8647db3d307c297448c91b328b1dabf19635fb94
[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 GIT_COLOR_FAINT, /* CONTEXT_DIM */
74 GIT_COLOR_FAINT_RED, /* OLD_DIM */
75 GIT_COLOR_FAINT_GREEN, /* NEW_DIM */
76 GIT_COLOR_BOLD, /* CONTEXT_BOLD */
77 GIT_COLOR_BOLD_RED, /* OLD_BOLD */
78 GIT_COLOR_BOLD_GREEN, /* NEW_BOLD */
79 };
80
81 static const char *color_diff_slots[] = {
82 [DIFF_CONTEXT] = "context",
83 [DIFF_METAINFO] = "meta",
84 [DIFF_FRAGINFO] = "frag",
85 [DIFF_FILE_OLD] = "old",
86 [DIFF_FILE_NEW] = "new",
87 [DIFF_COMMIT] = "commit",
88 [DIFF_WHITESPACE] = "whitespace",
89 [DIFF_FUNCINFO] = "func",
90 [DIFF_FILE_OLD_MOVED] = "oldMoved",
91 [DIFF_FILE_OLD_MOVED_ALT] = "oldMovedAlternative",
92 [DIFF_FILE_OLD_MOVED_DIM] = "oldMovedDimmed",
93 [DIFF_FILE_OLD_MOVED_ALT_DIM] = "oldMovedAlternativeDimmed",
94 [DIFF_FILE_NEW_MOVED] = "newMoved",
95 [DIFF_FILE_NEW_MOVED_ALT] = "newMovedAlternative",
96 [DIFF_FILE_NEW_MOVED_DIM] = "newMovedDimmed",
97 [DIFF_FILE_NEW_MOVED_ALT_DIM] = "newMovedAlternativeDimmed",
98 [DIFF_CONTEXT_DIM] = "contextDimmed",
99 [DIFF_FILE_OLD_DIM] = "oldDimmed",
100 [DIFF_FILE_NEW_DIM] = "newDimmed",
101 [DIFF_CONTEXT_BOLD] = "contextBold",
102 [DIFF_FILE_OLD_BOLD] = "oldBold",
103 [DIFF_FILE_NEW_BOLD] = "newBold",
104 };
105
106 static NORETURN void die_want_option(const char *option_name)
107 {
108 die(_("option '%s' requires a value"), option_name);
109 }
110
111 define_list_config_array_extra(color_diff_slots, {"plain"});
112
113 static int parse_diff_color_slot(const char *var)
114 {
115 if (!strcasecmp(var, "plain"))
116 return DIFF_CONTEXT;
117 return LOOKUP_CONFIG(color_diff_slots, var);
118 }
119
120 static int parse_dirstat_params(struct diff_options *options, const char *params_string,
121 struct strbuf *errmsg)
122 {
123 char *params_copy = xstrdup(params_string);
124 struct string_list params = STRING_LIST_INIT_NODUP;
125 int ret = 0;
126 int i;
127
128 if (*params_copy)
129 string_list_split_in_place(&params, params_copy, ',', -1);
130 for (i = 0; i < params.nr; i++) {
131 const char *p = params.items[i].string;
132 if (!strcmp(p, "changes")) {
133 options->flags.dirstat_by_line = 0;
134 options->flags.dirstat_by_file = 0;
135 } else if (!strcmp(p, "lines")) {
136 options->flags.dirstat_by_line = 1;
137 options->flags.dirstat_by_file = 0;
138 } else if (!strcmp(p, "files")) {
139 options->flags.dirstat_by_line = 0;
140 options->flags.dirstat_by_file = 1;
141 } else if (!strcmp(p, "noncumulative")) {
142 options->flags.dirstat_cumulative = 0;
143 } else if (!strcmp(p, "cumulative")) {
144 options->flags.dirstat_cumulative = 1;
145 } else if (isdigit(*p)) {
146 char *end;
147 int permille = strtoul(p, &end, 10) * 10;
148 if (*end == '.' && isdigit(*++end)) {
149 /* only use first digit */
150 permille += *end - '0';
151 /* .. and ignore any further digits */
152 while (isdigit(*++end))
153 ; /* nothing */
154 }
155 if (!*end)
156 options->dirstat_permille = permille;
157 else {
158 strbuf_addf(errmsg, _(" Failed to parse dirstat cut-off percentage '%s'\n"),
159 p);
160 ret++;
161 }
162 } else {
163 strbuf_addf(errmsg, _(" Unknown dirstat parameter '%s'\n"), p);
164 ret++;
165 }
166
167 }
168 string_list_clear(&params, 0);
169 free(params_copy);
170 return ret;
171 }
172
173 static int parse_submodule_params(struct diff_options *options, const char *value)
174 {
175 if (!strcmp(value, "log"))
176 options->submodule_format = DIFF_SUBMODULE_LOG;
177 else if (!strcmp(value, "short"))
178 options->submodule_format = DIFF_SUBMODULE_SHORT;
179 else if (!strcmp(value, "diff"))
180 options->submodule_format = DIFF_SUBMODULE_INLINE_DIFF;
181 else
182 return -1;
183 return 0;
184 }
185
186 int git_config_rename(const char *var, const char *value)
187 {
188 if (!value)
189 return DIFF_DETECT_RENAME;
190 if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
191 return DIFF_DETECT_COPY;
192 return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
193 }
194
195 long parse_algorithm_value(const char *value)
196 {
197 if (!value)
198 return -1;
199 else if (!strcasecmp(value, "myers") || !strcasecmp(value, "default"))
200 return 0;
201 else if (!strcasecmp(value, "minimal"))
202 return XDF_NEED_MINIMAL;
203 else if (!strcasecmp(value, "patience"))
204 return XDF_PATIENCE_DIFF;
205 else if (!strcasecmp(value, "histogram"))
206 return XDF_HISTOGRAM_DIFF;
207 return -1;
208 }
209
210 static int parse_one_token(const char **arg, const char *token)
211 {
212 const char *rest;
213 if (skip_prefix(*arg, token, &rest) && (!*rest || *rest == ',')) {
214 *arg = rest;
215 return 1;
216 }
217 return 0;
218 }
219
220 static int parse_ws_error_highlight(const char *arg)
221 {
222 const char *orig_arg = arg;
223 unsigned val = 0;
224
225 while (*arg) {
226 if (parse_one_token(&arg, "none"))
227 val = 0;
228 else if (parse_one_token(&arg, "default"))
229 val = WSEH_NEW;
230 else if (parse_one_token(&arg, "all"))
231 val = WSEH_NEW | WSEH_OLD | WSEH_CONTEXT;
232 else if (parse_one_token(&arg, "new"))
233 val |= WSEH_NEW;
234 else if (parse_one_token(&arg, "old"))
235 val |= WSEH_OLD;
236 else if (parse_one_token(&arg, "context"))
237 val |= WSEH_CONTEXT;
238 else {
239 return -1 - (int)(arg - orig_arg);
240 }
241 if (*arg)
242 arg++;
243 }
244 return val;
245 }
246
247 /*
248 * These are to give UI layer defaults.
249 * The core-level commands such as git-diff-files should
250 * never be affected by the setting of diff.renames
251 * the user happens to have in the configuration file.
252 */
253 void init_diff_ui_defaults(void)
254 {
255 diff_detect_rename_default = DIFF_DETECT_RENAME;
256 }
257
258 int git_diff_heuristic_config(const char *var, const char *value, void *cb)
259 {
260 if (!strcmp(var, "diff.indentheuristic"))
261 diff_indent_heuristic = git_config_bool(var, value);
262 return 0;
263 }
264
265 static int parse_color_moved(const char *arg)
266 {
267 switch (git_parse_maybe_bool(arg)) {
268 case 0:
269 return COLOR_MOVED_NO;
270 case 1:
271 return COLOR_MOVED_DEFAULT;
272 default:
273 break;
274 }
275
276 if (!strcmp(arg, "no"))
277 return COLOR_MOVED_NO;
278 else if (!strcmp(arg, "plain"))
279 return COLOR_MOVED_PLAIN;
280 else if (!strcmp(arg, "blocks"))
281 return COLOR_MOVED_BLOCKS;
282 else if (!strcmp(arg, "zebra"))
283 return COLOR_MOVED_ZEBRA;
284 else if (!strcmp(arg, "default"))
285 return COLOR_MOVED_DEFAULT;
286 else if (!strcmp(arg, "dimmed-zebra"))
287 return COLOR_MOVED_ZEBRA_DIM;
288 else if (!strcmp(arg, "dimmed_zebra"))
289 return COLOR_MOVED_ZEBRA_DIM;
290 else
291 return error(_("color moved setting must be one of 'no', 'default', 'blocks', 'zebra', 'dimmed-zebra', 'plain'"));
292 }
293
294 static int parse_color_moved_ws(const char *arg)
295 {
296 int ret = 0;
297 struct string_list l = STRING_LIST_INIT_DUP;
298 struct string_list_item *i;
299
300 string_list_split(&l, arg, ',', -1);
301
302 for_each_string_list_item(i, &l) {
303 struct strbuf sb = STRBUF_INIT;
304 strbuf_addstr(&sb, i->string);
305 strbuf_trim(&sb);
306
307 if (!strcmp(sb.buf, "ignore-space-change"))
308 ret |= XDF_IGNORE_WHITESPACE_CHANGE;
309 else if (!strcmp(sb.buf, "ignore-space-at-eol"))
310 ret |= XDF_IGNORE_WHITESPACE_AT_EOL;
311 else if (!strcmp(sb.buf, "ignore-all-space"))
312 ret |= XDF_IGNORE_WHITESPACE;
313 else if (!strcmp(sb.buf, "allow-indentation-change"))
314 ret |= COLOR_MOVED_WS_ALLOW_INDENTATION_CHANGE;
315 else
316 error(_("ignoring unknown color-moved-ws mode '%s'"), sb.buf);
317
318 strbuf_release(&sb);
319 }
320
321 if ((ret & COLOR_MOVED_WS_ALLOW_INDENTATION_CHANGE) &&
322 (ret & XDF_WHITESPACE_FLAGS))
323 die(_("color-moved-ws: allow-indentation-change cannot be combined with other white space modes"));
324
325 string_list_clear(&l, 0);
326
327 return ret;
328 }
329
330 int git_diff_ui_config(const char *var, const char *value, void *cb)
331 {
332 if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
333 diff_use_color_default = git_config_colorbool(var, value);
334 return 0;
335 }
336 if (!strcmp(var, "diff.colormoved")) {
337 int cm = parse_color_moved(value);
338 if (cm < 0)
339 return -1;
340 diff_color_moved_default = cm;
341 return 0;
342 }
343 if (!strcmp(var, "diff.colormovedws")) {
344 int cm = parse_color_moved_ws(value);
345 if (cm < 0)
346 return -1;
347 diff_color_moved_ws_default = cm;
348 return 0;
349 }
350 if (!strcmp(var, "diff.context")) {
351 diff_context_default = git_config_int(var, value);
352 if (diff_context_default < 0)
353 return -1;
354 return 0;
355 }
356 if (!strcmp(var, "diff.interhunkcontext")) {
357 diff_interhunk_context_default = git_config_int(var, value);
358 if (diff_interhunk_context_default < 0)
359 return -1;
360 return 0;
361 }
362 if (!strcmp(var, "diff.renames")) {
363 diff_detect_rename_default = git_config_rename(var, value);
364 return 0;
365 }
366 if (!strcmp(var, "diff.autorefreshindex")) {
367 diff_auto_refresh_index = git_config_bool(var, value);
368 return 0;
369 }
370 if (!strcmp(var, "diff.mnemonicprefix")) {
371 diff_mnemonic_prefix = git_config_bool(var, value);
372 return 0;
373 }
374 if (!strcmp(var, "diff.noprefix")) {
375 diff_no_prefix = git_config_bool(var, value);
376 return 0;
377 }
378 if (!strcmp(var, "diff.statgraphwidth")) {
379 diff_stat_graph_width = git_config_int(var, value);
380 return 0;
381 }
382 if (!strcmp(var, "diff.external"))
383 return git_config_string(&external_diff_cmd_cfg, var, value);
384 if (!strcmp(var, "diff.wordregex"))
385 return git_config_string(&diff_word_regex_cfg, var, value);
386 if (!strcmp(var, "diff.orderfile"))
387 return git_config_pathname(&diff_order_file_cfg, var, value);
388
389 if (!strcmp(var, "diff.ignoresubmodules"))
390 handle_ignore_submodules_arg(&default_diff_options, value);
391
392 if (!strcmp(var, "diff.submodule")) {
393 if (parse_submodule_params(&default_diff_options, value))
394 warning(_("Unknown value for 'diff.submodule' config variable: '%s'"),
395 value);
396 return 0;
397 }
398
399 if (!strcmp(var, "diff.algorithm")) {
400 diff_algorithm = parse_algorithm_value(value);
401 if (diff_algorithm < 0)
402 return -1;
403 return 0;
404 }
405
406 if (!strcmp(var, "diff.wserrorhighlight")) {
407 int val = parse_ws_error_highlight(value);
408 if (val < 0)
409 return -1;
410 ws_error_highlight_default = val;
411 return 0;
412 }
413
414 if (git_color_config(var, value, cb) < 0)
415 return -1;
416
417 return git_diff_basic_config(var, value, cb);
418 }
419
420 int git_diff_basic_config(const char *var, const char *value, void *cb)
421 {
422 const char *name;
423
424 if (!strcmp(var, "diff.renamelimit")) {
425 diff_rename_limit_default = git_config_int(var, value);
426 return 0;
427 }
428
429 if (userdiff_config(var, value) < 0)
430 return -1;
431
432 if (skip_prefix(var, "diff.color.", &name) ||
433 skip_prefix(var, "color.diff.", &name)) {
434 int slot = parse_diff_color_slot(name);
435 if (slot < 0)
436 return 0;
437 if (!value)
438 return config_error_nonbool(var);
439 return color_parse(value, diff_colors[slot]);
440 }
441
442 /* like GNU diff's --suppress-blank-empty option */
443 if (!strcmp(var, "diff.suppressblankempty") ||
444 /* for backwards compatibility */
445 !strcmp(var, "diff.suppress-blank-empty")) {
446 diff_suppress_blank_empty = git_config_bool(var, value);
447 return 0;
448 }
449
450 if (!strcmp(var, "diff.dirstat")) {
451 struct strbuf errmsg = STRBUF_INIT;
452 default_diff_options.dirstat_permille = diff_dirstat_permille_default;
453 if (parse_dirstat_params(&default_diff_options, value, &errmsg))
454 warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
455 errmsg.buf);
456 strbuf_release(&errmsg);
457 diff_dirstat_permille_default = default_diff_options.dirstat_permille;
458 return 0;
459 }
460
461 if (git_diff_heuristic_config(var, value, cb) < 0)
462 return -1;
463
464 return git_default_config(var, value, cb);
465 }
466
467 static char *quote_two(const char *one, const char *two)
468 {
469 int need_one = quote_c_style(one, NULL, NULL, 1);
470 int need_two = quote_c_style(two, NULL, NULL, 1);
471 struct strbuf res = STRBUF_INIT;
472
473 if (need_one + need_two) {
474 strbuf_addch(&res, '"');
475 quote_c_style(one, &res, NULL, 1);
476 quote_c_style(two, &res, NULL, 1);
477 strbuf_addch(&res, '"');
478 } else {
479 strbuf_addstr(&res, one);
480 strbuf_addstr(&res, two);
481 }
482 return strbuf_detach(&res, NULL);
483 }
484
485 static const char *external_diff(void)
486 {
487 static const char *external_diff_cmd = NULL;
488 static int done_preparing = 0;
489
490 if (done_preparing)
491 return external_diff_cmd;
492 external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
493 if (!external_diff_cmd)
494 external_diff_cmd = external_diff_cmd_cfg;
495 done_preparing = 1;
496 return external_diff_cmd;
497 }
498
499 /*
500 * Keep track of files used for diffing. Sometimes such an entry
501 * refers to a temporary file, sometimes to an existing file, and
502 * sometimes to "/dev/null".
503 */
504 static struct diff_tempfile {
505 /*
506 * filename external diff should read from, or NULL if this
507 * entry is currently not in use:
508 */
509 const char *name;
510
511 char hex[GIT_MAX_HEXSZ + 1];
512 char mode[10];
513
514 /*
515 * If this diff_tempfile instance refers to a temporary file,
516 * this tempfile object is used to manage its lifetime.
517 */
518 struct tempfile *tempfile;
519 } diff_temp[2];
520
521 struct emit_callback {
522 int color_diff;
523 unsigned ws_rule;
524 int blank_at_eof_in_preimage;
525 int blank_at_eof_in_postimage;
526 int lno_in_preimage;
527 int lno_in_postimage;
528 const char **label_path;
529 struct diff_words_data *diff_words;
530 struct diff_options *opt;
531 struct strbuf *header;
532 };
533
534 static int count_lines(const char *data, int size)
535 {
536 int count, ch, completely_empty = 1, nl_just_seen = 0;
537 count = 0;
538 while (0 < size--) {
539 ch = *data++;
540 if (ch == '\n') {
541 count++;
542 nl_just_seen = 1;
543 completely_empty = 0;
544 }
545 else {
546 nl_just_seen = 0;
547 completely_empty = 0;
548 }
549 }
550 if (completely_empty)
551 return 0;
552 if (!nl_just_seen)
553 count++; /* no trailing newline */
554 return count;
555 }
556
557 static int fill_mmfile(struct repository *r, mmfile_t *mf,
558 struct diff_filespec *one)
559 {
560 if (!DIFF_FILE_VALID(one)) {
561 mf->ptr = (char *)""; /* does not matter */
562 mf->size = 0;
563 return 0;
564 }
565 else if (diff_populate_filespec(r, one, 0))
566 return -1;
567
568 mf->ptr = one->data;
569 mf->size = one->size;
570 return 0;
571 }
572
573 /* like fill_mmfile, but only for size, so we can avoid retrieving blob */
574 static unsigned long diff_filespec_size(struct repository *r,
575 struct diff_filespec *one)
576 {
577 if (!DIFF_FILE_VALID(one))
578 return 0;
579 diff_populate_filespec(r, one, CHECK_SIZE_ONLY);
580 return one->size;
581 }
582
583 static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
584 {
585 char *ptr = mf->ptr;
586 long size = mf->size;
587 int cnt = 0;
588
589 if (!size)
590 return cnt;
591 ptr += size - 1; /* pointing at the very end */
592 if (*ptr != '\n')
593 ; /* incomplete line */
594 else
595 ptr--; /* skip the last LF */
596 while (mf->ptr < ptr) {
597 char *prev_eol;
598 for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
599 if (*prev_eol == '\n')
600 break;
601 if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
602 break;
603 cnt++;
604 ptr = prev_eol - 1;
605 }
606 return cnt;
607 }
608
609 static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
610 struct emit_callback *ecbdata)
611 {
612 int l1, l2, at;
613 unsigned ws_rule = ecbdata->ws_rule;
614 l1 = count_trailing_blank(mf1, ws_rule);
615 l2 = count_trailing_blank(mf2, ws_rule);
616 if (l2 <= l1) {
617 ecbdata->blank_at_eof_in_preimage = 0;
618 ecbdata->blank_at_eof_in_postimage = 0;
619 return;
620 }
621 at = count_lines(mf1->ptr, mf1->size);
622 ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
623
624 at = count_lines(mf2->ptr, mf2->size);
625 ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
626 }
627
628 static void emit_line_0(struct diff_options *o,
629 const char *set_sign, const char *set, unsigned reverse, const char *reset,
630 int first, const char *line, int len)
631 {
632 int has_trailing_newline, has_trailing_carriage_return;
633 int needs_reset = 0; /* at the end of the line */
634 FILE *file = o->file;
635
636 fputs(diff_line_prefix(o), file);
637
638 has_trailing_newline = (len > 0 && line[len-1] == '\n');
639 if (has_trailing_newline)
640 len--;
641
642 has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
643 if (has_trailing_carriage_return)
644 len--;
645
646 if (!len && !first)
647 goto end_of_line;
648
649 if (reverse && want_color(o->use_color)) {
650 fputs(GIT_COLOR_REVERSE, file);
651 needs_reset = 1;
652 }
653
654 if (set_sign) {
655 fputs(set_sign, file);
656 needs_reset = 1;
657 }
658
659 if (first)
660 fputc(first, file);
661
662 if (!len)
663 goto end_of_line;
664
665 if (set) {
666 if (set_sign && set != set_sign)
667 fputs(reset, file);
668 fputs(set, file);
669 needs_reset = 1;
670 }
671 fwrite(line, len, 1, file);
672 needs_reset = 1; /* 'line' may contain color codes. */
673
674 end_of_line:
675 if (needs_reset)
676 fputs(reset, file);
677 if (has_trailing_carriage_return)
678 fputc('\r', file);
679 if (has_trailing_newline)
680 fputc('\n', file);
681 }
682
683 static void emit_line(struct diff_options *o, const char *set, const char *reset,
684 const char *line, int len)
685 {
686 emit_line_0(o, set, NULL, 0, reset, 0, line, len);
687 }
688
689 enum diff_symbol {
690 DIFF_SYMBOL_BINARY_DIFF_HEADER,
691 DIFF_SYMBOL_BINARY_DIFF_HEADER_DELTA,
692 DIFF_SYMBOL_BINARY_DIFF_HEADER_LITERAL,
693 DIFF_SYMBOL_BINARY_DIFF_BODY,
694 DIFF_SYMBOL_BINARY_DIFF_FOOTER,
695 DIFF_SYMBOL_STATS_SUMMARY_NO_FILES,
696 DIFF_SYMBOL_STATS_SUMMARY_ABBREV,
697 DIFF_SYMBOL_STATS_SUMMARY_INSERTS_DELETES,
698 DIFF_SYMBOL_STATS_LINE,
699 DIFF_SYMBOL_WORD_DIFF,
700 DIFF_SYMBOL_STAT_SEP,
701 DIFF_SYMBOL_SUMMARY,
702 DIFF_SYMBOL_SUBMODULE_ADD,
703 DIFF_SYMBOL_SUBMODULE_DEL,
704 DIFF_SYMBOL_SUBMODULE_UNTRACKED,
705 DIFF_SYMBOL_SUBMODULE_MODIFIED,
706 DIFF_SYMBOL_SUBMODULE_HEADER,
707 DIFF_SYMBOL_SUBMODULE_ERROR,
708 DIFF_SYMBOL_SUBMODULE_PIPETHROUGH,
709 DIFF_SYMBOL_REWRITE_DIFF,
710 DIFF_SYMBOL_BINARY_FILES,
711 DIFF_SYMBOL_HEADER,
712 DIFF_SYMBOL_FILEPAIR_PLUS,
713 DIFF_SYMBOL_FILEPAIR_MINUS,
714 DIFF_SYMBOL_WORDS_PORCELAIN,
715 DIFF_SYMBOL_WORDS,
716 DIFF_SYMBOL_CONTEXT,
717 DIFF_SYMBOL_CONTEXT_INCOMPLETE,
718 DIFF_SYMBOL_PLUS,
719 DIFF_SYMBOL_MINUS,
720 DIFF_SYMBOL_NO_LF_EOF,
721 DIFF_SYMBOL_CONTEXT_FRAGINFO,
722 DIFF_SYMBOL_CONTEXT_MARKER,
723 DIFF_SYMBOL_SEPARATOR
724 };
725 /*
726 * Flags for content lines:
727 * 0..12 are whitespace rules
728 * 13-15 are WSEH_NEW | WSEH_OLD | WSEH_CONTEXT
729 * 16 is marking if the line is blank at EOF
730 */
731 #define DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF (1<<16)
732 #define DIFF_SYMBOL_MOVED_LINE (1<<17)
733 #define DIFF_SYMBOL_MOVED_LINE_ALT (1<<18)
734 #define DIFF_SYMBOL_MOVED_LINE_UNINTERESTING (1<<19)
735 #define DIFF_SYMBOL_CONTENT_WS_MASK (WSEH_NEW | WSEH_OLD | WSEH_CONTEXT | WS_RULE_MASK)
736
737 /*
738 * This struct is used when we need to buffer the output of the diff output.
739 *
740 * NEEDSWORK: Instead of storing a copy of the line, add an offset pointer
741 * into the pre/post image file. This pointer could be a union with the
742 * line pointer. By storing an offset into the file instead of the literal line,
743 * we can decrease the memory footprint for the buffered output. At first we
744 * may want to only have indirection for the content lines, but we could also
745 * enhance the state for emitting prefabricated lines, e.g. the similarity
746 * score line or hunk/file headers would only need to store a number or path
747 * and then the output can be constructed later on depending on state.
748 */
749 struct emitted_diff_symbol {
750 const char *line;
751 int len;
752 int flags;
753 enum diff_symbol s;
754 };
755 #define EMITTED_DIFF_SYMBOL_INIT {NULL}
756
757 struct emitted_diff_symbols {
758 struct emitted_diff_symbol *buf;
759 int nr, alloc;
760 };
761 #define EMITTED_DIFF_SYMBOLS_INIT {NULL, 0, 0}
762
763 static void append_emitted_diff_symbol(struct diff_options *o,
764 struct emitted_diff_symbol *e)
765 {
766 struct emitted_diff_symbol *f;
767
768 ALLOC_GROW(o->emitted_symbols->buf,
769 o->emitted_symbols->nr + 1,
770 o->emitted_symbols->alloc);
771 f = &o->emitted_symbols->buf[o->emitted_symbols->nr++];
772
773 memcpy(f, e, sizeof(struct emitted_diff_symbol));
774 f->line = e->line ? xmemdupz(e->line, e->len) : NULL;
775 }
776
777 struct moved_entry {
778 struct hashmap_entry ent;
779 const struct emitted_diff_symbol *es;
780 struct moved_entry *next_line;
781 };
782
783 /**
784 * The struct ws_delta holds white space differences between moved lines, i.e.
785 * between '+' and '-' lines that have been detected to be a move.
786 * The string contains the difference in leading white spaces, before the
787 * rest of the line is compared using the white space config for move
788 * coloring. The current_longer indicates if the first string in the
789 * comparision is longer than the second.
790 */
791 struct ws_delta {
792 char *string;
793 unsigned int current_longer : 1;
794 };
795 #define WS_DELTA_INIT { NULL, 0 }
796
797 struct moved_block {
798 struct moved_entry *match;
799 struct ws_delta wsd;
800 };
801
802 static void moved_block_clear(struct moved_block *b)
803 {
804 FREE_AND_NULL(b->wsd.string);
805 b->match = NULL;
806 }
807
808 static int compute_ws_delta(const struct emitted_diff_symbol *a,
809 const struct emitted_diff_symbol *b,
810 struct ws_delta *out)
811 {
812 const struct emitted_diff_symbol *longer = a->len > b->len ? a : b;
813 const struct emitted_diff_symbol *shorter = a->len > b->len ? b : a;
814 int d = longer->len - shorter->len;
815
816 if (strncmp(longer->line + d, shorter->line, shorter->len))
817 return 0;
818
819 out->string = xmemdupz(longer->line, d);
820 out->current_longer = (a == longer);
821
822 return 1;
823 }
824
825 static int cmp_in_block_with_wsd(const struct diff_options *o,
826 const struct moved_entry *cur,
827 const struct moved_entry *match,
828 struct moved_block *pmb,
829 int n)
830 {
831 struct emitted_diff_symbol *l = &o->emitted_symbols->buf[n];
832 int al = cur->es->len, cl = l->len;
833 const char *a = cur->es->line,
834 *b = match->es->line,
835 *c = l->line;
836
837 int wslen;
838
839 /*
840 * We need to check if 'cur' is equal to 'match'.
841 * As those are from the same (+/-) side, we do not need to adjust for
842 * indent changes. However these were found using fuzzy matching
843 * so we do have to check if they are equal.
844 */
845 if (strcmp(a, b))
846 return 1;
847
848 if (!pmb->wsd.string)
849 /*
850 * The white space delta is not active? This can happen
851 * when we exit early in this function.
852 */
853 return 1;
854
855 /*
856 * The indent changes of the block are known and stored in
857 * pmb->wsd; however we need to check if the indent changes of the
858 * current line are still the same as before.
859 *
860 * To do so we need to compare 'l' to 'cur', adjusting the
861 * one of them for the white spaces, depending which was longer.
862 */
863
864 wslen = strlen(pmb->wsd.string);
865 if (pmb->wsd.current_longer) {
866 c += wslen;
867 cl -= wslen;
868 } else {
869 a += wslen;
870 al -= wslen;
871 }
872
873 if (al != cl || memcmp(a, c, al))
874 return 1;
875
876 return 0;
877 }
878
879 static int moved_entry_cmp(const void *hashmap_cmp_fn_data,
880 const void *entry,
881 const void *entry_or_key,
882 const void *keydata)
883 {
884 const struct diff_options *diffopt = hashmap_cmp_fn_data;
885 const struct moved_entry *a = entry;
886 const struct moved_entry *b = entry_or_key;
887 unsigned flags = diffopt->color_moved_ws_handling
888 & XDF_WHITESPACE_FLAGS;
889
890 if (diffopt->color_moved_ws_handling &
891 COLOR_MOVED_WS_ALLOW_INDENTATION_CHANGE)
892 /*
893 * As there is not specific white space config given,
894 * we'd need to check for a new block, so ignore all
895 * white space. The setup of the white space
896 * configuration for the next block is done else where
897 */
898 flags |= XDF_IGNORE_WHITESPACE;
899
900 return !xdiff_compare_lines(a->es->line, a->es->len,
901 b->es->line, b->es->len,
902 flags);
903 }
904
905 static struct moved_entry *prepare_entry(struct diff_options *o,
906 int line_no)
907 {
908 struct moved_entry *ret = xmalloc(sizeof(*ret));
909 struct emitted_diff_symbol *l = &o->emitted_symbols->buf[line_no];
910 unsigned flags = o->color_moved_ws_handling & XDF_WHITESPACE_FLAGS;
911
912 ret->ent.hash = xdiff_hash_string(l->line, l->len, flags);
913 ret->es = l;
914 ret->next_line = NULL;
915
916 return ret;
917 }
918
919 static void add_lines_to_move_detection(struct diff_options *o,
920 struct hashmap *add_lines,
921 struct hashmap *del_lines)
922 {
923 struct moved_entry *prev_line = NULL;
924
925 int n;
926 for (n = 0; n < o->emitted_symbols->nr; n++) {
927 struct hashmap *hm;
928 struct moved_entry *key;
929
930 switch (o->emitted_symbols->buf[n].s) {
931 case DIFF_SYMBOL_PLUS:
932 hm = add_lines;
933 break;
934 case DIFF_SYMBOL_MINUS:
935 hm = del_lines;
936 break;
937 default:
938 prev_line = NULL;
939 continue;
940 }
941
942 key = prepare_entry(o, n);
943 if (prev_line && prev_line->es->s == o->emitted_symbols->buf[n].s)
944 prev_line->next_line = key;
945
946 hashmap_add(hm, key);
947 prev_line = key;
948 }
949 }
950
951 static void pmb_advance_or_null(struct diff_options *o,
952 struct moved_entry *match,
953 struct hashmap *hm,
954 struct moved_block *pmb,
955 int pmb_nr)
956 {
957 int i;
958 for (i = 0; i < pmb_nr; i++) {
959 struct moved_entry *prev = pmb[i].match;
960 struct moved_entry *cur = (prev && prev->next_line) ?
961 prev->next_line : NULL;
962 if (cur && !hm->cmpfn(o, cur, match, NULL)) {
963 pmb[i].match = cur;
964 } else {
965 pmb[i].match = NULL;
966 }
967 }
968 }
969
970 static void pmb_advance_or_null_multi_match(struct diff_options *o,
971 struct moved_entry *match,
972 struct hashmap *hm,
973 struct moved_block *pmb,
974 int pmb_nr, int n)
975 {
976 int i;
977 char *got_match = xcalloc(1, pmb_nr);
978
979 for (; match; match = hashmap_get_next(hm, match)) {
980 for (i = 0; i < pmb_nr; i++) {
981 struct moved_entry *prev = pmb[i].match;
982 struct moved_entry *cur = (prev && prev->next_line) ?
983 prev->next_line : NULL;
984 if (!cur)
985 continue;
986 if (!cmp_in_block_with_wsd(o, cur, match, &pmb[i], n))
987 got_match[i] |= 1;
988 }
989 }
990
991 for (i = 0; i < pmb_nr; i++) {
992 if (got_match[i]) {
993 /* Advance to the next line */
994 pmb[i].match = pmb[i].match->next_line;
995 } else {
996 moved_block_clear(&pmb[i]);
997 }
998 }
999
1000 free(got_match);
1001 }
1002
1003 static int shrink_potential_moved_blocks(struct moved_block *pmb,
1004 int pmb_nr)
1005 {
1006 int lp, rp;
1007
1008 /* Shrink the set of potential block to the remaining running */
1009 for (lp = 0, rp = pmb_nr - 1; lp <= rp;) {
1010 while (lp < pmb_nr && pmb[lp].match)
1011 lp++;
1012 /* lp points at the first NULL now */
1013
1014 while (rp > -1 && !pmb[rp].match)
1015 rp--;
1016 /* rp points at the last non-NULL */
1017
1018 if (lp < pmb_nr && rp > -1 && lp < rp) {
1019 pmb[lp] = pmb[rp];
1020 pmb[rp].match = NULL;
1021 pmb[rp].wsd.string = NULL;
1022 rp--;
1023 lp++;
1024 }
1025 }
1026
1027 /* Remember the number of running sets */
1028 return rp + 1;
1029 }
1030
1031 /*
1032 * If o->color_moved is COLOR_MOVED_PLAIN, this function does nothing.
1033 *
1034 * Otherwise, if the last block has fewer alphanumeric characters than
1035 * COLOR_MOVED_MIN_ALNUM_COUNT, unset DIFF_SYMBOL_MOVED_LINE on all lines in
1036 * that block.
1037 *
1038 * The last block consists of the (n - block_length)'th line up to but not
1039 * including the nth line.
1040 *
1041 * NEEDSWORK: This uses the same heuristic as blame_entry_score() in blame.c.
1042 * Think of a way to unify them.
1043 */
1044 static void adjust_last_block(struct diff_options *o, int n, int block_length)
1045 {
1046 int i, alnum_count = 0;
1047 if (o->color_moved == COLOR_MOVED_PLAIN)
1048 return;
1049 for (i = 1; i < block_length + 1; i++) {
1050 const char *c = o->emitted_symbols->buf[n - i].line;
1051 for (; *c; c++) {
1052 if (!isalnum(*c))
1053 continue;
1054 alnum_count++;
1055 if (alnum_count >= COLOR_MOVED_MIN_ALNUM_COUNT)
1056 return;
1057 }
1058 }
1059 for (i = 1; i < block_length + 1; i++)
1060 o->emitted_symbols->buf[n - i].flags &= ~DIFF_SYMBOL_MOVED_LINE;
1061 }
1062
1063 /* Find blocks of moved code, delegate actual coloring decision to helper */
1064 static void mark_color_as_moved(struct diff_options *o,
1065 struct hashmap *add_lines,
1066 struct hashmap *del_lines)
1067 {
1068 struct moved_block *pmb = NULL; /* potentially moved blocks */
1069 int pmb_nr = 0, pmb_alloc = 0;
1070 int n, flipped_block = 1, block_length = 0;
1071
1072
1073 for (n = 0; n < o->emitted_symbols->nr; n++) {
1074 struct hashmap *hm = NULL;
1075 struct moved_entry *key;
1076 struct moved_entry *match = NULL;
1077 struct emitted_diff_symbol *l = &o->emitted_symbols->buf[n];
1078
1079 switch (l->s) {
1080 case DIFF_SYMBOL_PLUS:
1081 hm = del_lines;
1082 key = prepare_entry(o, n);
1083 match = hashmap_get(hm, key, NULL);
1084 free(key);
1085 break;
1086 case DIFF_SYMBOL_MINUS:
1087 hm = add_lines;
1088 key = prepare_entry(o, n);
1089 match = hashmap_get(hm, key, NULL);
1090 free(key);
1091 break;
1092 default:
1093 flipped_block = 1;
1094 }
1095
1096 if (!match) {
1097 int i;
1098
1099 adjust_last_block(o, n, block_length);
1100 for(i = 0; i < pmb_nr; i++)
1101 moved_block_clear(&pmb[i]);
1102 pmb_nr = 0;
1103 block_length = 0;
1104 continue;
1105 }
1106
1107 l->flags |= DIFF_SYMBOL_MOVED_LINE;
1108
1109 if (o->color_moved == COLOR_MOVED_PLAIN)
1110 continue;
1111
1112 if (o->color_moved_ws_handling &
1113 COLOR_MOVED_WS_ALLOW_INDENTATION_CHANGE)
1114 pmb_advance_or_null_multi_match(o, match, hm, pmb, pmb_nr, n);
1115 else
1116 pmb_advance_or_null(o, match, hm, pmb, pmb_nr);
1117
1118 pmb_nr = shrink_potential_moved_blocks(pmb, pmb_nr);
1119
1120 if (pmb_nr == 0) {
1121 /*
1122 * The current line is the start of a new block.
1123 * Setup the set of potential blocks.
1124 */
1125 for (; match; match = hashmap_get_next(hm, match)) {
1126 ALLOC_GROW(pmb, pmb_nr + 1, pmb_alloc);
1127 if (o->color_moved_ws_handling &
1128 COLOR_MOVED_WS_ALLOW_INDENTATION_CHANGE) {
1129 if (compute_ws_delta(l, match->es,
1130 &pmb[pmb_nr].wsd))
1131 pmb[pmb_nr++].match = match;
1132 } else {
1133 pmb[pmb_nr].wsd.string = NULL;
1134 pmb[pmb_nr++].match = match;
1135 }
1136 }
1137
1138 flipped_block = (flipped_block + 1) % 2;
1139
1140 adjust_last_block(o, n, block_length);
1141 block_length = 0;
1142 }
1143
1144 block_length++;
1145
1146 if (flipped_block && o->color_moved != COLOR_MOVED_BLOCKS)
1147 l->flags |= DIFF_SYMBOL_MOVED_LINE_ALT;
1148 }
1149 adjust_last_block(o, n, block_length);
1150
1151 for(n = 0; n < pmb_nr; n++)
1152 moved_block_clear(&pmb[n]);
1153 free(pmb);
1154 }
1155
1156 #define DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK \
1157 (DIFF_SYMBOL_MOVED_LINE | DIFF_SYMBOL_MOVED_LINE_ALT)
1158 static void dim_moved_lines(struct diff_options *o)
1159 {
1160 int n;
1161 for (n = 0; n < o->emitted_symbols->nr; n++) {
1162 struct emitted_diff_symbol *prev = (n != 0) ?
1163 &o->emitted_symbols->buf[n - 1] : NULL;
1164 struct emitted_diff_symbol *l = &o->emitted_symbols->buf[n];
1165 struct emitted_diff_symbol *next =
1166 (n < o->emitted_symbols->nr - 1) ?
1167 &o->emitted_symbols->buf[n + 1] : NULL;
1168
1169 /* Not a plus or minus line? */
1170 if (l->s != DIFF_SYMBOL_PLUS && l->s != DIFF_SYMBOL_MINUS)
1171 continue;
1172
1173 /* Not a moved line? */
1174 if (!(l->flags & DIFF_SYMBOL_MOVED_LINE))
1175 continue;
1176
1177 /*
1178 * If prev or next are not a plus or minus line,
1179 * pretend they don't exist
1180 */
1181 if (prev && prev->s != DIFF_SYMBOL_PLUS &&
1182 prev->s != DIFF_SYMBOL_MINUS)
1183 prev = NULL;
1184 if (next && next->s != DIFF_SYMBOL_PLUS &&
1185 next->s != DIFF_SYMBOL_MINUS)
1186 next = NULL;
1187
1188 /* Inside a block? */
1189 if ((prev &&
1190 (prev->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK) ==
1191 (l->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK)) &&
1192 (next &&
1193 (next->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK) ==
1194 (l->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK))) {
1195 l->flags |= DIFF_SYMBOL_MOVED_LINE_UNINTERESTING;
1196 continue;
1197 }
1198
1199 /* Check if we are at an interesting bound: */
1200 if (prev && (prev->flags & DIFF_SYMBOL_MOVED_LINE) &&
1201 (prev->flags & DIFF_SYMBOL_MOVED_LINE_ALT) !=
1202 (l->flags & DIFF_SYMBOL_MOVED_LINE_ALT))
1203 continue;
1204 if (next && (next->flags & DIFF_SYMBOL_MOVED_LINE) &&
1205 (next->flags & DIFF_SYMBOL_MOVED_LINE_ALT) !=
1206 (l->flags & DIFF_SYMBOL_MOVED_LINE_ALT))
1207 continue;
1208
1209 /*
1210 * The boundary to prev and next are not interesting,
1211 * so this line is not interesting as a whole
1212 */
1213 l->flags |= DIFF_SYMBOL_MOVED_LINE_UNINTERESTING;
1214 }
1215 }
1216
1217 static void emit_line_ws_markup(struct diff_options *o,
1218 const char *set_sign, const char *set,
1219 const char *reset,
1220 int sign_index, const char *line, int len,
1221 unsigned ws_rule, int blank_at_eof)
1222 {
1223 const char *ws = NULL;
1224 int sign = o->output_indicators[sign_index];
1225
1226 if (o->ws_error_highlight & ws_rule) {
1227 ws = diff_get_color_opt(o, DIFF_WHITESPACE);
1228 if (!*ws)
1229 ws = NULL;
1230 }
1231
1232 if (!ws && !set_sign)
1233 emit_line_0(o, set, NULL, 0, reset, sign, line, len);
1234 else if (!ws) {
1235 emit_line_0(o, set_sign, set, !!set_sign, reset, sign, line, len);
1236 } else if (blank_at_eof)
1237 /* Blank line at EOF - paint '+' as well */
1238 emit_line_0(o, ws, NULL, 0, reset, sign, line, len);
1239 else {
1240 /* Emit just the prefix, then the rest. */
1241 emit_line_0(o, set_sign ? set_sign : set, NULL, !!set_sign, reset,
1242 sign, "", 0);
1243 ws_check_emit(line, len, ws_rule,
1244 o->file, set, reset, ws);
1245 }
1246 }
1247
1248 static void emit_diff_symbol_from_struct(struct diff_options *o,
1249 struct emitted_diff_symbol *eds)
1250 {
1251 static const char *nneof = " No newline at end of file\n";
1252 const char *context, *reset, *set, *set_sign, *meta, *fraginfo;
1253 struct strbuf sb = STRBUF_INIT;
1254
1255 enum diff_symbol s = eds->s;
1256 const char *line = eds->line;
1257 int len = eds->len;
1258 unsigned flags = eds->flags;
1259
1260 switch (s) {
1261 case DIFF_SYMBOL_NO_LF_EOF:
1262 context = diff_get_color_opt(o, DIFF_CONTEXT);
1263 reset = diff_get_color_opt(o, DIFF_RESET);
1264 putc('\n', o->file);
1265 emit_line_0(o, context, NULL, 0, reset, '\\',
1266 nneof, strlen(nneof));
1267 break;
1268 case DIFF_SYMBOL_SUBMODULE_HEADER:
1269 case DIFF_SYMBOL_SUBMODULE_ERROR:
1270 case DIFF_SYMBOL_SUBMODULE_PIPETHROUGH:
1271 case DIFF_SYMBOL_STATS_SUMMARY_INSERTS_DELETES:
1272 case DIFF_SYMBOL_SUMMARY:
1273 case DIFF_SYMBOL_STATS_LINE:
1274 case DIFF_SYMBOL_BINARY_DIFF_BODY:
1275 case DIFF_SYMBOL_CONTEXT_FRAGINFO:
1276 emit_line(o, "", "", line, len);
1277 break;
1278 case DIFF_SYMBOL_CONTEXT_INCOMPLETE:
1279 case DIFF_SYMBOL_CONTEXT_MARKER:
1280 context = diff_get_color_opt(o, DIFF_CONTEXT);
1281 reset = diff_get_color_opt(o, DIFF_RESET);
1282 emit_line(o, context, reset, line, len);
1283 break;
1284 case DIFF_SYMBOL_SEPARATOR:
1285 fprintf(o->file, "%s%c",
1286 diff_line_prefix(o),
1287 o->line_termination);
1288 break;
1289 case DIFF_SYMBOL_CONTEXT:
1290 set = diff_get_color_opt(o, DIFF_CONTEXT);
1291 reset = diff_get_color_opt(o, DIFF_RESET);
1292 set_sign = NULL;
1293 if (o->flags.dual_color_diffed_diffs) {
1294 char c = !len ? 0 : line[0];
1295
1296 if (c == '+')
1297 set = diff_get_color_opt(o, DIFF_FILE_NEW);
1298 else if (c == '@')
1299 set = diff_get_color_opt(o, DIFF_FRAGINFO);
1300 else if (c == '-')
1301 set = diff_get_color_opt(o, DIFF_FILE_OLD);
1302 }
1303 emit_line_ws_markup(o, set_sign, set, reset,
1304 OUTPUT_INDICATOR_CONTEXT, line, len,
1305 flags & (DIFF_SYMBOL_CONTENT_WS_MASK), 0);
1306 break;
1307 case DIFF_SYMBOL_PLUS:
1308 switch (flags & (DIFF_SYMBOL_MOVED_LINE |
1309 DIFF_SYMBOL_MOVED_LINE_ALT |
1310 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING)) {
1311 case DIFF_SYMBOL_MOVED_LINE |
1312 DIFF_SYMBOL_MOVED_LINE_ALT |
1313 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1314 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED_ALT_DIM);
1315 break;
1316 case DIFF_SYMBOL_MOVED_LINE |
1317 DIFF_SYMBOL_MOVED_LINE_ALT:
1318 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED_ALT);
1319 break;
1320 case DIFF_SYMBOL_MOVED_LINE |
1321 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1322 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED_DIM);
1323 break;
1324 case DIFF_SYMBOL_MOVED_LINE:
1325 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED);
1326 break;
1327 default:
1328 set = diff_get_color_opt(o, DIFF_FILE_NEW);
1329 }
1330 reset = diff_get_color_opt(o, DIFF_RESET);
1331 if (!o->flags.dual_color_diffed_diffs)
1332 set_sign = NULL;
1333 else {
1334 char c = !len ? 0 : line[0];
1335
1336 set_sign = set;
1337 if (c == '-')
1338 set = diff_get_color_opt(o, DIFF_FILE_OLD_BOLD);
1339 else if (c == '@')
1340 set = diff_get_color_opt(o, DIFF_FRAGINFO);
1341 else if (c == '+')
1342 set = diff_get_color_opt(o, DIFF_FILE_NEW_BOLD);
1343 else
1344 set = diff_get_color_opt(o, DIFF_CONTEXT_BOLD);
1345 flags &= ~DIFF_SYMBOL_CONTENT_WS_MASK;
1346 }
1347 emit_line_ws_markup(o, set_sign, set, reset,
1348 OUTPUT_INDICATOR_NEW, line, len,
1349 flags & DIFF_SYMBOL_CONTENT_WS_MASK,
1350 flags & DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF);
1351 break;
1352 case DIFF_SYMBOL_MINUS:
1353 switch (flags & (DIFF_SYMBOL_MOVED_LINE |
1354 DIFF_SYMBOL_MOVED_LINE_ALT |
1355 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING)) {
1356 case DIFF_SYMBOL_MOVED_LINE |
1357 DIFF_SYMBOL_MOVED_LINE_ALT |
1358 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1359 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED_ALT_DIM);
1360 break;
1361 case DIFF_SYMBOL_MOVED_LINE |
1362 DIFF_SYMBOL_MOVED_LINE_ALT:
1363 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED_ALT);
1364 break;
1365 case DIFF_SYMBOL_MOVED_LINE |
1366 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1367 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED_DIM);
1368 break;
1369 case DIFF_SYMBOL_MOVED_LINE:
1370 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED);
1371 break;
1372 default:
1373 set = diff_get_color_opt(o, DIFF_FILE_OLD);
1374 }
1375 reset = diff_get_color_opt(o, DIFF_RESET);
1376 if (!o->flags.dual_color_diffed_diffs)
1377 set_sign = NULL;
1378 else {
1379 char c = !len ? 0 : line[0];
1380
1381 set_sign = set;
1382 if (c == '+')
1383 set = diff_get_color_opt(o, DIFF_FILE_NEW_DIM);
1384 else if (c == '@')
1385 set = diff_get_color_opt(o, DIFF_FRAGINFO);
1386 else if (c == '-')
1387 set = diff_get_color_opt(o, DIFF_FILE_OLD_DIM);
1388 else
1389 set = diff_get_color_opt(o, DIFF_CONTEXT_DIM);
1390 }
1391 emit_line_ws_markup(o, set_sign, set, reset,
1392 OUTPUT_INDICATOR_OLD, line, len,
1393 flags & DIFF_SYMBOL_CONTENT_WS_MASK, 0);
1394 break;
1395 case DIFF_SYMBOL_WORDS_PORCELAIN:
1396 context = diff_get_color_opt(o, DIFF_CONTEXT);
1397 reset = diff_get_color_opt(o, DIFF_RESET);
1398 emit_line(o, context, reset, line, len);
1399 fputs("~\n", o->file);
1400 break;
1401 case DIFF_SYMBOL_WORDS:
1402 context = diff_get_color_opt(o, DIFF_CONTEXT);
1403 reset = diff_get_color_opt(o, DIFF_RESET);
1404 /*
1405 * Skip the prefix character, if any. With
1406 * diff_suppress_blank_empty, there may be
1407 * none.
1408 */
1409 if (line[0] != '\n') {
1410 line++;
1411 len--;
1412 }
1413 emit_line(o, context, reset, line, len);
1414 break;
1415 case DIFF_SYMBOL_FILEPAIR_PLUS:
1416 meta = diff_get_color_opt(o, DIFF_METAINFO);
1417 reset = diff_get_color_opt(o, DIFF_RESET);
1418 fprintf(o->file, "%s%s+++ %s%s%s\n", diff_line_prefix(o), meta,
1419 line, reset,
1420 strchr(line, ' ') ? "\t" : "");
1421 break;
1422 case DIFF_SYMBOL_FILEPAIR_MINUS:
1423 meta = diff_get_color_opt(o, DIFF_METAINFO);
1424 reset = diff_get_color_opt(o, DIFF_RESET);
1425 fprintf(o->file, "%s%s--- %s%s%s\n", diff_line_prefix(o), meta,
1426 line, reset,
1427 strchr(line, ' ') ? "\t" : "");
1428 break;
1429 case DIFF_SYMBOL_BINARY_FILES:
1430 case DIFF_SYMBOL_HEADER:
1431 fprintf(o->file, "%s", line);
1432 break;
1433 case DIFF_SYMBOL_BINARY_DIFF_HEADER:
1434 fprintf(o->file, "%sGIT binary patch\n", diff_line_prefix(o));
1435 break;
1436 case DIFF_SYMBOL_BINARY_DIFF_HEADER_DELTA:
1437 fprintf(o->file, "%sdelta %s\n", diff_line_prefix(o), line);
1438 break;
1439 case DIFF_SYMBOL_BINARY_DIFF_HEADER_LITERAL:
1440 fprintf(o->file, "%sliteral %s\n", diff_line_prefix(o), line);
1441 break;
1442 case DIFF_SYMBOL_BINARY_DIFF_FOOTER:
1443 fputs(diff_line_prefix(o), o->file);
1444 fputc('\n', o->file);
1445 break;
1446 case DIFF_SYMBOL_REWRITE_DIFF:
1447 fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
1448 reset = diff_get_color_opt(o, DIFF_RESET);
1449 emit_line(o, fraginfo, reset, line, len);
1450 break;
1451 case DIFF_SYMBOL_SUBMODULE_ADD:
1452 set = diff_get_color_opt(o, DIFF_FILE_NEW);
1453 reset = diff_get_color_opt(o, DIFF_RESET);
1454 emit_line(o, set, reset, line, len);
1455 break;
1456 case DIFF_SYMBOL_SUBMODULE_DEL:
1457 set = diff_get_color_opt(o, DIFF_FILE_OLD);
1458 reset = diff_get_color_opt(o, DIFF_RESET);
1459 emit_line(o, set, reset, line, len);
1460 break;
1461 case DIFF_SYMBOL_SUBMODULE_UNTRACKED:
1462 fprintf(o->file, "%sSubmodule %s contains untracked content\n",
1463 diff_line_prefix(o), line);
1464 break;
1465 case DIFF_SYMBOL_SUBMODULE_MODIFIED:
1466 fprintf(o->file, "%sSubmodule %s contains modified content\n",
1467 diff_line_prefix(o), line);
1468 break;
1469 case DIFF_SYMBOL_STATS_SUMMARY_NO_FILES:
1470 emit_line(o, "", "", " 0 files changed\n",
1471 strlen(" 0 files changed\n"));
1472 break;
1473 case DIFF_SYMBOL_STATS_SUMMARY_ABBREV:
1474 emit_line(o, "", "", " ...\n", strlen(" ...\n"));
1475 break;
1476 case DIFF_SYMBOL_WORD_DIFF:
1477 fprintf(o->file, "%.*s", len, line);
1478 break;
1479 case DIFF_SYMBOL_STAT_SEP:
1480 fputs(o->stat_sep, o->file);
1481 break;
1482 default:
1483 BUG("unknown diff symbol");
1484 }
1485 strbuf_release(&sb);
1486 }
1487
1488 static void emit_diff_symbol(struct diff_options *o, enum diff_symbol s,
1489 const char *line, int len, unsigned flags)
1490 {
1491 struct emitted_diff_symbol e = {line, len, flags, s};
1492
1493 if (o->emitted_symbols)
1494 append_emitted_diff_symbol(o, &e);
1495 else
1496 emit_diff_symbol_from_struct(o, &e);
1497 }
1498
1499 void diff_emit_submodule_del(struct diff_options *o, const char *line)
1500 {
1501 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_DEL, line, strlen(line), 0);
1502 }
1503
1504 void diff_emit_submodule_add(struct diff_options *o, const char *line)
1505 {
1506 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_ADD, line, strlen(line), 0);
1507 }
1508
1509 void diff_emit_submodule_untracked(struct diff_options *o, const char *path)
1510 {
1511 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_UNTRACKED,
1512 path, strlen(path), 0);
1513 }
1514
1515 void diff_emit_submodule_modified(struct diff_options *o, const char *path)
1516 {
1517 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_MODIFIED,
1518 path, strlen(path), 0);
1519 }
1520
1521 void diff_emit_submodule_header(struct diff_options *o, const char *header)
1522 {
1523 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_HEADER,
1524 header, strlen(header), 0);
1525 }
1526
1527 void diff_emit_submodule_error(struct diff_options *o, const char *err)
1528 {
1529 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_ERROR, err, strlen(err), 0);
1530 }
1531
1532 void diff_emit_submodule_pipethrough(struct diff_options *o,
1533 const char *line, int len)
1534 {
1535 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_PIPETHROUGH, line, len, 0);
1536 }
1537
1538 static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
1539 {
1540 if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
1541 ecbdata->blank_at_eof_in_preimage &&
1542 ecbdata->blank_at_eof_in_postimage &&
1543 ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
1544 ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
1545 return 0;
1546 return ws_blank_line(line, len, ecbdata->ws_rule);
1547 }
1548
1549 static void emit_add_line(const char *reset,
1550 struct emit_callback *ecbdata,
1551 const char *line, int len)
1552 {
1553 unsigned flags = WSEH_NEW | ecbdata->ws_rule;
1554 if (new_blank_line_at_eof(ecbdata, line, len))
1555 flags |= DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF;
1556
1557 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_PLUS, line, len, flags);
1558 }
1559
1560 static void emit_del_line(const char *reset,
1561 struct emit_callback *ecbdata,
1562 const char *line, int len)
1563 {
1564 unsigned flags = WSEH_OLD | ecbdata->ws_rule;
1565 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_MINUS, line, len, flags);
1566 }
1567
1568 static void emit_context_line(const char *reset,
1569 struct emit_callback *ecbdata,
1570 const char *line, int len)
1571 {
1572 unsigned flags = WSEH_CONTEXT | ecbdata->ws_rule;
1573 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_CONTEXT, line, len, flags);
1574 }
1575
1576 static void emit_hunk_header(struct emit_callback *ecbdata,
1577 const char *line, int len)
1578 {
1579 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
1580 const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
1581 const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
1582 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1583 const char *reverse = ecbdata->color_diff ? GIT_COLOR_REVERSE : "";
1584 static const char atat[2] = { '@', '@' };
1585 const char *cp, *ep;
1586 struct strbuf msgbuf = STRBUF_INIT;
1587 int org_len = len;
1588 int i = 1;
1589
1590 /*
1591 * As a hunk header must begin with "@@ -<old>, +<new> @@",
1592 * it always is at least 10 bytes long.
1593 */
1594 if (len < 10 ||
1595 memcmp(line, atat, 2) ||
1596 !(ep = memmem(line + 2, len - 2, atat, 2))) {
1597 emit_diff_symbol(ecbdata->opt,
1598 DIFF_SYMBOL_CONTEXT_MARKER, line, len, 0);
1599 return;
1600 }
1601 ep += 2; /* skip over @@ */
1602
1603 /* The hunk header in fraginfo color */
1604 if (ecbdata->opt->flags.dual_color_diffed_diffs)
1605 strbuf_addstr(&msgbuf, reverse);
1606 strbuf_addstr(&msgbuf, frag);
1607 strbuf_add(&msgbuf, line, ep - line);
1608 strbuf_addstr(&msgbuf, reset);
1609
1610 /*
1611 * trailing "\r\n"
1612 */
1613 for ( ; i < 3; i++)
1614 if (line[len - i] == '\r' || line[len - i] == '\n')
1615 len--;
1616
1617 /* blank before the func header */
1618 for (cp = ep; ep - line < len; ep++)
1619 if (*ep != ' ' && *ep != '\t')
1620 break;
1621 if (ep != cp) {
1622 strbuf_addstr(&msgbuf, context);
1623 strbuf_add(&msgbuf, cp, ep - cp);
1624 strbuf_addstr(&msgbuf, reset);
1625 }
1626
1627 if (ep < line + len) {
1628 strbuf_addstr(&msgbuf, func);
1629 strbuf_add(&msgbuf, ep, line + len - ep);
1630 strbuf_addstr(&msgbuf, reset);
1631 }
1632
1633 strbuf_add(&msgbuf, line + len, org_len - len);
1634 strbuf_complete_line(&msgbuf);
1635 emit_diff_symbol(ecbdata->opt,
1636 DIFF_SYMBOL_CONTEXT_FRAGINFO, msgbuf.buf, msgbuf.len, 0);
1637 strbuf_release(&msgbuf);
1638 }
1639
1640 static struct diff_tempfile *claim_diff_tempfile(void) {
1641 int i;
1642 for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
1643 if (!diff_temp[i].name)
1644 return diff_temp + i;
1645 BUG("diff is failing to clean up its tempfiles");
1646 }
1647
1648 static void remove_tempfile(void)
1649 {
1650 int i;
1651 for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
1652 if (is_tempfile_active(diff_temp[i].tempfile))
1653 delete_tempfile(&diff_temp[i].tempfile);
1654 diff_temp[i].name = NULL;
1655 }
1656 }
1657
1658 static void add_line_count(struct strbuf *out, int count)
1659 {
1660 switch (count) {
1661 case 0:
1662 strbuf_addstr(out, "0,0");
1663 break;
1664 case 1:
1665 strbuf_addstr(out, "1");
1666 break;
1667 default:
1668 strbuf_addf(out, "1,%d", count);
1669 break;
1670 }
1671 }
1672
1673 static void emit_rewrite_lines(struct emit_callback *ecb,
1674 int prefix, const char *data, int size)
1675 {
1676 const char *endp = NULL;
1677 const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
1678
1679 while (0 < size) {
1680 int len;
1681
1682 endp = memchr(data, '\n', size);
1683 len = endp ? (endp - data + 1) : size;
1684 if (prefix != '+') {
1685 ecb->lno_in_preimage++;
1686 emit_del_line(reset, ecb, data, len);
1687 } else {
1688 ecb->lno_in_postimage++;
1689 emit_add_line(reset, ecb, data, len);
1690 }
1691 size -= len;
1692 data += len;
1693 }
1694 if (!endp)
1695 emit_diff_symbol(ecb->opt, DIFF_SYMBOL_NO_LF_EOF, NULL, 0, 0);
1696 }
1697
1698 static void emit_rewrite_diff(const char *name_a,
1699 const char *name_b,
1700 struct diff_filespec *one,
1701 struct diff_filespec *two,
1702 struct userdiff_driver *textconv_one,
1703 struct userdiff_driver *textconv_two,
1704 struct diff_options *o)
1705 {
1706 int lc_a, lc_b;
1707 static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
1708 const char *a_prefix, *b_prefix;
1709 char *data_one, *data_two;
1710 size_t size_one, size_two;
1711 struct emit_callback ecbdata;
1712 struct strbuf out = STRBUF_INIT;
1713
1714 if (diff_mnemonic_prefix && o->flags.reverse_diff) {
1715 a_prefix = o->b_prefix;
1716 b_prefix = o->a_prefix;
1717 } else {
1718 a_prefix = o->a_prefix;
1719 b_prefix = o->b_prefix;
1720 }
1721
1722 name_a += (*name_a == '/');
1723 name_b += (*name_b == '/');
1724
1725 strbuf_reset(&a_name);
1726 strbuf_reset(&b_name);
1727 quote_two_c_style(&a_name, a_prefix, name_a, 0);
1728 quote_two_c_style(&b_name, b_prefix, name_b, 0);
1729
1730 size_one = fill_textconv(o->repo, textconv_one, one, &data_one);
1731 size_two = fill_textconv(o->repo, textconv_two, two, &data_two);
1732
1733 memset(&ecbdata, 0, sizeof(ecbdata));
1734 ecbdata.color_diff = want_color(o->use_color);
1735 ecbdata.ws_rule = whitespace_rule(o->repo->index, name_b);
1736 ecbdata.opt = o;
1737 if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
1738 mmfile_t mf1, mf2;
1739 mf1.ptr = (char *)data_one;
1740 mf2.ptr = (char *)data_two;
1741 mf1.size = size_one;
1742 mf2.size = size_two;
1743 check_blank_at_eof(&mf1, &mf2, &ecbdata);
1744 }
1745 ecbdata.lno_in_preimage = 1;
1746 ecbdata.lno_in_postimage = 1;
1747
1748 lc_a = count_lines(data_one, size_one);
1749 lc_b = count_lines(data_two, size_two);
1750
1751 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_MINUS,
1752 a_name.buf, a_name.len, 0);
1753 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_PLUS,
1754 b_name.buf, b_name.len, 0);
1755
1756 strbuf_addstr(&out, "@@ -");
1757 if (!o->irreversible_delete)
1758 add_line_count(&out, lc_a);
1759 else
1760 strbuf_addstr(&out, "?,?");
1761 strbuf_addstr(&out, " +");
1762 add_line_count(&out, lc_b);
1763 strbuf_addstr(&out, " @@\n");
1764 emit_diff_symbol(o, DIFF_SYMBOL_REWRITE_DIFF, out.buf, out.len, 0);
1765 strbuf_release(&out);
1766
1767 if (lc_a && !o->irreversible_delete)
1768 emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
1769 if (lc_b)
1770 emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
1771 if (textconv_one)
1772 free((char *)data_one);
1773 if (textconv_two)
1774 free((char *)data_two);
1775 }
1776
1777 struct diff_words_buffer {
1778 mmfile_t text;
1779 unsigned long alloc;
1780 struct diff_words_orig {
1781 const char *begin, *end;
1782 } *orig;
1783 int orig_nr, orig_alloc;
1784 };
1785
1786 static void diff_words_append(char *line, unsigned long len,
1787 struct diff_words_buffer *buffer)
1788 {
1789 ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
1790 line++;
1791 len--;
1792 memcpy(buffer->text.ptr + buffer->text.size, line, len);
1793 buffer->text.size += len;
1794 buffer->text.ptr[buffer->text.size] = '\0';
1795 }
1796
1797 struct diff_words_style_elem {
1798 const char *prefix;
1799 const char *suffix;
1800 const char *color; /* NULL; filled in by the setup code if
1801 * color is enabled */
1802 };
1803
1804 struct diff_words_style {
1805 enum diff_words_type type;
1806 struct diff_words_style_elem new_word, old_word, ctx;
1807 const char *newline;
1808 };
1809
1810 static struct diff_words_style diff_words_styles[] = {
1811 { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
1812 { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
1813 { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
1814 };
1815
1816 struct diff_words_data {
1817 struct diff_words_buffer minus, plus;
1818 const char *current_plus;
1819 int last_minus;
1820 struct diff_options *opt;
1821 regex_t *word_regex;
1822 enum diff_words_type type;
1823 struct diff_words_style *style;
1824 };
1825
1826 static int fn_out_diff_words_write_helper(struct diff_options *o,
1827 struct diff_words_style_elem *st_el,
1828 const char *newline,
1829 size_t count, const char *buf)
1830 {
1831 int print = 0;
1832 struct strbuf sb = STRBUF_INIT;
1833
1834 while (count) {
1835 char *p = memchr(buf, '\n', count);
1836 if (print)
1837 strbuf_addstr(&sb, diff_line_prefix(o));
1838
1839 if (p != buf) {
1840 const char *reset = st_el->color && *st_el->color ?
1841 GIT_COLOR_RESET : NULL;
1842 if (st_el->color && *st_el->color)
1843 strbuf_addstr(&sb, st_el->color);
1844 strbuf_addstr(&sb, st_el->prefix);
1845 strbuf_add(&sb, buf, p ? p - buf : count);
1846 strbuf_addstr(&sb, st_el->suffix);
1847 if (reset)
1848 strbuf_addstr(&sb, reset);
1849 }
1850 if (!p)
1851 goto out;
1852
1853 strbuf_addstr(&sb, newline);
1854 count -= p + 1 - buf;
1855 buf = p + 1;
1856 print = 1;
1857 if (count) {
1858 emit_diff_symbol(o, DIFF_SYMBOL_WORD_DIFF,
1859 sb.buf, sb.len, 0);
1860 strbuf_reset(&sb);
1861 }
1862 }
1863
1864 out:
1865 if (sb.len)
1866 emit_diff_symbol(o, DIFF_SYMBOL_WORD_DIFF,
1867 sb.buf, sb.len, 0);
1868 strbuf_release(&sb);
1869 return 0;
1870 }
1871
1872 /*
1873 * '--color-words' algorithm can be described as:
1874 *
1875 * 1. collect the minus/plus lines of a diff hunk, divided into
1876 * minus-lines and plus-lines;
1877 *
1878 * 2. break both minus-lines and plus-lines into words and
1879 * place them into two mmfile_t with one word for each line;
1880 *
1881 * 3. use xdiff to run diff on the two mmfile_t to get the words level diff;
1882 *
1883 * And for the common parts of the both file, we output the plus side text.
1884 * diff_words->current_plus is used to trace the current position of the plus file
1885 * which printed. diff_words->last_minus is used to trace the last minus word
1886 * printed.
1887 *
1888 * For '--graph' to work with '--color-words', we need to output the graph prefix
1889 * on each line of color words output. Generally, there are two conditions on
1890 * which we should output the prefix.
1891 *
1892 * 1. diff_words->last_minus == 0 &&
1893 * diff_words->current_plus == diff_words->plus.text.ptr
1894 *
1895 * that is: the plus text must start as a new line, and if there is no minus
1896 * word printed, a graph prefix must be printed.
1897 *
1898 * 2. diff_words->current_plus > diff_words->plus.text.ptr &&
1899 * *(diff_words->current_plus - 1) == '\n'
1900 *
1901 * that is: a graph prefix must be printed following a '\n'
1902 */
1903 static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
1904 {
1905 if ((diff_words->last_minus == 0 &&
1906 diff_words->current_plus == diff_words->plus.text.ptr) ||
1907 (diff_words->current_plus > diff_words->plus.text.ptr &&
1908 *(diff_words->current_plus - 1) == '\n')) {
1909 return 1;
1910 } else {
1911 return 0;
1912 }
1913 }
1914
1915 static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
1916 {
1917 struct diff_words_data *diff_words = priv;
1918 struct diff_words_style *style = diff_words->style;
1919 int minus_first, minus_len, plus_first, plus_len;
1920 const char *minus_begin, *minus_end, *plus_begin, *plus_end;
1921 struct diff_options *opt = diff_words->opt;
1922 const char *line_prefix;
1923
1924 if (line[0] != '@' || parse_hunk_header(line, len,
1925 &minus_first, &minus_len, &plus_first, &plus_len))
1926 return;
1927
1928 assert(opt);
1929 line_prefix = diff_line_prefix(opt);
1930
1931 /* POSIX requires that first be decremented by one if len == 0... */
1932 if (minus_len) {
1933 minus_begin = diff_words->minus.orig[minus_first].begin;
1934 minus_end =
1935 diff_words->minus.orig[minus_first + minus_len - 1].end;
1936 } else
1937 minus_begin = minus_end =
1938 diff_words->minus.orig[minus_first].end;
1939
1940 if (plus_len) {
1941 plus_begin = diff_words->plus.orig[plus_first].begin;
1942 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
1943 } else
1944 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
1945
1946 if (color_words_output_graph_prefix(diff_words)) {
1947 fputs(line_prefix, diff_words->opt->file);
1948 }
1949 if (diff_words->current_plus != plus_begin) {
1950 fn_out_diff_words_write_helper(diff_words->opt,
1951 &style->ctx, style->newline,
1952 plus_begin - diff_words->current_plus,
1953 diff_words->current_plus);
1954 }
1955 if (minus_begin != minus_end) {
1956 fn_out_diff_words_write_helper(diff_words->opt,
1957 &style->old_word, style->newline,
1958 minus_end - minus_begin, minus_begin);
1959 }
1960 if (plus_begin != plus_end) {
1961 fn_out_diff_words_write_helper(diff_words->opt,
1962 &style->new_word, style->newline,
1963 plus_end - plus_begin, plus_begin);
1964 }
1965
1966 diff_words->current_plus = plus_end;
1967 diff_words->last_minus = minus_first;
1968 }
1969
1970 /* This function starts looking at *begin, and returns 0 iff a word was found. */
1971 static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
1972 int *begin, int *end)
1973 {
1974 if (word_regex && *begin < buffer->size) {
1975 regmatch_t match[1];
1976 if (!regexec_buf(word_regex, buffer->ptr + *begin,
1977 buffer->size - *begin, 1, match, 0)) {
1978 char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
1979 '\n', match[0].rm_eo - match[0].rm_so);
1980 *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
1981 *begin += match[0].rm_so;
1982 return *begin >= *end;
1983 }
1984 return -1;
1985 }
1986
1987 /* find the next word */
1988 while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
1989 (*begin)++;
1990 if (*begin >= buffer->size)
1991 return -1;
1992
1993 /* find the end of the word */
1994 *end = *begin + 1;
1995 while (*end < buffer->size && !isspace(buffer->ptr[*end]))
1996 (*end)++;
1997
1998 return 0;
1999 }
2000
2001 /*
2002 * This function splits the words in buffer->text, stores the list with
2003 * newline separator into out, and saves the offsets of the original words
2004 * in buffer->orig.
2005 */
2006 static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
2007 regex_t *word_regex)
2008 {
2009 int i, j;
2010 long alloc = 0;
2011
2012 out->size = 0;
2013 out->ptr = NULL;
2014
2015 /* fake an empty "0th" word */
2016 ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
2017 buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
2018 buffer->orig_nr = 1;
2019
2020 for (i = 0; i < buffer->text.size; i++) {
2021 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
2022 return;
2023
2024 /* store original boundaries */
2025 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
2026 buffer->orig_alloc);
2027 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
2028 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
2029 buffer->orig_nr++;
2030
2031 /* store one word */
2032 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
2033 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
2034 out->ptr[out->size + j - i] = '\n';
2035 out->size += j - i + 1;
2036
2037 i = j - 1;
2038 }
2039 }
2040
2041 /* this executes the word diff on the accumulated buffers */
2042 static void diff_words_show(struct diff_words_data *diff_words)
2043 {
2044 xpparam_t xpp;
2045 xdemitconf_t xecfg;
2046 mmfile_t minus, plus;
2047 struct diff_words_style *style = diff_words->style;
2048
2049 struct diff_options *opt = diff_words->opt;
2050 const char *line_prefix;
2051
2052 assert(opt);
2053 line_prefix = diff_line_prefix(opt);
2054
2055 /* special case: only removal */
2056 if (!diff_words->plus.text.size) {
2057 emit_diff_symbol(diff_words->opt, DIFF_SYMBOL_WORD_DIFF,
2058 line_prefix, strlen(line_prefix), 0);
2059 fn_out_diff_words_write_helper(diff_words->opt,
2060 &style->old_word, style->newline,
2061 diff_words->minus.text.size,
2062 diff_words->minus.text.ptr);
2063 diff_words->minus.text.size = 0;
2064 return;
2065 }
2066
2067 diff_words->current_plus = diff_words->plus.text.ptr;
2068 diff_words->last_minus = 0;
2069
2070 memset(&xpp, 0, sizeof(xpp));
2071 memset(&xecfg, 0, sizeof(xecfg));
2072 diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
2073 diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
2074 xpp.flags = 0;
2075 /* as only the hunk header will be parsed, we need a 0-context */
2076 xecfg.ctxlen = 0;
2077 if (xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
2078 &xpp, &xecfg))
2079 die("unable to generate word diff");
2080 free(minus.ptr);
2081 free(plus.ptr);
2082 if (diff_words->current_plus != diff_words->plus.text.ptr +
2083 diff_words->plus.text.size) {
2084 if (color_words_output_graph_prefix(diff_words))
2085 emit_diff_symbol(diff_words->opt, DIFF_SYMBOL_WORD_DIFF,
2086 line_prefix, strlen(line_prefix), 0);
2087 fn_out_diff_words_write_helper(diff_words->opt,
2088 &style->ctx, style->newline,
2089 diff_words->plus.text.ptr + diff_words->plus.text.size
2090 - diff_words->current_plus, diff_words->current_plus);
2091 }
2092 diff_words->minus.text.size = diff_words->plus.text.size = 0;
2093 }
2094
2095 /* In "color-words" mode, show word-diff of words accumulated in the buffer */
2096 static void diff_words_flush(struct emit_callback *ecbdata)
2097 {
2098 struct diff_options *wo = ecbdata->diff_words->opt;
2099
2100 if (ecbdata->diff_words->minus.text.size ||
2101 ecbdata->diff_words->plus.text.size)
2102 diff_words_show(ecbdata->diff_words);
2103
2104 if (wo->emitted_symbols) {
2105 struct diff_options *o = ecbdata->opt;
2106 struct emitted_diff_symbols *wol = wo->emitted_symbols;
2107 int i;
2108
2109 /*
2110 * NEEDSWORK:
2111 * Instead of appending each, concat all words to a line?
2112 */
2113 for (i = 0; i < wol->nr; i++)
2114 append_emitted_diff_symbol(o, &wol->buf[i]);
2115
2116 for (i = 0; i < wol->nr; i++)
2117 free((void *)wol->buf[i].line);
2118
2119 wol->nr = 0;
2120 }
2121 }
2122
2123 static void diff_filespec_load_driver(struct diff_filespec *one,
2124 struct index_state *istate)
2125 {
2126 /* Use already-loaded driver */
2127 if (one->driver)
2128 return;
2129
2130 if (S_ISREG(one->mode))
2131 one->driver = userdiff_find_by_path(istate, one->path);
2132
2133 /* Fallback to default settings */
2134 if (!one->driver)
2135 one->driver = userdiff_find_by_name("default");
2136 }
2137
2138 static const char *userdiff_word_regex(struct diff_filespec *one,
2139 struct index_state *istate)
2140 {
2141 diff_filespec_load_driver(one, istate);
2142 return one->driver->word_regex;
2143 }
2144
2145 static void init_diff_words_data(struct emit_callback *ecbdata,
2146 struct diff_options *orig_opts,
2147 struct diff_filespec *one,
2148 struct diff_filespec *two)
2149 {
2150 int i;
2151 struct diff_options *o = xmalloc(sizeof(struct diff_options));
2152 memcpy(o, orig_opts, sizeof(struct diff_options));
2153
2154 ecbdata->diff_words =
2155 xcalloc(1, sizeof(struct diff_words_data));
2156 ecbdata->diff_words->type = o->word_diff;
2157 ecbdata->diff_words->opt = o;
2158
2159 if (orig_opts->emitted_symbols)
2160 o->emitted_symbols =
2161 xcalloc(1, sizeof(struct emitted_diff_symbols));
2162
2163 if (!o->word_regex)
2164 o->word_regex = userdiff_word_regex(one, o->repo->index);
2165 if (!o->word_regex)
2166 o->word_regex = userdiff_word_regex(two, o->repo->index);
2167 if (!o->word_regex)
2168 o->word_regex = diff_word_regex_cfg;
2169 if (o->word_regex) {
2170 ecbdata->diff_words->word_regex = (regex_t *)
2171 xmalloc(sizeof(regex_t));
2172 if (regcomp(ecbdata->diff_words->word_regex,
2173 o->word_regex,
2174 REG_EXTENDED | REG_NEWLINE))
2175 die("invalid regular expression: %s",
2176 o->word_regex);
2177 }
2178 for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
2179 if (o->word_diff == diff_words_styles[i].type) {
2180 ecbdata->diff_words->style =
2181 &diff_words_styles[i];
2182 break;
2183 }
2184 }
2185 if (want_color(o->use_color)) {
2186 struct diff_words_style *st = ecbdata->diff_words->style;
2187 st->old_word.color = diff_get_color_opt(o, DIFF_FILE_OLD);
2188 st->new_word.color = diff_get_color_opt(o, DIFF_FILE_NEW);
2189 st->ctx.color = diff_get_color_opt(o, DIFF_CONTEXT);
2190 }
2191 }
2192
2193 static void free_diff_words_data(struct emit_callback *ecbdata)
2194 {
2195 if (ecbdata->diff_words) {
2196 diff_words_flush(ecbdata);
2197 free (ecbdata->diff_words->opt->emitted_symbols);
2198 free (ecbdata->diff_words->opt);
2199 free (ecbdata->diff_words->minus.text.ptr);
2200 free (ecbdata->diff_words->minus.orig);
2201 free (ecbdata->diff_words->plus.text.ptr);
2202 free (ecbdata->diff_words->plus.orig);
2203 if (ecbdata->diff_words->word_regex) {
2204 regfree(ecbdata->diff_words->word_regex);
2205 free(ecbdata->diff_words->word_regex);
2206 }
2207 FREE_AND_NULL(ecbdata->diff_words);
2208 }
2209 }
2210
2211 const char *diff_get_color(int diff_use_color, enum color_diff ix)
2212 {
2213 if (want_color(diff_use_color))
2214 return diff_colors[ix];
2215 return "";
2216 }
2217
2218 const char *diff_line_prefix(struct diff_options *opt)
2219 {
2220 struct strbuf *msgbuf;
2221 if (!opt->output_prefix)
2222 return "";
2223
2224 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
2225 return msgbuf->buf;
2226 }
2227
2228 static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
2229 {
2230 const char *cp;
2231 unsigned long allot;
2232 size_t l = len;
2233
2234 cp = line;
2235 allot = l;
2236 while (0 < l) {
2237 (void) utf8_width(&cp, &l);
2238 if (!cp)
2239 break; /* truncated in the middle? */
2240 }
2241 return allot - l;
2242 }
2243
2244 static void find_lno(const char *line, struct emit_callback *ecbdata)
2245 {
2246 const char *p;
2247 ecbdata->lno_in_preimage = 0;
2248 ecbdata->lno_in_postimage = 0;
2249 p = strchr(line, '-');
2250 if (!p)
2251 return; /* cannot happen */
2252 ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
2253 p = strchr(p, '+');
2254 if (!p)
2255 return; /* cannot happen */
2256 ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
2257 }
2258
2259 static void fn_out_consume(void *priv, char *line, unsigned long len)
2260 {
2261 struct emit_callback *ecbdata = priv;
2262 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
2263 struct diff_options *o = ecbdata->opt;
2264
2265 o->found_changes = 1;
2266
2267 if (ecbdata->header) {
2268 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
2269 ecbdata->header->buf, ecbdata->header->len, 0);
2270 strbuf_reset(ecbdata->header);
2271 ecbdata->header = NULL;
2272 }
2273
2274 if (ecbdata->label_path[0]) {
2275 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_MINUS,
2276 ecbdata->label_path[0],
2277 strlen(ecbdata->label_path[0]), 0);
2278 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_PLUS,
2279 ecbdata->label_path[1],
2280 strlen(ecbdata->label_path[1]), 0);
2281 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
2282 }
2283
2284 if (diff_suppress_blank_empty
2285 && len == 2 && line[0] == ' ' && line[1] == '\n') {
2286 line[0] = '\n';
2287 len = 1;
2288 }
2289
2290 if (line[0] == '@') {
2291 if (ecbdata->diff_words)
2292 diff_words_flush(ecbdata);
2293 len = sane_truncate_line(ecbdata, line, len);
2294 find_lno(line, ecbdata);
2295 emit_hunk_header(ecbdata, line, len);
2296 return;
2297 }
2298
2299 if (ecbdata->diff_words) {
2300 enum diff_symbol s =
2301 ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN ?
2302 DIFF_SYMBOL_WORDS_PORCELAIN : DIFF_SYMBOL_WORDS;
2303 if (line[0] == '-') {
2304 diff_words_append(line, len,
2305 &ecbdata->diff_words->minus);
2306 return;
2307 } else if (line[0] == '+') {
2308 diff_words_append(line, len,
2309 &ecbdata->diff_words->plus);
2310 return;
2311 } else if (starts_with(line, "\\ ")) {
2312 /*
2313 * Eat the "no newline at eof" marker as if we
2314 * saw a "+" or "-" line with nothing on it,
2315 * and return without diff_words_flush() to
2316 * defer processing. If this is the end of
2317 * preimage, more "+" lines may come after it.
2318 */
2319 return;
2320 }
2321 diff_words_flush(ecbdata);
2322 emit_diff_symbol(o, s, line, len, 0);
2323 return;
2324 }
2325
2326 switch (line[0]) {
2327 case '+':
2328 ecbdata->lno_in_postimage++;
2329 emit_add_line(reset, ecbdata, line + 1, len - 1);
2330 break;
2331 case '-':
2332 ecbdata->lno_in_preimage++;
2333 emit_del_line(reset, ecbdata, line + 1, len - 1);
2334 break;
2335 case ' ':
2336 ecbdata->lno_in_postimage++;
2337 ecbdata->lno_in_preimage++;
2338 emit_context_line(reset, ecbdata, line + 1, len - 1);
2339 break;
2340 default:
2341 /* incomplete line at the end */
2342 ecbdata->lno_in_preimage++;
2343 emit_diff_symbol(o, DIFF_SYMBOL_CONTEXT_INCOMPLETE,
2344 line, len, 0);
2345 break;
2346 }
2347 }
2348
2349 static void pprint_rename(struct strbuf *name, const char *a, const char *b)
2350 {
2351 const char *old_name = a;
2352 const char *new_name = b;
2353 int pfx_length, sfx_length;
2354 int pfx_adjust_for_slash;
2355 int len_a = strlen(a);
2356 int len_b = strlen(b);
2357 int a_midlen, b_midlen;
2358 int qlen_a = quote_c_style(a, NULL, NULL, 0);
2359 int qlen_b = quote_c_style(b, NULL, NULL, 0);
2360
2361 if (qlen_a || qlen_b) {
2362 quote_c_style(a, name, NULL, 0);
2363 strbuf_addstr(name, " => ");
2364 quote_c_style(b, name, NULL, 0);
2365 return;
2366 }
2367
2368 /* Find common prefix */
2369 pfx_length = 0;
2370 while (*old_name && *new_name && *old_name == *new_name) {
2371 if (*old_name == '/')
2372 pfx_length = old_name - a + 1;
2373 old_name++;
2374 new_name++;
2375 }
2376
2377 /* Find common suffix */
2378 old_name = a + len_a;
2379 new_name = b + len_b;
2380 sfx_length = 0;
2381 /*
2382 * If there is a common prefix, it must end in a slash. In
2383 * that case we let this loop run 1 into the prefix to see the
2384 * same slash.
2385 *
2386 * If there is no common prefix, we cannot do this as it would
2387 * underrun the input strings.
2388 */
2389 pfx_adjust_for_slash = (pfx_length ? 1 : 0);
2390 while (a + pfx_length - pfx_adjust_for_slash <= old_name &&
2391 b + pfx_length - pfx_adjust_for_slash <= new_name &&
2392 *old_name == *new_name) {
2393 if (*old_name == '/')
2394 sfx_length = len_a - (old_name - a);
2395 old_name--;
2396 new_name--;
2397 }
2398
2399 /*
2400 * pfx{mid-a => mid-b}sfx
2401 * {pfx-a => pfx-b}sfx
2402 * pfx{sfx-a => sfx-b}
2403 * name-a => name-b
2404 */
2405 a_midlen = len_a - pfx_length - sfx_length;
2406 b_midlen = len_b - pfx_length - sfx_length;
2407 if (a_midlen < 0)
2408 a_midlen = 0;
2409 if (b_midlen < 0)
2410 b_midlen = 0;
2411
2412 strbuf_grow(name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
2413 if (pfx_length + sfx_length) {
2414 strbuf_add(name, a, pfx_length);
2415 strbuf_addch(name, '{');
2416 }
2417 strbuf_add(name, a + pfx_length, a_midlen);
2418 strbuf_addstr(name, " => ");
2419 strbuf_add(name, b + pfx_length, b_midlen);
2420 if (pfx_length + sfx_length) {
2421 strbuf_addch(name, '}');
2422 strbuf_add(name, a + len_a - sfx_length, sfx_length);
2423 }
2424 }
2425
2426 struct diffstat_t {
2427 int nr;
2428 int alloc;
2429 struct diffstat_file {
2430 char *from_name;
2431 char *name;
2432 char *print_name;
2433 const char *comments;
2434 unsigned is_unmerged:1;
2435 unsigned is_binary:1;
2436 unsigned is_renamed:1;
2437 unsigned is_interesting:1;
2438 uintmax_t added, deleted;
2439 } **files;
2440 };
2441
2442 static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
2443 const char *name_a,
2444 const char *name_b)
2445 {
2446 struct diffstat_file *x;
2447 x = xcalloc(1, sizeof(*x));
2448 ALLOC_GROW(diffstat->files, diffstat->nr + 1, diffstat->alloc);
2449 diffstat->files[diffstat->nr++] = x;
2450 if (name_b) {
2451 x->from_name = xstrdup(name_a);
2452 x->name = xstrdup(name_b);
2453 x->is_renamed = 1;
2454 }
2455 else {
2456 x->from_name = NULL;
2457 x->name = xstrdup(name_a);
2458 }
2459 return x;
2460 }
2461
2462 static void diffstat_consume(void *priv, char *line, unsigned long len)
2463 {
2464 struct diffstat_t *diffstat = priv;
2465 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
2466
2467 if (line[0] == '+')
2468 x->added++;
2469 else if (line[0] == '-')
2470 x->deleted++;
2471 }
2472
2473 const char mime_boundary_leader[] = "------------";
2474
2475 static int scale_linear(int it, int width, int max_change)
2476 {
2477 if (!it)
2478 return 0;
2479 /*
2480 * make sure that at least one '-' or '+' is printed if
2481 * there is any change to this path. The easiest way is to
2482 * scale linearly as if the alloted width is one column shorter
2483 * than it is, and then add 1 to the result.
2484 */
2485 return 1 + (it * (width - 1) / max_change);
2486 }
2487
2488 static void show_graph(struct strbuf *out, char ch, int cnt,
2489 const char *set, const char *reset)
2490 {
2491 if (cnt <= 0)
2492 return;
2493 strbuf_addstr(out, set);
2494 strbuf_addchars(out, ch, cnt);
2495 strbuf_addstr(out, reset);
2496 }
2497
2498 static void fill_print_name(struct diffstat_file *file)
2499 {
2500 struct strbuf pname = STRBUF_INIT;
2501
2502 if (file->print_name)
2503 return;
2504
2505 if (file->is_renamed)
2506 pprint_rename(&pname, file->from_name, file->name);
2507 else
2508 quote_c_style(file->name, &pname, NULL, 0);
2509
2510 if (file->comments)
2511 strbuf_addf(&pname, " (%s)", file->comments);
2512
2513 file->print_name = strbuf_detach(&pname, NULL);
2514 }
2515
2516 static void print_stat_summary_inserts_deletes(struct diff_options *options,
2517 int files, int insertions, int deletions)
2518 {
2519 struct strbuf sb = STRBUF_INIT;
2520
2521 if (!files) {
2522 assert(insertions == 0 && deletions == 0);
2523 emit_diff_symbol(options, DIFF_SYMBOL_STATS_SUMMARY_NO_FILES,
2524 NULL, 0, 0);
2525 return;
2526 }
2527
2528 strbuf_addf(&sb,
2529 (files == 1) ? " %d file changed" : " %d files changed",
2530 files);
2531
2532 /*
2533 * For binary diff, the caller may want to print "x files
2534 * changed" with insertions == 0 && deletions == 0.
2535 *
2536 * Not omitting "0 insertions(+), 0 deletions(-)" in this case
2537 * is probably less confusing (i.e skip over "2 files changed
2538 * but nothing about added/removed lines? Is this a bug in Git?").
2539 */
2540 if (insertions || deletions == 0) {
2541 strbuf_addf(&sb,
2542 (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
2543 insertions);
2544 }
2545
2546 if (deletions || insertions == 0) {
2547 strbuf_addf(&sb,
2548 (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
2549 deletions);
2550 }
2551 strbuf_addch(&sb, '\n');
2552 emit_diff_symbol(options, DIFF_SYMBOL_STATS_SUMMARY_INSERTS_DELETES,
2553 sb.buf, sb.len, 0);
2554 strbuf_release(&sb);
2555 }
2556
2557 void print_stat_summary(FILE *fp, int files,
2558 int insertions, int deletions)
2559 {
2560 struct diff_options o;
2561 memset(&o, 0, sizeof(o));
2562 o.file = fp;
2563
2564 print_stat_summary_inserts_deletes(&o, files, insertions, deletions);
2565 }
2566
2567 static void show_stats(struct diffstat_t *data, struct diff_options *options)
2568 {
2569 int i, len, add, del, adds = 0, dels = 0;
2570 uintmax_t max_change = 0, max_len = 0;
2571 int total_files = data->nr, count;
2572 int width, name_width, graph_width, number_width = 0, bin_width = 0;
2573 const char *reset, *add_c, *del_c;
2574 int extra_shown = 0;
2575 const char *line_prefix = diff_line_prefix(options);
2576 struct strbuf out = STRBUF_INIT;
2577
2578 if (data->nr == 0)
2579 return;
2580
2581 count = options->stat_count ? options->stat_count : data->nr;
2582
2583 reset = diff_get_color_opt(options, DIFF_RESET);
2584 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
2585 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
2586
2587 /*
2588 * Find the longest filename and max number of changes
2589 */
2590 for (i = 0; (i < count) && (i < data->nr); i++) {
2591 struct diffstat_file *file = data->files[i];
2592 uintmax_t change = file->added + file->deleted;
2593
2594 if (!file->is_interesting && (change == 0)) {
2595 count++; /* not shown == room for one more */
2596 continue;
2597 }
2598 fill_print_name(file);
2599 len = strlen(file->print_name);
2600 if (max_len < len)
2601 max_len = len;
2602
2603 if (file->is_unmerged) {
2604 /* "Unmerged" is 8 characters */
2605 bin_width = bin_width < 8 ? 8 : bin_width;
2606 continue;
2607 }
2608 if (file->is_binary) {
2609 /* "Bin XXX -> YYY bytes" */
2610 int w = 14 + decimal_width(file->added)
2611 + decimal_width(file->deleted);
2612 bin_width = bin_width < w ? w : bin_width;
2613 /* Display change counts aligned with "Bin" */
2614 number_width = 3;
2615 continue;
2616 }
2617
2618 if (max_change < change)
2619 max_change = change;
2620 }
2621 count = i; /* where we can stop scanning in data->files[] */
2622
2623 /*
2624 * We have width = stat_width or term_columns() columns total.
2625 * We want a maximum of min(max_len, stat_name_width) for the name part.
2626 * We want a maximum of min(max_change, stat_graph_width) for the +- part.
2627 * We also need 1 for " " and 4 + decimal_width(max_change)
2628 * for " | NNNN " and one the empty column at the end, altogether
2629 * 6 + decimal_width(max_change).
2630 *
2631 * If there's not enough space, we will use the smaller of
2632 * stat_name_width (if set) and 5/8*width for the filename,
2633 * and the rest for constant elements + graph part, but no more
2634 * than stat_graph_width for the graph part.
2635 * (5/8 gives 50 for filename and 30 for the constant parts + graph
2636 * for the standard terminal size).
2637 *
2638 * In other words: stat_width limits the maximum width, and
2639 * stat_name_width fixes the maximum width of the filename,
2640 * and is also used to divide available columns if there
2641 * aren't enough.
2642 *
2643 * Binary files are displayed with "Bin XXX -> YYY bytes"
2644 * instead of the change count and graph. This part is treated
2645 * similarly to the graph part, except that it is not
2646 * "scaled". If total width is too small to accommodate the
2647 * guaranteed minimum width of the filename part and the
2648 * separators and this message, this message will "overflow"
2649 * making the line longer than the maximum width.
2650 */
2651
2652 if (options->stat_width == -1)
2653 width = term_columns() - strlen(line_prefix);
2654 else
2655 width = options->stat_width ? options->stat_width : 80;
2656 number_width = decimal_width(max_change) > number_width ?
2657 decimal_width(max_change) : number_width;
2658
2659 if (options->stat_graph_width == -1)
2660 options->stat_graph_width = diff_stat_graph_width;
2661
2662 /*
2663 * Guarantee 3/8*16==6 for the graph part
2664 * and 5/8*16==10 for the filename part
2665 */
2666 if (width < 16 + 6 + number_width)
2667 width = 16 + 6 + number_width;
2668
2669 /*
2670 * First assign sizes that are wanted, ignoring available width.
2671 * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
2672 * starting from "XXX" should fit in graph_width.
2673 */
2674 graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
2675 if (options->stat_graph_width &&
2676 options->stat_graph_width < graph_width)
2677 graph_width = options->stat_graph_width;
2678
2679 name_width = (options->stat_name_width > 0 &&
2680 options->stat_name_width < max_len) ?
2681 options->stat_name_width : max_len;
2682
2683 /*
2684 * Adjust adjustable widths not to exceed maximum width
2685 */
2686 if (name_width + number_width + 6 + graph_width > width) {
2687 if (graph_width > width * 3/8 - number_width - 6) {
2688 graph_width = width * 3/8 - number_width - 6;
2689 if (graph_width < 6)
2690 graph_width = 6;
2691 }
2692
2693 if (options->stat_graph_width &&
2694 graph_width > options->stat_graph_width)
2695 graph_width = options->stat_graph_width;
2696 if (name_width > width - number_width - 6 - graph_width)
2697 name_width = width - number_width - 6 - graph_width;
2698 else
2699 graph_width = width - number_width - 6 - name_width;
2700 }
2701
2702 /*
2703 * From here name_width is the width of the name area,
2704 * and graph_width is the width of the graph area.
2705 * max_change is used to scale graph properly.
2706 */
2707 for (i = 0; i < count; i++) {
2708 const char *prefix = "";
2709 struct diffstat_file *file = data->files[i];
2710 char *name = file->print_name;
2711 uintmax_t added = file->added;
2712 uintmax_t deleted = file->deleted;
2713 int name_len;
2714
2715 if (!file->is_interesting && (added + deleted == 0))
2716 continue;
2717
2718 /*
2719 * "scale" the filename
2720 */
2721 len = name_width;
2722 name_len = strlen(name);
2723 if (name_width < name_len) {
2724 char *slash;
2725 prefix = "...";
2726 len -= 3;
2727 name += name_len - len;
2728 slash = strchr(name, '/');
2729 if (slash)
2730 name = slash;
2731 }
2732
2733 if (file->is_binary) {
2734 strbuf_addf(&out, " %s%-*s |", prefix, len, name);
2735 strbuf_addf(&out, " %*s", number_width, "Bin");
2736 if (!added && !deleted) {
2737 strbuf_addch(&out, '\n');
2738 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2739 out.buf, out.len, 0);
2740 strbuf_reset(&out);
2741 continue;
2742 }
2743 strbuf_addf(&out, " %s%"PRIuMAX"%s",
2744 del_c, deleted, reset);
2745 strbuf_addstr(&out, " -> ");
2746 strbuf_addf(&out, "%s%"PRIuMAX"%s",
2747 add_c, added, reset);
2748 strbuf_addstr(&out, " bytes\n");
2749 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2750 out.buf, out.len, 0);
2751 strbuf_reset(&out);
2752 continue;
2753 }
2754 else if (file->is_unmerged) {
2755 strbuf_addf(&out, " %s%-*s |", prefix, len, name);
2756 strbuf_addstr(&out, " Unmerged\n");
2757 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2758 out.buf, out.len, 0);
2759 strbuf_reset(&out);
2760 continue;
2761 }
2762
2763 /*
2764 * scale the add/delete
2765 */
2766 add = added;
2767 del = deleted;
2768
2769 if (graph_width <= max_change) {
2770 int total = scale_linear(add + del, graph_width, max_change);
2771 if (total < 2 && add && del)
2772 /* width >= 2 due to the sanity check */
2773 total = 2;
2774 if (add < del) {
2775 add = scale_linear(add, graph_width, max_change);
2776 del = total - add;
2777 } else {
2778 del = scale_linear(del, graph_width, max_change);
2779 add = total - del;
2780 }
2781 }
2782 strbuf_addf(&out, " %s%-*s |", prefix, len, name);
2783 strbuf_addf(&out, " %*"PRIuMAX"%s",
2784 number_width, added + deleted,
2785 added + deleted ? " " : "");
2786 show_graph(&out, '+', add, add_c, reset);
2787 show_graph(&out, '-', del, del_c, reset);
2788 strbuf_addch(&out, '\n');
2789 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2790 out.buf, out.len, 0);
2791 strbuf_reset(&out);
2792 }
2793
2794 for (i = 0; i < data->nr; i++) {
2795 struct diffstat_file *file = data->files[i];
2796 uintmax_t added = file->added;
2797 uintmax_t deleted = file->deleted;
2798
2799 if (file->is_unmerged ||
2800 (!file->is_interesting && (added + deleted == 0))) {
2801 total_files--;
2802 continue;
2803 }
2804
2805 if (!file->is_binary) {
2806 adds += added;
2807 dels += deleted;
2808 }
2809 if (i < count)
2810 continue;
2811 if (!extra_shown)
2812 emit_diff_symbol(options,
2813 DIFF_SYMBOL_STATS_SUMMARY_ABBREV,
2814 NULL, 0, 0);
2815 extra_shown = 1;
2816 }
2817
2818 print_stat_summary_inserts_deletes(options, total_files, adds, dels);
2819 strbuf_release(&out);
2820 }
2821
2822 static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
2823 {
2824 int i, adds = 0, dels = 0, total_files = data->nr;
2825
2826 if (data->nr == 0)
2827 return;
2828
2829 for (i = 0; i < data->nr; i++) {
2830 int added = data->files[i]->added;
2831 int deleted = data->files[i]->deleted;
2832
2833 if (data->files[i]->is_unmerged ||
2834 (!data->files[i]->is_interesting && (added + deleted == 0))) {
2835 total_files--;
2836 } else if (!data->files[i]->is_binary) { /* don't count bytes */
2837 adds += added;
2838 dels += deleted;
2839 }
2840 }
2841 print_stat_summary_inserts_deletes(options, total_files, adds, dels);
2842 }
2843
2844 static void show_numstat(struct diffstat_t *data, struct diff_options *options)
2845 {
2846 int i;
2847
2848 if (data->nr == 0)
2849 return;
2850
2851 for (i = 0; i < data->nr; i++) {
2852 struct diffstat_file *file = data->files[i];
2853
2854 fprintf(options->file, "%s", diff_line_prefix(options));
2855
2856 if (file->is_binary)
2857 fprintf(options->file, "-\t-\t");
2858 else
2859 fprintf(options->file,
2860 "%"PRIuMAX"\t%"PRIuMAX"\t",
2861 file->added, file->deleted);
2862 if (options->line_termination) {
2863 fill_print_name(file);
2864 if (!file->is_renamed)
2865 write_name_quoted(file->name, options->file,
2866 options->line_termination);
2867 else {
2868 fputs(file->print_name, options->file);
2869 putc(options->line_termination, options->file);
2870 }
2871 } else {
2872 if (file->is_renamed) {
2873 putc('\0', options->file);
2874 write_name_quoted(file->from_name, options->file, '\0');
2875 }
2876 write_name_quoted(file->name, options->file, '\0');
2877 }
2878 }
2879 }
2880
2881 struct dirstat_file {
2882 const char *name;
2883 unsigned long changed;
2884 };
2885
2886 struct dirstat_dir {
2887 struct dirstat_file *files;
2888 int alloc, nr, permille, cumulative;
2889 };
2890
2891 static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
2892 unsigned long changed, const char *base, int baselen)
2893 {
2894 unsigned long sum_changes = 0;
2895 unsigned int sources = 0;
2896 const char *line_prefix = diff_line_prefix(opt);
2897
2898 while (dir->nr) {
2899 struct dirstat_file *f = dir->files;
2900 int namelen = strlen(f->name);
2901 unsigned long changes;
2902 char *slash;
2903
2904 if (namelen < baselen)
2905 break;
2906 if (memcmp(f->name, base, baselen))
2907 break;
2908 slash = strchr(f->name + baselen, '/');
2909 if (slash) {
2910 int newbaselen = slash + 1 - f->name;
2911 changes = gather_dirstat(opt, dir, changed, f->name, newbaselen);
2912 sources++;
2913 } else {
2914 changes = f->changed;
2915 dir->files++;
2916 dir->nr--;
2917 sources += 2;
2918 }
2919 sum_changes += changes;
2920 }
2921
2922 /*
2923 * We don't report dirstat's for
2924 * - the top level
2925 * - or cases where everything came from a single directory
2926 * under this directory (sources == 1).
2927 */
2928 if (baselen && sources != 1) {
2929 if (sum_changes) {
2930 int permille = sum_changes * 1000 / changed;
2931 if (permille >= dir->permille) {
2932 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
2933 permille / 10, permille % 10, baselen, base);
2934 if (!dir->cumulative)
2935 return 0;
2936 }
2937 }
2938 }
2939 return sum_changes;
2940 }
2941
2942 static int dirstat_compare(const void *_a, const void *_b)
2943 {
2944 const struct dirstat_file *a = _a;
2945 const struct dirstat_file *b = _b;
2946 return strcmp(a->name, b->name);
2947 }
2948
2949 static void show_dirstat(struct diff_options *options)
2950 {
2951 int i;
2952 unsigned long changed;
2953 struct dirstat_dir dir;
2954 struct diff_queue_struct *q = &diff_queued_diff;
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 < q->nr; i++) {
2964 struct diff_filepair *p = q->queue[i];
2965 const char *name;
2966 unsigned long copied, added, damage;
2967
2968 name = p->two->path ? p->two->path : p->one->path;
2969
2970 if (p->one->oid_valid && p->two->oid_valid &&
2971 oideq(&p->one->oid, &p->two->oid)) {
2972 /*
2973 * The SHA1 has not changed, so pre-/post-content is
2974 * identical. We can therefore skip looking at the
2975 * file contents altogether.
2976 */
2977 damage = 0;
2978 goto found_damage;
2979 }
2980
2981 if (options->flags.dirstat_by_file) {
2982 /*
2983 * In --dirstat-by-file mode, we don't really need to
2984 * look at the actual file contents at all.
2985 * The fact that the SHA1 changed is enough for us to
2986 * add this file to the list of results
2987 * (with each file contributing equal damage).
2988 */
2989 damage = 1;
2990 goto found_damage;
2991 }
2992
2993 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
2994 diff_populate_filespec(options->repo, p->one, 0);
2995 diff_populate_filespec(options->repo, p->two, 0);
2996 diffcore_count_changes(options->repo,
2997 p->one, p->two, NULL, NULL,
2998 &copied, &added);
2999 diff_free_filespec_data(p->one);
3000 diff_free_filespec_data(p->two);
3001 } else if (DIFF_FILE_VALID(p->one)) {
3002 diff_populate_filespec(options->repo, p->one, CHECK_SIZE_ONLY);
3003 copied = added = 0;
3004 diff_free_filespec_data(p->one);
3005 } else if (DIFF_FILE_VALID(p->two)) {
3006 diff_populate_filespec(options->repo, p->two, CHECK_SIZE_ONLY);
3007 copied = 0;
3008 added = p->two->size;
3009 diff_free_filespec_data(p->two);
3010 } else
3011 continue;
3012
3013 /*
3014 * Original minus copied is the removed material,
3015 * added is the new material. They are both damages
3016 * made to the preimage.
3017 * If the resulting damage is zero, we know that
3018 * diffcore_count_changes() considers the two entries to
3019 * be identical, but since the oid changed, we
3020 * know that there must have been _some_ kind of change,
3021 * so we force all entries to have damage > 0.
3022 */
3023 damage = (p->one->size - copied) + added;
3024 if (!damage)
3025 damage = 1;
3026
3027 found_damage:
3028 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
3029 dir.files[dir.nr].name = name;
3030 dir.files[dir.nr].changed = damage;
3031 changed += damage;
3032 dir.nr++;
3033 }
3034
3035 /* This can happen even with many files, if everything was renames */
3036 if (!changed)
3037 return;
3038
3039 /* Show all directories with more than x% of the changes */
3040 QSORT(dir.files, dir.nr, dirstat_compare);
3041 gather_dirstat(options, &dir, changed, "", 0);
3042 }
3043
3044 static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
3045 {
3046 int i;
3047 unsigned long changed;
3048 struct dirstat_dir dir;
3049
3050 if (data->nr == 0)
3051 return;
3052
3053 dir.files = NULL;
3054 dir.alloc = 0;
3055 dir.nr = 0;
3056 dir.permille = options->dirstat_permille;
3057 dir.cumulative = options->flags.dirstat_cumulative;
3058
3059 changed = 0;
3060 for (i = 0; i < data->nr; i++) {
3061 struct diffstat_file *file = data->files[i];
3062 unsigned long damage = file->added + file->deleted;
3063 if (file->is_binary)
3064 /*
3065 * binary files counts bytes, not lines. Must find some
3066 * way to normalize binary bytes vs. textual lines.
3067 * The following heuristic assumes that there are 64
3068 * bytes per "line".
3069 * This is stupid and ugly, but very cheap...
3070 */
3071 damage = DIV_ROUND_UP(damage, 64);
3072 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
3073 dir.files[dir.nr].name = file->name;
3074 dir.files[dir.nr].changed = damage;
3075 changed += damage;
3076 dir.nr++;
3077 }
3078
3079 /* This can happen even with many files, if everything was renames */
3080 if (!changed)
3081 return;
3082
3083 /* Show all directories with more than x% of the changes */
3084 QSORT(dir.files, dir.nr, dirstat_compare);
3085 gather_dirstat(options, &dir, changed, "", 0);
3086 }
3087
3088 static void free_diffstat_info(struct diffstat_t *diffstat)
3089 {
3090 int i;
3091 for (i = 0; i < diffstat->nr; i++) {
3092 struct diffstat_file *f = diffstat->files[i];
3093 free(f->print_name);
3094 free(f->name);
3095 free(f->from_name);
3096 free(f);
3097 }
3098 free(diffstat->files);
3099 }
3100
3101 struct checkdiff_t {
3102 const char *filename;
3103 int lineno;
3104 int conflict_marker_size;
3105 struct diff_options *o;
3106 unsigned ws_rule;
3107 unsigned status;
3108 };
3109
3110 static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
3111 {
3112 char firstchar;
3113 int cnt;
3114
3115 if (len < marker_size + 1)
3116 return 0;
3117 firstchar = line[0];
3118 switch (firstchar) {
3119 case '=': case '>': case '<': case '|':
3120 break;
3121 default:
3122 return 0;
3123 }
3124 for (cnt = 1; cnt < marker_size; cnt++)
3125 if (line[cnt] != firstchar)
3126 return 0;
3127 /* line[1] thru line[marker_size-1] are same as firstchar */
3128 if (len < marker_size + 1 || !isspace(line[marker_size]))
3129 return 0;
3130 return 1;
3131 }
3132
3133 static void checkdiff_consume(void *priv, char *line, unsigned long len)
3134 {
3135 struct checkdiff_t *data = priv;
3136 int marker_size = data->conflict_marker_size;
3137 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
3138 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
3139 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
3140 char *err;
3141 const char *line_prefix;
3142
3143 assert(data->o);
3144 line_prefix = diff_line_prefix(data->o);
3145
3146 if (line[0] == '+') {
3147 unsigned bad;
3148 data->lineno++;
3149 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
3150 data->status |= 1;
3151 fprintf(data->o->file,
3152 "%s%s:%d: leftover conflict marker\n",
3153 line_prefix, data->filename, data->lineno);
3154 }
3155 bad = ws_check(line + 1, len - 1, data->ws_rule);
3156 if (!bad)
3157 return;
3158 data->status |= bad;
3159 err = whitespace_error_string(bad);
3160 fprintf(data->o->file, "%s%s:%d: %s.\n",
3161 line_prefix, data->filename, data->lineno, err);
3162 free(err);
3163 emit_line(data->o, set, reset, line, 1);
3164 ws_check_emit(line + 1, len - 1, data->ws_rule,
3165 data->o->file, set, reset, ws);
3166 } else if (line[0] == ' ') {
3167 data->lineno++;
3168 } else if (line[0] == '@') {
3169 char *plus = strchr(line, '+');
3170 if (plus)
3171 data->lineno = strtol(plus, NULL, 10) - 1;
3172 else
3173 die("invalid diff");
3174 }
3175 }
3176
3177 static unsigned char *deflate_it(char *data,
3178 unsigned long size,
3179 unsigned long *result_size)
3180 {
3181 int bound;
3182 unsigned char *deflated;
3183 git_zstream stream;
3184
3185 git_deflate_init(&stream, zlib_compression_level);
3186 bound = git_deflate_bound(&stream, size);
3187 deflated = xmalloc(bound);
3188 stream.next_out = deflated;
3189 stream.avail_out = bound;
3190
3191 stream.next_in = (unsigned char *)data;
3192 stream.avail_in = size;
3193 while (git_deflate(&stream, Z_FINISH) == Z_OK)
3194 ; /* nothing */
3195 git_deflate_end(&stream);
3196 *result_size = stream.total_out;
3197 return deflated;
3198 }
3199
3200 static void emit_binary_diff_body(struct diff_options *o,
3201 mmfile_t *one, mmfile_t *two)
3202 {
3203 void *cp;
3204 void *delta;
3205 void *deflated;
3206 void *data;
3207 unsigned long orig_size;
3208 unsigned long delta_size;
3209 unsigned long deflate_size;
3210 unsigned long data_size;
3211
3212 /* We could do deflated delta, or we could do just deflated two,
3213 * whichever is smaller.
3214 */
3215 delta = NULL;
3216 deflated = deflate_it(two->ptr, two->size, &deflate_size);
3217 if (one->size && two->size) {
3218 delta = diff_delta(one->ptr, one->size,
3219 two->ptr, two->size,
3220 &delta_size, deflate_size);
3221 if (delta) {
3222 void *to_free = delta;
3223 orig_size = delta_size;
3224 delta = deflate_it(delta, delta_size, &delta_size);
3225 free(to_free);
3226 }
3227 }
3228
3229 if (delta && delta_size < deflate_size) {
3230 char *s = xstrfmt("%lu", orig_size);
3231 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_HEADER_DELTA,
3232 s, strlen(s), 0);
3233 free(s);
3234 free(deflated);
3235 data = delta;
3236 data_size = delta_size;
3237 } else {
3238 char *s = xstrfmt("%lu", two->size);
3239 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_HEADER_LITERAL,
3240 s, strlen(s), 0);
3241 free(s);
3242 free(delta);
3243 data = deflated;
3244 data_size = deflate_size;
3245 }
3246
3247 /* emit data encoded in base85 */
3248 cp = data;
3249 while (data_size) {
3250 int len;
3251 int bytes = (52 < data_size) ? 52 : data_size;
3252 char line[71];
3253 data_size -= bytes;
3254 if (bytes <= 26)
3255 line[0] = bytes + 'A' - 1;
3256 else
3257 line[0] = bytes - 26 + 'a' - 1;
3258 encode_85(line + 1, cp, bytes);
3259 cp = (char *) cp + bytes;
3260
3261 len = strlen(line);
3262 line[len++] = '\n';
3263 line[len] = '\0';
3264
3265 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_BODY,
3266 line, len, 0);
3267 }
3268 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_FOOTER, NULL, 0, 0);
3269 free(data);
3270 }
3271
3272 static void emit_binary_diff(struct diff_options *o,
3273 mmfile_t *one, mmfile_t *two)
3274 {
3275 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_HEADER, NULL, 0, 0);
3276 emit_binary_diff_body(o, one, two);
3277 emit_binary_diff_body(o, two, one);
3278 }
3279
3280 int diff_filespec_is_binary(struct repository *r,
3281 struct diff_filespec *one)
3282 {
3283 if (one->is_binary == -1) {
3284 diff_filespec_load_driver(one, r->index);
3285 if (one->driver->binary != -1)
3286 one->is_binary = one->driver->binary;
3287 else {
3288 if (!one->data && DIFF_FILE_VALID(one))
3289 diff_populate_filespec(r, one, CHECK_BINARY);
3290 if (one->is_binary == -1 && one->data)
3291 one->is_binary = buffer_is_binary(one->data,
3292 one->size);
3293 if (one->is_binary == -1)
3294 one->is_binary = 0;
3295 }
3296 }
3297 return one->is_binary;
3298 }
3299
3300 static const struct userdiff_funcname *
3301 diff_funcname_pattern(struct diff_options *o, struct diff_filespec *one)
3302 {
3303 diff_filespec_load_driver(one, o->repo->index);
3304 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
3305 }
3306
3307 void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
3308 {
3309 if (!options->a_prefix)
3310 options->a_prefix = a;
3311 if (!options->b_prefix)
3312 options->b_prefix = b;
3313 }
3314
3315 struct userdiff_driver *get_textconv(struct index_state *istate,
3316 struct diff_filespec *one)
3317 {
3318 if (!DIFF_FILE_VALID(one))
3319 return NULL;
3320
3321 diff_filespec_load_driver(one, istate);
3322 return userdiff_get_textconv(one->driver);
3323 }
3324
3325 static void builtin_diff(const char *name_a,
3326 const char *name_b,
3327 struct diff_filespec *one,
3328 struct diff_filespec *two,
3329 const char *xfrm_msg,
3330 int must_show_header,
3331 struct diff_options *o,
3332 int complete_rewrite)
3333 {
3334 mmfile_t mf1, mf2;
3335 const char *lbl[2];
3336 char *a_one, *b_two;
3337 const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
3338 const char *reset = diff_get_color_opt(o, DIFF_RESET);
3339 const char *a_prefix, *b_prefix;
3340 struct userdiff_driver *textconv_one = NULL;
3341 struct userdiff_driver *textconv_two = NULL;
3342 struct strbuf header = STRBUF_INIT;
3343 const char *line_prefix = diff_line_prefix(o);
3344
3345 diff_set_mnemonic_prefix(o, "a/", "b/");
3346 if (o->flags.reverse_diff) {
3347 a_prefix = o->b_prefix;
3348 b_prefix = o->a_prefix;
3349 } else {
3350 a_prefix = o->a_prefix;
3351 b_prefix = o->b_prefix;
3352 }
3353
3354 if (o->submodule_format == DIFF_SUBMODULE_LOG &&
3355 (!one->mode || S_ISGITLINK(one->mode)) &&
3356 (!two->mode || S_ISGITLINK(two->mode))) {
3357 show_submodule_summary(o, one->path ? one->path : two->path,
3358 &one->oid, &two->oid,
3359 two->dirty_submodule);
3360 return;
3361 } else if (o->submodule_format == DIFF_SUBMODULE_INLINE_DIFF &&
3362 (!one->mode || S_ISGITLINK(one->mode)) &&
3363 (!two->mode || S_ISGITLINK(two->mode))) {
3364 show_submodule_inline_diff(o, one->path ? one->path : two->path,
3365 &one->oid, &two->oid,
3366 two->dirty_submodule);
3367 return;
3368 }
3369
3370 if (o->flags.allow_textconv) {
3371 textconv_one = get_textconv(o->repo->index, one);
3372 textconv_two = get_textconv(o->repo->index, two);
3373 }
3374
3375 /* Never use a non-valid filename anywhere if at all possible */
3376 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
3377 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
3378
3379 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
3380 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
3381 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
3382 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
3383 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
3384 if (lbl[0][0] == '/') {
3385 /* /dev/null */
3386 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
3387 if (xfrm_msg)
3388 strbuf_addstr(&header, xfrm_msg);
3389 must_show_header = 1;
3390 }
3391 else if (lbl[1][0] == '/') {
3392 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
3393 if (xfrm_msg)
3394 strbuf_addstr(&header, xfrm_msg);
3395 must_show_header = 1;
3396 }
3397 else {
3398 if (one->mode != two->mode) {
3399 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
3400 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
3401 must_show_header = 1;
3402 }
3403 if (xfrm_msg)
3404 strbuf_addstr(&header, xfrm_msg);
3405
3406 /*
3407 * we do not run diff between different kind
3408 * of objects.
3409 */
3410 if ((one->mode ^ two->mode) & S_IFMT)
3411 goto free_ab_and_return;
3412 if (complete_rewrite &&
3413 (textconv_one || !diff_filespec_is_binary(o->repo, one)) &&
3414 (textconv_two || !diff_filespec_is_binary(o->repo, two))) {
3415 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3416 header.buf, header.len, 0);
3417 strbuf_reset(&header);
3418 emit_rewrite_diff(name_a, name_b, one, two,
3419 textconv_one, textconv_two, o);
3420 o->found_changes = 1;
3421 goto free_ab_and_return;
3422 }
3423 }
3424
3425 if (o->irreversible_delete && lbl[1][0] == '/') {
3426 emit_diff_symbol(o, DIFF_SYMBOL_HEADER, header.buf,
3427 header.len, 0);
3428 strbuf_reset(&header);
3429 goto free_ab_and_return;
3430 } else if (!o->flags.text &&
3431 ( (!textconv_one && diff_filespec_is_binary(o->repo, one)) ||
3432 (!textconv_two && diff_filespec_is_binary(o->repo, two)) )) {
3433 struct strbuf sb = STRBUF_INIT;
3434 if (!one->data && !two->data &&
3435 S_ISREG(one->mode) && S_ISREG(two->mode) &&
3436 !o->flags.binary) {
3437 if (oideq(&one->oid, &two->oid)) {
3438 if (must_show_header)
3439 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3440 header.buf, header.len,
3441 0);
3442 goto free_ab_and_return;
3443 }
3444 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3445 header.buf, header.len, 0);
3446 strbuf_addf(&sb, "%sBinary files %s and %s differ\n",
3447 diff_line_prefix(o), lbl[0], lbl[1]);
3448 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_FILES,
3449 sb.buf, sb.len, 0);
3450 strbuf_release(&sb);
3451 goto free_ab_and_return;
3452 }
3453 if (fill_mmfile(o->repo, &mf1, one) < 0 ||
3454 fill_mmfile(o->repo, &mf2, two) < 0)
3455 die("unable to read files to diff");
3456 /* Quite common confusing case */
3457 if (mf1.size == mf2.size &&
3458 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
3459 if (must_show_header)
3460 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3461 header.buf, header.len, 0);
3462 goto free_ab_and_return;
3463 }
3464 emit_diff_symbol(o, DIFF_SYMBOL_HEADER, header.buf, header.len, 0);
3465 strbuf_reset(&header);
3466 if (o->flags.binary)
3467 emit_binary_diff(o, &mf1, &mf2);
3468 else {
3469 strbuf_addf(&sb, "%sBinary files %s and %s differ\n",
3470 diff_line_prefix(o), lbl[0], lbl[1]);
3471 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_FILES,
3472 sb.buf, sb.len, 0);
3473 strbuf_release(&sb);
3474 }
3475 o->found_changes = 1;
3476 } else {
3477 /* Crazy xdl interfaces.. */
3478 const char *diffopts = getenv("GIT_DIFF_OPTS");
3479 const char *v;
3480 xpparam_t xpp;
3481 xdemitconf_t xecfg;
3482 struct emit_callback ecbdata;
3483 const struct userdiff_funcname *pe;
3484
3485 if (must_show_header) {
3486 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3487 header.buf, header.len, 0);
3488 strbuf_reset(&header);
3489 }
3490
3491 mf1.size = fill_textconv(o->repo, textconv_one, one, &mf1.ptr);
3492 mf2.size = fill_textconv(o->repo, textconv_two, two, &mf2.ptr);
3493
3494 pe = diff_funcname_pattern(o, one);
3495 if (!pe)
3496 pe = diff_funcname_pattern(o, two);
3497
3498 memset(&xpp, 0, sizeof(xpp));
3499 memset(&xecfg, 0, sizeof(xecfg));
3500 memset(&ecbdata, 0, sizeof(ecbdata));
3501 if (o->flags.suppress_diff_headers)
3502 lbl[0] = NULL;
3503 ecbdata.label_path = lbl;
3504 ecbdata.color_diff = want_color(o->use_color);
3505 ecbdata.ws_rule = whitespace_rule(o->repo->index, name_b);
3506 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
3507 check_blank_at_eof(&mf1, &mf2, &ecbdata);
3508 ecbdata.opt = o;
3509 if (header.len && !o->flags.suppress_diff_headers)
3510 ecbdata.header = &header;
3511 xpp.flags = o->xdl_opts;
3512 xpp.anchors = o->anchors;
3513 xpp.anchors_nr = o->anchors_nr;
3514 xecfg.ctxlen = o->context;
3515 xecfg.interhunkctxlen = o->interhunkcontext;
3516 xecfg.flags = XDL_EMIT_FUNCNAMES;
3517 if (o->flags.funccontext)
3518 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
3519 if (pe)
3520 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
3521 if (!diffopts)
3522 ;
3523 else if (skip_prefix(diffopts, "--unified=", &v))
3524 xecfg.ctxlen = strtoul(v, NULL, 10);
3525 else if (skip_prefix(diffopts, "-u", &v))
3526 xecfg.ctxlen = strtoul(v, NULL, 10);
3527 if (o->word_diff)
3528 init_diff_words_data(&ecbdata, o, one, two);
3529 if (xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
3530 &xpp, &xecfg))
3531 die("unable to generate diff for %s", one->path);
3532 if (o->word_diff)
3533 free_diff_words_data(&ecbdata);
3534 if (textconv_one)
3535 free(mf1.ptr);
3536 if (textconv_two)
3537 free(mf2.ptr);
3538 xdiff_clear_find_func(&xecfg);
3539 }
3540
3541 free_ab_and_return:
3542 strbuf_release(&header);
3543 diff_free_filespec_data(one);
3544 diff_free_filespec_data(two);
3545 free(a_one);
3546 free(b_two);
3547 return;
3548 }
3549
3550 static char *get_compact_summary(const struct diff_filepair *p, int is_renamed)
3551 {
3552 if (!is_renamed) {
3553 if (p->status == DIFF_STATUS_ADDED) {
3554 if (S_ISLNK(p->two->mode))
3555 return "new +l";
3556 else if ((p->two->mode & 0777) == 0755)
3557 return "new +x";
3558 else
3559 return "new";
3560 } else if (p->status == DIFF_STATUS_DELETED)
3561 return "gone";
3562 }
3563 if (S_ISLNK(p->one->mode) && !S_ISLNK(p->two->mode))
3564 return "mode -l";
3565 else if (!S_ISLNK(p->one->mode) && S_ISLNK(p->two->mode))
3566 return "mode +l";
3567 else if ((p->one->mode & 0777) == 0644 &&
3568 (p->two->mode & 0777) == 0755)
3569 return "mode +x";
3570 else if ((p->one->mode & 0777) == 0755 &&
3571 (p->two->mode & 0777) == 0644)
3572 return "mode -x";
3573 return NULL;
3574 }
3575
3576 static void builtin_diffstat(const char *name_a, const char *name_b,
3577 struct diff_filespec *one,
3578 struct diff_filespec *two,
3579 struct diffstat_t *diffstat,
3580 struct diff_options *o,
3581 struct diff_filepair *p)
3582 {
3583 mmfile_t mf1, mf2;
3584 struct diffstat_file *data;
3585 int same_contents;
3586 int complete_rewrite = 0;
3587
3588 if (!DIFF_PAIR_UNMERGED(p)) {
3589 if (p->status == DIFF_STATUS_MODIFIED && p->score)
3590 complete_rewrite = 1;
3591 }
3592
3593 data = diffstat_add(diffstat, name_a, name_b);
3594 data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
3595 if (o->flags.stat_with_summary)
3596 data->comments = get_compact_summary(p, data->is_renamed);
3597
3598 if (!one || !two) {
3599 data->is_unmerged = 1;
3600 return;
3601 }
3602
3603 same_contents = oideq(&one->oid, &two->oid);
3604
3605 if (diff_filespec_is_binary(o->repo, one) ||
3606 diff_filespec_is_binary(o->repo, two)) {
3607 data->is_binary = 1;
3608 if (same_contents) {
3609 data->added = 0;
3610 data->deleted = 0;
3611 } else {
3612 data->added = diff_filespec_size(o->repo, two);
3613 data->deleted = diff_filespec_size(o->repo, one);
3614 }
3615 }
3616
3617 else if (complete_rewrite) {
3618 diff_populate_filespec(o->repo, one, 0);
3619 diff_populate_filespec(o->repo, two, 0);
3620 data->deleted = count_lines(one->data, one->size);
3621 data->added = count_lines(two->data, two->size);
3622 }
3623
3624 else if (!same_contents) {
3625 /* Crazy xdl interfaces.. */
3626 xpparam_t xpp;
3627 xdemitconf_t xecfg;
3628
3629 if (fill_mmfile(o->repo, &mf1, one) < 0 ||
3630 fill_mmfile(o->repo, &mf2, two) < 0)
3631 die("unable to read files to diff");
3632
3633 memset(&xpp, 0, sizeof(xpp));
3634 memset(&xecfg, 0, sizeof(xecfg));
3635 xpp.flags = o->xdl_opts;
3636 xpp.anchors = o->anchors;
3637 xpp.anchors_nr = o->anchors_nr;
3638 xecfg.ctxlen = o->context;
3639 xecfg.interhunkctxlen = o->interhunkcontext;
3640 if (xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
3641 &xpp, &xecfg))
3642 die("unable to generate diffstat for %s", one->path);
3643 }
3644
3645 diff_free_filespec_data(one);
3646 diff_free_filespec_data(two);
3647 }
3648
3649 static void builtin_checkdiff(const char *name_a, const char *name_b,
3650 const char *attr_path,
3651 struct diff_filespec *one,
3652 struct diff_filespec *two,
3653 struct diff_options *o)
3654 {
3655 mmfile_t mf1, mf2;
3656 struct checkdiff_t data;
3657
3658 if (!two)
3659 return;
3660
3661 memset(&data, 0, sizeof(data));
3662 data.filename = name_b ? name_b : name_a;
3663 data.lineno = 0;
3664 data.o = o;
3665 data.ws_rule = whitespace_rule(o->repo->index, attr_path);
3666 data.conflict_marker_size = ll_merge_marker_size(o->repo->index, attr_path);
3667
3668 if (fill_mmfile(o->repo, &mf1, one) < 0 ||
3669 fill_mmfile(o->repo, &mf2, two) < 0)
3670 die("unable to read files to diff");
3671
3672 /*
3673 * All the other codepaths check both sides, but not checking
3674 * the "old" side here is deliberate. We are checking the newly
3675 * introduced changes, and as long as the "new" side is text, we
3676 * can and should check what it introduces.
3677 */
3678 if (diff_filespec_is_binary(o->repo, two))
3679 goto free_and_return;
3680 else {
3681 /* Crazy xdl interfaces.. */
3682 xpparam_t xpp;
3683 xdemitconf_t xecfg;
3684
3685 memset(&xpp, 0, sizeof(xpp));
3686 memset(&xecfg, 0, sizeof(xecfg));
3687 xecfg.ctxlen = 1; /* at least one context line */
3688 xpp.flags = 0;
3689 if (xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
3690 &xpp, &xecfg))
3691 die("unable to generate checkdiff for %s", one->path);
3692
3693 if (data.ws_rule & WS_BLANK_AT_EOF) {
3694 struct emit_callback ecbdata;
3695 int blank_at_eof;
3696
3697 ecbdata.ws_rule = data.ws_rule;
3698 check_blank_at_eof(&mf1, &mf2, &ecbdata);
3699 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
3700
3701 if (blank_at_eof) {
3702 static char *err;
3703 if (!err)
3704 err = whitespace_error_string(WS_BLANK_AT_EOF);
3705 fprintf(o->file, "%s:%d: %s.\n",
3706 data.filename, blank_at_eof, err);
3707 data.status = 1; /* report errors */