtoul(end+1, &end, 10);
3736 if (*end == ',')
3737 count = strtoul(end+1, &end, 10);
3738 }
3739
3740 /* Important! This checks all the error cases! */
3741 if (*end)
3742 return 0;
3743 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3744 options->stat_name_width = name_width;
3745 options->stat_graph_width = graph_width;
3746 options->stat_width = width;
3747 options->stat_count = count;
3748 return argcount;
3749 }
3750
3751 static int parse_dirstat_opt(struct diff_options *options, const char *params)
3752 {
3753 struct strbuf errmsg = STRBUF_INIT;
3754 if (parse_dirstat_params(options, params, &errmsg))
3755 die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3756 errmsg.buf);
3757 strbuf_release(&errmsg);
3758 /*
3759 * The caller knows a dirstat-related option is given from the command
3760 * line; allow it to say "return this_function();"
3761 */
3762 options->output_format |= DIFF_FORMAT_DIRSTAT;
3763 return 1;
3764 }
3765
3766 static int parse_submodule_opt(struct diff_options *options, const char *value)
3767 {
3768 if (parse_submodule_params(options, value))
3769 die(_("Failed to parse --submodule option parameter: '%s'"),
3770 value);
3771 return 1;
3772 }
3773
3774 static const char diff_status_letters[] = {
3775 DIFF_STATUS_ADDED,
3776 DIFF_STATUS_COPIED,
3777 DIFF_STATUS_DELETED,
3778 DIFF_STATUS_MODIFIED,
3779 DIFF_STATUS_RENAMED,
3780 DIFF_STATUS_TYPE_CHANGED,
3781 DIFF_STATUS_UNKNOWN,
3782 DIFF_STATUS_UNMERGED,
3783 DIFF_STATUS_FILTER_AON,
3784 DIFF_STATUS_FILTER_BROKEN,
3785 '\0',
3786 };
3787
3788 static unsigned int filter_bit['Z' + 1];
3789
3790 static void prepare_filter_bits(void)
3791 {
3792 int i;
3793
3794 if (!filter_bit[DIFF_STATUS_ADDED]) {
3795 for (i = 0; diff_status_letters[i]; i++)
3796 filter_bit[(int) diff_status_letters[i]] = (1 << i);
3797 }
3798 }
3799
3800 static unsigned filter_bit_tst(char status, const struct diff_options *opt)
3801 {
3802 return opt->filter & filter_bit[(int) status];
3803 }
3804
3805 static int parse_diff_filter_opt(const char *optarg, struct diff_options *opt)
3806 {
3807 int i, optch;
3808
3809 prepare_filter_bits();
3810
3811 /*
3812 * If there is a negation e.g. 'd' in the input, and we haven't
3813 * initialized the filter field with another --diff-filter, start
3814 * from full set of bits, except for AON.
3815 */
3816 if (!opt->filter) {
3817 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3818 if (optch < 'a' || 'z' < optch)
3819 continue;
3820 opt->filter = (1 << (ARRAY_SIZE(diff_status_letters) - 1)) - 1;
3821 opt->filter &= ~filter_bit[DIFF_STATUS_FILTER_AON];
3822 break;
3823 }
3824 }
3825
3826 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3827 unsigned int bit;
3828 int negate;
3829
3830 if ('a' <= optch && optch <= 'z') {
3831 negate = 1;
3832 optch = toupper(optch);
3833 } else {
3834 negate = 0;
3835 }
3836
3837 bit = (0 <= optch && optch <= 'Z') ? filter_bit[optch] : 0;
3838 if (!bit)
3839 return optarg[i];
3840 if (negate)
3841 opt->filter &= ~bit;
3842 else
3843 opt->filter |= bit;
3844 }
3845 return 0;
3846 }
3847
3848 static void enable_patch_output(int *fmt) {
3849 *fmt &= ~DIFF_FORMAT_NO_OUTPUT;
3850 *fmt |= DIFF_FORMAT_PATCH;
3851 }
3852
3853 static int parse_ws_error_highlight_opt(struct diff_options *opt, const char *arg)
3854 {
3855 int val = parse_ws_error_highlight(arg);
3856
3857 if (val < 0) {
3858 error("unknown value after ws-error-highlight=%.*s",
3859 -1 - val, arg);
3860 return 0;
3861 }
3862 opt->ws_error_highlight = val;
3863 return 1;
3864 }
3865
3866 int diff_opt_parse(struct diff_options *options,
3867 const char **av, int ac, const char *prefix)
3868 {
3869 const char *arg = av[0];
3870 const char *optarg;
3871 int argcount;
3872
3873 if (!prefix)
3874 prefix = "";
3875
3876 /* Output format options */
3877 if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch")
3878 || opt_arg(arg, 'U', "unified", &options->context))
3879 enable_patch_output(&options->output_format);
3880 else if (!strcmp(arg, "--raw"))
3881 options->output_format |= DIFF_FORMAT_RAW;
3882 else if (!strcmp(arg, "--patch-with-raw")) {
3883 enable_patch_output(&options->output_format);
3884 options->output_format |= DIFF_FORMAT_RAW;
3885 } else if (!strcmp(arg, "--numstat"))
3886 options->output_format |= DIFF_FORMAT_NUMSTAT;
3887 else if (!strcmp(arg, "--shortstat"))
3888 options->output_format |= DIFF_FORMAT_SHORTSTAT;
3889 else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3890 return parse_dirstat_opt(options, "");
3891 else if (skip_prefix(arg, "-X", &arg))
3892 return parse_dirstat_opt(options, arg);
3893 else if (skip_prefix(arg, "--dirstat=", &arg))
3894 return parse_dirstat_opt(options, arg);
3895 else if (!strcmp(arg, "--cumulative"))
3896 return parse_dirstat_opt(options, "cumulative");
3897 else if (!strcmp(arg, "--dirstat-by-file"))
3898 return parse_dirstat_opt(options, "files");
3899 else if (skip_prefix(arg, "--dirstat-by-file=", &arg)) {
3900 parse_dirstat_opt(options, "files");
3901 return parse_dirstat_opt(options, arg);
3902 }
3903 else if (!strcmp(arg, "--check"))
3904 options->output_format |= DIFF_FORMAT_CHECKDIFF;
3905 else if (!strcmp(arg, "--summary"))
3906 options->output_format |= DIFF_FORMAT_SUMMARY;
3907 else if (!strcmp(arg, "--patch-with-stat")) {
3908 enable_patch_output(&options->output_format);
3909 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3910 } else if (!strcmp(arg, "--name-only"))
3911 options->output_format |= DIFF_FORMAT_NAME;
3912 else if (!strcmp(arg, "--name-status"))
3913 options->output_format |= DIFF_FORMAT_NAME_STATUS;
3914 else if (!strcmp(arg, "-s") || !strcmp(arg, "--no-patch"))
3915 options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3916 else if (starts_with(arg, "--stat"))
3917 /* --stat, --stat-width, --stat-name-width, or --stat-count */
3918 return stat_opt(options, av);
3919
3920 /* renames options */
3921 else if (starts_with(arg, "-B") || starts_with(arg, "--break-rewrites=") ||
3922 !strcmp(arg, "--break-rewrites")) {
3923 if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3924 return error("invalid argument to -B: %s", arg+2);
3925 }
3926 else if (starts_with(arg, "-M") || starts_with(arg, "--find-renames=") ||
3927 !strcmp(arg, "--find-renames")) {
3928 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3929 return error("invalid argument to -M: %s", arg+2);
3930 options->detect_rename = DIFF_DETECT_RENAME;
3931 }
3932 else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3933 options->irreversible_delete = 1;
3934 }
3935 else if (starts_with(arg, "-C") || starts_with(arg, "--find-copies=") ||
3936 !strcmp(arg, "--find-copies")) {
3937 if (options->detect_rename == DIFF_DETECT_COPY)
3938 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3939 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3940 return error("invalid argument to -C: %s", arg+2);
3941 options->detect_rename = DIFF_DETECT_COPY;
3942 }
3943 else if (!strcmp(arg, "--no-renames"))
3944 options->detect_rename = 0;
3945 else if (!strcmp(arg, "--rename-empty"))
3946 DIFF_OPT_SET(options, RENAME_EMPTY);
3947 else if (!strcmp(arg, "--no-rename-empty"))
3948 DIFF_OPT_CLR(options, RENAME_EMPTY);
3949 else if (!strcmp(arg, "--relative"))
3950 DIFF_OPT_SET(options, RELATIVE_NAME);
3951 else if (skip_prefix(arg, "--relative=", &arg)) {
3952 DIFF_OPT_SET(options, RELATIVE_NAME);
3953 options->prefix = arg;
3954 }
3955
3956 /* xdiff options */
3957 else if (!strcmp(arg, "--minimal"))
3958 DIFF_XDL_SET(options, NEED_MINIMAL);
3959 else if (!strcmp(arg, "--no-minimal"))
3960 DIFF_XDL_CLR(options, NEED_MINIMAL);
3961 else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3962 DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3963 else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3964 DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3965 else if (!strcmp(arg, "--ignore-space-at-eol"))
3966 DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3967 else if (!strcmp(arg, "--ignore-blank-lines"))
3968 DIFF_XDL_SET(options, IGNORE_BLANK_LINES);
3969 else if (!strcmp(arg, "--indent-heuristic"))
3970 DIFF_XDL_SET(options, INDENT_HEURISTIC);
3971 else if (!strcmp(arg, "--no-indent-heuristic"))
3972 DIFF_XDL_CLR(options, INDENT_HEURISTIC);
3973 else if (!strcmp(arg, "--patience"))
3974 options->xdl_opts = DIFF_WITH_ALG(options, PATIENCE_DIFF);
3975 else if (!strcmp(arg, "--histogram"))
3976 options->xdl_opts = DIFF_WITH_ALG(options, HISTOGRAM_DIFF);
3977 else if ((argcount = parse_long_opt("diff-algorithm", av, &optarg))) {
3978 long value = parse_algorithm_value(optarg);
3979 if (value < 0)
3980 return error("option diff-algorithm accepts \"myers\", "
3981 "\"minimal\", \"patience\" and \"histogram\"");
3982 /* clear out previous settings */
3983 DIFF_XDL_CLR(options, NEED_MINIMAL);
3984 options->xdl_opts &= ~XDF_DIFF_ALGORITHM_MASK;
3985 options->xdl_opts |= value;
3986 return argcount;
3987 }
3988
3989 /* flags options */
3990 else if (!strcmp(arg, "--binary")) {
3991 enable_patch_output(&options->output_format);
3992 DIFF_OPT_SET(options, BINARY);
3993 }
3994 else if (!strcmp(arg, "--full-index"))
3995 DIFF_OPT_SET(options, FULL_INDEX);
3996 else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
3997 DIFF_OPT_SET(options, TEXT);
3998 else if (!strcmp(arg, "-R"))
3999 DIFF_OPT_SET(options, REVERSE_DIFF);
4000 else if (!strcmp(arg, "--find-copies-harder"))
4001 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
4002 else if (!strcmp(arg, "--follow"))
4003 DIFF_OPT_SET(options, FOLLOW_RENAMES);
4004 else if (!strcmp(arg, "--no-follow")) {
4005 DIFF_OPT_CLR(options, FOLLOW_RENAMES);
4006 DIFF_OPT_CLR(options, DEFAULT_FOLLOW_RENAMES);
4007 } else if (!strcmp(arg, "--color"))
4008 options->use_color = 1;
4009 else if (skip_prefix(arg, "--color=", &arg)) {
4010 int value = git_config_colorbool(NULL, arg);
4011 if (value < 0)
4012 return error("option `color' expects \"always\", \"auto\", or \"never\"");
4013 options->use_color = value;
4014 }
4015 else if (!strcmp(arg, "--no-color"))
4016 options->use_color = 0;
4017 else if (!strcmp(arg, "--color-words")) {
4018 options->use_color = 1;
4019 options->word_diff = DIFF_WORDS_COLOR;
4020 }
4021 else if (skip_prefix(arg, "--color-words=", &arg)) {
4022 options->use_color = 1;
4023 options->word_diff = DIFF_WORDS_COLOR;
4024 options->word_regex = arg;
4025 }
4026 else if (!strcmp(arg, "--word-diff")) {
4027 if (options->word_diff == DIFF_WORDS_NONE)
4028 options->word_diff = DIFF_WORDS_PLAIN;
4029 }
4030 else if (skip_prefix(arg, "--word-diff=", &arg)) {
4031 if (!strcmp(arg, "plain"))
4032 options->word_diff = DIFF_WORDS_PLAIN;
4033 else if (!strcmp(arg, "color")) {
4034 options->use_color = 1;
4035 options->word_diff = DIFF_WORDS_COLOR;
4036 }
4037 else if (!strcmp(arg, "porcelain"))
4038 options->word_diff = DIFF_WORDS_PORCELAIN;
4039 else if (!strcmp(arg, "none"))
4040 options->word_diff = DIFF_WORDS_NONE;
4041 else
4042 die("bad --word-diff argument: %s", arg);
4043 }
4044 else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
4045 if (options->word_diff == DIFF_WORDS_NONE)
4046 options->word_diff = DIFF_WORDS_PLAIN;
4047 options->word_regex = optarg;
4048 return argcount;
4049 }
4050 else if (!strcmp(arg, "--exit-code"))
4051 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
4052 else if (!strcmp(arg, "--quiet"))
4053 DIFF_OPT_SET(options, QUICK);
4054 else if (!strcmp(arg, "--ext-diff"))
4055 DIFF_OPT_SET(options, ALLOW_EXTERNAL);
4056 else if (!strcmp(arg, "--no-ext-diff"))
4057 DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
4058 else if (!strcmp(arg, "--textconv"))
4059 DIFF_OPT_SET(options, ALLOW_TEXTCONV);
4060 else if (!strcmp(arg, "--no-textconv"))
4061 DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
4062 else if (!strcmp(arg, "--ignore-submodules")) {
4063 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
4064 handle_ignore_submodules_arg(options, "all");
4065 } else if (skip_prefix(arg, "--ignore-submodules=", &arg)) {
4066 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
4067 handle_ignore_submodules_arg(options, arg);
4068 } else if (!strcmp(arg, "--submodule"))
4069 options->submodule_format = DIFF_SUBMODULE_LOG;
4070 else if (skip_prefix(arg, "--submodule=", &arg))
4071 return parse_submodule_opt(options, arg);
4072 else if (skip_prefix(arg, "--ws-error-highlight=", &arg))
4073 return parse_ws_error_highlight_opt(options, arg);
4074 else if (!strcmp(arg, "--ita-invisible-in-index"))
4075 options->ita_invisible_in_index = 1;
4076 else if (!strcmp(arg, "--ita-visible-in-index"))
4077 options->ita_invisible_in_index = 0;
4078
4079 /* misc options */
4080 else if (!strcmp(arg, "-z"))
4081 options->line_termination = 0;
4082 else if ((argcount = short_opt('l', av, &optarg))) {
4083 options->rename_limit = strtoul(optarg, NULL, 10);
4084 return argcount;
4085 }
4086 else if ((argcount = short_opt('S', av, &optarg))) {
4087 options->pickaxe = optarg;
4088 options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
4089 return argcount;
4090 } else if ((argcount = short_opt('G', av, &optarg))) {
4091 options->pickaxe = optarg;
4092 options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
4093 return argcount;
4094 }
4095 else if (!strcmp(arg, "--pickaxe-all"))
4096 options->pickaxe_opts |= DIFF_PICKAXE_ALL;
4097 else if (!strcmp(arg, "--pickaxe-regex"))
4098 options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
4099 else if ((argcount = short_opt('O', av, &optarg))) {
4100 options->orderfile = prefix_filename(prefix, optarg);
4101 return argcount;
4102 }
4103 else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
4104 int offending = parse_diff_filter_opt(optarg, options);
4105 if (offending)
4106 die("unknown change class '%c' in --diff-filter=%s",
4107 offending, optarg);
4108 return argcount;
4109 }
4110 else if (!strcmp(arg, "--no-abbrev"))
4111 options->abbrev = 0;
4112 else if (!strcmp(arg, "--abbrev"))
4113 options->abbrev = DEFAULT_ABBREV;
4114 else if (skip_prefix(arg, "--abbrev=", &arg)) {
4115 options->abbrev = strtoul(arg, NULL, 10);
4116 if (options->abbrev < MINIMUM_ABBREV)
4117 options->abbrev = MINIMUM_ABBREV;
4118 else if (40 < options->abbrev)
4119 options->abbrev = 40;
4120 }
4121 else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
4122 options->a_prefix = optarg;
4123 return argcount;
4124 }
4125 else if ((argcount = parse_long_opt("line-prefix", av, &optarg))) {
4126 options->line_prefix = optarg;
4127 options->line_prefix_length = strlen(options->line_prefix);
4128 graph_setup_line_prefix(options);
4129 return argcount;
4130 }
4131 else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
4132 options->b_prefix = optarg;
4133 return argcount;
4134 }
4135 else if (!strcmp(arg, "--no-prefix"))
4136 options->a_prefix = options->b_prefix = "";
4137 else if (opt_arg(arg, '\0', "inter-hunk-context",
4138 &options->interhunkcontext))
4139 ;
4140 else if (!strcmp(arg, "-W"))
4141 DIFF_OPT_SET(options, FUNCCONTEXT);
4142 else if (!strcmp(arg, "--function-context"))
4143 DIFF_OPT_SET(options, FUNCCONTEXT);
4144 else if (!strcmp(arg, "--no-function-context"))
4145 DIFF_OPT_CLR(options, FUNCCONTEXT);
4146 else if ((argcount = parse_long_opt("output", av, &optarg))) {
4147 char *path = prefix_filename(prefix, optarg);
4148 options->file = xfopen(path, "w");
4149 options->close_file = 1;
4150 if (options->use_color != GIT_COLOR_ALWAYS)
4151 options->use_color = GIT_COLOR_NEVER;
4152 free(path);
4153 return argcount;
4154 } else
4155 return 0;
4156 return 1;
4157 }
4158
4159 int parse_rename_score(const char **cp_p)
4160 {
4161 unsigned long num, scale;
4162 int ch, dot;
4163 const char *cp = *cp_p;
4164
4165 num = 0;
4166 scale = 1;
4167 dot = 0;
4168 for (;;) {
4169 ch = *cp;
4170 if ( !dot && ch == '.' ) {
4171 scale = 1;
4172 dot = 1;
4173 } else if ( ch == '%' ) {
4174 scale = dot ? scale*100 : 100;
4175 cp++; /* % is always at the end */
4176 break;
4177 } else if ( ch >= '0' && ch <= '9' ) {
4178 if ( scale < 100000 ) {
4179 scale *= 10;
4180 num = (num*10) + (ch-'0');
4181 }
4182 } else {
4183 break;
4184 }
4185 cp++;
4186 }
4187 *cp_p = cp;
4188
4189 /* user says num divided by scale and we say internally that
4190 * is MAX_SCORE * num / scale.
4191 */
4192 return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
4193 }
4194
4195 static int diff_scoreopt_parse(const char *opt)
4196 {
4197 int opt1, opt2, cmd;
4198
4199 if (*opt++ != '-')
4200 return -1;
4201 cmd = *opt++;
4202 if (cmd == '-') {
4203 /* convert the long-form arguments into short-form versions */
4204 if (skip_prefix(opt, "break-rewrites", &opt)) {
4205 if (*opt == 0 || *opt++ == '=')
4206 cmd = 'B';
4207 } else if (skip_prefix(opt, "find-copies", &opt)) {
4208 if (*opt == 0 || *opt++ == '=')
4209 cmd = 'C';
4210 } else if (skip_prefix(opt, "find-renames", &opt)) {
4211 if (*opt == 0 || *opt++ == '=')
4212 cmd = 'M';
4213 }
4214 }
4215 if (cmd != 'M' && cmd != 'C' && cmd != 'B')
4216 return -1; /* that is not a -M, -C, or -B option */
4217
4218 opt1 = parse_rename_score(&opt);
4219 if (cmd != 'B')
4220 opt2 = 0;
4221 else {
4222 if (*opt == 0)
4223 opt2 = 0;
4224 else if (*opt != '/')
4225 return -1; /* we expect -B80/99 or -B80 */
4226 else {
4227 opt++;
4228 opt2 = parse_rename_score(&opt);
4229 }
4230 }
4231 if (*opt != 0)
4232 return -1;
4233 return opt1 | (opt2 << 16);
4234 }
4235
4236 struct diff_queue_struct diff_queued_diff;
4237
4238 void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
4239 {
4240 ALLOC_GROW(queue->queue, queue->nr + 1, queue->alloc);
4241 queue->queue[queue->nr++] = dp;
4242 }
4243
4244 struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
4245 struct diff_filespec *one,
4246 struct diff_filespec *two)
4247 {
4248 struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
4249 dp->one = one;
4250 dp->two = two;
4251 if (queue)
4252 diff_q(queue, dp);
4253 return dp;
4254 }
4255
4256 void diff_free_filepair(struct diff_filepair *p)
4257 {
4258 free_filespec(p->one);
4259 free_filespec(p->two);
4260 free(p);
4261 }
4262
4263 const char *diff_aligned_abbrev(const struct object_id *oid, int len)
4264 {
4265 int abblen;
4266 const char *abbrev;
4267
4268 if (len == GIT_SHA1_HEXSZ)
4269 return oid_to_hex(oid);
4270
4271 abbrev = diff_abbrev_oid(oid, len);
4272 abblen = strlen(abbrev);
4273
4274 /*
4275 * In well-behaved cases, where the abbbreviated result is the
4276 * same as the requested length, append three dots after the
4277 * abbreviation (hence the whole logic is limited to the case
4278 * where abblen < 37); when the actual abbreviated result is a
4279 * bit longer than the requested length, we reduce the number
4280 * of dots so that they match the well-behaved ones. However,
4281 * if the actual abbreviation is longer than the requested
4282 * length by more than three, we give up on aligning, and add
4283 * three dots anyway, to indicate that the output is not the
4284 * full object name. Yes, this may be suboptimal, but this
4285 * appears only in "diff --raw --abbrev" output and it is not
4286 * worth the effort to change it now. Note that this would
4287 * likely to work fine when the automatic sizing of default
4288 * abbreviation length is used--we would be fed -1 in "len" in
4289 * that case, and will end up always appending three-dots, but
4290 * the automatic sizing is supposed to give abblen that ensures
4291 * uniqueness across all objects (statistically speaking).
4292 */
4293 if (abblen < GIT_SHA1_HEXSZ - 3) {
4294 static char hex[GIT_MAX_HEXSZ + 1];
4295 if (len < abblen && abblen <= len + 2)
4296 xsnprintf(hex, sizeof(hex), "%s%.*s", abbrev, len+3-abblen, "..");
4297 else
4298 xsnprintf(hex, sizeof(hex), "%s...", abbrev);
4299 return hex;
4300 }
4301
4302 return oid_to_hex(oid);
4303 }
4304
4305 static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
4306 {
4307 int line_termination = opt->line_termination;
4308 int inter_name_termination = line_termination ? '\t' : '\0';
4309
4310 fprintf(opt->file, "%s", diff_line_prefix(opt));
4311 if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
4312 fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
4313 diff_aligned_abbrev(&p->one->oid, opt->abbrev));
4314 fprintf(opt->file, "%s ",
4315 diff_aligned_abbrev(&p->two->oid, opt->abbrev));
4316 }
4317 if (p->score) {
4318 fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
4319 inter_name_termination);
4320 } else {
4321 fprintf(opt->file, "%c%c", p->status, inter_name_termination);
4322 }
4323
4324 if (p->status == DIFF_STATUS_COPIED ||
4325 p->status == DIFF_STATUS_RENAMED) {
4326 const char *name_a, *name_b;
4327 name_a = p->one->path;
4328 name_b = p->two->path;
4329 strip_prefix(opt->prefix_length, &name_a, &name_b);
4330 write_name_quoted(name_a, opt->file, inter_name_termination);
4331 write_name_quoted(name_b, opt->file, line_termination);
4332 } else {
4333 const char *name_a, *name_b;
4334 name_a = p->one->mode ? p->one->path : p->two->path;
4335 name_b = NULL;
4336 strip_prefix(opt->prefix_length, &name_a, &name_b);
4337 write_name_quoted(name_a, opt->file, line_termination);
4338 }
4339 }
4340
4341 int diff_unmodified_pair(struct diff_filepair *p)
4342 {
4343 /* This function is written stricter than necessary to support
4344 * the currently implemented transformers, but the idea is to
4345 * let transformers to produce diff_filepairs any way they want,
4346 * and filter and clean them up here before producing the output.
4347 */
4348 struct diff_filespec *one = p->one, *two = p->two;
4349
4350 if (DIFF_PAIR_UNMERGED(p))
4351 return 0; /* unmerged is interesting */
4352
4353 /* deletion, addition, mode or type change
4354 * and rename are all interesting.
4355 */
4356 if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
4357 DIFF_PAIR_MODE_CHANGED(p) ||
4358 strcmp(one->path, two->path))
4359 return 0;
4360
4361 /* both are valid and point at the same path. that is, we are
4362 * dealing with a change.
4363 */
4364 if (one->oid_valid && two->oid_valid &&
4365 !oidcmp(&one->oid, &two->oid) &&
4366 !one->dirty_submodule && !two->dirty_submodule)
4367 return 1; /* no change */
4368 if (!one->oid_valid && !two->oid_valid)
4369 return 1; /* both look at the same file on the filesystem. */
4370 return 0;
4371 }
4372
4373 static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
4374 {
4375 if (diff_unmodified_pair(p))
4376 return;
4377
4378 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4379 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4380 return; /* no tree diffs in patch format */
4381
4382 run_diff(p, o);
4383 }
4384
4385 static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
4386 struct diffstat_t *diffstat)
4387 {
4388 if (diff_unmodified_pair(p))
4389 return;
4390
4391 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4392 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4393 return; /* no useful stat for tree diffs */
4394
4395 run_diffstat(p, o, diffstat);
4396 }
4397
4398 static void diff_flush_checkdiff(struct diff_filepair *p,
4399 struct diff_options *o)
4400 {
4401 if (diff_unmodified_pair(p))
4402 return;
4403
4404 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4405 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4406 return; /* nothing to check in tree diffs */
4407
4408 run_checkdiff(p, o);
4409 }
4410
4411 int diff_queue_is_empty(void)
4412 {
4413 struct diff_queue_struct *q = &diff_queued_diff;
4414 int i;
4415 for (i = 0; i < q->nr; i++)
4416 if (!diff_unmodified_pair(q->queue[i]))
4417 return 0;
4418 return 1;
4419 }
4420
4421 #if DIFF_DEBUG
4422 void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
4423 {
4424 fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
4425 x, one ? one : "",
4426 s->path,
4427 DIFF_FILE_VALID(s) ? "valid" : "invalid",
4428 s->mode,
4429 s->oid_valid ? oid_to_hex(&s->oid) : "");
4430 fprintf(stderr, "queue[%d] %s size %lu\n",
4431 x, one ? one : "",
4432 s->size);
4433 }
4434
4435 void diff_debug_filepair(const struct diff_filepair *p, int i)
4436 {
4437 diff_debug_filespec(p->one, i, "one");
4438 diff_debug_filespec(p->two, i, "two");
4439 fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
4440 p->score, p->status ? p->status : '?',
4441 p->one->rename_used, p->broken_pair);
4442 }
4443
4444 void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
4445 {
4446 int i;
4447 if (msg)
4448 fprintf(stderr, "%s\n", msg);
4449 fprintf(stderr, "q->nr = %d\n", q->nr);
4450 for (i = 0; i < q->nr; i++) {
4451 struct diff_filepair *p = q->queue[i];
4452 diff_debug_filepair(p, i);
4453 }
4454 }
4455 #endif
4456
4457 static void diff_resolve_rename_copy(void)
4458 {
4459 int i;
4460 struct diff_filepair *p;
4461 struct diff_queue_struct *q = &diff_queued_diff;
4462
4463 diff_debug_queue("resolve-rename-copy", q);
4464
4465 for (i = 0; i < q->nr; i++) {
4466 p = q->queue[i];
4467 p->status = 0; /* undecided */
4468 if (DIFF_PAIR_UNMERGED(p))
4469 p->status = DIFF_STATUS_UNMERGED;
4470 else if (!DIFF_FILE_VALID(p->one))
4471 p->status = DIFF_STATUS_ADDED;
4472 else if (!DIFF_FILE_VALID(p->two))
4473 p->status = DIFF_STATUS_DELETED;
4474 else if (DIFF_PAIR_TYPE_CHANGED(p))
4475 p->status = DIFF_STATUS_TYPE_CHANGED;
4476
4477 /* from this point on, we are dealing with a pair
4478 * whose both sides are valid and of the same type, i.e.
4479 * either in-place edit or rename/copy edit.
4480 */
4481 else if (DIFF_PAIR_RENAME(p)) {
4482 /*
4483 * A rename might have re-connected a broken
4484 * pair up, causing the pathnames to be the
4485 * same again. If so, that's not a rename at
4486 * all, just a modification..
4487 *
4488 * Otherwise, see if this source was used for
4489 * multiple renames, in which case we decrement
4490 * the count, and call it a copy.
4491 */
4492 if (!strcmp(p->one->path, p->two->path))
4493 p->status = DIFF_STATUS_MODIFIED;
4494 else if (--p->one->rename_used > 0)
4495 p->status = DIFF_STATUS_COPIED;
4496 else
4497 p->status = DIFF_STATUS_RENAMED;
4498 }
4499 else if (oidcmp(&p->one->oid, &p->two->oid) ||
4500 p->one->mode != p->two->mode ||
4501 p->one->dirty_submodule ||
4502 p->two->dirty_submodule ||
4503 is_null_oid(&p->one->oid))
4504 p->status = DIFF_STATUS_MODIFIED;
4505 else {
4506 /* This is a "no-change" entry and should not
4507 * happen anymore, but prepare for broken callers.
4508 */
4509 error("feeding unmodified %s to diffcore",
4510 p->one->path);
4511 p->status = DIFF_STATUS_UNKNOWN;
4512 }
4513 }
4514 diff_debug_queue("resolve-rename-copy done", q);
4515 }
4516
4517 static int check_pair_status(struct diff_filepair *p)
4518 {
4519 switch (p->status) {
4520 case DIFF_STATUS_UNKNOWN:
4521 return 0;
4522 case 0:
4523 die("internal error in diff-resolve-rename-copy");
4524 default:
4525 return 1;
4526 }
4527 }
4528
4529 static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
4530 {
4531 int fmt = opt->output_format;
4532
4533 if (fmt & DIFF_FORMAT_CHECKDIFF)
4534 diff_flush_checkdiff(p, opt);
4535 else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
4536 diff_flush_raw(p, opt);
4537 else if (fmt & DIFF_FORMAT_NAME) {
4538 const char *name_a, *name_b;
4539 name_a = p->two->path;
4540 name_b = NULL;
4541 strip_prefix(opt->prefix_length, &name_a, &name_b);
4542 fprintf(opt->file, "%s", diff_line_prefix(opt));
4543 write_name_quoted(name_a, opt->file, opt->line_termination);
4544 }
4545 }
4546
4547 static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
4548 {
4549 if (fs->mode)
4550 fprintf(file, " %s mode %06o ", newdelete, fs->mode);
4551 else
4552 fprintf(file, " %s ", newdelete);
4553 write_name_quoted(fs->path, file, '\n');
4554 }
4555
4556
4557 static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
4558 const char *line_prefix)
4559 {
4560 if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
4561 fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
4562 p->two->mode, show_name ? ' ' : '\n');
4563 if (show_name) {
4564 write_name_quoted(p->two->path, file, '\n');
4565 }
4566 }
4567 }
4568
4569 static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
4570 const char *line_prefix)
4571 {
4572 char *names = pprint_rename(p->one->path, p->two->path);
4573
4574 fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
4575 free(names);
4576 show_mode_change(file, p, 0, line_prefix);
4577 }
4578
4579 static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
4580 {
4581 FILE *file = opt->file;
4582 const char *line_prefix = diff_line_prefix(opt);
4583
4584 switch(p->status) {
4585 case DIFF_STATUS_DELETED:
4586 fputs(line_prefix, file);
4587 show_file_mode_name(file, "delete", p->one);
4588 break;
4589 case DIFF_STATUS_ADDED:
4590 fputs(line_prefix, file);
4591 show_file_mode_name(file, "create", p->two);
4592 break;
4593 case DIFF_STATUS_COPIED:
4594 fputs(line_prefix, file);
4595 show_rename_copy(file, "copy", p, line_prefix);
4596 break;
4597 case DIFF_STATUS_RENAMED:
4598 fputs(line_prefix, file);
4599 show_rename_copy(file, "rename", p, line_prefix);
4600 break;
4601 default:
4602 if (p->score) {
4603 fprintf(file, "%s rewrite ", line_prefix);
4604 write_name_quoted(p->two->path, file, ' ');
4605 fprintf(file, "(%d%%)\n", similarity_index(p));
4606 }
4607 show_mode_change(file, p, !p->score, line_prefix);
4608 break;
4609 }
4610 }
4611
4612 struct patch_id_t {
4613 git_SHA_CTX *ctx;
4614 int patchlen;
4615 };
4616
4617 static int remove_space(char *line, int len)
4618 {
4619 int i;
4620 char *dst = line;
4621 unsigned char c;
4622
4623 for (i = 0; i < len; i++)
4624 if (!isspace((c = line[i])))
4625 *dst++ = c;
4626
4627 return dst - line;
4628 }
4629
4630 static void patch_id_consume(void *priv, char *line, unsigned long len)
4631 {
4632 struct patch_id_t *data = priv;
4633 int new_len;
4634
4635 /* Ignore line numbers when computing the SHA1 of the patch */
4636 if (starts_with(line, "@@ -"))
4637 return;
4638
4639 new_len = remove_space(line, len);
4640
4641 git_SHA1_Update(data->ctx, line, new_len);
4642 data->patchlen += new_len;
4643 }
4644
4645 static void patch_id_add_string(git_SHA_CTX *ctx, const char *str)
4646 {
4647 git_SHA1_Update(ctx, str, strlen(str));
4648 }
4649
4650 static void patch_id_add_mode(git_SHA_CTX *ctx, unsigned mode)
4651 {
4652 /* large enough for 2^32 in octal */
4653 char buf[12];
4654 int len = xsnprintf(buf, sizeof(buf), "%06o", mode);
4655 git_SHA1_Update(ctx, buf, len);
4656 }
4657
4658 /* returns 0 upon success, and writes result into sha1 */
4659 static int diff_get_patch_id(struct diff_options *options, struct object_id *oid, int diff_header_only)
4660 {
4661 struct diff_queue_struct *q = &diff_queued_diff;
4662 int i;
4663 git_SHA_CTX ctx;
4664 struct patch_id_t data;
4665
4666 git_SHA1_Init(&ctx);
4667 memset(&data, 0, sizeof(struct patch_id_t));
4668 data.ctx = &ctx;
4669
4670 for (i = 0; i < q->nr; i++) {
4671 xpparam_t xpp;
4672 xdemitconf_t xecfg;
4673 mmfile_t mf1, mf2;
4674 struct diff_filepair *p = q->queue[i];
4675 int len1, len2;
4676
4677 memset(&xpp, 0, sizeof(xpp));
4678 memset(&xecfg, 0, sizeof(xecfg));
4679 if (p->status == 0)
4680 return error("internal diff status error");
4681 if (p->status == DIFF_STATUS_UNKNOWN)
4682 continue;
4683 if (diff_unmodified_pair(p))
4684 continue;
4685 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4686 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4687 continue;
4688 if (DIFF_PAIR_UNMERGED(p))
4689 continue;
4690
4691 diff_fill_oid_info(p->one);
4692 diff_fill_oid_info(p->two);
4693
4694 len1 = remove_space(p->one->path, strlen(p->one->path));
4695 len2 = remove_space(p->two->path, strlen(p->two->path));
4696 patch_id_add_string(&ctx, "diff--git");
4697 patch_id_add_string(&ctx, "a/");
4698 git_SHA1_Update(&ctx, p->one->path, len1);
4699 patch_id_add_string(&ctx, "b/");
4700 git_SHA1_Update(&ctx, p->two->path, len2);
4701
4702 if (p->one->mode == 0) {
4703 patch_id_add_string(&ctx, "newfilemode");
4704 patch_id_add_mode(&ctx, p->two->mode);
4705 patch_id_add_string(&ctx, "---/dev/null");
4706 patch_id_add_string(&ctx, "+++b/");
4707 git_SHA1_Update(&ctx, p->two->path, len2);
4708 } else if (p->two->mode == 0) {
4709 patch_id_add_string(&ctx, "deletedfilemode");
4710 patch_id_add_mode(&ctx, p->one->mode);
4711 patch_id_add_string(&ctx, "---a/");
4712 git_SHA1_Update(&ctx, p->one->path, len1);
4713 patch_id_add_string(&ctx, "+++/dev/null");
4714 } else {
4715 patch_id_add_string(&ctx, "---a/");
4716 git_SHA1_Update(&ctx, p->one->path, len1);
4717 patch_id_add_string(&ctx, "+++b/");
4718 git_SHA1_Update(&ctx, p->two->path, len2);
4719 }
4720
4721 if (diff_header_only)
4722 continue;
4723
4724 if (fill_mmfile(&mf1, p->one) < 0 ||
4725 fill_mmfile(&mf2, p->two) < 0)
4726 return error("unable to read files to diff");
4727
4728 if (diff_filespec_is_binary(p->one) ||
4729 diff_filespec_is_binary(p->two)) {
4730 git_SHA1_Update(&ctx, oid_to_hex(&p->one->oid),
4731 GIT_SHA1_HEXSZ);
4732 git_SHA1_Update(&ctx, oid_to_hex(&p->two->oid),
4733 GIT_SHA1_HEXSZ);
4734 continue;
4735 }
4736
4737 xpp.flags = 0;
4738 xecfg.ctxlen = 3;
4739 xecfg.flags = 0;
4740 if (xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4741 &xpp, &xecfg))
4742 return error("unable to generate patch-id diff for %s",
4743 p->one->path);
4744 }
4745
4746 git_SHA1_Final(oid->hash, &ctx);
4747 return 0;
4748 }
4749
4750 int diff_flush_patch_id(struct diff_options *options, struct object_id *oid, int diff_header_only)
4751 {
4752 struct diff_queue_struct *q = &diff_queued_diff;
4753 int i;
4754 int result = diff_get_patch_id(options, oid, diff_header_only);
4755
4756 for (i = 0; i < q->nr; i++)
4757 diff_free_filepair(q->queue[i]);
4758
4759 free(q->queue);
4760 DIFF_QUEUE_CLEAR(q);
4761
4762 return result;
4763 }
4764
4765 static int is_summary_empty(const struct diff_queue_struct *q)
4766 {
4767 int i;
4768
4769 for (i = 0; i < q->nr; i++) {
4770 const struct diff_filepair *p = q->queue[i];
4771
4772 switch (p->status) {
4773 case DIFF_STATUS_DELETED:
4774 case DIFF_STATUS_ADDED:
4775 case DIFF_STATUS_COPIED:
4776 case DIFF_STATUS_RENAMED:
4777 return 0;
4778 default:
4779 if (p->score)
4780 return 0;
4781 if (p->one->mode && p->two->mode &&
4782 p->one->mode != p->two->mode)
4783 return 0;
4784 break;
4785 }
4786 }
4787 return 1;
4788 }
4789
4790 static const char rename_limit_warning[] =
4791 N_("inexact rename detection was skipped due to too many files.");
4792
4793 static const char degrade_cc_to_c_warning[] =
4794 N_("only found copies from modified paths due to too many files.");
4795
4796 static const char rename_limit_advice[] =
4797 N_("you may want to set your %s variable to at least "
4798 "%d and retry the command.");
4799
4800 void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4801 {
4802 if (degraded_cc)
4803 warning(_(degrade_cc_to_c_warning));
4804 else if (needed)
4805 warning(_(rename_limit_warning));
4806 else
4807 return;
4808 if (0 < needed && needed < 32767)
4809 warning(_(rename_limit_advice), varname, needed);
4810 }
4811
4812 static void diff_flush_patch_all_file_pairs(struct diff_options *o)
4813 {
4814 int i;
4815 struct diff_queue_struct *q = &diff_queued_diff;
4816
4817 if (WSEH_NEW & WS_RULE_MASK)
4818 die("BUG: WS rules bit mask overlaps with diff symbol flags");
4819
4820 for (i = 0; i < q->nr; i++) {
4821 struct diff_filepair *p = q->queue[i];
4822 if (check_pair_status(p))
4823 diff_flush_patch(p, o);
4824 }
4825 }
4826
4827 void diff_flush(struct diff_options *options)
4828 {
4829 struct diff_queue_struct *q = &diff_queued_diff;
4830 int i, output_format = options->output_format;
4831 int separator = 0;
4832 int dirstat_by_line = 0;
4833
4834 /*
4835 * Order: raw, stat, summary, patch
4836 * or: name/name-status/checkdiff (other bits clear)
4837 */
4838 if (!q->nr)
4839 goto free_queue;
4840
4841 if (output_format & (DIFF_FORMAT_RAW |
4842 DIFF_FORMAT_NAME |
4843 DIFF_FORMAT_NAME_STATUS |
4844 DIFF_FORMAT_CHECKDIFF)) {
4845 for (i = 0; i < q->nr; i++) {
4846 struct diff_filepair *p = q->queue[i];
4847 if (check_pair_status(p))
4848 flush_one_pair(p, options);
4849 }
4850 separator++;
4851 }
4852
4853 if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4854 dirstat_by_line = 1;
4855
4856 if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4857 dirstat_by_line) {
4858 struct diffstat_t diffstat;
4859
4860 memset(&diffstat, 0, sizeof(struct diffstat_t));
4861 for (i = 0; i < q->nr; i++) {
4862 struct diff_filepair *p = q->queue[i];
4863 if (check_pair_status(p))
4864 diff_flush_stat(p, options, &diffstat);
4865 }
4866 if (output_format & DIFF_FORMAT_NUMSTAT)
4867 show_numstat(&diffstat, options);
4868 if (output_format & DIFF_FORMAT_DIFFSTAT)
4869 show_stats(&diffstat, options);
4870 if (output_format & DIFF_FORMAT_SHORTSTAT)
4871 show_shortstats(&diffstat, options);
4872 if (output_format & DIFF_FORMAT_DIRSTAT && dirstat_by_line)
4873 show_dirstat_by_line(&diffstat, options);
4874 free_diffstat_info(&diffstat);
4875 separator++;
4876 }
4877 if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4878 show_dirstat(options);
4879
4880 if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4881 for (i = 0; i < q->nr; i++) {
4882 diff_summary(options, q->queue[i]);
4883 }
4884 separator++;
4885 }
4886
4887 if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4888 DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4889 DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4890 /*
4891 * run diff_flush_patch for the exit status. setting
4892 * options->file to /dev/null should be safe, because we
4893 * aren't supposed to produce any output anyway.
4894 */
4895 if (options->close_file)
4896 fclose(options->file);
4897 options->file = xfopen("/dev/null", "w");
4898 options->close_file = 1;
4899 for (i = 0; i < q->nr; i++) {
4900 struct diff_filepair *p = q->queue[i];
4901 if (check_pair_status(p))
4902 diff_flush_patch(p, options);
4903 if (options->found_changes)
4904 break;
4905 }
4906 }
4907
4908 if (output_format & DIFF_FORMAT_PATCH) {
4909 if (separator) {
4910 emit_diff_symbol(options, DIFF_SYMBOL_SEPARATOR, NULL, 0, 0);
4911 if (options->stat_sep) {
4912 /* attach patch instead of inline */
4913 fputs(options->stat_sep, options->file);
4914 }
4915 }
4916
4917 diff_flush_patch_all_file_pairs(options);
4918 }
4919
4920 if (output_format & DIFF_FORMAT_CALLBACK)
4921 options->format_callback(q, options, options->format_callback_data);
4922
4923 for (i = 0; i < q->nr; i++)
4924 diff_free_filepair(q->queue[i]);
4925 free_queue:
4926 free(q->queue);
4927 DIFF_QUEUE_CLEAR(q);
4928 if (options->close_file)
4929 fclose(options->file);
4930
4931 /*
4932 * Report the content-level differences with HAS_CHANGES;
4933 * diff_addremove/diff_change does not set the bit when
4934 * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4935 */
4936 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4937 if (options->found_changes)
4938 DIFF_OPT_SET(options, HAS_CHANGES);
4939 else
4940 DIFF_OPT_CLR(options, HAS_CHANGES);
4941 }
4942 }
4943
4944 static int match_filter(const struct diff_options *options, const struct diff_filepair *p)
4945 {
4946 return (((p->status == DIFF_STATUS_MODIFIED) &&
4947 ((p->score &&
4948 filter_bit_tst(DIFF_STATUS_FILTER_BROKEN, options)) ||
4949 (!p->score &&
4950 filter_bit_tst(DIFF_STATUS_MODIFIED, options)))) ||
4951 ((p->status != DIFF_STATUS_MODIFIED) &&
4952 filter_bit_tst(p->status, options)));
4953 }
4954
4955 static void diffcore_apply_filter(struct diff_options *options)
4956 {
4957 int i;
4958 struct diff_queue_struct *q = &diff_queued_diff;
4959 struct diff_queue_struct outq;
4960
4961 DIFF_QUEUE_CLEAR(&outq);
4962
4963 if (!options->filter)
4964 return;
4965
4966 if (filter_bit_tst(DIFF_STATUS_FILTER_AON, options)) {
4967 int found;
4968 for (i = found = 0; !found && i < q->nr; i++) {
4969 if (match_filter(options, q->queue[i]))
4970 found++;
4971 }
4972 if (found)
4973 return;
4974
4975 /* otherwise we will clear the whole queue
4976 * by copying the empty outq at the end of this
4977 * function, but first clear the current entries
4978 * in the queue.
4979 */
4980 for (i = 0; i < q->nr; i++)
4981 diff_free_filepair(q->queue[i]);
4982 }
4983 else {
4984 /* Only the matching ones */
4985 for (i = 0; i < q->nr; i++) {
4986 struct diff_filepair *p = q->queue[i];
4987 if (match_filter(options, p))
4988 diff_q(&outq, p);
4989 else
4990 diff_free_filepair(p);
4991 }
4992 }
4993 free(q->queue);
4994 *q = outq;
4995 }
4996
4997 /* Check whether two filespecs with the same mode and size are identical */
4998 static int diff_filespec_is_identical(struct diff_filespec *one,
4999 struct diff_filespec *two)
5000 {
5001 if (S_ISGITLINK(one->mode))
5002 return 0;
5003 if (diff_populate_filespec(one, 0))
5004 return 0;
5005 if (diff_populate_filespec(two, 0))
5006 return 0;
5007 return !memcmp(one->data, two->data, one->size);
5008 }
5009
5010 static int diff_filespec_check_stat_unmatch(struct diff_filepair *p)
5011 {
5012 if (p->done_skip_stat_unmatch)
5013 return p->skip_stat_unmatch_result;
5014
5015 p->done_skip_stat_unmatch = 1;
5016 p->skip_stat_unmatch_result = 0;
5017 /*
5018 * 1. Entries that come from stat info dirtiness
5019 * always have both sides (iow, not create/delete),
5020 * one side of the object name is unknown, with
5021 * the same mode and size. Keep the ones that
5022 * do not match these criteria. They have real
5023 * differences.
5024 *
5025 * 2. At this point, the file is known to be modified,
5026 * with the same mode and size, and the object
5027 * name of one side is unknown. Need to inspect
5028 * the identical contents.
5029 */
5030 if (!DIFF_FILE_VALID(p->one) || /* (1) */
5031 !DIFF_FILE_VALID(p->two) ||
5032 (p->one->oid_valid && p->two->oid_valid) ||
5033 (p->one->mode != p->two->mode) ||
5034 diff_populate_filespec(p->one, CHECK_SIZE_ONLY) ||
5035 diff_populate_filespec(p->two, CHECK_SIZE_ONLY) ||
5036 (p->one->size != p->two->size) ||
5037 !diff_filespec_is_identical(p->one, p->two)) /* (2) */
5038 p->skip_stat_unmatch_result = 1;
5039 return p->skip_stat_unmatch_result;
5040 }
5041
5042 static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
5043 {
5044 int i;
5045 struct diff_queue_struct *q = &diff_queued_diff;
5046 struct diff_queue_struct outq;
5047 DIFF_QUEUE_CLEAR(&outq);
5048
5049 for (i = 0; i < q->nr; i++) {
5050 struct diff_filepair *p = q->queue[i];
5051
5052 if (diff_filespec_check_stat_unmatch(p))
5053 diff_q(&outq, p);
5054 else {
5055 /*
5056 * The caller can subtract 1 from skip_stat_unmatch
5057 * to determine how many paths were dirty only
5058 * due to stat info mismatch.
5059 */
5060 if (!DIFF_OPT_TST(diffopt, NO_INDEX))
5061 diffopt->skip_stat_unmatch++;
5062 diff_free_filepair(p);
5063 }
5064 }
5065 free(q->queue);
5066 *q = outq;
5067 }
5068
5069 static int diffnamecmp(const void *a_, const void *b_)
5070 {
5071 const struct diff_filepair *a = *((const struct diff_filepair **)a_);
5072 const struct diff_filepair *b = *((const struct diff_filepair **)b_);
5073 const char *name_a, *name_b;
5074
5075 name_a = a->one ? a->one->path : a->two->path;
5076 name_b = b->one ? b->one->path : b->two->path;
5077 return strcmp(name_a, name_b);
5078 }
5079
5080 void diffcore_fix_diff_index(struct diff_options *options)
5081 {
5082 struct diff_queue_struct *q = &diff_queued_diff;
5083 QSORT(q->queue, q->nr, diffnamecmp);
5084 }
5085
5086 void diffcore_std(struct diff_options *options)
5087 {
5088 /* NOTE please keep the following in sync with diff_tree_combined() */
5089 if (options->skip_stat_unmatch)
5090 diffcore_skip_stat_unmatch(options);
5091 if (!options->found_follow) {
5092 /* See try_to_follow_renames() in tree-diff.c */
5093 if (options->break_opt != -1)
5094 diffcore_break(options->break_opt);
5095 if (options->detect_rename)
5096 diffcore_rename(options);
5097 if (options->break_opt != -1)
5098 diffcore_merge_broken();
5099 }
5100 if (options->pickaxe)
5101 diffcore_pickaxe(options);
5102 if (options->orderfile)
5103 diffcore_order(options->orderfile);
5104 if (!options->found_follow)
5105 /* See try_to_follow_renames() in tree-diff.c */
5106 diff_resolve_rename_copy();
5107 diffcore_apply_filter(options);
5108
5109 if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5110 DIFF_OPT_SET(options, HAS_CHANGES);
5111 else
5112 DIFF_OPT_CLR(options, HAS_CHANGES);
5113
5114 options->found_follow = 0;
5115 }
5116
5117 int diff_result_code(struct diff_options *opt, int status)
5118 {
5119 int result = 0;
5120
5121 diff_warn_rename_limit("diff.renameLimit",
5122 opt->needed_rename_limit,
5123 opt->degraded_cc_to_c);
5124 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5125 !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
5126 return status;
5127 if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5128 DIFF_OPT_TST(opt, HAS_CHANGES))
5129 result |= 01;
5130 if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
5131 DIFF_OPT_TST(opt, CHECK_FAILED))
5132 result |= 02;
5133 return result;
5134 }
5135
5136 int diff_can_quit_early(struct diff_options *opt)
5137 {
5138 return (DIFF_OPT_TST(opt, QUICK) &&
5139 !opt->filter &&
5140 DIFF_OPT_TST(opt, HAS_CHANGES));
5141 }
5142
5143 /*
5144 * Shall changes to this submodule be ignored?
5145 *
5146 * Submodule changes can be configured to be ignored separately for each path,
5147 * but that configuration can be overridden from the command line.
5148 */
5149 static int is_submodule_ignored(const char *path, struct diff_options *options)
5150 {
5151 int ignored = 0;
5152 unsigned orig_flags = options->flags;
5153 if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
5154 set_diffopt_flags_from_submodule_config(options, path);
5155 if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
5156 ignored = 1;
5157 options->flags = orig_flags;
5158 return ignored;
5159 }
5160
5161 void diff_addremove(struct diff_options *options,
5162 int addremove, unsigned mode,
5163 const struct object_id *oid,
5164 int oid_valid,
5165 const char *concatpath, unsigned dirty_submodule)
5166 {
5167 struct diff_filespec *one, *two;
5168
5169 if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
5170 return;
5171
5172 /* This may look odd, but it is a preparation for
5173 * feeding "there are unchanged files which should
5174 * not produce diffs, but when you are doing copy
5175 * detection you would need them, so here they are"
5176 * entries to the diff-core. They will be prefixed
5177 * with something like '=' or '*' (I haven't decided
5178 * which but should not make any difference).
5179 * Feeding the same new and old to diff_change()
5180 * also has the same effect.
5181 * Before the final output happens, they are pruned after
5182 * merged into rename/copy pairs as appropriate.
5183 */
5184 if (DIFF_OPT_TST(options, REVERSE_DIFF))
5185 addremove = (addremove == '+' ? '-' :
5186 addremove == '-' ? '+' : addremove);
5187
5188 if (options->prefix &&
5189 strncmp(concatpath, options->prefix, options->prefix_length))
5190 return;
5191
5192 one = alloc_filespec(concatpath);
5193 two = alloc_filespec(concatpath);
5194
5195 if (addremove != '+')
5196 fill_filespec(one, oid, oid_valid, mode);
5197 if (addremove != '-') {
5198 fill_filespec(two, oid, oid_valid, mode);
5199 two->dirty_submodule = dirty_submodule;
5200 }
5201
5202 diff_queue(&diff_queued_diff, one, two);
5203 if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5204 DIFF_OPT_SET(options, HAS_CHANGES);
5205 }
5206
5207 void diff_change(struct diff_options *options,
5208 unsigned old_mode, unsigned new_mode,
5209 const struct object_id *old_oid,
5210 const struct object_id *new_oid,
5211 int old_oid_valid, int new_oid_valid,
5212 const char *concatpath,
5213 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
5214 {
5215 struct diff_filespec *one, *two;
5216 struct diff_filepair *p;
5217
5218 if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
5219 is_submodule_ignored(concatpath, options))
5220 return;
5221
5222 if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
5223 SWAP(old_mode, new_mode);
5224 SWAP(old_oid, new_oid);
5225 SWAP(old_oid_valid, new_oid_valid);
5226 SWAP(old_dirty_submodule, new_dirty_submodule);
5227 }
5228
5229 if (options->prefix &&
5230 strncmp(concatpath, options->prefix, options->prefix_length))
5231 return;
5232
5233 one = alloc_filespec(concatpath);
5234 two = alloc_filespec(concatpath);
5235 fill_filespec(one, old_oid, old_oid_valid, old_mode);
5236 fill_filespec(two, new_oid, new_oid_valid, new_mode);
5237 one->dirty_submodule = old_dirty_submodule;
5238 two->dirty_submodule = new_dirty_submodule;
5239 p = diff_queue(&diff_queued_diff, one, two);
5240
5241 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5242 return;
5243
5244 if (DIFF_OPT_TST(options, QUICK) && options->skip_stat_unmatch &&
5245 !diff_filespec_check_stat_unmatch(p))
5246 return;
5247
5248 DIFF_OPT_SET(options, HAS_CHANGES);
5249 }
5250
5251 struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
5252 {
5253 struct diff_filepair *pair;
5254 struct diff_filespec *one, *two;
5255
5256 if (options->prefix &&
5257 strncmp(path, options->prefix, options->prefix_length))
5258 return NULL;
5259
5260 one = alloc_filespec(path);
5261 two = alloc_filespec(path);
5262 pair = diff_queue(&diff_queued_diff, one, two);
5263 pair->is_unmerged = 1;
5264 return pair;
5265 }
5266
5267 static char *run_textconv(const char *pgm, struct diff_filespec *spec,
5268 size_t *outsize)
5269 {
5270 struct diff_tempfile *temp;
5271 const char *argv[3];
5272 const char **arg = argv;
5273 struct child_process child = CHILD_PROCESS_INIT;
5274 struct strbuf buf = STRBUF_INIT;
5275 int err = 0;
5276
5277 temp = prepare_temp_file(spec->path, spec);
5278 *arg++ = pgm;
5279 *arg++ = temp->name;
5280 *arg = NULL;
5281
5282 child.use_shell = 1;
5283 child.argv = argv;
5284 child.out = -1;
5285 if (start_command(&child)) {
5286 remove_tempfile();
5287 return NULL;
5288 }
5289
5290 if (strbuf_read(&buf, child.out, 0) < 0)
5291 err = error("error reading from textconv command '%s'", pgm);
5292 close(child.out);
5293
5294 if (finish_command(&child) || err) {
5295 strbuf_release(&buf);
5296 remove_tempfile();
5297 return NULL;
5298 }
5299 remove_tempfile();
5300
5301 return strbuf_detach(&buf, outsize);
5302 }
5303
5304 size_t fill_textconv(struct userdiff_driver *driver,
5305 struct diff_filespec *df,
5306 char **outbuf)
5307 {
5308 size_t size;
5309
5310 if (!driver) {
5311 if (!DIFF_FILE_VALID(df)) {
5312 *outbuf = "";
5313 return 0;
5314 }
5315 if (diff_populate_filespec(df, 0))
5316 die("unable to read files to diff");
5317 *outbuf = df->data;
5318 return df->size;
5319 }
5320
5321 if (!driver->textconv)
5322 die("BUG: fill_textconv called with non-textconv driver");
5323
5324 if (driver->textconv_cache && df->oid_valid) {
5325 *outbuf = notes_cache_get(driver->textconv_cache,
5326 &df->oid,
5327 &size);
5328 if (*outbuf)
5329 return size;
5330 }
5331
5332 *outbuf = run_textconv(driver->textconv, df, &size);
5333 if (!*outbuf)
5334 die("unable to read files to diff");
5335
5336 if (driver->textconv_cache && df->oid_valid) {
5337 /* ignore errors, as we might be in a readonly repository */
5338 notes_cache_put(driver->textconv_cache, &df->oid, *outbuf,
5339 size);
5340 /*
5341 * we could save up changes and flush them all at the end,
5342 * but we would need an extra call after all diffing is done.
5343 * Since generating a cache entry is the slow path anyway,
5344 * this extra overhead probably isn't a big deal.
5345 */
5346 notes_cache_write(driver->textconv_cache);
5347 }
5348
5349 return size;
5350 }
5351
5352 int textconv_object(const char *path,
5353 unsigned mode,
5354 const struct object_id *oid,
5355 int oid_valid,
5356 char **buf,
5357 unsigned long *buf_size)
5358 {
5359 struct diff_filespec *df;
5360 struct userdiff_driver *textconv;
5361
5362 df = alloc_filespec(path);
5363 fill_filespec(df, oid, oid_valid, mode);
5364 textconv = get_textconv(df);
5365 if (!textconv) {
5366 free_filespec(df);
5367 return 0;
5368 }
5369
5370 *buf_size = fill_textconv(textconv, df, buf);
5371 free_filespec(df);
5372 return 1;
5373 }
5374
5375 void setup_diff_pager(struct diff_options *opt)
5376 {
5377 /*
5378 * If the user asked for our exit code, then either they want --quiet
5379 * or --exit-code. We should definitely not bother with a pager in the
5380 * former case, as we will generate no output. Since we still properly
5381 * report our exit code even when a pager is run, we _could_ run a
5382 * pager with --exit-code. But since we have not done so historically,
5383 * and because it is easy to find people oneline advising "git diff
5384 * --exit-code" in hooks and other scripts, we do not do so.
5385 */
5386 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5387 check_pager_config("diff") != 0)
5388 setup_pager();
5389 }
Unnamed repository; edit this file 'description' to name the repository.