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#ifdef NO_FAST_WORKING_DIRECTORY
29#define FAST_WORKING_DIRECTORY 0
30#else
31#define FAST_WORKING_DIRECTORY 1
32#endif
33static int diff_detect_rename_default;
35static int diff_indent_heuristic = 1;
36static int diff_rename_limit_default = 400;
37static int diff_suppress_blank_empty;
38static int diff_use_color_default = -1;
39static int diff_color_moved_default;
40static int diff_color_moved_ws_default;
41static int diff_context_default = 3;
42static int diff_interhunk_context_default;
43static const char *diff_word_regex_cfg;
44static const char *external_diff_cmd_cfg;
45static const char *diff_order_file_cfg;
46int diff_auto_refresh_index = 1;
47static int diff_mnemonic_prefix;
48static int diff_no_prefix;
49static int diff_stat_graph_width;
50static int diff_dirstat_permille_default = 30;
51static struct diff_options default_diff_options;
52static long diff_algorithm;
53static unsigned ws_error_highlight_default = WSEH_NEW;
54static 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};
80static 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};
105static NORETURN void die_want_option(const char *option_name)
107{
108 die(_("option '%s' requires a value"), option_name);
109}
110define_list_config_array_extra(color_diff_slots, {"plain"});
112static 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}
119static 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 if (*params_copy)
129 string_list_split_in_place(¶ms, 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 }
168 string_list_clear(¶ms, 0);
169 free(params_copy);
170 return ret;
171}
172static 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}
185int 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}
194long 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}
209static 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}
219static int parse_ws_error_highlight(const char *arg)
221{
222 const char *orig_arg = arg;
223 unsigned val = 0;
224 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/*
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 */
253void init_diff_ui_defaults(void)
254{
255 diff_detect_rename_default = DIFF_DETECT_RENAME;
256}
257int 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}
264static 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 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}
293static unsigned 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 string_list_split(&l, arg, ',', -1);
301 for_each_string_list_item(i, &l) {
303 struct strbuf sb = STRBUF_INIT;
304 strbuf_addstr(&sb, i->string);
305 strbuf_trim(&sb);
306 if (!strcmp(sb.buf, "no"))
308 ret = 0;
309 else if (!strcmp(sb.buf, "ignore-space-change"))
310 ret |= XDF_IGNORE_WHITESPACE_CHANGE;
311 else if (!strcmp(sb.buf, "ignore-space-at-eol"))
312 ret |= XDF_IGNORE_WHITESPACE_AT_EOL;
313 else if (!strcmp(sb.buf, "ignore-all-space"))
314 ret |= XDF_IGNORE_WHITESPACE;
315 else if (!strcmp(sb.buf, "allow-indentation-change"))
316 ret |= COLOR_MOVED_WS_ALLOW_INDENTATION_CHANGE;
317 else {
318 ret |= COLOR_MOVED_WS_ERROR;
319 error(_("unknown color-moved-ws mode '%s', possible values are 'ignore-space-change', 'ignore-space-at-eol', 'ignore-all-space', 'allow-indentation-change'"), sb.buf);
320 }
321 strbuf_release(&sb);
323 }
324 if ((ret & COLOR_MOVED_WS_ALLOW_INDENTATION_CHANGE) &&
326 (ret & XDF_WHITESPACE_FLAGS)) {
327 error(_("color-moved-ws: allow-indentation-change cannot be combined with other whitespace modes"));
328 ret |= COLOR_MOVED_WS_ERROR;
329 }
330 string_list_clear(&l, 0);
332 return ret;
334}
335int git_diff_ui_config(const char *var, const char *value, void *cb)
337{
338 if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
339 diff_use_color_default = git_config_colorbool(var, value);
340 return 0;
341 }
342 if (!strcmp(var, "diff.colormoved")) {
343 int cm = parse_color_moved(value);
344 if (cm < 0)
345 return -1;
346 diff_color_moved_default = cm;
347 return 0;
348 }
349 if (!strcmp(var, "diff.colormovedws")) {
350 unsigned cm = parse_color_moved_ws(value);
351 if (cm & COLOR_MOVED_WS_ERROR)
352 return -1;
353 diff_color_moved_ws_default = cm;
354 return 0;
355 }
356 if (!strcmp(var, "diff.context")) {
357 diff_context_default = git_config_int(var, value);
358 if (diff_context_default < 0)
359 return -1;
360 return 0;
361 }
362 if (!strcmp(var, "diff.interhunkcontext")) {
363 diff_interhunk_context_default = git_config_int(var, value);
364 if (diff_interhunk_context_default < 0)
365 return -1;
366 return 0;
367 }
368 if (!strcmp(var, "diff.renames")) {
369 diff_detect_rename_default = git_config_rename(var, value);
370 return 0;
371 }
372 if (!strcmp(var, "diff.autorefreshindex")) {
373 diff_auto_refresh_index = git_config_bool(var, value);
374 return 0;
375 }
376 if (!strcmp(var, "diff.mnemonicprefix")) {
377 diff_mnemonic_prefix = git_config_bool(var, value);
378 return 0;
379 }
380 if (!strcmp(var, "diff.noprefix")) {
381 diff_no_prefix = git_config_bool(var, value);
382 return 0;
383 }
384 if (!strcmp(var, "diff.statgraphwidth")) {
385 diff_stat_graph_width = git_config_int(var, value);
386 return 0;
387 }
388 if (!strcmp(var, "diff.external"))
389 return git_config_string(&external_diff_cmd_cfg, var, value);
390 if (!strcmp(var, "diff.wordregex"))
391 return git_config_string(&diff_word_regex_cfg, var, value);
392 if (!strcmp(var, "diff.orderfile"))
393 return git_config_pathname(&diff_order_file_cfg, var, value);
394 if (!strcmp(var, "diff.ignoresubmodules"))
396 handle_ignore_submodules_arg(&default_diff_options, value);
397 if (!strcmp(var, "diff.submodule")) {
399 if (parse_submodule_params(&default_diff_options, value))
400 warning(_("Unknown value for 'diff.submodule' config variable: '%s'"),
401 value);
402 return 0;
403 }
404 if (!strcmp(var, "diff.algorithm")) {
406 diff_algorithm = parse_algorithm_value(value);
407 if (diff_algorithm < 0)
408 return -1;
409 return 0;
410 }
411 if (!strcmp(var, "diff.wserrorhighlight")) {
413 int val = parse_ws_error_highlight(value);
414 if (val < 0)
415 return -1;
416 ws_error_highlight_default = val;
417 return 0;
418 }
419 if (git_color_config(var, value, cb) < 0)
421 return -1;
422 return git_diff_basic_config(var, value, cb);
424}
425int git_diff_basic_config(const char *var, const char *value, void *cb)
427{
428 const char *name;
429 if (!strcmp(var, "diff.renamelimit")) {
431 diff_rename_limit_default = git_config_int(var, value);
432 return 0;
433 }
434 if (userdiff_config(var, value) < 0)
436 return -1;
437 if (skip_prefix(var, "diff.color.", &name) ||
439 skip_prefix(var, "color.diff.", &name)) {
440 int slot = parse_diff_color_slot(name);
441 if (slot < 0)
442 return 0;
443 if (!value)
444 return config_error_nonbool(var);
445 return color_parse(value, diff_colors[slot]);
446 }
447 /* like GNU diff's --suppress-blank-empty option */
449 if (!strcmp(var, "diff.suppressblankempty") ||
450 /* for backwards compatibility */
451 !strcmp(var, "diff.suppress-blank-empty")) {
452 diff_suppress_blank_empty = git_config_bool(var, value);
453 return 0;
454 }
455 if (!strcmp(var, "diff.dirstat")) {
457 struct strbuf errmsg = STRBUF_INIT;
458 default_diff_options.dirstat_permille = diff_dirstat_permille_default;
459 if (parse_dirstat_params(&default_diff_options, value, &errmsg))
460 warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
461 errmsg.buf);
462 strbuf_release(&errmsg);
463 diff_dirstat_permille_default = default_diff_options.dirstat_permille;
464 return 0;
465 }
466 if (git_diff_heuristic_config(var, value, cb) < 0)
468 return -1;
469 return git_default_config(var, value, cb);
471}
472static char *quote_two(const char *one, const char *two)
474{
475 int need_one = quote_c_style(one, NULL, NULL, 1);
476 int need_two = quote_c_style(two, NULL, NULL, 1);
477 struct strbuf res = STRBUF_INIT;
478 if (need_one + need_two) {
480 strbuf_addch(&res, '"');
481 quote_c_style(one, &res, NULL, 1);
482 quote_c_style(two, &res, NULL, 1);
483 strbuf_addch(&res, '"');
484 } else {
485 strbuf_addstr(&res, one);
486 strbuf_addstr(&res, two);
487 }
488 return strbuf_detach(&res, NULL);
489}
490static const char *external_diff(void)
492{
493 static const char *external_diff_cmd = NULL;
494 static int done_preparing = 0;
495 if (done_preparing)
497 return external_diff_cmd;
498 external_diff_cmd = xstrdup_or_null(getenv("GIT_EXTERNAL_DIFF"));
499 if (!external_diff_cmd)
500 external_diff_cmd = external_diff_cmd_cfg;
501 done_preparing = 1;
502 return external_diff_cmd;
503}
504/*
506 * Keep track of files used for diffing. Sometimes such an entry
507 * refers to a temporary file, sometimes to an existing file, and
508 * sometimes to "/dev/null".
509 */
510static struct diff_tempfile {
511 /*
512 * filename external diff should read from, or NULL if this
513 * entry is currently not in use:
514 */
515 const char *name;
516 char hex[GIT_MAX_HEXSZ + 1];
518 char mode[10];
519 /*
521 * If this diff_tempfile instance refers to a temporary file,
522 * this tempfile object is used to manage its lifetime.
523 */
524 struct tempfile *tempfile;
525} diff_temp[2];
526struct emit_callback {
528 int color_diff;
529 unsigned ws_rule;
530 int blank_at_eof_in_preimage;
531 int blank_at_eof_in_postimage;
532 int lno_in_preimage;
533 int lno_in_postimage;
534 const char **label_path;
535 struct diff_words_data *diff_words;
536 struct diff_options *opt;
537 struct strbuf *header;
538};
539static int count_lines(const char *data, int size)
541{
542 int count, ch, completely_empty = 1, nl_just_seen = 0;
543 count = 0;
544 while (0 < size--) {
545 ch = *data++;
546 if (ch == '\n') {
547 count++;
548 nl_just_seen = 1;
549 completely_empty = 0;
550 }
551 else {
552 nl_just_seen = 0;
553 completely_empty = 0;
554 }
555 }
556 if (completely_empty)
557 return 0;
558 if (!nl_just_seen)
559 count++; /* no trailing newline */
560 return count;
561}
562static int fill_mmfile(struct repository *r, mmfile_t *mf,
564 struct diff_filespec *one)
565{
566 if (!DIFF_FILE_VALID(one)) {
567 mf->ptr = (char *)""; /* does not matter */
568 mf->size = 0;
569 return 0;
570 }
571 else if (diff_populate_filespec(r, one, 0))
572 return -1;
573 mf->ptr = one->data;
575 mf->size = one->size;
576 return 0;
577}
578/* like fill_mmfile, but only for size, so we can avoid retrieving blob */
580static unsigned long diff_filespec_size(struct repository *r,
581 struct diff_filespec *one)
582{
583 if (!DIFF_FILE_VALID(one))
584 return 0;
585 diff_populate_filespec(r, one, CHECK_SIZE_ONLY);
586 return one->size;
587}
588static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
590{
591 char *ptr = mf->ptr;
592 long size = mf->size;
593 int cnt = 0;
594 if (!size)
596 return cnt;
597 ptr += size - 1; /* pointing at the very end */
598 if (*ptr != '\n')
599 ; /* incomplete line */
600 else
601 ptr--; /* skip the last LF */
602 while (mf->ptr < ptr) {
603 char *prev_eol;
604 for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
605 if (*prev_eol == '\n')
606 break;
607 if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
608 break;
609 cnt++;
610 ptr = prev_eol - 1;
611 }
612 return cnt;
613}
614static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
616 struct emit_callback *ecbdata)
617{
618 int l1, l2, at;
619 unsigned ws_rule = ecbdata->ws_rule;
620 l1 = count_trailing_blank(mf1, ws_rule);
621 l2 = count_trailing_blank(mf2, ws_rule);
622 if (l2 <= l1) {
623 ecbdata->blank_at_eof_in_preimage = 0;
624 ecbdata->blank_at_eof_in_postimage = 0;
625 return;
626 }
627 at = count_lines(mf1->ptr, mf1->size);
628 ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
629 at = count_lines(mf2->ptr, mf2->size);
631 ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
632}
633static void emit_line_0(struct diff_options *o,
635 const char *set_sign, const char *set, unsigned reverse, const char *reset,
636 int first, const char *line, int len)
637{
638 int has_trailing_newline, has_trailing_carriage_return;
639 int needs_reset = 0; /* at the end of the line */
640 FILE *file = o->file;
641 fputs(diff_line_prefix(o), file);
643 has_trailing_newline = (len > 0 && line[len-1] == '\n');
645 if (has_trailing_newline)
646 len--;
647 has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
649 if (has_trailing_carriage_return)
650 len--;
651 if (!len && !first)
653 goto end_of_line;
654 if (reverse && want_color(o->use_color)) {
656 fputs(GIT_COLOR_REVERSE, file);
657 needs_reset = 1;
658 }
659 if (set_sign) {
661 fputs(set_sign, file);
662 needs_reset = 1;
663 }
664 if (first)
666 fputc(first, file);
667 if (!len)
669 goto end_of_line;
670 if (set) {
672 if (set_sign && set != set_sign)
673 fputs(reset, file);
674 fputs(set, file);
675 needs_reset = 1;
676 }
677 fwrite(line, len, 1, file);
678 needs_reset = 1; /* 'line' may contain color codes. */
679end_of_line:
681 if (needs_reset)
682 fputs(reset, file);
683 if (has_trailing_carriage_return)
684 fputc('\r', file);
685 if (has_trailing_newline)
686 fputc('\n', file);
687}
688static void emit_line(struct diff_options *o, const char *set, const char *reset,
690 const char *line, int len)
691{
692 emit_line_0(o, set, NULL, 0, reset, 0, line, len);
693}
694enum diff_symbol {
696 DIFF_SYMBOL_BINARY_DIFF_HEADER,
697 DIFF_SYMBOL_BINARY_DIFF_HEADER_DELTA,
698 DIFF_SYMBOL_BINARY_DIFF_HEADER_LITERAL,
699 DIFF_SYMBOL_BINARY_DIFF_BODY,
700 DIFF_SYMBOL_BINARY_DIFF_FOOTER,
701 DIFF_SYMBOL_STATS_SUMMARY_NO_FILES,
702 DIFF_SYMBOL_STATS_SUMMARY_ABBREV,
703 DIFF_SYMBOL_STATS_SUMMARY_INSERTS_DELETES,
704 DIFF_SYMBOL_STATS_LINE,
705 DIFF_SYMBOL_WORD_DIFF,
706 DIFF_SYMBOL_STAT_SEP,
707 DIFF_SYMBOL_SUMMARY,
708 DIFF_SYMBOL_SUBMODULE_ADD,
709 DIFF_SYMBOL_SUBMODULE_DEL,
710 DIFF_SYMBOL_SUBMODULE_UNTRACKED,
711 DIFF_SYMBOL_SUBMODULE_MODIFIED,
712 DIFF_SYMBOL_SUBMODULE_HEADER,
713 DIFF_SYMBOL_SUBMODULE_ERROR,
714 DIFF_SYMBOL_SUBMODULE_PIPETHROUGH,
715 DIFF_SYMBOL_REWRITE_DIFF,
716 DIFF_SYMBOL_BINARY_FILES,
717 DIFF_SYMBOL_HEADER,
718 DIFF_SYMBOL_FILEPAIR_PLUS,
719 DIFF_SYMBOL_FILEPAIR_MINUS,
720 DIFF_SYMBOL_WORDS_PORCELAIN,
721 DIFF_SYMBOL_WORDS,
722 DIFF_SYMBOL_CONTEXT,
723 DIFF_SYMBOL_CONTEXT_INCOMPLETE,
724 DIFF_SYMBOL_PLUS,
725 DIFF_SYMBOL_MINUS,
726 DIFF_SYMBOL_NO_LF_EOF,
727 DIFF_SYMBOL_CONTEXT_FRAGINFO,
728 DIFF_SYMBOL_CONTEXT_MARKER,
729 DIFF_SYMBOL_SEPARATOR
730};
731/*
732 * Flags for content lines:
733 * 0..12 are whitespace rules
734 * 13-15 are WSEH_NEW | WSEH_OLD | WSEH_CONTEXT
735 * 16 is marking if the line is blank at EOF
736 */
737#define DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF (1<<16)
738#define DIFF_SYMBOL_MOVED_LINE (1<<17)
739#define DIFF_SYMBOL_MOVED_LINE_ALT (1<<18)
740#define DIFF_SYMBOL_MOVED_LINE_UNINTERESTING (1<<19)
741#define DIFF_SYMBOL_CONTENT_WS_MASK (WSEH_NEW | WSEH_OLD | WSEH_CONTEXT | WS_RULE_MASK)
742/*
744 * This struct is used when we need to buffer the output of the diff output.
745 *
746 * NEEDSWORK: Instead of storing a copy of the line, add an offset pointer
747 * into the pre/post image file. This pointer could be a union with the
748 * line pointer. By storing an offset into the file instead of the literal line,
749 * we can decrease the memory footprint for the buffered output. At first we
750 * may want to only have indirection for the content lines, but we could also
751 * enhance the state for emitting prefabricated lines, e.g. the similarity
752 * score line or hunk/file headers would only need to store a number or path
753 * and then the output can be constructed later on depending on state.
754 */
755struct emitted_diff_symbol {
756 const char *line;
757 int len;
758 int flags;
759 int indent_off; /* Offset to first non-whitespace character */
760 int indent_width; /* The visual width of the indentation */
761 enum diff_symbol s;
762};
763#define EMITTED_DIFF_SYMBOL_INIT {NULL}
764struct emitted_diff_symbols {
766 struct emitted_diff_symbol *buf;
767 int nr, alloc;
768};
769#define EMITTED_DIFF_SYMBOLS_INIT {NULL, 0, 0}
770static void append_emitted_diff_symbol(struct diff_options *o,
772 struct emitted_diff_symbol *e)
773{
774 struct emitted_diff_symbol *f;
775 ALLOC_GROW(o->emitted_symbols->buf,
777 o->emitted_symbols->nr + 1,
778 o->emitted_symbols->alloc);
779 f = &o->emitted_symbols->buf[o->emitted_symbols->nr++];
780 memcpy(f, e, sizeof(struct emitted_diff_symbol));
782 f->line = e->line ? xmemdupz(e->line, e->len) : NULL;
783}
784struct moved_entry {
786 struct hashmap_entry ent;
787 const struct emitted_diff_symbol *es;
788 struct moved_entry *next_line;
789};
790struct moved_block {
792 struct moved_entry *match;
793 int wsd; /* The whitespace delta of this block */
794};
795static void moved_block_clear(struct moved_block *b)
797{
798 memset(b, 0, sizeof(*b));
799}
800#define INDENT_BLANKLINE INT_MIN
802static void fill_es_indent_data(struct emitted_diff_symbol *es)
804{
805 unsigned int off = 0, i;
806 int width = 0, tab_width = es->flags & WS_TAB_WIDTH_MASK;
807 const char *s = es->line;
808 const int len = es->len;
809 /* skip any \v \f \r at start of indentation */
811 while (s[off] == '\f' || s[off] == '\v' ||
812 (s[off] == '\r' && off < len - 1))
813 off++;
814 /* calculate the visual width of indentation */
816 while(1) {
817 if (s[off] == ' ') {
818 width++;
819 off++;
820 } else if (s[off] == '\t') {
821 width += tab_width - (width % tab_width);
822 while (s[++off] == '\t')
823 width += tab_width;
824 } else {
825 break;
826 }
827 }
828 /* check if this line is blank */
830 for (i = off; i < len; i++)
831 if (!isspace(s[i]))
832 break;
833 if (i == len) {
835 es->indent_width = INDENT_BLANKLINE;
836 es->indent_off = len;
837 } else {
838 es->indent_off = off;
839 es->indent_width = width;
840 }
841}
842static int compute_ws_delta(const struct emitted_diff_symbol *a,
844 const struct emitted_diff_symbol *b,
845 int *out)
846{
847 int a_len = a->len,
848 b_len = b->len,
849 a_off = a->indent_off,
850 a_width = a->indent_width,
851 b_off = b->indent_off,
852 b_width = b->indent_width;
853 int delta;
854 if (a_width == INDENT_BLANKLINE && b_width == INDENT_BLANKLINE) {
856 *out = INDENT_BLANKLINE;
857 return 1;
858 }
859 if (a->s == DIFF_SYMBOL_PLUS)
861 delta = a_width - b_width;
862 else
863 delta = b_width - a_width;
864 if (a_len - a_off != b_len - b_off ||
866 memcmp(a->line + a_off, b->line + b_off, a_len - a_off))
867 return 0;
868 *out = delta;
870 return 1;
872}
873static int cmp_in_block_with_wsd(const struct diff_options *o,
875 const struct moved_entry *cur,
876 const struct moved_entry *match,
877 struct moved_block *pmb,
878 int n)
879{
880 struct emitted_diff_symbol *l = &o->emitted_symbols->buf[n];
881 int al = cur->es->len, bl = match->es->len, cl = l->len;
882 const char *a = cur->es->line,
883 *b = match->es->line,
884 *c = l->line;
885 int a_off = cur->es->indent_off,
886 a_width = cur->es->indent_width,
887 c_off = l->indent_off,
888 c_width = l->indent_width;
889 int delta;
890 /*
892 * We need to check if 'cur' is equal to 'match'. As those
893 * are from the same (+/-) side, we do not need to adjust for
894 * indent changes. However these were found using fuzzy
895 * matching so we do have to check if they are equal. Here we
896 * just check the lengths. We delay calling memcmp() to check
897 * the contents until later as if the length comparison for a
898 * and c fails we can avoid the call all together.
899 */
900 if (al != bl)
901 return 1;
902 /* If 'l' and 'cur' are both blank then they match. */
904 if (a_width == INDENT_BLANKLINE && c_width == INDENT_BLANKLINE)
905 return 0;
906 /*
908 * The indent changes of the block are known and stored in pmb->wsd;
909 * however we need to check if the indent changes of the current line
910 * match those of the current block and that the text of 'l' and 'cur'
911 * after the indentation match.
912 */
913 if (cur->es->s == DIFF_SYMBOL_PLUS)
914 delta = a_width - c_width;
915 else
916 delta = c_width - a_width;
917 /*
919 * If the previous lines of this block were all blank then set its
920 * whitespace delta.
921 */
922 if (pmb->wsd == INDENT_BLANKLINE)
923 pmb->wsd = delta;
924 return !(delta == pmb->wsd && al - a_off == cl - c_off &&
926 !memcmp(a, b, al) && !
927 memcmp(a + a_off, c + c_off, al - a_off));
928}
929static int moved_entry_cmp(const void *hashmap_cmp_fn_data,
931 const void *entry,
932 const void *entry_or_key,
933 const void *keydata)
934{
935 const struct diff_options *diffopt = hashmap_cmp_fn_data;
936 const struct moved_entry *a = entry;
937 const struct moved_entry *b = entry_or_key;
938 unsigned flags = diffopt->color_moved_ws_handling
939 & XDF_WHITESPACE_FLAGS;
940 if (diffopt->color_moved_ws_handling &
942 COLOR_MOVED_WS_ALLOW_INDENTATION_CHANGE)
943 /*
944 * As there is not specific white space config given,
945 * we'd need to check for a new block, so ignore all
946 * white space. The setup of the white space
947 * configuration for the next block is done else where
948 */
949 flags |= XDF_IGNORE_WHITESPACE;
950 return !xdiff_compare_lines(a->es->line, a->es->len,
952 b->es->line, b->es->len,
953 flags);
954}
955static struct moved_entry *prepare_entry(struct diff_options *o,
957 int line_no)
958{
959 struct moved_entry *ret = xmalloc(sizeof(*ret));
960 struct emitted_diff_symbol *l = &o->emitted_symbols->buf[line_no];
961 unsigned flags = o->color_moved_ws_handling & XDF_WHITESPACE_FLAGS;
962 ret->ent.hash = xdiff_hash_string(l->line, l->len, flags);
964 ret->es = l;
965 ret->next_line = NULL;
966 return ret;
968}
969static void add_lines_to_move_detection(struct diff_options *o,
971 struct hashmap *add_lines,
972 struct hashmap *del_lines)
973{
974 struct moved_entry *prev_line = NULL;
975 int n;
977 for (n = 0; n < o->emitted_symbols->nr; n++) {
978 struct hashmap *hm;
979 struct moved_entry *key;
980 switch (o->emitted_symbols->buf[n].s) {
982 case DIFF_SYMBOL_PLUS:
983 hm = add_lines;
984 break;
985 case DIFF_SYMBOL_MINUS:
986 hm = del_lines;
987 break;
988 default:
989 prev_line = NULL;
990 continue;
991 }
992 if (o->color_moved_ws_handling &
994 COLOR_MOVED_WS_ALLOW_INDENTATION_CHANGE)
995 fill_es_indent_data(&o->emitted_symbols->buf[n]);
996 key = prepare_entry(o, n);
997 if (prev_line && prev_line->es->s == o->emitted_symbols->buf[n].s)
998 prev_line->next_line = key;
999 hashmap_add(hm, key);
1001 prev_line = key;
1002 }
1003}
1004static void pmb_advance_or_null(struct diff_options *o,
1006 struct moved_entry *match,
1007 struct hashmap *hm,
1008 struct moved_block *pmb,
1009 int pmb_nr)
1010{
1011 int i;
1012 for (i = 0; i < pmb_nr; i++) {
1013 struct moved_entry *prev = pmb[i].match;
1014 struct moved_entry *cur = (prev && prev->next_line) ?
1015 prev->next_line : NULL;
1016 if (cur && !hm->cmpfn(o, cur, match, NULL)) {
1017 pmb[i].match = cur;
1018 } else {
1019 pmb[i].match = NULL;
1020 }
1021 }
1022}
1023static void pmb_advance_or_null_multi_match(struct diff_options *o,
1025 struct moved_entry *match,
1026 struct hashmap *hm,
1027 struct moved_block *pmb,
1028 int pmb_nr, int n)
1029{
1030 int i;
1031 char *got_match = xcalloc(1, pmb_nr);
1032 for (; match; match = hashmap_get_next(hm, match)) {
1034 for (i = 0; i < pmb_nr; i++) {
1035 struct moved_entry *prev = pmb[i].match;
1036 struct moved_entry *cur = (prev && prev->next_line) ?
1037 prev->next_line : NULL;
1038 if (!cur)
1039 continue;
1040 if (!cmp_in_block_with_wsd(o, cur, match, &pmb[i], n))
1041 got_match[i] |= 1;
1042 }
1043 }
1044 for (i = 0; i < pmb_nr; i++) {
1046 if (got_match[i]) {
1047 /* Advance to the next line */
1048 pmb[i].match = pmb[i].match->next_line;
1049 } else {
1050 moved_block_clear(&pmb[i]);
1051 }
1052 }
1053 free(got_match);
1055}
1056static int shrink_potential_moved_blocks(struct moved_block *pmb,
1058 int pmb_nr)
1059{
1060 int lp, rp;
1061 /* Shrink the set of potential block to the remaining running */
1063 for (lp = 0, rp = pmb_nr - 1; lp <= rp;) {
1064 while (lp < pmb_nr && pmb[lp].match)
1065 lp++;
1066 /* lp points at the first NULL now */
1067 while (rp > -1 && !pmb[rp].match)
1069 rp--;
1070 /* rp points at the last non-NULL */
1071 if (lp < pmb_nr && rp > -1 && lp < rp) {
1073 pmb[lp] = pmb[rp];
1074 memset(&pmb[rp], 0, sizeof(pmb[rp]));
1075 rp--;
1076 lp++;
1077 }
1078 }
1079 /* Remember the number of running sets */
1081 return rp + 1;
1082}
1083/*
1085 * If o->color_moved is COLOR_MOVED_PLAIN, this function does nothing.
1086 *
1087 * Otherwise, if the last block has fewer alphanumeric characters than
1088 * COLOR_MOVED_MIN_ALNUM_COUNT, unset DIFF_SYMBOL_MOVED_LINE on all lines in
1089 * that block.
1090 *
1091 * The last block consists of the (n - block_length)'th line up to but not
1092 * including the nth line.
1093 *
1094 * Returns 0 if the last block is empty or is unset by this function, non zero
1095 * otherwise.
1096 *
1097 * NEEDSWORK: This uses the same heuristic as blame_entry_score() in blame.c.
1098 * Think of a way to unify them.
1099 */
1100static int adjust_last_block(struct diff_options *o, int n, int block_length)
1101{
1102 int i, alnum_count = 0;
1103 if (o->color_moved == COLOR_MOVED_PLAIN)
1104 return block_length;
1105 for (i = 1; i < block_length + 1; i++) {
1106 const char *c = o->emitted_symbols->buf[n - i].line;
1107 for (; *c; c++) {
1108 if (!isalnum(*c))
1109 continue;
1110 alnum_count++;
1111 if (alnum_count >= COLOR_MOVED_MIN_ALNUM_COUNT)
1112 return 1;
1113 }
1114 }
1115 for (i = 1; i < block_length + 1; i++)
1116 o->emitted_symbols->buf[n - i].flags &= ~DIFF_SYMBOL_MOVED_LINE;
1117 return 0;
1118}
1119/* Find blocks of moved code, delegate actual coloring decision to helper */
1121static void mark_color_as_moved(struct diff_options *o,
1122 struct hashmap *add_lines,
1123 struct hashmap *del_lines)
1124{
1125 struct moved_block *pmb = NULL; /* potentially moved blocks */
1126 int pmb_nr = 0, pmb_alloc = 0;
1127 int n, flipped_block = 0, block_length = 0;
1128 for (n = 0; n < o->emitted_symbols->nr; n++) {
1131 struct hashmap *hm = NULL;
1132 struct moved_entry *key;
1133 struct moved_entry *match = NULL;
1134 struct emitted_diff_symbol *l = &o->emitted_symbols->buf[n];
1135 enum diff_symbol last_symbol = 0;
1136 switch (l->s) {
1138 case DIFF_SYMBOL_PLUS:
1139 hm = del_lines;
1140 key = prepare_entry(o, n);
1141 match = hashmap_get(hm, key, NULL);
1142 free(key);
1143 break;
1144 case DIFF_SYMBOL_MINUS:
1145 hm = add_lines;
1146 key = prepare_entry(o, n);
1147 match = hashmap_get(hm, key, NULL);
1148 free(key);
1149 break;
1150 default:
1151 flipped_block = 0;
1152 }
1153 if (!match) {
1155 int i;
1156 adjust_last_block(o, n, block_length);
1158 for(i = 0; i < pmb_nr; i++)
1159 moved_block_clear(&pmb[i]);
1160 pmb_nr = 0;
1161 block_length = 0;
1162 flipped_block = 0;
1163 last_symbol = l->s;
1164 continue;
1165 }
1166 if (o->color_moved == COLOR_MOVED_PLAIN) {
1168 last_symbol = l->s;
1169 l->flags |= DIFF_SYMBOL_MOVED_LINE;
1170 continue;
1171 }
1172 if (o->color_moved_ws_handling &
1174 COLOR_MOVED_WS_ALLOW_INDENTATION_CHANGE)
1175 pmb_advance_or_null_multi_match(o, match, hm, pmb, pmb_nr, n);
1176 else
1177 pmb_advance_or_null(o, match, hm, pmb, pmb_nr);
1178 pmb_nr = shrink_potential_moved_blocks(pmb, pmb_nr);
1180 if (pmb_nr == 0) {
1182 /*
1183 * The current line is the start of a new block.
1184 * Setup the set of potential blocks.
1185 */
1186 for (; match; match = hashmap_get_next(hm, match)) {
1187 ALLOC_GROW(pmb, pmb_nr + 1, pmb_alloc);
1188 if (o->color_moved_ws_handling &
1189 COLOR_MOVED_WS_ALLOW_INDENTATION_CHANGE) {
1190 if (compute_ws_delta(l, match->es,
1191 &pmb[pmb_nr].wsd))
1192 pmb[pmb_nr++].match = match;
1193 } else {
1194 pmb[pmb_nr].wsd = 0;
1195 pmb[pmb_nr++].match = match;
1196 }
1197 }
1198 if (adjust_last_block(o, n, block_length) &&
1200 pmb_nr && last_symbol != l->s)
1201 flipped_block = (flipped_block + 1) % 2;
1202 else
1203 flipped_block = 0;
1204 block_length = 0;
1206 }
1207 if (pmb_nr) {
1209 block_length++;
1210 l->flags |= DIFF_SYMBOL_MOVED_LINE;
1211 if (flipped_block && o->color_moved != COLOR_MOVED_BLOCKS)
1212 l->flags |= DIFF_SYMBOL_MOVED_LINE_ALT;
1213 }
1214 last_symbol = l->s;
1215 }
1216 adjust_last_block(o, n, block_length);
1217 for(n = 0; n < pmb_nr; n++)
1219 moved_block_clear(&pmb[n]);
1220 free(pmb);
1221}
1222#define DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK \
1224 (DIFF_SYMBOL_MOVED_LINE | DIFF_SYMBOL_MOVED_LINE_ALT)
1225static void dim_moved_lines(struct diff_options *o)
1226{
1227 int n;
1228 for (n = 0; n < o->emitted_symbols->nr; n++) {
1229 struct emitted_diff_symbol *prev = (n != 0) ?
1230 &o->emitted_symbols->buf[n - 1] : NULL;
1231 struct emitted_diff_symbol *l = &o->emitted_symbols->buf[n];
1232 struct emitted_diff_symbol *next =
1233 (n < o->emitted_symbols->nr - 1) ?
1234 &o->emitted_symbols->buf[n + 1] : NULL;
1235 /* Not a plus or minus line? */
1237 if (l->s != DIFF_SYMBOL_PLUS && l->s != DIFF_SYMBOL_MINUS)
1238 continue;
1239 /* Not a moved line? */
1241 if (!(l->flags & DIFF_SYMBOL_MOVED_LINE))
1242 continue;
1243 /*
1245 * If prev or next are not a plus or minus line,
1246 * pretend they don't exist
1247 */
1248 if (prev && prev->s != DIFF_SYMBOL_PLUS &&
1249 prev->s != DIFF_SYMBOL_MINUS)
1250 prev = NULL;
1251 if (next && next->s != DIFF_SYMBOL_PLUS &&
1252 next->s != DIFF_SYMBOL_MINUS)
1253 next = NULL;
1254 /* Inside a block? */
1256 if ((prev &&
1257 (prev->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK) ==
1258 (l->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK)) &&
1259 (next &&
1260 (next->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK) ==
1261 (l->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK))) {
1262 l->flags |= DIFF_SYMBOL_MOVED_LINE_UNINTERESTING;
1263 continue;
1264 }
1265 /* Check if we are at an interesting bound: */
1267 if (prev && (prev->flags & DIFF_SYMBOL_MOVED_LINE) &&
1268 (prev->flags & DIFF_SYMBOL_MOVED_LINE_ALT) !=
1269 (l->flags & DIFF_SYMBOL_MOVED_LINE_ALT))
1270 continue;
1271 if (next && (next->flags & DIFF_SYMBOL_MOVED_LINE) &&
1272 (next->flags & DIFF_SYMBOL_MOVED_LINE_ALT) !=
1273 (l->flags & DIFF_SYMBOL_MOVED_LINE_ALT))
1274 continue;
1275 /*
1277 * The boundary to prev and next are not interesting,
1278 * so this line is not interesting as a whole
1279 */
1280 l->flags |= DIFF_SYMBOL_MOVED_LINE_UNINTERESTING;
1281 }
1282}
1283static void emit_line_ws_markup(struct diff_options *o,
1285 const char *set_sign, const char *set,
1286 const char *reset,
1287 int sign_index, const char *line, int len,
1288 unsigned ws_rule, int blank_at_eof)
1289{
1290 const char *ws = NULL;
1291 int sign = o->output_indicators[sign_index];
1292 if (o->ws_error_highlight & ws_rule) {
1294 ws = diff_get_color_opt(o, DIFF_WHITESPACE);
1295 if (!*ws)
1296 ws = NULL;
1297 }
1298 if (!ws && !set_sign)
1300 emit_line_0(o, set, NULL, 0, reset, sign, line, len);
1301 else if (!ws) {
1302 emit_line_0(o, set_sign, set, !!set_sign, reset, sign, line, len);
1303 } else if (blank_at_eof)
1304 /* Blank line at EOF - paint '+' as well */
1305 emit_line_0(o, ws, NULL, 0, reset, sign, line, len);
1306 else {
1307 /* Emit just the prefix, then the rest. */
1308 emit_line_0(o, set_sign ? set_sign : set, NULL, !!set_sign, reset,
1309 sign, "", 0);
1310 ws_check_emit(line, len, ws_rule,
1311 o->file, set, reset, ws);
1312 }
1313}
1314static void emit_diff_symbol_from_struct(struct diff_options *o,
1316 struct emitted_diff_symbol *eds)
1317{
1318 static const char *nneof = " No newline at end of file\n";
1319 const char *context, *reset, *set, *set_sign, *meta, *fraginfo;
1320 struct strbuf sb = STRBUF_INIT;
1321 enum diff_symbol s = eds->s;
1323 const char *line = eds->line;
1324 int len = eds->len;
1325 unsigned flags = eds->flags;
1326 switch (s) {
1328 case DIFF_SYMBOL_NO_LF_EOF:
1329 context = diff_get_color_opt(o, DIFF_CONTEXT);
1330 reset = diff_get_color_opt(o, DIFF_RESET);
1331 putc('\n', o->file);
1332 emit_line_0(o, context, NULL, 0, reset, '\\',
1333 nneof, strlen(nneof));
1334 break;
1335 case DIFF_SYMBOL_SUBMODULE_HEADER:
1336 case DIFF_SYMBOL_SUBMODULE_ERROR:
1337 case DIFF_SYMBOL_SUBMODULE_PIPETHROUGH:
1338 case DIFF_SYMBOL_STATS_SUMMARY_INSERTS_DELETES:
1339 case DIFF_SYMBOL_SUMMARY:
1340 case DIFF_SYMBOL_STATS_LINE:
1341 case DIFF_SYMBOL_BINARY_DIFF_BODY:
1342 case DIFF_SYMBOL_CONTEXT_FRAGINFO:
1343 emit_line(o, "", "", line, len);
1344 break;
1345 case DIFF_SYMBOL_CONTEXT_INCOMPLETE:
1346 case DIFF_SYMBOL_CONTEXT_MARKER:
1347 context = diff_get_color_opt(o, DIFF_CONTEXT);
1348 reset = diff_get_color_opt(o, DIFF_RESET);
1349 emit_line(o, context, reset, line, len);
1350 break;
1351 case DIFF_SYMBOL_SEPARATOR:
1352 fprintf(o->file, "%s%c",
1353 diff_line_prefix(o),
1354 o->line_termination);
1355 break;
1356 case DIFF_SYMBOL_CONTEXT:
1357 set = diff_get_color_opt(o, DIFF_CONTEXT);
1358 reset = diff_get_color_opt(o, DIFF_RESET);
1359 set_sign = NULL;
1360 if (o->flags.dual_color_diffed_diffs) {
1361 char c = !len ? 0 : line[0];
1362 if (c == '+')
1364 set = diff_get_color_opt(o, DIFF_FILE_NEW);
1365 else if (c == '@')
1366 set = diff_get_color_opt(o, DIFF_FRAGINFO);
1367 else if (c == '-')
1368 set = diff_get_color_opt(o, DIFF_FILE_OLD);
1369 }
1370 emit_line_ws_markup(o, set_sign, set, reset,
1371 OUTPUT_INDICATOR_CONTEXT, line, len,
1372 flags & (DIFF_SYMBOL_CONTENT_WS_MASK), 0);
1373 break;
1374 case DIFF_SYMBOL_PLUS:
1375 switch (flags & (DIFF_SYMBOL_MOVED_LINE |
1376 DIFF_SYMBOL_MOVED_LINE_ALT |
1377 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING)) {
1378 case DIFF_SYMBOL_MOVED_LINE |
1379 DIFF_SYMBOL_MOVED_LINE_ALT |
1380 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1381 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED_ALT_DIM);
1382 break;
1383 case DIFF_SYMBOL_MOVED_LINE |
1384 DIFF_SYMBOL_MOVED_LINE_ALT:
1385 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED_ALT);
1386 break;
1387 case DIFF_SYMBOL_MOVED_LINE |
1388 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1389 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED_DIM);
1390 break;
1391 case DIFF_SYMBOL_MOVED_LINE:
1392 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED);
1393 break;
1394 default:
1395 set = diff_get_color_opt(o, DIFF_FILE_NEW);
1396 }
1397 reset = diff_get_color_opt(o, DIFF_RESET);
1398 if (!o->flags.dual_color_diffed_diffs)
1399 set_sign = NULL;
1400 else {
1401 char c = !len ? 0 : line[0];
1402 set_sign = set;
1404 if (c == '-')
1405 set = diff_get_color_opt(o, DIFF_FILE_OLD_BOLD);
1406 else if (c == '@')
1407 set = diff_get_color_opt(o, DIFF_FRAGINFO);
1408 else if (c == '+')
1409 set = diff_get_color_opt(o, DIFF_FILE_NEW_BOLD);
1410 else
1411 set = diff_get_color_opt(o, DIFF_CONTEXT_BOLD);
1412 flags &= ~DIFF_SYMBOL_CONTENT_WS_MASK;
1413 }
1414 emit_line_ws_markup(o, set_sign, set, reset,
1415 OUTPUT_INDICATOR_NEW, line, len,
1416 flags & DIFF_SYMBOL_CONTENT_WS_MASK,
1417 flags & DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF);
1418 break;
1419 case DIFF_SYMBOL_MINUS:
1420 switch (flags & (DIFF_SYMBOL_MOVED_LINE |
1421 DIFF_SYMBOL_MOVED_LINE_ALT |
1422 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING)) {
1423 case DIFF_SYMBOL_MOVED_LINE |
1424 DIFF_SYMBOL_MOVED_LINE_ALT |
1425 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1426 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED_ALT_DIM);
1427 break;
1428 case DIFF_SYMBOL_MOVED_LINE |
1429 DIFF_SYMBOL_MOVED_LINE_ALT:
1430 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED_ALT);
1431 break;
1432 case DIFF_SYMBOL_MOVED_LINE |
1433 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1434 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED_DIM);
1435 break;
1436 case DIFF_SYMBOL_MOVED_LINE:
1437 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED);
1438 break;
1439 default:
1440 set = diff_get_color_opt(o, DIFF_FILE_OLD);
1441 }
1442 reset = diff_get_color_opt(o, DIFF_RESET);
1443 if (!o->flags.dual_color_diffed_diffs)
1444 set_sign = NULL;
1445 else {
1446 char c = !len ? 0 : line[0];
1447 set_sign = set;
1449 if (c == '+')
1450 set = diff_get_color_opt(o, DIFF_FILE_NEW_DIM);
1451 else if (c == '@')
1452 set = diff_get_color_opt(o, DIFF_FRAGINFO);
1453 else if (c == '-')
1454 set = diff_get_color_opt(o, DIFF_FILE_OLD_DIM);
1455 else
1456 set = diff_get_color_opt(o, DIFF_CONTEXT_DIM);
1457 }
1458 emit_line_ws_markup(o, set_sign, set, reset,
1459 OUTPUT_INDICATOR_OLD, line, len,
1460 flags & DIFF_SYMBOL_CONTENT_WS_MASK, 0);
1461 break;
1462 case DIFF_SYMBOL_WORDS_PORCELAIN:
1463 context = diff_get_color_opt(o, DIFF_CONTEXT);
1464 reset = diff_get_color_opt(o, DIFF_RESET);
1465 emit_line(o, context, reset, line, len);
1466 fputs("~\n", o->file);
1467 break;
1468 case DIFF_SYMBOL_WORDS:
1469 context = diff_get_color_opt(o, DIFF_CONTEXT);
1470 reset = diff_get_color_opt(o, DIFF_RESET);
1471 /*
1472 * Skip the prefix character, if any. With
1473 * diff_suppress_blank_empty, there may be
1474 * none.
1475 */
1476 if (line[0] != '\n') {
1477 line++;
1478 len--;
1479 }
1480 emit_line(o, context, reset, line, len);
1481 break;
1482 case DIFF_SYMBOL_FILEPAIR_PLUS:
1483 meta = diff_get_color_opt(o, DIFF_METAINFO);
1484 reset = diff_get_color_opt(o, DIFF_RESET);
1485 fprintf(o->file, "%s%s+++ %s%s%s\n", diff_line_prefix(o), meta,
1486 line, reset,
1487 strchr(line, ' ') ? "\t" : "");
1488 break;
1489 case DIFF_SYMBOL_FILEPAIR_MINUS:
1490 meta = diff_get_color_opt(o, DIFF_METAINFO);
1491 reset = diff_get_color_opt(o, DIFF_RESET);
1492 fprintf(o->file, "%s%s--- %s%s%s\n", diff_line_prefix(o), meta,
1493 line, reset,
1494 strchr(line, ' ') ? "\t" : "");
1495 break;
1496 case DIFF_SYMBOL_BINARY_FILES:
1497 case DIFF_SYMBOL_HEADER:
1498 fprintf(o->file, "%s", line);
1499 break;
1500 case DIFF_SYMBOL_BINARY_DIFF_HEADER:
1501 fprintf(o->file, "%sGIT binary patch\n", diff_line_prefix(o));
1502 break;
1503 case DIFF_SYMBOL_BINARY_DIFF_HEADER_DELTA:
1504 fprintf(o->file, "%sdelta %s\n", diff_line_prefix(o), line);
1505 break;
1506 case DIFF_SYMBOL_BINARY_DIFF_HEADER_LITERAL:
1507 fprintf(o->file, "%sliteral %s\n", diff_line_prefix(o), line);
1508 break;
1509 case DIFF_SYMBOL_BINARY_DIFF_FOOTER:
1510 fputs(diff_line_prefix(o), o->file);
1511 fputc('\n', o->file);
1512 break;
1513 case DIFF_SYMBOL_REWRITE_DIFF:
1514 fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
1515 reset = diff_get_color_opt(o, DIFF_RESET);
1516 emit_line(o, fraginfo, reset, line, len);
1517 break;
1518 case DIFF_SYMBOL_SUBMODULE_ADD:
1519 set = diff_get_color_opt(o, DIFF_FILE_NEW);
1520 reset = diff_get_color_opt(o, DIFF_RESET);
1521 emit_line(o, set, reset, line, len);
1522 break;
1523 case DIFF_SYMBOL_SUBMODULE_DEL:
1524 set = diff_get_color_opt(o, DIFF_FILE_OLD);
1525 reset = diff_get_color_opt(o, DIFF_RESET);
1526 emit_line(o, set, reset, line, len);
1527 break;
1528 case DIFF_SYMBOL_SUBMODULE_UNTRACKED:
1529 fprintf(o->file, "%sSubmodule %s contains untracked content\n",
1530 diff_line_prefix(o), line);
1531 break;
1532 case DIFF_SYMBOL_SUBMODULE_MODIFIED:
1533 fprintf(o->file, "%sSubmodule %s contains modified content\n",
1534 diff_line_prefix(o), line);
1535 break;
1536 case DIFF_SYMBOL_STATS_SUMMARY_NO_FILES:
1537 emit_line(o, "", "", " 0 files changed\n",
1538 strlen(" 0 files changed\n"));
1539 break;
1540 case DIFF_SYMBOL_STATS_SUMMARY_ABBREV:
1541 emit_line(o, "", "", " ...\n", strlen(" ...\n"));
1542 break;
1543 case DIFF_SYMBOL_WORD_DIFF:
1544 fprintf(o->file, "%.*s", len, line);
1545 break;
1546 case DIFF_SYMBOL_STAT_SEP:
1547 fputs(o->stat_sep, o->file);
1548 break;
1549 default:
1550 BUG("unknown diff symbol");
1551 }
1552 strbuf_release(&sb);
1553}
1554static void emit_diff_symbol(struct diff_options *o, enum diff_symbol s,
1556 const char *line, int len, unsigned flags)
1557{
1558 struct emitted_diff_symbol e = {line, len, flags, 0, 0, s};
1559 if (o->emitted_symbols)
1561 append_emitted_diff_symbol(o, &e);
1562 else
1563 emit_diff_symbol_from_struct(o, &e);
1564}
1565void diff_emit_submodule_del(struct diff_options *o, const char *line)
1567{
1568 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_DEL, line, strlen(line), 0);
1569}
1570void diff_emit_submodule_add(struct diff_options *o, const char *line)
1572{
1573 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_ADD, line, strlen(line), 0);
1574}
1575void diff_emit_submodule_untracked(struct diff_options *o, const char *path)
1577{
1578 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_UNTRACKED,
1579 path, strlen(path), 0);
1580}
1581void diff_emit_submodule_modified(struct diff_options *o, const char *path)
1583{
1584 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_MODIFIED,
1585 path, strlen(path), 0);
1586}
1587void diff_emit_submodule_header(struct diff_options *o, const char *header)
1589{
1590 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_HEADER,
1591 header, strlen(header), 0);
1592}
1593void diff_emit_submodule_error(struct diff_options *o, const char *err)
1595{
1596 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_ERROR, err, strlen(err), 0);
1597}
1598void diff_emit_submodule_pipethrough(struct diff_options *o,
1600 const char *line, int len)
1601{
1602 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_PIPETHROUGH, line, len, 0);
1603}
1604static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
1606{
1607 if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
1608 ecbdata->blank_at_eof_in_preimage &&
1609 ecbdata->blank_at_eof_in_postimage &&
1610 ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
1611 ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
1612 return 0;
1613 return ws_blank_line(line, len, ecbdata->ws_rule);
1614}
1615static void emit_add_line(const char *reset,
1617 struct emit_callback *ecbdata,
1618 const char *line, int len)
1619{
1620 unsigned flags = WSEH_NEW | ecbdata->ws_rule;
1621 if (new_blank_line_at_eof(ecbdata, line, len))
1622 flags |= DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF;
1623 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_PLUS, line, len, flags);
1625}
1626static void emit_del_line(const char *reset,
1628 struct emit_callback *ecbdata,
1629 const char *line, int len)
1630{
1631 unsigned flags = WSEH_OLD | ecbdata->ws_rule;
1632 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_MINUS, line, len, flags);
1633}
1634static void emit_context_line(const char *reset,
1636 struct emit_callback *ecbdata,
1637 const char *line, int len)
1638{
1639 unsigned flags = WSEH_CONTEXT | ecbdata->ws_rule;
1640 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_CONTEXT, line, len, flags);
1641}
1642static void emit_hunk_header(struct emit_callback *ecbdata,
1644 const char *line, int len)
1645{
1646 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
1647 const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
1648 const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
1649 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1650 const char *reverse = ecbdata->color_diff ? GIT_COLOR_REVERSE : "";
1651 static const char atat[2] = { '@', '@' };
1652 const char *cp, *ep;
1653 struct strbuf msgbuf = STRBUF_INIT;
1654 int org_len = len;
1655 int i = 1;
1656 /*
1658 * As a hunk header must begin with "@@ -<old>, +<new> @@",
1659 * it always is at least 10 bytes long.
1660 */
1661 if (len < 10 ||
1662 memcmp(line, atat, 2) ||
1663 !(ep = memmem(line + 2, len - 2, atat, 2))) {
1664 emit_diff_symbol(ecbdata->opt,
1665 DIFF_SYMBOL_CONTEXT_MARKER, line, len, 0);
1666 return;
1667 }
1668 ep += 2; /* skip over @@ */
1669 /* The hunk header in fraginfo color */
1671 if (ecbdata->opt->flags.dual_color_diffed_diffs)
1672 strbuf_addstr(&msgbuf, reverse);
1673 strbuf_addstr(&msgbuf, frag);
1674 strbuf_add(&msgbuf, line, ep - line);
1675 strbuf_addstr(&msgbuf, reset);
1676 /*
1678 * trailing "\r\n"
1679 */
1680 for ( ; i < 3; i++)
1681 if (line[len - i] == '\r' || line[len - i] == '\n')
1682 len--;
1683 /* blank before the func header */
1685 for (cp = ep; ep - line < len; ep++)
1686 if (*ep != ' ' && *ep != '\t')
1687 break;
1688 if (ep != cp) {
1689 strbuf_addstr(&msgbuf, context);
1690 strbuf_add(&msgbuf, cp, ep - cp);
1691 strbuf_addstr(&msgbuf, reset);
1692 }
1693 if (ep < line + len) {
1695 strbuf_addstr(&msgbuf, func);
1696 strbuf_add(&msgbuf, ep, line + len - ep);
1697 strbuf_addstr(&msgbuf, reset);
1698 }
1699 strbuf_add(&msgbuf, line + len, org_len - len);
1701 strbuf_complete_line(&msgbuf);
1702 emit_diff_symbol(ecbdata->opt,
1703 DIFF_SYMBOL_CONTEXT_FRAGINFO, msgbuf.buf, msgbuf.len, 0);
1704 strbuf_release(&msgbuf);
1705}
1706static struct diff_tempfile *claim_diff_tempfile(void)
1708{
1709 int i;
1710 for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
1711 if (!diff_temp[i].name)
1712 return diff_temp + i;
1713 BUG("diff is failing to clean up its tempfiles");
1714}
1715static void remove_tempfile(void)
1717{
1718 int i;
1719 for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
1720 if (is_tempfile_active(diff_temp[i].tempfile))
1721 delete_tempfile(&diff_temp[i].tempfile);
1722 diff_temp[i].name = NULL;
1723 }
1724}
1725static void add_line_count(struct strbuf *out, int count)
1727{
1728 switch (count) {
1729 case 0:
1730 strbuf_addstr(out, "0,0");
1731 break;
1732 case 1:
1733 strbuf_addstr(out, "1");
1734 break;
1735 default:
1736 strbuf_addf(out, "1,%d", count);
1737 break;
1738 }
1739}
1740static void emit_rewrite_lines(struct emit_callback *ecb,
1742 int prefix, const char *data, int size)
1743{
1744 const char *endp = NULL;
1745 const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
1746 while (0 < size) {
1748 int len;
1749 endp = memchr(data, '\n', size);
1751 len = endp ? (endp - data + 1) : size;
1752 if (prefix != '+') {
1753 ecb->lno_in_preimage++;
1754 emit_del_line(reset, ecb, data, len);
1755 } else {
1756 ecb->lno_in_postimage++;
1757 emit_add_line(reset, ecb, data, len);
1758 }
1759 size -= len;
1760 data += len;
1761 }
1762 if (!endp)
1763 emit_diff_symbol(ecb->opt, DIFF_SYMBOL_NO_LF_EOF, NULL, 0, 0);
1764}
1765static void emit_rewrite_diff(const char *name_a,
1767 const char *name_b,
1768 struct diff_filespec *one,
1769 struct diff_filespec *two,
1770 struct userdiff_driver *textconv_one,
1771 struct userdiff_driver *textconv_two,
1772 struct diff_options *o)
1773{
1774 int lc_a, lc_b;
1775 static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
1776 const char *a_prefix, *b_prefix;
1777 char *data_one, *data_two;
1778 size_t size_one, size_two;
1779 struct emit_callback ecbdata;
1780 struct strbuf out = STRBUF_INIT;
1781 if (diff_mnemonic_prefix && o->flags.reverse_diff) {
1783 a_prefix = o->b_prefix;
1784 b_prefix = o->a_prefix;
1785 } else {
1786 a_prefix = o->a_prefix;
1787 b_prefix = o->b_prefix;
1788 }
1789 name_a += (*name_a == '/');
1791 name_b += (*name_b == '/');
1792 strbuf_reset(&a_name);
1794 strbuf_reset(&b_name);
1795 quote_two_c_style(&a_name, a_prefix, name_a, 0);
1796 quote_two_c_style(&b_name, b_prefix, name_b, 0);
1797 size_one = fill_textconv(o->repo, textconv_one, one, &data_one);
1799 size_two = fill_textconv(o->repo, textconv_two, two, &data_two);
1800 memset(&ecbdata, 0, sizeof(ecbdata));
1802 ecbdata.color_diff = want_color(o->use_color);
1803 ecbdata.ws_rule = whitespace_rule(o->repo->index, name_b);
1804 ecbdata.opt = o;
1805 if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
1806 mmfile_t mf1, mf2;
1807 mf1.ptr = (char *)data_one;
1808 mf2.ptr = (char *)data_two;
1809 mf1.size = size_one;
1810 mf2.size = size_two;
1811 check_blank_at_eof(&mf1, &mf2, &ecbdata);
1812 }
1813 ecbdata.lno_in_preimage = 1;
1814 ecbdata.lno_in_postimage = 1;
1815 lc_a = count_lines(data_one, size_one);
1817 lc_b = count_lines(data_two, size_two);
1818 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_MINUS,
1820 a_name.buf, a_name.len, 0);
1821 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_PLUS,
1822 b_name.buf, b_name.len, 0);
1823 strbuf_addstr(&out, "@@ -");
1825 if (!o->irreversible_delete)
1826 add_line_count(&out, lc_a);
1827 else
1828 strbuf_addstr(&out, "?,?");
1829 strbuf_addstr(&out, " +");
1830 add_line_count(&out, lc_b);
1831 strbuf_addstr(&out, " @@\n");
1832 emit_diff_symbol(o, DIFF_SYMBOL_REWRITE_DIFF, out.buf, out.len, 0);
1833 strbuf_release(&out);
1834 if (lc_a && !o->irreversible_delete)
1836 emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
1837 if (lc_b)
1838 emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
1839 if (textconv_one)
1840 free((char *)data_one);
1841 if (textconv_two)
1842 free((char *)data_two);
1843}
1844struct diff_words_buffer {
1846 mmfile_t text;
1847 unsigned long alloc;
1848 struct diff_words_orig {
1849 const char *begin, *end;
1850 } *orig;
1851 int orig_nr, orig_alloc;
1852};
1853static void diff_words_append(char *line, unsigned long len,
1855 struct diff_words_buffer *buffer)
1856{
1857 ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
1858 line++;
1859 len--;
1860 memcpy(buffer->text.ptr + buffer->text.size, line, len);
1861 buffer->text.size += len;
1862 buffer->text.ptr[buffer->text.size] = '\0';
1863}
1864struct diff_words_style_elem {
1866 const char *prefix;
1867 const char *suffix;
1868 const char *color; /* NULL; filled in by the setup code if
1869 * color is enabled */
1870};
1871struct diff_words_style {
1873 enum diff_words_type type;
1874 struct diff_words_style_elem new_word, old_word, ctx;
1875 const char *newline;
1876};
1877static struct diff_words_style diff_words_styles[] = {
1879 { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
1880 { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
1881 { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
1882};
1883struct diff_words_data {
1885 struct diff_words_buffer minus, plus;
1886 const char *current_plus;
1887 int last_minus;
1888 struct diff_options *opt;
1889 regex_t *word_regex;
1890 enum diff_words_type type;
1891 struct diff_words_style *style;
1892};
1893static int fn_out_diff_words_write_helper(struct diff_options *o,
1895 struct diff_words_style_elem *st_el,
1896 const char *newline,
1897 size_t count, const char *buf)
1898{
1899 int print = 0;
1900 struct strbuf sb = STRBUF_INIT;
1901 while (count) {
1903 char *p = memchr(buf, '\n', count);
1904 if (print)
1905 strbuf_addstr(&sb, diff_line_prefix(o));
1906 if (p != buf) {
1908 const char *reset = st_el->color && *st_el->color ?
1909 GIT_COLOR_RESET : NULL;
1910 if (st_el->color && *st_el->color)
1911 strbuf_addstr(&sb, st_el->color);
1912 strbuf_addstr(&sb, st_el->prefix);
1913 strbuf_add(&sb, buf, p ? p - buf : count);
1914 strbuf_addstr(&sb, st_el->suffix);
1915 if (reset)
1916 strbuf_addstr(&sb, reset);
1917 }
1918 if (!p)
1919 goto out;
1920 strbuf_addstr(&sb, newline);
1922 count -= p + 1 - buf;
1923 buf = p + 1;
1924 print = 1;
1925 if (count) {
1926 emit_diff_symbol(o, DIFF_SYMBOL_WORD_DIFF,
1927 sb.buf, sb.len, 0);
1928 strbuf_reset(&sb);
1929 }
1930 }
1931out:
1933 if (sb.len)
1934 emit_diff_symbol(o, DIFF_SYMBOL_WORD_DIFF,
1935 sb.buf, sb.len, 0);
1936 strbuf_release(&sb);
1937 return 0;
1938}
1939/*
1941 * '--color-words' algorithm can be described as:
1942 *
1943 * 1. collect the minus/plus lines of a diff hunk, divided into
1944 * minus-lines and plus-lines;
1945 *
1946 * 2. break both minus-lines and plus-lines into words and
1947 * place them into two mmfile_t with one word for each line;
1948 *
1949 * 3. use xdiff to run diff on the two mmfile_t to get the words level diff;
1950 *
1951 * And for the common parts of the both file, we output the plus side text.
1952 * diff_words->current_plus is used to trace the current position of the plus file
1953 * which printed. diff_words->last_minus is used to trace the last minus word
1954 * printed.
1955 *
1956 * For '--graph' to work with '--color-words', we need to output the graph prefix
1957 * on each line of color words output. Generally, there are two conditions on
1958 * which we should output the prefix.
1959 *
1960 * 1. diff_words->last_minus == 0 &&
1961 * diff_words->current_plus == diff_words->plus.text.ptr
1962 *
1963 * that is: the plus text must start as a new line, and if there is no minus
1964 * word printed, a graph prefix must be printed.
1965 *
1966 * 2. diff_words->current_plus > diff_words->plus.text.ptr &&
1967 * *(diff_words->current_plus - 1) == '\n'
1968 *
1969 * that is: a graph prefix must be printed following a '\n'
1970 */
1971static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
1972{
1973 if ((diff_words->last_minus == 0 &&
1974 diff_words->current_plus == diff_words->plus.text.ptr) ||
1975 (diff_words->current_plus > diff_words->plus.text.ptr &&
1976 *(diff_words->current_plus - 1) == '\n')) {
1977 return 1;
1978 } else {
1979 return 0;
1980 }
1981}
1982static void fn_out_diff_words_aux(void *priv,
1984 long minus_first, long minus_len,
1985 long plus_first, long plus_len,
1986 const char *func, long funclen)
1987{
1988 struct diff_words_data *diff_words = priv;
1989 struct diff_words_style *style = diff_words->style;
1990 const char *minus_begin, *minus_end, *plus_begin, *plus_end;
1991 struct diff_options *opt = diff_words->opt;
1992 const char *line_prefix;
1993 assert(opt);
1995 line_prefix = diff_line_prefix(opt);
1996 /* POSIX requires that first be decremented by one if len == 0... */
1998 if (minus_len) {
1999 minus_begin = diff_words->minus.orig[minus_first].begin;
2000 minus_end =
2001 diff_words->minus.orig[minus_first + minus_len - 1].end;
2002 } else
2003 minus_begin = minus_end =
2004 diff_words->minus.orig[minus_first].end;
2005 if (plus_len) {
2007 plus_begin = diff_words->plus.orig[plus_first].begin;
2008 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
2009 } else
2010 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
2011 if (color_words_output_graph_prefix(diff_words)) {
2013 fputs(line_prefix, diff_words->opt->file);
2014 }
2015 if (diff_words->current_plus != plus_begin) {
2016 fn_out_diff_words_write_helper(diff_words->opt,
2017 &style->ctx, style->newline,
2018 plus_begin - diff_words->current_plus,
2019 diff_words->current_plus);
2020 }
2021 if (minus_begin != minus_end) {
2022 fn_out_diff_words_write_helper(diff_words->opt,
2023 &style->old_word, style->newline,
2024 minus_end - minus_begin, minus_begin);
2025 }
2026 if (plus_begin != plus_end) {
2027 fn_out_diff_words_write_helper(diff_words->opt,
2028 &style->new_word, style->newline,
2029 plus_end - plus_begin, plus_begin);
2030 }
2031 diff_words->current_plus = plus_end;
2033 diff_words->last_minus = minus_first;
2034}
2035/* This function starts looking at *begin, and returns 0 iff a word was found. */
2037static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
2038 int *begin, int *end)
2039{
2040 if (word_regex && *begin < buffer->size) {
2041 regmatch_t match[1];
2042 if (!regexec_buf(word_regex, buffer->ptr + *begin,
2043 buffer->size - *begin, 1, match, 0)) {
2044 char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
2045 '\n', match[0].rm_eo - match[0].rm_so);
2046 *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
2047 *begin += match[0].rm_so;
2048 return *begin >= *end;
2049 }
2050 return -1;
2051 }
2052 /* find the next word */
2054 while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
2055 (*begin)++;
2056 if (*begin >= buffer->size)
2057 return -1;
2058 /* find the end of the word */
2060 *end = *begin + 1;
2061 while (*end < buffer->size && !isspace(buffer->ptr[*end]))
2062 (*end)++;
2063 return 0;
2065}
2066/*
2068 * This function splits the words in buffer->text, stores the list with
2069 * newline separator into out, and saves the offsets of the original words
2070 * in buffer->orig.
2071 */
2072static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
2073 regex_t *word_regex)
2074{
2075 int i, j;
2076 long alloc = 0;
2077 out->size = 0;
2079 out->ptr = NULL;
2080 /* fake an empty "0th" word */
2082 ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
2083 buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
2084 buffer->orig_nr = 1;
2085 for (i = 0; i < buffer->text.size; i++) {
2087 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
2088 return;
2089 /* store original boundaries */
2091 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
2092 buffer->orig_alloc);
2093 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
2094 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
2095 buffer->orig_nr++;
2096 /* store one word */
2098 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
2099 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
2100 out->ptr[out->size + j - i] = '\n';
2101 out->size += j - i + 1;
2102 i = j - 1;
2104 }
2105}
2106/* this executes the word diff on the accumulated buffers */
2108static void diff_words_show(struct diff_words_data *diff_words)
2109{
2110 xpparam_t xpp;
2111 xdemitconf_t xecfg;
2112 mmfile_t minus, plus;
2113 struct diff_words_style *style = diff_words->style;
2114 struct diff_options *opt = diff_words->opt;
2116 const char *line_prefix;
2117 assert(opt);
2119 line_prefix = diff_line_prefix(opt);
2120 /* special case: only removal */
2122 if (!diff_words->plus.text.size) {
2123 emit_diff_symbol(diff_words->opt, DIFF_SYMBOL_WORD_DIFF,
2124 line_prefix, strlen(line_prefix), 0);
2125 fn_out_diff_words_write_helper(diff_words->opt,
2126 &style->old_word, style->newline,
2127 diff_words->minus.text.size,
2128 diff_words->minus.text.ptr);
2129 diff_words->minus.text.size = 0;
2130 return;
2131 }
2132 diff_words->current_plus = diff_words->plus.text.ptr;
2134 diff_words->last_minus = 0;
2135 memset(&xpp, 0, sizeof(xpp));
2137 memset(&xecfg, 0, sizeof(xecfg));
2138 diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
2139 diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
2140 xpp.flags = 0;
2141 /* as only the hunk header will be parsed, we need a 0-context */
2142 xecfg.ctxlen = 0;
2143 if (xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, NULL,
2144 diff_words, &xpp, &xecfg))
2145 die("unable to generate word diff");
2146 free(minus.ptr);
2147 free(plus.ptr);
2148 if (diff_words->current_plus != diff_words->plus.text.ptr +
2149 diff_words->plus.text.size) {
2150 if (color_words_output_graph_prefix(diff_words))
2151 emit_diff_symbol(diff_words->opt, DIFF_SYMBOL_WORD_DIFF,
2152 line_prefix, strlen(line_prefix), 0);
2153 fn_out_diff_words_write_helper(diff_words->opt,
2154 &style->ctx, style->newline,
2155 diff_words->plus.text.ptr + diff_words->plus.text.size
2156 - diff_words->current_plus, diff_words->current_plus);
2157 }
2158 diff_words->minus.text.size = diff_words->plus.text.size = 0;
2159}
2160/* In "color-words" mode, show word-diff of words accumulated in the buffer */
2162static void diff_words_flush(struct emit_callback *ecbdata)
2163{
2164 struct diff_options *wo = ecbdata->diff_words->opt;
2165 if (ecbdata->diff_words->minus.text.size ||
2167 ecbdata->diff_words->plus.text.size)
2168 diff_words_show(ecbdata->diff_words);
2169 if (wo->emitted_symbols) {
2171 struct diff_options *o = ecbdata->opt;
2172 struct emitted_diff_symbols *wol = wo->emitted_symbols;
2173 int i;
2174 /*
2176 * NEEDSWORK:
2177 * Instead of appending each, concat all words to a line?
2178 */
2179 for (i = 0; i < wol->nr; i++)
2180 append_emitted_diff_symbol(o, &wol->buf[i]);
2181 for (i = 0; i < wol->nr; i++)
2183 free((void *)wol->buf[i].line);
2184 wol->nr = 0;
2186 }
2187}
2188static void diff_filespec_load_driver(struct diff_filespec *one,
2190 struct index_state *istate)
2191{
2192 /* Use already-loaded driver */
2193 if (one->driver)
2194 return;
2195 if (S_ISREG(one->mode))
2197 one->driver = userdiff_find_by_path(istate, one->path);
2198 /* Fallback to default settings */
2200 if (!one->driver)
2201 one->driver = userdiff_find_by_name("default");
2202}
2203static const char *userdiff_word_regex(struct diff_filespec *one,
2205 struct index_state *istate)
2206{
2207 diff_filespec_load_driver(one, istate);
2208 return one->driver->word_regex;
2209}
2210static void init_diff_words_data(struct emit_callback *ecbdata,
2212 struct diff_options *orig_opts,
2213 struct diff_filespec *one,
2214 struct diff_filespec *two)
2215{
2216 int i;
2217 struct diff_options *o = xmalloc(sizeof(struct diff_options));
2218 memcpy(o, orig_opts, sizeof(struct diff_options));
2219 ecbdata->diff_words =
2221 xcalloc(1, sizeof(struct diff_words_data));
2222 ecbdata->diff_words->type = o->word_diff;
2223 ecbdata->diff_words->opt = o;
2224 if (orig_opts->emitted_symbols)
2226 o->emitted_symbols =
2227 xcalloc(1, sizeof(struct emitted_diff_symbols));
2228 if (!o->word_regex)
2230 o->word_regex = userdiff_word_regex(one, o->repo->index);
2231 if (!o->word_regex)
2232 o->word_regex = userdiff_word_regex(two, o->repo->index);
2233 if (!o->word_regex)
2234 o->word_regex = diff_word_regex_cfg;
2235 if (o->word_regex) {
2236 ecbdata->diff_words->word_regex = (regex_t *)
2237 xmalloc(sizeof(regex_t));
2238 if (regcomp(ecbdata->diff_words->word_regex,
2239 o->word_regex,
2240 REG_EXTENDED | REG_NEWLINE))
2241 die("invalid regular expression: %s",
2242 o->word_regex);
2243 }
2244 for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
2245 if (o->word_diff == diff_words_styles[i].type) {
2246 ecbdata->diff_words->style =
2247 &diff_words_styles[i];
2248 break;
2249 }
2250 }
2251 if (want_color(o->use_color)) {
2252 struct diff_words_style *st = ecbdata->diff_words->style;
2253 st->old_word.color = diff_get_color_opt(o, DIFF_FILE_OLD);
2254 st->new_word.color = diff_get_color_opt(o, DIFF_FILE_NEW);
2255 st->ctx.color = diff_get_color_opt(o, DIFF_CONTEXT);
2256 }
2257}
2258static void free_diff_words_data(struct emit_callback *ecbdata)
2260{
2261 if (ecbdata->diff_words) {
2262 diff_words_flush(ecbdata);
2263 free (ecbdata->diff_words->opt->emitted_symbols);
2264 free (ecbdata->diff_words->opt);
2265 free (ecbdata->diff_words->minus.text.ptr);
2266 free (ecbdata->diff_words->minus.orig);
2267 free (ecbdata->diff_words->plus.text.ptr);
2268 free (ecbdata->diff_words->plus.orig);
2269 if (ecbdata->diff_words->word_regex) {
2270 regfree(ecbdata->diff_words->word_regex);
2271 free(ecbdata->diff_words->word_regex);
2272 }
2273 FREE_AND_NULL(ecbdata->diff_words);
2274 }
2275}
2276const char *diff_get_color(int diff_use_color, enum color_diff ix)
2278{
2279 if (want_color(diff_use_color))
2280 return diff_colors[ix];
2281 return "";
2282}
2283const char *diff_line_prefix(struct diff_options *opt)
2285{
2286 struct strbuf *msgbuf;
2287 if (!opt->output_prefix)
2288 return "";
2289 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
2291 return msgbuf->buf;
2292}
2293static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
2295{
2296 const char *cp;
2297 unsigned long allot;
2298 size_t l = len;
2299 cp = line;
2301 allot = l;
2302 while (0 < l) {
2303 (void) utf8_width(&cp, &l);
2304 if (!cp)
2305 break; /* truncated in the middle? */
2306 }
2307 return allot - l;
2308}
2309static void find_lno(const char *line, struct emit_callback *ecbdata)
2311{
2312 const char *p;
2313 ecbdata->lno_in_preimage = 0;
2314 ecbdata->lno_in_postimage = 0;
2315 p = strchr(line, '-');
2316 if (!p)
2317 return; /* cannot happen */
2318 ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
2319 p = strchr(p, '+');
2320 if (!p)
2321 return; /* cannot happen */
2322 ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
2323}
2324static void fn_out_consume(void *priv, char *line, unsigned long len)
2326{
2327 struct emit_callback *ecbdata = priv;
2328 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
2329 struct diff_options *o = ecbdata->opt;
2330 o->found_changes = 1;
2332 if (ecbdata->header) {
2334 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
2335 ecbdata->header->buf, ecbdata->header->len, 0);
2336 strbuf_reset(ecbdata->header);
2337 ecbdata->header = NULL;
2338 }
2339 if (ecbdata->label_path[0]) {
2341 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_MINUS,
2342 ecbdata->label_path[0],
2343 strlen(ecbdata->label_path[0]), 0);
2344 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_PLUS,
2345 ecbdata->label_path[1],
2346 strlen(ecbdata->label_path[1]), 0);
2347 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
2348 }
2349 if (diff_suppress_blank_empty
2351 && len == 2 && line[0] == ' ' && line[1] == '\n') {
2352 line[0] = '\n';
2353 len = 1;
2354 }
2355 if (line[0] == '@') {
2357 if (ecbdata->diff_words)
2358 diff_words_flush(ecbdata);
2359 len = sane_truncate_line(ecbdata, line, len);
2360 find_lno(line, ecbdata);
2361 emit_hunk_header(ecbdata, line, len);
2362 return;
2363 }
2364 if (ecbdata->diff_words) {
2366 enum diff_symbol s =
2367 ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN ?
2368 DIFF_SYMBOL_WORDS_PORCELAIN : DIFF_SYMBOL_WORDS;
2369 if (line[0] == '-') {
2370 diff_words_append(line, len,
2371 &ecbdata->diff_words->minus);
2372 return;
2373 } else if (line[0] == '+') {
2374 diff_words_append(line, len,
2375 &ecbdata->diff_words->plus);
2376 return;
2377 } else if (starts_with(line, "\\ ")) {
2378 /*
2379 * Eat the "no newline at eof" marker as if we
2380 * saw a "+" or "-" line with nothing on it,
2381 * and return without diff_words_flush() to
2382 * defer processing. If this is the end of
2383 * preimage, more "+" lines may come after it.
2384 */
2385 return;
2386 }
2387 diff_words_flush(ecbdata);
2388 emit_diff_symbol(o, s, line, len, 0);
2389 return;
2390 }
2391 switch (line[0]) {
2393 case '+':
2394 ecbdata->lno_in_postimage++;
2395 emit_add_line(reset, ecbdata, line + 1, len - 1);
2396 break;
2397 case '-':
2398 ecbdata->lno_in_preimage++;
2399 emit_del_line(reset, ecbdata, line + 1, len - 1);
2400 break;
2401 case ' ':
2402 ecbdata->lno_in_postimage++;
2403 ecbdata->lno_in_preimage++;
2404 emit_context_line(reset, ecbdata, line + 1, len - 1);
2405 break;
2406 default:
2407 /* incomplete line at the end */
2408 ecbdata->lno_in_preimage++;
2409 emit_diff_symbol(o, DIFF_SYMBOL_CONTEXT_INCOMPLETE,
2410 line, len, 0);
2411 break;
2412 }
2413}
2414static void pprint_rename(struct strbuf *name, const char *a, const char *b)
2416{
2417 const char *old_name = a;
2418 const char *new_name = b;
2419 int pfx_length, sfx_length;
2420 int pfx_adjust_for_slash;
2421 int len_a = strlen(a);
2422 int len_b = strlen(b);
2423 int a_midlen, b_midlen;
2424 int qlen_a = quote_c_style(a, NULL, NULL, 0);
2425 int qlen_b = quote_c_style(b, NULL, NULL, 0);
2426 if (qlen_a || qlen_b) {
2428 quote_c_style(a, name, NULL, 0);
2429 strbuf_addstr(name, " => ");
2430 quote_c_style(b, name, NULL, 0);
2431 return;
2432 }
2433 /* Find common prefix */
2435 pfx_length = 0;
2436 while (*old_name && *new_name && *old_name == *new_name) {
2437 if (*old_name == '/')
2438 pfx_length = old_name - a + 1;
2439 old_name++;
2440 new_name++;
2441 }
2442 /* Find common suffix */
2444 old_name = a + len_a;
2445 new_name = b + len_b;
2446 sfx_length = 0;
2447 /*
2448 * If there is a common prefix, it must end in a slash. In
2449 * that case we let this loop run 1 into the prefix to see the
2450 * same slash.
2451 *
2452 * If there is no common prefix, we cannot do this as it would
2453 * underrun the input strings.
2454 */
2455 pfx_adjust_for_slash = (pfx_length ? 1 : 0);
2456 while (a + pfx_length - pfx_adjust_for_slash <= old_name &&
2457 b + pfx_length - pfx_adjust_for_slash <= new_name &&
2458 *old_name == *new_name) {
2459 if (*old_name == '/')
2460 sfx_length = len_a - (old_name - a);
2461 old_name--;
2462 new_name--;
2463 }
2464 /*
2466 * pfx{mid-a => mid-b}sfx
2467 * {pfx-a => pfx-b}sfx
2468 * pfx{sfx-a => sfx-b}
2469 * name-a => name-b
2470 */
2471 a_midlen = len_a - pfx_length - sfx_length;
2472 b_midlen = len_b - pfx_length - sfx_length;
2473 if (a_midlen < 0)
2474 a_midlen = 0;
2475 if (b_midlen < 0)
2476 b_midlen = 0;
2477 strbuf_grow(name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
2479 if (pfx_length + sfx_length) {
2480 strbuf_add(name, a, pfx_length);
2481 strbuf_addch(name, '{');
2482 }
2483 strbuf_add(name, a + pfx_length, a_midlen);
2484 strbuf_addstr(name, " => ");
2485 strbuf_add(name, b + pfx_length, b_midlen);
2486 if (pfx_length + sfx_length) {
2487 strbuf_addch(name, '}');
2488 strbuf_add(name, a + len_a - sfx_length, sfx_length);
2489 }
2490}
2491struct diffstat_t {
2493 int nr;
2494 int alloc;
2495 struct diffstat_file {
2496 char *from_name;
2497 char *name;
2498 char *print_name;
2499 const char *comments;
2500 unsigned is_unmerged:1;
2501 unsigned is_binary:1;
2502 unsigned is_renamed:1;
2503 unsigned is_interesting:1;
2504 uintmax_t added, deleted;
2505 } **files;
2506};
2507static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
2509 const char *name_a,
2510 const char *name_b)
2511{
2512 struct diffstat_file *x;
2513 x = xcalloc(1, sizeof(*x));
2514 ALLOC_GROW(diffstat->files, diffstat->nr + 1, diffstat->alloc);
2515 diffstat->files[diffstat->nr++] = x;
2516 if (name_b) {
2517 x->from_name = xstrdup(name_a);
2518 x->name = xstrdup(name_b);
2519 x->is_renamed = 1;
2520 }
2521 else {
2522 x->from_name = NULL;
2523 x->name = xstrdup(name_a);
2524 }
2525 return x;
2526}
2527static void diffstat_consume(void *priv, char *line, unsigned long len)
2529{
2530 struct diffstat_t *diffstat = priv;
2531 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
2532 if (line[0] == '+')
2534 x->added++;
2535 else if (line[0] == '-')
2536 x->deleted++;
2537}
2538const char mime_boundary_leader[] = "------------";
2540static int scale_linear(int it, int width, int max_change)
2542{
2543 if (!it)
2544 return 0;
2545 /*
2546 * make sure that at least one '-' or '+' is printed if
2547 * there is any change to this path. The easiest way is to
2548 * scale linearly as if the alloted width is one column shorter
2549 * than it is, and then add 1 to the result.
2550 */
2551 return 1 + (it * (width - 1) / max_change);
2552}
2553static void show_graph(struct strbuf *out, char ch, int cnt,
2555 const char *set, const char *reset)
2556{
2557 if (cnt <= 0)
2558 return;
2559 strbuf_addstr(out, set);
2560 strbuf_addchars(out, ch, cnt);
2561 strbuf_addstr(out, reset);
2562}
2563static void fill_print_name(struct diffstat_file *file)
2565{
2566 struct strbuf pname = STRBUF_INIT;
2567 if (file->print_name)
2569 return;
2570 if (file->is_renamed)
2572 pprint_rename(&pname, file->from_name, file->name);
2573 else
2574 quote_c_style(file->name, &pname, NULL, 0);
2575 if (file->comments)
2577 strbuf_addf(&pname, " (%s)", file->comments);
2578 file->print_name = strbuf_detach(&pname, NULL);
2580}
2581static void print_stat_summary_inserts_deletes(struct diff_options *options,
2583 int files, int insertions, int deletions)
2584{
2585 struct strbuf sb = STRBUF_INIT;
2586 if (!files) {
2588 assert(insertions == 0 && deletions == 0);
2589 emit_diff_symbol(options, DIFF_SYMBOL_STATS_SUMMARY_NO_FILES,
2590 NULL, 0, 0);
2591 return;
2592 }
2593 strbuf_addf(&sb,
2595 (files == 1) ? " %d file changed" : " %d files changed",
2596 files);
2597 /*
2599 * For binary diff, the caller may want to print "x files
2600 * changed" with insertions == 0 && deletions == 0.
2601 *
2602 * Not omitting "0 insertions(+), 0 deletions(-)" in this case
2603 * is probably less confusing (i.e skip over "2 files changed
2604 * but nothing about added/removed lines? Is this a bug in Git?").
2605 */
2606 if (insertions || deletions == 0) {
2607 strbuf_addf(&sb,
2608 (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
2609 insertions);
2610 }
2611 if (deletions || insertions == 0) {
2613 strbuf_addf(&sb,
2614 (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
2615 deletions);
2616 }
2617 strbuf_addch(&sb, '\n');
2618 emit_diff_symbol(options, DIFF_SYMBOL_STATS_SUMMARY_INSERTS_DELETES,
2619 sb.buf, sb.len, 0);
2620 strbuf_release(&sb);
2621}
2622void print_stat_summary(FILE *fp, int files,
2624 int insertions, int deletions)
2625{
2626 struct diff_options o;
2627 memset(&o, 0, sizeof(o));
2628 o.file = fp;
2629 print_stat_summary_inserts_deletes(&o, files, insertions, deletions);
2631}
2632static void show_stats(struct diffstat_t *data, struct diff_options *options)
2634{
2635 int i, len, add, del, adds = 0, dels = 0;
2636 uintmax_t max_change = 0, max_len = 0;
2637 int total_files = data->nr, count;
2638 int width, name_width, graph_width, number_width = 0, bin_width = 0;
2639 const char *reset, *add_c, *del_c;
2640 int extra_shown = 0;
2641 const char *line_prefix = diff_line_prefix(options);
2642 struct strbuf out = STRBUF_INIT;
2643 if (data->nr == 0)
2645 return;
2646 count = options->stat_count ? options->stat_count : data->nr;
2648 reset = diff_get_color_opt(options, DIFF_RESET);
2650 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
2651 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
2652 /*
2654 * Find the longest filename and max number of changes
2655 */
2656 for (i = 0; (i < count) && (i < data->nr); i++) {
2657 struct diffstat_file *file = data->files[i];
2658 uintmax_t change = file->added + file->deleted;
2659 if (!file->is_interesting && (change == 0)) {
2661 count++; /* not shown == room for one more */
2662 continue;
2663 }
2664 fill_print_name(file);
2665 len = strlen(file->print_name);
2666 if (max_len < len)
2667 max_len = len;
2668 if (file->is_unmerged) {
2670 /* "Unmerged" is 8 characters */
2671 bin_width = bin_width < 8 ? 8 : bin_width;
2672 continue;
2673 }
2674 if (file->is_binary) {
2675 /* "Bin XXX -> YYY bytes" */
2676 int w = 14 + decimal_width(file->added)
2677 + decimal_width(file->deleted);
2678 bin_width = bin_width < w ? w : bin_width;
2679 /* Display change counts aligned with "Bin" */
2680 number_width = 3;
2681 continue;
2682 }
2683 if (max_change < change)
2685 max_change = change;
2686 }
2687 count = i; /* where we can stop scanning in data->files[] */
2688 /*
2690 * We have width = stat_width or term_columns() columns total.
2691 * We want a maximum of min(max_len, stat_name_width) for the name part.
2692 * We want a maximum of min(max_change, stat_graph_width) for the +- part.
2693 * We also need 1 for " " and 4 + decimal_width(max_change)
2694 * for " | NNNN " and one the empty column at the end, altogether
2695 * 6 + decimal_width(max_change).
2696 *
2697 * If there's not enough space, we will use the smaller of
2698 * stat_name_width (if set) and 5/8*width for the filename,
2699 * and the rest for constant elements + graph part, but no more
2700 * than stat_graph_width for the graph part.
2701 * (5/8 gives 50 for filename and 30 for the constant parts + graph
2702 * for the standard terminal size).
2703 *
2704 * In other words: stat_width limits the maximum width, and
2705 * stat_name_width fixes the maximum width of the filename,
2706 * and is also used to divide available columns if there
2707 * aren't enough.
2708 *
2709 * Binary files are displayed with "Bin XXX -> YYY bytes"
2710 * instead of the change count and graph. This part is treated
2711 * similarly to the graph part, except that it is not
2712 * "scaled". If total width is too small to accommodate the
2713 * guaranteed minimum width of the filename part and the
2714 * separators and this message, this message will "overflow"
2715 * making the line longer than the maximum width.
2716 */
2717 if (options->stat_width == -1)
2719 width = term_columns() - strlen(line_prefix);
2720 else
2721 width = options->stat_width ? options->stat_width : 80;
2722 number_width = decimal_width(max_change) > number_width ?
2723 decimal_width(max_change) : number_width;
2724 if (options->stat_graph_width == -1)
2726 options->stat_graph_width = diff_stat_graph_width;
2727 /*
2729 * Guarantee 3/8*16==6 for the graph part
2730 * and 5/8*16==10 for the filename part
2731 */
2732 if (width < 16 + 6 + number_width)
2733 width = 16 + 6 + number_width;
2734 /*
2736 * First assign sizes that are wanted, ignoring available width.
2737 * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
2738 * starting from "XXX" should fit in graph_width.
2739 */
2740 graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
2741 if (options->stat_graph_width &&
2742 options->stat_graph_width < graph_width)
2743 graph_width = options->stat_graph_width;
2744 name_width = (options->stat_name_width > 0 &&
2746 options->stat_name_width < max_len) ?
2747 options->stat_name_width : max_len;
2748 /*
2750 * Adjust adjustable widths not to exceed maximum width
2751 */
2752 if (name_width + number_width + 6 + graph_width > width) {
2753 if (graph_width > width * 3/8 - number_width - 6) {
2754 graph_width = width * 3/8 - number_width - 6;
2755 if (graph_width < 6)
2756 graph_width = 6;
2757 }
2758 if (options->stat_graph_width &&
2760 graph_width > options->stat_graph_width)
2761 graph_width = options->stat_graph_width;
2762 if (name_width > width - number_width - 6 - graph_width)
2763 name_width = width - number_width - 6 - graph_width;
2764 else
2765 graph_width = width - number_width - 6 - name_width;
2766 }
2767 /*
2769 * From here name_width is the width of the name area,
2770 * and graph_width is the width of the graph area.
2771 * max_change is used to scale graph properly.
2772 */
2773 for (i = 0; i < count; i++) {
2774 const char *prefix = "";
2775 struct diffstat_file *file = data->files[i];
2776 char *name = file->print_name;
2777 uintmax_t added = file->added;
2778 uintmax_t deleted = file->deleted;
2779 int name_len;
2780 if (!file->is_interesting && (added + deleted == 0))
2782 continue;
2783 /*
2785 * "scale" the filename
2786 */
2787 len = name_width;
2788 name_len = strlen(name);
2789 if (name_width < name_len) {
2790 char *slash;
2791 prefix = "...";
2792 len -= 3;
2793 name += name_len - len;
2794 slash = strchr(name, '/');
2795 if (slash)
2796 name = slash;
2797 }
2798 if (file->is_binary) {
2800 strbuf_addf(&out, " %s%-*s |", prefix, len, name);
2801 strbuf_addf(&out, " %*s", number_width, "Bin");
2802 if (!added && !deleted) {
2803 strbuf_addch(&out, '\n');
2804 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2805 out.buf, out.len, 0);
2806 strbuf_reset(&out);
2807 continue;
2808 }
2809 strbuf_addf(&out, " %s%"PRIuMAX"%s",
2810 del_c, deleted, reset);
2811 strbuf_addstr(&out, " -> ");
2812 strbuf_addf(&out, "%s%"PRIuMAX"%s",
2813 add_c, added, reset);
2814 strbuf_addstr(&out, " bytes\n");
2815 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2816 out.buf, out.len, 0);
2817 strbuf_reset(&out);
2818 continue;
2819 }
2820 else if (file->is_unmerged) {
2821 strbuf_addf(&out, " %s%-*s |", prefix, len, name);
2822 strbuf_addstr(&out, " Unmerged\n");
2823 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2824 out.buf, out.len, 0);
2825 strbuf_reset(&out);
2826 continue;
2827 }
2828 /*
2830 * scale the add/delete
2831 */
2832 add = added;
2833 del = deleted;
2834 if (graph_width <= max_change) {
2836 int total = scale_linear(add + del, graph_width, max_change);
2837 if (total < 2 && add && del)
2838 /* width >= 2 due to the sanity check */
2839 total = 2;
2840 if (add < del) {
2841 add = scale_linear(add, graph_width, max_change);
2842 del = total - add;
2843 } else {
2844 del = scale_linear(del, graph_width, max_change);
2845 add = total - del;
2846 }
2847 }
2848 strbuf_addf(&out, " %s%-*s |", prefix, len, name);
2849 strbuf_addf(&out, " %*"PRIuMAX"%s",
2850 number_width, added + deleted,
2851 added + deleted ? " " : "");
2852 show_graph(&out, '+', add, add_c, reset);
2853 show_graph(&out, '-', del, del_c, reset);
2854 strbuf_addch(&out, '\n');
2855 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2856 out.buf, out.len, 0);
2857 strbuf_reset(&out);
2858 }
2859 for (i = 0; i < data->nr; i++) {
2861 struct diffstat_file *file = data->files[i];
2862 uintmax_t added = file->added;
2863 uintmax_t deleted = file->deleted;
2864 if (file->is_unmerged ||
2866 (!file->is_interesting && (added + deleted == 0))) {
2867 total_files--;
2868 continue;
2869 }
2870 if (!file->is_binary) {
2872 adds += added;
2873 dels += deleted;
2874 }
2875 if (i < count)
2876 continue;
2877 if (!extra_shown)
2878 emit_diff_symbol(options,
2879 DIFF_SYMBOL_STATS_SUMMARY_ABBREV,
2880 NULL, 0, 0);
2881 extra_shown = 1;
2882 }
2883 print_stat_summary_inserts_deletes(options, total_files, adds, dels);
2885 strbuf_release(&out);
2886}
2887static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
2889{
2890 int i, adds = 0, dels = 0, total_files = data->nr;
2891 if (data->nr == 0)
2893 return;
2894 for (i = 0; i < data->nr; i++) {
2896 int added = data->files[i]->added;
2897 int deleted = data->files[i]->deleted;
2898 if (data->files[i]->is_unmerged ||
2900 (!data->files[i]->is_interesting && (added + deleted == 0))) {
2901 total_files--;
2902 } else if (!data->files[i]->is_binary) { /* don't count bytes */
2903 adds += added;
2904 dels += deleted;
2905 }
2906 }
2907 print_stat_summary_inserts_deletes(options, total_files, adds, dels);
2908}
2909static void show_numstat(struct diffstat_t *data, struct diff_options *options)
2911{
2912 int i;
2913 if (data->nr == 0)
2915 return;
2916 for (i = 0; i < data->nr; i++) {
2918 struct diffstat_file *file = data->files[i];
2919 fprintf(options->file, "%s", diff_line_prefix(options));
2921 if (file->is_binary)
2923 fprintf(options->file, "-\t-\t");
2924 else
2925 fprintf(options->file,
2926 "%"PRIuMAX"\t%"PRIuMAX"\t",
2927 file->added, file->deleted);
2928 if (options->line_termination) {
2929 fill_print_name(file);
2930 if (!file->is_renamed)
2931 write_name_quoted(file->name, options->file,
2932 options->line_termination);
2933 else {
2934 fputs(file->print_name, options->file);
2935 putc(options->line_termination, options->file);
2936 }
2937 } else {
2938 if (file->is_renamed) {
2939 putc('\0', options->file);
2940 write_name_quoted(file->from_name, options->file, '\0');
2941 }
2942 write_name_quoted(file->name, options->file, '\0');
2943 }
2944 }
2945}
2946struct dirstat_file {
2948 const char *name;
2949 unsigned long changed;
2950};
2951struct dirstat_dir {
2953 struct dirstat_file *files;
2954 int alloc, nr, permille, cumulative;
2955};
2956static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
2958 unsigned long changed, const char *base, int baselen)
2959{
2960 unsigned long sum_changes = 0;
2961 unsigned int sources = 0;
2962 const char *line_prefix = diff_line_prefix(opt);
2963 while (dir->nr) {
2965 struct dirstat_file *f = dir->files;
2966 int namelen = strlen(f->name);
2967 unsigned long changes;
2968 char *slash;
2969 if (namelen < baselen)
2971 break;
2972 if (memcmp(f->name, base, baselen))
2973 break;
2974 slash = strchr(f->name + baselen, '/');
2975 if (slash) {
2976 int newbaselen = slash + 1 - f->name;
2977 changes = gather_dirstat(opt, dir, changed, f->name, newbaselen);
2978 sources++;
2979 } else {
2980 changes = f->changed;
2981 dir->files++;
2982 dir->nr--;
2983 sources += 2;
2984 }
2985 sum_changes += changes;
2986 }
2987 /*
2989 * We don't report dirstat's for
2990 * - the top level
2991 * - or cases where everything came from a single directory
2992 * under this directory (sources == 1).
2993 */
2994 if (baselen && sources != 1) {
2995 if (sum_changes) {
2996 int permille = sum_changes * 1000 / changed;
2997 if (permille >= dir->permille) {
2998 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
2999 permille / 10, permille % 10, baselen, base);
3000 if (!dir->cumulative)
3001 return 0;
3002 }
3003 }
3004 }
3005 return sum_changes;
3006}
3007static int dirstat_compare(const void *_a, const void *_b)
3009{
3010 const struct dirstat_file *a = _a;
3011 const struct dirstat_file *b = _b;
3012 return strcmp(a->name, b->name);
3013}
3014static void show_dirstat(struct diff_options *options)
3016{
3017 int i;
3018 unsigned long changed;
3019 struct dirstat_dir dir;
3020 struct diff_queue_struct *q = &diff_queued_diff;
3021 dir.files = NULL;
3023 dir.alloc = 0;
3024 dir.nr = 0;
3025 dir.permille = options->dirstat_permille;
3026 dir.cumulative = options->flags.dirstat_cumulative;
3027 changed = 0;
3029 for (i = 0; i < q->nr; i++) {
3030 struct diff_filepair *p = q->queue[i];
3031 const char *name;
3032 unsigned long copied, added, damage;
3033 name = p->two->path ? p->two->path : p->one->path;
3035 if (p->one->oid_valid && p->two->oid_valid &&
3037 oideq(&p->one->oid, &p->two->oid)) {
3038 /*
3039 * The SHA1 has not changed, so pre-/post-content is
3040 * identical. We can therefore skip looking at the
3041 * file contents altogether.
3042 */
3043 damage = 0;
3044 goto found_damage;
3045 }
3046 if (options->flags.dirstat_by_file) {
3048 /*
3049 * In --dirstat-by-file mode, we don't really need to
3050 * look at the actual file contents at all.
3051 * The fact that the SHA1 changed is enough for us to
3052 * add this file to the list of results
3053 * (with each file contributing equal damage).
3054 */
3055 damage = 1;
3056 goto found_damage;
3057 }
3058 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
3060 diff_populate_filespec(options->repo, p->one, 0);
3061 diff_populate_filespec(options->repo, p->two, 0);
3062 diffcore_count_changes(options->repo,
3063 p->one, p->two, NULL, NULL,
3064 &copied, &added);
3065 diff_free_filespec_data(p->one);
3066 diff_free_filespec_data(p->two);
3067 } else if (DIFF_FILE_VALID(p->one)) {
3068 diff_populate_filespec(options->repo, p->one, CHECK_SIZE_ONLY);
3069 copied = added = 0;
3070 diff_free_filespec_data(p->one);
3071 } else if (DIFF_FILE_VALID(p->two)) {
3072 diff_populate_filespec(options->repo, p->two, CHECK_SIZE_ONLY);
3073 copied = 0;
3074 added = p->two->size;
3075 diff_free_filespec_data(p->two);
3076 } else
3077 continue;
3078 /*
3080 * Original minus copied is the removed material,
3081 * added is the new material. They are both damages
3082 * made to the preimage.
3083 * If the resulting damage is zero, we know that
3084 * diffcore_count_changes() considers the two entries to
3085 * be identical, but since the oid changed, we
3086 * know that there must have been _some_ kind of change,
3087 * so we force all entries to have damage > 0.
3088 */
3089 damage = (p->one->size - copied) + added;
3090 if (!damage)
3091 damage = 1;
3092found_damage:
3094 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
3095 dir.files[dir.nr].name = name;
3096 dir.files[dir.nr].changed = damage;
3097 changed += damage;
3098 dir.nr++;
3099 }
3100 /* This can happen even with many files, if everything was renames */
3102 if (!changed)
3103 return;
3104 /* Show all directories with more than x% of the changes */
3106 QSORT(dir.files, dir.nr, dirstat_compare);
3107 gather_dirstat(options, &dir, changed, "", 0);
3108}
3109static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
3111{
3112 int i;
3113 unsigned long changed;
3114 struct dirstat_dir dir;
3115 if (data->nr == 0)
3117 return;
3118 dir.files = NULL;
3120 dir.alloc = 0;
3121 dir.nr = 0;
3122 dir.permille = options->dirstat_permille;
3123 dir.cumulative = options->flags.dirstat_cumulative;
3124 changed = 0;
3126 for (i = 0; i < data->nr; i++) {
3127 struct diffstat_file *file = data->files[i];
3128 unsigned long damage = file->added + file->deleted;
3129 if (file->is_binary)
3130 /*
3131 * binary files counts bytes, not lines. Must find some
3132 * way to normalize binary bytes vs. textual lines.
3133 * The following heuristic assumes that there are 64
3134 * bytes per "line".
3135 * This is stupid and ugly, but very cheap...
3136 */
3137 damage = DIV_ROUND_UP(damage, 64);
3138 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
3139 dir.files[dir.nr].name = file->name;
3140 dir.files[dir.nr].changed = damage;
3141 changed += damage;
3142 dir.nr++;
3143 }
3144 /* This can happen even with many files, if everything was renames */
3146 if (!changed)
3147 return;
3148 /* Show all directories with more than x% of the changes */
3150 QSORT(dir.files, dir.nr, dirstat_compare);
3151 gather_dirstat(options, &dir, changed, "", 0);
3152}
3153static void free_diffstat_info(struct diffstat_t *diffstat)
3155{
3156 int i;
3157 for (i = 0; i < diffstat->nr; i++) {
3158 struct diffstat_file *f = diffstat->files[i];
3159 free(f->print_name);
3160 free(f->name);
3161 free(f->from_name);
3162 free(f);
3163 }
3164 free(diffstat->files);
3165}
3166struct checkdiff_t {
3168 const char *filename;
3169 int lineno;
3170 int conflict_marker_size;
3171 struct diff_options *o;
3172 unsigned ws_rule;
3173 unsigned status;
3174};
3175static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
3177{
3178 char firstchar;
3179 int cnt;
3180 if (len < marker_size + 1)
3182 return 0;
3183 firstchar = line[0];
3184 switch (firstchar) {
3185 case '=': case '>': case '<': case '|':
3186 break;
3187 default:
3188 return 0;
3189 }
3190 for (cnt = 1; cnt < marker_size; cnt++)
3191 if (line[cnt] != firstchar)
3192 return 0;
3193 /* line[1] thru line[marker_size-1] are same as firstchar */
3194 if (len < marker_size + 1 || !isspace(line[marker_size]))
3195 return 0;
3196 return 1;
3197}
3198static void checkdiff_consume_hunk(void *priv,
3200 long ob, long on, long nb, long nn,
3201 const char *func, long funclen)
3202{
3204 struct checkdiff_t *data = priv;
3205 data->lineno = nb - 1;
3206}
3207static void checkdiff_consume(void *priv, char *line, unsigned long len)
3209{
3210 struct checkdiff_t *data = priv;
3211 int marker_size = data->conflict_marker_size;
3212 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
3213 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
3214 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
3215 char *err;
3216 const char *line_prefix;
3217 assert(data->o);
3219 line_prefix = diff_line_prefix(data->o);
3220 if (line[0] == '+') {
3222 unsigned bad;
3223 data->lineno++;
3224 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
3225 data->status |= 1;
3226 fprintf(data->o->file,
3227 "%s%s:%d: leftover conflict marker\n",
3228 line_prefix, data->filename, data->lineno);
3229 }
3230 bad = ws_check(line + 1, len - 1, data->ws_rule);
3231 if (!bad)
3232 return;
3233 data->status |= bad;
3234 err = whitespace_error_string(bad);
3235 fprintf(data->o->file, "%s%s:%d: %s.\n",
3236 line_prefix, data->filename, data->lineno, err);
3237 free(err);
3238 emit_line(data->o, set, reset, line, 1);
3239 ws_check_emit(line + 1, len - 1, data->ws_rule,
3240 data->o->file, set, reset, ws);
3241 } else if (line[0] == ' ') {
3242 data->lineno++;
3243 }
3244}
3245static unsigned char *deflate_it(char *data,
3247 unsigned long size,
3248 unsigned long *result_size)
3249{
3250 int bound;
3251 unsigned char *deflated;
3252 git_zstream stream;
3253 git_deflate_init(&stream, zlib_compression_level);
3255 bound = git_deflate_bound(&stream, size);
3256 deflated = xmalloc(bound);
3257 stream.next_out = deflated;
3258 stream.avail_out = bound;
3259 stream.next_in = (unsigned char *)data;
3261 stream.avail_in = size;
3262 while (git_deflate(&stream, Z_FINISH) == Z_OK)
3263 ; /* nothing */
3264 git_deflate_end(&stream);
3265 *result_size = stream.total_out;
3266 return deflated;
3267}
3268static void emit_binary_diff_body(struct diff_options *o,
3270 mmfile_t *one, mmfile_t *two)
3271{
3272 void *cp;
3273 void *delta;
3274 void *deflated;
3275 void *data;
3276 unsigned long orig_size;
3277 unsigned long delta_size;
3278 unsigned long deflate_size;
3279 unsigned long data_size;
3280 /* We could do deflated delta, or we could do just deflated two,
3282 * whichever is smaller.
3283 */
3284 delta = NULL;
3285 deflated = deflate_it(two->ptr, two->size, &deflate_size);
3286 if (one->size && two->size) {
3287 delta = diff_delta(one->ptr, one->size,
3288 two->ptr, two->size,
3289 &delta_size, deflate_size);
3290 if (delta) {
3291 void *to_free = delta;
3292 orig_size = delta_size;
3293 delta = deflate_it(delta, delta_size, &delta_size);
3294 free(to_free);
3295 }
3296 }
3297 if (delta && delta_size < deflate_size) {
3299 char *s = xstrfmt("%"PRIuMAX , (uintmax_t)orig_size);
3300 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_HEADER_DELTA,
3301 s, strlen(s), 0);
3302 free(s);
3303 free(deflated);
3304 data = delta;
3305 data_size = delta_size;
3306 } else {
3307 char *s = xstrfmt("%lu", two->size);
3308 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_HEADER_LITERAL,
3309 s, strlen(s), 0);
3310 free(s);
3311 free(delta);
3312 data = deflated;
3313 data_size = deflate_size;
3314 }
3315 /* emit data encoded in base85 */
3317 cp = data;
3318 while (data_size) {
3319 int len;
3320 int bytes = (52 < data_size) ? 52 : data_size;
3321 char line[71];
3322 data_size -= bytes;
3323 if (bytes <= 26)
3324 line[0] = bytes + 'A' - 1;
3325 else
3326 line[0] = bytes - 26 + 'a' - 1;
3327 encode_85(line + 1, cp, bytes);
3328 cp = (char *) cp + bytes;
3329 len = strlen(line);
3331 line[len++] = '\n';
3332 line[len] = '\0';
3333 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_BODY,
3335 line, len, 0);
3336 }
3337 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_FOOTER, NULL, 0, 0);
3338 free(data);
3339}
3340static void emit_binary_diff(struct diff_options *o,
3342 mmfile_t *one, mmfile_t *two)
3343{
3344 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_HEADER, NULL, 0, 0);
3345 emit_binary_diff_body(o, one, two);
3346 emit_binary_diff_body(o, two, one);
3347}
3348int diff_filespec_is_binary(struct repository *r,
3350 struct diff_filespec *one)
3351{
3352 if (one->is_binary == -1) {
3353 diff_filespec_load_driver(one, r->index);
3354 if (one->driver->binary != -1)
3355 one->is_binary = one->driver->binary;
3356 else {
3357 if (!one->data && DIFF_FILE_VALID(one))
3358 diff_populate_filespec(r, one, CHECK_BINARY);
3359 if (one->is_binary == -1 && one->data)
3360 one->is_binary = buffer_is_binary(one->data,
3361 one->size);
3362 if (one->is_binary == -1)
3363 one->is_binary = 0;
3364 }
3365 }
3366 return one->is_binary;
3367}
3368static const struct userdiff_funcname *
3370diff_funcname_pattern(struct diff_options *o, struct diff_filespec *one)
3371{
3372 diff_filespec_load_driver(one, o->repo->index);
3373 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
3374}
3375void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
3377{
3378 if (!options->a_prefix)
3379 options->a_prefix = a;
3380 if (!options->b_prefix)
3381 options->b_prefix = b;
3382}
3383struct userdiff_driver *get_textconv(struct repository *r,
3385 struct diff_filespec *one)
3386{
3387 if (!DIFF_FILE_VALID(one))
3388 return NULL;
3389 diff_filespec_load_driver(one, r->index);
3391 return userdiff_get_textconv(r, one->driver);
3392}
3393static void builtin_diff(const char *name_a,
3395 const char *name_b,
3396 struct diff_filespec *one,
3397 struct diff_filespec *two,
3398 const char *xfrm_msg,
3399 int must_show_header,
3400 struct diff_options *o,
3401 int complete_rewrite)
3402{
3403 mmfile_t mf1, mf2;
3404 const char *lbl[2];
3405 char *a_one, *b_two;
3406 const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
3407 const char *reset = diff_get_color_opt(o, DIFF_RESET);
3408 const char *a_prefix, *b_prefix;
3409 struct userdiff_driver *textconv_one = NULL;
3410 struct userdiff_driver *textconv_two = NULL;
3411 struct strbuf header = STRBUF_INIT;
3412 const char *line_prefix = diff_line_prefix(o);
3413 diff_set_mnemonic_prefix(o, "a/", "b/");
3415 if (o->flags.reverse_diff) {
3416 a_prefix = o->b_prefix;
3417 b_prefix = o->a_prefix;
3418 } else {
3419 a_prefix = o->a_prefix;
3420 b_prefix = o->b_prefix;
3421 }
3422 if (o->submodule_format == DIFF_SUBMODULE_LOG &&
3424 (!one->mode || S_ISGITLINK(one->mode)) &&
3425 (!two->mode || S_ISGITLINK(two->mode))) {
3426 show_submodule_summary(o, one->path ? one->path : two->path,
3427 &one->oid, &two->oid,
3428 two->dirty_submodule);
3429 return;
3430 } else if (o->submodule_format == DIFF_SUBMODULE_INLINE_DIFF &&
3431 (!one->mode || S_ISGITLINK(one->mode)) &&
3432 (!two->mode || S_ISGITLINK(two->mode))) {
3433 show_submodule_inline_diff(o, one->path ? one->path : two->path,
3434 &one->oid, &two->oid,
3435 two->dirty_submodule);
3436 return;
3437 }
3438 if (o->flags.allow_textconv) {
3440 textconv_one = get_textconv(o->repo, one);
3441 textconv_two = get_textconv(o->repo, two);
3442 }
3443 /* Never use a non-valid filename anywhere if at all possible */
3445 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
3446 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
3447 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
3449 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
3450 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
3451 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
3452 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
3453 if (lbl[0][0] == '/') {
3454 /* /dev/null */
3455 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
3456 if (xfrm_msg)
3457 strbuf_addstr(&header, xfrm_msg);
3458 must_show_header = 1;
3459 }
3460 else if (lbl[1][0] == '/') {
3461 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
3462 if (xfrm_msg)
3463 strbuf_addstr(&header, xfrm_msg);
3464 must_show_header = 1;
3465 }
3466 else {
3467 if (one->mode != two->mode) {
3468 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
3469 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
3470 must_show_header = 1;
3471 }
3472 if (xfrm_msg)
3473 strbuf_addstr(&header, xfrm_msg);
3474 /*
3476 * we do not run diff between different kind
3477 * of objects.
3478 */
3479 if ((one->mode ^ two->mode) & S_IFMT)
3480 goto free_ab_and_return;
3481 if (complete_rewrite &&
3482 (textconv_one || !diff_filespec_is_binary(o->repo, one)) &&
3483 (textconv_two || !diff_filespec_is_binary(o->repo, two))) {
3484 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3485 header.buf, header.len, 0);
3486 strbuf_reset(&header);
3487 emit_rewrite_diff(name_a, name_b, one, two,
3488 textconv_one, textconv_two, o);
3489 o->found_changes = 1;
3490 goto free_ab_and_return;
3491 }
3492 }
3493 if (o->irreversible_delete && lbl[1][0] == '/') {
3495 emit_diff_symbol(o, DIFF_SYMBOL_HEADER, header.buf,
3496 header.len, 0);
3497 strbuf_reset(&header);
3498 goto free_ab_and_return;
3499 } else if (!o->flags.text &&
3500 ( (!textconv_one && diff_filespec_is_binary(o->repo, one)) ||
3501 (!textconv_two && diff_filespec_is_binary(o->repo, two)) )) {
3502 struct strbuf sb = STRBUF_INIT;
3503 if (!one->data && !two->data &&
3504 S_ISREG(one->mode) && S_ISREG(two->mode) &&
3505 !o->flags.binary) {
3506 if (oideq(&one->oid, &two->oid)) {
3507 if (must_show_header)
3508 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3509 header.buf, header.len,
3510 0);
3511 goto free_ab_and_return;
3512 }
3513 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3514 header.buf, header.len, 0);
3515 strbuf_addf(&sb, "%sBinary files %s and %s differ\n",
3516 diff_line_prefix(o), lbl[0], lbl[1]);
3517 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_FILES,
3518 sb.buf, sb.len, 0);
3519 strbuf_release(&sb);
3520 goto free_ab_and_return;
3521 }
3522 if (fill_mmfile(o->repo, &mf1, one) < 0 ||
3523 fill_mmfile(o->repo, &mf2, two) < 0)
3524 die("unable to read files to diff");
3525 /* Quite common confusing case */
3526 if (mf1.size == mf2.size &&
3527 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
3528 if (must_show_header)
3529 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3530 header.buf, header.len, 0);
3531 goto free_ab_and_return;
3532 }
3533 emit_diff_symbol(o, DIFF_SYMBOL_HEADER, header.buf, header.len, 0);
3534 strbuf_reset(&header);
3535 if (o->flags.binary)
3536 emit_binary_diff(o, &mf1, &mf2);
3537 else {
3538 strbuf_addf(&sb, "%sBinary files %s and %s differ\n",
3539 diff_line_prefix(o), lbl[0], lbl[1]);
3540 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_FILES,
3541 sb.buf, sb.len, 0);
3542 strbuf_release(&sb);
3543 }
3544 o->found_changes = 1;
3545 } else {
3546 /* Crazy xdl interfaces.. */
3547 const char *diffopts;
3548 const char *v;
3549 xpparam_t xpp;
3550 xdemitconf_t xecfg;
3551 struct emit_callback ecbdata;
3552 const struct userdiff_funcname *pe;
3553 if (must_show_header) {
3555 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3556 header.buf, header.len, 0);
3557 strbuf_reset(&header);
3558 }
3559 mf1.size = fill_textconv(o->repo, textconv_one, one, &mf1.ptr);
3561 mf2.size = fill_textconv(o->repo, textconv_two, two, &mf2.ptr);
3562 pe = diff_funcname_pattern(o, one);
3564 if (!pe)
3565 pe = diff_funcname_pattern(o, two);
3566 memset(&xpp, 0, sizeof(xpp));
3568 memset(&xecfg, 0, sizeof(xecfg));
3569 memset(&ecbdata, 0, sizeof(ecbdata));
3570 if (o->flags.suppress_diff_headers)
3571 lbl[0] = NULL;
3572 ecbdata.label_path = lbl;
3573 ecbdata.color_diff = want_color(o->use_color);
3574 ecbdata.ws_rule = whitespace_rule(o->repo->index, name_b);
3575 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
3576 check_blank_at_eof(&mf1, &mf2, &ecbdata);
3577 ecbdata.opt = o;
3578 if (header.len && !o->flags.suppress_diff_headers)
3579 ecbdata.header = &header;
3580 xpp.flags = o->xdl_opts;
3581 xpp.anchors = o->anchors;
3582 xpp.anchors_nr = o->anchors_nr;
3583 xecfg.ctxlen = o->context;
3584 xecfg.interhunkctxlen = o->interhunkcontext;
3585 xecfg.flags = XDL_EMIT_FUNCNAMES;
3586 if (o->flags.funccontext)
3587 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
3588 if (pe)
3589 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
3590 diffopts = getenv("GIT_DIFF_OPTS");
3592 if (!diffopts)
3593 ;
3594 else if (skip_prefix(diffopts, "--unified=", &v))
3595 xecfg.ctxlen = strtoul(v, NULL, 10);
3596 else if (skip_prefix(diffopts, "-u", &v))
3597 xecfg.ctxlen = strtoul(v, NULL, 10);
3598 if (o->word_diff)
3600 init_diff_words_data(&ecbdata, o, one, two);
3601 if (xdi_diff_outf(&mf1, &mf2, NULL, fn_out_consume,
3602 &ecbdata, &xpp, &xecfg))
3603 die("unable to generate diff for %s", one->path);
3604 if (o->word_diff)
3605 free_diff_words_data(&ecbdata);
3606 if (textconv_one)
3607 free(mf1.ptr);
3608 if (textconv_two)
3609 free(mf2.ptr);
3610 xdiff_clear_find_func(&xecfg);
3611 }
3612 free_ab_and_return:
3614 strbuf_release(&header);
3615 diff_free_filespec_data(one);
3616 diff_free_filespec_data(two);
3617 free(a_one);
3618 free(b_two);
3619 return;
3620}
3621static char *get_compact_summary(const struct diff_filepair *p, int is_renamed)
3623{
3624 if (!is_renamed) {
3625 if (p->status == DIFF_STATUS_ADDED) {
3626 if (S_ISLNK(p->two->mode))
3627 return "new +l";
3628 else if ((p->two->mode & 0777) == 0755)
3629 return "new +x";
3630 else
3631 return "new";
3632 } else if (p->status == DIFF_STATUS_DELETED)
3633 return "gone";
3634 }
3635 if (S_ISLNK(p->one->mode) && !S_ISLNK(p->two->mode))
3636 return "mode -l";
3637 else if (!S_ISLNK(p->one->mode) && S_ISLNK(p->two->mode))
3638 return "mode +l";
3639 else if ((p->one->mode & 0777) == 0644 &&
3640 (p->two->mode & 0777) == 0755)
3641 return "mode +x";
3642 else if ((p->one->mode & 0777) == 0755 &&
3643 (p->two->mode & 0777) == 0644)
3644 return "mode -x";
3645 return NULL;
3646}
3647static void builtin_diffstat(const char *name_a, const char *name_b,
3649 struct diff_filespec *one,
3650 struct diff_filespec *two,
3651 struct diffstat_t *diffstat,
3652 struct diff_options *o,
3653 struct diff_filepair *p)
3654{
3655 mmfile_t mf1, mf2;
3656 struct diffstat_file *data;
3657 int same_contents;
3658 int complete_rewrite = 0;
3659 if (!DIFF_PAIR_UNMERGED(p)) {
3661 if (p->status == DIFF_STATUS_MODIFIED && p->score)
3662 complete_rewrite = 1;
3663 }
3664 data = diffstat_add(diffstat, name_a, name_b);
3666 data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
3667 if (o->flags.stat_with_summary)
3668 data->comments = get_compact_summary(p, data->is_renamed);
3669 if (!one || !two) {
3671 data->is_unmerged = 1;
3672 return;
3673 }
3674 same_contents = oideq(&one->oid, &two->oid);
3676 if (diff_filespec_is_binary(o->repo, one) ||
3678 diff_filespec_is_binary(o->repo, two)) {
3679 data->is_binary = 1;
3680 if (same_contents) {
3681 data->added = 0;
3682 data->deleted = 0;
3683 } else {
3684 data->added = diff_filespec_size(o->repo, two);
3685 data->deleted = diff_filespec_size(o->repo, one);
3686 }
3687 }
3688 else if (complete_rewrite) {
3690 diff_populate_filespec(o->repo, one, 0);
3691 diff_populate_filespec(o->repo, two, 0);
3692 data->deleted = count_lines(one->data, one->size);
3693 data->added = count_lines(two->data, two->size);
3694 }
3695 else if (!same_contents) {
3697 /* Crazy xdl interfaces.. */
3698 xpparam_t xpp;
3699 xdemitconf_t xecfg;
3700 if (fill_mmfile(o->repo, &mf1, one) < 0 ||
3702 fill_mmfile(o->repo, &mf2, two) < 0)
3703 die("unable to read files to diff");
3704 memset(&xpp, 0, sizeof(xpp));
3706 memset(&xecfg, 0, sizeof(xecfg));
3707 xpp.flags = o->xdl_opts;
3708 xpp.anchors = o->anchors;
3709 xpp.anchors_nr = o->anchors_nr;
3710 xecfg.ctxlen = o->context;
3711 xecfg.interhunkctxlen = o->interhunkcontext;
3712 if (xdi_diff_outf(&mf1, &mf2, discard_hunk_line,
3713 diffstat_consume, diffstat, &xpp, &xecfg))
3714 die("unable to generate diffstat for %s", one->path);
3715 }
3716 diff_free_filespec_data(one);
3718 diff_free_filespec_data(two);
3719}
3720static void builtin_checkdiff(const char *name_a, const char *name_b,
3722 const char *attr_path,
3723 struct diff_filespec *one,
3724 struct diff_filespec *two,
3725 struct diff_options *o)
3726{
3727 mmfile_t mf1, mf2;
3728 struct checkdiff_t data;
3729 if (!two)
3731 return;
3732 memset(&data, 0, sizeof(data));
3734 data.filename = name_b ? name_b : name_a;
3735 data.lineno = 0;
3736 data.o = o;
3737 data.ws_rule = whitespace_rule(o->repo->index, attr_path);
3738 data.conflict_marker_size = ll_merge_marker_size(o->repo->index, attr_path);
3739 if (fill_mmfile(o->repo, &mf1, one) < 0 ||
3741 fill_mmfile(o->repo, &mf2, two) < 0)
3742 die("unable to read files to diff");
3743 /*
3745 * All the other codepaths check both sides, but not checking
3746 * the "old" side here is deliberate. We are checking the newly
3747 * introduced changes, and as long as the "new" side is text, we
3748 * can and should check what it introduces.
3749 */
3750 if (diff_filespec_is_binary(o->repo, two))
3751 goto free_and_return;
3752 else {
3753 /* Crazy xdl interfaces.. */
3754 xpparam_t xpp;
3755 xdemitconf_t xecfg;
3756 memset(&xpp, 0, sizeof(xpp));
3758 memset(&xecfg, 0, sizeof(xecfg));
3759 xecfg.ctxlen = 1; /* at least one context line */
3760 xpp.flags = 0;
3761 if (xdi_diff_outf(&mf1, &mf2, checkdiff_consume_hunk,
3762 checkdiff_consume, &data,
3763 &xpp, &xecfg))
3764 die("unable to generate checkdiff for %s", one->path);
3765 if (data.ws_rule & WS_BLANK_AT_EOF) {
3767 struct emit_callback ecbdata;
3768 int blank_at_eof;
3769 ecbdata.ws_rule = data.ws_rule;
3771 check_blank_at_eof(&mf1, &mf2, &ecbdata);
3772 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
3773 if (blank_at_eof) {
3775 static char *err;
3776 if (!err)
3777 err = whitespace_error_string(WS_BLANK_AT_EOF);
3778 fprintf(o->file, "%s:%d: %s.\n",
3779 data.filename, blank_at_eof, err);
3780 data.status = 1; /* report errors */
3781 }
3782 }
3783 }
3784 free_and_return:
3785 diff_free_filespec_data(one);
3786 diff_free_filespec_data(two);
3787 if (data.status)
3788 o->flags.check_failed = 1;
3789}
3790struct diff_filespec *alloc_filespec(const char *path)
3792{
3793 struct diff_filespec *spec;
3794 FLEXPTR_ALLOC_STR(spec, path, path);
3796 spec->count = 1;
3797 spec->is_binary = -1;
3798 return spec;
3799}
3800void free_filespec(struct diff_filespec *spec)
3802{
3803 if (!--spec->count) {
3804 diff_free_filespec_data(spec);
3805 free(spec);
3806 }
3807}
3808void fill_filespec(struct diff_filespec *spec, const struct object_id *oid,
3810 int oid_valid, unsigned short mode)
3811{
3812 if (mode) {
3813 spec->mode = canon_mode(mode);
3814 oidcpy(&spec->oid, oid);
3815 spec->oid_valid = oid_valid;
3816 }
3817}
3818/*
3820 * Given a name and sha1 pair, if the index tells us the file in
3821 * the work tree has that object contents, return true, so that
3822 * prepare_temp_file() does not have to inflate and extract.
3823 */
3824static int reuse_worktree_file(struct index_state *istate,
3825 const char *name,
3826 const struct object_id *oid,
3827 int want_file)
3828{
3829 const struct cache_entry *ce;
3830 struct stat st;
3831 int pos, len;
3832 /*
3834 * We do not read the cache ourselves here, because the
3835 * benchmark with my previous version that always reads cache
3836 * shows that it makes things worse for diff-tree comparing
3837 * two linux-2.6 kernel trees in an already checked out work
3838 * tree. This is because most diff-tree comparisons deal with
3839 * only a small number of files, while reading the cache is
3840 * expensive for a large project, and its cost outweighs the
3841 * savings we get by not inflating the object to a temporary
3842 * file. Practically, this code only helps when we are used
3843 * by diff-cache --cached, which does read the cache before
3844 * calling us.
3845 */
3846 if (!istate->cache)
3847 return 0;
3848 /* We want to avoid the working directory if our caller
3850 * doesn't need the data in a normal file, this system
3851 * is rather slow with its stat/open/mmap/close syscalls,
3852 * and the object is contained in a pack file. The pack
3853 * is probably already open and will be faster to obtain
3854 * the data through than the working directory. Loose
3855 * objects however would tend to be slower as they need
3856 * to be individually opened and inflated.
3857 */
3858 if (!FAST_WORKING_DIRECTORY && !want_file && has_object_pack(oid))
3859 return 0;
3860 /*
3862 * Similarly, if we'd have to convert the file contents anyway, that
3863 * makes the optimization not worthwhile.
3864 */
3865 if (!want_file && would_convert_to_git(istate, name))
3866 return 0;
3867 len = strlen(name);
3869 pos = index_name_pos(istate, name, len);
3870 if (pos < 0)
3871 return 0;
3872 ce = istate->cache[pos];
3873 /*
3875 * This is not the sha1 we are looking for, or
3876 * unreusable because it is not a regular file.
3877 */
3878 if (!oideq(oid, &ce->oid) || !S_ISREG(ce->ce_mode))
3879 return 0;
3880 /*
3882 * If ce is marked as "assume unchanged", there is no
3883 * guarantee that work tree matches what we are looking for.
3884 */
3885 if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
3886 return 0;
3887 /*
3889 * If ce matches the file in the work tree, we can reuse it.
3890 */
3891 if (ce_uptodate(ce) ||
3892 (!lstat(name, &st) && !ie_match_stat(istate, ce, &st, 0)))
3893 return 1;
3894 return 0;
3896}
3897static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
3899{
3900 struct strbuf buf = STRBUF_INIT;
3901 char *dirty = "";
3902 /* Are we looking at the work tree? */
3904 if (s->dirty_submodule)
3905 dirty = "-dirty";
3906 strbuf_addf(&buf, "Subproject commit %s%s\n",
3908 oid_to_hex(&s->oid), dirty);
3909 s->size = buf.len;
3910 if (size_only) {
3911 s->data = NULL;
3912 strbuf_release(&buf);
3913 } else {
3914 s->data = strbuf_detach(&buf, NULL);
3915 s->should_free = 1;
3916 }
3917 return 0;
3918}
3919/*
3921 * While doing rename detection and pickaxe operation, we may need to
3922 * grab the data for the blob (or file) for our own in-core comparison.
3923 * diff_filespec has data and size fields for this purpose.
3924 */
3925int diff_populate_filespec(struct repository *r,
3926 struct diff_filespec *s,
3927 unsigned int flags)
3928{
3929 int size_only = flags & CHECK_SIZE_ONLY;
3930 int err = 0;
3931 int conv_flags = global_conv_flags_eol;
3932 /*
3933 * demote FAIL to WARN to allow inspecting the situation
3934 * instead of refusing.
3935 */
3936 if (conv_flags & CONV_EOL_RNDTRP_DIE)
3937 conv_flags = CONV_EOL_RNDTRP_WARN;
3938 if (!DIFF_FILE_VALID(s))
3940 die("internal error: asking to populate invalid file.");
3941 if (S_ISDIR(s->mode))
3942 return -1;
3943 if (s->data)
3945 return 0;
3946 if (size_only && 0 < s->size)
3948 return 0;
3949 if (S_ISGITLINK(s->mode))
3951 return diff_populate_gitlink(s, size_only);
3952 if (!s->oid_valid ||
3954 reuse_worktree_file(r->index, s->path, &s->oid, 0)) {
3955 struct strbuf buf = STRBUF_INIT;
3956 struct stat st;
3957 int fd;
3958 if (lstat(s->path, &st) < 0) {
3960 err_empty:
3961 err = -1;
3962 empty:
3963 s->data = (char *)"";
3964 s->size = 0;
3965 return err;
3966 }
3967 s->size = xsize_t(st.st_size);
3968 if (!s->size)
3969 goto empty;
3970 if (S_ISLNK(st.st_mode)) {
3971 struct strbuf sb = STRBUF_INIT;
3972 if (strbuf_readlink(&sb, s->path, s->size))
3974 goto err_empty;
3975 s->size = sb.len;
3976 s->data = strbuf_detach(&sb, NULL);
3977 s->should_free = 1;
3978 return 0;
3979 }
3980 /*
3982 * Even if the caller would be happy with getting
3983 * only the size, we cannot return early at this
3984 * point if the path requires us to run the content
3985 * conversion.
3986 */
3987 if (size_only && !would_convert_to_git(r->index, s->path))
3988 return 0;
3989 /*
3991 * Note: this check uses xsize_t(st.st_size) that may
3992 * not be the true size of the blob after it goes
3993 * through convert_to_git(). This may not strictly be
3994 * correct, but the whole point of big_file_threshold
3995 * and is_binary check being that we want to avoid
3996 * opening the file and inspecting the contents, this
3997 * is probably fine.
3998 */
3999 if ((flags & CHECK_BINARY) &&
4000 s->size > big_file_threshold && s->is_binary == -1) {
4001 s->is_binary = 1;
4002 return 0;
4003 }
4004 fd = open(s->path, O_RDONLY);
4005 if (fd < 0)
4006 goto err_empty;
4007 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
4008 close(fd);
4009 s->should_munmap = 1;
4010 /*
4012 * Convert from working tree format to canonical git format
4013 */
4014 if (convert_to_git(r->index, s->path, s->data, s->size, &buf, conv_flags)) {
4015 size_t size = 0;
4016 munmap(s->data, s->size);
4017 s->should_munmap = 0;
4018 s->data = strbuf_detach(&buf, &size);
4019 s->size = size;
4020 s->should_free = 1;
4021 }
4022 }
4023 else {
4024 enum object_type type;
4025 if (size_only || (flags & CHECK_BINARY)) {
4026 type = oid_object_info(r, &s->oid, &s->size);
4027 if (type < 0)
4028 die("unable to read %s",
4029 oid_to_hex(&s->oid));
4030 if (size_only)
4031 return 0;
4032 if (s->size > big_file_threshold && s->is_binary == -1) {
4033 s->is_binary = 1;
4034 return 0;
4035 }
4036 }
4037 s->data = read_object_file(&s->oid, &type, &s->size);
4038 if (!s->data)
4039 die("unable to read %s", oid_to_hex(&s->oid));
4040 s->should_free = 1;
4041 }
4042 return 0;
4043}
4044void diff_free_filespec_blob(struct diff_filespec *s)
4046{
4047 if (s->should_free)
4048 free(s->data);
4049 else if (s->should_munmap)
4050 munmap(s->data, s->size);
4051 if (s->should_free || s->should_munmap) {
4053 s->should_free = s->should_munmap = 0;
4054 s->data = NULL;
4055 }
4056}
4057void diff_free_filespec_data(struct diff_filespec *s)
4059{
4060 diff_free_filespec_blob(s);
4061 FREE_AND_NULL(s->cnt_data);
4062}
4063static void prep_temp_blob(struct index_state *istate,
4065 const char *path, struct diff_tempfile *temp,
4066 void *blob,
4067 unsigned long size,
4068 const struct object_id *oid,
4069 int mode)
4070{
4071 struct strbuf buf = STRBUF_INIT;
4072 struct strbuf tempfile = STRBUF_INIT;
4073 char *path_dup = xstrdup(path);
4074 const char *base = basename(path_dup);
4075 /* Generate "XXXXXX_basename.ext" */
4077 strbuf_addstr(&tempfile, "XXXXXX_");
4078 strbuf_addstr(&tempfile, base);
4079 temp->tempfile = mks_tempfile_ts(tempfile.buf, strlen(base) + 1);
4081 if (!temp->tempfile)
4082 die_errno("unable to create temp-file");
4083 if (convert_to_working_tree(istate, path,
4084 (const char *)blob, (size_t)size, &buf)) {
4085 blob = buf.buf;
4086 size = buf.len;
4087 }
4088 if (write_in_full(temp->tempfile->fd, blob, size) < 0 ||
4089 close_tempfile_gently(temp->tempfile))
4090 die_errno("unable to write temp-file");
4091 temp->name = get_tempfile_path(temp->tempfile);
4092 oid_to_hex_r(temp->hex, oid);
4093 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", mode);
4094 strbuf_release(&buf);
4095 strbuf_release(&tempfile);
4096 free(path_dup);
4097}
4098static struct diff_tempfile *prepare_temp_file(struct repository *r,
4100 const char *name,
4101 struct diff_filespec *one)
4102{
4103 struct diff_tempfile *temp = claim_diff_tempfile();
4104 if (!DIFF_FILE_VALID(one)) {
4106 not_a_valid_file:
4107 /* A '-' entry produces this for file-2, and
4108 * a '+' entry produces this for file-1.
4109 */
4110 temp->name = "/dev/null";
4111 xsnprintf(temp->hex, sizeof(temp->hex), ".");
4112 xsnprintf(temp->mode, sizeof(temp->mode), ".");
4113 return temp;
4114 }
4115 if (!S_ISGITLINK(one->mode) &&
4117 (!one->oid_valid ||
4118 reuse_worktree_file(r->index, name, &one->oid, 1))) {
4119 struct stat st;
4120 if (lstat(name, &st) < 0) {
4121 if (errno == ENOENT)
4122 goto not_a_valid_file;
4123 die_errno("stat(%s)", name);
4124 }
4125 if (S_ISLNK(st.st_mode)) {
4126 struct strbuf sb = STRBUF_INIT;
4127 if (strbuf_readlink(&sb, name, st.st_size) < 0)
4128 die_errno("readlink(%s)", name);
4129 prep_temp_blob(r->index, name, temp, sb.buf, sb.len,
4130 (one->oid_valid ?
4131 &one->oid : &null_oid),
4132 (one->oid_valid ?
4133 one->mode : S_IFLNK));
4134 strbuf_release(&sb);
4135 }
4136 else {
4137 /* we can borrow from the file in the work tree */
4138 temp->name = name;
4139 if (!one->oid_valid)
4140 oid_to_hex_r(temp->hex, &null_oid);
4141 else
4142 oid_to_hex_r(temp->hex, &one->oid);
4143 /* Even though we may sometimes borrow the
4144 * contents from the work tree, we always want
4145 * one->mode. mode is trustworthy even when
4146 * !(one->oid_valid), as long as
4147 * DIFF_FILE_VALID(one).
4148 */
4149 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", one->mode);
4150 }
4151 return temp;
4152 }
4153 else {
4154 if (diff_populate_filespec(r, one, 0))
4155 die("cannot read data blob for %s", one->path);
4156 prep_temp_blob(r->index, name, temp,
4157 one->data, one->size,
4158 &one->oid, one->mode);
4159 }
4160 return temp;
4161}
4162static void add_external_diff_name(struct repository *r,
4164 struct argv_array *argv,
4165 const char *name,
4166 struct diff_filespec *df)
4167{
4168 struct diff_tempfile *temp = prepare_temp_file(r, name, df);
4169 argv_array_push(argv, temp->name);
4170 argv_array_push(argv, temp->hex);
4171 argv_array_push(argv, temp->mode);
4172}
4173/* An external diff command takes:
4175 *
4176 * diff-cmd name infile1 infile1-sha1 infile1-mode \
4177 * infile2 infile2-sha1 infile2-mode [ rename-to ]
4178 *
4179 */
4180static void run_external_diff(const char *pgm,
4181 const char *name,
4182 const char *other,
4183 struct diff_filespec *one,
4184 struct diff_filespec *two,
4185 const char *xfrm_msg,
4186 int complete_rewrite,
4187 struct diff_options *o)
4188{
4189 struct argv_array argv = ARGV_ARRAY_INIT;
4190 struct argv_array env = ARGV_ARRAY_INIT;
4191 struct diff_queue_struct *q = &diff_queued_diff;
4192 argv_array_push(&argv, pgm);
4194 argv_array_push(&argv, name);
4195 if (one && two) {
4197 add_external_diff_name(o->repo, &argv, name, one);
4198 if (!other)
4199 add_external_diff_name(o->repo, &argv, name, two);
4200 else {
4201 add_external_diff_name(o->repo, &argv, other, two);
4202 argv_array_push(&argv, other);
4203 argv_array_push(&argv, xfrm_msg);
4204 }
4205 }
4206 argv_array_pushf(&env, "GIT_DIFF_PATH_COUNTER=%d", ++o->diff_path_counter);
4208 argv_array_pushf(&env, "GIT_DIFF_PATH_TOTAL=%d", q->nr);
4209 if (run_command_v_opt_cd_env(argv.argv, RUN_USING_SHELL, NULL, env.argv))
4211 die(_("external diff died, stopping at %s"), name);
4212 remove_tempfile();
4214 argv_array_clear(&argv);
4215 argv_array_clear(&env);
4216}
4217static int similarity_index(struct diff_filepair *p)
4219{
4220 return p->score * 100 / MAX_SCORE;
4221}
4222static const char *diff_abbrev_oid(const struct object_id *oid, int abbrev)
4224{
4225 if (startup_info->have_repository)
4226 return find_unique_abbrev(oid, abbrev);
4227 else {
4228 char *hex = oid_to_hex(oid);
4229 if (abbrev < 0)
4230 abbrev = FALLBACK_DEFAULT_ABBREV;
4231 if (abbrev > the_hash_algo->hexsz)
4232 BUG("oid abbreviation out of range: %d", abbrev);
4233 if (abbrev)
4234 hex[abbrev] = '\0';
4235 return hex;
4236 }
4237}
4238static void fill_metainfo(struct strbuf *msg,
4240 const char *name,
4241 const char *other,
4242 struct diff_filespec *one,
4243 struct diff_filespec *two,
4244 struct diff_options *o,
4245 struct diff_filepair *p,
4246 int *must_show_header,
4247 int use_color)
4248{
4249 const char *set = diff_get_color(use_color, DIFF_METAINFO);
4250 const char *reset = diff_get_color(use_color, DIFF_RESET);
4251 const char *line_prefix = diff_line_prefix(o);
4252 *must_show_header = 1;
4254 strbuf_init(msg, PATH_MAX * 2 + 300);
4255 switch (p->status) {
4256 case DIFF_STATUS_COPIED:
4257 strbuf_addf(msg, "%s%ssimilarity index %d%%",
4258 line_prefix, set, similarity_index(p));
4259 strbuf_addf(msg, "%s\n%s%scopy from ",
4260 reset, line_prefix, set);
4261 quote_c_style(name, msg, NULL, 0);
4262 strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
4263 quote_c_style(other, msg, NULL, 0);
4264 strbuf_addf(msg, "%s\n", reset);
4265 break;
4266 case DIFF_STATUS_RENAMED:
4267 strbuf_addf(msg, "%s%ssimilarity index %d%%",
4268 line_prefix, set, similarity_index(p));
4269 strbuf_addf(msg, "%s\n%s%srename from ",
4270 reset, line_prefix, set);
4271 quote_c_style(name, msg, NULL, 0);
4272 strbuf_addf(msg, "%s\n%s%srename to ",
4273 reset, line_prefix, set);
4274 quote_c_style(other, msg, NULL, 0);
4275 strbuf_addf(msg, "%s\n", reset);
4276 break;
4277 case DIFF_STATUS_MODIFIED:
4278 if (p->score) {
4279 strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
4280 line_prefix,
4281 set, similarity_index(p), reset);
4282 break;
4283 }
4284 /* fallthru */
4285 default:
4286 *must_show_header = 0;
4287 }
4288 if (one && two && !oideq(&one->oid, &two->oid)) {
4289 const unsigned hexsz = the_hash_algo->hexsz;
4290 int abbrev = o->flags.full_index ? hexsz : DEFAULT_ABBREV;
4291 if (o->flags.binary) {
4293 mmfile_t mf;
4294 if ((!fill_mmfile(o->repo, &mf, one) &&
4295 diff_filespec_is_binary(o->repo, one)) ||
4296 (!fill_mmfile(o->repo, &mf, two) &&
4297 diff_filespec_is_binary(o->repo, two)))
4298 abbrev = hexsz;
4299 }
4300 strbuf_addf(msg, "%s%sindex %s..%s", line_prefix, set,
4301 diff_abbrev_oid(&one->oid, abbrev),
4302 diff_abbrev_oid(&two->oid, abbrev));
4303 if (one->mode == two->mode)
4304 strbuf_addf(msg, " %06o", one->mode);
4305 strbuf_addf(msg, "%s\n", reset);
4306 }
4307}
4308static void run_diff_cmd(const char *pgm,
4310 const char *name,
4311 const char *other,
4312 const char *attr_path,
4313 struct diff_filespec *one,
4314 struct diff_filespec *two,
4315 struct strbuf *msg,
4316 struct diff_options *o,
4317 struct diff_filepair *p)
4318{
4319 const char *xfrm_msg = NULL;
4320 int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
4321 int must_show_header = 0;
4322 if (o->flags.allow_external) {
4325 struct userdiff_driver *drv;
4326 drv = userdiff_find_by_path(o->repo->index, attr_path);
4328 if (drv && drv->external)
4329 pgm = drv->external;
4330 }
4331 if (msg) {
4333 /*
4334 * don't use colors when the header is intended for an
4335 * external diff driver
4336 */
4337 fill_metainfo(msg, name, other, one, two, o, p,
4338 &must_show_header,
4339 want_color(o->use_color) && !pgm);
4340 xfrm_msg = msg->len ? msg->buf : NULL;
4341 }
4342 if (pgm) {
4344 run_external_diff(pgm, name, other, one, two, xfrm_msg,
4345 complete_rewrite, o);
4346 return;
4347 }
4348 if (one && two)
4349 builtin_diff(name, other ? other : name,
4350 one, two, xfrm_msg, must_show_header,
4351 o, complete_rewrite);
4352 else
4353 fprintf(o->file, "* Unmerged path %s\n", name);
4354}
4355static void diff_fill_oid_info(struct diff_filespec *one, struct index_state *istate)
4357{
4358 if (DIFF_FILE_VALID(one)) {
4359 if (!one->oid_valid) {
4360 struct stat st;
4361 if (one->is_stdin) {
4362 oidclr(&one->oid);
4363 return;
4364 }
4365 if (lstat(one->path, &st) < 0)
4366 die_errno("stat '%s'", one->path);
4367 if (index_path(istate, &one->oid, one->path, &st, 0))
4368 die("cannot hash %s", one->path);
4369 }
4370 }
4371 else
4372 oidclr(&one->oid);
4373}
4374static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
4376{
4377 /* Strip the prefix but do not molest /dev/null and absolute paths */
4378 if (*namep && !is_absolute_path(*namep)) {
4379 *namep += prefix_length;
4380 if (**namep == '/')
4381 ++*namep;
4382 }
4383 if (*otherp && !is_absolute_path(*otherp)) {
4384 *otherp += prefix_length;
4385 if (**otherp == '/')
4386 ++*otherp;
4387 }
4388}
4389static void run_diff(struct diff_filepair *p, struct diff_options *o)
4391{
4392 const char *pgm = external_diff();
4393 struct strbuf msg;
4394 struct diff_filespec *one = p->one;
4395 struct diff_filespec *two = p->two;
4396 const char *name;
4397 const char *other;
4398 const char *attr_path;
4399 name = one->path;
4401 other = (strcmp(name, two->path) ? two->path : NULL);
4402 attr_path = name;
4403 if (o->prefix_length)
4404 strip_prefix(o->prefix_length, &name, &other);
4405 if (!o->flags.allow_external)
4407 pgm = NULL;
4408 if (DIFF_PAIR_UNMERGED(p)) {
4410 run_diff_cmd(pgm, name, NULL, attr_path,
4411 NULL, NULL, NULL, o, p);
4412 return;
4413 }
4414 diff_fill_oid_info(one, o->repo->index);
4416 diff_fill_oid_info(two, o->repo->index);
4417 if (!pgm &&
4419 DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
4420 (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
4421 /*
4422 * a filepair that changes between file and symlink
4423 * needs to be split into deletion and creation.
4424 */
4425 struct diff_filespec *null = alloc_filespec(two->path);
4426 run_diff_cmd(NULL, name, other, attr_path,
4427 one, null, &msg,
4428 o, p);
4429 free(null);
4430 strbuf_release(&msg);
4431 null = alloc_filespec(one->path);
4433 run_diff_cmd(NULL, name, other, attr_path,
4434 null, two, &msg, o, p);
4435 free(null);
4436 }
4437 else
4438 run_diff_cmd(pgm, name, other, attr_path,
4439 one, two, &msg, o, p);
4440 strbuf_release(&msg);
4442}
4443static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
4445 struct diffstat_t *diffstat)
4446{
4447 const char *name;
4448 const char *other;
4449 if (DIFF_PAIR_UNMERGED(p)) {
4451 /* unmerged */
4452 builtin_diffstat(p->one->path, NULL, NULL, NULL,
4453 diffstat, o, p);
4454 return;
4455 }
4456 name = p->one->path;
4458 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
4459 if (o->prefix_length)
4461 strip_prefix(o->prefix_length, &name, &other);
4462 diff_fill_oid_info(p->one, o->repo->index);
4464 diff_fill_oid_info(p->two, o->repo->index);
4465 builtin_diffstat(name, other, p->one, p->two,
4467 diffstat, o, p);
4468}
4469static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
4471{
4472 const char *name;
4473 const char *other;
4474 const char *attr_path;
4475 if (DIFF_PAIR_UNMERGED(p)) {
4477 /* unmerged */
4478 return;
4479 }
4480 name = p->one->path;
4482 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
4483 attr_path = other ? other : name;
4484 if (o->prefix_length)
4486 strip_prefix(o->prefix_length, &name, &other);
4487 diff_fill_oid_info(p->one, o->repo->index);
4489 diff_fill_oid_info(p->two, o->repo->index);
4490 builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
4492}
4493void repo_diff_setup(struct repository *r, struct diff_options *options)
4495{
4496 memcpy(options, &default_diff_options, sizeof(*options));
4497 options->file = stdout;
4499 options->repo = r;
4500 options->output_indicators[OUTPUT_INDICATOR_NEW] = '+';
4502 options->output_indicators[OUTPUT_INDICATOR_OLD] = '-';
4503 options->output_indicators[OUTPUT_INDICATOR_CONTEXT] = ' ';
4504 options->abbrev = DEFAULT_ABBREV;
4505 options->line_termination = '\n';
4506 options->break_opt = -1;
4507 options->rename_limit = -1;
4508 options->dirstat_permille = diff_dirstat_permille_default;
4509 options->context = diff_context_default;
4510 options->interhunkcontext = diff_interhunk_context_default;
4511 options->ws_error_highlight = ws_error_highlight_default;
4512 options->flags.rename_empty = 1;
4513 options->objfind = NULL;
4514 /* pathchange left =NULL by default */
4516 options->change = diff_change;
4517 options->add_remove = diff_addremove;
4518 options->use_color = diff_use_color_default;
4519 options->detect_rename = diff_detect_rename_default;
4520 options->xdl_opts |= diff_algorithm;
4521 if (diff_indent_heuristic)
4522 DIFF_XDL_SET(options, INDENT_HEURISTIC);
4523 options->orderfile = diff_order_file_cfg;
4525 if (diff_no_prefix) {
4527 options->a_prefix = options->b_prefix = "";
4528 } else if (!diff_mnemonic_prefix) {
4529 options->a_prefix = "a/";
4530 options->b_prefix = "b/";
4531 }
4532 options->color_moved = diff_color_moved_default;
4534 options->color_moved_ws_handling = diff_color_moved_ws_default;
4535}
4536void diff_setup_done(struct diff_options *options)
4538{
4539 unsigned check_mask = DIFF_FORMAT_NAME |
4540 DIFF_FORMAT_NAME_STATUS |
4541 DIFF_FORMAT_CHECKDIFF |
4542 DIFF_FORMAT_NO_OUTPUT;
4543 /*
4544 * This must be signed because we're comparing against a potentially
4545 * negative value.
4546 */
4547 const int hexsz = the_hash_algo->hexsz;
4548 if (options->set_default)
4550 options->set_default(options);
4551 if (HAS_MULTI_BITS(options->output_format & check_mask))
4553 die(_("--name-only, --name-status, --check and -s are mutually exclusive"));
4554 if (HAS_MULTI_BITS(options->pickaxe_opts & DIFF_PICKAXE_KINDS_MASK))
4556 die(_("-G, -S and --find-object are mutually exclusive"));
4557 /*
4559 * Most of the time we can say "there are changes"
4560 * only by checking if there are changed paths, but
4561 * --ignore-whitespace* options force us to look
4562 * inside contents.
4563 */
4564 if ((options->xdl_opts & XDF_WHITESPACE_FLAGS))
4566 options->flags.diff_from_contents = 1;
4567 else
4568 options->flags.diff_from_contents = 0;
4569 if (options->flags.find_copies_harder)
4571 options->detect_rename = DIFF_DETECT_COPY;
4572 if (!options->flags.relative_name)
4574 options->prefix = NULL;
4575 if (options->prefix)
4576 options->prefix_length = strlen(options->prefix);
4577 else
4578 options->prefix_length = 0;
4579 if (options->output_format & (DIFF_FORMAT_NAME |
4581 DIFF_FORMAT_NAME_STATUS |
4582 DIFF_FORMAT_CHECKDIFF |
4583 DIFF_FORMAT_NO_OUTPUT))
4584 options->output_format &= ~(DIFF_FORMAT_RAW |
4585 DIFF_FORMAT_NUMSTAT |
4586 DIFF_FORMAT_DIFFSTAT |
4587 DIFF_FORMAT_SHORTSTAT |
4588 DIFF_FORMAT_DIRSTAT |
4589 DIFF_FORMAT_SUMMARY |
4590 DIFF_FORMAT_PATCH);
4591 /*
4593 * These cases always need recursive; we do not drop caller-supplied
4594 * recursive bits for other formats here.
4595 */
4596 if (options->output_format & (DIFF_FORMAT_PATCH |
4597 DIFF_FORMAT_NUMSTAT |
4598 DIFF_FORMAT_DIFFSTAT |
4599 DIFF_FORMAT_SHORTSTAT |
4600 DIFF_FORMAT_DIRSTAT |
4601 DIFF_FORMAT_SUMMARY |
4602 DIFF_FORMAT_CHECKDIFF))
4603 options->flags.recursive = 1;
4604 /*
4605 * Also pickaxe would not work very well if you do not say recursive
4606 */
4607 if (options->pickaxe_opts & DIFF_PICKAXE_KINDS_MASK)
4608 options->flags.recursive = 1;
4609 /*
4610 * When patches are generated, submodules diffed against the work tree
4611 * must be checked for dirtiness too so it can be shown in the output
4612 */
4613 if (options->output_format & DIFF_FORMAT_PATCH)
4614 options->flags.dirty_submodules = 1;
4615 if (options->detect_rename && options->rename_limit < 0)
4617 options->rename_limit = diff_rename_limit_default;
4618 if (hexsz < options->abbrev)
4619 options->abbrev = hexsz; /* full */
4620 /*
4622 * It does not make sense to show the first hit we happened
4623 * to have found. It does not make sense not to return with
4624 * exit code in such a case either.
4625 */
4626 if (options->flags.quick) {
4627 options->output_format = DIFF_FORMAT_NO_OUTPUT;
4628 options->flags.exit_with_status = 1;
4629 }
4630 options->diff_path_counter = 0;
4632 if (options->flags.follow_renames && options->pathspec.nr != 1)
4634 die(_("--follow requires exactly one pathspec"));
4635 if (!options->use_color || external_diff())
4637 options->color_moved = 0;
4638}
4639static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
4641{
4642 char c, *eq;
4643 int len;
4644 if (*arg != '-')
4646 return 0;
4647 c = *++arg;
4648 if (!c)
4649 return 0;
4650 if (c == arg_short) {
4651 c = *++arg;
4652 if (!c)
4653 return 1;
4654 if (val && isdigit(c)) {
4655 char *end;
4656 int n = strtoul(arg, &end, 10);
4657 if (*end)
4658 return 0;
4659 *val = n;
4660 return 1;
4661 }
4662 return 0;
4663 }
4664 if (c != '-')
4665 return 0;
4666 arg++;
4667 eq = strchrnul(arg, '=');
4668 len = eq - arg;
4669 if (!len || strncmp(arg, arg_long, len))
4670 return 0;
4671 if (*eq) {
4672 int n;
4673 char *end;
4674 if (!isdigit(*++eq))
4675 return 0;
4676 n = strtoul(eq, &end, 10);
4677 if (*end)
4678 return 0;
4679 *val = n;
4680 }
4681 return 1;
4682}
4683static int diff_scoreopt_parse(const char *opt);
4685static inline int short_opt(char opt, const char **argv,
4687 const char **optarg)
4688{
4689 const char *arg = argv[0];
4690 if (arg[0] != '-' || arg[1] != opt)
4691 return 0;
4692 if (arg[2] != '\0') {
4693 *optarg = arg + 2;
4694 return 1;
4695 }
4696 if (!argv[1])
4697 die("Option '%c' requires a value", opt);
4698 *optarg = argv[1];
4699 return 2;
4700}
4701int parse_long_opt(const char *opt, const char **argv,
4703 const char **optarg)
4704{
4705 const char *arg = argv[0];
4706 if (!skip_prefix(arg, "--", &arg))
4707 return 0;
4708 if (!skip_prefix(arg, opt, &arg))
4709 return 0;
4710 if (*arg == '=') { /* stuck form: --option=value */
4711 *optarg = arg + 1;
4712 return 1;
4713 }
4714 if (*arg != '\0')
4715 return 0;
4716 /* separate form: --option value */
4717 if (!argv[1])
4718 die("Option '--%s' requires a value", opt);
4719 *optarg = argv[1];
4720 return 2;
4721}
4722static int stat_opt(struct diff_options *options, const char **av)
4724{
4725 const char *arg = av[0];
4726 char *end;
4727 int width = options->stat_width;
4728 int name_width = options->stat_name_width;
4729 int graph_width = options->stat_graph_width;
4730 int count = options->stat_count;
4731 int argcount = 1;
4732 if (!skip_prefix(arg, "--stat", &arg))
4734 BUG("stat option does not begin with --stat: %s", arg);
4735 end = (char *)arg;
4736 switch (*arg) {
4738 case '-':
4739 if (skip_prefix(arg, "-width", &arg)) {
4740 if (*arg == '=')
4741 width = strtoul(arg + 1, &end, 10);
4742 else if (!*arg && !av[1])
4743 die_want_option("--stat-width");
4744 else if (!*arg) {
4745 width = strtoul(av[1], &end, 10);
4746 argcount = 2;
4747 }
4748 } else if (skip_prefix(arg, "-name-width", &arg)) {
4749 if (*arg == '=')
4750 name_width = strtoul(arg + 1, &end, 10);
4751 else if (!*arg && !av[1])
4752 die_want_option("--stat-name-width");
4753 else if (!*arg) {
4754 name_width = strtoul(av[1], &end, 10);
4755 argcount = 2;
4756 }
4757 } else if (skip_prefix(arg, "-graph-width", &arg)) {
4758 if (*arg == '=')
4759 graph_width = strtoul(arg + 1, &end, 10);
4760 else if (!*arg && !av[1])
4761 die_want_option("--stat-graph-width");
4762 else if (!*arg) {
4763 graph_width = strtoul(av[1], &end, 10);
4764 argcount = 2;
4765 }
4766 } else if (skip_prefix(arg, "-count", &arg)) {
4767 if (*arg == '=')
4768 count = strtoul(arg + 1, &end, 10);
4769 else if (!*arg && !av[1])
4770 die_want_option("--stat-count");
4771 else if (!*arg) {
4772 count = strtoul(av[1], &end, 10);
4773 argcount = 2;
4774 }
4775 }
4776 break;
4777 case '=':
4778 width = strtoul(arg+1, &end, 10);
4779 if (*end == ',')
4780 name_width = strtoul(end+1, &end, 10);
4781 if (*end == ',')
4782 count = strtoul(end+1, &end, 10);
4783 }
4784 /* Important! This checks all the error cases! */
4786 if (*end)
4787 return 0;
4788 options->output_format |= DIFF_FORMAT_DIFFSTAT;
4789 options->stat_name_width = name_width;
4790 options->stat_graph_width = graph_width;
4791 options->stat_width = width;
4792 options->stat_count = count;
4793 return argcount;
4794}
4795static int parse_dirstat_opt(struct diff_options *options, const char *params)
4797{
4798 struct strbuf errmsg = STRBUF_INIT;
4799 if (parse_dirstat_params(options, params, &errmsg))
4800 die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
4801 errmsg.buf);
4802 strbuf_release(&errmsg);
4803 /*
4804 * The caller knows a dirstat-related option is given from the command
4805 * line; allow it to say "return this_function();"
4806 */
4807 options->output_format |= DIFF_FORMAT_DIRSTAT;
4808 return 1;
4809}
4810static int parse_submodule_opt(struct diff_options *options, const char *value)
4812{
4813 if (parse_submodule_params(options, value))
4814 die(_("Failed to parse --submodule option parameter: '%s'"),
4815 value);
4816 return 1;
4817}
4818static const char diff_status_letters[] = {
4820 DIFF_STATUS_ADDED,
4821 DIFF_STATUS_COPIED,
4822 DIFF_STATUS_DELETED,
4823 DIFF_STATUS_MODIFIED,
4824 DIFF_STATUS_RENAMED,
4825 DIFF_STATUS_TYPE_CHANGED,
4826 DIFF_STATUS_UNKNOWN,
4827 DIFF_STATUS_UNMERGED,
4828 DIFF_STATUS_FILTER_AON,
4829 DIFF_STATUS_FILTER_BROKEN,
4830 '\0',
4831};
4832static unsigned int filter_bit['Z' + 1];
4834static void prepare_filter_bits(void)
4836{
4837 int i;
4838 if (!filter_bit[DIFF_STATUS_ADDED]) {
4840 for (i = 0; diff_status_letters[i]; i++)
4841 filter_bit[(int) diff_status_letters[i]] = (1 << i);
4842 }
4843}
4844static unsigned filter_bit_tst(char status, const struct diff_options *opt)
4846{
4847 return opt->filter & filter_bit[(int) status];
4848}
4849static int parse_diff_filter_opt(const char *optarg, struct diff_options *opt)
4851{
4852 int i, optch;
4853 prepare_filter_bits();
4855 /*
4857 * If there is a negation e.g. 'd' in the input, and we haven't
4858 * initialized the filter field with another --diff-filter, start
4859 * from full set of bits, except for AON.
4860 */
4861 if (!opt->filter) {
4862 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
4863 if (optch < 'a' || 'z' < optch)
4864 continue;
4865 opt->filter = (1 << (ARRAY_SIZE(diff_status_letters) - 1)) - 1;
4866 opt->filter &= ~filter_bit[DIFF_STATUS_FILTER_AON];
4867 break;
4868 }
4869 }
4870 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
4872 unsigned int bit;
4873 int negate;
4874 if ('a' <= optch && optch <= 'z') {
4876 negate = 1;
4877 optch = toupper(optch);
4878 } else {
4879 negate = 0;
4880 }
4881 bit = (0 <= optch && optch <= 'Z') ? filter_bit[optch] : 0;
4883 if (!bit)
4884 return optarg[i];
4885 if (negate)
4886 opt->filter &= ~bit;
4887 else
4888 opt->filter |= bit;
4889 }
4890 return 0;
4891}
4892static void enable_patch_output(int *fmt)
4894{
4895 *fmt &= ~DIFF_FORMAT_NO_OUTPUT;
4896 *fmt |= DIFF_FORMAT_PATCH;
4897}
4898static int parse_ws_error_highlight_opt(struct diff_options *opt, const char *arg)
4900{
4901 int val = parse_ws_error_highlight(arg);
4902 if (val < 0) {
4904 error("unknown value after ws-error-highlight=%.*s",
4905 -1 - val, arg);
4906 return 0;
4907 }
4908 opt->ws_error_highlight = val;
4909 return 1;
4910}
4911static int parse_objfind_opt(struct diff_options *opt, const char *arg)
4913{
4914 struct object_id oid;
4915 if (get_oid(arg, &oid))
4917 return error("unable to resolve '%s'", arg);
4918 if (!opt->objfind)
4920 opt->objfind = xcalloc(1, sizeof(*opt->objfind));
4921 opt->pickaxe_opts |= DIFF_PICKAXE_KIND_OBJFIND;
4923 opt->flags.recursive = 1;
4924 opt->flags.tree_in_recursive = 1;
4925 oidset_insert(opt->objfind, &oid);
4926 return 1;
4927}
4928int diff_opt_parse(struct diff_options *options,
4930 const char **av, int ac, const char *prefix)
4931{
4932 const char *arg = av[0];
4933 const char *optarg;
4934 int argcount;
4935 if (!prefix)
4937 prefix = "";
4938 /* Output format options */
4940 if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch")
4941 || opt_arg(arg, 'U', "unified", &options->context))
4942 enable_patch_output(&options->output_format);
4943 else if (!strcmp(arg, "--raw"))
4944 options->output_format |= DIFF_FORMAT_RAW;
4945 else if (!strcmp(arg, "--patch-with-raw")) {
4946 enable_patch_output(&options->output_format);
4947 options->output_format |= DIFF_FORMAT_RAW;
4948 } else if (!strcmp(arg, "--numstat"))
4949 options->output_format |= DIFF_FORMAT_NUMSTAT;
4950 else if (!strcmp(arg, "--shortstat"))
4951 options->output_format |= DIFF_FORMAT_SHORTSTAT;
4952 else if (skip_prefix(arg, "-X", &arg) ||
4953 skip_to_optional_arg(arg, "--dirstat", &arg))
4954 return parse_dirstat_opt(options, arg);
4955 else if (!strcmp(arg, "--cumulative"))
4956 return parse_dirstat_opt(options, "cumulative");
4957 else if (skip_to_optional_arg(arg, "--dirstat-by-file", &arg)) {
4958 parse_dirstat_opt(options, "files");
4959 return parse_dirstat_opt(options, arg);
4960 }
4961 else if (!strcmp(arg, "--check"))
4962 options->output_format |= DIFF_FORMAT_CHECKDIFF;
4963 else if (!strcmp(arg, "--summary"))
4964 options->output_format |= DIFF_FORMAT_SUMMARY;
4965 else if (!strcmp(arg, "--patch-with-stat")) {
4966 enable_patch_output(&options->output_format);
4967 options->output_format |= DIFF_FORMAT_DIFFSTAT;
4968 } else if (!strcmp(arg, "--name-only"))
4969 options->output_format |= DIFF_FORMAT_NAME;
4970 else if (!strcmp(arg, "--name-status"))
4971 options->output_format |= DIFF_FORMAT_NAME_STATUS;
4972 else if (!strcmp(arg, "-s") || !strcmp(arg, "--no-patch"))
4973 options->output_format |= DIFF_FORMAT_NO_OUTPUT;
4974 else if (starts_with(arg, "--stat"))
4975 /* --stat, --stat-width, --stat-name-width, or --stat-count */
4976 return stat_opt(options, av);
4977 else if (!strcmp(arg, "--compact-summary")) {
4978 options->flags.stat_with_summary = 1;
4979 options->output_format |= DIFF_FORMAT_DIFFSTAT;
4980 } else if (!strcmp(arg, "--no-compact-summary"))
4981 options->flags.stat_with_summary = 0;
4982 else if (skip_prefix(arg, "--output-indicator-new=", &arg))
4983 options->output_indicators[OUTPUT_INDICATOR_NEW] = arg[0];
4984 else if (skip_prefix(arg, "--output-indicator-old=", &arg))
4985 options->output_indicators[OUTPUT_INDICATOR_OLD] = arg[0];
4986 else if (skip_prefix(arg, "--output-indicator-context=", &arg))
4987 options->output_indicators[OUTPUT_INDICATOR_CONTEXT] = arg[0];
4988 /* renames options */
4990 else if (starts_with(arg, "-B") ||
4991 skip_to_optional_arg(arg, "--break-rewrites", NULL)) {
4992 if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
4993 return error("invalid argument to -B: %s", arg+2);
4994 }
4995 else if (starts_with(arg, "-M") ||
4996 skip_to_optional_arg(arg, "--find-renames", NULL)) {
4997 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
4998 return error("invalid argument to -M: %s", arg+2);
4999 options->detect_rename = DIFF_DETECT_RENAME;
5000 }
5001 else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
5002 options->irreversible_delete = 1;
5003 }
5004 else if (starts_with(arg, "-C") ||
5005 skip_to_optional_arg(arg, "--find-copies", NULL)) {
5006 if (options->detect_rename == DIFF_DETECT_COPY)
5007 options->flags.find_copies_harder = 1;
5008 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
5009 return error("invalid argument to -C: %s", arg+2);
5010 options->detect_rename = DIFF_DETECT_COPY;
5011 }
5012 else if (!strcmp(arg, "--no-renames"))
5013 options->detect_rename = 0;
5014 else if (!strcmp(arg, "--rename-empty"))
5015 options->flags.rename_empty = 1;
5016 else if (!strcmp(arg, "--no-rename-empty"))
5017 options->flags.rename_empty = 0;
5018 else if (skip_to_optional_arg_default(arg, "--relative", &arg, NULL)) {
5019 options->flags.relative_name = 1;
5020 if (arg)
5021 options->prefix = arg;
5022 }
5023 /* xdiff options */
5025 else if (!strcmp(arg, "--minimal"))
5026 DIFF_XDL_SET(options, NEED_MINIMAL);
5027 else if (!strcmp(arg, "--no-minimal"))
5028 DIFF_XDL_CLR(options, NEED_MINIMAL);
5029 else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
5030 DIFF_XDL_SET(options, IGNORE_WHITESPACE);
5031 else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
5032 DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
5033 else if (!strcmp(arg, "--ignore-space-at-eol"))
5034 DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
5035 else if (!strcmp(arg, "--ignore-cr-at-eol"))
5036 DIFF_XDL_SET(options, IGNORE_CR_AT_EOL);
5037 else if (!strcmp(arg, "--ignore-blank-lines"))
5038 DIFF_XDL_SET(options, IGNORE_BLANK_LINES);
5039 else if (!strcmp(arg, "--indent-heuristic"))
5040 DIFF_XDL_SET(options, INDENT_HEURISTIC);
5041 else if (!strcmp(arg, "--no-indent-heuristic"))
5042 DIFF_XDL_CLR(options, INDENT_HEURISTIC);
5043 else if (!strcmp(arg, "--patience")) {
5044 int i;
5045 options->xdl_opts = DIFF_WITH_ALG(options, PATIENCE_DIFF);
5046 /*
5047 * Both --patience and --anchored use PATIENCE_DIFF
5048 * internally, so remove any anchors previously
5049 * specified.
5050 */
5051 for (i = 0; i < options->anchors_nr; i++)
5052 free(options->anchors[i]);
5053 options->anchors_nr = 0;
5054 } else if (!strcmp(arg, "--histogram"))
5055 options->xdl_opts = DIFF_WITH_ALG(options, HISTOGRAM_DIFF);
5056 else if ((argcount = parse_long_opt("diff-algorithm", av, &optarg))) {
5057 long value = parse_algorithm_value(optarg);
5058 if (value < 0)
5059 return error("option diff-algorithm accepts \"myers\", "
5060 "\"minimal\", \"patience\" and \"histogram\"");
5061 /* clear out previous settings */
5062 DIFF_XDL_CLR(options, NEED_MINIMAL);
5063 options->xdl_opts &= ~XDF_DIFF_ALGORITHM_MASK;
5064 options->xdl_opts |= value;
5065 return argcount;
5066 } else if (skip_prefix(arg, "--anchored=", &arg)) {
5067 options->xdl_opts = DIFF_WITH_ALG(options, PATIENCE_DIFF);
5068 ALLOC_GROW(options->anchors, options->anchors_nr + 1,
5069 options->anchors_alloc);
5070 options->anchors[options->anchors_nr++] = xstrdup(arg);
5071 }
5072 /* flags options */
5074 else if (!strcmp(arg, "--binary")) {
5075 enable_patch_output(&options->output_format);
5076 options->flags.binary = 1;
5077 }
5078 else if (!strcmp(arg, "--full-index"))
5079 options->flags.full_index = 1;
5080 else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
5081 options->flags.text = 1;
5082 else if (!strcmp(arg, "-R"))
5083 options->flags.reverse_diff = 1;
5084 else if (!strcmp(arg, "--find-copies-harder"))
5085 options->flags.find_copies_harder = 1;
5086 else if (!strcmp(arg, "--follow"))
5087 options->flags.follow_renames = 1;
5088 else if (!strcmp(arg, "--no-follow")) {
5089 options->flags.follow_renames = 0;
5090 options->flags.default_follow_renames = 0;
5091 } else if (skip_to_optional_arg_default(arg, "--color", &arg, "always")) {
5092 int value = git_config_colorbool(NULL, arg);
5093 if (value < 0)
5094 return error("option `color' expects \"always\", \"auto\", or \"never\"");
5095 options->use_color = value;
5096 }
5097 else if (!strcmp(arg, "--no-color"))
5098 options->use_color = 0;
5099 else if (!strcmp(arg, "--color-moved")) {
5100 if (diff_color_moved_default)
5101 options->color_moved = diff_color_moved_default;
5102 if (options->color_moved == COLOR_MOVED_NO)
5103 options->color_moved = COLOR_MOVED_DEFAULT;
5104 } else if (!strcmp(arg, "--no-color-moved"))
5105 options->color_moved = COLOR_MOVED_NO;
5106 else if (skip_prefix(arg, "--color-moved=", &arg)) {
5107 int cm = parse_color_moved(arg);
5108 if (cm < 0)
5109 return error("bad --color-moved argument: %s", arg);
5110 options->color_moved = cm;
5111 } else if (!strcmp(arg, "--no-color-moved-ws")) {
5112 options->color_moved_ws_handling = 0;
5113 } else if (skip_prefix(arg, "--color-moved-ws=", &arg)) {
5114 unsigned cm = parse_color_moved_ws(arg);
5115 if (cm & COLOR_MOVED_WS_ERROR)
5116 return -1;
5117 options->color_moved_ws_handling = cm;
5118 } else if (skip_to_optional_arg_default(arg, "--color-words", &options->word_regex, NULL)) {
5119 options->use_color = 1;
5120 options->word_diff = DIFF_WORDS_COLOR;
5121 }
5122 else if (!strcmp(arg, "--word-diff")) {
5123 if (options->word_diff == DIFF_WORDS_NONE)
5124 options->word_diff = DIFF_WORDS_PLAIN;
5125 }
5126 else if (skip_prefix(arg, "--word-diff=", &arg)) {
5127 if (!strcmp(arg, "plain"))
5128 options->word_diff = DIFF_WORDS_PLAIN;
5129 else if (!strcmp(arg, "color")) {
5130 options->use_color = 1;
5131 options->word_diff = DIFF_WORDS_COLOR;
5132 }
5133 else if (!strcmp(arg, "porcelain"))
5134 options->word_diff = DIFF_WORDS_PORCELAIN;
5135 else if (!strcmp(arg, "none"))
5136 options->word_diff = DIFF_WORDS_NONE;
5137 else
5138 die("bad --word-diff argument: %s", arg);
5139 }
5140 else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
5141 if (options->word_diff == DIFF_WORDS_NONE)
5142 options->word_diff = DIFF_WORDS_PLAIN;
5143 options->word_regex = optarg;
5144 return argcount;
5145 }
5146 else if (!strcmp(arg, "--exit-code"))
5147 options->flags.exit_with_status = 1;
5148 else if (!strcmp(arg, "--quiet"))
5149 options->flags.quick = 1;
5150 else if (!strcmp(arg, "--ext-diff"))
5151 options->flags.allow_external = 1;
5152 else if (!strcmp(arg, "--no-ext-diff"))
5153 options->flags.allow_external = 0;
5154 else if (!strcmp(arg, "--textconv")) {
5155 options->flags.allow_textconv = 1;
5156 options->flags.textconv_set_via_cmdline = 1;
5157 } else if (!strcmp(arg, "--no-textconv"))
5158 options->flags.allow_textconv = 0;
5159 else if (skip_to_optional_arg_default(arg, "--ignore-submodules", &arg, "all")) {
5160 options->flags.override_submodule_config = 1;
5161 handle_ignore_submodules_arg(options, arg);
5162 } else if (skip_to_optional_arg_default(arg, "--submodule", &arg, "log"))
5163 return parse_submodule_opt(options, arg);
5164 else if (skip_prefix(arg, "--ws-error-highlight=", &arg))
5165 return parse_ws_error_highlight_opt(options, arg);
5166 else if (!strcmp(arg, "--ita-invisible-in-index"))
5167 options->ita_invisible_in_index = 1;
5168 else if (!strcmp(arg, "--ita-visible-in-index"))
5169 options->ita_invisible_in_index = 0;
5170 /* misc options */
5172 else if (!strcmp(arg, "-z"))
5173 options->line_termination = 0;
5174 else if ((argcount = short_opt('l', av, &optarg))) {
5175 options->rename_limit = strtoul(optarg, NULL, 10);
5176 return argcount;
5177 }
5178 else if ((argcount = short_opt('S', av, &optarg))) {
5179 options->pickaxe = optarg;
5180 options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
5181 return argcount;
5182 } else if ((argcount = short_opt('G', av, &optarg))) {
5183 options->pickaxe = optarg;
5184 options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
5185 return argcount;
5186 }
5187 else if (!strcmp(arg, "--pickaxe-all"))
5188 options->pickaxe_opts |= DIFF_PICKAXE_ALL;
5189 else if (!strcmp(arg, "--pickaxe-regex"))
5190 options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
5191 else if ((argcount = short_opt('O', av, &optarg))) {
5192 options->orderfile = prefix_filename(prefix, optarg);
5193 return argcount;
5194 } else if (skip_prefix(arg, "--find-object=", &arg))
5195 return parse_objfind_opt(options, arg);
5196 else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
5197 int offending = parse_diff_filter_opt(optarg, options);
5198 if (offending)
5199 die("unknown change class '%c' in --diff-filter=%s",
5200 offending, optarg);
5201 return argcount;
5202 }
5203 else if (!strcmp(arg, "--no-abbrev"))
5204 options->abbrev = 0;
5205 else if (!strcmp(arg, "--abbrev"))
5206 options->abbrev = DEFAULT_ABBREV;
5207 else if (skip_prefix(arg, "--abbrev=", &arg)) {
5208 options->abbrev = strtoul(arg, NULL, 10);
5209 if (options->abbrev < MINIMUM_ABBREV)
5210 options->abbrev = MINIMUM_ABBREV;
5211 else if (the_hash_algo->hexsz < options->abbrev)
5212 options->abbrev = the_hash_algo->hexsz;
5213 }
5214 else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
5215 options->a_prefix = optarg;
5216 return argcount;
5217 }
5218 else if ((argcount = parse_long_opt("line-prefix", av, &optarg))) {
5219 options->line_prefix = optarg;
5220 options->line_prefix_length = strlen(options->line_prefix);
5221 graph_setup_line_prefix(options);
5222 return argcount;
5223 }
5224 else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
5225 options->b_prefix = optarg;
5226 return argcount;
5227 }
5228 else if (!strcmp(arg, "--no-prefix"))
5229 options->a_prefix = options->b_prefix = "";
5230 else if (opt_arg(arg, '\0', "inter-hunk-context",
5231 &options->interhunkcontext))
5232 ;
5233 else if (!strcmp(arg, "-W"))
5234 options->flags.funccontext = 1;
5235 else if (!strcmp(arg, "--function-context"))
5236 options->flags.funccontext = 1;
5237 else if (!strcmp(arg, "--no-function-context"))
5238 options->flags.funccontext = 0;
5239 else if ((argcount = parse_long_opt("output", av, &optarg))) {
5240 char *path = prefix_filename(prefix, optarg);
5241 options->file = xfopen(path, "w");
5242 options->close_file = 1;
5243 if (options->use_color != GIT_COLOR_ALWAYS)
5244 options->use_color = GIT_COLOR_NEVER;
5245 free(path);
5246 return argcount;
5247 } else
5248 return 0;
5249 return 1;
5250}
5251int parse_rename_score(const char **cp_p)
5253{
5254 unsigned long num, scale;
5255 int ch, dot;
5256 const char *cp = *cp_p;
5257 num = 0;
5259 scale = 1;
5260 dot = 0;
5261 for (;;) {
5262 ch = *cp;
5263 if ( !dot && ch == '.' ) {
5264 scale = 1;
5265 dot = 1;
5266 } else if ( ch == '%' ) {
5267 scale = dot ? scale*100 : 100;
5268 cp++; /* % is always at the end */
5269 break;
5270 } else if ( ch >= '0' && ch <= '9' ) {
5271 if ( scale < 100000 ) {
5272 scale *= 10;
5273 num = (num*10) + (ch-'0');
5274 }
5275 } else {
5276 break;
5277 }
5278 cp++;
5279 }
5280 *cp_p = cp;
5281 /* user says num divided by scale and we say internally that
5283 * is MAX_SCORE * num / scale.
5284 */
5285 return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
5286}
5287static int diff_scoreopt_parse(const char *opt)
5289{
5290 int opt1, opt2, cmd;
5291 if (*opt++ != '-')
5293 return -1;
5294 cmd = *opt++;
5295 if (cmd == '-') {
5296 /* convert the long-form arguments into short-form versions */
5297 if (skip_prefix(opt, "break-rewrites", &opt)) {
5298 if (*opt == 0 || *opt++ == '=')
5299 cmd = 'B';
5300 } else if (skip_prefix(opt, "find-copies", &opt)) {
5301 if (*opt == 0 || *opt++ == '=')
5302 cmd = 'C';
5303 } else if (skip_prefix(opt, "find-renames", &opt)) {
5304 if (*opt == 0 || *opt++ == '=')
5305 cmd = 'M';
5306 }
5307 }
5308 if (cmd != 'M' && cmd != 'C' && cmd != 'B')
5309 return -1; /* that is not a -M, -C, or -B option */
5310 opt1 = parse_rename_score(&opt);
5312 if (cmd != 'B')
5313 opt2 = 0;
5314 else {
5315 if (*opt == 0)
5316 opt2 = 0;
5317 else if (*opt != '/')
5318 return -1; /* we expect -B80/99 or -B80 */
5319 else {
5320 opt++;
5321 opt2 = parse_rename_score(&opt);
5322 }
5323 }
5324 if (*opt != 0)
5325 return -1;
5326 return opt1 | (opt2 << 16);
5327}
5328struct diff_queue_struct diff_queued_diff;
5330void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
5332{
5333 ALLOC_GROW(queue->queue, queue->nr + 1, queue->alloc);
5334 queue->queue[queue->nr++] = dp;
5335}
5336struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
5338 struct diff_filespec *one,
5339 struct diff_filespec *two)
5340{
5341 struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
5342 dp->one = one;
5343 dp->two = two;
5344 if (queue)
5345 diff_q(queue, dp);
5346 return dp;
5347}
5348void diff_free_filepair(struct diff_filepair *p)
5350{
5351 free_filespec(p->one);
5352 free_filespec(p->two);
5353 free(p);
5354}
5355const char *diff_aligned_abbrev(const struct object_id *oid, int len)
5357{
5358 int abblen;
5359 const char *abbrev;
5360 /* Do we want all 40 hex characters? */
5362 if (len == the_hash_algo->hexsz)
5363 return oid_to_hex(oid);
5364 /* An abbreviated value is fine, possibly followed by an ellipsis. */
5366 abbrev = diff_abbrev_oid(oid, len);
5367 if (!print_sha1_ellipsis())
5369 return abbrev;
5370 abblen = strlen(abbrev);
5372 /*
5374 * In well-behaved cases, where the abbreviated result is the
5375 * same as the requested length, append three dots after the
5376 * abbreviation (hence the whole logic is limited to the case
5377 * where abblen < 37); when the actual abbreviated result is a
5378 * bit longer than the requested length, we reduce the number
5379 * of dots so that they match the well-behaved ones. However,
5380 * if the actual abbreviation is longer than the requested
5381 * length by more than three, we give up on aligning, and add
5382 * three dots anyway, to indicate that the output is not the
5383 * full object name. Yes, this may be suboptimal, but this
5384 * appears only in "diff --raw --abbrev" output and it is not
5385 * worth the effort to change it now. Note that this would
5386 * likely to work fine when the automatic sizing of default
5387 * abbreviation length is used--we would be fed -1 in "len" in
5388 * that case, and will end up always appending three-dots, but
5389 * the automatic sizing is supposed to give abblen that ensures
5390 * uniqueness across all objects (statistically speaking).
5391 */
5392 if (abblen < the_hash_algo->hexsz - 3) {
5393 static char hex[GIT_MAX_HEXSZ + 1];
5394 if (len < abblen && abblen <= len + 2)
5395 xsnprintf(hex, sizeof(hex), "%s%.*s", abbrev, len+3-abblen, "..");
5396 else
5397 xsnprintf(hex, sizeof(hex), "%s...", abbrev);
5398 return hex;
5399 }
5400 return oid_to_hex(oid);
5402}
5403static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
5405{
5406 int line_termination = opt->line_termination;
5407 int inter_name_termination = line_termination ? '\t' : '\0';
5408 fprintf(opt->file, "%s", diff_line_prefix(opt));
5410 if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
5411 fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
5412 diff_aligned_abbrev(&p->one->oid, opt->abbrev));
5413 fprintf(opt->file, "%s ",
5414 diff_aligned_abbrev(&p->two->oid, opt->abbrev));
5415 }
5416 if (p->score) {
5417 fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
5418 inter_name_termination);
5419 } else {
5420 fprintf(opt->file, "%c%c", p->status, inter_name_termination);
5421 }
5422 if (p->status == DIFF_STATUS_COPIED ||
5424 p->status == DIFF_STATUS_RENAMED) {
5425 const char *name_a, *name_b;
5426 name_a = p->one->path;
5427 name_b = p->two->path;
5428 strip_prefix(opt->prefix_length, &name_a, &name_b);
5429 write_name_quoted(name_a, opt->file, inter_name_termination);
5430 write_name_quoted(name_b, opt->file, line_termination);
5431 } else {
5432 const char *name_a, *name_b;
5433 name_a = p->one->mode ? p->one->path : p->two->path;
5434 name_b = NULL;
5435 strip_prefix(opt->prefix_length, &name_a, &name_b);
5436 write_name_quoted(name_a, opt->file, line_termination);
5437 }
5438}
5439int diff_unmodified_pair(struct diff_filepair *p)
5441{
5442 /* This function is written stricter than necessary to support
5443 * the currently implemented transformers, but the idea is to
5444 * let transformers to produce diff_filepairs any way they want,
5445 * and filter and clean them up here before producing the output.
5446 */
5447 struct diff_filespec *one = p->one, *two = p->two;
5448 if (DIFF_PAIR_UNMERGED(p))
5450 return 0; /* unmerged is interesting */
5451 /* deletion, addition, mode or type change
5453 * and rename are all interesting.
5454 */
5455 if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
5456 DIFF_PAIR_MODE_CHANGED(p) ||
5457 strcmp(one->path, two->path))
5458 return 0;
5459 /* both are valid and point at the same path. that is, we are
5461 * dealing with a change.
5462 */
5463 if (one->oid_valid && two->oid_valid &&
5464 oideq(&one->oid, &two->oid) &&
5465 !one->dirty_submodule && !two->dirty_submodule)
5466 return 1; /* no change */
5467 if (!one->oid_valid && !two->oid_valid)
5468 return 1; /* both look at the same file on the filesystem. */
5469 return 0;
5470}
5471static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
5473{
5474 if (diff_unmodified_pair(p))
5475 return;
5476 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
5478 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
5479 return; /* no tree diffs in patch format */
5480 run_diff(p, o);
5482}
5483static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
5485 struct diffstat_t *diffstat)
5486{
5487 if (diff_unmodified_pair(p))
5488 return;
5489 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
5491 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
5492 return; /* no useful stat for tree diffs */
5493 run_diffstat(p, o, diffstat);
5495}
5496static void diff_flush_checkdiff(struct diff_filepair *p,
5498 struct diff_options *o)
5499{
5500 if (diff_unmodified_pair(p))
5501 return;
5502 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
5504 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
5505 return; /* nothing to check in tree diffs */
5506 run_checkdiff(p, o);
5508}
5509int diff_queue_is_empty(void)
5511{
5512 struct diff_queue_struct *q = &diff_queued_diff;
5513 int i;
5514 for (i = 0; i < q->nr; i++)
5515 if (!diff_unmodified_pair(q->queue[i]))
5516 return 0;
5517 return 1;
5518}
5519#if DIFF_DEBUG
5521void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
5522{
5523 fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
5524 x, one ? one : "",
5525 s->path,
5526 DIFF_FILE_VALID(s) ? "valid" : "invalid",
5527 s->mode,
5528 s->oid_valid ? oid_to_hex(&s->oid) : "");
5529 fprintf(stderr, "queue[%d] %s size %lu\n",
5530 x, one ? one : "",
5531 s->size);
5532}
5533void diff_debug_filepair(const struct diff_filepair *p, int i)
5535{
5536 diff_debug_filespec(p->one, i, "one");
5537 diff_debug_filespec(p->two, i, "two");
5538 fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
5539 p->score, p->status ? p->status : '?',
5540 p->one->rename_used, p->broken_pair);
5541}
5542void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
5544{
5545 int i;
5546 if (msg)
5547 fprintf(stderr, "%s\n", msg);
5548 fprintf(stderr, "q->nr = %d\n", q->nr);
5549 for (i = 0; i < q->nr; i++) {
5550 struct diff_filepair *p = q->queue[i];
5551 diff_debug_filepair(p, i);
5552 }
5553}
5554#endif
5555static void diff_resolve_rename_copy(void)
5557{
5558 int i;
5559 struct diff_filepair *p;
5560 struct diff_queue_struct *q = &diff_queued_diff;
5561 diff_debug_queue("resolve-rename-copy", q);
5563 for (i = 0; i < q->nr; i++) {
5565 p = q->queue[i];
5566 p->status = 0; /* undecided */
5567 if (DIFF_PAIR_UNMERGED(p))
5568 p->status = DIFF_STATUS_UNMERGED;
5569 else if (!DIFF_FILE_VALID(p->one))
5570 p->status = DIFF_STATUS_ADDED;
5571 else if (!DIFF_FILE_VALID(p->two))
5572 p->status = DIFF_STATUS_DELETED;
5573 else if (DIFF_PAIR_TYPE_CHANGED(p))
5574 p->status = DIFF_STATUS_TYPE_CHANGED;
5575 /* from this point on, we are dealing with a pair
5577 * whose both sides are valid and of the same type, i.e.
5578 * either in-place edit or rename/copy edit.
5579 */
5580 else if (DIFF_PAIR_RENAME(p)) {
5581 /*
5582 * A rename might have re-connected a broken
5583 * pair up, causing the pathnames to be the
5584 * same again. If so, that's not a rename at
5585 * all, just a modification..
5586 *
5587 * Otherwise, see if this source was used for
5588 * multiple renames, in which case we decrement
5589 * the count, and call it a copy.
5590 */
5591 if (!strcmp(p->one->path, p->two->path))
5592 p->status = DIFF_STATUS_MODIFIED;
5593 else if (--p->one->rename_used > 0)
5594 p->status = DIFF_STATUS_COPIED;
5595 else
5596 p->status = DIFF_STATUS_RENAMED;
5597 }
5598 else if (!oideq(&p->one->oid, &p->two->oid) ||
5599 p->one->mode != p->two->mode ||
5600 p->one->dirty_submodule ||
5601 p->two->dirty_submodule ||
5602 is_null_oid(&p->one->oid))
5603 p->status = DIFF_STATUS_MODIFIED;
5604 else {
5605 /* This is a "no-change" entry and should not
5606 * happen anymore, but prepare for broken callers.
5607 */
5608 error("feeding unmodified %s to diffcore",
5609 p->one->path);
5610 p->status = DIFF_STATUS_UNKNOWN;
5611 }
5612 }
5613 diff_debug_queue("resolve-rename-copy done", q);
5614}
5615static int check_pair_status(struct diff_filepair *p)
5617{
5618 switch (p->status) {
5619 case DIFF_STATUS_UNKNOWN:
5620 return 0;
5621 case 0:
5622 die("internal error in diff-resolve-rename-copy");
5623 default:
5624 return 1;
5625 }
5626}
5627static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
5629{
5630 int fmt = opt->output_format;
5631 if (fmt & DIFF_FORMAT_CHECKDIFF)
5633 diff_flush_checkdiff(p, opt);
5634 else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
5635 diff_flush_raw(p, opt);
5636 else if (fmt & DIFF_FORMAT_NAME) {
5637 const char *name_a, *name_b;
5638 name_a = p->two->path;
5639 name_b = NULL;
5640 strip_prefix(opt->prefix_length, &name_a, &name_b);
5641 fprintf(opt->file, "%s", diff_line_prefix(opt));
5642 write_name_quoted(name_a, opt->file, opt->line_termination);
5643 }
5644}
5645static void show_file_mode_name(struct diff_options *opt, const char *newdelete, struct diff_filespec *fs)
5647{
5648 struct strbuf sb = STRBUF_INIT;
5649 if (fs->mode)
5650 strbuf_addf(&sb, " %s mode %06o ", newdelete, fs->mode);
5651 else
5652 strbuf_addf(&sb, " %s ", newdelete);
5653 quote_c_style(fs->path, &sb, NULL, 0);
5655 strbuf_addch(&sb, '\n');
5656 emit_diff_symbol(opt, DIFF_SYMBOL_SUMMARY,
5657 sb.buf, sb.len, 0);
5658 strbuf_release(&sb);
5659}
5660static void show_mode_change(struct diff_options *opt, struct diff_filepair *p,
5662 int show_name)
5663{
5664 if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
5665 struct strbuf sb = STRBUF_INIT;
5666 strbuf_addf(&sb, " mode change %06o => %06o",
5667 p->one->mode, p->two->mode);
5668 if (show_name) {
5669 strbuf_addch(&sb, ' ');
5670 quote_c_style(p->two->path, &sb, NULL, 0);
5671 }
5672 strbuf_addch(&sb, '\n');
5673 emit_diff_symbol(opt, DIFF_SYMBOL_SUMMARY,
5674 sb.buf, sb.len, 0);
5675 strbuf_release(&sb);
5676 }
5677}
5678static void show_rename_copy(struct diff_options *opt, const char *renamecopy,
5680 struct diff_filepair *p)
5681{
5682 struct strbuf sb = STRBUF_INIT;
5683 struct strbuf names = STRBUF_INIT;
5684 pprint_rename(&names, p->one->path, p->two->path);
5686 strbuf_addf(&sb, " %s %s (%d%%)\n",
5687 renamecopy, names.buf, similarity_index(p));
5688 strbuf_release(&names);
5689 emit_diff_symbol(opt, DIFF_SYMBOL_SUMMARY,
5690 sb.buf, sb.len, 0);
5691 show_mode_change(opt, p, 0);
5692 strbuf_release(&sb);
5693}
5694static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
5696{
5697 switch(p->status) {
5698 case DIFF_STATUS_DELETED:
5699 show_file_mode_name(opt, "delete", p->one);
5700 break;
5701 case DIFF_STATUS_ADDED:
5702 show_file_mode_name(opt, "create", p->two);
5703 break;
5704 case DIFF_STATUS_COPIED:
5705 show_rename_copy(opt, "copy", p);
5706 break;
5707 case DIFF_STATUS_RENAMED:
5708 show_rename_copy(opt, "rename", p);
5709 break;
5710 default:
5711 if (p->score) {
5712 struct strbuf sb = STRBUF_INIT;
5713 strbuf_addstr(&sb, " rewrite ");
5714 quote_c_style(p->two->path, &sb, NULL, 0);
5715 strbuf_addf(&sb, " (%d%%)\n", similarity_index(p));
5716 emit_diff_symbol(opt, DIFF_SYMBOL_SUMMARY,
5717 sb.buf, sb.len, 0);
5718 strbuf_release(&sb);
5719 }
5720 show_mode_change(opt, p, !p->score);
5721 break;
5722 }
5723}
5724struct patch_id_t {
5726 git_SHA_CTX *ctx;
5727 int patchlen;
5728};
5729static int remove_space(char *line, int len)
5731{
5732 int i;
5733 char *dst = line;
5734 unsigned char c;
5735 for (i = 0; i < len; i++)
5737 if (!isspace((c = line[i])))
5738 *dst++ = c;
5739 return dst - line;
5741}
5742static void patch_id_consume(void *priv, char *line, unsigned long len)
5744{
5745 struct patch_id_t *data = priv;
5746 int new_len;
5747 new_len = remove_space(line, len);
5749 git_SHA1_Update(data->ctx, line, new_len);
5751 data->patchlen += new_len;
5752}
5753static void patch_id_add_string(git_SHA_CTX *ctx, const char *str)
5755{
5756 git_SHA1_Update(ctx, str, strlen(str));
5757}
5758static void patch_id_add_mode(git_SHA_CTX *ctx, unsigned mode)
5760{
5761 /* large enough for 2^32 in octal */
5762 char buf[12];
5763 int len = xsnprintf(buf, sizeof(buf), "%06o", mode);
5764 git_SHA1_Update(ctx, buf, len);
5765}
5766/* returns 0 upon success, and writes result into sha1 */
5768static int diff_get_patch_id(struct diff_options *options, struct object_id *oid, int diff_header_only)
5769{
5770 struct diff_queue_struct *q = &diff_queued_diff;
5771 int i;
5772 git_SHA_CTX ctx;
5773 struct patch_id_t data;
5774 git_SHA1_Init(&ctx);
5776 memset(&data, 0, sizeof(struct patch_id_t));
5777 data.ctx = &ctx;
5778 for (i = 0; i < q->nr; i++) {
5780 xpparam_t xpp;
5781 xdemitconf_t xecfg;
5782 mmfile_t mf1, mf2;
5783 struct diff_filepair *p = q->queue[i];
5784 int len1, len2;
5785 memset(&xpp, 0, sizeof(xpp));
5787 memset(&xecfg, 0, sizeof(xecfg));
5788 if (p->status == 0)
5789 return error("internal diff status error");
5790 if (p->status == DIFF_STATUS_UNKNOWN)
5791 continue;
5792 if (diff_unmodified_pair(p))
5793 continue;
5794 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
5795 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
5796 continue;
5797 if (DIFF_PAIR_UNMERGED(p))
5798 continue;
5799 diff_fill_oid_info(p->one, options->repo->index);
5801 diff_fill_oid_info(p->two, options->repo->index);
5802 len1 = remove_space(p->one->path, strlen(p->one->path));
5804 len2 = remove_space(p->two->path, strlen(p->two->path));
5805 patch_id_add_string(&ctx, "diff--git");
5806 patch_id_add_string(&ctx, "a/");
5807 git_SHA1_Update(&ctx, p->one->path, len1);
5808 patch_id_add_string(&ctx, "b/");
5809 git_SHA1_Update(&ctx, p->two->path, len2);
5810 if (p->one->mode == 0) {
5812 patch_id_add_string(&ctx, "newfilemode");
5813 patch_id_add_mode(&ctx, p->two->mode);
5814 patch_id_add_string(&ctx, "---/dev/null");
5815 patch_id_add_string(&ctx, "+++b/");
5816 git_SHA1_Update(&ctx, p->two->path, len2);
5817 } else if (p->two->mode == 0) {
5818 patch_id_add_string(&ctx, "deletedfilemode");
5819 patch_id_add_mode(&ctx, p->one->mode);
5820 patch_id_add_string(&ctx, "---a/");
5821 git_SHA1_Update(&ctx, p->one->path, len1);
5822 patch_id_add_string(&ctx, "+++/dev/null");
5823 } else {
5824 patch_id_add_string(&ctx, "---a/");
5825 git_SHA1_Update(&ctx, p->one->path, len1);
5826 patch_id_add_string(&ctx, "+++b/");
5827 git_SHA1_Update(&ctx, p->two->path, len2);
5828 }
5829 if (diff_header_only)
5831 continue;
5832 if (fill_mmfile(options->repo, &mf1, p->one) < 0 ||
5834 fill_mmfile(options->repo, &mf2, p->two) < 0)
5835 return error("unable to read files to diff");
5836 if (diff_filespec_is_binary(options->repo, p->one) ||
5838 diff_filespec_is_binary(options->repo, p->two)) {
5839 git_SHA1_Update(&ctx, oid_to_hex(&p->one->oid),
5840 GIT_SHA1_HEXSZ);
5841 git_SHA1_Update(&ctx, oid_to_hex(&p->two->oid),
5842 GIT_SHA1_HEXSZ);
5843 continue;
5844 }
5845 xpp.flags = 0;
5847 xecfg.ctxlen = 3;
5848 xecfg.flags = 0;
5849 if (xdi_diff_outf(&mf1, &mf2, discard_hunk_line,
5850 patch_id_consume, &data, &xpp, &xecfg))
5851 return error("unable to generate patch-id diff for %s",
5852 p->one->path);
5853 }
5854 git_SHA1_Final(oid->hash, &ctx);
5856 return 0;
5857}
5858int diff_flush_patch_id(struct diff_options *options, struct object_id *oid, int diff_header_only)
5860{
5861 struct diff_queue_struct *q = &diff_queued_diff;
5862 int i;
5863 int result = diff_get_patch_id(options, oid, diff_header_only);
5864 for (i = 0; i < q->nr; i++)
5866 diff_free_filepair(q->queue[i]);
5867 free(q->queue);
5869 DIFF_QUEUE_CLEAR(q);
5870 return result;
5872}
5873static int is_summary_empty(const struct diff_queue_struct *q)
5875{
5876 int i;
5877 for (i = 0; i < q->nr; i++) {
5879 const struct diff_filepair *p = q->queue[i];
5880 switch (p->status) {
5882 case DIFF_STATUS_DELETED:
5883 case DIFF_STATUS_ADDED:
5884 case DIFF_STATUS_COPIED:
5885 case DIFF_STATUS_RENAMED:
5886 return 0;
5887 default:
5888 if (p->score)
5889 return 0;
5890 if (p->one->mode && p->two->mode &&
5891 p->one->mode != p->two->mode)
5892 return 0;
5893 break;
5894 }
5895 }
5896 return 1;
5897}
5898static const char rename_limit_warning[] =
5900N_("inexact rename detection was skipped due to too many files.");
5901static const char degrade_cc_to_c_warning[] =
5903N_("only found copies from modified paths due to too many files.");
5904static const char rename_limit_advice[] =
5906N_("you may want to set your %s variable to at least "
5907 "%d and retry the command.");
5908void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
5910{
5911 fflush(stdout);
5912 if (degraded_cc)
5913 warning(_(degrade_cc_to_c_warning));
5914 else if (needed)
5915 warning(_(rename_limit_warning));
5916 else
5917 return;
5918 if (0 < needed)
5919 warning(_(rename_limit_advice), varname, needed);
5920}
5921static void diff_flush_patch_all_file_pairs(struct diff_options *o)
5923{
5924 int i;
5925 static struct emitted_diff_symbols esm = EMITTED_DIFF_SYMBOLS_INIT;
5926 struct diff_queue_struct *q = &diff_queued_diff;
5927 if (WSEH_NEW & WS_RULE_MASK)
5929 BUG("WS rules bit mask overlaps with diff symbol flags");
5930 if (o->color_moved)
5932 o->emitted_symbols = &esm;
5933 for (i = 0; i < q->nr; i++) {
5935 struct diff_filepair *p = q->queue[i];
5936 if (check_pair_status(p))
5937 diff_flush_patch(p, o);
5938 }
5939 if (o->emitted_symbols) {
5941 if (o->color_moved) {
5942 struct hashmap add_lines, del_lines;
5943 if (o->color_moved_ws_handling &
5945 COLOR_MOVED_WS_ALLOW_INDENTATION_CHANGE)
5946 o->color_moved_ws_handling |= XDF_IGNORE_WHITESPACE;
5947 hashmap_init(&del_lines, moved_entry_cmp, o, 0);
5949 hashmap_init(&add_lines, moved_entry_cmp, o, 0);
5950 add_lines_to_move_detection(o, &add_lines, &del_lines);
5952 mark_color_as_moved(o, &add_lines, &del_lines);
5953 if (o->color_moved == COLOR_MOVED_ZEBRA_DIM)
5954 dim_moved_lines(o);
5955 hashmap_free(&add_lines, 1);
5957 hashmap_free(&del_lines, 1);
5958 }
5959 for (i = 0; i < esm.nr; i++)
5961 emit_diff_symbol_from_struct(o, &esm.buf[i]);
5962 for (i = 0; i < esm.nr; i++)
5964 free((void *)esm.buf[i].line);
5965 esm.nr = 0;
5966 o->emitted_symbols = NULL;
5968 }
5969}
5970void diff_flush(struct diff_options *options)
5972{
5973 struct diff_queue_struct *q = &diff_queued_diff;
5974 int i, output_format = options->output_format;
5975 int separator = 0;
5976 int dirstat_by_line = 0;
5977 /*
5979 * Order: raw, stat, summary, patch
5980 * or: name/name-status/checkdiff (other bits clear)
5981 */
5982 if (!q->nr)
5983 goto free_queue;
5984 if (output_format & (DIFF_FORMAT_RAW |
5986 DIFF_FORMAT_NAME |
5987 DIFF_FORMAT_NAME_STATUS |
5988 DIFF_FORMAT_CHECKDIFF)) {
5989 for (i = 0; i < q->nr; i++) {
5990 struct diff_filepair *p = q->queue[i];
5991 if (check_pair_status(p))
5992 flush_one_pair(p, options);
5993 }
5994 separator++;
5995 }
5996 if (output_format & DIFF_FORMAT_DIRSTAT && options->flags.dirstat_by_line)
5998 dirstat_by_line = 1;
5999 if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
6001 dirstat_by_line) {
6002 struct diffstat_t diffstat;
6003 memset(&diffstat, 0, sizeof(struct diffstat_t));
6005 for (i = 0; i < q->nr; i++) {
6006 struct diff_filepair *p = q->queue[i];
6007 if (check_pair_status(p))
6008 diff_flush_stat(p, options, &diffstat);
6009 }
6010 if (output_format & DIFF_FORMAT_NUMSTAT)
6011 show_numstat(&diffstat, options);
6012 if (output_format & DIFF_FORMAT_DIFFSTAT)
6013 show_stats(&diffstat, options);
6014 if (output_format & DIFF_FORMAT_SHORTSTAT)
6015 show_shortstats(&diffstat, options);
6016 if (output_format & DIFF_FORMAT_DIRSTAT && dirstat_by_line)
6017 show_dirstat_by_line(&diffstat, options);
6018 free_diffstat_info(&diffstat);
6019 separator++;
6020 }
6021 if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
6022 show_dirstat(options);
6023 if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
6025 for (i = 0; i < q->nr; i++) {
6026 diff_summary(options, q->queue[i]);
6027 }
6028 separator++;
6029 }
6030 if (output_format & DIFF_FORMAT_NO_OUTPUT &&
6032 options->flags.exit_with_status &&
6033 options->flags.diff_from_contents) {
6034 /*
6035 * run diff_flush_patch for the exit status. setting
6036 * options->file to /dev/null should be safe, because we
6037 * aren't supposed to produce any output anyway.
6038 */
6039 if (options->close_file)
6040 fclose(options->file);
6041 options->file = xfopen("/dev/null", "w");
6042 options->close_file = 1;
6043 options->color_moved = 0;
6044 for (i = 0; i < q->nr; i++) {
6045 struct diff_filepair *p = q->queue[i];
6046 if (check_pair_status(p))
6047 diff_flush_patch(p, options);
6048 if (options->found_changes)
6049 break;
6050 }
6051 }
6052 if (output_format & DIFF_FORMAT_PATCH) {
6054 if (separator) {
6055 emit_diff_symbol(options, DIFF_SYMBOL_SEPARATOR, NULL, 0, 0);
6056 if (options->stat_sep)
6057 /* attach patch instead of inline */
6058 emit_diff_symbol(options, DIFF_SYMBOL_STAT_SEP,
6059 NULL, 0, 0);
6060 }
6061 diff_flush_patch_all_file_pairs(options);
6063 }
6064 if (output_format & DIFF_FORMAT_CALLBACK)
6066 options->format_callback(q, options, options->format_callback_data);
6067 for (i = 0; i < q->nr; i++)
6069 diff_free_filepair(q->queue[i]);
6070free_queue:
6071 free(q->queue);
6072 DIFF_QUEUE_CLEAR(q);
6073 if (options->close_file)
6074 fclose(options->file);
6075 /*
6077 * Report the content-level differences with HAS_CHANGES;
6078 * diff_addremove/diff_change does not set the bit when
6079 * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
6080 */
6081 if (options->flags.diff_from_contents) {
6082 if (options->found_changes)
6083 options->flags.has_changes = 1;
6084 else
6085 options->flags.has_changes = 0;
6086 }
6087}
6088static int match_filter(const struct diff_options *options, const struct diff_filepair *p)
6090{
6091 return (((p->status == DIFF_STATUS_MODIFIED) &&
6092 ((p->score &&
6093 filter_bit_tst(DIFF_STATUS_FILTER_BROKEN, options)) ||
6094 (!p->score &&
6095 filter_bit_tst(DIFF_STATUS_MODIFIED, options)))) ||
6096 ((p->status != DIFF_STATUS_MODIFIED) &&
6097 filter_bit_tst(p->status, options)));
6098}
6099static void diffcore_apply_filter(struct diff_options *options)
6101{
6102 int i;
6103 struct diff_queue_struct *q = &diff_queued_diff;
6104 struct diff_queue_struct outq;
6105 DIFF_QUEUE_CLEAR(&outq);
6107 if (!options->filter)
6109 return;
6110 if (filter_bit_tst(DIFF_STATUS_FILTER_AON, options)) {
6112 int found;
6113 for (i = found = 0; !found && i < q->nr; i++) {
6114 if (match_filter(options, q->queue[i]))
6115 found++;
6116 }
6117 if (found)
6118 return;
6119 /* otherwise we will clear the whole queue
6121 * by copying the empty outq at the end of this
6122 * function, but first clear the current entries
6123 * in the queue.
6124 */
6125 for (i = 0; i < q->nr; i++)
6126 diff_free_filepair(q->queue[i]);
6127 }
6128 else {
6129 /* Only the matching ones */
6130 for (i = 0; i < q->nr; i++) {
6131 struct diff_filepair *p = q->queue[i];
6132 if (match_filter(options, p))
6133 diff_q(&outq, p);
6134 else
6135 diff_free_filepair(p);
6136 }
6137 }
6138 free(q->queue);
6139 *q = outq;
6140}
6141/* Check whether two filespecs with the same mode and size are identical */
6143static int diff_filespec_is_identical(struct repository *r,
6144 struct diff_filespec *one,
6145 struct diff_filespec *two)
6146{
6147 if (S_ISGITLINK(one->mode))
6148 return 0;
6149 if (diff_populate_filespec(r, one, 0))
6150 return 0;
6151 if (diff_populate_filespec(r, two, 0))
6152 return 0;
6153 return !memcmp(one->data, two->data, one->size);
6154}
6155static int diff_filespec_check_stat_unmatch(struct repository *r,
6157 struct diff_filepair *p)
6158{
6159 if (p->done_skip_stat_unmatch)
6160 return p->skip_stat_unmatch_result;
6161 p->done_skip_stat_unmatch = 1;
6163 p->skip_stat_unmatch_result = 0;
6164 /*
6165 * 1. Entries that come from stat info dirtiness
6166 * always have both sides (iow, not create/delete),
6167 * one side of the object name is unknown, with
6168 * the same mode and size. Keep the ones that
6169 * do not match these criteria. They have real
6170 * differences.
6171 *
6172 * 2. At this point, the file is known to be modified,
6173 * with the same mode and size, and the object
6174 * name of one side is unknown. Need to inspect
6175 * the identical contents.
6176 */
6177 if (!DIFF_FILE_VALID(p->one) || /* (1) */
6178 !DIFF_FILE_VALID(p->two) ||
6179 (p->one->oid_valid && p->two->oid_valid) ||
6180 (p->one->mode != p->two->mode) ||
6181 diff_populate_filespec(r, p->one, CHECK_SIZE_ONLY) ||
6182 diff_populate_filespec(r, p->two, CHECK_SIZE_ONLY) ||
6183 (p->one->size != p->two->size) ||
6184 !diff_filespec_is_identical(r, p->one, p->two)) /* (2) */
6185 p->skip_stat_unmatch_result = 1;
6186 return p->skip_stat_unmatch_result;
6187}
6188static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
6190{
6191 int i;
6192 struct diff_queue_struct *q = &diff_queued_diff;
6193 struct diff_queue_struct outq;
6194 DIFF_QUEUE_CLEAR(&outq);
6195 for (i = 0; i < q->nr; i++) {
6197 struct diff_filepair *p = q->queue[i];
6198 if (diff_filespec_check_stat_unmatch(diffopt->repo, p))
6200 diff_q(&outq, p);
6201 else {
6202 /*
6203 * The caller can subtract 1 from skip_stat_unmatch
6204 * to determine how many paths were dirty only
6205 * due to stat info mismatch.
6206 */
6207 if (!diffopt->flags.no_index)
6208 diffopt->skip_stat_unmatch++;
6209 diff_free_filepair(p);
6210 }
6211 }
6212 free(q->queue);
6213 *q = outq;
6214}
6215static int diffnamecmp(const void *a_, const void *b_)
6217{
6218 const struct diff_filepair *a = *((const struct diff_filepair **)a_);
6219 const struct diff_filepair *b = *((const struct diff_filepair **)b_);
6220 const char *name_a, *name_b;
6221 name_a = a->one ? a->one->path : a->two->path;
6223 name_b = b->one ? b->one->path : b->two->path;
6224 return strcmp(name_a, name_b);
6225}
6226void diffcore_fix_diff_index(struct diff_options *options)
6228{
6229 struct diff_queue_struct *q = &diff_queued_diff;
6230 QSORT(q->queue, q->nr, diffnamecmp);
6231}
6232void diffcore_std(struct diff_options *options)
6234{
6235 /* NOTE please keep the following in sync with diff_tree_combined() */
6236 if (options->skip_stat_unmatch)
6237 diffcore_skip_stat_unmatch(options);
6238 if (!options->found_follow) {
6239 /* See try_to_follow_renames() in tree-diff.c */
6240 if (options->break_opt != -1)
6241 diffcore_break(options->repo,
6242 options->break_opt);
6243 if (options->detect_rename)
6244 diffcore_rename(options);
6245 if (options->break_opt != -1)
6246 diffcore_merge_broken();
6247 }
6248 if (options->pickaxe_opts & DIFF_PICKAXE_KINDS_MASK)
6249 diffcore_pickaxe(options);
6250 if (options->orderfile)
6251 diffcore_order(options->orderfile);
6252 if (!options->found_follow)
6253 /* See try_to_follow_renames() in tree-diff.c */
6254 diff_resolve_rename_copy();
6255 diffcore_apply_filter(options);
6256 if (diff_queued_diff.nr && !options->flags.diff_from_contents)
6258 options->flags.has_changes = 1;
6259 else
6260 options->flags.has_changes = 0;
6261 options->found_follow = 0;
6263}
6264int diff_result_code(struct diff_options *opt, int status)
6266{
6267 int result = 0;
6268 diff_warn_rename_limit("diff.renameLimit",
6270 opt->needed_rename_limit,
6271 opt->degraded_cc_to_c);
6272 if (!opt->flags.exit_with_status &&
6273 !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
6274 return status;
6275 if (opt->flags.exit_with_status &&
6276 opt->flags.has_changes)
6277 result |= 01;
6278 if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
6279 opt->flags.check_failed)
6280 result |= 02;
6281 return result;
6282}
6283int diff_can_quit_early(struct diff_options *opt)
6285{
6286 return (opt->flags.quick &&
6287 !opt->filter &&
6288 opt->flags.has_changes);
6289}
6290/*
6292 * Shall changes to this submodule be ignored?
6293 *
6294 * Submodule changes can be configured to be ignored separately for each path,
6295 * but that configuration can be overridden from the command line.
6296 */
6297static int is_submodule_ignored(const char *path, struct diff_options *options)
6298{
6299 int ignored = 0;
6300 struct diff_flags orig_flags = options->flags;
6301 if (!options->flags.override_submodule_config)
6302 set_diffopt_flags_from_submodule_config(options, path);
6303 if (options->flags.ignore_submodules)
6304 ignored = 1;
6305 options->flags = orig_flags;
6306 return ignored;
6307}
6308void diff_addremove(struct diff_options *options,
6310 int addremove, unsigned mode,
6311 const struct object_id *oid,
6312 int oid_valid,
6313 const char *concatpath, unsigned dirty_submodule)
6314{
6315 struct diff_filespec *one, *two;
6316 if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
6318 return;
6319 /* This may look odd, but it is a preparation for
6321 * feeding "there are unchanged files which should
6322 * not produce diffs, but when you are doing copy
6323 * detection you would need them, so here they are"
6324 * entries to the diff-core. They will be prefixed
6325 * with something like '=' or '*' (I haven't decided
6326 * which but should not make any difference).
6327 * Feeding the same new and old to diff_change()
6328 * also has the same effect.
6329 * Before the final output happens, they are pruned after
6330 * merged into rename/copy pairs as appropriate.
6331 */
6332 if (options->flags.reverse_diff)
6333 addremove = (addremove == '+' ? '-' :
6334 addremove == '-' ? '+' : addremove);
6335 if (options->prefix &&
6337 strncmp(concatpath, options->prefix, options->prefix_length))
6338 return;
6339 one = alloc_filespec(concatpath);
6341 two = alloc_filespec(concatpath);
6342 if (addremove != '+')
6344 fill_filespec(one, oid, oid_valid, mode);
6345 if (addremove != '-') {
6346 fill_filespec(two, oid, oid_valid, mode);
6347 two->dirty_submodule = dirty_submodule;
6348 }
6349 diff_queue(&diff_queued_diff, one, two);
6351 if (!options->flags.diff_from_contents)
6352 options->flags.has_changes = 1;
6353}
6354void diff_change(struct diff_options *options,
6356 unsigned old_mode, unsigned new_mode,
6357 const struct object_id *old_oid,
6358 const struct object_id *new_oid,
6359 int old_oid_valid, int new_oid_valid,
6360 const char *concatpath,
6361 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
6362{
6363 struct diff_filespec *one, *two;
6364 struct diff_filepair *p;
6365 if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
6367 is_submodule_ignored(concatpath, options))
6368 return;
6369 if (options->flags.reverse_diff) {
6371 SWAP(old_mode, new_mode);
6372 SWAP(old_oid, new_oid);
6373 SWAP(old_oid_valid, new_oid_valid);
6374 SWAP(old_dirty_submodule, new_dirty_submodule);
6375 }
6376 if (options->prefix &&
6378 strncmp(concatpath, options->prefix, options->prefix_length))
6379 return;
6380 one = alloc_filespec(concatpath);
6382 two = alloc_filespec(concatpath);
6383 fill_filespec(one, old_oid, old_oid_valid, old_mode);
6384 fill_filespec(two, new_oid, new_oid_valid, new_mode);
6385 one->dirty_submodule = old_dirty_submodule;
6386 two->dirty_submodule = new_dirty_submodule;
6387 p = diff_queue(&diff_queued_diff, one, two);
6388 if (options->flags.diff_from_contents)
6390 return;
6391 if (options->flags.quick && options->skip_stat_unmatch &&
6393 !diff_filespec_check_stat_unmatch(options->repo, p))
6394 return;
6395 options->flags.has_changes = 1;
6397}
6398struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
6400{
6401 struct diff_filepair *pair;
6402 struct diff_filespec *one, *two;
6403 if (options->prefix &&
6405 strncmp(path, options->prefix, options->prefix_length))
6406 return NULL;
6407 one = alloc_filespec(path);
6409 two = alloc_filespec(path);
6410 pair = diff_queue(&diff_queued_diff, one, two);
6411 pair->is_unmerged = 1;
6412 return pair;
6413}
6414static char *run_textconv(struct repository *r,
6416 const char *pgm,
6417 struct diff_filespec *spec,
6418 size_t *outsize)
6419{
6420 struct diff_tempfile *temp;
6421 const char *argv[3];
6422 const char **arg = argv;
6423 struct child_process child = CHILD_PROCESS_INIT;
6424 struct strbuf buf = STRBUF_INIT;
6425 int err = 0;
6426 temp = prepare_temp_file(r, spec->path, spec);
6428 *arg++ = pgm;
6429 *arg++ = temp->name;
6430 *arg = NULL;
6431 child.use_shell = 1;
6433 child.argv = argv;
6434 child.out = -1;
6435 if (start_command(&child)) {
6436 remove_tempfile();
6437 return NULL;
6438 }
6439 if (strbuf_read(&buf, child.out, 0) < 0)
6441 err = error("error reading from textconv command '%s'", pgm);
6442 close(child.out);
6443 if (finish_command(&child) || err) {
6445 strbuf_release(&buf);
6446 remove_tempfile();
6447 return NULL;
6448 }
6449 remove_tempfile();
6450 return strbuf_detach(&buf, outsize);
6452}
6453size_t fill_textconv(struct repository *r,
6455 struct userdiff_driver *driver,
6456 struct diff_filespec *df,
6457 char **outbuf)
6458{
6459 size_t size;
6460 if (!driver) {
6462 if (!DIFF_FILE_VALID(df)) {
6463 *outbuf = "";
6464 return 0;
6465 }
6466 if (diff_populate_filespec(r, df, 0))
6467 die("unable to read files to diff");
6468 *outbuf = df->data;
6469 return df->size;
6470 }
6471 if (!driver->textconv)
6473 BUG("fill_textconv called with non-textconv driver");
6474 if (driver->textconv_cache && df->oid_valid) {
6476 *outbuf = notes_cache_get(driver->textconv_cache,
6477 &df->oid,
6478 &size);
6479 if (*outbuf)
6480 return size;
6481 }
6482 *outbuf = run_textconv(r, driver->textconv, df, &size);
6484 if (!*outbuf)
6485 die("unable to read files to diff");
6486 if (driver->textconv_cache && df->oid_valid) {
6488 /* ignore errors, as we might be in a readonly repository */
6489 notes_cache_put(driver->textconv_cache, &df->oid, *outbuf,
6490 size);
6491 /*
6492 * we could save up changes and flush them all at the end,
6493 * but we would need an extra call after all diffing is done.
6494 * Since generating a cache entry is the slow path anyway,
6495 * this extra overhead probably isn't a big deal.
6496 */
6497 notes_cache_write(driver->textconv_cache);
6498 }
6499 return size;
6501}
6502int textconv_object(struct repository *r,
6504 const char *path,
6505 unsigned mode,
6506 const struct object_id *oid,
6507 int oid_valid,
6508 char **buf,
6509 unsigned long *buf_size)
6510{
6511 struct diff_filespec *df;
6512 struct userdiff_driver *textconv;
6513 df = alloc_filespec(path);
6515 fill_filespec(df, oid, oid_valid, mode);
6516 textconv = get_textconv(r, df);
6517 if (!textconv) {
6518 free_filespec(df);
6519 return 0;
6520 }
6521 *buf_size = fill_textconv(r, textconv, df, buf);
6523 free_filespec(df);
6524 return 1;
6525}
6526void setup_diff_pager(struct diff_options *opt)
6528{
6529 /*
6530 * If the user asked for our exit code, then either they want --quiet
6531 * or --exit-code. We should definitely not bother with a pager in the
6532 * former case, as we will generate no output. Since we still properly
6533 * report our exit code even when a pager is run, we _could_ run a
6534 * pager with --exit-code. But since we have not done so historically,
6535 * and because it is easy to find people oneline advising "git diff
6536 * --exit-code" in hooks and other scripts, we do not do so.
6537 */
6538 if (!opt->flags.exit_with_status &&
6539 check_pager_config("diff") != 0)
6540 setup_pager();
6541}