1/*
2 * Copyright (C) 2005 Junio C Hamano
3 */
4#include "cache.h"
5#include "config.h"
6#include "tempfile.h"
7#include "quote.h"
8#include "diff.h"
9#include "diffcore.h"
10#include "delta.h"
11#include "xdiff-interface.h"
12#include "color.h"
13#include "attr.h"
14#include "run-command.h"
15#include "utf8.h"
16#include "userdiff.h"
17#include "submodule-config.h"
18#include "submodule.h"
19#include "ll-merge.h"
20#include "string-list.h"
21#include "argv-array.h"
22#include "graph.h"
23
24#ifdef NO_FAST_WORKING_DIRECTORY
25#define FAST_WORKING_DIRECTORY 0
26#else
27#define FAST_WORKING_DIRECTORY 1
28#endif
29
30static int diff_detect_rename_default;
31static int diff_indent_heuristic = 1;
32static int diff_rename_limit_default = 400;
33static int diff_suppress_blank_empty;
34static int diff_use_color_default = -1;
35static int diff_context_default = 3;
36static int diff_interhunk_context_default;
37static const char *diff_word_regex_cfg;
38static const char *external_diff_cmd_cfg;
39static const char *diff_order_file_cfg;
40int diff_auto_refresh_index = 1;
41static int diff_mnemonic_prefix;
42static int diff_no_prefix;
43static int diff_stat_graph_width;
44static int diff_dirstat_permille_default = 30;
45static struct diff_options default_diff_options;
46static long diff_algorithm;
47static unsigned ws_error_highlight_default = WSEH_NEW;
48
49static char diff_colors[][COLOR_MAXLEN] = {
50 GIT_COLOR_RESET,
51 GIT_COLOR_NORMAL, /* CONTEXT */
52 GIT_COLOR_BOLD, /* METAINFO */
53 GIT_COLOR_CYAN, /* FRAGINFO */
54 GIT_COLOR_RED, /* OLD */
55 GIT_COLOR_GREEN, /* NEW */
56 GIT_COLOR_YELLOW, /* COMMIT */
57 GIT_COLOR_BG_RED, /* WHITESPACE */
58 GIT_COLOR_NORMAL, /* FUNCINFO */
59};
60
61static NORETURN void die_want_option(const char *option_name)
62{
63 die(_("option '%s' requires a value"), option_name);
64}
65
66static int parse_diff_color_slot(const char *var)
67{
68 if (!strcasecmp(var, "context") || !strcasecmp(var, "plain"))
69 return DIFF_CONTEXT;
70 if (!strcasecmp(var, "meta"))
71 return DIFF_METAINFO;
72 if (!strcasecmp(var, "frag"))
73 return DIFF_FRAGINFO;
74 if (!strcasecmp(var, "old"))
75 return DIFF_FILE_OLD;
76 if (!strcasecmp(var, "new"))
77 return DIFF_FILE_NEW;
78 if (!strcasecmp(var, "commit"))
79 return DIFF_COMMIT;
80 if (!strcasecmp(var, "whitespace"))
81 return DIFF_WHITESPACE;
82 if (!strcasecmp(var, "func"))
83 return DIFF_FUNCINFO;
84 return -1;
85}
86
87static int parse_dirstat_params(struct diff_options *options, const char *params_string,
88 struct strbuf *errmsg)
89{
90 char *params_copy = xstrdup(params_string);
91 struct string_list params = STRING_LIST_INIT_NODUP;
92 int ret = 0;
93 int i;
94
95 if (*params_copy)
96 string_list_split_in_place(¶ms, params_copy, ',', -1);
97 for (i = 0; i < params.nr; i++) {
98 const char *p = params.items[i].string;
99 if (!strcmp(p, "changes")) {
100 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
101 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
102 } else if (!strcmp(p, "lines")) {
103 DIFF_OPT_SET(options, DIRSTAT_BY_LINE);
104 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
105 } else if (!strcmp(p, "files")) {
106 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
107 DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
108 } else if (!strcmp(p, "noncumulative")) {
109 DIFF_OPT_CLR(options, DIRSTAT_CUMULATIVE);
110 } else if (!strcmp(p, "cumulative")) {
111 DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
112 } else if (isdigit(*p)) {
113 char *end;
114 int permille = strtoul(p, &end, 10) * 10;
115 if (*end == '.' && isdigit(*++end)) {
116 /* only use first digit */
117 permille += *end - '0';
118 /* .. and ignore any further digits */
119 while (isdigit(*++end))
120 ; /* nothing */
121 }
122 if (!*end)
123 options->dirstat_permille = permille;
124 else {
125 strbuf_addf(errmsg, _(" Failed to parse dirstat cut-off percentage '%s'\n"),
126 p);
127 ret++;
128 }
129 } else {
130 strbuf_addf(errmsg, _(" Unknown dirstat parameter '%s'\n"), p);
131 ret++;
132 }
133
134 }
135 string_list_clear(¶ms, 0);
136 free(params_copy);
137 return ret;
138}
139
140static int parse_submodule_params(struct diff_options *options, const char *value)
141{
142 if (!strcmp(value, "log"))
143 options->submodule_format = DIFF_SUBMODULE_LOG;
144 else if (!strcmp(value, "short"))
145 options->submodule_format = DIFF_SUBMODULE_SHORT;
146 else if (!strcmp(value, "diff"))
147 options->submodule_format = DIFF_SUBMODULE_INLINE_DIFF;
148 else
149 return -1;
150 return 0;
151}
152
153static int git_config_rename(const char *var, const char *value)
154{
155 if (!value)
156 return DIFF_DETECT_RENAME;
157 if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
158 return DIFF_DETECT_COPY;
159 return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
160}
161
162long parse_algorithm_value(const char *value)
163{
164 if (!value)
165 return -1;
166 else if (!strcasecmp(value, "myers") || !strcasecmp(value, "default"))
167 return 0;
168 else if (!strcasecmp(value, "minimal"))
169 return XDF_NEED_MINIMAL;
170 else if (!strcasecmp(value, "patience"))
171 return XDF_PATIENCE_DIFF;
172 else if (!strcasecmp(value, "histogram"))
173 return XDF_HISTOGRAM_DIFF;
174 return -1;
175}
176
177static int parse_one_token(const char **arg, const char *token)
178{
179 const char *rest;
180 if (skip_prefix(*arg, token, &rest) && (!*rest || *rest == ',')) {
181 *arg = rest;
182 return 1;
183 }
184 return 0;
185}
186
187static int parse_ws_error_highlight(const char *arg)
188{
189 const char *orig_arg = arg;
190 unsigned val = 0;
191
192 while (*arg) {
193 if (parse_one_token(&arg, "none"))
194 val = 0;
195 else if (parse_one_token(&arg, "default"))
196 val = WSEH_NEW;
197 else if (parse_one_token(&arg, "all"))
198 val = WSEH_NEW | WSEH_OLD | WSEH_CONTEXT;
199 else if (parse_one_token(&arg, "new"))
200 val |= WSEH_NEW;
201 else if (parse_one_token(&arg, "old"))
202 val |= WSEH_OLD;
203 else if (parse_one_token(&arg, "context"))
204 val |= WSEH_CONTEXT;
205 else {
206 return -1 - (int)(arg - orig_arg);
207 }
208 if (*arg)
209 arg++;
210 }
211 return val;
212}
213
214/*
215 * These are to give UI layer defaults.
216 * The core-level commands such as git-diff-files should
217 * never be affected by the setting of diff.renames
218 * the user happens to have in the configuration file.
219 */
220void init_diff_ui_defaults(void)
221{
222 diff_detect_rename_default = 1;
223}
224
225int git_diff_heuristic_config(const char *var, const char *value, void *cb)
226{
227 if (!strcmp(var, "diff.indentheuristic"))
228 diff_indent_heuristic = git_config_bool(var, value);
229 return 0;
230}
231
232int git_diff_ui_config(const char *var, const char *value, void *cb)
233{
234 if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
235 diff_use_color_default = git_config_colorbool(var, value);
236 return 0;
237 }
238 if (!strcmp(var, "diff.context")) {
239 diff_context_default = git_config_int(var, value);
240 if (diff_context_default < 0)
241 return -1;
242 return 0;
243 }
244 if (!strcmp(var, "diff.interhunkcontext")) {
245 diff_interhunk_context_default = git_config_int(var, value);
246 if (diff_interhunk_context_default < 0)
247 return -1;
248 return 0;
249 }
250 if (!strcmp(var, "diff.renames")) {
251 diff_detect_rename_default = git_config_rename(var, value);
252 return 0;
253 }
254 if (!strcmp(var, "diff.autorefreshindex")) {
255 diff_auto_refresh_index = git_config_bool(var, value);
256 return 0;
257 }
258 if (!strcmp(var, "diff.mnemonicprefix")) {
259 diff_mnemonic_prefix = git_config_bool(var, value);
260 return 0;
261 }
262 if (!strcmp(var, "diff.noprefix")) {
263 diff_no_prefix = git_config_bool(var, value);
264 return 0;
265 }
266 if (!strcmp(var, "diff.statgraphwidth")) {
267 diff_stat_graph_width = git_config_int(var, value);
268 return 0;
269 }
270 if (!strcmp(var, "diff.external"))
271 return git_config_string(&external_diff_cmd_cfg, var, value);
272 if (!strcmp(var, "diff.wordregex"))
273 return git_config_string(&diff_word_regex_cfg, var, value);
274 if (!strcmp(var, "diff.orderfile"))
275 return git_config_pathname(&diff_order_file_cfg, var, value);
276
277 if (!strcmp(var, "diff.ignoresubmodules"))
278 handle_ignore_submodules_arg(&default_diff_options, value);
279
280 if (!strcmp(var, "diff.submodule")) {
281 if (parse_submodule_params(&default_diff_options, value))
282 warning(_("Unknown value for 'diff.submodule' config variable: '%s'"),
283 value);
284 return 0;
285 }
286
287 if (!strcmp(var, "diff.algorithm")) {
288 diff_algorithm = parse_algorithm_value(value);
289 if (diff_algorithm < 0)
290 return -1;
291 return 0;
292 }
293
294 if (!strcmp(var, "diff.wserrorhighlight")) {
295 int val = parse_ws_error_highlight(value);
296 if (val < 0)
297 return -1;
298 ws_error_highlight_default = val;
299 return 0;
300 }
301
302 if (git_color_config(var, value, cb) < 0)
303 return -1;
304
305 return git_diff_basic_config(var, value, cb);
306}
307
308int git_diff_basic_config(const char *var, const char *value, void *cb)
309{
310 const char *name;
311
312 if (!strcmp(var, "diff.renamelimit")) {
313 diff_rename_limit_default = git_config_int(var, value);
314 return 0;
315 }
316
317 if (userdiff_config(var, value) < 0)
318 return -1;
319
320 if (skip_prefix(var, "diff.color.", &name) ||
321 skip_prefix(var, "color.diff.", &name)) {
322 int slot = parse_diff_color_slot(name);
323 if (slot < 0)
324 return 0;
325 if (!value)
326 return config_error_nonbool(var);
327 return color_parse(value, diff_colors[slot]);
328 }
329
330 /* like GNU diff's --suppress-blank-empty option */
331 if (!strcmp(var, "diff.suppressblankempty") ||
332 /* for backwards compatibility */
333 !strcmp(var, "diff.suppress-blank-empty")) {
334 diff_suppress_blank_empty = git_config_bool(var, value);
335 return 0;
336 }
337
338 if (!strcmp(var, "diff.dirstat")) {
339 struct strbuf errmsg = STRBUF_INIT;
340 default_diff_options.dirstat_permille = diff_dirstat_permille_default;
341 if (parse_dirstat_params(&default_diff_options, value, &errmsg))
342 warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
343 errmsg.buf);
344 strbuf_release(&errmsg);
345 diff_dirstat_permille_default = default_diff_options.dirstat_permille;
346 return 0;
347 }
348
349 if (starts_with(var, "submodule."))
350 return parse_submodule_config_option(var, value);
351
352 if (git_diff_heuristic_config(var, value, cb) < 0)
353 return -1;
354
355 return git_default_config(var, value, cb);
356}
357
358static char *quote_two(const char *one, const char *two)
359{
360 int need_one = quote_c_style(one, NULL, NULL, 1);
361 int need_two = quote_c_style(two, NULL, NULL, 1);
362 struct strbuf res = STRBUF_INIT;
363
364 if (need_one + need_two) {
365 strbuf_addch(&res, '"');
366 quote_c_style(one, &res, NULL, 1);
367 quote_c_style(two, &res, NULL, 1);
368 strbuf_addch(&res, '"');
369 } else {
370 strbuf_addstr(&res, one);
371 strbuf_addstr(&res, two);
372 }
373 return strbuf_detach(&res, NULL);
374}
375
376static const char *external_diff(void)
377{
378 static const char *external_diff_cmd = NULL;
379 static int done_preparing = 0;
380
381 if (done_preparing)
382 return external_diff_cmd;
383 external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
384 if (!external_diff_cmd)
385 external_diff_cmd = external_diff_cmd_cfg;
386 done_preparing = 1;
387 return external_diff_cmd;
388}
389
390/*
391 * Keep track of files used for diffing. Sometimes such an entry
392 * refers to a temporary file, sometimes to an existing file, and
393 * sometimes to "/dev/null".
394 */
395static struct diff_tempfile {
396 /*
397 * filename external diff should read from, or NULL if this
398 * entry is currently not in use:
399 */
400 const char *name;
401
402 char hex[GIT_MAX_HEXSZ + 1];
403 char mode[10];
404
405 /*
406 * If this diff_tempfile instance refers to a temporary file,
407 * this tempfile object is used to manage its lifetime.
408 */
409 struct tempfile tempfile;
410} diff_temp[2];
411
412typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
413
414struct emit_callback {
415 int color_diff;
416 unsigned ws_rule;
417 int blank_at_eof_in_preimage;
418 int blank_at_eof_in_postimage;
419 int lno_in_preimage;
420 int lno_in_postimage;
421 sane_truncate_fn truncate;
422 const char **label_path;
423 struct diff_words_data *diff_words;
424 struct diff_options *opt;
425 struct strbuf *header;
426};
427
428static int count_lines(const char *data, int size)
429{
430 int count, ch, completely_empty = 1, nl_just_seen = 0;
431 count = 0;
432 while (0 < size--) {
433 ch = *data++;
434 if (ch == '\n') {
435 count++;
436 nl_just_seen = 1;
437 completely_empty = 0;
438 }
439 else {
440 nl_just_seen = 0;
441 completely_empty = 0;
442 }
443 }
444 if (completely_empty)
445 return 0;
446 if (!nl_just_seen)
447 count++; /* no trailing newline */
448 return count;
449}
450
451static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
452{
453 if (!DIFF_FILE_VALID(one)) {
454 mf->ptr = (char *)""; /* does not matter */
455 mf->size = 0;
456 return 0;
457 }
458 else if (diff_populate_filespec(one, 0))
459 return -1;
460
461 mf->ptr = one->data;
462 mf->size = one->size;
463 return 0;
464}
465
466/* like fill_mmfile, but only for size, so we can avoid retrieving blob */
467static unsigned long diff_filespec_size(struct diff_filespec *one)
468{
469 if (!DIFF_FILE_VALID(one))
470 return 0;
471 diff_populate_filespec(one, CHECK_SIZE_ONLY);
472 return one->size;
473}
474
475static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
476{
477 char *ptr = mf->ptr;
478 long size = mf->size;
479 int cnt = 0;
480
481 if (!size)
482 return cnt;
483 ptr += size - 1; /* pointing at the very end */
484 if (*ptr != '\n')
485 ; /* incomplete line */
486 else
487 ptr--; /* skip the last LF */
488 while (mf->ptr < ptr) {
489 char *prev_eol;
490 for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
491 if (*prev_eol == '\n')
492 break;
493 if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
494 break;
495 cnt++;
496 ptr = prev_eol - 1;
497 }
498 return cnt;
499}
500
501static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
502 struct emit_callback *ecbdata)
503{
504 int l1, l2, at;
505 unsigned ws_rule = ecbdata->ws_rule;
506 l1 = count_trailing_blank(mf1, ws_rule);
507 l2 = count_trailing_blank(mf2, ws_rule);
508 if (l2 <= l1) {
509 ecbdata->blank_at_eof_in_preimage = 0;
510 ecbdata->blank_at_eof_in_postimage = 0;
511 return;
512 }
513 at = count_lines(mf1->ptr, mf1->size);
514 ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
515
516 at = count_lines(mf2->ptr, mf2->size);
517 ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
518}
519
520static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
521 int first, const char *line, int len)
522{
523 int has_trailing_newline, has_trailing_carriage_return;
524 int nofirst;
525 FILE *file = o->file;
526
527 fputs(diff_line_prefix(o), file);
528
529 if (len == 0) {
530 has_trailing_newline = (first == '\n');
531 has_trailing_carriage_return = (!has_trailing_newline &&
532 (first == '\r'));
533 nofirst = has_trailing_newline || has_trailing_carriage_return;
534 } else {
535 has_trailing_newline = (len > 0 && line[len-1] == '\n');
536 if (has_trailing_newline)
537 len--;
538 has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
539 if (has_trailing_carriage_return)
540 len--;
541 nofirst = 0;
542 }
543
544 if (len || !nofirst) {
545 fputs(set, file);
546 if (!nofirst)
547 fputc(first, file);
548 fwrite(line, len, 1, file);
549 fputs(reset, file);
550 }
551 if (has_trailing_carriage_return)
552 fputc('\r', file);
553 if (has_trailing_newline)
554 fputc('\n', file);
555}
556
557static void emit_line(struct diff_options *o, const char *set, const char *reset,
558 const char *line, int len)
559{
560 emit_line_0(o, set, reset, line[0], line+1, len-1);
561}
562
563enum diff_symbol {
564 DIFF_SYMBOL_WORDS_PORCELAIN,
565 DIFF_SYMBOL_WORDS,
566 DIFF_SYMBOL_CONTEXT,
567 DIFF_SYMBOL_PLUS,
568 DIFF_SYMBOL_MINUS,
569 DIFF_SYMBOL_NO_LF_EOF,
570 DIFF_SYMBOL_CONTEXT_FRAGINFO,
571 DIFF_SYMBOL_CONTEXT_MARKER,
572 DIFF_SYMBOL_SEPARATOR
573};
574/*
575 * Flags for content lines:
576 * 0..12 are whitespace rules
577 * 13-15 are WSEH_NEW | WSEH_OLD | WSEH_CONTEXT
578 * 16 is marking if the line is blank at EOF
579 */
580#define DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF (1<<16)
581#define DIFF_SYMBOL_CONTENT_WS_MASK (WSEH_NEW | WSEH_OLD | WSEH_CONTEXT | WS_RULE_MASK)
582
583static void emit_line_ws_markup(struct diff_options *o,
584 const char *set, const char *reset,
585 const char *line, int len, char sign,
586 unsigned ws_rule, int blank_at_eof)
587{
588 const char *ws = NULL;
589
590 if (o->ws_error_highlight & ws_rule) {
591 ws = diff_get_color_opt(o, DIFF_WHITESPACE);
592 if (!*ws)
593 ws = NULL;
594 }
595
596 if (!ws)
597 emit_line_0(o, set, reset, sign, line, len);
598 else if (blank_at_eof)
599 /* Blank line at EOF - paint '+' as well */
600 emit_line_0(o, ws, reset, sign, line, len);
601 else {
602 /* Emit just the prefix, then the rest. */
603 emit_line_0(o, set, reset, sign, "", 0);
604 ws_check_emit(line, len, ws_rule,
605 o->file, set, reset, ws);
606 }
607}
608
609static void emit_diff_symbol(struct diff_options *o, enum diff_symbol s,
610 const char *line, int len, unsigned flags)
611{
612 static const char *nneof = " No newline at end of file\n";
613 const char *context, *reset, *set;
614 switch (s) {
615 case DIFF_SYMBOL_NO_LF_EOF:
616 context = diff_get_color_opt(o, DIFF_CONTEXT);
617 reset = diff_get_color_opt(o, DIFF_RESET);
618 putc('\n', o->file);
619 emit_line_0(o, context, reset, '\\',
620 nneof, strlen(nneof));
621 break;
622 case DIFF_SYMBOL_CONTEXT_FRAGINFO:
623 emit_line(o, "", "", line, len);
624 break;
625 case DIFF_SYMBOL_CONTEXT_MARKER:
626 context = diff_get_color_opt(o, DIFF_CONTEXT);
627 reset = diff_get_color_opt(o, DIFF_RESET);
628 emit_line(o, context, reset, line, len);
629 break;
630 case DIFF_SYMBOL_SEPARATOR:
631 fprintf(o->file, "%s%c",
632 diff_line_prefix(o),
633 o->line_termination);
634 break;
635 case DIFF_SYMBOL_CONTEXT:
636 set = diff_get_color_opt(o, DIFF_CONTEXT);
637 reset = diff_get_color_opt(o, DIFF_RESET);
638 emit_line_ws_markup(o, set, reset, line, len, ' ',
639 flags & (DIFF_SYMBOL_CONTENT_WS_MASK), 0);
640 break;
641 case DIFF_SYMBOL_PLUS:
642 set = diff_get_color_opt(o, DIFF_FILE_NEW);
643 reset = diff_get_color_opt(o, DIFF_RESET);
644 emit_line_ws_markup(o, set, reset, line, len, '+',
645 flags & DIFF_SYMBOL_CONTENT_WS_MASK,
646 flags & DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF);
647 break;
648 case DIFF_SYMBOL_MINUS:
649 set = diff_get_color_opt(o, DIFF_FILE_OLD);
650 reset = diff_get_color_opt(o, DIFF_RESET);
651 emit_line_ws_markup(o, set, reset, line, len, '-',
652 flags & DIFF_SYMBOL_CONTENT_WS_MASK, 0);
653 break;
654 case DIFF_SYMBOL_WORDS_PORCELAIN:
655 context = diff_get_color_opt(o, DIFF_CONTEXT);
656 reset = diff_get_color_opt(o, DIFF_RESET);
657 emit_line(o, context, reset, line, len);
658 fputs("~\n", o->file);
659 break;
660 case DIFF_SYMBOL_WORDS:
661 context = diff_get_color_opt(o, DIFF_CONTEXT);
662 reset = diff_get_color_opt(o, DIFF_RESET);
663 /*
664 * Skip the prefix character, if any. With
665 * diff_suppress_blank_empty, there may be
666 * none.
667 */
668 if (line[0] != '\n') {
669 line++;
670 len--;
671 }
672 emit_line(o, context, reset, line, len);
673 break;
674 default:
675 die("BUG: unknown diff symbol");
676 }
677}
678
679static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
680{
681 if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
682 ecbdata->blank_at_eof_in_preimage &&
683 ecbdata->blank_at_eof_in_postimage &&
684 ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
685 ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
686 return 0;
687 return ws_blank_line(line, len, ecbdata->ws_rule);
688}
689
690static void emit_add_line(const char *reset,
691 struct emit_callback *ecbdata,
692 const char *line, int len)
693{
694 unsigned flags = WSEH_NEW | ecbdata->ws_rule;
695 if (new_blank_line_at_eof(ecbdata, line, len))
696 flags |= DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF;
697
698 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_PLUS, line, len, flags);
699}
700
701static void emit_del_line(const char *reset,
702 struct emit_callback *ecbdata,
703 const char *line, int len)
704{
705 unsigned flags = WSEH_OLD | ecbdata->ws_rule;
706 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_MINUS, line, len, flags);
707}
708
709static void emit_context_line(const char *reset,
710 struct emit_callback *ecbdata,
711 const char *line, int len)
712{
713 unsigned flags = WSEH_CONTEXT | ecbdata->ws_rule;
714 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_CONTEXT, line, len, flags);
715}
716
717static void emit_hunk_header(struct emit_callback *ecbdata,
718 const char *line, int len)
719{
720 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
721 const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
722 const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
723 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
724 static const char atat[2] = { '@', '@' };
725 const char *cp, *ep;
726 struct strbuf msgbuf = STRBUF_INIT;
727 int org_len = len;
728 int i = 1;
729
730 /*
731 * As a hunk header must begin with "@@ -<old>, +<new> @@",
732 * it always is at least 10 bytes long.
733 */
734 if (len < 10 ||
735 memcmp(line, atat, 2) ||
736 !(ep = memmem(line + 2, len - 2, atat, 2))) {
737 emit_diff_symbol(ecbdata->opt,
738 DIFF_SYMBOL_CONTEXT_MARKER, line, len, 0);
739 return;
740 }
741 ep += 2; /* skip over @@ */
742
743 /* The hunk header in fraginfo color */
744 strbuf_addstr(&msgbuf, frag);
745 strbuf_add(&msgbuf, line, ep - line);
746 strbuf_addstr(&msgbuf, reset);
747
748 /*
749 * trailing "\r\n"
750 */
751 for ( ; i < 3; i++)
752 if (line[len - i] == '\r' || line[len - i] == '\n')
753 len--;
754
755 /* blank before the func header */
756 for (cp = ep; ep - line < len; ep++)
757 if (*ep != ' ' && *ep != '\t')
758 break;
759 if (ep != cp) {
760 strbuf_addstr(&msgbuf, context);
761 strbuf_add(&msgbuf, cp, ep - cp);
762 strbuf_addstr(&msgbuf, reset);
763 }
764
765 if (ep < line + len) {
766 strbuf_addstr(&msgbuf, func);
767 strbuf_add(&msgbuf, ep, line + len - ep);
768 strbuf_addstr(&msgbuf, reset);
769 }
770
771 strbuf_add(&msgbuf, line + len, org_len - len);
772 strbuf_complete_line(&msgbuf);
773 emit_diff_symbol(ecbdata->opt,
774 DIFF_SYMBOL_CONTEXT_FRAGINFO, msgbuf.buf, msgbuf.len, 0);
775 strbuf_release(&msgbuf);
776}
777
778static struct diff_tempfile *claim_diff_tempfile(void) {
779 int i;
780 for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
781 if (!diff_temp[i].name)
782 return diff_temp + i;
783 die("BUG: diff is failing to clean up its tempfiles");
784}
785
786static void remove_tempfile(void)
787{
788 int i;
789 for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
790 if (is_tempfile_active(&diff_temp[i].tempfile))
791 delete_tempfile(&diff_temp[i].tempfile);
792 diff_temp[i].name = NULL;
793 }
794}
795
796static void print_line_count(FILE *file, int count)
797{
798 switch (count) {
799 case 0:
800 fprintf(file, "0,0");
801 break;
802 case 1:
803 fprintf(file, "1");
804 break;
805 default:
806 fprintf(file, "1,%d", count);
807 break;
808 }
809}
810
811static void emit_rewrite_lines(struct emit_callback *ecb,
812 int prefix, const char *data, int size)
813{
814 const char *endp = NULL;
815 const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
816
817 while (0 < size) {
818 int len;
819
820 endp = memchr(data, '\n', size);
821 len = endp ? (endp - data + 1) : size;
822 if (prefix != '+') {
823 ecb->lno_in_preimage++;
824 emit_del_line(reset, ecb, data, len);
825 } else {
826 ecb->lno_in_postimage++;
827 emit_add_line(reset, ecb, data, len);
828 }
829 size -= len;
830 data += len;
831 }
832 if (!endp)
833 emit_diff_symbol(ecb->opt, DIFF_SYMBOL_NO_LF_EOF, NULL, 0, 0);
834}
835
836static void emit_rewrite_diff(const char *name_a,
837 const char *name_b,
838 struct diff_filespec *one,
839 struct diff_filespec *two,
840 struct userdiff_driver *textconv_one,
841 struct userdiff_driver *textconv_two,
842 struct diff_options *o)
843{
844 int lc_a, lc_b;
845 const char *name_a_tab, *name_b_tab;
846 const char *metainfo = diff_get_color(o->use_color, DIFF_METAINFO);
847 const char *fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
848 const char *reset = diff_get_color(o->use_color, DIFF_RESET);
849 static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
850 const char *a_prefix, *b_prefix;
851 char *data_one, *data_two;
852 size_t size_one, size_two;
853 struct emit_callback ecbdata;
854 const char *line_prefix = diff_line_prefix(o);
855
856 if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
857 a_prefix = o->b_prefix;
858 b_prefix = o->a_prefix;
859 } else {
860 a_prefix = o->a_prefix;
861 b_prefix = o->b_prefix;
862 }
863
864 name_a += (*name_a == '/');
865 name_b += (*name_b == '/');
866 name_a_tab = strchr(name_a, ' ') ? "\t" : "";
867 name_b_tab = strchr(name_b, ' ') ? "\t" : "";
868
869 strbuf_reset(&a_name);
870 strbuf_reset(&b_name);
871 quote_two_c_style(&a_name, a_prefix, name_a, 0);
872 quote_two_c_style(&b_name, b_prefix, name_b, 0);
873
874 size_one = fill_textconv(textconv_one, one, &data_one);
875 size_two = fill_textconv(textconv_two, two, &data_two);
876
877 memset(&ecbdata, 0, sizeof(ecbdata));
878 ecbdata.color_diff = want_color(o->use_color);
879 ecbdata.ws_rule = whitespace_rule(name_b);
880 ecbdata.opt = o;
881 if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
882 mmfile_t mf1, mf2;
883 mf1.ptr = (char *)data_one;
884 mf2.ptr = (char *)data_two;
885 mf1.size = size_one;
886 mf2.size = size_two;
887 check_blank_at_eof(&mf1, &mf2, &ecbdata);
888 }
889 ecbdata.lno_in_preimage = 1;
890 ecbdata.lno_in_postimage = 1;
891
892 lc_a = count_lines(data_one, size_one);
893 lc_b = count_lines(data_two, size_two);
894 fprintf(o->file,
895 "%s%s--- %s%s%s\n%s%s+++ %s%s%s\n%s%s@@ -",
896 line_prefix, metainfo, a_name.buf, name_a_tab, reset,
897 line_prefix, metainfo, b_name.buf, name_b_tab, reset,
898 line_prefix, fraginfo);
899 if (!o->irreversible_delete)
900 print_line_count(o->file, lc_a);
901 else
902 fprintf(o->file, "?,?");
903 fprintf(o->file, " +");
904 print_line_count(o->file, lc_b);
905 fprintf(o->file, " @@%s\n", reset);
906 if (lc_a && !o->irreversible_delete)
907 emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
908 if (lc_b)
909 emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
910 if (textconv_one)
911 free((char *)data_one);
912 if (textconv_two)
913 free((char *)data_two);
914}
915
916struct diff_words_buffer {
917 mmfile_t text;
918 long alloc;
919 struct diff_words_orig {
920 const char *begin, *end;
921 } *orig;
922 int orig_nr, orig_alloc;
923};
924
925static void diff_words_append(char *line, unsigned long len,
926 struct diff_words_buffer *buffer)
927{
928 ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
929 line++;
930 len--;
931 memcpy(buffer->text.ptr + buffer->text.size, line, len);
932 buffer->text.size += len;
933 buffer->text.ptr[buffer->text.size] = '\0';
934}
935
936struct diff_words_style_elem {
937 const char *prefix;
938 const char *suffix;
939 const char *color; /* NULL; filled in by the setup code if
940 * color is enabled */
941};
942
943struct diff_words_style {
944 enum diff_words_type type;
945 struct diff_words_style_elem new, old, ctx;
946 const char *newline;
947};
948
949static struct diff_words_style diff_words_styles[] = {
950 { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
951 { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
952 { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
953};
954
955struct diff_words_data {
956 struct diff_words_buffer minus, plus;
957 const char *current_plus;
958 int last_minus;
959 struct diff_options *opt;
960 regex_t *word_regex;
961 enum diff_words_type type;
962 struct diff_words_style *style;
963};
964
965static int fn_out_diff_words_write_helper(FILE *fp,
966 struct diff_words_style_elem *st_el,
967 const char *newline,
968 size_t count, const char *buf,
969 const char *line_prefix)
970{
971 int print = 0;
972
973 while (count) {
974 char *p = memchr(buf, '\n', count);
975 if (print)
976 fputs(line_prefix, fp);
977 if (p != buf) {
978 if (st_el->color && fputs(st_el->color, fp) < 0)
979 return -1;
980 if (fputs(st_el->prefix, fp) < 0 ||
981 fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
982 fputs(st_el->suffix, fp) < 0)
983 return -1;
984 if (st_el->color && *st_el->color
985 && fputs(GIT_COLOR_RESET, fp) < 0)
986 return -1;
987 }
988 if (!p)
989 return 0;
990 if (fputs(newline, fp) < 0)
991 return -1;
992 count -= p + 1 - buf;
993 buf = p + 1;
994 print = 1;
995 }
996 return 0;
997}
998
999/*
1000 * '--color-words' algorithm can be described as:
1001 *
1002 * 1. collect the minus/plus lines of a diff hunk, divided into
1003 * minus-lines and plus-lines;
1004 *
1005 * 2. break both minus-lines and plus-lines into words and
1006 * place them into two mmfile_t with one word for each line;
1007 *
1008 * 3. use xdiff to run diff on the two mmfile_t to get the words level diff;
1009 *
1010 * And for the common parts of the both file, we output the plus side text.
1011 * diff_words->current_plus is used to trace the current position of the plus file
1012 * which printed. diff_words->last_minus is used to trace the last minus word
1013 * printed.
1014 *
1015 * For '--graph' to work with '--color-words', we need to output the graph prefix
1016 * on each line of color words output. Generally, there are two conditions on
1017 * which we should output the prefix.
1018 *
1019 * 1. diff_words->last_minus == 0 &&
1020 * diff_words->current_plus == diff_words->plus.text.ptr
1021 *
1022 * that is: the plus text must start as a new line, and if there is no minus
1023 * word printed, a graph prefix must be printed.
1024 *
1025 * 2. diff_words->current_plus > diff_words->plus.text.ptr &&
1026 * *(diff_words->current_plus - 1) == '\n'
1027 *
1028 * that is: a graph prefix must be printed following a '\n'
1029 */
1030static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
1031{
1032 if ((diff_words->last_minus == 0 &&
1033 diff_words->current_plus == diff_words->plus.text.ptr) ||
1034 (diff_words->current_plus > diff_words->plus.text.ptr &&
1035 *(diff_words->current_plus - 1) == '\n')) {
1036 return 1;
1037 } else {
1038 return 0;
1039 }
1040}
1041
1042static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
1043{
1044 struct diff_words_data *diff_words = priv;
1045 struct diff_words_style *style = diff_words->style;
1046 int minus_first, minus_len, plus_first, plus_len;
1047 const char *minus_begin, *minus_end, *plus_begin, *plus_end;
1048 struct diff_options *opt = diff_words->opt;
1049 const char *line_prefix;
1050
1051 if (line[0] != '@' || parse_hunk_header(line, len,
1052 &minus_first, &minus_len, &plus_first, &plus_len))
1053 return;
1054
1055 assert(opt);
1056 line_prefix = diff_line_prefix(opt);
1057
1058 /* POSIX requires that first be decremented by one if len == 0... */
1059 if (minus_len) {
1060 minus_begin = diff_words->minus.orig[minus_first].begin;
1061 minus_end =
1062 diff_words->minus.orig[minus_first + minus_len - 1].end;
1063 } else
1064 minus_begin = minus_end =
1065 diff_words->minus.orig[minus_first].end;
1066
1067 if (plus_len) {
1068 plus_begin = diff_words->plus.orig[plus_first].begin;
1069 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
1070 } else
1071 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
1072
1073 if (color_words_output_graph_prefix(diff_words)) {
1074 fputs(line_prefix, diff_words->opt->file);
1075 }
1076 if (diff_words->current_plus != plus_begin) {
1077 fn_out_diff_words_write_helper(diff_words->opt->file,
1078 &style->ctx, style->newline,
1079 plus_begin - diff_words->current_plus,
1080 diff_words->current_plus, line_prefix);
1081 if (*(plus_begin - 1) == '\n')
1082 fputs(line_prefix, diff_words->opt->file);
1083 }
1084 if (minus_begin != minus_end) {
1085 fn_out_diff_words_write_helper(diff_words->opt->file,
1086 &style->old, style->newline,
1087 minus_end - minus_begin, minus_begin,
1088 line_prefix);
1089 }
1090 if (plus_begin != plus_end) {
1091 fn_out_diff_words_write_helper(diff_words->opt->file,
1092 &style->new, style->newline,
1093 plus_end - plus_begin, plus_begin,
1094 line_prefix);
1095 }
1096
1097 diff_words->current_plus = plus_end;
1098 diff_words->last_minus = minus_first;
1099}
1100
1101/* This function starts looking at *begin, and returns 0 iff a word was found. */
1102static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
1103 int *begin, int *end)
1104{
1105 if (word_regex && *begin < buffer->size) {
1106 regmatch_t match[1];
1107 if (!regexec_buf(word_regex, buffer->ptr + *begin,
1108 buffer->size - *begin, 1, match, 0)) {
1109 char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
1110 '\n', match[0].rm_eo - match[0].rm_so);
1111 *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
1112 *begin += match[0].rm_so;
1113 return *begin >= *end;
1114 }
1115 return -1;
1116 }
1117
1118 /* find the next word */
1119 while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
1120 (*begin)++;
1121 if (*begin >= buffer->size)
1122 return -1;
1123
1124 /* find the end of the word */
1125 *end = *begin + 1;
1126 while (*end < buffer->size && !isspace(buffer->ptr[*end]))
1127 (*end)++;
1128
1129 return 0;
1130}
1131
1132/*
1133 * This function splits the words in buffer->text, stores the list with
1134 * newline separator into out, and saves the offsets of the original words
1135 * in buffer->orig.
1136 */
1137static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
1138 regex_t *word_regex)
1139{
1140 int i, j;
1141 long alloc = 0;
1142
1143 out->size = 0;
1144 out->ptr = NULL;
1145
1146 /* fake an empty "0th" word */
1147 ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
1148 buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
1149 buffer->orig_nr = 1;
1150
1151 for (i = 0; i < buffer->text.size; i++) {
1152 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
1153 return;
1154
1155 /* store original boundaries */
1156 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
1157 buffer->orig_alloc);
1158 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
1159 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
1160 buffer->orig_nr++;
1161
1162 /* store one word */
1163 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
1164 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
1165 out->ptr[out->size + j - i] = '\n';
1166 out->size += j - i + 1;
1167
1168 i = j - 1;
1169 }
1170}
1171
1172/* this executes the word diff on the accumulated buffers */
1173static void diff_words_show(struct diff_words_data *diff_words)
1174{
1175 xpparam_t xpp;
1176 xdemitconf_t xecfg;
1177 mmfile_t minus, plus;
1178 struct diff_words_style *style = diff_words->style;
1179
1180 struct diff_options *opt = diff_words->opt;
1181 const char *line_prefix;
1182
1183 assert(opt);
1184 line_prefix = diff_line_prefix(opt);
1185
1186 /* special case: only removal */
1187 if (!diff_words->plus.text.size) {
1188 fputs(line_prefix, diff_words->opt->file);
1189 fn_out_diff_words_write_helper(diff_words->opt->file,
1190 &style->old, style->newline,
1191 diff_words->minus.text.size,
1192 diff_words->minus.text.ptr, line_prefix);
1193 diff_words->minus.text.size = 0;
1194 return;
1195 }
1196
1197 diff_words->current_plus = diff_words->plus.text.ptr;
1198 diff_words->last_minus = 0;
1199
1200 memset(&xpp, 0, sizeof(xpp));
1201 memset(&xecfg, 0, sizeof(xecfg));
1202 diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
1203 diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
1204 xpp.flags = 0;
1205 /* as only the hunk header will be parsed, we need a 0-context */
1206 xecfg.ctxlen = 0;
1207 if (xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
1208 &xpp, &xecfg))
1209 die("unable to generate word diff");
1210 free(minus.ptr);
1211 free(plus.ptr);
1212 if (diff_words->current_plus != diff_words->plus.text.ptr +
1213 diff_words->plus.text.size) {
1214 if (color_words_output_graph_prefix(diff_words))
1215 fputs(line_prefix, diff_words->opt->file);
1216 fn_out_diff_words_write_helper(diff_words->opt->file,
1217 &style->ctx, style->newline,
1218 diff_words->plus.text.ptr + diff_words->plus.text.size
1219 - diff_words->current_plus, diff_words->current_plus,
1220 line_prefix);
1221 }
1222 diff_words->minus.text.size = diff_words->plus.text.size = 0;
1223}
1224
1225/* In "color-words" mode, show word-diff of words accumulated in the buffer */
1226static void diff_words_flush(struct emit_callback *ecbdata)
1227{
1228 if (ecbdata->diff_words->minus.text.size ||
1229 ecbdata->diff_words->plus.text.size)
1230 diff_words_show(ecbdata->diff_words);
1231}
1232
1233static void diff_filespec_load_driver(struct diff_filespec *one)
1234{
1235 /* Use already-loaded driver */
1236 if (one->driver)
1237 return;
1238
1239 if (S_ISREG(one->mode))
1240 one->driver = userdiff_find_by_path(one->path);
1241
1242 /* Fallback to default settings */
1243 if (!one->driver)
1244 one->driver = userdiff_find_by_name("default");
1245}
1246
1247static const char *userdiff_word_regex(struct diff_filespec *one)
1248{
1249 diff_filespec_load_driver(one);
1250 return one->driver->word_regex;
1251}
1252
1253static void init_diff_words_data(struct emit_callback *ecbdata,
1254 struct diff_options *orig_opts,
1255 struct diff_filespec *one,
1256 struct diff_filespec *two)
1257{
1258 int i;
1259 struct diff_options *o = xmalloc(sizeof(struct diff_options));
1260 memcpy(o, orig_opts, sizeof(struct diff_options));
1261
1262 ecbdata->diff_words =
1263 xcalloc(1, sizeof(struct diff_words_data));
1264 ecbdata->diff_words->type = o->word_diff;
1265 ecbdata->diff_words->opt = o;
1266 if (!o->word_regex)
1267 o->word_regex = userdiff_word_regex(one);
1268 if (!o->word_regex)
1269 o->word_regex = userdiff_word_regex(two);
1270 if (!o->word_regex)
1271 o->word_regex = diff_word_regex_cfg;
1272 if (o->word_regex) {
1273 ecbdata->diff_words->word_regex = (regex_t *)
1274 xmalloc(sizeof(regex_t));
1275 if (regcomp(ecbdata->diff_words->word_regex,
1276 o->word_regex,
1277 REG_EXTENDED | REG_NEWLINE))
1278 die ("Invalid regular expression: %s",
1279 o->word_regex);
1280 }
1281 for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1282 if (o->word_diff == diff_words_styles[i].type) {
1283 ecbdata->diff_words->style =
1284 &diff_words_styles[i];
1285 break;
1286 }
1287 }
1288 if (want_color(o->use_color)) {
1289 struct diff_words_style *st = ecbdata->diff_words->style;
1290 st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1291 st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1292 st->ctx.color = diff_get_color_opt(o, DIFF_CONTEXT);
1293 }
1294}
1295
1296static void free_diff_words_data(struct emit_callback *ecbdata)
1297{
1298 if (ecbdata->diff_words) {
1299 diff_words_flush(ecbdata);
1300 free (ecbdata->diff_words->opt);
1301 free (ecbdata->diff_words->minus.text.ptr);
1302 free (ecbdata->diff_words->minus.orig);
1303 free (ecbdata->diff_words->plus.text.ptr);
1304 free (ecbdata->diff_words->plus.orig);
1305 if (ecbdata->diff_words->word_regex) {
1306 regfree(ecbdata->diff_words->word_regex);
1307 free(ecbdata->diff_words->word_regex);
1308 }
1309 FREE_AND_NULL(ecbdata->diff_words);
1310 }
1311}
1312
1313const char *diff_get_color(int diff_use_color, enum color_diff ix)
1314{
1315 if (want_color(diff_use_color))
1316 return diff_colors[ix];
1317 return "";
1318}
1319
1320const char *diff_line_prefix(struct diff_options *opt)
1321{
1322 struct strbuf *msgbuf;
1323 if (!opt->output_prefix)
1324 return "";
1325
1326 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
1327 return msgbuf->buf;
1328}
1329
1330static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1331{
1332 const char *cp;
1333 unsigned long allot;
1334 size_t l = len;
1335
1336 if (ecb->truncate)
1337 return ecb->truncate(line, len);
1338 cp = line;
1339 allot = l;
1340 while (0 < l) {
1341 (void) utf8_width(&cp, &l);
1342 if (!cp)
1343 break; /* truncated in the middle? */
1344 }
1345 return allot - l;
1346}
1347
1348static void find_lno(const char *line, struct emit_callback *ecbdata)
1349{
1350 const char *p;
1351 ecbdata->lno_in_preimage = 0;
1352 ecbdata->lno_in_postimage = 0;
1353 p = strchr(line, '-');
1354 if (!p)
1355 return; /* cannot happen */
1356 ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1357 p = strchr(p, '+');
1358 if (!p)
1359 return; /* cannot happen */
1360 ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1361}
1362
1363static void fn_out_consume(void *priv, char *line, unsigned long len)
1364{
1365 struct emit_callback *ecbdata = priv;
1366 const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
1367 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1368 struct diff_options *o = ecbdata->opt;
1369 const char *line_prefix = diff_line_prefix(o);
1370
1371 o->found_changes = 1;
1372
1373 if (ecbdata->header) {
1374 fprintf(o->file, "%s", ecbdata->header->buf);
1375 strbuf_reset(ecbdata->header);
1376 ecbdata->header = NULL;
1377 }
1378
1379 if (ecbdata->label_path[0]) {
1380 const char *name_a_tab, *name_b_tab;
1381
1382 name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
1383 name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
1384
1385 fprintf(o->file, "%s%s--- %s%s%s\n",
1386 line_prefix, meta, ecbdata->label_path[0], reset, name_a_tab);
1387 fprintf(o->file, "%s%s+++ %s%s%s\n",
1388 line_prefix, meta, ecbdata->label_path[1], reset, name_b_tab);
1389 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1390 }
1391
1392 if (diff_suppress_blank_empty
1393 && len == 2 && line[0] == ' ' && line[1] == '\n') {
1394 line[0] = '\n';
1395 len = 1;
1396 }
1397
1398 if (line[0] == '@') {
1399 if (ecbdata->diff_words)
1400 diff_words_flush(ecbdata);
1401 len = sane_truncate_line(ecbdata, line, len);
1402 find_lno(line, ecbdata);
1403 emit_hunk_header(ecbdata, line, len);
1404 return;
1405 }
1406
1407 if (ecbdata->diff_words) {
1408 enum diff_symbol s =
1409 ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN ?
1410 DIFF_SYMBOL_WORDS_PORCELAIN : DIFF_SYMBOL_WORDS;
1411 if (line[0] == '-') {
1412 diff_words_append(line, len,
1413 &ecbdata->diff_words->minus);
1414 return;
1415 } else if (line[0] == '+') {
1416 diff_words_append(line, len,
1417 &ecbdata->diff_words->plus);
1418 return;
1419 } else if (starts_with(line, "\\ ")) {
1420 /*
1421 * Eat the "no newline at eof" marker as if we
1422 * saw a "+" or "-" line with nothing on it,
1423 * and return without diff_words_flush() to
1424 * defer processing. If this is the end of
1425 * preimage, more "+" lines may come after it.
1426 */
1427 return;
1428 }
1429 diff_words_flush(ecbdata);
1430 emit_diff_symbol(o, s, line, len, 0);
1431 return;
1432 }
1433
1434 switch (line[0]) {
1435 case '+':
1436 ecbdata->lno_in_postimage++;
1437 emit_add_line(reset, ecbdata, line + 1, len - 1);
1438 break;
1439 case '-':
1440 ecbdata->lno_in_preimage++;
1441 emit_del_line(reset, ecbdata, line + 1, len - 1);
1442 break;
1443 case ' ':
1444 ecbdata->lno_in_postimage++;
1445 ecbdata->lno_in_preimage++;
1446 emit_context_line(reset, ecbdata, line + 1, len - 1);
1447 break;
1448 default:
1449 /* incomplete line at the end */
1450 ecbdata->lno_in_preimage++;
1451 emit_line(o, diff_get_color(ecbdata->color_diff, DIFF_CONTEXT),
1452 reset, line, len);
1453 break;
1454 }
1455}
1456
1457static char *pprint_rename(const char *a, const char *b)
1458{
1459 const char *old = a;
1460 const char *new = b;
1461 struct strbuf name = STRBUF_INIT;
1462 int pfx_length, sfx_length;
1463 int pfx_adjust_for_slash;
1464 int len_a = strlen(a);
1465 int len_b = strlen(b);
1466 int a_midlen, b_midlen;
1467 int qlen_a = quote_c_style(a, NULL, NULL, 0);
1468 int qlen_b = quote_c_style(b, NULL, NULL, 0);
1469
1470 if (qlen_a || qlen_b) {
1471 quote_c_style(a, &name, NULL, 0);
1472 strbuf_addstr(&name, " => ");
1473 quote_c_style(b, &name, NULL, 0);
1474 return strbuf_detach(&name, NULL);
1475 }
1476
1477 /* Find common prefix */
1478 pfx_length = 0;
1479 while (*old && *new && *old == *new) {
1480 if (*old == '/')
1481 pfx_length = old - a + 1;
1482 old++;
1483 new++;
1484 }
1485
1486 /* Find common suffix */
1487 old = a + len_a;
1488 new = b + len_b;
1489 sfx_length = 0;
1490 /*
1491 * If there is a common prefix, it must end in a slash. In
1492 * that case we let this loop run 1 into the prefix to see the
1493 * same slash.
1494 *
1495 * If there is no common prefix, we cannot do this as it would
1496 * underrun the input strings.
1497 */
1498 pfx_adjust_for_slash = (pfx_length ? 1 : 0);
1499 while (a + pfx_length - pfx_adjust_for_slash <= old &&
1500 b + pfx_length - pfx_adjust_for_slash <= new &&
1501 *old == *new) {
1502 if (*old == '/')
1503 sfx_length = len_a - (old - a);
1504 old--;
1505 new--;
1506 }
1507
1508 /*
1509 * pfx{mid-a => mid-b}sfx
1510 * {pfx-a => pfx-b}sfx
1511 * pfx{sfx-a => sfx-b}
1512 * name-a => name-b
1513 */
1514 a_midlen = len_a - pfx_length - sfx_length;
1515 b_midlen = len_b - pfx_length - sfx_length;
1516 if (a_midlen < 0)
1517 a_midlen = 0;
1518 if (b_midlen < 0)
1519 b_midlen = 0;
1520
1521 strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1522 if (pfx_length + sfx_length) {
1523 strbuf_add(&name, a, pfx_length);
1524 strbuf_addch(&name, '{');
1525 }
1526 strbuf_add(&name, a + pfx_length, a_midlen);
1527 strbuf_addstr(&name, " => ");
1528 strbuf_add(&name, b + pfx_length, b_midlen);
1529 if (pfx_length + sfx_length) {
1530 strbuf_addch(&name, '}');
1531 strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1532 }
1533 return strbuf_detach(&name, NULL);
1534}
1535
1536struct diffstat_t {
1537 int nr;
1538 int alloc;
1539 struct diffstat_file {
1540 char *from_name;
1541 char *name;
1542 char *print_name;
1543 unsigned is_unmerged:1;
1544 unsigned is_binary:1;
1545 unsigned is_renamed:1;
1546 unsigned is_interesting:1;
1547 uintmax_t added, deleted;
1548 } **files;
1549};
1550
1551static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1552 const char *name_a,
1553 const char *name_b)
1554{
1555 struct diffstat_file *x;
1556 x = xcalloc(1, sizeof(*x));
1557 ALLOC_GROW(diffstat->files, diffstat->nr + 1, diffstat->alloc);
1558 diffstat->files[diffstat->nr++] = x;
1559 if (name_b) {
1560 x->from_name = xstrdup(name_a);
1561 x->name = xstrdup(name_b);
1562 x->is_renamed = 1;
1563 }
1564 else {
1565 x->from_name = NULL;
1566 x->name = xstrdup(name_a);
1567 }
1568 return x;
1569}
1570
1571static void diffstat_consume(void *priv, char *line, unsigned long len)
1572{
1573 struct diffstat_t *diffstat = priv;
1574 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1575
1576 if (line[0] == '+')
1577 x->added++;
1578 else if (line[0] == '-')
1579 x->deleted++;
1580}
1581
1582const char mime_boundary_leader[] = "------------";
1583
1584static int scale_linear(int it, int width, int max_change)
1585{
1586 if (!it)
1587 return 0;
1588 /*
1589 * make sure that at least one '-' or '+' is printed if
1590 * there is any change to this path. The easiest way is to
1591 * scale linearly as if the alloted width is one column shorter
1592 * than it is, and then add 1 to the result.
1593 */
1594 return 1 + (it * (width - 1) / max_change);
1595}
1596
1597static void show_name(FILE *file,
1598 const char *prefix, const char *name, int len)
1599{
1600 fprintf(file, " %s%-*s |", prefix, len, name);
1601}
1602
1603static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1604{
1605 if (cnt <= 0)
1606 return;
1607 fprintf(file, "%s", set);
1608 while (cnt--)
1609 putc(ch, file);
1610 fprintf(file, "%s", reset);
1611}
1612
1613static void fill_print_name(struct diffstat_file *file)
1614{
1615 char *pname;
1616
1617 if (file->print_name)
1618 return;
1619
1620 if (!file->is_renamed) {
1621 struct strbuf buf = STRBUF_INIT;
1622 if (quote_c_style(file->name, &buf, NULL, 0)) {
1623 pname = strbuf_detach(&buf, NULL);
1624 } else {
1625 pname = file->name;
1626 strbuf_release(&buf);
1627 }
1628 } else {
1629 pname = pprint_rename(file->from_name, file->name);
1630 }
1631 file->print_name = pname;
1632}
1633
1634int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1635{
1636 struct strbuf sb = STRBUF_INIT;
1637 int ret;
1638
1639 if (!files) {
1640 assert(insertions == 0 && deletions == 0);
1641 return fprintf(fp, "%s\n", " 0 files changed");
1642 }
1643
1644 strbuf_addf(&sb,
1645 (files == 1) ? " %d file changed" : " %d files changed",
1646 files);
1647
1648 /*
1649 * For binary diff, the caller may want to print "x files
1650 * changed" with insertions == 0 && deletions == 0.
1651 *
1652 * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1653 * is probably less confusing (i.e skip over "2 files changed
1654 * but nothing about added/removed lines? Is this a bug in Git?").
1655 */
1656 if (insertions || deletions == 0) {
1657 strbuf_addf(&sb,
1658 (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
1659 insertions);
1660 }
1661
1662 if (deletions || insertions == 0) {
1663 strbuf_addf(&sb,
1664 (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
1665 deletions);
1666 }
1667 strbuf_addch(&sb, '\n');
1668 ret = fputs(sb.buf, fp);
1669 strbuf_release(&sb);
1670 return ret;
1671}
1672
1673static void show_stats(struct diffstat_t *data, struct diff_options *options)
1674{
1675 int i, len, add, del, adds = 0, dels = 0;
1676 uintmax_t max_change = 0, max_len = 0;
1677 int total_files = data->nr, count;
1678 int width, name_width, graph_width, number_width = 0, bin_width = 0;
1679 const char *reset, *add_c, *del_c;
1680 const char *line_prefix = "";
1681 int extra_shown = 0;
1682
1683 if (data->nr == 0)
1684 return;
1685
1686 line_prefix = diff_line_prefix(options);
1687 count = options->stat_count ? options->stat_count : data->nr;
1688
1689 reset = diff_get_color_opt(options, DIFF_RESET);
1690 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1691 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1692
1693 /*
1694 * Find the longest filename and max number of changes
1695 */
1696 for (i = 0; (i < count) && (i < data->nr); i++) {
1697 struct diffstat_file *file = data->files[i];
1698 uintmax_t change = file->added + file->deleted;
1699
1700 if (!file->is_interesting && (change == 0)) {
1701 count++; /* not shown == room for one more */
1702 continue;
1703 }
1704 fill_print_name(file);
1705 len = strlen(file->print_name);
1706 if (max_len < len)
1707 max_len = len;
1708
1709 if (file->is_unmerged) {
1710 /* "Unmerged" is 8 characters */
1711 bin_width = bin_width < 8 ? 8 : bin_width;
1712 continue;
1713 }
1714 if (file->is_binary) {
1715 /* "Bin XXX -> YYY bytes" */
1716 int w = 14 + decimal_width(file->added)
1717 + decimal_width(file->deleted);
1718 bin_width = bin_width < w ? w : bin_width;
1719 /* Display change counts aligned with "Bin" */
1720 number_width = 3;
1721 continue;
1722 }
1723
1724 if (max_change < change)
1725 max_change = change;
1726 }
1727 count = i; /* where we can stop scanning in data->files[] */
1728
1729 /*
1730 * We have width = stat_width or term_columns() columns total.
1731 * We want a maximum of min(max_len, stat_name_width) for the name part.
1732 * We want a maximum of min(max_change, stat_graph_width) for the +- part.
1733 * We also need 1 for " " and 4 + decimal_width(max_change)
1734 * for " | NNNN " and one the empty column at the end, altogether
1735 * 6 + decimal_width(max_change).
1736 *
1737 * If there's not enough space, we will use the smaller of
1738 * stat_name_width (if set) and 5/8*width for the filename,
1739 * and the rest for constant elements + graph part, but no more
1740 * than stat_graph_width for the graph part.
1741 * (5/8 gives 50 for filename and 30 for the constant parts + graph
1742 * for the standard terminal size).
1743 *
1744 * In other words: stat_width limits the maximum width, and
1745 * stat_name_width fixes the maximum width of the filename,
1746 * and is also used to divide available columns if there
1747 * aren't enough.
1748 *
1749 * Binary files are displayed with "Bin XXX -> YYY bytes"
1750 * instead of the change count and graph. This part is treated
1751 * similarly to the graph part, except that it is not
1752 * "scaled". If total width is too small to accommodate the
1753 * guaranteed minimum width of the filename part and the
1754 * separators and this message, this message will "overflow"
1755 * making the line longer than the maximum width.
1756 */
1757
1758 if (options->stat_width == -1)
1759 width = term_columns() - strlen(line_prefix);
1760 else
1761 width = options->stat_width ? options->stat_width : 80;
1762 number_width = decimal_width(max_change) > number_width ?
1763 decimal_width(max_change) : number_width;
1764
1765 if (options->stat_graph_width == -1)
1766 options->stat_graph_width = diff_stat_graph_width;
1767
1768 /*
1769 * Guarantee 3/8*16==6 for the graph part
1770 * and 5/8*16==10 for the filename part
1771 */
1772 if (width < 16 + 6 + number_width)
1773 width = 16 + 6 + number_width;
1774
1775 /*
1776 * First assign sizes that are wanted, ignoring available width.
1777 * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
1778 * starting from "XXX" should fit in graph_width.
1779 */
1780 graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
1781 if (options->stat_graph_width &&
1782 options->stat_graph_width < graph_width)
1783 graph_width = options->stat_graph_width;
1784
1785 name_width = (options->stat_name_width > 0 &&
1786 options->stat_name_width < max_len) ?
1787 options->stat_name_width : max_len;
1788
1789 /*
1790 * Adjust adjustable widths not to exceed maximum width
1791 */
1792 if (name_width + number_width + 6 + graph_width > width) {
1793 if (graph_width > width * 3/8 - number_width - 6) {
1794 graph_width = width * 3/8 - number_width - 6;
1795 if (graph_width < 6)
1796 graph_width = 6;
1797 }
1798
1799 if (options->stat_graph_width &&
1800 graph_width > options->stat_graph_width)
1801 graph_width = options->stat_graph_width;
1802 if (name_width > width - number_width - 6 - graph_width)
1803 name_width = width - number_width - 6 - graph_width;
1804 else
1805 graph_width = width - number_width - 6 - name_width;
1806 }
1807
1808 /*
1809 * From here name_width is the width of the name area,
1810 * and graph_width is the width of the graph area.
1811 * max_change is used to scale graph properly.
1812 */
1813 for (i = 0; i < count; i++) {
1814 const char *prefix = "";
1815 struct diffstat_file *file = data->files[i];
1816 char *name = file->print_name;
1817 uintmax_t added = file->added;
1818 uintmax_t deleted = file->deleted;
1819 int name_len;
1820
1821 if (!file->is_interesting && (added + deleted == 0))
1822 continue;
1823
1824 /*
1825 * "scale" the filename
1826 */
1827 len = name_width;
1828 name_len = strlen(name);
1829 if (name_width < name_len) {
1830 char *slash;
1831 prefix = "...";
1832 len -= 3;
1833 name += name_len - len;
1834 slash = strchr(name, '/');
1835 if (slash)
1836 name = slash;
1837 }
1838
1839 if (file->is_binary) {
1840 fprintf(options->file, "%s", line_prefix);
1841 show_name(options->file, prefix, name, len);
1842 fprintf(options->file, " %*s", number_width, "Bin");
1843 if (!added && !deleted) {
1844 putc('\n', options->file);
1845 continue;
1846 }
1847 fprintf(options->file, " %s%"PRIuMAX"%s",
1848 del_c, deleted, reset);
1849 fprintf(options->file, " -> ");
1850 fprintf(options->file, "%s%"PRIuMAX"%s",
1851 add_c, added, reset);
1852 fprintf(options->file, " bytes");
1853 fprintf(options->file, "\n");
1854 continue;
1855 }
1856 else if (file->is_unmerged) {
1857 fprintf(options->file, "%s", line_prefix);
1858 show_name(options->file, prefix, name, len);
1859 fprintf(options->file, " Unmerged\n");
1860 continue;
1861 }
1862
1863 /*
1864 * scale the add/delete
1865 */
1866 add = added;
1867 del = deleted;
1868
1869 if (graph_width <= max_change) {
1870 int total = scale_linear(add + del, graph_width, max_change);
1871 if (total < 2 && add && del)
1872 /* width >= 2 due to the sanity check */
1873 total = 2;
1874 if (add < del) {
1875 add = scale_linear(add, graph_width, max_change);
1876 del = total - add;
1877 } else {
1878 del = scale_linear(del, graph_width, max_change);
1879 add = total - del;
1880 }
1881 }
1882 fprintf(options->file, "%s", line_prefix);
1883 show_name(options->file, prefix, name, len);
1884 fprintf(options->file, " %*"PRIuMAX"%s",
1885 number_width, added + deleted,
1886 added + deleted ? " " : "");
1887 show_graph(options->file, '+', add, add_c, reset);
1888 show_graph(options->file, '-', del, del_c, reset);
1889 fprintf(options->file, "\n");
1890 }
1891
1892 for (i = 0; i < data->nr; i++) {
1893 struct diffstat_file *file = data->files[i];
1894 uintmax_t added = file->added;
1895 uintmax_t deleted = file->deleted;
1896
1897 if (file->is_unmerged ||
1898 (!file->is_interesting && (added + deleted == 0))) {
1899 total_files--;
1900 continue;
1901 }
1902
1903 if (!file->is_binary) {
1904 adds += added;
1905 dels += deleted;
1906 }
1907 if (i < count)
1908 continue;
1909 if (!extra_shown)
1910 fprintf(options->file, "%s ...\n", line_prefix);
1911 extra_shown = 1;
1912 }
1913 fprintf(options->file, "%s", line_prefix);
1914 print_stat_summary(options->file, total_files, adds, dels);
1915}
1916
1917static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1918{
1919 int i, adds = 0, dels = 0, total_files = data->nr;
1920
1921 if (data->nr == 0)
1922 return;
1923
1924 for (i = 0; i < data->nr; i++) {
1925 int added = data->files[i]->added;
1926 int deleted= data->files[i]->deleted;
1927
1928 if (data->files[i]->is_unmerged ||
1929 (!data->files[i]->is_interesting && (added + deleted == 0))) {
1930 total_files--;
1931 } else if (!data->files[i]->is_binary) { /* don't count bytes */
1932 adds += added;
1933 dels += deleted;
1934 }
1935 }
1936 fprintf(options->file, "%s", diff_line_prefix(options));
1937 print_stat_summary(options->file, total_files, adds, dels);
1938}
1939
1940static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1941{
1942 int i;
1943
1944 if (data->nr == 0)
1945 return;
1946
1947 for (i = 0; i < data->nr; i++) {
1948 struct diffstat_file *file = data->files[i];
1949
1950 fprintf(options->file, "%s", diff_line_prefix(options));
1951
1952 if (file->is_binary)
1953 fprintf(options->file, "-\t-\t");
1954 else
1955 fprintf(options->file,
1956 "%"PRIuMAX"\t%"PRIuMAX"\t",
1957 file->added, file->deleted);
1958 if (options->line_termination) {
1959 fill_print_name(file);
1960 if (!file->is_renamed)
1961 write_name_quoted(file->name, options->file,
1962 options->line_termination);
1963 else {
1964 fputs(file->print_name, options->file);
1965 putc(options->line_termination, options->file);
1966 }
1967 } else {
1968 if (file->is_renamed) {
1969 putc('\0', options->file);
1970 write_name_quoted(file->from_name, options->file, '\0');
1971 }
1972 write_name_quoted(file->name, options->file, '\0');
1973 }
1974 }
1975}
1976
1977struct dirstat_file {
1978 const char *name;
1979 unsigned long changed;
1980};
1981
1982struct dirstat_dir {
1983 struct dirstat_file *files;
1984 int alloc, nr, permille, cumulative;
1985};
1986
1987static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1988 unsigned long changed, const char *base, int baselen)
1989{
1990 unsigned long this_dir = 0;
1991 unsigned int sources = 0;
1992 const char *line_prefix = diff_line_prefix(opt);
1993
1994 while (dir->nr) {
1995 struct dirstat_file *f = dir->files;
1996 int namelen = strlen(f->name);
1997 unsigned long this;
1998 char *slash;
1999
2000 if (namelen < baselen)
2001 break;
2002 if (memcmp(f->name, base, baselen))
2003 break;
2004 slash = strchr(f->name + baselen, '/');
2005 if (slash) {
2006 int newbaselen = slash + 1 - f->name;
2007 this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
2008 sources++;
2009 } else {
2010 this = f->changed;
2011 dir->files++;
2012 dir->nr--;
2013 sources += 2;
2014 }
2015 this_dir += this;
2016 }
2017
2018 /*
2019 * We don't report dirstat's for
2020 * - the top level
2021 * - or cases where everything came from a single directory
2022 * under this directory (sources == 1).
2023 */
2024 if (baselen && sources != 1) {
2025 if (this_dir) {
2026 int permille = this_dir * 1000 / changed;
2027 if (permille >= dir->permille) {
2028 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
2029 permille / 10, permille % 10, baselen, base);
2030 if (!dir->cumulative)
2031 return 0;
2032 }
2033 }
2034 }
2035 return this_dir;
2036}
2037
2038static int dirstat_compare(const void *_a, const void *_b)
2039{
2040 const struct dirstat_file *a = _a;
2041 const struct dirstat_file *b = _b;
2042 return strcmp(a->name, b->name);
2043}
2044
2045static void show_dirstat(struct diff_options *options)
2046{
2047 int i;
2048 unsigned long changed;
2049 struct dirstat_dir dir;
2050 struct diff_queue_struct *q = &diff_queued_diff;
2051
2052 dir.files = NULL;
2053 dir.alloc = 0;
2054 dir.nr = 0;
2055 dir.permille = options->dirstat_permille;
2056 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
2057
2058 changed = 0;
2059 for (i = 0; i < q->nr; i++) {
2060 struct diff_filepair *p = q->queue[i];
2061 const char *name;
2062 unsigned long copied, added, damage;
2063 int content_changed;
2064
2065 name = p->two->path ? p->two->path : p->one->path;
2066
2067 if (p->one->oid_valid && p->two->oid_valid)
2068 content_changed = oidcmp(&p->one->oid, &p->two->oid);
2069 else
2070 content_changed = 1;
2071
2072 if (!content_changed) {
2073 /*
2074 * The SHA1 has not changed, so pre-/post-content is
2075 * identical. We can therefore skip looking at the
2076 * file contents altogether.
2077 */
2078 damage = 0;
2079 goto found_damage;
2080 }
2081
2082 if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
2083 /*
2084 * In --dirstat-by-file mode, we don't really need to
2085 * look at the actual file contents at all.
2086 * The fact that the SHA1 changed is enough for us to
2087 * add this file to the list of results
2088 * (with each file contributing equal damage).
2089 */
2090 damage = 1;
2091 goto found_damage;
2092 }
2093
2094 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
2095 diff_populate_filespec(p->one, 0);
2096 diff_populate_filespec(p->two, 0);
2097 diffcore_count_changes(p->one, p->two, NULL, NULL,
2098 &copied, &added);
2099 diff_free_filespec_data(p->one);
2100 diff_free_filespec_data(p->two);
2101 } else if (DIFF_FILE_VALID(p->one)) {
2102 diff_populate_filespec(p->one, CHECK_SIZE_ONLY);
2103 copied = added = 0;
2104 diff_free_filespec_data(p->one);
2105 } else if (DIFF_FILE_VALID(p->two)) {
2106 diff_populate_filespec(p->two, CHECK_SIZE_ONLY);
2107 copied = 0;
2108 added = p->two->size;
2109 diff_free_filespec_data(p->two);
2110 } else
2111 continue;
2112
2113 /*
2114 * Original minus copied is the removed material,
2115 * added is the new material. They are both damages
2116 * made to the preimage.
2117 * If the resulting damage is zero, we know that
2118 * diffcore_count_changes() considers the two entries to
2119 * be identical, but since content_changed is true, we
2120 * know that there must have been _some_ kind of change,
2121 * so we force all entries to have damage > 0.
2122 */
2123 damage = (p->one->size - copied) + added;
2124 if (!damage)
2125 damage = 1;
2126
2127found_damage:
2128 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2129 dir.files[dir.nr].name = name;
2130 dir.files[dir.nr].changed = damage;
2131 changed += damage;
2132 dir.nr++;
2133 }
2134
2135 /* This can happen even with many files, if everything was renames */
2136 if (!changed)
2137 return;
2138
2139 /* Show all directories with more than x% of the changes */
2140 QSORT(dir.files, dir.nr, dirstat_compare);
2141 gather_dirstat(options, &dir, changed, "", 0);
2142}
2143
2144static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
2145{
2146 int i;
2147 unsigned long changed;
2148 struct dirstat_dir dir;
2149
2150 if (data->nr == 0)
2151 return;
2152
2153 dir.files = NULL;
2154 dir.alloc = 0;
2155 dir.nr = 0;
2156 dir.permille = options->dirstat_permille;
2157 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
2158
2159 changed = 0;
2160 for (i = 0; i < data->nr; i++) {
2161 struct diffstat_file *file = data->files[i];
2162 unsigned long damage = file->added + file->deleted;
2163 if (file->is_binary)
2164 /*
2165 * binary files counts bytes, not lines. Must find some
2166 * way to normalize binary bytes vs. textual lines.
2167 * The following heuristic assumes that there are 64
2168 * bytes per "line".
2169 * This is stupid and ugly, but very cheap...
2170 */
2171 damage = (damage + 63) / 64;
2172 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2173 dir.files[dir.nr].name = file->name;
2174 dir.files[dir.nr].changed = damage;
2175 changed += damage;
2176 dir.nr++;
2177 }
2178
2179 /* This can happen even with many files, if everything was renames */
2180 if (!changed)
2181 return;
2182
2183 /* Show all directories with more than x% of the changes */
2184 QSORT(dir.files, dir.nr, dirstat_compare);
2185 gather_dirstat(options, &dir, changed, "", 0);
2186}
2187
2188static void free_diffstat_info(struct diffstat_t *diffstat)
2189{
2190 int i;
2191 for (i = 0; i < diffstat->nr; i++) {
2192 struct diffstat_file *f = diffstat->files[i];
2193 if (f->name != f->print_name)
2194 free(f->print_name);
2195 free(f->name);
2196 free(f->from_name);
2197 free(f);
2198 }
2199 free(diffstat->files);
2200}
2201
2202struct checkdiff_t {
2203 const char *filename;
2204 int lineno;
2205 int conflict_marker_size;
2206 struct diff_options *o;
2207 unsigned ws_rule;
2208 unsigned status;
2209};
2210
2211static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2212{
2213 char firstchar;
2214 int cnt;
2215
2216 if (len < marker_size + 1)
2217 return 0;
2218 firstchar = line[0];
2219 switch (firstchar) {
2220 case '=': case '>': case '<': case '|':
2221 break;
2222 default:
2223 return 0;
2224 }
2225 for (cnt = 1; cnt < marker_size; cnt++)
2226 if (line[cnt] != firstchar)
2227 return 0;
2228 /* line[1] thru line[marker_size-1] are same as firstchar */
2229 if (len < marker_size + 1 || !isspace(line[marker_size]))
2230 return 0;
2231 return 1;
2232}
2233
2234static void checkdiff_consume(void *priv, char *line, unsigned long len)
2235{
2236 struct checkdiff_t *data = priv;
2237 int marker_size = data->conflict_marker_size;
2238 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2239 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2240 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2241 char *err;
2242 const char *line_prefix;
2243
2244 assert(data->o);
2245 line_prefix = diff_line_prefix(data->o);
2246
2247 if (line[0] == '+') {
2248 unsigned bad;
2249 data->lineno++;
2250 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2251 data->status |= 1;
2252 fprintf(data->o->file,
2253 "%s%s:%d: leftover conflict marker\n",
2254 line_prefix, data->filename, data->lineno);
2255 }
2256 bad = ws_check(line + 1, len - 1, data->ws_rule);
2257 if (!bad)
2258 return;
2259 data->status |= bad;
2260 err = whitespace_error_string(bad);
2261 fprintf(data->o->file, "%s%s:%d: %s.\n",
2262 line_prefix, data->filename, data->lineno, err);
2263 free(err);
2264 emit_line(data->o, set, reset, line, 1);
2265 ws_check_emit(line + 1, len - 1, data->ws_rule,
2266 data->o->file, set, reset, ws);
2267 } else if (line[0] == ' ') {
2268 data->lineno++;
2269 } else if (line[0] == '@') {
2270 char *plus = strchr(line, '+');
2271 if (plus)
2272 data->lineno = strtol(plus, NULL, 10) - 1;
2273 else
2274 die("invalid diff");
2275 }
2276}
2277
2278static unsigned char *deflate_it(char *data,
2279 unsigned long size,
2280 unsigned long *result_size)
2281{
2282 int bound;
2283 unsigned char *deflated;
2284 git_zstream stream;
2285
2286 git_deflate_init(&stream, zlib_compression_level);
2287 bound = git_deflate_bound(&stream, size);
2288 deflated = xmalloc(bound);
2289 stream.next_out = deflated;
2290 stream.avail_out = bound;
2291
2292 stream.next_in = (unsigned char *)data;
2293 stream.avail_in = size;
2294 while (git_deflate(&stream, Z_FINISH) == Z_OK)
2295 ; /* nothing */
2296 git_deflate_end(&stream);
2297 *result_size = stream.total_out;
2298 return deflated;
2299}
2300
2301static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two,
2302 const char *prefix)
2303{
2304 void *cp;
2305 void *delta;
2306 void *deflated;
2307 void *data;
2308 unsigned long orig_size;
2309 unsigned long delta_size;
2310 unsigned long deflate_size;
2311 unsigned long data_size;
2312
2313 /* We could do deflated delta, or we could do just deflated two,
2314 * whichever is smaller.
2315 */
2316 delta = NULL;
2317 deflated = deflate_it(two->ptr, two->size, &deflate_size);
2318 if (one->size && two->size) {
2319 delta = diff_delta(one->ptr, one->size,
2320 two->ptr, two->size,
2321 &delta_size, deflate_size);
2322 if (delta) {
2323 void *to_free = delta;
2324 orig_size = delta_size;
2325 delta = deflate_it(delta, delta_size, &delta_size);
2326 free(to_free);
2327 }
2328 }
2329
2330 if (delta && delta_size < deflate_size) {
2331 fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2332 free(deflated);
2333 data = delta;
2334 data_size = delta_size;
2335 }
2336 else {
2337 fprintf(file, "%sliteral %lu\n", prefix, two->size);
2338 free(delta);
2339 data = deflated;
2340 data_size = deflate_size;
2341 }
2342
2343 /* emit data encoded in base85 */
2344 cp = data;
2345 while (data_size) {
2346 int bytes = (52 < data_size) ? 52 : data_size;
2347 char line[70];
2348 data_size -= bytes;
2349 if (bytes <= 26)
2350 line[0] = bytes + 'A' - 1;
2351 else
2352 line[0] = bytes - 26 + 'a' - 1;
2353 encode_85(line + 1, cp, bytes);
2354 cp = (char *) cp + bytes;
2355 fprintf(file, "%s", prefix);
2356 fputs(line, file);
2357 fputc('\n', file);
2358 }
2359 fprintf(file, "%s\n", prefix);
2360 free(data);
2361}
2362
2363static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two,
2364 const char *prefix)
2365{
2366 fprintf(file, "%sGIT binary patch\n", prefix);
2367 emit_binary_diff_body(file, one, two, prefix);
2368 emit_binary_diff_body(file, two, one, prefix);
2369}
2370
2371int diff_filespec_is_binary(struct diff_filespec *one)
2372{
2373 if (one->is_binary == -1) {
2374 diff_filespec_load_driver(one);
2375 if (one->driver->binary != -1)
2376 one->is_binary = one->driver->binary;
2377 else {
2378 if (!one->data && DIFF_FILE_VALID(one))
2379 diff_populate_filespec(one, CHECK_BINARY);
2380 if (one->is_binary == -1 && one->data)
2381 one->is_binary = buffer_is_binary(one->data,
2382 one->size);
2383 if (one->is_binary == -1)
2384 one->is_binary = 0;
2385 }
2386 }
2387 return one->is_binary;
2388}
2389
2390static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2391{
2392 diff_filespec_load_driver(one);
2393 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2394}
2395
2396void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2397{
2398 if (!options->a_prefix)
2399 options->a_prefix = a;
2400 if (!options->b_prefix)
2401 options->b_prefix = b;
2402}
2403
2404struct userdiff_driver *get_textconv(struct diff_filespec *one)
2405{
2406 if (!DIFF_FILE_VALID(one))
2407 return NULL;
2408
2409 diff_filespec_load_driver(one);
2410 return userdiff_get_textconv(one->driver);
2411}
2412
2413static void builtin_diff(const char *name_a,
2414 const char *name_b,
2415 struct diff_filespec *one,
2416 struct diff_filespec *two,
2417 const char *xfrm_msg,
2418 int must_show_header,
2419 struct diff_options *o,
2420 int complete_rewrite)
2421{
2422 mmfile_t mf1, mf2;
2423 const char *lbl[2];
2424 char *a_one, *b_two;
2425 const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
2426 const char *reset = diff_get_color_opt(o, DIFF_RESET);
2427 const char *a_prefix, *b_prefix;
2428 struct userdiff_driver *textconv_one = NULL;
2429 struct userdiff_driver *textconv_two = NULL;
2430 struct strbuf header = STRBUF_INIT;
2431 const char *line_prefix = diff_line_prefix(o);
2432
2433 diff_set_mnemonic_prefix(o, "a/", "b/");
2434 if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2435 a_prefix = o->b_prefix;
2436 b_prefix = o->a_prefix;
2437 } else {
2438 a_prefix = o->a_prefix;
2439 b_prefix = o->b_prefix;
2440 }
2441
2442 if (o->submodule_format == DIFF_SUBMODULE_LOG &&
2443 (!one->mode || S_ISGITLINK(one->mode)) &&
2444 (!two->mode || S_ISGITLINK(two->mode))) {
2445 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2446 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2447 show_submodule_summary(o->file, one->path ? one->path : two->path,
2448 line_prefix,
2449 &one->oid, &two->oid,
2450 two->dirty_submodule,
2451 meta, del, add, reset);
2452 return;
2453 } else if (o->submodule_format == DIFF_SUBMODULE_INLINE_DIFF &&
2454 (!one->mode || S_ISGITLINK(one->mode)) &&
2455 (!two->mode || S_ISGITLINK(two->mode))) {
2456 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2457 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2458 show_submodule_inline_diff(o->file, one->path ? one->path : two->path,
2459 line_prefix,
2460 &one->oid, &two->oid,
2461 two->dirty_submodule,
2462 meta, del, add, reset, o);
2463 return;
2464 }
2465
2466 if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2467 textconv_one = get_textconv(one);
2468 textconv_two = get_textconv(two);
2469 }
2470
2471 /* Never use a non-valid filename anywhere if at all possible */
2472 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2473 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2474
2475 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2476 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2477 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2478 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2479 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
2480 if (lbl[0][0] == '/') {
2481 /* /dev/null */
2482 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
2483 if (xfrm_msg)
2484 strbuf_addstr(&header, xfrm_msg);
2485 must_show_header = 1;
2486 }
2487 else if (lbl[1][0] == '/') {
2488 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
2489 if (xfrm_msg)
2490 strbuf_addstr(&header, xfrm_msg);
2491 must_show_header = 1;
2492 }
2493 else {
2494 if (one->mode != two->mode) {
2495 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
2496 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
2497 must_show_header = 1;
2498 }
2499 if (xfrm_msg)
2500 strbuf_addstr(&header, xfrm_msg);
2501
2502 /*
2503 * we do not run diff between different kind
2504 * of objects.
2505 */
2506 if ((one->mode ^ two->mode) & S_IFMT)
2507 goto free_ab_and_return;
2508 if (complete_rewrite &&
2509 (textconv_one || !diff_filespec_is_binary(one)) &&
2510 (textconv_two || !diff_filespec_is_binary(two))) {
2511 fprintf(o->file, "%s", header.buf);
2512 strbuf_reset(&header);
2513 emit_rewrite_diff(name_a, name_b, one, two,
2514 textconv_one, textconv_two, o);
2515 o->found_changes = 1;
2516 goto free_ab_and_return;
2517 }
2518 }
2519
2520 if (o->irreversible_delete && lbl[1][0] == '/') {
2521 fprintf(o->file, "%s", header.buf);
2522 strbuf_reset(&header);
2523 goto free_ab_and_return;
2524 } else if (!DIFF_OPT_TST(o, TEXT) &&
2525 ( (!textconv_one && diff_filespec_is_binary(one)) ||
2526 (!textconv_two && diff_filespec_is_binary(two)) )) {
2527 if (!one->data && !two->data &&
2528 S_ISREG(one->mode) && S_ISREG(two->mode) &&
2529 !DIFF_OPT_TST(o, BINARY)) {
2530 if (!oidcmp(&one->oid, &two->oid)) {
2531 if (must_show_header)
2532 fprintf(o->file, "%s", header.buf);
2533 goto free_ab_and_return;
2534 }
2535 fprintf(o->file, "%s", header.buf);
2536 fprintf(o->file, "%sBinary files %s and %s differ\n",
2537 line_prefix, lbl[0], lbl[1]);
2538 goto free_ab_and_return;
2539 }
2540 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2541 die("unable to read files to diff");
2542 /* Quite common confusing case */
2543 if (mf1.size == mf2.size &&
2544 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2545 if (must_show_header)
2546 fprintf(o->file, "%s", header.buf);
2547 goto free_ab_and_return;
2548 }
2549 fprintf(o->file, "%s", header.buf);
2550 strbuf_reset(&header);
2551 if (DIFF_OPT_TST(o, BINARY))
2552 emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2553 else
2554 fprintf(o->file, "%sBinary files %s and %s differ\n",
2555 line_prefix, lbl[0], lbl[1]);
2556 o->found_changes = 1;
2557 } else {
2558 /* Crazy xdl interfaces.. */
2559 const char *diffopts = getenv("GIT_DIFF_OPTS");
2560 const char *v;
2561 xpparam_t xpp;
2562 xdemitconf_t xecfg;
2563 struct emit_callback ecbdata;
2564 const struct userdiff_funcname *pe;
2565
2566 if (must_show_header) {
2567 fprintf(o->file, "%s", header.buf);
2568 strbuf_reset(&header);
2569 }
2570
2571 mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2572 mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2573
2574 pe = diff_funcname_pattern(one);
2575 if (!pe)
2576 pe = diff_funcname_pattern(two);
2577
2578 memset(&xpp, 0, sizeof(xpp));
2579 memset(&xecfg, 0, sizeof(xecfg));
2580 memset(&ecbdata, 0, sizeof(ecbdata));
2581 ecbdata.label_path = lbl;
2582 ecbdata.color_diff = want_color(o->use_color);
2583 ecbdata.ws_rule = whitespace_rule(name_b);
2584 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2585 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2586 ecbdata.opt = o;
2587 ecbdata.header = header.len ? &header : NULL;
2588 xpp.flags = o->xdl_opts;
2589 xecfg.ctxlen = o->context;
2590 xecfg.interhunkctxlen = o->interhunkcontext;
2591 xecfg.flags = XDL_EMIT_FUNCNAMES;
2592 if (DIFF_OPT_TST(o, FUNCCONTEXT))
2593 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2594 if (pe)
2595 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2596 if (!diffopts)
2597 ;
2598 else if (skip_prefix(diffopts, "--unified=", &v))
2599 xecfg.ctxlen = strtoul(v, NULL, 10);
2600 else if (skip_prefix(diffopts, "-u", &v))
2601 xecfg.ctxlen = strtoul(v, NULL, 10);
2602 if (o->word_diff)
2603 init_diff_words_data(&ecbdata, o, one, two);
2604 if (xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2605 &xpp, &xecfg))
2606 die("unable to generate diff for %s", one->path);
2607 if (o->word_diff)
2608 free_diff_words_data(&ecbdata);
2609 if (textconv_one)
2610 free(mf1.ptr);
2611 if (textconv_two)
2612 free(mf2.ptr);
2613 xdiff_clear_find_func(&xecfg);
2614 }
2615
2616 free_ab_and_return:
2617 strbuf_release(&header);
2618 diff_free_filespec_data(one);
2619 diff_free_filespec_data(two);
2620 free(a_one);
2621 free(b_two);
2622 return;
2623}
2624
2625static void builtin_diffstat(const char *name_a, const char *name_b,
2626 struct diff_filespec *one,
2627 struct diff_filespec *two,
2628 struct diffstat_t *diffstat,
2629 struct diff_options *o,
2630 struct diff_filepair *p)
2631{
2632 mmfile_t mf1, mf2;
2633 struct diffstat_file *data;
2634 int same_contents;
2635 int complete_rewrite = 0;
2636
2637 if (!DIFF_PAIR_UNMERGED(p)) {
2638 if (p->status == DIFF_STATUS_MODIFIED && p->score)
2639 complete_rewrite = 1;
2640 }
2641
2642 data = diffstat_add(diffstat, name_a, name_b);
2643 data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
2644
2645 if (!one || !two) {
2646 data->is_unmerged = 1;
2647 return;
2648 }
2649
2650 same_contents = !oidcmp(&one->oid, &two->oid);
2651
2652 if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2653 data->is_binary = 1;
2654 if (same_contents) {
2655 data->added = 0;
2656 data->deleted = 0;
2657 } else {
2658 data->added = diff_filespec_size(two);
2659 data->deleted = diff_filespec_size(one);
2660 }
2661 }
2662
2663 else if (complete_rewrite) {
2664 diff_populate_filespec(one, 0);
2665 diff_populate_filespec(two, 0);
2666 data->deleted = count_lines(one->data, one->size);
2667 data->added = count_lines(two->data, two->size);
2668 }
2669
2670 else if (!same_contents) {
2671 /* Crazy xdl interfaces.. */
2672 xpparam_t xpp;
2673 xdemitconf_t xecfg;
2674
2675 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2676 die("unable to read files to diff");
2677
2678 memset(&xpp, 0, sizeof(xpp));
2679 memset(&xecfg, 0, sizeof(xecfg));
2680 xpp.flags = o->xdl_opts;
2681 xecfg.ctxlen = o->context;
2682 xecfg.interhunkctxlen = o->interhunkcontext;
2683 if (xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2684 &xpp, &xecfg))
2685 die("unable to generate diffstat for %s", one->path);
2686 }
2687
2688 diff_free_filespec_data(one);
2689 diff_free_filespec_data(two);
2690}
2691
2692static void builtin_checkdiff(const char *name_a, const char *name_b,
2693 const char *attr_path,
2694 struct diff_filespec *one,
2695 struct diff_filespec *two,
2696 struct diff_options *o)
2697{
2698 mmfile_t mf1, mf2;
2699 struct checkdiff_t data;
2700
2701 if (!two)
2702 return;
2703
2704 memset(&data, 0, sizeof(data));
2705 data.filename = name_b ? name_b : name_a;
2706 data.lineno = 0;
2707 data.o = o;
2708 data.ws_rule = whitespace_rule(attr_path);
2709 data.conflict_marker_size = ll_merge_marker_size(attr_path);
2710
2711 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2712 die("unable to read files to diff");
2713
2714 /*
2715 * All the other codepaths check both sides, but not checking
2716 * the "old" side here is deliberate. We are checking the newly
2717 * introduced changes, and as long as the "new" side is text, we
2718 * can and should check what it introduces.
2719 */
2720 if (diff_filespec_is_binary(two))
2721 goto free_and_return;
2722 else {
2723 /* Crazy xdl interfaces.. */
2724 xpparam_t xpp;
2725 xdemitconf_t xecfg;
2726
2727 memset(&xpp, 0, sizeof(xpp));
2728 memset(&xecfg, 0, sizeof(xecfg));
2729 xecfg.ctxlen = 1; /* at least one context line */
2730 xpp.flags = 0;
2731 if (xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2732 &xpp, &xecfg))
2733 die("unable to generate checkdiff for %s", one->path);
2734
2735 if (data.ws_rule & WS_BLANK_AT_EOF) {
2736 struct emit_callback ecbdata;
2737 int blank_at_eof;
2738
2739 ecbdata.ws_rule = data.ws_rule;
2740 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2741 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2742
2743 if (blank_at_eof) {
2744 static char *err;
2745 if (!err)
2746 err = whitespace_error_string(WS_BLANK_AT_EOF);
2747 fprintf(o->file, "%s:%d: %s.\n",
2748 data.filename, blank_at_eof, err);
2749 data.status = 1; /* report errors */
2750 }
2751 }
2752 }
2753 free_and_return:
2754 diff_free_filespec_data(one);
2755 diff_free_filespec_data(two);
2756 if (data.status)
2757 DIFF_OPT_SET(o, CHECK_FAILED);
2758}
2759
2760struct diff_filespec *alloc_filespec(const char *path)
2761{
2762 struct diff_filespec *spec;
2763
2764 FLEXPTR_ALLOC_STR(spec, path, path);
2765 spec->count = 1;
2766 spec->is_binary = -1;
2767 return spec;
2768}
2769
2770void free_filespec(struct diff_filespec *spec)
2771{
2772 if (!--spec->count) {
2773 diff_free_filespec_data(spec);
2774 free(spec);
2775 }
2776}
2777
2778void fill_filespec(struct diff_filespec *spec, const struct object_id *oid,
2779 int oid_valid, unsigned short mode)
2780{
2781 if (mode) {
2782 spec->mode = canon_mode(mode);
2783 oidcpy(&spec->oid, oid);
2784 spec->oid_valid = oid_valid;
2785 }
2786}
2787
2788/*
2789 * Given a name and sha1 pair, if the index tells us the file in
2790 * the work tree has that object contents, return true, so that
2791 * prepare_temp_file() does not have to inflate and extract.
2792 */
2793static int reuse_worktree_file(const char *name, const struct object_id *oid, int want_file)
2794{
2795 const struct cache_entry *ce;
2796 struct stat st;
2797 int pos, len;
2798
2799 /*
2800 * We do not read the cache ourselves here, because the
2801 * benchmark with my previous version that always reads cache
2802 * shows that it makes things worse for diff-tree comparing
2803 * two linux-2.6 kernel trees in an already checked out work
2804 * tree. This is because most diff-tree comparisons deal with
2805 * only a small number of files, while reading the cache is
2806 * expensive for a large project, and its cost outweighs the
2807 * savings we get by not inflating the object to a temporary
2808 * file. Practically, this code only helps when we are used
2809 * by diff-cache --cached, which does read the cache before
2810 * calling us.
2811 */
2812 if (!active_cache)
2813 return 0;
2814
2815 /* We want to avoid the working directory if our caller
2816 * doesn't need the data in a normal file, this system
2817 * is rather slow with its stat/open/mmap/close syscalls,
2818 * and the object is contained in a pack file. The pack
2819 * is probably already open and will be faster to obtain
2820 * the data through than the working directory. Loose
2821 * objects however would tend to be slower as they need
2822 * to be individually opened and inflated.
2823 */
2824 if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(oid->hash))
2825 return 0;
2826
2827 /*
2828 * Similarly, if we'd have to convert the file contents anyway, that
2829 * makes the optimization not worthwhile.
2830 */
2831 if (!want_file && would_convert_to_git(&the_index, name))
2832 return 0;
2833
2834 len = strlen(name);
2835 pos = cache_name_pos(name, len);
2836 if (pos < 0)
2837 return 0;
2838 ce = active_cache[pos];
2839
2840 /*
2841 * This is not the sha1 we are looking for, or
2842 * unreusable because it is not a regular file.
2843 */
2844 if (oidcmp(oid, &ce->oid) || !S_ISREG(ce->ce_mode))
2845 return 0;
2846
2847 /*
2848 * If ce is marked as "assume unchanged", there is no
2849 * guarantee that work tree matches what we are looking for.
2850 */
2851 if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2852 return 0;
2853
2854 /*
2855 * If ce matches the file in the work tree, we can reuse it.
2856 */
2857 if (ce_uptodate(ce) ||
2858 (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2859 return 1;
2860
2861 return 0;
2862}
2863
2864static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2865{
2866 struct strbuf buf = STRBUF_INIT;
2867 char *dirty = "";
2868
2869 /* Are we looking at the work tree? */
2870 if (s->dirty_submodule)
2871 dirty = "-dirty";
2872
2873 strbuf_addf(&buf, "Subproject commit %s%s\n",
2874 oid_to_hex(&s->oid), dirty);
2875 s->size = buf.len;
2876 if (size_only) {
2877 s->data = NULL;
2878 strbuf_release(&buf);
2879 } else {
2880 s->data = strbuf_detach(&buf, NULL);
2881 s->should_free = 1;
2882 }
2883 return 0;
2884}
2885
2886/*
2887 * While doing rename detection and pickaxe operation, we may need to
2888 * grab the data for the blob (or file) for our own in-core comparison.
2889 * diff_filespec has data and size fields for this purpose.
2890 */
2891int diff_populate_filespec(struct diff_filespec *s, unsigned int flags)
2892{
2893 int size_only = flags & CHECK_SIZE_ONLY;
2894 int err = 0;
2895 /*
2896 * demote FAIL to WARN to allow inspecting the situation
2897 * instead of refusing.
2898 */
2899 enum safe_crlf crlf_warn = (safe_crlf == SAFE_CRLF_FAIL
2900 ? SAFE_CRLF_WARN
2901 : safe_crlf);
2902
2903 if (!DIFF_FILE_VALID(s))
2904 die("internal error: asking to populate invalid file.");
2905 if (S_ISDIR(s->mode))
2906 return -1;
2907
2908 if (s->data)
2909 return 0;
2910
2911 if (size_only && 0 < s->size)
2912 return 0;
2913
2914 if (S_ISGITLINK(s->mode))
2915 return diff_populate_gitlink(s, size_only);
2916
2917 if (!s->oid_valid ||
2918 reuse_worktree_file(s->path, &s->oid, 0)) {
2919 struct strbuf buf = STRBUF_INIT;
2920 struct stat st;
2921 int fd;
2922
2923 if (lstat(s->path, &st) < 0) {
2924 if (errno == ENOENT) {
2925 err_empty:
2926 err = -1;
2927 empty:
2928 s->data = (char *)"";
2929 s->size = 0;
2930 return err;
2931 }
2932 }
2933 s->size = xsize_t(st.st_size);
2934 if (!s->size)
2935 goto empty;
2936 if (S_ISLNK(st.st_mode)) {
2937 struct strbuf sb = STRBUF_INIT;
2938
2939 if (strbuf_readlink(&sb, s->path, s->size))
2940 goto err_empty;
2941 s->size = sb.len;
2942 s->data = strbuf_detach(&sb, NULL);
2943 s->should_free = 1;
2944 return 0;
2945 }
2946
2947 /*
2948 * Even if the caller would be happy with getting
2949 * only the size, we cannot return early at this
2950 * point if the path requires us to run the content
2951 * conversion.
2952 */
2953 if (size_only && !would_convert_to_git(&the_index, s->path))
2954 return 0;
2955
2956 /*
2957 * Note: this check uses xsize_t(st.st_size) that may
2958 * not be the true size of the blob after it goes
2959 * through convert_to_git(). This may not strictly be
2960 * correct, but the whole point of big_file_threshold
2961 * and is_binary check being that we want to avoid
2962 * opening the file and inspecting the contents, this
2963 * is probably fine.
2964 */
2965 if ((flags & CHECK_BINARY) &&
2966 s->size > big_file_threshold && s->is_binary == -1) {
2967 s->is_binary = 1;
2968 return 0;
2969 }
2970 fd = open(s->path, O_RDONLY);
2971 if (fd < 0)
2972 goto err_empty;
2973 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2974 close(fd);
2975 s->should_munmap = 1;
2976
2977 /*
2978 * Convert from working tree format to canonical git format
2979 */
2980 if (convert_to_git(&the_index, s->path, s->data, s->size, &buf, crlf_warn)) {
2981 size_t size = 0;
2982 munmap(s->data, s->size);
2983 s->should_munmap = 0;
2984 s->data = strbuf_detach(&buf, &size);
2985 s->size = size;
2986 s->should_free = 1;
2987 }
2988 }
2989 else {
2990 enum object_type type;
2991 if (size_only || (flags & CHECK_BINARY)) {
2992 type = sha1_object_info(s->oid.hash, &s->size);
2993 if (type < 0)
2994 die("unable to read %s",
2995 oid_to_hex(&s->oid));
2996 if (size_only)
2997 return 0;
2998 if (s->size > big_file_threshold && s->is_binary == -1) {
2999 s->is_binary = 1;
3000 return 0;
3001 }
3002 }
3003 s->data = read_sha1_file(s->oid.hash, &type, &s->size);
3004 if (!s->data)
3005 die("unable to read %s", oid_to_hex(&s->oid));
3006 s->should_free = 1;
3007 }
3008 return 0;
3009}
3010
3011void diff_free_filespec_blob(struct diff_filespec *s)
3012{
3013 if (s->should_free)
3014 free(s->data);
3015 else if (s->should_munmap)
3016 munmap(s->data, s->size);
3017
3018 if (s->should_free || s->should_munmap) {
3019 s->should_free = s->should_munmap = 0;
3020 s->data = NULL;
3021 }
3022}
3023
3024void diff_free_filespec_data(struct diff_filespec *s)
3025{
3026 diff_free_filespec_blob(s);
3027 FREE_AND_NULL(s->cnt_data);
3028}
3029
3030static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
3031 void *blob,
3032 unsigned long size,
3033 const struct object_id *oid,
3034 int mode)
3035{
3036 int fd;
3037 struct strbuf buf = STRBUF_INIT;
3038 struct strbuf template = STRBUF_INIT;
3039 char *path_dup = xstrdup(path);
3040 const char *base = basename(path_dup);
3041
3042 /* Generate "XXXXXX_basename.ext" */
3043 strbuf_addstr(&template, "XXXXXX_");
3044 strbuf_addstr(&template, base);
3045
3046 fd = mks_tempfile_ts(&temp->tempfile, template.buf, strlen(base) + 1);
3047 if (fd < 0)
3048 die_errno("unable to create temp-file");
3049 if (convert_to_working_tree(path,
3050 (const char *)blob, (size_t)size, &buf)) {
3051 blob = buf.buf;
3052 size = buf.len;
3053 }
3054 if (write_in_full(fd, blob, size) != size)
3055 die_errno("unable to write temp-file");
3056 close_tempfile(&temp->tempfile);
3057 temp->name = get_tempfile_path(&temp->tempfile);
3058 oid_to_hex_r(temp->hex, oid);
3059 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", mode);
3060 strbuf_release(&buf);
3061 strbuf_release(&template);
3062 free(path_dup);
3063}
3064
3065static struct diff_tempfile *prepare_temp_file(const char *name,
3066 struct diff_filespec *one)
3067{
3068 struct diff_tempfile *temp = claim_diff_tempfile();
3069
3070 if (!DIFF_FILE_VALID(one)) {
3071 not_a_valid_file:
3072 /* A '-' entry produces this for file-2, and
3073 * a '+' entry produces this for file-1.
3074 */
3075 temp->name = "/dev/null";
3076 xsnprintf(temp->hex, sizeof(temp->hex), ".");
3077 xsnprintf(temp->mode, sizeof(temp->mode), ".");
3078 return temp;
3079 }
3080
3081 if (!S_ISGITLINK(one->mode) &&
3082 (!one->oid_valid ||
3083 reuse_worktree_file(name, &one->oid, 1))) {
3084 struct stat st;
3085 if (lstat(name, &st) < 0) {
3086 if (errno == ENOENT)
3087 goto not_a_valid_file;
3088 die_errno("stat(%s)", name);
3089 }
3090 if (S_ISLNK(st.st_mode)) {
3091 struct strbuf sb = STRBUF_INIT;
3092 if (strbuf_readlink(&sb, name, st.st_size) < 0)
3093 die_errno("readlink(%s)", name);
3094 prep_temp_blob(name, temp, sb.buf, sb.len,
3095 (one->oid_valid ?
3096 &one->oid : &null_oid),
3097 (one->oid_valid ?
3098 one->mode : S_IFLNK));
3099 strbuf_release(&sb);
3100 }
3101 else {
3102 /* we can borrow from the file in the work tree */
3103 temp->name = name;
3104 if (!one->oid_valid)
3105 oid_to_hex_r(temp->hex, &null_oid);
3106 else
3107 oid_to_hex_r(temp->hex, &one->oid);
3108 /* Even though we may sometimes borrow the
3109 * contents from the work tree, we always want
3110 * one->mode. mode is trustworthy even when
3111 * !(one->oid_valid), as long as
3112 * DIFF_FILE_VALID(one).
3113 */
3114 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", one->mode);
3115 }
3116 return temp;
3117 }
3118 else {
3119 if (diff_populate_filespec(one, 0))
3120 die("cannot read data blob for %s", one->path);
3121 prep_temp_blob(name, temp, one->data, one->size,
3122 &one->oid, one->mode);
3123 }
3124 return temp;
3125}
3126
3127static void add_external_diff_name(struct argv_array *argv,
3128 const char *name,
3129 struct diff_filespec *df)
3130{
3131 struct diff_tempfile *temp = prepare_temp_file(name, df);
3132 argv_array_push(argv, temp->name);
3133 argv_array_push(argv, temp->hex);
3134 argv_array_push(argv, temp->mode);
3135}
3136
3137/* An external diff command takes:
3138 *
3139 * diff-cmd name infile1 infile1-sha1 infile1-mode \
3140 * infile2 infile2-sha1 infile2-mode [ rename-to ]
3141 *
3142 */
3143static void run_external_diff(const char *pgm,
3144 const char *name,
3145 const char *other,
3146 struct diff_filespec *one,
3147 struct diff_filespec *two,
3148 const char *xfrm_msg,
3149 int complete_rewrite,
3150 struct diff_options *o)
3151{
3152 struct argv_array argv = ARGV_ARRAY_INIT;
3153 struct argv_array env = ARGV_ARRAY_INIT;
3154 struct diff_queue_struct *q = &diff_queued_diff;
3155
3156 argv_array_push(&argv, pgm);
3157 argv_array_push(&argv, name);
3158
3159 if (one && two) {
3160 add_external_diff_name(&argv, name, one);
3161 if (!other)
3162 add_external_diff_name(&argv, name, two);
3163 else {
3164 add_external_diff_name(&argv, other, two);
3165 argv_array_push(&argv, other);
3166 argv_array_push(&argv, xfrm_msg);
3167 }
3168 }
3169
3170 argv_array_pushf(&env, "GIT_DIFF_PATH_COUNTER=%d", ++o->diff_path_counter);
3171 argv_array_pushf(&env, "GIT_DIFF_PATH_TOTAL=%d", q->nr);
3172
3173 if (run_command_v_opt_cd_env(argv.argv, RUN_USING_SHELL, NULL, env.argv))
3174 die(_("external diff died, stopping at %s"), name);
3175
3176 remove_tempfile();
3177 argv_array_clear(&argv);
3178 argv_array_clear(&env);
3179}
3180
3181static int similarity_index(struct diff_filepair *p)
3182{
3183 return p->score * 100 / MAX_SCORE;
3184}
3185
3186static const char *diff_abbrev_oid(const struct object_id *oid, int abbrev)
3187{
3188 if (startup_info->have_repository)
3189 return find_unique_abbrev(oid->hash, abbrev);
3190 else {
3191 char *hex = oid_to_hex(oid);
3192 if (abbrev < 0)
3193 abbrev = FALLBACK_DEFAULT_ABBREV;
3194 if (abbrev > GIT_SHA1_HEXSZ)
3195 die("BUG: oid abbreviation out of range: %d", abbrev);
3196 if (abbrev)
3197 hex[abbrev] = '\0';
3198 return hex;
3199 }
3200}
3201
3202static void fill_metainfo(struct strbuf *msg,
3203 const char *name,
3204 const char *other,
3205 struct diff_filespec *one,
3206 struct diff_filespec *two,
3207 struct diff_options *o,
3208 struct diff_filepair *p,
3209 int *must_show_header,
3210 int use_color)
3211{
3212 const char *set = diff_get_color(use_color, DIFF_METAINFO);
3213 const char *reset = diff_get_color(use_color, DIFF_RESET);
3214 const char *line_prefix = diff_line_prefix(o);
3215
3216 *must_show_header = 1;
3217 strbuf_init(msg, PATH_MAX * 2 + 300);
3218 switch (p->status) {
3219 case DIFF_STATUS_COPIED:
3220 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3221 line_prefix, set, similarity_index(p));
3222 strbuf_addf(msg, "%s\n%s%scopy from ",
3223 reset, line_prefix, set);
3224 quote_c_style(name, msg, NULL, 0);
3225 strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
3226 quote_c_style(other, msg, NULL, 0);
3227 strbuf_addf(msg, "%s\n", reset);
3228 break;
3229 case DIFF_STATUS_RENAMED:
3230 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3231 line_prefix, set, similarity_index(p));
3232 strbuf_addf(msg, "%s\n%s%srename from ",
3233 reset, line_prefix, set);
3234 quote_c_style(name, msg, NULL, 0);
3235 strbuf_addf(msg, "%s\n%s%srename to ",
3236 reset, line_prefix, set);
3237 quote_c_style(other, msg, NULL, 0);
3238 strbuf_addf(msg, "%s\n", reset);
3239 break;
3240 case DIFF_STATUS_MODIFIED:
3241 if (p->score) {
3242 strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
3243 line_prefix,
3244 set, similarity_index(p), reset);
3245 break;
3246 }
3247 /* fallthru */
3248 default:
3249 *must_show_header = 0;
3250 }
3251 if (one && two && oidcmp(&one->oid, &two->oid)) {
3252 int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
3253
3254 if (DIFF_OPT_TST(o, BINARY)) {
3255 mmfile_t mf;
3256 if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
3257 (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
3258 abbrev = 40;
3259 }
3260 strbuf_addf(msg, "%s%sindex %s..%s", line_prefix, set,
3261 diff_abbrev_oid(&one->oid, abbrev),
3262 diff_abbrev_oid(&two->oid, abbrev));
3263 if (one->mode == two->mode)
3264 strbuf_addf(msg, " %06o", one->mode);
3265 strbuf_addf(msg, "%s\n", reset);
3266 }
3267}
3268
3269static void run_diff_cmd(const char *pgm,
3270 const char *name,
3271 const char *other,
3272 const char *attr_path,
3273 struct diff_filespec *one,
3274 struct diff_filespec *two,
3275 struct strbuf *msg,
3276 struct diff_options *o,
3277 struct diff_filepair *p)
3278{
3279 const char *xfrm_msg = NULL;
3280 int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
3281 int must_show_header = 0;
3282
3283
3284 if (DIFF_OPT_TST(o, ALLOW_EXTERNAL)) {
3285 struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
3286 if (drv && drv->external)
3287 pgm = drv->external;
3288 }
3289
3290 if (msg) {
3291 /*
3292 * don't use colors when the header is intended for an
3293 * external diff driver
3294 */
3295 fill_metainfo(msg, name, other, one, two, o, p,
3296 &must_show_header,
3297 want_color(o->use_color) && !pgm);
3298 xfrm_msg = msg->len ? msg->buf : NULL;
3299 }
3300
3301 if (pgm) {
3302 run_external_diff(pgm, name, other, one, two, xfrm_msg,
3303 complete_rewrite, o);
3304 return;
3305 }
3306 if (one && two)
3307 builtin_diff(name, other ? other : name,
3308 one, two, xfrm_msg, must_show_header,
3309 o, complete_rewrite);
3310 else
3311 fprintf(o->file, "* Unmerged path %s\n", name);
3312}
3313
3314static void diff_fill_oid_info(struct diff_filespec *one)
3315{
3316 if (DIFF_FILE_VALID(one)) {
3317 if (!one->oid_valid) {
3318 struct stat st;
3319 if (one->is_stdin) {
3320 oidclr(&one->oid);
3321 return;
3322 }
3323 if (lstat(one->path, &st) < 0)
3324 die_errno("stat '%s'", one->path);
3325 if (index_path(one->oid.hash, one->path, &st, 0))
3326 die("cannot hash %s", one->path);
3327 }
3328 }
3329 else
3330 oidclr(&one->oid);
3331}
3332
3333static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
3334{
3335 /* Strip the prefix but do not molest /dev/null and absolute paths */
3336 if (*namep && **namep != '/') {
3337 *namep += prefix_length;
3338 if (**namep == '/')
3339 ++*namep;
3340 }
3341 if (*otherp && **otherp != '/') {
3342 *otherp += prefix_length;
3343 if (**otherp == '/')
3344 ++*otherp;
3345 }
3346}
3347
3348static void run_diff(struct diff_filepair *p, struct diff_options *o)
3349{
3350 const char *pgm = external_diff();
3351 struct strbuf msg;
3352 struct diff_filespec *one = p->one;
3353 struct diff_filespec *two = p->two;
3354 const char *name;
3355 const char *other;
3356 const char *attr_path;
3357
3358 name = one->path;
3359 other = (strcmp(name, two->path) ? two->path : NULL);
3360 attr_path = name;
3361 if (o->prefix_length)
3362 strip_prefix(o->prefix_length, &name, &other);
3363
3364 if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
3365 pgm = NULL;
3366
3367 if (DIFF_PAIR_UNMERGED(p)) {
3368 run_diff_cmd(pgm, name, NULL, attr_path,
3369 NULL, NULL, NULL, o, p);
3370 return;
3371 }
3372
3373 diff_fill_oid_info(one);
3374 diff_fill_oid_info(two);
3375
3376 if (!pgm &&
3377 DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3378 (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3379 /*
3380 * a filepair that changes between file and symlink
3381 * needs to be split into deletion and creation.
3382 */
3383 struct diff_filespec *null = alloc_filespec(two->path);
3384 run_diff_cmd(NULL, name, other, attr_path,
3385 one, null, &msg, o, p);
3386 free(null);
3387 strbuf_release(&msg);
3388
3389 null = alloc_filespec(one->path);
3390 run_diff_cmd(NULL, name, other, attr_path,
3391 null, two, &msg, o, p);
3392 free(null);
3393 }
3394 else
3395 run_diff_cmd(pgm, name, other, attr_path,
3396 one, two, &msg, o, p);
3397
3398 strbuf_release(&msg);
3399}
3400
3401static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3402 struct diffstat_t *diffstat)
3403{
3404 const char *name;
3405 const char *other;
3406
3407 if (DIFF_PAIR_UNMERGED(p)) {
3408 /* unmerged */
3409 builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, p);
3410 return;
3411 }
3412
3413 name = p->one->path;
3414 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3415
3416 if (o->prefix_length)
3417 strip_prefix(o->prefix_length, &name, &other);
3418
3419 diff_fill_oid_info(p->one);
3420 diff_fill_oid_info(p->two);
3421
3422 builtin_diffstat(name, other, p->one, p->two, diffstat, o, p);
3423}
3424
3425static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3426{
3427 const char *name;
3428 const char *other;
3429 const char *attr_path;
3430
3431 if (DIFF_PAIR_UNMERGED(p)) {
3432 /* unmerged */
3433 return;
3434 }
3435
3436 name = p->one->path;
3437 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3438 attr_path = other ? other : name;
3439
3440 if (o->prefix_length)
3441 strip_prefix(o->prefix_length, &name, &other);
3442
3443 diff_fill_oid_info(p->one);
3444 diff_fill_oid_info(p->two);
3445
3446 builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3447}
3448
3449void diff_setup(struct diff_options *options)
3450{
3451 memcpy(options, &default_diff_options, sizeof(*options));
3452
3453 options->file = stdout;
3454
3455 options->abbrev = DEFAULT_ABBREV;
3456 options->line_termination = '\n';
3457 options->break_opt = -1;
3458 options->rename_limit = -1;
3459 options->dirstat_permille = diff_dirstat_permille_default;
3460 options->context = diff_context_default;
3461 options->interhunkcontext = diff_interhunk_context_default;
3462 options->ws_error_highlight = ws_error_highlight_default;
3463 DIFF_OPT_SET(options, RENAME_EMPTY);
3464
3465 /* pathchange left =NULL by default */
3466 options->change = diff_change;
3467 options->add_remove = diff_addremove;
3468 options->use_color = diff_use_color_default;
3469 options->detect_rename = diff_detect_rename_default;
3470 options->xdl_opts |= diff_algorithm;
3471 if (diff_indent_heuristic)
3472 DIFF_XDL_SET(options, INDENT_HEURISTIC);
3473
3474 options->orderfile = diff_order_file_cfg;
3475
3476 if (diff_no_prefix) {
3477 options->a_prefix = options->b_prefix = "";
3478 } else if (!diff_mnemonic_prefix) {
3479 options->a_prefix = "a/";
3480 options->b_prefix = "b/";
3481 }
3482}
3483
3484void diff_setup_done(struct diff_options *options)
3485{
3486 int count = 0;
3487
3488 if (options->set_default)
3489 options->set_default(options);
3490
3491 if (options->output_format & DIFF_FORMAT_NAME)
3492 count++;
3493 if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3494 count++;
3495 if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3496 count++;
3497 if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3498 count++;
3499 if (count > 1)
3500 die(_("--name-only, --name-status, --check and -s are mutually exclusive"));
3501
3502 /*
3503 * Most of the time we can say "there are changes"
3504 * only by checking if there are changed paths, but
3505 * --ignore-whitespace* options force us to look
3506 * inside contents.
3507 */
3508
3509 if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3510 DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3511 DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3512 DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3513 else
3514 DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3515
3516 if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3517 options->detect_rename = DIFF_DETECT_COPY;
3518
3519 if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3520 options->prefix = NULL;
3521 if (options->prefix)
3522 options->prefix_length = strlen(options->prefix);
3523 else
3524 options->prefix_length = 0;
3525
3526 if (options->output_format & (DIFF_FORMAT_NAME |
3527 DIFF_FORMAT_NAME_STATUS |
3528 DIFF_FORMAT_CHECKDIFF |
3529 DIFF_FORMAT_NO_OUTPUT))
3530 options->output_format &= ~(DIFF_FORMAT_RAW |
3531 DIFF_FORMAT_NUMSTAT |
3532 DIFF_FORMAT_DIFFSTAT |
3533 DIFF_FORMAT_SHORTSTAT |
3534 DIFF_FORMAT_DIRSTAT |
3535 DIFF_FORMAT_SUMMARY |
3536 DIFF_FORMAT_PATCH);
3537
3538 /*
3539 * These cases always need recursive; we do not drop caller-supplied
3540 * recursive bits for other formats here.
3541 */
3542 if (options->output_format & (DIFF_FORMAT_PATCH |
3543 DIFF_FORMAT_NUMSTAT |
3544 DIFF_FORMAT_DIFFSTAT |
3545 DIFF_FORMAT_SHORTSTAT |
3546 DIFF_FORMAT_DIRSTAT |
3547 DIFF_FORMAT_SUMMARY |
3548 DIFF_FORMAT_CHECKDIFF))
3549 DIFF_OPT_SET(options, RECURSIVE);
3550 /*
3551 * Also pickaxe would not work very well if you do not say recursive
3552 */
3553 if (options->pickaxe)
3554 DIFF_OPT_SET(options, RECURSIVE);
3555 /*
3556 * When patches are generated, submodules diffed against the work tree
3557 * must be checked for dirtiness too so it can be shown in the output
3558 */
3559 if (options->output_format & DIFF_FORMAT_PATCH)
3560 DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3561
3562 if (options->detect_rename && options->rename_limit < 0)
3563 options->rename_limit = diff_rename_limit_default;
3564 if (options->setup & DIFF_SETUP_USE_CACHE) {
3565 if (!active_cache)
3566 /* read-cache does not die even when it fails
3567 * so it is safe for us to do this here. Also
3568 * it does not smudge active_cache or active_nr
3569 * when it fails, so we do not have to worry about
3570 * cleaning it up ourselves either.
3571 */
3572 read_cache();
3573 }
3574 if (40 < options->abbrev)
3575 options->abbrev = 40; /* full */
3576
3577 /*
3578 * It does not make sense to show the first hit we happened
3579 * to have found. It does not make sense not to return with
3580 * exit code in such a case either.
3581 */
3582 if (DIFF_OPT_TST(options, QUICK)) {
3583 options->output_format = DIFF_FORMAT_NO_OUTPUT;
3584 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3585 }
3586
3587 options->diff_path_counter = 0;
3588
3589 if (DIFF_OPT_TST(options, FOLLOW_RENAMES) && options->pathspec.nr != 1)
3590 die(_("--follow requires exactly one pathspec"));
3591}
3592
3593static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3594{
3595 char c, *eq;
3596 int len;
3597
3598 if (*arg != '-')
3599 return 0;
3600 c = *++arg;
3601 if (!c)
3602 return 0;
3603 if (c == arg_short) {
3604 c = *++arg;
3605 if (!c)
3606 return 1;
3607 if (val && isdigit(c)) {
3608 char *end;
3609 int n = strtoul(arg, &end, 10);
3610 if (*end)
3611 return 0;
3612 *val = n;
3613 return 1;
3614 }
3615 return 0;
3616 }
3617 if (c != '-')
3618 return 0;
3619 arg++;
3620 eq = strchrnul(arg, '=');
3621 len = eq - arg;
3622 if (!len || strncmp(arg, arg_long, len))
3623 return 0;
3624 if (*eq) {
3625 int n;
3626 char *end;
3627 if (!isdigit(*++eq))
3628 return 0;
3629 n = strtoul(eq, &end, 10);
3630 if (*end)
3631 return 0;
3632 *val = n;
3633 }
3634 return 1;
3635}
3636
3637static int diff_scoreopt_parse(const char *opt);
3638
3639static inline int short_opt(char opt, const char **argv,
3640 const char **optarg)
3641{
3642 const char *arg = argv[0];
3643 if (arg[0] != '-' || arg[1] != opt)
3644 return 0;
3645 if (arg[2] != '\0') {
3646 *optarg = arg + 2;
3647 return 1;
3648 }
3649 if (!argv[1])
3650 die("Option '%c' requires a value", opt);
3651 *optarg = argv[1];
3652 return 2;
3653}
3654
3655int parse_long_opt(const char *opt, const char **argv,
3656 const char **optarg)
3657{
3658 const char *arg = argv[0];
3659 if (!skip_prefix(arg, "--", &arg))
3660 return 0;
3661 if (!skip_prefix(arg, opt, &arg))
3662 return 0;
3663 if (*arg == '=') { /* stuck form: --option=value */
3664 *optarg = arg + 1;
3665 return 1;
3666 }
3667 if (*arg != '\0')
3668 return 0;
3669 /* separate form: --option value */
3670 if (!argv[1])
3671 die("Option '--%s' requires a value", opt);
3672 *optarg = argv[1];
3673 return 2;
3674}
3675
3676static int stat_opt(struct diff_options *options, const char **av)
3677{
3678 const char *arg = av[0];
3679 char *end;
3680 int width = options->stat_width;
3681 int name_width = options->stat_name_width;
3682 int graph_width = options->stat_graph_width;
3683 int count = options->stat_count;
3684 int argcount = 1;
3685
3686 if (!skip_prefix(arg, "--stat", &arg))
3687 die("BUG: stat option does not begin with --stat: %s", arg);
3688 end = (char *)arg;
3689
3690 switch (*arg) {
3691 case '-':
3692 if (skip_prefix(arg, "-width", &arg)) {
3693 if (*arg == '=')
3694 width = strtoul(arg + 1, &end, 10);
3695 else if (!*arg && !av[1])
3696 die_want_option("--stat-width");
3697 else if (!*arg) {
3698 width = strtoul(av[1], &end, 10);
3699 argcount = 2;
3700 }
3701 } else if (skip_prefix(arg, "-name-width", &arg)) {
3702 if (*arg == '=')
3703 name_width = strtoul(arg + 1, &end, 10);
3704 else if (!*arg && !av[1])
3705 die_want_option("--stat-name-width");
3706 else if (!*arg) {
3707 name_width = strtoul(av[1], &end, 10);
3708 argcount = 2;
3709 }
3710 } else if (skip_prefix(arg, "-graph-width", &arg)) {
3711 if (*arg == '=')
3712 graph_width = strtoul(arg + 1, &end, 10);
3713 else if (!*arg && !av[1])
3714 die_want_option("--stat-graph-width");
3715 else if (!*arg) {
3716 graph_width = strtoul(av[1], &end, 10);
3717 argcount = 2;
3718 }
3719 } else if (skip_prefix(arg, "-count", &arg)) {
3720 if (*arg == '=')
3721 count = strtoul(arg + 1, &end, 10);
3722 else if (!*arg && !av[1])
3723 die_want_option("--stat-count");
3724 else if (!*arg) {
3725 count = strtoul(av[1], &end, 10);
3726 argcount = 2;
3727 }
3728 }
3729 break;
3730 case '=':
3731 width = strtoul(arg+1, &end, 10);
3732 if (*end == ',')
3733 name_width = strtoul(end+1, &end, 10);
3734 if (*end == ',')
3735 count = strtoul(end+1, &end, 10);
3736 }
3737
3738 /* Important! This checks all the error cases! */
3739 if (*end)
3740 return 0;
3741 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3742 options->stat_name_width = name_width;
3743 options->stat_graph_width = graph_width;
3744 options->stat_width = width;
3745 options->stat_count = count;
3746 return argcount;
3747}
3748
3749static int parse_dirstat_opt(struct diff_options *options, const char *params)
3750{
3751 struct strbuf errmsg = STRBUF_INIT;
3752 if (parse_dirstat_params(options, params, &errmsg))
3753 die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3754 errmsg.buf);
3755 strbuf_release(&errmsg);
3756 /*
3757 * The caller knows a dirstat-related option is given from the command
3758 * line; allow it to say "return this_function();"
3759 */
3760 options->output_format |= DIFF_FORMAT_DIRSTAT;
3761 return 1;
3762}
3763
3764static int parse_submodule_opt(struct diff_options *options, const char *value)
3765{
3766 if (parse_submodule_params(options, value))
3767 die(_("Failed to parse --submodule option parameter: '%s'"),
3768 value);
3769 return 1;
3770}
3771
3772static const char diff_status_letters[] = {
3773 DIFF_STATUS_ADDED,
3774 DIFF_STATUS_COPIED,
3775 DIFF_STATUS_DELETED,
3776 DIFF_STATUS_MODIFIED,
3777 DIFF_STATUS_RENAMED,
3778 DIFF_STATUS_TYPE_CHANGED,
3779 DIFF_STATUS_UNKNOWN,
3780 DIFF_STATUS_UNMERGED,
3781 DIFF_STATUS_FILTER_AON,
3782 DIFF_STATUS_FILTER_BROKEN,
3783 '\0',
3784};
3785
3786static unsigned int filter_bit['Z' + 1];
3787
3788static void prepare_filter_bits(void)
3789{
3790 int i;
3791
3792 if (!filter_bit[DIFF_STATUS_ADDED]) {
3793 for (i = 0; diff_status_letters[i]; i++)
3794 filter_bit[(int) diff_status_letters[i]] = (1 << i);
3795 }
3796}
3797
3798static unsigned filter_bit_tst(char status, const struct diff_options *opt)
3799{
3800 return opt->filter & filter_bit[(int) status];
3801}
3802
3803static int parse_diff_filter_opt(const char *optarg, struct diff_options *opt)
3804{
3805 int i, optch;
3806
3807 prepare_filter_bits();
3808
3809 /*
3810 * If there is a negation e.g. 'd' in the input, and we haven't
3811 * initialized the filter field with another --diff-filter, start
3812 * from full set of bits, except for AON.
3813 */
3814 if (!opt->filter) {
3815 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3816 if (optch < 'a' || 'z' < optch)
3817 continue;
3818 opt->filter = (1 << (ARRAY_SIZE(diff_status_letters) - 1)) - 1;
3819 opt->filter &= ~filter_bit[DIFF_STATUS_FILTER_AON];
3820 break;
3821 }
3822 }
3823
3824 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3825 unsigned int bit;
3826 int negate;
3827
3828 if ('a' <= optch && optch <= 'z') {
3829 negate = 1;
3830 optch = toupper(optch);
3831 } else {
3832 negate = 0;
3833 }
3834
3835 bit = (0 <= optch && optch <= 'Z') ? filter_bit[optch] : 0;
3836 if (!bit)
3837 return optarg[i];
3838 if (negate)
3839 opt->filter &= ~bit;
3840 else
3841 opt->filter |= bit;
3842 }
3843 return 0;
3844}
3845
3846static void enable_patch_output(int *fmt) {
3847 *fmt &= ~DIFF_FORMAT_NO_OUTPUT;
3848 *fmt |= DIFF_FORMAT_PATCH;
3849}
3850
3851static int parse_ws_error_highlight_opt(struct diff_options *opt, const char *arg)
3852{
3853 int val = parse_ws_error_highlight(arg);
3854
3855 if (val < 0) {
3856 error("unknown value after ws-error-highlight=%.*s",
3857 -1 - val, arg);
3858 return 0;
3859 }
3860 opt->ws_error_highlight = val;
3861 return 1;
3862}
3863
3864int diff_opt_parse(struct diff_options *options,
3865 const char **av, int ac, const char *prefix)
3866{
3867 const char *arg = av[0];
3868 const char *optarg;
3869 int argcount;
3870
3871 if (!prefix)
3872 prefix = "";
3873
3874 /* Output format options */
3875 if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch")
3876 || opt_arg(arg, 'U', "unified", &options->context))
3877 enable_patch_output(&options->output_format);
3878 else if (!strcmp(arg, "--raw"))
3879 options->output_format |= DIFF_FORMAT_RAW;
3880 else if (!strcmp(arg, "--patch-with-raw")) {
3881 enable_patch_output(&options->output_format);
3882 options->output_format |= DIFF_FORMAT_RAW;
3883 } else if (!strcmp(arg, "--numstat"))
3884 options->output_format |= DIFF_FORMAT_NUMSTAT;
3885 else if (!strcmp(arg, "--shortstat"))
3886 options->output_format |= DIFF_FORMAT_SHORTSTAT;
3887 else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3888 return parse_dirstat_opt(options, "");
3889 else if (skip_prefix(arg, "-X", &arg))
3890 return parse_dirstat_opt(options, arg);
3891 else if (skip_prefix(arg, "--dirstat=", &arg))
3892 return parse_dirstat_opt(options, arg);
3893 else if (!strcmp(arg, "--cumulative"))
3894 return parse_dirstat_opt(options, "cumulative");
3895 else if (!strcmp(arg, "--dirstat-by-file"))
3896 return parse_dirstat_opt(options, "files");
3897 else if (skip_prefix(arg, "--dirstat-by-file=", &arg)) {
3898 parse_dirstat_opt(options, "files");
3899 return parse_dirstat_opt(options, arg);
3900 }
3901 else if (!strcmp(arg, "--check"))
3902 options->output_format |= DIFF_FORMAT_CHECKDIFF;
3903 else if (!strcmp(arg, "--summary"))
3904 options->output_format |= DIFF_FORMAT_SUMMARY;
3905 else if (!strcmp(arg, "--patch-with-stat")) {
3906 enable_patch_output(&options->output_format);
3907 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3908 } else if (!strcmp(arg, "--name-only"))
3909 options->output_format |= DIFF_FORMAT_NAME;
3910 else if (!strcmp(arg, "--name-status"))
3911 options->output_format |= DIFF_FORMAT_NAME_STATUS;
3912 else if (!strcmp(arg, "-s") || !strcmp(arg, "--no-patch"))
3913 options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3914 else if (starts_with(arg, "--stat"))
3915 /* --stat, --stat-width, --stat-name-width, or --stat-count */
3916 return stat_opt(options, av);
3917
3918 /* renames options */
3919 else if (starts_with(arg, "-B") || starts_with(arg, "--break-rewrites=") ||
3920 !strcmp(arg, "--break-rewrites")) {
3921 if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3922 return error("invalid argument to -B: %s", arg+2);
3923 }
3924 else if (starts_with(arg, "-M") || starts_with(arg, "--find-renames=") ||
3925 !strcmp(arg, "--find-renames")) {
3926 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3927 return error("invalid argument to -M: %s", arg+2);
3928 options->detect_rename = DIFF_DETECT_RENAME;
3929 }
3930 else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3931 options->irreversible_delete = 1;
3932 }
3933 else if (starts_with(arg, "-C") || starts_with(arg, "--find-copies=") ||
3934 !strcmp(arg, "--find-copies")) {
3935 if (options->detect_rename == DIFF_DETECT_COPY)
3936 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3937 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3938 return error("invalid argument to -C: %s", arg+2);
3939 options->detect_rename = DIFF_DETECT_COPY;
3940 }
3941 else if (!strcmp(arg, "--no-renames"))
3942 options->detect_rename = 0;
3943 else if (!strcmp(arg, "--rename-empty"))
3944 DIFF_OPT_SET(options, RENAME_EMPTY);
3945 else if (!strcmp(arg, "--no-rename-empty"))
3946 DIFF_OPT_CLR(options, RENAME_EMPTY);
3947 else if (!strcmp(arg, "--relative"))
3948 DIFF_OPT_SET(options, RELATIVE_NAME);
3949 else if (skip_prefix(arg, "--relative=", &arg)) {
3950 DIFF_OPT_SET(options, RELATIVE_NAME);
3951 options->prefix = arg;
3952 }
3953
3954 /* xdiff options */
3955 else if (!strcmp(arg, "--minimal"))
3956 DIFF_XDL_SET(options, NEED_MINIMAL);
3957 else if (!strcmp(arg, "--no-minimal"))
3958 DIFF_XDL_CLR(options, NEED_MINIMAL);
3959 else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3960 DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3961 else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3962 DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3963 else if (!strcmp(arg, "--ignore-space-at-eol"))
3964 DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3965 else if (!strcmp(arg, "--ignore-blank-lines"))
3966 DIFF_XDL_SET(options, IGNORE_BLANK_LINES);
3967 else if (!strcmp(arg, "--indent-heuristic"))
3968 DIFF_XDL_SET(options, INDENT_HEURISTIC);
3969 else if (!strcmp(arg, "--no-indent-heuristic"))
3970 DIFF_XDL_CLR(options, INDENT_HEURISTIC);
3971 else if (!strcmp(arg, "--patience"))
3972 options->xdl_opts = DIFF_WITH_ALG(options, PATIENCE_DIFF);
3973 else if (!strcmp(arg, "--histogram"))
3974 options->xdl_opts = DIFF_WITH_ALG(options, HISTOGRAM_DIFF);
3975 else if ((argcount = parse_long_opt("diff-algorithm", av, &optarg))) {
3976 long value = parse_algorithm_value(optarg);
3977 if (value < 0)
3978 return error("option diff-algorithm accepts \"myers\", "
3979 "\"minimal\", \"patience\" and \"histogram\"");
3980 /* clear out previous settings */
3981 DIFF_XDL_CLR(options, NEED_MINIMAL);
3982 options->xdl_opts &= ~XDF_DIFF_ALGORITHM_MASK;
3983 options->xdl_opts |= value;
3984 return argcount;
3985 }
3986
3987 /* flags options */
3988 else if (!strcmp(arg, "--binary")) {
3989 enable_patch_output(&options->output_format);
3990 DIFF_OPT_SET(options, BINARY);
3991 }
3992 else if (!strcmp(arg, "--full-index"))
3993 DIFF_OPT_SET(options, FULL_INDEX);
3994 else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
3995 DIFF_OPT_SET(options, TEXT);
3996 else if (!strcmp(arg, "-R"))
3997 DIFF_OPT_SET(options, REVERSE_DIFF);
3998 else if (!strcmp(arg, "--find-copies-harder"))
3999 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
4000 else if (!strcmp(arg, "--follow"))
4001 DIFF_OPT_SET(options, FOLLOW_RENAMES);
4002 else if (!strcmp(arg, "--no-follow")) {
4003 DIFF_OPT_CLR(options, FOLLOW_RENAMES);
4004 DIFF_OPT_CLR(options, DEFAULT_FOLLOW_RENAMES);
4005 } else if (!strcmp(arg, "--color"))
4006 options->use_color = 1;
4007 else if (skip_prefix(arg, "--color=", &arg)) {
4008 int value = git_config_colorbool(NULL, arg);
4009 if (value < 0)
4010 return error("option `color' expects \"always\", \"auto\", or \"never\"");
4011 options->use_color = value;
4012 }
4013 else if (!strcmp(arg, "--no-color"))
4014 options->use_color = 0;
4015 else if (!strcmp(arg, "--color-words")) {
4016 options->use_color = 1;
4017 options->word_diff = DIFF_WORDS_COLOR;
4018 }
4019 else if (skip_prefix(arg, "--color-words=", &arg)) {
4020 options->use_color = 1;
4021 options->word_diff = DIFF_WORDS_COLOR;
4022 options->word_regex = arg;
4023 }
4024 else if (!strcmp(arg, "--word-diff")) {
4025 if (options->word_diff == DIFF_WORDS_NONE)
4026 options->word_diff = DIFF_WORDS_PLAIN;
4027 }
4028 else if (skip_prefix(arg, "--word-diff=", &arg)) {
4029 if (!strcmp(arg, "plain"))
4030 options->word_diff = DIFF_WORDS_PLAIN;
4031 else if (!strcmp(arg, "color")) {
4032 options->use_color = 1;
4033 options->word_diff = DIFF_WORDS_COLOR;
4034 }
4035 else if (!strcmp(arg, "porcelain"))
4036 options->word_diff = DIFF_WORDS_PORCELAIN;
4037 else if (!strcmp(arg, "none"))
4038 options->word_diff = DIFF_WORDS_NONE;
4039 else
4040 die("bad --word-diff argument: %s", arg);
4041 }
4042 else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
4043 if (options->word_diff == DIFF_WORDS_NONE)
4044 options->word_diff = DIFF_WORDS_PLAIN;
4045 options->word_regex = optarg;
4046 return argcount;
4047 }
4048 else if (!strcmp(arg, "--exit-code"))
4049 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
4050 else if (!strcmp(arg, "--quiet"))
4051 DIFF_OPT_SET(options, QUICK);
4052 else if (!strcmp(arg, "--ext-diff"))
4053 DIFF_OPT_SET(options, ALLOW_EXTERNAL);
4054 else if (!strcmp(arg, "--no-ext-diff"))
4055 DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
4056 else if (!strcmp(arg, "--textconv"))
4057 DIFF_OPT_SET(options, ALLOW_TEXTCONV);
4058 else if (!strcmp(arg, "--no-textconv"))
4059 DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
4060 else if (!strcmp(arg, "--ignore-submodules")) {
4061 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
4062 handle_ignore_submodules_arg(options, "all");
4063 } else if (skip_prefix(arg, "--ignore-submodules=", &arg)) {
4064 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
4065 handle_ignore_submodules_arg(options, arg);
4066 } else if (!strcmp(arg, "--submodule"))
4067 options->submodule_format = DIFF_SUBMODULE_LOG;
4068 else if (skip_prefix(arg, "--submodule=", &arg))
4069 return parse_submodule_opt(options, arg);
4070 else if (skip_prefix(arg, "--ws-error-highlight=", &arg))
4071 return parse_ws_error_highlight_opt(options, arg);
4072 else if (!strcmp(arg, "--ita-invisible-in-index"))
4073 options->ita_invisible_in_index = 1;
4074 else if (!strcmp(arg, "--ita-visible-in-index"))
4075 options->ita_invisible_in_index = 0;
4076
4077 /* misc options */
4078 else if (!strcmp(arg, "-z"))
4079 options->line_termination = 0;
4080 else if ((argcount = short_opt('l', av, &optarg))) {
4081 options->rename_limit = strtoul(optarg, NULL, 10);
4082 return argcount;
4083 }
4084 else if ((argcount = short_opt('S', av, &optarg))) {
4085 options->pickaxe = optarg;
4086 options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
4087 return argcount;
4088 } else if ((argcount = short_opt('G', av, &optarg))) {
4089 options->pickaxe = optarg;
4090 options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
4091 return argcount;
4092 }
4093 else if (!strcmp(arg, "--pickaxe-all"))
4094 options->pickaxe_opts |= DIFF_PICKAXE_ALL;
4095 else if (!strcmp(arg, "--pickaxe-regex"))
4096 options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
4097 else if ((argcount = short_opt('O', av, &optarg))) {
4098 options->orderfile = prefix_filename(prefix, optarg);
4099 return argcount;
4100 }
4101 else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
4102 int offending = parse_diff_filter_opt(optarg, options);
4103 if (offending)
4104 die("unknown change class '%c' in --diff-filter=%s",
4105 offending, optarg);
4106 return argcount;
4107 }
4108 else if (!strcmp(arg, "--no-abbrev"))
4109 options->abbrev = 0;
4110 else if (!strcmp(arg, "--abbrev"))
4111 options->abbrev = DEFAULT_ABBREV;
4112 else if (skip_prefix(arg, "--abbrev=", &arg)) {
4113 options->abbrev = strtoul(arg, NULL, 10);
4114 if (options->abbrev < MINIMUM_ABBREV)
4115 options->abbrev = MINIMUM_ABBREV;
4116 else if (40 < options->abbrev)
4117 options->abbrev = 40;
4118 }
4119 else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
4120 options->a_prefix = optarg;
4121 return argcount;
4122 }
4123 else if ((argcount = parse_long_opt("line-prefix", av, &optarg))) {
4124 options->line_prefix = optarg;
4125 options->line_prefix_length = strlen(options->line_prefix);
4126 graph_setup_line_prefix(options);
4127 return argcount;
4128 }
4129 else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
4130 options->b_prefix = optarg;
4131 return argcount;
4132 }
4133 else if (!strcmp(arg, "--no-prefix"))
4134 options->a_prefix = options->b_prefix = "";
4135 else if (opt_arg(arg, '\0', "inter-hunk-context",
4136 &options->interhunkcontext))
4137 ;
4138 else if (!strcmp(arg, "-W"))
4139 DIFF_OPT_SET(options, FUNCCONTEXT);
4140 else if (!strcmp(arg, "--function-context"))
4141 DIFF_OPT_SET(options, FUNCCONTEXT);
4142 else if (!strcmp(arg, "--no-function-context"))
4143 DIFF_OPT_CLR(options, FUNCCONTEXT);
4144 else if ((argcount = parse_long_opt("output", av, &optarg))) {
4145 char *path = prefix_filename(prefix, optarg);
4146 options->file = xfopen(path, "w");
4147 options->close_file = 1;
4148 if (options->use_color != GIT_COLOR_ALWAYS)
4149 options->use_color = GIT_COLOR_NEVER;
4150 free(path);
4151 return argcount;
4152 } else
4153 return 0;
4154 return 1;
4155}
4156
4157int parse_rename_score(const char **cp_p)
4158{
4159 unsigned long num, scale;
4160 int ch, dot;
4161 const char *cp = *cp_p;
4162
4163 num = 0;
4164 scale = 1;
4165 dot = 0;
4166 for (;;) {
4167 ch = *cp;
4168 if ( !dot && ch == '.' ) {
4169 scale = 1;
4170 dot = 1;
4171 } else if ( ch == '%' ) {
4172 scale = dot ? scale*100 : 100;
4173 cp++; /* % is always at the end */
4174 break;
4175 } else if ( ch >= '0' && ch <= '9' ) {
4176 if ( scale < 100000 ) {
4177 scale *= 10;
4178 num = (num*10) + (ch-'0');
4179 }
4180 } else {
4181 break;
4182 }
4183 cp++;
4184 }
4185 *cp_p = cp;
4186
4187 /* user says num divided by scale and we say internally that
4188 * is MAX_SCORE * num / scale.
4189 */
4190 return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
4191}
4192
4193static int diff_scoreopt_parse(const char *opt)
4194{
4195 int opt1, opt2, cmd;
4196
4197 if (*opt++ != '-')
4198 return -1;
4199 cmd = *opt++;
4200 if (cmd == '-') {
4201 /* convert the long-form arguments into short-form versions */
4202 if (skip_prefix(opt, "break-rewrites", &opt)) {
4203 if (*opt == 0 || *opt++ == '=')
4204 cmd = 'B';
4205 } else if (skip_prefix(opt, "find-copies", &opt)) {
4206 if (*opt == 0 || *opt++ == '=')
4207 cmd = 'C';
4208 } else if (skip_prefix(opt, "find-renames", &opt)) {
4209 if (*opt == 0 || *opt++ == '=')
4210 cmd = 'M';
4211 }
4212 }
4213 if (cmd != 'M' && cmd != 'C' && cmd != 'B')
4214 return -1; /* that is not a -M, -C, or -B option */
4215
4216 opt1 = parse_rename_score(&opt);
4217 if (cmd != 'B')
4218 opt2 = 0;
4219 else {
4220 if (*opt == 0)
4221 opt2 = 0;
4222 else if (*opt != '/')
4223 return -1; /* we expect -B80/99 or -B80 */
4224 else {
4225 opt++;
4226 opt2 = parse_rename_score(&opt);
4227 }
4228 }
4229 if (*opt != 0)
4230 return -1;
4231 return opt1 | (opt2 << 16);
4232}
4233
4234struct diff_queue_struct diff_queued_diff;
4235
4236void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
4237{
4238 ALLOC_GROW(queue->queue, queue->nr + 1, queue->alloc);
4239 queue->queue[queue->nr++] = dp;
4240}
4241
4242struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
4243 struct diff_filespec *one,
4244 struct diff_filespec *two)
4245{
4246 struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
4247 dp->one = one;
4248 dp->two = two;
4249 if (queue)
4250 diff_q(queue, dp);
4251 return dp;
4252}
4253
4254void diff_free_filepair(struct diff_filepair *p)
4255{
4256 free_filespec(p->one);
4257 free_filespec(p->two);
4258 free(p);
4259}
4260
4261const char *diff_aligned_abbrev(const struct object_id *oid, int len)
4262{
4263 int abblen;
4264 const char *abbrev;
4265
4266 if (len == GIT_SHA1_HEXSZ)
4267 return oid_to_hex(oid);
4268
4269 abbrev = diff_abbrev_oid(oid, len);
4270 abblen = strlen(abbrev);
4271
4272 /*
4273 * In well-behaved cases, where the abbbreviated result is the
4274 * same as the requested length, append three dots after the
4275 * abbreviation (hence the whole logic is limited to the case
4276 * where abblen < 37); when the actual abbreviated result is a
4277 * bit longer than the requested length, we reduce the number
4278 * of dots so that they match the well-behaved ones. However,
4279 * if the actual abbreviation is longer than the requested
4280 * length by more than three, we give up on aligning, and add
4281 * three dots anyway, to indicate that the output is not the
4282 * full object name. Yes, this may be suboptimal, but this
4283 * appears only in "diff --raw --abbrev" output and it is not
4284 * worth the effort to change it now. Note that this would
4285 * likely to work fine when the automatic sizing of default
4286 * abbreviation length is used--we would be fed -1 in "len" in
4287 * that case, and will end up always appending three-dots, but
4288 * the automatic sizing is supposed to give abblen that ensures
4289 * uniqueness across all objects (statistically speaking).
4290 */
4291 if (abblen < GIT_SHA1_HEXSZ - 3) {
4292 static char hex[GIT_MAX_HEXSZ + 1];
4293 if (len < abblen && abblen <= len + 2)
4294 xsnprintf(hex, sizeof(hex), "%s%.*s", abbrev, len+3-abblen, "..");
4295 else
4296 xsnprintf(hex, sizeof(hex), "%s...", abbrev);
4297 return hex;
4298 }
4299
4300 return oid_to_hex(oid);
4301}
4302
4303static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
4304{
4305 int line_termination = opt->line_termination;
4306 int inter_name_termination = line_termination ? '\t' : '\0';
4307
4308 fprintf(opt->file, "%s", diff_line_prefix(opt));
4309 if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
4310 fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
4311 diff_aligned_abbrev(&p->one->oid, opt->abbrev));
4312 fprintf(opt->file, "%s ",
4313 diff_aligned_abbrev(&p->two->oid, opt->abbrev));
4314 }
4315 if (p->score) {
4316 fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
4317 inter_name_termination);
4318 } else {
4319 fprintf(opt->file, "%c%c", p->status, inter_name_termination);
4320 }
4321
4322 if (p->status == DIFF_STATUS_COPIED ||
4323 p->status == DIFF_STATUS_RENAMED) {
4324 const char *name_a, *name_b;
4325 name_a = p->one->path;
4326 name_b = p->two->path;
4327 strip_prefix(opt->prefix_length, &name_a, &name_b);
4328 write_name_quoted(name_a, opt->file, inter_name_termination);
4329 write_name_quoted(name_b, opt->file, line_termination);
4330 } else {
4331 const char *name_a, *name_b;
4332 name_a = p->one->mode ? p->one->path : p->two->path;
4333 name_b = NULL;
4334 strip_prefix(opt->prefix_length, &name_a, &name_b);
4335 write_name_quoted(name_a, opt->file, line_termination);
4336 }
4337}
4338
4339int diff_unmodified_pair(struct diff_filepair *p)
4340{
4341 /* This function is written stricter than necessary to support
4342 * the currently implemented transformers, but the idea is to
4343 * let transformers to produce diff_filepairs any way they want,
4344 * and filter and clean them up here before producing the output.
4345 */
4346 struct diff_filespec *one = p->one, *two = p->two;
4347
4348 if (DIFF_PAIR_UNMERGED(p))
4349 return 0; /* unmerged is interesting */
4350
4351 /* deletion, addition, mode or type change
4352 * and rename are all interesting.
4353 */
4354 if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
4355 DIFF_PAIR_MODE_CHANGED(p) ||
4356 strcmp(one->path, two->path))
4357 return 0;
4358
4359 /* both are valid and point at the same path. that is, we are
4360 * dealing with a change.
4361 */
4362 if (one->oid_valid && two->oid_valid &&
4363 !oidcmp(&one->oid, &two->oid) &&
4364 !one->dirty_submodule && !two->dirty_submodule)
4365 return 1; /* no change */
4366 if (!one->oid_valid && !two->oid_valid)
4367 return 1; /* both look at the same file on the filesystem. */
4368 return 0;
4369}
4370
4371static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
4372{
4373 if (diff_unmodified_pair(p))
4374 return;
4375
4376 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4377 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4378 return; /* no tree diffs in patch format */
4379
4380 run_diff(p, o);
4381}
4382
4383static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
4384 struct diffstat_t *diffstat)
4385{
4386 if (diff_unmodified_pair(p))
4387 return;
4388
4389 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4390 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4391 return; /* no useful stat for tree diffs */
4392
4393 run_diffstat(p, o, diffstat);
4394}
4395
4396static void diff_flush_checkdiff(struct diff_filepair *p,
4397 struct diff_options *o)
4398{
4399 if (diff_unmodified_pair(p))
4400 return;
4401
4402 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4403 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4404 return; /* nothing to check in tree diffs */
4405
4406 run_checkdiff(p, o);
4407}
4408
4409int diff_queue_is_empty(void)
4410{
4411 struct diff_queue_struct *q = &diff_queued_diff;
4412 int i;
4413 for (i = 0; i < q->nr; i++)
4414 if (!diff_unmodified_pair(q->queue[i]))
4415 return 0;
4416 return 1;
4417}
4418
4419#if DIFF_DEBUG
4420void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
4421{
4422 fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
4423 x, one ? one : "",
4424 s->path,
4425 DIFF_FILE_VALID(s) ? "valid" : "invalid",
4426 s->mode,
4427 s->oid_valid ? oid_to_hex(&s->oid) : "");
4428 fprintf(stderr, "queue[%d] %s size %lu\n",
4429 x, one ? one : "",
4430 s->size);
4431}
4432
4433void diff_debug_filepair(const struct diff_filepair *p, int i)
4434{
4435 diff_debug_filespec(p->one, i, "one");
4436 diff_debug_filespec(p->two, i, "two");
4437 fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
4438 p->score, p->status ? p->status : '?',
4439 p->one->rename_used, p->broken_pair);
4440}
4441
4442void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
4443{
4444 int i;
4445 if (msg)
4446 fprintf(stderr, "%s\n", msg);
4447 fprintf(stderr, "q->nr = %d\n", q->nr);
4448 for (i = 0; i < q->nr; i++) {
4449 struct diff_filepair *p = q->queue[i];
4450 diff_debug_filepair(p, i);
4451 }
4452}
4453#endif
4454
4455static void diff_resolve_rename_copy(void)
4456{
4457 int i;
4458 struct diff_filepair *p;
4459 struct diff_queue_struct *q = &diff_queued_diff;
4460
4461 diff_debug_queue("resolve-rename-copy", q);
4462
4463 for (i = 0; i < q->nr; i++) {
4464 p = q->queue[i];
4465 p->status = 0; /* undecided */
4466 if (DIFF_PAIR_UNMERGED(p))
4467 p->status = DIFF_STATUS_UNMERGED;
4468 else if (!DIFF_FILE_VALID(p->one))
4469 p->status = DIFF_STATUS_ADDED;
4470 else if (!DIFF_FILE_VALID(p->two))
4471 p->status = DIFF_STATUS_DELETED;
4472 else if (DIFF_PAIR_TYPE_CHANGED(p))
4473 p->status = DIFF_STATUS_TYPE_CHANGED;
4474
4475 /* from this point on, we are dealing with a pair
4476 * whose both sides are valid and of the same type, i.e.
4477 * either in-place edit or rename/copy edit.
4478 */
4479 else if (DIFF_PAIR_RENAME(p)) {
4480 /*
4481 * A rename might have re-connected a broken
4482 * pair up, causing the pathnames to be the
4483 * same again. If so, that's not a rename at
4484 * all, just a modification..
4485 *
4486 * Otherwise, see if this source was used for
4487 * multiple renames, in which case we decrement
4488 * the count, and call it a copy.
4489 */
4490 if (!strcmp(p->one->path, p->two->path))
4491 p->status = DIFF_STATUS_MODIFIED;
4492 else if (--p->one->rename_used > 0)
4493 p->status = DIFF_STATUS_COPIED;
4494 else
4495 p->status = DIFF_STATUS_RENAMED;
4496 }
4497 else if (oidcmp(&p->one->oid, &p->two->oid) ||
4498 p->one->mode != p->two->mode ||
4499 p->one->dirty_submodule ||
4500 p->two->dirty_submodule ||
4501 is_null_oid(&p->one->oid))
4502 p->status = DIFF_STATUS_MODIFIED;
4503 else {
4504 /* This is a "no-change" entry and should not
4505 * happen anymore, but prepare for broken callers.
4506 */
4507 error("feeding unmodified %s to diffcore",
4508 p->one->path);
4509 p->status = DIFF_STATUS_UNKNOWN;
4510 }
4511 }
4512 diff_debug_queue("resolve-rename-copy done", q);
4513}
4514
4515static int check_pair_status(struct diff_filepair *p)
4516{
4517 switch (p->status) {
4518 case DIFF_STATUS_UNKNOWN:
4519 return 0;
4520 case 0:
4521 die("internal error in diff-resolve-rename-copy");
4522 default:
4523 return 1;
4524 }
4525}
4526
4527static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
4528{
4529 int fmt = opt->output_format;
4530
4531 if (fmt & DIFF_FORMAT_CHECKDIFF)
4532 diff_flush_checkdiff(p, opt);
4533 else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
4534 diff_flush_raw(p, opt);
4535 else if (fmt & DIFF_FORMAT_NAME) {
4536 const char *name_a, *name_b;
4537 name_a = p->two->path;
4538 name_b = NULL;
4539 strip_prefix(opt->prefix_length, &name_a, &name_b);
4540 fprintf(opt->file, "%s", diff_line_prefix(opt));
4541 write_name_quoted(name_a, opt->file, opt->line_termination);
4542 }
4543}
4544
4545static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
4546{
4547 if (fs->mode)
4548 fprintf(file, " %s mode %06o ", newdelete, fs->mode);
4549 else
4550 fprintf(file, " %s ", newdelete);
4551 write_name_quoted(fs->path, file, '\n');
4552}
4553
4554
4555static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
4556 const char *line_prefix)
4557{
4558 if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
4559 fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
4560 p->two->mode, show_name ? ' ' : '\n');
4561 if (show_name) {
4562 write_name_quoted(p->two->path, file, '\n');
4563 }
4564 }
4565}
4566
4567static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
4568 const char *line_prefix)
4569{
4570 char *names = pprint_rename(p->one->path, p->two->path);
4571
4572 fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
4573 free(names);
4574 show_mode_change(file, p, 0, line_prefix);
4575}
4576
4577static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
4578{
4579 FILE *file = opt->file;
4580 const char *line_prefix = diff_line_prefix(opt);
4581
4582 switch(p->status) {
4583 case DIFF_STATUS_DELETED:
4584 fputs(line_prefix, file);
4585 show_file_mode_name(file, "delete", p->one);
4586 break;
4587 case DIFF_STATUS_ADDED:
4588 fputs(line_prefix, file);
4589 show_file_mode_name(file, "create", p->two);
4590 break;
4591 case DIFF_STATUS_COPIED:
4592 fputs(line_prefix, file);
4593 show_rename_copy(file, "copy", p, line_prefix);
4594 break;
4595 case DIFF_STATUS_RENAMED:
4596 fputs(line_prefix, file);
4597 show_rename_copy(file, "rename", p, line_prefix);
4598 break;
4599 default:
4600 if (p->score) {
4601 fprintf(file, "%s rewrite ", line_prefix);
4602 write_name_quoted(p->two->path, file, ' ');
4603 fprintf(file, "(%d%%)\n", similarity_index(p));
4604 }
4605 show_mode_change(file, p, !p->score, line_prefix);
4606 break;
4607 }
4608}
4609
4610struct patch_id_t {
4611 git_SHA_CTX *ctx;
4612 int patchlen;
4613};
4614
4615static int remove_space(char *line, int len)
4616{
4617 int i;
4618 char *dst = line;
4619 unsigned char c;
4620
4621 for (i = 0; i < len; i++)
4622 if (!isspace((c = line[i])))
4623 *dst++ = c;
4624
4625 return dst - line;
4626}
4627
4628static void patch_id_consume(void *priv, char *line, unsigned long len)
4629{
4630 struct patch_id_t *data = priv;
4631 int new_len;
4632
4633 /* Ignore line numbers when computing the SHA1 of the patch */
4634 if (starts_with(line, "@@ -"))
4635 return;
4636
4637 new_len = remove_space(line, len);
4638
4639 git_SHA1_Update(data->ctx, line, new_len);
4640 data->patchlen += new_len;
4641}
4642
4643static void patch_id_add_string(git_SHA_CTX *ctx, const char *str)
4644{
4645 git_SHA1_Update(ctx, str, strlen(str));
4646}
4647
4648static void patch_id_add_mode(git_SHA_CTX *ctx, unsigned mode)
4649{
4650 /* large enough for 2^32 in octal */
4651 char buf[12];
4652 int len = xsnprintf(buf, sizeof(buf), "%06o", mode);
4653 git_SHA1_Update(ctx, buf, len);
4654}
4655
4656/* returns 0 upon success, and writes result into sha1 */
4657static int diff_get_patch_id(struct diff_options *options, struct object_id *oid, int diff_header_only)
4658{
4659 struct diff_queue_struct *q = &diff_queued_diff;
4660 int i;
4661 git_SHA_CTX ctx;
4662 struct patch_id_t data;
4663
4664 git_SHA1_Init(&ctx);
4665 memset(&data, 0, sizeof(struct patch_id_t));
4666 data.ctx = &ctx;
4667
4668 for (i = 0; i < q->nr; i++) {
4669 xpparam_t xpp;
4670 xdemitconf_t xecfg;
4671 mmfile_t mf1, mf2;
4672 struct diff_filepair *p = q->queue[i];
4673 int len1, len2;
4674
4675 memset(&xpp, 0, sizeof(xpp));
4676 memset(&xecfg, 0, sizeof(xecfg));
4677 if (p->status == 0)
4678 return error("internal diff status error");
4679 if (p->status == DIFF_STATUS_UNKNOWN)
4680 continue;
4681 if (diff_unmodified_pair(p))
4682 continue;
4683 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4684 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4685 continue;
4686 if (DIFF_PAIR_UNMERGED(p))
4687 continue;
4688
4689 diff_fill_oid_info(p->one);
4690 diff_fill_oid_info(p->two);
4691
4692 len1 = remove_space(p->one->path, strlen(p->one->path));
4693 len2 = remove_space(p->two->path, strlen(p->two->path));
4694 patch_id_add_string(&ctx, "diff--git");
4695 patch_id_add_string(&ctx, "a/");
4696 git_SHA1_Update(&ctx, p->one->path, len1);
4697 patch_id_add_string(&ctx, "b/");
4698 git_SHA1_Update(&ctx, p->two->path, len2);
4699
4700 if (p->one->mode == 0) {
4701 patch_id_add_string(&ctx, "newfilemode");
4702 patch_id_add_mode(&ctx, p->two->mode);
4703 patch_id_add_string(&ctx, "---/dev/null");
4704 patch_id_add_string(&ctx, "+++b/");
4705 git_SHA1_Update(&ctx, p->two->path, len2);
4706 } else if (p->two->mode == 0) {
4707 patch_id_add_string(&ctx, "deletedfilemode");
4708 patch_id_add_mode(&ctx, p->one->mode);
4709 patch_id_add_string(&ctx, "---a/");
4710 git_SHA1_Update(&ctx, p->one->path, len1);
4711 patch_id_add_string(&ctx, "+++/dev/null");
4712 } else {
4713 patch_id_add_string(&ctx, "---a/");
4714 git_SHA1_Update(&ctx, p->one->path, len1);
4715 patch_id_add_string(&ctx, "+++b/");
4716 git_SHA1_Update(&ctx, p->two->path, len2);
4717 }
4718
4719 if (diff_header_only)
4720 continue;
4721
4722 if (fill_mmfile(&mf1, p->one) < 0 ||
4723 fill_mmfile(&mf2, p->two) < 0)
4724 return error("unable to read files to diff");
4725
4726 if (diff_filespec_is_binary(p->one) ||
4727 diff_filespec_is_binary(p->two)) {
4728 git_SHA1_Update(&ctx, oid_to_hex(&p->one->oid),
4729 GIT_SHA1_HEXSZ);
4730 git_SHA1_Update(&ctx, oid_to_hex(&p->two->oid),
4731 GIT_SHA1_HEXSZ);
4732 continue;
4733 }
4734
4735 xpp.flags = 0;
4736 xecfg.ctxlen = 3;
4737 xecfg.flags = 0;
4738 if (xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4739 &xpp, &xecfg))
4740 return error("unable to generate patch-id diff for %s",
4741 p->one->path);
4742 }
4743
4744 git_SHA1_Final(oid->hash, &ctx);
4745 return 0;
4746}
4747
4748int diff_flush_patch_id(struct diff_options *options, struct object_id *oid, int diff_header_only)
4749{
4750 struct diff_queue_struct *q = &diff_queued_diff;
4751 int i;
4752 int result = diff_get_patch_id(options, oid, diff_header_only);
4753
4754 for (i = 0; i < q->nr; i++)
4755 diff_free_filepair(q->queue[i]);
4756
4757 free(q->queue);
4758 DIFF_QUEUE_CLEAR(q);
4759
4760 return result;
4761}
4762
4763static int is_summary_empty(const struct diff_queue_struct *q)
4764{
4765 int i;
4766
4767 for (i = 0; i < q->nr; i++) {
4768 const struct diff_filepair *p = q->queue[i];
4769
4770 switch (p->status) {
4771 case DIFF_STATUS_DELETED:
4772 case DIFF_STATUS_ADDED:
4773 case DIFF_STATUS_COPIED:
4774 case DIFF_STATUS_RENAMED:
4775 return 0;
4776 default:
4777 if (p->score)
4778 return 0;
4779 if (p->one->mode && p->two->mode &&
4780 p->one->mode != p->two->mode)
4781 return 0;
4782 break;
4783 }
4784 }
4785 return 1;
4786}
4787
4788static const char rename_limit_warning[] =
4789N_("inexact rename detection was skipped due to too many files.");
4790
4791static const char degrade_cc_to_c_warning[] =
4792N_("only found copies from modified paths due to too many files.");
4793
4794static const char rename_limit_advice[] =
4795N_("you may want to set your %s variable to at least "
4796 "%d and retry the command.");
4797
4798void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4799{
4800 if (degraded_cc)
4801 warning(_(degrade_cc_to_c_warning));
4802 else if (needed)
4803 warning(_(rename_limit_warning));
4804 else
4805 return;
4806 if (0 < needed && needed < 32767)
4807 warning(_(rename_limit_advice), varname, needed);
4808}
4809
4810static void diff_flush_patch_all_file_pairs(struct diff_options *o)
4811{
4812 int i;
4813 struct diff_queue_struct *q = &diff_queued_diff;
4814
4815 if (WSEH_NEW & WS_RULE_MASK)
4816 die("BUG: WS rules bit mask overlaps with diff symbol flags");
4817
4818 for (i = 0; i < q->nr; i++) {
4819 struct diff_filepair *p = q->queue[i];
4820 if (check_pair_status(p))
4821 diff_flush_patch(p, o);
4822 }
4823}
4824
4825void diff_flush(struct diff_options *options)
4826{
4827 struct diff_queue_struct *q = &diff_queued_diff;
4828 int i, output_format = options->output_format;
4829 int separator = 0;
4830 int dirstat_by_line = 0;
4831
4832 /*
4833 * Order: raw, stat, summary, patch
4834 * or: name/name-status/checkdiff (other bits clear)
4835 */
4836 if (!q->nr)
4837 goto free_queue;
4838
4839 if (output_format & (DIFF_FORMAT_RAW |
4840 DIFF_FORMAT_NAME |
4841 DIFF_FORMAT_NAME_STATUS |
4842 DIFF_FORMAT_CHECKDIFF)) {
4843 for (i = 0; i < q->nr; i++) {
4844 struct diff_filepair *p = q->queue[i];
4845 if (check_pair_status(p))
4846 flush_one_pair(p, options);
4847 }
4848 separator++;
4849 }
4850
4851 if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4852 dirstat_by_line = 1;
4853
4854 if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4855 dirstat_by_line) {
4856 struct diffstat_t diffstat;
4857
4858 memset(&diffstat, 0, sizeof(struct diffstat_t));
4859 for (i = 0; i < q->nr; i++) {
4860 struct diff_filepair *p = q->queue[i];
4861 if (check_pair_status(p))
4862 diff_flush_stat(p, options, &diffstat);
4863 }
4864 if (output_format & DIFF_FORMAT_NUMSTAT)
4865 show_numstat(&diffstat, options);
4866 if (output_format & DIFF_FORMAT_DIFFSTAT)
4867 show_stats(&diffstat, options);
4868 if (output_format & DIFF_FORMAT_SHORTSTAT)
4869 show_shortstats(&diffstat, options);
4870 if (output_format & DIFF_FORMAT_DIRSTAT && dirstat_by_line)
4871 show_dirstat_by_line(&diffstat, options);
4872 free_diffstat_info(&diffstat);
4873 separator++;
4874 }
4875 if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4876 show_dirstat(options);
4877
4878 if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4879 for (i = 0; i < q->nr; i++) {
4880 diff_summary(options, q->queue[i]);
4881 }
4882 separator++;
4883 }
4884
4885 if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4886 DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4887 DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4888 /*
4889 * run diff_flush_patch for the exit status. setting
4890 * options->file to /dev/null should be safe, because we
4891 * aren't supposed to produce any output anyway.
4892 */
4893 if (options->close_file)
4894 fclose(options->file);
4895 options->file = xfopen("/dev/null", "w");
4896 options->close_file = 1;
4897 for (i = 0; i < q->nr; i++) {
4898 struct diff_filepair *p = q->queue[i];
4899 if (check_pair_status(p))
4900 diff_flush_patch(p, options);
4901 if (options->found_changes)
4902 break;
4903 }
4904 }
4905
4906 if (output_format & DIFF_FORMAT_PATCH) {
4907 if (separator) {
4908 emit_diff_symbol(options, DIFF_SYMBOL_SEPARATOR, NULL, 0, 0);
4909 if (options->stat_sep) {
4910 /* attach patch instead of inline */
4911 fputs(options->stat_sep, options->file);
4912 }
4913 }
4914
4915 diff_flush_patch_all_file_pairs(options);
4916 }
4917
4918 if (output_format & DIFF_FORMAT_CALLBACK)
4919 options->format_callback(q, options, options->format_callback_data);
4920
4921 for (i = 0; i < q->nr; i++)
4922 diff_free_filepair(q->queue[i]);
4923free_queue:
4924 free(q->queue);
4925 DIFF_QUEUE_CLEAR(q);
4926 if (options->close_file)
4927 fclose(options->file);
4928
4929 /*
4930 * Report the content-level differences with HAS_CHANGES;
4931 * diff_addremove/diff_change does not set the bit when
4932 * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4933 */
4934 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4935 if (options->found_changes)
4936 DIFF_OPT_SET(options, HAS_CHANGES);
4937 else
4938 DIFF_OPT_CLR(options, HAS_CHANGES);
4939 }
4940}
4941
4942static int match_filter(const struct diff_options *options, const struct diff_filepair *p)
4943{
4944 return (((p->status == DIFF_STATUS_MODIFIED) &&
4945 ((p->score &&
4946 filter_bit_tst(DIFF_STATUS_FILTER_BROKEN, options)) ||
4947 (!p->score &&
4948 filter_bit_tst(DIFF_STATUS_MODIFIED, options)))) ||
4949 ((p->status != DIFF_STATUS_MODIFIED) &&
4950 filter_bit_tst(p->status, options)));
4951}
4952
4953static void diffcore_apply_filter(struct diff_options *options)
4954{
4955 int i;
4956 struct diff_queue_struct *q = &diff_queued_diff;
4957 struct diff_queue_struct outq;
4958
4959 DIFF_QUEUE_CLEAR(&outq);
4960
4961 if (!options->filter)
4962 return;
4963
4964 if (filter_bit_tst(DIFF_STATUS_FILTER_AON, options)) {
4965 int found;
4966 for (i = found = 0; !found && i < q->nr; i++) {
4967 if (match_filter(options, q->queue[i]))
4968 found++;
4969 }
4970 if (found)
4971 return;
4972
4973 /* otherwise we will clear the whole queue
4974 * by copying the empty outq at the end of this
4975 * function, but first clear the current entries
4976 * in the queue.
4977 */
4978 for (i = 0; i < q->nr; i++)
4979 diff_free_filepair(q->queue[i]);
4980 }
4981 else {
4982 /* Only the matching ones */
4983 for (i = 0; i < q->nr; i++) {
4984 struct diff_filepair *p = q->queue[i];
4985 if (match_filter(options, p))
4986 diff_q(&outq, p);
4987 else
4988 diff_free_filepair(p);
4989 }
4990 }
4991 free(q->queue);
4992 *q = outq;
4993}
4994
4995/* Check whether two filespecs with the same mode and size are identical */
4996static int diff_filespec_is_identical(struct diff_filespec *one,
4997 struct diff_filespec *two)
4998{
4999 if (S_ISGITLINK(one->mode))
5000 return 0;
5001 if (diff_populate_filespec(one, 0))
5002 return 0;
5003 if (diff_populate_filespec(two, 0))
5004 return 0;
5005 return !memcmp(one->data, two->data, one->size);
5006}
5007
5008static int diff_filespec_check_stat_unmatch(struct diff_filepair *p)
5009{
5010 if (p->done_skip_stat_unmatch)
5011 return p->skip_stat_unmatch_result;
5012
5013 p->done_skip_stat_unmatch = 1;
5014 p->skip_stat_unmatch_result = 0;
5015 /*
5016 * 1. Entries that come from stat info dirtiness
5017 * always have both sides (iow, not create/delete),
5018 * one side of the object name is unknown, with
5019 * the same mode and size. Keep the ones that
5020 * do not match these criteria. They have real
5021 * differences.
5022 *
5023 * 2. At this point, the file is known to be modified,
5024 * with the same mode and size, and the object
5025 * name of one side is unknown. Need to inspect
5026 * the identical contents.
5027 */
5028 if (!DIFF_FILE_VALID(p->one) || /* (1) */
5029 !DIFF_FILE_VALID(p->two) ||
5030 (p->one->oid_valid && p->two->oid_valid) ||
5031 (p->one->mode != p->two->mode) ||
5032 diff_populate_filespec(p->one, CHECK_SIZE_ONLY) ||
5033 diff_populate_filespec(p->two, CHECK_SIZE_ONLY) ||
5034 (p->one->size != p->two->size) ||
5035 !diff_filespec_is_identical(p->one, p->two)) /* (2) */
5036 p->skip_stat_unmatch_result = 1;
5037 return p->skip_stat_unmatch_result;
5038}
5039
5040static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
5041{
5042 int i;
5043 struct diff_queue_struct *q = &diff_queued_diff;
5044 struct diff_queue_struct outq;
5045 DIFF_QUEUE_CLEAR(&outq);
5046
5047 for (i = 0; i < q->nr; i++) {
5048 struct diff_filepair *p = q->queue[i];
5049
5050 if (diff_filespec_check_stat_unmatch(p))
5051 diff_q(&outq, p);
5052 else {
5053 /*
5054 * The caller can subtract 1 from skip_stat_unmatch
5055 * to determine how many paths were dirty only
5056 * due to stat info mismatch.
5057 */
5058 if (!DIFF_OPT_TST(diffopt, NO_INDEX))
5059 diffopt->skip_stat_unmatch++;
5060 diff_free_filepair(p);
5061 }
5062 }
5063 free(q->queue);
5064 *q = outq;
5065}
5066
5067static int diffnamecmp(const void *a_, const void *b_)
5068{
5069 const struct diff_filepair *a = *((const struct diff_filepair **)a_);
5070 const struct diff_filepair *b = *((const struct diff_filepair **)b_);
5071 const char *name_a, *name_b;
5072
5073 name_a = a->one ? a->one->path : a->two->path;
5074 name_b = b->one ? b->one->path : b->two->path;
5075 return strcmp(name_a, name_b);
5076}
5077
5078void diffcore_fix_diff_index(struct diff_options *options)
5079{
5080 struct diff_queue_struct *q = &diff_queued_diff;
5081 QSORT(q->queue, q->nr, diffnamecmp);
5082}
5083
5084void diffcore_std(struct diff_options *options)
5085{
5086 /* NOTE please keep the following in sync with diff_tree_combined() */
5087 if (options->skip_stat_unmatch)
5088 diffcore_skip_stat_unmatch(options);
5089 if (!options->found_follow) {
5090 /* See try_to_follow_renames() in tree-diff.c */
5091 if (options->break_opt != -1)
5092 diffcore_break(options->break_opt);
5093 if (options->detect_rename)
5094 diffcore_rename(options);
5095 if (options->break_opt != -1)
5096 diffcore_merge_broken();
5097 }
5098 if (options->pickaxe)
5099 diffcore_pickaxe(options);
5100 if (options->orderfile)
5101 diffcore_order(options->orderfile);
5102 if (!options->found_follow)
5103 /* See try_to_follow_renames() in tree-diff.c */
5104 diff_resolve_rename_copy();
5105 diffcore_apply_filter(options);
5106
5107 if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5108 DIFF_OPT_SET(options, HAS_CHANGES);
5109 else
5110 DIFF_OPT_CLR(options, HAS_CHANGES);
5111
5112 options->found_follow = 0;
5113}
5114
5115int diff_result_code(struct diff_options *opt, int status)
5116{
5117 int result = 0;
5118
5119 diff_warn_rename_limit("diff.renameLimit",
5120 opt->needed_rename_limit,
5121 opt->degraded_cc_to_c);
5122 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5123 !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
5124 return status;
5125 if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5126 DIFF_OPT_TST(opt, HAS_CHANGES))
5127 result |= 01;
5128 if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
5129 DIFF_OPT_TST(opt, CHECK_FAILED))
5130 result |= 02;
5131 return result;
5132}
5133
5134int diff_can_quit_early(struct diff_options *opt)
5135{
5136 return (DIFF_OPT_TST(opt, QUICK) &&
5137 !opt->filter &&
5138 DIFF_OPT_TST(opt, HAS_CHANGES));
5139}
5140
5141/*
5142 * Shall changes to this submodule be ignored?
5143 *
5144 * Submodule changes can be configured to be ignored separately for each path,
5145 * but that configuration can be overridden from the command line.
5146 */
5147static int is_submodule_ignored(const char *path, struct diff_options *options)
5148{
5149 int ignored = 0;
5150 unsigned orig_flags = options->flags;
5151 if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
5152 set_diffopt_flags_from_submodule_config(options, path);
5153 if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
5154 ignored = 1;
5155 options->flags = orig_flags;
5156 return ignored;
5157}
5158
5159void diff_addremove(struct diff_options *options,
5160 int addremove, unsigned mode,
5161 const struct object_id *oid,
5162 int oid_valid,
5163 const char *concatpath, unsigned dirty_submodule)
5164{
5165 struct diff_filespec *one, *two;
5166
5167 if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
5168 return;
5169
5170 /* This may look odd, but it is a preparation for
5171 * feeding "there are unchanged files which should
5172 * not produce diffs, but when you are doing copy
5173 * detection you would need them, so here they are"
5174 * entries to the diff-core. They will be prefixed
5175 * with something like '=' or '*' (I haven't decided
5176 * which but should not make any difference).
5177 * Feeding the same new and old to diff_change()
5178 * also has the same effect.
5179 * Before the final output happens, they are pruned after
5180 * merged into rename/copy pairs as appropriate.
5181 */
5182 if (DIFF_OPT_TST(options, REVERSE_DIFF))
5183 addremove = (addremove == '+' ? '-' :
5184 addremove == '-' ? '+' : addremove);
5185
5186 if (options->prefix &&
5187 strncmp(concatpath, options->prefix, options->prefix_length))
5188 return;
5189
5190 one = alloc_filespec(concatpath);
5191 two = alloc_filespec(concatpath);
5192
5193 if (addremove != '+')
5194 fill_filespec(one, oid, oid_valid, mode);
5195 if (addremove != '-') {
5196 fill_filespec(two, oid, oid_valid, mode);
5197 two->dirty_submodule = dirty_submodule;
5198 }
5199
5200 diff_queue(&diff_queued_diff, one, two);
5201 if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5202 DIFF_OPT_SET(options, HAS_CHANGES);
5203}
5204
5205void diff_change(struct diff_options *options,
5206 unsigned old_mode, unsigned new_mode,
5207 const struct object_id *old_oid,
5208 const struct object_id *new_oid,
5209 int old_oid_valid, int new_oid_valid,
5210 const char *concatpath,
5211 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
5212{
5213 struct diff_filespec *one, *two;
5214 struct diff_filepair *p;
5215
5216 if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
5217 is_submodule_ignored(concatpath, options))
5218 return;
5219
5220 if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
5221 SWAP(old_mode, new_mode);
5222 SWAP(old_oid, new_oid);
5223 SWAP(old_oid_valid, new_oid_valid);
5224 SWAP(old_dirty_submodule, new_dirty_submodule);
5225 }
5226
5227 if (options->prefix &&
5228 strncmp(concatpath, options->prefix, options->prefix_length))
5229 return;
5230
5231 one = alloc_filespec(concatpath);
5232 two = alloc_filespec(concatpath);
5233 fill_filespec(one, old_oid, old_oid_valid, old_mode);
5234 fill_filespec(two, new_oid, new_oid_valid, new_mode);
5235 one->dirty_submodule = old_dirty_submodule;
5236 two->dirty_submodule = new_dirty_submodule;
5237 p = diff_queue(&diff_queued_diff, one, two);
5238
5239 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5240 return;
5241
5242 if (DIFF_OPT_TST(options, QUICK) && options->skip_stat_unmatch &&
5243 !diff_filespec_check_stat_unmatch(p))
5244 return;
5245
5246 DIFF_OPT_SET(options, HAS_CHANGES);
5247}
5248
5249struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
5250{
5251 struct diff_filepair *pair;
5252 struct diff_filespec *one, *two;
5253
5254 if (options->prefix &&
5255 strncmp(path, options->prefix, options->prefix_length))
5256 return NULL;
5257
5258 one = alloc_filespec(path);
5259 two = alloc_filespec(path);
5260 pair = diff_queue(&diff_queued_diff, one, two);
5261 pair->is_unmerged = 1;
5262 return pair;
5263}
5264
5265static char *run_textconv(const char *pgm, struct diff_filespec *spec,
5266 size_t *outsize)
5267{
5268 struct diff_tempfile *temp;
5269 const char *argv[3];
5270 const char **arg = argv;
5271 struct child_process child = CHILD_PROCESS_INIT;
5272 struct strbuf buf = STRBUF_INIT;
5273 int err = 0;
5274
5275 temp = prepare_temp_file(spec->path, spec);
5276 *arg++ = pgm;
5277 *arg++ = temp->name;
5278 *arg = NULL;
5279
5280 child.use_shell = 1;
5281 child.argv = argv;
5282 child.out = -1;
5283 if (start_command(&child)) {
5284 remove_tempfile();
5285 return NULL;
5286 }
5287
5288 if (strbuf_read(&buf, child.out, 0) < 0)
5289 err = error("error reading from textconv command '%s'", pgm);
5290 close(child.out);
5291
5292 if (finish_command(&child) || err) {
5293 strbuf_release(&buf);
5294 remove_tempfile();
5295 return NULL;
5296 }
5297 remove_tempfile();
5298
5299 return strbuf_detach(&buf, outsize);
5300}
5301
5302size_t fill_textconv(struct userdiff_driver *driver,
5303 struct diff_filespec *df,
5304 char **outbuf)
5305{
5306 size_t size;
5307
5308 if (!driver) {
5309 if (!DIFF_FILE_VALID(df)) {
5310 *outbuf = "";
5311 return 0;
5312 }
5313 if (diff_populate_filespec(df, 0))
5314 die("unable to read files to diff");
5315 *outbuf = df->data;
5316 return df->size;
5317 }
5318
5319 if (!driver->textconv)
5320 die("BUG: fill_textconv called with non-textconv driver");
5321
5322 if (driver->textconv_cache && df->oid_valid) {
5323 *outbuf = notes_cache_get(driver->textconv_cache,
5324 &df->oid,
5325 &size);
5326 if (*outbuf)
5327 return size;
5328 }
5329
5330 *outbuf = run_textconv(driver->textconv, df, &size);
5331 if (!*outbuf)
5332 die("unable to read files to diff");
5333
5334 if (driver->textconv_cache && df->oid_valid) {
5335 /* ignore errors, as we might be in a readonly repository */
5336 notes_cache_put(driver->textconv_cache, &df->oid, *outbuf,
5337 size);
5338 /*
5339 * we could save up changes and flush them all at the end,
5340 * but we would need an extra call after all diffing is done.
5341 * Since generating a cache entry is the slow path anyway,
5342 * this extra overhead probably isn't a big deal.
5343 */
5344 notes_cache_write(driver->textconv_cache);
5345 }
5346
5347 return size;
5348}
5349
5350int textconv_object(const char *path,
5351 unsigned mode,
5352 const struct object_id *oid,
5353 int oid_valid,
5354 char **buf,
5355 unsigned long *buf_size)
5356{
5357 struct diff_filespec *df;
5358 struct userdiff_driver *textconv;
5359
5360 df = alloc_filespec(path);
5361 fill_filespec(df, oid, oid_valid, mode);
5362 textconv = get_textconv(df);
5363 if (!textconv) {
5364 free_filespec(df);
5365 return 0;
5366 }
5367
5368 *buf_size = fill_textconv(textconv, df, buf);
5369 free_filespec(df);
5370 return 1;
5371}
5372
5373void setup_diff_pager(struct diff_options *opt)
5374{
5375 /*
5376 * If the user asked for our exit code, then either they want --quiet
5377 * or --exit-code. We should definitely not bother with a pager in the
5378 * former case, as we will generate no output. Since we still properly
5379 * report our exit code even when a pager is run, we _could_ run a
5380 * pager with --exit-code. But since we have not done so historically,
5381 * and because it is easy to find people oneline advising "git diff
5382 * --exit-code" in hooks and other scripts, we do not do so.
5383 */
5384 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5385 check_pager_config("diff") != 0)
5386 setup_pager();
5387}