Merge branch 'nd/no-the-index'
[git/git.git] / diff.c
1 /*
2 * Copyright (C) 2005 Junio C Hamano
3 */
4 #include "cache.h"
5 #include "config.h"
6 #include "tempfile.h"
7 #include "quote.h"
8 #include "diff.h"
9 #include "diffcore.h"
10 #include "delta.h"
11 #include "xdiff-interface.h"
12 #include "color.h"
13 #include "attr.h"
14 #include "run-command.h"
15 #include "utf8.h"
16 #include "object-store.h"
17 #include "userdiff.h"
18 #include "submodule-config.h"
19 #include "submodule.h"
20 #include "hashmap.h"
21 #include "ll-merge.h"
22 #include "string-list.h"
23 #include "argv-array.h"
24 #include "graph.h"
25 #include "packfile.h"
26 #include "help.h"
27
28 #ifdef NO_FAST_WORKING_DIRECTORY
29 #define FAST_WORKING_DIRECTORY 0
30 #else
31 #define FAST_WORKING_DIRECTORY 1
32 #endif
33
34 static int diff_detect_rename_default;
35 static int diff_indent_heuristic = 1;
36 static int diff_rename_limit_default = 400;
37 static int diff_suppress_blank_empty;
38 static int diff_use_color_default = -1;
39 static int diff_color_moved_default;
40 static int diff_color_moved_ws_default;
41 static int diff_context_default = 3;
42 static int diff_interhunk_context_default;
43 static const char *diff_word_regex_cfg;
44 static const char *external_diff_cmd_cfg;
45 static const char *diff_order_file_cfg;
46 int diff_auto_refresh_index = 1;
47 static int diff_mnemonic_prefix;
48 static int diff_no_prefix;
49 static int diff_stat_graph_width;
50 static int diff_dirstat_permille_default = 30;
51 static struct diff_options default_diff_options;
52 static long diff_algorithm;
53 static unsigned ws_error_highlight_default = WSEH_NEW;
54
55 static char diff_colors[][COLOR_MAXLEN] = {
56 GIT_COLOR_RESET,
57 GIT_COLOR_NORMAL, /* CONTEXT */
58 GIT_COLOR_BOLD, /* METAINFO */
59 GIT_COLOR_CYAN, /* FRAGINFO */
60 GIT_COLOR_RED, /* OLD */
61 GIT_COLOR_GREEN, /* NEW */
62 GIT_COLOR_YELLOW, /* COMMIT */
63 GIT_COLOR_BG_RED, /* WHITESPACE */
64 GIT_COLOR_NORMAL, /* FUNCINFO */
65 GIT_COLOR_BOLD_MAGENTA, /* OLD_MOVED */
66 GIT_COLOR_BOLD_BLUE, /* OLD_MOVED ALTERNATIVE */
67 GIT_COLOR_FAINT, /* OLD_MOVED_DIM */
68 GIT_COLOR_FAINT_ITALIC, /* OLD_MOVED_ALTERNATIVE_DIM */
69 GIT_COLOR_BOLD_CYAN, /* NEW_MOVED */
70 GIT_COLOR_BOLD_YELLOW, /* NEW_MOVED ALTERNATIVE */
71 GIT_COLOR_FAINT, /* NEW_MOVED_DIM */
72 GIT_COLOR_FAINT_ITALIC, /* NEW_MOVED_ALTERNATIVE_DIM */
73 };
74
75 static const char *color_diff_slots[] = {
76 [DIFF_CONTEXT] = "context",
77 [DIFF_METAINFO] = "meta",
78 [DIFF_FRAGINFO] = "frag",
79 [DIFF_FILE_OLD] = "old",
80 [DIFF_FILE_NEW] = "new",
81 [DIFF_COMMIT] = "commit",
82 [DIFF_WHITESPACE] = "whitespace",
83 [DIFF_FUNCINFO] = "func",
84 [DIFF_FILE_OLD_MOVED] = "oldMoved",
85 [DIFF_FILE_OLD_MOVED_ALT] = "oldMovedAlternative",
86 [DIFF_FILE_OLD_MOVED_DIM] = "oldMovedDimmed",
87 [DIFF_FILE_OLD_MOVED_ALT_DIM] = "oldMovedAlternativeDimmed",
88 [DIFF_FILE_NEW_MOVED] = "newMoved",
89 [DIFF_FILE_NEW_MOVED_ALT] = "newMovedAlternative",
90 [DIFF_FILE_NEW_MOVED_DIM] = "newMovedDimmed",
91 [DIFF_FILE_NEW_MOVED_ALT_DIM] = "newMovedAlternativeDimmed",
92 };
93
94 static NORETURN void die_want_option(const char *option_name)
95 {
96 die(_("option '%s' requires a value"), option_name);
97 }
98
99 define_list_config_array_extra(color_diff_slots, {"plain"});
100
101 static int parse_diff_color_slot(const char *var)
102 {
103 if (!strcasecmp(var, "plain"))
104 return DIFF_CONTEXT;
105 return LOOKUP_CONFIG(color_diff_slots, var);
106 }
107
108 static int parse_dirstat_params(struct diff_options *options, const char *params_string,
109 struct strbuf *errmsg)
110 {
111 char *params_copy = xstrdup(params_string);
112 struct string_list params = STRING_LIST_INIT_NODUP;
113 int ret = 0;
114 int i;
115
116 if (*params_copy)
117 string_list_split_in_place(&params, params_copy, ',', -1);
118 for (i = 0; i < params.nr; i++) {
119 const char *p = params.items[i].string;
120 if (!strcmp(p, "changes")) {
121 options->flags.dirstat_by_line = 0;
122 options->flags.dirstat_by_file = 0;
123 } else if (!strcmp(p, "lines")) {
124 options->flags.dirstat_by_line = 1;
125 options->flags.dirstat_by_file = 0;
126 } else if (!strcmp(p, "files")) {
127 options->flags.dirstat_by_line = 0;
128 options->flags.dirstat_by_file = 1;
129 } else if (!strcmp(p, "noncumulative")) {
130 options->flags.dirstat_cumulative = 0;
131 } else if (!strcmp(p, "cumulative")) {
132 options->flags.dirstat_cumulative = 1;
133 } else if (isdigit(*p)) {
134 char *end;
135 int permille = strtoul(p, &end, 10) * 10;
136 if (*end == '.' && isdigit(*++end)) {
137 /* only use first digit */
138 permille += *end - '0';
139 /* .. and ignore any further digits */
140 while (isdigit(*++end))
141 ; /* nothing */
142 }
143 if (!*end)
144 options->dirstat_permille = permille;
145 else {
146 strbuf_addf(errmsg, _(" Failed to parse dirstat cut-off percentage '%s'\n"),
147 p);
148 ret++;
149 }
150 } else {
151 strbuf_addf(errmsg, _(" Unknown dirstat parameter '%s'\n"), p);
152 ret++;
153 }
154
155 }
156 string_list_clear(&params, 0);
157 free(params_copy);
158 return ret;
159 }
160
161 static int parse_submodule_params(struct diff_options *options, const char *value)
162 {
163 if (!strcmp(value, "log"))
164 options->submodule_format = DIFF_SUBMODULE_LOG;
165 else if (!strcmp(value, "short"))
166 options->submodule_format = DIFF_SUBMODULE_SHORT;
167 else if (!strcmp(value, "diff"))
168 options->submodule_format = DIFF_SUBMODULE_INLINE_DIFF;
169 else
170 return -1;
171 return 0;
172 }
173
174 int git_config_rename(const char *var, const char *value)
175 {
176 if (!value)
177 return DIFF_DETECT_RENAME;
178 if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
179 return DIFF_DETECT_COPY;
180 return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
181 }
182
183 long parse_algorithm_value(const char *value)
184 {
185 if (!value)
186 return -1;
187 else if (!strcasecmp(value, "myers") || !strcasecmp(value, "default"))
188 return 0;
189 else if (!strcasecmp(value, "minimal"))
190 return XDF_NEED_MINIMAL;
191 else if (!strcasecmp(value, "patience"))
192 return XDF_PATIENCE_DIFF;
193 else if (!strcasecmp(value, "histogram"))
194 return XDF_HISTOGRAM_DIFF;
195 return -1;
196 }
197
198 static int parse_one_token(const char **arg, const char *token)
199 {
200 const char *rest;
201 if (skip_prefix(*arg, token, &rest) && (!*rest || *rest == ',')) {
202 *arg = rest;
203 return 1;
204 }
205 return 0;
206 }
207
208 static int parse_ws_error_highlight(const char *arg)
209 {
210 const char *orig_arg = arg;
211 unsigned val = 0;
212
213 while (*arg) {
214 if (parse_one_token(&arg, "none"))
215 val = 0;
216 else if (parse_one_token(&arg, "default"))
217 val = WSEH_NEW;
218 else if (parse_one_token(&arg, "all"))
219 val = WSEH_NEW | WSEH_OLD | WSEH_CONTEXT;
220 else if (parse_one_token(&arg, "new"))
221 val |= WSEH_NEW;
222 else if (parse_one_token(&arg, "old"))
223 val |= WSEH_OLD;
224 else if (parse_one_token(&arg, "context"))
225 val |= WSEH_CONTEXT;
226 else {
227 return -1 - (int)(arg - orig_arg);
228 }
229 if (*arg)
230 arg++;
231 }
232 return val;
233 }
234
235 /*
236 * These are to give UI layer defaults.
237 * The core-level commands such as git-diff-files should
238 * never be affected by the setting of diff.renames
239 * the user happens to have in the configuration file.
240 */
241 void init_diff_ui_defaults(void)
242 {
243 diff_detect_rename_default = DIFF_DETECT_RENAME;
244 }
245
246 int git_diff_heuristic_config(const char *var, const char *value, void *cb)
247 {
248 if (!strcmp(var, "diff.indentheuristic"))
249 diff_indent_heuristic = git_config_bool(var, value);
250 return 0;
251 }
252
253 static int parse_color_moved(const char *arg)
254 {
255 switch (git_parse_maybe_bool(arg)) {
256 case 0:
257 return COLOR_MOVED_NO;
258 case 1:
259 return COLOR_MOVED_DEFAULT;
260 default:
261 break;
262 }
263
264 if (!strcmp(arg, "no"))
265 return COLOR_MOVED_NO;
266 else if (!strcmp(arg, "plain"))
267 return COLOR_MOVED_PLAIN;
268 else if (!strcmp(arg, "blocks"))
269 return COLOR_MOVED_BLOCKS;
270 else if (!strcmp(arg, "zebra"))
271 return COLOR_MOVED_ZEBRA;
272 else if (!strcmp(arg, "default"))
273 return COLOR_MOVED_DEFAULT;
274 else if (!strcmp(arg, "dimmed-zebra"))
275 return COLOR_MOVED_ZEBRA_DIM;
276 else if (!strcmp(arg, "dimmed_zebra"))
277 return COLOR_MOVED_ZEBRA_DIM;
278 else
279 return error(_("color moved setting must be one of 'no', 'default', 'blocks', 'zebra', 'dimmed-zebra', 'plain'"));
280 }
281
282 static int parse_color_moved_ws(const char *arg)
283 {
284 int ret = 0;
285 struct string_list l = STRING_LIST_INIT_DUP;
286 struct string_list_item *i;
287
288 string_list_split(&l, arg, ',', -1);
289
290 for_each_string_list_item(i, &l) {
291 struct strbuf sb = STRBUF_INIT;
292 strbuf_addstr(&sb, i->string);
293 strbuf_trim(&sb);
294
295 if (!strcmp(sb.buf, "ignore-space-change"))
296 ret |= XDF_IGNORE_WHITESPACE_CHANGE;
297 else if (!strcmp(sb.buf, "ignore-space-at-eol"))
298 ret |= XDF_IGNORE_WHITESPACE_AT_EOL;
299 else if (!strcmp(sb.buf, "ignore-all-space"))
300 ret |= XDF_IGNORE_WHITESPACE;
301 else if (!strcmp(sb.buf, "allow-indentation-change"))
302 ret |= COLOR_MOVED_WS_ALLOW_INDENTATION_CHANGE;
303 else
304 error(_("ignoring unknown color-moved-ws mode '%s'"), sb.buf);
305
306 strbuf_release(&sb);
307 }
308
309 if ((ret & COLOR_MOVED_WS_ALLOW_INDENTATION_CHANGE) &&
310 (ret & XDF_WHITESPACE_FLAGS))
311 die(_("color-moved-ws: allow-indentation-change cannot be combined with other white space modes"));
312
313 string_list_clear(&l, 0);
314
315 return ret;
316 }
317
318 int git_diff_ui_config(const char *var, const char *value, void *cb)
319 {
320 if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
321 diff_use_color_default = git_config_colorbool(var, value);
322 return 0;
323 }
324 if (!strcmp(var, "diff.colormoved")) {
325 int cm = parse_color_moved(value);
326 if (cm < 0)
327 return -1;
328 diff_color_moved_default = cm;
329 return 0;
330 }
331 if (!strcmp(var, "diff.colormovedws")) {
332 int cm = parse_color_moved_ws(value);
333 if (cm < 0)
334 return -1;
335 diff_color_moved_ws_default = cm;
336 return 0;
337 }
338 if (!strcmp(var, "diff.context")) {
339 diff_context_default = git_config_int(var, value);
340 if (diff_context_default < 0)
341 return -1;
342 return 0;
343 }
344 if (!strcmp(var, "diff.interhunkcontext")) {
345 diff_interhunk_context_default = git_config_int(var, value);
346 if (diff_interhunk_context_default < 0)
347 return -1;
348 return 0;
349 }
350 if (!strcmp(var, "diff.renames")) {
351 diff_detect_rename_default = git_config_rename(var, value);
352 return 0;
353 }
354 if (!strcmp(var, "diff.autorefreshindex")) {
355 diff_auto_refresh_index = git_config_bool(var, value);
356 return 0;
357 }
358 if (!strcmp(var, "diff.mnemonicprefix")) {
359 diff_mnemonic_prefix = git_config_bool(var, value);
360 return 0;
361 }
362 if (!strcmp(var, "diff.noprefix")) {
363 diff_no_prefix = git_config_bool(var, value);
364 return 0;
365 }
366 if (!strcmp(var, "diff.statgraphwidth")) {
367 diff_stat_graph_width = git_config_int(var, value);
368 return 0;
369 }
370 if (!strcmp(var, "diff.external"))
371 return git_config_string(&external_diff_cmd_cfg, var, value);
372 if (!strcmp(var, "diff.wordregex"))
373 return git_config_string(&diff_word_regex_cfg, var, value);
374 if (!strcmp(var, "diff.orderfile"))
375 return git_config_pathname(&diff_order_file_cfg, var, value);
376
377 if (!strcmp(var, "diff.ignoresubmodules"))
378 handle_ignore_submodules_arg(&default_diff_options, value);
379
380 if (!strcmp(var, "diff.submodule")) {
381 if (parse_submodule_params(&default_diff_options, value))
382 warning(_("Unknown value for 'diff.submodule' config variable: '%s'"),
383 value);
384 return 0;
385 }
386
387 if (!strcmp(var, "diff.algorithm")) {
388 diff_algorithm = parse_algorithm_value(value);
389 if (diff_algorithm < 0)
390 return -1;
391 return 0;
392 }
393
394 if (!strcmp(var, "diff.wserrorhighlight")) {
395 int val = parse_ws_error_highlight(value);
396 if (val < 0)
397 return -1;
398 ws_error_highlight_default = val;
399 return 0;
400 }
401
402 if (git_color_config(var, value, cb) < 0)
403 return -1;
404
405 return git_diff_basic_config(var, value, cb);
406 }
407
408 int git_diff_basic_config(const char *var, const char *value, void *cb)
409 {
410 const char *name;
411
412 if (!strcmp(var, "diff.renamelimit")) {
413 diff_rename_limit_default = git_config_int(var, value);
414 return 0;
415 }
416
417 if (userdiff_config(var, value) < 0)
418 return -1;
419
420 if (skip_prefix(var, "diff.color.", &name) ||
421 skip_prefix(var, "color.diff.", &name)) {
422 int slot = parse_diff_color_slot(name);
423 if (slot < 0)
424 return 0;
425 if (!value)
426 return config_error_nonbool(var);
427 return color_parse(value, diff_colors[slot]);
428 }
429
430 /* like GNU diff's --suppress-blank-empty option */
431 if (!strcmp(var, "diff.suppressblankempty") ||
432 /* for backwards compatibility */
433 !strcmp(var, "diff.suppress-blank-empty")) {
434 diff_suppress_blank_empty = git_config_bool(var, value);
435 return 0;
436 }
437
438 if (!strcmp(var, "diff.dirstat")) {
439 struct strbuf errmsg = STRBUF_INIT;
440 default_diff_options.dirstat_permille = diff_dirstat_permille_default;
441 if (parse_dirstat_params(&default_diff_options, value, &errmsg))
442 warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
443 errmsg.buf);
444 strbuf_release(&errmsg);
445 diff_dirstat_permille_default = default_diff_options.dirstat_permille;
446 return 0;
447 }
448
449 if (git_diff_heuristic_config(var, value, cb) < 0)
450 return -1;
451
452 return git_default_config(var, value, cb);
453 }
454
455 static char *quote_two(const char *one, const char *two)
456 {
457 int need_one = quote_c_style(one, NULL, NULL, 1);
458 int need_two = quote_c_style(two, NULL, NULL, 1);
459 struct strbuf res = STRBUF_INIT;
460
461 if (need_one + need_two) {
462 strbuf_addch(&res, '"');
463 quote_c_style(one, &res, NULL, 1);
464 quote_c_style(two, &res, NULL, 1);
465 strbuf_addch(&res, '"');
466 } else {
467 strbuf_addstr(&res, one);
468 strbuf_addstr(&res, two);
469 }
470 return strbuf_detach(&res, NULL);
471 }
472
473 static const char *external_diff(void)
474 {
475 static const char *external_diff_cmd = NULL;
476 static int done_preparing = 0;
477
478 if (done_preparing)
479 return external_diff_cmd;
480 external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
481 if (!external_diff_cmd)
482 external_diff_cmd = external_diff_cmd_cfg;
483 done_preparing = 1;
484 return external_diff_cmd;
485 }
486
487 /*
488 * Keep track of files used for diffing. Sometimes such an entry
489 * refers to a temporary file, sometimes to an existing file, and
490 * sometimes to "/dev/null".
491 */
492 static struct diff_tempfile {
493 /*
494 * filename external diff should read from, or NULL if this
495 * entry is currently not in use:
496 */
497 const char *name;
498
499 char hex[GIT_MAX_HEXSZ + 1];
500 char mode[10];
501
502 /*
503 * If this diff_tempfile instance refers to a temporary file,
504 * this tempfile object is used to manage its lifetime.
505 */
506 struct tempfile *tempfile;
507 } diff_temp[2];
508
509 struct emit_callback {
510 int color_diff;
511 unsigned ws_rule;
512 int blank_at_eof_in_preimage;
513 int blank_at_eof_in_postimage;
514 int lno_in_preimage;
515 int lno_in_postimage;
516 const char **label_path;
517 struct diff_words_data *diff_words;
518 struct diff_options *opt;
519 struct strbuf *header;
520 };
521
522 static int count_lines(const char *data, int size)
523 {
524 int count, ch, completely_empty = 1, nl_just_seen = 0;
525 count = 0;
526 while (0 < size--) {
527 ch = *data++;
528 if (ch == '\n') {
529 count++;
530 nl_just_seen = 1;
531 completely_empty = 0;
532 }
533 else {
534 nl_just_seen = 0;
535 completely_empty = 0;
536 }
537 }
538 if (completely_empty)
539 return 0;
540 if (!nl_just_seen)
541 count++; /* no trailing newline */
542 return count;
543 }
544
545 static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
546 {
547 if (!DIFF_FILE_VALID(one)) {
548 mf->ptr = (char *)""; /* does not matter */
549 mf->size = 0;
550 return 0;
551 }
552 else if (diff_populate_filespec(one, 0))
553 return -1;
554
555 mf->ptr = one->data;
556 mf->size = one->size;
557 return 0;
558 }
559
560 /* like fill_mmfile, but only for size, so we can avoid retrieving blob */
561 static unsigned long diff_filespec_size(struct diff_filespec *one)
562 {
563 if (!DIFF_FILE_VALID(one))
564 return 0;
565 diff_populate_filespec(one, CHECK_SIZE_ONLY);
566 return one->size;
567 }
568
569 static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
570 {
571 char *ptr = mf->ptr;
572 long size = mf->size;
573 int cnt = 0;
574
575 if (!size)
576 return cnt;
577 ptr += size - 1; /* pointing at the very end */
578 if (*ptr != '\n')
579 ; /* incomplete line */
580 else
581 ptr--; /* skip the last LF */
582 while (mf->ptr < ptr) {
583 char *prev_eol;
584 for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
585 if (*prev_eol == '\n')
586 break;
587 if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
588 break;
589 cnt++;
590 ptr = prev_eol - 1;
591 }
592 return cnt;
593 }
594
595 static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
596 struct emit_callback *ecbdata)
597 {
598 int l1, l2, at;
599 unsigned ws_rule = ecbdata->ws_rule;
600 l1 = count_trailing_blank(mf1, ws_rule);
601 l2 = count_trailing_blank(mf2, ws_rule);
602 if (l2 <= l1) {
603 ecbdata->blank_at_eof_in_preimage = 0;
604 ecbdata->blank_at_eof_in_postimage = 0;
605 return;
606 }
607 at = count_lines(mf1->ptr, mf1->size);
608 ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
609
610 at = count_lines(mf2->ptr, mf2->size);
611 ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
612 }
613
614 static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
615 int first, const char *line, int len)
616 {
617 int has_trailing_newline, has_trailing_carriage_return;
618 int nofirst;
619 FILE *file = o->file;
620
621 fputs(diff_line_prefix(o), file);
622
623 if (len == 0) {
624 has_trailing_newline = (first == '\n');
625 has_trailing_carriage_return = (!has_trailing_newline &&
626 (first == '\r'));
627 nofirst = has_trailing_newline || has_trailing_carriage_return;
628 } else {
629 has_trailing_newline = (len > 0 && line[len-1] == '\n');
630 if (has_trailing_newline)
631 len--;
632 has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
633 if (has_trailing_carriage_return)
634 len--;
635 nofirst = 0;
636 }
637
638 if (len || !nofirst) {
639 fputs(set, file);
640 if (!nofirst)
641 fputc(first, file);
642 fwrite(line, len, 1, file);
643 fputs(reset, file);
644 }
645 if (has_trailing_carriage_return)
646 fputc('\r', file);
647 if (has_trailing_newline)
648 fputc('\n', file);
649 }
650
651 static void emit_line(struct diff_options *o, const char *set, const char *reset,
652 const char *line, int len)
653 {
654 emit_line_0(o, set, reset, line[0], line+1, len-1);
655 }
656
657 enum diff_symbol {
658 DIFF_SYMBOL_BINARY_DIFF_HEADER,
659 DIFF_SYMBOL_BINARY_DIFF_HEADER_DELTA,
660 DIFF_SYMBOL_BINARY_DIFF_HEADER_LITERAL,
661 DIFF_SYMBOL_BINARY_DIFF_BODY,
662 DIFF_SYMBOL_BINARY_DIFF_FOOTER,
663 DIFF_SYMBOL_STATS_SUMMARY_NO_FILES,
664 DIFF_SYMBOL_STATS_SUMMARY_ABBREV,
665 DIFF_SYMBOL_STATS_SUMMARY_INSERTS_DELETES,
666 DIFF_SYMBOL_STATS_LINE,
667 DIFF_SYMBOL_WORD_DIFF,
668 DIFF_SYMBOL_STAT_SEP,
669 DIFF_SYMBOL_SUMMARY,
670 DIFF_SYMBOL_SUBMODULE_ADD,
671 DIFF_SYMBOL_SUBMODULE_DEL,
672 DIFF_SYMBOL_SUBMODULE_UNTRACKED,
673 DIFF_SYMBOL_SUBMODULE_MODIFIED,
674 DIFF_SYMBOL_SUBMODULE_HEADER,
675 DIFF_SYMBOL_SUBMODULE_ERROR,
676 DIFF_SYMBOL_SUBMODULE_PIPETHROUGH,
677 DIFF_SYMBOL_REWRITE_DIFF,
678 DIFF_SYMBOL_BINARY_FILES,
679 DIFF_SYMBOL_HEADER,
680 DIFF_SYMBOL_FILEPAIR_PLUS,
681 DIFF_SYMBOL_FILEPAIR_MINUS,
682 DIFF_SYMBOL_WORDS_PORCELAIN,
683 DIFF_SYMBOL_WORDS,
684 DIFF_SYMBOL_CONTEXT,
685 DIFF_SYMBOL_CONTEXT_INCOMPLETE,
686 DIFF_SYMBOL_PLUS,
687 DIFF_SYMBOL_MINUS,
688 DIFF_SYMBOL_NO_LF_EOF,
689 DIFF_SYMBOL_CONTEXT_FRAGINFO,
690 DIFF_SYMBOL_CONTEXT_MARKER,
691 DIFF_SYMBOL_SEPARATOR
692 };
693 /*
694 * Flags for content lines:
695 * 0..12 are whitespace rules
696 * 13-15 are WSEH_NEW | WSEH_OLD | WSEH_CONTEXT
697 * 16 is marking if the line is blank at EOF
698 */
699 #define DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF (1<<16)
700 #define DIFF_SYMBOL_MOVED_LINE (1<<17)
701 #define DIFF_SYMBOL_MOVED_LINE_ALT (1<<18)
702 #define DIFF_SYMBOL_MOVED_LINE_UNINTERESTING (1<<19)
703 #define DIFF_SYMBOL_CONTENT_WS_MASK (WSEH_NEW | WSEH_OLD | WSEH_CONTEXT | WS_RULE_MASK)
704
705 /*
706 * This struct is used when we need to buffer the output of the diff output.
707 *
708 * NEEDSWORK: Instead of storing a copy of the line, add an offset pointer
709 * into the pre/post image file. This pointer could be a union with the
710 * line pointer. By storing an offset into the file instead of the literal line,
711 * we can decrease the memory footprint for the buffered output. At first we
712 * may want to only have indirection for the content lines, but we could also
713 * enhance the state for emitting prefabricated lines, e.g. the similarity
714 * score line or hunk/file headers would only need to store a number or path
715 * and then the output can be constructed later on depending on state.
716 */
717 struct emitted_diff_symbol {
718 const char *line;
719 int len;
720 int flags;
721 enum diff_symbol s;
722 };
723 #define EMITTED_DIFF_SYMBOL_INIT {NULL}
724
725 struct emitted_diff_symbols {
726 struct emitted_diff_symbol *buf;
727 int nr, alloc;
728 };
729 #define EMITTED_DIFF_SYMBOLS_INIT {NULL, 0, 0}
730
731 static void append_emitted_diff_symbol(struct diff_options *o,
732 struct emitted_diff_symbol *e)
733 {
734 struct emitted_diff_symbol *f;
735
736 ALLOC_GROW(o->emitted_symbols->buf,
737 o->emitted_symbols->nr + 1,
738 o->emitted_symbols->alloc);
739 f = &o->emitted_symbols->buf[o->emitted_symbols->nr++];
740
741 memcpy(f, e, sizeof(struct emitted_diff_symbol));
742 f->line = e->line ? xmemdupz(e->line, e->len) : NULL;
743 }
744
745 struct moved_entry {
746 struct hashmap_entry ent;
747 const struct emitted_diff_symbol *es;
748 struct moved_entry *next_line;
749 struct ws_delta *wsd;
750 };
751
752 /**
753 * The struct ws_delta holds white space differences between moved lines, i.e.
754 * between '+' and '-' lines that have been detected to be a move.
755 * The string contains the difference in leading white spaces, before the
756 * rest of the line is compared using the white space config for move
757 * coloring. The current_longer indicates if the first string in the
758 * comparision is longer than the second.
759 */
760 struct ws_delta {
761 char *string;
762 unsigned int current_longer : 1;
763 };
764 #define WS_DELTA_INIT { NULL, 0 }
765
766 static int compute_ws_delta(const struct emitted_diff_symbol *a,
767 const struct emitted_diff_symbol *b,
768 struct ws_delta *out)
769 {
770 const struct emitted_diff_symbol *longer = a->len > b->len ? a : b;
771 const struct emitted_diff_symbol *shorter = a->len > b->len ? b : a;
772 int d = longer->len - shorter->len;
773
774 out->string = xmemdupz(longer->line, d);
775 out->current_longer = (a == longer);
776
777 return !strncmp(longer->line + d, shorter->line, shorter->len);
778 }
779
780 static int cmp_in_block_with_wsd(const struct diff_options *o,
781 const struct moved_entry *cur,
782 const struct moved_entry *match,
783 struct moved_entry *pmb,
784 int n)
785 {
786 struct emitted_diff_symbol *l = &o->emitted_symbols->buf[n];
787 int al = cur->es->len, cl = l->len;
788 const char *a = cur->es->line,
789 *b = match->es->line,
790 *c = l->line;
791
792 int wslen;
793
794 /*
795 * We need to check if 'cur' is equal to 'match'.
796 * As those are from the same (+/-) side, we do not need to adjust for
797 * indent changes. However these were found using fuzzy matching
798 * so we do have to check if they are equal.
799 */
800 if (strcmp(a, b))
801 return 1;
802
803 if (!pmb->wsd)
804 /*
805 * No white space delta was carried forward? This can happen
806 * when we exit early in this function and do not carry
807 * forward ws.
808 */
809 return 1;
810
811 /*
812 * The indent changes of the block are known and carried forward in
813 * pmb->wsd; however we need to check if the indent changes of the
814 * current line are still the same as before.
815 *
816 * To do so we need to compare 'l' to 'cur', adjusting the
817 * one of them for the white spaces, depending which was longer.
818 */
819
820 wslen = strlen(pmb->wsd->string);
821 if (pmb->wsd->current_longer) {
822 c += wslen;
823 cl -= wslen;
824 } else {
825 a += wslen;
826 al -= wslen;
827 }
828
829 if (strcmp(a, c))
830 return 1;
831
832 return 0;
833 }
834
835 static int moved_entry_cmp(const void *hashmap_cmp_fn_data,
836 const void *entry,
837 const void *entry_or_key,
838 const void *keydata)
839 {
840 const struct diff_options *diffopt = hashmap_cmp_fn_data;
841 const struct moved_entry *a = entry;
842 const struct moved_entry *b = entry_or_key;
843 unsigned flags = diffopt->color_moved_ws_handling
844 & XDF_WHITESPACE_FLAGS;
845
846 if (diffopt->color_moved_ws_handling &
847 COLOR_MOVED_WS_ALLOW_INDENTATION_CHANGE)
848 /*
849 * As there is not specific white space config given,
850 * we'd need to check for a new block, so ignore all
851 * white space. The setup of the white space
852 * configuration for the next block is done else where
853 */
854 flags |= XDF_IGNORE_WHITESPACE;
855
856 return !xdiff_compare_lines(a->es->line, a->es->len,
857 b->es->line, b->es->len,
858 flags);
859 }
860
861 static struct moved_entry *prepare_entry(struct diff_options *o,
862 int line_no)
863 {
864 struct moved_entry *ret = xmalloc(sizeof(*ret));
865 struct emitted_diff_symbol *l = &o->emitted_symbols->buf[line_no];
866 unsigned flags = o->color_moved_ws_handling & XDF_WHITESPACE_FLAGS;
867
868 ret->ent.hash = xdiff_hash_string(l->line, l->len, flags);
869 ret->es = l;
870 ret->next_line = NULL;
871 ret->wsd = NULL;
872
873 return ret;
874 }
875
876 static void add_lines_to_move_detection(struct diff_options *o,
877 struct hashmap *add_lines,
878 struct hashmap *del_lines)
879 {
880 struct moved_entry *prev_line = NULL;
881
882 int n;
883 for (n = 0; n < o->emitted_symbols->nr; n++) {
884 struct hashmap *hm;
885 struct moved_entry *key;
886
887 switch (o->emitted_symbols->buf[n].s) {
888 case DIFF_SYMBOL_PLUS:
889 hm = add_lines;
890 break;
891 case DIFF_SYMBOL_MINUS:
892 hm = del_lines;
893 break;
894 default:
895 prev_line = NULL;
896 continue;
897 }
898
899 key = prepare_entry(o, n);
900 if (prev_line && prev_line->es->s == o->emitted_symbols->buf[n].s)
901 prev_line->next_line = key;
902
903 hashmap_add(hm, key);
904 prev_line = key;
905 }
906 }
907
908 static void pmb_advance_or_null(struct diff_options *o,
909 struct moved_entry *match,
910 struct hashmap *hm,
911 struct moved_entry **pmb,
912 int pmb_nr)
913 {
914 int i;
915 for (i = 0; i < pmb_nr; i++) {
916 struct moved_entry *prev = pmb[i];
917 struct moved_entry *cur = (prev && prev->next_line) ?
918 prev->next_line : NULL;
919 if (cur && !hm->cmpfn(o, cur, match, NULL)) {
920 pmb[i] = cur;
921 } else {
922 pmb[i] = NULL;
923 }
924 }
925 }
926
927 static void pmb_advance_or_null_multi_match(struct diff_options *o,
928 struct moved_entry *match,
929 struct hashmap *hm,
930 struct moved_entry **pmb,
931 int pmb_nr, int n)
932 {
933 int i;
934 char *got_match = xcalloc(1, pmb_nr);
935
936 for (; match; match = hashmap_get_next(hm, match)) {
937 for (i = 0; i < pmb_nr; i++) {
938 struct moved_entry *prev = pmb[i];
939 struct moved_entry *cur = (prev && prev->next_line) ?
940 prev->next_line : NULL;
941 if (!cur)
942 continue;
943 if (!cmp_in_block_with_wsd(o, cur, match, pmb[i], n))
944 got_match[i] |= 1;
945 }
946 }
947
948 for (i = 0; i < pmb_nr; i++) {
949 if (got_match[i]) {
950 /* Carry the white space delta forward */
951 pmb[i]->next_line->wsd = pmb[i]->wsd;
952 pmb[i] = pmb[i]->next_line;
953 } else
954 pmb[i] = NULL;
955 }
956 }
957
958 static int shrink_potential_moved_blocks(struct moved_entry **pmb,
959 int pmb_nr)
960 {
961 int lp, rp;
962
963 /* Shrink the set of potential block to the remaining running */
964 for (lp = 0, rp = pmb_nr - 1; lp <= rp;) {
965 while (lp < pmb_nr && pmb[lp])
966 lp++;
967 /* lp points at the first NULL now */
968
969 while (rp > -1 && !pmb[rp])
970 rp--;
971 /* rp points at the last non-NULL */
972
973 if (lp < pmb_nr && rp > -1 && lp < rp) {
974 pmb[lp] = pmb[rp];
975 if (pmb[rp]->wsd) {
976 free(pmb[rp]->wsd->string);
977 FREE_AND_NULL(pmb[rp]->wsd);
978 }
979 pmb[rp] = NULL;
980 rp--;
981 lp++;
982 }
983 }
984
985 /* Remember the number of running sets */
986 return rp + 1;
987 }
988
989 /*
990 * If o->color_moved is COLOR_MOVED_PLAIN, this function does nothing.
991 *
992 * Otherwise, if the last block has fewer alphanumeric characters than
993 * COLOR_MOVED_MIN_ALNUM_COUNT, unset DIFF_SYMBOL_MOVED_LINE on all lines in
994 * that block.
995 *
996 * The last block consists of the (n - block_length)'th line up to but not
997 * including the nth line.
998 *
999 * NEEDSWORK: This uses the same heuristic as blame_entry_score() in blame.c.
1000 * Think of a way to unify them.
1001 */
1002 static void adjust_last_block(struct diff_options *o, int n, int block_length)
1003 {
1004 int i, alnum_count = 0;
1005 if (o->color_moved == COLOR_MOVED_PLAIN)
1006 return;
1007 for (i = 1; i < block_length + 1; i++) {
1008 const char *c = o->emitted_symbols->buf[n - i].line;
1009 for (; *c; c++) {
1010 if (!isalnum(*c))
1011 continue;
1012 alnum_count++;
1013 if (alnum_count >= COLOR_MOVED_MIN_ALNUM_COUNT)
1014 return;
1015 }
1016 }
1017 for (i = 1; i < block_length + 1; i++)
1018 o->emitted_symbols->buf[n - i].flags &= ~DIFF_SYMBOL_MOVED_LINE;
1019 }
1020
1021 /* Find blocks of moved code, delegate actual coloring decision to helper */
1022 static void mark_color_as_moved(struct diff_options *o,
1023 struct hashmap *add_lines,
1024 struct hashmap *del_lines)
1025 {
1026 struct moved_entry **pmb = NULL; /* potentially moved blocks */
1027 int pmb_nr = 0, pmb_alloc = 0;
1028 int n, flipped_block = 1, block_length = 0;
1029
1030
1031 for (n = 0; n < o->emitted_symbols->nr; n++) {
1032 struct hashmap *hm = NULL;
1033 struct moved_entry *key;
1034 struct moved_entry *match = NULL;
1035 struct emitted_diff_symbol *l = &o->emitted_symbols->buf[n];
1036
1037 switch (l->s) {
1038 case DIFF_SYMBOL_PLUS:
1039 hm = del_lines;
1040 key = prepare_entry(o, n);
1041 match = hashmap_get(hm, key, NULL);
1042 free(key);
1043 break;
1044 case DIFF_SYMBOL_MINUS:
1045 hm = add_lines;
1046 key = prepare_entry(o, n);
1047 match = hashmap_get(hm, key, NULL);
1048 free(key);
1049 break;
1050 default:
1051 flipped_block = 1;
1052 }
1053
1054 if (!match) {
1055 adjust_last_block(o, n, block_length);
1056 pmb_nr = 0;
1057 block_length = 0;
1058 continue;
1059 }
1060
1061 l->flags |= DIFF_SYMBOL_MOVED_LINE;
1062
1063 if (o->color_moved == COLOR_MOVED_PLAIN)
1064 continue;
1065
1066 if (o->color_moved_ws_handling &
1067 COLOR_MOVED_WS_ALLOW_INDENTATION_CHANGE)
1068 pmb_advance_or_null_multi_match(o, match, hm, pmb, pmb_nr, n);
1069 else
1070 pmb_advance_or_null(o, match, hm, pmb, pmb_nr);
1071
1072 pmb_nr = shrink_potential_moved_blocks(pmb, pmb_nr);
1073
1074 if (pmb_nr == 0) {
1075 /*
1076 * The current line is the start of a new block.
1077 * Setup the set of potential blocks.
1078 */
1079 for (; match; match = hashmap_get_next(hm, match)) {
1080 ALLOC_GROW(pmb, pmb_nr + 1, pmb_alloc);
1081 if (o->color_moved_ws_handling &
1082 COLOR_MOVED_WS_ALLOW_INDENTATION_CHANGE) {
1083 struct ws_delta *wsd = xmalloc(sizeof(*match->wsd));
1084 if (compute_ws_delta(l, match->es, wsd)) {
1085 match->wsd = wsd;
1086 pmb[pmb_nr++] = match;
1087 } else
1088 free(wsd);
1089 } else {
1090 pmb[pmb_nr++] = match;
1091 }
1092 }
1093
1094 flipped_block = (flipped_block + 1) % 2;
1095
1096 adjust_last_block(o, n, block_length);
1097 block_length = 0;
1098 }
1099
1100 block_length++;
1101
1102 if (flipped_block && o->color_moved != COLOR_MOVED_BLOCKS)
1103 l->flags |= DIFF_SYMBOL_MOVED_LINE_ALT;
1104 }
1105 adjust_last_block(o, n, block_length);
1106
1107 free(pmb);
1108 }
1109
1110 #define DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK \
1111 (DIFF_SYMBOL_MOVED_LINE | DIFF_SYMBOL_MOVED_LINE_ALT)
1112 static void dim_moved_lines(struct diff_options *o)
1113 {
1114 int n;
1115 for (n = 0; n < o->emitted_symbols->nr; n++) {
1116 struct emitted_diff_symbol *prev = (n != 0) ?
1117 &o->emitted_symbols->buf[n - 1] : NULL;
1118 struct emitted_diff_symbol *l = &o->emitted_symbols->buf[n];
1119 struct emitted_diff_symbol *next =
1120 (n < o->emitted_symbols->nr - 1) ?
1121 &o->emitted_symbols->buf[n + 1] : NULL;
1122
1123 /* Not a plus or minus line? */
1124 if (l->s != DIFF_SYMBOL_PLUS && l->s != DIFF_SYMBOL_MINUS)
1125 continue;
1126
1127 /* Not a moved line? */
1128 if (!(l->flags & DIFF_SYMBOL_MOVED_LINE))
1129 continue;
1130
1131 /*
1132 * If prev or next are not a plus or minus line,
1133 * pretend they don't exist
1134 */
1135 if (prev && prev->s != DIFF_SYMBOL_PLUS &&
1136 prev->s != DIFF_SYMBOL_MINUS)
1137 prev = NULL;
1138 if (next && next->s != DIFF_SYMBOL_PLUS &&
1139 next->s != DIFF_SYMBOL_MINUS)
1140 next = NULL;
1141
1142 /* Inside a block? */
1143 if ((prev &&
1144 (prev->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK) ==
1145 (l->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK)) &&
1146 (next &&
1147 (next->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK) ==
1148 (l->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK))) {
1149 l->flags |= DIFF_SYMBOL_MOVED_LINE_UNINTERESTING;
1150 continue;
1151 }
1152
1153 /* Check if we are at an interesting bound: */
1154 if (prev && (prev->flags & DIFF_SYMBOL_MOVED_LINE) &&
1155 (prev->flags & DIFF_SYMBOL_MOVED_LINE_ALT) !=
1156 (l->flags & DIFF_SYMBOL_MOVED_LINE_ALT))
1157 continue;
1158 if (next && (next->flags & DIFF_SYMBOL_MOVED_LINE) &&
1159 (next->flags & DIFF_SYMBOL_MOVED_LINE_ALT) !=
1160 (l->flags & DIFF_SYMBOL_MOVED_LINE_ALT))
1161 continue;
1162
1163 /*
1164 * The boundary to prev and next are not interesting,
1165 * so this line is not interesting as a whole
1166 */
1167 l->flags |= DIFF_SYMBOL_MOVED_LINE_UNINTERESTING;
1168 }
1169 }
1170
1171 static void emit_line_ws_markup(struct diff_options *o,
1172 const char *set, const char *reset,
1173 const char *line, int len, char sign,
1174 unsigned ws_rule, int blank_at_eof)
1175 {
1176 const char *ws = NULL;
1177
1178 if (o->ws_error_highlight & ws_rule) {
1179 ws = diff_get_color_opt(o, DIFF_WHITESPACE);
1180 if (!*ws)
1181 ws = NULL;
1182 }
1183
1184 if (!ws)
1185 emit_line_0(o, set, reset, sign, line, len);
1186 else if (blank_at_eof)
1187 /* Blank line at EOF - paint '+' as well */
1188 emit_line_0(o, ws, reset, sign, line, len);
1189 else {
1190 /* Emit just the prefix, then the rest. */
1191 emit_line_0(o, set, reset, sign, "", 0);
1192 ws_check_emit(line, len, ws_rule,
1193 o->file, set, reset, ws);
1194 }
1195 }
1196
1197 static void emit_diff_symbol_from_struct(struct diff_options *o,
1198 struct emitted_diff_symbol *eds)
1199 {
1200 static const char *nneof = " No newline at end of file\n";
1201 const char *context, *reset, *set, *meta, *fraginfo;
1202 struct strbuf sb = STRBUF_INIT;
1203
1204 enum diff_symbol s = eds->s;
1205 const char *line = eds->line;
1206 int len = eds->len;
1207 unsigned flags = eds->flags;
1208
1209 switch (s) {
1210 case DIFF_SYMBOL_NO_LF_EOF:
1211 context = diff_get_color_opt(o, DIFF_CONTEXT);
1212 reset = diff_get_color_opt(o, DIFF_RESET);
1213 putc('\n', o->file);
1214 emit_line_0(o, context, reset, '\\',
1215 nneof, strlen(nneof));
1216 break;
1217 case DIFF_SYMBOL_SUBMODULE_HEADER:
1218 case DIFF_SYMBOL_SUBMODULE_ERROR:
1219 case DIFF_SYMBOL_SUBMODULE_PIPETHROUGH:
1220 case DIFF_SYMBOL_STATS_SUMMARY_INSERTS_DELETES:
1221 case DIFF_SYMBOL_SUMMARY:
1222 case DIFF_SYMBOL_STATS_LINE:
1223 case DIFF_SYMBOL_BINARY_DIFF_BODY:
1224 case DIFF_SYMBOL_CONTEXT_FRAGINFO:
1225 emit_line(o, "", "", line, len);
1226 break;
1227 case DIFF_SYMBOL_CONTEXT_INCOMPLETE:
1228 case DIFF_SYMBOL_CONTEXT_MARKER:
1229 context = diff_get_color_opt(o, DIFF_CONTEXT);
1230 reset = diff_get_color_opt(o, DIFF_RESET);
1231 emit_line(o, context, reset, line, len);
1232 break;
1233 case DIFF_SYMBOL_SEPARATOR:
1234 fprintf(o->file, "%s%c",
1235 diff_line_prefix(o),
1236 o->line_termination);
1237 break;
1238 case DIFF_SYMBOL_CONTEXT:
1239 set = diff_get_color_opt(o, DIFF_CONTEXT);
1240 reset = diff_get_color_opt(o, DIFF_RESET);
1241 emit_line_ws_markup(o, set, reset, line, len, ' ',
1242 flags & (DIFF_SYMBOL_CONTENT_WS_MASK), 0);
1243 break;
1244 case DIFF_SYMBOL_PLUS:
1245 switch (flags & (DIFF_SYMBOL_MOVED_LINE |
1246 DIFF_SYMBOL_MOVED_LINE_ALT |
1247 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING)) {
1248 case DIFF_SYMBOL_MOVED_LINE |
1249 DIFF_SYMBOL_MOVED_LINE_ALT |
1250 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1251 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED_ALT_DIM);
1252 break;
1253 case DIFF_SYMBOL_MOVED_LINE |
1254 DIFF_SYMBOL_MOVED_LINE_ALT:
1255 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED_ALT);
1256 break;
1257 case DIFF_SYMBOL_MOVED_LINE |
1258 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1259 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED_DIM);
1260 break;
1261 case DIFF_SYMBOL_MOVED_LINE:
1262 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED);
1263 break;
1264 default:
1265 set = diff_get_color_opt(o, DIFF_FILE_NEW);
1266 }
1267 reset = diff_get_color_opt(o, DIFF_RESET);
1268 emit_line_ws_markup(o, set, reset, line, len, '+',
1269 flags & DIFF_SYMBOL_CONTENT_WS_MASK,
1270 flags & DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF);
1271 break;
1272 case DIFF_SYMBOL_MINUS:
1273 switch (flags & (DIFF_SYMBOL_MOVED_LINE |
1274 DIFF_SYMBOL_MOVED_LINE_ALT |
1275 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING)) {
1276 case DIFF_SYMBOL_MOVED_LINE |
1277 DIFF_SYMBOL_MOVED_LINE_ALT |
1278 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1279 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED_ALT_DIM);
1280 break;
1281 case DIFF_SYMBOL_MOVED_LINE |
1282 DIFF_SYMBOL_MOVED_LINE_ALT:
1283 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED_ALT);
1284 break;
1285 case DIFF_SYMBOL_MOVED_LINE |
1286 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1287 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED_DIM);
1288 break;
1289 case DIFF_SYMBOL_MOVED_LINE:
1290 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED);
1291 break;
1292 default:
1293 set = diff_get_color_opt(o, DIFF_FILE_OLD);
1294 }
1295 reset = diff_get_color_opt(o, DIFF_RESET);
1296 emit_line_ws_markup(o, set, reset, line, len, '-',
1297 flags & DIFF_SYMBOL_CONTENT_WS_MASK, 0);
1298 break;
1299 case DIFF_SYMBOL_WORDS_PORCELAIN:
1300 context = diff_get_color_opt(o, DIFF_CONTEXT);
1301 reset = diff_get_color_opt(o, DIFF_RESET);
1302 emit_line(o, context, reset, line, len);
1303 fputs("~\n", o->file);
1304 break;
1305 case DIFF_SYMBOL_WORDS:
1306 context = diff_get_color_opt(o, DIFF_CONTEXT);
1307 reset = diff_get_color_opt(o, DIFF_RESET);
1308 /*
1309 * Skip the prefix character, if any. With
1310 * diff_suppress_blank_empty, there may be
1311 * none.
1312 */
1313 if (line[0] != '\n') {
1314 line++;
1315 len--;
1316 }
1317 emit_line(o, context, reset, line, len);
1318 break;
1319 case DIFF_SYMBOL_FILEPAIR_PLUS:
1320 meta = diff_get_color_opt(o, DIFF_METAINFO);
1321 reset = diff_get_color_opt(o, DIFF_RESET);
1322 fprintf(o->file, "%s%s+++ %s%s%s\n", diff_line_prefix(o), meta,
1323 line, reset,
1324 strchr(line, ' ') ? "\t" : "");
1325 break;
1326 case DIFF_SYMBOL_FILEPAIR_MINUS:
1327 meta = diff_get_color_opt(o, DIFF_METAINFO);
1328 reset = diff_get_color_opt(o, DIFF_RESET);
1329 fprintf(o->file, "%s%s--- %s%s%s\n", diff_line_prefix(o), meta,
1330 line, reset,
1331 strchr(line, ' ') ? "\t" : "");
1332 break;
1333 case DIFF_SYMBOL_BINARY_FILES:
1334 case DIFF_SYMBOL_HEADER:
1335 fprintf(o->file, "%s", line);
1336 break;
1337 case DIFF_SYMBOL_BINARY_DIFF_HEADER:
1338 fprintf(o->file, "%sGIT binary patch\n", diff_line_prefix(o));
1339 break;
1340 case DIFF_SYMBOL_BINARY_DIFF_HEADER_DELTA:
1341 fprintf(o->file, "%sdelta %s\n", diff_line_prefix(o), line);
1342 break;
1343 case DIFF_SYMBOL_BINARY_DIFF_HEADER_LITERAL:
1344 fprintf(o->file, "%sliteral %s\n", diff_line_prefix(o), line);
1345 break;
1346 case DIFF_SYMBOL_BINARY_DIFF_FOOTER:
1347 fputs(diff_line_prefix(o), o->file);
1348 fputc('\n', o->file);
1349 break;
1350 case DIFF_SYMBOL_REWRITE_DIFF:
1351 fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
1352 reset = diff_get_color_opt(o, DIFF_RESET);
1353 emit_line(o, fraginfo, reset, line, len);
1354 break;
1355 case DIFF_SYMBOL_SUBMODULE_ADD:
1356 set = diff_get_color_opt(o, DIFF_FILE_NEW);
1357 reset = diff_get_color_opt(o, DIFF_RESET);
1358 emit_line(o, set, reset, line, len);
1359 break;
1360 case DIFF_SYMBOL_SUBMODULE_DEL:
1361 set = diff_get_color_opt(o, DIFF_FILE_OLD);
1362 reset = diff_get_color_opt(o, DIFF_RESET);
1363 emit_line(o, set, reset, line, len);
1364 break;
1365 case DIFF_SYMBOL_SUBMODULE_UNTRACKED:
1366 fprintf(o->file, "%sSubmodule %s contains untracked content\n",
1367 diff_line_prefix(o), line);
1368 break;
1369 case DIFF_SYMBOL_SUBMODULE_MODIFIED:
1370 fprintf(o->file, "%sSubmodule %s contains modified content\n",
1371 diff_line_prefix(o), line);
1372 break;
1373 case DIFF_SYMBOL_STATS_SUMMARY_NO_FILES:
1374 emit_line(o, "", "", " 0 files changed\n",
1375 strlen(" 0 files changed\n"));
1376 break;
1377 case DIFF_SYMBOL_STATS_SUMMARY_ABBREV:
1378 emit_line(o, "", "", " ...\n", strlen(" ...\n"));
1379 break;
1380 case DIFF_SYMBOL_WORD_DIFF:
1381 fprintf(o->file, "%.*s", len, line);
1382 break;
1383 case DIFF_SYMBOL_STAT_SEP:
1384 fputs(o->stat_sep, o->file);
1385 break;
1386 default:
1387 BUG("unknown diff symbol");
1388 }
1389 strbuf_release(&sb);
1390 }
1391
1392 static void emit_diff_symbol(struct diff_options *o, enum diff_symbol s,
1393 const char *line, int len, unsigned flags)
1394 {
1395 struct emitted_diff_symbol e = {line, len, flags, s};
1396
1397 if (o->emitted_symbols)
1398 append_emitted_diff_symbol(o, &e);
1399 else
1400 emit_diff_symbol_from_struct(o, &e);
1401 }
1402
1403 void diff_emit_submodule_del(struct diff_options *o, const char *line)
1404 {
1405 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_DEL, line, strlen(line), 0);
1406 }
1407
1408 void diff_emit_submodule_add(struct diff_options *o, const char *line)
1409 {
1410 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_ADD, line, strlen(line), 0);
1411 }
1412
1413 void diff_emit_submodule_untracked(struct diff_options *o, const char *path)
1414 {
1415 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_UNTRACKED,
1416 path, strlen(path), 0);
1417 }
1418
1419 void diff_emit_submodule_modified(struct diff_options *o, const char *path)
1420 {
1421 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_MODIFIED,
1422 path, strlen(path), 0);
1423 }
1424
1425 void diff_emit_submodule_header(struct diff_options *o, const char *header)
1426 {
1427 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_HEADER,
1428 header, strlen(header), 0);
1429 }
1430
1431 void diff_emit_submodule_error(struct diff_options *o, const char *err)
1432 {
1433 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_ERROR, err, strlen(err), 0);
1434 }
1435
1436 void diff_emit_submodule_pipethrough(struct diff_options *o,
1437 const char *line, int len)
1438 {
1439 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_PIPETHROUGH, line, len, 0);
1440 }
1441
1442 static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
1443 {
1444 if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
1445 ecbdata->blank_at_eof_in_preimage &&
1446 ecbdata->blank_at_eof_in_postimage &&
1447 ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
1448 ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
1449 return 0;
1450 return ws_blank_line(line, len, ecbdata->ws_rule);
1451 }
1452
1453 static void emit_add_line(const char *reset,
1454 struct emit_callback *ecbdata,
1455 const char *line, int len)
1456 {
1457 unsigned flags = WSEH_NEW | ecbdata->ws_rule;
1458 if (new_blank_line_at_eof(ecbdata, line, len))
1459 flags |= DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF;
1460
1461 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_PLUS, line, len, flags);
1462 }
1463
1464 static void emit_del_line(const char *reset,
1465 struct emit_callback *ecbdata,
1466 const char *line, int len)
1467 {
1468 unsigned flags = WSEH_OLD | ecbdata->ws_rule;
1469 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_MINUS, line, len, flags);
1470 }
1471
1472 static void emit_context_line(const char *reset,
1473 struct emit_callback *ecbdata,
1474 const char *line, int len)
1475 {
1476 unsigned flags = WSEH_CONTEXT | ecbdata->ws_rule;
1477 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_CONTEXT, line, len, flags);
1478 }
1479
1480 static void emit_hunk_header(struct emit_callback *ecbdata,
1481 const char *line, int len)
1482 {
1483 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
1484 const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
1485 const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
1486 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1487 static const char atat[2] = { '@', '@' };
1488 const char *cp, *ep;
1489 struct strbuf msgbuf = STRBUF_INIT;
1490 int org_len = len;
1491 int i = 1;
1492
1493 /*
1494 * As a hunk header must begin with "@@ -<old>, +<new> @@",
1495 * it always is at least 10 bytes long.
1496 */
1497 if (len < 10 ||
1498 memcmp(line, atat, 2) ||
1499 !(ep = memmem(line + 2, len - 2, atat, 2))) {
1500 emit_diff_symbol(ecbdata->opt,
1501 DIFF_SYMBOL_CONTEXT_MARKER, line, len, 0);
1502 return;
1503 }
1504 ep += 2; /* skip over @@ */
1505
1506 /* The hunk header in fraginfo color */
1507 strbuf_addstr(&msgbuf, frag);
1508 strbuf_add(&msgbuf, line, ep - line);
1509 strbuf_addstr(&msgbuf, reset);
1510
1511 /*
1512 * trailing "\r\n"
1513 */
1514 for ( ; i < 3; i++)
1515 if (line[len - i] == '\r' || line[len - i] == '\n')
1516 len--;
1517
1518 /* blank before the func header */
1519 for (cp = ep; ep - line < len; ep++)
1520 if (*ep != ' ' && *ep != '\t')
1521 break;
1522 if (ep != cp) {
1523 strbuf_addstr(&msgbuf, context);
1524 strbuf_add(&msgbuf, cp, ep - cp);
1525 strbuf_addstr(&msgbuf, reset);
1526 }
1527
1528 if (ep < line + len) {
1529 strbuf_addstr(&msgbuf, func);
1530 strbuf_add(&msgbuf, ep, line + len - ep);
1531 strbuf_addstr(&msgbuf, reset);
1532 }
1533
1534 strbuf_add(&msgbuf, line + len, org_len - len);
1535 strbuf_complete_line(&msgbuf);
1536 emit_diff_symbol(ecbdata->opt,
1537 DIFF_SYMBOL_CONTEXT_FRAGINFO, msgbuf.buf, msgbuf.len, 0);
1538 strbuf_release(&msgbuf);
1539 }
1540
1541 static struct diff_tempfile *claim_diff_tempfile(void) {
1542 int i;
1543 for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
1544 if (!diff_temp[i].name)
1545 return diff_temp + i;
1546 BUG("diff is failing to clean up its tempfiles");
1547 }
1548
1549 static void remove_tempfile(void)
1550 {
1551 int i;
1552 for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
1553 if (is_tempfile_active(diff_temp[i].tempfile))
1554 delete_tempfile(&diff_temp[i].tempfile);
1555 diff_temp[i].name = NULL;
1556 }
1557 }
1558
1559 static void add_line_count(struct strbuf *out, int count)
1560 {
1561 switch (count) {
1562 case 0:
1563 strbuf_addstr(out, "0,0");
1564 break;
1565 case 1:
1566 strbuf_addstr(out, "1");
1567 break;
1568 default:
1569 strbuf_addf(out, "1,%d", count);
1570 break;
1571 }
1572 }
1573
1574 static void emit_rewrite_lines(struct emit_callback *ecb,
1575 int prefix, const char *data, int size)
1576 {
1577 const char *endp = NULL;
1578 const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
1579
1580 while (0 < size) {
1581 int len;
1582
1583 endp = memchr(data, '\n', size);
1584 len = endp ? (endp - data + 1) : size;
1585 if (prefix != '+') {
1586 ecb->lno_in_preimage++;
1587 emit_del_line(reset, ecb, data, len);
1588 } else {
1589 ecb->lno_in_postimage++;
1590 emit_add_line(reset, ecb, data, len);
1591 }
1592 size -= len;
1593 data += len;
1594 }
1595 if (!endp)
1596 emit_diff_symbol(ecb->opt, DIFF_SYMBOL_NO_LF_EOF, NULL, 0, 0);
1597 }
1598
1599 static void emit_rewrite_diff(const char *name_a,
1600 const char *name_b,
1601 struct diff_filespec *one,
1602 struct diff_filespec *two,
1603 struct userdiff_driver *textconv_one,
1604 struct userdiff_driver *textconv_two,
1605 struct diff_options *o)
1606 {
1607 int lc_a, lc_b;
1608 static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
1609 const char *a_prefix, *b_prefix;
1610 char *data_one, *data_two;
1611 size_t size_one, size_two;
1612 struct emit_callback ecbdata;
1613 struct strbuf out = STRBUF_INIT;
1614
1615 if (diff_mnemonic_prefix && o->flags.reverse_diff) {
1616 a_prefix = o->b_prefix;
1617 b_prefix = o->a_prefix;
1618 } else {
1619 a_prefix = o->a_prefix;
1620 b_prefix = o->b_prefix;
1621 }
1622
1623 name_a += (*name_a == '/');
1624 name_b += (*name_b == '/');
1625
1626 strbuf_reset(&a_name);
1627 strbuf_reset(&b_name);
1628 quote_two_c_style(&a_name, a_prefix, name_a, 0);
1629 quote_two_c_style(&b_name, b_prefix, name_b, 0);
1630
1631 size_one = fill_textconv(textconv_one, one, &data_one);
1632 size_two = fill_textconv(textconv_two, two, &data_two);
1633
1634 memset(&ecbdata, 0, sizeof(ecbdata));
1635 ecbdata.color_diff = want_color(o->use_color);
1636 ecbdata.ws_rule = whitespace_rule(name_b);
1637 ecbdata.opt = o;
1638 if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
1639 mmfile_t mf1, mf2;
1640 mf1.ptr = (char *)data_one;
1641 mf2.ptr = (char *)data_two;
1642 mf1.size = size_one;
1643 mf2.size = size_two;
1644 check_blank_at_eof(&mf1, &mf2, &ecbdata);
1645 }
1646 ecbdata.lno_in_preimage = 1;
1647 ecbdata.lno_in_postimage = 1;
1648
1649 lc_a = count_lines(data_one, size_one);
1650 lc_b = count_lines(data_two, size_two);
1651
1652 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_MINUS,
1653 a_name.buf, a_name.len, 0);
1654 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_PLUS,
1655 b_name.buf, b_name.len, 0);
1656
1657 strbuf_addstr(&out, "@@ -");
1658 if (!o->irreversible_delete)
1659 add_line_count(&out, lc_a);
1660 else
1661 strbuf_addstr(&out, "?,?");
1662 strbuf_addstr(&out, " +");
1663 add_line_count(&out, lc_b);
1664 strbuf_addstr(&out, " @@\n");
1665 emit_diff_symbol(o, DIFF_SYMBOL_REWRITE_DIFF, out.buf, out.len, 0);
1666 strbuf_release(&out);
1667
1668 if (lc_a && !o->irreversible_delete)
1669 emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
1670 if (lc_b)
1671 emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
1672 if (textconv_one)
1673 free((char *)data_one);
1674 if (textconv_two)
1675 free((char *)data_two);
1676 }
1677
1678 struct diff_words_buffer {
1679 mmfile_t text;
1680 unsigned long alloc;
1681 struct diff_words_orig {
1682 const char *begin, *end;
1683 } *orig;
1684 int orig_nr, orig_alloc;
1685 };
1686
1687 static void diff_words_append(char *line, unsigned long len,
1688 struct diff_words_buffer *buffer)
1689 {
1690 ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
1691 line++;
1692 len--;
1693 memcpy(buffer->text.ptr + buffer->text.size, line, len);
1694 buffer->text.size += len;
1695 buffer->text.ptr[buffer->text.size] = '\0';
1696 }
1697
1698 struct diff_words_style_elem {
1699 const char *prefix;
1700 const char *suffix;
1701 const char *color; /* NULL; filled in by the setup code if
1702 * color is enabled */
1703 };
1704
1705 struct diff_words_style {
1706 enum diff_words_type type;
1707 struct diff_words_style_elem new_word, old_word, ctx;
1708 const char *newline;
1709 };
1710
1711 static struct diff_words_style diff_words_styles[] = {
1712 { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
1713 { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
1714 { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
1715 };
1716
1717 struct diff_words_data {
1718 struct diff_words_buffer minus, plus;
1719 const char *current_plus;
1720 int last_minus;
1721 struct diff_options *opt;
1722 regex_t *word_regex;
1723 enum diff_words_type type;
1724 struct diff_words_style *style;
1725 };
1726
1727 static int fn_out_diff_words_write_helper(struct diff_options *o,
1728 struct diff_words_style_elem *st_el,
1729 const char *newline,
1730 size_t count, const char *buf)
1731 {
1732 int print = 0;
1733 struct strbuf sb = STRBUF_INIT;
1734
1735 while (count) {
1736 char *p = memchr(buf, '\n', count);
1737 if (print)
1738 strbuf_addstr(&sb, diff_line_prefix(o));
1739
1740 if (p != buf) {
1741 const char *reset = st_el->color && *st_el->color ?
1742 GIT_COLOR_RESET : NULL;
1743 if (st_el->color && *st_el->color)
1744 strbuf_addstr(&sb, st_el->color);
1745 strbuf_addstr(&sb, st_el->prefix);
1746 strbuf_add(&sb, buf, p ? p - buf : count);
1747 strbuf_addstr(&sb, st_el->suffix);
1748 if (reset)
1749 strbuf_addstr(&sb, reset);
1750 }
1751 if (!p)
1752 goto out;
1753
1754 strbuf_addstr(&sb, newline);
1755 count -= p + 1 - buf;
1756 buf = p + 1;
1757 print = 1;
1758 if (count) {
1759 emit_diff_symbol(o, DIFF_SYMBOL_WORD_DIFF,
1760 sb.buf, sb.len, 0);
1761 strbuf_reset(&sb);
1762 }
1763 }
1764
1765 out:
1766 if (sb.len)
1767 emit_diff_symbol(o, DIFF_SYMBOL_WORD_DIFF,
1768 sb.buf, sb.len, 0);
1769 strbuf_release(&sb);
1770 return 0;
1771 }
1772
1773 /*
1774 * '--color-words' algorithm can be described as:
1775 *
1776 * 1. collect the minus/plus lines of a diff hunk, divided into
1777 * minus-lines and plus-lines;
1778 *
1779 * 2. break both minus-lines and plus-lines into words and
1780 * place them into two mmfile_t with one word for each line;
1781 *
1782 * 3. use xdiff to run diff on the two mmfile_t to get the words level diff;
1783 *
1784 * And for the common parts of the both file, we output the plus side text.
1785 * diff_words->current_plus is used to trace the current position of the plus file
1786 * which printed. diff_words->last_minus is used to trace the last minus word
1787 * printed.
1788 *
1789 * For '--graph' to work with '--color-words', we need to output the graph prefix
1790 * on each line of color words output. Generally, there are two conditions on
1791 * which we should output the prefix.
1792 *
1793 * 1. diff_words->last_minus == 0 &&
1794 * diff_words->current_plus == diff_words->plus.text.ptr
1795 *
1796 * that is: the plus text must start as a new line, and if there is no minus
1797 * word printed, a graph prefix must be printed.
1798 *
1799 * 2. diff_words->current_plus > diff_words->plus.text.ptr &&
1800 * *(diff_words->current_plus - 1) == '\n'
1801 *
1802 * that is: a graph prefix must be printed following a '\n'
1803 */
1804 static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
1805 {
1806 if ((diff_words->last_minus == 0 &&
1807 diff_words->current_plus == diff_words->plus.text.ptr) ||
1808 (diff_words->current_plus > diff_words->plus.text.ptr &&
1809 *(diff_words->current_plus - 1) == '\n')) {
1810 return 1;
1811 } else {
1812 return 0;
1813 }
1814 }
1815
1816 static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
1817 {
1818 struct diff_words_data *diff_words = priv;
1819 struct diff_words_style *style = diff_words->style;
1820 int minus_first, minus_len, plus_first, plus_len;
1821 const char *minus_begin, *minus_end, *plus_begin, *plus_end;
1822 struct diff_options *opt = diff_words->opt;
1823 const char *line_prefix;
1824
1825 if (line[0] != '@' || parse_hunk_header(line, len,
1826 &minus_first, &minus_len, &plus_first, &plus_len))
1827 return;
1828
1829 assert(opt);
1830 line_prefix = diff_line_prefix(opt);
1831
1832 /* POSIX requires that first be decremented by one if len == 0... */
1833 if (minus_len) {
1834 minus_begin = diff_words->minus.orig[minus_first].begin;
1835 minus_end =
1836 diff_words->minus.orig[minus_first + minus_len - 1].end;
1837 } else
1838 minus_begin = minus_end =
1839 diff_words->minus.orig[minus_first].end;
1840
1841 if (plus_len) {
1842 plus_begin = diff_words->plus.orig[plus_first].begin;
1843 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
1844 } else
1845 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
1846
1847 if (color_words_output_graph_prefix(diff_words)) {
1848 fputs(line_prefix, diff_words->opt->file);
1849 }
1850 if (diff_words->current_plus != plus_begin) {
1851 fn_out_diff_words_write_helper(diff_words->opt,
1852 &style->ctx, style->newline,
1853 plus_begin - diff_words->current_plus,
1854 diff_words->current_plus);
1855 }
1856 if (minus_begin != minus_end) {
1857 fn_out_diff_words_write_helper(diff_words->opt,
1858 &style->old_word, style->newline,
1859 minus_end - minus_begin, minus_begin);
1860 }
1861 if (plus_begin != plus_end) {
1862 fn_out_diff_words_write_helper(diff_words->opt,
1863 &style->new_word, style->newline,
1864 plus_end - plus_begin, plus_begin);
1865 }
1866
1867 diff_words->current_plus = plus_end;
1868 diff_words->last_minus = minus_first;
1869 }
1870
1871 /* This function starts looking at *begin, and returns 0 iff a word was found. */
1872 static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
1873 int *begin, int *end)
1874 {
1875 if (word_regex && *begin < buffer->size) {
1876 regmatch_t match[1];
1877 if (!regexec_buf(word_regex, buffer->ptr + *begin,
1878 buffer->size - *begin, 1, match, 0)) {
1879 char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
1880 '\n', match[0].rm_eo - match[0].rm_so);
1881 *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
1882 *begin += match[0].rm_so;
1883 return *begin >= *end;
1884 }
1885 return -1;
1886 }
1887
1888 /* find the next word */
1889 while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
1890 (*begin)++;
1891 if (*begin >= buffer->size)
1892 return -1;
1893
1894 /* find the end of the word */
1895 *end = *begin + 1;
1896 while (*end < buffer->size && !isspace(buffer->ptr[*end]))
1897 (*end)++;
1898
1899 return 0;
1900 }
1901
1902 /*
1903 * This function splits the words in buffer->text, stores the list with
1904 * newline separator into out, and saves the offsets of the original words
1905 * in buffer->orig.
1906 */
1907 static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
1908 regex_t *word_regex)
1909 {
1910 int i, j;
1911 long alloc = 0;
1912
1913 out->size = 0;
1914 out->ptr = NULL;
1915
1916 /* fake an empty "0th" word */
1917 ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
1918 buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
1919 buffer->orig_nr = 1;
1920
1921 for (i = 0; i < buffer->text.size; i++) {
1922 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
1923 return;
1924
1925 /* store original boundaries */
1926 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
1927 buffer->orig_alloc);
1928 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
1929 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
1930 buffer->orig_nr++;
1931
1932 /* store one word */
1933 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
1934 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
1935 out->ptr[out->size + j - i] = '\n';
1936 out->size += j - i + 1;
1937
1938 i = j - 1;
1939 }
1940 }
1941
1942 /* this executes the word diff on the accumulated buffers */
1943 static void diff_words_show(struct diff_words_data *diff_words)
1944 {
1945 xpparam_t xpp;
1946 xdemitconf_t xecfg;
1947 mmfile_t minus, plus;
1948 struct diff_words_style *style = diff_words->style;
1949
1950 struct diff_options *opt = diff_words->opt;
1951 const char *line_prefix;
1952
1953 assert(opt);
1954 line_prefix = diff_line_prefix(opt);
1955
1956 /* special case: only removal */
1957 if (!diff_words->plus.text.size) {
1958 emit_diff_symbol(diff_words->opt, DIFF_SYMBOL_WORD_DIFF,
1959 line_prefix, strlen(line_prefix), 0);
1960 fn_out_diff_words_write_helper(diff_words->opt,
1961 &style->old_word, style->newline,
1962 diff_words->minus.text.size,
1963 diff_words->minus.text.ptr);
1964 diff_words->minus.text.size = 0;
1965 return;
1966 }
1967
1968 diff_words->current_plus = diff_words->plus.text.ptr;
1969 diff_words->last_minus = 0;
1970
1971 memset(&xpp, 0, sizeof(xpp));
1972 memset(&xecfg, 0, sizeof(xecfg));
1973 diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
1974 diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
1975 xpp.flags = 0;
1976 /* as only the hunk header will be parsed, we need a 0-context */
1977 xecfg.ctxlen = 0;
1978 if (xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
1979 &xpp, &xecfg))
1980 die("unable to generate word diff");
1981 free(minus.ptr);
1982 free(plus.ptr);
1983 if (diff_words->current_plus != diff_words->plus.text.ptr +
1984 diff_words->plus.text.size) {
1985 if (color_words_output_graph_prefix(diff_words))
1986 emit_diff_symbol(diff_words->opt, DIFF_SYMBOL_WORD_DIFF,
1987 line_prefix, strlen(line_prefix), 0);
1988 fn_out_diff_words_write_helper(diff_words->opt,
1989 &style->ctx, style->newline,
1990 diff_words->plus.text.ptr + diff_words->plus.text.size
1991 - diff_words->current_plus, diff_words->current_plus);
1992 }
1993 diff_words->minus.text.size = diff_words->plus.text.size = 0;
1994 }
1995
1996 /* In "color-words" mode, show word-diff of words accumulated in the buffer */
1997 static void diff_words_flush(struct emit_callback *ecbdata)
1998 {
1999 struct diff_options *wo = ecbdata->diff_words->opt;
2000
2001 if (ecbdata->diff_words->minus.text.size ||
2002 ecbdata->diff_words->plus.text.size)
2003 diff_words_show(ecbdata->diff_words);
2004
2005 if (wo->emitted_symbols) {
2006 struct diff_options *o = ecbdata->opt;
2007 struct emitted_diff_symbols *wol = wo->emitted_symbols;
2008 int i;
2009
2010 /*
2011 * NEEDSWORK:
2012 * Instead of appending each, concat all words to a line?
2013 */
2014 for (i = 0; i < wol->nr; i++)
2015 append_emitted_diff_symbol(o, &wol->buf[i]);
2016
2017 for (i = 0; i < wol->nr; i++)
2018 free((void *)wol->buf[i].line);
2019
2020 wol->nr = 0;
2021 }
2022 }
2023
2024 static void diff_filespec_load_driver(struct diff_filespec *one)
2025 {
2026 /* Use already-loaded driver */
2027 if (one->driver)
2028 return;
2029
2030 if (S_ISREG(one->mode))
2031 one->driver = userdiff_find_by_path(one->path);
2032
2033 /* Fallback to default settings */
2034 if (!one->driver)
2035 one->driver = userdiff_find_by_name("default");
2036 }
2037
2038 static const char *userdiff_word_regex(struct diff_filespec *one)
2039 {
2040 diff_filespec_load_driver(one);
2041 return one->driver->word_regex;
2042 }
2043
2044 static void init_diff_words_data(struct emit_callback *ecbdata,
2045 struct diff_options *orig_opts,
2046 struct diff_filespec *one,
2047 struct diff_filespec *two)
2048 {
2049 int i;
2050 struct diff_options *o = xmalloc(sizeof(struct diff_options));
2051 memcpy(o, orig_opts, sizeof(struct diff_options));
2052
2053 ecbdata->diff_words =
2054 xcalloc(1, sizeof(struct diff_words_data));
2055 ecbdata->diff_words->type = o->word_diff;
2056 ecbdata->diff_words->opt = o;
2057
2058 if (orig_opts->emitted_symbols)
2059 o->emitted_symbols =
2060 xcalloc(1, sizeof(struct emitted_diff_symbols));
2061
2062 if (!o->word_regex)
2063 o->word_regex = userdiff_word_regex(one);
2064 if (!o->word_regex)
2065 o->word_regex = userdiff_word_regex(two);
2066 if (!o->word_regex)
2067 o->word_regex = diff_word_regex_cfg;
2068 if (o->word_regex) {
2069 ecbdata->diff_words->word_regex = (regex_t *)
2070 xmalloc(sizeof(regex_t));
2071 if (regcomp(ecbdata->diff_words->word_regex,
2072 o->word_regex,
2073 REG_EXTENDED | REG_NEWLINE))
2074 die("invalid regular expression: %s",
2075 o->word_regex);
2076 }
2077 for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
2078 if (o->word_diff == diff_words_styles[i].type) {
2079 ecbdata->diff_words->style =
2080 &diff_words_styles[i];
2081 break;
2082 }
2083 }
2084 if (want_color(o->use_color)) {
2085 struct diff_words_style *st = ecbdata->diff_words->style;
2086 st->old_word.color = diff_get_color_opt(o, DIFF_FILE_OLD);
2087 st->new_word.color = diff_get_color_opt(o, DIFF_FILE_NEW);
2088 st->ctx.color = diff_get_color_opt(o, DIFF_CONTEXT);
2089 }
2090 }
2091
2092 static void free_diff_words_data(struct emit_callback *ecbdata)
2093 {
2094 if (ecbdata->diff_words) {
2095 diff_words_flush(ecbdata);
2096 free (ecbdata->diff_words->opt->emitted_symbols);
2097 free (ecbdata->diff_words->opt);
2098 free (ecbdata->diff_words->minus.text.ptr);
2099 free (ecbdata->diff_words->minus.orig);
2100 free (ecbdata->diff_words->plus.text.ptr);
2101 free (ecbdata->diff_words->plus.orig);
2102 if (ecbdata->diff_words->word_regex) {
2103 regfree(ecbdata->diff_words->word_regex);
2104 free(ecbdata->diff_words->word_regex);
2105 }
2106 FREE_AND_NULL(ecbdata->diff_words);
2107 }
2108 }
2109
2110 const char *diff_get_color(int diff_use_color, enum color_diff ix)
2111 {
2112 if (want_color(diff_use_color))
2113 return diff_colors[ix];
2114 return "";
2115 }
2116
2117 const char *diff_line_prefix(struct diff_options *opt)
2118 {
2119 struct strbuf *msgbuf;
2120 if (!opt->output_prefix)
2121 return "";
2122
2123 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
2124 return msgbuf->buf;
2125 }
2126
2127 static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
2128 {
2129 const char *cp;
2130 unsigned long allot;
2131 size_t l = len;
2132
2133 cp = line;
2134 allot = l;
2135 while (0 < l) {
2136 (void) utf8_width(&cp, &l);
2137 if (!cp)
2138 break; /* truncated in the middle? */
2139 }
2140 return allot - l;
2141 }
2142
2143 static void find_lno(const char *line, struct emit_callback *ecbdata)
2144 {
2145 const char *p;
2146 ecbdata->lno_in_preimage = 0;
2147 ecbdata->lno_in_postimage = 0;
2148 p = strchr(line, '-');
2149 if (!p)
2150 return; /* cannot happen */
2151 ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
2152 p = strchr(p, '+');
2153 if (!p)
2154 return; /* cannot happen */
2155 ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
2156 }
2157
2158 static void fn_out_consume(void *priv, char *line, unsigned long len)
2159 {
2160 struct emit_callback *ecbdata = priv;
2161 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
2162 struct diff_options *o = ecbdata->opt;
2163
2164 o->found_changes = 1;
2165
2166 if (ecbdata->header) {
2167 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
2168 ecbdata->header->buf, ecbdata->header->len, 0);
2169 strbuf_reset(ecbdata->header);
2170 ecbdata->header = NULL;
2171 }
2172
2173 if (ecbdata->label_path[0]) {
2174 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_MINUS,
2175 ecbdata->label_path[0],
2176 strlen(ecbdata->label_path[0]), 0);
2177 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_PLUS,
2178 ecbdata->label_path[1],
2179 strlen(ecbdata->label_path[1]), 0);
2180 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
2181 }
2182
2183 if (diff_suppress_blank_empty
2184 && len == 2 && line[0] == ' ' && line[1] == '\n') {
2185 line[0] = '\n';
2186 len = 1;
2187 }
2188
2189 if (line[0] == '@') {
2190 if (ecbdata->diff_words)
2191 diff_words_flush(ecbdata);
2192 len = sane_truncate_line(ecbdata, line, len);
2193 find_lno(line, ecbdata);
2194 emit_hunk_header(ecbdata, line, len);
2195 return;
2196 }
2197
2198 if (ecbdata->diff_words) {
2199 enum diff_symbol s =
2200 ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN ?
2201 DIFF_SYMBOL_WORDS_PORCELAIN : DIFF_SYMBOL_WORDS;
2202 if (line[0] == '-') {
2203 diff_words_append(line, len,
2204 &ecbdata->diff_words->minus);
2205 return;
2206 } else if (line[0] == '+') {
2207 diff_words_append(line, len,
2208 &ecbdata->diff_words->plus);
2209 return;
2210 } else if (starts_with(line, "\\ ")) {
2211 /*
2212 * Eat the "no newline at eof" marker as if we
2213 * saw a "+" or "-" line with nothing on it,
2214 * and return without diff_words_flush() to
2215 * defer processing. If this is the end of
2216 * preimage, more "+" lines may come after it.
2217 */
2218 return;
2219 }
2220 diff_words_flush(ecbdata);
2221 emit_diff_symbol(o, s, line, len, 0);
2222 return;
2223 }
2224
2225 switch (line[0]) {
2226 case '+':
2227 ecbdata->lno_in_postimage++;
2228 emit_add_line(reset, ecbdata, line + 1, len - 1);
2229 break;
2230 case '-':
2231 ecbdata->lno_in_preimage++;
2232 emit_del_line(reset, ecbdata, line + 1, len - 1);
2233 break;
2234 case ' ':
2235 ecbdata->lno_in_postimage++;
2236 ecbdata->lno_in_preimage++;
2237 emit_context_line(reset, ecbdata, line + 1, len - 1);
2238 break;
2239 default:
2240 /* incomplete line at the end */
2241 ecbdata->lno_in_preimage++;
2242 emit_diff_symbol(o, DIFF_SYMBOL_CONTEXT_INCOMPLETE,
2243 line, len, 0);
2244 break;
2245 }
2246 }
2247
2248 static void pprint_rename(struct strbuf *name, const char *a, const char *b)
2249 {
2250 const char *old_name = a;
2251 const char *new_name = b;
2252 int pfx_length, sfx_length;
2253 int pfx_adjust_for_slash;
2254 int len_a = strlen(a);
2255 int len_b = strlen(b);
2256 int a_midlen, b_midlen;
2257 int qlen_a = quote_c_style(a, NULL, NULL, 0);
2258 int qlen_b = quote_c_style(b, NULL, NULL, 0);
2259
2260 if (qlen_a || qlen_b) {
2261 quote_c_style(a, name, NULL, 0);
2262 strbuf_addstr(name, " => ");
2263 quote_c_style(b, name, NULL, 0);
2264 return;
2265 }
2266
2267 /* Find common prefix */
2268 pfx_length = 0;
2269 while (*old_name && *new_name && *old_name == *new_name) {
2270 if (*old_name == '/')
2271 pfx_length = old_name - a + 1;
2272 old_name++;
2273 new_name++;
2274 }
2275
2276 /* Find common suffix */
2277 old_name = a + len_a;
2278 new_name = b + len_b;
2279 sfx_length = 0;
2280 /*
2281 * If there is a common prefix, it must end in a slash. In
2282 * that case we let this loop run 1 into the prefix to see the
2283 * same slash.
2284 *
2285 * If there is no common prefix, we cannot do this as it would
2286 * underrun the input strings.
2287 */
2288 pfx_adjust_for_slash = (pfx_length ? 1 : 0);
2289 while (a + pfx_length - pfx_adjust_for_slash <= old_name &&
2290 b + pfx_length - pfx_adjust_for_slash <= new_name &&
2291 *old_name == *new_name) {
2292 if (*old_name == '/')
2293 sfx_length = len_a - (old_name - a);
2294 old_name--;
2295 new_name--;
2296 }
2297
2298 /*
2299 * pfx{mid-a => mid-b}sfx
2300 * {pfx-a => pfx-b}sfx
2301 * pfx{sfx-a => sfx-b}
2302 * name-a => name-b
2303 */
2304 a_midlen = len_a - pfx_length - sfx_length;
2305 b_midlen = len_b - pfx_length - sfx_length;
2306 if (a_midlen < 0)
2307 a_midlen = 0;
2308 if (b_midlen < 0)
2309 b_midlen = 0;
2310
2311 strbuf_grow(name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
2312 if (pfx_length + sfx_length) {
2313 strbuf_add(name, a, pfx_length);
2314 strbuf_addch(name, '{');
2315 }
2316 strbuf_add(name, a + pfx_length, a_midlen);
2317 strbuf_addstr(name, " => ");
2318 strbuf_add(name, b + pfx_length, b_midlen);
2319 if (pfx_length + sfx_length) {
2320 strbuf_addch(name, '}');
2321 strbuf_add(name, a + len_a - sfx_length, sfx_length);
2322 }
2323 }
2324
2325 struct diffstat_t {
2326 int nr;
2327 int alloc;
2328 struct diffstat_file {
2329 char *from_name;
2330 char *name;
2331 char *print_name;
2332 const char *comments;
2333 unsigned is_unmerged:1;
2334 unsigned is_binary:1;
2335 unsigned is_renamed:1;
2336 unsigned is_interesting:1;
2337 uintmax_t added, deleted;
2338 } **files;
2339 };
2340
2341 static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
2342 const char *name_a,
2343 const char *name_b)
2344 {
2345 struct diffstat_file *x;
2346 x = xcalloc(1, sizeof(*x));
2347 ALLOC_GROW(diffstat->files, diffstat->nr + 1, diffstat->alloc);
2348 diffstat->files[diffstat->nr++] = x;
2349 if (name_b) {
2350 x->from_name = xstrdup(name_a);
2351 x->name = xstrdup(name_b);
2352 x->is_renamed = 1;
2353 }
2354 else {
2355 x->from_name = NULL;
2356 x->name = xstrdup(name_a);
2357 }
2358 return x;
2359 }
2360
2361 static void diffstat_consume(void *priv, char *line, unsigned long len)
2362 {
2363 struct diffstat_t *diffstat = priv;
2364 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
2365
2366 if (line[0] == '+')
2367 x->added++;
2368 else if (line[0] == '-')
2369 x->deleted++;
2370 }
2371
2372 const char mime_boundary_leader[] = "------------";
2373
2374 static int scale_linear(int it, int width, int max_change)
2375 {
2376 if (!it)
2377 return 0;
2378 /*
2379 * make sure that at least one '-' or '+' is printed if
2380 * there is any change to this path. The easiest way is to
2381 * scale linearly as if the alloted width is one column shorter
2382 * than it is, and then add 1 to the result.
2383 */
2384 return 1 + (it * (width - 1) / max_change);
2385 }
2386
2387 static void show_graph(struct strbuf *out, char ch, int cnt,
2388 const char *set, const char *reset)
2389 {
2390 if (cnt <= 0)
2391 return;
2392 strbuf_addstr(out, set);
2393 strbuf_addchars(out, ch, cnt);
2394 strbuf_addstr(out, reset);
2395 }
2396
2397 static void fill_print_name(struct diffstat_file *file)
2398 {
2399 struct strbuf pname = STRBUF_INIT;
2400
2401 if (file->print_name)
2402 return;
2403
2404 if (file->is_renamed)
2405 pprint_rename(&pname, file->from_name, file->name);
2406 else
2407 quote_c_style(file->name, &pname, NULL, 0);
2408
2409 if (file->comments)
2410 strbuf_addf(&pname, " (%s)", file->comments);
2411
2412 file->print_name = strbuf_detach(&pname, NULL);
2413 }
2414
2415 static void print_stat_summary_inserts_deletes(struct diff_options *options,
2416 int files, int insertions, int deletions)
2417 {
2418 struct strbuf sb = STRBUF_INIT;
2419
2420 if (!files) {
2421 assert(insertions == 0 && deletions == 0);
2422 emit_diff_symbol(options, DIFF_SYMBOL_STATS_SUMMARY_NO_FILES,
2423 NULL, 0, 0);
2424 return;
2425 }
2426
2427 strbuf_addf(&sb,
2428 (files == 1) ? " %d file changed" : " %d files changed",
2429 files);
2430
2431 /*
2432 * For binary diff, the caller may want to print "x files
2433 * changed" with insertions == 0 && deletions == 0.
2434 *
2435 * Not omitting "0 insertions(+), 0 deletions(-)" in this case
2436 * is probably less confusing (i.e skip over "2 files changed
2437 * but nothing about added/removed lines? Is this a bug in Git?").
2438 */
2439 if (insertions || deletions == 0) {
2440 strbuf_addf(&sb,
2441 (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
2442 insertions);
2443 }
2444
2445 if (deletions || insertions == 0) {
2446 strbuf_addf(&sb,
2447 (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
2448 deletions);
2449 }
2450 strbuf_addch(&sb, '\n');
2451 emit_diff_symbol(options, DIFF_SYMBOL_STATS_SUMMARY_INSERTS_DELETES,
2452 sb.buf, sb.len, 0);
2453 strbuf_release(&sb);
2454 }
2455
2456 void print_stat_summary(FILE *fp, int files,
2457 int insertions, int deletions)
2458 {
2459 struct diff_options o;
2460 memset(&o, 0, sizeof(o));
2461 o.file = fp;
2462
2463 print_stat_summary_inserts_deletes(&o, files, insertions, deletions);
2464 }
2465
2466 static void show_stats(struct diffstat_t *data, struct diff_options *options)
2467 {
2468 int i, len, add, del, adds = 0, dels = 0;
2469 uintmax_t max_change = 0, max_len = 0;
2470 int total_files = data->nr, count;
2471 int width, name_width, graph_width, number_width = 0, bin_width = 0;
2472 const char *reset, *add_c, *del_c;
2473 int extra_shown = 0;
2474 const char *line_prefix = diff_line_prefix(options);
2475 struct strbuf out = STRBUF_INIT;
2476
2477 if (data->nr == 0)
2478 return;
2479
2480 count = options->stat_count ? options->stat_count : data->nr;
2481
2482 reset = diff_get_color_opt(options, DIFF_RESET);
2483 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
2484 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
2485
2486 /*
2487 * Find the longest filename and max number of changes
2488 */
2489 for (i = 0; (i < count) && (i < data->nr); i++) {
2490 struct diffstat_file *file = data->files[i];
2491 uintmax_t change = file->added + file->deleted;
2492
2493 if (!file->is_interesting && (change == 0)) {
2494 count++; /* not shown == room for one more */
2495 continue;
2496 }
2497 fill_print_name(file);
2498 len = strlen(file->print_name);
2499 if (max_len < len)
2500 max_len = len;
2501
2502 if (file->is_unmerged) {
2503 /* "Unmerged" is 8 characters */
2504 bin_width = bin_width < 8 ? 8 : bin_width;
2505 continue;
2506 }
2507 if (file->is_binary) {
2508 /* "Bin XXX -> YYY bytes" */
2509 int w = 14 + decimal_width(file->added)
2510 + decimal_width(file->deleted);
2511 bin_width = bin_width < w ? w : bin_width;
2512 /* Display change counts aligned with "Bin" */
2513 number_width = 3;
2514 continue;
2515 }
2516
2517 if (max_change < change)
2518 max_change = change;
2519 }
2520 count = i; /* where we can stop scanning in data->files[] */
2521
2522 /*
2523 * We have width = stat_width or term_columns() columns total.
2524 * We want a maximum of min(max_len, stat_name_width) for the name part.
2525 * We want a maximum of min(max_change, stat_graph_width) for the +- part.
2526 * We also need 1 for " " and 4 + decimal_width(max_change)
2527 * for " | NNNN " and one the empty column at the end, altogether
2528 * 6 + decimal_width(max_change).
2529 *
2530 * If there's not enough space, we will use the smaller of
2531 * stat_name_width (if set) and 5/8*width for the filename,
2532 * and the rest for constant elements + graph part, but no more
2533 * than stat_graph_width for the graph part.
2534 * (5/8 gives 50 for filename and 30 for the constant parts + graph
2535 * for the standard terminal size).
2536 *
2537 * In other words: stat_width limits the maximum width, and
2538 * stat_name_width fixes the maximum width of the filename,
2539 * and is also used to divide available columns if there
2540 * aren't enough.
2541 *
2542 * Binary files are displayed with "Bin XXX -> YYY bytes"
2543 * instead of the change count and graph. This part is treated
2544 * similarly to the graph part, except that it is not
2545 * "scaled". If total width is too small to accommodate the
2546 * guaranteed minimum width of the filename part and the
2547 * separators and this message, this message will "overflow"
2548 * making the line longer than the maximum width.
2549 */
2550
2551 if (options->stat_width == -1)
2552 width = term_columns() - strlen(line_prefix);
2553 else
2554 width = options->stat_width ? options->stat_width : 80;
2555 number_width = decimal_width(max_change) > number_width ?
2556 decimal_width(max_change) : number_width;
2557
2558 if (options->stat_graph_width == -1)
2559 options->stat_graph_width = diff_stat_graph_width;
2560
2561 /*
2562 * Guarantee 3/8*16==6 for the graph part
2563 * and 5/8*16==10 for the filename part
2564 */
2565 if (width < 16 + 6 + number_width)
2566 width = 16 + 6 + number_width;
2567
2568 /*
2569 * First assign sizes that are wanted, ignoring available width.
2570 * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
2571 * starting from "XXX" should fit in graph_width.
2572 */
2573 graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
2574 if (options->stat_graph_width &&
2575 options->stat_graph_width < graph_width)
2576 graph_width = options->stat_graph_width;
2577
2578 name_width = (options->stat_name_width > 0 &&
2579 options->stat_name_width < max_len) ?
2580 options->stat_name_width : max_len;
2581
2582 /*
2583 * Adjust adjustable widths not to exceed maximum width
2584 */
2585 if (name_width + number_width + 6 + graph_width > width) {
2586 if (graph_width > width * 3/8 - number_width - 6) {
2587 graph_width = width * 3/8 - number_width - 6;
2588 if (graph_width < 6)
2589 graph_width = 6;
2590 }
2591
2592 if (options->stat_graph_width &&
2593 graph_width > options->stat_graph_width)
2594 graph_width = options->stat_graph_width;
2595 if (name_width > width - number_width - 6 - graph_width)
2596 name_width = width - number_width - 6 - graph_width;
2597 else
2598 graph_width = width - number_width - 6 - name_width;
2599 }
2600
2601 /*
2602 * From here name_width is the width of the name area,
2603 * and graph_width is the width of the graph area.
2604 * max_change is used to scale graph properly.
2605 */
2606 for (i = 0; i < count; i++) {
2607 const char *prefix = "";
2608 struct diffstat_file *file = data->files[i];
2609 char *name = file->print_name;
2610 uintmax_t added = file->added;
2611 uintmax_t deleted = file->deleted;
2612 int name_len;
2613
2614 if (!file->is_interesting && (added + deleted == 0))
2615 continue;
2616
2617 /*
2618 * "scale" the filename
2619 */
2620 len = name_width;
2621 name_len = strlen(name);
2622 if (name_width < name_len) {
2623 char *slash;
2624 prefix = "...";
2625 len -= 3;
2626 name += name_len - len;
2627 slash = strchr(name, '/');
2628 if (slash)
2629 name = slash;
2630 }
2631
2632 if (file->is_binary) {
2633 strbuf_addf(&out, " %s%-*s |", prefix, len, name);
2634 strbuf_addf(&out, " %*s", number_width, "Bin");
2635 if (!added && !deleted) {
2636 strbuf_addch(&out, '\n');
2637 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2638 out.buf, out.len, 0);
2639 strbuf_reset(&out);
2640 continue;
2641 }
2642 strbuf_addf(&out, " %s%"PRIuMAX"%s",
2643 del_c, deleted, reset);
2644 strbuf_addstr(&out, " -> ");
2645 strbuf_addf(&out, "%s%"PRIuMAX"%s",
2646 add_c, added, reset);
2647 strbuf_addstr(&out, " bytes\n");
2648 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2649 out.buf, out.len, 0);
2650 strbuf_reset(&out);
2651 continue;
2652 }
2653 else if (file->is_unmerged) {
2654 strbuf_addf(&out, " %s%-*s |", prefix, len, name);
2655 strbuf_addstr(&out, " Unmerged\n");
2656 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2657 out.buf, out.len, 0);
2658 strbuf_reset(&out);
2659 continue;
2660 }
2661
2662 /*
2663 * scale the add/delete
2664 */
2665 add = added;
2666 del = deleted;
2667
2668 if (graph_width <= max_change) {
2669 int total = scale_linear(add + del, graph_width, max_change);
2670 if (total < 2 && add && del)
2671 /* width >= 2 due to the sanity check */
2672 total = 2;
2673 if (add < del) {
2674 add = scale_linear(add, graph_width, max_change);
2675 del = total - add;
2676 } else {
2677 del = scale_linear(del, graph_width, max_change);
2678 add = total - del;
2679 }
2680 }
2681 strbuf_addf(&out, " %s%-*s |", prefix, len, name);
2682 strbuf_addf(&out, " %*"PRIuMAX"%s",
2683 number_width, added + deleted,
2684 added + deleted ? " " : "");
2685 show_graph(&out, '+', add, add_c, reset);
2686 show_graph(&out, '-', del, del_c, reset);
2687 strbuf_addch(&out, '\n');
2688 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2689 out.buf, out.len, 0);
2690 strbuf_reset(&out);
2691 }
2692
2693 for (i = 0; i < data->nr; i++) {
2694 struct diffstat_file *file = data->files[i];
2695 uintmax_t added = file->added;
2696 uintmax_t deleted = file->deleted;
2697
2698 if (file->is_unmerged ||
2699 (!file->is_interesting && (added + deleted == 0))) {
2700 total_files--;
2701 continue;
2702 }
2703
2704 if (!file->is_binary) {
2705 adds += added;
2706 dels += deleted;
2707 }
2708 if (i < count)
2709 continue;
2710 if (!extra_shown)
2711 emit_diff_symbol(options,
2712 DIFF_SYMBOL_STATS_SUMMARY_ABBREV,
2713 NULL, 0, 0);
2714 extra_shown = 1;
2715 }
2716
2717 print_stat_summary_inserts_deletes(options, total_files, adds, dels);
2718 strbuf_release(&out);
2719 }
2720
2721 static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
2722 {
2723 int i, adds = 0, dels = 0, total_files = data->nr;
2724
2725 if (data->nr == 0)
2726 return;
2727
2728 for (i = 0; i < data->nr; i++) {
2729 int added = data->files[i]->added;
2730 int deleted = data->files[i]->deleted;
2731
2732 if (data->files[i]->is_unmerged ||
2733 (!data->files[i]->is_interesting && (added + deleted == 0))) {
2734 total_files--;
2735 } else if (!data->files[i]->is_binary) { /* don't count bytes */
2736 adds += added;
2737 dels += deleted;
2738 }
2739 }
2740 print_stat_summary_inserts_deletes(options, total_files, adds, dels);
2741 }
2742
2743 static void show_numstat(struct diffstat_t *data, struct diff_options *options)
2744 {
2745 int i;
2746
2747 if (data->nr == 0)
2748 return;
2749
2750 for (i = 0; i < data->nr; i++) {
2751 struct diffstat_file *file = data->files[i];
2752
2753 fprintf(options->file, "%s", diff_line_prefix(options));
2754
2755 if (file->is_binary)
2756 fprintf(options->file, "-\t-\t");
2757 else
2758 fprintf(options->file,
2759 "%"PRIuMAX"\t%"PRIuMAX"\t",
2760 file->added, file->deleted);
2761 if (options->line_termination) {
2762 fill_print_name(file);
2763 if (!file->is_renamed)
2764 write_name_quoted(file->name, options->file,
2765 options->line_termination);
2766 else {
2767 fputs(file->print_name, options->file);
2768 putc(options->line_termination, options->file);
2769 }
2770 } else {
2771 if (file->is_renamed) {
2772 putc('\0', options->file);
2773 write_name_quoted(file->from_name, options->file, '\0');
2774 }
2775 write_name_quoted(file->name, options->file, '\0');
2776 }
2777 }
2778 }
2779
2780 struct dirstat_file {
2781 const char *name;
2782 unsigned long changed;
2783 };
2784
2785 struct dirstat_dir {
2786 struct dirstat_file *files;
2787 int alloc, nr, permille, cumulative;
2788 };
2789
2790 static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
2791 unsigned long changed, const char *base, int baselen)
2792 {
2793 unsigned long sum_changes = 0;
2794 unsigned int sources = 0;
2795 const char *line_prefix = diff_line_prefix(opt);
2796
2797 while (dir->nr) {
2798 struct dirstat_file *f = dir->files;
2799 int namelen = strlen(f->name);
2800 unsigned long changes;
2801 char *slash;
2802
2803 if (namelen < baselen)
2804 break;
2805 if (memcmp(f->name, base, baselen))
2806 break;
2807 slash = strchr(f->name + baselen, '/');
2808 if (slash) {
2809 int newbaselen = slash + 1 - f->name;
2810 changes = gather_dirstat(opt, dir, changed, f->name, newbaselen);
2811 sources++;
2812 } else {
2813 changes = f->changed;
2814 dir->files++;
2815 dir->nr--;
2816 sources += 2;
2817 }
2818 sum_changes += changes;
2819 }
2820
2821 /*
2822 * We don't report dirstat's for
2823 * - the top level
2824 * - or cases where everything came from a single directory
2825 * under this directory (sources == 1).
2826 */
2827 if (baselen && sources != 1) {
2828 if (sum_changes) {
2829 int permille = sum_changes * 1000 / changed;
2830 if (permille >= dir->permille) {
2831 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
2832 permille / 10, permille % 10, baselen, base);
2833 if (!dir->cumulative)
2834 return 0;
2835 }
2836 }
2837 }
2838 return sum_changes;
2839 }
2840
2841 static int dirstat_compare(const void *_a, const void *_b)
2842 {
2843 const struct dirstat_file *a = _a;
2844 const struct dirstat_file *b = _b;
2845 return strcmp(a->name, b->name);
2846 }
2847
2848 static void show_dirstat(struct diff_options *options)
2849 {
2850 int i;
2851 unsigned long changed;
2852 struct dirstat_dir dir;
2853 struct diff_queue_struct *q = &diff_queued_diff;
2854
2855 dir.files = NULL;
2856 dir.alloc = 0;
2857 dir.nr = 0;
2858 dir.permille = options->dirstat_permille;
2859 dir.cumulative = options->flags.dirstat_cumulative;
2860
2861 changed = 0;
2862 for (i = 0; i < q->nr; i++) {
2863 struct diff_filepair *p = q->queue[i];
2864 const char *name;
2865 unsigned long copied, added, damage;
2866 int content_changed;
2867
2868 name = p->two->path ? p->two->path : p->one->path;
2869
2870 if (p->one->oid_valid && p->two->oid_valid)
2871 content_changed = oidcmp(&p->one->oid, &p->two->oid);
2872 else
2873 content_changed = 1;
2874
2875 if (!content_changed) {
2876 /*
2877 * The SHA1 has not changed, so pre-/post-content is
2878 * identical. We can therefore skip looking at the
2879 * file contents altogether.
2880 */
2881 damage = 0;
2882 goto found_damage;
2883 }
2884
2885 if (options->flags.dirstat_by_file) {
2886 /*
2887 * In --dirstat-by-file mode, we don't really need to
2888 * look at the actual file contents at all.
2889 * The fact that the SHA1 changed is enough for us to
2890 * add this file to the list of results
2891 * (with each file contributing equal damage).
2892 */
2893 damage = 1;
2894 goto found_damage;
2895 }
2896
2897 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
2898 diff_populate_filespec(p->one, 0);
2899 diff_populate_filespec(p->two, 0);
2900 diffcore_count_changes(p->one, p->two, NULL, NULL,
2901 &copied, &added);
2902 diff_free_filespec_data(p->one);
2903 diff_free_filespec_data(p->two);
2904 } else if (DIFF_FILE_VALID(p->one)) {
2905 diff_populate_filespec(p->one, CHECK_SIZE_ONLY);
2906 copied = added = 0;
2907 diff_free_filespec_data(p->one);
2908 } else if (DIFF_FILE_VALID(p->two)) {
2909 diff_populate_filespec(p->two, CHECK_SIZE_ONLY);
2910 copied = 0;
2911 added = p->two->size;
2912 diff_free_filespec_data(p->two);
2913 } else
2914 continue;
2915
2916 /*
2917 * Original minus copied is the removed material,
2918 * added is the new material. They are both damages
2919 * made to the preimage.
2920 * If the resulting damage is zero, we know that
2921 * diffcore_count_changes() considers the two entries to
2922 * be identical, but since content_changed is true, we
2923 * know that there must have been _some_ kind of change,
2924 * so we force all entries to have damage > 0.
2925 */
2926 damage = (p->one->size - copied) + added;
2927 if (!damage)
2928 damage = 1;
2929
2930 found_damage:
2931 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2932 dir.files[dir.nr].name = name;
2933 dir.files[dir.nr].changed = damage;
2934 changed += damage;
2935 dir.nr++;
2936 }
2937
2938 /* This can happen even with many files, if everything was renames */
2939 if (!changed)
2940 return;
2941
2942 /* Show all directories with more than x% of the changes */
2943 QSORT(dir.files, dir.nr, dirstat_compare);
2944 gather_dirstat(options, &dir, changed, "", 0);
2945 }
2946
2947 static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
2948 {
2949 int i;
2950 unsigned long changed;
2951 struct dirstat_dir dir;
2952
2953 if (data->nr == 0)
2954 return;
2955
2956 dir.files = NULL;
2957 dir.alloc = 0;
2958 dir.nr = 0;
2959 dir.permille = options->dirstat_permille;
2960 dir.cumulative = options->flags.dirstat_cumulative;
2961
2962 changed = 0;
2963 for (i = 0; i < data->nr; i++) {
2964 struct diffstat_file *file = data->files[i];
2965 unsigned long damage = file->added + file->deleted;
2966 if (file->is_binary)
2967 /*
2968 * binary files counts bytes, not lines. Must find some
2969 * way to normalize binary bytes vs. textual lines.
2970 * The following heuristic assumes that there are 64
2971 * bytes per "line".
2972 * This is stupid and ugly, but very cheap...
2973 */
2974 damage = DIV_ROUND_UP(damage, 64);
2975 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2976 dir.files[dir.nr].name = file->name;
2977 dir.files[dir.nr].changed = damage;
2978 changed += damage;
2979 dir.nr++;
2980 }
2981
2982 /* This can happen even with many files, if everything was renames */
2983 if (!changed)
2984 return;
2985
2986 /* Show all directories with more than x% of the changes */
2987 QSORT(dir.files, dir.nr, dirstat_compare);
2988 gather_dirstat(options, &dir, changed, "", 0);
2989 }
2990
2991 static void free_diffstat_info(struct diffstat_t *diffstat)
2992 {
2993 int i;
2994 for (i = 0; i < diffstat->nr; i++) {
2995 struct diffstat_file *f = diffstat->files[i];
2996 free(f->print_name);
2997 free(f->name);
2998 free(f->from_name);
2999 free(f);
3000 }
3001 free(diffstat->files);
3002 }
3003
3004 struct checkdiff_t {
3005 const char *filename;
3006 int lineno;
3007 int conflict_marker_size;
3008 struct diff_options *o;
3009 unsigned ws_rule;
3010 unsigned status;
3011 };
3012
3013 static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
3014 {
3015 char firstchar;
3016 int cnt;
3017
3018 if (len < marker_size + 1)
3019 return 0;
3020 firstchar = line[0];
3021 switch (firstchar) {
3022 case '=': case '>': case '<': case '|':
3023 break;
3024 default:
3025 return 0;
3026 }
3027 for (cnt = 1; cnt < marker_size; cnt++)
3028 if (line[cnt] != firstchar)
3029 return 0;
3030 /* line[1] thru line[marker_size-1] are same as firstchar */
3031 if (len < marker_size + 1 || !isspace(line[marker_size]))
3032 return 0;
3033 return 1;
3034 }
3035
3036 static void checkdiff_consume(void *priv, char *line, unsigned long len)
3037 {
3038 struct checkdiff_t *data = priv;
3039 int marker_size = data->conflict_marker_size;
3040 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
3041 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
3042 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
3043 char *err;
3044 const char *line_prefix;
3045
3046 assert(data->o);
3047 line_prefix = diff_line_prefix(data->o);
3048
3049 if (line[0] == '+') {
3050 unsigned bad;
3051 data->lineno++;
3052 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
3053 data->status |= 1;
3054 fprintf(data->o->file,
3055 "%s%s:%d: leftover conflict marker\n",
3056 line_prefix, data->filename, data->lineno);
3057 }
3058 bad = ws_check(line + 1, len - 1, data->ws_rule);
3059 if (!bad)
3060 return;
3061 data->status |= bad;
3062 err = whitespace_error_string(bad);
3063 fprintf(data->o->file, "%s%s:%d: %s.\n",
3064 line_prefix, data->filename, data->lineno, err);
3065 free(err);
3066 emit_line(data->o, set, reset, line, 1);
3067 ws_check_emit(line + 1, len - 1, data->ws_rule,
3068 data->o->file, set, reset, ws);
3069 } else if (line[0] == ' ') {
3070 data->lineno++;
3071 } else if (line[0] == '@') {
3072 char *plus = strchr(line, '+');
3073 if (plus)
3074 data->lineno = strtol(plus, NULL, 10) - 1;
3075 else
3076 die("invalid diff");
3077 }
3078 }
3079
3080 static unsigned char *deflate_it(char *data,
3081 unsigned long size,
3082 unsigned long *result_size)
3083 {
3084 int bound;
3085 unsigned char *deflated;
3086 git_zstream stream;
3087
3088 git_deflate_init(&stream, zlib_compression_level);
3089 bound = git_deflate_bound(&stream, size);
3090 deflated = xmalloc(bound);
3091 stream.next_out = deflated;
3092 stream.avail_out = bound;
3093
3094 stream.next_in = (unsigned char *)data;
3095 stream.avail_in = size;
3096 while (git_deflate(&stream, Z_FINISH) == Z_OK)
3097 ; /* nothing */
3098 git_deflate_end(&stream);
3099 *result_size = stream.total_out;
3100 return deflated;
3101 }
3102
3103 static void emit_binary_diff_body(struct diff_options *o,
3104 mmfile_t *one, mmfile_t *two)
3105 {
3106 void *cp;
3107 void *delta;
3108 void *deflated;
3109 void *data;
3110 unsigned long orig_size;
3111 unsigned long delta_size;
3112 unsigned long deflate_size;
3113 unsigned long data_size;
3114
3115 /* We could do deflated delta, or we could do just deflated two,
3116 * whichever is smaller.
3117 */
3118 delta = NULL;
3119 deflated = deflate_it(two->ptr, two->size, &deflate_size);
3120 if (one->size && two->size) {
3121 delta = diff_delta(one->ptr, one->size,
3122 two->ptr, two->size,
3123 &delta_size, deflate_size);
3124 if (delta) {
3125 void *to_free = delta;
3126 orig_size = delta_size;
3127 delta = deflate_it(delta, delta_size, &delta_size);
3128 free(to_free);
3129 }
3130 }
3131
3132 if (delta && delta_size < deflate_size) {
3133 char *s = xstrfmt("%lu", orig_size);
3134 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_HEADER_DELTA,
3135 s, strlen(s), 0);
3136 free(s);
3137 free(deflated);
3138 data = delta;
3139 data_size = delta_size;
3140 } else {
3141 char *s = xstrfmt("%lu", two->size);
3142 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_HEADER_LITERAL,
3143 s, strlen(s), 0);
3144 free(s);
3145 free(delta);
3146 data = deflated;
3147 data_size = deflate_size;
3148 }
3149
3150 /* emit data encoded in base85 */
3151 cp = data;
3152 while (data_size) {
3153 int len;
3154 int bytes = (52 < data_size) ? 52 : data_size;
3155 char line[71];
3156 data_size -= bytes;
3157 if (bytes <= 26)
3158 line[0] = bytes + 'A' - 1;
3159 else
3160 line[0] = bytes - 26 + 'a' - 1;
3161 encode_85(line + 1, cp, bytes);
3162 cp = (char *) cp + bytes;
3163
3164 len = strlen(line);
3165 line[len++] = '\n';
3166 line[len] = '\0';
3167
3168 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_BODY,
3169 line, len, 0);
3170 }
3171 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_FOOTER, NULL, 0, 0);
3172 free(data);
3173 }
3174
3175 static void emit_binary_diff(struct diff_options *o,
3176 mmfile_t *one, mmfile_t *two)
3177 {
3178 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_HEADER, NULL, 0, 0);
3179 emit_binary_diff_body(o, one, two);
3180 emit_binary_diff_body(o, two, one);
3181 }
3182
3183 int diff_filespec_is_binary(struct diff_filespec *one)
3184 {
3185 if (one->is_binary == -1) {
3186 diff_filespec_load_driver(one);
3187 if (one->driver->binary != -1)
3188 one->is_binary = one->driver->binary;
3189 else {
3190 if (!one->data && DIFF_FILE_VALID(one))
3191 diff_populate_filespec(one, CHECK_BINARY);
3192 if (one->is_binary == -1 && one->data)
3193 one->is_binary = buffer_is_binary(one->data,
3194 one->size);
3195 if (one->is_binary == -1)
3196 one->is_binary = 0;
3197 }
3198 }
3199 return one->is_binary;
3200 }
3201
3202 static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
3203 {
3204 diff_filespec_load_driver(one);
3205 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
3206 }
3207
3208 void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
3209 {
3210 if (!options->a_prefix)
3211 options->a_prefix = a;
3212 if (!options->b_prefix)
3213 options->b_prefix = b;
3214 }
3215
3216 struct userdiff_driver *get_textconv(struct diff_filespec *one)
3217 {
3218 if (!DIFF_FILE_VALID(one))
3219 return NULL;
3220
3221 diff_filespec_load_driver(one);
3222 return userdiff_get_textconv(one->driver);
3223 }
3224
3225 static void builtin_diff(const char *name_a,
3226 const char *name_b,
3227 struct diff_filespec *one,
3228 struct diff_filespec *two,
3229 const char *xfrm_msg,
3230 int must_show_header,
3231 struct diff_options *o,
3232 int complete_rewrite)
3233 {
3234 mmfile_t mf1, mf2;
3235 const char *lbl[2];
3236 char *a_one, *b_two;
3237 const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
3238 const char *reset = diff_get_color_opt(o, DIFF_RESET);
3239 const char *a_prefix, *b_prefix;
3240 struct userdiff_driver *textconv_one = NULL;
3241 struct userdiff_driver *textconv_two = NULL;
3242 struct strbuf header = STRBUF_INIT;
3243 const char *line_prefix = diff_line_prefix(o);
3244
3245 diff_set_mnemonic_prefix(o, "a/", "b/");
3246 if (o->flags.reverse_diff) {
3247 a_prefix = o->b_prefix;
3248 b_prefix = o->a_prefix;
3249 } else {
3250 a_prefix = o->a_prefix;
3251 b_prefix = o->b_prefix;
3252 }
3253
3254 if (o->submodule_format == DIFF_SUBMODULE_LOG &&
3255 (!one->mode || S_ISGITLINK(one->mode)) &&
3256 (!two->mode || S_ISGITLINK(two->mode))) {
3257 show_submodule_summary(o, one->path ? one->path : two->path,
3258 &one->oid, &two->oid,
3259 two->dirty_submodule);
3260 return;
3261 } else if (o->submodule_format == DIFF_SUBMODULE_INLINE_DIFF &&
3262 (!one->mode || S_ISGITLINK(one->mode)) &&
3263 (!two->mode || S_ISGITLINK(two->mode))) {
3264 show_submodule_inline_diff(o, one->path ? one->path : two->path,
3265 &one->oid, &two->oid,
3266 two->dirty_submodule);
3267 return;
3268 }
3269
3270 if (o->flags.allow_textconv) {
3271 textconv_one = get_textconv(one);
3272 textconv_two = get_textconv(two);
3273 }
3274
3275 /* Never use a non-valid filename anywhere if at all possible */
3276 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
3277 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
3278
3279 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
3280 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
3281 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
3282 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
3283 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
3284 if (lbl[0][0] == '/') {
3285 /* /dev/null */
3286 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
3287 if (xfrm_msg)
3288 strbuf_addstr(&header, xfrm_msg);
3289 must_show_header = 1;
3290 }
3291 else if (lbl[1][0] == '/') {
3292 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
3293 if (xfrm_msg)
3294 strbuf_addstr(&header, xfrm_msg);
3295 must_show_header = 1;
3296 }
3297 else {
3298 if (one->mode != two->mode) {
3299 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
3300 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
3301 must_show_header = 1;
3302 }
3303 if (xfrm_msg)
3304 strbuf_addstr(&header, xfrm_msg);
3305
3306 /*
3307 * we do not run diff between different kind
3308 * of objects.
3309 */
3310 if ((one->mode ^ two->mode) & S_IFMT)
3311 goto free_ab_and_return;
3312 if (complete_rewrite &&
3313 (textconv_one || !diff_filespec_is_binary(one)) &&
3314 (textconv_two || !diff_filespec_is_binary(two))) {
3315 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3316 header.buf, header.len, 0);
3317 strbuf_reset(&header);
3318 emit_rewrite_diff(name_a, name_b, one, two,
3319 textconv_one, textconv_two, o);
3320 o->found_changes = 1;
3321 goto free_ab_and_return;
3322 }
3323 }
3324
3325 if (o->irreversible_delete && lbl[1][0] == '/') {
3326 emit_diff_symbol(o, DIFF_SYMBOL_HEADER, header.buf,
3327 header.len, 0);
3328 strbuf_reset(&header);
3329 goto free_ab_and_return;
3330 } else if (!o->flags.text &&
3331 ( (!textconv_one && diff_filespec_is_binary(one)) ||
3332 (!textconv_two && diff_filespec_is_binary(two)) )) {
3333 struct strbuf sb = STRBUF_INIT;
3334 if (!one->data && !two->data &&
3335 S_ISREG(one->mode) && S_ISREG(two->mode) &&
3336 !o->flags.binary) {
3337 if (!oidcmp(&one->oid, &two->oid)) {
3338 if (must_show_header)
3339 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3340 header.buf, header.len,
3341 0);
3342 goto free_ab_and_return;
3343 }
3344 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3345 header.buf, header.len, 0);
3346 strbuf_addf(&sb, "%sBinary files %s and %s differ\n",
3347 diff_line_prefix(o), lbl[0], lbl[1]);
3348 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_FILES,
3349 sb.buf, sb.len, 0);
3350 strbuf_release(&sb);
3351 goto free_ab_and_return;
3352 }
3353 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
3354 die("unable to read files to diff");
3355 /* Quite common confusing case */
3356 if (mf1.size == mf2.size &&
3357 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
3358 if (must_show_header)
3359 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3360 header.buf, header.len, 0);
3361 goto free_ab_and_return;
3362 }
3363 emit_diff_symbol(o, DIFF_SYMBOL_HEADER, header.buf, header.len, 0);
3364 strbuf_reset(&header);
3365 if (o->flags.binary)
3366 emit_binary_diff(o, &mf1, &mf2);
3367 else {
3368 strbuf_addf(&sb, "%sBinary files %s and %s differ\n",
3369 diff_line_prefix(o), lbl[0], lbl[1]);
3370 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_FILES,
3371 sb.buf, sb.len, 0);
3372 strbuf_release(&sb);
3373 }
3374 o->found_changes = 1;
3375 } else {
3376 /* Crazy xdl interfaces.. */
3377 const char *diffopts = getenv("GIT_DIFF_OPTS");
3378 const char *v;
3379 xpparam_t xpp;
3380 xdemitconf_t xecfg;
3381 struct emit_callback ecbdata;
3382 const struct userdiff_funcname *pe;
3383
3384 if (must_show_header) {
3385 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3386 header.buf, header.len, 0);
3387 strbuf_reset(&header);
3388 }
3389
3390 mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
3391 mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
3392
3393 pe = diff_funcname_pattern(one);
3394 if (!pe)
3395 pe = diff_funcname_pattern(two);
3396
3397 memset(&xpp, 0, sizeof(xpp));
3398 memset(&xecfg, 0, sizeof(xecfg));
3399 memset(&ecbdata, 0, sizeof(ecbdata));
3400 ecbdata.label_path = lbl;
3401 ecbdata.color_diff = want_color(o->use_color);
3402 ecbdata.ws_rule = whitespace_rule(name_b);
3403 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
3404 check_blank_at_eof(&mf1, &mf2, &ecbdata);
3405 ecbdata.opt = o;
3406 ecbdata.header = header.len ? &header : NULL;
3407 xpp.flags = o->xdl_opts;
3408 xpp.anchors = o->anchors;
3409 xpp.anchors_nr = o->anchors_nr;
3410 xecfg.ctxlen = o->context;
3411 xecfg.interhunkctxlen = o->interhunkcontext;
3412 xecfg.flags = XDL_EMIT_FUNCNAMES;
3413 if (o->flags.funccontext)
3414 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
3415 if (pe)
3416 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
3417 if (!diffopts)
3418 ;
3419 else if (skip_prefix(diffopts, "--unified=", &v))
3420 xecfg.ctxlen = strtoul(v, NULL, 10);
3421 else if (skip_prefix(diffopts, "-u", &v))
3422 xecfg.ctxlen = strtoul(v, NULL, 10);
3423 if (o->word_diff)
3424 init_diff_words_data(&ecbdata, o, one, two);
3425 if (xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
3426 &xpp, &xecfg))
3427 die("unable to generate diff for %s", one->path);
3428 if (o->word_diff)
3429 free_diff_words_data(&ecbdata);
3430 if (textconv_one)
3431 free(mf1.ptr);
3432 if (textconv_two)
3433 free(mf2.ptr);
3434 xdiff_clear_find_func(&xecfg);
3435 }
3436
3437 free_ab_and_return:
3438 strbuf_release(&header);
3439 diff_free_filespec_data(one);
3440 diff_free_filespec_data(two);
3441 free(a_one);
3442 free(b_two);
3443 return;
3444 }
3445
3446 static char *get_compact_summary(const struct diff_filepair *p, int is_renamed)
3447 {
3448 if (!is_renamed) {
3449 if (p->status == DIFF_STATUS_ADDED) {
3450 if (S_ISLNK(p->two->mode))
3451 return "new +l";
3452 else if ((p->two->mode & 0777) == 0755)
3453 return "new +x";
3454 else
3455 return "new";
3456 } else if (p->status == DIFF_STATUS_DELETED)
3457 return "gone";
3458 }
3459 if (S_ISLNK(p->one->mode) && !S_ISLNK(p->two->mode))
3460 return "mode -l";
3461 else if (!S_ISLNK(p->one->mode) && S_ISLNK(p->two->mode))
3462 return "mode +l";
3463 else if ((p->one->mode & 0777) == 0644 &&
3464 (p->two->mode & 0777) == 0755)
3465 return "mode +x";
3466 else if ((p->one->mode & 0777) == 0755 &&
3467 (p->two->mode & 0777) == 0644)
3468 return "mode -x";
3469 return NULL;
3470 }
3471
3472 static void builtin_diffstat(const char *name_a, const char *name_b,
3473 struct diff_filespec *one,
3474 struct diff_filespec *two,
3475 struct diffstat_t *diffstat,
3476 struct diff_options *o,
3477 struct diff_filepair *p)
3478 {
3479 mmfile_t mf1, mf2;
3480 struct diffstat_file *data;
3481 int same_contents;
3482 int complete_rewrite = 0;
3483
3484 if (!DIFF_PAIR_UNMERGED(p)) {
3485 if (p->status == DIFF_STATUS_MODIFIED && p->score)
3486 complete_rewrite = 1;
3487 }
3488
3489 data = diffstat_add(diffstat, name_a, name_b);
3490 data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
3491 if (o->flags.stat_with_summary)
3492 data->comments = get_compact_summary(p, data->is_renamed);
3493
3494 if (!one || !two) {
3495 data->is_unmerged = 1;
3496 return;
3497 }
3498
3499 same_contents = !oidcmp(&one->oid, &two->oid);
3500
3501 if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
3502 data->is_binary = 1;
3503 if (same_contents) {
3504 data->added = 0;
3505 data->deleted = 0;
3506 } else {
3507 data->added = diff_filespec_size(two);
3508 data->deleted = diff_filespec_size(one);
3509 }
3510 }
3511
3512 else if (complete_rewrite) {
3513 diff_populate_filespec(one, 0);
3514 diff_populate_filespec(two, 0);
3515 data->deleted = count_lines(one->data, one->size);
3516 data->added = count_lines(two->data, two->size);
3517 }
3518
3519 else if (!same_contents) {
3520 /* Crazy xdl interfaces.. */
3521 xpparam_t xpp;
3522 xdemitconf_t xecfg;
3523
3524 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
3525 die("unable to read files to diff");
3526
3527 memset(&xpp, 0, sizeof(xpp));
3528 memset(&xecfg, 0, sizeof(xecfg));
3529 xpp.flags = o->xdl_opts;
3530 xpp.anchors = o->anchors;
3531 xpp.anchors_nr = o->anchors_nr;
3532 xecfg.ctxlen = o->context;
3533 xecfg.interhunkctxlen = o->interhunkcontext;
3534 if (xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
3535 &xpp, &xecfg))
3536 die("unable to generate diffstat for %s", one->path);
3537 }
3538
3539 diff_free_filespec_data(one);
3540 diff_free_filespec_data(two);
3541 }
3542
3543 static void builtin_checkdiff(const char *name_a, const char *name_b,
3544 const char *attr_path,
3545 struct diff_filespec *one,
3546 struct diff_filespec *two,
3547 struct diff_options *o)
3548 {
3549 mmfile_t mf1, mf2;
3550 struct checkdiff_t data;
3551
3552 if (!two)
3553 return;
3554
3555 memset(&data, 0, sizeof(data));
3556 data.filename = name_b ? name_b : name_a;
3557 data.lineno = 0;
3558 data.o = o;
3559 data.ws_rule = whitespace_rule(attr_path);
3560 data.conflict_marker_size = ll_merge_marker_size(attr_path);
3561
3562 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
3563 die("unable to read files to diff");
3564
3565 /*
3566 * All the other codepaths check both sides, but not checking
3567 * the "old" side here is deliberate. We are checking the newly
3568 * introduced changes, and as long as the "new" side is text, we
3569 * can and should check what it introduces.
3570 */
3571 if (diff_filespec_is_binary(two))
3572 goto free_and_return;
3573 else {
3574 /* Crazy xdl interfaces.. */
3575 xpparam_t xpp;
3576 xdemitconf_t xecfg;
3577
3578 memset(&xpp, 0, sizeof(xpp));
3579 memset(&xecfg, 0, sizeof(xecfg));
3580 xecfg.ctxlen = 1; /* at least one context line */
3581 xpp.flags = 0;
3582 if (xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
3583 &xpp, &xecfg))
3584 die("unable to generate checkdiff for %s", one->path);
3585
3586 if (data.ws_rule & WS_BLANK_AT_EOF) {
3587 struct emit_callback ecbdata;
3588 int blank_at_eof;
3589
3590 ecbdata.ws_rule = data.ws_rule;
3591 check_blank_at_eof(&mf1, &mf2, &ecbdata);
3592 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
3593
3594 if (blank_at_eof) {
3595 static char *err;
3596 if (!err)
3597 err = whitespace_error_string(WS_BLANK_AT_EOF);
3598 fprintf(o->file, "%s:%d: %s.\n",
3599 data.filename, blank_at_eof, err);
3600 data.status = 1; /* report errors */
3601 }
3602 }
3603 }
3604 free_and_return:
3605 diff_free_filespec_data(one);
3606 diff_free_filespec_data(two);
3607 if (data.status)
3608 o->flags.check_failed = 1;
3609 }
3610
3611 struct diff_filespec *alloc_filespec(const char *path)
3612 {
3613 struct diff_filespec *spec;
3614
3615 FLEXPTR_ALLOC_STR(spec, path, path);
3616 spec->count = 1;
3617 spec->is_binary = -1;
3618 return spec;
3619 }
3620
3621 void free_filespec(struct diff_filespec *spec)
3622 {
3623 if (!--spec->count) {
3624 diff_free_filespec_data(spec);
3625 free(spec);
3626 }
3627 }
3628
3629 void fill_filespec(struct diff_filespec *spec, const struct object_id *oid,
3630 int oid_valid, unsigned short mode)
3631 {
3632 if (mode) {
3633 spec->mode = canon_mode(mode);
3634 oidcpy(&spec->oid, oid);
3635 spec->oid_valid = oid_valid;
3636 }
3637 }
3638
3639 /*
3640 * Given a name and sha1 pair, if the index tells us the file in
3641 * the work tree has that object contents, return true, so that
3642 * prepare_temp_file() does not have to inflate and extract.
3643 */
3644 static int reuse_worktree_file(const char *name, const struct object_id *oid, int want_file)
3645 {
3646 const struct cache_entry *ce;
3647 struct stat st;
3648 int pos, len;
3649
3650 /*
3651 * We do not read the cache ourselves here, because the
3652 * benchmark with my previous version that always reads cache
3653 * shows that it makes things worse for diff-tree comparing
3654 * two linux-2.6 kernel trees in an already checked out work
3655 * tree. This is because most diff-tree comparisons deal with
3656 * only a small number of files, while reading the cache is
3657 * expensive for a large project, and its cost outweighs the
3658 * savings we get by not inflating the object to a temporary
3659 * file. Practically, this code only helps when we are used
3660 * by diff-cache --cached, which does read the cache before
3661 * calling us.
3662 */
3663 if (!active_cache)
3664 return 0;
3665
3666 /* We want to avoid the working directory if our caller
3667 * doesn't need the data in a normal file, this system
3668 * is rather slow with its stat/open/mmap/close syscalls,
3669 * and the object is contained in a pack file. The pack
3670 * is probably already open and will be faster to obtain
3671 * the data through than the working directory. Loose
3672 * objects however would tend to be slower as they need
3673 * to be individually opened and inflated.
3674 */
3675 if (!FAST_WORKING_DIRECTORY && !want_file && has_object_pack(oid))
3676 return 0;
3677
3678 /*
3679 * Similarly, if we'd have to convert the file contents anyway, that
3680 * makes the optimization not worthwhile.
3681 */
3682 if (!want_file && would_convert_to_git(&the_index, name))
3683 return 0;
3684
3685 len = strlen(name);
3686 pos = cache_name_pos(name, len);
3687 if (pos < 0)
3688 return 0;
3689 ce = active_cache[pos];
3690
3691 /*
3692 * This is not the sha1 we are looking for, or
3693 * unreusable because it is not a regular file.
3694 */
3695 if (oidcmp(oid, &ce->oid) || !S_ISREG(ce->ce_mode))
3696 return 0;
3697
3698 /*
3699 * If ce is marked as "assume unchanged", there is no
3700 * guarantee that work tree matches what we are looking for.
3701 */
3702 if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
3703 return 0;
3704
3705 /*
3706 * If ce matches the file in the work tree, we can reuse it.
3707 */
3708 if (ce_uptodate(ce) ||
3709 (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
3710 return 1;
3711
3712 return 0;
3713 }
3714
3715 static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
3716 {
3717 struct strbuf buf = STRBUF_INIT;
3718 char *dirty = "";
3719
3720 /* Are we looking at the work tree? */
3721 if (s->dirty_submodule)
3722 dirty = "-dirty";
3723
3724 strbuf_addf(&buf, "Subproject commit %s%s\n",
3725 oid_to_hex(&s->oid), dirty);
3726 s->size = buf.len;
3727 if (size_only) {
3728 s->data = NULL;
3729 strbuf_release(&buf);
3730 } else {
3731 s->data = strbuf_detach(&buf, NULL);
3732 s->should_free = 1;
3733 }
3734 return 0;
3735 }
3736
3737 /*
3738 * While doing rename detection and pickaxe operation, we may need to
3739 * grab the data for the blob (or file) for our own in-core comparison.
3740 * diff_filespec has data and size fields for this purpose.
3741 */
3742 int diff_populate_filespec(struct diff_filespec *s, unsigned int flags)
3743 {
3744 int size_only = flags & CHECK_SIZE_ONLY;