?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
diff.c: emit_diff_symbol learns DIFF_SYMBOL_FILEPAIR_{PLUS, MINUS}
[git.git] / diff.c
blob49b45fef29143b4e21e185e255247320ea2c4b90
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_FILEPAIR_PLUS,
565 DIFF_SYMBOL_FILEPAIR_MINUS,
566 DIFF_SYMBOL_WORDS_PORCELAIN,
567 DIFF_SYMBOL_WORDS,
568 DIFF_SYMBOL_CONTEXT,
569 DIFF_SYMBOL_CONTEXT_INCOMPLETE,
570 DIFF_SYMBOL_PLUS,
571 DIFF_SYMBOL_MINUS,
572 DIFF_SYMBOL_NO_LF_EOF,
573 DIFF_SYMBOL_CONTEXT_FRAGINFO,
574 DIFF_SYMBOL_CONTEXT_MARKER,
575 DIFF_SYMBOL_SEPARATOR
576 };
577 /*
578 * Flags for content lines:
579 * 0..12 are whitespace rules
580 * 13-15 are WSEH_NEW | WSEH_OLD | WSEH_CONTEXT
581 * 16 is marking if the line is blank at EOF
582 */
583 #define DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF (1<<16)
584 #define DIFF_SYMBOL_CONTENT_WS_MASK (WSEH_NEW | WSEH_OLD | WSEH_CONTEXT | WS_RULE_MASK)
585
586 static void emit_line_ws_markup(struct diff_options *o,
587 const char *set, const char *reset,
588 const char *line, int len, char sign,
589 unsigned ws_rule, int blank_at_eof)
590 {
591 const char *ws = NULL;
592
593 if (o->ws_error_highlight & ws_rule) {
594 ws = diff_get_color_opt(o, DIFF_WHITESPACE);
595 if (!*ws)
596 ws = NULL;
597 }
598
599 if (!ws)
600 emit_line_0(o, set, reset, sign, line, len);
601 else if (blank_at_eof)
602 /* Blank line at EOF - paint '+' as well */
603 emit_line_0(o, ws, reset, sign, line, len);
604 else {
605 /* Emit just the prefix, then the rest. */
606 emit_line_0(o, set, reset, sign, "", 0);
607 ws_check_emit(line, len, ws_rule,
608 o->file, set, reset, ws);
609 }
610 }
611
612 static void emit_diff_symbol(struct diff_options *o, enum diff_symbol s,
613 const char *line, int len, unsigned flags)
614 {
615 static const char *nneof = " No newline at end of file\n";
616 const char *context, *reset, *set, *meta;
617 switch (s) {
618 case DIFF_SYMBOL_NO_LF_EOF:
619 context = diff_get_color_opt(o, DIFF_CONTEXT);
620 reset = diff_get_color_opt(o, DIFF_RESET);
621 putc('\n', o->file);
622 emit_line_0(o, context, reset, '\\',
623 nneof, strlen(nneof));
624 break;
625 case DIFF_SYMBOL_CONTEXT_FRAGINFO:
626 emit_line(o, "", "", line, len);
627 break;
628 case DIFF_SYMBOL_CONTEXT_INCOMPLETE:
629 case DIFF_SYMBOL_CONTEXT_MARKER:
630 context = diff_get_color_opt(o, DIFF_CONTEXT);
631 reset = diff_get_color_opt(o, DIFF_RESET);
632 emit_line(o, context, reset, line, len);
633 break;
634 case DIFF_SYMBOL_SEPARATOR:
635 fprintf(o->file, "%s%c",
636 diff_line_prefix(o),
637 o->line_termination);
638 break;
639 case DIFF_SYMBOL_CONTEXT:
640 set = diff_get_color_opt(o, DIFF_CONTEXT);
641 reset = diff_get_color_opt(o, DIFF_RESET);
642 emit_line_ws_markup(o, set, reset, line, len, ' ',
643 flags & (DIFF_SYMBOL_CONTENT_WS_MASK), 0);
644 break;
645 case DIFF_SYMBOL_PLUS:
646 set = diff_get_color_opt(o, DIFF_FILE_NEW);
647 reset = diff_get_color_opt(o, DIFF_RESET);
648 emit_line_ws_markup(o, set, reset, line, len, '+',
649 flags & DIFF_SYMBOL_CONTENT_WS_MASK,
650 flags & DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF);
651 break;
652 case DIFF_SYMBOL_MINUS:
653 set = diff_get_color_opt(o, DIFF_FILE_OLD);
654 reset = diff_get_color_opt(o, DIFF_RESET);
655 emit_line_ws_markup(o, set, reset, line, len, '-',
656 flags & DIFF_SYMBOL_CONTENT_WS_MASK, 0);
657 break;
658 case DIFF_SYMBOL_WORDS_PORCELAIN:
659 context = diff_get_color_opt(o, DIFF_CONTEXT);
660 reset = diff_get_color_opt(o, DIFF_RESET);
661 emit_line(o, context, reset, line, len);
662 fputs("~\n", o->file);
663 break;
664 case DIFF_SYMBOL_WORDS:
665 context = diff_get_color_opt(o, DIFF_CONTEXT);
666 reset = diff_get_color_opt(o, DIFF_RESET);
667 /*
668 * Skip the prefix character, if any. With
669 * diff_suppress_blank_empty, there may be
670 * none.
671 */
672 if (line[0] != '\n') {
673 line++;
674 len--;
675 }
676 emit_line(o, context, reset, line, len);
677 break;
678 case DIFF_SYMBOL_FILEPAIR_PLUS:
679 meta = diff_get_color_opt(o, DIFF_METAINFO);
680 reset = diff_get_color_opt(o, DIFF_RESET);
681 fprintf(o->file, "%s%s+++ %s%s%s\n", diff_line_prefix(o), meta,
682 line, reset,
683 strchr(line, ' ') ? "\t" : "");
684 break;
685 case DIFF_SYMBOL_FILEPAIR_MINUS:
686 meta = diff_get_color_opt(o, DIFF_METAINFO);
687 reset = diff_get_color_opt(o, DIFF_RESET);
688 fprintf(o->file, "%s%s--- %s%s%s\n", diff_line_prefix(o), meta,
689 line, reset,
690 strchr(line, ' ') ? "\t" : "");
691 break;
692 default:
693 die("BUG: unknown diff symbol");
694 }
695 }
696
697 static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
698 {
699 if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
700 ecbdata->blank_at_eof_in_preimage &&
701 ecbdata->blank_at_eof_in_postimage &&
702 ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
703 ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
704 return 0;
705 return ws_blank_line(line, len, ecbdata->ws_rule);
706 }
707
708 static void emit_add_line(const char *reset,
709 struct emit_callback *ecbdata,
710 const char *line, int len)
711 {
712 unsigned flags = WSEH_NEW | ecbdata->ws_rule;
713 if (new_blank_line_at_eof(ecbdata, line, len))
714 flags |= DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF;
715
716 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_PLUS, line, len, flags);
717 }
718
719 static void emit_del_line(const char *reset,
720 struct emit_callback *ecbdata,
721 const char *line, int len)
722 {
723 unsigned flags = WSEH_OLD | ecbdata->ws_rule;
724 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_MINUS, line, len, flags);
725 }
726
727 static void emit_context_line(const char *reset,
728 struct emit_callback *ecbdata,
729 const char *line, int len)
730 {
731 unsigned flags = WSEH_CONTEXT | ecbdata->ws_rule;
732 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_CONTEXT, line, len, flags);
733 }
734
735 static void emit_hunk_header(struct emit_callback *ecbdata,
736 const char *line, int len)
737 {
738 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
739 const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
740 const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
741 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
742 static const char atat[2] = { '@', '@' };
743 const char *cp, *ep;
744 struct strbuf msgbuf = STRBUF_INIT;
745 int org_len = len;
746 int i = 1;
747
748 /*
749 * As a hunk header must begin with "@@ -<old>, +<new> @@",
750 * it always is at least 10 bytes long.
751 */
752 if (len < 10 ||
753 memcmp(line, atat, 2) ||
754 !(ep = memmem(line + 2, len - 2, atat, 2))) {
755 emit_diff_symbol(ecbdata->opt,
756 DIFF_SYMBOL_CONTEXT_MARKER, line, len, 0);
757 return;
758 }
759 ep += 2; /* skip over @@ */
760
761 /* The hunk header in fraginfo color */
762 strbuf_addstr(&msgbuf, frag);
763 strbuf_add(&msgbuf, line, ep - line);
764 strbuf_addstr(&msgbuf, reset);
765
766 /*
767 * trailing "\r\n"
768 */
769 for ( ; i < 3; i++)
770 if (line[len - i] == '\r' || line[len - i] == '\n')
771 len--;
772
773 /* blank before the func header */
774 for (cp = ep; ep - line < len; ep++)
775 if (*ep != ' ' && *ep != '\t')
776 break;
777 if (ep != cp) {
778 strbuf_addstr(&msgbuf, context);
779 strbuf_add(&msgbuf, cp, ep - cp);
780 strbuf_addstr(&msgbuf, reset);
781 }
782
783 if (ep < line + len) {
784 strbuf_addstr(&msgbuf, func);
785 strbuf_add(&msgbuf, ep, line + len - ep);
786 strbuf_addstr(&msgbuf, reset);
787 }
788
789 strbuf_add(&msgbuf, line + len, org_len - len);
790 strbuf_complete_line(&msgbuf);
791 emit_diff_symbol(ecbdata->opt,
792 DIFF_SYMBOL_CONTEXT_FRAGINFO, msgbuf.buf, msgbuf.len, 0);
793 strbuf_release(&msgbuf);
794 }
795
796 static struct diff_tempfile *claim_diff_tempfile(void) {
797 int i;
798 for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
799 if (!diff_temp[i].name)
800 return diff_temp + i;
801 die("BUG: diff is failing to clean up its tempfiles");
802 }
803
804 static void remove_tempfile(void)
805 {
806 int i;
807 for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
808 if (is_tempfile_active(&diff_temp[i].tempfile))
809 delete_tempfile(&diff_temp[i].tempfile);
810 diff_temp[i].name = NULL;
811 }
812 }
813
814 static void print_line_count(FILE *file, int count)
815 {
816 switch (count) {
817 case 0:
818 fprintf(file, "0,0");
819 break;
820 case 1:
821 fprintf(file, "1");
822 break;
823 default:
824 fprintf(file, "1,%d", count);
825 break;
826 }
827 }
828
829 static void emit_rewrite_lines(struct emit_callback *ecb,
830 int prefix, const char *data, int size)
831 {
832 const char *endp = NULL;
833 const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
834
835 while (0 < size) {
836 int len;
837
838 endp = memchr(data, '\n', size);
839 len = endp ? (endp - data + 1) : size;
840 if (prefix != '+') {
841 ecb->lno_in_preimage++;
842 emit_del_line(reset, ecb, data, len);
843 } else {
844 ecb->lno_in_postimage++;
845 emit_add_line(reset, ecb, data, len);
846 }
847 size -= len;
848 data += len;
849 }
850 if (!endp)
851 emit_diff_symbol(ecb->opt, DIFF_SYMBOL_NO_LF_EOF, NULL, 0, 0);
852 }
853
854 static void emit_rewrite_diff(const char *name_a,
855 const char *name_b,
856 struct diff_filespec *one,
857 struct diff_filespec *two,
858 struct userdiff_driver *textconv_one,
859 struct userdiff_driver *textconv_two,
860 struct diff_options *o)
861 {
862 int lc_a, lc_b;
863 const char *fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
864 const char *reset = diff_get_color(o->use_color, DIFF_RESET);
865 static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
866 const char *a_prefix, *b_prefix;
867 char *data_one, *data_two;
868 size_t size_one, size_two;
869 struct emit_callback ecbdata;
870 const char *line_prefix = diff_line_prefix(o);
871
872 if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
873 a_prefix = o->b_prefix;
874 b_prefix = o->a_prefix;
875 } else {
876 a_prefix = o->a_prefix;
877 b_prefix = o->b_prefix;
878 }
879
880 name_a += (*name_a == '/');
881 name_b += (*name_b == '/');
882
883 strbuf_reset(&a_name);
884 strbuf_reset(&b_name);
885 quote_two_c_style(&a_name, a_prefix, name_a, 0);
886 quote_two_c_style(&b_name, b_prefix, name_b, 0);
887
888 size_one = fill_textconv(textconv_one, one, &data_one);
889 size_two = fill_textconv(textconv_two, two, &data_two);
890
891 memset(&ecbdata, 0, sizeof(ecbdata));
892 ecbdata.color_diff = want_color(o->use_color);
893 ecbdata.ws_rule = whitespace_rule(name_b);
894 ecbdata.opt = o;
895 if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
896 mmfile_t mf1, mf2;
897 mf1.ptr = (char *)data_one;
898 mf2.ptr = (char *)data_two;
899 mf1.size = size_one;
900 mf2.size = size_two;
901 check_blank_at_eof(&mf1, &mf2, &ecbdata);
902 }
903 ecbdata.lno_in_preimage = 1;
904 ecbdata.lno_in_postimage = 1;
905
906 lc_a = count_lines(data_one, size_one);
907 lc_b = count_lines(data_two, size_two);
908
909 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_MINUS,
910 a_name.buf, a_name.len, 0);
911 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_PLUS,
912 b_name.buf, b_name.len, 0);
913
914 fprintf(o->file, "%s%s@@ -", line_prefix, fraginfo);
915 if (!o->irreversible_delete)
916 print_line_count(o->file, lc_a);
917 else
918 fprintf(o->file, "?,?");
919 fprintf(o->file, " +");
920 print_line_count(o->file, lc_b);
921 fprintf(o->file, " @@%s\n", reset);
922 if (lc_a && !o->irreversible_delete)
923 emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
924 if (lc_b)
925 emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
926 if (textconv_one)
927 free((char *)data_one);
928 if (textconv_two)
929 free((char *)data_two);
930 }
931
932 struct diff_words_buffer {
933 mmfile_t text;
934 long alloc;
935 struct diff_words_orig {
936 const char *begin, *end;
937 } *orig;
938 int orig_nr, orig_alloc;
939 };
940
941 static void diff_words_append(char *line, unsigned long len,
942 struct diff_words_buffer *buffer)
943 {
944 ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
945 line++;
946 len--;
947 memcpy(buffer->text.ptr + buffer->text.size, line, len);
948 buffer->text.size += len;
949 buffer->text.ptr[buffer->text.size] = '\0';
950 }
951
952 struct diff_words_style_elem {
953 const char *prefix;
954 const char *suffix;
955 const char *color; /* NULL; filled in by the setup code if
956 * color is enabled */
957 };
958
959 struct diff_words_style {
960 enum diff_words_type type;
961 struct diff_words_style_elem new, old, ctx;
962 const char *newline;
963 };
964
965 static struct diff_words_style diff_words_styles[] = {
966 { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
967 { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
968 { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
969 };
970
971 struct diff_words_data {
972 struct diff_words_buffer minus, plus;
973 const char *current_plus;
974 int last_minus;
975 struct diff_options *opt;
976 regex_t *word_regex;
977 enum diff_words_type type;
978 struct diff_words_style *style;
979 };
980
981 static int fn_out_diff_words_write_helper(FILE *fp,
982 struct diff_words_style_elem *st_el,
983 const char *newline,
984 size_t count, const char *buf,
985 const char *line_prefix)
986 {
987 int print = 0;
988
989 while (count) {
990 char *p = memchr(buf, '\n', count);
991 if (print)
992 fputs(line_prefix, fp);
993 if (p != buf) {
994 if (st_el->color && fputs(st_el->color, fp) < 0)
995 return -1;
996 if (fputs(st_el->prefix, fp) < 0 ||
997 fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
998 fputs(st_el->suffix, fp) < 0)
999 return -1;
1000 if (st_el->color && *st_el->color
1001 && fputs(GIT_COLOR_RESET, fp) < 0)
1002 return -1;
1003 }
1004 if (!p)
1005 return 0;
1006 if (fputs(newline, fp) < 0)
1007 return -1;
1008 count -= p + 1 - buf;
1009 buf = p + 1;
1010 print = 1;
1011 }
1012 return 0;
1013 }
1014
1015 /*
1016 * '--color-words' algorithm can be described as:
1017 *
1018 * 1. collect the minus/plus lines of a diff hunk, divided into
1019 * minus-lines and plus-lines;
1020 *
1021 * 2. break both minus-lines and plus-lines into words and
1022 * place them into two mmfile_t with one word for each line;
1023 *
1024 * 3. use xdiff to run diff on the two mmfile_t to get the words level diff;
1025 *
1026 * And for the common parts of the both file, we output the plus side text.
1027 * diff_words->current_plus is used to trace the current position of the plus file
1028 * which printed. diff_words->last_minus is used to trace the last minus word
1029 * printed.
1030 *
1031 * For '--graph' to work with '--color-words', we need to output the graph prefix
1032 * on each line of color words output. Generally, there are two conditions on
1033 * which we should output the prefix.
1034 *
1035 * 1. diff_words->last_minus == 0 &&
1036 * diff_words->current_plus == diff_words->plus.text.ptr
1037 *
1038 * that is: the plus text must start as a new line, and if there is no minus
1039 * word printed, a graph prefix must be printed.
1040 *
1041 * 2. diff_words->current_plus > diff_words->plus.text.ptr &&
1042 * *(diff_words->current_plus - 1) == '\n'
1043 *
1044 * that is: a graph prefix must be printed following a '\n'
1045 */
1046 static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
1047 {
1048 if ((diff_words->last_minus == 0 &&
1049 diff_words->current_plus == diff_words->plus.text.ptr) ||
1050 (diff_words->current_plus > diff_words->plus.text.ptr &&
1051 *(diff_words->current_plus - 1) == '\n')) {
1052 return 1;
1053 } else {
1054 return 0;
1055 }
1056 }
1057
1058 static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
1059 {
1060 struct diff_words_data *diff_words = priv;
1061 struct diff_words_style *style = diff_words->style;
1062 int minus_first, minus_len, plus_first, plus_len;
1063 const char *minus_begin, *minus_end, *plus_begin, *plus_end;
1064 struct diff_options *opt = diff_words->opt;
1065 const char *line_prefix;
1066
1067 if (line[0] != '@' || parse_hunk_header(line, len,
1068 &minus_first, &minus_len, &plus_first, &plus_len))
1069 return;
1070
1071 assert(opt);
1072 line_prefix = diff_line_prefix(opt);
1073
1074 /* POSIX requires that first be decremented by one if len == 0... */
1075 if (minus_len) {
1076 minus_begin = diff_words->minus.orig[minus_first].begin;
1077 minus_end =
1078 diff_words->minus.orig[minus_first + minus_len - 1].end;
1079 } else
1080 minus_begin = minus_end =
1081 diff_words->minus.orig[minus_first].end;
1082
1083 if (plus_len) {
1084 plus_begin = diff_words->plus.orig[plus_first].begin;
1085 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
1086 } else
1087 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
1088
1089 if (color_words_output_graph_prefix(diff_words)) {
1090 fputs(line_prefix, diff_words->opt->file);
1091 }
1092 if (diff_words->current_plus != plus_begin) {
1093 fn_out_diff_words_write_helper(diff_words->opt->file,
1094 &style->ctx, style->newline,
1095 plus_begin - diff_words->current_plus,
1096 diff_words->current_plus, line_prefix);
1097 if (*(plus_begin - 1) == '\n')
1098 fputs(line_prefix, diff_words->opt->file);
1099 }
1100 if (minus_begin != minus_end) {
1101 fn_out_diff_words_write_helper(diff_words->opt->file,
1102 &style->old, style->newline,
1103 minus_end - minus_begin, minus_begin,
1104 line_prefix);
1105 }
1106 if (plus_begin != plus_end) {
1107 fn_out_diff_words_write_helper(diff_words->opt->file,
1108 &style->new, style->newline,
1109 plus_end - plus_begin, plus_begin,
1110 line_prefix);
1111 }
1112
1113 diff_words->current_plus = plus_end;
1114 diff_words->last_minus = minus_first;
1115 }
1116
1117 /* This function starts looking at *begin, and returns 0 iff a word was found. */
1118 static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
1119 int *begin, int *end)
1120 {
1121 if (word_regex && *begin < buffer->size) {
1122 regmatch_t match[1];
1123 if (!regexec_buf(word_regex, buffer->ptr + *begin,
1124 buffer->size - *begin, 1, match, 0)) {
1125 char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
1126 '\n', match[0].rm_eo - match[0].rm_so);
1127 *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
1128 *begin += match[0].rm_so;
1129 return *begin >= *end;
1130 }
1131 return -1;
1132 }
1133
1134 /* find the next word */
1135 while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
1136 (*begin)++;
1137 if (*begin >= buffer->size)
1138 return -1;
1139
1140 /* find the end of the word */
1141 *end = *begin + 1;
1142 while (*end < buffer->size && !isspace(buffer->ptr[*end]))
1143 (*end)++;
1144
1145 return 0;
1146 }
1147
1148 /*
1149 * This function splits the words in buffer->text, stores the list with
1150 * newline separator into out, and saves the offsets of the original words
1151 * in buffer->orig.
1152 */
1153 static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
1154 regex_t *word_regex)
1155 {
1156 int i, j;
1157 long alloc = 0;
1158
1159 out->size = 0;
1160 out->ptr = NULL;
1161
1162 /* fake an empty "0th" word */
1163 ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
1164 buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
1165 buffer->orig_nr = 1;
1166
1167 for (i = 0; i < buffer->text.size; i++) {
1168 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
1169 return;
1170
1171 /* store original boundaries */
1172 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
1173 buffer->orig_alloc);
1174 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
1175 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
1176 buffer->orig_nr++;
1177
1178 /* store one word */
1179 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
1180 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
1181 out->ptr[out->size + j - i] = '\n';
1182 out->size += j - i + 1;
1183
1184 i = j - 1;
1185 }
1186 }
1187
1188 /* this executes the word diff on the accumulated buffers */
1189 static void diff_words_show(struct diff_words_data *diff_words)
1190 {
1191 xpparam_t xpp;
1192 xdemitconf_t xecfg;
1193 mmfile_t minus, plus;
1194 struct diff_words_style *style = diff_words->style;
1195
1196 struct diff_options *opt = diff_words->opt;
1197 const char *line_prefix;
1198
1199 assert(opt);
1200 line_prefix = diff_line_prefix(opt);
1201
1202 /* special case: only removal */
1203 if (!diff_words->plus.text.size) {
1204 fputs(line_prefix, diff_words->opt->file);
1205 fn_out_diff_words_write_helper(diff_words->opt->file,
1206 &style->old, style->newline,
1207 diff_words->minus.text.size,
1208 diff_words->minus.text.ptr, line_prefix);
1209 diff_words->minus.text.size = 0;
1210 return;
1211 }
1212
1213 diff_words->current_plus = diff_words->plus.text.ptr;
1214 diff_words->last_minus = 0;
1215
1216 memset(&xpp, 0, sizeof(xpp));
1217 memset(&xecfg, 0, sizeof(xecfg));
1218 diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
1219 diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
1220 xpp.flags = 0;
1221 /* as only the hunk header will be parsed, we need a 0-context */
1222 xecfg.ctxlen = 0;
1223 if (xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
1224 &xpp, &xecfg))
1225 die("unable to generate word diff");
1226 free(minus.ptr);
1227 free(plus.ptr);
1228 if (diff_words->current_plus != diff_words->plus.text.ptr +
1229 diff_words->plus.text.size) {
1230 if (color_words_output_graph_prefix(diff_words))
1231 fputs(line_prefix, diff_words->opt->file);
1232 fn_out_diff_words_write_helper(diff_words->opt->file,
1233 &style->ctx, style->newline,
1234 diff_words->plus.text.ptr + diff_words->plus.text.size
1235 - diff_words->current_plus, diff_words->current_plus,
1236 line_prefix);
1237 }
1238 diff_words->minus.text.size = diff_words->plus.text.size = 0;
1239 }
1240
1241 /* In "color-words" mode, show word-diff of words accumulated in the buffer */
1242 static void diff_words_flush(struct emit_callback *ecbdata)
1243 {
1244 if (ecbdata->diff_words->minus.text.size ||
1245 ecbdata->diff_words->plus.text.size)
1246 diff_words_show(ecbdata->diff_words);
1247 }
1248
1249 static void diff_filespec_load_driver(struct diff_filespec *one)
1250 {
1251 /* Use already-loaded driver */
1252 if (one->driver)
1253 return;
1254
1255 if (S_ISREG(one->mode))
1256 one->driver = userdiff_find_by_path(one->path);
1257
1258 /* Fallback to default settings */
1259 if (!one->driver)
1260 one->driver = userdiff_find_by_name("default");
1261 }
1262
1263 static const char *userdiff_word_regex(struct diff_filespec *one)
1264 {
1265 diff_filespec_load_driver(one);
1266 return one->driver->word_regex;
1267 }
1268
1269 static void init_diff_words_data(struct emit_callback *ecbdata,
1270 struct diff_options *orig_opts,
1271 struct diff_filespec *one,
1272 struct diff_filespec *two)
1273 {
1274 int i;
1275 struct diff_options *o = xmalloc(sizeof(struct diff_options));
1276 memcpy(o, orig_opts, sizeof(struct diff_options));
1277
1278 ecbdata->diff_words =
1279 xcalloc(1, sizeof(struct diff_words_data));
1280 ecbdata->diff_words->type = o->word_diff;
1281 ecbdata->diff_words->opt = o;
1282 if (!o->word_regex)
1283 o->word_regex = userdiff_word_regex(one);
1284 if (!o->word_regex)
1285 o->word_regex = userdiff_word_regex(two);
1286 if (!o->word_regex)
1287 o->word_regex = diff_word_regex_cfg;
1288 if (o->word_regex) {
1289 ecbdata->diff_words->word_regex = (regex_t *)
1290 xmalloc(sizeof(regex_t));
1291 if (regcomp(ecbdata->diff_words->word_regex,
1292 o->word_regex,
1293 REG_EXTENDED | REG_NEWLINE))
1294 die ("Invalid regular expression: %s",
1295 o->word_regex);
1296 }
1297 for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1298 if (o->word_diff == diff_words_styles[i].type) {
1299 ecbdata->diff_words->style =
1300 &diff_words_styles[i];
1301 break;
1302 }
1303 }
1304 if (want_color(o->use_color)) {
1305 struct diff_words_style *st = ecbdata->diff_words->style;
1306 st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1307 st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1308 st->ctx.color = diff_get_color_opt(o, DIFF_CONTEXT);
1309 }
1310 }
1311
1312 static void free_diff_words_data(struct emit_callback *ecbdata)
1313 {
1314 if (ecbdata->diff_words) {
1315 diff_words_flush(ecbdata);
1316 free (ecbdata->diff_words->opt);
1317 free (ecbdata->diff_words->minus.text.ptr);
1318 free (ecbdata->diff_words->minus.orig);
1319 free (ecbdata->diff_words->plus.text.ptr);
1320 free (ecbdata->diff_words->plus.orig);
1321 if (ecbdata->diff_words->word_regex) {
1322 regfree(ecbdata->diff_words->word_regex);
1323 free(ecbdata->diff_words->word_regex);
1324 }
1325 FREE_AND_NULL(ecbdata->diff_words);
1326 }
1327 }
1328
1329 const char *diff_get_color(int diff_use_color, enum color_diff ix)
1330 {
1331 if (want_color(diff_use_color))
1332 return diff_colors[ix];
1333 return "";
1334 }
1335
1336 const char *diff_line_prefix(struct diff_options *opt)
1337 {
1338 struct strbuf *msgbuf;
1339 if (!opt->output_prefix)
1340 return "";
1341
1342 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
1343 return msgbuf->buf;
1344 }
1345
1346 static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1347 {
1348 const char *cp;
1349 unsigned long allot;
1350 size_t l = len;
1351
1352 if (ecb->truncate)
1353 return ecb->truncate(line, len);
1354 cp = line;
1355 allot = l;
1356 while (0 < l) {
1357 (void) utf8_width(&cp, &l);
1358 if (!cp)
1359 break; /* truncated in the middle? */
1360 }
1361 return allot - l;
1362 }
1363
1364 static void find_lno(const char *line, struct emit_callback *ecbdata)
1365 {
1366 const char *p;
1367 ecbdata->lno_in_preimage = 0;
1368 ecbdata->lno_in_postimage = 0;
1369 p = strchr(line, '-');
1370 if (!p)
1371 return; /* cannot happen */
1372 ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1373 p = strchr(p, '+');
1374 if (!p)
1375 return; /* cannot happen */
1376 ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1377 }
1378
1379 static void fn_out_consume(void *priv, char *line, unsigned long len)
1380 {
1381 struct emit_callback *ecbdata = priv;
1382 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1383 struct diff_options *o = ecbdata->opt;
1384
1385 o->found_changes = 1;
1386
1387 if (ecbdata->header) {
1388 fprintf(o->file, "%s", ecbdata->header->buf);
1389 strbuf_reset(ecbdata->header);
1390 ecbdata->header = NULL;
1391 }
1392
1393 if (ecbdata->label_path[0]) {
1394 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_MINUS,
1395 ecbdata->label_path[0],
1396 strlen(ecbdata->label_path[0]), 0);
1397 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_PLUS,
1398 ecbdata->label_path[1],
1399 strlen(ecbdata->label_path[1]), 0);
1400 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1401 }
1402
1403 if (diff_suppress_blank_empty
1404 && len == 2 && line[0] == ' ' && line[1] == '\n') {
1405 line[0] = '\n';
1406 len = 1;
1407 }
1408
1409 if (line[0] == '@') {
1410 if (ecbdata->diff_words)
1411 diff_words_flush(ecbdata);
1412 len = sane_truncate_line(ecbdata, line, len);
1413 find_lno(line, ecbdata);
1414 emit_hunk_header(ecbdata, line, len);
1415 return;
1416 }
1417
1418 if (ecbdata->diff_words) {
1419 enum diff_symbol s =
1420 ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN ?
1421 DIFF_SYMBOL_WORDS_PORCELAIN : DIFF_SYMBOL_WORDS;
1422 if (line[0] == '-') {
1423 diff_words_append(line, len,
1424 &ecbdata->diff_words->minus);
1425 return;
1426 } else if (line[0] == '+') {
1427 diff_words_append(line, len,
1428 &ecbdata->diff_words->plus);
1429 return;
1430 } else if (starts_with(line, "\\ ")) {
1431 /*
1432 * Eat the "no newline at eof" marker as if we
1433 * saw a "+" or "-" line with nothing on it,
1434 * and return without diff_words_flush() to
1435 * defer processing. If this is the end of
1436 * preimage, more "+" lines may come after it.
1437 */
1438 return;
1439 }
1440 diff_words_flush(ecbdata);
1441 emit_diff_symbol(o, s, line, len, 0);
1442 return;
1443 }
1444
1445 switch (line[0]) {
1446 case '+':
1447 ecbdata->lno_in_postimage++;
1448 emit_add_line(reset, ecbdata, line + 1, len - 1);
1449 break;
1450 case '-':
1451 ecbdata->lno_in_preimage++;
1452 emit_del_line(reset, ecbdata, line + 1, len - 1);
1453 break;
1454 case ' ':
1455 ecbdata->lno_in_postimage++;
1456 ecbdata->lno_in_preimage++;
1457 emit_context_line(reset, ecbdata, line + 1, len - 1);
1458 break;
1459 default:
1460 /* incomplete line at the end */
1461 ecbdata->lno_in_preimage++;
1462 emit_diff_symbol(o, DIFF_SYMBOL_CONTEXT_INCOMPLETE,
1463 line, len, 0);
1464 break;
1465 }
1466 }
1467
1468 static char *pprint_rename(const char *a, const char *b)
1469 {
1470 const char *old = a;
1471 const char *new = b;
1472 struct strbuf name = STRBUF_INIT;
1473 int pfx_length, sfx_length;
1474 int pfx_adjust_for_slash;
1475 int len_a = strlen(a);
1476 int len_b = strlen(b);
1477 int a_midlen, b_midlen;
1478 int qlen_a = quote_c_style(a, NULL, NULL, 0);
1479 int qlen_b = quote_c_style(b, NULL, NULL, 0);
1480
1481 if (qlen_a || qlen_b) {
1482 quote_c_style(a, &name, NULL, 0);
1483 strbuf_addstr(&name, " => ");
1484 quote_c_style(b, &name, NULL, 0);
1485 return strbuf_detach(&name, NULL);
1486 }
1487
1488 /* Find common prefix */
1489 pfx_length = 0;
1490 while (*old && *new && *old == *new) {
1491 if (*old == '/')
1492 pfx_length = old - a + 1;
1493 old++;
1494 new++;
1495 }
1496
1497 /* Find common suffix */
1498 old = a + len_a;
1499 new = b + len_b;
1500 sfx_length = 0;
1501 /*
1502 * If there is a common prefix, it must end in a slash. In
1503 * that case we let this loop run 1 into the prefix to see the
1504 * same slash.
1505 *
1506 * If there is no common prefix, we cannot do this as it would
1507 * underrun the input strings.
1508 */
1509 pfx_adjust_for_slash = (pfx_length ? 1 : 0);
1510 while (a + pfx_length - pfx_adjust_for_slash <= old &&
1511 b + pfx_length - pfx_adjust_for_slash <= new &&
1512 *old == *new) {
1513 if (*old == '/')
1514 sfx_length = len_a - (old - a);
1515 old--;
1516 new--;
1517 }
1518
1519 /*
1520 * pfx{mid-a => mid-b}sfx
1521 * {pfx-a => pfx-b}sfx
1522 * pfx{sfx-a => sfx-b}
1523 * name-a => name-b
1524 */
1525 a_midlen = len_a - pfx_length - sfx_length;
1526 b_midlen = len_b - pfx_length - sfx_length;
1527 if (a_midlen < 0)
1528 a_midlen = 0;
1529 if (b_midlen < 0)
1530 b_midlen = 0;
1531
1532 strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1533 if (pfx_length + sfx_length) {
1534 strbuf_add(&name, a, pfx_length);
1535 strbuf_addch(&name, '{');
1536 }
1537 strbuf_add(&name, a + pfx_length, a_midlen);
1538 strbuf_addstr(&name, " => ");
1539 strbuf_add(&name, b + pfx_length, b_midlen);
1540 if (pfx_length + sfx_length) {
1541 strbuf_addch(&name, '}');
1542 strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1543 }
1544 return strbuf_detach(&name, NULL);
1545 }
1546
1547 struct diffstat_t {
1548 int nr;
1549 int alloc;
1550 struct diffstat_file {
1551 char *from_name;
1552 char *name;
1553 char *print_name;
1554 unsigned is_unmerged:1;
1555 unsigned is_binary:1;
1556 unsigned is_renamed:1;
1557 unsigned is_interesting:1;
1558 uintmax_t added, deleted;
1559 } **files;
1560 };
1561
1562 static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1563 const char *name_a,
1564 const char *name_b)
1565 {
1566 struct diffstat_file *x;
1567 x = xcalloc(1, sizeof(*x));
1568 ALLOC_GROW(diffstat->files, diffstat->nr + 1, diffstat->alloc);
1569 diffstat->files[diffstat->nr++] = x;
1570 if (name_b) {
1571 x->from_name = xstrdup(name_a);
1572 x->name = xstrdup(name_b);
1573 x->is_renamed = 1;
1574 }
1575 else {
1576 x->from_name = NULL;
1577 x->name = xstrdup(name_a);
1578 }
1579 return x;
1580 }
1581
1582 static void diffstat_consume(void *priv, char *line, unsigned long len)
1583 {
1584 struct diffstat_t *diffstat = priv;
1585 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1586
1587 if (line[0] == '+')
1588 x->added++;
1589 else if (line[0] == '-')
1590 x->deleted++;
1591 }
1592
1593 const char mime_boundary_leader[] = "------------";
1594
1595 static int scale_linear(int it, int width, int max_change)
1596 {
1597 if (!it)
1598 return 0;
1599 /*
1600 * make sure that at least one '-' or '+' is printed if
1601 * there is any change to this path. The easiest way is to
1602 * scale linearly as if the alloted width is one column shorter
1603 * than it is, and then add 1 to the result.
1604 */
1605 return 1 + (it * (width - 1) / max_change);
1606 }
1607
1608 static void show_name(FILE *file,
1609 const char *prefix, const char *name, int len)
1610 {
1611 fprintf(file, " %s%-*s |", prefix, len, name);
1612 }
1613
1614 static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1615 {
1616 if (cnt <= 0)
1617 return;
1618 fprintf(file, "%s", set);
1619 while (cnt--)
1620 putc(ch, file);
1621 fprintf(file, "%s", reset);
1622 }
1623
1624 static void fill_print_name(struct diffstat_file *file)
1625 {
1626 char *pname;
1627
1628 if (file->print_name)
1629 return;
1630
1631 if (!file->is_renamed) {
1632 struct strbuf buf = STRBUF_INIT;
1633 if (quote_c_style(file->name, &buf, NULL, 0)) {
1634 pname = strbuf_detach(&buf, NULL);
1635 } else {
1636 pname = file->name;
1637 strbuf_release(&buf);
1638 }
1639 } else {
1640 pname = pprint_rename(file->from_name, file->name);
1641 }
1642 file->print_name = pname;
1643 }
1644
1645 int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1646 {
1647 struct strbuf sb = STRBUF_INIT;
1648 int ret;
1649
1650 if (!files) {
1651 assert(insertions == 0 && deletions == 0);
1652 return fprintf(fp, "%s\n", " 0 files changed");
1653 }
1654
1655 strbuf_addf(&sb,
1656 (files == 1) ? " %d file changed" : " %d files changed",
1657 files);
1658
1659 /*
1660 * For binary diff, the caller may want to print "x files
1661 * changed" with insertions == 0 && deletions == 0.
1662 *
1663 * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1664 * is probably less confusing (i.e skip over "2 files changed
1665 * but nothing about added/removed lines? Is this a bug in Git?").
1666 */
1667 if (insertions || deletions == 0) {
1668 strbuf_addf(&sb,
1669 (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
1670 insertions);
1671 }
1672
1673 if (deletions || insertions == 0) {
1674 strbuf_addf(&sb,
1675 (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
1676 deletions);
1677 }
1678 strbuf_addch(&sb, '\n');
1679 ret = fputs(sb.buf, fp);
1680 strbuf_release(&sb);
1681 return ret;
1682 }
1683
1684 static void show_stats(struct diffstat_t *data, struct diff_options *options)
1685 {
1686 int i, len, add, del, adds = 0, dels = 0;
1687 uintmax_t max_change = 0, max_len = 0;
1688 int total_files = data->nr, count;
1689 int width, name_width, graph_width, number_width = 0, bin_width = 0;
1690 const char *reset, *add_c, *del_c;
1691 const char *line_prefix = "";
1692 int extra_shown = 0;
1693
1694 if (data->nr == 0)
1695 return;
1696
1697 line_prefix = diff_line_prefix(options);
1698 count = options->stat_count ? options->stat_count : data->nr;
1699
1700 reset = diff_get_color_opt(options, DIFF_RESET);
1701 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1702 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1703
1704 /*
1705 * Find the longest filename and max number of changes
1706 */
1707 for (i = 0; (i < count) && (i < data->nr); i++) {
1708 struct diffstat_file *file = data->files[i];
1709 uintmax_t change = file->added + file->deleted;
1710
1711 if (!file->is_interesting && (change == 0)) {
1712 count++; /* not shown == room for one more */
1713 continue;
1714 }
1715 fill_print_name(file);
1716 len = strlen(file->print_name);
1717 if (max_len < len)
1718 max_len = len;
1719
1720 if (file->is_unmerged) {
1721 /* "Unmerged" is 8 characters */
1722 bin_width = bin_width < 8 ? 8 : bin_width;
1723 continue;
1724 }
1725 if (file->is_binary) {
1726 /* "Bin XXX -> YYY bytes" */
1727 int w = 14 + decimal_width(file->added)
1728 + decimal_width(file->deleted);
1729 bin_width = bin_width < w ? w : bin_width;
1730 /* Display change counts aligned with "Bin" */
1731 number_width = 3;
1732 continue;
1733 }
1734
1735 if (max_change < change)
1736 max_change = change;
1737 }
1738 count = i; /* where we can stop scanning in data->files[] */
1739
1740 /*
1741 * We have width = stat_width or term_columns() columns total.
1742 * We want a maximum of min(max_len, stat_name_width) for the name part.
1743 * We want a maximum of min(max_change, stat_graph_width) for the +- part.
1744 * We also need 1 for " " and 4 + decimal_width(max_change)
1745 * for " | NNNN " and one the empty column at the end, altogether
1746 * 6 + decimal_width(max_change).
1747 *
1748 * If there's not enough space, we will use the smaller of
1749 * stat_name_width (if set) and 5/8*width for the filename,
1750 * and the rest for constant elements + graph part, but no more
1751 * than stat_graph_width for the graph part.
1752 * (5/8 gives 50 for filename and 30 for the constant parts + graph
1753 * for the standard terminal size).
1754 *
1755 * In other words: stat_width limits the maximum width, and
1756 * stat_name_width fixes the maximum width of the filename,
1757 * and is also used to divide available columns if there
1758 * aren't enough.
1759 *
1760 * Binary files are displayed with "Bin XXX -> YYY bytes"
1761 * instead of the change count and graph. This part is treated
1762 * similarly to the graph part, except that it is not
1763 * "scaled". If total width is too small to accommodate the
1764 * guaranteed minimum width of the filename part and the
1765 * separators and this message, this message will "overflow"
1766 * making the line longer than the maximum width.
1767 */
1768
1769 if (options->stat_width == -1)
1770 width = term_columns() - strlen(line_prefix);
1771 else
1772 width = options->stat_width ? options->stat_width : 80;
1773 number_width = decimal_width(max_change) > number_width ?
1774 decimal_width(max_change) : number_width;
1775
1776 if (options->stat_graph_width == -1)
1777 options->stat_graph_width = diff_stat_graph_width;
1778
1779 /*
1780 * Guarantee 3/8*16==6 for the graph part
1781 * and 5/8*16==10 for the filename part
1782 */
1783 if (width < 16 + 6 + number_width)
1784 width = 16 + 6 + number_width;
1785
1786 /*
1787 * First assign sizes that are wanted, ignoring available width.
1788 * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
1789 * starting from "XXX" should fit in graph_width.
1790 */
1791 graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
1792 if (options->stat_graph_width &&
1793 options->stat_graph_width < graph_width)
1794 graph_width = options->stat_graph_width;
1795
1796 name_width = (options->stat_name_width > 0 &&
1797 options->stat_name_width < max_len) ?
1798 options->stat_name_width : max_len;
1799
1800 /*
1801 * Adjust adjustable widths not to exceed maximum width
1802 */
1803 if (name_width + number_width + 6 + graph_width > width) {
1804 if (graph_width > width * 3/8 - number_width - 6) {
1805 graph_width = width * 3/8 - number_width - 6;
1806 if (graph_width < 6)
1807 graph_width = 6;
1808 }
1809
1810 if (options->stat_graph_width &&
1811 graph_width > options->stat_graph_width)
1812 graph_width = options->stat_graph_width;
1813 if (name_width > width - number_width - 6 - graph_width)
1814 name_width = width - number_width - 6 - graph_width;
1815 else
1816 graph_width = width - number_width - 6 - name_width;
1817 }
1818
1819 /*
1820 * From here name_width is the width of the name area,
1821 * and graph_width is the width of the graph area.
1822 * max_change is used to scale graph properly.
1823 */
1824 for (i = 0; i < count; i++) {
1825 const char *prefix = "";
1826 struct diffstat_file *file = data->files[i];
1827 char *name = file->print_name;
1828 uintmax_t added = file->added;
1829 uintmax_t deleted = file->deleted;
1830 int name_len;
1831
1832 if (!file->is_interesting && (added + deleted == 0))
1833 continue;
1834
1835 /*
1836 * "scale" the filename
1837 */
1838 len = name_width;
1839 name_len = strlen(name);
1840 if (name_width < name_len) {
1841 char *slash;
1842 prefix = "...";
1843 len -= 3;
1844 name += name_len - len;
1845 slash = strchr(name, '/');
1846 if (slash)
1847 name = slash;
1848 }
1849
1850 if (file->is_binary) {
1851 fprintf(options->file, "%s", line_prefix);
1852 show_name(options->file, prefix, name, len);
1853 fprintf(options->file, " %*s", number_width, "Bin");
1854 if (!added && !deleted) {
1855 putc('\n', options->file);
1856 continue;
1857 }
1858 fprintf(options->file, " %s%"PRIuMAX"%s",
1859 del_c, deleted, reset);
1860 fprintf(options->file, " -> ");
1861 fprintf(options->file, "%s%"PRIuMAX"%s",
1862 add_c, added, reset);
1863 fprintf(options->file, " bytes");
1864 fprintf(options->file, "\n");
1865 continue;
1866 }
1867 else if (file->is_unmerged) {
1868 fprintf(options->file, "%s", line_prefix);
1869 show_name(options->file, prefix, name, len);
1870 fprintf(options->file, " Unmerged\n");
1871 continue;
1872 }
1873
1874 /*
1875 * scale the add/delete
1876 */
1877 add = added;
1878 del = deleted;
1879
1880 if (graph_width <= max_change) {
1881 int total = scale_linear(add + del, graph_width, max_change);
1882 if (total < 2 && add && del)
1883 /* width >= 2 due to the sanity check */
1884 total = 2;
1885 if (add < del) {
1886 add = scale_linear(add, graph_width, max_change);
1887 del = total - add;
1888 } else {
1889 del = scale_linear(del, graph_width, max_change);
1890 add = total - del;
1891 }
1892 }
1893 fprintf(options->file, "%s", line_prefix);
1894 show_name(options->file, prefix, name, len);
1895 fprintf(options->file, " %*"PRIuMAX"%s",
1896 number_width, added + deleted,
1897 added + deleted ? " " : "");
1898 show_graph(options->file, '+', add, add_c, reset);
1899 show_graph(options->file, '-', del, del_c, reset);
1900 fprintf(options->file, "\n");
1901 }
1902
1903 for (i = 0; i < data->nr; i++) {
1904 struct diffstat_file *file = data->files[i];
1905 uintmax_t added = file->added;
1906 uintmax_t deleted = file->deleted;
1907
1908 if (file->is_unmerged ||
1909 (!file->is_interesting && (added + deleted == 0))) {
1910 total_files--;
1911 continue;
1912 }
1913
1914 if (!file->is_binary) {
1915 adds += added;
1916 dels += deleted;
1917 }
1918 if (i < count)
1919 continue;
1920 if (!extra_shown)
1921 fprintf(options->file, "%s ...\n", line_prefix);
1922 extra_shown = 1;
1923 }
1924 fprintf(options->file, "%s", line_prefix);
1925 print_stat_summary(options->file, total_files, adds, dels);
1926 }
1927
1928 static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1929 {
1930 int i, adds = 0, dels = 0, total_files = data->nr;
1931
1932 if (data->nr == 0)
1933 return;
1934
1935 for (i = 0; i < data->nr; i++) {
1936 int added = data->files[i]->added;
1937 int deleted= data->files[i]->deleted;
1938
1939 if (data->files[i]->is_unmerged ||
1940 (!data->files[i]->is_interesting && (added + deleted == 0))) {
1941 total_files--;
1942 } else if (!data->files[i]->is_binary) { /* don't count bytes */
1943 adds += added;
1944 dels += deleted;
1945 }
1946 }
1947 fprintf(options->file, "%s", diff_line_prefix(options));
1948 print_stat_summary(options->file, total_files, adds, dels);
1949 }
1950
1951 static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1952 {
1953 int i;
1954
1955 if (data->nr == 0)
1956 return;
1957
1958 for (i = 0; i < data->nr; i++) {
1959 struct diffstat_file *file = data->files[i];
1960
1961 fprintf(options->file, "%s", diff_line_prefix(options));
1962
1963 if (file->is_binary)
1964 fprintf(options->file, "-\t-\t");
1965 else
1966 fprintf(options->file,
1967 "%"PRIuMAX"\t%"PRIuMAX"\t",
1968 file->added, file->deleted);
1969 if (options->line_termination) {
1970 fill_print_name(file);
1971 if (!file->is_renamed)
1972 write_name_quoted(file->name, options->file,
1973 options->line_termination);
1974 else {
1975 fputs(file->print_name, options->file);
1976 putc(options->line_termination, options->file);
1977 }
1978 } else {
1979 if (file->is_renamed) {
1980 putc('\0', options->file);
1981 write_name_quoted(file->from_name, options->file, '\0');
1982 }
1983 write_name_quoted(file->name, options->file, '\0');
1984 }
1985 }
1986 }
1987
1988 struct dirstat_file {
1989 const char *name;
1990 unsigned long changed;
1991 };
1992
1993 struct dirstat_dir {
1994 struct dirstat_file *files;
1995 int alloc, nr, permille, cumulative;
1996 };
1997
1998 static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1999 unsigned long changed, const char *base, int baselen)
2000 {
2001 unsigned long this_dir = 0;
2002 unsigned int sources = 0;
2003 const char *line_prefix = diff_line_prefix(opt);
2004
2005 while (dir->nr) {
2006 struct dirstat_file *f = dir->files;
2007 int namelen = strlen(f->name);
2008 unsigned long this;
2009 char *slash;
2010
2011 if (namelen < baselen)
2012 break;
2013 if (memcmp(f->name, base, baselen))
2014 break;
2015 slash = strchr(f->name + baselen, '/');
2016 if (slash) {
2017 int newbaselen = slash + 1 - f->name;
2018 this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
2019 sources++;
2020 } else {
2021 this = f->changed;
2022 dir->files++;
2023 dir->nr--;
2024 sources += 2;
2025 }
2026 this_dir += this;
2027 }
2028
2029 /*
2030 * We don't report dirstat's for
2031 * - the top level
2032 * - or cases where everything came from a single directory
2033 * under this directory (sources == 1).
2034 */
2035 if (baselen && sources != 1) {
2036 if (this_dir) {
2037 int permille = this_dir * 1000 / changed;
2038 if (permille >= dir->permille) {
2039 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
2040 permille / 10, permille % 10, baselen, base);
2041 if (!dir->cumulative)
2042 return 0;
2043 }
2044 }
2045 }
2046 return this_dir;
2047 }
2048
2049 static int dirstat_compare(const void *_a, const void *_b)
2050 {
2051 const struct dirstat_file *a = _a;
2052 const struct dirstat_file *b = _b;
2053 return strcmp(a->name, b->name);
2054 }
2055
2056 static void show_dirstat(struct diff_options *options)
2057 {
2058 int i;
2059 unsigned long changed;
2060 struct dirstat_dir dir;
2061 struct diff_queue_struct *q = &diff_queued_diff;
2062
2063 dir.files = NULL;
2064 dir.alloc = 0;
2065 dir.nr = 0;
2066 dir.permille = options->dirstat_permille;
2067 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
2068
2069 changed = 0;
2070 for (i = 0; i < q->nr; i++) {
2071 struct diff_filepair *p = q->queue[i];
2072 const char *name;
2073 unsigned long copied, added, damage;
2074 int content_changed;
2075
2076 name = p->two->path ? p->two->path : p->one->path;
2077
2078 if (p->one->oid_valid && p->two->oid_valid)
2079 content_changed = oidcmp(&p->one->oid, &p->two->oid);
2080 else
2081 content_changed = 1;
2082
2083 if (!content_changed) {
2084 /*
2085 * The SHA1 has not changed, so pre-/post-content is
2086 * identical. We can therefore skip looking at the
2087 * file contents altogether.
2088 */
2089 damage = 0;
2090 goto found_damage;
2091 }
2092
2093 if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
2094 /*
2095 * In --dirstat-by-file mode, we don't really need to
2096 * look at the actual file contents at all.
2097 * The fact that the SHA1 changed is enough for us to
2098 * add this file to the list of results
2099 * (with each file contributing equal damage).
2100 */
2101 damage = 1;
2102 goto found_damage;
2103 }
2104
2105 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
2106 diff_populate_filespec(p->one, 0);
2107 diff_populate_filespec(p->two, 0);
2108 diffcore_count_changes(p->one, p->two, NULL, NULL,
2109 &copied, &added);
2110 diff_free_filespec_data(p->one);
2111 diff_free_filespec_data(p->two);
2112 } else if (DIFF_FILE_VALID(p->one)) {
2113 diff_populate_filespec(p->one, CHECK_SIZE_ONLY);
2114 copied = added = 0;
2115 diff_free_filespec_data(p->one);
2116 } else if (DIFF_FILE_VALID(p->two)) {
2117 diff_populate_filespec(p->two, CHECK_SIZE_ONLY);
2118 copied = 0;
2119 added = p->two->size;
2120 diff_free_filespec_data(p->two);
2121 } else
2122 continue;
2123
2124 /*
2125 * Original minus copied is the removed material,
2126 * added is the new material. They are both damages
2127 * made to the preimage.
2128 * If the resulting damage is zero, we know that
2129 * diffcore_count_changes() considers the two entries to
2130 * be identical, but since content_changed is true, we
2131 * know that there must have been _some_ kind of change,
2132 * so we force all entries to have damage > 0.
2133 */
2134 damage = (p->one->size - copied) + added;
2135 if (!damage)
2136 damage = 1;
2137
2138 found_damage:
2139 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2140 dir.files[dir.nr].name = name;
2141 dir.files[dir.nr].changed = damage;
2142 changed += damage;
2143 dir.nr++;
2144 }
2145
2146 /* This can happen even with many files, if everything was renames */
2147 if (!changed)
2148 return;
2149
2150 /* Show all directories with more than x% of the changes */
2151 QSORT(dir.files, dir.nr, dirstat_compare);
2152 gather_dirstat(options, &dir, changed, "", 0);
2153 }
2154
2155 static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
2156 {
2157 int i;
2158 unsigned long changed;
2159 struct dirstat_dir dir;
2160
2161 if (data->nr == 0)
2162 return;
2163
2164 dir.files = NULL;
2165 dir.alloc = 0;
2166 dir.nr = 0;
2167 dir.permille = options->dirstat_permille;
2168 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
2169
2170 changed = 0;
2171 for (i = 0; i < data->nr; i++) {
2172 struct diffstat_file *file = data->files[i];
2173 unsigned long damage = file->added + file->deleted;
2174 if (file->is_binary)
2175 /*
2176 * binary files counts bytes, not lines. Must find some
2177 * way to normalize binary bytes vs. textual lines.
2178 * The following heuristic assumes that there are 64
2179 * bytes per "line".
2180 * This is stupid and ugly, but very cheap...
2181 */
2182 damage = (damage + 63) / 64;
2183 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2184 dir.files[dir.nr].name = file->name;
2185 dir.files[dir.nr].changed = damage;
2186 changed += damage;
2187 dir.nr++;
2188 }
2189
2190 /* This can happen even with many files, if everything was renames */
2191 if (!changed)
2192 return;
2193
2194 /* Show all directories with more than x% of the changes */
2195 QSORT(dir.files, dir.nr, dirstat_compare);
2196 gather_dirstat(options, &dir, changed, "", 0);
2197 }
2198
2199 static void free_diffstat_info(struct diffstat_t *diffstat)
2200 {
2201 int i;
2202 for (i = 0; i < diffstat->nr; i++) {
2203 struct diffstat_file *f = diffstat->files[i];
2204 if (f->name != f->print_name)
2205 free(f->print_name);
2206 free(f->name);
2207 free(f->from_name);
2208 free(f);
2209 }
2210 free(diffstat->files);
2211 }
2212
2213 struct checkdiff_t {
2214 const char *filename;
2215 int lineno;
2216 int conflict_marker_size;
2217 struct diff_options *o;
2218 unsigned ws_rule;
2219 unsigned status;
2220 };
2221
2222 static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2223 {
2224 char firstchar;
2225 int cnt;
2226
2227 if (len < marker_size + 1)
2228 return 0;
2229 firstchar = line[0];
2230 switch (firstchar) {
2231 case '=': case '>': case '<': case '|':
2232 break;
2233 default:
2234 return 0;
2235 }
2236 for (cnt = 1; cnt < marker_size; cnt++)
2237 if (line[cnt] != firstchar)
2238 return 0;
2239 /* line[1] thru line[marker_size-1] are same as firstchar */
2240 if (len < marker_size + 1 || !isspace(line[marker_size]))
2241 return 0;
2242 return 1;
2243 }
2244
2245 static void checkdiff_consume(void *priv, char *line, unsigned long len)
2246 {
2247 struct checkdiff_t *data = priv;
2248 int marker_size = data->conflict_marker_size;
2249 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2250 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2251 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2252 char *err;
2253 const char *line_prefix;
2254
2255 assert(data->o);
2256 line_prefix = diff_line_prefix(data->o);
2257
2258 if (line[0] == '+') {
2259 unsigned bad;
2260 data->lineno++;
2261 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2262 data->status |= 1;
2263 fprintf(data->o->file,
2264 "%s%s:%d: leftover conflict marker\n",
2265 line_prefix, data->filename, data->lineno);
2266 }
2267 bad = ws_check(line + 1, len - 1, data->ws_rule);
2268 if (!bad)
2269 return;
2270 data->status |= bad;
2271 err = whitespace_error_string(bad);
2272 fprintf(data->o->file, "%s%s:%d: %s.\n",
2273 line_prefix, data->filename, data->lineno, err);
2274 free(err);
2275 emit_line(data->o, set, reset, line, 1);
2276 ws_check_emit(line + 1, len - 1, data->ws_rule,
2277 data->o->file, set, reset, ws);
2278 } else if (line[0] == ' ') {
2279 data->lineno++;
2280 } else if (line[0] == '@') {
2281 char *plus = strchr(line, '+');
2282 if (plus)
2283 data->lineno = strtol(plus, NULL, 10) - 1;
2284 else
2285 die("invalid diff");
2286 }
2287 }
2288
2289 static unsigned char *deflate_it(char *data,
2290 unsigned long size,
2291 unsigned long *result_size)
2292 {
2293 int bound;
2294 unsigned char *deflated;
2295 git_zstream stream;
2296
2297 git_deflate_init(&stream, zlib_compression_level);
2298 bound = git_deflate_bound(&stream, size);
2299 deflated = xmalloc(bound);
2300 stream.next_out = deflated;
2301 stream.avail_out = bound;
2302
2303 stream.next_in = (unsigned char *)data;
2304 stream.avail_in = size;
2305 while (git_deflate(&stream, Z_FINISH) == Z_OK)
2306 ; /* nothing */
2307 git_deflate_end(&stream);
2308 *result_size = stream.total_out;
2309 return deflated;
2310 }
2311
2312 static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two,
2313 const char *prefix)
2314 {
2315 void *cp;
2316 void *delta;
2317 void *deflated;
2318 void *data;
2319 unsigned long orig_size;
2320 unsigned long delta_size;
2321 unsigned long deflate_size;
2322 unsigned long data_size;
2323
2324 /* We could do deflated delta, or we could do just deflated two,
2325 * whichever is smaller.
2326 */
2327 delta = NULL;
2328 deflated = deflate_it(two->ptr, two->size, &deflate_size);
2329 if (one->size && two->size) {
2330 delta = diff_delta(one->ptr, one->size,
2331 two->ptr, two->size,
2332 &delta_size, deflate_size);
2333 if (delta) {
2334 void *to_free = delta;
2335 orig_size = delta_size;
2336 delta = deflate_it(delta, delta_size, &delta_size);
2337 free(to_free);
2338 }
2339 }
2340
2341 if (delta && delta_size < deflate_size) {
2342 fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2343 free(deflated);
2344 data = delta;
2345 data_size = delta_size;
2346 }
2347 else {
2348 fprintf(file, "%sliteral %lu\n", prefix, two->size);
2349 free(delta);
2350 data = deflated;
2351 data_size = deflate_size;
2352 }
2353
2354 /* emit data encoded in base85 */
2355 cp = data;
2356 while (data_size) {
2357 int bytes = (52 < data_size) ? 52 : data_size;
2358 char line[70];
2359 data_size -= bytes;
2360 if (bytes <= 26)
2361 line[0] = bytes + 'A' - 1;
2362 else
2363 line[0] = bytes - 26 + 'a' - 1;
2364 encode_85(line + 1, cp, bytes);
2365 cp = (char *) cp + bytes;
2366 fprintf(file, "%s", prefix);
2367 fputs(line, file);
2368 fputc('\n', file);
2369 }
2370 fprintf(file, "%s\n", prefix);
2371 free(data);
2372 }
2373
2374 static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two,
2375 const char *prefix)
2376 {
2377 fprintf(file, "%sGIT binary patch\n", prefix);
2378 emit_binary_diff_body(file, one, two, prefix);
2379 emit_binary_diff_body(file, two, one, prefix);
2380 }
2381
2382 int diff_filespec_is_binary(struct diff_filespec *one)
2383 {
2384 if (one->is_binary == -1) {
2385 diff_filespec_load_driver(one);
2386 if (one->driver->binary != -1)
2387 one->is_binary = one->driver->binary;
2388 else {
2389 if (!one->data && DIFF_FILE_VALID(one))
2390 diff_populate_filespec(one, CHECK_BINARY);
2391 if (one->is_binary == -1 && one->data)
2392 one->is_binary = buffer_is_binary(one->data,
2393 one->size);
2394 if (one->is_binary == -1)
2395 one->is_binary = 0;
2396 }
2397 }
2398 return one->is_binary;
2399 }
2400
2401 static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2402 {
2403 diff_filespec_load_driver(one);
2404 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2405 }
2406
2407 void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2408 {
2409 if (!options->a_prefix)
2410 options->a_prefix = a;
2411 if (!options->b_prefix)
2412 options->b_prefix = b;
2413 }
2414
2415 struct userdiff_driver *get_textconv(struct diff_filespec *one)
2416 {
2417 if (!DIFF_FILE_VALID(one))
2418 return NULL;
2419
2420 diff_filespec_load_driver(one);
2421 return userdiff_get_textconv(one->driver);
2422 }
2423
2424 static void builtin_diff(const char *name_a,
2425 const char *name_b,
2426 struct diff_filespec *one,
2427 struct diff_filespec *two,
2428 const char *xfrm_msg,
2429 int must_show_header,
2430 struct diff_options *o,
2431 int complete_rewrite)
2432 {
2433 mmfile_t mf1, mf2;
2434 const char *lbl[2];
2435 char *a_one, *b_two;
2436 const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
2437 const char *reset = diff_get_color_opt(o, DIFF_RESET);
2438 const char *a_prefix, *b_prefix;
2439 struct userdiff_driver *textconv_one = NULL;
2440 struct userdiff_driver *textconv_two = NULL;
2441 struct strbuf header = STRBUF_INIT;
2442 const char *line_prefix = diff_line_prefix(o);
2443
2444 diff_set_mnemonic_prefix(o, "a/", "b/");
2445 if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2446 a_prefix = o->b_prefix;
2447 b_prefix = o->a_prefix;
2448 } else {
2449 a_prefix = o->a_prefix;
2450 b_prefix = o->b_prefix;
2451 }
2452
2453 if (o->submodule_format == DIFF_SUBMODULE_LOG &&
2454 (!one->mode || S_ISGITLINK(one->mode)) &&
2455 (!two->mode || S_ISGITLINK(two->mode))) {
2456 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2457 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2458 show_submodule_summary(o->file, one->path ? one->path : two->path,
2459 line_prefix,
2460 &one->oid, &two->oid,
2461 two->dirty_submodule,
2462 meta, del, add, reset);
2463 return;
2464 } else if (o->submodule_format == DIFF_SUBMODULE_INLINE_DIFF &&
2465 (!one->mode || S_ISGITLINK(one->mode)) &&
2466 (!two->mode || S_ISGITLINK(two->mode))) {
2467 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2468 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2469 show_submodule_inline_diff(o->file, one->path ? one->path : two->path,
2470 line_prefix,
2471 &one->oid, &two->oid,
2472 two->dirty_submodule,
2473 meta, del, add, reset, o);
2474 return;
2475 }
2476
2477 if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2478 textconv_one = get_textconv(one);
2479 textconv_two = get_textconv(two);
2480 }
2481
2482 /* Never use a non-valid filename anywhere if at all possible */
2483 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2484 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2485
2486 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2487 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2488 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2489 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2490 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
2491 if (lbl[0][0] == '/') {
2492 /* /dev/null */
2493 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
2494 if (xfrm_msg)
2495 strbuf_addstr(&header, xfrm_msg);
2496 must_show_header = 1;
2497 }
2498 else if (lbl[1][0] == '/') {
2499 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
2500 if (xfrm_msg)
2501 strbuf_addstr(&header, xfrm_msg);
2502 must_show_header = 1;
2503 }
2504 else {
2505 if (one->mode != two->mode) {
2506 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
2507 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
2508 must_show_header = 1;
2509 }
2510 if (xfrm_msg)
2511 strbuf_addstr(&header, xfrm_msg);
2512
2513 /*
2514 * we do not run diff between different kind
2515 * of objects.
2516 */
2517 if ((one->mode ^ two->mode) & S_IFMT)
2518 goto free_ab_and_return;
2519 if (complete_rewrite &&
2520 (textconv_one || !diff_filespec_is_binary(one)) &&
2521 (textconv_two || !diff_filespec_is_binary(two))) {
2522 fprintf(o->file, "%s", header.buf);
2523 strbuf_reset(&header);
2524 emit_rewrite_diff(name_a, name_b, one, two,
2525 textconv_one, textconv_two, o);
2526 o->found_changes = 1;
2527 goto free_ab_and_return;
2528 }
2529 }
2530
2531 if (o->irreversible_delete && lbl[1][0] == '/') {
2532 fprintf(o->file, "%s", header.buf);
2533 strbuf_reset(&header);
2534 goto free_ab_and_return;
2535 } else if (!DIFF_OPT_TST(o, TEXT) &&
2536 ( (!textconv_one && diff_filespec_is_binary(one)) ||
2537 (!textconv_two && diff_filespec_is_binary(two)) )) {
2538 if (!one->data && !two->data &&
2539 S_ISREG(one->mode) && S_ISREG(two->mode) &&
2540 !DIFF_OPT_TST(o, BINARY)) {
2541 if (!oidcmp(&one->oid, &two->oid)) {
2542 if (must_show_header)
2543 fprintf(o->file, "%s", header.buf);
2544 goto free_ab_and_return;
2545 }
2546 fprintf(o->file, "%s", header.buf);
2547 fprintf(o->file, "%sBinary files %s and %s differ\n",
2548 line_prefix, lbl[0], lbl[1]);
2549 goto free_ab_and_return;
2550 }
2551 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2552 die("unable to read files to diff");
2553 /* Quite common confusing case */
2554 if (mf1.size == mf2.size &&
2555 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2556 if (must_show_header)
2557 fprintf(o->file, "%s", header.buf);
2558 goto free_ab_and_return;
2559 }
2560 fprintf(o->file, "%s", header.buf);
2561 strbuf_reset(&header);
2562 if (DIFF_OPT_TST(o, BINARY))
2563 emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2564 else
2565 fprintf(o->file, "%sBinary files %s and %s differ\n",
2566 line_prefix, lbl[0], lbl[1]);
2567 o->found_changes = 1;
2568 } else {
2569 /* Crazy xdl interfaces.. */
2570 const char *diffopts = getenv("GIT_DIFF_OPTS");
2571 const char *v;
2572 xpparam_t xpp;
2573 xdemitconf_t xecfg;
2574 struct emit_callback ecbdata;
2575 const struct userdiff_funcname *pe;
2576
2577 if (must_show_header) {
2578 fprintf(o->file, "%s", header.buf);
2579 strbuf_reset(&header);
2580 }
2581
2582 mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2583 mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2584
2585 pe = diff_funcname_pattern(one);
2586 if (!pe)
2587 pe = diff_funcname_pattern(two);
2588
2589 memset(&xpp, 0, sizeof(xpp));
2590 memset(&xecfg, 0, sizeof(xecfg));
2591 memset(&ecbdata, 0, sizeof(ecbdata));
2592 ecbdata.label_path = lbl;
2593 ecbdata.color_diff = want_color(o->use_color);
2594 ecbdata.ws_rule = whitespace_rule(name_b);
2595 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2596 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2597 ecbdata.opt = o;
2598 ecbdata.header = header.len ? &header : NULL;
2599 xpp.flags = o->xdl_opts;
2600 xecfg.ctxlen = o->context;
2601 xecfg.interhunkctxlen = o->interhunkcontext;
2602 xecfg.flags = XDL_EMIT_FUNCNAMES;
2603 if (DIFF_OPT_TST(o, FUNCCONTEXT))
2604 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2605 if (pe)
2606 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2607 if (!diffopts)
2608 ;
2609 else if (skip_prefix(diffopts, "--unified=", &v))
2610 xecfg.ctxlen = strtoul(v, NULL, 10);
2611 else if (skip_prefix(diffopts, "-u", &v))
2612 xecfg.ctxlen = strtoul(v, NULL, 10);
2613 if (o->word_diff)
2614 init_diff_words_data(&ecbdata, o, one, two);
2615 if (xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2616 &xpp, &xecfg))
2617 die("unable to generate diff for %s", one->path);
2618 if (o->word_diff)
2619 free_diff_words_data(&ecbdata);
2620 if (textconv_one)
2621 free(mf1.ptr);
2622 if (textconv_two)
2623 free(mf2.ptr);
2624 xdiff_clear_find_func(&xecfg);
2625 }
2626
2627 free_ab_and_return:
2628 strbuf_release(&header);
2629 diff_free_filespec_data(one);
2630 diff_free_filespec_data(two);
2631 free(a_one);
2632 free(b_two);
2633 return;
2634 }
2635
2636 static void builtin_diffstat(const char *name_a, const char *name_b,
2637 struct diff_filespec *one,
2638 struct diff_filespec *two,
2639 struct diffstat_t *diffstat,
2640 struct diff_options *o,
2641 struct diff_filepair *p)
2642 {
2643 mmfile_t mf1, mf2;
2644 struct diffstat_file *data;
2645 int same_contents;
2646 int complete_rewrite = 0;
2647
2648 if (!DIFF_PAIR_UNMERGED(p)) {
2649 if (p->status == DIFF_STATUS_MODIFIED && p->score)
2650 complete_rewrite = 1;
2651 }
2652
2653 data = diffstat_add(diffstat, name_a, name_b);
2654 data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
2655
2656 if (!one || !two) {
2657 data->is_unmerged = 1;
2658 return;
2659 }
2660
2661 same_contents = !oidcmp(&one->oid, &two->oid);
2662
2663 if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2664 data->is_binary = 1;
2665 if (same_contents) {
2666 data->added = 0;
2667 data->deleted = 0;
2668 } else {
2669 data->added = diff_filespec_size(two);
2670 data->deleted = diff_filespec_size(one);
2671 }
2672 }
2673
2674 else if (complete_rewrite) {
2675 diff_populate_filespec(one, 0);
2676 diff_populate_filespec(two, 0);
2677 data->deleted = count_lines(one->data, one->size);
2678 data->added = count_lines(two->data, two->size);
2679 }
2680
2681 else if (!same_contents) {
2682 /* Crazy xdl interfaces.. */
2683 xpparam_t xpp;
2684 xdemitconf_t xecfg;
2685
2686 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2687 die("unable to read files to diff");
2688
2689 memset(&xpp, 0, sizeof(xpp));
2690 memset(&xecfg, 0, sizeof(xecfg));
2691 xpp.flags = o->xdl_opts;
2692 xecfg.ctxlen = o->context;
2693 xecfg.interhunkctxlen = o->interhunkcontext;
2694 if (xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2695 &xpp, &xecfg))
2696 die("unable to generate diffstat for %s", one->path);
2697 }
2698
2699 diff_free_filespec_data(one);
2700 diff_free_filespec_data(two);
2701 }
2702
2703 static void builtin_checkdiff(const char *name_a, const char *name_b,
2704 const char *attr_path,
2705 struct diff_filespec *one,
2706 struct diff_filespec *two,
2707 struct diff_options *o)
2708 {
2709 mmfile_t mf1, mf2;
2710 struct checkdiff_t data;
2711
2712 if (!two)
2713 return;
2714
2715 memset(&data, 0, sizeof(data));
2716 data.filename = name_b ? name_b : name_a;
2717 data.lineno = 0;
2718 data.o = o;
2719 data.ws_rule = whitespace_rule(attr_path);
2720 data.conflict_marker_size = ll_merge_marker_size(attr_path);
2721
2722 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2723 die("unable to read files to diff");
2724
2725 /*
2726 * All the other codepaths check both sides, but not checking
2727 * the "old" side here is deliberate. We are checking the newly
2728 * introduced changes, and as long as the "new" side is text, we
2729 * can and should check what it introduces.
2730 */
2731 if (diff_filespec_is_binary(two))
2732 goto free_and_return;
2733 else {
2734 /* Crazy xdl interfaces.. */
2735 xpparam_t xpp;
2736 xdemitconf_t xecfg;
2737
2738 memset(&xpp, 0, sizeof(xpp));
2739 memset(&xecfg, 0, sizeof(xecfg));
2740 xecfg.ctxlen = 1; /* at least one context line */
2741 xpp.flags = 0;
2742 if (xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2743 &xpp, &xecfg))
2744 die("unable to generate checkdiff for %s", one->path);
2745
2746 if (data.ws_rule & WS_BLANK_AT_EOF) {
2747 struct emit_callback ecbdata;
2748 int blank_at_eof;
2749
2750 ecbdata.ws_rule = data.ws_rule;
2751 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2752 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2753
2754 if (blank_at_eof) {
2755 static char *err;
2756 if (!err)
2757 err = whitespace_error_string(WS_BLANK_AT_EOF);
2758 fprintf(o->file, "%s:%d: %s.\n",
2759 data.filename, blank_at_eof, err);
2760 data.status = 1; /* report errors */
2761 }
2762 }
2763 }
2764 free_and_return:
2765 diff_free_filespec_data(one);
2766 diff_free_filespec_data(two);
2767 if (data.status)
2768 DIFF_OPT_SET(o, CHECK_FAILED);
2769 }
2770
2771 struct diff_filespec *alloc_filespec(const char *path)
2772 {
2773 struct diff_filespec *spec;
2774
2775 FLEXPTR_ALLOC_STR(spec, path, path);
2776 spec->count = 1;
2777 spec->is_binary = -1;
2778 return spec;
2779 }
2780
2781 void free_filespec(struct diff_filespec *spec)
2782 {
2783 if (!--spec->count) {
2784 diff_free_filespec_data(spec);
2785 free(spec);
2786 }
2787 }
2788
2789 void fill_filespec(struct diff_filespec *spec, const struct object_id *oid,
2790 int oid_valid, unsigned short mode)
2791 {
2792 if (mode) {
2793 spec->mode = canon_mode(mode);
2794 oidcpy(&spec->oid, oid);
2795 spec->oid_valid = oid_valid;
2796 }
2797 }
2798
2799 /*
2800 * Given a name and sha1 pair, if the index tells us the file in
2801 * the work tree has that object contents, return true, so that
2802 * prepare_temp_file() does not have to inflate and extract.
2803 */
2804 static int reuse_worktree_file(const char *name, const struct object_id *oid, int want_file)
2805 {
2806 const struct cache_entry *ce;
2807 struct stat st;
2808 int pos, len;
2809
2810 /*
2811 * We do not read the cache ourselves here, because the
2812 * benchmark with my previous version that always reads cache
2813 * shows that it makes things worse for diff-tree comparing
2814 * two linux-2.6 kernel trees in an already checked out work
2815 * tree. This is because most diff-tree comparisons deal with
2816 * only a small number of files, while reading the cache is
2817 * expensive for a large project, and its cost outweighs the
2818 * savings we get by not inflating the object to a temporary
2819 * file. Practically, this code only helps when we are used
2820 * by diff-cache --cached, which does read the cache before
2821 * calling us.
2822 */
2823 if (!active_cache)
2824 return 0;
2825
2826 /* We want to avoid the working directory if our caller
2827 * doesn't need the data in a normal file, this system
2828 * is rather slow with its stat/open/mmap/close syscalls,
2829 * and the object is contained in a pack file. The pack
2830 * is probably already open and will be faster to obtain
2831 * the data through than the working directory. Loose
2832 * objects however would tend to be slower as they need
2833 * to be individually opened and inflated.
2834 */
2835 if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(oid->hash))
2836 return 0;
2837
2838 /*
2839 * Similarly, if we'd have to convert the file contents anyway, that
2840 * makes the optimization not worthwhile.
2841 */
2842 if (!want_file && would_convert_to_git(&the_index, name))
2843 return 0;
2844
2845 len = strlen(name);
2846 pos = cache_name_pos(name, len);
2847 if (pos < 0)
2848 return 0;
2849 ce = active_cache[pos];
2850
2851 /*
2852 * This is not the sha1 we are looking for, or
2853 * unreusable because it is not a regular file.
2854 */
2855 if (oidcmp(oid, &ce->oid) || !S_ISREG(ce->ce_mode))
2856 return 0;
2857
2858 /*
2859 * If ce is marked as "assume unchanged", there is no
2860 * guarantee that work tree matches what we are looking for.
2861 */
2862 if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2863 return 0;
2864
2865 /*
2866 * If ce matches the file in the work tree, we can reuse it.
2867 */
2868 if (ce_uptodate(ce) ||
2869 (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2870 return 1;
2871
2872 return 0;
2873 }
2874
2875 static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2876 {
2877 struct strbuf buf = STRBUF_INIT;
2878 char *dirty = "";
2879
2880 /* Are we looking at the work tree? */
2881 if (s->dirty_submodule)
2882 dirty = "-dirty";
2883
2884 strbuf_addf(&buf, "Subproject commit %s%s\n",
2885 oid_to_hex(&s->oid), dirty);
2886 s->size = buf.len;
2887 if (size_only) {
2888 s->data = NULL;
2889 strbuf_release(&buf);
2890 } else {
2891 s->data = strbuf_detach(&buf, NULL);
2892 s->should_free = 1;
2893 }
2894 return 0;
2895 }
2896
2897 /*
2898 * While doing rename detection and pickaxe operation, we may need to
2899 * grab the data for the blob (or file) for our own in-core comparison.
2900 * diff_filespec has data and size fields for this purpose.
2901 */
2902 int diff_populate_filespec(struct diff_filespec *s, unsigned int flags)
2903 {
2904 int size_only = flags & CHECK_SIZE_ONLY;
2905 int err = 0;
2906 /*
2907 * demote FAIL to WARN to allow inspecting the situation
2908 * instead of refusing.
2909 */
2910 enum safe_crlf crlf_warn = (safe_crlf == SAFE_CRLF_FAIL
2911 ? SAFE_CRLF_WARN
2912 : safe_crlf);
2913
2914 if (!DIFF_FILE_VALID(s))
2915 die("internal error: asking to populate invalid file.");
2916 if (S_ISDIR(s->mode))
2917 return -1;
2918
2919 if (s->data)
2920 return 0;
2921
2922 if (size_only && 0 < s->size)
2923 return 0;
2924
2925 if (S_ISGITLINK(s->mode))
2926 return diff_populate_gitlink(s, size_only);
2927
2928 if (!s->oid_valid ||
2929 reuse_worktree_file(s->path, &s->oid, 0)) {
2930 struct strbuf buf = STRBUF_INIT;
2931 struct stat st;
2932 int fd;
2933
2934 if (lstat(s->path, &st) < 0) {
2935 if (errno == ENOENT) {
2936 err_empty:
2937 err = -1;
2938 empty:
2939 s->data = (char *)"";
2940 s->size = 0;
2941 return err;
2942 }
2943 }
2944 s->size = xsize_t(st.st_size);
2945 if (!s->size)
2946 goto empty;
2947 if (S_ISLNK(st.st_mode)) {
2948 struct strbuf sb = STRBUF_INIT;
2949
2950 if (strbuf_readlink(&sb, s->path, s->size))
2951 goto err_empty;
2952 s->size = sb.len;
2953 s->data = strbuf_detach(&sb, NULL);
2954 s->should_free = 1;
2955 return 0;
2956 }
2957
2958 /*
2959 * Even if the caller would be happy with getting
2960 * only the size, we cannot return early at this
2961 * point if the path requires us to run the content
2962 * conversion.
2963 */
2964 if (size_only && !would_convert_to_git(&the_index, s->path))
2965 return 0;
2966
2967 /*
2968 * Note: this check uses xsize_t(st.st_size) that may
2969 * not be the true size of the blob after it goes
2970 * through convert_to_git(). This may not strictly be
2971 * correct, but the whole point of big_file_threshold
2972 * and is_binary check being that we want to avoid
2973 * opening the file and inspecting the contents, this
2974 * is probably fine.
2975 */
2976 if ((flags & CHECK_BINARY) &&
2977 s->size > big_file_threshold && s->is_binary == -1) {
2978 s->is_binary = 1;
2979 return 0;
2980 }
2981 fd = open(s->path, O_RDONLY);
2982 if (fd < 0)
2983 goto err_empty;
2984 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2985 close(fd);
2986 s->should_munmap = 1;
2987
2988 /*
2989 * Convert from working tree format to canonical git format
2990 */
2991 if (convert_to_git(&the_index, s->path, s->data, s->size, &buf, crlf_warn)) {
2992 size_t size = 0;
2993 munmap(s->data, s->size);
2994 s->should_munmap = 0;
2995 s->data = strbuf_detach(&buf, &size);
2996 s->size = size;
2997 s->should_free = 1;
2998 }
2999 }
3000 else {
3001 enum object_type type;
3002 if (size_only || (flags & CHECK_BINARY)) {
3003 type = sha1_object_info(s->oid.hash, &s->size);
3004 if (type < 0)
3005 die("unable to read %s",
3006 oid_to_hex(&s->oid));
3007 if (size_only)
3008 return 0;
3009 if (s->size > big_file_threshold && s->is_binary == -1) {
3010 s->is_binary = 1;
3011 return 0;
3012 }
3013 }
3014 s->data = read_sha1_file(s->oid.hash, &type, &s->size);
3015 if (!s->data)
3016 die("unable to read %s", oid_to_hex(&s->oid));
3017 s->should_free = 1;
3018 }
3019 return 0;
3020 }
3021
3022 void diff_free_filespec_blob(struct diff_filespec *s)
3023 {
3024 if (s->should_free)
3025 free(s->data);
3026 else if (s->should_munmap)
3027 munmap(s->data, s->size);
3028
3029 if (s->should_free || s->should_munmap) {
3030 s->should_free = s->should_munmap = 0;
3031 s->data = NULL;
3032 }
3033 }
3034
3035 void diff_free_filespec_data(struct diff_filespec *s)
3036 {
3037 diff_free_filespec_blob(s);
3038 FREE_AND_NULL(s->cnt_data);
3039 }
3040
3041 static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
3042 void *blob,
3043 unsigned long size,
3044 const struct object_id *oid,
3045 int mode)
3046 {
3047 int fd;
3048 struct strbuf buf = STRBUF_INIT;
3049 struct strbuf template = STRBUF_INIT;
3050 char *path_dup = xstrdup(path);
3051 const char *base = basename(path_dup);
3052
3053 /* Generate "XXXXXX_basename.ext" */
3054 strbuf_addstr(&template, "XXXXXX_");
3055 strbuf_addstr(&template, base);
3056
3057 fd = mks_tempfile_ts(&temp->tempfile, template.buf, strlen(base) + 1);
3058 if (fd < 0)
3059 die_errno("unable to create temp-file");
3060 if (convert_to_working_tree(path,
3061 (const char *)blob, (size_t)size, &buf)) {
3062 blob = buf.buf;
3063 size = buf.len;
3064 }
3065 if (write_in_full(fd, blob, size) != size)
3066 die_errno("unable to write temp-file");
3067 close_tempfile(&temp->tempfile);
3068 temp->name = get_tempfile_path(&temp->tempfile);
3069 oid_to_hex_r(temp->hex, oid);
3070 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", mode);
3071 strbuf_release(&buf);
3072 strbuf_release(&template);
3073 free(path_dup);
3074 }
3075
3076 static struct diff_tempfile *prepare_temp_file(const char *name,
3077 struct diff_filespec *one)
3078 {
3079 struct diff_tempfile *temp = claim_diff_tempfile();
3080
3081 if (!DIFF_FILE_VALID(one)) {
3082 not_a_valid_file:
3083 /* A '-' entry produces this for file-2, and
3084 * a '+' entry produces this for file-1.
3085 */
3086 temp->name = "/dev/null";
3087 xsnprintf(temp->hex, sizeof(temp->hex), ".");
3088 xsnprintf(temp->mode, sizeof(temp->mode), ".");
3089 return temp;
3090 }
3091
3092 if (!S_ISGITLINK(one->mode) &&
3093 (!one->oid_valid ||
3094 reuse_worktree_file(name, &one->oid, 1))) {
3095 struct stat st;
3096 if (lstat(name, &st) < 0) {
3097 if (errno == ENOENT)
3098 goto not_a_valid_file;
3099 die_errno("stat(%s)", name);
3100 }
3101 if (S_ISLNK(st.st_mode)) {
3102 struct strbuf sb = STRBUF_INIT;
3103 if (strbuf_readlink(&sb, name, st.st_size) < 0)
3104 die_errno("readlink(%s)", name);
3105 prep_temp_blob(name, temp, sb.buf, sb.len,
3106 (one->oid_valid ?
3107 &one->oid : &null_oid),
3108 (one->oid_valid ?
3109 one->mode : S_IFLNK));
3110 strbuf_release(&sb);
3111 }
3112 else {
3113 /* we can borrow from the file in the work tree */
3114 temp->name = name;
3115 if (!one->oid_valid)
3116 oid_to_hex_r(temp->hex, &null_oid);
3117 else
3118 oid_to_hex_r(temp->hex, &one->oid);
3119 /* Even though we may sometimes borrow the
3120 * contents from the work tree, we always want
3121 * one->mode. mode is trustworthy even when
3122 * !(one->oid_valid), as long as
3123 * DIFF_FILE_VALID(one).
3124 */
3125 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", one->mode);
3126 }
3127 return temp;
3128 }
3129 else {
3130 if (diff_populate_filespec(one, 0))
3131 die("cannot read data blob for %s", one->path);
3132 prep_temp_blob(name, temp, one->data, one->size,
3133 &one->oid, one->mode);
3134 }
3135 return temp;
3136 }
3137
3138 static void add_external_diff_name(struct argv_array *argv,
3139 const char *name,
3140 struct diff_filespec *df)
3141 {
3142 struct diff_tempfile *temp = prepare_temp_file(name, df);
3143 argv_array_push(argv, temp->name);
3144 argv_array_push(argv, temp->hex);
3145 argv_array_push(argv, temp->mode);
3146 }
3147
3148 /* An external diff command takes:
3149 *
3150 * diff-cmd name infile1 infile1-sha1 infile1-mode \
3151 * infile2 infile2-sha1 infile2-mode [ rename-to ]
3152 *
3153 */
3154 static void run_external_diff(const char *pgm,
3155 const char *name,
3156 const char *other,
3157 struct diff_filespec *one,
3158 struct diff_filespec *two,
3159 const char *xfrm_msg,
3160 int complete_rewrite,
3161 struct diff_options *o)
3162 {
3163 struct argv_array argv = ARGV_ARRAY_INIT;
3164 struct argv_array env = ARGV_ARRAY_INIT;
3165 struct diff_queue_struct *q = &diff_queued_diff;
3166
3167 argv_array_push(&argv, pgm);
3168 argv_array_push(&argv, name);
3169
3170 if (one && two) {
3171 add_external_diff_name(&argv, name, one);
3172 if (!other)
3173 add_external_diff_name(&argv, name, two);
3174 else {
3175 add_external_diff_name(&argv, other, two);
3176 argv_array_push(&argv, other);
3177 argv_array_push(&argv, xfrm_msg);
3178 }
3179 }
3180
3181 argv_array_pushf(&env, "GIT_DIFF_PATH_COUNTER=%d", ++o->diff_path_counter);
3182 argv_array_pushf(&env, "GIT_DIFF_PATH_TOTAL=%d", q->nr);
3183
3184 if (run_command_v_opt_cd_env(argv.argv, RUN_USING_SHELL, NULL, env.argv))
3185 die(_("external diff died, stopping at %s"), name);
3186
3187 remove_tempfile();
3188 argv_array_clear(&argv);
3189 argv_array_clear(&env);
3190 }
3191
3192 static int similarity_index(struct diff_filepair *p)
3193 {
3194 return p->score * 100 / MAX_SCORE;
3195 }
3196
3197 static const char *diff_abbrev_oid(const struct object_id *oid, int abbrev)
3198 {
3199 if (startup_info->have_repository)
3200 return find_unique_abbrev(oid->hash, abbrev);
3201 else {
3202 char *hex = oid_to_hex(oid);
3203 if (abbrev < 0)
3204 abbrev = FALLBACK_DEFAULT_ABBREV;
3205 if (abbrev > GIT_SHA1_HEXSZ)
3206 die("BUG: oid abbreviation out of range: %d", abbrev);
3207 if (abbrev)
3208 hex[abbrev] = '\0';
3209 return hex;
3210 }
3211 }
3212
3213 static void fill_metainfo(struct strbuf *msg,
3214 const char *name,
3215 const char *other,
3216 struct diff_filespec *one,
3217 struct diff_filespec *two,
3218 struct diff_options *o,
3219 struct diff_filepair *p,
3220 int *must_show_header,
3221 int use_color)
3222 {
3223 const char *set = diff_get_color(use_color, DIFF_METAINFO);
3224 const char *reset = diff_get_color(use_color, DIFF_RESET);
3225 const char *line_prefix = diff_line_prefix(o);
3226
3227 *must_show_header = 1;
3228 strbuf_init(msg, PATH_MAX * 2 + 300);
3229 switch (p->status) {
3230 case DIFF_STATUS_COPIED:
3231 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3232 line_prefix, set, similarity_index(p));
3233 strbuf_addf(msg, "%s\n%s%scopy from ",
3234 reset, line_prefix, set);
3235 quote_c_style(name, msg, NULL, 0);
3236 strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
3237 quote_c_style(other, msg, NULL, 0);
3238 strbuf_addf(msg, "%s\n", reset);
3239 break;
3240 case DIFF_STATUS_RENAMED:
3241 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3242 line_prefix, set, similarity_index(p));
3243 strbuf_addf(msg, "%s\n%s%srename from ",
3244 reset, line_prefix, set);
3245 quote_c_style(name, msg, NULL, 0);
3246 strbuf_addf(msg, "%s\n%s%srename to ",
3247 reset, line_prefix, set);
3248 quote_c_style(other, msg, NULL, 0);
3249 strbuf_addf(msg, "%s\n", reset);
3250 break;
3251 case DIFF_STATUS_MODIFIED:
3252 if (p->score) {
3253 strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
3254 line_prefix,
3255 set, similarity_index(p), reset);
3256 break;
3257 }
3258 /* fallthru */
3259 default:
3260 *must_show_header = 0;
3261 }
3262 if (one && two && oidcmp(&one->oid, &two->oid)) {
3263 int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
3264
3265 if (DIFF_OPT_TST(o, BINARY)) {
3266 mmfile_t mf;
3267 if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
3268 (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
3269 abbrev = 40;
3270 }
3271 strbuf_addf(msg, "%s%sindex %s..%s", line_prefix, set,
3272 diff_abbrev_oid(&one->oid, abbrev),
3273 diff_abbrev_oid(&two->oid, abbrev));
3274 if (one->mode == two->mode)
3275 strbuf_addf(msg, " %06o", one->mode);
3276 strbuf_addf(msg, "%s\n", reset);
3277 }
3278 }
3279
3280 static void run_diff_cmd(const char *pgm,
3281 const char *name,
3282 const char *other,
3283 const char *attr_path,
3284 struct diff_filespec *one,
3285 struct diff_filespec *two,
3286 struct strbuf *msg,
3287 struct diff_options *o,
3288 struct diff_filepair *p)
3289 {
3290 const char *xfrm_msg = NULL;
3291 int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
3292 int must_show_header = 0;
3293
3294
3295 if (DIFF_OPT_TST(o, ALLOW_EXTERNAL)) {
3296 struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
3297 if (drv && drv->external)
3298 pgm = drv->external;
3299 }
3300
3301 if (msg) {
3302 /*
3303 * don't use colors when the header is intended for an
3304 * external diff driver
3305 */
3306 fill_metainfo(msg, name, other, one, two, o, p,
3307 &must_show_header,
3308 want_color(o->use_color) && !pgm);
3309 xfrm_msg = msg->len ? msg->buf : NULL;
3310 }
3311
3312 if (pgm) {
3313 run_external_diff(pgm, name, other, one, two, xfrm_msg,
3314 complete_rewrite, o);
3315 return;
3316 }
3317 if (one && two)
3318 builtin_diff(name, other ? other : name,
3319 one, two, xfrm_msg, must_show_header,
3320 o, complete_rewrite);
3321 else
3322 fprintf(o->file, "* Unmerged path %s\n", name);
3323 }
3324
3325 static void diff_fill_oid_info(struct diff_filespec *one)
3326 {
3327 if (DIFF_FILE_VALID(one)) {
3328 if (!one->oid_valid) {
3329 struct stat st;
3330 if (one->is_stdin) {
3331 oidclr(&one->oid);
3332 return;
3333 }
3334 if (lstat(one->path, &st) < 0)
3335 die_errno("stat '%s'", one->path);
3336 if (index_path(one->oid.hash, one->path, &st, 0))
3337 die("cannot hash %s", one->path);
3338 }
3339 }
3340 else
3341 oidclr(&one->oid);
3342 }
3343
3344 static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
3345 {
3346 /* Strip the prefix but do not molest /dev/null and absolute paths */
3347 if (*namep && **namep != '/') {
3348 *namep += prefix_length;
3349 if (**namep == '/')
3350 ++*namep;
3351 }
3352 if (*otherp && **otherp != '/') {
3353 *otherp += prefix_length;
3354 if (**otherp == '/')
3355 ++*otherp;
3356 }
3357 }
3358
3359 static void run_diff(struct diff_filepair *p, struct diff_options *o)
3360 {
3361 const char *pgm = external_diff();
3362 struct strbuf msg;
3363 struct diff_filespec *one = p->one;
3364 struct diff_filespec *two = p->two;
3365 const char *name;
3366 const char *other;
3367 const char *attr_path;
3368
3369 name = one->path;
3370 other = (strcmp(name, two->path) ? two->path : NULL);
3371 attr_path = name;
3372 if (o->prefix_length)
3373 strip_prefix(o->prefix_length, &name, &other);
3374
3375 if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
3376 pgm = NULL;
3377
3378 if (DIFF_PAIR_UNMERGED(p)) {
3379 run_diff_cmd(pgm, name, NULL, attr_path,
3380 NULL, NULL, NULL, o, p);
3381 return;
3382 }
3383
3384 diff_fill_oid_info(one);
3385 diff_fill_oid_info(two);
3386
3387 if (!pgm &&
3388 DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3389 (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3390 /*
3391 * a filepair that changes between file and symlink
3392 * needs to be split into deletion and creation.
3393 */
3394 struct diff_filespec *null = alloc_filespec(two->path);
3395 run_diff_cmd(NULL, name, other, attr_path,
3396 one, null, &msg, o, p);
3397 free(null);
3398 strbuf_release(&msg);
3399
3400 null = alloc_filespec(one->path);
3401 run_diff_cmd(NULL, name, other, attr_path,
3402 null, two, &msg, o, p);
3403 free(null);
3404 }
3405 else
3406 run_diff_cmd(pgm, name, other, attr_path,
3407 one, two, &msg, o, p);
3408
3409 strbuf_release(&msg);
3410 }
3411
3412 static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3413 struct diffstat_t *diffstat)
3414 {
3415 const char *name;
3416 const char *other;
3417
3418 if (DIFF_PAIR_UNMERGED(p)) {
3419 /* unmerged */
3420 builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, p);
3421 return;
3422 }
3423
3424 name = p->one->path;
3425 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3426
3427 if (o->prefix_length)
3428 strip_prefix(o->prefix_length, &name, &other);
3429
3430 diff_fill_oid_info(p->one);
3431 diff_fill_oid_info(p->two);
3432
3433 builtin_diffstat(name, other, p->one, p->two, diffstat, o, p);
3434 }
3435
3436 static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3437 {
3438 const char *name;
3439 const char *other;
3440 const char *attr_path;
3441
3442 if (DIFF_PAIR_UNMERGED(p)) {
3443 /* unmerged */
3444 return;
3445 }
3446
3447 name = p->one->path;
3448 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3449 attr_path = other ? other : name;
3450
3451 if (o->prefix_length)
3452 strip_prefix(o->prefix_length, &name, &other);
3453
3454 diff_fill_oid_info(p->one);
3455 diff_fill_oid_info(p->two);
3456
3457 builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3458 }
3459
3460 void diff_setup(struct diff_options *options)
3461 {
3462 memcpy(options, &default_diff_options, sizeof(*options));
3463
3464 options->file = stdout;
3465
3466 options->abbrev = DEFAULT_ABBREV;
3467 options->line_termination = '\n';
3468 options->break_opt = -1;
3469 options->rename_limit = -1;
3470 options->dirstat_permille = diff_dirstat_permille_default;
3471 options->context = diff_context_default;
3472 options->interhunkcontext = diff_interhunk_context_default;
3473 options->ws_error_highlight = ws_error_highlight_default;
3474 DIFF_OPT_SET(options, RENAME_EMPTY);
3475
3476 /* pathchange left =NULL by default */
3477 options->change = diff_change;
3478 options->add_remove = diff_addremove;
3479 options->use_color = diff_use_color_default;
3480 options->detect_rename = diff_detect_rename_default;
3481 options->xdl_opts |= diff_algorithm;
3482 if (diff_indent_heuristic)
3483 DIFF_XDL_SET(options, INDENT_HEURISTIC);
3484
3485 options->orderfile = diff_order_file_cfg;
3486
3487 if (diff_no_prefix) {
3488 options->a_prefix = options->b_prefix = "";
3489 } else if (!diff_mnemonic_prefix) {
3490 options->a_prefix = "a/";
3491 options->b_prefix = "b/";
3492 }
3493 }
3494
3495 void diff_setup_done(struct diff_options *options)
3496 {
3497 int count = 0;
3498
3499 if (options->set_default)
3500 options->set_default(options);
3501
3502 if (options->output_format & DIFF_FORMAT_NAME)
3503 count++;
3504 if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3505 count++;
3506 if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3507 count++;
3508 if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3509 count++;
3510 if (count > 1)
3511 die(_("--name-only, --name-status, --check and -s are mutually exclusive"));
3512
3513 /*
3514 * Most of the time we can say "there are changes"
3515 * only by checking if there are changed paths, but
3516 * --ignore-whitespace* options force us to look
3517 * inside contents.
3518 */
3519
3520 if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3521 DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3522 DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3523 DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3524 else
3525 DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3526
3527 if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3528 options->detect_rename = DIFF_DETECT_COPY;
3529
3530 if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3531 options->prefix = NULL;
3532 if (options->prefix)
3533 options->prefix_length = strlen(options->prefix);
3534 else
3535 options->prefix_length = 0;
3536
3537 if (options->output_format & (DIFF_FORMAT_NAME |
3538 DIFF_FORMAT_NAME_STATUS |
3539 DIFF_FORMAT_CHECKDIFF |
3540 DIFF_FORMAT_NO_OUTPUT))
3541 options->output_format &= ~(DIFF_FORMAT_RAW |
3542 DIFF_FORMAT_NUMSTAT |
3543 DIFF_FORMAT_DIFFSTAT |
3544 DIFF_FORMAT_SHORTSTAT |
3545 DIFF_FORMAT_DIRSTAT |
3546 DIFF_FORMAT_SUMMARY |
3547 DIFF_FORMAT_PATCH);
3548
3549 /*
3550 * These cases always need recursive; we do not drop caller-supplied
3551 * recursive bits for other formats here.
3552 */
3553 if (options->output_format & (DIFF_FORMAT_PATCH |
3554 DIFF_FORMAT_NUMSTAT |
3555 DIFF_FORMAT_DIFFSTAT |
3556 DIFF_FORMAT_SHORTSTAT |
3557 DIFF_FORMAT_DIRSTAT |
3558 DIFF_FORMAT_SUMMARY |
3559 DIFF_FORMAT_CHECKDIFF))
3560 DIFF_OPT_SET(options, RECURSIVE);
3561 /*
3562 * Also pickaxe would not work very well if you do not say recursive
3563 */
3564 if (options->pickaxe)
3565 DIFF_OPT_SET(options, RECURSIVE);
3566 /*
3567 * When patches are generated, submodules diffed against the work tree
3568 * must be checked for dirtiness too so it can be shown in the output
3569 */
3570 if (options->output_format & DIFF_FORMAT_PATCH)
3571 DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3572
3573 if (options->detect_rename && options->rename_limit < 0)
3574 options->rename_limit = diff_rename_limit_default;
3575 if (options->setup & DIFF_SETUP_USE_CACHE) {
3576 if (!active_cache)
3577 /* read-cache does not die even when it fails
3578 * so it is safe for us to do this here. Also
3579 * it does not smudge active_cache or active_nr
3580 * when it fails, so we do not have to worry about
3581 * cleaning it up ourselves either.
3582 */
3583 read_cache();
3584 }
3585 if (40 < options->abbrev)
3586 options->abbrev = 40; /* full */
3587
3588 /*
3589 * It does not make sense to show the first hit we happened
3590 * to have found. It does not make sense not to return with
3591 * exit code in such a case either.
3592 */
3593 if (DIFF_OPT_TST(options, QUICK)) {
3594 options->output_format = DIFF_FORMAT_NO_OUTPUT;
3595 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3596 }
3597
3598 options->diff_path_counter = 0;
3599
3600 if (DIFF_OPT_TST(options, FOLLOW_RENAMES) && options->pathspec.nr != 1)
3601 die(_("--follow requires exactly one pathspec"));
3602 }
3603
3604 static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3605 {
3606 char c, *eq;
3607 int len;
3608
3609 if (*arg != '-')
3610 return 0;
3611 c = *++arg;
3612 if (!c)
3613 return 0;
3614 if (c == arg_short) {
3615 c = *++arg;
3616 if (!c)
3617 return 1;
3618 if (val && isdigit(c)) {
3619 char *end;
3620 int n = strtoul(arg, &end, 10);
3621 if (*end)
3622 return 0;
3623 *val = n;
3624 return 1;
3625 }
3626 return 0;
3627 }
3628 if (c != '-')
3629 return 0;
3630 arg++;
3631 eq = strchrnul(arg, '=');
3632 len = eq - arg;
3633 if (!len || strncmp(arg, arg_long, len))
3634 return 0;
3635 if (*eq) {
3636 int n;
3637 char *end;
3638 if (!isdigit(*++eq))
3639 return 0;
3640 n = strtoul(eq, &end, 10);
3641 if (*end)
3642 return 0;
3643 *val = n;
3644 }
3645 return 1;
3646 }
3647
3648 static int diff_scoreopt_parse(const char *opt);
3649
3650 static inline int short_opt(char opt, const char **argv,
3651 const char **optarg)
3652 {
3653 const char *arg = argv[0];
3654 if (arg[0] != '-' || arg[1] != opt)
3655 return 0;
3656 if (arg[2] != '\0') {
3657 *optarg = arg + 2;
3658 return 1;
3659 }
3660 if (!argv[1])
3661 die("Option '%c' requires a value", opt);
3662 *optarg = argv[1];
3663 return 2;
3664 }
3665
3666 int parse_long_opt(const char *opt, const char **argv,
3667 const char **optarg)
3668 {
3669 const char *arg = argv[0];
3670 if (!skip_prefix(arg, "--", &arg))
3671 return 0;
3672 if (!skip_prefix(arg, opt, &arg))
3673 return 0;
3674 if (*arg == '=') { /* stuck form: --option=value */
3675 *optarg = arg + 1;
3676 return 1;
3677 }
3678 if (*arg != '\0')
3679 return 0;
3680 /* separate form: --option value */
3681 if (!argv[1])
3682 die("Option '--%s' requires a value", opt);
3683 *optarg = argv[1];
3684 return 2;
3685 }
3686
3687 static int stat_opt(struct diff_options *options, const char **av)
3688 {
3689 const char *arg = av[0];
3690 char *end;
3691 int width = options->stat_width;
3692 int name_width = options->stat_name_width;
3693 int graph_width = options->stat_graph_width;
3694 int count = options->stat_count;
3695 int argcount = 1;
3696
3697 if (!skip_prefix(arg, "--stat", &arg))
3698 die("BUG: stat option does not begin with --stat: %s", arg);
3699 end = (char *)arg;
3700
3701 switch (*arg) {
3702 case '-':
3703 if (skip_prefix(arg, "-width", &arg)) {
3704 if (*arg == '=')
3705 width = strtoul(arg + 1, &end, 10);
3706 else if (!*arg && !av[1])
3707 die_want_option("--stat-width");
3708 else if (!*arg) {
3709 width = strtoul(av[1], &end, 10);
3710 argcount = 2;
3711 }
3712 } else if (skip_prefix(arg, "-name-width", &arg)) {
3713 if (*arg == '=')
3714 name_width = strtoul(arg + 1, &end, 10);
3715 else if (!*arg && !av[1])
3716 die_want_option("--stat-name-width");
3717 else if (!*arg) {
3718 name_width = strtoul(av[1], &end, 10);
3719 argcount = 2;
3720 }
3721 } else if (skip_prefix(arg, "-graph-width", &arg)) {
3722 if (*arg == '=')
3723 graph_width = strtoul(arg + 1, &end, 10);
3724 else if (!*arg && !av[1])
3725 die_want_option("--stat-graph-width");
3726 else if (!*arg) {
3727 graph_width = strtoul(av[1], &end, 10);
3728 argcount = 2;
3729 }
3730 } else if (skip_prefix(arg, "-count", &arg)) {
3731 if (*arg == '=')
3732 count = strtoul(arg + 1, &end, 10);
3733 else if (!*arg && !av[1])
3734 die_want_option("--stat-count");
3735 else if (!*arg) {
3736 count = strtoul(av[1], &end, 10);
3737 argcount = 2;
3738 }
3739 }
3740 break;
3741 case '=':
3742 width = strtoul(arg+1, &end, 10);
3743 if (*end == ',')
3744 name_width = strtoul(end+1, &end, 10);
3745 if (*end == ',')
3746 count = strtoul(end+1, &end, 10);
3747 }
3748
3749 /* Important! This checks all the error cases! */
3750 if (*end)
3751 return 0;
3752 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3753 options->stat_name_width = name_width;
3754 options->stat_graph_width = graph_width;
3755 options->stat_width = width;
3756 options->stat_count = count;
3757 return argcount;
3758 }
3759
3760 static int parse_dirstat_opt(struct diff_options *options, const char *params)
3761 {
3762 struct strbuf errmsg = STRBUF_INIT;
3763 if (parse_dirstat_params(options, params, &errmsg))
3764 die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3765 errmsg.buf);
3766 strbuf_release(&errmsg);
3767 /*
3768 * The caller knows a dirstat-related option is given from the command
3769 * line; allow it to say "return this_function();"
3770 */
3771 options->output_format |= DIFF_FORMAT_DIRSTAT;
3772 return 1;
3773 }
3774
3775 static int parse_submodule_opt(struct diff_options *options, const char *value)
3776 {
3777 if (parse_submodule_params(options, value))
3778 die(_("Failed to parse --submodule option parameter: '%s'"),
3779 value);
3780 return 1;
3781 }
3782
3783 static const char diff_status_letters[] = {
3784 DIFF_STATUS_ADDED,
3785 DIFF_STATUS_COPIED,
3786 DIFF_STATUS_DELETED,
3787 DIFF_STATUS_MODIFIED,
3788 DIFF_STATUS_RENAMED,
3789 DIFF_STATUS_TYPE_CHANGED,
3790 DIFF_STATUS_UNKNOWN,
3791 DIFF_STATUS_UNMERGED,
3792 DIFF_STATUS_FILTER_AON,
3793 DIFF_STATUS_FILTER_BROKEN,
3794 '\0',
3795 };
3796
3797 static unsigned int filter_bit['Z' + 1];
3798
3799 static void prepare_filter_bits(void)
3800 {
3801 int i;
3802
3803 if (!filter_bit[DIFF_STATUS_ADDED]) {
3804 for (i = 0; diff_status_letters[i]; i++)
3805 filter_bit[(int) diff_status_letters[i]] = (1 << i);
3806 }
3807 }
3808
3809 static unsigned filter_bit_tst(char status, const struct diff_options *opt)
3810 {
3811 return opt->filter & filter_bit[(int) status];
3812 }
3813
3814 static int parse_diff_filter_opt(const char *optarg, struct diff_options *opt)
3815 {
3816 int i, optch;
3817
3818 prepare_filter_bits();
3819
3820 /*
3821 * If there is a negation e.g. 'd' in the input, and we haven't
3822 * initialized the filter field with another --diff-filter, start
3823 * from full set of bits, except for AON.
3824 */
3825 if (!opt->filter) {
3826 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3827 if (optch < 'a' || 'z' < optch)
3828 continue;
3829 opt->filter = (1 << (ARRAY_SIZE(diff_status_letters) - 1)) - 1;
3830 opt->filter &= ~filter_bit[DIFF_STATUS_FILTER_AON];
3831 break;
3832 }
3833 }
3834
3835 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3836 unsigned int bit;
3837 int negate;
3838
3839 if ('a' <= optch && optch <= 'z') {
3840 negate = 1;
3841 optch = toupper(optch);
3842 } else {
3843 negate = 0;
3844 }
3845
3846 bit = (0 <= optch && optch <= 'Z') ? filter_bit[optch] : 0;
3847 if (!bit)
3848 return optarg[i];
3849 if (negate)
3850 opt->filter &= ~bit;
3851 else
3852 opt->filter |= bit;
3853 }
3854 return 0;
3855 }
3856
3857 static void enable_patch_output(int *fmt) {
3858 *fmt &= ~DIFF_FORMAT_NO_OUTPUT;
3859 *fmt |= DIFF_FORMAT_PATCH;
3860 }
3861
3862 static int parse_ws_error_highlight_opt(struct diff_options *opt, const char *arg)
3863 {
3864 int val = parse_ws_error_highlight(arg);
3865
3866 if (val < 0) {
3867 error("unknown value after ws-error-highlight=%.*s",
3868 -1 - val, arg);
3869 return 0;
3870 }
3871 opt->ws_error_highlight = val;
3872 return 1;
3873 }
3874
3875 int diff_opt_parse(struct diff_options *options,
3876 const char **av, int ac, const char *prefix)
3877 {
3878 const char *arg = av[0];
3879 const char *optarg;
3880 int argcount;
3881
3882 if (!prefix)
3883 prefix = "";
3884
3885 /* Output format options */
3886 if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch")
3887 || opt_arg(arg, 'U', "unified", &options->context))
3888 enable_patch_output(&options->output_format);
3889 else if (!strcmp(arg, "--raw"))
3890 options->output_format |= DIFF_FORMAT_RAW;
3891 else if (!strcmp(arg, "--patch-with-raw")) {
3892 enable_patch_output(&options->output_format);
3893 options->output_format |= DIFF_FORMAT_RAW;
3894 } else if (!strcmp(arg, "--numstat"))
3895 options->output_format |= DIFF_FORMAT_NUMSTAT;
3896 else if (!strcmp(arg, "--shortstat"))
3897 options->output_format |= DIFF_FORMAT_SHORTSTAT;
3898 else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3899 return parse_dirstat_opt(options, "");
3900 else if (skip_prefix(arg, "-X", &arg))
3901 return parse_dirstat_opt(options, arg);
3902 else if (skip_prefix(arg, "--dirstat=", &arg))
3903 return parse_dirstat_opt(options, arg);
3904 else if (!strcmp(arg, "--cumulative"))
3905 return parse_dirstat_opt(options, "cumulative");
3906 else if (!strcmp(arg, "--dirstat-by-file"))
3907 return parse_dirstat_opt(options, "files");
3908 else if (skip_prefix(arg, "--dirstat-by-file=", &arg)) {
3909 parse_dirstat_opt(options, "files");
3910 return parse_dirstat_opt(options, arg);
3911 }
3912 else if (!strcmp(arg, "--check"))
3913 options->output_format |= DIFF_FORMAT_CHECKDIFF;
3914 else if (!strcmp(arg, "--summary"))
3915 options->output_format |= DIFF_FORMAT_SUMMARY;
3916 else if (!strcmp(arg, "--patch-with-stat")) {
3917 enable_patch_output(&options->output_format);
3918 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3919 } else if (!strcmp(arg, "--name-only"))
3920 options->output_format |= DIFF_FORMAT_NAME;
3921 else if (!strcmp(arg, "--name-status"))
3922 options->output_format |= DIFF_FORMAT_NAME_STATUS;
3923 else if (!strcmp(arg, "-s") || !strcmp(arg, "--no-patch"))
3924 options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3925 else if (starts_with(arg, "--stat"))
3926 /* --stat, --stat-width, --stat-name-width, or --stat-count */
3927 return stat_opt(options, av);
3928
3929 /* renames options */
3930 else if (starts_with(arg, "-B") || starts_with(arg, "--break-rewrites=") ||
3931 !strcmp(arg, "--break-rewrites")) {
3932 if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3933 return error("invalid argument to -B: %s", arg+2);
3934 }
3935 else if (starts_with(arg, "-M") || starts_with(arg, "--find-renames=") ||
3936 !strcmp(arg, "--find-renames")) {
3937 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3938 return error("invalid argument to -M: %s", arg+2);
3939 options->detect_rename = DIFF_DETECT_RENAME;
3940 }
3941 else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3942 options->irreversible_delete = 1;
3943 }
3944 else if (starts_with(arg, "-C") || starts_with(arg, "--find-copies=") ||
3945 !strcmp(arg, "--find-copies")) {
3946 if (options->detect_rename == DIFF_DETECT_COPY)
3947 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3948 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3949 return error("invalid argument to -C: %s", arg+2);
3950 options->detect_rename = DIFF_DETECT_COPY;
3951 }
3952 else if (!strcmp(arg, "--no-renames"))
3953 options->detect_rename = 0;
3954 else if (!strcmp(arg, "--rename-empty"))
3955 DIFF_OPT_SET(options, RENAME_EMPTY);
3956 else if (!strcmp(arg, "--no-rename-empty"))
3957 DIFF_OPT_CLR(options, RENAME_EMPTY);
3958 else if (!strcmp(arg, "--relative"))
3959 DIFF_OPT_SET(options, RELATIVE_NAME);
3960 else if (skip_prefix(arg, "--relative=", &arg)) {
3961 DIFF_OPT_SET(options, RELATIVE_NAME);
3962 options->prefix = arg;
3963 }
3964
3965 /* xdiff options */
3966 else if (!strcmp(arg, "--minimal"))
3967 DIFF_XDL_SET(options, NEED_MINIMAL);
3968 else if (!strcmp(arg, "--no-minimal"))
3969 DIFF_XDL_CLR(options, NEED_MINIMAL);
3970 else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3971 DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3972 else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3973 DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3974 else if (!strcmp(arg, "--ignore-space-at-eol"))
3975 DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3976 else if (!strcmp(arg, "--ignore-blank-lines"))
3977 DIFF_XDL_SET(options, IGNORE_BLANK_LINES);
3978 else if (!strcmp(arg, "--indent-heuristic"))
3979 DIFF_XDL_SET(options, INDENT_HEURISTIC);
3980 else if (!strcmp(arg, "--no-indent-heuristic"))
3981 DIFF_XDL_CLR(options, INDENT_HEURISTIC);
3982 else if (!strcmp(arg, "--patience"))
3983 options->xdl_opts = DIFF_WITH_ALG(options, PATIENCE_DIFF);
3984 else if (!strcmp(arg, "--histogram"))
3985 options->xdl_opts = DIFF_WITH_ALG(options, HISTOGRAM_DIFF);
3986 else if ((argcount = parse_long_opt("diff-algorithm", av, &optarg))) {
3987 long value = parse_algorithm_value(optarg);
3988 if (value < 0)
3989 return error("option diff-algorithm accepts \"myers\", "
3990 "\"minimal\", \"patience\" and \"histogram\"");
3991 /* clear out previous settings */
3992 DIFF_XDL_CLR(options, NEED_MINIMAL);
3993 options->xdl_opts &= ~XDF_DIFF_ALGORITHM_MASK;
3994 options->xdl_opts |= value;
3995 return argcount;
3996 }
3997
3998 /* flags options */
3999 else if (!strcmp(arg, "--binary")) {
4000 enable_patch_output(&options->output_format);
4001 DIFF_OPT_SET(options, BINARY);
4002 }
4003 else if (!strcmp(arg, "--full-index"))
4004 DIFF_OPT_SET(options, FULL_INDEX);
4005 else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
4006 DIFF_OPT_SET(options, TEXT);
4007 else if (!strcmp(arg, "-R"))
4008 DIFF_OPT_SET(options, REVERSE_DIFF);
4009 else if (!strcmp(arg, "--find-copies-harder"))
4010 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
4011 else if (!strcmp(arg, "--follow"))
4012 DIFF_OPT_SET(options, FOLLOW_RENAMES);
4013 else if (!strcmp(arg, "--no-follow")) {
4014 DIFF_OPT_CLR(options, FOLLOW_RENAMES);
4015 DIFF_OPT_CLR(options, DEFAULT_FOLLOW_RENAMES);
4016 } else if (!strcmp(arg, "--color"))
4017 options->use_color = 1;
4018 else if (skip_prefix(arg, "--color=", &arg)) {
4019 int value = git_config_colorbool(NULL, arg);
4020 if (value < 0)
4021 return error("option `color' expects \"always\", \"auto\", or \"never\"");
4022 options->use_color = value;
4023 }
4024 else if (!strcmp(arg, "--no-color"))
4025 options->use_color = 0;
4026 else if (!strcmp(arg, "--color-words")) {
4027 options->use_color = 1;
4028 options->word_diff = DIFF_WORDS_COLOR;
4029 }
4030 else if (skip_prefix(arg, "--color-words=", &arg)) {
4031 options->use_color = 1;
4032 options->word_diff = DIFF_WORDS_COLOR;
4033 options->word_regex = arg;
4034 }
4035 else if (!strcmp(arg, "--word-diff")) {
4036 if (options->word_diff == DIFF_WORDS_NONE)
4037 options->word_diff = DIFF_WORDS_PLAIN;
4038 }
4039 else if (skip_prefix(arg, "--word-diff=", &arg)) {
4040 if (!strcmp(arg, "plain"))
4041 options->word_diff = DIFF_WORDS_PLAIN;
4042 else if (!strcmp(arg, "color")) {
4043 options->use_color = 1;
4044 options->word_diff = DIFF_WORDS_COLOR;
4045 }
4046 else if (!strcmp(arg, "porcelain"))
4047 options->word_diff = DIFF_WORDS_PORCELAIN;
4048 else if (!strcmp(arg, "none"))
4049 options->word_diff = DIFF_WORDS_NONE;
4050 else
4051 die("bad --word-diff argument: %s", arg);
4052 }
4053 else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
4054 if (options->word_diff == DIFF_WORDS_NONE)
4055 options->word_diff = DIFF_WORDS_PLAIN;
4056 options->word_regex = optarg;
4057 return argcount;
4058 }
4059 else if (!strcmp(arg, "--exit-code"))
4060 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
4061 else if (!strcmp(arg, "--quiet"))
4062 DIFF_OPT_SET(options, QUICK);
4063 else if (!strcmp(arg, "--ext-diff"))
4064 DIFF_OPT_SET(options, ALLOW_EXTERNAL);
4065 else if (!strcmp(arg, "--no-ext-diff"))
4066 DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
4067 else if (!strcmp(arg, "--textconv"))
4068 DIFF_OPT_SET(options, ALLOW_TEXTCONV);
4069 else if (!strcmp(arg, "--no-textconv"))
4070 DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
4071 else if (!strcmp(arg, "--ignore-submodules")) {
4072 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
4073 handle_ignore_submodules_arg(options, "all");
4074 } else if (skip_prefix(arg, "--ignore-submodules=", &arg)) {
4075 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
4076 handle_ignore_submodules_arg(options, arg);
4077 } else if (!strcmp(arg, "--submodule"))
4078 options->submodule_format = DIFF_SUBMODULE_LOG;
4079 else if (skip_prefix(arg, "--submodule=", &arg))
4080 return parse_submodule_opt(options, arg);
4081 else if (skip_prefix(arg, "--ws-error-highlight=", &arg))
4082 return parse_ws_error_highlight_opt(options, arg);
4083 else if (!strcmp(arg, "--ita-invisible-in-index"))
4084 options->ita_invisible_in_index = 1;
4085 else if (!strcmp(arg, "--ita-visible-in-index"))
4086 options->ita_invisible_in_index = 0;
4087
4088 /* misc options */
4089 else if (!strcmp(arg, "-z"))
4090 options->line_termination = 0;
4091 else if ((argcount = short_opt('l', av, &optarg))) {
4092 options->rename_limit = strtoul(optarg, NULL, 10);
4093 return argcount;
4094 }
4095 else if ((argcount = short_opt('S', av, &optarg))) {
4096 options->pickaxe = optarg;
4097 options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
4098 return argcount;
4099 } else if ((argcount = short_opt('G', av, &optarg))) {
4100 options->pickaxe = optarg;
4101 options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
4102 return argcount;
4103 }
4104 else if (!strcmp(arg, "--pickaxe-all"))
4105 options->pickaxe_opts |= DIFF_PICKAXE_ALL;
4106 else if (!strcmp(arg, "--pickaxe-regex"))
4107 options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
4108 else if ((argcount = short_opt('O', av, &optarg))) {
4109 options->orderfile = prefix_filename(prefix, optarg);
4110 return argcount;
4111 }
4112 else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
4113 int offending = parse_diff_filter_opt(optarg, options);
4114 if (offending)
4115 die("unknown change class '%c' in --diff-filter=%s",
4116 offending, optarg);
4117 return argcount;
4118 }
4119 else if (!strcmp(arg, "--no-abbrev"))
4120 options->abbrev = 0;
4121 else if (!strcmp(arg, "--abbrev"))
4122 options->abbrev = DEFAULT_ABBREV;
4123 else if (skip_prefix(arg, "--abbrev=", &arg)) {
4124 options->abbrev = strtoul(arg, NULL, 10);
4125 if (options->abbrev < MINIMUM_ABBREV)
4126 options->abbrev = MINIMUM_ABBREV;
4127 else if (40 < options->abbrev)
4128 options->abbrev = 40;
4129 }
4130 else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
4131 options->a_prefix = optarg;
4132 return argcount;
4133 }
4134 else if ((argcount = parse_long_opt("line-prefix", av, &optarg))) {
4135 options->line_prefix = optarg;
4136 options->line_prefix_length = strlen(options->line_prefix);
4137 graph_setup_line_prefix(options);
4138 return argcount;
4139 }
4140 else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
4141 options->b_prefix = optarg;
4142 return argcount;
4143 }
4144 else if (!strcmp(arg, "--no-prefix"))
4145 options->a_prefix = options->b_prefix = "";
4146 else if (opt_arg(arg, '\0', "inter-hunk-context",
4147 &options->interhunkcontext))
4148 ;
4149 else if (!strcmp(arg, "-W"))
4150 DIFF_OPT_SET(options, FUNCCONTEXT);
4151 else if (!strcmp(arg, "--function-context"))
4152 DIFF_OPT_SET(options, FUNCCONTEXT);
4153 else if (!strcmp(arg, "--no-function-context"))
4154 DIFF_OPT_CLR(options, FUNCCONTEXT);
4155 else if ((argcount = parse_long_opt("output", av, &optarg))) {
4156 char *path = prefix_filename(prefix, optarg);
4157 options->file = xfopen(path, "w");
4158 options->close_file = 1;
4159 if (options->use_color != GIT_COLOR_ALWAYS)
4160 options->use_color = GIT_COLOR_NEVER;
4161 free(path);
4162 return argcount;
4163 } else
4164 return 0;
4165 return 1;
4166 }
4167
4168 int parse_rename_score(const char **cp_p)
4169 {
4170 unsigned long num, scale;
4171 int ch, dot;
4172 const char *cp = *cp_p;
4173
4174 num = 0;
4175 scale = 1;
4176 dot = 0;
4177 for (;;) {
4178 ch = *cp;
4179 if ( !dot && ch == '.' ) {
4180 scale = 1;
4181 dot = 1;
4182 } else if ( ch == '%' ) {
4183 scale = dot ? scale*100 : 100;
4184 cp++; /* % is always at the end */
4185 break;
4186 } else if ( ch >= '0' && ch <= '9' ) {
4187 if ( scale < 100000 ) {
4188 scale *= 10;
4189 num = (num*10) + (ch-'0');
4190 }
4191 } else {
4192 break;
4193 }
4194 cp++;
4195 }
4196 *cp_p = cp;
4197
4198 /* user says num divided by scale and we say internally that
4199 * is MAX_SCORE * num / scale.
4200 */
4201 return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
4202 }
4203
4204 static int diff_scoreopt_parse(const char *opt)
4205 {
4206 int opt1, opt2, cmd;
4207
4208 if (*opt++ != '-')
4209 return -1;
4210 cmd = *opt++;
4211 if (cmd == '-') {
4212 /* convert the long-form arguments into short-form versions */
4213 if (skip_prefix(opt, "break-rewrites", &opt)) {
4214 if (*opt == 0 || *opt++ == '=')
4215 cmd = 'B';
4216 } else if (skip_prefix(opt, "find-copies", &opt)) {
4217 if (*opt == 0 || *opt++ == '=')
4218 cmd = 'C';
4219 } else if (skip_prefix(opt, "find-renames", &opt)) {
4220 if (*opt == 0 || *opt++ == '=')
4221 cmd = 'M';
4222 }
4223 }
4224 if (cmd != 'M' && cmd != 'C' && cmd != 'B')
4225 return -1; /* that is not a -M, -C, or -B option */
4226
4227 opt1 = parse_rename_score(&opt);
4228 if (cmd != 'B')
4229 opt2 = 0;
4230 else {
4231 if (*opt == 0)
4232 opt2 = 0;
4233 else if (*opt != '/')
4234 return -1; /* we expect -B80/99 or -B80 */
4235 else {
4236 opt++;
4237 opt2 = parse_rename_score(&opt);
4238 }
4239 }
4240 if (*opt != 0)
4241 return -1;
4242 return opt1 | (opt2 << 16);
4243 }
4244
4245 struct diff_queue_struct diff_queued_diff;
4246
4247 void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
4248 {
4249 ALLOC_GROW(queue->queue, queue->nr + 1, queue->alloc);
4250 queue->queue[queue->nr++] = dp;
4251 }
4252
4253 struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
4254 struct diff_filespec *one,
4255 struct diff_filespec *two)
4256 {
4257 struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
4258 dp->one = one;
4259 dp->two = two;
4260 if (queue)
4261 diff_q(queue, dp);
4262 return dp;
4263 }
4264
4265 void diff_free_filepair(struct diff_filepair *p)
4266 {
4267 free_filespec(p->one);
4268 free_filespec(p->two);
4269 free(p);
4270 }
4271
4272 const char *diff_aligned_abbrev(const struct object_id *oid, int len)
4273 {
4274 int abblen;
4275 const char *abbrev;
4276
4277 if (len == GIT_SHA1_HEXSZ)
4278 return oid_to_hex(oid);
4279
4280 abbrev = diff_abbrev_oid(oid, len);
4281 abblen = strlen(abbrev);
4282
4283 /*
4284 * In well-behaved cases, where the abbbreviated result is the
4285 * same as the requested length, append three dots after the
4286 * abbreviation (hence the whole logic is limited to the case
4287 * where abblen < 37); when the actual abbreviated result is a
4288 * bit longer than the requested length, we reduce the number
4289 * of dots so that they match the well-behaved ones. However,
4290 * if the actual abbreviation is longer than the requested
4291 * length by more than three, we give up on aligning, and add
4292 * three dots anyway, to indicate that the output is not the
4293 * full object name. Yes, this may be suboptimal, but this
4294 * appears only in "diff --raw --abbrev" output and it is not
4295 * worth the effort to change it now. Note that this would
4296 * likely to work fine when the automatic sizing of default
4297 * abbreviation length is used--we would be fed -1 in "len" in
4298 * that case, and will end up always appending three-dots, but
4299 * the automatic sizing is supposed to give abblen that ensures
4300 * uniqueness across all objects (statistically speaking).
4301 */
4302 if (abblen < GIT_SHA1_HEXSZ - 3) {
4303 static char hex[GIT_MAX_HEXSZ + 1];
4304 if (len < abblen && abblen <= len + 2)
4305 xsnprintf(hex, sizeof(hex), "%s%.*s", abbrev, len+3-abblen, "..");
4306 else
4307 xsnprintf(hex, sizeof(hex), "%s...", abbrev);
4308 return hex;
4309 }
4310
4311 return oid_to_hex(oid);
4312 }
4313
4314 static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
4315 {
4316 int line_termination = opt->line_termination;
4317 int inter_name_termination = line_termination ? '\t' : '\0';
4318
4319 fprintf(opt->file, "%s", diff_line_prefix(opt));
4320 if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
4321 fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
4322 diff_aligned_abbrev(&p->one->oid, opt->abbrev));
4323 fprintf(opt->file, "%s ",
4324 diff_aligned_abbrev(&p->two->oid, opt->abbrev));
4325 }
4326 if (p->score) {
4327 fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
4328 inter_name_termination);
4329 } else {
4330 fprintf(opt->file, "%c%c", p->status, inter_name_termination);
4331 }
4332
4333 if (p->status == DIFF_STATUS_COPIED ||
4334 p->status == DIFF_STATUS_RENAMED) {
4335 const char *name_a, *name_b;
4336 name_a = p->one->path;
4337 name_b = p->two->path;
4338 strip_prefix(opt->prefix_length, &name_a, &name_b);
4339 write_name_quoted(name_a, opt->file, inter_name_termination);
4340 write_name_quoted(name_b, opt->file, line_termination);
4341 } else {
4342 const char *name_a, *name_b;
4343 name_a = p->one->mode ? p->one->path : p->two->path;
4344 name_b = NULL;
4345 strip_prefix(opt->prefix_length, &name_a, &name_b);
4346 write_name_quoted(name_a, opt->file, line_termination);
4347 }
4348 }
4349
4350 int diff_unmodified_pair(struct diff_filepair *p)
4351 {
4352 /* This function is written stricter than necessary to support
4353 * the currently implemented transformers, but the idea is to
4354 * let transformers to produce diff_filepairs any way they want,
4355 * and filter and clean them up here before producing the output.
4356 */
4357 struct diff_filespec *one = p->one, *two = p->two;
4358
4359 if (DIFF_PAIR_UNMERGED(p))
4360 return 0; /* unmerged is interesting */
4361
4362 /* deletion, addition, mode or type change
4363 * and rename are all interesting.
4364 */
4365 if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
4366 DIFF_PAIR_MODE_CHANGED(p) ||
4367 strcmp(one->path, two->path))
4368 return 0;
4369
4370 /* both are valid and point at the same path. that is, we are
4371 * dealing with a change.
4372 */
4373 if (one->oid_valid && two->oid_valid &&
4374 !oidcmp(&one->oid, &two->oid) &&
4375 !one->dirty_submodule && !two->dirty_submodule)
4376 return 1; /* no change */
4377 if (!one->oid_valid && !two->oid_valid)
4378 return 1; /* both look at the same file on the filesystem. */
4379 return 0;
4380 }
4381
4382 static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
4383 {
4384 if (diff_unmodified_pair(p))
4385 return;
4386
4387 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4388 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4389 return; /* no tree diffs in patch format */
4390
4391 run_diff(p, o);
4392 }
4393
4394 static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
4395 struct diffstat_t *diffstat)
4396 {
4397 if (diff_unmodified_pair(p))
4398 return;
4399
4400 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4401 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4402 return; /* no useful stat for tree diffs */
4403
4404 run_diffstat(p, o, diffstat);
4405 }
4406
4407 static void diff_flush_checkdiff(struct diff_filepair *p,
4408 struct diff_options *o)
4409 {
4410 if (diff_unmodified_pair(p))
4411 return;
4412
4413 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4414 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4415 return; /* nothing to check in tree diffs */
4416
4417 run_checkdiff(p, o);
4418 }
4419
4420 int diff_queue_is_empty(void)
4421 {
4422 struct diff_queue_struct *q = &diff_queued_diff;
4423 int i;
4424 for (i = 0; i < q->nr; i++)
4425 if (!diff_unmodified_pair(q->queue[i]))
4426 return 0;
4427 return 1;
4428 }
4429
4430 #if DIFF_DEBUG
4431 void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
4432 {
4433 fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
4434 x, one ? one : "",
4435 s->path,
4436 DIFF_FILE_VALID(s) ? "valid" : "invalid",
4437 s->mode,
4438 s->oid_valid ? oid_to_hex(&s->oid) : "");
4439 fprintf(stderr, "queue[%d] %s size %lu\n",
4440 x, one ? one : "",
4441 s->size);
4442 }
4443
4444 void diff_debug_filepair(const struct diff_filepair *p, int i)
4445 {
4446 diff_debug_filespec(p->one, i, "one");
4447 diff_debug_filespec(p->two, i, "two");
4448 fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
4449 p->score, p->status ? p->status : '?',
4450 p->one->rename_used, p->broken_pair);
4451 }
4452
4453 void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
4454 {
4455 int i;
4456 if (msg)
4457 fprintf(stderr, "%s\n", msg);
4458 fprintf(stderr, "q->nr = %d\n", q->nr);
4459 for (i = 0; i < q->nr; i++) {
4460 struct diff_filepair *p = q->queue[i];
4461 diff_debug_filepair(p, i);
4462 }
4463 }
4464 #endif
4465
4466 static void diff_resolve_rename_copy(void)
4467 {
4468 int i;
4469 struct diff_filepair *p;
4470 struct diff_queue_struct *q = &diff_queued_diff;
4471
4472 diff_debug_queue("resolve-rename-copy", q);
4473
4474 for (i = 0; i < q->nr; i++) {
4475 p = q->queue[i];
4476 p->status = 0; /* undecided */
4477 if (DIFF_PAIR_UNMERGED(p))
4478 p->status = DIFF_STATUS_UNMERGED;
4479 else if (!DIFF_FILE_VALID(p->one))
4480 p->status = DIFF_STATUS_ADDED;
4481 else if (!DIFF_FILE_VALID(p->two))
4482 p->status = DIFF_STATUS_DELETED;
4483 else if (DIFF_PAIR_TYPE_CHANGED(p))
4484 p->status = DIFF_STATUS_TYPE_CHANGED;
4485
4486 /* from this point on, we are dealing with a pair
4487 * whose both sides are valid and of the same type, i.e.
4488 * either in-place edit or rename/copy edit.
4489 */
4490 else if (DIFF_PAIR_RENAME(p)) {
4491 /*
4492 * A rename might have re-connected a broken
4493 * pair up, causing the pathnames to be the
4494 * same again. If so, that's not a rename at
4495 * all, just a modification..
4496 *
4497 * Otherwise, see if this source was used for
4498 * multiple renames, in which case we decrement
4499 * the count, and call it a copy.
4500 */
4501 if (!strcmp(p->one->path, p->two->path))
4502 p->status = DIFF_STATUS_MODIFIED;
4503 else if (--p->one->rename_used > 0)
4504 p->status = DIFF_STATUS_COPIED;
4505 else
4506 p->status = DIFF_STATUS_RENAMED;
4507 }
4508 else if (oidcmp(&p->one->oid, &p->two->oid) ||
4509 p->one->mode != p->two->mode ||
4510 p->one->dirty_submodule ||
4511 p->two->dirty_submodule ||
4512 is_null_oid(&p->one->oid))
4513 p->status = DIFF_STATUS_MODIFIED;
4514 else {
4515 /* This is a "no-change" entry and should not
4516 * happen anymore, but prepare for broken callers.
4517 */
4518 error("feeding unmodified %s to diffcore",
4519 p->one->path);
4520 p->status = DIFF_STATUS_UNKNOWN;
4521 }
4522 }
4523 diff_debug_queue("resolve-rename-copy done", q);
4524 }
4525
4526 static int check_pair_status(struct diff_filepair *p)
4527 {
4528 switch (p->status) {
4529 case DIFF_STATUS_UNKNOWN:
4530 return 0;
4531 case 0:
4532 die("internal error in diff-resolve-rename-copy");
4533 default:
4534 return 1;
4535 }
4536 }
4537
4538 static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
4539 {
4540 int fmt = opt->output_format;
4541
4542 if (fmt & DIFF_FORMAT_CHECKDIFF)
4543 diff_flush_checkdiff(p, opt);
4544 else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
4545 diff_flush_raw(p, opt);
4546 else if (fmt & DIFF_FORMAT_NAME) {
4547 const char *name_a, *name_b;
4548 name_a = p->two->path;
4549 name_b = NULL;
4550 strip_prefix(opt->prefix_length, &name_a, &name_b);
4551 fprintf(opt->file, "%s", diff_line_prefix(opt));
4552 write_name_quoted(name_a, opt->file, opt->line_termination);
4553 }
4554 }
4555
4556 static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
4557 {
4558 if (fs->mode)
4559 fprintf(file, " %s mode %06o ", newdelete, fs->mode);
4560 else
4561 fprintf(file, " %s ", newdelete);
4562 write_name_quoted(fs->path, file, '\n');
4563 }
4564
4565
4566 static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
4567 const char *line_prefix)
4568 {
4569 if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
4570 fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
4571 p->two->mode, show_name ? ' ' : '\n');
4572 if (show_name) {
4573 write_name_quoted(p->two->path, file, '\n');
4574 }
4575 }
4576 }
4577
4578 static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
4579 const char *line_prefix)
4580 {
4581 char *names = pprint_rename(p->one->path, p->two->path);
4582
4583 fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
4584 free(names);
4585 show_mode_change(file, p, 0, line_prefix);
4586 }
4587
4588 static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
4589 {
4590 FILE *file = opt->file;
4591 const char *line_prefix = diff_line_prefix(opt);
4592
4593 switch(p->status) {
4594 case DIFF_STATUS_DELETED:
4595 fputs(line_prefix, file);
4596 show_file_mode_name(file, "delete", p->one);
4597 break;
4598 case DIFF_STATUS_ADDED:
4599 fputs(line_prefix, file);
4600 show_file_mode_name(file, "create", p->two);
4601 break;
4602 case DIFF_STATUS_COPIED:
4603 fputs(line_prefix, file);
4604 show_rename_copy(file, "copy", p, line_prefix);
4605 break;
4606 case DIFF_STATUS_RENAMED:
4607 fputs(line_prefix, file);
4608 show_rename_copy(file, "rename", p, line_prefix);
4609 break;
4610 default:
4611 if (p->score) {
4612 fprintf(file, "%s rewrite ", line_prefix);
4613 write_name_quoted(p->two->path, file, ' ');
4614 fprintf(file, "(%d%%)\n", similarity_index(p));
4615 }
4616 show_mode_change(file, p, !p->score, line_prefix);
4617 break;
4618 }
4619 }
4620
4621 struct patch_id_t {
4622 git_SHA_CTX *ctx;
4623 int patchlen;
4624 };
4625
4626 static int remove_space(char *line, int len)
4627 {
4628 int i;
4629 char *dst = line;
4630 unsigned char c;
4631
4632 for (i = 0; i < len; i++)
4633 if (!isspace((c = line[i])))
4634 *dst++ = c;
4635
4636 return dst - line;
4637 }
4638
4639 static void patch_id_consume(void *priv, char *line, unsigned long len)
4640 {
4641 struct patch_id_t *data = priv;
4642 int new_len;
4643
4644 /* Ignore line numbers when computing the SHA1 of the patch */
4645 if (starts_with(line, "@@ -"))
4646 return;
4647
4648 new_len = remove_space(line, len);
4649
4650 git_SHA1_Update(data->ctx, line, new_len);
4651 data->patchlen += new_len;
4652 }
4653
4654 static void patch_id_add_string(git_SHA_CTX *ctx, const char *str)
4655 {
4656 git_SHA1_Update(ctx, str, strlen(str));
4657 }
4658
4659 static void patch_id_add_mode(git_SHA_CTX *ctx, unsigned mode)
4660 {
4661 /* large enough for 2^32 in octal */
4662 char buf[12];
4663 int len = xsnprintf(buf, sizeof(buf), "%06o", mode);
4664 git_SHA1_Update(ctx, buf, len);
4665 }
4666
4667 /* returns 0 upon success, and writes result into sha1 */
4668 static int diff_get_patch_id(struct diff_options *options, struct object_id *oid, int diff_header_only)
4669 {
4670 struct diff_queue_struct *q = &diff_queued_diff;
4671 int i;
4672 git_SHA_CTX ctx;
4673 struct patch_id_t data;
4674
4675 git_SHA1_Init(&ctx);
4676 memset(&data, 0, sizeof(struct patch_id_t));
4677 data.ctx = &ctx;
4678
4679 for (i = 0; i < q->nr; i++) {
4680 xpparam_t xpp;
4681 xdemitconf_t xecfg;
4682 mmfile_t mf1, mf2;
4683 struct diff_filepair *p = q->queue[i];
4684 int len1, len2;
4685
4686 memset(&xpp, 0, sizeof(xpp));
4687 memset(&xecfg, 0, sizeof(xecfg));
4688 if (p->status == 0)
4689 return error("internal diff status error");
4690 if (p->status == DIFF_STATUS_UNKNOWN)
4691 continue;
4692 if (diff_unmodified_pair(p))
4693 continue;
4694 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4695 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4696 continue;
4697 if (DIFF_PAIR_UNMERGED(p))
4698 continue;
4699
4700 diff_fill_oid_info(p->one);
4701 diff_fill_oid_info(p->two);
4702
4703 len1 = remove_space(p->one->path, strlen(p->one->path));
4704 len2 = remove_space(p->two->path, strlen(p->two->path));
4705 patch_id_add_string(&ctx, "diff--git");
4706 patch_id_add_string(&ctx, "a/");
4707 git_SHA1_Update(&ctx, p->one->path, len1);
4708 patch_id_add_string(&ctx, "b/");
4709 git_SHA1_Update(&ctx, p->two->path, len2);
4710
4711 if (p->one->mode == 0) {
4712 patch_id_add_string(&ctx, "newfilemode");
4713 patch_id_add_mode(&ctx, p->two->mode);
4714 patch_id_add_string(&ctx, "---/dev/null");
4715 patch_id_add_string(&ctx, "+++b/");
4716 git_SHA1_Update(&ctx, p->two->path, len2);
4717 } else if (p->two->mode == 0) {
4718 patch_id_add_string(&ctx, "deletedfilemode");
4719 patch_id_add_mode(&ctx, p->one->mode);
4720 patch_id_add_string(&ctx, "---a/");
4721 git_SHA1_Update(&ctx, p->one->path, len1);
4722 patch_id_add_string(&ctx, "+++/dev/null");
4723 } else {
4724 patch_id_add_string(&ctx, "---a/");
4725 git_SHA1_Update(&ctx, p->one->path, len1);
4726 patch_id_add_string(&ctx, "+++b/");
4727 git_SHA1_Update(&ctx, p->two->path, len2);
4728 }
4729
4730 if (diff_header_only)
4731 continue;
4732
4733 if (fill_mmfile(&mf1, p->one) < 0 ||
4734 fill_mmfile(&mf2, p->two) < 0)
4735 return error("unable to read files to diff");
4736
4737 if (diff_filespec_is_binary(p->one) ||
4738 diff_filespec_is_binary(p->two)) {
4739 git_SHA1_Update(&ctx, oid_to_hex(&p->one->oid),
4740 GIT_SHA1_HEXSZ);
4741 git_SHA1_Update(&ctx, oid_to_hex(&p->two->oid),
4742 GIT_SHA1_HEXSZ);
4743 continue;
4744 }
4745
4746 xpp.flags = 0;
4747 xecfg.ctxlen = 3;
4748 xecfg.flags = 0;
4749 if (xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4750 &xpp, &xecfg))
4751 return error("unable to generate patch-id diff for %s",
4752 p->one->path);
4753 }
4754
4755 git_SHA1_Final(oid->hash, &ctx);
4756 return 0;
4757 }
4758
4759 int diff_flush_patch_id(struct diff_options *options, struct object_id *oid, int diff_header_only)
4760 {
4761 struct diff_queue_struct *q = &diff_queued_diff;
4762 int i;
4763 int result = diff_get_patch_id(options, oid, diff_header_only);
4764
4765 for (i = 0; i < q->nr; i++)
4766 diff_free_filepair(q->queue[i]);
4767
4768 free(q->queue);
4769 DIFF_QUEUE_CLEAR(q);
4770
4771 return result;
4772 }
4773
4774 static int is_summary_empty(const struct diff_queue_struct *q)
4775 {
4776 int i;
4777
4778 for (i = 0; i < q->nr; i++) {
4779 const struct diff_filepair *p = q->queue[i];
4780
4781 switch (p->status) {
4782 case DIFF_STATUS_DELETED:
4783 case DIFF_STATUS_ADDED:
4784 case DIFF_STATUS_COPIED:
4785 case DIFF_STATUS_RENAMED:
4786 return 0;
4787 default:
4788 if (p->score)
4789 return 0;
4790 if (p->one->mode && p->two->mode &&
4791 p->one->mode != p->two->mode)
4792 return 0;
4793 break;
4794 }
4795 }
4796 return 1;
4797 }
4798
4799 static const char rename_limit_warning[] =
4800 N_("inexact rename detection was skipped due to too many files.");
4801
4802 static const char degrade_cc_to_c_warning[] =
4803 N_("only found copies from modified paths due to too many files.");
4804
4805 static const char rename_limit_advice[] =
4806 N_("you may want to set your %s variable to at least "
4807 "%d and retry the command.");
4808
4809 void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4810 {
4811 if (degraded_cc)
4812 warning(_(degrade_cc_to_c_warning));
4813 else if (needed)
4814 warning(_(rename_limit_warning));
4815 else
4816 return;
4817 if (0 < needed && needed < 32767)
4818 warning(_(rename_limit_advice), varname, needed);
4819 }
4820
4821 static void diff_flush_patch_all_file_pairs(struct diff_options *o)
4822 {
4823 int i;
4824 struct diff_queue_struct *q = &diff_queued_diff;
4825
4826 if (WSEH_NEW & WS_RULE_MASK)
4827 die("BUG: WS rules bit mask overlaps with diff symbol flags");
4828
4829 for (i = 0; i < q->nr; i++) {
4830 struct diff_filepair *p = q->queue[i];
4831 if (check_pair_status(p))
4832 diff_flush_patch(p, o);
4833 }
4834 }
4835
4836 void diff_flush(struct diff_options *options)
4837 {
4838 struct diff_queue_struct *q = &diff_queued_diff;
4839 int i, output_format = options->output_format;
4840 int separator = 0;
4841 int dirstat_by_line = 0;
4842
4843 /*
4844 * Order: raw, stat, summary, patch
4845 * or: name/name-status/checkdiff (other bits clear)
4846 */
4847 if (!q->nr)
4848 goto free_queue;
4849
4850 if (output_format & (DIFF_FORMAT_RAW |
4851 DIFF_FORMAT_NAME |
4852 DIFF_FORMAT_NAME_STATUS |
4853 DIFF_FORMAT_CHECKDIFF)) {
4854 for (i = 0; i < q->nr; i++) {
4855 struct diff_filepair *p = q->queue[i];
4856 if (check_pair_status(p))
4857 flush_one_pair(p, options);
4858 }
4859 separator++;
4860 }
4861
4862 if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4863 dirstat_by_line = 1;
4864
4865 if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4866 dirstat_by_line) {
4867 struct diffstat_t diffstat;
4868
4869 memset(&diffstat, 0, sizeof(struct diffstat_t));
4870 for (i = 0; i < q->nr; i++) {
4871 struct diff_filepair *p = q->queue[i];
4872 if (check_pair_status(p))
4873 diff_flush_stat(p, options, &diffstat);
4874 }
4875 if (output_format & DIFF_FORMAT_NUMSTAT)
4876 show_numstat(&diffstat, options);
4877 if (output_format & DIFF_FORMAT_DIFFSTAT)
4878 show_stats(&diffstat, options);
4879 if (output_format & DIFF_FORMAT_SHORTSTAT)
4880 show_shortstats(&diffstat, options);
4881 if (output_format & DIFF_FORMAT_DIRSTAT && dirstat_by_line)
4882 show_dirstat_by_line(&diffstat, options);
4883 free_diffstat_info(&diffstat);
4884 separator++;
4885 }
4886 if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4887 show_dirstat(options);
4888
4889 if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4890 for (i = 0; i < q->nr; i++) {
4891 diff_summary(options, q->queue[i]);
4892 }
4893 separator++;
4894 }
4895
4896 if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4897 DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4898 DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4899 /*
4900 * run diff_flush_patch for the exit status. setting
4901 * options->file to /dev/null should be safe, because we
4902 * aren't supposed to produce any output anyway.
4903 */
4904 if (options->close_file)
4905 fclose(options->file);
4906 options->file = xfopen("/dev/null", "w");
4907 options->close_file = 1;
4908 for (i = 0; i < q->nr; i++) {
4909 struct diff_filepair *p = q->queue[i];
4910 if (check_pair_status(p))
4911 diff_flush_patch(p, options);
4912 if (options->found_changes)
4913 break;
4914 }
4915 }
4916
4917 if (output_format & DIFF_FORMAT_PATCH) {
4918 if (separator) {
4919 emit_diff_symbol(options, DIFF_SYMBOL_SEPARATOR, NULL, 0, 0);
4920 if (options->stat_sep) {
4921 /* attach patch instead of inline */
4922 fputs(options->stat_sep, options->file);
4923 }
4924 }
4925
4926 diff_flush_patch_all_file_pairs(options);
4927 }
4928
4929 if (output_format & DIFF_FORMAT_CALLBACK)
4930 options->format_callback(q, options, options->format_callback_data);
4931
4932 for (i = 0; i < q->nr; i++)
4933 diff_free_filepair(q->queue[i]);
4934 free_queue:
4935 free(q->queue);
4936 DIFF_QUEUE_CLEAR(q);
4937 if (options->close_file)
4938 fclose(options->file);
4939
4940 /*
4941 * Report the content-level differences with HAS_CHANGES;
4942 * diff_addremove/diff_change does not set the bit when
4943 * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4944 */
4945 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4946 if (options->found_changes)
4947 DIFF_OPT_SET(options, HAS_CHANGES);
4948 else
4949 DIFF_OPT_CLR(options, HAS_CHANGES);
4950 }
4951 }
4952
4953 static int match_filter(const struct diff_options *options, const struct diff_filepair *p)
4954 {
4955 return (((p->status == DIFF_STATUS_MODIFIED) &&
4956 ((p->score &&
4957 filter_bit_tst(DIFF_STATUS_FILTER_BROKEN, options)) ||
4958 (!p->score &&
4959 filter_bit_tst(DIFF_STATUS_MODIFIED, options)))) ||
4960 ((p->status != DIFF_STATUS_MODIFIED) &&
4961 filter_bit_tst(p->status, options)));
4962 }
4963
4964 static void diffcore_apply_filter(struct diff_options *options)
4965 {
4966 int i;
4967 struct diff_queue_struct *q = &diff_queued_diff;
4968 struct diff_queue_struct outq;
4969
4970 DIFF_QUEUE_CLEAR(&outq);
4971
4972 if (!options->filter)
4973 return;
4974
4975 if (filter_bit_tst(DIFF_STATUS_FILTER_AON, options)) {
4976 int found;
4977 for (i = found = 0; !found && i < q->nr; i++) {
4978 if (match_filter(options, q->queue[i]))
4979 found++;
4980 }
4981 if (found)
4982 return;
4983
4984 /* otherwise we will clear the whole queue
4985 * by copying the empty outq at the end of this
4986 * function, but first clear the current entries
4987 * in the queue.
4988 */
4989 for (i = 0; i < q->nr; i++)
4990 diff_free_filepair(q->queue[i]);
4991 }
4992 else {
4993 /* Only the matching ones */
4994 for (i = 0; i < q->nr; i++) {
4995 struct diff_filepair *p = q->queue[i];
4996 if (match_filter(options, p))
4997 diff_q(&outq, p);
4998 else
4999 diff_free_filepair(p);
5000 }
5001 }
5002 free(q->queue);
5003 *q = outq;
5004 }
5005
5006 /* Check whether two filespecs with the same mode and size are identical */
5007 static int diff_filespec_is_identical(struct diff_filespec *one,
5008 struct diff_filespec *two)
5009 {
5010 if (S_ISGITLINK(one->mode))
5011 return 0;
5012 if (diff_populate_filespec(one, 0))
5013 return 0;
5014 if (diff_populate_filespec(two, 0))
5015 return 0;
5016 return !memcmp(one->data, two->data, one->size);
5017 }
5018
5019 static int diff_filespec_check_stat_unmatch(struct diff_filepair *p)
5020 {
5021 if (p->done_skip_stat_unmatch)
5022 return p->skip_stat_unmatch_result;
5023
5024 p->done_skip_stat_unmatch = 1;
5025 p->skip_stat_unmatch_result = 0;
5026 /*
5027 * 1. Entries that come from stat info dirtiness
5028 * always have both sides (iow, not create/delete),
5029 * one side of the object name is unknown, with
5030 * the same mode and size. Keep the ones that
5031 * do not match these criteria. They have real
5032 * differences.
5033 *
5034 * 2. At this point, the file is known to be modified,
5035 * with the same mode and size, and the object
5036 * name of one side is unknown. Need to inspect
5037 * the identical contents.
5038 */
5039 if (!DIFF_FILE_VALID(p->one) || /* (1) */
5040 !DIFF_FILE_VALID(p->two) ||
5041 (p->one->oid_valid && p->two->oid_valid) ||
5042 (p->one->mode != p->two->mode) ||
5043 diff_populate_filespec(p->one, CHECK_SIZE_ONLY) ||
5044 diff_populate_filespec(p->two, CHECK_SIZE_ONLY) ||
5045 (p->one->size != p->two->size) ||
5046 !diff_filespec_is_identical(p->one, p->two)) /* (2) */
5047 p->skip_stat_unmatch_result = 1;
5048 return p->skip_stat_unmatch_result;
5049 }
5050
5051 static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
5052 {
5053 int i;
5054 struct diff_queue_struct *q = &diff_queued_diff;
5055 struct diff_queue_struct outq;
5056 DIFF_QUEUE_CLEAR(&outq);
5057
5058 for (i = 0; i < q->nr; i++) {
5059 struct diff_filepair *p = q->queue[i];
5060
5061 if (diff_filespec_check_stat_unmatch(p))
5062 diff_q(&outq, p);
5063 else {
5064 /*
5065 * The caller can subtract 1 from skip_stat_unmatch
5066 * to determine how many paths were dirty only
5067 * due to stat info mismatch.
5068 */
5069 if (!DIFF_OPT_TST(diffopt, NO_INDEX))
5070 diffopt->skip_stat_unmatch++;
5071 diff_free_filepair(p);
5072 }
5073 }
5074 free(q->queue);
5075 *q = outq;
5076 }
5077
5078 static int diffnamecmp(const void *a_, const void *b_)
5079 {
5080 const struct diff_filepair *a = *((const struct diff_filepair **)a_);
5081 const struct diff_filepair *b = *((const struct diff_filepair **)b_);
5082 const char *name_a, *name_b;
5083
5084 name_a = a->one ? a->one->path : a->two->path;
5085 name_b = b->one ? b->one->path : b->two->path;
5086 return strcmp(name_a, name_b);
5087 }
5088
5089 void diffcore_fix_diff_index(struct diff_options *options)
5090 {
5091 struct diff_queue_struct *q = &diff_queued_diff;
5092 QSORT(q->queue, q->nr, diffnamecmp);
5093 }
5094
5095 void diffcore_std(struct diff_options *options)
5096 {
5097 /* NOTE please keep the following in sync with diff_tree_combined() */
5098 if (options->skip_stat_unmatch)
5099 diffcore_skip_stat_unmatch(options);
5100 if (!options->found_follow) {
5101 /* See try_to_follow_renames() in tree-diff.c */
5102 if (options->break_opt != -1)
5103 diffcore_break(options->break_opt);
5104 if (options->detect_rename)
5105 diffcore_rename(options);
5106 if (options->break_opt != -1)
5107 diffcore_merge_broken();
5108 }
5109 if (options->pickaxe)
5110 diffcore_pickaxe(options);
5111 if (options->orderfile)
5112 diffcore_order(options->orderfile);
5113 if (!options->found_follow)
5114 /* See try_to_follow_renames() in tree-diff.c */
5115 diff_resolve_rename_copy();
5116 diffcore_apply_filter(options);
5117
5118 if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5119 DIFF_OPT_SET(options, HAS_CHANGES);
5120 else
5121 DIFF_OPT_CLR(options, HAS_CHANGES);
5122
5123 options->found_follow = 0;
5124 }
5125
5126 int diff_result_code(struct diff_options *opt, int status)
5127 {
5128 int result = 0;
5129
5130 diff_warn_rename_limit("diff.renameLimit",
5131 opt->needed_rename_limit,
5132 opt->degraded_cc_to_c);
5133 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5134 !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
5135 return status;
5136 if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5137 DIFF_OPT_TST(opt, HAS_CHANGES))
5138 result |= 01;
5139 if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
5140 DIFF_OPT_TST(opt, CHECK_FAILED))
5141 result |= 02;
5142 return result;
5143 }
5144
5145 int diff_can_quit_early(struct diff_options *opt)
5146 {
5147 return (DIFF_OPT_TST(opt, QUICK) &&
5148 !opt->filter &&
5149 DIFF_OPT_TST(opt, HAS_CHANGES));
5150 }
5151
5152 /*
5153 * Shall changes to this submodule be ignored?
5154 *
5155 * Submodule changes can be configured to be ignored separately for each path,
5156 * but that configuration can be overridden from the command line.
5157 */
5158 static int is_submodule_ignored(const char *path, struct diff_options *options)
5159 {
5160 int ignored = 0;
5161 unsigned orig_flags = options->flags;
5162 if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
5163 set_diffopt_flags_from_submodule_config(options, path);
5164 if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
5165 ignored = 1;
5166 options->flags = orig_flags;
5167 return ignored;
5168 }
5169
5170 void diff_addremove(struct diff_options *options,
5171 int addremove, unsigned mode,
5172 const struct object_id *oid,
5173 int oid_valid,
5174 const char *concatpath, unsigned dirty_submodule)
5175 {
5176 struct diff_filespec *one, *two;
5177
5178 if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
5179 return;
5180
5181 /* This may look odd, but it is a preparation for
5182 * feeding "there are unchanged files which should
5183 * not produce diffs, but when you are doing copy
5184 * detection you would need them, so here they are"
5185 * entries to the diff-core. They will be prefixed
5186 * with something like '=' or '*' (I haven't decided
5187 * which but should not make any difference).
5188 * Feeding the same new and old to diff_change()
5189 * also has the same effect.
5190 * Before the final output happens, they are pruned after
5191 * merged into rename/copy pairs as appropriate.
5192 */
5193 if (DIFF_OPT_TST(options, REVERSE_DIFF))
5194 addremove = (addremove == '+' ? '-' :
5195 addremove == '-' ? '+' : addremove);
5196
5197 if (options->prefix &&
5198 strncmp(concatpath, options->prefix, options->prefix_length))
5199 return;
5200
5201 one = alloc_filespec(concatpath);
5202 two = alloc_filespec(concatpath);
5203
5204 if (addremove != '+')
5205 fill_filespec(one, oid, oid_valid, mode);
5206 if (addremove != '-') {
5207 fill_filespec(two, oid, oid_valid, mode);
5208 two->dirty_submodule = dirty_submodule;
5209 }
5210
5211 diff_queue(&diff_queued_diff, one, two);
5212 if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5213 DIFF_OPT_SET(options, HAS_CHANGES);
5214 }
5215
5216 void diff_change(struct diff_options *options,
5217 unsigned old_mode, unsigned new_mode,
5218 const struct object_id *old_oid,
5219 const struct object_id *new_oid,
5220 int old_oid_valid, int new_oid_valid,
5221 const char *concatpath,
5222 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
5223 {
5224 struct diff_filespec *one, *two;
5225 struct diff_filepair *p;
5226
5227 if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
5228 is_submodule_ignored(concatpath, options))
5229 return;
5230
5231 if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
5232 SWAP(old_mode, new_mode);
5233 SWAP(old_oid, new_oid);
5234 SWAP(old_oid_valid, new_oid_valid);
5235 SWAP(old_dirty_submodule, new_dirty_submodule);
5236 }
5237
5238 if (options->prefix &&
5239 strncmp(concatpath, options->prefix, options->prefix_length))
5240 return;
5241
5242 one = alloc_filespec(concatpath);
5243 two = alloc_filespec(concatpath);
5244 fill_filespec(one, old_oid, old_oid_valid, old_mode);
5245 fill_filespec(two, new_oid, new_oid_valid, new_mode);
5246 one->dirty_submodule = old_dirty_submodule;
5247 two->dirty_submodule = new_dirty_submodule;
5248 p = diff_queue(&diff_queued_diff, one, two);
5249
5250 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5251 return;
5252
5253 if (DIFF_OPT_TST(options, QUICK) && options->skip_stat_unmatch &&
5254 !diff_filespec_check_stat_unmatch(p))
5255 return;
5256
5257 DIFF_OPT_SET(options, HAS_CHANGES);
5258 }
5259
5260 struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
5261 {
5262 struct diff_filepair *pair;
5263 struct diff_filespec *one, *two;
5264
5265 if (options->prefix &&
5266 strncmp(path, options->prefix, options->prefix_length))
5267 return NULL;
5268
5269 one = alloc_filespec(path);
5270 two = alloc_filespec(path);
5271 pair = diff_queue(&diff_queued_diff, one, two);
5272 pair->is_unmerged = 1;
5273 return pair;
5274 }
5275
5276 static char *run_textconv(const char *pgm, struct diff_filespec *spec,
5277 size_t *outsize)
5278 {
5279 struct diff_tempfile *temp;
5280 const char *argv[3];
5281 const char **arg = argv;
5282 struct child_process child = CHILD_PROCESS_INIT;
5283 struct strbuf buf = STRBUF_INIT;
5284 int err = 0;
5285
5286 temp = prepare_temp_file(spec->path, spec);
5287 *arg++ = pgm;
5288 *arg++ = temp->name;
5289 *arg = NULL;
5290
5291 child.use_shell = 1;
5292 child.argv = argv;
5293 child.out = -1;
5294 if (start_command(&child)) {
5295 remove_tempfile();
5296 return NULL;
5297 }
5298
5299 if (strbuf_read(&buf, child.out, 0) < 0)
5300 err = error("error reading from textconv command '%s'", pgm);
5301 close(child.out);
5302
5303 if (finish_command(&child) || err) {
5304 strbuf_release(&buf);
5305 remove_tempfile();
5306 return NULL;
5307 }
5308 remove_tempfile();
5309
5310 return strbuf_detach(&buf, outsize);
5311 }
5312
5313 size_t fill_textconv(struct userdiff_driver *driver,
5314 struct diff_filespec *df,
5315 char **outbuf)
5316 {
5317 size_t size;
5318
5319 if (!driver) {
5320 if (!DIFF_FILE_VALID(df)) {
5321 *outbuf = "";
5322 return 0;
5323 }
5324 if (diff_populate_filespec(df, 0))
5325 die("unable to read files to diff");
5326 *outbuf = df->data;
5327 return df->size;
5328 }
5329
5330 if (!driver->textconv)
5331 die("BUG: fill_textconv called with non-textconv driver");
5332
5333 if (driver->textconv_cache && df->oid_valid) {
5334 *outbuf = notes_cache_get(driver->textconv_cache,
5335 &df->oid,
5336 &size);
5337 if (*outbuf)
5338 return size;
5339 }
5340
5341 *outbuf = run_textconv(driver->textconv, df, &size);
5342 if (!*outbuf)
5343 die("unable to read files to diff");
5344
5345 if (driver->textconv_cache && df->oid_valid) {
5346 /* ignore errors, as we might be in a readonly repository */
5347 notes_cache_put(driver->textconv_cache, &df->oid, *outbuf,
5348 size);
5349 /*
5350 * we could save up changes and flush them all at the end,
5351 * but we would need an extra call after all diffing is done.
5352 * Since generating a cache entry is the slow path anyway,
5353 * this extra overhead probably isn't a big deal.
5354 */
5355 notes_cache_write(driver->textconv_cache);
5356 }
5357
5358 return size;
5359 }
5360
5361 int textconv_object(const char *path,
5362 unsigned mode,
5363 const struct object_id *oid,
5364 int oid_valid,
5365 char **buf,
5366 unsigned long *buf_size)
5367 {
5368 struct diff_filespec *df;
5369 struct userdiff_driver *textconv;
5370
5371 df = alloc_filespec(path);
5372 fill_filespec(df, oid, oid_valid, mode);
5373 textconv = get_textconv(df);
5374 if (!textconv) {
5375 free_filespec(df);
5376 return 0;
5377 }
5378
5379 *buf_size = fill_textconv(textconv, df, buf);
5380 free_filespec(df);
5381 return 1;
5382 }
5383
5384 void setup_diff_pager(struct diff_options *opt)
5385 {
5386 /*
5387 * If the user asked for our exit code, then either they want --quiet
5388 * or --exit-code. We should definitely not bother with a pager in the
5389 * former case, as we will generate no output. Since we still properly
5390 * report our exit code even when a pager is run, we _could_ run a
5391 * pager with --exit-code. But since we have not done so historically,
5392 * and because it is easy to find people oneline advising "git diff
5393 * --exit-code" in hooks and other scripts, we do not do so.
5394 */
5395 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5396 check_pager_config("diff") != 0)
5397 setup_pager();
5398 }
The Best VCS Around