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