5c428e02b6a3fc49475a0ac05cad634708276059
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_REWRITE_DIFF,
565 DIFF_SYMBOL_BINARY_FILES,
566 DIFF_SYMBOL_HEADER,
567 DIFF_SYMBOL_FILEPAIR_PLUS,
568 DIFF_SYMBOL_FILEPAIR_MINUS,
569 DIFF_SYMBOL_WORDS_PORCELAIN,
570 DIFF_SYMBOL_WORDS,
571 DIFF_SYMBOL_CONTEXT,
572 DIFF_SYMBOL_CONTEXT_INCOMPLETE,
573 DIFF_SYMBOL_PLUS,
574 DIFF_SYMBOL_MINUS,
575 DIFF_SYMBOL_NO_LF_EOF,
576 DIFF_SYMBOL_CONTEXT_FRAGINFO,
577 DIFF_SYMBOL_CONTEXT_MARKER,
578 DIFF_SYMBOL_SEPARATOR
579};
580/*
581 * Flags for content lines:
582 * 0..12 are whitespace rules
583 * 13-15 are WSEH_NEW | WSEH_OLD | WSEH_CONTEXT
584 * 16 is marking if the line is blank at EOF
585 */
586#define DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF (1<<16)
587#define DIFF_SYMBOL_CONTENT_WS_MASK (WSEH_NEW | WSEH_OLD | WSEH_CONTEXT | WS_RULE_MASK)
588
589static void emit_line_ws_markup(struct diff_options *o,
590 const char *set, const char *reset,
591 const char *line, int len, char sign,
592 unsigned ws_rule, int blank_at_eof)
593{
594 const char *ws = NULL;
595
596 if (o->ws_error_highlight & ws_rule) {
597 ws = diff_get_color_opt(o, DIFF_WHITESPACE);
598 if (!*ws)
599 ws = NULL;
600 }
601
602 if (!ws)
603 emit_line_0(o, set, reset, sign, line, len);
604 else if (blank_at_eof)
605 /* Blank line at EOF - paint '+' as well */
606 emit_line_0(o, ws, reset, sign, line, len);
607 else {
608 /* Emit just the prefix, then the rest. */
609 emit_line_0(o, set, reset, sign, "", 0);
610 ws_check_emit(line, len, ws_rule,
611 o->file, set, reset, ws);
612 }
613}
614
615static void emit_diff_symbol(struct diff_options *o, enum diff_symbol s,
616 const char *line, int len, unsigned flags)
617{
618 static const char *nneof = " No newline at end of file\n";
619 const char *context, *reset, *set, *meta, *fraginfo;
620 switch (s) {
621 case DIFF_SYMBOL_NO_LF_EOF:
622 context = diff_get_color_opt(o, DIFF_CONTEXT);
623 reset = diff_get_color_opt(o, DIFF_RESET);
624 putc('\n', o->file);
625 emit_line_0(o, context, reset, '\\',
626 nneof, strlen(nneof));
627 break;
628 case DIFF_SYMBOL_CONTEXT_FRAGINFO:
629 emit_line(o, "", "", line, len);
630 break;
631 case DIFF_SYMBOL_CONTEXT_INCOMPLETE:
632 case DIFF_SYMBOL_CONTEXT_MARKER:
633 context = diff_get_color_opt(o, DIFF_CONTEXT);
634 reset = diff_get_color_opt(o, DIFF_RESET);
635 emit_line(o, context, reset, line, len);
636 break;
637 case DIFF_SYMBOL_SEPARATOR:
638 fprintf(o->file, "%s%c",
639 diff_line_prefix(o),
640 o->line_termination);
641 break;
642 case DIFF_SYMBOL_CONTEXT:
643 set = diff_get_color_opt(o, DIFF_CONTEXT);
644 reset = diff_get_color_opt(o, DIFF_RESET);
645 emit_line_ws_markup(o, set, reset, line, len, ' ',
646 flags & (DIFF_SYMBOL_CONTENT_WS_MASK), 0);
647 break;
648 case DIFF_SYMBOL_PLUS:
649 set = diff_get_color_opt(o, DIFF_FILE_NEW);
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,
653 flags & DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF);
654 break;
655 case DIFF_SYMBOL_MINUS:
656 set = diff_get_color_opt(o, DIFF_FILE_OLD);
657 reset = diff_get_color_opt(o, DIFF_RESET);
658 emit_line_ws_markup(o, set, reset, line, len, '-',
659 flags & DIFF_SYMBOL_CONTENT_WS_MASK, 0);
660 break;
661 case DIFF_SYMBOL_WORDS_PORCELAIN:
662 context = diff_get_color_opt(o, DIFF_CONTEXT);
663 reset = diff_get_color_opt(o, DIFF_RESET);
664 emit_line(o, context, reset, line, len);
665 fputs("~\n", o->file);
666 break;
667 case DIFF_SYMBOL_WORDS:
668 context = diff_get_color_opt(o, DIFF_CONTEXT);
669 reset = diff_get_color_opt(o, DIFF_RESET);
670 /*
671 * Skip the prefix character, if any. With
672 * diff_suppress_blank_empty, there may be
673 * none.
674 */
675 if (line[0] != '\n') {
676 line++;
677 len--;
678 }
679 emit_line(o, context, reset, line, len);
680 break;
681 case DIFF_SYMBOL_FILEPAIR_PLUS:
682 meta = diff_get_color_opt(o, DIFF_METAINFO);
683 reset = diff_get_color_opt(o, DIFF_RESET);
684 fprintf(o->file, "%s%s+++ %s%s%s\n", diff_line_prefix(o), meta,
685 line, reset,
686 strchr(line, ' ') ? "\t" : "");
687 break;
688 case DIFF_SYMBOL_FILEPAIR_MINUS:
689 meta = diff_get_color_opt(o, DIFF_METAINFO);
690 reset = diff_get_color_opt(o, DIFF_RESET);
691 fprintf(o->file, "%s%s--- %s%s%s\n", diff_line_prefix(o), meta,
692 line, reset,
693 strchr(line, ' ') ? "\t" : "");
694 break;
695 case DIFF_SYMBOL_BINARY_FILES:
696 case DIFF_SYMBOL_HEADER:
697 fprintf(o->file, "%s", line);
698 break;
699 case DIFF_SYMBOL_REWRITE_DIFF:
700 fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
701 reset = diff_get_color_opt(o, DIFF_RESET);
702 emit_line(o, fraginfo, reset, line, len);
703 break;
704 default:
705 die("BUG: unknown diff symbol");
706 }
707}
708
709static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
710{
711 if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
712 ecbdata->blank_at_eof_in_preimage &&
713 ecbdata->blank_at_eof_in_postimage &&
714 ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
715 ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
716 return 0;
717 return ws_blank_line(line, len, ecbdata->ws_rule);
718}
719
720static void emit_add_line(const char *reset,
721 struct emit_callback *ecbdata,
722 const char *line, int len)
723{
724 unsigned flags = WSEH_NEW | ecbdata->ws_rule;
725 if (new_blank_line_at_eof(ecbdata, line, len))
726 flags |= DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF;
727
728 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_PLUS, line, len, flags);
729}
730
731static void emit_del_line(const char *reset,
732 struct emit_callback *ecbdata,
733 const char *line, int len)
734{
735 unsigned flags = WSEH_OLD | ecbdata->ws_rule;
736 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_MINUS, line, len, flags);
737}
738
739static void emit_context_line(const char *reset,
740 struct emit_callback *ecbdata,
741 const char *line, int len)
742{
743 unsigned flags = WSEH_CONTEXT | ecbdata->ws_rule;
744 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_CONTEXT, line, len, flags);
745}
746
747static void emit_hunk_header(struct emit_callback *ecbdata,
748 const char *line, int len)
749{
750 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
751 const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
752 const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
753 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
754 static const char atat[2] = { '@', '@' };
755 const char *cp, *ep;
756 struct strbuf msgbuf = STRBUF_INIT;
757 int org_len = len;
758 int i = 1;
759
760 /*
761 * As a hunk header must begin with "@@ -<old>, +<new> @@",
762 * it always is at least 10 bytes long.
763 */
764 if (len < 10 ||
765 memcmp(line, atat, 2) ||
766 !(ep = memmem(line + 2, len - 2, atat, 2))) {
767 emit_diff_symbol(ecbdata->opt,
768 DIFF_SYMBOL_CONTEXT_MARKER, line, len, 0);
769 return;
770 }
771 ep += 2; /* skip over @@ */
772
773 /* The hunk header in fraginfo color */
774 strbuf_addstr(&msgbuf, frag);
775 strbuf_add(&msgbuf, line, ep - line);
776 strbuf_addstr(&msgbuf, reset);
777
778 /*
779 * trailing "\r\n"
780 */
781 for ( ; i < 3; i++)
782 if (line[len - i] == '\r' || line[len - i] == '\n')
783 len--;
784
785 /* blank before the func header */
786 for (cp = ep; ep - line < len; ep++)
787 if (*ep != ' ' && *ep != '\t')
788 break;
789 if (ep != cp) {
790 strbuf_addstr(&msgbuf, context);
791 strbuf_add(&msgbuf, cp, ep - cp);
792 strbuf_addstr(&msgbuf, reset);
793 }
794
795 if (ep < line + len) {
796 strbuf_addstr(&msgbuf, func);
797 strbuf_add(&msgbuf, ep, line + len - ep);
798 strbuf_addstr(&msgbuf, reset);
799 }
800
801 strbuf_add(&msgbuf, line + len, org_len - len);
802 strbuf_complete_line(&msgbuf);
803 emit_diff_symbol(ecbdata->opt,
804 DIFF_SYMBOL_CONTEXT_FRAGINFO, msgbuf.buf, msgbuf.len, 0);
805 strbuf_release(&msgbuf);
806}
807
808static struct diff_tempfile *claim_diff_tempfile(void) {
809 int i;
810 for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
811 if (!diff_temp[i].name)
812 return diff_temp + i;
813 die("BUG: diff is failing to clean up its tempfiles");
814}
815
816static void remove_tempfile(void)
817{
818 int i;
819 for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
820 if (is_tempfile_active(&diff_temp[i].tempfile))
821 delete_tempfile(&diff_temp[i].tempfile);
822 diff_temp[i].name = NULL;
823 }
824}
825
826static void add_line_count(struct strbuf *out, int count)
827{
828 switch (count) {
829 case 0:
830 strbuf_addstr(out, "0,0");
831 break;
832 case 1:
833 strbuf_addstr(out, "1");
834 break;
835 default:
836 strbuf_addf(out, "1,%d", count);
837 break;
838 }
839}
840
841static void emit_rewrite_lines(struct emit_callback *ecb,
842 int prefix, const char *data, int size)
843{
844 const char *endp = NULL;
845 const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
846
847 while (0 < size) {
848 int len;
849
850 endp = memchr(data, '\n', size);
851 len = endp ? (endp - data + 1) : size;
852 if (prefix != '+') {
853 ecb->lno_in_preimage++;
854 emit_del_line(reset, ecb, data, len);
855 } else {
856 ecb->lno_in_postimage++;
857 emit_add_line(reset, ecb, data, len);
858 }
859 size -= len;
860 data += len;
861 }
862 if (!endp)
863 emit_diff_symbol(ecb->opt, DIFF_SYMBOL_NO_LF_EOF, NULL, 0, 0);
864}
865
866static void emit_rewrite_diff(const char *name_a,
867 const char *name_b,
868 struct diff_filespec *one,
869 struct diff_filespec *two,
870 struct userdiff_driver *textconv_one,
871 struct userdiff_driver *textconv_two,
872 struct diff_options *o)
873{
874 int lc_a, lc_b;
875 static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
876 const char *a_prefix, *b_prefix;
877 char *data_one, *data_two;
878 size_t size_one, size_two;
879 struct emit_callback ecbdata;
880 struct strbuf out = STRBUF_INIT;
881
882 if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
883 a_prefix = o->b_prefix;
884 b_prefix = o->a_prefix;
885 } else {
886 a_prefix = o->a_prefix;
887 b_prefix = o->b_prefix;
888 }
889
890 name_a += (*name_a == '/');
891 name_b += (*name_b == '/');
892
893 strbuf_reset(&a_name);
894 strbuf_reset(&b_name);
895 quote_two_c_style(&a_name, a_prefix, name_a, 0);
896 quote_two_c_style(&b_name, b_prefix, name_b, 0);
897
898 size_one = fill_textconv(textconv_one, one, &data_one);
899 size_two = fill_textconv(textconv_two, two, &data_two);
900
901 memset(&ecbdata, 0, sizeof(ecbdata));
902 ecbdata.color_diff = want_color(o->use_color);
903 ecbdata.ws_rule = whitespace_rule(name_b);
904 ecbdata.opt = o;
905 if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
906 mmfile_t mf1, mf2;
907 mf1.ptr = (char *)data_one;
908 mf2.ptr = (char *)data_two;
909 mf1.size = size_one;
910 mf2.size = size_two;
911 check_blank_at_eof(&mf1, &mf2, &ecbdata);
912 }
913 ecbdata.lno_in_preimage = 1;
914 ecbdata.lno_in_postimage = 1;
915
916 lc_a = count_lines(data_one, size_one);
917 lc_b = count_lines(data_two, size_two);
918
919 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_MINUS,
920 a_name.buf, a_name.len, 0);
921 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_PLUS,
922 b_name.buf, b_name.len, 0);
923
924 strbuf_addstr(&out, "@@ -");
925 if (!o->irreversible_delete)
926 add_line_count(&out, lc_a);
927 else
928 strbuf_addstr(&out, "?,?");
929 strbuf_addstr(&out, " +");
930 add_line_count(&out, lc_b);
931 strbuf_addstr(&out, " @@\n");
932 emit_diff_symbol(o, DIFF_SYMBOL_REWRITE_DIFF, out.buf, out.len, 0);
933 strbuf_release(&out);
934
935 if (lc_a && !o->irreversible_delete)
936 emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
937 if (lc_b)
938 emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
939 if (textconv_one)
940 free((char *)data_one);
941 if (textconv_two)
942 free((char *)data_two);
943}
944
945struct diff_words_buffer {
946 mmfile_t text;
947 long alloc;
948 struct diff_words_orig {
949 const char *begin, *end;
950 } *orig;
951 int orig_nr, orig_alloc;
952};
953
954static void diff_words_append(char *line, unsigned long len,
955 struct diff_words_buffer *buffer)
956{
957 ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
958 line++;
959 len--;
960 memcpy(buffer->text.ptr + buffer->text.size, line, len);
961 buffer->text.size += len;
962 buffer->text.ptr[buffer->text.size] = '\0';
963}
964
965struct diff_words_style_elem {
966 const char *prefix;
967 const char *suffix;
968 const char *color; /* NULL; filled in by the setup code if
969 * color is enabled */
970};
971
972struct diff_words_style {
973 enum diff_words_type type;
974 struct diff_words_style_elem new, old, ctx;
975 const char *newline;
976};
977
978static struct diff_words_style diff_words_styles[] = {
979 { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
980 { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
981 { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
982};
983
984struct diff_words_data {
985 struct diff_words_buffer minus, plus;
986 const char *current_plus;
987 int last_minus;
988 struct diff_options *opt;
989 regex_t *word_regex;
990 enum diff_words_type type;
991 struct diff_words_style *style;
992};
993
994static int fn_out_diff_words_write_helper(FILE *fp,
995 struct diff_words_style_elem *st_el,
996 const char *newline,
997 size_t count, const char *buf,
998 const char *line_prefix)
999{
1000 int print = 0;
1001
1002 while (count) {
1003 char *p = memchr(buf, '\n', count);
1004 if (print)
1005 fputs(line_prefix, fp);
1006 if (p != buf) {
1007 if (st_el->color && fputs(st_el->color, fp) < 0)
1008 return -1;
1009 if (fputs(st_el->prefix, fp) < 0 ||
1010 fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
1011 fputs(st_el->suffix, fp) < 0)
1012 return -1;
1013 if (st_el->color && *st_el->color
1014 && fputs(GIT_COLOR_RESET, fp) < 0)
1015 return -1;
1016 }
1017 if (!p)
1018 return 0;
1019 if (fputs(newline, fp) < 0)
1020 return -1;
1021 count -= p + 1 - buf;
1022 buf = p + 1;
1023 print = 1;
1024 }
1025 return 0;
1026}
1027
1028/*
1029 * '--color-words' algorithm can be described as:
1030 *
1031 * 1. collect the minus/plus lines of a diff hunk, divided into
1032 * minus-lines and plus-lines;
1033 *
1034 * 2. break both minus-lines and plus-lines into words and
1035 * place them into two mmfile_t with one word for each line;
1036 *
1037 * 3. use xdiff to run diff on the two mmfile_t to get the words level diff;
1038 *
1039 * And for the common parts of the both file, we output the plus side text.
1040 * diff_words->current_plus is used to trace the current position of the plus file
1041 * which printed. diff_words->last_minus is used to trace the last minus word
1042 * printed.
1043 *
1044 * For '--graph' to work with '--color-words', we need to output the graph prefix
1045 * on each line of color words output. Generally, there are two conditions on
1046 * which we should output the prefix.
1047 *
1048 * 1. diff_words->last_minus == 0 &&
1049 * diff_words->current_plus == diff_words->plus.text.ptr
1050 *
1051 * that is: the plus text must start as a new line, and if there is no minus
1052 * word printed, a graph prefix must be printed.
1053 *
1054 * 2. diff_words->current_plus > diff_words->plus.text.ptr &&
1055 * *(diff_words->current_plus - 1) == '\n'
1056 *
1057 * that is: a graph prefix must be printed following a '\n'
1058 */
1059static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
1060{
1061 if ((diff_words->last_minus == 0 &&
1062 diff_words->current_plus == diff_words->plus.text.ptr) ||
1063 (diff_words->current_plus > diff_words->plus.text.ptr &&
1064 *(diff_words->current_plus - 1) == '\n')) {
1065 return 1;
1066 } else {
1067 return 0;
1068 }
1069}
1070
1071static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
1072{
1073 struct diff_words_data *diff_words = priv;
1074 struct diff_words_style *style = diff_words->style;
1075 int minus_first, minus_len, plus_first, plus_len;
1076 const char *minus_begin, *minus_end, *plus_begin, *plus_end;
1077 struct diff_options *opt = diff_words->opt;
1078 const char *line_prefix;
1079
1080 if (line[0] != '@' || parse_hunk_header(line, len,
1081 &minus_first, &minus_len, &plus_first, &plus_len))
1082 return;
1083
1084 assert(opt);
1085 line_prefix = diff_line_prefix(opt);
1086
1087 /* POSIX requires that first be decremented by one if len == 0... */
1088 if (minus_len) {
1089 minus_begin = diff_words->minus.orig[minus_first].begin;
1090 minus_end =
1091 diff_words->minus.orig[minus_first + minus_len - 1].end;
1092 } else
1093 minus_begin = minus_end =
1094 diff_words->minus.orig[minus_first].end;
1095
1096 if (plus_len) {
1097 plus_begin = diff_words->plus.orig[plus_first].begin;
1098 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
1099 } else
1100 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
1101
1102 if (color_words_output_graph_prefix(diff_words)) {
1103 fputs(line_prefix, diff_words->opt->file);
1104 }
1105 if (diff_words->current_plus != plus_begin) {
1106 fn_out_diff_words_write_helper(diff_words->opt->file,
1107 &style->ctx, style->newline,
1108 plus_begin - diff_words->current_plus,
1109 diff_words->current_plus, line_prefix);
1110 if (*(plus_begin - 1) == '\n')
1111 fputs(line_prefix, diff_words->opt->file);
1112 }
1113 if (minus_begin != minus_end) {
1114 fn_out_diff_words_write_helper(diff_words->opt->file,
1115 &style->old, style->newline,
1116 minus_end - minus_begin, minus_begin,
1117 line_prefix);
1118 }
1119 if (plus_begin != plus_end) {
1120 fn_out_diff_words_write_helper(diff_words->opt->file,
1121 &style->new, style->newline,
1122 plus_end - plus_begin, plus_begin,
1123 line_prefix);
1124 }
1125
1126 diff_words->current_plus = plus_end;
1127 diff_words->last_minus = minus_first;
1128}
1129
1130/* This function starts looking at *begin, and returns 0 iff a word was found. */
1131static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
1132 int *begin, int *end)
1133{
1134 if (word_regex && *begin < buffer->size) {
1135 regmatch_t match[1];
1136 if (!regexec_buf(word_regex, buffer->ptr + *begin,
1137 buffer->size - *begin, 1, match, 0)) {
1138 char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
1139 '\n', match[0].rm_eo - match[0].rm_so);
1140 *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
1141 *begin += match[0].rm_so;
1142 return *begin >= *end;
1143 }
1144 return -1;
1145 }
1146
1147 /* find the next word */
1148 while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
1149 (*begin)++;
1150 if (*begin >= buffer->size)
1151 return -1;
1152
1153 /* find the end of the word */
1154 *end = *begin + 1;
1155 while (*end < buffer->size && !isspace(buffer->ptr[*end]))
1156 (*end)++;
1157
1158 return 0;
1159}
1160
1161/*
1162 * This function splits the words in buffer->text, stores the list with
1163 * newline separator into out, and saves the offsets of the original words
1164 * in buffer->orig.
1165 */
1166static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
1167 regex_t *word_regex)
1168{
1169 int i, j;
1170 long alloc = 0;
1171
1172 out->size = 0;
1173 out->ptr = NULL;
1174
1175 /* fake an empty "0th" word */
1176 ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
1177 buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
1178 buffer->orig_nr = 1;
1179
1180 for (i = 0; i < buffer->text.size; i++) {
1181 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
1182 return;
1183
1184 /* store original boundaries */
1185 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
1186 buffer->orig_alloc);
1187 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
1188 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
1189 buffer->orig_nr++;
1190
1191 /* store one word */
1192 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
1193 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
1194 out->ptr[out->size + j - i] = '\n';
1195 out->size += j - i + 1;
1196
1197 i = j - 1;
1198 }
1199}
1200
1201/* this executes the word diff on the accumulated buffers */
1202static void diff_words_show(struct diff_words_data *diff_words)
1203{
1204 xpparam_t xpp;
1205 xdemitconf_t xecfg;
1206 mmfile_t minus, plus;
1207 struct diff_words_style *style = diff_words->style;
1208
1209 struct diff_options *opt = diff_words->opt;
1210 const char *line_prefix;
1211
1212 assert(opt);
1213 line_prefix = diff_line_prefix(opt);
1214
1215 /* special case: only removal */
1216 if (!diff_words->plus.text.size) {
1217 fputs(line_prefix, diff_words->opt->file);
1218 fn_out_diff_words_write_helper(diff_words->opt->file,
1219 &style->old, style->newline,
1220 diff_words->minus.text.size,
1221 diff_words->minus.text.ptr, line_prefix);
1222 diff_words->minus.text.size = 0;
1223 return;
1224 }
1225
1226 diff_words->current_plus = diff_words->plus.text.ptr;
1227 diff_words->last_minus = 0;
1228
1229 memset(&xpp, 0, sizeof(xpp));
1230 memset(&xecfg, 0, sizeof(xecfg));
1231 diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
1232 diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
1233 xpp.flags = 0;
1234 /* as only the hunk header will be parsed, we need a 0-context */
1235 xecfg.ctxlen = 0;
1236 if (xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
1237 &xpp, &xecfg))
1238 die("unable to generate word diff");
1239 free(minus.ptr);
1240 free(plus.ptr);
1241 if (diff_words->current_plus != diff_words->plus.text.ptr +
1242 diff_words->plus.text.size) {
1243 if (color_words_output_graph_prefix(diff_words))
1244 fputs(line_prefix, diff_words->opt->file);
1245 fn_out_diff_words_write_helper(diff_words->opt->file,
1246 &style->ctx, style->newline,
1247 diff_words->plus.text.ptr + diff_words->plus.text.size
1248 - diff_words->current_plus, diff_words->current_plus,
1249 line_prefix);
1250 }
1251 diff_words->minus.text.size = diff_words->plus.text.size = 0;
1252}
1253
1254/* In "color-words" mode, show word-diff of words accumulated in the buffer */
1255static void diff_words_flush(struct emit_callback *ecbdata)
1256{
1257 if (ecbdata->diff_words->minus.text.size ||
1258 ecbdata->diff_words->plus.text.size)
1259 diff_words_show(ecbdata->diff_words);
1260}
1261
1262static void diff_filespec_load_driver(struct diff_filespec *one)
1263{
1264 /* Use already-loaded driver */
1265 if (one->driver)
1266 return;
1267
1268 if (S_ISREG(one->mode))
1269 one->driver = userdiff_find_by_path(one->path);
1270
1271 /* Fallback to default settings */
1272 if (!one->driver)
1273 one->driver = userdiff_find_by_name("default");
1274}
1275
1276static const char *userdiff_word_regex(struct diff_filespec *one)
1277{
1278 diff_filespec_load_driver(one);
1279 return one->driver->word_regex;
1280}
1281
1282static void init_diff_words_data(struct emit_callback *ecbdata,
1283 struct diff_options *orig_opts,
1284 struct diff_filespec *one,
1285 struct diff_filespec *two)
1286{
1287 int i;
1288 struct diff_options *o = xmalloc(sizeof(struct diff_options));
1289 memcpy(o, orig_opts, sizeof(struct diff_options));
1290
1291 ecbdata->diff_words =
1292 xcalloc(1, sizeof(struct diff_words_data));
1293 ecbdata->diff_words->type = o->word_diff;
1294 ecbdata->diff_words->opt = o;
1295 if (!o->word_regex)
1296 o->word_regex = userdiff_word_regex(one);
1297 if (!o->word_regex)
1298 o->word_regex = userdiff_word_regex(two);
1299 if (!o->word_regex)
1300 o->word_regex = diff_word_regex_cfg;
1301 if (o->word_regex) {
1302 ecbdata->diff_words->word_regex = (regex_t *)
1303 xmalloc(sizeof(regex_t));
1304 if (regcomp(ecbdata->diff_words->word_regex,
1305 o->word_regex,
1306 REG_EXTENDED | REG_NEWLINE))
1307 die ("Invalid regular expression: %s",
1308 o->word_regex);
1309 }
1310 for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1311 if (o->word_diff == diff_words_styles[i].type) {
1312 ecbdata->diff_words->style =
1313 &diff_words_styles[i];
1314 break;
1315 }
1316 }
1317 if (want_color(o->use_color)) {
1318 struct diff_words_style *st = ecbdata->diff_words->style;
1319 st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1320 st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1321 st->ctx.color = diff_get_color_opt(o, DIFF_CONTEXT);
1322 }
1323}
1324
1325static void free_diff_words_data(struct emit_callback *ecbdata)
1326{
1327 if (ecbdata->diff_words) {
1328 diff_words_flush(ecbdata);
1329 free (ecbdata->diff_words->opt);
1330 free (ecbdata->diff_words->minus.text.ptr);
1331 free (ecbdata->diff_words->minus.orig);
1332 free (ecbdata->diff_words->plus.text.ptr);
1333 free (ecbdata->diff_words->plus.orig);
1334 if (ecbdata->diff_words->word_regex) {
1335 regfree(ecbdata->diff_words->word_regex);
1336 free(ecbdata->diff_words->word_regex);
1337 }
1338 FREE_AND_NULL(ecbdata->diff_words);
1339 }
1340}
1341
1342const char *diff_get_color(int diff_use_color, enum color_diff ix)
1343{
1344 if (want_color(diff_use_color))
1345 return diff_colors[ix];
1346 return "";
1347}
1348
1349const char *diff_line_prefix(struct diff_options *opt)
1350{
1351 struct strbuf *msgbuf;
1352 if (!opt->output_prefix)
1353 return "";
1354
1355 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
1356 return msgbuf->buf;
1357}
1358
1359static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1360{
1361 const char *cp;
1362 unsigned long allot;
1363 size_t l = len;
1364
1365 if (ecb->truncate)
1366 return ecb->truncate(line, len);
1367 cp = line;
1368 allot = l;
1369 while (0 < l) {
1370 (void) utf8_width(&cp, &l);
1371 if (!cp)
1372 break; /* truncated in the middle? */
1373 }
1374 return allot - l;
1375}
1376
1377static void find_lno(const char *line, struct emit_callback *ecbdata)
1378{
1379 const char *p;
1380 ecbdata->lno_in_preimage = 0;
1381 ecbdata->lno_in_postimage = 0;
1382 p = strchr(line, '-');
1383 if (!p)
1384 return; /* cannot happen */
1385 ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1386 p = strchr(p, '+');
1387 if (!p)
1388 return; /* cannot happen */
1389 ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1390}
1391
1392static void fn_out_consume(void *priv, char *line, unsigned long len)
1393{
1394 struct emit_callback *ecbdata = priv;
1395 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1396 struct diff_options *o = ecbdata->opt;
1397
1398 o->found_changes = 1;
1399
1400 if (ecbdata->header) {
1401 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
1402 ecbdata->header->buf, ecbdata->header->len, 0);
1403 strbuf_reset(ecbdata->header);
1404 ecbdata->header = NULL;
1405 }
1406
1407 if (ecbdata->label_path[0]) {
1408 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_MINUS,
1409 ecbdata->label_path[0],
1410 strlen(ecbdata->label_path[0]), 0);
1411 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_PLUS,
1412 ecbdata->label_path[1],
1413 strlen(ecbdata->label_path[1]), 0);
1414 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1415 }
1416
1417 if (diff_suppress_blank_empty
1418 && len == 2 && line[0] == ' ' && line[1] == '\n') {
1419 line[0] = '\n';
1420 len = 1;
1421 }
1422
1423 if (line[0] == '@') {
1424 if (ecbdata->diff_words)
1425 diff_words_flush(ecbdata);
1426 len = sane_truncate_line(ecbdata, line, len);
1427 find_lno(line, ecbdata);
1428 emit_hunk_header(ecbdata, line, len);
1429 return;
1430 }
1431
1432 if (ecbdata->diff_words) {
1433 enum diff_symbol s =
1434 ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN ?
1435 DIFF_SYMBOL_WORDS_PORCELAIN : DIFF_SYMBOL_WORDS;
1436 if (line[0] == '-') {
1437 diff_words_append(line, len,
1438 &ecbdata->diff_words->minus);
1439 return;
1440 } else if (line[0] == '+') {
1441 diff_words_append(line, len,
1442 &ecbdata->diff_words->plus);
1443 return;
1444 } else if (starts_with(line, "\\ ")) {
1445 /*
1446 * Eat the "no newline at eof" marker as if we
1447 * saw a "+" or "-" line with nothing on it,
1448 * and return without diff_words_flush() to
1449 * defer processing. If this is the end of
1450 * preimage, more "+" lines may come after it.
1451 */
1452 return;
1453 }
1454 diff_words_flush(ecbdata);
1455 emit_diff_symbol(o, s, line, len, 0);
1456 return;
1457 }
1458
1459 switch (line[0]) {
1460 case '+':
1461 ecbdata->lno_in_postimage++;
1462 emit_add_line(reset, ecbdata, line + 1, len - 1);
1463 break;
1464 case '-':
1465 ecbdata->lno_in_preimage++;
1466 emit_del_line(reset, ecbdata, line + 1, len - 1);
1467 break;
1468 case ' ':
1469 ecbdata->lno_in_postimage++;
1470 ecbdata->lno_in_preimage++;
1471 emit_context_line(reset, ecbdata, line + 1, len - 1);
1472 break;
1473 default:
1474 /* incomplete line at the end */
1475 ecbdata->lno_in_preimage++;
1476 emit_diff_symbol(o, DIFF_SYMBOL_CONTEXT_INCOMPLETE,
1477 line, len, 0);
1478 break;
1479 }
1480}
1481
1482static char *pprint_rename(const char *a, const char *b)
1483{
1484 const char *old = a;
1485 const char *new = b;
1486 struct strbuf name = STRBUF_INIT;
1487 int pfx_length, sfx_length;
1488 int pfx_adjust_for_slash;
1489 int len_a = strlen(a);
1490 int len_b = strlen(b);
1491 int a_midlen, b_midlen;
1492 int qlen_a = quote_c_style(a, NULL, NULL, 0);
1493 int qlen_b = quote_c_style(b, NULL, NULL, 0);
1494
1495 if (qlen_a || qlen_b) {
1496 quote_c_style(a, &name, NULL, 0);
1497 strbuf_addstr(&name, " => ");
1498 quote_c_style(b, &name, NULL, 0);
1499 return strbuf_detach(&name, NULL);
1500 }
1501
1502 /* Find common prefix */
1503 pfx_length = 0;
1504 while (*old && *new && *old == *new) {
1505 if (*old == '/')
1506 pfx_length = old - a + 1;
1507 old++;
1508 new++;
1509 }
1510
1511 /* Find common suffix */
1512 old = a + len_a;
1513 new = b + len_b;
1514 sfx_length = 0;
1515 /*
1516 * If there is a common prefix, it must end in a slash. In
1517 * that case we let this loop run 1 into the prefix to see the
1518 * same slash.
1519 *
1520 * If there is no common prefix, we cannot do this as it would
1521 * underrun the input strings.
1522 */
1523 pfx_adjust_for_slash = (pfx_length ? 1 : 0);
1524 while (a + pfx_length - pfx_adjust_for_slash <= old &&
1525 b + pfx_length - pfx_adjust_for_slash <= new &&
1526 *old == *new) {
1527 if (*old == '/')
1528 sfx_length = len_a - (old - a);
1529 old--;
1530 new--;
1531 }
1532
1533 /*
1534 * pfx{mid-a => mid-b}sfx
1535 * {pfx-a => pfx-b}sfx
1536 * pfx{sfx-a => sfx-b}
1537 * name-a => name-b
1538 */
1539 a_midlen = len_a - pfx_length - sfx_length;
1540 b_midlen = len_b - pfx_length - sfx_length;
1541 if (a_midlen < 0)
1542 a_midlen = 0;
1543 if (b_midlen < 0)
1544 b_midlen = 0;
1545
1546 strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1547 if (pfx_length + sfx_length) {
1548 strbuf_add(&name, a, pfx_length);
1549 strbuf_addch(&name, '{');
1550 }
1551 strbuf_add(&name, a + pfx_length, a_midlen);
1552 strbuf_addstr(&name, " => ");
1553 strbuf_add(&name, b + pfx_length, b_midlen);
1554 if (pfx_length + sfx_length) {
1555 strbuf_addch(&name, '}');
1556 strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1557 }
1558 return strbuf_detach(&name, NULL);
1559}
1560
1561struct diffstat_t {
1562 int nr;
1563 int alloc;
1564 struct diffstat_file {
1565 char *from_name;
1566 char *name;
1567 char *print_name;
1568 unsigned is_unmerged:1;
1569 unsigned is_binary:1;
1570 unsigned is_renamed:1;
1571 unsigned is_interesting:1;
1572 uintmax_t added, deleted;
1573 } **files;
1574};
1575
1576static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1577 const char *name_a,
1578 const char *name_b)
1579{
1580 struct diffstat_file *x;
1581 x = xcalloc(1, sizeof(*x));
1582 ALLOC_GROW(diffstat->files, diffstat->nr + 1, diffstat->alloc);
1583 diffstat->files[diffstat->nr++] = x;
1584 if (name_b) {
1585 x->from_name = xstrdup(name_a);
1586 x->name = xstrdup(name_b);
1587 x->is_renamed = 1;
1588 }
1589 else {
1590 x->from_name = NULL;
1591 x->name = xstrdup(name_a);
1592 }
1593 return x;
1594}
1595
1596static void diffstat_consume(void *priv, char *line, unsigned long len)
1597{
1598 struct diffstat_t *diffstat = priv;
1599 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1600
1601 if (line[0] == '+')
1602 x->added++;
1603 else if (line[0] == '-')
1604 x->deleted++;
1605}
1606
1607const char mime_boundary_leader[] = "------------";
1608
1609static int scale_linear(int it, int width, int max_change)
1610{
1611 if (!it)
1612 return 0;
1613 /*
1614 * make sure that at least one '-' or '+' is printed if
1615 * there is any change to this path. The easiest way is to
1616 * scale linearly as if the alloted width is one column shorter
1617 * than it is, and then add 1 to the result.
1618 */
1619 return 1 + (it * (width - 1) / max_change);
1620}
1621
1622static void show_name(FILE *file,
1623 const char *prefix, const char *name, int len)
1624{
1625 fprintf(file, " %s%-*s |", prefix, len, name);
1626}
1627
1628static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1629{
1630 if (cnt <= 0)
1631 return;
1632 fprintf(file, "%s", set);
1633 while (cnt--)
1634 putc(ch, file);
1635 fprintf(file, "%s", reset);
1636}
1637
1638static void fill_print_name(struct diffstat_file *file)
1639{
1640 char *pname;
1641
1642 if (file->print_name)
1643 return;
1644
1645 if (!file->is_renamed) {
1646 struct strbuf buf = STRBUF_INIT;
1647 if (quote_c_style(file->name, &buf, NULL, 0)) {
1648 pname = strbuf_detach(&buf, NULL);
1649 } else {
1650 pname = file->name;
1651 strbuf_release(&buf);
1652 }
1653 } else {
1654 pname = pprint_rename(file->from_name, file->name);
1655 }
1656 file->print_name = pname;
1657}
1658
1659int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1660{
1661 struct strbuf sb = STRBUF_INIT;
1662 int ret;
1663
1664 if (!files) {
1665 assert(insertions == 0 && deletions == 0);
1666 return fprintf(fp, "%s\n", " 0 files changed");
1667 }
1668
1669 strbuf_addf(&sb,
1670 (files == 1) ? " %d file changed" : " %d files changed",
1671 files);
1672
1673 /*
1674 * For binary diff, the caller may want to print "x files
1675 * changed" with insertions == 0 && deletions == 0.
1676 *
1677 * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1678 * is probably less confusing (i.e skip over "2 files changed
1679 * but nothing about added/removed lines? Is this a bug in Git?").
1680 */
1681 if (insertions || deletions == 0) {
1682 strbuf_addf(&sb,
1683 (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
1684 insertions);
1685 }
1686
1687 if (deletions || insertions == 0) {
1688 strbuf_addf(&sb,
1689 (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
1690 deletions);
1691 }
1692 strbuf_addch(&sb, '\n');
1693 ret = fputs(sb.buf, fp);
1694 strbuf_release(&sb);
1695 return ret;
1696}
1697
1698static void show_stats(struct diffstat_t *data, struct diff_options *options)
1699{
1700 int i, len, add, del, adds = 0, dels = 0;
1701 uintmax_t max_change = 0, max_len = 0;
1702 int total_files = data->nr, count;
1703 int width, name_width, graph_width, number_width = 0, bin_width = 0;
1704 const char *reset, *add_c, *del_c;
1705 const char *line_prefix = "";
1706 int extra_shown = 0;
1707
1708 if (data->nr == 0)
1709 return;
1710
1711 line_prefix = diff_line_prefix(options);
1712 count = options->stat_count ? options->stat_count : data->nr;
1713
1714 reset = diff_get_color_opt(options, DIFF_RESET);
1715 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1716 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1717
1718 /*
1719 * Find the longest filename and max number of changes
1720 */
1721 for (i = 0; (i < count) && (i < data->nr); i++) {
1722 struct diffstat_file *file = data->files[i];
1723 uintmax_t change = file->added + file->deleted;
1724
1725 if (!file->is_interesting && (change == 0)) {
1726 count++; /* not shown == room for one more */
1727 continue;
1728 }
1729 fill_print_name(file);
1730 len = strlen(file->print_name);
1731 if (max_len < len)
1732 max_len = len;
1733
1734 if (file->is_unmerged) {
1735 /* "Unmerged" is 8 characters */
1736 bin_width = bin_width < 8 ? 8 : bin_width;
1737 continue;
1738 }
1739 if (file->is_binary) {
1740 /* "Bin XXX -> YYY bytes" */
1741 int w = 14 + decimal_width(file->added)
1742 + decimal_width(file->deleted);
1743 bin_width = bin_width < w ? w : bin_width;
1744 /* Display change counts aligned with "Bin" */
1745 number_width = 3;
1746 continue;
1747 }
1748
1749 if (max_change < change)
1750 max_change = change;
1751 }
1752 count = i; /* where we can stop scanning in data->files[] */
1753
1754 /*
1755 * We have width = stat_width or term_columns() columns total.
1756 * We want a maximum of min(max_len, stat_name_width) for the name part.
1757 * We want a maximum of min(max_change, stat_graph_width) for the +- part.
1758 * We also need 1 for " " and 4 + decimal_width(max_change)
1759 * for " | NNNN " and one the empty column at the end, altogether
1760 * 6 + decimal_width(max_change).
1761 *
1762 * If there's not enough space, we will use the smaller of
1763 * stat_name_width (if set) and 5/8*width for the filename,
1764 * and the rest for constant elements + graph part, but no more
1765 * than stat_graph_width for the graph part.
1766 * (5/8 gives 50 for filename and 30 for the constant parts + graph
1767 * for the standard terminal size).
1768 *
1769 * In other words: stat_width limits the maximum width, and
1770 * stat_name_width fixes the maximum width of the filename,
1771 * and is also used to divide available columns if there
1772 * aren't enough.
1773 *
1774 * Binary files are displayed with "Bin XXX -> YYY bytes"
1775 * instead of the change count and graph. This part is treated
1776 * similarly to the graph part, except that it is not
1777 * "scaled". If total width is too small to accommodate the
1778 * guaranteed minimum width of the filename part and the
1779 * separators and this message, this message will "overflow"
1780 * making the line longer than the maximum width.
1781 */
1782
1783 if (options->stat_width == -1)
1784 width = term_columns() - strlen(line_prefix);
1785 else
1786 width = options->stat_width ? options->stat_width : 80;
1787 number_width = decimal_width(max_change) > number_width ?
1788 decimal_width(max_change) : number_width;
1789
1790 if (options->stat_graph_width == -1)
1791 options->stat_graph_width = diff_stat_graph_width;
1792
1793 /*
1794 * Guarantee 3/8*16==6 for the graph part
1795 * and 5/8*16==10 for the filename part
1796 */
1797 if (width < 16 + 6 + number_width)
1798 width = 16 + 6 + number_width;
1799
1800 /*
1801 * First assign sizes that are wanted, ignoring available width.
1802 * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
1803 * starting from "XXX" should fit in graph_width.
1804 */
1805 graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
1806 if (options->stat_graph_width &&
1807 options->stat_graph_width < graph_width)
1808 graph_width = options->stat_graph_width;
1809
1810 name_width = (options->stat_name_width > 0 &&
1811 options->stat_name_width < max_len) ?
1812 options->stat_name_width : max_len;
1813
1814 /*
1815 * Adjust adjustable widths not to exceed maximum width
1816 */
1817 if (name_width + number_width + 6 + graph_width > width) {
1818 if (graph_width > width * 3/8 - number_width - 6) {
1819 graph_width = width * 3/8 - number_width - 6;
1820 if (graph_width < 6)
1821 graph_width = 6;
1822 }
1823
1824 if (options->stat_graph_width &&
1825 graph_width > options->stat_graph_width)
1826 graph_width = options->stat_graph_width;
1827 if (name_width > width - number_width - 6 - graph_width)
1828 name_width = width - number_width - 6 - graph_width;
1829 else
1830 graph_width = width - number_width - 6 - name_width;
1831 }
1832
1833 /*
1834 * From here name_width is the width of the name area,
1835 * and graph_width is the width of the graph area.
1836 * max_change is used to scale graph properly.
1837 */
1838 for (i = 0; i < count; i++) {
1839 const char *prefix = "";
1840 struct diffstat_file *file = data->files[i];
1841 char *name = file->print_name;
1842 uintmax_t added = file->added;
1843 uintmax_t deleted = file->deleted;
1844 int name_len;
1845
1846 if (!file->is_interesting && (added + deleted == 0))
1847 continue;
1848
1849 /*
1850 * "scale" the filename
1851 */
1852 len = name_width;
1853 name_len = strlen(name);
1854 if (name_width < name_len) {
1855 char *slash;
1856 prefix = "...";
1857 len -= 3;
1858 name += name_len - len;
1859 slash = strchr(name, '/');
1860 if (slash)
1861 name = slash;
1862 }
1863
1864 if (file->is_binary) {
1865 fprintf(options->file, "%s", line_prefix);
1866 show_name(options->file, prefix, name, len);
1867 fprintf(options->file, " %*s", number_width, "Bin");
1868 if (!added && !deleted) {
1869 putc('\n', options->file);
1870 continue;
1871 }
1872 fprintf(options->file, " %s%"PRIuMAX"%s",
1873 del_c, deleted, reset);
1874 fprintf(options->file, " -> ");
1875 fprintf(options->file, "%s%"PRIuMAX"%s",
1876 add_c, added, reset);
1877 fprintf(options->file, " bytes");
1878 fprintf(options->file, "\n");
1879 continue;
1880 }
1881 else if (file->is_unmerged) {
1882 fprintf(options->file, "%s", line_prefix);
1883 show_name(options->file, prefix, name, len);
1884 fprintf(options->file, " Unmerged\n");
1885 continue;
1886 }
1887
1888 /*
1889 * scale the add/delete
1890 */
1891 add = added;
1892 del = deleted;
1893
1894 if (graph_width <= max_change) {
1895 int total = scale_linear(add + del, graph_width, max_change);
1896 if (total < 2 && add && del)
1897 /* width >= 2 due to the sanity check */
1898 total = 2;
1899 if (add < del) {
1900 add = scale_linear(add, graph_width, max_change);
1901 del = total - add;
1902 } else {
1903 del = scale_linear(del, graph_width, max_change);
1904 add = total - del;
1905 }
1906 }
1907 fprintf(options->file, "%s", line_prefix);
1908 show_name(options->file, prefix, name, len);
1909 fprintf(options->file, " %*"PRIuMAX"%s",
1910 number_width, added + deleted,
1911 added + deleted ? " " : "");
1912 show_graph(options->file, '+', add, add_c, reset);
1913 show_graph(options->file, '-', del, del_c, reset);
1914 fprintf(options->file, "\n");
1915 }
1916
1917 for (i = 0; i < data->nr; i++) {
1918 struct diffstat_file *file = data->files[i];
1919 uintmax_t added = file->added;
1920 uintmax_t deleted = file->deleted;
1921
1922 if (file->is_unmerged ||
1923 (!file->is_interesting && (added + deleted == 0))) {
1924 total_files--;
1925 continue;
1926 }
1927
1928 if (!file->is_binary) {
1929 adds += added;
1930 dels += deleted;
1931 }
1932 if (i < count)
1933 continue;
1934 if (!extra_shown)
1935 fprintf(options->file, "%s ...\n", line_prefix);
1936 extra_shown = 1;
1937 }
1938 fprintf(options->file, "%s", line_prefix);
1939 print_stat_summary(options->file, total_files, adds, dels);
1940}
1941
1942static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1943{
1944 int i, adds = 0, dels = 0, total_files = data->nr;
1945
1946 if (data->nr == 0)
1947 return;
1948
1949 for (i = 0; i < data->nr; i++) {
1950 int added = data->files[i]->added;
1951 int deleted= data->files[i]->deleted;
1952
1953 if (data->files[i]->is_unmerged ||
1954 (!data->files[i]->is_interesting && (added + deleted == 0))) {
1955 total_files--;
1956 } else if (!data->files[i]->is_binary) { /* don't count bytes */
1957 adds += added;
1958 dels += deleted;
1959 }
1960 }
1961 fprintf(options->file, "%s", diff_line_prefix(options));
1962 print_stat_summary(options->file, total_files, adds, dels);
1963}
1964
1965static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1966{
1967 int i;
1968
1969 if (data->nr == 0)
1970 return;
1971
1972 for (i = 0; i < data->nr; i++) {
1973 struct diffstat_file *file = data->files[i];
1974
1975 fprintf(options->file, "%s", diff_line_prefix(options));
1976
1977 if (file->is_binary)
1978 fprintf(options->file, "-\t-\t");
1979 else
1980 fprintf(options->file,
1981 "%"PRIuMAX"\t%"PRIuMAX"\t",
1982 file->added, file->deleted);
1983 if (options->line_termination) {
1984 fill_print_name(file);
1985 if (!file->is_renamed)
1986 write_name_quoted(file->name, options->file,
1987 options->line_termination);
1988 else {
1989 fputs(file->print_name, options->file);
1990 putc(options->line_termination, options->file);
1991 }
1992 } else {
1993 if (file->is_renamed) {
1994 putc('\0', options->file);
1995 write_name_quoted(file->from_name, options->file, '\0');
1996 }
1997 write_name_quoted(file->name, options->file, '\0');
1998 }
1999 }
2000}
2001
2002struct dirstat_file {
2003 const char *name;
2004 unsigned long changed;
2005};
2006
2007struct dirstat_dir {
2008 struct dirstat_file *files;
2009 int alloc, nr, permille, cumulative;
2010};
2011
2012static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
2013 unsigned long changed, const char *base, int baselen)
2014{
2015 unsigned long this_dir = 0;
2016 unsigned int sources = 0;
2017 const char *line_prefix = diff_line_prefix(opt);
2018
2019 while (dir->nr) {
2020 struct dirstat_file *f = dir->files;
2021 int namelen = strlen(f->name);
2022 unsigned long this;
2023 char *slash;
2024
2025 if (namelen < baselen)
2026 break;
2027 if (memcmp(f->name, base, baselen))
2028 break;
2029 slash = strchr(f->name + baselen, '/');
2030 if (slash) {
2031 int newbaselen = slash + 1 - f->name;
2032 this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
2033 sources++;
2034 } else {
2035 this = f->changed;
2036 dir->files++;
2037 dir->nr--;
2038 sources += 2;
2039 }
2040 this_dir += this;
2041 }
2042
2043 /*
2044 * We don't report dirstat's for
2045 * - the top level
2046 * - or cases where everything came from a single directory
2047 * under this directory (sources == 1).
2048 */
2049 if (baselen && sources != 1) {
2050 if (this_dir) {
2051 int permille = this_dir * 1000 / changed;
2052 if (permille >= dir->permille) {
2053 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
2054 permille / 10, permille % 10, baselen, base);
2055 if (!dir->cumulative)
2056 return 0;
2057 }
2058 }
2059 }
2060 return this_dir;
2061}
2062
2063static int dirstat_compare(const void *_a, const void *_b)
2064{
2065 const struct dirstat_file *a = _a;
2066 const struct dirstat_file *b = _b;
2067 return strcmp(a->name, b->name);
2068}
2069
2070static void show_dirstat(struct diff_options *options)
2071{
2072 int i;
2073 unsigned long changed;
2074 struct dirstat_dir dir;
2075 struct diff_queue_struct *q = &diff_queued_diff;
2076
2077 dir.files = NULL;
2078 dir.alloc = 0;
2079 dir.nr = 0;
2080 dir.permille = options->dirstat_permille;
2081 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
2082
2083 changed = 0;
2084 for (i = 0; i < q->nr; i++) {
2085 struct diff_filepair *p = q->queue[i];
2086 const char *name;
2087 unsigned long copied, added, damage;
2088 int content_changed;
2089
2090 name = p->two->path ? p->two->path : p->one->path;
2091
2092 if (p->one->oid_valid && p->two->oid_valid)
2093 content_changed = oidcmp(&p->one->oid, &p->two->oid);
2094 else
2095 content_changed = 1;
2096
2097 if (!content_changed) {
2098 /*
2099 * The SHA1 has not changed, so pre-/post-content is
2100 * identical. We can therefore skip looking at the
2101 * file contents altogether.
2102 */
2103 damage = 0;
2104 goto found_damage;
2105 }
2106
2107 if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
2108 /*
2109 * In --dirstat-by-file mode, we don't really need to
2110 * look at the actual file contents at all.
2111 * The fact that the SHA1 changed is enough for us to
2112 * add this file to the list of results
2113 * (with each file contributing equal damage).
2114 */
2115 damage = 1;
2116 goto found_damage;
2117 }
2118
2119 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
2120 diff_populate_filespec(p->one, 0);
2121 diff_populate_filespec(p->two, 0);
2122 diffcore_count_changes(p->one, p->two, NULL, NULL,
2123 &copied, &added);
2124 diff_free_filespec_data(p->one);
2125 diff_free_filespec_data(p->two);
2126 } else if (DIFF_FILE_VALID(p->one)) {
2127 diff_populate_filespec(p->one, CHECK_SIZE_ONLY);
2128 copied = added = 0;
2129 diff_free_filespec_data(p->one);
2130 } else if (DIFF_FILE_VALID(p->two)) {
2131 diff_populate_filespec(p->two, CHECK_SIZE_ONLY);
2132 copied = 0;
2133 added = p->two->size;
2134 diff_free_filespec_data(p->two);
2135 } else
2136 continue;
2137
2138 /*
2139 * Original minus copied is the removed material,
2140 * added is the new material. They are both damages
2141 * made to the preimage.
2142 * If the resulting damage is zero, we know that
2143 * diffcore_count_changes() considers the two entries to
2144 * be identical, but since content_changed is true, we
2145 * know that there must have been _some_ kind of change,
2146 * so we force all entries to have damage > 0.
2147 */
2148 damage = (p->one->size - copied) + added;
2149 if (!damage)
2150 damage = 1;
2151
2152found_damage:
2153 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2154 dir.files[dir.nr].name = name;
2155 dir.files[dir.nr].changed = damage;
2156 changed += damage;
2157 dir.nr++;
2158 }
2159
2160 /* This can happen even with many files, if everything was renames */
2161 if (!changed)
2162 return;
2163
2164 /* Show all directories with more than x% of the changes */
2165 QSORT(dir.files, dir.nr, dirstat_compare);
2166 gather_dirstat(options, &dir, changed, "", 0);
2167}
2168
2169static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
2170{
2171 int i;
2172 unsigned long changed;
2173 struct dirstat_dir dir;
2174
2175 if (data->nr == 0)
2176 return;
2177
2178 dir.files = NULL;
2179 dir.alloc = 0;
2180 dir.nr = 0;
2181 dir.permille = options->dirstat_permille;
2182 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
2183
2184 changed = 0;
2185 for (i = 0; i < data->nr; i++) {
2186 struct diffstat_file *file = data->files[i];
2187 unsigned long damage = file->added + file->deleted;
2188 if (file->is_binary)
2189 /*
2190 * binary files counts bytes, not lines. Must find some
2191 * way to normalize binary bytes vs. textual lines.
2192 * The following heuristic assumes that there are 64
2193 * bytes per "line".
2194 * This is stupid and ugly, but very cheap...
2195 */
2196 damage = (damage + 63) / 64;
2197 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2198 dir.files[dir.nr].name = file->name;
2199 dir.files[dir.nr].changed = damage;
2200 changed += damage;
2201 dir.nr++;
2202 }
2203
2204 /* This can happen even with many files, if everything was renames */
2205 if (!changed)
2206 return;
2207
2208 /* Show all directories with more than x% of the changes */
2209 QSORT(dir.files, dir.nr, dirstat_compare);
2210 gather_dirstat(options, &dir, changed, "", 0);
2211}
2212
2213static void free_diffstat_info(struct diffstat_t *diffstat)
2214{
2215 int i;
2216 for (i = 0; i < diffstat->nr; i++) {
2217 struct diffstat_file *f = diffstat->files[i];
2218 if (f->name != f->print_name)
2219 free(f->print_name);
2220 free(f->name);
2221 free(f->from_name);
2222 free(f);
2223 }
2224 free(diffstat->files);
2225}
2226
2227struct checkdiff_t {
2228 const char *filename;
2229 int lineno;
2230 int conflict_marker_size;
2231 struct diff_options *o;
2232 unsigned ws_rule;
2233 unsigned status;
2234};
2235
2236static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2237{
2238 char firstchar;
2239 int cnt;
2240
2241 if (len < marker_size + 1)
2242 return 0;
2243 firstchar = line[0];
2244 switch (firstchar) {
2245 case '=': case '>': case '<': case '|':
2246 break;
2247 default:
2248 return 0;
2249 }
2250 for (cnt = 1; cnt < marker_size; cnt++)
2251 if (line[cnt] != firstchar)
2252 return 0;
2253 /* line[1] thru line[marker_size-1] are same as firstchar */
2254 if (len < marker_size + 1 || !isspace(line[marker_size]))
2255 return 0;
2256 return 1;
2257}
2258
2259static void checkdiff_consume(void *priv, char *line, unsigned long len)
2260{
2261 struct checkdiff_t *data = priv;
2262 int marker_size = data->conflict_marker_size;
2263 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2264 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2265 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2266 char *err;
2267 const char *line_prefix;
2268
2269 assert(data->o);
2270 line_prefix = diff_line_prefix(data->o);
2271
2272 if (line[0] == '+') {
2273 unsigned bad;
2274 data->lineno++;
2275 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2276 data->status |= 1;
2277 fprintf(data->o->file,
2278 "%s%s:%d: leftover conflict marker\n",
2279 line_prefix, data->filename, data->lineno);
2280 }
2281 bad = ws_check(line + 1, len - 1, data->ws_rule);
2282 if (!bad)
2283 return;
2284 data->status |= bad;
2285 err = whitespace_error_string(bad);
2286 fprintf(data->o->file, "%s%s:%d: %s.\n",
2287 line_prefix, data->filename, data->lineno, err);
2288 free(err);
2289 emit_line(data->o, set, reset, line, 1);
2290 ws_check_emit(line + 1, len - 1, data->ws_rule,
2291 data->o->file, set, reset, ws);
2292 } else if (line[0] == ' ') {
2293 data->lineno++;
2294 } else if (line[0] == '@') {
2295 char *plus = strchr(line, '+');
2296 if (plus)
2297 data->lineno = strtol(plus, NULL, 10) - 1;
2298 else
2299 die("invalid diff");
2300 }
2301}
2302
2303static unsigned char *deflate_it(char *data,
2304 unsigned long size,
2305 unsigned long *result_size)
2306{
2307 int bound;
2308 unsigned char *deflated;
2309 git_zstream stream;
2310
2311 git_deflate_init(&stream, zlib_compression_level);
2312 bound = git_deflate_bound(&stream, size);
2313 deflated = xmalloc(bound);
2314 stream.next_out = deflated;
2315 stream.avail_out = bound;
2316
2317 stream.next_in = (unsigned char *)data;
2318 stream.avail_in = size;
2319 while (git_deflate(&stream, Z_FINISH) == Z_OK)
2320 ; /* nothing */
2321 git_deflate_end(&stream);
2322 *result_size = stream.total_out;
2323 return deflated;
2324}
2325
2326static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two,
2327 const char *prefix)
2328{
2329 void *cp;
2330 void *delta;
2331 void *deflated;
2332 void *data;
2333 unsigned long orig_size;
2334 unsigned long delta_size;
2335 unsigned long deflate_size;
2336 unsigned long data_size;
2337
2338 /* We could do deflated delta, or we could do just deflated two,
2339 * whichever is smaller.
2340 */
2341 delta = NULL;
2342 deflated = deflate_it(two->ptr, two->size, &deflate_size);
2343 if (one->size && two->size) {
2344 delta = diff_delta(one->ptr, one->size,
2345 two->ptr, two->size,
2346 &delta_size, deflate_size);
2347 if (delta) {
2348 void *to_free = delta;
2349 orig_size = delta_size;
2350 delta = deflate_it(delta, delta_size, &delta_size);
2351 free(to_free);
2352 }
2353 }
2354
2355 if (delta && delta_size < deflate_size) {
2356 fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2357 free(deflated);
2358 data = delta;
2359 data_size = delta_size;
2360 }
2361 else {
2362 fprintf(file, "%sliteral %lu\n", prefix, two->size);
2363 free(delta);
2364 data = deflated;
2365 data_size = deflate_size;
2366 }
2367
2368 /* emit data encoded in base85 */
2369 cp = data;
2370 while (data_size) {
2371 int bytes = (52 < data_size) ? 52 : data_size;
2372 char line[70];
2373 data_size -= bytes;
2374 if (bytes <= 26)
2375 line[0] = bytes + 'A' - 1;
2376 else
2377 line[0] = bytes - 26 + 'a' - 1;
2378 encode_85(line + 1, cp, bytes);
2379 cp = (char *) cp + bytes;
2380 fprintf(file, "%s", prefix);
2381 fputs(line, file);
2382 fputc('\n', file);
2383 }
2384 fprintf(file, "%s\n", prefix);
2385 free(data);
2386}
2387
2388static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two,
2389 const char *prefix)
2390{
2391 fprintf(file, "%sGIT binary patch\n", prefix);
2392 emit_binary_diff_body(file, one, two, prefix);
2393 emit_binary_diff_body(file, two, one, prefix);
2394}
2395
2396int diff_filespec_is_binary(struct diff_filespec *one)
2397{
2398 if (one->is_binary == -1) {
2399 diff_filespec_load_driver(one);
2400 if (one->driver->binary != -1)
2401 one->is_binary = one->driver->binary;
2402 else {
2403 if (!one->data && DIFF_FILE_VALID(one))
2404 diff_populate_filespec(one, CHECK_BINARY);
2405 if (one->is_binary == -1 && one->data)
2406 one->is_binary = buffer_is_binary(one->data,
2407 one->size);
2408 if (one->is_binary == -1)
2409 one->is_binary = 0;
2410 }
2411 }
2412 return one->is_binary;
2413}
2414
2415static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2416{
2417 diff_filespec_load_driver(one);
2418 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2419}
2420
2421void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2422{
2423 if (!options->a_prefix)
2424 options->a_prefix = a;
2425 if (!options->b_prefix)
2426 options->b_prefix = b;
2427}
2428
2429struct userdiff_driver *get_textconv(struct diff_filespec *one)
2430{
2431 if (!DIFF_FILE_VALID(one))
2432 return NULL;
2433
2434 diff_filespec_load_driver(one);
2435 return userdiff_get_textconv(one->driver);
2436}
2437
2438static void builtin_diff(const char *name_a,
2439 const char *name_b,
2440 struct diff_filespec *one,
2441 struct diff_filespec *two,
2442 const char *xfrm_msg,
2443 int must_show_header,
2444 struct diff_options *o,
2445 int complete_rewrite)
2446{
2447 mmfile_t mf1, mf2;
2448 const char *lbl[2];
2449 char *a_one, *b_two;
2450 const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
2451 const char *reset = diff_get_color_opt(o, DIFF_RESET);
2452 const char *a_prefix, *b_prefix;
2453 struct userdiff_driver *textconv_one = NULL;
2454 struct userdiff_driver *textconv_two = NULL;
2455 struct strbuf header = STRBUF_INIT;
2456 const char *line_prefix = diff_line_prefix(o);
2457
2458 diff_set_mnemonic_prefix(o, "a/", "b/");
2459 if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2460 a_prefix = o->b_prefix;
2461 b_prefix = o->a_prefix;
2462 } else {
2463 a_prefix = o->a_prefix;
2464 b_prefix = o->b_prefix;
2465 }
2466
2467 if (o->submodule_format == DIFF_SUBMODULE_LOG &&
2468 (!one->mode || S_ISGITLINK(one->mode)) &&
2469 (!two->mode || S_ISGITLINK(two->mode))) {
2470 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2471 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2472 show_submodule_summary(o->file, one->path ? one->path : two->path,
2473 line_prefix,
2474 &one->oid, &two->oid,
2475 two->dirty_submodule,
2476 meta, del, add, reset);
2477 return;
2478 } else if (o->submodule_format == DIFF_SUBMODULE_INLINE_DIFF &&
2479 (!one->mode || S_ISGITLINK(one->mode)) &&
2480 (!two->mode || S_ISGITLINK(two->mode))) {
2481 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2482 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2483 show_submodule_inline_diff(o->file, one->path ? one->path : two->path,
2484 line_prefix,
2485 &one->oid, &two->oid,
2486 two->dirty_submodule,
2487 meta, del, add, reset, o);
2488 return;
2489 }
2490
2491 if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2492 textconv_one = get_textconv(one);
2493 textconv_two = get_textconv(two);
2494 }
2495
2496 /* Never use a non-valid filename anywhere if at all possible */
2497 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2498 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2499
2500 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2501 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2502 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2503 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2504 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
2505 if (lbl[0][0] == '/') {
2506 /* /dev/null */
2507 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
2508 if (xfrm_msg)
2509 strbuf_addstr(&header, xfrm_msg);
2510 must_show_header = 1;
2511 }
2512 else if (lbl[1][0] == '/') {
2513 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
2514 if (xfrm_msg)
2515 strbuf_addstr(&header, xfrm_msg);
2516 must_show_header = 1;
2517 }
2518 else {
2519 if (one->mode != two->mode) {
2520 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
2521 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
2522 must_show_header = 1;
2523 }
2524 if (xfrm_msg)
2525 strbuf_addstr(&header, xfrm_msg);
2526
2527 /*
2528 * we do not run diff between different kind
2529 * of objects.
2530 */
2531 if ((one->mode ^ two->mode) & S_IFMT)
2532 goto free_ab_and_return;
2533 if (complete_rewrite &&
2534 (textconv_one || !diff_filespec_is_binary(one)) &&
2535 (textconv_two || !diff_filespec_is_binary(two))) {
2536 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
2537 header.buf, header.len, 0);
2538 strbuf_reset(&header);
2539 emit_rewrite_diff(name_a, name_b, one, two,
2540 textconv_one, textconv_two, o);
2541 o->found_changes = 1;
2542 goto free_ab_and_return;
2543 }
2544 }
2545
2546 if (o->irreversible_delete && lbl[1][0] == '/') {
2547 emit_diff_symbol(o, DIFF_SYMBOL_HEADER, header.buf,
2548 header.len, 0);
2549 strbuf_reset(&header);
2550 goto free_ab_and_return;
2551 } else if (!DIFF_OPT_TST(o, TEXT) &&
2552 ( (!textconv_one && diff_filespec_is_binary(one)) ||
2553 (!textconv_two && diff_filespec_is_binary(two)) )) {
2554 struct strbuf sb = STRBUF_INIT;
2555 if (!one->data && !two->data &&
2556 S_ISREG(one->mode) && S_ISREG(two->mode) &&
2557 !DIFF_OPT_TST(o, BINARY)) {
2558 if (!oidcmp(&one->oid, &two->oid)) {
2559 if (must_show_header)
2560 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
2561 header.buf, header.len,
2562 0);
2563 goto free_ab_and_return;
2564 }
2565 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
2566 header.buf, header.len, 0);
2567 strbuf_addf(&sb, "%sBinary files %s and %s differ\n",
2568 diff_line_prefix(o), lbl[0], lbl[1]);
2569 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_FILES,
2570 sb.buf, sb.len, 0);
2571 strbuf_release(&sb);
2572 goto free_ab_and_return;
2573 }
2574 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2575 die("unable to read files to diff");
2576 /* Quite common confusing case */
2577 if (mf1.size == mf2.size &&
2578 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2579 if (must_show_header)
2580 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
2581 header.buf, header.len, 0);
2582 goto free_ab_and_return;
2583 }
2584 emit_diff_symbol(o, DIFF_SYMBOL_HEADER, header.buf, header.len, 0);
2585 strbuf_reset(&header);
2586 if (DIFF_OPT_TST(o, BINARY))
2587 emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2588 else {
2589 strbuf_addf(&sb, "%sBinary files %s and %s differ\n",
2590 diff_line_prefix(o), lbl[0], lbl[1]);
2591 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_FILES,
2592 sb.buf, sb.len, 0);
2593 strbuf_release(&sb);
2594 }
2595 o->found_changes = 1;
2596 } else {
2597 /* Crazy xdl interfaces.. */
2598 const char *diffopts = getenv("GIT_DIFF_OPTS");
2599 const char *v;
2600 xpparam_t xpp;
2601 xdemitconf_t xecfg;
2602 struct emit_callback ecbdata;
2603 const struct userdiff_funcname *pe;
2604
2605 if (must_show_header) {
2606 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
2607 header.buf, header.len, 0);
2608 strbuf_reset(&header);
2609 }
2610
2611 mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2612 mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2613
2614 pe = diff_funcname_pattern(one);
2615 if (!pe)
2616 pe = diff_funcname_pattern(two);
2617
2618 memset(&xpp, 0, sizeof(xpp));
2619 memset(&xecfg, 0, sizeof(xecfg));
2620 memset(&ecbdata, 0, sizeof(ecbdata));
2621 ecbdata.label_path = lbl;
2622 ecbdata.color_diff = want_color(o->use_color);
2623 ecbdata.ws_rule = whitespace_rule(name_b);
2624 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2625 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2626 ecbdata.opt = o;
2627 ecbdata.header = header.len ? &header : NULL;
2628 xpp.flags = o->xdl_opts;
2629 xecfg.ctxlen = o->context;
2630 xecfg.interhunkctxlen = o->interhunkcontext;
2631 xecfg.flags = XDL_EMIT_FUNCNAMES;
2632 if (DIFF_OPT_TST(o, FUNCCONTEXT))
2633 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2634 if (pe)
2635 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2636 if (!diffopts)
2637 ;
2638 else if (skip_prefix(diffopts, "--unified=", &v))
2639 xecfg.ctxlen = strtoul(v, NULL, 10);
2640 else if (skip_prefix(diffopts, "-u", &v))
2641 xecfg.ctxlen = strtoul(v, NULL, 10);
2642 if (o->word_diff)
2643 init_diff_words_data(&ecbdata, o, one, two);
2644 if (xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2645 &xpp, &xecfg))
2646 die("unable to generate diff for %s", one->path);
2647 if (o->word_diff)
2648 free_diff_words_data(&ecbdata);
2649 if (textconv_one)
2650 free(mf1.ptr);
2651 if (textconv_two)
2652 free(mf2.ptr);
2653 xdiff_clear_find_func(&xecfg);
2654 }
2655
2656 free_ab_and_return:
2657 strbuf_release(&header);
2658 diff_free_filespec_data(one);
2659 diff_free_filespec_data(two);
2660 free(a_one);
2661 free(b_two);
2662 return;
2663}
2664
2665static void builtin_diffstat(const char *name_a, const char *name_b,
2666 struct diff_filespec *one,
2667 struct diff_filespec *two,
2668 struct diffstat_t *diffstat,
2669 struct diff_options *o,
2670 struct diff_filepair *p)
2671{
2672 mmfile_t mf1, mf2;
2673 struct diffstat_file *data;
2674 int same_contents;
2675 int complete_rewrite = 0;
2676
2677 if (!DIFF_PAIR_UNMERGED(p)) {
2678 if (p->status == DIFF_STATUS_MODIFIED && p->score)
2679 complete_rewrite = 1;
2680 }
2681
2682 data = diffstat_add(diffstat, name_a, name_b);
2683 data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
2684
2685 if (!one || !two) {
2686 data->is_unmerged = 1;
2687 return;
2688 }
2689
2690 same_contents = !oidcmp(&one->oid, &two->oid);
2691
2692 if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2693 data->is_binary = 1;
2694 if (same_contents) {
2695 data->added = 0;
2696 data->deleted = 0;
2697 } else {
2698 data->added = diff_filespec_size(two);
2699 data->deleted = diff_filespec_size(one);
2700 }
2701 }
2702
2703 else if (complete_rewrite) {
2704 diff_populate_filespec(one, 0);
2705 diff_populate_filespec(two, 0);
2706 data->deleted = count_lines(one->data, one->size);
2707 data->added = count_lines(two->data, two->size);
2708 }
2709
2710 else if (!same_contents) {
2711 /* Crazy xdl interfaces.. */
2712 xpparam_t xpp;
2713 xdemitconf_t xecfg;
2714
2715 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2716 die("unable to read files to diff");
2717
2718 memset(&xpp, 0, sizeof(xpp));
2719 memset(&xecfg, 0, sizeof(xecfg));
2720 xpp.flags = o->xdl_opts;
2721 xecfg.ctxlen = o->context;
2722 xecfg.interhunkctxlen = o->interhunkcontext;
2723 if (xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2724 &xpp, &xecfg))
2725 die("unable to generate diffstat for %s", one->path);
2726 }
2727
2728 diff_free_filespec_data(one);
2729 diff_free_filespec_data(two);
2730}
2731
2732static void builtin_checkdiff(const char *name_a, const char *name_b,
2733 const char *attr_path,
2734 struct diff_filespec *one,
2735 struct diff_filespec *two,
2736 struct diff_options *o)
2737{
2738 mmfile_t mf1, mf2;
2739 struct checkdiff_t data;
2740
2741 if (!two)
2742 return;
2743
2744 memset(&data, 0, sizeof(data));
2745 data.filename = name_b ? name_b : name_a;
2746 data.lineno = 0;
2747 data.o = o;
2748 data.ws_rule = whitespace_rule(attr_path);
2749 data.conflict_marker_size = ll_merge_marker_size(attr_path);
2750
2751 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2752 die("unable to read files to diff");
2753
2754 /*
2755 * All the other codepaths check both sides, but not checking
2756 * the "old" side here is deliberate. We are checking the newly
2757 * introduced changes, and as long as the "new" side is text, we
2758 * can and should check what it introduces.
2759 */
2760 if (diff_filespec_is_binary(two))
2761 goto free_and_return;
2762 else {
2763 /* Crazy xdl interfaces.. */
2764 xpparam_t xpp;
2765 xdemitconf_t xecfg;
2766
2767 memset(&xpp, 0, sizeof(xpp));
2768 memset(&xecfg, 0, sizeof(xecfg));
2769 xecfg.ctxlen = 1; /* at least one context line */
2770 xpp.flags = 0;
2771 if (xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2772 &xpp, &xecfg))
2773 die("unable to generate checkdiff for %s", one->path);
2774
2775 if (data.ws_rule & WS_BLANK_AT_EOF) {
2776 struct emit_callback ecbdata;
2777 int blank_at_eof;
2778
2779 ecbdata.ws_rule = data.ws_rule;
2780 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2781 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2782
2783 if (blank_at_eof) {
2784 static char *err;
2785 if (!err)
2786 err = whitespace_error_string(WS_BLANK_AT_EOF);
2787 fprintf(o->file, "%s:%d: %s.\n",
2788 data.filename, blank_at_eof, err);
2789 data.status = 1; /* report errors */
2790 }
2791 }
2792 }
2793 free_and_return:
2794 diff_free_filespec_data(one);
2795 diff_free_filespec_data(two);
2796 if (data.status)
2797 DIFF_OPT_SET(o, CHECK_FAILED);
2798}
2799
2800struct diff_filespec *alloc_filespec(const char *path)
2801{
2802 struct diff_filespec *spec;
2803
2804 FLEXPTR_ALLOC_STR(spec, path, path);
2805 spec->count = 1;
2806 spec->is_binary = -1;
2807 return spec;
2808}
2809
2810void free_filespec(struct diff_filespec *spec)
2811{
2812 if (!--spec->count) {
2813 diff_free_filespec_data(spec);
2814 free(spec);
2815 }
2816}
2817
2818void fill_filespec(struct diff_filespec *spec, const struct object_id *oid,
2819 int oid_valid, unsigned short mode)
2820{
2821 if (mode) {
2822 spec->mode = canon_mode(mode);
2823 oidcpy(&spec->oid, oid);
2824 spec->oid_valid = oid_valid;
2825 }
2826}
2827
2828/*
2829 * Given a name and sha1 pair, if the index tells us the file in
2830 * the work tree has that object contents, return true, so that
2831 * prepare_temp_file() does not have to inflate and extract.
2832 */
2833static int reuse_worktree_file(const char *name, const struct object_id *oid, int want_file)
2834{
2835 const struct cache_entry *ce;
2836 struct stat st;
2837 int pos, len;
2838
2839 /*
2840 * We do not read the cache ourselves here, because the
2841 * benchmark with my previous version that always reads cache
2842 * shows that it makes things worse for diff-tree comparing
2843 * two linux-2.6 kernel trees in an already checked out work
2844 * tree. This is because most diff-tree comparisons deal with
2845 * only a small number of files, while reading the cache is
2846 * expensive for a large project, and its cost outweighs the
2847 * savings we get by not inflating the object to a temporary
2848 * file. Practically, this code only helps when we are used
2849 * by diff-cache --cached, which does read the cache before
2850 * calling us.
2851 */
2852 if (!active_cache)
2853 return 0;
2854
2855 /* We want to avoid the working directory if our caller
2856 * doesn't need the data in a normal file, this system
2857 * is rather slow with its stat/open/mmap/close syscalls,
2858 * and the object is contained in a pack file. The pack
2859 * is probably already open and will be faster to obtain
2860 * the data through than the working directory. Loose
2861 * objects however would tend to be slower as they need
2862 * to be individually opened and inflated.
2863 */
2864 if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(oid->hash))
2865 return 0;
2866
2867 /*
2868 * Similarly, if we'd have to convert the file contents anyway, that
2869 * makes the optimization not worthwhile.
2870 */
2871 if (!want_file && would_convert_to_git(&the_index, name))
2872 return 0;
2873
2874 len = strlen(name);
2875 pos = cache_name_pos(name, len);
2876 if (pos < 0)
2877 return 0;
2878 ce = active_cache[pos];
2879
2880 /*
2881 * This is not the sha1 we are looking for, or
2882 * unreusable because it is not a regular file.
2883 */
2884 if (oidcmp(oid, &ce->oid) || !S_ISREG(ce->ce_mode))
2885 return 0;
2886
2887 /*
2888 * If ce is marked as "assume unchanged", there is no
2889 * guarantee that work tree matches what we are looking for.
2890 */
2891 if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2892 return 0;
2893
2894 /*
2895 * If ce matches the file in the work tree, we can reuse it.
2896 */
2897 if (ce_uptodate(ce) ||
2898 (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2899 return 1;
2900
2901 return 0;
2902}
2903
2904static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2905{
2906 struct strbuf buf = STRBUF_INIT;
2907 char *dirty = "";
2908
2909 /* Are we looking at the work tree? */
2910 if (s->dirty_submodule)
2911 dirty = "-dirty";
2912
2913 strbuf_addf(&buf, "Subproject commit %s%s\n",
2914 oid_to_hex(&s->oid), dirty);
2915 s->size = buf.len;
2916 if (size_only) {
2917 s->data = NULL;
2918 strbuf_release(&buf);
2919 } else {
2920 s->data = strbuf_detach(&buf, NULL);
2921 s->should_free = 1;
2922 }
2923 return 0;
2924}
2925
2926/*
2927 * While doing rename detection and pickaxe operation, we may need to
2928 * grab the data for the blob (or file) for our own in-core comparison.
2929 * diff_filespec has data and size fields for this purpose.
2930 */
2931int diff_populate_filespec(struct diff_filespec *s, unsigned int flags)
2932{
2933 int size_only = flags & CHECK_SIZE_ONLY;
2934 int err = 0;
2935 /*
2936 * demote FAIL to WARN to allow inspecting the situation
2937 * instead of refusing.
2938 */
2939 enum safe_crlf crlf_warn = (safe_crlf == SAFE_CRLF_FAIL
2940 ? SAFE_CRLF_WARN
2941 : safe_crlf);
2942
2943 if (!DIFF_FILE_VALID(s))
2944 die("internal error: asking to populate invalid file.");
2945 if (S_ISDIR(s->mode))
2946 return -1;
2947
2948 if (s->data)
2949 return 0;
2950
2951 if (size_only && 0 < s->size)
2952 return 0;
2953
2954 if (S_ISGITLINK(s->mode))
2955 return diff_populate_gitlink(s, size_only);
2956
2957 if (!s->oid_valid ||
2958 reuse_worktree_file(s->path, &s->oid, 0)) {
2959 struct strbuf buf = STRBUF_INIT;
2960 struct stat st;
2961 int fd;
2962
2963 if (lstat(s->path, &st) < 0) {
2964 if (errno == ENOENT) {
2965 err_empty:
2966 err = -1;
2967 empty:
2968 s->data = (char *)"";
2969 s->size = 0;
2970 return err;
2971 }
2972 }
2973 s->size = xsize_t(st.st_size);
2974 if (!s->size)
2975 goto empty;
2976 if (S_ISLNK(st.st_mode)) {
2977 struct strbuf sb = STRBUF_INIT;
2978
2979 if (strbuf_readlink(&sb, s->path, s->size))
2980 goto err_empty;
2981 s->size = sb.len;
2982 s->data = strbuf_detach(&sb, NULL);
2983 s->should_free = 1;
2984 return 0;
2985 }
2986
2987 /*
2988 * Even if the caller would be happy with getting
2989 * only the size, we cannot return early at this
2990 * point if the path requires us to run the content
2991 * conversion.
2992 */
2993 if (size_only && !would_convert_to_git(&the_index, s->path))
2994 return 0;
2995
2996 /*
2997 * Note: this check uses xsize_t(st.st_size) that may
2998 * not be the true size of the blob after it goes
2999 * through convert_to_git(). This may not strictly be
3000 * correct, but the whole point of big_file_threshold
3001 * and is_binary check being that we want to avoid
3002 * opening the file and inspecting the contents, this
3003 * is probably fine.
3004 */
3005 if ((flags & CHECK_BINARY) &&
3006 s->size > big_file_threshold && s->is_binary == -1) {
3007 s->is_binary = 1;
3008 return 0;
3009 }
3010 fd = open(s->path, O_RDONLY);
3011 if (fd < 0)
3012 goto err_empty;
3013 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
3014 close(fd);
3015 s->should_munmap = 1;
3016
3017 /*
3018 * Convert from working tree format to canonical git format
3019 */
3020 if (convert_to_git(&the_index, s->path, s->data, s->size, &buf, crlf_warn)) {
3021 size_t size = 0;
3022 munmap(s->data, s->size);
3023 s->should_munmap = 0;
3024 s->data = strbuf_detach(&buf, &size);
3025 s->size = size;
3026 s->should_free = 1;
3027 }
3028 }
3029 else {
3030 enum object_type type;
3031 if (size_only || (flags & CHECK_BINARY)) {
3032 type = sha1_object_info(s->oid.hash, &s->size);
3033 if (type < 0)
3034 die("unable to read %s",
3035 oid_to_hex(&s->oid));
3036 if (size_only)
3037 return 0;
3038 if (s->size > big_file_threshold && s->is_binary == -1) {
3039 s->is_binary = 1;
3040 return 0;
3041 }
3042 }
3043 s->data = read_sha1_file(s->oid.hash, &type, &s->size);
3044 if (!s->data)
3045 die("unable to read %s", oid_to_hex(&s->oid));
3046 s->should_free = 1;
3047 }
3048 return 0;
3049}
3050
3051void diff_free_filespec_blob(struct diff_filespec *s)
3052{
3053 if (s->should_free)
3054 free(s->data);
3055 else if (s->should_munmap)
3056 munmap(s->data, s->size);
3057
3058 if (s->should_free || s->should_munmap) {
3059 s->should_free = s->should_munmap = 0;
3060 s->data = NULL;
3061 }
3062}
3063
3064void diff_free_filespec_data(struct diff_filespec *s)
3065{
3066 diff_free_filespec_blob(s);
3067 FREE_AND_NULL(s->cnt_data);
3068}
3069
3070static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
3071 void *blob,
3072 unsigned long size,
3073 const struct object_id *oid,
3074 int mode)
3075{
3076 int fd;
3077 struct strbuf buf = STRBUF_INIT;
3078 struct strbuf template = STRBUF_INIT;
3079 char *path_dup = xstrdup(path);
3080 const char *base = basename(path_dup);
3081
3082 /* Generate "XXXXXX_basename.ext" */
3083 strbuf_addstr(&template, "XXXXXX_");
3084 strbuf_addstr(&template, base);
3085
3086 fd = mks_tempfile_ts(&temp->tempfile, template.buf, strlen(base) + 1);
3087 if (fd < 0)
3088 die_errno("unable to create temp-file");
3089 if (convert_to_working_tree(path,
3090 (const char *)blob, (size_t)size, &buf)) {
3091 blob = buf.buf;
3092 size = buf.len;
3093 }
3094 if (write_in_full(fd, blob, size) != size)
3095 die_errno("unable to write temp-file");
3096 close_tempfile(&temp->tempfile);
3097 temp->name = get_tempfile_path(&temp->tempfile);
3098 oid_to_hex_r(temp->hex, oid);
3099 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", mode);
3100 strbuf_release(&buf);
3101 strbuf_release(&template);
3102 free(path_dup);
3103}
3104
3105static struct diff_tempfile *prepare_temp_file(const char *name,
3106 struct diff_filespec *one)
3107{
3108 struct diff_tempfile *temp = claim_diff_tempfile();
3109
3110 if (!DIFF_FILE_VALID(one)) {
3111 not_a_valid_file:
3112 /* A '-' entry produces this for file-2, and
3113 * a '+' entry produces this for file-1.
3114 */
3115 temp->name = "/dev/null";
3116 xsnprintf(temp->hex, sizeof(temp->hex), ".");
3117 xsnprintf(temp->mode, sizeof(temp->mode), ".");
3118 return temp;
3119 }
3120
3121 if (!S_ISGITLINK(one->mode) &&
3122 (!one->oid_valid ||
3123 reuse_worktree_file(name, &one->oid, 1))) {
3124 struct stat st;
3125 if (lstat(name, &st) < 0) {
3126 if (errno == ENOENT)
3127 goto not_a_valid_file;
3128 die_errno("stat(%s)", name);
3129 }
3130 if (S_ISLNK(st.st_mode)) {
3131 struct strbuf sb = STRBUF_INIT;
3132 if (strbuf_readlink(&sb, name, st.st_size) < 0)
3133 die_errno("readlink(%s)", name);
3134 prep_temp_blob(name, temp, sb.buf, sb.len,
3135 (one->oid_valid ?
3136 &one->oid : &null_oid),
3137 (one->oid_valid ?
3138 one->mode : S_IFLNK));
3139 strbuf_release(&sb);
3140 }
3141 else {
3142 /* we can borrow from the file in the work tree */
3143 temp->name = name;
3144 if (!one->oid_valid)
3145 oid_to_hex_r(temp->hex, &null_oid);
3146 else
3147 oid_to_hex_r(temp->hex, &one->oid);
3148 /* Even though we may sometimes borrow the
3149 * contents from the work tree, we always want
3150 * one->mode. mode is trustworthy even when
3151 * !(one->oid_valid), as long as
3152 * DIFF_FILE_VALID(one).
3153 */
3154 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", one->mode);
3155 }
3156 return temp;
3157 }
3158 else {
3159 if (diff_populate_filespec(one, 0))
3160 die("cannot read data blob for %s", one->path);
3161 prep_temp_blob(name, temp, one->data, one->size,
3162 &one->oid, one->mode);
3163 }
3164 return temp;
3165}
3166
3167static void add_external_diff_name(struct argv_array *argv,
3168 const char *name,
3169 struct diff_filespec *df)
3170{
3171 struct diff_tempfile *temp = prepare_temp_file(name, df);
3172 argv_array_push(argv, temp->name);
3173 argv_array_push(argv, temp->hex);
3174 argv_array_push(argv, temp->mode);
3175}
3176
3177/* An external diff command takes:
3178 *
3179 * diff-cmd name infile1 infile1-sha1 infile1-mode \
3180 * infile2 infile2-sha1 infile2-mode [ rename-to ]
3181 *
3182 */
3183static void run_external_diff(const char *pgm,
3184 const char *name,
3185 const char *other,
3186 struct diff_filespec *one,
3187 struct diff_filespec *two,
3188 const char *xfrm_msg,
3189 int complete_rewrite,
3190 struct diff_options *o)
3191{
3192 struct argv_array argv = ARGV_ARRAY_INIT;
3193 struct argv_array env = ARGV_ARRAY_INIT;
3194 struct diff_queue_struct *q = &diff_queued_diff;
3195
3196 argv_array_push(&argv, pgm);
3197 argv_array_push(&argv, name);
3198
3199 if (one && two) {
3200 add_external_diff_name(&argv, name, one);
3201 if (!other)
3202 add_external_diff_name(&argv, name, two);
3203 else {
3204 add_external_diff_name(&argv, other, two);
3205 argv_array_push(&argv, other);
3206 argv_array_push(&argv, xfrm_msg);
3207 }
3208 }
3209
3210 argv_array_pushf(&env, "GIT_DIFF_PATH_COUNTER=%d", ++o->diff_path_counter);
3211 argv_array_pushf(&env, "GIT_DIFF_PATH_TOTAL=%d", q->nr);
3212
3213 if (run_command_v_opt_cd_env(argv.argv, RUN_USING_SHELL, NULL, env.argv))
3214 die(_("external diff died, stopping at %s"), name);
3215
3216 remove_tempfile();
3217 argv_array_clear(&argv);
3218 argv_array_clear(&env);
3219}
3220
3221static int similarity_index(struct diff_filepair *p)
3222{
3223 return p->score * 100 / MAX_SCORE;
3224}
3225
3226static const char *diff_abbrev_oid(const struct object_id *oid, int abbrev)
3227{
3228 if (startup_info->have_repository)
3229 return find_unique_abbrev(oid->hash, abbrev);
3230 else {
3231 char *hex = oid_to_hex(oid);
3232 if (abbrev < 0)
3233 abbrev = FALLBACK_DEFAULT_ABBREV;
3234 if (abbrev > GIT_SHA1_HEXSZ)
3235 die("BUG: oid abbreviation out of range: %d", abbrev);
3236 if (abbrev)
3237 hex[abbrev] = '\0';
3238 return hex;
3239 }
3240}
3241
3242static void fill_metainfo(struct strbuf *msg,
3243 const char *name,
3244 const char *other,
3245 struct diff_filespec *one,
3246 struct diff_filespec *two,
3247 struct diff_options *o,
3248 struct diff_filepair *p,
3249 int *must_show_header,
3250 int use_color)
3251{
3252 const char *set = diff_get_color(use_color, DIFF_METAINFO);
3253 const char *reset = diff_get_color(use_color, DIFF_RESET);
3254 const char *line_prefix = diff_line_prefix(o);
3255
3256 *must_show_header = 1;
3257 strbuf_init(msg, PATH_MAX * 2 + 300);
3258 switch (p->status) {
3259 case DIFF_STATUS_COPIED:
3260 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3261 line_prefix, set, similarity_index(p));
3262 strbuf_addf(msg, "%s\n%s%scopy from ",
3263 reset, line_prefix, set);
3264 quote_c_style(name, msg, NULL, 0);
3265 strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
3266 quote_c_style(other, msg, NULL, 0);
3267 strbuf_addf(msg, "%s\n", reset);
3268 break;
3269 case DIFF_STATUS_RENAMED:
3270 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3271 line_prefix, set, similarity_index(p));
3272 strbuf_addf(msg, "%s\n%s%srename from ",
3273 reset, line_prefix, set);
3274 quote_c_style(name, msg, NULL, 0);
3275 strbuf_addf(msg, "%s\n%s%srename to ",
3276 reset, line_prefix, set);
3277 quote_c_style(other, msg, NULL, 0);
3278 strbuf_addf(msg, "%s\n", reset);
3279 break;
3280 case DIFF_STATUS_MODIFIED:
3281 if (p->score) {
3282 strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
3283 line_prefix,
3284 set, similarity_index(p), reset);
3285 break;
3286 }
3287 /* fallthru */
3288 default:
3289 *must_show_header = 0;
3290 }
3291 if (one && two && oidcmp(&one->oid, &two->oid)) {
3292 int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
3293
3294 if (DIFF_OPT_TST(o, BINARY)) {
3295 mmfile_t mf;
3296 if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
3297 (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
3298 abbrev = 40;
3299 }
3300 strbuf_addf(msg, "%s%sindex %s..%s", line_prefix, set,
3301 diff_abbrev_oid(&one->oid, abbrev),
3302 diff_abbrev_oid(&two->oid, abbrev));
3303 if (one->mode == two->mode)
3304 strbuf_addf(msg, " %06o", one->mode);
3305 strbuf_addf(msg, "%s\n", reset);
3306 }
3307}
3308
3309static void run_diff_cmd(const char *pgm,
3310 const char *name,
3311 const char *other,
3312 const char *attr_path,
3313 struct diff_filespec *one,
3314 struct diff_filespec *two,
3315 struct strbuf *msg,
3316 struct diff_options *o,
3317 struct diff_filepair *p)
3318{
3319 const char *xfrm_msg = NULL;
3320 int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
3321 int must_show_header = 0;
3322
3323
3324 if (DIFF_OPT_TST(o, ALLOW_EXTERNAL)) {
3325 struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
3326 if (drv && drv->external)
3327 pgm = drv->external;
3328 }
3329
3330 if (msg) {
3331 /*
3332 * don't use colors when the header is intended for an
3333 * external diff driver
3334 */
3335 fill_metainfo(msg, name, other, one, two, o, p,
3336 &must_show_header,
3337 want_color(o->use_color) && !pgm);
3338 xfrm_msg = msg->len ? msg->buf : NULL;
3339 }
3340
3341 if (pgm) {
3342 run_external_diff(pgm, name, other, one, two, xfrm_msg,
3343 complete_rewrite, o);
3344 return;
3345 }
3346 if (one && two)
3347 builtin_diff(name, other ? other : name,
3348 one, two, xfrm_msg, must_show_header,
3349 o, complete_rewrite);
3350 else
3351 fprintf(o->file, "* Unmerged path %s\n", name);
3352}
3353
3354static void diff_fill_oid_info(struct diff_filespec *one)
3355{
3356 if (DIFF_FILE_VALID(one)) {
3357 if (!one->oid_valid) {
3358 struct stat st;
3359 if (one->is_stdin) {
3360 oidclr(&one->oid);
3361 return;
3362 }
3363 if (lstat(one->path, &st) < 0)
3364 die_errno("stat '%s'", one->path);
3365 if (index_path(one->oid.hash, one->path, &st, 0))
3366 die("cannot hash %s", one->path);
3367 }
3368 }
3369 else
3370 oidclr(&one->oid);
3371}
3372
3373static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
3374{
3375 /* Strip the prefix but do not molest /dev/null and absolute paths */
3376 if (*namep && **namep != '/') {
3377 *namep += prefix_length;
3378 if (**namep == '/')
3379 ++*namep;
3380 }
3381 if (*otherp && **otherp != '/') {
3382 *otherp += prefix_length;
3383 if (**otherp == '/')
3384 ++*otherp;
3385 }
3386}
3387
3388static void run_diff(struct diff_filepair *p, struct diff_options *o)
3389{
3390 const char *pgm = external_diff();
3391 struct strbuf msg;
3392 struct diff_filespec *one = p->one;
3393 struct diff_filespec *two = p->two;
3394 const char *name;
3395 const char *other;
3396 const char *attr_path;
3397
3398 name = one->path;
3399 other = (strcmp(name, two->path) ? two->path : NULL);
3400 attr_path = name;
3401 if (o->prefix_length)
3402 strip_prefix(o->prefix_length, &name, &other);
3403
3404 if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
3405 pgm = NULL;
3406
3407 if (DIFF_PAIR_UNMERGED(p)) {
3408 run_diff_cmd(pgm, name, NULL, attr_path,
3409 NULL, NULL, NULL, o, p);
3410 return;
3411 }
3412
3413 diff_fill_oid_info(one);
3414 diff_fill_oid_info(two);
3415
3416 if (!pgm &&
3417 DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3418 (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3419 /*
3420 * a filepair that changes between file and symlink
3421 * needs to be split into deletion and creation.
3422 */
3423 struct diff_filespec *null = alloc_filespec(two->path);
3424 run_diff_cmd(NULL, name, other, attr_path,
3425 one, null, &msg, o, p);
3426 free(null);
3427 strbuf_release(&msg);
3428
3429 null = alloc_filespec(one->path);
3430 run_diff_cmd(NULL, name, other, attr_path,
3431 null, two, &msg, o, p);
3432 free(null);
3433 }
3434 else
3435 run_diff_cmd(pgm, name, other, attr_path,
3436 one, two, &msg, o, p);
3437
3438 strbuf_release(&msg);
3439}
3440
3441static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3442 struct diffstat_t *diffstat)
3443{
3444 const char *name;
3445 const char *other;
3446
3447 if (DIFF_PAIR_UNMERGED(p)) {
3448 /* unmerged */
3449 builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, p);
3450 return;
3451 }
3452
3453 name = p->one->path;
3454 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3455
3456 if (o->prefix_length)
3457 strip_prefix(o->prefix_length, &name, &other);
3458
3459 diff_fill_oid_info(p->one);
3460 diff_fill_oid_info(p->two);
3461
3462 builtin_diffstat(name, other, p->one, p->two, diffstat, o, p);
3463}
3464
3465static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3466{
3467 const char *name;
3468 const char *other;
3469 const char *attr_path;
3470
3471 if (DIFF_PAIR_UNMERGED(p)) {
3472 /* unmerged */
3473 return;
3474 }
3475
3476 name = p->one->path;
3477 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3478 attr_path = other ? other : name;
3479
3480 if (o->prefix_length)
3481 strip_prefix(o->prefix_length, &name, &other);
3482
3483 diff_fill_oid_info(p->one);
3484 diff_fill_oid_info(p->two);
3485
3486 builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3487}
3488
3489void diff_setup(struct diff_options *options)
3490{
3491 memcpy(options, &default_diff_options, sizeof(*options));
3492
3493 options->file = stdout;
3494
3495 options->abbrev = DEFAULT_ABBREV;
3496 options->line_termination = '\n';
3497 options->break_opt = -1;
3498 options->rename_limit = -1;
3499 options->dirstat_permille = diff_dirstat_permille_default;
3500 options->context = diff_context_default;
3501 options->interhunkcontext = diff_interhunk_context_default;
3502 options->ws_error_highlight = ws_error_highlight_default;
3503 DIFF_OPT_SET(options, RENAME_EMPTY);
3504
3505 /* pathchange left =NULL by default */
3506 options->change = diff_change;
3507 options->add_remove = diff_addremove;
3508 options->use_color = diff_use_color_default;
3509 options->detect_rename = diff_detect_rename_default;
3510 options->xdl_opts |= diff_algorithm;
3511 if (diff_indent_heuristic)
3512 DIFF_XDL_SET(options, INDENT_HEURISTIC);
3513
3514 options->orderfile = diff_order_file_cfg;
3515
3516 if (diff_no_prefix) {
3517 options->a_prefix = options->b_prefix = "";
3518 } else if (!diff_mnemonic_prefix) {
3519 options->a_prefix = "a/";
3520 options->b_prefix = "b/";
3521 }
3522}
3523
3524void diff_setup_done(struct diff_options *options)
3525{
3526 int count = 0;
3527
3528 if (options->set_default)
3529 options->set_default(options);
3530
3531 if (options->output_format & DIFF_FORMAT_NAME)
3532 count++;
3533 if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3534 count++;
3535 if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3536 count++;
3537 if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3538 count++;
3539 if (count > 1)
3540 die(_("--name-only, --name-status, --check and -s are mutually exclusive"));
3541
3542 /*
3543 * Most of the time we can say "there are changes"
3544 * only by checking if there are changed paths, but
3545 * --ignore-whitespace* options force us to look
3546 * inside contents.
3547 */
3548
3549 if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3550 DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3551 DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3552 DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3553 else
3554 DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3555
3556 if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3557 options->detect_rename = DIFF_DETECT_COPY;
3558
3559 if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3560 options->prefix = NULL;
3561 if (options->prefix)
3562 options->prefix_length = strlen(options->prefix);
3563 else
3564 options->prefix_length = 0;
3565
3566 if (options->output_format & (DIFF_FORMAT_NAME |
3567 DIFF_FORMAT_NAME_STATUS |
3568 DIFF_FORMAT_CHECKDIFF |
3569 DIFF_FORMAT_NO_OUTPUT))
3570 options->output_format &= ~(DIFF_FORMAT_RAW |
3571 DIFF_FORMAT_NUMSTAT |
3572 DIFF_FORMAT_DIFFSTAT |
3573 DIFF_FORMAT_SHORTSTAT |
3574 DIFF_FORMAT_DIRSTAT |
3575 DIFF_FORMAT_SUMMARY |
3576 DIFF_FORMAT_PATCH);
3577
3578 /*
3579 * These cases always need recursive; we do not drop caller-supplied
3580 * recursive bits for other formats here.
3581 */
3582 if (options->output_format & (DIFF_FORMAT_PATCH |
3583 DIFF_FORMAT_NUMSTAT |
3584 DIFF_FORMAT_DIFFSTAT |
3585 DIFF_FORMAT_SHORTSTAT |
3586 DIFF_FORMAT_DIRSTAT |
3587 DIFF_FORMAT_SUMMARY |
3588 DIFF_FORMAT_CHECKDIFF))
3589 DIFF_OPT_SET(options, RECURSIVE);
3590 /*
3591 * Also pickaxe would not work very well if you do not say recursive
3592 */
3593 if (options->pickaxe)
3594 DIFF_OPT_SET(options, RECURSIVE);
3595 /*
3596 * When patches are generated, submodules diffed against the work tree
3597 * must be checked for dirtiness too so it can be shown in the output
3598 */
3599 if (options->output_format & DIFF_FORMAT_PATCH)
3600 DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3601
3602 if (options->detect_rename && options->rename_limit < 0)
3603 options->rename_limit = diff_rename_limit_default;
3604 if (options->setup & DIFF_SETUP_USE_CACHE) {
3605 if (!active_cache)
3606 /* read-cache does not die even when it fails
3607 * so it is safe for us to do this here. Also
3608 * it does not smudge active_cache or active_nr
3609 * when it fails, so we do not have to worry about
3610 * cleaning it up ourselves either.
3611 */
3612 read_cache();
3613 }
3614 if (40 < options->abbrev)
3615 options->abbrev = 40; /* full */
3616
3617 /*
3618 * It does not make sense to show the first hit we happened
3619 * to have found. It does not make sense not to return with
3620 * exit code in such a case either.
3621 */
3622 if (DIFF_OPT_TST(options, QUICK)) {
3623 options->output_format = DIFF_FORMAT_NO_OUTPUT;
3624 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3625 }
3626
3627 options->diff_path_counter = 0;
3628
3629 if (DIFF_OPT_TST(options, FOLLOW_RENAMES) && options->pathspec.nr != 1)
3630 die(_("--follow requires exactly one pathspec"));
3631}
3632
3633static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3634{
3635 char c, *eq;
3636 int len;
3637
3638 if (*arg != '-')
3639 return 0;
3640 c = *++arg;
3641 if (!c)
3642 return 0;
3643 if (c == arg_short) {
3644 c = *++arg;
3645 if (!c)
3646 return 1;
3647 if (val && isdigit(c)) {
3648 char *end;
3649 int n = strtoul(arg, &end, 10);
3650 if (*end)
3651 return 0;
3652 *val = n;
3653 return 1;
3654 }
3655 return 0;
3656 }
3657 if (c != '-')
3658 return 0;
3659 arg++;
3660 eq = strchrnul(arg, '=');
3661 len = eq - arg;
3662 if (!len || strncmp(arg, arg_long, len))
3663 return 0;
3664 if (*eq) {
3665 int n;
3666 char *end;
3667 if (!isdigit(*++eq))
3668 return 0;
3669 n = strtoul(eq, &end, 10);
3670 if (*end)
3671 return 0;
3672 *val = n;
3673 }
3674 return 1;
3675}
3676
3677static int diff_scoreopt_parse(const char *opt);
3678
3679static inline int short_opt(char opt, const char **argv,
3680 const char **optarg)
3681{
3682 const char *arg = argv[0];
3683 if (arg[0] != '-' || arg[1] != opt)
3684 return 0;
3685 if (arg[2] != '\0') {
3686 *optarg = arg + 2;
3687 return 1;
3688 }
3689 if (!argv[1])
3690 die("Option '%c' requires a value", opt);
3691 *optarg = argv[1];
3692 return 2;
3693}
3694
3695int parse_long_opt(const char *opt, const char **argv,
3696 const char **optarg)
3697{
3698 const char *arg = argv[0];
3699 if (!skip_prefix(arg, "--", &arg))
3700 return 0;
3701 if (!skip_prefix(arg, opt, &arg))
3702 return 0;
3703 if (*arg == '=') { /* stuck form: --option=value */
3704 *optarg = arg + 1;
3705 return 1;
3706 }
3707 if (*arg != '\0')
3708 return 0;
3709 /* separate form: --option value */
3710 if (!argv[1])
3711 die("Option '--%s' requires a value", opt);
3712 *optarg = argv[1];
3713 return 2;
3714}
3715
3716static int stat_opt(struct diff_options *options, const char **av)
3717{
3718 const char *arg = av[0];
3719 char *end;
3720 int width = options->stat_width;
3721 int name_width = options->stat_name_width;
3722 int graph_width = options->stat_graph_width;
3723 int count = options->stat_count;
3724 int argcount = 1;
3725
3726 if (!skip_prefix(arg, "--stat", &arg))
3727 die("BUG: stat option does not begin with --stat: %s", arg);
3728 end = (char *)arg;
3729
3730 switch (*arg) {
3731 case '-':
3732 if (skip_prefix(arg, "-width", &arg)) {
3733 if (*arg == '=')
3734 width = strtoul(arg + 1, &end, 10);
3735 else if (!*arg && !av[1])
3736 die_want_option("--stat-width");
3737 else if (!*arg) {
3738 width = strtoul(av[1], &end, 10);
3739 argcount = 2;
3740 }
3741 } else if (skip_prefix(arg, "-name-width", &arg)) {
3742 if (*arg == '=')
3743 name_width = strtoul(arg + 1, &end, 10);
3744 else if (!*arg && !av[1])
3745 die_want_option("--stat-name-width");
3746 else if (!*arg) {
3747 name_width = strtoul(av[1], &end, 10);
3748 argcount = 2;
3749 }
3750 } else if (skip_prefix(arg, "-graph-width", &arg)) {
3751 if (*arg == '=')
3752 graph_width = strtoul(arg + 1, &end, 10);
3753 else if (!*arg && !av[1])
3754 die_want_option("--stat-graph-width");
3755 else if (!*arg) {
3756 graph_width = strtoul(av[1], &end, 10);
3757 argcount = 2;
3758 }
3759 } else if (skip_prefix(arg, "-count", &arg)) {
3760 if (*arg == '=')
3761 count = strtoul(arg + 1, &end, 10);
3762 else if (!*arg && !av[1])
3763 die_want_option("--stat-count");
3764 else if (!*arg) {
3765 count = strtoul(av[1], &end, 10);
3766 argcount = 2;
3767 }
3768 }
3769 break;
3770 case '=':
3771 width = strtoul(arg+1, &end, 10);
3772 if (*end == ',')
3773 name_width = strtoul(end+1, &end, 10);
3774 if (*end == ',')
3775 count = strtoul(end+1, &end, 10);
3776 }
3777
3778 /* Important! This checks all the error cases! */
3779 if (*end)
3780 return 0;
3781 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3782 options->stat_name_width = name_width;
3783 options->stat_graph_width = graph_width;
3784 options->stat_width = width;
3785 options->stat_count = count;
3786 return argcount;
3787}
3788
3789static int parse_dirstat_opt(struct diff_options *options, const char *params)
3790{
3791 struct strbuf errmsg = STRBUF_INIT;
3792 if (parse_dirstat_params(options, params, &errmsg))
3793 die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3794 errmsg.buf);
3795 strbuf_release(&errmsg);
3796 /*
3797 * The caller knows a dirstat-related option is given from the command
3798 * line; allow it to say "return this_function();"
3799 */
3800 options->output_format |= DIFF_FORMAT_DIRSTAT;
3801 return 1;
3802}
3803
3804static int parse_submodule_opt(struct diff_options *options, const char *value)
3805{
3806 if (parse_submodule_params(options, value))
3807 die(_("Failed to parse --submodule option parameter: '%s'"),
3808 value);
3809 return 1;
3810}
3811
3812static const char diff_status_letters[] = {
3813 DIFF_STATUS_ADDED,
3814 DIFF_STATUS_COPIED,
3815 DIFF_STATUS_DELETED,
3816 DIFF_STATUS_MODIFIED,
3817 DIFF_STATUS_RENAMED,
3818 DIFF_STATUS_TYPE_CHANGED,
3819 DIFF_STATUS_UNKNOWN,
3820 DIFF_STATUS_UNMERGED,
3821 DIFF_STATUS_FILTER_AON,
3822 DIFF_STATUS_FILTER_BROKEN,
3823 '\0',
3824};
3825
3826static unsigned int filter_bit['Z' + 1];
3827
3828static void prepare_filter_bits(void)
3829{
3830 int i;
3831
3832 if (!filter_bit[DIFF_STATUS_ADDED]) {
3833 for (i = 0; diff_status_letters[i]; i++)
3834 filter_bit[(int) diff_status_letters[i]] = (1 << i);
3835 }
3836}
3837
3838static unsigned filter_bit_tst(char status, const struct diff_options *opt)
3839{
3840 return opt->filter & filter_bit[(int) status];
3841}
3842
3843static int parse_diff_filter_opt(const char *optarg, struct diff_options *opt)
3844{
3845 int i, optch;
3846
3847 prepare_filter_bits();
3848
3849 /*
3850 * If there is a negation e.g. 'd' in the input, and we haven't
3851 * initialized the filter field with another --diff-filter, start
3852 * from full set of bits, except for AON.
3853 */
3854 if (!opt->filter) {
3855 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3856 if (optch < 'a' || 'z' < optch)
3857 continue;
3858 opt->filter = (1 << (ARRAY_SIZE(diff_status_letters) - 1)) - 1;
3859 opt->filter &= ~filter_bit[DIFF_STATUS_FILTER_AON];
3860 break;
3861 }
3862 }
3863
3864 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3865 unsigned int bit;
3866 int negate;
3867
3868 if ('a' <= optch && optch <= 'z') {
3869 negate = 1;
3870 optch = toupper(optch);
3871 } else {
3872 negate = 0;
3873 }
3874
3875 bit = (0 <= optch && optch <= 'Z') ? filter_bit[optch] : 0;
3876 if (!bit)
3877 return optarg[i];
3878 if (negate)
3879 opt->filter &= ~bit;
3880 else
3881 opt->filter |= bit;
3882 }
3883 return 0;
3884}
3885
3886static void enable_patch_output(int *fmt) {
3887 *fmt &= ~DIFF_FORMAT_NO_OUTPUT;
3888 *fmt |= DIFF_FORMAT_PATCH;
3889}
3890
3891static int parse_ws_error_highlight_opt(struct diff_options *opt, const char *arg)
3892{
3893 int val = parse_ws_error_highlight(arg);
3894
3895 if (val < 0) {
3896 error("unknown value after ws-error-highlight=%.*s",
3897 -1 - val, arg);
3898 return 0;
3899 }
3900 opt->ws_error_highlight = val;
3901 return 1;
3902}
3903
3904int diff_opt_parse(struct diff_options *options,
3905 const char **av, int ac, const char *prefix)
3906{
3907 const char *arg = av[0];
3908 const char *optarg;
3909 int argcount;
3910
3911 if (!prefix)
3912 prefix = "";
3913
3914 /* Output format options */
3915 if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch")
3916 || opt_arg(arg, 'U', "unified", &options->context))
3917 enable_patch_output(&options->output_format);
3918 else if (!strcmp(arg, "--raw"))
3919 options->output_format |= DIFF_FORMAT_RAW;
3920 else if (!strcmp(arg, "--patch-with-raw")) {
3921 enable_patch_output(&options->output_format);
3922 options->output_format |= DIFF_FORMAT_RAW;
3923 } else if (!strcmp(arg, "--numstat"))
3924 options->output_format |= DIFF_FORMAT_NUMSTAT;
3925 else if (!strcmp(arg, "--shortstat"))
3926 options->output_format |= DIFF_FORMAT_SHORTSTAT;
3927 else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3928 return parse_dirstat_opt(options, "");
3929 else if (skip_prefix(arg, "-X", &arg))
3930 return parse_dirstat_opt(options, arg);
3931 else if (skip_prefix(arg, "--dirstat=", &arg))
3932 return parse_dirstat_opt(options, arg);
3933 else if (!strcmp(arg, "--cumulative"))
3934 return parse_dirstat_opt(options, "cumulative");
3935 else if (!strcmp(arg, "--dirstat-by-file"))
3936 return parse_dirstat_opt(options, "files");
3937 else if (skip_prefix(arg, "--dirstat-by-file=", &arg)) {
3938 parse_dirstat_opt(options, "files");
3939 return parse_dirstat_opt(options, arg);
3940 }
3941 else if (!strcmp(arg, "--check"))
3942 options->output_format |= DIFF_FORMAT_CHECKDIFF;
3943 else if (!strcmp(arg, "--summary"))
3944 options->output_format |= DIFF_FORMAT_SUMMARY;
3945 else if (!strcmp(arg, "--patch-with-stat")) {
3946 enable_patch_output(&options->output_format);
3947 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3948 } else if (!strcmp(arg, "--name-only"))
3949 options->output_format |= DIFF_FORMAT_NAME;
3950 else if (!strcmp(arg, "--name-status"))
3951 options->output_format |= DIFF_FORMAT_NAME_STATUS;
3952 else if (!strcmp(arg, "-s") || !strcmp(arg, "--no-patch"))
3953 options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3954 else if (starts_with(arg, "--stat"))
3955 /* --stat, --stat-width, --stat-name-width, or --stat-count */
3956 return stat_opt(options, av);
3957
3958 /* renames options */
3959 else if (starts_with(arg, "-B") || starts_with(arg, "--break-rewrites=") ||
3960 !strcmp(arg, "--break-rewrites")) {
3961 if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3962 return error("invalid argument to -B: %s", arg+2);
3963 }
3964 else if (starts_with(arg, "-M") || starts_with(arg, "--find-renames=") ||
3965 !strcmp(arg, "--find-renames")) {
3966 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3967 return error("invalid argument to -M: %s", arg+2);
3968 options->detect_rename = DIFF_DETECT_RENAME;
3969 }
3970 else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3971 options->irreversible_delete = 1;
3972 }
3973 else if (starts_with(arg, "-C") || starts_with(arg, "--find-copies=") ||
3974 !strcmp(arg, "--find-copies")) {
3975 if (options->detect_rename == DIFF_DETECT_COPY)
3976 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3977 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3978 return error("invalid argument to -C: %s", arg+2);
3979 options->detect_rename = DIFF_DETECT_COPY;
3980 }
3981 else if (!strcmp(arg, "--no-renames"))
3982 options->detect_rename = 0;
3983 else if (!strcmp(arg, "--rename-empty"))
3984 DIFF_OPT_SET(options, RENAME_EMPTY);
3985 else if (!strcmp(arg, "--no-rename-empty"))
3986 DIFF_OPT_CLR(options, RENAME_EMPTY);
3987 else if (!strcmp(arg, "--relative"))
3988 DIFF_OPT_SET(options, RELATIVE_NAME);
3989 else if (skip_prefix(arg, "--relative=", &arg)) {
3990 DIFF_OPT_SET(options, RELATIVE_NAME);
3991 options->prefix = arg;
3992 }
3993
3994 /* xdiff options */
3995 else if (!strcmp(arg, "--minimal"))
3996 DIFF_XDL_SET(options, NEED_MINIMAL);
3997 else if (!strcmp(arg, "--no-minimal"))
3998 DIFF_XDL_CLR(options, NEED_MINIMAL);
3999 else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
4000 DIFF_XDL_SET(options, IGNORE_WHITESPACE);
4001 else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
4002 DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
4003 else if (!strcmp(arg, "--ignore-space-at-eol"))
4004 DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
4005 else if (!strcmp(arg, "--ignore-blank-lines"))
4006 DIFF_XDL_SET(options, IGNORE_BLANK_LINES);
4007 else if (!strcmp(arg, "--indent-heuristic"))
4008 DIFF_XDL_SET(options, INDENT_HEURISTIC);
4009 else if (!strcmp(arg, "--no-indent-heuristic"))
4010 DIFF_XDL_CLR(options, INDENT_HEURISTIC);
4011 else if (!strcmp(arg, "--patience"))
4012 options->xdl_opts = DIFF_WITH_ALG(options, PATIENCE_DIFF);
4013 else if (!strcmp(arg, "--histogram"))
4014 options->xdl_opts = DIFF_WITH_ALG(options, HISTOGRAM_DIFF);
4015 else if ((argcount = parse_long_opt("diff-algorithm", av, &optarg))) {
4016 long value = parse_algorithm_value(optarg);
4017 if (value < 0)
4018 return error("option diff-algorithm accepts \"myers\", "
4019 "\"minimal\", \"patience\" and \"histogram\"");
4020 /* clear out previous settings */
4021 DIFF_XDL_CLR(options, NEED_MINIMAL);
4022 options->xdl_opts &= ~XDF_DIFF_ALGORITHM_MASK;
4023 options->xdl_opts |= value;
4024 return argcount;
4025 }
4026
4027 /* flags options */
4028 else if (!strcmp(arg, "--binary")) {
4029 enable_patch_output(&options->output_format);
4030 DIFF_OPT_SET(options, BINARY);
4031 }
4032 else if (!strcmp(arg, "--full-index"))
4033 DIFF_OPT_SET(options, FULL_INDEX);
4034 else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
4035 DIFF_OPT_SET(options, TEXT);
4036 else if (!strcmp(arg, "-R"))
4037 DIFF_OPT_SET(options, REVERSE_DIFF);
4038 else if (!strcmp(arg, "--find-copies-harder"))
4039 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
4040 else if (!strcmp(arg, "--follow"))
4041 DIFF_OPT_SET(options, FOLLOW_RENAMES);
4042 else if (!strcmp(arg, "--no-follow")) {
4043 DIFF_OPT_CLR(options, FOLLOW_RENAMES);
4044 DIFF_OPT_CLR(options, DEFAULT_FOLLOW_RENAMES);
4045 } else if (!strcmp(arg, "--color"))
4046 options->use_color = 1;
4047 else if (skip_prefix(arg, "--color=", &arg)) {
4048 int value = git_config_colorbool(NULL, arg);
4049 if (value < 0)
4050 return error("option `color' expects \"always\", \"auto\", or \"never\"");
4051 options->use_color = value;
4052 }
4053 else if (!strcmp(arg, "--no-color"))
4054 options->use_color = 0;
4055 else if (!strcmp(arg, "--color-words")) {
4056 options->use_color = 1;
4057 options->word_diff = DIFF_WORDS_COLOR;
4058 }
4059 else if (skip_prefix(arg, "--color-words=", &arg)) {
4060 options->use_color = 1;
4061 options->word_diff = DIFF_WORDS_COLOR;
4062 options->word_regex = arg;
4063 }
4064 else if (!strcmp(arg, "--word-diff")) {
4065 if (options->word_diff == DIFF_WORDS_NONE)
4066 options->word_diff = DIFF_WORDS_PLAIN;
4067 }
4068 else if (skip_prefix(arg, "--word-diff=", &arg)) {
4069 if (!strcmp(arg, "plain"))
4070 options->word_diff = DIFF_WORDS_PLAIN;
4071 else if (!strcmp(arg, "color")) {
4072 options->use_color = 1;
4073 options->word_diff = DIFF_WORDS_COLOR;
4074 }
4075 else if (!strcmp(arg, "porcelain"))
4076 options->word_diff = DIFF_WORDS_PORCELAIN;
4077 else if (!strcmp(arg, "none"))
4078 options->word_diff = DIFF_WORDS_NONE;
4079 else
4080 die("bad --word-diff argument: %s", arg);
4081 }
4082 else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
4083 if (options->word_diff == DIFF_WORDS_NONE)
4084 options->word_diff = DIFF_WORDS_PLAIN;
4085 options->word_regex = optarg;
4086 return argcount;
4087 }
4088 else if (!strcmp(arg, "--exit-code"))
4089 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
4090 else if (!strcmp(arg, "--quiet"))
4091 DIFF_OPT_SET(options, QUICK);
4092 else if (!strcmp(arg, "--ext-diff"))
4093 DIFF_OPT_SET(options, ALLOW_EXTERNAL);
4094 else if (!strcmp(arg, "--no-ext-diff"))
4095 DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
4096 else if (!strcmp(arg, "--textconv"))
4097 DIFF_OPT_SET(options, ALLOW_TEXTCONV);
4098 else if (!strcmp(arg, "--no-textconv"))
4099 DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
4100 else if (!strcmp(arg, "--ignore-submodules")) {
4101 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
4102 handle_ignore_submodules_arg(options, "all");
4103 } else if (skip_prefix(arg, "--ignore-submodules=", &arg)) {
4104 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
4105 handle_ignore_submodules_arg(options, arg);
4106 } else if (!strcmp(arg, "--submodule"))
4107 options->submodule_format = DIFF_SUBMODULE_LOG;
4108 else if (skip_prefix(arg, "--submodule=", &arg))
4109 return parse_submodule_opt(options, arg);
4110 else if (skip_prefix(arg, "--ws-error-highlight=", &arg))
4111 return parse_ws_error_highlight_opt(options, arg);
4112 else if (!strcmp(arg, "--ita-invisible-in-index"))
4113 options->ita_invisible_in_index = 1;
4114 else if (!strcmp(arg, "--ita-visible-in-index"))
4115 options->ita_invisible_in_index = 0;
4116
4117 /* misc options */
4118 else if (!strcmp(arg, "-z"))
4119 options->line_termination = 0;
4120 else if ((argcount = short_opt('l', av, &optarg))) {
4121 options->rename_limit = strtoul(optarg, NULL, 10);
4122 return argcount;
4123 }
4124 else if ((argcount = short_opt('S', av, &optarg))) {
4125 options->pickaxe = optarg;
4126 options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
4127 return argcount;
4128 } else if ((argcount = short_opt('G', av, &optarg))) {
4129 options->pickaxe = optarg;
4130 options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
4131 return argcount;
4132 }
4133 else if (!strcmp(arg, "--pickaxe-all"))
4134 options->pickaxe_opts |= DIFF_PICKAXE_ALL;
4135 else if (!strcmp(arg, "--pickaxe-regex"))
4136 options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
4137 else if ((argcount = short_opt('O', av, &optarg))) {
4138 options->orderfile = prefix_filename(prefix, optarg);
4139 return argcount;
4140 }
4141 else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
4142 int offending = parse_diff_filter_opt(optarg, options);
4143 if (offending)
4144 die("unknown change class '%c' in --diff-filter=%s",
4145 offending, optarg);
4146 return argcount;
4147 }
4148 else if (!strcmp(arg, "--no-abbrev"))
4149 options->abbrev = 0;
4150 else if (!strcmp(arg, "--abbrev"))
4151 options->abbrev = DEFAULT_ABBREV;
4152 else if (skip_prefix(arg, "--abbrev=", &arg)) {
4153 options->abbrev = strtoul(arg, NULL, 10);
4154 if (options->abbrev < MINIMUM_ABBREV)
4155 options->abbrev = MINIMUM_ABBREV;
4156 else if (40 < options->abbrev)
4157 options->abbrev = 40;
4158 }
4159 else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
4160 options->a_prefix = optarg;
4161 return argcount;
4162 }
4163 else if ((argcount = parse_long_opt("line-prefix", av, &optarg))) {
4164 options->line_prefix = optarg;
4165 options->line_prefix_length = strlen(options->line_prefix);
4166 graph_setup_line_prefix(options);
4167 return argcount;
4168 }
4169 else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
4170 options->b_prefix = optarg;
4171 return argcount;
4172 }
4173 else if (!strcmp(arg, "--no-prefix"))
4174 options->a_prefix = options->b_prefix = "";
4175 else if (opt_arg(arg, '\0', "inter-hunk-context",
4176 &options->interhunkcontext))
4177 ;
4178 else if (!strcmp(arg, "-W"))
4179 DIFF_OPT_SET(options, FUNCCONTEXT);
4180 else if (!strcmp(arg, "--function-context"))
4181 DIFF_OPT_SET(options, FUNCCONTEXT);
4182 else if (!strcmp(arg, "--no-function-context"))
4183 DIFF_OPT_CLR(options, FUNCCONTEXT);
4184 else if ((argcount = parse_long_opt("output", av, &optarg))) {
4185 char *path = prefix_filename(prefix, optarg);
4186 options->file = xfopen(path, "w");
4187 options->close_file = 1;
4188 if (options->use_color != GIT_COLOR_ALWAYS)
4189 options->use_color = GIT_COLOR_NEVER;
4190 free(path);
4191 return argcount;
4192 } else
4193 return 0;
4194 return 1;
4195}
4196
4197int parse_rename_score(const char **cp_p)
4198{
4199 unsigned long num, scale;
4200 int ch, dot;
4201 const char *cp = *cp_p;
4202
4203 num = 0;
4204 scale = 1;
4205 dot = 0;
4206 for (;;) {
4207 ch = *cp;
4208 if ( !dot && ch == '.' ) {
4209 scale = 1;
4210 dot = 1;
4211 } else if ( ch == '%' ) {
4212 scale = dot ? scale*100 : 100;
4213 cp++; /* % is always at the end */
4214 break;
4215 } else if ( ch >= '0' && ch <= '9' ) {
4216 if ( scale < 100000 ) {
4217 scale *= 10;
4218 num = (num*10) + (ch-'0');
4219 }
4220 } else {
4221 break;
4222 }
4223 cp++;
4224 }
4225 *cp_p = cp;
4226
4227 /* user says num divided by scale and we say internally that
4228 * is MAX_SCORE * num / scale.
4229 */
4230 return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
4231}
4232
4233static int diff_scoreopt_parse(const char *opt)
4234{
4235 int opt1, opt2, cmd;
4236
4237 if (*opt++ != '-')
4238 return -1;
4239 cmd = *opt++;
4240 if (cmd == '-') {
4241 /* convert the long-form arguments into short-form versions */
4242 if (skip_prefix(opt, "break-rewrites", &opt)) {
4243 if (*opt == 0 || *opt++ == '=')
4244 cmd = 'B';
4245 } else if (skip_prefix(opt, "find-copies", &opt)) {
4246 if (*opt == 0 || *opt++ == '=')
4247 cmd = 'C';
4248 } else if (skip_prefix(opt, "find-renames", &opt)) {
4249 if (*opt == 0 || *opt++ == '=')
4250 cmd = 'M';
4251 }
4252 }
4253 if (cmd != 'M' && cmd != 'C' && cmd != 'B')
4254 return -1; /* that is not a -M, -C, or -B option */
4255
4256 opt1 = parse_rename_score(&opt);
4257 if (cmd != 'B')
4258 opt2 = 0;
4259 else {
4260 if (*opt == 0)
4261 opt2 = 0;
4262 else if (*opt != '/')
4263 return -1; /* we expect -B80/99 or -B80 */
4264 else {
4265 opt++;
4266 opt2 = parse_rename_score(&opt);
4267 }
4268 }
4269 if (*opt != 0)
4270 return -1;
4271 return opt1 | (opt2 << 16);
4272}
4273
4274struct diff_queue_struct diff_queued_diff;
4275
4276void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
4277{
4278 ALLOC_GROW(queue->queue, queue->nr + 1, queue->alloc);
4279 queue->queue[queue->nr++] = dp;
4280}
4281
4282struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
4283 struct diff_filespec *one,
4284 struct diff_filespec *two)
4285{
4286 struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
4287 dp->one = one;
4288 dp->two = two;
4289 if (queue)
4290 diff_q(queue, dp);
4291 return dp;
4292}
4293
4294void diff_free_filepair(struct diff_filepair *p)
4295{
4296 free_filespec(p->one);
4297 free_filespec(p->two);
4298 free(p);
4299}
4300
4301const char *diff_aligned_abbrev(const struct object_id *oid, int len)
4302{
4303 int abblen;
4304 const char *abbrev;
4305
4306 if (len == GIT_SHA1_HEXSZ)
4307 return oid_to_hex(oid);
4308
4309 abbrev = diff_abbrev_oid(oid, len);
4310 abblen = strlen(abbrev);
4311
4312 /*
4313 * In well-behaved cases, where the abbbreviated result is the
4314 * same as the requested length, append three dots after the
4315 * abbreviation (hence the whole logic is limited to the case
4316 * where abblen < 37); when the actual abbreviated result is a
4317 * bit longer than the requested length, we reduce the number
4318 * of dots so that they match the well-behaved ones. However,
4319 * if the actual abbreviation is longer than the requested
4320 * length by more than three, we give up on aligning, and add
4321 * three dots anyway, to indicate that the output is not the
4322 * full object name. Yes, this may be suboptimal, but this
4323 * appears only in "diff --raw --abbrev" output and it is not
4324 * worth the effort to change it now. Note that this would
4325 * likely to work fine when the automatic sizing of default
4326 * abbreviation length is used--we would be fed -1 in "len" in
4327 * that case, and will end up always appending three-dots, but
4328 * the automatic sizing is supposed to give abblen that ensures
4329 * uniqueness across all objects (statistically speaking).
4330 */
4331 if (abblen < GIT_SHA1_HEXSZ - 3) {
4332 static char hex[GIT_MAX_HEXSZ + 1];
4333 if (len < abblen && abblen <= len + 2)
4334 xsnprintf(hex, sizeof(hex), "%s%.*s", abbrev, len+3-abblen, "..");
4335 else
4336 xsnprintf(hex, sizeof(hex), "%s...", abbrev);
4337 return hex;
4338 }
4339
4340 return oid_to_hex(oid);
4341}
4342
4343static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
4344{
4345 int line_termination = opt->line_termination;
4346 int inter_name_termination = line_termination ? '\t' : '\0';
4347
4348 fprintf(opt->file, "%s", diff_line_prefix(opt));
4349 if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
4350 fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
4351 diff_aligned_abbrev(&p->one->oid, opt->abbrev));
4352 fprintf(opt->file, "%s ",
4353 diff_aligned_abbrev(&p->two->oid, opt->abbrev));
4354 }
4355 if (p->score) {
4356 fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
4357 inter_name_termination);
4358 } else {
4359 fprintf(opt->file, "%c%c", p->status, inter_name_termination);
4360 }
4361
4362 if (p->status == DIFF_STATUS_COPIED ||
4363 p->status == DIFF_STATUS_RENAMED) {
4364 const char *name_a, *name_b;
4365 name_a = p->one->path;
4366 name_b = p->two->path;
4367 strip_prefix(opt->prefix_length, &name_a, &name_b);
4368 write_name_quoted(name_a, opt->file, inter_name_termination);
4369 write_name_quoted(name_b, opt->file, line_termination);
4370 } else {
4371 const char *name_a, *name_b;
4372 name_a = p->one->mode ? p->one->path : p->two->path;
4373 name_b = NULL;
4374 strip_prefix(opt->prefix_length, &name_a, &name_b);
4375 write_name_quoted(name_a, opt->file, line_termination);
4376 }
4377}
4378
4379int diff_unmodified_pair(struct diff_filepair *p)
4380{
4381 /* This function is written stricter than necessary to support
4382 * the currently implemented transformers, but the idea is to
4383 * let transformers to produce diff_filepairs any way they want,
4384 * and filter and clean them up here before producing the output.
4385 */
4386 struct diff_filespec *one = p->one, *two = p->two;
4387
4388 if (DIFF_PAIR_UNMERGED(p))
4389 return 0; /* unmerged is interesting */
4390
4391 /* deletion, addition, mode or type change
4392 * and rename are all interesting.
4393 */
4394 if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
4395 DIFF_PAIR_MODE_CHANGED(p) ||
4396 strcmp(one->path, two->path))
4397 return 0;
4398
4399 /* both are valid and point at the same path. that is, we are
4400 * dealing with a change.
4401 */
4402 if (one->oid_valid && two->oid_valid &&
4403 !oidcmp(&one->oid, &two->oid) &&
4404 !one->dirty_submodule && !two->dirty_submodule)
4405 return 1; /* no change */
4406 if (!one->oid_valid && !two->oid_valid)
4407 return 1; /* both look at the same file on the filesystem. */
4408 return 0;
4409}
4410
4411static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
4412{
4413 if (diff_unmodified_pair(p))
4414 return;
4415
4416 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4417 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4418 return; /* no tree diffs in patch format */
4419
4420 run_diff(p, o);
4421}
4422
4423static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
4424 struct diffstat_t *diffstat)
4425{
4426 if (diff_unmodified_pair(p))
4427 return;
4428
4429 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4430 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4431 return; /* no useful stat for tree diffs */
4432
4433 run_diffstat(p, o, diffstat);
4434}
4435
4436static void diff_flush_checkdiff(struct diff_filepair *p,
4437 struct diff_options *o)
4438{
4439 if (diff_unmodified_pair(p))
4440 return;
4441
4442 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4443 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4444 return; /* nothing to check in tree diffs */
4445
4446 run_checkdiff(p, o);
4447}
4448
4449int diff_queue_is_empty(void)
4450{
4451 struct diff_queue_struct *q = &diff_queued_diff;
4452 int i;
4453 for (i = 0; i < q->nr; i++)
4454 if (!diff_unmodified_pair(q->queue[i]))
4455 return 0;
4456 return 1;
4457}
4458
4459#if DIFF_DEBUG
4460void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
4461{
4462 fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
4463 x, one ? one : "",
4464 s->path,
4465 DIFF_FILE_VALID(s) ? "valid" : "invalid",
4466 s->mode,
4467 s->oid_valid ? oid_to_hex(&s->oid) : "");
4468 fprintf(stderr, "queue[%d] %s size %lu\n",
4469 x, one ? one : "",
4470 s->size);
4471}
4472
4473void diff_debug_filepair(const struct diff_filepair *p, int i)
4474{
4475 diff_debug_filespec(p->one, i, "one");
4476 diff_debug_filespec(p->two, i, "two");
4477 fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
4478 p->score, p->status ? p->status : '?',
4479 p->one->rename_used, p->broken_pair);
4480}
4481
4482void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
4483{
4484 int i;
4485 if (msg)
4486 fprintf(stderr, "%s\n", msg);
4487 fprintf(stderr, "q->nr = %d\n", q->nr);
4488 for (i = 0; i < q->nr; i++) {
4489 struct diff_filepair *p = q->queue[i];
4490 diff_debug_filepair(p, i);
4491 }
4492}
4493#endif
4494
4495static void diff_resolve_rename_copy(void)
4496{
4497 int i;
4498 struct diff_filepair *p;
4499 struct diff_queue_struct *q = &diff_queued_diff;
4500
4501 diff_debug_queue("resolve-rename-copy", q);
4502
4503 for (i = 0; i < q->nr; i++) {
4504 p = q->queue[i];
4505 p->status = 0; /* undecided */
4506 if (DIFF_PAIR_UNMERGED(p))
4507 p->status = DIFF_STATUS_UNMERGED;
4508 else if (!DIFF_FILE_VALID(p->one))
4509 p->status = DIFF_STATUS_ADDED;
4510 else if (!DIFF_FILE_VALID(p->two))
4511 p->status = DIFF_STATUS_DELETED;
4512 else if (DIFF_PAIR_TYPE_CHANGED(p))
4513 p->status = DIFF_STATUS_TYPE_CHANGED;
4514
4515 /* from this point on, we are dealing with a pair
4516 * whose both sides are valid and of the same type, i.e.
4517 * either in-place edit or rename/copy edit.
4518 */
4519 else if (DIFF_PAIR_RENAME(p)) {
4520 /*
4521 * A rename might have re-connected a broken
4522 * pair up, causing the pathnames to be the
4523 * same again. If so, that's not a rename at
4524 * all, just a modification..
4525 *
4526 * Otherwise, see if this source was used for
4527 * multiple renames, in which case we decrement
4528 * the count, and call it a copy.
4529 */
4530 if (!strcmp(p->one->path, p->two->path))
4531 p->status = DIFF_STATUS_MODIFIED;
4532 else if (--p->one->rename_used > 0)
4533 p->status = DIFF_STATUS_COPIED;
4534 else
4535 p->status = DIFF_STATUS_RENAMED;
4536 }
4537 else if (oidcmp(&p->one->oid, &p->two->oid) ||
4538 p->one->mode != p->two->mode ||
4539 p->one->dirty_submodule ||
4540 p->two->dirty_submodule ||
4541 is_null_oid(&p->one->oid))
4542 p->status = DIFF_STATUS_MODIFIED;
4543 else {
4544 /* This is a "no-change" entry and should not
4545 * happen anymore, but prepare for broken callers.
4546 */
4547 error("feeding unmodified %s to diffcore",
4548 p->one->path);
4549 p->status = DIFF_STATUS_UNKNOWN;
4550 }
4551 }
4552 diff_debug_queue("resolve-rename-copy done", q);
4553}
4554
4555static int check_pair_status(struct diff_filepair *p)
4556{
4557 switch (p->status) {
4558 case DIFF_STATUS_UNKNOWN:
4559 return 0;
4560 case 0:
4561 die("internal error in diff-resolve-rename-copy");
4562 default:
4563 return 1;
4564 }
4565}
4566
4567static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
4568{
4569 int fmt = opt->output_format;
4570
4571 if (fmt & DIFF_FORMAT_CHECKDIFF)
4572 diff_flush_checkdiff(p, opt);
4573 else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
4574 diff_flush_raw(p, opt);
4575 else if (fmt & DIFF_FORMAT_NAME) {
4576 const char *name_a, *name_b;
4577 name_a = p->two->path;
4578 name_b = NULL;
4579 strip_prefix(opt->prefix_length, &name_a, &name_b);
4580 fprintf(opt->file, "%s", diff_line_prefix(opt));
4581 write_name_quoted(name_a, opt->file, opt->line_termination);
4582 }
4583}
4584
4585static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
4586{
4587 if (fs->mode)
4588 fprintf(file, " %s mode %06o ", newdelete, fs->mode);
4589 else
4590 fprintf(file, " %s ", newdelete);
4591 write_name_quoted(fs->path, file, '\n');
4592}
4593
4594
4595static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
4596 const char *line_prefix)
4597{
4598 if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
4599 fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
4600 p->two->mode, show_name ? ' ' : '\n');
4601 if (show_name) {
4602 write_name_quoted(p->two->path, file, '\n');
4603 }
4604 }
4605}
4606
4607static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
4608 const char *line_prefix)
4609{
4610 char *names = pprint_rename(p->one->path, p->two->path);
4611
4612 fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
4613 free(names);
4614 show_mode_change(file, p, 0, line_prefix);
4615}
4616
4617static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
4618{
4619 FILE *file = opt->file;
4620 const char *line_prefix = diff_line_prefix(opt);
4621
4622 switch(p->status) {
4623 case DIFF_STATUS_DELETED:
4624 fputs(line_prefix, file);
4625 show_file_mode_name(file, "delete", p->one);
4626 break;
4627 case DIFF_STATUS_ADDED:
4628 fputs(line_prefix, file);
4629 show_file_mode_name(file, "create", p->two);
4630 break;
4631 case DIFF_STATUS_COPIED:
4632 fputs(line_prefix, file);
4633 show_rename_copy(file, "copy", p, line_prefix);
4634 break;
4635 case DIFF_STATUS_RENAMED:
4636 fputs(line_prefix, file);
4637 show_rename_copy(file, "rename", p, line_prefix);
4638 break;
4639 default:
4640 if (p->score) {
4641 fprintf(file, "%s rewrite ", line_prefix);
4642 write_name_quoted(p->two->path, file, ' ');
4643 fprintf(file, "(%d%%)\n", similarity_index(p));
4644 }
4645 show_mode_change(file, p, !p->score, line_prefix);
4646 break;
4647 }
4648}
4649
4650struct patch_id_t {
4651 git_SHA_CTX *ctx;
4652 int patchlen;
4653};
4654
4655static int remove_space(char *line, int len)
4656{
4657 int i;
4658 char *dst = line;
4659 unsigned char c;
4660
4661 for (i = 0; i < len; i++)
4662 if (!isspace((c = line[i])))
4663 *dst++ = c;
4664
4665 return dst - line;
4666}
4667
4668static void patch_id_consume(void *priv, char *line, unsigned long len)
4669{
4670 struct patch_id_t *data = priv;
4671 int new_len;
4672
4673 /* Ignore line numbers when computing the SHA1 of the patch */
4674 if (starts_with(line, "@@ -"))
4675 return;
4676
4677 new_len = remove_space(line, len);
4678
4679 git_SHA1_Update(data->ctx, line, new_len);
4680 data->patchlen += new_len;
4681}
4682
4683static void patch_id_add_string(git_SHA_CTX *ctx, const char *str)
4684{
4685 git_SHA1_Update(ctx, str, strlen(str));
4686}
4687
4688static void patch_id_add_mode(git_SHA_CTX *ctx, unsigned mode)
4689{
4690 /* large enough for 2^32 in octal */
4691 char buf[12];
4692 int len = xsnprintf(buf, sizeof(buf), "%06o", mode);
4693 git_SHA1_Update(ctx, buf, len);
4694}
4695
4696/* returns 0 upon success, and writes result into sha1 */
4697static int diff_get_patch_id(struct diff_options *options, struct object_id *oid, int diff_header_only)
4698{
4699 struct diff_queue_struct *q = &diff_queued_diff;
4700 int i;
4701 git_SHA_CTX ctx;
4702 struct patch_id_t data;
4703
4704 git_SHA1_Init(&ctx);
4705 memset(&data, 0, sizeof(struct patch_id_t));
4706 data.ctx = &ctx;
4707
4708 for (i = 0; i < q->nr; i++) {
4709 xpparam_t xpp;
4710 xdemitconf_t xecfg;
4711 mmfile_t mf1, mf2;
4712 struct diff_filepair *p = q->queue[i];
4713 int len1, len2;
4714
4715 memset(&xpp, 0, sizeof(xpp));
4716 memset(&xecfg, 0, sizeof(xecfg));
4717 if (p->status == 0)
4718 return error("internal diff status error");
4719 if (p->status == DIFF_STATUS_UNKNOWN)
4720 continue;
4721 if (diff_unmodified_pair(p))
4722 continue;
4723 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4724 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4725 continue;
4726 if (DIFF_PAIR_UNMERGED(p))
4727 continue;
4728
4729 diff_fill_oid_info(p->one);
4730 diff_fill_oid_info(p->two);
4731
4732 len1 = remove_space(p->one->path, strlen(p->one->path));
4733 len2 = remove_space(p->two->path, strlen(p->two->path));
4734 patch_id_add_string(&ctx, "diff--git");
4735 patch_id_add_string(&ctx, "a/");
4736 git_SHA1_Update(&ctx, p->one->path, len1);
4737 patch_id_add_string(&ctx, "b/");
4738 git_SHA1_Update(&ctx, p->two->path, len2);
4739
4740 if (p->one->mode == 0) {
4741 patch_id_add_string(&ctx, "newfilemode");
4742 patch_id_add_mode(&ctx, p->two->mode);
4743 patch_id_add_string(&ctx, "---/dev/null");
4744 patch_id_add_string(&ctx, "+++b/");
4745 git_SHA1_Update(&ctx, p->two->path, len2);
4746 } else if (p->two->mode == 0) {
4747 patch_id_add_string(&ctx, "deletedfilemode");
4748 patch_id_add_mode(&ctx, p->one->mode);
4749 patch_id_add_string(&ctx, "---a/");
4750 git_SHA1_Update(&ctx, p->one->path, len1);
4751 patch_id_add_string(&ctx, "+++/dev/null");
4752 } else {
4753 patch_id_add_string(&ctx, "---a/");
4754 git_SHA1_Update(&ctx, p->one->path, len1);
4755 patch_id_add_string(&ctx, "+++b/");
4756 git_SHA1_Update(&ctx, p->two->path, len2);
4757 }
4758
4759 if (diff_header_only)
4760 continue;
4761
4762 if (fill_mmfile(&mf1, p->one) < 0 ||
4763 fill_mmfile(&mf2, p->two) < 0)
4764 return error("unable to read files to diff");
4765
4766 if (diff_filespec_is_binary(p->one) ||
4767 diff_filespec_is_binary(p->two)) {
4768 git_SHA1_Update(&ctx, oid_to_hex(&p->one->oid),
4769 GIT_SHA1_HEXSZ);
4770 git_SHA1_Update(&ctx, oid_to_hex(&p->two->oid),
4771 GIT_SHA1_HEXSZ);
4772 continue;
4773 }
4774
4775 xpp.flags = 0;
4776 xecfg.ctxlen = 3;
4777 xecfg.flags = 0;
4778 if (xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4779 &xpp, &xecfg))
4780 return error("unable to generate patch-id diff for %s",
4781 p->one->path);
4782 }
4783
4784 git_SHA1_Final(oid->hash, &ctx);
4785 return 0;
4786}
4787
4788int diff_flush_patch_id(struct diff_options *options, struct object_id *oid, int diff_header_only)
4789{
4790 struct diff_queue_struct *q = &diff_queued_diff;
4791 int i;
4792 int result = diff_get_patch_id(options, oid, diff_header_only);
4793
4794 for (i = 0; i < q->nr; i++)
4795 diff_free_filepair(q->queue[i]);
4796
4797 free(q->queue);
4798 DIFF_QUEUE_CLEAR(q);
4799
4800 return result;
4801}
4802
4803static int is_summary_empty(const struct diff_queue_struct *q)
4804{
4805 int i;
4806
4807 for (i = 0; i < q->nr; i++) {
4808 const struct diff_filepair *p = q->queue[i];
4809
4810 switch (p->status) {
4811 case DIFF_STATUS_DELETED:
4812 case DIFF_STATUS_ADDED:
4813 case DIFF_STATUS_COPIED:
4814 case DIFF_STATUS_RENAMED:
4815 return 0;
4816 default:
4817 if (p->score)
4818 return 0;
4819 if (p->one->mode && p->two->mode &&
4820 p->one->mode != p->two->mode)
4821 return 0;
4822 break;
4823 }
4824 }
4825 return 1;
4826}
4827
4828static const char rename_limit_warning[] =
4829N_("inexact rename detection was skipped due to too many files.");
4830
4831static const char degrade_cc_to_c_warning[] =
4832N_("only found copies from modified paths due to too many files.");
4833
4834static const char rename_limit_advice[] =
4835N_("you may want to set your %s variable to at least "
4836 "%d and retry the command.");
4837
4838void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4839{
4840 if (degraded_cc)
4841 warning(_(degrade_cc_to_c_warning));
4842 else if (needed)
4843 warning(_(rename_limit_warning));
4844 else
4845 return;
4846 if (0 < needed && needed < 32767)
4847 warning(_(rename_limit_advice), varname, needed);
4848}
4849
4850static void diff_flush_patch_all_file_pairs(struct diff_options *o)
4851{
4852 int i;
4853 struct diff_queue_struct *q = &diff_queued_diff;
4854
4855 if (WSEH_NEW & WS_RULE_MASK)
4856 die("BUG: WS rules bit mask overlaps with diff symbol flags");
4857
4858 for (i = 0; i < q->nr; i++) {
4859 struct diff_filepair *p = q->queue[i];
4860 if (check_pair_status(p))
4861 diff_flush_patch(p, o);
4862 }
4863}
4864
4865void diff_flush(struct diff_options *options)
4866{
4867 struct diff_queue_struct *q = &diff_queued_diff;
4868 int i, output_format = options->output_format;
4869 int separator = 0;
4870 int dirstat_by_line = 0;
4871
4872 /*
4873 * Order: raw, stat, summary, patch
4874 * or: name/name-status/checkdiff (other bits clear)
4875 */
4876 if (!q->nr)
4877 goto free_queue;
4878
4879 if (output_format & (DIFF_FORMAT_RAW |
4880 DIFF_FORMAT_NAME |
4881 DIFF_FORMAT_NAME_STATUS |
4882 DIFF_FORMAT_CHECKDIFF)) {
4883 for (i = 0; i < q->nr; i++) {
4884 struct diff_filepair *p = q->queue[i];
4885 if (check_pair_status(p))
4886 flush_one_pair(p, options);
4887 }
4888 separator++;
4889 }
4890
4891 if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4892 dirstat_by_line = 1;
4893
4894 if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4895 dirstat_by_line) {
4896 struct diffstat_t diffstat;
4897
4898 memset(&diffstat, 0, sizeof(struct diffstat_t));
4899 for (i = 0; i < q->nr; i++) {
4900 struct diff_filepair *p = q->queue[i];
4901 if (check_pair_status(p))
4902 diff_flush_stat(p, options, &diffstat);
4903 }
4904 if (output_format & DIFF_FORMAT_NUMSTAT)
4905 show_numstat(&diffstat, options);
4906 if (output_format & DIFF_FORMAT_DIFFSTAT)
4907 show_stats(&diffstat, options);
4908 if (output_format & DIFF_FORMAT_SHORTSTAT)
4909 show_shortstats(&diffstat, options);
4910 if (output_format & DIFF_FORMAT_DIRSTAT && dirstat_by_line)
4911 show_dirstat_by_line(&diffstat, options);
4912 free_diffstat_info(&diffstat);
4913 separator++;
4914 }
4915 if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4916 show_dirstat(options);
4917
4918 if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4919 for (i = 0; i < q->nr; i++) {
4920 diff_summary(options, q->queue[i]);
4921 }
4922 separator++;
4923 }
4924
4925 if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4926 DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4927 DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4928 /*
4929 * run diff_flush_patch for the exit status. setting
4930 * options->file to /dev/null should be safe, because we
4931 * aren't supposed to produce any output anyway.
4932 */
4933 if (options->close_file)
4934 fclose(options->file);
4935 options->file = xfopen("/dev/null", "w");
4936 options->close_file = 1;
4937 for (i = 0; i < q->nr; i++) {
4938 struct diff_filepair *p = q->queue[i];
4939 if (check_pair_status(p))
4940 diff_flush_patch(p, options);
4941 if (options->found_changes)
4942 break;
4943 }
4944 }
4945
4946 if (output_format & DIFF_FORMAT_PATCH) {
4947 if (separator) {
4948 emit_diff_symbol(options, DIFF_SYMBOL_SEPARATOR, NULL, 0, 0);
4949 if (options->stat_sep) {
4950 /* attach patch instead of inline */
4951 fputs(options->stat_sep, options->file);
4952 }
4953 }
4954
4955 diff_flush_patch_all_file_pairs(options);
4956 }
4957
4958 if (output_format & DIFF_FORMAT_CALLBACK)
4959 options->format_callback(q, options, options->format_callback_data);
4960
4961 for (i = 0; i < q->nr; i++)
4962 diff_free_filepair(q->queue[i]);
4963free_queue:
4964 free(q->queue);
4965 DIFF_QUEUE_CLEAR(q);
4966 if (options->close_file)
4967 fclose(options->file);
4968
4969 /*
4970 * Report the content-level differences with HAS_CHANGES;
4971 * diff_addremove/diff_change does not set the bit when
4972 * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4973 */
4974 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4975 if (options->found_changes)
4976 DIFF_OPT_SET(options, HAS_CHANGES);
4977 else
4978 DIFF_OPT_CLR(options, HAS_CHANGES);
4979 }
4980}
4981
4982static int match_filter(const struct diff_options *options, const struct diff_filepair *p)
4983{
4984 return (((p->status == DIFF_STATUS_MODIFIED) &&
4985 ((p->score &&
4986 filter_bit_tst(DIFF_STATUS_FILTER_BROKEN, options)) ||
4987 (!p->score &&
4988 filter_bit_tst(DIFF_STATUS_MODIFIED, options)))) ||
4989 ((p->status != DIFF_STATUS_MODIFIED) &&
4990 filter_bit_tst(p->status, options)));
4991}
4992
4993static void diffcore_apply_filter(struct diff_options *options)
4994{
4995 int i;
4996 struct diff_queue_struct *q = &diff_queued_diff;
4997 struct diff_queue_struct outq;
4998
4999 DIFF_QUEUE_CLEAR(&outq);
5000
5001 if (!options->filter)
5002 return;
5003
5004 if (filter_bit_tst(DIFF_STATUS_FILTER_AON, options)) {
5005 int found;
5006 for (i = found = 0; !found && i < q->nr; i++) {
5007 if (match_filter(options, q->queue[i]))
5008 found++;
5009 }
5010 if (found)
5011 return;
5012
5013 /* otherwise we will clear the whole queue
5014 * by copying the empty outq at the end of this
5015 * function, but first clear the current entries
5016 * in the queue.
5017 */
5018 for (i = 0; i < q->nr; i++)
5019 diff_free_filepair(q->queue[i]);
5020 }
5021 else {
5022 /* Only the matching ones */
5023 for (i = 0; i < q->nr; i++) {
5024 struct diff_filepair *p = q->queue[i];
5025 if (match_filter(options, p))
5026 diff_q(&outq, p);
5027 else
5028 diff_free_filepair(p);
5029 }
5030 }
5031 free(q->queue);
5032 *q = outq;
5033}
5034
5035/* Check whether two filespecs with the same mode and size are identical */
5036static int diff_filespec_is_identical(struct diff_filespec *one,
5037 struct diff_filespec *two)
5038{
5039 if (S_ISGITLINK(one->mode))
5040 return 0;
5041 if (diff_populate_filespec(one, 0))
5042 return 0;
5043 if (diff_populate_filespec(two, 0))
5044 return 0;
5045 return !memcmp(one->data, two->data, one->size);
5046}
5047
5048static int diff_filespec_check_stat_unmatch(struct diff_filepair *p)
5049{
5050 if (p->done_skip_stat_unmatch)
5051 return p->skip_stat_unmatch_result;
5052
5053 p->done_skip_stat_unmatch = 1;
5054 p->skip_stat_unmatch_result = 0;
5055 /*
5056 * 1. Entries that come from stat info dirtiness
5057 * always have both sides (iow, not create/delete),
5058 * one side of the object name is unknown, with
5059 * the same mode and size. Keep the ones that
5060 * do not match these criteria. They have real
5061 * differences.
5062 *
5063 * 2. At this point, the file is known to be modified,
5064 * with the same mode and size, and the object
5065 * name of one side is unknown. Need to inspect
5066 * the identical contents.
5067 */
5068 if (!DIFF_FILE_VALID(p->one) || /* (1) */
5069 !DIFF_FILE_VALID(p->two) ||
5070 (p->one->oid_valid && p->two->oid_valid) ||
5071 (p->one->mode != p->two->mode) ||
5072 diff_populate_filespec(p->one, CHECK_SIZE_ONLY) ||
5073 diff_populate_filespec(p->two, CHECK_SIZE_ONLY) ||
5074 (p->one->size != p->two->size) ||
5075 !diff_filespec_is_identical(p->one, p->two)) /* (2) */
5076 p->skip_stat_unmatch_result = 1;
5077 return p->skip_stat_unmatch_result;
5078}
5079
5080static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
5081{
5082 int i;
5083 struct diff_queue_struct *q = &diff_queued_diff;
5084 struct diff_queue_struct outq;
5085 DIFF_QUEUE_CLEAR(&outq);
5086
5087 for (i = 0; i < q->nr; i++) {
5088 struct diff_filepair *p = q->queue[i];
5089
5090 if (diff_filespec_check_stat_unmatch(p))
5091 diff_q(&outq, p);
5092 else {
5093 /*
5094 * The caller can subtract 1 from skip_stat_unmatch
5095 * to determine how many paths were dirty only
5096 * due to stat info mismatch.
5097 */
5098 if (!DIFF_OPT_TST(diffopt, NO_INDEX))
5099 diffopt->skip_stat_unmatch++;
5100 diff_free_filepair(p);
5101 }
5102 }
5103 free(q->queue);
5104 *q = outq;
5105}
5106
5107static int diffnamecmp(const void *a_, const void *b_)
5108{
5109 const struct diff_filepair *a = *((const struct diff_filepair **)a_);
5110 const struct diff_filepair *b = *((const struct diff_filepair **)b_);
5111 const char *name_a, *name_b;
5112
5113 name_a = a->one ? a->one->path : a->two->path;
5114 name_b = b->one ? b->one->path : b->two->path;
5115 return strcmp(name_a, name_b);
5116}
5117
5118void diffcore_fix_diff_index(struct diff_options *options)
5119{
5120 struct diff_queue_struct *q = &diff_queued_diff;
5121 QSORT(q->queue, q->nr, diffnamecmp);
5122}
5123
5124void diffcore_std(struct diff_options *options)
5125{
5126 /* NOTE please keep the following in sync with diff_tree_combined() */
5127 if (options->skip_stat_unmatch)
5128 diffcore_skip_stat_unmatch(options);
5129 if (!options->found_follow) {
5130 /* See try_to_follow_renames() in tree-diff.c */
5131 if (options->break_opt != -1)
5132 diffcore_break(options->break_opt);
5133 if (options->detect_rename)
5134 diffcore_rename(options);
5135 if (options->break_opt != -1)
5136 diffcore_merge_broken();
5137 }
5138 if (options->pickaxe)
5139 diffcore_pickaxe(options);
5140 if (options->orderfile)
5141 diffcore_order(options->orderfile);
5142 if (!options->found_follow)
5143 /* See try_to_follow_renames() in tree-diff.c */
5144 diff_resolve_rename_copy();
5145 diffcore_apply_filter(options);
5146
5147 if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5148 DIFF_OPT_SET(options, HAS_CHANGES);
5149 else
5150 DIFF_OPT_CLR(options, HAS_CHANGES);
5151
5152 options->found_follow = 0;
5153}
5154
5155int diff_result_code(struct diff_options *opt, int status)
5156{
5157 int result = 0;
5158
5159 diff_warn_rename_limit("diff.renameLimit",
5160 opt->needed_rename_limit,
5161 opt->degraded_cc_to_c);
5162 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5163 !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
5164 return status;
5165 if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5166 DIFF_OPT_TST(opt, HAS_CHANGES))
5167 result |= 01;
5168 if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
5169 DIFF_OPT_TST(opt, CHECK_FAILED))
5170 result |= 02;
5171 return result;
5172}
5173
5174int diff_can_quit_early(struct diff_options *opt)
5175{
5176 return (DIFF_OPT_TST(opt, QUICK) &&
5177 !opt->filter &&
5178 DIFF_OPT_TST(opt, HAS_CHANGES));
5179}
5180
5181/*
5182 * Shall changes to this submodule be ignored?
5183 *
5184 * Submodule changes can be configured to be ignored separately for each path,
5185 * but that configuration can be overridden from the command line.
5186 */
5187static int is_submodule_ignored(const char *path, struct diff_options *options)
5188{
5189 int ignored = 0;
5190 unsigned orig_flags = options->flags;
5191 if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
5192 set_diffopt_flags_from_submodule_config(options, path);
5193 if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
5194 ignored = 1;
5195 options->flags = orig_flags;
5196 return ignored;
5197}
5198
5199void diff_addremove(struct diff_options *options,
5200 int addremove, unsigned mode,
5201 const struct object_id *oid,
5202 int oid_valid,
5203 const char *concatpath, unsigned dirty_submodule)
5204{
5205 struct diff_filespec *one, *two;
5206
5207 if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
5208 return;
5209
5210 /* This may look odd, but it is a preparation for
5211 * feeding "there are unchanged files which should
5212 * not produce diffs, but when you are doing copy
5213 * detection you would need them, so here they are"
5214 * entries to the diff-core. They will be prefixed
5215 * with something like '=' or '*' (I haven't decided
5216 * which but should not make any difference).
5217 * Feeding the same new and old to diff_change()
5218 * also has the same effect.
5219 * Before the final output happens, they are pruned after
5220 * merged into rename/copy pairs as appropriate.
5221 */
5222 if (DIFF_OPT_TST(options, REVERSE_DIFF))
5223 addremove = (addremove == '+' ? '-' :
5224 addremove == '-' ? '+' : addremove);
5225
5226 if (options->prefix &&
5227 strncmp(concatpath, options->prefix, options->prefix_length))
5228 return;
5229
5230 one = alloc_filespec(concatpath);
5231 two = alloc_filespec(concatpath);
5232
5233 if (addremove != '+')
5234 fill_filespec(one, oid, oid_valid, mode);
5235 if (addremove != '-') {
5236 fill_filespec(two, oid, oid_valid, mode);
5237 two->dirty_submodule = dirty_submodule;
5238 }
5239
5240 diff_queue(&diff_queued_diff, one, two);
5241 if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5242 DIFF_OPT_SET(options, HAS_CHANGES);
5243}
5244
5245void diff_change(struct diff_options *options,
5246 unsigned old_mode, unsigned new_mode,
5247 const struct object_id *old_oid,
5248 const struct object_id *new_oid,
5249 int old_oid_valid, int new_oid_valid,
5250 const char *concatpath,
5251 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
5252{
5253 struct diff_filespec *one, *two;
5254 struct diff_filepair *p;
5255
5256 if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
5257 is_submodule_ignored(concatpath, options))
5258 return;
5259
5260 if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
5261 SWAP(old_mode, new_mode);
5262 SWAP(old_oid, new_oid);
5263 SWAP(old_oid_valid, new_oid_valid);
5264 SWAP(old_dirty_submodule, new_dirty_submodule);
5265 }
5266
5267 if (options->prefix &&
5268 strncmp(concatpath, options->prefix, options->prefix_length))
5269 return;
5270
5271 one = alloc_filespec(concatpath);
5272 two = alloc_filespec(concatpath);
5273 fill_filespec(one, old_oid, old_oid_valid, old_mode);
5274 fill_filespec(two, new_oid, new_oid_valid, new_mode);
5275 one->dirty_submodule = old_dirty_submodule;
5276 two->dirty_submodule = new_dirty_submodule;
5277 p = diff_queue(&diff_queued_diff, one, two);
5278
5279 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5280 return;
5281
5282 if (DIFF_OPT_TST(options, QUICK) && options->skip_stat_unmatch &&
5283 !diff_filespec_check_stat_unmatch(p))
5284 return;
5285
5286 DIFF_OPT_SET(options, HAS_CHANGES);
5287}
5288
5289struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
5290{
5291 struct diff_filepair *pair;
5292 struct diff_filespec *one, *two;
5293
5294 if (options->prefix &&
5295 strncmp(path, options->prefix, options->prefix_length))
5296 return NULL;
5297
5298 one = alloc_filespec(path);
5299 two = alloc_filespec(path);
5300 pair = diff_queue(&diff_queued_diff, one, two);
5301 pair->is_unmerged = 1;
5302 return pair;
5303}
5304
5305static char *run_textconv(const char *pgm, struct diff_filespec *spec,
5306 size_t *outsize)
5307{
5308 struct diff_tempfile *temp;
5309 const char *argv[3];
5310 const char **arg = argv;
5311 struct child_process child = CHILD_PROCESS_INIT;
5312 struct strbuf buf = STRBUF_INIT;
5313 int err = 0;
5314
5315 temp = prepare_temp_file(spec->path, spec);
5316 *arg++ = pgm;
5317 *arg++ = temp->name;
5318 *arg = NULL;
5319
5320 child.use_shell = 1;
5321 child.argv = argv;
5322 child.out = -1;
5323 if (start_command(&child)) {
5324 remove_tempfile();
5325 return NULL;
5326 }
5327
5328 if (strbuf_read(&buf, child.out, 0) < 0)
5329 err = error("error reading from textconv command '%s'", pgm);
5330 close(child.out);
5331
5332 if (finish_command(&child) || err) {
5333 strbuf_release(&buf);
5334 remove_tempfile();
5335 return NULL;
5336 }
5337 remove_tempfile();
5338
5339 return strbuf_detach(&buf, outsize);
5340}
5341
5342size_t fill_textconv(struct userdiff_driver *driver,
5343 struct diff_filespec *df,
5344 char **outbuf)
5345{
5346 size_t size;
5347
5348 if (!driver) {
5349 if (!DIFF_FILE_VALID(df)) {
5350 *outbuf = "";
5351 return 0;
5352 }
5353 if (diff_populate_filespec(df, 0))
5354 die("unable to read files to diff");
5355 *outbuf = df->data;
5356 return df->size;
5357 }
5358
5359 if (!driver->textconv)
5360 die("BUG: fill_textconv called with non-textconv driver");
5361
5362 if (driver->textconv_cache && df->oid_valid) {
5363 *outbuf = notes_cache_get(driver->textconv_cache,
5364 &df->oid,
5365 &size);
5366 if (*outbuf)
5367 return size;
5368 }
5369
5370 *outbuf = run_textconv(driver->textconv, df, &size);
5371 if (!*outbuf)
5372 die("unable to read files to diff");
5373
5374 if (driver->textconv_cache && df->oid_valid) {
5375 /* ignore errors, as we might be in a readonly repository */
5376 notes_cache_put(driver->textconv_cache, &df->oid, *outbuf,
5377 size);
5378 /*
5379 * we could save up changes and flush them all at the end,
5380 * but we would need an extra call after all diffing is done.
5381 * Since generating a cache entry is the slow path anyway,
5382 * this extra overhead probably isn't a big deal.
5383 */
5384 notes_cache_write(driver->textconv_cache);
5385 }
5386
5387 return size;
5388}
5389
5390int textconv_object(const char *path,
5391 unsigned mode,
5392 const struct object_id *oid,
5393 int oid_valid,
5394 char **buf,
5395 unsigned long *buf_size)
5396{
5397 struct diff_filespec *df;
5398 struct userdiff_driver *textconv;
5399
5400 df = alloc_filespec(path);
5401 fill_filespec(df, oid, oid_valid, mode);
5402 textconv = get_textconv(df);
5403 if (!textconv) {
5404 free_filespec(df);
5405 return 0;
5406 }
5407
5408 *buf_size = fill_textconv(textconv, df, buf);
5409 free_filespec(df);
5410 return 1;
5411}
5412
5413void setup_diff_pager(struct diff_options *opt)
5414{
5415 /*
5416 * If the user asked for our exit code, then either they want --quiet
5417 * or --exit-code. We should definitely not bother with a pager in the
5418 * former case, as we will generate no output. Since we still properly
5419 * report our exit code even when a pager is run, we _could_ run a
5420 * pager with --exit-code. But since we have not done so historically,
5421 * and because it is easy to find people oneline advising "git diff
5422 * --exit-code" in hooks and other scripts, we do not do so.
5423 */
5424 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5425 check_pager_config("diff") != 0)
5426 setup_pager();
5427}